xref: /linux/fs/proc/base.c (revision 6e8331ac6973435b1e7604c30f2ad394035b46e1)
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/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/namespace.h>
63 #include <linux/mm.h>
64 #include <linux/smp_lock.h>
65 #include <linux/rcupdate.h>
66 #include <linux/kallsyms.h>
67 #include <linux/mount.h>
68 #include <linux/security.h>
69 #include <linux/ptrace.h>
70 #include <linux/seccomp.h>
71 #include <linux/cpuset.h>
72 #include <linux/audit.h>
73 #include <linux/poll.h>
74 #include "internal.h"
75 
76 /* NOTE:
77  *	Implementing inode permission operations in /proc is almost
78  *	certainly an error.  Permission checks need to happen during
79  *	each system call not at open time.  The reason is that most of
80  *	what we wish to check for permissions in /proc varies at runtime.
81  *
82  *	The classic example of a problem is opening file descriptors
83  *	in /proc for a task before it execs a suid executable.
84  */
85 
86 /*
87  * For hysterical raisins we keep the same inumbers as in the old procfs.
88  * Feel free to change the macro below - just keep the range distinct from
89  * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
90  * As soon as we'll get a separate superblock we will be able to forget
91  * about magical ranges too.
92  */
93 
94 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
95 
96 enum pid_directory_inos {
97 	PROC_TGID_INO = 2,
98 	PROC_TGID_TASK,
99 	PROC_TGID_STATUS,
100 	PROC_TGID_MEM,
101 #ifdef CONFIG_SECCOMP
102 	PROC_TGID_SECCOMP,
103 #endif
104 	PROC_TGID_CWD,
105 	PROC_TGID_ROOT,
106 	PROC_TGID_EXE,
107 	PROC_TGID_FD,
108 	PROC_TGID_ENVIRON,
109 	PROC_TGID_AUXV,
110 	PROC_TGID_CMDLINE,
111 	PROC_TGID_STAT,
112 	PROC_TGID_STATM,
113 	PROC_TGID_MAPS,
114 	PROC_TGID_NUMA_MAPS,
115 	PROC_TGID_MOUNTS,
116 	PROC_TGID_MOUNTSTATS,
117 	PROC_TGID_WCHAN,
118 #ifdef CONFIG_MMU
119 	PROC_TGID_SMAPS,
120 #endif
121 #ifdef CONFIG_SCHEDSTATS
122 	PROC_TGID_SCHEDSTAT,
123 #endif
124 #ifdef CONFIG_CPUSETS
125 	PROC_TGID_CPUSET,
126 #endif
127 #ifdef CONFIG_SECURITY
128 	PROC_TGID_ATTR,
129 	PROC_TGID_ATTR_CURRENT,
130 	PROC_TGID_ATTR_PREV,
131 	PROC_TGID_ATTR_EXEC,
132 	PROC_TGID_ATTR_FSCREATE,
133 	PROC_TGID_ATTR_KEYCREATE,
134 	PROC_TGID_ATTR_SOCKCREATE,
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 	PROC_TID_ATTR_SOCKCREATE,
178 #endif
179 #ifdef CONFIG_AUDITSYSCALL
180 	PROC_TID_LOGINUID,
181 #endif
182 	PROC_TID_OOM_SCORE,
183 	PROC_TID_OOM_ADJUST,
184 
185 	/* Add new entries before this */
186 	PROC_TID_FD_DIR = 0x8000,	/* 0x8000-0xffff */
187 };
188 
189 /* Worst case buffer size needed for holding an integer. */
190 #define PROC_NUMBUF 10
191 
192 struct pid_entry {
193 	int type;
194 	int len;
195 	char *name;
196 	mode_t mode;
197 };
198 
199 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
200 
201 static struct pid_entry tgid_base_stuff[] = {
202 	E(PROC_TGID_TASK,      "task",    S_IFDIR|S_IRUGO|S_IXUGO),
203 	E(PROC_TGID_FD,        "fd",      S_IFDIR|S_IRUSR|S_IXUSR),
204 	E(PROC_TGID_ENVIRON,   "environ", S_IFREG|S_IRUSR),
205 	E(PROC_TGID_AUXV,      "auxv",	  S_IFREG|S_IRUSR),
206 	E(PROC_TGID_STATUS,    "status",  S_IFREG|S_IRUGO),
207 	E(PROC_TGID_CMDLINE,   "cmdline", S_IFREG|S_IRUGO),
208 	E(PROC_TGID_STAT,      "stat",    S_IFREG|S_IRUGO),
209 	E(PROC_TGID_STATM,     "statm",   S_IFREG|S_IRUGO),
210 	E(PROC_TGID_MAPS,      "maps",    S_IFREG|S_IRUGO),
211 #ifdef CONFIG_NUMA
212 	E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
213 #endif
214 	E(PROC_TGID_MEM,       "mem",     S_IFREG|S_IRUSR|S_IWUSR),
215 #ifdef CONFIG_SECCOMP
216 	E(PROC_TGID_SECCOMP,   "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
217 #endif
218 	E(PROC_TGID_CWD,       "cwd",     S_IFLNK|S_IRWXUGO),
219 	E(PROC_TGID_ROOT,      "root",    S_IFLNK|S_IRWXUGO),
220 	E(PROC_TGID_EXE,       "exe",     S_IFLNK|S_IRWXUGO),
221 	E(PROC_TGID_MOUNTS,    "mounts",  S_IFREG|S_IRUGO),
222 	E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
223 #ifdef CONFIG_MMU
224 	E(PROC_TGID_SMAPS,     "smaps",   S_IFREG|S_IRUGO),
225 #endif
226 #ifdef CONFIG_SECURITY
227 	E(PROC_TGID_ATTR,      "attr",    S_IFDIR|S_IRUGO|S_IXUGO),
228 #endif
229 #ifdef CONFIG_KALLSYMS
230 	E(PROC_TGID_WCHAN,     "wchan",   S_IFREG|S_IRUGO),
231 #endif
232 #ifdef CONFIG_SCHEDSTATS
233 	E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
234 #endif
235 #ifdef CONFIG_CPUSETS
236 	E(PROC_TGID_CPUSET,    "cpuset",  S_IFREG|S_IRUGO),
237 #endif
238 	E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
239 	E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
240 #ifdef CONFIG_AUDITSYSCALL
241 	E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
242 #endif
243 	{0,0,NULL,0}
244 };
245 static struct pid_entry tid_base_stuff[] = {
246 	E(PROC_TID_FD,         "fd",      S_IFDIR|S_IRUSR|S_IXUSR),
247 	E(PROC_TID_ENVIRON,    "environ", S_IFREG|S_IRUSR),
248 	E(PROC_TID_AUXV,       "auxv",	  S_IFREG|S_IRUSR),
249 	E(PROC_TID_STATUS,     "status",  S_IFREG|S_IRUGO),
250 	E(PROC_TID_CMDLINE,    "cmdline", S_IFREG|S_IRUGO),
251 	E(PROC_TID_STAT,       "stat",    S_IFREG|S_IRUGO),
252 	E(PROC_TID_STATM,      "statm",   S_IFREG|S_IRUGO),
253 	E(PROC_TID_MAPS,       "maps",    S_IFREG|S_IRUGO),
254 #ifdef CONFIG_NUMA
255 	E(PROC_TID_NUMA_MAPS,  "numa_maps",    S_IFREG|S_IRUGO),
256 #endif
257 	E(PROC_TID_MEM,        "mem",     S_IFREG|S_IRUSR|S_IWUSR),
258 #ifdef CONFIG_SECCOMP
259 	E(PROC_TID_SECCOMP,    "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
260 #endif
261 	E(PROC_TID_CWD,        "cwd",     S_IFLNK|S_IRWXUGO),
262 	E(PROC_TID_ROOT,       "root",    S_IFLNK|S_IRWXUGO),
263 	E(PROC_TID_EXE,        "exe",     S_IFLNK|S_IRWXUGO),
264 	E(PROC_TID_MOUNTS,     "mounts",  S_IFREG|S_IRUGO),
265 #ifdef CONFIG_MMU
266 	E(PROC_TID_SMAPS,      "smaps",   S_IFREG|S_IRUGO),
267 #endif
268 #ifdef CONFIG_SECURITY
269 	E(PROC_TID_ATTR,       "attr",    S_IFDIR|S_IRUGO|S_IXUGO),
270 #endif
271 #ifdef CONFIG_KALLSYMS
272 	E(PROC_TID_WCHAN,      "wchan",   S_IFREG|S_IRUGO),
273 #endif
274 #ifdef CONFIG_SCHEDSTATS
275 	E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
276 #endif
277 #ifdef CONFIG_CPUSETS
278 	E(PROC_TID_CPUSET,     "cpuset",  S_IFREG|S_IRUGO),
279 #endif
280 	E(PROC_TID_OOM_SCORE,  "oom_score",S_IFREG|S_IRUGO),
281 	E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
282 #ifdef CONFIG_AUDITSYSCALL
283 	E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
284 #endif
285 	{0,0,NULL,0}
286 };
287 
288 #ifdef CONFIG_SECURITY
289 static struct pid_entry tgid_attr_stuff[] = {
290 	E(PROC_TGID_ATTR_CURRENT,  "current",  S_IFREG|S_IRUGO|S_IWUGO),
291 	E(PROC_TGID_ATTR_PREV,     "prev",     S_IFREG|S_IRUGO),
292 	E(PROC_TGID_ATTR_EXEC,     "exec",     S_IFREG|S_IRUGO|S_IWUGO),
293 	E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
294 	E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
295 	E(PROC_TGID_ATTR_SOCKCREATE, "sockcreate", S_IFREG|S_IRUGO|S_IWUGO),
296 	{0,0,NULL,0}
297 };
298 static struct pid_entry tid_attr_stuff[] = {
299 	E(PROC_TID_ATTR_CURRENT,   "current",  S_IFREG|S_IRUGO|S_IWUGO),
300 	E(PROC_TID_ATTR_PREV,      "prev",     S_IFREG|S_IRUGO),
301 	E(PROC_TID_ATTR_EXEC,      "exec",     S_IFREG|S_IRUGO|S_IWUGO),
302 	E(PROC_TID_ATTR_FSCREATE,  "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
303 	E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
304 	E(PROC_TID_ATTR_SOCKCREATE, "sockcreate", S_IFREG|S_IRUGO|S_IWUGO),
305 	{0,0,NULL,0}
306 };
307 #endif
308 
309 #undef E
310 
311 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
312 {
313 	struct task_struct *task = get_proc_task(inode);
314 	struct files_struct *files = NULL;
315 	struct file *file;
316 	int fd = proc_fd(inode);
317 
318 	if (task) {
319 		files = get_files_struct(task);
320 		put_task_struct(task);
321 	}
322 	if (files) {
323 		/*
324 		 * We are not taking a ref to the file structure, so we must
325 		 * hold ->file_lock.
