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/slab.h>
26 #include <linux/file.h>
27 #include <linux/mman.h>
28 #include <linux/a.out.h>
29 #include <linux/stat.h>
30 #include <linux/fcntl.h>
31 #include <linux/smp_lock.h>
32 #include <linux/init.h>
33 #include <linux/pagemap.h>
34 #include <linux/highmem.h>
35 #include <linux/spinlock.h>
36 #include <linux/key.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/tracehook.h>
45 #include <linux/mount.h>
46 #include <linux/security.h>
47 #include <linux/syscalls.h>
48 #include <linux/rmap.h>
49 #include <linux/acct.h>
50 #include <linux/cn_proc.h>
51 #include <linux/audit.h>
52 #include <linux/vs_base.h>
53 #include <linux/vs_memory.h>
54 #include <linux/vs_cvirt.h>
56 #include <asm/uaccess.h>
57 #include <asm/mmu_context.h>
60 #include <linux/kmod.h>
64 char core_pattern[65] = "core";
65 int suid_dumpable = 0;
67 EXPORT_SYMBOL(suid_dumpable);
68 /* The maximal length of core_pattern is also specified in sysctl.c */
70 static struct linux_binfmt *formats;
71 static DEFINE_RWLOCK(binfmt_lock);
73 int register_binfmt(struct linux_binfmt * fmt)
75 struct linux_binfmt ** tmp = &formats;
81 write_lock(&binfmt_lock);
84 write_unlock(&binfmt_lock);
91 write_unlock(&binfmt_lock);
95 EXPORT_SYMBOL(register_binfmt);
97 int unregister_binfmt(struct linux_binfmt * fmt)
99 struct linux_binfmt ** tmp = &formats;
101 write_lock(&binfmt_lock);
105 write_unlock(&binfmt_lock);
110 write_unlock(&binfmt_lock);
114 EXPORT_SYMBOL(unregister_binfmt);
116 static inline void put_binfmt(struct linux_binfmt * fmt)
118 module_put(fmt->module);
122 * Note that a shared library must be both readable and executable due to
125 * Also note that we take the address to load from from the file itself.
127 asmlinkage long sys_uselib(const char __user * library)
133 error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
138 if (!S_ISREG(nd.dentry->d_inode->i_mode))
141 error = vfs_permission(&nd, MAY_READ | MAY_EXEC);
145 file = nameidata_to_filp(&nd, O_RDONLY);
146 error = PTR_ERR(file);
152 struct linux_binfmt * fmt;
154 read_lock(&binfmt_lock);
155 for (fmt = formats ; fmt ; fmt = fmt->next) {
156 if (!fmt->load_shlib)
158 if (!try_module_get(fmt->module))
160 read_unlock(&binfmt_lock);
161 error = fmt->load_shlib(file);
162 read_lock(&binfmt_lock);
164 if (error != -ENOEXEC)
167 read_unlock(&binfmt_lock);
173 release_open_intent(&nd);
179 * count() counts the number of strings in array ARGV.
181 static int count(char __user * __user * argv, int max)
189 if (get_user(p, argv))
203 * 'copy_strings()' copies argument/environment strings from user
204 * memory to free pages in kernel mem. These are in a format ready
205 * to be put directly into the top of new user memory.
207 static int copy_strings(int argc, char __user * __user * argv,
208 struct linux_binprm *bprm)
210 struct page *kmapped_page = NULL;
219 if (get_user(str, argv+argc) ||
220 !(len = strnlen_user(str, bprm->p))) {
231 /* XXX: add architecture specific overflow check here. */
236 int offset, bytes_to_copy;
239 offset = pos % PAGE_SIZE;
241 page = bprm->page[i];
244 page = alloc_page(GFP_HIGHUSER);
245 bprm->page[i] = page;
253 if (page != kmapped_page) {
255 kunmap(kmapped_page);
257 kaddr = kmap(kmapped_page);
260 memset(kaddr, 0, offset);
261 bytes_to_copy = PAGE_SIZE - offset;
262 if (bytes_to_copy > len) {
265 memset(kaddr+offset+len, 0,
266 PAGE_SIZE-offset-len);
268 err = copy_from_user(kaddr+offset, str, bytes_to_copy);
274 pos += bytes_to_copy;
275 str += bytes_to_copy;
276 len -= bytes_to_copy;
282 kunmap(kmapped_page);
287 * Like copy_strings, but get argv and its values from kernel memory.
