4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/printk.h>
77 #include <linux/cgroup.h>
78 #include <linux/cpuset.h>
79 #include <linux/audit.h>
80 #include <linux/poll.h>
81 #include <linux/nsproxy.h>
82 #include <linux/oom.h>
83 #include <linux/elf.h>
84 #include <linux/pid_namespace.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
87 #include <linux/slab.h>
88 #include <linux/flex_array.h>
89 #include <linux/posix-timers.h>
90 #ifdef CONFIG_HARDWALL
91 #include <asm/hardwall.h>
93 #include <trace/events/oom.h>
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
111 const struct inode_operations
*iop
;
112 const struct file_operations
*fop
;
116 #define NOD(NAME, MODE, IOP, FOP, OP) { \
118 .len = sizeof(NAME) - 1, \
125 #define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127 #define LNK(NAME, get_link) \
128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
129 &proc_pid_link_inode_operations, NULL, \
130 { .proc_get_link = get_link } )
131 #define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133 #define ONE(NAME, MODE, show) \
134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
136 { .proc_show = show } )
139 * Count the number of hardlinks for the pid_entry table, excluding the .
142 static unsigned int pid_entry_count_dirs(const struct pid_entry
*entries
,
149 for (i
= 0; i
< n
; ++i
) {
150 if (S_ISDIR(entries
[i
].mode
))
157 static int get_task_root(struct task_struct
*task
, struct path
*root
)
159 int result
= -ENOENT
;
163 get_fs_root(task
->fs
, root
);
170 static int proc_cwd_link(struct dentry
*dentry
, struct path
*path
)
172 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
173 int result
= -ENOENT
;
178 get_fs_pwd(task
->fs
, path
);
182 put_task_struct(task
);
187 static int proc_root_link(struct dentry
*dentry
, struct path
*path
)
189 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
190 int result
= -ENOENT
;
193 result
= get_task_root(task
, path
);
194 put_task_struct(task
);
199 static ssize_t
proc_pid_cmdline_read(struct file
*file
, char __user
*buf
,
200 size_t _count
, loff_t
*pos
)
202 struct task_struct
*tsk
;
203 struct mm_struct
*mm
;
205 unsigned long count
= _count
;
206 unsigned long arg_start
, arg_end
, env_start
, env_end
;
207 unsigned long len1
, len2
, len
;
214 tsk
= get_proc_task(file_inode(file
));
217 mm
= get_task_mm(tsk
);
218 put_task_struct(tsk
);
221 /* Check if process spawned far enough to have cmdline. */
227 page
= (char *)__get_free_page(GFP_TEMPORARY
);
233 down_read(&mm
->mmap_sem
);
234 arg_start
= mm
->arg_start
;
235 arg_end
= mm
->arg_end
;
236 env_start
= mm
->env_start
;
237 env_end
= mm
->env_end
;
238 up_read(&mm
->mmap_sem
);
240 BUG_ON(arg_start
> arg_end
);
241 BUG_ON(env_start
> env_end
);
243 len1
= arg_end
- arg_start
;
244 len2
= env_end
- env_start
;
252 * Inherently racy -- command line shares address space
253 * with code and data.
255 rv
= access_remote_vm(mm
, arg_end
- 1, &c
, 1, 0);
262 /* Command line (set of strings) occupies whole ARGV. */
266 p
= arg_start
+ *pos
;
268 while (count
> 0 && len
> 0) {
272 _count
= min3(count
, len
, PAGE_SIZE
);
273 nr_read
= access_remote_vm(mm
, p
, page
, _count
, 0);
279 if (copy_to_user(buf
, page
, nr_read
)) {
292 * Command line (1 string) occupies ARGV and maybe
295 if (len1
+ len2
<= *pos
)
300 p
= arg_start
+ *pos
;
302 while (count
> 0 && len
> 0) {
303 unsigned int _count
, l
;
307 _count
= min3(count
, len
, PAGE_SIZE
);
308 nr_read
= access_remote_vm(mm
, p
, page
, _count
, 0);
315 * Command line can be shorter than whole ARGV
316 * even if last "marker" byte says it is not.
319 l
= strnlen(page
, nr_read
);
325 if (copy_to_user(buf
, page
, nr_read
)) {
341 * Command line (1 string) occupies ARGV and
345 p
= env_start
+ *pos
- len1
;
346 len
= len1
+ len2
- *pos
;
351 while (count
> 0 && len
> 0) {
352 unsigned int _count
, l
;
356 _count
= min3(count
, len
, PAGE_SIZE
);
357 nr_read
= access_remote_vm(mm
, p
, page
, _count
, 0);
365 l
= strnlen(page
, nr_read
);
371 if (copy_to_user(buf
, page
, nr_read
)) {
390 free_page((unsigned long)page
);
398 static const struct file_operations proc_pid_cmdline_ops
= {
399 .read
= proc_pid_cmdline_read
,
400 .llseek
= generic_file_llseek
,
403 static int proc_pid_auxv(struct seq_file
*m
, struct pid_namespace
*ns
,
404 struct pid
*pid
, struct task_struct
*task
)
406 struct mm_struct
*mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
407 if (mm
&& !IS_ERR(mm
)) {
408 unsigned int nwords
= 0;
411 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
412 seq_write(m
, mm
->saved_auxv
, nwords
* sizeof(mm
->saved_auxv
[0]));
420 #ifdef CONFIG_KALLSYMS
422 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
423 * Returns the resolved symbol. If that fails, simply return the address.
425 static int proc_pid_wchan(struct seq_file
*m
, struct pid_namespace
*ns
,
426 struct pid
*pid
, struct task_struct
*task
)
429 char symname
[KSYM_NAME_LEN
];
431 wchan
= get_wchan(task
);
433 if (wchan
&& ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)
434 && !lookup_symbol_name(wchan
, symname
))
435 seq_printf(m
, "%s", symname
);
441 #endif /* CONFIG_KALLSYMS */
443 static int lock_trace(struct task_struct
*task
)
445 int err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
448 if (!ptrace_may_access(task
, PTRACE_MODE_ATTACH_FSCREDS
)) {
449 mutex_unlock(&task
->signal
->cred_guard_mutex
);
455 static void unlock_trace(struct task_struct
*task
)
457 mutex_unlock(&task
->signal
->cred_guard_mutex
);
460 #ifdef CONFIG_STACKTRACE
462 #define MAX_STACK_TRACE_DEPTH 64
464 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
465 struct pid
*pid
, struct task_struct
*task
)
467 struct stack_trace trace
;
468 unsigned long *entries
;
472 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
476 trace
.nr_entries
= 0;
477 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
478 trace
.entries
= entries
;
481 err
= lock_trace(task
);
483 save_stack_trace_tsk(task
, &trace
);
485 for (i
= 0; i
< trace
.nr_entries
; i
++) {
486 seq_printf(m
, "[<%pK>] %pS\n",
487 (void *)entries
[i
], (void *)entries
[i
]);
497 #ifdef CONFIG_SCHED_INFO
499 * Provides /proc/PID/schedstat
501 static int proc_pid_schedstat(struct seq_file
*m
, struct pid_namespace
*ns
,
502 struct pid
*pid
, struct task_struct
*task
)
504 if (unlikely(!sched_info_on()))
505 seq_printf(m
, "0 0 0\n");
507 seq_printf(m
, "%llu %llu %lu\n",
508 (unsigned long long)task
->se
.sum_exec_runtime
,
509 (unsigned long long)task
->sched_info
.run_delay
,
510 task
->sched_info
.pcount
);
516 #ifdef CONFIG_LATENCYTOP
517 static int lstats_show_proc(struct seq_file
*m
, void *v
)
520 struct inode
*inode
= m
->private;
521 struct task_struct
*task
= get_proc_task(inode
);
525 seq_puts(m
, "Latency Top version : v0.1\n");
526 for (i
= 0; i
< 32; i
++) {
527 struct latency_record
*lr
= &task
->latency_record
[i
];
528 if (lr
->backtrace
[0]) {
530 seq_printf(m
, "%i %li %li",
531 lr
->count
, lr
->time
, lr
->max
);
532 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
533 unsigned long bt
= lr
->backtrace
[q
];
538 seq_printf(m
, " %ps", (void *)bt
);
544 put_task_struct(task
);
548 static int lstats_open(struct inode
*inode
, struct file
*file
)
550 return single_open(file
, lstats_show_proc
, inode
);
553 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
554 size_t count
, loff_t
*offs
)
556 struct task_struct
*task
= get_proc_task(file_inode(file
));
560 clear_all_latency_tracing(task
);
561 put_task_struct(task
);
566 static const struct file_operations proc_lstats_operations
= {
569 .write
= lstats_write
,
571 .release
= single_release
,
576 static int proc_oom_score(struct seq_file
*m
, struct pid_namespace
*ns
,
577 struct pid
*pid
, struct task_struct
*task
)
579 unsigned long totalpages
= totalram_pages
+ total_swap_pages
;
580 unsigned long points
= 0;
582 read_lock(&tasklist_lock
);
584 points
= oom_badness(task
, NULL
, NULL
, totalpages
) *
586 read_unlock(&tasklist_lock
);
587 seq_printf(m
, "%lu\n", points
);
597 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
598 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
599 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
600 [RLIMIT_DATA
] = {"Max data size", "bytes"},
601 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
602 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
603 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
604 [RLIMIT_NPROC
] = {"Max processes", "processes"},
605 [RLIMIT_NOFILE
