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/fdtable.h>
29 #include <linux/stat.h>
30 #include <linux/fcntl.h>
31 #include <linux/smp_lock.h>
32 #include <linux/swap.h>
33 #include <linux/string.h>
34 #include <linux/init.h>
35 #include <linux/pagemap.h>
36 #include <linux/perf_event.h>
37 #include <linux/highmem.h>
38 #include <linux/spinlock.h>
39 #include <linux/key.h>
40 #include <linux/personality.h>
41 #include <linux/binfmts.h>
42 #include <linux/utsname.h>
43 #include <linux/pid_namespace.h>
44 #include <linux/module.h>
45 #include <linux/namei.h>
46 #include <linux/proc_fs.h>
47 #include <linux/mount.h>
48 #include <linux/security.h>
49 #include <linux/syscalls.h>
50 #include <linux/tsacct_kern.h>
51 #include <linux/cn_proc.h>
52 #include <linux/audit.h>
53 #include <linux/tracehook.h>
54 #include <linux/kmod.h>
55 #include <linux/fsnotify.h>
56 #include <linux/fs_struct.h>
57 #include <linux/pipe_fs_i.h>
59 #include <asm/uaccess.h>
60 #include <asm/mmu_context.h>
65 char core_pattern
[CORENAME_MAX_SIZE
] = "core";
66 unsigned int core_pipe_limit
;
67 int suid_dumpable
= 0;
69 /* The maximal length of core_pattern is also specified in sysctl.c */
71 static LIST_HEAD(formats
);
72 static DEFINE_RWLOCK(binfmt_lock
);
74 int __register_binfmt(struct linux_binfmt
* fmt
, int insert
)
78 write_lock(&binfmt_lock
);
79 insert
? list_add(&fmt
->lh
, &formats
) :
80 list_add_tail(&fmt
->lh
, &formats
);
81 write_unlock(&binfmt_lock
);
85 EXPORT_SYMBOL(__register_binfmt
);
87 void unregister_binfmt(struct linux_binfmt
* fmt
)
89 write_lock(&binfmt_lock
);
91 write_unlock(&binfmt_lock
);
94 EXPORT_SYMBOL(unregister_binfmt
);
96 static inline void put_binfmt(struct linux_binfmt
* fmt
)
98 module_put(fmt
->module
);
102 * Note that a shared library must be both readable and executable due to
105 * Also note that we take the address to load from from the file itself.
107 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
110 char *tmp
= getname(library
);
111 int error
= PTR_ERR(tmp
);
116 file
= do_filp_open(AT_FDCWD
, tmp
,
117 O_LARGEFILE
| O_RDONLY
| FMODE_EXEC
, 0,
118 MAY_READ
| MAY_EXEC
| MAY_OPEN
);
120 error
= PTR_ERR(file
);
125 if (!S_ISREG(file
->f_path
.dentry
->d_inode
->i_mode
))
129 if (file
->f_path
.mnt
->mnt_flags
& MNT_NOEXEC
)
132 fsnotify_open(file
->f_path
.dentry
);
136 struct linux_binfmt
* fmt
;
138 read_lock(&binfmt_lock
);
139 list_for_each_entry(fmt
, &formats
, lh
) {
140 if (!fmt
->load_shlib
)
142 if (!try_module_get(fmt
->module
))
144 read_unlock(&binfmt_lock
);
145 error
= fmt
->load_shlib(file
);
146 read_lock(&binfmt_lock
);
148 if (error
!= -ENOEXEC
)
151 read_unlock(&binfmt_lock
);
161 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
167 #ifdef CONFIG_STACK_GROWSUP
169 ret
= expand_stack_downwards(bprm
->vma
, pos
);
174 ret
= get_user_pages(current
, bprm
->mm
, pos
,
175 1, write
, 1, &page
, NULL
);
180 unsigned long size
= bprm
->vma
->vm_end
- bprm
->vma
->vm_start
;
184 * We've historically supported up to 32 pages (ARG_MAX)
185 * of argument strings even with small stacks
191 * Limit to 1/4-th the stack size for the argv+env strings.
193 * - the remaining binfmt code will not run out of stack space,
194 * - the program will have a reasonable amount of stack left
197 rlim
= current
->signal
->rlim
;
198 if (size
> ACCESS_ONCE(rlim
[RLIMIT_STACK
].rlim_cur
) / 4) {
207 static void put_arg_page(struct page
*page
)
212 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
216 static void free_arg_pages(struct linux_binprm
*bprm
)
220 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
223 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
226 static int __bprm_mm_init(struct linux_binprm
*bprm
)
229 struct vm_area_struct
*vma
= NULL
;
230 struct mm_struct
*mm
= bprm
->mm
;
232 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
236 down_write(&mm
->mmap_sem
);
240 * Place the stack at the largest stack address the architecture
241 * supports. Later, we'll move this to an appropriate place. We don't
242 * use STACK_TOP because that can depend on attributes which aren't
245 BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
246 vma
->vm_end
= STACK_TOP_MAX
;
247 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
248 vma
->vm_flags
= VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
249 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
250 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
251 err
= insert_vm_struct(mm
, vma
);
255 mm
->stack_vm
= mm
->total_vm
= 1;
256 up_write(&mm
->mmap_sem
);
257 bprm
->p
= vma
->vm_end
- sizeof(void *);
260 up_write(&mm
->mmap_sem
);
262 kmem_cache_free(vm_area_cachep
, vma
);
266 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
268 return len
<= MAX_ARG_STRLEN
;
273 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
278 page
= bprm
->page
[pos
/ PAGE_SIZE
];
279 if (!page
&& write
) {
280 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
283 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
289 static void put_arg_page(struct page
*page
)
293 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
296 __free_page(bprm
->page
[i
]);
297 bprm
->page
[i
] = NULL
;
301 static void free_arg_pages(struct linux_binprm
*bprm
)
305 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
306 free_arg_page(bprm
, i
);
309 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
314 static int __bprm_mm_init(struct linux_binprm
*bprm
)
316 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
320 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
322 return len
<= bprm
->p
;
325 #endif /* CONFIG_MMU */
328 * Create a new mm_struct and populate it with a temporary stack
329 * vm_area_struct. We don't have enough context at this point to set the stack
330 * flags, permissions, and offset, so we use temporary values. We'll update
331 * them later in setup_arg_pages().
