projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
[PATCH] proc: don't lock task_structs indefinitely
[deliverable/linux.git] / fs / proc / base.c
1 /*
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.
14 *
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.
48 */
49
50 #include <asm/uaccess.h>
51
52 #include <linux/config.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/string.h>
61 #include <linux/seq_file.h>
62 #include <linux/namei.h>
63 #include <linux/namespace.h>
64 #include <linux/mm.h>
65 #include <linux/smp_lock.h>
66 #include <linux/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/mount.h>
69 #include <linux/security.h>
70 #include <linux/ptrace.h>
71 #include <linux/seccomp.h>
72 #include <linux/cpuset.h>
73 #include <linux/audit.h>
74 #include <linux/poll.h>
75 #include "internal.h"
76
77 /* NOTE:
78 * Implementing inode permission operations in /proc is almost
79 * certainly an error. Permission checks need to happen during
80 * each system call not at open time. The reason is that most of
81 * what we wish to check for permissions in /proc varies at runtime.
82 *
83 * The classic example of a problem is opening file descriptors
84 * in /proc for a task before it execs a suid executable.
85 */
86
87 /*
88 * For hysterical raisins we keep the same inumbers as in the old procfs.
89 * Feel free to change the macro below - just keep the range distinct from
90 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
91 * As soon as we'll get a separate superblock we will be able to forget
92 * about magical ranges too.
93 */
94
95 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
96
97 enum pid_directory_inos {
98 PROC_TGID_INO = 2,
99 PROC_TGID_TASK,
100 PROC_TGID_STATUS,
101 PROC_TGID_MEM,
102 #ifdef CONFIG_SECCOMP
103 PROC_TGID_SECCOMP,
104 #endif
105 PROC_TGID_CWD,
106 PROC_TGID_ROOT,
107 PROC_TGID_EXE,
108 PROC_TGID_FD,
109 PROC_TGID_ENVIRON,
110 PROC_TGID_AUXV,
111 PROC_TGID_CMDLINE,
112 PROC_TGID_STAT,
113 PROC_TGID_STATM,
114 PROC_TGID_MAPS,
115 PROC_TGID_NUMA_MAPS,
116 PROC_TGID_MOUNTS,
117 PROC_TGID_MOUNTSTATS,
118 PROC_TGID_WCHAN,
119 #ifdef CONFIG_MMU
120 PROC_TGID_SMAPS,
121 #endif
122 #ifdef CONFIG_SCHEDSTATS
123 PROC_TGID_SCHEDSTAT,
124 #endif
125 #ifdef CONFIG_CPUSETS
126 PROC_TGID_CPUSET,
127 #endif
128 #ifdef CONFIG_SECURITY
129 PROC_TGID_ATTR,
130 PROC_TGID_ATTR_CURRENT,
131 PROC_TGID_ATTR_PREV,
132 PROC_TGID_ATTR_EXEC,
133 PROC_TGID_ATTR_FSCREATE,
134 PROC_TGID_ATTR_KEYCREATE,
135 #endif
136 #ifdef CONFIG_AUDITSYSCALL
137 PROC_TGID_LOGINUID,
138 #endif
139 PROC_TGID_OOM_SCORE,
140 PROC_TGID_OOM_ADJUST,
141 PROC_TID_INO,
142 PROC_TID_STATUS,
143 PROC_TID_MEM,
144 #ifdef CONFIG_SECCOMP
145 PROC_TID_SECCOMP,
146 #endif
147 PROC_TID_CWD,
148 PROC_TID_ROOT,
149 PROC_TID_EXE,
150 PROC_TID_FD,
151 PROC_TID_ENVIRON,
152 PROC_TID_AUXV,
153 PROC_TID_CMDLINE,
154 PROC_TID_STAT,
155 PROC_TID_STATM,
156 PROC_TID_MAPS,
157 PROC_TID_NUMA_MAPS,
158 PROC_TID_MOUNTS,
159 PROC_TID_MOUNTSTATS,
160 PROC_TID_WCHAN,
161 #ifdef CONFIG_MMU
162 PROC_TID_SMAPS,
163 #endif
164 #ifdef CONFIG_SCHEDSTATS
165 PROC_TID_SCHEDSTAT,
166 #endif
167 #ifdef CONFIG_CPUSETS
168 PROC_TID_CPUSET,
169 #endif
170 #ifdef CONFIG_SECURITY
171 PROC_TID_ATTR,
172 PROC_TID_ATTR_CURRENT,
173 PROC_TID_ATTR_PREV,
174 PROC_TID_ATTR_EXEC,
175 PROC_TID_ATTR_FSCREATE,
176 PROC_TID_ATTR_KEYCREATE,
177 #endif
178 #ifdef CONFIG_AUDITSYSCALL
179 PROC_TID_LOGINUID,
180 #endif
181 PROC_TID_OOM_SCORE,
182 PROC_TID_OOM_ADJUST,
183
184 /* Add new entries before this */
185 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
186 };
187
188 /* Worst case buffer size needed for holding an integer. */
189 #define PROC_NUMBUF 10
190
191 struct pid_entry {
192 int type;
193 int len;
194 char *name;
195 mode_t mode;
196 };
197
198 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
199
200 static struct pid_entry tgid_base_stuff[] = {
201 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
202 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
203 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
204 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
205 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
206 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
207 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
208 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
209 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
210 #ifdef CONFIG_NUMA
211 E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
212 #endif
213 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
214 #ifdef CONFIG_SECCOMP
215 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
216 #endif
217 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
218 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
219 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
220 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
221 E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
222 #ifdef CONFIG_MMU
223 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
224 #endif
225 #ifdef CONFIG_SECURITY
226 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
227 #endif
228 #ifdef CONFIG_KALLSYMS
229 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
230 #endif
231 #ifdef CONFIG_SCHEDSTATS
232 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
233 #endif
234 #ifdef CONFIG_CPUSETS
235 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
236 #endif
237 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
238 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
239 #ifdef CONFIG_AUDITSYSCALL
240 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
241 #endif
242 {0,0,NULL,0}
243 };
244 static struct pid_entry tid_base_stuff[] = {
245 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
246 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
247 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
248 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
249 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
250 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
251 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
252 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
253 #ifdef CONFIG_NUMA
254 E(PROC_TID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
255 #endif
256 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
257 #ifdef CONFIG_SECCOMP
258 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
259 #endif
260 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
261 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
262 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
263 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
264 #ifdef CONFIG_MMU
265 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
266 #endif
267 #ifdef CONFIG_SECURITY
268 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
269 #endif
270 #ifdef CONFIG_KALLSYMS
271 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
272 #endif
273 #ifdef CONFIG_SCHEDSTATS
274 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
275 #endif
276 #ifdef CONFIG_CPUSETS
277 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
278 #endif
279 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
280 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
281 #ifdef CONFIG_AUDITSYSCALL
282 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
283 #endif
284 {0,0,NULL,0}
285 };
286
287 #ifdef CONFIG_SECURITY
288 static struct pid_entry tgid_attr_stuff[] = {
289 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
290 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
291 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
292 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
293 E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
294 {0,0,NULL,0}
295 };
296 static struct pid_entry tid_attr_stuff[] = {
297 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
298 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
299 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
300 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
301 E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
302 {0,0,NULL,0}
303 };
304 #endif
305
306 #undef E
307
308 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
309 {
310 struct task_struct *task = get_proc_task(inode);
311 struct files_struct *files = NULL;
312 struct file *file;
313 int fd = proc_fd(inode);
314
315 if (task) {
316 files = get_files_struct(task);
317 put_task_struct(task);
318 }
319 if (files) {
320 /*
321 * We are not taking a ref to the file structure, so we must
322 * hold ->file_lock.
