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>
84 #include <linux/slab.h>
88 * Implementing inode permission operations in /proc is almost
89 * certainly an error. Permission checks need to happen during
90 * each system call not at open time. The reason is that most of
91 * what we wish to check for permissions in /proc varies at runtime.
93 * The classic example of a problem is opening file descriptors
94 * in /proc for a task before it execs a suid executable.
101 const struct inode_operations *iop;
102 const struct file_operations *fop;
106 #define NOD(NAME, MODE, IOP, FOP, OP) { \
108 .len = sizeof(NAME) - 1, \
115 #define DIR(NAME, MODE, iops, fops) \
116 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
117 #define LNK(NAME, get_link) \
118 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
119 &proc_pid_link_inode_operations, NULL, \
120 { .proc_get_link = get_link } )
121 #define REG(NAME, MODE, fops) \
122 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
123 #define INF(NAME, MODE, read) \
124 NOD(NAME, (S_IFREG|(MODE)), \
125 NULL, &proc_info_file_operations, \
126 { .proc_read = read } )
127 #define ONE(NAME, MODE, show) \
128 NOD(NAME, (S_IFREG|(MODE)), \
129 NULL, &proc_single_file_operations, \
130 { .proc_show = show } )
133 * Count the number of hardlinks for the pid_entry table, excluding the .
136 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 for (i = 0; i < n; ++i) {
144 if (S_ISDIR(entries[i].mode))
151 static int get_fs_path(struct task_struct *task, struct path *path, bool root)
153 struct fs_struct *fs;
154 int result = -ENOENT;
159 read_lock(&fs->lock);
160 *path = root ? fs->root : fs->pwd;
162 read_unlock(&fs->lock);
169 static int proc_cwd_link(struct inode *inode, struct path *path)
171 struct task_struct *task = get_proc_task(inode);
172 int result = -ENOENT;
175 result = get_fs_path(task, path, 0);
176 put_task_struct(task);
181 static int proc_root_link(struct inode *inode, struct path *path)
183 struct task_struct *task = get_proc_task(inode);
184 int result = -ENOENT;
187 result = get_fs_path(task, path, 1);
188 put_task_struct(task);
194 * Return zero if current may access user memory in @task, -error if not.
196 static int check_mem_permission(struct task_struct *task)
199 * A task can always look at itself, in case it chooses
200 * to use system calls instead of load instructions.
206 * If current is actively ptrace'ing, and would also be
207 * permitted to freshly attach with ptrace now, permit it.
209 if (task_is_stopped_or_traced(task)) {
212 match = (tracehook_tracer_task(task) == current);
214 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
219 * Noone else is allowed.
224 struct mm_struct *mm_for_maps(struct task_struct *task)
226 struct mm_struct *mm;
228 if (mutex_lock_killable(&task->cred_guard_mutex))
231 mm = get_task_mm(task);
232 if (mm && mm != current->mm &&
233 !ptrace_may_access(task, PTRACE_MODE_READ)) {
237 mutex_unlock(&task->cred_guard_mutex);
242 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
246 struct mm_struct *mm = get_task_mm(task);
250 goto out_mm; /* Shh! No looking before we're done */
252 len = mm->arg_end - mm->arg_start;
257 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
259 // If the nul at the end of args has been overwritten, then
260 // assume application is using setproctitle(3).
261 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
262 len = strnlen(buffer, res);
266 len = mm->env_end - mm->env_start;
267 if (len > PAGE_SIZE - res)
268 len = PAGE_SIZE - res;
269 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
270 res = strnlen(buffer, res);
279 static int proc_pid_auxv(struct task_struct *task, char *buffer)
282 struct mm_struct *mm = get_task_mm(task);
284 unsigned int nwords = 0;
287 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
288 res = nwords * sizeof(mm->saved_auxv[0]);
291 memcpy(buffer, mm->saved_auxv, res);
298 #ifdef CONFIG_KALLSYMS
300 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
301 * Returns the resolved symbol. If that fails, simply return the address.
303 static int proc_pid_wchan(struct task_struct *task, char *buffer)
306 char symname[KSYM_NAME_LEN];
308 wchan = get_wchan(task);
310 if (lookup_symbol_name(wchan, symname) < 0)
311 if (!ptrace_may_access(task, PTRACE_MODE_READ))
314 return sprintf(buffer, "%lu", wchan);
316 return sprintf(buffer, "%s", symname);
318 #endif /* CONFIG_KALLSYMS */
320 #ifdef CONFIG_STACKTRACE
322 #define MAX_STACK_TRACE_DEPTH 64
324 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
325 struct pid *pid, struct task_struct *task)
327 struct stack_trace trace;
328 unsigned long *entries;
331 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
335 trace.nr_entries = 0;
336 trace.max_entries = MAX_STACK_TRACE_DEPTH;
337 trace.entries = entries;
339 save_stack_trace_tsk(task, &trace);
341 for (i = 0; i < trace.nr_entries; i++) {
342 seq_printf(m, "[<%p>] %pS\n",
343 (void *)entries[i], (void *)entries[i]);
351 #ifdef CONFIG_SCHEDSTATS
353 * Provides /proc/PID/schedstat
355 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
357 return sprintf(buffer, "%llu %llu %lu\n",
358 (unsigned long long)task->se.sum_exec_runtime,
359 (unsigned long long)task->sched_info.run_delay,
360 task->sched_info.pcount);
364 #ifdef CONFIG_LATENCYTOP
365 static int lstats_show_proc(struct seq_file *m, void *v)
368 struct inode *inode = m->private;
369 struct task_struct *task = get_proc_task(inode);
373 seq_puts(m, "Latency Top version : v0.1\n");
374 for (i = 0; i < 32; i++) {
375 if (task->latency_record[i].backtrace[0]) {
377 seq_printf(m, "%i %li %li ",
378 task->latency_record[i].count,
379 task->latency_record[i].time,
380 task->latency_record[i].max);
381 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
382 char sym[KSYM_SYMBOL_LEN];
384 if (!task->latency_record[i].backtrace[q])
386 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
388 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
389 c = strchr(sym, '+');
392 seq_printf(m, "%s ", sym);
398 put_task_struct(task);
402 static int lstats_open(struct inode *inode, struct file *file)
404 return single_open(file, lstats_show_proc, inode);
407 static ssize_t lstats_write(struct file *file, const char __user *buf,
408 size_t count, loff_t *offs)
410 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
414 clear_all_latency_tracing(task);
415 put_task_struct(task);
420 static const struct file_operations proc_lstats_operations = {
423 .write = lstats_write,
425 .release = single_release,
430 /* The badness from the OOM killer */
431 unsigned long badness(struct task_struct *p, unsigned long uptime);
432 static int proc_oom_score(struct task_struct *task, char *buffer)
434 unsigned long points = 0;
435 struct timespec uptime;
437 do_posix_clock_monotonic_gettime(&uptime);
438 read_lock(&tasklist_lock);
440 points = badness(task, uptime.tv_sec);
441 read_unlock(&tasklist_lock);
442 return sprintf(buffer, "%lu\n", points);
450 static const struct limit_names lnames[RLIM_NLIMITS] = {
451 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
452 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
453 [RLIMIT_DATA] = {"Max data size", "bytes"},