333 int bprm_mm_init(struct linux_binprm
*bprm
)
336 struct mm_struct
*mm
= NULL
;
338 bprm
->mm
= mm
= mm_alloc();
343 err
= init_new_context(current
, mm
);
347 err
= __bprm_mm_init(bprm
);
363 * count() counts the number of strings in array ARGV.
365 static int count(char __user
* __user
* argv
, int max
)
373 if (get_user(p
, argv
))
381 if (fatal_signal_pending(current
))
382 return -ERESTARTNOHAND
;
390 * 'copy_strings()' copies argument/environment strings from the old
391 * processes's memory to the new process's stack. The call to get_user_pages()
392 * ensures the destination page is created and not swapped out.
394 static int copy_strings(int argc
, char __user
* __user
* argv
,
395 struct linux_binprm
*bprm
)
397 struct page
*kmapped_page
= NULL
;
399 unsigned long kpos
= 0;
407 if (get_user(str
, argv
+argc
) ||
408 !(len
= strnlen_user(str
, MAX_ARG_STRLEN
))) {
413 if (!valid_arg_len(bprm
, len
)) {
418 /* We're going to work our way backwords. */
424 int offset
, bytes_to_copy
;
426 if (fatal_signal_pending(current
)) {
427 ret
= -ERESTARTNOHAND
;
432 offset
= pos
% PAGE_SIZE
;
436 bytes_to_copy
= offset
;
437 if (bytes_to_copy
> len
)
440 offset
-= bytes_to_copy
;
441 pos
-= bytes_to_copy
;
442 str
-= bytes_to_copy
;
443 len
-= bytes_to_copy
;
445 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
448 page
= get_arg_page(bprm
, pos
, 1);
455 flush_kernel_dcache_page(kmapped_page
);
456 kunmap(kmapped_page
);
457 put_arg_page(kmapped_page
);
460 kaddr
= kmap(kmapped_page
);
461 kpos
= pos
& PAGE_MASK
;
462 flush_arg_page(bprm
, kpos
, kmapped_page
);
464 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
473 flush_kernel_dcache_page(kmapped_page
);
474 kunmap(kmapped_page
);
475 put_arg_page(kmapped_page
);
481 * Like copy_strings, but get argv and its values from kernel memory.
483 int copy_strings_kernel(int argc
,char ** argv
, struct linux_binprm
*bprm
)
486 mm_segment_t oldfs
= get_fs();
488 r
= copy_strings(argc
, (char __user
* __user
*)argv
, bprm
);
492 EXPORT_SYMBOL(copy_strings_kernel
);
497 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
498 * the binfmt code determines where the new stack should reside, we shift it to
499 * its final location. The process proceeds as follows:
501 * 1) Use shift to calculate the new vma endpoints.
502 * 2) Extend vma to cover both the old and new ranges. This ensures the
503 * arguments passed to subsequent functions are consistent.
504 * 3) Move vma's page tables to the new range.
505 * 4) Free up any cleared pgd range.
506 * 5) Shrink the vma to cover only the new range.
508 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
510 struct mm_struct
*mm
= vma
->vm_mm
;
511 unsigned long old_start
= vma
->vm_start
;
512 unsigned long old_end
= vma
->vm_end
;
513 unsigned long length
= old_end
- old_start
;
514 unsigned long new_start
= old_start
- shift
;
515 unsigned long new_end
= old_end
- shift
;
516 struct mmu_gather
*tlb
;
518 BUG_ON(new_start
> new_end
);
521 * ensure there are no vmas between where we want to go
524 if (vma
!= find_vma(mm
, new_start
))
528 * cover the whole range: [new_start, old_end)
530 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
534 * move the page tables downwards, on failure we rely on
535 * process cleanup to remove whatever mess we made.
537 if (length
!= move_page_tables(vma
, old_start
,
538 vma
, new_start
, length
))
542 tlb
= tlb_gather_mmu(mm
, 0);
543 if (new_end
> old_start
) {
545 * when the old and new regions overlap clear from new_end.
547 free_pgd_range(tlb
, new_end
, old_end
, new_end
,
548 vma
->vm_next
? vma
->vm_next
->vm_start
: 0);
551 * otherwise, clean from old_start; this is done to not touch
552 * the address space in [new_end, old_start) some architectures
553 * have constraints on va-space that make this illegal (IA64) -
554 * for the others its just a little faster.
556 free_pgd_range(tlb
, old_start
, old_end
, new_end
,
557 vma
->vm_next
? vma
->vm_next
->vm_start
: 0);
559 tlb_finish_mmu(tlb
, new_end
, old_end
);
562 * Shrink the vma to just the new range. Always succeeds.
564 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
570 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
571 * the stack is optionally relocated, and some extra space is added.
