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/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/fs_struct.h>
85 #include <linux/slab.h>
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
102 const struct inode_operations
*iop
;
103 const struct file_operations
*fop
;
107 #define NOD(NAME, MODE, IOP, FOP, OP) { \
109 .len = sizeof(NAME) - 1, \
116 #define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118 #define LNK(NAME, get_link) \
119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
120 &proc_pid_link_inode_operations, NULL, \
121 { .proc_get_link = get_link } )
122 #define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124 #define INF(NAME, MODE, read) \
125 NOD(NAME, (S_IFREG|(MODE)), \
126 NULL, &proc_info_file_operations, \
127 { .proc_read = read } )
128 #define ONE(NAME, MODE, show) \
129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
131 { .proc_show = show } )
134 * Count the number of hardlinks for the pid_entry table, excluding the .
137 static unsigned int pid_entry_count_dirs(const struct pid_entry
*entries
,
144 for (i
= 0; i
< n
; ++i
) {
145 if (S_ISDIR(entries
[i
].mode
))
152 static int get_task_root(struct task_struct
*task
, struct path
*root
)
154 int result
= -ENOENT
;
158 get_fs_root(task
->fs
, root
);
165 static int proc_cwd_link(struct inode
*inode
, struct path
*path
)
167 struct task_struct
*task
= get_proc_task(inode
);
168 int result
= -ENOENT
;
173 get_fs_pwd(task
->fs
, path
);
177 put_task_struct(task
);
182 static int proc_root_link(struct inode
*inode
, struct path
*path
)
184 struct task_struct
*task
= get_proc_task(inode
);
185 int result
= -ENOENT
;
188 result
= get_task_root(task
, path
);
189 put_task_struct(task
);
195 * Return zero if current may access user memory in @task, -error if not.
197 static int check_mem_permission(struct task_struct
*task
)
200 * A task can always look at itself, in case it chooses
201 * to use system calls instead of load instructions.
207 * If current is actively ptrace'ing, and would also be
208 * permitted to freshly attach with ptrace now, permit it.
210 if (task_is_stopped_or_traced(task
)) {
213 match
= (tracehook_tracer_task(task
) == current
);
215 if (match
&& ptrace_may_access(task
, PTRACE_MODE_ATTACH
))
220 * Noone else is allowed.
225 struct mm_struct
*mm_for_maps(struct task_struct
*task
)
227 struct mm_struct
*mm
;
229 if (mutex_lock_killable(&task
->cred_guard_mutex
))
232 mm
= get_task_mm(task
);
233 if (mm
&& mm
!= current
->mm
&&
234 !ptrace_may_access(task
, PTRACE_MODE_READ
)) {
238 mutex_unlock(&task
->cred_guard_mutex
);
243 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
247 struct mm_struct
*mm
= get_task_mm(task
);
251 goto out_mm
; /* Shh! No looking before we're done */
253 len
= mm
->arg_end
- mm
->arg_start
;
258 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
260 // If the nul at the end of args has been overwritten, then
261 // assume application is using setproctitle(3).
262 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
263 len
= strnlen(buffer
, res
);
267 len
= mm
->env_end
- mm
->env_start
;
268 if (len
> PAGE_SIZE
- res
)
269 len
= PAGE_SIZE
- res
;
270 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
271 res
= strnlen(buffer
, res
);
280 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
283 struct mm_struct
*mm
= get_task_mm(task
);
285 unsigned int nwords
= 0;
288 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
289 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
292 memcpy(buffer
, mm
->saved_auxv
, res
);
299 #ifdef CONFIG_KALLSYMS
301 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
302 * Returns the resolved symbol. If that fails, simply return the address.
304 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
307 char symname
[KSYM_NAME_LEN
];
309 wchan
= get_wchan(task
);
311 if (lookup_symbol_name(wchan
, symname
) < 0)
312 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
315 return sprintf(buffer
, "%lu", wchan
);
317 return sprintf(buffer
, "%s", symname
);
319 #endif /* CONFIG_KALLSYMS */
321 #ifdef CONFIG_STACKTRACE
323 #define MAX_STACK_TRACE_DEPTH 64
325 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
326 struct pid
*pid
, struct task_struct
*task
)
328 struct stack_trace trace
;
329 unsigned long *entries
;
332 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
336 trace
.nr_entries
= 0;
337 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
338 trace
.entries
= entries
;
340 save_stack_trace_tsk(task
, &trace
);
342 for (i
= 0; i
< trace
.nr_entries
; i
++) {
343 seq_printf(m
, "[<%p>] %pS\n",
344 (void *)entries
[i
], (void *)entries
[i
]);
352 #ifdef CONFIG_SCHEDSTATS
354 * Provides /proc/PID/schedstat
356 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
358 return sprintf(buffer
, "%llu %llu %lu\n",
359 (unsigned long long)task
->se
.sum_exec_runtime
,
360 (unsigned long long)task
->sched_info
.run_delay
,
361 task
->sched_info
.pcount
);
365 #ifdef CONFIG_LATENCYTOP
366 static int lstats_show_proc(struct seq_file
*m
, void *v
)
369 struct inode
*inode
= m
->private;
370 struct task_struct
*task
= get_proc_task(inode
);
374 seq_puts(m
, "Latency Top version : v0.1\n");
375 for (i
= 0; i
< 32; i
++) {
376 if (task
->latency_record
[i
].backtrace
[0]) {
378 seq_printf(m
, "%i %li %li ",
379 task
->latency_record
[i
].count
,
380 task
->latency_record
[i
].time
,
381 task
->latency_record
[i
].max
);
382 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
383 char sym
[KSYM_SYMBOL_LEN
];
385 if (!task
->latency_record
[i
].backtrace
[q
])
387 if (task
->latency_record
[i
].backtrace
[q
] == ULONG_MAX
)
389 sprint_symbol(sym
, task
->latency_record
[i
].backtrace
[q
]);
390 c
= strchr(sym
, '+');
393 seq_printf(m
, "%s ", sym
);
399 put_task_struct(task
);
403 static int lstats_open(struct inode
*inode
, struct file
*file
)
405 return single_open(file
, lstats_show_proc
, inode
);
408 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
409 size_t count
, loff_t
*offs
)
411 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
415 clear_all_latency_tracing(task
);
416 put_task_struct(task
);
421 static const struct file_operations proc_lstats_operations
= {
424 .write
= lstats_write
,
426 .release
= single_release
,
431 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
433 unsigned long points
= 0;
435 read_lock(&tasklist_lock
);
437 points
= oom_badness(task
, NULL
, NULL
,
438 totalram_pages
+ total_swap_pages
);
439 read_unlock(&tasklist_lock
);
440 return sprintf(buffer
, "%lu\n", points
);
448 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
449 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
450 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
451 [RLIMIT_DATA
] = {"Max data size", "bytes"},
452 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
453 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
454 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
455 [RLIMIT_NPROC
] = {"Max processes", "processes"},
456 [RLIMIT_NOFILE
] = {"Max open files", "files"},
457 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
458 [RLIMIT_AS
] = {"Max address space", "bytes"},
459 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
460 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
461 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
462 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
463 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
464 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
467 /* Display limits for a process */
468 static int proc_pid_limits(struct task_struct
*task
, char *buffer
)
473 char *bufptr
= buffer
;
475 struct rlimit rlim
[RLIM_NLIMITS
];
477 if (!lock_task_sighand(task
, &flags
))
479 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
480 unlock_task_sighand(task
, &flags
);
483 * print the file header
485 count
+= sprintf(&bufptr
[count
], "%-25s %-20s %-20s %-10s\n",
486 "Limit", "Soft Limit", "Hard Limit", "Units");
488 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
489 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
490 count
+= sprintf(&bufptr
[count
], "%-25s %-20s ",
491 lnames
[i
].name
, "unlimited");
493 count
+= sprintf(&bufptr
[count
], "%-25s %-20lu ",
494 lnames
[i
].name
, rlim
[i
].rlim_cur
);
496 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
497 count
+= sprintf(&bufptr
[count
], "%-20s ", "unlimited");
