process.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852
  1. /*
  2. * linux/arch/i386/kernel/process.c
  3. *
  4. * Copyright (C) 1995 Linus Torvalds
  5. *
  6. * Pentium III FXSR, SSE support
  7. * Gareth Hughes <gareth@valinux.com>, May 2000
  8. */
  9. /*
  10. * This file handles the architecture-dependent parts of process handling..
  11. */
  12. #include <stdarg.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/smp_lock.h>
  21. #include <linux/stddef.h>
  22. #include <linux/slab.h>
  23. #include <linux/vmalloc.h>
  24. #include <linux/user.h>
  25. #include <linux/a.out.h>
  26. #include <linux/interrupt.h>
  27. #include <linux/config.h>
  28. #include <linux/utsname.h>
  29. #include <linux/delay.h>
  30. #include <linux/reboot.h>
  31. #include <linux/init.h>
  32. #include <linux/mc146818rtc.h>
  33. #include <linux/module.h>
  34. #include <linux/kallsyms.h>
  35. #include <linux/ptrace.h>
  36. #include <linux/random.h>
  37. #include <asm/uaccess.h>
  38. #include <asm/pgtable.h>
  39. #include <asm/system.h>
  40. #include <asm/io.h>
  41. #include <asm/ldt.h>
  42. #include <asm/processor.h>
  43. #include <asm/i387.h>
  44. #include <asm/irq.h>
  45. #include <asm/desc.h>
  46. #ifdef CONFIG_MATH_EMULATION
  47. #include <asm/math_emu.h>
  48. #endif
  49. #include <linux/irq.h>
  50. #include <linux/err.h>
  51. asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
  52. static int hlt_counter;
  53. unsigned long boot_option_idle_override = 0;
  54. EXPORT_SYMBOL(boot_option_idle_override);
  55. /*
  56. * Return saved PC of a blocked thread.
  57. */
  58. unsigned long thread_saved_pc(struct task_struct *tsk)
  59. {
  60. return ((unsigned long *)tsk->thread.esp)[3];
  61. }
  62. /*
  63. * Powermanagement idle function, if any..
  64. */
  65. void (*pm_idle)(void);
  66. static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
  67. void disable_hlt(void)
  68. {
  69. hlt_counter++;
  70. }
  71. EXPORT_SYMBOL(disable_hlt);
  72. void enable_hlt(void)
  73. {
  74. hlt_counter--;
  75. }
  76. EXPORT_SYMBOL(enable_hlt);
  77. /*
  78. * We use this if we don't have any better
  79. * idle routine..
  80. */
  81. void default_idle(void)
  82. {
  83. if (!hlt_counter && boot_cpu_data.hlt_works_ok) {
  84. local_irq_disable();
  85. if (!need_resched())
  86. safe_halt();
  87. else
  88. local_irq_enable();
  89. } else {
  90. cpu_relax();
  91. }
  92. }
  93. /*
  94. * On SMP it's slightly faster (but much more power-consuming!)
  95. * to poll the ->work.need_resched flag instead of waiting for the
  96. * cross-CPU IPI to arrive. Use this option with caution.
