diff options
Diffstat (limited to 'arch/x86/kernel/process_32.c')
-rw-r--r-- | arch/x86/kernel/process_32.c | 419 |
1 files changed, 167 insertions, 252 deletions
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index 46d391d49de..968371ab223 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c | |||
@@ -55,6 +55,7 @@ | |||
55 | 55 | ||
56 | #include <asm/tlbflush.h> | 56 | #include <asm/tlbflush.h> |
57 | #include <asm/cpu.h> | 57 | #include <asm/cpu.h> |
58 | #include <asm/kdebug.h> | ||
58 | 59 | ||
59 | asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); | 60 | asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); |
60 | 61 | ||
@@ -74,7 +75,7 @@ EXPORT_PER_CPU_SYMBOL(cpu_number); | |||
74 | */ | 75 | */ |
75 | unsigned long thread_saved_pc(struct task_struct *tsk) | 76 | unsigned long thread_saved_pc(struct task_struct *tsk) |
76 | { | 77 | { |
77 | return ((unsigned long *)tsk->thread.esp)[3]; | 78 | return ((unsigned long *)tsk->thread.sp)[3]; |
78 | } | 79 | } |
79 | 80 | ||
80 | /* | 81 | /* |
@@ -113,10 +114,19 @@ void default_idle(void) | |||
113 | smp_mb(); | 114 | smp_mb(); |
114 | 115 | ||
115 | local_irq_disable(); | 116 | local_irq_disable(); |
116 | if (!need_resched()) | 117 | if (!need_resched()) { |
118 | ktime_t t0, t1; | ||
119 | u64 t0n, t1n; | ||
120 | |||
121 | t0 = ktime_get(); | ||
122 | t0n = ktime_to_ns(t0); | ||
117 | safe_halt(); /* enables interrupts racelessly */ | 123 | safe_halt(); /* enables interrupts racelessly */ |
118 | else | 124 | local_irq_disable(); |
119 | local_irq_enable(); | 125 | t1 = ktime_get(); |
126 | t1n = ktime_to_ns(t1); | ||
127 | sched_clock_idle_wakeup_event(t1n - t0n); | ||
128 | } | ||
129 | local_irq_enable(); | ||
120 | current_thread_info()->status |= TS_POLLING; | 130 | current_thread_info()->status |= TS_POLLING; |
121 | } else { | 131 | } else { |
122 | /* loop is done by the caller */ | 132 | /* loop is done by the caller */ |
@@ -132,7 +142,7 @@ EXPORT_SYMBOL(default_idle); | |||
132 | * to poll the ->work.need_resched flag instead of waiting for the | 142 | * to poll the ->work.need_resched flag instead of waiting for the |
133 | * cross-CPU IPI to arrive. Use this option with caution. | 143 | * cross-CPU IPI to arrive. Use this option with caution. |
134 | */ | 144 | */ |
135 | static void poll_idle (void) | 145 | static void poll_idle(void) |
136 | { | 146 | { |
137 | cpu_relax(); | 147 | cpu_relax(); |
138 | } | 148 | } |
@@ -188,6 +198,9 @@ void cpu_idle(void) | |||
188 | rmb(); | 198 | rmb(); |
189 | idle = pm_idle; | 199 | idle = pm_idle; |
190 | 200 | ||
201 | if (rcu_pending(cpu)) | ||
202 | rcu_check_callbacks(cpu, 0); | ||
203 | |||
191 | if (!idle) | 204 | if (!idle) |
192 | idle = default_idle; | 205 | idle = default_idle; |
193 | 206 | ||
@@ -255,13 +268,13 @@ EXPORT_SYMBOL_GPL(cpu_idle_wait); | |||
255 | * New with Core Duo processors, MWAIT can take some hints based on CPU | 268 | * New with Core Duo processors, MWAIT can take some hints based on CPU |
256 | * capability. | 269 | * capability. |
257 | */ | 270 | */ |
258 | void mwait_idle_with_hints(unsigned long eax, unsigned long ecx) | 271 | void mwait_idle_with_hints(unsigned long ax, unsigned long cx) |
259 | { | 272 | { |
260 | if (!need_resched()) { | 273 | if (!need_resched()) { |
261 | __monitor((void *)¤t_thread_info()->flags, 0, 0); | 274 | __monitor((void *)¤t_thread_info()->flags, 0, 0); |
262 | smp_mb(); | 275 | smp_mb(); |
263 | if (!need_resched()) | 276 | if (!need_resched()) |
264 | __mwait(eax, ecx); | 277 | __mwait(ax, cx); |
265 | } | 278 | } |
266 | } | 279 | } |
267 | 280 | ||
@@ -272,19 +285,37 @@ static void mwait_idle(void) | |||
272 | mwait_idle_with_hints(0, 0); | 285 | mwait_idle_with_hints(0, 0); |
273 | } | 286 | } |
