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/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/posix-timers.h>
96 #include <linux/time_namespace.h>
97 #include <linux/resctrl.h>
98 #include <linux/cn_proc.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
)
390 if (ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
391 wchan
= get_wchan(task
);
396 seq_printf(m
, "%ps", (void *) wchan
);
402 #endif /* CONFIG_KALLSYMS */
404 static int lock_trace(struct task_struct
*task
)
406 int err
= down_read_killable(&task
->signal
->exec_update_lock
);
409 if (!ptrace_may_access(task
, PTRACE_MODE_ATTACH_FSCREDS
)) {
410 up_read(&task
->signal
->exec_update_lock
);
416 static void unlock_trace(struct task_struct
*task
)
418 up_read(&task
->signal
->exec_update_lock
);
421 #ifdef CONFIG_STACKTRACE
423 #define MAX_STACK_TRACE_DEPTH 64
425 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
426 struct pid
*pid
, struct task_struct
*task
)
428 unsigned long *entries
;
432 * The ability to racily run the kernel stack unwinder on a running task
433 * and then observe the unwinder output is scary; while it is useful for
434 * debugging kernel issues, it can also allow an attacker to leak kernel
436 * Doing this in a manner that is at least safe from races would require
437 * some work to ensure that the remote task can not be scheduled; and
438 * even then, this would still expose the unwinder as local attack
440 * Therefore, this interface is restricted to root.
442 if (!file_ns_capable(m
->file
, &init_user_ns
, CAP_SYS_ADMIN
))
445 entries
= kmalloc_array(MAX_STACK_TRACE_DEPTH
, sizeof(*entries
),
450 err
= lock_trace(task
);
452 unsigned int i
, nr_entries
;
454 nr_entries
= stack_trace_save_tsk(task
, entries
,
455 MAX_STACK_TRACE_DEPTH
, 0);
457 for (i
= 0; i
< nr_entries
; i
++) {
458 seq_printf(m
, "[<0>] %pB\n", (void *)entries
[i
]);
469 #ifdef CONFIG_SCHED_INFO
471 * Provides /proc/PID/schedstat
473 static int proc_pid_schedstat(struct seq_file
*m
, struct pid_namespace
*ns
,
474 struct pid
*pid
, struct task_struct
*task
)
476 if (unlikely(!sched_info_on()))
477 seq_puts(m
, "0 0 0\n");
479 seq_printf(m
, "%llu %llu %lu\n",
480 (unsigned long long)task
->se
.sum_exec_runtime
,
481 (unsigned long long)task
->sched_info
.run_delay
,
482 task
->sched_info
.pcount
);
488 #ifdef CONFIG_LATENCYTOP
489 static int lstats_show_proc(struct seq_file
*m
, void *v
)
492 struct inode
*inode
= m
->private;
493 struct task_struct
*task
= get_proc_task(inode
);
497 seq_puts(m
, "Latency Top version : v0.1\n");
498 for (i
= 0; i
< LT_SAVECOUNT
; i
++) {
499 struct latency_record
*lr
= &task
->latency_record
[i
];
500 if (lr
->backtrace
[0]) {
502 seq_printf(m
, "%i %li %li",
503 lr
->count
, lr
->time
, lr
->max
);
504 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
505 unsigned long bt
= lr
->backtrace
[q
];
509 seq_printf(m
, " %ps", (void *)bt
);
515 put_task_struct(task
);
519 static int lstats_open(struct inode
*inode
, struct file
*file
)
521 return single_open(file
, lstats_show_proc
, inode
);
524 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
525 size_t count
, loff_t
*offs
)
527 struct task_struct
*task
= get_proc_task(file_inode(file
));
531 clear_tsk_latency_tracing(task
);
532 put_task_struct(task
);
537 static const struct file_operations proc_lstats_operations
= {
540 .write
= lstats_write
,
542 .release
= single_release
,
547 static int proc_oom_score(struct seq_file
*m
, struct pid_namespace
*ns
,
548 struct pid
*pid
, struct task_struct
*task
)
550 unsigned long totalpages
= totalram_pages() + total_swap_pages
;
551 unsigned long points
= 0;
554 badness
= oom_badness(task
, totalpages
);
556 * Special case OOM_SCORE_ADJ_MIN for all others scale the
557 * badness value into [0, 2000] range which we have been
558 * exporting for a long time so userspace might depend on it.
560 if (badness
!= LONG_MIN
)
561 points
= (1000 + badness
* 1000 / (long)totalpages
) * 2 / 3;
563 seq_printf(m
, "%lu\n", points
);
573 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
574 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
575 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
576 [RLIMIT_DATA
] = {"Max data size", "bytes"},
577 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
578 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
579 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
580 [RLIMIT_NPROC
] = {"Max processes", "processes"},
581 [RLIMIT_NOFILE
] = {"Max open files", "files"},
582 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
583 [RLIMIT_AS
] = {"Max address space", "bytes"},
584 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
585 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
586 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
587 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
588 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
589 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
592 /* Display limits for a process */
593 static int proc_pid_limits(struct seq_file
*m
, struct pid_namespace
*ns
,
594 struct pid
*pid
, struct task_struct
*task
)
599 struct rlimit rlim
[RLIM_NLIMITS
];
601 if (!lock_task_sighand(task
, &flags
))
603 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
604 unlock_task_sighand(task
, &flags
);
607 * print the file header
614 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
615 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
616 seq_printf(m
, "%-25s %-20s ",
617 lnames
[i
].name
, "unlimited");
619 seq_printf(m
, "%-25s %-20lu ",
620 lnames
[i
].name
, rlim
[i
].rlim_cur
);
622 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
623 seq_printf(m
, "%-20s ", "unlimited");
625 seq_printf(m
, "%-20lu ", rlim
[i
].rlim_max
);
628 seq_printf(m
, "%-10s\n", lnames
[i
].unit
);
636 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
637 static int proc_pid_syscall(struct seq_file
*m
, struct pid_namespace
*ns
,
638 struct pid
*pid
, struct task_struct
*task
)
640 struct syscall_info info
;
641 u64
*args
= &info
.data
.args
[0];
644 res
= lock_trace(task
);
648 if (task_current_syscall(task
, &info
))
649 seq_puts(m
, "running\n");
650 else if (info
.data
.nr
< 0)
651 seq_printf(m
, "%d 0x%llx 0x%llx\n",
652 info
.data
.nr
, info
.sp
, info
.data
.instruction_pointer
);
655 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
657 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
658 info
.sp
, info
.data
.instruction_pointer
);
663 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
665 /************************************************************************/
666 /* Here the fs part begins */
667 /************************************************************************/
669 /* permission checks */
670 static int proc_fd_access_allowed(struct inode
*inode
)
672 struct task_struct
*task
;
674 /* Allow access to a task's file descriptors if it is us or we
675 * may use ptrace attach to the process and find out that
678 task
= get_proc_task(inode
);
680 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
681 put_task_struct(task
);
686 int proc_setattr(struct user_namespace
*mnt_userns
, struct dentry
*dentry
,
690 struct inode
*inode
= d_inode(dentry
);
692 if (attr
->ia_valid
& ATTR_MODE
)
695 error
= setattr_prepare(&init_user_ns
, dentry
, attr
);
699 setattr_copy(&init_user_ns
, inode
, attr
);
700 mark_inode_dirty(inode
);
705 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
706 * or euid/egid (for hide_pid_min=2)?
