1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * proc base directory handling functions
9 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
10 * Instead of using magical inumbers to determine the kind of object
11 * we allocate and fill in-core inodes upon lookup. They don't even
12 * go into icache. We cache the reference to task_struct upon lookup too.
13 * Eventually it should become a filesystem in its own. We don't use the
14 * rest of procfs anymore.
20 * Bruna Moreira <bruna.moreira@indt.org.br>
21 * Edjard Mota <edjard.mota@indt.org.br>
22 * Ilias Biris <ilias.biris@indt.org.br>
23 * Mauricio Lin <mauricio.lin@indt.org.br>
25 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
27 * A new process specific entry (smaps) included in /proc. It shows the
28 * size of rss for each memory area. The maps entry lacks information
29 * about physical memory size (rss) for each mapped file, i.e.,
30 * rss information for executables and library files.
31 * This additional information is useful for any tools that need to know
32 * about physical memory consumption for a process specific library.
36 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
37 * Pud inclusion in the page table walking.
41 * 10LE Instituto Nokia de Tecnologia - INdT:
42 * A better way to walks through the page table as suggested by Hugh Dickins.
44 * Simo Piiroinen <simo.piiroinen@nokia.com>:
45 * Smaps information related to shared, private, clean and dirty pages.
47 * Paul Mundt <paul.mundt@nokia.com>:
48 * Overall revision about smaps.
51 #include <linux/uaccess.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/task_io_accounting_ops.h>
58 #include <linux/init.h>
59 #include <linux/capability.h>
60 #include <linux/file.h>
61 #include <linux/fdtable.h>
62 #include <linux/string.h>
63 #include <linux/seq_file.h>
64 #include <linux/namei.h>
65 #include <linux/mnt_namespace.h>
67 #include <linux/swap.h>
68 #include <linux/rcupdate.h>
69 #include <linux/kallsyms.h>
70 #include <linux/stacktrace.h>
71 #include <linux/resource.h>
72 #include <linux/module.h>
73 #include <linux/mount.h>
74 #include <linux/security.h>
75 #include <linux/ptrace.h>
76 #include <linux/tracehook.h>
77 #include <linux/printk.h>
78 #include <linux/cache.h>
79 #include <linux/cgroup.h>
80 #include <linux/cpuset.h>
81 #include <linux/audit.h>
82 #include <linux/poll.h>
83 #include <linux/nsproxy.h>
84 #include <linux/oom.h>
85 #include <linux/elf.h>
86 #include <linux/pid_namespace.h>
87 #include <linux/user_namespace.h>
88 #include <linux/fs_struct.h>
89 #include <linux/slab.h>
90 #include <linux/sched/autogroup.h>
91 #include <linux/sched/mm.h>
92 #include <linux/sched/coredump.h>
93 #include <linux/sched/debug.h>
94 #include <linux/sched/stat.h>
95 #include <linux/flex_array.h>
96 #include <linux/posix-timers.h>
97 #include <trace/events/oom.h>
101 #include "../../lib/kstrtox.h"
104 * Implementing inode permission operations in /proc is almost
105 * certainly an error. Permission checks need to happen during
106 * each system call not at open time. The reason is that most of
107 * what we wish to check for permissions in /proc varies at runtime.
109 * The classic example of a problem is opening file descriptors
110 * in /proc for a task before it execs a suid executable.
113 static u8 nlink_tid __ro_after_init
;
114 static u8 nlink_tgid __ro_after_init
;
120 const struct inode_operations
*iop
;
121 const struct file_operations
*fop
;
125 #define NOD(NAME, MODE, IOP, FOP, OP) { \
127 .len = sizeof(NAME) - 1, \
134 #define DIR(NAME, MODE, iops, fops) \
135 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
136 #define LNK(NAME, get_link) \
137 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
138 &proc_pid_link_inode_operations, NULL, \
139 { .proc_get_link = get_link } )
140 #define REG(NAME, MODE, fops) \
141 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
142 #define ONE(NAME, MODE, show) \
143 NOD(NAME, (S_IFREG|(MODE)), \
144 NULL, &proc_single_file_operations, \
145 { .proc_show = show } )
148 * Count the number of hardlinks for the pid_entry table, excluding the .
151 static unsigned int __init
pid_entry_nlink(const struct pid_entry
*entries
,
158 for (i
= 0; i
< n
; ++i
) {
159 if (S_ISDIR(entries
[i
].mode
))
166 static int get_task_root(struct task_struct
*task
, struct path
*root
)
168 int result
= -ENOENT
;
172 get_fs_root(task
->fs
, root
);
179 static int proc_cwd_link(struct dentry
*dentry
, struct path
*path
)
181 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
182 int result
= -ENOENT
;
187 get_fs_pwd(task
->fs
, path
);
191 put_task_struct(task
);
196 static int proc_root_link(struct dentry
*dentry
, struct path
*path
)
198 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
199 int result
= -ENOENT
;
202 result
= get_task_root(task
, path
);
203 put_task_struct(task
);
208 static ssize_t
proc_pid_cmdline_read(struct file
*file
, char __user
*buf
,
209 size_t _count
, loff_t
*pos
)
211 struct task_struct
*tsk
;
212 struct mm_struct
*mm
;
214 unsigned long count
= _count
;
215 unsigned long arg_start
, arg_end
, env_start
, env_end
;
216 unsigned long len1
, len2
, len
;
223 tsk
= get_proc_task(file_inode(file
));
226 mm
= get_task_mm(tsk
);
227 put_task_struct(tsk
);
230 /* Check if process spawned far enough to have cmdline. */
236 page
= (char *)__get_free_page(GFP_KERNEL
);
242 down_read(&mm
->mmap_sem
);
243 arg_start
= mm
->arg_start
;
244 arg_end
= mm
->arg_end
;
245 env_start
= mm
->env_start
;
246 env_end
= mm
->env_end
;
247 up_read(&mm
->mmap_sem
);
249 BUG_ON(arg_start
> arg_end
);
250 BUG_ON(env_start
> env_end
);
252 len1
= arg_end
- arg_start
;
253 len2
= env_end
- env_start
;
261 * Inherently racy -- command line shares address space
262 * with code and data.
264 rv
= access_remote_vm(mm
, arg_end
- 1, &c
, 1, FOLL_ANON
);
271 /* Command line (set of strings) occupies whole ARGV. */
275 p
= arg_start
+ *pos
;
277 while (count
> 0 && len
> 0) {
281 _count
= min3(count
, len
, PAGE_SIZE
);
282 nr_read
= access_remote_vm(mm
, p
, page
, _count
, FOLL_ANON
);
288 if (copy_to_user(buf
, page
, nr_read
)) {
301 * Command line (1 string) occupies ARGV and
308 { .p
= arg_start
, .len
= len1
},
309 { .p
= env_start
, .len
= len2
},
315 while (i
< 2 && pos1
>= cmdline
[i
].len
) {
316 pos1
-= cmdline
[i
].len
;
320 p
= cmdline
[i
].p
+ pos1
;
321 len
= cmdline
[i
].len
- pos1
;
322 while (count
> 0 && len
> 0) {
323 unsigned int _count
, l
;
327 _count
= min3(count
, len
, PAGE_SIZE
);
328 nr_read
= access_remote_vm(mm
, p
, page
, _count
, FOLL_ANON
);
335 * Command line can be shorter than whole ARGV
336 * even if last "marker" byte says it is not.
339 l
= strnlen(page
, nr_read
);
345 if (copy_to_user(buf
, page
, nr_read
)) {
360 /* Only first chunk can be read partially. */
367 free_page((unsigned long)page
);
375 static const struct file_operations proc_pid_cmdline_ops
= {
376 .read
= proc_pid_cmdline_read
,
377 .llseek
= generic_file_llseek
,
380 #ifdef CONFIG_KALLSYMS
382 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
383 * Returns the resolved symbol. If that fails, simply return the address.
