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/rmap-locking.h>
49 #include <asm/uaccess.h>
50 #include <asm/pgalloc.h>
51 #include <asm/mmu_context.h>
54 #include <linux/kmod.h>
58 char core_pattern
[65] = "core";
59 /* The maximal length of core_pattern is also specified in sysctl.c */
61 static struct linux_binfmt
*formats
;
62 static rwlock_t binfmt_lock
= RW_LOCK_UNLOCKED
;
64 int register_binfmt(struct linux_binfmt
* fmt
)
66 struct linux_binfmt
** tmp
= &formats
;
72 write_lock(&binfmt_lock
);
75 write_unlock(&binfmt_lock
);
82 write_unlock(&binfmt_lock
);
86 int unregister_binfmt(struct linux_binfmt
* fmt
)
88 struct linux_binfmt
** tmp
= &formats
;
90 write_lock(&binfmt_lock
);
94 write_unlock(&binfmt_lock
);
99 write_unlock(&binfmt_lock
);
103 static inline void put_binfmt(struct linux_binfmt
* fmt
)
105 module_put(fmt
->module
);
109 * Note that a shared library must be both readable and executable due to
112 * Also note that we take the address to load from from the file itself.
114 asmlinkage
long sys_uselib(const char __user
* library
)
120 error
= user_path_walk(library
, &nd
);
125 if (!S_ISREG(nd
.dentry
->d_inode
->i_mode
))
128 error
= permission(nd
.dentry
->d_inode
, MAY_READ
| MAY_EXEC
);
132 file
= dentry_open(nd
.dentry
, nd
.mnt
, O_RDONLY
);
133 error
= PTR_ERR(file
);
139 struct linux_binfmt
* fmt
;
141 read_lock(&binfmt_lock
);
142 for (fmt
= formats
; fmt
; fmt
= fmt
->next
) {
143 if (!fmt
->load_shlib
)
145 if (!try_module_get(fmt
->module
))
147 read_unlock(&binfmt_lock
);
148 error
= fmt
->load_shlib(file
);
149 read_lock(&binfmt_lock
);
151 if (error
!= -ENOEXEC
)
154 read_unlock(&binfmt_lock
);
165 * count() counts the number of strings in array ARGV.
167 static int count(char __user
* __user
* argv
, int max
)
175 if (get_user(p
, argv
))
188 * 'copy_strings()' copies argument/environment strings from user
189 * memory to free pages in kernel mem. These are in a format ready
190 * to be put directly into the top of new user memory.
192 int copy_strings(int argc
,char __user
* __user
* argv
, struct linux_binprm
*bprm
)
194 struct page
*kmapped_page
= NULL
;
203 if (get_user(str
, argv
+argc
) ||
204 !(len
= strnlen_user(str
, bprm
->p
))) {
215 /* XXX: add architecture specific overflow check here. */
220 int offset
, bytes_to_copy
;
223 offset
= pos
% PAGE_SIZE
;
225 page
= bprm
->page
[i
];
228 page
= alloc_page(GFP_HIGHUSER
);
229 bprm
->page
[i
] = page
;
237 if (page
!= kmapped_page
) {
239 kunmap(kmapped_page
);
241 kaddr
= kmap(kmapped_page
);
244 memset(kaddr
, 0, offset
);
245 bytes_to_copy
= PAGE_SIZE
- offset
;
246 if (bytes_to_copy
> len
) {
249 memset(kaddr
+offset
+len
, 0,
250 PAGE_SIZE
-offset
-len
);
252 err
= copy_from_user(kaddr
+offset
, str
, bytes_to_copy
);
258 pos
+= bytes_to_copy
;
259 str
+= bytes_to_copy
;
260 len
-= bytes_to_copy
;
266 kunmap(kmapped_page
);
271 * Like copy_strings, but get argv and its values from kernel memory.