274 | 287 | ||
288 | static int __cpuinit mwait_usable(const struct cpuinfo_x86 *c) | ||
289 | { | ||
290 | if (force_mwait) | ||
291 | return 1; | ||
292 | /* Any C1 states supported? */ | ||
293 | return c->cpuid_level >= 5 && ((cpuid_edx(5) >> 4) & 0xf) > 0; | ||
294 | } | ||
295 | |||
275 | void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) | 296 | void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) |
276 | { | 297 | { |
277 | if (cpu_has(c, X86_FEATURE_MWAIT)) { | 298 | static int selected; |
278 | printk("monitor/mwait feature present.\n"); | 299 | |
300 | if (selected) | ||
301 | return; | ||
302 | #ifdef CONFIG_X86_SMP | ||
303 | if (pm_idle == poll_idle && smp_num_siblings > 1) { | ||
304 | printk(KERN_WARNING "WARNING: polling idle and HT enabled," | ||
305 | " performance may degrade.\n"); | ||
306 | } | ||
307 | #endif | ||
308 | if (cpu_has(c, X86_FEATURE_MWAIT) && mwait_usable(c)) { | ||
279 | /* | 309 | /* |
280 | * Skip, if setup has overridden idle. | 310 | * Skip, if setup has overridden idle. |
281 | * One CPU supports mwait => All CPUs supports mwait | 311 | * One CPU supports mwait => All CPUs supports mwait |
282 | */ | 312 | */ |
283 | if (!pm_idle) { | 313 | if (!pm_idle) { |
284 | printk("using mwait in idle threads.\n"); | 314 | printk(KERN_INFO "using mwait in idle threads.\n"); |
285 | pm_idle = mwait_idle; | 315 | pm_idle = mwait_idle; |
286 | } | 316 | } |
287 | } | 317 | } |
318 | selected = 1; | ||
288 | } | 319 | } |
289 | 320 | ||
290 | static int __init idle_setup(char *str) | 321 | static int __init idle_setup(char *str) |
@@ -292,10 +323,6 @@ static int __init idle_setup(char *str) | |||
292 | if (!strcmp(str, "poll")) { | 323 | if (!strcmp(str, "poll")) { |
293 | printk("using polling idle threads.\n"); | 324 | printk("using polling idle threads.\n"); |
294 | pm_idle = poll_idle; | 325 | pm_idle = poll_idle; |
295 | #ifdef CONFIG_X86_SMP | ||
296 | if (smp_num_siblings > 1) | ||
297 | printk("WARNING: polling idle and HT enabled, performance may degrade.\n"); | ||
298 | #endif | ||
299 | } else if (!strcmp(str, "mwait")) | 326 | } else if (!strcmp(str, "mwait")) |
300 | force_mwait = 1; | 327 | force_mwait = 1; |
301 | else | 328 | else |
@@ -310,15 +337,15 @@ void __show_registers(struct pt_regs *regs, int all) | |||
310 | { | 337 | { |
311 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; | 338 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; |
312 | unsigned long d0, d1, d2, d3, d6, d7; | 339 | unsigned long d0, d1, d2, d3, d6, d7; |
313 | unsigned long esp; | 340 | unsigned long sp; |
314 | unsigned short ss, gs; | 341 | unsigned short ss, gs; |
315 | 342 | ||
316 | if (user_mode_vm(regs)) { | 343 | if (user_mode_vm(regs)) { |
317 | esp = regs->esp; | 344 | sp = regs->sp; |
318 | ss = regs->xss & 0xffff; | 345 | ss = regs->ss & 0xffff; |
319 | savesegment(gs, gs); | 346 | savesegment(gs, gs); |
320 | } else { | 347 | } else { |
321 | esp = (unsigned long) (®s->esp); | 348 | sp = (unsigned long) (®s->sp); |
322 | savesegment(ss, ss); | 349 | savesegment(ss, ss); |
323 | savesegment(gs, gs); | 350 | savesegment(gs, gs); |
324 | } | 351 | } |
@@ -331,17 +358,17 @@ void __show_registers(struct pt_regs *regs, int all) | |||
331 | init_utsname()->version); | 358 | init_utsname()->version); |
332 | 359 | ||
333 | printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", | 360 | printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", |
334 | 0xffff & regs->xcs, regs->eip, regs->eflags, | 361 | 0xffff & regs->cs, regs->ip, regs->flags, |
335 | smp_processor_id()); | 362 | smp_processor_id()); |
336 | print_symbol("EIP is at %s\n", regs->eip); | 363 | print_symbol("EIP is at %s\n", regs->ip); |
337 | 364 | ||