326 		 */
327 		spin_lock(&files->file_lock);
328 		file = fcheck_files(files, fd);
329 		if (file) {
330 			*mnt = mntget(file->f_vfsmnt);
331 			*dentry = dget(file->f_dentry);
332 			spin_unlock(&files->file_lock);
333 			put_files_struct(files);
334 			return 0;
335 		}
336 		spin_unlock(&files->file_lock);
337 		put_files_struct(files);
338 	}
339 	return -ENOENT;
340 }
341 
342 static struct fs_struct *get_fs_struct(struct task_struct *task)
343 {
344 	struct fs_struct *fs;
345 	task_lock(task);
346 	fs = task->fs;
347 	if(fs)
348 		atomic_inc(&fs->count);
349 	task_unlock(task);
350 	return fs;
351 }
352 
353 static int get_nr_threads(struct task_struct *tsk)
354 {
355 	/* Must be called with the rcu_read_lock held */
356 	unsigned long flags;
357 	int count = 0;
358 
359 	if (lock_task_sighand(tsk, &flags)) {
360 		count = atomic_read(&tsk->signal->count);
361 		unlock_task_sighand(tsk, &flags);
362 	}
363 	return count;
364 }
365 
366 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
367 {
368 	struct task_struct *task = get_proc_task(inode);
369 	struct fs_struct *fs = NULL;
370 	int result = -ENOENT;
371 
372 	if (task) {
373 		fs = get_fs_struct(task);
374 		put_task_struct(task);
375 	}
376 	if (fs) {
377 		read_lock(&fs->lock);
378 		*mnt = mntget(fs->pwdmnt);
379 		*dentry = dget(fs->pwd);
380 		read_unlock(&fs->lock);
381 		result = 0;
382 		put_fs_struct(fs);
383 	}
384 	return result;
385 }
386 
387 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
388 {
389 	struct task_struct *task = get_proc_task(inode);
390 	struct fs_struct *fs = NULL;
391 	int result = -ENOENT;
392 
393 	if (task) {
394 		fs = get_fs_struct(task);
395 		put_task_struct(task);
396 	}
397 	if (fs) {
398 		read_lock(&fs->lock);
399 		*mnt = mntget(fs->rootmnt);
400 		*dentry = dget(fs->root);
401 		read_unlock(&fs->lock);
402 		result = 0;
403 		put_fs_struct(fs);
404 	}
405 	return result;
406 }
407 
408 #define MAY_PTRACE(task) \
409 	(task == current || \
410 	(task->parent == current && \
411 	(task->ptrace & PT_PTRACED) && \
412 	 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
413 	 security_ptrace(current,task) == 0))
414 
415 static int proc_pid_environ(struct task_struct *task, char * buffer)
416 {
417 	int res = 0;
418 	struct mm_struct *mm = get_task_mm(task);
419 	if (mm) {
420 		unsigned int len = mm->env_end - mm->env_start;
421 		if (len > PAGE_SIZE)
422 			len = PAGE_SIZE;
423 		res = access_process_vm(task, mm->env_start, buffer, len, 0);
424 		if (!ptrace_may_attach(task))
425 			res = -ESRCH;
426 		mmput(mm);
427 	}
428 	return res;
429 }
430 
431 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
432 {
433 	int res = 0;
434 	unsigned int len;
435 	struct mm_struct *mm = get_task_mm(task);
436 	if (!mm)
437 		goto out;
438 	if (!mm->arg_end)
439 		goto out_mm;	/* Shh! No looking before we're done */
440 
441  	len = mm->arg_end - mm->arg_start;
442 
443 	if (len > PAGE_SIZE)
444 		len = PAGE_SIZE;
445 
446 	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
447 
448 	// If the nul at the end of args has been overwritten, then
449 	// assume application is using setproctitle(3).
450 	if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
451 		len = strnlen(buffer, res);
452 		if (len < res) {
453 		    res = len;
454 		} else {
455 			len = mm->env_end - mm->env_start;
456 			if (len > PAGE_SIZE - res)
457 				len = PAGE_SIZE - res;
458 			res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
459 			res = strnlen(buffer, res);
460 		}
461 	}
462 out_mm:
463 	mmput(mm);
464 out:
465 	return res;
466 }
467 
468 static int proc_pid_auxv(struct task_struct *task, char *buffer)
469 {
470 	int res = 0;
471 	struct mm_struct *mm = get_task_mm(task);
472 	if (mm) {
473 		unsigned int nwords = 0;
474 		do
475 			nwords += 2;
476 		while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
477 		res = nwords * sizeof(mm->saved_auxv[0]);
478 		if (res > PAGE_SIZE)
479 			res = PAGE_SIZE;
480 		memcpy(buffer, mm->saved_auxv, res);
481 		mmput(mm);
482 	}
483 	return res;
484 }
485 
486 
487 #ifdef CONFIG_KALLSYMS
488 /*
489  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
490  * Returns the resolved symbol.  If that fails, simply return the address.
491  */
492 static int proc_pid_wchan(struct task_struct *task, char *buffer)
493 {
494 	char *modname;
495 	const char *sym_name;
496 	unsigned long wchan, size, offset;
497 	char namebuf[KSYM_NAME_LEN+1];
498 
499 	wchan = get_wchan(task);
500 
501 	sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
502 	if (sym_name)
503 		return sprintf(buffer, "%s", sym_name);
504 	return sprintf(buffer, "%lu", wchan);
505 }
506 #endif /* CONFIG_KALLSYMS */
507 
508 #ifdef CONFIG_SCHEDSTATS
509 /*
510  * Provides /proc/PID/schedstat
511  */
512 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
513 {
514 	return sprintf(buffer, "%lu %lu %lu\n",
515 			task->sched_info.cpu_time,
516 			task->sched_info.run_delay,
517 			task->sched_info.pcnt);
518 }
519 #endif
520 
521 /* The badness from the OOM killer */
522 unsigned long badness(struct task_struct *p, unsigned long uptime);
523 static int proc_oom_score(struct task_struct *task, char *buffer)
524 {
525 	unsigned long points;
526 	struct timespec uptime;
527 
528 	do_posix_clock_monotonic_gettime(&uptime);
529 	points = badness(task, uptime.tv_sec);
530 	return sprintf(buffer, "%lu\n", points);
531 }
532 
533 /************************************************************************/
534 /*                       Here the fs part begins                        */
535 /************************************************************************/
536 
537 /* permission checks */
538 static int proc_fd_access_allowed(struct inode *inode)
539 {
540 	struct task_struct *task;
541 	int allowed = 0;
542 	/* Allow access to a task's file descriptors if it is us or we
543 	 * may use ptrace attach to the process and find out that
544 	 * information.
