process_32.c 8.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337
  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. * This file handles the architecture-dependent parts of process handling..
  9. */
  10. #include <linux/cpu.h>
  11. #include <linux/errno.h>
  12. #include <linux/sched.h>
  13. #include <linux/fs.h>
  14. #include <linux/kernel.h>
  15. #include <linux/mm.h>
  16. #include <linux/elfcore.h>
  17. #include <linux/smp.h>
  18. #include <linux/stddef.h>
  19. #include <linux/slab.h>
  20. #include <linux/vmalloc.h>
  21. #include <linux/user.h>
  22. #include <linux/interrupt.h>
  23. #include <linux/delay.h>
  24. #include <linux/reboot.h>
  25. #include <linux/init.h>
  26. #include <linux/mc146818rtc.h>
  27. #include <linux/module.h>
  28. #include <linux/kallsyms.h>
  29. #include <linux/ptrace.h>
  30. #include <linux/personality.h>
  31. #include <linux/percpu.h>
  32. #include <linux/prctl.h>
  33. #include <linux/ftrace.h>
  34. #include <linux/uaccess.h>
  35. #include <linux/io.h>
  36. #include <linux/kdebug.h>
  37. #include <asm/pgtable.h>
  38. #include <asm/ldt.h>
  39. #include <asm/processor.h>
  40. #include <asm/i387.h>
  41. #include <asm/fpu-internal.h>
  42. #include <asm/desc.h>
  43. #ifdef CONFIG_MATH_EMULATION
  44. #include <asm/math_emu.h>
  45. #endif
  46. #include <linux/err.h>
  47. #include <asm/tlbflush.h>
  48. #include <asm/cpu.h>
  49. #include <asm/idle.h>
  50. #include <asm/syscalls.h>
  51. #include <asm/debugreg.h>
  52. #include <asm/switch_to.h>
  53. asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
  54. /*
  55. * Return saved PC of a blocked thread.
  56. */
  57. unsigned long thread_saved_pc(struct task_struct *tsk)
  58. {
  59. return ((unsigned long *)tsk->thread.sp)[3];
  60. }
  61. void __show_regs(struct pt_regs *regs, int all)
  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. if (user_mode_vm(regs)) {
  68. sp = regs->sp;
  69. ss = regs->ss & 0xffff;
  70. gs = get_user_gs(regs);
  71. } else {
  72. sp = kernel_stack_pointer(regs);
  73. savesegment(ss, ss);
  74. savesegment(gs, gs);
  75. }
  76. show_regs_common();
  77. printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
  78. (u16)regs->cs, regs->ip, regs->flags,
  79. smp_processor_id());
  80. print_symbol("EIP is at %s\n", regs->ip);
  81. printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
  82. regs->ax, regs->bx, regs->cx, regs->dx);
  83. printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
  84. regs->si, regs->di, regs->bp, sp);
  85. printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
  86. (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
  87. if (!all)
  88. return;
  89. cr0 = read_cr0();
  90. cr2 = read_cr2();
  91. cr3 = read_cr3();
  92. cr4 = read_cr4_safe();
  93. printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
  94. cr0, cr2, cr3, cr4);
  95. get_debugreg(d0, 0);
  96. get_debugreg(d1, 1);
  97. get_debugreg(d2, 2);
  98. get_debugreg(d3, 3);
  99. printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
  100. d0, d1, d2, d3);
  101. get_debugreg(d6, 6);
  102. get_debugreg(d7, 7);
  103. printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
  104. d6, d7);
  105. }
  106. void release_thread(struct task_struct *dead_task)
  107. {
  108. BUG_ON(dead_task->mm);
  109. release_vm86_irqs(dead_task);
  110. }
  111. /*
  112. * This gets called before we allocate a new thread and copy
  113. * the current task into it.
  114. */
  115. void prepare_to_copy(struct task_struct *tsk)
  116. {
  117. unlazy_fpu(tsk);
  118. }
  119. int copy_thread(unsigned long clone_flags, unsigned long sp,
  120. unsigned long unused,
  121. struct task_struct *p, struct pt_regs *regs)
  122. {
  123. struct pt_regs *childregs;
  124. struct task_struct *tsk;
  125. int err;
  126. childregs = task_pt_regs(p);
  127. *childregs = *regs;
  128. childregs->ax = 0;
  129. childregs->sp = sp;
  130. p->thread.sp = (unsigned long) childregs;
  131. p->thread.sp0 = (unsigned long) (childregs+1);
  132. p->thread.ip = (unsigned long) ret_from_fork;
  133. task_user_gs(p) = get_user_gs(regs);
  134. p->fpu_counter = 0;
  135. p->thread.io_bitmap_ptr = NULL;
  136. tsk = current;
  137. err = -ENOMEM;
  138. memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
  139. if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
  140. p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
  141. IO_BITMAP_BYTES, GFP_KERNEL);
  142. if (!p->thread.io_bitmap_ptr) {
  143. p->thread.io_bitmap_max = 0;
  144. return -ENOMEM;
  145. }
  146. set_tsk_thread_flag(p, TIF_IO_BITMAP);
  147. }
  148. err = 0;
  149. /*
  150. * Set a new TLS for the child thread?