289 int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
292 mm_segment_t oldfs = get_fs();
294 r = copy_strings(argc, (char __user * __user *)argv, bprm);
299 EXPORT_SYMBOL(copy_strings_kernel);
303 * This routine is used to map in a page into an address space: needed by
304 * execve() for the initial stack and environment pages.
306 * vma->vm_mm->mmap_sem is held for writing.
308 void install_arg_page(struct vm_area_struct *vma,
309 struct page *page, unsigned long address)
311 struct mm_struct *mm = vma->vm_mm;
315 if (unlikely(anon_vma_prepare(vma)))
318 flush_dcache_page(page);
319 pte = get_locked_pte(mm, address, &ptl);
322 if (!pte_none(*pte)) {
323 pte_unmap_unlock(pte, ptl);
326 inc_mm_counter(mm, anon_rss);
327 lru_cache_add_active(page);
328 set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte(
329 page, vma->vm_page_prot))));
330 page_add_new_anon_rmap(page, vma, address);
331 pte_unmap_unlock(pte, ptl);
333 /* no need for flush_tlb */
337 force_sig(SIGKILL, current);
340 #define EXTRA_STACK_VM_PAGES 20 /* random */
342 int setup_arg_pages(struct linux_binprm *bprm,
343 unsigned long stack_top,
344 int executable_stack)
346 unsigned long stack_base;
347 struct vm_area_struct *mpnt;
348 struct mm_struct *mm = current->mm;
352 #ifdef CONFIG_STACK_GROWSUP
353 /* Move the argument and environment strings to the bottom of the
359 /* Start by shifting all the pages down */
361 for (j = 0; j < MAX_ARG_PAGES; j++) {
362 struct page *page = bprm->page[j];
365 bprm->page[i++] = page;
368 /* Now move them within their pages */
369 offset = bprm->p % PAGE_SIZE;
370 to = kmap(bprm->page[0]);
371 for (j = 1; j < i; j++) {
372 memmove(to, to + offset, PAGE_SIZE - offset);
373 from = kmap(bprm->page[j]);
374 memcpy(to + PAGE_SIZE - offset, from, offset);
375 kunmap(bprm->page[j - 1]);
378 memmove(to, to + offset, PAGE_SIZE - offset);
379 kunmap(bprm->page[j - 1]);
381 /* Limit stack size to 1GB */
382 stack_base = current->signal->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 /* Adjust bprm->p to point to the end of the strings. */
388 bprm->p = stack_base + PAGE_SIZE * i - offset;
390 mm->arg_start = stack_base;
391 arg_size = i << PAGE_SHIFT;
393 /* zero pages that were copied above */
394 while (i < MAX_ARG_PAGES)
395 bprm->page[i++] = NULL;
397 stack_base = arch_align_stack(stack_top - MAX_ARG_PAGES*PAGE_SIZE);
398 stack_base = PAGE_ALIGN(stack_base);
399 bprm->p += stack_base;
400 mm->arg_start = bprm->p;
401 arg_size = stack_top - (PAGE_MASK & (unsigned long) mm->arg_start);
404 arg_size += EXTRA_STACK_VM_PAGES * PAGE_SIZE;
407 bprm->loader += stack_base;
408 bprm->exec += stack_base;
410 mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
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 = stack_base + arg_size;
423 mpnt->vm_end = stack_top;
424 mpnt->vm_start = mpnt->vm_end - arg_size;
426 /* Adjust stack execute permissions; explicitly enable
427 * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X
428 * and leave alone (arch default) otherwise. */
429 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
430 mpnt->vm_flags = VM_STACK_FLAGS | VM_EXEC;
431 else if (executable_stack == EXSTACK_DISABLE_X)
432 mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC;
434 mpnt->vm_flags = VM_STACK_FLAGS;
