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/generic-radix-tree.h>
63 #include <linux/string.h>
64 #include <linux/seq_file.h>
65 #include <linux/namei.h>
66 #include <linux/mnt_namespace.h>
68 #include <linux/swap.h>
69 #include <linux/rcupdate.h>
70 #include <linux/kallsyms.h>
71 #include <linux/stacktrace.h>
72 #include <linux/resource.h>
73 #include <linux/module.h>
74 #include <linux/mount.h>
75 #include <linux/security.h>
76 #include <linux/ptrace.h>
77 #include <linux/tracehook.h>
78 #include <linux/printk.h>
79 #include <linux/cache.h>
80 #include <linux/cgroup.h>
81 #include <linux/cpuset.h>
82 #include <linux/audit.h>
83 #include <linux/poll.h>
84 #include <linux/nsproxy.h>
85 #include <linux/oom.h>
86 #include <linux/elf.h>
87 #include <linux/pid_namespace.h>
88 #include <linux/user_namespace.h>
89 #include <linux/fs_struct.h>
90 #include <linux/slab.h>
91 #include <linux/sched/autogroup.h>
92 #include <linux/sched/mm.h>
93 #include <linux/sched/coredump.h>
94 #include <linux/sched/debug.h>
95 #include <linux/sched/stat.h>
96 #include <linux/posix-timers.h>
97 #include <linux/time_namespace.h>
98 #include <linux/resctrl.h>
99 #include <trace/events/oom.h>
100 #include "internal.h"
103 #include "../../lib/kstrtox.h"
106 * Implementing inode permission operations in /proc is almost
107 * certainly an error. Permission checks need to happen during
108 * each system call not at open time. The reason is that most of
109 * what we wish to check for permissions in /proc varies at runtime.
111 * The classic example of a problem is opening file descriptors
112 * in /proc for a task before it execs a suid executable.
115 static u8 nlink_tid __ro_after_init
;
116 static u8 nlink_tgid __ro_after_init
;
122 const struct inode_operations
*iop
;
123 const struct file_operations
*fop
;
127 #define NOD(NAME, MODE, IOP, FOP, OP) { \
129 .len = sizeof(NAME) - 1, \
136 #define DIR(NAME, MODE, iops, fops) \
137 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
138 #define LNK(NAME, get_link) \
139 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
140 &proc_pid_link_inode_operations, NULL, \
141 { .proc_get_link = get_link } )
142 #define REG(NAME, MODE, fops) \
143 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
144 #define ONE(NAME, MODE, show) \
145 NOD(NAME, (S_IFREG|(MODE)), \
146 NULL, &proc_single_file_operations, \
147 { .proc_show = show } )
148 #define ATTR(LSM, NAME, MODE) \
149 NOD(NAME, (S_IFREG|(MODE)), \
150 NULL, &proc_pid_attr_operations, \
154 * Count the number of hardlinks for the pid_entry table, excluding the .
157 static unsigned int __init
pid_entry_nlink(const struct pid_entry
*entries
,
164 for (i
= 0; i
< n
; ++i
) {
165 if (S_ISDIR(entries
[i
].mode
))
172 static int get_task_root(struct task_struct
*task
, struct path
*root
)
174 int result
= -ENOENT
;
178 get_fs_root(task
->fs
, root
);
185 static int proc_cwd_link(struct dentry
*dentry
, struct path
*path
)
187 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
188 int result
= -ENOENT
;
193 get_fs_pwd(task
->fs
, path
);
197 put_task_struct(task
);
202 static int proc_root_link(struct dentry
*dentry
, struct path
*path
)
204 struct task_struct
*task
= get_proc_task(d_inode(dentry
));
205 int result
= -ENOENT
;
208 result
= get_task_root(task
, path
);
209 put_task_struct(task
);
215 * If the user used setproctitle(), we just get the string from
216 * user space at arg_start, and limit it to a maximum of one page.
218 static ssize_t
get_mm_proctitle(struct mm_struct
*mm
, char __user
*buf
,
219 size_t count
, unsigned long pos
,
220 unsigned long arg_start
)
225 if (pos
>= PAGE_SIZE
)
228 page
= (char *)__get_free_page(GFP_KERNEL
);
233 got
= access_remote_vm(mm
, arg_start
, page
, PAGE_SIZE
, FOLL_ANON
);
235 int len
= strnlen(page
, got
);
237 /* Include the NUL character if it was found */
245 len
-= copy_to_user(buf
, page
+pos
, len
);
251 free_page((unsigned long)page
);
255 static ssize_t
get_mm_cmdline(struct mm_struct
*mm
, char __user
*buf
,
256 size_t count
, loff_t
*ppos
)
258 unsigned long arg_start
, arg_end
, env_start
, env_end
;
259 unsigned long pos
, len
;
262 /* Check if process spawned far enough to have cmdline. */
266 spin_lock(&mm
->arg_lock
);
267 arg_start
= mm
->arg_start
;
268 arg_end
= mm
->arg_end
;
269 env_start
= mm
->env_start
;
270 env_end
= mm
->env_end
;
271 spin_unlock(&mm
->arg_lock
);
273 if (arg_start
>= arg_end
)
277 * We allow setproctitle() to overwrite the argument
278 * strings, and overflow past the original end. But
279 * only when it overflows into the environment area.
281 if (env_start
!= arg_end
|| env_end
< env_start
)
282 env_start
= env_end
= arg_end
;
283 len
= env_end
- arg_start
;
285 /* We're not going to care if "*ppos" has high bits set */
289 if (count
> len
- pos
)
295 * Magical special case: if the argv[] end byte is not
296 * zero, the user has overwritten it with setproctitle(3).
298 * Possible future enhancement: do this only once when
299 * pos is 0, and set a flag in the 'struct file'.
301 if (access_remote_vm(mm
, arg_end
-1, &c
, 1, FOLL_ANON
) == 1 && c
)
302 return get_mm_proctitle(mm
, buf
, count
, pos
, arg_start
);
305 * For the non-setproctitle() case we limit things strictly
306 * to the [arg_start, arg_end[ range.
309 if (pos
< arg_start
|| pos
>= arg_end
)
311 if (count
> arg_end
- pos
)
312 count
= arg_end
- pos
;
314 page
= (char *)__get_free_page(GFP_KERNEL
);
321 size_t size
= min_t(size_t, PAGE_SIZE
, count
);
323 got
= access_remote_vm(mm
, pos
, page
, size
, FOLL_ANON
);
326 got
-= copy_to_user(buf
, page
, got
);
327 if (unlikely(!got
)) {
338 free_page((unsigned long)page
);
342 static ssize_t
get_task_cmdline(struct task_struct
*tsk
, char __user
*buf
,
343 size_t count
, loff_t
*pos
)
345 struct mm_struct
*mm
;
348 mm
= get_task_mm(tsk
);
352 ret
= get_mm_cmdline(mm
, buf
, count
, pos
);
357 static ssize_t
proc_pid_cmdline_read(struct file
*file
, char __user
*buf
,
358 size_t count
, loff_t
*pos
)
360 struct task_struct
*tsk
;
365 tsk
= get_proc_task(file_inode(file
));
368 ret
= get_task_cmdline(tsk
, buf
, count
, pos
);
369 put_task_struct(tsk
);
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
->exec_update_mutex
);
411 if (!ptrace_may_access(task
, PTRACE_MODE_ATTACH_FSCREDS
)) {
412 mutex_unlock(&task
->signal
->exec_update_mutex
);
418 static void unlock_trace(struct task_struct
*task
)
420 mutex_unlock(&task
->signal
->exec_update_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 unsigned long *entries
;
434 * The ability to racily run the kernel stack unwinder on a running task
435 * and then observe the unwinder output is scary; while it is useful for
436 * debugging kernel issues, it can also allow an attacker to leak kernel
438 * Doing this in a manner that is at least safe from races would require
439 * some work to ensure that the remote task can not be scheduled; and
440 * even then, this would still expose the unwinder as local attack
442 * Therefore, this interface is restricted to root.
