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, struct mem_cgroup *mem,
432 nodemask_t *nodemask, unsigned long uptime);
433 static int proc_oom_score(struct task_struct *task, char *buffer)
435 unsigned long points = 0;
436 struct timespec uptime;
438 do_posix_clock_monotonic_gettime(&uptime);
439 read_lock(&tasklist_lock);
441 points = badness(task, NULL, NULL, uptime.tv_sec);
442 read_unlock(&tasklist_lock);
443 return sprintf(buffer, "%lu\n", points);
451 static const struct limit_names lnames[RLIM_NLIMITS] = {
452 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
453 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
454 [RLIMIT_DATA] = {"Max data size", "bytes"},
455 [RLIMIT_STACK] = {"Max stack size", "bytes"},
456 [RLIMIT_CORE] = {"Max core file size", "bytes"},
457 [RLIMIT_RSS] = {"Max resident set", "bytes"},
458 [RLIMIT_NPROC] = {"Max processes", "processes"},
459 [RLIMIT_NOFILE] = {"Max open files", "files"},
460 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
461 [RLIMIT_AS] = {"Max address space", "bytes"},
462 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
463 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
464 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
465 [RLIMIT_NICE] = {"Max nice priority", NULL},
466 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
467 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
470 /* Display limits for a process */
471 static int proc_pid_limits(struct task_struct *task, char *buffer)
476 char *bufptr = buffer;
478 struct rlimit rlim[RLIM_NLIMITS];
480 if (!lock_task_sighand(task, &flags))
482 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
483 unlock_task_sighand(task, &flags);
486 * print the file header
488 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
489 "Limit", "Soft Limit", "Hard Limit", "Units");
491 for (i = 0; i < RLIM_NLIMITS; i++) {
492 if (rlim[i].rlim_cur == RLIM_INFINITY)
493 count += sprintf(&bufptr[count], "%-25s %-20s ",
494 lnames[i].name, "unlimited");
496 count += sprintf(&bufptr[count], "%-25s %-20lu ",
497 lnames[i].name, rlim[i].rlim_cur);
499 if (rlim[i].rlim_max == RLIM_INFINITY)
500 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
502 count += sprintf(&bufptr[count], "%-20lu ",
506 count += sprintf(&bufptr[count], "%-10s\n",
509 count += sprintf(&bufptr[count], "\n");
515 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
516 static int proc_pid_syscall(struct task_struct *task, char *buffer)
519 unsigned long args[6], sp, pc;
521 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
522 return sprintf(buffer, "running\n");
525 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
527 return sprintf(buffer,
528 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
530 args[0], args[1], args[2], args[3], args[4], args[5],
533 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
535 /************************************************************************/
536 /* Here the fs part begins */
537 /************************************************************************/
539 /* permission checks */
540 static int proc_fd_access_allowed(struct inode *inode)
542 struct task_struct *task;
544 /* Allow access to a task's file descriptors if it is us or we
545 * may use ptrace attach to the process and find out that
548 task = get_proc_task(inode);
550 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
551 put_task_struct(task);
556 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
559 struct inode *inode = dentry->d_inode;
561 if (attr->ia_valid & ATTR_MODE)
564 error = inode_change_ok(inode, attr);
566 error = inode_setattr(inode, attr);
570 static const struct inode_operations proc_def_inode_operations = {
571 .setattr = proc_setattr,
574 static int mounts_open_common(struct inode *inode, struct file *file,
575 const struct seq_operations *op)
577 struct task_struct *task = get_proc_task(inode);
579 struct mnt_namespace *ns = NULL;
581 struct proc_mounts *p;
586 nsp = task_nsproxy(task);
593 if (ns && get_fs_path(task, &root, 1) == 0)
595 put_task_struct(task);
604 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
608 file->private_data = &p->m;
609 ret = seq_open(file, op);
616 p->event = ns->event;
630 static int mounts_release(struct inode *inode, struct file *file)
632 struct proc_mounts *p = file->private_data;
635 return seq_release(inode, file);
638 static unsigned mounts_poll(struct file *file, poll_table *wait)
640 struct proc_mounts *p = file->private_data;
641 unsigned res = POLLIN | POLLRDNORM;
643 poll_wait(file, &p->ns->poll, wait);
644 if (mnt_had_events(p))
645 res |= POLLERR | POLLPRI;
650 static int mounts_open(struct inode *inode, struct file *file)
652 return mounts_open_common(inode, file, &mounts_op);
655 static const struct file_operations proc_mounts_operations = {
659 .release = mounts_release,
663 static int mountinfo_open(struct inode *inode, struct file *file)
665 return mounts_open_common(inode, file, &mountinfo_op);
668 static const struct file_operations proc_mountinfo_operations = {
669 .open = mountinfo_open,
672 .release = mounts_release,
676 static int mountstats_open(struct inode *inode, struct file *file)
678 return mounts_open_common(inode, file, &mountstats_op);
681 static const struct file_operations proc_mountstats_operations = {
682 .open = mountstats_open,
685 .release = mounts_release,
688 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
690 static ssize_t proc_info_read(struct file * file, char __user * buf,
691 size_t count, loff_t *ppos)
693 struct inode * inode = file->f_path.dentry->d_inode;
696 struct task_struct *task = get_proc_task(inode);
702 if (count > PROC_BLOCK_SIZE)
703 count = PROC_BLOCK_SIZE;
706 if (!(page = __get_free_page(GFP_TEMPORARY)))
709 length = PROC_I(inode)->op.proc_read(task, (char*)page);
712 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
715 put_task_struct(task);
720 static const struct file_operations proc_info_file_operations = {
721 .read = proc_info_read,
722 .llseek = generic_file_llseek,
725 static int proc_single_show(struct seq_file *m, void *v)
727 struct inode *inode = m->private;
728 struct pid_namespace *ns;
730 struct task_struct *task;
733 ns = inode->i_sb->s_fs_info;
734 pid = proc_pid(inode);
735 task = get_pid_task(pid, PIDTYPE_PID);
739 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
741 put_task_struct(task);
745 static int proc_single_open(struct inode *inode, struct file *filp)
748 ret = single_open(filp, proc_single_show, NULL);
750 struct seq_file *m = filp->private_data;
757 static const struct file_operations proc_single_file_operations = {
758 .open = proc_single_open,
761 .release = single_release,
764 static int mem_open(struct inode* inode, struct file* file)
766 file->private_data = (void*)((long)current->self_exec_id);
770 static ssize_t mem_read(struct file * file, char __user * buf,
771 size_t count, loff_t *ppos)
773 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
775 unsigned long src = *ppos;
777 struct mm_struct *mm;
782 if (check_mem_permission(task))
786 page = (char *)__get_free_page(GFP_TEMPORARY);
792 mm = get_task_mm(task);
798 if (file->private_data != (void*)((long)current->self_exec_id))
804 int this_len, retval;
806 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
807 retval = access_process_vm(task, src, page, this_len, 0);
808 if (!retval || check_mem_permission(task)) {
814 if (copy_to_user(buf, page, retval)) {
829 free_page((unsigned long) page);
831 put_task_struct(task);
836 #define mem_write NULL
839 /* This is a security hazard */
840 static ssize_t mem_write(struct file * file, const char __user *buf,
841 size_t count, loff_t *ppos)
845 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
846 unsigned long dst = *ppos;