454 [RLIMIT_STACK] = {"Max stack size", "bytes"},
455 [RLIMIT_CORE] = {"Max core file size", "bytes"},
456 [RLIMIT_RSS] = {"Max resident set", "bytes"},
457 [RLIMIT_NPROC] = {"Max processes", "processes"},
458 [RLIMIT_NOFILE] = {"Max open files", "files"},
459 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
460 [RLIMIT_AS] = {"Max address space", "bytes"},
461 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
462 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
463 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
464 [RLIMIT_NICE] = {"Max nice priority", NULL},
465 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
466 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
469 /* Display limits for a process */
470 static int proc_pid_limits(struct task_struct *task, char *buffer)
475 char *bufptr = buffer;
477 struct rlimit rlim[RLIM_NLIMITS];
479 if (!lock_task_sighand(task, &flags))
481 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
482 unlock_task_sighand(task, &flags);
485 * print the file header
487 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
488 "Limit", "Soft Limit", "Hard Limit", "Units");
490 for (i = 0; i < RLIM_NLIMITS; i++) {
491 if (rlim[i].rlim_cur == RLIM_INFINITY)
492 count += sprintf(&bufptr[count], "%-25s %-20s ",
493 lnames[i].name, "unlimited");
495 count += sprintf(&bufptr[count], "%-25s %-20lu ",
496 lnames[i].name, rlim[i].rlim_cur);
498 if (rlim[i].rlim_max == RLIM_INFINITY)
499 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
501 count += sprintf(&bufptr[count], "%-20lu ",
505 count += sprintf(&bufptr[count], "%-10s\n",
508 count += sprintf(&bufptr[count], "\n");
514 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
515 static int proc_pid_syscall(struct task_struct *task, char *buffer)
518 unsigned long args[6], sp, pc;
520 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
521 return sprintf(buffer, "running\n");
524 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
526 return sprintf(buffer,
527 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
529 args[0], args[1], args[2], args[3], args[4], args[5],
532 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
534 /************************************************************************/
535 /* Here the fs part begins */
536 /************************************************************************/
538 /* permission checks */
539 static int proc_fd_access_allowed(struct inode *inode)
541 struct task_struct *task;
543 /* Allow access to a task's file descriptors if it is us or we
544 * may use ptrace attach to the process and find out that
547 task = get_proc_task(inode);
549 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
550 put_task_struct(task);
555 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
558 struct inode *inode = dentry->d_inode;
560 if (attr->ia_valid & ATTR_MODE)
563 error = inode_change_ok(inode, attr);
567 if ((attr->ia_valid & ATTR_SIZE) &&
568 attr->ia_size != i_size_read(inode)) {
569 error = vmtruncate(inode, attr->ia_size);
574 setattr_copy(inode, attr);
575 mark_inode_dirty(inode);
579 static const struct inode_operations proc_def_inode_operations = {
580 .setattr = proc_setattr,
583 static int mounts_open_common(struct inode *inode, struct file *file,
584 const struct seq_operations *op)
586 struct task_struct *task = get_proc_task(inode);
588 struct mnt_namespace *ns = NULL;
590 struct proc_mounts *p;
595 nsp = task_nsproxy(task);
602 if (ns && get_fs_path(task, &root, 1) == 0)
604 put_task_struct(task);
613 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
617 file->private_data = &p->m;
618 ret = seq_open(file, op);
625 p->event = ns->event;
639 static int mounts_release(struct inode *inode, struct file *file)
641 struct proc_mounts *p = file->private_data;
644 return seq_release(inode, file);
647 static unsigned mounts_poll(struct file *file, poll_table *wait)
649 struct proc_mounts *p = file->private_data;
650 unsigned res = POLLIN | POLLRDNORM;
652 poll_wait(file, &p->ns->poll, wait);
653 if (mnt_had_events(p))
654 res |= POLLERR | POLLPRI;
659 static int mounts_open(struct inode *inode, struct file *file)
661 return mounts_open_common(inode, file, &mounts_op);
664 static const struct file_operations proc_mounts_operations = {
668 .release = mounts_release,
672 static int mountinfo_open(struct inode *inode, struct file *file)
674 return mounts_open_common(inode, file, &mountinfo_op);
677 static const struct file_operations proc_mountinfo_operations = {
678 .open = mountinfo_open,
681 .release = mounts_release,
685 static int mountstats_open(struct inode *inode, struct file *file)
687 return mounts_open_common(inode, file, &mountstats_op);
690 static const struct file_operations proc_mountstats_operations = {
691 .open = mountstats_open,
694 .release = mounts_release,
697 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
699 static ssize_t proc_info_read(struct file * file, char __user * buf,
700 size_t count, loff_t *ppos)
702 struct inode * inode = file->f_path.dentry->d_inode;
705 struct task_struct *task = get_proc_task(inode);
711 if (count > PROC_BLOCK_SIZE)
712 count = PROC_BLOCK_SIZE;
715 if (!(page = __get_free_page(GFP_TEMPORARY)))
718 length = PROC_I(inode)->op.proc_read(task, (char*)page);
721 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
724 put_task_struct(task);
729 static const struct file_operations proc_info_file_operations = {
730 .read = proc_info_read,
731 .llseek = generic_file_llseek,
734 static int proc_single_show(struct seq_file *m, void *v)
736 struct inode *inode = m->private;
737 struct pid_namespace *ns;
739 struct task_struct *task;
742 ns = inode->i_sb->s_fs_info;
743 pid = proc_pid(inode);
744 task = get_pid_task(pid, PIDTYPE_PID);
748 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
750 put_task_struct(task);
754 static int proc_single_open(struct inode *inode, struct file *filp)
757 ret = single_open(filp, proc_single_show, NULL);
759 struct seq_file *m = filp->private_data;
766 static const struct file_operations proc_single_file_operations = {
767 .open = proc_single_open,
770 .release = single_release,
773 static int mem_open(struct inode* inode, struct file* file)
775 file->private_data = (void*)((long)current->self_exec_id);
779 static ssize_t mem_read(struct file * file, char __user * buf,
780 size_t count, loff_t *ppos)
782 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
784 unsigned long src = *ppos;
786 struct mm_struct *mm;
791 if (check_mem_permission(task))
795 page = (char *)__get_free_page(GFP_TEMPORARY);
801 mm = get_task_mm(task);
807 if (file->private_data != (void*)((long)current->self_exec_id))
813 int this_len, retval;
815 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 retval = access_process_vm(task, src, page, this_len, 0);
817 if (!retval || check_mem_permission(task)) {
823 if (copy_to_user(buf, page, retval)) {
838 free_page((unsigned long) page);
840 put_task_struct(task);
845 #define mem_write NULL
848 /* This is a security hazard */
849 static ssize_t mem_write(struct file * file, const char __user *buf,
850 size_t count, loff_t *ppos)
854 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
855 unsigned long dst = *ppos;
