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
= inode
->i_sb
->s_fs_info
;
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
;
738 struct task_struct
*task
;
741 ns
= inode
->i_sb
->s_fs_info
;
742 pid
= proc_pid(inode
);
743 task
= get_pid_task(pid
, PIDTYPE_PID
);
747 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
749 put_task_struct(task
);
753 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
755 return single_open(filp
, proc_single_show
, inode
);
758 static const struct file_operations proc_single_file_operations
= {
759 .open
= proc_single_open
,
762 .release
= single_release
,
766 struct mm_struct
*proc_mem_open(struct inode
*inode
, unsigned int mode
)
768 struct task_struct
*task
= get_proc_task(inode
);
769 struct mm_struct
*mm
= ERR_PTR(-ESRCH
);
772 mm
= mm_access(task
, mode
| PTRACE_MODE_FSCREDS
);
773 put_task_struct(task
);
775 if (!IS_ERR_OR_NULL(mm
)) {
776 /* ensure this mm_struct can't be freed */
778 /* but do not pin its memory */
786 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
788 struct mm_struct
*mm
= proc_mem_open(inode
, mode
);
793 file
->private_data
= mm
;
797 static int mem_open(struct inode
*inode
, struct file
*file
)
799 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
801 /* OK to pass negative loff_t, we can catch out-of-range */
802 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
807 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
808 size_t count
, loff_t
*ppos
, int write
)
810 struct mm_struct
*mm
= file
->private_data
;
811 unsigned long addr
= *ppos
;
819 page
= (char *)__get_free_page(GFP_KERNEL
);
824 if (!mmget_not_zero(mm
))
827 flags
= FOLL_FORCE
| (write
? FOLL_WRITE
: 0);
830 int this_len
= min_t(int, count
, PAGE_SIZE
);
832 if (write
&& copy_from_user(page
, buf
, this_len
)) {
837 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, flags
);
844 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
858 free_page((unsigned long) page
);
862 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
863 size_t count
, loff_t
*ppos
)
865 return mem_rw(file
, buf
, count
, ppos
, 0);
868 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
869 size_t count
, loff_t
*ppos
)
871 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
874 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
878 file
->f_pos
= offset
;
881 file
->f_pos
+= offset
;
886 force_successful_syscall_return();
890 static int mem_release(struct inode
*inode
, struct file
*file
)
892 struct mm_struct
*mm
= file
->private_data
;
898 static const struct file_operations proc_mem_operations
= {
903 .release
= mem_release
,
906 static int environ_open(struct inode
*inode
, struct file
*file
)
908 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
911 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
912 size_t count
, loff_t
*ppos
)
915 unsigned long src
= *ppos
;
917 struct mm_struct
*mm
= file
->private_data
;
918 unsigned long env_start
, env_end
;
920 /* Ensure the process spawned far enough to have an environment. */
921 if (!mm
|| !mm
->env_end
)
924 page
= (char *)__get_free_page(GFP_KERNEL
);
929 if (!mmget_not_zero(mm
))
932 down_read(&mm
->mmap_sem
);
933 env_start
= mm
->env_start
;
934 env_end
= mm
->env_end
;
935 up_read(&mm
->mmap_sem
);
938 size_t this_len
, max_len
;
941 if (src
>= (env_end
- env_start
))
944 this_len
= env_end
- (env_start
+ src
);
946 max_len
= min_t(size_t, PAGE_SIZE
, count
);
947 this_len
= min(max_len
, this_len
);
949 retval
= access_remote_vm(mm
, (env_start
+ src
), page
, this_len
, FOLL_ANON
);
956 if (copy_to_user(buf
, page
, retval
)) {
970 free_page((unsigned long) page
);
974 static const struct file_operations proc_environ_operations
= {
975 .open
= environ_open
,
976 .read
= environ_read
,
977 .llseek
= generic_file_llseek
,
978 .release
= mem_release
,
981 static int auxv_open(struct inode
*inode
, struct file
*file
)
983 return __mem_open(inode
, file
, PTRACE_MODE_READ_FSCREDS
);
986 static ssize_t
auxv_read(struct file
*file
, char __user
*buf
,
987 size_t count
, loff_t
*ppos
)
989 struct mm_struct
*mm
= file
->private_data
;
990 unsigned int nwords
= 0;
996 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
997 return simple_read_from_buffer(buf
, count
, ppos
, mm
->saved_auxv
,
998 nwords
* sizeof(mm
->saved_auxv
[0]));
1001 static const struct file_operations proc_auxv_operations
= {
1004 .llseek
= generic_file_llseek
,
1005 .release
= mem_release
,
1008 static ssize_t
oom_adj_read(struct file
*file
, char __user
*buf
, size_t count
,
1011 struct task_struct
*task
= get_proc_task(file_inode(file
));
1012 char buffer
[PROC_NUMBUF
];
1013 int oom_adj
= OOM_ADJUST_MIN
;
1018 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MAX
)
1019 oom_adj
= OOM_ADJUST_MAX
;
1021 oom_adj
= (task
->signal
->oom_score_adj
* -OOM_DISABLE
) /
1023 put_task_struct(task
);
1024 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1025 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1028 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1030 static DEFINE_MUTEX(oom_adj_mutex
);
1031 struct mm_struct
*mm
= NULL
;
1032 struct task_struct
*task
;
1035 task
= get_proc_task(file_inode(file
));
1039 mutex_lock(&oom_adj_mutex
);
1041 if (oom_adj
< task
->signal
->oom_score_adj
&&
1042 !capable(CAP_SYS_RESOURCE
)) {
1047 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1048 * /proc/pid/oom_score_adj instead.
1050 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1051 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1054 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1055 !capable(CAP_SYS_RESOURCE
)) {
1062 * Make sure we will check other processes sharing the mm if this is
1063 * not vfrok which wants its own oom_score_adj.
1064 * pin the mm so it doesn't go away and get reused after task_unlock
1066 if (!task
->vfork_done
) {
1067 struct task_struct
*p
= find_lock_task_mm(task
);
1070 if (atomic_read(&p
->mm
->mm_users
) > 1) {
1078 task
->signal
->oom_score_adj
= oom_adj
;
1079 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1080 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1081 trace_oom_score_adj_update(task
);
1084 struct task_struct
*p
;
1087 for_each_process(p
) {
1088 if (same_thread_group(task
, p
))
1091 /* do not touch kernel threads or the global init */
1092 if (p
->flags
& PF_KTHREAD
|| is_global_init(p
))
1096 if (!p
->vfork_done
&& process_shares_mm(p
, mm
)) {
1097 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",
1098 task_pid_nr(p
), p
->comm
,
1099 p
->signal
->oom_score_adj
, oom_adj
,
1100 task_pid_nr(task
), task
->comm
);
1101 p
->signal
->oom_score_adj
= oom_adj
;
1102 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1103 p
->signal
->oom_score_adj_min
= (short)oom_adj
;
1111 mutex_unlock(&oom_adj_mutex
);
1112 put_task_struct(task
);
1117 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1118 * kernels. The effective policy is defined by oom_score_adj, which has a
1119 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1120 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1121 * Processes that become oom disabled via oom_adj will still be oom disabled
1122 * with this implementation.
1124 * oom_adj cannot be removed since existing userspace binaries use it.
