arch/x86/kernel/process_32.c

   1 /*
   2  *  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
  12 #include <stdarg.h>
  13
  14 #include <linux/cpu.h>
  15 #include <linux/errno.h>
  16 #include <linux/sched.h>
  17 #include <linux/fs.h>
  18 #include <linux/kernel.h>
  19 #include <linux/mm.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/interrupt.h>
  27 #include <linux/utsname.h>
  28 #include <linux/delay.h>
  29 #include <linux/reboot.h>
  30 #include <linux/init.h>
  31 #include <linux/mc146818rtc.h>
  32 #include <linux/module.h>
  33 #include <linux/kallsyms.h>
  34 #include <linux/ptrace.h>
  35 #include <linux/random.h>
  36 #include <linux/personality.h>
  37 #include <linux/tick.h>
  38 #include <linux/percpu.h>
  39 #include <linux/prctl.h>
  40 #include <linux/dmi.h>
  41
  42 #include <asm/uaccess.h>
  43 #include <asm/pgtable.h>
  44 #include <asm/system.h>
  45 #include <asm/io.h>
  46 #include <asm/ldt.h>
  47 #include <asm/processor.h>
  48 #include <asm/i387.h>
  49 #include <asm/desc.h>
  50 #ifdef CONFIG_MATH_EMULATION
  51 #include <asm/math_emu.h>
  52 #endif
  53
  54 #include <linux/err.h>
  55
  56 #include <asm/tlbflush.h>
  57 #include <asm/cpu.h>
  58 #include <asm/kdebug.h>
  59 #include <asm/idle.h>
  60 #include <asm/syscalls.h>
  61 #include <asm/smp.h>
  62
  63 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
  64
  65 DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
  66 EXPORT_PER_CPU_SYMBOL(current_task);
  67
  68 DEFINE_PER_CPU(int, cpu_number);
  69 EXPORT_PER_CPU_SYMBOL(cpu_number);
  70
  71 /*
  72  * Return saved PC of a blocked thread.
  73  */
  74 unsigned long thread_saved_pc(struct task_struct *tsk)
  75 {
  76         return ((unsigned long *)tsk->thread.sp)[3];
  77 }
  78
  79 #ifdef CONFIG_HOTPLUG_CPU
  80 #include <asm/nmi.h>
  81
  82 static void cpu_exit_clear(void)
  83 {
  84         int cpu = raw_smp_processor_id();
  85
  86         idle_task_exit();
  87
  88         cpu_uninit();
  89         irq_ctx_exit(cpu);
  90
  91         cpu_clear(cpu, cpu_callout_map);
  92         cpu_clear(cpu, cpu_callin_map);
  93
  94         numa_remove_cpu(cpu);
  95         c1e_remove_cpu(cpu);
  96 }
  97
  98 /* We don't actually take CPU down, just spin without interrupts. */
  99 static inline void play_dead(void)
 100 {
 101         /* This must be done before dead CPU ack */
 102         cpu_exit_clear();
 103         mb();
 104         /* Ack it */
 105         __get_cpu_var(cpu_state) = CPU_DEAD;
 106
 107         /*
 108          * With physical CPU hotplug, we should halt the cpu
 109          */
 110         local_irq_disable();
 111         /* mask all interrupts, flush any and all caches, and halt */
 112         wbinvd_halt();
 113 }
 114 #else
 115 static inline void play_dead(void)
 116 {
 117         BUG();
 118 }
 119 #endif /* CONFIG_HOTPLUG_CPU */
 120
 121 /*
 122  * The idle thread. There's no useful work to be
 123  * done, so just try to conserve power and have a
 124  * low exit latency (ie sit in a loop waiting for
 125  * somebody to say that they'd like to reschedule)
 126  */
 127 void cpu_idle(void)
 128 {
 129         int cpu = smp_processor_id();
 130
 131         current_thread_info()->status |= TS_POLLING;
 132
 133         /* endless idle loop with no priority at all */
 134         while (1) {
 135                 tick_nohz_stop_sched_tick(1);
 136                 while (!need_resched()) {
 137
 138                         check_pgt_cache();
 139                         rmb();
 140
 141                         if (rcu_pending(cpu))
 142                                 rcu_check_callbacks(cpu, 0);
 143
 144                         if (cpu_is_offline(cpu))
 145                                 play_dead();
 146
 147                         local_irq_disable();
 148                         __get_cpu_var(irq_stat).idle_timestamp = jiffies;
 149                         /* Don't trace irqs off for idle */
 150                         stop_critical_timings();
