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>
86 * Implementing inode permission operations in /proc is almost
87 * certainly an error. Permission checks need to happen during
88 * each system call not at open time. The reason is that most of
89 * what we wish to check for permissions in /proc varies at runtime.
91 * The classic example of a problem is opening file descriptors
92 * in /proc for a task before it execs a suid executable.
99 const struct inode_operations
*iop
;
100 const struct file_operations
*fop
;
104 #define NOD(NAME, MODE, IOP, FOP, OP) { \
106 .len = sizeof(NAME) - 1, \
113 #define DIR(NAME, MODE, iops, fops) \
114 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
115 #define LNK(NAME, get_link) \
116 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
117 &proc_pid_link_inode_operations, NULL, \
118 { .proc_get_link = get_link } )
119 #define REG(NAME, MODE, fops) \
120 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
121 #define INF(NAME, MODE, read) \
122 NOD(NAME, (S_IFREG|(MODE)), \
123 NULL, &proc_info_file_operations, \
124 { .proc_read = read } )
125 #define ONE(NAME, MODE, show) \
126 NOD(NAME, (S_IFREG|(MODE)), \
127 NULL, &proc_single_file_operations, \
128 { .proc_show = show } )
131 * Count the number of hardlinks for the pid_entry table, excluding the .
134 static unsigned int pid_entry_count_dirs(const struct pid_entry
*entries
,
141 for (i
= 0; i
< n
; ++i
) {
142 if (S_ISDIR(entries
[i
].mode
))
149 static int get_fs_path(struct task_struct
*task
, struct path
*path
, bool root
)
151 struct fs_struct
*fs
;
152 int result
= -ENOENT
;
157 read_lock(&fs
->lock
);
158 *path
= root
? fs
->root
: fs
->pwd
;
160 read_unlock(&fs
->lock
);
167 static int get_nr_threads(struct task_struct
*tsk
)
172 if (lock_task_sighand(tsk
, &flags
)) {
173 count
= atomic_read(&tsk
->signal
->count
);
174 unlock_task_sighand(tsk
, &flags
);
179 static int proc_cwd_link(struct inode
*inode
, struct path
*path
)
181 struct task_struct
*task
= get_proc_task(inode
);
182 int result
= -ENOENT
;
185 result
= get_fs_path(task
, path
, 0);
186 put_task_struct(task
);
191 static int proc_root_link(struct inode
*inode
, struct path
*path
)
193 struct task_struct
*task
= get_proc_task(inode
);
194 int result
= -ENOENT
;
197 result
= get_fs_path(task
, path
, 1);
198 put_task_struct(task
);
204 * Return zero if current may access user memory in @task, -error if not.
206 static int check_mem_permission(struct task_struct
*task
)
209 * A task can always look at itself, in case it chooses
210 * to use system calls instead of load instructions.
216 * If current is actively ptrace'ing, and would also be
217 * permitted to freshly attach with ptrace now, permit it.
219 if (task_is_stopped_or_traced(task
)) {
222 match
= (tracehook_tracer_task(task
) == current
);
224 if (match
&& ptrace_may_access(task
, PTRACE_MODE_ATTACH
))
229 * Noone else is allowed.
234 struct mm_struct
*mm_for_maps(struct task_struct
*task
)
236 struct mm_struct
*mm
= get_task_mm(task
);
239 down_read(&mm
->mmap_sem
);
243 if (task
->mm
!= current
->mm
&&
244 __ptrace_may_access(task
, PTRACE_MODE_READ
) < 0)
250 up_read(&mm
->mmap_sem
);
255 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
259 struct mm_struct
*mm
= get_task_mm(task
);
263 goto out_mm
; /* Shh! No looking before we're done */
265 len
= mm
->arg_end
- mm
->arg_start
;
270 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
272 // If the nul at the end of args has been overwritten, then
273 // assume application is using setproctitle(3).
274 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
275 len
= strnlen(buffer
, res
);
279 len
= mm
->env_end
- mm
->env_start
;
280 if (len
> PAGE_SIZE
- res
)
281 len
= PAGE_SIZE
- res
;
282 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
283 res
= strnlen(buffer
, res
);
292 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
295 struct mm_struct
*mm
= get_task_mm(task
);
297 unsigned int nwords
= 0;
300 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
301 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
304 memcpy(buffer
, mm
->saved_auxv
, res
);
311 #ifdef CONFIG_KALLSYMS
313 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
314 * Returns the resolved symbol. If that fails, simply return the address.
316 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
319 char symname
[KSYM_NAME_LEN
];
321 wchan
= get_wchan(task
);
323 if (lookup_symbol_name(wchan
, symname
) < 0)
324 return sprintf(buffer
, "%lu", wchan
);
326 return sprintf(buffer
, "%s", symname
);
328 #endif /* CONFIG_KALLSYMS */
330 #ifdef CONFIG_STACKTRACE
332 #define MAX_STACK_TRACE_DEPTH 64
334 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
335 struct pid
*pid
, struct task_struct
*task
)
337 struct stack_trace trace
;
338 unsigned long *entries
;
341 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
345 trace
.nr_entries
= 0;
346 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
347 trace
.entries
= entries
;
349 save_stack_trace_tsk(task
, &trace
);
351 for (i
= 0; i
< trace
.nr_entries
; i
++) {
352 seq_printf(m
, "[<%p>] %pS\n",
353 (void *)entries
[i
], (void *)entries
[i
]);
361 #ifdef CONFIG_SCHEDSTATS
363 * Provides /proc/PID/schedstat
365 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
367 return sprintf(buffer
, "%llu %llu %lu\n",
368 (unsigned long long)task
->se
.sum_exec_runtime
,
369 (unsigned long long)task
->sched_info
.run_delay
,
370 task
->sched_info
.pcount
);
374 #ifdef CONFIG_LATENCYTOP
375 static int lstats_show_proc(struct seq_file
*m
, void *v
)
378 struct inode
*inode
= m
->private;
379 struct task_struct
*task
= get_proc_task(inode
);
383 seq_puts(m
, "Latency Top version : v0.1\n");
384 for (i
= 0; i
< 32; i
++) {
385 if (task
->latency_record
[i
].backtrace
[0]) {
387 seq_printf(m
, "%i %li %li ",
388 task
->latency_record
[i
].count
,
389 task
->latency_record
[i
].time
,
390 task
->latency_record
[i
].max
);
391 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
392 char sym
[KSYM_SYMBOL_LEN
];
394 if (!task
->latency_record
[i
].backtrace
[q
])
396 if (task
->latency_record
[i
].backtrace
[q
] == ULONG_MAX
)
398 sprint_symbol(sym
, task
->latency_record
[i
].backtrace
[q
]);
399 c
= strchr(sym
, '+');
402 seq_printf(m
, "%s ", sym
);
408 put_task_struct(task
);
412 static int lstats_open(struct inode
*inode
, struct file
*file
)
414 return single_open(file
, lstats_show_proc
, inode
);
417 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
418 size_t count
, loff_t
*offs
)
420 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
424 clear_all_latency_tracing(task
);
425 put_task_struct(task
);
430 static const struct file_operations proc_lstats_operations
= {
433 .write
= lstats_write
,
435 .release
= single_release
,
440 /* The badness from the OOM killer */
441 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
442 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
444 unsigned long points
;
445 struct timespec uptime
;
447 do_posix_clock_monotonic_gettime(&uptime
);
448 read_lock(&tasklist_lock
);
449 points
= badness(task
, uptime