545 	 */
546 	task = get_proc_task(inode);
547 	if (task) {
548 		allowed = ptrace_may_attach(task);
549 		put_task_struct(task);
550 	}
551 	return allowed;
552 }
553 
554 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
555 {
556 	int error;
557 	struct inode *inode = dentry->d_inode;
558 
559 	if (attr->ia_valid & ATTR_MODE)
560 		return -EPERM;
561 
562 	error = inode_change_ok(inode, attr);
563 	if (!error) {
564 		error = security_inode_setattr(dentry, attr);
565 		if (!error)
566 			error = inode_setattr(inode, attr);
567 	}
568 	return error;
569 }
570 
571 static struct inode_operations proc_def_inode_operations = {
572 	.setattr	= proc_setattr,
573 };
574 
575 extern struct seq_operations mounts_op;
576 struct proc_mounts {
577 	struct seq_file m;
578 	int event;
579 };
580 
581 static int mounts_open(struct inode *inode, struct file *file)
582 {
583 	struct task_struct *task = get_proc_task(inode);
584 	struct namespace *namespace = NULL;
585 	struct proc_mounts *p;
586 	int ret = -EINVAL;
587 
588 	if (task) {
589 		task_lock(task);
590 		namespace = task->namespace;
591 		if (namespace)
592 			get_namespace(namespace);
593 		task_unlock(task);
594 		put_task_struct(task);
595 	}
596 
597 	if (namespace) {
598 		ret = -ENOMEM;
599 		p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
600 		if (p) {
601 			file->private_data = &p->m;
602 			ret = seq_open(file, &mounts_op);
603 			if (!ret) {
604 				p->m.private = namespace;
605 				p->event = namespace->event;
606 				return 0;
607 			}
608 			kfree(p);
609 		}
610 		put_namespace(namespace);
611 	}
612 	return ret;
613 }
614 
615 static int mounts_release(struct inode *inode, struct file *file)
616 {
617 	struct seq_file *m = file->private_data;
618 	struct namespace *namespace = m->private;
619 	put_namespace(namespace);
620 	return seq_release(inode, file);
621 }
622 
623 static unsigned mounts_poll(struct file *file, poll_table *wait)
624 {
625 	struct proc_mounts *p = file->private_data;
626 	struct namespace *ns = p->m.private;
627 	unsigned res = 0;
628 
629 	poll_wait(file, &ns->poll, wait);
630 
631 	spin_lock(&vfsmount_lock);
632 	if (p->event != ns->event) {
633 		p->event = ns->event;
634 		res = POLLERR;
635 	}
636 	spin_unlock(&vfsmount_lock);
637 
638 	return res;
639 }
640 
641 static struct file_operations proc_mounts_operations = {
642 	.open		= mounts_open,
643 	.read		= seq_read,
644 	.llseek		= seq_lseek,
645 	.release	= mounts_release,
646 	.poll		= mounts_poll,
647 };
648 
649 extern struct seq_operations mountstats_op;
650 static int mountstats_open(struct inode *inode, struct file *file)
651 {
652 	int ret = seq_open(file, &mountstats_op);
653 
654 	if (!ret) {
655 		struct seq_file *m = file->private_data;
656 		struct namespace *namespace = NULL;
657 		struct task_struct *task = get_proc_task(inode);
658 
659 		if (task) {
660 			task_lock(task);
661 			namespace = task->namespace;
662 			if (namespace)
663 				get_namespace(namespace);
664 			task_unlock(task);
665 			put_task_struct(task);
666 		}
667 
668 		if (namespace)
669 			m->private = namespace;
670 		else {
671 			seq_release(inode, file);
672 			ret = -EINVAL;
673 		}
674 	}
675 	return ret;
676 }
677 
678 static struct file_operations proc_mountstats_operations = {
679 	.open		= mountstats_open,
680 	.read		= seq_read,
681 	.llseek		= seq_lseek,
682 	.release	= mounts_release,
683 };
684 
685 #define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
686 
687 static ssize_t proc_info_read(struct file * file, char __user * buf,
688 			  size_t count, loff_t *ppos)
689 {
690 	struct inode * inode = file->f_dentry->d_inode;
691 	unsigned long page;
692 	ssize_t length;
693 	struct task_struct *task = get_proc_task(inode);
694 
695 	length = -ESRCH;
696 	if (!task)
697 		goto out_no_task;
698 
699 	if (count > PROC_BLOCK_SIZE)
700 		count = PROC_BLOCK_SIZE;
701 
702 	length = -ENOMEM;
703 	if (!(page = __get_free_page(GFP_KERNEL)))
704 		goto out;
705 
706 	length = PROC_I(inode)->op.proc_read(task, (char*)page);
707 
708 	if (length >= 0)
709 		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
710 	free_page(page);
711 out:
712 	put_task_struct(task);
713 out_no_task:
714 	return length;
715 }
716 
717 static struct file_operations proc_info_file_operations = {
718 	.read		= proc_info_read,
719 };
720 
721 static int mem_open(struct inode* inode, struct file* file)
722 {
723 	file->private_data = (void*)((long)current->self_exec_id);
724 	return 0;
725 }
726 
727 static ssize_t mem_read(struct file * file, char __user * buf,
728 			size_t count, loff_t *ppos)
729 {
730 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
731 	char *page;
732 	unsigned long src = *ppos;
733 	int ret = -ESRCH;
734 	struct mm_struct *mm;
735 
736 	if (!task)
737 		goto out_no_task;
738 
739 	if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
740 		goto out;
741 
742 	ret = -ENOMEM;
743 	page = (char *)__get_free_page(GFP_USER);
744 	if (!page)
745 		goto out;
746 
747 	ret = 0;
748 
749 	mm = get_task_mm(task);
750 	if (!mm)
751 		goto out_free;
752 
753 	ret = -EIO;
754 
755 	if (file->private_data != (void*)((long)current->self_exec_id))
756 		goto out_put;
757 
758 	ret = 0;
759 
760 	while (count > 0) {
761 		int this_len, retval;
762 
763 		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
764 		retval = access_process_vm(task, src, page, this_len, 0);
765 		if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
766 			if (!ret)
767 				ret = -EIO;
768 			break;
769 		}
770 
771 		if (copy_to_user(buf, page, retval)) {
772 			ret = -EFAULT;
773 			break;
774 		}
775 
776 		ret += retval;
777 		src += retval;
778 		buf += retval;
779 		count -= retval;
780 	}
781 	*ppos = src;
782 
783 out_put:
784 	mmput(mm);
785 out_free:
786 	free_page((unsigned long) page);
787 out:
788 	put_task_struct(task);
789 out_no_task:
790 	return ret;
791 }
792 
793 #define mem_write NULL
794 
795 #ifndef mem_write
796 /* This is a security hazard */
797 static ssize_t mem_write(struct file * file, const char * buf,
798 			 size_t count, loff_t *ppos)
799 {
800 	int copied = 0;
801 	char *page;
802 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
803 	unsigned long dst = *ppos;
804 
805 	copied = -ESRCH;
806 	if (!task)
807 		goto out_no_task;
808 
809 	if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
810 		goto out;
811 
812 	copied = -ENOMEM;
813 	page = (char *)__get_free_page(GFP_USER);
814 	if (!page)
815 		goto out;
816 
817 	while (count > 0) {
818 		int this_len, retval;
819 
820 		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
821 		if (copy_from_user(page, buf, this_len)) {
822 			copied = -EFAULT;
823 			break;
824 		}
825 		retval = access_process_vm(task, dst, page, this_len, 1);
826 		if (!retval) {
827 			if (!copied)
828 				copied = -EIO;
829 			break;
830 		}
831 		copied += retval;
832 		buf += retval;
833 		dst += retval;
834 		count -= retval;
835 	}
836 	*ppos = dst;
837 	free_page((unsigned long) page);
838 out:
839 	put_task_struct(task);
840 out_no_task:
841 	return copied;
842 }
843 #endif
844 
845 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
846 {
847 	switch (orig) {
848 	case 0:
849 		file->f_pos = offset;
850 		break;
851 	case 1:
852 		file->f_pos += offset;
853 		break;
854 	default:
855 		return -EINVAL;
856 	}
857 	force_successful_syscall_return();
858 	return file->f_pos;
859 }
860 
861 static struct file_operations proc_mem_operations = {
862 	.llseek		= mem_lseek,
863 	.read		= mem_read,
864 	.write		= mem_write,
865 	.open		= mem_open,
866 };
867 
868 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
869 				size_t count, loff_t *ppos)
870 {
871 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
872 	char buffer[PROC_NUMBUF];
873 	size_t len;
874 	int oom_adjust;
875 	loff_t __ppos = *ppos;
876 
877 	if (!task)
878 		return -ESRCH;
879 	oom_adjust = task->oomkilladj;
880 	put_task_struct(task);
881 
882 	len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
883 	if (__ppos >= len)
884 		return 0;
885 	if (count > len-__ppos)
886 		count = len-__ppos;
887 	if (copy_to_user(buf, buffer + __ppos, count))
888 		return -EFAULT;
889 	*ppos = __ppos + count;
890 	return count;
891 }
892 
893 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
894 				size_t count, loff_t *ppos)
895 {
896 	struct task_struct *task;
897 	char buffer[PROC_NUMBUF], *end;
898 	int oom_adjust;
899 
900 	if (!capable(CAP_SYS_RESOURCE))
901 		return -EPERM;
902 	memset(buffer, 0, sizeof(buffer));
903 	if (count > sizeof(buffer) - 1)
904 		count = sizeof(buffer) - 1;
905 	if (copy_from_user(buffer, buf, count))
906 		return -EFAULT;
907 	oom_adjust = simple_strtol(buffer, &end, 0);
908 	if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
909 		return -EINVAL;
910 	if (*end == '\n')
911 		end++;
912 	task = get_proc_task(file->f_dentry->d_inode);
913 	if (!task)
914 		return -ESRCH;
915 	task->oomkilladj = oom_adjust;
916 	put_task_struct(task);
917 	if (end - buffer == 0)
918 		return -EIO;
919 	return end - buffer;
920 }
921 
922 static struct file_operations proc_oom_adjust_operations = {
923 	.read		= oom_adjust_read,
924 	.write		= oom_adjust_write,
925 };
926 
927 #ifdef CONFIG_AUDITSYSCALL
928 #define TMPBUFLEN 21
929 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
930 				  size_t count, loff_t *ppos)
931 {
932 	struct inode * inode = file->f_dentry->d_inode;
933 	struct task_struct *task = get_proc_task(inode);
934 	ssize_t length;
935 	char tmpbuf[TMPBUFLEN];
936 
937 	if (!task)
938 		return -ESRCH;
939 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
940 				audit_get_loginuid(task->audit_context));