  97. */
  98. static void poll_idle (void)
  99. {
  100. int oldval;
  101. local_irq_enable();
  102. /*
  103. * Deal with another CPU just having chosen a thread to
  104. * run here:
  105. */
  106. oldval = test_and_clear_thread_flag(TIF_NEED_RESCHED);
  107. if (!oldval) {
  108. set_thread_flag(TIF_POLLING_NRFLAG);
  109. asm volatile(
  110. "2:"
  111. "testl %0, %1;"
  112. "rep; nop;"
  113. "je 2b;"
  114. : : "i"(_TIF_NEED_RESCHED), "m" (current_thread_info()->flags));
  115. clear_thread_flag(TIF_POLLING_NRFLAG);
  116. } else {
  117. set_need_resched();
  118. }
  119. }
  120. /*
  121. * The idle thread. There's no useful work to be
  122. * done, so just try to conserve power and have a
  123. * low exit latency (ie sit in a loop waiting for
  124. * somebody to say that they'd like to reschedule)
  125. */
  126. void cpu_idle (void)
  127. {
  128. /* endless idle loop with no priority at all */
  129. while (1) {
  130. while (!need_resched()) {
  131. void (*idle)(void);
  132. if (__get_cpu_var(cpu_idle_state))
  133. __get_cpu_var(cpu_idle_state) = 0;
  134. rmb();
  135. idle = pm_idle;
  136. if (!idle)
  137. idle = default_idle;
  138. __get_cpu_var(irq_stat).idle_timestamp = jiffies;
  139. idle();
  140. }
  141. schedule();
  142. }
  143. }
  144. void cpu_idle_wait(void)
  145. {
  146. unsigned int cpu, this_cpu = get_cpu();
  147. cpumask_t map;
  148. set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
  149. put_cpu();
  150. cpus_clear(map);
  151. for_each_online_cpu(cpu) {
  152. per_cpu(cpu_idle_state, cpu) = 1;
  153. cpu_set(cpu, map);
  154. }
  155. __get_cpu_var(cpu_idle_state) = 0;
  156. wmb();
  157. do {
  158. ssleep(1);
  159. for_each_online_cpu(cpu) {
  160. if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
  161. cpu_clear(cpu, map);
  162. }
  163. cpus_and(map, map, cpu_online_map);
  164. } while (!cpus_empty(map));
  165. }
  166. EXPORT_SYMBOL_GPL(cpu_idle_wait);
  167. /*
  168. * This uses new MONITOR/MWAIT instructions on P4 processors with PNI,
  169. * which can obviate IPI to trigger checking of need_resched.
  170. * We execute MONITOR against need_resched and enter optimized wait state
  171. * through MWAIT. Whenever someone changes need_resched, we would be woken
  172. * up from MWAIT (without an IPI).
  173. */
  174. static void mwait_idle(void)
  175. {
  176. local_irq_enable();
  177. if (!need_resched()) {
  178. set_thread_flag(TIF_POLLING_NRFLAG);
  179. do {
  180. __monitor((void *)&current_thread_info()->flags, 0, 0);
  181. if (need_resched())
  182. break;
  183. __mwait(0, 0);
  184. } while (!need_resched());
  185. clear_thread_flag(TIF_POLLING_NRFLAG);
  186. }
  187. }
  188. void __init select_idle_routine(const struct cpuinfo_x86 *c)
  189. {
  190. if (cpu_has(c, X86_FEATURE_MWAIT)) {
  191. printk("monitor/mwait feature present.\n");
  192. /*
  193. * Skip, if setup has overridden idle.
  194. * One CPU supports mwait => All CPUs supports mwait
  195. */
  196. if (!pm_idle) {
  197. printk("using mwait in idle threads.\n");
  198. pm_idle = mwait_idle;
  199. }
  200. }
  201. }
  202. static int __init idle_setup (char *str)
  203. {
  204. if (!strncmp(str, "poll", 4)) {
  205. printk("using polling idle threads.\n");
  206. pm_idle = poll_idle;
  207. #ifdef CONFIG_X86_SMP
  208. if (smp_num_siblings > 1)
  209. printk("WARNING: polling idle and HT enabled, performance may degrade.\n");
  210. #endif
  211. } else if (!strncmp(str, "halt", 4)) {
  212. printk("using halt in idle threads.\n");
  213. pm_idle = default_idle;
  214. }
  215. boot_option_idle_override = 1;
  216. return 1;
  217. }
  218. __setup("idle=", idle_setup);
  219. void show_regs(struct pt_regs * regs)
  220. {
  221. unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
  222. printk("\n");
  223. printk("Pid: %d, comm: %20s\n", current->pid, current->comm);
  224. printk("EIP: %04x:[<%08lx>] CPU: %d\n",0xffff & regs->xcs,regs->eip, smp_processor_id());
  225. print_symbol("EIP is at %s\n", regs->eip);
  226. if (regs->xcs & 3)
  227. printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp);
  228. printk(" EFLAGS: %08lx %s (%s)\n",
  229. regs->eflags, print_tainted(), system_utsname.release);
  230. printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
  231. regs->eax,regs->ebx,regs->ecx,regs->edx);
  232. printk("ESI: %08lx EDI: %08lx EBP: %08lx",
  233. regs->esi, regs->edi, regs->ebp);
  234. printk(" DS: %04x ES: %04x\n",
  235. 0xffff & regs->xds,0xffff & regs->xes);
  236. __asm__("movl %%cr0, %0": "=r" (cr0));
  237. __asm__("movl %%cr2, %0": "=r" (cr2));
  238. __asm__("movl %%cr3, %0": "=r" (cr3));
  239. /* This could fault if %cr4 does not exist */
  240. __asm__("1: movl %%cr4, %0 \n"
  241. "2: \n"
  242. ".section __ex_table,\"a\" \n"
  243. ".long 1b,2b \n"
  244. ".previous \n"
  245. : "=r" (cr4): "0" (0));
  246. printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", cr0, cr2, cr3, cr4);
  247. show_trace(NULL, &regs->esp);
  248. }
  249. /*
  250. * This gets run with %ebx containing the
  251. * function to call, and %edx containing
  252. * the "args".