385 static int proc_pid_wchan(struct seq_file
*m
, struct pid_namespace
*ns
,
386 struct pid
*pid
, struct task_struct
*task
)
389 char symname
[KSYM_NAME_LEN
];
391 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
394 wchan
= get_wchan(task
);
395 if (wchan
&& !lookup_symbol_name(wchan
, symname
)) {
396 seq_puts(m
, symname
);
404 #endif /* CONFIG_KALLSYMS */
406 static int lock_trace(struct task_struct
*task
)
408 int err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
411 if (!ptrace_may_access(task
, PTRACE_MODE_ATTACH_FSCREDS
)) {
412 mutex_unlock(&task
->signal
->cred_guard_mutex
);
418 static void unlock_trace(struct task_struct
*task
)
420 mutex_unlock(&task
->signal
->cred_guard_mutex
);
423 #ifdef CONFIG_STACKTRACE
425 #define MAX_STACK_TRACE_DEPTH 64
427 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
428 struct pid
*pid
, struct task_struct
*task
)
430 struct stack_trace trace
;
431 unsigned long *entries
;
435 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
439 trace
.nr_entries
= 0;
440 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
441 trace
.entries
= entries
;
444 err
= lock_trace(task
);
446 save_stack_trace_tsk(task
, &trace
);
448 for (i
= 0; i
< trace
.nr_entries
; i
++) {
449 seq_printf(m
, "[<0>] %pB\n", (void *)entries
[i
]);
459 #ifdef CONFIG_SCHED_INFO
461 * Provides /proc/PID/schedstat
463 static int proc_pid_schedstat(struct seq_file
*m
, struct pid_namespace
*ns
,
464 struct pid
*pid
, struct task_struct
*task
)
466 if (unlikely(!sched_info_on()))
467 seq_printf(m
, "0 0 0\n");
469 seq_printf(m
, "%llu %llu %lu\n",
470 (unsigned long long)task
->se
.sum_exec_runtime
,
471 (unsigned long long)task
->sched_info
.run_delay
,
472 task
->sched_info
.pcount
);
478 #ifdef CONFIG_LATENCYTOP
479 static int lstats_show_proc(struct seq_file
*m
, void *v
)
482 struct inode
*inode
= m
->private;
483 struct task_struct
*task
= get_proc_task(inode
);
487 seq_puts(m
, "Latency Top version : v0.1\n");
488 for (i
= 0; i
< 32; i
++) {
489 struct latency_record
*lr
= &task
->latency_record
[i
];
490 if (lr
->backtrace
[0]) {
492 seq_printf(m
, "%i %li %li",
493 lr
->count
, lr
->time
, lr
->max
);
494 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
495 unsigned long bt
= lr
->backtrace
[q
];
500 seq_printf(m
, " %ps", (void *)bt
);
506 put_task_struct(task
);
510 static int lstats_open(struct inode
*inode
, struct file
*file
)
512 return single_open(file
, lstats_show_proc
, inode
);
515 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
516 size_t count
, loff_t
*offs
)
518 struct task_struct
*task
= get_proc_task(file_inode(file
));
522 clear_all_latency_tracing(task
);
523 put_task_struct(task
);
528 static const struct file_operations proc_lstats_operations
= {
531 .write
= lstats_write
,
533 .release
= single_release
,
538 static int proc_oom_score(struct seq_file
*m
, struct pid_namespace
*ns
,
539 struct pid
*pid
, struct task_struct
*task
)
541 unsigned long totalpages
= totalram_pages
+ total_swap_pages
;
542 unsigned long points
= 0;
544 points
= oom_badness(task
, NULL
, NULL
, totalpages
) *
546 seq_printf(m
, "%lu\n", points
);
556 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
557 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
558 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
559 [RLIMIT_DATA
] = {"Max data size", "bytes"},
560 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
561 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
562 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
563 [RLIMIT_NPROC
] = {"Max processes", "processes"},
564 [RLIMIT_NOFILE
] = {"Max open files", "files"},
565 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
566 [RLIMIT_AS
] = {"Max address space", "bytes"},
567 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
568 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
569 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
570 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
571 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
572 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
575 /* Display limits for a process */
576 static int proc_pid_limits(struct seq_file
*m
, struct pid_namespace
*ns
,
577 struct pid
*pid
, struct task_struct
*task
)
582 struct rlimit rlim
[RLIM_NLIMITS
];
584 if (!lock_task_sighand(task
, &flags
))
586 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
587 unlock_task_sighand(task
, &flags
);
590 * print the file header
592 seq_printf(m
, "%-25s %-20s %-20s %-10s\n",
593 "Limit", "Soft Limit", "Hard Limit", "Units");
595 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
596 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
597 seq_printf(m
, "%-25s %-20s ",
598 lnames
[i
].name
, "unlimited");
600 seq_printf(m
, "%-25s %-20lu ",
601 lnames
[i
].name
, rlim
[i
].rlim_cur
);
603 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
604 seq_printf(m
, "%-20s ", "unlimited");
606 seq_printf(m
, "%-20lu ", rlim
[i
].rlim_max
);
609 seq_printf(m
, "%-10s\n", lnames
[i
].unit
);
617 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
618 static int proc_pid_syscall(struct seq_file
*m
, struct pid_namespace
*ns
,
619 struct pid
*pid
, struct task_struct
*task
)
622 unsigned long args
[6], sp
, pc
;
625 res
= lock_trace(task
);
629 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
630 seq_puts(m
, "running\n");
632 seq_printf(m
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
635 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
637 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
643 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
645 /************************************************************************/
646 /* Here the fs part begins */
647 /************************************************************************/
649 /* permission checks */
650 static int proc_fd_access_allowed(struct inode
*inode
)
652 struct task_struct
*task
;
654 /* Allow access to a task's file descriptors if it is us or we
655 * may use ptrace attach to the process and find out that
658 task
= get_proc_task(inode
);
660 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
661 put_task_struct(task
);
666 int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
669 struct inode
*inode
= d_inode(dentry
);
671 if (attr
->ia_valid
& ATTR_MODE
)
674 error
= setattr_prepare(dentry
, attr
);
678 setattr_copy(inode
, attr
);
679 mark_inode_dirty(inode
);
684 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
685 * or euid/egid (for hide_pid_min=2)?
687 static bool has_pid_permissions(struct pid_namespace
*pid
,
688 struct task_struct
*task
,
691 if (pid
->hide_pid
< hide_pid_min
)
693 if (in_group_p(pid
->pid_gid
))
695 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
699 static int proc_pid_permission(struct inode
*inode
, int mask
)
701 struct pid_namespace
*pid
= proc_pid_ns(inode
);
702 struct task_struct
*task
;
705 task
= get_proc_task(inode
);
708 has_perms
= has_pid_permissions(pid
, task
, HIDEPID_NO_ACCESS
);
709 put_task_struct(task
);
712 if (pid
->hide_pid
== HIDEPID_INVISIBLE
) {
714 * Let's make getdents(), stat(), and open()
715 * consistent with each other. If a process
716 * may not stat() a file, it shouldn't be seen
724 return generic_permission(inode
, mask
);
729 static const struct inode_operations proc_def_inode_operations
= {
730 .setattr
= proc_setattr
,
733 static int proc_single_show(struct seq_file
*m
, void *v
)
735 struct inode
*inode
= m
->private;
736 struct pid_namespace
*ns
= proc_pid_ns(inode
);
737 struct pid
*pid
= proc_pid(inode
);
738 struct task_struct
*task
;
741 task
= get_pid_task(pid
, PIDTYPE_PID
);
745 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
747 put_task_struct(task
);
751 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
753 return single_open(filp
, proc_single_show
, inode
);
756 static const struct file_operations proc_single_file_operations
= {
757 .open
= proc_single_open
,
760 .release
= single_release
,
764 struct mm_struct
*proc_mem_open(struct inode
*inode
, unsigned int mode
)
766 struct task_struct
*task
= get_proc_task(inode
);
767 struct mm_struct
*mm
= ERR_PTR(-ESRCH
);
770 mm
= mm_access(task
, mode
| PTRACE_MODE_FSCREDS
);
771 put_task_struct(task
);
773 if (!IS_ERR_OR_NULL(mm
)) {
774 /* ensure this mm_struct can't be freed */
776 /* but do not pin its memory */
784 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
786 struct mm_struct
*mm
= proc_mem_open(inode
, mode
);
791 file
->private_data
= mm
;
795 static int mem_open(struct inode
*inode
, struct file
*file
)
797 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
799 /* OK to pass negative loff_t, we can catch out-of-range */
800 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
805 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
806 size_t count
, loff_t
*ppos
, int write
)
808 struct mm_struct
*mm
= file
->private_data
;
809 unsigned long addr
= *ppos
;
817 page
= (char *)__get_free_page(GFP_KERNEL
);
822 if (!mmget_not_zero(mm
))
825 flags
= FOLL_FORCE
| (write
? FOLL_WRITE
: 0);
828 int this_len
= min_t(int, count
, PAGE_SIZE
);
830 if (write
&& copy_from_user(page
, buf
, this_len
)) {
835 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, flags
);
842 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
856 free_page((unsigned long) page
);
860 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
861 size_t count
, loff_t
*ppos
)
863 return mem_rw(file
, buf
, count
, ppos
, 0);
866 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
867 size_t count
, loff_t
*ppos
)
869 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
872 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
876 file
->f_pos
= offset
;
879 file
->f_pos
+= offset
;
884 force_successful_syscall_return();
888 static int mem_release(struct inode
*inode
, struct file
*file
)
890 struct mm_struct
*mm
= file
->private_data
;
896 static const struct file_operations proc_mem_operations
= {
901 .release
= mem_release
,
904 static int environ_open(struct inode
*inode
, struct file
*file
)
906 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
909 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
910 size_t count
, loff_t
*ppos
)
913 unsigned long src
= *ppos
;
915 struct mm_struct
*mm
= file
->private_data
;
916 unsigned long env_start
, env_end
;
918 /* Ensure the process spawned far enough to have an environment. */
919 if (!mm
|| !mm
->env_end
)
922 page
= (char *)__get_free_page(GFP_KERNEL
);
927 if (!mmget_not_zero(mm
))
930 down_read(&mm
->mmap_sem
);
931 env_start
= mm
->env_start
;
932 env_end
= mm
->env_end
;
933 up_read(&mm
->mmap_sem
);
936 size_t this_len
, max_len
;
939 if (src
>= (env_end
- env_start
))
942 this_len
= env_end
- (env_start
+ src
);
944 max_len
= min_t(size_t, PAGE_SIZE
, count
);
945 this_len
= min(max_len
, this_len
);
947 retval
= access_remote_vm(mm
, (env_start
+ src
), page
, this_len
, FOLL_ANON
);
954 if (copy_to_user(buf
, page
, retval
)) {
968 free_page((unsigned long) page
);
972 static const struct file_operations proc_environ_operations
= {
973 .open
= environ_open
,
974 .read
= environ_read
,
975 .llseek
= generic_file_llseek
,
976 .release
= mem_release
,
979 static int auxv_open(struct inode
*inode
, struct file
*file
)
981 return __mem_open(inode
, file
, PTRACE_MODE_READ_FSCREDS
);
984 static ssize_t
auxv_read(struct file
*file
, char __user
*buf
,
985 size_t count
, loff_t
*ppos
)
987 struct mm_struct
*mm
= file
->private_data
;
988 unsigned int nwords
= 0;
994 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
995 return simple_read_from_buffer(buf
, count
, ppos
, mm
->saved_auxv
,
996 nwords
* sizeof(mm
->saved_auxv
[0]));
999 static const struct file_operations proc_auxv_operations
= {
1002 .llseek
= generic_file_llseek
,
1003 .release
= mem_release
,
1006 static ssize_t
oom_adj_read(struct file
*file
, char __user
*buf
, size_t count
,
1009 struct task_struct
*task
= get_proc_task(file_inode(file
));
1010 char buffer
[PROC_NUMBUF
];
1011 int oom_adj
= OOM_ADJUST_MIN
;
1016 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MAX
)
1017 oom_adj
= OOM_ADJUST_MAX
;
1019 oom_adj
= (task
->signal
->oom_score_adj
* -OOM_DISABLE
) /
1021 put_task_struct(task
);
1022 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1023 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1026 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1028 static DEFINE_MUTEX(oom_adj_mutex
);
1029 struct mm_struct
*mm
= NULL
;
1030 struct task_struct
*task
;
1033 task
= get_proc_task(file_inode(file
));
1037 mutex_lock(&oom_adj_mutex
);
1039 if (oom_adj
< task
->signal
->oom_score_adj
&&
1040 !capable(CAP_SYS_RESOURCE
)) {
1045 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1046 * /proc/pid/oom_score_adj instead.