941 	put_task_struct(task);
942 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
943 }
944 
945 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
946 				   size_t count, loff_t *ppos)
947 {
948 	struct inode * inode = file->f_dentry->d_inode;
949 	char *page, *tmp;
950 	ssize_t length;
951 	uid_t loginuid;
952 
953 	if (!capable(CAP_AUDIT_CONTROL))
954 		return -EPERM;
955 
956 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
957 		return -EPERM;
958 
959 	if (count >= PAGE_SIZE)
960 		count = PAGE_SIZE - 1;
961 
962 	if (*ppos != 0) {
963 		/* No partial writes. */
964 		return -EINVAL;
965 	}
966 	page = (char*)__get_free_page(GFP_USER);
967 	if (!page)
968 		return -ENOMEM;
969 	length = -EFAULT;
970 	if (copy_from_user(page, buf, count))
971 		goto out_free_page;
972 
973 	page[count] = '\0';
974 	loginuid = simple_strtoul(page, &tmp, 10);
975 	if (tmp == page) {
976 		length = -EINVAL;
977 		goto out_free_page;
978 
979 	}
980 	length = audit_set_loginuid(current, loginuid);
981 	if (likely(length == 0))
982 		length = count;
983 
984 out_free_page:
985 	free_page((unsigned long) page);
986 	return length;
987 }
988 
989 static struct file_operations proc_loginuid_operations = {
990 	.read		= proc_loginuid_read,
991 	.write		= proc_loginuid_write,
992 };
993 #endif
994 
995 #ifdef CONFIG_SECCOMP
996 static ssize_t seccomp_read(struct file *file, char __user *buf,
997 			    size_t count, loff_t *ppos)
998 {
999 	struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
1000 	char __buf[20];
1001 	loff_t __ppos = *ppos;
1002 	size_t len;
1003 
1004 	if (!tsk)
1005 		return -ESRCH;
1006 	/* no need to print the trailing zero, so use only len */
1007 	len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
1008 	put_task_struct(tsk);
1009 	if (__ppos >= len)
1010 		return 0;
1011 	if (count > len - __ppos)
1012 		count = len - __ppos;
1013 	if (copy_to_user(buf, __buf + __ppos, count))
1014 		return -EFAULT;
1015 	*ppos = __ppos + count;
1016 	return count;
1017 }
1018 
1019 static ssize_t seccomp_write(struct file *file, const char __user *buf,
1020 			     size_t count, loff_t *ppos)
1021 {
1022 	struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
1023 	char __buf[20], *end;
1024 	unsigned int seccomp_mode;
1025 	ssize_t result;
1026 
1027 	result = -ESRCH;
1028 	if (!tsk)
1029 		goto out_no_task;
1030 
1031 	/* can set it only once to be even more secure */
1032 	result = -EPERM;
1033 	if (unlikely(tsk->seccomp.mode))
1034 		goto out;
1035 
1036 	result = -EFAULT;
1037 	memset(__buf, 0, sizeof(__buf));
1038 	count = min(count, sizeof(__buf) - 1);
1039 	if (copy_from_user(__buf, buf, count))
1040 		goto out;
1041 
1042 	seccomp_mode = simple_strtoul(__buf, &end, 0);
1043 	if (*end == '\n')
1044 		end++;
1045 	result = -EINVAL;
1046 	if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1047 		tsk->seccomp.mode = seccomp_mode;
1048 		set_tsk_thread_flag(tsk, TIF_SECCOMP);
1049 	} else
1050 		goto out;
1051 	result = -EIO;
1052 	if (unlikely(!(end - __buf)))
1053 		goto out;
1054 	result = end - __buf;
1055 out:
1056 	put_task_struct(tsk);
1057 out_no_task:
1058 	return result;
1059 }
1060 
1061 static struct file_operations proc_seccomp_operations = {
1062 	.read		= seccomp_read,
1063 	.write		= seccomp_write,
1064 };
1065 #endif /* CONFIG_SECCOMP */
1066 
1067 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1068 {
1069 	struct inode *inode = dentry->d_inode;
1070 	int error = -EACCES;
1071 
1072 	/* We don't need a base pointer in the /proc filesystem */
1073 	path_release(nd);
1074 
1075 	/* Are we allowed to snoop on the tasks file descriptors? */
1076 	if (!proc_fd_access_allowed(inode))
1077 		goto out;
1078 
1079 	error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1080 	nd->last_type = LAST_BIND;
1081 out:
1082 	return ERR_PTR(error);
1083 }
1084 
1085 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1086 			    char __user *buffer, int buflen)
1087 {
1088 	struct inode * inode;
1089 	char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1090 	int len;
1091 
1092 	if (!tmp)
1093 		return -ENOMEM;
1094 
1095 	inode = dentry->d_inode;
1096 	path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1097 	len = PTR_ERR(path);
1098 	if (IS_ERR(path))
1099 		goto out;
1100 	len = tmp + PAGE_SIZE - 1 - path;
1101 
1102 	if (len > buflen)
1103 		len = buflen;
1104 	if (copy_to_user(buffer, path, len))
1105 		len = -EFAULT;
1106  out:
1107 	free_page((unsigned long)tmp);
1108 	return len;
1109 }
1110 
1111 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1112 {
1113 	int error = -EACCES;
1114 	struct inode *inode = dentry->d_inode;
1115 	struct dentry *de;
1116 	struct vfsmount *mnt = NULL;
1117 
1118 	/* Are we allowed to snoop on the tasks file descriptors? */
1119 	if (!proc_fd_access_allowed(inode))
1120 		goto out;
1121 
1122 	error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1123 	if (error)
1124 		goto out;
1125 
1126 	error = do_proc_readlink(de, mnt, buffer, buflen);
1127 	dput(de);
1128 	mntput(mnt);
1129 out:
1130 	return error;
1131 }
1132 
1133 static struct inode_operations proc_pid_link_inode_operations = {
1134 	.readlink	= proc_pid_readlink,
1135 	.follow_link	= proc_pid_follow_link,
1136 	.setattr	= proc_setattr,
1137 };
1138 
1139 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1140 {
1141 	struct dentry *dentry = filp->f_dentry;
1142 	struct inode *inode = dentry->d_inode;
1143 	struct task_struct *p = get_proc_task(inode);
1144 	unsigned int fd, tid, ino;
1145 	int retval;
1146 	char buf[PROC_NUMBUF];
1147 	struct files_struct * files;
1148 	struct fdtable *fdt;
1149 
1150 	retval = -ENOENT;
1151 	if (!p)
1152 		goto out_no_task;
1153 	retval = 0;
1154 	tid = p->pid;
1155 
1156 	fd = filp->f_pos;
1157 	switch (fd) {
1158 		case 0:
1159 			if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1160 				goto out;
1161 			filp->f_pos++;
1162 		case 1:
1163 			ino = parent_ino(dentry);
1164 			if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1165 				goto out;
1166 			filp->f_pos++;
1167 		default:
1168 			files = get_files_struct(p);
1169 			if (!files)
1170 				goto out;
1171 			rcu_read_lock();
1172 			fdt = files_fdtable(files);
1173 			for (fd = filp->f_pos-2;
1174 			     fd < fdt->max_fds;
1175 			     fd++, filp->f_pos++) {
1176 				unsigned int i,j;
1177 
1178 				if (!fcheck_files(files, fd))
1179 					continue;
1180 				rcu_read_unlock();
1181 
1182 				j = PROC_NUMBUF;
1183 				i = fd;
1184 				do {
1185 					j--;
1186 					buf[j] = '0' + (i % 10);
1187 					i /= 10;
1188 				} while (i);
1189 
1190 				ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1191 				if (filldir(dirent, buf+j, PROC_NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1192 					rcu_read_lock();
1193 					break;
1194 				}
1195 				rcu_read_lock();
1196 			}
1197 			rcu_read_unlock();
1198 			put_files_struct(files);
1199 	}
1200 out:
1201 	put_task_struct(p);
1202 out_no_task:
1203 	return retval;
1204 }
1205 
1206 static int proc_pident_readdir(struct file *filp,
1207 		void *dirent, filldir_t filldir,
1208 		struct pid_entry *ents, unsigned int nents)
1209 {
1210 	int i;
1211 	int pid;
1212 	struct dentry *dentry = filp->f_dentry;
1213 	struct inode *inode = dentry->d_inode;
1214 	struct task_struct *task = get_proc_task(inode);
1215 	struct pid_entry *p;
1216 	ino_t ino;
1217 	int ret;
1218 
1219 	ret = -ENOENT;
1220 	if (!task)
1221 		goto out;
1222 
1223 	ret = 0;
1224 	pid = task->pid;
1225 	put_task_struct(task);
1226 	i = filp->f_pos;
1227 	switch (i) {
1228 	case 0:
1229 		ino = inode->i_ino;
1230 		if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1231 			goto out;
1232 		i++;
1233 		filp->f_pos++;
1234 		/* fall through */
1235 	case 1:
1236 		ino = parent_ino(dentry);
1237 		if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1238 			goto out;
1239 		i++;
1240 		filp->f_pos++;
1241 		/* fall through */
1242 	default:
1243 		i -= 2;
1244 		if (i >= nents) {
1245 			ret = 1;
1246 			goto out;
1247 		}
1248 		p = ents + i;
1249 		while (p->name) {
1250 			if (filldir(dirent, p->name, p->len, filp->f_pos,
1251 				    fake_ino(pid, p->type), p->mode >> 12) < 0)
1252 				goto out;
1253 			filp->f_pos++;
1254 			p++;
1255 		}
1256 	}
1257 
1258 	ret = 1;
1259 out:
1260 	return ret;
1261 }
1262 
1263 static int proc_tgid_base_readdir(struct file * filp,
1264 			     void * dirent, filldir_t filldir)
1265 {
1266 	return proc_pident_readdir(filp,dirent,filldir,
1267 				   tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1268 }
1269 
1270 static int proc_tid_base_readdir(struct file * filp,
1271 			     void * dirent, filldir_t filldir)
1272 {
1273 	return proc_pident_readdir(filp,dirent,filldir,
1274 				   tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1275 }
1276 
1277 /* building an inode */
1278 
1279 static int task_dumpable(struct task_struct *task)
1280 {
1281 	int dumpable = 0;
1282 	struct mm_struct *mm;
1283 
1284 	task_lock(task);
1285 	mm = task->mm;
1286 	if (mm)
1287 		dumpable = mm->dumpable;
1288 	task_unlock(task);
1289 	if(dumpable == 1)
1290 		return 1;
1291 	return 0;
1292 }
1293 
1294 
1295 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1296 {
1297 	struct inode * inode;
1298 	struct proc_inode *ei;
1299 
1300 	/* We need a new inode */
1301 
1302 	inode = new_inode(sb);
1303 	if (!inode)
1304 		goto out;
1305 
1306 	/* Common stuff */
1307 	ei = PROC_I(inode);
1308 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1309 	inode->i_ino = fake_ino(task->pid, ino);
1310 	inode->i_op = &proc_def_inode_operations;
1311 
1312 	/*
1313 	 * grab the reference to task.
