blob: 6f1236c29c4b575ec429dcc34d23629321ca5c0a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * 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
Zwane Mwaikambof3705132005-06-25 14:54:50 -070012#include <linux/cpu.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/errno.h>
14#include <linux/sched.h>
15#include <linux/fs.h>
16#include <linux/kernel.h>
17#include <linux/mm.h>
18#include <linux/elfcore.h>
19#include <linux/smp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/stddef.h>
21#include <linux/slab.h>
22#include <linux/vmalloc.h>
23#include <linux/user.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/interrupt.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/delay.h>
26#include <linux/reboot.h>
27#include <linux/init.h>
28#include <linux/mc146818rtc.h>
29#include <linux/module.h>
30#include <linux/kallsyms.h>
31#include <linux/ptrace.h>
Andi Kleenc16b63e02006-09-26 10:52:28 +020032#include <linux/personality.h>
Jeremy Fitzhardinge7c3576d2007-05-02 19:27:16 +020033#include <linux/percpu.h>
Erik Bosman529e25f2008-04-14 00:24:18 +020034#include <linux/prctl.h>
Frederic Weisbecker8b96f012008-12-06 03:40:00 +010035#include <linux/ftrace.h>
Jaswinder Singh Rajputbefa9e72009-01-04 16:18:56 +053036#include <linux/uaccess.h>
37#include <linux/io.h>
38#include <linux/kdebug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#include <asm/pgtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <asm/ldt.h>
42#include <asm/processor.h>
43#include <asm/i387.h>
Linus Torvalds1361b832012-02-21 13:19:22 -080044#include <asm/fpu-internal.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <asm/desc.h>
46#ifdef CONFIG_MATH_EMULATION
47#include <asm/math_emu.h>
48#endif
49
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/err.h>
51
Zwane Mwaikambof3705132005-06-25 14:54:50 -070052#include <asm/tlbflush.h>
53#include <asm/cpu.h>
Marc Dionne1eda8142008-09-23 22:40:02 -040054#include <asm/idle.h>
Jaswinder Singhbbc1f692008-07-21 21:34:13 +053055#include <asm/syscalls.h>
K.Prasad66cb5912009-06-01 23:44:55 +053056#include <asm/debugreg.h>
David Howellsf05e7982012-03-28 18:11:12 +010057#include <asm/switch_to.h>
Zwane Mwaikambof3705132005-06-25 14:54:50 -070058
Linus Torvalds1da177e2005-04-16 15:20:36 -070059asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
Al Viro7076aad2012-09-10 16:44:54 -040060asmlinkage void ret_from_kernel_thread(void) __asm__("ret_from_kernel_thread");
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Linus Torvalds1da177e2005-04-16 15:20:36 -070062/*
63 * Return saved PC of a blocked thread.
64 */
65unsigned long thread_saved_pc(struct task_struct *tsk)
66{
H. Peter Anvinfaca6222008-01-30 13:31:02 +010067 return ((unsigned long *)tsk->thread.sp)[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -070068}
69
Pekka Enberge2ce07c2008-04-03 16:40:48 +030070void __show_regs(struct pt_regs *regs, int all)
Linus Torvalds1da177e2005-04-16 15:20:36 -070071{
72 unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
Alan Sternbb1995d2007-07-21 17:10:42 +020073 unsigned long d0, d1, d2, d3, d6, d7;
H. Peter Anvin65ea5b02008-01-30 13:30:56 +010074 unsigned long sp;
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +020075 unsigned short ss, gs;
76
77 if (user_mode_vm(regs)) {
H. Peter Anvin65ea5b02008-01-30 13:30:56 +010078 sp = regs->sp;
79 ss = regs->ss & 0xffff;
Tejun Heod9a89a22009-02-09 22:17:40 +090080 gs = get_user_gs(regs);
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +020081 } else {
H. Peter Anvindef3c5d2009-10-12 14:09:07 -070082 sp = kernel_stack_pointer(regs);
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +020083 savesegment(ss, ss);
84 savesegment(gs, gs);
85 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
Pekka Enbergd015a092009-12-28 10:26:59 +020087 printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
Harvey Harrison92bc2052008-02-08 12:09:56 -080088 (u16)regs->cs, regs->ip, regs->flags,
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +020089 smp_processor_id());
H. Peter Anvin65ea5b02008-01-30 13:30:56 +010090 print_symbol("EIP is at %s\n", regs->ip);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
Pekka Enbergd015a092009-12-28 10:26:59 +020092 printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
H. Peter Anvin65ea5b02008-01-30 13:30:56 +010093 regs->ax, regs->bx, regs->cx, regs->dx);
Pekka Enbergd015a092009-12-28 10:26:59 +020094 printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
H. Peter Anvin65ea5b02008-01-30 13:30:56 +010095 regs->si, regs->di, regs->bp, sp);
Pekka Enbergd015a092009-12-28 10:26:59 +020096 printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
