4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * #!-checking implemented by tytso.
11 * Demand-loading implemented 01.12.91 - no need to read anything but
12 * the header into memory. The inode of the executable is put into
13 * "current->executable", and page faults do the actual loading. Clean.
15 * Once more I can proudly say that linux stood up to being changed: it
16 * was less than 2 hours work to get demand-loading completely implemented.
18 * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead,
19 * current->executable is only used by the procfs. This allows a dispatch
20 * table to check for several different types of binary formats. We keep
21 * trying until we recognize the file or we run out of supported binary
25 #include <linux/config.h>
26 #include <linux/slab.h>
27 #include <linux/file.h>
28 #include <linux/mman.h>
29 #include <linux/a.out.h>
30 #include <linux/stat.h>
31 #include <linux/fcntl.h>
32 #include <linux/smp_lock.h>
33 #include <linux/init.h>
34 #include <linux/pagemap.h>
35 #include <linux/highmem.h>
36 #include <linux/spinlock.h>
37 #include <linux/personality.h>
38 #include <linux/binfmts.h>
39 #include <linux/swap.h>
40 #include <linux/utsname.h>
41 #include <linux/module.h>
42 #include <linux/namei.h>
43 #include <linux/proc_fs.h>
44 #include <linux/ptrace.h>
45 #include <linux/mount.h>
46 #include <linux/security.h>
47 #include <linux/syscalls.h>
48 #include <linux/rmap.h>
50 #include <asm/uaccess.h>
51 #include <asm/pgalloc.h>
52 #include <asm/mmu_context.h>
55 #include <linux/kmod.h>
59 char core_pattern[65] = "core";
60 /* The maximal length of core_pattern is also specified in sysctl.c */
62 static struct linux_binfmt *formats;
63 static rwlock_t binfmt_lock = RW_LOCK_UNLOCKED;
65 int register_binfmt(struct linux_binfmt * fmt)
67 struct linux_binfmt ** tmp = &formats;
73 write_lock(&binfmt_lock);
76 write_unlock(&binfmt_lock);
83 write_unlock(&binfmt_lock);
87 EXPORT_SYMBOL(register_binfmt);
89 int unregister_binfmt(struct linux_binfmt * fmt)
91 struct linux_binfmt ** tmp = &formats;
93 write_lock(&binfmt_lock);
97 write_unlock(&binfmt_lock);
102 write_unlock(&binfmt_lock);
106 EXPORT_SYMBOL(unregister_binfmt);
108 static inline void put_binfmt(struct linux_binfmt * fmt)
110 module_put(fmt->module);
114 * Note that a shared library must be both readable and executable due to
117 * Also note that we take the address to load from from the file itself.
119 asmlinkage long sys_uselib(const char __user * library)
125 nd.intent.open.flags = FMODE_READ;
126 error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
131 if (!S_ISREG(nd.dentry->d_inode->i_mode))
134 error = permission(nd.dentry->d_inode, MAY_READ | MAY_EXEC, &nd);
138 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
139 error = PTR_ERR(file);
145 struct linux_binfmt * fmt;
147 read_lock(&binfmt_lock);
148 for (fmt = formats ; fmt ; fmt = fmt->next) {
149 if (!fmt->load_shlib)
151 if (!try_module_get(fmt->module))
153 read_unlock(&binfmt_lock);
154 error = fmt->load_shlib(file);
155 read_lock(&binfmt_lock);
157 if (error != -ENOEXEC)
160 read_unlock(&binfmt_lock);
171 * count() counts the number of strings in array ARGV.
173 static int count(char __user * __user * argv, int max)
181 if (get_user(p, argv))
194 * 'copy_strings()' copies argument/environment strings from user
195 * memory to free pages in kernel mem. These are in a format ready
196 * to be put directly into the top of new user memory.