.tv_sec
);
450 read_unlock(&tasklist_lock
);
451 return sprintf(buffer
, "%lu\n", points
);
459 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
460 [RLIMIT_CPU
] = {"Max cpu time", "ms"},
461 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
462 [RLIMIT_DATA
] = {"Max data size", "bytes"},
463 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
464 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
465 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
466 [RLIMIT_NPROC
] = {"Max processes", "processes"},
467 [RLIMIT_NOFILE
] = {"Max open files", "files"},
468 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
469 [RLIMIT_AS
] = {"Max address space", "bytes"},
470 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
471 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
472 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
473 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
474 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
475 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
478 /* Display limits for a process */
479 static int proc_pid_limits(struct task_struct
*task
, char *buffer
)
484 char *bufptr
= buffer
;
486 struct rlimit rlim
[RLIM_NLIMITS
];
488 if (!lock_task_sighand(task
, &flags
))
490 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
491 unlock_task_sighand(task
, &flags
);
494 * print the file header
496 count
+= sprintf(&bufptr
[count
], "%-25s %-20s %-20s %-10s\n",
497 "Limit", "Soft Limit", "Hard Limit", "Units");
499 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
500 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
501 count
+= sprintf(&bufptr
[count
], "%-25s %-20s ",
502 lnames
[i
].name
, "unlimited");
504 count
+= sprintf(&bufptr
[count
], "%-25s %-20lu ",
505 lnames
[i
].name
, rlim
[i
].rlim_cur
);
507 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
508 count
+= sprintf(&bufptr
[count
], "%-20s ", "unlimited");
510 count
+= sprintf(&bufptr
[count
], "%-20lu ",
514 count
+= sprintf(&bufptr
[count
], "%-10s\n",
517 count
+= sprintf(&bufptr
[count
], "\n");
523 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
524 static int proc_pid_syscall(struct task_struct
*task
, char *buffer
)
527 unsigned long args
[6], sp
, pc
;
529 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
530 return sprintf(buffer
, "running\n");
533 return sprintf(buffer
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
535 return sprintf(buffer
,
536 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
538 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
541 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
543 /************************************************************************/
544 /* Here the fs part begins */
545 /************************************************************************/
547 /* permission checks */
548 static int proc_fd_access_allowed(struct inode
*inode
)
550 struct task_struct
*task
;
552 /* Allow access to a task's file descriptors if it is us or we
553 * may use ptrace attach to the process and find out that
556 task
= get_proc_task(inode
);
558 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ
);
559 put_task_struct(task
);
564 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
567 struct inode
*inode
= dentry
->d_inode
;
569 if (attr
->ia_valid
& ATTR_MODE
)
572 error
= inode_change_ok(inode
, attr
);
574 error
= inode_setattr(inode
, attr
);
578 static const struct inode_operations proc_def_inode_operations
= {
579 .setattr
= proc_setattr
,
582 static int mounts_open_common(struct inode
*inode
, struct file
*file
,
583 const struct seq_operations
*op
)
585 struct task_struct
*task
= get_proc_task(inode
);
587 struct mnt_namespace
*ns
= NULL
;
589 struct proc_mounts
*p
;
594 nsp
= task_nsproxy(task
);
601 if (ns
&& get_fs_path(task
, &root
, 1) == 0)
603 put_task_struct(task
);
612 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
616 file
->private_data
= &p
->m
;
617 ret
= seq_open(file
, op
);
624 p
->event
= ns
->event
;
638 static int mounts_release(struct inode
*inode
, struct file
*file
)
640 struct proc_mounts
*p
= file
->private_data
;
643 return seq_release(inode
, file
);
646 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
648 struct proc_mounts
*p
= file
->private_data
;
649 struct mnt_namespace
*ns
= p
->ns
;
652 poll_wait(file
, &ns
->poll
, wait
);
654 spin_lock(&vfsmount_lock
);
655 if (p
->event
!= ns
->event
) {
656 p
->event
= ns
->event
;
659 spin_unlock(&vfsmount_lock
);
664 static int mounts_open(struct inode
*inode
, struct file
*file
)
666 return mounts_open_common(inode
, file
, &mounts_op
);
669 static const struct file_operations proc_mounts_operations
= {
673 .release
= mounts_release
,
677 static int mountinfo_open(struct inode
*inode
, struct file
*file
)
679 return mounts_open_common(inode
, file
, &mountinfo_op
);
682 static const struct file_operations proc_mountinfo_operations
= {
683 .open
= mountinfo_open
,
686 .release
= mounts_release
,
690 static int mountstats_open(struct inode
*inode
, struct file
*file
)
692 return mounts_open_common(inode
, file
, &mountstats_op
);
695 static const struct file_operations proc_mountstats_operations
= {
696 .open
= mountstats_open
,
699 .release
= mounts_release
,
702 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
704 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
705 size_t count
, loff_t
*ppos
)
707 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
710 struct task_struct
*task
= get_proc_task(inode
);
716 if (count
> PROC_BLOCK_SIZE
)
717 count
= PROC_BLOCK_SIZE
;
720 if (!(page
= __get_free_page(GFP_TEMPORARY
)))
723 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
726 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
729 put_task_struct(task
);
734 static const struct file_operations proc_info_file_operations
= {
735 .read
= proc_info_read
,
738 static int proc_single_show(struct seq_file
*m
, void *v
)
740 struct inode
*inode
= m
->private;
741 struct pid_namespace
*ns
;
743 struct task_struct
*task
;
746 ns
= inode
->i_sb
->s_fs_info
;
747 pid
= proc_pid(inode
);
748 task
= get_pid_task(pid
, PIDTYPE_PID
);
752 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
754 put_task_struct(task
);
758 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
761 ret
= single_open(filp
, proc_single_show
, NULL
);
763 struct seq_file
*m
= filp
->private_data
;
770 static const struct file_operations proc_single_file_operations
= {
771 .open
= proc_single_open
,
774 .release
= single_release
,
777 static int mem_open(struct inode
* inode
, struct file
* file
)
779 file
->private_data
= (void*)((long)current
->self_exec_id
);
783 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
784 size_t count
, loff_t
*ppos
)
786 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
788 unsigned long src
= *ppos
;
790 struct mm_struct
*mm
;
795 if (check_mem_permission(task
))
799 page
= (char *)__get_free_page(GFP_TEMPORARY
);
805 mm
= get_task_mm(task
);
811 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
817 int this_len
, retval
;
