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..
14 #include <linux/cpu.h>
15 #include <linux/errno.h>
16 #include <linux/sched.h>
18 #include <linux/kernel.h>
20 #include <linux/elfcore.h>
21 #include <linux/smp.h>
22 #include <linux/stddef.h>
23 #include <linux/slab.h>
24 #include <linux/vmalloc.h>
25 #include <linux/user.h>
26 #include <linux/a.out.h>
27 #include <linux/interrupt.h>
28 #include <linux/utsname.h>
29 #include <linux/delay.h>
30 #include <linux/reboot.h>
31 #include <linux/init.h>
32 #include <linux/mc146818rtc.h>
33 #include <linux/module.h>
34 #include <linux/kallsyms.h>
35 #include <linux/ptrace.h>
36 #include <linux/random.h>
37 #include <linux/personality.h>
38 #include <linux/tick.h>
39 #include <linux/percpu.h>
41 #include <asm/uaccess.h>
42 #include <asm/pgtable.h>
43 #include <asm/system.h>
46 #include <asm/processor.h>
50 #ifdef CONFIG_MATH_EMULATION
51 #include <asm/math_emu.h>
54 #include <linux/err.h>
56 #include <asm/tlbflush.h>
59 asmlinkage
void ret_from_fork(void) __asm__("ret_from_fork");
61 static int hlt_counter
;
63 unsigned long boot_option_idle_override
= 0;
64 EXPORT_SYMBOL(boot_option_idle_override
);
66 DEFINE_PER_CPU(struct task_struct
*, current_task
) = &init_task
;
67 EXPORT_PER_CPU_SYMBOL(current_task
);
69 DEFINE_PER_CPU(int, cpu_number
);
70 EXPORT_PER_CPU_SYMBOL(cpu_number
);
73 * Return saved PC of a blocked thread.
75 unsigned long thread_saved_pc(struct task_struct
*tsk
)
77 return ((unsigned long *)tsk
->thread
.esp
)[3];
81 * Powermanagement idle function, if any..
83 void (*pm_idle
)(void);
84 EXPORT_SYMBOL(pm_idle
);
85 static DEFINE_PER_CPU(unsigned int, cpu_idle_state
);
87 void disable_hlt(void)
92 EXPORT_SYMBOL(disable_hlt
);
99 EXPORT_SYMBOL(enable_hlt
);
102 * We use this if we don't have any better
105 void default_idle(void)
107 if (!hlt_counter
&& boot_cpu_data
.hlt_works_ok
) {
108 current_thread_info()->status
&= ~TS_POLLING
;
110 * TS_POLLING-cleared state must be visible before we
117 safe_halt(); /* enables interrupts racelessly */
120 current_thread_info()->status
|= TS_POLLING
;
122 /* loop is done by the caller */
126 #ifdef CONFIG_APM_MODULE
127 EXPORT_SYMBOL(default_idle
);
131 * On SMP it's slightly faster (but much more power-consuming!)
132 * to poll the ->work.need_resched flag instead of waiting for the
133 * cross-CPU IPI to arrive. Use this option with caution.
