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 file
->private_data
= mm
;
703 static int mem_open(struct inode
*inode
, struct file
*file
)
705 int ret
= __mem_open(inode
, file
, PTRACE_MODE_ATTACH
);
707 /* OK to pass negative loff_t, we can catch out-of-range */
708 file
->f_mode
|= FMODE_UNSIGNED_OFFSET
;
713 static ssize_t
mem_rw(struct file
*file
, char __user
*buf
,
714 size_t count
, loff_t
*ppos
, int write
)
716 struct mm_struct
*mm
= file
->private_data
;
717 unsigned long addr
= *ppos
;
724 page
= (char *)__get_free_page(GFP_TEMPORARY
);
729 if (!atomic_inc_not_zero(&mm
->mm_users
))
733 int this_len
= min_t(int, count
, PAGE_SIZE
);
735 if (write
&& copy_from_user(page
, buf
, this_len
)) {
740 this_len
= access_remote_vm(mm
, addr
, page
, this_len
, write
);
747 if (!write
&& copy_to_user(buf
, page
, this_len
)) {
761 free_page((unsigned long) page
);
765 static ssize_t
mem_read(struct file
*file
, char __user
*buf
,
766 size_t count
, loff_t
*ppos
)
768 return mem_rw(file
, buf
, count
, ppos
, 0);
771 static ssize_t
mem_write(struct file
*file
, const char __user
*buf
,
772 size_t count
, loff_t
*ppos
)
774 return mem_rw(file
, (char __user
*)buf
, count
, ppos
, 1);
777 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
781 file
->f_pos
= offset
;
784 file
->f_pos
+= offset
;
789 force_successful_syscall_return();
793 static int mem_release(struct inode
*inode
, struct file
*file
)
795 struct mm_struct
*mm
= file
->private_data
;
801 static const struct file_operations proc_mem_operations
= {
806 .release
= mem_release
,
809 static int environ_open(struct inode
*inode
, struct file
*file
)
811 return __mem_open(inode
, file
, PTRACE_MODE_READ
);
814 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
815 size_t count
, loff_t
*ppos
)
818 unsigned long src
= *ppos
;
820 struct mm_struct
*mm
= file
->private_data
;
825 page
= (char *)__get_free_page(GFP_TEMPORARY
);
830 if (!atomic_inc_not_zero(&mm
->mm_users
))
833 size_t this_len
, max_len
;
836 if (src
>= (mm
->env_end
- mm
->env_start
))
839 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
841 max_len
= min_t(size_t, PAGE_SIZE
, count
);
842 this_len
= min(max_len
, this_len
);
844 retval
= access_remote_vm(mm
, (mm
->env_start
+ src
),
852 if (copy_to_user(buf
, page
, retval
)) {
866 free_page((unsigned long) page
);
870 static const struct file_operations proc_environ_operations
= {
871 .open
= environ_open
,
872 .read
= environ_read
,
873 .llseek
= generic_file_llseek
,
874 .release
= mem_release
,
877 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
878 size_t count
, loff_t
*ppos
)
880 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
881 char buffer
[PROC_NUMBUF
];
883 int oom_adjust
= OOM_DISABLE
;
889 if (lock_task_sighand(task
, &flags
)) {
890 oom_adjust
= task
->signal
->oom_adj
;
891 unlock_task_sighand(task
, &flags
);
894 put_task_struct(task
);
896 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
898 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
901 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
902 size_t count
, loff_t
*ppos
)
904 struct task_struct
*task
;
905 char buffer
[PROC_NUMBUF
];
910 memset(buffer
, 0, sizeof(buffer
));
911 if (count
> sizeof(buffer
) - 1)
912 count
= sizeof(buffer
) - 1;
913 if (copy_from_user(buffer
, buf
, count
)) {
918 err
= kstrtoint(strstrip(buffer
), 0, &oom_adjust
);
921 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
922 oom_adjust
!= OOM_DISABLE
) {
927 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
939 if (!lock_task_sighand(task
, &flags
)) {
944 if (oom_adjust
< task
->signal
->oom_adj
&& !capable(CAP_SYS_RESOURCE
)) {
950 * Warn that /proc/pid/oom_adj is deprecated, see
951 * Documentation/feature-removal-schedule.txt.
953 printk_once(KERN_WARNING
"%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
954 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
956 task
->signal
->oom_adj
= oom_adjust
;
958 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
959 * value is always attainable.
961 if (task
->signal
->oom_adj
== OOM_ADJUST_MAX
)
962 task
->signal
->oom_score_adj
= OOM_SCORE_ADJ_MAX
;
964 task
->signal
->oom_score_adj
= (oom_adjust
* OOM_SCORE_ADJ_MAX
) /
966 trace_oom_score_adj_update(task
);
968 unlock_task_sighand(task
, &flags
);
971 put_task_struct(task
);
973 return err
< 0 ? err
: count
;
976 static const struct file_operations proc_oom_adjust_operations
= {
977 .read
= oom_adjust_read
,
978 .write
= oom_adjust_write
,
979 .llseek
= generic_file_llseek
,
982 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
983 size_t count
, loff_t
*ppos
)
985 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
986 char buffer
[PROC_NUMBUF
];
987 int oom_score_adj
= OOM_SCORE_ADJ_MIN
;
993 if (lock_task_sighand(task
, &flags
)) {
994 oom_score_adj
= task
->signal
->oom_score_adj
;
995 unlock_task_sighand(task
, &flags
);
997 put_task_struct(task
);
998 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_score_adj
);
999 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1002 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1003 size_t count
, loff_t
*ppos
)
1005 struct task_struct
*task
;
1006 char buffer
[PROC_NUMBUF
];
1007 unsigned long flags
;
1011 memset(buffer
, 0, sizeof(buffer
));
1012 if (count
> sizeof(buffer
) - 1)
1013 count
= sizeof(buffer
) - 1;
1014 if (copy_from_user(buffer
, buf
, count
)) {
1019 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1022 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1023 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1028 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1040 if (!lock_task_sighand(task
, &flags
)) {
1045 if (oom_score_adj
< task
->signal
->oom_score_adj_min
&&
1046 !capable(CAP_SYS_RESOURCE
)) {
1051 task
->signal
->oom_score_adj
= oom_score_adj
;
1052 if (has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1053 task
->signal
->oom_score_adj_min
= oom_score_adj
;
1054 trace_oom_score_adj_update(task
);
1056 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1057 * always attainable.