708 static bool has_pid_permissions(struct proc_fs_info
*fs_info
,
709 struct task_struct
*task
,
710 enum proc_hidepid hide_pid_min
)
713 * If 'hidpid' mount option is set force a ptrace check,
714 * we indicate that we are using a filesystem syscall
715 * by passing PTRACE_MODE_READ_FSCREDS
717 if (fs_info
->hide_pid
== HIDEPID_NOT_PTRACEABLE
)
718 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
720 if (fs_info
->hide_pid
< hide_pid_min
)
722 if (in_group_p(fs_info
->pid_gid
))
724 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
728 static int proc_pid_permission(struct user_namespace
*mnt_userns
,
729 struct inode
*inode
, int mask
)
731 struct proc_fs_info
*fs_info
= proc_sb_info(inode
->i_sb
);
732 struct task_struct
*task
;
735 task
= get_proc_task(inode
);
738 has_perms
= has_pid_permissions(fs_info
, task
, HIDEPID_NO_ACCESS
);
739 put_task_struct(task
);
742 if (fs_info
->hide_pid
== HIDEPID_INVISIBLE
) {
744 * Let's make getdents(), stat(), and open()
745 * consistent with each other. If a process
746 * may not stat() a file, it shouldn't be seen
754 return generic_permission(&init_user_ns
, inode
, mask
);
759 static const struct inode_operations proc_def_inode_operations
= {
760 .setattr
= proc_setattr
,
763 static int proc_single_show(struct seq_file
*m
, void *v
)
765 struct inode
*inode
= m
->private;
766 struct pid_namespace
*ns
= proc_pid_ns(inode
->i_sb
);
767 struct pid
*pid
= proc_pid(inode
);
768 struct task_struct
*task
;
771 task
= get_pid_task(pid
, PIDTYPE_PID
);
775 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
777 put_task_struct(task
);
781 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
783 return single_open(filp
, proc_single_show
, inode
);
786 static const struct file_operations proc_single_file_operations
= {
787 .open
= proc_single_open
,
790 .release
= single_release
,
794 struct mm_struct
*proc_mem_open(struct inode
*inode
, unsigned int mode
)
796 struct task_struct
*task
= get_proc_task(inode
);
797 struct mm_struct
*mm
= ERR_PTR(-ESRCH
);
800 mm
= mm_access(task
, mode
| PTRACE_MODE_FSCREDS
);
801 put_task_struct(task
);
803 if (!IS_ERR_OR_NULL(mm
)) {
804 /* ensure this mm_struct can't be freed */
806 /* but do not pin its memory */
814 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
816 struct mm_struct
*mm
= proc_mem_open(inode
, mode
);
821 file
->private_data
= mm
;
825 static int mem_open(struct inode
*inode
, struct file
*file
)
827 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
829 /* OK to pass negative loff_t, we can catch out-of-range */
830 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
835 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
836 size_t count
, loff_t
*ppos
, int write
)
838 struct mm_struct
*mm
= file
->private_data
;
839 unsigned long addr
= *ppos
;
847 page
= (char *)__get_free_page(GFP_KERNEL
);
852 if (!mmget_not_zero(mm
))
855 flags
= FOLL_FORCE
| (write
? FOLL_WRITE
: 0);
858 size_t this_len
= min_t(size_t, count
, PAGE_SIZE
);
860 if (write
&& copy_from_user(page
, buf
, this_len
)) {
865 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, flags
);
872 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
886 free_page((unsigned long) page
);
890 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
891 size_t count
, loff_t
*ppos
)
893 return mem_rw(file
, buf
, count
, ppos
, 0);
896 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
897 size_t count
, loff_t
*ppos
)
899 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
902 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
906 file
->f_pos
= offset
;
909 file
->f_pos
+= offset
;
914 force_successful_syscall_return();
918 static int mem_release(struct inode
*inode
, struct file
*file
)
920 struct mm_struct
*mm
= file
->private_data
;
926 static const struct file_operations proc_mem_operations
= {
931 .release
= mem_release
,
934 static int environ_open(struct inode
*inode
, struct file
*file
)
936 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
939 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
940 size_t count
, loff_t
*ppos
)
943 unsigned long src
= *ppos
;
945 struct mm_struct
*mm
= file
->private_data
;
946 unsigned long env_start
, env_end
;
948 /* Ensure the process spawned far enough to have an environment. */
949 if (!mm
|| !mm
->env_end
)
952 page
= (char *)__get_free_page(GFP_KERNEL
);
957 if (!mmget_not_zero(mm
))
960 spin_lock(&mm
->arg_lock
);
961 env_start
= mm
->env_start
;
962 env_end
= mm
->env_end
;
963 spin_unlock(&mm
->arg_lock
);
966 size_t this_len
, max_len
;
969 if (src
>= (env_end
- env_start
))
972 this_len
= env_end
- (env_start
+ src
);
974 max_len
= min_t(size_t, PAGE_SIZE
, count
);
975 this_len
= min(max_len
, this_len
);
977 retval
= access_remote_vm(mm
, (env_start
+ src
), page
, this_len
, FOLL_ANON
);
984 if (copy_to_user(buf
, page
, retval
)) {
998 free_page((unsigned long) page
);
1002 static const struct file_operations proc_environ_operations
= {
1003 .open
= environ_open
,
1004 .read
= environ_read
,
1005 .llseek
= generic_file_llseek
,
1006 .release
= mem_release
,
1009 static int auxv_open(struct inode
*inode
, struct file
*file
)
1011 return __mem_open(inode
, file
, PTRACE_MODE_READ_FSCREDS
);
1014 static ssize_t
auxv_read(struct file
*file
, char __user
*buf
,
1015 size_t count
, loff_t
*ppos
)
1017 struct mm_struct
*mm
= file
->private_data
;
1018 unsigned int nwords
= 0;
1024 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
1025 return simple_read_from_buffer(buf
, count
, ppos
, mm
->saved_auxv
,
1026 nwords
* sizeof(mm
->saved_auxv
[0]));
1029 static const struct file_operations proc_auxv_operations
= {
1032 .llseek
= generic_file_llseek
,
1033 .release
= mem_release
,
1036 static ssize_t
oom_adj_read(struct file
*file
, char __user
*buf
, size_t count
,
1039 struct task_struct
*task
= get_proc_task(file_inode(file
));
1040 char buffer
[PROC_NUMBUF
];
1041 int oom_adj
= OOM_ADJUST_MIN
;
1046 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MAX
)
1047 oom_adj
= OOM_ADJUST_MAX
;
1049 oom_adj
= (task
->signal
->oom_score_adj
* -OOM_DISABLE
) /
1051 put_task_struct(task
);
1052 if (oom_adj
> OOM_ADJUST_MAX
)
1053 oom_adj
= OOM_ADJUST_MAX
;
1054 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1055 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1058 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1060 struct mm_struct
*mm
= NULL
;
1061 struct task_struct
*task
;
1064 task
= get_proc_task(file_inode(file
));
1068 mutex_lock(&oom_adj_mutex
);
1070 if (oom_adj
< task
->signal
->oom_score_adj
&&
1071 !capable(CAP_SYS_RESOURCE
)) {
1076 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1077 * /proc/pid/oom_score_adj instead.
1079 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1080 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1083 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1084 !capable(CAP_SYS_RESOURCE
)) {
1091 * Make sure we will check other processes sharing the mm if this is
1092 * not vfrok which wants its own oom_score_adj.
1093 * pin the mm so it doesn't go away and get reused after task_unlock
1095 if (!task
->vfork_done
) {
1096 struct task_struct
*p
= find_lock_task_mm(task
);
1099 if (test_bit(MMF_MULTIPROCESS
, &p
->mm
->flags
)) {
1107 task
->signal
->oom_score_adj
= oom_adj
;
1108 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1109 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1110 trace_oom_score_adj_update(task
);
1113 struct task_struct
*p
;
1116 for_each_process(p
) {
1117 if (same_thread_group(task
, p
))
1120 /* do not touch kernel threads or the global init */
1121 if (p
->flags
& PF_KTHREAD
|| is_global_init(p
))
1125 if (!p
->vfork_done
&& process_shares_mm(p
, mm
)) {
1126 p
->signal
->oom_score_adj
= oom_adj
;
1127 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1128 p
->signal
->oom_score_adj_min
= (short)oom_adj
;
1136 mutex_unlock(&oom_adj_mutex
);
1137 put_task_struct(task
);
1142 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1143 * kernels. The effective policy is defined by oom_score_adj, which has a
1144 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1145 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1146 * Processes that become oom disabled via oom_adj will still be oom disabled
1147 * with this implementation.
1149 * oom_adj cannot be removed since existing userspace binaries use it.
1151 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1152 size_t count
, loff_t
*ppos
)
1154 char buffer
[PROC_NUMBUF
];
1158 memset(buffer
, 0, sizeof(buffer
));
1159 if (count
> sizeof(buffer
) - 1)
1160 count
= sizeof(buffer
) - 1;
1161 if (copy_from_user(buffer
, buf
, count
)) {
1166 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1169 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1170 oom_adj
!= OOM_DISABLE
) {
1176 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1177 * value is always attainable.