135 static void poll_idle (void)
140 #ifdef CONFIG_HOTPLUG_CPU
142 /* We don't actually take CPU down, just spin without interrupts. */
143 static inline void play_dead(void)
145 /* This must be done before dead CPU ack */
150 __get_cpu_var(cpu_state
) = CPU_DEAD
;
153 * With physical CPU hotplug, we should halt the cpu
160 static inline void play_dead(void)
164 #endif /* CONFIG_HOTPLUG_CPU */
167 * The idle thread. There's no useful work to be
168 * done, so just try to conserve power and have a
169 * low exit latency (ie sit in a loop waiting for
170 * somebody to say that they'd like to reschedule)
174 int cpu
= smp_processor_id();
176 current_thread_info()->status
|= TS_POLLING
;
178 /* endless idle loop with no priority at all */
180 tick_nohz_stop_sched_tick();
181 while (!need_resched()) {
184 if (__get_cpu_var(cpu_idle_state
))
185 __get_cpu_var(cpu_idle_state
) = 0;
194 if (cpu_is_offline(cpu
))
197 __get_cpu_var(irq_stat
).idle_timestamp
= jiffies
;
200 tick_nohz_restart_sched_tick();
201 preempt_enable_no_resched();
207 static void do_nothing(void *unused
)
211 void cpu_idle_wait(void)
213 unsigned int cpu
, this_cpu
= get_cpu();
214 cpumask_t map
, tmp
= current
->cpus_allowed
;
216 set_cpus_allowed(current
, cpumask_of_cpu(this_cpu
));
220 for_each_online_cpu(cpu
) {
221 per_cpu(cpu_idle_state
, cpu
) = 1;
225 __get_cpu_var(cpu_idle_state
) = 0;
230 for_each_online_cpu(cpu
) {
231 if (cpu_isset(cpu
, map
) && !per_cpu(cpu_idle_state
, cpu
))
234 cpus_and(map
, map
, cpu_online_map
);
236 * We waited 1 sec, if a CPU still did not call idle
237 * it may be because it is in idle and not waking up
238 * because it has nothing to do.
239 * Give all the remaining CPUS a kick.
241 smp_call_function_mask(map
, do_nothing
, 0, 0);
242 } while (!cpus_empty(map
));
244 set_cpus_allowed(current
, tmp
);
246 EXPORT_SYMBOL_GPL(cpu_idle_wait
);
249 * This uses new MONITOR/MWAIT instructions on P4 processors with PNI,
250 * which can obviate IPI to trigger checking of need_resched.
251 * We execute MONITOR against need_resched and enter optimized wait state
252 * through MWAIT. Whenever someone changes need_resched, we would be woken
253 * up from MWAIT (without an IPI).
255 * New with Core Duo processors, MWAIT can take some hints based on CPU
258 void mwait_idle_with_hints(unsigned long eax
, unsigned long ecx
)
260 if (!need_resched()) {
261 __monitor((void *)¤t_thread_info()->flags
, 0, 0);
268 /* Default MONITOR/MWAIT with no hints, used for default C1 state */
269 static void mwait_idle(void)
272 mwait_idle_with_hints(0, 0);
275 void __cpuinit
select_idle_routine(const struct cpuinfo_x86
*c
)
277 if (cpu_has(c
, X86_FEATURE_MWAIT
)) {
278 printk("monitor/mwait feature present.\n");
280 * Skip, if setup has overridden idle.
281 * One CPU supports mwait => All CPUs supports mwait
284 printk("using mwait in idle threads.\n");
285 pm_idle
= mwait_idle
;
290 static int __init
idle_setup(char *str
)
292 if (!strcmp(str
, "poll")) {
293 printk("using polling idle threads.\n");
295 #ifdef CONFIG_X86_SMP
296 if (smp_num_siblings
> 1)
297 printk("WARNING: polling idle and HT enabled, performance may degrade.\n");
299 } else if (!strcmp(str
, "mwait"))
304 boot_option_idle_override
= 1;
307 early_param("idle", idle_setup
);
309 void __show_registers(struct pt_regs
*regs
, int all
)
311 unsigned long cr0
= 0L, cr2
= 0L, cr3
= 0L, cr4
= 0L;
312 unsigned long d0
, d1
, d2
, d3
, d6
, d7
;
314 unsigned short ss
, gs
;
316 if (user_mode_vm(regs
)) {
318 ss
= regs
->xss
& 0xffff;
321 esp
= (unsigned long) (®s
->esp
);
327 printk("Pid: %d, comm: %s %s (%s %.*s)\n",
328 task_pid_nr(current
), current
->comm
,
329 print_tainted(), init_utsname()->release
,
330 (int)strcspn(init_utsname()->version