1048 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1049 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1052 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1053 !capable(CAP_SYS_RESOURCE
)) {
1060 * Make sure we will check other processes sharing the mm if this is
1061 * not vfrok which wants its own oom_score_adj.
1062 * pin the mm so it doesn't go away and get reused after task_unlock
1064 if (!task
->vfork_done
) {
1065 struct task_struct
*p
= find_lock_task_mm(task
);
1068 if (atomic_read(&p
->mm
->mm_users
) > 1) {
1076 task
->signal
->oom_score_adj
= oom_adj
;
1077 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1078 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1079 trace_oom_score_adj_update(task
);
1082 struct task_struct
*p
;
1085 for_each_process(p
) {
1086 if (same_thread_group(task
, p
))
1089 /* do not touch kernel threads or the global init */
1090 if (p
->flags
& PF_KTHREAD
|| is_global_init(p
))
1094 if (!p
->vfork_done
&& process_shares_mm(p
, mm
)) {
1095 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",
1096 task_pid_nr(p
), p
->comm
,
1097 p
->signal
->oom_score_adj
, oom_adj
,
1098 task_pid_nr(task
), task
->comm
);
1099 p
->signal
->oom_score_adj
= oom_adj
;
1100 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1101 p
->signal
->oom_score_adj_min
= (short)oom_adj
;
1109 mutex_unlock(&oom_adj_mutex
);
1110 put_task_struct(task
);
1115 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1116 * kernels. The effective policy is defined by oom_score_adj, which has a
1117 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1118 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1119 * Processes that become oom disabled via oom_adj will still be oom disabled
1120 * with this implementation.
1122 * oom_adj cannot be removed since existing userspace binaries use it.
1124 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1125 size_t count
, loff_t
*ppos
)
1127 char buffer
[PROC_NUMBUF
];
1131 memset(buffer
, 0, sizeof(buffer
));
1132 if (count
> sizeof(buffer
) - 1)
1133 count
= sizeof(buffer
) - 1;
1134 if (copy_from_user(buffer
, buf
, count
)) {
1139 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1142 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1143 oom_adj
!= OOM_DISABLE
) {
1149 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1150 * value is always attainable.
1152 if (oom_adj
== OOM_ADJUST_MAX
)
1153 oom_adj
= OOM_SCORE_ADJ_MAX
;
1155 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1157 err
= __set_oom_adj(file
, oom_adj
, true);
1159 return err
< 0 ? err
: count
;
1162 static const struct file_operations proc_oom_adj_operations
= {
1163 .read
= oom_adj_read
,
1164 .write
= oom_adj_write
,
1165 .llseek
= generic_file_llseek
,
1168 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1169 size_t count
, loff_t
*ppos
)
1171 struct task_struct
*task
= get_proc_task(file_inode(file
));
1172 char buffer
[PROC_NUMBUF
];
1173 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1178 oom_score_adj
= task
->signal
->oom_score_adj
;
1179 put_task_struct(task
);
1180 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1181 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1184 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1185 size_t count
, loff_t
*ppos
)
1187 char buffer
[PROC_NUMBUF
];
1191 memset(buffer
, 0, sizeof(buffer
));
1192 if (count
> sizeof(buffer
) - 1)
1193 count
= sizeof(buffer
) - 1;
1194 if (copy_from_user(buffer
, buf
, count
)) {
1199 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1202 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1203 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1208 err
= __set_oom_adj(file
, oom_score_adj
, false);
1210 return err
< 0 ? err
: count
;
1213 static const struct file_operations proc_oom_score_adj_operations
= {
1214 .read
= oom_score_adj_read
,
1215 .write
= oom_score_adj_write
,
1216 .llseek
= default_llseek
,
1219 #ifdef CONFIG_AUDITSYSCALL
1220 #define TMPBUFLEN 11
1221 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1222 size_t count
, loff_t
*ppos
)
1224 struct inode
* inode
= file_inode(file
);
1225 struct task_struct
*task
= get_proc_task(inode
);
1227 char tmpbuf
[TMPBUFLEN
];
1231 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1232 from_kuid(file
->f_cred
->user_ns
,
1233 audit_get_loginuid(task
)));
1234 put_task_struct(task
);
1235 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1238 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1239 size_t count
, loff_t
*ppos
)
1241 struct inode
* inode
= file_inode(file
);
1247 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1254 /* No partial writes. */
1258 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1262 /* is userspace tring to explicitly UNSET the loginuid? */
1263 if (loginuid
== AUDIT_UID_UNSET
) {
1264 kloginuid
= INVALID_UID
;
1266 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1267 if (!uid_valid(kloginuid
))
1271 rv
= audit_set_loginuid(kloginuid
);
1277 static const struct file_operations proc_loginuid_operations
= {
1278 .read
= proc_loginuid_read
,
1279 .write
= proc_loginuid_write
,
1280 .llseek
= generic_file_llseek
,
1283 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1284 size_t count
, loff_t
*ppos
)
1286 struct inode
* inode
= file_inode(file
);
1287 struct task_struct
*task
= get_proc_task(inode
);
1289 char tmpbuf
[TMPBUFLEN
];
1293 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1294 audit_get_sessionid(task
));
1295 put_task_struct(task
);
1296 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1299 static const struct file_operations proc_sessionid_operations
= {
1300 .read
= proc_sessionid_read
,
1301 .llseek
= generic_file_llseek
,
1305 #ifdef CONFIG_FAULT_INJECTION
1306 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1307 size_t count
, loff_t
*ppos
)
1309 struct task_struct
*task
= get_proc_task(file_inode(file
));
1310 char buffer
[PROC_NUMBUF
];
1316 make_it_fail
= task
->make_it_fail
;
1317 put_task_struct(task
);
1319 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1321 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1324 static ssize_t
proc_fault_inject_write(struct file
* file
,
1325 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1327 struct task_struct
*task
;
1328 char buffer
[PROC_NUMBUF
];
1332 if (!capable(CAP_SYS_RESOURCE
))
1334 memset(buffer
, 0, sizeof(buffer
));
1335 if (count
> sizeof(buffer
) - 1)
1336 count
= sizeof(buffer
) - 1;
1337 if (copy_from_user(buffer
, buf
, count
))
1339 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1342 if (make_it_fail
< 0 || make_it_fail
> 1)
1345 task
= get_proc_task(file_inode(file
));
1348 task
->make_it_fail
= make_it_fail
;
1349 put_task_struct(task
);
1354 static const struct file_operations proc_fault_inject_operations
= {
1355 .read
= proc_fault_inject_read
,
1356 .write
= proc_fault_inject_write
,
1357 .llseek
= generic_file_llseek
,
1360 static ssize_t
proc_fail_nth_write(struct file
*file
, const char __user
*buf
,
1361 size_t count
, loff_t
*ppos
)
1363 struct task_struct
*task
;
1367 err
= kstrtouint_from_user(buf
, count
, 0, &n
);
1371 task
= get_proc_task(file_inode(file
));
1375 put_task_struct(task
);
1380 static ssize_t
proc_fail_nth_read(struct file
*file
, char __user
*buf
,
1381 size_t count
, loff_t
*ppos
)
1383 struct task_struct
*task
;
1384 char numbuf
[PROC_NUMBUF
];
1387 task
= get_proc_task(file_inode(file
));
1390 len
= snprintf(numbuf
, sizeof(numbuf
), "%u\n", task
->fail_nth
);
1391 len
= simple_read_from_buffer(buf
, count
, ppos
, numbuf
, len
);
1392 put_task_struct(task
);
1397 static const struct file_operations proc_fail_nth_operations
= {
1398 .read
= proc_fail_nth_read
,
1399 .write
= proc_fail_nth_write
,
1404 #ifdef CONFIG_SCHED_DEBUG
1406 * Print out various scheduling related per-task fields:
1408 static int sched_show(struct seq_file
*m
, void *v
)
1410 struct inode
*inode
= m
->private;
1411 struct pid_namespace
*ns