819 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
820 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
821 if (!retval
|| check_mem_permission(task
)) {
827 if (copy_to_user(buf
, page
, retval
)) {
842 free_page((unsigned long) page
);
844 put_task_struct(task
);
849 #define mem_write NULL
852 /* This is a security hazard */
853 static ssize_t
mem_write(struct file
* file
, const char __user
*buf
,
854 size_t count
, loff_t
*ppos
)
858 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
859 unsigned long dst
= *ppos
;
865 if (check_mem_permission(task
))
869 page
= (char *)__get_free_page(GFP_TEMPORARY
);
875 int this_len
, retval
;
877 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
878 if (copy_from_user(page
, buf
, this_len
)) {
882 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
894 free_page((unsigned long) page
);
896 put_task_struct(task
);
902 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
906 file
->f_pos
= offset
;
909 file
->f_pos
+= offset
;
914 force_successful_syscall_return();
918 static const struct file_operations proc_mem_operations
= {
925 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
926 size_t count
, loff_t
*ppos
)
928 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
930 unsigned long src
= *ppos
;
932 struct mm_struct
*mm
;
937 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
941 page
= (char *)__get_free_page(GFP_TEMPORARY
);
947 mm
= get_task_mm(task
);
952 int this_len
, retval
, max_len
;
954 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
959 max_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
960 this_len
= (this_len
> max_len
) ? max_len
: this_len
;
962 retval
= access_process_vm(task
, (mm
->env_start
+ src
),
970 if (copy_to_user(buf
, page
, retval
)) {
984 free_page((unsigned long) page
);
986 put_task_struct(task
);
991 static const struct file_operations proc_environ_operations
= {
992 .read
= environ_read
,
995 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
996 size_t count
, loff_t
*ppos
)
998 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
999 char buffer
[PROC_NUMBUF
];
1005 oom_adjust
= task
->oomkilladj
;
1006 put_task_struct(task
);
1008 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
1010 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1013 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
1014 size_t count
, loff_t
*ppos
)
1016 struct task_struct
*task
;
1017 char buffer
[PROC_NUMBUF
], *end
;
1020 memset(buffer
, 0, sizeof(buffer
));
1021 if (count
> sizeof(buffer
) - 1)
1022 count
= sizeof(buffer
) - 1;
1023 if (copy_from_user(buffer
, buf
, count
))
1025 oom_adjust
= simple_strtol(buffer
, &end
, 0);
1026 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
1027 oom_adjust
!= OOM_DISABLE
)
1031 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1034 if (oom_adjust
< task
->oomkilladj
&& !capable(CAP_SYS_RESOURCE
)) {
1035 put_task_struct(task
);
1038 task
->oomkilladj
= oom_adjust
;
1039 put_task_struct(task
);
1040 if (end
- buffer
== 0)
1042 return end
- buffer
;
1045 static const struct file_operations proc_oom_adjust_operations
= {
1046 .read
= oom_adjust_read
,
1047 .write
= oom_adjust_write
,
1050 #ifdef CONFIG_AUDITSYSCALL
1051 #define TMPBUFLEN 21
1052 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1053 size_t count
, loff_t
*ppos
)
1055 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1056 struct task_struct
*task
= get_proc_task(inode
);
1058 char tmpbuf
[TMPBUFLEN
];
1062 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1063 audit_get_loginuid(task
));
1064 put_task_struct(task
);
1065 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1068 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1069 size_t count
, loff_t
*ppos
)
1071 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1076 if (!capable(CAP_AUDIT_CONTROL
))
1079 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
1082 if (count
>= PAGE_SIZE
)
1083 count
= PAGE_SIZE
- 1;
1086 /* No partial writes. */
1089 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1093 if (copy_from_user(page
, buf
, count
))
1097 loginuid
= simple_strtoul(page
, &tmp
, 10);
1103 length
= audit_set_loginuid(current
, loginuid
);
1104 if (likely(length
== 0))
1108 free_page((unsigned long) page
);
1112 static const struct file_operations proc_loginuid_operations
= {
1113 .read
= proc_loginuid_read
,
1114 .write
= proc_loginuid_write
,
1117 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1118 size_t count
, loff_t
*ppos
)
1120 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1121 struct task_struct
*task
= get_proc_task(inode
);
1123 char tmpbuf
[TMPBUFLEN
];
1127 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1128 audit_get_sessionid(task
));
1129 put_task_struct(task
);
1130 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1133 static const struct file_operations proc_sessionid_operations
= {
1134 .read
= proc_sessionid_read
,
1138 #ifdef CONFIG_FAULT_INJECTION
1139 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1140 size_t count
, loff_t
*ppos
)
1142 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1143 char buffer
[PROC_NUMBUF
];
1149 make_it_fail
= task
->make_it_fail
;
1150 put_task_struct(task
);
1152 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1154 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1157 static ssize_t
proc_fault_inject_write(struct file
* file
,
1158 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1160 struct task_struct
*task
;
1161 char buffer
[PROC_NUMBUF
], *end
;
1164 if (!capable(CAP_SYS_RESOURCE
))
1166 memset(buffer
, 0, sizeof(buffer
));
1167 if (count
> sizeof(buffer
) - 1)
1168 count
= sizeof(buffer
) - 1;
1169 if (copy_from_user(buffer
, buf
, count
))
1171 make_it_fail
= simple_strtol(buffer
, &end
, 0);
1174 task
= get_proc_task(file
->f_dentry
->d_inode
);
1177 task
->make_it_fail
= make_it_fail
;
1178 put_task_struct(task
);
1179 if (end
- buffer
== 0)
1181 return end
- buffer
;
1184 static const struct file_operations proc_fault_inject_operations
= {
1185 .read
= proc_fault_inject_read
,
1186 .write
= proc_fault_inject_write
,
1191 #ifdef CONFIG_SCHED_DEBUG
1193 * Print out various scheduling related per-task fields:
1195 static int sched_show(struct seq_file
*m
, void *v
)
1197 struct inode
*inode
= m
->private;
1198 struct task_struct
*p
;
1200 p
= get_proc_task(inode
);
1203 proc_sched_show_task(p
, m
);
1211 sched_write(struct file
*file
, const char __user
*buf
,
1212 size_t count
, loff_t
*offset
)
1214 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1215 struct task_struct
*p
;
1217 p
= get_proc_task(inode
);
1220 proc_sched_set_task(p
);
1227 static int sched_open(struct inode
*inode
, struct file
*filp
)
1231 ret
= single_open(filp
, sched_show
, NULL
);
1233 struct seq_file
*m
= filp
->private_data
;
1240 static const struct file_operations proc_pid_sched_operations
= {
1243 .write
= sched_write
,
1244 .llseek
= seq_lseek
,
1245 .release
= single_release
,
1251 * We added or removed a vma mapping the executable. The vmas are only mapped
1252 * during exec and are not mapped with the mmap system call.