198 int copy_strings(int argc,char __user * __user * argv, struct linux_binprm *bprm)
200 struct page *kmapped_page = NULL;
209 if (get_user(str, argv+argc) ||
210 !(len = strnlen_user(str, bprm->p))) {
221 /* XXX: add architecture specific overflow check here. */
226 int offset, bytes_to_copy;
229 offset = pos % PAGE_SIZE;
231 page = bprm->page[i];
234 page = alloc_page(GFP_HIGHUSER);
235 bprm->page[i] = page;
243 if (page != kmapped_page) {
245 kunmap(kmapped_page);
247 kaddr = kmap(kmapped_page);
250 memset(kaddr, 0, offset);
251 bytes_to_copy = PAGE_SIZE - offset;
252 if (bytes_to_copy > len) {
255 memset(kaddr+offset+len, 0,
256 PAGE_SIZE-offset-len);
258 err = copy_from_user(kaddr+offset, str, bytes_to_copy);
264 pos += bytes_to_copy;
265 str += bytes_to_copy;
266 len -= bytes_to_copy;
272 kunmap(kmapped_page);
277 * Like copy_strings, but get argv and its values from kernel memory.
279 int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
282 mm_segment_t oldfs = get_fs();
284 r = copy_strings(argc, (char __user * __user *)argv, bprm);
289 EXPORT_SYMBOL(copy_strings_kernel);
293 * This routine is used to map in a page into an address space: needed by
294 * execve() for the initial stack and environment pages.
296 * vma->vm_mm->mmap_sem is held for writing.
298 void install_arg_page(struct vm_area_struct *vma,
299 struct page *page, unsigned long address)
301 struct mm_struct *mm = vma->vm_mm;
306 if (unlikely(anon_vma_prepare(vma)))
309 flush_dcache_page(page);
310 pgd = pgd_offset(mm, address);
312 spin_lock(&mm->page_table_lock);
313 pmd = pmd_alloc(mm, pgd, address);
316 pte = pte_alloc_map(mm, pmd, address);
319 if (!pte_none(*pte)) {
325 lru_cache_add_active(page);
326 set_pte(pte, pte_mkdirty(pte_mkwrite(mk_pte(
327 page, vma->vm_page_prot))));
328 page_add_anon_rmap(page, vma, address);
330 spin_unlock(&mm->page_table_lock);
332 /* no need for flush_tlb */
335 spin_unlock(&mm->page_table_lock);
338 force_sig(SIGKILL, current);
341 int setup_arg_pages(struct linux_binprm *bprm, int executable_stack)
343 unsigned long stack_base;
344 struct vm_area_struct *mpnt;
345 struct mm_struct *mm = current->mm;
349 #ifdef CONFIG_STACK_GROWSUP
350 /* Move the argument and environment strings to the bottom of the
356 /* Start by shifting all the pages down */
358 for (j = 0; j < MAX_ARG_PAGES; j++) {
359 struct page *page = bprm->page[j];
362 bprm->page[i++] = page;
365 /* Now move them within their pages */
366 offset = bprm->p % PAGE_SIZE;
367 to = kmap(bprm->page[0]);
368 for (j = 1; j < i; j++) {
369 memmove(to, to + offset, PAGE_SIZE - offset);
370 from = kmap(bprm->page[j]);
371 memcpy(to + PAGE_SIZE - offset, from, offset);
372 kunmap(bprm->page[j - 1]);
375 memmove(to, to + offset, PAGE_SIZE - offset);
376 kunmap(bprm->page[j - 1]);
378 /* Adjust bprm->p to point to the end of the strings. */
379 bprm->p = PAGE_SIZE * i - offset;
381 /* Limit stack size to 1GB */
382 stack_base = current->rlim[RLIMIT_STACK].rlim_max;
383 if (stack_base > (1 << 30))
384 stack_base = 1 << 30;
385 stack_base = PAGE_ALIGN(STACK_TOP - stack_base);
387 mm->arg_start = stack_base;
388 arg_size = i << PAGE_SHIFT;
390 /* zero pages that were copied above */
391 while (i < MAX_ARG_PAGES)
392 bprm->page[i++] = NULL;
394 stack_base = STACK_TOP - MAX_ARG_PAGES * PAGE_SIZE;
395 mm->arg_start = bprm->p + stack_base;
396 arg_size = STACK_TOP - (PAGE_MASK & (unsigned long) mm->arg_start);
399 bprm->p += stack_base;
401 bprm->loader += stack_base;
402 bprm->exec += stack_base;
404 mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
408 if (security_vm_enough_memory(arg_size >> PAGE_SHIFT) ||
409 !vx_vmpages_avail(mm, arg_size >> PAGE_SHIFT)) {
410 kmem_cache_free(vm_area_cachep, mpnt);
414 memset(mpnt, 0, sizeof(*mpnt));
416 down_write(&mm->mmap_sem);
419 #ifdef CONFIG_STACK_GROWSUP
420 mpnt->vm_start = stack_base;
421 mpnt->vm_end = PAGE_MASK &
422 (PAGE_SIZE - 1 + (unsigned long) bprm->p);
424 mpnt->vm_start = PAGE_MASK & (unsigned long) bprm->p;
425 mpnt->vm_end = STACK_TOP;
427 /* Adjust stack execute permissions; explicitly enable