1059 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MIN
)
1060 task
->signal
->oom_adj
= OOM_DISABLE
;
1062 task
->signal
->oom_adj
= (oom_score_adj
* OOM_ADJUST_MAX
) /
1065 unlock_task_sighand(task
, &flags
);
1068 put_task_struct(task
);
1070 return err
< 0 ? err
: count
;
1073 static const struct file_operations proc_oom_score_adj_operations
= {
1074 .read
= oom_score_adj_read
,
1075 .write
= oom_score_adj_write
,
1076 .llseek
= default_llseek
,
1079 #ifdef CONFIG_AUDITSYSCALL
1080 #define TMPBUFLEN 21
1081 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1082 size_t count
, loff_t
*ppos
)
1084 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1085 struct task_struct
*task
= get_proc_task(inode
);
1087 char tmpbuf
[TMPBUFLEN
];
1091 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1092 audit_get_loginuid(task
));
1093 put_task_struct(task
);
1094 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1097 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1098 size_t count
, loff_t
*ppos
)
1100 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1106 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1112 if (count
>= PAGE_SIZE
)
1113 count
= PAGE_SIZE
- 1;
1116 /* No partial writes. */
1119 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1123 if (copy_from_user(page
, buf
, count
))
1127 loginuid
= simple_strtoul(page
, &tmp
, 10);
1133 length
= audit_set_loginuid(loginuid
);
1134 if (likely(length
== 0))
1138 free_page((unsigned long) page
);
1142 static const struct file_operations proc_loginuid_operations
= {
1143 .read
= proc_loginuid_read
,
1144 .write
= proc_loginuid_write
,
1145 .llseek
= generic_file_llseek
,
1148 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1149 size_t count
, loff_t
*ppos
)
1151 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1152 struct task_struct
*task
= get_proc_task(inode
);
1154 char tmpbuf
[TMPBUFLEN
];
1158 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1159 audit_get_sessionid(task
));
1160 put_task_struct(task
);
1161 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1164 static const struct file_operations proc_sessionid_operations
= {
1165 .read
= proc_sessionid_read
,
1166 .llseek
= generic_file_llseek
,
1170 #ifdef CONFIG_FAULT_INJECTION
1171 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1172 size_t count
, loff_t
*ppos
)
1174 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1175 char buffer
[PROC_NUMBUF
];
1181 make_it_fail
= task
->make_it_fail
;
1182 put_task_struct(task
);
1184 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1186 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1189 static ssize_t
proc_fault_inject_write(struct file
* file
,
1190 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1192 struct task_struct
*task
;
1193 char buffer
[PROC_NUMBUF
], *end
;
1196 if (!capable(CAP_SYS_RESOURCE
))
1198 memset(buffer
, 0, sizeof(buffer
));
1199 if (count
> sizeof(buffer
) - 1)
1200 count
= sizeof(buffer
) - 1;
1201 if (copy_from_user(buffer
, buf
, count
))
1203 make_it_fail
= simple_strtol(strstrip(buffer
), &end
, 0);
1206 task
= get_proc_task(file
->f_dentry
->d_inode
);
1209 task
->make_it_fail
= make_it_fail
;
1210 put_task_struct(task
);
1215 static const struct file_operations proc_fault_inject_operations
= {
1216 .read
= proc_fault_inject_read
,
1217 .write
= proc_fault_inject_write
,
1218 .llseek
= generic_file_llseek
,
1223 #ifdef CONFIG_SCHED_DEBUG
1225 * Print out various scheduling related per-task fields:
1227 static int sched_show(struct seq_file
*m
, void *v
)
1229 struct inode
*inode
= m
->private;
1230 struct task_struct
*p
;
1232 p
= get_proc_task(inode
);
1235 proc_sched_show_task(p
, m
);
1243 sched_write(struct file
*file
, const char __user
*buf
,
1244 size_t count
, loff_t
*offset
)
1246 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1247 struct task_struct
*p
;
1249 p
= get_proc_task(inode
);
1252 proc_sched_set_task(p
);
1259 static int sched_open(struct inode
*inode
, struct file
*filp
)
1261 return single_open(filp
, sched_show
, inode
);
1264 static const struct file_operations proc_pid_sched_operations
= {
1267 .write
= sched_write
,
1268 .llseek
= seq_lseek
,
1269 .release
= single_release
,
1274 #ifdef CONFIG_SCHED_AUTOGROUP
1276 * Print out autogroup related information:
1278 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1280 struct inode
*inode
= m
->private;
1281 struct task_struct
*p
;
1283 p
= get_proc_task(inode
);
1286 proc_sched_autogroup_show_task(p
, m
);
1294 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1295 size_t count
, loff_t
*offset
)
1297 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1298 struct task_struct
*p
;
1299 char buffer
[PROC_NUMBUF
];
1303 memset(buffer
, 0, sizeof(buffer
));
1304 if (count
> sizeof(buffer
) - 1)
1305 count
= sizeof(buffer
) - 1;
1306 if (copy_from_user(buffer
, buf
, count
))
1309 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1313 p
= get_proc_task(inode
);
1317 err
= proc_sched_autogroup_set_nice(p
, nice
);
1326 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1330 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1332 struct seq_file
*m
= filp
->private_data
;
1339 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1340 .open
= sched_autogroup_open
,
1342 .write
= sched_autogroup_write
,
1343 .llseek
= seq_lseek
,
1344 .release
= single_release
,
1347 #endif /* CONFIG_SCHED_AUTOGROUP */
1349 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1350 size_t count
, loff_t
*offset
)
1352 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1353 struct task_struct
*p
;
1354 char buffer
[TASK_COMM_LEN
];
1356 memset(buffer
, 0, sizeof(buffer
));
1357 if (count
> sizeof(buffer
) - 1)
1358 count
= sizeof(buffer
) - 1;
1359 if (copy_from_user(buffer
, buf
, count
))
1362 p
= get_proc_task(inode
);
1366 if (same_thread_group(current
, p
))
1367 set_task_comm(p
, buffer
);
1376 static int comm_show(struct seq_file
*m
, void *v
)
1378 struct inode
*inode
= m
->private;
1379 struct task_struct
*p
;
1381 p
= get_proc_task(inode
);
1386 seq_printf(m
, "%s\n", p
->comm
);
1394 static int comm_open(struct inode
*inode
, struct file
*filp
)
1396 return single_open(filp
, comm_show
, inode
);
1399 static const struct file_operations proc_pid_set_comm_operations
= {
1402 .write
= comm_write
,
1403 .llseek
= seq_lseek
,
1404 .release
= single_release
,
1407 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1409 struct task_struct
*task
;
1410 struct mm_struct
*mm
;
1411 struct file
*exe_file
;
1413 task
= get_proc_task(dentry
->d_inode
);
1416 mm
= get_task_mm(task
);
1417 put_task_struct(task
);
1420 exe_file
= get_mm_exe_file(mm
);
1423 *exe_path
= exe_file
->f_path
;
1424 path_get(&exe_file
->f_path
);
1431 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1433 struct inode
*inode
= dentry
->d_inode
;
1435 int error
= -EACCES
;
1437 /* Are we allowed to snoop on the tasks file descriptors? */
1438 if (!proc_fd_access_allowed(inode
))
1441 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1445 nd_jump_link(nd
, &path
);
1448 return ERR_PTR(error
);
1451 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1453 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1460 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1461 len
= PTR_ERR(pathname
);
1462 if (IS_ERR(pathname
))
1464 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1468 if (copy_to_user(buffer
, pathname
, len
))
1471 free_page((unsigned long)tmp
);
1475 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1477 int error
= -EACCES
;
1478 struct inode
*inode
= dentry
->d_inode
;
1481 /* Are we allowed to snoop on the tasks file descriptors? */
1482 if (!proc_fd_access_allowed(inode
))
1485 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1489 error
= do_proc_readlink(&path
, buffer
, buflen
);
1495 static const struct inode_operations proc_pid_link_inode_operations
= {
1496 .readlink
= proc_pid_readlink
,
1497 .follow_link
= proc_pid_follow_link
,
1498 .setattr
= proc_setattr
,
1502 /* building an inode */
1504 static int task_dumpable(struct task_struct
*task
)
1507 struct mm_struct
*mm
;
1512 dumpable
= get_dumpable(mm
);
1519 struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1521 struct inode
* inode
;
1522 struct proc_inode
*ei
;
1523 const struct cred
*cred
;
1525 /* We need a new inode */
1527 inode
= new_inode(sb
);
1533 inode
->i_ino
= get_next_ino();
1534 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1535 inode
->i_op
= &proc_def_inode_operations
;
1538 * grab the reference to task.
