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/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/stacktrace.h>
69 #include <linux/resource.h>
70 #include <linux/module.h>
71 #include <linux/mount.h>
72 #include <linux/security.h>
73 #include <linux/ptrace.h>
74 #include <linux/tracehook.h>
75 #include <linux/cgroup.h>
76 #include <linux/cpuset.h>
77 #include <linux/audit.h>
78 #include <linux/poll.h>
79 #include <linux/nsproxy.h>
80 #include <linux/oom.h>
81 #include <linux/elf.h>
82 #include <linux/pid_namespace.h>
83 #include <linux/fs_struct.h>
87 * Implementing inode permission operations in /proc is almost
88 * certainly an error. Permission checks need to happen during
89 * each system call not at open time. The reason is that most of
90 * what we wish to check for permissions in /proc varies at runtime.
92 * The classic example of a problem is opening file descriptors
93 * in /proc for a task before it execs a suid executable.
100 const struct inode_operations *iop;
101 const struct file_operations *fop;
105 #define NOD(NAME, MODE, IOP, FOP, OP) { \
107 .len = sizeof(NAME) - 1, \
114 #define DIR(NAME, MODE, iops, fops) \
115 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
116 #define LNK(NAME, get_link) \
117 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
118 &proc_pid_link_inode_operations, NULL, \
119 { .proc_get_link = get_link } )
120 #define REG(NAME, MODE, fops) \
121 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
122 #define INF(NAME, MODE, read) \
123 NOD(NAME, (S_IFREG|(MODE)), \
124 NULL, &proc_info_file_operations, \
125 { .proc_read = read } )
126 #define ONE(NAME, MODE, show) \
127 NOD(NAME, (S_IFREG|(MODE)), \
128 NULL, &proc_single_file_operations, \
129 { .proc_show = show } )
132 * Count the number of hardlinks for the pid_entry table, excluding the .
135 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
142 for (i = 0; i < n; ++i) {
143 if (S_ISDIR(entries[i].mode))
150 static int get_fs_path(struct task_struct *task, struct path *path, bool root)
152 struct fs_struct *fs;
153 int result = -ENOENT;
158 read_lock(&fs->lock);
159 *path = root ? fs->root : fs->pwd;
161 read_unlock(&fs->lock);
168 static int get_nr_threads(struct task_struct *tsk)
173 if (lock_task_sighand(tsk, &flags)) {
174 count = atomic_read(&tsk->signal->count);
175 unlock_task_sighand(tsk, &flags);
180 static int proc_cwd_link(struct inode *inode, struct path *path)
182 struct task_struct *task = get_proc_task(inode);
183 int result = -ENOENT;
186 result = get_fs_path(task, path, 0);
187 put_task_struct(task);
192 static int proc_root_link(struct inode *inode, struct path *path)
194 struct task_struct *task = get_proc_task(inode);
195 int result = -ENOENT;
198 result = get_fs_path(task, path, 1);
199 put_task_struct(task);
205 * Return zero if current may access user memory in @task, -error if not.
207 static int check_mem_permission(struct task_struct *task)
210 * A task can always look at itself, in case it chooses
211 * to use system calls instead of load instructions.
217 * If current is actively ptrace'ing, and would also be
218 * permitted to freshly attach with ptrace now, permit it.
220 if (task_is_stopped_or_traced(task)) {
223 match = (tracehook_tracer_task(task) == current);
225 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
230 * Noone else is allowed.
235 struct mm_struct *mm_for_maps(struct task_struct *task)
237 struct mm_struct *mm = get_task_mm(task);
240 if (mm != current->mm) {
242 * task->mm can be changed before security check,
243 * in that case we must notice the change after.
245 if (!ptrace_may_access(task, PTRACE_MODE_READ) ||
251 down_read(&mm->mmap_sem);
255 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
259 struct mm_struct *mm = get_task_mm(task);
263 goto out_mm; /* Shh! No looking before we're done */
265 len = mm->arg_end - mm->arg_start;
270 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
272 // If the nul at the end of args has been overwritten, then
273 // assume application is using setproctitle(3).
274 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
275 len = strnlen(buffer, res);
279 len = mm->env_end - mm->env_start;
280 if (len > PAGE_SIZE - res)
281 len = PAGE_SIZE - res;
282 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
283 res = strnlen(buffer, res);
292 static int proc_pid_auxv(struct task_struct *task, char *buffer)
295 struct mm_struct *mm = get_task_mm(task);
297 unsigned int nwords = 0;
300 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
301 res = nwords * sizeof(mm->saved_auxv[0]);
304 memcpy(buffer, mm->saved_auxv, res);
311 #ifdef CONFIG_KALLSYMS
313 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
314 * Returns the resolved symbol. If that fails, simply return the address.
316 static int proc_pid_wchan(struct task_struct *task, char *buffer)
319 char symname[KSYM_NAME_LEN];
321 wchan = get_wchan(task);
323 if (lookup_symbol_name(wchan, symname) < 0)
324 if (!ptrace_may_access(task, PTRACE_MODE_READ))
327 return sprintf(buffer, "%lu", wchan);
329 return sprintf(buffer, "%s", symname);
331 #endif /* CONFIG_KALLSYMS */
333 #ifdef CONFIG_STACKTRACE
335 #define MAX_STACK_TRACE_DEPTH 64
337 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
338 struct pid *pid, struct task_struct *task)
340 struct stack_trace trace;
341 unsigned long *entries;
344 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
348 trace.nr_entries = 0;
349 trace.max_entries = MAX_STACK_TRACE_DEPTH;
350 trace.entries = entries;
352 save_stack_trace_tsk(task, &trace);
354 for (i = 0; i < trace.nr_entries; i++) {
355 seq_printf(m, "[<%p>] %pS\n",
356 (void *)entries[i], (void *)entries[i]);
364 #ifdef CONFIG_SCHEDSTATS
366 * Provides /proc/PID/schedstat
368 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
370 return sprintf(buffer, "%llu %llu %lu\n",
371 (unsigned long long)task->se.sum_exec_runtime,
372 (unsigned long long)task->sched_info.run_delay,
373 task->sched_info.pcount);
377 #ifdef CONFIG_LATENCYTOP
378 static int lstats_show_proc(struct seq_file *m, void *v)
381 struct inode *inode = m->private;
382 struct task_struct *task = get_proc_task(inode);
386 seq_puts(m, "Latency Top version : v0.1\n");
387 for (i = 0; i < 32; i++) {
388 if (task->latency_record[i].backtrace[0]) {
390 seq_printf(m, "%i %li %li ",
391 task->latency_record[i].count,
392 task->latency_record[i].time,
393 task->latency_record[i].max);
394 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
395 char sym[KSYM_SYMBOL_LEN];
397 if (!task->latency_record[i].backtrace[q])
399 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
401 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
402 c = strchr(sym, '+');
405 seq_printf(m, "%s ", sym);
411 put_task_struct(task);
415 static int lstats_open(struct inode *inode, struct file *file)
417 return single_open(file, lstats_show_proc, inode);
420 static ssize_t lstats_write(struct file *file, const char __user *buf,
421 size_t count, loff_t *offs)
423 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
427 clear_all_latency_tracing(task);
428 put_task_struct(task);
433 static const struct file_operations proc_lstats_operations = {
