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
];
1006 oom_adjust
= task
->oomkilladj
;
1007 put_task_struct(task
);
1009 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
1011 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1014 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
1015 size_t count
, loff_t
*ppos
)
1017 struct task_struct
*task
;
1018 char buffer
[PROC_NUMBUF
], *end
;
1021 memset(buffer
, 0, sizeof(buffer
));
1022 if (count
> sizeof(buffer
) - 1)
1023 count
= sizeof(buffer
) - 1;
1024 if (copy_from_user(buffer
, buf
, count
))
1026 oom_adjust
= simple_strtol(buffer
, &end
, 0);
1027 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
1028 oom_adjust
!= OOM_DISABLE
)
1032 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1035 if (oom_adjust
< task
->oomkilladj
&& !capable(CAP_SYS_RESOURCE
)) {
1036 put_task_struct(task
);
1039 task
->oomkilladj
= oom_adjust
;
1040 put_task_struct(task
);
1041 if (end
- buffer
== 0)
1043 return end
- buffer
;
1046 static const struct file_operations proc_oom_adjust_operations
= {
1047 .read
= oom_adjust_read
,
1048 .write
= oom_adjust_write
,
1051 #ifdef CONFIG_AUDITSYSCALL
1052 #define TMPBUFLEN 21
1053 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1054 size_t count
, loff_t
*ppos
)
1056 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1057 struct task_struct
*task
= get_proc_task(inode
);
1059 char tmpbuf
[TMPBUFLEN
];
1063 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1064 audit_get_loginuid(task
));
1065 put_task_struct(task
);
1066 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1069 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1070 size_t count
, loff_t
*ppos
)
1072 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1077 if (!capable(CAP_AUDIT_CONTROL
))
1080 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
1083 if (count
>= PAGE_SIZE
)
1084 count
= PAGE_SIZE
- 1;
1087 /* No partial writes. */
1090 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1094 if (copy_from_user(page
, buf
, count
))
1098 loginuid
= simple_strtoul(page
, &tmp
, 10);
1104 length
= audit_set_loginuid(current
, loginuid
);
1105 if (likely(length
== 0))
1109 free_page((unsigned long) page
);
1113 static const struct file_operations proc_loginuid_operations
= {
1114 .read
= proc_loginuid_read
,
1115 .write
= proc_loginuid_write
,
1118 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1119 size_t count
, loff_t
*ppos
)
1121 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1122 struct task_struct
*task
= get_proc_task(inode
);
1124 char tmpbuf
[TMPBUFLEN
];
1128 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1129 audit_get_sessionid(task
));
1130 put_task_struct(task
);
1131 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1134 static const struct file_operations proc_sessionid_operations
= {
1135 .read
= proc_sessionid_read
,
1139 #ifdef CONFIG_FAULT_INJECTION
1140 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1141 size_t count
, loff_t
*ppos
)
1143 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1144 char buffer
[PROC_NUMBUF
];
1150 make_it_fail
= task
->make_it_fail
;
1151 put_task_struct(task
);
1153 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1155 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1158 static ssize_t
proc_fault_inject_write(struct file
* file
,
1159 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1161 struct task_struct
*task
;
1162 char buffer
[PROC_NUMBUF
], *end
;
1165 if (!capable(CAP_SYS_RESOURCE
))
1167 memset(buffer
, 0, sizeof(buffer
));
1168 if (count
> sizeof(buffer
) - 1)
1169 count
= sizeof(buffer
) - 1;
1170 if (copy_from_user(buffer
, buf
, count
))
1172 make_it_fail
= simple_strtol(buffer
, &end
, 0);
1175 task
= get_proc_task(file
->f_dentry
->d_inode
);
1178 task
->make_it_fail
= make_it_fail
;
1179 put_task_struct(task
);
1180 if (end
- buffer
== 0)
1182 return end
- buffer
;
1185 static const struct file_operations proc_fault_inject_operations
= {
1186 .read
= proc_fault_inject_read
,
1187 .write
= proc_fault_inject_write
,
1192 #ifdef CONFIG_SCHED_DEBUG
1194 * Print out various scheduling related per-task fields:
1196 static int sched_show(struct seq_file
*m
, void *v
)
1198 struct inode
*inode
= m
->private;
1199 struct task_struct
*p
;
1201 p
= get_proc_task(inode
);
1204 proc_sched_show_task(p
, m
);
1212 sched_write(struct file
*file
, const char __user
*buf
,
1213 size_t count
, loff_t
*offset
)
1215 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1216 struct task_struct
*p
;
1218 p
= get_proc_task(inode
);
1221 proc_sched_set_task(p
);
1228 static int sched_open(struct inode
*inode
, struct file
*filp
)
1232 ret
= single_open(filp
, sched_show
, NULL
);
1234 struct seq_file
*m
= filp
->private_data
;
1241 static const struct file_operations proc_pid_sched_operations
= {
1244 .write
= sched_write
,
1245 .llseek
= seq_lseek
,
1246 .release
= single_release
,
1252 * We added or removed a vma mapping the executable. The vmas are only mapped
1253 * during exec and are not mapped with the mmap system call.
