]> git.ipfire.org Git - ipfire-2.x.git/blob - src/patches/suse-2.6.27.31/patches.xen/xen3-patch-2.6.19
projects / ipfire-2.x.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Add a patch to fix Intel E100 wake-on-lan problems.
[ipfire-2.x.git] / src / patches / suse-2.6.27.31 / patches.xen / xen3-patch-2.6.19
1 From: www.kernel.org
2 Subject: Linux 2.6.19
3 Patch-mainline: 2.6.19
4
5 Automatically created from "patches.kernel.org/patch-2.6.19" by xen-port-patches.py
6
7 Acked-by: jbeulich@novell.com
8
9 --- sle11-2009-05-14.orig/arch/x86/Kconfig 2009-04-20 11:36:10.000000000 +0200
10 +++ sle11-2009-05-14/arch/x86/Kconfig 2009-03-04 11:28:34.000000000 +0100
11 @@ -415,6 +415,7 @@ config SCHED_NO_NO_OMIT_FRAME_POINTER
12
13 menuconfig PARAVIRT_GUEST
14 bool "Paravirtualized guest support"
15 + depends on !X86_XEN && !X86_64_XEN
16 help
17 Say Y here to get to see options related to running Linux under
18 various hypervisors. This option alone does not add any kernel code.
19 --- sle11-2009-05-14.orig/arch/x86/kernel/apic_32-xen.c 2009-05-14 11:02:43.000000000 +0200
20 +++ sle11-2009-05-14/arch/x86/kernel/apic_32-xen.c 2009-03-04 11:28:34.000000000 +0100
21 @@ -54,7 +54,6 @@ static cpumask_t timer_bcast_ipi;
22 /*
23 * Knob to control our willingness to enable the local APIC.
24 */
25 -int enable_local_apic __initdata = 0; /* -1=force-disable, +1=force-enable */
26
27 /*
28 * Debug level
29 @@ -102,7 +101,7 @@ int get_physical_broadcast(void)
30
31 #ifndef CONFIG_XEN
32 #ifndef CONFIG_SMP
33 -static void up_apic_timer_interrupt_call(struct pt_regs *regs)
34 +static void up_apic_timer_interrupt_call(void)
35 {
36 int cpu = smp_processor_id();
37
38 @@ -111,11 +110,11 @@ static void up_apic_timer_interrupt_call
39 */
40 per_cpu(irq_stat, cpu).apic_timer_irqs++;
41
42 - smp_local_timer_interrupt(regs);
43 + smp_local_timer_interrupt();
44 }
45 #endif
46
47 -void smp_send_timer_broadcast_ipi(struct pt_regs *regs)
48 +void smp_send_timer_broadcast_ipi(void)
49 {
50 cpumask_t mask;
51
52 @@ -128,7 +127,7 @@ void smp_send_timer_broadcast_ipi(struct
53 * We can directly call the apic timer interrupt handler
54 * in UP case. Minus all irq related functions
55 */
56 - up_apic_timer_interrupt_call(regs);
57 + up_apic_timer_interrupt_call();
58 #endif
59 }
60 }
61 --- sle11-2009-05-14.orig/arch/x86/kernel/cpu/common-xen.c 2009-05-14 11:02:43.000000000 +0200
62 +++ sle11-2009-05-14/arch/x86/kernel/cpu/common-xen.c 2009-03-04 11:28:34.000000000 +0100
63 @@ -43,7 +43,7 @@ struct cpu_dev * cpu_devs[X86_VENDOR_NUM
64
65 extern int disable_pse;
66
67 -static void default_init(struct cpuinfo_x86 * c)
68 +static void __cpuinit default_init(struct cpuinfo_x86 * c)
69 {
70 /* Not much we can do here... */
71 /* Check if at least it has cpuid */
72 @@ -56,7 +56,7 @@ static void default_init(struct cpuinfo_
73 }
74 }
75
76 -static struct cpu_dev default_cpu = {
77 +static struct cpu_dev __cpuinitdata default_cpu = {
78 .c_init = default_init,
79 .c_vendor = "Unknown",
80 };
81 @@ -191,7 +191,16 @@ static void __cpuinit get_cpu_vendor(str
82
83 static int __init x86_fxsr_setup(char * s)
84 {
85 + /* Tell all the other CPU's to not use it... */
86 disable_x86_fxsr = 1;
87 +
88 + /*
89 + * ... and clear the bits early in the boot_cpu_data
90 + * so that the bootup process doesn't try to do this
91 + * either.
92 + */
93 + clear_bit(X86_FEATURE_FXSR, boot_cpu_data.x86_capability);
94 + clear_bit(X86_FEATURE_XMM, boot_cpu_data.x86_capability);
95 return 1;
96 }
97 __setup("nofxsr", x86_fxsr_setup);
98 @@ -272,7 +281,7 @@ static void __init early_cpu_detect(void
99 }
100 }
101
102 -void __cpuinit generic_identify(struct cpuinfo_x86 * c)
103 +static void __cpuinit generic_identify(struct cpuinfo_x86 * c)
104 {
105 u32 tfms, xlvl;
106 int ebx;
107 @@ -698,8 +707,7 @@ old_gdt:
108 */
109 atomic_inc(&init_mm.mm_count);
110 current->active_mm = &init_mm;
111 - if (current->mm)
112 - BUG();
113 + BUG_ON(current->mm);
114 enter_lazy_tlb(&init_mm, current);
115
116 load_esp0(t, thread);
117 @@ -712,7 +720,7 @@ old_gdt:
118 #endif
119
120 /* Clear %fs and %gs. */
121 - asm volatile ("xorl %eax, %eax; movl %eax, %fs; movl %eax, %gs");
122 + asm volatile ("movl %0, %%fs; movl %0, %%gs" : : "r" (0));
123
124 /* Clear all 6 debug registers: */
125 set_debugreg(0, 0);
126 --- sle11-2009-05-14.orig/arch/x86/kernel/entry_32-xen.S 2009-05-14 11:01:46.000000000 +0200
127 +++ sle11-2009-05-14/arch/x86/kernel/entry_32-xen.S 2009-05-14 11:07:47.000000000 +0200
128 @@ -80,8 +80,12 @@ VM_MASK = 0x00020000
129 NMI_MASK = 0x80000000
130
131 #ifndef CONFIG_XEN
132 -#define DISABLE_INTERRUPTS cli
133 -#define ENABLE_INTERRUPTS sti
134 +/* These are replaces for paravirtualization */
135 +#define DISABLE_INTERRUPTS cli
136 +#define ENABLE_INTERRUPTS sti
137 +#define ENABLE_INTERRUPTS_SYSEXIT sti; sysexit
138 +#define INTERRUPT_RETURN iret
139 +#define GET_CR0_INTO_EAX movl %cr0, %eax
140 #else
141 /* Offsets into shared_info_t. */
142 #define evtchn_upcall_pending /* 0 */
143 @@ -99,15 +103,29 @@ NMI_MASK = 0x80000000
144
145 #define __DISABLE_INTERRUPTS movb $1,evtchn_upcall_mask(%esi)
146 #define __ENABLE_INTERRUPTS movb $0,evtchn_upcall_mask(%esi)
147 +#define __TEST_PENDING testb $0xFF,evtchn_upcall_pending(%esi)
148 #define DISABLE_INTERRUPTS GET_VCPU_INFO ; \
149 __DISABLE_INTERRUPTS
150 #define ENABLE_INTERRUPTS GET_VCPU_INFO ; \
151 __ENABLE_INTERRUPTS
152 -#define __TEST_PENDING testb $0xFF,evtchn_upcall_pending(%esi)
153 +#define ENABLE_INTERRUPTS_SYSEXIT __ENABLE_INTERRUPTS ; \
154 +sysexit_scrit: /**** START OF SYSEXIT CRITICAL REGION ****/ ; \
155 + __TEST_PENDING ; \
156 + jnz 14f # process more events if necessary... ; \
157 + movl ESI(%esp), %esi ; \
158 + sysexit ; \
159 +14: __DISABLE_INTERRUPTS ; \
160 + TRACE_IRQS_OFF ; \
161 +sysexit_ecrit: /**** END OF SYSEXIT CRITICAL REGION ****/ ; \
162 + push %esp ; \
163 + call evtchn_do_upcall ; \
164 + add $4,%esp ; \
165 + jmp ret_from_intr
166 +#define INTERRUPT_RETURN iret
167 #endif
168
169 #ifdef CONFIG_PREEMPT
170 -#define preempt_stop cli; TRACE_IRQS_OFF
171 +#define preempt_stop DISABLE_INTERRUPTS; TRACE_IRQS_OFF
172 #else
173 #define preempt_stop
174 #define resume_kernel restore_nocheck
175 @@ -206,18 +224,21 @@ NMI_MASK = 0x80000000
176
177 #define RING0_INT_FRAME \
178 CFI_STARTPROC simple;\
179 + CFI_SIGNAL_FRAME;\
180 CFI_DEF_CFA esp, 3*4;\
181 /*CFI_OFFSET cs, -2*4;*/\
182 CFI_OFFSET eip, -3*4
183
184 #define RING0_EC_FRAME \
185 CFI_STARTPROC simple;\
186 + CFI_SIGNAL_FRAME;\
187 CFI_DEF_CFA esp, 4*4;\
188 /*CFI_OFFSET cs, -2*4;*/\
189 CFI_OFFSET eip, -3*4
190
191 #define RING0_PTREGS_FRAME \
192 CFI_STARTPROC simple;\
193 + CFI_SIGNAL_FRAME;\
194 CFI_DEF_CFA esp, OLDESP-EBX;\
195 /*CFI_OFFSET cs, CS-OLDESP;*/\
196 CFI_OFFSET eip, EIP-OLDESP;\
197 @@ -263,8 +284,9 @@ ret_from_intr:
198 check_userspace:
199 movl EFLAGS(%esp), %eax # mix EFLAGS and CS
200 movb CS(%esp), %al
201 - testl $(VM_MASK | 2), %eax
202 - jz resume_kernel
203 + andl $(VM_MASK | SEGMENT_RPL_MASK), %eax
204 + cmpl $USER_RPL, %eax
205 + jb resume_kernel # not returning to v8086 or userspace
206 ENTRY(resume_userspace)
207 DISABLE_INTERRUPTS # make sure we don't miss an interrupt
208 # setting need_resched or sigpending
209 @@ -277,7 +299,7 @@ ENTRY(resume_userspace)
210
211 #ifdef CONFIG_PREEMPT
212 ENTRY(resume_kernel)
213 - cli
214 + DISABLE_INTERRUPTS
215 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
216 jnz restore_nocheck
217 need_resched:
218 @@ -297,6 +319,7 @@ need_resched:
219 # sysenter call handler stub
220 ENTRY(sysenter_entry)
221 CFI_STARTPROC simple
222 + CFI_SIGNAL_FRAME
223 CFI_DEF_CFA esp, 0
224 CFI_REGISTER esp, ebp
225 movl SYSENTER_stack_esp0(%esp),%esp
226 @@ -305,7 +328,7 @@ sysenter_past_esp:
227 * No need to follow this irqs on/off section: the syscall
228 * disabled irqs and here we enable it straight after entry:
229 */
230 - sti
231 + ENABLE_INTERRUPTS
232 pushl $(__USER_DS)
233 CFI_ADJUST_CFA_OFFSET 4
234 /*CFI_REL_OFFSET ss, 0*/
235 @@ -359,26 +382,8 @@ sysenter_past_esp:
236 movl EIP(%esp), %edx
237 movl OLDESP(%esp), %ecx
238 xorl %ebp,%ebp
239 -#ifdef CONFIG_XEN
240 TRACE_IRQS_ON
241 - __ENABLE_INTERRUPTS
242 -sysexit_scrit: /**** START OF SYSEXIT CRITICAL REGION ****/
243 - __TEST_PENDING
244 - jnz 14f # process more events if necessary...
245 - movl ESI(%esp), %esi
246 - sysexit
247 -14: __DISABLE_INTERRUPTS
248 - TRACE_IRQS_OFF
249 -sysexit_ecrit: /**** END OF SYSEXIT CRITICAL REGION ****/
250 - push %esp
251 - call evtchn_do_upcall
252 - add $4,%esp
253 - jmp ret_from_intr
254 -#else
255 - TRACE_IRQS_ON
256 - sti
257 - sysexit
258 -#endif /* !CONFIG_XEN */
259 + ENABLE_INTERRUPTS_SYSEXIT
260 CFI_ENDPROC
261
262 # pv sysenter call handler stub
263 @@ -444,8 +449,8 @@ restore_all:
264 # See comments in process.c:copy_thread() for details.
265 movb OLDSS(%esp), %ah
266 movb CS(%esp), %al
267 - andl $(VM_MASK | (4 << 8) | 3), %eax
268 - cmpl $((4 << 8) | 3), %eax
269 + andl $(VM_MASK | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax
270 + cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax
271 CFI_REMEMBER_STATE
272 je ldt_ss # returning to user-space with LDT SS
273 restore_nocheck:
274 @@ -467,12 +472,11 @@ restore_nocheck_notrace:
275 RESTORE_REGS
276 addl $4, %esp
277 CFI_ADJUST_CFA_OFFSET -4
278 -1: iret
279 +1: INTERRUPT_RETURN
280 .section .fixup,"ax"
281 iret_exc:
282 #ifndef CONFIG_XEN
283 - TRACE_IRQS_ON
284 - sti
285 + ENABLE_INTERRUPTS
286 #endif
287 pushl $0 # no error code
288 pushl $do_iret_error
289 @@ -498,7 +502,7 @@ ldt_ss:
290 * dosemu and wine happy. */
291 subl $8, %esp # reserve space for switch16 pointer
292 CFI_ADJUST_CFA_OFFSET 8
293 - cli
294 + DISABLE_INTERRUPTS
295 TRACE_IRQS_OFF
296 movl %esp, %eax
297 /* Set up the 16bit stack frame with switch32 pointer on top,
298 @@ -508,7 +512,7 @@ ldt_ss:
299 TRACE_IRQS_IRET
300 RESTORE_REGS
301 lss 20+4(%esp), %esp # switch to 16bit stack
302 -1: iret
303 +1: INTERRUPT_RETURN
304 .section __ex_table,"a"
305 .align 4
306 .long 1b,iret_exc
307 @@ -524,7 +528,7 @@ scrit: /**** START OF CRITICAL REGION **
308 RESTORE_REGS
309 addl $4, %esp
310 CFI_ADJUST_CFA_OFFSET -4
311 -1: iret
312 +1: INTERRUPT_RETURN
313 .section __ex_table,"a"
314 .align 4
315 .long 1b,iret_exc
316 @@ -713,11 +717,9 @@ ENTRY(name) \
317 #define UNWIND_ESPFIX_STACK
318 #endif
319
320 -ENTRY(divide_error)
321 - RING0_INT_FRAME
322 - pushl $0 # no error code
323 - CFI_ADJUST_CFA_OFFSET 4
324 - pushl $do_divide_error
325 +KPROBE_ENTRY(page_fault)
326 + RING0_EC_FRAME
327 + pushl $do_page_fault
328 CFI_ADJUST_CFA_OFFSET 4
329 ALIGN
330 error_code:
331 @@ -767,6 +769,7 @@ error_code:
332 call *%edi
333 jmp ret_from_exception
334 CFI_ENDPROC
335 +KPROBE_END(page_fault)
336
337 #ifdef CONFIG_XEN
338 # A note on the "critical region" in our callback handler.
339 @@ -790,9 +793,11 @@ ENTRY(hypervisor_callback)
340 pushl %eax
341 CFI_ADJUST_CFA_OFFSET 4
342 SAVE_ALL
343 - testb $2,CS(%esp)
344 + movl CS(%esp),%ecx
345 movl EIP(%esp),%eax
346 - jnz .Ldo_upcall
347 + andl $SEGMENT_RPL_MASK,%ecx
348 + cmpl $USER_RPL,%ecx
349 + jae .Ldo_upcall
350 cmpl $scrit,%eax
351 jb 0f
352 cmpl $ecrit,%eax
353 @@ -928,7 +933,7 @@ ENTRY(device_not_available)
354 CFI_ADJUST_CFA_OFFSET 4
355 SAVE_ALL
356 #ifndef CONFIG_XEN
357 - movl %cr0, %eax
358 + GET_CR0_INTO_EAX
359 testl $0x4, %eax # EM (math emulation bit)
360 je device_available_emulate
361 pushl $0 # temporary storage for ORIG_EIP
362 @@ -963,9 +968,15 @@ device_available_emulate:
363 jne ok; \
364 label: \
365 movl SYSENTER_stack_esp0+offset(%esp),%esp; \
366 + CFI_DEF_CFA esp, 0; \
367 + CFI_UNDEFINED eip; \
368 pushfl; \
369 + CFI_ADJUST_CFA_OFFSET 4; \
370 pushl $__KERNEL_CS; \
371 - pushl $sysenter_past_esp
372 + CFI_ADJUST_CFA_OFFSET 4; \
373 + pushl $sysenter_past_esp; \
374 + CFI_ADJUST_CFA_OFFSET 4; \
375 + CFI_REL_OFFSET eip, 0
376 #endif /* CONFIG_XEN */
377
378 KPROBE_ENTRY(debug)
379 @@ -984,7 +995,8 @@ debug_stack_correct:
380 call do_debug
381 jmp ret_from_exception
382 CFI_ENDPROC
383 - .previous .text
384 +KPROBE_END(debug)
385 +
386 #ifndef CONFIG_XEN
387 /*
388 * NMI is doubly nasty. It can happen _while_ we're handling
389 @@ -994,7 +1006,7 @@ debug_stack_correct:
390 * check whether we got an NMI on the debug path where the debug
391 * fault happened on the sysenter path.
392 */
393 -ENTRY(nmi)
394 +KPROBE_ENTRY(nmi)
395 RING0_INT_FRAME
396 pushl %eax
397 CFI_ADJUST_CFA_OFFSET 4
398 @@ -1019,6 +1031,7 @@ ENTRY(nmi)
399 cmpl $sysenter_entry,12(%esp)
400 je nmi_debug_stack_check
401 nmi_stack_correct:
402 + /* We have a RING0_INT_FRAME here */
403 pushl %eax
404 CFI_ADJUST_CFA_OFFSET 4
405 SAVE_ALL
406 @@ -1029,9 +1042,12 @@ nmi_stack_correct:
407 CFI_ENDPROC
408
409 nmi_stack_fixup:
410 + RING0_INT_FRAME
411 FIX_STACK(12,nmi_stack_correct, 1)
412 jmp nmi_stack_correct
413 +
414 nmi_debug_stack_check:
415 + /* We have a RING0_INT_FRAME here */
416 cmpw $__KERNEL_CS,16(%esp)
417 jne nmi_stack_correct
418 cmpl $debug,(%esp)
419 @@ -1042,8 +1058,10 @@ nmi_debug_stack_check:
420 jmp nmi_stack_correct
421
422 nmi_16bit_stack:
423 - RING0_INT_FRAME
424 - /* create the pointer to lss back */
425 + /* We have a RING0_INT_FRAME here.
426 + *
427 + * create the pointer to lss back
428 + */
429 pushl %ss
430 CFI_ADJUST_CFA_OFFSET 4
431 pushl %esp
432 @@ -1064,14 +1082,14 @@ nmi_16bit_stack:
433 call do_nmi
434 RESTORE_REGS
435 lss 12+4(%esp), %esp # back to 16bit stack
436 -1: iret
437 +1: INTERRUPT_RETURN
438 CFI_ENDPROC
439 .section __ex_table,"a"
440 .align 4
441 .long 1b,iret_exc
442 .previous
443 #else
444 -ENTRY(nmi)
445 +KPROBE_ENTRY(nmi)
446 RING0_INT_FRAME
447 pushl %eax
448 CFI_ADJUST_CFA_OFFSET 4
449 @@ -1083,6 +1101,7 @@ ENTRY(nmi)
450 jmp restore_all
451 CFI_ENDPROC
452 #endif
453 +KPROBE_END(nmi)
454
455 KPROBE_ENTRY(int3)
456 RING0_INT_FRAME
457 @@ -1094,7 +1113,7 @@ KPROBE_ENTRY(int3)
458 call do_int3
459 jmp ret_from_exception
460 CFI_ENDPROC
461 - .previous .text
462 +KPROBE_END(int3)
463
464 ENTRY(overflow)
465 RING0_INT_FRAME
466 @@ -1159,7 +1178,7 @@ KPROBE_ENTRY(general_protection)
467 CFI_ADJUST_CFA_OFFSET 4
468 jmp error_code
469 CFI_ENDPROC
470 - .previous .text
471 +KPROBE_END(general_protection)
472
473 ENTRY(alignment_check)
474 RING0_EC_FRAME
475 @@ -1168,13 +1187,14 @@ ENTRY(alignment_check)
476 jmp error_code
477 CFI_ENDPROC
478
479 -KPROBE_ENTRY(page_fault)
480 - RING0_EC_FRAME
481 - pushl $do_page_fault
482 +ENTRY(divide_error)
483 + RING0_INT_FRAME
484 + pushl $0 # no error code
485 + CFI_ADJUST_CFA_OFFSET 4
486 + pushl $do_divide_error
487 CFI_ADJUST_CFA_OFFSET 4
488 jmp error_code
489 CFI_ENDPROC
490 - .previous .text
491
492 #ifdef CONFIG_X86_MCE
493 ENTRY(machine_check)
494 @@ -1236,6 +1256,19 @@ ENTRY(fixup_4gb_segment)
495 jmp error_code
496 CFI_ENDPROC
497
498 +ENTRY(kernel_thread_helper)
499 + pushl $0 # fake return address for unwinder
500 + CFI_STARTPROC
501 + movl %edx,%eax
502 + push %edx
503 + CFI_ADJUST_CFA_OFFSET 4
504 + call *%ebx
505 + push %eax
506 + CFI_ADJUST_CFA_OFFSET 4
507 + call do_exit
508 + CFI_ENDPROC
509 +ENDPROC(kernel_thread_helper)
510 +
511 .section .rodata,"a"
512 #include "syscall_table.S"
513
514 --- sle11-2009-05-14.orig/arch/x86/kernel/head_32-xen.S 2009-05-14 11:02:43.000000000 +0200
515 +++ sle11-2009-05-14/arch/x86/kernel/head_32-xen.S 2009-03-04 11:28:34.000000000 +0100
516 @@ -62,7 +62,7 @@ ENTRY(startup_32)
517 movl %eax,%gs
518 cld # gcc2 wants the direction flag cleared at all times
519
520 - pushl %eax # fake return address
521 + pushl $0 # fake return address for unwinder
522 jmp start_kernel
523
524 #define HYPERCALL_PAGE_OFFSET 0x1000
525 --- sle11-2009-05-14.orig/arch/x86/kernel/io_apic_32-xen.c 2009-03-16 16:13:45.000000000 +0100
526 +++ sle11-2009-05-14/arch/x86/kernel/io_apic_32-xen.c 2009-03-04 11:28:34.000000000 +0100
527 @@ -31,6 +31,9 @@
528 #include <linux/acpi.h>
529 #include <linux/module.h>
530 #include <linux/sysdev.h>
531 +#include <linux/pci.h>
532 +#include <linux/msi.h>
533 +#include <linux/htirq.h>
534
535 #include <asm/io.h>
536 #include <asm/smp.h>
537 @@ -38,13 +41,15 @@
538 #include <asm/timer.h>
539 #include <asm/i8259.h>
540 #include <asm/nmi.h>
541 +#include <asm/msidef.h>
542 +#include <asm/hypertransport.h>
543
544 #include <mach_apic.h>
545 +#include <mach_apicdef.h>
546
547 #include "io_ports.h"
548
549 #ifdef CONFIG_XEN
550 -
551 #include <xen/interface/xen.h>
552 #include <xen/interface/physdev.h>
553 #include <xen/evtchn.h>
554 @@ -56,32 +61,7 @@
555
556 unsigned long io_apic_irqs;
557
558 -static inline unsigned int xen_io_apic_read(unsigned int apic, unsigned int reg)
559 -{
560 - struct physdev_apic apic_op;
561 - int ret;
562 -
563 - apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr;
564 - apic_op.reg = reg;
565 - ret = HYPERVISOR_physdev_op(PHYSDEVOP_apic_read, &apic_op);
566 - if (ret)
567 - return ret;
568 - return apic_op.value;
569 -}
570 -
571 -static inline void xen_io_apic_write(unsigned int apic, unsigned int reg, unsigned int value)
572 -{
573 - struct physdev_apic apic_op;
574 -
575 - apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr;
576 - apic_op.reg = reg;
577 - apic_op.value = value;
578 - WARN_ON(HYPERVISOR_physdev_op(PHYSDEVOP_apic_write, &apic_op));
579 -}
580 -
581 -#define io_apic_read(a,r) xen_io_apic_read(a,r)
582 -#define io_apic_write(a,r,v) xen_io_apic_write(a,r,v)
583 -
584 +#define clear_IO_APIC() ((void)0)
585 #endif /* CONFIG_XEN */
586
587 int (*ioapic_renumber_irq)(int ioapic, int irq);
588 @@ -108,7 +88,7 @@ int sis_apic_bug = -1;
589 */
590 int nr_ioapic_registers[MAX_IO_APICS];
591
592 -int disable_timer_pin_1 __initdata;
593 +static int disable_timer_pin_1 __initdata;
594
595 /*
596 * Rough estimation of how many shared IRQs there are, can
597 @@ -128,12 +108,124 @@ static struct irq_pin_list {
598 int apic, pin, next;
599 } irq_2_pin[PIN_MAP_SIZE];
600
601 -int vector_irq[NR_VECTORS] __read_mostly = { [0 ... NR_VECTORS - 1] = -1};
602 -#ifdef CONFIG_PCI_MSI
603 -#define vector_to_irq(vector) \
604 - (platform_legacy_irq(vector) ? vector : vector_irq[vector])
605 +#ifndef CONFIG_XEN
606 +struct io_apic {
607 + unsigned int index;
608 + unsigned int unused[3];
609 + unsigned int data;
610 +};
611 +
612 +static __attribute_const__ struct io_apic __iomem *io_apic_base(int idx)
613 +{
614 + return (void __iomem *) __fix_to_virt(FIX_IO_APIC_BASE_0 + idx)
615 + + (mp_ioapics[idx].mpc_apicaddr & ~PAGE_MASK);
616 +}
617 +#endif
618 +
619 +static inline unsigned int io_apic_read(unsigned int apic, unsigned int reg)
620 +{
621 +#ifndef CONFIG_XEN
622 + struct io_apic __iomem *io_apic = io_apic_base(apic);
623 + writel(reg, &io_apic->index);
624 + return readl(&io_apic->data);
625 +#else
626 + struct physdev_apic apic_op;
627 + int ret;
628 +
629 + apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr;
630 + apic_op.reg = reg;
631 + ret = HYPERVISOR_physdev_op(PHYSDEVOP_apic_read, &apic_op);
632 + if (ret)
633 + return ret;
634 + return apic_op.value;
635 +#endif
636 +}
637 +
638 +static inline void io_apic_write(unsigned int apic, unsigned int reg, unsigned int value)
639 +{
640 +#ifndef CONFIG_XEN
641 + struct io_apic __iomem *io_apic = io_apic_base(apic);
642 + writel(reg, &io_apic->index);
643 + writel(value, &io_apic->data);
644 +#else
645 + struct physdev_apic apic_op;
646 +
647 + apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr;
648 + apic_op.reg = reg;
649 + apic_op.value = value;
650 + WARN_ON(HYPERVISOR_physdev_op(PHYSDEVOP_apic_write, &apic_op));
651 +#endif
652 +}
653 +
654 +#ifndef CONFIG_XEN
655 +/*
656 + * Re-write a value: to be used for read-modify-write
657 + * cycles where the read already set up the index register.
658 + *
659 + * Older SiS APIC requires we rewrite the index register
660 + */
661 +static inline void io_apic_modify(unsigned int apic, unsigned int reg, unsigned int value)
662 +{
663 + volatile struct io_apic *io_apic = io_apic_base(apic);
664 + if (sis_apic_bug)
665 + writel(reg, &io_apic->index);
666 + writel(value, &io_apic->data);
667 +}
668 #else
669 -#define vector_to_irq(vector) (vector)
670 +#define io_apic_modify io_apic_write
671 +#endif
672 +
673 +union entry_union {
674 + struct { u32 w1, w2; };
675 + struct IO_APIC_route_entry entry;
676 +};
677 +
678 +#ifndef CONFIG_XEN
679 +static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin)
680 +{
681 + union entry_union eu;
682 + unsigned long flags;
683 + spin_lock_irqsave(&ioapic_lock, flags);
684 + eu.w1 = io_apic_read(apic, 0x10 + 2 * pin);
685 + eu.w2 = io_apic_read(apic, 0x11 + 2 * pin);
686 + spin_unlock_irqrestore(&ioapic_lock, flags);
687 + return eu.entry;
688 +}
689 +#endif
690 +
691 +/*
692 + * When we write a new IO APIC routing entry, we need to write the high
693 + * word first! If the mask bit in the low word is clear, we will enable
694 + * the interrupt, and we need to make sure the entry is fully populated
695 + * before that happens.
696 + */
697 +static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
698 +{
699 + unsigned long flags;
700 + union entry_union eu;
701 + eu.entry = e;
702 + spin_lock_irqsave(&ioapic_lock, flags);
703 + io_apic_write(apic, 0x11 + 2*pin, eu.w2);
704 + io_apic_write(apic, 0x10 + 2*pin, eu.w1);
705 + spin_unlock_irqrestore(&ioapic_lock, flags);
706 +}
707 +
708 +#ifndef CONFIG_XEN
709 +/*
710 + * When we mask an IO APIC routing entry, we need to write the low
711 + * word first, in order to set the mask bit before we change the
712 + * high bits!
713 + */
714 +static void ioapic_mask_entry(int apic, int pin)
715 +{
716 + unsigned long flags;
717 + union entry_union eu = { .entry.mask = 1 };
718 +
719 + spin_lock_irqsave(&ioapic_lock, flags);
720 + io_apic_write(apic, 0x10 + 2*pin, eu.w1);
721 + io_apic_write(apic, 0x11 + 2*pin, eu.w2);
722 + spin_unlock_irqrestore(&ioapic_lock, flags);
723 +}
724 #endif
725
726 /*
727 @@ -159,9 +251,7 @@ static void add_pin_to_irq(unsigned int
728 entry->pin = pin;
729 }
730
731 -#ifdef CONFIG_XEN
732 -#define clear_IO_APIC() ((void)0)
733 -#else
734 +#ifndef CONFIG_XEN
735 /*
736 * Reroute an IRQ to a different pin.
737 */
738 @@ -246,25 +336,16 @@ static void unmask_IO_APIC_irq (unsigned
739 static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
740 {
741 struct IO_APIC_route_entry entry;
742 - unsigned long flags;
743
744 /* Check delivery_mode to be sure we're not clearing an SMI pin */
745 - spin_lock_irqsave(&ioapic_lock, flags);
746 - *(((int*)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
747 - *(((int*)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
748 - spin_unlock_irqrestore(&ioapic_lock, flags);
749 + entry = ioapic_read_entry(apic, pin);
750 if (entry.delivery_mode == dest_SMI)
751 return;
752
753 /*
754 * Disable it in the IO-APIC irq-routing table:
755 */
756 - memset(&entry, 0, sizeof(entry));
757 - entry.mask = 1;
758 - spin_lock_irqsave(&ioapic_lock, flags);
759 - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry) + 0));
760 - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry) + 1));
761 - spin_unlock_irqrestore(&ioapic_lock, flags);
762 + ioapic_mask_entry(apic, pin);
763 }
764
765 static void clear_IO_APIC (void)
766 @@ -304,7 +385,7 @@ static void set_ioapic_affinity_irq(unsi
767 break;
768 entry = irq_2_pin + entry->next;
769 }
770 - set_irq_info(irq, cpumask);
771 + set_native_irq_info(irq, cpumask);
772 spin_unlock_irqrestore(&ioapic_lock, flags);
773 }
774
775 @@ -1212,43 +1293,43 @@ static inline int IO_APIC_irq_trigger(in
776 /* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */
777 u8 irq_vector[NR_IRQ_VECTORS] __read_mostly; /* = { FIRST_DEVICE_VECTOR , 0 }; */
778
779 -int assign_irq_vector(int irq)
780 +static int __assign_irq_vector(int irq)
781 {
782 - unsigned long flags;
783 int vector;
784 struct physdev_irq irq_op;
785
786 - BUG_ON(irq != AUTO_ASSIGN && (unsigned)irq >= NR_IRQ_VECTORS);
787 + BUG_ON((unsigned)irq >= NR_IRQ_VECTORS);
788
789 if (irq < PIRQ_BASE || irq - PIRQ_BASE >= NR_PIRQS)
790 return -EINVAL;
791
792 - spin_lock_irqsave(&vector_lock, flags);
793 -
794 - if (irq != AUTO_ASSIGN && IO_APIC_VECTOR(irq) > 0) {
795 - spin_unlock_irqrestore(&vector_lock, flags);
796 - return IO_APIC_VECTOR(irq);
797 - }
798 + if (irq_vector[irq] > 0)
799 + return irq_vector[irq];
800
801 irq_op.irq = irq;
802 - if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) {
803 - spin_unlock_irqrestore(&vector_lock, flags);
804 + if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op))
805 return -ENOSPC;
806 - }
807
808 vector = irq_op.vector;
809 - vector_irq[vector] = irq;
810 - if (irq != AUTO_ASSIGN)
811 - IO_APIC_VECTOR(irq) = vector;
812 + irq_vector[irq] = vector;
813 +
814 + return vector;
815 +}
816 +
817 +static int assign_irq_vector(int irq)
818 +{
819 + unsigned long flags;
820 + int vector;
821
822 + spin_lock_irqsave(&vector_lock, flags);
823 + vector = __assign_irq_vector(irq);
824 spin_unlock_irqrestore(&vector_lock, flags);
825
826 return vector;
827 }
828
829 #ifndef CONFIG_XEN
830 -static struct hw_interrupt_type ioapic_level_type;
831 -static struct hw_interrupt_type ioapic_edge_type;
832 +static struct irq_chip ioapic_chip;
833
834 #define IOAPIC_AUTO -1
835 #define IOAPIC_EDGE 0
836 @@ -1256,16 +1337,16 @@ static struct hw_interrupt_type ioapic_e
837
838 static void ioapic_register_intr(int irq, int vector, unsigned long trigger)
839 {
840 - unsigned idx;
841 -
842 - idx = use_pci_vector() && !platform_legacy_irq(irq) ? vector : irq;
843 -
844 if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
845 trigger == IOAPIC_LEVEL)
846 - irq_desc[idx].chip = &ioapic_level_type;
847 - else
848 - irq_desc[idx].chip = &ioapic_edge_type;
849 - set_intr_gate(vector, interrupt[idx]);
850 + set_irq_chip_and_handler_name(irq, &ioapic_chip,
851 + handle_fasteoi_irq, "fasteoi");
852 + else {
853 + irq_desc[irq].status |= IRQ_DELAYED_DISABLE;
854 + set_irq_chip_and_handler_name(irq, &ioapic_chip,
855 + handle_edge_irq, "edge");
856 + }
857 + set_intr_gate(vector, interrupt[irq]);
858 }
859 #else
860 #define ioapic_register_intr(irq, vector, trigger) evtchn_register_pirq(irq)
861 @@ -1336,9 +1417,8 @@ static void __init setup_IO_APIC_irqs(vo
862 if (!apic && (irq < 16))
863 disable_8259A_irq(irq);
864 }
865 + ioapic_write_entry(apic, pin, entry);
866 spin_lock_irqsave(&ioapic_lock, flags);
867 - io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1));
868 - io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0));
869 set_native_irq_info(irq, TARGET_CPUS);
870 spin_unlock_irqrestore(&ioapic_lock, flags);
871 }
872 @@ -1355,7 +1435,6 @@ static void __init setup_IO_APIC_irqs(vo
873 static void __init setup_ExtINT_IRQ0_pin(unsigned int apic, unsigned int pin, int vector)
874 {
875 struct IO_APIC_route_entry entry;
876 - unsigned long flags;
877
878 memset(&entry,0,sizeof(entry));
879
880 @@ -1380,15 +1459,13 @@ static void __init setup_ExtINT_IRQ0_pin
881 * The timer IRQ doesn't have to know that behind the
882 * scene we have a 8259A-master in AEOI mode ...
883 */
884 - irq_desc[0].chip = &ioapic_edge_type;
885 + irq_desc[0].chip = &ioapic_chip;
886 + set_irq_handler(0, handle_edge_irq);
887
888 /*
889 * Add it to the IO-APIC irq-routing table:
890 */
891 - spin_lock_irqsave(&ioapic_lock, flags);
892 - io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1));
893 - io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0));
894 - spin_unlock_irqrestore(&ioapic_lock, flags);
895 + ioapic_write_entry(apic, pin, entry);
896
897 enable_8259A_irq(0);
898 }
899 @@ -1498,10 +1575,7 @@ void __init print_IO_APIC(void)
900 for (i = 0; i <= reg_01.bits.entries; i++) {
901 struct IO_APIC_route_entry entry;
902
903 - spin_lock_irqsave(&ioapic_lock, flags);
904 - *(((int *)&entry)+0) = io_apic_read(apic, 0x10+i*2);
905 - *(((int *)&entry)+1) = io_apic_read(apic, 0x11+i*2);
906 - spin_unlock_irqrestore(&ioapic_lock, flags);
907 + entry = ioapic_read_entry(apic, i);
908
909 printk(KERN_DEBUG " %02x %03X %02X ",
910 i,
911 @@ -1521,17 +1595,12 @@ void __init print_IO_APIC(void)
912 );
913 }
914 }
915 - if (use_pci_vector())
916 - printk(KERN_INFO "Using vector-based indexing\n");
917 printk(KERN_DEBUG "IRQ to pin mappings:\n");
918 for (i = 0; i < NR_IRQS; i++) {
919 struct irq_pin_list *entry = irq_2_pin + i;
920 if (entry->pin < 0)
921 continue;
922 - if (use_pci_vector() && !platform_legacy_irq(i))
923 - printk(KERN_DEBUG "IRQ%d ", IO_APIC_VECTOR(i));
924 - else
925 - printk(KERN_DEBUG "IRQ%d ", i);
926 + printk(KERN_DEBUG "IRQ%d ", i);
927 for (;;) {
928 printk("-> %d:%d", entry->apic, entry->pin);
929 if (!entry->next)
930 @@ -1720,10 +1789,7 @@ static void __init enable_IO_APIC(void)
931 /* See if any of the pins is in ExtINT mode */
932 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
933 struct IO_APIC_route_entry entry;
934 - spin_lock_irqsave(&ioapic_lock, flags);
935 - *(((int *)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
936 - *(((int *)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
937 - spin_unlock_irqrestore(&ioapic_lock, flags);
938 + entry = ioapic_read_entry(apic, pin);
939
940
941 /* If the interrupt line is enabled and in ExtInt mode
942 @@ -1782,7 +1848,6 @@ void disable_IO_APIC(void)
943 */
944 if (ioapic_i8259.pin != -1) {
945 struct IO_APIC_route_entry entry;
946 - unsigned long flags;
947
948 memset(&entry, 0, sizeof(entry));
949 entry.mask = 0; /* Enabled */
950 @@ -1799,12 +1864,7 @@ void disable_IO_APIC(void)
951 /*
952 * Add it to the IO-APIC irq-routing table:
953 */
954 - spin_lock_irqsave(&ioapic_lock, flags);
955 - io_apic_write(ioapic_i8259.apic, 0x11+2*ioapic_i8259.pin,
956 - *(((int *)&entry)+1));
957 - io_apic_write(ioapic_i8259.apic, 0x10+2*ioapic_i8259.pin,
958 - *(((int *)&entry)+0));
959 - spin_unlock_irqrestore(&ioapic_lock, flags);
960 + ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry);
961 }
962 disconnect_bsp_APIC(ioapic_i8259.pin != -1);
963 #endif
964 @@ -1971,6 +2031,8 @@ static int __init timer_irq_works(void)
965 */
966
967 /*
968 + * Startup quirk:
969 + *
970 * Starting up a edge-triggered IO-APIC interrupt is
971 * nasty - we need to make sure that we get the edge.
972 * If it is already asserted for some reason, we need
973 @@ -1978,8 +2040,10 @@ static int __init timer_irq_works(void)
974 *
975 * This is not complete - we should be able to fake
976 * an edge even if it isn't on the 8259A...
977 + *
978 + * (We do this for level-triggered IRQs too - it cannot hurt.)
979 */
980 -static unsigned int startup_edge_ioapic_irq(unsigned int irq)
981 +static unsigned int startup_ioapic_irq(unsigned int irq)
982 {
983 int was_pending = 0;
984 unsigned long flags;
985 @@ -1996,47 +2060,18 @@ static unsigned int startup_edge_ioapic_
986 return was_pending;
987 }
988
989 -/*
990 - * Once we have recorded IRQ_PENDING already, we can mask the
991 - * interrupt for real. This prevents IRQ storms from unhandled
992 - * devices.
993 - */
994 -static void ack_edge_ioapic_irq(unsigned int irq)
995 -{
996 - move_irq(irq);
997 - if ((irq_desc[irq].status & (IRQ_PENDING | IRQ_DISABLED))
998 - == (IRQ_PENDING | IRQ_DISABLED))
999 - mask_IO_APIC_irq(irq);
1000 - ack_APIC_irq();
1001 -}
1002 -
1003 -/*
1004 - * Level triggered interrupts can just be masked,
1005 - * and shutting down and starting up the interrupt
1006 - * is the same as enabling and disabling them -- except
1007 - * with a startup need to return a "was pending" value.
1008 - *
1009 - * Level triggered interrupts are special because we
1010 - * do not touch any IO-APIC register while handling
1011 - * them. We ack the APIC in the end-IRQ handler, not
1012 - * in the start-IRQ-handler. Protection against reentrance
1013 - * from the same interrupt is still provided, both by the
1014 - * generic IRQ layer and by the fact that an unacked local
1015 - * APIC does not accept IRQs.
1016 - */
1017 -static unsigned int startup_level_ioapic_irq (unsigned int irq)
1018 +static void ack_ioapic_irq(unsigned int irq)
1019 {
1020 - unmask_IO_APIC_irq(irq);
1021 -
1022 - return 0; /* don't check for pending */
1023 + move_native_irq(irq);
1024 + ack_APIC_irq();
1025 }
1026
1027 -static void end_level_ioapic_irq (unsigned int irq)
1028 +static void ack_ioapic_quirk_irq(unsigned int irq)
1029 {
1030 unsigned long v;
1031 int i;
1032
1033 - move_irq(irq);
1034 + move_native_irq(irq);
1035 /*
1036 * It appears there is an erratum which affects at least version 0x11
1037 * of I/O APIC (that's the 82093AA and cores integrated into various
1038 @@ -2056,7 +2091,7 @@ static void end_level_ioapic_irq (unsign
1039 * operation to prevent an edge-triggered interrupt escaping meanwhile.
1040 * The idea is from Manfred Spraul. --macro
1041 */
1042 - i = IO_APIC_VECTOR(irq);
1043 + i = irq_vector[irq];
1044
1045 v = apic_read(APIC_TMR + ((i & ~0x1f) >> 1));
1046
1047 @@ -2071,104 +2106,24 @@ static void end_level_ioapic_irq (unsign
1048 }
1049 }
1050
1051 -#ifdef CONFIG_PCI_MSI
1052 -static unsigned int startup_edge_ioapic_vector(unsigned int vector)
1053 -{
1054 - int irq = vector_to_irq(vector);
1055 -
1056 - return startup_edge_ioapic_irq(irq);
1057 -}
1058 -
1059 -static void ack_edge_ioapic_vector(unsigned int vector)
1060 -{
1061 - int irq = vector_to_irq(vector);
1062 -
1063 - move_native_irq(vector);
1064 - ack_edge_ioapic_irq(irq);
1065 -}
1066 -
1067 -static unsigned int startup_level_ioapic_vector (unsigned int vector)
1068 -{
1069 - int irq = vector_to_irq(vector);
1070 -
1071 - return startup_level_ioapic_irq (irq);
1072 -}
1073 -
1074 -static void end_level_ioapic_vector (unsigned int vector)
1075 -{
1076 - int irq = vector_to_irq(vector);
1077 -
1078 - move_native_irq(vector);
1079 - end_level_ioapic_irq(irq);
1080 -}
1081 -
1082 -static void mask_IO_APIC_vector (unsigned int vector)
1083 -{
1084 - int irq = vector_to_irq(vector);
1085 -
1086 - mask_IO_APIC_irq(irq);
1087 -}
1088 -
1089 -static void unmask_IO_APIC_vector (unsigned int vector)
1090 +static int ioapic_retrigger_irq(unsigned int irq)
1091 {
1092 - int irq = vector_to_irq(vector);
1093 -
1094 - unmask_IO_APIC_irq(irq);
1095 -}
1096 -
1097 -#ifdef CONFIG_SMP
1098 -static void set_ioapic_affinity_vector (unsigned int vector,
1099 - cpumask_t cpu_mask)
1100 -{
1101 - int irq = vector_to_irq(vector);
1102 -
1103 - set_native_irq_info(vector, cpu_mask);
1104 - set_ioapic_affinity_irq(irq, cpu_mask);
1105 -}
1106 -#endif
1107 -#endif
1108 -
1109 -static int ioapic_retrigger(unsigned int irq)
1110 -{
1111 - send_IPI_self(IO_APIC_VECTOR(irq));
1112 + send_IPI_self(irq_vector[irq]);
1113
1114 return 1;
1115 }
1116
1117 -/*
1118 - * Level and edge triggered IO-APIC interrupts need different handling,
1119 - * so we use two separate IRQ descriptors. Edge triggered IRQs can be
1120 - * handled with the level-triggered descriptor, but that one has slightly
1121 - * more overhead. Level-triggered interrupts cannot be handled with the
1122 - * edge-triggered handler, without risking IRQ storms and other ugly
1123 - * races.
1124 - */
1125 -static struct hw_interrupt_type ioapic_edge_type __read_mostly = {
1126 - .typename = "IO-APIC-edge",
1127 - .startup = startup_edge_ioapic,
1128 - .shutdown = shutdown_edge_ioapic,
1129 - .enable = enable_edge_ioapic,
1130 - .disable = disable_edge_ioapic,
1131 - .ack = ack_edge_ioapic,
1132 - .end = end_edge_ioapic,
1133 -#ifdef CONFIG_SMP
1134 - .set_affinity = set_ioapic_affinity,
1135 -#endif
1136 - .retrigger = ioapic_retrigger,
1137 -};
1138 -
1139 -static struct hw_interrupt_type ioapic_level_type __read_mostly = {
1140 - .typename = "IO-APIC-level",
1141 - .startup = startup_level_ioapic,
1142 - .shutdown = shutdown_level_ioapic,
1143 - .enable = enable_level_ioapic,
1144 - .disable = disable_level_ioapic,
1145 - .ack = mask_and_ack_level_ioapic,
1146 - .end = end_level_ioapic,
1147 +static struct irq_chip ioapic_chip __read_mostly = {
1148 + .name = "IO-APIC",
1149 + .startup = startup_ioapic_irq,
1150 + .mask = mask_IO_APIC_irq,
1151 + .unmask = unmask_IO_APIC_irq,
1152 + .ack = ack_ioapic_irq,
1153 + .eoi = ack_ioapic_quirk_irq,
1154 #ifdef CONFIG_SMP
1155 - .set_affinity = set_ioapic_affinity,
1156 + .set_affinity = set_ioapic_affinity_irq,
1157 #endif
1158 - .retrigger = ioapic_retrigger,
1159 + .retrigger = ioapic_retrigger_irq,
1160 };
1161 #endif /* !CONFIG_XEN */
1162
1163 @@ -2189,12 +2144,7 @@ static inline void init_IO_APIC_traps(vo
1164 */
1165 for (irq = 0; irq < NR_IRQS ; irq++) {
1166 int tmp = irq;
1167 - if (use_pci_vector()) {
1168 - if (!platform_legacy_irq(tmp))
1169 - if ((tmp = vector_to_irq(tmp)) == -1)
1170 - continue;
1171 - }
1172 - if (IO_APIC_IRQ(tmp) && !IO_APIC_VECTOR(tmp)) {
1173 + if (IO_APIC_IRQ(tmp) && !irq_vector[tmp]) {
1174 /*
1175 * Hmm.. We don't have an entry for this,
1176 * so default to an old-fashioned 8259
1177 @@ -2205,22 +2155,23 @@ static inline void init_IO_APIC_traps(vo
1178 #ifndef CONFIG_XEN
1179 else
1180 /* Strange. Oh, well.. */
1181 - irq_desc[irq].chip = &no_irq_type;
1182 + irq_desc[irq].chip = &no_irq_chip;
1183 #endif
1184 }
1185 }
1186 }
1187
1188 #ifndef CONFIG_XEN
1189 -static void enable_lapic_irq (unsigned int irq)
1190 -{
1191 - unsigned long v;
1192 +/*
1193 + * The local APIC irq-chip implementation:
1194 + */
1195
1196 - v = apic_read(APIC_LVT0);
1197 - apic_write_around(APIC_LVT0, v & ~APIC_LVT_MASKED);
1198 +static void ack_apic(unsigned int irq)
1199 +{
1200 + ack_APIC_irq();
1201 }
1202
1203 -static void disable_lapic_irq (unsigned int irq)
1204 +static void mask_lapic_irq (unsigned int irq)
1205 {
1206 unsigned long v;
1207
1208 @@ -2228,21 +2179,19 @@ static void disable_lapic_irq (unsigned
1209 apic_write_around(APIC_LVT0, v | APIC_LVT_MASKED);
1210 }
1211
1212 -static void ack_lapic_irq (unsigned int irq)
1213 +static void unmask_lapic_irq (unsigned int irq)
1214 {
1215 - ack_APIC_irq();
1216 -}
1217 + unsigned long v;
1218
1219 -static void end_lapic_irq (unsigned int i) { /* nothing */ }
1220 + v = apic_read(APIC_LVT0);
1221 + apic_write_around(APIC_LVT0, v & ~APIC_LVT_MASKED);
1222 +}
1223
1224 -static struct hw_interrupt_type lapic_irq_type __read_mostly = {
1225 - .typename = "local-APIC-edge",
1226 - .startup = NULL, /* startup_irq() not used for IRQ0 */
1227 - .shutdown = NULL, /* shutdown_irq() not used for IRQ0 */
1228 - .enable = enable_lapic_irq,
1229 - .disable = disable_lapic_irq,
1230 - .ack = ack_lapic_irq,
1231 - .end = end_lapic_irq
1232 +static struct irq_chip lapic_chip __read_mostly = {
1233 + .name = "local-APIC-edge",
1234 + .mask = mask_lapic_irq,
1235 + .unmask = unmask_lapic_irq,
1236 + .eoi = ack_apic,
1237 };
1238
1239 static void setup_nmi (void)
1240 @@ -2275,17 +2224,13 @@ static inline void unlock_ExtINT_logic(v
1241 int apic, pin, i;
1242 struct IO_APIC_route_entry entry0, entry1;
1243 unsigned char save_control, save_freq_select;
1244 - unsigned long flags;
1245
1246 pin = find_isa_irq_pin(8, mp_INT);
1247 apic = find_isa_irq_apic(8, mp_INT);
1248 if (pin == -1)
1249 return;
1250
1251 - spin_lock_irqsave(&ioapic_lock, flags);
1252 - *(((int *)&entry0) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
1253 - *(((int *)&entry0) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
1254 - spin_unlock_irqrestore(&ioapic_lock, flags);
1255 + entry0 = ioapic_read_entry(apic, pin);
1256 clear_IO_APIC_pin(apic, pin);
1257
1258 memset(&entry1, 0, sizeof(entry1));
1259 @@ -2298,10 +2243,7 @@ static inline void unlock_ExtINT_logic(v
1260 entry1.trigger = 0;
1261 entry1.vector = 0;
1262
1263 - spin_lock_irqsave(&ioapic_lock, flags);
1264 - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry1) + 1));
1265 - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry1) + 0));
1266 - spin_unlock_irqrestore(&ioapic_lock, flags);
1267 + ioapic_write_entry(apic, pin, entry1);
1268
1269 save_control = CMOS_READ(RTC_CONTROL);
1270 save_freq_select = CMOS_READ(RTC_FREQ_SELECT);
1271 @@ -2320,10 +2262,7 @@ static inline void unlock_ExtINT_logic(v
1272 CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT);
1273 clear_IO_APIC_pin(apic, pin);
1274
1275 - spin_lock_irqsave(&ioapic_lock, flags);
1276 - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry0) + 1));
1277 - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry0) + 0));
1278 - spin_unlock_irqrestore(&ioapic_lock, flags);
1279 + ioapic_write_entry(apic, pin, entry0);
1280 }
1281
1282 int timer_uses_ioapic_pin_0;
1283 @@ -2423,7 +2362,8 @@ static inline void check_timer(void)
1284 printk(KERN_INFO "...trying to set up timer as Virtual Wire IRQ...");
1285
1286 disable_8259A_irq(0);
1287 - irq_desc[0].chip = &lapic_irq_type;
1288 + set_irq_chip_and_handler_name(0, &lapic_chip, handle_fasteoi_irq,
1289 + "fasteio");
1290 apic_write_around(APIC_LVT0, APIC_DM_FIXED | vector); /* Fixed mode */
1291 enable_8259A_irq(0);
1292
1293 @@ -2537,17 +2477,12 @@ static int ioapic_suspend(struct sys_dev
1294 {
1295 struct IO_APIC_route_entry *entry;
1296 struct sysfs_ioapic_data *data;
1297 - unsigned long flags;
1298 int i;
1299
1300 data = container_of(dev, struct sysfs_ioapic_data, dev);
1301 entry = data->entry;
1302 - spin_lock_irqsave(&ioapic_lock, flags);
1303 - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) {
1304 - *(((int *)entry) + 1) = io_apic_read(dev->id, 0x11 + 2 * i);
1305 - *(((int *)entry) + 0) = io_apic_read(dev->id, 0x10 + 2 * i);
1306 - }
1307 - spin_unlock_irqrestore(&ioapic_lock, flags);
1308 + for (i = 0; i < nr_ioapic_registers[dev->id]; i ++)
1309 + entry[i] = ioapic_read_entry(dev->id, i);
1310
1311 return 0;
1312 }
1313 @@ -2569,11 +2504,9 @@ static int ioapic_resume(struct sys_devi
1314 reg_00.bits.ID = mp_ioapics[dev->id].mpc_apicid;
1315 io_apic_write(dev->id, 0, reg_00.raw);
1316 }
1317 - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) {
1318 - io_apic_write(dev->id, 0x11+2*i, *(((int *)entry)+1));
1319 - io_apic_write(dev->id, 0x10+2*i, *(((int *)entry)+0));
1320 - }
1321 spin_unlock_irqrestore(&ioapic_lock, flags);
1322 + for (i = 0; i < nr_ioapic_registers[dev->id]; i ++)
1323 + ioapic_write_entry(dev->id, i, entry[i]);
1324
1325 return 0;
1326 }
1327 @@ -2619,8 +2552,240 @@ static int __init ioapic_init_sysfs(void
1328
1329 device_initcall(ioapic_init_sysfs);
1330
1331 +/*
1332 + * Dynamic irq allocate and deallocation
1333 + */
1334 +int create_irq(void)
1335 +{
1336 + /* Allocate an unused irq */
1337 + int irq, new, vector;
1338 + unsigned long flags;
1339 +
1340 + irq = -ENOSPC;
1341 + spin_lock_irqsave(&vector_lock, flags);
1342 + for (new = (NR_IRQS - 1); new >= 0; new--) {
1343 + if (platform_legacy_irq(new))
1344 + continue;
1345 + if (irq_vector[new] != 0)
1346 + continue;
1347 + vector = __assign_irq_vector(new);
1348 + if (likely(vector > 0))
1349 + irq = new;
1350 + break;
1351 + }
1352 + spin_unlock_irqrestore(&vector_lock, flags);
1353 +
1354 + if (irq >= 0) {
1355 + set_intr_gate(vector, interrupt[irq]);
1356 + dynamic_irq_init(irq);
1357 + }
1358 + return irq;
1359 +}
1360 +
1361 +void destroy_irq(unsigned int irq)
1362 +{
1363 + unsigned long flags;
1364 +
1365 + dynamic_irq_cleanup(irq);
1366 +
1367 + spin_lock_irqsave(&vector_lock, flags);
1368 + irq_vector[irq] = 0;
1369 + spin_unlock_irqrestore(&vector_lock, flags);
1370 +}
1371 +
1372 #endif /* CONFIG_XEN */
1373
1374 +/*
1375 + * MSI mesage composition
1376 + */
1377 +#if defined(CONFIG_PCI_MSI) && !defined(CONFIG_XEN)
1378 +static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_msg *msg)
1379 +{
1380 + int vector;
1381 + unsigned dest;
1382 +
1383 + vector = assign_irq_vector(irq);
1384 + if (vector >= 0) {
1385 + dest = cpu_mask_to_apicid(TARGET_CPUS);
1386 +
1387 + msg->address_hi = MSI_ADDR_BASE_HI;
1388 + msg->address_lo =
1389 + MSI_ADDR_BASE_LO |
1390 + ((INT_DEST_MODE == 0) ?
1391 + MSI_ADDR_DEST_MODE_PHYSICAL:
1392 + MSI_ADDR_DEST_MODE_LOGICAL) |
1393 + ((INT_DELIVERY_MODE != dest_LowestPrio) ?
1394 + MSI_ADDR_REDIRECTION_CPU:
1395 + MSI_ADDR_REDIRECTION_LOWPRI) |
1396 + MSI_ADDR_DEST_ID(dest);
1397 +
1398 + msg->data =
1399 + MSI_DATA_TRIGGER_EDGE |
1400 + MSI_DATA_LEVEL_ASSERT |
1401 + ((INT_DELIVERY_MODE != dest_LowestPrio) ?
1402 + MSI_DATA_DELIVERY_FIXED:
1403 + MSI_DATA_DELIVERY_LOWPRI) |
1404 + MSI_DATA_VECTOR(vector);
1405 + }
1406 + return vector;
1407 +}
1408 +
1409 +#ifdef CONFIG_SMP
1410 +static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
1411 +{
1412 + struct msi_msg msg;
1413 + unsigned int dest;
1414 + cpumask_t tmp;
1415 + int vector;
1416 +
1417 + cpus_and(tmp, mask, cpu_online_map);
1418 + if (cpus_empty(tmp))
1419 + tmp = TARGET_CPUS;
1420 +
1421 + vector = assign_irq_vector(irq);
1422 + if (vector < 0)
1423 + return;
1424 +
1425 + dest = cpu_mask_to_apicid(mask);
1426 +
1427 + read_msi_msg(irq, &msg);
1428 +
1429 + msg.data &= ~MSI_DATA_VECTOR_MASK;
1430 + msg.data |= MSI_DATA_VECTOR(vector);
1431 + msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
1432 + msg.address_lo |= MSI_ADDR_DEST_ID(dest);
1433 +
1434 + write_msi_msg(irq, &msg);
1435 + set_native_irq_info(irq, mask);
1436 +}
1437 +#endif /* CONFIG_SMP */
1438 +
1439 +/*
1440 + * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices,
1441 + * which implement the MSI or MSI-X Capability Structure.
1442 + */
1443 +static struct irq_chip msi_chip = {
1444 + .name = "PCI-MSI",
1445 + .unmask = unmask_msi_irq,
1446 + .mask = mask_msi_irq,
1447 + .ack = ack_ioapic_irq,
1448 +#ifdef CONFIG_SMP
1449 + .set_affinity = set_msi_irq_affinity,
1450 +#endif
1451 + .retrigger = ioapic_retrigger_irq,
1452 +};
1453 +
1454 +int arch_setup_msi_irq(unsigned int irq, struct pci_dev *dev)
1455 +{
1456 + struct msi_msg msg;
1457 + int ret;
1458 + ret = msi_compose_msg(dev, irq, &msg);
1459 + if (ret < 0)
1460 + return ret;
1461 +
1462 + write_msi_msg(irq, &msg);
1463 +
1464 + set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq,
1465 + "edge");
1466 +
1467 + return 0;
1468 +}
1469 +
1470 +void arch_teardown_msi_irq(unsigned int irq)
1471 +{
1472 + return;
1473 +}
1474 +
1475 +#endif /* CONFIG_PCI_MSI */
1476 +
1477 +/*
1478 + * Hypertransport interrupt support
1479 + */
1480 +#ifdef CONFIG_HT_IRQ
1481 +
1482 +#ifdef CONFIG_SMP
1483 +
1484 +static void target_ht_irq(unsigned int irq, unsigned int dest)
1485 +{
1486 + struct ht_irq_msg msg;
1487 + fetch_ht_irq_msg(irq, &msg);
1488 +
1489 + msg.address_lo &= ~(HT_IRQ_LOW_DEST_ID_MASK);
1490 + msg.address_hi &= ~(HT_IRQ_HIGH_DEST_ID_MASK);
1491 +
1492 + msg.address_lo |= HT_IRQ_LOW_DEST_ID(dest);
1493 + msg.address_hi |= HT_IRQ_HIGH_DEST_ID(dest);
1494 +
1495 + write_ht_irq_msg(irq, &msg);
1496 +}
1497 +
1498 +static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask)
1499 +{
1500 + unsigned int dest;
1501 + cpumask_t tmp;
1502 +
1503 + cpus_and(tmp, mask, cpu_online_map);
1504 + if (cpus_empty(tmp))
1505 + tmp = TARGET_CPUS;
1506 +
1507 + cpus_and(mask, tmp, CPU_MASK_ALL);
1508 +
1509 + dest = cpu_mask_to_apicid(mask);
1510 +
1511 + target_ht_irq(irq, dest);
1512 + set_native_irq_info(irq, mask);
1513 +}
1514 +#endif
1515 +
1516 +static struct irq_chip ht_irq_chip = {
1517 + .name = "PCI-HT",
1518 + .mask = mask_ht_irq,
1519 + .unmask = unmask_ht_irq,
1520 + .ack = ack_ioapic_irq,
1521 +#ifdef CONFIG_SMP
1522 + .set_affinity = set_ht_irq_affinity,
1523 +#endif
1524 + .retrigger = ioapic_retrigger_irq,
1525 +};
1526 +
1527 +int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
1528 +{
1529 + int vector;
1530 +
1531 + vector = assign_irq_vector(irq);
1532 + if (vector >= 0) {
1533 + struct ht_irq_msg msg;
1534 + unsigned dest;
1535 + cpumask_t tmp;
1536 +
1537 + cpus_clear(tmp);
1538 + cpu_set(vector >> 8, tmp);
1539 + dest = cpu_mask_to_apicid(tmp);
1540 +
1541 + msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest);
1542 +
1543 + msg.address_lo =
1544 + HT_IRQ_LOW_BASE |
1545 + HT_IRQ_LOW_DEST_ID(dest) |
1546 + HT_IRQ_LOW_VECTOR(vector) |
1547 + ((INT_DEST_MODE == 0) ?
1548 + HT_IRQ_LOW_DM_PHYSICAL :
1549 + HT_IRQ_LOW_DM_LOGICAL) |
1550 + HT_IRQ_LOW_RQEOI_EDGE |
1551 + ((INT_DELIVERY_MODE != dest_LowestPrio) ?
