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/user_namespace.h>
85 #include <linux/fs_struct.h>
86 #include <linux/slab.h>
87 #include <linux/flex_array.h>
88 #ifdef CONFIG_HARDWALL
89 #include <asm/hardwall.h>
91 #include <trace/events/oom.h>
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
108 const struct inode_operations
*iop
;
109 const struct file_operations
*fop
;
113 #define NOD(NAME, MODE, IOP, FOP, OP) { \
115 .len = sizeof(NAME) - 1, \
122 #define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124 #define LNK(NAME, get_link) \
125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
126 &proc_pid_link_inode_operations, NULL, \
127 { .proc_get_link = get_link } )
128 #define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130 #define INF(NAME, MODE, read) \
131 NOD(NAME, (S_IFREG|(MODE)), \
132 NULL, &proc_info_file_operations, \
133 { .proc_read = read } )
134 #define ONE(NAME, MODE, show) \
135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
137 { .proc_show = show } )
139 static int proc_fd_permission(struct inode
*inode
, int mask
);
142 * Count the number of hardlinks for the pid_entry table, excluding the .
145 static unsigned int pid_entry_count_dirs(const struct pid_entry
*entries
,
152 for (i
= 0; i
< n
; ++i
) {
153 if (S_ISDIR(entries
[i
].mode
))
160 static int get_task_root(struct task_struct
*task
, struct path
*root
)
162 int result
= -ENOENT
;
166 get_fs_root(task
->fs
, root
);
173 static int proc_cwd_link(struct dentry
*dentry
, struct path
*path
)
175 struct task_struct
*task
= get_proc_task(dentry
->d_inode
);
176 int result
= -ENOENT
;
181 get_fs_pwd(task
->fs
, path
);
185 put_task_struct(task
);
190 static int proc_root_link(struct dentry
*dentry
, struct path
*path
)
192 struct task_struct
*task
= get_proc_task(dentry
->d_inode
);
193 int result
= -ENOENT
;
196 result
= get_task_root(task
, path
);
197 put_task_struct(task
);
202 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
206 struct mm_struct
*mm
= get_task_mm(task
);
210 goto out_mm
; /* Shh! No looking before we're done */
212 len
= mm
->arg_end
- mm
->arg_start
;
217 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
222 len
= strnlen(buffer
, res
);
226 len
= mm
->env_end
- mm
->env_start
;
227 if (len
> PAGE_SIZE
- res
)
228 len
= PAGE_SIZE
- res
;
229 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
230 res
= strnlen(buffer
, res
);
239 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
241 struct mm_struct
*mm
= mm_access(task
, PTRACE_MODE_READ
);
242 int res
= PTR_ERR(mm
);
243 if (mm
&& !IS_ERR(mm
)) {
244 unsigned int nwords
= 0;
247 } while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
248 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
251 memcpy(buffer
, mm
->saved_auxv
, res
);
258 #ifdef CONFIG_KALLSYMS
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
263 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
266 char symname
[KSYM_NAME_LEN
];
268 wchan
= get_wchan(task
);
270 if (lookup_symbol_name(wchan
, symname
) < 0)
271 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
274 return sprintf(buffer
, "%lu", wchan
);
276 return sprintf(buffer
, "%s", symname
);
278 #endif /* CONFIG_KALLSYMS */
280 static int lock_trace(struct task_struct
*task
)
282 int err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
285 if (!ptrace_may_access(task
, PTRACE_MODE_ATTACH
)) {
286 mutex_unlock(&task
->signal
->cred_guard_mutex
);
292 static void unlock_trace(struct task_struct
*task
)
294 mutex_unlock(&task
->signal
->cred_guard_mutex
);
297 #ifdef CONFIG_STACKTRACE
299 #define MAX_STACK_TRACE_DEPTH 64
301 static int proc_pid_stack(struct seq_file
*m
, struct pid_namespace
*ns
,
302 struct pid
*pid
, struct task_struct
*task
)
304 struct stack_trace trace
;
305 unsigned long *entries
;
309 entries
= kmalloc(MAX_STACK_TRACE_DEPTH
* sizeof(*entries
), GFP_KERNEL
);
313 trace
.nr_entries
= 0;
314 trace
.max_entries
= MAX_STACK_TRACE_DEPTH
;
315 trace
.entries
= entries
;
318 err
= lock_trace(task
);
320 save_stack_trace_tsk(task
, &trace
);
322 for (i
= 0; i
< trace
.nr_entries
; i
++) {
323 seq_printf(m
, "[<%pK>] %pS\n",
324 (void *)entries
[i
], (void *)entries
[i
]);
334 #ifdef CONFIG_SCHEDSTATS
336 * Provides /proc/PID/schedstat
338 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
340 return sprintf(buffer
, "%llu %llu %lu\n",
341 (unsigned long long)task
->se
.sum_exec_runtime
,
342 (unsigned long long)task
->sched_info
.run_delay
,
343 task
->sched_info
.pcount
);
347 #ifdef CONFIG_LATENCYTOP
348 static int lstats_show_proc(struct seq_file
*m
, void *v
)
351 struct inode
*inode
= m
->private;
352 struct task_struct
*task
= get_proc_task(inode
);
356 seq_puts(m
, "Latency Top version : v0.1\n");
357 for (i
= 0; i
< 32; i
++) {
358 struct latency_record
*lr
= &task
->latency_record
[i
];
359 if (lr
->backtrace
[0]) {
361 seq_printf(m
, "%i %li %li",
362 lr
->count
, lr
->time
, lr
->max
);
363 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
364 unsigned long bt
= lr
->backtrace
[q
];
369 seq_printf(m
, " %ps", (void *)bt
);
375 put_task_struct(task
);
379 static int lstats_open(struct inode
*inode
, struct file
*file
)
381 return single_open(file
, lstats_show_proc
, inode
);
384 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
385 size_t count
, loff_t
*offs
)
387 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
391 clear_all_latency_tracing(task
);
392 put_task_struct(task
);
397 static const struct file_operations proc_lstats_operations
= {
400 .write
= lstats_write
,
402 .release
= single_release
,
407 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
409 unsigned long totalpages
= totalram_pages
+ total_swap_pages
;
410 unsigned long points
= 0;
412 read_lock(&tasklist_lock
);
414 points
= oom_badness(task
, NULL
, NULL
, totalpages
) *
416 read_unlock(&tasklist_lock
);
417 return sprintf(buffer
, "%lu\n", points
);
425 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
426 [RLIMIT_CPU
] = {"Max cpu time", "seconds"},
427 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
428 [RLIMIT_DATA
] = {"Max data size", "bytes"},
429 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC
] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE
] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS
] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
440 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
441 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
444 /* Display limits for a process */
445 static int proc_pid_limits(struct task_struct
*task
, char *buffer
)
450 char *bufptr
= buffer
;
452 struct rlimit rlim
[RLIM_NLIMITS
];
454 if (!lock_task_sighand(task
, &flags
))
456 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
457 unlock_task_sighand(task
, &flags
);
460 * print the file header
462 count
+= sprintf(&bufptr
[count
], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
465 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
466 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
467 count
+= sprintf(&bufptr
[count
], "%-25s %-20s ",
468 lnames
[i
].name
, "unlimited");
470 count
+= sprintf(&bufptr
[count
], "%-25s %-20lu ",
471 lnames
[i
].name
, rlim
[i
].rlim_cur
);
473 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
474 count
+= sprintf(&bufptr
[count
], "%-20s ", "unlimited");
476 count
+= sprintf(&bufptr
[count
], "%-20lu ",
480 count
+= sprintf(&bufptr
[count
], "%-10s\n",
483 count
+= sprintf(&bufptr
[count
], "\n");
489 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490 static int proc_pid_syscall(struct task_struct
*task
, char *buffer
)
493 unsigned long args
[6], sp
, pc
;
494 int res
= lock_trace(task
);
498 if (task_current_syscall(task
, &nr
, args
, 6, &sp
, &pc
))
499 res
= sprintf(buffer
, "running\n");
501 res
= sprintf(buffer
, "%ld 0x%lx 0x%lx\n", nr
, sp
, pc
);
503 res
= sprintf(buffer
,
504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
506 args
[0], args
[1], args
[2], args
[3], args
[4], args
[5],
511 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
513 /************************************************************************/
514 /* Here the fs part begins */
515 /************************************************************************/
517 /* permission checks */
518 static int proc_fd_access_allowed(struct inode
*inode
)
520 struct task_struct
*task
;
522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
526 task
= get_proc_task(inode
);
528 allowed
= ptrace_may_access(task
, PTRACE_MODE_READ
);
529 put_task_struct(task
);
534 int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
537 struct inode
*inode
= dentry
->d_inode
;
