~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/proc/base.c

Version: ~ [ linux-5.2-rc1 ] ~ [ linux-5.1.2 ] ~ [ linux-5.0.16 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.43 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.119 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.176 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.179 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.139 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.67 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.39.4 ] ~ [ linux-2.6.38.8 ] ~ [ linux-2.6.37.6 ] ~ [ linux-2.6.36.4 ] ~ [ linux-2.6.35.14 ] ~ [ linux-2.6.34.15 ] ~ [ linux-2.6.33.20 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0
  2 /*
  3  *  linux/fs/proc/base.c
  4  *
  5  *  Copyright (C) 1991, 1992 Linus Torvalds
  6  *
  7  *  proc base directory handling functions
  8  *
  9  *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
 10  *  Instead of using magical inumbers to determine the kind of object
 11  *  we allocate and fill in-core inodes upon lookup. They don't even
 12  *  go into icache. We cache the reference to task_struct upon lookup too.
 13  *  Eventually it should become a filesystem in its own. We don't use the
 14  *  rest of procfs anymore.
 15  *
 16  *
 17  *  Changelog:
 18  *  17-Jan-2005
 19  *  Allan Bezerra
 20  *  Bruna Moreira <bruna.moreira@indt.org.br>
 21  *  Edjard Mota <edjard.mota@indt.org.br>
 22  *  Ilias Biris <ilias.biris@indt.org.br>
 23  *  Mauricio Lin <mauricio.lin@indt.org.br>
 24  *
 25  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
 26  *
 27  *  A new process specific entry (smaps) included in /proc. It shows the
 28  *  size of rss for each memory area. The maps entry lacks information
 29  *  about physical memory size (rss) for each mapped file, i.e.,
 30  *  rss information for executables and library files.
 31  *  This additional information is useful for any tools that need to know
 32  *  about physical memory consumption for a process specific library.
 33  *
 34  *  Changelog:
 35  *  21-Feb-2005
 36  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
 37  *  Pud inclusion in the page table walking.
 38  *
 39  *  ChangeLog:
 40  *  10-Mar-2005
 41  *  10LE Instituto Nokia de Tecnologia - INdT:
 42  *  A better way to walks through the page table as suggested by Hugh Dickins.
 43  *
 44  *  Simo Piiroinen <simo.piiroinen@nokia.com>:
 45  *  Smaps information related to shared, private, clean and dirty pages.
 46  *
 47  *  Paul Mundt <paul.mundt@nokia.com>:
 48  *  Overall revision about smaps.
 49  */
 50 
 51 #include <linux/uaccess.h>
 52 
 53 #include <linux/errno.h>
 54 #include <linux/time.h>
 55 #include <linux/proc_fs.h>
 56 #include <linux/stat.h>
 57 #include <linux/task_io_accounting_ops.h>
 58 #include <linux/init.h>
 59 #include <linux/capability.h>
 60 #include <linux/file.h>
 61 #include <linux/fdtable.h>
 62 #include <linux/string.h>
 63 #include <linux/seq_file.h>
 64 #include <linux/namei.h>
 65 #include <linux/mnt_namespace.h>
 66 #include <linux/mm.h>
 67 #include <linux/swap.h>
 68 #include <linux/rcupdate.h>
 69 #include <linux/kallsyms.h>
 70 #include <linux/stacktrace.h>
 71 #include <linux/resource.h>
 72 #include <linux/module.h>
 73 #include <linux/mount.h>
 74 #include <linux/security.h>
 75 #include <linux/ptrace.h>
 76 #include <linux/tracehook.h>
 77 #include <linux/printk.h>
 78 #include <linux/cache.h>
 79 #include <linux/cgroup.h>
 80 #include <linux/cpuset.h>
 81 #include <linux/audit.h>
 82 #include <linux/poll.h>
 83 #include <linux/nsproxy.h>
 84 #include <linux/oom.h>
 85 #include <linux/elf.h>
 86 #include <linux/pid_namespace.h>
 87 #include <linux/user_namespace.h>
 88 #include <linux/fs_struct.h>
 89 #include <linux/slab.h>
 90 #include <linux/sched/autogroup.h>
 91 #include <linux/sched/mm.h>
 92 #include <linux/sched/coredump.h>
 93 #include <linux/sched/debug.h>
 94 #include <linux/sched/stat.h>
 95 #include <linux/flex_array.h>
 96 #include <linux/posix-timers.h>
 97 #ifdef CONFIG_HARDWALL
 98 #include <asm/hardwall.h>
 99 #endif
100 #include <trace/events/oom.h>
101 #include "internal.h"
102 #include "fd.h"
103 
104 #include "../../lib/kstrtox.h"
105 
106 /* NOTE:
107  *      Implementing inode permission operations in /proc is almost
108  *      certainly an error.  Permission checks need to happen during
109  *      each system call not at open time.  The reason is that most of
110  *      what we wish to check for permissions in /proc varies at runtime.
111  *
112  *      The classic example of a problem is opening file descriptors
113  *      in /proc for a task before it execs a suid executable.
114  */
115 
116 static u8 nlink_tid __ro_after_init;
117 static u8 nlink_tgid __ro_after_init;
118 
119 struct pid_entry {
120         const char *name;
121         unsigned int len;
122         umode_t mode;
123         const struct inode_operations *iop;
124         const struct file_operations *fop;
125         union proc_op op;
126 };
127 
128 #define NOD(NAME, MODE, IOP, FOP, OP) {                 \
129         .name = (NAME),                                 \
130         .len  = sizeof(NAME) - 1,                       \
131         .mode = MODE,                                   \
132         .iop  = IOP,                                    \
133         .fop  = FOP,                                    \
134         .op   = OP,                                     \
135 }
136 
137 #define DIR(NAME, MODE, iops, fops)     \
138         NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
139 #define LNK(NAME, get_link)                                     \
140         NOD(NAME, (S_IFLNK|S_IRWXUGO),                          \
141                 &proc_pid_link_inode_operations, NULL,          \
142                 { .proc_get_link = get_link } )
143 #define REG(NAME, MODE, fops)                           \
144         NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
145 #define ONE(NAME, MODE, show)                           \
146         NOD(NAME, (S_IFREG|(MODE)),                     \
147                 NULL, &proc_single_file_operations,     \
148                 { .proc_show = show } )
149 
150 /*
151  * Count the number of hardlinks for the pid_entry table, excluding the .
152  * and .. links.
153  */
154 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
155         unsigned int n)
156 {
157         unsigned int i;
158         unsigned int count;
159 
160         count = 2;
161         for (i = 0; i < n; ++i) {
162                 if (S_ISDIR(entries[i].mode))
163                         ++count;
164         }
165 
166         return count;
167 }
168 
169 static int get_task_root(struct task_struct *task, struct path *root)
170 {
171         int result = -ENOENT;
172 
173         task_lock(task);
174         if (task->fs) {
175                 get_fs_root(task->fs, root);
176                 result = 0;
177         }
178         task_unlock(task);
179         return result;
180 }
181 
182 static int proc_cwd_link(struct dentry *dentry, struct path *path)
183 {
184         struct task_struct *task = get_proc_task(d_inode(dentry));
185         int result = -ENOENT;
186 
187         if (task) {
188                 task_lock(task);
189                 if (task->fs) {
190                         get_fs_pwd(task->fs, path);
191                         result = 0;
192                 }
193                 task_unlock(task);
194                 put_task_struct(task);
195         }
196         return result;
197 }
198 
199 static int proc_root_link(struct dentry *dentry, struct path *path)
200 {
201         struct task_struct *task = get_proc_task(d_inode(dentry));
202         int result = -ENOENT;
203 
204         if (task) {
205                 result = get_task_root(task, path);
206                 put_task_struct(task);
207         }
208         return result;
209 }
210 
211 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
212                                      size_t _count, loff_t *pos)
213 {
214         struct task_struct *tsk;
215         struct mm_struct *mm;
216         char *page;
217         unsigned long count = _count;
218         unsigned long arg_start, arg_end, env_start, env_end;
219         unsigned long len1, len2, len;
220         unsigned long p;
221         char c;
222         ssize_t rv;
223 
224         BUG_ON(*pos < 0);
225 
226         tsk = get_proc_task(file_inode(file));
227         if (!tsk)
228                 return -ESRCH;
229         mm = get_task_mm(tsk);
230         put_task_struct(tsk);
231         if (!mm)
232                 return 0;
233         /* Check if process spawned far enough to have cmdline. */
234         if (!mm->env_end) {
235                 rv = 0;
236                 goto out_mmput;
237         }
238 
239         page = (char *)__get_free_page(GFP_KERNEL);
240         if (!page) {
241                 rv = -ENOMEM;
242                 goto out_mmput;
243         }
244 
245         down_read(&mm->mmap_sem);
246         arg_start = mm->arg_start;
247         arg_end = mm->arg_end;
248         env_start = mm->env_start;
249         env_end = mm->env_end;
250         up_read(&mm->mmap_sem);
251 
252         BUG_ON(arg_start > arg_end);
253         BUG_ON(env_start > env_end);
254 
255         len1 = arg_end - arg_start;
256         len2 = env_end - env_start;
257 
258         /* Empty ARGV. */
259         if (len1 == 0) {
260                 rv = 0;
261                 goto out_free_page;
262         }
263         /*
264          * Inherently racy -- command line shares address space
265          * with code and data.