1540 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1544 if (task_dumpable(task
)) {
1546 cred
= __task_cred(task
);
1547 inode
->i_uid
= cred
->euid
;
1548 inode
->i_gid
= cred
->egid
;
1551 security_task_to_inode(task
, inode
);
1561 int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1563 struct inode
*inode
= dentry
->d_inode
;
1564 struct task_struct
*task
;
1565 const struct cred
*cred
;
1566 struct pid_namespace
*pid
= dentry
->d_sb
->s_fs_info
;
1568 generic_fillattr(inode
, stat
);
1571 stat
->uid
= GLOBAL_ROOT_UID
;
1572 stat
->gid
= GLOBAL_ROOT_GID
;
1573 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1575 if (!has_pid_permissions(pid
, task
, 2)) {
1578 * This doesn't prevent learning whether PID exists,
1579 * it only makes getattr() consistent with readdir().
1583 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1584 task_dumpable(task
)) {
1585 cred
= __task_cred(task
);
1586 stat
->uid
= cred
->euid
;
1587 stat
->gid
= cred
->egid
;
1597 * Exceptional case: normally we are not allowed to unhash a busy
1598 * directory. In this case, however, we can do it - no aliasing problems
1599 * due to the way we treat inodes.
1601 * Rewrite the inode's ownerships here because the owning task may have
1602 * performed a setuid(), etc.
1604 * Before the /proc/pid/status file was created the only way to read
1605 * the effective uid of a /process was to stat /proc/pid. Reading
1606 * /proc/pid/status is slow enough that procps and other packages
1607 * kept stating /proc/pid. To keep the rules in /proc simple I have
1608 * made this apply to all per process world readable and executable
1611 int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1613 struct inode
*inode
;
1614 struct task_struct
*task
;
1615 const struct cred
*cred
;
1617 if (flags
& LOOKUP_RCU
)
1620 inode
= dentry
->d_inode
;
1621 task
= get_proc_task(inode
);
1624 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1625 task_dumpable(task
)) {
1627 cred
= __task_cred(task
);
1628 inode
->i_uid
= cred
->euid
;
1629 inode
->i_gid
= cred
->egid
;
1632 inode
->i_uid
= GLOBAL_ROOT_UID
;
1633 inode
->i_gid
= GLOBAL_ROOT_GID
;
1635 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1636 security_task_to_inode(task
, inode
);
1637 put_task_struct(task
);
1644 static int pid_delete_dentry(const struct dentry
* dentry
)
1646 /* Is the task we represent dead?
1647 * If so, then don't put the dentry on the lru list,
1648 * kill it immediately.
1650 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1653 const struct dentry_operations pid_dentry_operations
=
1655 .d_revalidate
= pid_revalidate
,
1656 .d_delete
= pid_delete_dentry
,
1662 * Fill a directory entry.
1664 * If possible create the dcache entry and derive our inode number and
1665 * file type from dcache entry.
1667 * Since all of the proc inode numbers are dynamically generated, the inode
1668 * numbers do not exist until the inode is cache. This means creating the
1669 * the dcache entry in readdir is necessary to keep the inode numbers
1670 * reported by readdir in sync with the inode numbers reported
1673 int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1674 const char *name
, int len
,
1675 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1677 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1678 struct inode
*inode
;
1681 unsigned type
= DT_UNKNOWN
;
1685 qname
.hash
= full_name_hash(name
, len
);
1687 child
= d_lookup(dir
, &qname
);
1690 new = d_alloc(dir
, &qname
);
1692 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1699 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1700 goto end_instantiate
;
1701 inode
= child
->d_inode
;
1704 type
= inode
->i_mode
>> 12;
1709 ino
= find_inode_number(dir
, &qname
);
1712 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1715 static unsigned name_to_int(struct dentry
*dentry
)
1717 const char *name
= dentry
->d_name
.name
;
1718 int len
= dentry
->d_name
.len
;
1721 if (len
> 1 && *name
== '0')
1724 unsigned c
= *name
++ - '0';
1727 if (n
>= (~0U-9)/10)
1737 #define PROC_FDINFO_MAX 64
1739 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1741 struct task_struct
*task
= get_proc_task(inode
);
1742 struct files_struct
*files
= NULL
;
1744 int fd
= proc_fd(inode
);
1747 files
= get_files_struct(task
);
1748 put_task_struct(task
);
1752 * We are not taking a ref to the file structure, so we must
1755 spin_lock(&files
->file_lock
);
1756 file
= fcheck_files(files
, fd
);
1758 unsigned int f_flags
;
1759 struct fdtable
*fdt
;
1761 fdt
= files_fdtable(files
);
1762 f_flags
= file
->f_flags
& ~O_CLOEXEC
;
1763 if (close_on_exec(fd
, fdt
))
1764 f_flags
|= O_CLOEXEC
;
1767 *path
= file
->f_path
;
1768 path_get(&file
->f_path
);
1771 snprintf(info
, PROC_FDINFO_MAX
,
1774 (long long) file
->f_pos
,
1776 spin_unlock(&files
->file_lock
);
1777 put_files_struct(files
);
1780 spin_unlock(&files
->file_lock
);
1781 put_files_struct(files
);
1786 static int proc_fd_link(struct dentry
*dentry
, struct path
*path
)
1788 return proc_fd_info(dentry
->d_inode
, path
, NULL
);
1791 static int tid_fd_revalidate(struct dentry
*dentry
, unsigned int flags
)
1793 struct inode
*inode
;
1794 struct task_struct
*task
;
1796 struct files_struct
*files
;
1797 const struct cred
*cred
;
1799 if (flags
& LOOKUP_RCU
)
1802 inode
= dentry
->d_inode
;
1803 task
= get_proc_task(inode
);
1804 fd
= proc_fd(inode
);
1807 files
= get_files_struct(task
);
1811 file
= fcheck_files(files
, fd
);
1813 unsigned f_mode
= file
->f_mode
;
1816 put_files_struct(files
);
1818 if (task_dumpable(task
)) {
1820 cred
= __task_cred(task
);
1821 inode
->i_uid
= cred
->euid
;
1822 inode
->i_gid
= cred
->egid
;
1825 inode
->i_uid
= GLOBAL_ROOT_UID
;
1826 inode
->i_gid
= GLOBAL_ROOT_GID
;
1829 if (S_ISLNK(inode
->i_mode
)) {
1830 unsigned i_mode
= S_IFLNK
;
1831 if (f_mode
& FMODE_READ
)
1832 i_mode
|= S_IRUSR
| S_IXUSR
;
1833 if (f_mode
& FMODE_WRITE
)
1834 i_mode
|= S_IWUSR
| S_IXUSR
;
1835 inode
->i_mode
= i_mode
;
1838 security_task_to_inode(task
, inode
);
1839 put_task_struct(task
);
1843 put_files_struct(files
);
1845 put_task_struct(task
);
1851 static const struct dentry_operations tid_fd_dentry_operations
=
1853 .d_revalidate
= tid_fd_revalidate
,
1854 .d_delete
= pid_delete_dentry
,
1857 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1858 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1860 unsigned fd
= (unsigned long)ptr
;
1861 struct inode
*inode
;
1862 struct proc_inode
*ei
;
1863 struct dentry
*error
= ERR_PTR(-ENOENT
);
1865 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1871 inode
->i_mode
= S_IFLNK
;
1872 inode
->i_op
= &proc_pid_link_inode_operations
;
1874 ei
->op
.proc_get_link
= proc_fd_link
;
1875 d_set_d_op(dentry
, &tid_fd_dentry_operations
);
1876 d_add(dentry
, inode
);
1877 /* Close the race of the process dying before we return the dentry */
1878 if (tid_fd_revalidate(dentry
, 0))
1885 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1886 struct dentry
*dentry
,
1887 instantiate_t instantiate
)
1889 struct task_struct
*task
= get_proc_task(dir
);
1890 unsigned fd
= name_to_int(dentry
);
1891 struct dentry
*result
= ERR_PTR(-ENOENT
);
1898 result
= instantiate(dir
, dentry
, task
, (void *)(unsigned long)fd
);
1900 put_task_struct(task
);
1905 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1906 filldir_t filldir
, instantiate_t instantiate
)
1908 struct dentry
*dentry
= filp
->f_path
.dentry
;
1909 struct inode
*inode
= dentry
->d_inode
;
1910 struct task_struct
*p
= get_proc_task(inode
);
1911 unsigned int fd
, ino
;
1913 struct files_struct
* files
;
1923 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1927 ino
= parent_ino(dentry
);
1928 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1932 files
= get_files_struct(p
);
1936 for (fd
= filp
->f_pos
-2;
1937 fd
< files_fdtable(files
)->max_fds
;
1938 fd
++, filp
->f_pos
++) {
1939 char name
[PROC_NUMBUF
];
1943 if (!fcheck_files(files
, fd
))
1947 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1948 rv
= proc_fill_cache(filp
, dirent
, filldir
,
1949 name
, len
, instantiate
, p
,
1950 (void *)(unsigned long)fd
);
1957 put_files_struct(files
);
1965 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1968 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1971 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1973 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1976 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1977 size_t len
, loff_t
*ppos
)
1979 char tmp
[PROC_FDINFO_MAX
];
1980 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1982 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1986 static const struct file_operations proc_fdinfo_file_operations
= {
1987 .open
= nonseekable_open
,
1988 .read
= proc_fdinfo_read
,
1989 .llseek
= no_llseek
,
1992 static const struct file_operations proc_fd_operations
= {
1993 .read
= generic_read_dir
,
1994 .readdir
= proc_readfd
,
1995 .llseek
= default_llseek
,
1998 #ifdef CONFIG_CHECKPOINT_RESTORE
2001 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2002 * which represent vma start and end addresses.