1314 	 */
1315 	ei->pid = get_pid(task->pids[PIDTYPE_PID].pid);
1316 	if (!ei->pid)
1317 		goto out_unlock;
1318 
1319 	inode->i_uid = 0;
1320 	inode->i_gid = 0;
1321 	if (task_dumpable(task)) {
1322 		inode->i_uid = task->euid;
1323 		inode->i_gid = task->egid;
1324 	}
1325 	security_task_to_inode(task, inode);
1326 
1327 out:
1328 	return inode;
1329 
1330 out_unlock:
1331 	iput(inode);
1332 	return NULL;
1333 }
1334 
1335 /* dentry stuff */
1336 
1337 /*
1338  *	Exceptional case: normally we are not allowed to unhash a busy
1339  * directory. In this case, however, we can do it - no aliasing problems
1340  * due to the way we treat inodes.
1341  *
1342  * Rewrite the inode's ownerships here because the owning task may have
1343  * performed a setuid(), etc.
1344  *
1345  * Before the /proc/pid/status file was created the only way to read
1346  * the effective uid of a /process was to stat /proc/pid.  Reading
1347  * /proc/pid/status is slow enough that procps and other packages
1348  * kept stating /proc/pid.  To keep the rules in /proc simple I have
1349  * made this apply to all per process world readable and executable
1350  * directories.
1351  */
1352 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1353 {
1354 	struct inode *inode = dentry->d_inode;
1355 	struct task_struct *task = get_proc_task(inode);
1356 	if (task) {
1357 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1358 		    task_dumpable(task)) {
1359 			inode->i_uid = task->euid;
1360 			inode->i_gid = task->egid;
1361 		} else {
1362 			inode->i_uid = 0;
1363 			inode->i_gid = 0;
1364 		}
1365 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1366 		security_task_to_inode(task, inode);
1367 		put_task_struct(task);
1368 		return 1;
1369 	}
1370 	d_drop(dentry);
1371 	return 0;
1372 }
1373 
1374 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1375 {
1376 	struct inode *inode = dentry->d_inode;
1377 	struct task_struct *task;
1378 	generic_fillattr(inode, stat);
1379 
1380 	rcu_read_lock();
1381 	stat->uid = 0;
1382 	stat->gid = 0;
1383 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1384 	if (task) {
1385 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1386 		    task_dumpable(task)) {
1387 			stat->uid = task->euid;
1388 			stat->gid = task->egid;
1389 		}
1390 	}
1391 	rcu_read_unlock();
1392 	return 0;
1393 }
1394 
1395 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1396 {
1397 	struct inode *inode = dentry->d_inode;
1398 	struct task_struct *task = get_proc_task(inode);
1399 	int fd = proc_fd(inode);
1400 	struct files_struct *files;
1401 
1402 	if (task) {
1403 		files = get_files_struct(task);
1404 		if (files) {
1405 			rcu_read_lock();
1406 			if (fcheck_files(files, fd)) {
1407 				rcu_read_unlock();
1408 				put_files_struct(files);
1409 				if (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 				inode->i_mode &= ~(S_ISUID | S_ISGID);
1417 				security_task_to_inode(task, inode);
1418 				put_task_struct(task);
1419 				return 1;
1420 			}
1421 			rcu_read_unlock();
1422 			put_files_struct(files);
1423 		}
1424 		put_task_struct(task);
1425 	}
1426 	d_drop(dentry);
1427 	return 0;
1428 }
1429 
1430 static int pid_delete_dentry(struct dentry * dentry)
1431 {
1432 	/* Is the task we represent dead?
1433 	 * If so, then don't put the dentry on the lru list,
1434 	 * kill it immediately.
1435 	 */
1436 	return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1437 }
1438 
1439 static struct dentry_operations tid_fd_dentry_operations =
1440 {
1441 	.d_revalidate	= tid_fd_revalidate,
1442 	.d_delete	= pid_delete_dentry,
1443 };
1444 
1445 static struct dentry_operations pid_dentry_operations =
1446 {
1447 	.d_revalidate	= pid_revalidate,
1448 	.d_delete	= pid_delete_dentry,
1449 };
1450 
1451 /* Lookups */
1452 
1453 static unsigned name_to_int(struct dentry *dentry)
1454 {
1455 	const char *name = dentry->d_name.name;
1456 	int len = dentry->d_name.len;
1457 	unsigned n = 0;
1458 
1459 	if (len > 1 && *name == '0')
1460 		goto out;
1461 	while (len-- > 0) {
1462 		unsigned c = *name++ - '0';
1463 		if (c > 9)
1464 			goto out;
1465 		if (n >= (~0U-9)/10)
1466 			goto out;
1467 		n *= 10;
1468 		n += c;
1469 	}
1470 	return n;
1471 out:
1472 	return ~0U;
1473 }
1474 
1475 /* SMP-safe */
1476 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1477 {
1478 	struct task_struct *task = get_proc_task(dir);
1479 	unsigned fd = name_to_int(dentry);
1480 	struct dentry *result = ERR_PTR(-ENOENT);
1481 	struct file * file;
1482 	struct files_struct * files;
1483 	struct inode *inode;
1484 	struct proc_inode *ei;
1485 
1486 	if (!task)
1487 		goto out_no_task;
1488 	if (fd == ~0U)
1489 		goto out;
1490 
1491 	inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1492 	if (!inode)
1493 		goto out;
1494 	ei = PROC_I(inode);
1495 	ei->fd = fd;
1496 	files = get_files_struct(task);
1497 	if (!files)
1498 		goto out_unlock;
1499 	inode->i_mode = S_IFLNK;
1500 
1501 	/*
1502 	 * We are not taking a ref to the file structure, so we must
1503 	 * hold ->file_lock.
1504 	 */
1505 	spin_lock(&files->file_lock);
1506 	file = fcheck_files(files, fd);
1507 	if (!file)
1508 		goto out_unlock2;
1509 	if (file->f_mode & 1)
1510 		inode->i_mode |= S_IRUSR | S_IXUSR;
1511 	if (file->f_mode & 2)
1512 		inode->i_mode |= S_IWUSR | S_IXUSR;
1513 	spin_unlock(&files->file_lock);
1514 	put_files_struct(files);
1515 	inode->i_op = &proc_pid_link_inode_operations;
1516 	inode->i_size = 64;
1517 	ei->op.proc_get_link = proc_fd_link;
1518 	dentry->d_op = &tid_fd_dentry_operations;
1519 	d_add(dentry, inode);
1520 	/* Close the race of the process dying before we return the dentry */
1521 	if (tid_fd_revalidate(dentry, NULL))
1522 		result = NULL;
1523 out:
1524 	put_task_struct(task);
1525 out_no_task:
1526 	return result;
1527 
1528 out_unlock2:
1529 	spin_unlock(&files->file_lock);
1530 	put_files_struct(files);
1531 out_unlock:
1532 	iput(inode);
1533 	goto out;
1534 }
1535 
1536 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1537 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1538 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
1539 
1540 static struct file_operations proc_fd_operations = {
1541 	.read		= generic_read_dir,
1542 	.readdir	= proc_readfd,
1543 };
1544 
1545 static struct file_operations proc_task_operations = {
1546 	.read		= generic_read_dir,
1547 	.readdir	= proc_task_readdir,
1548 };
1549 
1550 /*
1551  * proc directories can do almost nothing..