436 .write = lstats_write,
438 .release = single_release,
443 /* The badness from the OOM killer */
444 unsigned long badness(struct task_struct *p, unsigned long uptime);
445 static int proc_oom_score(struct task_struct *task, char *buffer)
447 unsigned long points;
448 struct timespec uptime;
450 do_posix_clock_monotonic_gettime(&uptime);
451 read_lock(&tasklist_lock);
452 points = badness(task, uptime.tv_sec);
453 read_unlock(&tasklist_lock);
454 return sprintf(buffer, "%lu\n", points);
462 static const struct limit_names lnames[RLIM_NLIMITS] = {
463 [RLIMIT_CPU] = {"Max cpu time", "ms"},
464 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
465 [RLIMIT_DATA] = {"Max data size", "bytes"},
466 [RLIMIT_STACK] = {"Max stack size", "bytes"},
467 [RLIMIT_CORE] = {"Max core file size", "bytes"},
468 [RLIMIT_RSS] = {"Max resident set", "bytes"},
469 [RLIMIT_NPROC] = {"Max processes", "processes"},
470 [RLIMIT_NOFILE] = {"Max open files", "files"},
471 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
472 [RLIMIT_AS] = {"Max address space", "bytes"},
473 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
474 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
475 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
476 [RLIMIT_NICE] = {"Max nice priority", NULL},
477 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
478 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
481 /* Display limits for a process */
482 static int proc_pid_limits(struct task_struct *task, char *buffer)
487 char *bufptr = buffer;
489 struct rlimit rlim[RLIM_NLIMITS];
491 if (!lock_task_sighand(task, &flags))
493 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
494 unlock_task_sighand(task, &flags);
497 * print the file header
499 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
500 "Limit", "Soft Limit", "Hard Limit", "Units");
502 for (i = 0; i < RLIM_NLIMITS; i++) {
503 if (rlim[i].rlim_cur == RLIM_INFINITY)
504 count += sprintf(&bufptr[count], "%-25s %-20s ",
505 lnames[i].name, "unlimited");
507 count += sprintf(&bufptr[count], "%-25s %-20lu ",
508 lnames[i].name, rlim[i].rlim_cur);
510 if (rlim[i].rlim_max == RLIM_INFINITY)
511 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
513 count += sprintf(&bufptr[count], "%-20lu ",
517 count += sprintf(&bufptr[count], "%-10s\n",
520 count += sprintf(&bufptr[count], "\n");
526 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
527 static int proc_pid_syscall(struct task_struct *task, char *buffer)
530 unsigned long args[6], sp, pc;
532 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
533 return sprintf(buffer, "running\n");
536 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
538 return sprintf(buffer,
539 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
541 args[0], args[1], args[2], args[3], args[4], args[5],
544 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
546 /************************************************************************/
547 /* Here the fs part begins */
548 /************************************************************************/
550 /* permission checks */
551 static int proc_fd_access_allowed(struct inode *inode)
553 struct task_struct *task;
555 /* Allow access to a task's file descriptors if it is us or we
556 * may use ptrace attach to the process and find out that
559 task = get_proc_task(inode);
561 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
562 put_task_struct(task);
567 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
570 struct inode *inode = dentry->d_inode;
572 if (attr->ia_valid & ATTR_MODE)
575 error = inode_change_ok(inode, attr);
577 error = inode_setattr(inode, attr);
581 static const struct inode_operations proc_def_inode_operations = {
582 .setattr = proc_setattr,
585 static int mounts_open_common(struct inode *inode, struct file *file,
586 const struct seq_operations *op)
588 struct task_struct *task = get_proc_task(inode);
590 struct mnt_namespace *ns = NULL;
592 struct proc_mounts *p;
597 nsp = task_nsproxy(task);
604 if (ns && get_fs_path(task, &root, 1) == 0)
606 put_task_struct(task);
615 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
619 file->private_data = &p->m;
620 ret = seq_open(file, op);
627 p->event = ns->event;
641 static int mounts_release(struct inode *inode, struct file *file)
643 struct proc_mounts *p = file->private_data;
646 return seq_release(inode, file);
649 static unsigned mounts_poll(struct file *file, poll_table *wait)
651 struct proc_mounts *p = file->private_data;
652 struct mnt_namespace *ns = p->ns;
653 unsigned res = POLLIN | POLLRDNORM;
655 poll_wait(file, &ns->poll, wait);
657 spin_lock(&vfsmount_lock);
658 if (p->event != ns->event) {
659 p->event = ns->event;
660 res |= POLLERR | POLLPRI;
662 spin_unlock(&vfsmount_lock);
667 static int mounts_open(struct inode *inode, struct file *file)
669 return mounts_open_common(inode, file, &mounts_op);
672 static const struct file_operations proc_mounts_operations = {
676 .release = mounts_release,
680 static int mountinfo_open(struct inode *inode, struct file *file)
682 return mounts_open_common(inode, file, &mountinfo_op);
685 static const struct file_operations proc_mountinfo_operations = {
686 .open = mountinfo_open,
689 .release = mounts_release,
693 static int mountstats_open(struct inode *inode, struct file *file)
695 return mounts_open_common(inode, file, &mountstats_op);
698 static const struct file_operations proc_mountstats_operations = {
699 .open = mountstats_open,
702 .release = mounts_release,
705 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
707 static ssize_t proc_info_read(struct file * file, char __user * buf,
708 size_t count, loff_t *ppos)
710 struct inode * inode = file->f_path.dentry->d_inode;
713 struct task_struct *task = get_proc_task(inode);
719 if (count > PROC_BLOCK_SIZE)
720 count = PROC_BLOCK_SIZE;
723 if (!(page = __get_free_page(GFP_TEMPORARY)))
726 length = PROC_I(inode)->op.proc_read(task, (char*)page);
729 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
732 put_task_struct(task);
737 static const struct file_operations proc_info_file_operations = {
738 .read = proc_info_read,
741 static int proc_single_show(struct seq_file *m, void *v)
743 struct inode *inode = m->private;
744 struct pid_namespace *ns;
746 struct task_struct *task;
749 ns = inode->i_sb->s_fs_info;
750 pid = proc_pid(inode);
751 task = get_pid_task(pid, PIDTYPE_PID);
755 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
757 put_task_struct(task);
761 static int proc_single_open(struct inode *inode, struct file *filp)
764 ret = single_open(filp, proc_single_show, NULL);
766 struct seq_file *m = filp->private_data;
773 static const struct file_operations proc_single_file_operations = {
774 .open = proc_single_open,
777 .release = single_release,
780 static int mem_open(struct inode* inode, struct file* file)
782 file->private_data = (void*)((long)current->self_exec_id);
786 static ssize_t mem_read(struct file * file, char __user * buf,
787 size_t count, loff_t *ppos)
789 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
791 unsigned long src = *ppos;
793 struct mm_struct *mm;
798 if (check_mem_permission(task))
802 page = (char *)__get_free_page(GFP_TEMPORARY);