266          */
267         rv = access_remote_vm(mm, arg_end - 1, &c, 1, FOLL_ANON);
268         if (rv <= 0)
269                 goto out_free_page;
270 
271         rv = 0;
272 
273         if (c == '\0') {
274                 /* Command line (set of strings) occupies whole ARGV. */
275                 if (len1 <= *pos)
276                         goto out_free_page;
277 
278                 p = arg_start + *pos;
279                 len = len1 - *pos;
280                 while (count > 0 && len > 0) {
281                         unsigned int _count;
282                         int nr_read;
283 
284                         _count = min3(count, len, PAGE_SIZE);
285                         nr_read = access_remote_vm(mm, p, page, _count, FOLL_ANON);
286                         if (nr_read < 0)
287                                 rv = nr_read;
288                         if (nr_read <= 0)
289                                 goto out_free_page;
290 
291                         if (copy_to_user(buf, page, nr_read)) {
292                                 rv = -EFAULT;
293                                 goto out_free_page;
294                         }
295 
296                         p       += nr_read;
297                         len     -= nr_read;
298                         buf     += nr_read;
299                         count   -= nr_read;
300                         rv      += nr_read;
301                 }
302         } else {
303                 /*
304                  * Command line (1 string) occupies ARGV and
305                  * extends into ENVP.
306                  */
307                 struct {
308                         unsigned long p;
309                         unsigned long len;
310                 } cmdline[2] = {
311                         { .p = arg_start, .len = len1 },
312                         { .p = env_start, .len = len2 },
313                 };
314                 loff_t pos1 = *pos;
315                 unsigned int i;
316 
317                 i = 0;
318                 while (i < 2 && pos1 >= cmdline[i].len) {
319                         pos1 -= cmdline[i].len;
320                         i++;
321                 }
322                 while (i < 2) {
323                         p = cmdline[i].p + pos1;
324                         len = cmdline[i].len - pos1;
325                         while (count > 0 && len > 0) {
326                                 unsigned int _count, l;
327                                 int nr_read;
328                                 bool final;
329 
330                                 _count = min3(count, len, PAGE_SIZE);
331                                 nr_read = access_remote_vm(mm, p, page, _count, FOLL_ANON);
332                                 if (nr_read < 0)
333                                         rv = nr_read;
334                                 if (nr_read <= 0)
335                                         goto out_free_page;
336 
337                                 /*
338                                  * Command line can be shorter than whole ARGV
339                                  * even if last "marker" byte says it is not.
340                                  */
341                                 final = false;
342                                 l = strnlen(page, nr_read);
343                                 if (l < nr_read) {
344                                         nr_read = l;
345                                         final = true;
346                                 }
347 
348                                 if (copy_to_user(buf, page, nr_read)) {
349                                         rv = -EFAULT;
350                                         goto out_free_page;
351                                 }
352 
353                                 p       += nr_read;
354                                 len     -= nr_read;
355                                 buf     += nr_read;
356                                 count   -= nr_read;
357                                 rv      += nr_read;
358 
359                                 if (final)
360                                         goto out_free_page;
361                         }
362 
363                         /* Only first chunk can be read partially. */
364                         pos1 = 0;
365                         i++;
366                 }
367         }
368 
369 out_free_page:
370         free_page((unsigned long)page);
371 out_mmput:
372         mmput(mm);
373         if (rv > 0)
374                 *pos += rv;
375         return rv;
376 }
377 
378 static const struct file_operations proc_pid_cmdline_ops = {
379         .read   = proc_pid_cmdline_read,
380         .llseek = generic_file_llseek,
381 };
382 
383 #ifdef CONFIG_KALLSYMS
384 /*
385  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
386  * Returns the resolved symbol.  If that fails, simply return the address.
387  */
388 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
389                           struct pid *pid, struct task_struct *task)
390 {
391         unsigned long wchan;
392         char symname[KSYM_NAME_LEN];
393 
394         wchan = get_wchan(task);
395 
396         if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
397                         && !lookup_symbol_name(wchan, symname))
398                 seq_printf(m, "%s", symname);
399         else
400                 seq_putc(m, '');
401 
402         return 0;
403 }
404 #endif /* CONFIG_KALLSYMS */
405 
406 static int lock_trace(struct task_struct *task)
407 {
408         int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
409         if (err)
410                 return err;
411         if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
412                 mutex_unlock(&task->signal->cred_guard_mutex);
413                 return -EPERM;
414         }
415         return 0;
416 }
417 
418 static void unlock_trace(struct task_struct *task)
419 {
420         mutex_unlock(&task->signal->cred_guard_mutex);
421 }
422 
423 #ifdef CONFIG_STACKTRACE
424 
425 #define MAX_STACK_TRACE_DEPTH   64
426 
427 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
428                           struct pid *pid, struct task_struct *task)
429 {
430         struct stack_trace trace;
431         unsigned long *entries;
432         int err;
433         int i;
434 
435         entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
436         if (!entries)
437                 return -ENOMEM;
438 
439         trace.nr_entries        = 0;
440         trace.max_entries       = MAX_STACK_TRACE_DEPTH;
441         trace.entries           = entries;
442         trace.skip              = 0;
443 
444         err = lock_trace(task);
445         if (!err) {
446                 save_stack_trace_tsk(task, &trace);
447 
448                 for (i = 0; i < trace.nr_entries; i++) {
449                         seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
450                 }
451                 unlock_trace(task);
452         }
453         kfree(entries);
454 
455         return err;
456 }
457 #endif
458 
459 #ifdef CONFIG_SCHED_INFO
460 /*
461  * Provides /proc/PID/schedstat
462  */
463 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
464                               struct pid *pid, struct task_struct *task)
465 {
466         if (unlikely(!sched_info_on()))
467                 seq_printf(m, "0 0 0\n");
468         else
469                 seq_printf(m, "%llu %llu %lu\n",
470                    (unsigned long long)task->se.sum_exec_runtime,
471                    (unsigned long long)task->sched_info.run_delay,
472                    task->sched_info.pcount);
473 
474         return 0;
475 }
476 #endif
477 
478 #ifdef CONFIG_LATENCYTOP
479 static int lstats_show_proc(struct seq_file *m, void *v)
480 {
481         int i;
482         struct inode *inode = m->private;
483         struct task_struct *task = get_proc_task(inode);
484 
485         if (!task)
486                 return -ESRCH;
487         seq_puts(m, "Latency Top version : v0.1\n");
488         for (i = 0; i < 32; i++) {
489                 struct latency_record *lr = &task->latency_record[i];
490                 if (lr->backtrace[0]) {
491                         int q;
492                         seq_printf(m, "%i %li %li",
493                                    lr->count, lr->time, lr->max);
494                         for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
495                                 unsigned long bt = lr->backtrace[q];
496                                 if (!bt)
497                                         break;
498                                 if (bt == ULONG_MAX)
499                                         break;
500                                 seq_printf(m, " %ps", (void *)bt);
501                         }
502                         seq_putc(m, '\n');
503                 }
504 
505         }
506         put_task_struct(task);
507         return 0;
508 }
509 
510 static int lstats_open(struct inode *inode, struct file *file)
511 {
512         return single_open(file, lstats_show_proc, inode);
513 }
514 
515 static ssize_t lstats_write(struct file *file, const char __user *buf,
516                             size_t count, loff_t *offs)
517 {
518         struct task_struct *task = get_proc_task(file_inode(file));
519 
520         if (!task)
521                 return -ESRCH;
522         clear_all_latency_tracing(task);
523         put_task_struct(task);
524 
525         return count;
526 }
527 
528 static const struct file_operations proc_lstats_operations = {
529         .open           = lstats_open,
530         .read           = seq_read,
531         .write          = lstats_write,
532         .llseek         = seq_lseek,
533         .release        = single_release,
534 };
535 
536 #endif
537 
538 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
539                           struct pid *pid, struct task_struct *task)
540 {
541         unsigned long totalpages = totalram_pages + total_swap_pages;
542         unsigned long points = 0;
543 
544         points = oom_badness(task, NULL, NULL, totalpages) *
545                                         1000 / totalpages;
546         seq_printf(m, "%lu\n", points);
547 
548         return 0;
549 }
550 
551 struct limit_names {
552         const char *name;
553         const char *unit;
554 };
555 
556 static const struct limit_names lnames[RLIM_NLIMITS] = {
557         [RLIMIT_CPU] = {"Max cpu time", "seconds"},