435 mpnt->vm_flags |= mm->def_flags;
436 mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7];
437 if ((ret = insert_vm_struct(mm, mpnt))) {
438 up_write(&mm->mmap_sem);
439 kmem_cache_free(vm_area_cachep, mpnt);
442 vx_vmpages_sub(mm, mm->total_vm - vma_pages(mpnt));
443 mm->stack_vm = mm->total_vm;
446 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
447 struct page *page = bprm->page[i];
449 bprm->page[i] = NULL;
450 install_arg_page(mpnt, page, stack_base);
452 stack_base += PAGE_SIZE;
454 up_write(&mm->mmap_sem);
459 EXPORT_SYMBOL(setup_arg_pages);
461 #define free_arg_pages(bprm) do { } while (0)
465 static inline void free_arg_pages(struct linux_binprm *bprm)
469 for (i = 0; i < MAX_ARG_PAGES; i++) {
471 __free_page(bprm->page[i]);
472 bprm->page[i] = NULL;
476 #endif /* CONFIG_MMU */
478 struct file *open_exec(const char *name)
484 err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
488 struct inode *inode = nd.dentry->d_inode;
489 file = ERR_PTR(-EACCES);
490 if (!(nd.mnt->mnt_flags & MNT_NOEXEC) &&
491 S_ISREG(inode->i_mode)) {
492 int err = vfs_permission(&nd, MAY_EXEC);
495 file = nameidata_to_filp(&nd, O_RDONLY);
497 err = deny_write_access(file);
507 release_open_intent(&nd);
513 EXPORT_SYMBOL(open_exec);
515 int kernel_read(struct file *file, unsigned long offset,
516 char *addr, unsigned long count)
524 /* The cast to a user pointer is valid due to the set_fs() */
525 result = vfs_read(file, (void __user *)addr, count, &pos);
530 EXPORT_SYMBOL(kernel_read);
532 static int exec_mmap(struct mm_struct *mm)
534 struct task_struct *tsk;
535 struct mm_struct * old_mm, *active_mm;
537 /* Notify parent that we're no longer interested in the old VM */
539 old_mm = current->mm;
540 mm_release(tsk, old_mm);
544 * Make sure that if there is a core dump in progress
545 * for the old mm, we get out and die instead of going
546 * through with the exec. We must hold mmap_sem around
547 * checking core_waiters and changing tsk->mm. The
548 * core-inducing thread will increment core_waiters for
549 * each thread whose ->mm == old_mm.
551 down_read(&old_mm->mmap_sem);
552 if (unlikely(old_mm->core_waiters)) {
553 up_read(&old_mm->mmap_sem);
558 active_mm = tsk->active_mm;
561 activate_mm(active_mm, mm);
563 arch_pick_mmap_layout(mm);
565 up_read(&old_mm->mmap_sem);
566 BUG_ON(active_mm != old_mm);
575 * This function makes sure the current process has its own signal table,
576 * so that flush_signal_handlers can later reset the handlers without
577 * disturbing other processes. (Other processes might share the signal
578 * table via the CLONE_SIGHAND option to clone().)
580 static int de_thread(struct task_struct *tsk)
582 struct signal_struct *sig = tsk->signal;
583 struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
584 spinlock_t *lock = &oldsighand->siglock;
585 struct task_struct *leader = NULL;
589 * If we don't share sighandlers, then we aren't sharing anything
590 * and we can just re-use it all.
592 if (atomic_read(&oldsighand->count) <= 1) {
593 BUG_ON(atomic_read(&sig->count) != 1);
598 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
602 if (thread_group_empty(current))
603 goto no_thread_group;
606 * Kill all other threads in the thread group.
607 * We must hold tasklist_lock to call zap_other_threads.
