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;
556 badness
= oom_badness(task
, totalpages
);
558 * Special case OOM_SCORE_ADJ_MIN for all others scale the
559 * badness value into [0, 2000] range which we have been
560 * exporting for a long time so userspace might depend on it.
562 if (badness
!= LONG_MIN
)
563 points
= (1000 + badness
* 1000 / (long)totalpages
) * 2 / 3;
565 seq_printf(m
, "%lu\n", points
);
575 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
576 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
577 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
578 [RLIMIT_DATA
] = {"Max data size", "bytes"},
579 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
580 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
581 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
582 [RLIMIT_NPROC
] = {"Max processes", "processes"},
583 [RLIMIT_NOFILE
] = {"Max open files", "files"},
584 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
585 [RLIMIT_AS
] = {"Max address space", "bytes"},
586 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
587 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
588 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
589 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
590 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
591 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
594 /* Display limits for a process */
595 static int proc_pid_limits(struct seq_file
*m
, struct pid_namespace
*ns
,
596 struct pid
*pid
, struct task_struct
*task
)
601 struct rlimit rlim
[RLIM_NLIMITS
];
603 if (!lock_task_sighand(task
, &flags
))
605 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
606 unlock_task_sighand(task
, &flags
);
609 * print the file header
616 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
617 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
618 seq_printf(m
, "%-25s %-20s ",
619 lnames
[i
].name
, "unlimited");
621 seq_printf(m
, "%-25s %-20lu ",
622 lnames
[i
].name
, rlim
[i
].rlim_cur
);
624 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
625 seq_printf(m
, "%-20s ", "unlimited");
627 seq_printf(m
, "%-20lu ", rlim
[i
].rlim_max
);
630 seq_printf(m
, "%-10s\n", lnames
[i
].unit
);
638 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
639 static int proc_pid_syscall(struct seq_file
*m
, struct pid_namespace
*ns
,
640 struct pid
*pid
, struct task_struct
*task
)
642 struct syscall_info info
;
643 u64
*args
= &info
.data
.args
[0];
646 res
= lock_trace(task
);
650 if (task_current_syscall(task
, &info
))
651 seq_puts(m
, "running\n");
652 else if (info
.data
.nr
< 0)
653 seq_printf(m
, "%d 0x%llx 0x%llx\n",
654 info
.data
.nr
, info
.sp
, info
.data
.instruction_pointer
);
657 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
659 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
660 info
.sp
, info
.data
.instruction_pointer
);
665 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
667 /************************************************************************/
668 /* Here the fs part begins */
669 /************************************************************************/
671 /* permission checks */
672 static int proc_fd_access_allowed(struct inode
*inode
)
674 struct task_struct
*task
;
676 /* Allow access to a task's file descriptors if it is us or we
677 * may use ptrace attach to the process and find out that
680 task
= get_proc_task(inode
);
682 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
683 put_task_struct(task
);
688 int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
691 struct inode
*inode
= d_inode(dentry
);
693 if (attr
->ia_valid
& ATTR_MODE
)
696 error
= setattr_prepare(dentry
, attr
);
700 setattr_copy(inode
, attr
);
701 mark_inode_dirty(inode
);
706 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
707 * or euid/egid (for hide_pid_min=2)?
709 static bool has_pid_permissions(struct proc_fs_info
*fs_info
,
710 struct task_struct
*task
,
711 enum proc_hidepid hide_pid_min
)
714 * If 'hidpid' mount option is set force a ptrace check,
715 * we indicate that we are using a filesystem syscall
716 * by passing PTRACE_MODE_READ_FSCREDS
718 if (fs_info
->hide_pid
== HIDEPID_NOT_PTRACEABLE
)
719 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
721 if (fs_info
->hide_pid
< hide_pid_min
)
723 if (in_group_p(fs_info
->pid_gid
))
725 return ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
);
729 static int proc_pid_permission(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(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 int this_len
= min_t(int, 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 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_adj
);
1053 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1056 static int __set_oom_adj(struct file
*file
, int oom_adj
, bool legacy
)
1058 static DEFINE_MUTEX(oom_adj_mutex
);
1059 struct mm_struct
*mm
= NULL
;
1060 struct task_struct
*task
;
1063 task
= get_proc_task(file_inode(file
));
1067 mutex_lock(&oom_adj_mutex
);
1069 if (oom_adj
< task
->signal
->oom_score_adj
&&
1070 !capable(CAP_SYS_RESOURCE
)) {
1075 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1076 * /proc/pid/oom_score_adj instead.
1078 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1079 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
1082 if ((short)oom_adj
< task
->signal
->oom_score_adj_min
&&
1083 !capable(CAP_SYS_RESOURCE
)) {
1090 * Make sure we will check other processes sharing the mm if this is
1091 * not vfrok which wants its own oom_score_adj.
1092 * pin the mm so it doesn't go away and get reused after task_unlock
1094 if (!task
->vfork_done
) {
1095 struct task_struct
*p
= find_lock_task_mm(task
);
1098 if (atomic_read(&p
->mm
->mm_users
) > 1) {
1106 task
->signal
->oom_score_adj
= oom_adj
;
1107 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1108 task
->signal
->oom_score_adj_min
= (short)oom_adj
;
1109 trace_oom_score_adj_update(task
);
1112 struct task_struct
*p
;
1115 for_each_process(p
) {
1116 if (same_thread_group(task
, p
))
1119 /* do not touch kernel threads or the global init */
1120 if (p
->flags
& PF_KTHREAD
|| is_global_init(p
))
1124 if (!p
->vfork_done
&& process_shares_mm(p
, mm
)) {
1125 p
->signal
->oom_score_adj
= oom_adj
;
1126 if (!legacy
&& has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1127 p
->signal
->oom_score_adj_min
= (short)oom_adj
;
1135 mutex_unlock(&oom_adj_mutex
);
1136 put_task_struct(task
);
1141 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1142 * kernels. The effective policy is defined by oom_score_adj, which has a
1143 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1144 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1145 * Processes that become oom disabled via oom_adj will still be oom disabled
1146 * with this implementation.
1148 * oom_adj cannot be removed since existing userspace binaries use it.
1150 static ssize_t
oom_adj_write(struct file
*file
, const char __user
*buf
,
1151 size_t count
, loff_t
*ppos
)
1153 char buffer
[PROC_NUMBUF
];
1157 memset(buffer
, 0, sizeof(buffer
));
1158 if (count
> sizeof(buffer
) - 1)
1159 count
= sizeof(buffer
) - 1;
1160 if (copy_from_user(buffer
, buf
, count
)) {
1165 err
= kstrtoint(strstrip(buffer
), 0, &oom_adj
);
1168 if ((oom_adj
< OOM_ADJUST_MIN
|| oom_adj
> OOM_ADJUST_MAX
) &&
1169 oom_adj
!= OOM_DISABLE
) {
1175 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1176 * value is always attainable.