1552 + HT_IRQ_LOW_MT_FIXED :
1553 + HT_IRQ_LOW_MT_ARBITRATED) |
1554 + HT_IRQ_LOW_IRQ_MASKED;
1555 +
1556 + write_ht_irq_msg(irq, &msg);
1557 +
1558 + set_irq_chip_and_handler_name(irq, &ht_irq_chip,
1559 + handle_edge_irq, "edge");
1560 + }
1561 + return vector;
1562 +}
1563 +#endif /* CONFIG_HT_IRQ */
1564 +
1565 /* --------------------------------------------------------------------------
1566 ACPI-based IOAPIC Configuration
1567 -------------------------------------------------------------------------- */
1568 @@ -2774,13 +2939,34 @@ int io_apic_set_pci_routing (int ioapic,
1569 if (!ioapic && (irq < 16))
1570 disable_8259A_irq(irq);
1571
1572 + ioapic_write_entry(ioapic, pin, entry);
1573 spin_lock_irqsave(&ioapic_lock, flags);
1574 - io_apic_write(ioapic, 0x11+2*pin, *(((int *)&entry)+1));
1575 - io_apic_write(ioapic, 0x10+2*pin, *(((int *)&entry)+0));
1576 - set_native_irq_info(use_pci_vector() ? entry.vector : irq, TARGET_CPUS);
1577 + set_native_irq_info(irq, TARGET_CPUS);
1578 spin_unlock_irqrestore(&ioapic_lock, flags);
1579
1580 return 0;
1581 }
1582
1583 #endif /* CONFIG_ACPI */
1584 +
1585 +static int __init parse_disable_timer_pin_1(char *arg)
1586 +{
1587 + disable_timer_pin_1 = 1;
1588 + return 0;
1589 +}
1590 +early_param("disable_timer_pin_1", parse_disable_timer_pin_1);
1591 +
1592 +static int __init parse_enable_timer_pin_1(char *arg)
1593 +{
1594 + disable_timer_pin_1 = -1;
1595 + return 0;
1596 +}
1597 +early_param("enable_timer_pin_1", parse_enable_timer_pin_1);
1598 +
1599 +static int __init parse_noapic(char *arg)
1600 +{
1601 + /* disable IO-APIC */
1602 + disable_ioapic_setup();
1603 + return 0;
1604 +}
1605 +early_param("noapic", parse_noapic);
1606 --- sle11-2009-05-14.orig/arch/x86/kernel/irq_32-xen.c 2009-05-14 11:02:43.000000000 +0200
1607 +++ sle11-2009-05-14/arch/x86/kernel/irq_32-xen.c 2009-03-04 11:28:34.000000000 +0100
1608 @@ -53,8 +53,10 @@ static union irq_ctx *softirq_ctx[NR_CPU
1609 */
1610 fastcall unsigned int do_IRQ(struct pt_regs *regs)
1611 {
1612 + struct pt_regs *old_regs;
1613 /* high bit used in ret_from_ code */
1614 int irq = ~regs->orig_eax;
1615 + struct irq_desc *desc = irq_desc + irq;
1616 #ifdef CONFIG_4KSTACKS
1617 union irq_ctx *curctx, *irqctx;
1618 u32 *isp;
1619 @@ -66,6 +68,7 @@ fastcall unsigned int do_IRQ(struct pt_r
1620 BUG();
1621 }
1622
1623 + old_regs = set_irq_regs(regs);
1624 /*irq_enter();*/
1625 #ifdef CONFIG_DEBUG_STACKOVERFLOW
1626 /* Debugging check for stack overflow: is there less than 1KB free? */
1627 @@ -110,19 +113,20 @@ fastcall unsigned int do_IRQ(struct pt_r
1628 (curctx->tinfo.preempt_count & SOFTIRQ_MASK);
1629
1630 asm volatile(
1631 - " xchgl %%ebx,%%esp \n"
1632 - " call __do_IRQ \n"
1633 + " xchgl %%ebx,%%esp \n"
1634 + " call *%%edi \n"
1635 " movl %%ebx,%%esp \n"
1636 : "=a" (arg1), "=d" (arg2), "=b" (ebx)
1637 - : "0" (irq), "1" (regs), "2" (isp)
1638 - : "memory", "cc", "ecx"
1639 + : "0" (irq), "1" (desc), "2" (isp),
1640 + "D" (desc->handle_irq)
1641 + : "memory", "cc"
1642 );
1643 } else
1644 #endif
1645 - __do_IRQ(irq, regs);
1646 + desc->handle_irq(irq, desc);
1647
1648 /*irq_exit();*/
1649 -
1650 + set_irq_regs(old_regs);
1651 return 1;
1652 }
1653
1654 @@ -253,7 +257,8 @@ int show_interrupts(struct seq_file *p,
1655 for_each_online_cpu(j)
1656 seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
1657 #endif
1658 - seq_printf(p, " %14s", irq_desc[i].chip->typename);
1659 + seq_printf(p, " %8s", irq_desc[i].chip->name);
1660 + seq_printf(p, "-%-8s", irq_desc[i].name);
1661 seq_printf(p, " %s", action->name);
1662
1663 for (action=action->next; action; action = action->next)
1664 --- sle11-2009-05-14.orig/arch/x86/kernel/ldt_32-xen.c 2009-05-14 11:02:43.000000000 +0200
1665 +++ sle11-2009-05-14/arch/x86/kernel/ldt_32-xen.c 2009-03-04 11:28:34.000000000 +0100
1666 @@ -1,5 +1,5 @@
1667 /*
1668 - * linux/kernel/ldt.c
1669 + * linux/arch/i386/kernel/ldt.c
1670 *
1671 * Copyright (C) 1992 Krishna Balasubramanian and Linus Torvalds
1672 * Copyright (C) 1999 Ingo Molnar <mingo@redhat.com>
1673 --- sle11-2009-05-14.orig/arch/x86/kernel/microcode-xen.c 2009-05-14 11:02:43.000000000 +0200
1674 +++ sle11-2009-05-14/arch/x86/kernel/microcode-xen.c 2009-03-04 11:28:34.000000000 +0100
1675 @@ -2,6 +2,7 @@
1676 * Intel CPU Microcode Update Driver for Linux
1677 *
1678 * Copyright (C) 2000-2004 Tigran Aivazian
1679 + * 2006 Shaohua Li <shaohua.li@intel.com>
1680 *
1681 * This driver allows to upgrade microcode on Intel processors
1682 * belonging to IA-32 family - PentiumPro, Pentium II,
1683 @@ -33,7 +34,9 @@
1684 #include <linux/spinlock.h>
1685 #include <linux/mm.h>
1686 #include <linux/mutex.h>
1687 -#include <linux/syscalls.h>
1688 +#include <linux/cpu.h>
1689 +#include <linux/firmware.h>
1690 +#include <linux/platform_device.h>
1691
1692 #include <asm/msr.h>
1693 #include <asm/uaccess.h>
1694 @@ -55,12 +58,7 @@ module_param(verbose, int, 0644);
1695 /* no concurrent ->write()s are allowed on /dev/cpu/microcode */
1696 static DEFINE_MUTEX(microcode_mutex);
1697
1698 -static int microcode_open (struct inode *unused1, struct file *unused2)
1699 -{
1700 - return capable(CAP_SYS_RAWIO) ? 0 : -EPERM;
1701 -}
1702 -
1703 -
1704 +#ifdef CONFIG_MICROCODE_OLD_INTERFACE
1705 static int do_microcode_update (const void __user *ubuf, size_t len)
1706 {
1707 int err;
1708 @@ -85,6 +83,11 @@ static int do_microcode_update (const vo
1709 return err;
1710 }
1711
1712 +static int microcode_open (struct inode *unused1, struct file *unused2)
1713 +{
1714 + return capable(CAP_SYS_RAWIO) ? 0 : -EPERM;
1715 +}
1716 +
1717 static ssize_t microcode_write (struct file *file, const char __user *buf, size_t len, loff_t *ppos)
1718 {
1719 ssize_t ret;
1720 @@ -117,7 +120,7 @@ static struct miscdevice microcode_dev =
1721 .fops = &microcode_fops,
1722 };
1723
1724 -static int __init microcode_init (void)
1725 +static int __init microcode_dev_init (void)
1726 {
1727 int error;
1728
1729 @@ -129,6 +132,68 @@ static int __init microcode_init (void)
1730 return error;
1731 }
1732
1733 + return 0;
1734 +}
1735 +
1736 +static void __exit microcode_dev_exit (void)
1737 +{
1738 + misc_deregister(&microcode_dev);
1739 +}
1740 +
1741 +MODULE_ALIAS_MISCDEV(MICROCODE_MINOR);
1742 +#else
1743 +#define microcode_dev_init() 0
1744 +#define microcode_dev_exit() do { } while(0)
1745 +#endif
1746 +
1747 +/* fake device for request_firmware */
1748 +static struct platform_device *microcode_pdev;
1749 +
1750 +static int request_microcode(void)
1751 +{
1752 + char name[30];
1753 + const struct cpuinfo_x86 *c = &boot_cpu_data;
1754 + const struct firmware *firmware;
1755 + int error;
1756 + struct xen_platform_op op;
1757 +
1758 + sprintf(name,"intel-ucode/%02x-%02x-%02x",
1759 + c->x86, c->x86_model, c->x86_mask);
1760 + error = request_firmware(&firmware, name, &microcode_pdev->dev);
1761 + if (error) {
1762 + pr_debug("ucode data file %s load failed\n", name);
1763 + return error;
1764 + }
1765 +
1766 + op.cmd = XENPF_microcode_update;
1767 + set_xen_guest_handle(op.u.microcode.data, (void *)firmware->data);
1768 + op.u.microcode.length = firmware->size;
1769 + error = HYPERVISOR_platform_op(&op);
1770 +
1771 + release_firmware(firmware);
1772 +
1773 + if (error)
1774 + pr_debug("ucode load failed\n");
1775 +
1776 + return error;
1777 +}
1778 +
1779 +static int __init microcode_init (void)
1780 +{
1781 + int error;
1782 +
1783 + error = microcode_dev_init();
1784 + if (error)
1785 + return error;
1786 + microcode_pdev = platform_device_register_simple("microcode", -1,
1787 + NULL, 0);
1788 + if (IS_ERR(microcode_pdev)) {
1789 + microcode_dev_exit();
1790 + return PTR_ERR(microcode_pdev);
1791 + }
1792 +
1793 + request_microcode();
1794 +
1795 printk(KERN_INFO
1796 "IA-32 Microcode Update Driver: v" MICROCODE_VERSION " <tigran@veritas.com>\n");
1797 return 0;
1798 @@ -136,9 +201,9 @@ static int __init microcode_init (void)
1799
1800 static void __exit microcode_exit (void)
1801 {
1802 - misc_deregister(&microcode_dev);
1803 + microcode_dev_exit();
1804 + platform_device_unregister(microcode_pdev);
1805 }
1806
1807 module_init(microcode_init)
1808 module_exit(microcode_exit)
1809 -MODULE_ALIAS_MISCDEV(MICROCODE_MINOR);
1810 --- sle11-2009-05-14.orig/arch/x86/kernel/mpparse_32-xen.c 2009-05-14 11:02:43.000000000 +0200
1811 +++ sle11-2009-05-14/arch/x86/kernel/mpparse_32-xen.c 2009-03-04 11:28:34.000000000 +0100
1812 @@ -30,6 +30,7 @@
1813 #include <asm/io_apic.h>
1814
1815 #include <mach_apic.h>
1816 +#include <mach_apicdef.h>
1817 #include <mach_mpparse.h>
1818 #include <bios_ebda.h>
1819
1820 @@ -68,7 +69,7 @@ unsigned int def_to_bigsmp = 0;
1821 /* Processor that is doing the boot up */
1822 unsigned int boot_cpu_physical_apicid = -1U;
1823 /* Internal processor count */
1824 -static unsigned int __devinitdata num_processors;
1825 +unsigned int __cpuinitdata num_processors;
1826
1827 /* Bitmask of physically existing CPUs */
1828 physid_mask_t phys_cpu_present_map;
1829 @@ -235,12 +236,14 @@ static void __init MP_bus_info (struct m
1830
1831 mpc_oem_bus_info(m, str, translation_table[mpc_record]);
1832
1833 +#if MAX_MP_BUSSES < 256
1834 if (m->mpc_busid >= MAX_MP_BUSSES) {
1835 printk(KERN_WARNING "MP table busid value (%d) for bustype %s "
1836 " is too large, max. supported is %d\n",
1837 m->mpc_busid, str, MAX_MP_BUSSES - 1);
1838 return;
1839 }
1840 +#endif
1841
1842 if (strncmp(str, BUSTYPE_ISA, sizeof(BUSTYPE_ISA)-1) == 0) {
1843 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA;
1844 @@ -300,19 +303,6 @@ static void __init MP_lintsrc_info (stru
1845 m->mpc_irqtype, m->mpc_irqflag & 3,
1846 (m->mpc_irqflag >> 2) &3, m->mpc_srcbusid,
1847 m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint);
1848 - /*
1849 - * Well it seems all SMP boards in existence
1850 - * use ExtINT/LVT1 == LINT0 and
1851 - * NMI/LVT2 == LINT1 - the following check
1852 - * will show us if this assumptions is false.
1853 - * Until then we do not have to add baggage.
1854 - */
1855 - if ((m->mpc_irqtype == mp_ExtINT) &&
1856 - (m->mpc_destapiclint != 0))
1857 - BUG();
1858 - if ((m->mpc_irqtype == mp_NMI) &&
1859 - (m->mpc_destapiclint != 1))
1860 - BUG();
1861 }
1862
1863 #ifdef CONFIG_X86_NUMAQ
1864 @@ -838,8 +828,7 @@ int es7000_plat;
1865
1866 #ifdef CONFIG_ACPI
1867
1868 -void __init mp_register_lapic_address (
1869 - u64 address)
1870 +void __init mp_register_lapic_address(u64 address)
1871 {
1872 #ifndef CONFIG_XEN
1873 mp_lapic_addr = (unsigned long) address;
1874 @@ -853,13 +842,10 @@ void __init mp_register_lapic_address (
1875 #endif
1876 }
1877
1878 -
1879 -void __devinit mp_register_lapic (
1880 - u8 id,
1881 - u8 enabled)
1882 +void __devinit mp_register_lapic (u8 id, u8 enabled)
1883 {
1884 struct mpc_config_processor processor;
1885 - int boot_cpu = 0;
1886 + int boot_cpu = 0;
1887
1888 if (MAX_APICS - id <= 0) {
1889 printk(KERN_WARNING "Processor #%d invalid (max %d)\n",
1890 @@ -898,11 +884,9 @@ static struct mp_ioapic_routing {
1891 u32 pin_programmed[4];
1892 } mp_ioapic_routing[MAX_IO_APICS];
1893
1894 -
1895 -static int mp_find_ioapic (
1896 - int gsi)
1897 +static int mp_find_ioapic (int gsi)
1898 {
1899 - int i = 0;
1900 + int i = 0;
1901
1902 /* Find the IOAPIC that manages this GSI. */
1903 for (i = 0; i < nr_ioapics; i++) {
1904 @@ -915,15 +899,11 @@ static int mp_find_ioapic (
1905
1906 return -1;
1907 }
1908 -
1909
1910 -void __init mp_register_ioapic (
1911 - u8 id,
1912 - u32 address,
1913 - u32 gsi_base)
1914 +void __init mp_register_ioapic(u8 id, u32 address, u32 gsi_base)
1915 {
1916 - int idx = 0;
1917 - int tmpid;
1918 + int idx = 0;
1919 + int tmpid;
1920
1921 if (nr_ioapics >= MAX_IO_APICS) {
1922 printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded "
1923 @@ -971,16 +951,10 @@ void __init mp_register_ioapic (
1924 mp_ioapics[idx].mpc_apicver, mp_ioapics[idx].mpc_apicaddr,
1925 mp_ioapic_routing[idx].gsi_base,
1926 mp_ioapic_routing[idx].gsi_end);
1927 -
1928 - return;
1929 }
1930
1931 -
1932 -void __init mp_override_legacy_irq (
1933 - u8 bus_irq,
1934 - u8 polarity,
1935 - u8 trigger,
1936 - u32 gsi)
1937 +void __init
1938 +mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
1939 {
1940 struct mpc_config_intsrc intsrc;
1941 int ioapic = -1;
1942 @@ -1018,15 +992,13 @@ void __init mp_override_legacy_irq (
1943 mp_irqs[mp_irq_entries] = intsrc;
1944 if (++mp_irq_entries == MAX_IRQ_SOURCES)
1945 panic("Max # of irq sources exceeded!\n");
1946 -
1947 - return;
1948 }
1949
1950 void __init mp_config_acpi_legacy_irqs (void)
1951 {
1952 struct mpc_config_intsrc intsrc;
1953 - int i = 0;
1954 - int ioapic = -1;
1955 + int i = 0;
1956 + int ioapic = -1;
1957
1958 /*
1959 * Fabricate the legacy ISA bus (bus #31).
1960 @@ -1095,12 +1067,12 @@ void __init mp_config_acpi_legacy_irqs (
1961
1962 #define MAX_GSI_NUM 4096
1963
1964 -int mp_register_gsi (u32 gsi, int triggering, int polarity)
1965 +int mp_register_gsi(u32 gsi, int triggering, int polarity)
1966 {
1967 - int ioapic = -1;
1968 - int ioapic_pin = 0;
1969 - int idx, bit = 0;
1970 - static int pci_irq = 16;
1971 + int ioapic = -1;
1972 + int ioapic_pin = 0;
1973 + int idx, bit = 0;
1974 + static int pci_irq = 16;
1975 /*
1976 * Mapping between Global System Interrups, which
1977 * represent all possible interrupts, and IRQs
1978 --- sle11-2009-05-14.orig/arch/x86/kernel/pci-dma-xen.c 2009-05-14 11:02:43.000000000 +0200
1979 +++ sle11-2009-05-14/arch/x86/kernel/pci-dma-xen.c 2009-03-04 11:28:34.000000000 +0100
1980 @@ -110,8 +110,7 @@ dma_map_sg(struct device *hwdev, struct
1981 {
1982 int i, rc;
1983
1984 - if (direction == DMA_NONE)
1985 - BUG();
1986 + BUG_ON(!valid_dma_direction(direction));
1987 WARN_ON(nents == 0 || sg[0].length == 0);
1988
1989 if (swiotlb) {
1990 @@ -142,7 +141,7 @@ dma_unmap_sg(struct device *hwdev, struc
1991 {
1992 int i;
1993
1994 - BUG_ON(direction == DMA_NONE);
1995 + BUG_ON(!valid_dma_direction(direction));
1996 if (swiotlb)
1997 swiotlb_unmap_sg(hwdev, sg, nents, direction);
1998 else {
1999 @@ -159,8 +158,7 @@ dma_map_page(struct device *dev, struct
2000 {
2001 dma_addr_t dma_addr;
2002
2003 - BUG_ON(direction == DMA_NONE);
2004 -
2005 + BUG_ON(!valid_dma_direction(direction));
2006 if (swiotlb) {
2007 dma_addr = swiotlb_map_page(
2008 dev, page, offset, size, direction);
2009 @@ -177,7 +175,7 @@ void
2010 dma_unmap_page(struct device *dev, dma_addr_t dma_address, size_t size,
2011 enum dma_data_direction direction)
2012 {
2013 - BUG_ON(direction == DMA_NONE);
2014 + BUG_ON(!valid_dma_direction(direction));
2015 if (swiotlb)
2016 swiotlb_unmap_page(dev, dma_address, size, direction);
2017 else
2018 @@ -359,8 +357,7 @@ dma_map_single(struct device *dev, void
2019 {
2020 dma_addr_t dma;
2021
2022 - if (direction == DMA_NONE)
2023 - BUG();
2024 + BUG_ON(!valid_dma_direction(direction));
2025 WARN_ON(size == 0);
2026
2027 if (swiotlb) {
2028 @@ -381,8 +378,7 @@ void
2029 dma_unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size,
2030 enum dma_data_direction direction)
2031 {
2032 - if (direction == DMA_NONE)
2033 - BUG();
2034 + BUG_ON(!valid_dma_direction(direction));
2035 if (swiotlb)
2036 swiotlb_unmap_single(dev, dma_addr, size, direction);
2037 else
2038 --- sle11-2009-05-14.orig/arch/x86/kernel/process_32-xen.c 2009-05-14 11:02:43.000000000 +0200
2039 +++ sle11-2009-05-14/arch/x86/kernel/process_32-xen.c 2009-03-04 11:28:34.000000000 +0100
2040 @@ -37,6 +37,7 @@
2041 #include <linux/kallsyms.h>
2042 #include <linux/ptrace.h>
2043 #include <linux/random.h>
2044 +#include <linux/personality.h>
2045
2046 #include <asm/uaccess.h>
2047 #include <asm/pgtable.h>
2048 @@ -186,7 +187,7 @@ void cpu_idle(void)
2049 void cpu_idle_wait(void)
2050 {
2051 unsigned int cpu, this_cpu = get_cpu();
2052 - cpumask_t map;
2053 + cpumask_t map, tmp = current->cpus_allowed;
2054
2055 set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
2056 put_cpu();
2057 @@ -208,6 +209,8 @@ void cpu_idle_wait(void)
2058 }
2059 cpus_and(map, map, cpu_online_map);
2060 } while (!cpus_empty(map));
2061 +
2062 + set_cpus_allowed(current, tmp);
2063 }
2064 EXPORT_SYMBOL_GPL(cpu_idle_wait);
2065
2066 @@ -240,9 +243,9 @@ void show_regs(struct pt_regs * regs)
2067 if (user_mode_vm(regs))
2068 printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp);
2069 printk(" EFLAGS: %08lx %s (%s %.*s)\n",
2070 - regs->eflags, print_tainted(), system_utsname.release,
2071 - (int)strcspn(system_utsname.version, " "),
2072 - system_utsname.version);
2073 + regs->eflags, print_tainted(), init_utsname()->release,
2074 + (int)strcspn(init_utsname()->version, " "),
2075 + init_utsname()->version);
2076 printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
2077 regs->eax,regs->ebx,regs->ecx,regs->edx);
2078 printk("ESI: %08lx EDI: %08lx EBP: %08lx",
2079 @@ -264,15 +267,6 @@ void show_regs(struct pt_regs * regs)
2080 * the "args".
2081 */
2082 extern void kernel_thread_helper(void);
2083 -__asm__(".section .text\n"
2084 - ".align 4\n"
2085 - "kernel_thread_helper:\n\t"
2086 - "movl %edx,%eax\n\t"
2087 - "pushl %edx\n\t"
2088 - "call *%ebx\n\t"
2089 - "pushl %eax\n\t"
2090 - "call do_exit\n"
2091 - ".previous");
2092
2093 /*
2094 * Create a kernel thread
2095 @@ -290,7 +284,7 @@ int kernel_thread(int (*fn)(void *), voi
2096 regs.xes = __USER_DS;
2097 regs.orig_eax = -1;
2098 regs.eip = (unsigned long) kernel_thread_helper;
2099 - regs.xcs = GET_KERNEL_CS();
2100 + regs.xcs = __KERNEL_CS | get_kernel_rpl();
2101 regs.eflags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2;
2102
2103 /* Ok, create the new process.. */
2104 @@ -369,13 +363,12 @@ int copy_thread(int nr, unsigned long cl
2105
2106 tsk = current;
2107 if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
2108 - p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL);
2109 + p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
2110 + IO_BITMAP_BYTES, GFP_KERNEL);
2111 if (!p->thread.io_bitmap_ptr) {
2112 p->thread.io_bitmap_max = 0;
2113 return -ENOMEM;
2114 }
2115 - memcpy(p->thread.io_bitmap_ptr, tsk->thread.io_bitmap_ptr,
2116 - IO_BITMAP_BYTES);
2117 set_tsk_thread_flag(p, TIF_IO_BITMAP);
2118 }
2119
2120 @@ -871,7 +864,7 @@ asmlinkage int sys_get_thread_area(struc
2121
2122 unsigned long arch_align_stack(unsigned long sp)
2123 {
2124 - if (randomize_va_space)
2125 + if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
2126 sp -= get_random_int() % 8192;
2127 return sp & ~0xf;
2128 }
2129 --- sle11-2009-05-14.orig/arch/x86/kernel/setup_32-xen.c 2009-05-14 11:02:43.000000000 +0200
2130 +++ sle11-2009-05-14/arch/x86/kernel/setup_32-xen.c 2009-03-04 11:28:34.000000000 +0100
2131 @@ -56,6 +56,7 @@
2132 #include <asm/apic.h>
2133 #include <asm/e820.h>
2134 #include <asm/mpspec.h>
2135 +#include <asm/mmzone.h>
2136 #include <asm/setup.h>
2137 #include <asm/arch_hooks.h>
2138 #include <asm/sections.h>
2139 @@ -83,9 +84,6 @@ static struct notifier_block xen_panic_b
2140 xen_panic_event, NULL, 0 /* try to go last */
2141 };
2142
2143 -extern char hypercall_page[PAGE_SIZE];
2144 -EXPORT_SYMBOL(hypercall_page);
2145 -
2146 int disable_pse __devinitdata = 0;
2147
2148 /*
2149 @@ -105,18 +103,6 @@ EXPORT_SYMBOL(boot_cpu_data);
2150
2151 unsigned long mmu_cr4_features;
2152
2153 -#ifdef CONFIG_ACPI
2154 - int acpi_disabled = 0;
2155 -#else
2156 - int acpi_disabled = 1;
2157 -#endif
2158 -EXPORT_SYMBOL(acpi_disabled);
2159 -
2160 -#ifdef CONFIG_ACPI
2161 -int __initdata acpi_force = 0;
2162 -extern acpi_interrupt_flags acpi_sci_flags;
2163 -#endif
2164 -
2165 /* for MCA, but anyone else can use it if they want */
2166 unsigned int machine_id;
2167 #ifdef CONFIG_MCA
2168 @@ -170,7 +156,6 @@ struct e820map machine_e820;
2169 #endif
2170
2171 extern void early_cpu_init(void);
2172 -extern void generic_apic_probe(char *);
2173 extern int root_mountflags;
2174
2175 unsigned long saved_videomode;
2176 @@ -243,9 +228,6 @@ static struct resource adapter_rom_resou
2177 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
2178 } };
2179
2180 -#define ADAPTER_ROM_RESOURCES \
2181 - (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0])
2182 -
2183 static struct resource video_rom_resource = {
2184 .name = "Video ROM",
2185 .start = 0xc0000,
2186 @@ -307,9 +289,6 @@ static struct resource standard_io_resou
2187 .flags = IORESOURCE_BUSY | IORESOURCE_IO
2188 } };
2189
2190 -#define STANDARD_IO_RESOURCES \
2191 - (sizeof standard_io_resources / sizeof standard_io_resources[0])
2192 -
2193 #define romsignature(x) (*(unsigned short *)(x) == 0xaa55)
2194
2195 static int __init romchecksum(unsigned char *rom, unsigned long length)
2196 @@ -372,7 +351,7 @@ static void __init probe_roms(void)
2197 }
2198
2199 /* check for adapter roms on 2k boundaries */
2200 - for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) {
2201 + for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper; start += 2048) {
2202 rom = isa_bus_to_virt(start);
2203 if (!romsignature(rom))
2204 continue;
2205 @@ -779,246 +758,152 @@ static inline void copy_edd(void)
2206 }
2207 #endif
2208
2209 -static void __init parse_cmdline_early (char ** cmdline_p)
2210 +static int __initdata user_defined_memmap = 0;
2211 +
2212 +/*
2213 + * "mem=nopentium" disables the 4MB page tables.
2214 + * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
2215 + * to <mem>, overriding the bios size.
2216 + * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
2217 + * <start> to <start>+<mem>, overriding the bios size.
2218 + *
2219 + * HPA tells me bootloaders need to parse mem=, so no new
2220 + * option should be mem= [also see Documentation/i386/boot.txt]
2221 + */
2222 +static int __init parse_mem(char *arg)
2223 {
2224 - char c = ' ', *to = command_line, *from = saved_command_line;
2225 - int len = 0, max_cmdline;
2226 - int userdef = 0;
2227 -
2228 - if ((max_cmdline = MAX_GUEST_CMDLINE) > COMMAND_LINE_SIZE)
2229 - max_cmdline = COMMAND_LINE_SIZE;
2230 - memcpy(saved_command_line, xen_start_info->cmd_line, max_cmdline);
2231 - /* Save unparsed command line copy for /proc/cmdline */
2232 - saved_command_line[max_cmdline-1] = '\0';
2233 -
2234 - for (;;) {
2235 - if (c != ' ')
2236 - goto next_char;
2237 - /*
2238 - * "mem=nopentium" disables the 4MB page tables.
2239 - * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
2240 - * to <mem>, overriding the bios size.
2241 - * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
2242 - * <start> to <start>+<mem>, overriding the bios size.
2243 - *
2244 - * HPA tells me bootloaders need to parse mem=, so no new
2245 - * option should be mem= [also see Documentation/i386/boot.txt]
2246 - */
2247 - if (!memcmp(from, "mem=", 4)) {
2248 - if (to != command_line)
2249 - to--;
2250 - if (!memcmp(from+4, "nopentium", 9)) {
2251 - from += 9+4;
2252 - clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability);
2253 - disable_pse = 1;
2254 - } else {
2255 - /* If the user specifies memory size, we
2256 - * limit the BIOS-provided memory map to
2257 - * that size. exactmap can be used to specify
2258 - * the exact map. mem=number can be used to
2259 - * trim the existing memory map.
2260 - */
2261 - unsigned long long mem_size;
2262 -
2263 - mem_size = memparse(from+4, &from);
2264 - limit_regions(mem_size);
2265 - userdef=1;
2266 - }
2267 - }
2268 + if (!arg)
2269 + return -EINVAL;
2270
2271 - else if (!memcmp(from, "memmap=", 7)) {
2272 - if (to != command_line)
2273 - to--;
2274 - if (!memcmp(from+7, "exactmap", 8)) {
2275 -#ifdef CONFIG_CRASH_DUMP
2276 - /* If we are doing a crash dump, we
2277 - * still need to know the real mem
2278 - * size before original memory map is
2279 - * reset.
2280 - */
2281 - find_max_pfn();
2282 - saved_max_pfn = max_pfn;
2283 -#endif
2284 - from += 8+7;
2285 - e820.nr_map = 0;
2286 - userdef = 1;
2287 - } else {
2288 - /* If the user specifies memory size, we
2289 - * limit the BIOS-provided memory map to
2290 - * that size. exactmap can be used to specify
2291 - * the exact map. mem=number can be used to
2292 - * trim the existing memory map.
2293 - */
2294 - unsigned long long start_at, mem_size;
2295 + if (strcmp(arg, "nopentium") == 0) {
2296 + clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability);
2297 + disable_pse = 1;
2298 + } else {
2299 + /* If the user specifies memory size, we
2300 + * limit the BIOS-provided memory map to
2301 + * that size. exactmap can be used to specify
2302 + * the exact map. mem=number can be used to
2303 + * trim the existing memory map.
2304 + */
2305 + unsigned long long mem_size;
2306
2307 - mem_size = memparse(from+7, &from);
2308 - if (*from == '@') {
2309 - start_at = memparse(from+1, &from);
2310 - add_memory_region(start_at, mem_size, E820_RAM);
2311 - } else if (*from == '#') {
2312 - start_at = memparse(from+1, &from);
2313 - add_memory_region(start_at, mem_size, E820_ACPI);
2314 - } else if (*from == '$') {
2315 - start_at = memparse(from+1, &from);
2316 - add_memory_region(start_at, mem_size, E820_RESERVED);
2317 - } else {
2318 - limit_regions(mem_size);
2319 - userdef=1;
2320 - }
2321 - }
2322 - }
2323 -
2324 - else if (!memcmp(from, "noexec=", 7))
2325 - noexec_setup(from + 7);
2326 + mem_size = memparse(arg, &arg);
2327 + limit_regions(mem_size);
2328 + user_defined_memmap = 1;
2329 + }
2330 + return 0;
2331 +}
2332 +early_param("mem", parse_mem);
2333
2334 +static int __init parse_memmap(char *arg)
2335 +{
2336 + if (!arg)
2337 + return -EINVAL;
2338
2339 -#ifdef CONFIG_X86_MPPARSE
2340 - /*
2341 - * If the BIOS enumerates physical processors before logical,
2342 - * maxcpus=N at enumeration-time can be used to disable HT.
2343 + if (strcmp(arg, "exactmap") == 0) {
2344 +#ifdef CONFIG_CRASH_DUMP
2345 + /* If we are doing a crash dump, we
2346 + * still need to know the real mem
2347 + * size before original memory map is
2348 + * reset.
2349 */
2350 - else if (!memcmp(from, "maxcpus=", 8)) {
2351 - extern unsigned int maxcpus;
2352 -
2353 - maxcpus = simple_strtoul(from + 8, NULL, 0);
2354 - }
2355 + find_max_pfn();
2356 + saved_max_pfn = max_pfn;
2357 #endif
2358 + e820.nr_map = 0;
2359 + user_defined_memmap = 1;
2360 + } else {
2361 + /* If the user specifies memory size, we
2362 + * limit the BIOS-provided memory map to
2363 + * that size. exactmap can be used to specify
2364 + * the exact map. mem=number can be used to
2365 + * trim the existing memory map.
2366 + */
2367 + unsigned long long start_at, mem_size;
2368
2369 -#ifdef CONFIG_ACPI
2370 - /* "acpi=off" disables both ACPI table parsing and interpreter */
2371 - else if (!memcmp(from, "acpi=off", 8)) {
2372 - disable_acpi();
2373 - }
2374 -
2375 - /* acpi=force to over-ride black-list */
2376 - else if (!memcmp(from, "acpi=force", 10)) {
2377 - acpi_force = 1;
2378 - acpi_ht = 1;
2379 - acpi_disabled = 0;
2380 - }
2381 -
2382 - /* acpi=strict disables out-of-spec workarounds */
2383 - else if (!memcmp(from, "acpi=strict", 11)) {
2384 - acpi_strict = 1;
2385 - }
2386 -
2387 - /* Limit ACPI just to boot-time to enable HT */
2388 - else if (!memcmp(from, "acpi=ht", 7)) {
2389 - if (!acpi_force)
2390 - disable_acpi();
2391 - acpi_ht = 1;
2392 - }
2393 -
2394 - /* "pci=noacpi" disable ACPI IRQ routing and PCI scan */
2395 - else if (!memcmp(from, "pci=noacpi", 10)) {
2396 - acpi_disable_pci();
2397 - }
2398 - /* "acpi=noirq" disables ACPI interrupt routing */
2399 - else if (!memcmp(from, "acpi=noirq", 10)) {
2400 - acpi_noirq_set();
2401 + mem_size = memparse(arg, &arg);
2402 + if (*arg == '@') {
2403 + start_at = memparse(arg+1, &arg);
2404 + add_memory_region(start_at, mem_size, E820_RAM);
2405 + } else if (*arg == '#') {
2406 + start_at = memparse(arg+1, &arg);
2407 + add_memory_region(start_at, mem_size, E820_ACPI);
2408 + } else if (*arg == '$') {
2409 + start_at = memparse(arg+1, &arg);
2410 + add_memory_region(start_at, mem_size, E820_RESERVED);
2411 + } else {
2412 + limit_regions(mem_size);
2413 + user_defined_memmap = 1;
2414 }
2415 + }
2416 + return 0;
2417 +}
2418 +early_param("memmap", parse_memmap);
2419
2420 - else if (!memcmp(from, "acpi_sci=edge", 13))
2421 - acpi_sci_flags.trigger = 1;
2422 -
2423 - else if (!memcmp(from, "acpi_sci=level", 14))
2424 - acpi_sci_flags.trigger = 3;
2425 +#ifdef CONFIG_PROC_VMCORE
2426 +/* elfcorehdr= specifies the location of elf core header
2427 + * stored by the crashed kernel.
2428 + */
2429 +static int __init parse_elfcorehdr(char *arg)
2430 +{
2431 + if (!arg)
2432 + return -EINVAL;
2433
2434 - else if (!memcmp(from, "acpi_sci=high", 13))
2435 - acpi_sci_flags.polarity = 1;
2436 + elfcorehdr_addr = memparse(arg, &arg);
2437 + return 0;
2438 +}
2439 +early_param("elfcorehdr", parse_elfcorehdr);
2440 +#endif /* CONFIG_PROC_VMCORE */
2441
2442 - else if (!memcmp(from, "acpi_sci=low", 12))
2443 - acpi_sci_flags.polarity = 3;
2444 +/*
2445 + * highmem=size forces highmem to be exactly 'size' bytes.
2446 + * This works even on boxes that have no highmem otherwise.
2447 + * This also works to reduce highmem size on bigger boxes.
2448 + */
2449 +static int __init parse_highmem(char *arg)
2450 +{
2451 + if (!arg)
2452 + return -EINVAL;
2453
2454 -#ifdef CONFIG_X86_IO_APIC
2455 - else if (!memcmp(from, "acpi_skip_timer_override", 24))
2456 - acpi_skip_timer_override = 1;
2457 + highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
2458 + return 0;
2459 +}
2460 +early_param("highmem", parse_highmem);
2461
2462 - if (!memcmp(from, "disable_timer_pin_1", 19))
2463 - disable_timer_pin_1 = 1;
2464 - if (!memcmp(from, "enable_timer_pin_1", 18))
2465 - disable_timer_pin_1 = -1;
2466 -
2467 - /* disable IO-APIC */
2468 - else if (!memcmp(from, "noapic", 6))
2469 - disable_ioapic_setup();
2470 -#endif /* CONFIG_X86_IO_APIC */
2471 -#endif /* CONFIG_ACPI */
2472 +/*
2473 + * vmalloc=size forces the vmalloc area to be exactly 'size'
2474 + * bytes. This can be used to increase (or decrease) the
2475 + * vmalloc area - the default is 128m.
2476 + */
2477 +static int __init parse_vmalloc(char *arg)
2478 +{
2479 + if (!arg)
2480 + return -EINVAL;
2481
2482 -#ifdef CONFIG_X86_LOCAL_APIC
2483 - /* enable local APIC */
2484 - else if (!memcmp(from, "lapic", 5))
2485 - lapic_enable();
2486 -
2487 - /* disable local APIC */
2488 - else if (!memcmp(from, "nolapic", 6))
2489 - lapic_disable();
2490 -#endif /* CONFIG_X86_LOCAL_APIC */
2491 + __VMALLOC_RESERVE = memparse(arg, &arg);
2492 + return 0;
2493 +}
2494 +early_param("vmalloc", parse_vmalloc);
2495
2496 -#ifdef CONFIG_KEXEC
2497 - /* crashkernel=size@addr specifies the location to reserve for
2498 - * a crash kernel. By reserving this memory we guarantee
2499 - * that linux never set's it up as a DMA target.
2500 - * Useful for holding code to do something appropriate
2501 - * after a kernel panic.
2502 - */
2503 - else if (!memcmp(from, "crashkernel=", 12)) {
2504 #ifndef CONFIG_XEN
2505 - unsigned long size, base;
2506 - size = memparse(from+12, &from);
2507 - if (*from == '@') {
2508 - base = memparse(from+1, &from);
2509 - /* FIXME: Do I want a sanity check
2510 - * to validate the memory range?
2511 - */
2512 - crashk_res.start = base;
2513 - crashk_res.end = base + size - 1;
2514 - }
2515 -#else
2516 - printk("Ignoring crashkernel command line, "
2517 - "parameter will be supplied by xen\n");
2518 -#endif
2519 - }
2520 -#endif
2521 -#ifdef CONFIG_PROC_VMCORE
2522 - /* elfcorehdr= specifies the location of elf core header
2523 - * stored by the crashed kernel.
2524 - */
2525 - else if (!memcmp(from, "elfcorehdr=", 11))
2526 - elfcorehdr_addr = memparse(from+11, &from);
2527 -#endif
2528 +/*
2529 + * reservetop=size reserves a hole at the top of the kernel address space which
2530 + * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
2531 + * so relocating the fixmap can be done before paging initialization.
2532 + */
2533 +static int __init parse_reservetop(char *arg)
2534 +{
2535 + unsigned long address;
2536
2537 - /*
2538 - * highmem=size forces highmem to be exactly 'size' bytes.
2539 - * This works even on boxes that have no highmem otherwise.
2540 - * This also works to reduce highmem size on bigger boxes.
2541 - */
2542 - else if (!memcmp(from, "highmem=", 8))
2543 - highmem_pages = memparse(from+8, &from) >> PAGE_SHIFT;
2544 -
2545 - /*
2546 - * vmalloc=size forces the vmalloc area to be exactly 'size'
2547 - * bytes. This can be used to increase (or decrease) the
2548 - * vmalloc area - the default is 128m.
2549 - */
2550 - else if (!memcmp(from, "vmalloc=", 8))
2551 - __VMALLOC_RESERVE = memparse(from+8, &from);
2552 + if (!arg)
2553 + return -EINVAL;
2554
2555 - next_char:
2556 - c = *(from++);
2557 - if (!c)
2558 - break;
2559 - if (COMMAND_LINE_SIZE <= ++len)
2560 - break;
2561 - *(to++) = c;
2562 - }
2563 - *to = '\0';
2564 - *cmdline_p = command_line;
2565 - if (userdef) {
2566 - printk(KERN_INFO "user-defined physical RAM map:\n");
2567 - print_memory_map("user");
2568 - }
2569 + address = memparse(arg, &arg);
2570 + reserve_top_address(address);
2571 + return 0;
2572 }
2573 +early_param("reservetop", parse_reservetop);
2574 +#endif
2575
2576 /*
2577 * Callback for efi_memory_walk.
2578 @@ -1039,7 +924,7 @@ efi_find_max_pfn(unsigned long start, un
2579 static int __init
2580 efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg)
2581 {
2582 - memory_present(0, start, end);
2583 + memory_present(0, PFN_UP(start), PFN_DOWN(end));
2584 return 0;
2585 }
2586
2587 @@ -1306,6 +1191,14 @@ static unsigned long __init setup_memory
2588 }
2589 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
2590 pages_to_mb(highend_pfn - highstart_pfn));
2591 + num_physpages = highend_pfn;
2592 + high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
2593 +#else
2594 + num_physpages = max_low_pfn;
2595 + high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
2596 +#endif
2597 +#ifdef CONFIG_FLATMEM
2598 + max_mapnr = num_physpages;
2599 #endif
2600 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
2601 pages_to_mb(max_low_pfn));
2602 @@ -1317,22 +1210,19 @@ static unsigned long __init setup_memory
2603
2604 void __init zone_sizes_init(void)
2605 {
2606 - unsigned long zones_size[MAX_NR_ZONES] = {0, 0, 0};
2607 - unsigned int max_dma, low;
2608 -
2609 - max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
2610 - low = max_low_pfn;
2611 -
2612 - if (low < max_dma)
2613 - zones_size[ZONE_DMA] = low;
2614 - else {
2615 - zones_size[ZONE_DMA] = max_dma;
2616 - zones_size[ZONE_NORMAL] = low - max_dma;
2617 + unsigned long max_zone_pfns[MAX_NR_ZONES];
2618 + memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
2619 + max_zone_pfns[ZONE_DMA] =
2620 + virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
2621 + max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
2622 #ifdef CONFIG_HIGHMEM
2623 - zones_size[ZONE_HIGHMEM] = highend_pfn - low;
2624 + max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
2625 + add_active_range(0, 0, highend_pfn);
2626 +#else
2627 + add_active_range(0, 0, max_low_pfn);
2628 #endif
2629 - }
2630 - free_area_init(zones_size);
2631 +
2632 + free_area_init_nodes(max_zone_pfns);
2633 }
2634 #else
2635 extern unsigned long __init setup_memory(void);
2636 @@ -1389,6 +1279,7 @@ void __init setup_bootmem_allocator(void
2637 */
2638 acpi_reserve_bootmem();
2639 #endif
2640 + numa_kva_reserve();
2641 #endif /* !CONFIG_XEN */
2642
2643 #ifdef CONFIG_BLK_DEV_INITRD
2644 @@ -1574,7 +1465,7 @@ static int __init request_standard_resou
2645 request_resource(&iomem_resource, &video_ram_resource);
2646
2647 /* request I/O space for devices used on all i[345]86 PCs */
2648 - for (i = 0; i < STANDARD_IO_RESOURCES; i++)
2649 + for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
2650 request_resource(&ioport_resource, &standard_io_resources[i]);
2651 return 0;
2652 }
2653 @@ -1705,17 +1596,19 @@ void __init setup_arch(char **cmdline_p)
2654 data_resource.start = virt_to_phys(_etext);
2655 data_resource.end = virt_to_phys(_edata)-1;
2656
2657 - parse_cmdline_early(cmdline_p);
2658 + if ((i = MAX_GUEST_CMDLINE) > COMMAND_LINE_SIZE)
2659 + i = COMMAND_LINE_SIZE;
2660 + memcpy(saved_command_line, xen_start_info->cmd_line, i);
2661 + saved_command_line[i - 1] = '\0';
2662 + parse_early_param();
2663
2664 -#ifdef CONFIG_EARLY_PRINTK
2665 - {
2666 - char *s = strstr(*cmdline_p, "earlyprintk=");
2667 - if (s) {
2668 - setup_early_printk(strchr(s, '=') + 1);
2669 - printk("early console enabled\n");
2670 - }
2671 + if (user_defined_memmap) {
2672 + printk(KERN_INFO "user-defined physical RAM map:\n");
2673 + print_memory_map("user");
2674 }
2675 -#endif
2676 +
2677 + strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE);
2678 + *cmdline_p = command_line;
2679
2680 max_low_pfn = setup_memory();
2681
2682 @@ -1822,7 +1715,7 @@ void __init setup_arch(char **cmdline_p)
2683 dmi_scan_machine();
2684
2685 #ifdef CONFIG_X86_GENERICARCH
2686 - generic_apic_probe(*cmdline_p);
2687 + generic_apic_probe();
2688 #endif
2689 if (efi_enabled)
2690 efi_map_memmap();
2691 @@ -1843,9 +1736,11 @@ void __init setup_arch(char **cmdline_p)
2692 acpi_boot_table_init();
2693 #endif
2694
2695 +#ifdef CONFIG_PCI
2696 #ifdef CONFIG_X86_IO_APIC
2697 check_acpi_pci(); /* Checks more than just ACPI actually */
2698 #endif
2699 +#endif
2700
2701 #ifdef CONFIG_ACPI
2702 acpi_boot_init();
2703 --- sle11-2009-05-14.orig/arch/x86/kernel/smp_32-xen.c 2009-05-14 11:02:43.000000000 +0200
2704 +++ sle11-2009-05-14/arch/x86/kernel/smp_32-xen.c 2009-03-04 11:28:34.000000000 +0100
2705 @@ -279,8 +279,7 @@ static inline void leave_mm (unsigned lo
2706 * 2) Leave the mm if we are in the lazy tlb mode.
2707 */
2708
2709 -irqreturn_t smp_invalidate_interrupt(int irq, void *dev_id,
2710 - struct pt_regs *regs)
2711 +irqreturn_t smp_invalidate_interrupt(int irq, void *dev_id)
2712 {
2713 unsigned long cpu;
2714
2715 @@ -567,16 +566,14 @@ void smp_send_stop(void)
2716 * all the work is done automatically when
2717 * we return from the interrupt.
2718 */
2719 -irqreturn_t smp_reschedule_interrupt(int irq, void *dev_id,
2720 - struct pt_regs *regs)
2721 +irqreturn_t smp_reschedule_interrupt(int irq, void *dev_id)
2722 {
2723
2724 return IRQ_HANDLED;
2725 }
2726
2727 #include <linux/kallsyms.h>
2728 -irqreturn_t smp_call_function_interrupt(int irq, void *dev_id,
2729 - struct pt_regs *regs)
2730 +irqreturn_t smp_call_function_interrupt(int irq, void *dev_id)
2731 {
2732 void (*func) (void *info) = call_data->func;
2733 void *info = call_data->info;
2734 @@ -603,3 +600,69 @@ irqreturn_t smp_call_function_interrupt(
2735 return IRQ_HANDLED;
2736 }
2737
2738 +/*
2739 + * this function sends a 'generic call function' IPI to one other CPU
2740 + * in the system.
2741 + *
2742 + * cpu is a standard Linux logical CPU number.
2743 + */
2744 +static void
2745 +__smp_call_function_single(int cpu, void (*func) (void *info), void *info,
2746 + int nonatomic, int wait)
2747 +{
2748 + struct call_data_struct data;
2749 + int cpus = 1;
2750 +
2751 + data.func = func;
2752 + data.info = info;
2753 + atomic_set(&data.started, 0);
2754 + data.wait = wait;
2755 + if (wait)
2756 + atomic_set(&data.finished, 0);
2757 +
2758 + call_data = &data;
2759 + wmb();
2760 + /* Send a message to all other CPUs and wait for them to respond */
2761 + send_IPI_mask(cpumask_of_cpu(cpu), CALL_FUNCTION_VECTOR);
2762 +
2763 + /* Wait for response */
2764 + while (atomic_read(&data.started) != cpus)
2765 + cpu_relax();
2766 +
2767 + if (!wait)
2768 + return;
2769 +
2770 + while (atomic_read(&data.finished) != cpus)
2771 + cpu_relax();
2772 +}
2773 +
2774 +/*
2775 + * smp_call_function_single - Run a function on another CPU
2776 + * @func: The function to run. This must be fast and non-blocking.
2777 + * @info: An arbitrary pointer to pass to the function.
2778 + * @nonatomic: Currently unused.
2779 + * @wait: If true, wait until function has completed on other CPUs.
2780 + *
2781 + * Retrurns 0 on success, else a negative status code.
2782 + *
2783 + * Does not return until the remote CPU is nearly ready to execute <func>
2784 + * or is or has executed.
2785 + */
2786 +
2787 +int smp_call_function_single(int cpu, void (*func) (void *info), void *info,
2788 + int nonatomic, int wait)
2789 +{
2790 + /* prevent preemption and reschedule on another processor */
2791 + int me = get_cpu();
2792 + if (cpu == me) {
2793 + WARN_ON(1);
2794 + put_cpu();
2795 + return -EBUSY;
2796 + }
2797 + spin_lock_bh(&call_lock);
2798 + __smp_call_function_single(cpu, func, info, nonatomic, wait);
2799 + spin_unlock_bh(&call_lock);
2800 + put_cpu();
2801 + return 0;
2802 +}
2803 +EXPORT_SYMBOL(smp_call_function_single);
2804 --- sle11-2009-05-14.orig/arch/x86/kernel/time_32-xen.c 2009-04-20 11:36:10.000000000 +0200
2805 +++ sle11-2009-05-14/arch/x86/kernel/time_32-xen.c 2009-03-24 10:08:00.000000000 +0100
2806 @@ -89,7 +89,6 @@ int pit_latch_buggy; /* ext
2807 unsigned long vxtime_hz = PIT_TICK_RATE;
2808 struct vxtime_data __vxtime __section_vxtime; /* for vsyscalls */
2809 volatile unsigned long __jiffies __section_jiffies = INITIAL_JIFFIES;
2810 -unsigned long __wall_jiffies __section_wall_jiffies = INITIAL_JIFFIES;
2811 struct timespec __xtime __section_xtime;
2812 struct timezone __sys_tz __section_sys_tz;
2813 #endif
2814 @@ -97,8 +96,6 @@ struct timezone __sys_tz __section_sys_t
2815 unsigned int cpu_khz; /* Detected as we calibrate the TSC */
2816 EXPORT_SYMBOL(cpu_khz);
2817
2818 -extern unsigned long wall_jiffies;
2819 -
2820 DEFINE_SPINLOCK(rtc_lock);
2821 EXPORT_SYMBOL(rtc_lock);
2822
2823 @@ -265,11 +262,10 @@ static void __update_wallclock(time_t se
2824 time_t wtm_sec, xtime_sec;
2825 u64 tmp, wc_nsec;
2826
2827 - /* Adjust wall-clock time base based on wall_jiffies ticks. */
2828 + /* Adjust wall-clock time base. */
2829 wc_nsec = processed_system_time;
2830 wc_nsec += sec * (u64)NSEC_PER_SEC;
2831 wc_nsec += nsec;
2832 - wc_nsec -= (jiffies - wall_jiffies) * (u64)NS_PER_TICK;
2833
2834 /* Split wallclock base into seconds and nanoseconds. */
2835 tmp = wc_nsec;
2836 @@ -392,16 +388,10 @@ void do_gettimeofday(struct timeval *tv)
2837 shadow = &per_cpu(shadow_time, cpu);
2838
2839 do {
2840 - unsigned long lost;
2841 -
2842 local_time_version = shadow->version;
2843 seq = read_seqbegin(&xtime_lock);
2844
2845 usec = get_usec_offset(shadow);
2846 - lost = jiffies - wall_jiffies;
2847 -
2848 - if (unlikely(lost))
2849 - usec += lost * (USEC_PER_SEC / HZ);
2850
2851 sec = xtime.tv_sec;
2852 usec += (xtime.tv_nsec / NSEC_PER_USEC);
2853 @@ -525,7 +515,7 @@ static void sync_xen_wallclock(unsigned
2854 write_seqlock_irq(&xtime_lock);
2855
2856 sec = xtime.tv_sec;
2857 - nsec = xtime.tv_nsec + ((jiffies - wall_jiffies) * (u64)NS_PER_TICK);
2858 + nsec = xtime.tv_nsec;
2859 __normalize_time(&sec, &nsec);
2860
2861 op.cmd = XENPF_settime;
2862 @@ -599,42 +589,49 @@ unsigned long long sched_clock(void)
2863 }
2864 #endif
2865
2866 -#if defined(CONFIG_SMP) && defined(CONFIG_FRAME_POINTER)
2867 unsigned long profile_pc(struct pt_regs *regs)
2868 {
2869 unsigned long pc = instruction_pointer(regs);
2870
2871 -#ifdef __x86_64__
2872 - /* Assume the lock function has either no stack frame or only a single word.
2873 - This checks if the address on the stack looks like a kernel text address.
2874 - There is a small window for false hits, but in that case the tick
2875 - is just accounted to the spinlock function.
2876 - Better would be to write these functions in assembler again
2877 - and check exactly. */
2878 +#if defined(CONFIG_SMP) || defined(__x86_64__)
2879 if (!user_mode_vm(regs) && in_lock_functions(pc)) {
2880 - char *v = *(char **)regs->rsp;
2881 - if ((v >= _stext && v <= _etext) ||
2882 - (v >= _sinittext && v <= _einittext) ||
2883 - (v >= (char *)MODULES_VADDR && v <= (char *)MODULES_END))
2884 - return (unsigned long)v;
2885 - return ((unsigned long *)regs->rsp)[1];
2886 +# ifdef CONFIG_FRAME_POINTER
2887 +# ifdef __i386__
2888 + return ((unsigned long *)regs->ebp)[1];
2889 +# else
2890 + return ((unsigned long *)regs->rbp)[1];
2891 +# endif
2892 +# else
2893 +# ifdef __i386__
2894 + unsigned long *sp;
2895 + if ((regs->xcs & 2) == 0)
2896 + sp = (unsigned long *)&regs->esp;
2897 + else
2898 + sp = (unsigned long *)regs->esp;
2899 +# else
2900 + unsigned long *sp = (unsigned long *)regs->rsp;
2901 +# endif
2902 + /* Return address is either directly at stack pointer
2903 + or above a saved eflags. Eflags has bits 22-31 zero,
2904 + kernel addresses don't. */
2905 + if (sp[0] >> 22)
2906 + return sp[0];
2907 + if (sp[1] >> 22)
2908 + return sp[1];
2909 +# endif
2910 }
2911 -#else
2912 - if (!user_mode_vm(regs) && in_lock_functions(pc))
2913 - return *(unsigned long *)(regs->ebp + 4);
2914 #endif
2915
2916 return pc;
2917 }
2918 EXPORT_SYMBOL(profile_pc);
2919 -#endif
2920
2921 /*
2922 * This is the same as the above, except we _also_ save the current
2923 * Time Stamp Counter value at the time of the timer interrupt, so that
2924 * we later on can estimate the time of day more exactly.
2925 */
2926 -irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
2927 +irqreturn_t timer_interrupt(int irq, void *dev_id)
2928 {
2929 s64 delta, delta_cpu, stolen, blocked;
2930 u64 sched_time;
2931 @@ -692,10 +689,14 @@ irqreturn_t timer_interrupt(int irq, voi
2932 }
2933
2934 /* System-wide jiffy work. */
2935 - while (delta >= NS_PER_TICK) {
2936 - delta -= NS_PER_TICK;
2937 - processed_system_time += NS_PER_TICK;
2938 - do_timer(regs);
2939 + if (delta >= NS_PER_TICK) {
2940 + do_div(delta, NS_PER_TICK);
2941 + processed_system_time += delta * NS_PER_TICK;
2942 + while (delta > HZ) {
2943 + do_timer(HZ);
2944 + delta -= HZ;
2945 + }
2946 + do_timer(delta);
2947 }
2948
2949 if (shadow_tv_version != HYPERVISOR_shared_info->wc_version) {
2950 @@ -740,7 +741,7 @@ irqreturn_t timer_interrupt(int irq, voi
2951 if (delta_cpu > 0) {
2952 do_div(delta_cpu, NS_PER_TICK);
2953 per_cpu(processed_system_time, cpu) += delta_cpu * NS_PER_TICK;
2954 - if (user_mode_vm(regs))
2955 + if (user_mode_vm(get_irq_regs()))
2956 account_user_time(current, (cputime_t)delta_cpu);
2957 else
2958 account_system_time(current, HARDIRQ_OFFSET,
2959 @@ -754,10 +755,10 @@ irqreturn_t timer_interrupt(int irq, voi
2960 /* Local timer processing (see update_process_times()). */
2961 run_local_timers();
2962 if (rcu_pending(cpu))
2963 - rcu_check_callbacks(cpu, user_mode_vm(regs));
2964 + rcu_check_callbacks(cpu, user_mode_vm(get_irq_regs()));
2965 scheduler_tick();
2966 run_posix_cpu_timers(current);
2967 - profile_tick(CPU_PROFILING, regs);
2968 + profile_tick(CPU_PROFILING);
2969
2970 return IRQ_HANDLED;
2971 }
2972 @@ -967,10 +968,11 @@ extern void (*late_time_init)(void);
2973 /* Duplicate of time_init() below, with hpet_enable part added */
2974 static void __init hpet_time_init(void)
2975 {
2976 - xtime.tv_sec = get_cmos_time();
2977 - xtime.tv_nsec = (INITIAL_JIFFIES % HZ) * (NSEC_PER_SEC / HZ);
2978 - set_normalized_timespec(&wall_to_monotonic,
2979 - -xtime.tv_sec, -xtime.tv_nsec);
2980 + struct timespec ts;
2981 + ts.tv_sec = get_cmos_time();
2982 + ts.tv_nsec = (INITIAL_JIFFIES % HZ) * (NSEC_PER_SEC / HZ);
2983 +
2984 + do_settimeofday(&ts);
2985
2986 if ((hpet_enable() >= 0) && hpet_use_timer) {
2987 printk("Using HPET for base-timer\n");
2988 --- sle11-2009-05-14.orig/arch/x86/kernel/traps_32-xen.c 2009-05-14 11:02:43.000000000 +0200
2989 +++ sle11-2009-05-14/arch/x86/kernel/traps_32-xen.c 2009-03-04 11:28:34.000000000 +0100
2990 @@ -28,6 +28,7 @@
2991 #include <linux/kprobes.h>
2992 #include <linux/kexec.h>
2993 #include <linux/unwind.h>
2994 +#include <linux/uaccess.h>
2995
2996 #ifdef CONFIG_EISA
2997 #include <linux/ioport.h>
2998 @@ -40,7 +41,6 @@
2999
3000 #include <asm/processor.h>
3001 #include <asm/system.h>
3002 -#include <asm/uaccess.h>
3003 #include <asm/io.h>
3004 #include <asm/atomic.h>
3005 #include <asm/debugreg.h>
3006 @@ -51,11 +51,14 @@
3007 #include <asm/smp.h>
3008 #include <asm/arch_hooks.h>
3009 #include <asm/kdebug.h>
3010 +#include <asm/stacktrace.h>
3011
3012 #include <linux/module.h>
3013
3014 #include "mach_traps.h"
3015
3016 +int panic_on_unrecovered_nmi;
3017 +
3018 asmlinkage int system_call(void);
3019
3020 struct desc_struct default_ldt[] = { { 0, 0 }, { 0, 0 }, { 0, 0 },
3021 @@ -124,62 +127,63 @@ static inline int valid_stack_ptr(struct
3022 p < (void *)tinfo + THREAD_SIZE - 3;
3023 }
3024
3025 -/*
3026 - * Print one address/symbol entries per line.
3027 - */
3028 -static inline void print_addr_and_symbol(unsigned long addr, char *log_lvl)
3029 -{
3030 - printk(" [<%08lx>] ", addr);
3031 -
3032 - print_symbol("%s\n", addr);
3033 -}
3034 -
3035 static inline unsigned long print_context_stack(struct thread_info *tinfo,
3036 unsigned long *stack, unsigned long ebp,
3037 - char *log_lvl)
3038 + struct stacktrace_ops *ops, void *data)
3039 {
3040 unsigned long addr;
3041
3042 #ifdef CONFIG_FRAME_POINTER
3043 while (valid_stack_ptr(tinfo, (void *)ebp)) {
3044 + unsigned long new_ebp;
3045 addr = *(unsigned long *)(ebp + 4);
3046 - print_addr_and_symbol(addr, log_lvl);
3047 + ops->address(data, addr);
3048 /*
3049 * break out of recursive entries (such as
3050 - * end_of_stack_stop_unwind_function):
3051 + * end_of_stack_stop_unwind_function). Also,
3052 + * we can never allow a frame pointer to
3053 + * move downwards!
3054 */
3055 - if (ebp == *(unsigned long *)ebp)
3056 + new_ebp = *(unsigned long *)ebp;
3057 + if (new_ebp <= ebp)
3058 break;
3059 - ebp = *(unsigned long *)ebp;
3060 + ebp = new_ebp;
3061 }
3062 #else
3063 while (valid_stack_ptr(tinfo, stack)) {
3064 addr = *stack++;
3065 if (__kernel_text_address(addr))
3066 - print_addr_and_symbol(addr, log_lvl);
3067 + ops->address(data, addr);
3068 }
3069 #endif
3070 return ebp;
3071 }
3072
3073 +struct ops_and_data {
3074 + struct stacktrace_ops *ops;
3075 + void *data;
3076 +};
3077 +
3078 static asmlinkage int
3079 -show_trace_unwind(struct unwind_frame_info *info, void *log_lvl)
3080 +dump_trace_unwind(struct unwind_frame_info *info, void *data)
3081 {
3082 + struct ops_and_data *oad = (struct ops_and_data *)data;
3083 int n = 0;
3084
3085 while (unwind(info) == 0 && UNW_PC(info)) {
3086 n++;
3087 - print_addr_and_symbol(UNW_PC(info), log_lvl);
3088 + oad->ops->address(oad->data, UNW_PC(info));
3089 if (arch_unw_user_mode(info))
3090 break;
3091 }
3092 return n;
3093 }
3094
3095 -static void show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
3096 - unsigned long *stack, char *log_lvl)
3097 +void dump_trace(struct task_struct *task, struct pt_regs *regs,
3098 + unsigned long *stack,
3099 + struct stacktrace_ops *ops, void *data)
3100 {
3101 - unsigned long ebp;
3102 + unsigned long ebp = 0;
3103
3104 if (!task)
3105 task = current;
3106 @@ -187,54 +191,116 @@ static void show_trace_log_lvl(struct ta
3107 if (call_trace >= 0) {
3108 int unw_ret = 0;
3109 struct unwind_frame_info info;
3110 + struct ops_and_data oad = { .ops = ops, .data = data };
3111
3112 if (regs) {
3113 if (unwind_init_frame_info(&info, task, regs) == 0)
3114 - unw_ret = show_trace_unwind(&info, log_lvl);
3115 + unw_ret = dump_trace_unwind(&info, &oad);
3116 } else if (task == current)
3117 - unw_ret = unwind_init_running(&info, show_trace_unwind, log_lvl);
3118 + unw_ret = unwind_init_running(&info, dump_trace_unwind, &oad);
3119 else {
3120 if (unwind_init_blocked(&info, task) == 0)
3121 - unw_ret = show_trace_unwind(&info, log_lvl);
3122 + unw_ret = dump_trace_unwind(&info, &oad);
3123 }
3124 if (unw_ret > 0) {
3125 if (call_trace == 1 && !arch_unw_user_mode(&info)) {
3126 - print_symbol("DWARF2 unwinder stuck at %s\n",
3127 + ops->warning_symbol(data, "DWARF2 unwinder stuck at %s\n",
3128 UNW_PC(&info));
3129 if (UNW_SP(&info) >= PAGE_OFFSET) {
3130 - printk("Leftover inexact backtrace:\n");
3131 + ops->warning(data, "Leftover inexact backtrace:\n");
3132 stack = (void *)UNW_SP(&info);
3133 + if (!stack)
3134 + return;
3135 + ebp = UNW_FP(&info);
3136 } else
3137 - printk("Full inexact backtrace again:\n");
3138 + ops->warning(data, "Full inexact backtrace again:\n");
3139 } else if (call_trace >= 1)
3140 return;
3141 else
3142 - printk("Full inexact backtrace again:\n");
3143 + ops->warning(data, "Full inexact backtrace again:\n");
3144 } else
3145 - printk("Inexact backtrace:\n");
3146 + ops->warning(data, "Inexact backtrace:\n");
3147 }
3148 -
3149 - if (task == current) {
3150 - /* Grab ebp right from our regs */
3151 - asm ("movl %%ebp, %0" : "=r" (ebp) : );
3152 - } else {
3153 - /* ebp is the last reg pushed by switch_to */
3154 - ebp = *(unsigned long *) task->thread.esp;
3155 + if (!stack) {
3156 + unsigned long dummy;
3157 + stack = &dummy;
3158 + if (task && task != current)
3159 + stack = (unsigned long *)task->thread.esp;
3160 + }
3161 +
3162 +#ifdef CONFIG_FRAME_POINTER
3163 + if (!ebp) {
3164 + if (task == current) {
3165 + /* Grab ebp right from our regs */
3166 + asm ("movl %%ebp, %0" : "=r" (ebp) : );
3167 + } else {
3168 + /* ebp is the last reg pushed by switch_to */
3169 + ebp = *(unsigned long *) task->thread.esp;
3170 + }
3171 }
3172 +#endif
3173
3174 while (1) {
3175 struct thread_info *context;
3176 context = (struct thread_info *)
3177 ((unsigned long)stack & (~(THREAD_SIZE - 1)));
3178 - ebp = print_context_stack(context, stack, ebp, log_lvl);
3179 + ebp = print_context_stack(context, stack, ebp, ops, data);
3180 + /* Should be after the line below, but somewhere
3181 + in early boot context comes out corrupted and we
3182 + can't reference it -AK */
3183 + if (ops->stack(data, "IRQ") < 0)
3184 + break;
3185 stack = (unsigned long*)context->previous_esp;
3186 if (!stack)
3187 break;
3188 - printk("%s =======================\n", log_lvl);
3189 }
3190 }
3191 +EXPORT_SYMBOL(dump_trace);
3192 +
3193 +static void
3194 +print_trace_warning_symbol(void *data, char *msg, unsigned long symbol)
3195 +{
3196 + printk(data);
3197 + print_symbol(msg, symbol);
3198 + printk("\n");
3199 +}
3200 +
3201 +static void print_trace_warning(void *data, char *msg)
3202 +{
3203 + printk("%s%s\n", (char *)data, msg);
3204 +}
3205
3206 -void show_trace(struct task_struct *task, struct pt_regs *regs, unsigned long * stack)
3207 +static int print_trace_stack(void *data, char *name)
3208 +{
3209 + return 0;
3210 +}
3211 +
3212 +/*
3213 + * Print one address/symbol entries per line.