1253 * Callers must hold down_write() on the mm's mmap_sem for these
1255 void added_exe_file_vma(struct mm_struct
*mm
)
1257 mm
->num_exe_file_vmas
++;
1260 void removed_exe_file_vma(struct mm_struct
*mm
)
1262 mm
->num_exe_file_vmas
--;
1263 if ((mm
->num_exe_file_vmas
== 0) && mm
->exe_file
){
1265 mm
->exe_file
= NULL
;
1270 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
1273 get_file(new_exe_file
);
1276 mm
->exe_file
= new_exe_file
;
1277 mm
->num_exe_file_vmas
= 0;
1280 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
1282 struct file
*exe_file
;
1284 /* We need mmap_sem to protect against races with removal of
1285 * VM_EXECUTABLE vmas */
1286 down_read(&mm
->mmap_sem
);
1287 exe_file
= mm
->exe_file
;
1290 up_read(&mm
->mmap_sem
);
1294 void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
1296 /* It's safe to write the exe_file pointer without exe_file_lock because
1297 * this is called during fork when the task is not yet in /proc */
1298 newmm
->exe_file
= get_mm_exe_file(oldmm
);
1301 static int proc_exe_link(struct inode
*inode
, struct path
*exe_path
)
1303 struct task_struct
*task
;
1304 struct mm_struct
*mm
;
1305 struct file
*exe_file
;
1307 task
= get_proc_task(inode
);
1310 mm
= get_task_mm(task
);
1311 put_task_struct(task
);
1314 exe_file
= get_mm_exe_file(mm
);
1317 *exe_path
= exe_file
->f_path
;
1318 path_get(&exe_file
->f_path
);
1325 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1327 struct inode
*inode
= dentry
->d_inode
;
1328 int error
= -EACCES
;
1330 /* We don't need a base pointer in the /proc filesystem */
1331 path_put(&nd
->path
);
1333 /* Are we allowed to snoop on the tasks file descriptors? */
1334 if (!proc_fd_access_allowed(inode
))
1337 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->path
);
1338 nd
->last_type
= LAST_BIND
;
1340 return ERR_PTR(error
);
1343 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1345 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1352 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1353 len
= PTR_ERR(pathname
);
1354 if (IS_ERR(pathname
))
1356 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1360 if (copy_to_user(buffer
, pathname
, len
))
1363 free_page((unsigned long)tmp
);
1367 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1369 int error
= -EACCES
;
1370 struct inode
*inode
= dentry
->d_inode
;
1373 /* Are we allowed to snoop on the tasks file descriptors? */
1374 if (!proc_fd_access_allowed(inode
))
1377 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &path
);
1381 error
= do_proc_readlink(&path
, buffer
, buflen
);
1387 static const struct inode_operations proc_pid_link_inode_operations
= {
1388 .readlink
= proc_pid_readlink
,
1389 .follow_link
= proc_pid_follow_link
,
1390 .setattr
= proc_setattr
,
1394 /* building an inode */
1396 static int task_dumpable(struct task_struct
*task
)
1399 struct mm_struct
*mm
;
1404 dumpable
= get_dumpable(mm
);
1412 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1414 struct inode
* inode
;
1415 struct proc_inode
*ei
;
1416 const struct cred
*cred
;
1418 /* We need a new inode */
1420 inode
= new_inode(sb
);
1426 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1427 inode
->i_op
= &proc_def_inode_operations
;
1430 * grab the reference to task.
1432 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1436 if (task_dumpable(task
)) {
1438 cred
= __task_cred(task
);
1439 inode
->i_uid
= cred
->euid
;
1440 inode
->i_gid
= cred
->egid
;
1443 security_task_to_inode(task
, inode
);
1453 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1455 struct inode
*inode
= dentry
->d_inode
;
1456 struct task_struct
*task
;
1457 const struct cred
*cred
;
1459 generic_fillattr(inode
, stat
);
1464 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1466 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1467 task_dumpable(task
)) {
1468 cred
= __task_cred(task
);
1469 stat
->uid
= cred
->euid
;
1470 stat
->gid
= cred
->egid
;
1480 * Exceptional case: normally we are not allowed to unhash a busy
1481 * directory. In this case, however, we can do it - no aliasing problems
1482 * due to the way we treat inodes.
1484 * Rewrite the inode's ownerships here because the owning task may have
1485 * performed a setuid(), etc.
1487 * Before the /proc/pid/status file was created the only way to read
1488 * the effective uid of a /process was to stat /proc/pid. Reading
1489 * /proc/pid/status is slow enough that procps and other packages
1490 * kept stating /proc/pid. To keep the rules in /proc simple I have
1491 * made this apply to all per process world readable and executable
1494 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1496 struct inode
*inode
= dentry
->d_inode
;
1497 struct task_struct
*task
= get_proc_task(inode
);
1498 const struct cred
*cred
;
1501 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1502 task_dumpable(task
)) {
1504 cred
= __task_cred(task
);
1505 inode
->i_uid
= cred
->euid
;
1506 inode
->i_gid
= cred
->egid
;
1512 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1513 security_task_to_inode(task
, inode
);
1514 put_task_struct(task
);
1521 static int pid_delete_dentry(struct dentry
* dentry
)
1523 /* Is the task we represent dead?
1524 * If so, then don't put the dentry on the lru list,
1525 * kill it immediately.
1527 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1530 static const struct dentry_operations pid_dentry_operations
=
1532 .d_revalidate
= pid_revalidate
,
1533 .d_delete
= pid_delete_dentry
,
1538 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1539 struct task_struct
*, const void *);
1542 * Fill a directory entry.
1544 * If possible create the dcache entry and derive our inode number and
1545 * file type from dcache entry.
1547 * Since all of the proc inode numbers are dynamically generated, the inode
1548 * numbers do not exist until the inode is cache. This means creating the
1549 * the dcache entry in readdir is necessary to keep the inode numbers
1550 * reported by readdir in sync with the inode numbers reported
1553 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1554 char *name
, int len
,
1555 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1557 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1558 struct inode
*inode
;
1561 unsigned type
= DT_UNKNOWN
;
1565 qname
.hash
= full_name_hash(name
, len
);
1567 child
= d_lookup(dir
, &qname
);
1570 new = d_alloc(dir
, &qname
);
1572 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1579 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1580 goto end_instantiate
;
1581 inode
= child
->d_inode
;
1584 type
= inode
->i_mode
>> 12;
1589 ino
= find_inode_number(dir
, &qname
);
1592 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1595 static unsigned name_to_int(struct dentry
*dentry
)
1597 const char *name
= dentry
->d_name
.name
;
1598 int len
= dentry
->d_name
.len
;
1601 if (len
> 1 && *name
== '0')
1604 unsigned c
= *name
++ - '0';
1607 if (n
>= (~0U-9)/10)
1617 #define PROC_FDINFO_MAX 64
1619 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1621 struct task_struct
*task
= get_proc_task(inode
);
1622 struct files_struct
*files
= NULL
;
1624 int fd
= proc_fd(inode
);
1627 files
= get_files_struct(task
);
1628 put_task_struct(task
);
1632 * We are not taking a ref to the file structure, so we must
