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>
58 #include <asm/kdebug.h>
60 asmlinkage
void ret_from_fork(void) __asm__("ret_from_fork");
62 static int hlt_counter
;
64 unsigned long boot_option_idle_override
= 0;
65 EXPORT_SYMBOL(boot_option_idle_override
);
67 DEFINE_PER_CPU(struct task_struct
*, current_task
) = &init_task
;
68 EXPORT_PER_CPU_SYMBOL(current_task
);
70 DEFINE_PER_CPU(int, cpu_number
);
71 EXPORT_PER_CPU_SYMBOL(cpu_number
);
74 * Return saved PC of a blocked thread.
76 unsigned long thread_saved_pc(struct task_struct
*tsk
)
78 return ((unsigned long *)tsk
->thread
.esp
)[3];
82 * Powermanagement idle function, if any..
84 void (*pm_idle
)(void);
85 EXPORT_SYMBOL(pm_idle
);
86 static DEFINE_PER_CPU(unsigned int, cpu_idle_state
);
88 void disable_hlt(void)
93 EXPORT_SYMBOL(disable_hlt
);
100 EXPORT_SYMBOL(enable_hlt
);
103 * We use this if we don't have any better
106 void default_idle(void)
108 if (!hlt_counter
&& boot_cpu_data
.hlt_works_ok
) {
109 current_thread_info()->status
&= ~TS_POLLING
;
111 * TS_POLLING-cleared state must be visible before we
117 if (!need_resched()) {
122 t0n
= ktime_to_ns(t0
);
123 safe_halt(); /* enables interrupts racelessly */
126 t1n
= ktime_to_ns(t1
);
127 sched_clock_idle_wakeup_event(t1n
- t0n
);
130 current_thread_info()->status
|= TS_POLLING
;
132 /* loop is done by the caller */
136 #ifdef CONFIG_APM_MODULE
137 EXPORT_SYMBOL(default_idle
);
141 * On SMP it's slightly faster (but much more power-consuming!)
142 * to poll the ->work.need_resched flag instead of waiting for the
143 * cross-CPU IPI to arrive. Use this option with caution.
145 static void poll_idle (void)
150 #ifdef CONFIG_HOTPLUG_CPU
152 /* We don't actually take CPU down, just spin without interrupts. */
153 static inline void play_dead(void)
155 /* This must be done before dead CPU ack */
160 __get_cpu_var(cpu_state
) = CPU_DEAD
;
163 * With physical CPU hotplug, we should halt the cpu
170 static inline void play_dead(void)
174 #endif /* CONFIG_HOTPLUG_CPU */
177 * The idle thread. There's no useful work to be
178 * done, so just try to conserve power and have a
179 * low exit latency (ie sit in a loop waiting for
180 * somebody to say that they'd like to reschedule)
184 int cpu
= smp_processor_id();
186 current_thread_info()->status
|= TS_POLLING
;
188 /* endless idle loop with no priority at all */
190 tick_nohz_stop_sched_tick();
191 while (!need_resched()) {
194 if (__get_cpu_var(cpu_idle_state
))
195 __get_cpu_var(cpu_idle_state
) = 0;
204 if (cpu_is_offline(cpu
))
207 __get_cpu_var(irq_stat
).idle_timestamp
= jiffies
;
210 tick_nohz_restart_sched_tick();
211 preempt_enable_no_resched();
217 static void do_nothing(void *unused
)
221 void cpu_idle_wait(void)
223 unsigned int cpu
, this_cpu
= get_cpu();
224 cpumask_t map
, tmp
= current
->cpus_allowed
;
226 set_cpus_allowed(current
, cpumask_of_cpu(this_cpu
));
230 for_each_online_cpu(cpu
) {
231 per_cpu(cpu_idle_state
, cpu
) = 1;
235 __get_cpu_var(cpu_idle_state
) = 0;
240 for_each_online_cpu(cpu
) {
241 if (cpu_isset(cpu
, map
) && !per_cpu(cpu_idle_state
, cpu
))
244 cpus_and(map
, map
, cpu_online_map
);
246 * We waited 1 sec, if a CPU still did not call idle
247 * it may be because it is in idle and not waking up
248 * because it has nothing to do.