338 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", | 365 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
339 | regs->eax, regs->ebx, regs->ecx, regs->edx); | 366 | regs->ax, regs->bx, regs->cx, regs->dx); |
340 | printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", | 367 | printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", |
341 | regs->esi, regs->edi, regs->ebp, esp); | 368 | regs->si, regs->di, regs->bp, sp); |
342 | printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", | 369 | printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", |
343 | regs->xds & 0xffff, regs->xes & 0xffff, | 370 | regs->ds & 0xffff, regs->es & 0xffff, |
344 | regs->xfs & 0xffff, gs, ss); | 371 | regs->fs & 0xffff, gs, ss); |
345 | 372 | ||
346 | if (!all) | 373 | if (!all) |
347 | return; | 374 | return; |
@@ -369,12 +396,12 @@ void __show_registers(struct pt_regs *regs, int all) | |||
369 | void show_regs(struct pt_regs *regs) | 396 | void show_regs(struct pt_regs *regs) |
370 | { | 397 | { |
371 | __show_registers(regs, 1); | 398 | __show_registers(regs, 1); |
372 | show_trace(NULL, regs, ®s->esp); | 399 | show_trace(NULL, regs, ®s->sp, regs->bp); |
373 | } | 400 | } |
374 | 401 | ||
375 | /* | 402 | /* |
376 | * This gets run with %ebx containing the | 403 | * This gets run with %bx containing the |
377 | * function to call, and %edx containing | 404 | * function to call, and %dx containing |
378 | * the "args". | 405 | * the "args". |
379 | */ | 406 | */ |
380 | extern void kernel_thread_helper(void); | 407 | extern void kernel_thread_helper(void); |
@@ -388,16 +415,16 @@ int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |||
388 | 415 | ||
389 | memset(®s, 0, sizeof(regs)); | 416 | memset(®s, 0, sizeof(regs)); |
390 | 417 | ||
391 | regs.ebx = (unsigned long) fn; | 418 | regs.bx = (unsigned long) fn; |
392 | regs.edx = (unsigned long) arg; | 419 | regs.dx = (unsigned long) arg; |
393 | 420 | ||
394 | regs.xds = __USER_DS; | 421 | regs.ds = __USER_DS; |
395 | regs.xes = __USER_DS; | 422 | regs.es = __USER_DS; |
396 | regs.xfs = __KERNEL_PERCPU; | 423 | regs.fs = __KERNEL_PERCPU; |
397 | regs.orig_eax = -1; | 424 | regs.orig_ax = -1; |
398 | regs.eip = (unsigned long) kernel_thread_helper; | 425 | regs.ip = (unsigned long) kernel_thread_helper; |
399 | regs.xcs = __KERNEL_CS | get_kernel_rpl(); | 426 | regs.cs = __KERNEL_CS | get_kernel_rpl(); |
400 | regs.eflags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; | 427 | regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; |
401 | 428 | ||
402 | /* Ok, create the new process.. */ | 429 | /* Ok, create the new process.. */ |
403 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); | 430 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); |
@@ -435,7 +462,12 @@ void flush_thread(void) | |||
435 | { | 462 | { |
436 | struct task_struct *tsk = current; | 463 | struct task_struct *tsk = current; |
437 | 464 | ||
438 | memset(tsk->thread.debugreg, 0, sizeof(unsigned long)*8); | 465 | tsk->thread.debugreg0 = 0; |
466 | tsk->thread.debugreg1 = 0; | ||
467 | tsk->thread.debugreg2 = 0; | ||
468 | tsk->thread.debugreg3 = 0; | ||
469 | tsk->thread.debugreg6 = 0; | ||
470 | tsk->thread.debugreg7 = 0; | ||
439 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); | 471 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
440 | clear_tsk_thread_flag(tsk, TIF_DEBUG); | 472 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
441 | /* | 473 | /* |
@@ -460,7 +492,7 @@ void prepare_to_copy(struct task_struct *tsk) | |||
460 | unlazy_fpu(tsk); | 492 | unlazy_fpu(tsk); |
461 | } | 493 | } |
462 | 494 | ||
463 | int copy_thread(int nr, unsigned long clone_flags, unsigned long esp, | 495 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
