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
, 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", "ms"},
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
];
1008 oom_adjust
= task
->mm
->oom_adj
;
1010 oom_adjust
= OOM_DISABLE
;
1012 put_task_struct(task
);
1014 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
1016 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1019 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
1020 size_t count
, loff_t
*ppos
)
1022 struct task_struct
*task
;
1023 char buffer
[PROC_NUMBUF
], *end
;
1026 memset(buffer
, 0, sizeof(buffer
));
1027 if (count
> sizeof(buffer
) - 1)
1028 count
= sizeof(buffer
) - 1;
1029 if (copy_from_user(buffer
, buf
, count
))
1031 oom_adjust
= simple_strtol(buffer
, &end
, 0);
1032 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
1033 oom_adjust
!= OOM_DISABLE
)
1037 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1043 put_task_struct(task
);
1046 if (oom_adjust
< task
->mm
->oom_adj
&& !capable(CAP_SYS_RESOURCE
)) {
1048 put_task_struct(task
);
1051 task
->mm
->oom_adj
= oom_adjust
;
1053 put_task_struct(task
);
1054 if (end
- buffer
== 0)
1056 return end
- buffer
;
1059 static const struct file_operations proc_oom_adjust_operations
= {
1060 .read
= oom_adjust_read
,
1061 .write
= oom_adjust_write
,
1064 #ifdef CONFIG_AUDITSYSCALL
1065 #define TMPBUFLEN 21
1066 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1067 size_t count
, loff_t
*ppos
)
1069 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1070 struct task_struct
*task
= get_proc_task(inode
);
1072 char tmpbuf
[TMPBUFLEN
];
1076 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1077 audit_get_loginuid(task
));
1078 put_task_struct(task
);
1079 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1082 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1083 size_t count
, loff_t
*ppos
)
1085 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1090 if (!capable(CAP_AUDIT_CONTROL
))
1093 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
1096 if (count
>= PAGE_SIZE
)
1097 count
= PAGE_SIZE
- 1;
1100 /* No partial writes. */
1103 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1107 if (copy_from_user(page
, buf
, count
))
1111 loginuid
= simple_strtoul(page
, &tmp
, 10);
1117 length
= audit_set_loginuid(current
, loginuid
);
1118 if (likely(length
== 0))
1122 free_page((unsigned long) page
);
1126 static const struct file_operations proc_loginuid_operations
= {
1127 .read
= proc_loginuid_read
,
1128 .write
= proc_loginuid_write
,
1131 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1132 size_t count
, loff_t
*ppos
)
1134 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1135 struct task_struct
*task
= get_proc_task(inode
);
1137 char tmpbuf
[TMPBUFLEN
];
1141 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1142 audit_get_sessionid(task
));
1143 put_task_struct(task
);
1144 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1147 static const struct file_operations proc_sessionid_operations
= {
1148 .read
= proc_sessionid_read
,
1152 #ifdef CONFIG_FAULT_INJECTION
1153 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1154 size_t count
, loff_t
*ppos
)
1156 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1157 char buffer
[PROC_NUMBUF
];
1163 make_it_fail
= task
->make_it_fail
;
1164 put_task_struct(task
);
1166 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1168 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1171 static ssize_t
proc_fault_inject_write(struct file
* file
,
1172 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1174 struct task_struct
*task
;
1175 char buffer
[PROC_NUMBUF
], *end
;
1178 if (!capable(CAP_SYS_RESOURCE
))
1180 memset(buffer
, 0, sizeof(buffer
));
1181 if (count
> sizeof(buffer
) - 1)
1182 count
= sizeof(buffer
) - 1;
1183 if (copy_from_user(buffer
, buf
, count
))
1185 make_it_fail
= simple_strtol(buffer
, &end
, 0);
1188 task
= get_proc_task(file
->f_dentry
->d_inode
);
1191 task
->make_it_fail
= make_it_fail
;
1192 put_task_struct(task
);
1193 if (end
- buffer
== 0)
1195 return end
- buffer
;
1198 static const struct file_operations proc_fault_inject_operations
= {
1199 .read
= proc_fault_inject_read
,
1200 .write
= proc_fault_inject_write
,
1205 #ifdef CONFIG_SCHED_DEBUG
1207 * Print out various scheduling related per-task fields:
1209 static int sched_show(struct seq_file
*m
, void *v
)
1211 struct inode
*inode
= m
->private;
1212 struct task_struct
*p
;
1214 p
= get_proc_task(inode
);
1217 proc_sched_show_task(p
, m
);
1225 sched_write(struct file
*file
, const char __user
*buf
,
1226 size_t count
, loff_t
*offset
)
1228 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1229 struct task_struct
*p
;
1231 p
= get_proc_task(inode
);
1234 proc_sched_set_task(p
);
1241 static int sched_open(struct inode
*inode
, struct file
*filp
)
1245 ret
= single_open(filp
, sched_show
, NULL
);
1247 struct seq_file
*m
= filp
->private_data
;
1254 static const struct file_operations proc_pid_sched_operations
= {
1257 .write
= sched_write
,
1258 .llseek
= seq_lseek
,
1259 .release
= single_release
,
1265 * We added or removed a vma mapping the executable. The vmas are only mapped
1266 * during exec and are not mapped with the mmap system call.