2004 static int dname_to_vma_addr(struct dentry
*dentry
,
2005 unsigned long *start
, unsigned long *end
)
2007 if (sscanf(dentry
->d_name
.name
, "%lx-%lx", start
, end
) != 2)
2013 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
2015 unsigned long vm_start
, vm_end
;
2016 bool exact_vma_exists
= false;
2017 struct mm_struct
*mm
= NULL
;
2018 struct task_struct
*task
;
2019 const struct cred
*cred
;
2020 struct inode
*inode
;
2023 if (flags
& LOOKUP_RCU
)
2026 if (!capable(CAP_SYS_ADMIN
)) {
2031 inode
= dentry
->d_inode
;
2032 task
= get_proc_task(inode
);
2036 mm
= mm_access(task
, PTRACE_MODE_READ
);
2037 if (IS_ERR_OR_NULL(mm
))
2040 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
2041 down_read(&mm
->mmap_sem
);
2042 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
2043 up_read(&mm
->mmap_sem
);
2048 if (exact_vma_exists
) {
2049 if (task_dumpable(task
)) {
2051 cred
= __task_cred(task
);
2052 inode
->i_uid
= cred
->euid
;
2053 inode
->i_gid
= cred
->egid
;
2056 inode
->i_uid
= GLOBAL_ROOT_UID
;
2057 inode
->i_gid
= GLOBAL_ROOT_GID
;
2059 security_task_to_inode(task
, inode
);
2064 put_task_struct(task
);
2073 static const struct dentry_operations tid_map_files_dentry_operations
= {
2074 .d_revalidate
= map_files_d_revalidate
,
2075 .d_delete
= pid_delete_dentry
,
2078 static int proc_map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2080 unsigned long vm_start
, vm_end
;
2081 struct vm_area_struct
*vma
;
2082 struct task_struct
*task
;
2083 struct mm_struct
*mm
;
2087 task
= get_proc_task(dentry
->d_inode
);
2091 mm
= get_task_mm(task
);
2092 put_task_struct(task
);
2096 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2100 down_read(&mm
->mmap_sem
);
2101 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2102 if (vma
&& vma
->vm_file
) {
2103 *path
= vma
->vm_file
->f_path
;
2107 up_read(&mm
->mmap_sem
);
2115 struct map_files_info
{
2118 unsigned char name
[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2121 static struct dentry
*
2122 proc_map_files_instantiate(struct inode
*dir
, struct dentry
*dentry
,
2123 struct task_struct
*task
, const void *ptr
)
2125 const struct file
*file
= ptr
;
2126 struct proc_inode
*ei
;
2127 struct inode
*inode
;
2130 return ERR_PTR(-ENOENT
);
2132 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2134 return ERR_PTR(-ENOENT
);
2137 ei
->op
.proc_get_link
= proc_map_files_get_link
;
2139 inode
->i_op
= &proc_pid_link_inode_operations
;
2141 inode
->i_mode
= S_IFLNK
;
2143 if (file
->f_mode
& FMODE_READ
)
2144 inode
->i_mode
|= S_IRUSR
;
2145 if (file
->f_mode
& FMODE_WRITE
)
2146 inode
->i_mode
|= S_IWUSR
;
2148 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2149 d_add(dentry
, inode
);
2154 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2155 struct dentry
*dentry
, unsigned int flags
)
2157 unsigned long vm_start
, vm_end
;
2158 struct vm_area_struct
*vma
;
2159 struct task_struct
*task
;
2160 struct dentry
*result
;
2161 struct mm_struct
*mm
;
2163 result
= ERR_PTR(-EACCES
);
2164 if (!capable(CAP_SYS_ADMIN
))
2167 result
= ERR_PTR(-ENOENT
);
2168 task
= get_proc_task(dir
);
2172 result
= ERR_PTR(-EACCES
);
2173 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
2176 result
= ERR_PTR(-ENOENT
);
2177 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2180 mm
= get_task_mm(task
);
2184 down_read(&mm
->mmap_sem
);
2185 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2189 result
= proc_map_files_instantiate(dir
, dentry
, task
, vma
->vm_file
);
2192 up_read(&mm
->mmap_sem
);
2195 put_task_struct(task
);
2200 static const struct inode_operations proc_map_files_inode_operations
= {
2201 .lookup
= proc_map_files_lookup
,
2202 .permission
= proc_fd_permission
,
2203 .setattr
= proc_setattr
,
2207 proc_map_files_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
2209 struct dentry
*dentry
= filp
->f_path
.dentry
;
2210 struct inode
*inode
= dentry
->d_inode
;
2211 struct vm_area_struct
*vma
;
2212 struct task_struct
*task
;
2213 struct mm_struct
*mm
;
2218 if (!capable(CAP_SYS_ADMIN
))
2222 task
= get_proc_task(inode
);
2227 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
2231 switch (filp
->f_pos
) {
2234 if (filldir(dirent
, ".", 1, 0, ino
, DT_DIR
) < 0)
2238 ino
= parent_ino(dentry
);
2239 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
2244 unsigned long nr_files
, pos
, i
;
2245 struct flex_array
*fa
= NULL
;
2246 struct map_files_info info
;
2247 struct map_files_info
*p
;
2249 mm
= get_task_mm(task
);
2252 down_read(&mm
->mmap_sem
);
2257 * We need two passes here:
2259 * 1) Collect vmas of mapped files with mmap_sem taken
2260 * 2) Release mmap_sem and instantiate entries
2262 * otherwise we get lockdep complained, since filldir()
2263 * routine might require mmap_sem taken in might_fault().