249 * Give all the remaining CPUS a kick.
251 smp_call_function_mask(map
, do_nothing
, 0, 0);
252 } while (!cpus_empty(map
));
254 set_cpus_allowed(current
, tmp
);
256 EXPORT_SYMBOL_GPL(cpu_idle_wait
);
259 * This uses new MONITOR/MWAIT instructions on P4 processors with PNI,
260 * which can obviate IPI to trigger checking of need_resched.
261 * We execute MONITOR against need_resched and enter optimized wait state
262 * through MWAIT. Whenever someone changes need_resched, we would be woken
263 * up from MWAIT (without an IPI).
265 * New with Core Duo processors, MWAIT can take some hints based on CPU
268 void mwait_idle_with_hints(unsigned long eax
, unsigned long ecx
)
270 if (!need_resched()) {
271 __monitor((void *)¤t_thread_info()->flags
, 0, 0);
278 /* Default MONITOR/MWAIT with no hints, used for default C1 state */
279 static void mwait_idle(void)
282 mwait_idle_with_hints(0, 0);
285 void __cpuinit
select_idle_routine(const struct cpuinfo_x86
*c
)
287 if (cpu_has(c
, X86_FEATURE_MWAIT
)) {
288 printk("monitor/mwait feature present.\n");
290 * Skip, if setup has overridden idle.
291 * One CPU supports mwait => All CPUs supports mwait
294 printk("using mwait in idle threads.\n");
295 pm_idle
= mwait_idle
;
300 static int __init
idle_setup(char *str
)
302 if (!strcmp(str
, "poll")) {
303 printk("using polling idle threads.\n");
305 #ifdef CONFIG_X86_SMP
306 if (smp_num_siblings
> 1)
307 printk("WARNING: polling idle and HT enabled, performance may degrade.\n");
309 } else if (!strcmp(str
, "mwait"))
314 boot_option_idle_override
= 1;
317 early_param("idle", idle_setup
);
319 void __show_registers(struct pt_regs
*regs
, int all
)
321 unsigned long cr0
= 0L, cr2
= 0L, cr3
= 0L, cr4
= 0L;
322 unsigned long d0
, d1
, d2
, d3
, d6
, d7
;
324 unsigned short ss
, gs
;
326 if (user_mode_vm(regs
)) {
328 ss
= regs
->xss
& 0xffff;
331 esp
= (unsigned long) (®s
->esp
);
337 printk("Pid: %d, comm: %s %s (%s %.*s)\n",
338 task_pid_nr(current
), current
->comm
,
339 print_tainted(), init_utsname()->release
,
340 (int)strcspn(init_utsname()->version
, " "),
341 init_utsname()->version
);
343 printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
344 0xffff & regs
->xcs
, regs
->eip
, regs
->eflags
,
346 print_symbol("EIP is at %s\n", regs
->eip
);
348 printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
349 regs
->eax
, regs
->ebx
, regs
->ecx
, regs
->edx
);
350 printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
351 regs
->esi
, regs
->edi
, regs
->ebp
, esp
);
352 printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
353 regs
->xds
& 0xffff, regs
->xes
& 0xffff,
354 regs
->xfs
& 0xffff, gs
, ss
);
362 cr4
= read_cr4_safe();
363 printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
370 printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
375 printk("DR6: %08lx DR7: %08lx\n",
379 void show_regs(struct pt_regs
*regs
)
381 __show_registers(regs
, 1);
382 show_trace(NULL
, regs
, ®s
->esp
);
386 * This gets run with %ebx containing the
387 * function to call, and %edx containing
390 extern void kernel_thread_helper(void);
393 * Create a kernel thread
395 int kernel_thread(int (*fn
)(void *), void * arg
, unsigned long flags
)
399 memset(®s
, 0, sizeof(regs
));
401 regs
.ebx
= (unsigned long) fn
;
402 regs
.edx
= (unsigned long) arg
;
404 regs
.xds
= __USER_DS
;
405 regs
.xes
= __USER_DS
;
406 regs
.xfs
= __KERNEL_PERCPU
;
408 regs
.eip
= (unsigned long) kernel_thread_helper
;
409 regs
.xcs
= __KERNEL_CS
| get_kernel_rpl();
410 regs
.eflags
= X86_EFLAGS_IF
| X86_EFLAGS_SF
| X86_EFLAGS_PF
| 0x2;
412 /* Ok, create the new process.. */
413 return do_fork(flags
| CLONE_VM
| CLONE_UNTRACED
, 0, ®s
, 0, NULL
, NULL
);
415 EXPORT_SYMBOL(kernel_thread
);
418 * Free current thread data structures etc..