464 | unsigned long unused, | 496 | unsigned long unused, |
465 | struct task_struct * p, struct pt_regs * regs) | 497 | struct task_struct * p, struct pt_regs * regs) |
466 | { | 498 | { |
@@ -470,15 +502,15 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp, | |||
470 | 502 | ||
471 | childregs = task_pt_regs(p); | 503 | childregs = task_pt_regs(p); |
472 | *childregs = *regs; | 504 | *childregs = *regs; |
473 | childregs->eax = 0; | 505 | childregs->ax = 0; |
474 | childregs->esp = esp; | 506 | childregs->sp = sp; |
475 | 507 | ||
476 | p->thread.esp = (unsigned long) childregs; | 508 | p->thread.sp = (unsigned long) childregs; |
477 | p->thread.esp0 = (unsigned long) (childregs+1); | 509 | p->thread.sp0 = (unsigned long) (childregs+1); |
478 | 510 | ||
479 | p->thread.eip = (unsigned long) ret_from_fork; | 511 | p->thread.ip = (unsigned long) ret_from_fork; |
480 | 512 | ||
481 | savesegment(gs,p->thread.gs); | 513 | savesegment(gs, p->thread.gs); |
482 | 514 | ||
483 | tsk = current; | 515 | tsk = current; |
484 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { | 516 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
@@ -491,32 +523,15 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp, | |||
491 | set_tsk_thread_flag(p, TIF_IO_BITMAP); | 523 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
492 | } | 524 | } |
493 | 525 | ||
526 | err = 0; | ||
527 | |||
494 | /* | 528 | /* |
495 | * Set a new TLS for the child thread? | 529 | * Set a new TLS for the child thread? |
496 | */ | 530 | */ |
497 | if (clone_flags & CLONE_SETTLS) { | 531 | if (clone_flags & CLONE_SETTLS) |
498 | struct desc_struct *desc; | 532 | err = do_set_thread_area(p, -1, |
499 | struct user_desc info; | 533 | (struct user_desc __user *)childregs->si, 0); |
500 | int idx; | ||
501 | |||
502 | err = -EFAULT; | ||
503 | if (copy_from_user(&info, (void __user *)childregs->esi, sizeof(info))) | ||
504 | goto out; | ||
505 | err = -EINVAL; | ||
506 | if (LDT_empty(&info)) | ||
507 | goto out; | ||
508 | |||
509 | idx = info.entry_number; | ||
510 | if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX) | ||
511 | goto out; | ||
512 | |||
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); | ||
516 | } | ||
517 | 534 | ||
518 | err = 0; | ||
519 | out: | ||
520 | if (err && p->thread.io_bitmap_ptr) { | 535 | if (err && p->thread.io_bitmap_ptr) { |
521 | kfree(p->thread.io_bitmap_ptr); | 536 | kfree(p->thread.io_bitmap_ptr); |
522 | p->thread.io_bitmap_max = 0; | 537 | p->thread.io_bitmap_max = 0; |
@@ -529,62 +544,52 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp, | |||
529 | */ | 544 | */ |
530 | void dump_thread(struct pt_regs * regs, struct user * dump) | 545 | void dump_thread(struct pt_regs * regs, struct user * dump) |
531 | { | 546 | { |
532 | int i; | 547 | u16 gs; |
533 | 548 | ||
534 | /* changed the size calculations - should hopefully work better. lbt */ | 549 | /* changed the size calculations - should hopefully work better. lbt */ |
535 | dump->magic = CMAGIC; | 550 | dump->magic = CMAGIC; |
536 | dump->start_code = 0; | 551 | dump->start_code = 0; |
537 | dump->start_stack = regs->esp & ~(PAGE_SIZE - 1); | 552 | dump->start_stack = regs->sp & ~(PAGE_SIZE - 1); |
538 | dump->u_tsize = ((unsigned long) current->mm->end_code) >> PAGE_SHIFT; | 553 | 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; | 554 | dump->u_dsize = ((unsigned long) (current->mm->brk + (PAGE_SIZE-1))) >> PAGE_SHIFT; |
540 | dump->u_dsize -= dump->u_tsize; | 555 | dump->u_dsize -= dump->u_tsize; |
541 | dump->u_ssize = 0; | 556 | dump->u_ssize = 0; |
542 | for (i = 0; i < 8; i++) | 557 | dump->u_debugreg[0] = current->thread.debugreg0; |
543 | dump->u_debugreg[i] = current->thread.debugreg[i]; | 558 | dump->u_debugreg[1] = current->thread.debugreg1; |
