1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright IBM Corp. 1999, 2012
5 * Author(s): Hartmut Penner (hp@de.ibm.com),
6 * Martin Schwidefsky (schwidefsky@de.ibm.com)
8 * Derived from "arch/i386/kernel/setup.c"
9 * Copyright (C) 1995, Linus Torvalds
13 * This file handles the architecture-dependent parts of initialization
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
19 #include <linux/errno.h>
20 #include <linux/export.h>
21 #include <linux/sched.h>
22 #include <linux/sched/task.h>
23 #include <linux/cpu.h>
24 #include <linux/kernel.h>
25 #include <linux/memblock.h>
27 #include <linux/stddef.h>
28 #include <linux/unistd.h>
29 #include <linux/ptrace.h>
30 #include <linux/random.h>
31 #include <linux/user.h>
32 #include <linux/tty.h>
33 #include <linux/ioport.h>
34 #include <linux/delay.h>
35 #include <linux/init.h>
36 #include <linux/initrd.h>
37 #include <linux/root_dev.h>
38 #include <linux/console.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/dma-contiguous.h>
41 #include <linux/device.h>
42 #include <linux/notifier.h>
43 #include <linux/pfn.h>
44 #include <linux/ctype.h>
45 #include <linux/reboot.h>
46 #include <linux/topology.h>
47 #include <linux/kexec.h>
48 #include <linux/crash_dump.h>
49 #include <linux/memory.h>
50 #include <linux/compat.h>
51 #include <linux/start_kernel.h>
53 #include <asm/boot_data.h>
55 #include <asm/facility.h>
57 #include <asm/mmu_context.h>
58 #include <asm/cpcmd.h>
59 #include <asm/lowcore.h>
63 #include <asm/ptrace.h>
64 #include <asm/sections.h>
65 #include <asm/ebcdic.h>
67 #include <asm/os_info.h>
69 #include <asm/stacktrace.h>
70 #include <asm/sysinfo.h>
72 #include <asm/alternative.h>
73 #include <asm/nospec-branch.h>
74 #include <asm/mem_detect.h>
76 #include <asm/asm-offsets.h>
82 unsigned int console_mode
= 0;
83 EXPORT_SYMBOL(console_mode
);
85 unsigned int console_devno
= -1;
86 EXPORT_SYMBOL(console_devno
);
88 unsigned int console_irq
= -1;
89 EXPORT_SYMBOL(console_irq
);
91 unsigned long elf_hwcap __read_mostly
= 0;
92 char elf_platform
[ELF_PLATFORM_SIZE
];
94 unsigned long int_hwcap
= 0;
96 int __bootdata(noexec_disabled
);
97 int __bootdata(memory_end_set
);
98 unsigned long __bootdata(memory_end
);
99 unsigned long __bootdata(vmalloc_size
);
100 unsigned long __bootdata(max_physmem_end
);
101 struct mem_detect_info
__bootdata(mem_detect
);
103 struct exception_table_entry
*__bootdata_preserved(__start_dma_ex_table
);
104 struct exception_table_entry
*__bootdata_preserved(__stop_dma_ex_table
);
105 unsigned long __bootdata_preserved(__swsusp_reset_dma
);
106 unsigned long __bootdata_preserved(__stext_dma
);
107 unsigned long __bootdata_preserved(__etext_dma
);
108 unsigned long __bootdata_preserved(__sdma
);
109 unsigned long __bootdata_preserved(__edma
);
110 unsigned long __bootdata_preserved(__kaslr_offset
);
111 unsigned int __bootdata_preserved(zlib_dfltcc_support
);
112 EXPORT_SYMBOL(zlib_dfltcc_support
);
114 unsigned long VMALLOC_START
;
115 EXPORT_SYMBOL(VMALLOC_START
);
117 unsigned long VMALLOC_END
;
118 EXPORT_SYMBOL(VMALLOC_END
);
120 struct page
*vmemmap
;
121 EXPORT_SYMBOL(vmemmap
);
123 unsigned long MODULES_VADDR
;
124 unsigned long MODULES_END
;
126 /* An array with a pointer to the lowcore of every CPU. */
127 struct lowcore
*lowcore_ptr
[NR_CPUS
];
128 EXPORT_SYMBOL(lowcore_ptr
);
131 * This is set up by the setup-routine at boot-time
132 * for S390 need to find out, what we have to setup
133 * using address 0x10400 ...
136 #include <asm/setup.h>
139 * condev= and conmode= setup parameter.