573 int setup_arg_pages(struct linux_binprm
*bprm
,
574 unsigned long stack_top
,
575 int executable_stack
)
578 unsigned long stack_shift
;
579 struct mm_struct
*mm
= current
->mm
;
580 struct vm_area_struct
*vma
= bprm
->vma
;
581 struct vm_area_struct
*prev
= NULL
;
582 unsigned long vm_flags
;
583 unsigned long stack_base
;
584 unsigned long stack_size
;
585 unsigned long stack_expand
;
586 unsigned long rlim_stack
;
588 #ifdef CONFIG_STACK_GROWSUP
589 /* Limit stack size to 1GB */
590 stack_base
= rlimit_max(RLIMIT_STACK
);
591 if (stack_base
> (1 << 30))
592 stack_base
= 1 << 30;
594 /* Make sure we didn't let the argument array grow too large. */
595 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
598 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
600 stack_shift
= vma
->vm_start
- stack_base
;
601 mm
->arg_start
= bprm
->p
- stack_shift
;
602 bprm
->p
= vma
->vm_end
- stack_shift
;
604 stack_top
= arch_align_stack(stack_top
);
605 stack_top
= PAGE_ALIGN(stack_top
);
607 if (unlikely(stack_top
< mmap_min_addr
) ||
608 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
611 stack_shift
= vma
->vm_end
- stack_top
;
613 bprm
->p
-= stack_shift
;
614 mm
->arg_start
= bprm
->p
;
618 bprm
->loader
-= stack_shift
;
619 bprm
->exec
-= stack_shift
;
621 down_write(&mm
->mmap_sem
);
622 vm_flags
= VM_STACK_FLAGS
;
625 * Adjust stack execute permissions; explicitly enable for
626 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
627 * (arch default) otherwise.
629 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
631 else if (executable_stack
== EXSTACK_DISABLE_X
)
632 vm_flags
&= ~VM_EXEC
;
633 vm_flags
|= mm
->def_flags
;
634 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
636 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
642 /* Move stack pages down in memory. */
644 ret
= shift_arg_pages(vma
, stack_shift
);
649 /* mprotect_fixup is overkill to remove the temporary stack flags */
650 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
652 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
653 stack_size
= vma
->vm_end
- vma
->vm_start
;
655 * Align this down to a page boundary as expand_stack
658 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
659 #ifdef CONFIG_STACK_GROWSUP
660 if (stack_size
+ stack_expand
> rlim_stack
)
661 stack_base
= vma
->vm_start
+ rlim_stack
;
663 stack_base
= vma
->vm_end
+ stack_expand
;
665 if (stack_size
+ stack_expand
> rlim_stack
)
666 stack_base
= vma
->vm_end
- rlim_stack
;
668 stack_base
= vma
->vm_start
- stack_expand
;
670 ret
= expand_stack(vma
, stack_base
);
675 up_write(&mm
->mmap_sem
);
678 EXPORT_SYMBOL(setup_arg_pages
);
680 #endif /* CONFIG_MMU */
682 struct file
*open_exec(const char *name
)
687 file
= do_filp_open(AT_FDCWD
, name
,
688 O_LARGEFILE
| O_RDONLY
| FMODE_EXEC
, 0,
689 MAY_EXEC
| MAY_OPEN
);
694 if (!S_ISREG(file
->f_path
.dentry
->d_inode
->i_mode
))
697 if (file
->f_path
.mnt
->mnt_flags
& MNT_NOEXEC
)
700 fsnotify_open(file
->f_path
.dentry
);
702 err
= deny_write_access(file
);
713 EXPORT_SYMBOL(open_exec
);
715 int kernel_read(struct file
*file
, loff_t offset
,
716 char *addr
, unsigned long count
)
724 /* The cast to a user pointer is valid due to the set_fs() */
725 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
730 EXPORT_SYMBOL(kernel_read
);
732 static int exec_mmap(struct mm_struct
*mm
)
734 struct task_struct
*tsk
;
735 struct mm_struct
* old_mm
, *active_mm
;
737 /* Notify parent that we're no longer interested in the old VM */
739 old_mm
= current
->mm
;
740 sync_mm_rss(tsk
, old_mm
);
741 mm_release(tsk
, old_mm
);
745 * Make sure that if there is a core dump in progress
746 * for the old mm, we get out and die instead of going
747 * through with the exec. We must hold mmap_sem around
748 * checking core_state and changing tsk->mm.
750 down_read(&old_mm
->mmap_sem
);
751 if (unlikely(old_mm
->core_state
)) {
752 up_read(&old_mm
->mmap_sem
);
757 active_mm
= tsk
->active_mm
;
760 activate_mm(active_mm
, mm
);
762 arch_pick_mmap_layout(mm
);
764 up_read(&old_mm
->mmap_sem
);
765 BUG_ON(active_mm
!= old_mm
);
766 mm_update_next_owner(old_mm
);
775 * This function makes sure the current process has its own signal table,
776 * so that flush_signal_handlers can later reset the handlers without
777 * disturbing other processes. (Other processes might share the signal
778 * table via the CLONE_SIGHAND option to clone().)
780 static int de_thread(struct task_struct
*tsk
)
782 struct signal_struct
*sig
= tsk
->signal
;
783 struct sighand_struct
*oldsighand
= tsk
->sighand
;
784 spinlock_t
*lock
= &oldsighand
->siglock
;
786 if (thread_group_empty(tsk
))
787 goto no_thread_group
;
790 * Kill all other threads in the thread group.
793 if (signal_group_exit(sig
)) {
795 * Another group action in progress, just
796 * return so that the signal is processed.
798 spin_unlock_irq(lock
);
802 sig
->group_exit_task
= tsk
;
803 sig
->notify_count
= zap_other_threads(tsk
);
804 if (!thread_group_leader(tsk
))
807 while (sig
->notify_count
) {
808 __set_current_state(TASK_UNINTERRUPTIBLE
);
809 spin_unlock_irq(lock
);
813 spin_unlock_irq(lock
);
816 * At this point all other threads have exited, all we have to
817 * do is to wait for the thread group leader to become inactive,
818 * and to assume its PID:
820 if (!thread_group_leader(tsk
)) {
821 struct task_struct
*leader
= tsk
->group_leader
;
823 sig
->notify_count
= -1; /* for exit_notify() */
825 write_lock_irq(&tasklist_lock
);
826 if (likely(leader
->exit_state
))
828 __set_current_state(TASK_UNINTERRUPTIBLE
);
829 write_unlock_irq(&tasklist_lock
);
834 * The only record we have of the real-time age of a
835 * process, regardless of execs it's done, is start_time.