= proc_pid_ns(inode
);
1412 struct task_struct
*p
;
1414 p
= get_proc_task(inode
);
1417 proc_sched_show_task(p
, ns
, m
);
1425 sched_write(struct file
*file
, const char __user
*buf
,
1426 size_t count
, loff_t
*offset
)
1428 struct inode
*inode
= file_inode(file
);
1429 struct task_struct
*p
;
1431 p
= get_proc_task(inode
);
1434 proc_sched_set_task(p
);
1441 static int sched_open(struct inode
*inode
, struct file
*filp
)
1443 return single_open(filp
, sched_show
, inode
);
1446 static const struct file_operations proc_pid_sched_operations
= {
1449 .write
= sched_write
,
1450 .llseek
= seq_lseek
,
1451 .release
= single_release
,
1456 #ifdef CONFIG_SCHED_AUTOGROUP
1458 * Print out autogroup related information:
1460 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1462 struct inode
*inode
= m
->private;
1463 struct task_struct
*p
;
1465 p
= get_proc_task(inode
);
1468 proc_sched_autogroup_show_task(p
, m
);
1476 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1477 size_t count
, loff_t
*offset
)
1479 struct inode
*inode
= file_inode(file
);
1480 struct task_struct
*p
;
1481 char buffer
[PROC_NUMBUF
];
1485 memset(buffer
, 0, sizeof(buffer
));
1486 if (count
> sizeof(buffer
) - 1)
1487 count
= sizeof(buffer
) - 1;
1488 if (copy_from_user(buffer
, buf
, count
))
1491 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1495 p
= get_proc_task(inode
);
1499 err
= proc_sched_autogroup_set_nice(p
, nice
);
1508 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1512 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1514 struct seq_file
*m
= filp
->private_data
;
1521 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1522 .open
= sched_autogroup_open
,
1524 .write
= sched_autogroup_write
,
1525 .llseek
= seq_lseek
,
1526 .release
= single_release
,
1529 #endif /* CONFIG_SCHED_AUTOGROUP */
1531 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1532 size_t count
, loff_t
*offset
)
1534 struct inode
*inode
= file_inode(file
);
1535 struct task_struct
*p
;
1536 char buffer
[TASK_COMM_LEN
];
1537 const size_t maxlen
= sizeof(buffer
) - 1;
1539 memset(buffer
, 0, sizeof(buffer
));
1540 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1543 p
= get_proc_task(inode
);
1547 if (same_thread_group(current
, p
))
1548 set_task_comm(p
, buffer
);
1557 static int comm_show(struct seq_file
*m
, void *v
)
1559 struct inode
*inode
= m
->private;
1560 struct task_struct
*p
;
1562 p
= get_proc_task(inode
);
1567 seq_printf(m
, "%s\n", p
->comm
);
1575 static int comm_open(struct inode
*inode
, struct file
*filp
)
1577 return single_open(filp
, comm_show
, inode
);
1580 static const struct file_operations proc_pid_set_comm_operations
= {
1583 .write
= comm_write
,
1584 .llseek
= seq_lseek
,
1585 .release
= single_release
,
1588 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1590 struct task_struct
*task
;
1591 struct file
*exe_file
;
1593 task
= get_proc_task(d_inode(dentry
));
1596 exe_file
= get_task_exe_file(task
);
1597 put_task_struct(task
);
1599 *exe_path
= exe_file
->f_path
;
1600 path_get(&exe_file
->f_path
);
1607 static const char *proc_pid_get_link(struct dentry
*dentry
,
1608 struct inode
*inode
,
1609 struct delayed_call
*done
)
1612 int error
= -EACCES
;
1615 return ERR_PTR(-ECHILD
);
1617 /* Are we allowed to snoop on the tasks file descriptors? */
1618 if (!proc_fd_access_allowed(inode
))
1621 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1625 nd_jump_link(&path
);
1628 return ERR_PTR(error
);
1631 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1633 char *tmp
= (char *)__get_free_page(GFP_KERNEL
);
1640 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1641 len
= PTR_ERR(pathname
);
1642 if (IS_ERR(pathname
))
1644 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1648 if (copy_to_user(buffer
, pathname
, len
))
1651 free_page((unsigned long)tmp
);
1655 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1657 int error
= -EACCES
;
1658 struct inode
*inode
= d_inode(dentry
);
1661 /* Are we allowed to snoop on the tasks file descriptors? */
1662 if (!proc_fd_access_allowed(inode
))
1665 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1669 error
= do_proc_readlink(&path
, buffer
, buflen
);
1675 const struct inode_operations proc_pid_link_inode_operations
= {
1676 .readlink
= proc_pid_readlink
,
1677 .get_link
= proc_pid_get_link
,
1678 .setattr
= proc_setattr
,
1682 /* building an inode */
1684 void task_dump_owner(struct task_struct
*task
, umode_t mode
,
1685 kuid_t
*ruid
, kgid_t
*rgid
)
1687 /* Depending on the state of dumpable compute who should own a
1688 * proc file for a task.
1690 const struct cred
*cred
;
1694 if (unlikely(task
->flags
& PF_KTHREAD
)) {
1695 *ruid
= GLOBAL_ROOT_UID
;
1696 *rgid
= GLOBAL_ROOT_GID
;
1700 /* Default to the tasks effective ownership */
1702 cred
= __task_cred(task
);
1708 * Before the /proc/pid/status file was created the only way to read
1709 * the effective uid of a /process was to stat /proc/pid. Reading
1710 * /proc/pid/status is slow enough that procps and other packages
1711 * kept stating /proc/pid. To keep the rules in /proc simple I have
1712 * made this apply to all per process world readable and executable
1715 if (mode
!= (S_IFDIR
|S_IRUGO
|S_IXUGO
)) {
1716 struct mm_struct
*mm
;
1719 /* Make non-dumpable tasks owned by some root */
1721 if (get_dumpable(mm
) != SUID_DUMP_USER
) {
1722 struct user_namespace
*user_ns
= mm
->user_ns
;
1724 uid
= make_kuid(user_ns
, 0);
1725 if (!uid_valid(uid
))
1726 uid
= GLOBAL_ROOT_UID
;
1728 gid
= make_kgid(user_ns
, 0);
1729 if (!gid_valid(gid
))
1730 gid
= GLOBAL_ROOT_GID
;
1733 uid
= GLOBAL_ROOT_UID
;
1734 gid
= GLOBAL_ROOT_GID
;
1742 struct inode
*proc_pid_make_inode(struct super_block
* sb
,
1743 struct task_struct
*task
, umode_t mode
)
1745 struct inode
* inode
;
1746 struct proc_inode
*ei
;
1748 /* We need a new inode */
1750 inode
= new_inode(sb
);
1756 inode
->i_mode
= mode
;
1757 inode
->i_ino
= get_next_ino();
1758 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= current_time(inode
);
1759 inode
->i_op
= &proc_def_inode_operations
;
1762 * grab the reference to task.
1764 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1768 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1769 security_task_to_inode(task
, inode
);
1779 int pid_getattr(const struct path
*path
, struct kstat
*stat
,
1780 u32 request_mask
, unsigned int query_flags
)
1782 struct inode
*inode
= d_inode(path
->dentry
);
1783 struct pid_namespace
*pid
= proc_pid_ns(inode
);
1784 struct task_struct
*task
;
1786 generic_fillattr(inode
, stat
);
1789 stat
->uid
= GLOBAL_ROOT_UID
;
1790 stat
->gid
= GLOBAL_ROOT_GID
;
1791 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1793 if (!has_pid_permissions(pid
, task
, HIDEPID_INVISIBLE
)) {
1796 * This doesn't prevent learning whether PID exists,
1797 * it only makes getattr() consistent with readdir().
1801 task_dump_owner(task
, inode
->i_mode
, &stat
->uid
, &stat
->gid
);
1810 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1812 void pid_update_inode(struct task_struct
*task
, struct inode
*inode
)
1814 task_dump_owner(task
, inode
->i_mode
, &inode
->i_uid
, &inode
->i_gid
);
1816 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1817 security_task_to_inode(task
, inode
);
1821 * Rewrite the inode's ownerships here because the owning task may have
1822 * performed a setuid(), etc.
1825 static int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1827 struct inode
*inode
;
1828 struct task_struct
*task
;
1830 if (flags
& LOOKUP_RCU
)
1833 inode
= d_inode(dentry
);
1834 task
= get_proc_task(inode
);
1837 pid_update_inode(task
, inode
);
1838 put_task_struct(task
);
1844 static inline bool proc_inode_is_dead(struct inode
*inode
)
1846 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
1849 int pid_delete_dentry(const struct dentry
*dentry
)
1851 /* Is the task we represent dead?
1852 * If so, then don't put the dentry on the lru list,
1853 * kill it immediately.