3214 + */
3215 +static void print_trace_address(void *data, unsigned long addr)
3216 +{
3217 + printk("%s [<%08lx>] ", (char *)data, addr);
3218 + print_symbol("%s\n", addr);
3219 +}
3220 +
3221 +static struct stacktrace_ops print_trace_ops = {
3222 + .warning = print_trace_warning,
3223 + .warning_symbol = print_trace_warning_symbol,
3224 + .stack = print_trace_stack,
3225 + .address = print_trace_address,
3226 +};
3227 +
3228 +static void
3229 +show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
3230 + unsigned long * stack, char *log_lvl)
3231 +{
3232 + dump_trace(task, regs, stack, &print_trace_ops, log_lvl);
3233 + printk("%s =======================\n", log_lvl);
3234 +}
3235 +
3236 +void show_trace(struct task_struct *task, struct pt_regs *regs,
3237 + unsigned long * stack)
3238 {
3239 show_trace_log_lvl(task, regs, stack, "");
3240 }
3241 @@ -297,12 +363,13 @@ void show_registers(struct pt_regs *regs
3242 ss = regs->xss & 0xffff;
3243 }
3244 print_modules();
3245 - printk(KERN_EMERG "CPU: %d\nEIP: %04x:[<%08lx>] %s VLI\n"
3246 - "EFLAGS: %08lx (%s %.*s) \n",
3247 + printk(KERN_EMERG "CPU: %d\n"
3248 + KERN_EMERG "EIP: %04x:[<%08lx>] %s VLI\n"
3249 + KERN_EMERG "EFLAGS: %08lx (%s %.*s)\n",
3250 smp_processor_id(), 0xffff & regs->xcs, regs->eip,
3251 - print_tainted(), regs->eflags, system_utsname.release,
3252 - (int)strcspn(system_utsname.version, " "),
3253 - system_utsname.version);
3254 + print_tainted(), regs->eflags, init_utsname()->release,
3255 + (int)strcspn(init_utsname()->version, " "),
3256 + init_utsname()->version);
3257 print_symbol(KERN_EMERG "EIP is at %s\n", regs->eip);
3258 printk(KERN_EMERG "eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n",
3259 regs->eax, regs->ebx, regs->ecx, regs->edx);
3260 @@ -319,6 +386,8 @@ void show_registers(struct pt_regs *regs
3261 */
3262 if (in_kernel) {
3263 u8 __user *eip;
3264 + int code_bytes = 64;
3265 + unsigned char c;
3266
3267 printk("\n" KERN_EMERG "Stack: ");
3268 show_stack_log_lvl(NULL, regs, (unsigned long *)esp, KERN_EMERG);
3269 @@ -326,9 +395,12 @@ void show_registers(struct pt_regs *regs
3270 printk(KERN_EMERG "Code: ");
3271
3272 eip = (u8 __user *)regs->eip - 43;
3273 - for (i = 0; i < 64; i++, eip++) {
3274 - unsigned char c;
3275 -
3276 + if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) {
3277 + /* try starting at EIP */
3278 + eip = (u8 __user *)regs->eip;
3279 + code_bytes = 32;
3280 + }
3281 + for (i = 0; i < code_bytes; i++, eip++) {
3282 if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) {
3283 printk(" Bad EIP value.");
3284 break;
3285 @@ -349,7 +421,7 @@ static void handle_BUG(struct pt_regs *r
3286
3287 if (eip < PAGE_OFFSET)
3288 return;
3289 - if (__get_user(ud2, (unsigned short __user *)eip))
3290 + if (probe_kernel_address((unsigned short __user *)eip, ud2))
3291 return;
3292 if (ud2 != 0x0b0f)
3293 return;
3294 @@ -362,7 +434,8 @@ static void handle_BUG(struct pt_regs *r
3295 char *file;
3296 char c;
3297
3298 - if (__get_user(line, (unsigned short __user *)(eip + 2)))
3299 + if (probe_kernel_address((unsigned short __user *)(eip + 2),
3300 + line))
3301 break;
3302 if (__get_user(file, (char * __user *)(eip + 4)) ||
3303 (unsigned long)file < PAGE_OFFSET || __get_user(c, file))
3304 @@ -604,18 +677,24 @@ gp_in_kernel:
3305 }
3306 }
3307
3308 -static void mem_parity_error(unsigned char reason, struct pt_regs * regs)
3309 +static __kprobes void
3310 +mem_parity_error(unsigned char reason, struct pt_regs * regs)
3311 {
3312 - printk(KERN_EMERG "Uhhuh. NMI received. Dazed and confused, but trying "
3313 - "to continue\n");
3314 + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on "
3315 + "CPU %d.\n", reason, smp_processor_id());
3316 printk(KERN_EMERG "You probably have a hardware problem with your RAM "
3317 "chips\n");
3318 + if (panic_on_unrecovered_nmi)
3319 + panic("NMI: Not continuing");
3320 +
3321 + printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
3322
3323 /* Clear and disable the memory parity error line. */
3324 clear_mem_error(reason);
3325 }
3326
3327 -static void io_check_error(unsigned char reason, struct pt_regs * regs)
3328 +static __kprobes void
3329 +io_check_error(unsigned char reason, struct pt_regs * regs)
3330 {
3331 printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n");
3332 show_registers(regs);
3333 @@ -624,7 +703,8 @@ static void io_check_error(unsigned char
3334 clear_io_check_error(reason);
3335 }
3336
3337 -static void unknown_nmi_error(unsigned char reason, struct pt_regs * regs)
3338 +static __kprobes void
3339 +unknown_nmi_error(unsigned char reason, struct pt_regs * regs)
3340 {
3341 #ifdef CONFIG_MCA
3342 /* Might actually be able to figure out what the guilty party
3343 @@ -634,15 +714,18 @@ static void unknown_nmi_error(unsigned c
3344 return;
3345 }
3346 #endif
3347 - printk("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
3348 - reason, smp_processor_id());
3349 - printk("Dazed and confused, but trying to continue\n");
3350 - printk("Do you have a strange power saving mode enabled?\n");
3351 + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on "
3352 + "CPU %d.\n", reason, smp_processor_id());
3353 + printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n");
3354 + if (panic_on_unrecovered_nmi)
3355 + panic("NMI: Not continuing");
3356 +
3357 + printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
3358 }
3359
3360 static DEFINE_SPINLOCK(nmi_print_lock);
3361
3362 -void die_nmi (struct pt_regs *regs, const char *msg)
3363 +void __kprobes die_nmi(struct pt_regs *regs, const char *msg)
3364 {
3365 if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 2, SIGINT) ==
3366 NOTIFY_STOP)
3367 @@ -674,7 +757,7 @@ void die_nmi (struct pt_regs *regs, cons
3368 do_exit(SIGSEGV);
3369 }
3370
3371 -static void default_do_nmi(struct pt_regs * regs)
3372 +static __kprobes void default_do_nmi(struct pt_regs * regs)
3373 {
3374 unsigned char reason = 0;
3375
3376 @@ -691,12 +774,12 @@ static void default_do_nmi(struct pt_reg
3377 * Ok, so this is none of the documented NMI sources,
3378 * so it must be the NMI watchdog.
3379 */
3380 - if (nmi_watchdog) {
3381 - nmi_watchdog_tick(regs);
3382 + if (nmi_watchdog_tick(regs, reason))
3383 return;
3384 - }
3385 + if (!do_nmi_callback(regs, smp_processor_id()))
3386 #endif
3387 - unknown_nmi_error(reason, regs);
3388 + unknown_nmi_error(reason, regs);
3389 +
3390 return;
3391 }
3392 if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
3393 @@ -712,14 +795,7 @@ static void default_do_nmi(struct pt_reg
3394 reassert_nmi();
3395 }
3396
3397 -static int dummy_nmi_callback(struct pt_regs * regs, int cpu)
3398 -{
3399 - return 0;
3400 -}
3401 -
3402 -static nmi_callback_t nmi_callback = dummy_nmi_callback;
3403 -
3404 -fastcall void do_nmi(struct pt_regs * regs, long error_code)
3405 +fastcall __kprobes void do_nmi(struct pt_regs * regs, long error_code)
3406 {
3407 int cpu;
3408
3409 @@ -729,25 +805,11 @@ fastcall void do_nmi(struct pt_regs * re
3410
3411 ++nmi_count(cpu);
3412
3413 - if (!rcu_dereference(nmi_callback)(regs, cpu))
3414 - default_do_nmi(regs);
3415 + default_do_nmi(regs);
3416
3417 nmi_exit();
3418 }
3419
3420 -void set_nmi_callback(nmi_callback_t callback)
3421 -{
3422 - vmalloc_sync_all();
3423 - rcu_assign_pointer(nmi_callback, callback);
3424 -}
3425 -EXPORT_SYMBOL_GPL(set_nmi_callback);
3426 -
3427 -void unset_nmi_callback(void)
3428 -{
3429 - nmi_callback = dummy_nmi_callback;
3430 -}
3431 -EXPORT_SYMBOL_GPL(unset_nmi_callback);
3432 -
3433 #ifdef CONFIG_KPROBES
3434 fastcall void __kprobes do_int3(struct pt_regs *regs, long error_code)
3435 {
3436 --- sle11-2009-05-14.orig/arch/x86/mach-xen/setup.c 2009-05-14 11:02:43.000000000 +0200
3437 +++ sle11-2009-05-14/arch/x86/mach-xen/setup.c 2009-03-04 11:28:34.000000000 +0100
3438 @@ -103,8 +103,10 @@ void __init pre_setup_arch_hook(void)
3439
3440 setup_xen_features();
3441
3442 - if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0)
3443 - set_fixaddr_top(pp.virt_start);
3444 + if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0) {
3445 + hypervisor_virt_start = pp.virt_start;
3446 + reserve_top_address(0UL - pp.virt_start);
3447 + }
3448
3449 if (HYPERVISOR_memory_op(XENMEM_machphys_mapping, &mapping) == 0) {
3450 machine_to_phys_mapping = (unsigned long *)mapping.v_start;
3451 --- sle11-2009-05-14.orig/arch/x86/mm/fault_32-xen.c 2009-05-14 11:02:43.000000000 +0200
3452 +++ sle11-2009-05-14/arch/x86/mm/fault_32-xen.c 2009-03-04 11:28:34.000000000 +0100
3453 @@ -27,21 +27,24 @@
3454 #include <asm/uaccess.h>
3455 #include <asm/desc.h>
3456 #include <asm/kdebug.h>
3457 +#include <asm/segment.h>
3458
3459 extern void die(const char *,struct pt_regs *,long);
3460
3461 -#ifdef CONFIG_KPROBES
3462 -ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
3463 +static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
3464 +
3465 int register_page_fault_notifier(struct notifier_block *nb)
3466 {
3467 vmalloc_sync_all();
3468 return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
3469 }
3470 +EXPORT_SYMBOL_GPL(register_page_fault_notifier);
3471
3472 int unregister_page_fault_notifier(struct notifier_block *nb)
3473 {
3474 return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
3475 }
3476 +EXPORT_SYMBOL_GPL(unregister_page_fault_notifier);
3477
3478 static inline int notify_page_fault(enum die_val val, const char *str,
3479 struct pt_regs *regs, long err, int trap, int sig)
3480 @@ -55,14 +58,6 @@ static inline int notify_page_fault(enum
3481 };
3482 return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
3483 }
3484 -#else
3485 -static inline int notify_page_fault(enum die_val val, const char *str,
3486 - struct pt_regs *regs, long err, int trap, int sig)
3487 -{
3488 - return NOTIFY_DONE;
3489 -}
3490 -#endif
3491 -
3492
3493 /*
3494 * Unlock any spinlocks which will prevent us from getting the
3495 @@ -119,10 +114,10 @@ static inline unsigned long get_segment_
3496 }
3497
3498 /* The standard kernel/user address space limit. */
3499 - *eip_limit = (seg & 2) ? USER_DS.seg : KERNEL_DS.seg;
3500 + *eip_limit = user_mode(regs) ? USER_DS.seg : KERNEL_DS.seg;
3501
3502 /* By far the most common cases. */
3503 - if (likely(seg == __USER_CS || seg == GET_KERNEL_CS()))
3504 + if (likely(SEGMENT_IS_FLAT_CODE(seg)))
3505 return eip;
3506
3507 /* Check the segment exists, is within the current LDT/GDT size,
3508 @@ -559,11 +554,7 @@ good_area:
3509 write = 0;
3510 switch (error_code & 3) {
3511 default: /* 3: write, present */
3512 -#ifdef TEST_VERIFY_AREA
3513 - if (regs->cs == GET_KERNEL_CS())
3514 - printk("WP fault at %08lx\n", regs->eip);
3515 -#endif
3516 - /* fall through */
3517 + /* fall through */
3518 case 2: /* write, not present */
3519 if (!(vma->vm_flags & VM_WRITE))
3520 goto bad_area;
3521 @@ -572,7 +563,7 @@ good_area:
3522 case 1: /* read, present */
3523 goto bad_area;
3524 case 0: /* read, not present */
3525 - if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
3526 + if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
3527 goto bad_area;
3528 }
3529
3530 @@ -704,7 +695,7 @@ no_context:
3531 */
3532 out_of_memory:
3533 up_read(&mm->mmap_sem);
3534 - if (tsk->pid == 1) {
3535 + if (is_init(tsk)) {
3536 yield();
3537 down_read(&mm->mmap_sem);
3538 goto survive;
3539 --- sle11-2009-05-14.orig/arch/x86/mm/highmem_32-xen.c 2009-05-14 11:02:43.000000000 +0200
3540 +++ sle11-2009-05-14/arch/x86/mm/highmem_32-xen.c 2009-03-04 11:28:34.000000000 +0100
3541 @@ -38,11 +38,9 @@ static void *__kmap_atomic(struct page *
3542
3543 idx = type + KM_TYPE_NR*smp_processor_id();
3544 vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
3545 -#ifdef CONFIG_DEBUG_HIGHMEM
3546 if (!pte_none(*(kmap_pte-idx)))
3547 BUG();
3548 -#endif
3549 - set_pte_at_sync(&init_mm, vaddr, kmap_pte-idx, mk_pte(page, prot));
3550 + set_pte_at(&init_mm, vaddr, kmap_pte-idx, mk_pte(page, prot));
3551
3552 return (void*) vaddr;
3553 }
3554 @@ -62,36 +60,26 @@ void *kmap_atomic_pte(struct page *page,
3555
3556 void kunmap_atomic(void *kvaddr, enum km_type type)
3557 {
3558 -#if defined(CONFIG_DEBUG_HIGHMEM) || defined(CONFIG_XEN)
3559 unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK;
3560 enum fixed_addresses idx = type + KM_TYPE_NR*smp_processor_id();
3561
3562 - if (vaddr < FIXADDR_START) { // FIXME
3563 +#ifdef CONFIG_DEBUG_HIGHMEM
3564 + if (vaddr >= PAGE_OFFSET && vaddr < (unsigned long)high_memory) {
3565 dec_preempt_count();
3566 preempt_check_resched();
3567 return;
3568 }
3569 -#endif
3570
3571 -#if defined(CONFIG_DEBUG_HIGHMEM)
3572 if (vaddr != __fix_to_virt(FIX_KMAP_BEGIN+idx))
3573 BUG();
3574 -
3575 - /*
3576 - * force other mappings to Oops if they'll try to access
3577 - * this pte without first remap it
3578 - */
3579 - pte_clear(&init_mm, vaddr, kmap_pte-idx);
3580 - __flush_tlb_one(vaddr);
3581 -#elif defined(CONFIG_XEN)
3582 +#endif
3583 /*
3584 - * We must ensure there are no dangling pagetable references when
3585 - * returning memory to Xen (decrease_reservation).
3586 - * XXX TODO: We could make this faster by only zapping when
3587 - * kmap_flush_unused is called but that is trickier and more invasive.
3588 + * Force other mappings to Oops if they'll try to access this pte
3589 + * without first remap it. Keeping stale mappings around is a bad idea
3590 + * also, in case the page changes cacheability attributes or becomes
3591 + * a protected page in a hypervisor.
3592 */
3593 - pte_clear(&init_mm, vaddr, kmap_pte-idx);
3594 -#endif
3595 + kpte_clear_flush(kmap_pte-idx, vaddr);
3596
3597 dec_preempt_count();
3598 preempt_check_resched();
3599 @@ -110,7 +98,6 @@ void *kmap_atomic_pfn(unsigned long pfn,
3600 idx = type + KM_TYPE_NR*smp_processor_id();
3601 vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
3602 set_pte(kmap_pte-idx, pfn_pte(pfn, kmap_prot));
3603 - __flush_tlb_one(vaddr);
3604
3605 return (void*) vaddr;
3606 }
3607 --- sle11-2009-05-14.orig/arch/x86/mm/hypervisor.c 2008-12-15 11:13:45.000000000 +0100
3608 +++ sle11-2009-05-14/arch/x86/mm/hypervisor.c 2009-03-04 11:28:34.000000000 +0100
3609 @@ -31,6 +31,7 @@
3610 */
3611
3612 #include <linux/sched.h>
3613 +#include <linux/hardirq.h>
3614 #include <linux/mm.h>
3615 #include <linux/vmalloc.h>
3616 #include <asm/page.h>
3617 @@ -44,6 +45,302 @@
3618 #include <asm/tlbflush.h>
3619 #include <linux/highmem.h>
3620
3621 +EXPORT_SYMBOL(hypercall_page);
3622 +
3623 +#define NR_MC BITS_PER_LONG
3624 +#define NR_MMU BITS_PER_LONG
3625 +#define NR_MMUEXT (BITS_PER_LONG / 4)
3626 +
3627 +DEFINE_PER_CPU(bool, xen_lazy_mmu);
3628 +EXPORT_PER_CPU_SYMBOL(xen_lazy_mmu);
3629 +struct lazy_mmu {
3630 + unsigned int nr_mc, nr_mmu, nr_mmuext;
3631 + multicall_entry_t mc[NR_MC];
3632 + mmu_update_t mmu[NR_MMU];
3633 + struct mmuext_op mmuext[NR_MMUEXT];
3634 +};
3635 +static DEFINE_PER_CPU(struct lazy_mmu, lazy_mmu);
3636 +
3637 +static inline bool use_lazy_mmu_mode(void)
3638 +{
3639 +#ifdef CONFIG_PREEMPT
3640 + if (!preempt_count())
3641 + return false;
3642 +#endif
3643 + return !irq_count();
3644 +}
3645 +
3646 +static void multicall_failed(const multicall_entry_t *mc, int rc)
3647 +{
3648 + printk(KERN_EMERG "hypercall#%lu(%lx, %lx, %lx, %lx)"
3649 + " failed: %d (caller %lx)\n",
3650 + mc->op, mc->args[0], mc->args[1], mc->args[2], mc->args[3],
3651 + rc, mc->args[5]);
3652 + BUG();
3653 +}
3654 +
3655 +int xen_multicall_flush(bool ret_last) {
3656 + struct lazy_mmu *lazy = &__get_cpu_var(lazy_mmu);
3657 + multicall_entry_t *mc = lazy->mc;
3658 + unsigned int count = lazy->nr_mc;
3659 +
3660 + if (!count || !use_lazy_mmu_mode())
3661 + return 0;
3662 +
3663 + lazy->nr_mc = 0;
3664 + lazy->nr_mmu = 0;
3665 + lazy->nr_mmuext = 0;
3666 +
3667 + if (count == 1) {
3668 + int rc = _hypercall(int, mc->op, mc->args[0], mc->args[1],
3669 + mc->args[2], mc->args[3], mc->args[4]);
3670 +
3671 + if (unlikely(rc)) {
3672 + if (ret_last)
3673 + return rc;
3674 + multicall_failed(mc, rc);
3675 + }
3676 + } else {
3677 + if (HYPERVISOR_multicall(mc, count))
3678 + BUG();
3679 + while (count-- > ret_last)
3680 + if (unlikely(mc++->result))
3681 + multicall_failed(mc - 1, mc[-1].result);
3682 + if (ret_last)
3683 + return mc->result;
3684 + }
3685 +
3686 + return 0;
3687 +}
3688 +EXPORT_SYMBOL(xen_multicall_flush);
3689 +
3690 +int xen_multi_update_va_mapping(unsigned long va, pte_t pte,
3691 + unsigned long uvmf)
3692 +{
3693 + struct lazy_mmu *lazy = &__get_cpu_var(lazy_mmu);
3694 + multicall_entry_t *mc;
3695 +
3696 + if (unlikely(!use_lazy_mmu_mode()))
3697 +#ifdef CONFIG_X86_PAE
3698 + return _hypercall4(int, update_va_mapping, va,
3699 + pte.pte_low, pte.pte_high, uvmf);
3700 +#else
3701 + return _hypercall3(int, update_va_mapping, va,
3702 + pte.pte, uvmf);
3703 +#endif
3704 +
3705 + if (unlikely(lazy->nr_mc == NR_MC))
3706 + xen_multicall_flush(false);
3707 +
3708 + mc = lazy->mc + lazy->nr_mc++;
3709 + mc->op = __HYPERVISOR_update_va_mapping;
3710 + mc->args[0] = va;
3711 +#ifndef CONFIG_X86_PAE
3712 + mc->args[1] = pte.pte;
3713 +#else
3714 + mc->args[1] = pte.pte_low;
3715 + mc->args[2] = pte.pte_high;
3716 +#endif
3717 + mc->args[MULTI_UVMFLAGS_INDEX] = uvmf;
3718 + mc->args[5] = (long)__builtin_return_address(0);
3719 +
3720 + return 0;
3721 +}
3722 +
3723 +static inline bool mmu_may_merge(const multicall_entry_t *mc,
3724 + unsigned int op, domid_t domid)
3725 +{
3726 + return mc->op == op && !mc->args[2] && mc->args[3] == domid;
3727 +}
3728 +
3729 +int xen_multi_mmu_update(mmu_update_t *src, unsigned int count,
3730 + unsigned int *success_count, domid_t domid)
3731 +{
3732 + struct lazy_mmu *lazy = &__get_cpu_var(lazy_mmu);
3733 + multicall_entry_t *mc = lazy->mc + lazy->nr_mc;
3734 + mmu_update_t *dst;
3735 + bool commit, merge;
3736 +
3737 + if (unlikely(!use_lazy_mmu_mode()))
3738 + return _hypercall4(int, mmu_update, src, count,
3739 + success_count, domid);
3740 +
3741 + commit = (lazy->nr_mmu + count) > NR_MMU || success_count;
3742 + merge = lazy->nr_mc && !commit
3743 + && mmu_may_merge(mc - 1, __HYPERVISOR_mmu_update, domid);
3744 + if (unlikely(lazy->nr_mc == NR_MC) && !merge) {
3745 + xen_multicall_flush(false);
3746 + mc = lazy->mc;
3747 + commit = count > NR_MMU || success_count;
3748 + }
3749 +
3750 + if (!lazy->nr_mc && unlikely(commit))
3751 + return _hypercall4(int, mmu_update, src, count,
3752 + success_count, domid);
3753 +
3754 + dst = lazy->mmu + lazy->nr_mmu;
3755 + lazy->nr_mmu += count;
3756 + if (merge) {
3757 + mc[-1].args[1] += count;
3758 + memcpy(dst, src, count * sizeof(*src));
3759 + } else {
3760 + ++lazy->nr_mc;
3761 + mc->op = __HYPERVISOR_mmu_update;
3762 + if (!commit) {
3763 + mc->args[0] = (unsigned long)dst;
3764 + memcpy(dst, src, count * sizeof(*src));
3765 + } else
3766 + mc->args[0] = (unsigned long)src;
3767 + mc->args[1] = count;
3768 + mc->args[2] = (unsigned long)success_count;
3769 + mc->args[3] = domid;
3770 + mc->args[5] = (long)__builtin_return_address(0);
3771 + }
3772 +
3773 + while (!commit && count--)
3774 + switch (src++->ptr & (sizeof(pteval_t) - 1)) {
3775 + case MMU_NORMAL_PT_UPDATE:
3776 + case MMU_PT_UPDATE_PRESERVE_AD:
3777 + break;
3778 + default:
3779 + commit = true;
3780 + break;
3781 + }
3782 +
3783 + return commit ? xen_multicall_flush(true) : 0;
3784 +}
3785 +
3786 +int xen_multi_mmuext_op(struct mmuext_op *src, unsigned int count,
3787 + unsigned int *success_count, domid_t domid)
3788 +{
3789 + struct lazy_mmu *lazy = &__get_cpu_var(lazy_mmu);
3790 + multicall_entry_t *mc;
3791 + struct mmuext_op *dst;
3792 + bool commit, merge;
3793 +
3794 + if (unlikely(!use_lazy_mmu_mode()))
3795 + return _hypercall4(int, mmuext_op, src, count,
3796 + success_count, domid);
3797 +
3798 + /*
3799 + * While it could be useful in theory, I've never seen the body of
3800 + * this conditional to be reached, hence it seems more reasonable
3801 + * to disable it for the time being.
3802 + */
3803 + if (0 && likely(count)
3804 + && likely(!success_count)
3805 + && likely(domid == DOMID_SELF)
3806 + && likely(lazy->nr_mc)
3807 + && lazy->mc[lazy->nr_mc - 1].op == __HYPERVISOR_update_va_mapping) {
3808 + unsigned long oldf, newf = UVMF_NONE;
3809 +
3810 + switch (src->cmd) {
3811 + case MMUEXT_TLB_FLUSH_ALL:
3812 + newf = UVMF_TLB_FLUSH | UVMF_ALL;
3813 + break;
3814 + case MMUEXT_INVLPG_ALL:
3815 + newf = UVMF_INVLPG | UVMF_ALL;
3816 + break;
3817 + case MMUEXT_TLB_FLUSH_MULTI:
3818 + newf = UVMF_TLB_FLUSH | UVMF_MULTI
3819 + | (unsigned long)src->arg2.vcpumask.p;
3820 + break;
3821 + case MMUEXT_INVLPG_MULTI:
3822 + newf = UVMF_INVLPG | UVMF_MULTI
3823 + | (unsigned long)src->arg2.vcpumask.p;
3824 + break;
3825 + case MMUEXT_TLB_FLUSH_LOCAL:
3826 + newf = UVMF_TLB_FLUSH | UVMF_LOCAL;
3827 + break;
3828 + case MMUEXT_INVLPG_LOCAL:
3829 + newf = UVMF_INVLPG | UVMF_LOCAL;
3830 + break;
3831 + }
3832 + mc = lazy->mc + lazy->nr_mc - 1;
3833 + oldf = mc->args[MULTI_UVMFLAGS_INDEX];
3834 + if (newf == UVMF_NONE || oldf == UVMF_NONE
3835 + || newf == (UVMF_TLB_FLUSH | UVMF_ALL))
3836 + ;
3837 + else if (oldf == (UVMF_TLB_FLUSH | UVMF_ALL))
3838 + newf = UVMF_TLB_FLUSH | UVMF_ALL;
3839 + else if ((newf & UVMF_FLUSHTYPE_MASK) == UVMF_INVLPG
3840 + && (oldf & UVMF_FLUSHTYPE_MASK) == UVMF_INVLPG
3841 + && ((src->arg1.linear_addr ^ mc->args[0])
3842 + >> PAGE_SHIFT))
3843 + newf = UVMF_NONE;
3844 + else if (((oldf | newf) & UVMF_ALL)
3845 + && !((oldf ^ newf) & UVMF_FLUSHTYPE_MASK))
3846 + newf |= UVMF_ALL;
3847 + else if ((oldf ^ newf) & ~UVMF_FLUSHTYPE_MASK)
3848 + newf = UVMF_NONE;
3849 + else if ((oldf & UVMF_FLUSHTYPE_MASK) == UVMF_TLB_FLUSH)
3850 + newf = (newf & ~UVMF_FLUSHTYPE_MASK) | UVMF_TLB_FLUSH;
3851 + else if ((newf & UVMF_FLUSHTYPE_MASK) != UVMF_TLB_FLUSH
3852 + && ((newf ^ oldf) & UVMF_FLUSHTYPE_MASK))
3853 + newf = UVMF_NONE;
3854 + if (newf != UVMF_NONE) {
3855 + mc->args[MULTI_UVMFLAGS_INDEX] = newf;
3856 + ++src;
3857 + if (!--count)
3858 + return 0;
3859 + }
3860 + }
3861 +
3862 + mc = lazy->mc + lazy->nr_mc;
3863 + commit = (lazy->nr_mmuext + count) > NR_MMUEXT || success_count;
3864 + merge = lazy->nr_mc && !commit
3865 + && mmu_may_merge(mc - 1, __HYPERVISOR_mmuext_op, domid);
3866 + if (unlikely(lazy->nr_mc == NR_MC) && !merge) {
3867 + xen_multicall_flush(false);
3868 + mc = lazy->mc;
3869 + commit = count > NR_MMUEXT || success_count;
3870 + }
3871 +
3872 + if (!lazy->nr_mc && unlikely(commit))
3873 + return _hypercall4(int, mmuext_op, src, count,
3874 + success_count, domid);
3875 +
3876 + dst = lazy->mmuext + lazy->nr_mmuext;
3877 + lazy->nr_mmuext += count;
3878 + if (merge) {
3879 + mc[-1].args[1] += count;
3880 + memcpy(dst, src, count * sizeof(*src));
3881 + } else {
3882 + ++lazy->nr_mc;
3883 + mc->op = __HYPERVISOR_mmuext_op;
3884 + if (!commit) {
3885 + mc->args[0] = (unsigned long)dst;
3886 + memcpy(dst, src, count * sizeof(*src));
3887 + } else
3888 + mc->args[0] = (unsigned long)src;
3889 + mc->args[1] = count;
3890 + mc->args[2] = (unsigned long)success_count;
3891 + mc->args[3] = domid;
3892 + mc->args[5] = (long)__builtin_return_address(0);
3893 + }
3894 +
3895 + while (!commit && count--)
3896 + switch (src++->cmd) {
3897 + case MMUEXT_PIN_L1_TABLE:
3898 + case MMUEXT_PIN_L2_TABLE:
3899 + case MMUEXT_PIN_L3_TABLE:
3900 + case MMUEXT_PIN_L4_TABLE:
3901 + case MMUEXT_UNPIN_TABLE:
3902 + case MMUEXT_TLB_FLUSH_LOCAL:
3903 + case MMUEXT_INVLPG_LOCAL:
3904 + case MMUEXT_TLB_FLUSH_MULTI:
3905 + case MMUEXT_INVLPG_MULTI:
3906 + case MMUEXT_TLB_FLUSH_ALL:
3907 + case MMUEXT_INVLPG_ALL:
3908 + break;
3909 + default:
3910 + commit = true;
3911 + break;
3912 + }
3913 +
3914 + return commit ? xen_multicall_flush(true) : 0;
3915 +}
3916 +
3917 void xen_l1_entry_update(pte_t *ptr, pte_t val)
3918 {
3919 mmu_update_t u;
3920 @@ -542,7 +839,8 @@ int write_ldt_entry(void *ldt, int entry
3921 #define MAX_BATCHED_FULL_PTES 32
3922
3923 int xen_change_pte_range(struct mm_struct *mm, pmd_t *pmd,
3924 - unsigned long addr, unsigned long end, pgprot_t newprot)
3925 + unsigned long addr, unsigned long end, pgprot_t newprot,
3926 + int dirty_accountable)
3927 {
3928 int rc = 0, i = 0;
3929 mmu_update_t u[MAX_BATCHED_FULL_PTES];
3930 @@ -555,10 +853,14 @@ int xen_change_pte_range(struct mm_struc
3931 pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
3932 do {
3933 if (pte_present(*pte)) {
3934 + pte_t ptent = pte_modify(*pte, newprot);
3935 +
3936 + if (dirty_accountable && pte_dirty(ptent))
3937 + ptent = pte_mkwrite(ptent);
3938 u[i].ptr = (__pmd_val(*pmd) & PHYSICAL_PAGE_MASK)
3939 | ((unsigned long)pte & ~PAGE_MASK)
3940 | MMU_PT_UPDATE_PRESERVE_AD;
3941 - u[i].val = __pte_val(pte_modify(*pte, newprot));
3942 + u[i].val = __pte_val(ptent);
3943 if (++i == MAX_BATCHED_FULL_PTES) {
3944 if ((rc = HYPERVISOR_mmu_update(
3945 &u[0], i, NULL, DOMID_SELF)) != 0)
3946 --- sle11-2009-05-14.orig/arch/x86/mm/init_32-xen.c 2009-05-14 11:02:43.000000000 +0200
3947 +++ sle11-2009-05-14/arch/x86/mm/init_32-xen.c 2009-03-04 11:28:34.000000000 +0100
3948 @@ -462,16 +462,22 @@ EXPORT_SYMBOL(__supported_pte_mask);
3949 * on Enable
3950 * off Disable
3951 */
3952 -void __init noexec_setup(const char *str)
3953 +static int __init noexec_setup(char *str)
3954 {
3955 - if (!strncmp(str, "on",2) && cpu_has_nx) {
3956 - __supported_pte_mask |= _PAGE_NX;
3957 - disable_nx = 0;
3958 - } else if (!strncmp(str,"off",3)) {
3959 + if (!str || !strcmp(str, "on")) {
3960 + if (cpu_has_nx) {
3961 + __supported_pte_mask |= _PAGE_NX;
3962 + disable_nx = 0;
3963 + }
3964 + } else if (!strcmp(str,"off")) {
3965 disable_nx = 1;
3966 __supported_pte_mask &= ~_PAGE_NX;
3967 - }
3968 + } else
3969 + return -EINVAL;
3970 +
3971 + return 0;
3972 }
3973 +early_param("noexec", noexec_setup);
3974
3975 int nx_enabled = 0;
3976 #ifdef CONFIG_X86_PAE
3977 @@ -514,6 +520,7 @@ int __init set_kernel_exec(unsigned long
3978 pte->pte_high &= ~(1 << (_PAGE_BIT_NX - 32));
3979 else
3980 pte->pte_high |= 1 << (_PAGE_BIT_NX - 32);
3981 + pte_update_defer(&init_mm, vaddr, pte);
3982 __flush_tlb_all();
3983 out:
3984 return ret;
3985 @@ -596,18 +603,6 @@ static void __init test_wp_bit(void)
3986 }
3987 }
3988
3989 -static void __init set_max_mapnr_init(void)
3990 -{
3991 -#ifdef CONFIG_HIGHMEM
3992 - num_physpages = highend_pfn;
3993 -#else
3994 - num_physpages = max_low_pfn;
3995 -#endif
3996 -#ifdef CONFIG_FLATMEM
3997 - max_mapnr = num_physpages;
3998 -#endif
3999 -}
4000 -
4001 static struct kcore_list kcore_mem, kcore_vmalloc;
4002
4003 void __init mem_init(void)
4004 @@ -623,8 +618,7 @@ void __init mem_init(void)
4005 #endif
4006
4007 #ifdef CONFIG_FLATMEM
4008 - if (!mem_map)
4009 - BUG();
4010 + BUG_ON(!mem_map);
4011 #endif
4012
4013 bad_ppro = ppro_with_ram_bug();
4014 @@ -639,17 +633,6 @@ void __init mem_init(void)
4015 }
4016 #endif
4017
4018 - set_max_mapnr_init();
4019 -
4020 -#ifdef CONFIG_HIGHMEM
4021 - high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
4022 -#else
4023 - high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
4024 -#endif
4025 - printk("vmalloc area: %lx-%lx, maxmem %lx\n",
4026 - VMALLOC_START,VMALLOC_END,MAXMEM);
4027 - BUG_ON(VMALLOC_START > VMALLOC_END);
4028 -
4029 /* this will put all low memory onto the freelists */
4030 totalram_pages += free_all_bootmem();
4031 /* XEN: init and count low-mem pages outside initial allocation. */
4032 @@ -687,6 +670,48 @@ void __init mem_init(void)
4033 (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10))
4034 );
4035
4036 +#if 1 /* double-sanity-check paranoia */
4037 + printk("virtual kernel memory layout:\n"
4038 + " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n"
4039 +#ifdef CONFIG_HIGHMEM
4040 + " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n"
4041 +#endif
4042 + " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n"
4043 + " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n"
4044 + " .init : 0x%08lx - 0x%08lx (%4ld kB)\n"
4045 + " .data : 0x%08lx - 0x%08lx (%4ld kB)\n"
4046 + " .text : 0x%08lx - 0x%08lx (%4ld kB)\n",
4047 + FIXADDR_START, FIXADDR_TOP,
4048 + (FIXADDR_TOP - FIXADDR_START) >> 10,
4049 +
4050 +#ifdef CONFIG_HIGHMEM
4051 + PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE,
4052 + (LAST_PKMAP*PAGE_SIZE) >> 10,
4053 +#endif
4054 +
4055 + VMALLOC_START, VMALLOC_END,
4056 + (VMALLOC_END - VMALLOC_START) >> 20,
4057 +
4058 + (unsigned long)__va(0), (unsigned long)high_memory,
4059 + ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20,
4060 +
4061 + (unsigned long)&__init_begin, (unsigned long)&__init_end,
4062 + ((unsigned long)&__init_end - (unsigned long)&__init_begin) >> 10,
4063 +
4064 + (unsigned long)&_etext, (unsigned long)&_edata,
4065 + ((unsigned long)&_edata - (unsigned long)&_etext) >> 10,
4066 +
4067 + (unsigned long)&_text, (unsigned long)&_etext,
4068 + ((unsigned long)&_etext - (unsigned long)&_text) >> 10);
4069 +
4070 +#ifdef CONFIG_HIGHMEM
4071 + BUG_ON(PKMAP_BASE+LAST_PKMAP*PAGE_SIZE > FIXADDR_START);
4072 + BUG_ON(VMALLOC_END > PKMAP_BASE);
4073 +#endif
4074 + BUG_ON(VMALLOC_START > VMALLOC_END);
4075 + BUG_ON((unsigned long)high_memory > VMALLOC_START);
4076 +#endif /* double-sanity-check paranoia */
4077 +
4078 #ifdef CONFIG_X86_PAE
4079 if (!cpu_has_pae)
4080 panic("cannot execute a PAE-enabled kernel on a PAE-less CPU!");
4081 @@ -717,7 +742,7 @@ void __init mem_init(void)
4082 int arch_add_memory(int nid, u64 start, u64 size)
4083 {
4084 struct pglist_data *pgdata = &contig_page_data;
4085 - struct zone *zone = pgdata->node_zones + MAX_NR_ZONES-1;
4086 + struct zone *zone = pgdata->node_zones + ZONE_HIGHMEM;
4087 unsigned long start_pfn = start >> PAGE_SHIFT;
4088 unsigned long nr_pages = size >> PAGE_SHIFT;
4089
4090 --- sle11-2009-05-14.orig/arch/x86/mm/ioremap_32-xen.c 2009-05-14 11:02:43.000000000 +0200
4091 +++ sle11-2009-05-14/arch/x86/mm/ioremap_32-xen.c 2009-03-04 11:28:34.000000000 +0100
4092 @@ -12,7 +12,7 @@
4093 #include <linux/init.h>
4094 #include <linux/slab.h>
4095 #include <linux/module.h>
4096 -#include <asm/io.h>
4097 +#include <linux/io.h>
4098 #include <asm/fixmap.h>
4099 #include <asm/cacheflush.h>
4100 #include <asm/tlbflush.h>
4101 @@ -118,7 +118,7 @@ int direct_remap_pfn_range(struct vm_are
4102 if (domid == DOMID_SELF)
4103 return -EINVAL;
4104
4105 - vma->vm_flags |= VM_IO | VM_RESERVED;
4106 + vma->vm_flags |= VM_IO | VM_RESERVED | VM_PFNMAP;
4107
4108 vma->vm_mm->context.has_foreign_mappings = 1;
4109
4110 @@ -203,6 +203,7 @@ void __iomem * __ioremap(unsigned long p
4111 void __iomem * addr;
4112 struct vm_struct * area;
4113 unsigned long offset, last_addr;
4114 + pgprot_t prot;
4115 domid_t domid = DOMID_IO;
4116
4117 /* Don't allow wraparound or zero size */
4118 @@ -234,6 +235,8 @@ void __iomem * __ioremap(unsigned long p
4119 domid = DOMID_SELF;
4120 }
4121
4122 + prot = __pgprot(_KERNPG_TABLE | flags);
4123 +
4124 /*
4125 * Mappings have to be page-aligned
4126 */
4127 @@ -249,10 +252,9 @@ void __iomem * __ioremap(unsigned long p
4128 return NULL;
4129 area->phys_addr = phys_addr;
4130 addr = (void __iomem *) area->addr;
4131 - flags |= _KERNPG_TABLE;
4132 if (__direct_remap_pfn_range(&init_mm, (unsigned long)addr,
4133 phys_addr>>PAGE_SHIFT,
4134 - size, __pgprot(flags), domid)) {
4135 + size, prot, domid)) {
4136 vunmap((void __force *) addr);
4137 return NULL;
4138 }
4139 --- sle11-2009-05-14.orig/arch/x86/mm/pgtable_32-xen.c 2008-12-01 11:25:57.000000000 +0100
4140 +++ sle11-2009-05-14/arch/x86/mm/pgtable_32-xen.c 2009-03-04 11:28:34.000000000 +0100
4141 @@ -68,7 +68,9 @@ void show_mem(void)
4142 printk(KERN_INFO "%lu pages writeback\n",
4143 global_page_state(NR_WRITEBACK));
4144 printk(KERN_INFO "%lu pages mapped\n", global_page_state(NR_FILE_MAPPED));
4145 - printk(KERN_INFO "%lu pages slab\n", global_page_state(NR_SLAB));
4146 + printk(KERN_INFO "%lu pages slab\n",
4147 + global_page_state(NR_SLAB_RECLAIMABLE) +
4148 + global_page_state(NR_SLAB_UNRECLAIMABLE));
4149 printk(KERN_INFO "%lu pages pagetables\n",
4150 global_page_state(NR_PAGETABLE));
4151 }
4152 @@ -108,18 +110,11 @@ void set_pmd_pfn(unsigned long vaddr, un
4153 __flush_tlb_one(vaddr);
4154 }
4155
4156 -static int nr_fixmaps = 0;
4157 +static int fixmaps;
4158 unsigned long hypervisor_virt_start = HYPERVISOR_VIRT_START;
4159 -unsigned long __FIXADDR_TOP = (HYPERVISOR_VIRT_START - 2 * PAGE_SIZE);
4160 +unsigned long __FIXADDR_TOP = (HYPERVISOR_VIRT_START - PAGE_SIZE);
4161 EXPORT_SYMBOL(__FIXADDR_TOP);
4162
4163 -void __init set_fixaddr_top(unsigned long top)
4164 -{
4165 - BUG_ON(nr_fixmaps > 0);
4166 - hypervisor_virt_start = top;
4167 - __FIXADDR_TOP = hypervisor_virt_start - 2 * PAGE_SIZE;
4168 -}
4169 -
4170 void __set_fixmap (enum fixed_addresses idx, maddr_t phys, pgprot_t flags)
4171 {
4172 unsigned long address = __fix_to_virt(idx);
4173 @@ -141,7 +136,21 @@ void __set_fixmap (enum fixed_addresses
4174 if (HYPERVISOR_update_va_mapping(address, pte,
4175 UVMF_INVLPG|UVMF_ALL))
4176 BUG();
4177 - nr_fixmaps++;
4178 + fixmaps++;
4179 +}
4180 +
4181 +/**
4182 + * reserve_top_address - reserves a hole in the top of kernel address space
4183 + * @reserve - size of hole to reserve
4184 + *
4185 + * Can be used to relocate the fixmap area and poke a hole in the top
4186 + * of kernel address space to make room for a hypervisor.
4187 + */
4188 +void __init reserve_top_address(unsigned long reserve)
4189 +{
4190 + BUG_ON(fixmaps > 0);
4191 + __FIXADDR_TOP = -reserve - PAGE_SIZE;
4192 + __VMALLOC_RESERVE += reserve;
4193 }
4194
4195 pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
4196 --- sle11-2009-05-14.orig/arch/x86/pci/irq-xen.c 2009-05-14 11:02:43.000000000 +0200
4197 +++ sle11-2009-05-14/arch/x86/pci/irq-xen.c 2009-03-04 11:28:34.000000000 +0100
4198 @@ -991,10 +991,6 @@ static void __init pcibios_fixup_irqs(vo
4199 pci_name(bridge), 'A' + pin, irq);
4200 }
4201 if (irq >= 0) {
4202 - if (use_pci_vector() &&
4203 - !platform_legacy_irq(irq))
4204 - irq = IO_APIC_VECTOR(irq);
4205 -
4206 printk(KERN_INFO "PCI->APIC IRQ transform: %s[%c] -> IRQ %d\n",
4207 pci_name(dev), 'A' + pin, irq);
4208 dev->irq = irq;
4209 @@ -1155,10 +1151,6 @@ static int pirq_enable_irq(struct pci_de
4210 }
4211 dev = temp_dev;
4212 if (irq >= 0) {
4213 -#ifdef CONFIG_PCI_MSI
4214 - if (!platform_legacy_irq(irq))
4215 - irq = IO_APIC_VECTOR(irq);
4216 -#endif
4217 printk(KERN_INFO "PCI->APIC IRQ transform: %s[%c] -> IRQ %d\n",
4218 pci_name(dev), 'A' + pin, irq);
4219 dev->irq = irq;
4220 @@ -1179,33 +1171,3 @@ static int pirq_enable_irq(struct pci_de
4221 }
4222 return 0;
4223 }
4224 -
4225 -int pci_vector_resources(int last, int nr_released)
4226 -{
4227 - int count = nr_released;
4228 -
4229 - int next = last;
4230 - int offset = (last % 8);
4231 -
4232 - while (next < FIRST_SYSTEM_VECTOR) {
4233 - next += 8;
4234 -#ifdef CONFIG_X86_64
4235 - if (next == IA32_SYSCALL_VECTOR)
4236 - continue;
4237 -#else
4238 - if (next == SYSCALL_VECTOR)
4239 - continue;
4240 -#endif
4241 - count++;
4242 - if (next >= FIRST_SYSTEM_VECTOR) {
4243 - if (offset%8) {
4244 - next = FIRST_DEVICE_VECTOR + offset;
4245 - offset++;
4246 - continue;
4247 - }
4248 - count--;
4249 - }
4250 - }
4251 -
4252 - return count;
4253 -}
4254 --- sle11-2009-05-14.orig/arch/x86/ia32/ia32entry-xen.S 2009-05-14 11:02:43.000000000 +0200
4255 +++ sle11-2009-05-14/arch/x86/ia32/ia32entry-xen.S 2009-03-04 11:28:34.000000000 +0100
4256 @@ -83,6 +83,7 @@
4257 */
4258 ENTRY(ia32_sysenter_target)
4259 CFI_STARTPROC32 simple
4260 + CFI_SIGNAL_FRAME
4261 CFI_DEF_CFA rsp,SS+8-RIP+16
4262 /*CFI_REL_OFFSET ss,SS-RIP+16*/
4263 CFI_REL_OFFSET rsp,RSP-RIP+16
4264 @@ -164,6 +165,7 @@ ENDPROC(ia32_sysenter_target)
4265 */
4266 ENTRY(ia32_cstar_target)
4267 CFI_STARTPROC32 simple
4268 + CFI_SIGNAL_FRAME
4269 CFI_DEF_CFA rsp,SS+8-RIP+16
4270 /*CFI_REL_OFFSET ss,SS-RIP+16*/
4271 CFI_REL_OFFSET rsp,RSP-RIP+16
4272 @@ -243,6 +245,7 @@ ia32_badarg:
4273
4274 ENTRY(ia32_syscall)
4275 CFI_STARTPROC simple
4276 + CFI_SIGNAL_FRAME
4277 CFI_DEF_CFA rsp,SS+8-RIP+16
4278 /*CFI_REL_OFFSET ss,SS-RIP+16*/
4279 CFI_REL_OFFSET rsp,RSP-RIP+16
4280 @@ -320,6 +323,7 @@ ENTRY(ia32_ptregs_common)
4281 popq %r11
4282 CFI_ENDPROC
4283 CFI_STARTPROC32 simple
4284 + CFI_SIGNAL_FRAME
4285 CFI_DEF_CFA rsp,SS+8-ARGOFFSET
4286 CFI_REL_OFFSET rax,RAX-ARGOFFSET
4287 CFI_REL_OFFSET rcx,RCX-ARGOFFSET
4288 @@ -653,8 +657,8 @@ ia32_sys_call_table:
4289 .quad sys_readlinkat /* 305 */
4290 .quad sys_fchmodat
4291 .quad sys_faccessat
4292 - .quad quiet_ni_syscall /* pselect6 for now */
4293 - .quad quiet_ni_syscall /* ppoll for now */
4294 + .quad compat_sys_pselect6
4295 + .quad compat_sys_ppoll
4296 .quad sys_unshare /* 310 */
4297 .quad compat_sys_set_robust_list
4298 .quad compat_sys_get_robust_list
4299 @@ -663,4 +667,5 @@ ia32_sys_call_table:
4300 .quad sys_tee
4301 .quad compat_sys_vmsplice
4302 .quad compat_sys_move_pages
4303 + .quad sys_getcpu
4304 ia32_syscall_end:
4305 --- sle11-2009-05-14.orig/arch/x86/kernel/Makefile 2009-04-20 11:36:10.000000000 +0200
4306 +++ sle11-2009-05-14/arch/x86/kernel/Makefile 2009-03-04 11:28:34.000000000 +0100
4307 @@ -104,9 +104,9 @@ obj-$(CONFIG_X86_XEN) += fixup.o
4308 ###
4309 # 64 bit specific files
4310 ifeq ($(CONFIG_X86_64),y)
4311 - obj-y += genapic_64.o genapic_flat_64.o genx2apic_uv_x.o tlb_uv.o
4312 - obj-$(CONFIG_X86_XEN_GENAPIC) += genapic_xen_64.o
4313 - obj-y += uv_sysfs.o
4314 + obj-$(CONFIG_X86_LOCAL_APIC) += genapic_64.o genapic_flat_64.o genx2apic_uv_x.o
4315 + obj-$(CONFIG_X86_XEN_GENAPIC) += genapic_64.o genapic_xen_64.o
4316 + obj-y += tlb_uv.o uv_sysfs.o
4317 obj-y += genx2apic_cluster.o
4318 obj-y += genx2apic_phys.o
4319 obj-y += bios_uv.o uv_irq.o
4320 @@ -124,5 +124,7 @@ ifeq ($(CONFIG_X86_64),y)
4321 pci-dma_64-$(CONFIG_XEN) += pci-dma_32.o
4322 endif
4323
4324 -disabled-obj-$(CONFIG_XEN) := i8253.o i8259_$(BITS).o reboot.o smpboot_$(BITS).o tsc_$(BITS).o
4325 +disabled-obj-$(CONFIG_XEN) := early-quirks.o i8253.o i8259_$(BITS).o reboot.o \
4326 + smpboot_$(BITS).o tsc_$(BITS).o
4327 +disabled-obj-$(CONFIG_XEN_UNPRIVILEGED_GUEST) += mpparse_64.o
4328 %/head_$(BITS).o %/head_$(BITS).s: $(if $(CONFIG_XEN),EXTRA_AFLAGS,dummy) :=
4329 --- sle11-2009-05-14.orig/arch/x86/kernel/apic_64-xen.c 2009-05-14 11:02:43.000000000 +0200
4330 +++ sle11-2009-05-14/arch/x86/kernel/apic_64-xen.c 2009-03-04 11:28:34.000000000 +0100
4331 @@ -43,7 +43,7 @@ int apic_verbosity;
4332 */
4333 void ack_bad_irq(unsigned int irq)
4334 {
4335 - printk("unexpected IRQ trap at vector %02x\n", irq);
4336 + printk("unexpected IRQ trap at irq %02x\n", irq);
4337 /*
4338 * Currently unexpected vectors happen only on SMP and APIC.
4339 * We _must_ ack these because every local APIC has only N
4340 @@ -62,19 +62,19 @@ int setup_profiling_timer(unsigned int m
4341 return -EINVAL;
4342 }
4343
4344 -void smp_local_timer_interrupt(struct pt_regs *regs)
4345 +void smp_local_timer_interrupt(void)
4346 {
4347 - profile_tick(CPU_PROFILING, regs);
4348 + profile_tick(CPU_PROFILING);
4349 #ifndef CONFIG_XEN
4350 #ifdef CONFIG_SMP
4351 - update_process_times(user_mode(regs));
4352 + update_process_times(user_mode(get_irq_regs()));
4353 #endif
4354 #endif
4355 /*
4356 * We take the 'long' return path, and there every subsystem
4357 * grabs the appropriate locks (kernel lock/ irq lock).
4358 *
4359 - * we might want to decouple profiling from the 'long path',
4360 + * We might want to decouple profiling from the 'long path',
4361 * and do the profiling totally in assembly.
4362 *
4363 * Currently this isn't too much of an issue (performance wise),
4364 @@ -92,6 +92,8 @@ void smp_local_timer_interrupt(struct pt
4365 */
4366 void smp_apic_timer_interrupt(struct pt_regs *regs)
4367 {
4368 + struct pt_regs *old_regs = set_irq_regs(regs);
4369 +
4370 /*
4371 * the NMI deadlock-detector uses this.
4372 */
4373 @@ -109,8 +111,9 @@ void smp_apic_timer_interrupt(struct pt_
4374 */
4375 exit_idle();
4376 irq_enter();
4377 - smp_local_timer_interrupt(regs);
4378 + smp_local_timer_interrupt();
4379 irq_exit();
4380 + set_irq_regs(old_regs);
4381 }
4382
4383 /*
4384 @@ -188,9 +191,8 @@ int disable_apic;
4385 int __init APIC_init_uniprocessor (void)
4386 {
4387 #ifdef CONFIG_X86_IO_APIC
4388 - if (smp_found_config)
4389 - if (!skip_ioapic_setup && nr_ioapics)
4390 - setup_IO_APIC();
4391 + if (smp_found_config && !skip_ioapic_setup && nr_ioapics)
4392 + setup_IO_APIC();
4393 #endif
4394
4395 return 1;
4396 --- sle11-2009-05-14.orig/arch/x86/kernel/e820_64-xen.c 2009-05-14 11:02:43.000000000 +0200
4397 +++ sle11-2009-05-14/arch/x86/kernel/e820_64-xen.c 2009-03-04 11:28:34.000000000 +0100
4398 @@ -16,6 +16,7 @@
4399 #include <linux/string.h>
4400 #include <linux/kexec.h>
4401 #include <linux/module.h>
4402 +#include <linux/mm.h>
4403
4404 #include <asm/pgtable.h>
4405 #include <asm/page.h>
4406 @@ -25,6 +26,11 @@
4407 #include <asm/sections.h>
4408 #include <xen/interface/memory.h>
4409
4410 +struct e820map e820 __initdata;
4411 +#ifdef CONFIG_XEN
4412 +struct e820map machine_e820 __initdata;
4413 +#endif
4414 +
4415 /*
4416 * PFN of last memory page.
4417 */
4418 @@ -41,14 +47,10 @@ unsigned long end_pfn_map;
4419 /*
4420 * Last pfn which the user wants to use.
4421 */
4422 -unsigned long end_user_pfn = MAXMEM>>PAGE_SHIFT;
4423 +static unsigned long __initdata end_user_pfn = MAXMEM>>PAGE_SHIFT;
4424
4425 extern struct resource code_resource, data_resource;
4426
4427 -#ifdef CONFIG_XEN
4428 -extern struct e820map machine_e820;
4429 -#endif
4430 -
4431 /* Check for some hardcoded bad areas that early boot is not allowed to touch */
4432 static inline int bad_addr(unsigned long *addrp, unsigned long size)
4433 {
4434 @@ -57,13 +59,13 @@ static inline int bad_addr(unsigned long
4435 #ifndef CONFIG_XEN
4436 /* various gunk below that needed for SMP startup */
4437 if (addr < 0x8000) {
4438 - *addrp = 0x8000;
4439 + *addrp = PAGE_ALIGN(0x8000);
4440 return 1;
4441 }
4442
4443 /* direct mapping tables of the kernel */
4444 if (last >= table_start<<PAGE_SHIFT && addr < table_end<<PAGE_SHIFT) {
4445 - *addrp = table_end << PAGE_SHIFT;
4446 + *addrp = PAGE_ALIGN(table_end << PAGE_SHIFT);
4447 return 1;
4448 }
4449
4450 @@ -71,23 +73,18 @@ static inline int bad_addr(unsigned long
4451 #ifdef CONFIG_BLK_DEV_INITRD
4452 if (LOADER_TYPE && INITRD_START && last >= INITRD_START &&
4453 addr < INITRD_START+INITRD_SIZE) {
4454 - *addrp = INITRD_START + INITRD_SIZE;
4455 + *addrp = PAGE_ALIGN(INITRD_START + INITRD_SIZE);
4456 return 1;
4457 }
4458 #endif
4459 - /* kernel code + 640k memory hole (later should not be needed, but
4460 - be paranoid for now) */
4461 - if (last >= 640*1024 && addr < 1024*1024) {
4462 - *addrp = 1024*1024;
4463 - return 1;
4464 - }
4465 - if (last >= __pa_symbol(&_text) && last < __pa_symbol(&_end)) {
4466 - *addrp = __pa_symbol(&_end);
4467 + /* kernel code */
4468 + if (last >= __pa_symbol(&_text) && addr < __pa_symbol(&_end)) {
4469 + *addrp = PAGE_ALIGN(__pa_symbol(&_end));
4470 return 1;
4471 }
4472
4473 if (last >= ebda_addr && addr < ebda_addr + ebda_size) {
4474 - *addrp = ebda_addr + ebda_size;
4475 + *addrp = PAGE_ALIGN(ebda_addr + ebda_size);
4476 return 1;
4477 }
4478
4479 @@ -184,7 +181,7 @@ unsigned long __init find_e820_area(unsi
4480 continue;
4481 while (bad_addr(&addr, size) && addr+size <= ei->addr+ei->size)
4482 ;
4483 - last = addr + size;
4484 + last = PAGE_ALIGN(addr) + size;
4485 if (last > ei->addr + ei->size)
4486 continue;
4487 if (last > end)
4488 @@ -194,59 +191,14 @@ unsigned long __init find_e820_area(unsi
4489 return -1UL;
4490 }
4491
4492 -/*
4493 - * Free bootmem based on the e820 table for a node.
4494 - */
4495 -void __init e820_bootmem_free(pg_data_t *pgdat, unsigned long start,unsigned long end)
4496 -{
4497 - int i;
4498 - for (i = 0; i < e820.nr_map; i++) {
4499 - struct e820entry *ei = &e820.map[i];
4500 - unsigned long last, addr;
4501 -
4502 - if (ei->type != E820_RAM ||
4503 - ei->addr+ei->size <= start ||
4504 - ei->addr >= end)
4505 - continue;
4506 -
4507 - addr = round_up(ei->addr, PAGE_SIZE);
4508 - if (addr < start)
4509 - addr = start;
4510 -
4511 - last = round_down(ei->addr + ei->size, PAGE_SIZE);
4512 - if (last >= end)
4513 - last = end;
4514 -
4515 - if (last > addr && last-addr >= PAGE_SIZE)
4516 - free_bootmem_node(pgdat, addr, last-addr);
4517 - }
4518 -}
4519 -
4520 /*
4521 * Find the highest page frame number we have available
4522 */
4523 unsigned long __init e820_end_of_ram(void)
4524 {
4525 - int i;
4526 unsigned long end_pfn = 0;
4527 + end_pfn = find_max_pfn_with_active_regions();
4528
4529 - for (i = 0; i < e820.nr_map; i++) {
4530 - struct e820entry *ei = &e820.map[i];
4531 - unsigned long start, end;
4532 -
4533 - start = round_up(ei->addr, PAGE_SIZE);
4534 - end = round_down(ei->addr + ei->size, PAGE_SIZE);
4535 - if (start >= end)
4536 - continue;
4537 - if (ei->type == E820_RAM) {
4538 - if (end > end_pfn<<PAGE_SHIFT)
4539 - end_pfn = end>>PAGE_SHIFT;
4540 - } else {
4541 - if (end > end_pfn_map<<PAGE_SHIFT)
4542 - end_pfn_map = end>>PAGE_SHIFT;
4543 - }
4544 - }
4545 -
4546 if (end_pfn > end_pfn_map)
4547 end_pfn_map = end_pfn;
4548 if (end_pfn_map > MAXMEM>>PAGE_SHIFT)
4549 @@ -256,43 +208,10 @@ unsigned long __init e820_end_of_ram(voi
4550 if (end_pfn > end_pfn_map)
4551 end_pfn = end_pfn_map;
4552
4553 + printk("end_pfn_map = %lu\n", end_pfn_map);
4554 return end_pfn;
4555 }
4556
4557 -/*
4558 - * Compute how much memory is missing in a range.
4559 - * Unlike the other functions in this file the arguments are in page numbers.
4560 - */
4561 -unsigned long __init
4562 -e820_hole_size(unsigned long start_pfn, unsigned long end_pfn)
4563 -{
4564 - unsigned long ram = 0;
4565 - unsigned long start = start_pfn << PAGE_SHIFT;
4566 - unsigned long end = end_pfn << PAGE_SHIFT;
4567 - int i;
4568 - for (i = 0; i < e820.nr_map; i++) {
4569 - struct e820entry *ei = &e820.map[i];
4570 - unsigned long last, addr;
4571 -
4572 - if (ei->type != E820_RAM ||
4573 - ei->addr+ei->size <= start ||
4574 - ei->addr >= end)
4575 - continue;
4576 -
4577 - addr = round_up(ei->addr, PAGE_SIZE);
4578 - if (addr < start)
4579 - addr = start;
4580 -
4581 - last = round_down(ei->addr + ei->size, PAGE_SIZE);
4582 - if (last >= end)
4583 - last = end;
4584 -
4585 - if (last > addr)
4586 - ram += last - addr;
4587 - }
4588 - return ((end - start) - ram) >> PAGE_SHIFT;
4589 -}
4590 -
4591 /*
4592 * Mark e820 reserved areas as busy for the resource manager.
4593 */
4594 @@ -333,6 +252,98 @@ void __init e820_reserve_resources(struc
4595 }
4596 }
4597
4598 +#ifndef CONFIG_XEN
4599 +/* Mark pages corresponding to given address range as nosave */
4600 +static void __init
4601 +e820_mark_nosave_range(unsigned long start, unsigned long end)
4602 +{
4603 + unsigned long pfn, max_pfn;
4604 +
4605 + if (start >= end)
4606 + return;
4607 +
4608 + printk("Nosave address range: %016lx - %016lx\n", start, end);
4609 + max_pfn = end >> PAGE_SHIFT;
4610 + for (pfn = start >> PAGE_SHIFT; pfn < max_pfn; pfn++)
4611 + if (pfn_valid(pfn))
4612 + SetPageNosave(pfn_to_page(pfn));
4613 +}
4614 +
4615 +/*
4616 + * Find the ranges of physical addresses that do not correspond to
4617 + * e820 RAM areas and mark the corresponding pages as nosave for software
4618 + * suspend and suspend to RAM.
4619 + *
4620 + * This function requires the e820 map to be sorted and without any
4621 + * overlapping entries and assumes the first e820 area to be RAM.
4622 + */
4623 +void __init e820_mark_nosave_regions(void)
4624 +{
4625 + int i;
4626 + unsigned long paddr;
4627 +
4628 + paddr = round_down(e820.map[0].addr + e820.map[0].size, PAGE_SIZE);
4629 + for (i = 1; i < e820.nr_map; i++) {
4630 + struct e820entry *ei = &e820.map[i];
4631 +
4632 + if (paddr < ei->addr)
4633 + e820_mark_nosave_range(paddr,
4634 + round_up(ei->addr, PAGE_SIZE));
4635 +
4636 + paddr = round_down(ei->addr + ei->size, PAGE_SIZE);
4637 + if (ei->type != E820_RAM)
4638 + e820_mark_nosave_range(round_up(ei->addr, PAGE_SIZE),
4639 + paddr);
4640 +
4641 + if (paddr >= (end_pfn << PAGE_SHIFT))
4642 + break;
4643 + }
4644 +}
4645 +#endif
4646 +
4647 +/* Walk the e820 map and register active regions within a node */
4648 +void __init
4649 +e820_register_active_regions(int nid, unsigned long start_pfn,
4650 + unsigned long end_pfn)
4651 +{
4652 + int i;
4653 + unsigned long ei_startpfn, ei_endpfn;
4654 + for (i = 0; i < e820.nr_map; i++) {
4655 + struct e820entry *ei = &e820.map[i];
4656 + ei_startpfn = round_up(ei->addr, PAGE_SIZE) >> PAGE_SHIFT;
4657 + ei_endpfn = round_down(ei->addr + ei->size, PAGE_SIZE)
4658 + >> PAGE_SHIFT;
4659 +
4660 + /* Skip map entries smaller than a page */
4661 + if (ei_startpfn >= ei_endpfn)
4662 + continue;
4663 +
4664 + /* Check if end_pfn_map should be updated */
4665 + if (ei->type != E820_RAM && ei_endpfn > end_pfn_map)
4666 + end_pfn_map = ei_endpfn;
4667 +
4668 + /* Skip if map is outside the node */
4669 + if (ei->type != E820_RAM ||
4670 + ei_endpfn <= start_pfn ||
4671 + ei_startpfn >= end_pfn)
4672 + continue;
4673 +
4674 + /* Check for overlaps */
4675 + if (ei_startpfn < start_pfn)
4676 + ei_startpfn = start_pfn;
4677 + if (ei_endpfn > end_pfn)
4678 + ei_endpfn = end_pfn;
4679 +
4680 + /* Obey end_user_pfn to save on memmap */
4681 + if (ei_startpfn >= end_user_pfn)
4682 + continue;
4683 + if (ei_endpfn > end_user_pfn)
4684 + ei_endpfn = end_user_pfn;
4685 +
4686 + add_active_range(nid, ei_startpfn, ei_endpfn);
4687 + }
4688 +}
4689 +
4690 /*
4691 * Add a memory region to the kernel e820 map.
4692 */
4693 @@ -553,13 +564,6 @@ static int __init sanitize_e820_map(stru
4694 * If we're lucky and live on a modern system, the setup code
4695 * will have given us a memory map that we can use to properly
4696 * set up memory. If we aren't, we'll fake a memory map.
4697 - *
4698 - * We check to see that the memory map contains at least 2 elements
4699 - * before we'll use it, because the detection code in setup.S may
4700 - * not be perfect and most every PC known to man has two memory
4701 - * regions: one from 0 to 640k, and one from 1mb up. (The IBM
4702 - * thinkpad 560x, for example, does not cooperate with the memory
4703 - * detection code.)
4704 */
4705 static int __init copy_e820_map(struct e820entry * biosmap, int nr_map)
4706 {
4707 @@ -581,27 +585,6 @@ static int __init copy_e820_map(struct e
4708 if (start > end)
4709 return -1;
4710
4711 -#ifndef CONFIG_XEN
4712 - /*
4713 - * Some BIOSes claim RAM in the 640k - 1M region.
4714 - * Not right. Fix it up.
4715 - *
4716 - * This should be removed on Hammer which is supposed to not
4717 - * have non e820 covered ISA mappings there, but I had some strange
4718 - * problems so it stays for now. -AK
4719 - */
4720 - if (type == E820_RAM) {
4721 - if (start < 0x100000ULL && end > 0xA0000ULL) {
4722 - if (start < 0xA0000ULL)
4723 - add_memory_region(start, 0xA0000ULL-start, type);
4724 - if (end <= 0x100000ULL)
4725 - continue;
4726 - start = 0x100000ULL;
4727 - size = end - start;
4728 - }
4729 - }
4730 -#endif
4731 -
4732 add_memory_region(start, size, type);
4733 } while (biosmap++,--nr_map);
4734
4735 @@ -622,11 +605,15 @@ static int __init copy_e820_map(struct e
4736 return 0;
4737 }
4738
4739 +void early_panic(char *msg)
4740 +{
4741 + early_printk(msg);
4742 + panic(msg);
4743 +}
4744 +
4745 #ifndef CONFIG_XEN
4746 void __init setup_memory_region(void)
4747 {
4748 - char *who = "BIOS-e820";
4749 -
4750 /*
4751 * Try to copy the BIOS-supplied E820-map.