1254 * Callers must hold down_write() on the mm's mmap_sem for these
1256 void added_exe_file_vma(struct mm_struct
*mm
)
1258 mm
->num_exe_file_vmas
++;
1261 void removed_exe_file_vma(struct mm_struct
*mm
)
1263 mm
->num_exe_file_vmas
--;
1264 if ((mm
->num_exe_file_vmas
== 0) && mm
->exe_file
){
1266 mm
->exe_file
= NULL
;
1271 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
1274 get_file(new_exe_file
);
1277 mm
->exe_file
= new_exe_file
;
1278 mm
->num_exe_file_vmas
= 0;
1281 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
1283 struct file
*exe_file
;
1285 /* We need mmap_sem to protect against races with removal of
1286 * VM_EXECUTABLE vmas */
1287 down_read(&mm
->mmap_sem
);
1288 exe_file
= mm
->exe_file
;
1291 up_read(&mm
->mmap_sem
);
1295 void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
1297 /* It's safe to write the exe_file pointer without exe_file_lock because
1298 * this is called during fork when the task is not yet in /proc */
1299 newmm
->exe_file
= get_mm_exe_file(oldmm
);
1302 static int proc_exe_link(struct inode
*inode
, struct path
*exe_path
)
1304 struct task_struct
*task
;
1305 struct mm_struct
*mm
;
1306 struct file
*exe_file
;
1308 task
= get_proc_task(inode
);
1311 mm
= get_task_mm(task
);
1312 put_task_struct(task
);
1315 exe_file
= get_mm_exe_file(mm
);
1318 *exe_path
= exe_file
->f_path
;
1319 path_get(&exe_file
->f_path
);
1326 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1328 struct inode
*inode
= dentry
->d_inode
;
1329 int error
= -EACCES
;
1331 /* We don't need a base pointer in the /proc filesystem */
1332 path_put(&nd
->path
);
1334 /* Are we allowed to snoop on the tasks file descriptors? */
1335 if (!proc_fd_access_allowed(inode
))
1338 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->path
);
1339 nd
->last_type
= LAST_BIND
;
1341 return ERR_PTR(error
);
1344 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1346 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1353 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1354 len
= PTR_ERR(pathname
);
1355 if (IS_ERR(pathname
))
1357 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1361 if (copy_to_user(buffer
, pathname
, len
))
1364 free_page((unsigned long)tmp
);
1368 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1370 int error
= -EACCES
;
1371 struct inode
*inode
= dentry
->d_inode
;
1374 /* Are we allowed to snoop on the tasks file descriptors? */
1375 if (!proc_fd_access_allowed(inode
))
1378 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &path
);
1382 error
= do_proc_readlink(&path
, buffer
, buflen
);
1388 static const struct inode_operations proc_pid_link_inode_operations
= {
1389 .readlink
= proc_pid_readlink
,
1390 .follow_link
= proc_pid_follow_link
,
1391 .setattr
= proc_setattr
,
1395 /* building an inode */
1397 static int task_dumpable(struct task_struct
*task
)
1400 struct mm_struct
*mm
;
1405 dumpable
= get_dumpable(mm
);
1413 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1415 struct inode
* inode
;
1416 struct proc_inode
*ei
;
1417 const struct cred
*cred
;
1419 /* We need a new inode */
1421 inode
= new_inode(sb
);
1427 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1428 inode
->i_op
= &proc_def_inode_operations
;
1431 * grab the reference to task.
1433 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1437 if (task_dumpable(task
)) {
1439 cred
= __task_cred(task
);
1440 inode
->i_uid
= cred
->euid
;
1441 inode
->i_gid
= cred
->egid
;
1444 security_task_to_inode(task
, inode
);
1454 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1456 struct inode
*inode
= dentry
->d_inode
;
1457 struct task_struct
*task
;
1458 const struct cred
*cred
;
1460 generic_fillattr(inode
, stat
);
1465 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1467 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1468 task_dumpable(task
)) {
1469 cred
= __task_cred(task
);
1470 stat
->uid
= cred
->euid
;
1471 stat
->gid
= cred
->egid
;
1481 * Exceptional case: normally we are not allowed to unhash a busy
1482 * directory. In this case, however, we can do it - no aliasing problems
1483 * due to the way we treat inodes.
1485 * Rewrite the inode's ownerships here because the owning task may have
1486 * performed a setuid(), etc.
1488 * Before the /proc/pid/status file was created the only way to read
1489 * the effective uid of a /process was to stat /proc/pid. Reading
1490 * /proc/pid/status is slow enough that procps and other packages
1491 * kept stating /proc/pid. To keep the rules in /proc simple I have
1492 * made this apply to all per process world readable and executable
1495 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1497 struct inode
*inode
= dentry
->d_inode
;
1498 struct task_struct
*task
= get_proc_task(inode
);
1499 const struct cred
*cred
;
1502 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1503 task_dumpable(task
)) {
1505 cred
= __task_cred(task
);
1506 inode
->i_uid
= cred
->euid
;
1507 inode
->i_gid
= cred
->egid
;
1513 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1514 security_task_to_inode(task
, inode
);
1515 put_task_struct(task
);
1522 static int pid_delete_dentry(struct dentry
* dentry
)
1524 /* Is the task we represent dead?
1525 * If so, then don't put the dentry on the lru list,
1526 * kill it immediately.
1528 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1531 static const struct dentry_operations pid_dentry_operations
=
1533 .d_revalidate
= pid_revalidate
,
1534 .d_delete
= pid_delete_dentry
,
1539 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1540 struct task_struct
*, const void *);
1543 * Fill a directory entry.
1545 * If possible create the dcache entry and derive our inode number and
1546 * file type from dcache entry.