273 int copy_strings_kernel(int argc
,char ** argv
, struct linux_binprm
*bprm
)
276 mm_segment_t oldfs
= get_fs();
278 r
= copy_strings(argc
, (char __user
* __user
*)argv
, bprm
);
285 * This routine is used to map in a page into an address space: needed by
286 * execve() for the initial stack and environment pages.
288 * tsk->mmap_sem is held for writing.
290 void put_dirty_page(struct task_struct
*tsk
, struct page
*page
,
291 unsigned long address
, pgprot_t prot
)
296 struct pte_chain
*pte_chain
;
298 if (page_count(page
) != 1)
299 printk(KERN_ERR
"mem_map disagrees with %p at %08lx\n",
302 pgd
= pgd_offset(tsk
->mm
, address
);
303 pte_chain
= pte_chain_alloc(GFP_KERNEL
);
306 spin_lock(&tsk
->mm
->page_table_lock
);
307 pmd
= pmd_alloc(tsk
->mm
, pgd
, address
);
310 pte
= pte_alloc_map(tsk
->mm
, pmd
, address
);
313 if (!pte_none(*pte
)) {
317 lru_cache_add_active(page
);
318 flush_dcache_page(page
);
319 set_pte(pte
, pte_mkdirty(pte_mkwrite(mk_pte(page
, prot
))));
320 pte_chain
= page_add_rmap(page
, pte
, pte_chain
);
323 spin_unlock(&tsk
->mm
->page_table_lock
);
325 /* no need for flush_tlb */
326 pte_chain_free(pte_chain
);
329 spin_unlock(&tsk
->mm
->page_table_lock
);
332 force_sig(SIGKILL
, tsk
);
333 pte_chain_free(pte_chain
);
337 int setup_arg_pages(struct linux_binprm
*bprm
)
339 unsigned long stack_base
;
340 struct vm_area_struct
*mpnt
;
341 struct mm_struct
*mm
= current
->mm
;
344 #ifdef CONFIG_STACK_GROWSUP
345 /* Move the argument and environment strings to the bottom of the
351 /* Start by shifting all the pages down */
353 for (j
= 0; j
< MAX_ARG_PAGES
; j
++) {
354 struct page
*page
= bprm
->page
[j
];
357 bprm
->page
[i
++] = page
;
360 /* Now move them within their pages */
361 offset
= bprm
->p
% PAGE_SIZE
;
362 to
= kmap(bprm
->page
[0]);
363 for (j
= 1; j
< i
; j
++) {
364 memmove(to
, to
+ offset
, PAGE_SIZE
- offset
);
365 from
= kmap(bprm
->page
[j
]);
366 memcpy(to
+ PAGE_SIZE
- offset
, from
, offset
);
367 kunmap(bprm
->page
[j
- 1]);
370 memmove(to
, to
+ offset
, PAGE_SIZE
- offset
);
371 kunmap(bprm
->page
[j
- 1]);
373 /* Adjust bprm->p to point to the end of the strings. */
374 bprm
->p
= PAGE_SIZE
* i
- offset
;
375 stack_base
= STACK_TOP
- current
->rlim
[RLIMIT_STACK
].rlim_max
;
376 mm
->arg_start
= stack_base
;
378 /* zero pages that were copied above */
379 while (i
< MAX_ARG_PAGES
)
380 bprm
->page
[i
++] = NULL
;
382 stack_base
= STACK_TOP
- MAX_ARG_PAGES
* PAGE_SIZE
;
383 mm
->arg_start
= bprm
->p
+ stack_base
;
386 bprm
->p
+= stack_base
;
388 bprm
->loader
+= stack_base
;
389 bprm
->exec
+= stack_base
;
391 mpnt
= kmem_cache_alloc(vm_area_cachep
, SLAB_KERNEL
);
395 if (!vm_enough_memory((STACK_TOP
- (PAGE_MASK
& (unsigned long) bprm
->p
))>>PAGE_SHIFT
)) {
396 kmem_cache_free(vm_area_cachep
, mpnt
);
400 down_write(&mm
->mmap_sem
);