1179 if (oom_adj
== OOM_ADJUST_MAX
)
1180 oom_adj
= OOM_SCORE_ADJ_MAX
;
1182 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1184 err
= __set_oom_adj(file
, oom_adj
, true);
1186 return err
< 0 ? err
: count
;
1189 static const struct file_operations proc_oom_adj_operations
= {
1190 .read
= oom_adj_read
,
1191 .write
= oom_adj_write
,
1192 .llseek
= generic_file_llseek
,
1195 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1196 size_t count
, loff_t
*ppos
)
1198 struct task_struct
*task
= get_proc_task(file_inode(file
));
1199 char buffer
[PROC_NUMBUF
];
1200 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1205 oom_score_adj
= task
->signal
->oom_score_adj
;
1206 put_task_struct(task
);
1207 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1208 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1211 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1212 size_t count
, loff_t
*ppos
)
1214 char buffer
[PROC_NUMBUF
];
1218 memset(buffer
, 0, sizeof(buffer
));
1219 if (count
> sizeof(buffer
) - 1)
1220 count
= sizeof(buffer
) - 1;
1221 if (copy_from_user(buffer
, buf
, count
)) {
1226 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1229 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1230 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1235 err
= __set_oom_adj(file
, oom_score_adj
, false);
1237 return err
< 0 ? err
: count
;
1240 static const struct file_operations proc_oom_score_adj_operations
= {
1241 .read
= oom_score_adj_read
,
1242 .write
= oom_score_adj_write
,
1243 .llseek
= default_llseek
,
1247 #define TMPBUFLEN 11
1248 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1249 size_t count
, loff_t
*ppos
)
1251 struct inode
* inode
= file_inode(file
);
1252 struct task_struct
*task
= get_proc_task(inode
);
1254 char tmpbuf
[TMPBUFLEN
];
1258 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1259 from_kuid(file
->f_cred
->user_ns
,
1260 audit_get_loginuid(task
)));
1261 put_task_struct(task
);
1262 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1265 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1266 size_t count
, loff_t
*ppos
)
1268 struct inode
* inode
= file_inode(file
);
1273 /* Don't let kthreads write their own loginuid */
1274 if (current
->flags
& PF_KTHREAD
)
1278 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1285 /* No partial writes. */
1289 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1293 /* is userspace tring to explicitly UNSET the loginuid? */
1294 if (loginuid
== AUDIT_UID_UNSET
) {
1295 kloginuid
= INVALID_UID
;
1297 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1298 if (!uid_valid(kloginuid
))
1302 rv
= audit_set_loginuid(kloginuid
);
1308 static const struct file_operations proc_loginuid_operations
= {
1309 .read
= proc_loginuid_read
,
1310 .write
= proc_loginuid_write
,
1311 .llseek
= generic_file_llseek
,
1314 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1315 size_t count
, loff_t
*ppos
)
1317 struct inode
* inode
= file_inode(file
);
1318 struct task_struct
*task
= get_proc_task(inode
);
1320 char tmpbuf
[TMPBUFLEN
];
1324 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1325 audit_get_sessionid(task
));
1326 put_task_struct(task
);
1327 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1330 static const struct file_operations proc_sessionid_operations
= {
1331 .read
= proc_sessionid_read
,
1332 .llseek
= generic_file_llseek
,
1336 #ifdef CONFIG_FAULT_INJECTION
1337 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1338 size_t count
, loff_t
*ppos
)
1340 struct task_struct
*task
= get_proc_task(file_inode(file
));
1341 char buffer
[PROC_NUMBUF
];
1347 make_it_fail
= task
->make_it_fail
;
1348 put_task_struct(task
);
1350 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1352 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1355 static ssize_t
proc_fault_inject_write(struct file
* file
,
1356 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1358 struct task_struct
*task
;
1359 char buffer
[PROC_NUMBUF
];
1363 if (!capable(CAP_SYS_RESOURCE
))
1365 memset(buffer
, 0, sizeof(buffer
));
1366 if (count
> sizeof(buffer
) - 1)
1367 count
= sizeof(buffer
) - 1;
1368 if (copy_from_user(buffer
, buf
, count
))
1370 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1373 if (make_it_fail
< 0 || make_it_fail
> 1)
1376 task
= get_proc_task(file_inode(file
));
1379 task
->make_it_fail
= make_it_fail
;
1380 put_task_struct(task
);
1385 static const struct file_operations proc_fault_inject_operations
= {
1386 .read
= proc_fault_inject_read
,
1387 .write
= proc_fault_inject_write
,
1388 .llseek
= generic_file_llseek
,
1391 static ssize_t
proc_fail_nth_write(struct file
*file
, const char __user
*buf
,
1392 size_t count
, loff_t
*ppos
)
1394 struct task_struct
*task
;
1398 err
= kstrtouint_from_user(buf
, count
, 0, &n
);
1402 task
= get_proc_task(file_inode(file
));
1406 put_task_struct(task
);
1411 static ssize_t
proc_fail_nth_read(struct file
*file
, char __user
*buf
,
1412 size_t count
, loff_t
*ppos
)
1414 struct task_struct
*task
;
1415 char numbuf
[PROC_NUMBUF
];
1418 task
= get_proc_task(file_inode(file
));
1421 len
= snprintf(numbuf
, sizeof(numbuf
), "%u\n", task
->fail_nth
);
1422 put_task_struct(task
);
1423 return simple_read_from_buffer(buf
, count
, ppos
, numbuf
, len
);
1426 static const struct file_operations proc_fail_nth_operations
= {
1427 .read
= proc_fail_nth_read
,
1428 .write
= proc_fail_nth_write
,
1433 #ifdef CONFIG_SCHED_DEBUG
1435 * Print out various scheduling related per-task fields:
1437 static int sched_show(struct seq_file
*m
, void *v
)
1439 struct inode
*inode
= m
->private;
1440 struct pid_namespace
*ns
= proc_pid_ns(inode
->i_sb
);
1441 struct task_struct
*p
;
1443 p
= get_proc_task(inode
);
1446 proc_sched_show_task(p
, ns
, m
);
1454 sched_write(struct file
*file
, const char __user
*buf
,
1455 size_t count
, loff_t
*offset
)
1457 struct inode
*inode
= file_inode(file
);
1458 struct task_struct
*p
;
1460 p
= get_proc_task(inode
);
1463 proc_sched_set_task(p
);
1470 static int sched_open(struct inode
*inode
, struct file
*filp
)
1472 return single_open(filp
, sched_show
, inode
);
1475 static const struct file_operations proc_pid_sched_operations
= {
1478 .write
= sched_write
,
1479 .llseek
= seq_lseek
,
1480 .release
= single_release
,
1485 #ifdef CONFIG_SCHED_AUTOGROUP
1487 * Print out autogroup related information:
1489 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1491 struct inode
*inode
= m
->private;
1492 struct task_struct
*p
;
1494 p
= get_proc_task(inode
);
1497 proc_sched_autogroup_show_task(p
, m
);
1505 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1506 size_t count
, loff_t
*offset
)
1508 struct inode
*inode
= file_inode(file
);
1509 struct task_struct
*p
;
1510 char buffer
[PROC_NUMBUF
];
1514 memset(buffer
, 0, sizeof(buffer
));
1515 if (count
> sizeof(buffer
) - 1)
1516 count
= sizeof(buffer
) - 1;
1517 if (copy_from_user(buffer
, buf
, count
))
1520 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1524 p
= get_proc_task(inode
);
1528 err
= proc_sched_autogroup_set_nice(p
, nice
);
1537 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1541 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1543 struct seq_file
*m
= filp
->private_data
;
1550 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1551 .open
= sched_autogroup_open
,
1553 .write
= sched_autogroup_write
,
1554 .llseek
= seq_lseek
,
1555 .release
= single_release
,
1558 #endif /* CONFIG_SCHED_AUTOGROUP */
1560 #ifdef CONFIG_TIME_NS
1561 static int timens_offsets_show(struct seq_file
*m
, void *v
)
1563 struct task_struct
*p
;
1565 p
= get_proc_task(file_inode(m
->file
));
1568 proc_timens_show_offsets(p
, m
);
1575 static ssize_t
timens_offsets_write(struct file
*file
, const char __user
*buf
,
1576 size_t count
, loff_t
*ppos
)
1578 struct inode
*inode
= file_inode(file
);
1579 struct proc_timens_offset offsets
[2];
1580 char *kbuf
= NULL
, *pos
, *next_line
;
1581 struct task_struct
*p
;
1584 /* Only allow < page size writes at the beginning of the file */
1585 if ((*ppos
!= 0) || (count
>= PAGE_SIZE
))
1588 /* Slurp in the user data */
1589 kbuf
= memdup_user_nul(buf
, count
);
1591 return PTR_ERR(kbuf
);
1593 /* Parse the user data */
1596 for (pos
= kbuf
; pos
; pos
= next_line
) {
1597 struct proc_timens_offset
*off
= &offsets
[noffsets
];
1601 /* Find the end of line and ensure we don't look past it */
1602 next_line
= strchr(pos
, '\n');
1606 if (*next_line
== '\0')
1610 err
= sscanf(pos
, "%9s %lld %lu", clock
,
1611 &off
->val
.tv_sec
, &off
->val
.tv_nsec
);
1612 if (err
!= 3 || off
->val
.tv_nsec
>= NSEC_PER_SEC
)
1615 clock
[sizeof(clock
) - 1] = 0;
1616 if (strcmp(clock
, "monotonic") == 0 ||
1617 strcmp(clock
, __stringify(CLOCK_MONOTONIC
)) == 0)
1618 off
->clockid
= CLOCK_MONOTONIC
;
1619 else if (strcmp(clock
, "boottime") == 0 ||
1620 strcmp(clock
, __stringify(CLOCK_BOOTTIME
)) == 0)
1621 off
->clockid
= CLOCK_BOOTTIME
;
1626 if (noffsets
== ARRAY_SIZE(offsets
)) {
1628 count
= next_line
- kbuf
;
1634 p
= get_proc_task(inode
);
1637 ret
= proc_timens_set_offset(file
, p
, offsets
, noffsets
);
1648 static int timens_offsets_open(struct inode
*inode
, struct file
*filp
)
1650 return single_open(filp
, timens_offsets_show
, inode
);
1653 static const struct file_operations proc_timens_offsets_operations
= {
1654 .open
= timens_offsets_open
,
1656 .write
= timens_offsets_write
,
1657 .llseek
= seq_lseek
,
1658 .release
= single_release
,
1660 #endif /* CONFIG_TIME_NS */
1662 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1663 size_t count
, loff_t
*offset
)
1665 struct inode
*inode
= file_inode(file
);
1666 struct task_struct
*p
;
1667 char buffer
[TASK_COMM_LEN
];
1668 const size_t maxlen
= sizeof(buffer
) - 1;
1670 memset(buffer
, 0, sizeof(buffer
));
1671 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1674 p
= get_proc_task(inode
);
1678 if (same_thread_group(current
, p
)) {
1679 set_task_comm(p
, buffer
);
1680 proc_comm_connector(p
);
1690 static int comm_show(struct seq_file
*m
, void *v
)
1692 struct inode
*inode
= m
->private;
1693 struct task_struct
*p
;
1695 p
= get_proc_task(inode
);
1699 proc_task_name(m
, p
, false);
1707 static int comm_open(struct inode
*inode
, struct file
*filp
)
1709 return single_open(filp
, comm_show
, inode
);
1712 static const struct file_operations proc_pid_set_comm_operations
= {
1715 .write
= comm_write
,
1716 .llseek
= seq_lseek
,
1717 .release
= single_release
,
1720 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1722 struct task_struct
*task
;
1723 struct file
*exe_file
;
1725 task
= get_proc_task(d_inode(dentry
));
1728 exe_file
= get_task_exe_file(task
);
1729 put_task_struct(task
);
1731 *exe_path
= exe_file
->f_path
;
1732 path_get(&exe_file
->f_path
);
1739 static const char *proc_pid_get_link(struct dentry
*dentry
,
1740 struct inode
*inode
,
1741 struct delayed_call
*done
)
1744 int error
= -EACCES
;
1747 return ERR_PTR(-ECHILD
);
1749 /* Are we allowed to snoop on the tasks file descriptors? */
1750 if (!proc_fd_access_allowed(inode
))
1753 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1757 error
= nd_jump_link(&path
);
1759 return ERR_PTR(error
);
1762 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1764 char *tmp
= (char *)__get_free_page(GFP_KERNEL
);
1771 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1772 len
= PTR_ERR(pathname
);
1773 if (IS_ERR(pathname
))
1775 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1779 if (copy_to_user(buffer
, pathname
, len
))
1782 free_page((unsigned long)tmp
);
1786 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1788 int error
= -EACCES
;
1789 struct inode
*inode
= d_inode(dentry
);
1792 /* Are we allowed to snoop on the tasks file descriptors? */
1793 if (!proc_fd_access_allowed(inode
))
1796 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1800 error
= do_proc_readlink(&path
, buffer
, buflen
);
1806 const struct inode_operations proc_pid_link_inode_operations
= {
1807 .readlink
= proc_pid_readlink
,
1808 .get_link
= proc_pid_get_link
,
1809 .setattr
= proc_setattr
,
1813 /* building an inode */
1815 void task_dump_owner(struct task_struct
*task
, umode_t mode
,
1816 kuid_t
*ruid
, kgid_t
*rgid
)
1818 /* Depending on the state of dumpable compute who should own a
1819 * proc file for a task.