836 * All the past CPU time is accumulated in signal_struct
837 * from sister threads now dead. But in this non-leader
838 * exec, nothing survives from the original leader thread,
839 * whose birth marks the true age of this process now.
840 * When we take on its identity by switching to its PID, we
841 * also take its birthdate (always earlier than our own).
843 tsk
->start_time
= leader
->start_time
;
845 BUG_ON(!same_thread_group(leader
, tsk
));
846 BUG_ON(has_group_leader_pid(tsk
));
848 * An exec() starts a new thread group with the
849 * TGID of the previous thread group. Rehash the
850 * two threads with a switched PID, and release
851 * the former thread group leader:
854 /* Become a process group leader with the old leader's pid.
855 * The old leader becomes a thread of the this thread group.
856 * Note: The old leader also uses this pid until release_task
857 * is called. Odd but simple and correct.
859 detach_pid(tsk
, PIDTYPE_PID
);
860 tsk
->pid
= leader
->pid
;
861 attach_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
862 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
863 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
865 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
866 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
868 tsk
->group_leader
= tsk
;
869 leader
->group_leader
= tsk
;
871 tsk
->exit_signal
= SIGCHLD
;
873 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
874 leader
->exit_state
= EXIT_DEAD
;
875 write_unlock_irq(&tasklist_lock
);
877 release_task(leader
);
880 sig
->group_exit_task
= NULL
;
881 sig
->notify_count
= 0;
885 setmax_mm_hiwater_rss(&sig
->maxrss
, current
->mm
);
888 flush_itimer_signals();
890 if (atomic_read(&oldsighand
->count
) != 1) {
891 struct sighand_struct
*newsighand
;
893 * This ->sighand is shared with the CLONE_SIGHAND
894 * but not CLONE_THREAD task, switch to the new one.
896 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
900 atomic_set(&newsighand
->count
, 1);
901 memcpy(newsighand
->action
, oldsighand
->action
,
902 sizeof(newsighand
->action
));
904 write_lock_irq(&tasklist_lock
);
905 spin_lock(&oldsighand
->siglock
);
906 rcu_assign_pointer(tsk
->sighand
, newsighand
);
907 spin_unlock(&oldsighand
->siglock
);
908 write_unlock_irq(&tasklist_lock
);
910 __cleanup_sighand(oldsighand
);
913 BUG_ON(!thread_group_leader(tsk
));
918 * These functions flushes out all traces of the currently running executable
919 * so that a new one can be started
921 static void flush_old_files(struct files_struct
* files
)
926 spin_lock(&files
->file_lock
);
928 unsigned long set
, i
;
932 fdt
= files_fdtable(files
);
933 if (i
>= fdt
->max_fds
)
935 set
= fdt
->close_on_exec
->fds_bits
[j
];
938 fdt
->close_on_exec
->fds_bits
[j
] = 0;
939 spin_unlock(&files
->file_lock
);
940 for ( ; set
; i
++,set
>>= 1) {
945 spin_lock(&files
->file_lock
);
948 spin_unlock(&files
->file_lock
);
951 char *get_task_comm(char *buf
, struct task_struct
*tsk
)
953 /* buf must be at least sizeof(tsk->comm) in size */
955 strncpy(buf
, tsk
->comm
, sizeof(tsk
->comm
));
960 void set_task_comm(struct task_struct
*tsk
, char *buf
)
965 * Threads may access current->comm without holding
966 * the task lock, so write the string carefully.
967 * Readers without a lock may see incomplete new
968 * names but are safe from non-terminating string reads.
970 memset(tsk
->comm
, 0, TASK_COMM_LEN
);
972 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
974 perf_event_comm(tsk
);
977 int flush_old_exec(struct linux_binprm
* bprm
)
982 * Make sure we have a private signal table and that
983 * we are unassociated from the previous thread group.
985 retval
= de_thread(current
);
989 set_mm_exe_file(bprm
->mm
, bprm
->file
);
992 * Release all of the old mmap stuff
994 retval
= exec_mmap(bprm
->mm
);
998 bprm
->mm
= NULL
; /* We're using it now */
1000 current
->flags
&= ~PF_RANDOMIZE
;
1002 current
->personality
&= ~bprm
->per_clear
;
1009 EXPORT_SYMBOL(flush_old_exec
);
1011 void setup_new_exec(struct linux_binprm
* bprm
)
1015 char tcomm
[sizeof(current
->comm
)];
1017 arch_pick_mmap_layout(current
->mm
);
1019 /* This is the point of no return */
1020 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1022 if (current_euid() == current_uid() && current_egid() == current_gid())
1023 set_dumpable(current
->mm
, 1);
1025 set_dumpable(current
->mm
, suid_dumpable
);
1027 name
= bprm
->filename
;
1029 /* Copies the binary name from after last slash */
1030 for (i
=0; (ch
= *(name
++)) != '\0';) {
1032 i
= 0; /* overwrite what we wrote */
1034 if (i
< (sizeof(tcomm
) - 1))
1038 set_task_comm(current
, tcomm
);
1040 /* Set the new mm task size. We have to do that late because it may
1041 * depend on TIF_32BIT which is only updated in flush_thread() on
1042 * some architectures like powerpc
1044 current
->mm
->task_size
= TASK_SIZE
;
1046 /* install the new credentials */
1047 if (bprm
->cred
->uid
!= current_euid() ||
1048 bprm
->cred
->gid
!= current_egid()) {
1049 current
->pdeath_signal
= 0;
1050 } else if (file_permission(bprm
->file
, MAY_READ
) ||
1051 bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
) {
1052 set_dumpable(current
->mm
, suid_dumpable
);
1056 * Flush performance counters when crossing a
1059 if (!get_dumpable(current
->mm
))
1060 perf_event_exit_task(current
);
1062 /* An exec changes our domain. We are no longer part of the thread
1065 current
->self_exec_id
++;
1067 flush_signal_handlers(current
, 0);
1068 flush_old_files(current
->files
);
1070 EXPORT_SYMBOL(setup_new_exec
);
1073 * Prepare credentials and lock ->cred_guard_mutex.