] = {"Max open files", "files"},
606 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
607 [RLIMIT_AS
] = {"Max address space", "bytes"},
608 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
609 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
610 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
611 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
612 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
613 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
616 /* Display limits for a process */
617 static int proc_pid_limits(struct seq_file
*m
, struct pid_namespace
*ns
,
618 struct pid
*pid
, struct task_struct
*task
)
623 struct rlimit rlim
[RLIM_NLIMITS
];
625 if (!lock_task_sighand(task
, &flags
))
627 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
628 unlock_task_sighand(task
, &flags
);
631 * print the file header
633 seq_printf(m
, "%-25s %-20s %-20s %-10s\n",
634 "Limit", "Soft Limit", "Hard Limit", "Units");
636 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
637 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
638 seq_printf(m
, "%-25s %-20s ",
639 lnames
[i
].name
, "unlimited");
641 seq_printf(m
, "%-25s %-20lu ",
642 lnames
[i
].name
, rlim
[i
].rlim_cur
);
644 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
645 seq_printf(m
, "%-20s ", "unlimited");
647 seq_printf(m
, "%-20lu ", rlim
[i
].rlim_max
);
650 seq_printf(m
, "%-10s\n", lnames
[i
].unit
);
658 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
659 static int proc_pid_syscall(struct seq_file
*m
, struct pid_namespace
*ns
,
660 struct pid
*pid
, struct task_struct
*task
)
663 unsigned long args
[6], sp
, pc
;
666 res
= lock_trace(task
);
670 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
671 seq_puts(m
, "running\n");
673 seq_printf(m
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
676 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
678 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
684 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
686 /************************************************************************/
687 /* Here the fs part begins */
688 /************************************************************************/
690 /* permission checks */
691 static int proc_fd_access_allowed(struct inode
*inode
)
693 struct task_struct
*task
;
695 /* Allow access to a task's file descriptors if it is us or we
696 * may use ptrace attach to the process and find out that
699 task
= get_proc_task(inode
);
701 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
702 put_task_struct(task
);
707 int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
710 struct inode
*inode
= d_inode(dentry
);
712 if (attr
->ia_valid
& ATTR_MODE
)
715 error
= inode_change_ok(inode
, attr
);
719 setattr_copy(inode
, attr
);
720 mark_inode_dirty(inode
);
725 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
726 * or euid/egid (for hide_pid_min=2)?
728 static bool has_pid_permissions(struct pid_namespace
*pid
,
729 struct task_struct
*task
,
732 if (pid
->hide_pid
< hide_pid_min
)
734 if (in_group_p(pid
->pid_gid
))
736 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
740 static int proc_pid_permission(struct inode
*inode
, int mask
)
742 struct pid_namespace
*pid
= inode
->i_sb
->s_fs_info
;
743 struct task_struct
*task
;
746 task
= get_proc_task(inode
);
749 has_perms
= has_pid_permissions(pid
, task
, 1);
750 put_task_struct(task
);
753 if (pid
->hide_pid
== 2) {
755 * Let's make getdents(), stat(), and open()
756 * consistent with each other. If a process
757 * may not stat() a file, it shouldn't be seen
765 return generic_permission(inode
, mask
);
770 static const struct inode_operations proc_def_inode_operations
= {
771 .setattr
= proc_setattr
,
774 static int proc_single_show(struct seq_file
*m
, void *v
)
776 struct inode
*inode
= m
->private;
777 struct pid_namespace
*ns
;
779 struct task_struct
*task
;
782 ns
= inode
->i_sb
->s_fs_info
;
783 pid
= proc_pid(inode
);
784 task
= get_pid_task(pid
, PIDTYPE_PID
);
788 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
790 put_task_struct(task
);
794 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
796 return single_open(filp
, proc_single_show
, inode
);
799 static const struct file_operations proc_single_file_operations
= {
800 .open
= proc_single_open
,
803 .release
= single_release
,
807 struct mm_struct
*proc_mem_open(struct inode
*inode
, unsigned int mode
)
809 struct task_struct
*task
= get_proc_task(inode
);
810 struct mm_struct
*mm
= ERR_PTR(-ESRCH
);
813 mm
= mm_access(task
, mode
| PTRACE_MODE_FSCREDS
);
814 put_task_struct(task
);
816 if (!IS_ERR_OR_NULL(mm
)) {
817 /* ensure this mm_struct can't be freed */
818 atomic_inc(&mm
->mm_count
);
819 /* but do not pin its memory */
827 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
829 struct mm_struct
*mm
= proc_mem_open(inode
, mode
);
834 file
->private_data
= mm
;
838 static int mem_open(struct inode
*inode
, struct file
*file
)
840 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
842 /* OK to pass negative loff_t, we can catch out-of-range */
843 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
848 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
849 size_t count
, loff_t
*ppos
, int write
)
851 struct mm_struct
*mm
= file
->private_data
;
852 unsigned long addr
= *ppos
;
859 page
= (char *)__get_free_page(GFP_TEMPORARY
);
864 if (!atomic_inc_not_zero(&mm
->mm_users
))
868 int this_len
= min_t(int, count
, PAGE_SIZE
);
870 if (write
&& copy_from_user(page
, buf
, this_len
)) {
875 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, write
);
882 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
896 free_page((unsigned long) page
);
900 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
901 size_t count
, loff_t
*ppos
)
903 return mem_rw(file
, buf
, count
, ppos
, 0);
906 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
907 size_t count
, loff_t
*ppos
)
909 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
912 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
916 file
->f_pos
= offset
;
919 file
->f_pos
+= offset
;
924 force_successful_syscall_return();
928 static int mem_release(struct inode
*inode
, struct file
*file
)
930 struct mm_struct
*mm
= file
->private_data
;
936 static const struct file_operations proc_mem_operations
= {
941 .release
= mem_release
,
944 static int environ_open(struct inode
*inode
, struct file
*file
)
946 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
949 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
950 size_t count
, loff_t
*ppos
)
953 unsigned long src
= *ppos
;
955 struct mm_struct
*mm
= file
->private_data
;
956 unsigned long env_start
, env_end
;
958 /* Ensure the process spawned far enough to have an environment. */
959 if (!mm
|| !mm
->env_end
)
962 page
= (char *)__get_free_page(GFP_TEMPORARY
);
967 if (!atomic_inc_not_zero(&mm
->mm_users
))
970 down_read(&mm
->mmap_sem
);
971 env_start
= mm
->env_start
;
972 env_end
= mm
->env_end
;
973 up_read(&mm
->mmap_sem
);
976 size_t this_len
, max_len
;
979 if (src
>= (env_end
- env_start
))
982 this_len
= env_end
- (env_start
+ src
);
984 max_len
= min_t(size_t, PAGE_SIZE
, count
);
985 this_len
= min(max_len
, this_len
);
987 retval
= access_remote_vm(mm
, (env_start
+ src
),
995 if (copy_to_user(buf
, page
, retval
)) {
1009 free_page((unsigned long) page
);
1013 static const struct file_operations proc_environ_operations
= {
1014 .open
= environ_open
,
1015 .read
= environ_read
,
1016 .llseek
= generic_file_llseek
,
1017 .release
= mem_release
,
1020 static ssize_t
oom_adj_read(struct file
*file
, char __user
*buf
, size_t count
,
1023 struct task_struct
*task
= get_proc_task(file_inode(file
));
1024 char buffer
[PROC_NUMBUF
];
1025 int oom_adj
= OOM_ADJUST_MIN
;
1030 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MAX
)
1031 oom_adj
= OOM_ADJUST_MAX
;
1033 oom_adj
= (task
->signal
->oom_score_adj
* -OOM_DISABLE
) /
1035 put_task_struct(task
);
1036 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1037 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1040 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1042 static DEFINE_MUTEX(oom_adj_mutex
);
1043 struct task_struct
*task
;
1046 task
= get_proc_task(file_inode(file
));
1050 mutex_lock(&oom_adj_mutex
);
1052 if (oom_adj
< task
->signal
->oom_score_adj
&&
1053 !capable(CAP_SYS_RESOURCE
)) {
1058 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1059 * /proc/pid/oom_score_adj instead.
