linux/arch/x86/kernel/process_32.c
<<
>>
Prefs
   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 <linux/cpu.h>
  13#include <linux/errno.h>
  14#include <linux/sched.h>
  15#include <linux/fs.h>
  16#include <linux/kernel.h>
  17#include <linux/mm.h>
  18#include <linux/elfcore.h>
  19#include <linux/smp.h>
  20#include <linux/stddef.h>
  21#include <linux/slab.h>
  22#include <linux/vmalloc.h>
  23#include <linux/user.h>
  24#include <linux/interrupt.h>
  25#include <linux/delay.h>
  26#include <linux/reboot.h>
  27#include <linux/init.h>
  28#include <linux/mc146818rtc.h>
  29#include <linux/module.h>
  30#include <linux/kallsyms.h>
  31#include <linux/ptrace.h>
  32#include <linux/personality.h>
  33#include <linux/percpu.h>
  34#include <linux/prctl.h>
  35#include <linux/ftrace.h>
  36#include <linux/uaccess.h>
  37#include <linux/io.h>
  38#include <linux/kdebug.h>
  39
  40#include <asm/pgtable.h>
  41#include <asm/ldt.h>
  42#include <asm/processor.h>
  43#include <asm/i387.h>
  44#include <asm/fpu-internal.h>
  45#include <asm/desc.h>
  46#ifdef CONFIG_MATH_EMULATION
  47#include <asm/math_emu.h>
  48#endif
  49
  50#include <linux/err.h>
  51
  52#include <asm/tlbflush.h>
  53#include <asm/cpu.h>
  54#include <asm/idle.h>
  55#include <asm/syscalls.h>
  56#include <asm/debugreg.h>
  57#include <asm/switch_to.h>
  58
  59asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
  60asmlinkage void ret_from_kernel_thread(void) __asm__("ret_from_kernel_thread");
  61
  62/*
  63 * Return saved PC of a blocked thread.
  64 */
  65unsigned long thread_saved_pc(struct task_struct *tsk)
  66{
  67        return ((unsigned long *)tsk->thread.sp)[3];
  68}
  69
  70void __show_regs(struct pt_regs *regs, int all)
  71{
  72        unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
  73        unsigned long d0, d1, d2, d3, d6, d7;
  74        unsigned long sp;
  75        unsigned short ss, gs;
  76
  77        if (user_mode_vm(regs)) {
  78                sp = regs->sp;
  79                ss = regs->ss & 0xffff;
  80                gs = get_user_gs(regs);
  81        } else {
  82                sp = kernel_stack_pointer(regs);
  83                savesegment(ss, ss);
  84                savesegment(gs, gs);
  85        }
  86
  87        show_regs_common();
  88
  89        printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
  90                        (u16)regs->cs, regs->ip, regs->flags,
  91                        smp_processor_id());
  92        print_symbol("EIP is at %s\n", regs->ip);
  93
  94        printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
  95                regs->ax, regs->bx, regs->cx, regs->dx);
  96        printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
  97                regs->si, regs->di, regs->bp, sp);
  98        printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
  99               (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
 100
 101        if (!all)
 102                return;
 103
 104        cr0 = read_cr0();
 105        cr2 = read_cr2();
 106        cr3 = read_cr3();
 107        cr4 = read_cr4_safe();
 108        printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
 109                        cr0, cr2, cr3, cr4);
 110
 111        get_debugreg(d0, 0);
 112        get_debugreg(d1, 1);
 113        get_debugreg(d2, 2);
 114        get_debugreg(d3, 3);
 115        printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
 116                        d0, d1, d2, d3);
 117
 118        get_debugreg(d6, 6);
 119        get_debugreg(d7, 7);
 120        printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
 121                        d6, d7);
 122}
 123
 124void release_thread(struct task_struct *dead_task)
 125{
 126        BUG_ON(dead_task->mm);
 127        release_vm86_irqs(dead_task);
 128}
 129
 130int copy_thread(unsigned long clone_flags, unsigned long sp,
 131        unsigned long arg, struct task_struct *p)
 132{
 133        struct pt_regs *childregs = task_pt_regs(p);