1821 const struct cred
*cred
;
1825 if (unlikely(task
->flags
& PF_KTHREAD
)) {
1826 *ruid
= GLOBAL_ROOT_UID
;
1827 *rgid
= GLOBAL_ROOT_GID
;
1831 /* Default to the tasks effective ownership */
1833 cred
= __task_cred(task
);
1839 * Before the /proc/pid/status file was created the only way to read
1840 * the effective uid of a /process was to stat /proc/pid. Reading
1841 * /proc/pid/status is slow enough that procps and other packages
1842 * kept stating /proc/pid. To keep the rules in /proc simple I have
1843 * made this apply to all per process world readable and executable
1846 if (mode
!= (S_IFDIR
|S_IRUGO
|S_IXUGO
)) {
1847 struct mm_struct
*mm
;
1850 /* Make non-dumpable tasks owned by some root */
1852 if (get_dumpable(mm
) != SUID_DUMP_USER
) {
1853 struct user_namespace
*user_ns
= mm
->user_ns
;
1855 uid
= make_kuid(user_ns
, 0);
1856 if (!uid_valid(uid
))
1857 uid
= GLOBAL_ROOT_UID
;
1859 gid
= make_kgid(user_ns
, 0);
1860 if (!gid_valid(gid
))
1861 gid
= GLOBAL_ROOT_GID
;
1864 uid
= GLOBAL_ROOT_UID
;
1865 gid
= GLOBAL_ROOT_GID
;
1873 void proc_pid_evict_inode(struct proc_inode
*ei
)
1875 struct pid
*pid
= ei
->pid
;
1877 if (S_ISDIR(ei
->vfs_inode
.i_mode
)) {
1878 spin_lock(&pid
->lock
);
1879 hlist_del_init_rcu(&ei
->sibling_inodes
);
1880 spin_unlock(&pid
->lock
);
1886 struct inode
*proc_pid_make_inode(struct super_block
* sb
,
1887 struct task_struct
*task
, umode_t mode
)
1889 struct inode
* inode
;
1890 struct proc_inode
*ei
;
1893 /* We need a new inode */
1895 inode
= new_inode(sb
);
1901 inode
->i_mode
= mode
;
1902 inode
->i_ino
= get_next_ino();
1903 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= current_time(inode
);
1904 inode
->i_op
= &proc_def_inode_operations
;
1907 * grab the reference to task.
1909 pid
= get_task_pid(task
, PIDTYPE_PID
);
1913 /* Let the pid remember us for quick removal */
1915 if (S_ISDIR(mode
)) {
1916 spin_lock(&pid
->lock
);
1917 hlist_add_head_rcu(&ei
->sibling_inodes
, &pid
->inodes
);
1918 spin_unlock(&pid
->lock
);
1921 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1922 security_task_to_inode(task
, inode
);
1932 int pid_getattr(struct user_namespace
*mnt_userns
, const struct path
*path
,
1933 struct kstat
*stat
, u32 request_mask
, unsigned int query_flags
)
1935 struct inode
*inode
= d_inode(path
->dentry
);
1936 struct proc_fs_info
*fs_info
= proc_sb_info(inode
->i_sb
);
1937 struct task_struct
*task
;
1939 generic_fillattr(&init_user_ns
, inode
, stat
);
1941 stat
->uid
= GLOBAL_ROOT_UID
;
1942 stat
->gid
= GLOBAL_ROOT_GID
;
1944 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1946 if (!has_pid_permissions(fs_info
, task
, HIDEPID_INVISIBLE
)) {
1949 * This doesn't prevent learning whether PID exists,
1950 * it only makes getattr() consistent with readdir().
1954 task_dump_owner(task
, inode
->i_mode
, &stat
->uid
, &stat
->gid
);
1963 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1965 void pid_update_inode(struct task_struct
*task
, struct inode
*inode
)
1967 task_dump_owner(task
, inode
->i_mode
, &inode
->i_uid
, &inode
->i_gid
);
1969 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1970 security_task_to_inode(task
, inode
);
1974 * Rewrite the inode's ownerships here because the owning task may have
1975 * performed a setuid(), etc.
1978 static int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1980 struct inode
*inode
;
1981 struct task_struct
*task
;
1983 if (flags
& LOOKUP_RCU
)
1986 inode
= d_inode(dentry
);
1987 task
= get_proc_task(inode
);
1990 pid_update_inode(task
, inode
);
1991 put_task_struct(task
);
1997 static inline bool proc_inode_is_dead(struct inode
*inode
)
1999 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
2002 int pid_delete_dentry(const struct dentry
*dentry
)
2004 /* Is the task we represent dead?
2005 * If so, then don't put the dentry on the lru list,
2006 * kill it immediately.
2008 return proc_inode_is_dead(d_inode(dentry
));
2011 const struct dentry_operations pid_dentry_operations
=
2013 .d_revalidate
= pid_revalidate
,
2014 .d_delete
= pid_delete_dentry
,
2020 * Fill a directory entry.
2022 * If possible create the dcache entry and derive our inode number and
2023 * file type from dcache entry.
2025 * Since all of the proc inode numbers are dynamically generated, the inode
2026 * numbers do not exist until the inode is cache. This means creating
2027 * the dcache entry in readdir is necessary to keep the inode numbers
2028 * reported by readdir in sync with the inode numbers reported
2031 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
2032 const char *name
, unsigned int len
,
2033 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
2035 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
2036 struct qstr qname
= QSTR_INIT(name
, len
);
2037 struct inode
*inode
;
2038 unsigned type
= DT_UNKNOWN
;
2041 child
= d_hash_and_lookup(dir
, &qname
);
2043 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
2044 child
= d_alloc_parallel(dir
, &qname
, &wq
);
2046 goto end_instantiate
;
2047 if (d_in_lookup(child
)) {
2049 res
= instantiate(child
, task
, ptr
);
2050 d_lookup_done(child
);
2051 if (unlikely(res
)) {
2055 goto end_instantiate
;
2059 inode
= d_inode(child
);
2061 type
= inode
->i_mode
>> 12;
2064 return dir_emit(ctx
, name
, len
, ino
, type
);
2068 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2069 * which represent vma start and end addresses.
2071 static int dname_to_vma_addr(struct dentry
*dentry
,
2072 unsigned long *start
, unsigned long *end
)
2074 const char *str
= dentry
->d_name
.name
;
2075 unsigned long long sval
, eval
;
2078 if (str
[0] == '0' && str
[1] != '-')
2080 len
= _parse_integer(str
, 16, &sval
);
2081 if (len
& KSTRTOX_OVERFLOW
)
2083 if (sval
!= (unsigned long)sval
)
2091 if (str
[0] == '0' && str
[1])
2093 len
= _parse_integer(str
, 16, &eval
);
2094 if (len
& KSTRTOX_OVERFLOW
)
2096 if (eval
!= (unsigned long)eval
)
2109 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
2111 unsigned long vm_start
, vm_end
;
2112 bool exact_vma_exists
= false;
2113 struct mm_struct
*mm
= NULL
;
2114 struct task_struct
*task
;
2115 struct inode
*inode
;
2118 if (flags
& LOOKUP_RCU
)
2121 inode
= d_inode(dentry
);
2122 task
= get_proc_task(inode
);
2126 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
2127 if (IS_ERR_OR_NULL(mm
))
2130 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
2131 status
= mmap_read_lock_killable(mm
);
2133 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
,
2135 mmap_read_unlock(mm
);
2141 if (exact_vma_exists
) {
2142 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
2144 security_task_to_inode(task
, inode
);
2149 put_task_struct(task
);
2155 static const struct dentry_operations tid_map_files_dentry_operations
= {
2156 .d_revalidate
= map_files_d_revalidate
,
2157 .d_delete
= pid_delete_dentry
,
2160 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2162 unsigned long vm_start
, vm_end
;
2163 struct vm_area_struct
*vma
;
2164 struct task_struct
*task
;
2165 struct mm_struct
*mm
;
2169 task
= get_proc_task(d_inode(dentry
));
2173 mm
= get_task_mm(task
);
2174 put_task_struct(task
);
2178 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2182 rc
= mmap_read_lock_killable(mm
);
2187 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2188 if (vma
&& vma
->vm_file
) {
2189 *path
= vma
->vm_file
->f_path
;
2193 mmap_read_unlock(mm
);
2201 struct map_files_info
{
2202 unsigned long start
;
2208 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2209 * to concerns about how the symlinks may be used to bypass permissions on
2210 * ancestor directories in the path to the file in question.
