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.
108 static u8 nlink_tgid
;
114 const struct inode_operations
*iop
;
115 const struct file_operations
*fop
;
119 #define NOD(NAME, MODE, IOP, FOP, OP) { \
121 .len = sizeof(NAME) - 1, \
128 #define DIR(NAME, MODE, iops, fops) \
129 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
130 #define LNK(NAME, get_link) \
131 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
132 &proc_pid_link_inode_operations, NULL, \
133 { .proc_get_link = get_link } )
134 #define REG(NAME, MODE, fops) \
135 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
136 #define ONE(NAME, MODE, show) \
137 NOD(NAME, (S_IFREG|(MODE)), \
138 NULL, &proc_single_file_operations, \
139 { .proc_show = show } )
142 * Count the number of hardlinks for the pid_entry table, excluding the .
145 static unsigned int __init
pid_entry_nlink(const struct pid_entry
*entries
,
152 for (i
= 0; i
< n
; ++i
) {
153 if (S_ISDIR(entries
[i
].mode
))
160 static int get_task_root(struct task_struct
*task
, struct path
*root
)
162 int result
= -ENOENT
;
166 get_fs_root(task
->fs
, root
);
173 static int proc_cwd_link(struct dentry
*dentry
, struct path
*path
)
175 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
176 int result
= -ENOENT
;
181 get_fs_pwd(task
->fs
, path
);
185 put_task_struct(task
);
190 static int proc_root_link(struct dentry
*dentry
, struct path
*path
)
192 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
193 int result
= -ENOENT
;
196 result
= get_task_root(task
, path
);
197 put_task_struct(task
);
202 static ssize_t
proc_pid_cmdline_read(struct file
*file
, char __user
*buf
,
203 size_t _count
, loff_t
*pos
)
205 struct task_struct
*tsk
;
206 struct mm_struct
*mm
;
208 unsigned long count
= _count
;
209 unsigned long arg_start
, arg_end
, env_start
, env_end
;
210 unsigned long len1
, len2
, len
;
217 tsk
= get_proc_task(file_inode(file
));
220 mm
= get_task_mm(tsk
);
221 put_task_struct(tsk
);
224 /* Check if process spawned far enough to have cmdline. */
230 page
= (char *)__get_free_page(GFP_TEMPORARY
);
236 down_read(&mm
->mmap_sem
);
237 arg_start
= mm
->arg_start
;
238 arg_end
= mm
->arg_end
;
239 env_start
= mm
->env_start
;
240 env_end
= mm
->env_end
;
241 up_read(&mm
->mmap_sem
);
243 BUG_ON(arg_start
> arg_end
);
244 BUG_ON(env_start
> env_end
);
246 len1
= arg_end
- arg_start
;
247 len2
= env_end
- env_start
;
255 * Inherently racy -- command line shares address space
256 * with code and data.
258 rv
= access_remote_vm(mm
, arg_end
- 1, &c
, 1, 0);
265 /* Command line (set of strings) occupies whole ARGV. */
269 p
= arg_start
+ *pos
;
271 while (count
> 0 && len
> 0) {
275 _count
= min3(count
, len
, PAGE_SIZE
);
276 nr_read
= access_remote_vm(mm
, p
, page
, _count
, 0);
282 if (copy_to_user(buf
, page
, nr_read
)) {
295 * Command line (1 string) occupies ARGV and maybe
298 if (len1
+ len2
<= *pos
)
303 p
= arg_start
+ *pos
;
305 while (count
> 0 && len
> 0) {
306 unsigned int _count
, l
;
310 _count
= min3(count
, len
, PAGE_SIZE
);
311 nr_read
= access_remote_vm(mm
, p
, page
, _count
, 0);
318 * Command line can be shorter than whole ARGV
319 * even if last "marker" byte says it is not.
322 l
= strnlen(page
, nr_read
);
328 if (copy_to_user(buf
, page
, nr_read
)) {
344 * Command line (1 string) occupies ARGV and
348 p
= env_start
+ *pos
- len1
;
349 len
= len1
+ len2
- *pos
;
354 while (count
> 0 && len
> 0) {
355 unsigned int _count
, l
;
359 _count
= min3(count
, len
, PAGE_SIZE
);
360 nr_read
= access_remote_vm(mm
, p
, page
, _count
, 0);
368 l
= strnlen(page
, nr_read
);
374 if (copy_to_user(buf
, page
, nr_read
)) {
393 free_page((unsigned long)page
);
401 static const struct file_operations proc_pid_cmdline_ops
= {
402 .read
= proc_pid_cmdline_read
,
403 .llseek
= generic_file_llseek
,
406 #ifdef CONFIG_KALLSYMS
408 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
409 * Returns the resolved symbol. If that fails, simply return the address.
411 static int proc_pid_wchan(struct seq_file
*m
, struct pid_namespace
*ns
,
412 struct pid
*pid
, struct task_struct
*task
)
415 char symname
[KSYM_NAME_LEN
];
417 wchan
= get_wchan(task
);
419 if (wchan
&& ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)
420 && !lookup_symbol_name(wchan
, symname
))
421 seq_printf(m
, "%s", symname
);
427 #endif /* CONFIG_KALLSYMS */
429 static int lock_trace(struct task_struct
*task
)
431 int err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
434 if (!ptrace_may_access(task
, PTRACE_MODE_ATTACH_FSCREDS
)) {
435 mutex_unlock(&task
->signal
->cred_guard_mutex
);
441 static void unlock_trace(struct task_struct
*task
)
443 mutex_unlock(&task
->signal
->cred_guard_mutex
);
446 #ifdef CONFIG_STACKTRACE
448 #define MAX_STACK_TRACE_DEPTH 64
450 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
451 struct pid
*pid
, struct task_struct
*task
)
453 struct stack_trace trace
;
454 unsigned long *entries
;
458 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
462 trace
.nr_entries
= 0;
463 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
464 trace
.entries
= entries
;
467 err
= lock_trace(task
);
469 save_stack_trace_tsk(task
, &trace
);
471 for (i
= 0; i
< trace
.nr_entries
; i
++) {
472 seq_printf(m
, "[<%pK>] %pB\n",
473 (void *)entries
[i
], (void *)entries
[i
]);
483 #ifdef CONFIG_SCHED_INFO
485 * Provides /proc/PID/schedstat
487 static int proc_pid_schedstat(struct seq_file
*m
, struct pid_namespace
*ns
,
488 struct pid
*pid
, struct task_struct
*task
)
490 if (unlikely(!sched_info_on()))
491 seq_printf(m
, "0 0 0\n");
493 seq_printf(m
, "%llu %llu %lu\n",
494 (unsigned long long)task
->se
.sum_exec_runtime
,
495 (unsigned long long)task
->sched_info
.run_delay
,
496 task
->sched_info
.pcount
);
502 #ifdef CONFIG_LATENCYTOP
503 static int lstats_show_proc(struct seq_file
*m
, void *v
)
506 struct inode
*inode
= m
->private;
507 struct task_struct
*task
= get_proc_task(inode
);
511 seq_puts(m
, "Latency Top version : v0.1\n");
512 for (i
= 0; i
< 32; i
++) {
513 struct latency_record
*lr
= &task
->latency_record
[i
];
514 if (lr
->backtrace
[0]) {
516 seq_printf(m
, "%i %li %li",
517 lr
->count
, lr
->time
, lr
->max
);
518 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
519 unsigned long bt
= lr
->backtrace
[q
];
524 seq_printf(m
, " %ps", (void *)bt
);
530 put_task_struct(task
);
534 static int lstats_open(struct inode
*inode
, struct file
*file
)
536 return single_open(file
, lstats_show_proc
, inode
);
539 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
540 size_t count
, loff_t
*offs
)
542 struct task_struct
*task
= get_proc_task(file_inode(file
));
546 clear_all_latency_tracing(task
);
547 put_task_struct(task
);
552 static const struct file_operations proc_lstats_operations
= {
555 .write
= lstats_write
,
557 .release
= single_release
,
562 static int proc_oom_score(struct seq_file
*m
, struct pid_namespace
*ns
,
563 struct pid
*pid
, struct task_struct
*task
)
565 unsigned long totalpages
= totalram_pages
+ total_swap_pages
;
566 unsigned long points
= 0;
568 points
= oom_badness(task
, NULL
, NULL
, totalpages
) *
570 seq_printf(m
, "%lu\n", points
);
580 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
581 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
582 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
583 [RLIMIT_DATA
] = {"Max data size", "bytes"},
584 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
585 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
586 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
587 [RLIMIT_NPROC
] = {"Max processes", "processes"},
588 [RLIMIT_NOFILE
] = {"Max open files", "files"},
589 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
590 [RLIMIT_AS
] = {"Max address space", "bytes"},
591 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
592 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
593 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
594 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
595 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
596 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
599 /* Display limits for a process */
600 static int proc_pid_limits(struct seq_file
*m
, struct pid_namespace
*ns
,
601 struct pid
*pid
, struct task_struct
*task
)
606 struct rlimit rlim
[RLIM_NLIMITS
];
608 if (!lock_task_sighand(task
, &flags
))
610 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
611 unlock_task_sighand(task
, &flags
);
614 * print the file header
616 seq_printf(m
, "%-25s %-20s %-20s %-10s\n",
617 "Limit", "Soft Limit", "Hard Limit", "Units");
619 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
620 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
621 seq_printf(m
, "%-25s %-20s ",
622 lnames
[i
].name
, "unlimited");
624 seq_printf(m
, "%-25s %-20lu ",
625 lnames
[i
].name
, rlim
[i
].rlim_cur
);
627 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
628 seq_printf(m
, "%-20s ", "unlimited");
630 seq_printf(m
, "%-20lu ", rlim
[i
].rlim_max
);
633 seq_printf(m
, "%-10s\n", lnames
[i
].unit
);
641 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
642 static int proc_pid_syscall(struct seq_file
*m
, struct pid_namespace
*ns
,
643 struct pid
*pid
, struct task_struct
*task
)
646 unsigned long args
[6], sp
, pc
;
649 res
= lock_trace(task
);
653 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
654 seq_puts(m
, "running\n");
656 seq_printf(m
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
659 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
661 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
667 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
669 /************************************************************************/
670 /* Here the fs part begins */
671 /************************************************************************/
673 /* permission checks */
674 static int proc_fd_access_allowed(struct inode
*inode
)
676 struct task_struct
*task
;
678 /* Allow access to a task's file descriptors if it is us or we
679 * may use ptrace attach to the process and find out that
682 task
= get_proc_task(inode
);
684 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
685 put_task_struct(task
);
690 int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
693 struct inode
*inode
= d_inode(dentry
);
695 if (attr
->ia_valid
& ATTR_MODE
)
698 error
= setattr_prepare(dentry
, attr
);
702 setattr_copy(inode
, attr
);
703 mark_inode_dirty(inode
);
708 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
709 * or euid/egid (for hide_pid_min=2)?