323 */
324 spin_lock(&files->file_lock);
325 file = fcheck_files(files, fd);
326 if (file) {
327 *mnt = mntget(file->f_vfsmnt);
328 *dentry = dget(file->f_dentry);
329 spin_unlock(&files->file_lock);
330 put_files_struct(files);
331 return 0;
332 }
333 spin_unlock(&files->file_lock);
334 put_files_struct(files);
335 }
336 return -ENOENT;
337 }
338
339 static struct fs_struct *get_fs_struct(struct task_struct *task)
340 {
341 struct fs_struct *fs;
342 task_lock(task);
343 fs = task->fs;
344 if(fs)
345 atomic_inc(&fs->count);
346 task_unlock(task);
347 return fs;
348 }
349
350 static int get_nr_threads(struct task_struct *tsk)
351 {
352 /* Must be called with the rcu_read_lock held */
353 unsigned long flags;
354 int count = 0;
355
356 if (lock_task_sighand(tsk, &flags)) {
357 count = atomic_read(&tsk->signal->count);
358 unlock_task_sighand(tsk, &flags);
359 }
360 return count;
361 }
362
363 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
364 {
365 struct task_struct *task = get_proc_task(inode);
366 struct fs_struct *fs = NULL;
367 int result = -ENOENT;
368
369 if (task) {
370 fs = get_fs_struct(task);
371 put_task_struct(task);
372 }
373 if (fs) {
374 read_lock(&fs->lock);
375 *mnt = mntget(fs->pwdmnt);
376 *dentry = dget(fs->pwd);
377 read_unlock(&fs->lock);
378 result = 0;
379 put_fs_struct(fs);
380 }
381 return result;
382 }
383
384 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
385 {
386 struct task_struct *task = get_proc_task(inode);
387 struct fs_struct *fs = NULL;
388 int result = -ENOENT;
389
390 if (task) {
391 fs = get_fs_struct(task);
392 put_task_struct(task);
393 }
394 if (fs) {
395 read_lock(&fs->lock);
396 *mnt = mntget(fs->rootmnt);
397 *dentry = dget(fs->root);
398 read_unlock(&fs->lock);
399 result = 0;
400 put_fs_struct(fs);
401 }
402 return result;
403 }
404
405 #define MAY_PTRACE(task) \
406 (task == current || \
407 (task->parent == current && \
408 (task->ptrace & PT_PTRACED) && \
409 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
410 security_ptrace(current,task) == 0))
411
412 static int proc_pid_environ(struct task_struct *task, char * buffer)
413 {
414 int res = 0;
415 struct mm_struct *mm = get_task_mm(task);
416 if (mm) {
417 unsigned int len = mm->env_end - mm->env_start;
418 if (len > PAGE_SIZE)
419 len = PAGE_SIZE;
420 res = access_process_vm(task, mm->env_start, buffer, len, 0);
421 if (!ptrace_may_attach(task))
422 res = -ESRCH;
423 mmput(mm);
424 }
425 return res;
426 }
427
428 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
429 {
430 int res = 0;
431 unsigned int len;
432 struct mm_struct *mm = get_task_mm(task);
433 if (!mm)
434 goto out;
435 if (!mm->arg_end)
436 goto out_mm; /* Shh! No looking before we're done */
437
438 len = mm->arg_end - mm->arg_start;
439
440 if (len > PAGE_SIZE)
441 len = PAGE_SIZE;
442
443 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
444
445 // If the nul at the end of args has been overwritten, then
446 // assume application is using setproctitle(3).
447 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
448 len = strnlen(buffer, res);
449 if (len < res) {
450 res = len;
451 } else {
452 len = mm->env_end - mm->env_start;
453 if (len > PAGE_SIZE - res)
454 len = PAGE_SIZE - res;
455 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
456 res = strnlen(buffer, res);
457 }
458 }
459 out_mm:
460 mmput(mm);
461 out:
462 return res;
463 }
464
465 static int proc_pid_auxv(struct task_struct *task, char *buffer)
466 {
467 int res = 0;
468 struct mm_struct *mm = get_task_mm(task);
469 if (mm) {
470 unsigned int nwords = 0;
471 do
472 nwords += 2;
473 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
474 res = nwords * sizeof(mm->saved_auxv[0]);
475 if (res > PAGE_SIZE)
476 res = PAGE_SIZE;
477 memcpy(buffer, mm->saved_auxv, res);
478 mmput(mm);
479 }
480 return res;
481 }
482
483
484 #ifdef CONFIG_KALLSYMS
485 /*
486 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
487 * Returns the resolved symbol. If that fails, simply return the address.
488 */
489 static int proc_pid_wchan(struct task_struct *task, char *buffer)
490 {
491 char *modname;
492 const char *sym_name;
493 unsigned long wchan, size, offset;
494 char namebuf[KSYM_NAME_LEN+1];
495
496 wchan = get_wchan(task);
497
498 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
499 if (sym_name)
500 return sprintf(buffer, "%s", sym_name);
501 return sprintf(buffer, "%lu", wchan);
502 }
503 #endif /* CONFIG_KALLSYMS */
504
505 #ifdef CONFIG_SCHEDSTATS
506 /*
507 * Provides /proc/PID/schedstat
508 */
509 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
510 {
511 return sprintf(buffer, "%lu %lu %lu\n",
512 task->sched_info.cpu_time,
513 task->sched_info.run_delay,
514 task->sched_info.pcnt);
515 }
516 #endif
517
518 /* The badness from the OOM killer */
519 unsigned long badness(struct task_struct *p, unsigned long uptime);
520 static int proc_oom_score(struct task_struct *task, char *buffer)
521 {
522 unsigned long points;
523 struct timespec uptime;
524
525 do_posix_clock_monotonic_gettime(&uptime);
526 points = badness(task, uptime.tv_sec);
527 return sprintf(buffer, "%lu\n", points);
528 }
529
530 /************************************************************************/
531 /* Here the fs part begins */
532 /************************************************************************/
533
534 /* permission checks */
535
536 /* If the process being read is separated by chroot from the reading process,
537 * don't let the reader access the threads.
538 */
539 static int proc_check_chroot(struct dentry *de, struct vfsmount *mnt)
540 {
541 struct dentry *base;
542 struct vfsmount *our_vfsmnt;
543 int res = 0;
544
545 read_lock(&current->fs->lock);
546 our_vfsmnt = mntget(current->fs->rootmnt);
547 base = dget(current->fs->root);
548 read_unlock(&current->fs->lock);
549
550 spin_lock(&vfsmount_lock);
551
552 while (mnt != our_vfsmnt) {
553 if (mnt == mnt->mnt_parent)
554 goto out;
555 de = mnt->mnt_mountpoint;
556 mnt = mnt->mnt_parent;
557 }
558
559 if (!is_subdir(de, base))
560 goto out;
561 spin_unlock(&vfsmount_lock);
562
563 exit:
564 dput(base);
565 mntput(our_vfsmnt);
566 return res;
567 out:
568 spin_unlock(&vfsmount_lock);
569 res = -EACCES;
570 goto exit;
571 }
572
573 extern struct seq_operations mounts_op;
574 struct proc_mounts {
575 struct seq_file m;
576 int event;
577 };
578
579 static int mounts_open(struct inode *inode, struct file *file)
580 {
581 struct task_struct *task = get_proc_task(inode);
582 struct namespace *namespace = NULL;
583 struct proc_mounts *p;
584 int ret = -EINVAL;
585
586 if (task) {
587 task_lock(task);
588 namespace = task->namespace;
589 if (namespace)
590 get_namespace(namespace);
591 task_unlock(task);
592 put_task_struct(task);
593 }
594
595 if (namespace) {
596 ret = -ENOMEM;
597 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
598 if (p) {
599 file->private_data = &p->m;
600 ret = seq_open(file, &mounts_op);
601 if (!ret) {
602 p->m.private = namespace;
603 p->event = namespace->event;
604 return 0;
605 }
606 kfree(p);
607 }
608 put_namespace(namespace);
609 }
610 return ret;
611 }
612
613 static int mounts_release(struct inode *inode, struct file *file)
614 {
615 struct seq_file *m = file->private_data;
616 struct namespace *namespace = m->private;
617 put_namespace(namespace);
618 return seq_release(inode, file);
619 }
620
621 static unsigned mounts_poll(struct file *file, poll_table *wait)
622 {
623 struct proc_mounts *p = file->private_data;
624 struct namespace *ns = p->m.private;
625 unsigned res = 0;
626
627 poll_wait(file, &ns->poll, wait);
628
629 spin_lock(&vfsmount_lock);
630 if (p->event != ns->event) {
631 p->event = ns->event;
632 res = POLLERR;
633 }
634 spin_unlock(&vfsmount_lock);
635
636 return res;
637 }
638
639 static struct file_operations proc_mounts_operations = {
640 .open = mounts_open,
641 .read = seq_read,
642 .llseek = seq_lseek,
643 .release = mounts_release,
644 .poll = mounts_poll,
645 };
646
647 extern struct seq_operations mountstats_op;
648 static int mountstats_open(struct inode *inode, struct file *file)
649 {
650 int ret = seq_open(file, &mountstats_op);
651
652 if (!ret) {
653 struct seq_file *m = file->private_data;
654 struct namespace *namespace = NULL;
655 struct task_struct *task = get_proc_task(inode);
656
657 if (task) {
658 task_lock(task);
659 namespace = task->namespace;
660 if (namespace)
661 get_namespace(namespace);
662 task_unlock(task);
663 put_task_struct(task);
664 }
665
666 if (namespace)
667 m->private = namespace;
668 else {
669 seq_release(inode, file);
670 ret = -EINVAL;
671 }
672 }
673 return ret;
674 }
675
676 static struct file_operations proc_mountstats_operations = {
677 .open = mountstats_open,
678 .read = seq_read,
679 .llseek = seq_lseek,
680 .release = mounts_release,
681 };
682
683 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
684
685 static ssize_t proc_info_read(struct file * file, char __user * buf,
686 size_t count, loff_t *ppos)
687 {
688 struct inode * inode = file->f_dentry->d_inode;
689 unsigned long page;
690 ssize_t length;