852 if (check_mem_permission(task))
856 page = (char *)__get_free_page(GFP_TEMPORARY);
862 int this_len, retval;
864 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
865 if (copy_from_user(page, buf, this_len)) {
869 retval = access_process_vm(task, dst, page, this_len, 1);
881 free_page((unsigned long) page);
883 put_task_struct(task);
889 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
893 file->f_pos = offset;
896 file->f_pos += offset;
901 force_successful_syscall_return();
905 static const struct file_operations proc_mem_operations = {
912 static ssize_t environ_read(struct file *file, char __user *buf,
913 size_t count, loff_t *ppos)
915 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
917 unsigned long src = *ppos;
919 struct mm_struct *mm;
924 if (!ptrace_may_access(task, PTRACE_MODE_READ))
928 page = (char *)__get_free_page(GFP_TEMPORARY);
934 mm = get_task_mm(task);
939 int this_len, retval, max_len;
941 this_len = mm->env_end - (mm->env_start + src);
946 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
947 this_len = (this_len > max_len) ? max_len : this_len;
949 retval = access_process_vm(task, (mm->env_start + src),
957 if (copy_to_user(buf, page, retval)) {
971 free_page((unsigned long) page);
973 put_task_struct(task);
978 static const struct file_operations proc_environ_operations = {
979 .read = environ_read,
980 .llseek = generic_file_llseek,
983 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
984 size_t count, loff_t *ppos)
986 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
987 char buffer[PROC_NUMBUF];
989 int oom_adjust = OOM_DISABLE;
995 if (lock_task_sighand(task, &flags)) {
996 oom_adjust = task->signal->oom_adj;
997 unlock_task_sighand(task, &flags);
1000 put_task_struct(task);
1002 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1004 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1007 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1008 size_t count, loff_t *ppos)
1010 struct task_struct *task;
1011 char buffer[PROC_NUMBUF];
1013 unsigned long flags;
1016 memset(buffer, 0, sizeof(buffer));
1017 if (count > sizeof(buffer) - 1)
1018 count = sizeof(buffer) - 1;
1019 if (copy_from_user(buffer, buf, count))
1022 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1025 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1026 oom_adjust != OOM_DISABLE)
1029 task = get_proc_task(file->f_path.dentry->d_inode);
1032 if (!lock_task_sighand(task, &flags)) {
1033 put_task_struct(task);
1037 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1038 unlock_task_sighand(task, &flags);
1039 put_task_struct(task);
1043 task->signal->oom_adj = oom_adjust;
1045 unlock_task_sighand(task, &flags);
1046 put_task_struct(task);
1051 static const struct file_operations proc_oom_adjust_operations = {
1052 .read = oom_adjust_read,
1053 .write = oom_adjust_write,
1054 .llseek = generic_file_llseek,
1057 #ifdef CONFIG_AUDITSYSCALL
1058 #define TMPBUFLEN 21
1059 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1060 size_t count, loff_t *ppos)
1062 struct inode * inode = file->f_path.dentry->d_inode;
1063 struct task_struct *task = get_proc_task(inode);
1065 char tmpbuf[TMPBUFLEN];
1069 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1070 audit_get_loginuid(task));
1071 put_task_struct(task);
1072 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1075 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1076 size_t count, loff_t *ppos)
1078 struct inode * inode = file->f_path.dentry->d_inode;
1083 if (!capable(CAP_AUDIT_CONTROL))
1087 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1093 if (count >= PAGE_SIZE)
1094 count = PAGE_SIZE - 1;
1097 /* No partial writes. */
1100 page = (char*)__get_free_page(GFP_TEMPORARY);
1104 if (copy_from_user(page, buf, count))
1108 loginuid = simple_strtoul(page, &tmp, 10);
1114 length = audit_set_loginuid(current, loginuid);
1115 if (likely(length == 0))
1119 free_page((unsigned long) page);
1123 static const struct file_operations proc_loginuid_operations = {
1124 .read = proc_loginuid_read,
1125 .write = proc_loginuid_write,
1126 .llseek = generic_file_llseek,
1129 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1130 size_t count, loff_t *ppos)
1132 struct inode * inode = file->f_path.dentry->d_inode;
1133 struct task_struct *task = get_proc_task(inode);
1135 char tmpbuf[TMPBUFLEN];
1139 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1140 audit_get_sessionid(task));
1141 put_task_struct(task);
1142 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1145 static const struct file_operations proc_sessionid_operations = {
1146 .read = proc_sessionid_read,
1147 .llseek = generic_file_llseek,
1151 #ifdef CONFIG_FAULT_INJECTION
1152 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1153 size_t count, loff_t *ppos)
1155 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1156 char buffer[PROC_NUMBUF];
1162 make_it_fail = task->make_it_fail;
1163 put_task_struct(task);
1165 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1167 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1170 static ssize_t proc_fault_inject_write(struct file * file,
1171 const char __user * buf, size_t count, loff_t *ppos)
1173 struct task_struct *task;
1174 char buffer[PROC_NUMBUF], *end;
1177 if (!capable(CAP_SYS_RESOURCE))
1179 memset(buffer, 0, sizeof(buffer));
1180 if (count > sizeof(buffer) - 1)
1181 count = sizeof(buffer) - 1;
1182 if (copy_from_user(buffer, buf, count))
1184 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1187 task = get_proc_task(file->f_dentry->d_inode);
1190 task->make_it_fail = make_it_fail;
1191 put_task_struct(task);
1196 static const struct file_operations proc_fault_inject_operations = {
1197 .read = proc_fault_inject_read,
1198 .write = proc_fault_inject_write,
1199 .llseek = generic_file_llseek,
1204 #ifdef CONFIG_SCHED_DEBUG
1206 * Print out various scheduling related per-task fields:
1208 static int sched_show(struct seq_file *m, void *v)
1210 struct inode *inode = m->private;
1211 struct task_struct *p;
1213 p = get_proc_task(inode);
1216 proc_sched_show_task(p, m);
1224 sched_write(struct file *file, const char __user *buf,
1225 size_t count, loff_t *offset)
1227 struct inode *inode = file->f_path.dentry->d_inode;
1228 struct task_struct *p;
1230 p = get_proc_task(inode);
1233 proc_sched_set_task(p);
1240 static int sched_open(struct inode *inode, struct file *filp)
1244 ret = single_open(filp, sched_show, NULL);
1246 struct seq_file *m = filp->private_data;
1253 static const struct file_operations proc_pid_sched_operations = {
1256 .write = sched_write,
1257 .llseek = seq_lseek,
1258 .release = single_release,
1263 static ssize_t comm_write(struct file *file, const char __user *buf,
1264 size_t count, loff_t *offset)
1266 struct inode *inode = file->f_path.dentry->d_inode;
1267 struct task_struct *p;
1268 char buffer[TASK_COMM_LEN];
1270 memset(buffer, 0, sizeof(buffer));
1271 if (count > sizeof(buffer) - 1)
1272 count = sizeof(buffer) - 1;
1273 if (copy_from_user(buffer, buf, count))
1276 p = get_proc_task(inode);
1280 if (same_thread_group(current, p))
1281 set_task_comm(p, buffer);
1290 static int comm_show(struct seq_file *m, void *v)
1292 struct inode *inode = m->private;
1293 struct task_struct *p;
1295 p = get_proc_task(inode);
1300 seq_printf(m, "%s\n", p->comm);
1308 static int comm_open(struct inode *inode, struct file *filp)
1312 ret = single_open(filp, comm_show, NULL);
1314 struct seq_file *m = filp->private_data;
1321 static const struct file_operations proc_pid_set_comm_operations = {
1324 .write = comm_write,
1325 .llseek = seq_lseek,
1326 .release = single_release,
1330 * We added or removed a vma mapping the executable. The vmas are only mapped
1331 * during exec and are not mapped with the mmap system call.
