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 size_t this_len
, max_len
;
833 if (src
>= (mm
->env_end
- mm
->env_start
))
836 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
838 max_len
= min_t(size_t, PAGE_SIZE
, count
);
839 this_len
= min(max_len
, this_len
);
841 retval
= access_remote_vm(mm
, (mm
->env_start
+ src
),
849 if (copy_to_user(buf
, page
, retval
)) {
863 free_page((unsigned long) page
);
867 static const struct file_operations proc_environ_operations
= {
868 .open
= environ_open
,
869 .read
= environ_read
,
870 .llseek
= generic_file_llseek
,
871 .release
= mem_release
,
874 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
875 size_t count
, loff_t
*ppos
)
877 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
878 char buffer
[PROC_NUMBUF
];
880 int oom_adjust
= OOM_DISABLE
;
886 if (lock_task_sighand(task
, &flags
)) {
887 oom_adjust
= task
->signal
->oom_adj
;
888 unlock_task_sighand(task
, &flags
);
891 put_task_struct(task
);
893 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
895 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
898 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
899 size_t count
, loff_t
*ppos
)
901 struct task_struct
*task
;
902 char buffer
[PROC_NUMBUF
];
907 memset(buffer
, 0, sizeof(buffer
));
908 if (count
> sizeof(buffer
) - 1)
909 count
= sizeof(buffer
) - 1;
910 if (copy_from_user(buffer
, buf
, count
)) {
915 err
= kstrtoint(strstrip(buffer
), 0, &oom_adjust
);
918 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
919 oom_adjust
!= OOM_DISABLE
) {
924 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
936 if (!lock_task_sighand(task
, &flags
)) {
941 if (oom_adjust
< task
->signal
->oom_adj
&& !capable(CAP_SYS_RESOURCE
)) {
947 * Warn that /proc/pid/oom_adj is deprecated, see
948 * Documentation/feature-removal-schedule.txt.
950 printk_once(KERN_WARNING
"%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
951 current
->comm
, task_pid_nr(current
), task_pid_nr(task
),
953 task
->signal
->oom_adj
= oom_adjust
;
955 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
956 * value is always attainable.
958 if (task
->signal
->oom_adj
== OOM_ADJUST_MAX
)
959 task
->signal
->oom_score_adj
= OOM_SCORE_ADJ_MAX
;
961 task
->signal
->oom_score_adj
= (oom_adjust
* OOM_SCORE_ADJ_MAX
) /
963 trace_oom_score_adj_update(task
);
965 unlock_task_sighand(task
, &flags
);
968 put_task_struct(task
);
970 return err
< 0 ? err
: count
;
973 static const struct file_operations proc_oom_adjust_operations
= {
974 .read
= oom_adjust_read
,
975 .write
= oom_adjust_write
,
976 .llseek
= generic_file_llseek
,
979 static ssize_t
oom_score_adj_read(struct file
*file
, char __user
*buf
,
980 size_t count
, loff_t
*ppos
)
982 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
983 char buffer
[PROC_NUMBUF
];
984 int oom_score_adj
= OOM_SCORE_ADJ_MIN
;
990 if (lock_task_sighand(task
, &flags
)) {
991 oom_score_adj
= task
->signal
->oom_score_adj
;
992 unlock_task_sighand(task
, &flags
);
994 put_task_struct(task
);
995 len
= snprintf(buffer
, sizeof(buffer
), "%d\n", oom_score_adj
);
996 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
999 static ssize_t
oom_score_adj_write(struct file
*file
, const char __user
*buf
,
1000 size_t count
, loff_t
*ppos
)
1002 struct task_struct
*task
;
1003 char buffer
[PROC_NUMBUF
];
1004 unsigned long flags
;
1008 memset(buffer
, 0, sizeof(buffer
));
1009 if (count
> sizeof(buffer
) - 1)
1010 count
= sizeof(buffer
) - 1;
1011 if (copy_from_user(buffer
, buf
, count
)) {
1016 err
= kstrtoint(strstrip(buffer
), 0, &oom_score_adj
);
1019 if (oom_score_adj
< OOM_SCORE_ADJ_MIN
||
1020 oom_score_adj
> OOM_SCORE_ADJ_MAX
) {
1025 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
1037 if (!lock_task_sighand(task
, &flags
)) {
1042 if (oom_score_adj
< task
->signal
->oom_score_adj_min
&&
1043 !capable(CAP_SYS_RESOURCE
)) {
1048 task
->signal
->oom_score_adj
= oom_score_adj
;
1049 if (has_capability_noaudit(current
, CAP_SYS_RESOURCE
))
1050 task
->signal
->oom_score_adj_min
= oom_score_adj
;
1051 trace_oom_score_adj_update(task
);
1053 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1054 * always attainable.
1056 if (task
->signal
->oom_score_adj
== OOM_SCORE_ADJ_MIN
)
1057 task
->signal
->oom_adj
= OOM_DISABLE
;
1059 task
->signal
->oom_adj
= (oom_score_adj
* OOM_ADJUST_MAX
) /
1062 unlock_task_sighand(task
, &flags
);
1065 put_task_struct(task
);
1067 return err
< 0 ? err
: count
;
1070 static const struct file_operations proc_oom_score_adj_operations
= {
1071 .read
= oom_score_adj_read
,
1072 .write
= oom_score_adj_write
,
1073 .llseek
= default_llseek
,
1076 #ifdef CONFIG_AUDITSYSCALL
1077 #define TMPBUFLEN 21
1078 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
1079 size_t count
, loff_t
*ppos
)
1081 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1082 struct task_struct
*task
= get_proc_task(inode
);
1084 char tmpbuf
[TMPBUFLEN
];
1088 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1089 audit_get_loginuid(task
));
1090 put_task_struct(task
);
1091 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1094 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1095 size_t count
, loff_t
*ppos
)
1097 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1103 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
)) {
1109 if (count
>= PAGE_SIZE
)
1110 count
= PAGE_SIZE
- 1;
1113 /* No partial writes. */
1116 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1120 if (copy_from_user(page
, buf
, count
))
1124 loginuid
= simple_strtoul(page
, &tmp
, 10);
1130 length
= audit_set_loginuid(loginuid
);
1131 if (likely(length
== 0))
1135 free_page((unsigned long) page
);
1139 static const struct file_operations proc_loginuid_operations
= {
1140 .read
= proc_loginuid_read
,
1141 .write
= proc_loginuid_write
,
1142 .llseek
= generic_file_llseek
,
1145 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1146 size_t count
, loff_t
*ppos
)
1148 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1149 struct task_struct
*task
= get_proc_task(inode
);
1151 char tmpbuf
[TMPBUFLEN
];
1155 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1156 audit_get_sessionid(task
));
1157 put_task_struct(task
);
1158 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1161 static const struct file_operations proc_sessionid_operations
= {
1162 .read
= proc_sessionid_read
,
1163 .llseek
= generic_file_llseek
,
1167 #ifdef CONFIG_FAULT_INJECTION
1168 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1169 size_t count
, loff_t
*ppos
)
1171 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1172 char buffer
[PROC_NUMBUF
];
1178 make_it_fail
= task
->make_it_fail
;
1179 put_task_struct(task
);
1181 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1183 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1186 static ssize_t
proc_fault_inject_write(struct file
* file
,
1187 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1189 struct task_struct
*task
;
1190 char buffer
[PROC_NUMBUF
], *end
;
1193 if (!capable(CAP_SYS_RESOURCE
))
1195 memset(buffer
, 0, sizeof(buffer
));
1196 if (count
> sizeof(buffer
) - 1)
1197 count
= sizeof(buffer
) - 1;
1198 if (copy_from_user(buffer
, buf
, count
))
1200 make_it_fail
= simple_strtol(strstrip(buffer
), &end
, 0);
1203 task
= get_proc_task(file
->f_dentry
->d_inode
);
1206 task
->make_it_fail
= make_it_fail
;
1207 put_task_struct(task
);
1212 static const struct file_operations proc_fault_inject_operations
= {
1213 .read
= proc_fault_inject_read
,
1214 .write
= proc_fault_inject_write
,
1215 .llseek
= generic_file_llseek
,
1220 #ifdef CONFIG_SCHED_DEBUG