Harvey Harrison92bc2052008-02-08 12:09:56 -080097 (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +020098
99 if (!all)
100 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
Zachary Amsden4bb0d3e2005-09-03 15:56:36 -0700102 cr0 = read_cr0();
103 cr2 = read_cr2();
104 cr3 = read_cr3();
Zachary Amsdenff6e8c02006-01-06 00:11:50 -0800105 cr4 = read_cr4_safe();
Pekka Enbergd015a092009-12-28 10:26:59 +0200106 printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +0200107 cr0, cr2, cr3, cr4);
Alan Sternbb1995d2007-07-21 17:10:42 +0200108
109 get_debugreg(d0, 0);
110 get_debugreg(d1, 1);
111 get_debugreg(d2, 2);
112 get_debugreg(d3, 3);
Alan Sternbb1995d2007-07-21 17:10:42 +0200113 get_debugreg(d6, 6);
114 get_debugreg(d7, 7);
Dave Jones4338774c2013-06-18 12:09:11 -0400115
116 /* Only print out debug registers if they are in their non-default state. */
117 if ((d0 == 0) && (d1 == 0) && (d2 == 0) && (d3 == 0) &&
118 (d6 == DR6_RESERVED) && (d7 == 0x400))
119 return;
120
121 printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
122 d0, d1, d2, d3);
Pekka Enbergd015a092009-12-28 10:26:59 +0200123 printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
Pavel Emelyanov9d975eb2007-10-19 20:35:03 +0200124 d6, d7);
125}
Alan Sternbb1995d2007-07-21 17:10:42 +0200126
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127void release_thread(struct task_struct *dead_task)
128{
Zachary Amsden26849272006-01-06 00:11:59 -0800129 BUG_ON(dead_task->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 release_vm86_irqs(dead_task);
131}
132
Alexey Dobriyan6f2c55b2009-04-02 16:56:59 -0700133int copy_thread(unsigned long clone_flags, unsigned long sp,
Al Viroafa86fc2012-10-22 22:51:14 -0400134 unsigned long arg, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135{
Al Viro7076aad2012-09-10 16:44:54 -0400136 struct pt_regs *childregs = task_pt_regs(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 struct task_struct *tsk;
138 int err;
139
H. Peter Anvinfaca6222008-01-30 13:31:02 +0100140 p->thread.sp = (unsigned long) childregs;
141 p->thread.sp0 = (unsigned long) (childregs+1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142
Al Viro1d4b4b22012-10-22 22:34:11 -0400143 if (unlikely(p->flags & PF_KTHREAD)) {
Al Viro7076aad2012-09-10 16:44:54 -0400144 /* kernel thread */
145 memset(childregs, 0, sizeof(struct pt_regs));
146 p->thread.ip = (unsigned long) ret_from_kernel_thread;
147 task_user_gs(p) = __KERNEL_STACK_CANARY;
148 childregs->ds = __USER_DS;
149 childregs->es = __USER_DS;
150 childregs->fs = __KERNEL_PERCPU;
151 childregs->bx = sp; /* function */
152 childregs->bp = arg;
153 childregs->orig_ax = -1;
154 childregs->cs = __KERNEL_CS | get_kernel_rpl();
H. Peter Anvin1adfa762013-04-27 16:10:11 -0700155 childregs->flags = X86_EFLAGS_IF | X86_EFLAGS_FIXED;
Vineet Guptac375f152013-11-12 15:08:46 -0800156 p->thread.fpu_counter = 0;
Al Viro7076aad2012-09-10 16:44:54 -0400157 p->thread.io_bitmap_ptr = NULL;
158 memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
159 return 0;
160 }
Al Viro1d4b4b22012-10-22 22:34:11 -0400161 *childregs = *current_pt_regs();
Al Viro7076aad2012-09-10 16:44:54 -0400162 childregs->ax = 0;
Al Viro1d4b4b22012-10-22 22:34:11 -0400163 if (sp)
164 childregs->sp = sp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
Al Viro7076aad2012-09-10 16:44:54 -0400166 p->thread.ip = (unsigned long) ret_from_fork;
Al Viro1d4b4b22012-10-22 22:34:11 -0400167 task_user_gs(p) = get_user_gs(current_pt_regs());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168
Vineet Guptac375f152013-11-12 15:08:46 -0800169 p->thread.fpu_counter = 0;
K.Prasad66cb5912009-06-01 23:44:55 +0530170 p->thread.io_bitmap_ptr = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 tsk = current;
K.Prasad66cb5912009-06-01 23:44:55 +0530172 err = -ENOMEM;
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +0200173
174 memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
K.Prasad66cb5912009-06-01 23:44:55 +0530175
Stephane Eranianb3cf2572006-07-09 21:12:39 -0400176 if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
Alexey Dobriyan52978be2006-09-30 23:27:21 -0700177 p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
178 IO_BITMAP_BYTES, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 if (!p->thread.io_bitmap_ptr) {
180 p->thread.io_bitmap_max = 0;
181 return -ENOMEM;
182 }
Stephane Eranianb3cf2572006-07-09 21:12:39 -0400183 set_tsk_thread_flag(p, TIF_IO_BITMAP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 }
185
Roland McGrathefd1ca52008-01-30 13:30:46 +0100186 err = 0;
187
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 /*
189 * Set a new TLS for the child thread?