 151                         pm_idle();
 152                         start_critical_timings();
 153                 }
 154                 tick_nohz_restart_sched_tick();
 155                 preempt_enable_no_resched();
 156                 schedule();
 157                 preempt_disable();
 158         }
 159 }
 160
 161 void __show_registers(struct pt_regs *regs, int all)
 162 {
 163         unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
 164         unsigned long d0, d1, d2, d3, d6, d7;
 165         unsigned long sp;
 166         unsigned short ss, gs;
 167         const char *board;
 168
 169         if (user_mode_vm(regs)) {
 170                 sp = regs->sp;
 171                 ss = regs->ss & 0xffff;
 172                 savesegment(gs, gs);
 173         } else {
 174                 sp = (unsigned long) (&regs->sp);
 175                 savesegment(ss, ss);
 176                 savesegment(gs, gs);
 177         }
 178
 179         printk("\n");
 180
 181         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 182         if (!board)
 183                 board = "";
 184         printk("Pid: %d, comm: %s %s (%s %.*s) %s\n",
 185                         task_pid_nr(current), current->comm,
 186                         print_tainted(), init_utsname()->release,
 187                         (int)strcspn(init_utsname()->version, " "),
 188                         init_utsname()->version, board);
 189
 190         printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
 191                         (u16)regs->cs, regs->ip, regs->flags,
 192                         smp_processor_id());
 193         print_symbol("EIP is at %s\n", regs->ip);
 194
 195         printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
 196                 regs->ax, regs->bx, regs->cx, regs->dx);
 197         printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
 198                 regs->si, regs->di, regs->bp, sp);
 199         printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
 200                (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
 201
 202         if (!all)
 203                 return;
 204
 205         cr0 = read_cr0();
 206         cr2 = read_cr2();
 207         cr3 = read_cr3();
 208         cr4 = read_cr4_safe();
 209         printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
 210                         cr0, cr2, cr3, cr4);
 211
 212         get_debugreg(d0, 0);
 213         get_debugreg(d1, 1);
 214         get_debugreg(d2, 2);
 215         get_debugreg(d3, 3);
 216         printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
 217                         d0, d1, d2, d3);
 218
 219         get_debugreg(d6, 6);
 220         get_debugreg(d7, 7);
 221         printk("DR6: %08lx DR7: %08lx\n",
 222                         d6, d7);
 223 }
 224
 225 void show_regs(struct pt_regs *regs)
 226 {
 227         __show_registers(regs, 1);
 228         show_trace(NULL, regs, &regs->sp, regs->bp);
 229 }
 230
 231 /*
 232  * This gets run with %bx containing the
 233  * function to call, and %dx containing
 234  * the "args".
 235  */
 236 extern void kernel_thread_helper(void);
 237
 238 /*
 239  * Create a kernel thread
 240  */
 241 int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
 242 {
 243         struct pt_regs regs;
 244
 245         memset(&regs, 0, sizeof(regs));
 246
 247         regs.bx = (unsigned long) fn;
 248         regs.dx = (unsigned long) arg;
 249
 250         regs.ds = __USER_DS;
 251         regs.es = __USER_DS;
 252         regs.fs = __KERNEL_PERCPU;
 253         regs.orig_ax = -1;
 254         regs.ip = (unsigned long) kernel_thread_helper;
 255         regs.cs = __KERNEL_CS | get_kernel_rpl();
 256         regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2;
 257
 258         /* Ok, create the new process.. */
 259         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
 260 }
 261 EXPORT_SYMBOL(kernel_thread);
 262
 263 /*
 264  * Free current thread data structures etc..