1855 return proc_inode_is_dead(d_inode(dentry
));
1858 const struct dentry_operations pid_dentry_operations
=
1860 .d_revalidate
= pid_revalidate
,
1861 .d_delete
= pid_delete_dentry
,
1867 * Fill a directory entry.
1869 * If possible create the dcache entry and derive our inode number and
1870 * file type from dcache entry.
1872 * Since all of the proc inode numbers are dynamically generated, the inode
1873 * numbers do not exist until the inode is cache. This means creating the
1874 * the dcache entry in readdir is necessary to keep the inode numbers
1875 * reported by readdir in sync with the inode numbers reported
1878 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
1879 const char *name
, int len
,
1880 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1882 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
1883 struct qstr qname
= QSTR_INIT(name
, len
);
1884 struct inode
*inode
;
1885 unsigned type
= DT_UNKNOWN
;
1888 child
= d_hash_and_lookup(dir
, &qname
);
1890 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
1891 child
= d_alloc_parallel(dir
, &qname
, &wq
);
1893 goto end_instantiate
;
1894 if (d_in_lookup(child
)) {
1896 res
= instantiate(child
, task
, ptr
);
1897 d_lookup_done(child
);
1899 goto end_instantiate
;
1900 if (unlikely(res
)) {
1906 inode
= d_inode(child
);
1908 type
= inode
->i_mode
>> 12;
1911 return dir_emit(ctx
, name
, len
, ino
, type
);
1915 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1916 * which represent vma start and end addresses.
1918 static int dname_to_vma_addr(struct dentry
*dentry
,
1919 unsigned long *start
, unsigned long *end
)
1921 const char *str
= dentry
->d_name
.name
;
1922 unsigned long long sval
, eval
;
1925 if (str
[0] == '0' && str
[1] != '-')
1927 len
= _parse_integer(str
, 16, &sval
);
1928 if (len
& KSTRTOX_OVERFLOW
)
1930 if (sval
!= (unsigned long)sval
)
1938 if (str
[0] == '0' && str
[1])
1940 len
= _parse_integer(str
, 16, &eval
);
1941 if (len
& KSTRTOX_OVERFLOW
)
1943 if (eval
!= (unsigned long)eval
)
1956 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
1958 unsigned long vm_start
, vm_end
;
1959 bool exact_vma_exists
= false;
1960 struct mm_struct
*mm
= NULL
;
1961 struct task_struct
*task
;
1962 struct inode
*inode
;
1965 if (flags
& LOOKUP_RCU
)
1968 inode
= d_inode(dentry
);
1969 task
= get_proc_task(inode
);
1973 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
1974 if (IS_ERR_OR_NULL(mm
))
1977 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
1978 down_read(&mm
->mmap_sem
);
1979 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
1980 up_read(&mm
->mmap_sem
);
1985 if (exact_vma_exists
) {
1986 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1988 security_task_to_inode(task
, inode
);
1993 put_task_struct(task
);
1999 static const struct dentry_operations tid_map_files_dentry_operations
= {
2000 .d_revalidate
= map_files_d_revalidate
,
2001 .d_delete
= pid_delete_dentry
,
2004 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2006 unsigned long vm_start
, vm_end
;
2007 struct vm_area_struct
*vma
;
2008 struct task_struct
*task
;
2009 struct mm_struct
*mm
;
2013 task
= get_proc_task(d_inode(dentry
));
2017 mm
= get_task_mm(task
);
2018 put_task_struct(task
);
2022 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2027 down_read(&mm
->mmap_sem
);
2028 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2029 if (vma
&& vma
->vm_file
) {
2030 *path
= vma
->vm_file
->f_path
;
2034 up_read(&mm
->mmap_sem
);
2042 struct map_files_info
{
2043 unsigned long start
;
2049 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2050 * symlinks may be used to bypass permissions on ancestor directories in the
2051 * path to the file in question.
2054 proc_map_files_get_link(struct dentry
*dentry
,
2055 struct inode
*inode
,
2056 struct delayed_call
*done
)
2058 if (!capable(CAP_SYS_ADMIN
))
2059 return ERR_PTR(-EPERM
);
2061 return proc_pid_get_link(dentry
, inode
, done
);
2065 * Identical to proc_pid_link_inode_operations except for get_link()
2067 static const struct inode_operations proc_map_files_link_inode_operations
= {
2068 .readlink
= proc_pid_readlink
,
2069 .get_link
= proc_map_files_get_link
,
2070 .setattr
= proc_setattr
,
2073 static struct dentry
*
2074 proc_map_files_instantiate(struct dentry
*dentry
,
2075 struct task_struct
*task
, const void *ptr
)
2077 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
2078 struct proc_inode
*ei
;
2079 struct inode
*inode
;
2081 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFLNK
|
2082 ((mode
& FMODE_READ
) ? S_IRUSR
: 0) |
2083 ((mode
& FMODE_WRITE
) ? S_IWUSR
: 0));
2085 return ERR_PTR(-ENOENT
);
2088 ei
->op
.proc_get_link
= map_files_get_link
;
2090 inode
->i_op
= &proc_map_files_link_inode_operations
;
2093 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2094 return d_splice_alias(inode
, dentry
);
2097 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2098 struct dentry
*dentry
, unsigned int flags
)
2100 unsigned long vm_start
, vm_end
;
2101 struct vm_area_struct
*vma
;
2102 struct task_struct
*task
;
2103 struct dentry
*result
;
2104 struct mm_struct
*mm
;
2106 result
= ERR_PTR(-ENOENT
);
2107 task
= get_proc_task(dir
);
2111 result
= ERR_PTR(-EACCES
);
2112 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2115 result
= ERR_PTR(-ENOENT
);
2116 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2119 mm
= get_task_mm(task
);
2123 down_read(&mm
->mmap_sem
);
2124 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2129 result
= proc_map_files_instantiate(dentry
, task
,
2130 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2133 up_read(&mm
->mmap_sem
);
2136 put_task_struct(task
);
2141 static const struct inode_operations proc_map_files_inode_operations
= {
2142 .lookup
= proc_map_files_lookup
,
2143 .permission
= proc_fd_permission
,
2144 .setattr
= proc_setattr
,
2148 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2150 struct vm_area_struct
*vma
;
2151 struct task_struct
*task
;
2152 struct mm_struct
*mm
;
2153 unsigned long nr_files
, pos
, i
;
2154 struct flex_array
*fa
= NULL
;
2155 struct map_files_info info
;
2156 struct map_files_info
*p
;
2160 task
= get_proc_task(file_inode(file
));
2165 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2169 if (!dir_emit_dots(file
, ctx
))
2172 mm
= get_task_mm(task
);
2175 down_read(&mm
->mmap_sem
);
2180 * We need two passes here:
2182 * 1) Collect vmas of mapped files with mmap_sem taken
2183 * 2) Release mmap_sem and instantiate entries
2185 * otherwise we get lockdep complained, since filldir()
2186 * routine might require mmap_sem taken in might_fault().