861 if (check_mem_permission(task))
865 page = (char *)__get_free_page(GFP_TEMPORARY);
871 int this_len, retval;
873 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 if (copy_from_user(page, buf, this_len)) {
878 retval = access_process_vm(task, dst, page, this_len, 1);
890 free_page((unsigned long) page);
892 put_task_struct(task);
898 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
902 file->f_pos = offset;
905 file->f_pos += offset;
910 force_successful_syscall_return();
914 static const struct file_operations proc_mem_operations = {
921 static ssize_t environ_read(struct file *file, char __user *buf,
922 size_t count, loff_t *ppos)
924 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
926 unsigned long src = *ppos;
928 struct mm_struct *mm;
933 if (!ptrace_may_access(task, PTRACE_MODE_READ))
937 page = (char *)__get_free_page(GFP_TEMPORARY);
943 mm = get_task_mm(task);
948 int this_len, retval, max_len;
950 this_len = mm->env_end - (mm->env_start + src);
955 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 this_len = (this_len > max_len) ? max_len : this_len;
958 retval = access_process_vm(task, (mm->env_start + src),
966 if (copy_to_user(buf, page, retval)) {
980 free_page((unsigned long) page);
982 put_task_struct(task);
987 static const struct file_operations proc_environ_operations = {
988 .read = environ_read,
989 .llseek = generic_file_llseek,
992 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
993 size_t count, loff_t *ppos)
995 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
996 char buffer[PROC_NUMBUF];
998 int oom_adjust = OOM_DISABLE;
1004 if (lock_task_sighand(task, &flags)) {
1005 oom_adjust = task->signal->oom_adj;
1006 unlock_task_sighand(task, &flags);
1009 put_task_struct(task);
1011 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1013 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1016 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1017 size_t count, loff_t *ppos)
1019 struct task_struct *task;
1020 char buffer[PROC_NUMBUF];
1022 unsigned long flags;
1025 memset(buffer, 0, sizeof(buffer));
1026 if (count > sizeof(buffer) - 1)
1027 count = sizeof(buffer) - 1;
1028 if (copy_from_user(buffer, buf, count))
1031 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1034 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1035 oom_adjust != OOM_DISABLE)
1038 task = get_proc_task(file->f_path.dentry->d_inode);
1041 if (!lock_task_sighand(task, &flags)) {
1042 put_task_struct(task);
1046 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1047 unlock_task_sighand(task, &flags);
1048 put_task_struct(task);
1052 task->signal->oom_adj = oom_adjust;
1054 unlock_task_sighand(task, &flags);
1055 put_task_struct(task);
1060 static const struct file_operations proc_oom_adjust_operations = {
1061 .read = oom_adjust_read,
1062 .write = oom_adjust_write,
1063 .llseek = generic_file_llseek,
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))
1096 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1102 if (count >= PAGE_SIZE)
1103 count = PAGE_SIZE - 1;
1106 /* No partial writes. */
1109 page = (char*)__get_free_page(GFP_TEMPORARY);
1113 if (copy_from_user(page, buf, count))
1117 loginuid = simple_strtoul(page, &tmp, 10);
1123 length = audit_set_loginuid(current, loginuid);
1124 if (likely(length == 0))
1128 free_page((unsigned long) page);
1132 static const struct file_operations proc_loginuid_operations = {
1133 .read = proc_loginuid_read,
1134 .write = proc_loginuid_write,
1135 .llseek = generic_file_llseek,
1138 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1139 size_t count, loff_t *ppos)
1141 struct inode * inode = file->f_path.dentry->d_inode;
1142 struct task_struct *task = get_proc_task(inode);
1144 char tmpbuf[TMPBUFLEN];
1148 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1149 audit_get_sessionid(task));
1150 put_task_struct(task);
1151 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1154 static const struct file_operations proc_sessionid_operations = {
1155 .read = proc_sessionid_read,
1156 .llseek = generic_file_llseek,
1160 #ifdef CONFIG_FAULT_INJECTION
1161 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1162 size_t count, loff_t *ppos)
1164 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1165 char buffer[PROC_NUMBUF];
1171 make_it_fail = task->make_it_fail;
1172 put_task_struct(task);
1174 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1176 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1179 static ssize_t proc_fault_inject_write(struct file * file,
1180 const char __user * buf, size_t count, loff_t *ppos)
1182 struct task_struct *task;
1183 char buffer[PROC_NUMBUF], *end;
1186 if (!capable(CAP_SYS_RESOURCE))
1188 memset(buffer, 0, sizeof(buffer));
1189 if (count > sizeof(buffer) - 1)
1190 count = sizeof(buffer) - 1;
1191 if (copy_from_user(buffer, buf, count))
1193 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1196 task = get_proc_task(file->f_dentry->d_inode);
1199 task->make_it_fail = make_it_fail;
1200 put_task_struct(task);
1205 static const struct file_operations proc_fault_inject_operations = {
1206 .read = proc_fault_inject_read,
1207 .write = proc_fault_inject_write,
1208 .llseek = generic_file_llseek,
1213 #ifdef CONFIG_SCHED_DEBUG
1215 * Print out various scheduling related per-task fields:
1217 static int sched_show(struct seq_file *m, void *v)
1219 struct inode *inode = m->private;
1220 struct task_struct *p;
1222 p = get_proc_task(inode);
1225 proc_sched_show_task(p, m);
1233 sched_write(struct file *file, const char __user *buf,
1234 size_t count, loff_t *offset)
1236 struct inode *inode = file->f_path.dentry->d_inode;
1237 struct task_struct *p;
1239 p = get_proc_task(inode);
1242 proc_sched_set_task(p);
1249 static int sched_open(struct inode *inode, struct file *filp)
1253 ret = single_open(filp, sched_show, NULL);
1255 struct seq_file *m = filp->private_data;
1262 static const struct file_operations proc_pid_sched_operations = {
1265 .write = sched_write,
1266 .llseek = seq_lseek,
1267 .release = single_release,
1272 static ssize_t comm_write(struct file *file, const char __user *buf,
1273 size_t count, loff_t *offset)
1275 struct inode *inode = file->f_path.dentry->d_inode;
1276 struct task_struct *p;
1277 char buffer[TASK_COMM_LEN];
1279 memset(buffer, 0, sizeof(buffer));
1280 if (count > sizeof(buffer) - 1)
1281 count = sizeof(buffer) - 1;
1282 if (copy_from_user(buffer, buf, count))
1285 p = get_proc_task(inode);
1289 if (same_thread_group(current, p))
1290 set_task_comm(p, buffer);
1299 static int comm_show(struct seq_file *m, void *v)
1301 struct inode *inode = m->private;
1302 struct task_struct *p;
1304 p = get_proc_task(inode);
1309 seq_printf(m, "%s\n", p->comm);
1317 static int comm_open(struct inode *inode, struct file *filp)
1321 ret = single_open(filp, comm_show, NULL);
1323 struct seq_file *m = filp->private_data;
1330 static const struct file_operations proc_pid_set_comm_operations = {
1333 .write = comm_write,
1334 .llseek = seq_lseek,
1335 .release = single_release,
1339 * We added or removed a vma mapping the executable. The vmas are only mapped
1340 * during exec and are not mapped with the mmap system call.