499 count
+= sprintf(&bufptr
[count
], "%-20lu ",
503 count
+= sprintf(&bufptr
[count
], "%-10s\n",
506 count
+= sprintf(&bufptr
[count
], "\n");
512 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
513 static int proc_pid_syscall(struct task_struct
*task
, char *buffer
)
516 unsigned long args
[6], sp
, pc
;
518 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
519 return sprintf(buffer
, "running\n");
522 return sprintf(buffer
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
524 return sprintf(buffer
,
525 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
527 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
530 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
532 /************************************************************************/
533 /* Here the fs part begins */
534 /************************************************************************/
536 /* permission checks */
537 static int proc_fd_access_allowed(struct inode
*inode
)
539 struct task_struct
*task
;
541 /* Allow access to a task's file descriptors if it is us or we
542 * may use ptrace attach to the process and find out that
545 task
= get_proc_task(inode
);
547 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ
);
548 put_task_struct(task
);
553 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
556 struct inode
*inode
= dentry
->d_inode
;
558 if (attr
->ia_valid
& ATTR_MODE
)
561 error
= inode_change_ok(inode
, attr
);
565 if ((attr
->ia_valid
& ATTR_SIZE
) &&
566 attr
->ia_size
!= i_size_read(inode
)) {
567 error
= vmtruncate(inode
, attr
->ia_size
);
572 setattr_copy(inode
, attr
);
573 mark_inode_dirty(inode
);
577 static const struct inode_operations proc_def_inode_operations
= {
578 .setattr
= proc_setattr
,
581 static int mounts_open_common(struct inode
*inode
, struct file
*file
,
582 const struct seq_operations
*op
)
584 struct task_struct
*task
= get_proc_task(inode
);
586 struct mnt_namespace
*ns
= NULL
;
588 struct proc_mounts
*p
;
593 nsp
= task_nsproxy(task
);
600 if (ns
&& get_task_root(task
, &root
) == 0)
602 put_task_struct(task
);
611 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
615 file
->private_data
= &p
->m
;
616 ret
= seq_open(file
, op
);
623 p
->event
= ns
->event
;
637 static int mounts_release(struct inode
*inode
, struct file
*file
)
639 struct proc_mounts
*p
= file
->private_data
;
642 return seq_release(inode
, file
);
645 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
647 struct proc_mounts
*p
= file
->private_data
;
648 unsigned res
= POLLIN
| POLLRDNORM
;
650 poll_wait(file
, &p
->ns
->poll
, wait
);
651 if (mnt_had_events(p
))
652 res
|= POLLERR
| POLLPRI
;
657 static int mounts_open(struct inode
*inode
, struct file
*file
)
659 return mounts_open_common(inode
, file
, &mounts_op
);
662 static const struct file_operations proc_mounts_operations
= {
666 .release
= mounts_release
,
670 static int mountinfo_open(struct inode
*inode
, struct file
*file
)
672 return mounts_open_common(inode
, file
, &mountinfo_op
);
675 static const struct file_operations proc_mountinfo_operations
= {
676 .open
= mountinfo_open
,
679 .release
= mounts_release
,
683 static int mountstats_open(struct inode
*inode
, struct file
*file
)
685 return mounts_open_common(inode
, file
, &mountstats_op
);
688 static const struct file_operations proc_mountstats_operations
= {
689 .open
= mountstats_open
,
692 .release
= mounts_release
,
695 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
697 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
698 size_t count
, loff_t
*ppos
)
700 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
703 struct task_struct
*task
= get_proc_task(inode
);
709 if (count
> PROC_BLOCK_SIZE
)
710 count
= PROC_BLOCK_SIZE
;
713 if (!(page
= __get_free_page(GFP_TEMPORARY
)))
716 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
719 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
722 put_task_struct(task
);
727 static const struct file_operations proc_info_file_operations
= {
728 .read
= proc_info_read
,
729 .llseek
= generic_file_llseek
,
732 static int proc_single_show(struct seq_file
*m
, void *v
)
734 struct inode
*inode
= m
->private;
735 struct pid_namespace
*ns
;
737 struct task_struct
*task
;
740 ns
= inode
->i_sb
->s_fs_info
;
741 pid
= proc_pid(inode
);
742 task
= get_pid_task(pid
, PIDTYPE_PID
);
746 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
748 put_task_struct(task
);
752 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
755 ret
= single_open(filp
, proc_single_show
, NULL
);
757 struct seq_file
*m
= filp
->private_data
;
764 static const struct file_operations proc_single_file_operations
= {
765 .open
= proc_single_open
,
768 .release
= single_release
,
771 static int mem_open(struct inode
* inode
, struct file
* file
)
773 file
->private_data
= (void*)((long)current
->self_exec_id
);
777 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
778 size_t count
, loff_t
*ppos
)
780 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
782 unsigned long src
= *ppos
;
784 struct mm_struct
*mm
;
789 if (check_mem_permission(task
))
793 page
= (char *)__get_free_page(GFP_TEMPORARY
);
799 mm
= get_task_mm(task
);
805 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
811 int this_len
, retval
;
813 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
814 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
815 if (!retval
|| check_mem_permission(task
)) {
821 if (copy_to_user(buf
, page
, retval
)) {
836 free_page((unsigned long) page
);
838 put_task_struct(task
);
843 #define mem_write NULL
846 /* This is a security hazard */
847 static ssize_t
mem_write(struct file
* file
, const char __user
*buf
,
848 size_t count
, loff_t
*ppos
)
852 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
853 unsigned long dst
= *ppos
;
859 if (check_mem_permission(task
))
863 page
= (char *)__get_free_page(GFP_TEMPORARY
);
869 int this_len
, retval
;
871 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
872 if (copy_from_user(page
, buf
, this_len
)) {
876 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
888 free_page((unsigned long) page
);
890 put_task_struct(task
);
896 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
900 file
->f_pos
= offset
;
903 file
->f_pos
+= offset
;
908 force_successful_syscall_return();
912 static const struct file_operations proc_mem_operations
= {
919 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
920 size_t count
, loff_t
*ppos
)
922 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
924 unsigned long src
= *ppos
;
926 struct mm_struct
*mm
;
931 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
935 page
= (char *)__get_free_page(GFP_TEMPORARY
);
941 mm
= get_task_mm(task
);
946 int this_len
, retval
, max_len
;
948 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
953 max_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
954 this_len
= (this_len
> max_len
) ? max_len
: this_len
;
956 retval
= access_process_vm(task
, (mm
->env_start
+ src
),
964 if (copy_to_user(buf
, page
, retval
)) {
978 free_page((unsigned long) page
);
980 put_task_struct(task
);
985 static const struct file_operations proc_environ_operations
= {
986 .read
= environ_read
,
987 .llseek
= generic_file_llseek
,
990 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
991 size_t count
, loff_t
*ppos
)
993 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
994 char buffer
[PROC_NUMBUF
];
996 int oom_adjust
= OOM_DISABLE
;
1002 if (lock_task_sighand(task
, &flags
)) {
1003 oom_adjust
= task
->signal
->oom_adj
;
1004 unlock_task_sighand(task
, &flags
);
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
];
1020 unsigned long flags
;
1023 memset(buffer
, 0, sizeof(buffer
));
1024 if (count
> sizeof(buffer
) - 1)
1025 count
= sizeof(buffer
) - 1;
1026 if (copy_from_user(buffer
, buf
, count
))
1029 err
= strict_strtol(strstrip(buffer
), 0, &oom_adjust
);
1032 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
1033 oom_adjust
!= OOM_DISABLE
)
1036 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1039 if (!lock_task_sighand(task
, &flags
)) {
1040 put_task_struct(task
);
1044 if (oom_adjust
< task
->signal
->oom_adj
&& !capable(CAP_SYS_RESOURCE
)) {
1045 unlock_task_sighand(task
, &flags
);
1046 put_task_struct(task
);
1051 * Warn that /proc/pid/oom_adj is deprecated, see
1052 * Documentation/feature-removal-schedule.txt.