2189 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2190 if (vma
->vm_file
&& ++pos
> ctx
->pos
)
2195 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2197 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2201 flex_array_free(fa
);
2202 up_read(&mm
->mmap_sem
);
2206 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2207 vma
= vma
->vm_next
) {
2210 if (++pos
<= ctx
->pos
)
2213 info
.start
= vma
->vm_start
;
2214 info
.end
= vma
->vm_end
;
2215 info
.mode
= vma
->vm_file
->f_mode
;
2216 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2220 up_read(&mm
->mmap_sem
);
2223 for (i
= 0; i
< nr_files
; i
++) {
2224 char buf
[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2227 p
= flex_array_get(fa
, i
);
2228 len
= snprintf(buf
, sizeof(buf
), "%lx-%lx", p
->start
, p
->end
);
2229 if (!proc_fill_cache(file
, ctx
,
2231 proc_map_files_instantiate
,
2233 (void *)(unsigned long)p
->mode
))
2238 flex_array_free(fa
);
2241 put_task_struct(task
);
2246 static const struct file_operations proc_map_files_operations
= {
2247 .read
= generic_read_dir
,
2248 .iterate_shared
= proc_map_files_readdir
,
2249 .llseek
= generic_file_llseek
,
2252 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2253 struct timers_private
{
2255 struct task_struct
*task
;
2256 struct sighand_struct
*sighand
;
2257 struct pid_namespace
*ns
;
2258 unsigned long flags
;
2261 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2263 struct timers_private
*tp
= m
->private;
2265 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2267 return ERR_PTR(-ESRCH
);
2269 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2271 return ERR_PTR(-ESRCH
);
2273 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2276 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2278 struct timers_private
*tp
= m
->private;
2279 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2282 static void timers_stop(struct seq_file
*m
, void *v
)
2284 struct timers_private
*tp
= m
->private;
2287 unlock_task_sighand(tp
->task
, &tp
->flags
);
2292 put_task_struct(tp
->task
);
2297 static int show_timer(struct seq_file
*m
, void *v
)
2299 struct k_itimer
*timer
;
2300 struct timers_private
*tp
= m
->private;
2302 static const char * const nstr
[] = {
2303 [SIGEV_SIGNAL
] = "signal",
2304 [SIGEV_NONE
] = "none",
2305 [SIGEV_THREAD
] = "thread",
2308 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2309 notify
= timer
->it_sigev_notify
;
2311 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2312 seq_printf(m
, "signal: %d/%px\n",
2313 timer
->sigq
->info
.si_signo
,
2314 timer
->sigq
->info
.si_value
.sival_ptr
);
2315 seq_printf(m
, "notify: %s/%s.%d\n",
2316 nstr
[notify
& ~SIGEV_THREAD_ID
],
2317 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2318 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2319 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2324 static const struct seq_operations proc_timers_seq_ops
= {
2325 .start
= timers_start
,
2326 .next
= timers_next
,
2327 .stop
= timers_stop
,
2331 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2333 struct timers_private
*tp
;
2335 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2336 sizeof(struct timers_private
));
2340 tp
->pid
= proc_pid(inode
);
2341 tp
->ns
= proc_pid_ns(inode
);
2345 static const struct file_operations proc_timers_operations
= {
2346 .open
= proc_timers_open
,
2348 .llseek
= seq_lseek
,
2349 .release
= seq_release_private
,
2353 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2354 size_t count
, loff_t
*offset
)
2356 struct inode
*inode
= file_inode(file
);
2357 struct task_struct
*p
;
2361 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2365 p
= get_proc_task(inode
);
2370 if (!capable(CAP_SYS_NICE
)) {
2375 err
= security_task_setscheduler(p
);
2384 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2386 p
->timer_slack_ns
= slack_ns
;
2395 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2397 struct inode
*inode
= m
->private;
2398 struct task_struct
*p
;
2401 p
= get_proc_task(inode
);
2407 if (!capable(CAP_SYS_NICE
)) {
2411 err
= security_task_getscheduler(p
);
2417 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2426 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2428 return single_open(filp
, timerslack_ns_show
, inode
);
2431 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2432 .open
= timerslack_ns_open
,
2434 .write
= timerslack_ns_write
,
2435 .llseek
= seq_lseek
,
2436 .release
= single_release
,
2439 static struct dentry
*proc_pident_instantiate(struct dentry
*dentry
,
2440 struct task_struct
*task
, const void *ptr
)
2442 const struct pid_entry
*p
= ptr
;
2443 struct inode
*inode
;
2444 struct proc_inode
*ei
;
2446 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, p
->mode
);
2448 return ERR_PTR(-ENOENT
);
2451 if (S_ISDIR(inode
->i_mode
))
2452 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2454 inode
->i_op
= p
->iop
;
2456 inode
->i_fop
= p
->fop
;
2458 pid_update_inode(task
, inode
);
2459 d_set_d_op(dentry
, &pid_dentry_operations
);
2460 return d_splice_alias(inode
, dentry
);
2463 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2464 struct dentry
*dentry
,
2465 const struct pid_entry
*ents
,
2468 struct task_struct
*task
= get_proc_task(dir
);
2469 const struct pid_entry
*p
, *last
;
2470 struct dentry
*res
= ERR_PTR(-ENOENT
);
2476 * Yes, it does not scale. And it should not. Don't add
2477 * new entries into /proc/<tgid>/ without very good reasons.
2479 last
= &ents
[nents
];
2480 for (p
= ents
; p
< last
; p
++) {
2481 if (p
->len
!= dentry
->d_name
.len
)
2483 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2489 res
= proc_pident_instantiate(dentry
, task
, p
);
2491 put_task_struct(task
);
2496 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2497 const struct pid_entry
*ents
, unsigned int nents
)
2499 struct task_struct
*task
= get_proc_task(file_inode(file
));
2500 const struct pid_entry
*p
;
2505 if (!dir_emit_dots(file
, ctx
))
2508 if (ctx
->pos
>= nents
+ 2)
2511 for (p
= ents
+ (ctx
->pos
- 2); p
< ents
+ nents
; p
++) {
2512 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2513 proc_pident_instantiate
, task
, p
))
2518 put_task_struct(task
);
2522 #ifdef CONFIG_SECURITY
2523 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2524 size_t count
, loff_t
*ppos
)
2526 struct inode
* inode
= file_inode(file
);
2529 struct task_struct
*task
= get_proc_task(inode
);
2534 length
= security_getprocattr(task
,
2535 (char*)file
->f_path
.dentry
->d_name
.name
,
2537 put_task_struct(task
);
2539 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2544 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2545 size_t count
, loff_t
*ppos
)
2547 struct inode
* inode
= file_inode(file
);
2550 struct task_struct
*task
= get_proc_task(inode
);
2556 /* A task may only write its own attributes. */
2558 if (current
!= task
)
2561 if (count
> PAGE_SIZE
)
2564 /* No partial writes. */
2569 page
= memdup_user(buf
, count
);
2571 length
= PTR_ERR(page
);
2575 /* Guard against adverse ptrace interaction */
2576 length
= mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
);
2580 length
= security_setprocattr(file
->f_path
.dentry
->d_name
.name
,
2582 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
2586 put_task_struct(task
);
2591 static const struct file_operations proc_pid_attr_operations
= {
2592 .read
= proc_pid_attr_read
,
2593 .write
= proc_pid_attr_write
,
2594 .llseek
= generic_file_llseek
,
2597 static const struct pid_entry attr_dir_stuff
[] = {
2598 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2599 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2600 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2601 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2602 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2603 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2606 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2608 return proc_pident_readdir(file
, ctx
,
2609 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2612 static const struct file_operations proc_attr_dir_operations
= {
2613 .read
= generic_read_dir
,
2614 .iterate_shared
= proc_attr_dir_readdir
,
2615 .llseek
= generic_file_llseek
,
2618 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2619 struct dentry
*dentry
, unsigned int flags
)
2621 return proc_pident_lookup(dir
, dentry
,
2622 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2625 static const struct inode_operations proc_attr_dir_inode_operations
= {
2626 .lookup
= proc_attr_dir_lookup
,
2627 .getattr
= pid_getattr
,
2628 .setattr
= proc_setattr
,
2633 #ifdef CONFIG_ELF_CORE
2634 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2635 size_t count
, loff_t
*ppos
)
2637 struct task_struct
*task
= get_proc_task(file_inode(file
));
2638 struct mm_struct
*mm
;
2639 char buffer
[PROC_NUMBUF
];
2647 mm
= get_task_mm(task
);
2649 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2650 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2651 MMF_DUMP_FILTER_SHIFT
));
2653 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2656 put_task_struct(task
);
2661 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2662 const char __user
*buf
,
2666 struct task_struct
*task
;
2667 struct mm_struct
*mm
;
2673 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2678 task
= get_proc_task(file_inode(file
));
2682 mm
= get_task_mm(task
);
2687 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2689 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2691 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2696 put_task_struct(task
);
2703 static const struct file_operations proc_coredump_filter_operations
= {
2704 .read
= proc_coredump_filter_read
,
2705 .write
= proc_coredump_filter_write
,
2706 .llseek
= generic_file_llseek
,
2710 #ifdef CONFIG_TASK_IO_ACCOUNTING
2711 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2713 struct task_io_accounting acct
= task
->ioac
;
2714 unsigned long flags
;
2717 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2721 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2726 if (whole
&& lock_task_sighand(task
, &flags
)) {
2727 struct task_struct
*t
= task
;
2729 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2730 while_each_thread(task
, t
)
2731 task_io_accounting_add(&acct
, &t
->ioac
);
2733 unlock_task_sighand(task
, &flags
);
2740 "read_bytes: %llu\n"
2741 "write_bytes: %llu\n"
2742 "cancelled_write_bytes: %llu\n",
2743 (unsigned long long)acct
.rchar
,
2744 (unsigned long long)acct
.wchar
,
2745 (unsigned long long)acct
.syscr
,
2746 (unsigned long long)acct
.syscw
,
2747 (unsigned long long)acct
.read_bytes
,
2748 (unsigned long long)acct
.write_bytes
,
2749 (unsigned long long)acct
.cancelled_write_bytes
);
2753 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2757 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2758 struct pid
*pid
, struct task_struct
*task
)
2760 return do_io_accounting(task
, m
, 0);
2763 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2764 struct pid
*pid
, struct task_struct
*task
)
2766 return do_io_accounting(task
, m
, 1);
2768 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2770 #ifdef CONFIG_USER_NS
2771 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2772 const struct seq_operations
*seq_ops
)
2774 struct user_namespace
*ns
= NULL
;
2775 struct task_struct
*task
;
2776 struct seq_file
*seq
;
2779 task
= get_proc_task(inode
);
2782 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2784 put_task_struct(task
);
2789 ret
= seq_open(file
, seq_ops
);
2793 seq
= file
->private_data
;
2803 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2805 struct seq_file
*seq
= file
->private_data
;
2806 struct user_namespace
*ns
= seq
->private;
2808 return seq_release(inode
, file
);