1222 * Print out various scheduling related per-task fields:
1224 static int sched_show(struct seq_file
*m
, void *v
)
1226 struct inode
*inode
= m
->private;
1227 struct task_struct
*p
;
1229 p
= get_proc_task(inode
);
1232 proc_sched_show_task(p
, m
);
1240 sched_write(struct file
*file
, const char __user
*buf
,
1241 size_t count
, loff_t
*offset
)
1243 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1244 struct task_struct
*p
;
1246 p
= get_proc_task(inode
);
1249 proc_sched_set_task(p
);
1256 static int sched_open(struct inode
*inode
, struct file
*filp
)
1258 return single_open(filp
, sched_show
, inode
);
1261 static const struct file_operations proc_pid_sched_operations
= {
1264 .write
= sched_write
,
1265 .llseek
= seq_lseek
,
1266 .release
= single_release
,
1271 #ifdef CONFIG_SCHED_AUTOGROUP
1273 * Print out autogroup related information:
1275 static int sched_autogroup_show(struct seq_file
*m
, void *v
)
1277 struct inode
*inode
= m
->private;
1278 struct task_struct
*p
;
1280 p
= get_proc_task(inode
);
1283 proc_sched_autogroup_show_task(p
, m
);
1291 sched_autogroup_write(struct file
*file
, const char __user
*buf
,
1292 size_t count
, loff_t
*offset
)
1294 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1295 struct task_struct
*p
;
1296 char buffer
[PROC_NUMBUF
];
1300 memset(buffer
, 0, sizeof(buffer
));
1301 if (count
> sizeof(buffer
) - 1)
1302 count
= sizeof(buffer
) - 1;
1303 if (copy_from_user(buffer
, buf
, count
))
1306 err
= kstrtoint(strstrip(buffer
), 0, &nice
);
1310 p
= get_proc_task(inode
);
1314 err
= proc_sched_autogroup_set_nice(p
, nice
);
1323 static int sched_autogroup_open(struct inode
*inode
, struct file
*filp
)
1327 ret
= single_open(filp
, sched_autogroup_show
, NULL
);
1329 struct seq_file
*m
= filp
->private_data
;
1336 static const struct file_operations proc_pid_sched_autogroup_operations
= {
1337 .open
= sched_autogroup_open
,
1339 .write
= sched_autogroup_write
,
1340 .llseek
= seq_lseek
,
1341 .release
= single_release
,
1344 #endif /* CONFIG_SCHED_AUTOGROUP */
1346 static ssize_t
comm_write(struct file
*file
, const char __user
*buf
,
1347 size_t count
, loff_t
*offset
)
1349 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1350 struct task_struct
*p
;
1351 char buffer
[TASK_COMM_LEN
];
1353 memset(buffer
, 0, sizeof(buffer
));
1354 if (count
> sizeof(buffer
) - 1)
1355 count
= sizeof(buffer
) - 1;
1356 if (copy_from_user(buffer
, buf
, count
))
1359 p
= get_proc_task(inode
);
1363 if (same_thread_group(current
, p
))
1364 set_task_comm(p
, buffer
);
1373 static int comm_show(struct seq_file
*m
, void *v
)
1375 struct inode
*inode
= m
->private;
1376 struct task_struct
*p
;
1378 p
= get_proc_task(inode
);
1383 seq_printf(m
, "%s\n", p
->comm
);
1391 static int comm_open(struct inode
*inode
, struct file
*filp
)
1393 return single_open(filp
, comm_show
, inode
);
1396 static const struct file_operations proc_pid_set_comm_operations
= {
1399 .write
= comm_write
,
1400 .llseek
= seq_lseek
,
1401 .release
= single_release
,
1404 static int proc_exe_link(struct dentry
*dentry
, struct path
*exe_path
)
1406 struct task_struct
*task
;
1407 struct mm_struct
*mm
;
1408 struct file
*exe_file
;
1410 task
= get_proc_task(dentry
->d_inode
);
1413 mm
= get_task_mm(task
);
1414 put_task_struct(task
);
1417 exe_file
= get_mm_exe_file(mm
);
1420 *exe_path
= exe_file
->f_path
;
1421 path_get(&exe_file
->f_path
);
1428 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1430 struct inode
*inode
= dentry
->d_inode
;
1432 int error
= -EACCES
;
1434 /* Are we allowed to snoop on the tasks file descriptors? */
1435 if (!proc_fd_access_allowed(inode
))
1438 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1442 nd_jump_link(nd
, &path
);
1445 return ERR_PTR(error
);
1448 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1450 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1457 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1458 len
= PTR_ERR(pathname
);
1459 if (IS_ERR(pathname
))
1461 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1465 if (copy_to_user(buffer
, pathname
, len
))
1468 free_page((unsigned long)tmp
);
1472 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1474 int error
= -EACCES
;
1475 struct inode
*inode
= dentry
->d_inode
;
1478 /* Are we allowed to snoop on the tasks file descriptors? */
1479 if (!proc_fd_access_allowed(inode
))
1482 error
= PROC_I(inode
)->op
.proc_get_link(dentry
, &path
);
1486 error
= do_proc_readlink(&path
, buffer
, buflen
);
1492 static const struct inode_operations proc_pid_link_inode_operations
= {
1493 .readlink
= proc_pid_readlink
,
1494 .follow_link
= proc_pid_follow_link
,
1495 .setattr
= proc_setattr
,
1499 /* building an inode */
1501 static int task_dumpable(struct task_struct
*task
)
1504 struct mm_struct
*mm
;
1509 dumpable
= get_dumpable(mm
);
1516 struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1518 struct inode
* inode
;
1519 struct proc_inode
*ei
;
1520 const struct cred
*cred
;
1522 /* We need a new inode */
1524 inode
= new_inode(sb
);
1530 inode
->i_ino
= get_next_ino();
1531 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1532 inode
->i_op
= &proc_def_inode_operations
;
1535 * grab the reference to task.
1537 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1541 if (task_dumpable(task
)) {
1543 cred
= __task_cred(task
);
1544 inode
->i_uid
= cred
->euid
;
1545 inode
->i_gid
= cred
->egid
;
1548 security_task_to_inode(task
, inode
);
1558 int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1560 struct inode
*inode
= dentry
->d_inode
;
1561 struct task_struct
*task
;
1562 const struct cred
*cred
;
1563 struct pid_namespace
*pid
= dentry
->d_sb
->s_fs_info
;
1565 generic_fillattr(inode
, stat
);
1568 stat
->uid
= GLOBAL_ROOT_UID
;
1569 stat
->gid
= GLOBAL_ROOT_GID
;
1570 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1572 if (!has_pid_permissions(pid
, task
, 2)) {
1575 * This doesn't prevent learning whether PID exists,
1576 * it only makes getattr() consistent with readdir().
1580 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1581 task_dumpable(task
)) {
1582 cred
= __task_cred(task
);
1583 stat
->uid
= cred
->euid
;
1584 stat
->gid
= cred
->egid
;
1594 * Exceptional case: normally we are not allowed to unhash a busy
1595 * directory. In this case, however, we can do it - no aliasing problems
1596 * due to the way we treat inodes.
1598 * Rewrite the inode's ownerships here because the owning task may have
1599 * performed a setuid(), etc.
1601 * Before the /proc/pid/status file was created the only way to read
1602 * the effective uid of a /process was to stat /proc/pid. Reading
1603 * /proc/pid/status is slow enough that procps and other packages
1604 * kept stating /proc/pid. To keep the rules in /proc simple I have
1605 * made this apply to all per process world readable and executable
1608 int pid_revalidate(struct dentry
*dentry
, unsigned int flags
)
1610 struct inode
*inode
;
1611 struct task_struct
*task
;
1612 const struct cred
*cred
;
1614 if (flags
& LOOKUP_RCU
)
1617 inode
= dentry
->d_inode
;
1618 task
= get_proc_task(inode
);
1621 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1622 task_dumpable(task
)) {
1624 cred
= __task_cred(task
);
1625 inode
->i_uid
= cred
->euid
;
1626 inode
->i_gid
= cred
->egid
;
1629 inode
->i_uid
= GLOBAL_ROOT_UID
;
1630 inode
->i_gid
= GLOBAL_ROOT_GID
;
1632 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1633 security_task_to_inode(task
, inode
);
1634 put_task_struct(task
);
1641 static int pid_delete_dentry(const struct dentry
* dentry
)
1643 /* Is the task we represent dead?
