4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/stacktrace.h>
69 #include <linux/resource.h>
70 #include <linux/module.h>
71 #include <linux/mount.h>
72 #include <linux/security.h>
73 #include <linux/ptrace.h>
74 #include <linux/tracehook.h>
75 #include <linux/cgroup.h>
76 #include <linux/cpuset.h>
77 #include <linux/audit.h>
78 #include <linux/poll.h>
79 #include <linux/nsproxy.h>
80 #include <linux/oom.h>
81 #include <linux/elf.h>
82 #include <linux/pid_namespace.h>
83 #include <linux/fs_struct.h>
87 * Implementing inode permission operations in /proc is almost
88 * certainly an error. Permission checks need to happen during
89 * each system call not at open time. The reason is that most of
90 * what we wish to check for permissions in /proc varies at runtime.
92 * The classic example of a problem is opening file descriptors
93 * in /proc for a task before it execs a suid executable.
100 const struct inode_operations
*iop
;
101 const struct file_operations
*fop
;
105 #define NOD(NAME, MODE, IOP, FOP, OP) { \
107 .len = sizeof(NAME) - 1, \
114 #define DIR(NAME, MODE, iops, fops) \
115 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
116 #define LNK(NAME, get_link) \
117 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
118 &proc_pid_link_inode_operations, NULL, \
119 { .proc_get_link = get_link } )
120 #define REG(NAME, MODE, fops) \
121 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
122 #define INF(NAME, MODE, read) \
123 NOD(NAME, (S_IFREG|(MODE)), \
124 NULL, &proc_info_file_operations, \
125 { .proc_read = read } )
126 #define ONE(NAME, MODE, show) \
127 NOD(NAME, (S_IFREG|(MODE)), \
128 NULL, &proc_single_file_operations, \
129 { .proc_show = show } )
132 * Count the number of hardlinks for the pid_entry table, excluding the .
135 static unsigned int pid_entry_count_dirs(const struct pid_entry
*entries
,
142 for (i
= 0; i
< n
; ++i
) {
143 if (S_ISDIR(entries
[i
].mode
))
150 static int get_fs_path(struct task_struct
*task
, struct path
*path
, bool root
)
152 struct fs_struct
*fs
;
153 int result
= -ENOENT
;
158 read_lock(&fs
->lock
);
159 *path
= root
? fs
->root
: fs
->pwd
;
161 read_unlock(&fs
->lock
);
168 static int get_nr_threads(struct task_struct
*tsk
)
173 if (lock_task_sighand(tsk
, &flags
)) {
174 count
= atomic_read(&tsk
->signal
->count
);
175 unlock_task_sighand(tsk
, &flags
);
180 static int proc_cwd_link(struct inode
*inode
, struct path
*path
)
182 struct task_struct
*task
= get_proc_task(inode
);
183 int result
= -ENOENT
;
186 result
= get_fs_path(task
, path
, 0);
187 put_task_struct(task
);
192 static int proc_root_link(struct inode
*inode
, struct path
*path
)
194 struct task_struct
*task
= get_proc_task(inode
);
195 int result
= -ENOENT
;
198 result
= get_fs_path(task
, path
, 1);
199 put_task_struct(task
);
205 * Return zero if current may access user memory in @task, -error if not.
207 static int check_mem_permission(struct task_struct
*task
)
210 * A task can always look at itself, in case it chooses
211 * to use system calls instead of load instructions.
217 * If current is actively ptrace'ing, and would also be
218 * permitted to freshly attach with ptrace now, permit it.
220 if (task_is_stopped_or_traced(task
)) {
223 match
= (tracehook_tracer_task(task
) == current
);
225 if (match
&& ptrace_may_access(task
, PTRACE_MODE_ATTACH
))
230 * Noone else is allowed.
235 struct mm_struct
*mm_for_maps(struct task_struct
*task
)
237 struct mm_struct
*mm
;
239 if (mutex_lock_killable(&task
->cred_guard_mutex
))
242 mm
= get_task_mm(task
);
243 if (mm
&& mm
!= current
->mm
&&
244 !ptrace_may_access(task
, PTRACE_MODE_READ
)) {
248 mutex_unlock(&task
->cred_guard_mutex
);
253 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
257 struct mm_struct
*mm
= get_task_mm(task
);
261 goto out_mm
; /* Shh! No looking before we're done */
263 len
= mm
->arg_end
- mm
->arg_start
;
268 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
270 // If the nul at the end of args has been overwritten, then
271 // assume application is using setproctitle(3).
272 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
273 len
= strnlen(buffer
, res
);
277 len
= mm
->env_end
- mm
->env_start
;
278 if (len
> PAGE_SIZE
- res
)
279 len
= PAGE_SIZE
- res
;
280 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
281 res
= strnlen(buffer
, res
);
290 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
293 struct mm_struct
*mm
= get_task_mm(task
);
295 unsigned int nwords
= 0;
298 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
299 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
302 memcpy(buffer
, mm
->saved_auxv
, res
);
309 #ifdef CONFIG_KALLSYMS
311 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
312 * Returns the resolved symbol. If that fails, simply return the address.
314 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
317 char symname
[KSYM_NAME_LEN
];
319 wchan
= get_wchan(task
);
321 if (lookup_symbol_name(wchan
, symname
) < 0)
322 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
325 return sprintf(buffer
, "%lu", wchan
);
327 return sprintf(buffer
, "%s", symname
);
329 #endif /* CONFIG_KALLSYMS */
331 #ifdef CONFIG_STACKTRACE
333 #define MAX_STACK_TRACE_DEPTH 64
335 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
336 struct pid
*pid
, struct task_struct
*task
)
338 struct stack_trace trace
;
339 unsigned long *entries
;
342 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
346 trace
.nr_entries
= 0;
347 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
348 trace
.entries
= entries
;
350 save_stack_trace_tsk(task
, &trace
);
352 for (i
= 0; i
< trace
.nr_entries
; i
++) {
353 seq_printf(m
, "[<%p>] %pS\n",
354 (void *)entries
[i
], (void *)entries
[i
]);
362 #ifdef CONFIG_SCHEDSTATS
364 * Provides /proc/PID/schedstat
366 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
368 return sprintf(buffer
, "%llu %llu %lu\n",
369 (unsigned long long)task
->se
.sum_exec_runtime
,
370 (unsigned long long)task
->sched_info
.run_delay
,
371 task
->sched_info
.pcount
);
375 #ifdef CONFIG_LATENCYTOP
376 static int lstats_show_proc(struct seq_file
*m
, void *v
)
379 struct inode
*inode
= m
->private;
380 struct task_struct
*task
= get_proc_task(inode
);
384 seq_puts(m
, "Latency Top version : v0.1\n");
385 for (i
= 0; i
< 32; i
++) {
386 if (task
->latency_record
[i
].backtrace
[0]) {
388 seq_printf(m
, "%i %li %li ",
389 task
->latency_record
[i
].count
,
390 task
->latency_record
[i
].time
,
391 task
->latency_record
[i
].max
);
392 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
393 char sym
[KSYM_SYMBOL_LEN
];
395 if (!task
->latency_record
[i
].backtrace
[q
])
397 if (task
->latency_record
[i
].backtrace
[q
] == ULONG_MAX
)
399 sprint_symbol(sym
, task
->latency_record
[i
].backtrace
[q
]);
400 c
= strchr(sym
, '+');
403 seq_printf(m
, "%s ", sym
);
409 put_task_struct(task
);
413 static int lstats_open(struct inode
*inode
, struct file
*file
)
415 return single_open(file
, lstats_show_proc
, inode
);
418 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
419 size_t count
, loff_t
*offs
)
421 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
425 clear_all_latency_tracing(task
);
426 put_task_struct(task
);
431 static const struct file_operations proc_lstats_operations
= {
434 .write
= lstats_write
,
436 .release
= single_release
,
441 /* The badness from the OOM killer */
442 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
443 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
445 unsigned long points
;
446 struct timespec uptime
;
448 do_posix_clock_monotonic_gettime(&uptime
);
449 read_lock(&tasklist_lock
);
450 points
= badness(task
->group_leader
, uptime
.tv_sec
);
451 read_unlock(&tasklist_lock
);
452 return sprintf(buffer