1341 * Callers must hold down_write() on the mm's mmap_sem for these
1343 void added_exe_file_vma(struct mm_struct *mm)
1345 mm->num_exe_file_vmas++;
1348 void removed_exe_file_vma(struct mm_struct *mm)
1350 mm->num_exe_file_vmas--;
1351 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1353 mm->exe_file = NULL;
1358 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1361 get_file(new_exe_file);
1364 mm->exe_file = new_exe_file;
1365 mm->num_exe_file_vmas = 0;
1368 struct file *get_mm_exe_file(struct mm_struct *mm)
1370 struct file *exe_file;
1372 /* We need mmap_sem to protect against races with removal of
1373 * VM_EXECUTABLE vmas */
1374 down_read(&mm->mmap_sem);
1375 exe_file = mm->exe_file;
1378 up_read(&mm->mmap_sem);
1382 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1384 /* It's safe to write the exe_file pointer without exe_file_lock because
1385 * this is called during fork when the task is not yet in /proc */
1386 newmm->exe_file = get_mm_exe_file(oldmm);
1389 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1391 struct task_struct *task;
1392 struct mm_struct *mm;
1393 struct file *exe_file;
1395 task = get_proc_task(inode);
1398 mm = get_task_mm(task);
1399 put_task_struct(task);
1402 exe_file = get_mm_exe_file(mm);
1405 *exe_path = exe_file->f_path;
1406 path_get(&exe_file->f_path);
1413 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1415 struct inode *inode = dentry->d_inode;
1416 int error = -EACCES;
1418 /* We don't need a base pointer in the /proc filesystem */
1419 path_put(&nd->path);
1421 /* Are we allowed to snoop on the tasks file descriptors? */
1422 if (!proc_fd_access_allowed(inode))
1425 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1427 return ERR_PTR(error);
1430 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1432 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1439 pathname = d_path(path, tmp, PAGE_SIZE);
1440 len = PTR_ERR(pathname);
1441 if (IS_ERR(pathname))
1443 len = tmp + PAGE_SIZE - 1 - pathname;
1447 if (copy_to_user(buffer, pathname, len))
1450 free_page((unsigned long)tmp);
1454 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1456 int error = -EACCES;
1457 struct inode *inode = dentry->d_inode;
1460 /* Are we allowed to snoop on the tasks file descriptors? */
1461 if (!proc_fd_access_allowed(inode))
1464 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1468 error = do_proc_readlink(&path, buffer, buflen);
1474 static const struct inode_operations proc_pid_link_inode_operations = {
1475 .readlink = proc_pid_readlink,
1476 .follow_link = proc_pid_follow_link,
1477 .setattr = proc_setattr,
1481 /* building an inode */
1483 static int task_dumpable(struct task_struct *task)
1486 struct mm_struct *mm;
1491 dumpable = get_dumpable(mm);
1499 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1501 struct inode * inode;
1502 struct proc_inode *ei;
1503 const struct cred *cred;
1505 /* We need a new inode */
1507 inode = new_inode(sb);
1513 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1514 inode->i_op = &proc_def_inode_operations;
1517 * grab the reference to task.
1519 ei->pid = get_task_pid(task, PIDTYPE_PID);
1523 if (task_dumpable(task)) {
1525 cred = __task_cred(task);
1526 inode->i_uid = cred->euid;
1527 inode->i_gid = cred->egid;
1530 security_task_to_inode(task, inode);
1540 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1542 struct inode *inode = dentry->d_inode;
1543 struct task_struct *task;
1544 const struct cred *cred;
1546 generic_fillattr(inode, stat);
1551 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1553 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1554 task_dumpable(task)) {
1555 cred = __task_cred(task);
1556 stat->uid = cred->euid;
1557 stat->gid = cred->egid;
1567 * Exceptional case: normally we are not allowed to unhash a busy
1568 * directory. In this case, however, we can do it - no aliasing problems
1569 * due to the way we treat inodes.
1571 * Rewrite the inode's ownerships here because the owning task may have
1572 * performed a setuid(), etc.
1574 * Before the /proc/pid/status file was created the only way to read
1575 * the effective uid of a /process was to stat /proc/pid. Reading
1576 * /proc/pid/status is slow enough that procps and other packages
1577 * kept stating /proc/pid. To keep the rules in /proc simple I have
1578 * made this apply to all per process world readable and executable
1581 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1583 struct inode *inode = dentry->d_inode;
1584 struct task_struct *task = get_proc_task(inode);
1585 const struct cred *cred;
1588 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1589 task_dumpable(task)) {
1591 cred = __task_cred(task);
1592 inode->i_uid = cred->euid;
1593 inode->i_gid = cred->egid;
1599 inode->i_mode &= ~(S_ISUID | S_ISGID);
1600 security_task_to_inode(task, inode);
1601 put_task_struct(task);
1608 static int pid_delete_dentry(struct dentry * dentry)
1610 /* Is the task we represent dead?
1611 * If so, then don't put the dentry on the lru list,
1612 * kill it immediately.
1614 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1617 static const struct dentry_operations pid_dentry_operations =
1619 .d_revalidate = pid_revalidate,
1620 .d_delete = pid_delete_dentry,
1625 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1626 struct task_struct *, const void *);
1629 * Fill a directory entry.
1631 * If possible create the dcache entry and derive our inode number and
1632 * file type from dcache entry.
1634 * Since all of the proc inode numbers are dynamically generated, the inode
1635 * numbers do not exist until the inode is cache. This means creating the
1636 * the dcache entry in readdir is necessary to keep the inode numbers
1637 * reported by readdir in sync with the inode numbers reported
1640 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1641 char *name, int len,
1642 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1644 struct dentry *child, *dir = filp->f_path.dentry;
1645 struct inode *inode;
1648 unsigned type = DT_UNKNOWN;
1652 qname.hash = full_name_hash(name, len);
1654 child = d_lookup(dir, &qname);
1657 new = d_alloc(dir, &qname);
1659 child = instantiate(dir->d_inode, new, task, ptr);
1666 if (!child || IS_ERR(child) || !child->d_inode)
1667 goto end_instantiate;
1668 inode = child->d_inode;
1671 type = inode->i_mode >> 12;
1676 ino = find_inode_number(dir, &qname);
1679 return filldir(dirent, name, len, filp->f_pos, ino, type);
1682 static unsigned name_to_int(struct dentry *dentry)
1684 const char *name = dentry->d_name.name;
1685 int len = dentry->d_name.len;
1688 if (len > 1 && *name == '0')
1691 unsigned c = *name++ - '0';
1694 if (n >= (~0U-9)/10)
1704 #define PROC_FDINFO_MAX 64
1706 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1708 struct task_struct *task = get_proc_task(inode);
1709 struct files_struct *files = NULL;
1711 int fd = proc_fd(inode);
1714 files = get_files_struct(task);
1715 put_task_struct(task);
1719 * We are not taking a ref to the file structure, so we must
1722 spin_lock(&files->file_lock);