691 struct task_struct *task = get_proc_task(inode);
692
693 length = -ESRCH;
694 if (!task)
695 goto out_no_task;
696
697 if (count > PROC_BLOCK_SIZE)
698 count = PROC_BLOCK_SIZE;
699
700 length = -ENOMEM;
701 if (!(page = __get_free_page(GFP_KERNEL)))
702 goto out;
703
704 length = PROC_I(inode)->op.proc_read(task, (char*)page);
705
706 if (length >= 0)
707 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
708 free_page(page);
709 out:
710 put_task_struct(task);
711 out_no_task:
712 return length;
713 }
714
715 static struct file_operations proc_info_file_operations = {
716 .read = proc_info_read,
717 };
718
719 static int mem_open(struct inode* inode, struct file* file)
720 {
721 file->private_data = (void*)((long)current->self_exec_id);
722 return 0;
723 }
724
725 static ssize_t mem_read(struct file * file, char __user * buf,
726 size_t count, loff_t *ppos)
727 {
728 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
729 char *page;
730 unsigned long src = *ppos;
731 int ret = -ESRCH;
732 struct mm_struct *mm;
733
734 if (!task)
735 goto out_no_task;
736
737 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
738 goto out;
739
740 ret = -ENOMEM;
741 page = (char *)__get_free_page(GFP_USER);
742 if (!page)
743 goto out;
744
745 ret = 0;
746
747 mm = get_task_mm(task);
748 if (!mm)
749 goto out_free;
750
751 ret = -EIO;
752
753 if (file->private_data != (void*)((long)current->self_exec_id))
754 goto out_put;
755
756 ret = 0;
757
758 while (count > 0) {
759 int this_len, retval;
760
761 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
762 retval = access_process_vm(task, src, page, this_len, 0);
763 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
764 if (!ret)
765 ret = -EIO;
766 break;
767 }
768
769 if (copy_to_user(buf, page, retval)) {
770 ret = -EFAULT;
771 break;
772 }
773
774 ret += retval;
775 src += retval;
776 buf += retval;
777 count -= retval;
778 }
779 *ppos = src;
780
781 out_put:
782 mmput(mm);
783 out_free:
784 free_page((unsigned long) page);
785 out:
786 put_task_struct(task);
787 out_no_task:
788 return ret;
789 }
790
791 #define mem_write NULL
792
793 #ifndef mem_write
794 /* This is a security hazard */
795 static ssize_t mem_write(struct file * file, const char * buf,
796 size_t count, loff_t *ppos)
797 {
798 int copied = 0;
799 char *page;
800 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
801 unsigned long dst = *ppos;
802
803 copied = -ESRCH;
804 if (!task)
805 goto out_no_task;
806
807 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
808 goto out;
809
810 copied = -ENOMEM;
811 page = (char *)__get_free_page(GFP_USER);
812 if (!page)
813 goto out;
814
815 while (count > 0) {
816 int this_len, retval;
817
818 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
819 if (copy_from_user(page, buf, this_len)) {
820 copied = -EFAULT;
821 break;
822 }
823 retval = access_process_vm(task, dst, page, this_len, 1);
824 if (!retval) {
825 if (!copied)
826 copied = -EIO;
827 break;
828 }
829 copied += retval;
830 buf += retval;
831 dst += retval;
832 count -= retval;
833 }
834 *ppos = dst;
835 free_page((unsigned long) page);
836 out:
837 put_task_struct(task);
838 out_no_task:
839 return copied;
840 }
841 #endif
842
843 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
844 {
845 switch (orig) {
846 case 0:
847 file->f_pos = offset;
848 break;
849 case 1:
850 file->f_pos += offset;
851 break;
852 default:
853 return -EINVAL;
854 }
855 force_successful_syscall_return();
856 return file->f_pos;
857 }
858
859 static struct file_operations proc_mem_operations = {
860 .llseek = mem_lseek,
861 .read = mem_read,
862 .write = mem_write,
863 .open = mem_open,
864 };
865
866 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
867 size_t count, loff_t *ppos)
868 {
869 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
870 char buffer[PROC_NUMBUF];
871 size_t len;
872 int oom_adjust;
873 loff_t __ppos = *ppos;
874
875 if (!task)
876 return -ESRCH;
877 oom_adjust = task->oomkilladj;
878 put_task_struct(task);
879
880 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
881 if (__ppos >= len)
882 return 0;
883 if (count > len-__ppos)
884 count = len-__ppos;
885 if (copy_to_user(buf, buffer + __ppos, count))
886 return -EFAULT;
887 *ppos = __ppos + count;
888 return count;
889 }
890
891 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
892 size_t count, loff_t *ppos)
893 {
894 struct task_struct *task;
895 char buffer[PROC_NUMBUF], *end;
896 int oom_adjust;
897
898 if (!capable(CAP_SYS_RESOURCE))
899 return -EPERM;
900 memset(buffer, 0, sizeof(buffer));
901 if (count > sizeof(buffer) - 1)
902 count = sizeof(buffer) - 1;
903 if (copy_from_user(buffer, buf, count))
904 return -EFAULT;
905 oom_adjust = simple_strtol(buffer, &end, 0);
906 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
907 return -EINVAL;
908 if (*end == '\n')
909 end++;
910 task = get_proc_task(file->f_dentry->d_inode);
911 if (!task)
912 return -ESRCH;
913 task->oomkilladj = oom_adjust;
914 put_task_struct(task);
915 if (end - buffer == 0)
916 return -EIO;
917 return end - buffer;
918 }
919
920 static struct file_operations proc_oom_adjust_operations = {
921 .read = oom_adjust_read,
922 .write = oom_adjust_write,
923 };
924
925 #ifdef CONFIG_AUDITSYSCALL
926 #define TMPBUFLEN 21
927 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
928 size_t count, loff_t *ppos)
929 {
930 struct inode * inode = file->f_dentry->d_inode;
931 struct task_struct *task = get_proc_task(inode);
932 ssize_t length;
933 char tmpbuf[TMPBUFLEN];
934
935 if (!task)
936 return -ESRCH;
937 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
938 audit_get_loginuid(task->audit_context));
939 put_task_struct(task);
940 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
941 }
942
943 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
944 size_t count, loff_t *ppos)
945 {
946 struct inode * inode = file->f_dentry->d_inode;
947 char *page, *tmp;
948 ssize_t length;
949 uid_t loginuid;
950
951 if (!capable(CAP_AUDIT_CONTROL))
952 return -EPERM;
953
954 if (current != proc_tref(inode)->task)
955 return -EPERM;
956
957 if (count >= PAGE_SIZE)
958 count = PAGE_SIZE - 1;
959
960 if (*ppos != 0) {
961 /* No partial writes. */
962 return -EINVAL;
963 }
964 page = (char*)__get_free_page(GFP_USER);
965 if (!page)
966 return -ENOMEM;
967 length = -EFAULT;
968 if (copy_from_user(page, buf, count))
969 goto out_free_page;
970
971 page[count] = '\0';
972 loginuid = simple_strtoul(page, &tmp, 10);
973 if (tmp == page) {
974 length = -EINVAL;
975 goto out_free_page;
976
977 }
978 length = audit_set_loginuid(current, loginuid);
979 if (likely(length == 0))
980 length = count;
981
982 out_free_page:
983 free_page((unsigned long) page);
984 return length;
985 }
986
987 static struct file_operations proc_loginuid_operations = {
988 .read = proc_loginuid_read,
989 .write = proc_loginuid_write,
990 };
991 #endif
992
993 #ifdef CONFIG_SECCOMP
994 static ssize_t seccomp_read(struct file *file, char __user *buf,
995 size_t count, loff_t *ppos)
996 {
997 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
998 char __buf[20];
999 loff_t __ppos = *ppos;
1000 size_t len;
1001
1002 if (!tsk)
1003 return -ESRCH;
1004 /* no need to print the trailing zero, so use only len */
1005 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
1006 put_task_struct(tsk);
1007 if (__ppos >= len)
1008 return 0;
1009 if (count > len - __ppos)
1010 count = len - __ppos;
1011 if (copy_to_user(buf, __buf + __ppos, count))
1012 return -EFAULT;
1013 *ppos = __ppos + count;
1014 return count;
1015 }
1016
1017 static ssize_t seccomp_write(struct file *file, const char __user *buf,
1018 size_t count, loff_t *ppos)
1019 {
1020 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
1021 char __buf[20], *end;
1022 unsigned int seccomp_mode;
1023 ssize_t result;
1024
1025 result = -ESRCH;
1026 if (!tsk)
1027 goto out_no_task;
1028
1029 /* can set it only once to be even more secure */
1030 result = -EPERM;
1031 if (unlikely(tsk->seccomp.mode))
1032 goto out;
1033
1034 result = -EFAULT;
1035 memset(__buf, 0, sizeof(__buf));
1036 count = min(count, sizeof(__buf) - 1);
1037 if (copy_from_user(__buf, buf, count))
1038 goto out;
1039
1040 seccomp_mode = simple_strtoul(__buf, &end, 0);
1041 if (*end == '\n')
1042 end++;
1043 result = -EINVAL;
1044 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1045 tsk->seccomp.mode = seccomp_mode;
1046 set_tsk_thread_flag(tsk, TIF_SECCOMP);
1047 } else
1048 goto out;
1049 result = -EIO;
1050 if (unlikely(!(end - __buf)))
1051 goto out;
1052 result = end - __buf;
1053 out:
1054 put_task_struct(tsk);
1055 out_no_task:
1056 return result;
1057 }
1058
1059 static struct file_operations proc_seccomp_operations = {
1060 .read = seccomp_read,
1061 .write = seccomp_write,
1062 };
1063 #endif /* CONFIG_SECCOMP */
1064
1065 static int proc_check_dentry_visible(struct inode *inode,
1066 struct dentry *dentry, struct vfsmount *mnt)
1067 {
1068 /* Verify that the current process can already see the
1069 * file pointed at by the file descriptor.