1178 if (oom_adj
== OOM_ADJUST_MAX
)
1179 oom_adj
= OOM_SCORE_ADJ_MAX
;
1181 oom_adj
= (oom_adj
* OOM_SCORE_ADJ_MAX
) / -OOM_DISABLE
;
1183 err
= __set_oom_adj(file
, oom_adj
, true);
1185 return err
< 0 ? err
: count
;
1188 static const struct file_operations proc_oom_adj_operations
= {
1189 .read
= oom_adj_read
,
1190 .write
= oom_adj_write
,
1191 .llseek
= generic_file_llseek
,
1194 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1195 size_t count
, loff_t
*ppos
)
1197 struct task_struct
*task
= get_proc_task(file_inode(file
));
1198 char buffer
[PROC_NUMBUF
];
1199 short oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1204 oom_score_adj
= task
->signal
->oom_score_adj
;
1205 put_task_struct(task
);
1206 len
= snprintf(buffer
, sizeof(buffer
), "%hd\n", oom_score_adj
);
1207 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1210 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1211 size_t count
, loff_t
*ppos
)
1213 char buffer
[PROC_NUMBUF
];
1217 memset(buffer
, 0, sizeof(buffer
));
1218 if (count
> sizeof(buffer
) - 1)
1219 count
= sizeof(buffer
) - 1;
1220 if (copy_from_user(buffer
, buf
, count
)) {
1225 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1228 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1229 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1234 err
= __set_oom_adj(file
, oom_score_adj
, false);
1236 return err
< 0 ? err
: count
;
1239 static const struct file_operations proc_oom_score_adj_operations
= {
1240 .read
= oom_score_adj_read
,
1241 .write
= oom_score_adj_write
,
1242 .llseek
= default_llseek
,
1246 #define TMPBUFLEN 11
1247 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1248 size_t count
, loff_t
*ppos
)
1250 struct inode
* inode
= file_inode(file
);
1251 struct task_struct
*task
= get_proc_task(inode
);
1253 char tmpbuf
[TMPBUFLEN
];
1257 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1258 from_kuid(file
->f_cred
->user_ns
,
1259 audit_get_loginuid(task
)));
1260 put_task_struct(task
);
1261 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1264 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1265 size_t count
, loff_t
*ppos
)
1267 struct inode
* inode
= file_inode(file
);
1273 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1280 /* No partial writes. */
1284 rv
= kstrtou32_from_user(buf
, count
, 10, &loginuid
);
1288 /* is userspace tring to explicitly UNSET the loginuid? */
1289 if (loginuid
== AUDIT_UID_UNSET
) {
1290 kloginuid
= INVALID_UID
;
1292 kloginuid
= make_kuid(file
->f_cred
->user_ns
, loginuid
);
1293 if (!uid_valid(kloginuid
))
1297 rv
= audit_set_loginuid(kloginuid
);
1303 static const struct file_operations proc_loginuid_operations
= {
1304 .read
= proc_loginuid_read
,
1305 .write
= proc_loginuid_write
,
1306 .llseek
= generic_file_llseek
,
1309 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1310 size_t count
, loff_t
*ppos
)
1312 struct inode
* inode
= file_inode(file
);
1313 struct task_struct
*task
= get_proc_task(inode
);
1315 char tmpbuf
[TMPBUFLEN
];
1319 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1320 audit_get_sessionid(task
));
1321 put_task_struct(task
);
1322 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1325 static const struct file_operations proc_sessionid_operations
= {
1326 .read
= proc_sessionid_read
,
1327 .llseek
= generic_file_llseek
,
1331 #ifdef CONFIG_FAULT_INJECTION
1332 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1333 size_t count
, loff_t
*ppos
)
1335 struct task_struct
*task
= get_proc_task(file_inode(file
));
1336 char buffer
[PROC_NUMBUF
];
1342 make_it_fail
= task
->make_it_fail
;
1343 put_task_struct(task
);
1345 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1347 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1350 static ssize_t
proc_fault_inject_write(struct file
* file
,
1351 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1353 struct task_struct
*task
;
1354 char buffer
[PROC_NUMBUF
];
1358 if (!capable(CAP_SYS_RESOURCE
))
1360 memset(buffer
, 0, sizeof(buffer
));
1361 if (count
> sizeof(buffer
) - 1)
1362 count
= sizeof(buffer
) - 1;
1363 if (copy_from_user(buffer
, buf
, count
))
1365 rv
= kstrtoint(strstrip(buffer
), 0, &make_it_fail
);
1368 if (make_it_fail
< 0 || make_it_fail
> 1)
1371 task
= get_proc_task(file_inode(file
));
1374 task
->make_it_fail
= make_it_fail
;
1375 put_task_struct(task
);
1380 static const struct file_operations proc_fault_inject_operations
= {
1381 .read
= proc_fault_inject_read
,
1382 .write
= proc_fault_inject_write
,
1383 .llseek
= generic_file_llseek
,
1386 static ssize_t
proc_fail_nth_write(struct file
*file
, const char __user
*buf
,
1387 size_t count
, loff_t
*ppos
)
1389 struct task_struct
*task
;
1393 err
= kstrtouint_from_user(buf
, count
, 0, &n
);
1397 task
= get_proc_task(file_inode(file
));
1401 put_task_struct(task
);
1406 static ssize_t
proc_fail_nth_read(struct file
*file
, char __user
*buf
,
1407 size_t count
, loff_t
*ppos
)
1409 struct task_struct
*task
;
1410 char numbuf
[PROC_NUMBUF
];
1413 task
= get_proc_task(file_inode(file
));
1416 len
= snprintf(numbuf
, sizeof(numbuf
), "%u\n", task
->fail_nth
);
1417 put_task_struct(task
);
1418 return simple_read_from_buffer(buf
, count
, ppos
, numbuf
, len
);
1421 static const struct file_operations proc_fail_nth_operations
= {
1422 .read
= proc_fail_nth_read
,
1423 .write
= proc_fail_nth_write
,
1428 #ifdef CONFIG_SCHED_DEBUG
1430 * Print out various scheduling related per-task fields:
1432 static int sched_show(struct seq_file
*m
, void *v
)
1434 struct inode
*inode
= m
->private;
1435 struct pid_namespace
*ns
= proc_pid_ns(inode
->i_sb
);
1436 struct task_struct
*p
;
1438 p
= get_proc_task(inode
);
1441 proc_sched_show_task(p
, ns
, m
);
1449 sched_write(struct file
*file
, const char __user
*buf
,
1450 size_t count
, loff_t
*offset
)
1452 struct inode
*inode
= file_inode(file
);
1453 struct task_struct
*p
;
1455 p
= get_proc_task(inode
);
1458 proc_sched_set_task(p
);
1465 static int sched_open(struct inode
*inode
, struct file
*filp
)
1467 return single_open(filp
, sched_show
, inode
);
1470 static const struct file_operations proc_pid_sched_operations
= {
1473 .write
= sched_write
,
1474 .llseek
= seq_lseek
,
1475 .release
= single_release
,
1480 #ifdef CONFIG_SCHED_AUTOGROUP
1482 * Print out autogroup related information:
1484 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1486 struct inode
*inode
= m
->private;
1487 struct task_struct
*p
;
1489 p
= get_proc_task(inode
);
1492 proc_sched_autogroup_show_task(p
, m
);
1500 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1501 size_t count
, loff_t
*offset
)
1503 struct inode
*inode
= file_inode(file
);
1504 struct task_struct
*p
;
1505 char buffer
[PROC_NUMBUF
];
1509 memset(buffer
, 0, sizeof(buffer
));
1510 if (count
> sizeof(buffer
) - 1)
1511 count
= sizeof(buffer
) - 1;
1512 if (copy_from_user(buffer
, buf
, count
))
1515 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1519 p
= get_proc_task(inode
);
1523 err
= proc_sched_autogroup_set_nice(p
, nice
);
1532 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1536 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1538 struct seq_file
*m
= filp
->private_data
;
1545 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1546 .open
= sched_autogroup_open
,
1548 .write
= sched_autogroup_write
,
1549 .llseek
= seq_lseek
,
1550 .release
= single_release
,
1553 #endif /* CONFIG_SCHED_AUTOGROUP */
1555 #ifdef CONFIG_TIME_NS
1556 static int timens_offsets_show(struct seq_file
*m
, void *v
)
1558 struct task_struct
*p
;
1560 p
= get_proc_task(file_inode(m
->file
));
1563 proc_timens_show_offsets(p
, m
);
1570 static ssize_t
timens_offsets_write(struct file
*file
, const char __user
*buf
,
1571 size_t count
, loff_t
*ppos
)
1573 struct inode
*inode
= file_inode(file
);
1574 struct proc_timens_offset offsets
[2];
1575 char *kbuf
= NULL
, *pos
, *next_line
;
1576 struct task_struct
*p
;
1579 /* Only allow < page size writes at the beginning of the file */
1580 if ((*ppos
!= 0) || (count
>= PAGE_SIZE
))
1583 /* Slurp in the user data */
1584 kbuf
= memdup_user_nul(buf
, count
);
1586 return PTR_ERR(kbuf
);
1588 /* Parse the user data */
1591 for (pos
= kbuf
; pos
; pos
= next_line
) {
1592 struct proc_timens_offset
*off
= &offsets
[noffsets
];
1596 /* Find the end of line and ensure we don't look past it */
1597 next_line
= strchr(pos
, '\n');
1601 if (*next_line
== '\0')
1605 err
= sscanf(pos
, "%9s %lld %lu", clock
,
1606 &off
->val
.tv_sec
, &off
->val
.tv_nsec
);
1607 if (err
!= 3 || off
->val
.tv_nsec
>= NSEC_PER_SEC
)
1610 clock
[sizeof(clock
) - 1] = 0;
1611 if (strcmp(clock
, "monotonic") == 0 ||
1612 strcmp(clock
, __stringify(CLOCK_MONOTONIC
)) == 0)
1613 off
->clockid
= CLOCK_MONOTONIC
;
1614 else if (strcmp(clock
, "boottime") == 0 ||
1615 strcmp(clock