1723 file = fcheck_files(files, fd);
1726 *path = file->f_path;
1727 path_get(&file->f_path);
1730 snprintf(info, PROC_FDINFO_MAX,
1733 (long long) file->f_pos,
1735 spin_unlock(&files->file_lock);
1736 put_files_struct(files);
1739 spin_unlock(&files->file_lock);
1740 put_files_struct(files);
1745 static int proc_fd_link(struct inode *inode, struct path *path)
1747 return proc_fd_info(inode, path, NULL);
1750 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1752 struct inode *inode = dentry->d_inode;
1753 struct task_struct *task = get_proc_task(inode);
1754 int fd = proc_fd(inode);
1755 struct files_struct *files;
1756 const struct cred *cred;
1759 files = get_files_struct(task);
1762 if (fcheck_files(files, fd)) {
1764 put_files_struct(files);
1765 if (task_dumpable(task)) {
1767 cred = __task_cred(task);
1768 inode->i_uid = cred->euid;
1769 inode->i_gid = cred->egid;
1775 inode->i_mode &= ~(S_ISUID | S_ISGID);
1776 security_task_to_inode(task, inode);
1777 put_task_struct(task);
1781 put_files_struct(files);
1783 put_task_struct(task);
1789 static const struct dentry_operations tid_fd_dentry_operations =
1791 .d_revalidate = tid_fd_revalidate,
1792 .d_delete = pid_delete_dentry,
1795 static struct dentry *proc_fd_instantiate(struct inode *dir,
1796 struct dentry *dentry, struct task_struct *task, const void *ptr)
1798 unsigned fd = *(const unsigned *)ptr;
1800 struct files_struct *files;
1801 struct inode *inode;
1802 struct proc_inode *ei;
1803 struct dentry *error = ERR_PTR(-ENOENT);
1805 inode = proc_pid_make_inode(dir->i_sb, task);
1810 files = get_files_struct(task);
1813 inode->i_mode = S_IFLNK;
1816 * We are not taking a ref to the file structure, so we must
1819 spin_lock(&files->file_lock);
1820 file = fcheck_files(files, fd);
1823 if (file->f_mode & FMODE_READ)
1824 inode->i_mode |= S_IRUSR | S_IXUSR;
1825 if (file->f_mode & FMODE_WRITE)
1826 inode->i_mode |= S_IWUSR | S_IXUSR;
1827 spin_unlock(&files->file_lock);
1828 put_files_struct(files);
1830 inode->i_op = &proc_pid_link_inode_operations;
1832 ei->op.proc_get_link = proc_fd_link;
1833 dentry->d_op = &tid_fd_dentry_operations;
1834 d_add(dentry, inode);
1835 /* Close the race of the process dying before we return the dentry */
1836 if (tid_fd_revalidate(dentry, NULL))
1842 spin_unlock(&files->file_lock);
1843 put_files_struct(files);
1849 static struct dentry *proc_lookupfd_common(struct inode *dir,
1850 struct dentry *dentry,
1851 instantiate_t instantiate)
1853 struct task_struct *task = get_proc_task(dir);
1854 unsigned fd = name_to_int(dentry);
1855 struct dentry *result = ERR_PTR(-ENOENT);
1862 result = instantiate(dir, dentry, task, &fd);
1864 put_task_struct(task);
1869 static int proc_readfd_common(struct file * filp, void * dirent,
1870 filldir_t filldir, instantiate_t instantiate)
1872 struct dentry *dentry = filp->f_path.dentry;
1873 struct inode *inode = dentry->d_inode;
1874 struct task_struct *p = get_proc_task(inode);
1875 unsigned int fd, ino;
1877 struct files_struct * files;
1887 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1891 ino = parent_ino(dentry);
1892 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1896 files = get_files_struct(p);
1900 for (fd = filp->f_pos-2;
1901 fd < files_fdtable(files)->max_fds;
1902 fd++, filp->f_pos++) {
1903 char name[PROC_NUMBUF];
1906 if (!fcheck_files(files, fd))
1910 len = snprintf(name, sizeof(name), "%d", fd);
1911 if (proc_fill_cache(filp, dirent, filldir,
1912 name, len, instantiate,
1920 put_files_struct(files);
1928 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1929 struct nameidata *nd)
1931 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1934 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1936 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1939 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1940 size_t len, loff_t *ppos)
1942 char tmp[PROC_FDINFO_MAX];
1943 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1945 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1949 static const struct file_operations proc_fdinfo_file_operations = {
1950 .open = nonseekable_open,
1951 .read = proc_fdinfo_read,
1954 static const struct file_operations proc_fd_operations = {
1955 .read = generic_read_dir,
1956 .readdir = proc_readfd,
1960 * /proc/pid/fd needs a special permission handler so that a process can still
1961 * access /proc/self/fd after it has executed a setuid().
1963 static int proc_fd_permission(struct inode *inode, int mask)
1967 rv = generic_permission(inode, mask, NULL);
1970 if (task_pid(current) == proc_pid(inode))
1976 * proc directories can do almost nothing..
1978 static const struct inode_operations proc_fd_inode_operations = {
1979 .lookup = proc_lookupfd,
1980 .permission = proc_fd_permission,
1981 .setattr = proc_setattr,
1984 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1985 struct dentry *dentry, struct task_struct *task, const void *ptr)
1987 unsigned fd = *(unsigned *)ptr;
1988 struct inode *inode;
1989 struct proc_inode *ei;
1990 struct dentry *error = ERR_PTR(-ENOENT);
1992 inode = proc_pid_make_inode(dir->i_sb, task);
1997 inode->i_mode = S_IFREG | S_IRUSR;
1998 inode->i_fop = &proc_fdinfo_file_operations;
1999 dentry->d_op = &tid_fd_dentry_operations;
2000 d_add(dentry, inode);
2001 /* Close the race of the process dying before we return the dentry */
2002 if (tid_fd_revalidate(dentry, NULL))
2009 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2010 struct dentry *dentry,
2011 struct nameidata *nd)
2013 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2016 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2018 return proc_readfd_common(filp, dirent, filldir,
2019 proc_fdinfo_instantiate);
2022 static const struct file_operations proc_fdinfo_operations = {
2023 .read = generic_read_dir,
2024 .readdir = proc_readfdinfo,
2028 * proc directories can do almost nothing..
2030 static const struct inode_operations proc_fdinfo_inode_operations = {
2031 .lookup = proc_lookupfdinfo,
2032 .setattr = proc_setattr,
2036 static struct dentry *proc_pident_instantiate(struct inode *dir,
2037 struct dentry *dentry, struct task_struct *task, const void *ptr)
2039 const struct pid_entry *p = ptr;
2040 struct inode *inode;
2041 struct proc_inode *ei;
2042 struct dentry *error = ERR_PTR(-ENOENT);
2044 inode = proc_pid_make_inode(dir->i_sb, task);
2049 inode->i_mode = p->mode;
2050 if (S_ISDIR(inode->i_mode))
2051 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2053 inode->i_op = p->iop;
2055 inode->i_fop = p->fop;
2057 dentry->d_op = &pid_dentry_operations;
2058 d_add(dentry, inode);
2059 /* Close the race of the process dying before we return the dentry */
2060 if (pid_revalidate(dentry, NULL))
2066 static struct dentry *proc_pident_lookup(struct inode *dir,
2067 struct dentry *dentry,
2068 const struct pid_entry *ents,
2071 struct dentry *error;
2072 struct task_struct *task = get_proc_task(dir);
2073 const struct pid_entry *p, *last;
2075 error = ERR_PTR(-ENOENT);
2081 * Yes, it does not scale. And it should not. Don't add
2082 * new entries into /proc/<tgid>/ without very good reasons.