1061 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1062 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1065 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1066 !capable(CAP_SYS_RESOURCE
)) {
1072 task
->signal
->oom_score_adj
= oom_adj
;
1073 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1074 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1075 trace_oom_score_adj_update(task
);
1077 mutex_unlock(&oom_adj_mutex
);
1078 put_task_struct(task
);
1083 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1084 * kernels. The effective policy is defined by oom_score_adj, which has a
1085 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1086 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1087 * Processes that become oom disabled via oom_adj will still be oom disabled
1088 * with this implementation.
1090 * oom_adj cannot be removed since existing userspace binaries use it.
1092 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1093 size_t count
, loff_t
*ppos
)
1095 char buffer
[PROC_NUMBUF
];
1099 memset(buffer
, 0, sizeof(buffer
));
1100 if (count
> sizeof(buffer
) - 1)
1101 count
= sizeof(buffer
) - 1;
1102 if (copy_from_user(buffer
, buf
, count
)) {
1107 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1110 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1111 oom_adj
!= OOM_DISABLE
) {
1117 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1118 * value is always attainable.
1120 if (oom_adj
== OOM_ADJUST_MAX
)
1121 oom_adj
= OOM_SCORE_ADJ_MAX
;
1123 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1125 err
= __set_oom_adj(file
, oom_adj
, true);
1127 return err
< 0 ? err
: count
;
1130 static const struct file_operations proc_oom_adj_operations
= {
1131 .read
= oom_adj_read
,
1132 .write
= oom_adj_write
,
1133 .llseek
= generic_file_llseek
,
1136 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1137 size_t count
, loff_t
*ppos
)
1139 struct task_struct
*task
= get_proc_task(file_inode(file
));
1140 char buffer
[PROC_NUMBUF
];
1141 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1146 oom_score_adj
= task
->signal
->oom_score_adj
;
1147 put_task_struct(task
);
1148 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1149 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1152 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1153 size_t count
, loff_t
*ppos
)
1155 char buffer
[PROC_NUMBUF
];
1159 memset(buffer
, 0, sizeof(buffer
));
1160 if (count
> sizeof(buffer
) - 1)
1161 count
= sizeof(buffer
) - 1;
1162 if (copy_from_user(buffer
, buf
, count
)) {
1167 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1170 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1171 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1176 err
= __set_oom_adj(file
, oom_score_adj
, false);
1178 return err
< 0 ? err
: count
;
1181 static const struct file_operations proc_oom_score_adj_operations
= {
1182 .read
= oom_score_adj_read
,
1183 .write
= oom_score_adj_write
,
1184 .llseek
= default_llseek
,
1187 #ifdef CONFIG_AUDITSYSCALL
1188 #define TMPBUFLEN 21
1189 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1190 size_t count
, loff_t
*ppos
)
1192 struct inode
* inode
= file_inode(file
);
1193 struct task_struct
*task
= get_proc_task(inode
);
1195 char tmpbuf
[TMPBUFLEN
];
1199 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1200 from_kuid(file
->f_cred
->user_ns
,
1201 audit_get_loginuid(task
)));
1202 put_task_struct(task
);
1203 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1206 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1207 size_t count
, loff_t
*ppos
)
1209 struct inode
* inode
= file_inode(file
);
1215 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1222 /* No partial writes. */
1226 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1230 /* is userspace tring to explicitly UNSET the loginuid? */
1231 if (loginuid
== AUDIT_UID_UNSET
) {
1232 kloginuid
= INVALID_UID
;
1234 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1235 if (!uid_valid(kloginuid
))
1239 rv
= audit_set_loginuid(kloginuid
);
1245 static const struct file_operations proc_loginuid_operations
= {
1246 .read
= proc_loginuid_read
,
1247 .write
= proc_loginuid_write
,
1248 .llseek
= generic_file_llseek
,
1251 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1252 size_t count
, loff_t
*ppos
)
1254 struct inode
* inode
= file_inode(file
);
1255 struct task_struct
*task
= get_proc_task(inode
);
1257 char tmpbuf
[TMPBUFLEN
];
1261 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1262 audit_get_sessionid(task
));
1263 put_task_struct(task
);
1264 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1267 static const struct file_operations proc_sessionid_operations
= {
1268 .read
= proc_sessionid_read
,
1269 .llseek
= generic_file_llseek
,
1273 #ifdef CONFIG_FAULT_INJECTION
1274 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1275 size_t count
, loff_t
*ppos
)
1277 struct task_struct
*task
= get_proc_task(file_inode(file
));
1278 char buffer
[PROC_NUMBUF
];
1284 make_it_fail
= task
->make_it_fail
;
1285 put_task_struct(task
);
1287 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1289 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1292 static ssize_t
proc_fault_inject_write(struct file
* file
,
1293 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1295 struct task_struct
*task
;
1296 char buffer
[PROC_NUMBUF
];
1300 if (!capable(CAP_SYS_RESOURCE
))
1302 memset(buffer
, 0, sizeof(buffer
));
1303 if (count
> sizeof(buffer
) - 1)
1304 count
= sizeof(buffer
) - 1;
1305 if (copy_from_user(buffer
, buf
, count
))
1307 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1310 if (make_it_fail
< 0 || make_it_fail
> 1)
1313 task
= get_proc_task(file_inode(file
));
1316 task
->make_it_fail
= make_it_fail
;
1317 put_task_struct(task
);
1322 static const struct file_operations proc_fault_inject_operations
= {
1323 .read
= proc_fault_inject_read
,
1324 .write
= proc_fault_inject_write
,
1325 .llseek
= generic_file_llseek
,
1330 #ifdef CONFIG_SCHED_DEBUG
1332 * Print out various scheduling related per-task fields:
1334 static int sched_show(struct seq_file
*m
, void *v
)
1336 struct inode
*inode
= m
->private;
1337 struct task_struct
*p
;
1339 p
= get_proc_task(inode
);
1342 proc_sched_show_task(p
, m
);
1350 sched_write(struct file
*file
, const char __user
*buf
,
1351 size_t count
, loff_t
*offset
)
1353 struct inode
*inode
= file_inode(file
);
1354 struct task_struct
*p
;
1356 p
= get_proc_task(inode
);
1359 proc_sched_set_task(p
);
1366 static int sched_open(struct inode
*inode
, struct file
*filp
)
1368 return single_open(filp
, sched_show
, inode
);
1371 static const struct file_operations proc_pid_sched_operations
= {
1374 .write
= sched_write
,
1375 .llseek
= seq_lseek
,
1376 .release
= single_release
,
1381 #ifdef CONFIG_SCHED_AUTOGROUP
1383 * Print out autogroup related information:
1385 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1387 struct inode
*inode
= m
->private;
1388 struct task_struct
*p
;
1390 p
= get_proc_task(inode
);
1393 proc_sched_autogroup_show_task(p
, m
);
1401 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1402 size_t count
, loff_t
*offset
)
1404 struct inode
*inode
= file_inode(file
);
1405 struct task_struct
*p
;
1406 char buffer
[PROC_NUMBUF
];
1410 memset(buffer
, 0, sizeof(buffer
));
1411 if (count
> sizeof(buffer
) - 1)
1412 count
= sizeof(buffer
) - 1;
1413 if (copy_from_user(buffer
, buf
, count
))
1416 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1420 p
= get_proc_task(inode
);
1424 err
= proc_sched_autogroup_set_nice(p
, nice
);
1433 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1437 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1439 struct seq_file
*m
= filp
->private_data
;
1446 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1447 .open
= sched_autogroup_open
,
1449 .write
= sched_autogroup_write
,
1450 .llseek
= seq_lseek
,
1451 .release
= single_release
,
1454 #endif /* CONFIG_SCHED_AUTOGROUP */
1456 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1457 size_t count
, loff_t
*offset
)
1459 struct inode
*inode
= file_inode(file
);
1460 struct task_struct
*p
;
1461 char buffer
[TASK_COMM_LEN
];
1462 const size_t maxlen
= sizeof(buffer
) - 1;
1464 memset(buffer
, 0, sizeof(buffer
));
1465 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1468 p
= get_proc_task(inode
);
1472 if (same_thread_group(current
, p
))
1473 set_task_comm(p
, buffer
);
1482 static int comm_show(struct seq_file
*m
, void *v
)
1484 struct inode
*inode
= m
->private;
1485 struct task_struct
*p
;
1487 p
= get_proc_task(inode
);
1492 seq_printf(m