, __stringify(CLOCK_BOOTTIME
)) == 0)
1616 off
->clockid
= CLOCK_BOOTTIME
;
1621 if (noffsets
== ARRAY_SIZE(offsets
)) {
1623 count
= next_line
- kbuf
;
1629 p
= get_proc_task(inode
);
1632 ret
= proc_timens_set_offset(file
, p
, offsets
, noffsets
);
1643 static int timens_offsets_open(struct inode
*inode
, struct file
*filp
)
1645 return single_open(filp
, timens_offsets_show
, inode
);
1648 static const struct file_operations proc_timens_offsets_operations
= {
1649 .open
= timens_offsets_open
,
1651 .write
= timens_offsets_write
,
1652 .llseek
= seq_lseek
,
1653 .release
= single_release
,
1655 #endif /* CONFIG_TIME_NS */
1657 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1658 size_t count
, loff_t
*offset
)
1660 struct inode
*inode
= file_inode(file
);
1661 struct task_struct
*p
;
1662 char buffer
[TASK_COMM_LEN
];
1663 const size_t maxlen
= sizeof(buffer
) - 1;
1665 memset(buffer
, 0, sizeof(buffer
));
1666 if (copy_from_user(buffer
, buf
, count
> maxlen
? maxlen
: count
))
1669 p
= get_proc_task(inode
);
1673 if (same_thread_group(current
, p
))
1674 set_task_comm(p
, buffer
);
1683 static int comm_show(struct seq_file
*m
, void *v
)
1685 struct inode
*inode
= m
->private;
1686 struct task_struct
*p
;
1688 p
= get_proc_task(inode
);
1692 proc_task_name(m
, p
, false);
1700 static int comm_open(struct inode
*inode
, struct file
*filp
)
1702 return single_open(filp
, comm_show
, inode
);
1705 static const struct file_operations proc_pid_set_comm_operations
= {
1708 .write
= comm_write
,
1709 .llseek
= seq_lseek
,
1710 .release
= single_release
,
1713 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1715 struct task_struct
*task
;
1716 struct file
*exe_file
;
1718 task
= get_proc_task(d_inode(dentry
));
1721 exe_file
= get_task_exe_file(task
);
1722 put_task_struct(task
);
1724 *exe_path
= exe_file
->f_path
;
1725 path_get(&exe_file
->f_path
);
1732 static const char *proc_pid_get_link(struct dentry
*dentry
,
1733 struct inode
*inode
,
1734 struct delayed_call
*done
)
1737 int error
= -EACCES
;
1740 return ERR_PTR(-ECHILD
);
1742 /* Are we allowed to snoop on the tasks file descriptors? */
1743 if (!proc_fd_access_allowed(inode
))
1746 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1750 error
= nd_jump_link(&path
);
1752 return ERR_PTR(error
);
1755 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1757 char *tmp
= (char *)__get_free_page(GFP_KERNEL
);
1764 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1765 len
= PTR_ERR(pathname
);
1766 if (IS_ERR(pathname
))
1768 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1772 if (copy_to_user(buffer
, pathname
, len
))
1775 free_page((unsigned long)tmp
);
1779 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1781 int error
= -EACCES
;
1782 struct inode
*inode
= d_inode(dentry
);
1785 /* Are we allowed to snoop on the tasks file descriptors? */
1786 if (!proc_fd_access_allowed(inode
))
1789 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1793 error
= do_proc_readlink(&path
, buffer
, buflen
);
1799 const struct inode_operations proc_pid_link_inode_operations
= {
1800 .readlink
= proc_pid_readlink
,
1801 .get_link
= proc_pid_get_link
,
1802 .setattr
= proc_setattr
,
1806 /* building an inode */
1808 void task_dump_owner(struct task_struct
*task
, umode_t mode
,
1809 kuid_t
*ruid
, kgid_t
*rgid
)
1811 /* Depending on the state of dumpable compute who should own a
1812 * proc file for a task.
1814 const struct cred
*cred
;
1818 if (unlikely(task
->flags
& PF_KTHREAD
)) {
1819 *ruid
= GLOBAL_ROOT_UID
;
1820 *rgid
= GLOBAL_ROOT_GID
;
1824 /* Default to the tasks effective ownership */
1826 cred
= __task_cred(task
);
1832 * Before the /proc/pid/status file was created the only way to read
1833 * the effective uid of a /process was to stat /proc/pid. Reading
1834 * /proc/pid/status is slow enough that procps and other packages
1835 * kept stating /proc/pid. To keep the rules in /proc simple I have
1836 * made this apply to all per process world readable and executable
1839 if (mode
!= (S_IFDIR
|S_IRUGO
|S_IXUGO
)) {
1840 struct mm_struct
*mm
;
1843 /* Make non-dumpable tasks owned by some root */
1845 if (get_dumpable(mm
) != SUID_DUMP_USER
) {
1846 struct user_namespace
*user_ns
= mm
->user_ns
;
1848 uid
= make_kuid(user_ns
, 0);
1849 if (!uid_valid(uid
))
1850 uid
= GLOBAL_ROOT_UID
;
1852 gid
= make_kgid(user_ns
, 0);
1853 if (!gid_valid(gid
))
1854 gid
= GLOBAL_ROOT_GID
;
1857 uid
= GLOBAL_ROOT_UID
;
1858 gid
= GLOBAL_ROOT_GID
;
1866 void proc_pid_evict_inode(struct proc_inode
*ei
)
1868 struct pid
*pid
= ei
->pid
;
1870 if (S_ISDIR(ei
->vfs_inode
.i_mode
)) {
1871 spin_lock(&pid
->lock
);
1872 hlist_del_init_rcu(&ei
->sibling_inodes
);
1873 spin_unlock(&pid
->lock
);
1879 struct inode
*proc_pid_make_inode(struct super_block
* sb
,
1880 struct task_struct
*task
, umode_t mode
)
1882 struct inode
* inode
;
1883 struct proc_inode
*ei
;
1886 /* We need a new inode */
1888 inode
= new_inode(sb
);
1894 inode
->i_mode
= mode
;
1895 inode
->i_ino
= get_next_ino();
1896 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= current_time(inode
);
1897 inode
->i_op
= &proc_def_inode_operations
;
1900 * grab the reference to task.
1902 pid
= get_task_pid(task
, PIDTYPE_PID
);
1906 /* Let the pid remember us for quick removal */
1908 if (S_ISDIR(mode
)) {
1909 spin_lock(&pid
->lock
);
1910 hlist_add_head_rcu(&ei
->sibling_inodes
, &pid
->inodes
);
1911 spin_unlock(&pid
->lock
);
1914 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
1915 security_task_to_inode(task
, inode
);
1925 int pid_getattr(const struct path
*path
, struct kstat
*stat
,
1926 u32 request_mask
, unsigned int query_flags
)
1928 struct inode
*inode
= d_inode(path
->dentry
);
1929 struct proc_fs_info
*fs_info
= proc_sb_info(inode
->i_sb
);
1930 struct task_struct
*task
;
1932 generic_fillattr(inode
, stat
);
1934 stat
->uid
= GLOBAL_ROOT_UID
;
1935 stat
->gid
= GLOBAL_ROOT_GID
;
1937 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1939 if (!has_pid_permissions(fs_info
, task
, HIDEPID_INVISIBLE
)) {
1942 * This doesn't prevent learning whether PID exists,
1943 * it only makes getattr() consistent with readdir().
1947 task_dump_owner(task
, inode
->i_mode
, &stat
->uid
, &stat
->gid
);
1956 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1958 void pid_update_inode(struct task_struct
*task
, struct inode
*inode
)
1960 task_dump_owner(task
, inode
->i_mode
, &inode
->i_uid
, &inode
->i_gid
);
1962 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1963 security_task_to_inode(task
, inode
);
1967 * Rewrite the inode's ownerships here because the owning task may have
1968 * performed a setuid(), etc.
1971 static int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1973 struct inode
*inode
;
1974 struct task_struct
*task
;
1976 if (flags
& LOOKUP_RCU
)
1979 inode
= d_inode(dentry
);
1980 task
= get_proc_task(inode
);
1983 pid_update_inode(task
, inode
);
1984 put_task_struct(task
);
1990 static inline bool proc_inode_is_dead(struct inode
*inode
)
1992 return !proc_pid(inode
)->tasks
[PIDTYPE_PID
].first
;
1995 int pid_delete_dentry(const struct dentry
*dentry
)
1997 /* Is the task we represent dead?
1998 * If so, then don't put the dentry on the lru list,
1999 * kill it immediately.
2001 return proc_inode_is_dead(d_inode(dentry
));
2004 const struct dentry_operations pid_dentry_operations
=
2006 .d_revalidate
= pid_revalidate
,
2007 .d_delete
= pid_delete_dentry
,
2013 * Fill a directory entry.
2015 * If possible create the dcache entry and derive our inode number and
2016 * file type from dcache entry.
2018 * Since all of the proc inode numbers are dynamically generated, the inode
2019 * numbers do not exist until the inode is cache. This means creating the
2020 * the dcache entry in readdir is necessary to keep the inode numbers
2021 * reported by readdir in sync with the inode numbers reported
2024 bool proc_fill_cache(struct file
*file
, struct dir_context
*ctx
,
2025 const char *name
, unsigned int len
,
2026 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
2028 struct dentry
*child
, *dir
= file
->f_path
.dentry
;
2029 struct qstr qname
= QSTR_INIT(name
, len
);
2030 struct inode
*inode
;
2031 unsigned type
= DT_UNKNOWN
;
2034 child
= d_hash_and_lookup(dir
, &qname
);
2036 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq
);
2037 child
= d_alloc_parallel(dir
, &qname
, &wq
);
2039 goto end_instantiate
;
2040 if (d_in_lookup(child
)) {
2042 res
= instantiate(child
, task
, ptr
);
2043 d_lookup_done(child
);
2044 if (unlikely(res
)) {
2048 goto end_instantiate
;
2052 inode
= d_inode(child
);
2054 type
= inode
->i_mode
>> 12;
2057 return dir_emit(ctx
, name
, len
, ino
, type
);
2061 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2062 * which represent vma start and end addresses.