444 if (!file_ns_capable(m
->file
, &init_user_ns
, CAP_SYS_ADMIN
))
447 entries
= kmalloc_array(MAX_STACK_TRACE_DEPTH
, sizeof(*entries
),
452 err
= lock_trace(task
);
454 unsigned int i
, nr_entries
;
456 nr_entries
= stack_trace_save_tsk(task
, entries
,
457 MAX_STACK_TRACE_DEPTH
, 0);
459 for (i
= 0; i
< nr_entries
; i
++) {
460 seq_printf(m
, "[<0>] %pB\n", (void *)entries
[i
]);
471 #ifdef CONFIG_SCHED_INFO
473 * Provides /proc/PID/schedstat
475 static int proc_pid_schedstat(struct seq_file
*m
, struct pid_namespace
*ns
,
476 struct pid
*pid
, struct task_struct
*task
)
478 if (unlikely(!sched_info_on()))
479 seq_puts(m
, "0 0 0\n");
481 seq_printf(m
, "%llu %llu %lu\n",
482 (unsigned long long)task
->se
.sum_exec_runtime
,
483 (unsigned long long)task
->sched_info
.run_delay
,
484 task
->sched_info
.pcount
);
490 #ifdef CONFIG_LATENCYTOP
491 static int lstats_show_proc(struct seq_file
*m
, void *v
)
494 struct inode
*inode
= m
->private;
495 struct task_struct
*task
= get_proc_task(inode
);
499 seq_puts(m
, "Latency Top version : v0.1\n");
500 for (i
= 0; i
< LT_SAVECOUNT
; i
++) {
501 struct latency_record
*lr
= &task
->latency_record
[i
];
502 if (lr
->backtrace
[0]) {
504 seq_printf(m
, "%i %li %li",
505 lr
->count
, lr
->time
, lr
->max
);
506 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
507 unsigned long bt
= lr
->backtrace
[q
];
511 seq_printf(m
, " %ps", (void *)bt
);
517 put_task_struct(task
);
521 static int lstats_open(struct inode
*inode
, struct file
*file
)
523 return single_open(file
, lstats_show_proc
, inode
);
526 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
527 size_t count
, loff_t
*offs
)
529 struct task_struct
*task
= get_proc_task(file_inode(file
));
533 clear_tsk_latency_tracing(task
);
534 put_task_struct(task
);
539 static const struct file_operations proc_lstats_operations
= {
542 .write
= lstats_write
,
544 .release
= single_release
,
549 static int proc_oom_score(struct seq_file
*m
, struct pid_namespace
*ns
,
550 struct pid
*pid
, struct task_struct
*task
)
552 unsigned long totalpages
= totalram_pages() + total_swap_pages
;
553 unsigned long points
= 0;
555 points
= oom_badness(task
, totalpages
) * 1000 / totalpages
;
556 seq_printf(m
, "%lu\n", points
);
566 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
567 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
568 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
569 [RLIMIT_DATA
] = {"Max data size", "bytes"},
570 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
571 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
572 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
573 [RLIMIT_NPROC
] = {"Max processes", "processes"},
574 [RLIMIT_NOFILE
] = {"Max open files", "files"},
575 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
576 [RLIMIT_AS
] = {"Max address space", "bytes"},
577 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
578 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
579 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
580 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
581 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
582 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
585 /* Display limits for a process */
586 static int proc_pid_limits(struct seq_file
*m
, struct pid_namespace
*ns
,
587 struct pid
*pid
, struct task_struct
*task
)
592 struct rlimit rlim
[RLIM_NLIMITS
];
594 if (!lock_task_sighand(task
, &flags
))
596 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
597 unlock_task_sighand(task
, &flags
);
600 * print the file header
607 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
608 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
609 seq_printf(m
, "%-25s %-20s ",
610 lnames
[i
].name
, "unlimited");
612 seq_printf(m
, "%-25s %-20lu ",
613 lnames
[i
].name
, rlim
[i
].rlim_cur
);
615 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
616 seq_printf(m
, "%-20s ", "unlimited");
618 seq_printf(m
, "%-20lu ", rlim
[i
].rlim_max
);
621 seq_printf(m
, "%-10s\n", lnames
[i
].unit
);
629 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
630 static int proc_pid_syscall(struct seq_file
*m
, struct pid_namespace
*ns
,
631 struct pid
*pid
, struct task_struct
*task
)
633 struct syscall_info info
;
634 u64
*args
= &info
.data
.args
[0];
637 res
= lock_trace(task
);
641 if (task_current_syscall(task
, &info
))
642 seq_puts(m
, "running\n");
643 else if (info
.data
.nr
< 0)
644 seq_printf(m
, "%d 0x%llx 0x%llx\n",
645 info
.data
.nr
, info
.sp
, info
.data
.instruction_pointer
);
648 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
650 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
651 info
.sp
, info
.data
.instruction_pointer
);
656 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
658 /************************************************************************/
659 /* Here the fs part begins */
660 /************************************************************************/
662 /* permission checks */
663 static int proc_fd_access_allowed(struct inode
*inode
)
665 struct task_struct
*task
;
667 /* Allow access to a task's file descriptors if it is us or we
668 * may use ptrace attach to the process and find out that
671 task
= get_proc_task(inode
);
673 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
674 put_task_struct(task
);
679 int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
682 struct inode
*inode
= d_inode(dentry
);
684 if (attr
->ia_valid
& ATTR_MODE
)
687 error
= setattr_prepare(dentry
, attr
);
691 setattr_copy(inode
, attr
);
692 mark_inode_dirty(inode
);
697 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
698 * or euid/egid (for hide_pid_min=2)?
700 static bool has_pid_permissions(struct pid_namespace
*pid
,
701 struct task_struct
*task
,
704 if (pid
->hide_pid
< hide_pid_min
)
706 if (in_group_p(pid
->pid_gid
))
708 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
712 static int proc_pid_permission(struct inode
*inode
, int mask
)
714 struct pid_namespace
*pid
= proc_pid_ns(inode
);
715 struct task_struct
*task
;
718 task
= get_proc_task(inode
);
721 has_perms
= has_pid_permissions(pid
, task
, HIDEPID_NO_ACCESS
);
722 put_task_struct(task
);
725 if (pid
->hide_pid
== HIDEPID_INVISIBLE
) {
727 * Let's make getdents(), stat(), and open()
728 * consistent with each other. If a process
729 * may not stat() a file, it shouldn't be seen
737 return generic_permission(inode
, mask
);
742 static const struct inode_operations proc_def_inode_operations
= {
743 .setattr
= proc_setattr
,
746 static int proc_single_show(struct seq_file
*m
, void *v
)
748 struct inode
*inode
= m
->private;
749 struct pid_namespace
*ns
= proc_pid_ns(inode
);
750 struct pid
*pid
= proc_pid(inode
);
751 struct task_struct
*task
;
754 task
= get_pid_task(pid
, PIDTYPE_PID
);
758 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
760 put_task_struct(task
);
764 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
766 return single_open(filp
, proc_single_show
, inode
);
769 static const struct file_operations proc_single_file_operations
= {
770 .open
= proc_single_open
,
773 .release
= single_release
,
777 struct mm_struct
*proc_mem_open(struct inode
*inode
, unsigned int mode
)
779 struct task_struct
*task
= get_proc_task(inode
);
780 struct mm_struct
*mm
= ERR_PTR(-ESRCH
);
783 mm
= mm_access(task
, mode
| PTRACE_MODE_FSCREDS
);
784 put_task_struct(task
);
786 if (!IS_ERR_OR_NULL(mm
)) {
787 /* ensure this mm_struct can't be freed */
789 /* but do not pin its memory */
797 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
799 struct mm_struct
*mm
= proc_mem_open(inode
, mode
);
804 file
->private_data
= mm
;
808 static int mem_open(struct inode
*inode
, struct file
*file
)
810 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
812 /* OK to pass negative loff_t, we can catch out-of-range */
813 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
818 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
819 size_t count
, loff_t
*ppos
, int write
)
821 struct mm_struct
*mm
= file
->private_data
;
822 unsigned long addr
= *ppos
;
830 page
= (char *)__get_free_page(GFP_KERNEL
);
835 if (!mmget_not_zero(mm
))
838 flags
= FOLL_FORCE
| (write
? FOLL_WRITE
: 0);
841 int this_len
= min_t(int, count
, PAGE_SIZE
);
843 if (write
&& copy_from_user(page
, buf
, this_len
)) {
848 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, flags
);
855 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
869 free_page((unsigned long) page
);
873 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
874 size_t count
, loff_t
*ppos
)
876 return mem_rw(file
, buf
, count
, ppos
, 0);
879 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
880 size_t count
, loff_t
*ppos
)
882 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
885 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
889 file
->f_pos
= offset
;
892 file
->f_pos
+= offset
;
897 force_successful_syscall_return();
901 static int mem_release(struct inode
*inode
, struct file
*file
)
903 struct mm_struct
*mm
= file
->private_data
;
909 static const struct file_operations proc_mem_operations
= {
914 .release
= mem_release
,
917 static int environ_open(struct inode
*inode
, struct file
*file
)
919 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
922 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
923 size_t count
, loff_t
*ppos
)
926 unsigned long src
= *ppos
;
928 struct mm_struct
*mm
= file
->private_data
;
929 unsigned long env_start
, env_end
;
931 /* Ensure the process spawned far enough to have an environment. */
932 if (!mm
|| !mm
->env_end
)
935 page
= (char *)__get_free_page(GFP_KERNEL
);
940 if (!mmget_not_zero(mm
))
943 spin_lock(&mm
->arg_lock
);
944 env_start
= mm
->env_start
;
945 env_end
= mm
->env_end
;
946 spin_unlock(&mm
->arg_lock
);
949 size_t this_len
, max_len
;
952 if (src
>= (env_end
- env_start
))
955 this_len
= env_end
- (env_start
+ src
);
957 max_len
= min_t(size_t, PAGE_SIZE
, count
);
958 this_len
= min(max_len
, this_len
);
960 retval
= access_remote_vm(mm
, (env_start
+ src
), page
, this_len
, FOLL_ANON
);
967 if (copy_to_user(buf
, page
, retval
)) {
981 free_page((unsigned long) page
);
985 static const struct file_operations proc_environ_operations
= {
986 .open
= environ_open
,
987 .read
= environ_read
,
988 .llseek
= generic_file_llseek
,
989 .release
= mem_release
,
992 static int auxv_open(struct inode
*inode
, struct file
*file
)
994 return __mem_open(inode
, file
, PTRACE_MODE_READ_FSCREDS
);
997 static ssize_t
auxv_read(struct file
*file
, char __user
*buf
,
998 size_t count
, loff_t
*ppos
)
1000 struct mm_struct
*mm
= file
->private_data
;
1001 unsigned int nwords
= 0;
1007 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
1008 return simple_read_from_buffer(buf
, count
, ppos
, mm
->saved_auxv
,
1009 nwords
* sizeof(mm
->saved_auxv
[0]));
1012 static const struct file_operations proc_auxv_operations
= {
1015 .llseek
= generic_file_llseek
,
1016 .release
= mem_release
,
1019 static ssize_t
oom_adj_read(struct file
*file
, char __user
*buf
, size_t count
,
1022 struct task_struct
*task
= get_proc_task(file_inode(file
));
1023 char buffer
[PROC_NUMBUF
];
1024 int oom_adj
= OOM_ADJUST_MIN
;
1029 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MAX
)
1030 oom_adj
= OOM_ADJUST_MAX
;
1032 oom_adj
= (task
->signal
->oom_score_adj
* -OOM_DISABLE
) /
1034 put_task_struct(task
);
1035 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1036 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1039 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1041 static DEFINE_MUTEX(oom_adj_mutex
);
1042 struct mm_struct
*mm
= NULL
;
1043 struct task_struct
*task
;
1046 task
= get_proc_task(file_inode(file
));
1050 mutex_lock(&oom_adj_mutex
);
1052 if (oom_adj
< task
->signal
->oom_score_adj
&&
1053 !capable(CAP_SYS_RESOURCE
)) {
1058 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1059 * /proc/pid/oom_score_adj instead.