, "%s\n", p
->comm
);
1500 static int comm_open(struct inode
*inode
, struct file
*filp
)
1502 return single_open(filp
, comm_show
, inode
);
1505 static const struct file_operations proc_pid_set_comm_operations
= {
1508 .write
= comm_write
,
1509 .llseek
= seq_lseek
,
1510 .release
= single_release
,
1513 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1515 struct task_struct
*task
;
1516 struct mm_struct
*mm
;
1517 struct file
*exe_file
;
1519 task
= get_proc_task(d_inode(dentry
));
1522 mm
= get_task_mm(task
);
1523 put_task_struct(task
);
1526 exe_file
= get_mm_exe_file(mm
);
1529 *exe_path
= exe_file
->f_path
;
1530 path_get(&exe_file
->f_path
);
1537 static const char *proc_pid_get_link(struct dentry
*dentry
,
1538 struct inode
*inode
,
1539 struct delayed_call
*done
)
1542 int error
= -EACCES
;
1545 return ERR_PTR(-ECHILD
);
1547 /* Are we allowed to snoop on the tasks file descriptors? */
1548 if (!proc_fd_access_allowed(inode
))
1551 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1555 nd_jump_link(&path
);
1558 return ERR_PTR(error
);
1561 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1563 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1570 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1571 len
= PTR_ERR(pathname
);
1572 if (IS_ERR(pathname
))
1574 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1578 if (copy_to_user(buffer
, pathname
, len
))
1581 free_page((unsigned long)tmp
);
1585 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1587 int error
= -EACCES
;
1588 struct inode
*inode
= d_inode(dentry
);
1591 /* Are we allowed to snoop on the tasks file descriptors? */
1592 if (!proc_fd_access_allowed(inode
))
1595 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1599 error
= do_proc_readlink(&path
, buffer
, buflen
);
1605 const struct inode_operations proc_pid_link_inode_operations
= {
1606 .readlink
= proc_pid_readlink
,
1607 .get_link
= proc_pid_get_link
,
1608 .setattr
= proc_setattr
,
1612 /* building an inode */
1614 struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1616 struct inode
* inode
;
1617 struct proc_inode
*ei
;
1618 const struct cred
*cred
;
1620 /* We need a new inode */
1622 inode
= new_inode(sb
);
1628 inode
->i_ino
= get_next_ino();
1629 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1630 inode
->i_op
= &proc_def_inode_operations
;
1633 * grab the reference to task.
1635 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1639 if (task_dumpable(task
)) {
1641 cred
= __task_cred(task
);
1642 inode
->i_uid
= cred
->euid
;
1643 inode
->i_gid
= cred
->egid
;
1646 security_task_to_inode(task
, inode
);
1656 int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1658 struct inode
*inode
= d_inode(dentry
);
1659 struct task_struct
*task
;
1660 const struct cred
*cred
;
1661 struct pid_namespace
*pid
= dentry
->d_sb
->s_fs_info
;
1663 generic_fillattr(inode
, stat
);
1666 stat
->uid
= GLOBAL_ROOT_UID
;
1667 stat
->gid
= GLOBAL_ROOT_GID
;
1668 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1670 if (!has_pid_permissions(pid
, task
, 2)) {
1673 * This doesn't prevent learning whether PID exists,
1674 * it only makes getattr() consistent with readdir().
1678 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1679 task_dumpable(task
)) {
1680 cred
= __task_cred(task
);
1681 stat
->uid
= cred
->euid
;
1682 stat
->gid
= cred
->egid
;
1692 * Exceptional case: normally we are not allowed to unhash a busy
1693 * directory. In this case, however, we can do it - no aliasing problems
1694 * due to the way we treat inodes.
1696 * Rewrite the inode's ownerships here because the owning task may have
1697 * performed a setuid(), etc.
1699 * Before the /proc/pid/status file was created the only way to read
1700 * the effective uid of a /process was to stat /proc/pid. Reading
1701 * /proc/pid/status is slow enough that procps and other packages
1702 * kept stating /proc/pid. To keep the rules in /proc simple I have
1703 * made this apply to all per process world readable and executable
1706 int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1708 struct inode
*inode
;
1709 struct task_struct
*task
;
1710 const struct cred
*cred
;
1712 if (flags
& LOOKUP_RCU
)
1715 inode
= d_inode(dentry
);
1716 task
= get_proc_task(inode
);
1719 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1720 task_dumpable(task
)) {
1722 cred
= __task_cred(task
);
1723 inode
->i_uid
= cred
->euid
;
1724 inode
->i_gid
= cred
->egid
;
1727 inode
->i_uid
= GLOBAL_ROOT_UID
;
1728 inode
->i_gid
= GLOBAL_ROOT_GID
;
1730 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1731 security_task_to_inode(task
, inode
);
1732 put_task_struct(task
);
1738 static inline bool proc_inode_is_dead(struct inode
*inode
)
1740 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
1743 int pid_delete_dentry(const struct dentry
*dentry
)
1745 /* Is the task we represent dead?
1746 * If so, then don't put the dentry on the lru list,
1747 * kill it immediately.
1749 return proc_inode_is_dead(d_inode(dentry
));
1752 const struct dentry_operations pid_dentry_operations
=
1754 .d_revalidate
= pid_revalidate
,
1755 .d_delete
= pid_delete_dentry
,
1761 * Fill a directory entry.
1763 * If possible create the dcache entry and derive our inode number and
1764 * file type from dcache entry.
1766 * Since all of the proc inode numbers are dynamically generated, the inode
1767 * numbers do not exist until the inode is cache. This means creating the
1768 * the dcache entry in readdir is necessary to keep the inode numbers
1769 * reported by readdir in sync with the inode numbers reported
1772 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
1773 const char *name
, int len
,
1774 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1776 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
1777 struct qstr qname
= QSTR_INIT(name
, len
);
1778 struct inode
*inode
;
1782 child
= d_hash_and_lookup(dir
, &qname
);
1784 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
1785 child
= d_alloc_parallel(dir
, &qname
, &wq
);
1787 goto end_instantiate
;
1788 if (d_in_lookup(child
)) {
1789 int err
= instantiate(d_inode(dir
), child
, task
, ptr
);
1790 d_lookup_done(child
);
1793 goto end_instantiate
;
1797 inode
= d_inode(child
);
1799 type
= inode
->i_mode
>> 12;
1801 return dir_emit(ctx
, name
, len
, ino
, type
);
1804 return dir_emit(ctx
, name
, len
, 1, DT_UNKNOWN
);
1808 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1809 * which represent vma start and end addresses.
1811 static int dname_to_vma_addr(struct dentry
*dentry
,
1812 unsigned long *start
, unsigned long *end
)
1814 if (sscanf(dentry
->d_name
.name
, "%lx-%lx", start
, end
) != 2)
1820 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
1822 unsigned long vm_start
, vm_end
;
1823 bool exact_vma_exists
= false;
1824 struct mm_struct
*mm
= NULL
;
1825 struct task_struct
*task
;
1826 const struct cred
*cred
;
1827 struct inode
*inode
;
1830 if (flags
& LOOKUP_RCU
)
1833 inode
= d_inode(dentry
);
1834 task
= get_proc_task(inode
);
1838 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
1839 if (IS_ERR_OR_NULL(mm
))
1842 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
1843 down_read(&mm
->mmap_sem
);
1844 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
1845 up_read(&mm
->mmap_sem
);
1850 if (exact_vma_exists
) {
1851 if (task_dumpable(task
)) {
1853 cred
= __task_cred(task
);
1854 inode
->i_uid
= cred
->euid
;
1855 inode
->i_gid
= cred
->egid
;
1858 inode
->i_uid
= GLOBAL_ROOT_UID
;
1859 inode
->i_gid
= GLOBAL_ROOT_GID
;
1861 security_task_to_inode(task
, inode
);
1866 put_task_struct(task
);
1872 static const struct dentry_operations tid_map_files_dentry_operations
= {
1873 .d_revalidate
= map_files_d_revalidate
,
1874 .d_delete
= pid_delete_dentry
,
1877 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
1879 unsigned long vm_start
, vm_end
;
1880 struct vm_area_struct
*vma
;
1881 struct task_struct
*task
;
1882 struct mm_struct
*mm
;
1886 task
= get_proc_task(d_inode(dentry
));
1890 mm
= get_task_mm(task
);
1891 put_task_struct(task
);
1895 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
1900 down_read(&mm
->mmap_sem
);
1901 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
1902 if (vma
&& vma
->vm_file
) {
1903 *path
= vma
->vm_file
->f_path
;
1907 up_read(&mm
->mmap_sem
);
1915 struct map_files_info
{
1918 unsigned char name
[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1922 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1923 * symlinks may be used to bypass permissions on ancestor directories in the
1924 * path to the file in question.