711 static bool has_pid_permissions(struct pid_namespace
*pid
,
712 struct task_struct
*task
,
715 if (pid
->hide_pid
< hide_pid_min
)
717 if (in_group_p(pid
->pid_gid
))
719 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
723 static int proc_pid_permission(struct inode
*inode
, int mask
)
725 struct pid_namespace
*pid
= inode
->i_sb
->s_fs_info
;
726 struct task_struct
*task
;
729 task
= get_proc_task(inode
);
732 has_perms
= has_pid_permissions(pid
, task
, 1);
733 put_task_struct(task
);
736 if (pid
->hide_pid
== 2) {
738 * Let's make getdents(), stat(), and open()
739 * consistent with each other. If a process
740 * may not stat() a file, it shouldn't be seen
748 return generic_permission(inode
, mask
);
753 static const struct inode_operations proc_def_inode_operations
= {
754 .setattr
= proc_setattr
,
757 static int proc_single_show(struct seq_file
*m
, void *v
)
759 struct inode
*inode
= m
->private;
760 struct pid_namespace
*ns
;
762 struct task_struct
*task
;
765 ns
= inode
->i_sb
->s_fs_info
;
766 pid
= proc_pid(inode
);
767 task
= get_pid_task(pid
, PIDTYPE_PID
);
771 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
773 put_task_struct(task
);
777 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
779 return single_open(filp
, proc_single_show
, inode
);
782 static const struct file_operations proc_single_file_operations
= {
783 .open
= proc_single_open
,
786 .release
= single_release
,
790 struct mm_struct
*proc_mem_open(struct inode
*inode
, unsigned int mode
)
792 struct task_struct
*task
= get_proc_task(inode
);
793 struct mm_struct
*mm
= ERR_PTR(-ESRCH
);
796 mm
= mm_access(task
, mode
| PTRACE_MODE_FSCREDS
);
797 put_task_struct(task
);
799 if (!IS_ERR_OR_NULL(mm
)) {
800 /* ensure this mm_struct can't be freed */
801 atomic_inc(&mm
->mm_count
);
802 /* but do not pin its memory */
810 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
812 struct mm_struct
*mm
= proc_mem_open(inode
, mode
);
817 file
->private_data
= mm
;
821 static int mem_open(struct inode
*inode
, struct file
*file
)
823 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
825 /* OK to pass negative loff_t, we can catch out-of-range */
826 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
831 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
832 size_t count
, loff_t
*ppos
, int write
)
834 struct mm_struct
*mm
= file
->private_data
;
835 unsigned long addr
= *ppos
;
843 page
= (char *)__get_free_page(GFP_TEMPORARY
);
848 if (!atomic_inc_not_zero(&mm
->mm_users
))
851 /* Maybe we should limit FOLL_FORCE to actual ptrace users? */
857 int this_len
= min_t(int, count
, PAGE_SIZE
);
859 if (write
&& copy_from_user(page
, buf
, this_len
)) {
864 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, flags
);
871 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
885 free_page((unsigned long) page
);
889 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
890 size_t count
, loff_t
*ppos
)
892 return mem_rw(file
, buf
, count
, ppos
, 0);
895 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
896 size_t count
, loff_t
*ppos
)
898 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
901 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
905 file
->f_pos
= offset
;
908 file
->f_pos
+= offset
;
913 force_successful_syscall_return();
917 static int mem_release(struct inode
*inode
, struct file
*file
)
919 struct mm_struct
*mm
= file
->private_data
;
925 static const struct file_operations proc_mem_operations
= {
930 .release
= mem_release
,
933 static int environ_open(struct inode
*inode
, struct file
*file
)
935 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
938 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
939 size_t count
, loff_t
*ppos
)
942 unsigned long src
= *ppos
;
944 struct mm_struct
*mm
= file
->private_data
;
945 unsigned long env_start
, env_end
;
947 /* Ensure the process spawned far enough to have an environment. */
948 if (!mm
|| !mm
->env_end
)
951 page
= (char *)__get_free_page(GFP_TEMPORARY
);
956 if (!atomic_inc_not_zero(&mm
->mm_users
))
959 down_read(&mm
->mmap_sem
);
960 env_start
= mm
->env_start
;
961 env_end
= mm
->env_end
;
962 up_read(&mm
->mmap_sem
);
965 size_t this_len
, max_len
;
968 if (src
>= (env_end
- env_start
))
971 this_len
= env_end
- (env_start
+ src
);
973 max_len
= min_t(size_t, PAGE_SIZE
, count
);
974 this_len
= min(max_len
, this_len
);
976 retval
= access_remote_vm(mm
, (env_start
+ src
), page
, this_len
, 0);
983 if (copy_to_user(buf
, page
, retval
)) {
997 free_page((unsigned long) page
);
1001 static const struct file_operations proc_environ_operations
= {
1002 .open
= environ_open
,
1003 .read
= environ_read
,
1004 .llseek
= generic_file_llseek
,
1005 .release
= mem_release
,
1008 static int auxv_open(struct inode
*inode
, struct file
*file
)
1010 return __mem_open(inode
, file
, PTRACE_MODE_READ_FSCREDS
);
1013 static ssize_t
auxv_read(struct file
*file
, char __user
*buf
,
1014 size_t count
, loff_t
*ppos
)
1016 struct mm_struct
*mm
= file
->private_data
;
1017 unsigned int nwords
= 0;
1023 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
1024 return simple_read_from_buffer(buf
, count
, ppos
, mm
->saved_auxv
,
1025 nwords
* sizeof(mm
->saved_auxv
[0]));
1028 static const struct file_operations proc_auxv_operations
= {
1031 .llseek
= generic_file_llseek
,
1032 .release
= mem_release
,
1035 static ssize_t
oom_adj_read(struct file
*file
, char __user
*buf
, size_t count
,
1038 struct task_struct
*task
= get_proc_task(file_inode(file
));
1039 char buffer
[PROC_NUMBUF
];
1040 int oom_adj
= OOM_ADJUST_MIN
;
1045 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MAX
)
1046 oom_adj
= OOM_ADJUST_MAX
;
1048 oom_adj
= (task
->signal
->oom_score_adj
* -OOM_DISABLE
) /
1050 put_task_struct(task
);
1051 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1052 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1055 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1057 static DEFINE_MUTEX(oom_adj_mutex
);
1058 struct mm_struct
*mm
= NULL
;
1059 struct task_struct
*task
;
1062 task
= get_proc_task(file_inode(file
));
1066 mutex_lock(&oom_adj_mutex
);
1068 if (oom_adj
< task
->signal
->oom_score_adj
&&
1069 !capable(CAP_SYS_RESOURCE
)) {
1074 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1075 * /proc/pid/oom_score_adj instead.