, "%lu\n", points
);
460 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
461 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
462 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
463 [RLIMIT_DATA
] = {"Max data size", "bytes"},
464 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
465 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
466 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
467 [RLIMIT_NPROC
] = {"Max processes", "processes"},
468 [RLIMIT_NOFILE
] = {"Max open files", "files"},
469 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
470 [RLIMIT_AS
] = {"Max address space", "bytes"},
471 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
472 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
473 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
474 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
475 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
476 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
479 /* Display limits for a process */
480 static int proc_pid_limits(struct task_struct
*task
, char *buffer
)
485 char *bufptr
= buffer
;
487 struct rlimit rlim
[RLIM_NLIMITS
];
489 if (!lock_task_sighand(task
, &flags
))
491 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
492 unlock_task_sighand(task
, &flags
);
495 * print the file header
497 count
+= sprintf(&bufptr
[count
], "%-25s %-20s %-20s %-10s\n",
498 "Limit", "Soft Limit", "Hard Limit", "Units");
500 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
501 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
502 count
+= sprintf(&bufptr
[count
], "%-25s %-20s ",
503 lnames
[i
].name
, "unlimited");
505 count
+= sprintf(&bufptr
[count
], "%-25s %-20lu ",
506 lnames
[i
].name
, rlim
[i
].rlim_cur
);
508 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
509 count
+= sprintf(&bufptr
[count
], "%-20s ", "unlimited");
511 count
+= sprintf(&bufptr
[count
], "%-20lu ",
515 count
+= sprintf(&bufptr
[count
], "%-10s\n",
518 count
+= sprintf(&bufptr
[count
], "\n");
524 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
525 static int proc_pid_syscall(struct task_struct
*task
, char *buffer
)
528 unsigned long args
[6], sp
, pc
;
530 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
531 return sprintf(buffer
, "running\n");
534 return sprintf(buffer
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
536 return sprintf(buffer
,
537 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
539 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
542 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
544 /************************************************************************/
545 /* Here the fs part begins */
546 /************************************************************************/
548 /* permission checks */
549 static int proc_fd_access_allowed(struct inode
*inode
)
551 struct task_struct
*task
;
553 /* Allow access to a task's file descriptors if it is us or we
554 * may use ptrace attach to the process and find out that
557 task
= get_proc_task(inode
);
559 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ
);
560 put_task_struct(task
);
565 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
568 struct inode
*inode
= dentry
->d_inode
;
570 if (attr
->ia_valid
& ATTR_MODE
)
573 error
= inode_change_ok(inode
, attr
);
575 error
= inode_setattr(inode
, attr
);
579 static const struct inode_operations proc_def_inode_operations
= {
580 .setattr
= proc_setattr
,
583 static int mounts_open_common(struct inode
*inode
, struct file
*file
,
584 const struct seq_operations
*op
)
586 struct task_struct
*task
= get_proc_task(inode
);
588 struct mnt_namespace
*ns
= NULL
;
590 struct proc_mounts
*p
;
595 nsp
= task_nsproxy(task
);
602 if (ns
&& get_fs_path(task
, &root
, 1) == 0)
604 put_task_struct(task
);
613 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
617 file
->private_data
= &p
->m
;
618 ret
= seq_open(file
, op
);
625 p
->event
= ns
->event
;
639 static int mounts_release(struct inode
*inode
, struct file
*file
)
641 struct proc_mounts
*p
= file
->private_data
;
644 return seq_release(inode
, file
);
647 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
649 struct proc_mounts
*p
= file
->private_data
;
650 struct mnt_namespace
*ns
= p
->ns
;
651 unsigned res
= POLLIN
| POLLRDNORM
;
653 poll_wait(file
, &ns
->poll
, wait
);
655 spin_lock(&vfsmount_lock
);
656 if (p
->event
!= ns
->event
) {
657 p
->event
= ns
->event
;
658 res
|= POLLERR
| POLLPRI
;
660 spin_unlock(&vfsmount_lock
);
665 static int mounts_open(struct inode
*inode
, struct file
*file
)
667 return mounts_open_common(inode
, file
, &mounts_op
);
670 static const struct file_operations proc_mounts_operations
= {
674 .release
= mounts_release
,
678 static int mountinfo_open(struct inode
*inode
, struct file
*file
)
680 return mounts_open_common(inode
, file
, &mountinfo_op
);
683 static const struct file_operations proc_mountinfo_operations
= {
684 .open
= mountinfo_open
,
687 .release
= mounts_release
,
691 static int mountstats_open(struct inode
*inode
, struct file
*file
)
693 return mounts_open_common(inode
, file
, &mountstats_op
);
696 static const struct file_operations proc_mountstats_operations
= {
697 .open
= mountstats_open
,
700 .release
= mounts_release
,
703 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
705 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
706 size_t count
, loff_t
*ppos
)
708 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
711 struct task_struct
*task
= get_proc_task(inode
);
717 if (count
> PROC_BLOCK_SIZE
)
718 count
= PROC_BLOCK_SIZE
;
721 if (!(page
= __get_free_page(GFP_TEMPORARY
)))
724 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
727 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
730 put_task_struct(task
);
735 static const struct file_operations proc_info_file_operations
= {
736 .read
= proc_info_read
,
739 static int proc_single_show(struct seq_file
*m
, void *v
)
741 struct inode
*inode
= m
->private;
742 struct pid_namespace
*ns
;
744 struct task_struct
*task
;
747 ns
= inode
->i_sb
->s_fs_info
;
748 pid
= proc_pid(inode
);
749 task
= get_pid_task(pid
, PIDTYPE_PID
);
753 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
755 put_task_struct(task
);
759 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
762 ret
= single_open(filp
, proc_single_show
, NULL
);
764 struct seq_file
*m
= filp
->private_data
;
771 static const struct file_operations proc_single_file_operations
= {
772 .open
= proc_single_open
,
775 .release
= single_release
,
778 static int mem_open(struct inode
* inode
, struct file
* file
)
780 file
->private_data
= (void*)((long)current
->self_exec_id
);
784 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
785 size_t count
, loff_t
*ppos
)
787 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
789 unsigned long src
= *ppos
;
791 struct mm_struct
*mm
;
796 if (check_mem_permission(task
))
800 page
= (char *)__get_free_page(GFP_TEMPORARY
);
806 mm
= get_task_mm(task
);
812 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
818 int this_len
, retval
;
820 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
821 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
822 if (!retval
|| check_mem_permission(task
)) {
828 if (copy_to_user(buf
, page
, retval
)) {
843 free_page((unsigned long) page
);
845 put_task_struct(task
);
850 #define mem_write NULL
853 /* This is a security hazard */
854 static ssize_t
mem_write(struct file
* file
, const char __user
*buf
,
855 size_t count
, loff_t
*ppos
)
859 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
860 unsigned long dst
= *ppos
;
866 if (check_mem_permission(task
))
870 page
= (char *)__get_free_page(GFP_TEMPORARY
);
876 int this_len
, retval