2084 last = &ents[nents - 1];
2085 for (p = ents; p <= last; p++) {
2086 if (p->len != dentry->d_name.len)
2088 if (!memcmp(dentry->d_name.name, p->name, p->len))
2094 error = proc_pident_instantiate(dir, dentry, task, p);
2096 put_task_struct(task);
2101 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2102 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2104 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2105 proc_pident_instantiate, task, p);
2108 static int proc_pident_readdir(struct file *filp,
2109 void *dirent, filldir_t filldir,
2110 const struct pid_entry *ents, unsigned int nents)
2113 struct dentry *dentry = filp->f_path.dentry;
2114 struct inode *inode = dentry->d_inode;
2115 struct task_struct *task = get_proc_task(inode);
2116 const struct pid_entry *p, *last;
2129 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2135 ino = parent_ino(dentry);
2136 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2148 last = &ents[nents - 1];
2150 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2159 put_task_struct(task);
2164 #ifdef CONFIG_SECURITY
2165 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2166 size_t count, loff_t *ppos)
2168 struct inode * inode = file->f_path.dentry->d_inode;
2171 struct task_struct *task = get_proc_task(inode);
2176 length = security_getprocattr(task,
2177 (char*)file->f_path.dentry->d_name.name,
2179 put_task_struct(task);
2181 length = simple_read_from_buffer(buf, count, ppos, p, length);
2186 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2187 size_t count, loff_t *ppos)
2189 struct inode * inode = file->f_path.dentry->d_inode;
2192 struct task_struct *task = get_proc_task(inode);
2197 if (count > PAGE_SIZE)
2200 /* No partial writes. */
2206 page = (char*)__get_free_page(GFP_TEMPORARY);
2211 if (copy_from_user(page, buf, count))
2214 /* Guard against adverse ptrace interaction */
2215 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2219 length = security_setprocattr(task,
2220 (char*)file->f_path.dentry->d_name.name,
2221 (void*)page, count);
2222 mutex_unlock(&task->cred_guard_mutex);
2224 free_page((unsigned long) page);
2226 put_task_struct(task);
2231 static const struct file_operations proc_pid_attr_operations = {
2232 .read = proc_pid_attr_read,
2233 .write = proc_pid_attr_write,
2234 .llseek = generic_file_llseek,
2237 static const struct pid_entry attr_dir_stuff[] = {
2238 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2239 REG("prev", S_IRUGO, proc_pid_attr_operations),
2240 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2241 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2242 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2243 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2246 static int proc_attr_dir_readdir(struct file * filp,
2247 void * dirent, filldir_t filldir)
2249 return proc_pident_readdir(filp,dirent,filldir,
2250 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2253 static const struct file_operations proc_attr_dir_operations = {
2254 .read = generic_read_dir,
2255 .readdir = proc_attr_dir_readdir,
2258 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2259 struct dentry *dentry, struct nameidata *nd)
2261 return proc_pident_lookup(dir, dentry,
2262 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2265 static const struct inode_operations proc_attr_dir_inode_operations = {
2266 .lookup = proc_attr_dir_lookup,
2267 .getattr = pid_getattr,
2268 .setattr = proc_setattr,
2273 #ifdef CONFIG_ELF_CORE
2274 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2275 size_t count, loff_t *ppos)
2277 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2278 struct mm_struct *mm;
2279 char buffer[PROC_NUMBUF];
2287 mm = get_task_mm(task);
2289 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2290 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2291 MMF_DUMP_FILTER_SHIFT));
2293 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2296 put_task_struct(task);
2301 static ssize_t proc_coredump_filter_write(struct file *file,
2302 const char __user *buf,
2306 struct task_struct *task;
2307 struct mm_struct *mm;
2308 char buffer[PROC_NUMBUF], *end;
2315 memset(buffer, 0, sizeof(buffer));
2316 if (count > sizeof(buffer) - 1)
2317 count = sizeof(buffer) - 1;
2318 if (copy_from_user(buffer, buf, count))
2322 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2325 if (end - buffer == 0)
2329 task = get_proc_task(file->f_dentry->d_inode);
2334 mm = get_task_mm(task);
2338 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2340 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2342 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2347 put_task_struct(task);
2352 static const struct file_operations proc_coredump_filter_operations = {
2353 .read = proc_coredump_filter_read,
2354 .write = proc_coredump_filter_write,
2355 .llseek = generic_file_llseek,
2362 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2365 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2366 pid_t tgid = task_tgid_nr_ns(current, ns);
2367 char tmp[PROC_NUMBUF];
2370 sprintf(tmp, "%d", tgid);
2371 return vfs_readlink(dentry,buffer,buflen,tmp);
2374 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2376 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2377 pid_t tgid = task_tgid_nr_ns(current, ns);
2378 char *name = ERR_PTR(-ENOENT);
2382 name = ERR_PTR(-ENOMEM);
2384 sprintf(name, "%d", tgid);
2386 nd_set_link(nd, name);
2390 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2393 char *s = nd_get_link(nd);
2398 static const struct inode_operations proc_self_inode_operations = {
2399 .readlink = proc_self_readlink,
2400 .follow_link = proc_self_follow_link,
2401 .put_link = proc_self_put_link,
2407 * These are the directory entries in the root directory of /proc
2408 * that properly belong to the /proc filesystem, as they describe
2409 * describe something that is process related.
2411 static const struct pid_entry proc_base_stuff[] = {
2412 NOD("self", S_IFLNK|S_IRWXUGO,
2413 &proc_self_inode_operations, NULL, {}),
2417 * Exceptional case: normally we are not allowed to unhash a busy
2418 * directory. In this case, however, we can do it - no aliasing problems
2419 * due to the way we treat inodes.
2421 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2423 struct inode *inode = dentry->d_inode;
2424 struct task_struct *task = get_proc_task(inode);
2426 put_task_struct(task);
2433 static const struct dentry_operations proc_base_dentry_operations =
2435 .d_revalidate = proc_base_revalidate,
2436 .d_delete = pid_delete_dentry,
2439 static struct dentry *proc_base_instantiate(struct inode *dir,
2440 struct dentry *dentry, struct task_struct *task, const void *ptr)
2442 const struct pid_entry *p = ptr;
2443 struct inode *inode;
2444 struct proc_inode *ei;
2445 struct dentry *error;
2447 /* Allocate the inode */
2448 error = ERR_PTR(-ENOMEM);
2449 inode = new_inode(dir->i_sb);
2453 /* Initialize the inode */
2455 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2458 * grab the reference to the task.
2460 ei->pid = get_task_pid(task, PIDTYPE_PID);
2464 inode->i_mode = p->mode;
2465 if (S_ISDIR(inode->i_mode))
2467 if (S_ISLNK(inode->i_mode))
2470 inode->i_op = p->iop;
2472 inode->i_fop = p->fop;
2474 dentry->d_op = &proc_base_dentry_operations;
2475 d_add(dentry, inode);
2484 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2486 struct dentry *error;
2487 struct task_struct *task = get_proc_task(dir);
2488 const struct pid_entry *p, *last;
2490 error = ERR_PTR(-ENOENT);
2495 /* Lookup the directory entry */
2496 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2497 for (p = proc_base_stuff; p <= last; p++) {
2498 if (p->len != dentry->d_name.len)
2500 if (!memcmp(dentry->d_name.name, p->name, p->len))
2506 error = proc_base_instantiate(dir, dentry, task, p);
2509 put_task_struct(task);
2514 static int proc_base_fill_cache(struct file *filp, void *dirent,
2515 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2517 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2518 proc_base_instantiate, task, p);
2521 #ifdef CONFIG_TASK_IO_ACCOUNTING
2522 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2524 struct task_io_accounting acct = task->ioac;
2525 unsigned long flags;
2527 if (whole && lock_task_sighand(task, &flags)) {
2528 struct task_struct *t = task;
2530 task_io_accounting_add(&acct, &task->signal->ioac);
2531 while_each_thread(task, t)