2213 proc_map_files_get_link(struct dentry
*dentry
,
2214 struct inode
*inode
,
2215 struct delayed_call
*done
)
2217 if (!checkpoint_restore_ns_capable(&init_user_ns
))
2218 return ERR_PTR(-EPERM
);
2220 return proc_pid_get_link(dentry
, inode
, done
);
2224 * Identical to proc_pid_link_inode_operations except for get_link()
2226 static const struct inode_operations proc_map_files_link_inode_operations
= {
2227 .readlink
= proc_pid_readlink
,
2228 .get_link
= proc_map_files_get_link
,
2229 .setattr
= proc_setattr
,
2232 static struct dentry
*
2233 proc_map_files_instantiate(struct dentry
*dentry
,
2234 struct task_struct
*task
, const void *ptr
)
2236 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
2237 struct proc_inode
*ei
;
2238 struct inode
*inode
;
2240 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFLNK
|
2241 ((mode
& FMODE_READ
) ? S_IRUSR
: 0) |
2242 ((mode
& FMODE_WRITE
) ? S_IWUSR
: 0));
2244 return ERR_PTR(-ENOENT
);
2247 ei
->op
.proc_get_link
= map_files_get_link
;
2249 inode
->i_op
= &proc_map_files_link_inode_operations
;
2252 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2253 return d_splice_alias(inode
, dentry
);
2256 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2257 struct dentry
*dentry
, unsigned int flags
)
2259 unsigned long vm_start
, vm_end
;
2260 struct vm_area_struct
*vma
;
2261 struct task_struct
*task
;
2262 struct dentry
*result
;
2263 struct mm_struct
*mm
;
2265 result
= ERR_PTR(-ENOENT
);
2266 task
= get_proc_task(dir
);
2270 result
= ERR_PTR(-EACCES
);
2271 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2274 result
= ERR_PTR(-ENOENT
);
2275 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2278 mm
= get_task_mm(task
);
2282 result
= ERR_PTR(-EINTR
);
2283 if (mmap_read_lock_killable(mm
))
2286 result
= ERR_PTR(-ENOENT
);
2287 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2292 result
= proc_map_files_instantiate(dentry
, task
,
2293 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2296 mmap_read_unlock(mm
);
2300 put_task_struct(task
);
2305 static const struct inode_operations proc_map_files_inode_operations
= {
2306 .lookup
= proc_map_files_lookup
,
2307 .permission
= proc_fd_permission
,
2308 .setattr
= proc_setattr
,
2312 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2314 struct vm_area_struct
*vma
;
2315 struct task_struct
*task
;
2316 struct mm_struct
*mm
;
2317 unsigned long nr_files
, pos
, i
;
2318 GENRADIX(struct map_files_info
) fa
;
2319 struct map_files_info
*p
;
2325 task
= get_proc_task(file_inode(file
));
2330 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2334 if (!dir_emit_dots(file
, ctx
))
2337 mm
= get_task_mm(task
);
2341 ret
= mmap_read_lock_killable(mm
);
2350 * We need two passes here:
2352 * 1) Collect vmas of mapped files with mmap_lock taken
2353 * 2) Release mmap_lock and instantiate entries
2355 * otherwise we get lockdep complained, since filldir()
2356 * routine might require mmap_lock taken in might_fault().
2359 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2362 if (++pos
<= ctx
->pos
)
2365 p
= genradix_ptr_alloc(&fa
, nr_files
++, GFP_KERNEL
);
2368 mmap_read_unlock(mm
);
2373 p
->start
= vma
->vm_start
;
2374 p
->end
= vma
->vm_end
;
2375 p
->mode
= vma
->vm_file
->f_mode
;
2377 mmap_read_unlock(mm
);
2380 for (i
= 0; i
< nr_files
; i
++) {
2381 char buf
[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2384 p
= genradix_ptr(&fa
, i
);
2385 len
= snprintf(buf
, sizeof(buf
), "%lx-%lx", p
->start
, p
->end
);
2386 if (!proc_fill_cache(file
, ctx
,
2388 proc_map_files_instantiate
,
2390 (void *)(unsigned long)p
->mode
))
2396 put_task_struct(task
);
2402 static const struct file_operations proc_map_files_operations
= {
2403 .read
= generic_read_dir
,
2404 .iterate_shared
= proc_map_files_readdir
,
2405 .llseek
= generic_file_llseek
,
2408 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2409 struct timers_private
{
2411 struct task_struct
*task
;
2412 struct sighand_struct
*sighand
;
2413 struct pid_namespace
*ns
;
2414 unsigned long flags
;
2417 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2419 struct timers_private
*tp
= m
->private;
2421 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2423 return ERR_PTR(-ESRCH
);
2425 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2427 return ERR_PTR(-ESRCH
);
2429 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2432 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2434 struct timers_private
*tp
= m
->private;
2435 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2438 static void timers_stop(struct seq_file
*m
, void *v
)
2440 struct timers_private
*tp
= m
->private;
2443 unlock_task_sighand(tp
->task
, &tp
->flags
);
2448 put_task_struct(tp
->task
);
2453 static int show_timer(struct seq_file
*m
, void *v
)
2455 struct k_itimer
*timer
;
2456 struct timers_private
*tp
= m
->private;
2458 static const char * const nstr
[] = {
2459 [SIGEV_SIGNAL
] = "signal",
2460 [SIGEV_NONE
] = "none",
2461 [SIGEV_THREAD
] = "thread",
2464 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2465 notify
= timer
->it_sigev_notify
;
2467 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2468 seq_printf(m
, "signal: %d/%px\n",
2469 timer
->sigq
->info
.si_signo
,
2470 timer
->sigq
->info
.si_value
.sival_ptr
);
2471 seq_printf(m
, "notify: %s/%s.%d\n",
2472 nstr
[notify
& ~SIGEV_THREAD_ID
],
2473 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2474 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2475 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2480 static const struct seq_operations proc_timers_seq_ops
= {
2481 .start
= timers_start
,
2482 .next
= timers_next
,
2483 .stop
= timers_stop
,
2487 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2489 struct timers_private
*tp
;
2491 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2492 sizeof(struct timers_private
));
2496 tp
->pid
= proc_pid(inode
);
2497 tp
->ns
= proc_pid_ns(inode
->i_sb
);
2501 static const struct file_operations proc_timers_operations
= {
2502 .open
= proc_timers_open
,
2504 .llseek
= seq_lseek
,
2505 .release
= seq_release_private
,
2509 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2510 size_t count
, loff_t
*offset
)
2512 struct inode
*inode
= file_inode(file
);
2513 struct task_struct
*p
;
2517 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2521 p
= get_proc_task(inode
);
2527 if (!ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
)) {
2534 err
= security_task_setscheduler(p
);
2543 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2545 p
->timer_slack_ns
= slack_ns
;
2554 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2556 struct inode
*inode
= m
->private;
2557 struct task_struct
*p
;
2560 p
= get_proc_task(inode
);
2566 if (!ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
)) {
2573 err
= security_task_getscheduler(p
);
2579 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2588 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2590 return single_open(filp
, timerslack_ns_show
, inode
);
2593 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2594 .open
= timerslack_ns_open
,
2596 .write
= timerslack_ns_write
,
2597 .llseek
= seq_lseek
,
2598 .release
= single_release
,
2601 static struct dentry
*proc_pident_instantiate(struct dentry
*dentry
,
2602 struct task_struct
*task
, const void *ptr
)
2604 const struct pid_entry
*p
= ptr
;
2605 struct inode
*inode
;
2606 struct proc_inode
*ei
;
2608 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, p
->mode
);
2610 return ERR_PTR(-ENOENT
);
2613 if (S_ISDIR(inode
->i_mode
))
2614 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2616 inode
->i_op
= p
->iop
;
2618 inode
->i_fop
= p
->fop
;
2620 pid_update_inode(task
, inode
);
2621 d_set_d_op(dentry
, &pid_dentry_operations
);
2622 return d_splice_alias(inode
, dentry
);
2625 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2626 struct dentry
*dentry
,
2627 const struct pid_entry
*p
,
2628 const struct pid_entry
*end
)
2630 struct task_struct
*task
= get_proc_task(dir
);
2631 struct dentry
*res
= ERR_PTR(-ENOENT
);
2637 * Yes, it does not scale. And it should not. Don't add
2638 * new entries into /proc/<tgid>/ without very good reasons.