4752 *
4753 @@ -634,24 +621,10 @@ void __init setup_memory_region(void)
4754 * the next section from 1mb->appropriate_mem_k
4755 */
4756 sanitize_e820_map(E820_MAP, &E820_MAP_NR);
4757 - if (copy_e820_map(E820_MAP, E820_MAP_NR) < 0) {
4758 - unsigned long mem_size;
4759 -
4760 - /* compare results from other methods and take the greater */
4761 - if (ALT_MEM_K < EXT_MEM_K) {
4762 - mem_size = EXT_MEM_K;
4763 - who = "BIOS-88";
4764 - } else {
4765 - mem_size = ALT_MEM_K;
4766 - who = "BIOS-e801";
4767 - }
4768 -
4769 - e820.nr_map = 0;
4770 - add_memory_region(0, LOWMEMSIZE(), E820_RAM);
4771 - add_memory_region(HIGH_MEMORY, mem_size << 10, E820_RAM);
4772 - }
4773 + if (copy_e820_map(E820_MAP, E820_MAP_NR) < 0)
4774 + early_panic("Cannot find a valid memory map");
4775 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
4776 - e820_print_map(who);
4777 + e820_print_map("BIOS-e820");
4778 }
4779
4780 #else /* CONFIG_XEN */
4781 @@ -683,20 +656,23 @@ void __init setup_memory_region(void)
4782
4783 sanitize_e820_map(map, (char *)&memmap.nr_entries);
4784
4785 - BUG_ON(copy_e820_map(map, (char)memmap.nr_entries) < 0);
4786 + if (copy_e820_map(map, (char)memmap.nr_entries) < 0)
4787 + early_panic("Cannot find a valid memory map");
4788
4789 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
4790 e820_print_map("Xen");
4791 }
4792 #endif
4793
4794 -void __init parse_memopt(char *p, char **from)
4795 -{
4796 +static int __init parse_memopt(char *p)
4797 +{
4798 int i;
4799 unsigned long current_end;
4800 unsigned long end;
4801
4802 - end_user_pfn = memparse(p, from);
4803 + if (!p)
4804 + return -EINVAL;
4805 + end_user_pfn = memparse(p, &p);
4806 end_user_pfn >>= PAGE_SHIFT;
4807
4808 end = end_user_pfn<<PAGE_SHIFT;
4809 @@ -713,27 +689,61 @@ void __init parse_memopt(char *p, char *
4810 else
4811 add_memory_region(current_end, end - current_end, E820_RAM);
4812 }
4813 +
4814 + return 0;
4815 }
4816 +early_param("mem", parse_memopt);
4817 +
4818 +static int userdef __initdata;
4819
4820 -void __init parse_memmapopt(char *p, char **from)
4821 +static int __init parse_memmap_opt(char *p)
4822 {
4823 + char *oldp;
4824 unsigned long long start_at, mem_size;
4825
4826 - mem_size = memparse(p, from);
4827 - p = *from;
4828 + if (!strcmp(p, "exactmap")) {
4829 +#ifdef CONFIG_CRASH_DUMP
4830 + /* If we are doing a crash dump, we
4831 + * still need to know the real mem
4832 + * size before original memory map is
4833 + * reset.
4834 + */
4835 + e820_register_active_regions(0, 0, -1UL);
4836 + saved_max_pfn = e820_end_of_ram();
4837 + remove_all_active_ranges();
4838 +#endif
4839 + end_pfn_map = 0;
4840 + e820.nr_map = 0;
4841 + userdef = 1;
4842 + return 0;
4843 + }
4844 +
4845 + oldp = p;
4846 + mem_size = memparse(p, &p);
4847 + if (p == oldp)
4848 + return -EINVAL;
4849 if (*p == '@') {
4850 - start_at = memparse(p+1, from);
4851 + start_at = memparse(p+1, &p);
4852 add_memory_region(start_at, mem_size, E820_RAM);
4853 } else if (*p == '#') {
4854 - start_at = memparse(p+1, from);
4855 + start_at = memparse(p+1, &p);
4856 add_memory_region(start_at, mem_size, E820_ACPI);
4857 } else if (*p == '$') {
4858 - start_at = memparse(p+1, from);
4859 + start_at = memparse(p+1, &p);
4860 add_memory_region(start_at, mem_size, E820_RESERVED);
4861 } else {
4862 end_user_pfn = (mem_size >> PAGE_SHIFT);
4863 }
4864 - p = *from;
4865 + return *p == '\0' ? 0 : -EINVAL;
4866 +}
4867 +early_param("memmap", parse_memmap_opt);
4868 +
4869 +void finish_e820_parsing(void)
4870 +{
4871 + if (userdef) {
4872 + printk(KERN_INFO "user-defined physical RAM map:\n");
4873 + e820_print_map("user");
4874 + }
4875 }
4876
4877 unsigned long pci_mem_start = 0xaeedbabe;
4878 --- sle11-2009-05-14.orig/arch/x86/kernel/early_printk-xen.c 2009-05-14 11:02:43.000000000 +0200
4879 +++ sle11-2009-05-14/arch/x86/kernel/early_printk-xen.c 2009-03-04 11:28:34.000000000 +0100
4880 @@ -244,20 +244,16 @@ void early_printk(const char *fmt, ...)
4881
4882 static int __initdata keep_early;
4883
4884 -int __init setup_early_printk(char *opt)
4885 +static int __init setup_early_printk(char *buf)
4886 {
4887 - char *space;
4888 - char buf[256];
4889 + if (!buf)
4890 + return 0;
4891
4892 if (early_console_initialized)
4893 - return 1;
4894 -
4895 - strlcpy(buf,opt,sizeof(buf));
4896 - space = strchr(buf, ' ');
4897 - if (space)
4898 - *space = 0;
4899 + return 0;
4900 + early_console_initialized = 1;
4901
4902 - if (strstr(buf,"keep"))
4903 + if (strstr(buf, "keep"))
4904 keep_early = 1;
4905
4906 if (!strncmp(buf, "serial", 6)) {
4907 @@ -281,11 +277,12 @@ int __init setup_early_printk(char *opt)
4908 early_console = &simnow_console;
4909 keep_early = 1;
4910 }
4911 - early_console_initialized = 1;
4912 register_console(early_console);
4913 return 0;
4914 }
4915
4916 +early_param("earlyprintk", setup_early_printk);
4917 +
4918 void __init disable_early_printk(void)
4919 {
4920 if (!early_console_initialized || !early_console)
4921 @@ -299,4 +296,3 @@ void __init disable_early_printk(void)
4922 }
4923 }
4924
4925 -__setup("earlyprintk=", setup_early_printk);
4926 --- sle11-2009-05-14.orig/arch/x86/kernel/entry_64-xen.S 2009-05-14 11:02:43.000000000 +0200
4927 +++ sle11-2009-05-14/arch/x86/kernel/entry_64-xen.S 2009-03-04 11:28:34.000000000 +0100
4928 @@ -4,9 +4,6 @@
4929 * Copyright (C) 1991, 1992 Linus Torvalds
4930 * Copyright (C) 2000, 2001, 2002 Andi Kleen SuSE Labs
4931 * Copyright (C) 2000 Pavel Machek <pavel@suse.cz>
4932 - *
4933 - * $Id$
4934 - *
4935 * Jun Nakajima <jun.nakajima@intel.com>
4936 * Asit Mallick <asit.k.mallick@intel.com>
4937 * Modified for Xen
4938 @@ -26,15 +23,25 @@
4939 * at the top of the kernel process stack.
4940 * - partial stack frame: partially saved registers upto R11.
4941 * - full stack frame: Like partial stack frame, but all register saved.
4942 - *
4943 - * TODO:
4944 - * - schedule it carefully for the final hardware.
4945 + *
4946 + * Some macro usage:
4947 + * - CFI macros are used to generate dwarf2 unwind information for better
4948 + * backtraces. They don't change any code.
4949 + * - SAVE_ALL/RESTORE_ALL - Save/restore all registers
4950 + * - SAVE_ARGS/RESTORE_ARGS - Save/restore registers that C functions modify.
4951 + * There are unfortunately lots of special cases where some registers
4952 + * not touched. The macro is a big mess that should be cleaned up.
4953 + * - SAVE_REST/RESTORE_REST - Handle the registers not saved by SAVE_ARGS.
4954 + * Gives a full stack frame.
4955 + * - ENTRY/END Define functions in the symbol table.
4956 + * - FIXUP_TOP_OF_STACK/RESTORE_TOP_OF_STACK - Fix up the hardware stack
4957 + * frame that is otherwise undefined after a SYSCALL
4958 + * - TRACE_IRQ_* - Trace hard interrupt state for lock debugging.
4959 + * - errorentry/paranoidentry/zeroentry - Define exception entry points.
4960 */
4961
4962 -#define ASSEMBLY 1
4963 #include <linux/linkage.h>
4964 #include <asm/segment.h>
4965 -#include <asm/smp.h>
4966 #include <asm/cache.h>
4967 #include <asm/errno.h>
4968 #include <asm/dwarf2.h>
4969 @@ -117,6 +124,7 @@ NMI_MASK = 0x80000000
4970 .macro CFI_DEFAULT_STACK start=1,adj=0
4971 .if \start
4972 CFI_STARTPROC simple
4973 + CFI_SIGNAL_FRAME
4974 CFI_DEF_CFA rsp,SS+8 - \adj*ARGOFFSET
4975 .else
4976 CFI_DEF_CFA_OFFSET SS+8 - \adj*ARGOFFSET
4977 @@ -207,6 +215,7 @@ END(ret_from_fork)
4978 */
4979 .macro _frame ref
4980 CFI_STARTPROC simple
4981 + CFI_SIGNAL_FRAME
4982 CFI_DEF_CFA rsp,SS+8-\ref
4983 /*CFI_REL_OFFSET ss,SS-\ref*/
4984 CFI_REL_OFFSET rsp,RSP-\ref
4985 @@ -334,6 +343,8 @@ tracesys:
4986 LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */
4987 RESTORE_REST
4988 cmpq $__NR_syscall_max,%rax
4989 + movq $-ENOSYS,%rcx
4990 + cmova %rcx,%rax
4991 ja 1f
4992 movq %r10,%rcx /* fixup for C */
4993 call *sys_call_table(,%rax,8)
4994 @@ -349,6 +360,7 @@ END(system_call)
4995 */
4996 ENTRY(int_ret_from_sys_call)
4997 CFI_STARTPROC simple
4998 + CFI_SIGNAL_FRAME
4999 CFI_DEF_CFA rsp,SS+8-ARGOFFSET
5000 /*CFI_REL_OFFSET ss,SS-ARGOFFSET*/
5001 CFI_REL_OFFSET rsp,RSP-ARGOFFSET
5002 @@ -583,8 +595,7 @@ retint_signal:
5003 #ifdef CONFIG_PREEMPT
5004 /* Returning to kernel space. Check if we need preemption */
5005 /* rcx: threadinfo. interrupts off. */
5006 - .p2align
5007 -retint_kernel:
5008 +ENTRY(retint_kernel)
5009 cmpl $0,threadinfo_preempt_count(%rcx)
5010 jnz retint_restore_args
5011 bt $TIF_NEED_RESCHED,threadinfo_flags(%rcx)
5012 @@ -644,7 +655,6 @@ ENTRY(call_function_interrupt)
5013 END(call_function_interrupt)
5014 #endif
5015
5016 -#ifdef CONFIG_X86_LOCAL_APIC
5017 ENTRY(apic_timer_interrupt)
5018 apicinterrupt LOCAL_TIMER_VECTOR,smp_apic_timer_interrupt
5019 END(apic_timer_interrupt)
5020 @@ -656,7 +666,6 @@ END(error_interrupt)
5021 ENTRY(spurious_interrupt)
5022 apicinterrupt SPURIOUS_APIC_VECTOR,smp_spurious_interrupt
5023 END(spurious_interrupt)
5024 -#endif
5025 #endif /* !CONFIG_XEN */
5026
5027 /*
5028 @@ -755,7 +764,9 @@ paranoid_exit\trace:
5029 testl $3,CS(%rsp)
5030 jnz paranoid_userspace\trace
5031 paranoid_swapgs\trace:
5032 + .if \trace
5033 TRACE_IRQS_IRETQ 0
5034 + .endif
5035 swapgs
5036 paranoid_restore\trace:
5037 RESTORE_ALL 8
5038 @@ -802,7 +813,7 @@ paranoid_schedule\trace:
5039 * Exception entry point. This expects an error code/orig_rax on the stack
5040 * and the exception handler in %rax.
5041 */
5042 -ENTRY(error_entry)
5043 +KPROBE_ENTRY(error_entry)
5044 _frame RDI
5045 CFI_REL_OFFSET rax,0
5046 /* rdi slot contains rax, oldrax contains error code */
5047 @@ -896,7 +907,7 @@ error_kernelspace:
5048 jmp error_sti
5049 #endif
5050 CFI_ENDPROC
5051 -END(error_entry)
5052 +KPROBE_END(error_entry)
5053
5054 ENTRY(hypervisor_callback)
5055 zeroentry do_hypervisor_callback
5056 @@ -936,26 +947,6 @@ ENTRY(do_hypervisor_callback) # do_hyp
5057 CFI_ENDPROC
5058 END(do_hypervisor_callback)
5059
5060 -#ifdef CONFIG_X86_LOCAL_APIC
5061 -KPROBE_ENTRY(nmi)
5062 - zeroentry do_nmi_callback
5063 -ENTRY(do_nmi_callback)
5064 - CFI_STARTPROC
5065 - addq $8, %rsp
5066 - CFI_ENDPROC
5067 - CFI_DEFAULT_STACK
5068 - call do_nmi
5069 - orl $NMI_MASK,EFLAGS(%rsp)
5070 - RESTORE_REST
5071 - XEN_BLOCK_EVENTS(%rsi)
5072 - TRACE_IRQS_OFF
5073 - GET_THREAD_INFO(%rcx)
5074 - jmp retint_restore_args
5075 - CFI_ENDPROC
5076 - .previous .text
5077 -END(nmi)
5078 -#endif
5079 -
5080 ALIGN
5081 restore_all_enable_events:
5082 CFI_DEFAULT_STACK adj=1
5083 @@ -1121,7 +1112,7 @@ ENDPROC(child_rip)
5084 * do_sys_execve asm fallback arguments:
5085 * rdi: name, rsi: argv, rdx: envp, fake frame on the stack
5086 */
5087 -ENTRY(execve)
5088 +ENTRY(kernel_execve)
5089 CFI_STARTPROC
5090 FAKE_STACK_FRAME $0
5091 SAVE_ALL
5092 @@ -1135,12 +1126,11 @@ ENTRY(execve)
5093 UNFAKE_STACK_FRAME
5094 ret
5095 CFI_ENDPROC
5096 -ENDPROC(execve)
5097 +ENDPROC(kernel_execve)
5098
5099 KPROBE_ENTRY(page_fault)
5100 errorentry do_page_fault
5101 -END(page_fault)
5102 - .previous .text
5103 +KPROBE_END(page_fault)
5104
5105 ENTRY(coprocessor_error)
5106 zeroentry do_coprocessor_error
5107 @@ -1162,25 +1152,25 @@ KPROBE_ENTRY(debug)
5108 zeroentry do_debug
5109 /* paranoidexit
5110 CFI_ENDPROC */
5111 -END(debug)
5112 - .previous .text
5113 +KPROBE_END(debug)
5114
5115 -#if 0
5116 - /* runs on exception stack */
5117 KPROBE_ENTRY(nmi)
5118 - INTR_FRAME
5119 - pushq $-1
5120 - CFI_ADJUST_CFA_OFFSET 8
5121 - paranoidentry do_nmi, 0, 0
5122 -#ifdef CONFIG_TRACE_IRQFLAGS
5123 - paranoidexit 0
5124 -#else
5125 - jmp paranoid_exit1
5126 - CFI_ENDPROC
5127 -#endif
5128 -END(nmi)
5129 - .previous .text
5130 -#endif
5131 + zeroentry do_nmi_callback
5132 +KPROBE_END(nmi)
5133 +do_nmi_callback:
5134 + CFI_STARTPROC
5135 + addq $8, %rsp
5136 + CFI_ENDPROC
5137 + CFI_DEFAULT_STACK
5138 + call do_nmi
5139 + orl $NMI_MASK,EFLAGS(%rsp)
5140 + RESTORE_REST
5141 + XEN_BLOCK_EVENTS(%rsi)
5142 + TRACE_IRQS_OFF
5143 + GET_THREAD_INFO(%rcx)
5144 + jmp retint_restore_args
5145 + CFI_ENDPROC
5146 +END(do_nmi_callback)
5147
5148 KPROBE_ENTRY(int3)
5149 /* INTR_FRAME
5150 @@ -1189,8 +1179,7 @@ KPROBE_ENTRY(int3)
5151 zeroentry do_int3
5152 /* jmp paranoid_exit1
5153 CFI_ENDPROC */
5154 -END(int3)
5155 - .previous .text
5156 +KPROBE_END(int3)
5157
5158 ENTRY(overflow)
5159 zeroentry do_overflow
5160 @@ -1241,8 +1230,7 @@ END(stack_segment)
5161
5162 KPROBE_ENTRY(general_protection)
5163 errorentry do_general_protection
5164 -END(general_protection)
5165 - .previous .text
5166 +KPROBE_END(general_protection)
5167
5168 ENTRY(alignment_check)
5169 errorentry do_alignment_check
5170 --- sle11-2009-05-14.orig/arch/x86/kernel/genapic_xen_64.c 2009-05-14 11:02:43.000000000 +0200
5171 +++ sle11-2009-05-14/arch/x86/kernel/genapic_xen_64.c 2009-03-04 11:28:34.000000000 +0100
5172 @@ -71,6 +71,13 @@ static cpumask_t xen_target_cpus(void)
5173 return cpu_online_map;
5174 }
5175
5176 +static cpumask_t xen_vector_allocation_domain(int cpu)
5177 +{
5178 + cpumask_t domain = CPU_MASK_NONE;
5179 + cpu_set(cpu, domain);
5180 + return domain;
5181 +}
5182 +
5183 /*
5184 * Set up the logical destination ID.
5185 * Do nothing, not called now.
5186 @@ -147,8 +154,8 @@ struct genapic apic_xen = {
5187 .int_delivery_mode = dest_LowestPrio,
5188 #endif
5189 .int_dest_mode = (APIC_DEST_LOGICAL != 0),
5190 - .int_delivery_dest = APIC_DEST_LOGICAL | APIC_DM_LOWEST,
5191 .target_cpus = xen_target_cpus,
5192 + .vector_allocation_domain = xen_vector_allocation_domain,
5193 #ifdef CONFIG_XEN_PRIVILEGED_GUEST
5194 .apic_id_registered = xen_apic_id_registered,
5195 #endif
5196 --- sle11-2009-05-14.orig/arch/x86/kernel/head_64-xen.S 2009-05-14 11:02:43.000000000 +0200
5197 +++ sle11-2009-05-14/arch/x86/kernel/head_64-xen.S 2009-03-04 11:28:34.000000000 +0100
5198 @@ -5,9 +5,6 @@
5199 * Copyright (C) 2000 Pavel Machek <pavel@suse.cz>
5200 * Copyright (C) 2000 Karsten Keil <kkeil@suse.de>
5201 * Copyright (C) 2001,2002 Andi Kleen <ak@suse.de>
5202 - *
5203 - * $Id: head.S,v 1.49 2002/03/19 17:39:25 ak Exp $
5204 - *
5205 * Jun Nakajima <jun.nakajima@intel.com>
5206 * Modified for Xen
5207 */
5208 @@ -149,7 +146,7 @@ ENTRY(cpu_gdt_table)
5209 .quad 0,0 /* TSS */
5210 .quad 0,0 /* LDT */
5211 .quad 0,0,0 /* three TLS descriptors */
5212 - .quad 0 /* unused */
5213 + .quad 0x0000f40000000000 /* node/CPU stored in limit */
5214 gdt_end:
5215 /* asm/segment.h:GDT_ENTRIES must match this */
5216 /* This should be a multiple of the cache line size */
5217 --- sle11-2009-05-14.orig/arch/x86/kernel/head64-xen.c 2009-05-14 11:02:43.000000000 +0200
5218 +++ sle11-2009-05-14/arch/x86/kernel/head64-xen.c 2009-03-04 11:28:34.000000000 +0100
5219 @@ -7,6 +7,9 @@
5220 * Modified for Xen.
5221 */
5222
5223 +/* PDA is not ready to be used until the end of x86_64_start_kernel(). */
5224 +#define arch_use_lazy_mmu_mode() false
5225 +
5226 #include <linux/init.h>
5227 #include <linux/linkage.h>
5228 #include <linux/types.h>
5229 @@ -54,11 +57,9 @@ static void __init copy_bootdata(char *r
5230 new_data = *(int *) (x86_boot_params + NEW_CL_POINTER);
5231 if (!new_data) {
5232 if (OLD_CL_MAGIC != * (u16 *) OLD_CL_MAGIC_ADDR) {
5233 - printk("so old bootloader that it does not support commandline?!\n");
5234 return;
5235 }
5236 new_data = OLD_CL_BASE_ADDR + * (u16 *) OLD_CL_OFFSET;
5237 - printk("old bootloader convention, maybe loadlin?\n");
5238 }
5239 command_line = (char *) ((u64)(new_data));
5240 memcpy(saved_command_line, command_line, COMMAND_LINE_SIZE);
5241 @@ -70,25 +71,6 @@ static void __init copy_bootdata(char *r
5242 memcpy(saved_command_line, xen_start_info->cmd_line, max_cmdline);
5243 saved_command_line[max_cmdline-1] = '\0';
5244 #endif
5245 - printk("Bootdata ok (command line is %s)\n", saved_command_line);
5246 -}
5247 -
5248 -static void __init setup_boot_cpu_data(void)
5249 -{
5250 - unsigned int dummy, eax;
5251 -
5252 - /* get vendor info */
5253 - cpuid(0, (unsigned int *)&boot_cpu_data.cpuid_level,
5254 - (unsigned int *)&boot_cpu_data.x86_vendor_id[0],
5255 - (unsigned int *)&boot_cpu_data.x86_vendor_id[8],
5256 - (unsigned int *)&boot_cpu_data.x86_vendor_id[4]);
5257 -
5258 - /* get cpu type */
5259 - cpuid(1, &eax, &dummy, &dummy,
5260 - (unsigned int *) &boot_cpu_data.x86_capability);
5261 - boot_cpu_data.x86 = (eax >> 8) & 0xf;
5262 - boot_cpu_data.x86_model = (eax >> 4) & 0xf;
5263 - boot_cpu_data.x86_mask = eax & 0xf;
5264 }
5265
5266 #include <xen/interface/memory.h>
5267 @@ -101,7 +83,6 @@ void __init x86_64_start_kernel(char * r
5268 {
5269 struct xen_machphys_mapping mapping;
5270 unsigned long machine_to_phys_nr_ents;
5271 - char *s;
5272 int i;
5273
5274 setup_xen_features();
5275 @@ -128,10 +109,7 @@ void __init x86_64_start_kernel(char * r
5276 asm volatile("lidt %0" :: "m" (idt_descr));
5277 #endif
5278
5279 - /*
5280 - * This must be called really, really early:
5281 - */
5282 - lockdep_init();
5283 + early_printk("Kernel alive\n");
5284
5285 for (i = 0; i < NR_CPUS; i++)
5286 cpu_pda(i) = &boot_cpu_pda[i];
5287 @@ -141,22 +119,5 @@ void __init x86_64_start_kernel(char * r
5288 #ifdef CONFIG_SMP
5289 cpu_set(0, cpu_online_map);
5290 #endif
5291 - s = strstr(saved_command_line, "earlyprintk=");
5292 - if (s != NULL)
5293 - setup_early_printk(strchr(s, '=') + 1);
5294 -#ifdef CONFIG_NUMA
5295 - s = strstr(saved_command_line, "numa=");
5296 - if (s != NULL)
5297 - numa_setup(s+5);
5298 -#endif
5299 -#ifdef CONFIG_X86_IO_APIC
5300 - if (strstr(saved_command_line, "disableapic"))
5301 - disable_apic = 1;
5302 -#endif
5303 - /* You need early console to see that */
5304 - if (__pa_symbol(&_end) >= KERNEL_TEXT_SIZE)
5305 - panic("Kernel too big for kernel mapping\n");
5306 -
5307 - setup_boot_cpu_data();
5308 start_kernel();
5309 }
5310 --- sle11-2009-05-14.orig/arch/x86/kernel/io_apic_64-xen.c 2009-03-16 16:13:45.000000000 +0100
5311 +++ sle11-2009-05-14/arch/x86/kernel/io_apic_64-xen.c 2009-03-04 11:28:34.000000000 +0100
5312 @@ -26,9 +26,12 @@
5313 #include <linux/delay.h>
5314 #include <linux/sched.h>
5315 #include <linux/smp_lock.h>
5316 +#include <linux/pci.h>
5317 #include <linux/mc146818rtc.h>
5318 #include <linux/acpi.h>
5319 #include <linux/sysdev.h>
5320 +#include <linux/msi.h>
5321 +#include <linux/htirq.h>
5322 #ifdef CONFIG_ACPI
5323 #include <acpi/acpi_bus.h>
5324 #endif
5325 @@ -41,6 +44,10 @@
5326 #include <asm/acpi.h>
5327 #include <asm/dma.h>
5328 #include <asm/nmi.h>
5329 +#include <asm/msidef.h>
5330 +#include <asm/hypertransport.h>
5331 +
5332 +static int assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result);
5333
5334 #define __apicdebuginit __init
5335
5336 @@ -48,17 +55,30 @@ int sis_apic_bug; /* not actually suppor
5337
5338 static int no_timer_check;
5339
5340 -int disable_timer_pin_1 __initdata;
5341 +static int disable_timer_pin_1 __initdata;
5342
5343 -#ifndef CONFIG_XEN
5344 -int timer_over_8254 __initdata = 0;
5345 +#ifdef CONFIG_XEN
5346 +#include <xen/interface/xen.h>
5347 +#include <xen/interface/physdev.h>
5348 +#include <xen/evtchn.h>
5349 +
5350 +/* Fake i8259 */
5351 +#define make_8259A_irq(_irq) (io_apic_irqs &= ~(1UL<<(_irq)))
5352 +#define disable_8259A_irq(_irq) ((void)0)
5353 +#define i8259A_irq_pending(_irq) (0)
5354 +
5355 +unsigned long io_apic_irqs;
5356 +
5357 +#define clear_IO_APIC() ((void)0)
5358 +#else
5359 +int timer_over_8254 __initdata = 1;
5360
5361 /* Where if anywhere is the i8259 connect in external int mode */
5362 static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };
5363 #endif
5364
5365 static DEFINE_SPINLOCK(ioapic_lock);
5366 -static DEFINE_SPINLOCK(vector_lock);
5367 +DEFINE_SPINLOCK(vector_lock);
5368
5369 /*
5370 * # of IRQ routing registers
5371 @@ -83,29 +103,27 @@ static struct irq_pin_list {
5372 short apic, pin, next;
5373 } irq_2_pin[PIN_MAP_SIZE];
5374
5375 -int vector_irq[NR_VECTORS] __read_mostly = { [0 ... NR_VECTORS - 1] = -1};
5376 -#ifdef CONFIG_PCI_MSI
5377 -#define vector_to_irq(vector) \
5378 - (platform_legacy_irq(vector) ? vector : vector_irq[vector])
5379 -#else
5380 -#define vector_to_irq(vector) (vector)
5381 -#endif
5382 -
5383 -#ifdef CONFIG_XEN
5384 -
5385 -#include <xen/interface/xen.h>
5386 -#include <xen/interface/physdev.h>
5387 -#include <xen/evtchn.h>
5388 -
5389 -/* Fake i8259 */
5390 -#define make_8259A_irq(_irq) (io_apic_irqs &= ~(1UL<<(_irq)))
5391 -#define disable_8259A_irq(_irq) ((void)0)
5392 -#define i8259A_irq_pending(_irq) (0)
5393 +#ifndef CONFIG_XEN
5394 +struct io_apic {
5395 + unsigned int index;
5396 + unsigned int unused[3];
5397 + unsigned int data;
5398 +};
5399
5400 -unsigned long io_apic_irqs;
5401 +static __attribute_const__ struct io_apic __iomem *io_apic_base(int idx)
5402 +{
5403 + return (void __iomem *) __fix_to_virt(FIX_IO_APIC_BASE_0 + idx)
5404 + + (mp_ioapics[idx].mpc_apicaddr & ~PAGE_MASK);
5405 +}
5406 +#endif
5407
5408 -static inline unsigned int xen_io_apic_read(unsigned int apic, unsigned int reg)
5409 +static inline unsigned int io_apic_read(unsigned int apic, unsigned int reg)
5410 {
5411 +#ifndef CONFIG_XEN
5412 + struct io_apic __iomem *io_apic = io_apic_base(apic);
5413 + writel(reg, &io_apic->index);
5414 + return readl(&io_apic->data);
5415 +#else
5416 struct physdev_apic apic_op;
5417 int ret;
5418
5419 @@ -115,31 +133,133 @@ static inline unsigned int xen_io_apic_r
5420 if (ret)
5421 return ret;
5422 return apic_op.value;
5423 +#endif
5424 }
5425
5426 -static inline void xen_io_apic_write(unsigned int apic, unsigned int reg, unsigned int value)
5427 +static inline void io_apic_write(unsigned int apic, unsigned int reg, unsigned int value)
5428 {
5429 +#ifndef CONFIG_XEN
5430 + struct io_apic __iomem *io_apic = io_apic_base(apic);
5431 + writel(reg, &io_apic->index);
5432 + writel(value, &io_apic->data);
5433 +#else
5434 struct physdev_apic apic_op;
5435
5436 apic_op.apic_physbase = mp_ioapics[apic].mpc_apicaddr;
5437 apic_op.reg = reg;
5438 apic_op.value = value;
5439 WARN_ON(HYPERVISOR_physdev_op(PHYSDEVOP_apic_write, &apic_op));
5440 +#endif
5441 +}
5442 +
5443 +#ifndef CONFIG_XEN
5444 +/*
5445 + * Re-write a value: to be used for read-modify-write
5446 + * cycles where the read already set up the index register.
5447 + */
5448 +static inline void io_apic_modify(unsigned int apic, unsigned int value)
5449 +{
5450 + struct io_apic __iomem *io_apic = io_apic_base(apic);
5451 + writel(value, &io_apic->data);
5452 }
5453 +#else
5454 +#define io_apic_modify io_apic_write
5455 +#endif
5456
5457 -#define io_apic_read(a,r) xen_io_apic_read(a,r)
5458 -#define io_apic_write(a,r,v) xen_io_apic_write(a,r,v)
5459 +/*
5460 + * Synchronize the IO-APIC and the CPU by doing
5461 + * a dummy read from the IO-APIC
5462 + */
5463 +static inline void io_apic_sync(unsigned int apic)
5464 +{
5465 +#ifndef CONFIG_XEN
5466 + struct io_apic __iomem *io_apic = io_apic_base(apic);
5467 + readl(&io_apic->data);
5468 +#endif
5469 +}
5470
5471 -#define clear_IO_APIC() ((void)0)
5472 +union entry_union {
5473 + struct { u32 w1, w2; };
5474 + struct IO_APIC_route_entry entry;
5475 +};
5476
5477 -#else
5478 +#ifndef CONFIG_XEN
5479 +static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin)
5480 +{
5481 + union entry_union eu;
5482 + unsigned long flags;
5483 + spin_lock_irqsave(&ioapic_lock, flags);
5484 + eu.w1 = io_apic_read(apic, 0x10 + 2 * pin);
5485 + eu.w2 = io_apic_read(apic, 0x11 + 2 * pin);
5486 + spin_unlock_irqrestore(&ioapic_lock, flags);
5487 + return eu.entry;
5488 +}
5489 +#endif
5490 +
5491 +/*
5492 + * When we write a new IO APIC routing entry, we need to write the high
5493 + * word first! If the mask bit in the low word is clear, we will enable
5494 + * the interrupt, and we need to make sure the entry is fully populated
5495 + * before that happens.
5496 + */
5497 +static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
5498 +{
5499 + unsigned long flags;
5500 + union entry_union eu;
5501 + eu.entry = e;
5502 + spin_lock_irqsave(&ioapic_lock, flags);
5503 + io_apic_write(apic, 0x11 + 2*pin, eu.w2);
5504 + io_apic_write(apic, 0x10 + 2*pin, eu.w1);
5505 + spin_unlock_irqrestore(&ioapic_lock, flags);
5506 +}
5507 +
5508 +#ifndef CONFIG_XEN
5509 +/*
5510 + * When we mask an IO APIC routing entry, we need to write the low
5511 + * word first, in order to set the mask bit before we change the
5512 + * high bits!
5513 + */
5514 +static void ioapic_mask_entry(int apic, int pin)
5515 +{
5516 + unsigned long flags;
5517 + union entry_union eu = { .entry.mask = 1 };
5518 +
5519 + spin_lock_irqsave(&ioapic_lock, flags);
5520 + io_apic_write(apic, 0x10 + 2*pin, eu.w1);
5521 + io_apic_write(apic, 0x11 + 2*pin, eu.w2);
5522 + spin_unlock_irqrestore(&ioapic_lock, flags);
5523 +}
5524
5525 #ifdef CONFIG_SMP
5526 +static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, u8 vector)
5527 +{
5528 + int apic, pin;
5529 + struct irq_pin_list *entry = irq_2_pin + irq;
5530 +
5531 + BUG_ON(irq >= NR_IRQS);
5532 + for (;;) {
5533 + unsigned int reg;
5534 + apic = entry->apic;
5535 + pin = entry->pin;
5536 + if (pin == -1)
5537 + break;
5538 + io_apic_write(apic, 0x11 + pin*2, dest);
5539 + reg = io_apic_read(apic, 0x10 + pin*2);
5540 + reg &= ~0x000000ff;
5541 + reg |= vector;
5542 + io_apic_modify(apic, reg);
5543 + if (!entry->next)
5544 + break;
5545 + entry = irq_2_pin + entry->next;
5546 + }
5547 +}
5548 +
5549 static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
5550 {
5551 unsigned long flags;
5552 unsigned int dest;
5553 cpumask_t tmp;
5554 + int vector;
5555
5556 cpus_and(tmp, mask, cpu_online_map);
5557 if (cpus_empty(tmp))
5558 @@ -147,7 +267,11 @@ static void set_ioapic_affinity_irq(unsi
5559
5560 cpus_and(mask, tmp, CPU_MASK_ALL);
5561
5562 - dest = cpu_mask_to_apicid(mask);
5563 + vector = assign_irq_vector(irq, mask, &tmp);
5564 + if (vector < 0)
5565 + return;
5566 +
5567 + dest = cpu_mask_to_apicid(tmp);
5568
5569 /*
5570 * Only the high 8 bits are valid.
5571 @@ -155,13 +279,12 @@ static void set_ioapic_affinity_irq(unsi
5572 dest = SET_APIC_LOGICAL_ID(dest);
5573
5574 spin_lock_irqsave(&ioapic_lock, flags);
5575 - __DO_ACTION(1, = dest, )
5576 - set_irq_info(irq, mask);
5577 + __target_IO_APIC_irq(irq, dest, vector);
5578 + set_native_irq_info(irq, mask);
5579 spin_unlock_irqrestore(&ioapic_lock, flags);
5580 }
5581 #endif
5582 -
5583 -#endif /* !CONFIG_XEN */
5584 +#endif
5585
5586 /*
5587 * The common case is 1:1 IRQ<->pin mappings. Sometimes there are
5588 @@ -241,24 +364,15 @@ static void unmask_IO_APIC_irq (unsigned
5589 static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
5590 {
5591 struct IO_APIC_route_entry entry;
5592 - unsigned long flags;
5593
5594 /* Check delivery_mode to be sure we're not clearing an SMI pin */
5595 - spin_lock_irqsave(&ioapic_lock, flags);
5596 - *(((int*)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
5597 - *(((int*)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
5598 - spin_unlock_irqrestore(&ioapic_lock, flags);
5599 + entry = ioapic_read_entry(apic, pin);
5600 if (entry.delivery_mode == dest_SMI)
5601 return;
5602 /*
5603 * Disable it in the IO-APIC irq-routing table:
5604 */
5605 - memset(&entry, 0, sizeof(entry));
5606 - entry.mask = 1;
5607 - spin_lock_irqsave(&ioapic_lock, flags);
5608 - io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry) + 0));
5609 - io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry) + 1));
5610 - spin_unlock_irqrestore(&ioapic_lock, flags);
5611 + ioapic_mask_entry(apic, pin);
5612 }
5613
5614 static void clear_IO_APIC (void)
5615 @@ -272,16 +386,6 @@ static void clear_IO_APIC (void)
5616
5617 #endif /* !CONFIG_XEN */
5618
5619 -static u8 gsi_2_irq[NR_IRQ_VECTORS] = { [0 ... NR_IRQ_VECTORS-1] = 0xFF };
5620 -
5621 -/*
5622 - * support for broken MP BIOSs, enables hand-redirection of PIRQ0-7 to
5623 - * specific CPU-side IRQs.
5624 - */
5625 -
5626 -#define MAX_PIRQS 8
5627 -static int pirq_entries [MAX_PIRQS];
5628 -static int pirqs_enabled;
5629 int skip_ioapic_setup;
5630 int ioapic_force;
5631
5632 @@ -290,18 +394,17 @@ int ioapic_force;
5633 static int __init disable_ioapic_setup(char *str)
5634 {
5635 skip_ioapic_setup = 1;
5636 - return 1;
5637 + return 0;
5638 }
5639 +early_param("noapic", disable_ioapic_setup);
5640
5641 -static int __init enable_ioapic_setup(char *str)
5642 +/* Actually the next is obsolete, but keep it for paranoid reasons -AK */
5643 +static int __init disable_timer_pin_setup(char *arg)
5644 {
5645 - ioapic_force = 1;
5646 - skip_ioapic_setup = 0;
5647 + disable_timer_pin_1 = 1;
5648 return 1;
5649 }
5650 -
5651 -__setup("noapic", disable_ioapic_setup);
5652 -__setup("apic", enable_ioapic_setup);
5653 +__setup("disable_timer_pin_1", disable_timer_pin_setup);
5654
5655 #ifndef CONFIG_XEN
5656 static int __init setup_disable_8254_timer(char *s)
5657 @@ -319,137 +422,6 @@ __setup("disable_8254_timer", setup_disa
5658 __setup("enable_8254_timer", setup_enable_8254_timer);
5659 #endif /* !CONFIG_XEN */
5660
5661 -#include <asm/pci-direct.h>
5662 -#include <linux/pci_ids.h>
5663 -#include <linux/pci.h>
5664 -
5665 -
5666 -#ifdef CONFIG_ACPI
5667 -
5668 -static int nvidia_hpet_detected __initdata;
5669 -
5670 -static int __init nvidia_hpet_check(unsigned long phys, unsigned long size)
5671 -{
5672 - nvidia_hpet_detected = 1;
5673 - return 0;
5674 -}
5675 -#endif
5676 -
5677 -/* Temporary Hack. Nvidia and VIA boards currently only work with IO-APIC
5678 - off. Check for an Nvidia or VIA PCI bridge and turn it off.
5679 - Use pci direct infrastructure because this runs before the PCI subsystem.
5680 -
5681 - Can be overwritten with "apic"
5682 -
5683 - And another hack to disable the IOMMU on VIA chipsets.
5684 -
5685 - ... and others. Really should move this somewhere else.
5686 -
5687 - Kludge-O-Rama. */
5688 -void __init check_ioapic(void)
5689 -{
5690 - int num,slot,func;
5691 - /* Poor man's PCI discovery */
5692 - for (num = 0; num < 32; num++) {
5693 - for (slot = 0; slot < 32; slot++) {
5694 - for (func = 0; func < 8; func++) {
5695 - u32 class;
5696 - u32 vendor;
5697 - u8 type;
5698 - class = read_pci_config(num,slot,func,
5699 - PCI_CLASS_REVISION);
5700 - if (class == 0xffffffff)
5701 - break;
5702 -
5703 - if ((class >> 16) != PCI_CLASS_BRIDGE_PCI)
5704 - continue;
5705 -
5706 - vendor = read_pci_config(num, slot, func,
5707 - PCI_VENDOR_ID);
5708 - vendor &= 0xffff;
5709 - switch (vendor) {
5710 - case PCI_VENDOR_ID_VIA:
5711 -#ifdef CONFIG_IOMMU
5712 - if ((end_pfn > MAX_DMA32_PFN ||
5713 - force_iommu) &&
5714 - !iommu_aperture_allowed) {
5715 - printk(KERN_INFO
5716 - "Looks like a VIA chipset. Disabling IOMMU. Override with \"iommu=allowed\"\n");
5717 - iommu_aperture_disabled = 1;
5718 - }
5719 -#endif
5720 - return;
5721 - case PCI_VENDOR_ID_NVIDIA:
5722 -#ifdef CONFIG_ACPI
5723 - /*
5724 - * All timer overrides on Nvidia are
5725 - * wrong unless HPET is enabled.
5726 - */
5727 - nvidia_hpet_detected = 0;
5728 - acpi_table_parse(ACPI_HPET,
5729 - nvidia_hpet_check);
5730 - if (nvidia_hpet_detected == 0) {
5731 - acpi_skip_timer_override = 1;
5732 - printk(KERN_INFO "Nvidia board "
5733 - "detected. Ignoring ACPI "
5734 - "timer override.\n");
5735 - }
5736 -#endif
5737 - /* RED-PEN skip them on mptables too? */
5738 - return;
5739 - case PCI_VENDOR_ID_ATI:
5740 -
5741 - /* This should be actually default, but
5742 - for 2.6.16 let's do it for ATI only where
5743 - it's really needed. */
5744 -#ifndef CONFIG_XEN
5745 - if (timer_over_8254 == 1) {
5746 - timer_over_8254 = 0;
5747 - printk(KERN_INFO
5748 - "ATI board detected. Disabling timer routing over 8254.\n");
5749 - }
5750 -#endif
5751 - return;
5752 - }
5753 -
5754 -
5755 - /* No multi-function device? */
5756 - type = read_pci_config_byte(num,slot,func,
5757 - PCI_HEADER_TYPE);
5758 - if (!(type & 0x80))
5759 - break;
5760 - }
5761 - }
5762 - }
5763 -}
5764 -
5765 -static int __init ioapic_pirq_setup(char *str)
5766 -{
5767 - int i, max;
5768 - int ints[MAX_PIRQS+1];
5769 -
5770 - get_options(str, ARRAY_SIZE(ints), ints);
5771 -
5772 - for (i = 0; i < MAX_PIRQS; i++)
5773 - pirq_entries[i] = -1;
5774 -
5775 - pirqs_enabled = 1;
5776 - apic_printk(APIC_VERBOSE, "PIRQ redirection, working around broken MP-BIOS.\n");
5777 - max = MAX_PIRQS;
5778 - if (ints[0] < MAX_PIRQS)
5779 - max = ints[0];
5780 -
5781 - for (i = 0; i < max; i++) {
5782 - apic_printk(APIC_VERBOSE, "... PIRQ%d -> IRQ %d\n", i, ints[i+1]);
5783 - /*
5784 - * PIRQs are mapped upside down, usually.
5785 - */
5786 - pirq_entries[MAX_PIRQS-i-1] = ints[i+1];
5787 - }
5788 - return 1;
5789 -}
5790 -
5791 -__setup("pirq=", ioapic_pirq_setup);
5792
5793 /*
5794 * Find the IRQ entry number of a certain pin.
5795 @@ -479,9 +451,7 @@ static int __init find_isa_irq_pin(int i
5796 for (i = 0; i < mp_irq_entries; i++) {
5797 int lbus = mp_irqs[i].mpc_srcbus;
5798
5799 - if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA ||
5800 - mp_bus_id_to_type[lbus] == MP_BUS_EISA ||
5801 - mp_bus_id_to_type[lbus] == MP_BUS_MCA) &&
5802 + if (test_bit(lbus, mp_bus_not_pci) &&
5803 (mp_irqs[i].mpc_irqtype == type) &&
5804 (mp_irqs[i].mpc_srcbusirq == irq))
5805
5806 @@ -497,9 +467,7 @@ static int __init find_isa_irq_apic(int
5807 for (i = 0; i < mp_irq_entries; i++) {
5808 int lbus = mp_irqs[i].mpc_srcbus;
5809
5810 - if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA ||
5811 - mp_bus_id_to_type[lbus] == MP_BUS_EISA ||
5812 - mp_bus_id_to_type[lbus] == MP_BUS_MCA) &&
5813 + if (test_bit(lbus, mp_bus_not_pci) &&
5814 (mp_irqs[i].mpc_irqtype == type) &&
5815 (mp_irqs[i].mpc_srcbusirq == irq))
5816 break;
5817 @@ -540,7 +508,7 @@ int IO_APIC_get_PCI_irq_vector(int bus,
5818 mp_irqs[i].mpc_dstapic == MP_APIC_ALL)
5819 break;
5820
5821 - if ((mp_bus_id_to_type[lbus] == MP_BUS_PCI) &&
5822 + if (!test_bit(lbus, mp_bus_not_pci) &&
5823 !mp_irqs[i].mpc_irqtype &&
5824 (bus == lbus) &&
5825 (slot == ((mp_irqs[i].mpc_srcbusirq >> 2) & 0x1f))) {
5826 @@ -563,27 +531,6 @@ int IO_APIC_get_PCI_irq_vector(int bus,
5827 return best_guess;
5828 }
5829
5830 -/*
5831 - * EISA Edge/Level control register, ELCR
5832 - */
5833 -static int EISA_ELCR(unsigned int irq)
5834 -{
5835 - if (irq < 16) {
5836 - unsigned int port = 0x4d0 + (irq >> 3);
5837 - return (inb(port) >> (irq & 7)) & 1;
5838 - }
5839 - apic_printk(APIC_VERBOSE, "Broken MPtable reports ISA irq %d\n", irq);
5840 - return 0;
5841 -}
5842 -
5843 -/* EISA interrupts are always polarity zero and can be edge or level
5844 - * trigger depending on the ELCR value. If an interrupt is listed as
5845 - * EISA conforming in the MP table, that means its trigger type must
5846 - * be read in from the ELCR */
5847 -
5848 -#define default_EISA_trigger(idx) (EISA_ELCR(mp_irqs[idx].mpc_srcbusirq))
5849 -#define default_EISA_polarity(idx) (0)
5850 -
5851 /* ISA interrupts are always polarity zero edge triggered,
5852 * when listed as conforming in the MP table. */
5853
5854 @@ -596,12 +543,6 @@ static int EISA_ELCR(unsigned int irq)
5855 #define default_PCI_trigger(idx) (1)
5856 #define default_PCI_polarity(idx) (1)
5857
5858 -/* MCA interrupts are always polarity zero level triggered,
5859 - * when listed as conforming in the MP table. */
5860 -
5861 -#define default_MCA_trigger(idx) (1)
5862 -#define default_MCA_polarity(idx) (0)
5863 -
5864 static int __init MPBIOS_polarity(int idx)
5865 {
5866 int bus = mp_irqs[idx].mpc_srcbus;
5867 @@ -613,38 +554,11 @@ static int __init MPBIOS_polarity(int id
5868 switch (mp_irqs[idx].mpc_irqflag & 3)
5869 {
5870 case 0: /* conforms, ie. bus-type dependent polarity */
5871 - {
5872 - switch (mp_bus_id_to_type[bus])
5873 - {
5874 - case MP_BUS_ISA: /* ISA pin */
5875 - {
5876 - polarity = default_ISA_polarity(idx);
5877 - break;
5878 - }
5879 - case MP_BUS_EISA: /* EISA pin */
5880 - {
5881 - polarity = default_EISA_polarity(idx);
5882 - break;
5883 - }
5884 - case MP_BUS_PCI: /* PCI pin */
5885 - {
5886 - polarity = default_PCI_polarity(idx);
5887 - break;
5888 - }
5889 - case MP_BUS_MCA: /* MCA pin */
5890 - {
5891 - polarity = default_MCA_polarity(idx);
5892 - break;
5893 - }
5894 - default:
5895 - {
5896 - printk(KERN_WARNING "broken BIOS!!\n");
5897 - polarity = 1;
5898 - break;
5899 - }
5900 - }
5901 + if (test_bit(bus, mp_bus_not_pci))
5902 + polarity = default_ISA_polarity(idx);
5903 + else
5904 + polarity = default_PCI_polarity(idx);
5905 break;
5906 - }
5907 case 1: /* high active */
5908 {
5909 polarity = 0;
5910 @@ -682,38 +596,11 @@ static int MPBIOS_trigger(int idx)
5911 switch ((mp_irqs[idx].mpc_irqflag>>2) & 3)
5912 {
5913 case 0: /* conforms, ie. bus-type dependent */
5914 - {
5915 - switch (mp_bus_id_to_type[bus])
5916 - {
5917 - case MP_BUS_ISA: /* ISA pin */
5918 - {
5919 - trigger = default_ISA_trigger(idx);
5920 - break;
5921 - }
5922 - case MP_BUS_EISA: /* EISA pin */
5923 - {
5924 - trigger = default_EISA_trigger(idx);
5925 - break;
5926 - }
5927 - case MP_BUS_PCI: /* PCI pin */
5928 - {
5929 - trigger = default_PCI_trigger(idx);
5930 - break;
5931 - }
5932 - case MP_BUS_MCA: /* MCA pin */
5933 - {
5934 - trigger = default_MCA_trigger(idx);
5935 - break;
5936 - }
5937 - default:
5938 - {
5939 - printk(KERN_WARNING "broken BIOS!!\n");
5940 - trigger = 1;
5941 - break;
5942 - }
5943 - }
5944 + if (test_bit(bus, mp_bus_not_pci))
5945 + trigger = default_ISA_trigger(idx);
5946 + else
5947 + trigger = default_PCI_trigger(idx);
5948 break;
5949 - }
5950 case 1: /* edge */
5951 {
5952 trigger = 0;
5953 @@ -750,64 +637,6 @@ static inline int irq_trigger(int idx)
5954 return MPBIOS_trigger(idx);
5955 }
5956
5957 -static int next_irq = 16;
5958 -
5959 -/*
5960 - * gsi_irq_sharing -- Name overload! "irq" can be either a legacy IRQ
5961 - * in the range 0-15, a linux IRQ in the range 0-223, or a GSI number
5962 - * from ACPI, which can reach 800 in large boxen.
5963 - *
5964 - * Compact the sparse GSI space into a sequential IRQ series and reuse
5965 - * vectors if possible.
5966 - */
5967 -int gsi_irq_sharing(int gsi)
5968 -{
5969 - int i, tries, vector;
5970 -
5971 - BUG_ON(gsi >= NR_IRQ_VECTORS);
5972 -
5973 - if (platform_legacy_irq(gsi))
5974 - return gsi;
5975 -
5976 - if (gsi_2_irq[gsi] != 0xFF)
5977 - return (int)gsi_2_irq[gsi];
5978 -
5979 - tries = NR_IRQS;
5980 - try_again:
5981 - vector = assign_irq_vector(gsi);
5982 -
5983 - /*
5984 - * Sharing vectors means sharing IRQs, so scan irq_vectors for previous
5985 - * use of vector and if found, return that IRQ. However, we never want
5986 - * to share legacy IRQs, which usually have a different trigger mode
5987 - * than PCI.
5988 - */
5989 - for (i = 0; i < NR_IRQS; i++)
5990 - if (IO_APIC_VECTOR(i) == vector)
5991 - break;
5992 - if (platform_legacy_irq(i)) {
5993 - if (--tries >= 0) {
5994 - IO_APIC_VECTOR(i) = 0;
5995 - goto try_again;
5996 - }
5997 - panic("gsi_irq_sharing: didn't find an IRQ using vector 0x%02X for GSI %d", vector, gsi);
5998 - }
5999 - if (i < NR_IRQS) {
6000 - gsi_2_irq[gsi] = i;
6001 - printk(KERN_INFO "GSI %d sharing vector 0x%02X and IRQ %d\n",
6002 - gsi, vector, i);
6003 - return i;
6004 - }
6005 -
6006 - i = next_irq++;
6007 - BUG_ON(i >= NR_IRQS);
6008 - gsi_2_irq[gsi] = i;
6009 - IO_APIC_VECTOR(i) = vector;
6010 - printk(KERN_INFO "GSI %d assigned vector 0x%02X and IRQ %d\n",
6011 - gsi, vector, i);
6012 - return i;
6013 -}
6014 -
6015 static int pin_2_irq(int idx, int apic, int pin)
6016 {
6017 int irq, i;
6018 @@ -819,49 +648,16 @@ static int pin_2_irq(int idx, int apic,
6019 if (mp_irqs[idx].mpc_dstirq != pin)
6020 printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n");
6021
6022 - switch (mp_bus_id_to_type[bus])
6023 - {
6024 - case MP_BUS_ISA: /* ISA pin */
6025 - case MP_BUS_EISA:
6026 - case MP_BUS_MCA:
6027 - {
6028 - irq = mp_irqs[idx].mpc_srcbusirq;
6029 - break;
6030 - }
6031 - case MP_BUS_PCI: /* PCI pin */
6032 - {
6033 - /*
6034 - * PCI IRQs are mapped in order
6035 - */
6036 - i = irq = 0;
6037 - while (i < apic)
6038 - irq += nr_ioapic_registers[i++];
6039 - irq += pin;
6040 - irq = gsi_irq_sharing(irq);
6041 - break;
6042 - }
6043 - default:
6044 - {
6045 - printk(KERN_ERR "unknown bus type %d.\n",bus);
6046 - irq = 0;
6047 - break;
6048 - }
6049 - }
6050 - BUG_ON(irq >= NR_IRQS);
6051 -
6052 - /*
6053 - * PCI IRQ command line redirection. Yes, limits are hardcoded.
6054 - */
6055 - if ((pin >= 16) && (pin <= 23)) {
6056 - if (pirq_entries[pin-16] != -1) {
6057 - if (!pirq_entries[pin-16]) {
6058 - apic_printk(APIC_VERBOSE, "disabling PIRQ%d\n", pin-16);
6059 - } else {
6060 - irq = pirq_entries[pin-16];
6061 - apic_printk(APIC_VERBOSE, "using PIRQ%d -> IRQ %d\n",
6062 - pin-16, irq);
6063 - }
6064 - }
6065 + if (test_bit(bus, mp_bus_not_pci)) {
6066 + irq = mp_irqs[idx].mpc_srcbusirq;
6067 + } else {
6068 + /*
6069 + * PCI IRQs are mapped in order
6070 + */
6071 + i = irq = 0;
6072 + while (i < apic)
6073 + irq += nr_ioapic_registers[i++];
6074 + irq += pin;
6075 }
6076 BUG_ON(irq >= NR_IRQS);
6077 return irq;
6078 @@ -885,46 +681,71 @@ static inline int IO_APIC_irq_trigger(in
6079 }
6080
6081 /* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */
6082 -u8 irq_vector[NR_IRQ_VECTORS] __read_mostly;
6083 +static u8 irq_vector[NR_IRQ_VECTORS] __read_mostly;
6084
6085 -int assign_irq_vector(int irq)
6086 +static int __assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result)
6087 {
6088 - unsigned long flags;
6089 int vector;
6090 struct physdev_irq irq_op;
6091
6092 - BUG_ON(irq != AUTO_ASSIGN && (unsigned)irq >= NR_IRQ_VECTORS);
6093 + BUG_ON((unsigned)irq >= NR_IRQ_VECTORS);
6094
6095 if (irq < PIRQ_BASE || irq - PIRQ_BASE >= NR_PIRQS)
6096 return -EINVAL;
6097
6098 - spin_lock_irqsave(&vector_lock, flags);
6099 + cpus_and(*result, mask, cpu_online_map);
6100
6101 - if (irq != AUTO_ASSIGN && IO_APIC_VECTOR(irq) > 0) {
6102 - spin_unlock_irqrestore(&vector_lock, flags);
6103 - return IO_APIC_VECTOR(irq);
6104 - }
6105 + if (irq_vector[irq] > 0)
6106 + return irq_vector[irq];
6107
6108 irq_op.irq = irq;
6109 - if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) {
6110 - spin_unlock_irqrestore(&vector_lock, flags);
6111 + if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op))
6112 return -ENOSPC;
6113 - }
6114
6115 vector = irq_op.vector;
6116 - vector_irq[vector] = irq;
6117 - if (irq != AUTO_ASSIGN)
6118 - IO_APIC_VECTOR(irq) = vector;
6119 + irq_vector[irq] = vector;
6120
6121 - spin_unlock_irqrestore(&vector_lock, flags);
6122 + return vector;
6123 +}
6124 +
6125 +static int assign_irq_vector(int irq, cpumask_t mask, cpumask_t *result)
6126 +{
6127 + int vector;
6128 + unsigned long flags;
6129
6130 + spin_lock_irqsave(&vector_lock, flags);
6131 + vector = __assign_irq_vector(irq, mask, result);
6132 + spin_unlock_irqrestore(&vector_lock, flags);
6133 return vector;
6134 }
6135
6136 -extern void (*interrupt[NR_IRQS])(void);
6137 #ifndef CONFIG_XEN
6138 -static struct hw_interrupt_type ioapic_level_type;
6139 -static struct hw_interrupt_type ioapic_edge_type;
6140 +void __setup_vector_irq(int cpu)
6141 +{
6142 + /* Initialize vector_irq on a new cpu */
6143 + /* This function must be called with vector_lock held */
6144 + int irq, vector;
6145 +
6146 + /* Mark the inuse vectors */
6147 + for (irq = 0; irq < NR_IRQ_VECTORS; ++irq) {
6148 + if (!cpu_isset(cpu, irq_domain[irq]))
6149 + continue;
6150 + vector = irq_vector[irq];
6151 + per_cpu(vector_irq, cpu)[vector] = irq;
6152 + }
6153 + /* Mark the free vectors */
6154 + for (vector = 0; vector < NR_VECTORS; ++vector) {
6155 + irq = per_cpu(vector_irq, cpu)[vector];
6156 + if (irq < 0)
6157 + continue;
6158 + if (!cpu_isset(cpu, irq_domain[irq]))
6159 + per_cpu(vector_irq, cpu)[vector] = -1;
6160 + }
6161 +}
6162 +
6163 +extern void (*interrupt[NR_IRQS])(void);
6164 +
6165 +static struct irq_chip ioapic_chip;
6166
6167 #define IOAPIC_AUTO -1
6168 #define IOAPIC_EDGE 0
6169 @@ -932,16 +753,15 @@ static struct hw_interrupt_type ioapic_e
6170
6171 static void ioapic_register_intr(int irq, int vector, unsigned long trigger)
6172 {
6173 - unsigned idx;
6174 -
6175 - idx = use_pci_vector() && !platform_legacy_irq(irq) ? vector : irq;
6176 -
6177 if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
6178 trigger == IOAPIC_LEVEL)
6179 - irq_desc[idx].chip = &ioapic_level_type;
6180 - else
6181 - irq_desc[idx].chip = &ioapic_edge_type;
6182 - set_intr_gate(vector, interrupt[idx]);
6183 + set_irq_chip_and_handler_name(irq, &ioapic_chip,
6184 + handle_fasteoi_irq, "fasteoi");
6185 + else {
6186 + irq_desc[irq].status |= IRQ_DELAYED_DISABLE;
6187 + set_irq_chip_and_handler_name(irq, &ioapic_chip,
6188 + handle_edge_irq, "edge");
6189 + }
6190 }
6191 #else
6192 #define ioapic_register_intr(irq, vector, trigger) evtchn_register_pirq(irq)
6193 @@ -994,16 +814,21 @@ static void __init setup_IO_APIC_irqs(vo
6194 continue;
6195
6196 if (IO_APIC_IRQ(irq)) {
6197 - vector = assign_irq_vector(irq);
6198 + cpumask_t mask;
6199 + vector = assign_irq_vector(irq, TARGET_CPUS, &mask);
6200 + if (vector < 0)
6201 + continue;
6202 +
6203 + entry.dest.logical.logical_dest = cpu_mask_to_apicid(mask);
6204 entry.vector = vector;
6205
6206 ioapic_register_intr(irq, vector, IOAPIC_AUTO);
6207 if (!apic && (irq < 16))
6208 disable_8259A_irq(irq);
6209 }
6210 + ioapic_write_entry(apic, pin, entry);
6211 +
6212 spin_lock_irqsave(&ioapic_lock, flags);
6213 - io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1));
6214 - io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0));
6215 set_native_irq_info(irq, TARGET_CPUS);
6216 spin_unlock_irqrestore(&ioapic_lock, flags);
6217 }
6218 @@ -1046,7 +871,7 @@ static void __init setup_ExtINT_IRQ0_pin
6219 * The timer IRQ doesn't have to know that behind the
6220 * scene we have a 8259A-master in AEOI mode ...
6221 */
6222 - irq_desc[0].chip = &ioapic_edge_type;
6223 + set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge");
6224
6225 /*
6226 * Add it to the IO-APIC irq-routing table:
6227 @@ -1142,10 +967,7 @@ void __apicdebuginit print_IO_APIC(void)
6228 for (i = 0; i <= reg_01.bits.entries; i++) {
6229 struct IO_APIC_route_entry entry;
6230
6231 - spin_lock_irqsave(&ioapic_lock, flags);
6232 - *(((int *)&entry)+0) = io_apic_read(apic, 0x10+i*2);
6233 - *(((int *)&entry)+1) = io_apic_read(apic, 0x11+i*2);
6234 - spin_unlock_irqrestore(&ioapic_lock, flags);
6235 + entry = ioapic_read_entry(apic, i);
6236
6237 printk(KERN_DEBUG " %02x %03X %02X ",
6238 i,
6239 @@ -1165,17 +987,12 @@ void __apicdebuginit print_IO_APIC(void)
6240 );
6241 }
6242 }
6243 - if (use_pci_vector())
6244 - printk(KERN_INFO "Using vector-based indexing\n");
6245 printk(KERN_DEBUG "IRQ to pin mappings:\n");
6246 for (i = 0; i < NR_IRQS; i++) {
6247 struct irq_pin_list *entry = irq_2_pin + i;
6248 if (entry->pin < 0)
6249 continue;
6250 - if (use_pci_vector() && !platform_legacy_irq(i))
6251 - printk(KERN_DEBUG "IRQ%d ", IO_APIC_VECTOR(i));
6252 - else
6253 - printk(KERN_DEBUG "IRQ%d ", i);
6254 + printk(KERN_DEBUG "IRQ%d ", i);
6255 for (;;) {
6256 printk("-> %d:%d", entry->apic, entry->pin);
6257 if (!entry->next)
6258 @@ -1339,9 +1156,6 @@ static void __init enable_IO_APIC(void)
6259 irq_2_pin[i].pin = -1;
6260 irq_2_pin[i].next = 0;
6261 }
6262 - if (!pirqs_enabled)
6263 - for (i = 0; i < MAX_PIRQS; i++)
6264 - pirq_entries[i] = -1;
6265
6266 /*
6267 * The number of IO-APIC IRQ registers (== #pins):
6268 @@ -1358,11 +1172,7 @@ static void __init enable_IO_APIC(void)
6269 /* See if any of the pins is in ExtINT mode */
6270 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
6271 struct IO_APIC_route_entry entry;
6272 - spin_lock_irqsave(&ioapic_lock, flags);
6273 - *(((int *)&entry) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
6274 - *(((int *)&entry) + 1) = io_apic_read(apic, 0x11 + 2 * pin);
6275 - spin_unlock_irqrestore(&ioapic_lock, flags);
6276 -
6277 + entry = ioapic_read_entry(apic, pin);
6278
6279 /* If the interrupt line is enabled and in ExtInt mode
6280 * I have found the pin where the i8259 is connected.
6281 @@ -1416,7 +1226,6 @@ void disable_IO_APIC(void)
6282 */
6283 if (ioapic_i8259.pin != -1) {
6284 struct IO_APIC_route_entry entry;
6285 - unsigned long flags;
6286
6287 memset(&entry, 0, sizeof(entry));
6288 entry.mask = 0; /* Enabled */
6289 @@ -1433,12 +1242,7 @@ void disable_IO_APIC(void)
6290 /*
6291 * Add it to the IO-APIC irq-routing table:
6292 */
6293 - spin_lock_irqsave(&ioapic_lock, flags);
6294 - io_apic_write(ioapic_i8259.apic, 0x11+2*ioapic_i8259.pin,
6295 - *(((int *)&entry)+1));
6296 - io_apic_write(ioapic_i8259.apic, 0x10+2*ioapic_i8259.pin,
6297 - *(((int *)&entry)+0));
6298 - spin_unlock_irqrestore(&ioapic_lock, flags);
6299 + ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry);
6300 }
6301
6302 disconnect_bsp_APIC(ioapic_i8259.pin != -1);
6303 @@ -1446,76 +1250,6 @@ void disable_IO_APIC(void)
6304 }
6305
6306 /*
6307 - * function to set the IO-APIC physical IDs based on the
6308 - * values stored in the MPC table.
6309 - *
6310 - * by Matt Domsch <Matt_Domsch@dell.com> Tue Dec 21 12:25:05 CST 1999
6311 - */
6312 -
6313 -#ifndef CONFIG_XEN
6314 -static void __init setup_ioapic_ids_from_mpc (void)
6315 -{
6316 - union IO_APIC_reg_00 reg_00;
6317 - int apic;
6318 - int i;
6319 - unsigned char old_id;
6320 - unsigned long flags;
6321 -
6322 - /*
6323 - * Set the IOAPIC ID to the value stored in the MPC table.
6324 - */
6325 - for (apic = 0; apic < nr_ioapics; apic++) {
6326 -
6327 - /* Read the register 0 value */
6328 - spin_lock_irqsave(&ioapic_lock, flags);
6329 - reg_00.raw = io_apic_read(apic, 0);
6330 - spin_unlock_irqrestore(&ioapic_lock, flags);
6331 -
6332 - old_id = mp_ioapics[apic].mpc_apicid;
6333 -
6334 -
6335 - printk(KERN_INFO "Using IO-APIC %d\n", mp_ioapics[apic].mpc_apicid);
6336 -
6337 -
6338 - /*
6339 - * We need to adjust the IRQ routing table
6340 - * if the ID changed.