1061 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1062 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1065 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1066 !capable(CAP_SYS_RESOURCE
)) {
1073 * Make sure we will check other processes sharing the mm if this is
1074 * not vfrok which wants its own oom_score_adj.
1075 * pin the mm so it doesn't go away and get reused after task_unlock
1077 if (!task
->vfork_done
) {
1078 struct task_struct
*p
= find_lock_task_mm(task
);
1081 if (atomic_read(&p
->mm
->mm_users
) > 1) {
1089 task
->signal
->oom_score_adj
= oom_adj
;
1090 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1091 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1092 trace_oom_score_adj_update(task
);
1095 struct task_struct
*p
;
1098 for_each_process(p
) {
1099 if (same_thread_group(task
, p
))
1102 /* do not touch kernel threads or the global init */
1103 if (p
->flags
& PF_KTHREAD
|| is_global_init(p
))
1107 if (!p
->vfork_done
&& process_shares_mm(p
, mm
)) {
1108 p
->signal
->oom_score_adj
= oom_adj
;
1109 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1110 p
->signal
->oom_score_adj_min
= (short)oom_adj
;
1118 mutex_unlock(&oom_adj_mutex
);
1119 put_task_struct(task
);
1124 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1125 * kernels. The effective policy is defined by oom_score_adj, which has a
1126 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1127 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1128 * Processes that become oom disabled via oom_adj will still be oom disabled
1129 * with this implementation.
1131 * oom_adj cannot be removed since existing userspace binaries use it.
1133 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1134 size_t count
, loff_t
*ppos
)
1136 char buffer
[PROC_NUMBUF
];
1140 memset(buffer
, 0, sizeof(buffer
));
1141 if (count
> sizeof(buffer
) - 1)
1142 count
= sizeof(buffer
) - 1;
1143 if (copy_from_user(buffer
, buf
, count
)) {
1148 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1151 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1152 oom_adj
!= OOM_DISABLE
) {
1158 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1159 * value is always attainable.
1161 if (oom_adj
== OOM_ADJUST_MAX
)
1162 oom_adj
= OOM_SCORE_ADJ_MAX
;
1164 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1166 err
= __set_oom_adj(file
, oom_adj
, true);
1168 return err
< 0 ? err
: count
;
1171 static const struct file_operations proc_oom_adj_operations
= {
1172 .read
= oom_adj_read
,
1173 .write
= oom_adj_write
,
1174 .llseek
= generic_file_llseek
,
1177 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1178 size_t count
, loff_t
*ppos
)
1180 struct task_struct
*task
= get_proc_task(file_inode(file
));
1181 char buffer
[PROC_NUMBUF
];
1182 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1187 oom_score_adj
= task
->signal
->oom_score_adj
;
1188 put_task_struct(task
);
1189 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1190 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1193 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1194 size_t count
, loff_t
*ppos
)
1196 char buffer
[PROC_NUMBUF
];
1200 memset(buffer
, 0, sizeof(buffer
));
1201 if (count
> sizeof(buffer
) - 1)
1202 count
= sizeof(buffer
) - 1;
1203 if (copy_from_user(buffer
, buf
, count
)) {
1208 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1211 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1212 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1217 err
= __set_oom_adj(file
, oom_score_adj
, false);
1219 return err
< 0 ? err
: count
;
1222 static const struct file_operations proc_oom_score_adj_operations
= {
1223 .read
= oom_score_adj_read
,
1224 .write
= oom_score_adj_write
,
1225 .llseek
= default_llseek
,
1229 #define TMPBUFLEN 11
1230 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1231 size_t count
, loff_t
*ppos
)
1233 struct inode
* inode
= file_inode(file
);
1234 struct task_struct
*task
= get_proc_task(inode
);
1236 char tmpbuf
[TMPBUFLEN
];
1240 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1241 from_kuid(file
->f_cred
->user_ns
,
1242 audit_get_loginuid(task
)));
1243 put_task_struct(task
);
1244 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1247 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1248 size_t count
, loff_t
*ppos
)
1250 struct inode
* inode
= file_inode(file
);
1256 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1263 /* No partial writes. */
1267 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1271 /* is userspace tring to explicitly UNSET the loginuid? */
1272 if (loginuid
== AUDIT_UID_UNSET
) {
1273 kloginuid
= INVALID_UID
;
1275 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1276 if (!uid_valid(kloginuid
))
1280 rv
= audit_set_loginuid(kloginuid
);
1286 static const struct file_operations proc_loginuid_operations
= {
1287 .read
= proc_loginuid_read
,
1288 .write
= proc_loginuid_write
,
1289 .llseek
= generic_file_llseek
,
1292 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1293 size_t count
, loff_t
*ppos
)
1295 struct inode
* inode
= file_inode(file
);
1296 struct task_struct
*task
= get_proc_task(inode
);
1298 char tmpbuf
[TMPBUFLEN
];
1302 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1303 audit_get_sessionid(task
));
1304 put_task_struct(task
);
1305 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1308 static const struct file_operations proc_sessionid_operations
= {
1309 .read
= proc_sessionid_read
,
1310 .llseek
= generic_file_llseek
,
1314 #ifdef CONFIG_FAULT_INJECTION
1315 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1316 size_t count
, loff_t
*ppos
)
1318 struct task_struct
*task
= get_proc_task(file_inode(file
));
1319 char buffer
[PROC_NUMBUF
];
1325 make_it_fail
= task
->make_it_fail
;
1326 put_task_struct(task
);
1328 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1330 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1333 static ssize_t
proc_fault_inject_write(struct file
* file
,
1334 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1336 struct task_struct
*task
;
1337 char buffer
[PROC_NUMBUF
];
1341 if (!capable(CAP_SYS_RESOURCE
))
1343 memset(buffer
, 0, sizeof(buffer
));
1344 if (count
> sizeof(buffer
) - 1)
1345 count
= sizeof(buffer
) - 1;
1346 if (copy_from_user(buffer
, buf
, count
))
1348 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1351 if (make_it_fail
< 0 || make_it_fail
> 1)
1354 task
= get_proc_task(file_inode(file
));
1357 task
->make_it_fail
= make_it_fail
;
1358 put_task_struct(task
);
1363 static const struct file_operations proc_fault_inject_operations
= {
1364 .read
= proc_fault_inject_read
,
1365 .write
= proc_fault_inject_write
,
1366 .llseek
= generic_file_llseek
,
1369 static ssize_t
proc_fail_nth_write(struct file
*file
, const char __user
*buf
,
1370 size_t count
, loff_t
*ppos
)
1372 struct task_struct
*task
;
1376 err
= kstrtouint_from_user(buf
, count
, 0, &n
);
1380 task
= get_proc_task(file_inode(file
));
1384 put_task_struct(task
);
1389 static ssize_t
proc_fail_nth_read(struct file
*file
, char __user
*buf
,
1390 size_t count
, loff_t
*ppos
)
1392 struct task_struct
*task
;
1393 char numbuf
[PROC_NUMBUF
];
1396 task
= get_proc_task(file_inode(file
));
1399 len
= snprintf(numbuf
, sizeof(numbuf
), "%u\n", task
->fail_nth
);
1400 put_task_struct(task
);
1401 return simple_read_from_buffer(buf
, count
, ppos
, numbuf
, len
);
1404 static const struct file_operations proc_fail_nth_operations
= {
1405 .read
= proc_fail_nth_read
,
1406 .write
= proc_fail_nth_write
,
1411 #ifdef CONFIG_SCHED_DEBUG
1413 * Print out various scheduling related per-task fields:
1415 static int sched_show(struct seq_file
*m
, void *v
)
1417 struct inode
*inode
= m
->private;
1418 struct pid_namespace
*ns
= proc_pid_ns(inode
);
1419 struct task_struct
*p
;
1421 p
= get_proc_task(inode
);
1424 proc_sched_show_task(p
, ns
, m
);
1432 sched_write(struct file
*file
, const char __user
*buf
,
1433 size_t count
, loff_t
*offset
)
1435 struct inode
*inode
= file_inode(file
);
1436 struct task_struct
*p
;
1438 p
= get_proc_task(inode
);
1441 proc_sched_set_task(p
);
1448 static int sched_open(struct inode
*inode
, struct file
*filp
)
1450 return single_open(filp
, sched_show
, inode
);
1453 static const struct file_operations proc_pid_sched_operations
= {
1456 .write
= sched_write
,
1457 .llseek
= seq_lseek
,
1458 .release
= single_release
,
1463 #ifdef CONFIG_SCHED_AUTOGROUP
1465 * Print out autogroup related information:
1467 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1469 struct inode
*inode
= m
->private;
1470 struct task_struct
*p
;
1472 p
= get_proc_task(inode
);
1475 proc_sched_autogroup_show_task(p
, m
);
1483 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1484 size_t count
, loff_t
*offset
)
1486 struct inode
*inode
= file_inode(file
);
1487 struct task_struct
*p
;
1488 char buffer
[PROC_NUMBUF
];
1492 memset(buffer
, 0, sizeof(buffer
));
1493 if (count
> sizeof(buffer
) - 1)
1494 count
= sizeof(buffer
) - 1;
1495 if (copy_from_user(buffer
, buf
, count
))