808 mm = get_task_mm(task);
814 if (file->private_data != (void*)((long)current->self_exec_id))
820 int this_len, retval;
822 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
823 retval = access_process_vm(task, src, page, this_len, 0);
824 if (!retval || check_mem_permission(task)) {
830 if (copy_to_user(buf, page, retval)) {
845 free_page((unsigned long) page);
847 put_task_struct(task);
852 #define mem_write NULL
855 /* This is a security hazard */
856 static ssize_t mem_write(struct file * file, const char __user *buf,
857 size_t count, loff_t *ppos)
861 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
862 unsigned long dst = *ppos;
868 if (check_mem_permission(task))
872 page = (char *)__get_free_page(GFP_TEMPORARY);
878 int this_len, retval;
880 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
881 if (copy_from_user(page, buf, this_len)) {
885 retval = access_process_vm(task, dst, page, this_len, 1);
897 free_page((unsigned long) page);
899 put_task_struct(task);
905 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
909 file->f_pos = offset;
912 file->f_pos += offset;
917 force_successful_syscall_return();
921 static const struct file_operations proc_mem_operations = {
928 static ssize_t environ_read(struct file *file, char __user *buf,
929 size_t count, loff_t *ppos)
931 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
933 unsigned long src = *ppos;
935 struct mm_struct *mm;
940 if (!ptrace_may_access(task, PTRACE_MODE_READ))
944 page = (char *)__get_free_page(GFP_TEMPORARY);
950 mm = get_task_mm(task);
955 int this_len, retval, max_len;
957 this_len = mm->env_end - (mm->env_start + src);
962 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
963 this_len = (this_len > max_len) ? max_len : this_len;
965 retval = access_process_vm(task, (mm->env_start + src),
973 if (copy_to_user(buf, page, retval)) {
987 free_page((unsigned long) page);
989 put_task_struct(task);
994 static const struct file_operations proc_environ_operations = {
995 .read = environ_read,
998 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
999 size_t count, loff_t *ppos)
1001 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1002 char buffer[PROC_NUMBUF];
1010 oom_adjust = task->mm->oom_adj;
1012 oom_adjust = OOM_DISABLE;
1014 put_task_struct(task);
1016 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1018 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1021 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1022 size_t count, loff_t *ppos)
1024 struct task_struct *task;
1025 char buffer[PROC_NUMBUF], *end;
1028 memset(buffer, 0, sizeof(buffer));
1029 if (count > sizeof(buffer) - 1)
1030 count = sizeof(buffer) - 1;
1031 if (copy_from_user(buffer, buf, count))
1033 oom_adjust = simple_strtol(buffer, &end, 0);
1034 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1035 oom_adjust != OOM_DISABLE)
1039 task = get_proc_task(file->f_path.dentry->d_inode);
1045 put_task_struct(task);
1048 if (oom_adjust < task->mm->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1050 put_task_struct(task);
1053 task->mm->oom_adj = oom_adjust;
1055 put_task_struct(task);
1056 if (end - buffer == 0)
1058 return end - buffer;
1061 static const struct file_operations proc_oom_adjust_operations = {
1062 .read = oom_adjust_read,
1063 .write = oom_adjust_write,
1066 #ifdef CONFIG_AUDITSYSCALL
1067 #define TMPBUFLEN 21
1068 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1069 size_t count, loff_t *ppos)
1071 struct inode * inode = file->f_path.dentry->d_inode;
1072 struct task_struct *task = get_proc_task(inode);
1074 char tmpbuf[TMPBUFLEN];
1078 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1079 audit_get_loginuid(task));
1080 put_task_struct(task);
1081 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1084 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1085 size_t count, loff_t *ppos)
1087 struct inode * inode = file->f_path.dentry->d_inode;
1092 if (!capable(CAP_AUDIT_CONTROL))
1095 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
1098 if (count >= PAGE_SIZE)
1099 count = PAGE_SIZE - 1;
1102 /* No partial writes. */
1105 page = (char*)__get_free_page(GFP_TEMPORARY);
1109 if (copy_from_user(page, buf, count))
1113 loginuid = simple_strtoul(page, &tmp, 10);
1119 length = audit_set_loginuid(current, loginuid);
1120 if (likely(length == 0))
1124 free_page((unsigned long) page);
1128 static const struct file_operations proc_loginuid_operations = {
1129 .read = proc_loginuid_read,
1130 .write = proc_loginuid_write,
1133 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1134 size_t count, loff_t *ppos)
1136 struct inode * inode = file->f_path.dentry->d_inode;
1137 struct task_struct *task = get_proc_task(inode);
1139 char tmpbuf[TMPBUFLEN];
1143 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1144 audit_get_sessionid(task));
1145 put_task_struct(task);
1146 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1149 static const struct file_operations proc_sessionid_operations = {
1150 .read = proc_sessionid_read,
1154 #ifdef CONFIG_FAULT_INJECTION
1155 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1156 size_t count, loff_t *ppos)
1158 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1159 char buffer[PROC_NUMBUF];
1165 make_it_fail = task->make_it_fail;
1166 put_task_struct(task);
1168 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1170 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1173 static ssize_t proc_fault_inject_write(struct file * file,
1174 const char __user * buf, size_t count, loff_t *ppos)
1176 struct task_struct *task;
1177 char buffer[PROC_NUMBUF], *end;
1180 if (!capable(CAP_SYS_RESOURCE))
1182 memset(buffer, 0, sizeof(buffer));
1183 if (count > sizeof(buffer) - 1)
1184 count = sizeof(buffer) - 1;
1185 if (copy_from_user(buffer, buf, count))
1187 make_it_fail = simple_strtol(buffer, &end, 0);
1190 task = get_proc_task(file->f_dentry->d_inode);
1193 task->make_it_fail = make_it_fail;
1194 put_task_struct(task);
1195 if (end - buffer == 0)
1197 return end - buffer;
1200 static const struct file_operations proc_fault_inject_operations = {
1201 .read = proc_fault_inject_read,
1202 .write = proc_fault_inject_write,
1207 #ifdef CONFIG_SCHED_DEBUG
1209 * Print out various scheduling related per-task fields:
1211 static int sched_show(struct seq_file *m, void *v)
1213 struct inode *inode = m->private;
1214 struct task_struct *p;
1216 p = get_proc_task(inode);
1219 proc_sched_show_task(p, m);
1227 sched_write(struct file *file, const char __user *buf,
1228 size_t count, loff_t *offset)
1230 struct inode *inode = file->f_path.dentry->d_inode;
1231 struct task_struct *p;
1233 p = get_proc_task(inode);
1236 proc_sched_set_task(p);
1243 static int sched_open(struct inode *inode, struct file *filp)
1247 ret = single_open(filp, sched_show, NULL);
1249 struct seq_file *m = filp->private_data;
1256 static const struct file_operations proc_pid_sched_operations = {
1259 .write = sched_write,
1260 .llseek = seq_lseek,
1261 .release = single_release,
1267 * We added or removed a vma mapping the executable. The vmas are only mapped
1268 * during exec and are not mapped with the mmap system call.