1074 * install_exec_creds() commits the new creds and drops the lock.
1075 * Or, if exec fails before, free_bprm() should release ->cred and
1078 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1080 if (mutex_lock_interruptible(¤t
->cred_guard_mutex
))
1081 return -ERESTARTNOINTR
;
1083 bprm
->cred
= prepare_exec_creds();
1084 if (likely(bprm
->cred
))
1087 mutex_unlock(¤t
->cred_guard_mutex
);
1091 void free_bprm(struct linux_binprm
*bprm
)
1093 free_arg_pages(bprm
);
1095 mutex_unlock(¤t
->cred_guard_mutex
);
1096 abort_creds(bprm
->cred
);
1102 * install the new credentials for this executable
1104 void install_exec_creds(struct linux_binprm
*bprm
)
1106 security_bprm_committing_creds(bprm
);
1108 commit_creds(bprm
->cred
);
1111 * cred_guard_mutex must be held at least to this point to prevent
1112 * ptrace_attach() from altering our determination of the task's
1113 * credentials; any time after this it may be unlocked.
1115 security_bprm_committed_creds(bprm
);
1116 mutex_unlock(¤t
->cred_guard_mutex
);
1118 EXPORT_SYMBOL(install_exec_creds
);
1121 * determine how safe it is to execute the proposed program
1122 * - the caller must hold current->cred_guard_mutex to protect against
1125 int check_unsafe_exec(struct linux_binprm
*bprm
)
1127 struct task_struct
*p
= current
, *t
;
1131 bprm
->unsafe
= tracehook_unsafe_exec(p
);
1134 write_lock(&p
->fs
->lock
);
1136 for (t
= next_thread(p
); t
!= p
; t
= next_thread(t
)) {
1142 if (p
->fs
->users
> n_fs
) {
1143 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1146 if (!p
->fs
->in_exec
) {
1151 write_unlock(&p
->fs
->lock
);
1157 * Fill the binprm structure from the inode.
1158 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1160 * This may be called multiple times for binary chains (scripts for example).
1162 int prepare_binprm(struct linux_binprm
*bprm
)
1165 struct inode
* inode
= bprm
->file
->f_path
.dentry
->d_inode
;
1168 mode
= inode
->i_mode
;
1169 if (bprm
->file
->f_op
== NULL
)
1172 /* clear any previous set[ug]id data from a previous binary */
1173 bprm
->cred
->euid
= current_euid();
1174 bprm
->cred
->egid
= current_egid();
1176 if (!(bprm
->file
->f_path
.mnt
->mnt_flags
& MNT_NOSUID
)) {
1178 if (mode
& S_ISUID
) {
1179 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1180 bprm
->cred
->euid
= inode
->i_uid
;
1185 * If setgid is set but no group execute bit then this
1186 * is a candidate for mandatory locking, not a setgid
1189 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1190 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1191 bprm
->cred
->egid
= inode
->i_gid
;
1195 /* fill in binprm security blob */
1196 retval
= security_bprm_set_creds(bprm
);
1199 bprm
->cred_prepared
= 1;
1201 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1202 return kernel_read(bprm
->file
, 0, bprm
->buf
, BINPRM_BUF_SIZE
);
1205 EXPORT_SYMBOL(prepare_binprm
);
1208 * Arguments are '\0' separated strings found at the location bprm->p
1209 * points to; chop off the first by relocating brpm->p to right after
1210 * the first '\0' encountered.
1212 int remove_arg_zero(struct linux_binprm
*bprm
)
1215 unsigned long offset
;
1223 offset
= bprm
->p
& ~PAGE_MASK
;
1224 page
= get_arg_page(bprm
, bprm
->p
, 0);
1229 kaddr
= kmap_atomic(page
, KM_USER0
);
1231 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1232 offset
++, bprm
->p
++)
1235 kunmap_atomic(kaddr
, KM_USER0
);
1238 if (offset
== PAGE_SIZE
)
1239 free_arg_page(bprm
, (bprm
->p
>> PAGE_SHIFT
) - 1);
1240 } while (offset
== PAGE_SIZE
);
1249 EXPORT_SYMBOL(remove_arg_zero
);
1252 * cycle the list of binary formats handler, until one recognizes the image
1254 int search_binary_handler(struct linux_binprm
*bprm
,struct pt_regs
*regs
)
1256 unsigned int depth
= bprm
->recursion_depth
;
1258 struct linux_binfmt
*fmt
;
1260 retval
= security_bprm_check(bprm
);
1264 /* kernel module loader fixup */
1265 /* so we don't try to load run modprobe in kernel space. */
1268 retval
= audit_bprm(bprm
);
1273 for (try=0; try<2; try++) {
1274 read_lock(&binfmt_lock
);
1275 list_for_each_entry(fmt
, &formats
, lh
) {
1276 int (*fn
)(struct linux_binprm
*, struct pt_regs
*) = fmt
->load_binary
;
1279 if (!try_module_get(fmt
->module
))
1281 read_unlock(&binfmt_lock
);
1282 retval
= fn(bprm
, regs
);
1284 * Restore the depth counter to its starting value
1285 * in this call, so we don't have to rely on every
1286 * load_binary function to restore it on return.