559 | dump->u_debugreg[2] = current->thread.debugreg2; | ||
560 | dump->u_debugreg[3] = current->thread.debugreg3; | ||
561 | dump->u_debugreg[4] = 0; | ||
562 | dump->u_debugreg[5] = 0; | ||
563 | dump->u_debugreg[6] = current->thread.debugreg6; | ||
564 | dump->u_debugreg[7] = current->thread.debugreg7; | ||
544 | 565 | ||
545 | if (dump->start_stack < TASK_SIZE) | 566 | if (dump->start_stack < TASK_SIZE) |
546 | dump->u_ssize = ((unsigned long) (TASK_SIZE - dump->start_stack)) >> PAGE_SHIFT; | 567 | dump->u_ssize = ((unsigned long) (TASK_SIZE - dump->start_stack)) >> PAGE_SHIFT; |
547 | 568 | ||
548 | dump->regs.ebx = regs->ebx; | 569 | dump->regs.bx = regs->bx; |
549 | dump->regs.ecx = regs->ecx; | 570 | dump->regs.cx = regs->cx; |
550 | dump->regs.edx = regs->edx; | 571 | dump->regs.dx = regs->dx; |
551 | dump->regs.esi = regs->esi; | 572 | dump->regs.si = regs->si; |
552 | dump->regs.edi = regs->edi; | 573 | dump->regs.di = regs->di; |
553 | dump->regs.ebp = regs->ebp; | 574 | dump->regs.bp = regs->bp; |
554 | dump->regs.eax = regs->eax; | 575 | dump->regs.ax = regs->ax; |
555 | dump->regs.ds = regs->xds; | 576 | dump->regs.ds = (u16)regs->ds; |
556 | dump->regs.es = regs->xes; | 577 | dump->regs.es = (u16)regs->es; |
557 | dump->regs.fs = regs->xfs; | 578 | dump->regs.fs = (u16)regs->fs; |
558 | savesegment(gs,dump->regs.gs); | 579 | savesegment(gs,gs); |
559 | dump->regs.orig_eax = regs->orig_eax; | 580 | dump->regs.orig_ax = regs->orig_ax; |
560 | dump->regs.eip = regs->eip; | 581 | dump->regs.ip = regs->ip; |
561 | dump->regs.cs = regs->xcs; | 582 | dump->regs.cs = (u16)regs->cs; |
562 | dump->regs.eflags = regs->eflags; | 583 | dump->regs.flags = regs->flags; |
563 | dump->regs.esp = regs->esp; | 584 | dump->regs.sp = regs->sp; |
564 | dump->regs.ss = regs->xss; | 585 | dump->regs.ss = (u16)regs->ss; |
565 | 586 | ||
566 | dump->u_fpvalid = dump_fpu (regs, &dump->i387); | 587 | dump->u_fpvalid = dump_fpu (regs, &dump->i387); |
567 | } | 588 | } |
568 | EXPORT_SYMBOL(dump_thread); | 589 | EXPORT_SYMBOL(dump_thread); |
569 | 590 | ||
570 | /* | ||
571 | * Capture the user space registers if the task is not running (in user space) | ||
572 | */ | ||
573 | int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs) | ||
574 | { | ||
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; | ||
580 | |||
581 | elf_core_copy_regs(regs, &ptregs); | ||
582 | |||
583 | return 1; | ||
584 | } | ||
585 | |||
586 | #ifdef CONFIG_SECCOMP | 591 | #ifdef CONFIG_SECCOMP |
587 | void hard_disable_TSC(void) | 592 | static void hard_disable_TSC(void) |
588 | { | 593 | { |
589 | write_cr4(read_cr4() | X86_CR4_TSD); | 594 | write_cr4(read_cr4() | X86_CR4_TSD); |
590 | } | 595 | } |
@@ -599,7 +604,7 @@ void disable_TSC(void) | |||
599 | hard_disable_TSC(); | 604 | hard_disable_TSC(); |
600 | preempt_enable(); | 605 | preempt_enable(); |
601 | } | 606 | } |
602 | void hard_enable_TSC(void) | 607 | static void hard_enable_TSC(void) |
603 | { | 608 | { |
604 | write_cr4(read_cr4() & ~X86_CR4_TSD); | 609 | write_cr4(read_cr4() & ~X86_CR4_TSD); |
605 | } | 610 | } |
@@ -609,18 +614,32 @@ static noinline void | |||
609 | __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | 614 | __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, |
610 | struct tss_struct *tss) | 615 | struct tss_struct *tss) |
611 | { | 616 | { |
612 | struct thread_struct *next; | 617 | struct thread_struct *prev, *next; |
618 | unsigned long debugctl; | ||
613 | 619 | ||
620 | prev = &prev_p->thread; | ||
614 | next = &next_p->thread; | 621 | next = &next_p->thread; |
615 | 622 | ||
623 | debugctl = prev->debugctlmsr; | ||
624 | if (next->ds_area_msr != prev->ds_area_msr) { | ||
625 | /* we clear debugctl to make sure DS | ||