6341 - */
6342 - if (old_id != mp_ioapics[apic].mpc_apicid)
6343 - for (i = 0; i < mp_irq_entries; i++)
6344 - if (mp_irqs[i].mpc_dstapic == old_id)
6345 - mp_irqs[i].mpc_dstapic
6346 - = mp_ioapics[apic].mpc_apicid;
6347 -
6348 - /*
6349 - * Read the right value from the MPC table and
6350 - * write it into the ID register.
6351 - */
6352 - apic_printk(APIC_VERBOSE,KERN_INFO "...changing IO-APIC physical APIC ID to %d ...",
6353 - mp_ioapics[apic].mpc_apicid);
6354 -
6355 - reg_00.bits.ID = mp_ioapics[apic].mpc_apicid;
6356 - spin_lock_irqsave(&ioapic_lock, flags);
6357 - io_apic_write(apic, 0, reg_00.raw);
6358 - spin_unlock_irqrestore(&ioapic_lock, flags);
6359 -
6360 - /*
6361 - * Sanity check
6362 - */
6363 - spin_lock_irqsave(&ioapic_lock, flags);
6364 - reg_00.raw = io_apic_read(apic, 0);
6365 - spin_unlock_irqrestore(&ioapic_lock, flags);
6366 - if (reg_00.bits.ID != mp_ioapics[apic].mpc_apicid)
6367 - printk("could not set ID!\n");
6368 - else
6369 - apic_printk(APIC_VERBOSE," ok.\n");
6370 - }
6371 -}
6372 -#else
6373 -static void __init setup_ioapic_ids_from_mpc(void) { }
6374 -#endif
6375 -
6376 -/*
6377 * There is a nasty bug in some older SMP boards, their mptable lies
6378 * about the timer IRQ. We do the following to work around the situation:
6379 *
6380 @@ -1569,7 +1303,7 @@ static int __init timer_irq_works(void)
6381 * an edge even if it isn't on the 8259A...
6382 */
6383
6384 -static unsigned int startup_edge_ioapic_irq(unsigned int irq)
6385 +static unsigned int startup_ioapic_irq(unsigned int irq)
6386 {
6387 int was_pending = 0;
6388 unsigned long flags;
6389 @@ -1586,107 +1320,19 @@ static unsigned int startup_edge_ioapic_
6390 return was_pending;
6391 }
6392
6393 -/*
6394 - * Once we have recorded IRQ_PENDING already, we can mask the
6395 - * interrupt for real. This prevents IRQ storms from unhandled
6396 - * devices.
6397 - */
6398 -static void ack_edge_ioapic_irq(unsigned int irq)
6399 -{
6400 - move_irq(irq);
6401 - if ((irq_desc[irq].status & (IRQ_PENDING | IRQ_DISABLED))
6402 - == (IRQ_PENDING | IRQ_DISABLED))
6403 - mask_IO_APIC_irq(irq);
6404 - ack_APIC_irq();
6405 -}
6406 -
6407 -/*
6408 - * Level triggered interrupts can just be masked,
6409 - * and shutting down and starting up the interrupt
6410 - * is the same as enabling and disabling them -- except
6411 - * with a startup need to return a "was pending" value.
6412 - *
6413 - * Level triggered interrupts are special because we
6414 - * do not touch any IO-APIC register while handling
6415 - * them. We ack the APIC in the end-IRQ handler, not
6416 - * in the start-IRQ-handler. Protection against reentrance
6417 - * from the same interrupt is still provided, both by the
6418 - * generic IRQ layer and by the fact that an unacked local
6419 - * APIC does not accept IRQs.
6420 - */
6421 -static unsigned int startup_level_ioapic_irq (unsigned int irq)
6422 -{
6423 - unmask_IO_APIC_irq(irq);
6424 -
6425 - return 0; /* don't check for pending */
6426 -}
6427 -
6428 -static void end_level_ioapic_irq (unsigned int irq)
6429 -{
6430 - move_irq(irq);
6431 - ack_APIC_irq();
6432 -}
6433 -
6434 -#ifdef CONFIG_PCI_MSI
6435 -static unsigned int startup_edge_ioapic_vector(unsigned int vector)
6436 -{
6437 - int irq = vector_to_irq(vector);
6438 -
6439 - return startup_edge_ioapic_irq(irq);
6440 -}
6441 -
6442 -static void ack_edge_ioapic_vector(unsigned int vector)
6443 -{
6444 - int irq = vector_to_irq(vector);
6445 -
6446 - move_native_irq(vector);
6447 - ack_edge_ioapic_irq(irq);
6448 -}
6449 -
6450 -static unsigned int startup_level_ioapic_vector (unsigned int vector)
6451 -{
6452 - int irq = vector_to_irq(vector);
6453 -
6454 - return startup_level_ioapic_irq (irq);
6455 -}
6456 -
6457 -static void end_level_ioapic_vector (unsigned int vector)
6458 -{
6459 - int irq = vector_to_irq(vector);
6460 -
6461 - move_native_irq(vector);
6462 - end_level_ioapic_irq(irq);
6463 -}
6464 -
6465 -static void mask_IO_APIC_vector (unsigned int vector)
6466 -{
6467 - int irq = vector_to_irq(vector);
6468 -
6469 - mask_IO_APIC_irq(irq);
6470 -}
6471 -
6472 -static void unmask_IO_APIC_vector (unsigned int vector)
6473 +static int ioapic_retrigger_irq(unsigned int irq)
6474 {
6475 - int irq = vector_to_irq(vector);
6476 -
6477 - unmask_IO_APIC_irq(irq);
6478 -}
6479 -
6480 -#ifdef CONFIG_SMP
6481 -static void set_ioapic_affinity_vector (unsigned int vector,
6482 - cpumask_t cpu_mask)
6483 -{
6484 - int irq = vector_to_irq(vector);
6485 + cpumask_t mask;
6486 + unsigned vector;
6487 + unsigned long flags;
6488
6489 - set_native_irq_info(vector, cpu_mask);
6490 - set_ioapic_affinity_irq(irq, cpu_mask);
6491 -}
6492 -#endif // CONFIG_SMP
6493 -#endif // CONFIG_PCI_MSI
6494 + spin_lock_irqsave(&vector_lock, flags);
6495 + vector = irq_vector[irq];
6496 + cpus_clear(mask);
6497 + cpu_set(first_cpu(irq_domain[irq]), mask);
6498
6499 -static int ioapic_retrigger(unsigned int irq)
6500 -{
6501 - send_IPI_self(IO_APIC_VECTOR(irq));
6502 + send_IPI_mask(mask, vector);
6503 + spin_unlock_irqrestore(&vector_lock, flags);
6504
6505 return 1;
6506 }
6507 @@ -1700,32 +1346,47 @@ static int ioapic_retrigger(unsigned int
6508 * races.
6509 */
6510
6511 -static struct hw_interrupt_type ioapic_edge_type __read_mostly = {
6512 - .typename = "IO-APIC-edge",
6513 - .startup = startup_edge_ioapic,
6514 - .shutdown = shutdown_edge_ioapic,
6515 - .enable = enable_edge_ioapic,
6516 - .disable = disable_edge_ioapic,
6517 - .ack = ack_edge_ioapic,
6518 - .end = end_edge_ioapic,
6519 -#ifdef CONFIG_SMP
6520 - .set_affinity = set_ioapic_affinity,
6521 +static void ack_apic_edge(unsigned int irq)
6522 +{
6523 + move_native_irq(irq);
6524 + ack_APIC_irq();
6525 +}
6526 +
6527 +static void ack_apic_level(unsigned int irq)
6528 +{
6529 + int do_unmask_irq = 0;
6530 +
6531 +#if defined(CONFIG_GENERIC_PENDING_IRQ) || defined(CONFIG_IRQBALANCE)
6532 + /* If we are moving the irq we need to mask it */
6533 + if (unlikely(irq_desc[irq].status & IRQ_MOVE_PENDING)) {
6534 + do_unmask_irq = 1;
6535 + mask_IO_APIC_irq(irq);
6536 + }
6537 #endif
6538 - .retrigger = ioapic_retrigger,
6539 -};
6540
6541 -static struct hw_interrupt_type ioapic_level_type __read_mostly = {
6542 - .typename = "IO-APIC-level",
6543 - .startup = startup_level_ioapic,
6544 - .shutdown = shutdown_level_ioapic,
6545 - .enable = enable_level_ioapic,
6546 - .disable = disable_level_ioapic,
6547 - .ack = mask_and_ack_level_ioapic,
6548 - .end = end_level_ioapic,
6549 + /*
6550 + * We must acknowledge the irq before we move it or the acknowledge will
6551 + * not propogate properly.
6552 + */
6553 + ack_APIC_irq();
6554 +
6555 + /* Now we can move and renable the irq */
6556 + move_masked_irq(irq);
6557 + if (unlikely(do_unmask_irq))
6558 + unmask_IO_APIC_irq(irq);
6559 +}
6560 +
6561 +static struct irq_chip ioapic_chip __read_mostly = {
6562 + .name = "IO-APIC",
6563 + .startup = startup_ioapic_irq,
6564 + .mask = mask_IO_APIC_irq,
6565 + .unmask = unmask_IO_APIC_irq,
6566 + .ack = ack_apic_edge,
6567 + .eoi = ack_apic_level,
6568 #ifdef CONFIG_SMP
6569 - .set_affinity = set_ioapic_affinity,
6570 + .set_affinity = set_ioapic_affinity_irq,
6571 #endif
6572 - .retrigger = ioapic_retrigger,
6573 + .retrigger = ioapic_retrigger_irq,
6574 };
6575 #endif /* !CONFIG_XEN */
6576
6577 @@ -1746,12 +1407,7 @@ static inline void init_IO_APIC_traps(vo
6578 */
6579 for (irq = 0; irq < NR_IRQS ; irq++) {
6580 int tmp = irq;
6581 - if (use_pci_vector()) {
6582 - if (!platform_legacy_irq(tmp))
6583 - if ((tmp = vector_to_irq(tmp)) == -1)
6584 - continue;
6585 - }
6586 - if (IO_APIC_IRQ(tmp) && !IO_APIC_VECTOR(tmp)) {
6587 + if (IO_APIC_IRQ(tmp) && !irq_vector[tmp]) {
6588 /*
6589 * Hmm.. We don't have an entry for this,
6590 * so default to an old-fashioned 8259
6591 @@ -1762,7 +1418,7 @@ static inline void init_IO_APIC_traps(vo
6592 #ifndef CONFIG_XEN
6593 else
6594 /* Strange. Oh, well.. */
6595 - irq_desc[irq].chip = &no_irq_type;
6596 + irq_desc[irq].chip = &no_irq_chip;
6597 #endif
6598 }
6599 }
6600 @@ -1883,8 +1539,6 @@ static inline void unlock_ExtINT_logic(v
6601 spin_unlock_irqrestore(&ioapic_lock, flags);
6602 }
6603
6604 -int timer_uses_ioapic_pin_0;
6605 -
6606 /*
6607 * This code may look a bit paranoid, but it's supposed to cooperate with
6608 * a wide range of boards and BIOS bugs. Fortunately only the timer IRQ
6609 @@ -1897,13 +1551,13 @@ static inline void check_timer(void)
6610 {
6611 int apic1, pin1, apic2, pin2;
6612 int vector;
6613 + cpumask_t mask;
6614
6615 /*
6616 * get/set the timer IRQ vector:
6617 */
6618 disable_8259A_irq(0);
6619 - vector = assign_irq_vector(0);
6620 - set_intr_gate(vector, interrupt[0]);
6621 + vector = assign_irq_vector(0, TARGET_CPUS, &mask);
6622
6623 /*
6624 * Subtle, code in do_timer_interrupt() expects an AEOI
6625 @@ -1922,9 +1576,6 @@ static inline void check_timer(void)
6626 pin2 = ioapic_i8259.pin;
6627 apic2 = ioapic_i8259.apic;
6628
6629 - if (pin1 == 0)
6630 - timer_uses_ioapic_pin_0 = 1;
6631 -
6632 apic_printk(APIC_VERBOSE,KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n",
6633 vector, apic1, pin1, apic2, pin2);
6634
6635 @@ -2039,11 +1690,6 @@ void __init setup_IO_APIC(void)
6636
6637 apic_printk(APIC_VERBOSE, "ENABLING IO-APIC IRQs\n");
6638
6639 - /*
6640 - * Set up the IO-APIC IRQ routing table.
6641 - */
6642 - if (!acpi_ioapic)
6643 - setup_ioapic_ids_from_mpc();
6644 #ifndef CONFIG_XEN
6645 sync_Arb_IDs();
6646 #endif /* !CONFIG_XEN */
6647 @@ -2066,17 +1712,12 @@ static int ioapic_suspend(struct sys_dev
6648 {
6649 struct IO_APIC_route_entry *entry;
6650 struct sysfs_ioapic_data *data;
6651 - unsigned long flags;
6652 int i;
6653
6654 data = container_of(dev, struct sysfs_ioapic_data, dev);
6655 entry = data->entry;
6656 - spin_lock_irqsave(&ioapic_lock, flags);
6657 - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) {
6658 - *(((int *)entry) + 1) = io_apic_read(dev->id, 0x11 + 2 * i);
6659 - *(((int *)entry) + 0) = io_apic_read(dev->id, 0x10 + 2 * i);
6660 - }
6661 - spin_unlock_irqrestore(&ioapic_lock, flags);
6662 + for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ )
6663 + *entry = ioapic_read_entry(dev->id, i);
6664
6665 return 0;
6666 }
6667 @@ -2098,11 +1739,9 @@ static int ioapic_resume(struct sys_devi
6668 reg_00.bits.ID = mp_ioapics[dev->id].mpc_apicid;
6669 io_apic_write(dev->id, 0, reg_00.raw);
6670 }
6671 - for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) {
6672 - io_apic_write(dev->id, 0x11+2*i, *(((int *)entry)+1));
6673 - io_apic_write(dev->id, 0x10+2*i, *(((int *)entry)+0));
6674 - }
6675 spin_unlock_irqrestore(&ioapic_lock, flags);
6676 + for (i = 0; i < nr_ioapic_registers[dev->id]; i++)
6677 + ioapic_write_entry(dev->id, i, entry[i]);
6678
6679 return 0;
6680 }
6681 @@ -2148,28 +1787,254 @@ static int __init ioapic_init_sysfs(void
6682
6683 device_initcall(ioapic_init_sysfs);
6684
6685 +/*
6686 + * Dynamic irq allocate and deallocation
6687 + */
6688 +int create_irq(void)
6689 +{
6690 + /* Allocate an unused irq */
6691 + int irq;
6692 + int new;
6693 + int vector = 0;
6694 + unsigned long flags;
6695 + cpumask_t mask;
6696 +
6697 + irq = -ENOSPC;
6698 + spin_lock_irqsave(&vector_lock, flags);
6699 + for (new = (NR_IRQS - 1); new >= 0; new--) {
6700 + if (platform_legacy_irq(new))
6701 + continue;
6702 + if (irq_vector[new] != 0)
6703 + continue;
6704 + vector = __assign_irq_vector(new, TARGET_CPUS, &mask);
6705 + if (likely(vector > 0))
6706 + irq = new;
6707 + break;
6708 + }
6709 + spin_unlock_irqrestore(&vector_lock, flags);
6710 +
6711 + if (irq >= 0) {
6712 + dynamic_irq_init(irq);
6713 + }
6714 + return irq;
6715 +}
6716 +
6717 +void destroy_irq(unsigned int irq)
6718 +{
6719 + unsigned long flags;
6720 +
6721 + dynamic_irq_cleanup(irq);
6722 +
6723 + spin_lock_irqsave(&vector_lock, flags);
6724 + irq_vector[irq] = 0;
6725 + spin_unlock_irqrestore(&vector_lock, flags);
6726 +}
6727 +
6728 #endif /* CONFIG_XEN */
6729
6730 -/* --------------------------------------------------------------------------
6731 - ACPI-based IOAPIC Configuration
6732 - -------------------------------------------------------------------------- */
6733 +/*
6734 + * MSI mesage composition
6735 + */
6736 +#if defined(CONFIG_PCI_MSI) && !defined(CONFIG_XEN)
6737 +static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_msg *msg)
6738 +{
6739 + int vector;
6740 + unsigned dest;
6741 + cpumask_t tmp;
6742
6743 -#ifdef CONFIG_ACPI
6744 + vector = assign_irq_vector(irq, TARGET_CPUS, &tmp);
6745 + if (vector >= 0) {
6746 + dest = cpu_mask_to_apicid(tmp);
6747 +
6748 + msg->address_hi = MSI_ADDR_BASE_HI;
6749 + msg->address_lo =
6750 + MSI_ADDR_BASE_LO |
6751 + ((INT_DEST_MODE == 0) ?
6752 + MSI_ADDR_DEST_MODE_PHYSICAL:
6753 + MSI_ADDR_DEST_MODE_LOGICAL) |
6754 + ((INT_DELIVERY_MODE != dest_LowestPrio) ?
6755 + MSI_ADDR_REDIRECTION_CPU:
6756 + MSI_ADDR_REDIRECTION_LOWPRI) |
6757 + MSI_ADDR_DEST_ID(dest);
6758 +
6759 + msg->data =
6760 + MSI_DATA_TRIGGER_EDGE |
6761 + MSI_DATA_LEVEL_ASSERT |
6762 + ((INT_DELIVERY_MODE != dest_LowestPrio) ?
6763 + MSI_DATA_DELIVERY_FIXED:
6764 + MSI_DATA_DELIVERY_LOWPRI) |
6765 + MSI_DATA_VECTOR(vector);
6766 + }
6767 + return vector;
6768 +}
6769
6770 -#define IO_APIC_MAX_ID 0xFE
6771 +#ifdef CONFIG_SMP
6772 +static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
6773 +{
6774 + struct msi_msg msg;
6775 + unsigned int dest;
6776 + cpumask_t tmp;
6777 + int vector;
6778 +
6779 + cpus_and(tmp, mask, cpu_online_map);
6780 + if (cpus_empty(tmp))
6781 + tmp = TARGET_CPUS;
6782 +
6783 + cpus_and(mask, tmp, CPU_MASK_ALL);
6784 +
6785 + vector = assign_irq_vector(irq, mask, &tmp);
6786 + if (vector < 0)
6787 + return;
6788 +
6789 + dest = cpu_mask_to_apicid(tmp);
6790 +
6791 + read_msi_msg(irq, &msg);
6792 +
6793 + msg.data &= ~MSI_DATA_VECTOR_MASK;
6794 + msg.data |= MSI_DATA_VECTOR(vector);
6795 + msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
6796 + msg.address_lo |= MSI_ADDR_DEST_ID(dest);
6797
6798 -int __init io_apic_get_version (int ioapic)
6799 + write_msi_msg(irq, &msg);
6800 + set_native_irq_info(irq, mask);
6801 +}
6802 +#endif /* CONFIG_SMP */
6803 +
6804 +/*
6805 + * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices,
6806 + * which implement the MSI or MSI-X Capability Structure.
6807 + */
6808 +static struct irq_chip msi_chip = {
6809 + .name = "PCI-MSI",
6810 + .unmask = unmask_msi_irq,
6811 + .mask = mask_msi_irq,
6812 + .ack = ack_apic_edge,
6813 +#ifdef CONFIG_SMP
6814 + .set_affinity = set_msi_irq_affinity,
6815 +#endif
6816 + .retrigger = ioapic_retrigger_irq,
6817 +};
6818 +
6819 +int arch_setup_msi_irq(unsigned int irq, struct pci_dev *dev)
6820 {
6821 - union IO_APIC_reg_01 reg_01;
6822 - unsigned long flags;
6823 + struct msi_msg msg;
6824 + int ret;
6825 + ret = msi_compose_msg(dev, irq, &msg);
6826 + if (ret < 0)
6827 + return ret;
6828
6829 - spin_lock_irqsave(&ioapic_lock, flags);
6830 - reg_01.raw = io_apic_read(ioapic, 1);
6831 - spin_unlock_irqrestore(&ioapic_lock, flags);
6832 + write_msi_msg(irq, &msg);
6833
6834 - return reg_01.bits.version;
6835 + set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge");
6836 +
6837 + return 0;
6838 +}
6839 +
6840 +void arch_teardown_msi_irq(unsigned int irq)
6841 +{
6842 + return;
6843 }
6844
6845 +#endif /* CONFIG_PCI_MSI */
6846 +
6847 +/*
6848 + * Hypertransport interrupt support
6849 + */
6850 +#ifdef CONFIG_HT_IRQ
6851 +
6852 +#ifdef CONFIG_SMP
6853 +
6854 +static void target_ht_irq(unsigned int irq, unsigned int dest, u8 vector)
6855 +{
6856 + struct ht_irq_msg msg;
6857 + fetch_ht_irq_msg(irq, &msg);
6858 +
6859 + msg.address_lo &= ~(HT_IRQ_LOW_VECTOR_MASK | HT_IRQ_LOW_DEST_ID_MASK);
6860 + msg.address_hi &= ~(HT_IRQ_HIGH_DEST_ID_MASK);
6861 +
6862 + msg.address_lo |= HT_IRQ_LOW_VECTOR(vector) | HT_IRQ_LOW_DEST_ID(dest);
6863 + msg.address_hi |= HT_IRQ_HIGH_DEST_ID(dest);
6864 +
6865 + write_ht_irq_msg(irq, &msg);
6866 +}
6867 +
6868 +static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask)
6869 +{
6870 + unsigned int dest;
6871 + cpumask_t tmp;
6872 + int vector;
6873 +
6874 + cpus_and(tmp, mask, cpu_online_map);
6875 + if (cpus_empty(tmp))
6876 + tmp = TARGET_CPUS;
6877 +
6878 + cpus_and(mask, tmp, CPU_MASK_ALL);
6879 +
6880 + vector = assign_irq_vector(irq, mask, &tmp);
6881 + if (vector < 0)
6882 + return;
6883 +
6884 + dest = cpu_mask_to_apicid(tmp);
6885 +
6886 + target_ht_irq(irq, dest, vector);
6887 + set_native_irq_info(irq, mask);
6888 +}
6889 +#endif
6890 +
6891 +static struct irq_chip ht_irq_chip = {
6892 + .name = "PCI-HT",
6893 + .mask = mask_ht_irq,
6894 + .unmask = unmask_ht_irq,
6895 + .ack = ack_apic_edge,
6896 +#ifdef CONFIG_SMP
6897 + .set_affinity = set_ht_irq_affinity,
6898 +#endif
6899 + .retrigger = ioapic_retrigger_irq,
6900 +};
6901 +
6902 +int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
6903 +{
6904 + int vector;
6905 + cpumask_t tmp;
6906 +
6907 + vector = assign_irq_vector(irq, TARGET_CPUS, &tmp);
6908 + if (vector >= 0) {
6909 + struct ht_irq_msg msg;
6910 + unsigned dest;
6911 +
6912 + dest = cpu_mask_to_apicid(tmp);
6913 +
6914 + msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest);
6915 +
6916 + msg.address_lo =
6917 + HT_IRQ_LOW_BASE |
6918 + HT_IRQ_LOW_DEST_ID(dest) |
6919 + HT_IRQ_LOW_VECTOR(vector) |
6920 + ((INT_DEST_MODE == 0) ?
6921 + HT_IRQ_LOW_DM_PHYSICAL :
6922 + HT_IRQ_LOW_DM_LOGICAL) |
6923 + HT_IRQ_LOW_RQEOI_EDGE |
6924 + ((INT_DELIVERY_MODE != dest_LowestPrio) ?
6925 + HT_IRQ_LOW_MT_FIXED :
6926 + HT_IRQ_LOW_MT_ARBITRATED) |
6927 + HT_IRQ_LOW_IRQ_MASKED;
6928 +
6929 + write_ht_irq_msg(irq, &msg);
6930 +
6931 + set_irq_chip_and_handler_name(irq, &ht_irq_chip,
6932 + handle_edge_irq, "edge");
6933 + }
6934 + return vector;
6935 +}
6936 +#endif /* CONFIG_HT_IRQ */
6937 +
6938 +/* --------------------------------------------------------------------------
6939 + ACPI-based IOAPIC Configuration
6940 + -------------------------------------------------------------------------- */
6941 +
6942 +#ifdef CONFIG_ACPI
6943 +
6944 +#define IO_APIC_MAX_ID 0xFE
6945
6946 int __init io_apic_get_redir_entries (int ioapic)
6947 {
6948 @@ -2188,6 +2053,8 @@ int io_apic_set_pci_routing (int ioapic,
6949 {
6950 struct IO_APIC_route_entry entry;
6951 unsigned long flags;
6952 + int vector;
6953 + cpumask_t mask;
6954
6955 if (!IO_APIC_IRQ(irq)) {
6956 apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
6957 @@ -2196,6 +2063,17 @@ int io_apic_set_pci_routing (int ioapic,
6958 }
6959
6960 /*
6961 + * IRQs < 16 are already in the irq_2_pin[] map
6962 + */
6963 + if (irq >= 16)
6964 + add_pin_to_irq(irq, ioapic, pin);
6965 +
6966 +
6967 + vector = assign_irq_vector(irq, TARGET_CPUS, &mask);
6968 + if (vector < 0)
6969 + return vector;
6970 +
6971 + /*
6972 * Generate a PCI IRQ routing entry and program the IOAPIC accordingly.
6973 * Note that we mask (disable) IRQs now -- these get enabled when the
6974 * corresponding device driver registers for this IRQ.
6975 @@ -2205,19 +2083,11 @@ int io_apic_set_pci_routing (int ioapic,
6976
6977 entry.delivery_mode = INT_DELIVERY_MODE;
6978 entry.dest_mode = INT_DEST_MODE;
6979 - entry.dest.logical.logical_dest = cpu_mask_to_apicid(TARGET_CPUS);
6980 + entry.dest.logical.logical_dest = cpu_mask_to_apicid(mask);
6981 entry.trigger = edge_level;
6982 entry.polarity = active_high_low;
6983 entry.mask = 1; /* Disabled (masked) */
6984 -
6985 - irq = gsi_irq_sharing(irq);
6986 - /*
6987 - * IRQs < 16 are already in the irq_2_pin[] map
6988 - */
6989 - if (irq >= 16)
6990 - add_pin_to_irq(irq, ioapic, pin);
6991 -
6992 - entry.vector = assign_irq_vector(irq);
6993 + entry.vector = vector & 0xff;
6994
6995 apic_printk(APIC_VERBOSE,KERN_DEBUG "IOAPIC[%d]: Set PCI routing entry (%d-%d -> 0x%x -> "
6996 "IRQ %d Mode:%i Active:%i)\n", ioapic,
6997 @@ -2229,10 +2099,10 @@ int io_apic_set_pci_routing (int ioapic,
6998 if (!ioapic && (irq < 16))
6999 disable_8259A_irq(irq);
7000
7001 + ioapic_write_entry(ioapic, pin, entry);
7002 +
7003 spin_lock_irqsave(&ioapic_lock, flags);
7004 - io_apic_write(ioapic, 0x11+2*pin, *(((int *)&entry)+1));
7005 - io_apic_write(ioapic, 0x10+2*pin, *(((int *)&entry)+0));
7006 - set_native_irq_info(use_pci_vector() ? entry.vector : irq, TARGET_CPUS);
7007 + set_native_irq_info(irq, TARGET_CPUS);
7008 spin_unlock_irqrestore(&ioapic_lock, flags);
7009
7010 return 0;
7011 --- sle11-2009-05-14.orig/arch/x86/kernel/ioport_64-xen.c 2009-05-14 11:02:43.000000000 +0200
7012 +++ sle11-2009-05-14/arch/x86/kernel/ioport_64-xen.c 2009-03-04 11:28:34.000000000 +0100
7013 @@ -58,6 +58,7 @@ asmlinkage long sys_ioperm(unsigned long
7014
7015 memset(bitmap, 0xff, IO_BITMAP_BYTES);
7016 t->io_bitmap_ptr = bitmap;
7017 + set_thread_flag(TIF_IO_BITMAP);
7018
7019 set_xen_guest_handle(set_iobitmap.bitmap, (char *)bitmap);
7020 set_iobitmap.nr_ports = IO_BITMAP_BITS;
7021 --- sle11-2009-05-14.orig/arch/x86/kernel/irq_64-xen.c 2009-05-14 11:02:43.000000000 +0200
7022 +++ sle11-2009-05-14/arch/x86/kernel/irq_64-xen.c 2009-03-04 11:28:34.000000000 +0100
7023 @@ -20,11 +20,6 @@
7024 #include <asm/idle.h>
7025
7026 atomic_t irq_err_count;
7027 -#ifdef CONFIG_X86_IO_APIC
7028 -#ifdef APIC_MISMATCH_DEBUG
7029 -atomic_t irq_mis_count;
7030 -#endif
7031 -#endif
7032
7033 #ifdef CONFIG_DEBUG_STACKOVERFLOW
7034 /*
7035 @@ -79,7 +74,8 @@ int show_interrupts(struct seq_file *p,
7036 for_each_online_cpu(j)
7037 seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
7038 #endif
7039 - seq_printf(p, " %14s", irq_desc[i].chip->typename);
7040 + seq_printf(p, " %8s", irq_desc[i].chip->name);
7041 + seq_printf(p, "-%-8s", irq_desc[i].name);
7042
7043 seq_printf(p, " %s", action->name);
7044 for (action=action->next; action; action = action->next)
7045 @@ -99,11 +95,6 @@ skip:
7046 seq_putc(p, '\n');
7047 #endif
7048 seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
7049 -#ifdef CONFIG_X86_IO_APIC
7050 -#ifdef APIC_MISMATCH_DEBUG
7051 - seq_printf(p, "MIS: %10u\n", atomic_read(&irq_mis_count));
7052 -#endif
7053 -#endif
7054 }
7055 return 0;
7056 }
7057 @@ -114,24 +105,28 @@ skip:
7058 * handlers).
7059 */
7060 asmlinkage unsigned int do_IRQ(struct pt_regs *regs)
7061 -{
7062 +{
7063 + struct pt_regs *old_regs = set_irq_regs(regs);
7064 +
7065 /* high bit used in ret_from_ code */
7066 unsigned irq = ~regs->orig_rax;
7067
7068 - if (unlikely(irq >= NR_IRQS)) {
7069 - printk(KERN_EMERG "%s: cannot handle IRQ %d\n",
7070 - __FUNCTION__, irq);
7071 - BUG();
7072 - }
7073 -
7074 /*exit_idle();*/
7075 /*irq_enter();*/
7076 +
7077 #ifdef CONFIG_DEBUG_STACKOVERFLOW
7078 stack_overflow_check(regs);
7079 #endif
7080 - __do_IRQ(irq, regs);
7081 +
7082 + if (likely(irq < NR_IRQS))
7083 + generic_handle_irq(irq);
7084 + else
7085 + printk(KERN_EMERG "%s: %d.%d No irq handler for irq\n",
7086 + __func__, smp_processor_id(), irq);
7087 +
7088 /*irq_exit();*/
7089
7090 + set_irq_regs(old_regs);
7091 return 1;
7092 }
7093
7094 @@ -192,6 +187,6 @@ EXPORT_SYMBOL(do_softirq);
7095 */
7096 void ack_bad_irq(unsigned int irq)
7097 {
7098 - printk("unexpected IRQ trap at vector %02x\n", irq);
7099 + printk("unexpected IRQ trap at irq %02x\n", irq);
7100 }
7101 #endif
7102 --- sle11-2009-05-14.orig/arch/x86/kernel/mpparse_64-xen.c 2009-05-14 11:02:43.000000000 +0200
7103 +++ sle11-2009-05-14/arch/x86/kernel/mpparse_64-xen.c 2009-03-04 11:28:34.000000000 +0100
7104 @@ -41,8 +41,7 @@ int acpi_found_madt;
7105 * Various Linux-internal data structures created from the
7106 * MP-table.
7107 */
7108 -unsigned char apic_version [MAX_APICS];
7109 -unsigned char mp_bus_id_to_type [MAX_MP_BUSSES] = { [0 ... MAX_MP_BUSSES-1] = -1 };
7110 +DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
7111 int mp_bus_id_to_pci_bus [MAX_MP_BUSSES] = { [0 ... MAX_MP_BUSSES-1] = -1 };
7112
7113 static int mp_current_pci_id = 0;
7114 @@ -56,7 +55,6 @@ struct mpc_config_intsrc mp_irqs[MAX_IRQ
7115 int mp_irq_entries;
7116
7117 int nr_ioapics;
7118 -int pic_mode;
7119 unsigned long mp_lapic_addr = 0;
7120
7121
7122 @@ -71,19 +69,6 @@ unsigned disabled_cpus __initdata;
7123 /* Bitmask of physically existing CPUs */
7124 physid_mask_t phys_cpu_present_map = PHYSID_MASK_NONE;
7125
7126 -/* ACPI MADT entry parsing functions */
7127 -#ifdef CONFIG_ACPI
7128 -extern struct acpi_boot_flags acpi_boot;
7129 -#ifdef CONFIG_X86_LOCAL_APIC
7130 -extern int acpi_parse_lapic (acpi_table_entry_header *header);
7131 -extern int acpi_parse_lapic_addr_ovr (acpi_table_entry_header *header);
7132 -extern int acpi_parse_lapic_nmi (acpi_table_entry_header *header);
7133 -#endif /*CONFIG_X86_LOCAL_APIC*/
7134 -#ifdef CONFIG_X86_IO_APIC
7135 -extern int acpi_parse_ioapic (acpi_table_entry_header *header);
7136 -#endif /*CONFIG_X86_IO_APIC*/
7137 -#endif /*CONFIG_ACPI*/
7138 -
7139 u8 bios_cpu_apicid[NR_CPUS] = { [0 ... NR_CPUS-1] = BAD_APICID };
7140
7141
7142 @@ -109,24 +94,20 @@ static int __init mpf_checksum(unsigned
7143 static void __cpuinit MP_processor_info (struct mpc_config_processor *m)
7144 {
7145 int cpu;
7146 - unsigned char ver;
7147 cpumask_t tmp_map;
7148 + char *bootup_cpu = "";
7149
7150 if (!(m->mpc_cpuflag & CPU_ENABLED)) {
7151 disabled_cpus++;
7152 return;
7153 }
7154 -
7155 - printk(KERN_INFO "Processor #%d %d:%d APIC version %d\n",
7156 - m->mpc_apicid,
7157 - (m->mpc_cpufeature & CPU_FAMILY_MASK)>>8,
7158 - (m->mpc_cpufeature & CPU_MODEL_MASK)>>4,
7159 - m->mpc_apicver);
7160 -
7161 if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) {
7162 - Dprintk(" Bootup CPU\n");
7163 + bootup_cpu = " (Bootup-CPU)";
7164 boot_cpu_id = m->mpc_apicid;
7165 }
7166 +
7167 + printk(KERN_INFO "Processor #%d%s\n", m->mpc_apicid, bootup_cpu);
7168 +
7169 if (num_processors >= NR_CPUS) {
7170 printk(KERN_WARNING "WARNING: NR_CPUS limit of %i reached."
7171 " Processor ignored.\n", NR_CPUS);
7172 @@ -137,24 +118,7 @@ static void __cpuinit MP_processor_info
7173 cpus_complement(tmp_map, cpu_present_map);
7174 cpu = first_cpu(tmp_map);
7175
7176 -#if MAX_APICS < 255
7177 - if ((int)m->mpc_apicid > MAX_APICS) {
7178 - printk(KERN_ERR "Processor #%d INVALID. (Max ID: %d).\n",
7179 - m->mpc_apicid, MAX_APICS);
7180 - return;
7181 - }
7182 -#endif
7183 - ver = m->mpc_apicver;
7184 -
7185 physid_set(m->mpc_apicid, phys_cpu_present_map);
7186 - /*
7187 - * Validate version
7188 - */
7189 - if (ver == 0x0) {
7190 - printk(KERN_ERR "BIOS bug, APIC version is 0 for CPU#%d! fixing up to 0x10. (tell your hw vendor)\n", m->mpc_apicid);
7191 - ver = 0x10;
7192 - }
7193 - apic_version[m->mpc_apicid] = ver;
7194 if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) {
7195 /*
7196 * bios_cpu_apicid is required to have processors listed
7197 @@ -185,37 +149,42 @@ static void __init MP_bus_info (struct m
7198 Dprintk("Bus #%d is %s\n", m->mpc_busid, str);
7199
7200 if (strncmp(str, "ISA", 3) == 0) {
7201 - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA;
7202 - } else if (strncmp(str, "EISA", 4) == 0) {
7203 - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_EISA;
7204 + set_bit(m->mpc_busid, mp_bus_not_pci);
7205 } else if (strncmp(str, "PCI", 3) == 0) {
7206 - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_PCI;
7207 + clear_bit(m->mpc_busid, mp_bus_not_pci);
7208 mp_bus_id_to_pci_bus[m->mpc_busid] = mp_current_pci_id;
7209 mp_current_pci_id++;
7210 - } else if (strncmp(str, "MCA", 3) == 0) {
7211 - mp_bus_id_to_type[m->mpc_busid] = MP_BUS_MCA;
7212 } else {
7213 printk(KERN_ERR "Unknown bustype %s\n", str);
7214 }
7215 }
7216
7217 +static int bad_ioapic(unsigned long address)
7218 +{
7219 + if (nr_ioapics >= MAX_IO_APICS) {
7220 + printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded "
7221 + "(found %d)\n", MAX_IO_APICS, nr_ioapics);
7222 + panic("Recompile kernel with bigger MAX_IO_APICS!\n");
7223 + }
7224 + if (!address) {
7225 + printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address"
7226 + " found in table, skipping!\n");
7227 + return 1;
7228 + }
7229 + return 0;
7230 +}
7231 +
7232 static void __init MP_ioapic_info (struct mpc_config_ioapic *m)
7233 {
7234 if (!(m->mpc_flags & MPC_APIC_USABLE))
7235 return;
7236
7237 - printk("I/O APIC #%d Version %d at 0x%X.\n",
7238 - m->mpc_apicid, m->mpc_apicver, m->mpc_apicaddr);
7239 - if (nr_ioapics >= MAX_IO_APICS) {
7240 - printk(KERN_ERR "Max # of I/O APICs (%d) exceeded (found %d).\n",
7241 - MAX_IO_APICS, nr_ioapics);
7242 - panic("Recompile kernel with bigger MAX_IO_APICS!.\n");
7243 - }
7244 - if (!m->mpc_apicaddr) {
7245 - printk(KERN_ERR "WARNING: bogus zero I/O APIC address"
7246 - " found in MP table, skipping!\n");
7247 + printk("I/O APIC #%d at 0x%X.\n",
7248 + m->mpc_apicid, m->mpc_apicaddr);
7249 +
7250 + if (bad_ioapic(m->mpc_apicaddr))
7251 return;
7252 - }
7253 +
7254 mp_ioapics[nr_ioapics] = *m;
7255 nr_ioapics++;
7256 }
7257 @@ -239,19 +208,6 @@ static void __init MP_lintsrc_info (stru
7258 m->mpc_irqtype, m->mpc_irqflag & 3,
7259 (m->mpc_irqflag >> 2) &3, m->mpc_srcbusid,
7260 m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint);
7261 - /*
7262 - * Well it seems all SMP boards in existence
7263 - * use ExtINT/LVT1 == LINT0 and
7264 - * NMI/LVT2 == LINT1 - the following check
7265 - * will show us if this assumptions is false.
7266 - * Until then we do not have to add baggage.
7267 - */
7268 - if ((m->mpc_irqtype == mp_ExtINT) &&
7269 - (m->mpc_destapiclint != 0))
7270 - BUG();
7271 - if ((m->mpc_irqtype == mp_NMI) &&
7272 - (m->mpc_destapiclint != 1))
7273 - BUG();
7274 }
7275
7276 /*
7277 @@ -265,7 +221,7 @@ static int __init smp_read_mpc(struct mp
7278 unsigned char *mpt=((unsigned char *)mpc)+count;
7279
7280 if (memcmp(mpc->mpc_signature,MPC_SIGNATURE,4)) {
7281 - printk("SMP mptable: bad signature [%c%c%c%c]!\n",
7282 + printk("MPTABLE: bad signature [%c%c%c%c]!\n",
7283 mpc->mpc_signature[0],
7284 mpc->mpc_signature[1],
7285 mpc->mpc_signature[2],
7286 @@ -273,31 +229,31 @@ static int __init smp_read_mpc(struct mp
7287 return 0;
7288 }
7289 if (mpf_checksum((unsigned char *)mpc,mpc->mpc_length)) {
7290 - printk("SMP mptable: checksum error!\n");
7291 + printk("MPTABLE: checksum error!\n");
7292 return 0;
7293 }
7294 if (mpc->mpc_spec!=0x01 && mpc->mpc_spec!=0x04) {
7295 - printk(KERN_ERR "SMP mptable: bad table version (%d)!!\n",
7296 + printk(KERN_ERR "MPTABLE: bad table version (%d)!!\n",
7297 mpc->mpc_spec);
7298 return 0;
7299 }
7300 if (!mpc->mpc_lapic) {
7301 - printk(KERN_ERR "SMP mptable: null local APIC address!\n");
7302 + printk(KERN_ERR "MPTABLE: null local APIC address!\n");
7303 return 0;
7304 }
7305 memcpy(str,mpc->mpc_oem,8);
7306 - str[8]=0;
7307 - printk(KERN_INFO "OEM ID: %s ",str);
7308 + str[8] = 0;
7309 + printk(KERN_INFO "MPTABLE: OEM ID: %s ",str);
7310
7311 memcpy(str,mpc->mpc_productid,12);
7312 - str[12]=0;
7313 - printk("Product ID: %s ",str);
7314 + str[12] = 0;
7315 + printk("MPTABLE: Product ID: %s ",str);
7316
7317 - printk("APIC at: 0x%X\n",mpc->mpc_lapic);
7318 + printk("MPTABLE: APIC at: 0x%X\n",mpc->mpc_lapic);
7319
7320 /* save the local APIC address, it might be non-default */
7321 if (!acpi_lapic)
7322 - mp_lapic_addr = mpc->mpc_lapic;
7323 + mp_lapic_addr = mpc->mpc_lapic;
7324
7325 /*
7326 * Now process the configuration blocks.
7327 @@ -309,7 +265,7 @@ static int __init smp_read_mpc(struct mp
7328 struct mpc_config_processor *m=
7329 (struct mpc_config_processor *)mpt;
7330 if (!acpi_lapic)
7331 - MP_processor_info(m);
7332 + MP_processor_info(m);
7333 mpt += sizeof(*m);
7334 count += sizeof(*m);
7335 break;
7336 @@ -328,8 +284,8 @@ static int __init smp_read_mpc(struct mp
7337 struct mpc_config_ioapic *m=
7338 (struct mpc_config_ioapic *)mpt;
7339 MP_ioapic_info(m);
7340 - mpt+=sizeof(*m);
7341 - count+=sizeof(*m);
7342 + mpt += sizeof(*m);
7343 + count += sizeof(*m);
7344 break;
7345 }
7346 case MP_INTSRC:
7347 @@ -338,8 +294,8 @@ static int __init smp_read_mpc(struct mp
7348 (struct mpc_config_intsrc *)mpt;
7349
7350 MP_intsrc_info(m);
7351 - mpt+=sizeof(*m);
7352 - count+=sizeof(*m);
7353 + mpt += sizeof(*m);
7354 + count += sizeof(*m);
7355 break;
7356 }
7357 case MP_LINTSRC:
7358 @@ -347,15 +303,15 @@ static int __init smp_read_mpc(struct mp
7359 struct mpc_config_lintsrc *m=
7360 (struct mpc_config_lintsrc *)mpt;
7361 MP_lintsrc_info(m);
7362 - mpt+=sizeof(*m);
7363 - count+=sizeof(*m);
7364 + mpt += sizeof(*m);
7365 + count += sizeof(*m);
7366 break;
7367 }
7368 }
7369 }
7370 clustered_apic_check();
7371 if (!num_processors)
7372 - printk(KERN_ERR "SMP mptable: no processors registered!\n");
7373 + printk(KERN_ERR "MPTABLE: no processors registered!\n");
7374 return num_processors;
7375 }
7376
7377 @@ -451,13 +407,10 @@ static inline void __init construct_defa
7378 * 2 CPUs, numbered 0 & 1.
7379 */
7380 processor.mpc_type = MP_PROCESSOR;
7381 - /* Either an integrated APIC or a discrete 82489DX. */
7382 - processor.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01;
7383 + processor.mpc_apicver = 0;
7384 processor.mpc_cpuflag = CPU_ENABLED;
7385 - processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) |
7386 - (boot_cpu_data.x86_model << 4) |
7387 - boot_cpu_data.x86_mask;
7388 - processor.mpc_featureflag = boot_cpu_data.x86_capability[0];
7389 + processor.mpc_cpufeature = 0;
7390 + processor.mpc_featureflag = 0;
7391 processor.mpc_reserved[0] = 0;
7392 processor.mpc_reserved[1] = 0;
7393 for (i = 0; i < 2; i++) {
7394 @@ -476,14 +429,6 @@ static inline void __init construct_defa
7395 case 5:
7396 memcpy(bus.mpc_bustype, "ISA ", 6);
7397 break;
7398 - case 2:
7399 - case 6:
7400 - case 3:
7401 - memcpy(bus.mpc_bustype, "EISA ", 6);
7402 - break;
7403 - case 4:
7404 - case 7:
7405 - memcpy(bus.mpc_bustype, "MCA ", 6);
7406 }
7407 MP_bus_info(&bus);
7408 if (mpc_default_type > 4) {
7409 @@ -494,7 +439,7 @@ static inline void __init construct_defa
7410
7411 ioapic.mpc_type = MP_IOAPIC;
7412 ioapic.mpc_apicid = 2;
7413 - ioapic.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01;
7414 + ioapic.mpc_apicver = 0;
7415 ioapic.mpc_flags = MPC_APIC_USABLE;
7416 ioapic.mpc_apicaddr = 0xFEC00000;
7417 MP_ioapic_info(&ioapic);
7418 @@ -537,13 +482,6 @@ void __init get_smp_config (void)
7419 printk(KERN_INFO "Using ACPI for processor (LAPIC) configuration information\n");
7420
7421 printk("Intel MultiProcessor Specification v1.%d\n", mpf->mpf_specification);
7422 - if (mpf->mpf_feature2 & (1<<7)) {
7423 - printk(KERN_INFO " IMCR and PIC compatibility mode.\n");
7424 - pic_mode = 1;
7425 - } else {
7426 - printk(KERN_INFO " Virtual Wire compatibility mode.\n");
7427 - pic_mode = 0;
7428 - }
7429
7430 /*
7431 * Now see if we need to read further.
7432 @@ -620,7 +558,7 @@ static int __init smp_scan_config (unsig
7433 return 0;
7434 }
7435
7436 -void __init find_intel_smp (void)
7437 +void __init find_smp_config(void)
7438 {
7439 unsigned int address;
7440
7441 @@ -637,9 +575,7 @@ void __init find_intel_smp (void)
7442 smp_scan_config(0xF0000,0x10000))
7443 return;
7444 /*
7445 - * If it is an SMP machine we should know now, unless the
7446 - * configuration is in an EISA/MCA bus machine with an
7447 - * extended bios data area.
7448 + * If it is an SMP machine we should know now.
7449 *
7450 * there is a real-mode segmented pointer pointing to the
7451 * 4K EBDA area at 0x40E, calculate and scan it here.
7452 @@ -660,64 +596,38 @@ void __init find_intel_smp (void)
7453 printk(KERN_INFO "No mptable found.\n");
7454 }
7455
7456 -/*
7457 - * - Intel MP Configuration Table
7458 - */
7459 -void __init find_smp_config (void)
7460 -{
7461 -#ifdef CONFIG_X86_LOCAL_APIC
7462 - find_intel_smp();
7463 -#endif
7464 -}
7465 -
7466 -
7467 /* --------------------------------------------------------------------------
7468 ACPI-based MP Configuration
7469 -------------------------------------------------------------------------- */
7470
7471 #ifdef CONFIG_ACPI
7472
7473 -void __init mp_register_lapic_address (
7474 - u64 address)
7475 +void __init mp_register_lapic_address(u64 address)
7476 {
7477 #ifndef CONFIG_XEN
7478 mp_lapic_addr = (unsigned long) address;
7479 -
7480 set_fixmap_nocache(FIX_APIC_BASE, mp_lapic_addr);
7481 -
7482 if (boot_cpu_id == -1U)
7483 boot_cpu_id = GET_APIC_ID(apic_read(APIC_ID));
7484 -
7485 - Dprintk("Boot CPU = %d\n", boot_cpu_physical_apicid);
7486 #endif
7487 }
7488
7489 -
7490 -void __cpuinit mp_register_lapic (
7491 - u8 id,
7492 - u8 enabled)
7493 +void __cpuinit mp_register_lapic (u8 id, u8 enabled)
7494 {
7495 struct mpc_config_processor processor;
7496 int boot_cpu = 0;
7497
7498 - if (id >= MAX_APICS) {
7499 - printk(KERN_WARNING "Processor #%d invalid (max %d)\n",
7500 - id, MAX_APICS);
7501 - return;
7502 - }
7503 -
7504 - if (id == boot_cpu_physical_apicid)
7505 + if (id == boot_cpu_id)
7506 boot_cpu = 1;
7507
7508 #ifndef CONFIG_XEN
7509 processor.mpc_type = MP_PROCESSOR;
7510 processor.mpc_apicid = id;
7511 - processor.mpc_apicver = GET_APIC_VERSION(apic_read(APIC_LVR));
7512 + processor.mpc_apicver = 0;
7513 processor.mpc_cpuflag = (enabled ? CPU_ENABLED : 0);
7514 processor.mpc_cpuflag |= (boot_cpu ? CPU_BOOTPROCESSOR : 0);
7515 - processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) |
7516 - (boot_cpu_data.x86_model << 4) | boot_cpu_data.x86_mask;
7517 - processor.mpc_featureflag = boot_cpu_data.x86_capability[0];
7518 + processor.mpc_cpufeature = 0;
7519 + processor.mpc_featureflag = 0;
7520 processor.mpc_reserved[0] = 0;
7521 processor.mpc_reserved[1] = 0;
7522 #endif
7523 @@ -725,8 +635,6 @@ void __cpuinit mp_register_lapic (
7524 MP_processor_info(&processor);
7525 }
7526
7527 -#ifdef CONFIG_X86_IO_APIC
7528 -
7529 #define MP_ISA_BUS 0
7530 #define MP_MAX_IOAPIC_PIN 127
7531
7532 @@ -737,11 +645,9 @@ static struct mp_ioapic_routing {
7533 u32 pin_programmed[4];
7534 } mp_ioapic_routing[MAX_IO_APICS];
7535
7536 -
7537 -static int mp_find_ioapic (
7538 - int gsi)
7539 +static int mp_find_ioapic(int gsi)
7540 {
7541 - int i = 0;
7542 + int i = 0;
7543
7544 /* Find the IOAPIC that manages this GSI. */
7545 for (i = 0; i < nr_ioapics; i++) {
7546 @@ -751,28 +657,15 @@ static int mp_find_ioapic (
7547 }
7548
7549 printk(KERN_ERR "ERROR: Unable to locate IOAPIC for GSI %d\n", gsi);
7550 -
7551 return -1;
7552 }
7553 -
7554
7555 -void __init mp_register_ioapic (
7556 - u8 id,
7557 - u32 address,
7558 - u32 gsi_base)
7559 +void __init mp_register_ioapic(u8 id, u32 address, u32 gsi_base)
7560 {
7561 - int idx = 0;
7562 + int idx = 0;
7563
7564 - if (nr_ioapics >= MAX_IO_APICS) {
7565 - printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded "
7566 - "(found %d)\n", MAX_IO_APICS, nr_ioapics);
7567 - panic("Recompile kernel with bigger MAX_IO_APICS!\n");
7568 - }
7569 - if (!address) {
7570 - printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address"
7571 - " found in MADT table, skipping!\n");
7572 + if (bad_ioapic(address))
7573 return;
7574 - }
7575
7576 idx = nr_ioapics++;
7577
7578 @@ -784,7 +677,7 @@ void __init mp_register_ioapic (
7579 set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address);
7580 #endif
7581 mp_ioapics[idx].mpc_apicid = id;
7582 - mp_ioapics[idx].mpc_apicver = io_apic_get_version(idx);
7583 + mp_ioapics[idx].mpc_apicver = 0;
7584
7585 /*
7586 * Build basic IRQ lookup table to facilitate gsi->io_apic lookups
7587 @@ -795,21 +688,15 @@ void __init mp_register_ioapic (
7588 mp_ioapic_routing[idx].gsi_end = gsi_base +
7589 io_apic_get_redir_entries(idx);
7590
7591 - printk(KERN_INFO "IOAPIC[%d]: apic_id %d, version %d, address 0x%x, "
7592 + printk(KERN_INFO "IOAPIC[%d]: apic_id %d, address 0x%x, "
7593 "GSI %d-%d\n", idx, mp_ioapics[idx].mpc_apicid,
7594 - mp_ioapics[idx].mpc_apicver, mp_ioapics[idx].mpc_apicaddr,
7595 + mp_ioapics[idx].mpc_apicaddr,
7596 mp_ioapic_routing[idx].gsi_start,
7597 mp_ioapic_routing[idx].gsi_end);
7598 -
7599 - return;
7600 }
7601
7602 -
7603 -void __init mp_override_legacy_irq (
7604 - u8 bus_irq,
7605 - u8 polarity,
7606 - u8 trigger,
7607 - u32 gsi)
7608 +void __init
7609 +mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
7610 {
7611 struct mpc_config_intsrc intsrc;
7612 int ioapic = -1;
7613 @@ -847,22 +734,18 @@ void __init mp_override_legacy_irq (
7614 mp_irqs[mp_irq_entries] = intsrc;
7615 if (++mp_irq_entries == MAX_IRQ_SOURCES)
7616 panic("Max # of irq sources exceeded!\n");
7617 -
7618 - return;
7619 }
7620
7621 -
7622 -void __init mp_config_acpi_legacy_irqs (void)
7623 +void __init mp_config_acpi_legacy_irqs(void)
7624 {
7625 struct mpc_config_intsrc intsrc;
7626 - int i = 0;
7627 - int ioapic = -1;
7628 + int i = 0;
7629 + int ioapic = -1;
7630
7631 /*
7632 * Fabricate the legacy ISA bus (bus #31).
7633 */
7634 - mp_bus_id_to_type[MP_ISA_BUS] = MP_BUS_ISA;
7635 - Dprintk("Bus #%d is ISA\n", MP_ISA_BUS);
7636 + set_bit(MP_ISA_BUS, mp_bus_not_pci);
7637
7638 /*
7639 * Locate the IOAPIC that manages the ISA IRQs (0-15).
7640 @@ -915,24 +798,13 @@ void __init mp_config_acpi_legacy_irqs (
7641 if (++mp_irq_entries == MAX_IRQ_SOURCES)
7642 panic("Max # of irq sources exceeded!\n");
7643 }
7644 -
7645 - return;
7646 }
7647
7648 -#define MAX_GSI_NUM 4096
7649 -
7650 int mp_register_gsi(u32 gsi, int triggering, int polarity)
7651 {
7652 - int ioapic = -1;
7653 - int ioapic_pin = 0;
7654 - int idx, bit = 0;
7655 - static int pci_irq = 16;
7656 - /*
7657 - * Mapping between Global System Interrupts, which
7658 - * represent all possible interrupts, to the IRQs
7659 - * assigned to actual devices.
7660 - */
7661 - static int gsi_to_irq[MAX_GSI_NUM];
7662 + int ioapic = -1;
7663 + int ioapic_pin = 0;
7664 + int idx, bit = 0;
7665
7666 if (acpi_irq_model != ACPI_IRQ_MODEL_IOAPIC)
7667 return gsi;
7668 @@ -965,47 +837,14 @@ int mp_register_gsi(u32 gsi, int trigger
7669 if ((1<<bit) & mp_ioapic_routing[ioapic].pin_programmed[idx]) {
7670 Dprintk(KERN_DEBUG "Pin %d-%d already programmed\n",
7671 mp_ioapic_routing[ioapic].apic_id, ioapic_pin);
7672 - return gsi_to_irq[gsi];
7673 + return gsi;
7674 }
7675
7676 mp_ioapic_routing[ioapic].pin_programmed[idx] |= (1<<bit);
7677
7678 - if (triggering == ACPI_LEVEL_SENSITIVE) {
7679 - /*
7680 - * For PCI devices assign IRQs in order, avoiding gaps
7681 - * due to unused I/O APIC pins.
7682 - */
7683 - int irq = gsi;
7684 - if (gsi < MAX_GSI_NUM) {
7685 - /*
7686 - * Retain the VIA chipset work-around (gsi > 15), but
7687 - * avoid a problem where the 8254 timer (IRQ0) is setup
7688 - * via an override (so it's not on pin 0 of the ioapic),
7689 - * and at the same time, the pin 0 interrupt is a PCI
7690 - * type. The gsi > 15 test could cause these two pins
7691 - * to be shared as IRQ0, and they are not shareable.
7692 - * So test for this condition, and if necessary, avoid
7693 - * the pin collision.
7694 - */
7695 - if (gsi > 15 || (gsi == 0 && !timer_uses_ioapic_pin_0))
7696 - gsi = pci_irq++;
7697 - /*
7698 - * Don't assign IRQ used by ACPI SCI
7699 - */
7700 - if (gsi == acpi_fadt.sci_int)
7701 - gsi = pci_irq++;
7702 - gsi_to_irq[irq] = gsi;
7703 - } else {
7704 - printk(KERN_ERR "GSI %u is too high\n", gsi);
7705 - return gsi;
7706 - }
7707 - }
7708 -
7709 io_apic_set_pci_routing(ioapic, ioapic_pin, gsi,
7710 triggering == ACPI_EDGE_SENSITIVE ? 0 : 1,
7711 polarity == ACPI_ACTIVE_HIGH ? 0 : 1);
7712 return gsi;
7713 }
7714 -
7715 -#endif /*CONFIG_X86_IO_APIC*/
7716 #endif /*CONFIG_ACPI*/
7717 --- sle11-2009-05-14.orig/arch/x86/kernel/process_64-xen.c 2009-05-14 11:02:43.000000000 +0200
7718 +++ sle11-2009-05-14/arch/x86/kernel/process_64-xen.c 2009-03-04 11:28:34.000000000 +0100
7719 @@ -89,25 +89,24 @@ void idle_notifier_unregister(struct not
7720 }
7721 EXPORT_SYMBOL(idle_notifier_unregister);
7722
7723 -enum idle_state { CPU_IDLE, CPU_NOT_IDLE };
7724 -static DEFINE_PER_CPU(enum idle_state, idle_state) = CPU_NOT_IDLE;
7725 -
7726 void enter_idle(void)
7727 {
7728 - __get_cpu_var(idle_state) = CPU_IDLE;
7729 + write_pda(isidle, 1);
7730 atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL);
7731 }
7732
7733 static void __exit_idle(void)
7734 {
7735 - __get_cpu_var(idle_state) = CPU_NOT_IDLE;
7736 + if (test_and_clear_bit_pda(0, isidle) == 0)
7737 + return;
7738 atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL);
7739 }
7740
7741 /* Called from interrupts to signify idle end */
7742 void exit_idle(void)
7743 {
7744 - if (current->pid | read_pda(irqcount))
7745 + /* idle loop has pid 0 */
7746 + if (current->pid)
7747 return;
7748 __exit_idle();
7749 }
7750 @@ -184,6 +183,9 @@ void cpu_idle (void)
7751 play_dead();
7752 enter_idle();
7753 idle();
7754 + /* In many cases the interrupt that ended idle
7755 + has already called exit_idle. But some idle
7756 + loops can be woken up without interrupt. */
7757 __exit_idle();
7758 }
7759
7760 @@ -196,7 +198,7 @@ void cpu_idle (void)
7761 void cpu_idle_wait(void)
7762 {
7763 unsigned int cpu, this_cpu = get_cpu();
7764 - cpumask_t map;
7765 + cpumask_t map, tmp = current->cpus_allowed;
7766
7767 set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
7768 put_cpu();
7769 @@ -219,6 +221,8 @@ void cpu_idle_wait(void)
7770 }
7771 cpus_and(map, map, cpu_online_map);
7772 } while (!cpus_empty(map));
7773 +
7774 + set_cpus_allowed(current, tmp);
7775 }
7776 EXPORT_SYMBOL_GPL(cpu_idle_wait);
7777
7778 @@ -250,9 +254,9 @@ void __show_regs(struct pt_regs * regs)
7779 print_modules();
7780 printk("Pid: %d, comm: %.20s %s %s %.*s\n",
7781 current->pid, current->comm, print_tainted(),
7782 - system_utsname.release,
7783 - (int)strcspn(system_utsname.version, " "),
7784 - system_utsname.version);
7785 + init_utsname()->release,
7786 + (int)strcspn(init_utsname()->version, " "),
7787 + init_utsname()->version);
7788 printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->rip);
7789 printk_address(regs->rip);
7790 printk("RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->rsp,
7791 @@ -310,6 +314,7 @@ void exit_thread(void)
7792
7793 kfree(t->io_bitmap_ptr);
7794 t->io_bitmap_ptr = NULL;
7795 + clear_thread_flag(TIF_IO_BITMAP);
7796 /*
7797 * Careful, clear this in the TSS too:
7798 */
7799 @@ -340,6 +345,7 @@ void flush_thread(void)
7800 if (t->flags & _TIF_IA32)
7801 current_thread_info()->status |= TS_COMPAT;
7802 }
7803 + t->flags &= ~_TIF_DEBUG;
7804
7805 tsk->thread.debugreg0 = 0;
7806 tsk->thread.debugreg1 = 0;
7807 @@ -432,7 +438,7 @@ int copy_thread(int nr, unsigned long cl
7808 asm("mov %%es,%0" : "=m" (p->thread.es));
7809 asm("mov %%ds,%0" : "=m" (p->thread.ds));
7810
7811 - if (unlikely(me->thread.io_bitmap_ptr != NULL)) {
7812 + if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) {
7813 p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL);
7814 if (!p->thread.io_bitmap_ptr) {
7815 p->thread.io_bitmap_max = 0;
7816 @@ -440,6 +446,7 @@ int copy_thread(int nr, unsigned long cl
7817 }
7818 memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr,
7819 IO_BITMAP_BYTES);
7820 + set_tsk_thread_flag(p, TIF_IO_BITMAP);
7821 }
7822
7823 /*
7824 @@ -474,6 +481,30 @@ static inline void __save_init_fpu( stru
7825 }
7826
7827 /*
7828 + * This special macro can be used to load a debugging register
7829 + */
7830 +#define loaddebug(thread,r) set_debugreg(thread->debugreg ## r, r)
7831 +
7832 +static inline void __switch_to_xtra(struct task_struct *prev_p,
7833 + struct task_struct *next_p)
7834 +{
7835 + struct thread_struct *prev, *next;
7836 +
7837 + prev = &prev_p->thread,
7838 + next = &next_p->thread;
7839 +
7840 + if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
7841 + loaddebug(next, 0);
7842 + loaddebug(next, 1);
7843 + loaddebug(next, 2);
7844 + loaddebug(next, 3);
7845 + /* no 4 and 5 */
7846 + loaddebug(next, 6);
7847 + loaddebug(next, 7);
7848 + }
7849 +}
7850 +
7851 +/*
7852 * switch_to(x,y) should switch tasks from x to y.