2640 for (; p
< end
; p
++) {
2641 if (p
->len
!= dentry
->d_name
.len
)
2643 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
)) {
2644 res
= proc_pident_instantiate(dentry
, task
, p
);
2648 put_task_struct(task
);
2653 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2654 const struct pid_entry
*ents
, unsigned int nents
)
2656 struct task_struct
*task
= get_proc_task(file_inode(file
));
2657 const struct pid_entry
*p
;
2662 if (!dir_emit_dots(file
, ctx
))
2665 if (ctx
->pos
>= nents
+ 2)
2668 for (p
= ents
+ (ctx
->pos
- 2); p
< ents
+ nents
; p
++) {
2669 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2670 proc_pident_instantiate
, task
, p
))
2675 put_task_struct(task
);
2679 #ifdef CONFIG_SECURITY
2680 static int proc_pid_attr_open(struct inode
*inode
, struct file
*file
)
2682 file
->private_data
= NULL
;
2683 __mem_open(inode
, file
, PTRACE_MODE_READ_FSCREDS
);
2687 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2688 size_t count
, loff_t
*ppos
)
2690 struct inode
* inode
= file_inode(file
);
2693 struct task_struct
*task
= get_proc_task(inode
);
2698 length
= security_getprocattr(task
, PROC_I(inode
)->op
.lsm
,
2699 (char*)file
->f_path
.dentry
->d_name
.name
,
2701 put_task_struct(task
);
2703 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2708 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2709 size_t count
, loff_t
*ppos
)
2711 struct inode
* inode
= file_inode(file
);
2712 struct task_struct
*task
;
2716 /* A task may only write when it was the opener. */
2717 if (file
->private_data
!= current
->mm
)
2721 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
2726 /* A task may only write its own attributes. */
2727 if (current
!= task
) {
2731 /* Prevent changes to overridden credentials. */
2732 if (current_cred() != current_real_cred()) {
2738 if (count
> PAGE_SIZE
)
2741 /* No partial writes. */
2745 page
= memdup_user(buf
, count
);
2751 /* Guard against adverse ptrace interaction */
2752 rv
= mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
);
2756 rv
= security_setprocattr(PROC_I(inode
)->op
.lsm
,
2757 file
->f_path
.dentry
->d_name
.name
, page
,
2759 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
2766 static const struct file_operations proc_pid_attr_operations
= {
2767 .open
= proc_pid_attr_open
,
2768 .read
= proc_pid_attr_read
,
2769 .write
= proc_pid_attr_write
,
2770 .llseek
= generic_file_llseek
,
2771 .release
= mem_release
,
2774 #define LSM_DIR_OPS(LSM) \
2775 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2776 struct dir_context *ctx) \
2778 return proc_pident_readdir(filp, ctx, \
2779 LSM##_attr_dir_stuff, \
2780 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2783 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2784 .read = generic_read_dir, \
2785 .iterate = proc_##LSM##_attr_dir_iterate, \
2786 .llseek = default_llseek, \
2789 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2790 struct dentry *dentry, unsigned int flags) \
2792 return proc_pident_lookup(dir, dentry, \
2793 LSM##_attr_dir_stuff, \
2794 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2797 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2798 .lookup = proc_##LSM##_attr_dir_lookup, \
2799 .getattr = pid_getattr, \
2800 .setattr = proc_setattr, \
2803 #ifdef CONFIG_SECURITY_SMACK
2804 static const struct pid_entry smack_attr_dir_stuff
[] = {
2805 ATTR("smack", "current", 0666),
2810 #ifdef CONFIG_SECURITY_APPARMOR
2811 static const struct pid_entry apparmor_attr_dir_stuff
[] = {
2812 ATTR("apparmor", "current", 0666),
2813 ATTR("apparmor", "prev", 0444),
2814 ATTR("apparmor", "exec", 0666),
2816 LSM_DIR_OPS(apparmor
);
2819 static const struct pid_entry attr_dir_stuff
[] = {
2820 ATTR(NULL
, "current", 0666),
2821 ATTR(NULL
, "prev", 0444),
2822 ATTR(NULL
, "exec", 0666),
2823 ATTR(NULL
, "fscreate", 0666),
2824 ATTR(NULL
, "keycreate", 0666),
2825 ATTR(NULL
, "sockcreate", 0666),
2826 #ifdef CONFIG_SECURITY_SMACK
2828 proc_smack_attr_dir_inode_ops
, proc_smack_attr_dir_ops
),
2830 #ifdef CONFIG_SECURITY_APPARMOR
2831 DIR("apparmor", 0555,
2832 proc_apparmor_attr_dir_inode_ops
, proc_apparmor_attr_dir_ops
),
2836 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2838 return proc_pident_readdir(file
, ctx
,
2839 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2842 static const struct file_operations proc_attr_dir_operations
= {
2843 .read
= generic_read_dir
,
2844 .iterate_shared
= proc_attr_dir_readdir
,
2845 .llseek
= generic_file_llseek
,
2848 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2849 struct dentry
*dentry
, unsigned int flags
)
2851 return proc_pident_lookup(dir
, dentry
,
2853 attr_dir_stuff
+ ARRAY_SIZE(attr_dir_stuff
));
2856 static const struct inode_operations proc_attr_dir_inode_operations
= {
2857 .lookup
= proc_attr_dir_lookup
,
2858 .getattr
= pid_getattr
,
2859 .setattr
= proc_setattr
,
2864 #ifdef CONFIG_ELF_CORE
2865 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2866 size_t count
, loff_t
*ppos
)
2868 struct task_struct
*task
= get_proc_task(file_inode(file
));
2869 struct mm_struct
*mm
;
2870 char buffer
[PROC_NUMBUF
];
2878 mm
= get_task_mm(task
);
2880 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2881 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2882 MMF_DUMP_FILTER_SHIFT
));
2884 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2887 put_task_struct(task
);
2892 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2893 const char __user
*buf
,
2897 struct task_struct
*task
;
2898 struct mm_struct
*mm
;
2904 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2909 task
= get_proc_task(file_inode(file
));
2913 mm
= get_task_mm(task
);
2918 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2920 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2922 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2927 put_task_struct(task
);
2934 static const struct file_operations proc_coredump_filter_operations
= {
2935 .read
= proc_coredump_filter_read
,
2936 .write
= proc_coredump_filter_write
,
2937 .llseek
= generic_file_llseek
,
2941 #ifdef CONFIG_TASK_IO_ACCOUNTING
2942 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2944 struct task_io_accounting acct
= task
->ioac
;
2945 unsigned long flags
;
2948 result
= down_read_killable(&task
->signal
->exec_update_lock
);
2952 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2957 if (whole
&& lock_task_sighand(task
, &flags
)) {
2958 struct task_struct
*t
= task
;
2960 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2961 while_each_thread(task
, t
)
2962 task_io_accounting_add(&acct
, &t
->ioac
);
2964 unlock_task_sighand(task
, &flags
);
2971 "read_bytes: %llu\n"
2972 "write_bytes: %llu\n"
2973 "cancelled_write_bytes: %llu\n",
2974 (unsigned long long)acct
.rchar
,
2975 (unsigned long long)acct
.wchar
,
2976 (unsigned long long)acct
.syscr
,
2977 (unsigned long long)acct
.syscw
,
2978 (unsigned long long)acct
.read_bytes
,
2979 (unsigned long long)acct
.write_bytes
,
2980 (unsigned long long)acct
.cancelled_write_bytes
);
2984 up_read(&task
->signal
->exec_update_lock
);
2988 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2989 struct pid
*pid
, struct task_struct
*task
)
2991 return do_io_accounting(task
, m
, 0);
2994 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2995 struct pid
*pid
, struct task_struct
*task
)
2997 return do_io_accounting(task
, m
, 1);
2999 #endif /* CONFIG_TASK_IO_ACCOUNTING */
3001 #ifdef CONFIG_USER_NS
3002 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
3003 const struct seq_operations
*seq_ops
)
3005 struct user_namespace
*ns
= NULL
;
3006 struct task_struct
*task
;
3007 struct seq_file
*seq
;
3010 task
= get_proc_task(inode
);
3013 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
3015 put_task_struct(task
);
3020 ret
= seq_open(file
, seq_ops
);
3024 seq
= file
->private_data
;
3034 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
3036 struct seq_file
*seq
= file
->private_data
;
3037 struct user_namespace
*ns
= seq
->private;
3039 return seq_release(inode
, file
);
3042 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
3044 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
3047 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
3049 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
3052 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
3054 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
3057 static const struct file_operations proc_uid_map_operations
= {
3058 .open
= proc_uid_map_open
,
3059 .write
= proc_uid_map_write
,
3061 .llseek
= seq_lseek
,
3062 .release
= proc_id_map_release
,
3065 static const struct file_operations proc_gid_map_operations
= {
3066 .open
= proc_gid_map_open
,
3067 .write
= proc_gid_map_write
,
3069 .llseek
= seq_lseek
,
3070 .release
= proc_id_map_release
,
3073 static const struct file_operations proc_projid_map_operations
= {
3074 .open
= proc_projid_map_open
,
3075 .write
= proc_projid_map_write
,
3077 .llseek
= seq_lseek
,
3078 .release
= proc_id_map_release
,
3081 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
3083 struct user_namespace
*ns
= NULL
;
3084 struct task_struct
*task
;
3088 task
= get_proc_task(inode
);
3091 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
3093 put_task_struct(task
);
3098 if (file
->f_mode
& FMODE_WRITE
) {
3100 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
3104 ret
= single_open(file
, &proc_setgroups_show
, ns
);
3115 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
3117 struct seq_file
*seq
= file
->private_data
;
3118 struct user_namespace
*ns
= seq
->private;
3119 int ret
= single_release(inode
, file
);
3124 static const struct file_operations proc_setgroups_operations
= {
3125 .open
= proc_setgroups_open
,
3126 .write
= proc_setgroups_write
,
3128 .llseek
= seq_lseek
,
3129 .release
= proc_setgroups_release
,
3131 #endif /* CONFIG_USER_NS */
3133 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
3134 struct pid
*pid
, struct task_struct
*task
)
3136 int err
= lock_trace(task
);
3138 seq_printf(m