2266 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2267 if (vma
->vm_file
&& ++pos
> filp
->f_pos
)
2272 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2274 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2278 flex_array_free(fa
);
2279 up_read(&mm
->mmap_sem
);
2283 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2284 vma
= vma
->vm_next
) {
2287 if (++pos
<= filp
->f_pos
)
2290 get_file(vma
->vm_file
);
2291 info
.file
= vma
->vm_file
;
2292 info
.len
= snprintf(info
.name
,
2293 sizeof(info
.name
), "%lx-%lx",
2294 vma
->vm_start
, vma
->vm_end
);
2295 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2299 up_read(&mm
->mmap_sem
);
2301 for (i
= 0; i
< nr_files
; i
++) {
2302 p
= flex_array_get(fa
, i
);
2303 ret
= proc_fill_cache(filp
, dirent
, filldir
,
2305 proc_map_files_instantiate
,
2312 for (; i
< nr_files
; i
++) {
2314 * In case of error don't forget
2315 * to put rest of file refs.
2317 p
= flex_array_get(fa
, i
);
2321 flex_array_free(fa
);
2327 put_task_struct(task
);
2332 static const struct file_operations proc_map_files_operations
= {
2333 .read
= generic_read_dir
,
2334 .readdir
= proc_map_files_readdir
,
2335 .llseek
= default_llseek
,
2338 #endif /* CONFIG_CHECKPOINT_RESTORE */
2341 * /proc/pid/fd needs a special permission handler so that a process can still
2342 * access /proc/self/fd after it has executed a setuid().
2344 static int proc_fd_permission(struct inode
*inode
, int mask
)
2346 int rv
= generic_permission(inode
, mask
);
2349 if (task_pid(current
) == proc_pid(inode
))
2355 * proc directories can do almost nothing..
2357 static const struct inode_operations proc_fd_inode_operations
= {
2358 .lookup
= proc_lookupfd
,
2359 .permission
= proc_fd_permission
,
2360 .setattr
= proc_setattr
,
2363 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
2364 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2366 unsigned fd
= (unsigned long)ptr
;
2367 struct inode
*inode
;
2368 struct proc_inode
*ei
;
2369 struct dentry
*error
= ERR_PTR(-ENOENT
);
2371 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2376 inode
->i_mode
= S_IFREG
| S_IRUSR
;
2377 inode
->i_fop
= &proc_fdinfo_file_operations
;
2378 d_set_d_op(dentry
, &tid_fd_dentry_operations
);
2379 d_add(dentry
, inode
);
2380 /* Close the race of the process dying before we return the dentry */
2381 if (tid_fd_revalidate(dentry
, 0))
2388 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
2389 struct dentry
*dentry
,
2392 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
2395 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
2397 return proc_readfd_common(filp
, dirent
, filldir
,
2398 proc_fdinfo_instantiate
);
2401 static const struct file_operations proc_fdinfo_operations
= {
2402 .read
= generic_read_dir
,
2403 .readdir
= proc_readfdinfo
,
2404 .llseek
= default_llseek
,
2408 * proc directories can do almost nothing..
2410 static const struct inode_operations proc_fdinfo_inode_operations
= {
2411 .lookup
= proc_lookupfdinfo
,
2412 .setattr
= proc_setattr
,
2416 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
2417 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2419 const struct pid_entry
*p
= ptr
;
2420 struct inode
*inode
;
2421 struct proc_inode
*ei
;
2422 struct dentry
*error
= ERR_PTR(-ENOENT
);
2424 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2429 inode
->i_mode
= p
->mode
;
2430 if (S_ISDIR(inode
->i_mode
))
2431 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2433 inode
->i_op
= p
->iop
;
2435 inode
->i_fop
= p
->fop
;
2437 d_set_d_op(dentry
, &pid_dentry_operations
);
2438 d_add(dentry
, inode
);
2439 /* Close the race of the process dying before we return the dentry */
2440 if (pid_revalidate(dentry
, 0))
2446 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2447 struct dentry
*dentry
,
2448 const struct pid_entry
*ents
,
2451 struct dentry
*error
;
2452 struct task_struct
*task
= get_proc_task(dir
);
2453 const struct pid_entry
*p
, *last
;
2455 error
= ERR_PTR(-ENOENT
);
2461 * Yes, it does not scale. And it should not. Don't add
2462 * new entries into /proc/<tgid>/ without very good reasons.
2464 last
= &ents
[nents
- 1];
2465 for (p
= ents
; p
<= last
; p
++) {
2466 if (p
->len
!= dentry
->d_name
.len
)
2468 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2474 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2476 put_task_struct(task
);
2481 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2482 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2484 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2485 proc_pident_instantiate
, task
, p
);
2488 static int proc_pident_readdir(struct file
*filp
,
2489 void *dirent
, filldir_t filldir
,
2490 const struct pid_entry
*ents
, unsigned int nents
)
2493 struct dentry
*dentry
= filp
->f_path
.dentry
;
2494 struct inode
*inode
= dentry
->d_inode
;
2495 struct task_struct
*task
= get_proc_task(inode
);
2496 const struct pid_entry
*p
, *last
;
2509 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2515 ino
= parent_ino(dentry
);
2516 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2528 last
= &ents
[nents
- 1];
2530 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2539 put_task_struct(task
);
2544 #ifdef CONFIG_SECURITY
2545 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2546 size_t count
, loff_t
*ppos
)
2548 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2551 struct task_struct
*task
= get_proc_task(inode
);
2556 length
= security_getprocattr(task
,
2557 (char*)file
->f_path
.dentry
->d_name
.name
,
2559 put_task_struct(task
);
2561 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2566 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2567 size_t count
, loff_t
*ppos
)
2569 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2572 struct task_struct
*task
= get_proc_task(inode
);
2577 if (count
> PAGE_SIZE
)
2580 /* No partial writes. */
2586 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2591 if (copy_from_user(page
, buf
, count
))
2594 /* Guard against adverse ptrace interaction */
2595 length
= mutex_lock_interruptible(&task
->signal
->cred_guard_mutex
);
2599 length
= security_setprocattr(task
,
2600 (char*)file
->f_path
.dentry
->d_name
.name
,
2601 (void*)page
, count
);
2602 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2604 free_page((unsigned long) page
);
2606 put_task_struct(task
);
2611 static const struct file_operations proc_pid_attr_operations
= {
2612 .read
= proc_pid_attr_read
,
2613 .write
= proc_pid_attr_write
,
2614 .llseek
= generic_file_llseek
,
2617 static const struct pid_entry attr_dir_stuff
[] = {
2618 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2619 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2620 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2621 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2622 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2623 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2626 static int proc_attr_dir_readdir(struct file
* filp
,
2627 void * dirent
, filldir_t filldir
)
2629 return proc_pident_readdir(filp
,dirent
,filldir
,
2630 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2633 static const struct file_operations proc_attr_dir_operations
= {
2634 .read
= generic_read_dir
,
2635 .readdir
= proc_attr_dir_readdir
,
2636 .llseek
= default_llseek
,
2639 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2640 struct dentry
*dentry
, unsigned int flags
)
2642 return proc_pident_lookup(dir
, dentry
,
2643 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2646 static const struct inode_operations proc_attr_dir_inode_operations
= {
2647 .lookup
= proc_attr_dir_lookup
,
2648 .getattr
= pid_getattr
,
2649 .setattr
= proc_setattr
,
2654 #ifdef CONFIG_ELF_CORE
2655 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2656 size_t count
, loff_t
*ppos
)
2658 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2659 struct mm_struct
*mm
;
2660 char buffer
[PROC_NUMBUF
];
2668 mm
= get_task_mm(task
);
2670 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2671 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2672 MMF_DUMP_FILTER_SHIFT
));
2674 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2677 put_task_struct(task
);
2682 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2683 const char __user
*buf
,
2687 struct task_struct
*task
;
2688 struct mm_struct
*mm
;
2689 char buffer
[PROC_NUMBUF
], *end
;
2696 memset(buffer
, 0, sizeof(buffer
));
2697 if (count
> sizeof(buffer
) - 1)
2698 count
= sizeof(buffer
) - 1;
2699 if (copy_from_user(buffer
, buf
, count
))
2703 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2706 if (end
- buffer
== 0)
2710 task
= get_proc_task(file
->f_dentry
->d_inode
);
2715 mm
= get_task_mm(task
);
2719 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2721 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2723 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2728 put_task_struct(task
);
2733 static const struct file_operations proc_coredump_filter_operations
= {
2734 .read
= proc_coredump_filter_read
,
2735 .write
= proc_coredump_filter_write
,
2736 .llseek
= generic_file_llseek
,
2743 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2746 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2747 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2748 char tmp
[PROC_NUMBUF
];
2751 sprintf(tmp
, "%d", tgid
);
2752 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2755 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2757 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2758 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2759 char *name
= ERR_PTR(-ENOENT
);
2763 name
= ERR_PTR(-ENOMEM
);
2765 sprintf(name
, "%d", tgid
);
2767 nd_set_link(nd
, name
);
2771 static void proc_self_put_link(struct dentry
*dentry
, struct nameidata
*nd
,
2774 char *s
= nd_get_link(nd
);
2779 static const struct inode_operations proc_self_inode_operations
= {
2780 .readlink
= proc_self_readlink
,
2781 .follow_link
= proc_self_follow_link
,
2782 .put_link
= proc_self_put_link
,
2788 * These are the directory entries in the root directory of /proc
2789 * that properly belong to the /proc filesystem, as they describe
2790 * describe something that is process related.
