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>
49 #include <linux/ckrm.h>
50 #include <linux/vs_memory.h>
52 #include <asm/uaccess.h>
53 #include <asm/mmu_context.h>
56 #include <linux/kmod.h>
60 char core_pattern[65] = "core";
61 /* The maximal length of core_pattern is also specified in sysctl.c */
63 static struct linux_binfmt *formats;
64 static rwlock_t binfmt_lock = RW_LOCK_UNLOCKED;
66 int register_binfmt(struct linux_binfmt * fmt)
68 struct linux_binfmt ** tmp = &formats;
74 write_lock(&binfmt_lock);
77 write_unlock(&binfmt_lock);
84 write_unlock(&binfmt_lock);
88 EXPORT_SYMBOL(register_binfmt);
90 int unregister_binfmt(struct linux_binfmt * fmt)
92 struct linux_binfmt ** tmp = &formats;
94 write_lock(&binfmt_lock);
98 write_unlock(&binfmt_lock);
103 write_unlock(&binfmt_lock);
107 EXPORT_SYMBOL(unregister_binfmt);
109 static inline void put_binfmt(struct linux_binfmt * fmt)
111 module_put(fmt->module);
115 * Note that a shared library must be both readable and executable due to
118 * Also note that we take the address to load from from the file itself.
120 asmlinkage long sys_uselib(const char __user * library)
126 nd.intent.open.flags = FMODE_READ;
127 error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
132 if (!S_ISREG(nd.dentry->d_inode->i_mode))
135 error = permission(nd.dentry->d_inode, MAY_READ | MAY_EXEC, &nd);
139 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
140 error = PTR_ERR(file);
146 struct linux_binfmt * fmt;
148 read_lock(&binfmt_lock);
149 for (fmt = formats ; fmt ; fmt = fmt->next) {
150 if (!fmt->load_shlib)
152 if (!try_module_get(fmt->module))
154 read_unlock(&binfmt_lock);
155 error = fmt->load_shlib(file);
156 read_lock(&binfmt_lock);
158 if (error != -ENOEXEC)
161 read_unlock(&binfmt_lock);
172 * count() counts the number of strings in array ARGV.
174 static int count(char __user * __user * argv, int max)
182 if (get_user(p, argv))
195 * 'copy_strings()' copies argument/environment strings from user
196 * memory to free pages in kernel mem. These are in a format ready
197 * to be put directly into the top of new user memory.
199 int copy_strings(int argc,char __user * __user * argv, struct linux_binprm *bprm)
201 struct page *kmapped_page = NULL;
210 if (get_user(str, argv+argc) ||
211 !(len = strnlen_user(str, bprm->p))) {
222 /* XXX: add architecture specific overflow check here. */
227 int offset, bytes_to_copy;
230 offset = pos % PAGE_SIZE;
232 page = bprm->page[i];
235 page = alloc_page(GFP_HIGHUSER);
236 bprm->page[i] = page;
244 if (page != kmapped_page) {
246 kunmap(kmapped_page);
248 kaddr = kmap(kmapped_page);
251 memset(kaddr, 0, offset);
252 bytes_to_copy = PAGE_SIZE - offset;
253 if (bytes_to_copy > len) {
256 memset(kaddr+offset+len, 0,
257 PAGE_SIZE-offset-len);
259 err = copy_from_user(kaddr+offset, str, bytes_to_copy);
265 pos += bytes_to_copy;
266 str += bytes_to_copy;
267 len -= bytes_to_copy;
273 kunmap(kmapped_page);
278 * Like copy_strings, but get argv and its values from kernel memory.
280 int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
283 mm_segment_t oldfs = get_fs();
285 r = copy_strings(argc, (char __user * __user *)argv, bprm);
290 EXPORT_SYMBOL(copy_strings_kernel);
294 * This routine is used to map in a page into an address space: needed by
295 * execve() for the initial stack and environment pages.
297 * vma->vm_mm->mmap_sem is held for writing.