1269 * Callers must hold down_write() on the mm's mmap_sem for these
1271 void added_exe_file_vma(struct mm_struct *mm)
1273 mm->num_exe_file_vmas++;
1276 void removed_exe_file_vma(struct mm_struct *mm)
1278 mm->num_exe_file_vmas--;
1279 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1281 mm->exe_file = NULL;
1286 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1289 get_file(new_exe_file);
1292 mm->exe_file = new_exe_file;
1293 mm->num_exe_file_vmas = 0;
1296 struct file *get_mm_exe_file(struct mm_struct *mm)
1298 struct file *exe_file;
1300 /* We need mmap_sem to protect against races with removal of
1301 * VM_EXECUTABLE vmas */
1302 down_read(&mm->mmap_sem);
1303 exe_file = mm->exe_file;
1306 up_read(&mm->mmap_sem);
1310 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1312 /* It's safe to write the exe_file pointer without exe_file_lock because
1313 * this is called during fork when the task is not yet in /proc */
1314 newmm->exe_file = get_mm_exe_file(oldmm);
1317 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1319 struct task_struct *task;
1320 struct mm_struct *mm;
1321 struct file *exe_file;
1323 task = get_proc_task(inode);
1326 mm = get_task_mm(task);
1327 put_task_struct(task);
1330 exe_file = get_mm_exe_file(mm);
1333 *exe_path = exe_file->f_path;
1334 path_get(&exe_file->f_path);
1341 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1343 struct inode *inode = dentry->d_inode;
1344 int error = -EACCES;
1346 /* We don't need a base pointer in the /proc filesystem */
1347 path_put(&nd->path);
1349 /* Are we allowed to snoop on the tasks file descriptors? */
1350 if (!proc_fd_access_allowed(inode))
1353 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1354 nd->last_type = LAST_BIND;
1356 return ERR_PTR(error);
1359 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1361 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1368 pathname = d_path(path, tmp, PAGE_SIZE);
1369 len = PTR_ERR(pathname);
1370 if (IS_ERR(pathname))
1372 len = tmp + PAGE_SIZE - 1 - pathname;
1376 if (copy_to_user(buffer, pathname, len))
1379 free_page((unsigned long)tmp);
1383 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1385 int error = -EACCES;
1386 struct inode *inode = dentry->d_inode;
1389 /* Are we allowed to snoop on the tasks file descriptors? */
1390 if (!proc_fd_access_allowed(inode))
1393 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1397 error = do_proc_readlink(&path, buffer, buflen);
1403 static const struct inode_operations proc_pid_link_inode_operations = {
1404 .readlink = proc_pid_readlink,
1405 .follow_link = proc_pid_follow_link,
1406 .setattr = proc_setattr,
1410 /* building an inode */
1412 static int task_dumpable(struct task_struct *task)
1415 struct mm_struct *mm;
1420 dumpable = get_dumpable(mm);
1428 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1430 struct inode * inode;
1431 struct proc_inode *ei;
1432 const struct cred *cred;
1434 /* We need a new inode */
1436 inode = new_inode(sb);
1442 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1443 inode->i_op = &proc_def_inode_operations;
1446 * grab the reference to task.
1448 ei->pid = get_task_pid(task, PIDTYPE_PID);
1452 if (task_dumpable(task)) {
1454 cred = __task_cred(task);
1455 inode->i_uid = cred->euid;
1456 inode->i_gid = cred->egid;
1459 security_task_to_inode(task, inode);
1469 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1471 struct inode *inode = dentry->d_inode;
1472 struct task_struct *task;
1473 const struct cred *cred;
1475 generic_fillattr(inode, stat);
1480 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1482 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1483 task_dumpable(task)) {
1484 cred = __task_cred(task);
1485 stat->uid = cred->euid;
1486 stat->gid = cred->egid;
1496 * Exceptional case: normally we are not allowed to unhash a busy
1497 * directory. In this case, however, we can do it - no aliasing problems
1498 * due to the way we treat inodes.
1500 * Rewrite the inode's ownerships here because the owning task may have
1501 * performed a setuid(), etc.
1503 * Before the /proc/pid/status file was created the only way to read
1504 * the effective uid of a /process was to stat /proc/pid. Reading
1505 * /proc/pid/status is slow enough that procps and other packages
1506 * kept stating /proc/pid. To keep the rules in /proc simple I have
1507 * made this apply to all per process world readable and executable
1510 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1512 struct inode *inode = dentry->d_inode;
1513 struct task_struct *task = get_proc_task(inode);
1514 const struct cred *cred;
1517 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1518 task_dumpable(task)) {
1520 cred = __task_cred(task);
1521 inode->i_uid = cred->euid;
1522 inode->i_gid = cred->egid;
1528 inode->i_mode &= ~(S_ISUID | S_ISGID);
1529 security_task_to_inode(task, inode);
1530 put_task_struct(task);
1537 static int pid_delete_dentry(struct dentry * dentry)
1539 /* Is the task we represent dead?
1540 * If so, then don't put the dentry on the lru list,
1541 * kill it immediately.
1543 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1546 static const struct dentry_operations pid_dentry_operations =
1548 .d_revalidate = pid_revalidate,
1549 .d_delete = pid_delete_dentry,
1554 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1555 struct task_struct *, const void *);
1558 * Fill a directory entry.
1560 * If possible create the dcache entry and derive our inode number and
1561 * file type from dcache entry.
1563 * Since all of the proc inode numbers are dynamically generated, the inode
1564 * numbers do not exist until the inode is cache. This means creating the
1565 * the dcache entry in readdir is necessary to keep the inode numbers
1566 * reported by readdir in sync with the inode numbers reported
1569 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1570 char *name, int len,
1571 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1573 struct dentry *child, *dir = filp->f_path.dentry;
1574 struct inode *inode;
1577 unsigned type = DT_UNKNOWN;
1581 qname.hash = full_name_hash(name, len);
1583 child = d_lookup(dir, &qname);
1586 new = d_alloc(dir, &qname);
1588 child = instantiate(dir->d_inode, new, task, ptr);
1595 if (!child || IS_ERR(child) || !child->d_inode)
1596 goto end_instantiate;
1597 inode = child->d_inode;
1600 type = inode->i_mode >> 12;
1605 ino = find_inode_number(dir, &qname);
1608 return filldir(dirent, name, len, filp->f_pos, ino, type);
1611 static unsigned name_to_int(struct dentry *dentry)
1613 const char *name = dentry->d_name.name;
1614 int len = dentry->d_name.len;
1617 if (len > 1 && *name == '0')
1620 unsigned c = *name++ - '0';
1623 if (n >= (~0U-9)/10)
1633 #define PROC_FDINFO_MAX 64
1635 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1637 struct task_struct *task = get_proc_task(inode);
1638 struct files_struct *files = NULL;
1640 int fd = proc_fd(inode);
1643 files = get_files_struct(task);
1644 put_task_struct(task);
1648 * We are not taking a ref to the file structure, so we must
1651 spin_lock(&files->file_lock);
