blob: 3f3d7aeb071257a15afaa9a65f88642e307f80f3 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
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.
Mauricio Line070ad42005-09-03 15:55:10 -070014 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
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>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
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.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
Linus Torvalds1da177e2005-04-16 15:20:36 -070048 */
49
50#include <asm/uaccess.h>
51
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
Andrea Righi59954772008-07-27 17:29:15 +020056#include <linux/task_io_accounting_ops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/init.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080058#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040060#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080064#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/mm.h>
David Rientjesa63d83f2010-08-09 17:19:46 -070066#include <linux/swap.h>
Dipankar Sarmab8359962005-09-09 13:04:14 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/kallsyms.h>
Ken Chen2ec220e2008-11-10 11:26:08 +030069#include <linux/stacktrace.h>
Neil Hormand85f50d2007-10-18 23:40:37 -070070#include <linux/resource.h>
Kees Cook5096add2007-05-08 00:26:04 -070071#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
Roland McGrath0d094ef2008-07-25 19:45:49 -070075#include <linux/tracehook.h>
Andrew Morton87ebdc02013-02-27 17:03:16 -080076#include <linux/printk.h>
Paul Menagea4243162007-10-18 23:39:35 -070077#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070078#include <linux/cpuset.h>
79#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050080#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070081#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070082#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070083#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070084#include <linux/pid_namespace.h>
Eric W. Biederman22d917d2011-11-17 00:11:58 -080085#include <linux/user_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040086#include <linux/fs_struct.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090087#include <linux/slab.h>
Pavel Emelyanov640708a2012-01-10 15:11:23 -080088#include <linux/flex_array.h>
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +040089#include <linux/posix-timers.h>
Chris Metcalff133ecc2011-05-26 12:40:09 -040090#ifdef CONFIG_HARDWALL
91#include <asm/hardwall.h>
92#endif
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080093#include <trace/events/oom.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include "internal.h"
Cyrill Gorcunovfaf60af2012-08-23 14:43:24 +040095#include "fd.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070097/* NOTE:
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
102 *
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
105 */
106
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107struct pid_entry {
Alexey Dobriyancedbcca2014-08-08 14:21:33 -0700108 const char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -0700109 int len;
Al Virod161a132011-07-24 03:36:29 -0400110 umode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800111 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800112 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700113 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114};
115
Eric W. Biederman61a28782006-10-02 02:18:49 -0700116#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700117 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700118 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700119 .mode = MODE, \
120 .iop = IOP, \
121 .fop = FOP, \
122 .op = OP, \
123}
124
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300125#define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700129 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300130 { .proc_get_link = get_link } )
131#define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300133#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300136 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Vegard Nossumaed54172008-06-05 22:46:53 -0700138/*
139 * Count the number of hardlinks for the pid_entry table, excluding the .
140 * and .. links.
141 */
142static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 unsigned int n)
144{
145 unsigned int i;
146 unsigned int count;
147
148 count = 0;
149 for (i = 0; i < n; ++i) {
150 if (S_ISDIR(entries[i].mode))
151 ++count;
152 }
153
154 return count;
155}
156
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200157static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000159 int result = -ENOENT;
160
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700161 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200162 if (task->fs) {
163 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000164 result = 0;
165 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700166 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000167 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700168}
169
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800170static int proc_cwd_link(struct dentry *dentry, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700171{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800172 struct task_struct *task = get_proc_task(dentry->d_inode);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700173 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700174
175 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200176 task_lock(task);
177 if (task->fs) {
178 get_fs_pwd(task->fs, path);
179 result = 0;
180 }
181 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700182 put_task_struct(task);
183 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 return result;
185}
186
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800187static int proc_root_link(struct dentry *dentry, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800189 struct task_struct *task = get_proc_task(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700191
192 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200193 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700194 put_task_struct(task);
195 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 return result;
197}
198
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -0700199static int proc_pid_cmdline(struct seq_file *m, struct pid_namespace *ns,
200 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -0700202 /*
203 * Rely on struct seq_operations::show() being called once
204 * per internal buffer allocation. See single_open(), traverse().
205 */
206 BUG_ON(m->size < PAGE_SIZE);
207 m->count += get_cmdline(task, m->buf, PAGE_SIZE);
208 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209}
210
Alexey Dobriyanf9ea5362014-08-08 14:21:35 -0700211static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
212 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
Cong Wange7dcd992012-05-31 16:26:17 -0700214 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
Al Viro2fadaef2011-02-15 22:52:11 -0500215 if (mm && !IS_ERR(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300217 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300219 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Alexey Dobriyanf9ea5362014-08-08 14:21:35 -0700220 seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 mmput(mm);
Alexey Dobriyanf9ea5362014-08-08 14:21:35 -0700222 return 0;
223 } else
224 return PTR_ERR(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225}
226
227
228#ifdef CONFIG_KALLSYMS
229/*
230 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
231 * Returns the resolved symbol. If that fails, simply return the address.
232 */
Alexey Dobriyanedfcd602014-08-08 14:21:44 -0700233static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
234 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700236 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700237 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239 wchan = get_wchan(task);
240
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700241 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600242 if (!ptrace_may_access(task, PTRACE_MODE_READ))
243 return 0;
244 else
Alexey Dobriyanedfcd602014-08-08 14:21:44 -0700245 return seq_printf(m, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700246 else
Alexey Dobriyanedfcd602014-08-08 14:21:44 -0700247 return seq_printf(m, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248}
249#endif /* CONFIG_KALLSYMS */
250
Al Viroa9712bc2011-03-23 15:52:50 -0400251static int lock_trace(struct task_struct *task)
252{
253 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
254 if (err)
255 return err;
256 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
257 mutex_unlock(&task->signal->cred_guard_mutex);
258 return -EPERM;
259 }
260 return 0;
261}
262
263static void unlock_trace(struct task_struct *task)
264{
265 mutex_unlock(&task->signal->cred_guard_mutex);
266}
267
Ken Chen2ec220e2008-11-10 11:26:08 +0300268#ifdef CONFIG_STACKTRACE
269
270#define MAX_STACK_TRACE_DEPTH 64
271
272static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
273 struct pid *pid, struct task_struct *task)
274{
275 struct stack_trace trace;
276 unsigned long *entries;
Al Viroa9712bc2011-03-23 15:52:50 -0400277 int err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300278 int i;
279
280 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
281 if (!entries)
282 return -ENOMEM;
283
284 trace.nr_entries = 0;
285 trace.max_entries = MAX_STACK_TRACE_DEPTH;
286 trace.entries = entries;
287 trace.skip = 0;
Ken Chen2ec220e2008-11-10 11:26:08 +0300288
Al Viroa9712bc2011-03-23 15:52:50 -0400289 err = lock_trace(task);
290 if (!err) {
291 save_stack_trace_tsk(task, &trace);
292
293 for (i = 0; i < trace.nr_entries; i++) {
Linus Torvaldsb81a6182011-03-23 20:51:42 -0700294 seq_printf(m, "[<%pK>] %pS\n",
Al Viroa9712bc2011-03-23 15:52:50 -0400295 (void *)entries[i], (void *)entries[i]);
296 }
297 unlock_trace(task);
Ken Chen2ec220e2008-11-10 11:26:08 +0300298 }
299 kfree(entries);
300
Al Viroa9712bc2011-03-23 15:52:50 -0400301 return err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300302}
303#endif
304
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305#ifdef CONFIG_SCHEDSTATS
306/*
307 * Provides /proc/PID/schedstat
308 */
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -0700309static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
310 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311{
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -0700312 return seq_printf(m, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100313 (unsigned long long)task->se.sum_exec_runtime,
314 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200315 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316}
317#endif
318
Arjan van de Ven97455122008-01-25 21:08:34 +0100319#ifdef CONFIG_LATENCYTOP
320static int lstats_show_proc(struct seq_file *m, void *v)
321{
322 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800323 struct inode *inode = m->private;
324 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100325
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800326 if (!task)
327 return -ESRCH;
328 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100329 for (i = 0; i < 32; i++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800330 struct latency_record *lr = &task->latency_record[i];
331 if (lr->backtrace[0]) {
Arjan van de Ven97455122008-01-25 21:08:34 +0100332 int q;
Joe Perches34e49d42011-01-12 17:00:30 -0800333 seq_printf(m, "%i %li %li",
334 lr->count, lr->time, lr->max);
Arjan van de Ven97455122008-01-25 21:08:34 +0100335 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800336 unsigned long bt = lr->backtrace[q];
337 if (!bt)
Arjan van de Ven97455122008-01-25 21:08:34 +0100338 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800339 if (bt == ULONG_MAX)
Arjan van de Ven97455122008-01-25 21:08:34 +0100340 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800341 seq_printf(m, " %ps", (void *)bt);
Arjan van de Ven97455122008-01-25 21:08:34 +0100342 }
Alexey Dobriyan9d6de122011-01-12 17:00:32 -0800343 seq_putc(m, '\n');
Arjan van de Ven97455122008-01-25 21:08:34 +0100344 }
345
346 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800347 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100348 return 0;
349}
350
351static int lstats_open(struct inode *inode, struct file *file)
352{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800353 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800354}
355
Arjan van de Ven97455122008-01-25 21:08:34 +0100356static ssize_t lstats_write(struct file *file, const char __user *buf,
357 size_t count, loff_t *offs)
358{
Al Viro496ad9a2013-01-23 17:07:38 -0500359 struct task_struct *task = get_proc_task(file_inode(file));
Arjan van de Ven97455122008-01-25 21:08:34 +0100360
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800361 if (!task)
362 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100363 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800364 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100365
366 return count;
367}
368
369static const struct file_operations proc_lstats_operations = {
370 .open = lstats_open,
371 .read = seq_read,
372 .write = lstats_write,
373 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800374 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100375};
376
377#endif
378
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -0700379static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
380 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381{
David Rientjesa7f638f2012-05-29 15:06:47 -0700382 unsigned long totalpages = totalram_pages + total_swap_pages;
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200383 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700385 read_lock(&tasklist_lock);
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200386 if (pid_alive(task))
David Rientjesa7f638f2012-05-29 15:06:47 -0700387 points = oom_badness(task, NULL, NULL, totalpages) *
388 1000 / totalpages;
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700389 read_unlock(&tasklist_lock);
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -0700390 return seq_printf(m, "%lu\n", points);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391}
392
Neil Hormand85f50d2007-10-18 23:40:37 -0700393struct limit_names {
Alexey Dobriyancedbcca2014-08-08 14:21:33 -0700394 const char *name;
395 const char *unit;
Neil Hormand85f50d2007-10-18 23:40:37 -0700396};
397
398static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700399 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700400 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
401 [RLIMIT_DATA] = {"Max data size", "bytes"},
402 [RLIMIT_STACK] = {"Max stack size", "bytes"},
403 [RLIMIT_CORE] = {"Max core file size", "bytes"},
404 [RLIMIT_RSS] = {"Max resident set", "bytes"},
405 [RLIMIT_NPROC] = {"Max processes", "processes"},
406 [RLIMIT_NOFILE] = {"Max open files", "files"},
407 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
408 [RLIMIT_AS] = {"Max address space", "bytes"},
409 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
410 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
411 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
412 [RLIMIT_NICE] = {"Max nice priority", NULL},
413 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800414 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700415};
416
417/* Display limits for a process */
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700418static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
419 struct pid *pid, struct task_struct *task)
Neil Hormand85f50d2007-10-18 23:40:37 -0700420{
421 unsigned int i;
Neil Hormand85f50d2007-10-18 23:40:37 -0700422 unsigned long flags;
Neil Hormand85f50d2007-10-18 23:40:37 -0700423
424 struct rlimit rlim[RLIM_NLIMITS];
425
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400426 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700427 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700428 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
429 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700430
431 /*
432 * print the file header
433 */
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700434 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
Neil Hormand85f50d2007-10-18 23:40:37 -0700435 "Limit", "Soft Limit", "Hard Limit", "Units");
436
437 for (i = 0; i < RLIM_NLIMITS; i++) {
438 if (rlim[i].rlim_cur == RLIM_INFINITY)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700439 seq_printf(m, "%-25s %-20s ",
Neil Hormand85f50d2007-10-18 23:40:37 -0700440 lnames[i].name, "unlimited");
441 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700442 seq_printf(m, "%-25s %-20lu ",
Neil Hormand85f50d2007-10-18 23:40:37 -0700443 lnames[i].name, rlim[i].rlim_cur);
444
445 if (rlim[i].rlim_max == RLIM_INFINITY)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700446 seq_printf(m, "%-20s ", "unlimited");
Neil Hormand85f50d2007-10-18 23:40:37 -0700447 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700448 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
Neil Hormand85f50d2007-10-18 23:40:37 -0700449
450 if (lnames[i].unit)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700451 seq_printf(m, "%-10s\n", lnames[i].unit);
Neil Hormand85f50d2007-10-18 23:40:37 -0700452 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700453 seq_putc(m, '\n');
Neil Hormand85f50d2007-10-18 23:40:37 -0700454 }
455
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700456 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700457}
458
Roland McGrathebcb6732008-07-25 19:46:00 -0700459#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700460static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
461 struct pid *pid, struct task_struct *task)
Roland McGrathebcb6732008-07-25 19:46:00 -0700462{
463 long nr;
464 unsigned long args[6], sp, pc;
Al Viroa9712bc2011-03-23 15:52:50 -0400465 int res = lock_trace(task);
466 if (res)
467 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700468
469 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700470 seq_puts(m, "running\n");
Al Viroa9712bc2011-03-23 15:52:50 -0400471 else if (nr < 0)
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700472 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400473 else
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700474 seq_printf(m,
Roland McGrathebcb6732008-07-25 19:46:00 -0700475 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
476 nr,
477 args[0], args[1], args[2], args[3], args[4], args[5],
478 sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400479 unlock_trace(task);
480 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700481}
482#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
483
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484/************************************************************************/
485/* Here the fs part begins */
486/************************************************************************/
487
488/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700489static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700491 struct task_struct *task;
492 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700493 /* Allow access to a task's file descriptors if it is us or we
494 * may use ptrace attach to the process and find out that
495 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700496 */
497 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700498 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400499 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700500 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700501 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700502 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503}
504
Eric W. Biederman6b4e3062010-03-07 16:41:34 -0800505int proc_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700506{
507 int error;
508 struct inode *inode = dentry->d_inode;
509
510 if (attr->ia_valid & ATTR_MODE)
511 return -EPERM;
512
513 error = inode_change_ok(inode, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200514 if (error)
515 return error;
516
Christoph Hellwig10257742010-06-04 11:30:02 +0200517 setattr_copy(inode, attr);
518 mark_inode_dirty(inode);
519 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700520}
521
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800522/*
523 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
524 * or euid/egid (for hide_pid_min=2)?