1548 * Since all of the proc inode numbers are dynamically generated, the inode
1549 * numbers do not exist until the inode is cache. This means creating the
1550 * the dcache entry in readdir is necessary to keep the inode numbers
1551 * reported by readdir in sync with the inode numbers reported
1554 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1555 char *name
, int len
,
1556 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1558 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1559 struct inode
*inode
;
1562 unsigned type
= DT_UNKNOWN
;
1566 qname
.hash
= full_name_hash(name
, len
);
1568 child
= d_lookup(dir
, &qname
);
1571 new = d_alloc(dir
, &qname
);
1573 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1580 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1581 goto end_instantiate
;
1582 inode
= child
->d_inode
;
1585 type
= inode
->i_mode
>> 12;
1590 ino
= find_inode_number(dir
, &qname
);
1593 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1596 static unsigned name_to_int(struct dentry
*dentry
)
1598 const char *name
= dentry
->d_name
.name
;
1599 int len
= dentry
->d_name
.len
;
1602 if (len
> 1 && *name
== '0')
1605 unsigned c
= *name
++ - '0';
1608 if (n
>= (~0U-9)/10)
1618 #define PROC_FDINFO_MAX 64
1620 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1622 struct task_struct
*task
= get_proc_task(inode
);
1623 struct files_struct
*files
= NULL
;
1625 int fd
= proc_fd(inode
);
1628 files
= get_files_struct(task
);
1629 put_task_struct(task
);
1633 * We are not taking a ref to the file structure, so we must
1636 spin_lock(&files
->file_lock
);
1637 file
= fcheck_files(files
, fd
);
1640 *path
= file
->f_path
;
1641 path_get(&file
->f_path
);
1644 snprintf(info
, PROC_FDINFO_MAX
,
1647 (long long) file
->f_pos
,
1649 spin_unlock(&files
->file_lock
);
1650 put_files_struct(files
);
1653 spin_unlock(&files
->file_lock
);
1654 put_files_struct(files
);
1659 static int proc_fd_link(struct inode
*inode
, struct path
*path
)
1661 return proc_fd_info(inode
, path
, NULL
);
1664 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1666 struct inode
*inode
= dentry
->d_inode
;
1667 struct task_struct
*task
= get_proc_task(inode
);
1668 int fd
= proc_fd(inode
);
1669 struct files_struct
*files
;
1670 const struct cred
*cred
;
1673 files
= get_files_struct(task
);
1676 if (fcheck_files(files
, fd
)) {
1678 put_files_struct(files
);
1679 if (task_dumpable(task
)) {
1681 cred
= __task_cred(task
);
1682 inode
->i_uid
= cred
->euid
;
1683 inode
->i_gid
= cred
->egid
;
1689 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1690 security_task_to_inode(task
, inode
);
1691 put_task_struct(task
);
1695 put_files_struct(files
);
1697 put_task_struct(task
);
1703 static const struct dentry_operations tid_fd_dentry_operations
=
1705 .d_revalidate
= tid_fd_revalidate
,
1706 .d_delete
= pid_delete_dentry
,
1709 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1710 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1712 unsigned fd
= *(const unsigned *)ptr
;
1714 struct files_struct
*files
;
1715 struct inode
*inode
;
1716 struct proc_inode
*ei
;
1717 struct dentry
*error
= ERR_PTR(-ENOENT
);
1719 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1724 files
= get_files_struct(task
);
1727 inode
->i_mode
= S_IFLNK
;
1730 * We are not taking a ref to the file structure, so we must
1733 spin_lock(&files
->file_lock
);
1734 file
= fcheck_files(files
, fd
);
1737 if (file
->f_mode
& FMODE_READ
)
1738 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1739 if (file
->f_mode
& FMODE_WRITE
)
1740 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1741 spin_unlock(&files
->file_lock
);
1742 put_files_struct(files
);
1744 inode
->i_op
= &proc_pid_link_inode_operations
;
1746 ei
->op
.proc_get_link
= proc_fd_link
;
1747 dentry
->d_op
= &tid_fd_dentry_operations
;
1748 d_add(dentry
, inode
);
1749 /* Close the race of the process dying before we return the dentry */
1750 if (tid_fd_revalidate(dentry
, NULL
))
1756 spin_unlock(&files
->file_lock
);
1757 put_files_struct(files
);
1763 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1764 struct dentry
*dentry
,
1765 instantiate_t instantiate
)
1767 struct task_struct
*task
= get_proc_task(dir
);
1768 unsigned fd
= name_to_int(dentry
);
1769 struct dentry
*result
= ERR_PTR(-ENOENT
);
1776 result
= instantiate(dir
, dentry
, task
, &fd
);
1778 put_task_struct(task
);
1783 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1784 filldir_t filldir
, instantiate_t instantiate
)
1786 struct dentry
*dentry
= filp
->f_path
.dentry
;
1787 struct inode
*inode
= dentry
->d_inode
;
1788 struct task_struct
*p
= get_proc_task(inode
);
1789 unsigned int fd
, ino
;
1791 struct files_struct
* files
;
1801 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1805 ino
= parent_ino(dentry
);
1806 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1810 files
= get_files_struct(p
);
1814 for (fd
= filp
->f_pos
-2;
1815 fd
< files_fdtable(files
)->max_fds
;
1816 fd
++, filp
->f_pos
++) {
1817 char name
[PROC_NUMBUF
];
1820 if (!fcheck_files(files
, fd
))
1824 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1825 if (proc_fill_cache(filp
, dirent
, filldir
,
1826 name
, len
, instantiate
,
1834 put_files_struct(files
);
1842 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1843 struct nameidata
*nd
)
1845 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1848 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1850 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1853 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1854 size_t len
, loff_t
*ppos
)
1856 char tmp
[PROC_FDINFO_MAX
];
1857 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1859 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1863 static const struct file_operations proc_fdinfo_file_operations
= {
1864 .open
= nonseekable_open
,
1865 .read
= proc_fdinfo_read
,
1868 static const struct file_operations proc_fd_operations
= {
1869 .read
= generic_read_dir
,
1870 .readdir
= proc_readfd
,
1874 * /proc/pid/fd needs a special permission handler so that a process can still
1875 * access /proc/self/fd after it has executed a setuid().