1288 bprm
->recursion_depth
= depth
;
1291 tracehook_report_exec(fmt
, bprm
, regs
);
1293 allow_write_access(bprm
->file
);
1297 current
->did_exec
= 1;
1298 proc_exec_connector(current
);
1301 read_lock(&binfmt_lock
);
1303 if (retval
!= -ENOEXEC
|| bprm
->mm
== NULL
)
1306 read_unlock(&binfmt_lock
);
1310 read_unlock(&binfmt_lock
);
1311 if (retval
!= -ENOEXEC
|| bprm
->mm
== NULL
) {
1313 #ifdef CONFIG_MODULES
1315 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1316 if (printable(bprm
->buf
[0]) &&
1317 printable(bprm
->buf
[1]) &&
1318 printable(bprm
->buf
[2]) &&
1319 printable(bprm
->buf
[3]))
1320 break; /* -ENOEXEC */
1321 request_module("binfmt-%04x", *(unsigned short *)(&bprm
->buf
[2]));
1328 EXPORT_SYMBOL(search_binary_handler
);
1331 * sys_execve() executes a new program.
1333 int do_execve(char * filename
,
1334 char __user
*__user
*argv
,
1335 char __user
*__user
*envp
,
1336 struct pt_regs
* regs
)
1338 struct linux_binprm
*bprm
;
1340 struct files_struct
*displaced
;
1344 retval
= unshare_files(&displaced
);
1349 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1353 retval
= prepare_bprm_creds(bprm
);
1357 retval
= check_unsafe_exec(bprm
);
1360 clear_in_exec
= retval
;
1361 current
->in_execve
= 1;
1363 file
= open_exec(filename
);
1364 retval
= PTR_ERR(file
);
1371 bprm
->filename
= filename
;
1372 bprm
->interp
= filename
;
1374 retval
= bprm_mm_init(bprm
);
1378 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1379 if ((retval
= bprm
->argc
) < 0)
1382 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1383 if ((retval
= bprm
->envc
) < 0)
1386 retval
= prepare_binprm(bprm
);
1390 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1394 bprm
->exec
= bprm
->p
;
1395 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1399 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1403 current
->flags
&= ~PF_KTHREAD
;
1404 retval
= search_binary_handler(bprm
,regs
);
1408 /* execve succeeded */
1409 current
->fs
->in_exec
= 0;
1410 current
->in_execve
= 0;
1411 acct_update_integrals(current
);
1414 put_files_struct(displaced
);
1423 allow_write_access(bprm
->file
);
1429 current
->fs
->in_exec
= 0;
1430 current
->in_execve
= 0;
1437 reset_files_struct(displaced
);
1442 void set_binfmt(struct linux_binfmt
*new)
1444 struct mm_struct
*mm
= current
->mm
;
1447 module_put(mm
->binfmt
->module
);
1451 __module_get(new->module
);
1454 EXPORT_SYMBOL(set_binfmt
);
1456 /* format_corename will inspect the pattern parameter, and output a
1457 * name into corename, which must have space for at least
1458 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1460 static int format_corename(char *corename
, long signr
)
1462 const struct cred
*cred
= current_cred();
1463 const char *pat_ptr
= core_pattern
;
1464 int ispipe
= (*pat_ptr
== '|');
1465 char *out_ptr
= corename
;
1466 char *const out_end
= corename
+ CORENAME_MAX_SIZE
;
1468 int pid_in_pattern
= 0;
1470 /* Repeat as long as we have more pattern to process and more output
1473 if (*pat_ptr
!= '%') {
1474 if (out_ptr
== out_end
)
1476 *out_ptr
++ = *pat_ptr
++;
1478 switch (*++pat_ptr
) {
1481 /* Double percent, output one percent */
1483 if (out_ptr
== out_end
)
1490 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1491 "%d", task_tgid_vnr(current
));
1492 if (rc
> out_end
- out_ptr
)
1498 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1500 if (rc
> out_end
- out_ptr
)
1506 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1508 if (rc
> out_end
- out_ptr
)
1512 /* signal that caused the coredump */
1514 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1516 if (rc
> out_end
- out_ptr
)
1520 /* UNIX time of coredump */
1523 do_gettimeofday(&tv
);
1524 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1526 if (rc
> out_end
- out_ptr
)
1533 down_read(&uts_sem
);
1534 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1535 "%s", utsname()->nodename
);
1537 if (rc
> out_end
- out_ptr
)
1543 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1544 "%s", current
->comm
);
1545 if (rc
> out_end
- out_ptr
)
1549 /* core limit size */
1551 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1552 "%lu", rlimit(RLIMIT_CORE
));
1553 if (rc
> out_end
- out_ptr
)
1563 /* Backward compatibility with core_uses_pid:
1565 * If core_pattern does not include a %p (as is the default)
1566 * and core_uses_pid is set, then .%pid will be appended to
1567 * the filename. Do not do this for piped commands. */
1568 if (!ispipe
&& !pid_in_pattern
&& core_uses_pid
) {
1569 rc
= snprintf(out_ptr
, out_end
- out_ptr
,
1570 ".%d", task_tgid_vnr(current
));
1571 if (rc
> out_end
- out_ptr
)
1580 static int zap_process(struct task_struct
*start
, int exit_code
)
1582 struct task_struct
*t
;
1585 start
->signal
->flags
= SIGNAL_GROUP_EXIT
;
1586 start
->signal
->group_exit_code
= exit_code
;
1587 start
->signal
->group_stop_count
= 0;
1591 if (t
!= current
&& t
->mm
) {
1592 sigaddset(&t
->pending
.signal
, SIGKILL
);
1593 signal_wake_up(t
, 1);
1596 } while_each_thread(start
, t
);
1601 static inline int zap_threads(struct task_struct
*tsk
, struct mm_struct
*mm
,
1602 struct core_state
*core_state
, int exit_code
)
1604 struct task_struct
*g
, *p
;
1605 unsigned long flags
;
1608 spin_lock_irq(&tsk
->sighand
->siglock
);
1609 if (!signal_group_exit(tsk
->signal
)) {
1610 mm
->core_state
= core_state
;
1611 nr
= zap_process(tsk
, exit_code
);
1613 spin_unlock_irq(&tsk
->sighand
->siglock
);
1614 if (unlikely(nr
< 0))
1617 if (atomic_read(&mm
->mm_users
) == nr
+ 1)
1620 * We should find and kill all tasks which use this mm, and we should
1621 * count them correctly into ->nr_threads. We don't take tasklist
1622 * lock, but this is safe wrt:
1625 * None of sub-threads can fork after zap_process(leader). All
1626 * processes which were created before this point should be
1627 * visible to zap_threads() because copy_process() adds the new
1628 * process to the tail of init_task.tasks list, and lock/unlock
1629 * of ->siglock provides a memory barrier.