1077 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1078 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1081 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1082 !capable(CAP_SYS_RESOURCE
)) {
1089 * Make sure we will check other processes sharing the mm if this is
1090 * not vfrok which wants its own oom_score_adj.
1091 * pin the mm so it doesn't go away and get reused after task_unlock
1093 if (!task
->vfork_done
) {
1094 struct task_struct
*p
= find_lock_task_mm(task
);
1097 if (atomic_read(&p
->mm
->mm_users
) > 1) {
1099 atomic_inc(&mm
->mm_count
);
1105 task
->signal
->oom_score_adj
= oom_adj
;
1106 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1107 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1108 trace_oom_score_adj_update(task
);
1111 struct task_struct
*p
;
1114 for_each_process(p
) {
1115 if (same_thread_group(task
, p
))
1118 /* do not touch kernel threads or the global init */
1119 if (p
->flags
& PF_KTHREAD
|| is_global_init(p
))
1123 if (!p
->vfork_done
&& process_shares_mm(p
, mm
)) {
1124 pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1125 task_pid_nr(p
), p
->comm
,
1126 p
->signal
->oom_score_adj
, oom_adj
,
1127 task_pid_nr(task
), task
->comm
);
1128 p
->signal
->oom_score_adj
= oom_adj
;
1129 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1130 p
->signal
->oom_score_adj_min
= (short)oom_adj
;
1138 mutex_unlock(&oom_adj_mutex
);
1139 put_task_struct(task
);
1144 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1145 * kernels. The effective policy is defined by oom_score_adj, which has a
1146 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1147 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1148 * Processes that become oom disabled via oom_adj will still be oom disabled
1149 * with this implementation.
1151 * oom_adj cannot be removed since existing userspace binaries use it.
1153 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1154 size_t count
, loff_t
*ppos
)
1156 char buffer
[PROC_NUMBUF
];
1160 memset(buffer
, 0, sizeof(buffer
));
1161 if (count
> sizeof(buffer
) - 1)
1162 count
= sizeof(buffer
) - 1;
1163 if (copy_from_user(buffer
, buf
, count
)) {
1168 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1171 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1172 oom_adj
!= OOM_DISABLE
) {
1178 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1179 * value is always attainable.
1181 if (oom_adj
== OOM_ADJUST_MAX
)
1182 oom_adj
= OOM_SCORE_ADJ_MAX
;
1184 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1186 err
= __set_oom_adj(file
, oom_adj
, true);
1188 return err
< 0 ? err
: count
;
1191 static const struct file_operations proc_oom_adj_operations
= {
1192 .read
= oom_adj_read
,
1193 .write
= oom_adj_write
,
1194 .llseek
= generic_file_llseek
,
1197 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1198 size_t count
, loff_t
*ppos
)
1200 struct task_struct
*task
= get_proc_task(file_inode(file
));
1201 char buffer
[PROC_NUMBUF
];
1202 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1207 oom_score_adj
= task
->signal
->oom_score_adj
;
1208 put_task_struct(task
);
1209 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1210 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1213 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1214 size_t count
, loff_t
*ppos
)
1216 char buffer
[PROC_NUMBUF
];
1220 memset(buffer
, 0, sizeof(buffer
));
1221 if (count
> sizeof(buffer
) - 1)
1222 count
= sizeof(buffer
) - 1;
1223 if (copy_from_user(buffer
, buf
, count
)) {
1228 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1231 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1232 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1237 err
= __set_oom_adj(file
, oom_score_adj
, false);
1239 return err
< 0 ? err
: count
;
1242 static const struct file_operations proc_oom_score_adj_operations
= {
1243 .read
= oom_score_adj_read
,
1244 .write
= oom_score_adj_write
,
1245 .llseek
= default_llseek
,
1248 #ifdef CONFIG_AUDITSYSCALL
1249 #define TMPBUFLEN 21
1250 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1251 size_t count
, loff_t
*ppos
)
1253 struct inode
* inode
= file_inode(file
);
1254 struct task_struct
*task
= get_proc_task(inode
);
1256 char tmpbuf
[TMPBUFLEN
];
1260 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1261 from_kuid(file
->f_cred
->user_ns
,
1262 audit_get_loginuid(task
)));
1263 put_task_struct(task
);
1264 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1267 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1268 size_t count
, loff_t
*ppos
)
1270 struct inode
* inode
= file_inode(file
);
1276 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1283 /* No partial writes. */
1287 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1291 /* is userspace tring to explicitly UNSET the loginuid? */
1292 if (loginuid
== AUDIT_UID_UNSET
) {
1293 kloginuid
= INVALID_UID
;
1295 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1296 if (!uid_valid(kloginuid
))
1300 rv
= audit_set_loginuid(kloginuid
);
1306 static const struct file_operations proc_loginuid_operations
= {
1307 .read
= proc_loginuid_read
,
1308 .write
= proc_loginuid_write
,
1309 .llseek
= generic_file_llseek
,
1312 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1313 size_t count
, loff_t
*ppos
)
1315 struct inode
* inode
= file_inode(file
);
1316 struct task_struct
*task
= get_proc_task(inode
);
1318 char tmpbuf
[TMPBUFLEN
];
1322 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1323 audit_get_sessionid(task
));
1324 put_task_struct(task
);
1325 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1328 static const struct file_operations proc_sessionid_operations
= {
1329 .read
= proc_sessionid_read
,
1330 .llseek
= generic_file_llseek
,
1334 #ifdef CONFIG_FAULT_INJECTION
1335 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1336 size_t count
, loff_t
*ppos
)
1338 struct task_struct
*task
= get_proc_task(file_inode(file
));
1339 char buffer
[PROC_NUMBUF
];
1345 make_it_fail
= task
->make_it_fail
;
1346 put_task_struct(task
);
1348 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1350 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1353 static ssize_t
proc_fault_inject_write(struct file
* file
,
1354 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1356 struct task_struct
*task
;
1357 char buffer
[PROC_NUMBUF
];
1361 if (!capable(CAP_SYS_RESOURCE
))
1363 memset(buffer
, 0, sizeof(buffer
));
1364 if (count
> sizeof(buffer
) - 1)
1365 count
= sizeof(buffer
) - 1;
1366 if (copy_from_user(buffer
, buf
, count
))
1368 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1371 if (make_it_fail
< 0 || make_it_fail
> 1)
1374 task
= get_proc_task(file_inode(file
));
1377 task
->make_it_fail
= make_it_fail
;
1378 put_task_struct(task
);
1383 static const struct file_operations proc_fault_inject_operations
= {
1384 .read
= proc_fault_inject_read
,
1385 .write
= proc_fault_inject_write
,
1386 .llseek
= generic_file_llseek
,
1391 #ifdef CONFIG_SCHED_DEBUG
1393 * Print out various scheduling related per-task fields:
1395 static int sched_show(struct seq_file
*m
, void *v
)
1397 struct inode
*inode
= m
->private;
1398 struct task_struct
*p
;
1400 p
= get_proc_task(inode
);
1403 proc_sched_show_task(p
, m
);
1411 sched_write(struct file
*file
, const char __user
*buf
,
1412 size_t count
, loff_t
*offset
)
1414 struct inode
*inode
= file_inode(file
);
1415 struct task_struct
*p
;
1417 p
= get_proc_task(inode
);
1420 proc_sched_set_task(p
);
1427 static int sched_open(struct inode
*inode
, struct file
*filp
)
1429 return single_open(filp
, sched_show
, inode
);
1432 static const struct file_operations proc_pid_sched_operations
= {
1435 .write
= sched_write
,
1436 .llseek
= seq_lseek
,
1437 .release
= single_release
,
1442 #ifdef CONFIG_SCHED_AUTOGROUP
1444 * Print out autogroup related information:
1446 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1448 struct inode
*inode
= m
->private;
1449 struct task_struct
*p
;
1451 p
= get_proc_task(inode
);
1454 proc_sched_autogroup_show_task(p