2064 static int dname_to_vma_addr(struct dentry
*dentry
,
2065 unsigned long *start
, unsigned long *end
)
2067 const char *str
= dentry
->d_name
.name
;
2068 unsigned long long sval
, eval
;
2071 if (str
[0] == '0' && str
[1] != '-')
2073 len
= _parse_integer(str
, 16, &sval
);
2074 if (len
& KSTRTOX_OVERFLOW
)
2076 if (sval
!= (unsigned long)sval
)
2084 if (str
[0] == '0' && str
[1])
2086 len
= _parse_integer(str
, 16, &eval
);
2087 if (len
& KSTRTOX_OVERFLOW
)
2089 if (eval
!= (unsigned long)eval
)
2102 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
2104 unsigned long vm_start
, vm_end
;
2105 bool exact_vma_exists
= false;
2106 struct mm_struct
*mm
= NULL
;
2107 struct task_struct
*task
;
2108 struct inode
*inode
;
2111 if (flags
& LOOKUP_RCU
)
2114 inode
= d_inode(dentry
);
2115 task
= get_proc_task(inode
);
2119 mm
= mm_access(task
, PTRACE_MODE_READ_FSCREDS
);
2120 if (IS_ERR_OR_NULL(mm
))
2123 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
2124 status
= mmap_read_lock_killable(mm
);
2126 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
,
2128 mmap_read_unlock(mm
);
2134 if (exact_vma_exists
) {
2135 task_dump_owner(task
, 0, &inode
->i_uid
, &inode
->i_gid
);
2137 security_task_to_inode(task
, inode
);
2142 put_task_struct(task
);
2148 static const struct dentry_operations tid_map_files_dentry_operations
= {
2149 .d_revalidate
= map_files_d_revalidate
,
2150 .d_delete
= pid_delete_dentry
,
2153 static int map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2155 unsigned long vm_start
, vm_end
;
2156 struct vm_area_struct
*vma
;
2157 struct task_struct
*task
;
2158 struct mm_struct
*mm
;
2162 task
= get_proc_task(d_inode(dentry
));
2166 mm
= get_task_mm(task
);
2167 put_task_struct(task
);
2171 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2175 rc
= mmap_read_lock_killable(mm
);
2180 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2181 if (vma
&& vma
->vm_file
) {
2182 *path
= vma
->vm_file
->f_path
;
2186 mmap_read_unlock(mm
);
2194 struct map_files_info
{
2195 unsigned long start
;
2201 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2202 * to concerns about how the symlinks may be used to bypass permissions on
2203 * ancestor directories in the path to the file in question.
2206 proc_map_files_get_link(struct dentry
*dentry
,
2207 struct inode
*inode
,
2208 struct delayed_call
*done
)
2210 if (!checkpoint_restore_ns_capable(&init_user_ns
))
2211 return ERR_PTR(-EPERM
);
2213 return proc_pid_get_link(dentry
, inode
, done
);
2217 * Identical to proc_pid_link_inode_operations except for get_link()
2219 static const struct inode_operations proc_map_files_link_inode_operations
= {
2220 .readlink
= proc_pid_readlink
,
2221 .get_link
= proc_map_files_get_link
,
2222 .setattr
= proc_setattr
,
2225 static struct dentry
*
2226 proc_map_files_instantiate(struct dentry
*dentry
,
2227 struct task_struct
*task
, const void *ptr
)
2229 fmode_t mode
= (fmode_t
)(unsigned long)ptr
;
2230 struct proc_inode
*ei
;
2231 struct inode
*inode
;
2233 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFLNK
|
2234 ((mode
& FMODE_READ
) ? S_IRUSR
: 0) |
2235 ((mode
& FMODE_WRITE
) ? S_IWUSR
: 0));
2237 return ERR_PTR(-ENOENT
);
2240 ei
->op
.proc_get_link
= map_files_get_link
;
2242 inode
->i_op
= &proc_map_files_link_inode_operations
;
2245 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2246 return d_splice_alias(inode
, dentry
);
2249 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2250 struct dentry
*dentry
, unsigned int flags
)
2252 unsigned long vm_start
, vm_end
;
2253 struct vm_area_struct
*vma
;
2254 struct task_struct
*task
;
2255 struct dentry
*result
;
2256 struct mm_struct
*mm
;
2258 result
= ERR_PTR(-ENOENT
);
2259 task
= get_proc_task(dir
);
2263 result
= ERR_PTR(-EACCES
);
2264 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2267 result
= ERR_PTR(-ENOENT
);
2268 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2271 mm
= get_task_mm(task
);
2275 result
= ERR_PTR(-EINTR
);
2276 if (mmap_read_lock_killable(mm
))
2279 result
= ERR_PTR(-ENOENT
);
2280 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2285 result
= proc_map_files_instantiate(dentry
, task
,
2286 (void *)(unsigned long)vma
->vm_file
->f_mode
);
2289 mmap_read_unlock(mm
);
2293 put_task_struct(task
);
2298 static const struct inode_operations proc_map_files_inode_operations
= {
2299 .lookup
= proc_map_files_lookup
,
2300 .permission
= proc_fd_permission
,
2301 .setattr
= proc_setattr
,
2305 proc_map_files_readdir(struct file
*file
, struct dir_context
*ctx
)
2307 struct vm_area_struct
*vma
;
2308 struct task_struct
*task
;
2309 struct mm_struct
*mm
;
2310 unsigned long nr_files
, pos
, i
;
2311 GENRADIX(struct map_files_info
) fa
;
2312 struct map_files_info
*p
;
2318 task
= get_proc_task(file_inode(file
));
2323 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
))
2327 if (!dir_emit_dots(file
, ctx
))
2330 mm
= get_task_mm(task
);
2334 ret
= mmap_read_lock_killable(mm
);
2343 * We need two passes here:
2345 * 1) Collect vmas of mapped files with mmap_lock taken
2346 * 2) Release mmap_lock and instantiate entries
2348 * otherwise we get lockdep complained, since filldir()
2349 * routine might require mmap_lock taken in might_fault().