, "%08x\n", task
->personality
);
3144 #ifdef CONFIG_LIVEPATCH
3145 static int proc_pid_patch_state(struct seq_file
*m
, struct pid_namespace
*ns
,
3146 struct pid
*pid
, struct task_struct
*task
)
3148 seq_printf(m
, "%d\n", task
->patch_state
);
3151 #endif /* CONFIG_LIVEPATCH */
3153 #ifdef CONFIG_STACKLEAK_METRICS
3154 static int proc_stack_depth(struct seq_file
*m
, struct pid_namespace
*ns
,
3155 struct pid
*pid
, struct task_struct
*task
)
3157 unsigned long prev_depth
= THREAD_SIZE
-
3158 (task
->prev_lowest_stack
& (THREAD_SIZE
- 1));
3159 unsigned long depth
= THREAD_SIZE
-
3160 (task
->lowest_stack
& (THREAD_SIZE
- 1));
3162 seq_printf(m
, "previous stack depth: %lu\nstack depth: %lu\n",
3166 #endif /* CONFIG_STACKLEAK_METRICS */
3171 static const struct file_operations proc_task_operations
;
3172 static const struct inode_operations proc_task_inode_operations
;
3174 static const struct pid_entry tgid_base_stuff
[] = {
3175 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
3176 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3177 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
3178 DIR("fdinfo", S_IRUGO
|S_IXUGO
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3179 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3181 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3183 REG("environ", S_IRUSR
, proc_environ_operations
),
3184 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3185 ONE("status", S_IRUGO
, proc_pid_status
),
3186 ONE("personality", S_IRUSR
, proc_pid_personality
),
3187 ONE("limits", S_IRUGO
, proc_pid_limits
),
3188 #ifdef CONFIG_SCHED_DEBUG
3189 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3191 #ifdef CONFIG_SCHED_AUTOGROUP
3192 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
3194 #ifdef CONFIG_TIME_NS
3195 REG("timens_offsets", S_IRUGO
|S_IWUSR
, proc_timens_offsets_operations
),
3197 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3198 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3199 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3201 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3202 ONE("stat", S_IRUGO
, proc_tgid_stat
),
3203 ONE("statm", S_IRUGO
, proc_pid_statm
),
3204 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3206 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3208 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3209 LNK("cwd", proc_cwd_link
),
3210 LNK("root", proc_root_link
),
3211 LNK("exe", proc_exe_link
),
3212 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3213 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3214 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
3215 #ifdef CONFIG_PROC_PAGE_MONITOR
3216 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3217 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3218 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3219 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3221 #ifdef CONFIG_SECURITY
3222 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3224 #ifdef CONFIG_KALLSYMS
3225 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3227 #ifdef CONFIG_STACKTRACE
3228 ONE("stack", S_IRUSR
, proc_pid_stack
),
3230 #ifdef CONFIG_SCHED_INFO
3231 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3233 #ifdef CONFIG_LATENCYTOP
3234 REG("latency", S_IRUGO
, proc_lstats_operations
),
3236 #ifdef CONFIG_PROC_PID_CPUSET
3237 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3239 #ifdef CONFIG_CGROUPS
3240 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3242 #ifdef CONFIG_PROC_CPU_RESCTRL
3243 ONE("cpu_resctrl_groups", S_IRUGO
, proc_resctrl_show
),
3245 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3246 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3247 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3249 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3250 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3252 #ifdef CONFIG_FAULT_INJECTION
3253 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3254 REG("fail-nth", 0644, proc_fail_nth_operations
),
3256 #ifdef CONFIG_ELF_CORE
3257 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3259 #ifdef CONFIG_TASK_IO_ACCOUNTING
3260 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
3262 #ifdef CONFIG_USER_NS
3263 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3264 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3265 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3266 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3268 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3269 REG("timers", S_IRUGO
, proc_timers_operations
),
3271 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
3272 #ifdef CONFIG_LIVEPATCH
3273 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3275 #ifdef CONFIG_STACKLEAK_METRICS
3276 ONE("stack_depth", S_IRUGO
, proc_stack_depth
),
3278 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3279 ONE("arch_status", S_IRUGO
, proc_pid_arch_status
),
3281 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3282 ONE("seccomp_cache", S_IRUSR
, proc_pid_seccomp_cache
),
3286 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3288 return proc_pident_readdir(file
, ctx
,
3289 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3292 static const struct file_operations proc_tgid_base_operations
= {
3293 .read
= generic_read_dir
,
3294 .iterate_shared
= proc_tgid_base_readdir
,
3295 .llseek
= generic_file_llseek
,
3298 struct pid
*tgid_pidfd_to_pid(const struct file
*file
)
3300 if (file
->f_op
!= &proc_tgid_base_operations
)
3301 return ERR_PTR(-EBADF
);
3303 return proc_pid(file_inode(file
));
3306 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3308 return proc_pident_lookup(dir
, dentry
,
3310 tgid_base_stuff
+ ARRAY_SIZE(tgid_base_stuff
));
3313 static const struct inode_operations proc_tgid_base_inode_operations
= {
3314 .lookup
= proc_tgid_base_lookup
,
3315 .getattr
= pid_getattr
,
3316 .setattr
= proc_setattr
,
3317 .permission
= proc_pid_permission
,
3321 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3322 * @pid: pid that should be flushed.
3324 * This function walks a list of inodes (that belong to any proc
3325 * filesystem) that are attached to the pid and flushes them from
3328 * It is safe and reasonable to cache /proc entries for a task until
3329 * that task exits. After that they just clog up the dcache with
3330 * useless entries, possibly causing useful dcache entries to be
3331 * flushed instead. This routine is provided to flush those useless
3332 * dcache entries when a process is reaped.
3334 * NOTE: This routine is just an optimization so it does not guarantee
3335 * that no dcache entries will exist after a process is reaped
3336 * it just makes it very unlikely that any will persist.
3339 void proc_flush_pid(struct pid
*pid
)
3341 proc_invalidate_siblings_dcache(&pid
->inodes
, &pid
->lock
);
3344 static struct dentry
*proc_pid_instantiate(struct dentry
* dentry
,
3345 struct task_struct
*task
, const void *ptr
)
3347 struct inode
*inode
;
3349 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3351 return ERR_PTR(-ENOENT
);
3353 inode
->i_op
= &proc_tgid_base_inode_operations
;
3354 inode
->i_fop
= &proc_tgid_base_operations
;
3355 inode
->i_flags
|=S_IMMUTABLE
;
3357 set_nlink(inode
, nlink_tgid
);
3358 pid_update_inode(task
, inode
);
3360 d_set_d_op(dentry
, &pid_dentry_operations
);
3361 return d_splice_alias(inode
, dentry
);
3364 struct dentry
*proc_pid_lookup(struct dentry
*dentry
, unsigned int flags
)
3366 struct task_struct
*task
;
3368 struct proc_fs_info
*fs_info
;
3369 struct pid_namespace
*ns
;
3370 struct dentry
*result
= ERR_PTR(-ENOENT
);
3372 tgid
= name_to_int(&dentry
->d_name
);
3376 fs_info
= proc_sb_info(dentry
->d_sb
);
3377 ns
= fs_info
->pid_ns
;
3379 task
= find_task_by_pid_ns(tgid
, ns
);
3381 get_task_struct(task
);
3386 /* Limit procfs to only ptraceable tasks */
3387 if (fs_info
->hide_pid
== HIDEPID_NOT_PTRACEABLE
) {
3388 if (!has_pid_permissions(fs_info
, task
, HIDEPID_NO_ACCESS
))
3392 result
= proc_pid_instantiate(dentry
, task
, NULL
);
3394 put_task_struct(task
);
3400 * Find the first task with tgid >= tgid
3405 struct task_struct
*task
;
3407 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3412 put_task_struct(iter
.task
);
3416 pid
= find_ge_pid(iter
.tgid
, ns
);
3418 iter
.tgid
= pid_nr_ns(pid
, ns
);
3419 iter
.task
= pid_task(pid
, PIDTYPE_TGID
);
3424 get_task_struct(iter
.task
);
3430 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3432 /* for the /proc/ directory itself, after non-process stuff has been done */
3433 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3435 struct tgid_iter iter
;
3436 struct proc_fs_info
*fs_info
= proc_sb_info(file_inode(file
)->i_sb
);
3437 struct pid_namespace
*ns
= proc_pid_ns(file_inode(file
)->i_sb
);
3438 loff_t pos
= ctx
->pos
;
3440 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3443 if (pos
== TGID_OFFSET
- 2) {
3444 struct inode
*inode
= d_inode(fs_info
->proc_self
);
3445 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3447 ctx
->pos
= pos
= pos
+ 1;
3449 if (pos
== TGID_OFFSET
- 1) {
3450 struct inode
*inode
= d_inode(fs_info
->proc_thread_self
);
3451 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3453 ctx
->pos
= pos
= pos
+ 1;
3455 iter
.tgid
= pos
- TGID_OFFSET
;
3457 for (iter
= next_tgid(ns
, iter
);
3459 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3464 if (!has_pid_permissions(fs_info
, iter
.task
, HIDEPID_INVISIBLE
))
3467 len
= snprintf(name
, sizeof(name
), "%u", iter
.tgid
);
3468 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3469 if (!proc_fill_cache(file
, ctx
, name
, len
,
3470 proc_pid_instantiate
, iter
.task
, NULL
)) {
3471 put_task_struct(iter
.task
);
3475 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3480 * proc_tid_comm_permission is a special permission function exclusively
3481 * used for the node /proc/<pid>/task/<tid>/comm.
3482 * It bypasses generic permission checks in the case where a task of the same
3483 * task group attempts to access the node.