1126 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1127 size_t count
, loff_t
*ppos
)
1129 char buffer
[PROC_NUMBUF
];
1133 memset(buffer
, 0, sizeof(buffer
));
1134 if (count
> sizeof(buffer
) - 1)
1135 count
= sizeof(buffer
) - 1;
1136 if (copy_from_user(buffer
, buf
, count
)) {
1141 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1144 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1145 oom_adj
!= OOM_DISABLE
) {
1151 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1152 * value is always attainable.
1154 if (oom_adj
== OOM_ADJUST_MAX
)
1155 oom_adj
= OOM_SCORE_ADJ_MAX
;
1157 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1159 err
= __set_oom_adj(file
, oom_adj
, true);
1161 return err
< 0 ? err
: count
;
1164 static const struct file_operations proc_oom_adj_operations
= {
1165 .read
= oom_adj_read
,
1166 .write
= oom_adj_write
,
1167 .llseek
= generic_file_llseek
,
1170 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1171 size_t count
, loff_t
*ppos
)
1173 struct task_struct
*task
= get_proc_task(file_inode(file
));
1174 char buffer
[PROC_NUMBUF
];
1175 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1180 oom_score_adj
= task
->signal
->oom_score_adj
;
1181 put_task_struct(task
);
1182 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1183 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1186 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1187 size_t count
, loff_t
*ppos
)
1189 char buffer
[PROC_NUMBUF
];
1193 memset(buffer
, 0, sizeof(buffer
));
1194 if (count
> sizeof(buffer
) - 1)
1195 count
= sizeof(buffer
) - 1;
1196 if (copy_from_user(buffer
, buf
, count
)) {
1201 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1204 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1205 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1210 err
= __set_oom_adj(file
, oom_score_adj
, false);
1212 return err
< 0 ? err
: count
;
1215 static const struct file_operations proc_oom_score_adj_operations
= {
1216 .read
= oom_score_adj_read
,
1217 .write
= oom_score_adj_write
,
1218 .llseek
= default_llseek
,
1221 #ifdef CONFIG_AUDITSYSCALL
1222 #define TMPBUFLEN 11
1223 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1224 size_t count
, loff_t
*ppos
)
1226 struct inode
* inode
= file_inode(file
);
1227 struct task_struct
*task
= get_proc_task(inode
);
1229 char tmpbuf
[TMPBUFLEN
];
1233 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1234 from_kuid(file
->f_cred
->user_ns
,
1235 audit_get_loginuid(task
)));
1236 put_task_struct(task
);
1237 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1240 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1241 size_t count
, loff_t
*ppos
)
1243 struct inode
* inode
= file_inode(file
);
1249 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1256 /* No partial writes. */
1260 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1264 /* is userspace tring to explicitly UNSET the loginuid? */
1265 if (loginuid
== AUDIT_UID_UNSET
) {
1266 kloginuid
= INVALID_UID
;
1268 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1269 if (!uid_valid(kloginuid
))
1273 rv
= audit_set_loginuid(kloginuid
);
1279 static const struct file_operations proc_loginuid_operations
= {
1280 .read
= proc_loginuid_read
,
1281 .write
= proc_loginuid_write
,
1282 .llseek
= generic_file_llseek
,
1285 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1286 size_t count
, loff_t
*ppos
)
1288 struct inode
* inode
= file_inode(file
);
1289 struct task_struct
*task
= get_proc_task(inode
);
1291 char tmpbuf
[TMPBUFLEN
];
1295 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1296 audit_get_sessionid(task
));
1297 put_task_struct(task
);
1298 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1301 static const struct file_operations proc_sessionid_operations
= {
1302 .read
= proc_sessionid_read
,
1303 .llseek
= generic_file_llseek
,
1307 #ifdef CONFIG_FAULT_INJECTION
1308 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1309 size_t count
, loff_t
*ppos
)
1311 struct task_struct
*task
= get_proc_task(file_inode(file
));
1312 char buffer
[PROC_NUMBUF
];
1318 make_it_fail
= task
->make_it_fail
;
1319 put_task_struct(task
);
1321 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1323 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1326 static ssize_t
proc_fault_inject_write(struct file
* file
,
1327 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1329 struct task_struct
*task
;
1330 char buffer
[PROC_NUMBUF
];
1334 if (!capable(CAP_SYS_RESOURCE
))
1336 memset(buffer
, 0, sizeof(buffer
));
1337 if (count
> sizeof(buffer
) - 1)
1338 count
= sizeof(buffer
) - 1;
1339 if (copy_from_user(buffer
, buf
, count
))
1341 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1344 if (make_it_fail
< 0 || make_it_fail
> 1)
1347 task
= get_proc_task(file_inode(file
));
1350 task
->make_it_fail
= make_it_fail
;
1351 put_task_struct(task
);
1356 static const struct file_operations proc_fault_inject_operations
= {
1357 .read
= proc_fault_inject_read
,
1358 .write
= proc_fault_inject_write
,
1359 .llseek
= generic_file_llseek
,
1362 static ssize_t
proc_fail_nth_write(struct file
*file
, const char __user
*buf
,
1363 size_t count
, loff_t
*ppos
)
1365 struct task_struct
*task
;
1369 err
= kstrtouint_from_user(buf
, count
, 0, &n
);
1373 task
= get_proc_task(file_inode(file
));
1377 put_task_struct(task
);
1382 static ssize_t
proc_fail_nth_read(struct file
*file
, char __user
*buf
,
1383 size_t count
, loff_t
*ppos
)
1385 struct task_struct
*task
;
1386 char numbuf
[PROC_NUMBUF
];
1389 task
= get_proc_task(file_inode(file
));
1392 len
= snprintf(numbuf
, sizeof(numbuf
), "%u\n", task
->fail_nth
);
1393 len
= simple_read_from_buffer(buf
, count
, ppos
, numbuf
, len
);
1394 put_task_struct(task
);
1399 static const struct file_operations proc_fail_nth_operations
= {
1400 .read
= proc_fail_nth_read
,
1401 .write
= proc_fail_nth_write
,
1406 #ifdef CONFIG_SCHED_DEBUG
1408 * Print out various scheduling related per-task fields:
1410 static int sched_show(struct seq_file
*m
, void *v
)
1412 struct inode
*inode
= m
->private;
1413 struct pid_namespace
*ns
= inode
->i_sb
->s_fs_info
;
1414 struct task_struct
*p
;
1416 p
= get_proc_task(inode
);
1419 proc_sched_show_task(p
, ns
, m
);
1427 sched_write(struct file
*file
, const char __user
*buf
,
1428 size_t count
, loff_t
*offset
)
1430 struct inode
*inode
= file_inode(file
);
1431 struct task_struct
*p
;
1433 p
= get_proc_task(inode
);
1436 proc_sched_set_task(p
);
1443 static int sched_open(struct inode
*inode
, struct file
*filp
)
1445 return single_open(filp
, sched_show
, inode
);
1448 static const struct file_operations proc_pid_sched_operations
= {
1451 .write
= sched_write
,
1452 .llseek
= seq_lseek
,
1453 .release
= single_release
,
1458 #ifdef CONFIG_SCHED_AUTOGROUP
1460 * Print out autogroup related information:
1462 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1464 struct inode
*inode
= m
->private;
1465 struct task_struct
*p
;
1467 p
= get_proc_task(inode
);
1470 proc_sched_autogroup_show_task(p
, m
);