626 | * is not in use when we change it */ | ||
627 | debugctl = 0; | ||
628 | wrmsrl(MSR_IA32_DEBUGCTLMSR, 0); | ||
629 | wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0); | ||
630 | } | ||
631 | |||
632 | if (next->debugctlmsr != debugctl) | ||
633 | wrmsr(MSR_IA32_DEBUGCTLMSR, next->debugctlmsr, 0); | ||
634 | |||
616 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { | 635 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
617 | set_debugreg(next->debugreg[0], 0); | 636 | set_debugreg(next->debugreg0, 0); |
618 | set_debugreg(next->debugreg[1], 1); | 637 | set_debugreg(next->debugreg1, 1); |
619 | set_debugreg(next->debugreg[2], 2); | 638 | set_debugreg(next->debugreg2, 2); |
620 | set_debugreg(next->debugreg[3], 3); | 639 | set_debugreg(next->debugreg3, 3); |
621 | /* no 4 and 5 */ | 640 | /* no 4 and 5 */ |
622 | set_debugreg(next->debugreg[6], 6); | 641 | set_debugreg(next->debugreg6, 6); |
623 | set_debugreg(next->debugreg[7], 7); | 642 | set_debugreg(next->debugreg7, 7); |
624 | } | 643 | } |
625 | 644 | ||
626 | #ifdef CONFIG_SECCOMP | 645 | #ifdef CONFIG_SECCOMP |
@@ -634,6 +653,13 @@ __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |||
634 | } | 653 | } |
635 | #endif | 654 | #endif |
636 | 655 | ||
656 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) | ||
657 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | ||
658 | |||
659 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | ||
660 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | ||
661 | |||
662 | |||
637 | if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { | 663 | if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
638 | /* | 664 | /* |
639 | * Disable the bitmap via an invalid offset. We still cache | 665 | * Disable the bitmap via an invalid offset. We still cache |
@@ -687,11 +713,11 @@ __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |||
687 | * More important, however, is the fact that this allows us much | 713 | * More important, however, is the fact that this allows us much |
688 | * more flexibility. | 714 | * more flexibility. |
689 | * | 715 | * |
690 | * The return value (in %eax) will be the "prev" task after | 716 | * The return value (in %ax) will be the "prev" task after |
691 | * the task-switch, and shows up in ret_from_fork in entry.S, | 717 | * the task-switch, and shows up in ret_from_fork in entry.S, |
692 | * for example. | 718 | * for example. |
693 | */ | 719 | */ |
694 | struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct task_struct *next_p) | 720 | struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
695 | { | 721 | { |
696 | struct thread_struct *prev = &prev_p->thread, | 722 | struct thread_struct *prev = &prev_p->thread, |
697 | *next = &next_p->thread; | 723 | *next = &next_p->thread; |
@@ -710,7 +736,7 @@ struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct tas | |||
710 | /* | 736 | /* |
711 | * Reload esp0. | 737 | * Reload esp0. |
712 | */ | 738 | */ |
713 | load_esp0(tss, next); | 739 | load_sp0(tss, next); |
714 | 740 | ||
715 | /* | 741 | /* |
716 | * Save away %gs. No need to save %fs, as it was saved on the | 742 | * Save away %gs. No need to save %fs, as it was saved on the |
@@ -774,7 +800,7 @@ struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct tas | |||
774 | 800 | ||
775 | asmlinkage int sys_fork(struct pt_regs regs) | 801 | asmlinkage int sys_fork(struct pt_regs regs) |
776 | { | 802 | { |
777 | return do_fork(SIGCHLD, regs.esp, ®s, 0, NULL, NULL); | 803 | return do_fork(SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
778 | } | 804 | } |
779 | 805 | ||
780 | asmlinkage int sys_clone(struct pt_regs regs) | 806 | asmlinkage int sys_clone(struct pt_regs regs) |