1652 file = fcheck_files(files, fd);
1655 *path = file->f_path;
1656 path_get(&file->f_path);
1659 snprintf(info, PROC_FDINFO_MAX,
1662 (long long) file->f_pos,
1664 spin_unlock(&files->file_lock);
1665 put_files_struct(files);
1668 spin_unlock(&files->file_lock);
1669 put_files_struct(files);
1674 static int proc_fd_link(struct inode *inode, struct path *path)
1676 return proc_fd_info(inode, path, NULL);
1679 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1681 struct inode *inode = dentry->d_inode;
1682 struct task_struct *task = get_proc_task(inode);
1683 int fd = proc_fd(inode);
1684 struct files_struct *files;
1685 const struct cred *cred;
1688 files = get_files_struct(task);
1691 if (fcheck_files(files, fd)) {
1693 put_files_struct(files);
1694 if (task_dumpable(task)) {
1696 cred = __task_cred(task);
1697 inode->i_uid = cred->euid;
1698 inode->i_gid = cred->egid;
1704 inode->i_mode &= ~(S_ISUID | S_ISGID);
1705 security_task_to_inode(task, inode);
1706 put_task_struct(task);
1710 put_files_struct(files);
1712 put_task_struct(task);
1718 static const struct dentry_operations tid_fd_dentry_operations =
1720 .d_revalidate = tid_fd_revalidate,
1721 .d_delete = pid_delete_dentry,
1724 static struct dentry *proc_fd_instantiate(struct inode *dir,
1725 struct dentry *dentry, struct task_struct *task, const void *ptr)
1727 unsigned fd = *(const unsigned *)ptr;
1729 struct files_struct *files;
1730 struct inode *inode;
1731 struct proc_inode *ei;
1732 struct dentry *error = ERR_PTR(-ENOENT);
1734 inode = proc_pid_make_inode(dir->i_sb, task);
1739 files = get_files_struct(task);
1742 inode->i_mode = S_IFLNK;
1745 * We are not taking a ref to the file structure, so we must
1748 spin_lock(&files->file_lock);
1749 file = fcheck_files(files, fd);
1752 if (file->f_mode & FMODE_READ)
1753 inode->i_mode |= S_IRUSR | S_IXUSR;
1754 if (file->f_mode & FMODE_WRITE)
1755 inode->i_mode |= S_IWUSR | S_IXUSR;
1756 spin_unlock(&files->file_lock);
1757 put_files_struct(files);
1759 inode->i_op = &proc_pid_link_inode_operations;
1761 ei->op.proc_get_link = proc_fd_link;
1762 dentry->d_op = &tid_fd_dentry_operations;
1763 d_add(dentry, inode);
1764 /* Close the race of the process dying before we return the dentry */
1765 if (tid_fd_revalidate(dentry, NULL))
1771 spin_unlock(&files->file_lock);
1772 put_files_struct(files);
1778 static struct dentry *proc_lookupfd_common(struct inode *dir,
1779 struct dentry *dentry,
1780 instantiate_t instantiate)
1782 struct task_struct *task = get_proc_task(dir);
1783 unsigned fd = name_to_int(dentry);
1784 struct dentry *result = ERR_PTR(-ENOENT);
1791 result = instantiate(dir, dentry, task, &fd);
1793 put_task_struct(task);
1798 static int proc_readfd_common(struct file * filp, void * dirent,
1799 filldir_t filldir, instantiate_t instantiate)
1801 struct dentry *dentry = filp->f_path.dentry;
1802 struct inode *inode = dentry->d_inode;
1803 struct task_struct *p = get_proc_task(inode);
1804 unsigned int fd, ino;
1806 struct files_struct * files;
1816 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1820 ino = parent_ino(dentry);
1821 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1825 files = get_files_struct(p);
1829 for (fd = filp->f_pos-2;
1830 fd < files_fdtable(files)->max_fds;
1831 fd++, filp->f_pos++) {
1832 char name[PROC_NUMBUF];
1835 if (!fcheck_files(files, fd))
1839 len = snprintf(name, sizeof(name), "%d", fd);
1840 if (proc_fill_cache(filp, dirent, filldir,
1841 name, len, instantiate,
1849 put_files_struct(files);
1857 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1858 struct nameidata *nd)
1860 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1863 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1865 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1868 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1869 size_t len, loff_t *ppos)
1871 char tmp[PROC_FDINFO_MAX];
1872 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1874 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1878 static const struct file_operations proc_fdinfo_file_operations = {
1879 .open = nonseekable_open,
1880 .read = proc_fdinfo_read,
1883 static const struct file_operations proc_fd_operations = {
1884 .read = generic_read_dir,
1885 .readdir = proc_readfd,
1889 * /proc/pid/fd needs a special permission handler so that a process can still
1890 * access /proc/self/fd after it has executed a setuid().
1892 static int proc_fd_permission(struct inode *inode, int mask)
1896 rv = generic_permission(inode, mask, NULL);
1899 if (task_pid(current) == proc_pid(inode))
1905 * proc directories can do almost nothing..
1907 static const struct inode_operations proc_fd_inode_operations = {
1908 .lookup = proc_lookupfd,
1909 .permission = proc_fd_permission,
1910 .setattr = proc_setattr,
1913 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1914 struct dentry *dentry, struct task_struct *task, const void *ptr)
1916 unsigned fd = *(unsigned *)ptr;
1917 struct inode *inode;
1918 struct proc_inode *ei;
1919 struct dentry *error = ERR_PTR(-ENOENT);
1921 inode = proc_pid_make_inode(dir->i_sb, task);
1926 inode->i_mode = S_IFREG | S_IRUSR;
1927 inode->i_fop = &proc_fdinfo_file_operations;
1928 dentry->d_op = &tid_fd_dentry_operations;
1929 d_add(dentry, inode);
1930 /* Close the race of the process dying before we return the dentry */
1931 if (tid_fd_revalidate(dentry, NULL))
1938 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1939 struct dentry *dentry,
1940 struct nameidata *nd)
1942 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1945 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1947 return proc_readfd_common(filp, dirent, filldir,
1948 proc_fdinfo_instantiate);
1951 static const struct file_operations proc_fdinfo_operations = {
1952 .read = generic_read_dir,
1953 .readdir = proc_readfdinfo,
1957 * proc directories can do almost nothing..
1959 static const struct inode_operations proc_fdinfo_inode_operations = {
1960 .lookup = proc_lookupfdinfo,
1961 .setattr = proc_setattr,
1965 static struct dentry *proc_pident_instantiate(struct inode *dir,
1966 struct dentry *dentry, struct task_struct *task, const void *ptr)
1968 const struct pid_entry *p = ptr;
1969 struct inode *inode;
1970 struct proc_inode *ei;
1971 struct dentry *error = ERR_PTR(-ENOENT);
1973 inode = proc_pid_make_inode(dir->i_sb, task);
1978 inode->i_mode = p->mode;
1979 if (S_ISDIR(inode->i_mode))
1980 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1982 inode->i_op = p->iop;
1984 inode->i_fop = p->fop;
1986 dentry->d_op = &pid_dentry_operations;
1987 d_add(dentry, inode);
1988 /* Close the race of the process dying before we return the dentry */
1989 if (pid_revalidate(dentry, NULL))
1995 static struct dentry *proc_pident_lookup(struct inode *dir,
1996 struct dentry *dentry,
1997 const struct pid_entry *ents,
2000 struct dentry *error;
2001 struct task_struct *task = get_proc_task(dir);
2002 const struct pid_entry *p, *last;
2004 error = ERR_PTR(-ENOENT);
2010 * Yes, it does not scale. And it should not. Don't add
2011 * new entries into /proc/<tgid>/ without very good reasons.