1644 * If so, then don't put the dentry on the lru list,
1645 * kill it immediately.
1647 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1650 const struct dentry_operations pid_dentry_operations
=
1652 .d_revalidate
= pid_revalidate
,
1653 .d_delete
= pid_delete_dentry
,
1659 * Fill a directory entry.
1661 * If possible create the dcache entry and derive our inode number and
1662 * file type from dcache entry.
1664 * Since all of the proc inode numbers are dynamically generated, the inode
1665 * numbers do not exist until the inode is cache. This means creating the
1666 * the dcache entry in readdir is necessary to keep the inode numbers
1667 * reported by readdir in sync with the inode numbers reported
1670 int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1671 const char *name
, int len
,
1672 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1674 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1675 struct inode
*inode
;
1678 unsigned type
= DT_UNKNOWN
;
1682 qname
.hash
= full_name_hash(name
, len
);
1684 child
= d_lookup(dir
, &qname
);
1687 new = d_alloc(dir
, &qname
);
1689 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1696 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1697 goto end_instantiate
;
1698 inode
= child
->d_inode
;
1701 type
= inode
->i_mode
>> 12;
1706 ino
= find_inode_number(dir
, &qname
);
1709 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1712 static unsigned name_to_int(struct dentry
*dentry
)
1714 const char *name
= dentry
->d_name
.name
;
1715 int len
= dentry
->d_name
.len
;
1718 if (len
> 1 && *name
== '0')
1721 unsigned c
= *name
++ - '0';
1724 if (n
>= (~0U-9)/10)
1734 #define PROC_FDINFO_MAX 64
1736 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1738 struct task_struct
*task
= get_proc_task(inode
);
1739 struct files_struct
*files
= NULL
;
1741 int fd
= proc_fd(inode
);
1744 files
= get_files_struct(task
);
1745 put_task_struct(task
);
1749 * We are not taking a ref to the file structure, so we must
1752 spin_lock(&files
->file_lock
);
1753 file
= fcheck_files(files
, fd
);
1755 unsigned int f_flags
;
1756 struct fdtable
*fdt
;
1758 fdt
= files_fdtable(files
);
1759 f_flags
= file
->f_flags
& ~O_CLOEXEC
;
1760 if (close_on_exec(fd
, fdt
))
1761 f_flags
|= O_CLOEXEC
;
1764 *path
= file
->f_path
;
1765 path_get(&file
->f_path
);
1768 snprintf(info
, PROC_FDINFO_MAX
,
1771 (long long) file
->f_pos
,
1773 spin_unlock(&files
->file_lock
);
1774 put_files_struct(files
);
1777 spin_unlock(&files
->file_lock
);
1778 put_files_struct(files
);
1783 static int proc_fd_link(struct dentry
*dentry
, struct path
*path
)
1785 return proc_fd_info(dentry
->d_inode
, path
, NULL
);
1788 static int tid_fd_revalidate(struct dentry
*dentry
, unsigned int flags
)
1790 struct inode
*inode
;
1791 struct task_struct
*task
;
1793 struct files_struct
*files
;
1794 const struct cred
*cred
;
1796 if (flags
& LOOKUP_RCU
)
1799 inode
= dentry
->d_inode
;
1800 task
= get_proc_task(inode
);
1801 fd
= proc_fd(inode
);
1804 files
= get_files_struct(task
);
1808 file
= fcheck_files(files
, fd
);
1810 unsigned f_mode
= file
->f_mode
;
1813 put_files_struct(files
);
1815 if (task_dumpable(task
)) {
1817 cred
= __task_cred(task
);
1818 inode
->i_uid
= cred
->euid
;
1819 inode
->i_gid
= cred
->egid
;
1822 inode
->i_uid
= GLOBAL_ROOT_UID
;
1823 inode
->i_gid
= GLOBAL_ROOT_GID
;
1826 if (S_ISLNK(inode
->i_mode
)) {
1827 unsigned i_mode
= S_IFLNK
;
1828 if (f_mode
& FMODE_READ
)
1829 i_mode
|= S_IRUSR
| S_IXUSR
;
1830 if (f_mode
& FMODE_WRITE
)
1831 i_mode
|= S_IWUSR
| S_IXUSR
;
1832 inode
->i_mode
= i_mode
;
1835 security_task_to_inode(task
, inode
);
1836 put_task_struct(task
);
1840 put_files_struct(files
);
1842 put_task_struct(task
);
1848 static const struct dentry_operations tid_fd_dentry_operations
=
1850 .d_revalidate
= tid_fd_revalidate
,
1851 .d_delete
= pid_delete_dentry
,
1854 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1855 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1857 unsigned fd
= (unsigned long)ptr
;
1858 struct inode
*inode
;
1859 struct proc_inode
*ei
;
1860 struct dentry
*error
= ERR_PTR(-ENOENT
);
1862 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1868 inode
->i_mode
= S_IFLNK
;
1869 inode
->i_op
= &proc_pid_link_inode_operations
;
1871 ei
->op
.proc_get_link
= proc_fd_link
;
1872 d_set_d_op(dentry
, &tid_fd_dentry_operations
);
1873 d_add(dentry
, inode
);
1874 /* Close the race of the process dying before we return the dentry */
1875 if (tid_fd_revalidate(dentry
, 0))
1882 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1883 struct dentry
*dentry
,
1884 instantiate_t instantiate
)
1886 struct task_struct
*task
= get_proc_task(dir
);
1887 unsigned fd
= name_to_int(dentry
);
1888 struct dentry
*result
= ERR_PTR(-ENOENT
);
1895 result
= instantiate(dir
, dentry
, task
, (void *)(unsigned long)fd
);
1897 put_task_struct(task
);
1902 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1903 filldir_t filldir
, instantiate_t instantiate
)
1905 struct dentry
*dentry
= filp
->f_path
.dentry
;
1906 struct inode
*inode
= dentry
->d_inode
;
1907 struct task_struct
*p
= get_proc_task(inode
);
1908 unsigned int fd
, ino
;
1910 struct files_struct
* files
;
1920 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1924 ino
= parent_ino(dentry
);
1925 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1929 files
= get_files_struct(p
);
1933 for (fd
= filp
->f_pos
-2;
1934 fd
< files_fdtable(files
)->max_fds
;
1935 fd
++, filp
->f_pos
++) {
1936 char name
[PROC_NUMBUF
];
1940 if (!fcheck_files(files
, fd
))
1944 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1945 rv
= proc_fill_cache(filp
, dirent
, filldir
,
1946 name
, len
, instantiate
, p
,
1947 (void *)(unsigned long)fd
);
1954 put_files_struct(files
);
1962 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1965 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1968 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1970 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1973 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1974 size_t len
, loff_t
*ppos
)
1976 char tmp
[PROC_FDINFO_MAX
];
1977 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1979 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1983 static const struct file_operations proc_fdinfo_file_operations
= {
1984 .open
= nonseekable_open
,
1985 .read
= proc_fdinfo_read
,
1986 .llseek
= no_llseek
,
1989 static const struct file_operations proc_fd_operations
= {
1990 .read
= generic_read_dir
,
1991 .readdir
= proc_readfd
,
1992 .llseek
= default_llseek
,
1995 #ifdef CONFIG_CHECKPOINT_RESTORE
1998 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1999 * which represent vma start and end addresses.
