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/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/mnt_namespace.h>
64 #include <linux/rcupdate.h>
65 #include <linux/kallsyms.h>
66 #include <linux/resource.h>
67 #include <linux/module.h>
68 #include <linux/mount.h>
69 #include <linux/security.h>
70 #include <linux/ptrace.h>
71 #include <linux/cgroup.h>
72 #include <linux/cpuset.h>
73 #include <linux/audit.h>
74 #include <linux/poll.h>
75 #include <linux/nsproxy.h>
76 #include <linux/oom.h>
77 #include <linux/elf.h>
78 #include <linux/pid_namespace.h>
82 * Implementing inode permission operations in /proc is almost
83 * certainly an error. Permission checks need to happen during
84 * each system call not at open time. The reason is that most of
85 * what we wish to check for permissions in /proc varies at runtime.
87 * The classic example of a problem is opening file descriptors
88 * in /proc for a task before it execs a suid executable.
95 const struct inode_operations
*iop
;
96 const struct file_operations
*fop
;
100 #define NOD(NAME, MODE, IOP, FOP, OP) { \
102 .len = sizeof(NAME) - 1, \
109 #define DIR(NAME, MODE, OTYPE) \
110 NOD(NAME, (S_IFDIR|(MODE)), \
111 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
113 #define LNK(NAME, OTYPE) \
114 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
115 &proc_pid_link_inode_operations, NULL, \
116 { .proc_get_link = &proc_##OTYPE##_link } )
117 #define REG(NAME, MODE, OTYPE) \
118 NOD(NAME, (S_IFREG|(MODE)), NULL, \
119 &proc_##OTYPE##_operations, {})
120 #define INF(NAME, MODE, OTYPE) \
121 NOD(NAME, (S_IFREG|(MODE)), \
122 NULL, &proc_info_file_operations, \
123 { .proc_read = &proc_##OTYPE } )
124 #define ONE(NAME, MODE, OTYPE) \
125 NOD(NAME, (S_IFREG|(MODE)), \
126 NULL, &proc_single_file_operations, \
127 { .proc_show = &proc_##OTYPE } )
130 EXPORT_SYMBOL(maps_protect
);
132 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
134 struct fs_struct
*fs
;
138 atomic_inc(&fs
->count
);
143 static int get_nr_threads(struct task_struct
*tsk
)
145 /* Must be called with the rcu_read_lock held */
149 if (lock_task_sighand(tsk
, &flags
)) {
150 count
= atomic_read(&tsk
->signal
->count
);
151 unlock_task_sighand(tsk
, &flags
);
156 static int proc_cwd_link(struct inode
*inode
, struct path
*path
)
158 struct task_struct
*task
= get_proc_task(inode
);
159 struct fs_struct
*fs
= NULL
;
160 int result
= -ENOENT
;
163 fs
= get_fs_struct(task
);
164 put_task_struct(task
);
167 read_lock(&fs
->lock
);
170 read_unlock(&fs
->lock
);
177 static int proc_root_link(struct inode
*inode
, struct path
*path
)
179 struct task_struct
*task
= get_proc_task(inode
);
180 struct fs_struct
*fs
= NULL
;
181 int result
= -ENOENT
;
184 fs
= get_fs_struct(task
);
185 put_task_struct(task
);
188 read_lock(&fs
->lock
);
191 read_unlock(&fs
->lock
);
198 #define MAY_PTRACE(task) \
199 (task == current || \
200 (task->parent == current && \
201 (task->ptrace & PT_PTRACED) && \
202 (task_is_stopped_or_traced(task)) && \
203 security_ptrace(current,task) == 0))
205 struct mm_struct
*mm_for_maps(struct task_struct
*task
)
207 struct mm_struct
*mm
= get_task_mm(task
);
210 down_read(&mm
->mmap_sem
);
214 if (task
->mm
!= current
->mm
&& __ptrace_may_attach(task
) < 0)
220 up_read(&mm
->mmap_sem
);
225 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
229 struct mm_struct
*mm
= get_task_mm(task
);
233 goto out_mm
; /* Shh! No looking before we're done */
235 len
= mm
->arg_end
- mm
->arg_start
;
240 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
242 // If the nul at the end of args has been overwritten, then
243 // assume application is using setproctitle(3).
244 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
245 len
= strnlen(buffer
, res
);
249 len
= mm
->env_end
- mm
->env_start
;
250 if (len
> PAGE_SIZE
- res
)
251 len
= PAGE_SIZE
- res
;
252 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
253 res
= strnlen(buffer
, res
);
262 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
265 struct mm_struct
*mm
= get_task_mm(task
);
267 unsigned int nwords
= 0;
270 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
271 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
274 memcpy(buffer
, mm
->saved_auxv
, res
);
281 #ifdef CONFIG_KALLSYMS
283 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
284 * Returns the resolved symbol. If that fails, simply return the address.
286 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
289 char symname
[KSYM_NAME_LEN
];
291 wchan
= get_wchan(task
);
293 if (lookup_symbol_name(wchan
, symname
) < 0)
294 return sprintf(buffer
, "%lu", wchan
);
296 return sprintf(buffer
, "%s", symname
);
298 #endif /* CONFIG_KALLSYMS */
300 #ifdef CONFIG_SCHEDSTATS
302 * Provides /proc/PID/schedstat
304 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
306 return sprintf(buffer
, "%llu %llu %lu\n",
307 task
->sched_info
.cpu_time
,
308 task
->sched_info
.run_delay
,
309 task
->sched_info
.pcount
);
313 #ifdef CONFIG_LATENCYTOP
314 static int lstats_show_proc(struct seq_file
*m
, void *v
)
317 struct inode
*inode
= m
->private;
318 struct task_struct
*task
= get_proc_task(inode
);
322 seq_puts(m
, "Latency Top version : v0.1\n");
323 for (i
= 0; i
< 32; i
++) {
324 if (task
->latency_record
[i
].backtrace
[0]) {
326 seq_printf(m
, "%i %li %li ",
327 task
->latency_record
[i
].count
,
328 task
->latency_record
[i
].time
,
329 task
->latency_record
[i
].max
);
330 for (q
= 0; q
< LT_BACKTRACEDEPTH
; q
++) {
331 char sym
[KSYM_NAME_LEN
];
333 if (!task
->latency_record
[i
].backtrace
[q
])
335 if (task
->latency_record
[i
].backtrace
[q
] == ULONG_MAX
)
337 sprint_symbol(sym
, task
->latency_record
[i
].backtrace
[q
]);
338 c
= strchr(sym
, '+');
341 seq_printf(m
, "%s ", sym
);
347 put_task_struct(task
);
351 static int lstats_open(struct inode
*inode
, struct file
*file
)
353 return single_open(file
, lstats_show_proc
, inode
);
356 static ssize_t
lstats_write(struct file
*file
, const char __user
*buf
,
357 size_t count
, loff_t
*offs
)
359 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
363 clear_all_latency_tracing(task
);
364 put_task_struct(task
);
369 static const struct file_operations proc_lstats_operations
= {
372 .write
= lstats_write
,
374 .release
= single_release
,
379 /* The badness from the OOM killer */
380 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
381 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
383 unsigned long points
;
384 struct timespec uptime
;
386 do_posix_clock_monotonic_gettime(&uptime
);
387 read_lock(&tasklist_lock
);
388 points
= badness(task
, uptime
.tv_sec
);
389 read_unlock(&tasklist_lock
);
390 return sprintf(buffer
, "%lu\n", points
);
398 static const struct limit_names lnames
[RLIM_NLIMITS
] = {
399 [RLIMIT_CPU
] = {"Max cpu time", "ms"},
400 [RLIMIT_FSIZE
] = {"Max file size", "bytes"},
401 [RLIMIT_DATA
] = {"Max data size", "bytes"},
402 [RLIMIT_STACK
] = {"Max stack size", "bytes"},
403 [RLIMIT_CORE
] = {"Max core file size", "bytes"},
404 [RLIMIT_RSS
] = {"Max resident set", "bytes"},
405 [RLIMIT_NPROC
] = {"Max processes", "processes"},
406 [RLIMIT_NOFILE
] = {"Max open files", "files"},
407 [RLIMIT_MEMLOCK
] = {"Max locked memory", "bytes"},
408 [RLIMIT_AS
] = {"Max address space", "bytes"},
409 [RLIMIT_LOCKS
] = {"Max file locks", "locks"},
410 [RLIMIT_SIGPENDING