1070 * This prevents /proc from being an accidental information leak.
1071 *
1072 * This prevents access to files that are not visible do to
1073 * being on the otherside of a chroot, in a different
1074 * namespace, or are simply process local (like pipes).
1075 */
1076 struct task_struct *task;
1077 struct files_struct *task_files, *files;
1078 int error = -EACCES;
1079
1080 /* See if the the two tasks share a commone set of
1081 * file descriptors. If so everything is visible.
1082 */
1083 task = get_proc_task(inode);
1084 if (!task)
1085 goto out;
1086 files = get_files_struct(current);
1087 task_files = get_files_struct(task);
1088 if (files && task_files && (files == task_files))
1089 error = 0;
1090 if (task_files)
1091 put_files_struct(task_files);
1092 if (files)
1093 put_files_struct(files);
1094 put_task_struct(task);
1095 if (!error)
1096 goto out;
1097
1098 /* If the two tasks don't share a common set of file
1099 * descriptors see if the destination dentry is already
1100 * visible in the current tasks filesystem namespace.
1101 */
1102 error = proc_check_chroot(dentry, mnt);
1103 out:
1104 return error;
1105
1106 }
1107
1108 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1109 {
1110 struct inode *inode = dentry->d_inode;
1111 int error = -EACCES;
1112
1113 /* We don't need a base pointer in the /proc filesystem */
1114 path_release(nd);
1115
1116 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1117 goto out;
1118
1119 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1120 nd->last_type = LAST_BIND;
1121 if (error)
1122 goto out;
1123
1124 /* Only return files this task can already see */
1125 error = proc_check_dentry_visible(inode, nd->dentry, nd->mnt);
1126 if (error)
1127 path_release(nd);
1128 out:
1129 return ERR_PTR(error);
1130 }
1131
1132 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1133 char __user *buffer, int buflen)
1134 {
1135 struct inode * inode;
1136 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1137 int len;
1138
1139 if (!tmp)
1140 return -ENOMEM;
1141
1142 inode = dentry->d_inode;
1143 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1144 len = PTR_ERR(path);
1145 if (IS_ERR(path))
1146 goto out;
1147 len = tmp + PAGE_SIZE - 1 - path;
1148
1149 if (len > buflen)
1150 len = buflen;
1151 if (copy_to_user(buffer, path, len))
1152 len = -EFAULT;
1153 out:
1154 free_page((unsigned long)tmp);
1155 return len;
1156 }
1157
1158 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1159 {
1160 int error = -EACCES;
1161 struct inode *inode = dentry->d_inode;
1162 struct dentry *de;
1163 struct vfsmount *mnt = NULL;
1164
1165
1166 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1167 goto out;
1168
1169 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1170 if (error)
1171 goto out;
1172
1173 /* Only return files this task can already see */
1174 error = proc_check_dentry_visible(inode, de, mnt);
1175 if (error)
1176 goto out_put;
1177
1178 error = do_proc_readlink(de, mnt, buffer, buflen);
1179 out_put:
1180 dput(de);
1181 mntput(mnt);
1182 out:
1183 return error;
1184 }
1185
1186 static struct inode_operations proc_pid_link_inode_operations = {
1187 .readlink = proc_pid_readlink,
1188 .follow_link = proc_pid_follow_link
1189 };
1190
1191 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1192 {
1193 struct dentry *dentry = filp->f_dentry;
1194 struct inode *inode = dentry->d_inode;
1195 struct task_struct *p = get_proc_task(inode);
1196 unsigned int fd, tid, ino;
1197 int retval;
1198 char buf[PROC_NUMBUF];
1199 struct files_struct * files;
1200 struct fdtable *fdt;
1201
1202 retval = -ENOENT;
1203 if (!p)
1204 goto out_no_task;
1205 retval = 0;
1206 tid = p->pid;
1207
1208 fd = filp->f_pos;
1209 switch (fd) {
1210 case 0:
1211 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1212 goto out;
1213 filp->f_pos++;
1214 case 1:
1215 ino = parent_ino(dentry);
1216 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1217 goto out;
1218 filp->f_pos++;
1219 default:
1220 files = get_files_struct(p);
1221 if (!files)
1222 goto out;
1223 rcu_read_lock();
1224 fdt = files_fdtable(files);
1225 for (fd = filp->f_pos-2;
1226 fd < fdt->max_fds;
1227 fd++, filp->f_pos++) {
1228 unsigned int i,j;
1229
1230 if (!fcheck_files(files, fd))
1231 continue;
1232 rcu_read_unlock();
1233
1234 j = PROC_NUMBUF;
1235 i = fd;
1236 do {
1237 j--;
1238 buf[j] = '0' + (i % 10);
1239 i /= 10;
1240 } while (i);
1241
1242 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1243 if (filldir(dirent, buf+j, PROC_NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1244 rcu_read_lock();
1245 break;
1246 }
1247 rcu_read_lock();
1248 }
1249 rcu_read_unlock();
1250 put_files_struct(files);
1251 }
1252 out:
1253 put_task_struct(p);
1254 out_no_task:
1255 return retval;
1256 }
1257
1258 static int proc_pident_readdir(struct file *filp,
1259 void *dirent, filldir_t filldir,
1260 struct pid_entry *ents, unsigned int nents)
1261 {
1262 int i;
1263 int pid;
1264 struct dentry *dentry = filp->f_dentry;
1265 struct inode *inode = dentry->d_inode;
1266 struct task_struct *task = get_proc_task(inode);
1267 struct pid_entry *p;
1268 ino_t ino;
1269 int ret;
1270
1271 ret = -ENOENT;
1272 if (!task)
1273 goto out;
1274
1275 ret = 0;
1276 pid = task->pid;
1277 put_task_struct(task);
1278 i = filp->f_pos;
1279 switch (i) {
1280 case 0:
1281 ino = inode->i_ino;
1282 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1283 goto out;
1284 i++;
1285 filp->f_pos++;
1286 /* fall through */
1287 case 1:
1288 ino = parent_ino(dentry);
1289 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1290 goto out;
1291 i++;
1292 filp->f_pos++;
1293 /* fall through */
1294 default:
1295 i -= 2;
1296 if (i >= nents) {
1297 ret = 1;
1298 goto out;
1299 }
1300 p = ents + i;
1301 while (p->name) {
1302 if (filldir(dirent, p->name, p->len, filp->f_pos,
1303 fake_ino(pid, p->type), p->mode >> 12) < 0)
1304 goto out;
1305 filp->f_pos++;
1306 p++;
1307 }
1308 }
1309
1310 ret = 1;
1311 out:
1312 return ret;
1313 }
1314
1315 static int proc_tgid_base_readdir(struct file * filp,
1316 void * dirent, filldir_t filldir)
1317 {
1318 return proc_pident_readdir(filp,dirent,filldir,
1319 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1320 }
1321
1322 static int proc_tid_base_readdir(struct file * filp,
1323 void * dirent, filldir_t filldir)
1324 {
1325 return proc_pident_readdir(filp,dirent,filldir,
1326 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1327 }
1328
1329 /* building an inode */
1330
1331 static int task_dumpable(struct task_struct *task)
1332 {
1333 int dumpable = 0;
1334 struct mm_struct *mm;
1335
1336 task_lock(task);
1337 mm = task->mm;
1338 if (mm)
1339 dumpable = mm->dumpable;
1340 task_unlock(task);
1341 if(dumpable == 1)
1342 return 1;
1343 return 0;
1344 }
1345
1346
1347 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1348 {
1349 struct inode * inode;
1350 struct proc_inode *ei;
1351
1352 /* We need a new inode */
1353
1354 inode = new_inode(sb);
1355 if (!inode)
1356 goto out;
1357
1358 /* Common stuff */
1359 ei = PROC_I(inode);
1360 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1361 inode->i_ino = fake_ino(task->pid, ino);
1362
1363 /*
1364 * grab the reference to task.