2001 static int dname_to_vma_addr(struct dentry
*dentry
,
2002 unsigned long *start
, unsigned long *end
)
2004 if (sscanf(dentry
->d_name
.name
, "%lx-%lx", start
, end
) != 2)
2010 static int map_files_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
2012 unsigned long vm_start
, vm_end
;
2013 bool exact_vma_exists
= false;
2014 struct mm_struct
*mm
= NULL
;
2015 struct task_struct
*task
;
2016 const struct cred
*cred
;
2017 struct inode
*inode
;
2020 if (flags
& LOOKUP_RCU
)
2023 if (!capable(CAP_SYS_ADMIN
)) {
2028 inode
= dentry
->d_inode
;
2029 task
= get_proc_task(inode
);
2033 mm
= mm_access(task
, PTRACE_MODE_READ
);
2034 if (IS_ERR_OR_NULL(mm
))
2037 if (!dname_to_vma_addr(dentry
, &vm_start
, &vm_end
)) {
2038 down_read(&mm
->mmap_sem
);
2039 exact_vma_exists
= !!find_exact_vma(mm
, vm_start
, vm_end
);
2040 up_read(&mm
->mmap_sem
);
2045 if (exact_vma_exists
) {
2046 if (task_dumpable(task
)) {
2048 cred
= __task_cred(task
);
2049 inode
->i_uid
= cred
->euid
;
2050 inode
->i_gid
= cred
->egid
;
2053 inode
->i_uid
= GLOBAL_ROOT_UID
;
2054 inode
->i_gid
= GLOBAL_ROOT_GID
;
2056 security_task_to_inode(task
, inode
);
2061 put_task_struct(task
);
2070 static const struct dentry_operations tid_map_files_dentry_operations
= {
2071 .d_revalidate
= map_files_d_revalidate
,
2072 .d_delete
= pid_delete_dentry
,
2075 static int proc_map_files_get_link(struct dentry
*dentry
, struct path
*path
)
2077 unsigned long vm_start
, vm_end
;
2078 struct vm_area_struct
*vma
;
2079 struct task_struct
*task
;
2080 struct mm_struct
*mm
;
2084 task
= get_proc_task(dentry
->d_inode
);
2088 mm
= get_task_mm(task
);
2089 put_task_struct(task
);
2093 rc
= dname_to_vma_addr(dentry
, &vm_start
, &vm_end
);
2097 down_read(&mm
->mmap_sem
);
2098 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2099 if (vma
&& vma
->vm_file
) {
2100 *path
= vma
->vm_file
->f_path
;
2104 up_read(&mm
->mmap_sem
);
2112 struct map_files_info
{
2115 unsigned char name
[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2118 static struct dentry
*
2119 proc_map_files_instantiate(struct inode
*dir
, struct dentry
*dentry
,
2120 struct task_struct
*task
, const void *ptr
)
2122 const struct file
*file
= ptr
;
2123 struct proc_inode
*ei
;
2124 struct inode
*inode
;
2127 return ERR_PTR(-ENOENT
);
2129 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2131 return ERR_PTR(-ENOENT
);
2134 ei
->op
.proc_get_link
= proc_map_files_get_link
;
2136 inode
->i_op
= &proc_pid_link_inode_operations
;
2138 inode
->i_mode
= S_IFLNK
;
2140 if (file
->f_mode
& FMODE_READ
)
2141 inode
->i_mode
|= S_IRUSR
;
2142 if (file
->f_mode
& FMODE_WRITE
)
2143 inode
->i_mode
|= S_IWUSR
;
2145 d_set_d_op(dentry
, &tid_map_files_dentry_operations
);
2146 d_add(dentry
, inode
);
2151 static struct dentry
*proc_map_files_lookup(struct inode
*dir
,
2152 struct dentry
*dentry
, unsigned int flags
)
2154 unsigned long vm_start
, vm_end
;
2155 struct vm_area_struct
*vma
;
2156 struct task_struct
*task
;
2157 struct dentry
*result
;
2158 struct mm_struct
*mm
;
2160 result
= ERR_PTR(-EACCES
);
2161 if (!capable(CAP_SYS_ADMIN
))
2164 result
= ERR_PTR(-ENOENT
);
2165 task
= get_proc_task(dir
);
2169 result
= ERR_PTR(-EACCES
);
2170 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
2173 result
= ERR_PTR(-ENOENT
);
2174 if (dname_to_vma_addr(dentry
, &vm_start
, &vm_end
))
2177 mm
= get_task_mm(task
);
2181 down_read(&mm
->mmap_sem
);
2182 vma
= find_exact_vma(mm
, vm_start
, vm_end
);
2186 result
= proc_map_files_instantiate(dir
, dentry
, task
, vma
->vm_file
);
2189 up_read(&mm
->mmap_sem
);
2192 put_task_struct(task
);
2197 static const struct inode_operations proc_map_files_inode_operations
= {
2198 .lookup
= proc_map_files_lookup
,
2199 .permission
= proc_fd_permission
,
2200 .setattr
= proc_setattr
,
2204 proc_map_files_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
2206 struct dentry
*dentry
= filp
->f_path
.dentry
;
2207 struct inode
*inode
= dentry
->d_inode
;
2208 struct vm_area_struct
*vma
;
2209 struct task_struct
*task
;
2210 struct mm_struct
*mm
;
2215 if (!capable(CAP_SYS_ADMIN
))
2219 task
= get_proc_task(inode
);
2224 if (!ptrace_may_access(task
, PTRACE_MODE_READ
))
2228 switch (filp
->f_pos
) {
2231 if (filldir(dirent
, ".", 1, 0, ino
, DT_DIR
) < 0)
2235 ino
= parent_ino(dentry
);
2236 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
2241 unsigned long nr_files
, pos
, i
;
2242 struct flex_array
*fa
= NULL
;
2243 struct map_files_info info
;
2244 struct map_files_info
*p
;
2246 mm
= get_task_mm(task
);
2249 down_read(&mm
->mmap_sem
);
2254 * We need two passes here:
2256 * 1) Collect vmas of mapped files with mmap_sem taken
2257 * 2) Release mmap_sem and instantiate entries
2259 * otherwise we get lockdep complained, since filldir()
2260 * routine might require mmap_sem taken in might_fault().
2263 for (vma
= mm
->mmap
, pos
= 2; vma
; vma
= vma
->vm_next
) {
2264 if (vma
->vm_file
&& ++pos
> filp
->f_pos
)
2269 fa
= flex_array_alloc(sizeof(info
), nr_files
,
2271 if (!fa
|| flex_array_prealloc(fa
, 0, nr_files
,
2275 flex_array_free(fa
);
2276 up_read(&mm
->mmap_sem
);
2280 for (i
= 0, vma
= mm
->mmap
, pos
= 2; vma
;
2281 vma
= vma
->vm_next
) {
2284 if (++pos
<= filp
->f_pos
)
2287 get_file(vma
->vm_file
);
2288 info
.file
= vma
->vm_file
;
2289 info
.len
= snprintf(info
.name
,
2290 sizeof(info
.name
), "%lx-%lx",
2291 vma
->vm_start
, vma
->vm_end
);
2292 if (flex_array_put(fa
, i
++, &info
, GFP_KERNEL
))
2296 up_read(&mm
->mmap_sem
);
2298 for (i
= 0; i
< nr_files
; i
++) {
2299 p
= flex_array_get(fa
, i
);
2300 ret
= proc_fill_cache(filp
, dirent
, filldir
,
2302 proc_map_files_instantiate
,
2309 for (; i
< nr_files
; i
++) {
2311 * In case of error don't forget
2312 * to put rest of file refs.
2314 p
= flex_array_get(fa
, i
);
2318 flex_array_free(fa
);
2324 put_task_struct(task
);
2329 static const struct file_operations proc_map_files_operations
= {
2330 .read
= generic_read_dir
,
2331 .readdir
= proc_map_files_readdir
,
2332 .llseek
= default_llseek
,
2335 #endif /* CONFIG_CHECKPOINT_RESTORE */
2338 * /proc/pid/fd needs a special permission handler so that a process can still
2339 * access /proc/self/fd after it has executed a setuid().
2341 static int proc_fd_permission(struct inode
*inode
, int mask
)
2343 int rv
= generic_permission(inode
, mask
);
2346 if (task_pid(current
) == proc_pid(inode
))
2352 * proc directories can do almost nothing..
2354 static const struct inode_operations proc_fd_inode_operations
= {
2355 .lookup
= proc_lookupfd
,
2356 .permission
= proc_fd_permission
,
2357 .setattr
= proc_setattr
,
2360 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
2361 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2363 unsigned fd
= (unsigned long)ptr
;
2364 struct inode
*inode
;
2365 struct proc_inode
*ei
;
2366 struct dentry
*error
= ERR_PTR(-ENOENT
);
2368 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2373 inode
->i_mode
= S_IFREG
| S_IRUSR
;
2374 inode
->i_fop
= &proc_fdinfo_file_operations
;
2375 d_set_d_op(dentry
, &tid_fd_dentry_operations
);
2376 d_add(dentry
, inode
);
2377 /* Close the race of the process dying before we return the dentry */
2378 if (tid_fd_revalidate(dentry
, 0))
2385 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
2386 struct dentry
*dentry
,
2389 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
2392 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
2394 return proc_readfd_common(filp
, dirent
, filldir
,
2395 proc_fdinfo_instantiate
);
2398 static const struct file_operations proc_fdinfo_operations
= {
2399 .read
= generic_read_dir
,
2400 .readdir
= proc_readfdinfo
,
2401 .llseek
= default_llseek
,
2405 * proc directories can do almost nothing..