539 if (attr
->ia_valid
& ATTR_MODE
)
542 error
= inode_change_ok(inode
, attr
);
546 if ((attr
->ia_valid
& ATTR_SIZE
) &&
547 attr
->ia_size
!= i_size_read(inode
)) {
548 error
= vmtruncate(inode
, attr
->ia_size
);
553 setattr_copy(inode
, attr
);
554 mark_inode_dirty(inode
);
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
562 static bool has_pid_permissions(struct pid_namespace
*pid
,
563 struct task_struct
*task
,
566 if (pid
->hide_pid
< hide_pid_min
)
568 if (in_group_p(pid
->pid_gid
))
570 return ptrace_may_access(task
, PTRACE_MODE_READ
);
574 static int proc_pid_permission(struct inode
*inode
, int mask
)
576 struct pid_namespace
*pid
= inode
->i_sb
->s_fs_info
;
577 struct task_struct
*task
;
580 task
= get_proc_task(inode
);
583 has_perms
= has_pid_permissions(pid
, task
, 1);
584 put_task_struct(task
);
587 if (pid
->hide_pid
== 2) {
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
599 return generic_permission(inode
, mask
);
604 static const struct inode_operations proc_def_inode_operations
= {
605 .setattr
= proc_setattr
,
608 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
610 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
611 size_t count
, loff_t
*ppos
)
613 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
616 struct task_struct
*task
= get_proc_task(inode
);
622 if (count
> PROC_BLOCK_SIZE
)
623 count
= PROC_BLOCK_SIZE
;
626 if (!(page
= __get_free_page(GFP_TEMPORARY
)))
629 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
632 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
635 put_task_struct(task
);
640 static const struct file_operations proc_info_file_operations
= {
641 .read
= proc_info_read
,
642 .llseek
= generic_file_llseek
,
645 static int proc_single_show(struct seq_file
*m
, void *v
)
647 struct inode
*inode
= m
->private;
648 struct pid_namespace
*ns
;
650 struct task_struct
*task
;
653 ns
= inode
->i_sb
->s_fs_info
;
654 pid
= proc_pid(inode
);
655 task
= get_pid_task(pid
, PIDTYPE_PID
);
659 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
661 put_task_struct(task
);
665 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
667 return single_open(filp
, proc_single_show
, inode
);
670 static const struct file_operations proc_single_file_operations
= {
671 .open
= proc_single_open
,
674 .release
= single_release
,
677 static int __mem_open(struct inode
*inode
, struct file
*file
, unsigned int mode
)
679 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
680 struct mm_struct
*mm
;
685 mm
= mm_access(task
, mode
);
686 put_task_struct(task
);
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm
->mm_count
);
694 /* but do not pin its memory */
698 /* OK to pass negative loff_t, we can catch out-of-range */
699 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
700 file
->private_data
= mm
;
705 static int mem_open(struct inode
*inode
, struct file
*file
)
707 return __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
710 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
711 size_t count
, loff_t
*ppos
, int write
)
713 struct mm_struct
*mm
= file
->private_data
;
714 unsigned long addr
= *ppos
;
721 page
= (char *)__get_free_page(GFP_TEMPORARY
);
726 if (!atomic_inc_not_zero(&mm
->mm_users
))
730 int this_len
= min_t(int, count
, PAGE_SIZE
);
732 if (write
&& copy_from_user(page
, buf
, this_len
)) {
737 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, write
);
744 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
758 free_page((unsigned long) page
);
762 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
763 size_t count
, loff_t
*ppos
)
765 return mem_rw(file
, buf
, count
, ppos
, 0);
768 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
769 size_t count
, loff_t
*ppos
)
771 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
774 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
778 file
->f_pos
= offset
;
781 file
->f_pos
+= offset
;
786 force_successful_syscall_return();
790 static int mem_release(struct inode
*inode
, struct file
*file
)
792 struct mm_struct
*mm
= file
->private_data
;
798 static const struct file_operations proc_mem_operations
= {
803 .release
= mem_release
,
806 static int environ_open(struct inode
*inode
, struct file
*file
)
808 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
811 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
812 size_t count
, loff_t
*ppos
)
815 unsigned long src
= *ppos
;
817 struct mm_struct
*mm
= file
->private_data
;
822 page
= (char *)__get_free_page(GFP_TEMPORARY
);
827 if (!atomic_inc_not_zero(&mm
->mm_users
))
830 int this_len
, retval
, max_len
;
832 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
837 max_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
838 this_len
= (this_len
> max_len
) ? max_len
: this_len
;
840 retval
= access_remote_vm(mm
, (mm
->env_start
+ src
),
848 if (copy_to_user(buf
, page
, retval
)) {
862 free_page((unsigned long) page
);
866 static const struct file_operations proc_environ_operations
= {
867 .open
= environ_open
,
868 .read
= environ_read
,
869 .llseek
= generic_file_llseek
,
870 .release
= mem_release
,
873 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
874 size_t count
, loff_t
*ppos
)
876 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
877 char buffer
[PROC_NUMBUF
];
879 int oom_adjust
= OOM_DISABLE
;
885 if (lock_task_sighand(task
, &flags
)) {
886 oom_adjust
= task
->signal
->oom_adj
;
887 unlock_task_sighand(task
, &flags
);
890 put_task_struct(task
);
892 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
894 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
897 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
898 size_t count
, loff_t
*ppos
)
900 struct task_struct
*task
;
901 char buffer
[PROC_NUMBUF
];
906 memset(buffer
, 0, sizeof(buffer
));
907 if (count
> sizeof(buffer
) - 1)
908 count
= sizeof(buffer
) - 1;
909 if (copy_from_user(buffer
, buf
, count
)) {
914 err
= kstrtoint(strstrip(buffer
), 0, &oom_adjust
);
917 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
918 oom_adjust
!= OOM_DISABLE
) {
923 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
935 if (!lock_task_sighand(task
, &flags
)) {
940 if (oom_adjust
< task
->signal
->oom_adj
&& !capable(CAP_SYS_RESOURCE
)) {
946 * Warn that /proc/pid/oom_adj is deprecated, see
947 * Documentation/feature-removal-schedule.txt.
949 printk_once(KERN_WARNING
"%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
950 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
952 task
->signal
->oom_adj
= oom_adjust
;
954 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 * value is always attainable.
957 if (task
->signal
->oom_adj
== OOM_ADJUST_MAX
)
958 task
->signal
->oom_score_adj
= OOM_SCORE_ADJ_MAX
;
960 task
->signal
->oom_score_adj
= (oom_adjust
* OOM_SCORE_ADJ_MAX
) /
962 trace_oom_score_adj_update(task
);
964 unlock_task_sighand(task
, &flags
);
967 put_task_struct(task
);
969 return err
< 0 ? err
: count
;
972 static const struct file_operations proc_oom_adjust_operations
= {
973 .read
= oom_adjust_read
,
974 .write
= oom_adjust_write
,
975 .llseek
= generic_file_llseek
,
978 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
979 size_t count
, loff_t
*ppos
)
981 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
982 char buffer
[PROC_NUMBUF
];
983 int oom_score_adj
= OOM_SCORE_ADJ_MIN
;
989 if (lock_task_sighand(task
, &flags
)) {
990 oom_score_adj
= task
->signal
->oom_score_adj
;
991 unlock_task_sighand(task
, &flags
);
993 put_task_struct(task
);
994 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_score_adj
);
995 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
998 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
999 size_t count
, loff_t
*ppos
)
1001 struct task_struct
*task
;
1002 char buffer
[PROC_NUMBUF
];
1003 unsigned long flags
;
1007 memset(buffer
, 0, sizeof(buffer
));
1008 if (count
> sizeof(buffer
) - 1)
1009 count
= sizeof(buffer
) - 1;
1010 if (copy_from_user(buffer
, buf
, count
)) {
1015 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1018 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1019 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1024 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1036 if (!lock_task_sighand(task
, &flags
)) {
1041 if (oom_score_adj
< task
->signal
->oom_score_adj_min
&&
1042 !capable(CAP_SYS_RESOURCE
)) {
1047 task
->signal
->oom_score_adj
= oom_score_adj
;
1048 if (has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1049 task
->signal
->oom_score_adj_min
= oom_score_adj
;
1050 trace_oom_score_adj_update(task
);
1052 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 * always attainable.