, " "),
331 init_utsname()->version
);
333 printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
334 0xffff & regs
->xcs
, regs
->eip
, regs
->eflags
,
336 print_symbol("EIP is at %s\n", regs
->eip
);
338 printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
339 regs
->eax
, regs
->ebx
, regs
->ecx
, regs
->edx
);
340 printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
341 regs
->esi
, regs
->edi
, regs
->ebp
, esp
);
342 printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
343 regs
->xds
& 0xffff, regs
->xes
& 0xffff,
344 regs
->xfs
& 0xffff, gs
, ss
);
352 cr4
= read_cr4_safe();
353 printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
360 printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
365 printk("DR6: %08lx DR7: %08lx\n",
369 void show_regs(struct pt_regs
*regs
)
371 __show_registers(regs
, 1);
372 show_trace(NULL
, regs
, ®s
->esp
);
376 * This gets run with %ebx containing the
377 * function to call, and %edx containing
380 extern void kernel_thread_helper(void);
383 * Create a kernel thread
385 int kernel_thread(int (*fn
)(void *), void * arg
, unsigned long flags
)
389 memset(®s
, 0, sizeof(regs
));
391 regs
.ebx
= (unsigned long) fn
;
392 regs
.edx
= (unsigned long) arg
;
394 regs
.xds
= __USER_DS
;
395 regs
.xes
= __USER_DS
;
396 regs
.xfs
= __KERNEL_PERCPU
;
398 regs
.eip
= (unsigned long) kernel_thread_helper
;
399 regs
.xcs
= __KERNEL_CS
| get_kernel_rpl();
400 regs
.eflags
= X86_EFLAGS_IF
| X86_EFLAGS_SF
| X86_EFLAGS_PF
| 0x2;
402 /* Ok, create the new process.. */
403 return do_fork(flags
| CLONE_VM
| CLONE_UNTRACED
, 0, ®s
, 0, NULL
, NULL
);
405 EXPORT_SYMBOL(kernel_thread
);
408 * Free current thread data structures etc..
410 void exit_thread(void)
412 /* The process may have allocated an io port bitmap... nuke it. */
413 if (unlikely(test_thread_flag(TIF_IO_BITMAP
))) {
414 struct task_struct
*tsk
= current
;
415 struct thread_struct
*t
= &tsk
->thread
;
417 struct tss_struct
*tss
= &per_cpu(init_tss
, cpu
);
419 kfree(t
->io_bitmap_ptr
);
420 t
->io_bitmap_ptr
= NULL
;
421 clear_thread_flag(TIF_IO_BITMAP
);
423 * Careful, clear this in the TSS too:
425 memset(tss
->io_bitmap
, 0xff, tss
->io_bitmap_max
);
426 t
->io_bitmap_max
= 0;
427 tss
->io_bitmap_owner
= NULL
;
428 tss
->io_bitmap_max
= 0;
429 tss
->x86_tss
.io_bitmap_base
= INVALID_IO_BITMAP_OFFSET
;
434 void flush_thread(void)
436 struct task_struct
*tsk
= current
;
438 memset(tsk
->thread
.debugreg
, 0, sizeof(unsigned long)*8);
439 memset(tsk
->thread
.tls_array
, 0, sizeof(tsk
->thread
.tls_array
));
440 clear_tsk_thread_flag(tsk
, TIF_DEBUG
);
442 * Forget coprocessor state..
448 void release_thread(struct task_struct
*dead_task
)
450 BUG_ON(dead_task
->mm
);
451 release_vm86_irqs(dead_task
);
455 * This gets called before we allocate a new thread and copy
456 * the current task into it.
458 void prepare_to_copy(struct task_struct
*tsk
)
463 int copy_thread(int nr
, unsigned long clone_flags
, unsigned long esp
,
464 unsigned long unused
,
465 struct task_struct
* p
, struct pt_regs
* regs
)
467 struct pt_regs
* childregs
;
468 struct task_struct
*tsk
;
471 childregs
= task_pt_regs(p
);
474 childregs
->esp
= esp
;
476 p
->thread
.esp
= (unsigned long) childregs
;
477 p
->thread
.esp0
= (unsigned long) (childregs
+1);
479 p
->thread
.eip
= (unsigned long) ret_from_fork
;
481 savesegment(gs
,p
->thread
.gs
);
484 if (unlikely(test_tsk_thread_flag(tsk
, TIF_IO_BITMAP
))) {
485 p
->thread
.io_bitmap_ptr
= kmemdup(tsk
->thread
.io_bitmap_ptr
,
486 IO_BITMAP_BYTES
, GFP_KERNEL
);
487 if (!p
->thread
.io_bitmap_ptr
) {
488 p
->thread
.io_bitmap_max
= 0;
491 set_tsk_thread_flag(p
, TIF_IO_BITMAP
);
495 * Set a new TLS for the child thread?