2407 static const struct inode_operations proc_fdinfo_inode_operations
= {
2408 .lookup
= proc_lookupfdinfo
,
2409 .setattr
= proc_setattr
,
2413 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
2414 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2416 const struct pid_entry
*p
= ptr
;
2417 struct inode
*inode
;
2418 struct proc_inode
*ei
;
2419 struct dentry
*error
= ERR_PTR(-ENOENT
);
2421 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2426 inode
->i_mode
= p
->mode
;
2427 if (S_ISDIR(inode
->i_mode
))
2428 set_nlink(inode
, 2); /* Use getattr to fix if necessary */
2430 inode
->i_op
= p
->iop
;
2432 inode
->i_fop
= p
->fop
;
2434 d_set_d_op(dentry
, &pid_dentry_operations
);
2435 d_add(dentry
, inode
);
2436 /* Close the race of the process dying before we return the dentry */
2437 if (pid_revalidate(dentry
, 0))
2443 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
2444 struct dentry
*dentry
,
2445 const struct pid_entry
*ents
,
2448 struct dentry
*error
;
2449 struct task_struct
*task
= get_proc_task(dir
);
2450 const struct pid_entry
*p
, *last
;
2452 error
= ERR_PTR(-ENOENT
);
2458 * Yes, it does not scale. And it should not. Don't add
2459 * new entries into /proc/<tgid>/ without very good reasons.
2461 last
= &ents
[nents
- 1];
2462 for (p
= ents
; p
<= last
; p
++) {
2463 if (p
->len
!= dentry
->d_name
.len
)
2465 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2471 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
2473 put_task_struct(task
);
2478 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
2479 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2481 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2482 proc_pident_instantiate
, task
, p
);
2485 static int proc_pident_readdir(struct file
*filp
,
2486 void *dirent
, filldir_t filldir
,
2487 const struct pid_entry
*ents
, unsigned int nents
)
2490 struct dentry
*dentry
= filp
->f_path
.dentry
;
2491 struct inode
*inode
= dentry
->d_inode
;
2492 struct task_struct
*task
= get_proc_task(inode
);
2493 const struct pid_entry
*p
, *last
;
2506 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
2512 ino
= parent_ino(dentry
);
2513 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
2525 last
= &ents
[nents
- 1];
2527 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
2536 put_task_struct(task
);
2541 #ifdef CONFIG_SECURITY
2542 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
2543 size_t count
, loff_t
*ppos
)
2545 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2548 struct task_struct
*task
= get_proc_task(inode
);
2553 length
= security_getprocattr(task
,
2554 (char*)file
->f_path
.dentry
->d_name
.name
,
2556 put_task_struct(task
);
2558 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
2563 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
2564 size_t count
, loff_t
*ppos
)
2566 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
2569 struct task_struct
*task
= get_proc_task(inode
);
2574 if (count
> PAGE_SIZE
)
2577 /* No partial writes. */
2583 page
= (char*)__get_free_page(GFP_TEMPORARY
);
2588 if (copy_from_user(page
, buf
, count
))
2591 /* Guard against adverse ptrace interaction */
2592 length
= mutex_lock_interruptible(&task
->signal
->cred_guard_mutex
);
2596 length
= security_setprocattr(task
,
2597 (char*)file
->f_path
.dentry
->d_name
.name
,
2598 (void*)page
, count
);
2599 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2601 free_page((unsigned long) page
);
2603 put_task_struct(task
);
2608 static const struct file_operations proc_pid_attr_operations
= {
2609 .read
= proc_pid_attr_read
,
2610 .write
= proc_pid_attr_write
,
2611 .llseek
= generic_file_llseek
,
2614 static const struct pid_entry attr_dir_stuff
[] = {
2615 REG("current", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2616 REG("prev", S_IRUGO
, proc_pid_attr_operations
),
2617 REG("exec", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2618 REG("fscreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2619 REG("keycreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2620 REG("sockcreate", S_IRUGO
|S_IWUGO
, proc_pid_attr_operations
),
2623 static int proc_attr_dir_readdir(struct file
* filp
,
2624 void * dirent
, filldir_t filldir
)
2626 return proc_pident_readdir(filp
,dirent
,filldir
,
2627 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2630 static const struct file_operations proc_attr_dir_operations
= {
2631 .read
= generic_read_dir
,
2632 .readdir
= proc_attr_dir_readdir
,
2633 .llseek
= default_llseek
,
2636 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2637 struct dentry
*dentry
, unsigned int flags
)
2639 return proc_pident_lookup(dir
, dentry
,
2640 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2643 static const struct inode_operations proc_attr_dir_inode_operations
= {
2644 .lookup
= proc_attr_dir_lookup
,
2645 .getattr
= pid_getattr
,
2646 .setattr
= proc_setattr
,
2651 #ifdef CONFIG_ELF_CORE
2652 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2653 size_t count
, loff_t
*ppos
)
2655 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2656 struct mm_struct
*mm
;
2657 char buffer
[PROC_NUMBUF
];
2665 mm
= get_task_mm(task
);
2667 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2668 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2669 MMF_DUMP_FILTER_SHIFT
));
2671 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2674 put_task_struct(task
);
2679 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2680 const char __user
*buf
,
2684 struct task_struct
*task
;
2685 struct mm_struct
*mm
;
2686 char buffer
[PROC_NUMBUF
], *end
;
2693 memset(buffer
, 0, sizeof(buffer
));
2694 if (count
> sizeof(buffer
) - 1)
2695 count
= sizeof(buffer
) - 1;
2696 if (copy_from_user(buffer
, buf
, count
))
2700 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2703 if (end
- buffer
== 0)
2707 task
= get_proc_task(file
->f_dentry
->d_inode
);
2712 mm
= get_task_mm(task
);
2716 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2718 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2720 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2725 put_task_struct(task
);
2730 static const struct file_operations proc_coredump_filter_operations
= {
2731 .read
= proc_coredump_filter_read
,
2732 .write
= proc_coredump_filter_write
,
2733 .llseek
= generic_file_llseek
,
2740 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2743 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2744 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2745 char tmp
[PROC_NUMBUF
];
2748 sprintf(tmp
, "%d", tgid
);
2749 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2752 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2754 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2755 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2756 char *name
= ERR_PTR(-ENOENT
);
2760 name
= ERR_PTR(-ENOMEM
);
2762 sprintf(name
, "%d", tgid
);
2764 nd_set_link(nd
, name
);
2768 static void proc_self_put_link(struct dentry
*dentry
, struct nameidata
*nd
,
2771 char *s
= nd_get_link(nd
);
2776 static const struct inode_operations proc_self_inode_operations
= {
2777 .readlink
= proc_self_readlink
,
2778 .follow_link
= proc_self_follow_link
,
2779 .put_link
= proc_self_put_link
,
2785 * These are the directory entries in the root directory of /proc
2786 * that properly belong to the /proc filesystem, as they describe
2787 * describe something that is process related.
2789 static const struct pid_entry proc_base_stuff
[] = {
2790 NOD("self", S_IFLNK
|S_IRWXUGO
,
2791 &proc_self_inode_operations
, NULL
, {}),
2794 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2795 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2797 const struct pid_entry
*p
= ptr
;
2798 struct inode
*inode
;
2799 struct proc_inode
*ei
;
2800 struct dentry
*error
;
2802 /* Allocate the inode */
2803 error
= ERR_PTR(-ENOMEM
);
2804 inode
= new_inode(dir
->i_sb
);
2808 /* Initialize the inode */
2810 inode
->i_ino
= get_next_ino();
2811 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2814 * grab the reference to the task.