1927 proc_map_files_get_link(struct dentry
*dentry
,
1928 struct inode
*inode
,
1929 struct delayed_call
*done
)
1931 if (!capable(CAP_SYS_ADMIN
))
1932 return ERR_PTR(-EPERM
);
1934 return proc_pid_get_link(dentry
, inode
, done
);
1938 * Identical to proc_pid_link_inode_operations except for get_link()
1940 static const struct inode_operations proc_map_files_link_inode_operations
= {
1941 .readlink
= proc_pid_readlink
,
1942 .get_link
= proc_map_files_get_link
,
1943 .setattr
= proc_setattr
,
1947 proc_map_files_instantiate(struct inode
*dir
, struct dentry
*dentry
,
1948 struct task_struct
*task
, const void *ptr
)
1950 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
1951 struct proc_inode
*ei
;
1952 struct inode
*inode
;
1954 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1959 ei
->op
.proc_get_link
= map_files_get_link
;
1961 inode
->i_op
= &proc_map_files_link_inode_operations
;
1963 inode
->i_mode
= S_IFLNK
;
1965 if (mode
& FMODE_READ
)
1966 inode
->i_mode
|= S_IRUSR
;
1967 if (mode
& FMODE_WRITE
)
1968 inode
->i_mode
|= S_IWUSR
;
1970 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
1971 d_add(dentry
, inode
);
1976 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
1977 struct dentry
*dentry
, unsigned int flags
)
1979 unsigned long vm_start
, vm_end
;
1980 struct vm_area_struct
*vma
;
1981 struct task_struct
*task
;
1983 struct mm_struct
*mm
;
1986 task
= get_proc_task(dir
);
1991 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
1995 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
1998 mm
= get_task_mm(task
);
2002 down_read(&mm
->mmap_sem
);
2003 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2008 result
= proc_map_files_instantiate(dir
, dentry
, task
,
2009 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2012 up_read(&mm
->mmap_sem
);
2015 put_task_struct(task
);
2017 return ERR_PTR(result
);
2020 static const struct inode_operations proc_map_files_inode_operations
= {
2021 .lookup
= proc_map_files_lookup
,
2022 .permission
= proc_fd_permission
,
2023 .setattr
= proc_setattr
,
2027 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2029 struct vm_area_struct
*vma
;
2030 struct task_struct
*task
;
2031 struct mm_struct
*mm
;
2032 unsigned long nr_files
, pos
, i
;
2033 struct flex_array
*fa
= NULL
;
2034 struct map_files_info info
;
2035 struct map_files_info
*p
;
2039 task
= get_proc_task(file_inode(file
));
2044 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2048 if (!dir_emit_dots(file
, ctx
))
2051 mm
= get_task_mm(task
);
2054 down_read(&mm
->mmap_sem
);
2059 * We need two passes here:
2061 * 1) Collect vmas of mapped files with mmap_sem taken
2062 * 2) Release mmap_sem and instantiate entries
2064 * otherwise we get lockdep complained, since filldir()
2065 * routine might require mmap_sem taken in might_fault().
2068 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2069 if (vma
->vm_file
&& ++pos
> ctx
->pos
)
2074 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2076 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2080 flex_array_free(fa
);
2081 up_read(&mm
->mmap_sem
);
2085 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2086 vma
= vma
->vm_next
) {
2089 if (++pos
<= ctx
->pos
)
2092 info
.mode
= vma
->vm_file
->f_mode
;
2093 info
.len
= snprintf(info
.name
,
2094 sizeof(info
.name
), "%lx-%lx",
2095 vma
->vm_start
, vma
->vm_end
);
2096 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2100 up_read(&mm
->mmap_sem
);
2102 for (i
= 0; i
< nr_files
; i
++) {
2103 p
= flex_array_get(fa
, i
);
2104 if (!proc_fill_cache(file
, ctx
,
2106 proc_map_files_instantiate
,
2108 (void *)(unsigned long)p
->mode
))
2113 flex_array_free(fa
);
2117 put_task_struct(task
);
2122 static const struct file_operations proc_map_files_operations
= {
2123 .read
= generic_read_dir
,
2124 .iterate_shared
= proc_map_files_readdir
,
2125 .llseek
= generic_file_llseek
,
2128 #ifdef CONFIG_CHECKPOINT_RESTORE
2129 struct timers_private
{
2131 struct task_struct
*task
;
2132 struct sighand_struct
*sighand
;
2133 struct pid_namespace
*ns
;
2134 unsigned long flags
;
2137 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2139 struct timers_private
*tp
= m
->private;
2141 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2143 return ERR_PTR(-ESRCH
);
2145 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2147 return ERR_PTR(-ESRCH
);
2149 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2152 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2154 struct timers_private
*tp
= m
->private;
2155 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2158 static void timers_stop(struct seq_file
*m
, void *v
)
2160 struct timers_private
*tp
= m
->private;
2163 unlock_task_sighand(tp
->task
, &tp
->flags
);
2168 put_task_struct(tp
->task
);
2173 static int show_timer(struct seq_file
*m
, void *v
)
2175 struct k_itimer
*timer
;
2176 struct timers_private
*tp
= m
->private;
2178 static const char * const nstr
[] = {
2179 [SIGEV_SIGNAL
] = "signal",
2180 [SIGEV_NONE
] = "none",
2181 [SIGEV_THREAD
] = "thread",
2184 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2185 notify
= timer
->it_sigev_notify
;
2187 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2188 seq_printf(m
, "signal: %d/%p\n",
2189 timer
->sigq
->info
.si_signo
,
2190 timer
->sigq
->info
.si_value
.sival_ptr
);
2191 seq_printf(m
, "notify: %s/%s.%d\n",
2192 nstr
[notify
& ~SIGEV_THREAD_ID
],
2193 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2194 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2195 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2200 static const struct seq_operations proc_timers_seq_ops
= {
2201 .start
= timers_start
,
2202 .next
= timers_next
,
2203 .stop
= timers_stop
,
2207 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2209 struct timers_private
*tp
;
2211 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2212 sizeof(struct timers_private
));
2216 tp
->pid
= proc_pid(inode
);
2217 tp
->ns
= inode
->i_sb
->s_fs_info
;
2221 static const struct file_operations proc_timers_operations
= {
2222 .open
= proc_timers_open
,
2224 .llseek
= seq_lseek
,
2225 .release
= seq_release_private
,
2229 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2230 size_t count
, loff_t
*offset
)
2232 struct inode
*inode
= file_inode(file
);
2233 struct task_struct
*p
;
2237 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2241 p
= get_proc_task(inode
);
2245 if (ptrace_may_access(p
, PTRACE_MODE_ATTACH_FSCREDS
)) {
2248 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2250 p
->timer_slack_ns
= slack_ns
;
2260 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2262 struct inode
*inode
= m
->private;
2263 struct task_struct
*p
;
2266 p
= get_proc_task(inode
);
2270 if (ptrace_may_access(p
, PTRACE_MODE_ATTACH_FSCREDS
)) {
2272 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2282 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2284 return single_open(filp
, timerslack_ns_show
, inode
);
2287 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2288 .open
= timerslack_ns_open
,
2290 .write
= timerslack_ns_write
,
2291 .llseek
= seq_lseek
,
2292 .release
= single_release
,
2295 static int proc_pident_instantiate(struct inode
*dir
,
2296 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2298 const struct pid_entry
*p
= ptr
;
2299 struct inode
*inode
;
2300 struct proc_inode
*ei
;
2302 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2307 inode
->i_mode
= p
->mode
;
2308 if (S_ISDIR(inode
->i_mode
))
2309 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2311 inode
->i_op
= p
->iop
;
2313 inode
->i_fop
= p
->fop
;
2315 d_set_d_op(dentry
, &pid_dentry_operations
);
2316 d_add(dentry
, inode
);
2317 /* Close the race of the process dying before we return the dentry */
2318 if (pid_revalidate(dentry
, 0))
2324 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2325 struct dentry
*dentry
,
2326 const struct pid_entry
*ents
,
2330 struct task_struct
*task
= get_proc_task(dir
);
2331 const struct pid_entry
*p
, *last
;
2339 * Yes, it does not scale. And it should not. Don't add
2340 * new entries into /proc/<tgid>/ without very good reasons.