1267 * Callers must hold down_write() on the mm's mmap_sem for these
1269 void added_exe_file_vma(struct mm_struct
*mm
)
1271 mm
->num_exe_file_vmas
++;
1274 void removed_exe_file_vma(struct mm_struct
*mm
)
1276 mm
->num_exe_file_vmas
--;
1277 if ((mm
->num_exe_file_vmas
== 0) && mm
->exe_file
){
1279 mm
->exe_file
= NULL
;
1284 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
1287 get_file(new_exe_file
);
1290 mm
->exe_file
= new_exe_file
;
1291 mm
->num_exe_file_vmas
= 0;
1294 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
1296 struct file
*exe_file
;
1298 /* We need mmap_sem to protect against races with removal of
1299 * VM_EXECUTABLE vmas */
1300 down_read(&mm
->mmap_sem
);
1301 exe_file
= mm
->exe_file
;
1304 up_read(&mm
->mmap_sem
);
1308 void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
1310 /* It's safe to write the exe_file pointer without exe_file_lock because
1311 * this is called during fork when the task is not yet in /proc */
1312 newmm
->exe_file
= get_mm_exe_file(oldmm
);
1315 static int proc_exe_link(struct inode
*inode
, struct path
*exe_path
)
1317 struct task_struct
*task
;
1318 struct mm_struct
*mm
;
1319 struct file
*exe_file
;
1321 task
= get_proc_task(inode
);
1324 mm
= get_task_mm(task
);
1325 put_task_struct(task
);
1328 exe_file
= get_mm_exe_file(mm
);
1331 *exe_path
= exe_file
->f_path
;
1332 path_get(&exe_file
->f_path
);
1339 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1341 struct inode
*inode
= dentry
->d_inode
;
1342 int error
= -EACCES
;
1344 /* We don't need a base pointer in the /proc filesystem */
1345 path_put(&nd
->path
);
1347 /* Are we allowed to snoop on the tasks file descriptors? */
1348 if (!proc_fd_access_allowed(inode
))
1351 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->path
);
1352 nd
->last_type
= LAST_BIND
;
1354 return ERR_PTR(error
);
1357 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1359 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1366 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1367 len
= PTR_ERR(pathname
);
1368 if (IS_ERR(pathname
))
1370 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1374 if (copy_to_user(buffer
, pathname
, len
))
1377 free_page((unsigned long)tmp
);
1381 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1383 int error
= -EACCES
;
1384 struct inode
*inode
= dentry
->d_inode
;
1387 /* Are we allowed to snoop on the tasks file descriptors? */
1388 if (!proc_fd_access_allowed(inode
))
1391 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &path
);
1395 error
= do_proc_readlink(&path
, buffer
, buflen
);
1401 static const struct inode_operations proc_pid_link_inode_operations
= {
1402 .readlink
= proc_pid_readlink
,
1403 .follow_link
= proc_pid_follow_link
,
1404 .setattr
= proc_setattr
,
1408 /* building an inode */
1410 static int task_dumpable(struct task_struct
*task
)
1413 struct mm_struct
*mm
;
1418 dumpable
= get_dumpable(mm
);
1426 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1428 struct inode
* inode
;
1429 struct proc_inode
*ei
;
1430 const struct cred
*cred
;
1432 /* We need a new inode */
1434 inode
= new_inode(sb
);
1440 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1441 inode
->i_op
= &proc_def_inode_operations
;
1444 * grab the reference to task.
1446 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1450 if (task_dumpable(task
)) {
1452 cred
= __task_cred(task
);
1453 inode
->i_uid
= cred
->euid
;
1454 inode
->i_gid
= cred
->egid
;
1457 security_task_to_inode(task
, inode
);
1467 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1469 struct inode
*inode
= dentry
->d_inode
;
1470 struct task_struct
*task
;
1471 const struct cred
*cred
;
1473 generic_fillattr(inode
, stat
);
1478 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1480 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1481 task_dumpable(task
)) {
1482 cred
= __task_cred(task
);
1483 stat
->uid
= cred
->euid
;
1484 stat
->gid
= cred
->egid
;
1494 * Exceptional case: normally we are not allowed to unhash a busy
1495 * directory. In this case, however, we can do it - no aliasing problems
1496 * due to the way we treat inodes.
1498 * Rewrite the inode's ownerships here because the owning task may have
1499 * performed a setuid(), etc.
1501 * Before the /proc/pid/status file was created the only way to read
1502 * the effective uid of a /process was to stat /proc/pid. Reading
1503 * /proc/pid/status is slow enough that procps and other packages
1504 * kept stating /proc/pid. To keep the rules in /proc simple I have
1505 * made this apply to all per process world readable and executable
1508 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1510 struct inode
*inode
= dentry
->d_inode
;
1511 struct task_struct
*task
= get_proc_task(inode
);
1512 const struct cred
*cred
;
1515 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1516 task_dumpable(task
)) {
1518 cred
= __task_cred(task
);
1519 inode
->i_uid
= cred
->euid
;
1520 inode
->i_gid
= cred
->egid
;
1526 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1527 security_task_to_inode(task
, inode
);
1528 put_task_struct(task
);
1535 static int pid_delete_dentry(struct dentry
* dentry
)
1537 /* Is the task we represent dead?
1538 * If so, then don't put the dentry on the lru list,
1539 * kill it immediately.
1541 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1544 static const struct dentry_operations pid_dentry_operations
=
1546 .d_revalidate
= pid_revalidate
,
1547 .d_delete
= pid_delete_dentry
,
1552 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1553 struct task_struct
*, const void *);
1556 * Fill a directory entry.
1558 * If possible create the dcache entry and derive our inode number and
1559 * file type from dcache entry.