403 #ifdef CONFIG_STACK_GROWSUP
404 mpnt
->vm_start
= stack_base
;
405 mpnt
->vm_end
= PAGE_MASK
&
406 (PAGE_SIZE
- 1 + (unsigned long) bprm
->p
);
408 mpnt
->vm_start
= PAGE_MASK
& (unsigned long) bprm
->p
;
409 mpnt
->vm_end
= STACK_TOP
;
411 mpnt
->vm_page_prot
= protection_map
[VM_STACK_FLAGS
& 0x7];
412 mpnt
->vm_flags
= VM_STACK_FLAGS
;
415 mpnt
->vm_file
= NULL
;
416 INIT_LIST_HEAD(&mpnt
->shared
);
417 mpnt
->vm_private_data
= (void *) 0;
418 insert_vm_struct(mm
, mpnt
);
419 mm
->total_vm
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
422 for (i
= 0 ; i
< MAX_ARG_PAGES
; i
++) {
423 struct page
*page
= bprm
->page
[i
];
425 bprm
->page
[i
] = NULL
;
426 put_dirty_page(current
, page
, stack_base
,
429 stack_base
+= PAGE_SIZE
;
431 up_write(&mm
->mmap_sem
);
436 #define free_arg_pages(bprm) do { } while (0)
440 static inline void free_arg_pages(struct linux_binprm
*bprm
)
444 for (i
= 0 ; i
< MAX_ARG_PAGES
; i
++) {
446 __free_page(bprm
->page
[i
]);
447 bprm
->page
[i
] = NULL
;
451 #endif /* CONFIG_MMU */
453 struct file
*open_exec(const char *name
)
456 int err
= path_lookup(name
, LOOKUP_FOLLOW
, &nd
);
457 struct file
*file
= ERR_PTR(err
);
460 struct inode
*inode
= nd
.dentry
->d_inode
;
461 file
= ERR_PTR(-EACCES
);
462 if (!(nd
.mnt
->mnt_flags
& MNT_NOEXEC
) &&
463 S_ISREG(inode
->i_mode
)) {
464 int err
= permission(inode
, MAY_EXEC
);
465 if (!err
&& !(inode
->i_mode
& 0111))
469 file
= dentry_open(nd
.dentry
, nd
.mnt
, O_RDONLY
);
471 err
= deny_write_access(file
);
486 int kernel_read(struct file
*file
, unsigned long offset
,
487 char *addr
, unsigned long count
)
495 /* The cast to a user pointer is valid due to the set_fs() */
496 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
501 static int exec_mmap(struct mm_struct
*mm
)
503 struct task_struct
*tsk
;
504 struct mm_struct
* old_mm
, *active_mm
;
506 /* Add it to the list of mm's */
507 spin_lock(&mmlist_lock
);
508 list_add(&mm
->mmlist
, &init_mm
.mmlist
);
510 spin_unlock(&mmlist_lock
);
512 /* Notify parent that we're no longer interested in the old VM */
514 old_mm
= current
->mm
;
515 mm_release(tsk
, old_mm
);
518 active_mm
= tsk
->active_mm
;
521 activate_mm(active_mm
, mm
);
524 if (active_mm
!= old_mm
) BUG();
533 * This function makes sure the current process has its own signal table,
534 * so that flush_signal_handlers can later reset the handlers without
535 * disturbing other processes. (Other processes might share the signal
536 * table via the CLONE_SIGHAND option to clone().)
538 static inline int de_thread(struct task_struct
*tsk
)
540 struct signal_struct
*newsig
, *oldsig
= tsk
->signal
;
541 struct sighand_struct
*newsighand
, *oldsighand
= tsk
->sighand
;
542 spinlock_t
*lock
= &oldsighand
->siglock
;
546 * If we don't share sighandlers, then we aren't sharing anything
547 * and we can just re-use it all.