1552  */
1553 static struct inode_operations proc_fd_inode_operations = {
1554 	.lookup		= proc_lookupfd,
1555 	.setattr	= proc_setattr,
1556 };
1557 
1558 static struct inode_operations proc_task_inode_operations = {
1559 	.lookup		= proc_task_lookup,
1560 	.getattr	= proc_task_getattr,
1561 	.setattr	= proc_setattr,
1562 };
1563 
1564 #ifdef CONFIG_SECURITY
1565 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1566 				  size_t count, loff_t *ppos)
1567 {
1568 	struct inode * inode = file->f_dentry->d_inode;
1569 	unsigned long page;
1570 	ssize_t length;
1571 	struct task_struct *task = get_proc_task(inode);
1572 
1573 	length = -ESRCH;
1574 	if (!task)
1575 		goto out_no_task;
1576 
1577 	if (count > PAGE_SIZE)
1578 		count = PAGE_SIZE;
1579 	length = -ENOMEM;
1580 	if (!(page = __get_free_page(GFP_KERNEL)))
1581 		goto out;
1582 
1583 	length = security_getprocattr(task,
1584 				      (char*)file->f_dentry->d_name.name,
1585 				      (void*)page, count);
1586 	if (length >= 0)
1587 		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1588 	free_page(page);
1589 out:
1590 	put_task_struct(task);
1591 out_no_task:
1592 	return length;
1593 }
1594 
1595 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1596 				   size_t count, loff_t *ppos)
1597 {
1598 	struct inode * inode = file->f_dentry->d_inode;
1599 	char *page;
1600 	ssize_t length;
1601 	struct task_struct *task = get_proc_task(inode);
1602 
1603 	length = -ESRCH;
1604 	if (!task)
1605 		goto out_no_task;
1606 	if (count > PAGE_SIZE)
1607 		count = PAGE_SIZE;
1608 
1609 	/* No partial writes. */
1610 	length = -EINVAL;
1611 	if (*ppos != 0)
1612 		goto out;
1613 
1614 	length = -ENOMEM;
1615 	page = (char*)__get_free_page(GFP_USER);
1616 	if (!page)
1617 		goto out;
1618 
1619 	length = -EFAULT;
1620 	if (copy_from_user(page, buf, count))
1621 		goto out_free;
1622 
1623 	length = security_setprocattr(task,
1624 				      (char*)file->f_dentry->d_name.name,
1625 				      (void*)page, count);
1626 out_free:
1627 	free_page((unsigned long) page);
1628 out:
1629 	put_task_struct(task);
1630 out_no_task:
1631 	return length;
1632 }
1633 
1634 static struct file_operations proc_pid_attr_operations = {
1635 	.read		= proc_pid_attr_read,
1636 	.write		= proc_pid_attr_write,
1637 };
1638 
1639 static struct file_operations proc_tid_attr_operations;
1640 static struct inode_operations proc_tid_attr_inode_operations;
1641 static struct file_operations proc_tgid_attr_operations;
1642 static struct inode_operations proc_tgid_attr_inode_operations;
1643 #endif
1644 
1645 /* SMP-safe */
1646 static struct dentry *proc_pident_lookup(struct inode *dir,
1647 					 struct dentry *dentry,
1648 					 struct pid_entry *ents)
1649 {
1650 	struct inode *inode;
1651 	struct dentry *error;
1652 	struct task_struct *task = get_proc_task(dir);
1653 	struct pid_entry *p;
1654 	struct proc_inode *ei;
1655 
1656 	error = ERR_PTR(-ENOENT);
1657 	inode = NULL;
1658 
1659 	if (!task)
1660 		goto out_no_task;
1661 
1662 	for (p = ents; p->name; p++) {
1663 		if (p->len != dentry->d_name.len)
1664 			continue;
1665 		if (!memcmp(dentry->d_name.name, p->name, p->len))
1666 			break;
1667 	}
1668 	if (!p->name)
1669 		goto out;
1670 
1671 	error = ERR_PTR(-EINVAL);
1672 	inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1673 	if (!inode)
1674 		goto out;
1675 
1676 	ei = PROC_I(inode);
1677 	inode->i_mode = p->mode;
1678 	/*
1679 	 * Yes, it does not scale. And it should not. Don't add
1680 	 * new entries into /proc/<tgid>/ without very good reasons.
1681 	 */
1682 	switch(p->type) {
1683 		case PROC_TGID_TASK:
1684 			inode->i_nlink = 2;
1685 			inode->i_op = &proc_task_inode_operations;
1686 			inode->i_fop = &proc_task_operations;
1687 			break;
1688 		case PROC_TID_FD:
1689 		case PROC_TGID_FD:
1690 			inode->i_nlink = 2;
1691 			inode->i_op = &proc_fd_inode_operations;
1692 			inode->i_fop = &proc_fd_operations;
1693 			break;
1694 		case PROC_TID_EXE:
1695 		case PROC_TGID_EXE:
1696 			inode->i_op = &proc_pid_link_inode_operations;
1697 			ei->op.proc_get_link = proc_exe_link;
1698 			break;
1699 		case PROC_TID_CWD:
1700 		case PROC_TGID_CWD:
1701 			inode->i_op = &proc_pid_link_inode_operations;
1702 			ei->op.proc_get_link = proc_cwd_link;
1703 			break;
1704 		case PROC_TID_ROOT:
1705 		case PROC_TGID_ROOT:
1706 			inode->i_op = &proc_pid_link_inode_operations;
1707 			ei->op.proc_get_link = proc_root_link;
1708 			break;
1709 		case PROC_TID_ENVIRON:
1710 		case PROC_TGID_ENVIRON:
1711 			inode->i_fop = &proc_info_file_operations;
1712 			ei->op.proc_read = proc_pid_environ;
1713 			break;
1714 		case PROC_TID_AUXV:
1715 		case PROC_TGID_AUXV:
1716 			inode->i_fop = &proc_info_file_operations;
1717 			ei->op.proc_read = proc_pid_auxv;
1718 			break;
1719 		case PROC_TID_STATUS:
1720 		case PROC_TGID_STATUS:
1721 			inode->i_fop = &proc_info_file_operations;
1722 			ei->op.proc_read = proc_pid_status;
1723 			break;
1724 		case PROC_TID_STAT:
1725 			inode->i_fop = &proc_info_file_operations;
1726 			ei->op.proc_read = proc_tid_stat;
1727 			break;
1728 		case PROC_TGID_STAT:
1729 			inode->i_fop = &proc_info_file_operations;
1730 			ei->op.proc_read = proc_tgid_stat;
1731 			break;
1732 		case PROC_TID_CMDLINE:
1733 		case PROC_TGID_CMDLINE:
1734 			inode->i_fop = &proc_info_file_operations;
1735 			ei->op.proc_read = proc_pid_cmdline;
1736 			break;
1737 		case PROC_TID_STATM:
1738 		case PROC_TGID_STATM:
1739 			inode->i_fop = &proc_info_file_operations;
1740 			ei->op.proc_read = proc_pid_statm;
1741 			break;
1742 		case PROC_TID_MAPS:
1743 		case PROC_TGID_MAPS:
1744 			inode->i_fop = &proc_maps_operations;
1745 			break;
1746 #ifdef CONFIG_NUMA
1747 		case PROC_TID_NUMA_MAPS:
1748 		case PROC_TGID_NUMA_MAPS:
1749 			inode->i_fop = &proc_numa_maps_operations;
1750 			break;
1751 #endif
1752 		case PROC_TID_MEM:
1753 		case PROC_TGID_MEM:
1754 			inode->i_fop = &proc_mem_operations;
1755 			break;
1756 #ifdef CONFIG_SECCOMP
1757 		case PROC_TID_SECCOMP:
1758 		case PROC_TGID_SECCOMP:
1759 			inode->i_fop = &proc_seccomp_operations;
1760 			break;
1761 #endif /* CONFIG_SECCOMP */
1762 		case PROC_TID_MOUNTS:
1763 		case PROC_TGID_MOUNTS:
1764 			inode->i_fop = &proc_mounts_operations;
1765 			break;
1766 #ifdef CONFIG_MMU
1767 		case PROC_TID_SMAPS:
1768 		case PROC_TGID_SMAPS:
1769 			inode->i_fop = &proc_smaps_operations;
1770 			break;
1771 #endif
1772 		case PROC_TID_MOUNTSTATS:
1773 		case PROC_TGID_MOUNTSTATS:
1774 			inode->i_fop = &proc_mountstats_operations;
1775 			break;
1776 #ifdef CONFIG_SECURITY
1777 		case PROC_TID_ATTR:
1778 			inode->i_nlink = 2;
1779 			inode->i_op = &proc_tid_attr_inode_operations;
1780 			inode->i_fop = &proc_tid_attr_operations;
1781 			break;
1782 		case PROC_TGID_ATTR:
1783 			inode->i_nlink = 2;
1784 			inode->i_op = &proc_tgid_attr_inode_operations;
1785 			inode->i_fop = &proc_tgid_attr_operations;
1786 			break;
1787 		case PROC_TID_ATTR_CURRENT:
1788 		case PROC_TGID_ATTR_CURRENT:
1789 		case PROC_TID_ATTR_PREV:
1790 		case PROC_TGID_ATTR_PREV:
1791 		case PROC_TID_ATTR_EXEC:
1792 		case PROC_TGID_ATTR_EXEC:
1793 		case PROC_TID_ATTR_FSCREATE:
1794 		case PROC_TGID_ATTR_FSCREATE:
1795 		case PROC_TID_ATTR_KEYCREATE:
1796 		case PROC_TGID_ATTR_KEYCREATE:
1797 		case PROC_TID_ATTR_SOCKCREATE:
1798 		case PROC_TGID_ATTR_SOCKCREATE:
1799 			inode->i_fop = &proc_pid_attr_operations;
1800 			break;
1801 #endif
1802 #ifdef CONFIG_KALLSYMS
1803 		case PROC_TID_WCHAN:
1804 		case PROC_TGID_WCHAN:
1805 			inode->i_fop = &proc_info_file_operations;
1806 			ei->op.proc_read = proc_pid_wchan;
1807 			break;
1808 #endif
1809 #ifdef CONFIG_SCHEDSTATS
1810 		case PROC_TID_SCHEDSTAT:
1811 		case PROC_TGID_SCHEDSTAT:
1812 			inode->i_fop = &proc_info_file_operations;
1813 			ei->op.proc_read = proc_pid_schedstat;
1814 			break;
1815 #endif
1816 #ifdef CONFIG_CPUSETS
1817 		case PROC_TID_CPUSET:
1818 		case PROC_TGID_CPUSET:
1819 			inode->i_fop = &proc_cpuset_operations;
1820 			break;
1821 #endif
1822 		case PROC_TID_OOM_SCORE:
1823 		case PROC_TGID_OOM_SCORE:
1824 			inode->i_fop = &proc_info_file_operations;
1825 			ei->op.proc_read = proc_oom_score;
1826 			break;
1827 		case PROC_TID_OOM_ADJUST:
1828 		case PROC_TGID_OOM_ADJUST:
1829 			inode->i_fop = &proc_oom_adjust_operations;
1830 			break;
1831 #ifdef CONFIG_AUDITSYSCALL
1832 		case PROC_TID_LOGINUID:
1833 		case PROC_TGID_LOGINUID:
1834 			inode->i_fop = &proc_loginuid_operations;
1835 			break;
1836 #endif
1837 		default:
1838 			printk("procfs: impossible type (%d)",p->type);
1839 			iput(inode);
1840 			error = ERR_PTR(-EINVAL);
1841 			goto out;
1842 	}
1843 	dentry->d_op = &pid_dentry_operations;