558         [RLIMIT_FSIZE] = {"Max file size", "bytes"},
559         [RLIMIT_DATA] = {"Max data size", "bytes"},
560         [RLIMIT_STACK] = {"Max stack size", "bytes"},
561         [RLIMIT_CORE] = {"Max core file size", "bytes"},
562         [RLIMIT_RSS] = {"Max resident set", "bytes"},
563         [RLIMIT_NPROC] = {"Max processes", "processes"},
564         [RLIMIT_NOFILE] = {"Max open files", "files"},
565         [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
566         [RLIMIT_AS] = {"Max address space", "bytes"},
567         [RLIMIT_LOCKS] = {"Max file locks", "locks"},
568         [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
569         [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
570         [RLIMIT_NICE] = {"Max nice priority", NULL},
571         [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
572         [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
573 };
574 
575 /* Display limits for a process */
576 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
577                            struct pid *pid, struct task_struct *task)
578 {
579         unsigned int i;
580         unsigned long flags;
581 
582         struct rlimit rlim[RLIM_NLIMITS];
583 
584         if (!lock_task_sighand(task, &flags))
585                 return 0;
586         memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
587         unlock_task_sighand(task, &flags);
588 
589         /*
590          * print the file header
591          */
592        seq_printf(m, "%-25s %-20s %-20s %-10s\n",
593                   "Limit", "Soft Limit", "Hard Limit", "Units");
594 
595         for (i = 0; i < RLIM_NLIMITS; i++) {
596                 if (rlim[i].rlim_cur == RLIM_INFINITY)
597                         seq_printf(m, "%-25s %-20s ",
598                                    lnames[i].name, "unlimited");
599                 else
600                         seq_printf(m, "%-25s %-20lu ",
601                                    lnames[i].name, rlim[i].rlim_cur);
602 
603                 if (rlim[i].rlim_max == RLIM_INFINITY)
604                         seq_printf(m, "%-20s ", "unlimited");
605                 else
606                         seq_printf(m, "%-20lu ", rlim[i].rlim_max);
607 
608                 if (lnames[i].unit)
609                         seq_printf(m, "%-10s\n", lnames[i].unit);
610                 else
611                         seq_putc(m, '\n');
612         }
613 
614         return 0;
615 }
616 
617 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
618 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
619                             struct pid *pid, struct task_struct *task)
620 {
621         long nr;
622         unsigned long args[6], sp, pc;
623         int res;
624 
625         res = lock_trace(task);
626         if (res)
627                 return res;
628 
629         if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
630                 seq_puts(m, "running\n");
631         else if (nr < 0)
632                 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
633         else
634                 seq_printf(m,
635                        "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
636                        nr,
637                        args[0], args[1], args[2], args[3], args[4], args[5],
638                        sp, pc);
639         unlock_trace(task);
640 
641         return 0;
642 }
643 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
644 
645 /************************************************************************/
646 /*                       Here the fs part begins                        */
647 /************************************************************************/
648 
649 /* permission checks */
650 static int proc_fd_access_allowed(struct inode *inode)
651 {
652         struct task_struct *task;
653         int allowed = 0;
654         /* Allow access to a task's file descriptors if it is us or we
655          * may use ptrace attach to the process and find out that
656          * information.
657          */
658         task = get_proc_task(inode);
659         if (task) {
660                 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
661                 put_task_struct(task);
662         }
663         return allowed;
664 }
665 
666 int proc_setattr(struct dentry *dentry, struct iattr *attr)
667 {
668         int error;
669         struct inode *inode = d_inode(dentry);
670 
671         if (attr->ia_valid & ATTR_MODE)
672                 return -EPERM;
673 
674         error = setattr_prepare(dentry, attr);
675         if (error)
676                 return error;
677 
678         setattr_copy(inode, attr);
679         mark_inode_dirty(inode);
680         return 0;
681 }
682 
683 /*
684  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
685  * or euid/egid (for hide_pid_min=2)?
686  */
687 static bool has_pid_permissions(struct pid_namespace *pid,
688                                  struct task_struct *task,
689                                  int hide_pid_min)
690 {
691         if (pid->hide_pid < hide_pid_min)
692                 return true;
693         if (in_group_p(pid->pid_gid))
694                 return true;
695         return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
696 }
697 
698 
699 static int proc_pid_permission(struct inode *inode, int mask)
700 {
701         struct pid_namespace *pid = inode->i_sb->s_fs_info;
702         struct task_struct *task;
703         bool has_perms;
704 
705         task = get_proc_task(inode);
706         if (!task)
707                 return -ESRCH;
708         has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
709         put_task_struct(task);
710 
711         if (!has_perms) {
712                 if (pid->hide_pid == HIDEPID_INVISIBLE) {
713                         /*
714                          * Let's make getdents(), stat(), and open()
715                          * consistent with each other.  If a process
716                          * may not stat() a file, it shouldn't be seen
717                          * in procfs at all.
718                          */
719                         return -ENOENT;
720                 }
721 
722                 return -EPERM;
723         }
724         return generic_permission(inode, mask);
725 }
726 
727 
728 
729 static const struct inode_operations proc_def_inode_operations = {
730         .setattr        = proc_setattr,
731 };
732 
733 static int proc_single_show(struct seq_file *m, void *v)
734 {
735         struct inode *inode = m->private;
736         struct pid_namespace *ns;
737         struct pid *pid;
738         struct task_struct *task;
739         int ret;
740 
741         ns = inode->i_sb->s_fs_info;
742         pid = proc_pid(inode);
743         task = get_pid_task(pid, PIDTYPE_PID);
744         if (!task)
745                 return -ESRCH;
746 
747         ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
748 
749         put_task_struct(task);
750         return ret;
751 }
752 
753 static int proc_single_open(struct inode *inode, struct file *filp)
754 {
755         return single_open(filp, proc_single_show, inode);
756 }
757 
758 static const struct file_operations proc_single_file_operations = {
759         .open           = proc_single_open,
760         .read           = seq_read,
761         .llseek         = seq_lseek,
762         .release        = single_release,
763 };
764 
765 
766 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
767 {
768         struct task_struct *task = get_proc_task(inode);
769         struct mm_struct *mm = ERR_PTR(-ESRCH);
770 
771         if (task) {
772                 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
773                 put_task_struct(task);
774 
775                 if (!IS_ERR_OR_NULL(mm)) {
776                         /* ensure this mm_struct can't be freed */
777                         mmgrab(mm);
778                         /* but do not pin its memory */
779                         mmput(mm);
780                 }
781         }
782 
783         return mm;
784 }
785 
786 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
787 {
788         struct mm_struct *mm = proc_mem_open(inode, mode);
789 
790         if (IS_ERR(mm))
791                 return PTR_ERR(mm);
792 
793         file->private_data = mm;
794         return 0;
795 }
796 
797 static int mem_open(struct inode *inode, struct file *file)
798 {
799         int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
800 
801         /* OK to pass negative loff_t, we can catch out-of-range */
802         file->f_mode |= FMODE_UNSIGNED_OFFSET;
803 
804         return ret;
805 }
806 
807 static ssize_t mem_rw(struct file *file, char __user *buf,
808                         size_t count, loff_t *ppos, int write)
809 {
810         struct mm_struct *mm = file->private_data;
811         unsigned long addr = *ppos;
812         ssize_t copied;
813         char *page;
814         unsigned int flags;
815 
816         if (!mm)
817                 return 0;
818 
819         page = (char *)__get_free_page(GFP_KERNEL);
820         if (!page)
821                 return -ENOMEM;
822 
823         copied = 0;
824         if (!mmget_not_zero(mm))
825                 goto free;
826 
827         flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
828 
829         while (count > 0) {
830                 int this_len = min_t(int, count, PAGE_SIZE);
831 
832                 if (write && copy_from_user(page, buf, this_len)) {
833                         copied = -EFAULT;
834                         break;
835                 }
836 
837                 this_len = access_remote_vm(mm, addr, page, this_len, flags);
838                 if (!this_len) {
839                         if (!copied)
840                                 copied = -EIO;
841                         break;
842                 }
843 
844                 if (!write && copy_to_user(buf, page, this_len)) {
845                         copied = -EFAULT;
846                         break;
847                 }
848 
849                 buf += this_len;
850                 addr += this_len;
851                 copied += this_len;
852                 count -= this_len;
853         }
854         *ppos = addr;
855 
856         mmput(mm);
857 free:
858         free_page((unsigned long) page);
859         return copied;
860 }
861 
862 static ssize_t mem_read(struct file *file, char __user *buf,
863                         size_t count, loff_t *ppos)
864 {