2816 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2820 inode
->i_mode
= p
->mode
;
2821 if (S_ISDIR(inode
->i_mode
))
2822 set_nlink(inode
, 2);
2823 if (S_ISLNK(inode
->i_mode
))
2826 inode
->i_op
= p
->iop
;
2828 inode
->i_fop
= p
->fop
;
2830 d_add(dentry
, inode
);
2839 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2841 struct dentry
*error
;
2842 struct task_struct
*task
= get_proc_task(dir
);
2843 const struct pid_entry
*p
, *last
;
2845 error
= ERR_PTR(-ENOENT
);
2850 /* Lookup the directory entry */
2851 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2852 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2853 if (p
->len
!= dentry
->d_name
.len
)
2855 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2861 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2864 put_task_struct(task
);
2869 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2870 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2872 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2873 proc_base_instantiate
, task
, p
);
2876 #ifdef CONFIG_TASK_IO_ACCOUNTING
2877 static int do_io_accounting(struct task_struct
*task
, char *buffer
, int whole
)
2879 struct task_io_accounting acct
= task
->ioac
;
2880 unsigned long flags
;
2883 result
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
2887 if (!ptrace_may_access(task
, PTRACE_MODE_READ
)) {
2892 if (whole
&& lock_task_sighand(task
, &flags
)) {
2893 struct task_struct
*t
= task
;
2895 task_io_accounting_add(&acct
, &task
->signal
->ioac
);
2896 while_each_thread(task
, t
)
2897 task_io_accounting_add(&acct
, &t
->ioac
);
2899 unlock_task_sighand(task
, &flags
);
2901 result
= sprintf(buffer
,
2906 "read_bytes: %llu\n"
2907 "write_bytes: %llu\n"
2908 "cancelled_write_bytes: %llu\n",
2909 (unsigned long long)acct
.rchar
,
2910 (unsigned long long)acct
.wchar
,
2911 (unsigned long long)acct
.syscr
,
2912 (unsigned long long)acct
.syscw
,
2913 (unsigned long long)acct
.read_bytes
,
2914 (unsigned long long)acct
.write_bytes
,
2915 (unsigned long long)acct
.cancelled_write_bytes
);
2917 mutex_unlock(&task
->signal
->cred_guard_mutex
);
2921 static int proc_tid_io_accounting(struct task_struct
*task
, char *buffer
)
2923 return do_io_accounting(task
, buffer
, 0);
2926 static int proc_tgid_io_accounting(struct task_struct
*task
, char *buffer
)
2928 return do_io_accounting(task
, buffer
, 1);
2930 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2932 #ifdef CONFIG_USER_NS
2933 static int proc_id_map_open(struct inode
*inode
, struct file
*file
,
2934 struct seq_operations
*seq_ops
)
2936 struct user_namespace
*ns
= NULL
;
2937 struct task_struct
*task
;
2938 struct seq_file
*seq
;
2941 task
= get_proc_task(inode
);
2944 ns
= get_user_ns(task_cred_xxx(task
, user_ns
));
2946 put_task_struct(task
);
2951 ret
= seq_open(file
, seq_ops
);
2955 seq
= file
->private_data
;
2965 static int proc_id_map_release(struct inode
*inode
, struct file
*file
)
2967 struct seq_file
*seq
= file
->private_data
;
2968 struct user_namespace
*ns
= seq
->private;
2970 return seq_release(inode
, file
);
2973 static int proc_uid_map_open(struct inode
*inode
, struct file
*file
)
2975 return proc_id_map_open(inode
, file
, &proc_uid_seq_operations
);
2978 static int proc_gid_map_open(struct inode
*inode
, struct file
*file
)
2980 return proc_id_map_open(inode
, file
, &proc_gid_seq_operations
);
2983 static const struct file_operations proc_uid_map_operations
= {
2984 .open
= proc_uid_map_open
,
2985 .write
= proc_uid_map_write
,
2987 .llseek
= seq_lseek
,
2988 .release
= proc_id_map_release
,
2991 static const struct file_operations proc_gid_map_operations
= {
2992 .open
= proc_gid_map_open
,
2993 .write
= proc_gid_map_write
,
2995 .llseek
= seq_lseek
,
2996 .release
= proc_id_map_release
,
2998 #endif /* CONFIG_USER_NS */
3000 static int proc_pid_personality(struct seq_file
*m
, struct pid_namespace
*ns
,
3001 struct pid
*pid
, struct task_struct
*task
)
3003 int err
= lock_trace(task
);
3005 seq_printf(m
, "%08x\n", task
->personality
);
3014 static const struct file_operations proc_task_operations
;
3015 static const struct inode_operations proc_task_inode_operations
;
3017 static const struct pid_entry tgid_base_stuff
[] = {
3018 DIR("task", S_IRUGO
|S_IXUGO
, proc_task_inode_operations
, proc_task_operations
),
3019 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3020 #ifdef CONFIG_CHECKPOINT_RESTORE
3021 DIR("map_files", S_IRUSR
|S_IXUSR
, proc_map_files_inode_operations
, proc_map_files_operations
),
3023 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3024 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3026 DIR("net", S_IRUGO
|S_IXUGO
, proc_net_inode_operations
, proc_net_operations
),
3028 REG("environ", S_IRUSR
, proc_environ_operations
),
3029 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3030 ONE("status", S_IRUGO
, proc_pid_status
),
3031 ONE("personality", S_IRUGO
, proc_pid_personality
),
3032 INF("limits", S_IRUGO
, proc_pid_limits
),
3033 #ifdef CONFIG_SCHED_DEBUG
3034 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3036 #ifdef CONFIG_SCHED_AUTOGROUP
3037 REG("autogroup", S_IRUGO
|S_IWUSR
, proc_pid_sched_autogroup_operations
),
3039 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3040 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3041 INF("syscall", S_IRUGO
, proc_pid_syscall
),
3043 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3044 ONE("stat", S_IRUGO
, proc_tgid_stat
),
3045 ONE("statm", S_IRUGO
, proc_pid_statm
),
3046 REG("maps", S_IRUGO
, proc_pid_maps_operations
),
3048 REG("numa_maps", S_IRUGO
, proc_pid_numa_maps_operations
),
3050 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3051 LNK("cwd", proc_cwd_link
),
3052 LNK("root", proc_root_link
),
3053 LNK("exe", proc_exe_link
),
3054 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3055 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3056 REG("mountstats", S_IRUSR
, proc_mountstats_operations
),
3057 #ifdef CONFIG_PROC_PAGE_MONITOR
3058 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3059 REG("smaps", S_IRUGO
, proc_pid_smaps_operations
),
3060 REG("pagemap", S_IRUGO
, proc_pagemap_operations
),
3062 #ifdef CONFIG_SECURITY
3063 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3065 #ifdef CONFIG_KALLSYMS
3066 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3068 #ifdef CONFIG_STACKTRACE
3069 ONE("stack", S_IRUGO
, proc_pid_stack
),
3071 #ifdef CONFIG_SCHEDSTATS
3072 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3074 #ifdef CONFIG_LATENCYTOP
3075 REG("latency", S_IRUGO
, proc_lstats_operations
),
3077 #ifdef CONFIG_PROC_PID_CPUSET
3078 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3080 #ifdef CONFIG_CGROUPS
3081 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3083 INF("oom_score", S_IRUGO
, proc_oom_score
),
3084 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3085 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3086 #ifdef CONFIG_AUDITSYSCALL
3087 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3088 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3090 #ifdef CONFIG_FAULT_INJECTION
3091 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3093 #ifdef CONFIG_ELF_CORE
3094 REG("coredump_filter", S_IRUGO
|S_IWUSR
, proc_coredump_filter_operations
),
3096 #ifdef CONFIG_TASK_IO_ACCOUNTING
3097 INF("io", S_IRUSR
, proc_tgid_io_accounting
),
3099 #ifdef CONFIG_HARDWALL
3100 INF("hardwall", S_IRUGO
, proc_pid_hardwall
),
3102 #ifdef CONFIG_USER_NS
3103 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3104 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3108 static int proc_tgid_base_readdir(struct file
* filp
,
3109 void * dirent
, filldir_t filldir
)
3111 return proc_pident_readdir(filp
,dirent
,filldir
,
3112 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
3115 static const struct file_operations proc_tgid_base_operations
= {
3116 .read
= generic_read_dir
,
3117 .readdir
= proc_tgid_base_readdir
,
3118 .llseek
= default_llseek
,
3121 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3123 return proc_pident_lookup(dir
, dentry
,
3124 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
3127 static const struct inode_operations proc_tgid_base_inode_operations
= {
3128 .lookup
= proc_tgid_base_lookup
,
3129 .getattr
= pid_getattr
,
3130 .setattr
= proc_setattr
,
3131 .permission
= proc_pid_permission
,
3134 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
3136 struct dentry
*dentry
, *leader
, *dir
;
3137 char buf
[PROC_NUMBUF
];
3141 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3142 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3144 shrink_dcache_parent(dentry
);
3150 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
3151 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
3156 name
.len
= strlen(name
.name
);
3157 dir
= d_hash_and_lookup(leader
, &name
);
3159 goto out_put_leader
;
3162 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
3163 dentry
= d_hash_and_lookup(dir
, &name
);
3165 shrink_dcache_parent(dentry
);
3178 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3179 * @task: task that should be flushed.