1877 static int proc_fd_permission(struct inode
*inode
, int mask
)
1881 rv
= generic_permission(inode
, mask
, NULL
);
1884 if (task_pid(current
) == proc_pid(inode
))
1890 * proc directories can do almost nothing..
1892 static const struct inode_operations proc_fd_inode_operations
= {
1893 .lookup
= proc_lookupfd
,
1894 .permission
= proc_fd_permission
,
1895 .setattr
= proc_setattr
,
1898 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
1899 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1901 unsigned fd
= *(unsigned *)ptr
;
1902 struct inode
*inode
;
1903 struct proc_inode
*ei
;
1904 struct dentry
*error
= ERR_PTR(-ENOENT
);
1906 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1911 inode
->i_mode
= S_IFREG
| S_IRUSR
;
1912 inode
->i_fop
= &proc_fdinfo_file_operations
;
1913 dentry
->d_op
= &tid_fd_dentry_operations
;
1914 d_add(dentry
, inode
);
1915 /* Close the race of the process dying before we return the dentry */
1916 if (tid_fd_revalidate(dentry
, NULL
))
1923 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
1924 struct dentry
*dentry
,
1925 struct nameidata
*nd
)
1927 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
1930 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
1932 return proc_readfd_common(filp
, dirent
, filldir
,
1933 proc_fdinfo_instantiate
);
1936 static const struct file_operations proc_fdinfo_operations
= {
1937 .read
= generic_read_dir
,
1938 .readdir
= proc_readfdinfo
,
1942 * proc directories can do almost nothing..
1944 static const struct inode_operations proc_fdinfo_inode_operations
= {
1945 .lookup
= proc_lookupfdinfo
,
1946 .setattr
= proc_setattr
,
1950 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
1951 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1953 const struct pid_entry
*p
= ptr
;
1954 struct inode
*inode
;
1955 struct proc_inode
*ei
;
1956 struct dentry
*error
= ERR_PTR(-ENOENT
);
1958 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1963 inode
->i_mode
= p
->mode
;
1964 if (S_ISDIR(inode
->i_mode
))
1965 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1967 inode
->i_op
= p
->iop
;
1969 inode
->i_fop
= p
->fop
;
1971 dentry
->d_op
= &pid_dentry_operations
;
1972 d_add(dentry
, inode
);
1973 /* Close the race of the process dying before we return the dentry */
1974 if (pid_revalidate(dentry
, NULL
))
1980 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1981 struct dentry
*dentry
,
1982 const struct pid_entry
*ents
,
1985 struct dentry
*error
;
1986 struct task_struct
*task
= get_proc_task(dir
);
1987 const struct pid_entry
*p
, *last
;
1989 error
= ERR_PTR(-ENOENT
);
1995 * Yes, it does not scale. And it should not. Don't add
1996 * new entries into /proc/<tgid>/ without very good reasons.
1998 last
= &ents
[nents
- 1];
1999 for (p
= ents
; p
<= last
; p
++) {
2000 if (p
->len
!= dentry
->d_name
.len
)
2002 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2008 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2010 put_task_struct(task
);
2015 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2016 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2018 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2019 proc_pident_instantiate
, task
, p
);
2022 static int proc_pident_readdir(struct file
*filp
,
2023 void *dirent
, filldir_t filldir
,
2024 const struct pid_entry
*ents
, unsigned int nents
)
2027 struct dentry
*dentry
= filp
->f_path
.dentry
;
2028 struct inode
*inode
= dentry
->d_inode
;
2029 struct task_struct
*task
= get_proc_task(inode
);
2030 const struct pid_entry
*p
, *last
;
2043 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2049 ino
= parent_ino(dentry
);
2050 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2062 last
= &ents
[nents
- 1];
2064 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2073 put_task_struct(task
);
2078 #ifdef CONFIG_SECURITY
2079 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2080 size_t count
, loff_t
*ppos
)
2082 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2085 struct task_struct
*task
= get_proc_task(inode
);
2090 length
= security_getprocattr(task
,
2091 (char*)file
->f_path
.dentry
->d_name
.name
,
2093 put_task_struct(task
);
2095 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2100 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2101 size_t count
, loff_t
*ppos
)
2103 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2106 struct task_struct
*task
= get_proc_task(inode
);
2111 if (count
> PAGE_SIZE
)
2114 /* No partial writes. */
2120 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2125 if (copy_from_user(page
, buf
, count
))
2128 /* Guard against adverse ptrace interaction */
2129 length
= mutex_lock_interruptible(&task
->cred_guard_mutex
);
2133 length
= security_setprocattr(task
,
2134 (char*)file
->f_path
.dentry
->d_name
.name
,
2135 (void*)page
, count
);
2136 mutex_unlock(&task
->cred_guard_mutex
);