497 if (clone_flags
& CLONE_SETTLS
) {
498 struct desc_struct
*desc
;
499 struct user_desc info
;
503 if (copy_from_user(&info
, (void __user
*)childregs
->esi
, sizeof(info
)))
506 if (LDT_empty(&info
))
509 idx
= info
.entry_number
;
510 if (idx
< GDT_ENTRY_TLS_MIN
|| idx
> GDT_ENTRY_TLS_MAX
)
513 desc
= p
->thread
.tls_array
+ idx
- GDT_ENTRY_TLS_MIN
;
514 desc
->a
= LDT_entry_a(&info
);
515 desc
->b
= LDT_entry_b(&info
);
520 if (err
&& p
->thread
.io_bitmap_ptr
) {
521 kfree(p
->thread
.io_bitmap_ptr
);
522 p
->thread
.io_bitmap_max
= 0;
528 * fill in the user structure for a core dump..
530 void dump_thread(struct pt_regs
* regs
, struct user
* dump
)
534 /* changed the size calculations - should hopefully work better. lbt */
535 dump
->magic
= CMAGIC
;
536 dump
->start_code
= 0;
537 dump
->start_stack
= regs
->esp
& ~(PAGE_SIZE
- 1);
538 dump
->u_tsize
= ((unsigned long) current
->mm
->end_code
) >> PAGE_SHIFT
;
539 dump
->u_dsize
= ((unsigned long) (current
->mm
->brk
+ (PAGE_SIZE
-1))) >> PAGE_SHIFT
;
540 dump
->u_dsize
-= dump
->u_tsize
;
542 for (i
= 0; i
< 8; i
++)
543 dump
->u_debugreg
[i
] = current
->thread
.debugreg
[i
];
545 if (dump
->start_stack
< TASK_SIZE
)
546 dump
->u_ssize
= ((unsigned long) (TASK_SIZE
- dump
->start_stack
)) >> PAGE_SHIFT
;
548 dump
->regs
.ebx
= regs
->ebx
;
549 dump
->regs
.ecx
= regs
->ecx
;
550 dump
->regs
.edx
= regs
->edx
;
551 dump
->regs
.esi
= regs
->esi
;
552 dump
->regs
.edi
= regs
->edi
;
553 dump
->regs
.ebp
= regs
->ebp
;
554 dump
->regs
.eax
= regs
->eax
;
555 dump
->regs
.ds
= regs
->xds
;
556 dump
->regs
.es
= regs
->xes
;
557 dump
->regs
.fs
= regs
->xfs
;
558 savesegment(gs
,dump
->regs
.gs
);
559 dump
->regs
.orig_eax
= regs
->orig_eax
;
560 dump
->regs
.eip
= regs
->eip
;
561 dump
->regs
.cs
= regs
->xcs
;
562 dump
->regs
.eflags
= regs
->eflags
;
563 dump
->regs
.esp
= regs
->esp
;
564 dump
->regs
.ss
= regs
->xss
;
566 dump
->u_fpvalid
= dump_fpu (regs
, &dump
->i387
);
568 EXPORT_SYMBOL(dump_thread
);
571 * Capture the user space registers if the task is not running (in user space)
573 int dump_task_regs(struct task_struct
*tsk
, elf_gregset_t
*regs
)
575 struct pt_regs ptregs
= *task_pt_regs(tsk
);
576 ptregs
.xcs
&= 0xffff;
577 ptregs
.xds
&= 0xffff;
578 ptregs
.xes
&= 0xffff;
579 ptregs
.xss
&= 0xffff;
581 elf_core_copy_regs(regs
, &ptregs
);
586 #ifdef CONFIG_SECCOMP
587 void hard_disable_TSC(void)
589 write_cr4(read_cr4() | X86_CR4_TSD
);
591 void disable_TSC(void)
594 if (!test_and_set_thread_flag(TIF_NOTSC
))
596 * Must flip the CPU state synchronously with
597 * TIF_NOTSC in the current running context.