2792 static const struct pid_entry proc_base_stuff
[] = {
2793 NOD("self", S_IFLNK
|S_IRWXUGO
,
2794 &proc_self_inode_operations
, NULL
, {}),
2797 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2798 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2800 const struct pid_entry
*p
= ptr
;
2801 struct inode
*inode
;
2802 struct proc_inode
*ei
;
2803 struct dentry
*error
;
2805 /* Allocate the inode */
2806 error
= ERR_PTR(-ENOMEM
);
2807 inode
= new_inode(dir
->i_sb
);
2811 /* Initialize the inode */
2813 inode
->i_ino
= get_next_ino();
2814 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2817 * grab the reference to the task.
2819 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2823 inode
->i_mode
= p
->mode
;
2824 if (S_ISDIR(inode
->i_mode
))
2825 set_nlink(inode
, 2);
2826 if (S_ISLNK(inode
->i_mode
))
2829 inode
->i_op
= p
->iop
;
2831 inode
->i_fop
= p
->fop
;
2833 d_add(dentry
, inode
);
2842 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2844 struct dentry
*error
;
2845 struct task_struct
*task
= get_proc_task(dir
);
2846 const struct pid_entry
*p
, *last
;
2848 error
= ERR_PTR(-ENOENT
);
2853 /* Lookup the directory entry */
2854 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2855 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2856 if (p
->len
!= dentry
->d_name
.len
)
2858 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2864 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2867 put_task_struct(task
);
2872 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2873 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2875 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2876 proc_base_instantiate
, task
, p
);
2879 #ifdef CONFIG_TASK_IO_ACCOUNTING
2880 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2882 struct task_io_accounting acct
= task
->ioac
;
2883 unsigned long flags
;
2886 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2890 if (!ptrace_may_access(task
, PTRACE_MODE_READ
)) {
2895 if (whole
&& lock_task_sighand(task
, &flags
)) {
2896 struct task_struct
*t
= task
;
2898 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2899 while_each_thread(task
, t
)
2900 task_io_accounting_add(&acct
, &t
->ioac
);
2902 unlock_task_sighand(task
, &flags
);
2904 result
= sprintf(buffer
,
2909 "read_bytes: %llu\n"
2910 "write_bytes: %llu\n"
2911 "cancelled_write_bytes: %llu\n",
2912 (unsigned long long)acct
.rchar
,
2913 (unsigned long long)acct
.wchar
,
2914 (unsigned long long)acct
.syscr
,
2915 (unsigned long long)acct
.syscw
,
2916 (unsigned long long)acct
.read_bytes
,
2917 (unsigned long long)acct
.write_bytes
,
2918 (unsigned long long)acct
.cancelled_write_bytes
);
2920 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2924 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2926 return do_io_accounting(task
, buffer
, 0);
2929 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2931 return do_io_accounting(task
, buffer
, 1);
2933 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2935 #ifdef CONFIG_USER_NS
2936 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2937 struct seq_operations
*seq_ops
)
2939 struct user_namespace
*ns
= NULL
;
2940 struct task_struct
*task
;
2941 struct seq_file
*seq
;
2944 task
= get_proc_task(inode
);
2947 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2949 put_task_struct(task
);
2954 ret
= seq_open(file
, seq_ops
);
2958 seq
= file
->private_data
;
2968 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2970 struct seq_file
*seq
= file
->private_data
;
2971 struct user_namespace
*ns
= seq
->private;
2973 return seq_release(inode
, file
);
2976 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2978 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2981 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2983 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2986 static const struct file_operations proc_uid_map_operations
= {
2987 .open
= proc_uid_map_open
,
2988 .write
= proc_uid_map_write
,
2990 .llseek
= seq_lseek
,
2991 .release
= proc_id_map_release
,
2994 static const struct file_operations proc_gid_map_operations
= {
2995 .open
= proc_gid_map_open
,
2996 .write
= proc_gid_map_write
,
2998 .llseek
= seq_lseek
,
2999 .release
= proc_id_map_release
,
3001 #endif /* CONFIG_USER_NS */
3003 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
3004 struct pid
*pid
, struct task_struct
*task
)
3006 int err
= lock_trace(task
);
3008 seq_printf(m
, "%08x\n", task
->personality
);
3017 static const struct file_operations proc_task_operations
;
3018 static const struct inode_operations proc_task_inode_operations
;
3020 static const struct pid_entry tgid_base_stuff
[] = {
3021 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
3022 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3023 #ifdef CONFIG_CHECKPOINT_RESTORE
3024 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
3026 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3027 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3029 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3031 REG("environ", S_IRUSR
, proc_environ_operations
),
3032 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3033 ONE("status", S_IRUGO
, proc_pid_status
),
3034 ONE("personality", S_IRUGO
, proc_pid_personality
),
3035 INF("limits", S_IRUGO
, proc_pid_limits
),
3036 #ifdef CONFIG_SCHED_DEBUG
3037 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3039 #ifdef CONFIG_SCHED_AUTOGROUP
3040 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
3042 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3043 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3044 INF("syscall", S_IRUGO
, proc_pid_syscall
),
3046 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3047 ONE("stat", S_IRUGO
, proc_tgid_stat
),
3048 ONE("statm", S_IRUGO
, proc_pid_statm
),
3049 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3051 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3053 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3054 LNK("cwd", proc_cwd_link
),
3055 LNK("root", proc_root_link
),
3056 LNK("exe", proc_exe_link
),
3057 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3058 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3059 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
3060 #ifdef CONFIG_PROC_PAGE_MONITOR