2138 free_page((unsigned long) page
);
2140 put_task_struct(task
);
2145 static const struct file_operations proc_pid_attr_operations
= {
2146 .read
= proc_pid_attr_read
,
2147 .write
= proc_pid_attr_write
,
2150 static const struct pid_entry attr_dir_stuff
[] = {
2151 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2152 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2153 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2154 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2155 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2156 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2159 static int proc_attr_dir_readdir(struct file
* filp
,
2160 void * dirent
, filldir_t filldir
)
2162 return proc_pident_readdir(filp
,dirent
,filldir
,
2163 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2166 static const struct file_operations proc_attr_dir_operations
= {
2167 .read
= generic_read_dir
,
2168 .readdir
= proc_attr_dir_readdir
,
2171 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2172 struct dentry
*dentry
, struct nameidata
*nd
)
2174 return proc_pident_lookup(dir
, dentry
,
2175 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2178 static const struct inode_operations proc_attr_dir_inode_operations
= {
2179 .lookup
= proc_attr_dir_lookup
,
2180 .getattr
= pid_getattr
,
2181 .setattr
= proc_setattr
,
2186 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2187 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2188 size_t count
, loff_t
*ppos
)
2190 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2191 struct mm_struct
*mm
;
2192 char buffer
[PROC_NUMBUF
];
2200 mm
= get_task_mm(task
);
2202 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2203 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2204 MMF_DUMP_FILTER_SHIFT
));
2206 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2209 put_task_struct(task
);
2214 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2215 const char __user
*buf
,
2219 struct task_struct
*task
;
2220 struct mm_struct
*mm
;
2221 char buffer
[PROC_NUMBUF
], *end
;
2228 memset(buffer
, 0, sizeof(buffer
));
2229 if (count
> sizeof(buffer
) - 1)
2230 count
= sizeof(buffer
) - 1;
2231 if (copy_from_user(buffer
, buf
, count
))
2235 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2238 if (end
- buffer
== 0)
2242 task
= get_proc_task(file
->f_dentry
->d_inode
);
2247 mm
= get_task_mm(task
);
2251 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2253 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2255 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2260 put_task_struct(task
);
2265 static const struct file_operations proc_coredump_filter_operations
= {
2266 .read
= proc_coredump_filter_read
,
2267 .write
= proc_coredump_filter_write
,
2274 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2277 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2278 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2279 char tmp
[PROC_NUMBUF
];
2282 sprintf(tmp
, "%d", tgid
);
2283 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2286 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2288 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2289 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2290 char tmp
[PROC_NUMBUF
];
2292 return ERR_PTR(-ENOENT
);
2293 sprintf(tmp
, "%d", task_tgid_nr_ns(current
, ns
));
2294 return ERR_PTR(vfs_follow_link(nd
,tmp
));
2297 static const struct inode_operations proc_self_inode_operations
= {
2298 .readlink
= proc_self_readlink
,
2299 .follow_link
= proc_self_follow_link
,
2305 * These are the directory entries in the root directory of /proc
2306 * that properly belong to the /proc filesystem, as they describe
2307 * describe something that is process related.
2309 static const struct pid_entry proc_base_stuff
[] = {
2310 NOD("self", S_IFLNK
|S_IRWXUGO
,
2311 &proc_self_inode_operations
, NULL
, {}),
2315 * Exceptional case: normally we are not allowed to unhash a busy
2316 * directory. In this case, however, we can do it - no aliasing problems
2317 * due to the way we treat inodes.
2319 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
2321 struct inode
*inode
= dentry
->d_inode
;
2322 struct task_struct
*task
= get_proc_task(inode
);
2324 put_task_struct(task
);
2331 static const struct dentry_operations proc_base_dentry_operations
=
2333 .d_revalidate
= proc_base_revalidate
,
2334 .d_delete
= pid_delete_dentry
,
2337 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2338 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2340 const struct pid_entry
*p
= ptr
;
2341 struct inode
*inode
;
2342 struct proc_inode
*ei
;
2343 struct dentry
*error
= ERR_PTR(-EINVAL
);
2345 /* Allocate the inode */
2346 error
= ERR_PTR(-ENOMEM
);
2347 inode
= new_inode(dir
->i_sb
);
2351 /* Initialize the inode */
2353 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2356 * grab the reference to the task.