1054 printk_once(KERN_WARNING
"%s (%d): /proc/%d/oom_adj is deprecated, "
1055 "please use /proc/%d/oom_score_adj instead.\n",
1056 current
->comm
, task_pid_nr(current
),
1057 task_pid_nr(task
), task_pid_nr(task
));
1058 task
->signal
->oom_adj
= oom_adjust
;
1060 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1061 * value is always attainable.
1063 if (task
->signal
->oom_adj
== OOM_ADJUST_MAX
)
1064 task
->signal
->oom_score_adj
= OOM_SCORE_ADJ_MAX
;
1066 task
->signal
->oom_score_adj
= (oom_adjust
* OOM_SCORE_ADJ_MAX
) /
1068 unlock_task_sighand(task
, &flags
);
1069 put_task_struct(task
);
1074 static const struct file_operations proc_oom_adjust_operations
= {
1075 .read
= oom_adjust_read
,
1076 .write
= oom_adjust_write
,
1077 .llseek
= generic_file_llseek
,
1080 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
1081 size_t count
, loff_t
*ppos
)
1083 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1084 char buffer
[PROC_NUMBUF
];
1085 int oom_score_adj
= OOM_SCORE_ADJ_MIN
;
1086 unsigned long flags
;
1091 if (lock_task_sighand(task
, &flags
)) {
1092 oom_score_adj
= task
->signal
->oom_score_adj
;
1093 unlock_task_sighand(task
, &flags
);
1095 put_task_struct(task
);
1096 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_score_adj
);
1097 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1100 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1101 size_t count
, loff_t
*ppos
)
1103 struct task_struct
*task
;
1104 char buffer
[PROC_NUMBUF
];
1105 unsigned long flags
;
1109 memset(buffer
, 0, sizeof(buffer
));
1110 if (count
> sizeof(buffer
) - 1)
1111 count
= sizeof(buffer
) - 1;
1112 if (copy_from_user(buffer
, buf
, count
))
1115 err
= strict_strtol(strstrip(buffer
), 0, &oom_score_adj
);
1118 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1119 oom_score_adj
> OOM_SCORE_ADJ_MAX
)
1122 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1125 if (!lock_task_sighand(task
, &flags
)) {
1126 put_task_struct(task
);
1129 if (oom_score_adj
< task
->signal
->oom_score_adj
&&
1130 !capable(CAP_SYS_RESOURCE
)) {
1131 unlock_task_sighand(task
, &flags
);
1132 put_task_struct(task
);
1136 task
->signal
->oom_score_adj
= oom_score_adj
;
1138 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1139 * always attainable.
1141 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MIN
)
1142 task
->signal
->oom_adj
= OOM_DISABLE
;
1144 task
->signal
->oom_adj
= (oom_score_adj
* OOM_ADJUST_MAX
) /
1146 unlock_task_sighand(task
, &flags
);
1147 put_task_struct(task
);
1151 static const struct file_operations proc_oom_score_adj_operations
= {
1152 .read
= oom_score_adj_read
,
1153 .write
= oom_score_adj_write
,
1156 #ifdef CONFIG_AUDITSYSCALL
1157 #define TMPBUFLEN 21
1158 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1159 size_t count
, loff_t
*ppos
)
1161 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1162 struct task_struct
*task
= get_proc_task(inode
);
1164 char tmpbuf
[TMPBUFLEN
];
1168 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1169 audit_get_loginuid(task
));
1170 put_task_struct(task
);
1171 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1174 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1175 size_t count
, loff_t
*ppos
)
1177 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1182 if (!capable(CAP_AUDIT_CONTROL
))
1186 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1192 if (count
>= PAGE_SIZE
)
1193 count
= PAGE_SIZE
- 1;
1196 /* No partial writes. */
1199 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1203 if (copy_from_user(page
, buf
, count
))
1207 loginuid
= simple_strtoul(page
, &tmp
, 10);
1213 length
= audit_set_loginuid(current
, loginuid
);
1214 if (likely(length
== 0))
1218 free_page((unsigned long) page
);
1222 static const struct file_operations proc_loginuid_operations
= {
1223 .read
= proc_loginuid_read
,
1224 .write
= proc_loginuid_write
,
1225 .llseek
= generic_file_llseek
,
1228 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1229 size_t count
, loff_t
*ppos
)
1231 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1232 struct task_struct
*task
= get_proc_task(inode
);
1234 char tmpbuf
[TMPBUFLEN
];
1238 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1239 audit_get_sessionid(task
));
1240 put_task_struct(task
);
1241 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1244 static const struct file_operations proc_sessionid_operations
= {
1245 .read
= proc_sessionid_read
,
1246 .llseek
= generic_file_llseek
,
1250 #ifdef CONFIG_FAULT_INJECTION
1251 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1252 size_t count
, loff_t
*ppos
)
1254 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1255 char buffer
[PROC_NUMBUF
];
1261 make_it_fail
= task
->make_it_fail
;
1262 put_task_struct(task
);
1264 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1266 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1269 static ssize_t
proc_fault_inject_write(struct file
* file
,
1270 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1272 struct task_struct
*task
;
1273 char buffer
[PROC_NUMBUF
], *end
;
1276 if (!capable(CAP_SYS_RESOURCE
))
1278 memset(buffer
, 0, sizeof(buffer
));
1279 if (count
> sizeof(buffer
) - 1)
1280 count
= sizeof(buffer
) - 1;
1281 if (copy_from_user(buffer
, buf
, count
))
1283 make_it_fail
= simple_strtol(strstrip(buffer
), &end
, 0);
1286 task
= get_proc_task(file
->f_dentry
->d_inode
);
1289 task
->make_it_fail
= make_it_fail
;
1290 put_task_struct(task
);
1295 static const struct file_operations proc_fault_inject_operations
= {
1296 .read
= proc_fault_inject_read
,
1297 .write
= proc_fault_inject_write
,
1298 .llseek
= generic_file_llseek
,
1303 #ifdef CONFIG_SCHED_DEBUG
1305 * Print out various scheduling related per-task fields:
1307 static int sched_show(struct seq_file
*m
, void *v
)
1309 struct inode
*inode
= m
->private;
1310 struct task_struct
*p
;
1312 p
= get_proc_task(inode
);
1315 proc_sched_show_task(p
, m
);
1323 sched_write(struct file
*file
, const char __user
*buf
,
1324 size_t count
, loff_t
*offset
)
1326 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1327 struct task_struct
*p
;
1329 p
= get_proc_task(inode
);
1332 proc_sched_set_task(p
);
1339 static int sched_open(struct inode
*inode
, struct file
*filp
)
1343 ret
= single_open(filp
, sched_show
, NULL
);
1345 struct seq_file
*m
= filp
->private_data
;
1352 static const struct file_operations proc_pid_sched_operations
= {
1355 .write
= sched_write
,
1356 .llseek
= seq_lseek
,
1357 .release
= single_release
,
1362 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1363 size_t count
, loff_t
*offset
)
1365 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1366 struct task_struct
*p
;
1367 char buffer
[TASK_COMM_LEN
];
1369 memset(buffer
, 0, sizeof(buffer
));
1370 if (count
> sizeof(buffer
) - 1)
1371 count
= sizeof(buffer
) - 1;
1372 if (copy_from_user(buffer
, buf
, count
))
1375 p
= get_proc_task(inode
);
1379 if (same_thread_group(current
, p
))
1380 set_task_comm(p
, buffer
);
1389 static int comm_show(struct seq_file
*m
, void *v
)
1391 struct inode
*inode
= m
->private;
1392 struct task_struct
*p
;
1394 p
= get_proc_task(inode
);
1399 seq_printf(m
, "%s\n", p
->comm
);
1407 static int comm_open(struct inode
*inode
, struct file
*filp
)
1411 ret
= single_open(filp
, comm_show
, NULL
);
1413 struct seq_file
*m
= filp
->private_data
;
1420 static const struct file_operations proc_pid_set_comm_operations
= {
1423 .write
= comm_write
,
1424 .llseek
= seq_lseek
,
1425 .release
= single_release
,
1429 * We added or removed a vma mapping the executable. The vmas are only mapped
1430 * during exec and are not mapped with the mmap system call.