420 void exit_thread(void)
422 /* The process may have allocated an io port bitmap... nuke it. */
423 if (unlikely(test_thread_flag(TIF_IO_BITMAP
))) {
424 struct task_struct
*tsk
= current
;
425 struct thread_struct
*t
= &tsk
->thread
;
427 struct tss_struct
*tss
= &per_cpu(init_tss
, cpu
);
429 kfree(t
->io_bitmap_ptr
);
430 t
->io_bitmap_ptr
= NULL
;
431 clear_thread_flag(TIF_IO_BITMAP
);
433 * Careful, clear this in the TSS too:
435 memset(tss
->io_bitmap
, 0xff, tss
->io_bitmap_max
);
436 t
->io_bitmap_max
= 0;
437 tss
->io_bitmap_owner
= NULL
;
438 tss
->io_bitmap_max
= 0;
439 tss
->x86_tss
.io_bitmap_base
= INVALID_IO_BITMAP_OFFSET
;
444 void flush_thread(void)
446 struct task_struct
*tsk
= current
;
448 memset(tsk
->thread
.debugreg
, 0, sizeof(unsigned long)*8);
449 memset(tsk
->thread
.tls_array
, 0, sizeof(tsk
->thread
.tls_array
));
450 clear_tsk_thread_flag(tsk
, TIF_DEBUG
);
452 * Forget coprocessor state..
458 void release_thread(struct task_struct
*dead_task
)
460 BUG_ON(dead_task
->mm
);
461 release_vm86_irqs(dead_task
);
465 * This gets called before we allocate a new thread and copy
466 * the current task into it.
468 void prepare_to_copy(struct task_struct
*tsk
)
473 int copy_thread(int nr
, unsigned long clone_flags
, unsigned long esp
,
474 unsigned long unused
,
475 struct task_struct
* p
, struct pt_regs
* regs
)
477 struct pt_regs
* childregs
;
478 struct task_struct
*tsk
;
481 childregs
= task_pt_regs(p
);
484 childregs
->esp
= esp
;
486 p
->thread
.esp
= (unsigned long) childregs
;
487 p
->thread
.esp0
= (unsigned long) (childregs
+1);
489 p
->thread
.eip
= (unsigned long) ret_from_fork
;
491 savesegment(gs
,p
->thread
.gs
);
494 if (unlikely(test_tsk_thread_flag(tsk
, TIF_IO_BITMAP
))) {
495 p
->thread
.io_bitmap_ptr
= kmemdup(tsk
->thread
.io_bitmap_ptr
,
496 IO_BITMAP_BYTES
, GFP_KERNEL
);
497 if (!p
->thread
.io_bitmap_ptr
) {
498 p
->thread
.io_bitmap_max
= 0;
501 set_tsk_thread_flag(p
, TIF_IO_BITMAP
);
505 * Set a new TLS for the child thread?