525 */
526static bool has_pid_permissions(struct pid_namespace *pid,
527 struct task_struct *task,
528 int hide_pid_min)
529{
530 if (pid->hide_pid < hide_pid_min)
531 return true;
532 if (in_group_p(pid->pid_gid))
533 return true;
534 return ptrace_may_access(task, PTRACE_MODE_READ);
535}
536
537
538static int proc_pid_permission(struct inode *inode, int mask)
539{
540 struct pid_namespace *pid = inode->i_sb->s_fs_info;
541 struct task_struct *task;
542 bool has_perms;
543
544 task = get_proc_task(inode);
Xiaotian Fenga2ef9902012-01-12 17:17:08 -0800545 if (!task)
546 return -ESRCH;
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800547 has_perms = has_pid_permissions(pid, task, 1);
548 put_task_struct(task);
549
550 if (!has_perms) {
551 if (pid->hide_pid == 2) {
552 /*
553 * Let's make getdents(), stat(), and open()
554 * consistent with each other. If a process
555 * may not stat() a file, it shouldn't be seen
556 * in procfs at all.
557 */
558 return -ENOENT;
559 }
560
561 return -EPERM;
562 }
563 return generic_permission(inode, mask);
564}
565
566
567
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800568static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700569 .setattr = proc_setattr,
570};
571
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800572static int proc_single_show(struct seq_file *m, void *v)
573{
574 struct inode *inode = m->private;
575 struct pid_namespace *ns;
576 struct pid *pid;
577 struct task_struct *task;
578 int ret;
579
580 ns = inode->i_sb->s_fs_info;
581 pid = proc_pid(inode);
582 task = get_pid_task(pid, PIDTYPE_PID);
583 if (!task)
584 return -ESRCH;
585
586 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
587
588 put_task_struct(task);
589 return ret;
590}
591
592static int proc_single_open(struct inode *inode, struct file *filp)
593{
Jovi Zhangc6a34052011-01-12 17:00:34 -0800594 return single_open(filp, proc_single_show, inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800595}
596
597static const struct file_operations proc_single_file_operations = {
598 .open = proc_single_open,
599 .read = seq_read,
600 .llseek = seq_lseek,
601 .release = single_release,
602};
603
Oleg Nesterov5381e162014-10-09 15:25:24 -0700604
605struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
606{
607 struct task_struct *task = get_proc_task(inode);
608 struct mm_struct *mm = ERR_PTR(-ESRCH);
609
610 if (task) {
611 mm = mm_access(task, mode);
612 put_task_struct(task);
613
614 if (!IS_ERR_OR_NULL(mm)) {
615 /* ensure this mm_struct can't be freed */
616 atomic_inc(&mm->mm_count);
617 /* but do not pin its memory */
618 mmput(mm);
619 }
620 }
621
622 return mm;
623}
624
Cong Wangb409e572012-05-31 16:26:17 -0700625static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626{
Oleg Nesterov5381e162014-10-09 15:25:24 -0700627 struct mm_struct *mm = proc_mem_open(inode, mode);
Linus Torvaldse2683372012-01-17 15:21:19 -0800628
629 if (IS_ERR(mm))
630 return PTR_ERR(mm);
631
Linus Torvaldse2683372012-01-17 15:21:19 -0800632 file->private_data = mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 return 0;
634}
635
Cong Wangb409e572012-05-31 16:26:17 -0700636static int mem_open(struct inode *inode, struct file *file)
637{
Djalal Harounibc452b42012-07-30 14:42:28 -0700638 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
639
640 /* OK to pass negative loff_t, we can catch out-of-range */
641 file->f_mode |= FMODE_UNSIGNED_OFFSET;
642
643 return ret;
Cong Wangb409e572012-05-31 16:26:17 -0700644}
645
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100646static ssize_t mem_rw(struct file *file, char __user *buf,
647 size_t count, loff_t *ppos, int write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648{
Linus Torvaldse2683372012-01-17 15:21:19 -0800649 struct mm_struct *mm = file->private_data;
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100650 unsigned long addr = *ppos;
651 ssize_t copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 char *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Linus Torvaldse2683372012-01-17 15:21:19 -0800654 if (!mm)
655 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Mel Gormane12ba742007-10-16 01:25:52 -0700657 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 if (!page)
Linus Torvaldse2683372012-01-17 15:21:19 -0800659 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700661 copied = 0;
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100662 if (!atomic_inc_not_zero(&mm->mm_users))
663 goto free;
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 while (count > 0) {
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100666 int this_len = min_t(int, count, PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100668 if (write && copy_from_user(page, buf, this_len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 copied = -EFAULT;
670 break;
671 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100672
673 this_len = access_remote_vm(mm, addr, page, this_len, write);
674 if (!this_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 if (!copied)
676 copied = -EIO;
677 break;
678 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100679
680 if (!write && copy_to_user(buf, page, this_len)) {
681 copied = -EFAULT;
682 break;
683 }
684
685 buf += this_len;
686 addr += this_len;
687 copied += this_len;
688 count -= this_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100690 *ppos = addr;
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700691
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100692 mmput(mm);
693free:
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700694 free_page((unsigned long) page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 return copied;
696}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100698static ssize_t mem_read(struct file *file, char __user *buf,
699 size_t count, loff_t *ppos)
700{
701 return mem_rw(file, buf, count, ppos, 0);
702}
703
704static ssize_t mem_write(struct file *file, const char __user *buf,
705 size_t count, loff_t *ppos)
706{
707 return mem_rw(file, (char __user*)buf, count, ppos, 1);
708}
709
Matt Mackall85863e42008-02-04 22:29:04 -0800710loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711{
712 switch (orig) {
713 case 0:
714 file->f_pos = offset;
715 break;
716 case 1:
717 file->f_pos += offset;
718 break;
719 default:
720 return -EINVAL;
721 }
722 force_successful_syscall_return();
723 return file->f_pos;
724}
725
Linus Torvaldse2683372012-01-17 15:21:19 -0800726static int mem_release(struct inode *inode, struct file *file)
727{
728 struct mm_struct *mm = file->private_data;
Oleg Nesterov71879d32012-01-31 17:14:38 +0100729 if (mm)
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100730 mmdrop(mm);
Linus Torvaldse2683372012-01-17 15:21:19 -0800731 return 0;
732}
733
Arjan van de Ven00977a52007-02-12 00:55:34 -0800734static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 .llseek = mem_lseek,
736 .read = mem_read,
737 .write = mem_write,
738 .open = mem_open,
Linus Torvaldse2683372012-01-17 15:21:19 -0800739 .release = mem_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740};
741
Cong Wangb409e572012-05-31 16:26:17 -0700742static int environ_open(struct inode *inode, struct file *file)
743{
744 return __mem_open(inode, file, PTRACE_MODE_READ);
745}
746
James Pearson315e28c2007-10-16 23:30:17 -0700747static ssize_t environ_read(struct file *file, char __user *buf,
748 size_t count, loff_t *ppos)
749{
James Pearson315e28c2007-10-16 23:30:17 -0700750 char *page;
751 unsigned long src = *ppos;
Cong Wangb409e572012-05-31 16:26:17 -0700752 int ret = 0;
753 struct mm_struct *mm = file->private_data;
James Pearson315e28c2007-10-16 23:30:17 -0700754
Cong Wangb409e572012-05-31 16:26:17 -0700755 if (!mm)
756 return 0;
James Pearson315e28c2007-10-16 23:30:17 -0700757
James Pearson315e28c2007-10-16 23:30:17 -0700758 page = (char *)__get_free_page(GFP_TEMPORARY);
759 if (!page)
Cong Wangb409e572012-05-31 16:26:17 -0700760 return -ENOMEM;
James Pearson315e28c2007-10-16 23:30:17 -0700761
Al Virod6f64b82011-02-15 22:26:01 -0500762 ret = 0;
Cong Wangb409e572012-05-31 16:26:17 -0700763 if (!atomic_inc_not_zero(&mm->mm_users))
764 goto free;
James Pearson315e28c2007-10-16 23:30:17 -0700765 while (count > 0) {
Djalal Harounie8905ec2012-07-30 14:42:26 -0700766 size_t this_len, max_len;
767 int retval;
768
769 if (src >= (mm->env_end - mm->env_start))
770 break;
James Pearson315e28c2007-10-16 23:30:17 -0700771
772 this_len = mm->env_end - (mm->env_start + src);
773
Djalal Harounie8905ec2012-07-30 14:42:26 -0700774 max_len = min_t(size_t, PAGE_SIZE, count);
775 this_len = min(max_len, this_len);
James Pearson315e28c2007-10-16 23:30:17 -0700776
Cong Wangb409e572012-05-31 16:26:17 -0700777 retval = access_remote_vm(mm, (mm->env_start + src),
James Pearson315e28c2007-10-16 23:30:17 -0700778 page, this_len, 0);
779
780 if (retval <= 0) {
781 ret = retval;
782 break;
783 }
784
785 if (copy_to_user(buf, page, retval)) {
786 ret = -EFAULT;
787 break;
788 }
789
790 ret += retval;
791 src += retval;
792 buf += retval;
793 count -= retval;
794 }
795 *ppos = src;
James Pearson315e28c2007-10-16 23:30:17 -0700796 mmput(mm);
Cong Wangb409e572012-05-31 16:26:17 -0700797
798free:
James Pearson315e28c2007-10-16 23:30:17 -0700799 free_page((unsigned long) page);
James Pearson315e28c2007-10-16 23:30:17 -0700800 return ret;
801}
802
803static const struct file_operations proc_environ_operations = {
Cong Wangb409e572012-05-31 16:26:17 -0700804 .open = environ_open,
James Pearson315e28c2007-10-16 23:30:17 -0700805 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100806 .llseek = generic_file_llseek,
Cong Wangb409e572012-05-31 16:26:17 -0700807 .release = mem_release,
James Pearson315e28c2007-10-16 23:30:17 -0700808};
809
David Rientjesfa0cbbf2012-11-12 17:53:04 -0800810static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
811 loff_t *ppos)
812{
Al Viro496ad9a2013-01-23 17:07:38 -0500813 struct task_struct *task = get_proc_task(file_inode(file));
David Rientjesfa0cbbf2012-11-12 17:53:04 -0800814 char buffer[PROC_NUMBUF];
815 int oom_adj = OOM_ADJUST_MIN;
816 size_t len;
817 unsigned long flags;
818
819 if (!task)
820 return -ESRCH;
821 if (lock_task_sighand(task, &flags)) {
822 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
823 oom_adj = OOM_ADJUST_MAX;
824 else
825 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
826 OOM_SCORE_ADJ_MAX;
827 unlock_task_sighand(task, &flags);
828 }
829 put_task_struct(task);
830 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
831 return simple_read_from_buffer(buf, count, ppos, buffer, len);
832}
833
834static ssize_t oom_adj_write(struct file *file, const char __user *buf,
835 size_t count, loff_t *ppos)
836{
837 struct task_struct *task;
838 char buffer[PROC_NUMBUF];
839 int oom_adj;
840 unsigned long flags;
841 int err;
842
843 memset(buffer, 0, sizeof(buffer));
844 if (count > sizeof(buffer) - 1)
845 count = sizeof(buffer) - 1;
846 if (copy_from_user(buffer, buf, count)) {
847 err = -EFAULT;
848 goto out;
849 }
850
851 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
852 if (err)
853 goto out;
854 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
855 oom_adj != OOM_DISABLE) {
856 err = -EINVAL;
857 goto out;
858 }
859
Al Viro496ad9a2013-01-23 17:07:38 -0500860 task = get_proc_task(file_inode(file));
David Rientjesfa0cbbf2012-11-12 17:53:04 -0800861 if (!task) {
862 err = -ESRCH;
863 goto out;
864 }
865
866 task_lock(task);
867 if (!task->mm) {
868 err = -EINVAL;
869 goto err_task_lock;
870 }
871
872 if (!lock_task_sighand(task, &flags)) {
873 err = -ESRCH;
874 goto err_task_lock;
875 }
876
877 /*
878 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
879 * value is always attainable.