;
878 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
879 if (copy_from_user(page
, buf
, this_len
)) {
883 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
895 free_page((unsigned long) page
);
897 put_task_struct(task
);
903 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
907 file
->f_pos
= offset
;
910 file
->f_pos
+= offset
;
915 force_successful_syscall_return();
919 static const struct file_operations proc_mem_operations
= {
926 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
927 size_t count
, loff_t
*ppos
)
929 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
931 unsigned long src
= *ppos
;
933 struct mm_struct
*mm
;
938 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
942 page
= (char *)__get_free_page(GFP_TEMPORARY
);
948 mm
= get_task_mm(task
);
953 int this_len
, retval
, max_len
;
955 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
960 max_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
961 this_len
= (this_len
> max_len
) ? max_len
: this_len
;
963 retval
= access_process_vm(task
, (mm
->env_start
+ src
),
971 if (copy_to_user(buf
, page
, retval
)) {
985 free_page((unsigned long) page
);
987 put_task_struct(task
);
992 static const struct file_operations proc_environ_operations
= {
993 .read
= environ_read
,
996 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
997 size_t count
, loff_t
*ppos
)
999 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1000 char buffer
[PROC_NUMBUF
];
1002 int oom_adjust
= OOM_DISABLE
;
1003 unsigned long flags
;
1008 if (lock_task_sighand(task
, &flags
)) {
1009 oom_adjust
= task
->signal
->oom_adj
;
1010 unlock_task_sighand(task
, &flags
);
1013 put_task_struct(task
);
1015 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
1017 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1020 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
1021 size_t count
, loff_t
*ppos
)
1023 struct task_struct
*task
;
1024 char buffer
[PROC_NUMBUF
];
1026 unsigned long flags
;
1029 memset(buffer
, 0, sizeof(buffer
));
1030 if (count
> sizeof(buffer
) - 1)
1031 count
= sizeof(buffer
) - 1;
1032 if (copy_from_user(buffer
, buf
, count
))
1035 err
= strict_strtol(strstrip(buffer
), 0, &oom_adjust
);
1038 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
1039 oom_adjust
!= OOM_DISABLE
)
1042 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1045 if (!lock_task_sighand(task
, &flags
)) {
1046 put_task_struct(task
);
1050 if (oom_adjust
< task
->signal
->oom_adj
&& !capable(CAP_SYS_RESOURCE
)) {
1051 unlock_task_sighand(task
, &flags
);
1052 put_task_struct(task
);
1056 task
->signal
->oom_adj
= oom_adjust
;
1058 unlock_task_sighand(task
, &flags
);
1059 put_task_struct(task
);
1064 static const struct file_operations proc_oom_adjust_operations
= {
1065 .read
= oom_adjust_read
,
1066 .write
= oom_adjust_write
,
1069 #ifdef CONFIG_AUDITSYSCALL
1070 #define TMPBUFLEN 21
1071 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1072 size_t count
, loff_t
*ppos
)
1074 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1075 struct task_struct
*task
= get_proc_task(inode
);
1077 char tmpbuf
[TMPBUFLEN
];
1081 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1082 audit_get_loginuid(task
));
1083 put_task_struct(task
);
1084 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1087 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1088 size_t count
, loff_t
*ppos
)
1090 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1095 if (!capable(CAP_AUDIT_CONTROL
))
1099 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1105 if (count
>= PAGE_SIZE
)
1106 count
= PAGE_SIZE
- 1;
1109 /* No partial writes. */
1112 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1116 if (copy_from_user(page
, buf
, count
))
1120 loginuid
= simple_strtoul(page
, &tmp
, 10);
1126 length
= audit_set_loginuid(current
, loginuid
);
1127 if (likely(length
== 0))
1131 free_page((unsigned long) page
);
1135 static const struct file_operations proc_loginuid_operations
= {
1136 .read
= proc_loginuid_read
,
1137 .write
= proc_loginuid_write
,
1140 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1141 size_t count
, loff_t
*ppos
)
1143 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1144 struct task_struct
*task
= get_proc_task(inode
);
1146 char tmpbuf
[TMPBUFLEN
];
1150 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1151 audit_get_sessionid(task
));
1152 put_task_struct(task
);
1153 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1156 static const struct file_operations proc_sessionid_operations
= {
1157 .read
= proc_sessionid_read
,
1161 #ifdef CONFIG_FAULT_INJECTION
1162 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1163 size_t count
, loff_t
*ppos
)
1165 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1166 char buffer
[PROC_NUMBUF
];
1172 make_it_fail
= task
->make_it_fail
;
1173 put_task_struct(task
);
1175 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1177 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1180 static ssize_t
proc_fault_inject_write(struct file
* file
,
1181 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1183 struct task_struct
*task
;
1184 char buffer
[PROC_NUMBUF
], *end
;
1187 if (!capable(CAP_SYS_RESOURCE
))
1189 memset(buffer
, 0, sizeof(buffer
));
1190 if (count
> sizeof(buffer
) - 1)
1191 count
= sizeof(buffer
) - 1;
1192 if (copy_from_user(buffer
, buf
, count
))
1194 make_it_fail
= simple_strtol(strstrip(buffer
), &end
, 0);
1197 task
= get_proc_task(file
->f_dentry
->d_inode
);
1200 task
->make_it_fail
= make_it_fail
;
1201 put_task_struct(task
);
1206 static const struct file_operations proc_fault_inject_operations
= {
1207 .read
= proc_fault_inject_read
,
1208 .write
= proc_fault_inject_write
,
1213 #ifdef CONFIG_SCHED_DEBUG
1215 * Print out various scheduling related per-task fields:
1217 static int sched_show(struct seq_file
*m
, void *v
)
1219 struct inode
*inode
= m
->private;
1220 struct task_struct
*p
;
1222 p
= get_proc_task(inode
);
1225 proc_sched_show_task(p
, m
);
1233 sched_write(struct file
*file
, const char __user
*buf
,
1234 size_t count
, loff_t
*offset
)
1236 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1237 struct task_struct
*p
;
1239 p
= get_proc_task(inode
);
1242 proc_sched_set_task(p
);
1249 static int sched_open(struct inode
*inode
, struct file
*filp
)
1253 ret
= single_open(filp
, sched_show
, NULL
);
1255 struct seq_file
*m
= filp
->private_data
;
1262 static const struct file_operations proc_pid_sched_operations
= {
1265 .write
= sched_write
,
1266 .llseek
= seq_lseek
,
1267 .release
= single_release
,
1272 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1273 size_t count
, loff_t
*offset
)
1275 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1276 struct task_struct
*p
;
1277 char buffer
[TASK_COMM_LEN
];
1279 memset(buffer
, 0, sizeof(buffer
));
1280 if (count
> sizeof(buffer
) - 1)
1281 count
= sizeof(buffer
) - 1;
1282 if (copy_from_user(buffer
, buf
, count
))
1285 p
= get_proc_task(inode
);
1289 if (same_thread_group(current
, p
))
1290 set_task_comm(p
, buffer
);
1299 static int comm_show(struct seq_file
*m
, void *v
)
1301 struct inode
*inode
= m
->private;
1302 struct task_struct
*p
;
1304 p
= get_proc_task(inode
);
1309 seq_printf(m
, "%s\n", p
->comm
);
1317 static int comm_open(struct inode
*inode
, struct file
*filp
)
1321 ret
= single_open(filp
, comm_show
, NULL
);
1323 struct seq_file
*m
= filp
->private_data
;
1330 static const struct file_operations proc_pid_set_comm_operations
= {
1333 .write
= comm_write
,
1334 .llseek
= seq_lseek
,
1335 .release
= single_release
,
1339 * We added or removed a vma mapping the executable. The vmas are only mapped
1340 * during exec and are not mapped with the mmap system call.