507 if (clone_flags
& CLONE_SETTLS
) {
508 struct desc_struct
*desc
;
509 struct user_desc info
;
513 if (copy_from_user(&info
, (void __user
*)childregs
->esi
, sizeof(info
)))
516 if (LDT_empty(&info
))
519 idx
= info
.entry_number
;
520 if (idx
< GDT_ENTRY_TLS_MIN
|| idx
> GDT_ENTRY_TLS_MAX
)
523 desc
= p
->thread
.tls_array
+ idx
- GDT_ENTRY_TLS_MIN
;
524 desc
->a
= LDT_entry_a(&info
);
525 desc
->b
= LDT_entry_b(&info
);
530 if (err
&& p
->thread
.io_bitmap_ptr
) {
531 kfree(p
->thread
.io_bitmap_ptr
);
532 p
->thread
.io_bitmap_max
= 0;
538 * fill in the user structure for a core dump..
540 void dump_thread(struct pt_regs
* regs
, struct user
* dump
)
544 /* changed the size calculations - should hopefully work better. lbt */
545 dump
->magic
= CMAGIC
;
546 dump
->start_code
= 0;
547 dump
->start_stack
= regs
->esp
& ~(PAGE_SIZE
- 1);
548 dump
->u_tsize
= ((unsigned long) current
->mm
->end_code
) >> PAGE_SHIFT
;
549 dump
->u_dsize
= ((unsigned long) (current
->mm
->brk
+ (PAGE_SIZE
-1))) >> PAGE_SHIFT
;
550 dump
->u_dsize
-= dump
->u_tsize
;
552 for (i
= 0; i
< 8; i
++)
553 dump
->u_debugreg
[i
] = current
->thread
.debugreg
[i
];
555 if (dump
->start_stack
< TASK_SIZE
)
556 dump
->u_ssize
= ((unsigned long) (TASK_SIZE
- dump
->start_stack
)) >> PAGE_SHIFT
;
558 dump
->regs
.ebx
= regs
->ebx
;
559 dump
->regs
.ecx
= regs
->ecx
;
560 dump
->regs
.edx
= regs
->edx
;
561 dump
->regs
.esi
= regs
->esi
;
562 dump
->regs
.edi
= regs
->edi
;
563 dump
->regs
.ebp
= regs
->ebp
;
564 dump
->regs
.eax
= regs
->eax
;
565 dump
->regs
.ds
= regs
->xds
;
566 dump
->regs
.es
= regs
->xes
;
567 dump
->regs
.fs
= regs
->xfs
;
568 savesegment(gs
,dump
->regs
.gs
);
569 dump
->regs
.orig_eax
= regs
->orig_eax
;
570 dump
->regs
.eip
= regs
->eip
;
571 dump
->regs
.cs
= regs
->xcs
;
572 dump
->regs
.eflags
= regs
->eflags
;
573 dump
->regs
.esp
= regs
->esp
;
574 dump
->regs
.ss
= regs
->xss
;
576 dump
->u_fpvalid
= dump_fpu (regs
, &dump
->i387
);
578 EXPORT_SYMBOL(dump_thread
);
581 * Capture the user space registers if the task is not running (in user space)
583 int dump_task_regs(struct task_struct
*tsk
, elf_gregset_t
*regs
)
585 struct pt_regs ptregs
= *task_pt_regs(tsk
);
586 ptregs
.xcs
&= 0xffff;
587 ptregs
.xds
&= 0xffff;
588 ptregs
.xes
&= 0xffff;
589 ptregs
.xss
&= 0xffff;
591 elf_core_copy_regs(regs
, &ptregs
);
596 #ifdef CONFIG_SECCOMP
597 void hard_disable_TSC(void)
599 write_cr4(read_cr4() | X86_CR4_TSD
);
601 void disable_TSC(void)
604 if (!test_and_set_thread_flag(TIF_NOTSC
))
606 * Must flip the CPU state synchronously with
607 * TIF_NOTSC in the current running context.