1478 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1479 size_t count
, loff_t
*offset
)
1481 struct inode
*inode
= file_inode(file
);
1482 struct task_struct
*p
;
1483 char buffer
[PROC_NUMBUF
];
1487 memset(buffer
, 0, sizeof(buffer
));
1488 if (count
> sizeof(buffer
) - 1)
1489 count
= sizeof(buffer
) - 1;
1490 if (copy_from_user(buffer
, buf
, count
))
1493 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1497 p
= get_proc_task(inode
);
1501 err
= proc_sched_autogroup_set_nice(p
, nice
);
1510 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1514 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1516 struct seq_file
*m
= filp
->private_data
;
1523 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1524 .open
= sched_autogroup_open
,
1526 .write
= sched_autogroup_write
,
1527 .llseek
= seq_lseek
,
1528 .release
= single_release
,
1531 #endif /* CONFIG_SCHED_AUTOGROUP */
1533 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1534 size_t count
, loff_t
*offset
)
1536 struct inode
*inode
= file_inode(file
);
1537 struct task_struct
*p
;
1538 char buffer
[TASK_COMM_LEN
];
1539 const size_t maxlen
= sizeof(buffer
) - 1;
1541 memset(buffer
, 0, sizeof(buffer
));
1542 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1545 p
= get_proc_task(inode
);
1549 if (same_thread_group(current
, p
))
1550 set_task_comm(p
, buffer
);
1559 static int comm_show(struct seq_file
*m
, void *v
)
1561 struct inode
*inode
= m
->private;
1562 struct task_struct
*p
;
1564 p
= get_proc_task(inode
);
1569 seq_printf(m
, "%s\n", p
->comm
);
1577 static int comm_open(struct inode
*inode
, struct file
*filp
)
1579 return single_open(filp
, comm_show
, inode
);
1582 static const struct file_operations proc_pid_set_comm_operations
= {
1585 .write
= comm_write
,
1586 .llseek
= seq_lseek
,
1587 .release
= single_release
,
1590 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1592 struct task_struct
*task
;
1593 struct file
*exe_file
;
1595 task
= get_proc_task(d_inode(dentry
));
1598 exe_file
= get_task_exe_file(task
);
1599 put_task_struct(task
);
1601 *exe_path
= exe_file
->f_path
;
1602 path_get(&exe_file
->f_path
);
1609 static const char *proc_pid_get_link(struct dentry
*dentry
,
1610 struct inode
*inode
,
1611 struct delayed_call
*done
)
1614 int error
= -EACCES
;
1617 return ERR_PTR(-ECHILD
);
1619 /* Are we allowed to snoop on the tasks file descriptors? */
1620 if (!proc_fd_access_allowed(inode
))
1623 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1627 nd_jump_link(&path
);
1630 return ERR_PTR(error
);
1633 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1635 char *tmp
= (char *)__get_free_page(GFP_KERNEL
);
1642 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1643 len
= PTR_ERR(pathname
);
1644 if (IS_ERR(pathname
))
1646 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1650 if (copy_to_user(buffer
, pathname
, len
))
1653 free_page((unsigned long)tmp
);
1657 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1659 int error
= -EACCES
;
1660 struct inode
*inode
= d_inode(dentry
);
1663 /* Are we allowed to snoop on the tasks file descriptors? */
1664 if (!proc_fd_access_allowed(inode
))
1667 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1671 error
= do_proc_readlink(&path
, buffer
, buflen
);
1677 const struct inode_operations proc_pid_link_inode_operations
= {
1678 .readlink
= proc_pid_readlink
,
1679 .get_link
= proc_pid_get_link
,
1680 .setattr
= proc_setattr
,
1684 /* building an inode */
1686 void task_dump_owner(struct task_struct
*task
, umode_t mode
,
1687 kuid_t
*ruid
, kgid_t
*rgid
)
1689 /* Depending on the state of dumpable compute who should own a
1690 * proc file for a task.
1692 const struct cred
*cred
;
1696 if (unlikely(task
->flags
& PF_KTHREAD
)) {
1697 *ruid
= GLOBAL_ROOT_UID
;
1698 *rgid
= GLOBAL_ROOT_GID
;
1702 /* Default to the tasks effective ownership */
1704 cred
= __task_cred(task
);
1710 * Before the /proc/pid/status file was created the only way to read
1711 * the effective uid of a /process was to stat /proc/pid. Reading
1712 * /proc/pid/status is slow enough that procps and other packages
1713 * kept stating /proc/pid. To keep the rules in /proc simple I have
1714 * made this apply to all per process world readable and executable
1717 if (mode
!= (S_IFDIR
|S_IRUGO
|S_IXUGO
)) {
1718 struct mm_struct
*mm
;
1721 /* Make non-dumpable tasks owned by some root */
1723 if (get_dumpable(mm
) != SUID_DUMP_USER
) {
1724 struct user_namespace
*user_ns
= mm
->user_ns
;
1726 uid
= make_kuid(user_ns
, 0);
1727 if (!uid_valid(uid
))
1728 uid
= GLOBAL_ROOT_UID
;
1730 gid
= make_kgid(user_ns
, 0);
1731 if (!gid_valid(gid
))
1732 gid
= GLOBAL_ROOT_GID
;
1735 uid
= GLOBAL_ROOT_UID
;
1736 gid
= GLOBAL_ROOT_GID
;
1744 struct inode
*proc_pid_make_inode(struct super_block
* sb
,
1745 struct task_struct
*task
, umode_t mode
)
1747 struct inode
* inode
;
1748 struct proc_inode
*ei
;
1750 /* We need a new inode */
1752 inode
= new_inode(sb
);
1758 inode
->i_mode
= mode
;
1759 inode
->i_ino
= get_next_ino();
1760 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= current_time(inode
);
1761 inode
->i_op
= &proc_def_inode_operations
;
1764 * grab the reference to task.
1766 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1770 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1771 security_task_to_inode(task
, inode
);
1781 int pid_getattr(const struct path
*path
, struct kstat
*stat
,
1782 u32 request_mask
, unsigned int query_flags
)
1784 struct inode
*inode
= d_inode(path
->dentry
);
1785 struct task_struct
*task
;
1786 struct pid_namespace
*pid
= path
->dentry
->d_sb
->s_fs_info
;
1788 generic_fillattr(inode
, stat
);
1791 stat
->uid
= GLOBAL_ROOT_UID
;
1792 stat
->gid
= GLOBAL_ROOT_GID
;
1793 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1795 if (!has_pid_permissions(pid
, task
, HIDEPID_INVISIBLE
)) {
1798 * This doesn't prevent learning whether PID exists,
1799 * it only makes getattr() consistent with readdir().
1803 task_dump_owner(task
, inode
->i_mode
, &stat
->uid
, &stat
->gid
);
1812 * Exceptional case: normally we are not allowed to unhash a busy
1813 * directory. In this case, however, we can do it - no aliasing problems
1814 * due to the way we treat inodes.
1816 * Rewrite the inode's ownerships here because the owning task may have
1817 * performed a setuid(), etc.
1820 int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1822 struct inode
*inode
;
1823 struct task_struct
*task
;
1825 if (flags
& LOOKUP_RCU
)
1828 inode
= d_inode(dentry
);
1829 task
= get_proc_task(inode
);
1832 task_dump_owner(task
, inode
->i_mode
, &inode
->i_uid
, &inode
->i_gid
);
1834 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1835 security_task_to_inode(task
, inode
);
1836 put_task_struct(task
);
1842 static inline bool proc_inode_is_dead(struct inode
*inode
)
1844 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
1847 int pid_delete_dentry(const struct dentry
*dentry
)
1849 /* Is the task we represent dead?
1850 * If so, then don't put the dentry on the lru list,
1851 * kill it immediately.