428 * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X
429 * and leave alone (arch default) otherwise. */
430 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
431 mpnt->vm_flags = VM_STACK_FLAGS | VM_EXEC;
432 else if (executable_stack == EXSTACK_DISABLE_X)
433 mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC;
435 mpnt->vm_flags = VM_STACK_FLAGS;
436 mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7];
437 insert_vm_struct(mm, mpnt);
438 // mm->total_vm = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
439 vx_vmpages_sub(mm, mm->total_vm -
440 ((mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT));
443 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
444 struct page *page = bprm->page[i];
446 bprm->page[i] = NULL;
447 install_arg_page(mpnt, page, stack_base);
449 stack_base += PAGE_SIZE;
451 up_write(&mm->mmap_sem);
456 EXPORT_SYMBOL(setup_arg_pages);
458 #define free_arg_pages(bprm) do { } while (0)
462 static inline void free_arg_pages(struct linux_binprm *bprm)
466 for (i = 0; i < MAX_ARG_PAGES; i++) {
468 __free_page(bprm->page[i]);
469 bprm->page[i] = NULL;
473 #endif /* CONFIG_MMU */
475 struct file *open_exec(const char *name)
481 nd.intent.open.flags = FMODE_READ;
482 err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
486 struct inode *inode = nd.dentry->d_inode;
487 file = ERR_PTR(-EACCES);
488 if (!(nd.mnt->mnt_flags & MNT_NOEXEC) &&
489 S_ISREG(inode->i_mode)) {
490 int err = permission(inode, MAY_EXEC, &nd);
491 if (!err && !(inode->i_mode & 0111))
495 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
497 err = deny_write_access(file);
512 EXPORT_SYMBOL(open_exec);
514 int kernel_read(struct file *file, unsigned long offset,
515 char *addr, unsigned long count)
523 /* The cast to a user pointer is valid due to the set_fs() */
524 result = vfs_read(file, (void __user *)addr, count, &pos);
529 EXPORT_SYMBOL(kernel_read);
531 static int exec_mmap(struct mm_struct *mm)
533 struct task_struct *tsk;
534 struct mm_struct * old_mm, *active_mm;
536 /* Add it to the list of mm's */
537 spin_lock(&mmlist_lock);
538 list_add(&mm->mmlist, &init_mm.mmlist);
540 spin_unlock(&mmlist_lock);
542 /* Notify parent that we're no longer interested in the old VM */
544 old_mm = current->mm;
545 mm_release(tsk, old_mm);
548 active_mm = tsk->active_mm;
551 activate_mm(active_mm, mm);
554 if (active_mm != old_mm) BUG();
563 * This function makes sure the current process has its own signal table,
564 * so that flush_signal_handlers can later reset the handlers without
565 * disturbing other processes. (Other processes might share the signal
566 * table via the CLONE_SIGHAND option to clone().)
568 static inline int de_thread(struct task_struct *tsk)
570 struct signal_struct *newsig, *oldsig = tsk->signal;
571 struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
572 spinlock_t *lock = &oldsighand->siglock;
576 * If we don't share sighandlers, then we aren't sharing anything
577 * and we can just re-use it all.
579 if (atomic_read(&oldsighand->count) <= 1)
582 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
586 spin_lock_init(&newsighand->siglock);
587 atomic_set(&newsighand->count, 1);
588 memcpy(newsighand->action, oldsighand->action, sizeof(newsighand->action));
591 * See if we need to allocate a new signal structure
594 if (atomic_read(&oldsig->count) > 1) {
595 newsig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
597 kmem_cache_free(sighand_cachep, newsighand);
600 atomic_set(&newsig->count, 1);
601 newsig->group_exit = 0;
602 newsig->group_exit_code = 0;
603 newsig->group_exit_task = NULL;
604 newsig->group_stop_count = 0;
605 newsig->curr_target = NULL;
606 init_sigpending(&newsig->shared_pending);
607 INIT_LIST_HEAD(&newsig->posix_timers);
609 newsig->tty = oldsig->tty;
610 newsig->pgrp = oldsig->pgrp;
611 newsig->session = oldsig->session;
612 newsig->leader = oldsig->leader;
613 newsig->tty_old_pgrp = oldsig->tty_old_pgrp;
616 if (thread_group_empty(current))
617 goto no_thread_group;
620 * Kill all other threads in the thread group.