865         return mem_rw(file, buf, count, ppos, 0);
866 }
867 
868 static ssize_t mem_write(struct file *file, const char __user *buf,
869                          size_t count, loff_t *ppos)
870 {
871         return mem_rw(file, (char __user*)buf, count, ppos, 1);
872 }
873 
874 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
875 {
876         switch (orig) {
877         case 0:
878                 file->f_pos = offset;
879                 break;
880         case 1:
881                 file->f_pos += offset;
882                 break;
883         default:
884                 return -EINVAL;
885         }
886         force_successful_syscall_return();
887         return file->f_pos;
888 }
889 
890 static int mem_release(struct inode *inode, struct file *file)
891 {
892         struct mm_struct *mm = file->private_data;
893         if (mm)
894                 mmdrop(mm);
895         return 0;
896 }
897 
898 static const struct file_operations proc_mem_operations = {
899         .llseek         = mem_lseek,
900         .read           = mem_read,
901         .write          = mem_write,
902         .open           = mem_open,
903         .release        = mem_release,
904 };
905 
906 static int environ_open(struct inode *inode, struct file *file)
907 {
908         return __mem_open(inode, file, PTRACE_MODE_READ);
909 }
910 
911 static ssize_t environ_read(struct file *file, char __user *buf,
912                         size_t count, loff_t *ppos)
913 {
914         char *page;
915         unsigned long src = *ppos;
916         int ret = 0;
917         struct mm_struct *mm = file->private_data;
918         unsigned long env_start, env_end;
919 
920         /* Ensure the process spawned far enough to have an environment. */
921         if (!mm || !mm->env_end)
922                 return 0;
923 
924         page = (char *)__get_free_page(GFP_KERNEL);
925         if (!page)
926                 return -ENOMEM;
927 
928         ret = 0;
929         if (!mmget_not_zero(mm))
930                 goto free;
931 
932         down_read(&mm->mmap_sem);
933         env_start = mm->env_start;
934         env_end = mm->env_end;
935         up_read(&mm->mmap_sem);
936 
937         while (count > 0) {
938                 size_t this_len, max_len;
939                 int retval;
940 
941                 if (src >= (env_end - env_start))
942                         break;
943 
944                 this_len = env_end - (env_start + src);
945 
946                 max_len = min_t(size_t, PAGE_SIZE, count);
947                 this_len = min(max_len, this_len);
948 
949                 retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
950 
951                 if (retval <= 0) {
952                         ret = retval;
953                         break;
954                 }
955 
956                 if (copy_to_user(buf, page, retval)) {
957                         ret = -EFAULT;
958                         break;
959                 }
960 
961                 ret += retval;
962                 src += retval;
963                 buf += retval;
964                 count -= retval;
965         }
966         *ppos = src;
967         mmput(mm);
968 
969 free:
970         free_page((unsigned long) page);
971         return ret;
972 }
973 
974 static const struct file_operations proc_environ_operations = {
975         .open           = environ_open,
976         .read           = environ_read,
977         .llseek         = generic_file_llseek,
978         .release        = mem_release,
979 };
980 
981 static int auxv_open(struct inode *inode, struct file *file)
982 {
983         return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
984 }
985 
986 static ssize_t auxv_read(struct file *file, char __user *buf,
987                         size_t count, loff_t *ppos)
988 {
989         struct mm_struct *mm = file->private_data;
990         unsigned int nwords = 0;
991 
992         if (!mm)
993                 return 0;
994         do {
995                 nwords += 2;
996         } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
997         return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
998                                        nwords * sizeof(mm->saved_auxv[0]));
999 }
1000 
1001 static const struct file_operations proc_auxv_operations = {
1002         .open           = auxv_open,
1003         .read           = auxv_read,
1004         .llseek         = generic_file_llseek,
1005         .release        = mem_release,
1006 };
1007 
1008 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1009                             loff_t *ppos)
1010 {
1011         struct task_struct *task = get_proc_task(file_inode(file));
1012         char buffer[PROC_NUMBUF];
1013         int oom_adj = OOM_ADJUST_MIN;
1014         size_t len;
1015 
1016         if (!task)
1017                 return -ESRCH;
1018         if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1019                 oom_adj = OOM_ADJUST_MAX;
1020         else
1021                 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1022                           OOM_SCORE_ADJ_MAX;
1023         put_task_struct(task);
1024         len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1025         return simple_read_from_buffer(buf, count, ppos, buffer, len);
1026 }
1027 
1028 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1029 {
1030         static DEFINE_MUTEX(oom_adj_mutex);
1031         struct mm_struct *mm = NULL;
1032         struct task_struct *task;
1033         int err = 0;
1034 
1035         task = get_proc_task(file_inode(file));
1036         if (!task)
1037                 return -ESRCH;
1038 
1039         mutex_lock(&oom_adj_mutex);
1040         if (legacy) {
1041                 if (oom_adj < task->signal->oom_score_adj &&
1042                                 !capable(CAP_SYS_RESOURCE)) {
1043                         err = -EACCES;
1044                         goto err_unlock;
1045                 }
1046                 /*
1047                  * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1048                  * /proc/pid/oom_score_adj instead.
1049                  */
1050                 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1051                           current->comm, task_pid_nr(current), task_pid_nr(task),
1052                           task_pid_nr(task));
1053         } else {
1054                 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1055                                 !capable(CAP_SYS_RESOURCE)) {
1056                         err = -EACCES;
1057                         goto err_unlock;
1058                 }
1059         }
1060 
1061         /*
1062          * Make sure we will check other processes sharing the mm if this is
1063          * not vfrok which wants its own oom_score_adj.
1064          * pin the mm so it doesn't go away and get reused after task_unlock
1065          */
1066         if (!task->vfork_done) {
1067                 struct task_struct *p = find_lock_task_mm(task);
1068 
1069                 if (p) {
1070                         if (atomic_read(&p->mm->mm_users) > 1) {
1071                                 mm = p->mm;
1072                                 mmgrab(mm);
1073                         }
1074                         task_unlock(p);
1075                 }
1076         }
1077 
1078         task->signal->oom_score_adj = oom_adj;
1079         if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1080                 task->signal->oom_score_adj_min = (short)oom_adj;
1081         trace_oom_score_adj_update(task);
1082 
1083         if (mm) {
1084                 struct task_struct *p;
1085 
1086                 rcu_read_lock();
1087                 for_each_process(p) {
1088                         if (same_thread_group(task, p))
1089                                 continue;
1090 
1091                         /* do not touch kernel threads or the global init */
1092                         if (p->flags & PF_KTHREAD || is_global_init(p))
1093                                 continue;
1094 
1095                         task_lock(p);
1096                         if (!p->vfork_done && process_shares_mm(p, mm)) {
1097                                 pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1098                                                 task_pid_nr(p), p->comm,
1099                                                 p->signal->oom_score_adj, oom_adj,
1100                                                 task_pid_nr(task), task->comm);
1101                                 p->signal->oom_score_adj = oom_adj;
1102                                 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1103                                         p->signal->oom_score_adj_min = (short)oom_adj;
1104                         }
1105                         task_unlock(p);
1106                 }
1107                 rcu_read_unlock();
1108                 mmdrop(mm);
1109         }
1110 err_unlock:
1111         mutex_unlock(&oom_adj_mutex);
1112         put_task_struct(task);
1113         return err;
1114 }
1115 
1116 /*
1117  * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1118  * kernels.  The effective policy is defined by oom_score_adj, which has a
1119  * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1120  * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1121  * Processes that become oom disabled via oom_adj will still be oom disabled
1122  * with this implementation.
1123  *
1124  * oom_adj cannot be removed since existing userspace binaries use it.
1125  */
1126 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1127                              size_t count, loff_t *ppos)
1128 {
1129         char buffer[PROC_NUMBUF];
1130         int oom_adj;
1131         int err;
1132 
1133         memset(buffer, 0, sizeof(buffer));
1134         if (count > sizeof(buffer) - 1)
1135                 count = sizeof(buffer) - 1;
1136         if (copy_from_user(buffer, buf, count)) {
1137                 err = -EFAULT;
1138                 goto out;
1139         }
1140 
1141         err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1142         if (err)
1143                 goto out;
1144         if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1145              oom_adj != OOM_DISABLE) {
1146                 err = -EINVAL;
1147                 goto out;
1148         }
1149 
1150         /*
1151          * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1152          * value is always attainable.