1498 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1502 p
= get_proc_task(inode
);
1506 err
= proc_sched_autogroup_set_nice(p
, nice
);
1515 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1519 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1521 struct seq_file
*m
= filp
->private_data
;
1528 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1529 .open
= sched_autogroup_open
,
1531 .write
= sched_autogroup_write
,
1532 .llseek
= seq_lseek
,
1533 .release
= single_release
,
1536 #endif /* CONFIG_SCHED_AUTOGROUP */
1538 #ifdef CONFIG_TIME_NS
1539 static int timens_offsets_show(struct seq_file
*m
, void *v
)
1541 struct task_struct
*p
;
1543 p
= get_proc_task(file_inode(m
->file
));
1546 proc_timens_show_offsets(p
, m
);
1553 static ssize_t
timens_offsets_write(struct file
*file
, const char __user
*buf
,
1554 size_t count
, loff_t
*ppos
)
1556 struct inode
*inode
= file_inode(file
);
1557 struct proc_timens_offset offsets
[2];
1558 char *kbuf
= NULL
, *pos
, *next_line
;
1559 struct task_struct
*p
;
1562 /* Only allow < page size writes at the beginning of the file */
1563 if ((*ppos
!= 0) || (count
>= PAGE_SIZE
))
1566 /* Slurp in the user data */
1567 kbuf
= memdup_user_nul(buf
, count
);
1569 return PTR_ERR(kbuf
);
1571 /* Parse the user data */
1574 for (pos
= kbuf
; pos
; pos
= next_line
) {
1575 struct proc_timens_offset
*off
= &offsets
[noffsets
];
1579 /* Find the end of line and ensure we don't look past it */
1580 next_line
= strchr(pos
, '\n');
1584 if (*next_line
== '\0')
1588 err
= sscanf(pos
, "%9s %lld %lu", clock
,
1589 &off
->val
.tv_sec
, &off
->val
.tv_nsec
);
1590 if (err
!= 3 || off
->val
.tv_nsec
>= NSEC_PER_SEC
)
1593 clock
[sizeof(clock
) - 1] = 0;
1594 if (strcmp(clock
, "monotonic") == 0 ||
1595 strcmp(clock
, __stringify(CLOCK_MONOTONIC
)) == 0)
1596 off
->clockid
= CLOCK_MONOTONIC
;
1597 else if (strcmp(clock
, "boottime") == 0 ||
1598 strcmp(clock
, __stringify(CLOCK_BOOTTIME
)) == 0)
1599 off
->clockid
= CLOCK_BOOTTIME
;
1604 if (noffsets
== ARRAY_SIZE(offsets
)) {
1606 count
= next_line
- kbuf
;
1612 p
= get_proc_task(inode
);
1615 ret
= proc_timens_set_offset(file
, p
, offsets
, noffsets
);
1626 static int timens_offsets_open(struct inode
*inode
, struct file
*filp
)
1628 return single_open(filp
, timens_offsets_show
, inode
);
1631 static const struct file_operations proc_timens_offsets_operations
= {
1632 .open
= timens_offsets_open
,
1634 .write
= timens_offsets_write
,
1635 .llseek
= seq_lseek
,
1636 .release
= single_release
,
1638 #endif /* CONFIG_TIME_NS */
1640 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1641 size_t count
, loff_t
*offset
)
1643 struct inode
*inode
= file_inode(file
);
1644 struct task_struct
*p
;
1645 char buffer
[TASK_COMM_LEN
];
1646 const size_t maxlen
= sizeof(buffer
) - 1;
1648 memset(buffer
, 0, sizeof(buffer
));
1649 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1652 p
= get_proc_task(inode
);
1656 if (same_thread_group(current
, p
))
1657 set_task_comm(p
, buffer
);
1666 static int comm_show(struct seq_file
*m
, void *v
)
1668 struct inode
*inode
= m
->private;
1669 struct task_struct
*p
;
1671 p
= get_proc_task(inode
);
1675 proc_task_name(m
, p
, false);
1683 static int comm_open(struct inode
*inode
, struct file
*filp
)
1685 return single_open(filp
, comm_show
, inode
);
1688 static const struct file_operations proc_pid_set_comm_operations
= {
1691 .write
= comm_write
,
1692 .llseek
= seq_lseek
,
1693 .release
= single_release
,
1696 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1698 struct task_struct
*task
;
1699 struct file
*exe_file
;
1701 task
= get_proc_task(d_inode(dentry
));
1704 exe_file
= get_task_exe_file(task
);
1705 put_task_struct(task
);
1707 *exe_path
= exe_file
->f_path
;
1708 path_get(&exe_file
->f_path
);
1715 static const char *proc_pid_get_link(struct dentry
*dentry
,
1716 struct inode
*inode
,
1717 struct delayed_call
*done
)
1720 int error
= -EACCES
;
1723 return ERR_PTR(-ECHILD
);
1725 /* Are we allowed to snoop on the tasks file descriptors? */
1726 if (!proc_fd_access_allowed(inode
))
1729 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1733 error
= nd_jump_link(&path
);
1735 return ERR_PTR(error
);
1738 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1740 char *tmp
= (char *)__get_free_page(GFP_KERNEL
);
1747 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1748 len
= PTR_ERR(pathname
);
1749 if (IS_ERR(pathname
))
1751 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1755 if (copy_to_user(buffer
, pathname
, len
))
1758 free_page((unsigned long)tmp
);
1762 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1764 int error
= -EACCES
;
1765 struct inode
*inode
= d_inode(dentry
);
1768 /* Are we allowed to snoop on the tasks file descriptors? */
1769 if (!proc_fd_access_allowed(inode
))
1772 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1776 error
= do_proc_readlink(&path
, buffer
, buflen
);
1782 const struct inode_operations proc_pid_link_inode_operations
= {
1783 .readlink
= proc_pid_readlink
,
1784 .get_link
= proc_pid_get_link
,
1785 .setattr
= proc_setattr
,
1789 /* building an inode */
1791 void task_dump_owner(struct task_struct
*task
, umode_t mode
,
1792 kuid_t
*ruid
, kgid_t
*rgid
)
1794 /* Depending on the state of dumpable compute who should own a
1795 * proc file for a task.
1797 const struct cred
*cred
;
1801 if (unlikely(task
->flags
& PF_KTHREAD
)) {
1802 *ruid
= GLOBAL_ROOT_UID
;
1803 *rgid
= GLOBAL_ROOT_GID
;
1807 /* Default to the tasks effective ownership */
1809 cred
= __task_cred(task
);
1815 * Before the /proc/pid/status file was created the only way to read
1816 * the effective uid of a /process was to stat /proc/pid. Reading
1817 * /proc/pid/status is slow enough that procps and other packages
1818 * kept stating /proc/pid. To keep the rules in /proc simple I have
1819 * made this apply to all per process world readable and executable
1822 if (mode
!= (S_IFDIR
|S_IRUGO
|S_IXUGO
)) {
1823 struct mm_struct
*mm
;
1826 /* Make non-dumpable tasks owned by some root */
1828 if (get_dumpable(mm
) != SUID_DUMP_USER
) {
1829 struct user_namespace
*user_ns
= mm
->user_ns
;
1831 uid
= make_kuid(user_ns
, 0);
1832 if (!uid_valid(uid
))
1833 uid
= GLOBAL_ROOT_UID
;
1835 gid
= make_kgid(user_ns
, 0);
1836 if (!gid_valid(gid
))
1837 gid
= GLOBAL_ROOT_GID
;
1840 uid
= GLOBAL_ROOT_UID
;
1841 gid
= GLOBAL_ROOT_GID
;
1849 void proc_pid_evict_inode(struct proc_inode
*ei
)
1851 struct pid
*pid
= ei
->pid
;
1853 if (S_ISDIR(ei
->vfs_inode
.i_mode
)) {
1854 spin_lock(&pid
->lock
);
1855 hlist_del_init_rcu(&ei
->sibling_inodes
);
1856 spin_unlock(&pid
->lock
);
1862 struct inode
*proc_pid_make_inode(struct super_block
* sb
,
1863 struct task_struct
*task
, umode_t mode
)
1865 struct inode
* inode
;
1866 struct proc_inode
*ei
;
1869 /* We need a new inode */
1871 inode
= new_inode(sb
);
1877 inode
->i_mode
= mode
;
1878 inode
->i_ino
= get_next_ino();
1879 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= current_time(inode
);
1880 inode
->i_op
= &proc_def_inode_operations
;
1883 * grab the reference to task.
1885 pid
= get_task_pid(task
, PIDTYPE_PID
);
1889 /* Let the pid remember us for quick removal */
1891 if (S_ISDIR(mode
)) {
1892 spin_lock(&pid
->lock
);
1893 hlist_add_head_rcu(&ei
->sibling_inodes
, &pid
->inodes
);
1894 spin_unlock(&pid
->lock
);
1897 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1898 security_task_to_inode(task
, inode
);
1908 int pid_getattr(const struct path
*path
, struct kstat
*stat
,
1909 u32 request_mask
, unsigned int query_flags
)
1911 struct inode
*inode
= d_inode(path
->dentry
);
1912 struct pid_namespace
*pid
= proc_pid_ns(inode
);
1913 struct task_struct
*task
;
1915 generic_fillattr(inode
, stat
);
1917 stat
->uid
= GLOBAL_ROOT_UID
;
1918 stat
->gid
= GLOBAL_ROOT_GID
;
1920 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1922 if (!has_pid_permissions(pid
, task
, HIDEPID_INVISIBLE
)) {
1925 * This doesn't prevent learning whether PID exists,
1926 * it only makes getattr() consistent with readdir().
1930 task_dump_owner(task
, inode
->i_mode
, &stat
->uid
, &stat
->gid
);
1939 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1941 void pid_update_inode(struct task_struct
*task
, struct inode
*inode
)
1943 task_dump_owner(task
, inode
->i_mode
, &inode
->i_uid
, &inode
->i_gid
);
1945 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1946 security_task_to_inode(task
, inode
);
1950 * Rewrite the inode's ownerships here because the owning task may have
1951 * performed a setuid(), etc.
1954 static int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1956 struct inode
*inode
;
1957 struct task_struct
*task
;
1959 if (flags
& LOOKUP_RCU
)
1962 inode
= d_inode(dentry
);
1963 task
= get_proc_task(inode
);
1966 pid_update_inode(task
, inode
);
1967 put_task_struct(task
);
1973 static inline bool proc_inode_is_dead(struct inode
*inode
)
1975 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
1978 int pid_delete_dentry(const struct dentry
*dentry
)
1980 /* Is the task we represent dead?