621 * We must hold tasklist_lock to call zap_other_threads.
623 read_lock(&tasklist_lock);
625 if (oldsig->group_exit) {
627 * Another group action in progress, just
628 * return so that the signal is processed.
630 spin_unlock_irq(lock);
631 read_unlock(&tasklist_lock);
632 kmem_cache_free(sighand_cachep, newsighand);
634 kmem_cache_free(signal_cachep, newsig);
637 oldsig->group_exit = 1;
638 zap_other_threads(current);
639 read_unlock(&tasklist_lock);
642 * Account for the thread group leader hanging around:
645 if (current->pid == current->tgid)
647 while (atomic_read(&oldsig->count) > count) {
648 oldsig->group_exit_task = current;
649 oldsig->notify_count = count;
650 __set_current_state(TASK_UNINTERRUPTIBLE);
651 spin_unlock_irq(lock);
655 spin_unlock_irq(lock);
658 * At this point all other threads have exited, all we have to
659 * do is to wait for the thread group leader to become inactive,
660 * and to assume its PID:
662 if (current->pid != current->tgid) {
663 struct task_struct *leader = current->group_leader, *parent;
664 struct dentry *proc_dentry1, *proc_dentry2;
665 unsigned long state, ptrace;
668 * Wait for the thread group leader to be a zombie.
669 * It should already be zombie at this point, most
672 while (leader->state != TASK_ZOMBIE)
675 spin_lock(&leader->proc_lock);
676 spin_lock(¤t->proc_lock);
677 proc_dentry1 = proc_pid_unhash(current);
678 proc_dentry2 = proc_pid_unhash(leader);
679 write_lock_irq(&tasklist_lock);
681 if (leader->tgid != current->tgid)
683 if (current->pid == current->tgid)
686 * An exec() starts a new thread group with the
687 * TGID of the previous thread group. Rehash the
688 * two threads with a switched PID, and release
689 * the former thread group leader:
691 ptrace = leader->ptrace;
692 parent = leader->parent;
694 ptrace_unlink(current);
695 ptrace_unlink(leader);
696 remove_parent(current);
697 remove_parent(leader);
699 switch_exec_pids(leader, current);
701 current->parent = current->real_parent = leader->real_parent;
702 leader->parent = leader->real_parent = child_reaper;
703 current->group_leader = current;
704 leader->group_leader = leader;
706 add_parent(current, current->parent);
707 add_parent(leader, leader->parent);
709 current->ptrace = ptrace;
710 __ptrace_link(current, parent);
713 list_del(¤t->tasks);
714 list_add_tail(¤t->tasks, &init_task.tasks);
715 current->exit_signal = SIGCHLD;
716 state = leader->state;
718 write_unlock_irq(&tasklist_lock);
719 spin_unlock(&leader->proc_lock);
720 spin_unlock(¤t->proc_lock);
721 proc_pid_flush(proc_dentry1);
722 proc_pid_flush(proc_dentry2);
724 if (state != TASK_ZOMBIE)
726 release_task(leader);
731 write_lock_irq(&tasklist_lock);
732 spin_lock(&oldsighand->siglock);
733 spin_lock(&newsighand->siglock);
735 if (current == oldsig->curr_target)
736 oldsig->curr_target = next_thread(current);
738 current->signal = newsig;
739 current->sighand = newsighand;
740 init_sigpending(¤t->pending);
743 spin_unlock(&newsighand->siglock);
744 spin_unlock(&oldsighand->siglock);
745 write_unlock_irq(&tasklist_lock);
747 if (newsig && atomic_dec_and_test(&oldsig->count))
748 kmem_cache_free(signal_cachep, oldsig);
750 if (atomic_dec_and_test(&oldsighand->count))
751 kmem_cache_free(sighand_cachep, oldsighand);
753 if (!thread_group_empty(current))
755 if (current->tgid != current->pid)
761 * These functions flushes out all traces of the currently running executable
762 * so that a new one can be started
765 static inline void flush_old_files(struct files_struct * files)
769 spin_lock(&files->file_lock);
771 unsigned long set, i;
775 if (i >= files->max_fds || i >= files->max_fdset)
777 set = files->close_on_exec->fds_bits[j];
780 files->close_on_exec->fds_bits[j] = 0;
781 spin_unlock(&files->file_lock);
782 for ( ; set ; i++,set >>= 1) {
787 spin_lock(&files->file_lock);
790 spin_unlock(&files->file_lock);
793 int flush_old_exec(struct linux_binprm * bprm)
797 struct files_struct *files;
800 * Make sure we have a private signal table and that
801 * we are unassociated from the previous thread group.