1431 * Callers must hold down_write() on the mm's mmap_sem for these
1433 void added_exe_file_vma(struct mm_struct
*mm
)
1435 mm
->num_exe_file_vmas
++;
1438 void removed_exe_file_vma(struct mm_struct
*mm
)
1440 mm
->num_exe_file_vmas
--;
1441 if ((mm
->num_exe_file_vmas
== 0) && mm
->exe_file
){
1443 mm
->exe_file
= NULL
;
1448 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
1451 get_file(new_exe_file
);
1454 mm
->exe_file
= new_exe_file
;
1455 mm
->num_exe_file_vmas
= 0;
1458 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
1460 struct file
*exe_file
;
1462 /* We need mmap_sem to protect against races with removal of
1463 * VM_EXECUTABLE vmas */
1464 down_read(&mm
->mmap_sem
);
1465 exe_file
= mm
->exe_file
;
1468 up_read(&mm
->mmap_sem
);
1472 void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
1474 /* It's safe to write the exe_file pointer without exe_file_lock because
1475 * this is called during fork when the task is not yet in /proc */
1476 newmm
->exe_file
= get_mm_exe_file(oldmm
);
1479 static int proc_exe_link(struct inode
*inode
, struct path
*exe_path
)
1481 struct task_struct
*task
;
1482 struct mm_struct
*mm
;
1483 struct file
*exe_file
;
1485 task
= get_proc_task(inode
);
1488 mm
= get_task_mm(task
);
1489 put_task_struct(task
);
1492 exe_file
= get_mm_exe_file(mm
);
1495 *exe_path
= exe_file
->f_path
;
1496 path_get(&exe_file
->f_path
);
1503 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1505 struct inode
*inode
= dentry
->d_inode
;
1506 int error
= -EACCES
;
1508 /* We don't need a base pointer in the /proc filesystem */
1509 path_put(&nd
->path
);
1511 /* Are we allowed to snoop on the tasks file descriptors? */
1512 if (!proc_fd_access_allowed(inode
))
1515 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->path
);
1517 return ERR_PTR(error
);
1520 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1522 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1529 pathname
= d_path_with_unreachable(path
, tmp
, PAGE_SIZE
);
1530 len
= PTR_ERR(pathname
);
1531 if (IS_ERR(pathname
))
1533 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1537 if (copy_to_user(buffer
, pathname
, len
))
1540 free_page((unsigned long)tmp
);
1544 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1546 int error
= -EACCES
;
1547 struct inode
*inode
= dentry
->d_inode
;
1550 /* Are we allowed to snoop on the tasks file descriptors? */
1551 if (!proc_fd_access_allowed(inode
))
1554 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &path
);
1558 error
= do_proc_readlink(&path
, buffer
, buflen
);
1564 static const struct inode_operations proc_pid_link_inode_operations
= {
1565 .readlink
= proc_pid_readlink
,
1566 .follow_link
= proc_pid_follow_link
,
1567 .setattr
= proc_setattr
,
1571 /* building an inode */
1573 static int task_dumpable(struct task_struct
*task
)
1576 struct mm_struct
*mm
;
1581 dumpable
= get_dumpable(mm
);
1589 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1591 struct inode
* inode
;
1592 struct proc_inode
*ei
;
1593 const struct cred
*cred
;
1595 /* We need a new inode */
1597 inode
= new_inode(sb
);
1603 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1604 inode
->i_op
= &proc_def_inode_operations
;
1607 * grab the reference to task.
1609 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1613 if (task_dumpable(task
)) {
1615 cred
= __task_cred(task
);
1616 inode
->i_uid
= cred
->euid
;
1617 inode
->i_gid
= cred
->egid
;
1620 security_task_to_inode(task
, inode
);
1630 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1632 struct inode
*inode
= dentry
->d_inode
;
1633 struct task_struct
*task
;
1634 const struct cred
*cred
;
1636 generic_fillattr(inode
, stat
);
1641 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1643 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1644 task_dumpable(task
)) {
1645 cred
= __task_cred(task
);
1646 stat
->uid
= cred
->euid
;
1647 stat
->gid
= cred
->egid
;
1657 * Exceptional case: normally we are not allowed to unhash a busy
1658 * directory. In this case, however, we can do it - no aliasing problems
1659 * due to the way we treat inodes.
1661 * Rewrite the inode's ownerships here because the owning task may have
1662 * performed a setuid(), etc.
1664 * Before the /proc/pid/status file was created the only way to read
1665 * the effective uid of a /process was to stat /proc/pid. Reading
1666 * /proc/pid/status is slow enough that procps and other packages
1667 * kept stating /proc/pid. To keep the rules in /proc simple I have
1668 * made this apply to all per process world readable and executable
1671 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1673 struct inode
*inode
= dentry
->d_inode
;
1674 struct task_struct
*task
= get_proc_task(inode
);
1675 const struct cred
*cred
;
1678 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1679 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
);
1698 static int pid_delete_dentry(struct dentry
* dentry
)
1700 /* Is the task we represent dead?
1701 * If so, then don't put the dentry on the lru list,
1702 * kill it immediately.
1704 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1707 static const struct dentry_operations pid_dentry_operations
=
1709 .d_revalidate
= pid_revalidate
,
1710 .d_delete
= pid_delete_dentry
,
1715 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1716 struct task_struct
*, const void *);
1719 * Fill a directory entry.
1721 * If possible create the dcache entry and derive our inode number and
1722 * file type from dcache entry.