1981 * If so, then don't put the dentry on the lru list,
1982 * kill it immediately.
1984 return proc_inode_is_dead(d_inode(dentry
));
1987 const struct dentry_operations pid_dentry_operations
=
1989 .d_revalidate
= pid_revalidate
,
1990 .d_delete
= pid_delete_dentry
,
1996 * Fill a directory entry.
1998 * If possible create the dcache entry and derive our inode number and
1999 * file type from dcache entry.
2001 * Since all of the proc inode numbers are dynamically generated, the inode
2002 * numbers do not exist until the inode is cache. This means creating the
2003 * the dcache entry in readdir is necessary to keep the inode numbers
2004 * reported by readdir in sync with the inode numbers reported
2007 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
2008 const char *name
, unsigned int len
,
2009 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
2011 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
2012 struct qstr qname
= QSTR_INIT(name
, len
);
2013 struct inode
*inode
;
2014 unsigned type
= DT_UNKNOWN
;
2017 child
= d_hash_and_lookup(dir
, &qname
);
2019 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
2020 child
= d_alloc_parallel(dir
, &qname
, &wq
);
2022 goto end_instantiate
;
2023 if (d_in_lookup(child
)) {
2025 res
= instantiate(child
, task
, ptr
);
2026 d_lookup_done(child
);
2027 if (unlikely(res
)) {
2031 goto end_instantiate
;
2035 inode
= d_inode(child
);
2037 type
= inode
->i_mode
>> 12;
2040 return dir_emit(ctx
, name
, len
, ino
, type
);
2044 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2045 * which represent vma start and end addresses.
2047 static int dname_to_vma_addr(struct dentry
*dentry
,
2048 unsigned long *start
, unsigned long *end
)
2050 const char *str
= dentry
->d_name
.name
;
2051 unsigned long long sval
, eval
;
2054 if (str
[0] == '0' && str
[1] != '-')
2056 len
= _parse_integer(str
, 16, &sval
);
2057 if (len
& KSTRTOX_OVERFLOW
)
2059 if (sval
!= (unsigned long)sval
)
2067 if (str
[0] == '0' && str
[1])
2069 len
= _parse_integer(str
, 16, &eval
);
2070 if (len
& KSTRTOX_OVERFLOW
)
2072 if (eval
!= (unsigned long)eval
)
2085 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
2087 unsigned long vm_start
, vm_end
;
2088 bool exact_vma_exists
= false;
2089 struct mm_struct
*mm
= NULL
;
2090 struct task_struct
*task
;
2091 struct inode
*inode
;
2094 if (flags
& LOOKUP_RCU
)
2097 inode
= d_inode(dentry
);
2098 task
= get_proc_task(inode
);
2102 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
2103 if (IS_ERR_OR_NULL(mm
))
2106 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
2107 status
= down_read_killable(&mm
->mmap_sem
);
2109 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
,
2111 up_read(&mm
->mmap_sem
);
2117 if (exact_vma_exists
) {
2118 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
2120 security_task_to_inode(task
, inode
);
2125 put_task_struct(task
);
2131 static const struct dentry_operations tid_map_files_dentry_operations
= {
2132 .d_revalidate
= map_files_d_revalidate
,
2133 .d_delete
= pid_delete_dentry
,
2136 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2138 unsigned long vm_start
, vm_end
;
2139 struct vm_area_struct
*vma
;
2140 struct task_struct
*task
;
2141 struct mm_struct
*mm
;
2145 task
= get_proc_task(d_inode(dentry
));
2149 mm
= get_task_mm(task
);
2150 put_task_struct(task
);
2154 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2158 rc
= down_read_killable(&mm
->mmap_sem
);
2163 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2164 if (vma
&& vma
->vm_file
) {
2165 *path
= vma
->vm_file
->f_path
;
2169 up_read(&mm
->mmap_sem
);
2177 struct map_files_info
{
2178 unsigned long start
;
2184 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2185 * symlinks may be used to bypass permissions on ancestor directories in the
2186 * path to the file in question.
2189 proc_map_files_get_link(struct dentry
*dentry
,
2190 struct inode
*inode
,
2191 struct delayed_call
*done
)
2193 if (!capable(CAP_SYS_ADMIN
))
2194 return ERR_PTR(-EPERM
);
2196 return proc_pid_get_link(dentry
, inode
, done
);
2200 * Identical to proc_pid_link_inode_operations except for get_link()
2202 static const struct inode_operations proc_map_files_link_inode_operations
= {
2203 .readlink
= proc_pid_readlink
,
2204 .get_link
= proc_map_files_get_link
,
2205 .setattr
= proc_setattr
,
2208 static struct dentry
*
2209 proc_map_files_instantiate(struct dentry
*dentry
,
2210 struct task_struct
*task
, const void *ptr
)
2212 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
2213 struct proc_inode
*ei
;
2214 struct inode
*inode
;
2216 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFLNK
|
2217 ((mode
& FMODE_READ
) ? S_IRUSR
: 0) |
2218 ((mode
& FMODE_WRITE
) ? S_IWUSR
: 0));
2220 return ERR_PTR(-ENOENT
);
2223 ei
->op
.proc_get_link
= map_files_get_link
;
2225 inode
->i_op
= &proc_map_files_link_inode_operations
;
2228 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2229 return d_splice_alias(inode
, dentry
);
2232 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2233 struct dentry
*dentry
, unsigned int flags
)
2235 unsigned long vm_start
, vm_end
;
2236 struct vm_area_struct
*vma
;
2237 struct task_struct
*task
;
2238 struct dentry
*result
;
2239 struct mm_struct
*mm
;
2241 result
= ERR_PTR(-ENOENT
);
2242 task
= get_proc_task(dir
);
2246 result
= ERR_PTR(-EACCES
);
2247 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2250 result
= ERR_PTR(-ENOENT
);
2251 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2254 mm
= get_task_mm(task
);
2258 result
= ERR_PTR(-EINTR
);
2259 if (down_read_killable(&mm
->mmap_sem
))
2262 result
= ERR_PTR(-ENOENT
);
2263 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2268 result
= proc_map_files_instantiate(dentry
, task
,
2269 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2272 up_read(&mm
->mmap_sem
);
2276 put_task_struct(task
);
2281 static const struct inode_operations proc_map_files_inode_operations
= {
2282 .lookup
= proc_map_files_lookup
,
2283 .permission
= proc_fd_permission
,
2284 .setattr
= proc_setattr
,
2288 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2290 struct vm_area_struct
*vma
;
2291 struct task_struct
*task
;
2292 struct mm_struct
*mm
;
2293 unsigned long nr_files
, pos
, i
;
2294 GENRADIX(struct map_files_info
) fa
;
2295 struct map_files_info
*p
;
2301 task
= get_proc_task(file_inode(file
));
2306 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2310 if (!dir_emit_dots(file
, ctx
))
2313 mm
= get_task_mm(task
);
2317 ret
= down_read_killable(&mm
->mmap_sem
);
2326 * We need two passes here:
2328 * 1) Collect vmas of mapped files with mmap_sem taken
2329 * 2) Release mmap_sem and instantiate entries
2331 * otherwise we get lockdep complained, since filldir()
2332 * routine might require mmap_sem taken in might_fault().
2335 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2338 if (++pos
<= ctx
->pos
)
2341 p
= genradix_ptr_alloc(&fa
, nr_files
++, GFP_KERNEL
);
2344 up_read(&mm
->mmap_sem
);
2349 p
->start
= vma
->vm_start
;
2350 p
->end
= vma
->vm_end
;
2351 p
->mode
= vma
->vm_file
->f_mode
;
2353 up_read(&mm
->mmap_sem
);
2356 for (i
= 0; i
< nr_files
; i
++) {
2357 char buf
[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2360 p
= genradix_ptr(&fa
, i
);
2361 len
= snprintf(buf
, sizeof(buf
), "%lx-%lx", p
->start
, p
->end
);
2362 if (!proc_fill_cache(file
, ctx
,
2364 proc_map_files_instantiate
,
2366 (void *)(unsigned long)p
->mode
))
2372 put_task_struct(task
);
2378 static const struct file_operations proc_map_files_operations
= {
2379 .read
= generic_read_dir
,
2380 .iterate_shared
= proc_map_files_readdir
,
2381 .llseek
= generic_file_llseek
,
2384 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2385 struct timers_private
{
2387 struct task_struct
*task
;
2388 struct sighand_struct
*sighand
;
2389 struct pid_namespace
*ns
;
2390 unsigned long flags
;
2393 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2395 struct timers_private
*tp
= m
->private;
2397 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2399 return ERR_PTR(-ESRCH
);
2401 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2403 return ERR_PTR(-ESRCH
);
2405 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2408 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2410 struct timers_private
*tp
= m
->private;
2411 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2414 static void timers_stop(struct seq_file
*m
, void *v
)
2416 struct timers_private
*tp
= m
->private;
2419 unlock_task_sighand(tp
->task
, &tp
->flags
);
2424 put_task_struct(tp
->task
);
2429 static int show_timer(struct seq_file
*m
, void *v
)
2431 struct k_itimer
*timer
;
2432 struct timers_private
*tp
= m
->private;
2434 static const char * const nstr
[] = {
2435 [SIGEV_SIGNAL
] = "signal",
2436 [SIGEV_NONE
] = "none",
2437 [SIGEV_THREAD
] = "thread",
2440 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2441 notify
= timer
->it_sigev_notify
;
2443 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2444 seq_printf(m
, "signal: %d/%px\n",
2445 timer
->sigq
->info
.si_signo
,
2446 timer
->sigq
->info
.si_value
.sival_ptr
);
2447 seq_printf(m
, "notify: %s/%s.%d\n",
2448 nstr
[notify
& ~SIGEV_THREAD_ID
],
2449 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2450 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2451 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2456 static const struct seq_operations proc_timers_seq_ops
= {
2457 .start
= timers_start
,
2458 .next
= timers_next
,
2459 .stop
= timers_stop
,
2463 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2465 struct timers_private
*tp
;
2467 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2468 sizeof(struct timers_private
));
2472 tp
->pid
= proc_pid(inode
);
2473 tp
->ns
= proc_pid_ns(inode
);
2477 static const struct file_operations proc_timers_operations
= {
2478 .open
= proc_timers_open
,
2480 .llseek
= seq_lseek
,
2481 .release
= seq_release_private
,
2485 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2486 size_t count
, loff_t
*offset
)
2488 struct inode
*inode
= file_inode(file
);
2489 struct task_struct
*p
;
2493 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2497 p
= get_proc_task(inode
);
2503 if (!ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
)) {
2510 err
= security_task_setscheduler(p
);
2519 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2521 p
->timer_slack_ns
= slack_ns
;
2530 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2532 struct inode
*inode
= m
->private;
2533 struct task_struct
*p
;
2536 p
= get_proc_task(inode
);
2542 if (!ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
)) {
2549 err
= security_task_getscheduler(p
);
2555 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2564 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2566 return single_open(filp
, timerslack_ns_show
, inode
);
2569 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2570 .open
= timerslack_ns_open
,
2572 .write
= timerslack_ns_write
,
2573 .llseek
= seq_lseek
,
2574 .release
= single_release
,
2577 static struct dentry
*proc_pident_instantiate(struct dentry
*dentry
,
2578 struct task_struct
*task
, const void *ptr
)
2580 const struct pid_entry
*p
= ptr
;
2581 struct inode
*inode
;
2582 struct proc_inode
*ei
;
2584 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, p
->mode
);
2586 return ERR_PTR(-ENOENT
);
2589 if (S_ISDIR(inode
->i_mode
))
2590 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2592 inode
->i_op
= p
->iop
;
2594 inode
->i_fop
= p
->fop
;
2596 pid_update_inode(task
, inode
);
2597 d_set_d_op(dentry
, &pid_dentry_operations
);
2598 return d_splice_alias(inode
, dentry
);
2601 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2602 struct dentry
*dentry
,
2603 const struct pid_entry
*p
,
2604 const struct pid_entry
*end
)
2606 struct task_struct
*task
= get_proc_task(dir
);
2607 struct dentry
*res
= ERR_PTR(-ENOENT
);
2613 * Yes, it does not scale. And it should not. Don't add
2614 * new entries into /proc/<tgid>/ without very good reasons.