1365 */
1366 ei->tref = tref_get_by_task(task);
1367 if (!tref_task(ei->tref))
1368 goto out_unlock;
1369
1370 inode->i_uid = 0;
1371 inode->i_gid = 0;
1372 if (task_dumpable(task)) {
1373 inode->i_uid = task->euid;
1374 inode->i_gid = task->egid;
1375 }
1376 security_task_to_inode(task, inode);
1377
1378 out:
1379 return inode;
1380
1381 out_unlock:
1382 iput(inode);
1383 return NULL;
1384 }
1385
1386 /* dentry stuff */
1387
1388 /*
1389 * Exceptional case: normally we are not allowed to unhash a busy
1390 * directory. In this case, however, we can do it - no aliasing problems
1391 * due to the way we treat inodes.
1392 *
1393 * Rewrite the inode's ownerships here because the owning task may have
1394 * performed a setuid(), etc.
1395 *
1396 * Before the /proc/pid/status file was created the only way to read
1397 * the effective uid of a /process was to stat /proc/pid. Reading
1398 * /proc/pid/status is slow enough that procps and other packages
1399 * kept stating /proc/pid. To keep the rules in /proc simple I have
1400 * made this apply to all per process world readable and executable
1401 * directories.
1402 */
1403 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1404 {
1405 struct inode *inode = dentry->d_inode;
1406 struct task_struct *task = get_proc_task(inode);
1407 if (task) {
1408 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1409 task_dumpable(task)) {
1410 inode->i_uid = task->euid;
1411 inode->i_gid = task->egid;
1412 } else {
1413 inode->i_uid = 0;
1414 inode->i_gid = 0;
1415 }
1416 security_task_to_inode(task, inode);
1417 put_task_struct(task);
1418 return 1;
1419 }
1420 d_drop(dentry);
1421 return 0;
1422 }
1423
1424 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1425 {
1426 struct inode *inode = dentry->d_inode;
1427 struct task_struct *task;
1428 generic_fillattr(inode, stat);
1429
1430 rcu_read_lock();
1431 stat->uid = 0;
1432 stat->gid = 0;
1433 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1434 if (task) {
1435 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1436 task_dumpable(task)) {
1437 stat->uid = task->euid;
1438 stat->gid = task->egid;
1439 }
1440 }
1441 rcu_read_unlock();
1442 return 0;
1443 }
1444
1445 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1446 {
1447 struct inode *inode = dentry->d_inode;
1448 struct task_struct *task = get_proc_task(inode);
1449 int fd = proc_fd(inode);
1450 struct files_struct *files;
1451
1452 if (task) {
1453 files = get_files_struct(task);
1454 if (files) {
1455 rcu_read_lock();
1456 if (fcheck_files(files, fd)) {
1457 rcu_read_unlock();
1458 put_files_struct(files);
1459 if (task_dumpable(task)) {
1460 inode->i_uid = task->euid;
1461 inode->i_gid = task->egid;
1462 } else {
1463 inode->i_uid = 0;
1464 inode->i_gid = 0;
1465 }
1466 security_task_to_inode(task, inode);
1467 put_task_struct(task);
1468 return 1;
1469 }
1470 rcu_read_unlock();
1471 put_files_struct(files);
1472 }
1473 put_task_struct(task);
1474 }
1475 d_drop(dentry);
1476 return 0;
1477 }
1478
1479 static int pid_delete_dentry(struct dentry * dentry)
1480 {
1481 /* Is the task we represent dead?
1482 * If so, then don't put the dentry on the lru list,
1483 * kill it immediately.
1484 */
1485 return !proc_tref(dentry->d_inode)->task;
1486 }
1487
1488 static struct dentry_operations tid_fd_dentry_operations =
1489 {
1490 .d_revalidate = tid_fd_revalidate,
1491 .d_delete = pid_delete_dentry,
1492 };
1493
1494 static struct dentry_operations pid_dentry_operations =
1495 {
1496 .d_revalidate = pid_revalidate,
1497 .d_delete = pid_delete_dentry,
1498 };
1499
1500 /* Lookups */
1501
1502 static unsigned name_to_int(struct dentry *dentry)
1503 {
1504 const char *name = dentry->d_name.name;
1505 int len = dentry->d_name.len;
1506 unsigned n = 0;
1507
1508 if (len > 1 && *name == '0')
1509 goto out;
1510 while (len-- > 0) {
1511 unsigned c = *name++ - '0';
1512 if (c > 9)
1513 goto out;
1514 if (n >= (~0U-9)/10)
1515 goto out;
1516 n *= 10;
1517 n += c;
1518 }
1519 return n;
1520 out:
1521 return ~0U;
1522 }
1523
1524 /* SMP-safe */
1525 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1526 {
1527 struct task_struct *task = get_proc_task(dir);
1528 unsigned fd = name_to_int(dentry);
1529 struct dentry *result = ERR_PTR(-ENOENT);
1530 struct file * file;
1531 struct files_struct * files;
1532 struct inode *inode;
1533 struct proc_inode *ei;
1534
1535 if (!task)
1536 goto out_no_task;
1537 if (fd == ~0U)
1538 goto out;
1539
1540 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1541 if (!inode)
1542 goto out;
1543 ei = PROC_I(inode);
1544 ei->fd = fd;
1545 files = get_files_struct(task);
1546 if (!files)
1547 goto out_unlock;
1548 inode->i_mode = S_IFLNK;
1549
1550 /*
1551 * We are not taking a ref to the file structure, so we must
1552 * hold ->file_lock.
1553 */
1554 spin_lock(&files->file_lock);
1555 file = fcheck_files(files, fd);
1556 if (!file)
1557 goto out_unlock2;
1558 if (file->f_mode & 1)
1559 inode->i_mode |= S_IRUSR | S_IXUSR;
1560 if (file->f_mode & 2)
1561 inode->i_mode |= S_IWUSR | S_IXUSR;
1562 spin_unlock(&files->file_lock);
1563 put_files_struct(files);
1564 inode->i_op = &proc_pid_link_inode_operations;
1565 inode->i_size = 64;
1566 ei->op.proc_get_link = proc_fd_link;
1567 dentry->d_op = &tid_fd_dentry_operations;
1568 d_add(dentry, inode);
1569 /* Close the race of the process dying before we return the dentry */
1570 if (tid_fd_revalidate(dentry, NULL))
1571 result = NULL;
1572 out:
1573 put_task_struct(task);
1574 out_no_task:
1575 return result;
1576
1577 out_unlock2:
1578 spin_unlock(&files->file_lock);
1579 put_files_struct(files);
1580 out_unlock:
1581 iput(inode);
1582 goto out;
1583 }
1584
1585 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1586 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1587 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
1588
1589 static struct file_operations proc_fd_operations = {
1590 .read = generic_read_dir,
1591 .readdir = proc_readfd,
1592 };
1593
1594 static struct file_operations proc_task_operations = {
1595 .read = generic_read_dir,
1596 .readdir = proc_task_readdir,
1597 };
1598
1599 /*
1600 * proc directories can do almost nothing..