880 */
881 if (oom_adj == OOM_ADJUST_MAX)
882 oom_adj = OOM_SCORE_ADJ_MAX;
883 else
884 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
885
886 if (oom_adj < task->signal->oom_score_adj &&
887 !capable(CAP_SYS_RESOURCE)) {
888 err = -EACCES;
889 goto err_sighand;
890 }
891
892 /*
893 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
894 * /proc/pid/oom_score_adj instead.
895 */
Andrew Morton87ebdc02013-02-27 17:03:16 -0800896 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
David Rientjesfa0cbbf2012-11-12 17:53:04 -0800897 current->comm, task_pid_nr(current), task_pid_nr(task),
898 task_pid_nr(task));
899
900 task->signal->oom_score_adj = oom_adj;
901 trace_oom_score_adj_update(task);
902err_sighand:
903 unlock_task_sighand(task, &flags);
904err_task_lock:
905 task_unlock(task);
906 put_task_struct(task);
907out:
908 return err < 0 ? err : count;
909}
910
911static const struct file_operations proc_oom_adj_operations = {
912 .read = oom_adj_read,
913 .write = oom_adj_write,
914 .llseek = generic_file_llseek,
915};
916
David Rientjesa63d83f2010-08-09 17:19:46 -0700917static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
918 size_t count, loff_t *ppos)
919{
Al Viro496ad9a2013-01-23 17:07:38 -0500920 struct task_struct *task = get_proc_task(file_inode(file));
David Rientjesa63d83f2010-08-09 17:19:46 -0700921 char buffer[PROC_NUMBUF];
David Rientjesa9c58b902012-12-11 16:02:54 -0800922 short oom_score_adj = OOM_SCORE_ADJ_MIN;
David Rientjesa63d83f2010-08-09 17:19:46 -0700923 unsigned long flags;
924 size_t len;
925
926 if (!task)
927 return -ESRCH;
928 if (lock_task_sighand(task, &flags)) {
929 oom_score_adj = task->signal->oom_score_adj;
930 unlock_task_sighand(task, &flags);
931 }
932 put_task_struct(task);
David Rientjesa9c58b902012-12-11 16:02:54 -0800933 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -0700934 return simple_read_from_buffer(buf, count, ppos, buffer, len);
935}
936
937static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
938 size_t count, loff_t *ppos)
939{
940 struct task_struct *task;
941 char buffer[PROC_NUMBUF];
942 unsigned long flags;
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -0700943 int oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -0700944 int err;
945
946 memset(buffer, 0, sizeof(buffer));
947 if (count > sizeof(buffer) - 1)
948 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -0700949 if (copy_from_user(buffer, buf, count)) {
950 err = -EFAULT;
951 goto out;
952 }
David Rientjesa63d83f2010-08-09 17:19:46 -0700953
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -0700954 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -0700955 if (err)
David Rientjes723548b2010-10-26 14:21:25 -0700956 goto out;
David Rientjesa63d83f2010-08-09 17:19:46 -0700957 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
David Rientjes723548b2010-10-26 14:21:25 -0700958 oom_score_adj > OOM_SCORE_ADJ_MAX) {
959 err = -EINVAL;
960 goto out;
961 }
David Rientjesa63d83f2010-08-09 17:19:46 -0700962
Al Viro496ad9a2013-01-23 17:07:38 -0500963 task = get_proc_task(file_inode(file));
David Rientjes723548b2010-10-26 14:21:25 -0700964 if (!task) {
965 err = -ESRCH;
966 goto out;
967 }
David Rientjesa63d83f2010-08-09 17:19:46 -0700968
Ying Han3d5992d2010-10-26 14:21:23 -0700969 task_lock(task);
970 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -0700971 err = -EINVAL;
972 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -0700973 }
David Rientjesd19d5472010-10-26 14:21:26 -0700974
975 if (!lock_task_sighand(task, &flags)) {
976 err = -ESRCH;
977 goto err_task_lock;
978 }
979
David Rientjesa9c58b902012-12-11 16:02:54 -0800980 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
David Rientjesd19d5472010-10-26 14:21:26 -0700981 !capable(CAP_SYS_RESOURCE)) {
982 err = -EACCES;
983 goto err_sighand;
984 }
985
David Rientjesa9c58b902012-12-11 16:02:54 -0800986 task->signal->oom_score_adj = (short)oom_score_adj;
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -0800987 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
David Rientjesa9c58b902012-12-11 16:02:54 -0800988 task->signal->oom_score_adj_min = (short)oom_score_adj;
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -0800989 trace_oom_score_adj_update(task);
Davidlohr Bueso01dc52e2012-10-08 16:29:30 -0700990
David Rientjes723548b2010-10-26 14:21:25 -0700991err_sighand:
David Rientjesa63d83f2010-08-09 17:19:46 -0700992 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -0700993err_task_lock:
994 task_unlock(task);
David Rientjesa63d83f2010-08-09 17:19:46 -0700995 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -0700996out:
997 return err < 0 ? err : count;
David Rientjesa63d83f2010-08-09 17:19:46 -0700998}
999
1000static const struct file_operations proc_oom_score_adj_operations = {
1001 .read = oom_score_adj_read,
1002 .write = oom_score_adj_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001003 .llseek = default_llseek,
David Rientjesa63d83f2010-08-09 17:19:46 -07001004};
1005
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006#ifdef CONFIG_AUDITSYSCALL
1007#define TMPBUFLEN 21
1008static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1009 size_t count, loff_t *ppos)
1010{
Al Viro496ad9a2013-01-23 17:07:38 -05001011 struct inode * inode = file_inode(file);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001012 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 ssize_t length;
1014 char tmpbuf[TMPBUFLEN];
1015
Eric W. Biederman99f89552006-06-26 00:25:55 -07001016 if (!task)
1017 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001019 from_kuid(file->f_cred->user_ns,
1020 audit_get_loginuid(task)));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001021 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1023}
1024
1025static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1026 size_t count, loff_t *ppos)
1027{
Al Viro496ad9a2013-01-23 17:07:38 -05001028 struct inode * inode = file_inode(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 char *page, *tmp;
1030 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 uid_t loginuid;
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001032 kuid_t kloginuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001034 rcu_read_lock();
1035 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1036 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 return -EPERM;
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001038 }
1039 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040
Al Viroe0182902006-05-18 08:28:02 -04001041 if (count >= PAGE_SIZE)
1042 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
1044 if (*ppos != 0) {
1045 /* No partial writes. */
1046 return -EINVAL;
1047 }
Mel Gormane12ba742007-10-16 01:25:52 -07001048 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 if (!page)
1050 return -ENOMEM;
1051 length = -EFAULT;
1052 if (copy_from_user(page, buf, count))
1053 goto out_free_page;
1054
Al Viroe0182902006-05-18 08:28:02 -04001055 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 loginuid = simple_strtoul(page, &tmp, 10);
1057 if (tmp == page) {
1058 length = -EINVAL;
1059 goto out_free_page;
1060
1061 }
Eric Paris81407c82013-05-24 09:49:14 -04001062
1063 /* is userspace tring to explicitly UNSET the loginuid? */
1064 if (loginuid == AUDIT_UID_UNSET) {
1065 kloginuid = INVALID_UID;
1066 } else {
1067 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1068 if (!uid_valid(kloginuid)) {
1069 length = -EINVAL;
1070 goto out_free_page;
1071 }
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001072 }
1073
1074 length = audit_set_loginuid(kloginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 if (likely(length == 0))
1076 length = count;
1077
1078out_free_page:
1079 free_page((unsigned long) page);
1080 return length;
1081}
1082
Arjan van de Ven00977a52007-02-12 00:55:34 -08001083static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 .read = proc_loginuid_read,
1085 .write = proc_loginuid_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001086 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087};
Eric Paris1e0bd752008-03-13 08:15:31 -04001088
1089static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1090 size_t count, loff_t *ppos)
1091{
Al Viro496ad9a2013-01-23 17:07:38 -05001092 struct inode * inode = file_inode(file);
Eric Paris1e0bd752008-03-13 08:15:31 -04001093 struct task_struct *task = get_proc_task(inode);
1094 ssize_t length;
1095 char tmpbuf[TMPBUFLEN];
1096
1097 if (!task)
1098 return -ESRCH;
1099 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1100 audit_get_sessionid(task));
1101 put_task_struct(task);
1102 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1103}
1104
1105static const struct file_operations proc_sessionid_operations = {
1106 .read = proc_sessionid_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001107 .llseek = generic_file_llseek,
Eric Paris1e0bd752008-03-13 08:15:31 -04001108};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109#endif
1110
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001111#ifdef CONFIG_FAULT_INJECTION
1112static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1113 size_t count, loff_t *ppos)
1114{
Al Viro496ad9a2013-01-23 17:07:38 -05001115 struct task_struct *task = get_proc_task(file_inode(file));
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001116 char buffer[PROC_NUMBUF];
1117 size_t len;
1118 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001119
1120 if (!task)
1121 return -ESRCH;
1122 make_it_fail = task->make_it_fail;
1123 put_task_struct(task);
1124
1125 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001126
1127 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001128}
1129
1130static ssize_t proc_fault_inject_write(struct file * file,
1131 const char __user * buf, size_t count, loff_t *ppos)
1132{
1133 struct task_struct *task;
1134 char buffer[PROC_NUMBUF], *end;
1135 int make_it_fail;
1136
1137 if (!capable(CAP_SYS_RESOURCE))
1138 return -EPERM;
1139 memset(buffer, 0, sizeof(buffer));
1140 if (count > sizeof(buffer) - 1)
1141 count = sizeof(buffer) - 1;
1142 if (copy_from_user(buffer, buf, count))
1143 return -EFAULT;
Vincent Licba8aaf2009-09-22 16:45:38 -07001144 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1145 if (*end)
1146 return -EINVAL;
Dave Jones16caed32014-04-07 15:39:15 -07001147 if (make_it_fail < 0 || make_it_fail > 1)
1148 return -EINVAL;
1149
Al Viro496ad9a2013-01-23 17:07:38 -05001150 task = get_proc_task(file_inode(file));
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001151 if (!task)
1152 return -ESRCH;
1153 task->make_it_fail = make_it_fail;
1154 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001155
1156 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001157}
1158
Arjan van de Ven00977a52007-02-12 00:55:34 -08001159static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001160 .read = proc_fault_inject_read,
1161 .write = proc_fault_inject_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001162 .llseek = generic_file_llseek,
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001163};
1164#endif
1165
Arjan van de Ven97455122008-01-25 21:08:34 +01001166
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001167#ifdef CONFIG_SCHED_DEBUG
1168/*
1169 * Print out various scheduling related per-task fields:
1170 */
1171static int sched_show(struct seq_file *m, void *v)
1172{
1173 struct inode *inode = m->private;
1174 struct task_struct *p;
1175
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001176 p = get_proc_task(inode);
1177 if (!p)
1178 return -ESRCH;
1179 proc_sched_show_task(p, m);
1180
1181 put_task_struct(p);
1182
1183 return 0;
1184}
1185
1186static ssize_t
1187sched_write(struct file *file, const char __user *buf,
1188 size_t count, loff_t *offset)
1189{
Al Viro496ad9a2013-01-23 17:07:38 -05001190 struct inode *inode = file_inode(file);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001191 struct task_struct *p;
1192
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001193 p = get_proc_task(inode);
1194 if (!p)
1195 return -ESRCH;
1196 proc_sched_set_task(p);
1197
1198 put_task_struct(p);
1199
1200 return count;
1201}
1202
1203static int sched_open(struct inode *inode, struct file *filp)
1204{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001205 return single_open(filp, sched_show, inode);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001206}
1207
1208static const struct file_operations proc_pid_sched_operations = {
1209 .open = sched_open,
1210 .read = seq_read,
1211 .write = sched_write,
1212 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001213 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001214};
1215
1216#endif
1217
Mike Galbraith5091faa2010-11-30 14:18:03 +01001218#ifdef CONFIG_SCHED_AUTOGROUP
1219/*
1220 * Print out autogroup related information:
1221 */
1222static int sched_autogroup_show(struct seq_file *m, void *v)
1223{
1224 struct inode *inode = m->private;
1225 struct task_struct *p;
1226
1227 p = get_proc_task(inode);
1228 if (!p)
1229 return -ESRCH;
1230 proc_sched_autogroup_show_task(p, m);
1231
1232 put_task_struct(p);
1233
1234 return 0;
1235}
1236
1237static ssize_t
1238sched_autogroup_write(struct file *file, const char __user *buf,
1239 size_t count, loff_t *offset)
1240{
Al Viro496ad9a2013-01-23 17:07:38 -05001241 struct inode *inode = file_inode(file);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001242 struct task_struct *p;
1243 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001244 int nice;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001245 int err;
1246
1247 memset(buffer, 0, sizeof(buffer));
1248 if (count > sizeof(buffer) - 1)
1249 count = sizeof(buffer) - 1;
1250 if (copy_from_user(buffer, buf, count))
1251 return -EFAULT;
1252
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001253 err = kstrtoint(strstrip(buffer), 0, &nice);
1254 if (err < 0)
1255 return err;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001256
1257 p = get_proc_task(inode);
1258 if (!p)
1259 return -ESRCH;