1055 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MIN
)
1056 task
->signal
->oom_adj
= OOM_DISABLE
;
1058 task
->signal
->oom_adj
= (oom_score_adj
* OOM_ADJUST_MAX
) /
1061 unlock_task_sighand(task
, &flags
);
1064 put_task_struct(task
);
1066 return err
< 0 ? err
: count
;
1069 static const struct file_operations proc_oom_score_adj_operations
= {
1070 .read
= oom_score_adj_read
,
1071 .write
= oom_score_adj_write
,
1072 .llseek
= default_llseek
,
1075 #ifdef CONFIG_AUDITSYSCALL
1076 #define TMPBUFLEN 21
1077 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1078 size_t count
, loff_t
*ppos
)
1080 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1081 struct task_struct
*task
= get_proc_task(inode
);
1083 char tmpbuf
[TMPBUFLEN
];
1087 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1088 audit_get_loginuid(task
));
1089 put_task_struct(task
);
1090 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1093 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1094 size_t count
, loff_t
*ppos
)
1096 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1102 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1108 if (count
>= PAGE_SIZE
)
1109 count
= PAGE_SIZE
- 1;
1112 /* No partial writes. */
1115 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1119 if (copy_from_user(page
, buf
, count
))
1123 loginuid
= simple_strtoul(page
, &tmp
, 10);
1129 length
= audit_set_loginuid(loginuid
);
1130 if (likely(length
== 0))
1134 free_page((unsigned long) page
);
1138 static const struct file_operations proc_loginuid_operations
= {
1139 .read
= proc_loginuid_read
,
1140 .write
= proc_loginuid_write
,
1141 .llseek
= generic_file_llseek
,
1144 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1145 size_t count
, loff_t
*ppos
)
1147 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1148 struct task_struct
*task
= get_proc_task(inode
);
1150 char tmpbuf
[TMPBUFLEN
];
1154 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1155 audit_get_sessionid(task
));
1156 put_task_struct(task
);
1157 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1160 static const struct file_operations proc_sessionid_operations
= {
1161 .read
= proc_sessionid_read
,
1162 .llseek
= generic_file_llseek
,
1166 #ifdef CONFIG_FAULT_INJECTION
1167 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1168 size_t count
, loff_t
*ppos
)
1170 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1171 char buffer
[PROC_NUMBUF
];
1177 make_it_fail
= task
->make_it_fail
;
1178 put_task_struct(task
);
1180 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1182 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1185 static ssize_t
proc_fault_inject_write(struct file
* file
,
1186 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1188 struct task_struct
*task
;
1189 char buffer
[PROC_NUMBUF
], *end
;
1192 if (!capable(CAP_SYS_RESOURCE
))
1194 memset(buffer
, 0, sizeof(buffer
));
1195 if (count
> sizeof(buffer
) - 1)
1196 count
= sizeof(buffer
) - 1;
1197 if (copy_from_user(buffer
, buf
, count
))
1199 make_it_fail
= simple_strtol(strstrip(buffer
), &end
, 0);
1202 task
= get_proc_task(file
->f_dentry
->d_inode
);
1205 task
->make_it_fail
= make_it_fail
;
1206 put_task_struct(task
);
1211 static const struct file_operations proc_fault_inject_operations
= {
1212 .read
= proc_fault_inject_read
,
1213 .write
= proc_fault_inject_write
,
1214 .llseek
= generic_file_llseek
,
1219 #ifdef CONFIG_SCHED_DEBUG
1221 * Print out various scheduling related per-task fields:
1223 static int sched_show(struct seq_file
*m
, void *v
)
1225 struct inode
*inode
= m
->private;
1226 struct task_struct
*p
;
1228 p
= get_proc_task(inode
);
1231 proc_sched_show_task(p
, m
);
1239 sched_write(struct file
*file
, const char __user
*buf
,
1240 size_t count
, loff_t
*offset
)
1242 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1243 struct task_struct
*p
;
1245 p
= get_proc_task(inode
);
1248 proc_sched_set_task(p
);
1255 static int sched_open(struct inode
*inode
, struct file
*filp
)
1257 return single_open(filp
, sched_show
, inode
);
1260 static const struct file_operations proc_pid_sched_operations
= {
1263 .write
= sched_write
,
1264 .llseek
= seq_lseek
,
1265 .release
= single_release
,
1270 #ifdef CONFIG_SCHED_AUTOGROUP
1272 * Print out autogroup related information:
1274 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1276 struct inode
*inode
= m
->private;
1277 struct task_struct
*p
;
1279 p
= get_proc_task(inode
);
1282 proc_sched_autogroup_show_task(p
, m
);
1290 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1291 size_t count
, loff_t
*offset
)
1293 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1294 struct task_struct
*p
;
1295 char buffer
[PROC_NUMBUF
];
1299 memset(buffer
, 0, sizeof(buffer
));
1300 if (count
> sizeof(buffer
) - 1)
1301 count
= sizeof(buffer
) - 1;
1302 if (copy_from_user(buffer
, buf
, count
))
1305 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1309 p
= get_proc_task(inode
);
1313 err
= proc_sched_autogroup_set_nice(p
, nice
);
1322 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1326 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1328 struct seq_file
*m
= filp
->private_data
;
1335 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1336 .open
= sched_autogroup_open
,
1338 .write
= sched_autogroup_write
,
1339 .llseek
= seq_lseek
,
1340 .release
= single_release
,
1343 #endif /* CONFIG_SCHED_AUTOGROUP */
1345 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1346 size_t count
, loff_t
*offset
)
1348 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1349 struct task_struct
*p
;
1350 char buffer
[TASK_COMM_LEN
];
1352 memset(buffer
, 0, sizeof(buffer
));
1353 if (count
> sizeof(buffer
) - 1)
1354 count
= sizeof(buffer
) - 1;
1355 if (copy_from_user(buffer
, buf
, count
))
1358 p
= get_proc_task(inode
);
1362 if (same_thread_group(current
, p
))
1363 set_task_comm(p
, buffer
);
1372 static int comm_show(struct seq_file
*m
, void *v
)
1374 struct inode
*inode
= m
->private;
1375 struct task_struct
*p
;
1377 p
= get_proc_task(inode
);
1382 seq_printf(m
, "%s\n", p
->comm
);
1390 static int comm_open(struct inode
*inode
, struct file
*filp
)
1392 return single_open(filp
, comm_show
, inode
);
1395 static const struct file_operations proc_pid_set_comm_operations
= {
1398 .write
= comm_write
,
1399 .llseek
= seq_lseek
,
1400 .release
= single_release
,
1403 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1405 struct task_struct
*task
;
1406 struct mm_struct
*mm
;
1407 struct file
*exe_file
;
1409 task
= get_proc_task(dentry
->d_inode
);
1412 mm
= get_task_mm(task
);
1413 put_task_struct(task
);
1416 exe_file
= get_mm_exe_file(mm
);
1419 *exe_path
= exe_file
->f_path
;
1420 path_get(&exe_file
->f_path
);
1427 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1429 struct inode
*inode
= dentry
->d_inode
;
1431 int error
= -EACCES
;
1433 /* Are we allowed to snoop on the tasks file descriptors? */
1434 if (!proc_fd_access_allowed(inode
))
1437 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1441 nd_jump_link(nd
, &path
);
1444 return ERR_PTR(error
);
1447 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1449 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1456 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1457 len
= PTR_ERR(pathname
);
1458 if (IS_ERR(pathname
))
1460 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1464 if (copy_to_user(buffer
, pathname
, len
))
1467 free_page((unsigned long)tmp
);
1471 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1473 int error
= -EACCES
;
1474 struct inode
*inode
= dentry
->d_inode
;
1477 /* Are we allowed to snoop on the tasks file descriptors? */
1478 if (!proc_fd_access_allowed(inode
))
1481 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1485 error
= do_proc_readlink(&path
, buffer
, buflen
);
1491 static const struct inode_operations proc_pid_link_inode_operations
= {
1492 .readlink
= proc_pid_readlink
,
1493 .follow_link
= proc_pid_follow_link
,
1494 .setattr
= proc_setattr
,
1498 /* building an inode */
1500 static int task_dumpable(struct task_struct
*task
)
1503 struct mm_struct
*mm
;
1508 dumpable
= get_dumpable(mm
);
1515 struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1517 struct inode
* inode
;
1518 struct proc_inode
*ei
;
1519 const struct cred
*cred
;
1521 /* We need a new inode */
1523 inode
= new_inode(sb
);
1529 inode
->i_ino
= get_next_ino();
1530 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1531 inode
->i_op
= &proc_def_inode_operations
;
1534 * grab the reference to task.
1536 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1540 if (task_dumpable(task
)) {
1542 cred
= __task_cred(task
);
1543 inode
->i_uid
= cred
->euid
;
1544 inode
->i_gid
= cred
->egid
;
1547 security_task_to_inode(task
, inode
);
1557 int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1559 struct inode
*inode
= dentry
->d_inode
;
1560 struct task_struct
*task
;
1561 const struct cred
*cred
;
1562 struct pid_namespace
*pid
= dentry
->d_sb
->s_fs_info
;
1564 generic_fillattr(inode
, stat
);
1567 stat
->uid
= GLOBAL_ROOT_UID
;
1568 stat
->gid
= GLOBAL_ROOT_GID
;
1569 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1571 if (!has_pid_permissions(pid
, task
, 2)) {
1574 * This doesn't prevent learning whether PID exists,
1575 * it only makes getattr() consistent with readdir().
1579 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1580 task_dumpable(task
)) {
1581 cred
= __task_cred(task
);
1582 stat
->uid
= cred
->euid
;
1583 stat
->gid
= cred
->egid
;
1593 * Exceptional case: normally we are not allowed to unhash a busy
1594 * directory. In this case, however, we can do it - no aliasing problems
1595 * due to the way we treat inodes.