3181 * When flushing dentries from proc, one needs to flush them from global
3182 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3183 * in. This call is supposed to do all of this job.
3185 * Looks in the dcache for
3187 * /proc/@tgid/task/@pid
3188 * if either directory is present flushes it and all of it'ts children
3191 * It is safe and reasonable to cache /proc entries for a task until
3192 * that task exits. After that they just clog up the dcache with
3193 * useless entries, possibly causing useful dcache entries to be
3194 * flushed instead. This routine is proved to flush those useless
3195 * dcache entries at process exit time.
3197 * NOTE: This routine is just an optimization so it does not guarantee
3198 * that no dcache entries will exist at process exit time it
3199 * just makes it very unlikely that any will persist.
3202 void proc_flush_task(struct task_struct
*task
)
3205 struct pid
*pid
, *tgid
;
3208 pid
= task_pid(task
);
3209 tgid
= task_tgid(task
);
3211 for (i
= 0; i
<= pid
->level
; i
++) {
3212 upid
= &pid
->numbers
[i
];
3213 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
3214 tgid
->numbers
[i
].nr
);
3217 upid
= &pid
->numbers
[pid
->level
];
3219 pid_ns_release_proc(upid
->ns
);
3222 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
3223 struct dentry
* dentry
,
3224 struct task_struct
*task
, const void *ptr
)
3226 struct dentry
*error
= ERR_PTR(-ENOENT
);
3227 struct inode
*inode
;
3229 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3233 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3234 inode
->i_op
= &proc_tgid_base_inode_operations
;
3235 inode
->i_fop
= &proc_tgid_base_operations
;
3236 inode
->i_flags
|=S_IMMUTABLE
;
3238 set_nlink(inode
, 2 + pid_entry_count_dirs(tgid_base_stuff
,
3239 ARRAY_SIZE(tgid_base_stuff
)));
3241 d_set_d_op(dentry
, &pid_dentry_operations
);
3243 d_add(dentry
, inode
);
3244 /* Close the race of the process dying before we return the dentry */
3245 if (pid_revalidate(dentry
, 0))
3251 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3253 struct dentry
*result
;
3254 struct task_struct
*task
;
3256 struct pid_namespace
*ns
;
3258 result
= proc_base_lookup(dir
, dentry
);
3259 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
3262 tgid
= name_to_int(dentry
);
3266 ns
= dentry
->d_sb
->s_fs_info
;
3268 task
= find_task_by_pid_ns(tgid
, ns
);
3270 get_task_struct(task
);
3275 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
3276 put_task_struct(task
);
3282 * Find the first task with tgid >= tgid
3287 struct task_struct
*task
;
3289 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
3294 put_task_struct(iter
.task
);
3298 pid
= find_ge_pid(iter
.tgid
, ns
);
3300 iter
.tgid
= pid_nr_ns(pid
, ns
);
3301 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
3302 /* What we to know is if the pid we have find is the
3303 * pid of a thread_group_leader. Testing for task
3304 * being a thread_group_leader is the obvious thing
3305 * todo but there is a window when it fails, due to
3306 * the pid transfer logic in de_thread.
3308 * So we perform the straight forward test of seeing
3309 * if the pid we have found is the pid of a thread
3310 * group leader, and don't worry if the task we have
3311 * found doesn't happen to be a thread group leader.
3312 * As we don't care in the case of readdir.
3314 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
3318 get_task_struct(iter
.task
);
3324 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3326 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3327 struct tgid_iter iter
)
3329 char name
[PROC_NUMBUF
];
3330 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
3331 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3332 proc_pid_instantiate
, iter
.task
, NULL
);
3335 static int fake_filldir(void *buf
, const char *name
, int namelen
,
3336 loff_t offset
, u64 ino
, unsigned d_type
)
3341 /* for the /proc/ directory itself, after non-process stuff has been done */
3342 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3345 struct task_struct
*reaper
;
3346 struct tgid_iter iter
;
3347 struct pid_namespace
*ns
;
3348 filldir_t __filldir
;
3350 if (filp
->f_pos
>= PID_MAX_LIMIT
+ TGID_OFFSET
)
3352 nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
3354 reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
3358 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
3359 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
3360 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
3364 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3366 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
3367 for (iter
= next_tgid(ns
, iter
);
3369 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
3370 if (has_pid_permissions(ns
, iter
.task
, 2))
3371 __filldir
= filldir
;
3373 __filldir
= fake_filldir
;
3375 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
3376 if (proc_pid_fill_cache(filp
, dirent
, __filldir
, iter
) < 0) {
3377 put_task_struct(iter
.task
);
3381 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
3383 put_task_struct(reaper
);
3391 static const struct pid_entry tid_base_stuff
[] = {
3392 DIR("fd", S_IRUSR
|S_IXUSR
, proc_fd_inode_operations
, proc_fd_operations
),
3393 DIR("fdinfo", S_IRUSR
|S_IXUSR
, proc_fdinfo_inode_operations
, proc_fdinfo_operations
),
3394 DIR("ns", S_IRUSR
|S_IXUGO
, proc_ns_dir_inode_operations
, proc_ns_dir_operations
),
3395 REG("environ", S_IRUSR
, proc_environ_operations
),
3396 INF("auxv", S_IRUSR
, proc_pid_auxv
),
3397 ONE("status", S_IRUGO
, proc_pid_status
),
3398 ONE("personality", S_IRUGO
, proc_pid_personality
),
3399 INF("limits", S_IRUGO
, proc_pid_limits
),
3400 #ifdef CONFIG_SCHED_DEBUG
3401 REG("sched", S_IRUGO
|S_IWUSR
, proc_pid_sched_operations
),
3403 REG("comm", S_IRUGO
|S_IWUSR
, proc_pid_set_comm_operations
),
3404 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3405 INF("syscall", S_IRUGO
, proc_pid_syscall
),
3407 INF("cmdline", S_IRUGO
, proc_pid_cmdline
),
3408 ONE("stat", S_IRUGO
, proc_tid_stat
),
3409 ONE("statm", S_IRUGO
, proc_pid_statm
),
3410 REG("maps", S_IRUGO
, proc_tid_maps_operations
),
3411 #ifdef CONFIG_CHECKPOINT_RESTORE
3412 REG("children", S_IRUGO
, proc_tid_children_operations
),
3415 REG("numa_maps", S_IRUGO
, proc_tid_numa_maps_operations
),
3417 REG("mem", S_IRUSR
|S_IWUSR
, proc_mem_operations
),
3418 LNK("cwd", proc_cwd_link
),
3419 LNK("root", proc_root_link
),
3420 LNK("exe", proc_exe_link
),
3421 REG("mounts", S_IRUGO
, proc_mounts_operations
),
3422 REG("mountinfo", S_IRUGO
, proc_mountinfo_operations
),
3423 #ifdef CONFIG_PROC_PAGE_MONITOR
3424 REG("clear_refs", S_IWUSR
, proc_clear_refs_operations
),
3425 REG("smaps", S_IRUGO
, proc_tid_smaps_operations
),
3426 REG("pagemap", S_IRUGO
, proc_pagemap_operations
),
3428 #ifdef CONFIG_SECURITY
3429 DIR("attr", S_IRUGO
|S_IXUGO
, proc_attr_dir_inode_operations
, proc_attr_dir_operations
),
3431 #ifdef CONFIG_KALLSYMS
3432 INF("wchan", S_IRUGO
, proc_pid_wchan
),