1724 * Since all of the proc inode numbers are dynamically generated, the inode
1725 * numbers do not exist until the inode is cache. This means creating the
1726 * the dcache entry in readdir is necessary to keep the inode numbers
1727 * reported by readdir in sync with the inode numbers reported
1730 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1731 char *name
, int len
,
1732 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1734 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1735 struct inode
*inode
;
1738 unsigned type
= DT_UNKNOWN
;
1742 qname
.hash
= full_name_hash(name
, len
);
1744 child
= d_lookup(dir
, &qname
);
1747 new = d_alloc(dir
, &qname
);
1749 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1756 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1757 goto end_instantiate
;
1758 inode
= child
->d_inode
;
1761 type
= inode
->i_mode
>> 12;
1766 ino
= find_inode_number(dir
, &qname
);
1769 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1772 static unsigned name_to_int(struct dentry
*dentry
)
1774 const char *name
= dentry
->d_name
.name
;
1775 int len
= dentry
->d_name
.len
;
1778 if (len
> 1 && *name
== '0')
1781 unsigned c
= *name
++ - '0';
1784 if (n
>= (~0U-9)/10)
1794 #define PROC_FDINFO_MAX 64
1796 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1798 struct task_struct
*task
= get_proc_task(inode
);
1799 struct files_struct
*files
= NULL
;
1801 int fd
= proc_fd(inode
);
1804 files
= get_files_struct(task
);
1805 put_task_struct(task
);
1809 * We are not taking a ref to the file structure, so we must
1812 spin_lock(&files
->file_lock
);
1813 file
= fcheck_files(files
, fd
);
1816 *path
= file
->f_path
;
1817 path_get(&file
->f_path
);
1820 snprintf(info
, PROC_FDINFO_MAX
,
1823 (long long) file
->f_pos
,
1825 spin_unlock(&files
->file_lock
);
1826 put_files_struct(files
);
1829 spin_unlock(&files
->file_lock
);
1830 put_files_struct(files
);
1835 static int proc_fd_link(struct inode
*inode
, struct path
*path
)
1837 return proc_fd_info(inode
, path
, NULL
);
1840 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1842 struct inode
*inode
= dentry
->d_inode
;
1843 struct task_struct
*task
= get_proc_task(inode
);
1844 int fd
= proc_fd(inode
);
1845 struct files_struct
*files
;
1846 const struct cred
*cred
;
1849 files
= get_files_struct(task
);
1852 if (fcheck_files(files
, fd
)) {
1854 put_files_struct(files
);
1855 if (task_dumpable(task
)) {
1857 cred
= __task_cred(task
);
1858 inode
->i_uid
= cred
->euid
;
1859 inode
->i_gid
= cred
->egid
;
1865 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1866 security_task_to_inode(task
, inode
);
1867 put_task_struct(task
);
1871 put_files_struct(files
);
1873 put_task_struct(task
);
1879 static const struct dentry_operations tid_fd_dentry_operations
=
1881 .d_revalidate
= tid_fd_revalidate
,
1882 .d_delete
= pid_delete_dentry
,
1885 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1886 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1888 unsigned fd
= *(const unsigned *)ptr
;
1890 struct files_struct
*files
;
1891 struct inode
*inode
;
1892 struct proc_inode
*ei
;
1893 struct dentry
*error
= ERR_PTR(-ENOENT
);
1895 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1900 files
= get_files_struct(task
);
1903 inode
->i_mode
= S_IFLNK
;
1906 * We are not taking a ref to the file structure, so we must
1909 spin_lock(&files
->file_lock
);
1910 file
= fcheck_files(files
, fd
);
1913 if (file
->f_mode
& FMODE_READ
)
1914 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1915 if (file
->f_mode
& FMODE_WRITE
)
1916 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1917 spin_unlock(&files
->file_lock
);
1918 put_files_struct(files
);
1920 inode
->i_op
= &proc_pid_link_inode_operations
;
1922 ei
->op
.proc_get_link
= proc_fd_link
;
1923 dentry
->d_op
= &tid_fd_dentry_operations
;
1924 d_add(dentry
, inode
);
1925 /* Close the race of the process dying before we return the dentry */
1926 if (tid_fd_revalidate(dentry
, NULL
))
1932 spin_unlock(&files
->file_lock
);
1933 put_files_struct(files
);
1939 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1940 struct dentry
*dentry
,
1941 instantiate_t instantiate
)
1943 struct task_struct
*task
= get_proc_task(dir
);
1944 unsigned fd
= name_to_int(dentry
);
1945 struct dentry
*result
= ERR_PTR(-ENOENT
);
1952 result
= instantiate(dir
, dentry
, task
, &fd
);
1954 put_task_struct(task
);
1959 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1960 filldir_t filldir
, instantiate_t instantiate
)
1962 struct dentry
*dentry
= filp
->f_path
.dentry
;
1963 struct inode
*inode
= dentry
->d_inode
;
1964 struct task_struct
*p
= get_proc_task(inode
);
1965 unsigned int fd
, ino
;
1967 struct files_struct
* files
;
1977 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1981 ino
= parent_ino(dentry
);
1982 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1986 files
= get_files_struct(p
);
1990 for (fd
= filp
->f_pos
-2;
1991 fd
< files_fdtable(files
)->max_fds
;
1992 fd
++, filp
->f_pos
++) {
1993 char name
[PROC_NUMBUF
];
1996 if (!fcheck_files(files
, fd
))
2000 len
= snprintf(name
, sizeof(name
), "%d", fd
);
2001 if (proc_fill_cache(filp
, dirent
, filldir
,
2002 name
, len
, instantiate
,
2010 put_files_struct(files
);
2018 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
2019 struct nameidata
*nd
)
2021 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
2024 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
2026 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
2029 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
2030 size_t len
, loff_t
*ppos
)
2032 char tmp
[PROC_FDINFO_MAX
];
2033 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
2035 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
2039 static const struct file_operations proc_fdinfo_file_operations
= {
2040 .open
= nonseekable_open
,
2041 .read
= proc_fdinfo_read
,
2044 static const struct file_operations proc_fd_operations
= {
2045 .read
= generic_read_dir
,
2046 .readdir
= proc_readfd
,
2050 * /proc/pid/fd needs a special permission handler so that a process can still
2051 * access /proc/self/fd after it has executed a setuid().
2053 static int proc_fd_permission(struct inode
*inode
, int mask
)
2057 rv
= generic_permission(inode
, mask
, NULL
);
2060 if (task_pid(current
) == proc_pid(inode
))
2066 * proc directories can do almost nothing..
2068 static const struct inode_operations proc_fd_inode_operations
= {
2069 .lookup
= proc_lookupfd
,
2070 .permission
= proc_fd_permission
,
2071 .setattr
= proc_setattr
,
2074 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
2075 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2077 unsigned fd
= *(unsigned *)ptr
;
2078 struct inode
*inode
;
2079 struct proc_inode
*ei
;
2080 struct dentry
*error
= ERR_PTR(-ENOENT
);
2082 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2087 inode
->i_mode
= S_IFREG
| S_IRUSR
;
2088 inode
->i_fop
= &proc_fdinfo_file_operations
;
2089 dentry
->d_op
= &tid_fd_dentry_operations
;
2090 d_add(dentry
, inode
);
2091 /* Close the race of the process dying before we return the dentry */
2092 if (tid_fd_revalidate(dentry
, NULL
))
2099 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
2100 struct dentry
*dentry
,
2101 struct nameidata
*nd
)
2103 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
2106 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
2108 return proc_readfd_common(filp
, dirent
, filldir
,
2109 proc_fdinfo_instantiate
);
2112 static const struct file_operations proc_fdinfo_operations
= {
2113 .read
= generic_read_dir
,
2114 .readdir
= proc_readfdinfo
,
2118 * proc directories can do almost nothing..
2120 static const struct inode_operations proc_fdinfo_inode_operations
= {
2121 .lookup
= proc_lookupfdinfo
,
2122 .setattr
= proc_setattr
,
2126 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
2127 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2129 const struct pid_entry
*p
= ptr
;
2130 struct inode
*inode
;
2131 struct proc_inode
*ei
;
2132 struct dentry
*error
= ERR_PTR(-ENOENT
);
2134 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2139 inode
->i_mode
= p
->mode
;
2140 if (S_ISDIR(inode
->i_mode
))
2141 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
2143 inode
->i_op
= p
->iop
;
2145 inode
->i_fop
= p
->fop
;
2147 dentry
->d_op
= &pid_dentry_operations
;
2148 d_add(dentry
, inode
);
2149 /* Close the race of the process dying before we return the dentry */
2150 if (pid_revalidate(dentry
, NULL
))
2156 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2157 struct dentry
*dentry
,
2158 const struct pid_entry
*ents
,
2161 struct dentry
*error
;
2162 struct task_struct
*task
= get_proc_task(dir
);
2163 const struct pid_entry
*p
, *last
;
2165 error
= ERR_PTR(-ENOENT
);
2171 * Yes, it does not scale. And it should not. Don't add
2172 * new entries into /proc/<tgid>/ without very good reasons.