1597 * Rewrite the inode's ownerships here because the owning task may have
1598 * performed a setuid(), etc.
1600 * Before the /proc/pid/status file was created the only way to read
1601 * the effective uid of a /process was to stat /proc/pid. Reading
1602 * /proc/pid/status is slow enough that procps and other packages
1603 * kept stating /proc/pid. To keep the rules in /proc simple I have
1604 * made this apply to all per process world readable and executable
1607 int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1609 struct inode
*inode
;
1610 struct task_struct
*task
;
1611 const struct cred
*cred
;
1613 if (flags
& LOOKUP_RCU
)
1616 inode
= dentry
->d_inode
;
1617 task
= get_proc_task(inode
);
1620 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1621 task_dumpable(task
)) {
1623 cred
= __task_cred(task
);
1624 inode
->i_uid
= cred
->euid
;
1625 inode
->i_gid
= cred
->egid
;
1628 inode
->i_uid
= GLOBAL_ROOT_UID
;
1629 inode
->i_gid
= GLOBAL_ROOT_GID
;
1631 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1632 security_task_to_inode(task
, inode
);
1633 put_task_struct(task
);
1640 static int pid_delete_dentry(const struct dentry
* dentry
)
1642 /* Is the task we represent dead?
1643 * If so, then don't put the dentry on the lru list,
1644 * kill it immediately.
1646 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1649 const struct dentry_operations pid_dentry_operations
=
1651 .d_revalidate
= pid_revalidate
,
1652 .d_delete
= pid_delete_dentry
,
1658 * Fill a directory entry.
1660 * If possible create the dcache entry and derive our inode number and
1661 * file type from dcache entry.
1663 * Since all of the proc inode numbers are dynamically generated, the inode
1664 * numbers do not exist until the inode is cache. This means creating the
1665 * the dcache entry in readdir is necessary to keep the inode numbers
1666 * reported by readdir in sync with the inode numbers reported
1669 int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1670 const char *name
, int len
,
1671 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1673 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1674 struct inode
*inode
;
1677 unsigned type
= DT_UNKNOWN
;
1681 qname
.hash
= full_name_hash(name
, len
);
1683 child
= d_lookup(dir
, &qname
);
1686 new = d_alloc(dir
, &qname
);
1688 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1695 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1696 goto end_instantiate
;
1697 inode
= child
->d_inode
;
1700 type
= inode
->i_mode
>> 12;
1705 ino
= find_inode_number(dir
, &qname
);
1708 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1711 static unsigned name_to_int(struct dentry
*dentry
)
1713 const char *name
= dentry
->d_name
.name
;
1714 int len
= dentry
->d_name
.len
;
1717 if (len
> 1 && *name
== '0')
1720 unsigned c
= *name
++ - '0';
1723 if (n
>= (~0U-9)/10)
1733 #define PROC_FDINFO_MAX 64
1735 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1737 struct task_struct
*task
= get_proc_task(inode
);
1738 struct files_struct
*files
= NULL
;
1740 int fd
= proc_fd(inode
);
1743 files
= get_files_struct(task
);
1744 put_task_struct(task
);
1748 * We are not taking a ref to the file structure, so we must
1751 spin_lock(&files
->file_lock
);
1752 file
= fcheck_files(files
, fd
);
1754 unsigned int f_flags
;
1755 struct fdtable
*fdt
;
1757 fdt
= files_fdtable(files
);
1758 f_flags
= file
->f_flags
& ~O_CLOEXEC
;
1759 if (close_on_exec(fd
, fdt
))
1760 f_flags
|= O_CLOEXEC
;
1763 *path
= file
->f_path
;
1764 path_get(&file
->f_path
);
1767 snprintf(info
, PROC_FDINFO_MAX
,
1770 (long long) file
->f_pos
,
1772 spin_unlock(&files
->file_lock
);
1773 put_files_struct(files
);
1776 spin_unlock(&files
->file_lock
);
1777 put_files_struct(files
);
1782 static int proc_fd_link(struct dentry
*dentry
, struct path
*path
)
1784 return proc_fd_info(dentry
->d_inode
, path
, NULL
);
1787 static int tid_fd_revalidate(struct dentry
*dentry
, unsigned int flags
)
1789 struct inode
*inode
;
1790 struct task_struct
*task
;
1792 struct files_struct
*files
;
1793 const struct cred
*cred
;
1795 if (flags
& LOOKUP_RCU
)
1798 inode
= dentry
->d_inode
;
1799 task
= get_proc_task(inode
);
1800 fd
= proc_fd(inode
);
1803 files
= get_files_struct(task
);
1807 file
= fcheck_files(files
, fd
);
1809 unsigned f_mode
= file
->f_mode
;
1812 put_files_struct(files
);
1814 if (task_dumpable(task
)) {
1816 cred
= __task_cred(task
);
1817 inode
->i_uid
= cred
->euid
;
1818 inode
->i_gid
= cred
->egid
;
1821 inode
->i_uid
= GLOBAL_ROOT_UID
;
1822 inode
->i_gid
= GLOBAL_ROOT_GID
;
1825 if (S_ISLNK(inode
->i_mode
)) {
1826 unsigned i_mode
= S_IFLNK
;
1827 if (f_mode
& FMODE_READ
)
1828 i_mode
|= S_IRUSR
| S_IXUSR
;
1829 if (f_mode
& FMODE_WRITE
)
1830 i_mode
|= S_IWUSR
| S_IXUSR
;
1831 inode
->i_mode
= i_mode
;
1834 security_task_to_inode(task
, inode
);
1835 put_task_struct(task
);
1839 put_files_struct(files
);
1841 put_task_struct(task
);
1847 static const struct dentry_operations tid_fd_dentry_operations
=
1849 .d_revalidate
= tid_fd_revalidate
,
1850 .d_delete
= pid_delete_dentry
,
1853 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1854 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1856 unsigned fd
= (unsigned long)ptr
;
1857 struct inode
*inode
;
1858 struct proc_inode
*ei
;
1859 struct dentry
*error
= ERR_PTR(-ENOENT
);
1861 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1867 inode
->i_mode
= S_IFLNK
;
1868 inode
->i_op
= &proc_pid_link_inode_operations
;
1870 ei
->op
.proc_get_link
= proc_fd_link
;
1871 d_set_d_op(dentry
, &tid_fd_dentry_operations
);
1872 d_add(dentry
, inode
);
1873 /* Close the race of the process dying before we return the dentry */
1874 if (tid_fd_revalidate(dentry
, 0))
1881 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1882 struct dentry
*dentry
,
1883 instantiate_t instantiate
)
1885 struct task_struct
*task
= get_proc_task(dir
);
1886 unsigned fd
= name_to_int(dentry
);
1887 struct dentry
*result
= ERR_PTR(-ENOENT
);
1894 result
= instantiate(dir
, dentry
, task
, (void *)(unsigned long)fd
);
1896 put_task_struct(task
);
1901 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1902 filldir_t filldir
, instantiate_t instantiate
)
1904 struct dentry
*dentry
= filp
->f_path
.dentry
;
1905 struct inode
*inode
= dentry
->d_inode
;
1906 struct task_struct
*p
= get_proc_task(inode
);
1907 unsigned int fd
, ino
;
1909 struct files_struct
* files
;
1919 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1923 ino
= parent_ino(dentry
);
1924 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1928 files
= get_files_struct(p
);
1932 for (fd
= filp
->f_pos
-2;
1933 fd
< files_fdtable(files
)->max_fds
;
1934 fd
++, filp
->f_pos
++) {
1935 char name
[PROC_NUMBUF
];
1939 if (!fcheck_files(files
, fd
))
1943 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1944 rv
= proc_fill_cache(filp
, dirent
, filldir
,
1945 name
, len
, instantiate
, p
,
1946 (void *)(unsigned long)fd
);
1953 put_files_struct(files
);
1961 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1964 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1967 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1969 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1972 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1973 size_t len
, loff_t
*ppos
)
1975 char tmp
[PROC_FDINFO_MAX
];
1976 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1978 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1982 static const struct file_operations proc_fdinfo_file_operations
= {
1983 .open
= nonseekable_open
,
1984 .read
= proc_fdinfo_read
,
1985 .llseek
= no_llseek
,
1988 static const struct file_operations proc_fd_operations
= {
1989 .read
= generic_read_dir
,
1990 .readdir
= proc_readfd
,
1991 .llseek
= default_llseek
,
1994 #ifdef CONFIG_CHECKPOINT_RESTORE
1997 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1998 * which represent vma start and end addresses.