2342 last
= &ents
[nents
- 1];
2343 for (p
= ents
; p
<= last
; p
++) {
2344 if (p
->len
!= dentry
->d_name
.len
)
2346 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2352 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2354 put_task_struct(task
);
2356 return ERR_PTR(error
);
2359 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2360 const struct pid_entry
*ents
, unsigned int nents
)
2362 struct task_struct
*task
= get_proc_task(file_inode(file
));
2363 const struct pid_entry
*p
;
2368 if (!dir_emit_dots(file
, ctx
))
2371 if (ctx
->pos
>= nents
+ 2)
2374 for (p
= ents
+ (ctx
->pos
- 2); p
<= ents
+ nents
- 1; p
++) {
2375 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2376 proc_pident_instantiate
, task
, p
))
2381 put_task_struct(task
);
2385 #ifdef CONFIG_SECURITY
2386 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2387 size_t count
, loff_t
*ppos
)
2389 struct inode
* inode
= file_inode(file
);
2392 struct task_struct
*task
= get_proc_task(inode
);
2397 length
= security_getprocattr(task
,
2398 (char*)file
->f_path
.dentry
->d_name
.name
,
2400 put_task_struct(task
);
2402 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2407 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2408 size_t count
, loff_t
*ppos
)
2410 struct inode
* inode
= file_inode(file
);
2413 struct task_struct
*task
= get_proc_task(inode
);
2418 if (count
> PAGE_SIZE
)
2421 /* No partial writes. */
2426 page
= memdup_user(buf
, count
);
2428 length
= PTR_ERR(page
);
2432 /* Guard against adverse ptrace interaction */
2433 length
= mutex_lock_interruptible(&task
->signal
->cred_guard_mutex
);
2437 length
= security_setprocattr(task
,
2438 (char*)file
->f_path
.dentry
->d_name
.name
,
2440 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2444 put_task_struct(task
);
2449 static const struct file_operations proc_pid_attr_operations
= {
2450 .read
= proc_pid_attr_read
,
2451 .write
= proc_pid_attr_write
,
2452 .llseek
= generic_file_llseek
,
2455 static const struct pid_entry attr_dir_stuff
[] = {
2456 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2457 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2458 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2459 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2460 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2461 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2464 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2466 return proc_pident_readdir(file
, ctx
,
2467 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2470 static const struct file_operations proc_attr_dir_operations
= {
2471 .read
= generic_read_dir
,
2472 .iterate_shared
= proc_attr_dir_readdir
,
2473 .llseek
= generic_file_llseek
,
2476 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2477 struct dentry
*dentry
, unsigned int flags
)
2479 return proc_pident_lookup(dir
, dentry
,
2480 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2483 static const struct inode_operations proc_attr_dir_inode_operations
= {
2484 .lookup
= proc_attr_dir_lookup
,
2485 .getattr
= pid_getattr
,
2486 .setattr
= proc_setattr
,
2491 #ifdef CONFIG_ELF_CORE
2492 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2493 size_t count
, loff_t
*ppos
)
2495 struct task_struct
*task
= get_proc_task(file_inode(file
));
2496 struct mm_struct
*mm
;
2497 char buffer
[PROC_NUMBUF
];
2505 mm
= get_task_mm(task
);
2507 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2508 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2509 MMF_DUMP_FILTER_SHIFT
));
2511 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2514 put_task_struct(task
);
2519 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2520 const char __user
*buf
,
2524 struct task_struct
*task
;
2525 struct mm_struct
*mm
;
2531 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2536 task
= get_proc_task(file_inode(file
));
2540 mm
= get_task_mm(task
);
2545 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2547 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2549 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2554 put_task_struct(task
);
2561 static const struct file_operations proc_coredump_filter_operations
= {
2562 .read
= proc_coredump_filter_read
,
2563 .write
= proc_coredump_filter_write
,
2564 .llseek
= generic_file_llseek
,
2568 #ifdef CONFIG_TASK_IO_ACCOUNTING
2569 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2571 struct task_io_accounting acct
= task
->ioac
;
2572 unsigned long flags
;
2575 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2579 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2584 if (whole
&& lock_task_sighand(task
, &flags
)) {
2585 struct task_struct
*t
= task
;
2587 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2588 while_each_thread(task
, t
)
2589 task_io_accounting_add(&acct
, &t
->ioac
);
2591 unlock_task_sighand(task
, &flags
);
2598 "read_bytes: %llu\n"
2599 "write_bytes: %llu\n"
2600 "cancelled_write_bytes: %llu\n",
2601 (unsigned long long)acct
.rchar
,
2602 (unsigned long long)acct
.wchar
,
2603 (unsigned long long)acct
.syscr
,
2604 (unsigned long long)acct
.syscw
,
2605 (unsigned long long)acct
.read_bytes
,
2606 (unsigned long long)acct
.write_bytes
,
2607 (unsigned long long)acct
.cancelled_write_bytes
);
2611 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2615 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2616 struct pid
*pid
, struct task_struct
*task
)
2618 return do_io_accounting(task
, m
, 0);
2621 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2622 struct pid
*pid
, struct task_struct
*task
)
2624 return do_io_accounting(task
, m
, 1);
2626 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2628 #ifdef CONFIG_USER_NS
2629 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2630 const struct seq_operations
*seq_ops
)
2632 struct user_namespace
*ns
= NULL
;
2633 struct task_struct
*task
;
2634 struct seq_file
*seq
;
2637 task
= get_proc_task(inode
);
2640 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2642 put_task_struct(task
);
2647 ret
= seq_open(file
, seq_ops
);
2651 seq
= file
->private_data
;
2661 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2663 struct seq_file
*seq
= file
->private_data
;
2664 struct user_namespace
*ns
= seq
->private;
2666 return seq_release(inode
, file
);
2669 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2671 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2674 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2676 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2679 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
2681 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
2684 static const struct file_operations proc_uid_map_operations
= {
2685 .open
= proc_uid_map_open
,
2686 .write
= proc_uid_map_write
,
2688 .llseek
= seq_lseek
,
2689 .release
= proc_id_map_release
,
2692 static const struct file_operations proc_gid_map_operations
= {
2693 .open
= proc_gid_map_open
,
2694 .write
= proc_gid_map_write
,
2696 .llseek
= seq_lseek
,
2697 .release
= proc_id_map_release
,
2700 static const struct file_operations proc_projid_map_operations
= {
2701 .open
= proc_projid_map_open
,
2702 .write
= proc_projid_map_write
,
2704 .llseek
= seq_lseek
,
2705 .release
= proc_id_map_release
,
2708 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
2710 struct user_namespace
*ns
= NULL
;
2711 struct task_struct
*task
;
2715 task
= get_proc_task(inode
);
2718 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2720 put_task_struct(task
);
2725 if (file
->f_mode
& FMODE_WRITE
) {
2727 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
2731 ret
= single_open(file
, &proc_setgroups_show
, ns
);
2742 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
2744 struct seq_file
*seq
= file
->private_data
;
2745 struct user_namespace
*ns
= seq
->private;
2746 int ret
= single_release(inode
, file
);
2751 static const struct file_operations proc_setgroups_operations
= {
2752 .open
= proc_setgroups_open
,
2753 .write
= proc_setgroups_write
,
2755 .llseek
= seq_lseek
,
2756 .release
= proc_setgroups_release
,
2758 #endif /* CONFIG_USER_NS */
2760 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2761 struct pid
*pid
, struct task_struct
*task
)
2763 int err
= lock_trace(task
);
2765 seq_printf(m
, "%08x\n", task
->personality
);
2774 static const struct file_operations proc_task_operations
;
2775 static const struct inode_operations proc_task_inode_operations
;
2777 static const struct pid_entry tgid_base_stuff
[] = {
2778 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2779 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2780 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
2781 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2782 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
2784 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2786 REG("environ", S_IRUSR
, proc_environ_operations
),
2787 ONE("auxv", S_IRUSR
, proc_pid_auxv
),
2788 ONE("status", S_IRUGO
, proc_pid_status
),
2789 ONE("personality", S_IRUSR
, proc_pid_personality
),
2790 ONE("limits", S_IRUGO
, proc_pid_limits
),
2791 #ifdef CONFIG_SCHED_DEBUG
2792 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2794 #ifdef CONFIG_SCHED_AUTOGROUP
2795 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