1853 return proc_inode_is_dead(d_inode(dentry
));
1856 const struct dentry_operations pid_dentry_operations
=
1858 .d_revalidate
= pid_revalidate
,
1859 .d_delete
= pid_delete_dentry
,
1865 * Fill a directory entry.
1867 * If possible create the dcache entry and derive our inode number and
1868 * file type from dcache entry.
1870 * Since all of the proc inode numbers are dynamically generated, the inode
1871 * numbers do not exist until the inode is cache. This means creating the
1872 * the dcache entry in readdir is necessary to keep the inode numbers
1873 * reported by readdir in sync with the inode numbers reported
1876 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
1877 const char *name
, int len
,
1878 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1880 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
1881 struct qstr qname
= QSTR_INIT(name
, len
);
1882 struct inode
*inode
;
1886 child
= d_hash_and_lookup(dir
, &qname
);
1888 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
1889 child
= d_alloc_parallel(dir
, &qname
, &wq
);
1891 goto end_instantiate
;
1892 if (d_in_lookup(child
)) {
1893 int err
= instantiate(d_inode(dir
), child
, task
, ptr
);
1894 d_lookup_done(child
);
1897 goto end_instantiate
;
1901 inode
= d_inode(child
);
1903 type
= inode
->i_mode
>> 12;
1905 return dir_emit(ctx
, name
, len
, ino
, type
);
1908 return dir_emit(ctx
, name
, len
, 1, DT_UNKNOWN
);
1912 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1913 * which represent vma start and end addresses.
1915 static int dname_to_vma_addr(struct dentry
*dentry
,
1916 unsigned long *start
, unsigned long *end
)
1918 const char *str
= dentry
->d_name
.name
;
1919 unsigned long long sval
, eval
;
1922 if (str
[0] == '0' && str
[1] != '-')
1924 len
= _parse_integer(str
, 16, &sval
);
1925 if (len
& KSTRTOX_OVERFLOW
)
1927 if (sval
!= (unsigned long)sval
)
1935 if (str
[0] == '0' && str
[1])
1937 len
= _parse_integer(str
, 16, &eval
);
1938 if (len
& KSTRTOX_OVERFLOW
)
1940 if (eval
!= (unsigned long)eval
)
1953 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
1955 unsigned long vm_start
, vm_end
;
1956 bool exact_vma_exists
= false;
1957 struct mm_struct
*mm
= NULL
;
1958 struct task_struct
*task
;
1959 struct inode
*inode
;
1962 if (flags
& LOOKUP_RCU
)
1965 inode
= d_inode(dentry
);
1966 task
= get_proc_task(inode
);
1970 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
1971 if (IS_ERR_OR_NULL(mm
))
1974 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
1975 down_read(&mm
->mmap_sem
);
1976 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
1977 up_read(&mm
->mmap_sem
);
1982 if (exact_vma_exists
) {
1983 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1985 security_task_to_inode(task
, inode
);
1990 put_task_struct(task
);
1996 static const struct dentry_operations tid_map_files_dentry_operations
= {
1997 .d_revalidate
= map_files_d_revalidate
,
1998 .d_delete
= pid_delete_dentry
,
2001 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2003 unsigned long vm_start
, vm_end
;
2004 struct vm_area_struct
*vma
;
2005 struct task_struct
*task
;
2006 struct mm_struct
*mm
;
2010 task
= get_proc_task(d_inode(dentry
));
2014 mm
= get_task_mm(task
);
2015 put_task_struct(task
);
2019 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2024 down_read(&mm
->mmap_sem
);
2025 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2026 if (vma
&& vma
->vm_file
) {
2027 *path
= vma
->vm_file
->f_path
;
2031 up_read(&mm
->mmap_sem
);
2039 struct map_files_info
{
2040 unsigned long start
;
2046 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2047 * symlinks may be used to bypass permissions on ancestor directories in the
2048 * path to the file in question.
2051 proc_map_files_get_link(struct dentry
*dentry
,
2052 struct inode
*inode
,
2053 struct delayed_call
*done
)
2055 if (!capable(CAP_SYS_ADMIN
))
2056 return ERR_PTR(-EPERM
);
2058 return proc_pid_get_link(dentry
, inode
, done
);
2062 * Identical to proc_pid_link_inode_operations except for get_link()
2064 static const struct inode_operations proc_map_files_link_inode_operations
= {
2065 .readlink
= proc_pid_readlink
,
2066 .get_link
= proc_map_files_get_link
,
2067 .setattr
= proc_setattr
,
2071 proc_map_files_instantiate(struct inode
*dir
, struct dentry
*dentry
,
2072 struct task_struct
*task
, const void *ptr
)
2074 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
2075 struct proc_inode
*ei
;
2076 struct inode
*inode
;
2078 inode
= proc_pid_make_inode(dir
->i_sb
, task
, S_IFLNK
|
2079 ((mode
& FMODE_READ
) ? S_IRUSR
: 0) |
2080 ((mode
& FMODE_WRITE
) ? S_IWUSR
: 0));
2085 ei
->op
.proc_get_link
= map_files_get_link
;
2087 inode
->i_op
= &proc_map_files_link_inode_operations
;
2090 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2091 d_add(dentry
, inode
);
2096 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2097 struct dentry
*dentry
, unsigned int flags
)
2099 unsigned long vm_start
, vm_end
;
2100 struct vm_area_struct
*vma
;
2101 struct task_struct
*task
;
2103 struct mm_struct
*mm
;
2106 task
= get_proc_task(dir
);
2111 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2115 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2118 mm
= get_task_mm(task
);
2122 down_read(&mm
->mmap_sem
);
2123 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2128 result
= proc_map_files_instantiate(dir
, dentry
, task
,
2129 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2132 up_read(&mm
->mmap_sem
);
2135 put_task_struct(task
);
2137 return ERR_PTR(result
);
2140 static const struct inode_operations proc_map_files_inode_operations
= {
2141 .lookup
= proc_map_files_lookup
,
2142 .permission
= proc_fd_permission
,
2143 .setattr
= proc_setattr
,
2147 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2149 struct vm_area_struct
*vma
;
2150 struct task_struct
*task
;
2151 struct mm_struct
*mm
;
2152 unsigned long nr_files
, pos
, i
;
2153 struct flex_array
*fa
= NULL
;
2154 struct map_files_info info
;
2155 struct map_files_info
*p
;
2159 task
= get_proc_task(file_inode(file
));
2164 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2168 if (!dir_emit_dots(file
, ctx
))
2171 mm
= get_task_mm(task
);
2174 down_read(&mm
->mmap_sem
);
2179 * We need two passes here:
2181 * 1) Collect vmas of mapped files with mmap_sem taken
2182 * 2) Release mmap_sem and instantiate entries
2184 * otherwise we get lockdep complained, since filldir()
2185 * routine might require mmap_sem taken in might_fault().