1260
Hiroshi Shimamoto2e5b5b32012-02-23 17:41:27 +09001261 err = proc_sched_autogroup_set_nice(p, nice);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001262 if (err)
1263 count = err;
1264
1265 put_task_struct(p);
1266
1267 return count;
1268}
1269
1270static int sched_autogroup_open(struct inode *inode, struct file *filp)
1271{
1272 int ret;
1273
1274 ret = single_open(filp, sched_autogroup_show, NULL);
1275 if (!ret) {
1276 struct seq_file *m = filp->private_data;
1277
1278 m->private = inode;
1279 }
1280 return ret;
1281}
1282
1283static const struct file_operations proc_pid_sched_autogroup_operations = {
1284 .open = sched_autogroup_open,
1285 .read = seq_read,
1286 .write = sched_autogroup_write,
1287 .llseek = seq_lseek,
1288 .release = single_release,
1289};
1290
1291#endif /* CONFIG_SCHED_AUTOGROUP */
1292
john stultz4614a696b2009-12-14 18:00:05 -08001293static ssize_t comm_write(struct file *file, const char __user *buf,
1294 size_t count, loff_t *offset)
1295{
Al Viro496ad9a2013-01-23 17:07:38 -05001296 struct inode *inode = file_inode(file);
john stultz4614a696b2009-12-14 18:00:05 -08001297 struct task_struct *p;
1298 char buffer[TASK_COMM_LEN];
David Rientjes830e0fc2013-04-30 15:28:18 -07001299 const size_t maxlen = sizeof(buffer) - 1;
john stultz4614a696b2009-12-14 18:00:05 -08001300
1301 memset(buffer, 0, sizeof(buffer));
David Rientjes830e0fc2013-04-30 15:28:18 -07001302 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
john stultz4614a696b2009-12-14 18:00:05 -08001303 return -EFAULT;
1304
1305 p = get_proc_task(inode);
1306 if (!p)
1307 return -ESRCH;
1308
1309 if (same_thread_group(current, p))
1310 set_task_comm(p, buffer);
1311 else
1312 count = -EINVAL;
1313
1314 put_task_struct(p);
1315
1316 return count;
1317}
1318
1319static int comm_show(struct seq_file *m, void *v)
1320{
1321 struct inode *inode = m->private;
1322 struct task_struct *p;
1323
1324 p = get_proc_task(inode);
1325 if (!p)
1326 return -ESRCH;
1327
1328 task_lock(p);
1329 seq_printf(m, "%s\n", p->comm);
1330 task_unlock(p);
1331
1332 put_task_struct(p);
1333
1334 return 0;
1335}
1336
1337static int comm_open(struct inode *inode, struct file *filp)
1338{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001339 return single_open(filp, comm_show, inode);
john stultz4614a696b2009-12-14 18:00:05 -08001340}
1341
1342static const struct file_operations proc_pid_set_comm_operations = {
1343 .open = comm_open,
1344 .read = seq_read,
1345 .write = comm_write,
1346 .llseek = seq_lseek,
1347 .release = single_release,
1348};
1349
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001350static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
Matt Helsley925d1c42008-04-29 01:01:36 -07001351{
1352 struct task_struct *task;
1353 struct mm_struct *mm;
1354 struct file *exe_file;
1355
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001356 task = get_proc_task(dentry->d_inode);
Matt Helsley925d1c42008-04-29 01:01:36 -07001357 if (!task)
1358 return -ENOENT;
1359 mm = get_task_mm(task);
1360 put_task_struct(task);
1361 if (!mm)
1362 return -ENOENT;
1363 exe_file = get_mm_exe_file(mm);
1364 mmput(mm);
1365 if (exe_file) {
1366 *exe_path = exe_file->f_path;
1367 path_get(&exe_file->f_path);
1368 fput(exe_file);
1369 return 0;
1370 } else
1371 return -ENOENT;
1372}
1373
Al Viro008b1502005-08-20 00:17:39 +01001374static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375{
1376 struct inode *inode = dentry->d_inode;
Christoph Hellwig408ef012012-06-18 10:47:03 -04001377 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 int error = -EACCES;
1379
Eric W. Biederman778c1142006-06-26 00:25:58 -07001380 /* Are we allowed to snoop on the tasks file descriptors? */
1381 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
Christoph Hellwig408ef012012-06-18 10:47:03 -04001384 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1385 if (error)
1386 goto out;
1387
Christoph Hellwigb5fb63c12012-06-18 10:47:04 -04001388 nd_jump_link(nd, &path);
Christoph Hellwig408ef012012-06-18 10:47:03 -04001389 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390out:
Al Viro008b1502005-08-20 00:17:39 +01001391 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392}
1393
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001394static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395{
Mel Gormane12ba742007-10-16 01:25:52 -07001396 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001397 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 int len;
1399
1400 if (!tmp)
1401 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001402
Eric W. Biederman7b2a69b2010-12-05 15:51:21 -08001403 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001404 len = PTR_ERR(pathname);
1405 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001407 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
1409 if (len > buflen)
1410 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001411 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 len = -EFAULT;
1413 out:
1414 free_page((unsigned long)tmp);
1415 return len;
1416}
1417
1418static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1419{
1420 int error = -EACCES;
1421 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001422 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
Eric W. Biederman778c1142006-06-26 00:25:58 -07001424 /* Are we allowed to snoop on the tasks file descriptors? */
1425 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001428 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 if (error)
1430 goto out;
1431
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001432 error = do_proc_readlink(&path, buffer, buflen);
1433 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 return error;
1436}
1437
Cyrill Gorcunovfaf60af2012-08-23 14:43:24 +04001438const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001440 .follow_link = proc_pid_follow_link,
1441 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442};
1443
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001444
1445/* building an inode */
1446
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001447struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001448{
1449 struct inode * inode;
1450 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001451 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001452
1453 /* We need a new inode */
1454
1455 inode = new_inode(sb);
1456 if (!inode)
1457 goto out;
1458
1459 /* Common stuff */
1460 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04001461 inode->i_ino = get_next_ino();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001462 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001463 inode->i_op = &proc_def_inode_operations;
1464
1465 /*
1466 * grab the reference to task.
1467 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001468 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001469 if (!ei->pid)
1470 goto out_unlock;
1471
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001472 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001473 rcu_read_lock();
1474 cred = __task_cred(task);
1475 inode->i_uid = cred->euid;
1476 inode->i_gid = cred->egid;
1477 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001478 }
1479 security_task_to_inode(task, inode);
1480
1481out:
1482 return inode;
1483
1484out_unlock:
1485 iput(inode);
1486 return NULL;
1487}
1488
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001489int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001490{
1491 struct inode *inode = dentry->d_inode;
1492 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001493 const struct cred *cred;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08001494 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
David Howellsc69e8d92008-11-14 10:39:19 +11001495
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001496 generic_fillattr(inode, stat);
1497
1498 rcu_read_lock();
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001499 stat->uid = GLOBAL_ROOT_UID;
1500 stat->gid = GLOBAL_ROOT_GID;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001501 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1502 if (task) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -08001503 if (!has_pid_permissions(pid, task, 2)) {
1504 rcu_read_unlock();
1505 /*
1506 * This doesn't prevent learning whether PID exists,
1507 * it only makes getattr() consistent with readdir().
1508 */
1509 return -ENOENT;
1510 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001511 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1512 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001513 cred = __task_cred(task);
1514 stat->uid = cred->euid;
1515 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001516 }
1517 }
1518 rcu_read_unlock();
1519 return 0;
1520}
1521
1522/* dentry stuff */
1523
1524/*
1525 * Exceptional case: normally we are not allowed to unhash a busy
1526 * directory. In this case, however, we can do it - no aliasing problems
1527 * due to the way we treat inodes.
1528 *
1529 * Rewrite the inode's ownerships here because the owning task may have
1530 * performed a setuid(), etc.
1531 *
1532 * Before the /proc/pid/status file was created the only way to read
1533 * the effective uid of a /process was to stat /proc/pid. Reading
1534 * /proc/pid/status is slow enough that procps and other packages
1535 * kept stating /proc/pid. To keep the rules in /proc simple I have
1536 * made this apply to all per process world readable and executable
1537 * directories.
1538 */
Al Viro0b728e12012-06-10 16:03:43 -04001539int pid_revalidate(struct dentry *dentry, unsigned int flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001540{
Nick Piggin34286d62011-01-07 17:49:57 +11001541 struct inode *inode;
1542 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001543 const struct cred *cred;
1544
Al Viro0b728e12012-06-10 16:03:43 -04001545 if (flags & LOOKUP_RCU)
Nick Piggin34286d62011-01-07 17:49:57 +11001546 return -ECHILD;
1547
1548 inode = dentry->d_inode;
1549 task = get_proc_task(inode);
1550
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001551 if (task) {
1552 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1553 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001554 rcu_read_lock();
1555 cred = __task_cred(task);
1556 inode->i_uid = cred->euid;
1557 inode->i_gid = cred->egid;
1558 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001559 } else {
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001560 inode->i_uid = GLOBAL_ROOT_UID;
1561 inode->i_gid = GLOBAL_ROOT_GID;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001562 }
1563 inode->i_mode &= ~(S_ISUID | S_ISGID);
1564 security_task_to_inode(task, inode);
1565 put_task_struct(task);
1566 return 1;
1567 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001568 return 0;
1569}
1570
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08001571static inline bool proc_inode_is_dead(struct inode *inode)
1572{
1573 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1574}
1575
David Howells1dd704b2013-04-12 01:08:50 +01001576int pid_delete_dentry(const struct dentry *dentry)
1577{
1578 /* Is the task we represent dead?
1579 * If so, then don't put the dentry on the lru list,
1580 * kill it immediately.
1581 */
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08001582 return proc_inode_is_dead(dentry->d_inode);
David Howells1dd704b2013-04-12 01:08:50 +01001583}
1584
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001585const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001586{
1587 .d_revalidate = pid_revalidate,
1588 .d_delete = pid_delete_dentry,
1589};
1590
1591/* Lookups */
1592
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001593/*
1594 * Fill a directory entry.
1595 *
1596 * If possible create the dcache entry and derive our inode number and
1597 * file type from dcache entry.
1598 *
1599 * Since all of the proc inode numbers are dynamically generated, the inode
1600 * numbers do not exist until the inode is cache. This means creating the
1601 * the dcache entry in readdir is necessary to keep the inode numbers
1602 * reported by readdir in sync with the inode numbers reported
1603 * by stat.
1604 */
Al Virof0c3b502013-05-16 12:07:31 -04001605bool proc_fill_cache(struct file *file, struct dir_context *ctx,
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001606 const char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001607 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001608{
Al Virof0c3b502013-05-16 12:07:31 -04001609 struct dentry *child, *dir = file->f_path.dentry;
Al Viro1df98b82013-06-15 11:33:10 +04001610 struct qstr qname = QSTR_INIT(name, len);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001611 struct inode *inode;
Al Viro1df98b82013-06-15 11:33:10 +04001612 unsigned type;
1613 ino_t ino;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001614
Al Viro1df98b82013-06-15 11:33:10 +04001615 child = d_hash_and_lookup(dir, &qname);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001616 if (!child) {
Al Viro1df98b82013-06-15 11:33:10 +04001617 child = d_alloc(dir, &qname);
1618 if (!child)
1619 goto end_instantiate;
1620 if (instantiate(dir->d_inode, child, task, ptr) < 0) {
1621 dput(child);
1622 goto end_instantiate;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001623 }
1624 }
Eric W. Biederman61a28782006-10-02 02:18:49 -07001625 inode = child->d_inode;
Al Viro147ce692013-06-15 10:26:35 +04001626 ino = inode->i_ino;
1627 type = inode->i_mode >> 12;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001628 dput(child);
Al Virof0c3b502013-05-16 12:07:31 -04001629 return dir_emit(ctx, name, len, ino, type);
Al Viro1df98b82013-06-15 11:33:10 +04001630
1631end_instantiate:
1632 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001633}
1634
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001635#ifdef CONFIG_CHECKPOINT_RESTORE
1636
1637/*
1638 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1639 * which represent vma start and end addresses.