2811 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2813 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2816 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2818 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2821 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
2823 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
2826 static const struct file_operations proc_uid_map_operations
= {
2827 .open
= proc_uid_map_open
,
2828 .write
= proc_uid_map_write
,
2830 .llseek
= seq_lseek
,
2831 .release
= proc_id_map_release
,
2834 static const struct file_operations proc_gid_map_operations
= {
2835 .open
= proc_gid_map_open
,
2836 .write
= proc_gid_map_write
,
2838 .llseek
= seq_lseek
,
2839 .release
= proc_id_map_release
,
2842 static const struct file_operations proc_projid_map_operations
= {
2843 .open
= proc_projid_map_open
,
2844 .write
= proc_projid_map_write
,
2846 .llseek
= seq_lseek
,
2847 .release
= proc_id_map_release
,
2850 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
2852 struct user_namespace
*ns
= NULL
;
2853 struct task_struct
*task
;
2857 task
= get_proc_task(inode
);
2860 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2862 put_task_struct(task
);
2867 if (file
->f_mode
& FMODE_WRITE
) {
2869 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
2873 ret
= single_open(file
, &proc_setgroups_show
, ns
);
2884 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
2886 struct seq_file
*seq
= file
->private_data
;
2887 struct user_namespace
*ns
= seq
->private;
2888 int ret
= single_release(inode
, file
);
2893 static const struct file_operations proc_setgroups_operations
= {
2894 .open
= proc_setgroups_open
,
2895 .write
= proc_setgroups_write
,
2897 .llseek
= seq_lseek
,
2898 .release
= proc_setgroups_release
,
2900 #endif /* CONFIG_USER_NS */
2902 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2903 struct pid
*pid
, struct task_struct
*task
)
2905 int err
= lock_trace(task
);
2907 seq_printf(m
, "%08x\n", task
->personality
);
2913 #ifdef CONFIG_LIVEPATCH
2914 static int proc_pid_patch_state(struct seq_file
*m
, struct pid_namespace
*ns
,
2915 struct pid
*pid
, struct task_struct
*task
)
2917 seq_printf(m
, "%d\n", task
->patch_state
);
2920 #endif /* CONFIG_LIVEPATCH */
2925 static const struct file_operations proc_task_operations
;
2926 static const struct inode_operations proc_task_inode_operations
;
2928 static const struct pid_entry tgid_base_stuff
[] = {
2929 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2930 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2931 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
2932 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2933 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
2935 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2937 REG("environ", S_IRUSR
, proc_environ_operations
),
2938 REG("auxv", S_IRUSR
, proc_auxv_operations
),
2939 ONE("status", S_IRUGO
, proc_pid_status
),
2940 ONE("personality", S_IRUSR
, proc_pid_personality
),
2941 ONE("limits", S_IRUGO
, proc_pid_limits
),
2942 #ifdef CONFIG_SCHED_DEBUG
2943 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2945 #ifdef CONFIG_SCHED_AUTOGROUP
2946 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
2948 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2949 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2950 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
2952 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
2953 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2954 ONE("statm", S_IRUGO
, proc_pid_statm
),
2955 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
2957 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
2959 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2960 LNK("cwd", proc_cwd_link
),
2961 LNK("root", proc_root_link
),
2962 LNK("exe", proc_exe_link
),
2963 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2964 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2965 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2966 #ifdef CONFIG_PROC_PAGE_MONITOR
2967 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2968 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
2969 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
2970 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2972 #ifdef CONFIG_SECURITY
2973 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2975 #ifdef CONFIG_KALLSYMS
2976 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
2978 #ifdef CONFIG_STACKTRACE
2979 ONE("stack", S_IRUSR
, proc_pid_stack
),
2981 #ifdef CONFIG_SCHED_INFO
2982 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
2984 #ifdef CONFIG_LATENCYTOP
2985 REG("latency", S_IRUGO
, proc_lstats_operations
),
2987 #ifdef CONFIG_PROC_PID_CPUSET
2988 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
2990 #ifdef CONFIG_CGROUPS
2991 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
2993 ONE("oom_score", S_IRUGO
, proc_oom_score
),
2994 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
2995 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
2996 #ifdef CONFIG_AUDITSYSCALL
2997 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2998 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3000 #ifdef CONFIG_FAULT_INJECTION
3001 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3002 REG("fail-nth", 0644, proc_fail_nth_operations
),
3004 #ifdef CONFIG_ELF_CORE
3005 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3007 #ifdef CONFIG_TASK_IO_ACCOUNTING
3008 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
3010 #ifdef CONFIG_USER_NS
3011 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3012 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3013 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3014 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3016 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3017 REG("timers", S_IRUGO
, proc_timers_operations
),
3019 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
3020 #ifdef CONFIG_LIVEPATCH
3021 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3025 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3027 return proc_pident_readdir(file
, ctx
,
3028 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3031 static const struct file_operations proc_tgid_base_operations
= {
3032 .read
= generic_read_dir
,
3033 .iterate_shared
= proc_tgid_base_readdir
,
3034 .llseek
= generic_file_llseek
,
3037 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3039 return proc_pident_lookup(dir
, dentry
,
3040 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3043 static const struct inode_operations proc_tgid_base_inode_operations
= {
3044 .lookup
= proc_tgid_base_lookup
,
3045 .getattr
= pid_getattr
,
3046 .setattr
= proc_setattr
,
3047 .permission
= proc_pid_permission
,
3050 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
3052 struct dentry
*dentry
, *leader
, *dir
;
3057 name
.len
= snprintf(buf
, sizeof(buf
), "%u", pid
);
3058 /* no ->d_hash() rejects on procfs */
3059 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3061 d_invalidate(dentry
);
3069 name
.len
= snprintf(buf
, sizeof(buf
), "%u", tgid
);
3070 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3075 name
.len
= strlen(name
.name
);
3076 dir
= d_hash_and_lookup(leader
, &name
);
3078 goto out_put_leader
;
3081 name
.len
= snprintf(buf
, sizeof(buf
), "%u", pid
);
3082 dentry
= d_hash_and_lookup(dir
, &name
);
3084 d_invalidate(dentry
);
3096 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3097 * @task: task that should be flushed.
3099 * When flushing dentries from proc, one needs to flush them from global
3100 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3101 * in. This call is supposed to do all of this job.
3103 * Looks in the dcache for
3105 * /proc/@tgid/task/@pid
3106 * if either directory is present flushes it and all of it'ts children
3109 * It is safe and reasonable to cache /proc entries for a task until
3110 * that task exits. After that they just clog up the dcache with
3111 * useless entries, possibly causing useful dcache entries to be
3112 * flushed instead. This routine is proved to flush those useless
3113 * dcache entries at process exit time.
3115 * NOTE: This routine is just an optimization so it does not guarantee
3116 * that no dcache entries will exist at process exit time it
3117 * just makes it very unlikely that any will persist.
3120 void proc_flush_task(struct task_struct
*task
)
3123 struct pid
*pid
, *tgid
;
3126 pid
= task_pid(task
);
3127 tgid
= task_tgid(task
);
3129 for (i
= 0; i
<= pid
->level
; i
++) {
3130 upid
= &pid
->numbers
[i
];
3131 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
3132 tgid
->numbers
[i
].nr
);
3136 static struct dentry
*proc_pid_instantiate(struct dentry
* dentry
,
3137 struct task_struct
*task
, const void *ptr
)
3139 struct inode
*inode
;
3141 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3143 return ERR_PTR(-ENOENT
);
3145 inode
->i_op
= &proc_tgid_base_inode_operations
;
3146 inode
->i_fop
= &proc_tgid_base_operations
;
3147 inode
->i_flags
|=S_IMMUTABLE
;
3149 set_nlink(inode
, nlink_tgid
);
3150 pid_update_inode(task
, inode
);
3152 d_set_d_op(dentry
, &pid_dentry_operations
);
3153 return d_splice_alias(inode
, dentry
);
3156 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3158 struct task_struct
*task
;
3160 struct pid_namespace
*ns
;
3161 struct dentry
*result
= ERR_PTR(-ENOENT
);
3163 tgid
= name_to_int(&dentry
->d_name
);
3167 ns
= dentry
->d_sb
->s_fs_info
;
3169 task
= find_task_by_pid_ns(tgid
, ns
);
3171 get_task_struct(task
);
3176 result
= proc_pid_instantiate(dentry
, task
, NULL
);
3177 put_task_struct(task
);
3183 * Find the first task with tgid >= tgid
3188 struct task_struct
*task
;
3190 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3195 put_task_struct(iter
.task
);
3199 pid
= find_ge_pid(iter
.tgid
, ns
);
3201 iter
.tgid
= pid_nr_ns(pid
, ns
);
3202 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3203 /* What we to know is if the pid we have find is the
3204 * pid of a thread_group_leader. Testing for task
3205 * being a thread_group_leader is the obvious thing
3206 * todo but there is a window when it fails, due to
3207 * the pid transfer logic in de_thread.
3209 * So we perform the straight forward test of seeing
3210 * if the pid we have found is the pid of a thread
3211 * group leader, and don't worry if the task we have
3212 * found doesn't happen to be a thread group leader.
3213 * As we don't care in the case of readdir.
3215 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3219 get_task_struct(iter
.task
);
3225 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3227 /* for the /proc/ directory itself, after non-process stuff has been done */
3228 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3230 struct tgid_iter iter
;
3231 struct pid_namespace
*ns
= proc_pid_ns(file_inode(file
));
3232 loff_t pos
= ctx
->pos
;
3234 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3237 if (pos
== TGID_OFFSET
- 2) {
3238 struct inode
*inode
= d_inode(ns
->proc_self
);
3239 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3241 ctx
->pos
= pos
= pos
+ 1;
3243 if (pos
== TGID_OFFSET
- 1) {
3244 struct inode
*inode
= d_inode(ns
->proc_thread_self
);
3245 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3247 ctx
->pos
= pos
= pos
+ 1;
3249 iter
.tgid
= pos
- TGID_OFFSET
;
3251 for (iter
= next_tgid(ns
, iter
);
3253 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3258 if (!has_pid_permissions(ns
, iter
.task
, HIDEPID_INVISIBLE
))
3261 len
= snprintf(name
, sizeof(name
), "%u", iter
.tgid
);
3262 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3263 if (!proc_fill_cache(file
, ctx
, name
, len
,
3264 proc_pid_instantiate
, iter
.task
, NULL
)) {
3265 put_task_struct(iter
.task
);
3269 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3274 * proc_tid_comm_permission is a special permission function exclusively
3275 * used for the node /proc/<pid>/task/<tid>/comm.