1632 * The caller holds mm->mmap_sem. This means that the task which
1633 * uses this mm can't pass exit_mm(), so it can't exit or clear
1637 * It does list_replace_rcu(&leader->tasks, ¤t->tasks),
1638 * we must see either old or new leader, this does not matter.
1639 * However, it can change p->sighand, so lock_task_sighand(p)
1640 * must be used. Since p->mm != NULL and we hold ->mmap_sem
1643 * Note also that "g" can be the old leader with ->mm == NULL
1644 * and already unhashed and thus removed from ->thread_group.
1645 * This is OK, __unhash_process()->list_del_rcu() does not
1646 * clear the ->next pointer, we will find the new leader via
1650 for_each_process(g
) {
1651 if (g
== tsk
->group_leader
)
1653 if (g
->flags
& PF_KTHREAD
)
1658 if (unlikely(p
->mm
== mm
)) {
1659 lock_task_sighand(p
, &flags
);
1660 nr
+= zap_process(p
, exit_code
);
1661 unlock_task_sighand(p
, &flags
);
1665 } while_each_thread(g
, p
);
1669 atomic_set(&core_state
->nr_threads
, nr
);
1673 static int coredump_wait(int exit_code
, struct core_state
*core_state
)
1675 struct task_struct
*tsk
= current
;
1676 struct mm_struct
*mm
= tsk
->mm
;
1677 struct completion
*vfork_done
;
1678 int core_waiters
= -EBUSY
;
1680 init_completion(&core_state
->startup
);
1681 core_state
->dumper
.task
= tsk
;
1682 core_state
->dumper
.next
= NULL
;
1684 down_write(&mm
->mmap_sem
);
1685 if (!mm
->core_state
)
1686 core_waiters
= zap_threads(tsk
, mm
, core_state
, exit_code
);
1687 up_write(&mm
->mmap_sem
);
1689 if (unlikely(core_waiters
< 0))
1693 * Make sure nobody is waiting for us to release the VM,
1694 * otherwise we can deadlock when we wait on each other
1696 vfork_done
= tsk
->vfork_done
;
1698 tsk
->vfork_done
= NULL
;
1699 complete(vfork_done
);
1703 wait_for_completion(&core_state
->startup
);
1705 return core_waiters
;
1708 static void coredump_finish(struct mm_struct
*mm
)
1710 struct core_thread
*curr
, *next
;
1711 struct task_struct
*task
;
1713 next
= mm
->core_state
->dumper
.next
;
1714 while ((curr
= next
) != NULL
) {
1718 * see exit_mm(), curr->task must not see
1719 * ->task == NULL before we read ->next.
1723 wake_up_process(task
);
1726 mm
->core_state
= NULL
;
1730 * set_dumpable converts traditional three-value dumpable to two flags and
1731 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1732 * these bits are not changed atomically. So get_dumpable can observe the
1733 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1734 * return either old dumpable or new one by paying attention to the order of
1735 * modifying the bits.
1737 * dumpable | mm->flags (binary)
1738 * old new | initial interim final
1739 * ---------+-----------------------
1747 * (*) get_dumpable regards interim value of 10 as 11.
1749 void set_dumpable(struct mm_struct
*mm
, int value
)
1753 clear_bit(MMF_DUMPABLE
, &mm
->flags
);
1755 clear_bit(MMF_DUMP_SECURELY
, &mm
->flags
);
1758 set_bit(MMF_DUMPABLE
, &mm
->flags
);
1760 clear_bit(MMF_DUMP_SECURELY
, &mm
->flags
);
1763 set_bit(MMF_DUMP_SECURELY
, &mm
->flags
);
1765 set_bit(MMF_DUMPABLE
, &mm
->flags
);
1770 static int __get_dumpable(unsigned long mm_flags
)
1774 ret
= mm_flags
& MMF_DUMPABLE_MASK
;
1775 return (ret
>= 2) ? 2 : ret
;
1778 int get_dumpable(struct mm_struct
*mm
)
1780 return __get_dumpable(mm
->flags
);
1783 static void wait_for_dump_helpers(struct file
*file
)
1785 struct pipe_inode_info
*pipe
;
1787 pipe
= file
->f_path
.dentry
->d_inode
->i_pipe
;
1793 while ((pipe
->readers
> 1) && (!signal_pending(current
))) {
1794 wake_up_interruptible_sync(&pipe
->wait
);
1795 kill_fasync(&pipe
->fasync_readers
, SIGIO
, POLL_IN
);
1808 * helper function to customize the process used
1809 * to collect the core in userspace. Specifically
1810 * it sets up a pipe and installs it as fd 0 (stdin)
1811 * for the process. Returns 0 on success, or
1812 * PTR_ERR on failure.