1332 * Callers must hold down_write() on the mm's mmap_sem for these
1334 void added_exe_file_vma(struct mm_struct *mm)
1336 mm->num_exe_file_vmas++;
1339 void removed_exe_file_vma(struct mm_struct *mm)
1341 mm->num_exe_file_vmas--;
1342 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1344 mm->exe_file = NULL;
1349 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1352 get_file(new_exe_file);
1355 mm->exe_file = new_exe_file;
1356 mm->num_exe_file_vmas = 0;
1359 struct file *get_mm_exe_file(struct mm_struct *mm)
1361 struct file *exe_file;
1363 /* We need mmap_sem to protect against races with removal of
1364 * VM_EXECUTABLE vmas */
1365 down_read(&mm->mmap_sem);
1366 exe_file = mm->exe_file;
1369 up_read(&mm->mmap_sem);
1373 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1375 /* It's safe to write the exe_file pointer without exe_file_lock because
1376 * this is called during fork when the task is not yet in /proc */
1377 newmm->exe_file = get_mm_exe_file(oldmm);
1380 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1382 struct task_struct *task;
1383 struct mm_struct *mm;
1384 struct file *exe_file;
1386 task = get_proc_task(inode);
1389 mm = get_task_mm(task);
1390 put_task_struct(task);
1393 exe_file = get_mm_exe_file(mm);
1396 *exe_path = exe_file->f_path;
1397 path_get(&exe_file->f_path);
1404 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1406 struct inode *inode = dentry->d_inode;
1407 int error = -EACCES;
1409 /* We don't need a base pointer in the /proc filesystem */
1410 path_put(&nd->path);
1412 /* Are we allowed to snoop on the tasks file descriptors? */
1413 if (!proc_fd_access_allowed(inode))
1416 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1418 return ERR_PTR(error);
1421 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1423 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1430 pathname = d_path(path, tmp, PAGE_SIZE);
1431 len = PTR_ERR(pathname);
1432 if (IS_ERR(pathname))
1434 len = tmp + PAGE_SIZE - 1 - pathname;
1438 if (copy_to_user(buffer, pathname, len))
1441 free_page((unsigned long)tmp);
1445 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1447 int error = -EACCES;
1448 struct inode *inode = dentry->d_inode;
1451 /* Are we allowed to snoop on the tasks file descriptors? */
1452 if (!proc_fd_access_allowed(inode))
1455 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1459 error = do_proc_readlink(&path, buffer, buflen);
1465 static const struct inode_operations proc_pid_link_inode_operations = {
1466 .readlink = proc_pid_readlink,
1467 .follow_link = proc_pid_follow_link,
1468 .setattr = proc_setattr,
1472 /* building an inode */
1474 static int task_dumpable(struct task_struct *task)
1477 struct mm_struct *mm;
1482 dumpable = get_dumpable(mm);
1490 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1492 struct inode * inode;
1493 struct proc_inode *ei;
1494 const struct cred *cred;
1496 /* We need a new inode */
1498 inode = new_inode(sb);
1504 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1505 inode->i_op = &proc_def_inode_operations;
1508 * grab the reference to task.
1510 ei->pid = get_task_pid(task, PIDTYPE_PID);
1514 if (task_dumpable(task)) {
1516 cred = __task_cred(task);
1517 inode->i_uid = cred->euid;
1518 inode->i_gid = cred->egid;
1521 security_task_to_inode(task, inode);
1531 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1533 struct inode *inode = dentry->d_inode;
1534 struct task_struct *task;
1535 const struct cred *cred;
1537 generic_fillattr(inode, stat);
1542 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1544 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1545 task_dumpable(task)) {
1546 cred = __task_cred(task);
1547 stat->uid = cred->euid;
1548 stat->gid = cred->egid;
1558 * Exceptional case: normally we are not allowed to unhash a busy
1559 * directory. In this case, however, we can do it - no aliasing problems
1560 * due to the way we treat inodes.
1562 * Rewrite the inode's ownerships here because the owning task may have
1563 * performed a setuid(), etc.
1565 * Before the /proc/pid/status file was created the only way to read
1566 * the effective uid of a /process was to stat /proc/pid. Reading
1567 * /proc/pid/status is slow enough that procps and other packages
1568 * kept stating /proc/pid. To keep the rules in /proc simple I have
1569 * made this apply to all per process world readable and executable
1572 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1574 struct inode *inode = dentry->d_inode;
1575 struct task_struct *task = get_proc_task(inode);
1576 const struct cred *cred;
1579 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1580 task_dumpable(task)) {
1582 cred = __task_cred(task);
1583 inode->i_uid = cred->euid;
1584 inode->i_gid = cred->egid;
1590 inode->i_mode &= ~(S_ISUID | S_ISGID);
1591 security_task_to_inode(task, inode);
1592 put_task_struct(task);
1599 static int pid_delete_dentry(struct dentry * dentry)
1601 /* Is the task we represent dead?
1602 * If so, then don't put the dentry on the lru list,
1603 * kill it immediately.
1605 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1608 static const struct dentry_operations pid_dentry_operations =
1610 .d_revalidate = pid_revalidate,
1611 .d_delete = pid_delete_dentry,
1616 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1617 struct task_struct *, const void *);
1620 * Fill a directory entry.