609 read_lock(&tasklist_lock);
611 if (sig->flags & SIGNAL_GROUP_EXIT) {
613 * Another group action in progress, just
614 * return so that the signal is processed.
616 spin_unlock_irq(lock);
617 read_unlock(&tasklist_lock);
618 kmem_cache_free(sighand_cachep, newsighand);
623 * child_reaper ignores SIGKILL, change it now.
624 * Reparenting needs write_lock on tasklist_lock,
625 * so it is safe to do it under read_lock.
627 if (unlikely(current->group_leader == child_reaper))
628 child_reaper = current;
630 zap_other_threads(current);
631 read_unlock(&tasklist_lock);
634 * Account for the thread group leader hanging around:
637 if (!thread_group_leader(current)) {
640 * The SIGALRM timer survives the exec, but needs to point
641 * at us as the new group leader now. We have a race with
642 * a timer firing now getting the old leader, so we need to
643 * synchronize with any firing (by calling del_timer_sync)
644 * before we can safely let the old group leader die.
647 spin_unlock_irq(lock);
648 if (hrtimer_cancel(&sig->real_timer))
649 hrtimer_restart(&sig->real_timer);
652 while (atomic_read(&sig->count) > count) {
653 sig->group_exit_task = current;
654 sig->notify_count = count;
655 __set_current_state(TASK_UNINTERRUPTIBLE);
656 spin_unlock_irq(lock);
660 sig->group_exit_task = NULL;
661 sig->notify_count = 0;
662 spin_unlock_irq(lock);
665 * At this point all other threads have exited, all we have to
666 * do is to wait for the thread group leader to become inactive,
667 * and to assume its PID:
669 if (!thread_group_leader(current)) {
671 * Wait for the thread group leader to be a zombie.
672 * It should already be zombie at this point, most
675 leader = current->group_leader;
676 while (leader->exit_state != EXIT_ZOMBIE)
680 * The only record we have of the real-time age of a
681 * process, regardless of execs it's done, is start_time.
682 * All the past CPU time is accumulated in signal_struct
683 * from sister threads now dead. But in this non-leader
684 * exec, nothing survives from the original leader thread,
685 * whose birth marks the true age of this process now.
686 * When we take on its identity by switching to its PID, we
687 * also take its birthdate (always earlier than our own).
689 current->start_time = leader->start_time;
691 write_lock_irq(&tasklist_lock);
693 BUG_ON(leader->tgid != current->tgid);
694 BUG_ON(current->pid == current->tgid);
696 * An exec() starts a new thread group with the
697 * TGID of the previous thread group. Rehash the
698 * two threads with a switched PID, and release
699 * the former thread group leader:
702 /* Become a process group leader with the old leader's pid.
703 * Note: The old leader also uses thispid until release_task
704 * is called. Odd but simple and correct.
706 detach_pid(current, PIDTYPE_PID);
707 current->pid = leader->pid;
708 attach_pid(current, PIDTYPE_PID, current->pid);
709 attach_pid(current, PIDTYPE_PGID, current->signal->pgrp);
710 attach_pid(current, PIDTYPE_SID, current->signal->session);
711 list_replace_rcu(&leader->tasks, ¤t->tasks);
713 current->group_leader = current;
714 leader->group_leader = current;
716 /* Reduce leader to a thread */
717 detach_pid(leader, PIDTYPE_PGID);
718 detach_pid(leader, PIDTYPE_SID);
720 current->exit_signal = SIGCHLD;
722 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
723 leader->exit_state = EXIT_DEAD;
725 write_unlock_irq(&tasklist_lock);
729 * There may be one thread left which is just exiting,
730 * but it's safe to stop telling the group to kill themselves.
737 release_task(leader);
739 BUG_ON(atomic_read(&sig->count) != 1);
741 if (atomic_read(&oldsighand->count) == 1) {
743 * Now that we nuked the rest of the thread group,
744 * it turns out we are not sharing sighand any more either.
745 * So we can just keep it.
747 kmem_cache_free(sighand_cachep, newsighand);
750 * Move our state over to newsighand and switch it in.