142 static int __init
condev_setup(char *str
)
146 vdev
= simple_strtoul(str
, &str
, 0);
147 if (vdev
>= 0 && vdev
< 65536) {
148 console_devno
= vdev
;
154 __setup("condev=", condev_setup
);
156 static void __init
set_preferred_console(void)
158 if (CONSOLE_IS_3215
|| CONSOLE_IS_SCLP
)
159 add_preferred_console("ttyS", 0, NULL
);
160 else if (CONSOLE_IS_3270
)
161 add_preferred_console("tty3270", 0, NULL
);
162 else if (CONSOLE_IS_VT220
)
163 add_preferred_console("ttyS", 1, NULL
);
164 else if (CONSOLE_IS_HVC
)
165 add_preferred_console("hvc", 0, NULL
);
168 static int __init
conmode_setup(char *str
)
170 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
171 if (!strcmp(str
, "hwc") || !strcmp(str
, "sclp"))
174 #if defined(CONFIG_TN3215_CONSOLE)
175 if (!strcmp(str
, "3215"))
178 #if defined(CONFIG_TN3270_CONSOLE)
179 if (!strcmp(str
, "3270"))
182 set_preferred_console();
186 __setup("conmode=", conmode_setup
);
188 static void __init
conmode_default(void)
190 char query_buffer
[1024];
194 cpcmd("QUERY CONSOLE", query_buffer
, 1024, NULL
);
195 console_devno
= simple_strtoul(query_buffer
+ 5, NULL
, 16);
196 ptr
= strstr(query_buffer
, "SUBCHANNEL =");
197 console_irq
= simple_strtoul(ptr
+ 13, NULL
, 16);
198 cpcmd("QUERY TERM", query_buffer
, 1024, NULL
);
199 ptr
= strstr(query_buffer
, "CONMODE");
201 * Set the conmode to 3215 so that the device recognition
202 * will set the cu_type of the console to 3215. If the
203 * conmode is 3270 and we don't set it back then both
204 * 3215 and the 3270 driver will try to access the console
205 * device (3215 as console and 3270 as normal tty).
207 cpcmd("TERM CONMODE 3215", NULL
, 0, NULL
);
209 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
214 if (str_has_prefix(ptr
+ 8, "3270")) {
215 #if defined(CONFIG_TN3270_CONSOLE)
217 #elif defined(CONFIG_TN3215_CONSOLE)
219 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
222 } else if (str_has_prefix(ptr
+ 8, "3215")) {
223 #if defined(CONFIG_TN3215_CONSOLE)
225 #elif defined(CONFIG_TN3270_CONSOLE)
227 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
231 } else if (MACHINE_IS_KVM
) {
232 if (sclp
.has_vt220
&& IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE
))
234 else if (sclp
.has_linemode
&& IS_ENABLED(CONFIG_SCLP_CONSOLE
))
239 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
245 #ifdef CONFIG_CRASH_DUMP
246 static void __init
setup_zfcpdump(void)
248 if (ipl_info
.type
!= IPL_TYPE_FCP_DUMP
)
252 strcat(boot_command_line
, " cio_ignore=all,!ipldev,!condev");
253 console_loglevel
= 2;
256 static inline void setup_zfcpdump(void) {}
257 #endif /* CONFIG_CRASH_DUMP */
260 * Reboot, halt and power_off stubs. They just call _machine_restart,
261 * _machine_halt or _machine_power_off.
264 void machine_restart(char *command
)
266 if ((!in_interrupt() && !in_atomic()) || oops_in_progress
)
268 * Only unblank the console if we are called in enabled
269 * context or a bust_spinlocks cleared the way for us.
272 _machine_restart(command
);
275 void machine_halt(void)
277 if (!in_interrupt() || oops_in_progress
)
279 * Only unblank the console if we are called in enabled
280 * context or a bust_spinlocks cleared the way for us.
286 void machine_power_off(void)
288 if (!in_interrupt() || oops_in_progress
)
290 * Only unblank the console if we are called in enabled
291 * context or a bust_spinlocks cleared the way for us.
294 _machine_power_off();
298 * Dummy power off function.