1622 * If possible create the dcache entry and derive our inode number and
1623 * file type from dcache entry.
1625 * Since all of the proc inode numbers are dynamically generated, the inode
1626 * numbers do not exist until the inode is cache. This means creating the
1627 * the dcache entry in readdir is necessary to keep the inode numbers
1628 * reported by readdir in sync with the inode numbers reported
1631 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1632 char *name, int len,
1633 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1635 struct dentry *child, *dir = filp->f_path.dentry;
1636 struct inode *inode;
1639 unsigned type = DT_UNKNOWN;
1643 qname.hash = full_name_hash(name, len);
1645 child = d_lookup(dir, &qname);
1648 new = d_alloc(dir, &qname);
1650 child = instantiate(dir->d_inode, new, task, ptr);
1657 if (!child || IS_ERR(child) || !child->d_inode)
1658 goto end_instantiate;
1659 inode = child->d_inode;
1662 type = inode->i_mode >> 12;
1667 ino = find_inode_number(dir, &qname);
1670 return filldir(dirent, name, len, filp->f_pos, ino, type);
1673 static unsigned name_to_int(struct dentry *dentry)
1675 const char *name = dentry->d_name.name;
1676 int len = dentry->d_name.len;
1679 if (len > 1 && *name == '0')
1682 unsigned c = *name++ - '0';
1685 if (n >= (~0U-9)/10)
1695 #define PROC_FDINFO_MAX 64
1697 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1699 struct task_struct *task = get_proc_task(inode);
1700 struct files_struct *files = NULL;
1702 int fd = proc_fd(inode);
1705 files = get_files_struct(task);
1706 put_task_struct(task);
1710 * We are not taking a ref to the file structure, so we must
1713 spin_lock(&files->file_lock);
1714 file = fcheck_files(files, fd);
1717 *path = file->f_path;
1718 path_get(&file->f_path);
1721 snprintf(info, PROC_FDINFO_MAX,
1724 (long long) file->f_pos,
1726 spin_unlock(&files->file_lock);
1727 put_files_struct(files);
1730 spin_unlock(&files->file_lock);
1731 put_files_struct(files);
1736 static int proc_fd_link(struct inode *inode, struct path *path)
1738 return proc_fd_info(inode, path, NULL);
1741 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1743 struct inode *inode = dentry->d_inode;
1744 struct task_struct *task = get_proc_task(inode);
1745 int fd = proc_fd(inode);
1746 struct files_struct *files;
1747 const struct cred *cred;
1750 files = get_files_struct(task);
1753 if (fcheck_files(files, fd)) {
1755 put_files_struct(files);
1756 if (task_dumpable(task)) {
1758 cred = __task_cred(task);
1759 inode->i_uid = cred->euid;
1760 inode->i_gid = cred->egid;
1766 inode->i_mode &= ~(S_ISUID | S_ISGID);
1767 security_task_to_inode(task, inode);
1768 put_task_struct(task);
1772 put_files_struct(files);
1774 put_task_struct(task);
1780 static const struct dentry_operations tid_fd_dentry_operations =
1782 .d_revalidate = tid_fd_revalidate,
1783 .d_delete = pid_delete_dentry,
1786 static struct dentry *proc_fd_instantiate(struct inode *dir,
1787 struct dentry *dentry, struct task_struct *task, const void *ptr)
1789 unsigned fd = *(const unsigned *)ptr;
1791 struct files_struct *files;
1792 struct inode *inode;
1793 struct proc_inode *ei;
1794 struct dentry *error = ERR_PTR(-ENOENT);
1796 inode = proc_pid_make_inode(dir->i_sb, task);
1801 files = get_files_struct(task);
1804 inode->i_mode = S_IFLNK;
1807 * We are not taking a ref to the file structure, so we must
1810 spin_lock(&files->file_lock);
1811 file = fcheck_files(files, fd);
1814 if (file->f_mode & FMODE_READ)
1815 inode->i_mode |= S_IRUSR | S_IXUSR;
1816 if (file->f_mode & FMODE_WRITE)
1817 inode->i_mode |= S_IWUSR | S_IXUSR;
1818 spin_unlock(&files->file_lock);
1819 put_files_struct(files);
1821 inode->i_op = &proc_pid_link_inode_operations;
1823 ei->op.proc_get_link = proc_fd_link;
1824 dentry->d_op = &tid_fd_dentry_operations;
1825 d_add(dentry, inode);
1826 /* Close the race of the process dying before we return the dentry */
1827 if (tid_fd_revalidate(dentry, NULL))
1833 spin_unlock(&files->file_lock);
1834 put_files_struct(files);
1840 static struct dentry *proc_lookupfd_common(struct inode *dir,
1841 struct dentry *dentry,
1842 instantiate_t instantiate)
1844 struct task_struct *task = get_proc_task(dir);
1845 unsigned fd = name_to_int(dentry);
1846 struct dentry *result = ERR_PTR(-ENOENT);
1853 result = instantiate(dir, dentry, task, &fd);
1855 put_task_struct(task);
1860 static int proc_readfd_common(struct file * filp, void * dirent,
1861 filldir_t filldir, instantiate_t instantiate)
1863 struct dentry *dentry = filp->f_path.dentry;
1864 struct inode *inode = dentry->d_inode;
1865 struct task_struct *p = get_proc_task(inode);
1866 unsigned int fd, ino;
1868 struct files_struct * files;
1878 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1882 ino = parent_ino(dentry);
1883 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1887 files = get_files_struct(p);
1891 for (fd = filp->f_pos-2;
1892 fd < files_fdtable(files)->max_fds;
1893 fd++, filp->f_pos++) {
1894 char name[PROC_NUMBUF];
1897 if (!fcheck_files(files, fd))
1901 len = snprintf(name, sizeof(name), "%d", fd);
1902 if (proc_fill_cache(filp, dirent, filldir,
1903 name, len, instantiate,
1911 put_files_struct(files);
1919 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1920 struct nameidata *nd)
1922 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1925 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1927 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1930 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1931 size_t len, loff_t *ppos)
1933 char tmp[PROC_FDINFO_MAX];
1934 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1936 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1940 static const struct file_operations proc_fdinfo_file_operations = {
1941 .open = nonseekable_open,
1942 .read = proc_fdinfo_read,
1945 static const struct file_operations proc_fd_operations = {
1946 .read = generic_read_dir,
1947 .readdir = proc_readfd,
1951 * /proc/pid/fd needs a special permission handler so that a process can still
1952 * access /proc/self/fd after it has executed a setuid().
1954 static int proc_fd_permission(struct inode *inode, int mask)
1958 rv = generic_permission(inode, mask, NULL);
1961 if (task_pid(current) == proc_pid(inode))
1967 * proc directories can do almost nothing..