 265  */
 266 void exit_thread(void)
 267 {
 268         /* The process may have allocated an io port bitmap... nuke it. */
 269         if (unlikely(test_thread_flag(TIF_IO_BITMAP))) {
 270                 struct task_struct *tsk = current;
 271                 struct thread_struct *t = &tsk->thread;
 272                 int cpu = get_cpu();
 273                 struct tss_struct *tss = &per_cpu(init_tss, cpu);
 274
 275                 kfree(t->io_bitmap_ptr);
 276                 t->io_bitmap_ptr = NULL;
 277                 clear_thread_flag(TIF_IO_BITMAP);
 278                 /*
 279                  * Careful, clear this in the TSS too:
 280                  */
 281                 memset(tss->io_bitmap, 0xff, tss->io_bitmap_max);
 282                 t->io_bitmap_max = 0;
 283                 tss->io_bitmap_owner = NULL;
 284                 tss->io_bitmap_max = 0;
 285                 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
 286                 put_cpu();
 287         }
 288 #ifdef CONFIG_X86_DS
 289         /* Free any DS contexts that have not been properly released. */
 290         if (unlikely(current->thread.ds_ctx)) {
 291                 /* we clear debugctl to make sure DS is not used. */
 292                 update_debugctlmsr(0);
 293                 ds_free(current->thread.ds_ctx);
 294         }
 295 #endif /* CONFIG_X86_DS */
 296 }
 297
 298 void flush_thread(void)
 299 {
 300         struct task_struct *tsk = current;
 301
 302         tsk->thread.debugreg0 = 0;
 303         tsk->thread.debugreg1 = 0;
 304         tsk->thread.debugreg2 = 0;
 305         tsk->thread.debugreg3 = 0;
 306         tsk->thread.debugreg6 = 0;
 307         tsk->thread.debugreg7 = 0;
 308         memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array));
 309         clear_tsk_thread_flag(tsk, TIF_DEBUG);
 310         /*
 311          * Forget coprocessor state..
 312          */
 313         tsk->fpu_counter = 0;
 314         clear_fpu(tsk);
 315         clear_used_math();
 316 }
 317
 318 void release_thread(struct task_struct *dead_task)
 319 {
 320         BUG_ON(dead_task->mm);
 321         release_vm86_irqs(dead_task);
 322 }
 323
 324 /*
 325  * This gets called before we allocate a new thread and copy
 326  * the current task into it.
 327  */
 328 void prepare_to_copy(struct task_struct *tsk)
 329 {
 330         unlazy_fpu(tsk);
 331 }
 332
 333 int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 334         unsigned long unused,
 335         struct task_struct * p, struct pt_regs * regs)
 336 {
 337         struct pt_regs * childregs;
 338         struct task_struct *tsk;
 339         int err;
 340
 341         childregs = task_pt_regs(p);
 342         *childregs = *regs;
 343         childregs->ax = 0;
 344         childregs->sp = sp;
 345
 346         p->thread.sp = (unsigned long) childregs;
 347         p->thread.sp0 = (unsigned long) (childregs+1);
 348
 349         p->thread.ip = (unsigned long) ret_from_fork;
 350
 351         savesegment(gs, p->thread.gs);
 352
 353         tsk = current;
 354         if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
 355                 p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
 356                                                 IO_BITMAP_BYTES, GFP_KERNEL);
 357                 if (!p->thread.io_bitmap_ptr) {
 358                         p->thread.io_bitmap_max = 0;
 359                         return -ENOMEM;
 360                 }
 361                 set_tsk_thread_flag(p, TIF_IO_BITMAP);
 362         }
 363
 364         err = 0;
 365
 366         /*
 367          * Set a new TLS for the child thread?
 368          */
 369         if (clone_flags & CLONE_SETTLS)
 370                 err = do_set_thread_area(p, -1,
 371                         (struct user_desc __user *)childregs->si, 0);
 372
 373         if (err && p->thread.io_bitmap_ptr) {
 374                 kfree(p->thread.io_bitmap_ptr);
 375                 p->thread.io_bitmap_max = 0;
 376         }
 377         return err;
 378 }
 379
 380 void
 381 start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
 382 {
 383         __asm__("movl %0, %%gs" :: "r"(0));
 384         regs->fs                = 0;
 385         set_fs(USER_DS);
 386         regs->ds                = __USER_DS;
 387         regs->es                = __USER_DS;
 388         regs->ss                = __USER_DS;
 389         regs->cs                = __USER_CS;
 390         regs->ip                = new_ip;
 391         regs->sp                = new_sp;
 392         /*
 393          * Free the old FP and other extended state
 394          */
 395         free_thread_xstate(current);
 396 }
 397 EXPORT_SYMBOL_GPL(start_thread);
 398
 399 static void hard_disable_TSC(void)
 400 {
 401         write_cr4(read_cr4() | X86_CR4_TSD);
 402 }
 403
 404 void disable_TSC(void)
 405 {
 406         preempt_disable();
 407         if (!test_and_set_thread_flag(TIF_NOTSC))
 408                 /*
 409                  * Must flip the CPU state synchronously with
 410                  * TIF_NOTSC in the current running context.