, m
);
1462 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1463 size_t count
, loff_t
*offset
)
1465 struct inode
*inode
= file_inode(file
);
1466 struct task_struct
*p
;
1467 char buffer
[PROC_NUMBUF
];
1471 memset(buffer
, 0, sizeof(buffer
));
1472 if (count
> sizeof(buffer
) - 1)
1473 count
= sizeof(buffer
) - 1;
1474 if (copy_from_user(buffer
, buf
, count
))
1477 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1481 p
= get_proc_task(inode
);
1485 err
= proc_sched_autogroup_set_nice(p
, nice
);
1494 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1498 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1500 struct seq_file
*m
= filp
->private_data
;
1507 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1508 .open
= sched_autogroup_open
,
1510 .write
= sched_autogroup_write
,
1511 .llseek
= seq_lseek
,
1512 .release
= single_release
,
1515 #endif /* CONFIG_SCHED_AUTOGROUP */
1517 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1518 size_t count
, loff_t
*offset
)
1520 struct inode
*inode
= file_inode(file
);
1521 struct task_struct
*p
;
1522 char buffer
[TASK_COMM_LEN
];
1523 const size_t maxlen
= sizeof(buffer
) - 1;
1525 memset(buffer
, 0, sizeof(buffer
));
1526 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1529 p
= get_proc_task(inode
);
1533 if (same_thread_group(current
, p
))
1534 set_task_comm(p
, buffer
);
1543 static int comm_show(struct seq_file
*m
, void *v
)
1545 struct inode
*inode
= m
->private;
1546 struct task_struct
*p
;
1548 p
= get_proc_task(inode
);
1553 seq_printf(m
, "%s\n", p
->comm
);
1561 static int comm_open(struct inode
*inode
, struct file
*filp
)
1563 return single_open(filp
, comm_show
, inode
);
1566 static const struct file_operations proc_pid_set_comm_operations
= {
1569 .write
= comm_write
,
1570 .llseek
= seq_lseek
,
1571 .release
= single_release
,
1574 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1576 struct task_struct
*task
;
1577 struct file
*exe_file
;
1579 task
= get_proc_task(d_inode(dentry
));
1582 exe_file
= get_task_exe_file(task
);
1583 put_task_struct(task
);
1585 *exe_path
= exe_file
->f_path
;
1586 path_get(&exe_file
->f_path
);
1593 static const char *proc_pid_get_link(struct dentry
*dentry
,
1594 struct inode
*inode
,
1595 struct delayed_call
*done
)
1598 int error
= -EACCES
;
1601 return ERR_PTR(-ECHILD
);
1603 /* Are we allowed to snoop on the tasks file descriptors? */
1604 if (!proc_fd_access_allowed(inode
))
1607 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1611 nd_jump_link(&path
);
1614 return ERR_PTR(error
);
1617 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1619 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1626 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1627 len
= PTR_ERR(pathname
);
1628 if (IS_ERR(pathname
))
1630 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1634 if (copy_to_user(buffer
, pathname
, len
))
1637 free_page((unsigned long)tmp
);
1641 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1643 int error
= -EACCES
;
1644 struct inode
*inode
= d_inode(dentry
);
1647 /* Are we allowed to snoop on the tasks file descriptors? */
1648 if (!proc_fd_access_allowed(inode
))
1651 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1655 error
= do_proc_readlink(&path
, buffer
, buflen
);
1661 const struct inode_operations proc_pid_link_inode_operations
= {
1662 .readlink
= proc_pid_readlink
,
1663 .get_link
= proc_pid_get_link
,
1664 .setattr
= proc_setattr
,
1668 /* building an inode */
1670 struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1672 struct inode
* inode
;
1673 struct proc_inode
*ei
;
1674 const struct cred
*cred
;
1676 /* We need a new inode */
1678 inode
= new_inode(sb
);
1684 inode
->i_ino
= get_next_ino();
1685 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= current_time(inode
);
1686 inode
->i_op
= &proc_def_inode_operations
;
1689 * grab the reference to task.
1691 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1695 if (task_dumpable(task
)) {
1697 cred
= __task_cred(task
);
1698 inode
->i_uid
= cred
->euid
;
1699 inode
->i_gid
= cred
->egid
;
1702 security_task_to_inode(task
, inode
);
1712 int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1714 struct inode
*inode
= d_inode(dentry
);
1715 struct task_struct
*task
;
1716 const struct cred
*cred
;
1717 struct pid_namespace
*pid
= dentry
->d_sb
->s_fs_info
;
1719 generic_fillattr(inode
, stat
);
1722 stat
->uid
= GLOBAL_ROOT_UID
;
1723 stat
->gid
= GLOBAL_ROOT_GID
;
1724 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1726 if (!has_pid_permissions(pid
, task
, 2)) {
1729 * This doesn't prevent learning whether PID exists,
1730 * it only makes getattr() consistent with readdir().
1734 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1735 task_dumpable(task
)) {
1736 cred
= __task_cred(task
);
1737 stat
->uid
= cred
->euid
;
1738 stat
->gid
= cred
->egid
;
1748 * Exceptional case: normally we are not allowed to unhash a busy
1749 * directory. In this case, however, we can do it - no aliasing problems
1750 * due to the way we treat inodes.
1752 * Rewrite the inode's ownerships here because the owning task may have
1753 * performed a setuid(), etc.
1755 * Before the /proc/pid/status file was created the only way to read
1756 * the effective uid of a /process was to stat /proc/pid. Reading
1757 * /proc/pid/status is slow enough that procps and other packages
1758 * kept stating /proc/pid. To keep the rules in /proc simple I have
1759 * made this apply to all per process world readable and executable
1762 int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1764 struct inode
*inode
;
1765 struct task_struct
*task
;
1766 const struct cred
*cred
;
1768 if (flags
& LOOKUP_RCU
)
1771 inode
= d_inode(dentry
);
1772 task
= get_proc_task(inode
);
1775 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1776 task_dumpable(task
)) {
1778 cred
= __task_cred(task
);
1779 inode
->i_uid
= cred
->euid
;
1780 inode
->i_gid
= cred
->egid
;
1783 inode
->i_uid
= GLOBAL_ROOT_UID
;
1784 inode
->i_gid
= GLOBAL_ROOT_GID
;
1786 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1787 security_task_to_inode(task
, inode
);
1788 put_task_struct(task
);
1794 static inline bool proc_inode_is_dead(struct inode
*inode
)
1796 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
1799 int pid_delete_dentry(const struct dentry
*dentry
)
1801 /* Is the task we represent dead?
1802 * If so, then don't put the dentry on the lru list,
1803 * kill it immediately.
1805 return proc_inode_is_dead(d_inode(dentry
));
1808 const struct dentry_operations pid_dentry_operations
=
1810 .d_revalidate
= pid_revalidate
,
1811 .d_delete
= pid_delete_dentry
,
1817 * Fill a directory entry.
1819 * If possible create the dcache entry and derive our inode number and
1820 * file type from dcache entry.
1822 * Since all of the proc inode numbers are dynamically generated, the inode
1823 * numbers do not exist until the inode is cache. This means creating the
1824 * the dcache entry in readdir is necessary to keep the inode numbers
1825 * reported by readdir in sync with the inode numbers reported
1828 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
1829 const char *name
, int len
,
1830 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1832 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
1833 struct qstr qname
= QSTR_INIT(name
, len
);
1834 struct inode
*inode
;
1838 child
= d_hash_and_lookup(dir
, &qname
);
1840 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
1841 child
= d_alloc_parallel(dir
, &qname
, &wq
);
1843 goto end_instantiate
;
1844 if (d_in_lookup(child
)) {
1845 int err
= instantiate(d_inode(dir
), child
, task
, ptr
);
1846 d_lookup_done(child
);
1849 goto end_instantiate
;
1853 inode
= d_inode(child
);
1855 type
= inode
->i_mode
>> 12;
1857 return dir_emit(ctx
, name
, len
, ino
, type
);
1860 return dir_emit(ctx
, name
, len
, 1, DT_UNKNOWN
);
1864 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1865 * which represent vma start and end addresses.