1969 static const struct inode_operations proc_fd_inode_operations = {
1970 .lookup = proc_lookupfd,
1971 .permission = proc_fd_permission,
1972 .setattr = proc_setattr,
1975 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1976 struct dentry *dentry, struct task_struct *task, const void *ptr)
1978 unsigned fd = *(unsigned *)ptr;
1979 struct inode *inode;
1980 struct proc_inode *ei;
1981 struct dentry *error = ERR_PTR(-ENOENT);
1983 inode = proc_pid_make_inode(dir->i_sb, task);
1988 inode->i_mode = S_IFREG | S_IRUSR;
1989 inode->i_fop = &proc_fdinfo_file_operations;
1990 dentry->d_op = &tid_fd_dentry_operations;
1991 d_add(dentry, inode);
1992 /* Close the race of the process dying before we return the dentry */
1993 if (tid_fd_revalidate(dentry, NULL))
2000 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2001 struct dentry *dentry,
2002 struct nameidata *nd)
2004 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2007 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2009 return proc_readfd_common(filp, dirent, filldir,
2010 proc_fdinfo_instantiate);
2013 static const struct file_operations proc_fdinfo_operations = {
2014 .read = generic_read_dir,
2015 .readdir = proc_readfdinfo,
2019 * proc directories can do almost nothing..
2021 static const struct inode_operations proc_fdinfo_inode_operations = {
2022 .lookup = proc_lookupfdinfo,
2023 .setattr = proc_setattr,
2027 static struct dentry *proc_pident_instantiate(struct inode *dir,
2028 struct dentry *dentry, struct task_struct *task, const void *ptr)
2030 const struct pid_entry *p = ptr;
2031 struct inode *inode;
2032 struct proc_inode *ei;
2033 struct dentry *error = ERR_PTR(-ENOENT);
2035 inode = proc_pid_make_inode(dir->i_sb, task);
2040 inode->i_mode = p->mode;
2041 if (S_ISDIR(inode->i_mode))
2042 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2044 inode->i_op = p->iop;
2046 inode->i_fop = p->fop;
2048 dentry->d_op = &pid_dentry_operations;
2049 d_add(dentry, inode);
2050 /* Close the race of the process dying before we return the dentry */
2051 if (pid_revalidate(dentry, NULL))
2057 static struct dentry *proc_pident_lookup(struct inode *dir,
2058 struct dentry *dentry,
2059 const struct pid_entry *ents,
2062 struct dentry *error;
2063 struct task_struct *task = get_proc_task(dir);
2064 const struct pid_entry *p, *last;
2066 error = ERR_PTR(-ENOENT);
2072 * Yes, it does not scale. And it should not. Don't add
2073 * new entries into /proc/<tgid>/ without very good reasons.
2075 last = &ents[nents - 1];
2076 for (p = ents; p <= last; p++) {
2077 if (p->len != dentry->d_name.len)
2079 if (!memcmp(dentry->d_name.name, p->name, p->len))
2085 error = proc_pident_instantiate(dir, dentry, task, p);
2087 put_task_struct(task);
2092 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2093 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2095 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2096 proc_pident_instantiate, task, p);
2099 static int proc_pident_readdir(struct file *filp,
2100 void *dirent, filldir_t filldir,
2101 const struct pid_entry *ents, unsigned int nents)
2104 struct dentry *dentry = filp->f_path.dentry;
2105 struct inode *inode = dentry->d_inode;
2106 struct task_struct *task = get_proc_task(inode);
2107 const struct pid_entry *p, *last;
2120 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2126 ino = parent_ino(dentry);
2127 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2139 last = &ents[nents - 1];
2141 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2150 put_task_struct(task);
2155 #ifdef CONFIG_SECURITY
2156 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2157 size_t count, loff_t *ppos)
2159 struct inode * inode = file->f_path.dentry->d_inode;
2162 struct task_struct *task = get_proc_task(inode);
2167 length = security_getprocattr(task,
2168 (char*)file->f_path.dentry->d_name.name,
2170 put_task_struct(task);
2172 length = simple_read_from_buffer(buf, count, ppos, p, length);
2177 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2178 size_t count, loff_t *ppos)
2180 struct inode * inode = file->f_path.dentry->d_inode;
2183 struct task_struct *task = get_proc_task(inode);
2188 if (count > PAGE_SIZE)
2191 /* No partial writes. */
2197 page = (char*)__get_free_page(GFP_TEMPORARY);
2202 if (copy_from_user(page, buf, count))
2205 /* Guard against adverse ptrace interaction */
2206 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2210 length = security_setprocattr(task,
2211 (char*)file->f_path.dentry->d_name.name,
2212 (void*)page, count);
2213 mutex_unlock(&task->cred_guard_mutex);
2215 free_page((unsigned long) page);
2217 put_task_struct(task);
2222 static const struct file_operations proc_pid_attr_operations = {
2223 .read = proc_pid_attr_read,
2224 .write = proc_pid_attr_write,
2225 .llseek = generic_file_llseek,
2228 static const struct pid_entry attr_dir_stuff[] = {
2229 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2230 REG("prev", S_IRUGO, proc_pid_attr_operations),
2231 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2232 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2233 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2234 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2237 static int proc_attr_dir_readdir(struct file * filp,
2238 void * dirent, filldir_t filldir)
2240 return proc_pident_readdir(filp,dirent,filldir,
2241 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2244 static const struct file_operations proc_attr_dir_operations = {
2245 .read = generic_read_dir,
2246 .readdir = proc_attr_dir_readdir,
2249 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2250 struct dentry *dentry, struct nameidata *nd)
2252 return proc_pident_lookup(dir, dentry,
2253 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2256 static const struct inode_operations proc_attr_dir_inode_operations = {
2257 .lookup = proc_attr_dir_lookup,
2258 .getattr = pid_getattr,
2259 .setattr = proc_setattr,
2264 #ifdef CONFIG_ELF_CORE
2265 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2266 size_t count, loff_t *ppos)
2268 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2269 struct mm_struct *mm;
2270 char buffer[PROC_NUMBUF];
2278 mm = get_task_mm(task);
2280 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2281 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2282 MMF_DUMP_FILTER_SHIFT));
2284 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2287 put_task_struct(task);
2292 static ssize_t proc_coredump_filter_write(struct file *file,
2293 const char __user *buf,
2297 struct task_struct *task;
2298 struct mm_struct *mm;
2299 char buffer[PROC_NUMBUF], *end;
2306 memset(buffer, 0, sizeof(buffer));
2307 if (count > sizeof(buffer) - 1)
2308 count = sizeof(buffer) - 1;
2309 if (copy_from_user(buffer, buf, count))
2313 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2316 if (end - buffer == 0)
2320 task = get_proc_task(file->f_dentry->d_inode);
2325 mm = get_task_mm(task);
2329 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2331 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2333 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2338 put_task_struct(task);
2343 static const struct file_operations proc_coredump_filter_operations = {
2344 .read = proc_coredump_filter_read,
2345 .write = proc_coredump_filter_write,
2346 .llseek = generic_file_llseek,
2353 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2356 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2357 pid_t tgid = task_tgid_nr_ns(current, ns);
2358 char tmp[PROC_NUMBUF];
2361 sprintf(tmp, "%d", tgid);
2362 return vfs_readlink(dentry,buffer,buflen,tmp);
2365 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2367 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2368 pid_t tgid = task_tgid_nr_ns(current, ns);
2369 char *name = ERR_PTR(-ENOENT);
2373 name = ERR_PTR(-ENOMEM);
2375 sprintf(name, "%d", tgid);
2377 nd_set_link(nd, name);
2381 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2384 char *s = nd_get_link(nd);
2389 static const struct inode_operations proc_self_inode_operations = {
2390 .readlink = proc_self_readlink,
2391 .follow_link = proc_self_follow_link,
2392 .put_link = proc_self_put_link,
2398 * These are the directory entries in the root directory of /proc
2399 * that properly belong to the /proc filesystem, as they describe
2400 * describe something that is process related.