299 void install_arg_page(struct vm_area_struct *vma,
300 struct page *page, unsigned long address)
302 struct mm_struct *mm = vma->vm_mm;
307 if (unlikely(anon_vma_prepare(vma)))
310 flush_dcache_page(page);
311 pgd = pgd_offset(mm, address);
313 spin_lock(&mm->page_table_lock);
314 pmd = pmd_alloc(mm, pgd, address);
317 pte = pte_alloc_map(mm, pmd, address);
320 if (!pte_none(*pte)) {
326 lru_cache_add_active(page);
327 set_pte(pte, pte_mkdirty(pte_mkwrite(mk_pte(
328 page, vma->vm_page_prot))));
329 page_add_anon_rmap(page, vma, address);
331 spin_unlock(&mm->page_table_lock);
333 /* no need for flush_tlb */
336 spin_unlock(&mm->page_table_lock);
339 force_sig(SIGKILL, current);
342 int setup_arg_pages(struct linux_binprm *bprm, int executable_stack)
344 unsigned long stack_base;
345 struct vm_area_struct *mpnt;
346 struct mm_struct *mm = current->mm;
350 #ifdef CONFIG_STACK_GROWSUP
351 /* Move the argument and environment strings to the bottom of the
357 /* Start by shifting all the pages down */
359 for (j = 0; j < MAX_ARG_PAGES; j++) {
360 struct page *page = bprm->page[j];
363 bprm->page[i++] = page;
366 /* Now move them within their pages */
367 offset = bprm->p % PAGE_SIZE;
368 to = kmap(bprm->page[0]);
369 for (j = 1; j < i; j++) {
370 memmove(to, to + offset, PAGE_SIZE - offset);
371 from = kmap(bprm->page[j]);
372 memcpy(to + PAGE_SIZE - offset, from, offset);
373 kunmap(bprm->page[j - 1]);
376 memmove(to, to + offset, PAGE_SIZE - offset);
377 kunmap(bprm->page[j - 1]);
379 /* Adjust bprm->p to point to the end of the strings. */
380 bprm->p = PAGE_SIZE * i - offset;
382 /* Limit stack size to 1GB */
383 stack_base = current->rlim[RLIMIT_STACK].rlim_max;
384 if (stack_base > (1 << 30))
385 stack_base = 1 << 30;
386 stack_base = PAGE_ALIGN(STACK_TOP - stack_base);
388 mm->arg_start = stack_base;
389 arg_size = i << PAGE_SHIFT;
391 /* zero pages that were copied above */
392 while (i < MAX_ARG_PAGES)
393 bprm->page[i++] = NULL;
395 #ifdef __HAVE_ARCH_ALIGN_STACK
396 stack_base = arch_align_stack(STACK_TOP - MAX_ARG_PAGES*PAGE_SIZE);
397 stack_base = PAGE_ALIGN(stack_base);
399 stack_base = STACK_TOP - MAX_ARG_PAGES * PAGE_SIZE;
401 mm->arg_start = bprm->p + stack_base;
402 arg_size = STACK_TOP - (PAGE_MASK & (unsigned long) mm->arg_start);
405 bprm->p += stack_base;
407 bprm->loader += stack_base;
408 bprm->exec += stack_base;
410 mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
414 if (security_vm_enough_memory(arg_size >> PAGE_SHIFT) ||
415 !vx_vmpages_avail(mm, arg_size >> PAGE_SHIFT)) {
416 kmem_cache_free(vm_area_cachep, mpnt);
420 memset(mpnt, 0, sizeof(*mpnt));
422 down_write(&mm->mmap_sem);
425 #ifdef CONFIG_STACK_GROWSUP
426 mpnt->vm_start = stack_base;
427 mpnt->vm_end = PAGE_MASK &
428 (PAGE_SIZE - 1 + (unsigned long) bprm->p);
430 mpnt->vm_start = PAGE_MASK & (unsigned long) bprm->p;
431 mpnt->vm_end = STACK_TOP;
433 /* Adjust stack execute permissions; explicitly enable
434 * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X
435 * and leave alone (arch default) otherwise. */
436 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
437 mpnt->vm_flags = VM_STACK_FLAGS | VM_EXEC;
438 else if (executable_stack == EXSTACK_DISABLE_X)
439 mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC;
441 mpnt->vm_flags = VM_STACK_FLAGS;
442 mpnt->vm_flags |= mm->def_flags;
443 mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7];
444 insert_vm_struct(mm, mpnt);
445 // mm->total_vm = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
446 vx_vmpages_sub(mm, mm->total_vm -
447 ((mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT));