300 void (*pm_power_off
)(void) = machine_power_off
;
301 EXPORT_SYMBOL_GPL(pm_power_off
);
303 void *restart_stack
__section(.data
);
305 unsigned long stack_alloc(void)
307 #ifdef CONFIG_VMAP_STACK
308 return (unsigned long)__vmalloc_node(THREAD_SIZE
, THREAD_SIZE
,
309 THREADINFO_GFP
, NUMA_NO_NODE
,
310 __builtin_return_address(0));
312 return __get_free_pages(GFP_KERNEL
, THREAD_SIZE_ORDER
);
316 void stack_free(unsigned long stack
)
318 #ifdef CONFIG_VMAP_STACK
319 vfree((void *) stack
);
321 free_pages(stack
, THREAD_SIZE_ORDER
);
325 int __init
arch_early_irq_init(void)
329 stack
= __get_free_pages(GFP_KERNEL
, THREAD_SIZE_ORDER
);
331 panic("Couldn't allocate async stack");
332 S390_lowcore
.async_stack
= stack
+ STACK_INIT_OFFSET
;
336 static int __init
async_stack_realloc(void)
338 unsigned long old
, new;
340 old
= S390_lowcore
.async_stack
- STACK_INIT_OFFSET
;
343 panic("Couldn't allocate async stack");
344 S390_lowcore
.async_stack
= new + STACK_INIT_OFFSET
;
345 free_pages(old
, THREAD_SIZE_ORDER
);
348 early_initcall(async_stack_realloc
);
350 void __init
arch_call_rest_init(void)
354 stack
= stack_alloc();
356 panic("Couldn't allocate kernel stack");
357 current
->stack
= (void *) stack
;
358 #ifdef CONFIG_VMAP_STACK
359 current
->stack_vm_area
= (void *) stack
;
361 set_task_stack_end_magic(current
);
362 stack
+= STACK_INIT_OFFSET
;
363 S390_lowcore
.kernel_stack
= stack
;
364 CALL_ON_STACK_NORETURN(rest_init
, stack
);
367 static void __init
setup_lowcore_dat_off(void)
372 * Setup lowcore for boot cpu
374 BUILD_BUG_ON(sizeof(struct lowcore
) != LC_PAGES
* PAGE_SIZE
);
375 lc
= memblock_alloc_low(sizeof(*lc
), sizeof(*lc
));
377 panic("%s: Failed to allocate %zu bytes align=%zx\n",
378 __func__
, sizeof(*lc
), sizeof(*lc
));
380 lc
->restart_psw
.mask
= PSW_KERNEL_BITS
;
381 lc
->restart_psw
.addr
= (unsigned long) restart_int_handler
;
382 lc
->external_new_psw
.mask
= PSW_KERNEL_BITS
| PSW_MASK_MCHECK
;
383 lc
->external_new_psw
.addr
= (unsigned long) ext_int_handler
;
384 lc
->svc_new_psw
.mask
= PSW_KERNEL_BITS
|
385 PSW_MASK_IO
| PSW_MASK_EXT
| PSW_MASK_MCHECK
;
386 lc
->svc_new_psw
.addr
= (unsigned long) system_call
;
387 lc
->program_new_psw
.mask
= PSW_KERNEL_BITS
| PSW_MASK_MCHECK
;
388 lc
->program_new_psw
.addr
= (unsigned long) pgm_check_handler
;
389 lc
->mcck_new_psw
.mask
= PSW_KERNEL_BITS
;
390 lc
->mcck_new_psw
.addr
= (unsigned long) mcck_int_handler
;
391 lc
->io_new_psw
.mask
= PSW_KERNEL_BITS
| PSW_MASK_MCHECK
;
392 lc
->io_new_psw
.addr
= (unsigned long) io_int_handler
;
393 lc
->clock_comparator
= clock_comparator_max
;
394 lc
->nodat_stack
= ((unsigned long) &init_thread_union
)
395 + THREAD_SIZE
- STACK_FRAME_OVERHEAD
- sizeof(struct pt_regs
);
396 lc
->current_task
= (unsigned long)&init_task
;
398 lc
->machine_flags
= S390_lowcore
.machine_flags
;
399 lc
->preempt_count
= S390_lowcore
.preempt_count
;
400 lc
->stfl_fac_list
= S390_lowcore
.stfl_fac_list
;
401 memcpy(lc
->stfle_fac_list
, S390_lowcore
.stfle_fac_list
,
402 sizeof(lc
->stfle_fac_list
));
403 memcpy(lc
->alt_stfle_fac_list
, S390_lowcore
.alt_stfle_fac_list
,
404 sizeof(lc
->alt_stfle_fac_list
));
405 nmi_alloc_boot_cpu(lc
);
406 vdso_alloc_boot_cpu(lc
);
407 lc
->sync_enter_timer
= S390_lowcore
.sync_enter_timer
;
408 lc
->async_enter_timer
= S390_lowcore
.async_enter_timer
;
409 lc
->exit_timer
= S390_lowcore
.exit_timer
;
410 lc
->user_timer
= S390_lowcore
.user_timer
;
411 lc
->system_timer
= S390_lowcore
.system_timer
;
412 lc
->steal_timer
= S390_lowcore
.steal_timer
;
413 lc
->last_update_timer
= S390_lowcore
.last_update_timer
;
414 lc
->last_update_clock
= S390_lowcore
.last_update_clock
;
417 * Allocate the global restart stack which is the same for
418 * all CPUs in cast *one* of them does a PSW restart.
420 restart_stack
= memblock_alloc(THREAD_SIZE
, THREAD_SIZE
);
422 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
423 __func__
, THREAD_SIZE
, THREAD_SIZE
);
424 restart_stack
+= STACK_INIT_OFFSET
;
427 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
428 * restart data to the absolute zero lowcore. This is necessary if
429 * PSW restart is done on an offline CPU that has lowcore zero.