1601 */
1602 static struct inode_operations proc_fd_inode_operations = {
1603 .lookup = proc_lookupfd,
1604 };
1605
1606 static struct inode_operations proc_task_inode_operations = {
1607 .lookup = proc_task_lookup,
1608 .getattr = proc_task_getattr,
1609 };
1610
1611 #ifdef CONFIG_SECURITY
1612 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1613 size_t count, loff_t *ppos)
1614 {
1615 struct inode * inode = file->f_dentry->d_inode;
1616 unsigned long page;
1617 ssize_t length;
1618 struct task_struct *task = get_proc_task(inode);
1619
1620 length = -ESRCH;
1621 if (!task)
1622 goto out_no_task;
1623
1624 if (count > PAGE_SIZE)
1625 count = PAGE_SIZE;
1626 length = -ENOMEM;
1627 if (!(page = __get_free_page(GFP_KERNEL)))
1628 goto out;
1629
1630 length = security_getprocattr(task,
1631 (char*)file->f_dentry->d_name.name,
1632 (void*)page, count);
1633 if (length >= 0)
1634 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1635 free_page(page);
1636 out:
1637 put_task_struct(task);
1638 out_no_task:
1639 return length;
1640 }
1641
1642 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1643 size_t count, loff_t *ppos)
1644 {
1645 struct inode * inode = file->f_dentry->d_inode;
1646 char *page;
1647 ssize_t length;
1648 struct task_struct *task = get_proc_task(inode);
1649
1650 length = -ESRCH;
1651 if (!task)
1652 goto out_no_task;
1653 if (count > PAGE_SIZE)
1654 count = PAGE_SIZE;
1655
1656 /* No partial writes. */
1657 length = -EINVAL;
1658 if (*ppos != 0)
1659 goto out;
1660
1661 length = -ENOMEM;
1662 page = (char*)__get_free_page(GFP_USER);
1663 if (!page)
1664 goto out;
1665
1666 length = -EFAULT;
1667 if (copy_from_user(page, buf, count))
1668 goto out_free;
1669
1670 length = security_setprocattr(task,
1671 (char*)file->f_dentry->d_name.name,
1672 (void*)page, count);
1673 out_free:
1674 free_page((unsigned long) page);
1675 out:
1676 put_task_struct(task);
1677 out_no_task:
1678 return length;
1679 }
1680
1681 static struct file_operations proc_pid_attr_operations = {
1682 .read = proc_pid_attr_read,
1683 .write = proc_pid_attr_write,
1684 };
1685
1686 static struct file_operations proc_tid_attr_operations;
1687 static struct inode_operations proc_tid_attr_inode_operations;
1688 static struct file_operations proc_tgid_attr_operations;
1689 static struct inode_operations proc_tgid_attr_inode_operations;
1690 #endif
1691
1692 /* SMP-safe */
1693 static struct dentry *proc_pident_lookup(struct inode *dir,
1694 struct dentry *dentry,
1695 struct pid_entry *ents)
1696 {
1697 struct inode *inode;
1698 struct dentry *error;
1699 struct task_struct *task = get_proc_task(dir);
1700 struct pid_entry *p;
1701 struct proc_inode *ei;
1702
1703 error = ERR_PTR(-ENOENT);
1704 inode = NULL;
1705
1706 if (!task)
1707 goto out_no_task;
1708
1709 for (p = ents; p->name; p++) {
1710 if (p->len != dentry->d_name.len)
1711 continue;
1712 if (!memcmp(dentry->d_name.name, p->name, p->len))
1713 break;
1714 }
1715 if (!p->name)
1716 goto out;
1717
1718 error = ERR_PTR(-EINVAL);
1719 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1720 if (!inode)
1721 goto out;
1722
1723 ei = PROC_I(inode);
1724 inode->i_mode = p->mode;
1725 /*
1726 * Yes, it does not scale. And it should not. Don't add
1727 * new entries into /proc/<tgid>/ without very good reasons.
1728 */
1729 switch(p->type) {
1730 case PROC_TGID_TASK:
1731 inode->i_nlink = 2;
1732 inode->i_op = &proc_task_inode_operations;
1733 inode->i_fop = &proc_task_operations;
1734 break;
1735 case PROC_TID_FD:
1736 case PROC_TGID_FD:
1737 inode->i_nlink = 2;
1738 inode->i_op = &proc_fd_inode_operations;
1739 inode->i_fop = &proc_fd_operations;
1740 break;
1741 case PROC_TID_EXE:
1742 case PROC_TGID_EXE:
1743 inode->i_op = &proc_pid_link_inode_operations;
1744 ei->op.proc_get_link = proc_exe_link;
1745 break;
1746 case PROC_TID_CWD:
1747 case PROC_TGID_CWD:
1748 inode->i_op = &proc_pid_link_inode_operations;
1749 ei->op.proc_get_link = proc_cwd_link;
1750 break;
1751 case PROC_TID_ROOT:
1752 case PROC_TGID_ROOT:
1753 inode->i_op = &proc_pid_link_inode_operations;
1754 ei->op.proc_get_link = proc_root_link;
1755 break;
1756 case PROC_TID_ENVIRON:
1757 case PROC_TGID_ENVIRON:
1758 inode->i_fop = &proc_info_file_operations;
1759 ei->op.proc_read = proc_pid_environ;
1760 break;
1761 case PROC_TID_AUXV:
1762 case PROC_TGID_AUXV:
1763 inode->i_fop = &proc_info_file_operations;
1764 ei->op.proc_read = proc_pid_auxv;
1765 break;
1766 case PROC_TID_STATUS:
1767 case PROC_TGID_STATUS:
1768 inode->i_fop = &proc_info_file_operations;
1769 ei->op.proc_read = proc_pid_status;
1770 break;
1771 case PROC_TID_STAT:
1772 inode->i_fop = &proc_info_file_operations;
1773 ei->op.proc_read = proc_tid_stat;
1774 break;
1775 case PROC_TGID_STAT:
1776 inode->i_fop = &proc_info_file_operations;
1777 ei->op.proc_read = proc_tgid_stat;
1778 break;
1779 case PROC_TID_CMDLINE:
1780 case PROC_TGID_CMDLINE:
1781 inode->i_fop = &proc_info_file_operations;
1782 ei->op.proc_read = proc_pid_cmdline;
1783 break;
1784 case PROC_TID_STATM:
1785 case PROC_TGID_STATM:
1786 inode->i_fop = &proc_info_file_operations;
1787 ei->op.proc_read = proc_pid_statm;
1788 break;
1789 case PROC_TID_MAPS:
1790 case PROC_TGID_MAPS:
1791 inode->i_fop = &proc_maps_operations;
1792 break;
1793 #ifdef CONFIG_NUMA
1794 case PROC_TID_NUMA_MAPS:
1795 case PROC_TGID_NUMA_MAPS:
1796 inode->i_fop = &proc_numa_maps_operations;
1797 break;
1798 #endif
1799 case PROC_TID_MEM:
1800 case PROC_TGID_MEM:
1801 inode->i_fop = &proc_mem_operations;
1802 break;
1803 #ifdef CONFIG_SECCOMP
1804 case PROC_TID_SECCOMP:
1805 case PROC_TGID_SECCOMP:
1806 inode->i_fop = &proc_seccomp_operations;
1807 break;
1808 #endif /* CONFIG_SECCOMP */
1809 case PROC_TID_MOUNTS:
1810 case PROC_TGID_MOUNTS:
1811 inode->i_fop = &proc_mounts_operations;
1812 break;
1813 #ifdef CONFIG_MMU
1814 case PROC_TID_SMAPS:
1815 case PROC_TGID_SMAPS:
1816 inode->i_fop = &proc_smaps_operations;
1817 break;
1818 #endif
1819 case PROC_TID_MOUNTSTATS:
1820 case PROC_TGID_MOUNTSTATS:
1821 inode->i_fop = &proc_mountstats_operations;
1822 break;
1823 #ifdef CONFIG_SECURITY
1824 case PROC_TID_ATTR:
1825 inode->i_nlink = 2;
1826 inode->i_op = &proc_tid_attr_inode_operations;
1827 inode->i_fop = &proc_tid_attr_operations;
1828 break;
1829 case PROC_TGID_ATTR:
1830 inode->i_nlink = 2;
1831 inode->i_op = &proc_tgid_attr_inode_operations;
1832 inode->i_fop = &proc_tgid_attr_operations;
1833 break;
1834 case PROC_TID_ATTR_CURRENT:
1835 case PROC_TGID_ATTR_CURRENT:
1836 case PROC_TID_ATTR_PREV:
1837 case PROC_TGID_ATTR_PREV:
1838 case PROC_TID_ATTR_EXEC:
1839 case PROC_TGID_ATTR_EXEC:
1840 case PROC_TID_ATTR_FSCREATE:
1841 case PROC_TGID_ATTR_FSCREATE:
1842 case PROC_TID_ATTR_KEYCREATE:
1843 case PROC_TGID_ATTR_KEYCREATE:
1844 inode->i_fop = &proc_pid_attr_operations;
1845 break;
1846 #endif
1847 #ifdef CONFIG_KALLSYMS
1848 case PROC_TID_WCHAN:
1849 case PROC_TGID_WCHAN:
1850 inode->i_fop = &proc_info_file_operations;
1851 ei->op.proc_read = proc_pid_wchan;
1852 break;
1853 #endif
1854 #ifdef CONFIG_SCHEDSTATS
1855 case PROC_TID_SCHEDSTAT:
1856 case PROC_TGID_SCHEDSTAT:
1857 inode->i_fop = &proc_info_file_operations;
1858 ei->op.proc_read = proc_pid_schedstat;
1859 break;
1860 #endif
1861 #ifdef CONFIG_CPUSETS
1862 case PROC_TID_CPUSET:
1863 case PROC_TGID_CPUSET:
1864 inode->i_fop = &proc_cpuset_operations;
1865 break;
1866 #endif
1867 case PROC_TID_OOM_SCORE:
1868 case PROC_TGID_OOM_SCORE:
1869 inode->i_fop = &proc_info_file_operations;
1870 ei->op.proc_read = proc_oom_score;
1871 break;
1872 case PROC_TID_OOM_ADJUST:
1873 case PROC_TGID_OOM_ADJUST:
1874 inode->i_fop = &proc_oom_adjust_operations;
1875 break;
1876 #ifdef CONFIG_AUDITSYSCALL
1877 case PROC_TID_LOGINUID:
1878 case PROC_TGID_LOGINUID:
1879 inode->i_fop = &proc_loginuid_operations;
1880 break;
1881 #endif
1882 default:
1883 printk("procfs: impossible type (%d)",p->type);
1884 iput(inode);
1885 error = ERR_PTR(-EINVAL);
1886 goto out;
1887 }
1888 dentry->d_op = &pid_dentry_operations;
1889 d_add(dentry, inode);
1890 /* Close the race of the process dying before we return the dentry */