1561 * Since all of the proc inode numbers are dynamically generated, the inode
1562 * numbers do not exist until the inode is cache. This means creating the
1563 * the dcache entry in readdir is necessary to keep the inode numbers
1564 * reported by readdir in sync with the inode numbers reported
1567 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1568 char *name
, int len
,
1569 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1571 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1572 struct inode
*inode
;
1575 unsigned type
= DT_UNKNOWN
;
1579 qname
.hash
= full_name_hash(name
, len
);
1581 child
= d_lookup(dir
, &qname
);
1584 new = d_alloc(dir
, &qname
);
1586 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1593 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1594 goto end_instantiate
;
1595 inode
= child
->d_inode
;
1598 type
= inode
->i_mode
>> 12;
1603 ino
= find_inode_number(dir
, &qname
);
1606 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1609 static unsigned name_to_int(struct dentry
*dentry
)
1611 const char *name
= dentry
->d_name
.name
;
1612 int len
= dentry
->d_name
.len
;
1615 if (len
> 1 && *name
== '0')
1618 unsigned c
= *name
++ - '0';
1621 if (n
>= (~0U-9)/10)
1631 #define PROC_FDINFO_MAX 64
1633 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1635 struct task_struct
*task
= get_proc_task(inode
);
1636 struct files_struct
*files
= NULL
;
1638 int fd
= proc_fd(inode
);
1641 files
= get_files_struct(task
);
1642 put_task_struct(task
);
1646 * We are not taking a ref to the file structure, so we must
1649 spin_lock(&files
->file_lock
);
1650 file
= fcheck_files(files
, fd
);
1653 *path
= file
->f_path
;
1654 path_get(&file
->f_path
);
1657 snprintf(info
, PROC_FDINFO_MAX
,
1660 (long long) file
->f_pos
,
1662 spin_unlock(&files
->file_lock
);
1663 put_files_struct(files
);
1666 spin_unlock(&files
->file_lock
);
1667 put_files_struct(files
);
1672 static int proc_fd_link(struct inode
*inode
, struct path
*path
)
1674 return proc_fd_info(inode
, path
, NULL
);
1677 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1679 struct inode
*inode
= dentry
->d_inode
;
1680 struct task_struct
*task
= get_proc_task(inode
);
1681 int fd
= proc_fd(inode
);
1682 struct files_struct
*files
;
1683 const struct cred
*cred
;
1686 files
= get_files_struct(task
);
1689 if (fcheck_files(files
, fd
)) {
1691 put_files_struct(files
);
1692 if (task_dumpable(task
)) {
1694 cred
= __task_cred(task
);
1695 inode
->i_uid
= cred
->euid
;
1696 inode
->i_gid
= cred
->egid
;
1702 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1703 security_task_to_inode(task
, inode
);
1704 put_task_struct(task
);
1708 put_files_struct(files
);
1710 put_task_struct(task
);
1716 static const struct dentry_operations tid_fd_dentry_operations
=
1718 .d_revalidate
= tid_fd_revalidate
,
1719 .d_delete
= pid_delete_dentry
,
1722 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1723 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1725 unsigned fd
= *(const unsigned *)ptr
;
1727 struct files_struct
*files
;
1728 struct inode
*inode
;
1729 struct proc_inode
*ei
;
1730 struct dentry
*error
= ERR_PTR(-ENOENT
);
1732 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1737 files
= get_files_struct(task
);
1740 inode
->i_mode
= S_IFLNK
;
1743 * We are not taking a ref to the file structure, so we must
1746 spin_lock(&files
->file_lock
);
1747 file
= fcheck_files(files
, fd
);
1750 if (file
->f_mode
& FMODE_READ
)
1751 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1752 if (file
->f_mode
& FMODE_WRITE
)
1753 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1754 spin_unlock(&files
->file_lock
);
1755 put_files_struct(files
);
1757 inode
->i_op
= &proc_pid_link_inode_operations
;
1759 ei
->op
.proc_get_link
= proc_fd_link
;
1760 dentry
->d_op
= &tid_fd_dentry_operations
;
1761 d_add(dentry
, inode
);
1762 /* Close the race of the process dying before we return the dentry */
1763 if (tid_fd_revalidate(dentry
, NULL
))
1769 spin_unlock(&files
->file_lock
);
1770 put_files_struct(files
);
1776 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1777 struct dentry
*dentry
,
1778 instantiate_t instantiate
)
1780 struct task_struct
*task
= get_proc_task(dir
);
1781 unsigned fd
= name_to_int(dentry
);
1782 struct dentry
*result
= ERR_PTR(-ENOENT
);
1789 result
= instantiate(dir
, dentry
, task
, &fd
);
1791 put_task_struct(task
);
1796 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1797 filldir_t filldir
, instantiate_t instantiate
)
1799 struct dentry
*dentry
= filp
->f_path
.dentry
;
1800 struct inode
*inode
= dentry
->d_inode
;
1801 struct task_struct
*p
= get_proc_task(inode
);
1802 unsigned int fd
, ino
;
1804 struct files_struct
* files
;
1814 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1818 ino
= parent_ino(dentry
);
1819 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1823 files
= get_files_struct(p
);
1827 for (fd
= filp
->f_pos
-2;
1828 fd
< files_fdtable(files
)->max_fds
;
1829 fd
++, filp
->f_pos
++) {
1830 char name
[PROC_NUMBUF
];
1833 if (!fcheck_files(files
, fd
))
1837 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1838 if (proc_fill_cache(filp
, dirent
, filldir
,
1839 name
, len
, instantiate
,
1847 put_files_struct(files
);
1855 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1856 struct nameidata
*nd
)
1858 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1861 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1863 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1866 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1867 size_t len
, loff_t
*ppos
)
1869 char tmp
[PROC_FDINFO_MAX
];
1870 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1872 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1876 static const struct file_operations proc_fdinfo_file_operations
= {
1877 .open
= nonseekable_open
,
1878 .read
= proc_fdinfo_read
,
1881 static const struct file_operations proc_fd_operations
= {
1882 .read
= generic_read_dir
,
1883 .readdir
= proc_readfd
,
1887 * /proc/pid/fd needs a special permission handler so that a process can still
1888 * access /proc/self/fd after it has executed a setuid().