2000 static int dname_to_vma_addr(struct dentry
*dentry
,
2001 unsigned long *start
, unsigned long *end
)
2003 if (sscanf(dentry
->d_name
.name
, "%lx-%lx", start
, end
) != 2)
2009 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
2011 unsigned long vm_start
, vm_end
;
2012 bool exact_vma_exists
= false;
2013 struct mm_struct
*mm
= NULL
;
2014 struct task_struct
*task
;
2015 const struct cred
*cred
;
2016 struct inode
*inode
;
2019 if (flags
& LOOKUP_RCU
)
2022 if (!capable(CAP_SYS_ADMIN
)) {
2027 inode
= dentry
->d_inode
;
2028 task
= get_proc_task(inode
);
2032 mm
= mm_access(task
, PTRACE_MODE_READ
);
2033 if (IS_ERR_OR_NULL(mm
))
2036 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
2037 down_read(&mm
->mmap_sem
);
2038 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
2039 up_read(&mm
->mmap_sem
);
2044 if (exact_vma_exists
) {
2045 if (task_dumpable(task
)) {
2047 cred
= __task_cred(task
);
2048 inode
->i_uid
= cred
->euid
;
2049 inode
->i_gid
= cred
->egid
;
2052 inode
->i_uid
= GLOBAL_ROOT_UID
;
2053 inode
->i_gid
= GLOBAL_ROOT_GID
;
2055 security_task_to_inode(task
, inode
);
2060 put_task_struct(task
);
2069 static const struct dentry_operations tid_map_files_dentry_operations
= {
2070 .d_revalidate
= map_files_d_revalidate
,
2071 .d_delete
= pid_delete_dentry
,
2074 static int proc_map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2076 unsigned long vm_start
, vm_end
;
2077 struct vm_area_struct
*vma
;
2078 struct task_struct
*task
;
2079 struct mm_struct
*mm
;
2083 task
= get_proc_task(dentry
->d_inode
);
2087 mm
= get_task_mm(task
);
2088 put_task_struct(task
);
2092 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2096 down_read(&mm
->mmap_sem
);
2097 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2098 if (vma
&& vma
->vm_file
) {
2099 *path
= vma
->vm_file
->f_path
;
2103 up_read(&mm
->mmap_sem
);
2111 struct map_files_info
{
2114 unsigned char name
[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2117 static struct dentry
*
2118 proc_map_files_instantiate(struct inode
*dir
, struct dentry
*dentry
,
2119 struct task_struct
*task
, const void *ptr
)
2121 const struct file
*file
= ptr
;
2122 struct proc_inode
*ei
;
2123 struct inode
*inode
;
2126 return ERR_PTR(-ENOENT
);
2128 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2130 return ERR_PTR(-ENOENT
);
2133 ei
->op
.proc_get_link
= proc_map_files_get_link
;
2135 inode
->i_op
= &proc_pid_link_inode_operations
;
2137 inode
->i_mode
= S_IFLNK
;
2139 if (file
->f_mode
& FMODE_READ
)
2140 inode
->i_mode
|= S_IRUSR
;
2141 if (file
->f_mode
& FMODE_WRITE
)
2142 inode
->i_mode
|= S_IWUSR
;
2144 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2145 d_add(dentry
, inode
);
2150 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2151 struct dentry
*dentry
, unsigned int flags
)
2153 unsigned long vm_start
, vm_end
;
2154 struct vm_area_struct
*vma
;
2155 struct task_struct
*task
;
2156 struct dentry
*result
;
2157 struct mm_struct
*mm
;
2159 result
= ERR_PTR(-EACCES
);
2160 if (!capable(CAP_SYS_ADMIN
))
2163 result
= ERR_PTR(-ENOENT
);
2164 task
= get_proc_task(dir
);
2168 result
= ERR_PTR(-EACCES
);
2169 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
2172 result
= ERR_PTR(-ENOENT
);
2173 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2176 mm
= get_task_mm(task
);
2180 down_read(&mm
->mmap_sem
);
2181 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2185 result
= proc_map_files_instantiate(dir
, dentry
, task
, vma
->vm_file
);
2188 up_read(&mm
->mmap_sem
);
2191 put_task_struct(task
);
2196 static const struct inode_operations proc_map_files_inode_operations
= {
2197 .lookup
= proc_map_files_lookup
,
2198 .permission
= proc_fd_permission
,
2199 .setattr
= proc_setattr
,
2203 proc_map_files_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
2205 struct dentry
*dentry
= filp
->f_path
.dentry
;
2206 struct inode
*inode
= dentry
->d_inode
;
2207 struct vm_area_struct
*vma
;
2208 struct task_struct
*task
;
2209 struct mm_struct
*mm
;
2214 if (!capable(CAP_SYS_ADMIN
))
2218 task
= get_proc_task(inode
);
2223 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
2227 switch (filp
->f_pos
) {
2230 if (filldir(dirent
, ".", 1, 0, ino
, DT_DIR
) < 0)
2234 ino
= parent_ino(dentry
);
2235 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
2240 unsigned long nr_files
, pos
, i
;
2241 struct flex_array
*fa
= NULL
;
2242 struct map_files_info info
;
2243 struct map_files_info
*p
;
2245 mm
= get_task_mm(task
);
2248 down_read(&mm
->mmap_sem
);
2253 * We need two passes here:
2255 * 1) Collect vmas of mapped files with mmap_sem taken
2256 * 2) Release mmap_sem and instantiate entries
2258 * otherwise we get lockdep complained, since filldir()
2259 * routine might require mmap_sem taken in might_fault().
2262 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2263 if (vma
->vm_file
&& ++pos
> filp
->f_pos
)
2268 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2270 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2274 flex_array_free(fa
);
2275 up_read(&mm
->mmap_sem
);
2279 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2280 vma
= vma
->vm_next
) {
2283 if (++pos
<= filp
->f_pos
)
2286 get_file(vma
->vm_file
);
2287 info
.file
= vma
->vm_file
;
2288 info
.len
= snprintf(info
.name
,
2289 sizeof(info
.name
), "%lx-%lx",
2290 vma
->vm_start
, vma
->vm_end
);
2291 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2295 up_read(&mm
->mmap_sem
);
2297 for (i
= 0; i
< nr_files
; i
++) {
2298 p
= flex_array_get(fa
, i
);
2299 ret
= proc_fill_cache(filp
, dirent
, filldir
,
2301 proc_map_files_instantiate
,
2308 for (; i
< nr_files
; i
++) {
2310 * In case of error don't forget
2311 * to put rest of file refs.
2313 p
= flex_array_get(fa
, i
);
2317 flex_array_free(fa
);
2323 put_task_struct(task
);
2328 static const struct file_operations proc_map_files_operations
= {
2329 .read
= generic_read_dir
,
2330 .readdir
= proc_map_files_readdir
,
2331 .llseek
= default_llseek
,
2334 #endif /* CONFIG_CHECKPOINT_RESTORE */
2337 * /proc/pid/fd needs a special permission handler so that a process can still
2338 * access /proc/self/fd after it has executed a setuid().
2340 static int proc_fd_permission(struct inode
*inode
, int mask
)
2342 int rv
= generic_permission(inode
, mask
);
2345 if (task_pid(current
) == proc_pid(inode
))
2351 * proc directories can do almost nothing..
2353 static const struct inode_operations proc_fd_inode_operations
= {
2354 .lookup
= proc_lookupfd
,
2355 .permission
= proc_fd_permission
,
2356 .setattr
= proc_setattr
,
2359 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
2360 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2362 unsigned fd
= (unsigned long)ptr
;
2363 struct inode
*inode
;
2364 struct proc_inode
*ei
;
2365 struct dentry
*error
= ERR_PTR(-ENOENT
);
2367 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2372 inode
->i_mode
= S_IFREG
| S_IRUSR
;
2373 inode
->i_fop
= &proc_fdinfo_file_operations
;
2374 d_set_d_op(dentry
, &tid_fd_dentry_operations
);
2375 d_add(dentry
, inode
);
2376 /* Close the race of the process dying before we return the dentry */
2377 if (tid_fd_revalidate(dentry
, 0))
2384 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
2385 struct dentry
*dentry
,
2388 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
2391 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
2393 return proc_readfd_common(filp
, dirent
, filldir
,
2394 proc_fdinfo_instantiate
);
2397 static const struct file_operations proc_fdinfo_operations
= {
2398 .read
= generic_read_dir
,
2399 .readdir
= proc_readfdinfo
,
2400 .llseek
= default_llseek
,
2404 * proc directories can do almost nothing..
2406 static const struct inode_operations proc_fdinfo_inode_operations
= {
2407 .lookup
= proc_lookupfdinfo
,
2408 .setattr
= proc_setattr
,
2412 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
2413 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2415 const struct pid_entry
*p
= ptr
;
2416 struct inode
*inode
;
2417 struct proc_inode
*ei
;
2418 struct dentry
*error
= ERR_PTR(-ENOENT
);
2420 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2425 inode
->i_mode
= p
->mode
;
2426 if (S_ISDIR(inode
->i_mode
))
2427 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2429 inode
->i_op
= p
->iop
;
2431 inode
->i_fop
= p
->fop
;
2433 d_set_d_op(dentry
, &pid_dentry_operations
);
2434 d_add(dentry
, inode
);
2435 /* Close the race of the process dying before we return the dentry */
2436 if (pid_revalidate(dentry
, 0))
2442 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2443 struct dentry
*dentry
,
2444 const struct pid_entry
*ents
,
2447 struct dentry
*error
;
2448 struct task_struct
*task
= get_proc_task(dir
);
2449 const struct pid_entry
*p
, *last
;
2451 error
= ERR_PTR(-ENOENT
);
2457 * Yes, it does not scale. And it should not. Don't add
2458 * new entries into /proc/<tgid>/ without very good reasons.