] = {"Max pending signals", "signals"},
411 [RLIMIT_MSGQUEUE
] = {"Max msgqueue size", "bytes"},
412 [RLIMIT_NICE
] = {"Max nice priority", NULL
},
413 [RLIMIT_RTPRIO
] = {"Max realtime priority", NULL
},
414 [RLIMIT_RTTIME
] = {"Max realtime timeout", "us"},
417 /* Display limits for a process */
418 static int proc_pid_limits(struct task_struct
*task
, char *buffer
)
423 char *bufptr
= buffer
;
425 struct rlimit rlim
[RLIM_NLIMITS
];
428 if (!lock_task_sighand(task
,&flags
)) {
432 memcpy(rlim
, task
->signal
->rlim
, sizeof(struct rlimit
) * RLIM_NLIMITS
);
433 unlock_task_sighand(task
, &flags
);
437 * print the file header
439 count
+= sprintf(&bufptr
[count
], "%-25s %-20s %-20s %-10s\n",
440 "Limit", "Soft Limit", "Hard Limit", "Units");
442 for (i
= 0; i
< RLIM_NLIMITS
; i
++) {
443 if (rlim
[i
].rlim_cur
== RLIM_INFINITY
)
444 count
+= sprintf(&bufptr
[count
], "%-25s %-20s ",
445 lnames
[i
].name
, "unlimited");
447 count
+= sprintf(&bufptr
[count
], "%-25s %-20lu ",
448 lnames
[i
].name
, rlim
[i
].rlim_cur
);
450 if (rlim
[i
].rlim_max
== RLIM_INFINITY
)
451 count
+= sprintf(&bufptr
[count
], "%-20s ", "unlimited");
453 count
+= sprintf(&bufptr
[count
], "%-20lu ",
457 count
+= sprintf(&bufptr
[count
], "%-10s\n",
460 count
+= sprintf(&bufptr
[count
], "\n");
466 /************************************************************************/
467 /* Here the fs part begins */
468 /************************************************************************/
470 /* permission checks */
471 static int proc_fd_access_allowed(struct inode
*inode
)
473 struct task_struct
*task
;
475 /* Allow access to a task's file descriptors if it is us or we
476 * may use ptrace attach to the process and find out that
479 task
= get_proc_task(inode
);
481 allowed
= ptrace_may_attach(task
);
482 put_task_struct(task
);
487 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
490 struct inode
*inode
= dentry
->d_inode
;
492 if (attr
->ia_valid
& ATTR_MODE
)
495 error
= inode_change_ok(inode
, attr
);
497 error
= inode_setattr(inode
, attr
);
501 static const struct inode_operations proc_def_inode_operations
= {
502 .setattr
= proc_setattr
,
505 static int mounts_open_common(struct inode
*inode
, struct file
*file
,
506 const struct seq_operations
*op
)
508 struct task_struct
*task
= get_proc_task(inode
);
510 struct mnt_namespace
*ns
= NULL
;
511 struct fs_struct
*fs
= NULL
;
513 struct proc_mounts
*p
;
518 nsp
= task_nsproxy(task
);
526 fs
= get_fs_struct(task
);
527 put_task_struct(task
);
535 read_lock(&fs
->lock
);
538 read_unlock(&fs
->lock
);
542 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
546 file
->private_data
= &p
->m
;
547 ret
= seq_open(file
, op
);
554 p
->event
= ns
->event
;
568 static int mounts_release(struct inode
*inode
, struct file
*file
)
570 struct proc_mounts
*p
= file
->private_data
;
573 return seq_release(inode
, file
);
576 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
578 struct proc_mounts
*p
= file
->private_data
;
579 struct mnt_namespace
*ns
= p
->ns
;
582 poll_wait(file
, &ns
->poll
, wait
);
584 spin_lock(&vfsmount_lock
);
585 if (p
->event
!= ns
->event
) {
586 p
->event
= ns
->event
;
589 spin_unlock(&vfsmount_lock
);
594 static int mounts_open(struct inode
*inode
, struct file
*file
)
596 return mounts_open_common(inode
, file
, &mounts_op
);
599 static const struct file_operations proc_mounts_operations
= {
603 .release
= mounts_release
,
607 static int mountinfo_open(struct inode
*inode
, struct file
*file
)
609 return mounts_open_common(inode
, file
, &mountinfo_op
);
612 static const struct file_operations proc_mountinfo_operations
= {
613 .open
= mountinfo_open
,
616 .release
= mounts_release
,
620 static int mountstats_open(struct inode
*inode
, struct file
*file
)
622 return mounts_open_common(inode
, file
, &mountstats_op
);
625 static const struct file_operations proc_mountstats_operations
= {
626 .open
= mountstats_open
,
629 .release
= mounts_release
,
632 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
634 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
635 size_t count
, loff_t
*ppos
)
637 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
640 struct task_struct
*task
= get_proc_task(inode
);
646 if (count
> PROC_BLOCK_SIZE
)
647 count
= PROC_BLOCK_SIZE
;
650 if (!(page
= __get_free_page(GFP_TEMPORARY
)))
653 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
656 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
659 put_task_struct(task
);
664 static const struct file_operations proc_info_file_operations
= {
665 .read
= proc_info_read
,
668 static int proc_single_show(struct seq_file
*m
, void *v
)
670 struct inode
*inode
= m
->private;
671 struct pid_namespace
*ns
;
673 struct task_struct
*task
;
676 ns
= inode
->i_sb
->s_fs_info
;
677 pid
= proc_pid(inode
);
678 task
= get_pid_task(pid
, PIDTYPE_PID
);
682 ret
= PROC_I(inode
)->op
.proc_show(m
, ns
, pid
, task
);
684 put_task_struct(task
);
688 static int proc_single_open(struct inode
*inode
, struct file
*filp
)
691 ret
= single_open(filp
, proc_single_show
, NULL
);
693 struct seq_file
*m
= filp
->private_data
;
700 static const struct file_operations proc_single_file_operations
= {
701 .open
= proc_single_open
,
704 .release
= single_release
,
707 static int mem_open(struct inode
* inode
, struct file
* file
)
709 file
->private_data
= (void*)((long)current
->self_exec_id
);
713 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
714 size_t count
, loff_t
*ppos
)
716 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
718 unsigned long src
= *ppos
;
720 struct mm_struct
*mm
;
725 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
729 page
= (char *)__get_free_page(GFP_TEMPORARY
);
735 mm
= get_task_mm(task
);
741 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
747 int this_len
, retval
;
749 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
750 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
751 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
757 if (copy_to_user(buf
, page
, retval
)) {
772 free_page((unsigned long) page
);
774 put_task_struct(task
);
779 #define mem_write NULL
782 /* This is a security hazard */
783 static ssize_t
mem_write(struct file
* file
, const char __user
*buf
,
784 size_t count
, loff_t
*ppos
)
788 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
789 unsigned long dst
= *ppos
;
795 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
799 page
= (char *)__get_free_page(GFP_TEMPORARY
);
805 int this_len
, retval
;
807 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
808 if (copy_from_user(page
, buf
, this_len
)) {
812 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
824 free_page((unsigned long) page
);
826 put_task_struct(task
);
832 loff_t
mem_lseek(struct file
*file
, loff_t offset
, int orig
)
836 file
->f_pos
= offset
;
839 file
->f_pos
+= offset
;
844 force_successful_syscall_return();
848 static const struct file_operations proc_mem_operations
= {
855 static ssize_t
environ_read(struct file
*file
, char __user
*buf
,
856 size_t count
, loff_t
*ppos
)
858 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
860 unsigned long src
= *ppos
;
862 struct mm_struct
*mm
;
867 if (!ptrace_may_attach(task
))
871 page
= (char *)__get_free_page(GFP_TEMPORARY
);
877 mm
= get_task_mm(task
);
882 int this_len
