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 <linux/cpu.h>
  13 #include <linux/errno.h>
  14 #include <linux/sched.h>
  15 #include <linux/sched/task.h>
  16 #include <linux/sched/task_stack.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/delay.h>
  28 #include <linux/reboot.h>
  29 #include <linux/mc146818rtc.h>
  30 #include <linux/export.h>
  31 #include <linux/kallsyms.h>
  32 #include <linux/ptrace.h>
  33 #include <linux/personality.h>
  34 #include <linux/percpu.h>
  35 #include <linux/prctl.h>
  36 #include <linux/ftrace.h>
  37 #include <linux/uaccess.h>
  38 #include <linux/io.h>
  39 #include <linux/kdebug.h>
  40 #include <linux/syscalls.h>
  41
  42 #include <asm/pgtable.h>
  43 #include <asm/ldt.h>
  44 #include <asm/processor.h>
  45 #include <asm/fpu/internal.h>
  46 #include <asm/desc.h>
  47
  48 #include <linux/err.h>
  49
  50 #include <asm/tlbflush.h>
  51 #include <asm/cpu.h>
  52 #include <asm/syscalls.h>
  53 #include <asm/debugreg.h>
  54 #include <asm/switch_to.h>
  55 #include <asm/vm86.h>
  56 #include <asm/resctrl_sched.h>
  57 #include <asm/proto.h>
  58
  59 #include "process.h"
  60
  61 void __show_regs(struct pt_regs *regs, enum show_regs_mode mode)
  62 {
  63         unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
  64         unsigned long d0, d1, d2, d3, d6, d7;
  65         unsigned long sp;
  66         unsigned short ss, gs;
  67
  68         if (user_mode(regs)) {
  69                 sp = regs->sp;
  70                 ss = regs->ss;
  71                 gs = get_user_gs(regs);
  72         } else {
  73                 sp = kernel_stack_pointer(regs);
  74                 savesegment(ss, ss);
  75                 savesegment(gs, gs);
  76         }
  77
  78         show_ip(regs, KERN_DEFAULT);
  79
  80         printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
  81                 regs->ax, regs->bx, regs->cx, regs->dx);
  82         printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
  83                 regs->si, regs->di, regs->bp, sp);
  84         printk(KERN_DEFAULT "DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x EFLAGS: %08lx\n",
  85                (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss, regs->flags);
  86
  87         if (mode != SHOW_REGS_ALL)
  88                 return;
  89
  90         cr0 = read_cr0();
  91         cr2 = read_cr2();
  92         cr3 = __read_cr3();
  93         cr4 = __read_cr4();
  94         printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
  95                         cr0, cr2, cr3, cr4);
  96
  97         get_debugreg(d0, 0);
  98         get_debugreg(d1, 1);
  99         get_debugreg(d2, 2);
 100         get_debugreg(d3, 3);
 101         get_debugreg(d6, 6);
 102         get_debugreg(d7, 7);
 103
 104         /* Only print out debug registers if they are in their non-default state. */
 105         if ((d0 == 0) && (d1 == 0) && (d2 == 0) && (d3 == 0) &&
 106             (d6 == DR6_RESERVED) && (d7 == 0x400))
 107                 return;
 108
 109         printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
 110                         d0, d1, d2, d3);
 111         printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
 112                         d6, d7);
 113 }
 114
 115 void release_thread(struct task_struct *dead_task)
 116 {
 117         BUG_ON(dead_task->mm);
 118         release_vm86_irqs(dead_task);
 119 }
 120
 121 int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
 122         unsigned long arg, struct task_struct *p, unsigned long tls)
 123 {
 124         struct pt_regs *childregs = task_pt_regs(p);
 125         struct fork_frame *fork_frame = container_of(childregs, struct fork_frame, regs);
 126         struct inactive_task_frame *frame = &fork_frame->frame;
 127         struct task_struct *tsk;
 128         int err;
 129
 130         frame->bp = 0;
 131         frame->ret_addr = (unsigned long) ret_from_fork;
 132         p->thread.sp = (unsigned long) fork_frame;
 133         p->thread.sp0 = (unsigned long) (childregs+1);
 134         memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
 135
 136         if (unlikely(p->flags & PF_KTHREAD)) {
 137                 /* kernel thread */
 138                 memset(childregs, 0, sizeof(struct pt_regs));
 139                 frame->bx = sp;         /* function */
 140                 frame->di = arg;
 141                 p->thread.io_bitmap_ptr = NULL;
 142                 return 0;
 143         }
 144         frame->bx = 0;
 145         *childregs = *current_pt_regs();
 146         childregs->ax = 0;
 147         if (sp)
 148                 childregs->sp = sp;
 149
 150         task_user_gs(p) = get_user_gs(current_pt_regs());
 151
 152         p->thread.io_bitmap_ptr = NULL;
 153         tsk = current;
 154         err = -ENOMEM;
 155
 156         if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
 157                 p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
 158                                                 IO_BITMAP_BYTES, GFP_KERNEL);
 159                 if (!p->thread.io_bitmap_ptr) {
 160                         p->thread.io_bitmap_max = 0;
 161                         return -ENOMEM;
 162                 }
 163                 set_tsk_thread_flag(p, TIF_IO_BITMAP);
 164         }
 165
 166         err = 0;
 167
 168         /*
 169          * Set a new TLS for the child thread?