1341 * Callers must hold down_write() on the mm's mmap_sem for these
1343 void added_exe_file_vma(struct mm_struct
*mm
)
1345 mm
->num_exe_file_vmas
++;
1348 void removed_exe_file_vma(struct mm_struct
*mm
)
1350 mm
->num_exe_file_vmas
--;
1351 if ((mm
->num_exe_file_vmas
== 0) && mm
->exe_file
){
1353 mm
->exe_file
= NULL
;
1358 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
1361 get_file(new_exe_file
);
1364 mm
->exe_file
= new_exe_file
;
1365 mm
->num_exe_file_vmas
= 0;
1368 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
1370 struct file
*exe_file
;
1372 /* We need mmap_sem to protect against races with removal of
1373 * VM_EXECUTABLE vmas */
1374 down_read(&mm
->mmap_sem
);
1375 exe_file
= mm
->exe_file
;
1378 up_read(&mm
->mmap_sem
);
1382 void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
1384 /* It's safe to write the exe_file pointer without exe_file_lock because
1385 * this is called during fork when the task is not yet in /proc */
1386 newmm
->exe_file
= get_mm_exe_file(oldmm
);
1389 static int proc_exe_link(struct inode
*inode
, struct path
*exe_path
)
1391 struct task_struct
*task
;
1392 struct mm_struct
*mm
;
1393 struct file
*exe_file
;
1395 task
= get_proc_task(inode
);
1398 mm
= get_task_mm(task
);
1399 put_task_struct(task
);
1402 exe_file
= get_mm_exe_file(mm
);
1405 *exe_path
= exe_file
->f_path
;
1406 path_get(&exe_file
->f_path
);
1413 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1415 struct inode
*inode
= dentry
->d_inode
;
1416 int error
= -EACCES
;
1418 /* We don't need a base pointer in the /proc filesystem */
1419 path_put(&nd
->path
);
1421 /* Are we allowed to snoop on the tasks file descriptors? */
1422 if (!proc_fd_access_allowed(inode
))
1425 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->path
);
1427 return ERR_PTR(error
);
1430 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1432 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1439 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1440 len
= PTR_ERR(pathname
);
1441 if (IS_ERR(pathname
))
1443 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1447 if (copy_to_user(buffer
, pathname
, len
))
1450 free_page((unsigned long)tmp
);
1454 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1456 int error
= -EACCES
;
1457 struct inode
*inode
= dentry
->d_inode
;
1460 /* Are we allowed to snoop on the tasks file descriptors? */
1461 if (!proc_fd_access_allowed(inode
))
1464 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &path
);
1468 error
= do_proc_readlink(&path
, buffer
, buflen
);
1474 static const struct inode_operations proc_pid_link_inode_operations
= {
1475 .readlink
= proc_pid_readlink
,
1476 .follow_link
= proc_pid_follow_link
,
1477 .setattr
= proc_setattr
,
1481 /* building an inode */
1483 static int task_dumpable(struct task_struct
*task
)
1486 struct mm_struct
*mm
;
1491 dumpable
= get_dumpable(mm
);
1499 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1501 struct inode
* inode
;
1502 struct proc_inode
*ei
;
1503 const struct cred
*cred
;
1505 /* We need a new inode */
1507 inode
= new_inode(sb
);
1513 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1514 inode
->i_op
= &proc_def_inode_operations
;
1517 * grab the reference to task.
1519 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1523 if (task_dumpable(task
)) {
1525 cred
= __task_cred(task
);
1526 inode
->i_uid
= cred
->euid
;
1527 inode
->i_gid
= cred
->egid
;
1530 security_task_to_inode(task
, inode
);
1540 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1542 struct inode
*inode
= dentry
->d_inode
;
1543 struct task_struct
*task
;
1544 const struct cred
*cred
;
1546 generic_fillattr(inode
, stat
);
1551 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1553 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1554 task_dumpable(task
)) {
1555 cred
= __task_cred(task
);
1556 stat
->uid
= cred
->euid
;
1557 stat
->gid
= cred
->egid
;
1567 * Exceptional case: normally we are not allowed to unhash a busy
1568 * directory. In this case, however, we can do it - no aliasing problems
1569 * due to the way we treat inodes.
1571 * Rewrite the inode's ownerships here because the owning task may have
1572 * performed a setuid(), etc.
1574 * Before the /proc/pid/status file was created the only way to read
1575 * the effective uid of a /process was to stat /proc/pid. Reading
1576 * /proc/pid/status is slow enough that procps and other packages
1577 * kept stating /proc/pid. To keep the rules in /proc simple I have
1578 * made this apply to all per process world readable and executable
1581 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1583 struct inode
*inode
= dentry
->d_inode
;
1584 struct task_struct
*task
= get_proc_task(inode
);
1585 const struct cred
*cred
;
1588 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1589 task_dumpable(task
)) {
1591 cred
= __task_cred(task
);
1592 inode
->i_uid
= cred
->euid
;
1593 inode
->i_gid
= cred
->egid
;
1599 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1600 security_task_to_inode(task
, inode
);
1601 put_task_struct(task
);
1608 static int pid_delete_dentry(struct dentry
* dentry
)
1610 /* Is the task we represent dead?
1611 * If so, then don't put the dentry on the lru list,
1612 * kill it immediately.
1614 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1617 static const struct dentry_operations pid_dentry_operations
=
1619 .d_revalidate
= pid_revalidate
,
1620 .d_delete
= pid_delete_dentry
,
1625 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1626 struct task_struct
*, const void *);
1629 * Fill a directory entry.
1631 * If possible create the dcache entry and derive our inode number and
1632 * file type from dcache entry.