602 void hard_enable_TSC(void)
604 write_cr4(read_cr4() & ~X86_CR4_TSD
);
606 #endif /* CONFIG_SECCOMP */
609 __switch_to_xtra(struct task_struct
*prev_p
, struct task_struct
*next_p
,
610 struct tss_struct
*tss
)
612 struct thread_struct
*next
;
614 next
= &next_p
->thread
;
616 if (test_tsk_thread_flag(next_p
, TIF_DEBUG
)) {
617 set_debugreg(next
->debugreg
[0], 0);
618 set_debugreg(next
->debugreg
[1], 1);
619 set_debugreg(next
->debugreg
[2], 2);
620 set_debugreg(next
->debugreg
[3], 3);
622 set_debugreg(next
->debugreg
[6], 6);
623 set_debugreg(next
->debugreg
[7], 7);
626 #ifdef CONFIG_SECCOMP
627 if (test_tsk_thread_flag(prev_p
, TIF_NOTSC
) ^
628 test_tsk_thread_flag(next_p
, TIF_NOTSC
)) {
629 /* prev and next are different */
630 if (test_tsk_thread_flag(next_p
, TIF_NOTSC
))
637 if (!test_tsk_thread_flag(next_p
, TIF_IO_BITMAP
)) {
639 * Disable the bitmap via an invalid offset. We still cache
640 * the previous bitmap owner and the IO bitmap contents:
642 tss
->x86_tss
.io_bitmap_base
= INVALID_IO_BITMAP_OFFSET
;
646 if (likely(next
== tss
->io_bitmap_owner
)) {
648 * Previous owner of the bitmap (hence the bitmap content)
649 * matches the next task, we dont have to do anything but
650 * to set a valid offset in the TSS:
652 tss
->x86_tss
.io_bitmap_base
= IO_BITMAP_OFFSET
;
656 * Lazy TSS's I/O bitmap copy. We set an invalid offset here
657 * and we let the task to get a GPF in case an I/O instruction
658 * is performed. The handler of the GPF will verify that the
659 * faulting task has a valid I/O bitmap and, it true, does the
660 * real copy and restart the instruction. This will save us
661 * redundant copies when the currently switched task does not
662 * perform any I/O during its timeslice.
664 tss
->x86_tss
.io_bitmap_base
= INVALID_IO_BITMAP_OFFSET_LAZY
;
668 * switch_to(x,yn) should switch tasks from x to y.
670 * We fsave/fwait so that an exception goes off at the right time
671 * (as a call from the fsave or fwait in effect) rather than to
672 * the wrong process. Lazy FP saving no longer makes any sense
673 * with modern CPU's, and this simplifies a lot of things (SMP
674 * and UP become the same).
676 * NOTE! We used to use the x86 hardware context switching. The
677 * reason for not using it any more becomes apparent when you
678 * try to recover gracefully from saved state that is no longer
679 * valid (stale segment register values in particular). With the
680 * hardware task-switch, there is no way to fix up bad state in
681 * a reasonable manner.
683 * The fact that Intel documents the hardware task-switching to
684 * be slow is a fairly red herring - this code is not noticeably
685 * faster. However, there _is_ some room for improvement here,
686 * so the performance issues may eventually be a valid point.
687 * More important, however, is the fact that this allows us much
690 * The return value (in %eax) will be the "prev" task after
691 * the task-switch, and shows up in ret_from_fork in entry.S,
694 struct task_struct fastcall
* __switch_to(struct task_struct
*prev_p
, struct task_struct
*next_p
)
696 struct thread_struct
*prev
= &prev_p
->thread
,
697 *next
= &next_p
->thread
;
698 int cpu
= smp_processor_id();
699 struct tss_struct
*tss
= &per_cpu(init_tss
, cpu
);
701 /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
703 __unlazy_fpu(prev_p
);
706 /* we're going to use this soon, after a few expensive things */
707 if (next_p
->fpu_counter
> 5)
708 prefetch(&next
->i387
.fxsave
);
713 load_esp0(tss
, next
);
716 * Save away %gs. No need to save %fs, as it was saved on the
717 * stack on entry. No need to save %es and %ds, as those are
718 * always kernel segments while inside the kernel. Doing this
719 * before setting the new TLS descriptors avoids the situation
720 * where we temporarily have non-reloadable segments in %fs
721 * and %gs. This could be an issue if the NMI handler ever
722 * used %fs or %gs (it does not today), or if the kernel is
723 * running inside of a hypervisor layer.