7853 *
7854 * This could still be optimized:
7855 @@ -501,6 +532,10 @@ __switch_to(struct task_struct *prev_p,
7856 #endif
7857 multicall_entry_t _mcl[8], *mcl = _mcl;
7858
7859 + /* we're going to use this soon, after a few expensive things */
7860 + if (next_p->fpu_counter>5)
7861 + prefetch(&next->i387.fxsave);
7862 +
7863 /*
7864 * This is basically '__unlazy_fpu', except that we queue a
7865 * multicall to indicate FPU task switch, rather than
7866 @@ -513,7 +548,8 @@ __switch_to(struct task_struct *prev_p,
7867 mcl->op = __HYPERVISOR_fpu_taskswitch;
7868 mcl->args[0] = 1;
7869 mcl++;
7870 - }
7871 + } else
7872 + prev_p->fpu_counter = 0;
7873
7874 /*
7875 * Reload esp0, LDT and the page table pointer:
7876 @@ -608,21 +644,29 @@ __switch_to(struct task_struct *prev_p,
7877 write_pda(oldrsp, next->userrsp);
7878 write_pda(pcurrent, next_p);
7879 write_pda(kernelstack,
7880 - task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET);
7881 + (unsigned long)task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET);
7882 +#ifdef CONFIG_CC_STACKPROTECTOR
7883 + write_pda(stack_canary, next_p->stack_canary);
7884 +
7885 + /*
7886 + * Build time only check to make sure the stack_canary is at
7887 + * offset 40 in the pda; this is a gcc ABI requirement
7888 + */
7889 + BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40);
7890 +#endif
7891
7892 /*
7893 * Now maybe reload the debug registers
7894 */
7895 - if (unlikely(next->debugreg7)) {
7896 - set_debugreg(next->debugreg0, 0);
7897 - set_debugreg(next->debugreg1, 1);
7898 - set_debugreg(next->debugreg2, 2);
7899 - set_debugreg(next->debugreg3, 3);
7900 - /* no 4 and 5 */
7901 - set_debugreg(next->debugreg6, 6);
7902 - set_debugreg(next->debugreg7, 7);
7903 - }
7904 + if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW))
7905 + __switch_to_xtra(prev_p, next_p);
7906
7907 + /* If the task has used fpu the last 5 timeslices, just do a full
7908 + * restore of the math state immediately to avoid the trap; the
7909 + * chances of needing FPU soon are obviously high now
7910 + */
7911 + if (next_p->fpu_counter>5)
7912 + math_state_restore();
7913 return prev_p;
7914 }
7915
7916 @@ -842,7 +886,7 @@ int dump_task_regs(struct task_struct *t
7917
7918 unsigned long arch_align_stack(unsigned long sp)
7919 {
7920 - if (randomize_va_space)
7921 + if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
7922 sp -= get_random_int() % 8192;
7923 return sp & ~0xf;
7924 }
7925 --- sle11-2009-05-14.orig/arch/x86/kernel/setup_64-xen.c 2009-05-14 11:02:43.000000000 +0200
7926 +++ sle11-2009-05-14/arch/x86/kernel/setup_64-xen.c 2009-03-04 11:28:34.000000000 +0100
7927 @@ -88,9 +88,6 @@ extern struct edid_info edid_info;
7928 shared_info_t *HYPERVISOR_shared_info = (shared_info_t *)empty_zero_page;
7929 EXPORT_SYMBOL(HYPERVISOR_shared_info);
7930
7931 -extern char hypercall_page[PAGE_SIZE];
7932 -EXPORT_SYMBOL(hypercall_page);
7933 -
7934 static int xen_panic_event(struct notifier_block *, unsigned long, void *);
7935 static struct notifier_block xen_panic_block = {
7936 xen_panic_event, NULL, 0 /* try to go last */
7937 @@ -118,16 +115,6 @@ EXPORT_SYMBOL(boot_cpu_data);
7938
7939 unsigned long mmu_cr4_features;
7940
7941 -int acpi_disabled;
7942 -EXPORT_SYMBOL(acpi_disabled);
7943 -#ifdef CONFIG_ACPI
7944 -extern int __initdata acpi_ht;
7945 -extern acpi_interrupt_flags acpi_sci_flags;
7946 -int __initdata acpi_force = 0;
7947 -#endif
7948 -
7949 -int acpi_numa __initdata;
7950 -
7951 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
7952 int bootloader_type;
7953
7954 @@ -151,10 +138,6 @@ struct sys_desc_table_struct {
7955
7956 struct edid_info edid_info;
7957 EXPORT_SYMBOL_GPL(edid_info);
7958 -struct e820map e820;
7959 -#ifdef CONFIG_XEN
7960 -struct e820map machine_e820;
7961 -#endif
7962
7963 extern int root_mountflags;
7964
7965 @@ -181,9 +164,6 @@ struct resource standard_io_resources[]
7966 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
7967 };
7968
7969 -#define STANDARD_IO_RESOURCES \
7970 - (sizeof standard_io_resources / sizeof standard_io_resources[0])
7971 -
7972 #define IORESOURCE_RAM (IORESOURCE_BUSY | IORESOURCE_MEM)
7973
7974 struct resource data_resource = {
7975 @@ -230,9 +210,6 @@ static struct resource adapter_rom_resou
7976 .flags = IORESOURCE_ROM }
7977 };
7978
7979 -#define ADAPTER_ROM_RESOURCES \
7980 - (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0])
7981 -
7982 static struct resource video_rom_resource = {
7983 .name = "Video ROM",
7984 .start = 0xc0000,
7985 @@ -309,7 +286,8 @@ static void __init probe_roms(void)
7986 }
7987
7988 /* check for adapter roms on 2k boundaries */
7989 - for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) {
7990 + for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper;
7991 + start += 2048) {
7992 rom = isa_bus_to_virt(start);
7993 if (!romsignature(rom))
7994 continue;
7995 @@ -329,187 +307,22 @@ static void __init probe_roms(void)
7996 }
7997 }
7998
7999 -/* Check for full argument with no trailing characters */
8000 -static int fullarg(char *p, char *arg)
8001 +#ifdef CONFIG_PROC_VMCORE
8002 +/* elfcorehdr= specifies the location of elf core header
8003 + * stored by the crashed kernel. This option will be passed
8004 + * by kexec loader to the capture kernel.
8005 + */
8006 +static int __init setup_elfcorehdr(char *arg)
8007 {
8008 - int l = strlen(arg);
8009 - return !memcmp(p, arg, l) && (p[l] == 0 || isspace(p[l]));
8010 + char *end;
8011 + if (!arg)
8012 + return -EINVAL;
8013 + elfcorehdr_addr = memparse(arg, &end);
8014 + return end > arg ? 0 : -EINVAL;
8015 }
8016 -
8017 -static __init void parse_cmdline_early (char ** cmdline_p)
8018 -{
8019 - char c = ' ', *to = command_line, *from = COMMAND_LINE;
8020 - int len = 0;
8021 - int userdef = 0;
8022 -
8023 - for (;;) {
8024 - if (c != ' ')
8025 - goto next_char;
8026 -
8027 -#ifdef CONFIG_SMP
8028 - /*
8029 - * If the BIOS enumerates physical processors before logical,
8030 - * maxcpus=N at enumeration-time can be used to disable HT.
8031 - */
8032 - else if (!memcmp(from, "maxcpus=", 8)) {
8033 - extern unsigned int maxcpus;
8034 -
8035 - maxcpus = simple_strtoul(from + 8, NULL, 0);
8036 - }
8037 -#endif
8038 -#ifdef CONFIG_ACPI
8039 - /* "acpi=off" disables both ACPI table parsing and interpreter init */
8040 - if (fullarg(from,"acpi=off"))
8041 - disable_acpi();
8042 -
8043 - if (fullarg(from, "acpi=force")) {
8044 - /* add later when we do DMI horrors: */
8045 - acpi_force = 1;
8046 - acpi_disabled = 0;
8047 - }
8048 -
8049 - /* acpi=ht just means: do ACPI MADT parsing
8050 - at bootup, but don't enable the full ACPI interpreter */
8051 - if (fullarg(from, "acpi=ht")) {
8052 - if (!acpi_force)
8053 - disable_acpi();
8054 - acpi_ht = 1;
8055 - }
8056 - else if (fullarg(from, "pci=noacpi"))
8057 - acpi_disable_pci();
8058 - else if (fullarg(from, "acpi=noirq"))
8059 - acpi_noirq_set();
8060 -
8061 - else if (fullarg(from, "acpi_sci=edge"))
8062 - acpi_sci_flags.trigger = 1;
8063 - else if (fullarg(from, "acpi_sci=level"))
8064 - acpi_sci_flags.trigger = 3;
8065 - else if (fullarg(from, "acpi_sci=high"))
8066 - acpi_sci_flags.polarity = 1;
8067 - else if (fullarg(from, "acpi_sci=low"))
8068 - acpi_sci_flags.polarity = 3;
8069 -
8070 - /* acpi=strict disables out-of-spec workarounds */
8071 - else if (fullarg(from, "acpi=strict")) {
8072 - acpi_strict = 1;
8073 - }
8074 -#ifdef CONFIG_X86_IO_APIC
8075 - else if (fullarg(from, "acpi_skip_timer_override"))
8076 - acpi_skip_timer_override = 1;
8077 -#endif
8078 -#endif
8079 -
8080 -#ifndef CONFIG_XEN
8081 - if (fullarg(from, "nolapic") || fullarg(from, "disableapic")) {
8082 - clear_bit(X86_FEATURE_APIC, boot_cpu_data.x86_capability);
8083 - disable_apic = 1;
8084 - }
8085 -
8086 - if (fullarg(from, "noapic"))
8087 - skip_ioapic_setup = 1;
8088 -
8089 - if (fullarg(from,"apic")) {
8090 - skip_ioapic_setup = 0;
8091 - ioapic_force = 1;
8092 - }
8093 -#endif
8094 -
8095 - if (!memcmp(from, "mem=", 4))
8096 - parse_memopt(from+4, &from);
8097 -
8098 - if (!memcmp(from, "memmap=", 7)) {
8099 - /* exactmap option is for used defined memory */
8100 - if (!memcmp(from+7, "exactmap", 8)) {
8101 -#ifdef CONFIG_CRASH_DUMP
8102 - /* If we are doing a crash dump, we
8103 - * still need to know the real mem
8104 - * size before original memory map is
8105 - * reset.
8106 - */
8107 - saved_max_pfn = e820_end_of_ram();
8108 -#endif
8109 - from += 8+7;
8110 - end_pfn_map = 0;
8111 - e820.nr_map = 0;
8112 - userdef = 1;
8113 - }
8114 - else {
8115 - parse_memmapopt(from+7, &from);
8116 - userdef = 1;
8117 - }
8118 - }
8119 -
8120 -#ifdef CONFIG_NUMA
8121 - if (!memcmp(from, "numa=", 5))
8122 - numa_setup(from+5);
8123 +early_param("elfcorehdr", setup_elfcorehdr);
8124 #endif
8125
8126 - if (!memcmp(from,"iommu=",6)) {
8127 - iommu_setup(from+6);
8128 - }
8129 -
8130 - if (fullarg(from,"oops=panic"))
8131 - panic_on_oops = 1;
8132 -
8133 - if (!memcmp(from, "noexec=", 7))
8134 - nonx_setup(from + 7);
8135 -
8136 -#ifdef CONFIG_KEXEC
8137 - /* crashkernel=size@addr specifies the location to reserve for
8138 - * a crash kernel. By reserving this memory we guarantee
8139 - * that linux never set's it up as a DMA target.
8140 - * Useful for holding code to do something appropriate
8141 - * after a kernel panic.
8142 - */
8143 - else if (!memcmp(from, "crashkernel=", 12)) {
8144 -#ifndef CONFIG_XEN
8145 - unsigned long size, base;
8146 - size = memparse(from+12, &from);
8147 - if (*from == '@') {
8148 - base = memparse(from+1, &from);
8149 - /* FIXME: Do I want a sanity check
8150 - * to validate the memory range?
8151 - */
8152 - crashk_res.start = base;
8153 - crashk_res.end = base + size - 1;
8154 - }
8155 -#else
8156 - printk("Ignoring crashkernel command line, "
8157 - "parameter will be supplied by xen\n");
8158 -#endif
8159 - }
8160 -#endif
8161 -
8162 -#ifdef CONFIG_PROC_VMCORE
8163 - /* elfcorehdr= specifies the location of elf core header
8164 - * stored by the crashed kernel. This option will be passed
8165 - * by kexec loader to the capture kernel.
8166 - */
8167 - else if(!memcmp(from, "elfcorehdr=", 11))
8168 - elfcorehdr_addr = memparse(from+11, &from);
8169 -#endif
8170 -
8171 -#if defined(CONFIG_HOTPLUG_CPU) && !defined(CONFIG_XEN)
8172 - else if (!memcmp(from, "additional_cpus=", 16))
8173 - setup_additional_cpus(from+16);
8174 -#endif
8175 -
8176 - next_char:
8177 - c = *(from++);
8178 - if (!c)
8179 - break;
8180 - if (COMMAND_LINE_SIZE <= ++len)
8181 - break;
8182 - *(to++) = c;
8183 - }
8184 - if (userdef) {
8185 - printk(KERN_INFO "user-defined physical RAM map:\n");
8186 - e820_print_map("user");
8187 - }
8188 - *to = '\0';
8189 - *cmdline_p = command_line;
8190 -}
8191 -
8192 #ifndef CONFIG_NUMA
8193 static void __init
8194 contig_initmem_init(unsigned long start_pfn, unsigned long end_pfn)
8195 @@ -521,10 +334,11 @@ contig_initmem_init(unsigned long start_
8196 if (bootmap == -1L)
8197 panic("Cannot find bootmem map of size %ld\n",bootmap_size);
8198 bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn);
8199 + e820_register_active_regions(0, start_pfn, end_pfn);
8200 #ifdef CONFIG_XEN
8201 - e820_bootmem_free(NODE_DATA(0), 0, xen_start_info->nr_pages<<PAGE_SHIFT);
8202 + free_bootmem_with_active_regions(0, xen_start_info->nr_pages);
8203 #else
8204 - e820_bootmem_free(NODE_DATA(0), 0, end_pfn << PAGE_SHIFT);
8205 + free_bootmem_with_active_regions(0, end_pfn);
8206 #endif
8207 reserve_bootmem(bootmap, bootmap_size);
8208 }
8209 @@ -587,6 +401,10 @@ static void discover_ebda(void)
8210 void __init setup_arch(char **cmdline_p)
8211 {
8212 #ifdef CONFIG_XEN
8213 + extern struct e820map machine_e820;
8214 +
8215 + printk(KERN_INFO "Command line: %s\n", saved_command_line);
8216 +
8217 /* Register a call for panic conditions. */
8218 atomic_notifier_chain_register(&panic_notifier_list, &xen_panic_block);
8219
8220 @@ -612,6 +430,8 @@ void __init setup_arch(char **cmdline_p)
8221
8222 ARCH_SETUP
8223 #else
8224 + printk(KERN_INFO "Command line: %s\n", saved_command_line);
8225 +
8226 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
8227 screen_info = SCREEN_INFO;
8228 edid_info = EDID_INFO;
8229 @@ -639,16 +459,22 @@ void __init setup_arch(char **cmdline_p)
8230 data_resource.start = virt_to_phys(&_etext);
8231 data_resource.end = virt_to_phys(&_edata)-1;
8232
8233 - parse_cmdline_early(cmdline_p);
8234 -
8235 early_identify_cpu(&boot_cpu_data);
8236
8237 + strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE);
8238 + *cmdline_p = command_line;
8239 +
8240 + parse_early_param();
8241 +
8242 + finish_e820_parsing();
8243 +
8244 + e820_register_active_regions(0, 0, -1UL);
8245 /*
8246 * partially used pages are not usable - thus
8247 * we are rounding upwards:
8248 */
8249 end_pfn = e820_end_of_ram();
8250 - num_physpages = end_pfn; /* for pfn_valid */
8251 + num_physpages = end_pfn;
8252
8253 check_efer();
8254
8255 @@ -659,6 +485,14 @@ void __init setup_arch(char **cmdline_p)
8256 if (is_initial_xendomain())
8257 dmi_scan_machine();
8258
8259 + /* How many end-of-memory variables you have, grandma! */
8260 + max_low_pfn = end_pfn;
8261 + max_pfn = end_pfn;
8262 + high_memory = (void *)__va(end_pfn * PAGE_SIZE - 1) + 1;
8263 +
8264 + /* Remove active ranges so rediscovery with NUMA-awareness happens */
8265 + remove_all_active_ranges();
8266 +
8267 #ifdef CONFIG_ACPI_NUMA
8268 /*
8269 * Parse SRAT to discover nodes.
8270 @@ -848,16 +682,16 @@ void __init setup_arch(char **cmdline_p)
8271 BUG();
8272 }
8273
8274 +#ifdef CONFIG_ACPI
8275 if (!is_initial_xendomain()) {
8276 acpi_disabled = 1;
8277 -#ifdef CONFIG_ACPI
8278 acpi_ht = 0;
8279 -#endif
8280 }
8281 #endif
8282 +#endif
8283
8284 -#ifndef CONFIG_XEN
8285 - check_ioapic();
8286 +#if defined(CONFIG_PCI) && !defined(CONFIG_XEN)
8287 + early_quirks();
8288 #endif
8289
8290 zap_low_mappings(0);
8291 @@ -907,6 +741,7 @@ void __init setup_arch(char **cmdline_p)
8292 e820_reserve_resources(machine_e820.map, machine_e820.nr_map);
8293 #else
8294 e820_reserve_resources(e820.map, e820.nr_map);
8295 + e820_mark_nosave_regions();
8296 #endif
8297
8298 request_resource(&iomem_resource, &video_ram_resource);
8299 @@ -914,7 +749,7 @@ void __init setup_arch(char **cmdline_p)
8300 {
8301 unsigned i;
8302 /* request I/O space for devices used on all i[345]86 PCs */
8303 - for (i = 0; i < STANDARD_IO_RESOURCES; i++)
8304 + for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
8305 request_resource(&ioport_resource, &standard_io_resources[i]);
8306 }
8307
8308 @@ -1098,7 +933,7 @@ static void __init amd_detect_cmp(struct
8309 #endif
8310 }
8311
8312 -static void __init init_amd(struct cpuinfo_x86 *c)
8313 +static void __cpuinit init_amd(struct cpuinfo_x86 *c)
8314 {
8315 unsigned level;
8316
8317 @@ -1154,6 +989,12 @@ static void __init init_amd(struct cpuin
8318
8319 /* Fix cpuid4 emulation for more */
8320 num_cache_leaves = 3;
8321 +
8322 + /* When there is only one core no need to synchronize RDTSC */
8323 + if (num_possible_cpus() == 1)
8324 + set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
8325 + else
8326 + clear_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
8327 }
8328
8329 static void __cpuinit detect_ht(struct cpuinfo_x86 *c)
8330 @@ -1235,8 +1076,7 @@ static void srat_detect_node(void)
8331 node = first_node(node_online_map);
8332 numa_set_node(cpu, node);
8333
8334 - if (acpi_numa > 0)
8335 - printk(KERN_INFO "CPU %d/%x -> Node %d\n", cpu, apicid, node);
8336 + printk(KERN_INFO "CPU %d/%x -> Node %d\n", cpu, apicid, node);
8337 #endif
8338 }
8339
8340 @@ -1270,6 +1110,8 @@ static void __cpuinit init_intel(struct
8341 if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
8342 (c->x86 == 0x6 && c->x86_model >= 0x0e))
8343 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
8344 + if (c->x86 == 6)
8345 + set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability);
8346 set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
8347 c->x86_max_cores = intel_num_cpu_cores(c);
8348
8349 @@ -1488,8 +1330,8 @@ static int show_cpuinfo(struct seq_file
8350
8351 /* Intel-defined (#2) */
8352 "pni", NULL, NULL, "monitor", "ds_cpl", "vmx", "smx", "est",
8353 - "tm2", NULL, "cid", NULL, NULL, "cx16", "xtpr", NULL,
8354 - NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
8355 + "tm2", "ssse3", "cid", NULL, NULL, "cx16", "xtpr", NULL,
8356 + NULL, NULL, "dca", NULL, NULL, NULL, NULL, NULL,
8357 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
8358
8359 /* VIA/Cyrix/Centaur-defined */
8360 --- sle11-2009-05-14.orig/arch/x86/kernel/setup64-xen.c 2009-04-20 11:36:10.000000000 +0200
8361 +++ sle11-2009-05-14/arch/x86/kernel/setup64-xen.c 2009-03-04 11:28:34.000000000 +0100
8362 @@ -31,7 +31,7 @@
8363 #include <asm/hypervisor.h>
8364 #endif
8365
8366 -char x86_boot_params[BOOT_PARAM_SIZE] __initdata = {0,};
8367 +char x86_boot_params[BOOT_PARAM_SIZE] __initdata;
8368
8369 cpumask_t cpu_initialized __cpuinitdata = CPU_MASK_NONE;
8370
8371 @@ -55,8 +55,10 @@ Control non executable mappings for 64bi
8372 on Enable(default)
8373 off Disable
8374 */
8375 -int __init nonx_setup(char *str)
8376 +static int __init nonx_setup(char *str)
8377 {
8378 + if (!str)
8379 + return -EINVAL;
8380 if (!strncmp(str, "on", 2)) {
8381 __supported_pte_mask |= _PAGE_NX;
8382 do_not_nx = 0;
8383 @@ -64,9 +66,9 @@ int __init nonx_setup(char *str)
8384 do_not_nx = 1;
8385 __supported_pte_mask &= ~_PAGE_NX;
8386 }
8387 - return 1;
8388 + return 0;
8389 }
8390 -__setup("noexec=", nonx_setup); /* parsed early actually */
8391 +early_param("noexec", nonx_setup);
8392
8393 int force_personality32 = 0;
8394
8395 @@ -102,12 +104,9 @@ void __init setup_per_cpu_areas(void)
8396 #endif
8397
8398 /* Copy section for each CPU (we discard the original) */
8399 - size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
8400 -#ifdef CONFIG_MODULES
8401 - if (size < PERCPU_ENOUGH_ROOM)
8402 - size = PERCPU_ENOUGH_ROOM;
8403 -#endif
8404 + size = PERCPU_ENOUGH_ROOM;
8405
8406 + printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n", size);
8407 for_each_cpu_mask (i, cpu_possible_map) {
8408 char *ptr;
8409
8410 @@ -169,7 +168,10 @@ void pda_init(int cpu)
8411 /* Setup up data that may be needed in __get_free_pages early */
8412 asm volatile("movl %0,%%fs ; movl %0,%%gs" :: "r" (0));
8413 #ifndef CONFIG_XEN
8414 + /* Memory clobbers used to order PDA accessed */
8415 + mb();
8416 wrmsrl(MSR_GS_BASE, pda);
8417 + mb();
8418 #else
8419 if (HYPERVISOR_set_segment_base(SEGBASE_GS_KERNEL,
8420 (unsigned long)pda))
8421 @@ -302,28 +304,17 @@ void __cpuinit cpu_init (void)
8422 * set up and load the per-CPU TSS
8423 */
8424 for (v = 0; v < N_EXCEPTION_STACKS; v++) {
8425 + static const unsigned int order[N_EXCEPTION_STACKS] = {
8426 + [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER,
8427 + [DEBUG_STACK - 1] = DEBUG_STACK_ORDER
8428 + };
8429 if (cpu) {
8430 - static const unsigned int order[N_EXCEPTION_STACKS] = {
8431 - [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER,
8432 - [DEBUG_STACK - 1] = DEBUG_STACK_ORDER
8433 - };
8434 -
8435 estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]);
8436 if (!estacks)
8437 panic("Cannot allocate exception stack %ld %d\n",
8438 v, cpu);
8439 }
8440 - switch (v + 1) {
8441 -#if DEBUG_STKSZ > EXCEPTION_STKSZ
8442 - case DEBUG_STACK:
8443 - cpu_pda(cpu)->debugstack = (unsigned long)estacks;
8444 - estacks += DEBUG_STKSZ;
8445 - break;
8446 -#endif
8447 - default:
8448 - estacks += EXCEPTION_STKSZ;
8449 - break;
8450 - }
8451 + estacks += PAGE_SIZE << order[v];
8452 orig_ist->ist[v] = t->ist[v] = (unsigned long)estacks;
8453 }
8454
8455 --- sle11-2009-05-14.orig/arch/x86/kernel/smp_64-xen.c 2009-05-14 11:02:43.000000000 +0200
8456 +++ sle11-2009-05-14/arch/x86/kernel/smp_64-xen.c 2009-03-04 11:28:34.000000000 +0100
8457 @@ -381,9 +381,8 @@ int smp_call_function_single (int cpu, v
8458 /* prevent preemption and reschedule on another processor */
8459 int me = get_cpu();
8460 if (cpu == me) {
8461 - WARN_ON(1);
8462 put_cpu();
8463 - return -EBUSY;
8464 + return 0;
8465 }
8466 spin_lock_bh(&call_lock);
8467 __smp_call_function_single(cpu, func, info, nonatomic, wait);
8468 @@ -501,7 +500,7 @@ void smp_send_stop(void)
8469 #ifndef CONFIG_XEN
8470 asmlinkage void smp_reschedule_interrupt(void)
8471 #else
8472 -asmlinkage irqreturn_t smp_reschedule_interrupt(void)
8473 +asmlinkage irqreturn_t smp_reschedule_interrupt(int irq, void *ctx)
8474 #endif
8475 {
8476 #ifndef CONFIG_XEN
8477 @@ -514,7 +513,7 @@ asmlinkage irqreturn_t smp_reschedule_in
8478 #ifndef CONFIG_XEN
8479 asmlinkage void smp_call_function_interrupt(void)
8480 #else
8481 -asmlinkage irqreturn_t smp_call_function_interrupt(void)
8482 +asmlinkage irqreturn_t smp_call_function_interrupt(int irq, void *ctx)
8483 #endif
8484 {
8485 void (*func) (void *info) = call_data->func;
8486 @@ -545,31 +544,3 @@ asmlinkage irqreturn_t smp_call_function
8487 return IRQ_HANDLED;
8488 #endif
8489 }
8490 -
8491 -int safe_smp_processor_id(void)
8492 -{
8493 -#ifdef CONFIG_XEN
8494 - return smp_processor_id();
8495 -#else
8496 - unsigned apicid, i;
8497 -
8498 - if (disable_apic)
8499 - return 0;
8500 -
8501 - apicid = hard_smp_processor_id();
8502 - if (apicid < NR_CPUS && x86_cpu_to_apicid[apicid] == apicid)
8503 - return apicid;
8504 -
8505 - for (i = 0; i < NR_CPUS; ++i) {
8506 - if (x86_cpu_to_apicid[i] == apicid)
8507 - return i;
8508 - }
8509 -
8510 - /* No entries in x86_cpu_to_apicid? Either no MPS|ACPI,
8511 - * or called too early. Either way, we must be CPU 0. */
8512 - if (x86_cpu_to_apicid[0] == BAD_APICID)
8513 - return 0;
8514 -
8515 - return 0; /* Should not happen */
8516 -#endif
8517 -}
8518 --- sle11-2009-05-14.orig/arch/x86/kernel/traps_64-xen.c 2009-05-14 11:02:43.000000000 +0200
8519 +++ sle11-2009-05-14/arch/x86/kernel/traps_64-xen.c 2009-03-04 11:28:34.000000000 +0100
8520 @@ -23,6 +23,7 @@
8521 #include <linux/delay.h>
8522 #include <linux/spinlock.h>
8523 #include <linux/interrupt.h>
8524 +#include <linux/kallsyms.h>
8525 #include <linux/module.h>
8526 #include <linux/moduleparam.h>
8527 #include <linux/nmi.h>
8528 @@ -45,6 +46,7 @@
8529 #include <asm/pda.h>
8530 #include <asm/proto.h>
8531 #include <asm/nmi.h>
8532 +#include <asm/stacktrace.h>
8533
8534 asmlinkage void divide_error(void);
8535 asmlinkage void debug(void);
8536 @@ -114,7 +116,6 @@ static int call_trace = 1;
8537 #endif
8538
8539 #ifdef CONFIG_KALLSYMS
8540 -# include <linux/kallsyms.h>
8541 void printk_address(unsigned long address)
8542 {
8543 unsigned long offset = 0, symsize;
8544 @@ -142,7 +143,7 @@ void printk_address(unsigned long addres
8545 #endif
8546
8547 static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
8548 - unsigned *usedp, const char **idp)
8549 + unsigned *usedp, char **idp)
8550 {
8551 #ifndef CONFIG_X86_NO_TSS
8552 static char ids[][8] = {
8553 @@ -162,26 +163,7 @@ static unsigned long *in_exception_stack
8554 * 'stack' is in one of them:
8555 */
8556 for (k = 0; k < N_EXCEPTION_STACKS; k++) {
8557 - unsigned long end;
8558 -
8559 - /*
8560 - * set 'end' to the end of the exception stack.
8561 - */
8562 - switch (k + 1) {
8563 - /*
8564 - * TODO: this block is not needed i think, because
8565 - * setup64.c:cpu_init() sets up t->ist[DEBUG_STACK]
8566 - * properly too.
8567 - */
8568 -#if DEBUG_STKSZ > EXCEPTION_STKSZ
8569 - case DEBUG_STACK:
8570 - end = cpu_pda(cpu)->debugstack + DEBUG_STKSZ;
8571 - break;
8572 -#endif
8573 - default:
8574 - end = per_cpu(orig_ist, cpu).ist[k];
8575 - break;
8576 - }
8577 + unsigned long end = per_cpu(orig_ist, cpu).ist[k];
8578 /*
8579 * Is 'stack' above this exception frame's end?
8580 * If yes then skip to the next frame.
8581 @@ -236,13 +218,19 @@ static unsigned long *in_exception_stack
8582 return NULL;
8583 }
8584
8585 -static int show_trace_unwind(struct unwind_frame_info *info, void *context)
8586 +struct ops_and_data {
8587 + struct stacktrace_ops *ops;
8588 + void *data;
8589 +};
8590 +
8591 +static int dump_trace_unwind(struct unwind_frame_info *info, void *context)
8592 {
8593 + struct ops_and_data *oad = (struct ops_and_data *)context;
8594 int n = 0;
8595
8596 while (unwind(info) == 0 && UNW_PC(info)) {
8597 n++;
8598 - printk_address(UNW_PC(info));
8599 + oad->ops->address(oad->data, UNW_PC(info));
8600 if (arch_unw_user_mode(info))
8601 break;
8602 }
8603 @@ -256,13 +244,19 @@ static int show_trace_unwind(struct unwi
8604 * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack
8605 */
8606
8607 -void show_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long * stack)
8608 +static inline int valid_stack_ptr(struct thread_info *tinfo, void *p)
8609 {
8610 - const unsigned cpu = safe_smp_processor_id();
8611 + void *t = (void *)tinfo;
8612 + return p > t && p < t + THREAD_SIZE - 3;
8613 +}
8614 +
8615 +void dump_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long * stack,
8616 + struct stacktrace_ops *ops, void *data)
8617 +{
8618 + const unsigned cpu = smp_processor_id();
8619 unsigned long *irqstack_end = (unsigned long *)cpu_pda(cpu)->irqstackptr;
8620 unsigned used = 0;
8621 -
8622 - printk("\nCall Trace:\n");
8623 + struct thread_info *tinfo;
8624
8625 if (!tsk)
8626 tsk = current;
8627 @@ -270,32 +264,47 @@ void show_trace(struct task_struct *tsk,
8628 if (call_trace >= 0) {
8629 int unw_ret = 0;
8630 struct unwind_frame_info info;
8631 + struct ops_and_data oad = { .ops = ops, .data = data };
8632
8633 if (regs) {
8634 if (unwind_init_frame_info(&info, tsk, regs) == 0)
8635 - unw_ret = show_trace_unwind(&info, NULL);
8636 + unw_ret = dump_trace_unwind(&info, &oad);
8637 } else if (tsk == current)
8638 - unw_ret = unwind_init_running(&info, show_trace_unwind, NULL);
8639 + unw_ret = unwind_init_running(&info, dump_trace_unwind, &oad);
8640 else {
8641 if (unwind_init_blocked(&info, tsk) == 0)
8642 - unw_ret = show_trace_unwind(&info, NULL);
8643 + unw_ret = dump_trace_unwind(&info, &oad);
8644 }
8645 if (unw_ret > 0) {
8646 if (call_trace == 1 && !arch_unw_user_mode(&info)) {
8647 - print_symbol("DWARF2 unwinder stuck at %s\n",
8648 + ops->warning_symbol(data, "DWARF2 unwinder stuck at %s\n",
8649 UNW_PC(&info));
8650 if ((long)UNW_SP(&info) < 0) {
8651 - printk("Leftover inexact backtrace:\n");
8652 + ops->warning(data, "Leftover inexact backtrace:\n");
8653 stack = (unsigned long *)UNW_SP(&info);
8654 + if (!stack)
8655 + return;
8656 } else
8657 - printk("Full inexact backtrace again:\n");
8658 + ops->warning(data, "Full inexact backtrace again:\n");
8659 } else if (call_trace >= 1)
8660 return;
8661 else
8662 - printk("Full inexact backtrace again:\n");
8663 + ops->warning(data, "Full inexact backtrace again:\n");
8664 } else
8665 - printk("Inexact backtrace:\n");
8666 + ops->warning(data, "Inexact backtrace:\n");
8667 }
8668 + if (!stack) {
8669 + unsigned long dummy;
8670 + stack = &dummy;
8671 + if (tsk && tsk != current)
8672 + stack = (unsigned long *)tsk->thread.rsp;
8673 + }
8674 + /*
8675 + * Align the stack pointer on word boundary, later loops
8676 + * rely on that (and corruption / debug info bugs can cause
8677 + * unaligned values here):
8678 + */
8679 + stack = (unsigned long *)((unsigned long)stack & ~(sizeof(long)-1));
8680
8681 /*
8682 * Print function call entries within a stack. 'cond' is the
8683 @@ -305,7 +314,9 @@ void show_trace(struct task_struct *tsk,
8684 #define HANDLE_STACK(cond) \
8685 do while (cond) { \
8686 unsigned long addr = *stack++; \
8687 - if (kernel_text_address(addr)) { \
8688 + if (oops_in_progress ? \
8689 + __kernel_text_address(addr) : \
8690 + kernel_text_address(addr)) { \
8691 /* \
8692 * If the address is either in the text segment of the \
8693 * kernel, or in the region which contains vmalloc'ed \
8694 @@ -314,7 +325,7 @@ void show_trace(struct task_struct *tsk,
8695 * down the cause of the crash will be able to figure \
8696 * out the call path that was taken. \
8697 */ \
8698 - printk_address(addr); \
8699 + ops->address(data, addr); \
8700 } \
8701 } while (0)
8702
8703 @@ -323,16 +334,17 @@ void show_trace(struct task_struct *tsk,
8704 * current stack address. If the stacks consist of nested
8705 * exceptions
8706 */
8707 - for ( ; ; ) {
8708 - const char *id;
8709 + for (;;) {
8710 + char *id;
8711 unsigned long *estack_end;
8712 estack_end = in_exception_stack(cpu, (unsigned long)stack,
8713 &used, &id);
8714
8715 if (estack_end) {
8716 - printk(" <%s>", id);
8717 + if (ops->stack(data, id) < 0)
8718 + break;
8719 HANDLE_STACK (stack < estack_end);
8720 - printk(" <EOE>");
8721 + ops->stack(data, "<EOE>");
8722 /*
8723 * We link to the next stack via the
8724 * second-to-last pointer (index -2 to end) in the
8725 @@ -347,7 +359,8 @@ void show_trace(struct task_struct *tsk,
8726 (IRQSTACKSIZE - 64) / sizeof(*irqstack);
8727
8728 if (stack >= irqstack && stack < irqstack_end) {
8729 - printk(" <IRQ>");
8730 + if (ops->stack(data, "IRQ") < 0)
8731 + break;
8732 HANDLE_STACK (stack < irqstack_end);
8733 /*
8734 * We link to the next stack (which would be
8735 @@ -356,7 +369,7 @@ void show_trace(struct task_struct *tsk,
8736 */
8737 stack = (unsigned long *) (irqstack_end[-1]);
8738 irqstack_end = NULL;
8739 - printk(" <EOI>");
8740 + ops->stack(data, "EOI");
8741 continue;
8742 }
8743 }
8744 @@ -364,19 +377,58 @@ void show_trace(struct task_struct *tsk,
8745 }
8746
8747 /*
8748 - * This prints the process stack:
8749 + * This handles the process stack:
8750 */
8751 - HANDLE_STACK (((long) stack & (THREAD_SIZE-1)) != 0);
8752 + tinfo = current_thread_info();
8753 + HANDLE_STACK (valid_stack_ptr(tinfo, stack));
8754 #undef HANDLE_STACK
8755 +}
8756 +EXPORT_SYMBOL(dump_trace);
8757
8758 +static void
8759 +print_trace_warning_symbol(void *data, char *msg, unsigned long symbol)
8760 +{
8761 + print_symbol(msg, symbol);
8762 printk("\n");
8763 }
8764
8765 -static void _show_stack(struct task_struct *tsk, struct pt_regs *regs, unsigned long * rsp)
8766 +static void print_trace_warning(void *data, char *msg)
8767 +{
8768 + printk("%s\n", msg);
8769 +}
8770 +
8771 +static int print_trace_stack(void *data, char *name)
8772 +{
8773 + printk(" <%s> ", name);
8774 + return 0;
8775 +}
8776 +
8777 +static void print_trace_address(void *data, unsigned long addr)
8778 +{
8779 + printk_address(addr);
8780 +}
8781 +
8782 +static struct stacktrace_ops print_trace_ops = {
8783 + .warning = print_trace_warning,
8784 + .warning_symbol = print_trace_warning_symbol,
8785 + .stack = print_trace_stack,
8786 + .address = print_trace_address,
8787 +};
8788 +
8789 +void
8790 +show_trace(struct task_struct *tsk, struct pt_regs *regs, unsigned long *stack)
8791 +{
8792 + printk("\nCall Trace:\n");
8793 + dump_trace(tsk, regs, stack, &print_trace_ops, NULL);
8794 + printk("\n");
8795 +}
8796 +
8797 +static void
8798 +_show_stack(struct task_struct *tsk, struct pt_regs *regs, unsigned long *rsp)
8799 {
8800 unsigned long *stack;
8801 int i;
8802 - const int cpu = safe_smp_processor_id();
8803 + const int cpu = smp_processor_id();
8804 unsigned long *irqstack_end = (unsigned long *) (cpu_pda(cpu)->irqstackptr);
8805 unsigned long *irqstack = (unsigned long *) (cpu_pda(cpu)->irqstackptr - IRQSTACKSIZE);
8806
8807 @@ -430,7 +482,7 @@ void show_registers(struct pt_regs *regs
8808 int i;
8809 int in_kernel = !user_mode(regs);
8810 unsigned long rsp;
8811 - const int cpu = safe_smp_processor_id();
8812 + const int cpu = smp_processor_id();
8813 struct task_struct *cur = cpu_pda(cpu)->pcurrent;
8814
8815 rsp = regs->rsp;
8816 @@ -505,9 +557,11 @@ static unsigned int die_nest_count;
8817
8818 unsigned __kprobes long oops_begin(void)
8819 {
8820 - int cpu = safe_smp_processor_id();
8821 + int cpu = smp_processor_id();
8822 unsigned long flags;
8823
8824 + oops_enter();
8825 +
8826 /* racy, but better than risking deadlock. */
8827 local_irq_save(flags);
8828 if (!spin_trylock(&die_lock)) {
8829 @@ -536,6 +590,7 @@ void __kprobes oops_end(unsigned long fl
8830 spin_unlock_irqrestore(&die_lock, flags);
8831 if (panic_on_oops)
8832 panic("Fatal exception");
8833 + oops_exit();
8834 }
8835
8836 void __kprobes __die(const char * str, struct pt_regs * regs, long err)
8837 @@ -572,8 +627,8 @@ void die(const char * str, struct pt_reg
8838 do_exit(SIGSEGV);
8839 }
8840
8841 -#ifdef CONFIG_X86_LOCAL_APIC
8842 -void __kprobes die_nmi(char *str, struct pt_regs *regs)
8843 +#if defined(CONFIG_X86_LOCAL_APIC) || defined(CONFIG_SYSCTL)
8844 +void __kprobes die_nmi(char *str, struct pt_regs *regs, int do_panic)
8845 {
8846 unsigned long flags = oops_begin();
8847
8848 @@ -581,13 +636,12 @@ void __kprobes die_nmi(char *str, struct
8849 * We are in trouble anyway, lets at least try
8850 * to get a message out.
8851 */
8852 - printk(str, safe_smp_processor_id());
8853 + printk(str, smp_processor_id());
8854 show_registers(regs);
8855 if (kexec_should_crash(current))
8856 crash_kexec(regs);
8857 - if (panic_on_timeout || panic_on_oops)
8858 - panic("nmi watchdog");
8859 - printk("console shuts up ...\n");
8860 + if (do_panic || panic_on_oops)
8861 + panic("Non maskable interrupt");
8862 oops_end(flags);
8863 nmi_exit();
8864 local_irq_enable();
8865 @@ -734,8 +788,15 @@ asmlinkage void __kprobes do_general_pro
8866 static __kprobes void
8867 mem_parity_error(unsigned char reason, struct pt_regs * regs)
8868 {
8869 - printk("Uhhuh. NMI received. Dazed and confused, but trying to continue\n");
8870 - printk("You probably have a hardware problem with your RAM chips\n");
8871 + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x.\n",
8872 + reason);
8873 + printk(KERN_EMERG "You probably have a hardware problem with your "
8874 + "RAM chips\n");
8875 +
8876 + if (panic_on_unrecovered_nmi)
8877 + panic("NMI: Not continuing");
8878 +
8879 + printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
8880
8881 #if 0 /* XEN */
8882 /* Clear and disable the memory parity error line. */
8883 @@ -762,9 +823,15 @@ io_check_error(unsigned char reason, str
8884
8885 static __kprobes void
8886 unknown_nmi_error(unsigned char reason, struct pt_regs * regs)
8887 -{ printk("Uhhuh. NMI received for unknown reason %02x.\n", reason);
8888 - printk("Dazed and confused, but trying to continue\n");
8889 - printk("Do you have a strange power saving mode enabled?\n");
8890 +{
8891 + printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x.\n",
8892 + reason);
8893 + printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n");
8894 +
8895 + if (panic_on_unrecovered_nmi)
8896 + panic("NMI: Not continuing");
8897 +
8898 + printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
8899 }
8900
8901 /* Runs on IST stack. This code must keep interrupts off all the time.
8902 @@ -789,12 +856,12 @@ asmlinkage __kprobes void default_do_nmi
8903 * Ok, so this is none of the documented NMI sources,
8904 * so it must be the NMI watchdog.
8905 */
8906 - if (nmi_watchdog > 0) {
8907 - nmi_watchdog_tick(regs,reason);
8908 + if (nmi_watchdog_tick(regs,reason))
8909 return;
8910 - }
8911 #endif
8912 - unknown_nmi_error(reason, regs);
8913 + if (!do_nmi_callback(regs,cpu))
8914 + unknown_nmi_error(reason, regs);
8915 +
8916 return;
8917 }
8918 if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
8919 @@ -1081,6 +1148,7 @@ asmlinkage void math_state_restore(void)
8920 init_fpu(me);
8921 restore_fpu_checking(&me->thread.i387.fxsave);
8922 task_thread_info(me)->status |= TS_USEDFPU;
8923 + me->fpu_counter++;
8924 }
8925
8926
8927 @@ -1141,24 +1209,30 @@ void __cpuinit smp_trap_init(trap_info_t
8928 }
8929
8930
8931 -/* Actual parsing is done early in setup.c. */
8932 -static int __init oops_dummy(char *s)
8933 +static int __init oops_setup(char *s)
8934 {
8935 - panic_on_oops = 1;
8936 - return 1;
8937 + if (!s)
8938 + return -EINVAL;
8939 + if (!strcmp(s, "panic"))
8940 + panic_on_oops = 1;
8941 + return 0;
8942 }
8943 -__setup("oops=", oops_dummy);
8944 +early_param("oops", oops_setup);
8945
8946 static int __init kstack_setup(char *s)
8947 {
8948 + if (!s)
8949 + return -EINVAL;
8950 kstack_depth_to_print = simple_strtoul(s,NULL,0);
8951 - return 1;
8952 + return 0;
8953 }
8954 -__setup("kstack=", kstack_setup);
8955 +early_param("kstack", kstack_setup);
8956
8957 #ifdef CONFIG_STACK_UNWIND
8958 static int __init call_trace_setup(char *s)
8959 {
8960 + if (!s)
8961 + return -EINVAL;
8962 if (strcmp(s, "old") == 0)
8963 call_trace = -1;
8964 else if (strcmp(s, "both") == 0)
8965 @@ -1167,7 +1241,7 @@ static int __init call_trace_setup(char
8966 call_trace = 1;
8967 else if (strcmp(s, "new") == 0)
8968 call_trace = 2;
8969 - return 1;
8970 + return 0;
8971 }
8972 -__setup("call_trace=", call_trace_setup);
8973 +early_param("call_trace", call_trace_setup);
8974 #endif
8975 --- sle11-2009-05-14.orig/arch/x86/kernel/vsyscall_64-xen.c 2009-05-14 11:02:43.000000000 +0200
8976 +++ sle11-2009-05-14/arch/x86/kernel/vsyscall_64-xen.c 2009-03-04 11:28:34.000000000 +0100
8977 @@ -26,6 +26,10 @@
8978 #include <linux/seqlock.h>
8979 #include <linux/jiffies.h>
8980 #include <linux/sysctl.h>
8981 +#include <linux/getcpu.h>
8982 +#include <linux/cpu.h>
8983 +#include <linux/smp.h>
8984 +#include <linux/notifier.h>
8985
8986 #include <asm/vsyscall.h>
8987 #include <asm/pgtable.h>
8988 @@ -33,11 +37,15 @@
8989 #include <asm/fixmap.h>
8990 #include <asm/errno.h>
8991 #include <asm/io.h>
8992 +#include <asm/segment.h>
8993 +#include <asm/desc.h>
8994 +#include <asm/topology.h>
8995
8996 #define __vsyscall(nr) __attribute__ ((unused,__section__(".vsyscall_" #nr)))
8997
8998 int __sysctl_vsyscall __section_sysctl_vsyscall = 1;
8999 seqlock_t __xtime_lock __section_xtime_lock = SEQLOCK_UNLOCKED;
9000 +int __vgetcpu_mode __section_vgetcpu_mode;
9001
9002 #include <asm/unistd.h>
9003
9004 @@ -61,8 +69,7 @@ static __always_inline void do_vgettimeo
9005 sequence = read_seqbegin(&__xtime_lock);
9006
9007 sec = __xtime.tv_sec;
9008 - usec = (__xtime.tv_nsec / 1000) +
9009 - (__jiffies - __wall_jiffies) * (1000000 / HZ);
9010 + usec = __xtime.tv_nsec / 1000;
9011
9012 if (__vxtime.mode != VXTIME_HPET) {
9013 t = get_cycles_sync();
9014 @@ -72,7 +79,8 @@ static __always_inline void do_vgettimeo
9015 __vxtime.tsc_quot) >> 32;
9016 /* See comment in x86_64 do_gettimeofday. */
9017 } else {
9018 - usec += ((readl((void *)fix_to_virt(VSYSCALL_HPET) + 0xf0) -
9019 + usec += ((readl((void __iomem *)
9020 + fix_to_virt(VSYSCALL_HPET) + 0xf0) -
9021 __vxtime.last) * __vxtime.quot) >> 32;
9022 }
9023 } while (read_seqretry(&__xtime_lock, sequence));
9024 @@ -127,9 +135,46 @@ time_t __vsyscall(1) vtime(time_t *t)
9025 return __xtime.tv_sec;
9026 }
9027
9028 -long __vsyscall(2) venosys_0(void)
9029 -{
9030 - return -ENOSYS;
9031 +/* Fast way to get current CPU and node.
9032 + This helps to do per node and per CPU caches in user space.
9033 + The result is not guaranteed without CPU affinity, but usually
9034 + works out because the scheduler tries to keep a thread on the same
9035 + CPU.
9036 +
9037 + tcache must point to a two element sized long array.
9038 + All arguments can be NULL. */
9039 +long __vsyscall(2)
9040 +vgetcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
9041 +{
9042 + unsigned int dummy, p;
9043 + unsigned long j = 0;
9044 +
9045 + /* Fast cache - only recompute value once per jiffies and avoid
9046 + relatively costly rdtscp/cpuid otherwise.
9047 + This works because the scheduler usually keeps the process
9048 + on the same CPU and this syscall doesn't guarantee its
9049 + results anyways.
9050 + We do this here because otherwise user space would do it on
9051 + its own in a likely inferior way (no access to jiffies).
9052 + If you don't like it pass NULL. */
9053 + if (tcache && tcache->blob[0] == (j = __jiffies)) {
9054 + p = tcache->blob[1];
9055 + } else if (__vgetcpu_mode == VGETCPU_RDTSCP) {
9056 + /* Load per CPU data from RDTSCP */
9057 + rdtscp(dummy, dummy, p);
9058 + } else {
9059 + /* Load per CPU data from GDT */
9060 + asm("lsl %1,%0" : "=r" (p) : "r" (__PER_CPU_SEG));
9061 + }
9062 + if (tcache) {
9063 + tcache->blob[0] = j;
9064 + tcache->blob[1] = p;
9065 + }
9066 + if (cpu)
9067 + *cpu = p & 0xfff;
9068 + if (node)
9069 + *node = p >> 12;
9070 + return 0;
9071 }
9072
9073 long __vsyscall(3) venosys_1(void)
9074 @@ -149,7 +194,8 @@ static int vsyscall_sysctl_change(ctl_ta
9075 void __user *buffer, size_t *lenp, loff_t *ppos)
9076 {
9077 extern u16 vsysc1, vsysc2;
9078 - u16 *map1, *map2;
9079 + u16 __iomem *map1;
9080 + u16 __iomem *map2;
9081 int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
9082 if (!write)
9083 return ret;
9084 @@ -164,11 +210,11 @@ static int vsyscall_sysctl_change(ctl_ta
9085 goto out;
9086 }
9087 if (!sysctl_vsyscall) {
9088 - *map1 = SYSCALL;
9089 - *map2 = SYSCALL;
9090 + writew(SYSCALL, map1);
9091 + writew(SYSCALL, map2);
9092 } else {
9093 - *map1 = NOP2;
9094 - *map2 = NOP2;
9095 + writew(NOP2, map1);
9096 + writew(NOP2, map2);
9097 }
9098 iounmap(map2);
9099 out:
9100 @@ -200,6 +246,48 @@ static ctl_table kernel_root_table2[] =
9101
9102 #endif
9103
9104 +/* Assume __initcall executes before all user space. Hopefully kmod
9105 + doesn't violate that. We'll find out if it does. */
9106 +static void __cpuinit vsyscall_set_cpu(int cpu)
9107 +{
9108 + unsigned long d;
9109 + unsigned long node = 0;
9110 +#ifdef CONFIG_NUMA
9111 + node = cpu_to_node[cpu];
9112 +#endif
9113 + if (cpu_has(&cpu_data[cpu], X86_FEATURE_RDTSCP))
9114 + write_rdtscp_aux((node << 12) | cpu);
9115 +
9116 + /* Store cpu number in limit so that it can be loaded quickly
9117 + in user space in vgetcpu.
9118 + 12 bits for the CPU and 8 bits for the node. */
9119 + d = 0x0f40000000000ULL;
9120 + d |= cpu;
9121 + d |= (node & 0xf) << 12;
9122 + d |= (node >> 4) << 48;
9123 + if (HYPERVISOR_update_descriptor(virt_to_machine(cpu_gdt(cpu)
9124 + + GDT_ENTRY_PER_CPU),
9125 + d))
9126 + BUG();
9127 +}
9128 +
9129 +static void __cpuinit cpu_vsyscall_init(void *arg)
9130 +{
9131 + /* preemption should be already off */
9132 + vsyscall_set_cpu(raw_smp_processor_id());
9133 +}
9134 +
9135 +#ifdef CONFIG_HOTPLUG_CPU
9136 +static int __cpuinit
9137 +cpu_vsyscall_notifier(struct notifier_block *n, unsigned long action, void *arg)
9138 +{
9139 + long cpu = (long)arg;
9140 + if (action == CPU_ONLINE)
9141 + smp_call_function_single(cpu, cpu_vsyscall_init, NULL, 0, 1);
9142 + return NOTIFY_DONE;
9143 +}
9144 +#endif
9145 +
9146 static void __init map_vsyscall(void)
9147 {
9148 extern char __vsyscall_0;
9149 @@ -214,13 +302,20 @@ static int __init vsyscall_init(void)
9150 VSYSCALL_ADDR(__NR_vgettimeofday)));
9151 BUG_ON((unsigned long) &vtime != VSYSCALL_ADDR(__NR_vtime));
9152 BUG_ON((VSYSCALL_ADDR(0) != __fix_to_virt(VSYSCALL_FIRST_PAGE)));
9153 + BUG_ON((unsigned long) &vgetcpu != VSYSCALL_ADDR(__NR_vgetcpu));
9154 map_vsyscall();
9155 #ifdef CONFIG_XEN
9156 sysctl_vsyscall = 0; /* disable vgettimeofay() */
9157 + if (boot_cpu_has(X86_FEATURE_RDTSCP))
9158 + vgetcpu_mode = VGETCPU_RDTSCP;
9159 + else
9160 + vgetcpu_mode = VGETCPU_LSL;
9161 #endif
9162 #ifdef CONFIG_SYSCTL
9163 register_sysctl_table(kernel_root_table2, 0);
9164 #endif
9165 + on_each_cpu(cpu_vsyscall_init, NULL, 0, 1);
9166 + hotcpu_notifier(cpu_vsyscall_notifier, 0);
9167 return 0;
9168 }
9169
9170 --- sle11-2009-05-14.orig/arch/x86/mm/fault_64-xen.c 2009-05-14 11:02:43.000000000 +0200
9171 +++ sle11-2009-05-14/arch/x86/mm/fault_64-xen.c 2009-03-04 11:28:34.000000000 +0100
9172 @@ -40,8 +40,7 @@
9173 #define PF_RSVD (1<<3)
9174 #define PF_INSTR (1<<4)
9175
9176 -#ifdef CONFIG_KPROBES
9177 -ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
9178 +static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
9179
9180 /* Hook to register for page fault notifications */
9181 int register_page_fault_notifier(struct notifier_block *nb)
9182 @@ -49,11 +48,13 @@ int register_page_fault_notifier(struct
9183 vmalloc_sync_all();
9184 return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
9185 }
9186 +EXPORT_SYMBOL_GPL(register_page_fault_notifier);
9187
9188 int unregister_page_fault_notifier(struct notifier_block *nb)
9189 {
9190 return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
9191 }
9192 +EXPORT_SYMBOL_GPL(unregister_page_fault_notifier);
9193
9194 static inline int notify_page_fault(enum die_val val, const char *str,
9195 struct pt_regs *regs, long err, int trap, int sig)
9196 @@ -67,13 +68,6 @@ static inline int notify_page_fault(enum
9197 };
9198 return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
9199 }
9200 -#else
9201 -static inline int notify_page_fault(enum die_val val, const char *str,
9202 - struct pt_regs *regs, long err, int trap, int sig)
9203 -{
9204 - return NOTIFY_DONE;
9205 -}
9206 -#endif
9207
9208 void bust_spinlocks(int yes)
9209 {
9210 @@ -102,7 +96,7 @@ void bust_spinlocks(int yes)
9211 static noinline int is_prefetch(struct pt_regs *regs, unsigned long addr,
9212 unsigned long error_code)
9213 {
9214 - unsigned char *instr;
9215 + unsigned char __user *instr;
9216 int scan_more = 1;
9217 int prefetch = 0;
9218 unsigned char *max_instr;
9219 @@ -111,7 +105,7 @@ static noinline int is_prefetch(struct p
9220 if (error_code & PF_INSTR)
9221 return 0;
9222
9223 - instr = (unsigned char *)convert_rip_to_linear(current, regs);
9224 + instr = (unsigned char __user *)convert_rip_to_linear(current, regs);
9225 max_instr = instr + 15;
9226
9227 if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
9228 @@ -122,7 +116,7 @@ static noinline int is_prefetch(struct p
9229 unsigned char instr_hi;
9230 unsigned char instr_lo;
9231
9232 - if (__get_user(opcode, instr))
9233 + if (__get_user(opcode, (char __user *)instr))
9234 break;
9235
9236 instr_hi = opcode & 0xf0;
9237 @@ -160,7 +154,7 @@ static noinline int is_prefetch(struct p
9238 case 0x00:
9239 /* Prefetch instruction is 0x0F0D or 0x0F18 */
9240 scan_more = 0;
9241 - if (__get_user(opcode, instr))
9242 + if (__get_user(opcode, (char __user *)instr))
9243 break;
9244 prefetch = (instr_lo == 0xF) &&
9245 (opcode == 0x0D || opcode == 0x18);
9246 @@ -176,7 +170,7 @@ static noinline int is_prefetch(struct p
9247 static int bad_address(void *p)
9248 {
9249 unsigned long dummy;
9250 - return __get_user(dummy, (unsigned long *)p);
9251 + return __get_user(dummy, (unsigned long __user *)p);
9252 }
9253
9254 void dump_pagetable(unsigned long address)
9255 @@ -248,7 +242,7 @@ static int is_errata93(struct pt_regs *r
9256
9257 int unhandled_signal(struct task_struct *tsk, int sig)
9258 {
9259 - if (tsk->pid == 1)
9260 + if (is_init(tsk))
9261 return 1;
9262 if (tsk->ptrace & PT_PTRACED)
9263 return 0;
9264 @@ -300,7 +294,7 @@ static int vmalloc_fault(unsigned long a
9265 if (pgd_none(*pgd))
9266 set_pgd(pgd, *pgd_ref);
9267 else
9268 - BUG_ON(pgd_page(*pgd) != pgd_page(*pgd_ref));
9269 + BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
9270
9271 /* Below here mismatches are bugs because these lower tables
9272 are shared */
9273 @@ -309,7 +303,7 @@ static int vmalloc_fault(unsigned long a
9274 pud_ref = pud_offset(pgd_ref, address);
9275 if (pud_none(*pud_ref))
9276 return -1;
9277 - if (pud_none(*pud) || pud_page(*pud) != pud_page(*pud_ref))
9278 + if (pud_none(*pud) || pud_page_vaddr(*pud) != pud_page_vaddr(*pud_ref))
9279 BUG();
9280 pmd = pmd_offset(pud, address);
9281 pmd_ref = pmd_offset(pud_ref, address);
9282 @@ -531,7 +525,7 @@ good_area:
9283 case PF_PROT: /* read, present */
9284 goto bad_area;
9285 case 0: /* read, not present */
9286 - if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
9287 + if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
9288 goto bad_area;
9289 }
9290
9291 @@ -647,7 +641,7 @@ no_context:
9292 */
9293 out_of_memory:
9294 up_read(&mm->mmap_sem);
9295 - if (current->pid == 1) {
9296 + if (is_init(current)) {
9297 yield();
9298 goto again;
9299 }
9300 @@ -702,7 +696,7 @@ void vmalloc_sync_all(void)
9301 if (pgd_none(*pgd))
9302 set_pgd(pgd, *pgd_ref);
9303 else
9304 - BUG_ON(pgd_page(*pgd) != pgd_page(*pgd_ref));
9305 + BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
9306 }
9307 spin_unlock(&pgd_lock);
9308 set_bit(pgd_index(address), insync);
9309 --- sle11-2009-05-14.orig/arch/x86/mm/init_64-xen.c 2009-05-14 11:02:43.000000000 +0200
9310 +++ sle11-2009-05-14/arch/x86/mm/init_64-xen.c 2009-03-04 11:28:34.000000000 +0100
9311 @@ -61,8 +61,6 @@ EXPORT_SYMBOL(__kernel_page_user);
9312
9313 int after_bootmem;
9314
9315 -static unsigned long dma_reserve __initdata;
9316 -
9317 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
9318 extern unsigned long start_pfn;
9319
9320 @@ -420,7 +418,6 @@ __init void *early_ioremap(unsigned long
9321
9322 /* actually usually some more */
9323 if (size >= LARGE_PAGE_SIZE) {
9324 - printk("SMBIOS area too long %lu\n", size);
9325 return NULL;
9326 }
9327 set_pmd(temp_mappings[0].pmd, __pmd(map | _KERNPG_TABLE | _PAGE_PSE));
9328 @@ -442,16 +439,24 @@ __init void early_iounmap(void *addr, un
9329 #endif
9330
9331 static void __meminit
9332 -phys_pmd_init(pmd_t *pmd, unsigned long address, unsigned long end)
9333 +phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
9334 {
9335 - int i, k;
9336 + int i = pmd_index(address);
9337
9338 - for (i = 0; i < PTRS_PER_PMD; pmd++, i++) {
9339 + for (; i < PTRS_PER_PMD; i++) {
9340 unsigned long pte_phys;
9341 + pmd_t *pmd = pmd_page + i;
9342 pte_t *pte, *pte_save;
9343 + int k;
9344
9345 if (address >= end)
9346 break;
9347 +
9348 + if (__pmd_val(*pmd)) {
9349 + address += PMD_SIZE;
9350 + continue;
9351 + }
9352 +
9353 pte = alloc_static_page(&pte_phys);
9354 pte_save = pte;
9355 for (k = 0; k < PTRS_PER_PTE; pte++, k++, address += PTE_SIZE) {
9356 @@ -478,40 +483,35 @@ phys_pmd_init(pmd_t *pmd, unsigned long
9357 static void __meminit
9358 phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end)
9359 {
9360 - pmd_t *pmd = pmd_offset(pud, (unsigned long)__va(address));
9361 -
9362 - if (pmd_none(*pmd)) {
9363 - spin_lock(&init_mm.page_table_lock);
9364 - phys_pmd_init(pmd, address, end);
9365 - spin_unlock(&init_mm.page_table_lock);
9366 - __flush_tlb_all();
9367 - }
9368 + pmd_t *pmd = pmd_offset(pud,0);
9369 + spin_lock(&init_mm.page_table_lock);
9370 + phys_pmd_init(pmd, address, end);
9371 + spin_unlock(&init_mm.page_table_lock);
9372 + __flush_tlb_all();
9373 }
9374
9375 -static void __meminit phys_pud_init(pud_t *pud, unsigned long address, unsigned long end)
9376 +static void __meminit phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
9377 {
9378 - long i = pud_index(address);
9379 -
9380 - pud = pud + i;
9381 -
9382 - if (after_bootmem && pud_val(*pud)) {
9383 - phys_pmd_update(pud, address, end);
9384 - return;
9385 - }
9386 + int i = pud_index(addr);
9387
9388 - for (; i < PTRS_PER_PUD; pud++, i++) {
9389 - unsigned long paddr, pmd_phys;
9390 + for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE ) {
9391 + unsigned long pmd_phys;
9392 + pud_t *pud = pud_page + pud_index(addr);
9393 pmd_t *pmd;
9394
9395 - paddr = (address & PGDIR_MASK) + i*PUD_SIZE;
9396 - if (paddr >= end)
9397 + if (addr >= end)
9398 break;
9399
9400 + if (__pud_val(*pud)) {
9401 + phys_pmd_update(pud, addr, end);
9402 + continue;
9403 + }
9404 +
9405 pmd = alloc_static_page(&pmd_phys);
9406
9407 spin_lock(&init_mm.page_table_lock);
9408 *pud = __pud(pmd_phys | _KERNPG_TABLE);
9409 - phys_pmd_init(pmd, paddr, end);
9410 + phys_pmd_init(pmd, addr, end);
9411 spin_unlock(&init_mm.page_table_lock);
9412
9413 early_make_page_readonly(pmd, XENFEAT_writable_page_tables);
9414 @@ -775,69 +775,18 @@ void __cpuinit zap_low_mappings(int cpu)
9415 #endif
9416 }
9417
9418 -/* Compute zone sizes for the DMA and DMA32 zones in a node. */
9419 -__init void
9420 -size_zones(unsigned long *z, unsigned long *h,
9421 - unsigned long start_pfn, unsigned long end_pfn)
9422 -{
9423 - int i;
9424 - unsigned long w;
9425 -
9426 - for (i = 0; i < MAX_NR_ZONES; i++)
9427 - z[i] = 0;
9428 -
9429 - if (start_pfn < MAX_DMA_PFN)
9430 - z[ZONE_DMA] = MAX_DMA_PFN - start_pfn;
9431 - if (start_pfn < MAX_DMA32_PFN) {
9432 - unsigned long dma32_pfn = MAX_DMA32_PFN;
9433 - if (dma32_pfn > end_pfn)
9434 - dma32_pfn = end_pfn;
9435 - z[ZONE_DMA32] = dma32_pfn - start_pfn;
9436 - }
9437 - z[ZONE_NORMAL] = end_pfn - start_pfn;
9438 -
9439 - /* Remove lower zones from higher ones. */
9440 - w = 0;
9441 - for (i = 0; i < MAX_NR_ZONES; i++) {
9442 - if (z[i])
9443 - z[i] -= w;
9444 - w += z[i];
9445 - }
9446 -
9447 - /* Compute holes */
9448 - w = start_pfn;
9449 - for (i = 0; i < MAX_NR_ZONES; i++) {
9450 - unsigned long s = w;
9451 - w += z[i];
9452 - h[i] = e820_hole_size(s, w);
9453 - }
9454 -
9455 - /* Add the space pace needed for mem_map to the holes too. */
9456 - for (i = 0; i < MAX_NR_ZONES; i++)
9457 - h[i] += (z[i] * sizeof(struct page)) / PAGE_SIZE;
9458 -
9459 - /* The 16MB DMA zone has the kernel and other misc mappings.
9460 - Account them too */
9461 - if (h[ZONE_DMA]) {
9462 - h[ZONE_DMA] += dma_reserve;
9463 - if (h[ZONE_DMA] >= z[ZONE_DMA]) {
9464 - printk(KERN_WARNING
9465 - "Kernel too large and filling up ZONE_DMA?\n");
9466 - h[ZONE_DMA] = z[ZONE_DMA];
9467 - }
9468 - }
9469 -}
9470 -
9471 #ifndef CONFIG_NUMA
9472 void __init paging_init(void)
9473 {
9474 - unsigned long zones[MAX_NR_ZONES], holes[MAX_NR_ZONES];
9475 + unsigned long max_zone_pfns[MAX_NR_ZONES];
9476 + memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
9477 + max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
9478 + max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
9479 + max_zone_pfns[ZONE_NORMAL] = end_pfn;
9480
9481 memory_present(0, 0, end_pfn);
9482 sparse_init();
9483 - size_zones(zones, holes, 0, end_pfn);
9484 - free_area_init_node(0, NODE_DATA(0), zones,
9485 - __pa(PAGE_OFFSET) >> PAGE_SHIFT, holes);
9486 + free_area_init_nodes(max_zone_pfns);
9487
9488 init_mm.context.pinned = 1;
9489 }
9490 @@ -891,36 +840,23 @@ void online_page(struct page *page)
9491
9492 #ifdef CONFIG_MEMORY_HOTPLUG
9493 /*
9494 - * XXX: memory_add_physaddr_to_nid() is to find node id from physical address
9495 - * via probe interface of sysfs. If acpi notifies hot-add event, then it
9496 - * can tell node id by searching dsdt. But, probe interface doesn't have
9497 - * node id. So, return 0 as node id at this time.
9498 - */
9499 -#ifdef CONFIG_NUMA
9500 -int memory_add_physaddr_to_nid(u64 start)
9501 -{
9502 - return 0;
9503 -}
9504 -#endif
9505 -
9506 -/*
9507 * Memory is added always to NORMAL zone. This means you will never get
9508 * additional DMA/DMA32 memory.
9509 */
9510 int arch_add_memory(int nid, u64 start, u64 size)
9511 {
9512 struct pglist_data *pgdat = NODE_DATA(nid);
9513 - struct zone *zone = pgdat->node_zones + MAX_NR_ZONES-2;
9514 + struct zone *zone = pgdat->node_zones + ZONE_NORMAL;
9515 unsigned long start_pfn = start >> PAGE_SHIFT;
9516 unsigned long nr_pages = size >> PAGE_SHIFT;
9517 int ret;
9518
9519 + init_memory_mapping(start, (start + size -1));
9520 +
9521 ret = __add_pages(zone, start_pfn, nr_pages);
9522 if (ret)
9523 goto error;
9524
9525 - init_memory_mapping(start, (start + size -1));
9526 -
9527 return ret;
9528 error:
9529 printk("%s: Problem encountered in __add_pages!\n", __func__);
9530 @@ -934,7 +870,17 @@ int remove_memory(u64 start, u64 size)
9531 }
9532 EXPORT_SYMBOL_GPL(remove_memory);
9533
9534 -#else /* CONFIG_MEMORY_HOTPLUG */
9535 +#if !defined(CONFIG_ACPI_NUMA) && defined(CONFIG_NUMA)
9536 +int memory_add_physaddr_to_nid(u64 start)
9537 +{
9538 + return 0;
9539 +}
9540 +EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
9541 +#endif
9542 +
9543 +#endif /* CONFIG_MEMORY_HOTPLUG */
9544 +
9545 +#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE
9546 /*
9547 * Memory Hotadd without sparsemem. The mem_maps have been allocated in advance,
9548 * just online the pages.