, retval
, max_len
;
884 this_len
= mm
->env_end
- (mm
->env_start
+ src
);
889 max_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
890 this_len
= (this_len
> max_len
) ? max_len
: this_len
;
892 retval
= access_process_vm(task
, (mm
->env_start
+ src
),
900 if (copy_to_user(buf
, page
, retval
)) {
914 free_page((unsigned long) page
);
916 put_task_struct(task
);
921 static const struct file_operations proc_environ_operations
= {
922 .read
= environ_read
,
925 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
926 size_t count
, loff_t
*ppos
)
928 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
929 char buffer
[PROC_NUMBUF
];
935 oom_adjust
= task
->oomkilladj
;
936 put_task_struct(task
);
938 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
940 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
943 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
944 size_t count
, loff_t
*ppos
)
946 struct task_struct
*task
;
947 char buffer
[PROC_NUMBUF
], *end
;
950 memset(buffer
, 0, sizeof(buffer
));
951 if (count
> sizeof(buffer
) - 1)
952 count
= sizeof(buffer
) - 1;
953 if (copy_from_user(buffer
, buf
, count
))
955 oom_adjust
= simple_strtol(buffer
, &end
, 0);
956 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
957 oom_adjust
!= OOM_DISABLE
)
961 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
964 if (oom_adjust
< task
->oomkilladj
&& !capable(CAP_SYS_RESOURCE
)) {
965 put_task_struct(task
);
968 task
->oomkilladj
= oom_adjust
;
969 put_task_struct(task
);
970 if (end
- buffer
== 0)
975 static const struct file_operations proc_oom_adjust_operations
= {
976 .read
= oom_adjust_read
,
977 .write
= oom_adjust_write
,
980 #ifdef CONFIG_AUDITSYSCALL
982 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
983 size_t count
, loff_t
*ppos
)
985 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
986 struct task_struct
*task
= get_proc_task(inode
);
988 char tmpbuf
[TMPBUFLEN
];
992 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
993 audit_get_loginuid(task
));
994 put_task_struct(task
);
995 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
998 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
999 size_t count
, loff_t
*ppos
)
1001 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1006 if (!capable(CAP_AUDIT_CONTROL
))
1009 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
1012 if (count
>= PAGE_SIZE
)
1013 count
= PAGE_SIZE
- 1;
1016 /* No partial writes. */
1019 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1023 if (copy_from_user(page
, buf
, count
))
1027 loginuid
= simple_strtoul(page
, &tmp
, 10);
1033 length
= audit_set_loginuid(current
, loginuid
);
1034 if (likely(length
== 0))
1038 free_page((unsigned long) page
);
1042 static const struct file_operations proc_loginuid_operations
= {
1043 .read
= proc_loginuid_read
,
1044 .write
= proc_loginuid_write
,
1047 static ssize_t
proc_sessionid_read(struct file
* file
, char __user
* buf
,
1048 size_t count
, loff_t
*ppos
)
1050 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1051 struct task_struct
*task
= get_proc_task(inode
);
1053 char tmpbuf
[TMPBUFLEN
];
1057 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
1058 audit_get_sessionid(task
));
1059 put_task_struct(task
);
1060 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1063 static const struct file_operations proc_sessionid_operations
= {
1064 .read
= proc_sessionid_read
,
1068 #ifdef CONFIG_FAULT_INJECTION
1069 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
1070 size_t count
, loff_t
*ppos
)
1072 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
1073 char buffer
[PROC_NUMBUF
];
1079 make_it_fail
= task
->make_it_fail
;
1080 put_task_struct(task
);
1082 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
1084 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
1087 static ssize_t
proc_fault_inject_write(struct file
* file
,
1088 const char __user
* buf
, size_t count
, loff_t
*ppos
)
1090 struct task_struct
*task
;
1091 char buffer
[PROC_NUMBUF
], *end
;
1094 if (!capable(CAP_SYS_RESOURCE
))
1096 memset(buffer
, 0, sizeof(buffer
));
1097 if (count
> sizeof(buffer
) - 1)
1098 count
= sizeof(buffer
) - 1;
1099 if (copy_from_user(buffer
, buf
, count
))
1101 make_it_fail
= simple_strtol(buffer
, &end
, 0);
1104 task
= get_proc_task(file
->f_dentry
->d_inode
);
1107 task
->make_it_fail
= make_it_fail
;
1108 put_task_struct(task
);
1109 if (end
- buffer
== 0)
1111 return end
- buffer
;
1114 static const struct file_operations proc_fault_inject_operations
= {
1115 .read
= proc_fault_inject_read
,
1116 .write
= proc_fault_inject_write
,
1121 #ifdef CONFIG_SCHED_DEBUG
1123 * Print out various scheduling related per-task fields:
1125 static int sched_show(struct seq_file
*m
, void *v
)
1127 struct inode
*inode
= m
->private;
1128 struct task_struct
*p
;
1132 p
= get_proc_task(inode
);
1135 proc_sched_show_task(p
, m
);
1143 sched_write(struct file
*file
, const char __user
*buf
,
1144 size_t count
, loff_t
*offset
)
1146 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1147 struct task_struct
*p
;
1151 p
= get_proc_task(inode
);
1154 proc_sched_set_task(p
);
1161 static int sched_open(struct inode
*inode
, struct file
*filp
)
1165 ret
= single_open(filp
, sched_show
, NULL
);
1167 struct seq_file
*m
= filp
->private_data
;
1174 static const struct file_operations proc_pid_sched_operations
= {
1177 .write
= sched_write
,
1178 .llseek
= seq_lseek
,
1179 .release
= single_release
,
1184 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1186 struct inode
*inode
= dentry
->d_inode
;
1187 int error
= -EACCES
;
1189 /* We don't need a base pointer in the /proc filesystem */
1190 path_put(&nd
->path
);
1192 /* Are we allowed to snoop on the tasks file descriptors? */
1193 if (!proc_fd_access_allowed(inode
))
1196 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->path
);
1197 nd
->last_type
= LAST_BIND
;
1199 return ERR_PTR(error
);
1202 static int do_proc_readlink(struct path
*path
, char __user
*buffer
, int buflen
)
1204 char *tmp
= (char*)__get_free_page(GFP_TEMPORARY
);
1211 pathname
= d_path(path
, tmp
, PAGE_SIZE
);
1212 len
= PTR_ERR(pathname
);
1213 if (IS_ERR(pathname
))
1215 len
= tmp
+ PAGE_SIZE
- 1 - pathname
;
1219 if (copy_to_user(buffer
, pathname
, len
))
1222 free_page((unsigned long)tmp
);
1226 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1228 int error
= -EACCES
;
1229 struct inode
*inode
= dentry
->d_inode
;
1232 /* Are we allowed to snoop on the tasks file descriptors? */
1233 if (!proc_fd_access_allowed(inode
))
1236 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &path
);
1240 error
= do_proc_readlink(&path
, buffer
, buflen
);
1246 static const struct inode_operations proc_pid_link_inode_operations
= {
1247 .readlink
= proc_pid_readlink
,
1248 .follow_link
= proc_pid_follow_link
,
1249 .setattr
= proc_setattr
,
1253 /* building an inode */
1255 static int task_dumpable(struct task_struct
*task
)
1258 struct mm_struct
*mm
;
1263 dumpable
= get_dumpable(mm
);
1271 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1273 struct inode
* inode
;
1274 struct proc_inode
*ei
;
1276 /* We need a new inode */
1278 inode
= new_inode(sb
);
1284 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1285 inode
->i_op
= &proc_def_inode_operations
;
1288 * grab the reference to task.