2532 task_io_accounting_add(&acct, &t->ioac);
2534 unlock_task_sighand(task, &flags);
2536 return sprintf(buffer,
2541 "read_bytes: %llu\n"
2542 "write_bytes: %llu\n"
2543 "cancelled_write_bytes: %llu\n",
2544 (unsigned long long)acct.rchar,
2545 (unsigned long long)acct.wchar,
2546 (unsigned long long)acct.syscr,
2547 (unsigned long long)acct.syscw,
2548 (unsigned long long)acct.read_bytes,
2549 (unsigned long long)acct.write_bytes,
2550 (unsigned long long)acct.cancelled_write_bytes);
2553 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2555 return do_io_accounting(task, buffer, 0);
2558 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2560 return do_io_accounting(task, buffer, 1);
2562 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2564 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2565 struct pid *pid, struct task_struct *task)
2567 seq_printf(m, "%08x\n", task->personality);
2574 static const struct file_operations proc_task_operations;
2575 static const struct inode_operations proc_task_inode_operations;
2577 static const struct pid_entry tgid_base_stuff[] = {
2578 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2579 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2580 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2582 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2584 REG("environ", S_IRUSR, proc_environ_operations),
2585 INF("auxv", S_IRUSR, proc_pid_auxv),
2586 ONE("status", S_IRUGO, proc_pid_status),
2587 ONE("personality", S_IRUSR, proc_pid_personality),
2588 INF("limits", S_IRUSR, proc_pid_limits),
2589 #ifdef CONFIG_SCHED_DEBUG
2590 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2592 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2593 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2594 INF("syscall", S_IRUSR, proc_pid_syscall),
2596 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2597 ONE("stat", S_IRUGO, proc_tgid_stat),
2598 ONE("statm", S_IRUGO, proc_pid_statm),
2599 REG("maps", S_IRUGO, proc_maps_operations),
2601 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2603 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2604 LNK("cwd", proc_cwd_link),
2605 LNK("root", proc_root_link),
2606 LNK("exe", proc_exe_link),
2607 REG("mounts", S_IRUGO, proc_mounts_operations),
2608 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2609 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2610 #ifdef CONFIG_PROC_PAGE_MONITOR
2611 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2612 REG("smaps", S_IRUGO, proc_smaps_operations),
2613 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2615 #ifdef CONFIG_SECURITY
2616 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2618 #ifdef CONFIG_KALLSYMS
2619 INF("wchan", S_IRUGO, proc_pid_wchan),
2621 #ifdef CONFIG_STACKTRACE
2622 ONE("stack", S_IRUSR, proc_pid_stack),
2624 #ifdef CONFIG_SCHEDSTATS
2625 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2627 #ifdef CONFIG_LATENCYTOP
2628 REG("latency", S_IRUGO, proc_lstats_operations),
2630 #ifdef CONFIG_PROC_PID_CPUSET
2631 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2633 #ifdef CONFIG_CGROUPS
2634 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2636 INF("oom_score", S_IRUGO, proc_oom_score),
2637 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2638 #ifdef CONFIG_AUDITSYSCALL
2639 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2640 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2642 #ifdef CONFIG_FAULT_INJECTION
2643 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2645 #ifdef CONFIG_ELF_CORE
2646 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2648 #ifdef CONFIG_TASK_IO_ACCOUNTING
2649 INF("io", S_IRUGO, proc_tgid_io_accounting),
2653 static int proc_tgid_base_readdir(struct file * filp,
2654 void * dirent, filldir_t filldir)
2656 return proc_pident_readdir(filp,dirent,filldir,
2657 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2660 static const struct file_operations proc_tgid_base_operations = {
2661 .read = generic_read_dir,
2662 .readdir = proc_tgid_base_readdir,
2665 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2666 return proc_pident_lookup(dir, dentry,
2667 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2670 static const struct inode_operations proc_tgid_base_inode_operations = {
2671 .lookup = proc_tgid_base_lookup,
2672 .getattr = pid_getattr,
2673 .setattr = proc_setattr,
2676 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2678 struct dentry *dentry, *leader, *dir;
2679 char buf[PROC_NUMBUF];
2683 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2684 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2686 shrink_dcache_parent(dentry);
2692 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2693 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2698 name.len = strlen(name.name);
2699 dir = d_hash_and_lookup(leader, &name);
2701 goto out_put_leader;
2704 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2705 dentry = d_hash_and_lookup(dir, &name);
2707 shrink_dcache_parent(dentry);
2720 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2721 * @task: task that should be flushed.
2723 * When flushing dentries from proc, one needs to flush them from global
2724 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2725 * in. This call is supposed to do all of this job.
2727 * Looks in the dcache for
2729 * /proc/@tgid/task/@pid
2730 * if either directory is present flushes it and all of it'ts children
2733 * It is safe and reasonable to cache /proc entries for a task until
2734 * that task exits. After that they just clog up the dcache with
2735 * useless entries, possibly causing useful dcache entries to be
2736 * flushed instead. This routine is proved to flush those useless
2737 * dcache entries at process exit time.
2739 * NOTE: This routine is just an optimization so it does not guarantee
2740 * that no dcache entries will exist at process exit time it
2741 * just makes it very unlikely that any will persist.
2744 void proc_flush_task(struct task_struct *task)
2747 struct pid *pid, *tgid;
2750 pid = task_pid(task);
2751 tgid = task_tgid(task);
2753 for (i = 0; i <= pid->level; i++) {
2754 upid = &pid->numbers[i];
2755 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2756 tgid->numbers[i].nr);
2759 upid = &pid->numbers[pid->level];
2761 pid_ns_release_proc(upid->ns);
2764 static struct dentry *proc_pid_instantiate(struct inode *dir,
2765 struct dentry * dentry,
2766 struct task_struct *task, const void *ptr)
2768 struct dentry *error = ERR_PTR(-ENOENT);
2769 struct inode *inode;
2771 inode = proc_pid_make_inode(dir->i_sb, task);
2775 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2776 inode->i_op = &proc_tgid_base_inode_operations;
2777 inode->i_fop = &proc_tgid_base_operations;
2778 inode->i_flags|=S_IMMUTABLE;
2780 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2781 ARRAY_SIZE(tgid_base_stuff));
2783 dentry->d_op = &pid_dentry_operations;
2785 d_add(dentry, inode);
2786 /* Close the race of the process dying before we return the dentry */
2787 if (pid_revalidate(dentry, NULL))
2793 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2795 struct dentry *result;
2796 struct task_struct *task;
2798 struct pid_namespace *ns;
2800 result = proc_base_lookup(dir, dentry);
2801 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2804 tgid = name_to_int(dentry);
2808 ns = dentry->d_sb->s_fs_info;
2810 task = find_task_by_pid_ns(tgid, ns);
2812 get_task_struct(task);
2817 result = proc_pid_instantiate(dir, dentry, task, NULL);
2818 put_task_struct(task);
2824 * Find the first task with tgid >= tgid
2829 struct task_struct *task;
2831 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2836 put_task_struct(iter.task);
2840 pid = find_ge_pid(iter.tgid, ns);
2842 iter.tgid = pid_nr_ns(pid, ns);
2843 iter.task = pid_task(pid, PIDTYPE_PID);
2844 /* What we to know is if the pid we have find is the
2845 * pid of a thread_group_leader. Testing for task
2846 * being a thread_group_leader is the obvious thing
2847 * todo but there is a window when it fails, due to
2848 * the pid transfer logic in de_thread.
2850 * So we perform the straight forward test of seeing
2851 * if the pid we have found is the pid of a thread
2852 * group leader, and don't worry if the task we have
2853 * found doesn't happen to be a thread group leader.
2854 * As we don't care in the case of readdir.