2013 last = &ents[nents - 1];
2014 for (p = ents; p <= last; p++) {
2015 if (p->len != dentry->d_name.len)
2017 if (!memcmp(dentry->d_name.name, p->name, p->len))
2023 error = proc_pident_instantiate(dir, dentry, task, p);
2025 put_task_struct(task);
2030 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2031 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2033 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2034 proc_pident_instantiate, task, p);
2037 static int proc_pident_readdir(struct file *filp,
2038 void *dirent, filldir_t filldir,
2039 const struct pid_entry *ents, unsigned int nents)
2042 struct dentry *dentry = filp->f_path.dentry;
2043 struct inode *inode = dentry->d_inode;
2044 struct task_struct *task = get_proc_task(inode);
2045 const struct pid_entry *p, *last;
2058 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2064 ino = parent_ino(dentry);
2065 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2077 last = &ents[nents - 1];
2079 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2088 put_task_struct(task);
2093 #ifdef CONFIG_SECURITY
2094 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2095 size_t count, loff_t *ppos)
2097 struct inode * inode = file->f_path.dentry->d_inode;
2100 struct task_struct *task = get_proc_task(inode);
2105 length = security_getprocattr(task,
2106 (char*)file->f_path.dentry->d_name.name,
2108 put_task_struct(task);
2110 length = simple_read_from_buffer(buf, count, ppos, p, length);
2115 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2116 size_t count, loff_t *ppos)
2118 struct inode * inode = file->f_path.dentry->d_inode;
2121 struct task_struct *task = get_proc_task(inode);
2126 if (count > PAGE_SIZE)
2129 /* No partial writes. */
2135 page = (char*)__get_free_page(GFP_TEMPORARY);
2140 if (copy_from_user(page, buf, count))
2143 /* Guard against adverse ptrace interaction */
2144 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2148 length = security_setprocattr(task,
2149 (char*)file->f_path.dentry->d_name.name,
2150 (void*)page, count);
2151 mutex_unlock(&task->cred_guard_mutex);
2153 free_page((unsigned long) page);
2155 put_task_struct(task);
2160 static const struct file_operations proc_pid_attr_operations = {
2161 .read = proc_pid_attr_read,
2162 .write = proc_pid_attr_write,
2165 static const struct pid_entry attr_dir_stuff[] = {
2166 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2167 REG("prev", S_IRUGO, proc_pid_attr_operations),
2168 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2169 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2170 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2171 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2174 static int proc_attr_dir_readdir(struct file * filp,
2175 void * dirent, filldir_t filldir)
2177 return proc_pident_readdir(filp,dirent,filldir,
2178 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2181 static const struct file_operations proc_attr_dir_operations = {
2182 .read = generic_read_dir,
2183 .readdir = proc_attr_dir_readdir,
2186 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2187 struct dentry *dentry, struct nameidata *nd)
2189 return proc_pident_lookup(dir, dentry,
2190 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2193 static const struct inode_operations proc_attr_dir_inode_operations = {
2194 .lookup = proc_attr_dir_lookup,
2195 .getattr = pid_getattr,
2196 .setattr = proc_setattr,
2201 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2202 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2203 size_t count, loff_t *ppos)
2205 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2206 struct mm_struct *mm;
2207 char buffer[PROC_NUMBUF];
2215 mm = get_task_mm(task);
2217 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2218 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2219 MMF_DUMP_FILTER_SHIFT));
2221 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2224 put_task_struct(task);
2229 static ssize_t proc_coredump_filter_write(struct file *file,
2230 const char __user *buf,
2234 struct task_struct *task;
2235 struct mm_struct *mm;
2236 char buffer[PROC_NUMBUF], *end;
2243 memset(buffer, 0, sizeof(buffer));
2244 if (count > sizeof(buffer) - 1)
2245 count = sizeof(buffer) - 1;
2246 if (copy_from_user(buffer, buf, count))
2250 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2253 if (end - buffer == 0)
2257 task = get_proc_task(file->f_dentry->d_inode);
2262 mm = get_task_mm(task);
2266 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2268 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2270 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2275 put_task_struct(task);
2280 static const struct file_operations proc_coredump_filter_operations = {
2281 .read = proc_coredump_filter_read,
2282 .write = proc_coredump_filter_write,
2289 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2292 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2293 pid_t tgid = task_tgid_nr_ns(current, ns);
2294 char tmp[PROC_NUMBUF];
2297 sprintf(tmp, "%d", tgid);
2298 return vfs_readlink(dentry,buffer,buflen,tmp);
2301 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2303 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2304 pid_t tgid = task_tgid_nr_ns(current, ns);
2305 char tmp[PROC_NUMBUF];
2307 return ERR_PTR(-ENOENT);
2308 sprintf(tmp, "%d", task_tgid_nr_ns(current, ns));
2309 return ERR_PTR(vfs_follow_link(nd,tmp));
2312 static const struct inode_operations proc_self_inode_operations = {
2313 .readlink = proc_self_readlink,
2314 .follow_link = proc_self_follow_link,
2320 * These are the directory entries in the root directory of /proc
2321 * that properly belong to the /proc filesystem, as they describe
2322 * describe something that is process related.
2324 static const struct pid_entry proc_base_stuff[] = {
2325 NOD("self", S_IFLNK|S_IRWXUGO,
2326 &proc_self_inode_operations, NULL, {}),
2330 * Exceptional case: normally we are not allowed to unhash a busy
2331 * directory. In this case, however, we can do it - no aliasing problems
2332 * due to the way we treat inodes.
2334 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2336 struct inode *inode = dentry->d_inode;
2337 struct task_struct *task = get_proc_task(inode);
2339 put_task_struct(task);
2346 static const struct dentry_operations proc_base_dentry_operations =
2348 .d_revalidate = proc_base_revalidate,
2349 .d_delete = pid_delete_dentry,
2352 static struct dentry *proc_base_instantiate(struct inode *dir,
2353 struct dentry *dentry, struct task_struct *task, const void *ptr)
2355 const struct pid_entry *p = ptr;
2356 struct inode *inode;
2357 struct proc_inode *ei;
2358 struct dentry *error = ERR_PTR(-EINVAL);
2360 /* Allocate the inode */
2361 error = ERR_PTR(-ENOMEM);
2362 inode = new_inode(dir->i_sb);
2366 /* Initialize the inode */
2368 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2371 * grab the reference to the task.
2373 ei->pid = get_task_pid(task, PIDTYPE_PID);
2377 inode->i_mode = p->mode;
2378 if (S_ISDIR(inode->i_mode))
2380 if (S_ISLNK(inode->i_mode))
2383 inode->i_op = p->iop;
2385 inode->i_fop = p->fop;
2387 dentry->d_op = &proc_base_dentry_operations;
2388 d_add(dentry, inode);
2397 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2399 struct dentry *error;
2400 struct task_struct *task = get_proc_task(dir);
2401 const struct pid_entry *p, *last;
2403 error = ERR_PTR(-ENOENT);
2408 /* Lookup the directory entry */
2409 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2410 for (p = proc_base_stuff; p <= last; p++) {
2411 if (p->len != dentry->d_name.len)
2413 if (!memcmp(dentry->d_name.name, p->name, p->len))
2419 error = proc_base_instantiate(dir, dentry, task, p);
2422 put_task_struct(task);
2427 static int proc_base_fill_cache(struct file *filp, void *dirent,
2428 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2430 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2431 proc_base_instantiate, task, p);
2434 #ifdef CONFIG_TASK_IO_ACCOUNTING
2435 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2437 struct task_io_accounting acct = task->ioac;
2438 unsigned long flags;
2440 if (whole && lock_task_sighand(task, &flags)) {
2441 struct task_struct *t = task;
2443 task_io_accounting_add(&acct, &task->signal->ioac);
2444 while_each_thread(task, t)
2445 task_io_accounting_add(&acct, &t->ioac);
2447 unlock_task_sighand(task, &flags);