1813 * Note that it also sets the core limit to 1. This
1814 * is a special value that we use to trap recursive
1817 static int umh_pipe_setup(struct subprocess_info
*info
)
1819 struct file
*rp
, *wp
;
1820 struct fdtable
*fdt
;
1821 struct coredump_params
*cp
= (struct coredump_params
*)info
->data
;
1822 struct files_struct
*cf
= current
->files
;
1824 wp
= create_write_pipe(0);
1828 rp
= create_read_pipe(wp
, 0);
1830 free_write_pipe(wp
);
1838 spin_lock(&cf
->file_lock
);
1839 fdt
= files_fdtable(cf
);
1840 FD_SET(0, fdt
->open_fds
);
1841 FD_CLR(0, fdt
->close_on_exec
);
1842 spin_unlock(&cf
->file_lock
);
1844 /* and disallow core files too */
1845 current
->signal
->rlim
[RLIMIT_CORE
] = (struct rlimit
){1, 1};
1850 void do_coredump(long signr
, int exit_code
, struct pt_regs
*regs
)
1852 struct core_state core_state
;
1853 char corename
[CORENAME_MAX_SIZE
+ 1];
1854 struct mm_struct
*mm
= current
->mm
;
1855 struct linux_binfmt
* binfmt
;
1856 const struct cred
*old_cred
;
1861 static atomic_t core_dump_count
= ATOMIC_INIT(0);
1862 struct coredump_params cprm
= {
1865 .limit
= rlimit(RLIMIT_CORE
),
1867 * We must use the same mm->flags while dumping core to avoid
1868 * inconsistency of bit flags, since this flag is not protected
1871 .mm_flags
= mm
->flags
,
1874 audit_core_dumps(signr
);
1876 binfmt
= mm
->binfmt
;
1877 if (!binfmt
|| !binfmt
->core_dump
)
1879 if (!__get_dumpable(cprm
.mm_flags
))
1882 cred
= prepare_creds();
1886 * We cannot trust fsuid as being the "true" uid of the
1887 * process nor do we know its entire history. We only know it
1888 * was tainted so we dump it as root in mode 2.
1890 if (__get_dumpable(cprm
.mm_flags
) == 2) {
1891 /* Setuid core dump mode */
1892 flag
= O_EXCL
; /* Stop rewrite attacks */
1893 cred
->fsuid
= 0; /* Dump root private */
1896 retval
= coredump_wait(exit_code
, &core_state
);
1900 old_cred
= override_creds(cred
);
1903 * Clear any false indication of pending signals that might
1904 * be seen by the filesystem code called to write the core file.
1906 clear_thread_flag(TIF_SIGPENDING
);
1909 * lock_kernel() because format_corename() is controlled by sysctl, which
1910 * uses lock_kernel()
1913 ispipe
= format_corename(corename
, signr
);
1920 if (cprm
.limit
== 1) {
1922 * Normally core limits are irrelevant to pipes, since
1923 * we're not writing to the file system, but we use
1924 * cprm.limit of 1 here as a speacial value. Any
1925 * non-1 limit gets set to RLIM_INFINITY below, but
1926 * a limit of 0 skips the dump. This is a consistent
1927 * way to catch recursive crashes. We can still crash
1928 * if the core_pattern binary sets RLIM_CORE = !1
1929 * but it runs as root, and can do lots of stupid things
1930 * Note that we use task_tgid_vnr here to grab the pid
1931 * of the process group leader. That way we get the
1932 * right pid if a thread in a multi-threaded
1933 * core_pattern process dies.
1936 "Process %d(%s) has RLIMIT_CORE set to 1\n",
1937 task_tgid_vnr(current
), current
->comm
);
1938 printk(KERN_WARNING
"Aborting core\n");
1941 cprm
.limit
= RLIM_INFINITY
;
1943 dump_count
= atomic_inc_return(&core_dump_count
);
1944 if (core_pipe_limit
&& (core_pipe_limit
< dump_count
)) {
1945 printk(KERN_WARNING
"Pid %d(%s) over core_pipe_limit\n",
1946 task_tgid_vnr(current
), current
->comm
);
1947 printk(KERN_WARNING
"Skipping core dump\n");
1948 goto fail_dropcount
;
1951 helper_argv
= argv_split(GFP_KERNEL
, corename
+1, NULL
);
1953 printk(KERN_WARNING
"%s failed to allocate memory\n",
1955 goto fail_dropcount
;
1958 retval
= call_usermodehelper_fns(helper_argv
[0], helper_argv
,
1959 NULL
, UMH_WAIT_EXEC
, umh_pipe_setup
,
1961 argv_free(helper_argv
);
1963 printk(KERN_INFO
"Core dump to %s pipe failed\n",
1968 struct inode
*inode
;
1970 if (cprm
.limit
< binfmt
->min_coredump
)
1973 cprm
.file
= filp_open(corename
,
1974 O_CREAT
| 2 | O_NOFOLLOW
| O_LARGEFILE
| flag
,
1976 if (IS_ERR(cprm
.file
))
1979 inode
= cprm
.file
->f_path
.dentry
->d_inode
;
1980 if (inode
->i_nlink
> 1)
1982 if (d_unhashed(cprm
.file
->f_path
.dentry
))
1985 * AK: actually i see no reason to not allow this for named
1986 * pipes etc, but keep the previous behaviour for now.
1988 if (!S_ISREG(inode
->i_mode
))
1991 * Dont allow local users get cute and trick others to coredump
1992 * into their pre-created files.
1994 if (inode
->i_uid
!= current_fsuid())
1996 if (!cprm
.file
->f_op
|| !cprm
.file
->f_op
->write
)
1998 if (do_truncate(cprm
.file
->f_path
.dentry
, 0, 0, cprm
.file
))
2002 retval
= binfmt
->core_dump(&cprm
);
2004 current
->signal
->group_exit_code
|= 0x80;
2006 if (ispipe
&& core_pipe_limit
)
2007 wait_for_dump_helpers(cprm
.file
);
2010 filp_close(cprm
.file
, NULL
);
2013 atomic_dec(&core_dump_count
);
2015 coredump_finish(mm
);
2016 revert_creds(old_cred
);