612 void hard_enable_TSC(void)
614 write_cr4(read_cr4() & ~X86_CR4_TSD
);
616 #endif /* CONFIG_SECCOMP */
619 __switch_to_xtra(struct task_struct
*prev_p
, struct task_struct
*next_p
,
620 struct tss_struct
*tss
)
622 struct thread_struct
*next
;
624 next
= &next_p
->thread
;
626 if (test_tsk_thread_flag(next_p
, TIF_DEBUG
)) {
627 set_debugreg(next
->debugreg
[0], 0);
628 set_debugreg(next
->debugreg
[1], 1);
629 set_debugreg(next
->debugreg
[2], 2);
630 set_debugreg(next
->debugreg
[3], 3);
632 set_debugreg(next
->debugreg
[6], 6);
633 set_debugreg(next
->debugreg
[7], 7);
636 #ifdef CONFIG_SECCOMP
637 if (test_tsk_thread_flag(prev_p
, TIF_NOTSC
) ^
638 test_tsk_thread_flag(next_p
, TIF_NOTSC
)) {
639 /* prev and next are different */
640 if (test_tsk_thread_flag(next_p
, TIF_NOTSC
))
647 if (!test_tsk_thread_flag(next_p
, TIF_IO_BITMAP
)) {
649 * Disable the bitmap via an invalid offset. We still cache
650 * the previous bitmap owner and the IO bitmap contents:
652 tss
->x86_tss
.io_bitmap_base
= INVALID_IO_BITMAP_OFFSET
;
656 if (likely(next
== tss
->io_bitmap_owner
)) {
658 * Previous owner of the bitmap (hence the bitmap content)
659 * matches the next task, we dont have to do anything but
660 * to set a valid offset in the TSS:
662 tss
->x86_tss
.io_bitmap_base
= IO_BITMAP_OFFSET
;
666 * Lazy TSS's I/O bitmap copy. We set an invalid offset here
667 * and we let the task to get a GPF in case an I/O instruction
668 * is performed. The handler of the GPF will verify that the
669 * faulting task has a valid I/O bitmap and, it true, does the
670 * real copy and restart the instruction. This will save us
671 * redundant copies when the currently switched task does not
672 * perform any I/O during its timeslice.
674 tss
->x86_tss
.io_bitmap_base
= INVALID_IO_BITMAP_OFFSET_LAZY
;
678 * switch_to(x,yn) should switch tasks from x to y.
680 * We fsave/fwait so that an exception goes off at the right time
681 * (as a call from the fsave or fwait in effect) rather than to
682 * the wrong process. Lazy FP saving no longer makes any sense
683 * with modern CPU's, and this simplifies a lot of things (SMP
684 * and UP become the same).
686 * NOTE! We used to use the x86 hardware context switching. The
687 * reason for not using it any more becomes apparent when you
688 * try to recover gracefully from saved state that is no longer
689 * valid (stale segment register values in particular). With the
690 * hardware task-switch, there is no way to fix up bad state in
691 * a reasonable manner.
693 * The fact that Intel documents the hardware task-switching to
694 * be slow is a fairly red herring - this code is not noticeably
695 * faster. However, there _is_ some room for improvement here,
696 * so the performance issues may eventually be a valid point.
697 * More important, however, is the fact that this allows us much
700 * The return value (in %eax) will be the "prev" task after
701 * the task-switch, and shows up in ret_from_fork in entry.S,
704 struct task_struct fastcall
* __switch_to(struct task_struct
*prev_p
, struct task_struct
*next_p
)
706 struct thread_struct
*prev
= &prev_p
->thread
,
707 *next
= &next_p
->thread
;
708 int cpu
= smp_processor_id();
709 struct tss_struct
*tss
= &per_cpu(init_tss
, cpu
);
711 /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
713 __unlazy_fpu(prev_p
);
716 /* we're going to use this soon, after a few expensive things */
717 if (next_p
->fpu_counter
> 5)
718 prefetch(&next
->i387
.fxsave
);
723 load_esp0(tss
, next
);
726 * Save away %gs. No need to save %fs, as it was saved on the
727 * stack on entry. No need to save %es and %ds, as those are
728 * always kernel segments while inside the kernel. Doing this
729 * before setting the new TLS descriptors avoids the situation
730 * where we temporarily have non-reloadable segments in %fs
731 * and %gs. This could be an issue if the NMI handler ever
732 * used %fs or %gs (it does not today), or if the kernel is
733 * running inside of a hypervisor layer.