1290 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1296 if (task_dumpable(task
)) {
1297 inode
->i_uid
= task
->euid
;
1298 inode
->i_gid
= task
->egid
;
1300 security_task_to_inode(task
, inode
);
1310 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1312 struct inode
*inode
= dentry
->d_inode
;
1313 struct task_struct
*task
;
1314 generic_fillattr(inode
, stat
);
1319 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1321 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1322 task_dumpable(task
)) {
1323 stat
->uid
= task
->euid
;
1324 stat
->gid
= task
->egid
;
1334 * Exceptional case: normally we are not allowed to unhash a busy
1335 * directory. In this case, however, we can do it - no aliasing problems
1336 * due to the way we treat inodes.
1338 * Rewrite the inode's ownerships here because the owning task may have
1339 * performed a setuid(), etc.
1341 * Before the /proc/pid/status file was created the only way to read
1342 * the effective uid of a /process was to stat /proc/pid. Reading
1343 * /proc/pid/status is slow enough that procps and other packages
1344 * kept stating /proc/pid. To keep the rules in /proc simple I have
1345 * made this apply to all per process world readable and executable
1348 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1350 struct inode
*inode
= dentry
->d_inode
;
1351 struct task_struct
*task
= get_proc_task(inode
);
1353 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1354 task_dumpable(task
)) {
1355 inode
->i_uid
= task
->euid
;
1356 inode
->i_gid
= task
->egid
;
1361 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1362 security_task_to_inode(task
, inode
);
1363 put_task_struct(task
);
1370 static int pid_delete_dentry(struct dentry
* dentry
)
1372 /* Is the task we represent dead?
1373 * If so, then don't put the dentry on the lru list,
1374 * kill it immediately.
1376 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1379 static struct dentry_operations pid_dentry_operations
=
1381 .d_revalidate
= pid_revalidate
,
1382 .d_delete
= pid_delete_dentry
,
1387 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1388 struct task_struct
*, const void *);
1391 * Fill a directory entry.
1393 * If possible create the dcache entry and derive our inode number and
1394 * file type from dcache entry.
1396 * Since all of the proc inode numbers are dynamically generated, the inode
1397 * numbers do not exist until the inode is cache. This means creating the
1398 * the dcache entry in readdir is necessary to keep the inode numbers
1399 * reported by readdir in sync with the inode numbers reported
1402 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1403 char *name
, int len
,
1404 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1406 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1407 struct inode
*inode
;
1410 unsigned type
= DT_UNKNOWN
;
1414 qname
.hash
= full_name_hash(name
, len
);
1416 child
= d_lookup(dir
, &qname
);
1419 new = d_alloc(dir
, &qname
);
1421 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1428 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1429 goto end_instantiate
;
1430 inode
= child
->d_inode
;
1433 type
= inode
->i_mode
>> 12;
1438 ino
= find_inode_number(dir
, &qname
);
1441 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1444 static unsigned name_to_int(struct dentry
*dentry
)
1446 const char *name
= dentry
->d_name
.name
;
1447 int len
= dentry
->d_name
.len
;
1450 if (len
> 1 && *name
== '0')
1453 unsigned c
= *name
++ - '0';
1456 if (n
>= (~0U-9)/10)
1466 #define PROC_FDINFO_MAX 64
1468 static int proc_fd_info(struct inode
*inode
, struct path
*path
, char *info
)
1470 struct task_struct
*task
= get_proc_task(inode
);
1471 struct files_struct
*files
= NULL
;
1473 int fd
= proc_fd(inode
);
1476 files
= get_files_struct(task
);
1477 put_task_struct(task
);
1481 * We are not taking a ref to the file structure, so we must
1484 spin_lock(&files
->file_lock
);
1485 file
= fcheck_files(files
, fd
);
1488 *path
= file
->f_path
;
1489 path_get(&file
->f_path
);
1492 snprintf(info
, PROC_FDINFO_MAX
,
1495 (long long) file
->f_pos
,
1497 spin_unlock(&files
->file_lock
);
1498 put_files_struct(files
);
1501 spin_unlock(&files
->file_lock
);
1502 put_files_struct(files
);
1507 static int proc_fd_link(struct inode
*inode
, struct path
*path
)
1509 return proc_fd_info(inode
, path
, NULL
);
1512 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1514 struct inode
*inode
= dentry
->d_inode
;
1515 struct task_struct
*task
= get_proc_task(inode
);
1516 int fd
= proc_fd(inode
);
1517 struct files_struct
*files
;
1520 files
= get_files_struct(task
);
1523 if (fcheck_files(files
, fd
)) {
1525 put_files_struct(files
);
1526 if (task_dumpable(task
)) {
1527 inode
->i_uid
= task
->euid
;
1528 inode
->i_gid
= task
->egid
;
1533 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1534 security_task_to_inode(task
, inode
);
1535 put_task_struct(task
);
1539 put_files_struct(files
);
1541 put_task_struct(task
);
1547 static struct dentry_operations tid_fd_dentry_operations
=
1549 .d_revalidate
= tid_fd_revalidate
,
1550 .d_delete
= pid_delete_dentry
,
1553 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1554 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1556 unsigned fd
= *(const unsigned *)ptr
;
1558 struct files_struct
*files
;
1559 struct inode
*inode
;
1560 struct proc_inode
*ei
;
1561 struct dentry
*error
= ERR_PTR(-ENOENT
);
1563 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1568 files
= get_files_struct(task
);
1571 inode
->i_mode
= S_IFLNK
;
1574 * We are not taking a ref to the file structure, so we must
1577 spin_lock(&files
->file_lock
);
1578 file
= fcheck_files(files
, fd
);
1581 if (file
->f_mode
& 1)
1582 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1583 if (file
->f_mode
& 2)
1584 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1585 spin_unlock(&files
->file_lock
);
1586 put_files_struct(files
);
1588 inode
->i_op
= &proc_pid_link_inode_operations
;
1590 ei
->op
.proc_get_link
= proc_fd_link
;
1591 dentry
->d_op
= &tid_fd_dentry_operations
;
1592 d_add(dentry
, inode
);
1593 /* Close the race of the process dying before we return the dentry */
1594 if (tid_fd_revalidate(dentry
, NULL
))
1600 spin_unlock(&files
->file_lock
);
1601 put_files_struct(files
);
1607 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1608 struct dentry
*dentry
,
1609 instantiate_t instantiate
)
1611 struct task_struct
*task
= get_proc_task(dir
);
1612 unsigned fd
= name_to_int(dentry
);
1613 struct dentry
*result
= ERR_PTR(-ENOENT
);
1620 result
= instantiate(dir
, dentry
, task
, &fd
);
1622 put_task_struct(task
);
1627 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1628 filldir_t filldir
, instantiate_t instantiate
)
1630 struct dentry
*dentry
= filp
->f_path
.dentry
;
1631 struct inode
*inode
= dentry
->d_inode
;
1632 struct task_struct
*p
= get_proc_task(inode
);
1633 unsigned int fd
, ino
;
1635 struct files_struct
* files
;
1645 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1649 ino
= parent_ino(dentry
);
1650 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1654 files
= get_files_struct(p
);
1658 for (fd
= filp
->f_pos
-2;
1659 fd
< files_fdtable(files
)->max_fds
;
1660 fd
++, filp
->f_pos
++) {
1661 char name
[PROC_NUMBUF
];
1664 if (!fcheck_files(files
, fd
))
1668 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1669 if (proc_fill_cache(filp
, dirent
, filldir
,
1670 name
, len
, instantiate
,
1678 put_files_struct(files
);
1686 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1687 struct nameidata
*nd
)
1689 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1692 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1694 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1697 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1698 size_t len
, loff_t
*ppos
)
1700 char tmp
[PROC_FDINFO_MAX
];
1701 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, tmp
);
1703 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1707 static const struct file_operations proc_fdinfo_file_operations
= {
1708 .open
= nonseekable_open
,
1709 .read
= proc_fdinfo_read
,
1712 static const struct file_operations proc_fd_operations
= {
1713 .read
= generic_read_dir
,
1714 .readdir
= proc_readfd
,
1718 * /proc/pid/fd needs a special permission handler so that a process can still
1719 * access /proc/self/fd after it has executed a setuid().