2358 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2362 inode
->i_mode
= p
->mode
;
2363 if (S_ISDIR(inode
->i_mode
))
2365 if (S_ISLNK(inode
->i_mode
))
2368 inode
->i_op
= p
->iop
;
2370 inode
->i_fop
= p
->fop
;
2372 dentry
->d_op
= &proc_base_dentry_operations
;
2373 d_add(dentry
, inode
);
2382 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2384 struct dentry
*error
;
2385 struct task_struct
*task
= get_proc_task(dir
);
2386 const struct pid_entry
*p
, *last
;
2388 error
= ERR_PTR(-ENOENT
);
2393 /* Lookup the directory entry */
2394 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2395 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2396 if (p
->len
!= dentry
->d_name
.len
)
2398 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2404 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2407 put_task_struct(task
);
2412 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2413 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2415 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2416 proc_base_instantiate
, task
, p
);
2419 #ifdef CONFIG_TASK_IO_ACCOUNTING
2420 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2422 struct task_io_accounting acct
= task
->ioac
;
2423 unsigned long flags
;
2425 if (whole
&& lock_task_sighand(task
, &flags
)) {
2426 struct task_struct
*t
= task
;
2428 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2429 while_each_thread(task
, t
)
2430 task_io_accounting_add(&acct
, &t
->ioac
);
2432 unlock_task_sighand(task
, &flags
);
2434 return sprintf(buffer
,
2439 "read_bytes: %llu\n"
2440 "write_bytes: %llu\n"
2441 "cancelled_write_bytes: %llu\n",
2442 (unsigned long long)acct
.rchar
,
2443 (unsigned long long)acct
.wchar
,
2444 (unsigned long long)acct
.syscr
,
2445 (unsigned long long)acct
.syscw
,
2446 (unsigned long long)acct
.read_bytes
,
2447 (unsigned long long)acct
.write_bytes
,
2448 (unsigned long long)acct
.cancelled_write_bytes
);
2451 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2453 return do_io_accounting(task
, buffer
, 0);
2456 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2458 return do_io_accounting(task
, buffer
, 1);
2460 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2462 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2463 struct pid
*pid
, struct task_struct
*task
)
2465 seq_printf(m
, "%08x\n", task
->personality
);
2472 static const struct file_operations proc_task_operations
;
2473 static const struct inode_operations proc_task_inode_operations
;
2475 static const struct pid_entry tgid_base_stuff
[] = {
2476 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2477 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2478 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2480 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2482 REG("environ", S_IRUSR
, proc_environ_operations
),
2483 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2484 ONE("status", S_IRUGO
, proc_pid_status
),
2485 ONE("personality", S_IRUSR
, proc_pid_personality
),
2486 INF("limits", S_IRUSR
, proc_pid_limits
),
2487 #ifdef CONFIG_SCHED_DEBUG
2488 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2490 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2491 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2493 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2494 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2495 ONE("statm", S_IRUGO
, proc_pid_statm
),
2496 REG("maps", S_IRUGO
, proc_maps_operations
),
2498 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2500 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2501 LNK("cwd", proc_cwd_link
),
2502 LNK("root", proc_root_link
),
2503 LNK("exe", proc_exe_link
),
2504 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2505 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2506 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2507 #ifdef CONFIG_PROC_PAGE_MONITOR
2508 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2509 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2510 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2512 #ifdef CONFIG_SECURITY
2513 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2515 #ifdef CONFIG_KALLSYMS
2516 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2518 #ifdef CONFIG_STACKTRACE
2519 ONE("stack", S_IRUSR
, proc_pid_stack
),
2521 #ifdef CONFIG_SCHEDSTATS
2522 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2524 #ifdef CONFIG_LATENCYTOP
2525 REG("latency", S_IRUGO
, proc_lstats_operations
),
2527 #ifdef CONFIG_PROC_PID_CPUSET
2528 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2530 #ifdef CONFIG_CGROUPS
2531 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2533 INF("oom_score", S_IRUGO
, proc_oom_score
),
2534 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2535 #ifdef CONFIG_AUDITSYSCALL
2536 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2537 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2539 #ifdef CONFIG_FAULT_INJECTION
2540 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2542 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2543 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2545 #ifdef CONFIG_TASK_IO_ACCOUNTING
2546 INF("io", S_IRUGO
, proc_tgid_io_accounting
),
2550 static int proc_tgid_base_readdir(struct file
* filp
,
2551 void * dirent
, filldir_t filldir
)
2553 return proc_pident_readdir(filp
,dirent
,filldir
,
2554 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2557 static const struct file_operations proc_tgid_base_operations
= {
2558 .read
= generic_read_dir
,
2559 .readdir
= proc_tgid_base_readdir
,
2562 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2563 return proc_pident_lookup(dir
, dentry
,
2564 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2567 static const struct inode_operations proc_tgid_base_inode_operations
= {
2568 .lookup
= proc_tgid_base_lookup
,
2569 .getattr
= pid_getattr
,
2570 .setattr
= proc_setattr
,
2573 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2575 struct dentry
*dentry
, *leader
, *dir
;
2576 char buf
[PROC_NUMBUF
];
2580 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2581 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2583 if (!(current
->flags
& PF_EXITING
))
2584 shrink_dcache_parent(dentry
);
2593 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2594 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2599 name
.len
= strlen(name
.name
);
2600 dir
= d_hash_and_lookup(leader
, &name
);
2602 goto out_put_leader
;
2605 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2606 dentry
= d_hash_and_lookup(dir
, &name
);
2608 shrink_dcache_parent(dentry
);
2621 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2622 * @task: task that should be flushed.
2624 * When flushing dentries from proc, one needs to flush them from global
2625 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2626 * in. This call is supposed to do all of this job.
2628 * Looks in the dcache for
2630 * /proc/@tgid/task/@pid
2631 * if either directory is present flushes it and all of it'ts children
2634 * It is safe and reasonable to cache /proc entries for a task until
2635 * that task exits. After that they just clog up the dcache with
2636 * useless entries, possibly causing useful dcache entries to be
2637 * flushed instead. This routine is proved to flush those useless
2638 * dcache entries at process exit time.
2640 * NOTE: This routine is just an optimization so it does not guarantee
2641 * that no dcache entries will exist at process exit time it
2642 * just makes it very unlikely that any will persist.