1153          */
1154         if (oom_adj == OOM_ADJUST_MAX)
1155                 oom_adj = OOM_SCORE_ADJ_MAX;
1156         else
1157                 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1158 
1159         err = __set_oom_adj(file, oom_adj, true);
1160 out:
1161         return err < 0 ? err : count;
1162 }
1163 
1164 static const struct file_operations proc_oom_adj_operations = {
1165         .read           = oom_adj_read,
1166         .write          = oom_adj_write,
1167         .llseek         = generic_file_llseek,
1168 };
1169 
1170 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1171                                         size_t count, loff_t *ppos)
1172 {
1173         struct task_struct *task = get_proc_task(file_inode(file));
1174         char buffer[PROC_NUMBUF];
1175         short oom_score_adj = OOM_SCORE_ADJ_MIN;
1176         size_t len;
1177 
1178         if (!task)
1179                 return -ESRCH;
1180         oom_score_adj = task->signal->oom_score_adj;
1181         put_task_struct(task);
1182         len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1183         return simple_read_from_buffer(buf, count, ppos, buffer, len);
1184 }
1185 
1186 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1187                                         size_t count, loff_t *ppos)
1188 {
1189         char buffer[PROC_NUMBUF];
1190         int oom_score_adj;
1191         int err;
1192 
1193         memset(buffer, 0, sizeof(buffer));
1194         if (count > sizeof(buffer) - 1)
1195                 count = sizeof(buffer) - 1;
1196         if (copy_from_user(buffer, buf, count)) {
1197                 err = -EFAULT;
1198                 goto out;
1199         }
1200 
1201         err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1202         if (err)
1203                 goto out;
1204         if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1205                         oom_score_adj > OOM_SCORE_ADJ_MAX) {
1206                 err = -EINVAL;
1207                 goto out;
1208         }
1209 
1210         err = __set_oom_adj(file, oom_score_adj, false);
1211 out:
1212         return err < 0 ? err : count;
1213 }
1214 
1215 static const struct file_operations proc_oom_score_adj_operations = {
1216         .read           = oom_score_adj_read,
1217         .write          = oom_score_adj_write,
1218         .llseek         = default_llseek,
1219 };
1220 
1221 #ifdef CONFIG_AUDITSYSCALL
1222 #define TMPBUFLEN 11
1223 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1224                                   size_t count, loff_t *ppos)
1225 {
1226         struct inode * inode = file_inode(file);
1227         struct task_struct *task = get_proc_task(inode);
1228         ssize_t length;
1229         char tmpbuf[TMPBUFLEN];
1230 
1231         if (!task)
1232                 return -ESRCH;
1233         length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1234                            from_kuid(file->f_cred->user_ns,
1235                                      audit_get_loginuid(task)));
1236         put_task_struct(task);
1237         return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1238 }
1239 
1240 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1241                                    size_t count, loff_t *ppos)
1242 {
1243         struct inode * inode = file_inode(file);
1244         uid_t loginuid;
1245         kuid_t kloginuid;
1246         int rv;
1247 
1248         rcu_read_lock();
1249         if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1250                 rcu_read_unlock();
1251                 return -EPERM;
1252         }
1253         rcu_read_unlock();
1254 
1255         if (*ppos != 0) {
1256                 /* No partial writes. */
1257                 return -EINVAL;
1258         }
1259 
1260         rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1261         if (rv < 0)
1262                 return rv;
1263 
1264         /* is userspace tring to explicitly UNSET the loginuid? */
1265         if (loginuid == AUDIT_UID_UNSET) {
1266                 kloginuid = INVALID_UID;
1267         } else {
1268                 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1269                 if (!uid_valid(kloginuid))
1270                         return -EINVAL;
1271         }
1272 
1273         rv = audit_set_loginuid(kloginuid);
1274         if (rv < 0)
1275                 return rv;
1276         return count;
1277 }
1278 
1279 static const struct file_operations proc_loginuid_operations = {
1280         .read           = proc_loginuid_read,
1281         .write          = proc_loginuid_write,
1282         .llseek         = generic_file_llseek,
1283 };
1284 
1285 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1286                                   size_t count, loff_t *ppos)
1287 {
1288         struct inode * inode = file_inode(file);
1289         struct task_struct *task = get_proc_task(inode);
1290         ssize_t length;
1291         char tmpbuf[TMPBUFLEN];
1292 
1293         if (!task)
1294                 return -ESRCH;
1295         length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1296                                 audit_get_sessionid(task));
1297         put_task_struct(task);
1298         return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1299 }
1300 
1301 static const struct file_operations proc_sessionid_operations = {
1302         .read           = proc_sessionid_read,
1303         .llseek         = generic_file_llseek,
1304 };
1305 #endif
1306 
1307 #ifdef CONFIG_FAULT_INJECTION
1308 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1309                                       size_t count, loff_t *ppos)
1310 {
1311         struct task_struct *task = get_proc_task(file_inode(file));
1312         char buffer[PROC_NUMBUF];
1313         size_t len;
1314         int make_it_fail;
1315 
1316         if (!task)
1317                 return -ESRCH;
1318         make_it_fail = task->make_it_fail;
1319         put_task_struct(task);
1320 
1321         len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1322 
1323         return simple_read_from_buffer(buf, count, ppos, buffer, len);
1324 }
1325 
1326 static ssize_t proc_fault_inject_write(struct file * file,
1327                         const char __user * buf, size_t count, loff_t *ppos)
1328 {
1329         struct task_struct *task;
1330         char buffer[PROC_NUMBUF];
1331         int make_it_fail;
1332         int rv;
1333 
1334         if (!capable(CAP_SYS_RESOURCE))
1335                 return -EPERM;
1336         memset(buffer, 0, sizeof(buffer));
1337         if (count > sizeof(buffer) - 1)
1338                 count = sizeof(buffer) - 1;
1339         if (copy_from_user(buffer, buf, count))
1340                 return -EFAULT;
1341         rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1342         if (rv < 0)
1343                 return rv;
1344         if (make_it_fail < 0 || make_it_fail > 1)
1345                 return -EINVAL;
1346 
1347         task = get_proc_task(file_inode(file));
1348         if (!task)
1349                 return -ESRCH;
1350         task->make_it_fail = make_it_fail;
1351         put_task_struct(task);
1352 
1353         return count;
1354 }
1355 
1356 static const struct file_operations proc_fault_inject_operations = {
1357         .read           = proc_fault_inject_read,
1358         .write          = proc_fault_inject_write,
1359         .llseek         = generic_file_llseek,
1360 };
1361 
1362 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1363                                    size_t count, loff_t *ppos)
1364 {
1365         struct task_struct *task;
1366         int err;
1367         unsigned int n;
1368 
1369         err = kstrtouint_from_user(buf, count, 0, &n);
1370         if (err)
1371                 return err;
1372 
1373         task = get_proc_task(file_inode(file));
1374         if (!task)
1375                 return -ESRCH;
1376         task->fail_nth = n;
1377         put_task_struct(task);
1378 
1379         return count;
1380 }
1381 
1382 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1383                                   size_t count, loff_t *ppos)
1384 {
1385         struct task_struct *task;
1386         char numbuf[PROC_NUMBUF];
1387         ssize_t len;
1388 
1389         task = get_proc_task(file_inode(file));
1390         if (!task)
1391                 return -ESRCH;
1392         len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1393         len = simple_read_from_buffer(buf, count, ppos, numbuf, len);
1394         put_task_struct(task);
1395 
1396         return len;
1397 }
1398 
1399 static const struct file_operations proc_fail_nth_operations = {
1400         .read           = proc_fail_nth_read,
1401         .write          = proc_fail_nth_write,
1402 };
1403 #endif
1404 
1405 
1406 #ifdef CONFIG_SCHED_DEBUG
1407 /*
1408  * Print out various scheduling related per-task fields:
1409  */
1410 static int sched_show(struct seq_file *m, void *v)
1411 {
1412         struct inode *inode = m->private;
1413         struct pid_namespace *ns = inode->i_sb->s_fs_info;
1414         struct task_struct *p;
1415 