450 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
451 struct page *page = bprm->page[i];
453 bprm->page[i] = NULL;
454 install_arg_page(mpnt, page, stack_base);
456 stack_base += PAGE_SIZE;
458 up_write(&mm->mmap_sem);
463 EXPORT_SYMBOL(setup_arg_pages);
465 #define free_arg_pages(bprm) do { } while (0)
469 static inline void free_arg_pages(struct linux_binprm *bprm)
473 for (i = 0; i < MAX_ARG_PAGES; i++) {
475 __free_page(bprm->page[i]);
476 bprm->page[i] = NULL;
480 #endif /* CONFIG_MMU */
482 struct file *open_exec(const char *name)
488 nd.intent.open.flags = FMODE_READ;
489 err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
493 struct inode *inode = nd.dentry->d_inode;
494 file = ERR_PTR(-EACCES);
495 if (!(nd.mnt->mnt_flags & MNT_NOEXEC) &&
496 S_ISREG(inode->i_mode)) {
497 int err = permission(inode, MAY_EXEC, &nd);
498 if (!err && !(inode->i_mode & 0111))
502 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
504 err = deny_write_access(file);
519 EXPORT_SYMBOL(open_exec);
521 int kernel_read(struct file *file, unsigned long offset,
522 char *addr, unsigned long count)
530 /* The cast to a user pointer is valid due to the set_fs() */
531 result = vfs_read(file, (void __user *)addr, count, &pos);
536 EXPORT_SYMBOL(kernel_read);
538 static int exec_mmap(struct mm_struct *mm)
540 struct task_struct *tsk;
541 struct mm_struct * old_mm, *active_mm;
543 /* Add it to the list of mm's */
544 spin_lock(&mmlist_lock);
545 list_add(&mm->mmlist, &init_mm.mmlist);
547 spin_unlock(&mmlist_lock);
549 /* Notify parent that we're no longer interested in the old VM */
551 old_mm = current->mm;
552 mm_release(tsk, old_mm);
555 active_mm = tsk->active_mm;
558 activate_mm(active_mm, mm);
560 arch_pick_mmap_layout(mm);
562 if (active_mm != old_mm) BUG();
571 * This function makes sure the current process has its own signal table,
572 * so that flush_signal_handlers can later reset the handlers without
573 * disturbing other processes. (Other processes might share the signal
574 * table via the CLONE_SIGHAND option to clone().)
576 static inline int de_thread(struct task_struct *tsk)
578 struct signal_struct *newsig, *oldsig = tsk->signal;
579 struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
580 spinlock_t *lock = &oldsighand->siglock;
584 * If we don't share sighandlers, then we aren't sharing anything
585 * and we can just re-use it all.
587 if (atomic_read(&oldsighand->count) <= 1)
590 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
594 spin_lock_init(&newsighand->siglock);
595 atomic_set(&newsighand->count, 1);
596 memcpy(newsighand->action, oldsighand->action, sizeof(newsighand->action));
599 * See if we need to allocate a new signal structure
602 if (atomic_read(&oldsig->count) > 1) {
603 newsig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
605 kmem_cache_free(sighand_cachep, newsighand);
608 atomic_set(&newsig->count, 1);
609 newsig->group_exit = 0;
610 newsig->group_exit_code = 0;
611 newsig->group_exit_task = NULL;
612 newsig->group_stop_count = 0;
613 newsig->curr_target = NULL;
614 init_sigpending(&newsig->shared_pending);
615 INIT_LIST_HEAD(&newsig->posix_timers);
617 newsig->tty = oldsig->tty;
618 newsig->pgrp = oldsig->pgrp;
619 newsig->session = oldsig->session;
620 newsig->leader = oldsig->leader;
621 newsig->tty_old_pgrp = oldsig->tty_old_pgrp;
624 if (thread_group_empty(current))
625 goto no_thread_group;
628 * Kill all other threads in the thread group.
629 * We must hold tasklist_lock to call zap_other_threads.
631 read_lock(&tasklist_lock);
633 if (oldsig->group_exit) {
635 * Another group action in progress, just
636 * return so that the signal is processed.