2645 void proc_flush_task(struct task_struct
*task
)
2648 struct pid
*pid
, *tgid
= NULL
;
2651 pid
= task_pid(task
);
2652 if (thread_group_leader(task
))
2653 tgid
= task_tgid(task
);
2655 for (i
= 0; i
<= pid
->level
; i
++) {
2656 upid
= &pid
->numbers
[i
];
2657 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2658 tgid
? tgid
->numbers
[i
].nr
: 0);
2661 upid
= &pid
->numbers
[pid
->level
];
2663 pid_ns_release_proc(upid
->ns
);
2666 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2667 struct dentry
* dentry
,
2668 struct task_struct
*task
, const void *ptr
)
2670 struct dentry
*error
= ERR_PTR(-ENOENT
);
2671 struct inode
*inode
;
2673 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2677 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2678 inode
->i_op
= &proc_tgid_base_inode_operations
;
2679 inode
->i_fop
= &proc_tgid_base_operations
;
2680 inode
->i_flags
|=S_IMMUTABLE
;
2682 inode
->i_nlink
= 2 + pid_entry_count_dirs(tgid_base_stuff
,
2683 ARRAY_SIZE(tgid_base_stuff
));
2685 dentry
->d_op
= &pid_dentry_operations
;
2687 d_add(dentry
, inode
);
2688 /* Close the race of the process dying before we return the dentry */
2689 if (pid_revalidate(dentry
, NULL
))
2695 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2697 struct dentry
*result
= ERR_PTR(-ENOENT
);
2698 struct task_struct
*task
;
2700 struct pid_namespace
*ns
;
2702 result
= proc_base_lookup(dir
, dentry
);
2703 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2706 tgid
= name_to_int(dentry
);
2710 ns
= dentry
->d_sb
->s_fs_info
;
2712 task
= find_task_by_pid_ns(tgid
, ns
);
2714 get_task_struct(task
);
2719 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2720 put_task_struct(task
);
2726 * Find the first task with tgid >= tgid
2731 struct task_struct
*task
;
2733 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
2738 put_task_struct(iter
.task
);
2742 pid
= find_ge_pid(iter
.tgid
, ns
);
2744 iter
.tgid
= pid_nr_ns(pid
, ns
);
2745 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
2746 /* What we to know is if the pid we have find is the
2747 * pid of a thread_group_leader. Testing for task
2748 * being a thread_group_leader is the obvious thing
2749 * todo but there is a window when it fails, due to
2750 * the pid transfer logic in de_thread.
2752 * So we perform the straight forward test of seeing
2753 * if the pid we have found is the pid of a thread
2754 * group leader, and don't worry if the task we have
2755 * found doesn't happen to be a thread group leader.
2756 * As we don't care in the case of readdir.
2758 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
2762 get_task_struct(iter
.task
);
2768 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2770 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2771 struct tgid_iter iter
)
2773 char name
[PROC_NUMBUF
];
2774 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
2775 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2776 proc_pid_instantiate
, iter
.task
, NULL
);
2779 /* for the /proc/ directory itself, after non-process stuff has been done */
2780 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2782 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2783 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2784 struct tgid_iter iter
;
2785 struct pid_namespace
*ns
;
2790 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2791 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2792 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2796 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2798 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
2799 for (iter
= next_tgid(ns
, iter
);
2801 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
2802 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
2803 if (proc_pid_fill_cache(filp
, dirent
, filldir
, iter
) < 0) {
2804 put_task_struct(iter
.task
);
2808 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2810 put_task_struct(reaper
);
2818 static const struct pid_entry tid_base_stuff
[] = {
2819 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2820 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fd_operations
),
2821 REG("environ", S_IRUSR
, proc_environ_operations
),
2822 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2823 ONE("status", S_IRUGO
, proc_pid_status
),
2824 ONE("personality", S_IRUSR
, proc_pid_personality
),
2825 INF("limits", S_IRUSR
, proc_pid_limits
),
2826 #ifdef CONFIG_SCHED_DEBUG
2827 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2829 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2830 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2832 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2833 ONE("stat", S_IRUGO
, proc_tid_stat
),
2834 ONE("statm", S_IRUGO
, proc_pid_statm
),
2835 REG("maps", S_IRUGO
, proc_maps_operations
),
2837 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2839 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2840 LNK("cwd", proc_cwd_link
),
2841 LNK("root", proc_root_link
),
2842 LNK("exe", proc_exe_link
),
2843 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2844 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2845 #ifdef CONFIG_PROC_PAGE_MONITOR
2846 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2847 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2848 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2850 #ifdef CONFIG_SECURITY
2851 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2853 #ifdef CONFIG_KALLSYMS
2854 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2856 #ifdef CONFIG_STACKTRACE
2857 ONE("stack", S_IRUSR
, proc_pid_stack
),
2859 #ifdef CONFIG_SCHEDSTATS
2860 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2862 #ifdef CONFIG_LATENCYTOP
2863 REG("latency", S_IRUGO
, proc_lstats_operations
),
2865 #ifdef CONFIG_PROC_PID_CPUSET
2866 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2868 #ifdef CONFIG_CGROUPS
2869 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2871 INF("oom_score", S_IRUGO
, proc_oom_score
),