3484 * The rationale behind this is that glibc and bionic access this node for
3485 * cross thread naming (pthread_set/getname_np(!self)). However, if
3486 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3487 * which locks out the cross thread naming implementation.
3488 * This function makes sure that the node is always accessible for members of
3489 * same thread group.
3491 static int proc_tid_comm_permission(struct user_namespace
*mnt_userns
,
3492 struct inode
*inode
, int mask
)
3494 bool is_same_tgroup
;
3495 struct task_struct
*task
;
3497 task
= get_proc_task(inode
);
3500 is_same_tgroup
= same_thread_group(current
, task
);
3501 put_task_struct(task
);
3503 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3504 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3505 * read or written by the members of the corresponding
3511 return generic_permission(&init_user_ns
, inode
, mask
);
3514 static const struct inode_operations proc_tid_comm_inode_operations
= {
3515 .permission
= proc_tid_comm_permission
,
3521 static const struct pid_entry tid_base_stuff
[] = {
3522 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3523 DIR("fdinfo", S_IRUGO
|S_IXUGO
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3524 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3526 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3528 REG("environ", S_IRUSR
, proc_environ_operations
),
3529 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3530 ONE("status", S_IRUGO
, proc_pid_status
),
3531 ONE("personality", S_IRUSR
, proc_pid_personality
),
3532 ONE("limits", S_IRUGO
, proc_pid_limits
),
3533 #ifdef CONFIG_SCHED_DEBUG
3534 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3536 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3537 &proc_tid_comm_inode_operations
,
3538 &proc_pid_set_comm_operations
, {}),
3539 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3540 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3542 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3543 ONE("stat", S_IRUGO
, proc_tid_stat
),
3544 ONE("statm", S_IRUGO
, proc_pid_statm
),
3545 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3546 #ifdef CONFIG_PROC_CHILDREN
3547 REG("children", S_IRUGO
, proc_tid_children_operations
),
3550 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3552 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3553 LNK("cwd", proc_cwd_link
),
3554 LNK("root", proc_root_link
),
3555 LNK("exe", proc_exe_link
),
3556 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3557 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3558 #ifdef CONFIG_PROC_PAGE_MONITOR
3559 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3560 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3561 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3562 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3564 #ifdef CONFIG_SECURITY
3565 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3567 #ifdef CONFIG_KALLSYMS
3568 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3570 #ifdef CONFIG_STACKTRACE
3571 ONE("stack", S_IRUSR
, proc_pid_stack
),
3573 #ifdef CONFIG_SCHED_INFO
3574 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3576 #ifdef CONFIG_LATENCYTOP
3577 REG("latency", S_IRUGO
, proc_lstats_operations
),
3579 #ifdef CONFIG_PROC_PID_CPUSET
3580 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3582 #ifdef CONFIG_CGROUPS
3583 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3585 #ifdef CONFIG_PROC_CPU_RESCTRL
3586 ONE("cpu_resctrl_groups", S_IRUGO
, proc_resctrl_show
),
3588 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3589 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3590 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3592 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3593 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3595 #ifdef CONFIG_FAULT_INJECTION
3596 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3597 REG("fail-nth", 0644, proc_fail_nth_operations
),
3599 #ifdef CONFIG_TASK_IO_ACCOUNTING
3600 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3602 #ifdef CONFIG_USER_NS
3603 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3604 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3605 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3606 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3608 #ifdef CONFIG_LIVEPATCH
3609 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3611 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3612 ONE("arch_status", S_IRUGO
, proc_pid_arch_status
),
3614 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3615 ONE("seccomp_cache", S_IRUSR
, proc_pid_seccomp_cache
),
3619 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3621 return proc_pident_readdir(file
, ctx
,
3622 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3625 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3627 return proc_pident_lookup(dir
, dentry
,
3629 tid_base_stuff
+ ARRAY_SIZE(tid_base_stuff
));
3632 static const struct file_operations proc_tid_base_operations
= {
3633 .read
= generic_read_dir
,
3634 .iterate_shared
= proc_tid_base_readdir
,
3635 .llseek
= generic_file_llseek
,
3638 static const struct inode_operations proc_tid_base_inode_operations
= {
3639 .lookup
= proc_tid_base_lookup
,
3640 .getattr
= pid_getattr
,
3641 .setattr
= proc_setattr
,
3644 static struct dentry
*proc_task_instantiate(struct dentry
*dentry
,
3645 struct task_struct
*task
, const void *ptr
)
3647 struct inode
*inode
;
3648 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3650 return ERR_PTR(-ENOENT
);
3652 inode
->i_op
= &proc_tid_base_inode_operations
;
3653 inode
->i_fop
= &proc_tid_base_operations
;
3654 inode
->i_flags
|= S_IMMUTABLE
;
3656 set_nlink(inode
, nlink_tid
);
3657 pid_update_inode(task
, inode
);
3659 d_set_d_op(dentry
, &pid_dentry_operations
);
3660 return d_splice_alias(inode
, dentry
);
3663 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3665 struct task_struct
*task
;
3666 struct task_struct
*leader
= get_proc_task(dir
);
3668 struct proc_fs_info
*fs_info
;
3669 struct pid_namespace
*ns
;
3670 struct dentry
*result
= ERR_PTR(-ENOENT
);
3675 tid
= name_to_int(&dentry
->d_name
);
3679 fs_info
= proc_sb_info(dentry
->d_sb
);
3680 ns
= fs_info
->pid_ns
;
3682 task
= find_task_by_pid_ns(tid
, ns
);
3684 get_task_struct(task
);
3688 if (!same_thread_group(leader
, task
))
3691 result
= proc_task_instantiate(dentry
, task
, NULL
);
3693 put_task_struct(task
);
3695 put_task_struct(leader
);
3701 * Find the first tid of a thread group to return to user space.
3703 * Usually this is just the thread group leader, but if the users
3704 * buffer was too small or there was a seek into the middle of the
3705 * directory we have more work todo.
3707 * In the case of a short read we start with find_task_by_pid.
3709 * In the case of a seek we start with the leader and walk nr
3712 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3713 struct pid_namespace
*ns
)
3715 struct task_struct
*pos
, *task
;
3716 unsigned long nr
= f_pos
;
3718 if (nr
!= f_pos
) /* 32bit overflow? */
3722 task
= pid_task(pid
, PIDTYPE_PID
);
3726 /* Attempt to start with the tid of a thread */
3728 pos
= find_task_by_pid_ns(tid
, ns
);
3729 if (pos
&& same_thread_group(pos
, task
))
3733 /* If nr exceeds the number of threads there is nothing todo */
3734 if (nr
>= get_nr_threads(task
))
3737 /* If we haven't found our starting place yet start
3738 * with the leader and walk nr threads forward.
3740 pos
= task
= task
->group_leader
;
3744 } while_each_thread(task
, pos
);
3749 get_task_struct(pos
);
3756 * Find the next thread in the thread list.
3757 * Return NULL if there is an error or no next thread.
3759 * The reference to the input task_struct is released.
3761 static struct task_struct
*next_tid(struct task_struct
*start
)
3763 struct task_struct
*pos
= NULL
;
3765 if (pid_alive(start
)) {
3766 pos
= next_thread(start
);
3767 if (thread_group_leader(pos
))
3770 get_task_struct(pos
);
3773 put_task_struct(start
);
3777 /* for the /proc/TGID/task/ directories */
3778 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3780 struct inode
*inode
= file_inode(file
);
3781 struct task_struct
*task
;
3782 struct pid_namespace
*ns
;
3785 if (proc_inode_is_dead(inode
))
3788 if (!dir_emit_dots(file
, ctx
))
3791 /* f_version caches the tgid value that the last readdir call couldn't
3792 * return. lseek aka telldir automagically resets f_version to 0.
3794 ns
= proc_pid_ns(inode
->i_sb
);
3795 tid
= (int)file
->f_version
;
3796 file
->f_version
= 0;
3797 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3799 task
= next_tid(task
), ctx
->pos
++) {
3802 tid
= task_pid_nr_ns(task
, ns
);
3803 len
= snprintf(name
, sizeof(name
), "%u", tid
);
3804 if (!proc_fill_cache(file
, ctx
, name
, len
,
3805 proc_task_instantiate
, task
, NULL
)) {
3806 /* returning this tgid failed, save it as the first
3807 * pid for the next readir call */
3808 file
->f_version
= (u64
)tid
;
3809 put_task_struct(task
);
3817 static int proc_task_getattr(struct user_namespace
*mnt_userns
,
3818 const struct path
*path
, struct kstat
*stat
,
3819 u32 request_mask
, unsigned int query_flags
)
3821 struct inode
*inode
= d_inode(path
->dentry
);
3822 struct task_struct
*p
= get_proc_task(inode
);
3823 generic_fillattr(&init_user_ns
, inode
, stat
);
3826 stat
->nlink
+= get_nr_threads(p
);
3833 static const struct inode_operations proc_task_inode_operations
= {
3834 .lookup
= proc_task_lookup
,
3835 .getattr
= proc_task_getattr
,
3836 .setattr
= proc_setattr
,
3837 .permission
= proc_pid_permission
,
3840 static const struct file_operations proc_task_operations
= {
3841 .read
= generic_read_dir
,
3842 .iterate_shared
= proc_task_readdir
,
3843 .llseek
= generic_file_llseek
,
3846 void __init
set_proc_pid_nlink(void)
3848 nlink_tid
= pid_entry_nlink(tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3849 nlink_tgid
= pid_entry_nlink(tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));