2352 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2355 if (++pos
<= ctx
->pos
)
2358 p
= genradix_ptr_alloc(&fa
, nr_files
++, GFP_KERNEL
);
2361 mmap_read_unlock(mm
);
2366 p
->start
= vma
->vm_start
;
2367 p
->end
= vma
->vm_end
;
2368 p
->mode
= vma
->vm_file
->f_mode
;
2370 mmap_read_unlock(mm
);
2373 for (i
= 0; i
< nr_files
; i
++) {
2374 char buf
[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2377 p
= genradix_ptr(&fa
, i
);
2378 len
= snprintf(buf
, sizeof(buf
), "%lx-%lx", p
->start
, p
->end
);
2379 if (!proc_fill_cache(file
, ctx
,
2381 proc_map_files_instantiate
,
2383 (void *)(unsigned long)p
->mode
))
2389 put_task_struct(task
);
2395 static const struct file_operations proc_map_files_operations
= {
2396 .read
= generic_read_dir
,
2397 .iterate_shared
= proc_map_files_readdir
,
2398 .llseek
= generic_file_llseek
,
2401 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2402 struct timers_private
{
2404 struct task_struct
*task
;
2405 struct sighand_struct
*sighand
;
2406 struct pid_namespace
*ns
;
2407 unsigned long flags
;
2410 static void *timers_start(struct seq_file
*m
, loff_t
*pos
)
2412 struct timers_private
*tp
= m
->private;
2414 tp
->task
= get_pid_task(tp
->pid
, PIDTYPE_PID
);
2416 return ERR_PTR(-ESRCH
);
2418 tp
->sighand
= lock_task_sighand(tp
->task
, &tp
->flags
);
2420 return ERR_PTR(-ESRCH
);
2422 return seq_list_start(&tp
->task
->signal
->posix_timers
, *pos
);
2425 static void *timers_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
2427 struct timers_private
*tp
= m
->private;
2428 return seq_list_next(v
, &tp
->task
->signal
->posix_timers
, pos
);
2431 static void timers_stop(struct seq_file
*m
, void *v
)
2433 struct timers_private
*tp
= m
->private;
2436 unlock_task_sighand(tp
->task
, &tp
->flags
);
2441 put_task_struct(tp
->task
);
2446 static int show_timer(struct seq_file
*m
, void *v
)
2448 struct k_itimer
*timer
;
2449 struct timers_private
*tp
= m
->private;
2451 static const char * const nstr
[] = {
2452 [SIGEV_SIGNAL
] = "signal",
2453 [SIGEV_NONE
] = "none",
2454 [SIGEV_THREAD
] = "thread",
2457 timer
= list_entry((struct list_head
*)v
, struct k_itimer
, list
);
2458 notify
= timer
->it_sigev_notify
;
2460 seq_printf(m
, "ID: %d\n", timer
->it_id
);
2461 seq_printf(m
, "signal: %d/%px\n",
2462 timer
->sigq
->info
.si_signo
,
2463 timer
->sigq
->info
.si_value
.sival_ptr
);
2464 seq_printf(m
, "notify: %s/%s.%d\n",
2465 nstr
[notify
& ~SIGEV_THREAD_ID
],
2466 (notify
& SIGEV_THREAD_ID
) ? "tid" : "pid",
2467 pid_nr_ns(timer
->it_pid
, tp
->ns
));
2468 seq_printf(m
, "ClockID: %d\n", timer
->it_clock
);
2473 static const struct seq_operations proc_timers_seq_ops
= {
2474 .start
= timers_start
,
2475 .next
= timers_next
,
2476 .stop
= timers_stop
,
2480 static int proc_timers_open(struct inode
*inode
, struct file
*file
)
2482 struct timers_private
*tp
;
2484 tp
= __seq_open_private(file
, &proc_timers_seq_ops
,
2485 sizeof(struct timers_private
));
2489 tp
->pid
= proc_pid(inode
);
2490 tp
->ns
= proc_pid_ns(inode
->i_sb
);
2494 static const struct file_operations proc_timers_operations
= {
2495 .open
= proc_timers_open
,
2497 .llseek
= seq_lseek
,
2498 .release
= seq_release_private
,
2502 static ssize_t
timerslack_ns_write(struct file
*file
, const char __user
*buf
,
2503 size_t count
, loff_t
*offset
)
2505 struct inode
*inode
= file_inode(file
);
2506 struct task_struct
*p
;
2510 err
= kstrtoull_from_user(buf
, count
, 10, &slack_ns
);
2514 p
= get_proc_task(inode
);
2520 if (!ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
)) {
2527 err
= security_task_setscheduler(p
);
2536 p
->timer_slack_ns
= p
->default_timer_slack_ns
;
2538 p
->timer_slack_ns
= slack_ns
;
2547 static int timerslack_ns_show(struct seq_file
*m
, void *v
)
2549 struct inode
*inode
= m
->private;
2550 struct task_struct
*p
;
2553 p
= get_proc_task(inode
);
2559 if (!ns_capable(__task_cred(p
)->user_ns
, CAP_SYS_NICE
)) {
2566 err
= security_task_getscheduler(p
);
2572 seq_printf(m
, "%llu\n", p
->timer_slack_ns
);
2581 static int timerslack_ns_open(struct inode
*inode
, struct file
*filp
)
2583 return single_open(filp
, timerslack_ns_show
, inode
);
2586 static const struct file_operations proc_pid_set_timerslack_ns_operations
= {
2587 .open
= timerslack_ns_open
,
2589 .write
= timerslack_ns_write
,
2590 .llseek
= seq_lseek
,
2591 .release
= single_release
,
2594 static struct dentry
*proc_pident_instantiate(struct dentry
*dentry
,
2595 struct task_struct
*task
, const void *ptr
)
2597 const struct pid_entry
*p
= ptr
;
2598 struct inode
*inode
;
2599 struct proc_inode
*ei
;
2601 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, p
->mode
);
2603 return ERR_PTR(-ENOENT
);
2606 if (S_ISDIR(inode
->i_mode
))
2607 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2609 inode
->i_op
= p
->iop
;
2611 inode
->i_fop
= p
->fop
;
2613 pid_update_inode(task
, inode
);
2614 d_set_d_op(dentry
, &pid_dentry_operations
);
2615 return d_splice_alias(inode
, dentry
);
2618 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2619 struct dentry
*dentry
,
2620 const struct pid_entry
*p
,
2621 const struct pid_entry
*end
)
2623 struct task_struct
*task
= get_proc_task(dir
);
2624 struct dentry
*res
= ERR_PTR(-ENOENT
);
2630 * Yes, it does not scale. And it should not. Don't add
2631 * new entries into /proc/<tgid>/ without very good reasons.
2633 for (; p
< end
; p
++) {
2634 if (p
->len
!= dentry
->d_name
.len
)
2636 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
)) {
2637 res
= proc_pident_instantiate(dentry
, task
, p
);
2641 put_task_struct(task
);
2646 static int proc_pident_readdir(struct file
*file
, struct dir_context
*ctx
,
2647 const struct pid_entry
*ents
, unsigned int nents
)
2649 struct task_struct
*task
= get_proc_task(file_inode(file
));
2650 const struct pid_entry
*p
;
2655 if (!dir_emit_dots(file
, ctx
))
2658 if (ctx
->pos
>= nents
+ 2)
2661 for (p
= ents
+ (ctx
->pos
- 2); p
< ents
+ nents
; p
++) {
2662 if (!proc_fill_cache(file
, ctx
, p
->name
, p
->len
,
2663 proc_pident_instantiate
, task
, p
))
2668 put_task_struct(task
);
2672 #ifdef CONFIG_SECURITY
2673 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2674 size_t count
, loff_t
*ppos
)
2676 struct inode
* inode
= file_inode(file
);
2679 struct task_struct
*task
= get_proc_task(inode
);
2684 length
= security_getprocattr(task
, PROC_I(inode
)->op
.lsm
,
2685 (char*)file
->f_path
.dentry
->d_name
.name
,
2687 put_task_struct(task
);
2689 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2694 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2695 size_t count
, loff_t
*ppos
)
2697 struct inode
* inode
= file_inode(file
);
2698 struct task_struct
*task
;
2703 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
2708 /* A task may only write its own attributes. */
2709 if (current
!= task
) {
2713 /* Prevent changes to overridden credentials. */
2714 if (current_cred() != current_real_cred()) {
2720 if (count
> PAGE_SIZE
)
2723 /* No partial writes. */
2727 page
= memdup_user(buf
, count
);
2733 /* Guard against adverse ptrace interaction */
2734 rv
= mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
);
2738 rv
= security_setprocattr(PROC_I(inode
)->op
.lsm
,
2739 file
->f_path
.dentry
->d_name
.name
, page
,
2741 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
2748 static const struct file_operations proc_pid_attr_operations
= {
2749 .read
= proc_pid_attr_read
,
2750 .write
= proc_pid_attr_write
,
2751 .llseek
= generic_file_llseek
,
2754 #define LSM_DIR_OPS(LSM) \
2755 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2756 struct dir_context *ctx) \
2758 return proc_pident_readdir(filp, ctx, \
2759 LSM##_attr_dir_stuff, \
2760 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2763 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2764 .read = generic_read_dir, \
2765 .iterate = proc_##LSM##_attr_dir_iterate, \
2766 .llseek = default_llseek, \
2769 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2770 struct dentry *dentry, unsigned int flags) \
2772 return proc_pident_lookup(dir, dentry, \
2773 LSM##_attr_dir_stuff, \
2774 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2777 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2778 .lookup = proc_##LSM##_attr_dir_lookup, \
2779 .getattr = pid_getattr, \
2780 .setattr = proc_setattr, \
2783 #ifdef CONFIG_SECURITY_SMACK
2784 static const struct pid_entry smack_attr_dir_stuff
[] = {
2785 ATTR("smack", "current", 0666),
2790 #ifdef CONFIG_SECURITY_APPARMOR
2791 static const struct pid_entry apparmor_attr_dir_stuff
[] = {
2792 ATTR("apparmor", "current", 0666),
2793 ATTR("apparmor", "prev", 0444),
2794 ATTR("apparmor", "exec", 0666),
2796 LSM_DIR_OPS(apparmor
);
2799 static const struct pid_entry attr_dir_stuff
[] = {
2800 ATTR(NULL
, "current", 0666),
2801 ATTR(NULL
, "prev", 0444),
2802 ATTR(NULL
, "exec", 0666),
2803 ATTR(NULL
, "fscreate", 0666),
2804 ATTR(NULL
, "keycreate", 0666),
2805 ATTR(NULL
, "sockcreate", 0666),
2806 #ifdef CONFIG_SECURITY_SMACK
2808 proc_smack_attr_dir_inode_ops
, proc_smack_attr_dir_ops
),
2810 #ifdef CONFIG_SECURITY_APPARMOR
2811 DIR("apparmor", 0555,
2812 proc_apparmor_attr_dir_inode_ops
, proc_apparmor_attr_dir_ops
),
2816 static int proc_attr_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
2818 return proc_pident_readdir(file
, ctx
,
2819 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2822 static const struct file_operations proc_attr_dir_operations
= {
2823 .read
= generic_read_dir
,
2824 .iterate_shared
= proc_attr_dir_readdir
,
2825 .llseek
= generic_file_llseek
,
2828 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2829 struct dentry
*dentry
, unsigned int flags
)
2831 return proc_pident_lookup(dir
, dentry
,
2833 attr_dir_stuff
+ ARRAY_SIZE(attr_dir_stuff
));
2836 static const struct inode_operations proc_attr_dir_inode_operations
= {
2837 .lookup
= proc_attr_dir_lookup
,
2838 .getattr
= pid_getattr
,
2839 .setattr
= proc_setattr
,
2844 #ifdef CONFIG_ELF_CORE
2845 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2846 size_t count
, loff_t
*ppos
)
2848 struct task_struct
*task
= get_proc_task(file_inode(file
));
2849 struct mm_struct
*mm
;
2850 char buffer
[PROC_NUMBUF
];
2858 mm
= get_task_mm(task
);
2860 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2861 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2862 MMF_DUMP_FILTER_SHIFT
));
2864 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2867 put_task_struct(task
);
2872 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2873 const char __user
*buf
,
2877 struct task_struct
*task
;
2878 struct mm_struct
*mm
;
2884 ret