2797 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2798 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2799 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
2801 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
2802 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2803 ONE("statm", S_IRUGO
, proc_pid_statm
),
2804 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
2806 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
2808 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2809 LNK("cwd", proc_cwd_link
),
2810 LNK("root", proc_root_link
),
2811 LNK("exe", proc_exe_link
),
2812 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2813 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2814 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2815 #ifdef CONFIG_PROC_PAGE_MONITOR
2816 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2817 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
2818 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2820 #ifdef CONFIG_SECURITY
2821 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2823 #ifdef CONFIG_KALLSYMS
2824 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
2826 #ifdef CONFIG_STACKTRACE
2827 ONE("stack", S_IRUSR
, proc_pid_stack
),
2829 #ifdef CONFIG_SCHED_INFO
2830 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
2832 #ifdef CONFIG_LATENCYTOP
2833 REG("latency", S_IRUGO
, proc_lstats_operations
),
2835 #ifdef CONFIG_PROC_PID_CPUSET
2836 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
2838 #ifdef CONFIG_CGROUPS
2839 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
2841 ONE("oom_score", S_IRUGO
, proc_oom_score
),
2842 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
2843 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
2844 #ifdef CONFIG_AUDITSYSCALL
2845 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2846 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2848 #ifdef CONFIG_FAULT_INJECTION
2849 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2851 #ifdef CONFIG_ELF_CORE
2852 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2854 #ifdef CONFIG_TASK_IO_ACCOUNTING
2855 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
2857 #ifdef CONFIG_HARDWALL
2858 ONE("hardwall", S_IRUGO
, proc_pid_hardwall
),
2860 #ifdef CONFIG_USER_NS
2861 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
2862 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
2863 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
2864 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
2866 #ifdef CONFIG_CHECKPOINT_RESTORE
2867 REG("timers", S_IRUGO
, proc_timers_operations
),
2869 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
2872 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
2874 return proc_pident_readdir(file
, ctx
,
2875 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2878 static const struct file_operations proc_tgid_base_operations
= {
2879 .read
= generic_read_dir
,
2880 .iterate_shared
= proc_tgid_base_readdir
,
2881 .llseek
= generic_file_llseek
,
2884 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
2886 return proc_pident_lookup(dir
, dentry
,
2887 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2890 static const struct inode_operations proc_tgid_base_inode_operations
= {
2891 .lookup
= proc_tgid_base_lookup
,
2892 .getattr
= pid_getattr
,
2893 .setattr
= proc_setattr
,
2894 .permission
= proc_pid_permission
,
2897 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2899 struct dentry
*dentry
, *leader
, *dir
;
2900 char buf
[PROC_NUMBUF
];
2904 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2905 /* no ->d_hash() rejects on procfs */
2906 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2908 d_invalidate(dentry
);
2916 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2917 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2922 name
.len
= strlen(name
.name
);
2923 dir
= d_hash_and_lookup(leader
, &name
);
2925 goto out_put_leader
;
2928 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2929 dentry
= d_hash_and_lookup(dir
, &name
);
2931 d_invalidate(dentry
);
2943 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2944 * @task: task that should be flushed.
2946 * When flushing dentries from proc, one needs to flush them from global
2947 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2948 * in. This call is supposed to do all of this job.
2950 * Looks in the dcache for
2952 * /proc/@tgid/task/@pid
2953 * if either directory is present flushes it and all of it'ts children
2956 * It is safe and reasonable to cache /proc entries for a task until
2957 * that task exits. After that they just clog up the dcache with
2958 * useless entries, possibly causing useful dcache entries to be
2959 * flushed instead. This routine is proved to flush those useless
2960 * dcache entries at process exit time.
2962 * NOTE: This routine is just an optimization so it does not guarantee
2963 * that no dcache entries will exist at process exit time it
2964 * just makes it very unlikely that any will persist.
2967 void proc_flush_task(struct task_struct
*task
)
2970 struct pid
*pid
, *tgid
;
2973 pid
= task_pid(task
);
2974 tgid
= task_tgid(task
);
2976 for (i
= 0; i
<= pid
->level
; i
++) {
2977 upid
= &pid
->numbers
[i
];
2978 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2979 tgid
->numbers
[i
].nr
);
2983 static int proc_pid_instantiate(struct inode
*dir
,
2984 struct dentry
* dentry
,
2985 struct task_struct
*task
, const void *ptr
)
2987 struct inode
*inode
;
2989 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2993 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2994 inode
->i_op
= &proc_tgid_base_inode_operations
;
2995 inode
->i_fop
= &proc_tgid_base_operations
;
2996 inode
->i_flags
|=S_IMMUTABLE
;
2998 set_nlink(inode
, 2 + pid_entry_count_dirs(tgid_base_stuff
,
2999 ARRAY_SIZE(tgid_base_stuff
)));
3001 d_set_d_op(dentry
, &pid_dentry_operations
);
3003 d_add(dentry
, inode
);
3004 /* Close the race of the process dying before we return the dentry */
3005 if (pid_revalidate(dentry
, 0))
3011 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3013 int result
= -ENOENT
;
3014 struct task_struct
*task
;
3016 struct pid_namespace
*ns
;
3018 tgid
= name_to_int(&dentry
->d_name
);
3022 ns
= dentry
->d_sb
->s_fs_info
;
3024 task
= find_task_by_pid_ns(tgid
, ns
);
3026 get_task_struct(task
);
3031 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
3032 put_task_struct(task
);
3034 return ERR_PTR(result
);
3038 * Find the first task with tgid >= tgid
3043 struct task_struct
*task
;
3045 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3050 put_task_struct(iter
.task
);
3054 pid
= find_ge_pid(iter
.tgid
, ns
);
3056 iter
.tgid
= pid_nr_ns(pid
, ns
);
3057 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3058 /* What we to know is if the pid we have find is the
3059 * pid of a thread_group_leader. Testing for task
3060 * being a thread_group_leader is the obvious thing
3061 * todo but there is a window when it fails, due to
3062 * the pid transfer logic in de_thread.
3064 * So we perform the straight forward test of seeing
3065 * if the pid we have found is the pid of a thread
3066 * group leader, and don't worry if the task we have
3067 * found doesn't happen to be a thread group leader.
3068 * As we don't care in the case of readdir.
3070 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3074 get_task_struct(iter
.task
);
3080 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3082 /* for the /proc/ directory itself, after non-process stuff has been done */
3083 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3085 struct tgid_iter iter
;
3086 struct pid_namespace
*ns
= file_inode(file
)->i_sb
->s_fs_info
;
3087 loff_t pos
= ctx
->pos
;
3089 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3092 if (pos
== TGID_OFFSET
- 2) {
3093 struct inode
*inode
= d_inode(ns
->proc_self
);
3094 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3096 ctx
->pos
= pos
= pos
+ 1;
3098 if (pos
== TGID_OFFSET
- 1) {
3099 struct inode
*inode
= d_inode(ns
->proc_thread_self
);
3100 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3102 ctx
->pos
= pos
= pos
+ 1;
3104 iter
.tgid
= pos
- TGID_OFFSET
;
3106 for (iter
= next_tgid(ns
, iter
);
3108 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3109 char name
[PROC_NUMBUF
];
3111 if (!has_pid_permissions(ns
, iter
.task
, 2))
3114 len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
3115 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3116 if (!proc_fill_cache(file
, ctx
, name
, len
,
3117 proc_pid_instantiate
, iter
.task
, NULL
)) {
3118 put_task_struct(iter
.task
);
3122 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3127 * proc_tid_comm_permission is a special permission function exclusively
3128 * used for the node /proc/<pid>/task/<tid>/comm.
3129 * It bypasses generic permission checks in the case where a task of the same
3130 * task group attempts to access the node.
3131 * The rationale behind this is that glibc and bionic access this node for
3132 * cross thread naming (pthread_set/getname_np(!self)). However, if
3133 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3134 * which locks out the cross thread naming implementation.