1721 static int proc_fd_permission(struct inode
*inode
, int mask
,
1722 struct nameidata
*nd
)
1726 rv
= generic_permission(inode
, mask
, NULL
);
1729 if (task_pid(current
) == proc_pid(inode
))
1735 * proc directories can do almost nothing..
1737 static const struct inode_operations proc_fd_inode_operations
= {
1738 .lookup
= proc_lookupfd
,
1739 .permission
= proc_fd_permission
,
1740 .setattr
= proc_setattr
,
1743 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
1744 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1746 unsigned fd
= *(unsigned *)ptr
;
1747 struct inode
*inode
;
1748 struct proc_inode
*ei
;
1749 struct dentry
*error
= ERR_PTR(-ENOENT
);
1751 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1756 inode
->i_mode
= S_IFREG
| S_IRUSR
;
1757 inode
->i_fop
= &proc_fdinfo_file_operations
;
1758 dentry
->d_op
= &tid_fd_dentry_operations
;
1759 d_add(dentry
, inode
);
1760 /* Close the race of the process dying before we return the dentry */
1761 if (tid_fd_revalidate(dentry
, NULL
))
1768 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
1769 struct dentry
*dentry
,
1770 struct nameidata
*nd
)
1772 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
1775 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
1777 return proc_readfd_common(filp
, dirent
, filldir
,
1778 proc_fdinfo_instantiate
);
1781 static const struct file_operations proc_fdinfo_operations
= {
1782 .read
= generic_read_dir
,
1783 .readdir
= proc_readfdinfo
,
1787 * proc directories can do almost nothing..
1789 static const struct inode_operations proc_fdinfo_inode_operations
= {
1790 .lookup
= proc_lookupfdinfo
,
1791 .setattr
= proc_setattr
,
1795 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
1796 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1798 const struct pid_entry
*p
= ptr
;
1799 struct inode
*inode
;
1800 struct proc_inode
*ei
;
1801 struct dentry
*error
= ERR_PTR(-EINVAL
);
1803 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1808 inode
->i_mode
= p
->mode
;
1809 if (S_ISDIR(inode
->i_mode
))
1810 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1812 inode
->i_op
= p
->iop
;
1814 inode
->i_fop
= p
->fop
;
1816 dentry
->d_op
= &pid_dentry_operations
;
1817 d_add(dentry
, inode
);
1818 /* Close the race of the process dying before we return the dentry */
1819 if (pid_revalidate(dentry
, NULL
))
1825 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1826 struct dentry
*dentry
,
1827 const struct pid_entry
*ents
,
1830 struct inode
*inode
;
1831 struct dentry
*error
;
1832 struct task_struct
*task
= get_proc_task(dir
);
1833 const struct pid_entry
*p
, *last
;
1835 error
= ERR_PTR(-ENOENT
);
1842 * Yes, it does not scale. And it should not. Don't add
1843 * new entries into /proc/<tgid>/ without very good reasons.
1845 last
= &ents
[nents
- 1];
1846 for (p
= ents
; p
<= last
; p
++) {
1847 if (p
->len
!= dentry
->d_name
.len
)
1849 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1855 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
1857 put_task_struct(task
);
1862 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
1863 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
1865 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
1866 proc_pident_instantiate
, task
, p
);
1869 static int proc_pident_readdir(struct file
*filp
,
1870 void *dirent
, filldir_t filldir
,
1871 const struct pid_entry
*ents
, unsigned int nents
)
1874 struct dentry
*dentry
= filp
->f_path
.dentry
;
1875 struct inode
*inode
= dentry
->d_inode
;
1876 struct task_struct
*task
= get_proc_task(inode
);
1877 const struct pid_entry
*p
, *last
;
1890 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1896 ino
= parent_ino(dentry
);
1897 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1909 last
= &ents
[nents
- 1];
1911 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
1920 put_task_struct(task
);
1925 #ifdef CONFIG_SECURITY
1926 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1927 size_t count
, loff_t
*ppos
)
1929 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1932 struct task_struct
*task
= get_proc_task(inode
);
1937 length
= security_getprocattr(task
,
1938 (char*)file
->f_path
.dentry
->d_name
.name
,
1940 put_task_struct(task
);
1942 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
1947 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1948 size_t count
, loff_t
*ppos
)
1950 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1953 struct task_struct
*task
= get_proc_task(inode
);
1958 if (count
> PAGE_SIZE
)
1961 /* No partial writes. */
1967 page
= (char*)__get_free_page(GFP_TEMPORARY
);
1972 if (copy_from_user(page
, buf
, count
))
1975 length
= security_setprocattr(task
,
1976 (char*)file
->f_path
.dentry
->d_name
.name
,
1977 (void*)page
, count
);
1979 free_page((unsigned long) page
);
1981 put_task_struct(task
);
1986 static const struct file_operations proc_pid_attr_operations
= {
1987 .read
= proc_pid_attr_read
,
1988 .write
= proc_pid_attr_write
,
1991 static const struct pid_entry attr_dir_stuff
[] = {
1992 REG("current", S_IRUGO
|S_IWUGO
, pid_attr
),
1993 REG("prev", S_IRUGO
, pid_attr
),
1994 REG("exec", S_IRUGO
|S_IWUGO
, pid_attr
),
1995 REG("fscreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1996 REG("keycreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1997 REG("sockcreate", S_IRUGO
|S_IWUGO
, pid_attr
),
2000 static int proc_attr_dir_readdir(struct file
* filp
,
2001 void * dirent
, filldir_t filldir
)
2003 return proc_pident_readdir(filp
,dirent
,filldir
,
2004 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
2007 static const struct file_operations proc_attr_dir_operations
= {
2008 .read
= generic_read_dir
,
2009 .readdir
= proc_attr_dir_readdir
,
2012 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
2013 struct dentry
*dentry
, struct nameidata
*nd
)
2015 return proc_pident_lookup(dir
, dentry
,
2016 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
2019 static const struct inode_operations proc_attr_dir_inode_operations
= {
2020 .lookup
= proc_attr_dir_lookup
,
2021 .getattr
= pid_getattr
,
2022 .setattr
= proc_setattr
,
2027 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2028 static ssize_t
proc_coredump_filter_read(struct file
*file
, char __user
*buf
,
2029 size_t count
, loff_t
*ppos
)
2031 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
2032 struct mm_struct
*mm
;
2033 char buffer
[PROC_NUMBUF
];
2041 mm
= get_task_mm(task
);
2043 len
= snprintf(buffer
, sizeof(buffer
), "%08lx\n",
2044 ((mm
->flags
& MMF_DUMP_FILTER_MASK
) >>
2045 MMF_DUMP_FILTER_SHIFT
));
2047 ret
= simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
2050 put_task_struct(task
);
2055 static ssize_t
proc_coredump_filter_write(struct file
*file
,
2056 const char __user
*buf
,
2060 struct task_struct
*task
;
2061 struct mm_struct
*mm
;
2062 char buffer
[PROC_NUMBUF
], *end
;
2069 memset(buffer
, 0, sizeof(buffer
));
2070 if (count
> sizeof(buffer
) - 1)
2071 count
= sizeof(buffer
) - 1;
2072 if (copy_from_user(buffer
, buf
, count
))
2076 val
= (unsigned int)simple_strtoul(buffer
, &end
, 0);
2079 if (end
- buffer
== 0)
2083 task
= get_proc_task(file
->f_dentry
->d_inode
);
2088 mm
= get_task_mm(task
);
2092 for (i
= 0, mask
= 1; i
< MMF_DUMP_FILTER_BITS
; i
++, mask
<<= 1) {
2094 set_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2096 clear_bit(i
+ MMF_DUMP_FILTER_SHIFT
, &mm
->flags
);
2101 put_task_struct(task
);
2106 static const struct file_operations proc_coredump_filter_operations
= {
2107 .read
= proc_coredump_filter_read
,
2108 .write
= proc_coredump_filter_write
,
2115 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
2118 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2119 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2120 char tmp
[PROC_NUMBUF
];
2123 sprintf(tmp
, "%d", tgid
);
2124 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
2127 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
2129 struct pid_namespace
*ns
= dentry
->d_sb
->s_fs_info
;
2130 pid_t tgid
= task_tgid_nr_ns(current
, ns
);
2131 char tmp
[PROC_NUMBUF
];
2133 return ERR_PTR(-ENOENT
);
2134 sprintf(tmp
, "%d", task_tgid_nr_ns(current
, ns
));
2135 return ERR_PTR(vfs_follow_link(nd
,tmp
));
2138 static const struct inode_operations proc_self_inode_operations
= {
2139 .readlink
= proc_self_readlink
,
2140 .follow_link
= proc_self_follow_link
,
2146 * These are the directory entries in the root directory of /proc
2147 * that properly belong to the /proc filesystem, as they describe
2148 * describe something that is process related.