1634 * Since all of the proc inode numbers are dynamically generated, the inode
1635 * numbers do not exist until the inode is cache. This means creating the
1636 * the dcache entry in readdir is necessary to keep the inode numbers
1637 * reported by readdir in sync with the inode numbers reported
1640 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1641 char *name
, int len
,
1642 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1644 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1645 struct inode
*inode
;
1648 unsigned type
= DT_UNKNOWN
;
1652 qname
.hash
= full_name_hash(name
, len
);
1654 child
= d_lookup(dir
, &qname
);
1657 new = d_alloc(dir
, &qname
);
1659 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1666 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1667 goto end_instantiate
;
1668 inode
= child
->d_inode
;
1671 type
= inode
->i_mode
>> 12;
1676 ino
= find_inode_number(dir
, &qname
);
1679 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1682 static unsigned name_to_int(struct dentry
*dentry
)
1684 const char *name
= dentry
->d_name
.name
;
1685 int len
= dentry
->d_name
.len
;
1688 if (len
> 1 && *name
== '0')
1691 unsigned c
= *name
++ - '0';
1694 if (n
>= (~0U-9)/10)
1704 #define PROC_FDINFO_MAX 64
1706 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1708 struct task_struct
*task
= get_proc_task(inode
);
1709 struct files_struct
*files
= NULL
;
1711 int fd
= proc_fd(inode
);
1714 files
= get_files_struct(task
);
1715 put_task_struct(task
);
1719 * We are not taking a ref to the file structure, so we must
1722 spin_lock(&files
->file_lock
);
1723 file
= fcheck_files(files
, fd
);
1726 *path
= file
->f_path
;
1727 path_get(&file
->f_path
);
1730 snprintf(info
, PROC_FDINFO_MAX
,
1733 (long long) file
->f_pos
,
1735 spin_unlock(&files
->file_lock
);
1736 put_files_struct(files
);
1739 spin_unlock(&files
->file_lock
);
1740 put_files_struct(files
);
1745 static int proc_fd_link(struct inode
*inode
, struct path
*path
)
1747 return proc_fd_info(inode
, path
, NULL
);
1750 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1752 struct inode
*inode
= dentry
->d_inode
;
1753 struct task_struct
*task
= get_proc_task(inode
);
1754 int fd
= proc_fd(inode
);
1755 struct files_struct
*files
;
1756 const struct cred
*cred
;
1759 files
= get_files_struct(task
);
1762 if (fcheck_files(files
, fd
)) {
1764 put_files_struct(files
);
1765 if (task_dumpable(task
)) {
1767 cred
= __task_cred(task
);
1768 inode
->i_uid
= cred
->euid
;
1769 inode
->i_gid
= cred
->egid
;
1775 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1776 security_task_to_inode(task
, inode
);
1777 put_task_struct(task
);
1781 put_files_struct(files
);
1783 put_task_struct(task
);
1789 static const struct dentry_operations tid_fd_dentry_operations
=
1791 .d_revalidate
= tid_fd_revalidate
,
1792 .d_delete
= pid_delete_dentry
,
1795 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1796 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1798 unsigned fd
= *(const unsigned *)ptr
;
1800 struct files_struct
*files
;
1801 struct inode
*inode
;
1802 struct proc_inode
*ei
;
1803 struct dentry
*error
= ERR_PTR(-ENOENT
);
1805 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1810 files
= get_files_struct(task
);
1813 inode
->i_mode
= S_IFLNK
;
1816 * We are not taking a ref to the file structure, so we must
1819 spin_lock(&files
->file_lock
);
1820 file
= fcheck_files(files
, fd
);
1823 if (file
->f_mode
& FMODE_READ
)
1824 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1825 if (file
->f_mode
& FMODE_WRITE
)
1826 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1827 spin_unlock(&files
->file_lock
);
1828 put_files_struct(files
);
1830 inode
->i_op
= &proc_pid_link_inode_operations
;
1832 ei
->op
.proc_get_link
= proc_fd_link
;
1833 dentry
->d_op
= &tid_fd_dentry_operations
;
1834 d_add(dentry
, inode
);
1835 /* Close the race of the process dying before we return the dentry */
1836 if (tid_fd_revalidate(dentry
, NULL
))
1842 spin_unlock(&files
->file_lock
);
1843 put_files_struct(files
);
1849 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1850 struct dentry
*dentry
,
1851 instantiate_t instantiate
)
1853 struct task_struct
*task
= get_proc_task(dir
);
1854 unsigned fd
= name_to_int(dentry
);
1855 struct dentry
*result
= ERR_PTR(-ENOENT
);
1862 result
= instantiate(dir
, dentry
, task
, &fd
);
1864 put_task_struct(task
);
1869 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1870 filldir_t filldir
, instantiate_t instantiate
)
1872 struct dentry
*dentry
= filp
->f_path
.dentry
;
1873 struct inode
*inode
= dentry
->d_inode
;
1874 struct task_struct
*p
= get_proc_task(inode
);
1875 unsigned int fd
, ino
;
1877 struct files_struct
* files
;
1887 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1891 ino
= parent_ino(dentry
);
1892 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1896 files
= get_files_struct(p
);
1900 for (fd
= filp
->f_pos
-2;
1901 fd
< files_fdtable(files
)->max_fds
;
1902 fd
++, filp
->f_pos
++) {
1903 char name
[PROC_NUMBUF
];
1906 if (!fcheck_files(files
, fd
))
1910 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1911 if (proc_fill_cache(filp
, dirent
, filldir
,
1912 name
, len
, instantiate
,
1920 put_files_struct(files
);
1928 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1929 struct nameidata
*nd
)
1931 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1934 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1936 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1939 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1940 size_t len
, loff_t
*ppos
)
1942 char tmp
[PROC_FDINFO_MAX
];
1943 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1945 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1949 static const struct file_operations proc_fdinfo_file_operations
= {
1950 .open
= nonseekable_open
,
1951 .read
= proc_fdinfo_read
,
1954 static const struct file_operations proc_fd_operations
= {
1955 .read
= generic_read_dir
,
1956 .readdir
= proc_readfd
,
1960 * /proc/pid/fd needs a special permission handler so that a process can still
1961 * access /proc/self/fd after it has executed a setuid().
1963 static int proc_fd_permission(struct inode
*inode
, int mask
)
1967 rv
= generic_permission(inode
, mask
, NULL
);
1970 if (task_pid(current
) == proc_pid(inode
))
1976 * proc directories can do almost nothing..
1978 static const struct inode_operations proc_fd_inode_operations
= {
1979 .lookup
= proc_lookupfd
,
1980 .permission
= proc_fd_permission
,
1981 .setattr
= proc_setattr
,
1984 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
1985 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1987 unsigned fd
= *(unsigned *)ptr
;
1988 struct inode
*inode
;
1989 struct proc_inode
*ei
;
1990 struct dentry
*error
= ERR_PTR(-ENOENT
);
1992 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1997 inode
->i_mode
= S_IFREG
| S_IRUSR
;
1998 inode
->i_fop
= &proc_fdinfo_file_operations
;
1999 dentry
->d_op
= &tid_fd_dentry_operations
;
2000 d_add(dentry
, inode
);
2001 /* Close the race of the process dying before we return the dentry */
2002 if (tid_fd_revalidate(dentry
, NULL
))
2009 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
2010 struct dentry
*dentry
,
2011 struct nameidata
*nd
)
2013 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
2016 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
2018 return proc_readfd_common(filp
, dirent
, filldir
,
2019 proc_fdinfo_instantiate
);
2022 static const struct file_operations proc_fdinfo_operations
= {
2023 .read
= generic_read_dir
,
2024 .readdir
= proc_readfdinfo
,
2028 * proc directories can do almost nothing..
2030 static const struct inode_operations proc_fdinfo_inode_operations
= {
2031 .lookup
= proc_lookupfdinfo
,
2032 .setattr
= proc_setattr
,
2036 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
2037 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2039 const struct pid_entry
*p
= ptr
;
2040 struct inode
*inode
;
2041 struct proc_inode
*ei
;
2042 struct dentry
*error
= ERR_PTR(-ENOENT
);
2044 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2049 inode
->i_mode
= p
->mode
;
2050 if (S_ISDIR(inode
->i_mode
))
2051 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
2053 inode
->i_op
= p
->iop
;
2055 inode
->i_fop
= p
->fop
;
2057 dentry
->d_op
= &pid_dentry_operations
;
2058 d_add(dentry
, inode
);
2059 /* Close the race of the process dying before we return the dentry */
2060 if (pid_revalidate(dentry
, NULL
))
2066 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2067 struct dentry
*dentry
,
2068 const struct pid_entry
*ents
,
2071 struct dentry
*error
;
2072 struct task_struct
*task
= get_proc_task(dir
);
2073 const struct pid_entry
*p
, *last
;
2075 error
= ERR_PTR(-ENOENT
);
2081 * Yes, it does not scale. And it should not. Don't add
2082 * new entries into /proc/<tgid>/ without very good reasons.