2188 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2189 if (vma
->vm_file
&& ++pos
> ctx
->pos
)
2194 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2196 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2200 flex_array_free(fa
);
2201 up_read(&mm
->mmap_sem
);
2205 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2206 vma
= vma
->vm_next
) {
2209 if (++pos
<= ctx
->pos
)
2212 info
.start
= vma
->vm_start
;
2213 info
.end
= vma
->vm_end
;
2214 info
.mode
= vma
->vm_file
->f_mode
;
2215 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2219 up_read(&mm
->mmap_sem
);
2222 for (i
= 0; i
< nr_files
; i
++) {
2223 char buf
[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2226 p
= flex_array_get(fa
, i
);
2227 len
= snprintf(buf
, sizeof(buf
), "%lx-%lx", p
->start
, p
->end
);
2228 if (!proc_fill_cache(file
, ctx
,
2230 proc_map_files_instantiate
,
2232 (void *)(unsigned long)p
->mode
))
2237 flex_array_free(fa
);
2240 put_task_struct(task
);
2245 static const struct file_operations proc_map_files_operations
= {
2246 .read
= generic_read_dir
,
2247 .iterate_shared
= proc_map_files_readdir
,
2248 .llseek
= generic_file_llseek
,
2251 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2252 struct timers_private
{
2254 struct task_struct
*task
;
2255 struct sighand_struct
*sighand
;
2256 struct pid_namespace
*ns
;
2257 unsigned long flags
;
2260 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2262 struct timers_private
*tp
= m
->private;
2264 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2266 return ERR_PTR(-ESRCH
);
2268 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2270 return ERR_PTR(-ESRCH
);
2272 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2275 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2277 struct timers_private
*tp
= m
->private;
2278 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2281 static void timers_stop(struct seq_file
*m
, void *v
)
2283 struct timers_private
*tp
= m
->private;
2286 unlock_task_sighand(tp
->task
, &tp
->flags
);
2291 put_task_struct(tp
->task
);
2296 static int show_timer(struct seq_file
*m
, void *v
)
2298 struct k_itimer
*timer
;
2299 struct timers_private
*tp
= m
->private;
2301 static const char * const nstr
[] = {
2302 [SIGEV_SIGNAL
] = "signal",
2303 [SIGEV_NONE
] = "none",
2304 [SIGEV_THREAD
] = "thread",
2307 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2308 notify
= timer
->it_sigev_notify
;
2310 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2311 seq_printf(m
, "signal: %d/%px\n",
2312 timer
->sigq
->info
.si_signo
,
2313 timer
->sigq
->info
.si_value
.sival_ptr
);
2314 seq_printf(m
, "notify: %s/%s.%d\n",
2315 nstr
[notify
& ~SIGEV_THREAD_ID
],
2316 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2317 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2318 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2323 static const struct seq_operations proc_timers_seq_ops
= {
2324 .start
= timers_start
,
2325 .next
= timers_next
,
2326 .stop
= timers_stop
,
2330 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2332 struct timers_private
*tp
;
2334 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2335 sizeof(struct timers_private
));
2339 tp
->pid
= proc_pid(inode
);
2340 tp
->ns
= inode
->i_sb
->s_fs_info
;
2344 static const struct file_operations proc_timers_operations
= {
2345 .open
= proc_timers_open
,
2347 .llseek
= seq_lseek
,
2348 .release
= seq_release_private
,
2352 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2353 size_t count
, loff_t
*offset
)
2355 struct inode
*inode
= file_inode(file
);
2356 struct task_struct
*p
;
2360 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2364 p
= get_proc_task(inode
);
2369 if (!capable(CAP_SYS_NICE
)) {
2374 err
= security_task_setscheduler(p
);
2383 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2385 p
->timer_slack_ns
= slack_ns
;
2394 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2396 struct inode
*inode
= m
->private;
2397 struct task_struct
*p
;
2400 p
= get_proc_task(inode
);
2406 if (!capable(CAP_SYS_NICE
)) {
2410 err
= security_task_getscheduler(p
);
2416 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2425 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2427 return single_open(filp
, timerslack_ns_show
, inode
);
2430 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2431 .open
= timerslack_ns_open
,
2433 .write
= timerslack_ns_write
,
2434 .llseek
= seq_lseek
,
2435 .release
= single_release
,
2438 static int proc_pident_instantiate(struct inode
*dir
,
2439 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2441 const struct pid_entry
*p
= ptr
;
2442 struct inode
*inode
;
2443 struct proc_inode
*ei
;
2445 inode
= proc_pid_make_inode(dir
->i_sb
, task
, p
->mode
);
2450 if (S_ISDIR(inode
->i_mode
))
2451 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2453 inode
->i_op
= p
->iop
;
2455 inode
->i_fop
= p
->fop
;
2457 d_set_d_op(dentry
, &pid_dentry_operations
);
2458 d_add(dentry
, inode
);
2459 /* Close the race of the process dying before we return the dentry */
2460 if (pid_revalidate(dentry
, 0))
2466 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2467 struct dentry
*dentry
,
2468 const struct pid_entry
*ents
,
2472 struct task_struct
*task
= get_proc_task(dir
);
2473 const struct pid_entry
*p
, *last
;
2481 * Yes, it does not scale. And it should not. Don't add
2482 * new entries into /proc/<tgid>/ without very good reasons.
2484 last
= &ents
[nents
];
2485 for (p
= ents
; p
< last
; p
++) {
2486 if (p
->len
!= dentry
->d_name
.len
)
2488 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2494 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2496 put_task_struct(task
);
2498 return ERR_PTR(error
);
2501 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2502 const struct pid_entry
*ents
, unsigned int nents
)
2504 struct task_struct
*task
= get_proc_task(file_inode(file
));
2505 const struct pid_entry
*p
;
2510 if (!dir_emit_dots(file
, ctx
))
2513 if (ctx
->pos
>= nents
+ 2)
2516 for (p
= ents
+ (ctx
->pos
- 2); p
< ents
+ nents
; p
++) {
2517 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2518 proc_pident_instantiate
, task
, p
))
2523 put_task_struct(task
);
2527 #ifdef CONFIG_SECURITY
2528 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2529 size_t count
, loff_t
*ppos
)
2531 struct inode
* inode
= file_inode(file
);
2534 struct task_struct
*task
= get_proc_task(inode
);
2539 length
= security_getprocattr(task
,
2540 (char*)file
->f_path
.dentry
->d_name
.name
,
2542 put_task_struct(task
);
2544 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2549 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2550 size_t count
, loff_t
*ppos
)
2552 struct inode
* inode
= file_inode(file
);
2555 struct task_struct
*task
= get_proc_task(inode
);
2561 /* A task may only write its own attributes. */
2563 if (current
!= task
)
2566 if (count
> PAGE_SIZE
)
2569 /* No partial writes. */
2574 page
= memdup_user(buf
, count
);
2576 length
= PTR_ERR(page
);
2580 /* Guard against adverse ptrace interaction */
2581 length
= mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
);
2585 length
= security_setprocattr(file
->f_path
.dentry
->d_name
.name
,
2587 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
2591 put_task_struct(task
);
2596 static const struct file_operations proc_pid_attr_operations
= {
2597 .read
= proc_pid_attr_read
,
2598 .write
= proc_pid_attr_write
,
2599 .llseek
= generic_file_llseek
,
2602 static const struct pid_entry attr_dir_stuff
[] = {
2603 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2604 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2605 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2606 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2607 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2608 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2611 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2613 return proc_pident_readdir(file
, ctx
,
2614 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2617 static const struct file_operations proc_attr_dir_operations
= {
2618 .read
= generic_read_dir
,
2619 .iterate_shared
= proc_attr_dir_readdir
,
2620 .llseek
= generic_file_llseek
,
2623 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2624 struct dentry
*dentry
, unsigned int flags
)
2626 return proc_pident_lookup(dir
, dentry
,
2627 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2630 static const struct inode_operations proc_attr_dir_inode_operations
= {
2631 .lookup
= proc_attr_dir_lookup
,
2632 .getattr
= pid_getattr
,
2633 .setattr
= proc_setattr
,
2638 #ifdef CONFIG_ELF_CORE
2639 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2640 size_t count
, loff_t
*ppos
)
2642 struct task_struct
*task
= get_proc_task(file_inode(file
));
2643 struct mm_struct
*mm
;
2644 char buffer
[PROC_NUMBUF
];
2652 mm
= get_task_mm(task
);
2654 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2655 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2656 MMF_DUMP_FILTER_SHIFT
));
2658 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2661 put_task_struct(task
);
2666 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2667 const char __user
*buf
,
2671 struct task_struct
*task
;
2672 struct mm_struct
*mm
;
2678 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2683 task
= get_proc_task(file_inode(file
));
2687 mm
= get_task_mm(task
);
2692 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2694 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2696 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2701 put_task_struct(task
);
2708 static const struct file_operations proc_coredump_filter_operations