431 lc
->restart_stack
= (unsigned long) restart_stack
;
432 lc
->restart_fn
= (unsigned long) do_restart
;
433 lc
->restart_data
= 0;
434 lc
->restart_source
= -1UL;
436 /* Setup absolute zero lowcore */
437 mem_assign_absolute(S390_lowcore
.restart_stack
, lc
->restart_stack
);
438 mem_assign_absolute(S390_lowcore
.restart_fn
, lc
->restart_fn
);
439 mem_assign_absolute(S390_lowcore
.restart_data
, lc
->restart_data
);
440 mem_assign_absolute(S390_lowcore
.restart_source
, lc
->restart_source
);
441 mem_assign_absolute(S390_lowcore
.restart_psw
, lc
->restart_psw
);
443 lc
->spinlock_lockval
= arch_spin_lockval(0);
444 lc
->spinlock_index
= 0;
445 arch_spin_lock_setup(0);
446 lc
->br_r1_trampoline
= 0x07f1; /* br %r1 */
447 lc
->return_lpswe
= gen_lpswe(__LC_RETURN_PSW
);
448 lc
->return_mcck_lpswe
= gen_lpswe(__LC_RETURN_MCCK_PSW
);
450 set_prefix((u32
)(unsigned long) lc
);
454 static void __init
setup_lowcore_dat_on(void)
456 __ctl_clear_bit(0, 28);
457 S390_lowcore
.external_new_psw
.mask
|= PSW_MASK_DAT
;
458 S390_lowcore
.svc_new_psw
.mask
|= PSW_MASK_DAT
;
459 S390_lowcore
.program_new_psw
.mask
|= PSW_MASK_DAT
;
460 S390_lowcore
.io_new_psw
.mask
|= PSW_MASK_DAT
;
461 __ctl_set_bit(0, 28);
464 static struct resource code_resource
= {
465 .name
= "Kernel code",
466 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
,
469 static struct resource data_resource
= {
470 .name
= "Kernel data",
471 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
,
474 static struct resource bss_resource
= {
475 .name
= "Kernel bss",
476 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
,
479 static struct resource __initdata
*standard_resources
[] = {
485 static void __init
setup_resources(void)
487 struct resource
*res
, *std_res
, *sub_res
;
488 struct memblock_region
*reg
;
491 code_resource
.start
= (unsigned long) _text
;
492 code_resource
.end
= (unsigned long) _etext
- 1;
493 data_resource
.start
= (unsigned long) _etext
;
494 data_resource
.end
= (unsigned long) _edata
- 1;
495 bss_resource
.start
= (unsigned long) __bss_start
;
496 bss_resource
.end
= (unsigned long) __bss_stop
- 1;
498 for_each_memblock(memory
, reg
) {
499 res
= memblock_alloc(sizeof(*res
), 8);
501 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
502 __func__
, sizeof(*res
), 8);
503 res
->flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
;
505 res
->name
= "System RAM";
506 res
->start
= reg
->base
;
507 res
->end
= reg
->base
+ reg
->size
- 1;
508 request_resource(&iomem_resource
, res
);
510 for (j
= 0; j
< ARRAY_SIZE(standard_resources
); j
++) {
511 std_res
= standard_resources
[j
];
512 if (std_res
->start
< res
->start
||
513 std_res
->start
> res
->end
)
515 if (std_res
->end
> res
->end
) {
516 sub_res
= memblock_alloc(sizeof(*sub_res
), 8);
518 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
519 __func__
, sizeof(*sub_res
), 8);
521 sub_res
->end
= res
->end
;
522 std_res
->start
= res
->end
+ 1;
523 request_resource(res
, sub_res
);
525 request_resource(res
, std_res
);
529 #ifdef CONFIG_CRASH_DUMP
531 * Re-add removed crash kernel memory as reserved memory. This makes
532 * sure it will be mapped with the identity mapping and struct pages
533 * will be created, so it can be resized later on.
534 * However add it later since the crash kernel resource should not be
535 * part of the System RAM resource.