638 spin_unlock_irq(lock);
639 read_unlock(&tasklist_lock);
640 kmem_cache_free(sighand_cachep, newsighand);
642 kmem_cache_free(signal_cachep, newsig);
645 oldsig->group_exit = 1;
646 zap_other_threads(current);
647 read_unlock(&tasklist_lock);
650 * Account for the thread group leader hanging around:
653 if (current->pid == current->tgid)
655 while (atomic_read(&oldsig->count) > count) {
656 oldsig->group_exit_task = current;
657 oldsig->notify_count = count;
658 __set_current_state(TASK_UNINTERRUPTIBLE);
659 spin_unlock_irq(lock);
663 spin_unlock_irq(lock);
666 * At this point all other threads have exited, all we have to
667 * do is to wait for the thread group leader to become inactive,
668 * and to assume its PID:
670 if (current->pid != current->tgid) {
671 struct task_struct *leader = current->group_leader, *parent;
672 struct dentry *proc_dentry1, *proc_dentry2;
673 unsigned long state, ptrace;
676 * Wait for the thread group leader to be a zombie.
677 * It should already be zombie at this point, most
680 while (leader->state != TASK_ZOMBIE)
683 spin_lock(&leader->proc_lock);
684 spin_lock(¤t->proc_lock);
685 proc_dentry1 = proc_pid_unhash(current);
686 proc_dentry2 = proc_pid_unhash(leader);
687 write_lock_irq(&tasklist_lock);
689 if (leader->tgid != current->tgid)
691 if (current->pid == current->tgid)
694 * An exec() starts a new thread group with the
695 * TGID of the previous thread group. Rehash the
696 * two threads with a switched PID, and release
697 * the former thread group leader:
699 ptrace = leader->ptrace;
700 parent = leader->parent;
702 ptrace_unlink(current);
703 ptrace_unlink(leader);
704 remove_parent(current);
705 remove_parent(leader);
707 switch_exec_pids(leader, current);
709 current->parent = current->real_parent = leader->real_parent;
710 leader->parent = leader->real_parent = child_reaper;
711 current->group_leader = current;
712 leader->group_leader = leader;
714 add_parent(current, current->parent);
715 add_parent(leader, leader->parent);
717 current->ptrace = ptrace;
718 __ptrace_link(current, parent);
721 list_del(¤t->tasks);
722 list_add_tail(¤t->tasks, &init_task.tasks);
723 current->exit_signal = SIGCHLD;
724 state = leader->state;
726 write_unlock_irq(&tasklist_lock);
727 spin_unlock(&leader->proc_lock);
728 spin_unlock(¤t->proc_lock);
729 proc_pid_flush(proc_dentry1);
730 proc_pid_flush(proc_dentry2);
732 if (state != TASK_ZOMBIE)
734 release_task(leader);
739 write_lock_irq(&tasklist_lock);
740 spin_lock(&oldsighand->siglock);
741 spin_lock(&newsighand->siglock);
743 if (current == oldsig->curr_target)
744 oldsig->curr_target = next_thread(current);
746 current->signal = newsig;
747 current->sighand = newsighand;
748 init_sigpending(¤t->pending);
751 spin_unlock(&newsighand->siglock);
752 spin_unlock(&oldsighand->siglock);
753 write_unlock_irq(&tasklist_lock);
755 if (newsig && atomic_dec_and_test(&oldsig->count))
756 kmem_cache_free(signal_cachep, oldsig);
758 if (atomic_dec_and_test(&oldsighand->count))
759 kmem_cache_free(sighand_cachep, oldsighand);
761 if (!thread_group_empty(current))
763 if (current->tgid != current->pid)
769 * These functions flushes out all traces of the currently running executable
770 * so that a new one can be started
773 static inline void flush_old_files(struct files_struct * files)
777 spin_lock(&files->file_lock);
779 unsigned long set, i;
783 if (i >= files->max_fds || i >= files->max_fdset)
785 set = files->close_on_exec->fds_bits[j];
788 files->close_on_exec->fds_bits[j] = 0;
789 spin_unlock(&files->file_lock);
790 for ( ; set ; i++,set >>= 1) {
795 spin_lock(&files->file_lock);
798 spin_unlock(&files->file_lock);
801 int flush_old_exec(struct linux_binprm * bprm)
805 struct files_struct *files;
808 * Make sure we have a private signal table and that
809 * we are unassociated from the previous thread group.