= kstrtouint_from_user(buf
, count
, 0, &val
);
2889 task
= get_proc_task(file_inode(file
));
2893 mm
= get_task_mm(task
);
2898 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2900 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2902 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2907 put_task_struct(task
);
2914 static const struct file_operations proc_coredump_filter_operations
= {
2915 .read
= proc_coredump_filter_read
,
2916 .write
= proc_coredump_filter_write
,
2917 .llseek
= generic_file_llseek
,
2921 #ifdef CONFIG_TASK_IO_ACCOUNTING
2922 static int do_io_accounting(struct task_struct
*task
, struct seq_file
*m
, int whole
)
2924 struct task_io_accounting acct
= task
->ioac
;
2925 unsigned long flags
;
2928 result
= mutex_lock_killable(&task
->signal
->exec_update_mutex
);
2932 if (!ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
)) {
2937 if (whole
&& lock_task_sighand(task
, &flags
)) {
2938 struct task_struct
*t
= task
;
2940 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2941 while_each_thread(task
, t
)
2942 task_io_accounting_add(&acct
, &t
->ioac
);
2944 unlock_task_sighand(task
, &flags
);
2951 "read_bytes: %llu\n"
2952 "write_bytes: %llu\n"
2953 "cancelled_write_bytes: %llu\n",
2954 (unsigned long long)acct
.rchar
,
2955 (unsigned long long)acct
.wchar
,
2956 (unsigned long long)acct
.syscr
,
2957 (unsigned long long)acct
.syscw
,
2958 (unsigned long long)acct
.read_bytes
,
2959 (unsigned long long)acct
.write_bytes
,
2960 (unsigned long long)acct
.cancelled_write_bytes
);
2964 mutex_unlock(&task
->signal
->exec_update_mutex
);
2968 static int proc_tid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2969 struct pid
*pid
, struct task_struct
*task
)
2971 return do_io_accounting(task
, m
, 0);
2974 static int proc_tgid_io_accounting(struct seq_file
*m
, struct pid_namespace
*ns
,
2975 struct pid
*pid
, struct task_struct
*task
)
2977 return do_io_accounting(task
, m
, 1);
2979 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2981 #ifdef CONFIG_USER_NS
2982 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2983 const struct seq_operations
*seq_ops
)
2985 struct user_namespace
*ns
= NULL
;
2986 struct task_struct
*task
;
2987 struct seq_file
*seq
;
2990 task
= get_proc_task(inode
);
2993 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2995 put_task_struct(task
);
3000 ret
= seq_open(file
, seq_ops
);
3004 seq
= file
->private_data
;
3014 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
3016 struct seq_file
*seq
= file
->private_data
;
3017 struct user_namespace
*ns
= seq
->private;
3019 return seq_release(inode
, file
);
3022 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
3024 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
3027 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
3029 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
3032 static int proc_projid_map_open(struct inode
*inode
, struct file
*file
)
3034 return proc_id_map_open(inode
, file
, &proc_projid_seq_operations
);
3037 static const struct file_operations proc_uid_map_operations
= {
3038 .open
= proc_uid_map_open
,
3039 .write
= proc_uid_map_write
,
3041 .llseek
= seq_lseek
,
3042 .release
= proc_id_map_release
,
3045 static const struct file_operations proc_gid_map_operations
= {
3046 .open
= proc_gid_map_open
,
3047 .write
= proc_gid_map_write
,
3049 .llseek
= seq_lseek
,
3050 .release
= proc_id_map_release
,
3053 static const struct file_operations proc_projid_map_operations
= {
3054 .open
= proc_projid_map_open
,
3055 .write
= proc_projid_map_write
,
3057 .llseek
= seq_lseek
,
3058 .release
= proc_id_map_release
,
3061 static int proc_setgroups_open(struct inode
*inode
, struct file
*file
)
3063 struct user_namespace
*ns
= NULL
;
3064 struct task_struct
*task
;
3068 task
= get_proc_task(inode
);
3071 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
3073 put_task_struct(task
);
3078 if (file
->f_mode
& FMODE_WRITE
) {
3080 if (!ns_capable(ns
, CAP_SYS_ADMIN
))
3084 ret
= single_open(file
, &proc_setgroups_show
, ns
);
3095 static int proc_setgroups_release(struct inode
*inode
, struct file
*file
)
3097 struct seq_file
*seq
= file
->private_data
;
3098 struct user_namespace
*ns
= seq
->private;
3099 int ret
= single_release(inode
, file
);
3104 static const struct file_operations proc_setgroups_operations
= {
3105 .open
= proc_setgroups_open
,
3106 .write
= proc_setgroups_write
,
3108 .llseek
= seq_lseek
,
3109 .release
= proc_setgroups_release
,
3111 #endif /* CONFIG_USER_NS */
3113 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
3114 struct pid
*pid
, struct task_struct
*task
)
3116 int err
= lock_trace(task
);
3118 seq_printf(m
, "%08x\n", task
->personality
);
3124 #ifdef CONFIG_LIVEPATCH
3125 static int proc_pid_patch_state(struct seq_file
*m
, struct pid_namespace
*ns
,
3126 struct pid
*pid
, struct task_struct
*task
)
3128 seq_printf(m
, "%d\n", task
->patch_state
);
3131 #endif /* CONFIG_LIVEPATCH */
3133 #ifdef CONFIG_STACKLEAK_METRICS
3134 static int proc_stack_depth(struct seq_file
*m
, struct pid_namespace
*ns
,
3135 struct pid
*pid
, struct task_struct
*task
)
3137 unsigned long prev_depth
= THREAD_SIZE
-
3138 (task
->prev_lowest_stack
& (THREAD_SIZE
- 1));
3139 unsigned long depth
= THREAD_SIZE
-
3140 (task
->lowest_stack
& (THREAD_SIZE
- 1));
3142 seq_printf(m
, "previous stack depth: %lu\nstack depth: %lu\n",
3146 #endif /* CONFIG_STACKLEAK_METRICS */
3151 static const struct file_operations proc_task_operations
;
3152 static const struct inode_operations proc_task_inode_operations
;
3154 static const struct pid_entry tgid_base_stuff
[] = {
3155 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
3156 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3157 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
3158 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3159 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3161 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3163 REG("environ", S_IRUSR
, proc_environ_operations
),
3164 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3165 ONE("status", S_IRUGO
, proc_pid_status
),
3166 ONE("personality", S_IRUSR
, proc_pid_personality
),
3167 ONE("limits", S_IRUGO
, proc_pid_limits
),
3168 #ifdef CONFIG_SCHED_DEBUG
3169 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3171 #ifdef CONFIG_SCHED_AUTOGROUP
3172 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
3174 #ifdef CONFIG_TIME_NS
3175 REG("timens_offsets", S_IRUGO
|S_IWUSR
, proc_timens_offsets_operations
),
3177 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3178 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3179 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3181 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3182 ONE("stat", S_IRUGO
, proc_tgid_stat
),
3183 ONE("statm", S_IRUGO
, proc_pid_statm
),
3184 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3186 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3188 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3189 LNK("cwd", proc_cwd_link
),
3190 LNK("root", proc_root_link
),
3191 LNK("exe", proc_exe_link
),
3192 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3193 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3194 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
3195 #ifdef CONFIG_PROC_PAGE_MONITOR
3196 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3197 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3198 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3199 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3201 #ifdef CONFIG_SECURITY
3202 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3204 #ifdef CONFIG_KALLSYMS
3205 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3207 #ifdef CONFIG_STACKTRACE
3208 ONE("stack", S_IRUSR
, proc_pid_stack
),
3210 #ifdef CONFIG_SCHED_INFO
3211 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3213 #ifdef CONFIG_LATENCYTOP
3214 REG("latency", S_IRUGO
, proc_lstats_operations
),
3216 #ifdef CONFIG_PROC_PID_CPUSET
3217 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3219 #ifdef CONFIG_CGROUPS
3220 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3222 #ifdef CONFIG_PROC_CPU_RESCTRL
3223 ONE("cpu_resctrl_groups", S_IRUGO
, proc_resctrl_show
),
3225 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3226 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3227 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3229 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3230 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3232 #ifdef CONFIG_FAULT_INJECTION
3233 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3234 REG("fail-nth", 0644, proc_fail_nth_operations
),
3236 #ifdef CONFIG_ELF_CORE
3237 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3239 #ifdef CONFIG_TASK_IO_ACCOUNTING
3240 ONE("io", S_IRUSR
, proc_tgid_io_accounting
),
3242 #ifdef CONFIG_USER_NS
3243 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3244 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3245 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3246 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3248 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3249 REG("timers", S_IRUGO
, proc_timers_operations
),
3251 REG("timerslack_ns", S_IRUGO
|S_IWUGO
, proc_pid_set_timerslack_ns_operations
),
3252 #ifdef CONFIG_LIVEPATCH
3253 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3255 #ifdef CONFIG_STACKLEAK_METRICS
3256 ONE("stack_depth", S_IRUGO
, proc_stack_depth
),
3258 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3259 ONE("arch_status", S_IRUGO
, proc_pid_arch_status
),
3263 static int proc_tgid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3265 return proc_pident_readdir(file
, ctx
,
3266 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3269 static const struct file_operations proc_tgid_base_operations
= {
3270 .read
= generic_read_dir
,
3271 .iterate_shared
= proc_tgid_base_readdir
,
3272 .llseek
= generic_file_llseek
,
3275 struct pid
*tgid_pidfd_to_pid(const struct file
*file
)
3277 if (file
->f_op
!= &proc_tgid_base_operations
)
3278 return ERR_PTR(-EBADF
);
3280 return proc_pid(file_inode(file
));
3283 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3285 return proc_pident_lookup(dir
, dentry
,
3287 tgid_base_stuff
+ ARRAY_SIZE(tgid_base_stuff
));
3290 static const struct inode_operations proc_tgid_base_inode_operations
= {
3291 .lookup
= proc_tgid_base_lookup
,
3292 .getattr
= pid_getattr
,
3293 .setattr
= proc_setattr
,
3294 .permission
= proc_pid_permission
,
3298 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3299 * @pid: pid that should be flushed.