@@ -783,12 +809,12 @@ asmlinkage int sys_clone(struct pt_regs regs) | |||
783 | unsigned long newsp; | 809 | unsigned long newsp; |
784 | int __user *parent_tidptr, *child_tidptr; | 810 | int __user *parent_tidptr, *child_tidptr; |
785 | 811 | ||
786 | clone_flags = regs.ebx; | 812 | clone_flags = regs.bx; |
787 | newsp = regs.ecx; | 813 | newsp = regs.cx; |
788 | parent_tidptr = (int __user *)regs.edx; | 814 | parent_tidptr = (int __user *)regs.dx; |
789 | child_tidptr = (int __user *)regs.edi; | 815 | child_tidptr = (int __user *)regs.di; |
790 | if (!newsp) | 816 | if (!newsp) |
791 | newsp = regs.esp; | 817 | newsp = regs.sp; |
792 | return do_fork(clone_flags, newsp, ®s, 0, parent_tidptr, child_tidptr); | 818 | return do_fork(clone_flags, newsp, ®s, 0, parent_tidptr, child_tidptr); |
793 | } | 819 | } |
794 | 820 | ||
@@ -804,7 +830,7 @@ asmlinkage int sys_clone(struct pt_regs regs) | |||
804 | */ | 830 | */ |
805 | asmlinkage int sys_vfork(struct pt_regs regs) | 831 | asmlinkage int sys_vfork(struct pt_regs regs) |
806 | { | 832 | { |
807 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.esp, ®s, 0, NULL, NULL); | 833 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
808 | } | 834 | } |
809 | 835 | ||
810 | /* | 836 | /* |
@@ -815,18 +841,15 @@ asmlinkage int sys_execve(struct pt_regs regs) | |||
815 | int error; | 841 | int error; |
816 | char * filename; | 842 | char * filename; |
817 | 843 | ||
818 | filename = getname((char __user *) regs.ebx); | 844 | filename = getname((char __user *) regs.bx); |
819 | error = PTR_ERR(filename); | 845 | error = PTR_ERR(filename); |
820 | if (IS_ERR(filename)) | 846 | if (IS_ERR(filename)) |
821 | goto out; | 847 | goto out; |
822 | error = do_execve(filename, | 848 | error = do_execve(filename, |
823 | (char __user * __user *) regs.ecx, | 849 | (char __user * __user *) regs.cx, |
824 | (char __user * __user *) regs.edx, | 850 | (char __user * __user *) regs.dx, |
825 | ®s); | 851 | ®s); |
826 | if (error == 0) { | 852 | if (error == 0) { |
827 | task_lock(current); | ||
828 | current->ptrace &= ~PT_DTRACE; | ||
829 | task_unlock(current); | ||
830 | /* Make sure we don't return using sysenter.. */ | 853 | /* Make sure we don't return using sysenter.. */ |
831 | set_thread_flag(TIF_IRET); | 854 | set_thread_flag(TIF_IRET); |
832 | } | 855 | } |
@@ -840,145 +863,37 @@ out: | |||
840 | 863 | ||
841 | unsigned long get_wchan(struct task_struct *p) | 864 | unsigned long get_wchan(struct task_struct *p) |
842 | { | 865 | { |
843 | unsigned long ebp, esp, eip; | 866 | unsigned long bp, sp, ip; |
844 | unsigned long stack_page; | 867 | unsigned long stack_page; |
845 | int count = 0; | 868 | int count = 0; |
846 | if (!p || p == current || p->state == TASK_RUNNING) | 869 | if (!p || p == current || p->state == TASK_RUNNING) |
847 | return 0; | 870 | return 0; |
848 | stack_page = (unsigned long)task_stack_page(p); | 871 | stack_page = (unsigned long)task_stack_page(p); |
849 | esp = p->thread.esp; | 872 | sp = p->thread.sp; |
850 | if (!stack_page || esp < stack_page || esp > top_esp+stack_page) | 873 | if (!stack_page || sp < stack_page || sp > top_esp+stack_page) |
851 | return 0; | 874 | return 0; |
852 | /* include/asm-i386/system.h:switch_to() pushes ebp last. */ | 875 | /* include/asm-i386/system.h:switch_to() pushes bp last. */ |
853 | ebp = *(unsigned long *) esp; | 876 | bp = *(unsigned long *) sp; |
854 | do { | 877 | do { |
855 | if (ebp < stack_page || ebp > top_ebp+stack_page) | 878 | if (bp < stack_page || bp > top_ebp+stack_page) |
856 | return 0; | 879 | return 0; |
857 | eip = *(unsigned long *) (ebp+4); | 880 | ip = *(unsigned long *) (bp+4); |
858 | if (!in_sched_functions(eip)) | 881 | if (!in_sched_functions(ip)) |
859 | return eip; | 882 | return ip; |