811 retval = de_thread(current);
816 * Make sure we have private file handles. Ask the
817 * fork helper to do the work for us and the exit
818 * helper to do the cleanup of the old one.
820 files = current->files; /* refcounted so safe to hold */
821 retval = unshare_files();
825 * Release all of the old mmap stuff
827 retval = exec_mmap(bprm->mm);
831 bprm->mm = NULL; /* We're using it now */
833 /* This is the point of no return */
835 put_files_struct(files);
837 current->sas_ss_sp = current->sas_ss_size = 0;
839 if (current->euid == current->uid && current->egid == current->gid)
840 current->mm->dumpable = 1;
841 name = bprm->filename;
842 for (i=0; (ch = *(name++)) != '\0';) {
847 current->comm[i++] = ch;
849 current->comm[i] = '\0';
851 current->flags &= ~PF_RELOCEXEC;
854 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
855 permission(bprm->file->f_dentry->d_inode,MAY_READ, NULL) ||
856 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP))
857 current->mm->dumpable = 0;
859 /* An exec changes our domain. We are no longer part of the thread
862 current->self_exec_id++;
864 flush_signal_handlers(current, 0);
865 flush_old_files(current->files);
870 put_files_struct(current->files);
871 current->files = files;
876 EXPORT_SYMBOL(flush_old_exec);
879 * Fill the binprm structure from the inode.
880 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
882 int prepare_binprm(struct linux_binprm *bprm)
885 struct inode * inode = bprm->file->f_dentry->d_inode;
888 mode = inode->i_mode;
890 * Check execute perms again - if the caller has CAP_DAC_OVERRIDE,
891 * vfs_permission lets a non-executable through
893 if (!(mode & 0111)) /* with at least _one_ execute bit set */
895 if (bprm->file->f_op == NULL)
898 bprm->e_uid = current->euid;
899 bprm->e_gid = current->egid;
901 if(!(bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)) {
903 if (mode & S_ISUID) {
904 current->personality &= ~PER_CLEAR_ON_SETID;
905 bprm->e_uid = inode->i_uid;
910 * If setgid is set but no group execute bit then this
911 * is a candidate for mandatory locking, not a setgid
914 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
915 current->personality &= ~PER_CLEAR_ON_SETID;
916 bprm->e_gid = inode->i_gid;
920 /* fill in binprm security blob */
921 retval = security_bprm_set(bprm);
925 memset(bprm->buf,0,BINPRM_BUF_SIZE);
926 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
929 EXPORT_SYMBOL(prepare_binprm);
931 static inline int unsafe_exec(struct task_struct *p)
934 if (p->ptrace & PT_PTRACED) {
935 if (p->ptrace & PT_PTRACE_CAP)
936 unsafe |= LSM_UNSAFE_PTRACE_CAP;
938 unsafe |= LSM_UNSAFE_PTRACE;
940 if (atomic_read(&p->fs->count) > 1 ||
941 atomic_read(&p->files->count) > 1 ||
942 atomic_read(&p->sighand->count) > 1)
943 unsafe |= LSM_UNSAFE_SHARE;
948 void compute_creds(struct linux_binprm *bprm)
952 unsafe = unsafe_exec(current);
953 security_bprm_apply_creds(bprm, unsafe);
954 task_unlock(current);
957 EXPORT_SYMBOL(compute_creds);
959 void remove_arg_zero(struct linux_binprm *bprm)
962 unsigned long offset;
966 offset = bprm->p % PAGE_SIZE;
969 while (bprm->p++, *(kaddr+offset++)) {
970 if (offset != PAGE_SIZE)
973 kunmap_atomic(kaddr, KM_USER0);
975 page = bprm->page[bprm->p/PAGE_SIZE];
976 kaddr = kmap_atomic(page, KM_USER0);
978 kunmap_atomic(kaddr, KM_USER0);
983 EXPORT_SYMBOL(remove_arg_zero);
986 * cycle the list of binary formats handler, until one recognizes the image
988 int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
991 struct linux_binfmt *fmt;
993 /* handle /sbin/loader.. */
995 struct exec * eh = (struct exec *) bprm->buf;