 134        struct task_struct *tsk;
 135        int err;
 136
 137        p->thread.sp = (unsigned long) childregs;
 138        p->thread.sp0 = (unsigned long) (childregs+1);
 139
 140        if (unlikely(p->flags & PF_KTHREAD)) {
 141                /* kernel thread */
 142                memset(childregs, 0, sizeof(struct pt_regs));
 143                p->thread.ip = (unsigned long) ret_from_kernel_thread;
 144                task_user_gs(p) = __KERNEL_STACK_CANARY;
 145                childregs->ds = __USER_DS;
 146                childregs->es = __USER_DS;
 147                childregs->fs = __KERNEL_PERCPU;
 148                childregs->bx = sp;     /* function */
 149                childregs->bp = arg;
 150                childregs->orig_ax = -1;
 151                childregs->cs = __KERNEL_CS | get_kernel_rpl();
 152                childregs->flags = X86_EFLAGS_IF | X86_EFLAGS_BIT1;
 153                p->fpu_counter = 0;
 154                p->thread.io_bitmap_ptr = NULL;
 155                memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
 156                return 0;
 157        }
 158        *childregs = *current_pt_regs();
 159        childregs->ax = 0;
 160        if (sp)
 161                childregs->sp = sp;
 162
 163        p->thread.ip = (unsigned long) ret_from_fork;
 164        task_user_gs(p) = get_user_gs(current_pt_regs());
 165
 166        p->fpu_counter = 0;
 167        p->thread.io_bitmap_ptr = NULL;
 168        tsk = current;
 169        err = -ENOMEM;
 170
 171        memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
 172
 173        if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
 174                p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
 175                                                IO_BITMAP_BYTES, GFP_KERNEL);
 176                if (!p->thread.io_bitmap_ptr) {
 177                        p->thread.io_bitmap_max = 0;
 178                        return -ENOMEM;
 179                }
 180                set_tsk_thread_flag(p, TIF_IO_BITMAP);
 181        }
 182
 183        err = 0;
 184
 185        /*
 186         * Set a new TLS for the child thread?
 187         */
 188        if (clone_flags & CLONE_SETTLS)
 189                err = do_set_thread_area(p, -1,
 190                        (struct user_desc __user *)childregs->si, 0);
 191
 192        if (err && p->thread.io_bitmap_ptr) {
 193                kfree(p->thread.io_bitmap_ptr);
 194                p->thread.io_bitmap_max = 0;
 195        }
 196        return err;
 197}
 198
 199void
 200start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
 201{
 202        set_user_gs(regs, 0);
 203        regs->fs                = 0;
 204        regs->ds                = __USER_DS;
 205        regs->es                = __USER_DS;
 206        regs->ss                = __USER_DS;
 207        regs->cs                = __USER_CS;
 208        regs->ip                = new_ip;
 209        regs->sp                = new_sp;
 210        regs->flags             = X86_EFLAGS_IF;
 211        /*
 212         * force it to the iret return path by making it look as if there was
 213         * some work pending.
 214         */
 215        set_thread_flag(TIF_NOTIFY_RESUME);
 216}
 217EXPORT_SYMBOL_GPL(start_thread);
 218
 219
 220/*
 221 *      switch_to(x,y) should switch tasks from x to y.
 222 *
 223 * We fsave/fwait so that an exception goes off at the right time
 224 * (as a call from the fsave or fwait in effect) rather than to
 225 * the wrong process. Lazy FP saving no longer makes any sense
 226 * with modern CPU's, and this simplifies a lot of things (SMP
 227 * and UP become the same).
 228 *
 229 * NOTE! We used to use the x86 hardware context switching. The
 230 * reason for not using it any more becomes apparent when you
 231 * try to recover gracefully from saved state that is no longer
 232 * valid (stale segment register values in particular). With the
 233 * hardware task-switch, there is no way to fix up bad state in
 234 * a reasonable manner.