 411                  */
 412                 hard_disable_TSC();
 413         preempt_enable();
 414 }
 415
 416 static void hard_enable_TSC(void)
 417 {
 418         write_cr4(read_cr4() & ~X86_CR4_TSD);
 419 }
 420
 421 static void enable_TSC(void)
 422 {
 423         preempt_disable();
 424         if (test_and_clear_thread_flag(TIF_NOTSC))
 425                 /*
 426                  * Must flip the CPU state synchronously with
 427                  * TIF_NOTSC in the current running context.
 428                  */
 429                 hard_enable_TSC();
 430         preempt_enable();
 431 }
 432
 433 int get_tsc_mode(unsigned long adr)
 434 {
 435         unsigned int val;
 436
 437         if (test_thread_flag(TIF_NOTSC))
 438                 val = PR_TSC_SIGSEGV;
 439         else
 440                 val = PR_TSC_ENABLE;
 441
 442         return put_user(val, (unsigned int __user *)adr);
 443 }
 444
 445 int set_tsc_mode(unsigned int val)
 446 {
 447         if (val == PR_TSC_SIGSEGV)
 448                 disable_TSC();
 449         else if (val == PR_TSC_ENABLE)
 450                 enable_TSC();
 451         else
 452                 return -EINVAL;
 453
 454         return 0;
 455 }
 456
 457 #ifdef CONFIG_X86_DS
 458 static int update_debugctl(struct thread_struct *prev,
 459                         struct thread_struct *next, unsigned long debugctl)
 460 {
 461         unsigned long ds_prev = 0;
 462         unsigned long ds_next = 0;
 463
 464         if (prev->ds_ctx)
 465                 ds_prev = (unsigned long)prev->ds_ctx->ds;
 466         if (next->ds_ctx)
 467                 ds_next = (unsigned long)next->ds_ctx->ds;
 468
 469         if (ds_next != ds_prev) {
 470                 /* we clear debugctl to make sure DS
 471                  * is not in use when we change it */
 472                 debugctl = 0;
 473                 update_debugctlmsr(0);
 474                 wrmsr(MSR_IA32_DS_AREA, ds_next, 0);
 475         }
 476         return debugctl;
 477 }
 478 #else
 479 static int update_debugctl(struct thread_struct *prev,
 480                         struct thread_struct *next, unsigned long debugctl)
 481 {
 482         return debugctl;
 483 }
 484 #endif /* CONFIG_X86_DS */
 485
 486 static noinline void
 487 __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 488                  struct tss_struct *tss)
 489 {
 490         struct thread_struct *prev, *next;
 491         unsigned long debugctl;
 492
 493         prev = &prev_p->thread;
 494         next = &next_p->thread;
 495
 496         debugctl = update_debugctl(prev, next, prev->debugctlmsr);
 497
 498         if (next->debugctlmsr != debugctl)
 499                 update_debugctlmsr(next->debugctlmsr);
 500
 501         if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
 502                 set_debugreg(next->debugreg0, 0);
 503                 set_debugreg(next->debugreg1, 1);
 504                 set_debugreg(next->debugreg2, 2);
 505                 set_debugreg(next->debugreg3, 3);
 506                 /* no 4 and 5 */
 507                 set_debugreg(next->debugreg6, 6);
 508                 set_debugreg(next->debugreg7, 7);
 509         }
 510
 511         if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^
 512             test_tsk_thread_flag(next_p, TIF_NOTSC)) {
 513                 /* prev and next are different */
 514                 if (test_tsk_thread_flag(next_p, TIF_NOTSC))
 515                         hard_disable_TSC();
 516                 else
 517                         hard_enable_TSC();
 518         }
 519
 520 #ifdef CONFIG_X86_PTRACE_BTS
 521         if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS))
 522                 ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS);
 523
 524         if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS))
 525                 ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES);
 526 #endif /* CONFIG_X86_PTRACE_BTS */
 527
 528
 529         if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) {
 530                 /*
 531                  * Disable the bitmap via an invalid offset. We still cache
 532                  * the previous bitmap owner and the IO bitmap contents:
 533                  */
 534                 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
 535                 return;
 536         }
 537
 538         if (likely(next == tss->io_bitmap_owner)) {
 539                 /*
 540                  * Previous owner of the bitmap (hence the bitmap content)
 541                  * matches the next task, we dont have to do anything but
 542                  * to set a valid offset in the TSS:
 543                  */
 544                 tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET;
 545                 return;
 546         }
 547         /*
 548          * Lazy TSS's I/O bitmap copy. We set an invalid offset here
 549          * and we let the task to get a GPF in case an I/O instruction
 550          * is performed.  The handler of the GPF will verify that the
 551          * faulting task has a valid I/O bitmap and, it true, does the
 552          * real copy and restart the instruction.  This will save us
 553          * redundant copies when the currently switched task does not
 554          * perform any I/O during its timeslice.