3301 * This function walks a list of inodes (that belong to any proc
3302 * filesystem) that are attached to the pid and flushes them from
3305 * It is safe and reasonable to cache /proc entries for a task until
3306 * that task exits. After that they just clog up the dcache with
3307 * useless entries, possibly causing useful dcache entries to be
3308 * flushed instead. This routine is provided to flush those useless
3309 * dcache entries when a process is reaped.
3311 * NOTE: This routine is just an optimization so it does not guarantee
3312 * that no dcache entries will exist after a process is reaped
3313 * it just makes it very unlikely that any will persist.
3316 void proc_flush_pid(struct pid
*pid
)
3318 proc_invalidate_siblings_dcache(&pid
->inodes
, &pid
->lock
);
3321 static struct dentry
*proc_pid_instantiate(struct dentry
* dentry
,
3322 struct task_struct
*task
, const void *ptr
)
3324 struct inode
*inode
;
3326 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3328 return ERR_PTR(-ENOENT
);
3330 inode
->i_op
= &proc_tgid_base_inode_operations
;
3331 inode
->i_fop
= &proc_tgid_base_operations
;
3332 inode
->i_flags
|=S_IMMUTABLE
;
3334 set_nlink(inode
, nlink_tgid
);
3335 pid_update_inode(task
, inode
);
3337 d_set_d_op(dentry
, &pid_dentry_operations
);
3338 return d_splice_alias(inode
, dentry
);
3341 struct dentry
*proc_pid_lookup(struct dentry
*dentry
, unsigned int flags
)
3343 struct task_struct
*task
;
3345 struct proc_fs_info
*fs_info
;
3346 struct pid_namespace
*ns
;
3347 struct dentry
*result
= ERR_PTR(-ENOENT
);
3349 tgid
= name_to_int(&dentry
->d_name
);
3353 fs_info
= proc_sb_info(dentry
->d_sb
);
3354 ns
= fs_info
->pid_ns
;
3356 task
= find_task_by_pid_ns(tgid
, ns
);
3358 get_task_struct(task
);
3363 /* Limit procfs to only ptraceable tasks */
3364 if (fs_info
->hide_pid
== HIDEPID_NOT_PTRACEABLE
) {
3365 if (!has_pid_permissions(fs_info
, task
, HIDEPID_NO_ACCESS
))
3369 result
= proc_pid_instantiate(dentry
, task
, NULL
);
3371 put_task_struct(task
);
3377 * Find the first task with tgid >= tgid
3382 struct task_struct
*task
;
3384 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3389 put_task_struct(iter
.task
);
3393 pid
= find_ge_pid(iter
.tgid
, ns
);
3395 iter
.tgid
= pid_nr_ns(pid
, ns
);
3396 iter
.task
= pid_task(pid
, PIDTYPE_TGID
);
3401 get_task_struct(iter
.task
);
3407 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3409 /* for the /proc/ directory itself, after non-process stuff has been done */
3410 int proc_pid_readdir(struct file
*file
, struct dir_context
*ctx
)
3412 struct tgid_iter iter
;
3413 struct proc_fs_info
*fs_info
= proc_sb_info(file_inode(file
)->i_sb
);
3414 struct pid_namespace
*ns
= proc_pid_ns(file_inode(file
)->i_sb
);
3415 loff_t pos
= ctx
->pos
;
3417 if (pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3420 if (pos
== TGID_OFFSET
- 2) {
3421 struct inode
*inode
= d_inode(fs_info
->proc_self
);
3422 if (!dir_emit(ctx
, "self", 4, inode
->i_ino
, DT_LNK
))
3424 ctx
->pos
= pos
= pos
+ 1;
3426 if (pos
== TGID_OFFSET
- 1) {
3427 struct inode
*inode
= d_inode(fs_info
->proc_thread_self
);
3428 if (!dir_emit(ctx
, "thread-self", 11, inode
->i_ino
, DT_LNK
))
3430 ctx
->pos
= pos
= pos
+ 1;
3432 iter
.tgid
= pos
- TGID_OFFSET
;
3434 for (iter
= next_tgid(ns
, iter
);
3436 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3441 if (!has_pid_permissions(fs_info
, iter
.task
, HIDEPID_INVISIBLE
))
3444 len
= snprintf(name
, sizeof(name
), "%u", iter
.tgid
);
3445 ctx
->pos
= iter
.tgid
+ TGID_OFFSET
;
3446 if (!proc_fill_cache(file
, ctx
, name
, len
,
3447 proc_pid_instantiate
, iter
.task
, NULL
)) {
3448 put_task_struct(iter
.task
);
3452 ctx
->pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3457 * proc_tid_comm_permission is a special permission function exclusively
3458 * used for the node /proc/<pid>/task/<tid>/comm.
3459 * It bypasses generic permission checks in the case where a task of the same
3460 * task group attempts to access the node.
3461 * The rationale behind this is that glibc and bionic access this node for
3462 * cross thread naming (pthread_set/getname_np(!self)). However, if
3463 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3464 * which locks out the cross thread naming implementation.
3465 * This function makes sure that the node is always accessible for members of
3466 * same thread group.