2872 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2873 #ifdef CONFIG_AUDITSYSCALL
2874 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2875 REG("sessionid", S_IRUSR
, proc_sessionid_operations
),
2877 #ifdef CONFIG_FAULT_INJECTION
2878 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2880 #ifdef CONFIG_TASK_IO_ACCOUNTING
2881 INF("io", S_IRUGO
, proc_tid_io_accounting
),
2885 static int proc_tid_base_readdir(struct file
* filp
,
2886 void * dirent
, filldir_t filldir
)
2888 return proc_pident_readdir(filp
,dirent
,filldir
,
2889 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2892 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2893 return proc_pident_lookup(dir
, dentry
,
2894 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
2897 static const struct file_operations proc_tid_base_operations
= {
2898 .read
= generic_read_dir
,
2899 .readdir
= proc_tid_base_readdir
,
2902 static const struct inode_operations proc_tid_base_inode_operations
= {
2903 .lookup
= proc_tid_base_lookup
,
2904 .getattr
= pid_getattr
,
2905 .setattr
= proc_setattr
,
2908 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
2909 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2911 struct dentry
*error
= ERR_PTR(-ENOENT
);
2912 struct inode
*inode
;
2913 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2917 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2918 inode
->i_op
= &proc_tid_base_inode_operations
;
2919 inode
->i_fop
= &proc_tid_base_operations
;
2920 inode
->i_flags
|=S_IMMUTABLE
;
2922 inode
->i_nlink
= 2 + pid_entry_count_dirs(tid_base_stuff
,
2923 ARRAY_SIZE(tid_base_stuff
));
2925 dentry
->d_op
= &pid_dentry_operations
;
2927 d_add(dentry
, inode
);
2928 /* Close the race of the process dying before we return the dentry */
2929 if (pid_revalidate(dentry
, NULL
))
2935 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2937 struct dentry
*result
= ERR_PTR(-ENOENT
);
2938 struct task_struct
*task
;
2939 struct task_struct
*leader
= get_proc_task(dir
);
2941 struct pid_namespace
*ns
;
2946 tid
= name_to_int(dentry
);
2950 ns
= dentry
->d_sb
->s_fs_info
;
2952 task
= find_task_by_pid_ns(tid
, ns
);
2954 get_task_struct(task
);
2958 if (!same_thread_group(leader
, task
))
2961 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
2963 put_task_struct(task
);
2965 put_task_struct(leader
);
2971 * Find the first tid of a thread group to return to user space.
2973 * Usually this is just the thread group leader, but if the users
2974 * buffer was too small or there was a seek into the middle of the
2975 * directory we have more work todo.
2977 * In the case of a short read we start with find_task_by_pid.
2979 * In the case of a seek we start with the leader and walk nr
2982 static struct task_struct
*first_tid(struct task_struct
*leader
,
2983 int tid
, int nr
, struct pid_namespace
*ns
)
2985 struct task_struct
*pos
;
2988 /* Attempt to start with the pid of a thread */
2989 if (tid
&& (nr
> 0)) {
2990 pos
= find_task_by_pid_ns(tid
, ns
);
2991 if (pos
&& (pos
->group_leader
== leader
))
2995 /* If nr exceeds the number of threads there is nothing todo */
2997 if (nr
&& nr
>= get_nr_threads(leader
))
3000 /* If we haven't found our starting place yet start
3001 * with the leader and walk nr threads forward.
3003 for (pos
= leader
; nr
> 0; --nr
) {
3004 pos
= next_thread(pos
);
3005 if (pos
== leader
) {
3011 get_task_struct(pos
);
3018 * Find the next thread in the thread list.
3019 * Return NULL if there is an error or no next thread.
3021 * The reference to the input task_struct is released.
3023 static struct task_struct
*next_tid(struct task_struct
*start
)
3025 struct task_struct
*pos
= NULL
;
3027 if (pid_alive(start
)) {
3028 pos
= next_thread(start
);
3029 if (thread_group_leader(pos
))
3032 get_task_struct(pos
);
3035 put_task_struct(start
);
3039 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3040 struct task_struct
*task
, int tid
)
3042 char name
[PROC_NUMBUF
];
3043 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3044 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3045 proc_task_instantiate
, task
, NULL
);
3048 /* for the /proc/TGID/task/ directories */
3049 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3051 struct dentry
*dentry
= filp
->f_path
.dentry
;
3052 struct inode
*inode
= dentry
->d_inode
;
3053 struct task_struct
*leader
= NULL
;
3054 struct task_struct
*task
;
3055 int retval
= -ENOENT
;
3058 struct pid_namespace
*ns
;
3060 task
= get_proc_task(inode
);
3064 if (pid_alive(task
)) {
3065 leader
= task
->group_leader
;
3066 get_task_struct(leader
);
3069 put_task_struct(task
);
3074 switch ((unsigned long)filp
->f_pos
) {
3077 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3082 ino
= parent_ino(dentry
);
3083 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3089 /* f_version caches the tgid value that the last readdir call couldn't
3090 * return. lseek aka telldir automagically resets f_version to 0.
3092 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3093 tid
= (int)filp
->f_version
;
3094 filp
->f_version
= 0;
3095 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3097 task
= next_tid(task
), filp
->f_pos
++) {
3098 tid
= task_pid_nr_ns(task
, ns
);
3099 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3100 /* returning this tgid failed, save it as the first
3101 * pid for the next readir call */
3102 filp
->f_version
= (u64
)tid
;
3103 put_task_struct(task
);
3108 put_task_struct(leader
);
3113 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3115 struct inode
*inode
= dentry
->d_inode
;
3116 struct task_struct
*p
= get_proc_task(inode
);
3117 generic_fillattr(inode
, stat
);
3120 stat
->nlink
+= get_nr_threads(p
);
3127 static const struct inode_operations proc_task_inode_operations
= {
3128 .lookup
= proc_task_lookup
,
3129 .getattr
= proc_task_getattr
,
3130 .setattr
= proc_setattr
,
3133 static const struct file_operations proc_task_operations
= {
3134 .read
= generic_read_dir
,
3135 .readdir
= proc_task_readdir
,