2 * Copyright (C) 1995 Linus Torvalds
4 * Pentium III FXSR, SSE support
5 * Gareth Hughes <gareth@valinux.com>, May 2000
9 * This file handles the architecture-dependent parts of process handling..
12 #include <linux/stackprotector.h>
13 #include <linux/cpu.h>
14 #include <linux/errno.h>
15 #include <linux/sched.h>
17 #include <linux/kernel.h>
19 #include <linux/elfcore.h>
20 #include <linux/smp.h>
21 #include <linux/stddef.h>
22 #include <linux/slab.h>
23 #include <linux/vmalloc.h>
24 #include <linux/user.h>
25 #include <linux/interrupt.h>
26 #include <linux/delay.h>
27 #include <linux/reboot.h>
28 #include <linux/init.h>
29 #include <linux/mc146818rtc.h>
30 #include <linux/module.h>
31 #include <linux/kallsyms.h>
32 #include <linux/ptrace.h>
33 #include <linux/personality.h>
34 #include <linux/tick.h>
35 #include <linux/percpu.h>
36 #include <linux/prctl.h>
37 #include <linux/ftrace.h>
38 #include <linux/uaccess.h>
40 #include <linux/kdebug.h>
41 #include <linux/cpuidle.h>
43 #include <asm/pgtable.h>
45 #include <asm/processor.h>
47 #include <asm/fpu-internal.h>
49 #ifdef CONFIG_MATH_EMULATION
50 #include <asm/math_emu.h>
53 #include <linux/err.h>
55 #include <asm/tlbflush.h>
58 #include <asm/syscalls.h>
59 #include <asm/debugreg.h>
61 #include <asm/switch_to.h>
63 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
66 * Return saved PC of a blocked thread.
68 unsigned long thread_saved_pc(struct task_struct *tsk)
70 return ((unsigned long *)tsk->thread.sp)[3];
74 static inline void play_dead(void)
81 * The idle thread. There's no useful work to be
82 * done, so just try to conserve power and have a
83 * low exit latency (ie sit in a loop waiting for
84 * somebody to say that they'd like to reschedule)
88 int cpu = smp_processor_id();
91 * If we're the non-boot CPU, nothing set the stack canary up
92 * for us. CPU0 already has it initialized but no harm in
93 * doing it again. This is a good place for updating it, as
94 * we wont ever return from this function (so the invalid
95 * canaries already on the stack wont ever trigger).
97 boot_init_stack_canary();
99 current_thread_info()->status |= TS_POLLING;
101 /* endless idle loop with no priority at all */
103 tick_nohz_idle_enter();
105 while (!need_resched()) {
110 if (cpu_is_offline(cpu))
115 /* Don't trace irqs off for idle */
116 stop_critical_timings();
117 if (cpuidle_idle_call())
119 start_critical_timings();
122 tick_nohz_idle_exit();
123 schedule_preempt_disabled();
127 void __show_regs(struct pt_regs *regs, int all)
129 unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
130 unsigned long d0, d1, d2, d3, d6, d7;
132 unsigned short ss, gs;
134 if (user_mode_vm(regs)) {
136 ss = regs->ss & 0xffff;
137 gs = get_user_gs(regs);
139 sp = kernel_stack_pointer(regs);
146 printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
147 (u16)regs->cs, regs->ip, regs->flags,
149 print_symbol("EIP is at %s\n", regs->ip);
151 printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
152 regs->ax, regs->bx, regs->cx, regs->dx);
153 printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
154 regs->si, regs->di, regs->bp, sp);
155 printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
156 (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
164 cr4 = read_cr4_safe();
165 printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
172 printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
177 printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
181 void release_thread(struct task_struct *dead_task)
183 BUG_ON(dead_task->mm);
184 release_vm86_irqs(dead_task);
188 * This gets called before we allocate a new thread and copy
189 * the current task into it.
191 void prepare_to_copy(struct task_struct *tsk)
196 int copy_thread(unsigned long clone_flags, unsigned long sp,
197 unsigned long unused,
198 struct task_struct *p, struct pt_regs *regs)
200 struct pt_regs *childregs;
201 struct task_struct *tsk;
204 childregs = task_pt_regs(p);
209 p->thread.sp = (unsigned long) childregs;
210 p->thread.sp0 = (unsigned long) (childregs+1);
212 p->thread.ip = (unsigned long) ret_from_fork;
214 task_user_gs(p) = get_user_gs(regs);
217 p->thread.io_bitmap_ptr = NULL;
221 memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
223 if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
224 p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
225 IO_BITMAP_BYTES, GFP_KERNEL);
226 if (!p->thread.io_bitmap_ptr) {
227 p->thread.io_bitmap_max = 0;
230 set_tsk_thread_flag(p, TIF_IO_BITMAP);
236 * Set a new TLS for the child thread?