 555          */
 556         tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY;
 557 }
 558
 559 /*
 560  *      switch_to(x,yn) should switch tasks from x to y.
 561  *
 562  * We fsave/fwait so that an exception goes off at the right time
 563  * (as a call from the fsave or fwait in effect) rather than to
 564  * the wrong process. Lazy FP saving no longer makes any sense
 565  * with modern CPU's, and this simplifies a lot of things (SMP
 566  * and UP become the same).
 567  *
 568  * NOTE! We used to use the x86 hardware context switching. The
 569  * reason for not using it any more becomes apparent when you
 570  * try to recover gracefully from saved state that is no longer
 571  * valid (stale segment register values in particular). With the
 572  * hardware task-switch, there is no way to fix up bad state in
 573  * a reasonable manner.
 574  *
 575  * The fact that Intel documents the hardware task-switching to
 576  * be slow is a fairly red herring - this code is not noticeably
 577  * faster. However, there _is_ some room for improvement here,
 578  * so the performance issues may eventually be a valid point.
 579  * More important, however, is the fact that this allows us much
 580  * more flexibility.
 581  *
 582  * The return value (in %ax) will be the "prev" task after
 583  * the task-switch, and shows up in ret_from_fork in entry.S,
 584  * for example.
 585  */
 586 struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 587 {
 588         struct thread_struct *prev = &prev_p->thread,
 589                                  *next = &next_p->thread;
 590         int cpu = smp_processor_id();
 591         struct tss_struct *tss = &per_cpu(init_tss, cpu);
 592
 593         /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
 594
 595         __unlazy_fpu(prev_p);
 596
 597
 598         /* we're going to use this soon, after a few expensive things */
 599         if (next_p->fpu_counter > 5)
 600                 prefetch(next->xstate);
 601
 602         /*
 603          * Reload esp0.
 604          */
 605         load_sp0(tss, next);
 606
 607         /*
 608          * Save away %gs. No need to save %fs, as it was saved on the
 609          * stack on entry.  No need to save %es and %ds, as those are
 610          * always kernel segments while inside the kernel.  Doing this
 611          * before setting the new TLS descriptors avoids the situation
 612          * where we temporarily have non-reloadable segments in %fs
 613          * and %gs.  This could be an issue if the NMI handler ever
 614          * used %fs or %gs (it does not today), or if the kernel is
 615          * running inside of a hypervisor layer.
 616          */
 617         savesegment(gs, prev->gs);
 618
 619         /*
 620          * Load the per-thread Thread-Local Storage descriptor.
 621          */
 622         load_TLS(next, cpu);
 623
 624         /*
 625          * Restore IOPL if needed.  In normal use, the flags restore
 626          * in the switch assembly will handle this.  But if the kernel
 627          * is running virtualized at a non-zero CPL, the popf will
 628          * not restore flags, so it must be done in a separate step.
 629          */
 630         if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
 631                 set_iopl_mask(next->iopl);
 632
 633         /*
 634          * Now maybe handle debug registers and/or IO bitmaps
 635          */
 636         if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
 637                      task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
 638                 __switch_to_xtra(prev_p, next_p, tss);
 639
 640         /*
 641          * Leave lazy mode, flushing any hypercalls made here.
 642          * This must be done before restoring TLS segments so
 643          * the GDT and LDT are properly updated, and must be
 644          * done before math_state_restore, so the TS bit is up
 645          * to date.