3434 #ifdef CONFIG_STACKTRACE
3435 ONE("stack", S_IRUGO
, proc_pid_stack
),
3437 #ifdef CONFIG_SCHEDSTATS
3438 INF("schedstat", S_IRUGO
, proc_pid_schedstat
),
3440 #ifdef CONFIG_LATENCYTOP
3441 REG("latency", S_IRUGO
, proc_lstats_operations
),
3443 #ifdef CONFIG_PROC_PID_CPUSET
3444 REG("cpuset", S_IRUGO
, proc_cpuset_operations
),
3446 #ifdef CONFIG_CGROUPS
3447 REG("cgroup", S_IRUGO
, proc_cgroup_operations
),
3449 INF("oom_score", S_IRUGO
, proc_oom_score
),
3450 REG("oom_adj", S_IRUGO
|S_IWUSR
, proc_oom_adjust_operations
),
3451 REG("oom_score_adj", S_IRUGO
|S_IWUSR
, proc_oom_score_adj_operations
),
3452 #ifdef CONFIG_AUDITSYSCALL
3453 REG("loginuid", S_IWUSR
|S_IRUGO
, proc_loginuid_operations
),
3454 REG("sessionid", S_IRUGO
, proc_sessionid_operations
),
3456 #ifdef CONFIG_FAULT_INJECTION
3457 REG("make-it-fail", S_IRUGO
|S_IWUSR
, proc_fault_inject_operations
),
3459 #ifdef CONFIG_TASK_IO_ACCOUNTING
3460 INF("io", S_IRUSR
, proc_tid_io_accounting
),
3462 #ifdef CONFIG_HARDWALL
3463 INF("hardwall", S_IRUGO
, proc_pid_hardwall
),
3465 #ifdef CONFIG_USER_NS
3466 REG("uid_map", S_IRUGO
|S_IWUSR
, proc_uid_map_operations
),
3467 REG("gid_map", S_IRUGO
|S_IWUSR
, proc_gid_map_operations
),
3471 static int proc_tid_base_readdir(struct file
* filp
,
3472 void * dirent
, filldir_t filldir
)
3474 return proc_pident_readdir(filp
,dirent
,filldir
,
3475 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
3478 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
3480 return proc_pident_lookup(dir
, dentry
,
3481 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
3484 static const struct file_operations proc_tid_base_operations
= {
3485 .read
= generic_read_dir
,
3486 .readdir
= proc_tid_base_readdir
,
3487 .llseek
= default_llseek
,
3490 static const struct inode_operations proc_tid_base_inode_operations
= {
3491 .lookup
= proc_tid_base_lookup
,
3492 .getattr
= pid_getattr
,
3493 .setattr
= proc_setattr
,
3496 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
3497 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
3499 struct dentry
*error
= ERR_PTR(-ENOENT
);
3500 struct inode
*inode
;
3501 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
3505 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
3506 inode
->i_op
= &proc_tid_base_inode_operations
;
3507 inode
->i_fop
= &proc_tid_base_operations
;
3508 inode
->i_flags
|=S_IMMUTABLE
;
3510 set_nlink(inode
, 2 + pid_entry_count_dirs(tid_base_stuff
,
3511 ARRAY_SIZE(tid_base_stuff
)));
3513 d_set_d_op(dentry
, &pid_dentry_operations
);
3515 d_add(dentry
, inode
);
3516 /* Close the race of the process dying before we return the dentry */
3517 if (pid_revalidate(dentry
, 0))
3523 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, unsigned int flags
)
3525 struct dentry
*result
= ERR_PTR(-ENOENT
);
3526 struct task_struct
*task
;
3527 struct task_struct
*leader
= get_proc_task(dir
);
3529 struct pid_namespace
*ns
;
3534 tid
= name_to_int(dentry
);
3538 ns
= dentry
->d_sb
->s_fs_info
;
3540 task
= find_task_by_pid_ns(tid
, ns
);
3542 get_task_struct(task
);
3546 if (!same_thread_group(leader
, task
))
3549 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
3551 put_task_struct(task
);
3553 put_task_struct(leader
);
3559 * Find the first tid of a thread group to return to user space.
3561 * Usually this is just the thread group leader, but if the users
3562 * buffer was too small or there was a seek into the middle of the
3563 * directory we have more work todo.
3565 * In the case of a short read we start with find_task_by_pid.
3567 * In the case of a seek we start with the leader and walk nr
3570 static struct task_struct
*first_tid(struct task_struct
*leader
,
3571 int tid
, int nr
, struct pid_namespace
*ns
)
3573 struct task_struct
*pos
;
3576 /* Attempt to start with the pid of a thread */
3577 if (tid
&& (nr
> 0)) {
3578 pos
= find_task_by_pid_ns(tid
, ns
);
3579 if (pos
&& (pos
->group_leader
== leader
))
3583 /* If nr exceeds the number of threads there is nothing todo */
3585 if (nr
&& nr
>= get_nr_threads(leader
))
3588 /* If we haven't found our starting place yet start
3589 * with the leader and walk nr threads forward.
3591 for (pos
= leader
; nr
> 0; --nr
) {
3592 pos
= next_thread(pos
);
3593 if (pos
== leader
) {
3599 get_task_struct(pos
);
3606 * Find the next thread in the thread list.
3607 * Return NULL if there is an error or no next thread.
3609 * The reference to the input task_struct is released.
3611 static struct task_struct
*next_tid(struct task_struct
*start
)
3613 struct task_struct
*pos
= NULL
;
3615 if (pid_alive(start
)) {
3616 pos
= next_thread(start
);
3617 if (thread_group_leader(pos
))
3620 get_task_struct(pos
);
3623 put_task_struct(start
);
3627 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
3628 struct task_struct
*task
, int tid
)
3630 char name
[PROC_NUMBUF
];
3631 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
3632 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
3633 proc_task_instantiate
, task
, NULL
);
3636 /* for the /proc/TGID/task/ directories */
3637 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
3639 struct dentry
*dentry
= filp
->f_path
.dentry
;
3640 struct inode
*inode
= dentry
->d_inode
;
3641 struct task_struct
*leader
= NULL
;
3642 struct task_struct
*task
;
3643 int retval
= -ENOENT
;
3646 struct pid_namespace
*ns
;
3648 task
= get_proc_task(inode
);
3652 if (pid_alive(task
)) {
3653 leader
= task
->group_leader
;
3654 get_task_struct(leader
);
3657 put_task_struct(task
);
3662 switch ((unsigned long)filp
->f_pos
) {
3665 if (filldir(dirent
, ".", 1, filp
->f_pos
, ino
, DT_DIR
) < 0)
3670 ino
= parent_ino(dentry
);
3671 if (filldir(dirent
, "..", 2, filp
->f_pos
, ino
, DT_DIR
) < 0)
3677 /* f_version caches the tgid value that the last readdir call couldn't
3678 * return. lseek aka telldir automagically resets f_version to 0.
3680 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
3681 tid
= (int)filp
->f_version
;
3682 filp
->f_version
= 0;
3683 for (task
= first_tid(leader
, tid
, filp
->f_pos
- 2, ns
);
3685 task
= next_tid(task
), filp
->f_pos
++) {
3686 tid
= task_pid_nr_ns(task
, ns
);
3687 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
3688 /* returning this tgid failed, save it as the first
3689 * pid for the next readir call */
3690 filp
->f_version
= (u64
)tid
;
3691 put_task_struct(task
);
3696 put_task_struct(leader
);
3701 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
3703 struct inode
*inode
= dentry
->d_inode
;
3704 struct task_struct
*p
= get_proc_task(inode
);
3705 generic_fillattr(inode
, stat
);
3708 stat
->nlink
+= get_nr_threads(p
);
3715 static const struct inode_operations proc_task_inode_operations
= {
3716 .lookup
= proc_task_lookup
,
3717 .getattr
= proc_task_getattr
,
3718 .setattr
= proc_setattr
,
3719 .permission
= proc_pid_permission
,
3722 static const struct file_operations proc_task_operations
= {
3723 .read
= generic_read_dir
,
3724 .readdir
= proc_task_readdir
,
3725 .llseek
= default_llseek
,