549 if (atomic_read(&oldsighand
->count
) <= 1)
552 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
556 spin_lock_init(&newsighand
->siglock
);
557 atomic_set(&newsighand
->count
, 1);
558 memcpy(newsighand
->action
, oldsighand
->action
, sizeof(newsighand
->action
));
561 * See if we need to allocate a new signal structure
564 if (atomic_read(&oldsig
->count
) > 1) {
565 newsig
= kmem_cache_alloc(signal_cachep
, GFP_KERNEL
);
567 kmem_cache_free(sighand_cachep
, newsighand
);
570 atomic_set(&newsig
->count
, 1);
571 newsig
->group_exit
= 0;
572 newsig
->group_exit_code
= 0;
573 newsig
->group_exit_task
= NULL
;
574 newsig
->group_stop_count
= 0;
575 newsig
->curr_target
= NULL
;
576 init_sigpending(&newsig
->shared_pending
);
579 if (thread_group_empty(current
))
580 goto no_thread_group
;
583 * Kill all other threads in the thread group.
584 * We must hold tasklist_lock to call zap_other_threads.
586 read_lock(&tasklist_lock
);
588 if (oldsig
->group_exit
) {
590 * Another group action in progress, just
591 * return so that the signal is processed.
593 spin_unlock_irq(lock
);
594 read_unlock(&tasklist_lock
);
595 kmem_cache_free(sighand_cachep
, newsighand
);
597 kmem_cache_free(signal_cachep
, newsig
);
600 oldsig
->group_exit
= 1;
601 zap_other_threads(current
);
602 read_unlock(&tasklist_lock
);
605 * Account for the thread group leader hanging around:
608 if (current
->pid
== current
->tgid
)
610 while (atomic_read(&oldsig
->count
) > count
) {
611 oldsig
->group_exit_task
= current
;
612 oldsig
->notify_count
= count
;
613 __set_current_state(TASK_UNINTERRUPTIBLE
);
614 spin_unlock_irq(lock
);
618 spin_unlock_irq(lock
);
621 * At this point all other threads have exited, all we have to
622 * do is to wait for the thread group leader to become inactive,
623 * and to assume its PID:
625 if (current
->pid
!= current
->tgid
) {
626 struct task_struct
*leader
= current
->group_leader
, *parent
;
627 struct dentry
*proc_dentry1
, *proc_dentry2
;
628 unsigned long state
, ptrace
;
631 * Wait for the thread group leader to be a zombie.
632 * It should already be zombie at this point, most
635 while (leader
->state
!= TASK_ZOMBIE
)
638 spin_lock(&leader
->proc_lock
);
639 spin_lock(¤t
->proc_lock
);
640 proc_dentry1
= proc_pid_unhash(current
);
641 proc_dentry2
= proc_pid_unhash(leader
);
642 write_lock_irq(&tasklist_lock
);
644 if (leader
->tgid
!= current
->tgid
)
646 if (current
->pid
== current
->tgid
)
649 * An exec() starts a new thread group with the
650 * TGID of the previous thread group. Rehash the
651 * two threads with a switched PID, and release
652 * the former thread group leader:
654 ptrace
= leader
->ptrace
;
655 parent
= leader
->parent
;
657 ptrace_unlink(current
);
658 ptrace_unlink(leader
);
659 remove_parent(current
);
660 remove_parent(leader
);
662 switch_exec_pids(leader
, current
);
664 current
->parent
= current
->real_parent
= leader
->real_parent
;
665 leader
->parent
= leader
->real_parent
= child_reaper
;
666 current
->group_leader
= current
;
667 leader
->group_leader
= leader
;
669 add_parent(current