 170          */
 171         if (clone_flags & CLONE_SETTLS)
 172                 err = do_set_thread_area(p, -1,
 173                         (struct user_desc __user *)tls, 0);
 174
 175         if (err && p->thread.io_bitmap_ptr) {
 176                 kfree(p->thread.io_bitmap_ptr);
 177                 p->thread.io_bitmap_max = 0;
 178         }
 179         return err;
 180 }
 181
 182 void
 183 start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
 184 {
 185         set_user_gs(regs, 0);
 186         regs->fs                = 0;
 187         regs->ds                = __USER_DS;
 188         regs->es                = __USER_DS;
 189         regs->ss                = __USER_DS;
 190         regs->cs                = __USER_CS;
 191         regs->ip                = new_ip;
 192         regs->sp                = new_sp;
 193         regs->flags             = X86_EFLAGS_IF;
 194         force_iret();
 195 }
 196 EXPORT_SYMBOL_GPL(start_thread);
 197
 198
 199 /*
 200  *      switch_to(x,y) should switch tasks from x to y.
 201  *
 202  * We fsave/fwait so that an exception goes off at the right time
 203  * (as a call from the fsave or fwait in effect) rather than to
 204  * the wrong process. Lazy FP saving no longer makes any sense
 205  * with modern CPU's, and this simplifies a lot of things (SMP
 206  * and UP become the same).
 207  *
 208  * NOTE! We used to use the x86 hardware context switching. The
 209  * reason for not using it any more becomes apparent when you
 210  * try to recover gracefully from saved state that is no longer
 211  * valid (stale segment register values in particular). With the
 212  * hardware task-switch, there is no way to fix up bad state in
 213  * a reasonable manner.
 214  *
 215  * The fact that Intel documents the hardware task-switching to
 216  * be slow is a fairly red herring - this code is not noticeably
 217  * faster. However, there _is_ some room for improvement here,
 218  * so the performance issues may eventually be a valid point.
 219  * More important, however, is the fact that this allows us much
 220  * more flexibility.
 221  *
 222  * The return value (in %ax) will be the "prev" task after
 223  * the task-switch, and shows up in ret_from_fork in entry.S,
 224  * for example.
 225  */
 226 __visible __notrace_funcgraph struct task_struct *
 227 __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 228 {
 229         struct thread_struct *prev = &prev_p->thread,
 230                              *next = &next_p->thread;
 231         struct fpu *prev_fpu = &prev->fpu;
 232         struct fpu *next_fpu = &next->fpu;
 233         int cpu = smp_processor_id();
 234
 235         /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
 236
 237         switch_fpu_prepare(prev_fpu, cpu);
 238
 239         /*
 240          * Save away %gs. No need to save %fs, as it was saved on the
 241          * stack on entry.  No need to save %es and %ds, as those are
 242          * always kernel segments while inside the kernel.  Doing this
 243          * before setting the new TLS descriptors avoids the situation
 244          * where we temporarily have non-reloadable segments in %fs
 245          * and %gs.  This could be an issue if the NMI handler ever
 246          * used %fs or %gs (it does not today), or if the kernel is
 247          * running inside of a hypervisor layer.
 248          */
 249         lazy_save_gs(prev->gs);
 250
 251         /*
 252          * Load the per-thread Thread-Local Storage descriptor.
 253          */
 254         load_TLS(next, cpu);
 255
 256         /*
 257          * Restore IOPL if needed.  In normal use, the flags restore
 258          * in the switch assembly will handle this.  But if the kernel
 259          * is running virtualized at a non-zero CPL, the popf will
 260          * not restore flags, so it must be done in a separate step.
 261          */
 262         if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
 263                 set_iopl_mask(next->iopl);
 264
 265         switch_to_extra(prev_p, next_p);
 266
 267         /*
 268          * Leave lazy mode, flushing any hypercalls made here.
 269          * This must be done before restoring TLS segments so
 270          * the GDT and LDT are properly updated, and must be
 271          * done before fpu__restore(), so the TS bit is up
 272          * to date.
 273          */
 274         arch_end_context_switch(next_p);
 275
 276         /*
 277          * Reload esp0 and cpu_current_top_of_stack.  This changes
 278          * current_thread_info().  Refresh the SYSENTER configuration in
 279          * case prev or next is vm86.
 280          */
 281         update_task_stack(next_p);
 282         refresh_sysenter_cs(next);
 283         this_cpu_write(cpu_current_top_of_stack,
 284                        (unsigned long)task_stack_page(next_p) +
 285                        THREAD_SIZE);
 286
 287         /*
 288          * Restore %gs if needed (which is common)
 289          */
 290         if (prev->gs | next->gs)
 291                 lazy_load_gs(next->gs);
 292
 293         switch_fpu_finish(next_fpu, cpu);
 294
 295         this_cpu_write(current_task, next_p);
 296
 297         /* Load the Intel cache allocation PQR MSR. */
 298         resctrl_sched_in();
 299
 300         return prev_p;
 301 }
 302
 303 SYSCALL_DEFINE2(arch_prctl, int, option, unsigned long, arg2)
 304 {
 305         return do_arch_prctl_common(current, option, arg2);
 306 }