1640 */
1641static int dname_to_vma_addr(struct dentry *dentry,
1642 unsigned long *start, unsigned long *end)
1643{
1644 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1645 return -EINVAL;
1646
1647 return 0;
1648}
1649
Al Viro0b728e12012-06-10 16:03:43 -04001650static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001651{
1652 unsigned long vm_start, vm_end;
1653 bool exact_vma_exists = false;
1654 struct mm_struct *mm = NULL;
1655 struct task_struct *task;
1656 const struct cred *cred;
1657 struct inode *inode;
1658 int status = 0;
1659
Al Viro0b728e12012-06-10 16:03:43 -04001660 if (flags & LOOKUP_RCU)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001661 return -ECHILD;
1662
1663 if (!capable(CAP_SYS_ADMIN)) {
Zhao Hongjiang41735812013-02-20 13:13:55 +11001664 status = -EPERM;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001665 goto out_notask;
1666 }
1667
1668 inode = dentry->d_inode;
1669 task = get_proc_task(inode);
1670 if (!task)
1671 goto out_notask;
1672
Cong Wang2344bec2012-05-31 16:26:18 -07001673 mm = mm_access(task, PTRACE_MODE_READ);
1674 if (IS_ERR_OR_NULL(mm))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001675 goto out;
1676
1677 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1678 down_read(&mm->mmap_sem);
1679 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1680 up_read(&mm->mmap_sem);
1681 }
1682
1683 mmput(mm);
1684
1685 if (exact_vma_exists) {
1686 if (task_dumpable(task)) {
1687 rcu_read_lock();
1688 cred = __task_cred(task);
1689 inode->i_uid = cred->euid;
1690 inode->i_gid = cred->egid;
1691 rcu_read_unlock();
1692 } else {
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001693 inode->i_uid = GLOBAL_ROOT_UID;
1694 inode->i_gid = GLOBAL_ROOT_GID;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001695 }
1696 security_task_to_inode(task, inode);
1697 status = 1;
1698 }
1699
1700out:
1701 put_task_struct(task);
1702
1703out_notask:
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001704 return status;
1705}
1706
1707static const struct dentry_operations tid_map_files_dentry_operations = {
1708 .d_revalidate = map_files_d_revalidate,
1709 .d_delete = pid_delete_dentry,
1710};
1711
1712static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1713{
1714 unsigned long vm_start, vm_end;
1715 struct vm_area_struct *vma;
1716 struct task_struct *task;
1717 struct mm_struct *mm;
1718 int rc;
1719
1720 rc = -ENOENT;
1721 task = get_proc_task(dentry->d_inode);
1722 if (!task)
1723 goto out;
1724
1725 mm = get_task_mm(task);
1726 put_task_struct(task);
1727 if (!mm)
1728 goto out;
1729
1730 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1731 if (rc)
1732 goto out_mmput;
1733
Artem Fetishev70335ab2014-03-10 15:49:45 -07001734 rc = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001735 down_read(&mm->mmap_sem);
1736 vma = find_exact_vma(mm, vm_start, vm_end);
1737 if (vma && vma->vm_file) {
1738 *path = vma->vm_file->f_path;
1739 path_get(path);
1740 rc = 0;
1741 }
1742 up_read(&mm->mmap_sem);
1743
1744out_mmput:
1745 mmput(mm);
1746out:
1747 return rc;
1748}
1749
1750struct map_files_info {
Al Viro7b540d02012-08-27 14:55:26 -04001751 fmode_t mode;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001752 unsigned long len;
1753 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1754};
1755
Al Viroc52a47a2013-06-15 11:15:20 +04001756static int
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001757proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1758 struct task_struct *task, const void *ptr)
1759{
Al Viro7b540d02012-08-27 14:55:26 -04001760 fmode_t mode = (fmode_t)(unsigned long)ptr;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001761 struct proc_inode *ei;
1762 struct inode *inode;
1763
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001764 inode = proc_pid_make_inode(dir->i_sb, task);
1765 if (!inode)
Al Viroc52a47a2013-06-15 11:15:20 +04001766 return -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001767
1768 ei = PROC_I(inode);
1769 ei->op.proc_get_link = proc_map_files_get_link;
1770
1771 inode->i_op = &proc_pid_link_inode_operations;
1772 inode->i_size = 64;
1773 inode->i_mode = S_IFLNK;
1774
Al Viro7b540d02012-08-27 14:55:26 -04001775 if (mode & FMODE_READ)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001776 inode->i_mode |= S_IRUSR;
Al Viro7b540d02012-08-27 14:55:26 -04001777 if (mode & FMODE_WRITE)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001778 inode->i_mode |= S_IWUSR;
1779
1780 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1781 d_add(dentry, inode);
1782
Al Viroc52a47a2013-06-15 11:15:20 +04001783 return 0;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001784}
1785
1786static struct dentry *proc_map_files_lookup(struct inode *dir,
Al Viro00cd8dd2012-06-10 17:13:09 -04001787 struct dentry *dentry, unsigned int flags)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001788{
1789 unsigned long vm_start, vm_end;
1790 struct vm_area_struct *vma;
1791 struct task_struct *task;
Al Viroc52a47a2013-06-15 11:15:20 +04001792 int result;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001793 struct mm_struct *mm;
1794
Al Viroc52a47a2013-06-15 11:15:20 +04001795 result = -EPERM;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001796 if (!capable(CAP_SYS_ADMIN))
1797 goto out;
1798
Al Viroc52a47a2013-06-15 11:15:20 +04001799 result = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001800 task = get_proc_task(dir);
1801 if (!task)
1802 goto out;
1803
Al Viroc52a47a2013-06-15 11:15:20 +04001804 result = -EACCES;
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07001805 if (!ptrace_may_access(task, PTRACE_MODE_READ))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001806 goto out_put_task;
1807
Al Viroc52a47a2013-06-15 11:15:20 +04001808 result = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001809 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07001810 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001811
1812 mm = get_task_mm(task);
1813 if (!mm)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07001814 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001815
1816 down_read(&mm->mmap_sem);
1817 vma = find_exact_vma(mm, vm_start, vm_end);
1818 if (!vma)
1819 goto out_no_vma;
1820
Stanislav Kinsbursky05f56482012-11-26 16:29:42 -08001821 if (vma->vm_file)
1822 result = proc_map_files_instantiate(dir, dentry, task,
1823 (void *)(unsigned long)vma->vm_file->f_mode);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001824
1825out_no_vma:
1826 up_read(&mm->mmap_sem);
1827 mmput(mm);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001828out_put_task:
1829 put_task_struct(task);
1830out:
Al Viroc52a47a2013-06-15 11:15:20 +04001831 return ERR_PTR(result);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001832}
1833
1834static const struct inode_operations proc_map_files_inode_operations = {
1835 .lookup = proc_map_files_lookup,
1836 .permission = proc_fd_permission,
1837 .setattr = proc_setattr,
1838};
1839
1840static int
Al Virof0c3b502013-05-16 12:07:31 -04001841proc_map_files_readdir(struct file *file, struct dir_context *ctx)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001842{
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001843 struct vm_area_struct *vma;
1844 struct task_struct *task;
1845 struct mm_struct *mm;
Al Virof0c3b502013-05-16 12:07:31 -04001846 unsigned long nr_files, pos, i;
1847 struct flex_array *fa = NULL;
1848 struct map_files_info info;
1849 struct map_files_info *p;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001850 int ret;
1851
Zhao Hongjiang41735812013-02-20 13:13:55 +11001852 ret = -EPERM;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001853 if (!capable(CAP_SYS_ADMIN))
1854 goto out;
1855
1856 ret = -ENOENT;
Al Virof0c3b502013-05-16 12:07:31 -04001857 task = get_proc_task(file_inode(file));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001858 if (!task)
1859 goto out;
1860
1861 ret = -EACCES;
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07001862 if (!ptrace_may_access(task, PTRACE_MODE_READ))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001863 goto out_put_task;
1864
1865 ret = 0;
Al Virof0c3b502013-05-16 12:07:31 -04001866 if (!dir_emit_dots(file, ctx))
1867 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001868
Al Virof0c3b502013-05-16 12:07:31 -04001869 mm = get_task_mm(task);
1870 if (!mm)
1871 goto out_put_task;
1872 down_read(&mm->mmap_sem);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001873
Al Virof0c3b502013-05-16 12:07:31 -04001874 nr_files = 0;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001875
Al Virof0c3b502013-05-16 12:07:31 -04001876 /*
1877 * We need two passes here:
1878 *
1879 * 1) Collect vmas of mapped files with mmap_sem taken
1880 * 2) Release mmap_sem and instantiate entries
1881 *
1882 * otherwise we get lockdep complained, since filldir()
1883 * routine might require mmap_sem taken in might_fault().
1884 */
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001885
Al Virof0c3b502013-05-16 12:07:31 -04001886 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1887 if (vma->vm_file && ++pos > ctx->pos)
1888 nr_files++;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001889 }
Al Virof0c3b502013-05-16 12:07:31 -04001890
1891 if (nr_files) {
1892 fa = flex_array_alloc(sizeof(info), nr_files,
1893 GFP_KERNEL);
1894 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1895 GFP_KERNEL)) {
1896 ret = -ENOMEM;
1897 if (fa)
1898 flex_array_free(fa);
1899 up_read(&mm->mmap_sem);
1900 mmput(mm);
1901 goto out_put_task;
1902 }
1903 for (i = 0, vma = mm->mmap, pos = 2; vma;
1904 vma = vma->vm_next) {
1905 if (!vma->vm_file)
1906 continue;
1907 if (++pos <= ctx->pos)
1908 continue;
1909
1910 info.mode = vma->vm_file->f_mode;
1911 info.len = snprintf(info.name,
1912 sizeof(info.name), "%lx-%lx",
1913 vma->vm_start, vma->vm_end);
1914 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1915 BUG();
1916 }
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001917 }
Al Virof0c3b502013-05-16 12:07:31 -04001918 up_read(&mm->mmap_sem);
1919
1920 for (i = 0; i < nr_files; i++) {
1921 p = flex_array_get(fa, i);
1922 if (!proc_fill_cache(file, ctx,
1923 p->name, p->len,
1924 proc_map_files_instantiate,
1925 task,
1926 (void *)(unsigned long)p->mode))
1927 break;
1928 ctx->pos++;
1929 }
1930 if (fa)
1931 flex_array_free(fa);
1932 mmput(mm);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001933
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001934out_put_task:
1935 put_task_struct(task);
1936out:
1937 return ret;
1938}
1939
1940static const struct file_operations proc_map_files_operations = {
1941 .read = generic_read_dir,
Al Virof0c3b502013-05-16 12:07:31 -04001942 .iterate = proc_map_files_readdir,
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001943 .llseek = default_llseek,
1944};
1945
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04001946struct timers_private {
1947 struct pid *pid;
1948 struct task_struct *task;
1949 struct sighand_struct *sighand;
Pavel Emelyanov57b80152013-03-11 13:13:08 +04001950 struct pid_namespace *ns;
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04001951 unsigned long flags;
1952};
1953
1954static void *timers_start(struct seq_file *m, loff_t *pos)
1955{
1956 struct timers_private *tp = m->private;
1957
1958 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
1959 if (!tp->task)
1960 return ERR_PTR(-ESRCH);
1961
1962 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
1963 if (!tp->sighand)
1964 return ERR_PTR(-ESRCH);
1965
1966 return seq_list_start(&tp->task->signal->posix_timers, *pos);
1967}
1968
1969static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
1970{
1971 struct timers_private *tp = m->private;
1972 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
1973}
1974
1975static void timers_stop(struct seq_file *m, void *v)
1976{
1977 struct timers_private *tp = m->private;
1978
1979 if (tp->sighand) {
1980 unlock_task_sighand(tp->task, &tp->flags);
1981 tp->sighand = NULL;
1982 }
1983
1984 if (tp->task) {
1985 put_task_struct(tp->task);
1986 tp->task = NULL;
1987 }
1988}
1989
1990static int show_timer(struct seq_file *m, void *v)
1991{
1992 struct k_itimer *timer;
Pavel Emelyanov57b80152013-03-11 13:13:08 +04001993 struct timers_private *tp = m->private;
1994 int notify;
Alexey Dobriyancedbcca2014-08-08 14:21:33 -07001995 static const char * const nstr[] = {
Pavel Emelyanov57b80152013-03-11 13:13:08 +04001996 [SIGEV_SIGNAL] = "signal",
1997 [SIGEV_NONE] = "none",
1998 [SIGEV_THREAD] = "thread",
1999 };
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002000
2001 timer = list_entry((struct list_head *)v, struct k_itimer, list);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002002 notify = timer->it_sigev_notify;
2003
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002004 seq_printf(m, "ID: %d\n", timer->it_id);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002005 seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2006 timer->sigq->info.si_value.sival_ptr);
2007 seq_printf(m, "notify: %s/%s.%d\n",
2008 nstr[notify & ~SIGEV_THREAD_ID],
2009 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2010 pid_nr_ns(timer->it_pid, tp->ns));
Pavel Tikhomirov15ef0292013-05-17 02:12:03 +04002011 seq_printf(m, "ClockID: %d\n", timer->it_clock);
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002012
2013 return 0;
2014}
2015
2016static const struct seq_operations proc_timers_seq_ops = {
2017 .start = timers_start,
2018 .next = timers_next,
2019 .stop = timers_stop,
2020 .show = show_timer,
2021};
2022
2023static int proc_timers_open(struct inode *inode, struct file *file)
2024{
2025 struct timers_private *tp;
2026
2027 tp = __seq_open_private(file, &proc_timers_seq_ops,
2028 sizeof(struct timers_private));
2029 if (!tp)
2030 return -ENOMEM;
2031
2032 tp->pid = proc_pid(inode);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002033 tp->ns = inode->i_sb->s_fs_info;
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002034 return 0;
2035}
2036
2037static const struct file_operations proc_timers_operations = {
2038 .open = proc_timers_open,
2039 .read = seq_read,
2040 .llseek = seq_lseek,
2041 .release = seq_release_private,
2042};
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002043#endif /* CONFIG_CHECKPOINT_RESTORE */
2044
Al Viroc52a47a2013-06-15 11:15:20 +04002045static int proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002046 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002047{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002048 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002049 struct inode *inode;
2050 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002051
Eric W. Biederman61a28782006-10-02 02:18:49 -07002052 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002053 if (!inode)
2054 goto out;
2055
2056 ei = PROC_I(inode);
2057 inode->i_mode = p->mode;
2058 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002059 set_nlink(inode, 2); /* Use getattr to fix if necessary */
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002060 if (p->iop)
2061 inode->i_op = p->iop;
2062 if (p->fop)
2063 inode->i_fop = p->fop;
2064 ei->op = p->op;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002065 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002066 d_add(dentry, inode);
2067 /* Close the race of the process dying before we return the dentry */
Al Viro0b728e12012-06-10 16:03:43 -04002068 if (pid_revalidate(dentry, 0))
Al Viroc52a47a2013-06-15 11:15:20 +04002069 return 0;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002070out:
Al Viroc52a47a2013-06-15 11:15:20 +04002071 return -ENOENT;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002072}
2073
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074static struct dentry *proc_pident_lookup(struct inode *dir,
2075 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002076 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002077 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078{
Al Viroc52a47a2013-06-15 11:15:20 +04002079 int error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002080 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002081 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082
Al Viroc52a47a2013-06-15 11:15:20 +04002083 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084
Eric W. Biederman99f89552006-06-26 00:25:55 -07002085 if (!task)
2086 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002088 /*
2089 * Yes, it does not scale. And it should not. Don't add
2090 * new entries into /proc/<tgid>/ without very good reasons.