9549 @@ -960,7 +906,7 @@ int __add_pages(struct zone *z, unsigned
9550 }
9551 return err;
9552 }
9553 -#endif /* CONFIG_MEMORY_HOTPLUG */
9554 +#endif
9555
9556 static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, kcore_modules,
9557 kcore_vsyscall;
9558 @@ -972,12 +918,6 @@ void __init mem_init(void)
9559
9560 pci_iommu_alloc();
9561
9562 - /* How many end-of-memory variables you have, grandma! */
9563 - max_low_pfn = end_pfn;
9564 - max_pfn = end_pfn;
9565 - num_physpages = end_pfn;
9566 - high_memory = (void *) __va(end_pfn * PAGE_SIZE);
9567 -
9568 /* clear the zero-page */
9569 memset(empty_zero_page, 0, PAGE_SIZE);
9570
9571 @@ -995,7 +935,8 @@ void __init mem_init(void)
9572 init_page_count(pfn_to_page(pfn));
9573 totalram_pages++;
9574 }
9575 - reservedpages = end_pfn - totalram_pages - e820_hole_size(0, end_pfn);
9576 + reservedpages = end_pfn - totalram_pages -
9577 + absent_pages_in_range(0, end_pfn);
9578
9579 after_bootmem = 1;
9580
9581 @@ -1102,15 +1043,32 @@ void free_initrd_mem(unsigned long start
9582
9583 void __init reserve_bootmem_generic(unsigned long phys, unsigned len)
9584 {
9585 - /* Should check here against the e820 map to avoid double free */
9586 #ifdef CONFIG_NUMA
9587 int nid = phys_to_nid(phys);
9588 +#endif
9589 + unsigned long pfn = phys >> PAGE_SHIFT;
9590 + if (pfn >= end_pfn) {
9591 + /* This can happen with kdump kernels when accessing firmware
9592 + tables. */
9593 + if (pfn < end_pfn_map)
9594 + return;
9595 + printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %u\n",
9596 + phys, len);
9597 + return;
9598 + }
9599 +
9600 + /* Should check here against the e820 map to avoid double free */
9601 +#ifdef CONFIG_NUMA
9602 reserve_bootmem_node(NODE_DATA(nid), phys, len);
9603 #else
9604 reserve_bootmem(phys, len);
9605 #endif
9606 - if (phys+len <= MAX_DMA_PFN*PAGE_SIZE)
9607 +#ifndef CONFIG_XEN
9608 + if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
9609 dma_reserve += len / PAGE_SIZE;
9610 + set_dma_reserve(dma_reserve);
9611 + }
9612 +#endif
9613 }
9614
9615 int kern_addr_valid(unsigned long addr)
9616 --- sle11-2009-05-14.orig/arch/x86/mm/pageattr_64-xen.c 2008-12-01 11:25:57.000000000 +0100
9617 +++ sle11-2009-05-14/arch/x86/mm/pageattr_64-xen.c 2009-03-04 11:28:34.000000000 +0100
9618 @@ -377,8 +377,8 @@ static void revert_page(unsigned long ad
9619 BUG_ON(pud_none(*pud));
9620 pmd = pmd_offset(pud, address);
9621 BUG_ON(__pmd_val(*pmd) & _PAGE_PSE);
9622 - pgprot_val(ref_prot) |= _PAGE_PSE;
9623 large_pte = mk_pte_phys(__pa(address) & LARGE_PAGE_MASK, ref_prot);
9624 + large_pte = pte_mkhuge(large_pte);
9625 set_pte((pte_t *)pmd, large_pte);
9626 }
9627
9628 @@ -388,32 +388,28 @@ __change_page_attr(unsigned long address
9629 {
9630 pte_t *kpte;
9631 struct page *kpte_page;
9632 - unsigned kpte_flags;
9633 pgprot_t ref_prot2;
9634 kpte = lookup_address(address);
9635 if (!kpte) return 0;
9636 kpte_page = virt_to_page(((unsigned long)kpte) & PAGE_MASK);
9637 - kpte_flags = pte_val(*kpte);
9638 if (pgprot_val(prot) != pgprot_val(ref_prot)) {
9639 - if ((kpte_flags & _PAGE_PSE) == 0) {
9640 + if (!pte_huge(*kpte)) {
9641 set_pte(kpte, pfn_pte(pfn, prot));
9642 } else {
9643 /*
9644 * split_large_page will take the reference for this
9645 * change_page_attr on the split page.
9646 */
9647 -
9648 struct page *split;
9649 - ref_prot2 = __pgprot(pgprot_val(pte_pgprot(*lookup_address(address))) & ~(1<<_PAGE_BIT_PSE));
9650 -
9651 + ref_prot2 = pte_pgprot(pte_clrhuge(*kpte));
9652 split = split_large_page(address, prot, ref_prot2);
9653 if (!split)
9654 return -ENOMEM;
9655 - set_pte(kpte,mk_pte(split, ref_prot2));
9656 + set_pte(kpte, mk_pte(split, ref_prot2));
9657 kpte_page = split;
9658 - }
9659 + }
9660 page_private(kpte_page)++;
9661 - } else if ((kpte_flags & _PAGE_PSE) == 0) {
9662 + } else if (!pte_huge(*kpte)) {
9663 set_pte(kpte, pfn_pte(pfn, ref_prot));
9664 BUG_ON(page_private(kpte_page) == 0);
9665 page_private(kpte_page)--;
9666 @@ -470,10 +466,12 @@ int change_page_attr_addr(unsigned long
9667 * lowmem */
9668 if (__pa(address) < KERNEL_TEXT_SIZE) {
9669 unsigned long addr2;
9670 - pgprot_t prot2 = prot;
9671 + pgprot_t prot2;
9672 addr2 = __START_KERNEL_map + __pa(address);
9673 - pgprot_val(prot2) &= ~_PAGE_NX;
9674 - err = __change_page_attr(addr2, pfn, prot2, PAGE_KERNEL_EXEC);
9675 + /* Make sure the kernel mappings stay executable */
9676 + prot2 = pte_pgprot(pte_mkexec(pfn_pte(0, prot)));
9677 + err = __change_page_attr(addr2, pfn, prot2,
9678 + PAGE_KERNEL_EXEC);
9679 }
9680 }
9681 up_write(&init_mm.mmap_sem);
9682 --- sle11-2009-05-14.orig/drivers/char/tpm/tpm_xen.c 2009-02-16 15:58:14.000000000 +0100
9683 +++ sle11-2009-05-14/drivers/char/tpm/tpm_xen.c 2009-03-04 11:28:34.000000000 +0100
9684 @@ -85,8 +85,7 @@ static struct tpm_private *my_priv;
9685
9686 /* local function prototypes */
9687 static irqreturn_t tpmif_int(int irq,
9688 - void *tpm_priv,
9689 - struct pt_regs *ptregs);
9690 + void *tpm_priv);
9691 static void tpmif_rx_action(unsigned long unused);
9692 static int tpmif_connect(struct xenbus_device *dev,
9693 struct tpm_private *tp,
9694 @@ -559,7 +558,7 @@ static void tpmif_rx_action(unsigned lon
9695 }
9696
9697
9698 -static irqreturn_t tpmif_int(int irq, void *tpm_priv, struct pt_regs *ptregs)
9699 +static irqreturn_t tpmif_int(int irq, void *tpm_priv)
9700 {
9701 struct tpm_private *tp = tpm_priv;
9702 unsigned long flags;
9703 --- sle11-2009-05-14.orig/drivers/pci/Kconfig 2009-05-14 11:02:43.000000000 +0200
9704 +++ sle11-2009-05-14/drivers/pci/Kconfig 2009-03-04 11:28:34.000000000 +0100
9705 @@ -48,7 +48,7 @@ config PCI_DEBUG
9706 config HT_IRQ
9707 bool "Interrupts on hypertransport devices"
9708 default y
9709 - depends on PCI && X86_LOCAL_APIC && X86_IO_APIC
9710 + depends on PCI && X86_LOCAL_APIC && X86_IO_APIC && !XEN
9711 help
9712 This allows native hypertransport devices to use interrupts.
9713
9714 --- sle11-2009-05-14.orig/drivers/pci/msi-xen.c 2009-05-14 11:02:43.000000000 +0200
9715 +++ sle11-2009-05-14/drivers/pci/msi-xen.c 2009-04-24 13:31:56.000000000 +0200
9716 @@ -6,6 +6,7 @@
9717 * Copyright (C) Tom Long Nguyen (tom.l.nguyen@intel.com)
9718 */
9719
9720 +#include <linux/err.h>
9721 #include <linux/mm.h>
9722 #include <linux/irq.h>
9723 #include <linux/interrupt.h>
9724 @@ -14,6 +15,7 @@
9725 #include <linux/smp_lock.h>
9726 #include <linux/pci.h>
9727 #include <linux/proc_fs.h>
9728 +#include <linux/msi.h>
9729
9730 #include <xen/evtchn.h>
9731
9732 @@ -26,14 +28,6 @@
9733
9734 static int pci_msi_enable = 1;
9735
9736 -static struct msi_ops *msi_ops;
9737 -
9738 -int msi_register(struct msi_ops *ops)
9739 -{
9740 - msi_ops = ops;
9741 - return 0;
9742 -}
9743 -
9744 static LIST_HEAD(msi_dev_head);
9745 DEFINE_SPINLOCK(msi_dev_lock);
9746
9747 @@ -406,9 +400,9 @@ void pci_restore_msix_state(struct pci_d
9748 * @dev: pointer to the pci_dev data structure of MSI device function
9749 *
9750 * Setup the MSI capability structure of device function with a single
9751 - * MSI vector, regardless of device function is capable of handling
9752 + * MSI irq, regardless of device function is capable of handling
9753 * multiple messages. A return of zero indicates the successful setup
9754 - * of an entry zero with the new MSI vector or non-zero for otherwise.
9755 + * of an entry zero with the new MSI irq or non-zero for otherwise.
9756 **/
9757 static int msi_capability_init(struct pci_dev *dev)
9758 {
9759 @@ -422,11 +416,11 @@ static int msi_capability_init(struct pc
9760 if (pirq < 0)
9761 return -EBUSY;
9762
9763 - dev->irq = pirq;
9764 /* Set MSI enabled bits */
9765 enable_msi_mode(dev, pos, PCI_CAP_ID_MSI);
9766 dev->msi_enabled = 1;
9767
9768 + dev->irq = pirq;
9769 return 0;
9770 }
9771
9772 @@ -437,8 +431,8 @@ static int msi_capability_init(struct pc
9773 * @nvec: number of @entries
9774 *
9775 * Setup the MSI-X capability structure of device function with a
9776 - * single MSI-X vector. A return of zero indicates the successful setup of
9777 - * requested MSI-X entries with allocated vectors or non-zero for otherwise.
9778 + * single MSI-X irq. A return of zero indicates the successful setup of
9779 + * requested MSI-X entries with allocated irqs or non-zero for otherwise.
9780 **/
9781 static int msix_capability_init(struct pci_dev *dev,
9782 struct msix_entry *entries, int nvec)
9783 @@ -480,12 +474,18 @@ static int msix_capability_init(struct p
9784 }
9785
9786 if (i != nvec) {
9787 + int avail = i - 1;
9788 for (j = --i; j >= 0; j--) {
9789 msi_unmap_pirq(dev, entries[j].vector);
9790 detach_pirq_entry(entries[j].entry, msi_dev_entry);
9791 entries[j].vector = 0;
9792 }
9793 - return -EBUSY;
9794 + /* If we had some success report the number of irqs
9795 + * we succeeded in setting up.
9796 + */
9797 + if (avail <= 0)
9798 + avail = -EBUSY;
9799 + return avail;
9800 }
9801
9802 enable_msi_mode(dev, pos, PCI_CAP_ID_MSIX);
9803 @@ -495,11 +495,40 @@ static int msix_capability_init(struct p
9804 }
9805
9806 /**
9807 + * pci_msi_supported - check whether MSI may be enabled on device
9808 + * @dev: pointer to the pci_dev data structure of MSI device function
9809 + *
9810 + * Look at global flags, the device itself, and its parent busses
9811 + * to return 0 if MSI are supported for the device.
9812 + **/
9813 +static
9814 +int pci_msi_supported(struct pci_dev * dev)
9815 +{
9816 + struct pci_bus *bus;
9817 +
9818 + /* MSI must be globally enabled and supported by the device */
9819 + if (!pci_msi_enable || !dev || dev->no_msi)
9820 + return -EINVAL;
9821 +
9822 + /* Any bridge which does NOT route MSI transactions from it's
9823 + * secondary bus to it's primary bus must set NO_MSI flag on
9824 + * the secondary pci_bus.
9825 + * We expect only arch-specific PCI host bus controller driver
9826 + * or quirks for specific PCI bridges to be setting NO_MSI.
9827 + */
9828 + for (bus = dev->bus; bus; bus = bus->parent)
9829 + if (bus->bus_flags & PCI_BUS_FLAGS_NO_MSI)
9830 + return -EINVAL;
9831 +
9832 + return 0;
9833 +}
9834 +
9835 +/**
9836 * pci_enable_msi - configure device's MSI capability structure
9837 * @dev: pointer to the pci_dev data structure of MSI device function
9838 *
9839 * Setup the MSI capability structure of device function with
9840 - * a single MSI vector upon its software driver call to request for
9841 + * a single MSI irq upon its software driver call to request for
9842 * MSI mode enabled on its hardware device function. A return of zero
9843 * indicates the successful setup of an entry zero with the new MSI
9844 * vector or non-zero for otherwise.
9845 @@ -507,18 +536,10 @@ static int msix_capability_init(struct p
9846 extern int pci_frontend_enable_msi(struct pci_dev *dev);
9847 int pci_enable_msi(struct pci_dev* dev)
9848 {
9849 - struct pci_bus *bus;
9850 - int pos, temp, status = -EINVAL;
9851 + int pos, temp, status;
9852
9853 - if (!pci_msi_enable || !dev)
9854 - return status;
9855 -
9856 - if (dev->no_msi)
9857 - return status;
9858 -
9859 - for (bus = dev->bus; bus; bus = bus->parent)
9860 - if (bus->bus_flags & PCI_BUS_FLAGS_NO_MSI)
9861 - return -EINVAL;
9862 + if (pci_msi_supported(dev) < 0)
9863 + return -EINVAL;
9864
9865 status = msi_init();
9866 if (status < 0)
9867 @@ -547,10 +568,10 @@ int pci_enable_msi(struct pci_dev* dev)
9868 if (!pos)
9869 return -EINVAL;
9870
9871 - /* Check whether driver already requested for MSI-X vectors */
9872 + /* Check whether driver already requested for MSI-X irqs */
9873 if (dev->msix_enabled) {
9874 printk(KERN_INFO "PCI: %s: Can't enable MSI. "
9875 - "Device already has MSI-X vectors assigned\n",
9876 + "Device already has MSI-X irq assigned\n",
9877 pci_name(dev));
9878 dev->irq = temp;
9879 return -EINVAL;
9880 @@ -602,36 +623,28 @@ void pci_disable_msi(struct pci_dev* dev
9881 * pci_enable_msix - configure device's MSI-X capability structure
9882 * @dev: pointer to the pci_dev data structure of MSI-X device function
9883 * @entries: pointer to an array of MSI-X entries
9884 - * @nvec: number of MSI-X vectors requested for allocation by device driver
9885 + * @nvec: number of MSI-X irqs requested for allocation by device driver
9886 *
9887 * Setup the MSI-X capability structure of device function with the number
9888 - * of requested vectors upon its software driver call to request for
9889 + * of requested irqs upon its software driver call to request for
9890 * MSI-X mode enabled on its hardware device function. A return of zero
9891 * indicates the successful configuration of MSI-X capability structure
9892 - * with new allocated MSI-X vectors. A return of < 0 indicates a failure.
9893 + * with new allocated MSI-X irqs. A return of < 0 indicates a failure.
9894 * Or a return of > 0 indicates that driver request is exceeding the number
9895 - * of vectors available. Driver should use the returned value to re-send
9896 + * of irqs available. Driver should use the returned value to re-send
9897 * its request.
9898 **/
9899 extern int pci_frontend_enable_msix(struct pci_dev *dev,
9900 struct msix_entry *entries, int nvec);
9901 int pci_enable_msix(struct pci_dev* dev, struct msix_entry *entries, int nvec)
9902 {
9903 - struct pci_bus *bus;
9904 int status, pos, nr_entries;
9905 int i, j, temp;
9906 u16 control;
9907
9908 - if (!pci_msi_enable || !dev || !entries)
9909 + if (!entries || pci_msi_supported(dev) < 0)
9910 return -EINVAL;
9911
9912 - if (dev->no_msi)
9913 - return -EINVAL;
9914 -
9915 - for (bus = dev->bus; bus; bus = bus->parent)
9916 - if (bus->bus_flags & PCI_BUS_FLAGS_NO_MSI)
9917 - return -EINVAL;
9918 -
9919 #ifdef CONFIG_XEN_PCIDEV_FRONTEND
9920 if (!is_initial_xendomain()) {
9921 struct msi_dev_list *msi_dev_entry;
9922 @@ -694,7 +707,7 @@ int pci_enable_msix(struct pci_dev* dev,
9923 /* Check whether driver already requested for MSI vector */
9924 if (dev->msi_enabled) {
9925 printk(KERN_INFO "PCI: %s: Can't enable MSI-X. "
9926 - "Device already has an MSI vector assigned\n",
9927 + "Device already has an MSI irq assigned\n",
9928 pci_name(dev));
9929 dev->irq = temp;
9930 return -EINVAL;
9931 @@ -757,11 +770,11 @@ void pci_disable_msix(struct pci_dev* de
9932 }
9933
9934 /**
9935 - * msi_remove_pci_irq_vectors - reclaim MSI(X) vectors to unused state
9936 + * msi_remove_pci_irq_vectors - reclaim MSI(X) irqs to unused state
9937 * @dev: pointer to the pci_dev data structure of MSI(X) device function
9938 *
9939 * Being called during hotplug remove, from which the device function
9940 - * is hot-removed. All previous assigned MSI/MSI-X vectors, if
9941 + * is hot-removed. All previous assigned MSI/MSI-X irqs, if
9942 * allocated for this device function, are reclaimed to unused state,
9943 * which may be used later on.
9944 **/
9945 --- sle11-2009-05-14.orig/drivers/xen/Kconfig 2008-12-05 08:43:56.000000000 +0100
9946 +++ sle11-2009-05-14/drivers/xen/Kconfig 2009-03-04 11:28:34.000000000 +0100
9947 @@ -287,6 +287,9 @@ endmenu
9948 config HAVE_IRQ_IGNORE_UNHANDLED
9949 def_bool y
9950
9951 +config GENERIC_HARDIRQS_NO__DO_IRQ
9952 + def_bool y
9953 +
9954 config NO_IDLE_HZ
9955 def_bool y
9956
9957 --- sle11-2009-05-14.orig/drivers/xen/balloon/balloon.c 2008-11-25 13:31:07.000000000 +0100
9958 +++ sle11-2009-05-14/drivers/xen/balloon/balloon.c 2009-03-04 11:28:34.000000000 +0100
9959 @@ -84,7 +84,7 @@ static unsigned long frame_list[PAGE_SIZ
9960 /* VM /proc information for memory */
9961 extern unsigned long totalram_pages;
9962
9963 -#ifndef MODULE
9964 +#if !defined(MODULE) && defined(CONFIG_HIGHMEM)
9965 extern unsigned long totalhigh_pages;
9966 #define inc_totalhigh_pages() (totalhigh_pages++)
9967 #define dec_totalhigh_pages() (totalhigh_pages--)
9968 --- sle11-2009-05-14.orig/drivers/xen/blkback/blkback.c 2008-12-01 11:21:10.000000000 +0100
9969 +++ sle11-2009-05-14/drivers/xen/blkback/blkback.c 2009-03-04 11:28:34.000000000 +0100
9970 @@ -288,7 +288,7 @@ static void blkif_notify_work(blkif_t *b
9971 wake_up(&blkif->wq);
9972 }
9973
9974 -irqreturn_t blkif_be_int(int irq, void *dev_id, struct pt_regs *regs)
9975 +irqreturn_t blkif_be_int(int irq, void *dev_id)
9976 {
9977 blkif_notify_work(dev_id);
9978 return IRQ_HANDLED;
9979 --- sle11-2009-05-14.orig/drivers/xen/blkback/common.h 2009-05-14 11:02:43.000000000 +0200
9980 +++ sle11-2009-05-14/drivers/xen/blkback/common.h 2009-03-04 11:28:34.000000000 +0100
9981 @@ -130,7 +130,7 @@ void blkif_interface_init(void);
9982
9983 void blkif_xenbus_init(void);
9984
9985 -irqreturn_t blkif_be_int(int irq, void *dev_id, struct pt_regs *regs);
9986 +irqreturn_t blkif_be_int(int irq, void *dev_id);
9987 int blkif_schedule(void *arg);
9988
9989 int blkback_barrier(struct xenbus_transaction xbt,
9990 --- sle11-2009-05-14.orig/drivers/xen/blkfront/blkfront.c 2009-03-05 15:42:00.000000000 +0100
9991 +++ sle11-2009-05-14/drivers/xen/blkfront/blkfront.c 2009-03-24 10:08:16.000000000 +0100
9992 @@ -70,7 +70,7 @@ static int setup_blkring(struct xenbus_d
9993
9994 static void kick_pending_request_queues(struct blkfront_info *);
9995
9996 -static irqreturn_t blkif_int(int irq, void *dev_id, struct pt_regs *ptregs);
9997 +static irqreturn_t blkif_int(int irq, void *dev_id);
9998 static void blkif_restart_queue(void *arg);
9999 static void blkif_recover(struct blkfront_info *);
10000 static void blkif_completion(struct blk_shadow *);
10001 @@ -707,7 +707,7 @@ void do_blkif_request(request_queue_t *r
10002 }
10003
10004
10005 -static irqreturn_t blkif_int(int irq, void *dev_id, struct pt_regs *ptregs)
10006 +static irqreturn_t blkif_int(int irq, void *dev_id)
10007 {
10008 struct request *req;
10009 blkif_response_t *bret;
10010 --- sle11-2009-05-14.orig/drivers/xen/blktap/blktap.c 2009-04-20 11:36:10.000000000 +0200
10011 +++ sle11-2009-05-14/drivers/xen/blktap/blktap.c 2009-04-20 11:37:34.000000000 +0200
10012 @@ -1222,7 +1222,7 @@ static void blkif_notify_work(blkif_t *b
10013 wake_up(&blkif->wq);
10014 }
10015
10016 -irqreturn_t tap_blkif_be_int(int irq, void *dev_id, struct pt_regs *regs)
10017 +irqreturn_t tap_blkif_be_int(int irq, void *dev_id)
10018 {
10019 blkif_notify_work(dev_id);
10020 return IRQ_HANDLED;
10021 --- sle11-2009-05-14.orig/drivers/xen/blktap/common.h 2009-05-14 11:02:43.000000000 +0200
10022 +++ sle11-2009-05-14/drivers/xen/blktap/common.h 2009-03-04 11:28:34.000000000 +0100
10023 @@ -113,7 +113,7 @@ void tap_blkif_interface_init(void);
10024
10025 void tap_blkif_xenbus_init(void);
10026
10027 -irqreturn_t tap_blkif_be_int(int irq, void *dev_id, struct pt_regs *regs);
10028 +irqreturn_t tap_blkif_be_int(int irq, void *dev_id);
10029 int tap_blkif_schedule(void *arg);
10030
10031 int dom_to_devid(domid_t domid, int xenbus_id, blkif_t *blkif);
10032 --- sle11-2009-05-14.orig/drivers/xen/console/console.c 2008-12-15 11:13:47.000000000 +0100
10033 +++ sle11-2009-05-14/drivers/xen/console/console.c 2009-03-04 11:28:34.000000000 +0100
10034 @@ -361,7 +361,7 @@ static struct tty_struct *xencons_tty;
10035 static int xencons_priv_irq;
10036 static char x_char;
10037
10038 -void xencons_rx(char *buf, unsigned len, struct pt_regs *regs)
10039 +void xencons_rx(char *buf, unsigned len)
10040 {
10041 int i;
10042 unsigned long flags;
10043 @@ -386,8 +386,7 @@ void xencons_rx(char *buf, unsigned len,
10044 if (time_before(jiffies, sysrq_timeout)) {
10045 spin_unlock_irqrestore(
10046 &xencons_lock, flags);
10047 - handle_sysrq(
10048 - buf[i], regs, xencons_tty);
10049 + handle_sysrq(buf[i], xencons_tty);
10050 spin_lock_irqsave(
10051 &xencons_lock, flags);
10052 continue;
10053 @@ -452,14 +451,13 @@ void xencons_tx(void)
10054 }
10055
10056 /* Privileged receive callback and transmit kicker. */
10057 -static irqreturn_t xencons_priv_interrupt(int irq, void *dev_id,
10058 - struct pt_regs *regs)
10059 +static irqreturn_t xencons_priv_interrupt(int irq, void *dev_id)
10060 {
10061 static char rbuf[16];
10062 int l;
10063
10064 while ((l = HYPERVISOR_console_io(CONSOLEIO_read, 16, rbuf)) > 0)
10065 - xencons_rx(rbuf, l, regs);
10066 + xencons_rx(rbuf, l);
10067
10068 xencons_tx();
10069
10070 @@ -647,7 +645,7 @@ static void xencons_close(struct tty_str
10071 spin_unlock_irqrestore(&xencons_lock, flags);
10072 }
10073
10074 -static struct tty_operations xencons_ops = {
10075 +static const struct tty_operations xencons_ops = {
10076 .open = xencons_open,
10077 .close = xencons_close,
10078 .write = xencons_write,
10079 --- sle11-2009-05-14.orig/drivers/xen/console/xencons_ring.c 2009-05-14 11:02:43.000000000 +0200
10080 +++ sle11-2009-05-14/drivers/xen/console/xencons_ring.c 2009-03-04 11:28:34.000000000 +0100
10081 @@ -83,7 +83,7 @@ int xencons_ring_send(const char *data,
10082 return sent;
10083 }
10084
10085 -static irqreturn_t handle_input(int irq, void *unused, struct pt_regs *regs)
10086 +static irqreturn_t handle_input(int irq, void *unused)
10087 {
10088 struct xencons_interface *intf = xencons_interface();
10089 XENCONS_RING_IDX cons, prod;
10090 @@ -94,7 +94,7 @@ static irqreturn_t handle_input(int irq,
10091 BUG_ON((prod - cons) > sizeof(intf->in));
10092
10093 while (cons != prod) {
10094 - xencons_rx(intf->in+MASK_XENCONS_IDX(cons,intf->in), 1, regs);
10095 + xencons_rx(intf->in+MASK_XENCONS_IDX(cons,intf->in), 1);
10096 cons++;
10097 }
10098
10099 --- sle11-2009-05-14.orig/drivers/xen/core/evtchn.c 2009-03-16 16:14:12.000000000 +0100
10100 +++ sle11-2009-05-14/drivers/xen/core/evtchn.c 2009-03-04 11:28:34.000000000 +0100
10101 @@ -516,7 +516,7 @@ static void unbind_from_irq(unsigned int
10102
10103 int bind_caller_port_to_irqhandler(
10104 unsigned int caller_port,
10105 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
10106 + irq_handler_t handler,
10107 unsigned long irqflags,
10108 const char *devname,
10109 void *dev_id)
10110 @@ -539,7 +539,7 @@ EXPORT_SYMBOL_GPL(bind_caller_port_to_ir
10111
10112 int bind_listening_port_to_irqhandler(
10113 unsigned int remote_domain,
10114 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
10115 + irq_handler_t handler,
10116 unsigned long irqflags,
10117 const char *devname,
10118 void *dev_id)
10119 @@ -563,7 +563,7 @@ EXPORT_SYMBOL_GPL(bind_listening_port_to
10120 int bind_interdomain_evtchn_to_irqhandler(
10121 unsigned int remote_domain,
10122 unsigned int remote_port,
10123 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
10124 + irq_handler_t handler,
10125 unsigned long irqflags,
10126 const char *devname,
10127 void *dev_id)
10128 @@ -587,7 +587,7 @@ EXPORT_SYMBOL_GPL(bind_interdomain_evtch
10129 int bind_virq_to_irqhandler(
10130 unsigned int virq,
10131 unsigned int cpu,
10132 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
10133 + irq_handler_t handler,
10134 unsigned long irqflags,
10135 const char *devname,
10136 void *dev_id)
10137 @@ -611,7 +611,7 @@ EXPORT_SYMBOL_GPL(bind_virq_to_irqhandle
10138 int bind_ipi_to_irqhandler(
10139 unsigned int ipi,
10140 unsigned int cpu,
10141 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
10142 + irq_handler_t handler,
10143 unsigned long irqflags,
10144 const char *devname,
10145 void *dev_id)
10146 @@ -696,15 +696,7 @@ static unsigned int startup_dynirq(unsig
10147 return 0;
10148 }
10149
10150 -static void shutdown_dynirq(unsigned int irq)
10151 -{
10152 - int evtchn = evtchn_from_irq(irq);
10153 -
10154 - if (VALID_EVTCHN(evtchn))
10155 - mask_evtchn(evtchn);
10156 -}
10157 -
10158 -static void enable_dynirq(unsigned int irq)
10159 +static void unmask_dynirq(unsigned int irq)
10160 {
10161 int evtchn = evtchn_from_irq(irq);
10162
10163 @@ -712,7 +704,7 @@ static void enable_dynirq(unsigned int i
10164 unmask_evtchn(evtchn);
10165 }
10166
10167 -static void disable_dynirq(unsigned int irq)
10168 +static void mask_dynirq(unsigned int irq)
10169 {
10170 int evtchn = evtchn_from_irq(irq);
10171
10172 @@ -740,12 +732,13 @@ static void end_dynirq(unsigned int irq)
10173 unmask_evtchn(evtchn);
10174 }
10175
10176 -static struct hw_interrupt_type dynirq_type = {
10177 - .typename = "Dynamic-irq",
10178 +static struct irq_chip dynirq_chip = {
10179 + .name = "Dynamic",
10180 .startup = startup_dynirq,
10181 - .shutdown = shutdown_dynirq,
10182 - .enable = enable_dynirq,
10183 - .disable = disable_dynirq,
10184 + .shutdown = mask_dynirq,
10185 + .mask = mask_dynirq,
10186 + .unmask = unmask_dynirq,
10187 + .mask_ack = ack_dynirq,
10188 .ack = ack_dynirq,
10189 .end = end_dynirq,
10190 #ifdef CONFIG_SMP
10191 @@ -859,12 +852,12 @@ static void shutdown_pirq(unsigned int i
10192 irq_info[irq] = mk_irq_info(IRQT_PIRQ, index_from_irq(irq), 0);
10193 }
10194
10195 -static void enable_pirq(unsigned int irq)
10196 +static void unmask_pirq(unsigned int irq)
10197 {
10198 startup_pirq(irq);
10199 }
10200
10201 -static void disable_pirq(unsigned int irq)
10202 +static void mask_pirq(unsigned int irq)
10203 {
10204 }
10205
10206 @@ -891,12 +884,13 @@ static void end_pirq(unsigned int irq)
10207 pirq_unmask_and_notify(evtchn, irq);
10208 }
10209
10210 -static struct hw_interrupt_type pirq_type = {
10211 - .typename = "Phys-irq",
10212 +static struct irq_chip pirq_chip = {
10213 + .name = "Phys",
10214 .startup = startup_pirq,
10215 .shutdown = shutdown_pirq,
10216 - .enable = enable_pirq,
10217 - .disable = disable_pirq,
10218 + .mask = mask_pirq,
10219 + .unmask = unmask_pirq,
10220 + .mask_ack = ack_pirq,
10221 .ack = ack_pirq,
10222 .end = end_pirq,
10223 #ifdef CONFIG_SMP
10224 @@ -1081,7 +1075,8 @@ void evtchn_register_pirq(int irq)
10225 if (identity_mapped_irq(irq) || type_from_irq(irq) != IRQT_UNBOUND)
10226 return;
10227 irq_info[irq] = mk_irq_info(IRQT_PIRQ, irq, 0);
10228 - irq_desc[irq].chip = &pirq_type;
10229 + set_irq_chip_and_handler_name(irq, &pirq_chip, handle_level_irq,
10230 + "level");
10231 }
10232
10233 int evtchn_map_pirq(int irq, int xen_pirq)
10234 @@ -1104,11 +1099,18 @@ int evtchn_map_pirq(int irq, int xen_pir
10235 spin_unlock(&irq_alloc_lock);
10236 if (irq < PIRQ_BASE)
10237 return -ENOSPC;
10238 - irq_desc[irq].chip = &pirq_type;
10239 + set_irq_chip_and_handler_name(irq, &pirq_chip,
10240 + handle_level_irq, "level");
10241 } else if (!xen_pirq) {
10242 if (unlikely(type_from_irq(irq) != IRQT_PIRQ))
10243 return -EINVAL;
10244 - irq_desc[irq].chip = &no_irq_type;
10245 + /*
10246 + * dynamic_irq_cleanup(irq) would seem to be the correct thing
10247 + * here, but cannot be used as we get here also during shutdown
10248 + * when a driver didn't free_irq() its MSI(-X) IRQ(s), which
10249 + * then causes a warning in dynamic_irq_cleanup().
10250 + */
10251 + set_irq_chip_and_handler(irq, NULL, NULL);
10252 irq_info[irq] = IRQ_UNBOUND;
10253 return 0;
10254 } else if (type_from_irq(irq) != IRQT_PIRQ
10255 @@ -1154,10 +1156,9 @@ void __init xen_init_IRQ(void)
10256 for (i = DYNIRQ_BASE; i < (DYNIRQ_BASE + NR_DYNIRQS); i++) {
10257 irq_bindcount[i] = 0;
10258
10259 - irq_desc[i].status = IRQ_DISABLED|IRQ_NOPROBE;
10260 - irq_desc[i].action = NULL;
10261 - irq_desc[i].depth = 1;
10262 - irq_desc[i].chip = &dynirq_type;
10263 + irq_desc[i].status |= IRQ_NOPROBE;
10264 + set_irq_chip_and_handler_name(i, &dynirq_chip,
10265 + handle_level_irq, "level");
10266 }
10267
10268 /* Phys IRQ space is statically bound (1:1 mapping). Nail refcnts. */
10269 @@ -1173,9 +1174,7 @@ void __init xen_init_IRQ(void)
10270 continue;
10271 #endif
10272
10273 - irq_desc[i].status = IRQ_DISABLED;
10274 - irq_desc[i].action = NULL;
10275 - irq_desc[i].depth = 1;
10276 - irq_desc[i].chip = &pirq_type;
10277 + set_irq_chip_and_handler_name(i, &pirq_chip,
10278 + handle_level_irq, "level");
10279 }
10280 }
10281 --- sle11-2009-05-14.orig/drivers/xen/core/reboot.c 2009-05-14 11:02:43.000000000 +0200
10282 +++ sle11-2009-05-14/drivers/xen/core/reboot.c 2009-03-04 11:28:34.000000000 +0100
10283 @@ -14,6 +14,7 @@
10284
10285 #ifdef HAVE_XEN_PLATFORM_COMPAT_H
10286 #include <xen/platform-compat.h>
10287 +#undef handle_sysrq
10288 #endif
10289
10290 MODULE_LICENSE("Dual BSD/GPL");
10291 @@ -231,7 +232,7 @@ static void sysrq_handler(struct xenbus_
10292
10293 #ifdef CONFIG_MAGIC_SYSRQ
10294 if (sysrq_key != '\0')
10295 - handle_sysrq(sysrq_key, NULL, NULL);
10296 + handle_sysrq(sysrq_key, NULL);
10297 #endif
10298 }
10299
10300 @@ -245,7 +246,7 @@ static struct xenbus_watch sysrq_watch =
10301 .callback = sysrq_handler
10302 };
10303
10304 -static irqreturn_t suspend_int(int irq, void* dev_id, struct pt_regs *ptregs)
10305 +static irqreturn_t suspend_int(int irq, void* dev_id)
10306 {
10307 switch_shutdown_state(SHUTDOWN_SUSPEND);
10308 return IRQ_HANDLED;
10309 --- sle11-2009-05-14.orig/drivers/xen/core/smpboot.c 2009-04-28 16:02:07.000000000 +0200
10310 +++ sle11-2009-05-14/drivers/xen/core/smpboot.c 2009-03-04 11:28:34.000000000 +0100
10311 @@ -25,8 +25,8 @@
10312 #include <xen/cpu_hotplug.h>
10313 #include <xen/xenbus.h>
10314
10315 -extern irqreturn_t smp_reschedule_interrupt(int, void *, struct pt_regs *);
10316 -extern irqreturn_t smp_call_function_interrupt(int, void *, struct pt_regs *);
10317 +extern irqreturn_t smp_reschedule_interrupt(int, void *);
10318 +extern irqreturn_t smp_call_function_interrupt(int, void *);
10319
10320 extern int local_setup_timer(unsigned int cpu);
10321 extern void local_teardown_timer(unsigned int cpu);
10322 @@ -62,8 +62,6 @@ EXPORT_SYMBOL(cpu_core_map);
10323 #if defined(__i386__)
10324 u8 x86_cpu_to_apicid[NR_CPUS] = { [0 ... NR_CPUS-1] = 0xff };
10325 EXPORT_SYMBOL(x86_cpu_to_apicid);
10326 -#elif !defined(CONFIG_X86_IO_APIC)
10327 -unsigned int maxcpus = NR_CPUS;
10328 #endif
10329
10330 void __init prefill_possible_map(void)
10331 --- sle11-2009-05-14.orig/drivers/xen/fbfront/xenfb.c 2009-02-16 15:59:55.000000000 +0100
10332 +++ sle11-2009-05-14/drivers/xen/fbfront/xenfb.c 2009-03-04 11:28:34.000000000 +0100
10333 @@ -524,8 +524,7 @@ static struct fb_ops xenfb_fb_ops = {
10334 .fb_set_par = xenfb_set_par,
10335 };
10336
10337 -static irqreturn_t xenfb_event_handler(int rq, void *dev_id,
10338 - struct pt_regs *regs)
10339 +static irqreturn_t xenfb_event_handler(int rq, void *dev_id)
10340 {
10341 /*
10342 * No in events recognized, simply ignore them all.
10343 --- sle11-2009-05-14.orig/drivers/xen/fbfront/xenkbd.c 2009-05-14 11:02:43.000000000 +0200
10344 +++ sle11-2009-05-14/drivers/xen/fbfront/xenkbd.c 2009-03-04 11:28:34.000000000 +0100
10345 @@ -46,7 +46,7 @@ static void xenkbd_disconnect_backend(st
10346 * to do that.
10347 */
10348
10349 -static irqreturn_t input_handler(int rq, void *dev_id, struct pt_regs *regs)
10350 +static irqreturn_t input_handler(int rq, void *dev_id)
10351 {
10352 struct xenkbd_info *info = dev_id;
10353 struct xenkbd_page *page = info->page;
10354 --- sle11-2009-05-14.orig/drivers/xen/gntdev/gntdev.c 2008-12-15 11:13:45.000000000 +0100
10355 +++ sle11-2009-05-14/drivers/xen/gntdev/gntdev.c 2009-03-04 11:28:34.000000000 +0100
10356 @@ -752,9 +752,6 @@ static pte_t gntdev_clear_pte(struct vm_
10357 BUG();
10358 }
10359
10360 - /* Copy the existing value of the PTE for returning. */
10361 - copy = *ptep;
10362 -
10363 /* Calculate the grant relating to this PTE. */
10364 slot_index = vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT);
10365
10366 @@ -769,6 +766,10 @@ static pte_t gntdev_clear_pte(struct vm_
10367 GNTDEV_INVALID_HANDLE &&
10368 !xen_feature(XENFEAT_auto_translated_physmap)) {
10369 /* NOT USING SHADOW PAGE TABLES. */
10370 +
10371 + /* Copy the existing value of the PTE for returning. */
10372 + copy = *ptep;
10373 +
10374 gnttab_set_unmap_op(&op, ptep_to_machine(ptep),
10375 GNTMAP_contains_pte,
10376 private_data->grants[slot_index]
10377 @@ -781,7 +782,7 @@ static pte_t gntdev_clear_pte(struct vm_
10378 op.status);
10379 } else {
10380 /* USING SHADOW PAGE TABLES. */
10381 - pte_clear_full(vma->vm_mm, addr, ptep, is_fullmm);
10382 + copy = ptep_get_and_clear_full(vma->vm_mm, addr, ptep, is_fullmm);
10383 }
10384
10385 /* Finally, we unmap the grant from kernel space. */
10386 @@ -809,7 +810,7 @@ static pte_t gntdev_clear_pte(struct vm_
10387 >> PAGE_SHIFT, INVALID_P2M_ENTRY);
10388
10389 } else {
10390 - pte_clear_full(vma->vm_mm, addr, ptep, is_fullmm);
10391 + copy = ptep_get_and_clear_full(vma->vm_mm, addr, ptep, is_fullmm);
10392 }
10393
10394 return copy;
10395 --- sle11-2009-05-14.orig/drivers/xen/netback/accel.c 2009-05-14 11:02:43.000000000 +0200
10396 +++ sle11-2009-05-14/drivers/xen/netback/accel.c 2009-03-04 11:28:34.000000000 +0100
10397 @@ -65,7 +65,7 @@ static int match_accelerator(struct xenb
10398
10399 if (IS_ERR(eth_name)) {
10400 /* Probably means not present */
10401 - DPRINTK("%s: no match due to xenbus_read accel error %d\n",
10402 + DPRINTK("%s: no match due to xenbus_read accel error %ld\n",
10403 __FUNCTION__, PTR_ERR(eth_name));
10404 return 0;
10405 } else {
10406 --- sle11-2009-05-14.orig/drivers/xen/netback/common.h 2009-05-14 11:02:43.000000000 +0200
10407 +++ sle11-2009-05-14/drivers/xen/netback/common.h 2009-03-04 11:28:34.000000000 +0100
10408 @@ -200,7 +200,7 @@ void netif_deschedule_work(netif_t *neti
10409
10410 int netif_be_start_xmit(struct sk_buff *skb, struct net_device *dev);
10411 struct net_device_stats *netif_be_get_stats(struct net_device *dev);
10412 -irqreturn_t netif_be_int(int irq, void *dev_id, struct pt_regs *regs);
10413 +irqreturn_t netif_be_int(int irq, void *dev_id);
10414
10415 static inline int netbk_can_queue(struct net_device *dev)
10416 {
10417 --- sle11-2009-05-14.orig/drivers/xen/netback/loopback.c 2009-05-14 11:02:43.000000000 +0200
10418 +++ sle11-2009-05-14/drivers/xen/netback/loopback.c 2009-03-04 11:28:34.000000000 +0100
10419 @@ -151,7 +151,7 @@ static int loopback_start_xmit(struct sk
10420 np->stats.rx_bytes += skb->len;
10421 np->stats.rx_packets++;
10422
10423 - if (skb->ip_summed == CHECKSUM_HW) {
10424 + if (skb->ip_summed == CHECKSUM_PARTIAL) {
10425 /* Defer checksum calculation. */
10426 skb->proto_csum_blank = 1;
10427 /* Must be a local packet: assert its integrity. */
10428 --- sle11-2009-05-14.orig/drivers/xen/netback/netback.c 2008-12-23 09:31:07.000000000 +0100
10429 +++ sle11-2009-05-14/drivers/xen/netback/netback.c 2009-03-04 11:28:34.000000000 +0100
10430 @@ -692,7 +692,7 @@ static void net_rx_action(unsigned long
10431 id = meta[npo.meta_cons].id;
10432 flags = nr_frags ? NETRXF_more_data : 0;
10433
10434 - if (skb->ip_summed == CHECKSUM_HW) /* local packet? */
10435 + if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */
10436 flags |= NETRXF_csum_blank | NETRXF_data_validated;
10437 else if (skb->proto_data_valid) /* remote but checksummed? */
10438 flags |= NETRXF_data_validated;
10439 @@ -1459,7 +1459,7 @@ static void netif_page_release(struct pa
10440 netif_idx_release(idx);
10441 }
10442
10443 -irqreturn_t netif_be_int(int irq, void *dev_id, struct pt_regs *regs)
10444 +irqreturn_t netif_be_int(int irq, void *dev_id)
10445 {
10446 netif_t *netif = dev_id;
10447
10448 @@ -1526,7 +1526,7 @@ static netif_rx_response_t *make_rx_resp
10449 }
10450
10451 #ifdef NETBE_DEBUG_INTERRUPT
10452 -static irqreturn_t netif_be_dbg(int irq, void *dev_id, struct pt_regs *regs)
10453 +static irqreturn_t netif_be_dbg(int irq, void *dev_id)
10454 {
10455 struct list_head *ent;
10456 netif_t *netif;
10457 --- sle11-2009-05-14.orig/drivers/xen/netfront/netfront.c 2009-04-09 14:41:33.000000000 +0200
10458 +++ sle11-2009-05-14/drivers/xen/netfront/netfront.c 2009-03-30 16:34:59.000000000 +0200
10459 @@ -136,7 +136,7 @@ static inline int netif_needs_gso(struct
10460 {
10461 return skb_is_gso(skb) &&
10462 (!skb_gso_ok(skb, dev->features) ||
10463 - unlikely(skb->ip_summed != CHECKSUM_HW));
10464 + unlikely(skb->ip_summed != CHECKSUM_PARTIAL));
10465 }
10466 #else
10467 #define HAVE_GSO 0
10468 @@ -222,7 +222,7 @@ static void network_tx_buf_gc(struct net
10469 static void network_alloc_rx_buffers(struct net_device *);
10470 static void send_fake_arp(struct net_device *);
10471
10472 -static irqreturn_t netif_int(int irq, void *dev_id, struct pt_regs *ptregs);
10473 +static irqreturn_t netif_int(int irq, void *dev_id);
10474
10475 #ifdef CONFIG_SYSFS
10476 static int xennet_sysfs_addif(struct net_device *netdev);
10477 @@ -992,7 +992,7 @@ static int network_start_xmit(struct sk_
10478 tx->flags = 0;
10479 extra = NULL;
10480
10481 - if (skb->ip_summed == CHECKSUM_HW) /* local packet? */
10482 + if (skb->ip_summed == CHECKSUM_PARTIAL) /* local packet? */
10483 tx->flags |= NETTXF_csum_blank | NETTXF_data_validated;
10484 #ifdef CONFIG_XEN
10485 if (skb->proto_data_valid) /* remote but checksummed? */
10486 @@ -1049,7 +1049,7 @@ static int network_start_xmit(struct sk_
10487 return 0;
10488 }
10489
10490 -static irqreturn_t netif_int(int irq, void *dev_id, struct pt_regs *ptregs)
10491 +static irqreturn_t netif_int(int irq, void *dev_id)
10492 {
10493 struct net_device *dev = dev_id;
10494 struct netfront_info *np = netdev_priv(dev);
10495 --- sle11-2009-05-14.orig/drivers/xen/pciback/pciback.h 2009-05-14 11:02:43.000000000 +0200
10496 +++ sle11-2009-05-14/drivers/xen/pciback/pciback.h 2009-03-04 11:28:34.000000000 +0100
10497 @@ -87,7 +87,7 @@ int pciback_publish_pci_roots(struct pci
10498 void pciback_release_devices(struct pciback_device *pdev);
10499
10500 /* Handles events from front-end */
10501 -irqreturn_t pciback_handle_event(int irq, void *dev_id, struct pt_regs *regs);
10502 +irqreturn_t pciback_handle_event(int irq, void *dev_id);
10503 void pciback_do_op(void *data);
10504
10505 int pciback_xenbus_register(void);
10506 --- sle11-2009-05-14.orig/drivers/xen/pciback/pciback_ops.c 2009-05-14 11:02:43.000000000 +0200
10507 +++ sle11-2009-05-14/drivers/xen/pciback/pciback_ops.c 2009-03-04 11:28:34.000000000 +0100
10508 @@ -107,7 +107,7 @@ void pciback_do_op(void *data)
10509 test_and_schedule_op(pdev);
10510 }
10511
10512 -irqreturn_t pciback_handle_event(int irq, void *dev_id, struct pt_regs *regs)
10513 +irqreturn_t pciback_handle_event(int irq, void *dev_id)
10514 {
10515 struct pciback_device *pdev = dev_id;
10516
10517 --- sle11-2009-05-14.orig/drivers/xen/pcifront/pci_op.c 2009-05-14 11:02:43.000000000 +0200
10518 +++ sle11-2009-05-14/drivers/xen/pcifront/pci_op.c 2009-03-04 11:28:34.000000000 +0100
10519 @@ -508,10 +508,16 @@ int __devinit pcifront_rescan_root(struc
10520
10521 d = pci_scan_single_device(b, devfn);
10522 if (d) {
10523 + int err;
10524 +
10525 dev_info(&pdev->xdev->dev, "New device on "
10526 "%04x:%02x:%02x.%02x found.\n", domain, bus,
10527 PCI_SLOT(devfn), PCI_FUNC(devfn));
10528 - pci_bus_add_device(d);
10529 + err = pci_bus_add_device(d);
10530 + if (err)
10531 + dev_err(&pdev->xdev->dev,
10532 + "error %d adding device, continuing.\n",
10533 + err);
10534 }
10535 }
10536
10537 --- sle11-2009-05-14.orig/drivers/xen/privcmd/compat_privcmd.c 2009-05-14 11:02:43.000000000 +0200
10538 +++ sle11-2009-05-14/drivers/xen/privcmd/compat_privcmd.c 2009-03-04 11:28:34.000000000 +0100
10539 @@ -18,7 +18,6 @@
10540 * Authors: Jimi Xenidis <jimix@watson.ibm.com>
10541 */
10542
10543 -#include <linux/config.h>
10544 #include <linux/compat.h>
10545 #include <linux/ioctl.h>
10546 #include <linux/syscalls.h>
10547 --- sle11-2009-05-14.orig/drivers/xen/privcmd/privcmd.c 2009-05-14 11:02:43.000000000 +0200
10548 +++ sle11-2009-05-14/drivers/xen/privcmd/privcmd.c 2009-03-04 11:28:34.000000000 +0100
10549 @@ -40,7 +40,7 @@ static int privcmd_enforce_singleshot_ma
10550 static long privcmd_ioctl(struct file *file,
10551 unsigned int cmd, unsigned long data)
10552 {
10553 - int ret = -ENOSYS;
10554 + long ret = -ENOSYS;
10555 void __user *udata = (void __user *) data;
10556
10557 switch (cmd) {
10558 @@ -50,42 +50,15 @@ static long privcmd_ioctl(struct file *f
10559 if (copy_from_user(&hypercall, udata, sizeof(hypercall)))
10560 return -EFAULT;
10561
10562 -#if defined(__i386__)
10563 +#ifdef CONFIG_X86
10564 if (hypercall.op >= (PAGE_SIZE >> 5))
10565 break;
10566 - __asm__ __volatile__ (
10567 - "pushl %%ebx; pushl %%ecx; pushl %%edx; "
10568 - "pushl %%esi; pushl %%edi; "
10569 - "movl 8(%%eax),%%ebx ;"
10570 - "movl 16(%%eax),%%ecx ;"
10571 - "movl 24(%%eax),%%edx ;"
10572 - "movl 32(%%eax),%%esi ;"
10573 - "movl 40(%%eax),%%edi ;"
10574 - "movl (%%eax),%%eax ;"
10575 - "shll $5,%%eax ;"
10576 - "addl $hypercall_page,%%eax ;"
10577 - "call *%%eax ;"
10578 - "popl %%edi; popl %%esi; popl %%edx; "
10579 - "popl %%ecx; popl %%ebx"
10580 - : "=a" (ret) : "0" (&hypercall) : "memory" );
10581 -#elif defined (__x86_64__)
10582 - if (hypercall.op < (PAGE_SIZE >> 5)) {
10583 - long ign1, ign2, ign3;
10584 - __asm__ __volatile__ (
10585 - "movq %8,%%r10; movq %9,%%r8;"
10586 - "shll $5,%%eax ;"
10587 - "addq $hypercall_page,%%rax ;"
10588 - "call *%%rax"
10589 - : "=a" (ret), "=D" (ign1),
10590 - "=S" (ign2), "=d" (ign3)
10591 - : "0" ((unsigned int)hypercall.op),
10592 - "1" (hypercall.arg[0]),
10593 - "2" (hypercall.arg[1]),
10594 - "3" (hypercall.arg[2]),
10595 - "g" (hypercall.arg[3]),
10596 - "g" (hypercall.arg[4])
10597 - : "r8", "r10", "memory" );
10598 - }
10599 + ret = _hypercall(long, (unsigned int)hypercall.op,
10600 + (unsigned long)hypercall.arg[0],
10601 + (unsigned long)hypercall.arg[1],
10602 + (unsigned long)hypercall.arg[2],
10603 + (unsigned long)hypercall.arg[3],
10604 + (unsigned long)hypercall.arg[4]);
10605 #else
10606 ret = privcmd_hypercall(&hypercall);
10607 #endif
10608 @@ -306,7 +279,7 @@ static int privcmd_mmap(struct file * fi
10609 return -ENOSYS;
10610
10611 /* DONTCOPY is essential for Xen as copy_page_range is broken. */
10612 - vma->vm_flags |= VM_RESERVED | VM_IO | VM_DONTCOPY;
10613 + vma->vm_flags |= VM_RESERVED | VM_IO | VM_PFNMAP | VM_DONTCOPY;
10614 vma->vm_ops = &privcmd_vm_ops;
10615 vma->vm_private_data = NULL;
10616
10617 --- sle11-2009-05-14.orig/drivers/xen/scsiback/common.h 2009-05-14 11:02:43.000000000 +0200
10618 +++ sle11-2009-05-14/drivers/xen/scsiback/common.h 2009-03-04 11:28:34.000000000 +0100
10619 @@ -142,7 +142,7 @@ typedef struct {
10620 #define VSCSIIF_TIMEOUT (900*HZ)
10621
10622
10623 -irqreturn_t scsiback_intr(int, void *, struct pt_regs *);
10624 +irqreturn_t scsiback_intr(int, void *);
10625 int scsiback_init_sring(struct vscsibk_info *info,
10626 unsigned long ring_ref, unsigned int evtchn);
10627 int scsiback_schedule(void *data);
10628 --- sle11-2009-05-14.orig/drivers/xen/scsiback/scsiback.c 2009-05-14 11:02:43.000000000 +0200
10629 +++ sle11-2009-05-14/drivers/xen/scsiback/scsiback.c 2009-03-04 11:28:34.000000000 +0100
10630 @@ -440,7 +440,7 @@ void scsiback_cmd_exec(pending_req_t *pe
10631 write = (data_dir == DMA_TO_DEVICE);
10632 rq = blk_get_request(pending_req->sdev->request_queue, write, GFP_KERNEL);
10633
10634 - rq->flags |= REQ_BLOCK_PC;
10635 + rq->cmd_type = REQ_TYPE_BLOCK_PC;
10636 rq->cmd_len = cmd_len;
10637 memcpy(rq->cmd, pending_req->cmnd, cmd_len);
10638
10639 @@ -484,7 +484,7 @@ static void scsiback_device_reset_exec(p
10640 }
10641
10642
10643 -irqreturn_t scsiback_intr(int irq, void *dev_id, struct pt_regs *regs)
10644 +irqreturn_t scsiback_intr(int irq, void *dev_id)
10645 {
10646 scsiback_notify_work((struct vscsibk_info *)dev_id);
10647 return IRQ_HANDLED;
10648 --- sle11-2009-05-14.orig/drivers/xen/scsifront/common.h 2009-05-14 11:02:43.000000000 +0200
10649 +++ sle11-2009-05-14/drivers/xen/scsifront/common.h 2009-03-04 11:28:34.000000000 +0100
10650 @@ -122,7 +122,7 @@ struct vscsifrnt_info {
10651 int scsifront_xenbus_init(void);
10652 void scsifront_xenbus_unregister(void);
10653 int scsifront_schedule(void *data);
10654 -irqreturn_t scsifront_intr(int irq, void *dev_id, struct pt_regs *ptregs);
10655 +irqreturn_t scsifront_intr(int irq, void *dev_id);
10656 int scsifront_cmd_done(struct vscsifrnt_info *info);
10657
10658
10659 --- sle11-2009-05-14.orig/drivers/xen/scsifront/scsifront.c 2009-05-14 11:02:43.000000000 +0200
10660 +++ sle11-2009-05-14/drivers/xen/scsifront/scsifront.c 2009-03-04 11:28:34.000000000 +0100
10661 @@ -100,7 +100,7 @@ static void scsifront_do_request(struct
10662 notify_remote_via_irq(irq);
10663 }
10664
10665 -irqreturn_t scsifront_intr(int irq, void *dev_id, struct pt_regs *ptregs)
10666 +irqreturn_t scsifront_intr(int irq, void *dev_id)
10667 {
10668 scsifront_notify_work((struct vscsifrnt_info *)dev_id);
10669 return IRQ_HANDLED;
10670 --- sle11-2009-05-14.orig/drivers/xen/sfc_netback/accel_xenbus.c 2009-05-14 11:02:43.000000000 +0200
10671 +++ sle11-2009-05-14/drivers/xen/sfc_netback/accel_xenbus.c 2009-03-04 11:28:34.000000000 +0100
10672 @@ -68,8 +68,7 @@ static void unlink_bend(struct netback_a
10673
10674
10675 /* Demultiplex a message IRQ from the frontend driver. */
10676 -static irqreturn_t msgirq_from_frontend(int irq, void *context,
10677 - struct pt_regs *unused)
10678 +static irqreturn_t msgirq_from_frontend(int irq, void *context)
10679 {
10680 struct xenbus_device *dev = context;
10681 struct netback_accel *bend = NETBACK_ACCEL_FROM_XENBUS_DEVICE(dev);
10682 @@ -84,8 +83,7 @@ static irqreturn_t msgirq_from_frontend(
10683 * functionally, but we need it to pass to the bind function, and may
10684 * get called spuriously
10685 */
10686 -static irqreturn_t netirq_from_frontend(int irq, void *context,
10687 - struct pt_regs *unused)
10688 +static irqreturn_t netirq_from_frontend(int irq, void *context)
10689 {
10690 VPRINTK("netirq %d from device %s\n", irq,
10691 ((struct xenbus_device *)context)->nodename);
10692 --- sle11-2009-05-14.orig/drivers/xen/sfc_netfront/accel.h 2009-04-09 14:41:38.000000000 +0200
10693 +++ sle11-2009-05-14/drivers/xen/sfc_netfront/accel.h 2009-03-30 16:34:56.000000000 +0200
10694 @@ -451,10 +451,8 @@ void netfront_accel_msg_tx_fastpath(netf
10695 u32 ip, u16 port, u8 protocol);
10696
10697 /* Process an IRQ received from back end driver */
10698 -irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context,
10699 - struct pt_regs *unused);
10700 -irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context,
10701 - struct pt_regs *unused);
10702 +irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context);
10703 +irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context);
10704
10705 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
10706 extern void netfront_accel_msg_from_bend(struct work_struct *context);
10707 --- sle11-2009-05-14.orig/drivers/xen/sfc_netfront/accel_msg.c 2009-05-14 11:02:43.000000000 +0200
10708 +++ sle11-2009-05-14/drivers/xen/sfc_netfront/accel_msg.c 2009-03-04 11:28:34.000000000 +0100
10709 @@ -490,8 +490,7 @@ void netfront_accel_msg_from_bend(void *
10710 }
10711
10712
10713 -irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context,
10714 - struct pt_regs *unused)
10715 +irqreturn_t netfront_accel_msg_channel_irq_from_bend(int irq, void *context)
10716 {
10717 netfront_accel_vnic *vnic = (netfront_accel_vnic *)context;
10718 VPRINTK("irq %d from device %s\n", irq, vnic->dev->nodename);
10719 @@ -502,8 +501,7 @@ irqreturn_t netfront_accel_msg_channel_i
10720 }
10721
10722 /* Process an interrupt received from the NIC via backend */
10723 -irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context,
10724 - struct pt_regs *unused)
10725 +irqreturn_t netfront_accel_net_channel_irq_from_bend(int irq, void *context)
10726 {
10727 netfront_accel_vnic *vnic = (netfront_accel_vnic *)context;
10728 struct net_device *net_dev = vnic->net_dev;
10729 --- sle11-2009-05-14.orig/drivers/xen/sfc_netfront/accel_tso.c 2009-05-14 11:02:43.000000000 +0200
10730 +++ sle11-2009-05-14/drivers/xen/sfc_netfront/accel_tso.c 2009-03-04 11:28:34.000000000 +0100
10731 @@ -363,7 +363,7 @@ int netfront_accel_enqueue_skb_tso(netfr
10732
10733 tso_check_safe(skb);
10734
10735 - if (skb->ip_summed != CHECKSUM_HW)
10736 + if (skb->ip_summed != CHECKSUM_PARTIAL)
10737 EPRINTK("Trying to TSO send a packet without HW checksum\n");
10738
10739 tso_start(&state, skb);
10740 --- sle11-2009-05-14.orig/drivers/xen/sfc_netfront/accel_vi.c 2009-04-09 14:41:38.000000000 +0200
10741 +++ sle11-2009-05-14/drivers/xen/sfc_netfront/accel_vi.c 2009-03-30 16:35:11.000000000 +0200
10742 @@ -461,7 +461,7 @@ netfront_accel_enqueue_skb_multi(netfron
10743
10744 frag_i = -1;
10745
10746 - if (skb->ip_summed == CHECKSUM_HW) {
10747 + if (skb->ip_summed == CHECKSUM_PARTIAL) {
10748 /* Set to zero to encourage falcon to work it out for us */
10749 *(u16*)(skb->h.raw + skb->csum) = 0;
10750 }
10751 @@ -580,7 +580,7 @@ netfront_accel_enqueue_skb_single(netfro
10752
10753 kva = buf->pkt_kva;
10754
10755 - if (skb->ip_summed == CHECKSUM_HW) {
10756 + if (skb->ip_summed == CHECKSUM_PARTIAL) {
10757 /* Set to zero to encourage falcon to work it out for us */
10758 *(u16*)(skb->h.raw + skb->csum) = 0;
10759 }
10760 --- sle11-2009-05-14.orig/drivers/xen/tpmback/common.h 2009-05-14 11:02:43.000000000 +0200
10761 +++ sle11-2009-05-14/drivers/xen/tpmback/common.h 2009-03-04 11:28:34.000000000 +0100
10762 @@ -61,7 +61,7 @@ void tpmif_deschedule_work(tpmif_t * tpm
10763 void tpmif_xenbus_init(void);
10764 void tpmif_xenbus_exit(void);
10765 int tpmif_map(tpmif_t *tpmif, unsigned long shared_page, unsigned int evtchn);
10766 -irqreturn_t tpmif_be_int(int irq, void *dev_id, struct pt_regs *regs);
10767 +irqreturn_t tpmif_be_int(int irq, void *dev_id);
10768
10769 long int tpmback_get_instance(struct backend_info *bi);
10770
10771 --- sle11-2009-05-14.orig/drivers/xen/tpmback/tpmback.c 2009-05-14 11:02:43.000000000 +0200
10772 +++ sle11-2009-05-14/drivers/xen/tpmback/tpmback.c 2009-03-04 11:28:34.000000000 +0100
10773 @@ -502,7 +502,7 @@ static ssize_t vtpm_op_read(struct file
10774 list_del(&pak->next);
10775 write_unlock_irqrestore(&dataex.pak_lock, flags);
10776
10777 - DPRINTK("size given by app: %d, available: %d\n", size, left);
10778 + DPRINTK("size given by app: %zu, available: %u\n", size, left);
10779
10780 ret_size = min_t(size_t, size, left);
10781
10782 @@ -899,7 +899,7 @@ static void tpm_tx_action(unsigned long
10783 }
10784 }
10785
10786 -irqreturn_t tpmif_be_int(int irq, void *dev_id, struct pt_regs *regs)
10787 +irqreturn_t tpmif_be_int(int irq, void *dev_id)
10788 {
10789 tpmif_t *tpmif = (tpmif_t *) dev_id;
10790
10791 --- sle11-2009-05-14.orig/drivers/xen/xenbus/xenbus_comms.c 2008-11-25 12:35:56.000000000 +0100
10792 +++ sle11-2009-05-14/drivers/xen/xenbus/xenbus_comms.c 2009-03-04 11:28:34.000000000 +0100
10793 @@ -55,7 +55,7 @@ static DECLARE_WORK(probe_work, xenbus_p
10794
10795 static DECLARE_WAIT_QUEUE_HEAD(xb_waitq);
10796
10797 -static irqreturn_t wake_waiting(int irq, void *unused, struct pt_regs *regs)
10798 +static irqreturn_t wake_waiting(int irq, void *unused)
10799 {
10800 if (unlikely(xenstored_ready == 0)) {
10801 xenstored_ready = 1;
10802 --- sle11-2009-05-14.orig/drivers/xen/xenoprof/xenoprofile.c 2009-05-14 11:02:43.000000000 +0200
10803 +++ sle11-2009-05-14/drivers/xen/xenoprof/xenoprofile.c 2009-03-04 11:28:34.000000000 +0100
10804 @@ -194,8 +194,7 @@ done:
10805 oprofile_add_domain_switch(COORDINATOR_DOMAIN);
10806 }
10807
10808 -static irqreturn_t
10809 -xenoprof_ovf_interrupt(int irq, void * dev_id, struct pt_regs * regs)
10810 +static irqreturn_t xenoprof_ovf_interrupt(int irq, void *dev_id)
10811 {
10812 struct xenoprof_buf * buf;
10813 static unsigned long flag;
10814 --- sle11-2009-05-14.orig/include/asm-generic/pgtable.h 2009-02-16 15:58:14.000000000 +0100
10815 +++ sle11-2009-05-14/include/asm-generic/pgtable.h 2009-03-04 11:28:34.000000000 +0100
10816 @@ -100,7 +100,7 @@ static inline void ptep_set_wrprotect(st
10817 #endif
10818
10819 #ifndef arch_change_pte_range
10820 -#define arch_change_pte_range(mm, pmd, addr, end, newprot) 0
10821 +#define arch_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) 0
10822 #endif
10823
10824 #ifndef __HAVE_ARCH_PTE_SAME
10825 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/desc_32.h 2009-05-14 11:02:43.000000000 +0200
10826 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/desc_32.h 2009-03-04 11:28:34.000000000 +0100
10827 @@ -32,52 +32,110 @@ static inline struct desc_struct *get_cp
10828 return (struct desc_struct *)per_cpu(cpu_gdt_descr, cpu).address;
10829 }
10830
10831 +/*
10832 + * This is the ldt that every process will get unless we need
10833 + * something other than this.