1635 spin_lock(&files
->file_lock
);
1636 file
= fcheck_files(files
, fd
);
1639 *path
= file
->f_path
;
1640 path_get(&file
->f_path
);
1643 snprintf(info
, PROC_FDINFO_MAX
,
1646 (long long) file
->f_pos
,
1648 spin_unlock(&files
->file_lock
);
1649 put_files_struct(files
);
1652 spin_unlock(&files
->file_lock
);
1653 put_files_struct(files
);
1658 static int proc_fd_link(struct inode
*inode
, struct path
*path
)
1660 return proc_fd_info(inode
, path
, NULL
);
1663 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1665 struct inode
*inode
= dentry
->d_inode
;
1666 struct task_struct
*task
= get_proc_task(inode
);
1667 int fd
= proc_fd(inode
);
1668 struct files_struct
*files
;
1669 const struct cred
*cred
;
1672 files
= get_files_struct(task
);
1675 if (fcheck_files(files
, fd
)) {
1677 put_files_struct(files
);
1678 if (task_dumpable(task
)) {
1680 cred
= __task_cred(task
);
1681 inode
->i_uid
= cred
->euid
;
1682 inode
->i_gid
= cred
->egid
;
1688 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1689 security_task_to_inode(task
, inode
);
1690 put_task_struct(task
);
1694 put_files_struct(files
);
1696 put_task_struct(task
);
1702 static const struct dentry_operations tid_fd_dentry_operations
=
1704 .d_revalidate
= tid_fd_revalidate
,
1705 .d_delete
= pid_delete_dentry
,
1708 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1709 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1711 unsigned fd
= *(const unsigned *)ptr
;
1713 struct files_struct
*files
;
1714 struct inode
*inode
;
1715 struct proc_inode
*ei
;
1716 struct dentry
*error
= ERR_PTR(-ENOENT
);
1718 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1723 files
= get_files_struct(task
);
1726 inode
->i_mode
= S_IFLNK
;
1729 * We are not taking a ref to the file structure, so we must
1732 spin_lock(&files
->file_lock
);
1733 file
= fcheck_files(files
, fd
);
1736 if (file
->f_mode
& FMODE_READ
)
1737 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1738 if (file
->f_mode
& FMODE_WRITE
)
1739 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1740 spin_unlock(&files
->file_lock
);
1741 put_files_struct(files
);
1743 inode
->i_op
= &proc_pid_link_inode_operations
;
1745 ei
->op
.proc_get_link
= proc_fd_link
;
1746 dentry
->d_op
= &tid_fd_dentry_operations
;
1747 d_add(dentry
, inode
);
1748 /* Close the race of the process dying before we return the dentry */
1749 if (tid_fd_revalidate(dentry
, NULL
))
1755 spin_unlock(&files
->file_lock
);
1756 put_files_struct(files
);
1762 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1763 struct dentry
*dentry
,
1764 instantiate_t instantiate
)
1766 struct task_struct
*task
= get_proc_task(dir
);
1767 unsigned fd
= name_to_int(dentry
);
1768 struct dentry
*result
= ERR_PTR(-ENOENT
);
1775 result
= instantiate(dir
, dentry
, task
, &fd
);
1777 put_task_struct(task
);
1782 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1783 filldir_t filldir
, instantiate_t instantiate
)
1785 struct dentry
*dentry
= filp
->f_path
.dentry
;
1786 struct inode
*inode
= dentry
->d_inode
;
1787 struct task_struct
*p
= get_proc_task(inode
);
1788 unsigned int fd
, ino
;
1790 struct files_struct
* files
;
1800 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1804 ino
= parent_ino(dentry
);
1805 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1809 files
= get_files_struct(p
);
1813 for (fd
= filp
->f_pos
-2;
1814 fd
< files_fdtable(files
)->max_fds
;
1815 fd
++, filp
->f_pos
++) {
1816 char name
[PROC_NUMBUF
];
1819 if (!fcheck_files(files
, fd
))
1823 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1824 if (proc_fill_cache(filp
, dirent
, filldir
,
1825 name
, len
, instantiate
,
1833 put_files_struct(files
);
1841 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1842 struct nameidata
*nd
)
1844 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1847 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1849 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1852 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1853 size_t len
, loff_t
*ppos
)
1855 char tmp
[PROC_FDINFO_MAX
];
1856 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1858 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1862 static const struct file_operations proc_fdinfo_file_operations
= {
1863 .open
= nonseekable_open
,
1864 .read
= proc_fdinfo_read
,
1867 static const struct file_operations proc_fd_operations
= {
1868 .read
= generic_read_dir
,
1869 .readdir
= proc_readfd
,
1873 * /proc/pid/fd needs a special permission handler so that a process can still
1874 * access /proc/self/fd after it has executed a setuid().
1876 static int proc_fd_permission(struct inode
*inode
, int mask
)
1880 rv
= generic_permission(inode
, mask
, NULL
);
1883 if (task_pid(current
) == proc_pid(inode
))
1889 * proc directories can do almost nothing..
1891 static const struct inode_operations proc_fd_inode_operations
= {
1892 .lookup
= proc_lookupfd
,
1893 .permission
= proc_fd_permission
,
1894 .setattr
= proc_setattr
,
1897 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
1898 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1900 unsigned fd
= *(unsigned *)ptr
;
1901 struct inode
*inode
;
1902 struct proc_inode
*ei
;
1903 struct dentry
*error
= ERR_PTR(-ENOENT
);
1905 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1910 inode
->i_mode
= S_IFREG
| S_IRUSR
;
1911 inode
->i_fop
= &proc_fdinfo_file_operations
;
1912 dentry
->d_op
= &tid_fd_dentry_operations
;
1913 d_add(dentry
, inode
);
1914 /* Close the race of the process dying before we return the dentry */
1915 if (tid_fd_revalidate(dentry
, NULL
))
1922 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
1923 struct dentry
*dentry
,
1924 struct nameidata
*nd
)
1926 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
1929 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
1931 return proc_readfd_common(filp
, dirent
, filldir
,
1932 proc_fdinfo_instantiate
);
1935 static const struct file_operations proc_fdinfo_operations
= {
1936 .read
= generic_read_dir
,
1937 .readdir
= proc_readfdinfo
,
1941 * proc directories can do almost nothing..
1943 static const struct inode_operations proc_fdinfo_inode_operations
= {
1944 .lookup
= proc_lookupfdinfo
,
1945 .setattr
= proc_setattr
,
1949 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
1950 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1952 const struct pid_entry
*p
= ptr
;
1953 struct inode
*inode
;
1954 struct proc_inode
*ei
;
1955 struct dentry
*error
= ERR_PTR(-EINVAL
);
1957 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1962 inode
->i_mode
= p
->mode
;
1963 if (S_ISDIR(inode
->i_mode
))
1964 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1966 inode
->i_op
= p
->iop
;
1968 inode
->i_fop
= p
->fop
;
1970 dentry
->d_op
= &pid_dentry_operations
;
1971 d_add(dentry
, inode
);
1972 /* Close the race of the process dying before we return the dentry */
1973 if (pid_revalidate(dentry
, NULL
))
1979 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1980 struct dentry
*dentry
,
1981 const struct pid_entry
*ents
,
1984 struct dentry
*error
;
1985 struct task_struct
*task
= get_proc_task(dir
);
1986 const struct pid_entry
*p
, *last
;
1988 error
= ERR_PTR(-ENOENT
);
1994 * Yes, it does not scale. And it should not. Don't add
1995 * new entries into /proc/<tgid>/ without very good reasons.