2856 if (!iter.task || !has_group_leader_pid(iter.task)) {
2860 get_task_struct(iter.task);
2866 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2868 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2869 struct tgid_iter iter)
2871 char name[PROC_NUMBUF];
2872 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2873 return proc_fill_cache(filp, dirent, filldir, name, len,
2874 proc_pid_instantiate, iter.task, NULL);
2877 /* for the /proc/ directory itself, after non-process stuff has been done */
2878 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2880 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2881 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2882 struct tgid_iter iter;
2883 struct pid_namespace *ns;
2888 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2889 const struct pid_entry *p = &proc_base_stuff[nr];
2890 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2894 ns = filp->f_dentry->d_sb->s_fs_info;
2896 iter.tgid = filp->f_pos - TGID_OFFSET;
2897 for (iter = next_tgid(ns, iter);
2899 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2900 filp->f_pos = iter.tgid + TGID_OFFSET;
2901 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2902 put_task_struct(iter.task);
2906 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2908 put_task_struct(reaper);
2916 static const struct pid_entry tid_base_stuff[] = {
2917 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2918 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2919 REG("environ", S_IRUSR, proc_environ_operations),
2920 INF("auxv", S_IRUSR, proc_pid_auxv),
2921 ONE("status", S_IRUGO, proc_pid_status),
2922 ONE("personality", S_IRUSR, proc_pid_personality),
2923 INF("limits", S_IRUSR, proc_pid_limits),
2924 #ifdef CONFIG_SCHED_DEBUG
2925 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2927 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2928 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2929 INF("syscall", S_IRUSR, proc_pid_syscall),
2931 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2932 ONE("stat", S_IRUGO, proc_tid_stat),
2933 ONE("statm", S_IRUGO, proc_pid_statm),
2934 REG("maps", S_IRUGO, proc_maps_operations),
2936 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2938 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2939 LNK("cwd", proc_cwd_link),
2940 LNK("root", proc_root_link),
2941 LNK("exe", proc_exe_link),
2942 REG("mounts", S_IRUGO, proc_mounts_operations),
2943 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2944 #ifdef CONFIG_PROC_PAGE_MONITOR
2945 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2946 REG("smaps", S_IRUGO, proc_smaps_operations),
2947 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2949 #ifdef CONFIG_SECURITY
2950 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2952 #ifdef CONFIG_KALLSYMS
2953 INF("wchan", S_IRUGO, proc_pid_wchan),
2955 #ifdef CONFIG_STACKTRACE
2956 ONE("stack", S_IRUSR, proc_pid_stack),
2958 #ifdef CONFIG_SCHEDSTATS
2959 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2961 #ifdef CONFIG_LATENCYTOP
2962 REG("latency", S_IRUGO, proc_lstats_operations),
2964 #ifdef CONFIG_PROC_PID_CPUSET
2965 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2967 #ifdef CONFIG_CGROUPS
2968 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2970 INF("oom_score", S_IRUGO, proc_oom_score),
2971 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2972 #ifdef CONFIG_AUDITSYSCALL
2973 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2974 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2976 #ifdef CONFIG_FAULT_INJECTION
2977 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2979 #ifdef CONFIG_TASK_IO_ACCOUNTING
2980 INF("io", S_IRUGO, proc_tid_io_accounting),
2984 static int proc_tid_base_readdir(struct file * filp,
2985 void * dirent, filldir_t filldir)
2987 return proc_pident_readdir(filp,dirent,filldir,
2988 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2991 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2992 return proc_pident_lookup(dir, dentry,
2993 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2996 static const struct file_operations proc_tid_base_operations = {
2997 .read = generic_read_dir,
2998 .readdir = proc_tid_base_readdir,
3001 static const struct inode_operations proc_tid_base_inode_operations = {
3002 .lookup = proc_tid_base_lookup,
3003 .getattr = pid_getattr,
3004 .setattr = proc_setattr,
3007 static struct dentry *proc_task_instantiate(struct inode *dir,
3008 struct dentry *dentry, struct task_struct *task, const void *ptr)
3010 struct dentry *error = ERR_PTR(-ENOENT);
3011 struct inode *inode;
3012 inode = proc_pid_make_inode(dir->i_sb, task);
3016 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3017 inode->i_op = &proc_tid_base_inode_operations;
3018 inode->i_fop = &proc_tid_base_operations;
3019 inode->i_flags|=S_IMMUTABLE;
3021 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3022 ARRAY_SIZE(tid_base_stuff));
3024 dentry->d_op = &pid_dentry_operations;
3026 d_add(dentry, inode);
3027 /* Close the race of the process dying before we return the dentry */
3028 if (pid_revalidate(dentry, NULL))
3034 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3036 struct dentry *result = ERR_PTR(-ENOENT);
3037 struct task_struct *task;
3038 struct task_struct *leader = get_proc_task(dir);
3040 struct pid_namespace *ns;
3045 tid = name_to_int(dentry);
3049 ns = dentry->d_sb->s_fs_info;
3051 task = find_task_by_pid_ns(tid, ns);
3053 get_task_struct(task);
3057 if (!same_thread_group(leader, task))
3060 result = proc_task_instantiate(dir, dentry, task, NULL);
3062 put_task_struct(task);
3064 put_task_struct(leader);
3070 * Find the first tid of a thread group to return to user space.
3072 * Usually this is just the thread group leader, but if the users
3073 * buffer was too small or there was a seek into the middle of the
3074 * directory we have more work todo.
3076 * In the case of a short read we start with find_task_by_pid.
3078 * In the case of a seek we start with the leader and walk nr
3081 static struct task_struct *first_tid(struct task_struct *leader,
3082 int tid, int nr, struct pid_namespace *ns)
3084 struct task_struct *pos;
3087 /* Attempt to start with the pid of a thread */
3088 if (tid && (nr > 0)) {
3089 pos = find_task_by_pid_ns(tid, ns);
3090 if (pos && (pos->group_leader == leader))
3094 /* If nr exceeds the number of threads there is nothing todo */
3096 if (nr && nr >= get_nr_threads(leader))
3099 /* If we haven't found our starting place yet start
3100 * with the leader and walk nr threads forward.
3102 for (pos = leader; nr > 0; --nr) {
3103 pos = next_thread(pos);
3104 if (pos == leader) {
3110 get_task_struct(pos);
3117 * Find the next thread in the thread list.
3118 * Return NULL if there is an error or no next thread.
3120 * The reference to the input task_struct is released.
3122 static struct task_struct *next_tid(struct task_struct *start)
3124 struct task_struct *pos = NULL;
3126 if (pid_alive(start)) {
3127 pos = next_thread(start);
3128 if (thread_group_leader(pos))
3131 get_task_struct(pos);
3134 put_task_struct(start);
3138 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3139 struct task_struct *task, int tid)
3141 char name[PROC_NUMBUF];
3142 int len = snprintf(name, sizeof(name), "%d", tid);
3143 return proc_fill_cache(filp, dirent, filldir, name, len,
3144 proc_task_instantiate, task, NULL);
3147 /* for the /proc/TGID/task/ directories */
3148 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3150 struct dentry *dentry = filp->f_path.dentry;
3151 struct inode *inode = dentry->d_inode;
3152 struct task_struct *leader = NULL;
3153 struct task_struct *task;
3154 int retval = -ENOENT;
3157 struct pid_namespace *ns;
3159 task = get_proc_task(inode);
3163 if (pid_alive(task)) {
3164 leader = task->group_leader;
3165 get_task_struct(leader);
3168 put_task_struct(task);
3173 switch ((unsigned long)filp->f_pos) {
3176 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3181 ino = parent_ino(dentry);
3182 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3188 /* f_version caches the tgid value that the last readdir call couldn't
3189 * return. lseek aka telldir automagically resets f_version to 0.
3191 ns = filp->f_dentry->d_sb->s_fs_info;
3192 tid = (int)filp->f_version;
3193 filp->f_version = 0;
3194 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3196 task = next_tid(task), filp->f_pos++) {
3197 tid = task_pid_nr_ns(task, ns);
3198 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3199 /* returning this tgid failed, save it as the first
3200 * pid for the next readir call */
3201 filp->f_version = (u64)tid;
3202 put_task_struct(task);
3207 put_task_struct(leader);
3212 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3214 struct inode *inode = dentry->d_inode;
3215 struct task_struct *p = get_proc_task(inode);
3216 generic_fillattr(inode, stat);
3219 stat->nlink += get_nr_threads(p);
3226 static const struct inode_operations proc_task_inode_operations = {
3227 .lookup = proc_task_lookup,
3228 .getattr = proc_task_getattr,
3229 .setattr = proc_setattr,
3232 static const struct file_operations proc_task_operations = {
3233 .read = generic_read_dir,
3234 .readdir = proc_task_readdir,