537 if (crashk_res
.end
) {
538 memblock_add_node(crashk_res
.start
, resource_size(&crashk_res
), 0);
539 memblock_reserve(crashk_res
.start
, resource_size(&crashk_res
));
540 insert_resource(&iomem_resource
, &crashk_res
);
545 static void __init
setup_memory_end(void)
547 unsigned long vmax
, tmp
;
549 /* Choose kernel address space layout: 3 or 4 levels. */
550 if (IS_ENABLED(CONFIG_KASAN
)) {
551 vmax
= IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING
)
555 tmp
= (memory_end
?: max_physmem_end
) / PAGE_SIZE
;
556 tmp
= tmp
* (sizeof(struct page
) + PAGE_SIZE
);
557 if (tmp
+ vmalloc_size
+ MODULES_LEN
<= _REGION2_SIZE
)
558 vmax
= _REGION2_SIZE
; /* 3-level kernel page table */
560 vmax
= _REGION1_SIZE
; /* 4-level kernel page table */
563 if (is_prot_virt_host())
564 adjust_to_uv_max(&vmax
);
566 /* module area is at the end of the kernel address space. */
568 MODULES_VADDR
= MODULES_END
- MODULES_LEN
;
569 VMALLOC_END
= MODULES_VADDR
;
570 VMALLOC_START
= VMALLOC_END
- vmalloc_size
;
572 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
573 tmp
= VMALLOC_START
/ (PAGE_SIZE
+ sizeof(struct page
));
574 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
575 tmp
= SECTION_ALIGN_UP(tmp
);
576 tmp
= VMALLOC_START
- tmp
* sizeof(struct page
);
577 tmp
&= ~((vmax
>> 11) - 1); /* align to page table level */
578 tmp
= min(tmp
, 1UL << MAX_PHYSMEM_BITS
);
579 vmemmap
= (struct page
*) tmp
;
581 /* Take care that memory_end is set and <= vmemmap */
582 memory_end
= min(memory_end
?: max_physmem_end
, (unsigned long)vmemmap
);
584 /* fit in kasan shadow memory region between 1:1 and vmemmap */
585 memory_end
= min(memory_end
, KASAN_SHADOW_START
);
586 vmemmap
= max(vmemmap
, (struct page
*)KASAN_SHADOW_END
);
588 max_pfn
= max_low_pfn
= PFN_DOWN(memory_end
);
589 memblock_remove(memory_end
, ULONG_MAX
);
591 pr_notice("The maximum memory size is %luMB\n", memory_end
>> 20);
594 #ifdef CONFIG_CRASH_DUMP
597 * When kdump is enabled, we have to ensure that no memory from
598 * the area [0 - crashkernel memory size] and
599 * [crashk_res.start - crashk_res.end] is set offline.
601 static int kdump_mem_notifier(struct notifier_block
*nb
,
602 unsigned long action
, void *data
)
604 struct memory_notify
*arg
= data
;
606 if (action
!= MEM_GOING_OFFLINE
)
608 if (arg
->start_pfn
< PFN_DOWN(resource_size(&crashk_res
)))
610 if (arg
->start_pfn
> PFN_DOWN(crashk_res
.end
))
612 if (arg
->start_pfn
+ arg
->nr_pages
- 1 < PFN_DOWN(crashk_res
.start
))
617 static struct notifier_block kdump_mem_nb
= {
618 .notifier_call
= kdump_mem_notifier
,
624 * Make sure that the area behind memory_end is protected
626 static void reserve_memory_end(void)
629 memblock_reserve(memory_end
, ULONG_MAX
);
633 * Make sure that oldmem, where the dump is stored, is protected
635 static void reserve_oldmem(void)
637 #ifdef CONFIG_CRASH_DUMP
639 /* Forget all memory above the running kdump system */
640 memblock_reserve(OLDMEM_SIZE
, (phys_addr_t
)ULONG_MAX
);
645 * Make sure that oldmem, where the dump is stored, is protected
647 static void remove_oldmem(void)
649 #ifdef CONFIG_CRASH_DUMP
651 /* Forget all memory above the running kdump system */
652 memblock_remove(OLDMEM_SIZE
, (phys_addr_t
)ULONG_MAX
);
657 * Reserve memory for kdump kernel to be loaded with kexec
659 static void __init
reserve_crashkernel(void)
661 #ifdef CONFIG_CRASH_DUMP
662 unsigned long long crash_base
, crash_size
;
663 phys_addr_t low
, high
;
666 rc
= parse_crashkernel(boot_command_line
, memory_end
, &crash_size
,
669 crash_base
= ALIGN(crash_base
, KEXEC_CRASH_MEM_ALIGN
);
670 crash_size
= ALIGN(crash_size
, KEXEC_CRASH_MEM_ALIGN
);
671 if (rc
|| crash_size
== 0)
674 if (memblock
.memory
.regions
[0].size
< crash_size
) {
675 pr_info("crashkernel reservation failed: %s\n",
676 "first memory chunk must be at least crashkernel size");
680 low
= crash_base
?: OLDMEM_BASE
;
681 high
= low