  253. */
  254. extern void kernel_thread_helper(void);
  255. __asm__(".section .text\n"
  256. ".align 4\n"
  257. "kernel_thread_helper:\n\t"
  258. "movl %edx,%eax\n\t"
  259. "pushl %edx\n\t"
  260. "call *%ebx\n\t"
  261. "pushl %eax\n\t"
  262. "call do_exit\n"
  263. ".previous");
  264. /*
  265. * Create a kernel thread
  266. */
  267. int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
  268. {
  269. struct pt_regs regs;
  270. memset(&regs, 0, sizeof(regs));
  271. regs.ebx = (unsigned long) fn;
  272. regs.edx = (unsigned long) arg;
  273. regs.xds = __USER_DS;
  274. regs.xes = __USER_DS;
  275. regs.orig_eax = -1;
  276. regs.eip = (unsigned long) kernel_thread_helper;
  277. regs.xcs = __KERNEL_CS;
  278. regs.eflags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2;
  279. /* Ok, create the new process.. */
  280. return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
  281. }
  282. /*
  283. * Free current thread data structures etc..
  284. */
  285. void exit_thread(void)
  286. {
  287. struct task_struct *tsk = current;
  288. struct thread_struct *t = &tsk->thread;
  289. /* The process may have allocated an io port bitmap... nuke it. */
  290. if (unlikely(NULL != t->io_bitmap_ptr)) {
  291. int cpu = get_cpu();
  292. struct tss_struct *tss = &per_cpu(init_tss, cpu);
  293. kfree(t->io_bitmap_ptr);
  294. t->io_bitmap_ptr = NULL;
  295. /*
  296. * Careful, clear this in the TSS too:
  297. */
  298. memset(tss->io_bitmap, 0xff, tss->io_bitmap_max);
  299. t->io_bitmap_max = 0;
  300. tss->io_bitmap_owner = NULL;
  301. tss->io_bitmap_max = 0;
  302. tss->io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
  303. put_cpu();
  304. }
  305. }
  306. void flush_thread(void)
  307. {
  308. struct task_struct *tsk = current;
  309. memset(tsk->thread.debugreg, 0, sizeof(unsigned long)*8);
  310. memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array));
  311. /*
  312. * Forget coprocessor state..
  313. */
  314. clear_fpu(tsk);
  315. clear_used_math();
  316. }
  317. void release_thread(struct task_struct *dead_task)
  318. {
  319. if (dead_task->mm) {
  320. // temporary debugging check
  321. if (dead_task->mm->context.size) {
  322. printk("WARNING: dead process %8s still has LDT? <%p/%d>\n",
  323. dead_task->comm,
  324. dead_task->mm->context.ldt,
  325. dead_task->mm->context.size);
  326. BUG();
  327. }
  328. }
  329. release_vm86_irqs(dead_task);
  330. }
  331. /*
  332. * This gets called before we allocate a new thread and copy
  333. * the current task into it.