1844 	d_add(dentry, inode);
1845 	/* Close the race of the process dying before we return the dentry */
1846 	if (pid_revalidate(dentry, NULL))
1847 		error = NULL;
1848 out:
1849 	put_task_struct(task);
1850 out_no_task:
1851 	return error;
1852 }
1853 
1854 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1855 	return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1856 }
1857 
1858 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1859 	return proc_pident_lookup(dir, dentry, tid_base_stuff);
1860 }
1861 
1862 static struct file_operations proc_tgid_base_operations = {
1863 	.read		= generic_read_dir,
1864 	.readdir	= proc_tgid_base_readdir,
1865 };
1866 
1867 static struct file_operations proc_tid_base_operations = {
1868 	.read		= generic_read_dir,
1869 	.readdir	= proc_tid_base_readdir,
1870 };
1871 
1872 static struct inode_operations proc_tgid_base_inode_operations = {
1873 	.lookup		= proc_tgid_base_lookup,
1874 	.getattr	= pid_getattr,
1875 	.setattr	= proc_setattr,
1876 };
1877 
1878 static struct inode_operations proc_tid_base_inode_operations = {
1879 	.lookup		= proc_tid_base_lookup,
1880 	.getattr	= pid_getattr,
1881 	.setattr	= proc_setattr,
1882 };
1883 
1884 #ifdef CONFIG_SECURITY
1885 static int proc_tgid_attr_readdir(struct file * filp,
1886 			     void * dirent, filldir_t filldir)
1887 {
1888 	return proc_pident_readdir(filp,dirent,filldir,
1889 				   tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1890 }
1891 
1892 static int proc_tid_attr_readdir(struct file * filp,
1893 			     void * dirent, filldir_t filldir)
1894 {
1895 	return proc_pident_readdir(filp,dirent,filldir,
1896 				   tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1897 }
1898 
1899 static struct file_operations proc_tgid_attr_operations = {
1900 	.read		= generic_read_dir,
1901 	.readdir	= proc_tgid_attr_readdir,
1902 };
1903 
1904 static struct file_operations proc_tid_attr_operations = {
1905 	.read		= generic_read_dir,
1906 	.readdir	= proc_tid_attr_readdir,
1907 };
1908 
1909 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1910 				struct dentry *dentry, struct nameidata *nd)
1911 {
1912 	return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1913 }
1914 
1915 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1916 				struct dentry *dentry, struct nameidata *nd)
1917 {
1918 	return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1919 }
1920 
1921 static struct inode_operations proc_tgid_attr_inode_operations = {
1922 	.lookup		= proc_tgid_attr_lookup,
1923 	.getattr	= pid_getattr,
1924 	.setattr	= proc_setattr,
1925 };
1926 
1927 static struct inode_operations proc_tid_attr_inode_operations = {
1928 	.lookup		= proc_tid_attr_lookup,
1929 	.getattr	= pid_getattr,
1930 	.setattr	= proc_setattr,
1931 };
1932 #endif
1933 
1934 /*
1935  * /proc/self:
1936  */
1937 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1938 			      int buflen)
1939 {
1940 	char tmp[PROC_NUMBUF];
1941 	sprintf(tmp, "%d", current->tgid);
1942 	return vfs_readlink(dentry,buffer,buflen,tmp);
1943 }
1944 
1945 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1946 {
1947 	char tmp[PROC_NUMBUF];
1948 	sprintf(tmp, "%d", current->tgid);
1949 	return ERR_PTR(vfs_follow_link(nd,tmp));
1950 }
1951 
1952 static struct inode_operations proc_self_inode_operations = {
1953 	.readlink	= proc_self_readlink,
1954 	.follow_link	= proc_self_follow_link,
1955 };
1956 
1957 /**
1958  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
1959  *
1960  * @task: task that should be flushed.
1961  *
1962  * Looks in the dcache for
1963  * /proc/@pid
1964  * /proc/@tgid/task/@pid
1965  * if either directory is present flushes it and all of it'ts children
1966  * from the dcache.
1967  *
1968  * It is safe and reasonable to cache /proc entries for a task until
1969  * that task exits.  After that they just clog up the dcache with
1970  * useless entries, possibly causing useful dcache entries to be
1971  * flushed instead.  This routine is proved to flush those useless
1972  * dcache entries at process exit time.
1973  *
1974  * NOTE: This routine is just an optimization so it does not guarantee
1975  *       that no dcache entries will exist at process exit time it
1976  *       just makes it very unlikely that any will persist.
1977  */
1978 void proc_flush_task(struct task_struct *task)
1979 {
1980 	struct dentry *dentry, *leader, *dir;
1981 	char buf[PROC_NUMBUF];
1982 	struct qstr name;
1983 
1984 	name.name = buf;
1985 	name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1986 	dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1987 	if (dentry) {
1988 		shrink_dcache_parent(dentry);
1989 		d_drop(dentry);
1990 		dput(dentry);
1991 	}
1992 
1993 	if (thread_group_leader(task))
1994 		goto out;
1995 
1996 	name.name = buf;
1997 	name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1998 	leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1999 	if (!leader)
2000 		goto out;
2001 
2002 	name.name = "task";
2003 	name.len = strlen(name.name);
2004 	dir = d_hash_and_lookup(leader, &name);
2005 	if (!dir)
2006 		goto out_put_leader;
2007 
2008 	name.name = buf;
2009 	name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2010 	dentry = d_hash_and_lookup(dir, &name);
2011 	if (dentry) {
2012 		shrink_dcache_parent(dentry);
2013 		d_drop(dentry);
2014 		dput(dentry);
2015 	}
2016 
2017 	dput(dir);
2018 out_put_leader:
2019 	dput(leader);
2020 out:
2021 	return;
2022 }
2023 
2024 /* SMP-safe */
2025 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2026 {
2027 	struct dentry *result = ERR_PTR(-ENOENT);
2028 	struct task_struct *task;
2029 	struct inode *inode;
2030 	struct proc_inode *ei;
2031 	unsigned tgid;
2032 
2033 	if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
2034 		inode = new_inode(dir->i_sb);
2035 		if (!inode)
2036 			return ERR_PTR(-ENOMEM);
2037 		ei = PROC_I(inode);
2038 		inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2039 		inode->i_ino = fake_ino(0, PROC_TGID_INO);
2040 		ei->pde = NULL;
2041 		inode->i_mode = S_IFLNK|S_IRWXUGO;
2042 		inode->i_uid = inode->i_gid = 0;
2043 		inode->i_size = 64;
2044 		inode->i_op = &proc_self_inode_operations;
2045 		d_add(dentry, inode);
2046 		return NULL;
2047 	}
2048 	tgid = name_to_int(dentry);
2049 	if (tgid == ~0U)
2050 		goto out;
2051 
2052 	rcu_read_lock();
2053 	task = find_task_by_pid(tgid);
2054 	if (task)
2055 		get_task_struct(task);
2056 	rcu_read_unlock();
2057 	if (!task)
2058 		goto out;
2059 
2060 	inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
2061 	if (!inode)
2062 		goto out_put_task;
2063 
2064 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2065 	inode->i_op = &proc_tgid_base_inode_operations;
2066 	inode->i_fop = &proc_tgid_base_operations;
2067 	inode->i_flags|=S_IMMUTABLE;
2068 #ifdef CONFIG_SECURITY
2069 	inode->i_nlink = 5;
2070 #else
2071 	inode->i_nlink = 4;
2072 #endif
2073 
2074 	dentry->d_op = &pid_dentry_operations;
2075 
2076 	d_add(dentry, inode);
2077 	/* Close the race of the process dying before we return the dentry */
2078 	if (pid_revalidate(dentry, NULL))
2079 		result = NULL;
2080 
2081 out_put_task:
2082 	put_task_struct(task);
2083 out:
2084 	return result;
2085 }
2086 
2087 /* SMP-safe */
2088 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2089 {
2090 	struct dentry *result = ERR_PTR(-ENOENT);
2091 	struct task_struct *task;
2092 	struct task_struct *leader = get_proc_task(dir);
2093 	struct inode *inode;
2094 	unsigned tid;
2095 
2096 	if (!leader)
2097 		goto out_no_task;
2098 
2099 	tid = name_to_int(dentry);
2100 	if (tid == ~0U)
2101 		goto out;
2102 
2103 	rcu_read_lock();
2104 	task = find_task_by_pid(tid);
2105 	if (task)
2106 		get_task_struct(task);
2107 	rcu_read_unlock();
2108 	if (!task)
2109 		goto out;
2110 	if (leader->tgid != task->tgid)
2111 		goto out_drop_task;
2112 
2113 	inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2114 
2115 
2116 	if (!inode)
2117 		goto out_drop_task;
2118 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2119 	inode->i_op = &proc_tid_base_inode_operations;
2120 	inode->i_fop = &proc_tid_base_operations;
2121 	inode->i_flags|=S_IMMUTABLE;
2122 #ifdef CONFIG_SECURITY
2123 	inode->i_nlink = 4;
2124 #else
2125 	inode->i_nlink = 3;
2126 #endif
2127 
2128 	dentry->d_op = &pid_dentry_operations;
2129 
2130 	d_add(dentry, inode);
2131 	/* Close the race of the process dying before we return the dentry */
2132 	if (pid_revalidate(dentry, NULL))
2133 		result = NULL;
2134 
2135 out_drop_task:
2136 	put_task_struct(task);
2137 out:
2138 	put_task_struct(leader);
2139 out_no_task:
2140 	return result;
2141 }
2142 
2143 /*
2144  * Find the first tgid to return to user space.