735 savesegment(gs
, prev
->gs
);
738 * Load the per-thread Thread-Local Storage descriptor.
743 * Restore IOPL if needed. In normal use, the flags restore
744 * in the switch assembly will handle this. But if the kernel
745 * is running virtualized at a non-zero CPL, the popf will
746 * not restore flags, so it must be done in a separate step.
748 if (get_kernel_rpl() && unlikely(prev
->iopl
!= next
->iopl
))
749 set_iopl_mask(next
->iopl
);
752 * Now maybe handle debug registers and/or IO bitmaps
754 if (unlikely(task_thread_info(prev_p
)->flags
& _TIF_WORK_CTXSW_PREV
||
755 task_thread_info(next_p
)->flags
& _TIF_WORK_CTXSW_NEXT
))
756 __switch_to_xtra(prev_p
, next_p
, tss
);
759 * Leave lazy mode, flushing any hypercalls made here.
760 * This must be done before restoring TLS segments so
761 * the GDT and LDT are properly updated, and must be
762 * done before math_state_restore, so the TS bit is up
765 arch_leave_lazy_cpu_mode();
767 /* If the task has used fpu the last 5 timeslices, just do a full
768 * restore of the math state immediately to avoid the trap; the
769 * chances of needing FPU soon are obviously high now
771 if (next_p
->fpu_counter
> 5)
772 math_state_restore();
775 * Restore %gs if needed (which is common)
777 if (prev
->gs
| next
->gs
)
778 loadsegment(gs
, next
->gs
);
780 x86_write_percpu(current_task
, next_p
);
785 asmlinkage
int sys_fork(struct pt_regs regs
)
787 return do_fork(SIGCHLD
, regs
.esp
, ®s
, 0, NULL
, NULL
);
790 asmlinkage
int sys_clone(struct pt_regs regs
)
792 unsigned long clone_flags
;
794 int __user
*parent_tidptr
, *child_tidptr
;
796 clone_flags
= regs
.ebx
;
798 parent_tidptr
= (int __user
*)regs
.edx
;
799 child_tidptr
= (int __user
*)regs
.edi
;
802 return do_fork(clone_flags
, newsp
, ®s
, 0, parent_tidptr
, child_tidptr
);
806 * This is trivial, and on the face of it looks like it
807 * could equally well be done in user mode.
809 * Not so, for quite unobvious reasons - register pressure.
810 * In user mode vfork() cannot have a stack frame, and if
811 * done by calling the "clone()" system call directly, you
812 * do not have enough call-clobbered registers to hold all
813 * the information you need.
815 asmlinkage
int sys_vfork(struct pt_regs regs
)
817 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, regs
.esp
, ®s
, 0, NULL
, NULL
);
821 * sys_execve() executes a new program.
823 asmlinkage
int sys_execve(struct pt_regs regs
)
828 filename
= getname((char __user
*) regs
.ebx
);
829 error
= PTR_ERR(filename
);
830 if (IS_ERR(filename
))
832 error
= do_execve(filename
,
833 (char __user
* __user
*) regs
.ecx
,
834 (char __user
* __user
*) regs
.edx
,
838 current
->ptrace
&= ~PT_DTRACE
;
839 task_unlock(current
);
840 /* Make sure we don't return using sysenter.. */
841 set_thread_flag(TIF_IRET
);
848 #define top_esp (THREAD_SIZE - sizeof(unsigned long))
849 #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
851 unsigned long get_wchan(struct task_struct
*p
)
853 unsigned long ebp
, esp
, eip
;
854 unsigned long stack_page
;
856 if (!p
|| p
== current
|| p
->state
== TASK_RUNNING
)
858 stack_page
= (unsigned long)task_stack_page(p
);
860 if (!stack_page
|| esp
< stack_page
|| esp
> top_esp
+stack_page
)
862 /* include/asm-i386/system.h:switch_to() pushes ebp last. */
863 ebp
= *(unsigned long *) esp
;
865 if (ebp
< stack_page
|| ebp
> top_ebp
+stack_page
)
867 eip
= *(unsigned long *) (ebp
+4);
868 if (!in_sched_functions(eip
))
870 ebp
= *(unsigned long *) ebp
;
871 } while (count
++ < 16);
876 * sys_alloc_thread_area: get a yet unused TLS descriptor index.