  334. */
  335. void prepare_to_copy(struct task_struct *tsk)
  336. {
  337. unlazy_fpu(tsk);
  338. }
  339. int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
  340. unsigned long unused,
  341. struct task_struct * p, struct pt_regs * regs)
  342. {
  343. struct pt_regs * childregs;
  344. struct task_struct *tsk;
  345. int err;
  346. childregs = ((struct pt_regs *) (THREAD_SIZE + (unsigned long) p->thread_info)) - 1;
  347. /*
  348. * The below -8 is to reserve 8 bytes on top of the ring0 stack.
  349. * This is necessary to guarantee that the entire "struct pt_regs"
  350. * is accessable even if the CPU haven't stored the SS/ESP registers
  351. * on the stack (interrupt gate does not save these registers
  352. * when switching to the same priv ring).
  353. * Therefore beware: accessing the xss/esp fields of the
  354. * "struct pt_regs" is possible, but they may contain the
  355. * completely wrong values.
  356. */
  357. childregs = (struct pt_regs *) ((unsigned long) childregs - 8);
  358. *childregs = *regs;
  359. childregs->eax = 0;
  360. childregs->esp = esp;
  361. p->thread.esp = (unsigned long) childregs;
  362. p->thread.esp0 = (unsigned long) (childregs+1);
  363. p->thread.eip = (unsigned long) ret_from_fork;
  364. savesegment(fs,p->thread.fs);
  365. savesegment(gs,p->thread.gs);
  366. tsk = current;
  367. if (unlikely(NULL != tsk->thread.io_bitmap_ptr)) {
  368. p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL);
  369. if (!p->thread.io_bitmap_ptr) {
  370. p->thread.io_bitmap_max = 0;
  371. return -ENOMEM;
  372. }
  373. memcpy(p->thread.io_bitmap_ptr, tsk->thread.io_bitmap_ptr,
  374. IO_BITMAP_BYTES);
  375. }
  376. /*
  377. * Set a new TLS for the child thread?
  378. */
  379. if (clone_flags & CLONE_SETTLS) {
  380. struct desc_struct *desc;
  381. struct user_desc info;
  382. int idx;
  383. err = -EFAULT;
  384. if (copy_from_user(&info, (void __user *)childregs->esi, sizeof(info)))
  385. goto out;
  386. err = -EINVAL;
  387. if (LDT_empty(&info))
  388. goto out;
  389. idx = info.entry_number;
  390. if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
  391. goto out;
  392. desc = p->thread.tls_array + idx - GDT_ENTRY_TLS_MIN;
  393. desc->a = LDT_entry_a(&info);
  394. desc->b = LDT_entry_b(&info);
  395. }
  396. err = 0;
  397. out:
  398. if (err && p->thread.io_bitmap_ptr) {
  399. kfree(p->thread.io_bitmap_ptr);
  400. p->thread.io_bitmap_max = 0;
  401. }
  402. return err;
  403. }
  404. /*
  405. * fill in the user structure for a core dump..
  406. */
  407. void dump_thread(struct pt_regs * regs, struct user * dump)
  408. {
  409. int i;
  410. /* changed the size calculations - should hopefully work better. lbt */
  411. dump->magic = CMAGIC;
  412. dump->start_code = 0;
  413. dump->start_stack = regs->esp & ~(PAGE_SIZE - 1);
  414. dump->u_tsize = ((unsigned long) current->mm->end_code) >> PAGE_SHIFT;
  415. dump->u_dsize = ((unsigned long) (current->mm->brk + (PAGE_SIZE-1))) >> PAGE_SHIFT;
  416. dump->u_dsize -= dump->u_tsize;
  417. dump->u_ssize = 0;
  418. for (i = 0; i < 8; i++)
  419. dump->u_debugreg[i] = current->thread.debugreg[i];
  420. if (dump->start_stack < TASK_SIZE)
  421. dump->u_ssize = ((unsigned long) (TASK_SIZE - dump->start_stack)) >> PAGE_SHIFT;
  422. dump->regs.ebx = regs->ebx;
  423. dump->regs.ecx = regs->ecx;
  424. dump->regs.edx = regs->edx;
  425. dump->regs.esi = regs->esi;
  426. dump->regs.edi = regs->edi;
  427. dump->regs.ebp = regs->ebp;
  428. dump->regs.eax = regs->eax;
  429. dump->regs.ds = regs->xds;
  430. dump->regs.es = regs->xes;
  431. savesegment(fs,dump->regs.fs);
  432. savesegment(gs,dump->regs.gs);
  433. dump->regs.orig_eax = regs->orig_eax;
  434. dump->regs.eip = regs->eip;
  435. dump->regs.cs = regs->xcs;