10834 + */
10835 +extern struct desc_struct default_ldt[];
10836 +extern struct desc_struct idt_table[];
10837 +extern void set_intr_gate(unsigned int irq, void * addr);
10838 +
10839 +static inline void pack_descriptor(__u32 *a, __u32 *b,
10840 + unsigned long base, unsigned long limit, unsigned char type, unsigned char flags)
10841 +{
10842 + *a = ((base & 0xffff) << 16) | (limit & 0xffff);
10843 + *b = (base & 0xff000000) | ((base & 0xff0000) >> 16) |
10844 + (limit & 0x000f0000) | ((type & 0xff) << 8) | ((flags & 0xf) << 20);
10845 +}
10846 +
10847 +static inline void pack_gate(__u32 *a, __u32 *b,
10848 + unsigned long base, unsigned short seg, unsigned char type, unsigned char flags)
10849 +{
10850 + *a = (seg << 16) | (base & 0xffff);
10851 + *b = (base & 0xffff0000) | ((type & 0xff) << 8) | (flags & 0xff);
10852 +}
10853 +
10854 +#define DESCTYPE_LDT 0x82 /* present, system, DPL-0, LDT */
10855 +#define DESCTYPE_TSS 0x89 /* present, system, DPL-0, 32-bit TSS */
10856 +#define DESCTYPE_TASK 0x85 /* present, system, DPL-0, task gate */
10857 +#define DESCTYPE_INT 0x8e /* present, system, DPL-0, interrupt gate */
10858 +#define DESCTYPE_TRAP 0x8f /* present, system, DPL-0, trap gate */
10859 +#define DESCTYPE_DPL3 0x60 /* DPL-3 */
10860 +#define DESCTYPE_S 0x10 /* !system */
10861 +
10862 #define load_TR_desc() __asm__ __volatile__("ltr %w0"::"q" (GDT_ENTRY_TSS*8))
10863 #define load_LDT_desc() __asm__ __volatile__("lldt %w0"::"q" (GDT_ENTRY_LDT*8))
10864
10865 #define load_gdt(dtr) __asm__ __volatile("lgdt %0"::"m" (*dtr))
10866 #define load_idt(dtr) __asm__ __volatile("lidt %0"::"m" (*dtr))
10867 -#define load_tr(tr) __asm__ __volatile("ltr %0"::"mr" (tr))
10868 -#define load_ldt(ldt) __asm__ __volatile("lldt %0"::"mr" (ldt))
10869 +#define load_tr(tr) __asm__ __volatile("ltr %0"::"m" (tr))
10870 +#define load_ldt(ldt) __asm__ __volatile("lldt %0"::"m" (ldt))
10871
10872 #define store_gdt(dtr) __asm__ ("sgdt %0":"=m" (*dtr))
10873 #define store_idt(dtr) __asm__ ("sidt %0":"=m" (*dtr))
10874 -#define store_tr(tr) __asm__ ("str %0":"=mr" (tr))
10875 -#define store_ldt(ldt) __asm__ ("sldt %0":"=mr" (ldt))
10876 +#define store_tr(tr) __asm__ ("str %0":"=m" (tr))
10877 +#define store_ldt(ldt) __asm__ ("sldt %0":"=m" (ldt))
10878
10879 -/*
10880 - * This is the ldt that every process will get unless we need
10881 - * something other than this.
10882 - */
10883 -extern struct desc_struct default_ldt[];
10884 -extern void set_intr_gate(unsigned int irq, void * addr);
10885 +#if TLS_SIZE != 24
10886 +# error update this code.
10887 +#endif
10888 +
10889 +static inline void load_TLS(struct thread_struct *t, unsigned int cpu)
10890 +{
10891 +#define C(i) if (HYPERVISOR_update_descriptor(virt_to_machine(&get_cpu_gdt_table(cpu)[GDT_ENTRY_TLS_MIN + i]), \
10892 + *(u64 *)&t->tls_array[i]) \
10893 + BUG()
10894 + C(0); C(1); C(2);
10895 +#undef C
10896 +}
10897
10898 -#define _set_tssldt_desc(n,addr,limit,type) \
10899 -__asm__ __volatile__ ("movw %w3,0(%2)\n\t" \
10900 - "movw %w1,2(%2)\n\t" \
10901 - "rorl $16,%1\n\t" \
10902 - "movb %b1,4(%2)\n\t" \
10903 - "movb %4,5(%2)\n\t" \
10904 - "movb $0,6(%2)\n\t" \
10905 - "movb %h1,7(%2)\n\t" \
10906 - "rorl $16,%1" \
10907 - : "=m"(*(n)) : "q" (addr), "r"(n), "ir"(limit), "i"(type))
10908 +#ifndef CONFIG_XEN
10909 +static inline void write_dt_entry(void *dt, int entry, __u32 entry_a, __u32 entry_b)
10910 +{
10911 + __u32 *lp = (__u32 *)((char *)dt + entry*8);
10912 + *lp = entry_a;
10913 + *(lp+1) = entry_b;
10914 +}
10915
10916 -#ifndef CONFIG_X86_NO_TSS
10917 -static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, void *addr)
10918 +#define write_ldt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
10919 +#define write_gdt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
10920 +#else
10921 +extern int write_ldt_entry(void *ldt, int entry, __u32 entry_a, __u32 entry_b);
10922 +extern int write_gdt_entry(void *gdt, int entry, __u32 entry_a, __u32 entry_b);
10923 +#endif
10924 +#ifndef CONFIG_X86_NO_IDT
10925 +#define write_idt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
10926 +
10927 +static inline void _set_gate(int gate, unsigned int type, void *addr, unsigned short seg)
10928 {
10929 - _set_tssldt_desc(&get_cpu_gdt_table(cpu)[entry], (int)addr,
10930 - offsetof(struct tss_struct, __cacheline_filler) - 1, 0x89);
10931 + __u32 a, b;
10932 + pack_gate(&a, &b, (unsigned long)addr, seg, type, 0);
10933 + write_idt_entry(idt_table, gate, a, b);
10934 }
10935 +#endif
10936
10937 -#define set_tss_desc(cpu,addr) __set_tss_desc(cpu, GDT_ENTRY_TSS, addr)
10938 +#ifndef CONFIG_X86_NO_TSS
10939 +static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, const void *addr)
10940 +{
10941 + __u32 a, b;
10942 + pack_descriptor(&a, &b, (unsigned long)addr,
10943 + offsetof(struct tss_struct, __cacheline_filler) - 1,
10944 + DESCTYPE_TSS, 0);
10945 + write_gdt_entry(get_cpu_gdt_table(cpu), entry, a, b);
10946 +}
10947 #endif
10948
10949 -static inline void set_ldt_desc(unsigned int cpu, void *addr, unsigned int size)
10950 +static inline void set_ldt_desc(unsigned int cpu, void *addr, unsigned int entries)
10951 {
10952 - _set_tssldt_desc(&get_cpu_gdt_table(cpu)[GDT_ENTRY_LDT], (int)addr, ((size << 3)-1), 0x82);
10953 + __u32 a, b;
10954 + pack_descriptor(&a, &b, (unsigned long)addr,
10955 + entries * sizeof(struct desc_struct) - 1,
10956 + DESCTYPE_LDT, 0);
10957 + write_gdt_entry(get_cpu_gdt_table(cpu), GDT_ENTRY_LDT, a, b);
10958 }
10959
10960 +#define set_tss_desc(cpu,addr) __set_tss_desc(cpu, GDT_ENTRY_TSS, addr)
10961 +
10962 #define LDT_entry_a(info) \
10963 ((((info)->base_addr & 0x0000ffff) << 16) | ((info)->limit & 0x0ffff))
10964
10965 @@ -103,21 +161,6 @@ static inline void set_ldt_desc(unsigned
10966 (info)->seg_not_present == 1 && \
10967 (info)->useable == 0 )
10968
10969 -extern int write_ldt_entry(void *ldt, int entry, __u32 entry_a, __u32 entry_b);
10970 -
10971 -#if TLS_SIZE != 24
10972 -# error update this code.
10973 -#endif
10974 -
10975 -static inline void load_TLS(struct thread_struct *t, unsigned int cpu)
10976 -{
10977 -#define C(i) if (HYPERVISOR_update_descriptor(virt_to_machine(&get_cpu_gdt_table(cpu)[GDT_ENTRY_TLS_MIN + i]), \
10978 - *(u64 *)&t->tls_array[i])) \
10979 - BUG();
10980 - C(0); C(1); C(2);
10981 -#undef C
10982 -}
10983 -
10984 static inline void clear_LDT(void)
10985 {
10986 int cpu = get_cpu();
10987 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/fixmap_32.h 2009-05-14 11:02:43.000000000 +0200
10988 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/fixmap_32.h 2009-03-04 11:28:34.000000000 +0100
10989 @@ -55,7 +55,7 @@ enum fixed_addresses {
10990 #ifdef CONFIG_X86_LOCAL_APIC
10991 FIX_APIC_BASE, /* local (CPU) APIC) -- required for SMP or not */
10992 #endif
10993 -#ifdef CONFIG_X86_IO_APIC
10994 +#if defined(CONFIG_X86_IO_APIC) && !defined(CONFIG_XEN)
10995 FIX_IO_APIC_BASE_0,
10996 FIX_IO_APIC_BASE_END = FIX_IO_APIC_BASE_0 + MAX_IO_APICS-1,
10997 #endif
10998 @@ -95,10 +95,9 @@ enum fixed_addresses {
10999 __end_of_fixed_addresses
11000 };
11001
11002 -extern void set_fixaddr_top(unsigned long top);
11003 -
11004 extern void __set_fixmap(enum fixed_addresses idx,
11005 maddr_t phys, pgprot_t flags);
11006 +extern void reserve_top_address(unsigned long reserve);
11007
11008 #define set_fixmap(idx, phys) \
11009 __set_fixmap(idx, phys, PAGE_KERNEL)
11010 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/hypercall_32.h 2009-05-14 11:02:43.000000000 +0200
11011 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/hypercall_32.h 2009-03-04 11:28:34.000000000 +0100
11012 @@ -128,6 +128,23 @@
11013 __res; \
11014 })
11015
11016 +#define _hypercall(type, op, a1, a2, a3, a4, a5) \
11017 +({ \
11018 + type __res; \
11019 + register typeof((a1)+0) __arg1 asm("ebx") = (a1); \
11020 + register typeof((a2)+0) __arg2 asm("ecx") = (a2); \
11021 + register typeof((a3)+0) __arg3 asm("edx") = (a3); \
11022 + register typeof((a4)+0) __arg4 asm("esi") = (a4); \
11023 + register typeof((a5)+0) __arg5 asm("edi") = (a5); \
11024 + asm volatile ( \
11025 + "call *%6" \
11026 + : "=a" (__res), "+r" (__arg1), "+r" (__arg2), \
11027 + "+r" (__arg3), "+r" (__arg4), "+r" (__arg5) \
11028 + : "0" (hypercall_page + (op) * 32) \
11029 + : "memory" ); \
11030 + __res; \
11031 +})
11032 +
11033 static inline int __must_check
11034 HYPERVISOR_set_trap_table(
11035 const trap_info_t *table)
11036 @@ -140,6 +157,8 @@ HYPERVISOR_mmu_update(
11037 mmu_update_t *req, unsigned int count, unsigned int *success_count,
11038 domid_t domid)
11039 {
11040 + if (arch_use_lazy_mmu_mode())
11041 + return xen_multi_mmu_update(req, count, success_count, domid);
11042 return _hypercall4(int, mmu_update, req, count, success_count, domid);
11043 }
11044
11045 @@ -148,6 +167,8 @@ HYPERVISOR_mmuext_op(
11046 struct mmuext_op *op, unsigned int count, unsigned int *success_count,
11047 domid_t domid)
11048 {
11049 + if (arch_use_lazy_mmu_mode())
11050 + return xen_multi_mmuext_op(op, count, success_count, domid);
11051 return _hypercall4(int, mmuext_op, op, count, success_count, domid);
11052 }
11053
11054 @@ -238,6 +259,8 @@ static inline int __must_check
11055 HYPERVISOR_memory_op(
11056 unsigned int cmd, void *arg)
11057 {
11058 + if (arch_use_lazy_mmu_mode())
11059 + xen_multicall_flush(false);
11060 return _hypercall2(int, memory_op, cmd, arg);
11061 }
11062
11063 @@ -253,6 +276,9 @@ HYPERVISOR_update_va_mapping(
11064 unsigned long va, pte_t new_val, unsigned long flags)
11065 {
11066 unsigned long pte_hi = 0;
11067 +
11068 + if (arch_use_lazy_mmu_mode())
11069 + return xen_multi_update_va_mapping(va, new_val, flags);
11070 #ifdef CONFIG_X86_PAE
11071 pte_hi = new_val.pte_high;
11072 #endif
11073 @@ -316,6 +342,8 @@ static inline int __must_check
11074 HYPERVISOR_grant_table_op(
11075 unsigned int cmd, void *uop, unsigned int count)
11076 {
11077 + if (arch_use_lazy_mmu_mode())
11078 + xen_multicall_flush(false);
11079 return _hypercall3(int, grant_table_op, cmd, uop, count);
11080 }
11081
11082 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/hypercall_64.h 2009-05-14 11:02:43.000000000 +0200
11083 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/hypercall_64.h 2009-03-04 11:28:34.000000000 +0100
11084 @@ -134,6 +134,23 @@
11085 __res; \
11086 })
11087
11088 +#define _hypercall(type, op, a1, a2, a3, a4, a5) \
11089 +({ \
11090 + type __res; \
11091 + register typeof((a1)+0) __arg1 asm("rdi") = (a1); \
11092 + register typeof((a2)+0) __arg2 asm("rsi") = (a2); \
11093 + register typeof((a3)+0) __arg3 asm("rdx") = (a3); \
11094 + register typeof((a4)+0) __arg4 asm("r10") = (a4); \
11095 + register typeof((a5)+0) __arg5 asm("r8") = (a5); \
11096 + asm volatile ( \
11097 + "call *%6" \
11098 + : "=a" (__res), "+r" (__arg1), "+r" (__arg2), \
11099 + "+r" (__arg3), "+r" (__arg4), "+r" (__arg5) \
11100 + : "0" (hypercall_page + (op) * 32) \
11101 + : "memory" ); \
11102 + __res; \
11103 +})
11104 +
11105 static inline int __must_check
11106 HYPERVISOR_set_trap_table(
11107 const trap_info_t *table)
11108 @@ -146,6 +163,8 @@ HYPERVISOR_mmu_update(
11109 mmu_update_t *req, unsigned int count, unsigned int *success_count,
11110 domid_t domid)
11111 {
11112 + if (arch_use_lazy_mmu_mode())
11113 + return xen_multi_mmu_update(req, count, success_count, domid);
11114 return _hypercall4(int, mmu_update, req, count, success_count, domid);
11115 }
11116
11117 @@ -154,6 +173,8 @@ HYPERVISOR_mmuext_op(
11118 struct mmuext_op *op, unsigned int count, unsigned int *success_count,
11119 domid_t domid)
11120 {
11121 + if (arch_use_lazy_mmu_mode())
11122 + return xen_multi_mmuext_op(op, count, success_count, domid);
11123 return _hypercall4(int, mmuext_op, op, count, success_count, domid);
11124 }
11125
11126 @@ -241,6 +262,8 @@ static inline int __must_check
11127 HYPERVISOR_memory_op(
11128 unsigned int cmd, void *arg)
11129 {
11130 + if (arch_use_lazy_mmu_mode())
11131 + xen_multicall_flush(false);
11132 return _hypercall2(int, memory_op, cmd, arg);
11133 }
11134
11135 @@ -255,6 +278,8 @@ static inline int __must_check
11136 HYPERVISOR_update_va_mapping(
11137 unsigned long va, pte_t new_val, unsigned long flags)
11138 {
11139 + if (arch_use_lazy_mmu_mode())
11140 + return xen_multi_update_va_mapping(va, new_val, flags);
11141 return _hypercall3(int, update_va_mapping, va, new_val.pte, flags);
11142 }
11143
11144 @@ -314,6 +339,8 @@ static inline int __must_check
11145 HYPERVISOR_grant_table_op(
11146 unsigned int cmd, void *uop, unsigned int count)
11147 {
11148 + if (arch_use_lazy_mmu_mode())
11149 + xen_multicall_flush(false);
11150 return _hypercall3(int, grant_table_op, cmd, uop, count);
11151 }
11152
11153 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/hypervisor.h 2009-05-14 11:02:43.000000000 +0200
11154 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/hypervisor.h 2009-03-04 11:28:34.000000000 +0100
11155 @@ -43,6 +43,7 @@
11156 #include <xen/interface/physdev.h>
11157 #include <xen/interface/sched.h>
11158 #include <xen/interface/nmi.h>
11159 +#include <asm/percpu.h>
11160 #include <asm/ptrace.h>
11161 #include <asm/page.h>
11162 #if defined(__i386__)
11163 @@ -135,7 +136,46 @@ void scrub_pages(void *, unsigned int);
11164 #define scrub_pages(_p,_n) ((void)0)
11165 #endif
11166
11167 -#include <xen/hypercall.h>
11168 +#if defined(CONFIG_XEN) && !defined(MODULE)
11169 +
11170 +DECLARE_PER_CPU(bool, xen_lazy_mmu);
11171 +
11172 +int xen_multicall_flush(bool);
11173 +
11174 +int __must_check xen_multi_update_va_mapping(unsigned long va, pte_t,
11175 + unsigned long flags);
11176 +int __must_check xen_multi_mmu_update(mmu_update_t *, unsigned int count,
11177 + unsigned int *success_count, domid_t);
11178 +int __must_check xen_multi_mmuext_op(struct mmuext_op *, unsigned int count,
11179 + unsigned int *success_count, domid_t);
11180 +
11181 +#define __HAVE_ARCH_ENTER_LAZY_MMU_MODE
11182 +static inline void arch_enter_lazy_mmu_mode(void)
11183 +{
11184 + __get_cpu_var(xen_lazy_mmu) = true;
11185 +}
11186 +
11187 +static inline void arch_leave_lazy_mmu_mode(void)
11188 +{
11189 + __get_cpu_var(xen_lazy_mmu) = false;
11190 + xen_multicall_flush(false);
11191 +}
11192 +
11193 +#if defined(CONFIG_X86_32)
11194 +#define arch_use_lazy_mmu_mode() unlikely(x86_read_percpu(xen_lazy_mmu))
11195 +#elif !defined(arch_use_lazy_mmu_mode)
11196 +#define arch_use_lazy_mmu_mode() unlikely(__get_cpu_var(xen_lazy_mmu))
11197 +#endif
11198 +
11199 +#else /* !CONFIG_XEN || MODULE */
11200 +
11201 +static inline void xen_multicall_flush(bool ignore) {}
11202 +#define arch_use_lazy_mmu_mode() false
11203 +#define xen_multi_update_va_mapping(...) ({ BUG(); -ENOSYS; })
11204 +#define xen_multi_mmu_update(...) ({ BUG(); -ENOSYS; })
11205 +#define xen_multi_mmuext_op(...) ({ BUG(); -ENOSYS; })
11206 +
11207 +#endif /* CONFIG_XEN && !MODULE */
11208
11209 #if defined(CONFIG_X86_64)
11210 #define MULTI_UVMFLAGS_INDEX 2
11211 @@ -147,11 +187,15 @@ void scrub_pages(void *, unsigned int);
11212
11213 #ifdef CONFIG_XEN
11214 #define is_running_on_xen() 1
11215 +extern char hypercall_page[PAGE_SIZE];
11216 #else
11217 extern char *hypercall_stubs;
11218 +#define hypercall_page hypercall_stubs
11219 #define is_running_on_xen() (!!hypercall_stubs)
11220 #endif
11221
11222 +#include <xen/hypercall.h>
11223 +
11224 static inline int
11225 HYPERVISOR_yield(
11226 void)
11227 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/pgtable-3level.h 2009-05-14 11:02:43.000000000 +0200
11228 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/pgtable-3level.h 2009-03-04 11:28:34.000000000 +0100
11229 @@ -53,7 +53,6 @@ static inline int pte_exec_kernel(pte_t
11230 * not possible, use pte_get_and_clear to obtain the old pte
11231 * value and then use set_pte to update it. -ben
11232 */
11233 -#define __HAVE_ARCH_SET_PTE_ATOMIC
11234
11235 static inline void set_pte(pte_t *ptep, pte_t pte)
11236 {
11237 @@ -70,14 +69,6 @@ static inline void set_pte(pte_t *ptep,
11238 set_pte((ptep), (pteval)); \
11239 } while (0)
11240
11241 -#define set_pte_at_sync(_mm,addr,ptep,pteval) do { \
11242 - if (((_mm) != current->mm && (_mm) != &init_mm) || \
11243 - HYPERVISOR_update_va_mapping((addr), (pteval), UVMF_INVLPG)) { \
11244 - set_pte((ptep), (pteval)); \
11245 - xen_invlpg((addr)); \
11246 - } \
11247 -} while (0)
11248 -
11249 #define set_pmd(pmdptr,pmdval) \
11250 xen_l2_entry_update((pmdptr), (pmdval))
11251 #define set_pud(pudptr,pudval) \
11252 @@ -94,7 +85,7 @@ static inline void pud_clear (pud_t * pu
11253 #define pud_page(pud) \
11254 ((struct page *) __va(pud_val(pud) & PAGE_MASK))
11255
11256 -#define pud_page_kernel(pud) \
11257 +#define pud_page_vaddr(pud) \
11258 ((unsigned long) __va(pud_val(pud) & PAGE_MASK))
11259
11260
11261 @@ -124,6 +115,7 @@ static inline void pte_clear(struct mm_s
11262
11263 #define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0)
11264
11265 +#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
11266 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
11267 {
11268 pte_t pte = *ptep;
11269 @@ -142,6 +134,7 @@ static inline pte_t ptep_get_and_clear(s
11270 return pte;
11271 }
11272
11273 +#define __HAVE_ARCH_PTEP_CLEAR_FLUSH
11274 #define ptep_clear_flush(vma, addr, ptep) \
11275 ({ \
11276 pte_t *__ptep = (ptep); \
11277 @@ -159,6 +152,7 @@ static inline pte_t ptep_get_and_clear(s
11278 __res; \
11279 })
11280
11281 +#define __HAVE_ARCH_PTE_SAME
11282 static inline int pte_same(pte_t a, pte_t b)
11283 {
11284 return a.pte_low == b.pte_low && a.pte_high == b.pte_high;
11285 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/pgtable_32.h 2008-12-15 11:13:45.000000000 +0100
11286 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/pgtable_32.h 2009-03-04 11:28:34.000000000 +0100
11287 @@ -260,31 +260,89 @@ static inline pte_t pte_mkhuge(pte_t pte
11288 # include <asm/pgtable-2level.h>
11289 #endif
11290
11291 -#define ptep_test_and_clear_dirty(vma, addr, ptep) \
11292 +/*
11293 + * Rules for using pte_update - it must be called after any PTE update which
11294 + * has not been done using the set_pte / clear_pte interfaces. It is used by
11295 + * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
11296 + * updates should either be sets, clears, or set_pte_atomic for P->P
11297 + * transitions, which means this hook should only be called for user PTEs.
11298 + * This hook implies a P->P protection or access change has taken place, which
11299 + * requires a subsequent TLB flush. The notification can optionally be delayed
11300 + * until the TLB flush event by using the pte_update_defer form of the
11301 + * interface, but care must be taken to assure that the flush happens while
11302 + * still holding the same page table lock so that the shadow and primary pages
11303 + * do not become out of sync on SMP.
11304 + */
11305 +#define pte_update(mm, addr, ptep) do { } while (0)
11306 +#define pte_update_defer(mm, addr, ptep) do { } while (0)
11307 +
11308 +
11309 +/*
11310 + * We only update the dirty/accessed state if we set
11311 + * the dirty bit by hand in the kernel, since the hardware
11312 + * will do the accessed bit for us, and we don't want to
11313 + * race with other CPU's that might be updating the dirty
11314 + * bit at the same time.
11315 + */
11316 +#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
11317 +#define ptep_set_access_flags(vma, address, ptep, entry, dirty) \
11318 +do { \
11319 + if (dirty) \
11320 + ptep_establish(vma, address, ptep, entry); \
11321 +} while (0)
11322 +
11323 +/*
11324 + * We don't actually have these, but we want to advertise them so that
11325 + * we can encompass the flush here.
11326 + */
11327 +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
11328 +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
11329 +
11330 +/*
11331 + * Rules for using ptep_establish: the pte MUST be a user pte, and
11332 + * must be a present->present transition.
11333 + */
11334 +#define __HAVE_ARCH_PTEP_ESTABLISH
11335 +#define ptep_establish(vma, address, ptep, pteval) \
11336 +do { \
11337 + if ( likely((vma)->vm_mm == current->mm) ) { \
11338 + BUG_ON(HYPERVISOR_update_va_mapping(address, \
11339 + pteval, \
11340 + (unsigned long)(vma)->vm_mm->cpu_vm_mask.bits| \
11341 + UVMF_INVLPG|UVMF_MULTI)); \
11342 + } else { \
11343 + xen_l1_entry_update(ptep, pteval); \
11344 + flush_tlb_page(vma, address); \
11345 + } \
11346 +} while (0)
11347 +
11348 +#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
11349 +#define ptep_clear_flush_dirty(vma, address, ptep) \
11350 ({ \
11351 pte_t __pte = *(ptep); \
11352 - int __ret = pte_dirty(__pte); \
11353 - if (__ret) { \
11354 - __pte = pte_mkclean(__pte); \
11355 - if ((vma)->vm_mm != current->mm || \
11356 - HYPERVISOR_update_va_mapping(addr, __pte, 0)) \
11357 - (ptep)->pte_low = __pte.pte_low; \
11358 - } \
11359 - __ret; \
11360 + int __dirty = pte_dirty(__pte); \
11361 + __pte = pte_mkclean(__pte); \
11362 + if (test_bit(PG_pinned, &virt_to_page((vma)->vm_mm->pgd)->flags)) \
11363 + ptep_set_access_flags(vma, address, ptep, __pte, __dirty); \
11364 + else if (__dirty) \
11365 + (ptep)->pte_low = __pte.pte_low; \
11366 + __dirty; \
11367 })
11368
11369 -#define ptep_test_and_clear_young(vma, addr, ptep) \
11370 +#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
11371 +#define ptep_clear_flush_young(vma, address, ptep) \
11372 ({ \
11373 pte_t __pte = *(ptep); \
11374 - int __ret = pte_young(__pte); \
11375 - if (__ret) \
11376 - __pte = pte_mkold(__pte); \
11377 - if ((vma)->vm_mm != current->mm || \
11378 - HYPERVISOR_update_va_mapping(addr, __pte, 0)) \
11379 - (ptep)->pte_low = __pte.pte_low; \
11380 - __ret; \
11381 + int __young = pte_young(__pte); \
11382 + __pte = pte_mkold(__pte); \
11383 + if (test_bit(PG_pinned, &virt_to_page((vma)->vm_mm->pgd)->flags)) \
11384 + ptep_set_access_flags(vma, address, ptep, __pte, __young); \
11385 + else if (__young) \
11386 + (ptep)->pte_low = __pte.pte_low; \
11387 + __young; \
11388 })
11389
11390 +#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
11391 #define ptep_get_and_clear_full(mm, addr, ptep, full) \
11392 ((full) ? ({ \
11393 pte_t __res = *(ptep); \
11394 @@ -296,6 +354,7 @@ static inline pte_t pte_mkhuge(pte_t pte
11395 }) : \
11396 ptep_get_and_clear(mm, addr, ptep))
11397
11398 +#define __HAVE_ARCH_PTEP_SET_WRPROTECT
11399 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
11400 {
11401 pte_t pte = *ptep;
11402 @@ -391,11 +450,11 @@ static inline pte_t pte_modify(pte_t pte
11403 #define pte_index(address) \
11404 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
11405 #define pte_offset_kernel(dir, address) \
11406 - ((pte_t *) pmd_page_kernel(*(dir)) + pte_index(address))
11407 + ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(address))
11408
11409 #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
11410
11411 -#define pmd_page_kernel(pmd) \
11412 +#define pmd_page_vaddr(pmd) \
11413 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
11414
11415 /*
11416 @@ -418,8 +477,6 @@ extern pte_t *lookup_address(unsigned lo
11417 static inline int set_kernel_exec(unsigned long vaddr, int enable) { return 0;}
11418 #endif
11419
11420 -extern void noexec_setup(const char *str);
11421 -
11422 #if defined(CONFIG_HIGHPTE)
11423 #define pte_offset_map(dir, address) \
11424 ((pte_t *)kmap_atomic_pte(pmd_page(*(dir)),KM_PTE0) + \
11425 @@ -437,37 +494,17 @@ extern void noexec_setup(const char *str
11426 #define pte_unmap_nested(pte) do { } while (0)
11427 #endif
11428
11429 -#define __HAVE_ARCH_PTEP_ESTABLISH
11430 -#define ptep_establish(vma, address, ptep, pteval) \
11431 - do { \
11432 - if ( likely((vma)->vm_mm == current->mm) ) { \
11433 - BUG_ON(HYPERVISOR_update_va_mapping(address, \
11434 - pteval, \
11435 - (unsigned long)(vma)->vm_mm->cpu_vm_mask.bits| \
11436 - UVMF_INVLPG|UVMF_MULTI)); \
11437 - } else { \
11438 - xen_l1_entry_update(ptep, pteval); \
11439 - flush_tlb_page(vma, address); \
11440 - } \
11441 - } while (0)
11442 +/* Clear a kernel PTE and flush it from the TLB */
11443 +#define kpte_clear_flush(ptep, vaddr) do { \
11444 + if (HYPERVISOR_update_va_mapping(vaddr, __pte(0), UVMF_INVLPG)) \
11445 + BUG(); \
11446 +} while (0)
11447
11448 /*
11449 * The i386 doesn't have any external MMU info: the kernel page
11450 * tables contain all the necessary information.
11451 - *
11452 - * Also, we only update the dirty/accessed state if we set
11453 - * the dirty bit by hand in the kernel, since the hardware
11454 - * will do the accessed bit for us, and we don't want to
11455 - * race with other CPU's that might be updating the dirty
11456 - * bit at the same time.
11457 */
11458 #define update_mmu_cache(vma,address,pte) do { } while (0)
11459 -#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
11460 -#define ptep_set_access_flags(vma, address, ptep, entry, dirty) \
11461 - do { \
11462 - if (dirty) \
11463 - ptep_establish(vma, address, ptep, entry); \
11464 - } while (0)
11465
11466 #include <xen/features.h>
11467 void make_lowmem_page_readonly(void *va, unsigned int feature);
11468 @@ -526,10 +563,11 @@ int touch_pte_range(struct mm_struct *mm
11469 unsigned long size);
11470
11471 int xen_change_pte_range(struct mm_struct *mm, pmd_t *pmd,
11472 - unsigned long addr, unsigned long end, pgprot_t newprot);
11473 + unsigned long addr, unsigned long end, pgprot_t newprot,
11474 + int dirty_accountable);
11475
11476 -#define arch_change_pte_range(mm, pmd, addr, end, newprot) \
11477 - xen_change_pte_range(mm, pmd, addr, end, newprot)
11478 +#define arch_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) \
11479 + xen_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable)
11480
11481 #define io_remap_pfn_range(vma,from,pfn,size,prot) \
11482 direct_remap_pfn_range(vma,from,pfn,size,prot,DOMID_IO)
11483 @@ -538,13 +576,6 @@ direct_remap_pfn_range(vma,from,pfn,size
11484 #define GET_IOSPACE(pfn) 0
11485 #define GET_PFN(pfn) (pfn)
11486
11487 -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
11488 -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
11489 -#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
11490 -#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
11491 -#define __HAVE_ARCH_PTEP_CLEAR_FLUSH
11492 -#define __HAVE_ARCH_PTEP_SET_WRPROTECT
11493 -#define __HAVE_ARCH_PTE_SAME
11494 #include <asm-generic/pgtable.h>
11495
11496 #endif /* _I386_PGTABLE_H */
11497 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/processor_32.h 2009-04-20 11:36:10.000000000 +0200
11498 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/processor_32.h 2009-03-04 11:28:34.000000000 +0100
11499 @@ -146,6 +146,18 @@ static inline void detect_ht(struct cpui
11500 #define X86_EFLAGS_VIP 0x00100000 /* Virtual Interrupt Pending */
11501 #define X86_EFLAGS_ID 0x00200000 /* CPUID detection flag */
11502
11503 +static inline void __cpuid(unsigned int *eax, unsigned int *ebx,
11504 + unsigned int *ecx, unsigned int *edx)
11505 +{
11506 + /* ecx is often an input as well as an output. */
11507 + __asm__(XEN_CPUID
11508 + : "=a" (*eax),
11509 + "=b" (*ebx),
11510 + "=c" (*ecx),
11511 + "=d" (*edx)
11512 + : "0" (*eax), "2" (*ecx));
11513 +}
11514 +
11515 /*
11516 * Generic CPUID function
11517 * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
11518 @@ -153,24 +165,18 @@ static inline void detect_ht(struct cpui
11519 */
11520 static inline void cpuid(unsigned int op, unsigned int *eax, unsigned int *ebx, unsigned int *ecx, unsigned int *edx)
11521 {
11522 - __asm__(XEN_CPUID
11523 - : "=a" (*eax),
11524 - "=b" (*ebx),
11525 - "=c" (*ecx),
11526 - "=d" (*edx)
11527 - : "0" (op), "c"(0));
11528 + *eax = op;
11529 + *ecx = 0;
11530 + __cpuid(eax, ebx, ecx, edx);
11531 }
11532
11533 /* Some CPUID calls want 'count' to be placed in ecx */
11534 static inline void cpuid_count(int op, int count, int *eax, int *ebx, int *ecx,
11535 - int *edx)
11536 + int *edx)
11537 {
11538 - __asm__(XEN_CPUID
11539 - : "=a" (*eax),
11540 - "=b" (*ebx),
11541 - "=c" (*ecx),
11542 - "=d" (*edx)
11543 - : "0" (op), "c" (count));
11544 + *eax = op;
11545 + *ecx = count;
11546 + __cpuid(eax, ebx, ecx, edx);
11547 }
11548
11549 /*
11550 @@ -178,42 +184,30 @@ static inline void cpuid_count(int op, i
11551 */
11552 static inline unsigned int cpuid_eax(unsigned int op)
11553 {
11554 - unsigned int eax;
11555 + unsigned int eax, ebx, ecx, edx;
11556
11557 - __asm__(XEN_CPUID
11558 - : "=a" (eax)
11559 - : "0" (op)
11560 - : "bx", "cx", "dx");
11561 + cpuid(op, &eax, &ebx, &ecx, &edx);
11562 return eax;
11563 }
11564 static inline unsigned int cpuid_ebx(unsigned int op)
11565 {
11566 - unsigned int eax, ebx;
11567 + unsigned int eax, ebx, ecx, edx;
11568
11569 - __asm__(XEN_CPUID
11570 - : "=a" (eax), "=b" (ebx)
11571 - : "0" (op)
11572 - : "cx", "dx" );
11573 + cpuid(op, &eax, &ebx, &ecx, &edx);
11574 return ebx;
11575 }
11576 static inline unsigned int cpuid_ecx(unsigned int op)
11577 {
11578 - unsigned int eax, ecx;
11579 + unsigned int eax, ebx, ecx, edx;
11580
11581 - __asm__(XEN_CPUID
11582 - : "=a" (eax), "=c" (ecx)
11583 - : "0" (op)
11584 - : "bx", "dx" );
11585 + cpuid(op, &eax, &ebx, &ecx, &edx);
11586 return ecx;
11587 }
11588 static inline unsigned int cpuid_edx(unsigned int op)
11589 {
11590 - unsigned int eax, edx;
11591 + unsigned int eax, ebx, ecx, edx;
11592
11593 - __asm__(XEN_CPUID
11594 - : "=a" (eax), "=d" (edx)
11595 - : "0" (op)
11596 - : "bx", "cx");
11597 + cpuid(op, &eax, &ebx, &ecx, &edx);
11598 return edx;
11599 }
11600
11601 @@ -315,6 +309,8 @@ static inline void __mwait(unsigned long
11602 : :"a" (eax), "c" (ecx));
11603 }
11604
11605 +extern void mwait_idle_with_hints(unsigned long eax, unsigned long ecx);
11606 +
11607 /* from system description table in BIOS. Mostly for MCA use, but
11608 others may find it useful. */
11609 extern unsigned int machine_id;
11610 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/segment_32.h 2009-05-14 11:02:43.000000000 +0200
11611 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/segment_32.h 2009-03-04 11:28:34.000000000 +0100
11612 @@ -61,11 +61,9 @@
11613
11614 #define GDT_ENTRY_KERNEL_CS (GDT_ENTRY_KERNEL_BASE + 0)
11615 #define __KERNEL_CS (GDT_ENTRY_KERNEL_CS * 8)
11616 -#define GET_KERNEL_CS() (__KERNEL_CS | (xen_feature(XENFEAT_supervisor_mode_kernel)?0:1) )
11617
11618 #define GDT_ENTRY_KERNEL_DS (GDT_ENTRY_KERNEL_BASE + 1)
11619 #define __KERNEL_DS (GDT_ENTRY_KERNEL_DS * 8)
11620 -#define GET_KERNEL_DS() (__KERNEL_DS | (xen_feature(XENFEAT_supervisor_mode_kernel)?0:1) )
11621
11622 #define GDT_ENTRY_TSS (GDT_ENTRY_KERNEL_BASE + 4)
11623 #define GDT_ENTRY_LDT (GDT_ENTRY_KERNEL_BASE + 5)
11624 @@ -85,6 +83,11 @@
11625
11626 #define GDT_SIZE (GDT_ENTRIES * 8)
11627
11628 +/* Matches __KERNEL_CS and __USER_CS (they must be 2 entries apart) */
11629 +#define SEGMENT_IS_FLAT_CODE(x) (((x) & 0xec) == GDT_ENTRY_KERNEL_CS * 8)
11630 +/* Matches PNP_CS32 and PNP_CS16 (they must be consecutive) */
11631 +#define SEGMENT_IS_PNP_CODE(x) (((x) & 0xf4) == GDT_ENTRY_PNPBIOS_BASE * 8)
11632 +
11633 /* Simple and small GDT entries for booting only */
11634
11635 #define GDT_ENTRY_BOOT_CS 2
11636 @@ -114,4 +117,16 @@
11637 */
11638 #define IDT_ENTRIES 256
11639
11640 +/* Bottom two bits of selector give the ring privilege level */
11641 +#define SEGMENT_RPL_MASK 0x3
11642 +/* Bit 2 is table indicator (LDT/GDT) */
11643 +#define SEGMENT_TI_MASK 0x4
11644 +
11645 +/* User mode is privilege level 3 */
11646 +#define USER_RPL 0x3
11647 +/* LDT segment has TI set, GDT has it cleared */
11648 +#define SEGMENT_LDT 0x4
11649 +#define SEGMENT_GDT 0x0
11650 +
11651 +#define get_kernel_rpl() (xen_feature(XENFEAT_supervisor_mode_kernel)?0:1)
11652 #endif
11653 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/smp_32.h 2009-05-14 11:02:43.000000000 +0200
11654 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/smp_32.h 2009-03-04 11:28:34.000000000 +0100
11655 @@ -79,25 +79,36 @@ static inline int hard_smp_processor_id(
11656 return GET_APIC_ID(*(unsigned long *)(APIC_BASE+APIC_ID));
11657 }
11658 #endif
11659 -
11660 -static __inline int logical_smp_processor_id(void)
11661 -{
11662 - /* we don't want to mark this access volatile - bad code generation */
11663 - return GET_APIC_LOGICAL_ID(*(unsigned long *)(APIC_BASE+APIC_LDR));
11664 -}
11665 -
11666 #endif
11667
11668 +#define safe_smp_processor_id() smp_processor_id()
11669 extern int __cpu_disable(void);
11670 extern void __cpu_die(unsigned int cpu);
11671 extern void prefill_possible_map(void);
11672 +extern unsigned int num_processors;
11673 +
11674 #endif /* !__ASSEMBLY__ */
11675
11676 #else /* CONFIG_SMP */
11677
11678 +#define safe_smp_processor_id() 0
11679 #define cpu_physical_id(cpu) boot_cpu_physical_apicid
11680
11681 #define NO_PROC_ID 0xFF /* No processor magic marker */
11682
11683 #endif
11684 +
11685 +#ifndef __ASSEMBLY__
11686 +
11687 +extern u8 apicid_2_node[];
11688 +
11689 +#ifdef CONFIG_X86_LOCAL_APIC
11690 +static __inline int logical_smp_processor_id(void)
11691 +{
11692 + /* we don't want to mark this access volatile - bad code generation */
11693 + return GET_APIC_LOGICAL_ID(*(unsigned long *)(APIC_BASE+APIC_LDR));
11694 +}
11695 +#endif
11696 +#endif
11697 +
11698 #endif
11699 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/system_32.h 2009-05-14 11:02:43.000000000 +0200
11700 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/system_32.h 2009-03-04 11:28:34.000000000 +0100
11701 @@ -267,6 +267,9 @@ static inline unsigned long __xchg(unsig
11702 #define cmpxchg(ptr,o,n)\
11703 ((__typeof__(*(ptr)))__cmpxchg((ptr),(unsigned long)(o),\
11704 (unsigned long)(n),sizeof(*(ptr))))
11705 +#define sync_cmpxchg(ptr,o,n)\
11706 + ((__typeof__(*(ptr)))__sync_cmpxchg((ptr),(unsigned long)(o),\
11707 + (unsigned long)(n),sizeof(*(ptr))))
11708 #endif
11709
11710 static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old,
11711 @@ -296,6 +299,39 @@ static inline unsigned long __cmpxchg(vo
11712 return old;
11713 }
11714
11715 +/*
11716 + * Always use locked operations when touching memory shared with a
11717 + * hypervisor, since the system may be SMP even if the guest kernel
11718 + * isn't.
11719 + */
11720 +static inline unsigned long __sync_cmpxchg(volatile void *ptr,
11721 + unsigned long old,
11722 + unsigned long new, int size)
11723 +{
11724 + unsigned long prev;
11725 + switch (size) {
11726 + case 1:
11727 + __asm__ __volatile__("lock; cmpxchgb %b1,%2"
11728 + : "=a"(prev)
11729 + : "q"(new), "m"(*__xg(ptr)), "0"(old)
11730 + : "memory");
11731 + return prev;
11732 + case 2:
11733 + __asm__ __volatile__("lock; cmpxchgw %w1,%2"
11734 + : "=a"(prev)
11735 + : "r"(new), "m"(*__xg(ptr)), "0"(old)
11736 + : "memory");
11737 + return prev;
11738 + case 4:
11739 + __asm__ __volatile__("lock; cmpxchgl %1,%2"
11740 + : "=a"(prev)
11741 + : "r"(new), "m"(*__xg(ptr)), "0"(old)
11742 + : "memory");
11743 + return prev;
11744 + }
11745 + return old;
11746 +}
11747 +
11748 #ifndef CONFIG_X86_CMPXCHG
11749 /*
11750 * Building a kernel capable running on 80386. It may be necessary to
11751 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/tlbflush_32.h 2009-05-14 11:02:43.000000000 +0200
11752 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/tlbflush_32.h 2009-03-04 11:28:34.000000000 +0100
11753 @@ -8,8 +8,6 @@
11754 #define __flush_tlb_global() xen_tlb_flush()
11755 #define __flush_tlb_all() xen_tlb_flush()
11756
11757 -extern unsigned long pgkern_mask;
11758 -
11759 #define cpu_has_invlpg (boot_cpu_data.x86 > 3)
11760
11761 #define __flush_tlb_single(addr) xen_invlpg(addr)
11762 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/dma-mapping_64.h 2009-05-14 11:02:43.000000000 +0200
11763 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/dma-mapping_64.h 2009-03-04 11:28:34.000000000 +0100
11764 @@ -55,13 +55,6 @@ extern dma_addr_t bad_dma_address;
11765 extern struct dma_mapping_ops* dma_ops;
11766 extern int iommu_merge;
11767
11768 -static inline int valid_dma_direction(int dma_direction)
11769 -{
11770 - return ((dma_direction == DMA_BIDIRECTIONAL) ||
11771 - (dma_direction == DMA_TO_DEVICE) ||
11772 - (dma_direction == DMA_FROM_DEVICE));
11773 -}
11774 -
11775 #if 0
11776 static inline int dma_mapping_error(dma_addr_t dma_addr)
11777 {
11778 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/fixmap_64.h 2009-05-14 11:02:43.000000000 +0200
11779 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/fixmap_64.h 2009-03-04 11:28:34.000000000 +0100
11780 @@ -41,7 +41,7 @@ enum fixed_addresses {
11781 #ifdef CONFIG_X86_LOCAL_APIC
11782 FIX_APIC_BASE, /* local (CPU) APIC) -- required for SMP or not */
11783 #endif
11784 -#ifdef CONFIG_X86_IO_APIC
11785 +#ifndef CONFIG_XEN
11786 FIX_IO_APIC_BASE_0,
11787 FIX_IO_APIC_BASE_END = FIX_IO_APIC_BASE_0 + MAX_IO_APICS-1,
11788 #endif
11789 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/pgtable_64.h 2009-04-20 11:36:10.000000000 +0200
11790 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/pgtable_64.h 2009-03-04 11:28:34.000000000 +0100
11791 @@ -44,12 +44,9 @@ extern unsigned long __supported_pte_mas
11792
11793 #define swapper_pg_dir init_level4_pgt
11794
11795 -extern int nonx_setup(char *str);
11796 extern void paging_init(void);
11797 extern void clear_kernel_mapping(unsigned long addr, unsigned long size);
11798
11799 -extern unsigned long pgkern_mask;
11800 -
11801 /*
11802 * ZERO_PAGE is a global shared page that is always zero: used
11803 * for zero-mapped memory areas etc..
11804 @@ -119,9 +116,6 @@ static inline void pgd_clear (pgd_t * pg
11805 set_pgd(__user_pgd(pgd), __pgd(0));
11806 }
11807
11808 -#define pud_page(pud) \
11809 - ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK))
11810 -
11811 #define pte_same(a, b) ((a).pte == (b).pte)
11812
11813 #define pte_pgprot(a) (__pgprot((a).pte & ~PHYSICAL_PAGE_MASK))
11814 @@ -333,7 +327,7 @@ static inline pte_t ptep_get_and_clear_f
11815 #define __LARGE_PTE (_PAGE_PSE|_PAGE_PRESENT)
11816 static inline int pte_user(pte_t pte) { return __pte_val(pte) & _PAGE_USER; }
11817 static inline int pte_read(pte_t pte) { return __pte_val(pte) & _PAGE_USER; }
11818 -static inline int pte_exec(pte_t pte) { return __pte_val(pte) & _PAGE_USER; }
11819 +static inline int pte_exec(pte_t pte) { return !(__pte_val(pte) & _PAGE_NX); }
11820 static inline int pte_dirty(pte_t pte) { return __pte_val(pte) & _PAGE_DIRTY; }
11821 static inline int pte_young(pte_t pte) { return __pte_val(pte) & _PAGE_ACCESSED; }
11822 static inline int pte_write(pte_t pte) { return __pte_val(pte) & _PAGE_RW; }
11823 @@ -346,29 +340,12 @@ static inline pte_t pte_mkclean(pte_t pt
11824 static inline pte_t pte_mkold(pte_t pte) { __pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
11825 static inline pte_t pte_wrprotect(pte_t pte) { __pte_val(pte) &= ~_PAGE_RW; return pte; }
11826 static inline pte_t pte_mkread(pte_t pte) { __pte_val(pte) |= _PAGE_USER; return pte; }
11827 -static inline pte_t pte_mkexec(pte_t pte) { __pte_val(pte) |= _PAGE_USER; return pte; }
11828 +static inline pte_t pte_mkexec(pte_t pte) { __pte_val(pte) &= ~_PAGE_NX; return pte; }
11829 static inline pte_t pte_mkdirty(pte_t pte) { __pte_val(pte) |= _PAGE_DIRTY; return pte; }
11830 static inline pte_t pte_mkyoung(pte_t pte) { __pte_val(pte) |= _PAGE_ACCESSED; return pte; }
11831 static inline pte_t pte_mkwrite(pte_t pte) { __pte_val(pte) |= _PAGE_RW; return pte; }
11832 static inline pte_t pte_mkhuge(pte_t pte) { __pte_val(pte) |= _PAGE_PSE; return pte; }
11833 -
11834 -#define ptep_test_and_clear_dirty(vma, addr, ptep) \
11835 -({ \
11836 - pte_t __pte = *(ptep); \
11837 - int __ret = pte_dirty(__pte); \
11838 - if (__ret) \
11839 - set_pte_at((vma)->vm_mm, addr, ptep, pte_mkclean(__pte)); \
11840 - __ret; \
11841 -})
11842 -
11843 -#define ptep_test_and_clear_young(vma, addr, ptep) \
11844 -({ \
11845 - pte_t __pte = *(ptep); \
11846 - int __ret = pte_young(__pte); \
11847 - if (__ret) \
11848 - set_pte_at((vma)->vm_mm, addr, ptep, pte_mkold(__pte)); \
11849 - __ret; \
11850 -})
11851 +static inline pte_t pte_clrhuge(pte_t pte) { __pte_val(pte) &= ~_PAGE_PSE; return pte; }
11852
11853 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
11854 {
11855 @@ -395,7 +372,8 @@ static inline int pmd_large(pmd_t pte) {
11856 /*
11857 * Level 4 access.
11858 */
11859 -#define pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & PTE_MASK))
11860 +#define pgd_page_vaddr(pgd) ((unsigned long) __va(pgd_val(pgd) & PTE_MASK))
11861 +#define pgd_page(pgd) (pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT))
11862 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
11863 #define pgd_offset(mm, addr) ((mm)->pgd + pgd_index(addr))
11864 #define pgd_offset_k(address) (init_level4_pgt + pgd_index(address))
11865 @@ -404,16 +382,18 @@ static inline int pmd_large(pmd_t pte) {
11866
11867 /* PUD - Level3 access */
11868 /* to find an entry in a page-table-directory. */
11869 +#define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & PHYSICAL_PAGE_MASK))
11870 +#define pud_page(pud) (pfn_to_page(pud_val(pud) >> PAGE_SHIFT))
11871 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
11872 -#define pud_offset(pgd, address) ((pud_t *) pgd_page(*(pgd)) + pud_index(address))
11873 +#define pud_offset(pgd, address) ((pud_t *) pgd_page_vaddr(*(pgd)) + pud_index(address))
11874 #define pud_present(pud) (__pud_val(pud) & _PAGE_PRESENT)
11875
11876 /* PMD - Level 2 access */
11877 -#define pmd_page_kernel(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK))
11878 +#define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & PTE_MASK))
11879 #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
11880
11881 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
11882 -#define pmd_offset(dir, address) ((pmd_t *) pud_page(*(dir)) + \
11883 +#define pmd_offset(dir, address) ((pmd_t *) pud_page_vaddr(*(dir)) + \
11884 pmd_index(address))
11885 #define pmd_none(x) (!__pmd_val(x))
11886 #if CONFIG_XEN_COMPAT <= 0x030002
11887 @@ -444,6 +424,7 @@ static inline pte_t mk_pte_phys(unsigned
11888 {
11889 unsigned long pteval;
11890 pteval = physpage | pgprot_val(pgprot);
11891 + pteval &= __supported_pte_mask;
11892 return __pte(pteval);
11893 }
11894
11895 @@ -465,7 +446,7 @@ static inline pte_t pte_modify(pte_t pte
11896
11897 #define pte_index(address) \
11898 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
11899 -#define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_kernel(*(dir)) + \
11900 +#define pte_offset_kernel(dir, address) ((pte_t *) pmd_page_vaddr(*(dir)) + \
11901 pte_index(address))
11902
11903 /* x86-64 always has all page tables mapped. */
11904 @@ -506,6 +487,40 @@ static inline pte_t pte_modify(pte_t pte
11905 ptep_establish(vma, address, ptep, entry); \
11906 } while (0)
11907
11908 +
11909 +/*
11910 + * i386 says: We don't actually have these, but we want to advertise
11911 + * them so that we can encompass the flush here.
11912 + */
11913 +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
11914 +#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
11915 +
11916 +#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
11917 +#define ptep_clear_flush_dirty(vma, address, ptep) \
11918 +({ \
11919 + pte_t __pte = *(ptep); \
11920 + int __dirty = pte_dirty(__pte); \
11921 + __pte = pte_mkclean(__pte); \
11922 + if ((vma)->vm_mm->context.pinned) \
11923 + ptep_set_access_flags(vma, address, ptep, __pte, __dirty); \
11924 + else if (__dirty) \
11925 + set_pte(ptep, __pte); \
11926 + __dirty; \
11927 +})
11928 +
11929 +#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
11930 +#define ptep_clear_flush_young(vma, address, ptep) \
11931 +({ \
11932 + pte_t __pte = *(ptep); \
11933 + int __young = pte_young(__pte); \
11934 + __pte = pte_mkold(__pte); \
11935 + if ((vma)->vm_mm->context.pinned) \
11936 + ptep_set_access_flags(vma, address, ptep, __pte, __young); \
11937 + else if (__young) \
11938 + set_pte(ptep, __pte); \
11939 + __young; \
11940 +})
11941 +
11942 /* Encode and de-code a swap entry */
11943 #define __swp_type(x) (((x).val >> 1) & 0x3f)
11944 #define __swp_offset(x) ((x).val >> 8)
11945 @@ -547,10 +562,11 @@ int touch_pte_range(struct mm_struct *mm
11946 unsigned long size);
11947
11948 int xen_change_pte_range(struct mm_struct *mm, pmd_t *pmd,
11949 - unsigned long addr, unsigned long end, pgprot_t newprot);
11950 + unsigned long addr, unsigned long end, pgprot_t newprot,
11951 + int dirty_accountable);
11952
11953 -#define arch_change_pte_range(mm, pmd, addr, end, newprot) \
11954 - xen_change_pte_range(mm, pmd, addr, end, newprot)
11955 +#define arch_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable) \
11956 + xen_change_pte_range(mm, pmd, addr, end, newprot, dirty_accountable)
11957
11958 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
11959 direct_remap_pfn_range(vma,vaddr,pfn,size,prot,DOMID_IO)
11960 @@ -572,8 +588,6 @@ int xen_change_pte_range(struct mm_struc
11961 #define kc_offset_to_vaddr(o) \
11962 (((o) & (1UL << (__VIRTUAL_MASK_SHIFT-1))) ? ((o) | (~__VIRTUAL_MASK)) : (o))
11963
11964 -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
11965 -#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
11966 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
11967 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
11968 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
11969 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/processor_64.h 2009-05-14 11:02:43.000000000 +0200
11970 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/processor_64.h 2009-03-04 11:28:34.000000000 +0100
11971 @@ -484,6 +484,8 @@ static inline void __mwait(unsigned long
11972 : :"a" (eax), "c" (ecx));
11973 }
11974
11975 +extern void mwait_idle_with_hints(unsigned long eax, unsigned long ecx);
11976 +
11977 #define stack_current() \
11978 ({ \
11979 struct thread_info *ti; \
11980 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/smp_64.h 2009-05-14 11:02:43.000000000 +0200
11981 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/smp_64.h 2009-03-04 11:28:34.000000000 +0100
11982 @@ -4,15 +4,12 @@
11983 /*
11984 * We need the APIC definitions automatically as part of 'smp.h'
11985 */
11986 -#ifndef __ASSEMBLY__
11987 #include <linux/threads.h>
11988 #include <linux/cpumask.h>
11989 #include <linux/bitops.h>
11990 extern int disable_apic;
11991 -#endif
11992
11993 #ifdef CONFIG_X86_LOCAL_APIC
11994 -#ifndef __ASSEMBLY__
11995 #include <asm/fixmap.h>
11996 #include <asm/mpspec.h>
11997 #ifdef CONFIG_X86_IO_APIC
11998 @@ -21,10 +18,8 @@ extern int disable_apic;
11999 #include <asm/apic.h>
12000 #include <asm/thread_info.h>
12001 #endif
12002 -#endif
12003
12004 #ifdef CONFIG_SMP
12005 -#ifndef ASSEMBLY
12006
12007 #include <asm/pda.h>
12008
12009 @@ -41,14 +36,11 @@ extern cpumask_t cpu_initialized;
12010
12011 extern void smp_alloc_memory(void);
12012 extern volatile unsigned long smp_invalidate_needed;
12013 -extern int pic_mode;
12014 extern void lock_ipi_call_lock(void);
12015 extern void unlock_ipi_call_lock(void);
12016 extern int smp_num_siblings;
12017 extern void smp_send_reschedule(int cpu);
12018 void smp_stop_cpu(void);
12019 -extern int smp_call_function_single(int cpuid, void (*func) (void *info),
12020 - void *info, int retry, int wait);
12021
12022 extern cpumask_t cpu_sibling_map[NR_CPUS];
12023 extern cpumask_t cpu_core_map[NR_CPUS];
12024 @@ -77,20 +69,16 @@ static inline int hard_smp_processor_id(
12025 }
12026 #endif
12027
12028 -extern int safe_smp_processor_id(void);
12029 extern int __cpu_disable(void);
12030 extern void __cpu_die(unsigned int cpu);
12031 extern void prefill_possible_map(void);
12032 extern unsigned num_processors;
12033 extern unsigned disabled_cpus;
12034
12035 -#endif /* !ASSEMBLY */
12036 -
12037 #define NO_PROC_ID 0xFF /* No processor magic marker */
12038
12039 #endif
12040
12041 -#ifndef ASSEMBLY
12042 /*
12043 * Some lowlevel functions might want to know about
12044 * the real APIC ID <-> CPU # mapping.
12045 @@ -114,11 +102,8 @@ static inline int cpu_present_to_apicid(
12046 }
12047 #endif
12048
12049 -#endif /* !ASSEMBLY */
12050 -
12051 #ifndef CONFIG_SMP
12052 #define stack_smp_processor_id() 0
12053 -#define safe_smp_processor_id() 0
12054 #define cpu_logical_map(x) (x)
12055 #else
12056 #include <asm/thread_info.h>
12057 @@ -130,7 +115,6 @@ static inline int cpu_present_to_apicid(
12058 })
12059 #endif
12060
12061 -#ifndef __ASSEMBLY__
12062 #ifdef CONFIG_X86_LOCAL_APIC
12063 static __inline int logical_smp_processor_id(void)
12064 {
12065 @@ -138,13 +122,18 @@ static __inline int logical_smp_processo
12066 return GET_APIC_LOGICAL_ID(*(unsigned long *)(APIC_BASE+APIC_LDR));
12067 }
12068 #endif
12069 -#endif
12070
12071 #ifdef CONFIG_SMP
12072 #define cpu_physical_id(cpu) x86_cpu_to_apicid[cpu]
12073 #else
12074 #define cpu_physical_id(cpu) boot_cpu_id
12075 -#endif
12076 -
12077 +static inline int smp_call_function_single(int cpuid, void (*func) (void *info),
12078 + void *info, int retry, int wait)
12079 +{
12080 + /* Disable interrupts here? */
12081 + func(info);
12082 + return 0;
12083 +}
12084 +#endif /* !CONFIG_SMP */
12085 #endif
12086
12087 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/system_64.h 2009-05-14 11:02:43.000000000 +0200
12088 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/system_64.h 2009-03-04 11:28:34.000000000 +0100
12089 @@ -24,6 +24,7 @@
12090 #define __EXTRA_CLOBBER \
12091 ,"rcx","rbx","rdx","r8","r9","r10","r11","r12","r13","r14","r15"
12092
12093 +/* Save restore flags to clear handle leaking NT */
12094 #define switch_to(prev,next,last) \
12095 asm volatile(SAVE_CONTEXT \
12096 "movq %%rsp,%P[threadrsp](%[prev])\n\t" /* save RSP */ \
12097 --- sle11-2009-05-14.orig/include/asm-x86/mach-xen/asm/tlbflush_64.h 2009-05-14 11:02:43.000000000 +0200
12098 +++ sle11-2009-05-14/include/asm-x86/mach-xen/asm/tlbflush_64.h 2009-03-04 11:28:34.000000000 +0100
12099 @@ -12,9 +12,6 @@
12100 */
12101 #define __flush_tlb_global() xen_tlb_flush()
12102
12103 -
12104 -extern unsigned long pgkern_mask;
12105 -
12106 #define __flush_tlb_all() __flush_tlb_global()
12107
12108 #define __flush_tlb_one(addr) xen_invlpg((unsigned long)addr)
12109 --- sle11-2009-05-14.orig/include/linux/skbuff.h 2009-02-16 15:58:14.000000000 +0100
12110 +++ sle11-2009-05-14/include/linux/skbuff.h 2009-03-04 11:28:34.000000000 +0100
12111 @@ -1771,5 +1771,12 @@ static inline void skb_forward_csum(stru
12112 }
12113
12114 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
12115 +
12116 +#ifdef CONFIG_XEN
12117 +int skb_checksum_setup(struct sk_buff *skb);
12118 +#else
12119 +static inline int skb_checksum_setup(struct sk_buff *skb) { return 0; }
12120 +#endif
12121 +
12122 #endif /* __KERNEL__ */
12123 #endif /* _LINUX_SKBUFF_H */
12124 --- sle11-2009-05-14.orig/include/xen/evtchn.h 2009-05-14 11:02:43.000000000 +0200
12125 +++ sle11-2009-05-14/include/xen/evtchn.h 2009-03-04 11:28:34.000000000 +0100
12126 @@ -54,34 +54,34 @@
12127 */
12128 int bind_caller_port_to_irqhandler(
12129 unsigned int caller_port,
12130 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
12131 + irq_handler_t handler,
12132 unsigned long irqflags,
12133 const char *devname,
12134 void *dev_id);
12135 int bind_listening_port_to_irqhandler(
12136 unsigned int remote_domain,
12137 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
12138 + irq_handler_t handler,
12139 unsigned long irqflags,
12140 const char *devname,
12141 void *dev_id);
12142 int bind_interdomain_evtchn_to_irqhandler(
12143 unsigned int remote_domain,
12144 unsigned int remote_port,
12145 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
12146 + irq_handler_t handler,
12147 unsigned long irqflags,
12148 const char *devname,
12149 void *dev_id);
12150 int bind_virq_to_irqhandler(
12151 unsigned int virq,
12152 unsigned int cpu,
12153 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
12154 + irq_handler_t handler,
12155 unsigned long irqflags,
12156 const char *devname,
12157 void *dev_id);
12158 int bind_ipi_to_irqhandler(
12159 unsigned int ipi,
12160 unsigned int cpu,
12161 - irqreturn_t (*handler)(int, void *, struct pt_regs *),
12162 + irq_handler_t handler,
12163 unsigned long irqflags,
12164 const char *devname,
12165 void *dev_id);
12166 --- sle11-2009-05-14.orig/include/xen/xencons.h 2009-05-14 11:02:43.000000000 +0200
12167 +++ sle11-2009-05-14/include/xen/xencons.h 2009-03-04 11:28:34.000000000 +0100
12168 @@ -8,7 +8,7 @@ void xencons_force_flush(void);
12169 void xencons_resume(void);
12170
12171 /* Interrupt work hooks. Receive data, or kick data out. */
12172 -void xencons_rx(char *buf, unsigned len, struct pt_regs *regs);
12173 +void xencons_rx(char *buf, unsigned len);
12174 void xencons_tx(void);
12175
12176 int xencons_ring_init(void);
12177 --- sle11-2009-05-14.orig/mm/mprotect.c 2009-02-16 15:58:14.000000000 +0100
12178 +++ sle11-2009-05-14/mm/mprotect.c 2009-03-04 11:28:34.000000000 +0100
12179 @@ -92,7 +92,7 @@ static inline void change_pmd_range(stru
12180 next = pmd_addr_end(addr, end);
12181 if (pmd_none_or_clear_bad(pmd))
12182 continue;
12183 - if (arch_change_pte_range(mm, pmd, addr, next, newprot))
12184 + if (arch_change_pte_range(mm, pmd, addr, next, newprot, dirty_accountable))
12185 continue;
12186 change_pte_range(mm, pmd, addr, next, newprot, dirty_accountable);
12187 } while (pmd++, addr = next, addr != end);
12188 --- sle11-2009-05-14.orig/net/core/dev.c 2009-02-16 15:58:14.000000000 +0100
12189 +++ sle11-2009-05-14/net/core/dev.c 2009-03-04 11:28:34.000000000 +0100
12190 @@ -1765,15 +1765,14 @@ inline int skb_checksum_setup(struct sk_
12191 }
12192 if ((skb->h.raw + skb->csum + 2) > skb->tail)
12193 goto out;
12194 - skb->ip_summed = CHECKSUM_HW;
12195 + skb->ip_summed = CHECKSUM_PARTIAL;
12196 skb->proto_csum_blank = 0;
12197 }
12198 return 0;
12199 out:
12200 return -EPROTO;
12201 }
12202 -#else
12203 -inline int skb_checksum_setup(struct sk_buff *skb) { return 0; }
12204 +EXPORT_SYMBOL(skb_checksum_setup);
12205 #endif
12206
12207 /**
12208 @@ -2327,7 +2326,7 @@ int netif_receive_skb(struct sk_buff *sk
12209 case CHECKSUM_UNNECESSARY:
12210 skb->proto_data_valid = 1;
12211 break;
12212 - case CHECKSUM_HW:
12213 + case CHECKSUM_PARTIAL:
12214 /* XXX Implement me. */
12215 default:
12216 skb->proto_data_valid = 0;
12217 @@ -4989,7 +4988,6 @@ EXPORT_SYMBOL(unregister_netdevice_notif
12218 EXPORT_SYMBOL(net_enable_timestamp);
12219 EXPORT_SYMBOL(net_disable_timestamp);
12220 EXPORT_SYMBOL(dev_get_flags);
12221 -EXPORT_SYMBOL(skb_checksum_setup);
12222
12223 #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
12224 EXPORT_SYMBOL(br_handle_frame_hook);
IPFire 2.x development tree