3135 * This function makes sure that the node is always accessible for members of
3136 * same thread group.
3138 static int proc_tid_comm_permission(struct inode
*inode
, int mask
)
3140 bool is_same_tgroup
;
3141 struct task_struct
*task
;
3143 task
= get_proc_task(inode
);
3146 is_same_tgroup
= same_thread_group(current
, task
);
3147 put_task_struct(task
);
3149 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3150 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3151 * read or written by the members of the corresponding
3157 return generic_permission(inode
, mask
);
3160 static const struct inode_operations proc_tid_comm_inode_operations
= {
3161 .permission
= proc_tid_comm_permission
,
3167 static const struct pid_entry tid_base_stuff
[] = {
3168 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3169 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3170 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3172 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3174 REG("environ", S_IRUSR
, proc_environ_operations
),
3175 ONE("auxv", S_IRUSR
, proc_pid_auxv
),
3176 ONE("status", S_IRUGO
, proc_pid_status
),
3177 ONE("personality", S_IRUSR
, proc_pid_personality
),
3178 ONE("limits", S_IRUGO
, proc_pid_limits
),
3179 #ifdef CONFIG_SCHED_DEBUG
3180 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3182 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3183 &proc_tid_comm_inode_operations
,
3184 &proc_pid_set_comm_operations
, {}),
3185 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3186 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3188 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3189 ONE("stat", S_IRUGO
, proc_tid_stat
),
3190 ONE("statm", S_IRUGO
, proc_pid_statm
),
3191 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3192 #ifdef CONFIG_PROC_CHILDREN
3193 REG("children", S_IRUGO
, proc_tid_children_operations
),
3196 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3198 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3199 LNK("cwd", proc_cwd_link
),
3200 LNK("root", proc_root_link
),
3201 LNK("exe", proc_exe_link
),
3202 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3203 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3204 #ifdef CONFIG_PROC_PAGE_MONITOR
3205 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3206 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3207 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3209 #ifdef CONFIG_SECURITY
3210 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3212 #ifdef CONFIG_KALLSYMS
3213 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3215 #ifdef CONFIG_STACKTRACE
3216 ONE("stack", S_IRUSR
, proc_pid_stack
),
3218 #ifdef CONFIG_SCHED_INFO
3219 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3221 #ifdef CONFIG_LATENCYTOP
3222 REG("latency", S_IRUGO
, proc_lstats_operations
),
3224 #ifdef CONFIG_PROC_PID_CPUSET
3225 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3227 #ifdef CONFIG_CGROUPS
3228 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3230 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3231 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3232 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3233 #ifdef CONFIG_AUDITSYSCALL
3234 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3235 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3237 #ifdef CONFIG_FAULT_INJECTION
3238 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3240 #ifdef CONFIG_TASK_IO_ACCOUNTING
3241 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3243 #ifdef CONFIG_HARDWALL
3244 ONE("hardwall", S_IRUGO
, proc_pid_hardwall
),
3246 #ifdef CONFIG_USER_NS
3247 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3248 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3249 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3250 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3254 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3256 return proc_pident_readdir(file
, ctx
,
3257 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3260 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3262 return proc_pident_lookup(dir
, dentry
,
3263 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3266 static const struct file_operations proc_tid_base_operations
= {
3267 .read
= generic_read_dir
,
3268 .iterate_shared
= proc_tid_base_readdir
,
3269 .llseek
= generic_file_llseek
,
3272 static const struct inode_operations proc_tid_base_inode_operations
= {
3273 .lookup
= proc_tid_base_lookup
,
3274 .getattr
= pid_getattr
,
3275 .setattr
= proc_setattr
,
3278 static int proc_task_instantiate(struct inode
*dir
,
3279 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3281 struct inode
*inode
;
3282 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3286 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3287 inode
->i_op
= &proc_tid_base_inode_operations
;
3288 inode
->i_fop
= &proc_tid_base_operations
;
3289 inode
->i_flags
|=S_IMMUTABLE
;
3291 set_nlink(inode
, 2 + pid_entry_count_dirs(tid_base_stuff
,
3292 ARRAY_SIZE(tid_base_stuff
)));
3294 d_set_d_op(dentry
, &pid_dentry_operations
);
3296 d_add(dentry
, inode
);
3297 /* Close the race of the process dying before we return the dentry */
3298 if (pid_revalidate(dentry
, 0))
3304 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3306 int result
= -ENOENT
;
3307 struct task_struct
*task
;
3308 struct task_struct
*leader
= get_proc_task(dir
);
3310 struct pid_namespace
*ns
;
3315 tid
= name_to_int(&dentry
->d_name
);
3319 ns
= dentry
->d_sb
->s_fs_info
;
3321 task
= find_task_by_pid_ns(tid
, ns
);
3323 get_task_struct(task
);
3327 if (!same_thread_group(leader
, task
))
3330 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3332 put_task_struct(task
);
3334 put_task_struct(leader
);
3336 return ERR_PTR(result
);
3340 * Find the first tid of a thread group to return to user space.
3342 * Usually this is just the thread group leader, but if the users
3343 * buffer was too small or there was a seek into the middle of the
3344 * directory we have more work todo.
3346 * In the case of a short read we start with find_task_by_pid.
3348 * In the case of a seek we start with the leader and walk nr
3351 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3352 struct pid_namespace
*ns
)
3354 struct task_struct
*pos
, *task
;
3355 unsigned long nr
= f_pos
;
3357 if (nr
!= f_pos
) /* 32bit overflow? */
3361 task
= pid_task(pid
, PIDTYPE_PID
);
3365 /* Attempt to start with the tid of a thread */
3367 pos
= find_task_by_pid_ns(tid
, ns
);
3368 if (pos
&& same_thread_group(pos
, task
))
3372 /* If nr exceeds the number of threads there is nothing todo */
3373 if (nr
>= get_nr_threads(task
))
3376 /* If we haven't found our starting place yet start
3377 * with the leader and walk nr threads forward.
3379 pos
= task
= task
->group_leader
;
3383 } while_each_thread(task
, pos
);
3388 get_task_struct(pos
);
3395 * Find the next thread in the thread list.
3396 * Return NULL if there is an error or no next thread.
3398 * The reference to the input task_struct is released.
3400 static struct task_struct
*next_tid(struct task_struct
*start
)
3402 struct task_struct
*pos
= NULL
;
3404 if (pid_alive(start
)) {
3405 pos
= next_thread(start
);
3406 if (thread_group_leader(pos
))
3409 get_task_struct(pos
);
3412 put_task_struct(start
);
3416 /* for the /proc/TGID/task/ directories */
3417 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3419 struct inode
*inode
= file_inode(file
);
3420 struct task_struct
*task
;
3421 struct pid_namespace
*ns
;
3424 if (proc_inode_is_dead(inode
))
3427 if (!dir_emit_dots(file
, ctx
))
3430 /* f_version caches the tgid value that the last readdir call couldn't
3431 * return. lseek aka telldir automagically resets f_version to 0.
3433 ns
= inode
->i_sb
->s_fs_info
;
3434 tid
= (int)file
->f_version
;
3435 file
->f_version
= 0;
3436 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3438 task
= next_tid(task
), ctx
->pos
++) {
3439 char name
[PROC_NUMBUF
];
3441 tid
= task_pid_nr_ns(task
, ns
);
3442 len
= snprintf(name
, sizeof(name
), "%d", tid
);
3443 if (!proc_fill_cache(file
, ctx
, name
, len
,
3444 proc_task_instantiate
, task
, NULL
)) {
3445 /* returning this tgid failed, save it as the first
3446 * pid for the next readir call */
3447 file
->f_version
= (u64
)tid
;
3448 put_task_struct(task
);
3456 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3458 struct inode
*inode
= d_inode(dentry
);
3459 struct task_struct
*p
= get_proc_task(inode
);
3460 generic_fillattr(inode
, stat
);
3463 stat
->nlink
+= get_nr_threads(p
);
3470 static const struct inode_operations proc_task_inode_operations
= {
3471 .lookup
= proc_task_lookup
,
3472 .getattr
= proc_task_getattr
,
3473 .setattr
= proc_setattr
,
3474 .permission
= proc_pid_permission
,
3477 static const struct file_operations proc_task_operations
= {
3478 .read
= generic_read_dir
,
3479 .iterate_shared
= proc_task_readdir
,
3480 .llseek
= generic_file_llseek
,