  436. dump->regs.eflags = regs->eflags;
  437. dump->regs.esp = regs->esp;
  438. dump->regs.ss = regs->xss;
  439. dump->u_fpvalid = dump_fpu (regs, &dump->i387);
  440. }
  441. /*
  442. * Capture the user space registers if the task is not running (in user space)
  443. */
  444. int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs)
  445. {
  446. struct pt_regs ptregs;
  447. ptregs = *(struct pt_regs *)
  448. ((unsigned long)tsk->thread_info+THREAD_SIZE - sizeof(ptregs));
  449. ptregs.xcs &= 0xffff;
  450. ptregs.xds &= 0xffff;
  451. ptregs.xes &= 0xffff;
  452. ptregs.xss &= 0xffff;
  453. elf_core_copy_regs(regs, &ptregs);
  454. return 1;
  455. }
  456. static inline void
  457. handle_io_bitmap(struct thread_struct *next, struct tss_struct *tss)
  458. {
  459. if (!next->io_bitmap_ptr) {
  460. /*
  461. * Disable the bitmap via an invalid offset. We still cache
  462. * the previous bitmap owner and the IO bitmap contents:
  463. */
  464. tss->io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
  465. return;
  466. }
  467. if (likely(next == tss->io_bitmap_owner)) {
  468. /*
  469. * Previous owner of the bitmap (hence the bitmap content)
  470. * matches the next task, we dont have to do anything but
  471. * to set a valid offset in the TSS:
  472. */
  473. tss->io_bitmap_base = IO_BITMAP_OFFSET;
  474. return;
  475. }
  476. /*
  477. * Lazy TSS's I/O bitmap copy. We set an invalid offset here
  478. * and we let the task to get a GPF in case an I/O instruction
  479. * is performed. The handler of the GPF will verify that the
  480. * faulting task has a valid I/O bitmap and, it true, does the
  481. * real copy and restart the instruction. This will save us
  482. * redundant copies when the currently switched task does not
  483. * perform any I/O during its timeslice.
  484. */
  485. tss->io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY;
  486. }
  487. /*
  488. * switch_to(x,yn) should switch tasks from x to y.
  489. *
  490. * We fsave/fwait so that an exception goes off at the right time
  491. * (as a call from the fsave or fwait in effect) rather than to
  492. * the wrong process. Lazy FP saving no longer makes any sense
  493. * with modern CPU's, and this simplifies a lot of things (SMP
  494. * and UP become the same).
  495. *
  496. * NOTE! We used to use the x86 hardware context switching. The
  497. * reason for not using it any more becomes apparent when you
  498. * try to recover gracefully from saved state that is no longer
  499. * valid (stale segment register values in particular). With the
  500. * hardware task-switch, there is no way to fix up bad state in
  501. * a reasonable manner.
  502. *
  503. * The fact that Intel documents the hardware task-switching to
  504. * be slow is a fairly red herring - this code is not noticeably
  505. * faster. However, there _is_ some room for improvement here,
  506. * so the performance issues may eventually be a valid point.
  507. * More important, however, is the fact that this allows us much
  508. * more flexibility.
  509. *
  510. * The return value (in %eax) will be the "prev" task after
  511. * the task-switch, and shows up in ret_from_fork in entry.S,
  512. * for example.
  513. */
  514. struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
  515. {
  516. struct thread_struct *prev = &prev_p->thread,
  517. *next = &next_p->thread;
  518. int cpu = smp_processor_id();
  519. struct tss_struct *tss = &per_cpu(init_tss, cpu);
  520. /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
  521. __unlazy_fpu(prev_p);
  522. /*
  523. * Reload esp0, LDT and the page table pointer:
  524. */
  525. load_esp0(tss, next);
  526. /*
  527. * Load the per-thread Thread-Local Storage descriptor.