2174 last
= &ents
[nents
- 1];
2175 for (p
= ents
; p
<= last
; p
++) {
2176 if (p
->len
!= dentry
->d_name
.len
)
2178 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2184 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2186 put_task_struct(task
);
2191 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2192 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2194 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2195 proc_pident_instantiate
, task
, p
);
2198 static int proc_pident_readdir(struct file
*filp
,
2199 void *dirent
, filldir_t filldir
,
2200 const struct pid_entry
*ents
, unsigned int nents
)
2203 struct dentry
*dentry
= filp
->f_path
.dentry
;
2204 struct inode
*inode
= dentry
->d_inode
;
2205 struct task_struct
*task
= get_proc_task(inode
);
2206 const struct pid_entry
*p
, *last
;
2219 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2225 ino
= parent_ino(dentry
);
2226 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2238 last
= &ents
[nents
- 1];
2240 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2249 put_task_struct(task
);
2254 #ifdef CONFIG_SECURITY
2255 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2256 size_t count
, loff_t
*ppos
)
2258 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2261 struct task_struct
*task
= get_proc_task(inode
);
2266 length
= security_getprocattr(task
,
2267 (char*)file
->f_path
.dentry
->d_name
.name
,
2269 put_task_struct(task
);
2271 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2276 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2277 size_t count
, loff_t
*ppos
)
2279 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2282 struct task_struct
*task
= get_proc_task(inode
);
2287 if (count
> PAGE_SIZE
)
2290 /* No partial writes. */
2296 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2301 if (copy_from_user(page
, buf
, count
))
2304 /* Guard against adverse ptrace interaction */
2305 length
= mutex_lock_interruptible(&task
->cred_guard_mutex
);
2309 length
= security_setprocattr(task
,
2310 (char*)file
->f_path
.dentry
->d_name
.name
,
2311 (void*)page
, count
);
2312 mutex_unlock(&task
->cred_guard_mutex
);
2314 free_page((unsigned long) page
);
2316 put_task_struct(task
);
2321 static const struct file_operations proc_pid_attr_operations
= {
2322 .read
= proc_pid_attr_read
,
2323 .write
= proc_pid_attr_write
,
2324 .llseek
= generic_file_llseek
,
2327 static const struct pid_entry attr_dir_stuff
[] = {
2328 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2329 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2330 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2331 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2332 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2333 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2336 static int proc_attr_dir_readdir(struct file
* filp
,
2337 void * dirent
, filldir_t filldir
)
2339 return proc_pident_readdir(filp
,dirent
,filldir
,
2340 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2343 static const struct file_operations proc_attr_dir_operations
= {
2344 .read
= generic_read_dir
,
2345 .readdir
= proc_attr_dir_readdir
,
2348 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2349 struct dentry
*dentry
, struct nameidata
*nd
)
2351 return proc_pident_lookup(dir
, dentry
,
2352 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2355 static const struct inode_operations proc_attr_dir_inode_operations
= {
2356 .lookup
= proc_attr_dir_lookup
,
2357 .getattr
= pid_getattr
,
2358 .setattr
= proc_setattr
,
2363 #ifdef CONFIG_ELF_CORE
2364 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2365 size_t count
, loff_t
*ppos
)
2367 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2368 struct mm_struct
*mm
;
2369 char buffer
[PROC_NUMBUF
];
2377 mm
= get_task_mm(task
);
2379 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2380 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2381 MMF_DUMP_FILTER_SHIFT
));
2383 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2386 put_task_struct(task
);
2391 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2392 const char __user
*buf
,
2396 struct task_struct
*task
;
2397 struct mm_struct
*mm
;
2398 char buffer
[PROC_NUMBUF
], *end
;
2405 memset(buffer
, 0, sizeof(buffer
));
2406 if (count
> sizeof(buffer
) - 1)
2407 count
= sizeof(buffer
) - 1;
2408 if (copy_from_user(buffer
, buf
, count
))
2412 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2415 if (end
- buffer
== 0)
2419 task
= get_proc_task(file
->f_dentry
->d_inode
);
2424 mm
= get_task_mm(task
);
2428 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2430 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2432 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2437 put_task_struct(task
);
2442 static const struct file_operations proc_coredump_filter_operations
= {
2443 .read
= proc_coredump_filter_read
,
2444 .write
= proc_coredump_filter_write
,
2445 .llseek
= generic_file_llseek
,
2452 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2455 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2456 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2457 char tmp
[PROC_NUMBUF
];
2460 sprintf(tmp
, "%d", tgid
);
2461 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2464 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2466 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2467 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2468 char *name
= ERR_PTR(-ENOENT
);
2472 name
= ERR_PTR(-ENOMEM
);
2474 sprintf(name
, "%d", tgid
);
2476 nd_set_link(nd
, name
);
2480 static void proc_self_put_link(struct dentry
*dentry
, struct nameidata
*nd
,
2483 char *s
= nd_get_link(nd
);
2488 static const struct inode_operations proc_self_inode_operations
= {
2489 .readlink
= proc_self_readlink
,
2490 .follow_link
= proc_self_follow_link
,
2491 .put_link
= proc_self_put_link
,
2497 * These are the directory entries in the root directory of /proc
2498 * that properly belong to the /proc filesystem, as they describe
2499 * describe something that is process related.
2501 static const struct pid_entry proc_base_stuff
[] = {
2502 NOD("self", S_IFLNK
|S_IRWXUGO
,
2503 &proc_self_inode_operations
, NULL
, {}),
2507 * Exceptional case: normally we are not allowed to unhash a busy
2508 * directory. In this case, however, we can do it - no aliasing problems
2509 * due to the way we treat inodes.
2511 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
2513 struct inode
*inode
= dentry
->d_inode
;
2514 struct task_struct
*task
= get_proc_task(inode
);
2516 put_task_struct(task
);
2523 static const struct dentry_operations proc_base_dentry_operations
=
2525 .d_revalidate
= proc_base_revalidate
,
2526 .d_delete
= pid_delete_dentry
,
2529 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2530 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2532 const struct pid_entry
*p
= ptr
;
2533 struct inode
*inode
;
2534 struct proc_inode
*ei
;
2535 struct dentry
*error
;
2537 /* Allocate the inode */
2538 error
= ERR_PTR(-ENOMEM
);
2539 inode
= new_inode(dir
->i_sb
);
2543 /* Initialize the inode */
2545 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2548 * grab the reference to the task.