1867 static int dname_to_vma_addr(struct dentry
*dentry
,
1868 unsigned long *start
, unsigned long *end
)
1870 if (sscanf(dentry
->d_name
.name
, "%lx-%lx", start
, end
) != 2)
1876 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
1878 unsigned long vm_start
, vm_end
;
1879 bool exact_vma_exists
= false;
1880 struct mm_struct
*mm
= NULL
;
1881 struct task_struct
*task
;
1882 const struct cred
*cred
;
1883 struct inode
*inode
;
1886 if (flags
& LOOKUP_RCU
)
1889 inode
= d_inode(dentry
);
1890 task
= get_proc_task(inode
);
1894 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
1895 if (IS_ERR_OR_NULL(mm
))
1898 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
1899 down_read(&mm
->mmap_sem
);
1900 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
1901 up_read(&mm
->mmap_sem
);
1906 if (exact_vma_exists
) {
1907 if (task_dumpable(task
)) {
1909 cred
= __task_cred(task
);
1910 inode
->i_uid
= cred
->euid
;
1911 inode
->i_gid
= cred
->egid
;
1914 inode
->i_uid
= GLOBAL_ROOT_UID
;
1915 inode
->i_gid
= GLOBAL_ROOT_GID
;
1917 security_task_to_inode(task
, inode
);
1922 put_task_struct(task
);
1928 static const struct dentry_operations tid_map_files_dentry_operations
= {
1929 .d_revalidate
= map_files_d_revalidate
,
1930 .d_delete
= pid_delete_dentry
,
1933 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
1935 unsigned long vm_start
, vm_end
;
1936 struct vm_area_struct
*vma
;
1937 struct task_struct
*task
;
1938 struct mm_struct
*mm
;
1942 task
= get_proc_task(d_inode(dentry
));
1946 mm
= get_task_mm(task
);
1947 put_task_struct(task
);
1951 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
1956 down_read(&mm
->mmap_sem
);
1957 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
1958 if (vma
&& vma
->vm_file
) {
1959 *path
= vma
->vm_file
->f_path
;
1963 up_read(&mm
->mmap_sem
);
1971 struct map_files_info
{
1974 unsigned char name
[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1978 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1979 * symlinks may be used to bypass permissions on ancestor directories in the
1980 * path to the file in question.
1983 proc_map_files_get_link(struct dentry
*dentry
,
1984 struct inode
*inode
,
1985 struct delayed_call
*done
)
1987 if (!capable(CAP_SYS_ADMIN
))
1988 return ERR_PTR(-EPERM
);
1990 return proc_pid_get_link(dentry
, inode
, done
);
1994 * Identical to proc_pid_link_inode_operations except for get_link()
1996 static const struct inode_operations proc_map_files_link_inode_operations
= {
1997 .readlink
= proc_pid_readlink
,
1998 .get_link
= proc_map_files_get_link
,
1999 .setattr
= proc_setattr
,
2003 proc_map_files_instantiate(struct inode
*dir
, struct dentry
*dentry
,
2004 struct task_struct
*task
, const void *ptr
)
2006 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
2007 struct proc_inode
*ei
;
2008 struct inode
*inode
;
2010 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2015 ei
->op
.proc_get_link
= map_files_get_link
;
2017 inode
->i_op
= &proc_map_files_link_inode_operations
;
2019 inode
->i_mode
= S_IFLNK
;
2021 if (mode
& FMODE_READ
)
2022 inode
->i_mode
|= S_IRUSR
;
2023 if (mode
& FMODE_WRITE
)
2024 inode
->i_mode
|= S_IWUSR
;
2026 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2027 d_add(dentry
, inode
);
2032 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2033 struct dentry
*dentry
, unsigned int flags
)
2035 unsigned long vm_start
, vm_end
;
2036 struct vm_area_struct
*vma
;
2037 struct task_struct
*task
;
2039 struct mm_struct
*mm
;
2042 task
= get_proc_task(dir
);
2047 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2051 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2054 mm
= get_task_mm(task
);
2058 down_read(&mm
->mmap_sem
);
2059 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2064 result
= proc_map_files_instantiate(dir
, dentry
, task
,
2065 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2068 up_read(&mm
->mmap_sem
);
2071 put_task_struct(task
);
2073 return ERR_PTR(result
);
2076 static const struct inode_operations proc_map_files_inode_operations
= {
2077 .lookup
= proc_map_files_lookup
,
2078 .permission
= proc_fd_permission
,
2079 .setattr
= proc_setattr
,
2083 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2085 struct vm_area_struct
*vma
;
2086 struct task_struct
*task
;
2087 struct mm_struct
*mm
;
2088 unsigned long nr_files
, pos
, i
;
2089 struct flex_array
*fa
= NULL
;
2090 struct map_files_info info
;
2091 struct map_files_info
*p
;
2095 task
= get_proc_task(file_inode(file
));
2100 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2104 if (!dir_emit_dots(file
, ctx
))
2107 mm
= get_task_mm(task
);
2110 down_read(&mm
->mmap_sem
);
2115 * We need two passes here:
2117 * 1) Collect vmas of mapped files with mmap_sem taken
2118 * 2) Release mmap_sem and instantiate entries
2120 * otherwise we get lockdep complained, since filldir()
2121 * routine might require mmap_sem taken in might_fault().
2124 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2125 if (vma
->vm_file
&& ++pos
> ctx
->pos
)
2130 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2132 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2136 flex_array_free(fa
);
2137 up_read(&mm
->mmap_sem
);
2141 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2142 vma
= vma
->vm_next
) {
2145 if (++pos
<= ctx
->pos
)
2148 info
.mode
= vma
->vm_file
->f_mode
;
2149 info
.len
= snprintf(info
.name
,
2150 sizeof(info
.name
), "%lx-%lx",
2151 vma
->vm_start
, vma
->vm_end
);
2152 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2156 up_read(&mm
->mmap_sem
);
2158 for (i
= 0; i
< nr_files
; i
++) {
2159 p
= flex_array_get(fa
, i
);
2160 if (!proc_fill_cache(file
, ctx
,
2162 proc_map_files_instantiate
,
2164 (void *)(unsigned long)p
->mode
))
2169 flex_array_free(fa
);
2173 put_task_struct(task
);
2178 static const struct file_operations proc_map_files_operations
= {
2179 .read
= generic_read_dir
,
2180 .iterate_shared
= proc_map_files_readdir
,
2181 .llseek
= generic_file_llseek
,
2184 #ifdef CONFIG_CHECKPOINT_RESTORE
2185 struct timers_private
{
2187 struct task_struct
*task
;
2188 struct sighand_struct
*sighand
;
2189 struct pid_namespace
*ns
;
2190 unsigned long flags
;
2193 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2195 struct timers_private
*tp
= m
->private;
2197 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2199 return ERR_PTR(-ESRCH
);
2201 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2203 return ERR_PTR(-ESRCH
);
2205 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2208 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2210 struct timers_private
*tp
= m
->private;
2211 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2214 static void timers_stop(struct seq_file
*m
, void *v
)
2216 struct timers_private
*tp
= m
->private;
2219 unlock_task_sighand(tp
->task
, &tp
->flags
);
2224 put_task_struct(tp
->task
);
2229 static int show_timer(struct seq_file
*m
, void *v
)
2231 struct k_itimer
*timer
;
2232 struct timers_private
*tp
= m
->private;
2234 static const char * const nstr
[] = {
2235 [SIGEV_SIGNAL
] = "signal",
2236 [SIGEV_NONE
] = "none",
2237 [SIGEV_THREAD
] = "thread",
2240 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2241 notify
= timer
->it_sigev_notify
;
2243 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2244 seq_printf(m
, "signal: %d/%p\n",
2245 timer
->sigq
->info
.si_signo
,
2246 timer
->sigq
->info
.si_value
.sival_ptr
);
2247 seq_printf(m
, "notify: %s/%s.%d\n",
2248 nstr
[notify
& ~SIGEV_THREAD_ID
],
2249 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2250 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2251 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2256 static const struct seq_operations proc_timers_seq_ops
= {
2257 .start
= timers_start
,
2258 .next
= timers_next
,
2259 .stop
= timers_stop
,
2263 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2265 struct timers_private
*tp
;
2267 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2268 sizeof(struct timers_private
));
2272 tp
->pid
= proc_pid(inode
);
2273 tp
->ns
= inode
->i_sb
->s_fs_info
;
2277 static const struct file_operations proc_timers_operations
= {
2278 .open
= proc_timers_open
,
2280 .llseek
= seq_lseek
,
2281 .release
= seq_release_private
,
2285 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2286 size_t count
, loff_t
*offset
)
2288 struct inode
*inode
= file_inode(file
);
2289 struct task_struct
*p
;
2293 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2297 p
= get_proc_task(inode
);
2302 if (!capable(CAP_SYS_NICE
)) {
2307 err
= security_task_setscheduler(p
);
2316 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2318 p
->timer_slack_ns
= slack_ns
;
2327 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2329 struct inode
*inode
= m
->private;
2330 struct task_struct
*p
;
2333 p
= get_proc_task(inode
);
2339 if (!capable(CAP_SYS_NICE
)) {
2343 err
= security_task_getscheduler(p
);
2349 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2358 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2360 return single_open(filp
, timerslack_ns_show
, inode
);
2363 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2364 .open
= timerslack_ns_open
,
2366 .write
= timerslack_ns_write
,
2367 .llseek
= seq_lseek
,
2368 .release
= single_release
,
2371 static int proc_pident_instantiate(struct inode
*dir
,
2372 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2374 const struct pid_entry
*p
= ptr
;
2375 struct inode
*inode
;
2376 struct proc_inode
*ei
;
2378 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2383 inode
->i_mode
= p
->mode
;
2384 if (S_ISDIR(inode
->i_mode
))
2385 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2387 inode
->i_op
= p
->iop
;
2389 inode
->i_fop
= p
->fop
;
2391 d_set_d_op(dentry
, &pid_dentry_operations
);
2392 d_add(dentry
, inode
);
2393 /* Close the race of the process dying before we return the dentry */
2394 if (pid_revalidate(dentry
, 0))
2400 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2401 struct dentry
*dentry
,
2402 const struct pid_entry
*ents
,
2406 struct task_struct
*task
= get_proc_task(dir
);
2407 const struct pid_entry
*p
, *last
;
2415 * Yes, it does not scale. And it should not. Don't add
2416 * new entries into /proc/<tgid>/ without very good reasons.