803 retval = de_thread(current);
808 * Make sure we have private file handles. Ask the
809 * fork helper to do the work for us and the exit
810 * helper to do the cleanup of the old one.
812 files = current->files; /* refcounted so safe to hold */
813 retval = unshare_files();
817 * Release all of the old mmap stuff
819 retval = exec_mmap(bprm->mm);
823 bprm->mm = NULL; /* We're using it now */
825 /* This is the point of no return */
827 put_files_struct(files);
829 current->sas_ss_sp = current->sas_ss_size = 0;
831 if (current->euid == current->uid && current->egid == current->gid)
832 current->mm->dumpable = 1;
833 name = bprm->filename;
834 for (i=0; (ch = *(name++)) != '\0';) {
839 current->comm[i++] = ch;
841 current->comm[i] = '\0';
845 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
846 permission(bprm->file->f_dentry->d_inode,MAY_READ, NULL))
847 current->mm->dumpable = 0;
849 /* An exec changes our domain. We are no longer part of the thread
852 current->self_exec_id++;
854 flush_signal_handlers(current, 0);
855 flush_old_files(current->files);
860 put_files_struct(current->files);
861 current->files = files;
866 EXPORT_SYMBOL(flush_old_exec);
869 * Fill the binprm structure from the inode.
870 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
872 int prepare_binprm(struct linux_binprm *bprm)
875 struct inode * inode = bprm->file->f_dentry->d_inode;
878 mode = inode->i_mode;
880 * Check execute perms again - if the caller has CAP_DAC_OVERRIDE,
881 * vfs_permission lets a non-executable through
883 if (!(mode & 0111)) /* with at least _one_ execute bit set */
885 if (bprm->file->f_op == NULL)
888 bprm->e_uid = current->euid;
889 bprm->e_gid = current->egid;
891 if(!(bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)) {
894 bprm->e_uid = inode->i_uid;
898 * If setgid is set but no group execute bit then this
899 * is a candidate for mandatory locking, not a setgid
902 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP))
903 bprm->e_gid = inode->i_gid;
906 /* fill in binprm security blob */
907 retval = security_bprm_set(bprm);
911 memset(bprm->buf,0,BINPRM_BUF_SIZE);
912 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
915 EXPORT_SYMBOL(prepare_binprm);
917 static inline int unsafe_exec(struct task_struct *p)
920 if (p->ptrace & PT_PTRACED) {
921 if (p->ptrace & PT_PTRACE_CAP)
922 unsafe |= LSM_UNSAFE_PTRACE_CAP;
924 unsafe |= LSM_UNSAFE_PTRACE;
926 if (atomic_read(&p->fs->count) > 1 ||
927 atomic_read(&p->files->count) > 1 ||
928 atomic_read(&p->sighand->count) > 1)
929 unsafe |= LSM_UNSAFE_SHARE;
934 void compute_creds(struct linux_binprm *bprm)
938 unsafe = unsafe_exec(current);
939 security_bprm_apply_creds(bprm, unsafe);
940 task_unlock(current);
943 EXPORT_SYMBOL(compute_creds);
945 void remove_arg_zero(struct linux_binprm *bprm)
948 unsigned long offset;
952 offset = bprm->p % PAGE_SIZE;
955 while (bprm->p++, *(kaddr+offset++)) {
956 if (offset != PAGE_SIZE)
959 kunmap_atomic(kaddr, KM_USER0);
961 page = bprm->page[bprm->p/PAGE_SIZE];
962 kaddr = kmap_atomic(page, KM_USER0);
964 kunmap_atomic(kaddr, KM_USER0);
969 EXPORT_SYMBOL(remove_arg_zero);
972 * cycle the list of binary formats handler, until one recognizes the image
974 int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
977 struct linux_binfmt *fmt;
979 /* handle /sbin/loader.. */
981 struct exec * eh = (struct exec *) bprm->buf;
983 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
984 (eh->fh.f_flags & 0x3000) == 0x3000)
987 unsigned long loader;