2616 for (; p
< end
; p
++) {
2617 if (p
->len
!= dentry
->d_name
.len
)
2619 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
)) {
2620 res
= proc_pident_instantiate(dentry
, task
, p
);
2624 put_task_struct(task
);
2629 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2630 const struct pid_entry
*ents
, unsigned int nents
)
2632 struct task_struct
*task
= get_proc_task(file_inode(file
));
2633 const struct pid_entry
*p
;
2638 if (!dir_emit_dots(file
, ctx
))
2641 if (ctx
->pos
>= nents
+ 2)
2644 for (p
= ents
+ (ctx
->pos
- 2); p
< ents
+ nents
; p
++) {
2645 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2646 proc_pident_instantiate
, task
, p
))
2651 put_task_struct(task
);
2655 #ifdef CONFIG_SECURITY
2656 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2657 size_t count
, loff_t
*ppos
)
2659 struct inode
* inode
= file_inode(file
);
2662 struct task_struct
*task
= get_proc_task(inode
);
2667 length
= security_getprocattr(task
, PROC_I(inode
)->op
.lsm
,
2668 (char*)file
->f_path
.dentry
->d_name
.name
,
2670 put_task_struct(task
);
2672 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2677 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2678 size_t count
, loff_t
*ppos
)
2680 struct inode
* inode
= file_inode(file
);
2681 struct task_struct
*task
;
2686 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
2691 /* A task may only write its own attributes. */
2692 if (current
!= task
) {
2696 /* Prevent changes to overridden credentials. */
2697 if (current_cred() != current_real_cred()) {
2703 if (count
> PAGE_SIZE
)
2706 /* No partial writes. */
2710 page
= memdup_user(buf
, count
);
2716 /* Guard against adverse ptrace interaction */
2717 rv
= mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
);
2721 rv
= security_setprocattr(PROC_I(inode
)->op
.lsm
,
2722 file
->f_path
.dentry
->d_name
.name
, page
,
2724 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
2731 static const struct file_operations proc_pid_attr_operations
= {
2732 .read
= proc_pid_attr_read
,
2733 .write
= proc_pid_attr_write
,
2734 .llseek
= generic_file_llseek
,
2737 #define LSM_DIR_OPS(LSM) \
2738 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2739 struct dir_context *ctx) \
2741 return proc_pident_readdir(filp, ctx, \
2742 LSM##_attr_dir_stuff, \
2743 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2746 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2747 .read = generic_read_dir, \
2748 .iterate = proc_##LSM##_attr_dir_iterate, \
2749 .llseek = default_llseek, \
2752 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2753 struct dentry *dentry, unsigned int flags) \
2755 return proc_pident_lookup(dir, dentry, \
2756 LSM##_attr_dir_stuff, \
2757 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2760 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2761 .lookup = proc_##LSM##_attr_dir_lookup, \
2762 .getattr = pid_getattr, \
2763 .setattr = proc_setattr, \
2766 #ifdef CONFIG_SECURITY_SMACK
2767 static const struct pid_entry smack_attr_dir_stuff
[] = {
2768 ATTR("smack", "current", 0666),
2773 static const struct pid_entry attr_dir_stuff
[] = {
2774 ATTR(NULL
, "current", 0666),
2775 ATTR(NULL
, "prev", 0444),
2776 ATTR(NULL
, "exec", 0666),
2777 ATTR(NULL
, "fscreate", 0666),
2778 ATTR(NULL
, "keycreate", 0666),
2779 ATTR(NULL
, "sockcreate", 0666),
2780 #ifdef CONFIG_SECURITY_SMACK
2782 proc_smack_attr_dir_inode_ops
, proc_smack_attr_dir_ops
),
2786 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2788 return proc_pident_readdir(file
, ctx
,
2789 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2792 static const struct file_operations proc_attr_dir_operations
= {
2793 .read
= generic_read_dir
,
2794 .iterate_shared
= proc_attr_dir_readdir
,
2795 .llseek
= generic_file_llseek
,
2798 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2799 struct dentry
*dentry
, unsigned int flags
)
2801 return proc_pident_lookup(dir
, dentry
,
2803 attr_dir_stuff
+ ARRAY_SIZE(attr_dir_stuff
));
2806 static const struct inode_operations proc_attr_dir_inode_operations
= {
2807 .lookup
= proc_attr_dir_lookup
,
2808 .getattr
= pid_getattr
,
2809 .setattr
= proc_setattr
,
2814 #ifdef CONFIG_ELF_CORE
2815 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2816 size_t count
, loff_t
*ppos
)
2818 struct task_struct
*task
= get_proc_task(file_inode(file
));
2819 struct mm_struct
*mm
;
2820 char buffer
[PROC_NUMBUF
];
2828 mm
= get_task_mm(task
);
2830 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2831 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2832 MMF_DUMP_FILTER_SHIFT
));
2834 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2837 put_task_struct(task
);
2842 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2843 const char __user
*buf
,
2847 struct task_struct
*task
;
2848 struct mm_struct
*mm
;
2854 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2859 task
= get_proc_task(file_inode(file
));
2863 mm
= get_task_mm(task
);
2868 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2870 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2872 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2877 put_task_struct(task
);
2884 static const struct file_operations proc_coredump_filter_operations
= {
2885 .read
= proc_coredump_filter_read
,
2886 .write
= proc_coredump_filter_write
,
2887 .llseek
= generic_file_llseek
,
2891 #ifdef CONFIG_TASK_IO_ACCOUNTING
2892 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2894 struct task_io_accounting acct
= task
->ioac
;
2895 unsigned long flags
;
2898 result
= mutex_lock_killable(&task
->signal
->exec_update_mutex
);
2902 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2907 if (whole
&& lock_task_sighand(task
, &flags
)) {
2908 struct task_struct
*t
= task
;
2910 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2911 while_each_thread(task
, t
)
2912 task_io_accounting_add(&acct
, &t
->ioac
);
2914 unlock_task_sighand(task
, &flags
);
2921 "read_bytes: %llu\n"
2922 "write_bytes: %llu\n"
2923 "cancelled_write_bytes: %llu\n",
2924 (unsigned long long)acct
.rchar
,
2925 (unsigned long long)acct
.wchar
,
2926 (unsigned long long)acct
.syscr
,
2927 (unsigned long long)acct
.syscw
,
2928 (unsigned long long)acct
.read_bytes
,
2929 (unsigned long long)acct
.write_bytes
,
2930 (unsigned long long)acct
.cancelled_write_bytes
);
2934 mutex_unlock(&task
->signal
->exec_update_mutex
);
2938 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2939 struct pid
*pid
, struct task_struct
*task
)
2941 return do_io_accounting(task
, m
, 0);
2944 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2945 struct pid
*pid
, struct task_struct
*task
)
2947 return do_io_accounting(task
, m
, 1);
2949 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2951 #ifdef CONFIG_USER_NS
2952 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2953 const struct seq_operations
*seq_ops
)
2955 struct user_namespace
*ns
= NULL
;
2956 struct task_struct
*task
;
2957 struct seq_file
*seq
;
2960 task
= get_proc_task(inode
);
2963 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2965 put_task_struct(task
);
2970 ret
= seq_open(file
, seq_ops
);
2974 seq
= file
->private_data
;
2984 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2986 struct seq_file
*seq
= file
->private_data
;
2987 struct user_namespace
*ns
= seq
->private;
2989 return seq_release(inode
, file
);
2992 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2994 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2997 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2999 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
3002 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
3004 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
3007 static const struct file_operations proc_uid_map_operations
= {
3008 .open
= proc_uid_map_open
,
3009 .write
= proc_uid_map_write
,
3011 .llseek
= seq_lseek
,
3012 .release
= proc_id_map_release
,
3015 static const struct file_operations proc_gid_map_operations
= {
3016 .open
= proc_gid_map_open
,
3017 .write
= proc_gid_map_write
,
3019 .llseek
= seq_lseek
,
3020 .release
= proc_id_map_release
,
3023 static const struct file_operations proc_projid_map_operations
= {
3024 .open
= proc_projid_map_open
,
3025 .write
= proc_projid_map_write
,
3027 .llseek
= seq_lseek
,
3028 .release
= proc_id_map_release
,
3031 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
3033 struct user_namespace
*ns
= NULL
;
3034 struct task_struct
*task
;
3038 task
= get_proc_task(inode
);
3041 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
3043 put_task_struct(task
);
3048 if (file
->f_mode
& FMODE_WRITE
) {
3050 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
3054 ret
= single_open(file
, &proc_setgroups_show
, ns
);
3065 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
3067 struct seq_file
*seq
= file
->private_data
;
3068 struct user_namespace
*ns
= seq
->private;
3069 int ret
= single_release(inode
, file
);