1416         p = get_proc_task(inode);
1417         if (!p)
1418                 return -ESRCH;
1419         proc_sched_show_task(p, ns, m);
1420 
1421         put_task_struct(p);
1422 
1423         return 0;
1424 }
1425 
1426 static ssize_t
1427 sched_write(struct file *file, const char __user *buf,
1428             size_t count, loff_t *offset)
1429 {
1430         struct inode *inode = file_inode(file);
1431         struct task_struct *p;
1432 
1433         p = get_proc_task(inode);
1434         if (!p)
1435                 return -ESRCH;
1436         proc_sched_set_task(p);
1437 
1438         put_task_struct(p);
1439 
1440         return count;
1441 }
1442 
1443 static int sched_open(struct inode *inode, struct file *filp)
1444 {
1445         return single_open(filp, sched_show, inode);
1446 }
1447 
1448 static const struct file_operations proc_pid_sched_operations = {
1449         .open           = sched_open,
1450         .read           = seq_read,
1451         .write          = sched_write,
1452         .llseek         = seq_lseek,
1453         .release        = single_release,
1454 };
1455 
1456 #endif
1457 
1458 #ifdef CONFIG_SCHED_AUTOGROUP
1459 /*
1460  * Print out autogroup related information:
1461  */
1462 static int sched_autogroup_show(struct seq_file *m, void *v)
1463 {
1464         struct inode *inode = m->private;
1465         struct task_struct *p;
1466 
1467         p = get_proc_task(inode);
1468         if (!p)
1469                 return -ESRCH;
1470         proc_sched_autogroup_show_task(p, m);
1471 
1472         put_task_struct(p);
1473 
1474         return 0;
1475 }
1476 
1477 static ssize_t
1478 sched_autogroup_write(struct file *file, const char __user *buf,
1479             size_t count, loff_t *offset)
1480 {
1481         struct inode *inode = file_inode(file);
1482         struct task_struct *p;
1483         char buffer[PROC_NUMBUF];
1484         int nice;
1485         int err;
1486 
1487         memset(buffer, 0, sizeof(buffer));
1488         if (count > sizeof(buffer) - 1)
1489                 count = sizeof(buffer) - 1;
1490         if (copy_from_user(buffer, buf, count))
1491                 return -EFAULT;
1492 
1493         err = kstrtoint(strstrip(buffer), 0, &nice);
1494         if (err < 0)
1495                 return err;
1496 
1497         p = get_proc_task(inode);
1498         if (!p)
1499                 return -ESRCH;
1500 
1501         err = proc_sched_autogroup_set_nice(p, nice);
1502         if (err)
1503                 count = err;
1504 
1505         put_task_struct(p);
1506 
1507         return count;
1508 }
1509 
1510 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1511 {
1512         int ret;
1513 
1514         ret = single_open(filp, sched_autogroup_show, NULL);
1515         if (!ret) {
1516                 struct seq_file *m = filp->private_data;
1517 
1518                 m->private = inode;
1519         }
1520         return ret;
1521 }
1522 
1523 static const struct file_operations proc_pid_sched_autogroup_operations = {
1524         .open           = sched_autogroup_open,
1525         .read           = seq_read,
1526         .write          = sched_autogroup_write,
1527         .llseek         = seq_lseek,
1528         .release        = single_release,
1529 };
1530 
1531 #endif /* CONFIG_SCHED_AUTOGROUP */
1532 
1533 static ssize_t comm_write(struct file *file, const char __user *buf,
1534                                 size_t count, loff_t *offset)
1535 {
1536         struct inode *inode = file_inode(file);
1537         struct task_struct *p;
1538         char buffer[TASK_COMM_LEN];
1539         const size_t maxlen = sizeof(buffer) - 1;
1540 
1541         memset(buffer, 0, sizeof(buffer));
1542         if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1543                 return -EFAULT;
1544 
1545         p = get_proc_task(inode);
1546         if (!p)
1547                 return -ESRCH;
1548 
1549         if (same_thread_group(current, p))
1550                 set_task_comm(p, buffer);
1551         else
1552                 count = -EINVAL;
1553 
1554         put_task_struct(p);
1555 
1556         return count;
1557 }
1558 
1559 static int comm_show(struct seq_file *m, void *v)
1560 {
1561         struct inode *inode = m->private;
1562         struct task_struct *p;
1563 
1564         p = get_proc_task(inode);
1565         if (!p)
1566                 return -ESRCH;
1567 
1568         task_lock(p);
1569         seq_printf(m, "%s\n", p->comm);
1570         task_unlock(p);
1571 
1572         put_task_struct(p);
1573 
1574         return 0;
1575 }
1576 
1577 static int comm_open(struct inode *inode, struct file *filp)
1578 {
1579         return single_open(filp, comm_show, inode);
1580 }
1581 
1582 static const struct file_operations proc_pid_set_comm_operations = {
1583         .open           = comm_open,
1584         .read           = seq_read,
1585         .write          = comm_write,
1586         .llseek         = seq_lseek,
1587         .release        = single_release,
1588 };
1589 
1590 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1591 {
1592         struct task_struct *task;
1593         struct file *exe_file;
1594 
1595         task = get_proc_task(d_inode(dentry));
1596         if (!task)
1597                 return -ENOENT;
1598         exe_file = get_task_exe_file(task);
1599         put_task_struct(task);
1600         if (exe_file) {
1601                 *exe_path = exe_file->f_path;
1602                 path_get(&exe_file->f_path);
1603                 fput(exe_file);
1604                 return 0;
1605         } else
1606                 return -ENOENT;
1607 }
1608 
1609 static const char *proc_pid_get_link(struct dentry *dentry,
1610                                      struct inode *inode,
1611                                      struct delayed_call *done)
1612 {
1613         struct path path;
1614         int error = -EACCES;
1615 
1616         if (!dentry)
1617                 return ERR_PTR(-ECHILD);
1618 
1619         /* Are we allowed to snoop on the tasks file descriptors? */
1620         if (!proc_fd_access_allowed(inode))
1621                 goto out;
1622 
1623         error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1624         if (error)
1625                 goto out;
1626 
1627         nd_jump_link(&path);
1628         return NULL;
1629 out:
1630         return ERR_PTR(error);
1631 }
1632 
1633 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1634 {
1635         char *tmp = (char *)__get_free_page(GFP_KERNEL);
1636         char *pathname;
1637         int len;
1638 
1639         if (!tmp)
1640                 return -ENOMEM;
1641 
1642         pathname = d_path(path, tmp, PAGE_SIZE);
1643         len = PTR_ERR(pathname);
1644         if (IS_ERR(pathname))
1645                 goto out;
1646         len = tmp + PAGE_SIZE - 1 - pathname;
1647 
1648         if (len > buflen)
1649                 len = buflen;
1650         if (copy_to_user(buffer, pathname, len))
1651                 len = -EFAULT;
1652  out:
1653         free_page((unsigned long)tmp);
1654         return len;
1655 }
1656 
1657 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1658 {
1659         int error = -EACCES;
1660         struct inode *inode = d_inode(dentry);
1661         struct path path;
1662 
1663         /* Are we allowed to snoop on the tasks file descriptors? */
1664         if (!proc_fd_access_allowed(inode))
1665                 goto out;
1666 
1667         error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1668         if (error)
1669                 goto out;
1670 
1671         error = do_proc_readlink(&path, buffer, buflen);
1672         path_put(&path);
1673 out:
1674         return error;
1675 }
1676 
1677 const struct inode_operations proc_pid_link_inode_operations = {
1678         .readlink       = proc_pid_readlink,
1679         .get_link       = proc_pid_get_link,
1680         .setattr        = proc_setattr,
1681 };
1682 
1683 
1684 /* building an inode */
1685 
1686 void task_dump_owner(struct task_struct *task, umode_t mode,
1687                      kuid_t *ruid, kgid_t *rgid)
1688 {
1689         /* Depending on the state of dumpable compute who should own a
1690          * proc file for a task.
1691          */
1692         const struct cred *cred;
1693         kuid_t uid;
1694         kgid_t gid;
1695 
1696         if (unlikely(task->flags & PF_KTHREAD)) {
1697                 *ruid = GLOBAL_ROOT_UID;
1698                 *rgid = GLOBAL_ROOT_GID;
1699                 return;
1700         }
1701 
1702         /* Default to the tasks effective ownership */
1703         rcu_read_lock();
1704         cred = __task_cred(task);
1705         uid = cred->euid;
1706         gid = cred->egid;
1707         rcu_read_unlock();
1708 
1709         /*
1710          * Before the /proc/pid/status file was created the only way to read
1711          * the effective uid of a /process was to stat /proc/pid.  Reading
1712          * /proc/pid/status is slow enough that procps and other packages
1713          * kept stating /proc/pid.  To keep the rules in /proc simple I have
1714          * made this apply to all per process world readable and executable
1715          * directories.