997 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
998 (eh->fh.f_flags & 0x3000) == 0x3000)
1001 unsigned long loader;
1003 allow_write_access(bprm->file);
1007 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1009 file = open_exec("/sbin/loader");
1010 retval = PTR_ERR(file);
1014 /* Remember if the application is TASO. */
1015 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1018 bprm->loader = loader;
1019 retval = prepare_binprm(bprm);
1022 /* should call search_binary_handler recursively here,
1023 but it does not matter */
1027 retval = security_bprm_check(bprm);
1031 /* kernel module loader fixup */
1032 /* so we don't try to load run modprobe in kernel space. */
1034 for (try=0; try<2; try++) {
1035 read_lock(&binfmt_lock);
1036 for (fmt = formats ; fmt ; fmt = fmt->next) {
1037 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1040 if (!try_module_get(fmt->module))
1042 read_unlock(&binfmt_lock);
1043 retval = fn(bprm, regs);
1046 allow_write_access(bprm->file);
1050 current->did_exec = 1;
1051 ckrm_cb_exec(bprm->filename);
1054 read_lock(&binfmt_lock);
1056 if (retval != -ENOEXEC || bprm->mm == NULL)
1059 read_unlock(&binfmt_lock);
1063 read_unlock(&binfmt_lock);
1064 if (retval != -ENOEXEC || bprm->mm == NULL) {
1068 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1069 if (printable(bprm->buf[0]) &&
1070 printable(bprm->buf[1]) &&
1071 printable(bprm->buf[2]) &&
1072 printable(bprm->buf[3]))
1073 break; /* -ENOEXEC */
1074 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1081 EXPORT_SYMBOL(search_binary_handler);
1084 * sys_execve() executes a new program.
1086 int do_execve(char * filename,
1087 char __user *__user *argv,
1088 char __user *__user *envp,
1089 struct pt_regs * regs)
1091 struct linux_binprm bprm;
1096 file = open_exec(filename);
1098 retval = PTR_ERR(file);
1102 sched_balance_exec();
1104 bprm.p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1105 memset(bprm.page, 0, MAX_ARG_PAGES*sizeof(bprm.page[0]));
1108 bprm.filename = filename;
1109 bprm.interp = filename;
1110 bprm.interp_flags = 0;
1111 bprm.interp_data = 0;
1115 bprm.security = NULL;
1116 bprm.mm = mm_alloc();
1121 retval = init_new_context(current, bprm.mm);
1125 bprm.argc = count(argv, bprm.p / sizeof(void *));
1126 if ((retval = bprm.argc) < 0)
1129 bprm.envc = count(envp, bprm.p / sizeof(void *));
1130 if ((retval = bprm.envc) < 0)
1133 retval = security_bprm_alloc(&bprm);
1137 retval = prepare_binprm(&bprm);
1141 retval = copy_strings_kernel(1, &bprm.filename, &bprm);
1146 retval = copy_strings(bprm.envc, envp, &bprm);
1150 retval = copy_strings(bprm.argc, argv, &bprm);
1154 retval = search_binary_handler(&bprm,regs);
1156 free_arg_pages(&bprm);
1158 /* execve success */
1159 security_bprm_free(&bprm);
1164 /* Something went wrong, return the inode and free the argument pages*/
1165 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1166 struct page * page = bprm.page[i];
1172 security_bprm_free(&bprm);
1180 allow_write_access(bprm.file);
1186 EXPORT_SYMBOL(do_execve);
1188 int set_binfmt(struct linux_binfmt *new)
1190 struct linux_binfmt *old = current->binfmt;
1193 if (!try_module_get(new->module))
1196 current->binfmt = new;
1198 module_put(old->module);
1202 EXPORT_SYMBOL(set_binfmt);
1204 #define CORENAME_MAX_SIZE 64
1206 /* format_corename will inspect the pattern parameter, and output a
1207 * name into corename, which must have space for at least
1208 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1210 void format_corename(char *corename, const char *pattern, long signr)
1212 const char *pat_ptr = pattern;