2150 static const struct pid_entry proc_base_stuff
[] = {
2151 NOD("self", S_IFLNK
|S_IRWXUGO
,
2152 &proc_self_inode_operations
, NULL
, {}),
2156 * Exceptional case: normally we are not allowed to unhash a busy
2157 * directory. In this case, however, we can do it - no aliasing problems
2158 * due to the way we treat inodes.
2160 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
2162 struct inode
*inode
= dentry
->d_inode
;
2163 struct task_struct
*task
= get_proc_task(inode
);
2165 put_task_struct(task
);
2172 static struct dentry_operations proc_base_dentry_operations
=
2174 .d_revalidate
= proc_base_revalidate
,
2175 .d_delete
= pid_delete_dentry
,
2178 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
2179 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2181 const struct pid_entry
*p
= ptr
;
2182 struct inode
*inode
;
2183 struct proc_inode
*ei
;
2184 struct dentry
*error
= ERR_PTR(-EINVAL
);
2186 /* Allocate the inode */
2187 error
= ERR_PTR(-ENOMEM
);
2188 inode
= new_inode(dir
->i_sb
);
2192 /* Initialize the inode */
2194 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2197 * grab the reference to the task.
2199 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
2205 inode
->i_mode
= p
->mode
;
2206 if (S_ISDIR(inode
->i_mode
))
2208 if (S_ISLNK(inode
->i_mode
))
2211 inode
->i_op
= p
->iop
;
2213 inode
->i_fop
= p
->fop
;
2215 dentry
->d_op
= &proc_base_dentry_operations
;
2216 d_add(dentry
, inode
);
2225 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
2227 struct dentry
*error
;
2228 struct task_struct
*task
= get_proc_task(dir
);
2229 const struct pid_entry
*p
, *last
;
2231 error
= ERR_PTR(-ENOENT
);
2236 /* Lookup the directory entry */
2237 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
2238 for (p
= proc_base_stuff
; p
<= last
; p
++) {
2239 if (p
->len
!= dentry
->d_name
.len
)
2241 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
2247 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
2250 put_task_struct(task
);
2255 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
2256 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
2258 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
2259 proc_base_instantiate
, task
, p
);
2262 #ifdef CONFIG_TASK_IO_ACCOUNTING
2263 static int proc_pid_io_accounting(struct task_struct
*task
, char *buffer
)
2265 return sprintf(buffer
,
2266 #ifdef CONFIG_TASK_XACCT
2272 "read_bytes: %llu\n"
2273 "write_bytes: %llu\n"
2274 "cancelled_write_bytes: %llu\n",
2275 #ifdef CONFIG_TASK_XACCT
2276 (unsigned long long)task
->rchar
,
2277 (unsigned long long)task
->wchar
,
2278 (unsigned long long)task
->syscr
,
2279 (unsigned long long)task
->syscw
,
2281 (unsigned long long)task
->ioac
.read_bytes
,
2282 (unsigned long long)task
->ioac
.write_bytes
,
2283 (unsigned long long)task
->ioac
.cancelled_write_bytes
);
2290 static const struct file_operations proc_task_operations
;
2291 static const struct inode_operations proc_task_inode_operations
;
2293 static const struct pid_entry tgid_base_stuff
[] = {
2294 DIR("task", S_IRUGO
|S_IXUGO
, task
),
2295 DIR("fd", S_IRUSR
|S_IXUSR
, fd
),
2296 DIR("fdinfo", S_IRUSR
|S_IXUSR
, fdinfo
),
2298 DIR("net", S_IRUGO
|S_IXUGO
, net
),
2300 REG("environ", S_IRUSR
, environ
),
2301 INF("auxv", S_IRUSR
, pid_auxv
),
2302 ONE("status", S_IRUGO
, pid_status
),
2303 INF("limits", S_IRUSR
, pid_limits
),
2304 #ifdef CONFIG_SCHED_DEBUG
2305 REG("sched", S_IRUGO
|S_IWUSR
, pid_sched
),
2307 INF("cmdline", S_IRUGO
, pid_cmdline
),
2308 ONE("stat", S_IRUGO
, tgid_stat
),
2309 ONE("statm", S_IRUGO
, pid_statm
),
2310 REG("maps", S_IRUGO
, maps
),
2312 REG("numa_maps", S_IRUGO
, numa_maps
),
2314 REG("mem", S_IRUSR
|S_IWUSR
, mem
),
2318 REG("mounts", S_IRUGO
, mounts
),
2319 REG("mountinfo", S_IRUGO
, mountinfo
),
2320 REG("mountstats", S_IRUSR
, mountstats
),
2321 #ifdef CONFIG_PROC_PAGE_MONITOR
2322 REG("clear_refs", S_IWUSR
, clear_refs
),
2323 REG("smaps", S_IRUGO
, smaps
),
2324 REG("pagemap", S_IRUSR
, pagemap
),
2326 #ifdef CONFIG_SECURITY
2327 DIR("attr", S_IRUGO
|S_IXUGO
, attr_dir
),
2329 #ifdef CONFIG_KALLSYMS
2330 INF("wchan", S_IRUGO
, pid_wchan
),
2332 #ifdef CONFIG_SCHEDSTATS
2333 INF("schedstat", S_IRUGO
, pid_schedstat
),
2335 #ifdef CONFIG_LATENCYTOP
2336 REG("latency", S_IRUGO
, lstats
),
2338 #ifdef CONFIG_PROC_PID_CPUSET
2339 REG("cpuset", S_IRUGO
, cpuset
),
2341 #ifdef CONFIG_CGROUPS
2342 REG("cgroup", S_IRUGO
, cgroup
),
2344 INF("oom_score", S_IRUGO
, oom_score
),
2345 REG("oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
2346 #ifdef CONFIG_AUDITSYSCALL
2347 REG("loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
2348 REG("sessionid", S_IRUSR
, sessionid
),
2350 #ifdef CONFIG_FAULT_INJECTION
2351 REG("make-it-fail", S_IRUGO
|S_IWUSR
, fault_inject
),
2353 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2354 REG("coredump_filter", S_IRUGO
|S_IWUSR
, coredump_filter
),
2356 #ifdef CONFIG_TASK_IO_ACCOUNTING
2357 INF("io", S_IRUGO
