Version:  2.0.40 2.2.26 2.4.37 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9

Linux/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/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/mc146818rtc.h>
 28 #include <linux/export.h>
 29 #include <linux/kallsyms.h>
 30 #include <linux/ptrace.h>
 31 #include <linux/personality.h>
 32 #include <linux/percpu.h>
 33 #include <linux/prctl.h>
 34 #include <linux/ftrace.h>
 35 #include <linux/uaccess.h>
 36 #include <linux/io.h>
 37 #include <linux/kdebug.h>
 38 
 39 #include <asm/pgtable.h>
 40 #include <asm/ldt.h>
 41 #include <asm/processor.h>
 42 #include <asm/fpu/internal.h>
 43 #include <asm/desc.h>
 44 #ifdef CONFIG_MATH_EMULATION
 45 #include <asm/math_emu.h>
 46 #endif
 47 
 48 #include <linux/err.h>
 49 
 50 #include <asm/tlbflush.h>
 51 #include <asm/cpu.h>
 52 #include <asm/idle.h>
 53 #include <asm/syscalls.h>
 54 #include <asm/debugreg.h>
 55 #include <asm/switch_to.h>
 56 #include <asm/vm86.h>
 57 
 58 void __show_regs(struct pt_regs *regs, int all)
 59 {
 60         unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
 61         unsigned long d0, d1, d2, d3, d6, d7;
 62         unsigned long sp;
 63         unsigned short ss, gs;
 64 
 65         if (user_mode(regs)) {
 66                 sp = regs->sp;
 67                 ss = regs->ss & 0xffff;
 68                 gs = get_user_gs(regs);
 69         } else {
 70                 sp = kernel_stack_pointer(regs);
 71                 savesegment(ss, ss);
 72                 savesegment(gs, gs);
 73         }
 74 
 75         printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
 76                         (u16)regs->cs, regs->ip, regs->flags,
 77                         smp_processor_id());
 78         print_symbol("EIP is at %s\n", regs->ip);
 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\n",
 85                (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
 86 
 87         if (!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         struct tss_struct *tss = &per_cpu(cpu_tss, cpu);
235         fpu_switch_t fpu_switch;
236 
237         /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
238 
239         fpu_switch = switch_fpu_prepare(prev_fpu, next_fpu, cpu);
240 
241         /*
242          * Save away %gs. No need to save %fs, as it was saved on the
243          * stack on entry.  No need to save %es and %ds, as those are
244          * always kernel segments while inside the kernel.  Doing this
245          * before setting the new TLS descriptors avoids the situation
246          * where we temporarily have non-reloadable segments in %fs
247          * and %gs.  This could be an issue if the NMI handler ever
248          * used %fs or %gs (it does not today), or if the kernel is
249          * running inside of a hypervisor layer.
250          */
251         lazy_save_gs(prev->gs);
252 
253         /*
254          * Load the per-thread Thread-Local Storage descriptor.
255          */
256         load_TLS(next, cpu);
257 
258         /*
259          * Restore IOPL if needed.  In normal use, the flags restore
260          * in the switch assembly will handle this.  But if the kernel
261          * is running virtualized at a non-zero CPL, the popf will
262          * not restore flags, so it must be done in a separate step.
263          */
264         if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
265                 set_iopl_mask(next->iopl);
266 
267         /*
268          * Now maybe handle debug registers and/or IO bitmaps
269          */
270         if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
271                      task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
272                 __switch_to_xtra(prev_p, next_p, tss);
273 
274         /*
275          * Leave lazy mode, flushing any hypercalls made here.
276          * This must be done before restoring TLS segments so
277          * the GDT and LDT are properly updated, and must be
278          * done before fpu__restore(), so the TS bit is up
279          * to date.
280          */
281         arch_end_context_switch(next_p);
282 
283         /*
284          * Reload esp0 and cpu_current_top_of_stack.  This changes
285          * current_thread_info().
286          */
287         load_sp0(tss, next);
288         this_cpu_write(cpu_current_top_of_stack,
289                        (unsigned long)task_stack_page(next_p) +
290                        THREAD_SIZE);
291 
292         /*
293          * Restore %gs if needed (which is common)
294          */
295         if (prev->gs | next->gs)
296                 lazy_load_gs(next->gs);
297 
298         switch_fpu_finish(next_fpu, fpu_switch);
299 
300         this_cpu_write(current_task, next_p);
301 
302         return prev_p;
303 }
304 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us