238 if (clone_flags & CLONE_SETTLS)
239 err = do_set_thread_area(p, -1,
240 (struct user_desc __user *)childregs->si, 0);
242 if (err && p->thread.io_bitmap_ptr) {
243 kfree(p->thread.io_bitmap_ptr);
244 p->thread.io_bitmap_max = 0;
250 start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
252 set_user_gs(regs, 0);
254 regs->ds = __USER_DS;
255 regs->es = __USER_DS;
256 regs->ss = __USER_DS;
257 regs->cs = __USER_CS;
261 * Free the old FP and other extended state
263 free_thread_xstate(current);
265 EXPORT_SYMBOL_GPL(start_thread);
269 * switch_to(x,y) should switch tasks from x to y.
271 * We fsave/fwait so that an exception goes off at the right time
272 * (as a call from the fsave or fwait in effect) rather than to
273 * the wrong process. Lazy FP saving no longer makes any sense
274 * with modern CPU's, and this simplifies a lot of things (SMP
275 * and UP become the same).
277 * NOTE! We used to use the x86 hardware context switching. The
278 * reason for not using it any more becomes apparent when you
279 * try to recover gracefully from saved state that is no longer
280 * valid (stale segment register values in particular). With the
281 * hardware task-switch, there is no way to fix up bad state in
282 * a reasonable manner.
284 * The fact that Intel documents the hardware task-switching to
285 * be slow is a fairly red herring - this code is not noticeably
286 * faster. However, there _is_ some room for improvement here,
287 * so the performance issues may eventually be a valid point.
288 * More important, however, is the fact that this allows us much
291 * The return value (in %ax) will be the "prev" task after
292 * the task-switch, and shows up in ret_from_fork in entry.S,
295 __notrace_funcgraph struct task_struct *
296 __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
298 struct thread_struct *prev = &prev_p->thread,
299 *next = &next_p->thread;
300 int cpu = smp_processor_id();
301 struct tss_struct *tss = &per_cpu(init_tss, cpu);
304 /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
306 fpu = switch_fpu_prepare(prev_p, next_p, cpu);
314 * Save away %gs. No need to save %fs, as it was saved on the
315 * stack on entry. No need to save %es and %ds, as those are
316 * always kernel segments while inside the kernel. Doing this
317 * before setting the new TLS descriptors avoids the situation
318 * where we temporarily have non-reloadable segments in %fs
319 * and %gs. This could be an issue if the NMI handler ever
320 * used %fs or %gs (it does not today), or if the kernel is
321 * running inside of a hypervisor layer.
323 lazy_save_gs(prev->gs);
326 * Load the per-thread Thread-Local Storage descriptor.
331 * Restore IOPL if needed. In normal use, the flags restore
332 * in the switch assembly will handle this. But if the kernel
333 * is running virtualized at a non-zero CPL, the popf will
334 * not restore flags, so it must be done in a separate step.
336 if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
337 set_iopl_mask(next->iopl);
340 * Now maybe handle debug registers and/or IO bitmaps
342 if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
343 task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
344 __switch_to_xtra(prev_p, next_p, tss);
347 * Leave lazy mode, flushing any hypercalls made here.
348 * This must be done before restoring TLS segments so
349 * the GDT and LDT are properly updated, and must be
350 * done before math_state_restore, so the TS bit is up
353 arch_end_context_switch(next_p);
356 * Restore %gs if needed (which is common)
358 if (prev->gs | next->gs)
359 lazy_load_gs(next->gs);
361 switch_fpu_finish(next_p, fpu);
363 percpu_write(current_task, next_p);
368 #define top_esp (THREAD_SIZE - sizeof(unsigned long))
369 #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
371 unsigned long get_wchan(struct task_struct *p)
373 unsigned long bp, sp, ip;
374 unsigned long stack_page;
376 if (!p || p == current || p->state == TASK_RUNNING)
378 stack_page = (unsigned long)task_stack_page(p);
380 if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
382 /* include/asm-i386/system.h:switch_to() pushes bp last. */
383 bp = *(unsigned long *) sp;
385 if (bp < stack_page || bp > top_ebp+stack_page)
387 ip = *(unsigned long *) (bp+4);
388 if (!in_sched_functions(ip))
390 bp = *(unsigned long *) bp;
391 } while (count++ < 16);