1213 char *out_ptr = corename;
1214 char *const out_end = corename + CORENAME_MAX_SIZE;
1216 int pid_in_pattern = 0;
1218 /* Repeat as long as we have more pattern to process and more output
1221 if (*pat_ptr != '%') {
1222 if (out_ptr == out_end)
1224 *out_ptr++ = *pat_ptr++;
1226 switch (*++pat_ptr) {
1229 /* Double percent, output one percent */
1231 if (out_ptr == out_end)
1238 rc = snprintf(out_ptr, out_end - out_ptr,
1239 "%d", current->tgid);
1240 if (rc > out_end - out_ptr)
1246 rc = snprintf(out_ptr, out_end - out_ptr,
1247 "%d", current->uid);
1248 if (rc > out_end - out_ptr)
1254 rc = snprintf(out_ptr, out_end - out_ptr,
1255 "%d", current->gid);
1256 if (rc > out_end - out_ptr)
1260 /* signal that caused the coredump */
1262 rc = snprintf(out_ptr, out_end - out_ptr,
1264 if (rc > out_end - out_ptr)
1268 /* UNIX time of coredump */
1271 do_gettimeofday(&tv);
1272 rc = snprintf(out_ptr, out_end - out_ptr,
1274 if (rc > out_end - out_ptr)
1281 down_read(&uts_sem);
1282 rc = snprintf(out_ptr, out_end - out_ptr,
1283 "%s", system_utsname.nodename);
1285 if (rc > out_end - out_ptr)
1291 rc = snprintf(out_ptr, out_end - out_ptr,
1292 "%s", current->comm);
1293 if (rc > out_end - out_ptr)
1303 /* Backward compatibility with core_uses_pid:
1305 * If core_pattern does not include a %p (as is the default)
1306 * and core_uses_pid is set, then .%pid will be appended to
1309 && (core_uses_pid || atomic_read(¤t->mm->mm_users) != 1)) {
1310 rc = snprintf(out_ptr, out_end - out_ptr,
1311 ".%d", current->tgid);
1312 if (rc > out_end - out_ptr)
1320 static void zap_threads (struct mm_struct *mm)
1322 struct task_struct *g, *p;
1323 struct task_struct *tsk = current;
1324 struct completion *vfork_done = tsk->vfork_done;
1327 * Make sure nobody is waiting for us to release the VM,
1328 * otherwise we can deadlock when we wait on each other
1331 tsk->vfork_done = NULL;
1332 complete(vfork_done);
1335 read_lock(&tasklist_lock);
1337 if (mm == p->mm && p != tsk) {
1338 force_sig_specific(SIGKILL, p);
1341 while_each_thread(g,p);
1343 read_unlock(&tasklist_lock);
1346 static void coredump_wait(struct mm_struct *mm)
1348 DECLARE_COMPLETION(startup_done);
1350 mm->core_waiters++; /* let other threads block */
1351 mm->core_startup_done = &startup_done;
1353 /* give other threads a chance to run: */
1357 if (--mm->core_waiters) {
1358 up_write(&mm->mmap_sem);
1359 wait_for_completion(&startup_done);
1361 up_write(&mm->mmap_sem);
1362 BUG_ON(mm->core_waiters);
1365 int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1367 char corename[CORENAME_MAX_SIZE + 1];
1368 struct mm_struct *mm = current->mm;
1369 struct linux_binfmt * binfmt;
1370 struct inode * inode;
1375 binfmt = current->binfmt;
1376 if (!binfmt || !binfmt->core_dump)
1378 down_write(&mm->mmap_sem);
1379 if (!mm->dumpable) {
1380 up_write(&mm->mmap_sem);
1384 init_completion(&mm->core_done);
1385 current->signal->group_exit = 1;
1386 current->signal->group_exit_code = exit_code;
1389 if (current->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1392 format_corename(corename, core_pattern, signr);
1393 file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE, 0600);
1396 inode = file->f_dentry->d_inode;
1397 if (inode->i_nlink > 1)
1398 goto close_fail; /* multiple links - don't dump */
1399 if (d_unhashed(file->f_dentry))
1402 if (!S_ISREG(inode->i_mode))
1406 if (!file->f_op->write)
1408 if (do_truncate(file->f_dentry, 0) != 0)
1411 retval = binfmt->core_dump(signr, regs, file);
1413 current->signal->group_exit_code |= 0x80;
1415 filp_close(file, NULL);
1417 complete_all(&mm->core_done);