2449 return sprintf(buffer,
2454 "read_bytes: %llu\n"
2455 "write_bytes: %llu\n"
2456 "cancelled_write_bytes: %llu\n",
2457 (unsigned long long)acct.rchar,
2458 (unsigned long long)acct.wchar,
2459 (unsigned long long)acct.syscr,
2460 (unsigned long long)acct.syscw,
2461 (unsigned long long)acct.read_bytes,
2462 (unsigned long long)acct.write_bytes,
2463 (unsigned long long)acct.cancelled_write_bytes);
2466 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2468 return do_io_accounting(task, buffer, 0);
2471 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2473 return do_io_accounting(task, buffer, 1);
2475 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2477 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2478 struct pid *pid, struct task_struct *task)
2480 seq_printf(m, "%08x\n", task->personality);
2487 static const struct file_operations proc_task_operations;
2488 static const struct inode_operations proc_task_inode_operations;
2490 static const struct pid_entry tgid_base_stuff[] = {
2491 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2492 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2493 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2495 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2497 REG("environ", S_IRUSR, proc_environ_operations),
2498 INF("auxv", S_IRUSR, proc_pid_auxv),
2499 ONE("status", S_IRUGO, proc_pid_status),
2500 ONE("personality", S_IRUSR, proc_pid_personality),
2501 INF("limits", S_IRUSR, proc_pid_limits),
2502 #ifdef CONFIG_SCHED_DEBUG
2503 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2505 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2506 INF("syscall", S_IRUSR, proc_pid_syscall),
2508 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2509 ONE("stat", S_IRUGO, proc_tgid_stat),
2510 ONE("statm", S_IRUGO, proc_pid_statm),
2511 REG("maps", S_IRUGO, proc_maps_operations),
2513 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2515 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2516 LNK("cwd", proc_cwd_link),
2517 LNK("root", proc_root_link),
2518 LNK("exe", proc_exe_link),
2519 REG("mounts", S_IRUGO, proc_mounts_operations),
2520 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2521 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2522 #ifdef CONFIG_PROC_PAGE_MONITOR
2523 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2524 REG("smaps", S_IRUGO, proc_smaps_operations),
2525 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2527 #ifdef CONFIG_SECURITY
2528 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2530 #ifdef CONFIG_KALLSYMS
2531 INF("wchan", S_IRUGO, proc_pid_wchan),
2533 #ifdef CONFIG_STACKTRACE
2534 ONE("stack", S_IRUSR, proc_pid_stack),
2536 #ifdef CONFIG_SCHEDSTATS
2537 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2539 #ifdef CONFIG_LATENCYTOP
2540 REG("latency", S_IRUGO, proc_lstats_operations),
2542 #ifdef CONFIG_PROC_PID_CPUSET
2543 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2545 #ifdef CONFIG_CGROUPS
2546 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2548 INF("oom_score", S_IRUGO, proc_oom_score),
2549 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2550 #ifdef CONFIG_AUDITSYSCALL
2551 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2552 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2554 #ifdef CONFIG_FAULT_INJECTION
2555 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2557 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2558 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2560 #ifdef CONFIG_TASK_IO_ACCOUNTING
2561 INF("io", S_IRUGO, proc_tgid_io_accounting),
2565 static int proc_tgid_base_readdir(struct file * filp,
2566 void * dirent, filldir_t filldir)
2568 return proc_pident_readdir(filp,dirent,filldir,
2569 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2572 static const struct file_operations proc_tgid_base_operations = {
2573 .read = generic_read_dir,
2574 .readdir = proc_tgid_base_readdir,
2577 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2578 return proc_pident_lookup(dir, dentry,
2579 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2582 static const struct inode_operations proc_tgid_base_inode_operations = {
2583 .lookup = proc_tgid_base_lookup,
2584 .getattr = pid_getattr,
2585 .setattr = proc_setattr,
2588 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2590 struct dentry *dentry, *leader, *dir;
2591 char buf[PROC_NUMBUF];
2595 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2596 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2598 if (!(current->flags & PF_EXITING))
2599 shrink_dcache_parent(dentry);
2608 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2609 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2614 name.len = strlen(name.name);
2615 dir = d_hash_and_lookup(leader, &name);
2617 goto out_put_leader;
2620 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2621 dentry = d_hash_and_lookup(dir, &name);
2623 shrink_dcache_parent(dentry);
2636 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2637 * @task: task that should be flushed.
2639 * When flushing dentries from proc, one needs to flush them from global
2640 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2641 * in. This call is supposed to do all of this job.
2643 * Looks in the dcache for
2645 * /proc/@tgid/task/@pid
2646 * if either directory is present flushes it and all of it'ts children
2649 * It is safe and reasonable to cache /proc entries for a task until
2650 * that task exits. After that they just clog up the dcache with
2651 * useless entries, possibly causing useful dcache entries to be
2652 * flushed instead. This routine is proved to flush those useless
2653 * dcache entries at process exit time.
2655 * NOTE: This routine is just an optimization so it does not guarantee
2656 * that no dcache entries will exist at process exit time it
2657 * just makes it very unlikely that any will persist.
2660 void proc_flush_task(struct task_struct *task)
2663 struct pid *pid, *tgid = NULL;
2666 pid = task_pid(task);
2667 if (thread_group_leader(task))
2668 tgid = task_tgid(task);
2670 for (i = 0; i <= pid->level; i++) {
2671 upid = &pid->numbers[i];
2672 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2673 tgid ? tgid->numbers[i].nr : 0);
2676 upid = &pid->numbers[pid->level];
2678 pid_ns_release_proc(upid->ns);
2681 static struct dentry *proc_pid_instantiate(struct inode *dir,
2682 struct dentry * dentry,
2683 struct task_struct *task, const void *ptr)
2685 struct dentry *error = ERR_PTR(-ENOENT);
2686 struct inode *inode;
2688 inode = proc_pid_make_inode(dir->i_sb, task);
2692 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2693 inode->i_op = &proc_tgid_base_inode_operations;
2694 inode->i_fop = &proc_tgid_base_operations;
2695 inode->i_flags|=S_IMMUTABLE;
2697 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2698 ARRAY_SIZE(tgid_base_stuff));
2700 dentry->d_op = &pid_dentry_operations;
2702 d_add(dentry, inode);
2703 /* Close the race of the process dying before we return the dentry */
2704 if (pid_revalidate(dentry, NULL))
2710 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2712 struct dentry *result = ERR_PTR(-ENOENT);
2713 struct task_struct *task;
2715 struct pid_namespace *ns;
2717 result = proc_base_lookup(dir, dentry);
2718 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2721 tgid = name_to_int(dentry);
2725 ns = dentry->d_sb->s_fs_info;
2727 task = find_task_by_pid_ns(tgid, ns);
2729 get_task_struct(task);
2734 result = proc_pid_instantiate(dir, dentry, task, NULL);
2735 put_task_struct(task);
2741 * Find the first task with tgid >= tgid
2746 struct task_struct *task;
2748 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2753 put_task_struct(iter.task);
2757 pid = find_ge_pid(iter.tgid, ns);
2759 iter.tgid = pid_nr_ns(pid, ns);
2760 iter.task = pid_task(pid, PIDTYPE_PID);
2761 /* What we to know is if the pid we have find is the
2762 * pid of a thread_group_leader. Testing for task
2763 * being a thread_group_leader is the obvious thing
2764 * todo but there is a window when it fails, due to
2765 * the pid transfer logic in de_thread.
2767 * So we perform the straight forward test of seeing
2768 * if the pid we have found is the pid of a thread
2769 * group leader, and don't worry if the task we have
2770 * found doesn't happen to be a thread group leader.
2771 * As we don't care in the case of readdir.