2091 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002092 last = &ents[nents - 1];
2093 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 if (p->len != dentry->d_name.len)
2095 continue;
2096 if (!memcmp(dentry->d_name.name, p->name, p->len))
2097 break;
2098 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002099 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 goto out;
2101
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002102 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002104 put_task_struct(task);
2105out_no_task:
Al Viroc52a47a2013-06-15 11:15:20 +04002106 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107}
2108
Al Virof0c3b502013-05-16 12:07:31 -04002109static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002110 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002111{
Al Virof0c3b502013-05-16 12:07:31 -04002112 struct task_struct *task = get_proc_task(file_inode(file));
2113 const struct pid_entry *p;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002114
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002115 if (!task)
Al Virof0c3b502013-05-16 12:07:31 -04002116 return -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002117
Al Virof0c3b502013-05-16 12:07:31 -04002118 if (!dir_emit_dots(file, ctx))
2119 goto out;
2120
2121 if (ctx->pos >= nents + 2)
2122 goto out;
2123
2124 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2125 if (!proc_fill_cache(file, ctx, p->name, p->len,
2126 proc_pident_instantiate, task, p))
2127 break;
2128 ctx->pos++;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002129 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002130out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002131 put_task_struct(task);
Al Virof0c3b502013-05-16 12:07:31 -04002132 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133}
2134
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002136static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2137 size_t count, loff_t *ppos)
2138{
Al Viro496ad9a2013-01-23 17:07:38 -05002139 struct inode * inode = file_inode(file);
Al Viro04ff9702007-03-12 16:17:58 +00002140 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002141 ssize_t length;
2142 struct task_struct *task = get_proc_task(inode);
2143
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002144 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002145 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002146
2147 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002148 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002149 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002150 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002151 if (length > 0)
2152 length = simple_read_from_buffer(buf, count, ppos, p, length);
2153 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002154 return length;
2155}
2156
2157static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2158 size_t count, loff_t *ppos)
2159{
Al Viro496ad9a2013-01-23 17:07:38 -05002160 struct inode * inode = file_inode(file);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002161 char *page;
2162 ssize_t length;
2163 struct task_struct *task = get_proc_task(inode);
2164
2165 length = -ESRCH;
2166 if (!task)
2167 goto out_no_task;
2168 if (count > PAGE_SIZE)
2169 count = PAGE_SIZE;
2170
2171 /* No partial writes. */
2172 length = -EINVAL;
2173 if (*ppos != 0)
2174 goto out;
2175
2176 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002177 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002178 if (!page)
2179 goto out;
2180
2181 length = -EFAULT;
2182 if (copy_from_user(page, buf, count))
2183 goto out_free;
2184
David Howells107db7c2009-05-08 13:55:27 +01002185 /* Guard against adverse ptrace interaction */
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002186 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
David Howells107db7c2009-05-08 13:55:27 +01002187 if (length < 0)
2188 goto out_free;
2189
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002190 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002191 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002192 (void*)page, count);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002193 mutex_unlock(&task->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002194out_free:
2195 free_page((unsigned long) page);
2196out:
2197 put_task_struct(task);
2198out_no_task:
2199 return length;
2200}
2201
Arjan van de Ven00977a52007-02-12 00:55:34 -08002202static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002203 .read = proc_pid_attr_read,
2204 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002205 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002206};
2207
Eric Dumazetc5141e62007-05-08 00:26:15 -07002208static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002209 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2210 REG("prev", S_IRUGO, proc_pid_attr_operations),
2211 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2212 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2213 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2214 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002215};
2216
Al Virof0c3b502013-05-16 12:07:31 -04002217static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218{
Al Virof0c3b502013-05-16 12:07:31 -04002219 return proc_pident_readdir(file, ctx,
2220 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221}
2222
Arjan van de Ven00977a52007-02-12 00:55:34 -08002223static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 .read = generic_read_dir,
Al Virof0c3b502013-05-16 12:07:31 -04002225 .iterate = proc_attr_dir_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002226 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227};
2228
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002229static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Al Viro00cd8dd2012-06-10 17:13:09 -04002230 struct dentry *dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002232 return proc_pident_lookup(dir, dentry,
2233 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234}
2235
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002236static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002237 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002238 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002239 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240};
2241
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242#endif
2243
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002244#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002245static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2246 size_t count, loff_t *ppos)
2247{
Al Viro496ad9a2013-01-23 17:07:38 -05002248 struct task_struct *task = get_proc_task(file_inode(file));
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002249 struct mm_struct *mm;
2250 char buffer[PROC_NUMBUF];
2251 size_t len;
2252 int ret;
2253
2254 if (!task)
2255 return -ESRCH;
2256
2257 ret = 0;
2258 mm = get_task_mm(task);
2259 if (mm) {
2260 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2261 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2262 MMF_DUMP_FILTER_SHIFT));
2263 mmput(mm);
2264 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2265 }
2266
2267 put_task_struct(task);
2268
2269 return ret;
2270}
2271
2272static ssize_t proc_coredump_filter_write(struct file *file,
2273 const char __user *buf,
2274 size_t count,
2275 loff_t *ppos)
2276{
2277 struct task_struct *task;
2278 struct mm_struct *mm;
2279 char buffer[PROC_NUMBUF], *end;
2280 unsigned int val;
2281 int ret;
2282 int i;
2283 unsigned long mask;
2284
2285 ret = -EFAULT;
2286 memset(buffer, 0, sizeof(buffer));
2287 if (count > sizeof(buffer) - 1)
2288 count = sizeof(buffer) - 1;
2289 if (copy_from_user(buffer, buf, count))
2290 goto out_no_task;
2291
2292 ret = -EINVAL;
2293 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2294 if (*end == '\n')
2295 end++;
2296 if (end - buffer == 0)
2297 goto out_no_task;
2298
2299 ret = -ESRCH;
Al Viro496ad9a2013-01-23 17:07:38 -05002300 task = get_proc_task(file_inode(file));
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002301 if (!task)
2302 goto out_no_task;
2303
2304 ret = end - buffer;
2305 mm = get_task_mm(task);
2306 if (!mm)
2307 goto out_no_mm;
2308
2309 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2310 if (val & mask)
2311 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2312 else
2313 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2314 }
2315
2316 mmput(mm);
2317 out_no_mm:
2318 put_task_struct(task);
2319 out_no_task:
2320 return ret;
2321}
2322
2323static const struct file_operations proc_coredump_filter_operations = {
2324 .read = proc_coredump_filter_read,
2325 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002326 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002327};
2328#endif
2329
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002330#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002331static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002332{
Andrea Righi940389b2008-07-28 00:48:12 +02002333 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002334 unsigned long flags;
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002335 int result;
Andrea Righi297c5d92008-07-25 01:48:49 -07002336
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002337 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2338 if (result)
2339 return result;
2340
2341 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2342 result = -EACCES;
2343 goto out_unlock;
2344 }
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002345
Andrea Righi59954772008-07-27 17:29:15 +02002346 if (whole && lock_task_sighand(task, &flags)) {
2347 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002348
Andrea Righi59954772008-07-27 17:29:15 +02002349 task_io_accounting_add(&acct, &task->signal->ioac);
2350 while_each_thread(task, t)
2351 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002352
Andrea Righi59954772008-07-27 17:29:15 +02002353 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002354 }
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002355 result = seq_printf(m,
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002356 "rchar: %llu\n"
2357 "wchar: %llu\n"
2358 "syscr: %llu\n"
2359 "syscw: %llu\n"
2360 "read_bytes: %llu\n"
2361 "write_bytes: %llu\n"
2362 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002363 (unsigned long long)acct.rchar,
2364 (unsigned long long)acct.wchar,
2365 (unsigned long long)acct.syscr,
2366 (unsigned long long)acct.syscw,
2367 (unsigned long long)acct.read_bytes,
2368 (unsigned long long)acct.write_bytes,
2369 (unsigned long long)acct.cancelled_write_bytes);
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002370out_unlock:
2371 mutex_unlock(&task->signal->cred_guard_mutex);
2372 return result;
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002373}
Andrea Righi297c5d92008-07-25 01:48:49 -07002374
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002375static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2376 struct pid *pid, struct task_struct *task)
Andrea Righi297c5d92008-07-25 01:48:49 -07002377{
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002378 return do_io_accounting(task, m, 0);
Andrea Righi297c5d92008-07-25 01:48:49 -07002379}
2380
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002381static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2382 struct pid *pid, struct task_struct *task)
Andrea Righi297c5d92008-07-25 01:48:49 -07002383{
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002384 return do_io_accounting(task, m, 1);
Andrea Righi297c5d92008-07-25 01:48:49 -07002385}
2386#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002387
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002388#ifdef CONFIG_USER_NS
2389static int proc_id_map_open(struct inode *inode, struct file *file,
Fabian Frederickccf94f12014-08-08 14:21:22 -07002390 const struct seq_operations *seq_ops)
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002391{
2392 struct user_namespace *ns = NULL;
2393 struct task_struct *task;
2394 struct seq_file *seq;
2395 int ret = -EINVAL;
2396
2397 task = get_proc_task(inode);
2398 if (task) {
2399 rcu_read_lock();
2400 ns = get_user_ns(task_cred_xxx(task, user_ns));
2401 rcu_read_unlock();
2402 put_task_struct(task);
2403 }
2404 if (!ns)
2405 goto err;
2406
2407 ret = seq_open(file, seq_ops);
2408 if (ret)
2409 goto err_put_ns;
2410
2411 seq = file->private_data;
2412 seq->private = ns;
2413
2414 return 0;
2415err_put_ns:
2416 put_user_ns(ns);
2417err:
2418 return ret;
2419}
2420
2421static int proc_id_map_release(struct inode *inode, struct file *file)
2422{
2423 struct seq_file *seq = file->private_data;
2424 struct user_namespace *ns = seq->private;
2425 put_user_ns(ns);
2426 return seq_release(inode, file);
2427}
2428
2429static int proc_uid_map_open(struct inode *inode, struct file *file)
2430{
2431 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2432}
2433
2434static int proc_gid_map_open(struct inode *inode, struct file *file)
2435{
2436 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2437}
2438
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002439static int proc_projid_map_open(struct inode *inode, struct file *file)
2440{
2441 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2442}
2443
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002444static const struct file_operations proc_uid_map_operations = {
2445 .open = proc_uid_map_open,
2446 .write = proc_uid_map_write,
2447 .read = seq_read,
2448 .llseek = seq_lseek,
2449 .release = proc_id_map_release,
2450};
2451
2452static const struct file_operations proc_gid_map_operations = {
2453 .open = proc_gid_map_open,
2454 .write = proc_gid_map_write,
2455 .read = seq_read,
2456 .llseek = seq_lseek,
2457 .release = proc_id_map_release,
2458};
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002459
2460static const struct file_operations proc_projid_map_operations = {
2461 .open = proc_projid_map_open,
2462 .write = proc_projid_map_write,
2463 .read = seq_read,
2464 .llseek = seq_lseek,
2465 .release = proc_id_map_release,
2466};
Eric W. Biederman9cc46512014-12-02 12:27:26 -06002467
2468static int proc_setgroups_open(struct inode *inode, struct file *file)
2469{
2470 struct user_namespace *ns = NULL;
2471 struct task_struct *task;
2472 int ret;
2473
2474 ret = -ESRCH;
2475 task = get_proc_task(inode);
2476 if (task) {
2477 rcu_read_lock();
2478 ns = get_user_ns(task_cred_xxx(task, user_ns));
2479 rcu_read_unlock();
2480 put_task_struct(task);
2481 }
2482 if (!ns)
2483 goto err;
2484
2485 if (file->f_mode & FMODE_WRITE) {
2486 ret = -EACCES;
2487 if (!ns_capable(ns, CAP_SYS_ADMIN))
2488 goto err_put_ns;
2489 }
2490
2491 ret = single_open(file, &proc_setgroups_show, ns);
2492 if (ret)
2493 goto err_put_ns;
2494
2495 return 0;
2496err_put_ns:
2497 put_user_ns(ns);
2498err:
2499 return ret;
2500}
2501
2502static int proc_setgroups_release(struct inode *inode, struct file *file)
2503{
2504 struct seq_file *seq = file->private_data;
2505 struct user_namespace *ns = seq->private;
2506 int ret = single_release(inode, file);
2507 put_user_ns(ns);
2508 return ret;
2509}
2510
2511static const struct file_operations proc_setgroups_operations = {
2512 .open = proc_setgroups_open,
2513 .write = proc_setgroups_write,
2514 .read = seq_read,
2515 .llseek = seq_lseek,
2516 .release = proc_setgroups_release,
2517};
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002518#endif /* CONFIG_USER_NS */
2519
Kees Cook47830722008-10-06 03:11:58 +04002520static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2521 struct pid *pid, struct task_struct *task)
