]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * Copyright (C) 1995 Linus Torvalds |
3 | * | |
4 | * Pentium III FXSR, SSE support | |
5 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
6 | */ | |
7 | ||
8 | /* | |
9 | * This file handles the architecture-dependent parts of process handling.. | |
10 | */ | |
11 | ||
f3705136 | 12 | #include <linux/cpu.h> |
1da177e4 LT |
13 | #include <linux/errno.h> |
14 | #include <linux/sched.h> | |
29930025 | 15 | #include <linux/sched/task.h> |
68db0cf1 | 16 | #include <linux/sched/task_stack.h> |
1da177e4 LT |
17 | #include <linux/fs.h> |
18 | #include <linux/kernel.h> | |
19 | #include <linux/mm.h> | |
20 | #include <linux/elfcore.h> | |
21 | #include <linux/smp.h> | |
1da177e4 LT |
22 | #include <linux/stddef.h> |
23 | #include <linux/slab.h> | |
24 | #include <linux/vmalloc.h> | |
25 | #include <linux/user.h> | |
1da177e4 | 26 | #include <linux/interrupt.h> |
1da177e4 LT |
27 | #include <linux/delay.h> |
28 | #include <linux/reboot.h> | |
1da177e4 | 29 | #include <linux/mc146818rtc.h> |
186f4360 | 30 | #include <linux/export.h> |
1da177e4 LT |
31 | #include <linux/kallsyms.h> |
32 | #include <linux/ptrace.h> | |
c16b63e0 | 33 | #include <linux/personality.h> |
7c3576d2 | 34 | #include <linux/percpu.h> |
529e25f6 | 35 | #include <linux/prctl.h> |
8b96f011 | 36 | #include <linux/ftrace.h> |
befa9e78 JSR |
37 | #include <linux/uaccess.h> |
38 | #include <linux/io.h> | |
39 | #include <linux/kdebug.h> | |
79170fda | 40 | #include <linux/syscalls.h> |
1da177e4 | 41 | |
1da177e4 | 42 | #include <asm/pgtable.h> |
1da177e4 LT |
43 | #include <asm/ldt.h> |
44 | #include <asm/processor.h> | |
78f7f1e5 | 45 | #include <asm/fpu/internal.h> |
1da177e4 LT |
46 | #include <asm/desc.h> |
47 | #ifdef CONFIG_MATH_EMULATION | |
48 | #include <asm/math_emu.h> | |
49 | #endif | |
50 | ||
1da177e4 LT |
51 | #include <linux/err.h> |
52 | ||
f3705136 ZM |
53 | #include <asm/tlbflush.h> |
54 | #include <asm/cpu.h> | |
bbc1f698 | 55 | #include <asm/syscalls.h> |
66cb5917 | 56 | #include <asm/debugreg.h> |
f05e798a | 57 | #include <asm/switch_to.h> |
ba3e127e | 58 | #include <asm/vm86.h> |
05830204 | 59 | #include <asm/intel_rdt_sched.h> |
79170fda | 60 | #include <asm/proto.h> |
f3705136 | 61 | |
e2ce07c8 | 62 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
63 | { |
64 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; | |
bb1995d5 | 65 | unsigned long d0, d1, d2, d3, d6, d7; |
65ea5b03 | 66 | unsigned long sp; |
9d975ebd PE |
67 | unsigned short ss, gs; |
68 | ||
f39b6f0e | 69 | if (user_mode(regs)) { |
65ea5b03 | 70 | sp = regs->sp; |
99504819 | 71 | ss = regs->ss; |
d9a89a26 | 72 | gs = get_user_gs(regs); |
9d975ebd | 73 | } else { |
def3c5d0 | 74 | sp = kernel_stack_pointer(regs); |
9d975ebd PE |
75 | savesegment(ss, ss); |
76 | savesegment(gs, gs); | |
77 | } | |
1da177e4 | 78 | |
bb5e5ce5 JP |
79 | printk(KERN_DEFAULT "EIP: %pS\n", (void *)regs->ip); |
80 | printk(KERN_DEFAULT "EFLAGS: %08lx CPU: %d\n", regs->flags, | |
5d9070b1 | 81 | raw_smp_processor_id()); |
1da177e4 | 82 | |
d015a092 | 83 | printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
65ea5b03 | 84 | regs->ax, regs->bx, regs->cx, regs->dx); |
d015a092 | 85 | printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", |