725 savesegment(gs
, prev
->gs
);
728 * Load the per-thread Thread-Local Storage descriptor.
733 * Restore IOPL if needed. In normal use, the flags restore
734 * in the switch assembly will handle this. But if the kernel
735 * is running virtualized at a non-zero CPL, the popf will
736 * not restore flags, so it must be done in a separate step.
738 if (get_kernel_rpl() && unlikely(prev
->iopl
!= next
->iopl
))
739 set_iopl_mask(next
->iopl
);
742 * Now maybe handle debug registers and/or IO bitmaps
744 if (unlikely(task_thread_info(prev_p
)->flags
& _TIF_WORK_CTXSW_PREV
||
745 task_thread_info(next_p
)->flags
& _TIF_WORK_CTXSW_NEXT
))
746 __switch_to_xtra(prev_p
, next_p
, tss
);
749 * Leave lazy mode, flushing any hypercalls made here.
750 * This must be done before restoring TLS segments so
751 * the GDT and LDT are properly updated, and must be
752 * done before math_state_restore, so the TS bit is up
755 arch_leave_lazy_cpu_mode();
757 /* If the task has used fpu the last 5 timeslices, just do a full
758 * restore of the math state immediately to avoid the trap; the
759 * chances of needing FPU soon are obviously high now
761 if (next_p
->fpu_counter
> 5)
762 math_state_restore();
765 * Restore %gs if needed (which is common)
767 if (prev
->gs
| next
->gs
)
768 loadsegment(gs
, next
->gs
);
770 x86_write_percpu(current_task
, next_p
);
775 asmlinkage
int sys_fork(struct pt_regs regs
)
777 return do_fork(SIGCHLD
, regs
.esp
, ®s
, 0, NULL
, NULL
);
780 asmlinkage
int sys_clone(struct pt_regs regs
)
782 unsigned long clone_flags
;
784 int __user
*parent_tidptr
, *child_tidptr
;
786 clone_flags
= regs
.ebx
;
788 parent_tidptr
= (int __user
*)regs
.edx
;
789 child_tidptr
= (int __user
*)regs
.edi
;
792 return do_fork(clone_flags
, newsp
, ®s
, 0, parent_tidptr
, child_tidptr
);
796 * This is trivial, and on the face of it looks like it
797 * could equally well be done in user mode.
799 * Not so, for quite unobvious reasons - register pressure.
800 * In user mode vfork() cannot have a stack frame, and if
801 * done by calling the "clone()" system call directly, you
802 * do not have enough call-clobbered registers to hold all
803 * the information you need.
805 asmlinkage
int sys_vfork(struct pt_regs regs
)
807 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, regs
.esp
, ®s
, 0, NULL
, NULL
);
811 * sys_execve() executes a new program.