2402 static const struct pid_entry proc_base_stuff[] = {
2403 NOD("self", S_IFLNK|S_IRWXUGO,
2404 &proc_self_inode_operations, NULL, {}),
2408 * Exceptional case: normally we are not allowed to unhash a busy
2409 * directory. In this case, however, we can do it - no aliasing problems
2410 * due to the way we treat inodes.
2412 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2414 struct inode *inode = dentry->d_inode;
2415 struct task_struct *task = get_proc_task(inode);
2417 put_task_struct(task);
2424 static const struct dentry_operations proc_base_dentry_operations =
2426 .d_revalidate = proc_base_revalidate,
2427 .d_delete = pid_delete_dentry,
2430 static struct dentry *proc_base_instantiate(struct inode *dir,
2431 struct dentry *dentry, struct task_struct *task, const void *ptr)
2433 const struct pid_entry *p = ptr;
2434 struct inode *inode;
2435 struct proc_inode *ei;
2436 struct dentry *error;
2438 /* Allocate the inode */
2439 error = ERR_PTR(-ENOMEM);
2440 inode = new_inode(dir->i_sb);
2444 /* Initialize the inode */
2446 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2449 * grab the reference to the task.
2451 ei->pid = get_task_pid(task, PIDTYPE_PID);
2455 inode->i_mode = p->mode;
2456 if (S_ISDIR(inode->i_mode))
2458 if (S_ISLNK(inode->i_mode))
2461 inode->i_op = p->iop;
2463 inode->i_fop = p->fop;
2465 dentry->d_op = &proc_base_dentry_operations;
2466 d_add(dentry, inode);
2475 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2477 struct dentry *error;
2478 struct task_struct *task = get_proc_task(dir);
2479 const struct pid_entry *p, *last;
2481 error = ERR_PTR(-ENOENT);
2486 /* Lookup the directory entry */
2487 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2488 for (p = proc_base_stuff; p <= last; p++) {
2489 if (p->len != dentry->d_name.len)
2491 if (!memcmp(dentry->d_name.name, p->name, p->len))
2497 error = proc_base_instantiate(dir, dentry, task, p);
2500 put_task_struct(task);
2505 static int proc_base_fill_cache(struct file *filp, void *dirent,
2506 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2508 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2509 proc_base_instantiate, task, p);
2512 #ifdef CONFIG_TASK_IO_ACCOUNTING
2513 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2515 struct task_io_accounting acct = task->ioac;
2516 unsigned long flags;
2518 if (whole && lock_task_sighand(task, &flags)) {
2519 struct task_struct *t = task;
2521 task_io_accounting_add(&acct, &task->signal->ioac);
2522 while_each_thread(task, t)
2523 task_io_accounting_add(&acct, &t->ioac);
2525 unlock_task_sighand(task, &flags);
2527 return sprintf(buffer,
2532 "read_bytes: %llu\n"
2533 "write_bytes: %llu\n"
2534 "cancelled_write_bytes: %llu\n",
2535 (unsigned long long)acct.rchar,
2536 (unsigned long long)acct.wchar,
2537 (unsigned long long)acct.syscr,
2538 (unsigned long long)acct.syscw,
2539 (unsigned long long)acct.read_bytes,
2540 (unsigned long long)acct.write_bytes,
2541 (unsigned long long)acct.cancelled_write_bytes);
2544 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2546 return do_io_accounting(task, buffer, 0);
2549 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2551 return do_io_accounting(task, buffer, 1);
2553 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2555 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2556 struct pid *pid, struct task_struct *task)
2558 seq_printf(m, "%08x\n", task->personality);
2565 static const struct file_operations proc_task_operations;
2566 static const struct inode_operations proc_task_inode_operations;
2568 static const struct pid_entry tgid_base_stuff[] = {
2569 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2570 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2571 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2573 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2575 REG("environ", S_IRUSR, proc_environ_operations),
2576 INF("auxv", S_IRUSR, proc_pid_auxv),
2577 ONE("status", S_IRUGO, proc_pid_status),
2578 ONE("personality", S_IRUSR, proc_pid_personality),
2579 INF("limits", S_IRUSR, proc_pid_limits),
2580 #ifdef CONFIG_SCHED_DEBUG
2581 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2583 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2584 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2585 INF("syscall", S_IRUSR, proc_pid_syscall),
2587 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2588 ONE("stat", S_IRUGO, proc_tgid_stat),
2589 ONE("statm", S_IRUGO, proc_pid_statm),
2590 REG("maps", S_IRUGO, proc_maps_operations),
2592 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2594 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2595 LNK("cwd", proc_cwd_link),
2596 LNK("root", proc_root_link),
2597 LNK("exe", proc_exe_link),
2598 REG("mounts", S_IRUGO, proc_mounts_operations),
2599 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2600 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2601 #ifdef CONFIG_PROC_PAGE_MONITOR
2602 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2603 REG("smaps", S_IRUGO, proc_smaps_operations),
2604 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2606 #ifdef CONFIG_SECURITY
2607 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2609 #ifdef CONFIG_KALLSYMS
2610 INF("wchan", S_IRUGO, proc_pid_wchan),
2612 #ifdef CONFIG_STACKTRACE
2613 ONE("stack", S_IRUSR, proc_pid_stack),
2615 #ifdef CONFIG_SCHEDSTATS
2616 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2618 #ifdef CONFIG_LATENCYTOP
2619 REG("latency", S_IRUGO, proc_lstats_operations),
2621 #ifdef CONFIG_PROC_PID_CPUSET
2622 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2624 #ifdef CONFIG_CGROUPS
2625 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2627 INF("oom_score", S_IRUGO, proc_oom_score),
2628 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2629 #ifdef CONFIG_AUDITSYSCALL
2630 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2631 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2633 #ifdef CONFIG_FAULT_INJECTION
2634 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2636 #ifdef CONFIG_ELF_CORE
2637 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2639 #ifdef CONFIG_TASK_IO_ACCOUNTING
2640 INF("io", S_IRUGO, proc_tgid_io_accounting),
2644 static int proc_tgid_base_readdir(struct file * filp,
2645 void * dirent, filldir_t filldir)
2647 return proc_pident_readdir(filp,dirent,filldir,
2648 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2651 static const struct file_operations proc_tgid_base_operations = {
2652 .read = generic_read_dir,
2653 .readdir = proc_tgid_base_readdir,
2656 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2657 return proc_pident_lookup(dir, dentry,
2658 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2661 static const struct inode_operations proc_tgid_base_inode_operations = {
2662 .lookup = proc_tgid_base_lookup,
2663 .getattr = pid_getattr,
2664 .setattr = proc_setattr,
2667 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2669 struct dentry *dentry, *leader, *dir;
2670 char buf[PROC_NUMBUF];
2674 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2675 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2677 shrink_dcache_parent(dentry);
2683 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2684 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2689 name.len = strlen(name.name);
2690 dir = d_hash_and_lookup(leader, &name);
2692 goto out_put_leader;
2695 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2696 dentry = d_hash_and_lookup(dir, &name);
2698 shrink_dcache_parent(dentry);
2711 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2712 * @task: task that should be flushed.