2084 last
= &ents
[nents
- 1];
2085 for (p
= ents
; p
<= last
; p
++) {
2086 if (p
->len
!= dentry
->d_name
.len
)
2088 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2094 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2096 put_task_struct(task
);
2101 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2102 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2104 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2105 proc_pident_instantiate
, task
, p
);
2108 static int proc_pident_readdir(struct file
*filp
,
2109 void *dirent
, filldir_t filldir
,
2110 const struct pid_entry
*ents
, unsigned int nents
)
2113 struct dentry
*dentry
= filp
->f_path
.dentry
;
2114 struct inode
*inode
= dentry
->d_inode
;
2115 struct task_struct
*task
= get_proc_task(inode
);
2116 const struct pid_entry
*p
, *last
;
2129 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2135 ino
= parent_ino(dentry
);
2136 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2148 last
= &ents
[nents
- 1];
2150 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2159 put_task_struct(task
);
2164 #ifdef CONFIG_SECURITY
2165 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2166 size_t count
, loff_t
*ppos
)
2168 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2171 struct task_struct
*task
= get_proc_task(inode
);
2176 length
= security_getprocattr(task
,
2177 (char*)file
->f_path
.dentry
->d_name
.name
,
2179 put_task_struct(task
);
2181 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2186 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2187 size_t count
, loff_t
*ppos
)
2189 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2192 struct task_struct
*task
= get_proc_task(inode
);
2197 if (count
> PAGE_SIZE
)
2200 /* No partial writes. */
2206 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2211 if (copy_from_user(page
, buf
, count
))
2214 /* Guard against adverse ptrace interaction */
2215 length
= mutex_lock_interruptible(&task
->cred_guard_mutex
);
2219 length
= security_setprocattr(task
,
2220 (char*)file
->f_path
.dentry
->d_name
.name
,
2221 (void*)page
, count
);
2222 mutex_unlock(&task
->cred_guard_mutex
);
2224 free_page((unsigned long) page
);
2226 put_task_struct(task
);
2231 static const struct file_operations proc_pid_attr_operations
= {
2232 .read
= proc_pid_attr_read
,
2233 .write
= proc_pid_attr_write
,
2236 static const struct pid_entry attr_dir_stuff
[] = {
2237 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2238 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2239 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2240 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2241 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2242 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2245 static int proc_attr_dir_readdir(struct file
* filp
,
2246 void * dirent
, filldir_t filldir
)
2248 return proc_pident_readdir(filp
,dirent
,filldir
,
2249 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2252 static const struct file_operations proc_attr_dir_operations
= {
2253 .read
= generic_read_dir
,
2254 .readdir
= proc_attr_dir_readdir
,
2257 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2258 struct dentry
*dentry
, struct nameidata
*nd
)
2260 return proc_pident_lookup(dir
, dentry
,
2261 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2264 static const struct inode_operations proc_attr_dir_inode_operations
= {
2265 .lookup
= proc_attr_dir_lookup
,
2266 .getattr
= pid_getattr
,
2267 .setattr
= proc_setattr
,
2272 #ifdef CONFIG_ELF_CORE
2273 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2274 size_t count
, loff_t
*ppos
)
2276 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2277 struct mm_struct
*mm
;
2278 char buffer
[PROC_NUMBUF
];
2286 mm
= get_task_mm(task
);
2288 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2289 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2290 MMF_DUMP_FILTER_SHIFT
));
2292 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2295 put_task_struct(task
);
2300 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2301 const char __user
*buf
,
2305 struct task_struct
*task
;
2306 struct mm_struct
*mm
;
2307 char buffer
[PROC_NUMBUF
], *end
;
2314 memset(buffer
, 0, sizeof(buffer
));
2315 if (count
> sizeof(buffer
) - 1)
2316 count
= sizeof(buffer
) - 1;
2317 if (copy_from_user(buffer
, buf
, count
))
2321 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2324 if (end
- buffer
== 0)
2328 task
= get_proc_task(file
->f_dentry
->d_inode
);
2333 mm
= get_task_mm(task
);
2337 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2339 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2341 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2346 put_task_struct(task
);
2351 static const struct file_operations proc_coredump_filter_operations
= {
2352 .read
= proc_coredump_filter_read
,
2353 .write
= proc_coredump_filter_write
,
2360 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2363 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2364 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2365 char tmp
[PROC_NUMBUF
];
2368 sprintf(tmp
, "%d", tgid
);
2369 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2372 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2374 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2375 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2376 char *name
= ERR_PTR(-ENOENT
);
2380 name
= ERR_PTR(-ENOMEM
);
2382 sprintf(name
, "%d", tgid
);
2384 nd_set_link(nd
, name
);
2388 static void proc_self_put_link(struct dentry
*dentry
, struct nameidata
*nd
,
2391 char *s
= nd_get_link(nd
);
2396 static const struct inode_operations proc_self_inode_operations
= {
2397 .readlink
= proc_self_readlink
,
2398 .follow_link
= proc_self_follow_link
,
2399 .put_link
= proc_self_put_link
,
2405 * These are the directory entries in the root directory of /proc
2406 * that properly belong to the /proc filesystem, as they describe
2407 * describe something that is process related.
2409 static const struct pid_entry proc_base_stuff
[] = {
2410 NOD("self", S_IFLNK
|S_IRWXUGO
,
2411 &proc_self_inode_operations
, NULL
, {}),
2415 * Exceptional case: normally we are not allowed to unhash a busy
2416 * directory. In this case, however, we can do it - no aliasing problems
2417 * due to the way we treat inodes.
2419 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
2421 struct inode
*inode
= dentry
->d_inode
;
2422 struct task_struct
*task
= get_proc_task(inode
);
2424 put_task_struct(task
);
2431 static const struct dentry_operations proc_base_dentry_operations
=
2433 .d_revalidate
= proc_base_revalidate
,
2434 .d_delete
= pid_delete_dentry
,
2437 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2438 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2440 const struct pid_entry
*p
= ptr
;
2441 struct inode
*inode
;
2442 struct proc_inode
*ei
;
2443 struct dentry
*error
= ERR_PTR(-EINVAL
);
2445 /* Allocate the inode */
2446 error
= ERR_PTR(-ENOMEM
);
2447 inode
= new_inode(dir
->i_sb
);
2451 /* Initialize the inode */
2453 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2456 * grab the reference to the task.