 235 *
 236 * The fact that Intel documents the hardware task-switching to
 237 * be slow is a fairly red herring - this code is not noticeably
 238 * faster. However, there _is_ some room for improvement here,
 239 * so the performance issues may eventually be a valid point.
 240 * More important, however, is the fact that this allows us much
 241 * more flexibility.
 242 *
 243 * The return value (in %ax) will be the "prev" task after
 244 * the task-switch, and shows up in ret_from_fork in entry.S,
 245 * for example.
 246 */
 247__notrace_funcgraph struct task_struct *
 248__switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 249{
 250        struct thread_struct *prev = &prev_p->thread,
 251                                 *next = &next_p->thread;
 252        int cpu = smp_processor_id();
 253        struct tss_struct *tss = &per_cpu(init_tss, cpu);
 254        fpu_switch_t fpu;
 255
 256        /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
 257
 258        fpu = switch_fpu_prepare(prev_p, next_p, cpu);
 259
 260        /*
 261         * Reload esp0.
 262         */
 263        load_sp0(tss, next);
 264
 265        /*
 266         * Save away %gs. No need to save %fs, as it was saved on the
 267         * stack on entry.  No need to save %es and %ds, as those are
 268         * always kernel segments while inside the kernel.  Doing this
 269         * before setting the new TLS descriptors avoids the situation
 270         * where we temporarily have non-reloadable segments in %fs
 271         * and %gs.  This could be an issue if the NMI handler ever
 272         * used %fs or %gs (it does not today), or if the kernel is
 273         * running inside of a hypervisor layer.
 274         */
 275        lazy_save_gs(prev->gs);
 276
 277        /*
 278         * Load the per-thread Thread-Local Storage descriptor.
 279         */
 280        load_TLS(next, cpu);
 281
 282        /*
 283         * Restore IOPL if needed.  In normal use, the flags restore
 284         * in the switch assembly will handle this.  But if the kernel
 285         * is running virtualized at a non-zero CPL, the popf will
 286         * not restore flags, so it must be done in a separate step.
 287         */
 288        if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
 289                set_iopl_mask(next->iopl);
 290
 291        /*
 292         * Now maybe handle debug registers and/or IO bitmaps
 293         */
 294        if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
 295                     task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
 296                __switch_to_xtra(prev_p, next_p, tss);
 297
 298        /*
 299         * Leave lazy mode, flushing any hypercalls made here.
 300         * This must be done before restoring TLS segments so
 301         * the GDT and LDT are properly updated, and must be
 302         * done before math_state_restore, so the TS bit is up
 303         * to date.
 304         */
 305        arch_end_context_switch(next_p);
 306
 307        /*
 308         * Restore %gs if needed (which is common)
 309         */
 310        if (prev->gs | next->gs)
 311                lazy_load_gs(next->gs);
 312
 313        switch_fpu_finish(next_p, fpu);
 314
 315        this_cpu_write(current_task, next_p);
 316
 317        return prev_p;
 318}
 319
 320#define top_esp                (THREAD_SIZE - sizeof(unsigned long))
 321#define top_ebp                (THREAD_SIZE - 2*sizeof(unsigned long))
 322
 323unsigned long get_wchan(struct task_struct *p)
 324{
 325        unsigned long bp, sp, ip;
 326        unsigned long stack_page;
 327        int count = 0;
 328        if (!p || p == current || p->state == TASK_RUNNING)
 329                return 0;
 330        stack_page = (unsigned long)task_stack_page(p);
 331        sp = p->thread.sp;
 332        if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
 333                return 0;
 334        /* include/asm-i386/system.h:switch_to() pushes bp last. */
 335        bp = *(unsigned long *) sp;
 336        do {
 337                if (bp < stack_page || bp > top_ebp+stack_page)
 338                        return 0;
 339                ip = *(unsigned long *) (bp+4);
 340                if (!in_sched_functions(ip))
 341                        return ip;
 342                bp = *(unsigned long *) bp;
 343        } while (count++ < 16);
 344        return 0;
 345}
 346
 347
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.