  528. */
  529. load_TLS(next, cpu);
  530. /*
  531. * Save away %fs and %gs. No need to save %es and %ds, as
  532. * those are always kernel segments while inside the kernel.
  533. */
  534. asm volatile("mov %%fs,%0":"=m" (prev->fs));
  535. asm volatile("mov %%gs,%0":"=m" (prev->gs));
  536. /*
  537. * Restore %fs and %gs if needed.
  538. */
  539. if (unlikely(prev->fs | prev->gs | next->fs | next->gs)) {
  540. loadsegment(fs, next->fs);
  541. loadsegment(gs, next->gs);
  542. }
  543. /*
  544. * Now maybe reload the debug registers
  545. */
  546. if (unlikely(next->debugreg[7])) {
  547. loaddebug(next, 0);
  548. loaddebug(next, 1);
  549. loaddebug(next, 2);
  550. loaddebug(next, 3);
  551. /* no 4 and 5 */
  552. loaddebug(next, 6);
  553. loaddebug(next, 7);
  554. }
  555. if (unlikely(prev->io_bitmap_ptr || next->io_bitmap_ptr))
  556. handle_io_bitmap(next, tss);
  557. return prev_p;
  558. }
  559. asmlinkage int sys_fork(struct pt_regs regs)
  560. {
  561. return do_fork(SIGCHLD, regs.esp, &regs, 0, NULL, NULL);
  562. }
  563. asmlinkage int sys_clone(struct pt_regs regs)
  564. {
  565. unsigned long clone_flags;
  566. unsigned long newsp;
  567. int __user *parent_tidptr, *child_tidptr;
  568. clone_flags = regs.ebx;
  569. newsp = regs.ecx;
  570. parent_tidptr = (int __user *)regs.edx;
  571. child_tidptr = (int __user *)regs.edi;
  572. if (!newsp)
  573. newsp = regs.esp;
  574. return do_fork(clone_flags, newsp, &regs, 0, parent_tidptr, child_tidptr);
  575. }
  576. /*
  577. * This is trivial, and on the face of it looks like it
  578. * could equally well be done in user mode.
  579. *
  580. * Not so, for quite unobvious reasons - register pressure.
  581. * In user mode vfork() cannot have a stack frame, and if
  582. * done by calling the "clone()" system call directly, you
  583. * do not have enough call-clobbered registers to hold all
  584. * the information you need.
  585. */
  586. asmlinkage int sys_vfork(struct pt_regs regs)
  587. {
  588. return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.esp, &regs, 0, NULL, NULL);
  589. }
  590. /*
  591. * sys_execve() executes a new program.
  592. */
  593. asmlinkage int sys_execve(struct pt_regs regs)
  594. {
  595. int error;
  596. char * filename;
  597. filename = getname((char __user *) regs.ebx);
  598. error = PTR_ERR(filename);
  599. if (IS_ERR(filename))
  600. goto out;
  601. error = do_execve(filename,
  602. (char __user * __user *) regs.ecx,
  603. (char __user * __user *) regs.edx,
  604. &regs);
  605. if (error == 0) {
  606. task_lock(current);
  607. current->ptrace &= ~PT_DTRACE;
  608. task_unlock(current);
  609. /* Make sure we don't return using sysenter.. */
  610. set_thread_flag(TIF_IRET);
  611. }
  612. putname(filename);
  613. out:
  614. return error;
  615. }
  616. #define top_esp (THREAD_SIZE - sizeof(unsigned long))
  617. #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
  618. unsigned long get_wchan(struct task_struct *p)
  619. {
  620. unsigned long ebp, esp, eip;
  621. unsigned long stack_page;
  622. int count = 0;
  623. if (!p || p == current || p->state == TASK_RUNNING)
  624. return 0;
  625. stack_page = (unsigned long)p->thread_info;
  626. esp = p->thread.esp;
  627. if (!stack_page || esp < stack_page || esp > top_esp+stack_page)
  628. return 0;
  629. /* include/asm-i386/system.h:switch_to() pushes ebp last. */
  630. ebp = *(unsigned long *) esp;
  631. do {
  632. if (ebp < stack_page || ebp > top_ebp+stack_page)
  633. return 0;
  634. eip = *(unsigned long *) (ebp+4);
  635. if (!in_sched_functions(eip))
  636. return eip;
  637. ebp = *(unsigned long *) ebp;
  638. } while (count++ < 16);
  639. return 0;
  640. }
  641. /*
  642. * sys_alloc_thread_area: get a yet unused TLS descriptor index.