2460 last
= &ents
[nents
- 1];
2461 for (p
= ents
; p
<= last
; p
++) {
2462 if (p
->len
!= dentry
->d_name
.len
)
2464 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2470 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2472 put_task_struct(task
);
2477 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2478 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2480 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2481 proc_pident_instantiate
, task
, p
);
2484 static int proc_pident_readdir(struct file
*filp
,
2485 void *dirent
, filldir_t filldir
,
2486 const struct pid_entry
*ents
, unsigned int nents
)
2489 struct dentry
*dentry
= filp
->f_path
.dentry
;
2490 struct inode
*inode
= dentry
->d_inode
;
2491 struct task_struct
*task
= get_proc_task(inode
);
2492 const struct pid_entry
*p
, *last
;
2505 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2511 ino
= parent_ino(dentry
);
2512 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2524 last
= &ents
[nents
- 1];
2526 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2535 put_task_struct(task
);
2540 #ifdef CONFIG_SECURITY
2541 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2542 size_t count
, loff_t
*ppos
)
2544 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2547 struct task_struct
*task
= get_proc_task(inode
);
2552 length
= security_getprocattr(task
,
2553 (char*)file
->f_path
.dentry
->d_name
.name
,
2555 put_task_struct(task
);
2557 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2562 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2563 size_t count
, loff_t
*ppos
)
2565 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2568 struct task_struct
*task
= get_proc_task(inode
);
2573 if (count
> PAGE_SIZE
)
2576 /* No partial writes. */
2582 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2587 if (copy_from_user(page
, buf
, count
))
2590 /* Guard against adverse ptrace interaction */
2591 length
= mutex_lock_interruptible(&task
->signal
->cred_guard_mutex
);
2595 length
= security_setprocattr(task
,
2596 (char*)file
->f_path
.dentry
->d_name
.name
,
2597 (void*)page
, count
);
2598 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2600 free_page((unsigned long) page
);
2602 put_task_struct(task
);
2607 static const struct file_operations proc_pid_attr_operations
= {
2608 .read
= proc_pid_attr_read
,
2609 .write
= proc_pid_attr_write
,
2610 .llseek
= generic_file_llseek
,
2613 static const struct pid_entry attr_dir_stuff
[] = {
2614 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2615 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2616 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2617 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2618 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2619 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2622 static int proc_attr_dir_readdir(struct file
* filp
,
2623 void * dirent
, filldir_t filldir
)
2625 return proc_pident_readdir(filp
,dirent
,filldir
,
2626 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2629 static const struct file_operations proc_attr_dir_operations
= {
2630 .read
= generic_read_dir
,
2631 .readdir
= proc_attr_dir_readdir
,
2632 .llseek
= default_llseek
,
2635 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2636 struct dentry
*dentry
, unsigned int flags
)
2638 return proc_pident_lookup(dir
, dentry
,
2639 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2642 static const struct inode_operations proc_attr_dir_inode_operations
= {
2643 .lookup
= proc_attr_dir_lookup
,
2644 .getattr
= pid_getattr
,
2645 .setattr
= proc_setattr
,
2650 #ifdef CONFIG_ELF_CORE
2651 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2652 size_t count
, loff_t
*ppos
)
2654 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2655 struct mm_struct
*mm
;
2656 char buffer
[PROC_NUMBUF
];
2664 mm
= get_task_mm(task
);
2666 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2667 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2668 MMF_DUMP_FILTER_SHIFT
));
2670 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2673 put_task_struct(task
);
2678 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2679 const char __user
*buf
,
2683 struct task_struct
*task
;
2684 struct mm_struct
*mm
;
2685 char buffer
[PROC_NUMBUF
], *end
;
2692 memset(buffer
, 0, sizeof(buffer
));
2693 if (count
> sizeof(buffer
) - 1)
2694 count
= sizeof(buffer
) - 1;
2695 if (copy_from_user(buffer
, buf
, count
))
2699 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2702 if (end
- buffer
== 0)
2706 task
= get_proc_task(file
->f_dentry
->d_inode
);
2711 mm
= get_task_mm(task
);
2715 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2717 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2719 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2724 put_task_struct(task
);
2729 static const struct file_operations proc_coredump_filter_operations
= {
2730 .read
= proc_coredump_filter_read
,
2731 .write
= proc_coredump_filter_write
,
2732 .llseek
= generic_file_llseek
,
2739 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2742 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2743 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2744 char tmp
[PROC_NUMBUF
];
2747 sprintf(tmp
, "%d", tgid
);
2748 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2751 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2753 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2754 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2755 char *name
= ERR_PTR(-ENOENT
);
2759 name
= ERR_PTR(-ENOMEM
);
2761 sprintf(name
, "%d", tgid
);
2763 nd_set_link(nd
, name
);
2767 static void proc_self_put_link(struct dentry
*dentry
, struct nameidata
*nd
,
2770 char *s
= nd_get_link(nd
);
2775 static const struct inode_operations proc_self_inode_operations
= {
2776 .readlink
= proc_self_readlink
,
2777 .follow_link
= proc_self_follow_link
,
2778 .put_link
= proc_self_put_link
,
2784 * These are the directory entries in the root directory of /proc
2785 * that properly belong to the /proc filesystem, as they describe
2786 * describe something that is process related.
2788 static const struct pid_entry proc_base_stuff
[] = {
2789 NOD("self", S_IFLNK
|S_IRWXUGO
,
2790 &proc_self_inode_operations
, NULL
, {}),
2793 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2794 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2796 const struct pid_entry
*p
= ptr
;
2797 struct inode
*inode
;
2798 struct proc_inode
*ei
;
2799 struct dentry
*error
;
2801 /* Allocate the inode */
2802 error
= ERR_PTR(-ENOMEM
);
2803 inode
= new_inode(dir
->i_sb
);
2807 /* Initialize the inode */
2809 inode
->i_ino
= get_next_ino();
2810 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2813 * grab the reference to the task.