860 | ebp = *(unsigned long *) ebp; | 883 | bp = *(unsigned long *) bp; |
861 | } while (count++ < 16); | 884 | } while (count++ < 16); |
862 | return 0; | 885 | return 0; |
863 | } | 886 | } |
864 | 887 | ||
865 | /* | ||
866 | * sys_alloc_thread_area: get a yet unused TLS descriptor index. | ||
867 | */ | ||
868 | static int get_free_idx(void) | ||
869 | { | ||
870 | struct thread_struct *t = ¤t->thread; | ||
871 | int idx; | ||
872 | |||
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; | ||
876 | return -ESRCH; | ||
877 | } | ||
878 | |||
879 | /* | ||
880 | * Set a given TLS descriptor: | ||
881 | */ | ||
882 | asmlinkage int sys_set_thread_area(struct user_desc __user *u_info) | ||
883 | { | ||
884 | struct thread_struct *t = ¤t->thread; | ||
885 | struct user_desc info; | ||
886 | struct desc_struct *desc; | ||
887 | int cpu, idx; | ||
888 | |||
889 | if (copy_from_user(&info, u_info, sizeof(info))) | ||
890 | return -EFAULT; | ||
891 | idx = info.entry_number; | ||
892 | |||
893 | /* | ||
894 | * index -1 means the kernel should try to find and | ||
895 | * allocate an empty descriptor: | ||
896 | */ | ||
897 | if (idx == -1) { | ||
898 | idx = get_free_idx(); | ||
899 | if (idx < 0) | ||
900 | return idx; | ||
901 | if (put_user(idx, &u_info->entry_number)) | ||
902 | return -EFAULT; | ||
903 | } | ||
904 | |||
905 | if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX) | ||
906 | return -EINVAL; | ||
907 | |||
908 | desc = t->tls_array + idx - GDT_ENTRY_TLS_MIN; | ||
909 | |||
910 | /* | ||
911 | * We must not get preempted while modifying the TLS. | ||
912 | */ | ||
913 | cpu = get_cpu(); | ||
914 | |||
915 | if (LDT_empty(&info)) { | ||
916 | desc->a = 0; | ||
917 | desc->b = 0; | ||
918 | } else { | ||
919 | desc->a = LDT_entry_a(&info); | ||
920 | desc->b = LDT_entry_b(&info); | ||
921 | } | ||
922 | load_TLS(t, cpu); | ||
923 | |||
924 | put_cpu(); | ||
925 | |||
926 | return 0; | ||
927 | } | ||
928 | |||
929 | /* | ||
930 | * Get the current Thread-Local Storage area: | ||
931 | */ | ||
932 | |||
933 | #define GET_BASE(desc) ( \ | ||
934 | (((desc)->a >> 16) & 0x0000ffff) | \ | ||
935 | (((desc)->b << 16) & 0x00ff0000) | \ | ||
936 | ( (desc)->b & 0xff000000) ) | ||
937 | |||
938 | #define GET_LIMIT(desc) ( \ | ||
939 | ((desc)->a & 0x0ffff) | \ | ||
940 | ((desc)->b & 0xf0000) ) | ||
941 | |||
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) | ||
948 | |||
949 | asmlinkage int sys_get_thread_area(struct user_desc __user *u_info) | ||
950 | { | ||
951 | struct user_desc info; | ||
952 | struct desc_struct *desc; | ||
953 | int idx; | ||
954 | |||
955 | if (get_user(idx, &u_info->entry_number)) | ||
956 | return -EFAULT; | ||
957 | if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX) | ||
958 | return -EINVAL; | ||
959 | |||
960 | memset(&info, 0, sizeof(info)); | ||
961 | |||
962 | desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN; | ||
963 | |||
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); | ||
973 | |||
974 | if (copy_to_user(u_info, &info, sizeof(info))) | ||
975 | return -EFAULT; | ||
976 | return 0; | ||
977 | } | ||
978 | |||
979 | unsigned long arch_align_stack(unsigned long sp) | 888 | unsigned long arch_align_stack(unsigned long sp) |
980 | { | 889 | { |
981 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) | 890 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
982 | sp -= get_random_int() % 8192; | 891 | sp -= get_random_int() % 8192; |
983 | return sp & ~0xf; | 892 | return sp & ~0xf; |
984 | } | 893 | } |
894 | |||
895 | unsigned long arch_randomize_brk(struct mm_struct *mm) | ||
896 | { | ||
897 | unsigned long range_end = mm->brk + 0x02000000; | ||
898 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | ||
899 | } | ||