  643. */
  644. static int get_free_idx(void)
  645. {
  646. struct thread_struct *t = &current->thread;
  647. int idx;
  648. for (idx = 0; idx < GDT_ENTRY_TLS_ENTRIES; idx++)
  649. if (desc_empty(t->tls_array + idx))
  650. return idx + GDT_ENTRY_TLS_MIN;
  651. return -ESRCH;
  652. }
  653. /*
  654. * Set a given TLS descriptor:
  655. */
  656. asmlinkage int sys_set_thread_area(struct user_desc __user *u_info)
  657. {
  658. struct thread_struct *t = &current->thread;
  659. struct user_desc info;
  660. struct desc_struct *desc;
  661. int cpu, idx;
  662. if (copy_from_user(&info, u_info, sizeof(info)))
  663. return -EFAULT;
  664. idx = info.entry_number;
  665. /*
  666. * index -1 means the kernel should try to find and
  667. * allocate an empty descriptor:
  668. */
  669. if (idx == -1) {
  670. idx = get_free_idx();
  671. if (idx < 0)
  672. return idx;
  673. if (put_user(idx, &u_info->entry_number))
  674. return -EFAULT;
  675. }
  676. if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
  677. return -EINVAL;
  678. desc = t->tls_array + idx - GDT_ENTRY_TLS_MIN;
  679. /*
  680. * We must not get preempted while modifying the TLS.
  681. */
  682. cpu = get_cpu();
  683. if (LDT_empty(&info)) {
  684. desc->a = 0;
  685. desc->b = 0;
  686. } else {
  687. desc->a = LDT_entry_a(&info);
  688. desc->b = LDT_entry_b(&info);
  689. }
  690. load_TLS(t, cpu);
  691. put_cpu();
  692. return 0;
  693. }
  694. /*
  695. * Get the current Thread-Local Storage area:
  696. */
  697. #define GET_BASE(desc) ( \
  698. (((desc)->a >> 16) & 0x0000ffff) | \
  699. (((desc)->b << 16) & 0x00ff0000) | \
  700. ( (desc)->b & 0xff000000) )
  701. #define GET_LIMIT(desc) ( \
  702. ((desc)->a & 0x0ffff) | \
  703. ((desc)->b & 0xf0000) )
  704. #define GET_32BIT(desc) (((desc)->b >> 22) & 1)
  705. #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3)
  706. #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1)
  707. #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1)
  708. #define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
  709. #define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
  710. asmlinkage int sys_get_thread_area(struct user_desc __user *u_info)
  711. {
  712. struct user_desc info;
  713. struct desc_struct *desc;
  714. int idx;
  715. if (get_user(idx, &u_info->entry_number))
  716. return -EFAULT;
  717. if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
  718. return -EINVAL;
  719. desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN;
  720. info.entry_number = idx;
  721. info.base_addr = GET_BASE(desc);
  722. info.limit = GET_LIMIT(desc);
  723. info.seg_32bit = GET_32BIT(desc);
  724. info.contents = GET_CONTENTS(desc);
  725. info.read_exec_only = !GET_WRITABLE(desc);
  726. info.limit_in_pages = GET_LIMIT_PAGES(desc);
  727. info.seg_not_present = !GET_PRESENT(desc);
  728. info.useable = GET_USEABLE(desc);
  729. if (copy_to_user(u_info, &info, sizeof(info)))
  730. return -EFAULT;
  731. return 0;
  732. }
  733. unsigned long arch_align_stack(unsigned long sp)
  734. {
  735. if (randomize_va_space)
  736. sp -= get_random_int() % 8192;
  737. return sp & ~0xf;
  738. }