2145  *
2146  * Usually this is just whatever follows &init_task, but if the users
2147  * buffer was too small to hold the full list or there was a seek into
2148  * the middle of the directory we have more work to do.
2149  *
2150  * In the case of a short read we start with find_task_by_pid.
2151  *
2152  * In the case of a seek we start with &init_task and walk nr
2153  * threads past it.
2154  */
2155 static struct task_struct *first_tgid(int tgid, unsigned int nr)
2156 {
2157 	struct task_struct *pos;
2158 	rcu_read_lock();
2159 	if (tgid && nr) {
2160 		pos = find_task_by_pid(tgid);
2161 		if (pos && thread_group_leader(pos))
2162 			goto found;
2163 	}
2164 	/* If nr exceeds the number of processes get out quickly */
2165 	pos = NULL;
2166 	if (nr && nr >= nr_processes())
2167 		goto done;
2168 
2169 	/* If we haven't found our starting place yet start with
2170 	 * the init_task and walk nr tasks forward.
2171 	 */
2172 	for (pos = next_task(&init_task); nr > 0; --nr) {
2173 		pos = next_task(pos);
2174 		if (pos == &init_task) {
2175 			pos = NULL;
2176 			goto done;
2177 		}
2178 	}
2179 found:
2180 	get_task_struct(pos);
2181 done:
2182 	rcu_read_unlock();
2183 	return pos;
2184 }
2185 
2186 /*
2187  * Find the next task in the task list.
2188  * Return NULL if we loop or there is any error.
2189  *
2190  * The reference to the input task_struct is released.
2191  */
2192 static struct task_struct *next_tgid(struct task_struct *start)
2193 {
2194 	struct task_struct *pos;
2195 	rcu_read_lock();
2196 	pos = start;
2197 	if (pid_alive(start))
2198 		pos = next_task(start);
2199 	if (pid_alive(pos) && (pos != &init_task)) {
2200 		get_task_struct(pos);
2201 		goto done;
2202 	}
2203 	pos = NULL;
2204 done:
2205 	rcu_read_unlock();
2206 	put_task_struct(start);
2207 	return pos;
2208 }
2209 
2210 /* for the /proc/ directory itself, after non-process stuff has been done */
2211 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2212 {
2213 	char buf[PROC_NUMBUF];
2214 	unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2215 	struct task_struct *task;
2216 	int tgid;
2217 
2218 	if (!nr) {
2219 		ino_t ino = fake_ino(0,PROC_TGID_INO);
2220 		if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2221 			return 0;
2222 		filp->f_pos++;
2223 		nr++;
2224 	}
2225 	nr -= 1;
2226 
2227 	/* f_version caches the tgid value that the last readdir call couldn't
2228 	 * return. lseek aka telldir automagically resets f_version to 0.
2229 	 */
2230 	tgid = filp->f_version;
2231 	filp->f_version = 0;
2232 	for (task = first_tgid(tgid, nr);
2233 	     task;
2234 	     task = next_tgid(task), filp->f_pos++) {
2235 		int len;
2236 		ino_t ino;
2237 		tgid = task->pid;
2238 		len = snprintf(buf, sizeof(buf), "%d", tgid);
2239 		ino = fake_ino(tgid, PROC_TGID_INO);
2240 		if (filldir(dirent, buf, len, filp->f_pos, ino, DT_DIR) < 0) {
2241 			/* returning this tgid failed, save it as the first
2242 			 * pid for the next readir call */
2243 			filp->f_version = tgid;
2244 			put_task_struct(task);
2245 			break;
2246 		}
2247 	}
2248 	return 0;
2249 }
2250 
2251 /*
2252  * Find the first tid of a thread group to return to user space.
2253  *
2254  * Usually this is just the thread group leader, but if the users
2255  * buffer was too small or there was a seek into the middle of the
2256  * directory we have more work todo.
2257  *
2258  * In the case of a short read we start with find_task_by_pid.
2259  *
2260  * In the case of a seek we start with the leader and walk nr
2261  * threads past it.
2262  */
2263 static struct task_struct *first_tid(struct task_struct *leader,
2264 					int tid, int nr)
2265 {
2266 	struct task_struct *pos;
2267 
2268 	rcu_read_lock();
2269 	/* Attempt to start with the pid of a thread */
2270 	if (tid && (nr > 0)) {
2271 		pos = find_task_by_pid(tid);
2272 		if (pos && (pos->group_leader == leader))
2273 			goto found;
2274 	}
2275 
2276 	/* If nr exceeds the number of threads there is nothing todo */
2277 	pos = NULL;
2278 	if (nr && nr >= get_nr_threads(leader))
2279 		goto out;
2280 
2281 	/* If we haven't found our starting place yet start
2282 	 * with the leader and walk nr threads forward.
2283 	 */
2284 	for (pos = leader; nr > 0; --nr) {
2285 		pos = next_thread(pos);
2286 		if (pos == leader) {
2287 			pos = NULL;
2288 			goto out;
2289 		}
2290 	}
2291 found:
2292 	get_task_struct(pos);
2293 out:
2294 	rcu_read_unlock();
2295 	return pos;
2296 }
2297 
2298 /*
2299  * Find the next thread in the thread list.
2300  * Return NULL if there is an error or no next thread.
2301  *
2302  * The reference to the input task_struct is released.
2303  */
2304 static struct task_struct *next_tid(struct task_struct *start)
2305 {
2306 	struct task_struct *pos = NULL;
2307 	rcu_read_lock();
2308 	if (pid_alive(start)) {
2309 		pos = next_thread(start);
2310 		if (thread_group_leader(pos))
2311 			pos = NULL;
2312 		else
2313 			get_task_struct(pos);
2314 	}
2315 	rcu_read_unlock();
2316 	put_task_struct(start);
2317 	return pos;
2318 }
2319 
2320 /* for the /proc/TGID/task/ directories */
2321 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2322 {
2323 	char buf[PROC_NUMBUF];
2324 	struct dentry *dentry = filp->f_dentry;
2325 	struct inode *inode = dentry->d_inode;
2326 	struct task_struct *leader = get_proc_task(inode);
2327 	struct task_struct *task;
2328 	int retval = -ENOENT;
2329 	ino_t ino;
2330 	int tid;
2331 	unsigned long pos = filp->f_pos;  /* avoiding "long long" filp->f_pos */
2332 
2333 	if (!leader)
2334 		goto out_no_task;
2335 	retval = 0;
2336 
2337 	switch (pos) {
2338 	case 0:
2339 		ino = inode->i_ino;
2340 		if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2341 			goto out;
2342 		pos++;
2343 		/* fall through */
2344 	case 1:
2345 		ino = parent_ino(dentry);
2346 		if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2347 			goto out;
2348 		pos++;
2349 		/* fall through */
2350 	}
2351 
2352 	/* f_version caches the tgid value that the last readdir call couldn't
2353 	 * return. lseek aka telldir automagically resets f_version to 0.
2354 	 */
2355 	tid = filp->f_version;
2356 	filp->f_version = 0;
2357 	for (task = first_tid(leader, tid, pos - 2);
2358 	     task;
2359 	     task = next_tid(task), pos++) {
2360 		int len;
2361 		tid = task->pid;
2362 		len = snprintf(buf, sizeof(buf), "%d", tid);
2363 		ino = fake_ino(tid, PROC_TID_INO);
2364 		if (filldir(dirent, buf, len, pos, ino, DT_DIR < 0)) {
2365 			/* returning this tgid failed, save it as the first
2366 			 * pid for the next readir call */
2367 			filp->f_version = tid;
2368 			put_task_struct(task);
2369 			break;
2370 		}
2371 	}
2372 out:
2373 	filp->f_pos = pos;
2374 	put_task_struct(leader);
2375 out_no_task:
2376 	return retval;
2377 }
2378 
2379 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2380 {
2381 	struct inode *inode = dentry->d_inode;
2382 	struct task_struct *p = get_proc_task(inode);
2383 	generic_fillattr(inode, stat);
2384 
2385 	if (p) {
2386 		rcu_read_lock();
2387 		stat->nlink += get_nr_threads(p);
2388 		rcu_read_unlock();
2389 		put_task_struct(p);
2390 	}
2391 
2392 	return 0;
2393 }
2394