878 static int get_free_idx(void)
880 struct thread_struct
*t
= ¤t
->thread
;
883 for (idx
= 0; idx
< GDT_ENTRY_TLS_ENTRIES
; idx
++)
884 if (desc_empty(t
->tls_array
+ idx
))
885 return idx
+ GDT_ENTRY_TLS_MIN
;
890 * Set a given TLS descriptor:
892 asmlinkage
int sys_set_thread_area(struct user_desc __user
*u_info
)
894 struct thread_struct
*t
= ¤t
->thread
;
895 struct user_desc info
;
896 struct desc_struct
*desc
;
899 if (copy_from_user(&info
, u_info
, sizeof(info
)))
901 idx
= info
.entry_number
;
904 * index -1 means the kernel should try to find and
905 * allocate an empty descriptor:
908 idx
= get_free_idx();
911 if (put_user(idx
, &u_info
->entry_number
))
915 if (idx
< GDT_ENTRY_TLS_MIN
|| idx
> GDT_ENTRY_TLS_MAX
)
918 desc
= t
->tls_array
+ idx
- GDT_ENTRY_TLS_MIN
;
921 * We must not get preempted while modifying the TLS.
925 if (LDT_empty(&info
)) {
929 desc
->a
= LDT_entry_a(&info
);
930 desc
->b
= LDT_entry_b(&info
);
940 * Get the current Thread-Local Storage area:
943 #define GET_BASE(desc) ( \
944 (((desc)->a >> 16) & 0x0000ffff) | \
945 (((desc)->b << 16) & 0x00ff0000) | \
946 ( (desc)->b & 0xff000000) )
948 #define GET_LIMIT(desc) ( \
949 ((desc)->a & 0x0ffff) | \
950 ((desc)->b & 0xf0000) )
952 #define GET_32BIT(desc) (((desc)->b >> 22) & 1)
953 #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3)
954 #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1)
955 #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1)
956 #define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
957 #define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
959 asmlinkage
int sys_get_thread_area(struct user_desc __user
*u_info
)
961 struct user_desc info
;
962 struct desc_struct
*desc
;
965 if (get_user(idx
, &u_info
->entry_number
))
967 if (idx
< GDT_ENTRY_TLS_MIN
|| idx
> GDT_ENTRY_TLS_MAX
)
970 memset(&info
, 0, sizeof(info
));
972 desc
= current
->thread
.tls_array
+ idx
- GDT_ENTRY_TLS_MIN
;
974 info
.entry_number
= idx
;
975 info
.base_addr
= GET_BASE(desc
);
976 info
.limit
= GET_LIMIT(desc
);
977 info
.seg_32bit
= GET_32BIT(desc
);
978 info
.contents
= GET_CONTENTS(desc
);
979 info
.read_exec_only
= !GET_WRITABLE(desc
);
980 info
.limit_in_pages
= GET_LIMIT_PAGES(desc
);
981 info
.seg_not_present
= !GET_PRESENT(desc
);
982 info
.useable
= GET_USEABLE(desc
);
984 if (copy_to_user(u_info
, &info
, sizeof(info
)))
989 unsigned long arch_align_stack(unsigned long sp
)
991 if (!(current
->personality
& ADDR_NO_RANDOMIZE
) && randomize_va_space
)
992 sp
-= get_random_int() % 8192;
996 unsigned long arch_randomize_brk(struct mm_struct
*mm
)
998 unsigned long range_end
= mm
->brk
+ 0x02000000;
999 return randomize_range(mm
->brk
, range_end
, 0) ? : mm
->brk
;