3468 static int proc_tid_comm_permission(struct inode
*inode
, int mask
)
3470 bool is_same_tgroup
;
3471 struct task_struct
*task
;
3473 task
= get_proc_task(inode
);
3476 is_same_tgroup
= same_thread_group(current
, task
);
3477 put_task_struct(task
);
3479 if (likely(is_same_tgroup
&& !(mask
& MAY_EXEC
))) {
3480 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3481 * read or written by the members of the corresponding
3487 return generic_permission(inode
, mask
);
3490 static const struct inode_operations proc_tid_comm_inode_operations
= {
3491 .permission
= proc_tid_comm_permission
,
3497 static const struct pid_entry tid_base_stuff
[] = {
3498 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3499 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3500 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3502 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3504 REG("environ", S_IRUSR
, proc_environ_operations
),
3505 REG("auxv", S_IRUSR
, proc_auxv_operations
),
3506 ONE("status", S_IRUGO
, proc_pid_status
),
3507 ONE("personality", S_IRUSR
, proc_pid_personality
),
3508 ONE("limits", S_IRUGO
, proc_pid_limits
),
3509 #ifdef CONFIG_SCHED_DEBUG
3510 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3512 NOD("comm", S_IFREG
|S_IRUGO
|S_IWUSR
,
3513 &proc_tid_comm_inode_operations
,
3514 &proc_pid_set_comm_operations
, {}),
3515 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3516 ONE("syscall", S_IRUSR
, proc_pid_syscall
),
3518 REG("cmdline", S_IRUGO
, proc_pid_cmdline_ops
),
3519 ONE("stat", S_IRUGO
, proc_tid_stat
),
3520 ONE("statm", S_IRUGO
, proc_pid_statm
),
3521 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3522 #ifdef CONFIG_PROC_CHILDREN
3523 REG("children", S_IRUGO
, proc_tid_children_operations
),
3526 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3528 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3529 LNK("cwd", proc_cwd_link
),
3530 LNK("root", proc_root_link
),
3531 LNK("exe", proc_exe_link
),
3532 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3533 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3534 #ifdef CONFIG_PROC_PAGE_MONITOR
3535 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3536 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3537 REG("smaps_rollup", S_IRUGO
, proc_pid_smaps_rollup_operations
),
3538 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3540 #ifdef CONFIG_SECURITY
3541 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3543 #ifdef CONFIG_KALLSYMS
3544 ONE("wchan", S_IRUGO
, proc_pid_wchan
),
3546 #ifdef CONFIG_STACKTRACE
3547 ONE("stack", S_IRUSR
, proc_pid_stack
),
3549 #ifdef CONFIG_SCHED_INFO
3550 ONE("schedstat", S_IRUGO
, proc_pid_schedstat
),
3552 #ifdef CONFIG_LATENCYTOP
3553 REG("latency", S_IRUGO
, proc_lstats_operations
),
3555 #ifdef CONFIG_PROC_PID_CPUSET
3556 ONE("cpuset", S_IRUGO
, proc_cpuset_show
),
3558 #ifdef CONFIG_CGROUPS
3559 ONE("cgroup", S_IRUGO
, proc_cgroup_show
),
3561 #ifdef CONFIG_PROC_CPU_RESCTRL
3562 ONE("cpu_resctrl_groups", S_IRUGO
, proc_resctrl_show
),
3564 ONE("oom_score", S_IRUGO
, proc_oom_score
),
3565 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adj_operations
),
3566 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3568 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3569 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3571 #ifdef CONFIG_FAULT_INJECTION
3572 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3573 REG("fail-nth", 0644, proc_fail_nth_operations
),
3575 #ifdef CONFIG_TASK_IO_ACCOUNTING
3576 ONE("io", S_IRUSR
, proc_tid_io_accounting
),
3578 #ifdef CONFIG_USER_NS
3579 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3580 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3581 REG("projid_map", S_IRUGO
|S_IWUSR
, proc_projid_map_operations
),
3582 REG("setgroups", S_IRUGO
|S_IWUSR
, proc_setgroups_operations
),
3584 #ifdef CONFIG_LIVEPATCH
3585 ONE("patch_state", S_IRUSR
, proc_pid_patch_state
),
3587 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3588 ONE("arch_status", S_IRUGO
, proc_pid_arch_status
),
3592 static int proc_tid_base_readdir(struct file
*file
, struct dir_context
*ctx
)
3594 return proc_pident_readdir(file
, ctx
,
3595 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3598 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3600 return proc_pident_lookup(dir
, dentry
,
3602 tid_base_stuff
+ ARRAY_SIZE(tid_base_stuff
));
3605 static const struct file_operations proc_tid_base_operations
= {
3606 .read
= generic_read_dir
,
3607 .iterate_shared
= proc_tid_base_readdir
,
3608 .llseek
= generic_file_llseek
,
3611 static const struct inode_operations proc_tid_base_inode_operations
= {
3612 .lookup
= proc_tid_base_lookup
,
3613 .getattr
= pid_getattr
,
3614 .setattr
= proc_setattr
,
3617 static struct dentry
*proc_task_instantiate(struct dentry
*dentry
,
3618 struct task_struct
*task
, const void *ptr
)
3620 struct inode
*inode
;
3621 inode
= proc_pid_make_inode(dentry
->d_sb
, task
, S_IFDIR
| S_IRUGO
| S_IXUGO
);
3623 return ERR_PTR(-ENOENT
);
3625 inode
->i_op
= &proc_tid_base_inode_operations
;
3626 inode
->i_fop
= &proc_tid_base_operations
;
3627 inode
->i_flags
|= S_IMMUTABLE
;
3629 set_nlink(inode
, nlink_tid
);
3630 pid_update_inode(task
, inode
);
3632 d_set_d_op(dentry
, &pid_dentry_operations
);
3633 return d_splice_alias(inode
, dentry
);
3636 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3638 struct task_struct
*task
;
3639 struct task_struct
*leader
= get_proc_task(dir
);
3641 struct proc_fs_info
*fs_info
;
3642 struct pid_namespace
*ns
;
3643 struct dentry
*result
= ERR_PTR(-ENOENT
);
3648 tid
= name_to_int(&dentry
->d_name
);
3652 fs_info
= proc_sb_info(dentry
->d_sb
);
3653 ns
= fs_info
->pid_ns
;
3655 task
= find_task_by_pid_ns(tid
, ns
);
3657 get_task_struct(task
);
3661 if (!same_thread_group(leader
, task
))
3664 result
= proc_task_instantiate(dentry
, task
, NULL
);
3666 put_task_struct(task
);
3668 put_task_struct(leader
);
3674 * Find the first tid of a thread group to return to user space.
3676 * Usually this is just the thread group leader, but if the users
3677 * buffer was too small or there was a seek into the middle of the
3678 * directory we have more work todo.
3680 * In the case of a short read we start with find_task_by_pid.
3682 * In the case of a seek we start with the leader and walk nr
3685 static struct task_struct
*first_tid(struct pid
*pid
, int tid
, loff_t f_pos
,
3686 struct pid_namespace
*ns
)
3688 struct task_struct
*pos
, *task
;
3689 unsigned long nr
= f_pos
;
3691 if (nr
!= f_pos
) /* 32bit overflow? */
3695 task
= pid_task(pid
, PIDTYPE_PID
);
3699 /* Attempt to start with the tid of a thread */
3701 pos
= find_task_by_pid_ns(tid
, ns
);
3702 if (pos
&& same_thread_group(pos
, task
))
3706 /* If nr exceeds the number of threads there is nothing todo */
3707 if (nr
>= get_nr_threads(task
))
3710 /* If we haven't found our starting place yet start
3711 * with the leader and walk nr threads forward.
3713 pos
= task
= task
->group_leader
;
3717 } while_each_thread(task
, pos
);
3722 get_task_struct(pos
);
3729 * Find the next thread in the thread list.
3730 * Return NULL if there is an error or no next thread.
3732 * The reference to the input task_struct is released.
3734 static struct task_struct
*next_tid(struct task_struct
*start
)
3736 struct task_struct
*pos
= NULL
;
3738 if (pid_alive(start
)) {
3739 pos
= next_thread(start
);
3740 if (thread_group_leader(pos
))
3743 get_task_struct(pos
);
3746 put_task_struct(start
);
3750 /* for the /proc/TGID/task/ directories */
3751 static int proc_task_readdir(struct file
*file
, struct dir_context
*ctx
)
3753 struct inode
*inode
= file_inode(file
);
3754 struct task_struct
*task
;
3755 struct pid_namespace
*ns
;
3758 if (proc_inode_is_dead(inode
))
3761 if (!dir_emit_dots(file
, ctx
))
3764 /* f_version caches the tgid value that the last readdir call couldn't
3765 * return. lseek aka telldir automagically resets f_version to 0.
3767 ns
= proc_pid_ns(inode
->i_sb
);
3768 tid
= (int)file
->f_version
;
3769 file
->f_version
= 0;
3770 for (task
= first_tid(proc_pid(inode
), tid
, ctx
->pos
- 2, ns
);
3772 task
= next_tid(task
), ctx
->pos
++) {
3775 tid
= task_pid_nr_ns(task
, ns
);
3776 len
= snprintf(name
, sizeof(name
), "%u", tid
);
3777 if (!proc_fill_cache(file
, ctx
, name
, len
,
3778 proc_task_instantiate
, task
, NULL
)) {
3779 /* returning this tgid failed, save it as the first
3780 * pid for the next readir call */
3781 file
->f_version
= (u64
)tid
;
3782 put_task_struct(task
);
3790 static int proc_task_getattr(const struct path
*path
, struct kstat
*stat
,
3791 u32 request_mask
, unsigned int query_flags
)
3793 struct inode
*inode
= d_inode(path
->dentry
);
3794 struct task_struct
*p
= get_proc_task(inode
);
3795 generic_fillattr(inode
, stat
);
3798 stat
->nlink
+= get_nr_threads(p
);
3805 static const struct inode_operations proc_task_inode_operations
= {
3806 .lookup
= proc_task_lookup
,
3807 .getattr
= proc_task_getattr
,
3808 .setattr
= proc_setattr
,
3809 .permission
= proc_pid_permission
,
3812 static const struct file_operations proc_task_operations
= {
3813 .read
= generic_read_dir
,
3814 .iterate_shared
= proc_task_readdir
,
3815 .llseek
= generic_file_llseek
,
3818 void __init
set_proc_pid_nlink(void)
3820 nlink_tid
= pid_entry_nlink(tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3821 nlink_tgid
= pid_entry_nlink(tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));