2522{
Al Viroa9712bc2011-03-23 15:52:50 -04002523 int err = lock_trace(task);
2524 if (!err) {
2525 seq_printf(m, "%08x\n", task->personality);
2526 unlock_trace(task);
2527 }
2528 return err;
Kees Cook47830722008-10-06 03:11:58 +04002529}
2530
Eric W. Biederman801199c2006-10-02 02:18:48 -07002531/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002532 * Thread groups
2533 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002534static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002535static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002536
Eric Dumazetc5141e62007-05-08 00:26:15 -07002537static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002538 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2539 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002540#ifdef CONFIG_CHECKPOINT_RESTORE
2541 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2542#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002543 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08002544 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002545#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002546 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002547#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002548 REG("environ", S_IRUSR, proc_environ_operations),
Alexey Dobriyanf9ea5362014-08-08 14:21:35 -07002549 ONE("auxv", S_IRUSR, proc_pid_auxv),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002550 ONE("status", S_IRUGO, proc_pid_status),
Djalal Harouni35a35042014-04-07 15:38:36 -07002551 ONE("personality", S_IRUSR, proc_pid_personality),
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -07002552 ONE("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002553#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002554 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002555#endif
Mike Galbraith5091faa2010-11-30 14:18:03 +01002556#ifdef CONFIG_SCHED_AUTOGROUP
2557 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2558#endif
john stultz4614a696b2009-12-14 18:00:05 -08002559 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002560#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -07002561 ONE("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002562#endif
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -07002563 ONE("cmdline", S_IRUGO, proc_pid_cmdline),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002564 ONE("stat", S_IRUGO, proc_tgid_stat),
2565 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002566 REG("maps", S_IRUGO, proc_pid_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002567#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002568 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002569#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002570 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2571 LNK("cwd", proc_cwd_link),
2572 LNK("root", proc_root_link),
2573 LNK("exe", proc_exe_link),
2574 REG("mounts", S_IRUGO, proc_mounts_operations),
2575 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2576 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002577#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002578 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002579 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
Djalal Harouni32ed74a2014-04-07 15:38:38 -07002580 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002581#endif
2582#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002583 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002584#endif
2585#ifdef CONFIG_KALLSYMS
Alexey Dobriyanedfcd602014-08-08 14:21:44 -07002586 ONE("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002587#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002588#ifdef CONFIG_STACKTRACE
Djalal Harouni35a35042014-04-07 15:38:36 -07002589 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002590#endif
2591#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -07002592 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002593#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002594#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002595 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002596#endif
Paul Menage8793d852007-10-18 23:39:39 -07002597#ifdef CONFIG_PROC_PID_CPUSET
Zefan Li52de4772014-09-18 16:03:36 +08002598 ONE("cpuset", S_IRUGO, proc_cpuset_show),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002599#endif
Paul Menagea4243162007-10-18 23:39:35 -07002600#ifdef CONFIG_CGROUPS
Zefan Li006f4ac2014-09-18 16:03:15 +08002601 ONE("cgroup", S_IRUGO, proc_cgroup_show),
Paul Menagea4243162007-10-18 23:39:35 -07002602#endif
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -07002603 ONE("oom_score", S_IRUGO, proc_oom_score),
David Rientjesfa0cbbf2012-11-12 17:53:04 -08002604 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07002605 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002606#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002607 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2608 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002609#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002610#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002611 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002612#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002613#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002614 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002615#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002616#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002617 ONE("io", S_IRUSR, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002618#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04002619#ifdef CONFIG_HARDWALL
Alexey Dobriyand962c142014-08-08 14:21:52 -07002620 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
Chris Metcalff133ecc2011-05-26 12:40:09 -04002621#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002622#ifdef CONFIG_USER_NS
2623 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2624 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002625 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
Eric W. Biederman9cc46512014-12-02 12:27:26 -06002626 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002627#endif
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002628#ifdef CONFIG_CHECKPOINT_RESTORE
2629 REG("timers", S_IRUGO, proc_timers_operations),
2630#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002631};
2632
Al Virof0c3b502013-05-16 12:07:31 -04002633static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002634{
Al Virof0c3b502013-05-16 12:07:31 -04002635 return proc_pident_readdir(file, ctx,
2636 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002637}
2638
Arjan van de Ven00977a52007-02-12 00:55:34 -08002639static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002640 .read = generic_read_dir,
Al Virof0c3b502013-05-16 12:07:31 -04002641 .iterate = proc_tgid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002642 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002643};
2644
Al Viro00cd8dd2012-06-10 17:13:09 -04002645static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2646{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002647 return proc_pident_lookup(dir, dentry,
2648 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002649}
2650
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002651static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002652 .lookup = proc_tgid_base_lookup,
2653 .getattr = pid_getattr,
2654 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08002655 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002656};
2657
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002658static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002660 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002661 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002662 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663
Eric W. Biederman48e64842006-06-26 00:25:48 -07002664 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002665 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Al Viro4f522a22013-02-11 23:20:37 -05002666 /* no ->d_hash() rejects on procfs */
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002667 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002668 if (dentry) {
Eric W. Biedermanbbd51922014-02-13 10:24:23 -08002669 d_invalidate(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002670 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002671 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002672
Oleg Nesterovc35a7f12014-12-10 15:54:56 -08002673 if (pid == tgid)
2674 return;
2675
Eric W. Biederman48e64842006-06-26 00:25:48 -07002676 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002677 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2678 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002679 if (!leader)
2680 goto out;
2681
2682 name.name = "task";
2683 name.len = strlen(name.name);
2684 dir = d_hash_and_lookup(leader, &name);
2685 if (!dir)
2686 goto out_put_leader;
2687
2688 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002689 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002690 dentry = d_hash_and_lookup(dir, &name);
2691 if (dentry) {
Eric W. Biedermanbbd51922014-02-13 10:24:23 -08002692 d_invalidate(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002693 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002694 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002695
2696 dput(dir);
2697out_put_leader:
2698 dput(leader);
2699out:
2700 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002701}
2702
Randy Dunlap0895e912007-10-21 21:00:10 -07002703/**
2704 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2705 * @task: task that should be flushed.
2706 *
2707 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002708 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002709 * in. This call is supposed to do all of this job.
2710 *
2711 * Looks in the dcache for
2712 * /proc/@pid
2713 * /proc/@tgid/task/@pid
2714 * if either directory is present flushes it and all of it'ts children
2715 * from the dcache.
2716 *
2717 * It is safe and reasonable to cache /proc entries for a task until
2718 * that task exits. After that they just clog up the dcache with
2719 * useless entries, possibly causing useful dcache entries to be
2720 * flushed instead. This routine is proved to flush those useless
2721 * dcache entries at process exit time.
2722 *
2723 * NOTE: This routine is just an optimization so it does not guarantee
2724 * that no dcache entries will exist at process exit time it
2725 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002726 */
2727
2728void proc_flush_task(struct task_struct *task)
2729{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002730 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002731 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002732 struct upid *upid;
2733
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002734 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002735 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002736
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002737 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002738 upid = &pid->numbers[i];
2739 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002740 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002741 }
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002742}
2743
Al Viroc52a47a2013-06-15 11:15:20 +04002744static int proc_pid_instantiate(struct inode *dir,
2745 struct dentry * dentry,
2746 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002747{
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002748 struct inode *inode;
2749
Eric W. Biederman61a28782006-10-02 02:18:49 -07002750 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002751 if (!inode)
2752 goto out;
2753
2754 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2755 inode->i_op = &proc_tgid_base_inode_operations;
2756 inode->i_fop = &proc_tgid_base_operations;
2757 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002758
Miklos Szeredibfe86842011-10-28 14:13:29 +02002759 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2760 ARRAY_SIZE(tgid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002761
Nick Pigginfb045ad2011-01-07 17:49:55 +11002762 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002763
2764 d_add(dentry, inode);
2765 /* Close the race of the process dying before we return the dentry */
Al Viro0b728e12012-06-10 16:03:43 -04002766 if (pid_revalidate(dentry, 0))
Al Viroc52a47a2013-06-15 11:15:20 +04002767 return 0;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002768out:
Al Viroc52a47a2013-06-15 11:15:20 +04002769 return -ENOENT;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002770}
2771
Al Viro00cd8dd2012-06-10 17:13:09 -04002772struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002773{
Alexey Dobriyan335eb532014-08-08 14:21:27 -07002774 int result = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002775 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002777 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002778
Alexey Dobriyandbcdb502014-08-08 14:21:25 -07002779 tgid = name_to_int(&dentry->d_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780 if (tgid == ~0U)
2781 goto out;
2782
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002783 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07002784 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002785 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002786 if (task)
2787 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002788 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789 if (!task)
2790 goto out;
2791
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002792 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002793 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794out:
Al Viroc52a47a2013-06-15 11:15:20 +04002795 return ERR_PTR(result);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002796}
2797
Linus Torvalds1da177e2005-04-16 15:20:36 -07002798/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002799 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002800 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002801 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002802struct tgid_iter {
2803 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002804 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002805};
2806static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2807{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002808 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002809
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002810 if (iter.task)
2811 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002812 rcu_read_lock();
2813retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002814 iter.task = NULL;
2815 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002816 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002817 iter.tgid = pid_nr_ns(pid, ns);
2818 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002819 /* What we to know is if the pid we have find is the
2820 * pid of a thread_group_leader. Testing for task
2821 * being a thread_group_leader is the obvious thing
2822 * todo but there is a window when it fails, due to
2823 * the pid transfer logic in de_thread.
2824 *
2825 * So we perform the straight forward test of seeing
2826 * if the pid we have found is the pid of a thread
2827 * group leader, and don't worry if the task we have
2828 * found doesn't happen to be a thread group leader.
2829 * As we don't care in the case of readdir.