190 */
Roland McGrathefd1ca52008-01-30 13:30:46 +0100191 if (clone_flags & CLONE_SETTLS)
192 err = do_set_thread_area(p, -1,
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100193 (struct user_desc __user *)childregs->si, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 if (err && p->thread.io_bitmap_ptr) {
196 kfree(p->thread.io_bitmap_ptr);
197 p->thread.io_bitmap_max = 0;
198 }
199 return err;
200}
201
Ingo Molnar513ad842008-02-21 05:18:40 +0100202void
203start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
204{
Tejun Heod9a89a22009-02-09 22:17:40 +0900205 set_user_gs(regs, 0);
Ingo Molnar513ad842008-02-21 05:18:40 +0100206 regs->fs = 0;
Ingo Molnar513ad842008-02-21 05:18:40 +0100207 regs->ds = __USER_DS;
208 regs->es = __USER_DS;
209 regs->ss = __USER_DS;
210 regs->cs = __USER_CS;
211 regs->ip = new_ip;
212 regs->sp = new_sp;
Al Viro6783eaa22012-08-02 23:05:11 +0400213 regs->flags = X86_EFLAGS_IF;
Suresh Siddhaaa283f42008-03-10 15:28:05 -0700214 /*
Al Viroe76623d2012-08-02 22:12:06 +0400215 * force it to the iret return path by making it look as if there was
216 * some work pending.
217 */
218 set_thread_flag(TIF_NOTIFY_RESUME);
Ingo Molnar513ad842008-02-21 05:18:40 +0100219}
220EXPORT_SYMBOL_GPL(start_thread);
221
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
223/*
Kamalesh Babulalea70ef32011-04-28 14:32:08 +0530224 * switch_to(x,y) should switch tasks from x to y.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 *
226 * We fsave/fwait so that an exception goes off at the right time
227 * (as a call from the fsave or fwait in effect) rather than to
228 * the wrong process. Lazy FP saving no longer makes any sense
229 * with modern CPU's, and this simplifies a lot of things (SMP
230 * and UP become the same).
231 *
232 * NOTE! We used to use the x86 hardware context switching. The
233 * reason for not using it any more becomes apparent when you
234 * try to recover gracefully from saved state that is no longer
235 * valid (stale segment register values in particular). With the
236 * hardware task-switch, there is no way to fix up bad state in
237 * a reasonable manner.
238 *
239 * The fact that Intel documents the hardware task-switching to
240 * be slow is a fairly red herring - this code is not noticeably
241 * faster. However, there _is_ some room for improvement here,
242 * so the performance issues may eventually be a valid point.
243 * More important, however, is the fact that this allows us much
244 * more flexibility.
245 *
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100246 * The return value (in %ax) will be the "prev" task after
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 * the task-switch, and shows up in ret_from_fork in entry.S,
248 * for example.
249 */
Andi Kleen35ea79032013-08-05 15:02:39 -0700250__visible __notrace_funcgraph struct task_struct *
Frederic Weisbecker8b96f012008-12-06 03:40:00 +0100251__switch_to(struct task_struct *prev_p, struct task_struct *next_p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252{
253 struct thread_struct *prev = &prev_p->thread,
254 *next = &next_p->thread;
255 int cpu = smp_processor_id();
256 struct tss_struct *tss = &per_cpu(init_tss, cpu);
Linus Torvalds34ddc812012-02-18 12:56:35 -0800257 fpu_switch_t fpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259 /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
260
Linus Torvalds7e168382012-02-19 13:27:00 -0800261 fpu = switch_fpu_prepare(prev_p, next_p, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
263 /*
Zachary Amsdene7a2ff52005-09-03 15:56:39 -0700264 * Reload esp0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 */
H. Peter Anvinfaca6222008-01-30 13:31:02 +0100266 load_sp0(tss, next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
268 /*
Jeremy Fitzhardinge464d1a72007-02-13 13:26:20 +0100269 * Save away %gs. No need to save %fs, as it was saved on the
Jeremy Fitzhardingef95d47c2006-12-07 02:14:02 +0100270 * stack on entry. No need to save %es and %ds, as those are
271 * always kernel segments while inside the kernel. Doing this
272 * before setting the new TLS descriptors avoids the situation
273 * where we temporarily have non-reloadable segments in %fs
274 * and %gs. This could be an issue if the NMI handler ever
275 * used %fs or %gs (it does not today), or if the kernel is
276 * running inside of a hypervisor layer.