= {
2709 .read
= proc_coredump_filter_read
,
2710 .write
= proc_coredump_filter_write
,
2711 .llseek
= generic_file_llseek
,
2715 #ifdef CONFIG_TASK_IO_ACCOUNTING
2716 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2718 struct task_io_accounting acct
= task
->ioac
;
2719 unsigned long flags
;
2722 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2726 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2731 if (whole
&& lock_task_sighand(task
, &flags
)) {
2732 struct task_struct
*t
= task
;
2734 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2735 while_each_thread(task
, t
)
2736 task_io_accounting_add(&acct
, &t
->ioac
);
2738 unlock_task_sighand(task
, &flags
);
2745 "read_bytes: %llu\n"
2746 "write_bytes: %llu\n"
2747 "cancelled_write_bytes: %llu\n",
2748 (unsigned long long)acct
.rchar
,
2749 (unsigned long long)acct
.wchar
,
2750 (unsigned long long)acct
.syscr
,
2751 (unsigned long long)acct
.syscw
,
2752 (unsigned long long)acct
.read_bytes
,
2753 (unsigned long long)acct
.write_bytes
,
2754 (unsigned long long)acct
.cancelled_write_bytes
);
2758 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2762 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2763 struct pid
*pid
, struct task_struct
*task
)
2765 return do_io_accounting(task
, m
, 0);
2768 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2769 struct pid
*pid
, struct task_struct
*task
)
2771 return do_io_accounting(task
, m
, 1);
2773 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2775 #ifdef CONFIG_USER_NS
2776 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2777 const struct seq_operations
*seq_ops
)
2779 struct user_namespace
*ns
= NULL
;
2780 struct task_struct
*task
;
2781 struct seq_file
*seq
;
2784 task
= get_proc_task(inode
);
2787 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2789 put_task_struct(task
);
2794 ret
= seq_open(file
, seq_ops
);
2798 seq
= file
->private_data
;
2808 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2810 struct seq_file
*seq
= file
->private_data
;
2811 struct user_namespace
*ns
= seq
->private;
2813 return seq_release(inode
, file
);
2816 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2818 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2821 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2823 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2826 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
2828 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
2831 static const struct file_operations proc_uid_map_operations
= {
2832 .open
= proc_uid_map_open
,
2833 .write
= proc_uid_map_write
,
2835 .llseek
= seq_lseek
,
2836 .release
= proc_id_map_release
,
2839 static const struct file_operations proc_gid_map_operations
= {
2840 .open
= proc_gid_map_open
,
2841 .write
= proc_gid_map_write
,
2843 .llseek
= seq_lseek
,
2844 .release
= proc_id_map_release
,
2847 static const struct file_operations proc_projid_map_operations
= {
2848 .open
= proc_projid_map_open
,
2849 .write
= proc_projid_map_write
,
2851 .llseek
= seq_lseek
,
2852 .release
= proc_id_map_release
,
2855 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
2857 struct user_namespace
*ns
= NULL
;
2858 struct task_struct
*task
;
2862 task
= get_proc_task(inode
);
2865 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2867 put_task_struct(task
);
2872 if (file
->f_mode
& FMODE_WRITE
) {
2874 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
2878 ret
= single_open(file
, &proc_setgroups_show
, ns
);
2889 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
2891 struct seq_file
*seq
= file
->private_data
;
2892 struct user_namespace
*ns
= seq
->private;
2893 int ret
= single_release(inode
, file
);
2898 static const struct file_operations proc_setgroups_operations
= {
2899 .open
= proc_setgroups_open
,
2900 .write
= proc_setgroups_write
,
2902 .llseek
= seq_lseek
,
2903 .release
= proc_setgroups_release
,
2905 #endif /* CONFIG_USER_NS */
2907 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2908 struct pid
*pid
, struct task_struct
*task
)
2910 int err
= lock_trace(task
);
2912 seq_printf(m
, "%08x\n", task
->personality
);
2918 #ifdef CONFIG_LIVEPATCH
2919 static int proc_pid_patch_state(struct seq_file
*m
, struct pid_namespace
*ns
,
2920 struct pid
*pid
, struct task_struct
*task
)
2922 seq_printf(m
, "%d\n", task
->patch_state
);
2925 #endif /* CONFIG_LIVEPATCH */
2930 static const struct file_operations proc_task_operations
;
2931 static const struct inode_operations proc_task_inode_operations
;
2933 static const struct pid_entry tgid_base_stuff
[] = {
2934 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2935 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2936 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
2937 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2938 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
2940 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2942 REG("environ", S_IRUSR
, proc_environ_operations
),
2943 REG("auxv", S_IRUSR
, proc_auxv_operations
),
2944 ONE("status", S_IRUGO
, proc_pid_status
),
2945 ONE("personality", S_IRUSR
, proc_pid_personality
),
2946 ONE("limits", S_IRUGO
, proc_pid_limits
),
2947 #ifdef CONFIG_SCHED_DEBUG
2948 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2950 #ifdef CONFIG_SCHED_AUTOGROUP
2951 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
2953 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2954 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2955 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
2957 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
2958 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2959 ONE("statm", S_IRUGO
, proc_pid_statm
),
2960 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
2962 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
2964 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2965 LNK("cwd", proc_cwd_link
),
2966 LNK("root", proc_root_link
),
2967 LNK("exe", proc_exe_link
),
2968 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2969 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2970 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2971 #ifdef CONFIG_PROC_PAGE_MONITOR
2972 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2973 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
2974 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
2975 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2977 #ifdef CONFIG_SECURITY
2978 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2980 #ifdef CONFIG_KALLSYMS
2981 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
2983 #ifdef CONFIG_STACKTRACE
2984 ONE("stack", S_IRUSR
, proc_pid_stack
),
2986 #ifdef CONFIG_SCHED_INFO
2987 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
2989 #ifdef CONFIG_LATENCYTOP
2990 REG("latency", S_IRUGO
, proc_lstats_operations
),
2992 #ifdef CONFIG_PROC_PID_CPUSET
2993 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
2995 #ifdef CONFIG_CGROUPS
2996 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
2998 ONE("oom_score", S_IRUGO
, proc_oom_score
),
2999 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3000 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3001 #ifdef CONFIG_AUDITSYSCALL
3002 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3003 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3005 #ifdef CONFIG_FAULT_INJECTION
3006 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3007 REG("fail-nth", 0644, proc_fail_nth_operations
),
3009 #ifdef CONFIG_ELF_CORE
3010 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3012 #ifdef CONFIG_TASK_IO_ACCOUNTING
3013 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
3015 #ifdef CONFIG_USER_NS
3016 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3017 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3018 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3019 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3021 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3022 REG("timers", S_IRUGO
, proc_timers_operations
),
3024 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
3025 #ifdef CONFIG_LIVEPATCH
3026 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3030 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3032 return proc_pident_readdir(file
, ctx
,
3033 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3036 static const struct file_operations proc_tgid_base_operations
= {
3037 .read
= generic_read_dir
,
3038 .iterate_shared
= proc_tgid_base_readdir
,
3039 .llseek
= generic_file_llseek
,
3042 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3044 return proc_pident_lookup(dir
, dentry
,
3045 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3048 static const struct inode_operations proc_tgid_base_inode_operations
= {
3049 .lookup
= proc_tgid_base_lookup
,
3050 .getattr
= pid_getattr
,
3051 .setattr
= proc_setattr
,
3052 .permission
= proc_pid_permission
,
3055 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
3057 struct dentry
*dentry
, *leader
, *dir
;
3062 name
.len
= snprintf(buf
, sizeof(buf
), "%u", pid
);
3063 /* no ->d_hash() rejects on procfs */
3064 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3066 d_invalidate(dentry
);
3074 name
.len
= snprintf(buf
, sizeof(buf
), "%u", tgid
);
3075 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3080 name
.len
= strlen(name
.name
);
3081 dir
= d_hash_and_lookup(leader
, &name
);
3083 goto out_put_leader
;
3086 name
.len
= snprintf(buf
, sizeof(buf
), "%u", pid
);
3087 dentry
= d_hash_and_lookup(dir
, &name
);
3089 d_invalidate(dentry
);
3101 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3102 * @task: task that should be flushed.