1890 static int proc_fd_permission(struct inode
*inode
, int mask
)
1894 rv
= generic_permission(inode
, mask
, NULL
);
1897 if (task_pid(current
) == proc_pid(inode
))
1903 * proc directories can do almost nothing..
1905 static const struct inode_operations proc_fd_inode_operations
= {
1906 .lookup
= proc_lookupfd
,
1907 .permission
= proc_fd_permission
,
1908 .setattr
= proc_setattr
,
1911 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
1912 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1914 unsigned fd
= *(unsigned *)ptr
;
1915 struct inode
*inode
;
1916 struct proc_inode
*ei
;
1917 struct dentry
*error
= ERR_PTR(-ENOENT
);
1919 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1924 inode
->i_mode
= S_IFREG
| S_IRUSR
;
1925 inode
->i_fop
= &proc_fdinfo_file_operations
;
1926 dentry
->d_op
= &tid_fd_dentry_operations
;
1927 d_add(dentry
, inode
);
1928 /* Close the race of the process dying before we return the dentry */
1929 if (tid_fd_revalidate(dentry
, NULL
))
1936 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
1937 struct dentry
*dentry
,
1938 struct nameidata
*nd
)
1940 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
1943 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
1945 return proc_readfd_common(filp
, dirent
, filldir
,
1946 proc_fdinfo_instantiate
);
1949 static const struct file_operations proc_fdinfo_operations
= {
1950 .read
= generic_read_dir
,
1951 .readdir
= proc_readfdinfo
,
1955 * proc directories can do almost nothing..
1957 static const struct inode_operations proc_fdinfo_inode_operations
= {
1958 .lookup
= proc_lookupfdinfo
,
1959 .setattr
= proc_setattr
,
1963 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
1964 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1966 const struct pid_entry
*p
= ptr
;
1967 struct inode
*inode
;
1968 struct proc_inode
*ei
;
1969 struct dentry
*error
= ERR_PTR(-ENOENT
);
1971 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1976 inode
->i_mode
= p
->mode
;
1977 if (S_ISDIR(inode
->i_mode
))
1978 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1980 inode
->i_op
= p
->iop
;
1982 inode
->i_fop
= p
->fop
;
1984 dentry
->d_op
= &pid_dentry_operations
;
1985 d_add(dentry
, inode
);
1986 /* Close the race of the process dying before we return the dentry */
1987 if (pid_revalidate(dentry
, NULL
))
1993 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1994 struct dentry
*dentry
,
1995 const struct pid_entry
*ents
,
1998 struct dentry
*error
;
1999 struct task_struct
*task
= get_proc_task(dir
);
2000 const struct pid_entry
*p
, *last
;
2002 error
= ERR_PTR(-ENOENT
);
2008 * Yes, it does not scale. And it should not. Don't add
2009 * new entries into /proc/<tgid>/ without very good reasons.
2011 last
= &ents
[nents
- 1];
2012 for (p
= ents
; p
<= last
; p
++) {
2013 if (p
->len
!= dentry
->d_name
.len
)
2015 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2021 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2023 put_task_struct(task
);
2028 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2029 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2031 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2032 proc_pident_instantiate
, task
, p
);
2035 static int proc_pident_readdir(struct file
*filp
,
2036 void *dirent
, filldir_t filldir
,
2037 const struct pid_entry
*ents
, unsigned int nents
)
2040 struct dentry
*dentry
= filp
->f_path
.dentry
;
2041 struct inode
*inode
= dentry
->d_inode
;
2042 struct task_struct
*task
= get_proc_task(inode
);
2043 const struct pid_entry
*p
, *last
;
2056 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2062 ino
= parent_ino(dentry
);
2063 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2075 last
= &ents
[nents
- 1];
2077 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2086 put_task_struct(task
);
2091 #ifdef CONFIG_SECURITY
2092 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2093 size_t count
, loff_t
*ppos
)
2095 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2098 struct task_struct
*task
= get_proc_task(inode
);
2103 length
= security_getprocattr(task
,
2104 (char*)file
->f_path
.dentry
->d_name
.name
,
2106 put_task_struct(task
);
2108 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2113 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2114 size_t count
, loff_t
*ppos
)
2116 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2119 struct task_struct
*task
= get_proc_task(inode
);
2124 if (count
> PAGE_SIZE
)
2127 /* No partial writes. */
2133 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2138 if (copy_from_user(page
, buf
, count
))
2141 /* Guard against adverse ptrace interaction */
2142 length
= mutex_lock_interruptible(&task
->cred_guard_mutex
);
2146 length
= security_setprocattr(task
,
2147 (char*)file
->f_path
.dentry
->d_name
.name
,
2148 (void*)page
, count
);
2149 mutex_unlock(&task
->cred_guard_mutex
);