1716          */
1717         if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1718                 struct mm_struct *mm;
1719                 task_lock(task);
1720                 mm = task->mm;
1721                 /* Make non-dumpable tasks owned by some root */
1722                 if (mm) {
1723                         if (get_dumpable(mm) != SUID_DUMP_USER) {
1724                                 struct user_namespace *user_ns = mm->user_ns;
1725 
1726                                 uid = make_kuid(user_ns, 0);
1727                                 if (!uid_valid(uid))
1728                                         uid = GLOBAL_ROOT_UID;
1729 
1730                                 gid = make_kgid(user_ns, 0);
1731                                 if (!gid_valid(gid))
1732                                         gid = GLOBAL_ROOT_GID;
1733                         }
1734                 } else {
1735                         uid = GLOBAL_ROOT_UID;
1736                         gid = GLOBAL_ROOT_GID;
1737                 }
1738                 task_unlock(task);
1739         }
1740         *ruid = uid;
1741         *rgid = gid;
1742 }
1743 
1744 struct inode *proc_pid_make_inode(struct super_block * sb,
1745                                   struct task_struct *task, umode_t mode)
1746 {
1747         struct inode * inode;
1748         struct proc_inode *ei;
1749 
1750         /* We need a new inode */
1751 
1752         inode = new_inode(sb);
1753         if (!inode)
1754                 goto out;
1755 
1756         /* Common stuff */
1757         ei = PROC_I(inode);
1758         inode->i_mode = mode;
1759         inode->i_ino = get_next_ino();
1760         inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1761         inode->i_op = &proc_def_inode_operations;
1762 
1763         /*
1764          * grab the reference to task.
1765          */
1766         ei->pid = get_task_pid(task, PIDTYPE_PID);
1767         if (!ei->pid)
1768                 goto out_unlock;
1769 
1770         task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1771         security_task_to_inode(task, inode);
1772 
1773 out:
1774         return inode;
1775 
1776 out_unlock:
1777         iput(inode);
1778         return NULL;
1779 }
1780 
1781 int pid_getattr(const struct path *path, struct kstat *stat,
1782                 u32 request_mask, unsigned int query_flags)
1783 {
1784         struct inode *inode = d_inode(path->dentry);
1785         struct task_struct *task;
1786         struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
1787 
1788         generic_fillattr(inode, stat);
1789 
1790         rcu_read_lock();
1791         stat->uid = GLOBAL_ROOT_UID;
1792         stat->gid = GLOBAL_ROOT_GID;
1793         task = pid_task(proc_pid(inode), PIDTYPE_PID);
1794         if (task) {
1795                 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
1796                         rcu_read_unlock();
1797                         /*
1798                          * This doesn't prevent learning whether PID exists,
1799                          * it only makes getattr() consistent with readdir().
1800                          */
1801                         return -ENOENT;
1802                 }
1803                 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1804         }
1805         rcu_read_unlock();
1806         return 0;
1807 }
1808 
1809 /* dentry stuff */
1810 
1811 /*
1812  *      Exceptional case: normally we are not allowed to unhash a busy
1813  * directory. In this case, however, we can do it - no aliasing problems
1814  * due to the way we treat inodes.
1815  *
1816  * Rewrite the inode's ownerships here because the owning task may have
1817  * performed a setuid(), etc.
1818  *
1819  */
1820 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1821 {
1822         struct inode *inode;
1823         struct task_struct *task;
1824 
1825         if (flags & LOOKUP_RCU)
1826                 return -ECHILD;
1827 
1828         inode = d_inode(dentry);
1829         task = get_proc_task(inode);
1830 
1831         if (task) {
1832                 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1833 
1834                 inode->i_mode &= ~(S_ISUID | S_ISGID);
1835                 security_task_to_inode(task, inode);
1836                 put_task_struct(task);
1837                 return 1;
1838         }
1839         return 0;
1840 }
1841 
1842 static inline bool proc_inode_is_dead(struct inode *inode)
1843 {
1844         return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1845 }
1846 
1847 int pid_delete_dentry(const struct dentry *dentry)
1848 {
1849         /* Is the task we represent dead?
1850          * If so, then don't put the dentry on the lru list,
1851          * kill it immediately.
1852          */
1853         return proc_inode_is_dead(d_inode(dentry));
1854 }
1855 
1856 const struct dentry_operations pid_dentry_operations =
1857 {
1858         .d_revalidate   = pid_revalidate,
1859         .d_delete       = pid_delete_dentry,
1860 };
1861 
1862 /* Lookups */
1863 
1864 /*
1865  * Fill a directory entry.
1866  *
1867  * If possible create the dcache entry and derive our inode number and
1868  * file type from dcache entry.
1869  *
1870  * Since all of the proc inode numbers are dynamically generated, the inode
1871  * numbers do not exist until the inode is cache.  This means creating the
1872  * the dcache entry in readdir is necessary to keep the inode numbers
1873  * reported by readdir in sync with the inode numbers reported
1874  * by stat.
1875  */
1876 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1877         const char *name, int len,
1878         instantiate_t instantiate, struct task_struct *task, const void *ptr)
1879 {
1880         struct dentry *child, *dir = file->f_path.dentry;
1881         struct qstr qname = QSTR_INIT(name, len);
1882         struct inode *inode;
1883         unsigned type;
1884         ino_t ino;
1885 
1886         child = d_hash_and_lookup(dir, &qname);
1887         if (!child) {
1888                 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1889                 child = d_alloc_parallel(dir, &qname, &wq);
1890                 if (IS_ERR(child))
1891                         goto end_instantiate;
1892                 if (d_in_lookup(child)) {
1893                         int err = instantiate(d_inode(dir), child, task, ptr);
1894                         d_lookup_done(child);
1895                         if (err < 0) {
1896                                 dput(child);
1897                                 goto end_instantiate;
1898                         }
1899                 }
1900         }
1901         inode = d_inode(child);
1902         ino = inode->i_ino;
1903         type = inode->i_mode >> 12;
1904         dput(child);
1905         return dir_emit(ctx, name, len, ino, type);
1906 
1907 end_instantiate:
1908         return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1909 }
1910 
1911 /*
1912  * dname_to_vma_addr - maps a dentry name into two unsigned longs
1913  * which represent vma start and end addresses.
1914  */
1915 static int dname_to_vma_addr(struct dentry *dentry,
1916                              unsigned long *start, unsigned long *end)
1917 {
1918         const char *str = dentry->d_name.name;
1919         unsigned long long sval, eval;
1920         unsigned int len;
1921 
1922         len = _parse_integer(str, 16, &sval);
1923         if (len & KSTRTOX_OVERFLOW)
1924                 return -EINVAL;
1925         if (sval != (unsigned long)sval)
1926                 return -EINVAL;
1927         str += len;
1928 
1929         if (*str != '-')
1930                 return -EINVAL;
1931         str++;
1932 
1933         len = _parse_integer(str, 16, &eval);
1934         if (len & KSTRTOX_OVERFLOW)
1935                 return -EINVAL;
1936         if (eval != (unsigned long)eval)
1937                 return -EINVAL;
1938         str += len;
1939 
1940         if (*str != '\0')
1941                 return -EINVAL;
1942 
1943         *start = sval;
1944         *end = eval;
1945 
1946         return 0;
1947 }
1948 
1949 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1950 {
1951         unsigned long vm_start, vm_end;
1952         bool exact_vma_exists = false;
1953         struct mm_struct *mm = NULL;
1954         struct task_struct *task;
1955         struct inode *inode;
1956         int status = 0;
1957 
1958         if (flags & LOOKUP_RCU)
1959                 return -ECHILD;
1960 
1961         inode = d_inode(dentry);
1962         task = get_proc_task(inode);
1963         if (!task)
1964                 goto out_notask;
1965 
1966         mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
1967         if (IS_ERR_OR_NULL(mm))
1968                 goto out;
1969 
1970         if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1971                 down_read(&mm->mmap_sem);
1972                 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1973                 up_read(&mm->mmap_sem);
1974         }
1975 
1976         mmput(mm);
1977 
1978         if (exact_vma_exists) {
1979                 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1980 
1981                 security_task_to_inode(task, inode);
1982                 status = 1;
1983         }
1984 
1985 out:
1986         put_task_struct(task);
1987 
1988 out_notask:
1989         return status;
1990 }
1991 
1992 static const struct dentry_operations tid_map_files_dentry_operations = {
1993         .d_revalidate   = map_files_d_revalidate,
1994         .d_delete       = pid_delete_dentry,
1995 };
1996 
1997 static int map_files_get_link(struct dentry *dentry, struct path *path)
1998 {
1999         unsigned long vm_start, vm_end;
2000         struct vm_area_struct *vma;
2001         struct task_struct *task;
2002         struct mm_struct *mm;
2003         int rc;
2004 
2005         rc = -ENOENT;
2006         task = get_proc_task(d_inode(dentry));
2007         if (!task)
2008                 goto out;
2009 
2010         mm = get_task_mm(task);
2011         put_task_struct(task);
2012         if (!mm)
2013                 goto out;
2014 
2015         rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2016         if (rc)
2017                 goto out_mmput;
2018 
2019         rc = -ENOENT;
2020         down_read(&mm->mmap_sem);
2021         vma = find_exact_vma(mm, vm_start, vm_end);
2022         if (vma && vma->vm_file) {
2023                 *path = vma->vm_file->f_path;
2024                 path_get(path);
2025                 rc = 0;
2026         }
2027         up_read(&mm->mmap_sem);
2028 
2029 out_mmput:
2030         mmput(mm);
2031 out:
2032         return rc;
2033 }
2034 
2035 struct map_files_info {
2036         unsigned long   start;
2037         unsigned long   end;
2038         fmode_t         mode;
2039 };
2040 
2041 /*
2042  * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2043  * symlinks may be used to bypass permissions on ancestor directories in the
2044  * path to the file in question.