3061 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3062 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3063 REG("pagemap", S_IRUGO
, proc_pagemap_operations
),
3065 #ifdef CONFIG_SECURITY
3066 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3068 #ifdef CONFIG_KALLSYMS
3069 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3071 #ifdef CONFIG_STACKTRACE
3072 ONE("stack", S_IRUGO
, proc_pid_stack
),
3074 #ifdef CONFIG_SCHEDSTATS
3075 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3077 #ifdef CONFIG_LATENCYTOP
3078 REG("latency", S_IRUGO
, proc_lstats_operations
),
3080 #ifdef CONFIG_PROC_PID_CPUSET
3081 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3083 #ifdef CONFIG_CGROUPS
3084 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3086 INF("oom_score", S_IRUGO
, proc_oom_score
),
3087 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3088 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3089 #ifdef CONFIG_AUDITSYSCALL
3090 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3091 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3093 #ifdef CONFIG_FAULT_INJECTION
3094 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3096 #ifdef CONFIG_ELF_CORE
3097 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3099 #ifdef CONFIG_TASK_IO_ACCOUNTING
3100 INF("io", S_IRUSR
, proc_tgid_io_accounting
),
3102 #ifdef CONFIG_HARDWALL
3103 INF("hardwall", S_IRUGO
, proc_pid_hardwall
),
3105 #ifdef CONFIG_USER_NS
3106 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3107 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3111 static int proc_tgid_base_readdir(struct file
* filp
,
3112 void * dirent
, filldir_t filldir
)
3114 return proc_pident_readdir(filp
,dirent
,filldir
,
3115 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
3118 static const struct file_operations proc_tgid_base_operations
= {
3119 .read
= generic_read_dir
,
3120 .readdir
= proc_tgid_base_readdir
,
3121 .llseek
= default_llseek
,
3124 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3126 return proc_pident_lookup(dir
, dentry
,
3127 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3130 static const struct inode_operations proc_tgid_base_inode_operations
= {
3131 .lookup
= proc_tgid_base_lookup
,
3132 .getattr
= pid_getattr
,
3133 .setattr
= proc_setattr
,
3134 .permission
= proc_pid_permission
,
3137 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
3139 struct dentry
*dentry
, *leader
, *dir
;
3140 char buf
[PROC_NUMBUF
];
3144 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3145 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3147 shrink_dcache_parent(dentry
);
3153 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
3154 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3159 name
.len
= strlen(name
.name
);
3160 dir
= d_hash_and_lookup(leader
, &name
);
3162 goto out_put_leader
;
3165 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3166 dentry
= d_hash_and_lookup(dir
, &name
);
3168 shrink_dcache_parent(dentry
);
3181 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3182 * @task: task that should be flushed.
3184 * When flushing dentries from proc, one needs to flush them from global
3185 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3186 * in. This call is supposed to do all of this job.
3188 * Looks in the dcache for
3190 * /proc/@tgid/task/@pid
3191 * if either directory is present flushes it and all of it'ts children
3194 * It is safe and reasonable to cache /proc entries for a task until
3195 * that task exits. After that they just clog up the dcache with
3196 * useless entries, possibly causing useful dcache entries to be
3197 * flushed instead. This routine is proved to flush those useless
3198 * dcache entries at process exit time.
3200 * NOTE: This routine is just an optimization so it does not guarantee
3201 * that no dcache entries will exist at process exit time it
3202 * just makes it very unlikely that any will persist.
3205 void proc_flush_task(struct task_struct
*task
)
3208 struct pid
*pid
, *tgid
;
3211 pid
= task_pid(task
);
3212 tgid
= task_tgid(task
);
3214 for (i
= 0; i
<= pid
->level
; i
++) {
3215 upid
= &pid
->numbers
[i
];
3216 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
3217 tgid
->numbers
[i
].nr
);
3220 upid
= &pid
->numbers
[pid
->level
];
3222 pid_ns_release_proc(upid
->ns
);
3225 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
3226 struct dentry
* dentry
,
3227 struct task_struct
*task
, const void *ptr
)
3229 struct dentry
*error
= ERR_PTR(-ENOENT
);
3230 struct inode
*inode
;
3232 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3236 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3237 inode
->i_op
= &proc_tgid_base_inode_operations
;
3238 inode
->i_fop
= &proc_tgid_base_operations
;
3239 inode
->i_flags
|=S_IMMUTABLE
;
3241 set_nlink(inode
, 2 + pid_entry_count_dirs(tgid_base_stuff
,
3242 ARRAY_SIZE(tgid_base_stuff
)));
3244 d_set_d_op(dentry
, &pid_dentry_operations
);
3246 d_add(dentry
, inode
);
3247 /* Close the race of the process dying before we return the dentry */
3248 if (pid_revalidate(dentry
, 0))
3254 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3256 struct dentry
*result
;
3257 struct task_struct
*task
;
3259 struct pid_namespace
*ns
;
3261 result
= proc_base_lookup(dir
, dentry
);
3262 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
3265 tgid
= name_to_int(dentry
);
3269 ns
= dentry
->d_sb
->s_fs_info
;
3271 task
= find_task_by_pid_ns(tgid
, ns
);
3273 get_task_struct(task
);
3278 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
3279 put_task_struct(task
);
3285 * Find the first task with tgid >= tgid
3290 struct task_struct
*task
;
3292 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3297 put_task_struct(iter
.task
);
3301 pid
= find_ge_pid(iter
.tgid
, ns
);
3303 iter
.tgid
= pid_nr_ns(pid
, ns
);
3304 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3305 /* What we to know is if the pid we have find is the
3306 * pid of a thread_group_leader. Testing for task
3307 * being a thread_group_leader is the obvious thing
3308 * todo but there is a window when it fails, due to
3309 * the pid transfer logic in de_thread.
3311 * So we perform the straight forward test of seeing
3312 * if the pid we have found is the pid of a thread
3313 * group leader, and don't worry if the task we have
3314 * found doesn't happen to be a thread group leader.
3315 * As we don't care in the case of readdir.