, current
->parent
);
670 add_parent(leader
, leader
->parent
);
672 current
->ptrace
= ptrace
;
673 __ptrace_link(current
, parent
);
676 list_del(¤t
->tasks
);
677 list_add_tail(¤t
->tasks
, &init_task
.tasks
);
678 current
->exit_signal
= SIGCHLD
;
679 state
= leader
->state
;
681 write_unlock_irq(&tasklist_lock
);
682 spin_unlock(&leader
->proc_lock
);
683 spin_unlock(¤t
->proc_lock
);
684 proc_pid_flush(proc_dentry1
);
685 proc_pid_flush(proc_dentry2
);
687 if (state
!= TASK_ZOMBIE
)
689 release_task(leader
);
694 write_lock_irq(&tasklist_lock
);
695 spin_lock(&oldsighand
->siglock
);
696 spin_lock(&newsighand
->siglock
);
698 if (current
== oldsig
->curr_target
)
699 oldsig
->curr_target
= next_thread(current
);
701 current
->signal
= newsig
;
702 current
->sighand
= newsighand
;
703 init_sigpending(¤t
->pending
);
706 spin_unlock(&newsighand
->siglock
);
707 spin_unlock(&oldsighand
->siglock
);
708 write_unlock_irq(&tasklist_lock
);
710 if (newsig
&& atomic_dec_and_test(&oldsig
->count
))
711 kmem_cache_free(signal_cachep
, oldsig
);
713 if (atomic_dec_and_test(&oldsighand
->count
))
714 kmem_cache_free(sighand_cachep
, oldsighand
);
716 if (!thread_group_empty(current
))
718 if (current
->tgid
!= current
->pid
)
724 * These functions flushes out all traces of the currently running executable
725 * so that a new one can be started
728 static inline void flush_old_files(struct files_struct
* files
)
732 spin_lock(&files
->file_lock
);
734 unsigned long set
, i
;
738 if (i
>= files
->max_fds
|| i
>= files
->max_fdset
)
740 set
= files
->close_on_exec
->fds_bits
[j
];
743 files
->close_on_exec
->fds_bits
[j
] = 0;
744 spin_unlock(&files
->file_lock
);
745 for ( ; set
; i
++,set
>>= 1) {
750 spin_lock(&files
->file_lock
);
753 spin_unlock(&files
->file_lock
);
756 int flush_old_exec(struct linux_binprm
* bprm
)
762 * Release all of the old mmap stuff
764 retval
= exec_mmap(bprm
->mm
);
768 * Make sure we have a private signal table and that
769 * we are unassociated from the previous thread group.
771 retval
= de_thread(current
);
775 /* This is the point of no return */
777 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
779 if (current
->euid
== current
->uid
&& current
->egid
== current
->gid
)
780 current
->mm
->dumpable
= 1;
781 name
= bprm
->filename
;
782 for (i
=0; (ch
= *(name
++)) != '\0';) {
787 current
->comm
[i
++] = ch
;
789 current
->comm
[i
] = '\0';
793 if (bprm
->e_uid
!= current
->euid
|| bprm
->e_gid
!= current
->egid
||
794 permission(bprm
->file
->f_dentry
->d_inode
,MAY_READ
))
795 current
->mm
->dumpable
= 0;
797 /* An exec changes our domain. We are no longer part of the thread
800 current
->self_exec_id
++;
802 flush_signal_handlers(current
, 0);
803 flush_old_files(current
->files
);
804 exit_itimers(current
);
813 * We mustn't allow tracing of suid binaries, unless
814 * the tracer has the capability to trace anything..
816 static inline int must_not_trace_exec(struct task_struct
* p
)
818 return (p
->ptrace
& PT_PTRACED
) && !(p
->ptrace
& PT_PTRACE_CAP
);
822 * Fill the binprm structure from the inode.