752 atomic_set(&newsighand->count, 1);
753 memcpy(newsighand->action, oldsighand->action,
754 sizeof(newsighand->action));
756 write_lock_irq(&tasklist_lock);
757 spin_lock(&oldsighand->siglock);
758 spin_lock_nested(&newsighand->siglock, SINGLE_DEPTH_NESTING);
760 rcu_assign_pointer(current->sighand, newsighand);
763 spin_unlock(&newsighand->siglock);
764 spin_unlock(&oldsighand->siglock);
765 write_unlock_irq(&tasklist_lock);
767 if (atomic_dec_and_test(&oldsighand->count))
768 kmem_cache_free(sighand_cachep, oldsighand);
771 BUG_ON(!thread_group_leader(current));
776 * These functions flushes out all traces of the currently running executable
777 * so that a new one can be started
780 static void flush_old_files(struct files_struct * files)
785 spin_lock(&files->file_lock);
787 unsigned long set, i;
791 fdt = files_fdtable(files);
792 if (i >= fdt->max_fds || i >= fdt->max_fdset)
794 set = fdt->close_on_exec->fds_bits[j];
797 fdt->close_on_exec->fds_bits[j] = 0;
798 spin_unlock(&files->file_lock);
799 for ( ; set ; i++,set >>= 1) {
804 spin_lock(&files->file_lock);
807 spin_unlock(&files->file_lock);
810 void get_task_comm(char *buf, struct task_struct *tsk)
812 /* buf must be at least sizeof(tsk->comm) in size */
814 strncpy(buf, tsk->comm, sizeof(tsk->comm));
818 void set_task_comm(struct task_struct *tsk, char *buf)
821 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
825 int flush_old_exec(struct linux_binprm * bprm)
829 struct files_struct *files;
830 char tcomm[sizeof(current->comm)];
833 * Make sure we have a private signal table and that
834 * we are unassociated from the previous thread group.
836 retval = de_thread(current);
841 * Make sure we have private file handles. Ask the
842 * fork helper to do the work for us and the exit
843 * helper to do the cleanup of the old one.
845 files = current->files; /* refcounted so safe to hold */
846 retval = unshare_files();
850 * Release all of the old mmap stuff
852 retval = exec_mmap(bprm->mm);
856 bprm->mm = NULL; /* We're using it now */
858 /* This is the point of no return */
859 put_files_struct(files);
861 current->sas_ss_sp = current->sas_ss_size = 0;
863 if (current->euid == current->uid && current->egid == current->gid)
864 current->mm->dumpable = 1;
866 current->mm->dumpable = suid_dumpable;
868 name = bprm->filename;
870 /* Copies the binary name from after last slash */
871 for (i=0; (ch = *(name++)) != '\0';) {
873 i = 0; /* overwrite what we wrote */
875 if (i < (sizeof(tcomm) - 1))
879 set_task_comm(current, tcomm);
881 current->flags &= ~PF_RANDOMIZE;
884 /* Set the new mm task size. We have to do that late because it may
885 * depend on TIF_32BIT which is only updated in flush_thread() on
886 * some architectures like powerpc
888 current->mm->task_size = TASK_SIZE;
890 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
891 file_permission(bprm->file, MAY_READ) ||
892 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) {
894 current->mm->dumpable = suid_dumpable;
897 /* An exec changes our domain. We are no longer part of the thread
900 current->self_exec_id++;
902 flush_signal_handlers(current, 0);
903 flush_old_files(current->files);
908 put_files_struct(current->files);
909 current->files = files;
914 EXPORT_SYMBOL(flush_old_exec);
917 * Fill the binprm structure from the inode.