813 asmlinkage
int sys_execve(struct pt_regs regs
)
818 filename
= getname((char __user
*) regs
.ebx
);
819 error
= PTR_ERR(filename
);
820 if (IS_ERR(filename
))
822 error
= do_execve(filename
,
823 (char __user
* __user
*) regs
.ecx
,
824 (char __user
* __user
*) regs
.edx
,
828 current
->ptrace
&= ~PT_DTRACE
;
829 task_unlock(current
);
830 /* Make sure we don't return using sysenter.. */
831 set_thread_flag(TIF_IRET
);
838 #define top_esp (THREAD_SIZE - sizeof(unsigned long))
839 #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
841 unsigned long get_wchan(struct task_struct
*p
)
843 unsigned long ebp
, esp
, eip
;
844 unsigned long stack_page
;
846 if (!p
|| p
== current
|| p
->state
== TASK_RUNNING
)
848 stack_page
= (unsigned long)task_stack_page(p
);
850 if (!stack_page
|| esp
< stack_page
|| esp
> top_esp
+stack_page
)
852 /* include/asm-i386/system.h:switch_to() pushes ebp last. */
853 ebp
= *(unsigned long *) esp
;
855 if (ebp
< stack_page
|| ebp
> top_ebp
+stack_page
)
857 eip
= *(unsigned long *) (ebp
+4);
858 if (!in_sched_functions(eip
))
860 ebp
= *(unsigned long *) ebp
;
861 } while (count
++ < 16);
866 * sys_alloc_thread_area: get a yet unused TLS descriptor index.
868 static int get_free_idx(void)
870 struct thread_struct
*t
= ¤t
->thread
;
873 for (idx
= 0; idx
< GDT_ENTRY_TLS_ENTRIES
; idx
++)
874 if (desc_empty(t
->tls_array
+ idx
))
875 return idx
+ GDT_ENTRY_TLS_MIN
;
880 * Set a given TLS descriptor:
882 asmlinkage
int sys_set_thread_area(struct user_desc __user
*u_info
)
884 struct thread_struct
*t
= ¤t
->thread
;
885 struct user_desc info
;
886 struct desc_struct
*desc
;
889 if (copy_from_user(&info
, u_info
, sizeof(info
)))
891 idx
= info
.entry_number
;
894 * index -1 means the kernel should try to find and
895 * allocate an empty descriptor:
898 idx
= get_free_idx();
901 if (put_user(idx
, &u_info
->entry_number
))
905 if (idx
< GDT_ENTRY_TLS_MIN
|| idx
> GDT_ENTRY_TLS_MAX
)
908 desc
= t
->tls_array
+ idx
- GDT_ENTRY_TLS_MIN
;
911 * We must not get preempted while modifying the TLS.
915 if (LDT_empty(&info
)) {
919 desc
->a
= LDT_entry_a(&info
);
920 desc
->b
= LDT_entry_b(&info
);
930 * Get the current Thread-Local Storage area:
933 #define GET_BASE(desc) ( \
934 (((desc)->a >> 16) & 0x0000ffff) | \
935 (((desc)->b << 16) & 0x00ff0000) | \
936 ( (desc)->b & 0xff000000) )
938 #define GET_LIMIT(desc) ( \
939 ((desc)->a & 0x0ffff) | \
940 ((desc)->b & 0xf0000) )
942 #define GET_32BIT(desc) (((desc)->b >> 22) & 1)
943 #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3)
944 #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1)
945 #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1)
946 #define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
947 #define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
949 asmlinkage
int sys_get_thread_area(struct user_desc __user
*u_info
)
951 struct user_desc info
;
952 struct desc_struct
*desc
;
955 if (get_user(idx
, &u_info
->entry_number
))
957 if (idx
< GDT_ENTRY_TLS_MIN
|| idx
> GDT_ENTRY_TLS_MAX
)
960 memset(&info
, 0, sizeof(info
));
962 desc
= current
->thread
.tls_array
+ idx
- GDT_ENTRY_TLS_MIN
;
964 info
.entry_number
= idx
;
965 info
.base_addr
= GET_BASE(desc
);
966 info
.limit
= GET_LIMIT(desc
);
967 info
.seg_32bit
= GET_32BIT(desc
);
968 info
.contents
= GET_CONTENTS(desc
);
969 info
.read_exec_only
= !GET_WRITABLE(desc
);
970 info
.limit_in_pages
= GET_LIMIT_PAGES(desc
);
971 info
.seg_not_present
= !GET_PRESENT(desc
);
972 info
.useable
= GET_USEABLE(desc
);
974 if (copy_to_user(u_info
, &info
, sizeof(info
)))
979 unsigned long arch_align_stack(unsigned long sp
)
981 if (!(current
->personality
& ADDR_NO_RANDOMIZE
) && randomize_va_space
)
982 sp
-= get_random_int() % 8192;