823 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
825 int prepare_binprm(struct linux_binprm
*bprm
)
828 struct inode
* inode
= bprm
->file
->f_dentry
->d_inode
;
831 mode
= inode
->i_mode
;
833 * Check execute perms again - if the caller has CAP_DAC_OVERRIDE,
834 * vfs_permission lets a non-executable through
836 if (!(mode
& 0111)) /* with at least _one_ execute bit set */
838 if (bprm
->file
->f_op
== NULL
)
841 bprm
->e_uid
= current
->euid
;
842 bprm
->e_gid
= current
->egid
;
844 if(!(bprm
->file
->f_vfsmnt
->mnt_flags
& MNT_NOSUID
)) {
847 bprm
->e_uid
= inode
->i_uid
;
851 * If setgid is set but no group execute bit then this
852 * is a candidate for mandatory locking, not a setgid
855 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
))
856 bprm
->e_gid
= inode
->i_gid
;
859 /* fill in binprm security blob */
860 retval
= security_bprm_set(bprm
);
864 memset(bprm
->buf
,0,BINPRM_BUF_SIZE
);
865 return kernel_read(bprm
->file
,0,bprm
->buf
,BINPRM_BUF_SIZE
);
869 * This function is used to produce the new IDs and capabilities
870 * from the old ones and the file's capabilities.
872 * The formula used for evolving capabilities is:
875 * (***) pP' = (fP & X) | (fI & pI)
876 * pE' = pP' & fE [NB. fE is 0 or ~0]
878 * I=Inheritable, P=Permitted, E=Effective // p=process, f=file
879 * ' indicates post-exec(), and X is the global 'cap_bset'.
883 void compute_creds(struct linux_binprm
*bprm
)
886 if (bprm
->e_uid
!= current
->uid
|| bprm
->e_gid
!= current
->gid
) {
887 current
->mm
->dumpable
= 0;
889 if (must_not_trace_exec(current
)
890 || atomic_read(¤t
->fs
->count
) > 1
891 || atomic_read(¤t
->files
->count
) > 1
892 || atomic_read(¤t
->sighand
->count
) > 1) {
893 if(!capable(CAP_SETUID
)) {
894 bprm
->e_uid
= current
->uid
;
895 bprm
->e_gid
= current
->gid
;
900 current
->suid
= current
->euid
= current
->fsuid
= bprm
->e_uid
;
901 current
->sgid
= current
->egid
= current
->fsgid
= bprm
->e_gid
;
903 task_unlock(current
);
905 security_bprm_compute_creds(bprm
);
908 void remove_arg_zero(struct linux_binprm
*bprm
)
911 unsigned long offset
;
915 offset
= bprm
->p
% PAGE_SIZE
;
918 while (bprm
->p
++, *(kaddr
+offset
++)) {
919 if (offset
!= PAGE_SIZE
)
924 page
= bprm
->page
[bprm
->p
/PAGE_SIZE
];
933 * cycle the list of binary formats handler, until one recognizes the image
935 int search_binary_handler(struct linux_binprm
*bprm
,struct pt_regs
*regs
)
938 struct linux_binfmt
*fmt
;
940 /* handle /sbin/loader.. */
942 struct exec
* eh
= (struct exec
*) bprm
->buf
;
944 if (!bprm
->loader
&& eh
->fh
.f_magic
== 0x183 &&
945 (eh
->fh
.f_flags
& 0x3000) == 0x3000)
948 unsigned long loader
;
950 allow_write_access(bprm
->file
);
954 loader
= PAGE_SIZE
*MAX_ARG_PAGES
-sizeof(void *);
956 file
= open_exec("/sbin/loader");
957 retval
= PTR_ERR(file
);
961 /* Remember if the application is TASO. */
962 bprm
->sh_bang
= eh
->ah
.entry
< 0x100000000;
965 bprm
->loader
= loader
;
966 retval
= prepare_binprm(bprm
);
969 /* should call search_binary_handler recursively here,
970 but it does not matter */
974 retval
= security_bprm_check(bprm
);
978 /* kernel module loader fixup */
979 /* so we don't try to load run modprobe in kernel space. */
981 for (try=0; try<2; try++) {
982 read_lock(&binfmt_lock
);
983 for (fmt
= formats
; fmt
; fmt
= fmt
->next
) {
984 int (*fn
)(struct linux_binprm
*, struct pt_regs
*) = fmt
->load_binary
;
987 if (!try_module_get(fmt
->module
))
989 read_unlock(&binfmt_lock
);
990 retval
= fn(bprm
, regs
);
993 allow_write_access(bprm
->file
);
997 current
->did_exec
= 1;
1000 read_lock(&binfmt_lock
);
1002 if (retval
!= -ENOEXEC
)
1005 read_unlock(&binfmt_lock
);
1009 read_unlock(&binfmt_lock
);
1010 if (retval
!= -ENOEXEC
) {
1014 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1015 if (printable(bprm
->buf
[0]) &&
1016 printable(bprm
->buf
[1]) &&
1017 printable(bprm
->buf
[2]) &&
1018 printable(bprm
->buf
[3]))
1019 break; /* -ENOEXEC */
1020 request_module("binfmt-%04x", *(unsigned short *)(&bprm
->buf
[2]));
1028 * sys_execve() executes a new program.