, pid_io_accounting
),
2361 static int proc_tgid_base_readdir(struct file
* filp
,
2362 void * dirent
, filldir_t filldir
)
2364 return proc_pident_readdir(filp
,dirent
,filldir
,
2365 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2368 static const struct file_operations proc_tgid_base_operations
= {
2369 .read
= generic_read_dir
,
2370 .readdir
= proc_tgid_base_readdir
,
2373 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2374 return proc_pident_lookup(dir
, dentry
,
2375 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2378 static const struct inode_operations proc_tgid_base_inode_operations
= {
2379 .lookup
= proc_tgid_base_lookup
,
2380 .getattr
= pid_getattr
,
2381 .setattr
= proc_setattr
,
2384 static void proc_flush_task_mnt(struct vfsmount
*mnt
, pid_t pid
, pid_t tgid
)
2386 struct dentry
*dentry
, *leader
, *dir
;
2387 char buf
[PROC_NUMBUF
];
2391 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2392 dentry
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2394 if (!(current
->flags
& PF_EXITING
))
2395 shrink_dcache_parent(dentry
);
2404 name
.len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2405 leader
= d_hash_and_lookup(mnt
->mnt_root
, &name
);
2410 name
.len
= strlen(name
.name
);
2411 dir
= d_hash_and_lookup(leader
, &name
);
2413 goto out_put_leader
;
2416 name
.len
= snprintf(buf
, sizeof(buf
), "%d", pid
);
2417 dentry
= d_hash_and_lookup(dir
, &name
);
2419 shrink_dcache_parent(dentry
);
2432 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2433 * @task: task that should be flushed.
2435 * When flushing dentries from proc, one needs to flush them from global
2436 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2437 * in. This call is supposed to do all of this job.
2439 * Looks in the dcache for
2441 * /proc/@tgid/task/@pid
2442 * if either directory is present flushes it and all of it'ts children
2445 * It is safe and reasonable to cache /proc entries for a task until
2446 * that task exits. After that they just clog up the dcache with
2447 * useless entries, possibly causing useful dcache entries to be
2448 * flushed instead. This routine is proved to flush those useless
2449 * dcache entries at process exit time.
2451 * NOTE: This routine is just an optimization so it does not guarantee
2452 * that no dcache entries will exist at process exit time it
2453 * just makes it very unlikely that any will persist.
2456 void proc_flush_task(struct task_struct
*task
)
2459 struct pid
*pid
, *tgid
= NULL
;
2462 pid
= task_pid(task
);
2463 if (thread_group_leader(task
))
2464 tgid
= task_tgid(task
);
2466 for (i
= 0; i
<= pid
->level
; i
++) {
2467 upid
= &pid
->numbers
[i
];
2468 proc_flush_task_mnt(upid
->ns
->proc_mnt
, upid
->nr
,
2469 tgid
? tgid
->numbers
[i
].nr
: 0);
2472 upid
= &pid
->numbers
[pid
->level
];
2474 pid_ns_release_proc(upid
->ns
);
2477 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2478 struct dentry
* dentry
,
2479 struct task_struct
*task
, const void *ptr
)
2481 struct dentry
*error
= ERR_PTR(-ENOENT
);
2482 struct inode
*inode
;
2484 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2488 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2489 inode
->i_op
= &proc_tgid_base_inode_operations
;
2490 inode
->i_fop
= &proc_tgid_base_operations
;
2491 inode
->i_flags
|=S_IMMUTABLE
;
2493 #ifdef CONFIG_SECURITY
2494 inode
->i_nlink
+= 1;
2497 dentry
->d_op
= &pid_dentry_operations
;
2499 d_add(dentry
, inode
);
2500 /* Close the race of the process dying before we return the dentry */
2501 if (pid_revalidate(dentry
, NULL
))
2507 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2509 struct dentry
*result
= ERR_PTR(-ENOENT
);
2510 struct task_struct
*task
;
2512 struct pid_namespace
*ns
;
2514 result
= proc_base_lookup(dir
, dentry
);
2515 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2518 tgid
= name_to_int(dentry
);
2522 ns
= dentry
->d_sb
->s_fs_info
;
2524 task
= find_task_by_pid_ns(tgid
, ns
);
2526 get_task_struct(task
);
2531 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2532 put_task_struct(task
);
2538 * Find the first task with tgid >= tgid
2543 struct task_struct
*task
;
2545 static struct tgid_iter
next_tgid(struct pid_namespace
*ns
, struct tgid_iter iter
)
2550 put_task_struct(iter
.task
);
2554 pid
= find_ge_pid(iter
.tgid
, ns
);
2556 iter
.tgid
= pid_nr_ns(pid
, ns
);
2557 iter
.task
= pid_task(pid
, PIDTYPE_PID
);
2558 /* What we to know is if the pid we have find is the
2559 * pid of a thread_group_leader. Testing for task
2560 * being a thread_group_leader is the obvious thing
2561 * todo but there is a window when it fails, due to
2562 * the pid transfer logic in de_thread.
2564 * So we perform the straight forward test of seeing
2565 * if the pid we have found is the pid of a thread
2566 * group leader, and don't worry if the task we have
2567 * found doesn't happen to be a thread group leader.
2568 * As we don't care in the case of readdir.