2151 free_page((unsigned long) page
);
2153 put_task_struct(task
);
2158 static const struct file_operations proc_pid_attr_operations
= {
2159 .read
= proc_pid_attr_read
,
2160 .write
= proc_pid_attr_write
,
2163 static const struct pid_entry attr_dir_stuff
[] = {
2164 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2165 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2166 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2167 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2168 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2169 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2172 static int proc_attr_dir_readdir(struct file
* filp
,
2173 void * dirent
, filldir_t filldir
)
2175 return proc_pident_readdir(filp
,dirent
,filldir
,
2176 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2179 static const struct file_operations proc_attr_dir_operations
= {
2180 .read
= generic_read_dir
,
2181 .readdir
= proc_attr_dir_readdir
,
2184 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2185 struct dentry
*dentry
, struct nameidata
*nd
)
2187 return proc_pident_lookup(dir
, dentry
,
2188 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2191 static const struct inode_operations proc_attr_dir_inode_operations
= {
2192 .lookup
= proc_attr_dir_lookup
,
2193 .getattr
= pid_getattr
,
2194 .setattr
= proc_setattr
,
2199 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2200 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2201 size_t count
, loff_t
*ppos
)
2203 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2204 struct mm_struct
*mm
;
2205 char buffer
[PROC_NUMBUF
];
2213 mm
= get_task_mm(task
);
2215 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2216 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2217 MMF_DUMP_FILTER_SHIFT
));
2219 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2222 put_task_struct(task
);
2227 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2228 const char __user
*buf
,
2232 struct task_struct
*task
;
2233 struct mm_struct
*mm
;
2234 char buffer
[PROC_NUMBUF
], *end
;
2241 memset(buffer
, 0, sizeof(buffer
));
2242 if (count
> sizeof(buffer
) - 1)
2243 count
= sizeof(buffer
) - 1;
2244 if (copy_from_user(buffer
, buf
, count
))
2248 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2251 if (end
- buffer
== 0)
2255 task
= get_proc_task(file
->f_dentry
->d_inode
);
2260 mm
= get_task_mm(task
);
2264 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2266 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2268 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2273 put_task_struct(task
);
2278 static const struct file_operations proc_coredump_filter_operations
= {
2279 .read
= proc_coredump_filter_read
,
2280 .write
= proc_coredump_filter_write
,
2287 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2290 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2291 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2292 char tmp
[PROC_NUMBUF
];
2295 sprintf(tmp
, "%d", tgid
);
2296 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2299 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2301 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2302 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2303 char tmp
[PROC_NUMBUF
];
2305 return ERR_PTR(-ENOENT
);
2306 sprintf(tmp
, "%d", task_tgid_nr_ns(current
, ns
));
2307 return ERR_PTR(vfs_follow_link(nd
,tmp
));
2310 static const struct inode_operations proc_self_inode_operations
= {
2311 .readlink
= proc_self_readlink
,
2312 .follow_link
= proc_self_follow_link
,
2318 * These are the directory entries in the root directory of /proc
2319 * that properly belong to the /proc filesystem, as they describe
2320 * describe something that is process related.
2322 static const struct pid_entry proc_base_stuff
[] = {
2323 NOD("self", S_IFLNK
|S_IRWXUGO
,
2324 &proc_self_inode_operations
, NULL
, {}),
2328 * Exceptional case: normally we are not allowed to unhash a busy
2329 * directory. In this case, however, we can do it - no aliasing problems
2330 * due to the way we treat inodes.
2332 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
2334 struct inode
*inode
= dentry
->d_inode
;
2335 struct task_struct
*task
= get_proc_task(inode
);
2337 put_task_struct(task
);
2344 static const struct dentry_operations proc_base_dentry_operations
=
2346 .d_revalidate
= proc_base_revalidate
,
2347 .d_delete
= pid_delete_dentry
,
2350 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2351 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2353 const struct pid_entry
*p
= ptr
;
2354 struct inode
*inode
;
2355 struct proc_inode
*ei
;
2356 struct dentry
*error
= ERR_PTR(-EINVAL
);
2358 /* Allocate the inode */
2359 error
= ERR_PTR(-ENOMEM
);
2360 inode
= new_inode(dir
->i_sb
);
2364 /* Initialize the inode */
2366 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2369 * grab the reference to the task.