2550 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2554 inode
->i_mode
= p
->mode
;
2555 if (S_ISDIR(inode
->i_mode
))
2557 if (S_ISLNK(inode
->i_mode
))
2560 inode
->i_op
= p
->iop
;
2562 inode
->i_fop
= p
->fop
;
2564 dentry
->d_op
= &proc_base_dentry_operations
;
2565 d_add(dentry
, inode
);
2574 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2576 struct dentry
*error
;
2577 struct task_struct
*task
= get_proc_task(dir
);
2578 const struct pid_entry
*p
, *last
;
2580 error
= ERR_PTR(-ENOENT
);
2585 /* Lookup the directory entry */
2586 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2587 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2588 if (p
->len
!= dentry
->d_name
.len
)
2590 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2596 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2599 put_task_struct(task
);
2604 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2605 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2607 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2608 proc_base_instantiate
, task
, p
);
2611 #ifdef CONFIG_TASK_IO_ACCOUNTING
2612 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2614 struct task_io_accounting acct
= task
->ioac
;
2615 unsigned long flags
;
2617 if (whole
&& lock_task_sighand(task
, &flags
)) {
2618 struct task_struct
*t
= task
;
2620 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2621 while_each_thread(task
, t
)
2622 task_io_accounting_add(&acct
, &t
->ioac
);
2624 unlock_task_sighand(task
, &flags
);
2626 return sprintf(buffer
,
2631 "read_bytes: %llu\n"
2632 "write_bytes: %llu\n"
2633 "cancelled_write_bytes: %llu\n",
2634 (unsigned long long)acct
.rchar
,
2635 (unsigned long long)acct
.wchar
,
2636 (unsigned long long)acct
.syscr
,
2637 (unsigned long long)acct
.syscw
,
2638 (unsigned long long)acct
.read_bytes
,
2639 (unsigned long long)acct
.write_bytes
,
2640 (unsigned long long)acct
.cancelled_write_bytes
);
2643 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2645 return do_io_accounting(task
, buffer
, 0);
2648 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2650 return do_io_accounting(task
, buffer
, 1);
2652 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2654 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
2655 struct pid
*pid
, struct task_struct
*task
)
2657 seq_printf(m
, "%08x\n", task
->personality
);
2664 static const struct file_operations proc_task_operations
;
2665 static const struct inode_operations proc_task_inode_operations
;
2667 static const struct pid_entry tgid_base_stuff
[] = {
2668 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
2669 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
2670 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
2672 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
2674 REG("environ", S_IRUSR
, proc_environ_operations
),
2675 INF("auxv", S_IRUSR
, proc_pid_auxv
),
2676 ONE("status", S_IRUGO
, proc_pid_status
),
2677 ONE("personality", S_IRUSR
, proc_pid_personality
),
2678 INF("limits", S_IRUSR
, proc_pid_limits
),
2679 #ifdef CONFIG_SCHED_DEBUG
2680 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
2682 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
2683 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2684 INF("syscall", S_IRUSR
, proc_pid_syscall
),
2686 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
2687 ONE("stat", S_IRUGO
, proc_tgid_stat
),
2688 ONE("statm", S_IRUGO
, proc_pid_statm
),
2689 REG("maps", S_IRUGO
, proc_maps_operations
),
2691 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
2693 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
2694 LNK("cwd", proc_cwd_link
),
2695 LNK("root", proc_root_link
),
2696 LNK("exe", proc_exe_link
),
2697 REG("mounts", S_IRUGO
, proc_mounts_operations
),
2698 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
2699 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
2700 #ifdef CONFIG_PROC_PAGE_MONITOR
2701 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
2702 REG("smaps", S_IRUGO
, proc_smaps_operations
),
2703 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
2705 #ifdef CONFIG_SECURITY
2706 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
2708 #ifdef CONFIG_KALLSYMS
2709 INF("wchan", S_IRUGO
, proc_pid_wchan
),
2711 #ifdef CONFIG_STACKTRACE
2712 ONE("stack", S_IRUSR
, proc_pid_stack
),
2714 #ifdef CONFIG_SCHEDSTATS
2715 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
2717 #ifdef CONFIG_LATENCYTOP
2718 REG("latency", S_IRUGO
, proc_lstats_operations
),
2720 #ifdef CONFIG_PROC_PID_CPUSET
2721 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
2723 #ifdef CONFIG_CGROUPS
2724 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
2726 INF("oom_score", S_IRUGO
, proc_oom_score
),
2727 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
2728 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
2729 #ifdef CONFIG_AUDITSYSCALL
2730 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
2731 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
2733 #ifdef CONFIG_FAULT_INJECTION
2734 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
2736 #ifdef CONFIG_ELF_CORE
2737 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
2739 #ifdef CONFIG_TASK_IO_ACCOUNTING
2740 INF("io", S_IRUGO
, proc_tgid_io_accounting
),
2744 static int proc_tgid_base_readdir(struct file
* filp
,
2745 void * dirent
, filldir_t filldir
)
2747 return proc_pident_readdir(filp
,dirent
,filldir
,
2748 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2751 static const struct file_operations proc_tgid_base_operations
= {
2752 .read
= generic_read_dir
,
2753 .readdir
= proc_tgid_base_readdir
,
2756 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2757 return proc_pident_lookup(dir
, dentry
,
2758 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2761 static const struct inode_operations proc_tgid_base_inode_operations
= {
2762 .lookup
= proc_tgid_base_lookup
,
2763 .getattr
= pid_getattr
,
2764 .setattr
= proc_setattr
,
2767 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2769 struct dentry
*dentry
, *leader
, *dir
;
2770 char buf
[PROC_NUMBUF
];
2774 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2775 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2777 shrink_dcache_parent(dentry
);
2783 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2784 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2789 name
.len
= strlen(name
.name
);
2790 dir
= d_hash_and_lookup(leader
, &name
);
2792 goto out_put_leader
;
2795 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2796 dentry
= d_hash_and_lookup(dir
, &name
);
2798 shrink_dcache_parent(dentry
);
2811 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2812 * @task: task that should be flushed.
2814 * When flushing dentries from proc, one needs to flush them from global
2815 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2816 * in. This call is supposed to do all of this job.
2818 * Looks in the dcache for
2820 * /proc/@tgid/task/@pid
2821 * if either directory is present flushes it and all of it'ts children
2824 * It is safe and reasonable to cache /proc entries for a task until
2825 * that task exits. After that they just clog up the dcache with
2826 * useless entries, possibly causing useful dcache entries to be
2827 * flushed instead. This routine is proved to flush those useless
2828 * dcache entries at process exit time.
2830 * NOTE: This routine is just an optimization so it does not guarantee
2831 * that no dcache entries will exist at process exit time it
2832 * just makes it very unlikely that any will persist.
2835 void proc_flush_task(struct task_struct
*task
)
2838 struct pid
*pid
, *tgid
;
2841 pid
= task_pid(task
);
2842 tgid
= task_tgid(task
);
2844 for (i
= 0; i
<= pid
->level
; i
++) {
2845 upid
= &pid
->numbers
[i
];
2846 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2847 tgid
->numbers
[i
].nr
);
2850 upid
= &pid
->numbers
[pid
->level
];
2852 pid_ns_release_proc(upid
->ns
);
2855 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2856 struct dentry
* dentry
,
2857 struct task_struct
*task
, const void *ptr
)
2859 struct dentry
*error
= ERR_PTR(-ENOENT
);
2860 struct inode
*inode
;
2862 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2866 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2867 inode
->i_op
= &proc_tgid_base_inode_operations
;
2868 inode
->i_fop
= &proc_tgid_base_operations
;
2869 inode
->i_flags
|=S_IMMUTABLE
;
2871 inode
->i_nlink
= 2 + pid_entry_count_dirs(tgid_base_stuff
,
2872 ARRAY_SIZE(tgid_base_stuff
));
2874 dentry
->d_op
= &pid_dentry_operations
;
2876 d_add(dentry
, inode
);
2877 /* Close the race of the process dying before we return the dentry */
2878 if (pid_revalidate(dentry
, NULL
))
2884 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2886 struct dentry
*result
;
2887 struct task_struct
*task
;
2889 struct pid_namespace
*ns
;
2891 result
= proc_base_lookup(dir
, dentry
);
2892 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2895 tgid
= name_to_int(dentry
);
2899 ns
= dentry
->d_sb
->s_fs_info
;
2901 task
= find_task_by_pid_ns(tgid
, ns
);
2903 get_task_struct(task
);
2908 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2909 put_task_struct(task
);
2915 * Find the first task with tgid >= tgid
2920 struct task_struct
*task
;
2922 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
2927 put_task_struct(iter
.task
);
2931 pid
= find_ge_pid(iter
.tgid
, ns
);
2933 iter
.tgid
= pid_nr_ns(pid
, ns
);
2934 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
2935 /* What we to know is if the pid we have find is the
2936 * pid of a thread_group_leader. Testing for task
2937 * being a thread_group_leader is the obvious thing
2938 * todo but there is a window when it fails, due to
2939 * the pid transfer logic in de_thread.
2941 * So we perform the straight forward test of seeing
2942 * if the pid we have found is the pid of a thread
2943 * group leader, and don't worry if the task we have
2944 * found doesn't happen to be a thread group leader.
2945 * As we don't care in the case of readdir.