+ crash_size
;
682 if (low
>= OLDMEM_BASE
&& high
<= OLDMEM_BASE
+ OLDMEM_SIZE
) {
683 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
686 /* Find suitable area in free memory */
687 low
= max_t(unsigned long, crash_size
, sclp
.hsa_size
);
688 high
= crash_base
? crash_base
+ crash_size
: ULONG_MAX
;
690 if (crash_base
&& crash_base
< low
) {
691 pr_info("crashkernel reservation failed: %s\n",
692 "crash_base too low");
695 low
= crash_base
?: low
;
696 crash_base
= memblock_find_in_range(low
, high
, crash_size
,
697 KEXEC_CRASH_MEM_ALIGN
);
701 pr_info("crashkernel reservation failed: %s\n",
702 "no suitable area found");
706 if (register_memory_notifier(&kdump_mem_nb
))
709 if (!OLDMEM_BASE
&& MACHINE_IS_VM
)
710 diag10_range(PFN_DOWN(crash_base
), PFN_DOWN(crash_size
));
711 crashk_res
.start
= crash_base
;
712 crashk_res
.end
= crash_base
+ crash_size
- 1;
713 memblock_remove(crash_base
, crash_size
);
714 pr_info("Reserving %lluMB of memory at %lluMB "
715 "for crashkernel (System RAM: %luMB)\n",
716 crash_size
>> 20, crash_base
>> 20,
717 (unsigned long)memblock
.memory
.total_size
>> 20);
718 os_info_crashkernel_add(crash_base
, crash_size
);
723 * Reserve the initrd from being used by memblock
725 static void __init
reserve_initrd(void)
727 #ifdef CONFIG_BLK_DEV_INITRD
728 if (!INITRD_START
|| !INITRD_SIZE
)
730 initrd_start
= INITRD_START
;
731 initrd_end
= initrd_start
+ INITRD_SIZE
;
732 memblock_reserve(INITRD_START
, INITRD_SIZE
);
737 * Reserve the memory area used to pass the certificate lists
739 static void __init
reserve_certificate_list(void)
741 if (ipl_cert_list_addr
)
742 memblock_reserve(ipl_cert_list_addr
, ipl_cert_list_size
);
745 static void __init
reserve_mem_detect_info(void)
747 unsigned long start
, size
;
749 get_mem_detect_reserved(&start
, &size
);
751 memblock_reserve(start
, size
);
754 static void __init
free_mem_detect_info(void)
756 unsigned long start
, size
;
758 get_mem_detect_reserved(&start
, &size
);
760 memblock_free(start
, size
);
763 static const char * __init
get_mem_info_source(void)
765 switch (mem_detect
.info_source
) {
766 case MEM_DETECT_SCLP_STOR_INFO
:
767 return "sclp storage info";
768 case MEM_DETECT_DIAG260
:
770 case MEM_DETECT_SCLP_READ_INFO
:
771 return "sclp read info";
772 case MEM_DETECT_BIN_SEARCH
:
773 return "binary search";
778 static void __init
memblock_add_mem_detect_info(void)
780 unsigned long start
, end
;
783 memblock_dbg("physmem info source: %s (%hhd)\n",
784 get_mem_info_source(), mem_detect
.info_source
);
785 /* keep memblock lists close to the kernel */
786 memblock_set_bottom_up(true);
787 for_each_mem_detect_block(i
, &start
, &end
) {
788 memblock_add(start
, end
- start
);
789 memblock_physmem_add(start
, end
- start
);
791 memblock_set_bottom_up(false);
792 memblock_set_node(0, ULONG_MAX
, &memblock
.memory
, 0);
797 * Check for initrd being in usable memory
799 static void __init
check_initrd(void)
801 #ifdef CONFIG_BLK_DEV_INITRD
802 if (INITRD_START
&& INITRD_SIZE
&&
803 !memblock_is_region_memory(INITRD_START
, INITRD_SIZE
)) {
804 pr_err("The initial RAM disk does not fit into the memory\n");
805 memblock_free(INITRD_START
, INITRD_SIZE
);
806 initrd_start
= initrd_end
= 0;
812 * Reserve memory used for lowcore/command line/kernel image.
814 static void __init
reserve_kernel(void)
816 unsigned long start_pfn
= PFN_UP(__pa(_end
));
818 memblock_reserve(0, HEAD_END
);
819 memblock_reserve((unsigned long)_stext
, PFN_PHYS(start_pfn
)
820 - (unsigned long)_stext
);
821 memblock_reserve(__sdma
, __edma
- __sdma
);
824 static void __init
setup_memory(void)
826 struct memblock_region
*reg
;
829 * Init storage key for present memory
831 for_each_memblock(memory
, reg
) {
832 storage_key_init_range(reg
->base
, reg
->base
+ reg
->size
);
834 psw_set_key(PAGE_DEFAULT_KEY
);
837 memblock_enforce_memory_limit(memblock_end_of_DRAM());
841 * Setup hardware capabilities.