2458 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2462 inode
->i_mode
= p
->mode
;
2463 if (S_ISDIR(inode
->i_mode
))
2465 if (S_ISLNK(inode
->i_mode
))
2468 inode
->i_op
= p
->iop
;
2470 inode
->i_fop
= p
->fop
;
2472 dentry
->d_op
= &proc_base_dentry_operations
;
2473 d_add(dentry
, inode
);
2482 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2484 struct dentry
*error
;
2485 struct task_struct
*task
= get_proc_task(dir
);
2486 const struct pid_entry
*p
, *last
;
2488 error
= ERR_PTR(-ENOENT
);
2493 /* Lookup the directory entry */
2494 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2495 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2496 if (p
->len
!= dentry
->d_name
.len
)
2498 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2504 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2507 put_task_struct(task
);
2512 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2513 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2515 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2516 proc_base_instantiate
, task
, p
);
2519 #ifdef CONFIG_TASK_IO_ACCOUNTING
2520 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2522 struct task_io_accounting acct
= task
->ioac
;
2523 unsigned long flags
;
2525 if (whole
&& lock_task_sighand(task
, &flags
)) {
2526 struct task_struct
*t
= task
;
2528 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2529 while_each_thread(task
, t
)
2530 task_io_accounting_add(&acct
, &t
->ioac
);
2532 unlock_task_sighand(task
, &flags
);
2534 return sprintf(buffer
,
2539 "read_bytes: %llu\n"
2540 "write_bytes: %llu\n"
2541 "cancelled_write_bytes: %llu\n",
2542 (unsigned long long)acct
.rchar
,
2543 (unsigned long long)acct
.wchar
,
2544 (unsigned long long)acct
.syscr
,
2545 (unsigned long long)acct
.syscw
,
2546 (unsigned long long)acct
.read_bytes
,
2547 (unsigned long long)acct
.write_bytes
,
2548 (unsigned long long)acct
.cancelled_write_bytes
);
2551 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2553 return do_io_accounting(task
, buffer
, 0);
2556 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2558 return do_io_accounting(task
, buffer
, 1);
2560 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2562 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2563 struct pid
*pid
, struct task_struct
*task
)
2565 seq_printf(m
, "%08x\n", task
->personality
);
2572 static const struct file_operations proc_task_operations
;
2573 static const struct inode_operations proc_task_inode_operations
;
2575 static const struct pid_entry tgid_base_stuff
[] = {
2576 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2577 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2578 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2580 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2582 REG("environ", S_IRUSR
, proc_environ_operations
),
2583 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2584 ONE("status", S_IRUGO
, proc_pid_status
),
2585 ONE("personality", S_IRUSR
, proc_pid_personality
),
2586 INF("limits", S_IRUSR
, proc_pid_limits
),
2587 #ifdef CONFIG_SCHED_DEBUG
2588 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2590 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2591 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2592 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2594 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2595 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2596 ONE("statm", S_IRUGO
, proc_pid_statm
),
2597 REG("maps", S_IRUGO
, proc_maps_operations
),
2599 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2601 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2602 LNK("cwd", proc_cwd_link
),
2603 LNK("root", proc_root_link
),
2604 LNK("exe", proc_exe_link
),
2605 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2606 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2607 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2608 #ifdef CONFIG_PROC_PAGE_MONITOR
2609 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2610 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2611 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2613 #ifdef CONFIG_SECURITY
2614 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2616 #ifdef CONFIG_KALLSYMS
2617 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2619 #ifdef CONFIG_STACKTRACE
2620 ONE("stack", S_IRUSR
, proc_pid_stack
),
2622 #ifdef CONFIG_SCHEDSTATS
2623 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2625 #ifdef CONFIG_LATENCYTOP
2626 REG("latency", S_IRUGO
, proc_lstats_operations
),
2628 #ifdef CONFIG_PROC_PID_CPUSET
2629 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2631 #ifdef CONFIG_CGROUPS
2632 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2634 INF("oom_score", S_IRUGO
, proc_oom_score
),
2635 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2636 #ifdef CONFIG_AUDITSYSCALL
2637 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2638 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2640 #ifdef CONFIG_FAULT_INJECTION
2641 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2643 #ifdef CONFIG_ELF_CORE
2644 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2646 #ifdef CONFIG_TASK_IO_ACCOUNTING
2647 INF("io", S_IRUGO
, proc_tgid_io_accounting
),
2651 static int proc_tgid_base_readdir(struct file
* filp
,
2652 void * dirent
, filldir_t filldir
)
2654 return proc_pident_readdir(filp
,dirent
,filldir
,
2655 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2658 static const struct file_operations proc_tgid_base_operations
= {
2659 .read
= generic_read_dir
,
2660 .readdir
= proc_tgid_base_readdir
,
2663 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2664 return proc_pident_lookup(dir
, dentry
,
2665 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2668 static const struct inode_operations proc_tgid_base_inode_operations
= {
2669 .lookup
= proc_tgid_base_lookup
,
2670 .getattr
= pid_getattr
,
2671 .setattr
= proc_setattr
,
2674 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2676 struct dentry
*dentry
, *leader
, *dir
;
2677 char buf
[PROC_NUMBUF
];
2681 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2682 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2684 shrink_dcache_parent(dentry
);
2690 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2691 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2696 name
.len
= strlen(name
.name
);
2697 dir
= d_hash_and_lookup(leader
, &name
);
2699 goto out_put_leader
;
2702 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2703 dentry
= d_hash_and_lookup(dir
, &name
);
2705 shrink_dcache_parent(dentry
);
2718 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2719 * @task: task that should be flushed.
2721 * When flushing dentries from proc, one needs to flush them from global
2722 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2723 * in. This call is supposed to do all of this job.
2725 * Looks in the dcache for
2727 * /proc/@tgid/task/@pid
2728 * if either directory is present flushes it and all of it'ts children
2731 * It is safe and reasonable to cache /proc entries for a task until
2732 * that task exits. After that they just clog up the dcache with
2733 * useless entries, possibly causing useful dcache entries to be
2734 * flushed instead. This routine is proved to flush those useless
2735 * dcache entries at process exit time.
2737 * NOTE: This routine is just an optimization so it does not guarantee
2738 * that no dcache entries will exist at process exit time it
2739 * just makes it very unlikely that any will persist.
2742 void proc_flush_task(struct task_struct
*task
)
2745 struct pid
*pid
, *tgid
;
2748 pid
= task_pid(task
);
2749 tgid
= task_tgid(task
);
2751 for (i
= 0; i
<= pid
->level
; i
++) {
2752 upid
= &pid
->numbers
[i
];
2753 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2754 tgid
->numbers
[i
].nr
);
2757 upid
= &pid
->numbers
[pid
->level
];
2759 pid_ns_release_proc(upid
->ns
);
2762 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2763 struct dentry
* dentry
,
2764 struct task_struct
*task
, const void *ptr
)
2766 struct dentry
*error
= ERR_PTR(-ENOENT
);
2767 struct inode
*inode
;
2769 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2773 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2774 inode
->i_op
= &proc_tgid_base_inode_operations
;
2775 inode
->i_fop
= &proc_tgid_base_operations
;
2776 inode
->i_flags
|=S_IMMUTABLE
;
2778 inode
->i_nlink
= 2 + pid_entry_count_dirs(tgid_base_stuff
,
2779 ARRAY_SIZE(tgid_base_stuff
));
2781 dentry
->d_op
= &pid_dentry_operations
;
2783 d_add(dentry
, inode
);
2784 /* Close the race of the process dying before we return the dentry */
2785 if (pid_revalidate(dentry
, NULL
))
2791 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2793 struct dentry
*result
= ERR_PTR(-ENOENT
);
2794 struct task_struct
*task
;
2796 struct pid_namespace
*ns
;
2798 result
= proc_base_lookup(dir
, dentry
);
2799 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2802 tgid
= name_to_int(dentry
);
2806 ns
= dentry
->d_sb
->s_fs_info
;
2808 task
= find_task_by_pid_ns(tgid
, ns
);
2810 get_task_struct(task
);
2815 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2816 put_task_struct(task
);
2822 * Find the first task with tgid >= tgid
2827 struct task_struct
*task
;
2829 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
2834 put_task_struct(iter
.task
);
2838 pid
= find_ge_pid(iter
.tgid
, ns
);
2840 iter
.tgid
= pid_nr_ns(pid
, ns
);
2841 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
2842 /* What we to know is if the pid we have find is the
2843 * pid of a thread_group_leader. Testing for task
2844 * being a thread_group_leader is the obvious thing
2845 * todo but there is a window when it fails, due to
2846 * the pid transfer logic in de_thread.
2848 * So we perform the straight forward test of seeing
2849 * if the pid we have found is the pid of a thread
2850 * group leader, and don't worry if the task we have
2851 * found doesn't happen to be a thread group leader.
2852 * As we don't care in the case of readdir.