1030 int do_execve(char * filename
,
1031 char __user
*__user
*argv
,
1032 char __user
*__user
*envp
,
1033 struct pt_regs
* regs
)
1035 struct linux_binprm bprm
;
1040 sched_balance_exec();
1042 file
= open_exec(filename
);
1044 retval
= PTR_ERR(file
);
1048 bprm
.p
= PAGE_SIZE
*MAX_ARG_PAGES
-sizeof(void *);
1049 memset(bprm
.page
, 0, MAX_ARG_PAGES
*sizeof(bprm
.page
[0]));
1052 bprm
.filename
= filename
;
1056 bprm
.security
= NULL
;
1057 bprm
.mm
= mm_alloc();
1062 retval
= init_new_context(current
, bprm
.mm
);
1066 bprm
.argc
= count(argv
, bprm
.p
/ sizeof(void *));
1067 if ((retval
= bprm
.argc
) < 0)
1070 bprm
.envc
= count(envp
, bprm
.p
/ sizeof(void *));
1071 if ((retval
= bprm
.envc
) < 0)
1074 retval
= security_bprm_alloc(&bprm
);
1078 retval
= prepare_binprm(&bprm
);
1082 retval
= copy_strings_kernel(1, &bprm
.filename
, &bprm
);
1087 retval
= copy_strings(bprm
.envc
, envp
, &bprm
);
1091 retval
= copy_strings(bprm
.argc
, argv
, &bprm
);
1095 retval
= search_binary_handler(&bprm
,regs
);
1097 free_arg_pages(&bprm
);
1099 /* execve success */
1100 security_bprm_free(&bprm
);
1105 /* Something went wrong, return the inode and free the argument pages*/
1106 for (i
= 0 ; i
< MAX_ARG_PAGES
; i
++) {
1107 struct page
* page
= bprm
.page
[i
];
1113 security_bprm_free(&bprm
);
1120 allow_write_access(bprm
.file
);
1126 int set_binfmt(struct linux_binfmt
*new)
1128 struct linux_binfmt
*old
= current
->binfmt
;
1131 if (!try_module_get(new->module
))
1134 current
->binfmt
= new;
1136 module_put(old
->module
);
1140 #define CORENAME_MAX_SIZE 64
1142 /* format_corename will inspect the pattern parameter, and output a
1143 * name into corename, which must have space for at least
1144 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1146 void format_corename(char *corename
, const char *pattern
, long signr
)
1148 const char *pat_ptr
= pattern
;
1149 char *out_ptr
= corename
;
1150 char *const out_end
= corename
+ CORENAME_MAX_SIZE
;
1152 int pid_in_pattern
= 0;
1154 /* Repeat as long as we have more pattern to process and more output
1157 if (*pat_ptr
!= '%') {
1158 if (out_ptr
== out_end
)
1160 *out_ptr
++ = *pat_ptr
++;
1162 switch (*++pat_ptr
) {
1165 /* Double percent, output one percent */
1167 if (out_ptr
== out_end
)
1174 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1175 "%d", current
->tgid
);
1176 if (rc
> out_end
- out_ptr
)
1182 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1183 "%d", current
->uid
);
1184 if (rc
> out_end
- out_ptr
)
1190 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1191 "%d", current
->gid
);
1192 if (rc
> out_end
- out_ptr
)
1196 /* signal that caused the coredump */
1198 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1200 if (rc