2371 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2375 inode
->i_mode
= p
->mode
;
2376 if (S_ISDIR(inode
->i_mode
))
2378 if (S_ISLNK(inode
->i_mode
))
2381 inode
->i_op
= p
->iop
;
2383 inode
->i_fop
= p
->fop
;
2385 dentry
->d_op
= &proc_base_dentry_operations
;
2386 d_add(dentry
, inode
);
2395 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2397 struct dentry
*error
;
2398 struct task_struct
*task
= get_proc_task(dir
);
2399 const struct pid_entry
*p
, *last
;
2401 error
= ERR_PTR(-ENOENT
);
2406 /* Lookup the directory entry */
2407 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2408 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2409 if (p
->len
!= dentry
->d_name
.len
)
2411 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2417 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2420 put_task_struct(task
);
2425 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2426 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2428 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2429 proc_base_instantiate
, task
, p
);
2432 #ifdef CONFIG_TASK_IO_ACCOUNTING
2433 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2435 struct task_io_accounting acct
= task
->ioac
;
2436 unsigned long flags
;
2438 if (whole
&& lock_task_sighand(task
, &flags
)) {
2439 struct task_struct
*t
= task
;
2441 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2442 while_each_thread(task
, t
)
2443 task_io_accounting_add(&acct
, &t
->ioac
);
2445 unlock_task_sighand(task
, &flags
);
2447 return sprintf(buffer
,
2452 "read_bytes: %llu\n"
2453 "write_bytes: %llu\n"
2454 "cancelled_write_bytes: %llu\n",
2455 (unsigned long long)acct
.rchar
,
2456 (unsigned long long)acct
.wchar
,
2457 (unsigned long long)acct
.syscr
,
2458 (unsigned long long)acct
.syscw
,
2459 (unsigned long long)acct
.read_bytes
,
2460 (unsigned long long)acct
.write_bytes
,
2461 (unsigned long long)acct
.cancelled_write_bytes
);
2464 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2466 return do_io_accounting(task
, buffer
, 0);
2469 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2471 return do_io_accounting(task
, buffer
, 1);
2473 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2475 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2476 struct pid
*pid
, struct task_struct
*task
)
2478 seq_printf(m
, "%08x\n", task
->personality
);
2485 static const struct file_operations proc_task_operations
;
2486 static const struct inode_operations proc_task_inode_operations
;
2488 static const struct pid_entry tgid_base_stuff
[] = {
2489 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2490 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2491 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2493 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2495 REG("environ", S_IRUSR
, proc_environ_operations
),
2496 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2497 ONE("status", S_IRUGO
, proc_pid_status
),
2498 ONE("personality", S_IRUSR
, proc_pid_personality
),
2499 INF("limits", S_IRUSR
, proc_pid_limits
),
2500 #ifdef CONFIG_SCHED_DEBUG
2501 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2503 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2504 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2506 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2507 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2508 ONE("statm", S_IRUGO
, proc_pid_statm
),
2509 REG("maps", S_IRUGO
, proc_maps_operations
),
2511 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2513 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2514 LNK("cwd", proc_cwd_link
),
2515 LNK("root", proc_root_link
),
2516 LNK("exe", proc_exe_link
),
2517 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2518 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2519 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2520 #ifdef CONFIG_PROC_PAGE_MONITOR
2521 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2522 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2523 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2525 #ifdef CONFIG_SECURITY
2526 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2528 #ifdef CONFIG_KALLSYMS
2529 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2531 #ifdef CONFIG_STACKTRACE
2532 ONE("stack", S_IRUSR
, proc_pid_stack
),
2534 #ifdef CONFIG_SCHEDSTATS
2535 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2537 #ifdef CONFIG_LATENCYTOP
2538 REG("latency", S_IRUGO
, proc_lstats_operations
),
2540 #ifdef CONFIG_PROC_PID_CPUSET
2541 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2543 #ifdef CONFIG_CGROUPS
2544 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2546 INF("oom_score", S_IRUGO
, proc_oom_score
),
2547 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2548 #ifdef CONFIG_AUDITSYSCALL
2549 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2550 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2552 #ifdef CONFIG_FAULT_INJECTION
2553 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2555 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2556 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2558 #ifdef CONFIG_TASK_IO_ACCOUNTING
2559 INF("io", S_IRUGO
, proc_tgid_io_accounting
),
2563 static int proc_tgid_base_readdir(struct file
* filp
,
2564 void * dirent
, filldir_t filldir
)
2566 return proc_pident_readdir(filp
,dirent
,filldir
,
2567 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2570 static const struct file_operations proc_tgid_base_operations
= {
2571 .read
= generic_read_dir
,
2572 .readdir
= proc_tgid_base_readdir
,
2575 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2576 return proc_pident_lookup(dir
, dentry
,
2577 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2580 static const struct inode_operations proc_tgid_base_inode_operations
= {
2581 .lookup
= proc_tgid_base_lookup
,
2582 .getattr
= pid_getattr
,
2583 .setattr
= proc_setattr
,
2586 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2588 struct dentry
*dentry
, *leader
, *dir
;
2589 char buf
[PROC_NUMBUF
];
2593 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2594 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2596 if (!(current
->flags
& PF_EXITING
))
2597 shrink_dcache_parent(dentry
);
2606 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2607 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2612 name
.len
= strlen(name
.name
);
2613 dir
= d_hash_and_lookup(leader
, &name
);
2615 goto out_put_leader
;
2618 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2619 dentry
= d_hash_and_lookup(dir
, &name
);
2621 shrink_dcache_parent(dentry
);
2634 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2635 * @task: task that should be flushed.
2637 * When flushing dentries from proc, one needs to flush them from global
2638 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2639 * in. This call is supposed to do all of this job.
2641 * Looks in the dcache for
2643 * /proc/@tgid/task/@pid
2644 * if either directory is present flushes it and all of it'ts children
2647 * It is safe and reasonable to cache /proc entries for a task until
2648 * that task exits. After that they just clog up the dcache with
2649 * useless entries, possibly causing useful dcache entries to be
2650 * flushed instead. This routine is proved to flush those useless
2651 * dcache entries at process exit time.
2653 * NOTE: This routine is just an optimization so it does not guarantee
2654 * that no dcache entries will exist at process exit time it
2655 * just makes it very unlikely that any will persist.