1891 if (pid_revalidate(dentry, NULL))
1892 error = NULL;
1893 out:
1894 put_task_struct(task);
1895 out_no_task:
1896 return error;
1897 }
1898
1899 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1900 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1901 }
1902
1903 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1904 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1905 }
1906
1907 static struct file_operations proc_tgid_base_operations = {
1908 .read = generic_read_dir,
1909 .readdir = proc_tgid_base_readdir,
1910 };
1911
1912 static struct file_operations proc_tid_base_operations = {
1913 .read = generic_read_dir,
1914 .readdir = proc_tid_base_readdir,
1915 };
1916
1917 static struct inode_operations proc_tgid_base_inode_operations = {
1918 .lookup = proc_tgid_base_lookup,
1919 .getattr = pid_getattr,
1920 };
1921
1922 static struct inode_operations proc_tid_base_inode_operations = {
1923 .lookup = proc_tid_base_lookup,
1924 .getattr = pid_getattr,
1925 };
1926
1927 #ifdef CONFIG_SECURITY
1928 static int proc_tgid_attr_readdir(struct file * filp,
1929 void * dirent, filldir_t filldir)
1930 {
1931 return proc_pident_readdir(filp,dirent,filldir,
1932 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1933 }
1934
1935 static int proc_tid_attr_readdir(struct file * filp,
1936 void * dirent, filldir_t filldir)
1937 {
1938 return proc_pident_readdir(filp,dirent,filldir,
1939 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1940 }
1941
1942 static struct file_operations proc_tgid_attr_operations = {
1943 .read = generic_read_dir,
1944 .readdir = proc_tgid_attr_readdir,
1945 };
1946
1947 static struct file_operations proc_tid_attr_operations = {
1948 .read = generic_read_dir,
1949 .readdir = proc_tid_attr_readdir,
1950 };
1951
1952 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1953 struct dentry *dentry, struct nameidata *nd)
1954 {
1955 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1956 }
1957
1958 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1959 struct dentry *dentry, struct nameidata *nd)
1960 {
1961 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1962 }
1963
1964 static struct inode_operations proc_tgid_attr_inode_operations = {
1965 .lookup = proc_tgid_attr_lookup,
1966 .getattr = pid_getattr,
1967 };
1968
1969 static struct inode_operations proc_tid_attr_inode_operations = {
1970 .lookup = proc_tid_attr_lookup,
1971 .getattr = pid_getattr,
1972 };
1973 #endif
1974
1975 /*
1976 * /proc/self:
1977 */
1978 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1979 int buflen)
1980 {
1981 char tmp[PROC_NUMBUF];
1982 sprintf(tmp, "%d", current->tgid);
1983 return vfs_readlink(dentry,buffer,buflen,tmp);
1984 }
1985
1986 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1987 {
1988 char tmp[PROC_NUMBUF];
1989 sprintf(tmp, "%d", current->tgid);
1990 return ERR_PTR(vfs_follow_link(nd,tmp));
1991 }
1992
1993 static struct inode_operations proc_self_inode_operations = {
1994 .readlink = proc_self_readlink,
1995 .follow_link = proc_self_follow_link,
1996 };
1997
1998 /**
1999 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2000 *
2001 * @task: task that should be flushed.
2002 *
2003 * Looks in the dcache for
2004 * /proc/@pid
2005 * /proc/@tgid/task/@pid
2006 * if either directory is present flushes it and all of it'ts children
2007 * from the dcache.
2008 *
2009 * It is safe and reasonable to cache /proc entries for a task until
2010 * that task exits. After that they just clog up the dcache with
2011 * useless entries, possibly causing useful dcache entries to be
2012 * flushed instead. This routine is proved to flush those useless
2013 * dcache entries at process exit time.
2014 *
2015 * NOTE: This routine is just an optimization so it does not guarantee
2016 * that no dcache entries will exist at process exit time it
2017 * just makes it very unlikely that any will persist.
2018 */
2019 void proc_flush_task(struct task_struct *task)
2020 {
2021 struct dentry *dentry, *leader, *dir;
2022 char buf[PROC_NUMBUF];
2023 struct qstr name;
2024
2025 name.name = buf;
2026 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2027 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2028 if (dentry) {
2029 shrink_dcache_parent(dentry);
2030 d_drop(dentry);
2031 dput(dentry);
2032 }
2033
2034 if (thread_group_leader(task))
2035 goto out;
2036
2037 name.name = buf;
2038 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2039 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2040 if (!leader)
2041 goto out;
2042
2043 name.name = "task";
2044 name.len = strlen(name.name);
2045 dir = d_hash_and_lookup(leader, &name);
2046 if (!dir)
2047 goto out_put_leader;
2048
2049 name.name = buf;
2050 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2051 dentry = d_hash_and_lookup(dir, &name);
2052 if (dentry) {
2053 shrink_dcache_parent(dentry);
2054 d_drop(dentry);
2055 dput(dentry);
2056 }
2057
2058 dput(dir);
2059 out_put_leader:
2060 dput(leader);
2061 out:
2062 return;
2063 }
2064
2065 /* SMP-safe */
2066 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2067 {
2068 struct dentry *result = ERR_PTR(-ENOENT);
2069 struct task_struct *task;
2070 struct inode *inode;
2071 struct proc_inode *ei;
2072 unsigned tgid;
2073
2074 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
2075 inode = new_inode(dir->i_sb);
2076 if (!inode)
2077 return ERR_PTR(-ENOMEM);
2078 ei = PROC_I(inode);
2079 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2080 inode->i_ino = fake_ino(0, PROC_TGID_INO);
2081 ei->pde = NULL;
2082 inode->i_mode = S_IFLNK|S_IRWXUGO;
2083 inode->i_uid = inode->i_gid = 0;
2084 inode->i_size = 64;
2085 inode->i_op = &proc_self_inode_operations;
2086 d_add(dentry, inode);
2087 return NULL;
2088 }
2089 tgid = name_to_int(dentry);
2090 if (tgid == ~0U)
2091 goto out;
2092
2093 rcu_read_lock();
2094 task = find_task_by_pid(tgid);
2095 if (task)
2096 get_task_struct(task);
2097 rcu_read_unlock();
2098 if (!task)
2099 goto out;
2100
2101 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
2102 if (!inode)
2103 goto out_put_task;
2104
2105 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2106 inode->i_op = &proc_tgid_base_inode_operations;
2107 inode->i_fop = &proc_tgid_base_operations;
2108 inode->i_flags|=S_IMMUTABLE;
2109 #ifdef CONFIG_SECURITY
2110 inode->i_nlink = 5;
2111 #else
2112 inode->i_nlink = 4;
2113 #endif
2114
2115 dentry->d_op = &pid_dentry_operations;
2116
2117 d_add(dentry, inode);
2118 /* Close the race of the process dying before we return the dentry */
2119 if (pid_revalidate(dentry, NULL))
2120 result = NULL;
2121
2122 out_put_task:
2123 put_task_struct(task);
2124 out:
2125 return result;
2126 }
2127
2128 /* SMP-safe */
2129 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2130 {
2131 struct dentry *result = ERR_PTR(-ENOENT);
2132 struct task_struct *task;
2133 struct task_struct *leader = get_proc_task(dir);
2134 struct inode *inode;
2135 unsigned tid;
2136
2137 if (!leader)
2138 goto out_no_task;
2139
2140 tid = name_to_int(dentry);
2141 if (tid == ~0U)
2142 goto out;
2143
2144 rcu_read_lock();
2145 task = find_task_by_pid(tid);
2146 if (task)
2147 get_task_struct(task);
2148 rcu_read_unlock();
2149 if (!task)
2150 goto out;
2151 if (leader->tgid != task->tgid)
2152 goto out_drop_task;
2153
2154 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2155
2156
2157 if (!inode)
2158 goto out_drop_task;
2159 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2160 inode->i_op = &proc_tid_base_inode_operations;
2161 inode->i_fop = &proc_tid_base_operations;
2162 inode->i_flags|=S_IMMUTABLE;
2163 #ifdef CONFIG_SECURITY
2164 inode->i_nlink = 4;
2165 #else
2166 inode->i_nlink = 3;
2167 #endif
2168
2169 dentry->d_op = &pid_dentry_operations;
2170
2171 d_add(dentry, inode);
2172 /* Close the race of the process dying before we return the dentry */
2173 if (pid_revalidate(dentry, NULL))
2174 result = NULL;
2175
2176 out_drop_task:
2177 put_task_struct(task);
2178 out:
2179 put_task_struct(leader);
2180 out_no_task:
2181 return result;
2182 }
2183
2184 /*
2185 * Find the first tgid to return to user space.