65ea5b03 | 86 | regs->si, regs->di, regs->bp, sp); |
d015a092 | 87 | printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", |
92bc2056 | 88 | (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss); |
9d975ebd PE |
89 | |
90 | if (!all) | |
91 | return; | |
1da177e4 | 92 | |
4bb0d3ec ZA |
93 | cr0 = read_cr0(); |
94 | cr2 = read_cr2(); | |
6c690ee1 | 95 | cr3 = __read_cr3(); |
1ef55be1 | 96 | cr4 = __read_cr4(); |
d015a092 | 97 | printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", |
9d975ebd | 98 | cr0, cr2, cr3, cr4); |
bb1995d5 AS |
99 | |
100 | get_debugreg(d0, 0); | |
101 | get_debugreg(d1, 1); | |
102 | get_debugreg(d2, 2); | |
103 | get_debugreg(d3, 3); | |
bb1995d5 AS |
104 | get_debugreg(d6, 6); |
105 | get_debugreg(d7, 7); | |
4338774c DJ |
106 | |
107 | /* Only print out debug registers if they are in their non-default state. */ | |
108 | if ((d0 == 0) && (d1 == 0) && (d2 == 0) && (d3 == 0) && | |
109 | (d6 == DR6_RESERVED) && (d7 == 0x400)) | |
110 | return; | |
111 | ||
112 | printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n", | |
113 | d0, d1, d2, d3); | |
d015a092 | 114 | printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n", |
9d975ebd PE |
115 | d6, d7); |
116 | } | |
bb1995d5 | 117 | |
1da177e4 LT |
118 | void release_thread(struct task_struct *dead_task) |
119 | { | |
2684927c | 120 | BUG_ON(dead_task->mm); |
1da177e4 LT |
121 | release_vm86_irqs(dead_task); |
122 | } | |
123 | ||
c1bd55f9 JT |
124 | int copy_thread_tls(unsigned long clone_flags, unsigned long sp, |
125 | unsigned long arg, struct task_struct *p, unsigned long tls) | |
1da177e4 | 126 | { |
7076aada | 127 | struct pt_regs *childregs = task_pt_regs(p); |
0100301b BG |
128 | struct fork_frame *fork_frame = container_of(childregs, struct fork_frame, regs); |
129 | struct inactive_task_frame *frame = &fork_frame->frame; | |
1da177e4 LT |
130 | struct task_struct *tsk; |
131 | int err; | |
132 | ||
0100301b | 133 | frame->bp = 0; |
616d2483 | 134 | frame->ret_addr = (unsigned long) ret_from_fork; |
0100301b | 135 | p->thread.sp = (unsigned long) fork_frame; |
7076aada | 136 | p->thread.sp0 = (unsigned long) (childregs+1); |
6f46b3ae | 137 | memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps)); |
7076aada | 138 | |
1d4b4b29 | 139 | if (unlikely(p->flags & PF_KTHREAD)) { |
7076aada AV |
140 | /* kernel thread */ |
141 | memset(childregs, 0, sizeof(struct pt_regs)); | |
616d2483 BG |
142 | frame->bx = sp; /* function */ |
143 | frame->di = arg; | |
7076aada | 144 | p->thread.io_bitmap_ptr = NULL; |
7076aada AV |
145 | return 0; |
146 | } | |
616d2483 | 147 | frame->bx = 0; |
1d4b4b29 | 148 | *childregs = *current_pt_regs(); |
65ea5b03 | 149 | childregs->ax = 0; |
1d4b4b29 AV |
150 | if (sp) |
151 | childregs->sp = sp; | |
f48d9663 | 152 | |
1d4b4b29 | 153 | task_user_gs(p) = get_user_gs(current_pt_regs()); |
1da177e4 | 154 | |
66cb5917 | 155 | p->thread.io_bitmap_ptr = NULL; |
1da177e4 | 156 | tsk = current; |
66cb5917 | 157 | err = -ENOMEM; |
24f1e32c | 158 | |
b3cf2576 | 159 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
52978be6 AD |
160 | p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr, |
161 | IO_BITMAP_BYTES, GFP_KERNEL); | |
1da177e4 LT |