2418 last
= &ents
[nents
];
2419 for (p
= ents
; p
< last
; p
++) {
2420 if (p
->len
!= dentry
->d_name
.len
)
2422 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2428 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2430 put_task_struct(task
);
2432 return ERR_PTR(error
);
2435 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2436 const struct pid_entry
*ents
, unsigned int nents
)
2438 struct task_struct
*task
= get_proc_task(file_inode(file
));
2439 const struct pid_entry
*p
;
2444 if (!dir_emit_dots(file
, ctx
))
2447 if (ctx
->pos
>= nents
+ 2)
2450 for (p
= ents
+ (ctx
->pos
- 2); p
< ents
+ nents
; p
++) {
2451 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2452 proc_pident_instantiate
, task
, p
))
2457 put_task_struct(task
);
2461 #ifdef CONFIG_SECURITY
2462 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2463 size_t count
, loff_t
*ppos
)
2465 struct inode
* inode
= file_inode(file
);
2468 struct task_struct
*task
= get_proc_task(inode
);
2473 length
= security_getprocattr(task
,
2474 (char*)file
->f_path
.dentry
->d_name
.name
,
2476 put_task_struct(task
);
2478 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2483 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2484 size_t count
, loff_t
*ppos
)
2486 struct inode
* inode
= file_inode(file
);
2489 struct task_struct
*task
= get_proc_task(inode
);
2494 if (count
> PAGE_SIZE
)
2497 /* No partial writes. */
2502 page
= memdup_user(buf
, count
);
2504 length
= PTR_ERR(page
);
2508 /* Guard against adverse ptrace interaction */
2509 length
= mutex_lock_interruptible(&task
->signal
->cred_guard_mutex
);
2513 length
= security_setprocattr(task
,
2514 (char*)file
->f_path
.dentry
->d_name
.name
,
2516 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2520 put_task_struct(task
);
2525 static const struct file_operations proc_pid_attr_operations
= {
2526 .read
= proc_pid_attr_read
,
2527 .write
= proc_pid_attr_write
,
2528 .llseek
= generic_file_llseek
,
2531 static const struct pid_entry attr_dir_stuff
[] = {
2532 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2533 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2534 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2535 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2536 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2537 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2540 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2542 return proc_pident_readdir(file
, ctx
,
2543 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2546 static const struct file_operations proc_attr_dir_operations
= {
2547 .read
= generic_read_dir
,
2548 .iterate_shared
= proc_attr_dir_readdir
,
2549 .llseek
= generic_file_llseek
,
2552 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2553 struct dentry
*dentry
, unsigned int flags
)
2555 return proc_pident_lookup(dir
, dentry
,
2556 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2559 static const struct inode_operations proc_attr_dir_inode_operations
= {
2560 .lookup
= proc_attr_dir_lookup
,
2561 .getattr
= pid_getattr
,
2562 .setattr
= proc_setattr
,
2567 #ifdef CONFIG_ELF_CORE
2568 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2569 size_t count
, loff_t
*ppos
)
2571 struct task_struct
*task
= get_proc_task(file_inode(file
));
2572 struct mm_struct
*mm
;
2573 char buffer
[PROC_NUMBUF
];
2581 mm
= get_task_mm(task
);
2583 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2584 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2585 MMF_DUMP_FILTER_SHIFT
));
2587 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2590 put_task_struct(task
);
2595 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2596 const char __user
*buf
,
2600 struct task_struct
*task
;
2601 struct mm_struct
*mm
;
2607 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2612 task
= get_proc_task(file_inode(file
));
2616 mm
= get_task_mm(task
);
2621 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2623 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2625 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2630 put_task_struct(task
);
2637 static const struct file_operations proc_coredump_filter_operations
= {
2638 .read
= proc_coredump_filter_read
,
2639 .write
= proc_coredump_filter_write
,
2640 .llseek
= generic_file_llseek
,
2644 #ifdef CONFIG_TASK_IO_ACCOUNTING
2645 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2647 struct task_io_accounting acct
= task
->ioac
;
2648 unsigned long flags
;
2651 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2655 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2660 if (whole
&& lock_task_sighand(task
, &flags
)) {
2661 struct task_struct
*t
= task
;
2663 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2664 while_each_thread(task
, t
)
2665 task_io_accounting_add(&acct
, &t
->ioac
);
2667 unlock_task_sighand(task
, &flags
);
2674 "read_bytes: %llu\n"
2675 "write_bytes: %llu\n"
2676 "cancelled_write_bytes: %llu\n",
2677 (unsigned long long)acct
.rchar
,
2678 (unsigned long long)acct
.wchar
,
2679 (unsigned long long)acct
.syscr
,
2680 (unsigned long long)acct
.syscw
,
2681 (unsigned long long)acct
.read_bytes
,
2682 (unsigned long long)acct
.write_bytes
,
2683 (unsigned long long)acct
.cancelled_write_bytes
);
2687 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2691 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2692 struct pid
*pid
, struct task_struct
*task
)
2694 return do_io_accounting(task
, m
, 0);
2697 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2698 struct pid
*pid
, struct task_struct
*task
)
2700 return do_io_accounting(task
, m
, 1);
2702 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2704 #ifdef CONFIG_USER_NS
2705 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2706 const struct seq_operations
*seq_ops
)
2708 struct user_namespace
*ns
= NULL
;
2709 struct task_struct
*task
;
2710 struct seq_file
*seq
;
2713 task
= get_proc_task(inode
);
2716 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2718 put_task_struct(task
);
2723 ret
= seq_open(file
, seq_ops
);
2727 seq
= file
->private_data
;
2737 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2739 struct seq_file
*seq
= file
->private_data
;
2740 struct user_namespace
*ns
= seq
->private;
2742 return seq_release(inode
, file
);
2745 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2747 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2750 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2752 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2755 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
2757 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
2760 static const struct file_operations proc_uid_map_operations
= {
2761 .open
= proc_uid_map_open
,
2762 .write
= proc_uid_map_write
,
2764 .llseek
= seq_lseek
,
2765 .release
= proc_id_map_release
,
2768 static const struct file_operations proc_gid_map_operations
= {
2769 .open
= proc_gid_map_open
,
2770 .write
= proc_gid_map_write
,
2772 .llseek
= seq_lseek
,
2773 .release
= proc_id_map_release
,
2776 static const struct file_operations proc_projid_map_operations
= {
2777 .open
= proc_projid_map_open
,
2778 .write
= proc_projid_map_write
,
2780 .llseek
= seq_lseek
,
2781 .release
= proc_id_map_release
,
2784 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
2786 struct user_namespace
*ns
= NULL
;
2787 struct task_struct
*task
;
2791 task
= get_proc_task(inode
);
2794 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2796 put_task_struct(task
);
2801 if (file
->f_mode
& FMODE_WRITE
) {
2803 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
2807 ret
= single_open(file
, &proc_setgroups_show
, ns
);
2818 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
2820 struct seq_file
*seq
= file
->private_data
;
2821 struct user_namespace
*ns
= seq
->private;
2822 int ret
= single_release(inode
, file
);
2827 static const struct file_operations proc_setgroups_operations
= {
2828 .open
= proc_setgroups_open
,
2829 .write
= proc_setgroups_write
,
2831 .llseek
= seq_lseek
,
2832 .release
= proc_setgroups_release
,
2834 #endif /* CONFIG_USER_NS */
2836 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2837 struct pid
*pid
, struct task_struct
*task
)
2839 int err
= lock_trace(task
);
2841 seq_printf(m
, "%08x\n", task
->personality
);
2850 static const struct file_operations proc_task_operations
;
2851 static const struct inode_operations proc_task_inode_operations
;
2853 static const struct pid_entry tgid_base_stuff
[] = {
2854 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2855 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2856 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
2857 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2858 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
2860 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2862 REG("environ", S_IRUSR
, proc_environ_operations
),
2863 REG("auxv", S_IRUSR
, proc_auxv_operations
),
2864 ONE("status", S_IRUGO
, proc_pid_status
),
2865 ONE("personality", S_IRUSR
, proc_pid_personality
),
2866 ONE("limits", S_IRUGO
, proc_pid_limits