3074 static const struct file_operations proc_setgroups_operations
= {
3075 .open
= proc_setgroups_open
,
3076 .write
= proc_setgroups_write
,
3078 .llseek
= seq_lseek
,
3079 .release
= proc_setgroups_release
,
3081 #endif /* CONFIG_USER_NS */
3083 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
3084 struct pid
*pid
, struct task_struct
*task
)
3086 int err
= lock_trace(task
);
3088 seq_printf(m
, "%08x\n", task
->personality
);
3094 #ifdef CONFIG_LIVEPATCH
3095 static int proc_pid_patch_state(struct seq_file
*m
, struct pid_namespace
*ns
,
3096 struct pid
*pid
, struct task_struct
*task
)
3098 seq_printf(m
, "%d\n", task
->patch_state
);
3101 #endif /* CONFIG_LIVEPATCH */
3103 #ifdef CONFIG_STACKLEAK_METRICS
3104 static int proc_stack_depth(struct seq_file
*m
, struct pid_namespace
*ns
,
3105 struct pid
*pid
, struct task_struct
*task
)
3107 unsigned long prev_depth
= THREAD_SIZE
-
3108 (task
->prev_lowest_stack
& (THREAD_SIZE
- 1));
3109 unsigned long depth
= THREAD_SIZE
-
3110 (task
->lowest_stack
& (THREAD_SIZE
- 1));
3112 seq_printf(m
, "previous stack depth: %lu\nstack depth: %lu\n",
3116 #endif /* CONFIG_STACKLEAK_METRICS */
3121 static const struct file_operations proc_task_operations
;
3122 static const struct inode_operations proc_task_inode_operations
;
3124 static const struct pid_entry tgid_base_stuff
[] = {
3125 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
3126 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3127 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
3128 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3129 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3131 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3133 REG("environ", S_IRUSR
, proc_environ_operations
),
3134 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3135 ONE("status", S_IRUGO
, proc_pid_status
),
3136 ONE("personality", S_IRUSR
, proc_pid_personality
),
3137 ONE("limits", S_IRUGO
, proc_pid_limits
),
3138 #ifdef CONFIG_SCHED_DEBUG
3139 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3141 #ifdef CONFIG_SCHED_AUTOGROUP
3142 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
3144 #ifdef CONFIG_TIME_NS
3145 REG("timens_offsets", S_IRUGO
|S_IWUSR
, proc_timens_offsets_operations
),
3147 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3148 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3149 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3151 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3152 ONE("stat", S_IRUGO
, proc_tgid_stat
),
3153 ONE("statm", S_IRUGO
, proc_pid_statm
),
3154 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3156 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3158 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3159 LNK("cwd", proc_cwd_link
),
3160 LNK("root", proc_root_link
),
3161 LNK("exe", proc_exe_link
),
3162 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3163 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3164 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
3165 #ifdef CONFIG_PROC_PAGE_MONITOR
3166 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3167 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3168 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3169 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3171 #ifdef CONFIG_SECURITY
3172 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3174 #ifdef CONFIG_KALLSYMS
3175 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3177 #ifdef CONFIG_STACKTRACE
3178 ONE("stack", S_IRUSR
, proc_pid_stack
),
3180 #ifdef CONFIG_SCHED_INFO
3181 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3183 #ifdef CONFIG_LATENCYTOP
3184 REG("latency", S_IRUGO
, proc_lstats_operations
),
3186 #ifdef CONFIG_PROC_PID_CPUSET
3187 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3189 #ifdef CONFIG_CGROUPS
3190 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3192 #ifdef CONFIG_PROC_CPU_RESCTRL
3193 ONE("cpu_resctrl_groups", S_IRUGO
, proc_resctrl_show
),
3195 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3196 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3197 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3199 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3200 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3202 #ifdef CONFIG_FAULT_INJECTION
3203 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3204 REG("fail-nth", 0644, proc_fail_nth_operations
),
3206 #ifdef CONFIG_ELF_CORE
3207 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3209 #ifdef CONFIG_TASK_IO_ACCOUNTING
3210 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
3212 #ifdef CONFIG_USER_NS
3213 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3214 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3215 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3216 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3218 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3219 REG("timers", S_IRUGO
, proc_timers_operations
),
3221 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
3222 #ifdef CONFIG_LIVEPATCH
3223 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3225 #ifdef CONFIG_STACKLEAK_METRICS
3226 ONE("stack_depth", S_IRUGO
, proc_stack_depth
),
3228 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3229 ONE("arch_status", S_IRUGO
, proc_pid_arch_status
),
3233 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3235 return proc_pident_readdir(file
, ctx
,
3236 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3239 static const struct file_operations proc_tgid_base_operations
= {
3240 .read
= generic_read_dir
,
3241 .iterate_shared
= proc_tgid_base_readdir
,
3242 .llseek
= generic_file_llseek
,
3245 struct pid
*tgid_pidfd_to_pid(const struct file
*file
)
3247 if (file
->f_op
!= &proc_tgid_base_operations
)
3248 return ERR_PTR(-EBADF
);
3250 return proc_pid(file_inode(file
));
3253 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3255 return proc_pident_lookup(dir
, dentry
,
3257 tgid_base_stuff
+ ARRAY_SIZE(tgid_base_stuff
));
3260 static const struct inode_operations proc_tgid_base_inode_operations
= {
3261 .lookup
= proc_tgid_base_lookup
,
3262 .getattr
= pid_getattr
,
3263 .setattr
= proc_setattr
,
3264 .permission
= proc_pid_permission
,
3268 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3269 * @pid: pid that should be flushed.
3271 * This function walks a list of inodes (that belong to any proc
3272 * filesystem) that are attached to the pid and flushes them from
3275 * It is safe and reasonable to cache /proc entries for a task until
3276 * that task exits. After that they just clog up the dcache with
3277 * useless entries, possibly causing useful dcache entries to be
3278 * flushed instead. This routine is provided to flush those useless
3279 * dcache entries when a process is reaped.
3281 * NOTE: This routine is just an optimization so it does not guarantee
3282 * that no dcache entries will exist after a process is reaped
3283 * it just makes it very unlikely that any will persist.
3286 void proc_flush_pid(struct pid
*pid
)
3288 proc_invalidate_siblings_dcache(&pid
->inodes
, &pid
->lock
);
3291 static struct dentry
*proc_pid_instantiate(struct dentry
* dentry
,
3292 struct task_struct
*task
, const void *ptr
)
3294 struct inode
*inode
;
3296 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3298 return ERR_PTR(-ENOENT
);
3300 inode
->i_op
= &proc_tgid_base_inode_operations
;
3301 inode
->i_fop
= &proc_tgid_base_operations
;
3302 inode
->i_flags
|=S_IMMUTABLE
;
3304 set_nlink(inode
, nlink_tgid
);
3305 pid_update_inode(task
, inode
);
3307 d_set_d_op(dentry
, &pid_dentry_operations
);
3308 return d_splice_alias(inode
, dentry
);
3311 struct dentry
*proc_pid_lookup(struct dentry
*dentry
, unsigned int flags
)
3313 struct task_struct
*task
;
3315 struct pid_namespace
*ns
;
3316 struct dentry
*result
= ERR_PTR(-ENOENT
);
3318 tgid
= name_to_int(&dentry
->d_name
);
3322 ns
= dentry
->d_sb
->s_fs_info
;
3324 task
= find_task_by_pid_ns(tgid
, ns
);
3326 get_task_struct(task
);
3331 result
= proc_pid_instantiate(dentry
, task
, NULL
);
3332 put_task_struct(task
);
3338 * Find the first task with tgid >= tgid
3343 struct task_struct
*task
;
3345 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3350 put_task_struct(iter
.task
);
3354 pid
= find_ge_pid(iter
.tgid
, ns
);
3356 iter
.tgid
= pid_nr_ns(pid
, ns
);
3357 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3358 /* What we to know is if the pid we have find is the
3359 * pid of a thread_group_leader. Testing for task
3360 * being a thread_group_leader is the obvious thing
3361 * todo but there is a window when it fails, due to
3362 * the pid transfer logic in de_thread.
3364 * So we perform the straight forward test of seeing
3365 * if the pid we have found is the pid of a thread
3366 * group leader, and don't worry if the task we have
3367 * found doesn't happen to be a thread group leader.
3368 * As we don't care in the case of readdir.