2854 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
2858 get_task_struct(iter
.task
);
2864 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2866 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2867 struct tgid_iter iter
)
2869 char name
[PROC_NUMBUF
];
2870 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
2871 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2872 proc_pid_instantiate
, iter
.task
, NULL
);
2875 /* for the /proc/ directory itself, after non-process stuff has been done */
2876 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2878 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2879 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2880 struct tgid_iter iter
;
2881 struct pid_namespace
*ns
;
2886 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2887 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2888 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2892 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2894 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
2895 for (iter
= next_tgid(ns
, iter
);
2897 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
2898 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
2899 if (proc_pid_fill_cache(filp
, dirent
, filldir
, iter
) < 0) {
2900 put_task_struct(iter
.task
);
2904 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2906 put_task_struct(reaper
);
2914 static const struct pid_entry tid_base_stuff
[] = {
2915 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2916 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fd_operations
),
2917 REG("environ", S_IRUSR
, proc_environ_operations
),
2918 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2919 ONE("status", S_IRUGO
, proc_pid_status
),
2920 ONE("personality", S_IRUSR
, proc_pid_personality
),
2921 INF("limits", S_IRUSR
, proc_pid_limits
),
2922 #ifdef CONFIG_SCHED_DEBUG
2923 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2925 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2926 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2927 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2929 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2930 ONE("stat", S_IRUGO
, proc_tid_stat
),
2931 ONE("statm", S_IRUGO
, proc_pid_statm
),
2932 REG("maps", S_IRUGO
, proc_maps_operations
),
2934 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2936 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2937 LNK("cwd", proc_cwd_link
),
2938 LNK("root", proc_root_link
),
2939 LNK("exe", proc_exe_link
),
2940 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2941 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2942 #ifdef CONFIG_PROC_PAGE_MONITOR
2943 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2944 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2945 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2947 #ifdef CONFIG_SECURITY
2948 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2950 #ifdef CONFIG_KALLSYMS
2951 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2953 #ifdef CONFIG_STACKTRACE
2954 ONE("stack", S_IRUSR
, proc_pid_stack
),
2956 #ifdef CONFIG_SCHEDSTATS
2957 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2959 #ifdef CONFIG_LATENCYTOP
2960 REG("latency", S_IRUGO
, proc_lstats_operations
),
2962 #ifdef CONFIG_PROC_PID_CPUSET
2963 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2965 #ifdef CONFIG_CGROUPS
2966 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2968 INF("oom_score", S_IRUGO
, proc_oom_score
),
2969 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2970 #ifdef CONFIG_AUDITSYSCALL
2971 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2972 REG("sessionid", S_IRUSR
, proc_sessionid_operations
),
2974 #ifdef CONFIG_FAULT_INJECTION
2975 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2977 #ifdef CONFIG_TASK_IO_ACCOUNTING
2978 INF("io", S_IRUGO
, proc_tid_io_accounting
),
2982 static int proc_tid_base_readdir(struct file
* filp
,
2983 void * dirent
, filldir_t filldir
)
2985 return proc_pident_readdir(filp
,dirent
,filldir
,
2986 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2989 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2990 return proc_pident_lookup(dir
, dentry
,
2991 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
2994 static const struct file_operations proc_tid_base_operations
= {
2995 .read
= generic_read_dir
,
2996 .readdir
= proc_tid_base_readdir
,
2999 static const struct inode_operations proc_tid_base_inode_operations
= {
3000 .lookup
= proc_tid_base_lookup
,
3001 .getattr
= pid_getattr
,
3002 .setattr
= proc_setattr
,
3005 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
3006 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3008 struct dentry
*error
= ERR_PTR(-ENOENT
);
3009 struct inode
*inode
;
3010 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3014 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3015 inode
->i_op
= &proc_tid_base_inode_operations
;
3016 inode
->i_fop
= &proc_tid_base_operations
;
3017 inode
->i_flags
|=S_IMMUTABLE
;
3019 inode
->i_nlink
= 2 + pid_entry_count_dirs(tid_base_stuff
,
3020 ARRAY_SIZE(tid_base_stuff
));
3022 dentry
->d_op
= &pid_dentry_operations
;
3024 d_add(dentry
, inode
);
3025 /* Close the race of the process dying before we return the dentry */
3026 if (pid_revalidate(dentry
, NULL
))
3032 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
3034 struct dentry
*result
= ERR_PTR(-ENOENT
);
3035 struct task_struct
*task
;
3036 struct task_struct
*leader
= get_proc_task(dir
);
3038 struct pid_namespace
*ns
;
3043 tid
= name_to_int(dentry
);
3047 ns
= dentry
->d_sb
->s_fs_info
;
3049 task
= find_task_by_pid_ns(tid
, ns
);
3051 get_task_struct(task
);
3055 if (!same_thread_group(leader
, task
))
3058 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3060 put_task_struct(task
);
3062 put_task_struct(leader
);
3068 * Find the first tid of a thread group to return to user space.
3070 * Usually this is just the thread group leader, but if the users
3071 * buffer was too small or there was a seek into the middle of the
3072 * directory we have more work todo.
3074 * In the case of a short read we start with find_task_by_pid.
3076 * In the case of a seek we start with the leader and walk nr
3079 static struct task_struct
*first_tid(struct task_struct
*leader
,
3080 int tid
, int nr
, struct pid_namespace
*ns
)
3082 struct task_struct
*pos
;
3085 /* Attempt to start with the pid of a thread */
3086 if (tid
&& (nr
> 0)) {
3087 pos
= find_task_by_pid_ns(tid
, ns
);
3088 if (pos
&& (pos
->group_leader
== leader
))
3092 /* If nr exceeds the number of threads there is nothing todo */
3094 if (nr
&& nr
>= get_nr_threads(leader
))
3097 /* If we haven't found our starting place yet start
3098 * with the leader and walk nr threads forward.
3100 for (pos
= leader
; nr
> 0; --nr
) {
3101 pos
= next_thread(pos
);
3102 if (pos
== leader
) {
3108 get_task_struct(pos
);
3115 * Find the next thread in the thread list.
3116 * Return NULL if there is an error or no next thread.
3118 * The reference to the input task_struct is released.
3120 static struct task_struct
*next_tid(struct task_struct
*start
)
3122 struct task_struct
*pos
= NULL
;
3124 if (pid_alive(start
)) {
3125 pos
= next_thread(start
);
3126 if (thread_group_leader(pos
))
3129 get_task_struct(pos
);
3132 put_task_struct(start
);
3136 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3137 struct task_struct
*task
, int tid
)
3139 char name
[PROC_NUMBUF
];
3140 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3141 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3142 proc_task_instantiate
, task
, NULL
);
3145 /* for the /proc/TGID/task/ directories */
3146 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3148 struct dentry
*dentry
= filp
->f_path
.dentry
;
3149 struct inode
*inode
= dentry
->d_inode
;
3150 struct task_struct
*leader
= NULL
;
3151 struct task_struct
*task
;
3152 int retval
= -ENOENT
;
3155 struct pid_namespace
*ns
;
3157 task
= get_proc_task(inode
);
3161 if (pid_alive(task
)) {
3162 leader
= task
->group_leader
;
3163 get_task_struct(leader
);
3166 put_task_struct(task
);
3171 switch ((unsigned long)filp
->f_pos
) {
3174 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3179 ino
= parent_ino(dentry
);
3180 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3186 /* f_version caches the tgid value that the last readdir call couldn't
3187 * return. lseek aka telldir automagically resets f_version to 0.
3189 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3190 tid
= (int)filp
->f_version
;
3191 filp
->f_version
= 0;
3192 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3194 task
= next_tid(task
), filp
->f_pos
++) {
3195 tid
= task_pid_nr_ns(task
, ns
);
3196 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3197 /* returning this tgid failed, save it as the first
3198 * pid for the next readir call */
3199 filp
->f_version
= (u64
)tid
;
3200 put_task_struct(task
);
3205 put_task_struct(leader
);
3210 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3212 struct inode
*inode
= dentry
->d_inode
;
3213 struct task_struct
*p
= get_proc_task(inode
);
3214 generic_fillattr(inode
, stat
);
3217 stat
->nlink
+= get_nr_threads(p
);
3224 static const struct inode_operations proc_task_inode_operations
= {
3225 .lookup
= proc_task_lookup
,
3226 .getattr
= proc_task_getattr
,
3227 .setattr
= proc_setattr
,
3230 static const struct file_operations proc_task_operations
= {
3231 .read
= generic_read_dir
,
3232 .readdir
= proc_task_readdir
,