3276 * It bypasses generic permission checks in the case where a task of the same
3277 * task group attempts to access the node.
3278 * The rationale behind this is that glibc and bionic access this node for
3279 * cross thread naming (pthread_set/getname_np(!self)). However, if
3280 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3281 * which locks out the cross thread naming implementation.
3282 * This function makes sure that the node is always accessible for members of
3283 * same thread group.
3285 static int proc_tid_comm_permission(struct inode
*inode
, int mask
)
3287 bool is_same_tgroup
;
3288 struct task_struct
*task
;
3290 task
= get_proc_task(inode
);
3293 is_same_tgroup
= same_thread_group(current
, task
);
3294 put_task_struct(task
);
3296 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3297 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3298 * read or written by the members of the corresponding
3304 return generic_permission(inode
, mask
);
3307 static const struct inode_operations proc_tid_comm_inode_operations
= {
3308 .permission
= proc_tid_comm_permission
,
3314 static const struct pid_entry tid_base_stuff
[] = {
3315 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3316 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3317 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3319 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3321 REG("environ", S_IRUSR
, proc_environ_operations
),
3322 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3323 ONE("status", S_IRUGO
, proc_pid_status
),
3324 ONE("personality", S_IRUSR
, proc_pid_personality
),
3325 ONE("limits", S_IRUGO
, proc_pid_limits
),
3326 #ifdef CONFIG_SCHED_DEBUG
3327 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3329 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3330 &proc_tid_comm_inode_operations
,
3331 &proc_pid_set_comm_operations
, {}),
3332 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3333 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3335 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3336 ONE("stat", S_IRUGO
, proc_tid_stat
),
3337 ONE("statm", S_IRUGO
, proc_pid_statm
),
3338 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3339 #ifdef CONFIG_PROC_CHILDREN
3340 REG("children", S_IRUGO
, proc_tid_children_operations
),
3343 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3345 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3346 LNK("cwd", proc_cwd_link
),
3347 LNK("root", proc_root_link
),
3348 LNK("exe", proc_exe_link
),
3349 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3350 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3351 #ifdef CONFIG_PROC_PAGE_MONITOR
3352 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3353 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3354 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3355 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3357 #ifdef CONFIG_SECURITY
3358 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3360 #ifdef CONFIG_KALLSYMS
3361 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3363 #ifdef CONFIG_STACKTRACE
3364 ONE("stack", S_IRUSR
, proc_pid_stack
),
3366 #ifdef CONFIG_SCHED_INFO
3367 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3369 #ifdef CONFIG_LATENCYTOP
3370 REG("latency", S_IRUGO
, proc_lstats_operations
),
3372 #ifdef CONFIG_PROC_PID_CPUSET
3373 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3375 #ifdef CONFIG_CGROUPS
3376 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3378 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3379 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3380 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3381 #ifdef CONFIG_AUDITSYSCALL
3382 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3383 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3385 #ifdef CONFIG_FAULT_INJECTION
3386 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3387 REG("fail-nth", 0644, proc_fail_nth_operations
),
3389 #ifdef CONFIG_TASK_IO_ACCOUNTING
3390 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3392 #ifdef CONFIG_USER_NS
3393 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3394 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3395 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3396 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3398 #ifdef CONFIG_LIVEPATCH
3399 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3403 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3405 return proc_pident_readdir(file
, ctx
,
3406 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3409 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3411 return proc_pident_lookup(dir
, dentry
,
3412 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3415 static const struct file_operations proc_tid_base_operations
= {
3416 .read
= generic_read_dir
,
3417 .iterate_shared
= proc_tid_base_readdir
,
3418 .llseek
= generic_file_llseek
,
3421 static const struct inode_operations proc_tid_base_inode_operations
= {
3422 .lookup
= proc_tid_base_lookup
,
3423 .getattr
= pid_getattr
,
3424 .setattr
= proc_setattr
,
3427 static struct dentry
*proc_task_instantiate(struct dentry
*dentry
,
3428 struct task_struct
*task
, const void *ptr
)
3430 struct inode
*inode
;
3431 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3433 return ERR_PTR(-ENOENT
);
3435 inode
->i_op
= &proc_tid_base_inode_operations
;
3436 inode
->i_fop
= &proc_tid_base_operations
;
3437 inode
->i_flags
|= S_IMMUTABLE
;
3439 set_nlink(inode
, nlink_tid
);
3440 pid_update_inode(task
, inode
);
3442 d_set_d_op(dentry
, &pid_dentry_operations
);
3443 return d_splice_alias(inode
, dentry
);
3446 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3448 struct task_struct
*task
;
3449 struct task_struct
*leader
= get_proc_task(dir
);
3451 struct pid_namespace
*ns
;
3452 struct dentry
*result
= ERR_PTR(-ENOENT
);
3457 tid
= name_to_int(&dentry
->d_name
);
3461 ns
= dentry
->d_sb
->s_fs_info
;
3463 task
= find_task_by_pid_ns(tid
, ns
);
3465 get_task_struct(task
);
3469 if (!same_thread_group(leader
, task
))
3472 result
= proc_task_instantiate(dentry
, task
, NULL
);
3474 put_task_struct(task
);
3476 put_task_struct(leader
);
3482 * Find the first tid of a thread group to return to user space.
3484 * Usually this is just the thread group leader, but if the users
3485 * buffer was too small or there was a seek into the middle of the
3486 * directory we have more work todo.
3488 * In the case of a short read we start with find_task_by_pid.
3490 * In the case of a seek we start with the leader and walk nr
3493 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3494 struct pid_namespace
*ns
)
3496 struct task_struct
*pos
, *task
;
3497 unsigned long nr
= f_pos
;
3499 if (nr
!= f_pos
) /* 32bit overflow? */
3503 task
= pid_task(pid
, PIDTYPE_PID
);
3507 /* Attempt to start with the tid of a thread */
3509 pos
= find_task_by_pid_ns(tid
, ns
);
3510 if (pos
&& same_thread_group(pos
, task
))
3514 /* If nr exceeds the number of threads there is nothing todo */
3515 if (nr
>= get_nr_threads(task
))
3518 /* If we haven't found our starting place yet start
3519 * with the leader and walk nr threads forward.
3521 pos
= task
= task
->group_leader
;
3525 } while_each_thread(task
, pos
);
3530 get_task_struct(pos
);
3537 * Find the next thread in the thread list.
3538 * Return NULL if there is an error or no next thread.
3540 * The reference to the input task_struct is released.
3542 static struct task_struct
*next_tid(struct task_struct
*start
)
3544 struct task_struct
*pos
= NULL
;
3546 if (pid_alive(start
)) {
3547 pos
= next_thread(start
);
3548 if (thread_group_leader(pos
))
3551 get_task_struct(pos
);
3554 put_task_struct(start
);
3558 /* for the /proc/TGID/task/ directories */
3559 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3561 struct inode
*inode
= file_inode(file
);
3562 struct task_struct
*task
;
3563 struct pid_namespace
*ns
;
3566 if (proc_inode_is_dead(inode
))
3569 if (!dir_emit_dots(file
, ctx
))
3572 /* f_version caches the tgid value that the last readdir call couldn't
3573 * return. lseek aka telldir automagically resets f_version to 0.
3575 ns
= proc_pid_ns(inode
);
3576 tid
= (int)file
->f_version
;
3577 file
->f_version
= 0;
3578 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3580 task
= next_tid(task
), ctx
->pos
++) {
3583 tid
= task_pid_nr_ns(task
, ns
);
3584 len
= snprintf(name
, sizeof(name
), "%u", tid
);
3585 if (!proc_fill_cache(file
, ctx
, name
, len
,
3586 proc_task_instantiate
, task
, NULL
)) {
3587 /* returning this tgid failed, save it as the first
3588 * pid for the next readir call */
3589 file
->f_version
= (u64
)tid
;
3590 put_task_struct(task
);
3598 static int proc_task_getattr(const struct path
*path
, struct kstat
*stat
,
3599 u32 request_mask
, unsigned int query_flags
)
3601 struct inode
*inode
= d_inode(path
->dentry
);
3602 struct task_struct
*p
= get_proc_task(inode
);
3603 generic_fillattr(inode
, stat
);
3606 stat
->nlink
+= get_nr_threads(p
);
3613 static const struct inode_operations proc_task_inode_operations
= {
3614 .lookup
= proc_task_lookup
,
3615 .getattr
= proc_task_getattr
,
3616 .setattr
= proc_setattr
,
3617 .permission
= proc_pid_permission
,
3620 static const struct file_operations proc_task_operations
= {
3621 .read
= generic_read_dir
,
3622 .iterate_shared
= proc_task_readdir
,
3623 .llseek
= generic_file_llseek
,
3626 void __init
set_proc_pid_nlink(void)
3628 nlink_tid
= pid_entry_nlink(tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3629 nlink_tgid
= pid_entry_nlink(tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));