3104 * When flushing dentries from proc, one needs to flush them from global
3105 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3106 * in. This call is supposed to do all of this job.
3108 * Looks in the dcache for
3110 * /proc/@tgid/task/@pid
3111 * if either directory is present flushes it and all of it'ts children
3114 * It is safe and reasonable to cache /proc entries for a task until
3115 * that task exits. After that they just clog up the dcache with
3116 * useless entries, possibly causing useful dcache entries to be
3117 * flushed instead. This routine is proved to flush those useless
3118 * dcache entries at process exit time.
3120 * NOTE: This routine is just an optimization so it does not guarantee
3121 * that no dcache entries will exist at process exit time it
3122 * just makes it very unlikely that any will persist.
3125 void proc_flush_task(struct task_struct
*task
)
3128 struct pid
*pid
, *tgid
;
3131 pid
= task_pid(task
);
3132 tgid
= task_tgid(task
);
3134 for (i
= 0; i
<= pid
->level
; i
++) {
3135 upid
= &pid
->numbers
[i
];
3136 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
3137 tgid
->numbers
[i
].nr
);
3141 static int proc_pid_instantiate(struct inode
*dir
,
3142 struct dentry
* dentry
,
3143 struct task_struct
*task
, const void *ptr
)
3145 struct inode
*inode
;
3147 inode
= proc_pid_make_inode(dir
->i_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3151 inode
->i_op
= &proc_tgid_base_inode_operations
;
3152 inode
->i_fop
= &proc_tgid_base_operations
;
3153 inode
->i_flags
|=S_IMMUTABLE
;
3155 set_nlink(inode
, nlink_tgid
);
3157 d_set_d_op(dentry
, &pid_dentry_operations
);
3159 d_add(dentry
, inode
);
3160 /* Close the race of the process dying before we return the dentry */
3161 if (pid_revalidate(dentry
, 0))
3167 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3169 int result
= -ENOENT
;
3170 struct task_struct
*task
;
3172 struct pid_namespace
*ns
;
3174 tgid
= name_to_int(&dentry
->d_name
);
3178 ns
= dentry
->d_sb
->s_fs_info
;
3180 task
= find_task_by_pid_ns(tgid
, ns
);
3182 get_task_struct(task
);
3187 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
3188 put_task_struct(task
);
3190 return ERR_PTR(result
);
3194 * Find the first task with tgid >= tgid
3199 struct task_struct
*task
;
3201 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3206 put_task_struct(iter
.task
);
3210 pid
= find_ge_pid(iter
.tgid
, ns
);
3212 iter
.tgid
= pid_nr_ns(pid
, ns
);
3213 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3214 /* What we to know is if the pid we have find is the
3215 * pid of a thread_group_leader. Testing for task
3216 * being a thread_group_leader is the obvious thing
3217 * todo but there is a window when it fails, due to
3218 * the pid transfer logic in de_thread.
3220 * So we perform the straight forward test of seeing
3221 * if the pid we have found is the pid of a thread
3222 * group leader, and don't worry if the task we have
3223 * found doesn't happen to be a thread group leader.
3224 * As we don't care in the case of readdir.
3226 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3230 get_task_struct(iter
.task
);
3236 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3238 /* for the /proc/ directory itself, after non-process stuff has been done */
3239 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3241 struct tgid_iter iter
;
3242 struct pid_namespace
*ns
= file_inode(file
)->i_sb
->s_fs_info
;
3243 loff_t pos
= ctx
->pos
;
3245 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3248 if (pos
== TGID_OFFSET
- 2) {
3249 struct inode
*inode
= d_inode(ns
->proc_self
);
3250 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3252 ctx
->pos
= pos
= pos
+ 1;
3254 if (pos
== TGID_OFFSET
- 1) {
3255 struct inode
*inode
= d_inode(ns
->proc_thread_self
);
3256 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3258 ctx
->pos
= pos
= pos
+ 1;
3260 iter
.tgid
= pos
- TGID_OFFSET
;
3262 for (iter
= next_tgid(ns
, iter
);
3264 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3269 if (!has_pid_permissions(ns
, iter
.task
, HIDEPID_INVISIBLE
))
3272 len
= snprintf(name
, sizeof(name
), "%u", iter
.tgid
);
3273 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3274 if (!proc_fill_cache(file
, ctx
, name
, len
,
3275 proc_pid_instantiate
, iter
.task
, NULL
)) {
3276 put_task_struct(iter
.task
);
3280 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3285 * proc_tid_comm_permission is a special permission function exclusively
3286 * used for the node /proc/<pid>/task/<tid>/comm.
3287 * It bypasses generic permission checks in the case where a task of the same
3288 * task group attempts to access the node.
3289 * The rationale behind this is that glibc and bionic access this node for
3290 * cross thread naming (pthread_set/getname_np(!self)). However, if
3291 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3292 * which locks out the cross thread naming implementation.
3293 * This function makes sure that the node is always accessible for members of
3294 * same thread group.