 646          */
 647         arch_leave_lazy_cpu_mode();
 648
 649         /* If the task has used fpu the last 5 timeslices, just do a full
 650          * restore of the math state immediately to avoid the trap; the
 651          * chances of needing FPU soon are obviously high now
 652          *
 653          * tsk_used_math() checks prevent calling math_state_restore(),
 654          * which can sleep in the case of !tsk_used_math()
 655          */
 656         if (tsk_used_math(next_p) && next_p->fpu_counter > 5)
 657                 math_state_restore();
 658
 659         /*
 660          * Restore %gs if needed (which is common)
 661          */
 662         if (prev->gs | next->gs)
 663                 loadsegment(gs, next->gs);
 664
 665         x86_write_percpu(current_task, next_p);
 666
 667         return prev_p;
 668 }
 669
 670 asmlinkage int sys_fork(struct pt_regs regs)
 671 {
 672         return do_fork(SIGCHLD, regs.sp, &regs, 0, NULL, NULL);
 673 }
 674
 675 asmlinkage int sys_clone(struct pt_regs regs)
 676 {
 677         unsigned long clone_flags;
 678         unsigned long newsp;
 679         int __user *parent_tidptr, *child_tidptr;
 680
 681         clone_flags = regs.bx;
 682         newsp = regs.cx;
 683         parent_tidptr = (int __user *)regs.dx;
 684         child_tidptr = (int __user *)regs.di;
 685         if (!newsp)
 686                 newsp = regs.sp;
 687         return do_fork(clone_flags, newsp, &regs, 0, parent_tidptr, child_tidptr);
 688 }
 689
 690 /*
 691  * This is trivial, and on the face of it looks like it
 692  * could equally well be done in user mode.
 693  *
 694  * Not so, for quite unobvious reasons - register pressure.
 695  * In user mode vfork() cannot have a stack frame, and if
 696  * done by calling the "clone()" system call directly, you
 697  * do not have enough call-clobbered registers to hold all
 698  * the information you need.
 699  */
 700 asmlinkage int sys_vfork(struct pt_regs regs)
 701 {
 702         return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, &regs, 0, NULL, NULL);
 703 }
 704
 705 /*
 706  * sys_execve() executes a new program.
 707  */
 708 asmlinkage int sys_execve(struct pt_regs regs)
 709 {
 710         int error;
 711         char * filename;
 712
 713         filename = getname((char __user *) regs.bx);
 714         error = PTR_ERR(filename);
 715         if (IS_ERR(filename))
 716                 goto out;
 717         error = do_execve(filename,
 718                         (char __user * __user *) regs.cx,
 719                         (char __user * __user *) regs.dx,
 720                         &regs);
 721         if (error == 0) {
 722                 /* Make sure we don't return using sysenter.. */
 723                 set_thread_flag(TIF_IRET);
 724         }
 725         putname(filename);
 726 out:
 727         return error;
 728 }
 729
 730 #define top_esp                (THREAD_SIZE - sizeof(unsigned long))
 731 #define top_ebp                (THREAD_SIZE - 2*sizeof(unsigned long))
 732
 733 unsigned long get_wchan(struct task_struct *p)
 734 {
 735         unsigned long bp, sp, ip;
 736         unsigned long stack_page;
 737         int count = 0;
 738         if (!p || p == current || p->state == TASK_RUNNING)
 739                 return 0;
 740         stack_page = (unsigned long)task_stack_page(p);
 741         sp = p->thread.sp;
 742         if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
 743                 return 0;
 744         /* include/asm-i386/system.h:switch_to() pushes bp last. */
 745         bp = *(unsigned long *) sp;
 746         do {
 747                 if (bp < stack_page || bp > top_ebp+stack_page)
 748                         return 0;
 749                 ip = *(unsigned long *) (bp+4);
 750                 if (!in_sched_functions(ip))
 751                         return ip;
 752                 bp = *(unsigned long *) bp;
 753         } while (count++ < 16);
 754         return 0;
 755 }
 756
 757 unsigned long arch_align_stack(unsigned long sp)
 758 {
 759         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 760                 sp -= get_random_int() % 8192;
 761         return sp & ~0xf;
 762 }
 763
 764 unsigned long arch_randomize_brk(struct mm_struct *mm)
 765 {
 766         unsigned long range_end = mm->brk + 0x02000000;
 767         return randomize_range(mm->brk, range_end, 0) ? : mm->brk;
 768 }