2815 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2819 inode
->i_mode
= p
->mode
;
2820 if (S_ISDIR(inode
->i_mode
))
2821 set_nlink(inode
, 2);
2822 if (S_ISLNK(inode
->i_mode
))
2825 inode
->i_op
= p
->iop
;
2827 inode
->i_fop
= p
->fop
;
2829 d_add(dentry
, inode
);
2838 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2840 struct dentry
*error
;
2841 struct task_struct
*task
= get_proc_task(dir
);
2842 const struct pid_entry
*p
, *last
;
2844 error
= ERR_PTR(-ENOENT
);
2849 /* Lookup the directory entry */
2850 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2851 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2852 if (p
->len
!= dentry
->d_name
.len
)
2854 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2860 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2863 put_task_struct(task
);
2868 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2869 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2871 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2872 proc_base_instantiate
, task
, p
);
2875 #ifdef CONFIG_TASK_IO_ACCOUNTING
2876 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2878 struct task_io_accounting acct
= task
->ioac
;
2879 unsigned long flags
;
2882 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2886 if (!ptrace_may_access(task
, PTRACE_MODE_READ
)) {
2891 if (whole
&& lock_task_sighand(task
, &flags
)) {
2892 struct task_struct
*t
= task
;
2894 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2895 while_each_thread(task
, t
)
2896 task_io_accounting_add(&acct
, &t
->ioac
);
2898 unlock_task_sighand(task
, &flags
);
2900 result
= sprintf(buffer
,
2905 "read_bytes: %llu\n"
2906 "write_bytes: %llu\n"
2907 "cancelled_write_bytes: %llu\n",
2908 (unsigned long long)acct
.rchar
,
2909 (unsigned long long)acct
.wchar
,
2910 (unsigned long long)acct
.syscr
,
2911 (unsigned long long)acct
.syscw
,
2912 (unsigned long long)acct
.read_bytes
,
2913 (unsigned long long)acct
.write_bytes
,
2914 (unsigned long long)acct
.cancelled_write_bytes
);
2916 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2920 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2922 return do_io_accounting(task
, buffer
, 0);
2925 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2927 return do_io_accounting(task
, buffer
, 1);
2929 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2931 #ifdef CONFIG_USER_NS
2932 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2933 struct seq_operations
*seq_ops
)
2935 struct user_namespace
*ns
= NULL
;
2936 struct task_struct
*task
;
2937 struct seq_file
*seq
;
2940 task
= get_proc_task(inode
);
2943 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2945 put_task_struct(task
);
2950 ret
= seq_open(file
, seq_ops
);
2954 seq
= file
->private_data
;
2964 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2966 struct seq_file
*seq
= file
->private_data
;
2967 struct user_namespace
*ns
= seq
->private;
2969 return seq_release(inode
, file
);
2972 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2974 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2977 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2979 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2982 static const struct file_operations proc_uid_map_operations
= {
2983 .open
= proc_uid_map_open
,
2984 .write
= proc_uid_map_write
,
2986 .llseek
= seq_lseek
,
2987 .release
= proc_id_map_release
,
2990 static const struct file_operations proc_gid_map_operations
= {
2991 .open
= proc_gid_map_open
,
2992 .write
= proc_gid_map_write
,
2994 .llseek
= seq_lseek
,
2995 .release
= proc_id_map_release
,
2997 #endif /* CONFIG_USER_NS */
2999 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
3000 struct pid
*pid
, struct task_struct
*task
)
3002 int err
= lock_trace(task
);
3004 seq_printf(m
, "%08x\n", task
->personality
);
3013 static const struct file_operations proc_task_operations
;
3014 static const struct inode_operations proc_task_inode_operations
;
3016 static const struct pid_entry tgid_base_stuff
[] = {
3017 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
3018 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3019 #ifdef CONFIG_CHECKPOINT_RESTORE
3020 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
3022 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3023 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3025 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3027 REG("environ", S_IRUSR
, proc_environ_operations
),
3028 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3029 ONE("status", S_IRUGO
, proc_pid_status
),
3030 ONE("personality", S_IRUGO
, proc_pid_personality
),
3031 INF("limits", S_IRUGO
, proc_pid_limits
),
3032 #ifdef CONFIG_SCHED_DEBUG
3033 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3035 #ifdef CONFIG_SCHED_AUTOGROUP
3036 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
3038 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3039 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3040 INF("syscall", S_IRUGO
, proc_pid_syscall
),
3042 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3043 ONE("stat", S_IRUGO
, proc_tgid_stat
),
3044 ONE("statm", S_IRUGO
, proc_pid_statm
),
3045 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3047 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3049 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3050 LNK("cwd", proc_cwd_link
),
3051 LNK("root", proc_root_link
),
3052 LNK("exe", proc_exe_link
),
3053 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3054 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3055 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
3056 #ifdef CONFIG_PROC_PAGE_MONITOR
3057 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3058 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3059 REG("pagemap", S_IRUGO
, proc_pagemap_operations
),
3061 #ifdef CONFIG_SECURITY
3062 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3064 #ifdef CONFIG_KALLSYMS
3065 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3067 #ifdef CONFIG_STACKTRACE
3068 ONE("stack", S_IRUGO
, proc_pid_stack
),
3070 #ifdef CONFIG_SCHEDSTATS
3071 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3073 #ifdef CONFIG_LATENCYTOP
3074 REG("latency", S_IRUGO
, proc_lstats_operations
),
3076 #ifdef CONFIG_PROC_PID_CPUSET
3077 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3079 #ifdef CONFIG_CGROUPS
3080 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3082 INF("oom_score", S_IRUGO
, proc_oom_score
),
3083 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3084 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3085 #ifdef CONFIG_AUDITSYSCALL
3086 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3087 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3089 #ifdef CONFIG_FAULT_INJECTION
3090 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3092 #ifdef CONFIG_ELF_CORE
3093 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3095 #ifdef CONFIG_TASK_IO_ACCOUNTING
3096 INF("io", S_IRUSR
, proc_tgid_io_accounting
),
3098 #ifdef CONFIG_HARDWALL
3099 INF("hardwall", S_IRUGO
, proc_pid_hardwall
),
3101 #ifdef CONFIG_USER_NS
3102 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3103 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3107 static int proc_tgid_base_readdir(struct file
* filp
,
3108 void * dirent
, filldir_t filldir
)
3110 return proc_pident_readdir(filp
,dirent
,filldir
,
3111 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
3114 static const struct file_operations proc_tgid_base_operations
= {
3115 .read
= generic_read_dir
,
3116 .readdir
= proc_tgid_base_readdir
,
3117 .llseek
= default_llseek
,
3120 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3122 return proc_pident_lookup(dir
, dentry
,
3123 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3126 static const struct inode_operations proc_tgid_base_inode_operations
= {
3127 .lookup
= proc_tgid_base_lookup
,
3128 .getattr
= pid_getattr
,
3129 .setattr
= proc_setattr
,
3130 .permission
= proc_pid_permission
,
3133 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
3135 struct dentry
*dentry
, *leader
, *dir
;
3136 char buf
[PROC_NUMBUF
];
3140 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3141 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3143 shrink_dcache_parent(dentry
);
3149 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
3150 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3155 name
.len
= strlen(name
.name
);
3156 dir
= d_hash_and_lookup(leader
, &name
);
3158 goto out_put_leader
;
3161 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3162 dentry
= d_hash_and_lookup(dir
, &name
);
3164 shrink_dcache_parent(dentry
);
3177 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3178 * @task: task that should be flushed.
3180 * When flushing dentries from proc, one needs to flush them from global
3181 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3182 * in. This call is supposed to do all of this job.
3184 * Looks in the dcache for
3186 * /proc/@tgid/task/@pid
3187 * if either directory is present flushes it and all of it'ts children
3190 * It is safe and reasonable to cache /proc entries for a task until
3191 * that task exits. After that they just clog up the dcache with
3192 * useless entries, possibly causing useful dcache entries to be
3193 * flushed instead. This routine is proved to flush those useless
3194 * dcache entries at process exit time.
3196 * NOTE: This routine is just an optimization so it does not guarantee
3197 * that no dcache entries will exist at process exit time it
3198 * just makes it very unlikely that any will persist.
3201 void proc_flush_task(struct task_struct
*task
)
3204 struct pid
*pid
, *tgid
;
3207 pid
= task_pid(task
);
3208 tgid
= task_tgid(task
);
3210 for (i
= 0; i
<= pid
->level
; i
++) {
3211 upid
= &pid
->numbers
[i
];
3212 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
3213 tgid
->numbers
[i
].nr
);
3216 upid
= &pid
->numbers
[pid
->level
];
3218 pid_ns_release_proc(upid
->ns
);
3221 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
3222 struct dentry
* dentry
,
3223 struct task_struct
*task
, const void *ptr
)
3225 struct dentry
*error
= ERR_PTR(-ENOENT
);
3226 struct inode
*inode
;
3228 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3232 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3233 inode
->i_op
= &proc_tgid_base_inode_operations
;
3234 inode
->i_fop
= &proc_tgid_base_operations
;
3235 inode
->i_flags
|=S_IMMUTABLE
;
3237 set_nlink(inode
, 2 + pid_entry_count_dirs(tgid_base_stuff
,
3238 ARRAY_SIZE(tgid_base_stuff
)));
3240 d_set_d_op(dentry
, &pid_dentry_operations
);
3242 d_add(dentry
, inode
);
3243 /* Close the race of the process dying before we return the dentry */
3244 if (pid_revalidate(dentry
, 0))
3250 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3252 struct dentry
*result
;
3253 struct task_struct
*task
;
3255 struct pid_namespace
*ns
;
3257 result
= proc_base_lookup(dir
, dentry
);
3258 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
3261 tgid
= name_to_int(dentry
);
3265 ns
= dentry
->d_sb
->s_fs_info
;
3267 task
= find_task_by_pid_ns(tgid
, ns
);
3269 get_task_struct(task
);
3274 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
3275 put_task_struct(task
);
3281 * Find the first task with tgid >= tgid
3286 struct task_struct
*task
;
3288 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3293 put_task_struct(iter
.task
);
3297 pid
= find_ge_pid(iter
.tgid
, ns
);
3299 iter
.tgid
= pid_nr_ns(pid
, ns
);
3300 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3301 /* What we to know is if the pid we have find is the
3302 * pid of a thread_group_leader. Testing for task
3303 * being a thread_group_leader is the obvious thing
3304 * todo but there is a window when it fails, due to
3305 * the pid transfer logic in de_thread.
3307 * So we perform the straight forward test of seeing
3308 * if the pid we have found is the pid of a thread
3309 * group leader, and don't worry if the task we have
3310 * found doesn't happen to be a thread group leader.
3311 * As we don't care in the case of readdir.