2714 * When flushing dentries from proc, one needs to flush them from global
2715 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2716 * in. This call is supposed to do all of this job.
2718 * Looks in the dcache for
2720 * /proc/@tgid/task/@pid
2721 * if either directory is present flushes it and all of it'ts children
2724 * It is safe and reasonable to cache /proc entries for a task until
2725 * that task exits. After that they just clog up the dcache with
2726 * useless entries, possibly causing useful dcache entries to be
2727 * flushed instead. This routine is proved to flush those useless
2728 * dcache entries at process exit time.
2730 * NOTE: This routine is just an optimization so it does not guarantee
2731 * that no dcache entries will exist at process exit time it
2732 * just makes it very unlikely that any will persist.
2735 void proc_flush_task(struct task_struct *task)
2738 struct pid *pid, *tgid;
2741 pid = task_pid(task);
2742 tgid = task_tgid(task);
2744 for (i = 0; i <= pid->level; i++) {
2745 upid = &pid->numbers[i];
2746 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2747 tgid->numbers[i].nr);
2750 upid = &pid->numbers[pid->level];
2752 pid_ns_release_proc(upid->ns);
2755 static struct dentry *proc_pid_instantiate(struct inode *dir,
2756 struct dentry * dentry,
2757 struct task_struct *task, const void *ptr)
2759 struct dentry *error = ERR_PTR(-ENOENT);
2760 struct inode *inode;
2762 inode = proc_pid_make_inode(dir->i_sb, task);
2766 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2767 inode->i_op = &proc_tgid_base_inode_operations;
2768 inode->i_fop = &proc_tgid_base_operations;
2769 inode->i_flags|=S_IMMUTABLE;
2771 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2772 ARRAY_SIZE(tgid_base_stuff));
2774 dentry->d_op = &pid_dentry_operations;
2776 d_add(dentry, inode);
2777 /* Close the race of the process dying before we return the dentry */
2778 if (pid_revalidate(dentry, NULL))
2784 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2786 struct dentry *result;
2787 struct task_struct *task;
2789 struct pid_namespace *ns;
2791 result = proc_base_lookup(dir, dentry);
2792 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2795 tgid = name_to_int(dentry);
2799 ns = dentry->d_sb->s_fs_info;
2801 task = find_task_by_pid_ns(tgid, ns);
2803 get_task_struct(task);
2808 result = proc_pid_instantiate(dir, dentry, task, NULL);
2809 put_task_struct(task);
2815 * Find the first task with tgid >= tgid
2820 struct task_struct *task;
2822 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2827 put_task_struct(iter.task);
2831 pid = find_ge_pid(iter.tgid, ns);
2833 iter.tgid = pid_nr_ns(pid, ns);
2834 iter.task = pid_task(pid, PIDTYPE_PID);
2835 /* What we to know is if the pid we have find is the
2836 * pid of a thread_group_leader. Testing for task
2837 * being a thread_group_leader is the obvious thing
2838 * todo but there is a window when it fails, due to
2839 * the pid transfer logic in de_thread.
2841 * So we perform the straight forward test of seeing
2842 * if the pid we have found is the pid of a thread
2843 * group leader, and don't worry if the task we have
2844 * found doesn't happen to be a thread group leader.
2845 * As we don't care in the case of readdir.