1997 last
= &ents
[nents
- 1];
1998 for (p
= ents
; p
<= last
; p
++) {
1999 if (p
->len
!= dentry
->d_name
.len
)
2001 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2007 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2009 put_task_struct(task
);
2014 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2015 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2017 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2018 proc_pident_instantiate
, task
, p
);
2021 static int proc_pident_readdir(struct file
*filp
,
2022 void *dirent
, filldir_t filldir
,
2023 const struct pid_entry
*ents
, unsigned int nents
)
2026 struct dentry
*dentry
= filp
->f_path
.dentry
;
2027 struct inode
*inode
= dentry
->d_inode
;
2028 struct task_struct
*task
= get_proc_task(inode
);
2029 const struct pid_entry
*p
, *last
;
2042 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2048 ino
= parent_ino(dentry
);
2049 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2061 last
= &ents
[nents
- 1];
2063 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2072 put_task_struct(task
);
2077 #ifdef CONFIG_SECURITY
2078 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2079 size_t count
, loff_t
*ppos
)
2081 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2084 struct task_struct
*task
= get_proc_task(inode
);
2089 length
= security_getprocattr(task
,
2090 (char*)file
->f_path
.dentry
->d_name
.name
,
2092 put_task_struct(task
);
2094 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2099 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2100 size_t count
, loff_t
*ppos
)
2102 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2105 struct task_struct
*task
= get_proc_task(inode
);
2110 if (count
> PAGE_SIZE
)
2113 /* No partial writes. */
2119 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2124 if (copy_from_user(page
, buf
, count
))
2127 length
= security_setprocattr(task
,
2128 (char*)file
->f_path
.dentry
->d_name
.name
,
2129 (void*)page
, count
);
2131 free_page((unsigned long) page
);
2133 put_task_struct(task
);
2138 static const struct file_operations proc_pid_attr_operations
= {
2139 .read
= proc_pid_attr_read
,
2140 .write
= proc_pid_attr_write
,
2143 static const struct pid_entry attr_dir_stuff
[] = {
2144 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2145 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2146 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2147 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2148 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2149 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2152 static int proc_attr_dir_readdir(struct file
* filp
,
2153 void * dirent
, filldir_t filldir
)
2155 return proc_pident_readdir(filp
,dirent
,filldir
,
2156 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2159 static const struct file_operations proc_attr_dir_operations
= {
2160 .read
= generic_read_dir
,
2161 .readdir
= proc_attr_dir_readdir
,
2164 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2165 struct dentry
*dentry
, struct nameidata
*nd
)
2167 return proc_pident_lookup(dir
, dentry
,
2168 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2171 static const struct inode_operations proc_attr_dir_inode_operations
= {
2172 .lookup
= proc_attr_dir_lookup
,
2173 .getattr
= pid_getattr
,
2174 .setattr
= proc_setattr
,
2179 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2180 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2181 size_t count
, loff_t
*ppos
)
2183 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2184 struct mm_struct
*mm
;
2185 char buffer
[PROC_NUMBUF
];
2193 mm
= get_task_mm(task
);
2195 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2196 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2197 MMF_DUMP_FILTER_SHIFT
));
2199 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2202 put_task_struct(task
);
2207 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2208 const char __user
*buf
,
2212 struct task_struct
*task
;
2213 struct mm_struct
*mm
;
2214 char buffer
[PROC_NUMBUF
], *end
;
2221 memset(buffer
, 0, sizeof(buffer
));
2222 if (count
> sizeof(buffer
) - 1)
2223 count
= sizeof(buffer
) - 1;
2224 if (copy_from_user(buffer
, buf
, count
))
2228 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2231 if (end
- buffer
== 0)
2235 task
= get_proc_task(file
->f_dentry
->d_inode
);
2240 mm
= get_task_mm(task
);
2244 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2246 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2248 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2253 put_task_struct(task
);
2258 static const struct file_operations proc_coredump_filter_operations
= {
2259 .read
= proc_coredump_filter_read
,
2260 .write
= proc_coredump_filter_write
,
2267 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2270 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2271 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2272 char tmp
[PROC_NUMBUF
];
2275 sprintf(tmp
, "%d", tgid
);
2276 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2279 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2281 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2282 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2283 char tmp
[PROC_NUMBUF
];
2285 return ERR_PTR(-ENOENT
);
2286 sprintf(tmp
, "%d", task_tgid_nr_ns(current
, ns
));
2287 return ERR_PTR(vfs_follow_link(nd
,tmp
));
2290 static const struct inode_operations proc_self_inode_operations
= {
2291 .readlink
= proc_self_readlink
,
2292 .follow_link
= proc_self_follow_link
,
2298 * These are the directory entries in the root directory of /proc
2299 * that properly belong to the /proc filesystem, as they describe
2300 * describe something that is process related.
2302 static const struct pid_entry proc_base_stuff
[] = {
2303 NOD("self", S_IFLNK
|S_IRWXUGO
,
2304 &proc_self_inode_operations
, NULL
, {}),
2308 * Exceptional case: normally we are not allowed to unhash a busy
2309 * directory. In this case, however, we can do it - no aliasing problems
2310 * due to the way we treat inodes.
2312 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
2314 struct inode
*inode
= dentry
->d_inode
;
2315 struct task_struct
*task
= get_proc_task(inode
);
2317 put_task_struct(task
);
2324 static const struct dentry_operations proc_base_dentry_operations
=
2326 .d_revalidate
= proc_base_revalidate
,
2327 .d_delete
= pid_delete_dentry
,
2330 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2331 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2333 const struct pid_entry
*p
= ptr
;
2334 struct inode
*inode
;
2335 struct proc_inode
*ei
;
2336 struct dentry
*error
= ERR_PTR(-EINVAL
);
2338 /* Allocate the inode */
2339 error
= ERR_PTR(-ENOMEM
);
2340 inode
= new_inode(dir
->i_sb
);
2344 /* Initialize the inode */
2346 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2349 * grab the reference to the task.
2351 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2355 inode
->i_mode
= p
->mode
;
2356 if (S_ISDIR(inode
->i_mode
))
2358 if (S_ISLNK(inode
->i_mode
))
2361 inode
->i_op
= p
->iop
;
2363 inode
->i_fop
= p
->fop
;
2365 dentry
->d_op
= &proc_base_dentry_operations
;
2366 d_add(dentry
, inode
);
2375 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2377 struct dentry
*error
;
2378 struct task_struct
*task
= get_proc_task(dir
);
2379 const struct pid_entry
*p
, *last
;
2381 error
= ERR_PTR(-ENOENT
);
2386 /* Lookup the directory entry */
2387 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2388 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2389 if (p
->len
!= dentry
->d_name
.len
)
2391 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2397 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2400 put_task_struct(task
);
2405 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2406 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2408 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2409 proc_base_instantiate
, task
, p
);
2412 #ifdef CONFIG_TASK_IO_ACCOUNTING
2413 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2415 struct task_io_accounting acct
= task
->ioac
;
2416 unsigned long flags
;
2418 if (whole
&& lock_task_sighand(task
, &flags
)) {
2419 struct task_struct
*t
= task
;
2421 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2422 while_each_thread(task
, t
)
2423 task_io_accounting_add(&acct
, &t
->ioac
);
2425 unlock_task_sighand(task
, &flags
);
2427 return sprintf(buffer
,
2432 "read_bytes: %llu\n"
2433 "write_bytes: %llu\n"
2434 "cancelled_write_bytes: %llu\n",