918 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
920 int prepare_binprm(struct linux_binprm *bprm)
923 struct inode * inode = bprm->file->f_dentry->d_inode;
926 mode = inode->i_mode;
927 if (bprm->file->f_op == NULL)
930 bprm->e_uid = current->euid;
931 bprm->e_gid = current->egid;
933 if(!(bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)) {
935 if (mode & S_ISUID) {
936 current->personality &= ~PER_CLEAR_ON_SETID;
937 bprm->e_uid = inode->i_uid;
942 * If setgid is set but no group execute bit then this
943 * is a candidate for mandatory locking, not a setgid
946 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
947 current->personality &= ~PER_CLEAR_ON_SETID;
948 bprm->e_gid = inode->i_gid;
952 /* fill in binprm security blob */
953 retval = security_bprm_set(bprm);
957 memset(bprm->buf,0,BINPRM_BUF_SIZE);
958 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
961 EXPORT_SYMBOL(prepare_binprm);
963 static int unsafe_exec(struct task_struct *p)
965 int unsafe = tracehook_unsafe_exec(p);
966 if (atomic_read(&p->fs->count) > 1 ||
967 atomic_read(&p->files->count) > 1 ||
968 atomic_read(&p->sighand->count) > 1)
969 unsafe |= LSM_UNSAFE_SHARE;
974 void compute_creds(struct linux_binprm *bprm)
978 if (bprm->e_uid != current->uid)
983 unsafe = unsafe_exec(current);
984 security_bprm_apply_creds(bprm, unsafe);
985 task_unlock(current);
986 security_bprm_post_apply_creds(bprm);
989 EXPORT_SYMBOL(compute_creds);
991 void remove_arg_zero(struct linux_binprm *bprm)
994 unsigned long offset;
998 offset = bprm->p % PAGE_SIZE;
1001 while (bprm->p++, *(kaddr+offset++)) {
1002 if (offset != PAGE_SIZE)
1005 kunmap_atomic(kaddr, KM_USER0);
1007 page = bprm->page[bprm->p/PAGE_SIZE];
1008 kaddr = kmap_atomic(page, KM_USER0);
1010 kunmap_atomic(kaddr, KM_USER0);
1015 EXPORT_SYMBOL(remove_arg_zero);
1018 * cycle the list of binary formats handler, until one recognizes the image
1020 int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1023 struct linux_binfmt *fmt;
1025 /* handle /sbin/loader.. */
1027 struct exec * eh = (struct exec *) bprm->buf;
1029 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1030 (eh->fh.f_flags & 0x3000) == 0x3000)
1033 unsigned long loader;
1035 allow_write_access(bprm->file);
1039 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1041 file = open_exec("/sbin/loader");
1042 retval = PTR_ERR(file);
1046 /* Remember if the application is TASO. */
1047 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1050 bprm->loader = loader;
1051 retval = prepare_binprm(bprm);
1054 /* should call search_binary_handler recursively here,
1055 but it does not matter */
1059 retval = security_bprm_check(bprm);
1063 /* kernel module loader fixup */
1064 /* so we don't try to load run modprobe in kernel space. */
1067 retval = audit_bprm(bprm);
1072 for (try=0; try<2; try++) {
1073 read_lock(&binfmt_lock);
1074 for (fmt = formats ; fmt ; fmt = fmt->next) {
1075 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1078 if (!try_module_get(fmt->module))
1080 read_unlock(&binfmt_lock);
1081 retval = fn(bprm, regs);
1084 allow_write_access(bprm->file);
1088 current->did_exec = 1;
1089 proc_exec_connector(current);
1090 tracehook_report_exec(bprm, regs);
1093 read_lock(&binfmt_lock);
1095 if (retval != -ENOEXEC || bprm->mm == NULL)
1098 read_unlock(&binfmt_lock);
1102 read_unlock(&binfmt_lock);
1103 if (retval != -ENOEXEC || bprm->mm == NULL) {
1107 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1108 if (printable(bprm->buf[0]) &&
1109 printable(bprm->buf[1]) &&
1110 printable(bprm->buf[2]) &&
1111 printable(bprm->buf[3]))
1112 break; /* -ENOEXEC */
1113 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1120 EXPORT_SYMBOL(search_binary_handler);
1123 * sys_execve() executes a new program.