843 static int __init
setup_hwcaps(void)
845 static const int stfl_bits
[6] = { 0, 2, 7, 17, 19, 21 };
850 * The store facility list bits numbers as found in the principles
851 * of operation are numbered with bit 1UL<<31 as number 0 to
852 * bit 1UL<<0 as number 31.
853 * Bit 0: instructions named N3, "backported" to esa-mode
854 * Bit 2: z/Architecture mode is active
855 * Bit 7: the store-facility-list-extended facility is installed
856 * Bit 17: the message-security assist is installed
857 * Bit 19: the long-displacement facility is installed
858 * Bit 21: the extended-immediate facility is installed
859 * Bit 22: extended-translation facility 3 is installed
860 * Bit 30: extended-translation facility 3 enhancement facility
861 * These get translated to:
862 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
863 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
864 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
865 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
867 for (i
= 0; i
< 6; i
++)
868 if (test_facility(stfl_bits
[i
]))
869 elf_hwcap
|= 1UL << i
;
871 if (test_facility(22) && test_facility(30))
872 elf_hwcap
|= HWCAP_S390_ETF3EH
;
875 * Check for additional facilities with store-facility-list-extended.
876 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
877 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
878 * as stored by stfl, bits 32-xxx contain additional facilities.
879 * How many facility words are stored depends on the number of
880 * doublewords passed to the instruction. The additional facilities
882 * Bit 42: decimal floating point facility is installed
883 * Bit 44: perform floating point operation facility is installed
885 * HWCAP_S390_DFP bit 6 (42 && 44).
887 if ((elf_hwcap
& (1UL << 2)) && test_facility(42) && test_facility(44))
888 elf_hwcap
|= HWCAP_S390_DFP
;
891 * Huge page support HWCAP_S390_HPAGE is bit 7.
893 if (MACHINE_HAS_EDAT1
)
894 elf_hwcap
|= HWCAP_S390_HPAGE
;
897 * 64-bit register support for 31-bit processes
898 * HWCAP_S390_HIGH_GPRS is bit 9.
900 elf_hwcap
|= HWCAP_S390_HIGH_GPRS
;
903 * Transactional execution support HWCAP_S390_TE is bit 10.
906 elf_hwcap
|= HWCAP_S390_TE
;
909 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
910 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
911 * instead of facility bit 129.
913 if (MACHINE_HAS_VX
) {
914 elf_hwcap
|= HWCAP_S390_VXRS
;
915 if (test_facility(134))
916 elf_hwcap
|= HWCAP_S390_VXRS_EXT
;
917 if (test_facility(135))
918 elf_hwcap
|= HWCAP_S390_VXRS_BCD
;
919 if (test_facility(148))
920 elf_hwcap
|= HWCAP_S390_VXRS_EXT2
;
921 if (test_facility(152))
922 elf_hwcap
|= HWCAP_S390_VXRS_PDE
;
924 if (test_facility(150))
925 elf_hwcap
|= HWCAP_S390_SORT
;
926 if (test_facility(151))
927 elf_hwcap
|= HWCAP_S390_DFLT
;
930 * Guarded storage support HWCAP_S390_GS is bit 12.
933 elf_hwcap
|= HWCAP_S390_GS
;
936 add_device_randomness(&cpu_id
, sizeof(cpu_id
));
937 switch (cpu_id
.machine
) {
940 default: /* Use "z900" as default for 64 bit kernels. */
941 strcpy(elf_platform
, "z900");
945 strcpy(elf_platform
, "z990");
949 strcpy(elf_platform
, "z9-109");
953 strcpy(elf_platform
, "z10");
957 strcpy(elf_platform
, "z196");
961 strcpy(elf_platform
, "zEC12");
965 strcpy(elf_platform
, "z13");
969 strcpy(elf_platform
, "z14");
973 strcpy(elf_platform
, "z15");
978 * Virtualization support HWCAP_INT_SIE is bit 0.
981 int_hwcap
|= HWCAP_INT_SIE
;
985 arch_initcall(setup_hwcaps
);
988 * Add system information as device randomness
990 static void __init
setup_randomness(void)
992 struct sysinfo_3_2_2
*vmms
;
994 vmms
= (struct sysinfo_3_2_2
*) memblock_phys_alloc(PAGE_SIZE
,
997 panic("Failed to allocate memory for sysinfo structure\n");
999 if (stsi(vmms
, 3, 2, 2) == 0 && vmms
->count
)
1000 add_device_randomness(&vmms
->vm
, sizeof(vmms
->vm
[0]) * vmms
->count
);
1001 memblock_free((unsigned long) vmms
, PAGE_SIZE
);
1005 * Find the correct size for the task_struct. This depends on
1006 * the size of the struct fpu at the end of the thread_struct
1007 * which is embedded in the task_struct.