2830 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002831 if (!iter.task || !has_group_leader_pid(iter.task)) {
2832 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002833 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002834 }
2835 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07002837 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002838 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002839}
2840
Eric W. Biederman00978752014-07-31 03:10:50 -07002841#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002842
2843/* for the /proc/ directory itself, after non-process stuff has been done */
Al Virof0c3b502013-05-16 12:07:31 -04002844int proc_pid_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845{
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002846 struct tgid_iter iter;
Al Viro3aa33772014-10-31 00:42:35 -04002847 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
Al Virof0c3b502013-05-16 12:07:31 -04002848 loff_t pos = ctx->pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002849
Al Viro021ada72013-03-29 19:27:05 -04002850 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
Al Virof0c3b502013-05-16 12:07:31 -04002851 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002852
Eric W. Biederman00978752014-07-31 03:10:50 -07002853 if (pos == TGID_OFFSET - 2) {
Al Virodb963162013-06-15 10:45:10 +04002854 struct inode *inode = ns->proc_self->d_inode;
2855 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
Al Virof0c3b502013-05-16 12:07:31 -04002856 return 0;
Eric W. Biederman00978752014-07-31 03:10:50 -07002857 ctx->pos = pos = pos + 1;
Al Viro021ada72013-03-29 19:27:05 -04002858 }
Eric W. Biederman00978752014-07-31 03:10:50 -07002859 if (pos == TGID_OFFSET - 1) {
2860 struct inode *inode = ns->proc_thread_self->d_inode;
2861 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
2862 return 0;
2863 ctx->pos = pos = pos + 1;
2864 }
2865 iter.tgid = pos - TGID_OFFSET;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002866 iter.task = NULL;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002867 for (iter = next_tgid(ns, iter);
2868 iter.task;
2869 iter.tgid += 1, iter = next_tgid(ns, iter)) {
Al Virof0c3b502013-05-16 12:07:31 -04002870 char name[PROC_NUMBUF];
2871 int len;
2872 if (!has_pid_permissions(ns, iter.task, 2))
2873 continue;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08002874
Al Virof0c3b502013-05-16 12:07:31 -04002875 len = snprintf(name, sizeof(name), "%d", iter.tgid);
2876 ctx->pos = iter.tgid + TGID_OFFSET;
2877 if (!proc_fill_cache(file, ctx, name, len,
2878 proc_pid_instantiate, iter.task, NULL)) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002879 put_task_struct(iter.task);
Al Virof0c3b502013-05-16 12:07:31 -04002880 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002881 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002882 }
Al Virof0c3b502013-05-16 12:07:31 -04002883 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002884 return 0;
2885}
2886
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002887/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002888 * Tasks
2889 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002890static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002891 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07002892 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08002893 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Eric W. Biederman6ba8ed72014-07-31 16:27:08 -07002894#ifdef CONFIG_NET
2895 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2896#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002897 REG("environ", S_IRUSR, proc_environ_operations),
Alexey Dobriyanf9ea5362014-08-08 14:21:35 -07002898 ONE("auxv", S_IRUSR, proc_pid_auxv),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002899 ONE("status", S_IRUGO, proc_pid_status),
Djalal Harouni35a35042014-04-07 15:38:36 -07002900 ONE("personality", S_IRUSR, proc_pid_personality),
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -07002901 ONE("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002902#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002903 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002904#endif
john stultz4614a696b2009-12-14 18:00:05 -08002905 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002906#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -07002907 ONE("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002908#endif
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -07002909 ONE("cmdline", S_IRUGO, proc_pid_cmdline),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002910 ONE("stat", S_IRUGO, proc_tid_stat),
2911 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002912 REG("maps", S_IRUGO, proc_tid_maps_operations),
Cyrill Gorcunov818411612012-05-31 16:26:43 -07002913#ifdef CONFIG_CHECKPOINT_RESTORE
2914 REG("children", S_IRUGO, proc_tid_children_operations),
2915#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002916#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002917 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002918#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002919 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2920 LNK("cwd", proc_cwd_link),
2921 LNK("root", proc_root_link),
2922 LNK("exe", proc_exe_link),
2923 REG("mounts", S_IRUGO, proc_mounts_operations),
2924 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002925#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002926 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002927 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
Djalal Harouni32ed74a2014-04-07 15:38:38 -07002928 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002929#endif
2930#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002931 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002932#endif
2933#ifdef CONFIG_KALLSYMS
Alexey Dobriyanedfcd602014-08-08 14:21:44 -07002934 ONE("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002935#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002936#ifdef CONFIG_STACKTRACE
Djalal Harouni35a35042014-04-07 15:38:36 -07002937 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002938#endif
2939#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -07002940 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002941#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002942#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002943 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002944#endif
Paul Menage8793d852007-10-18 23:39:39 -07002945#ifdef CONFIG_PROC_PID_CPUSET
Zefan Li52de4772014-09-18 16:03:36 +08002946 ONE("cpuset", S_IRUGO, proc_cpuset_show),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002947#endif
Paul Menagea4243162007-10-18 23:39:35 -07002948#ifdef CONFIG_CGROUPS
Zefan Li006f4ac2014-09-18 16:03:15 +08002949 ONE("cgroup", S_IRUGO, proc_cgroup_show),
Paul Menagea4243162007-10-18 23:39:35 -07002950#endif
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -07002951 ONE("oom_score", S_IRUGO, proc_oom_score),
David Rientjesfa0cbbf2012-11-12 17:53:04 -08002952 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07002953 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002954#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002955 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
Al Viro26ec3c62011-02-15 21:24:05 -05002956 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002957#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002958#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002959 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002960#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07002961#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002962 ONE("io", S_IRUSR, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07002963#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04002964#ifdef CONFIG_HARDWALL
Alexey Dobriyand962c142014-08-08 14:21:52 -07002965 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
Chris Metcalff133ecc2011-05-26 12:40:09 -04002966#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002967#ifdef CONFIG_USER_NS
2968 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2969 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002970 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
Eric W. Biederman9cc46512014-12-02 12:27:26 -06002971 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002972#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002973};
2974
Al Virof0c3b502013-05-16 12:07:31 -04002975static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002976{
Al Virof0c3b502013-05-16 12:07:31 -04002977 return proc_pident_readdir(file, ctx,
2978 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002979}
2980
Al Viro00cd8dd2012-06-10 17:13:09 -04002981static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2982{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002983 return proc_pident_lookup(dir, dentry,
2984 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002985}
2986
Arjan van de Ven00977a52007-02-12 00:55:34 -08002987static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002988 .read = generic_read_dir,
Al Virof0c3b502013-05-16 12:07:31 -04002989 .iterate = proc_tid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002990 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002991};
2992
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002993static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002994 .lookup = proc_tid_base_lookup,
2995 .getattr = pid_getattr,
2996 .setattr = proc_setattr,
2997};
2998
Al Viroc52a47a2013-06-15 11:15:20 +04002999static int proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003000 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003001{
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003002 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07003003 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003004
3005 if (!inode)
3006 goto out;
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;
Vegard Nossumaed54172008-06-05 22:46:53 -07003011
Miklos Szeredibfe86842011-10-28 14:13:29 +02003012 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3013 ARRAY_SIZE(tid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003014
Nick Pigginfb045ad2011-01-07 17:49:55 +11003015 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003016
3017 d_add(dentry, inode);
3018 /* Close the race of the process dying before we return the dentry */
Al Viro0b728e12012-06-10 16:03:43 -04003019 if (pid_revalidate(dentry, 0))
Al Viroc52a47a2013-06-15 11:15:20 +04003020 return 0;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003021out:
Al Viroc52a47a2013-06-15 11:15:20 +04003022 return -ENOENT;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003023}
3024
Al Viro00cd8dd2012-06-10 17:13:09 -04003025static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003026{
Al Viroc52a47a2013-06-15 11:15:20 +04003027 int result = -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003028 struct task_struct *task;
3029 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003030 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003031 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003032
3033 if (!leader)
3034 goto out_no_task;
3035
Alexey Dobriyandbcdb502014-08-08 14:21:25 -07003036 tid = name_to_int(&dentry->d_name);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003037 if (tid == ~0U)
3038 goto out;
3039
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003040 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003041 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003042 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003043 if (task)
3044 get_task_struct(task);
3045 rcu_read_unlock();
3046 if (!task)
3047 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003048 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003049 goto out_drop_task;
3050
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003051 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003052out_drop_task:
3053 put_task_struct(task);
3054out:
3055 put_task_struct(leader);
3056out_no_task:
Al Viroc52a47a2013-06-15 11:15:20 +04003057 return ERR_PTR(result);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003058}
3059
3060/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003061 * Find the first tid of a thread group to return to user space.
3062 *
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.
3066 *
3067 * In the case of a short read we start with find_task_by_pid.
3068 *
3069 * In the case of a seek we start with the leader and walk nr
3070 * threads past it.
3071 */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003072static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3073 struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003074{
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003075 struct task_struct *pos, *task;
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003076 unsigned long nr = f_pos;
3077
3078 if (nr != f_pos) /* 32bit overflow? */
3079 return NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003080
Eric W. Biedermancc288732006-06-26 00:26:01 -07003081 rcu_read_lock();
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003082 task = pid_task(pid, PIDTYPE_PID);
3083 if (!task)
3084 goto fail;
3085
3086 /* Attempt to start with the tid of a thread */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003087 if (tid && nr) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003088 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003089 if (pos && same_thread_group(pos, task))
Oleg Nesterova872ff02006-06-26 00:26:01 -07003090 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003091 }
3092
3093 /* If nr exceeds the number of threads there is nothing todo */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003094 if (nr >= get_nr_threads(task))
Oleg Nesterovc986c142014-01-23 15:55:38 -08003095 goto fail;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003096
3097 /* If we haven't found our starting place yet start
3098 * with the leader and walk nr threads forward.
3099 */
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003100 pos = task = task->group_leader;
Oleg Nesterovc986c142014-01-23 15:55:38 -08003101 do {
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003102 if (!nr--)
Oleg Nesterovc986c142014-01-23 15:55:38 -08003103 goto found;
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003104 } while_each_thread(task, pos);
Oleg Nesterovc986c142014-01-23 15:55:38 -08003105fail:
3106 pos = NULL;
3107 goto out;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003108found:
3109 get_task_struct(pos);
3110out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003111 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003112 return pos;
3113}
3114
3115/*
3116 * Find the next thread in the thread list.
3117 * Return NULL if there is an error or no next thread.
3118 *
3119 * The reference to the input task_struct is released.
3120 */
3121static struct task_struct *next_tid(struct task_struct *start)
3122{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003123 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003124 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003125 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003126 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003127 if (thread_group_leader(pos))
3128 pos = NULL;
3129 else
3130 get_task_struct(pos);
3131 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003132 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003133 put_task_struct(start);
3134 return pos;
3135}
3136
Linus Torvalds1da177e2005-04-16 15:20:36 -07003137/* for the /proc/TGID/task/ directories */
Al Virof0c3b502013-05-16 12:07:31 -04003138static int proc_task_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003139{
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003140 struct inode *inode = file_inode(file);
3141 struct task_struct *task;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003142 struct pid_namespace *ns;
Al Virof0c3b502013-05-16 12:07:31 -04003143 int tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003144
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003145 if (proc_inode_is_dead(inode))
Al Virof0c3b502013-05-16 12:07:31 -04003146 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003147
Al Virof0c3b502013-05-16 12:07:31 -04003148 if (!dir_emit_dots(file, ctx))
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003149 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003150
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003151 /* f_version caches the tgid value that the last readdir call couldn't
3152 * return. lseek aka telldir automagically resets f_version to 0.
3153 */
Al Viro3aa33772014-10-31 00:42:35 -04003154 ns = inode->i_sb->s_fs_info;
Al Virof0c3b502013-05-16 12:07:31 -04003155 tid = (int)file->f_version;
3156 file->f_version = 0;
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003157 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003158 task;
Al Virof0c3b502013-05-16 12:07:31 -04003159 task = next_tid(task), ctx->pos++) {
3160 char name[PROC_NUMBUF];
3161 int len;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003162 tid = task_pid_nr_ns(task, ns);
Al Virof0c3b502013-05-16 12:07:31 -04003163 len = snprintf(name, sizeof(name), "%d", tid);
3164 if (!proc_fill_cache(file, ctx, name, len,
3165 proc_task_instantiate, task, NULL)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003166 /* returning this tgid failed, save it as the first
3167 * pid for the next readir call */
Al Virof0c3b502013-05-16 12:07:31 -04003168 file->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003169 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003170 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003171 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003172 }
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003173
Al Virof0c3b502013-05-16 12:07:31 -04003174 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003175}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003176
3177static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3178{
3179 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003180 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003181 generic_fillattr(inode, stat);
3182
Eric W. Biederman99f89552006-06-26 00:25:55 -07003183 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003184 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003185 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003186 }
3187
3188 return 0;
3189}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003190
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003191static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003192 .lookup = proc_task_lookup,
3193 .getattr = proc_task_getattr,
3194 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003195 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003196};
3197
Arjan van de Ven00977a52007-02-12 00:55:34 -08003198static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003199 .read = generic_read_dir,
Al Virof0c3b502013-05-16 12:07:31 -04003200 .iterate = proc_task_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003201 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003202};