2773 if (!iter.task || !has_group_leader_pid(iter.task)) {
2777 get_task_struct(iter.task);
2783 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2785 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2786 struct tgid_iter iter)
2788 char name[PROC_NUMBUF];
2789 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2790 return proc_fill_cache(filp, dirent, filldir, name, len,
2791 proc_pid_instantiate, iter.task, NULL);
2794 /* for the /proc/ directory itself, after non-process stuff has been done */
2795 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2797 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2798 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2799 struct tgid_iter iter;
2800 struct pid_namespace *ns;
2805 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2806 const struct pid_entry *p = &proc_base_stuff[nr];
2807 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2811 ns = filp->f_dentry->d_sb->s_fs_info;
2813 iter.tgid = filp->f_pos - TGID_OFFSET;
2814 for (iter = next_tgid(ns, iter);
2816 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2817 filp->f_pos = iter.tgid + TGID_OFFSET;
2818 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2819 put_task_struct(iter.task);
2823 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2825 put_task_struct(reaper);
2833 static const struct pid_entry tid_base_stuff[] = {
2834 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2835 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2836 REG("environ", S_IRUSR, proc_environ_operations),
2837 INF("auxv", S_IRUSR, proc_pid_auxv),
2838 ONE("status", S_IRUGO, proc_pid_status),
2839 ONE("personality", S_IRUSR, proc_pid_personality),
2840 INF("limits", S_IRUSR, proc_pid_limits),
2841 #ifdef CONFIG_SCHED_DEBUG
2842 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2844 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2845 INF("syscall", S_IRUSR, proc_pid_syscall),
2847 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2848 ONE("stat", S_IRUGO, proc_tid_stat),
2849 ONE("statm", S_IRUGO, proc_pid_statm),
2850 REG("maps", S_IRUGO, proc_maps_operations),
2852 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2854 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2855 LNK("cwd", proc_cwd_link),
2856 LNK("root", proc_root_link),
2857 LNK("exe", proc_exe_link),
2858 REG("mounts", S_IRUGO, proc_mounts_operations),
2859 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2860 #ifdef CONFIG_PROC_PAGE_MONITOR
2861 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2862 REG("smaps", S_IRUGO, proc_smaps_operations),
2863 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2865 #ifdef CONFIG_SECURITY
2866 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2868 #ifdef CONFIG_KALLSYMS
2869 INF("wchan", S_IRUGO, proc_pid_wchan),
2871 #ifdef CONFIG_STACKTRACE
2872 ONE("stack", S_IRUSR, proc_pid_stack),
2874 #ifdef CONFIG_SCHEDSTATS
2875 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2877 #ifdef CONFIG_LATENCYTOP
2878 REG("latency", S_IRUGO, proc_lstats_operations),
2880 #ifdef CONFIG_PROC_PID_CPUSET
2881 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2883 #ifdef CONFIG_CGROUPS
2884 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2886 INF("oom_score", S_IRUGO, proc_oom_score),
2887 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2888 #ifdef CONFIG_AUDITSYSCALL
2889 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2890 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2892 #ifdef CONFIG_FAULT_INJECTION
2893 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2895 #ifdef CONFIG_TASK_IO_ACCOUNTING
2896 INF("io", S_IRUGO, proc_tid_io_accounting),
2900 static int proc_tid_base_readdir(struct file * filp,
2901 void * dirent, filldir_t filldir)
2903 return proc_pident_readdir(filp,dirent,filldir,
2904 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2907 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2908 return proc_pident_lookup(dir, dentry,
2909 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2912 static const struct file_operations proc_tid_base_operations = {
2913 .read = generic_read_dir,
2914 .readdir = proc_tid_base_readdir,
2917 static const struct inode_operations proc_tid_base_inode_operations = {
2918 .lookup = proc_tid_base_lookup,
2919 .getattr = pid_getattr,
2920 .setattr = proc_setattr,
2923 static struct dentry *proc_task_instantiate(struct inode *dir,
2924 struct dentry *dentry, struct task_struct *task, const void *ptr)
2926 struct dentry *error = ERR_PTR(-ENOENT);
2927 struct inode *inode;
2928 inode = proc_pid_make_inode(dir->i_sb, task);
2932 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2933 inode->i_op = &proc_tid_base_inode_operations;
2934 inode->i_fop = &proc_tid_base_operations;
2935 inode->i_flags|=S_IMMUTABLE;
2937 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
2938 ARRAY_SIZE(tid_base_stuff));
2940 dentry->d_op = &pid_dentry_operations;
2942 d_add(dentry, inode);
2943 /* Close the race of the process dying before we return the dentry */
2944 if (pid_revalidate(dentry, NULL))
2950 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2952 struct dentry *result = ERR_PTR(-ENOENT);
2953 struct task_struct *task;
2954 struct task_struct *leader = get_proc_task(dir);
2956 struct pid_namespace *ns;
2961 tid = name_to_int(dentry);
2965 ns = dentry->d_sb->s_fs_info;
2967 task = find_task_by_pid_ns(tid, ns);
2969 get_task_struct(task);
2973 if (!same_thread_group(leader, task))
2976 result = proc_task_instantiate(dir, dentry, task, NULL);
2978 put_task_struct(task);
2980 put_task_struct(leader);
2986 * Find the first tid of a thread group to return to user space.
2988 * Usually this is just the thread group leader, but if the users
2989 * buffer was too small or there was a seek into the middle of the
2990 * directory we have more work todo.
2992 * In the case of a short read we start with find_task_by_pid.
2994 * In the case of a seek we start with the leader and walk nr
2997 static struct task_struct *first_tid(struct task_struct *leader,
2998 int tid, int nr, struct pid_namespace *ns)
3000 struct task_struct *pos;
3003 /* Attempt to start with the pid of a thread */
3004 if (tid && (nr > 0)) {
3005 pos = find_task_by_pid_ns(tid, ns);
3006 if (pos && (pos->group_leader == leader))
3010 /* If nr exceeds the number of threads there is nothing todo */
3012 if (nr && nr >= get_nr_threads(leader))
3015 /* If we haven't found our starting place yet start
3016 * with the leader and walk nr threads forward.
3018 for (pos = leader; nr > 0; --nr) {
3019 pos = next_thread(pos);
3020 if (pos == leader) {
3026 get_task_struct(pos);
3033 * Find the next thread in the thread list.
3034 * Return NULL if there is an error or no next thread.
3036 * The reference to the input task_struct is released.
3038 static struct task_struct *next_tid(struct task_struct *start)
3040 struct task_struct *pos = NULL;
3042 if (pid_alive(start)) {
3043 pos = next_thread(start);
3044 if (thread_group_leader(pos))
3047 get_task_struct(pos);
3050 put_task_struct(start);
3054 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3055 struct task_struct *task, int tid)
3057 char name[PROC_NUMBUF];
3058 int len = snprintf(name, sizeof(name), "%d", tid);
3059 return proc_fill_cache(filp, dirent, filldir, name, len,
3060 proc_task_instantiate, task, NULL);
3063 /* for the /proc/TGID/task/ directories */
3064 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3066 struct dentry *dentry = filp->f_path.dentry;
3067 struct inode *inode = dentry->d_inode;
3068 struct task_struct *leader = NULL;
3069 struct task_struct *task;
3070 int retval = -ENOENT;
3073 struct pid_namespace *ns;
3075 task = get_proc_task(inode);
3079 if (pid_alive(task)) {
3080 leader = task->group_leader;
3081 get_task_struct(leader);
3084 put_task_struct(task);
3089 switch ((unsigned long)filp->f_pos) {
3092 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3097 ino = parent_ino(dentry);
3098 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3104 /* f_version caches the tgid value that the last readdir call couldn't
3105 * return. lseek aka telldir automagically resets f_version to 0.
3107 ns = filp->f_dentry->d_sb->s_fs_info;
3108 tid = (int)filp->f_version;
3109 filp->f_version = 0;
3110 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3112 task = next_tid(task), filp->f_pos++) {
3113 tid = task_pid_nr_ns(task, ns);
3114 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3115 /* returning this tgid failed, save it as the first
3116 * pid for the next readir call */
3117 filp->f_version = (u64)tid;
3118 put_task_struct(task);
3123 put_task_struct(leader);
3128 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3130 struct inode *inode = dentry->d_inode;
3131 struct task_struct *p = get_proc_task(inode);
3132 generic_fillattr(inode, stat);
3135 stat->nlink += get_nr_threads(p);
3142 static const struct inode_operations proc_task_inode_operations = {
3143 .lookup = proc_task_lookup,
3144 .getattr = proc_task_getattr,
3145 .setattr = proc_setattr,
3148 static const struct file_operations proc_task_operations = {
3149 .read = generic_read_dir,
3150 .readdir = proc_task_readdir,