2570 if (!iter
.task
|| !has_group_leader_pid(iter
.task
)) {
2574 get_task_struct(iter
.task
);
2580 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2582 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2583 struct tgid_iter iter
)
2585 char name
[PROC_NUMBUF
];
2586 int len
= snprintf(name
, sizeof(name
), "%d", iter
.tgid
);
2587 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2588 proc_pid_instantiate
, iter
.task
, NULL
);
2591 /* for the /proc/ directory itself, after non-process stuff has been done */
2592 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2594 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2595 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2596 struct tgid_iter iter
;
2597 struct pid_namespace
*ns
;
2602 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2603 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2604 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2608 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2610 iter
.tgid
= filp
->f_pos
- TGID_OFFSET
;
2611 for (iter
= next_tgid(ns
, iter
);
2613 iter
.tgid
+= 1, iter
= next_tgid(ns
, iter
)) {
2614 filp
->f_pos
= iter
.tgid
+ TGID_OFFSET
;
2615 if (proc_pid_fill_cache(filp
, dirent
, filldir
, iter
) < 0) {
2616 put_task_struct(iter
.task
);
2620 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2622 put_task_struct(reaper
);
2630 static const struct pid_entry tid_base_stuff
[] = {
2631 DIR("fd", S_IRUSR
|S_IXUSR
, fd
),
2632 DIR("fdinfo", S_IRUSR
|S_IXUSR
, fdinfo
),
2633 REG("environ", S_IRUSR
, environ
),
2634 INF("auxv", S_IRUSR
, pid_auxv
),
2635 ONE("status", S_IRUGO
, pid_status
),
2636 INF("limits", S_IRUSR
, pid_limits
),
2637 #ifdef CONFIG_SCHED_DEBUG
2638 REG("sched", S_IRUGO
|S_IWUSR
, pid_sched
),
2640 INF("cmdline", S_IRUGO
, pid_cmdline
),
2641 ONE("stat", S_IRUGO
, tid_stat
),
2642 ONE("statm", S_IRUGO
, pid_statm
),
2643 REG("maps", S_IRUGO
, maps
),
2645 REG("numa_maps", S_IRUGO
, numa_maps
),
2647 REG("mem", S_IRUSR
|S_IWUSR
, mem
),
2651 REG("mounts", S_IRUGO
, mounts
),
2652 REG("mountinfo", S_IRUGO
, mountinfo
),
2653 #ifdef CONFIG_PROC_PAGE_MONITOR
2654 REG("clear_refs", S_IWUSR
, clear_refs
),
2655 REG("smaps", S_IRUGO
, smaps
),
2656 REG("pagemap", S_IRUSR
, pagemap
),
2658 #ifdef CONFIG_SECURITY
2659 DIR("attr", S_IRUGO
|S_IXUGO
, attr_dir
),
2661 #ifdef CONFIG_KALLSYMS
2662 INF("wchan", S_IRUGO
, pid_wchan
),
2664 #ifdef CONFIG_SCHEDSTATS
2665 INF("schedstat", S_IRUGO
, pid_schedstat
),
2667 #ifdef CONFIG_LATENCYTOP
2668 REG("latency", S_IRUGO
, lstats
),
2670 #ifdef CONFIG_PROC_PID_CPUSET
2671 REG("cpuset", S_IRUGO
, cpuset
),
2673 #ifdef CONFIG_CGROUPS
2674 REG("cgroup", S_IRUGO
, cgroup
),
2676 INF("oom_score", S_IRUGO
, oom_score
),
2677 REG("oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
2678 #ifdef CONFIG_AUDITSYSCALL
2679 REG("loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
2680 REG("sessionid", S_IRUSR
, sessionid
),
2682 #ifdef CONFIG_FAULT_INJECTION
2683 REG("make-it-fail", S_IRUGO
|S_IWUSR
, fault_inject
),
2687 static int proc_tid_base_readdir(struct file
* filp
,
2688 void * dirent
, filldir_t filldir
)
2690 return proc_pident_readdir(filp
,dirent
,filldir
,
2691 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2694 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2695 return proc_pident_lookup(dir
, dentry
,
2696 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
2699 static const struct file_operations proc_tid_base_operations
= {
2700 .read
= generic_read_dir
,
2701 .readdir
= proc_tid_base_readdir
,
2704 static const struct inode_operations proc_tid_base_inode_operations
= {
2705 .lookup
= proc_tid_base_lookup
,
2706 .getattr
= pid_getattr
,
2707 .setattr
= proc_setattr
,
2710 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
2711 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2713 struct dentry
*error
= ERR_PTR(-ENOENT
);
2714 struct inode
*inode
;
2715 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2719 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2720 inode
->i_op
= &proc_tid_base_inode_operations
;
2721 inode
->i_fop
= &proc_tid_base_operations
;
2722 inode
->i_flags
|=S_IMMUTABLE
;
2724 #ifdef CONFIG_SECURITY
2725 inode
->i_nlink
+= 1;
2728 dentry
->d_op
= &pid_dentry_operations
;
2730 d_add(dentry
, inode
);
2731 /* Close the race of the process dying before we return the dentry */
2732 if (pid_revalidate(dentry
, NULL
))
2738 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2740 struct dentry
*result
= ERR_PTR(-ENOENT
);
2741 struct task_struct
*task
;
2742 struct task_struct
*leader
= get_proc_task(dir
);
2744 struct pid_namespace
*ns
;
2749 tid
= name_to_int(dentry
);
2753 ns
= dentry
->d_sb
->s_fs_info
;
2755 task
= find_task_by_pid_ns(tid
, ns
);
2757 get_task_struct(task
);
2761 if (!same_thread_group(leader
, task
))
2764 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
2766 put_task_struct(task
);
2768 put_task_struct(leader
);
2774 * Find the first tid of a thread group to return to user space.
2776 * Usually this is just the thread group leader, but if the users
2777 * buffer was too small or there was a seek into the middle of the
2778 * directory we have more work todo.
2780 * In the case of a short read we start with find_task_by_pid.
2782 * In the case of a seek we start with the leader and walk nr
2785 static struct task_struct
*first_tid(struct task_struct
*leader
,
2786 int tid
, int nr
, struct pid_namespace
*ns
)
2788 struct task_struct
*pos
;
2791 /* Attempt to start with the pid of a thread */
2792 if (tid
&& (nr
> 0)) {
2793 pos
= find_task_by_pid_ns(tid
, ns
);
2794 if (pos
&& (pos
->group_leader
== leader
))
2798 /* If nr exceeds the number of threads there is nothing todo */
2800 if (nr
&& nr
>= get_nr_threads(leader
))
2803 /* If we haven't found our starting place yet start
2804 * with the leader and walk nr threads forward.
2806 for (pos
= leader
; nr
> 0; --nr
) {
2807 pos
= next_thread(pos
);
2808 if (pos
== leader
) {
2814 get_task_struct(pos
);
2821 * Find the next thread in the thread list.
2822 * Return NULL if there is an error or no next thread.
2824 * The reference to the input task_struct is released.
2826 static struct task_struct
*next_tid(struct task_struct
*start
)
2828 struct task_struct
*pos
= NULL
;
2830 if (pid_alive(start
)) {
2831 pos
= next_thread(start
);
2832 if (thread_group_leader(pos
))
2835 get_task_struct(pos
);
2838 put_task_struct(start
);
2842 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2843 struct task_struct
*task
, int tid
)
2845 char name
[PROC_NUMBUF
];
2846 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
2847 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2848 proc_task_instantiate
, task
, NULL
);
2851 /* for the /proc/TGID/task/ directories */
2852 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2854 struct dentry
*dentry
= filp
->f_path
.dentry
;
2855 struct inode
*inode
= dentry
->d_inode
;
2856 struct task_struct
*leader
= NULL
;
2857 struct task_struct
*task
;
2858 int retval
= -ENOENT
;
2861 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2862 struct pid_namespace
*ns
;
2864 task
= get_proc_task(inode
);
2868 if (pid_alive(task
)) {
2869 leader
= task
->group_leader
;
2870 get_task_struct(leader
);
2873 put_task_struct(task
);
2881 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2886 ino
= parent_ino(dentry
);
2887 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2893 /* f_version caches the tgid value that the last readdir call couldn't
2894 * return. lseek aka telldir automagically resets f_version to 0.
2896 ns
= filp
->f_dentry
->d_sb
->s_fs_info
;
2897 tid
= (int)filp
->f_version
;
2898 filp
->f_version
= 0;
2899 for (task
= first_tid(leader
, tid
, pos
- 2, ns
);
2901 task
= next_tid(task
), pos
++) {
2902 tid
= task_pid_nr_ns(task
, ns
);
2903 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
2904 /* returning this tgid failed, save it as the first
2905 * pid for the next readir call */
2906 filp
->f_version
= (u64
)tid
;
2907 put_task_struct(task
);
2913 put_task_struct(leader
);
2918 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
2920 struct inode
*inode
= dentry
->d_inode
;
2921 struct task_struct
*p
= get_proc_task(inode
);
2922 generic_fillattr(inode
, stat
);
2926 stat
->nlink
+= get_nr_threads(p
);
2934 static const struct inode_operations proc_task_inode_operations
= {
2935 .lookup
= proc_task_lookup
,
2936 .getattr
= proc_task_getattr
,
2937 .setattr
= proc_setattr
,
2940 static const struct file_operations proc_task_operations
= {
2941 .read
= generic_read_dir
,
2942 .readdir
= proc_task_readdir
,