2947 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
2951 get_task_struct(iter
.task
);
2957 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2959 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2960 struct tgid_iter iter
)
2962 char name
[PROC_NUMBUF
];
2963 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
2964 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2965 proc_pid_instantiate
, iter
.task
, NULL
);
2968 /* for the /proc/ directory itself, after non-process stuff has been done */
2969 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2971 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2972 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2973 struct tgid_iter iter
;
2974 struct pid_namespace
*ns
;
2979 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2980 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2981 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2985 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2987 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
2988 for (iter
= next_tgid(ns
, iter
);
2990 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
2991 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
2992 if (proc_pid_fill_cache(filp
, dirent
, filldir
, iter
) < 0) {
2993 put_task_struct(iter
.task
);
2997 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2999 put_task_struct(reaper
);
3007 static const struct pid_entry tid_base_stuff
[] = {
3008 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3009 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3010 REG("environ", S_IRUSR
, proc_environ_operations
),
3011 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3012 ONE("status", S_IRUGO
, proc_pid_status
),
3013 ONE("personality", S_IRUSR
, proc_pid_personality
),
3014 INF("limits", S_IRUSR
, proc_pid_limits
),
3015 #ifdef CONFIG_SCHED_DEBUG
3016 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3018 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3019 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3020 INF("syscall", S_IRUSR
, proc_pid_syscall
),
3022 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3023 ONE("stat", S_IRUGO
, proc_tid_stat
),
3024 ONE("statm", S_IRUGO
, proc_pid_statm
),
3025 REG("maps", S_IRUGO
, proc_maps_operations
),
3027 REG("numa_maps", S_IRUGO
, proc_numa_maps_operations
),
3029 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3030 LNK("cwd", proc_cwd_link
),
3031 LNK("root", proc_root_link
),
3032 LNK("exe", proc_exe_link
),
3033 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3034 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3035 #ifdef CONFIG_PROC_PAGE_MONITOR
3036 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3037 REG("smaps", S_IRUGO
, proc_smaps_operations
),
3038 REG("pagemap", S_IRUSR
, proc_pagemap_operations
),
3040 #ifdef CONFIG_SECURITY
3041 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3043 #ifdef CONFIG_KALLSYMS
3044 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3046 #ifdef CONFIG_STACKTRACE
3047 ONE("stack", S_IRUSR
, proc_pid_stack
),
3049 #ifdef CONFIG_SCHEDSTATS
3050 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3052 #ifdef CONFIG_LATENCYTOP
3053 REG("latency", S_IRUGO
, proc_lstats_operations
),
3055 #ifdef CONFIG_PROC_PID_CPUSET
3056 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3058 #ifdef CONFIG_CGROUPS
3059 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3061 INF("oom_score", S_IRUGO
, proc_oom_score
),
3062 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3063 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3064 #ifdef CONFIG_AUDITSYSCALL
3065 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3066 REG("sessionid", S_IRUSR
, proc_sessionid_operations
),
3068 #ifdef CONFIG_FAULT_INJECTION
3069 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3071 #ifdef CONFIG_TASK_IO_ACCOUNTING
3072 INF("io", S_IRUGO
, proc_tid_io_accounting
),
3076 static int proc_tid_base_readdir(struct file
* filp
,
3077 void * dirent
, filldir_t filldir
)
3079 return proc_pident_readdir(filp
,dirent
,filldir
,
3080 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
3083 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
3084 return proc_pident_lookup(dir
, dentry
,
3085 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3088 static const struct file_operations proc_tid_base_operations
= {
3089 .read
= generic_read_dir
,
3090 .readdir
= proc_tid_base_readdir
,
3093 static const struct inode_operations proc_tid_base_inode_operations
= {
3094 .lookup
= proc_tid_base_lookup
,
3095 .getattr
= pid_getattr
,
3096 .setattr
= proc_setattr
,
3099 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
3100 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3102 struct dentry
*error
= ERR_PTR(-ENOENT
);
3103 struct inode
*inode
;
3104 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3108 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3109 inode
->i_op
= &proc_tid_base_inode_operations
;
3110 inode
->i_fop
= &proc_tid_base_operations
;
3111 inode
->i_flags
|=S_IMMUTABLE
;
3113 inode
->i_nlink
= 2 + pid_entry_count_dirs(tid_base_stuff
,
3114 ARRAY_SIZE(tid_base_stuff
));
3116 dentry
->d_op
= &pid_dentry_operations
;
3118 d_add(dentry
, inode
);
3119 /* Close the race of the process dying before we return the dentry */
3120 if (pid_revalidate(dentry
, NULL
))
3126 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
3128 struct dentry
*result
= ERR_PTR(-ENOENT
);
3129 struct task_struct
*task
;
3130 struct task_struct
*leader
= get_proc_task(dir
);
3132 struct pid_namespace
*ns
;
3137 tid
= name_to_int(dentry
);
3141 ns
= dentry
->d_sb
->s_fs_info
;
3143 task
= find_task_by_pid_ns(tid
, ns
);
3145 get_task_struct(task
);
3149 if (!same_thread_group(leader
, task
))
3152 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3154 put_task_struct(task
);
3156 put_task_struct(leader
);
3162 * Find the first tid of a thread group to return to user space.
3164 * Usually this is just the thread group leader, but if the users
3165 * buffer was too small or there was a seek into the middle of the
3166 * directory we have more work todo.
3168 * In the case of a short read we start with find_task_by_pid.
3170 * In the case of a seek we start with the leader and walk nr
3173 static struct task_struct
*first_tid(struct task_struct
*leader
,
3174 int tid
, int nr
, struct pid_namespace
*ns
)
3176 struct task_struct
*pos
;
3179 /* Attempt to start with the pid of a thread */
3180 if (tid
&& (nr
> 0)) {
3181 pos
= find_task_by_pid_ns(tid
, ns
);
3182 if (pos
&& (pos
->group_leader
== leader
))
3186 /* If nr exceeds the number of threads there is nothing todo */
3188 if (nr
&& nr
>= get_nr_threads(leader
))
3191 /* If we haven't found our starting place yet start
3192 * with the leader and walk nr threads forward.
3194 for (pos
= leader
; nr
> 0; --nr
) {
3195 pos
= next_thread(pos
);
3196 if (pos
== leader
) {
3202 get_task_struct(pos
);
3209 * Find the next thread in the thread list.
3210 * Return NULL if there is an error or no next thread.
3212 * The reference to the input task_struct is released.
3214 static struct task_struct
*next_tid(struct task_struct
*start
)
3216 struct task_struct
*pos
= NULL
;
3218 if (pid_alive(start
)) {
3219 pos
= next_thread(start
);
3220 if (thread_group_leader(pos
))
3223 get_task_struct(pos
);
3226 put_task_struct(start
);
3230 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3231 struct task_struct
*task
, int tid
)
3233 char name
[PROC_NUMBUF
];
3234 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3235 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3236 proc_task_instantiate
, task
, NULL
);
3239 /* for the /proc/TGID/task/ directories */
3240 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3242 struct dentry
*dentry
= filp
->f_path
.dentry
;
3243 struct inode
*inode
= dentry
->d_inode
;
3244 struct task_struct
*leader
= NULL
;
3245 struct task_struct
*task
;
3246 int retval
= -ENOENT
;
3249 struct pid_namespace
*ns
;
3251 task
= get_proc_task(inode
);
3255 if (pid_alive(task
)) {
3256 leader
= task
->group_leader
;
3257 get_task_struct(leader
);
3260 put_task_struct(task
);
3265 switch ((unsigned long)filp
->f_pos
) {
3268 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3273 ino
= parent_ino(dentry
);
3274 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3280 /* f_version caches the tgid value that the last readdir call couldn't
3281 * return. lseek aka telldir automagically resets f_version to 0.
3283 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3284 tid
= (int)filp
->f_version
;
3285 filp
->f_version
= 0;
3286 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3288 task
= next_tid(task
), filp
->f_pos
++) {
3289 tid
= task_pid_nr_ns(task
, ns
);
3290 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3291 /* returning this tgid failed, save it as the first
3292 * pid for the next readir call */
3293 filp
->f_version
= (u64
)tid
;
3294 put_task_struct(task
);
3299 put_task_struct(leader
);
3304 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3306 struct inode
*inode
= dentry
->d_inode
;
3307 struct task_struct
*p
= get_proc_task(inode
);
3308 generic_fillattr(inode
, stat
);
3311 stat
->nlink
+= get_nr_threads(p
);
3318 static const struct inode_operations proc_task_inode_operations
= {
3319 .lookup
= proc_task_lookup
,
3320 .getattr
= proc_task_getattr
,
3321 .setattr
= proc_setattr
,
3324 static const struct file_operations proc_task_operations
= {
3325 .read
= generic_read_dir
,
3326 .readdir
= proc_task_readdir
,