2186 *
2187 * Usually this is just whatever follows &init_task, but if the users
2188 * buffer was too small to hold the full list or there was a seek into
2189 * the middle of the directory we have more work to do.
2190 *
2191 * In the case of a short read we start with find_task_by_pid.
2192 *
2193 * In the case of a seek we start with &init_task and walk nr
2194 * threads past it.
2195 */
2196 static struct task_struct *first_tgid(int tgid, unsigned int nr)
2197 {
2198 struct task_struct *pos;
2199 rcu_read_lock();
2200 if (tgid && nr) {
2201 pos = find_task_by_pid(tgid);
2202 if (pos && thread_group_leader(pos))
2203 goto found;
2204 }
2205 /* If nr exceeds the number of processes get out quickly */
2206 pos = NULL;
2207 if (nr && nr >= nr_processes())
2208 goto done;
2209
2210 /* If we haven't found our starting place yet start with
2211 * the init_task and walk nr tasks forward.
2212 */
2213 for (pos = next_task(&init_task); nr > 0; --nr) {
2214 pos = next_task(pos);
2215 if (pos == &init_task) {
2216 pos = NULL;
2217 goto done;
2218 }
2219 }
2220 found:
2221 get_task_struct(pos);
2222 done:
2223 rcu_read_unlock();
2224 return pos;
2225 }
2226
2227 /*
2228 * Find the next task in the task list.
2229 * Return NULL if we loop or there is any error.
2230 *
2231 * The reference to the input task_struct is released.
2232 */
2233 static struct task_struct *next_tgid(struct task_struct *start)
2234 {
2235 struct task_struct *pos;
2236 rcu_read_lock();
2237 pos = start;
2238 if (pid_alive(start))
2239 pos = next_task(start);
2240 if (pid_alive(pos) && (pos != &init_task)) {
2241 get_task_struct(pos);
2242 goto done;
2243 }
2244 pos = NULL;
2245 done:
2246 rcu_read_unlock();
2247 put_task_struct(start);
2248 return pos;
2249 }
2250
2251 /* for the /proc/ directory itself, after non-process stuff has been done */
2252 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2253 {
2254 char buf[PROC_NUMBUF];
2255 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2256 struct task_struct *task;
2257 int tgid;
2258
2259 if (!nr) {
2260 ino_t ino = fake_ino(0,PROC_TGID_INO);
2261 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2262 return 0;
2263 filp->f_pos++;
2264 nr++;
2265 }
2266 nr -= 1;
2267
2268 /* f_version caches the tgid value that the last readdir call couldn't
2269 * return. lseek aka telldir automagically resets f_version to 0.
2270 */
2271 tgid = filp->f_version;
2272 filp->f_version = 0;
2273 for (task = first_tgid(tgid, nr);
2274 task;
2275 task = next_tgid(task), filp->f_pos++) {
2276 int len;
2277 ino_t ino;
2278 tgid = task->pid;
2279 len = snprintf(buf, sizeof(buf), "%d", tgid);
2280 ino = fake_ino(tgid, PROC_TGID_INO);
2281 if (filldir(dirent, buf, len, filp->f_pos, ino, DT_DIR) < 0) {
2282 /* returning this tgid failed, save it as the first
2283 * pid for the next readir call */
2284 filp->f_version = tgid;
2285 put_task_struct(task);
2286 break;
2287 }
2288 }
2289 return 0;
2290 }
2291
2292 /*
2293 * Find the first tid of a thread group to return to user space.
2294 *
2295 * Usually this is just the thread group leader, but if the users
2296 * buffer was too small or there was a seek into the middle of the
2297 * directory we have more work todo.
2298 *
2299 * In the case of a short read we start with find_task_by_pid.
2300 *
2301 * In the case of a seek we start with the leader and walk nr
2302 * threads past it.
2303 */
2304 static struct task_struct *first_tid(struct task_struct *leader, int tid, int nr)
2305 {
2306 struct task_struct *pos = NULL;
2307 read_lock(&tasklist_lock);
2308
2309 /* Attempt to start with the pid of a thread */
2310 if (tid && (nr > 0)) {
2311 pos = find_task_by_pid(tid);
2312 if (pos && (pos->group_leader != leader))
2313 pos = NULL;
2314 if (pos)
2315 nr = 0;
2316 }
2317
2318 /* If nr exceeds the number of threads there is nothing todo */
2319 if (nr) {
2320 if (nr >= get_nr_threads(leader))
2321 goto done;
2322 }
2323
2324 /* If we haven't found our starting place yet start with the
2325 * leader and walk nr threads forward.
2326 */
2327 if (!pos && (nr >= 0))
2328 pos = leader;
2329
2330 for (; pos && pid_alive(pos); pos = next_thread(pos)) {
2331 if (--nr > 0)
2332 continue;
2333 get_task_struct(pos);
2334 goto done;
2335 }
2336 pos = NULL;
2337 done:
2338 read_unlock(&tasklist_lock);
2339 return pos;
2340 }
2341
2342 /*
2343 * Find the next thread in the thread list.
2344 * Return NULL if there is an error or no next thread.
2345 *
2346 * The reference to the input task_struct is released.
2347 */
2348 static struct task_struct *next_tid(struct task_struct *start)
2349 {
2350 struct task_struct *pos;
2351 read_lock(&tasklist_lock);
2352 pos = start;
2353 if (pid_alive(start))
2354 pos = next_thread(start);
2355 if (pid_alive(pos) && (pos != start->group_leader))
2356 get_task_struct(pos);
2357 else
2358 pos = NULL;
2359 read_unlock(&tasklist_lock);
2360 put_task_struct(start);
2361 return pos;
2362 }
2363
2364 /* for the /proc/TGID/task/ directories */
2365 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2366 {
2367 char buf[PROC_NUMBUF];
2368 struct dentry *dentry = filp->f_dentry;
2369 struct inode *inode = dentry->d_inode;
2370 struct task_struct *leader = get_proc_task(inode);
2371 struct task_struct *task;
2372 int retval = -ENOENT;
2373 ino_t ino;
2374 int tid;
2375 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2376
2377 if (!leader)
2378 goto out_no_task;
2379 retval = 0;
2380
2381 switch (pos) {
2382 case 0:
2383 ino = inode->i_ino;
2384 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2385 goto out;
2386 pos++;
2387 /* fall through */
2388 case 1:
2389 ino = parent_ino(dentry);
2390 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2391 goto out;
2392 pos++;
2393 /* fall through */
2394 }
2395
2396 /* f_version caches the tgid value that the last readdir call couldn't
2397 * return. lseek aka telldir automagically resets f_version to 0.
2398 */
2399 tid = filp->f_version;
2400 filp->f_version = 0;
2401 for (task = first_tid(leader, tid, pos - 2);
2402 task;
2403 task = next_tid(task), pos++) {
2404 int len;
2405 tid = task->pid;
2406 len = snprintf(buf, sizeof(buf), "%d", tid);
2407 ino = fake_ino(tid, PROC_TID_INO);
2408 if (filldir(dirent, buf, len, pos, ino, DT_DIR < 0)) {
2409 /* returning this tgid failed, save it as the first
2410 * pid for the next readir call */
2411 filp->f_version = tid;
2412 put_task_struct(task);
2413 break;
2414 }
2415 }
2416 out:
2417 filp->f_pos = pos;
2418 put_task_struct(leader);
2419 out_no_task:
2420 return retval;
2421 }
2422
2423 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2424 {
2425 struct inode *inode = dentry->d_inode;
2426 struct task_struct *p = get_proc_task(inode);
2427 generic_fillattr(inode, stat);
2428
2429 if (p) {
2430 rcu_read_lock();
2431 stat->nlink += get_nr_threads(p);
2432 rcu_read_unlock();
2433 put_task_struct(p);
2434 }
2435
2436 return 0;
2437 }
Linux kernel - Deliverables
This page took 0.231852 seconds and 5 git commands to generate.