1125 int do_execve(char * filename,
1126 char __user *__user *argv,
1127 char __user *__user *envp,
1128 struct pt_regs * regs)
1130 struct linux_binprm *bprm;
1136 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
1140 file = open_exec(filename);
1141 retval = PTR_ERR(file);
1147 bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1150 bprm->filename = filename;
1151 bprm->interp = filename;
1152 bprm->mm = mm_alloc();
1157 retval = init_new_context(current, bprm->mm);
1161 bprm->argc = count(argv, bprm->p / sizeof(void *));
1162 if ((retval = bprm->argc) < 0)
1165 bprm->envc = count(envp, bprm->p / sizeof(void *));
1166 if ((retval = bprm->envc) < 0)
1169 retval = security_bprm_alloc(bprm);
1173 retval = prepare_binprm(bprm);
1177 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1181 bprm->exec = bprm->p;
1182 retval = copy_strings(bprm->envc, envp, bprm);
1186 retval = copy_strings(bprm->argc, argv, bprm);
1190 retval = search_binary_handler(bprm,regs);
1192 free_arg_pages(bprm);
1194 /* execve success */
1195 security_bprm_free(bprm);
1196 acct_update_integrals(current);
1202 /* Something went wrong, return the inode and free the argument pages*/
1203 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1204 struct page * page = bprm->page[i];
1210 security_bprm_free(bprm);
1218 allow_write_access(bprm->file);
1229 int set_binfmt(struct linux_binfmt *new)
1231 struct linux_binfmt *old = current->binfmt;
1234 if (!try_module_get(new->module))
1237 current->binfmt = new;
1239 module_put(old->module);
1243 EXPORT_SYMBOL(set_binfmt);
1245 #define CORENAME_MAX_SIZE 64
1247 /* format_corename will inspect the pattern parameter, and output a
1248 * name into corename, which must have space for at least
1249 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1251 static void format_corename(char *corename, const char *pattern, long signr)
1253 const char *pat_ptr = pattern;
1254 char *out_ptr = corename;
1255 char *const out_end = corename + CORENAME_MAX_SIZE;
1257 int pid_in_pattern = 0;
1259 /* Repeat as long as we have more pattern to process and more output
1262 if (*pat_ptr != '%') {
1263 if (out_ptr == out_end)
1265 *out_ptr++ = *pat_ptr++;
1267 switch (*++pat_ptr) {
1270 /* Double percent, output one percent */
1272 if (out_ptr == out_end)
1279 rc = snprintf(out_ptr, out_end - out_ptr,
1280 "%d", current->tgid);
1281 if (rc > out_end - out_ptr)
1287 rc = snprintf(out_ptr, out_end - out_ptr,
1288 "%d", current->uid);
1289 if (rc > out_end - out_ptr)
1295 rc = snprintf(out_ptr, out_end - out_ptr,
1296 "%d", current->gid);
1297 if (rc > out_end - out_ptr)
1301 /* signal that caused the coredump */
1303 rc = snprintf(out_ptr, out_end - out_ptr,
1305 if (rc > out_end - out_ptr)
1309 /* UNIX time of coredump */
1312 do_gettimeofday(&tv);
1313 rc = snprintf(out_ptr, out_end - out_ptr,
1315 if (rc > out_end - out_ptr)
1322 down_read(&uts_sem);
1323 rc = snprintf(out_ptr, out_end - out_ptr,
1324 "%s", vx_new_uts(nodename));
1326 if (rc > out_end - out_ptr)
1332 rc = snprintf(out_ptr, out_end - out_ptr,
1333 "%s", current->comm);
1334 if (rc > out_end - out_ptr)
1344 /* Backward compatibility with core_uses_pid:
1346 * If core_pattern does not include a %p (as is the default)
1347 * and core_uses_pid is set, then .%pid will be appended to
1350 && (core_uses_pid || atomic_read(¤t->mm->mm_users) != 1)) {
1351 rc = snprintf(out_ptr, out_end - out_ptr,
1352 ".%d", current->tgid);
1353 if (rc > out_end - out_ptr)
1361 static void zap_process(struct task_struct *start)