2435 (unsigned long long)acct
.rchar
,
2436 (unsigned long long)acct
.wchar
,
2437 (unsigned long long)acct
.syscr
,
2438 (unsigned long long)acct
.syscw
,
2439 (unsigned long long)acct
.read_bytes
,
2440 (unsigned long long)acct
.write_bytes
,
2441 (unsigned long long)acct
.cancelled_write_bytes
);
2444 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2446 return do_io_accounting(task
, buffer
, 0);
2449 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2451 return do_io_accounting(task
, buffer
, 1);
2453 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2455 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2456 struct pid
*pid
, struct task_struct
*task
)
2458 seq_printf(m
, "%08x\n", task
->personality
);
2465 static const struct file_operations proc_task_operations
;
2466 static const struct inode_operations proc_task_inode_operations
;
2468 static const struct pid_entry tgid_base_stuff
[] = {
2469 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2470 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2471 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2473 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2475 REG("environ", S_IRUSR
, proc_environ_operations
),
2476 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2477 ONE("status", S_IRUGO
, proc_pid_status
),
2478 ONE("personality", S_IRUSR
, proc_pid_personality
),
2479 INF("limits", S_IRUSR
, proc_pid_limits
),
2480 #ifdef CONFIG_SCHED_DEBUG
2481 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2483 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2484 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2486 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2487 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2488 ONE("statm", S_IRUGO
, proc_pid_statm
),
2489 REG("maps", S_IRUGO
, proc_maps_operations
),
2491 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2493 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2494 LNK("cwd", proc_cwd_link
),
2495 LNK("root", proc_root_link
),
2496 LNK("exe", proc_exe_link
),
2497 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2498 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2499 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2500 #ifdef CONFIG_PROC_PAGE_MONITOR
2501 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2502 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2503 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2505 #ifdef CONFIG_SECURITY
2506 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2508 #ifdef CONFIG_KALLSYMS
2509 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2511 #ifdef CONFIG_STACKTRACE
2512 ONE("stack", S_IRUSR
, proc_pid_stack
),
2514 #ifdef CONFIG_SCHEDSTATS
2515 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2517 #ifdef CONFIG_LATENCYTOP
2518 REG("latency", S_IRUGO
, proc_lstats_operations
),
2520 #ifdef CONFIG_PROC_PID_CPUSET
2521 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2523 #ifdef CONFIG_CGROUPS
2524 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2526 INF("oom_score", S_IRUGO
, proc_oom_score
),
2527 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2528 #ifdef CONFIG_AUDITSYSCALL
2529 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2530 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2532 #ifdef CONFIG_FAULT_INJECTION
2533 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2535 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2536 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2538 #ifdef CONFIG_TASK_IO_ACCOUNTING
2539 INF("io", S_IRUGO
, proc_tgid_io_accounting
),
2543 static int proc_tgid_base_readdir(struct file
* filp
,
2544 void * dirent
, filldir_t filldir
)
2546 return proc_pident_readdir(filp
,dirent
,filldir
,
2547 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2550 static const struct file_operations proc_tgid_base_operations
= {
2551 .read
= generic_read_dir
,
2552 .readdir
= proc_tgid_base_readdir
,
2555 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2556 return proc_pident_lookup(dir
, dentry
,
2557 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2560 static const struct inode_operations proc_tgid_base_inode_operations
= {
2561 .lookup
= proc_tgid_base_lookup
,
2562 .getattr
= pid_getattr
,
2563 .setattr
= proc_setattr
,
2566 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2568 struct dentry
*dentry
, *leader
, *dir
;
2569 char buf
[PROC_NUMBUF
];
2573 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2574 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2576 if (!(current
->flags
& PF_EXITING
))
2577 shrink_dcache_parent(dentry
);
2586 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2587 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2592 name
.len
= strlen(name
.name
);
2593 dir
= d_hash_and_lookup(leader
, &name
);
2595 goto out_put_leader
;
2598 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2599 dentry
= d_hash_and_lookup(dir
, &name
);
2601 shrink_dcache_parent(dentry
);
2614 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2615 * @task: task that should be flushed.
2617 * When flushing dentries from proc, one needs to flush them from global
2618 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2619 * in. This call is supposed to do all of this job.
2621 * Looks in the dcache for
2623 * /proc/@tgid/task/@pid
2624 * if either directory is present flushes it and all of it'ts children
2627 * It is safe and reasonable to cache /proc entries for a task until
2628 * that task exits. After that they just clog up the dcache with
2629 * useless entries, possibly causing useful dcache entries to be
2630 * flushed instead. This routine is proved to flush those useless
2631 * dcache entries at process exit time.
2633 * NOTE: This routine is just an optimization so it does not guarantee
2634 * that no dcache entries will exist at process exit time it
2635 * just makes it very unlikely that any will persist.
2638 void proc_flush_task(struct task_struct
*task
)
2641 struct pid
*pid
, *tgid
= NULL
;
2644 pid
= task_pid(task
);
2645 if (thread_group_leader(task
))
2646 tgid
= task_tgid(task
);
2648 for (i
= 0; i
<= pid
->level
; i
++) {
2649 upid
= &pid
->numbers
[i
];
2650 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2651 tgid
? tgid
->numbers
[i
].nr
: 0);
2654 upid
= &pid
->numbers
[pid
->level
];
2656 pid_ns_release_proc(upid
->ns
);
2659 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2660 struct dentry
* dentry
,
2661 struct task_struct
*task
, const void *ptr
)
2663 struct dentry
*error
= ERR_PTR(-ENOENT
);
2664 struct inode
*inode
;
2666 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2670 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2671 inode
->i_op
= &proc_tgid_base_inode_operations
;
2672 inode
->i_fop
= &proc_tgid_base_operations
;
2673 inode
->i_flags
|=S_IMMUTABLE
;
2675 inode
->i_nlink
= 2 + pid_entry_count_dirs(tgid_base_stuff
,
2676 ARRAY_SIZE(tgid_base_stuff
));
2678 dentry
->d_op
= &pid_dentry_operations
;
2680 d_add(dentry
, inode
);
2681 /* Close the race of the process dying before we return the dentry */
2682 if (pid_revalidate(dentry
, NULL
))
2688 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2690 struct dentry
*result
= ERR_PTR(-ENOENT
);
2691 struct task_struct
*task
;
2693 struct pid_namespace
*ns
;
2695 result
= proc_base_lookup(dir
, dentry
);
2696 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2699 tgid
= name_to_int(dentry
);
2703 ns
= dentry
->d_sb
->s_fs_info
;
2705 task
= find_task_by_pid_ns(tgid
, ns
);
2707 get_task_struct(task
);
2712 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2713 put_task_struct(task
);
2719 * Find the first task with tgid >= tgid
2724 struct task_struct
*task
;
2726 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
2731 put_task_struct(iter
.task
);
2735 pid
= find_ge_pid(iter
.tgid
, ns
);
2737 iter
.tgid
= pid_nr_ns(pid
, ns
);
2738 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
2739 /* What we to know is if the pid we have find is the
2740 * pid of a thread_group_leader. Testing for task
2741 * being a thread_group_leader is the obvious thing
2742 * todo but there is a window when it fails, due to
2743 * the pid transfer logic in de_thread.
2745 * So we perform the straight forward test of seeing
2746 * if the pid we have found is the pid of a thread
2747 * group leader, and don't worry if the task we have
2748 * found doesn't happen to be a thread group leader.
2749 * As we don't care in the case of readdir.