3296 static int proc_tid_comm_permission(struct inode
*inode
, int mask
)
3298 bool is_same_tgroup
;
3299 struct task_struct
*task
;
3301 task
= get_proc_task(inode
);
3304 is_same_tgroup
= same_thread_group(current
, task
);
3305 put_task_struct(task
);
3307 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3308 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3309 * read or written by the members of the corresponding
3315 return generic_permission(inode
, mask
);
3318 static const struct inode_operations proc_tid_comm_inode_operations
= {
3319 .permission
= proc_tid_comm_permission
,
3325 static const struct pid_entry tid_base_stuff
[] = {
3326 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3327 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3328 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3330 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3332 REG("environ", S_IRUSR
, proc_environ_operations
),
3333 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3334 ONE("status", S_IRUGO
, proc_pid_status
),
3335 ONE("personality", S_IRUSR
, proc_pid_personality
),
3336 ONE("limits", S_IRUGO
, proc_pid_limits
),
3337 #ifdef CONFIG_SCHED_DEBUG
3338 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3340 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3341 &proc_tid_comm_inode_operations
,
3342 &proc_pid_set_comm_operations
, {}),
3343 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3344 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3346 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3347 ONE("stat", S_IRUGO
, proc_tid_stat
),
3348 ONE("statm", S_IRUGO
, proc_pid_statm
),
3349 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3350 #ifdef CONFIG_PROC_CHILDREN
3351 REG("children", S_IRUGO
, proc_tid_children_operations
),
3354 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3356 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3357 LNK("cwd", proc_cwd_link
),
3358 LNK("root", proc_root_link
),
3359 LNK("exe", proc_exe_link
),
3360 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3361 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3362 #ifdef CONFIG_PROC_PAGE_MONITOR
3363 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3364 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3365 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3366 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3368 #ifdef CONFIG_SECURITY
3369 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3371 #ifdef CONFIG_KALLSYMS
3372 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3374 #ifdef CONFIG_STACKTRACE
3375 ONE("stack", S_IRUSR
, proc_pid_stack
),
3377 #ifdef CONFIG_SCHED_INFO
3378 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3380 #ifdef CONFIG_LATENCYTOP
3381 REG("latency", S_IRUGO
, proc_lstats_operations
),
3383 #ifdef CONFIG_PROC_PID_CPUSET
3384 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3386 #ifdef CONFIG_CGROUPS
3387 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3389 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3390 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3391 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3392 #ifdef CONFIG_AUDITSYSCALL
3393 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3394 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3396 #ifdef CONFIG_FAULT_INJECTION
3397 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3398 REG("fail-nth", 0644, proc_fail_nth_operations
),
3400 #ifdef CONFIG_TASK_IO_ACCOUNTING
3401 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3403 #ifdef CONFIG_USER_NS
3404 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3405 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3406 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3407 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3409 #ifdef CONFIG_LIVEPATCH
3410 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3414 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3416 return proc_pident_readdir(file
, ctx
,
3417 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3420 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3422 return proc_pident_lookup(dir
, dentry
,
3423 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3426 static const struct file_operations proc_tid_base_operations
= {
3427 .read
= generic_read_dir
,
3428 .iterate_shared
= proc_tid_base_readdir
,
3429 .llseek
= generic_file_llseek
,
3432 static const struct inode_operations proc_tid_base_inode_operations
= {
3433 .lookup
= proc_tid_base_lookup
,
3434 .getattr
= pid_getattr
,
3435 .setattr
= proc_setattr
,
3438 static int proc_task_instantiate(struct inode
*dir
,
3439 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3441 struct inode
*inode
;
3442 inode
= proc_pid_make_inode(dir
->i_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3446 inode
->i_op
= &proc_tid_base_inode_operations
;
3447 inode
->i_fop
= &proc_tid_base_operations
;
3448 inode
->i_flags
|=S_IMMUTABLE
;
3450 set_nlink(inode
, nlink_tid
);
3452 d_set_d_op(dentry
, &pid_dentry_operations
);
3454 d_add(dentry
, inode
);
3455 /* Close the race of the process dying before we return the dentry */
3456 if (pid_revalidate(dentry
, 0))
3462 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3464 int result
= -ENOENT
;
3465 struct task_struct
*task
;
3466 struct task_struct
*leader
= get_proc_task(dir
);
3468 struct pid_namespace
*ns
;
3473 tid
= name_to_int(&dentry
->d_name
);
3477 ns
= dentry
->d_sb
->s_fs_info
;
3479 task
= find_task_by_pid_ns(tid
, ns
);
3481 get_task_struct(task
);
3485 if (!same_thread_group(leader
, task
))
3488 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3490 put_task_struct(task
);
3492 put_task_struct(leader
);
3494 return ERR_PTR(result
);
3498 * Find the first tid of a thread group to return to user space.
3500 * Usually this is just the thread group leader, but if the users
3501 * buffer was too small or there was a seek into the middle of the
3502 * directory we have more work todo.
3504 * In the case of a short read we start with find_task_by_pid.
3506 * In the case of a seek we start with the leader and walk nr
3509 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3510 struct pid_namespace
*ns
)
3512 struct task_struct
*pos
, *task
;
3513 unsigned long nr
= f_pos
;
3515 if (nr
!= f_pos
) /* 32bit overflow? */
3519 task
= pid_task(pid
, PIDTYPE_PID
);
3523 /* Attempt to start with the tid of a thread */
3525 pos
= find_task_by_pid_ns(tid
, ns
);
3526 if (pos
&& same_thread_group(pos
, task
))
3530 /* If nr exceeds the number of threads there is nothing todo */
3531 if (nr
>= get_nr_threads(task
))
3534 /* If we haven't found our starting place yet start
3535 * with the leader and walk nr threads forward.
3537 pos
= task
= task
->group_leader
;
3541 } while_each_thread(task
, pos
);
3546 get_task_struct(pos
);
3553 * Find the next thread in the thread list.
3554 * Return NULL if there is an error or no next thread.
3556 * The reference to the input task_struct is released.
3558 static struct task_struct
*next_tid(struct task_struct
*start
)
3560 struct task_struct
*pos
= NULL
;
3562 if (pid_alive(start
)) {
3563 pos
= next_thread(start
);
3564 if (thread_group_leader(pos
))
3567 get_task_struct(pos
);
3570 put_task_struct(start
);
3574 /* for the /proc/TGID/task/ directories */
3575 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3577 struct inode
*inode
= file_inode(file
);
3578 struct task_struct
*task
;
3579 struct pid_namespace
*ns
;
3582 if (proc_inode_is_dead(inode
))
3585 if (!dir_emit_dots(file
, ctx
))
3588 /* f_version caches the tgid value that the last readdir call couldn't
3589 * return. lseek aka telldir automagically resets f_version to 0.
3591 ns
= inode
->i_sb
->s_fs_info
;
3592 tid
= (int)file
->f_version
;
3593 file
->f_version
= 0;
3594 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3596 task
= next_tid(task
), ctx
->pos
++) {
3599 tid
= task_pid_nr_ns(task
, ns
);
3600 len
= snprintf(name
, sizeof(name
), "%u", tid
);
3601 if (!proc_fill_cache(file
, ctx
, name
, len
,
3602 proc_task_instantiate
, task
, NULL
)) {
3603 /* returning this tgid failed, save it as the first
3604 * pid for the next readir call */
3605 file
->f_version
= (u64
)tid
;
3606 put_task_struct(task
);
3614 static int proc_task_getattr(const struct path
*path
, struct kstat
*stat
,
3615 u32 request_mask
, unsigned int query_flags
)
3617 struct inode
*inode
= d_inode(path
->dentry
);
3618 struct task_struct
*p
= get_proc_task(inode
);
3619 generic_fillattr(inode
, stat
);
3622 stat
->nlink
+= get_nr_threads(p
);
3629 static const struct inode_operations proc_task_inode_operations
= {
3630 .lookup
= proc_task_lookup
,
3631 .getattr
= proc_task_getattr
,
3632 .setattr
= proc_setattr
,
3633 .permission
= proc_pid_permission
,
3636 static const struct file_operations proc_task_operations
= {
3637 .read
= generic_read_dir
,
3638 .iterate_shared
= proc_task_readdir
,
3639 .llseek
= generic_file_llseek
,
3642 void __init
set_proc_pid_nlink(void)
3644 nlink_tid
= pid_entry_nlink(tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3645 nlink_tgid
= pid_entry_nlink(tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));