Zachary Amsdene7a2ff52005-09-03 15:56:39 -0700277 */
Tejun Heoccbeed32009-02-09 22:17:40 +0900278 lazy_save_gs(prev->gs);
Zachary Amsdene7a2ff52005-09-03 15:56:39 -0700279
280 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 * Load the per-thread Thread-Local Storage descriptor.
282 */
283 load_TLS(next, cpu);
284
285 /*
Zachary Amsden8b151142007-02-13 13:26:21 +0100286 * Restore IOPL if needed. In normal use, the flags restore
287 * in the switch assembly will handle this. But if the kernel
288 * is running virtualized at a non-zero CPL, the popf will
289 * not restore flags, so it must be done in a separate step.
290 */
291 if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
292 set_iopl_mask(next->iopl);
293
294 /*
Peter Zijlstrac2daa3b2013-08-14 14:51:00 +0200295 * If it were not for PREEMPT_ACTIVE we could guarantee that the
296 * preempt_count of all tasks was equal here and this would not be
297 * needed.
298 */
299 task_thread_info(prev_p)->saved_preempt_count = this_cpu_read(__preempt_count);
300 this_cpu_write(__preempt_count, task_thread_info(next_p)->saved_preempt_count);
301
302 /*
Stephane Eranianb3cf2572006-07-09 21:12:39 -0400303 * Now maybe handle debug registers and/or IO bitmaps
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 */
Andrea Arcangelicf99aba2007-07-15 23:41:33 -0700305 if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
306 task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
307 __switch_to_xtra(prev_p, next_p, tss);
Andrea Arcangeliffaa8bd2005-06-27 14:36:36 -0700308
Zachary Amsden9226d122007-02-13 13:26:21 +0100309 /*
310 * Leave lazy mode, flushing any hypercalls made here.
311 * This must be done before restoring TLS segments so
312 * the GDT and LDT are properly updated, and must be
313 * done before math_state_restore, so the TS bit is up
314 * to date.
315 */
Jeremy Fitzhardinge224101e2009-02-18 11:18:57 -0800316 arch_end_context_switch(next_p);
Zachary Amsden9226d122007-02-13 13:26:21 +0100317
Zachary Amsden9226d122007-02-13 13:26:21 +0100318 /*
319 * Restore %gs if needed (which is common)
320 */
321 if (prev->gs | next->gs)
Tejun Heoccbeed32009-02-09 22:17:40 +0900322 lazy_load_gs(next->gs);
Zachary Amsden9226d122007-02-13 13:26:21 +0100323
Linus Torvalds34ddc812012-02-18 12:56:35 -0800324 switch_fpu_finish(next_p, fpu);
325
Alex Shic6ae41e2012-05-11 15:35:27 +0800326 this_cpu_write(current_task, next_p);
Zachary Amsden9226d122007-02-13 13:26:21 +0100327
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 return prev_p;
329}
330
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331#define top_esp (THREAD_SIZE - sizeof(unsigned long))
332#define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
333
334unsigned long get_wchan(struct task_struct *p)
335{
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100336 unsigned long bp, sp, ip;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 unsigned long stack_page;
338 int count = 0;
339 if (!p || p == current || p->state == TASK_RUNNING)
340 return 0;
Al Viro65e0fdf2006-01-12 01:05:41 -0800341 stack_page = (unsigned long)task_stack_page(p);
H. Peter Anvinfaca6222008-01-30 13:31:02 +0100342 sp = p->thread.sp;
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100343 if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 return 0;
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100345 /* include/asm-i386/system.h:switch_to() pushes bp last. */
346 bp = *(unsigned long *) sp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 do {
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100348 if (bp < stack_page || bp > top_ebp+stack_page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 return 0;
H. Peter Anvin65ea5b02008-01-30 13:30:56 +0100350 ip = *(unsigned long *) (bp+4);
351 if (!in_sched_functions(ip))
352 return ip;
353 bp = *(unsigned long *) bp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 } while (count++ < 16);
355 return 0;
356}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357