),
2867 #ifdef CONFIG_SCHED_DEBUG
2868 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2870 #ifdef CONFIG_SCHED_AUTOGROUP
2871 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
2873 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2874 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2875 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
2877 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
2878 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2879 ONE("statm", S_IRUGO
, proc_pid_statm
),
2880 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
2882 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
2884 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2885 LNK("cwd", proc_cwd_link
),
2886 LNK("root", proc_root_link
),
2887 LNK("exe", proc_exe_link
),
2888 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2889 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2890 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2891 #ifdef CONFIG_PROC_PAGE_MONITOR
2892 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2893 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
2894 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2896 #ifdef CONFIG_SECURITY
2897 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2899 #ifdef CONFIG_KALLSYMS
2900 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
2902 #ifdef CONFIG_STACKTRACE
2903 ONE("stack", S_IRUSR
, proc_pid_stack
),
2905 #ifdef CONFIG_SCHED_INFO
2906 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
2908 #ifdef CONFIG_LATENCYTOP
2909 REG("latency", S_IRUGO
, proc_lstats_operations
),
2911 #ifdef CONFIG_PROC_PID_CPUSET
2912 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
2914 #ifdef CONFIG_CGROUPS
2915 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
2917 ONE("oom_score", S_IRUGO
, proc_oom_score
),
2918 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
2919 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
2920 #ifdef CONFIG_AUDITSYSCALL
2921 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2922 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2924 #ifdef CONFIG_FAULT_INJECTION
2925 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2927 #ifdef CONFIG_ELF_CORE
2928 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2930 #ifdef CONFIG_TASK_IO_ACCOUNTING
2931 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
2933 #ifdef CONFIG_HARDWALL
2934 ONE("hardwall", S_IRUGO
, proc_pid_hardwall
),
2936 #ifdef CONFIG_USER_NS
2937 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
2938 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
2939 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
2940 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
2942 #ifdef CONFIG_CHECKPOINT_RESTORE
2943 REG("timers", S_IRUGO
, proc_timers_operations
),
2945 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
2948 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
2950 return proc_pident_readdir(file
, ctx
,
2951 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2954 static const struct file_operations proc_tgid_base_operations
= {
2955 .read
= generic_read_dir
,
2956 .iterate_shared
= proc_tgid_base_readdir
,
2957 .llseek
= generic_file_llseek
,
2960 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
2962 return proc_pident_lookup(dir
, dentry
,
2963 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2966 static const struct inode_operations proc_tgid_base_inode_operations
= {
2967 .lookup
= proc_tgid_base_lookup
,
2968 .getattr
= pid_getattr
,
2969 .setattr
= proc_setattr
,
2970 .permission
= proc_pid_permission
,
2973 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2975 struct dentry
*dentry
, *leader
, *dir
;
2976 char buf
[PROC_NUMBUF
];
2980 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2981 /* no ->d_hash() rejects on procfs */
2982 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2984 d_invalidate(dentry
);
2992 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2993 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2998 name
.len
= strlen(name
.name
);
2999 dir
= d_hash_and_lookup(leader
, &name
);
3001 goto out_put_leader
;
3004 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3005 dentry
= d_hash_and_lookup(dir
, &name
);
3007 d_invalidate(dentry
);
3019 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3020 * @task: task that should be flushed.
3022 * When flushing dentries from proc, one needs to flush them from global
3023 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3024 * in. This call is supposed to do all of this job.
3026 * Looks in the dcache for
3028 * /proc/@tgid/task/@pid
3029 * if either directory is present flushes it and all of it'ts children
3032 * It is safe and reasonable to cache /proc entries for a task until
3033 * that task exits. After that they just clog up the dcache with
3034 * useless entries, possibly causing useful dcache entries to be
3035 * flushed instead. This routine is proved to flush those useless
3036 * dcache entries at process exit time.
3038 * NOTE: This routine is just an optimization so it does not guarantee
3039 * that no dcache entries will exist at process exit time it
3040 * just makes it very unlikely that any will persist.
3043 void proc_flush_task(struct task_struct
*task
)
3046 struct pid
*pid
, *tgid
;
3049 pid
= task_pid(task
);
3050 tgid
= task_tgid(task
);
3052 for (i
= 0; i
<= pid
->level
; i
++) {
3053 upid
= &pid
->numbers
[i
];
3054 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
3055 tgid
->numbers
[i
].nr
);
3059 static int proc_pid_instantiate(struct inode
*dir
,
3060 struct dentry
* dentry
,
3061 struct task_struct
*task
, const void *ptr
)
3063 struct inode
*inode
;
3065 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3069 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3070 inode
->i_op
= &proc_tgid_base_inode_operations
;
3071 inode
->i_fop
= &proc_tgid_base_operations
;
3072 inode
->i_flags
|=S_IMMUTABLE
;
3074 set_nlink(inode
, nlink_tgid
);
3076 d_set_d_op(dentry
, &pid_dentry_operations
);
3078 d_add(dentry
, inode
);
3079 /* Close the race of the process dying before we return the dentry */
3080 if (pid_revalidate(dentry
, 0))
3086 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3088 int result
= -ENOENT
;
3089 struct task_struct
*task
;
3091 struct pid_namespace
*ns
;
3093 tgid
= name_to_int(&dentry
->d_name
);
3097 ns
= dentry
->d_sb
->s_fs_info
;
3099 task
= find_task_by_pid_ns(tgid
, ns
);
3101 get_task_struct(task
);
3106 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
3107 put_task_struct(task
);
3109 return ERR_PTR(result
);
3113 * Find the first task with tgid >= tgid
3118 struct task_struct
*task
;
3120 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3125 put_task_struct(iter
.task
);
3129 pid
= find_ge_pid(iter
.tgid
, ns
);
3131 iter
.tgid
= pid_nr_ns(pid
, ns
);
3132 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3133 /* What we to know is if the pid we have find is the
3134 * pid of a thread_group_leader. Testing for task
3135 * being a thread_group_leader is the obvious thing
3136 * todo but there is a window when it fails, due to
3137 * the pid transfer logic in de_thread.
3139 * So we perform the straight forward test of seeing
3140 * if the pid we have found is the pid of a thread
3141 * group leader, and don't worry if the task we have
3142 * found doesn't happen to be a thread group leader.
3143 * As we don't care in the case of readdir.
3145 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3149 get_task_struct(iter
.task
);
3155 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3157 /* for the /proc/ directory itself, after non-process stuff has been done */
3158 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3160 struct tgid_iter iter
;
3161 struct pid_namespace
*ns
= file_inode(file
)->i_sb
->s_fs_info
;
3162 loff_t pos
= ctx
->pos
;
3164 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3167 if (pos
== TGID_OFFSET
- 2) {
3168 struct inode
*inode
= d_inode(ns
->proc_self
);
3169 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3171 ctx
->pos
= pos
= pos
+ 1;
3173 if (pos
== TGID_OFFSET
- 1) {
3174 struct inode
*inode
= d_inode(ns
->proc_thread_self
);
3175 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3177 ctx
->pos
= pos
= pos
+ 1;
3179 iter
.tgid
= pos
- TGID_OFFSET
;
3181 for (iter
= next_tgid(ns
, iter
);
3183 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3184 char name
[PROC_NUMBUF
];
3186 if (!has_pid_permissions(ns
, iter
.task
, 2))
3189 len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
3190 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3191 if (!proc_fill_cache(file
, ctx
, name
, len
,
3192 proc_pid_instantiate
, iter
.task
, NULL
)) {
3193 put_task_struct(iter
.task
);
3197 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3202 * proc_tid_comm_permission is a special permission function exclusively
3203 * used for the node /proc/<pid>/task/<tid>/comm.
3204 * It bypasses generic permission checks in the case where a task of the same
3205 * task group attempts to access the node.