> out_end
- out_ptr
)
1204 /* UNIX time of coredump */
1207 do_gettimeofday(&tv
);
1208 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1210 if (rc
> out_end
- out_ptr
)
1217 down_read(&uts_sem
);
1218 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1219 "%s", system_utsname
.nodename
);
1221 if (rc
> out_end
- out_ptr
)
1227 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1228 "%s", current
->comm
);
1229 if (rc
> out_end
- out_ptr
)
1239 /* Backward compatibility with core_uses_pid:
1241 * If core_pattern does not include a %p (as is the default)
1242 * and core_uses_pid is set, then .%pid will be appended to
1245 && (core_uses_pid
|| atomic_read(¤t
->mm
->mm_users
) != 1)) {
1246 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1247 ".%d", current
->tgid
);
1248 if (rc
> out_end
- out_ptr
)
1256 static void zap_threads (struct mm_struct
*mm
)
1258 struct task_struct
*g
, *p
;
1260 read_lock(&tasklist_lock
);
1262 if (mm
== p
->mm
&& p
!= current
) {
1263 force_sig_specific(SIGKILL
, p
);
1266 while_each_thread(g
,p
);
1268 read_unlock(&tasklist_lock
);
1271 static void coredump_wait(struct mm_struct
*mm
)
1273 DECLARE_COMPLETION(startup_done
);
1275 mm
->core_waiters
++; /* let other threads block */
1276 mm
->core_startup_done
= &startup_done
;
1278 /* give other threads a chance to run: */
1282 if (--mm
->core_waiters
) {
1283 up_write(&mm
->mmap_sem
);
1284 wait_for_completion(&startup_done
);
1286 up_write(&mm
->mmap_sem
);
1287 BUG_ON(mm
->core_waiters
);
1290 int do_coredump(long signr
, int exit_code
, struct pt_regs
* regs
)
1292 char corename
[CORENAME_MAX_SIZE
+ 1];
1293 struct mm_struct
*mm
= current
->mm
;
1294 struct linux_binfmt
* binfmt
;
1295 struct inode
* inode
;
1300 binfmt
= current
->binfmt
;
1301 if (!binfmt
|| !binfmt
->core_dump
)
1303 down_write(&mm
->mmap_sem
);
1304 if (!mm
->dumpable
) {
1305 up_write(&mm
->mmap_sem
);
1309 init_completion(&mm
->core_done
);
1310 current
->signal
->group_exit
= 1;
1311 current
->signal
->group_exit_code
= exit_code
;
1314 if (current
->rlim
[RLIMIT_CORE
].rlim_cur
< binfmt
->min_coredump
)
1317 format_corename(corename
, core_pattern
, signr
);
1318 file
= filp_open(corename
, O_CREAT
| 2 | O_NOFOLLOW
, 0600);
1321 inode
= file
->f_dentry
->d_inode
;
1322 if (inode
->i_nlink
> 1)
1323 goto close_fail
; /* multiple links - don't dump */
1324 if (d_unhashed(file
->f_dentry
))
1327 if (!S_ISREG(inode
->i_mode
))
1331 if (!file
->f_op
->write
)
1333 if (do_truncate(file
->f_dentry
, 0) != 0)
1336 retval
= binfmt
->core_dump(signr
, regs
, file
);
1338 current
->signal
->group_exit_code
|= 0x80;
1340 filp_close(file
, NULL
);
1342 complete_all(&mm
->core_done
);