3313 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3317 get_task_struct(iter
.task
);
3323 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3325 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3326 struct tgid_iter iter
)
3328 char name
[PROC_NUMBUF
];
3329 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
3330 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3331 proc_pid_instantiate
, iter
.task
, NULL
);
3334 static int fake_filldir(void *buf
, const char *name
, int namelen
,
3335 loff_t offset
, u64 ino
, unsigned d_type
)
3340 /* for the /proc/ directory itself, after non-process stuff has been done */
3341 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3344 struct task_struct
*reaper
;
3345 struct tgid_iter iter
;
3346 struct pid_namespace
*ns
;
3347 filldir_t __filldir
;
3349 if (filp
->f_pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3351 nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
3353 reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
3357 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
3358 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
3359 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
3363 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3365 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
3366 for (iter
= next_tgid(ns
, iter
);
3368 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3369 if (has_pid_permissions(ns
, iter
.task
, 2))
3370 __filldir
= filldir
;
3372 __filldir
= fake_filldir
;
3374 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
3375 if (proc_pid_fill_cache(filp
, dirent
, __filldir
, iter
) < 0) {
3376 put_task_struct(iter
.task
);
3380 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3382 put_task_struct(reaper
);
3390 static const struct pid_entry tid_base_stuff
[] = {
3391 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3392 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3393 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3394 REG("environ", S_IRUSR
, proc_environ_operations
),
3395 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3396 ONE("status", S_IRUGO
, proc_pid_status
),
3397 ONE("personality", S_IRUGO
, proc_pid_personality
),
3398 INF("limits", S_IRUGO
, proc_pid_limits
),
3399 #ifdef CONFIG_SCHED_DEBUG
3400 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3402 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3403 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3404 INF("syscall", S_IRUGO
, proc_pid_syscall
),
3406 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3407 ONE("stat", S_IRUGO
, proc_tid_stat
),
3408 ONE("statm", S_IRUGO
, proc_pid_statm
),
3409 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3410 #ifdef CONFIG_CHECKPOINT_RESTORE
3411 REG("children", S_IRUGO
, proc_tid_children_operations
),
3414 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3416 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3417 LNK("cwd", proc_cwd_link
),
3418 LNK("root", proc_root_link
),
3419 LNK("exe", proc_exe_link
),
3420 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3421 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3422 #ifdef CONFIG_PROC_PAGE_MONITOR
3423 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3424 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3425 REG("pagemap", S_IRUGO
, proc_pagemap_operations
),
3427 #ifdef CONFIG_SECURITY
3428 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3430 #ifdef CONFIG_KALLSYMS
3431 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3433 #ifdef CONFIG_STACKTRACE
3434 ONE("stack", S_IRUGO
, proc_pid_stack
),
3436 #ifdef CONFIG_SCHEDSTATS
3437 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3439 #ifdef CONFIG_LATENCYTOP
3440 REG("latency", S_IRUGO
, proc_lstats_operations
),
3442 #ifdef CONFIG_PROC_PID_CPUSET
3443 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3445 #ifdef CONFIG_CGROUPS
3446 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3448 INF("oom_score", S_IRUGO
, proc_oom_score
),
3449 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3450 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3451 #ifdef CONFIG_AUDITSYSCALL
3452 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3453 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3455 #ifdef CONFIG_FAULT_INJECTION
3456 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3458 #ifdef CONFIG_TASK_IO_ACCOUNTING
3459 INF("io", S_IRUSR
, proc_tid_io_accounting
),
3461 #ifdef CONFIG_HARDWALL
3462 INF("hardwall", S_IRUGO
, proc_pid_hardwall
),
3464 #ifdef CONFIG_USER_NS
3465 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3466 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3470 static int proc_tid_base_readdir(struct file
* filp
,
3471 void * dirent
, filldir_t filldir
)
3473 return proc_pident_readdir(filp
,dirent
,filldir
,
3474 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
3477 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3479 return proc_pident_lookup(dir
, dentry
,
3480 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3483 static const struct file_operations proc_tid_base_operations
= {
3484 .read
= generic_read_dir
,
3485 .readdir
= proc_tid_base_readdir
,
3486 .llseek
= default_llseek
,
3489 static const struct inode_operations proc_tid_base_inode_operations
= {
3490 .lookup
= proc_tid_base_lookup
,
3491 .getattr
= pid_getattr
,
3492 .setattr
= proc_setattr
,
3495 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
3496 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3498 struct dentry
*error
= ERR_PTR(-ENOENT
);
3499 struct inode
*inode
;
3500 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3504 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3505 inode
->i_op
= &proc_tid_base_inode_operations
;
3506 inode
->i_fop
= &proc_tid_base_operations
;
3507 inode
->i_flags
|=S_IMMUTABLE
;
3509 set_nlink(inode
, 2 + pid_entry_count_dirs(tid_base_stuff
,
3510 ARRAY_SIZE(tid_base_stuff
)));
3512 d_set_d_op(dentry
, &pid_dentry_operations
);
3514 d_add(dentry
, inode
);
3515 /* Close the race of the process dying before we return the dentry */
3516 if (pid_revalidate(dentry
, 0))
3522 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3524 struct dentry
*result
= ERR_PTR(-ENOENT
);
3525 struct task_struct
*task
;
3526 struct task_struct
*leader
= get_proc_task(dir
);
3528 struct pid_namespace
*ns
;
3533 tid
= name_to_int(dentry
);
3537 ns
= dentry
->d_sb
->s_fs_info
;
3539 task
= find_task_by_pid_ns(tid
, ns
);
3541 get_task_struct(task
);
3545 if (!same_thread_group(leader
, task
))
3548 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3550 put_task_struct(task
);
3552 put_task_struct(leader
);
3558 * Find the first tid of a thread group to return to user space.
3560 * Usually this is just the thread group leader, but if the users
3561 * buffer was too small or there was a seek into the middle of the
3562 * directory we have more work todo.
3564 * In the case of a short read we start with find_task_by_pid.
3566 * In the case of a seek we start with the leader and walk nr
3569 static struct task_struct
*first_tid(struct task_struct
*leader
,
3570 int tid
, int nr
, struct pid_namespace
*ns
)
3572 struct task_struct
*pos
;
3575 /* Attempt to start with the pid of a thread */
3576 if (tid
&& (nr
> 0)) {
3577 pos
= find_task_by_pid_ns(tid
, ns
);
3578 if (pos
&& (pos
->group_leader
== leader
))
3582 /* If nr exceeds the number of threads there is nothing todo */
3584 if (nr
&& nr
>= get_nr_threads(leader
))
3587 /* If we haven't found our starting place yet start
3588 * with the leader and walk nr threads forward.
3590 for (pos
= leader
; nr
> 0; --nr
) {
3591 pos
= next_thread(pos
);
3592 if (pos
== leader
) {
3598 get_task_struct(pos
);
3605 * Find the next thread in the thread list.
3606 * Return NULL if there is an error or no next thread.
3608 * The reference to the input task_struct is released.
3610 static struct task_struct
*next_tid(struct task_struct
*start
)
3612 struct task_struct
*pos
= NULL
;
3614 if (pid_alive(start
)) {
3615 pos
= next_thread(start
);
3616 if (thread_group_leader(pos
))
3619 get_task_struct(pos
);
3622 put_task_struct(start
);
3626 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3627 struct task_struct
*task
, int tid
)
3629 char name
[PROC_NUMBUF
];
3630 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3631 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3632 proc_task_instantiate
, task
, NULL
);
3635 /* for the /proc/TGID/task/ directories */
3636 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3638 struct dentry
*dentry
= filp
->f_path
.dentry
;
3639 struct inode
*inode
= dentry
->d_inode
;
3640 struct task_struct
*leader
= NULL
;
3641 struct task_struct
*task
;
3642 int retval
= -ENOENT
;
3645 struct pid_namespace
*ns
;
3647 task
= get_proc_task(inode
);
3651 if (pid_alive(task
)) {
3652 leader
= task
->group_leader
;
3653 get_task_struct(leader
);
3656 put_task_struct(task
);
3661 switch ((unsigned long)filp
->f_pos
) {
3664 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3669 ino
= parent_ino(dentry
);
3670 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3676 /* f_version caches the tgid value that the last readdir call couldn't
3677 * return. lseek aka telldir automagically resets f_version to 0.
3679 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3680 tid
= (int)filp
->f_version
;
3681 filp
->f_version
= 0;
3682 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3684 task
= next_tid(task
), filp
->f_pos
++) {
3685 tid
= task_pid_nr_ns(task
, ns
);
3686 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3687 /* returning this tgid failed, save it as the first
3688 * pid for the next readir call */
3689 filp
->f_version
= (u64
)tid
;
3690 put_task_struct(task
);
3695 put_task_struct(leader
);
3700 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3702 struct inode
*inode
= dentry
->d_inode
;
3703 struct task_struct
*p
= get_proc_task(inode
);
3704 generic_fillattr(inode
, stat
);
3707 stat
->nlink
+= get_nr_threads(p
);
3714 static const struct inode_operations proc_task_inode_operations
= {
3715 .lookup
= proc_task_lookup
,
3716 .getattr
= proc_task_getattr
,
3717 .setattr
= proc_setattr
,
3718 .permission
= proc_pid_permission
,
3721 static const struct file_operations proc_task_operations
= {
3722 .read
= generic_read_dir
,
3723 .readdir
= proc_task_readdir
,
3724 .llseek
= default_llseek
,