3206 * The rationale behind this is that glibc and bionic access this node for
3207 * cross thread naming (pthread_set/getname_np(!self)). However, if
3208 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3209 * which locks out the cross thread naming implementation.
3210 * This function makes sure that the node is always accessible for members of
3211 * same thread group.
3213 static int proc_tid_comm_permission(struct inode
*inode
, int mask
)
3215 bool is_same_tgroup
;
3216 struct task_struct
*task
;
3218 task
= get_proc_task(inode
);
3221 is_same_tgroup
= same_thread_group(current
, task
);
3222 put_task_struct(task
);
3224 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3225 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3226 * read or written by the members of the corresponding
3232 return generic_permission(inode
, mask
);
3235 static const struct inode_operations proc_tid_comm_inode_operations
= {
3236 .permission
= proc_tid_comm_permission
,
3242 static const struct pid_entry tid_base_stuff
[] = {
3243 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3244 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3245 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3247 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3249 REG("environ", S_IRUSR
, proc_environ_operations
),
3250 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3251 ONE("status", S_IRUGO
, proc_pid_status
),
3252 ONE("personality", S_IRUSR
, proc_pid_personality
),
3253 ONE("limits", S_IRUGO
, proc_pid_limits
),
3254 #ifdef CONFIG_SCHED_DEBUG
3255 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3257 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3258 &proc_tid_comm_inode_operations
,
3259 &proc_pid_set_comm_operations
, {}),
3260 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3261 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3263 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3264 ONE("stat", S_IRUGO
, proc_tid_stat
),
3265 ONE("statm", S_IRUGO
, proc_pid_statm
),
3266 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3267 #ifdef CONFIG_PROC_CHILDREN
3268 REG("children", S_IRUGO
, proc_tid_children_operations
),
3271 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3273 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3274 LNK("cwd", proc_cwd_link
),
3275 LNK("root", proc_root_link
),
3276 LNK("exe", proc_exe_link
),
3277 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3278 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3279 #ifdef CONFIG_PROC_PAGE_MONITOR
3280 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3281 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3282 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3284 #ifdef CONFIG_SECURITY
3285 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3287 #ifdef CONFIG_KALLSYMS
3288 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3290 #ifdef CONFIG_STACKTRACE
3291 ONE("stack", S_IRUSR
, proc_pid_stack
),
3293 #ifdef CONFIG_SCHED_INFO
3294 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3296 #ifdef CONFIG_LATENCYTOP
3297 REG("latency", S_IRUGO
, proc_lstats_operations
),
3299 #ifdef CONFIG_PROC_PID_CPUSET
3300 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3302 #ifdef CONFIG_CGROUPS
3303 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3305 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3306 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3307 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3308 #ifdef CONFIG_AUDITSYSCALL
3309 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3310 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3312 #ifdef CONFIG_FAULT_INJECTION
3313 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3315 #ifdef CONFIG_TASK_IO_ACCOUNTING
3316 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3318 #ifdef CONFIG_HARDWALL
3319 ONE("hardwall", S_IRUGO
, proc_pid_hardwall
),
3321 #ifdef CONFIG_USER_NS
3322 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3323 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3324 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3325 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3329 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3331 return proc_pident_readdir(file
, ctx
,
3332 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3335 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3337 return proc_pident_lookup(dir
, dentry
,
3338 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3341 static const struct file_operations proc_tid_base_operations
= {
3342 .read
= generic_read_dir
,
3343 .iterate_shared
= proc_tid_base_readdir
,
3344 .llseek
= generic_file_llseek
,
3347 static const struct inode_operations proc_tid_base_inode_operations
= {
3348 .lookup
= proc_tid_base_lookup
,
3349 .getattr
= pid_getattr
,
3350 .setattr
= proc_setattr
,
3353 static int proc_task_instantiate(struct inode
*dir
,
3354 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3356 struct inode
*inode
;
3357 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3361 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3362 inode
->i_op
= &proc_tid_base_inode_operations
;
3363 inode
->i_fop
= &proc_tid_base_operations
;
3364 inode
->i_flags
|=S_IMMUTABLE
;
3366 set_nlink(inode
, nlink_tid
);
3368 d_set_d_op(dentry
, &pid_dentry_operations
);
3370 d_add(dentry
, inode
);
3371 /* Close the race of the process dying before we return the dentry */
3372 if (pid_revalidate(dentry
, 0))
3378 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3380 int result
= -ENOENT
;
3381 struct task_struct
*task
;
3382 struct task_struct
*leader
= get_proc_task(dir
);
3384 struct pid_namespace
*ns
;
3389 tid
= name_to_int(&dentry
->d_name
);
3393 ns
= dentry
->d_sb
->s_fs_info
;
3395 task
= find_task_by_pid_ns(tid
, ns
);
3397 get_task_struct(task
);
3401 if (!same_thread_group(leader
, task
))
3404 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3406 put_task_struct(task
);
3408 put_task_struct(leader
);
3410 return ERR_PTR(result
);
3414 * Find the first tid of a thread group to return to user space.
3416 * Usually this is just the thread group leader, but if the users
3417 * buffer was too small or there was a seek into the middle of the
3418 * directory we have more work todo.
3420 * In the case of a short read we start with find_task_by_pid.
3422 * In the case of a seek we start with the leader and walk nr
3425 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3426 struct pid_namespace
*ns
)
3428 struct task_struct
*pos
, *task
;
3429 unsigned long nr
= f_pos
;
3431 if (nr
!= f_pos
) /* 32bit overflow? */
3435 task
= pid_task(pid
, PIDTYPE_PID
);
3439 /* Attempt to start with the tid of a thread */
3441 pos
= find_task_by_pid_ns(tid
, ns
);
3442 if (pos
&& same_thread_group(pos
, task
))
3446 /* If nr exceeds the number of threads there is nothing todo */
3447 if (nr
>= get_nr_threads(task
))
3450 /* If we haven't found our starting place yet start
3451 * with the leader and walk nr threads forward.
3453 pos
= task
= task
->group_leader
;
3457 } while_each_thread(task
, pos
);
3462 get_task_struct(pos
);
3469 * Find the next thread in the thread list.
3470 * Return NULL if there is an error or no next thread.
3472 * The reference to the input task_struct is released.
3474 static struct task_struct
*next_tid(struct task_struct
*start
)
3476 struct task_struct
*pos
= NULL
;
3478 if (pid_alive(start
)) {
3479 pos
= next_thread(start
);
3480 if (thread_group_leader(pos
))
3483 get_task_struct(pos
);
3486 put_task_struct(start
);
3490 /* for the /proc/TGID/task/ directories */
3491 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3493 struct inode
*inode
= file_inode(file
);
3494 struct task_struct
*task
;
3495 struct pid_namespace
*ns
;
3498 if (proc_inode_is_dead(inode
))
3501 if (!dir_emit_dots(file
, ctx
))
3504 /* f_version caches the tgid value that the last readdir call couldn't
3505 * return. lseek aka telldir automagically resets f_version to 0.
3507 ns
= inode
->i_sb
->s_fs_info
;
3508 tid
= (int)file
->f_version
;
3509 file
->f_version
= 0;
3510 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3512 task
= next_tid(task
), ctx
->pos
++) {
3513 char name
[PROC_NUMBUF
];
3515 tid
= task_pid_nr_ns(task
, ns
);
3516 len
= snprintf(name
, sizeof(name
), "%d", tid
);
3517 if (!proc_fill_cache(file
, ctx
, name
, len
,
3518 proc_task_instantiate
, task
, NULL
)) {
3519 /* returning this tgid failed, save it as the first
3520 * pid for the next readir call */
3521 file
->f_version
= (u64
)tid
;
3522 put_task_struct(task
);
3530 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3532 struct inode
*inode
= d_inode(dentry
);
3533 struct task_struct
*p
= get_proc_task(inode
);
3534 generic_fillattr(inode
, stat
);
3537 stat
->nlink
+= get_nr_threads(p
);
3544 static const struct inode_operations proc_task_inode_operations
= {
3545 .lookup
= proc_task_lookup
,
3546 .getattr
= proc_task_getattr
,
3547 .setattr
= proc_setattr
,
3548 .permission
= proc_pid_permission
,
3551 static const struct file_operations proc_task_operations
= {
3552 .read
= generic_read_dir
,
3553 .iterate_shared
= proc_task_readdir
,
3554 .llseek
= generic_file_llseek
,
3557 void __init
set_proc_pid_nlink(void)
3559 nlink_tid
= pid_entry_nlink(tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3560 nlink_tgid
= pid_entry_nlink(tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));