2751 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
2755 get_task_struct(iter
.task
);
2761 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2763 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2764 struct tgid_iter iter
)
2766 char name
[PROC_NUMBUF
];
2767 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
2768 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2769 proc_pid_instantiate
, iter
.task
, NULL
);
2772 /* for the /proc/ directory itself, after non-process stuff has been done */
2773 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2775 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2776 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2777 struct tgid_iter iter
;
2778 struct pid_namespace
*ns
;
2783 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2784 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2785 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2789 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2791 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
2792 for (iter
= next_tgid(ns
, iter
);
2794 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
2795 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
2796 if (proc_pid_fill_cache(filp
, dirent
, filldir
, iter
) < 0) {
2797 put_task_struct(iter
.task
);
2801 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2803 put_task_struct(reaper
);
2811 static const struct pid_entry tid_base_stuff
[] = {
2812 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2813 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fd_operations
),
2814 REG("environ", S_IRUSR
, proc_environ_operations
),
2815 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2816 ONE("status", S_IRUGO
, proc_pid_status
),
2817 ONE("personality", S_IRUSR
, proc_pid_personality
),
2818 INF("limits", S_IRUSR
, proc_pid_limits
),
2819 #ifdef CONFIG_SCHED_DEBUG
2820 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2822 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2823 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2825 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2826 ONE("stat", S_IRUGO
, proc_tid_stat
),
2827 ONE("statm", S_IRUGO
, proc_pid_statm
),
2828 REG("maps", S_IRUGO
, proc_maps_operations
),
2830 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2832 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2833 LNK("cwd", proc_cwd_link
),
2834 LNK("root", proc_root_link
),
2835 LNK("exe", proc_exe_link
),
2836 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2837 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2838 #ifdef CONFIG_PROC_PAGE_MONITOR
2839 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2840 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2841 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2843 #ifdef CONFIG_SECURITY
2844 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2846 #ifdef CONFIG_KALLSYMS
2847 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2849 #ifdef CONFIG_STACKTRACE
2850 ONE("stack", S_IRUSR
, proc_pid_stack
),
2852 #ifdef CONFIG_SCHEDSTATS
2853 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2855 #ifdef CONFIG_LATENCYTOP
2856 REG("latency", S_IRUGO
, proc_lstats_operations
),
2858 #ifdef CONFIG_PROC_PID_CPUSET
2859 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2861 #ifdef CONFIG_CGROUPS
2862 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2864 INF("oom_score", S_IRUGO
, proc_oom_score
),
2865 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2866 #ifdef CONFIG_AUDITSYSCALL
2867 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2868 REG("sessionid", S_IRUSR
, proc_sessionid_operations
),
2870 #ifdef CONFIG_FAULT_INJECTION
2871 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2873 #ifdef CONFIG_TASK_IO_ACCOUNTING
2874 INF("io", S_IRUGO
, proc_tid_io_accounting
),
2878 static int proc_tid_base_readdir(struct file
* filp
,
2879 void * dirent
, filldir_t filldir
)
2881 return proc_pident_readdir(filp
,dirent
,filldir
,
2882 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2885 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2886 return proc_pident_lookup(dir
, dentry
,
2887 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
2890 static const struct file_operations proc_tid_base_operations
= {
2891 .read
= generic_read_dir
,
2892 .readdir
= proc_tid_base_readdir
,
2895 static const struct inode_operations proc_tid_base_inode_operations
= {
2896 .lookup
= proc_tid_base_lookup
,
2897 .getattr
= pid_getattr
,
2898 .setattr
= proc_setattr
,
2901 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
2902 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2904 struct dentry
*error
= ERR_PTR(-ENOENT
);
2905 struct inode
*inode
;
2906 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2910 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2911 inode
->i_op
= &proc_tid_base_inode_operations
;
2912 inode
->i_fop
= &proc_tid_base_operations
;
2913 inode
->i_flags
|=S_IMMUTABLE
;
2915 inode
->i_nlink
= 2 + pid_entry_count_dirs(tid_base_stuff
,
2916 ARRAY_SIZE(tid_base_stuff
));
2918 dentry
->d_op
= &pid_dentry_operations
;
2920 d_add(dentry
, inode
);
2921 /* Close the race of the process dying before we return the dentry */
2922 if (pid_revalidate(dentry
, NULL
))
2928 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2930 struct dentry
*result
= ERR_PTR(-ENOENT
);
2931 struct task_struct
*task
;
2932 struct task_struct
*leader
= get_proc_task(dir
);
2934 struct pid_namespace
*ns
;
2939 tid
= name_to_int(dentry
);
2943 ns
= dentry
->d_sb
->s_fs_info
;
2945 task
= find_task_by_pid_ns(tid
, ns
);
2947 get_task_struct(task
);
2951 if (!same_thread_group(leader
, task
))
2954 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
2956 put_task_struct(task
);
2958 put_task_struct(leader
);
2964 * Find the first tid of a thread group to return to user space.
2966 * Usually this is just the thread group leader, but if the users
2967 * buffer was too small or there was a seek into the middle of the
2968 * directory we have more work todo.
2970 * In the case of a short read we start with find_task_by_pid.
2972 * In the case of a seek we start with the leader and walk nr
2975 static struct task_struct
*first_tid(struct task_struct
*leader
,
2976 int tid
, int nr
, struct pid_namespace
*ns
)
2978 struct task_struct
*pos
;
2981 /* Attempt to start with the pid of a thread */
2982 if (tid
&& (nr
> 0)) {
2983 pos
= find_task_by_pid_ns(tid
, ns
);
2984 if (pos
&& (pos
->group_leader
== leader
))
2988 /* If nr exceeds the number of threads there is nothing todo */
2990 if (nr
&& nr
>= get_nr_threads(leader
))
2993 /* If we haven't found our starting place yet start
2994 * with the leader and walk nr threads forward.
2996 for (pos
= leader
; nr
> 0; --nr
) {
2997 pos
= next_thread(pos
);
2998 if (pos
== leader
) {
3004 get_task_struct(pos
);
3011 * Find the next thread in the thread list.
3012 * Return NULL if there is an error or no next thread.
3014 * The reference to the input task_struct is released.
3016 static struct task_struct
*next_tid(struct task_struct
*start
)
3018 struct task_struct
*pos
= NULL
;
3020 if (pid_alive(start
)) {
3021 pos
= next_thread(start
);
3022 if (thread_group_leader(pos
))
3025 get_task_struct(pos
);
3028 put_task_struct(start
);
3032 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3033 struct task_struct
*task
, int tid
)
3035 char name
[PROC_NUMBUF
];
3036 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3037 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3038 proc_task_instantiate
, task
, NULL
);
3041 /* for the /proc/TGID/task/ directories */
3042 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3044 struct dentry
*dentry
= filp
->f_path
.dentry
;
3045 struct inode
*inode
= dentry
->d_inode
;
3046 struct task_struct
*leader
= NULL
;
3047 struct task_struct
*task
;
3048 int retval
= -ENOENT
;
3051 struct pid_namespace
*ns
;
3053 task
= get_proc_task(inode
);
3057 if (pid_alive(task
)) {
3058 leader
= task
->group_leader
;
3059 get_task_struct(leader
);
3062 put_task_struct(task
);
3067 switch ((unsigned long)filp
->f_pos
) {
3070 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3075 ino
= parent_ino(dentry
);
3076 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3082 /* f_version caches the tgid value that the last readdir call couldn't
3083 * return. lseek aka telldir automagically resets f_version to 0.
3085 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3086 tid
= (int)filp
->f_version
;
3087 filp
->f_version
= 0;
3088 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3090 task
= next_tid(task
), filp
->f_pos
++) {
3091 tid
= task_pid_nr_ns(task
, ns
);
3092 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3093 /* returning this tgid failed, save it as the first
3094 * pid for the next readir call */
3095 filp
->f_version
= (u64
)tid
;
3096 put_task_struct(task
);
3101 put_task_struct(leader
);
3106 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3108 struct inode
*inode
= dentry
->d_inode
;
3109 struct task_struct
*p
= get_proc_task(inode
);
3110 generic_fillattr(inode
, stat
);
3113 stat
->nlink
+= get_nr_threads(p
);
3120 static const struct inode_operations proc_task_inode_operations
= {
3121 .lookup
= proc_task_lookup
,
3122 .getattr
= proc_task_getattr
,
3123 .setattr
= proc_setattr
,
3126 static const struct file_operations proc_task_operations
= {
3127 .read
= generic_read_dir
,
3128 .readdir
= proc_task_readdir
,