1009 static void __init
setup_task_size(void)
1011 int task_size
= sizeof(struct task_struct
);
1013 if (!MACHINE_HAS_VX
) {
1014 task_size
-= sizeof(__vector128
) * __NUM_VXRS
;
1015 task_size
+= sizeof(freg_t
) * __NUM_FPRS
;
1017 arch_task_struct_size
= task_size
;
1021 * Issue diagnose 318 to set the control program name and
1024 static void __init
setup_control_program_code(void)
1026 union diag318_info diag318_info
= {
1032 if (!sclp
.has_diag318
)
1035 diag_stat_inc(DIAG_STAT_X318
);
1036 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info
.val
));
1040 * Print the component list from the IPL report
1042 static void __init
log_component_list(void)
1044 struct ipl_rb_component_entry
*ptr
, *end
;
1047 if (!early_ipl_comp_list_addr
)
1049 if (ipl_block
.hdr
.flags
& IPL_PL_FLAG_SIPL
)
1050 pr_info("Linux is running with Secure-IPL enabled\n");
1052 pr_info("Linux is running with Secure-IPL disabled\n");
1053 ptr
= (void *) early_ipl_comp_list_addr
;
1054 end
= (void *) ptr
+ early_ipl_comp_list_size
;
1055 pr_info("The IPL report contains the following components:\n");
1057 if (ptr
->flags
& IPL_RB_COMPONENT_FLAG_SIGNED
) {
1058 if (ptr
->flags
& IPL_RB_COMPONENT_FLAG_VERIFIED
)
1059 str
= "signed, verified";
1061 str
= "signed, verification failed";
1065 pr_info("%016llx - %016llx (%s)\n",
1066 ptr
->addr
, ptr
->addr
+ ptr
->len
, str
);
1072 * Setup function called from init/main.c just after the banner
1076 void __init
setup_arch(char **cmdline_p
)
1079 * print what head.S has found out about the machine
1082 pr_info("Linux is running as a z/VM "
1083 "guest operating system in 64-bit mode\n");
1084 else if (MACHINE_IS_KVM
)
1085 pr_info("Linux is running under KVM in 64-bit mode\n");
1086 else if (MACHINE_IS_LPAR
)
1087 pr_info("Linux is running natively in 64-bit mode\n");
1089 pr_info("Linux is running as a guest in 64-bit mode\n");
1091 log_component_list();
1093 /* Have one command line that is parsed and saved in /proc/cmdline */
1094 /* boot_command_line has been already set up in early.c */
1095 *cmdline_p
= boot_command_line
;
1097 ROOT_DEV
= Root_RAM0
;
1099 init_mm
.start_code
= (unsigned long) _text
;
1100 init_mm
.end_code
= (unsigned long) _etext
;
1101 init_mm
.end_data
= (unsigned long) _edata
;
1102 init_mm
.brk
= (unsigned long) _end
;
1104 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO
))
1105 nospec_auto_detect();
1107 parse_early_param();
1108 #ifdef CONFIG_CRASH_DUMP
1109 /* Deactivate elfcorehdr= kernel parameter */
1110 elfcorehdr_addr
= ELFCORE_ADDR_MAX
;
1116 setup_control_program_code();
1118 /* Do some memory reservations *before* memory is added to memblock */
1119 reserve_memory_end();
1123 reserve_certificate_list();
1124 reserve_mem_detect_info();
1125 memblock_allow_resize();
1127 /* Get information about *all* installed memory */
1128 memblock_add_mem_detect_info();
1130 free_mem_detect_info();
1134 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1135 * extra checks that HOLES_IN_ZONE would require.
1137 * Is this still required?
1139 memblock_trim_memory(1UL << (MAX_ORDER
- 1 + PAGE_SHIFT
));
1141 if (is_prot_virt_host())
1145 dma_contiguous_reserve(memory_end
);
1149 reserve_crashkernel();
1150 #ifdef CONFIG_CRASH_DUMP
1152 * Be aware that smp_save_dump_cpus() triggers a system reset.
1153 * Therefore CPU and device initialization should be done afterwards.
1155 smp_save_dump_cpus();
1159 setup_lowcore_dat_off();
1160 smp_fill_possible_mask();
1161 cpu_detect_mhz_feature();
1165 topology_init_early();
1168 * Create kernel page tables and switch to virtual addressing.
1173 * After paging_init created the kernel page table, the new PSWs
1174 * in lowcore can now run with DAT enabled.
1176 setup_lowcore_dat_on();
1178 /* Setup default console */
1180 set_preferred_console();
1182 apply_alternative_instructions();
1183 if (IS_ENABLED(CONFIG_EXPOLINE
))
1184 nospec_init_branches();
1186 /* Setup zfcpdump support */
1189 /* Add system specific data to the random pool */