2658 void proc_flush_task(struct task_struct
*task
)
2661 struct pid
*pid
, *tgid
= NULL
;
2664 pid
= task_pid(task
);
2665 if (thread_group_leader(task
))
2666 tgid
= task_tgid(task
);
2668 for (i
= 0; i
<= pid
->level
; i
++) {
2669 upid
= &pid
->numbers
[i
];
2670 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2671 tgid
? tgid
->numbers
[i
].nr
: 0);
2674 upid
= &pid
->numbers
[pid
->level
];
2676 pid_ns_release_proc(upid
->ns
);
2679 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2680 struct dentry
* dentry
,
2681 struct task_struct
*task
, const void *ptr
)
2683 struct dentry
*error
= ERR_PTR(-ENOENT
);
2684 struct inode
*inode
;
2686 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2690 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2691 inode
->i_op
= &proc_tgid_base_inode_operations
;
2692 inode
->i_fop
= &proc_tgid_base_operations
;
2693 inode
->i_flags
|=S_IMMUTABLE
;
2695 inode
->i_nlink
= 2 + pid_entry_count_dirs(tgid_base_stuff
,
2696 ARRAY_SIZE(tgid_base_stuff
));
2698 dentry
->d_op
= &pid_dentry_operations
;
2700 d_add(dentry
, inode
);
2701 /* Close the race of the process dying before we return the dentry */
2702 if (pid_revalidate(dentry
, NULL
))
2708 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2710 struct dentry
*result
= ERR_PTR(-ENOENT
);
2711 struct task_struct
*task
;
2713 struct pid_namespace
*ns
;
2715 result
= proc_base_lookup(dir
, dentry
);
2716 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2719 tgid
= name_to_int(dentry
);
2723 ns
= dentry
->d_sb
->s_fs_info
;
2725 task
= find_task_by_pid_ns(tgid
, ns
);
2727 get_task_struct(task
);
2732 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2733 put_task_struct(task
);
2739 * Find the first task with tgid >= tgid
2744 struct task_struct
*task
;
2746 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
2751 put_task_struct(iter
.task
);
2755 pid
= find_ge_pid(iter
.tgid
, ns
);
2757 iter
.tgid
= pid_nr_ns(pid
, ns
);
2758 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
2759 /* What we to know is if the pid we have find is the
2760 * pid of a thread_group_leader. Testing for task
2761 * being a thread_group_leader is the obvious thing
2762 * todo but there is a window when it fails, due to
2763 * the pid transfer logic in de_thread.
2765 * So we perform the straight forward test of seeing
2766 * if the pid we have found is the pid of a thread
2767 * group leader, and don't worry if the task we have
2768 * found doesn't happen to be a thread group leader.
2769 * As we don't care in the case of readdir.
2771 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
2775 get_task_struct(iter
.task
);
2781 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2783 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2784 struct tgid_iter iter
)
2786 char name
[PROC_NUMBUF
];
2787 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
2788 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2789 proc_pid_instantiate
, iter
.task
, NULL
);
2792 /* for the /proc/ directory itself, after non-process stuff has been done */
2793 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2795 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2796 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2797 struct tgid_iter iter
;
2798 struct pid_namespace
*ns
;
2803 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2804 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2805 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2809 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2811 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
2812 for (iter
= next_tgid(ns
, iter
);
2814 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
2815 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
2816 if (proc_pid_fill_cache(filp
, dirent
, filldir
, iter
) < 0) {
2817 put_task_struct(iter
.task
);
2821 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2823 put_task_struct(reaper
);
2831 static const struct pid_entry tid_base_stuff
[] = {
2832 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2833 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fd_operations
),
2834 REG("environ", S_IRUSR
, proc_environ_operations
),
2835 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2836 ONE("status", S_IRUGO
, proc_pid_status
),
2837 ONE("personality", S_IRUSR
, proc_pid_personality
),
2838 INF("limits", S_IRUSR
, proc_pid_limits
),
2839 #ifdef CONFIG_SCHED_DEBUG
2840 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2842 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2843 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2845 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2846 ONE("stat", S_IRUGO
, proc_tid_stat
),
2847 ONE("statm", S_IRUGO
, proc_pid_statm
),
2848 REG("maps", S_IRUGO
, proc_maps_operations
),
2850 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2852 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2853 LNK("cwd", proc_cwd_link
),
2854 LNK("root", proc_root_link
),
2855 LNK("exe", proc_exe_link
),
2856 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2857 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2858 #ifdef CONFIG_PROC_PAGE_MONITOR
2859 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2860 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2861 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2863 #ifdef CONFIG_SECURITY
2864 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2866 #ifdef CONFIG_KALLSYMS
2867 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2869 #ifdef CONFIG_STACKTRACE
2870 ONE("stack", S_IRUSR
, proc_pid_stack
),
2872 #ifdef CONFIG_SCHEDSTATS
2873 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2875 #ifdef CONFIG_LATENCYTOP
2876 REG("latency", S_IRUGO
, proc_lstats_operations
),
2878 #ifdef CONFIG_PROC_PID_CPUSET
2879 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2881 #ifdef CONFIG_CGROUPS
2882 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2884 INF("oom_score", S_IRUGO
, proc_oom_score
),