3370 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3374 get_task_struct(iter
.task
);
3380 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3382 /* for the /proc/ directory itself, after non-process stuff has been done */
3383 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3385 struct tgid_iter iter
;
3386 struct pid_namespace
*ns
= proc_pid_ns(file_inode(file
));
3387 loff_t pos
= ctx
->pos
;
3389 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3392 if (pos
== TGID_OFFSET
- 2) {
3393 struct inode
*inode
= d_inode(ns
->proc_self
);
3394 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3396 ctx
->pos
= pos
= pos
+ 1;
3398 if (pos
== TGID_OFFSET
- 1) {
3399 struct inode
*inode
= d_inode(ns
->proc_thread_self
);
3400 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3402 ctx
->pos
= pos
= pos
+ 1;
3404 iter
.tgid
= pos
- TGID_OFFSET
;
3406 for (iter
= next_tgid(ns
, iter
);
3408 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3413 if (!has_pid_permissions(ns
, iter
.task
, HIDEPID_INVISIBLE
))
3416 len
= snprintf(name
, sizeof(name
), "%u", iter
.tgid
);
3417 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3418 if (!proc_fill_cache(file
, ctx
, name
, len
,
3419 proc_pid_instantiate
, iter
.task
, NULL
)) {
3420 put_task_struct(iter
.task
);
3424 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3429 * proc_tid_comm_permission is a special permission function exclusively
3430 * used for the node /proc/<pid>/task/<tid>/comm.
3431 * It bypasses generic permission checks in the case where a task of the same
3432 * task group attempts to access the node.
3433 * The rationale behind this is that glibc and bionic access this node for
3434 * cross thread naming (pthread_set/getname_np(!self)). However, if
3435 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3436 * which locks out the cross thread naming implementation.
3437 * This function makes sure that the node is always accessible for members of
3438 * same thread group.
3440 static int proc_tid_comm_permission(struct inode
*inode
, int mask
)
3442 bool is_same_tgroup
;
3443 struct task_struct
*task
;
3445 task
= get_proc_task(inode
);
3448 is_same_tgroup
= same_thread_group(current
, task
);
3449 put_task_struct(task
);
3451 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3452 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3453 * read or written by the members of the corresponding
3459 return generic_permission(inode
, mask
);
3462 static const struct inode_operations proc_tid_comm_inode_operations
= {
3463 .permission
= proc_tid_comm_permission
,
3469 static const struct pid_entry tid_base_stuff
[] = {
3470 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3471 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3472 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3474 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3476 REG("environ", S_IRUSR
, proc_environ_operations
),
3477 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3478 ONE("status", S_IRUGO
, proc_pid_status
),
3479 ONE("personality", S_IRUSR
, proc_pid_personality
),
3480 ONE("limits", S_IRUGO
, proc_pid_limits
),
3481 #ifdef CONFIG_SCHED_DEBUG
3482 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3484 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3485 &proc_tid_comm_inode_operations
,
3486 &proc_pid_set_comm_operations
, {}),
3487 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3488 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3490 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3491 ONE("stat", S_IRUGO
, proc_tid_stat
),
3492 ONE("statm", S_IRUGO
, proc_pid_statm
),
3493 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3494 #ifdef CONFIG_PROC_CHILDREN
3495 REG("children", S_IRUGO
, proc_tid_children_operations
),
3498 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3500 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3501 LNK("cwd", proc_cwd_link
),
3502 LNK("root", proc_root_link
),
3503 LNK("exe", proc_exe_link
),
3504 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3505 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3506 #ifdef CONFIG_PROC_PAGE_MONITOR
3507 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3508 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3509 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3510 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3512 #ifdef CONFIG_SECURITY
3513 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3515 #ifdef CONFIG_KALLSYMS
3516 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3518 #ifdef CONFIG_STACKTRACE
3519 ONE("stack", S_IRUSR
, proc_pid_stack
),
3521 #ifdef CONFIG_SCHED_INFO
3522 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3524 #ifdef CONFIG_LATENCYTOP
3525 REG("latency", S_IRUGO
, proc_lstats_operations
),
3527 #ifdef CONFIG_PROC_PID_CPUSET
3528 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3530 #ifdef CONFIG_CGROUPS
3531 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3533 #ifdef CONFIG_PROC_CPU_RESCTRL
3534 ONE("cpu_resctrl_groups", S_IRUGO
, proc_resctrl_show
),
3536 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3537 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3538 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3540 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3541 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3543 #ifdef CONFIG_FAULT_INJECTION
3544 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3545 REG("fail-nth", 0644, proc_fail_nth_operations
),
3547 #ifdef CONFIG_TASK_IO_ACCOUNTING
3548 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3550 #ifdef CONFIG_USER_NS
3551 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3552 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3553 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3554 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3556 #ifdef CONFIG_LIVEPATCH
3557 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3559 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3560 ONE("arch_status", S_IRUGO
, proc_pid_arch_status
),
3564 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3566 return proc_pident_readdir(file
, ctx
,
3567 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3570 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3572 return proc_pident_lookup(dir
, dentry
,
3574 tid_base_stuff
+ ARRAY_SIZE(tid_base_stuff
));
3577 static const struct file_operations proc_tid_base_operations
= {
3578 .read
= generic_read_dir
,
3579 .iterate_shared
= proc_tid_base_readdir
,
3580 .llseek
= generic_file_llseek
,
3583 static const struct inode_operations proc_tid_base_inode_operations
= {
3584 .lookup
= proc_tid_base_lookup
,
3585 .getattr
= pid_getattr
,
3586 .setattr
= proc_setattr
,
3589 static struct dentry
*proc_task_instantiate(struct dentry
*dentry
,
3590 struct task_struct
*task
, const void *ptr
)
3592 struct inode
*inode
;
3593 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3595 return ERR_PTR(-ENOENT
);
3597 inode
->i_op
= &proc_tid_base_inode_operations
;
3598 inode
->i_fop
= &proc_tid_base_operations
;
3599 inode
->i_flags
|= S_IMMUTABLE
;
3601 set_nlink(inode
, nlink_tid
);
3602 pid_update_inode(task
, inode
);
3604 d_set_d_op(dentry
, &pid_dentry_operations
);
3605 return d_splice_alias(inode
, dentry
);
3608 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3610 struct task_struct
*task
;
3611 struct task_struct
*leader
= get_proc_task(dir
);
3613 struct pid_namespace
*ns
;
3614 struct dentry
*result
= ERR_PTR(-ENOENT
);
3619 tid
= name_to_int(&dentry
->d_name
);
3623 ns
= dentry
->d_sb
->s_fs_info
;
3625 task
= find_task_by_pid_ns(tid
, ns
);
3627 get_task_struct(task
);
3631 if (!same_thread_group(leader
, task
))
3634 result
= proc_task_instantiate(dentry
, task
, NULL
);
3636 put_task_struct(task
);
3638 put_task_struct(leader
);
3644 * Find the first tid of a thread group to return to user space.
3646 * Usually this is just the thread group leader, but if the users
3647 * buffer was too small or there was a seek into the middle of the
3648 * directory we have more work todo.
3650 * In the case of a short read we start with find_task_by_pid.
3652 * In the case of a seek we start with the leader and walk nr
3655 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3656 struct pid_namespace
*ns
)
3658 struct task_struct
*pos
, *task
;
3659 unsigned long nr
= f_pos
;
3661 if (nr
!= f_pos
) /* 32bit overflow? */
3665 task
= pid_task(pid
, PIDTYPE_PID
);
3669 /* Attempt to start with the tid of a thread */
3671 pos
= find_task_by_pid_ns(tid
, ns
);
3672 if (pos
&& same_thread_group(pos
, task
))
3676 /* If nr exceeds the number of threads there is nothing todo */
3677 if (nr
>= get_nr_threads(task
))
3680 /* If we haven't found our starting place yet start
3681 * with the leader and walk nr threads forward.
3683 pos
= task
= task
->group_leader
;
3687 } while_each_thread(task
, pos
);
3692 get_task_struct(pos
);
3699 * Find the next thread in the thread list.
3700 * Return NULL if there is an error or no next thread.
3702 * The reference to the input task_struct is released.
3704 static struct task_struct
*next_tid(struct task_struct
*start
)
3706 struct task_struct
*pos
= NULL
;
3708 if (pid_alive(start
)) {
3709 pos
= next_thread(start
);
3710 if (thread_group_leader(pos
))
3713 get_task_struct(pos
);
3716 put_task_struct(start
);
3720 /* for the /proc/TGID/task/ directories */
3721 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3723 struct inode
*inode
= file_inode(file
);
3724 struct task_struct
*task
;
3725 struct pid_namespace
*ns
;
3728 if (proc_inode_is_dead(inode
))
3731 if (!dir_emit_dots(file
, ctx
))
3734 /* f_version caches the tgid value that the last readdir call couldn't
3735 * return. lseek aka telldir automagically resets f_version to 0.
3737 ns
= proc_pid_ns(inode
);
3738 tid
= (int)file
->f_version
;
3739 file
->f_version
= 0;
3740 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3742 task
= next_tid(task
), ctx
->pos
++) {
3745 tid
= task_pid_nr_ns(task
, ns
);
3746 len
= snprintf(name
, sizeof(name
), "%u", tid
);
3747 if (!proc_fill_cache(file
, ctx
, name
, len
,
3748 proc_task_instantiate
, task
, NULL
)) {
3749 /* returning this tgid failed, save it as the first
3750 * pid for the next readir call */
3751 file
->f_version
= (u64
)tid
;
3752 put_task_struct(task
);
3760 static int proc_task_getattr(const struct path
*path
, struct kstat
*stat
,
3761 u32 request_mask
, unsigned int query_flags
)
3763 struct inode
*inode
= d_inode(path
->dentry
);
3764 struct task_struct
*p
= get_proc_task(inode
);
3765 generic_fillattr(inode
, stat
);
3768 stat
->nlink
+= get_nr_threads(p
);
3775 static const struct inode_operations proc_task_inode_operations
= {
3776 .lookup
= proc_task_lookup
,
3777 .getattr
= proc_task_getattr
,
3778 .setattr
= proc_setattr
,
3779 .permission
= proc_pid_permission
,
3782 static const struct file_operations proc_task_operations
= {
3783 .read
= generic_read_dir
,
3784 .iterate_shared
= proc_task_readdir
,
3785 .llseek
= generic_file_llseek
,
3788 void __init
set_proc_pid_nlink(void)
3790 nlink_tid
= pid_entry_nlink(tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3791 nlink_tgid
= pid_entry_nlink(tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));