989 allow_write_access(bprm->file);
993 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
995 file = open_exec("/sbin/loader");
996 retval = PTR_ERR(file);
1000 /* Remember if the application is TASO. */
1001 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1004 bprm->loader = loader;
1005 retval = prepare_binprm(bprm);
1008 /* should call search_binary_handler recursively here,
1009 but it does not matter */
1013 retval = security_bprm_check(bprm);
1017 /* kernel module loader fixup */
1018 /* so we don't try to load run modprobe in kernel space. */
1020 for (try=0; try<2; try++) {
1021 read_lock(&binfmt_lock);
1022 for (fmt = formats ; fmt ; fmt = fmt->next) {
1023 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1026 if (!try_module_get(fmt->module))
1028 read_unlock(&binfmt_lock);
1029 retval = fn(bprm, regs);
1032 allow_write_access(bprm->file);
1036 current->did_exec = 1;
1039 read_lock(&binfmt_lock);
1041 if (retval != -ENOEXEC || bprm->mm == NULL)
1044 read_unlock(&binfmt_lock);
1048 read_unlock(&binfmt_lock);
1049 if (retval != -ENOEXEC || bprm->mm == NULL) {
1053 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1054 if (printable(bprm->buf[0]) &&
1055 printable(bprm->buf[1]) &&
1056 printable(bprm->buf[2]) &&
1057 printable(bprm->buf[3]))
1058 break; /* -ENOEXEC */
1059 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1066 EXPORT_SYMBOL(search_binary_handler);
1069 * sys_execve() executes a new program.
1071 int do_execve(char * filename,
1072 char __user *__user *argv,
1073 char __user *__user *envp,
1074 struct pt_regs * regs)
1076 struct linux_binprm bprm;
1081 file = open_exec(filename);
1083 retval = PTR_ERR(file);
1087 sched_balance_exec();
1089 bprm.p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1090 memset(bprm.page, 0, MAX_ARG_PAGES*sizeof(bprm.page[0]));
1093 bprm.filename = filename;
1094 bprm.interp = filename;
1098 bprm.security = NULL;
1099 bprm.mm = mm_alloc();
1104 retval = init_new_context(current, bprm.mm);
1108 bprm.argc = count(argv, bprm.p / sizeof(void *));
1109 if ((retval = bprm.argc) < 0)
1112 bprm.envc = count(envp, bprm.p / sizeof(void *));
1113 if ((retval = bprm.envc) < 0)
1116 retval = security_bprm_alloc(&bprm);
1120 retval = prepare_binprm(&bprm);
1124 retval = copy_strings_kernel(1, &bprm.filename, &bprm);
1129 retval = copy_strings(bprm.envc, envp, &bprm);
1133 retval = copy_strings(bprm.argc, argv, &bprm);
1137 retval = search_binary_handler(&bprm,regs);
1139 free_arg_pages(&bprm);
1141 /* execve success */
1142 security_bprm_free(&bprm);
1147 /* Something went wrong, return the inode and free the argument pages*/
1148 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1149 struct page * page = bprm.page[i];
1155 security_bprm_free(&bprm);
1163 allow_write_access(bprm.file);
1169 EXPORT_SYMBOL(do_execve);
1171 int set_binfmt(struct linux_binfmt *new)
1173 struct linux_binfmt *old = current->binfmt;
1176 if (!try_module_get(new->module))
1179 current->binfmt = new;
1181 module_put(old->module);
1185 EXPORT_SYMBOL(set_binfmt);
1187 #define CORENAME_MAX_SIZE 64
1189 /* format_corename will inspect the pattern parameter, and output a
1190 * name into corename, which must have space for at least
1191 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1193 void format_corename(char *corename, const char *pattern, long signr)
1195 const char *pat_ptr = pattern;
1196 char *out_ptr = corename;
1197 char *const out_end = corename + CORENAME_MAX_SIZE;
1199 int pid_in_pattern = 0;