1363 struct task_struct *t;
1365 start->signal->flags = SIGNAL_GROUP_EXIT;
1366 start->signal->group_stop_count = 0;
1370 if (t != current && t->mm) {
1371 t->mm->core_waiters++;
1372 sigaddset(&t->pending.signal, SIGKILL);
1373 signal_wake_up(t, 1);
1375 } while ((t = next_thread(t)) != start);
1378 static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
1381 struct task_struct *g, *p;
1382 unsigned long flags;
1385 spin_lock_irq(&tsk->sighand->siglock);
1386 if (!(tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
1387 tsk->signal->group_exit_code = exit_code;
1391 spin_unlock_irq(&tsk->sighand->siglock);
1395 if (atomic_read(&mm->mm_users) == mm->core_waiters + 1)
1399 for_each_process(g) {
1400 if (g == tsk->group_leader)
1408 * p->sighand can't disappear, but
1409 * may be changed by de_thread()
1411 lock_task_sighand(p, &flags);
1413 unlock_task_sighand(p, &flags);
1417 } while ((p = next_thread(p)) != g);
1421 return mm->core_waiters;
1424 static int coredump_wait(int exit_code)
1426 struct task_struct *tsk = current;
1427 struct mm_struct *mm = tsk->mm;
1428 struct completion startup_done;
1429 struct completion *vfork_done;
1432 init_completion(&mm->core_done);
1433 init_completion(&startup_done);
1434 mm->core_startup_done = &startup_done;
1436 core_waiters = zap_threads(tsk, mm, exit_code);
1437 up_write(&mm->mmap_sem);
1439 if (unlikely(core_waiters < 0))
1443 * Make sure nobody is waiting for us to release the VM,
1444 * otherwise we can deadlock when we wait on each other
1446 vfork_done = tsk->vfork_done;
1448 tsk->vfork_done = NULL;
1449 complete(vfork_done);
1453 wait_for_completion(&startup_done);
1455 BUG_ON(mm->core_waiters);
1456 return core_waiters;
1459 int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1461 char corename[CORENAME_MAX_SIZE + 1];
1462 struct mm_struct *mm = current->mm;
1463 struct linux_binfmt * binfmt;
1464 struct inode * inode;
1467 int fsuid = current->fsuid;
1470 binfmt = current->binfmt;
1471 if (!binfmt || !binfmt->core_dump)
1473 if (current->tux_exit)
1474 current->tux_exit();
1475 down_write(&mm->mmap_sem);
1476 if (!mm->dumpable) {
1477 up_write(&mm->mmap_sem);
1482 * We cannot trust fsuid as being the "true" uid of the
1483 * process nor do we know its entire history. We only know it
1484 * was tainted so we dump it as root in mode 2.
1486 if (mm->dumpable == 2) { /* Setuid core dump mode */
1487 flag = O_EXCL; /* Stop rewrite attacks */
1488 current->fsuid = 0; /* Dump root private */
1492 retval = coredump_wait(exit_code);
1497 * Clear any false indication of pending signals that might
1498 * be seen by the filesystem code called to write the core file.
1500 clear_thread_flag(TIF_SIGPENDING);
1502 if (current->signal->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1506 * lock_kernel() because format_corename() is controlled by sysctl, which
1507 * uses lock_kernel()
1510 format_corename(corename, core_pattern, signr);
1512 file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag, 0600);
1515 inode = file->f_dentry->d_inode;
1516 if (inode->i_nlink > 1)
1517 goto close_fail; /* multiple links - don't dump */
1518 if (d_unhashed(file->f_dentry))
1521 if (!S_ISREG(inode->i_mode))
1525 if (!file->f_op->write)
1527 if (do_truncate(file->f_dentry, 0, 0, file) != 0)
1530 retval = binfmt->core_dump(signr, regs, file);
1533 current->signal->group_exit_code |= 0x80;
1535 filp_close(file, NULL);
1537 current->fsuid = fsuid;
1538 complete_all(&mm->core_done);