  151. */
  152. if (clone_flags & CLONE_SETTLS)
  153. err = do_set_thread_area(p, -1,
  154. (struct user_desc __user *)childregs->si, 0);
  155. if (err && p->thread.io_bitmap_ptr) {
  156. kfree(p->thread.io_bitmap_ptr);
  157. p->thread.io_bitmap_max = 0;
  158. }
  159. return err;
  160. }
  161. void
  162. start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
  163. {
  164. set_user_gs(regs, 0);
  165. regs->fs = 0;
  166. regs->ds = __USER_DS;
  167. regs->es = __USER_DS;
  168. regs->ss = __USER_DS;
  169. regs->cs = __USER_CS;
  170. regs->ip = new_ip;
  171. regs->sp = new_sp;
  172. /*
  173. * Free the old FP and other extended state
  174. */
  175. free_thread_xstate(current);
  176. }
  177. EXPORT_SYMBOL_GPL(start_thread);
  178. /*
  179. * switch_to(x,y) should switch tasks from x to y.
  180. *
  181. * We fsave/fwait so that an exception goes off at the right time
  182. * (as a call from the fsave or fwait in effect) rather than to
  183. * the wrong process. Lazy FP saving no longer makes any sense
  184. * with modern CPU's, and this simplifies a lot of things (SMP
  185. * and UP become the same).
  186. *
  187. * NOTE! We used to use the x86 hardware context switching. The
  188. * reason for not using it any more becomes apparent when you
  189. * try to recover gracefully from saved state that is no longer
  190. * valid (stale segment register values in particular). With the
  191. * hardware task-switch, there is no way to fix up bad state in
  192. * a reasonable manner.
  193. *
  194. * The fact that Intel documents the hardware task-switching to
  195. * be slow is a fairly red herring - this code is not noticeably
  196. * faster. However, there _is_ some room for improvement here,
  197. * so the performance issues may eventually be a valid point.
  198. * More important, however, is the fact that this allows us much
  199. * more flexibility.
  200. *
  201. * The return value (in %ax) will be the "prev" task after
  202. * the task-switch, and shows up in ret_from_fork in entry.S,
  203. * for example.
  204. */
  205. __notrace_funcgraph struct task_struct *
  206. __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
  207. {
  208. struct thread_struct *prev = &prev_p->thread,
  209. *next = &next_p->thread;
  210. int cpu = smp_processor_id();
  211. struct tss_struct *tss = &per_cpu(init_tss, cpu);
  212. fpu_switch_t fpu;
  213. /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
  214. fpu = switch_fpu_prepare(prev_p, next_p, cpu);
  215. /*
  216. * Reload esp0.
  217. */
  218. load_sp0(tss, next);
  219. /*
  220. * Save away %gs. No need to save %fs, as it was saved on the
  221. * stack on entry. No need to save %es and %ds, as those are
  222. * always kernel segments while inside the kernel. Doing this
  223. * before setting the new TLS descriptors avoids the situation
  224. * where we temporarily have non-reloadable segments in %fs
  225. * and %gs. This could be an issue if the NMI handler ever
  226. * used %fs or %gs (it does not today), or if the kernel is
  227. * running inside of a hypervisor layer.
  228. */
  229. lazy_save_gs(prev->gs);
  230. /*
  231. * Load the per-thread Thread-Local Storage descriptor.
  232. */
  233. load_TLS(next, cpu);
  234. /*
  235. * Restore IOPL if needed. In normal use, the flags restore
  236. * in the switch assembly will handle this. But if the kernel
  237. * is running virtualized at a non-zero CPL, the popf will
  238. * not restore flags, so it must be done in a separate step.
  239. */
  240. if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
  241. set_iopl_mask(next->iopl);
  242. /*
  243. * Now maybe handle debug registers and/or IO bitmaps
  244. */
  245. if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
  246. task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
  247. __switch_to_xtra(prev_p, next_p, tss);
  248. /*
  249. * Leave lazy mode, flushing any hypercalls made here.
  250. * This must be done before restoring TLS segments so
  251. * the GDT and LDT are properly updated, and must be
  252. * done before math_state_restore, so the TS bit is up
  253. * to date.
  254. */
  255. arch_end_context_switch(next_p);
  256. /*
  257. * Restore %gs if needed (which is common)
  258. */
  259. if (prev->gs | next->gs)
  260. lazy_load_gs(next->gs);
  261. switch_fpu_finish(next_p, fpu);
  262. percpu_write(current_task, next_p);
  263. return prev_p;
  264. }
  265. #define top_esp (THREAD_SIZE - sizeof(unsigned long))
  266. #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
  267. unsigned long get_wchan(struct task_struct *p)
  268. {
  269. unsigned long bp, sp, ip;
  270. unsigned long stack_page;
  271. int count = 0;
  272. if (!p || p == current || p->state == TASK_RUNNING)
  273. return 0;
  274. stack_page = (unsigned long)task_stack_page(p);
  275. sp = p->thread.sp;
  276. if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
  277. return 0;
  278. /* include/asm-i386/system.h:switch_to() pushes bp last. */
  279. bp = *(unsigned long *) sp;
  280. do {
  281. if (bp < stack_page || bp > top_ebp+stack_page)
  282. return 0;
  283. ip = *(unsigned long *) (bp+4);
  284. if (!in_sched_functions(ip))
  285. return ip;
  286. bp = *(unsigned long *) bp;
  287. } while (count++ < 16);
  288. return 0;
  289. }