1201 /* Repeat as long as we have more pattern to process and more output
1204 if (*pat_ptr != '%') {
1205 if (out_ptr == out_end)
1207 *out_ptr++ = *pat_ptr++;
1209 switch (*++pat_ptr) {
1212 /* Double percent, output one percent */
1214 if (out_ptr == out_end)
1221 rc = snprintf(out_ptr, out_end - out_ptr,
1222 "%d", current->tgid);
1223 if (rc > out_end - out_ptr)
1229 rc = snprintf(out_ptr, out_end - out_ptr,
1230 "%d", current->uid);
1231 if (rc > out_end - out_ptr)
1237 rc = snprintf(out_ptr, out_end - out_ptr,
1238 "%d", current->gid);
1239 if (rc > out_end - out_ptr)
1243 /* signal that caused the coredump */
1245 rc = snprintf(out_ptr, out_end - out_ptr,
1247 if (rc > out_end - out_ptr)
1251 /* UNIX time of coredump */
1254 do_gettimeofday(&tv);
1255 rc = snprintf(out_ptr, out_end - out_ptr,
1257 if (rc > out_end - out_ptr)
1264 down_read(&uts_sem);
1265 rc = snprintf(out_ptr, out_end - out_ptr,
1266 "%s", system_utsname.nodename);
1268 if (rc > out_end - out_ptr)
1274 rc = snprintf(out_ptr, out_end - out_ptr,
1275 "%s", current->comm);
1276 if (rc > out_end - out_ptr)
1286 /* Backward compatibility with core_uses_pid:
1288 * If core_pattern does not include a %p (as is the default)
1289 * and core_uses_pid is set, then .%pid will be appended to
1292 && (core_uses_pid || atomic_read(¤t->mm->mm_users) != 1)) {
1293 rc = snprintf(out_ptr, out_end - out_ptr,
1294 ".%d", current->tgid);
1295 if (rc > out_end - out_ptr)
1303 static void zap_threads (struct mm_struct *mm)
1305 struct task_struct *g, *p;
1306 struct task_struct *tsk = current;
1307 struct completion *vfork_done = tsk->vfork_done;
1310 * Make sure nobody is waiting for us to release the VM,
1311 * otherwise we can deadlock when we wait on each other
1314 tsk->vfork_done = NULL;
1315 complete(vfork_done);
1318 read_lock(&tasklist_lock);
1320 if (mm == p->mm && p != tsk) {
1321 force_sig_specific(SIGKILL, p);
1324 while_each_thread(g,p);
1326 read_unlock(&tasklist_lock);
1329 static void coredump_wait(struct mm_struct *mm)
1331 DECLARE_COMPLETION(startup_done);
1333 mm->core_waiters++; /* let other threads block */
1334 mm->core_startup_done = &startup_done;
1336 /* give other threads a chance to run: */
1340 if (--mm->core_waiters) {
1341 up_write(&mm->mmap_sem);
1342 wait_for_completion(&startup_done);
1344 up_write(&mm->mmap_sem);
1345 BUG_ON(mm->core_waiters);
1348 int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1350 char corename[CORENAME_MAX_SIZE + 1];
1351 struct mm_struct *mm = current->mm;
1352 struct linux_binfmt * binfmt;
1353 struct inode * inode;
1358 binfmt = current->binfmt;
1359 if (!binfmt || !binfmt->core_dump)
1361 down_write(&mm->mmap_sem);
1362 if (!mm->dumpable) {
1363 up_write(&mm->mmap_sem);
1367 init_completion(&mm->core_done);
1368 current->signal->group_exit = 1;
1369 current->signal->group_exit_code = exit_code;
1372 if (current->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1375 format_corename(corename, core_pattern, signr);
1376 file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE, 0600);
1379 inode = file->f_dentry->d_inode;
1380 if (inode->i_nlink > 1)
1381 goto close_fail; /* multiple links - don't dump */
1382 if (d_unhashed(file->f_dentry))
1385 if (!S_ISREG(inode->i_mode))
1389 if (!file->f_op->write)
1391 if (do_truncate(file->f_dentry, 0) != 0)
1394 retval = binfmt->core_dump(signr, regs, file);
1396 current->signal->group_exit_code |= 0x80;
1398 filp_close(file, NULL);
1400 complete_all(&mm->core_done);