2847 if (!iter.task || !has_group_leader_pid(iter.task)) {
2851 get_task_struct(iter.task);
2857 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2859 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2860 struct tgid_iter iter)
2862 char name[PROC_NUMBUF];
2863 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2864 return proc_fill_cache(filp, dirent, filldir, name, len,
2865 proc_pid_instantiate, iter.task, NULL);
2868 /* for the /proc/ directory itself, after non-process stuff has been done */
2869 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2871 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2872 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2873 struct tgid_iter iter;
2874 struct pid_namespace *ns;
2879 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2880 const struct pid_entry *p = &proc_base_stuff[nr];
2881 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2885 ns = filp->f_dentry->d_sb->s_fs_info;
2887 iter.tgid = filp->f_pos - TGID_OFFSET;
2888 for (iter = next_tgid(ns, iter);
2890 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2891 filp->f_pos = iter.tgid + TGID_OFFSET;
2892 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2893 put_task_struct(iter.task);
2897 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2899 put_task_struct(reaper);
2907 static const struct pid_entry tid_base_stuff[] = {
2908 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2909 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2910 REG("environ", S_IRUSR, proc_environ_operations),
2911 INF("auxv", S_IRUSR, proc_pid_auxv),
2912 ONE("status", S_IRUGO, proc_pid_status),
2913 ONE("personality", S_IRUSR, proc_pid_personality),
2914 INF("limits", S_IRUSR, proc_pid_limits),
2915 #ifdef CONFIG_SCHED_DEBUG
2916 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2918 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2919 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2920 INF("syscall", S_IRUSR, proc_pid_syscall),
2922 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2923 ONE("stat", S_IRUGO, proc_tid_stat),
2924 ONE("statm", S_IRUGO, proc_pid_statm),
2925 REG("maps", S_IRUGO, proc_maps_operations),
2927 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2929 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2930 LNK("cwd", proc_cwd_link),
2931 LNK("root", proc_root_link),
2932 LNK("exe", proc_exe_link),
2933 REG("mounts", S_IRUGO, proc_mounts_operations),
2934 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2935 #ifdef CONFIG_PROC_PAGE_MONITOR
2936 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2937 REG("smaps", S_IRUGO, proc_smaps_operations),
2938 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2940 #ifdef CONFIG_SECURITY
2941 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2943 #ifdef CONFIG_KALLSYMS
2944 INF("wchan", S_IRUGO, proc_pid_wchan),
2946 #ifdef CONFIG_STACKTRACE
2947 ONE("stack", S_IRUSR, proc_pid_stack),
2949 #ifdef CONFIG_SCHEDSTATS
2950 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2952 #ifdef CONFIG_LATENCYTOP
2953 REG("latency", S_IRUGO, proc_lstats_operations),
2955 #ifdef CONFIG_PROC_PID_CPUSET
2956 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2958 #ifdef CONFIG_CGROUPS
2959 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2961 INF("oom_score", S_IRUGO, proc_oom_score),
2962 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2963 #ifdef CONFIG_AUDITSYSCALL
2964 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2965 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2967 #ifdef CONFIG_FAULT_INJECTION
2968 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2970 #ifdef CONFIG_TASK_IO_ACCOUNTING
2971 INF("io", S_IRUGO, proc_tid_io_accounting),
2975 static int proc_tid_base_readdir(struct file * filp,
2976 void * dirent, filldir_t filldir)
2978 return proc_pident_readdir(filp,dirent,filldir,
2979 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2982 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2983 return proc_pident_lookup(dir, dentry,
2984 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2987 static const struct file_operations proc_tid_base_operations = {
2988 .read = generic_read_dir,
2989 .readdir = proc_tid_base_readdir,
2992 static const struct inode_operations proc_tid_base_inode_operations = {
2993 .lookup = proc_tid_base_lookup,
2994 .getattr = pid_getattr,
2995 .setattr = proc_setattr,
2998 static struct dentry *proc_task_instantiate(struct inode *dir,
2999 struct dentry *dentry, struct task_struct *task, const void *ptr)
3001 struct dentry *error = ERR_PTR(-ENOENT);
3002 struct inode *inode;
3003 inode = proc_pid_make_inode(dir->i_sb, task);
3007 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3008 inode->i_op = &proc_tid_base_inode_operations;
3009 inode->i_fop = &proc_tid_base_operations;
3010 inode->i_flags|=S_IMMUTABLE;
3012 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3013 ARRAY_SIZE(tid_base_stuff));
3015 dentry->d_op = &pid_dentry_operations;
3017 d_add(dentry, inode);
3018 /* Close the race of the process dying before we return the dentry */
3019 if (pid_revalidate(dentry, NULL))
3025 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3027 struct dentry *result = ERR_PTR(-ENOENT);
3028 struct task_struct *task;
3029 struct task_struct *leader = get_proc_task(dir);
3031 struct pid_namespace *ns;
3036 tid = name_to_int(dentry);
3040 ns = dentry->d_sb->s_fs_info;
3042 task = find_task_by_pid_ns(tid, ns);
3044 get_task_struct(task);
3048 if (!same_thread_group(leader, task))
3051 result = proc_task_instantiate(dir, dentry, task, NULL);
3053 put_task_struct(task);
3055 put_task_struct(leader);
3061 * Find the first tid of a thread group to return to user space.
3063 * Usually this is just the thread group leader, but if the users
3064 * buffer was too small or there was a seek into the middle of the
3065 * directory we have more work todo.
3067 * In the case of a short read we start with find_task_by_pid.
3069 * In the case of a seek we start with the leader and walk nr
3072 static struct task_struct *first_tid(struct task_struct *leader,
3073 int tid, int nr, struct pid_namespace *ns)
3075 struct task_struct *pos;
3078 /* Attempt to start with the pid of a thread */
3079 if (tid && (nr > 0)) {
3080 pos = find_task_by_pid_ns(tid, ns);
3081 if (pos && (pos->group_leader == leader))
3085 /* If nr exceeds the number of threads there is nothing todo */
3087 if (nr && nr >= get_nr_threads(leader))
3090 /* If we haven't found our starting place yet start
3091 * with the leader and walk nr threads forward.
3093 for (pos = leader; nr > 0; --nr) {
3094 pos = next_thread(pos);
3095 if (pos == leader) {
3101 get_task_struct(pos);
3108 * Find the next thread in the thread list.
3109 * Return NULL if there is an error or no next thread.
3111 * The reference to the input task_struct is released.
3113 static struct task_struct *next_tid(struct task_struct *start)
3115 struct task_struct *pos = NULL;
3117 if (pid_alive(start)) {
3118 pos = next_thread(start);
3119 if (thread_group_leader(pos))
3122 get_task_struct(pos);
3125 put_task_struct(start);
3129 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3130 struct task_struct *task, int tid)
3132 char name[PROC_NUMBUF];
3133 int len = snprintf(name, sizeof(name), "%d", tid);
3134 return proc_fill_cache(filp, dirent, filldir, name, len,
3135 proc_task_instantiate, task, NULL);
3138 /* for the /proc/TGID/task/ directories */
3139 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3141 struct dentry *dentry = filp->f_path.dentry;
3142 struct inode *inode = dentry->d_inode;
3143 struct task_struct *leader = NULL;
3144 struct task_struct *task;
3145 int retval = -ENOENT;
3148 struct pid_namespace *ns;
3150 task = get_proc_task(inode);
3154 if (pid_alive(task)) {
3155 leader = task->group_leader;
3156 get_task_struct(leader);
3159 put_task_struct(task);
3164 switch ((unsigned long)filp->f_pos) {
3167 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3172 ino = parent_ino(dentry);
3173 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3179 /* f_version caches the tgid value that the last readdir call couldn't
3180 * return. lseek aka telldir automagically resets f_version to 0.
3182 ns = filp->f_dentry->d_sb->s_fs_info;
3183 tid = (int)filp->f_version;
3184 filp->f_version = 0;
3185 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3187 task = next_tid(task), filp->f_pos++) {
3188 tid = task_pid_nr_ns(task, ns);
3189 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3190 /* returning this tgid failed, save it as the first
3191 * pid for the next readir call */
3192 filp->f_version = (u64)tid;
3193 put_task_struct(task);
3198 put_task_struct(leader);
3203 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3205 struct inode *inode = dentry->d_inode;
3206 struct task_struct *p = get_proc_task(inode);
3207 generic_fillattr(inode, stat);
3210 stat->nlink += get_nr_threads(p);
3217 static const struct inode_operations proc_task_inode_operations = {
3218 .lookup = proc_task_lookup,
3219 .getattr = proc_task_getattr,
3220 .setattr = proc_setattr,
3223 static const struct file_operations proc_task_operations = {
3224 .read = generic_read_dir,
3225 .readdir = proc_task_readdir,