3317 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3321 get_task_struct(iter
.task
);
3327 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3329 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3330 struct tgid_iter iter
)
3332 char name
[PROC_NUMBUF
];
3333 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
3334 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3335 proc_pid_instantiate
, iter
.task
, NULL
);
3338 static int fake_filldir(void *buf
, const char *name
, int namelen
,
3339 loff_t offset
, u64 ino
, unsigned d_type
)
3344 /* for the /proc/ directory itself, after non-process stuff has been done */
3345 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3348 struct task_struct
*reaper
;
3349 struct tgid_iter iter
;
3350 struct pid_namespace
*ns
;
3351 filldir_t __filldir
;
3353 if (filp
->f_pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3355 nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
3357 reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
3361 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
3362 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
3363 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
3367 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3369 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
3370 for (iter
= next_tgid(ns
, iter
);
3372 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3373 if (has_pid_permissions(ns
, iter
.task
, 2))
3374 __filldir
= filldir
;
3376 __filldir
= fake_filldir
;
3378 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
3379 if (proc_pid_fill_cache(filp
, dirent
, __filldir
, iter
) < 0) {
3380 put_task_struct(iter
.task
);
3384 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3386 put_task_struct(reaper
);
3394 static const struct pid_entry tid_base_stuff
[] = {
3395 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3396 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3397 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3398 REG("environ", S_IRUSR
, proc_environ_operations
),
3399 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3400 ONE("status", S_IRUGO
, proc_pid_status
),
3401 ONE("personality", S_IRUGO
, proc_pid_personality
),
3402 INF("limits", S_IRUGO
, proc_pid_limits
),
3403 #ifdef CONFIG_SCHED_DEBUG
3404 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3406 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3407 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3408 INF("syscall", S_IRUGO
, proc_pid_syscall
),
3410 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3411 ONE("stat", S_IRUGO
, proc_tid_stat
),
3412 ONE("statm", S_IRUGO
, proc_pid_statm
),
3413 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3414 #ifdef CONFIG_CHECKPOINT_RESTORE
3415 REG("children", S_IRUGO
, proc_tid_children_operations
),
3418 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3420 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3421 LNK("cwd", proc_cwd_link
),
3422 LNK("root", proc_root_link
),
3423 LNK("exe", proc_exe_link
),
3424 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3425 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3426 #ifdef CONFIG_PROC_PAGE_MONITOR
3427 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3428 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3429 REG("pagemap", S_IRUGO
, proc_pagemap_operations
),
3431 #ifdef CONFIG_SECURITY
3432 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3434 #ifdef CONFIG_KALLSYMS
3435 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3437 #ifdef CONFIG_STACKTRACE
3438 ONE("stack", S_IRUGO
, proc_pid_stack
),
3440 #ifdef CONFIG_SCHEDSTATS
3441 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3443 #ifdef CONFIG_LATENCYTOP
3444 REG("latency", S_IRUGO
, proc_lstats_operations
),
3446 #ifdef CONFIG_PROC_PID_CPUSET
3447 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3449 #ifdef CONFIG_CGROUPS
3450 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3452 INF("oom_score", S_IRUGO
, proc_oom_score
),
3453 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3454 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3455 #ifdef CONFIG_AUDITSYSCALL
3456 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3457 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3459 #ifdef CONFIG_FAULT_INJECTION
3460 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3462 #ifdef CONFIG_TASK_IO_ACCOUNTING
3463 INF("io", S_IRUSR
, proc_tid_io_accounting
),
3465 #ifdef CONFIG_HARDWALL
3466 INF("hardwall", S_IRUGO
, proc_pid_hardwall
),
3468 #ifdef CONFIG_USER_NS
3469 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3470 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3474 static int proc_tid_base_readdir(struct file
* filp
,
3475 void * dirent
, filldir_t filldir
)
3477 return proc_pident_readdir(filp
,dirent
,filldir
,
3478 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
3481 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3483 return proc_pident_lookup(dir
, dentry
,
3484 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3487 static const struct file_operations proc_tid_base_operations
= {
3488 .read
= generic_read_dir
,
3489 .readdir
= proc_tid_base_readdir
,
3490 .llseek
= default_llseek
,
3493 static const struct inode_operations proc_tid_base_inode_operations
= {
3494 .lookup
= proc_tid_base_lookup
,
3495 .getattr
= pid_getattr
,
3496 .setattr
= proc_setattr
,
3499 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
3500 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3502 struct dentry
*error
= ERR_PTR(-ENOENT
);
3503 struct inode
*inode
;
3504 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3508 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3509 inode
->i_op
= &proc_tid_base_inode_operations
;
3510 inode
->i_fop
= &proc_tid_base_operations
;
3511 inode
->i_flags
|=S_IMMUTABLE
;
3513 set_nlink(inode
, 2 + pid_entry_count_dirs(tid_base_stuff
,
3514 ARRAY_SIZE(tid_base_stuff
)));
3516 d_set_d_op(dentry
, &pid_dentry_operations
);
3518 d_add(dentry
, inode
);
3519 /* Close the race of the process dying before we return the dentry */
3520 if (pid_revalidate(dentry
, 0))
3526 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3528 struct dentry
*result
= ERR_PTR(-ENOENT
);
3529 struct task_struct
*task
;
3530 struct task_struct
*leader
= get_proc_task(dir
);
3532 struct pid_namespace
*ns
;
3537 tid
= name_to_int(dentry
);
3541 ns
= dentry
->d_sb
->s_fs_info
;
3543 task
= find_task_by_pid_ns(tid
, ns
);
3545 get_task_struct(task
);
3549 if (!same_thread_group(leader
, task
))
3552 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3554 put_task_struct(task
);
3556 put_task_struct(leader
);
3562 * Find the first tid of a thread group to return to user space.
3564 * Usually this is just the thread group leader, but if the users
3565 * buffer was too small or there was a seek into the middle of the
3566 * directory we have more work todo.
3568 * In the case of a short read we start with find_task_by_pid.
3570 * In the case of a seek we start with the leader and walk nr
3573 static struct task_struct
*first_tid(struct task_struct
*leader
,
3574 int tid
, int nr
, struct pid_namespace
*ns
)
3576 struct task_struct
*pos
;
3579 /* Attempt to start with the pid of a thread */
3580 if (tid
&& (nr
> 0)) {
3581 pos
= find_task_by_pid_ns(tid
, ns
);
3582 if (pos
&& (pos
->group_leader
== leader
))
3586 /* If nr exceeds the number of threads there is nothing todo */
3588 if (nr
&& nr
>= get_nr_threads(leader
))
3591 /* If we haven't found our starting place yet start
3592 * with the leader and walk nr threads forward.
3594 for (pos
= leader
; nr
> 0; --nr
) {
3595 pos
= next_thread(pos
);
3596 if (pos
== leader
) {
3602 get_task_struct(pos
);
3609 * Find the next thread in the thread list.
3610 * Return NULL if there is an error or no next thread.
3612 * The reference to the input task_struct is released.
3614 static struct task_struct
*next_tid(struct task_struct
*start
)
3616 struct task_struct
*pos
= NULL
;
3618 if (pid_alive(start
)) {
3619 pos
= next_thread(start
);
3620 if (thread_group_leader(pos
))
3623 get_task_struct(pos
);
3626 put_task_struct(start
);
3630 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3631 struct task_struct
*task
, int tid
)
3633 char name
[PROC_NUMBUF
];
3634 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3635 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3636 proc_task_instantiate
, task
, NULL
);
3639 /* for the /proc/TGID/task/ directories */
3640 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3642 struct dentry
*dentry
= filp
->f_path
.dentry
;
3643 struct inode
*inode
= dentry
->d_inode
;
3644 struct task_struct
*leader
= NULL
;
3645 struct task_struct
*task
;
3646 int retval
= -ENOENT
;
3649 struct pid_namespace
*ns
;
3651 task
= get_proc_task(inode
);
3655 if (pid_alive(task
)) {
3656 leader
= task
->group_leader
;
3657 get_task_struct(leader
);
3660 put_task_struct(task
);
3665 switch ((unsigned long)filp
->f_pos
) {
3668 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3673 ino
= parent_ino(dentry
);
3674 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3680 /* f_version caches the tgid value that the last readdir call couldn't
3681 * return. lseek aka telldir automagically resets f_version to 0.
3683 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3684 tid
= (int)filp
->f_version
;
3685 filp
->f_version
= 0;
3686 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3688 task
= next_tid(task
), filp
->f_pos
++) {
3689 tid
= task_pid_nr_ns(task
, ns
);
3690 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3691 /* returning this tgid failed, save it as the first
3692 * pid for the next readir call */
3693 filp
->f_version
= (u64
)tid
;
3694 put_task_struct(task
);
3699 put_task_struct(leader
);
3704 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3706 struct inode
*inode
= dentry
->d_inode
;
3707 struct task_struct
*p
= get_proc_task(inode
);
3708 generic_fillattr(inode
, stat
);
3711 stat
->nlink
+= get_nr_threads(p
);
3718 static const struct inode_operations proc_task_inode_operations
= {
3719 .lookup
= proc_task_lookup
,
3720 .getattr
= proc_task_getattr
,
3721 .setattr
= proc_setattr
,
3722 .permission
= proc_pid_permission
,
3725 static const struct file_operations proc_task_operations
= {
3726 .read
= generic_read_dir
,
3727 .readdir
= proc_task_readdir
,
3728 .llseek
= default_llseek
,