2885 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2886 #ifdef CONFIG_AUDITSYSCALL
2887 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2888 REG("sessionid", S_IRUSR
, proc_sessionid_operations
),
2890 #ifdef CONFIG_FAULT_INJECTION
2891 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2893 #ifdef CONFIG_TASK_IO_ACCOUNTING
2894 INF("io", S_IRUGO
, proc_tid_io_accounting
),
2898 static int proc_tid_base_readdir(struct file
* filp
,
2899 void * dirent
, filldir_t filldir
)
2901 return proc_pident_readdir(filp
,dirent
,filldir
,
2902 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2905 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2906 return proc_pident_lookup(dir
, dentry
,
2907 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
2910 static const struct file_operations proc_tid_base_operations
= {
2911 .read
= generic_read_dir
,
2912 .readdir
= proc_tid_base_readdir
,
2915 static const struct inode_operations proc_tid_base_inode_operations
= {
2916 .lookup
= proc_tid_base_lookup
,
2917 .getattr
= pid_getattr
,
2918 .setattr
= proc_setattr
,
2921 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
2922 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2924 struct dentry
*error
= ERR_PTR(-ENOENT
);
2925 struct inode
*inode
;
2926 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2930 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2931 inode
->i_op
= &proc_tid_base_inode_operations
;
2932 inode
->i_fop
= &proc_tid_base_operations
;
2933 inode
->i_flags
|=S_IMMUTABLE
;
2935 inode
->i_nlink
= 2 + pid_entry_count_dirs(tid_base_stuff
,
2936 ARRAY_SIZE(tid_base_stuff
));
2938 dentry
->d_op
= &pid_dentry_operations
;
2940 d_add(dentry
, inode
);
2941 /* Close the race of the process dying before we return the dentry */
2942 if (pid_revalidate(dentry
, NULL
))
2948 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2950 struct dentry
*result
= ERR_PTR(-ENOENT
);
2951 struct task_struct
*task
;
2952 struct task_struct
*leader
= get_proc_task(dir
);
2954 struct pid_namespace
*ns
;
2959 tid
= name_to_int(dentry
);
2963 ns
= dentry
->d_sb
->s_fs_info
;
2965 task
= find_task_by_pid_ns(tid
, ns
);
2967 get_task_struct(task
);
2971 if (!same_thread_group(leader
, task
))
2974 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
2976 put_task_struct(task
);
2978 put_task_struct(leader
);
2984 * Find the first tid of a thread group to return to user space.
2986 * Usually this is just the thread group leader, but if the users
2987 * buffer was too small or there was a seek into the middle of the
2988 * directory we have more work todo.
2990 * In the case of a short read we start with find_task_by_pid.
2992 * In the case of a seek we start with the leader and walk nr
2995 static struct task_struct
*first_tid(struct task_struct
*leader
,
2996 int tid
, int nr
, struct pid_namespace
*ns
)
2998 struct task_struct
*pos
;
3001 /* Attempt to start with the pid of a thread */
3002 if (tid
&& (nr
> 0)) {
3003 pos
= find_task_by_pid_ns(tid
, ns
);
3004 if (pos
&& (pos
->group_leader
== leader
))
3008 /* If nr exceeds the number of threads there is nothing todo */
3010 if (nr
&& nr
>= get_nr_threads(leader
))
3013 /* If we haven't found our starting place yet start
3014 * with the leader and walk nr threads forward.
3016 for (pos
= leader
; nr
> 0; --nr
) {
3017 pos
= next_thread(pos
);
3018 if (pos
== leader
) {
3024 get_task_struct(pos
);
3031 * Find the next thread in the thread list.
3032 * Return NULL if there is an error or no next thread.
3034 * The reference to the input task_struct is released.
3036 static struct task_struct
*next_tid(struct task_struct
*start
)
3038 struct task_struct
*pos
= NULL
;
3040 if (pid_alive(start
)) {
3041 pos
= next_thread(start
);
3042 if (thread_group_leader(pos
))
3045 get_task_struct(pos
);
3048 put_task_struct(start
);
3052 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3053 struct task_struct
*task
, int tid
)
3055 char name
[PROC_NUMBUF
];
3056 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3057 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3058 proc_task_instantiate
, task
, NULL
);
3061 /* for the /proc/TGID/task/ directories */
3062 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3064 struct dentry
*dentry
= filp
->f_path
.dentry
;
3065 struct inode
*inode
= dentry
->d_inode
;
3066 struct task_struct
*leader
= NULL
;
3067 struct task_struct
*task
;
3068 int retval
= -ENOENT
;
3071 struct pid_namespace
*ns
;
3073 task
= get_proc_task(inode
);
3077 if (pid_alive(task
)) {
3078 leader
= task
->group_leader
;
3079 get_task_struct(leader
);
3082 put_task_struct(task
);
3087 switch ((unsigned long)filp
->f_pos
) {
3090 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3095 ino
= parent_ino(dentry
);
3096 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3102 /* f_version caches the tgid value that the last readdir call couldn't
3103 * return. lseek aka telldir automagically resets f_version to 0.
3105 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3106 tid
= (int)filp
->f_version
;
3107 filp
->f_version
= 0;
3108 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3110 task
= next_tid(task
), filp
->f_pos
++) {
3111 tid
= task_pid_nr_ns(task
, ns
);
3112 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3113 /* returning this tgid failed, save it as the first
3114 * pid for the next readir call */
3115 filp
->f_version
= (u64
)tid
;
3116 put_task_struct(task
);
3121 put_task_struct(leader
);
3126 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3128 struct inode
*inode
= dentry
->d_inode
;
3129 struct task_struct
*p
= get_proc_task(inode
);
3130 generic_fillattr(inode
, stat
);
3133 stat
->nlink
+= get_nr_threads(p
);
3140 static const struct inode_operations proc_task_inode_operations
= {
3141 .lookup
= proc_task_lookup
,
3142 .getattr
= proc_task_getattr
,
3143 .setattr
= proc_setattr
,
3146 static const struct file_operations proc_task_operations
= {
3147 .read
= generic_read_dir
,
3148 .readdir
= proc_task_readdir
,