162 | if (!p->thread.io_bitmap_ptr) { |
163 | p->thread.io_bitmap_max = 0; | |
164 | return -ENOMEM; | |
165 | } | |
b3cf2576 | 166 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
1da177e4 LT |
167 | } |
168 | ||
efd1ca52 RM |
169 | err = 0; |
170 | ||
1da177e4 LT |
171 | /* |
172 | * Set a new TLS for the child thread? | |
173 | */ | |
efd1ca52 RM |
174 | if (clone_flags & CLONE_SETTLS) |
175 | err = do_set_thread_area(p, -1, | |
c1bd55f9 | 176 | (struct user_desc __user *)tls, 0); |
1da177e4 | 177 | |
1da177e4 LT |
178 | if (err && p->thread.io_bitmap_ptr) { |
179 | kfree(p->thread.io_bitmap_ptr); | |
180 | p->thread.io_bitmap_max = 0; | |
181 | } | |
182 | return err; | |
183 | } | |
184 | ||
513ad84b IM |
185 | void |
186 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
187 | { | |
d9a89a26 | 188 | set_user_gs(regs, 0); |
513ad84b | 189 | regs->fs = 0; |
513ad84b IM |
190 | regs->ds = __USER_DS; |
191 | regs->es = __USER_DS; | |
192 | regs->ss = __USER_DS; | |
193 | regs->cs = __USER_CS; | |
194 | regs->ip = new_ip; | |
195 | regs->sp = new_sp; | |
6783eaa2 | 196 | regs->flags = X86_EFLAGS_IF; |
1daeaa31 | 197 | force_iret(); |
513ad84b IM |
198 | } |
199 | EXPORT_SYMBOL_GPL(start_thread); | |
200 | ||
1da177e4 LT |
201 | |
202 | /* | |
ea70ef3d | 203 | * switch_to(x,y) should switch tasks from x to y. |
1da177e4 LT |
204 | * |
205 | * We fsave/fwait so that an exception goes off at the right time | |
206 | * (as a call from the fsave or fwait in effect) rather than to | |
207 | * the wrong process. Lazy FP saving no longer makes any sense | |
208 | * with modern CPU's, and this simplifies a lot of things (SMP | |
209 | * and UP become the same). | |
210 | * | |
211 | * NOTE! We used to use the x86 hardware context switching. The | |
212 | * reason for not using it any more becomes apparent when you | |
213 | * try to recover gracefully from saved state that is no longer | |
214 | * valid (stale segment register values in particular). With the | |
215 | * hardware task-switch, there is no way to fix up bad state in | |
216 | * a reasonable manner. | |
217 | * | |
218 | * The fact that Intel documents the hardware task-switching to | |
219 | * be slow is a fairly red herring - this code is not noticeably | |
220 | * faster. However, there _is_ some room for improvement here, | |
221 | * so the performance issues may eventually be a valid point. | |
222 | * More important, however, is the fact that this allows us much | |
223 | * more flexibility. | |
224 | * | |
65ea5b03 | 225 | * The return value (in %ax) will be the "prev" task after |
1da177e4 LT |
226 | * the task-switch, and shows up in ret_from_fork in entry.S, |
227 | * for example. | |
228 | */ | |
35ea7903 | 229 | __visible __notrace_funcgraph struct task_struct * |
8b96f011 | 230 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 LT |
231 | { |
232 | struct thread_struct *prev = &prev_p->thread, | |
384a23f9 IM |
233 | *next = &next_p->thread; |
234 | struct fpu *prev_fpu = &prev->fpu; | |
235 | struct fpu *next_fpu = &next->fpu; | |
1da177e4 | 236 | int cpu = smp_processor_id(); |
c482feef | 237 | struct tss_struct *tss = &per_cpu(cpu_tss_rw, cpu); |
1da177e4 LT |
238 | |
239 | /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */ | |
240 | ||
c474e507 | 241 | switch_fpu_prepare(prev_fpu, cpu); |
acc20761 | 242 | |
1da177e4 | 243 | /* |
464d1a78 | 244 | * Save away %gs. No need to save %fs, as it was saved on the |
f95d47ca JF |
245 | * stack on entry. No need to save %es and %ds, as those are |
246 | * always kernel segments while inside the kernel. Doing this | |
247 | * before setting the new TLS descriptors avoids the situation | |
248 | * where we temporarily have non-reloadable segments in %fs | |
249 | * and %gs. This could be an issue if the NMI handler ever | |
250 | * used %fs or %gs (it does not today), or if the kernel is | |
251 | * running inside of a hypervisor layer. | |
1da177e4 | 252 | */ |
ccbeed3a | 253 | lazy_save_gs(prev->gs); |
1da177e4 LT |
254 | |
255 | /* | |
e7a2ff59 | 256 | * Load the per-thread Thread-Local Storage descriptor. |
1da177e4 | 257 | */ |
e7a2ff59 | 258 | load_TLS(next, cpu); |
1da177e4 | 259 | |
8b151144 ZA |
260 | /* |
261 | * Restore IOPL if needed. In normal use, the flags restore | |
262 | * in the switch assembly will handle this. But if the kernel | |
263 | * is running virtualized at a non-zero CPL, the popf will | |
264 | * not restore flags, so it must be done in a separate step. | |
265 | */ | |
266 | if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl)) | |
267 | set_iopl_mask(next->iopl); | |
268 | ||
1da177e4 | 269 | /* |
b3cf2576 | 270 | * Now maybe handle debug registers and/or IO bitmaps |
1da177e4 | 271 | */ |
cf99abac AA |
272 | if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV || |
273 | task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT)) | |
274 | __switch_to_xtra(prev_p, next_p, tss); | |
ffaa8bd6 | 275 | |
9226d125 ZA |
276 | /* |
277 | * Leave lazy mode, flushing any hypercalls made here. | |
278 | * This must be done before restoring TLS segments so | |
279 | * the GDT and LDT are properly updated, and must be | |
3a0aee48 | 280 | * done before fpu__restore(), so the TS bit is up |
9226d125 ZA |
281 | * to date. |
282 | */ | |
224101ed | 283 | arch_end_context_switch(next_p); |
9226d125 | 284 | |
b27559a4 | 285 | /* |
fed7c3f0 | 286 | * Reload esp0 and cpu_current_top_of_stack. This changes |
bd7dc5a6 AL |
287 | * current_thread_info(). Refresh the SYSENTER configuration in |
288 | * case prev or next is vm86. | |
b27559a4 | 289 | */ |
46f5a10a | 290 | update_sp0(next_p); |
bd7dc5a6 | 291 | refresh_sysenter_cs(next); |
a7fcf28d AL |
292 | this_cpu_write(cpu_current_top_of_stack, |
293 | (unsigned long)task_stack_page(next_p) + | |
294 | THREAD_SIZE); | |
198d208d | 295 | |
9226d125 ZA |
296 | /* |
297 | * Restore %gs if needed (which is common) | |
298 | */ | |
299 | if (prev->gs | next->gs) | |
ccbeed3a | 300 | lazy_load_gs(next->gs); |
9226d125 | 301 | |
c474e507 | 302 | switch_fpu_finish(next_fpu, cpu); |
34ddc81a | 303 | |
c6ae41e7 | 304 | this_cpu_write(current_task, next_p); |
9226d125 | 305 | |
4f341a5e FY |
306 | /* Load the Intel cache allocation PQR MSR. */ |
307 | intel_rdt_sched_in(); | |
308 | ||
1da177e4 LT |
309 | return prev_p; |
310 | } | |
79170fda KH |
311 | |
312 | SYSCALL_DEFINE2(arch_prctl, int, option, unsigned long, arg2) | |
313 | { | |
314 | return do_arch_prctl_common(current, option, arg2); | |
315 | } |