2045  */
2046 static const char *
2047 proc_map_files_get_link(struct dentry *dentry,
2048                         struct inode *inode,
2049                         struct delayed_call *done)
2050 {
2051         if (!capable(CAP_SYS_ADMIN))
2052                 return ERR_PTR(-EPERM);
2053 
2054         return proc_pid_get_link(dentry, inode, done);
2055 }
2056 
2057 /*
2058  * Identical to proc_pid_link_inode_operations except for get_link()
2059  */
2060 static const struct inode_operations proc_map_files_link_inode_operations = {
2061         .readlink       = proc_pid_readlink,
2062         .get_link       = proc_map_files_get_link,
2063         .setattr        = proc_setattr,
2064 };
2065 
2066 static int
2067 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2068                            struct task_struct *task, const void *ptr)
2069 {
2070         fmode_t mode = (fmode_t)(unsigned long)ptr;
2071         struct proc_inode *ei;
2072         struct inode *inode;
2073 
2074         inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2075                                     ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2076                                     ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2077         if (!inode)
2078                 return -ENOENT;
2079 
2080         ei = PROC_I(inode);
2081         ei->op.proc_get_link = map_files_get_link;
2082 
2083         inode->i_op = &proc_map_files_link_inode_operations;
2084         inode->i_size = 64;
2085 
2086         d_set_d_op(dentry, &tid_map_files_dentry_operations);
2087         d_add(dentry, inode);
2088 
2089         return 0;
2090 }
2091 
2092 static struct dentry *proc_map_files_lookup(struct inode *dir,
2093                 struct dentry *dentry, unsigned int flags)
2094 {
2095         unsigned long vm_start, vm_end;
2096         struct vm_area_struct *vma;
2097         struct task_struct *task;
2098         int result;
2099         struct mm_struct *mm;
2100 
2101         result = -ENOENT;
2102         task = get_proc_task(dir);
2103         if (!task)
2104                 goto out;
2105 
2106         result = -EACCES;
2107         if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2108                 goto out_put_task;
2109 
2110         result = -ENOENT;
2111         if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2112                 goto out_put_task;
2113 
2114         mm = get_task_mm(task);
2115         if (!mm)
2116                 goto out_put_task;
2117 
2118         down_read(&mm->mmap_sem);
2119         vma = find_exact_vma(mm, vm_start, vm_end);
2120         if (!vma)
2121                 goto out_no_vma;
2122 
2123         if (vma->vm_file)
2124                 result = proc_map_files_instantiate(dir, dentry, task,
2125                                 (void *)(unsigned long)vma->vm_file->f_mode);
2126 
2127 out_no_vma:
2128         up_read(&mm->mmap_sem);
2129         mmput(mm);
2130 out_put_task:
2131         put_task_struct(task);
2132 out:
2133         return ERR_PTR(result);
2134 }
2135 
2136 static const struct inode_operations proc_map_files_inode_operations = {
2137         .lookup         = proc_map_files_lookup,
2138         .permission     = proc_fd_permission,
2139         .setattr        = proc_setattr,
2140 };
2141 
2142 static int
2143 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2144 {
2145         struct vm_area_struct *vma;
2146         struct task_struct *task;
2147         struct mm_struct *mm;
2148         unsigned long nr_files, pos, i;
2149         struct flex_array *fa = NULL;
2150         struct map_files_info info;
2151         struct map_files_info *p;
2152         int ret;
2153 
2154         ret = -ENOENT;
2155         task = get_proc_task(file_inode(file));
2156         if (!task)
2157                 goto out;
2158 
2159         ret = -EACCES;
2160         if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2161                 goto out_put_task;
2162 
2163         ret = 0;
2164         if (!dir_emit_dots(file, ctx))
2165                 goto out_put_task;
2166 
2167         mm = get_task_mm(task);
2168         if (!mm)
2169                 goto out_put_task;
2170         down_read(&mm->mmap_sem);
2171 
2172         nr_files = 0;
2173 
2174         /*
2175          * We need two passes here:
2176          *
2177          *  1) Collect vmas of mapped files with mmap_sem taken
2178          *  2) Release mmap_sem and instantiate entries
2179          *
2180          * otherwise we get lockdep complained, since filldir()
2181          * routine might require mmap_sem taken in might_fault().
2182          */
2183 
2184         for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2185                 if (vma->vm_file && ++pos > ctx->pos)
2186                         nr_files++;
2187         }
2188 
2189         if (nr_files) {
2190                 fa = flex_array_alloc(sizeof(info), nr_files,
2191                                         GFP_KERNEL);
2192                 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2193                                                 GFP_KERNEL)) {
2194                         ret = -ENOMEM;
2195                         if (fa)
2196                                 flex_array_free(fa);
2197                         up_read(&mm->mmap_sem);
2198                         mmput(mm);
2199                         goto out_put_task;
2200                 }
2201                 for (i = 0, vma = mm->mmap, pos = 2; vma;
2202                                 vma = vma->vm_next) {
2203                         if (!vma->vm_file)
2204                                 continue;
2205                         if (++pos <= ctx->pos)
2206                                 continue;
2207 
2208                         info.start = vma->vm_start;
2209                         info.end = vma->vm_end;
2210                         info.mode = vma->vm_file->f_mode;
2211                         if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2212                                 BUG();
2213                 }
2214         }
2215         up_read(&mm->mmap_sem);
2216 
2217         for (i = 0; i < nr_files; i++) {
2218                 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2219                 unsigned int len;
2220 
2221                 p = flex_array_get(fa, i);
2222                 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2223                 if (!proc_fill_cache(file, ctx,
2224                                       buf, len,
2225                                       proc_map_files_instantiate,
2226                                       task,
2227                                       (void *)(unsigned long)p->mode))
2228                         break;
2229                 ctx->pos++;
2230         }
2231         if (fa)
2232                 flex_array_free(fa);
2233         mmput(mm);
2234 
2235 out_put_task:
2236         put_task_struct(task);
2237 out:
2238         return ret;
2239 }
2240 
2241 static const struct file_operations proc_map_files_operations = {
2242         .read           = generic_read_dir,
2243         .iterate_shared = proc_map_files_readdir,
2244         .llseek         = generic_file_llseek,
2245 };
2246 
2247 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2248 struct timers_private {
2249         struct pid *pid;
2250         struct task_struct *task;
2251         struct sighand_struct *sighand;
2252         struct pid_namespace *ns;
2253         unsigned long flags;
2254 };
2255 
2256 static void *timers_start(struct seq_file *m, loff_t *pos)
2257 {
2258         struct timers_private *tp = m->private;
2259 
2260         tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2261         if (!tp->task)
2262                 return ERR_PTR(-ESRCH);
2263 
2264         tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2265         if (!tp->sighand)
2266                 return ERR_PTR(-ESRCH);
2267 
2268         return seq_list_start(&tp->task->signal->posix_timers, *pos);
2269 }
2270 
2271 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2272 {
2273         struct timers_private *tp = m->private;
2274         return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2275 }
2276 
2277 static void timers_stop(struct seq_file *m, void *v)
2278 {
2279         struct timers_private *tp = m->private;
2280 
2281         if (tp->sighand) {
2282                 unlock_task_sighand(tp->task, &tp->flags);
2283                 tp->sighand = NULL;
2284         }
2285 
2286         if (tp->task) {
2287                 put_task_struct(tp->task);
2288                 tp->task = NULL;
2289         }
2290 }
2291 
2292 static int show_timer(struct seq_file *m, void *v)
2293 {
2294         struct k_itimer *timer;
2295         struct timers_private *tp = m->private;
2296         int notify;
2297         static const char * const nstr[] = {
2298                 [SIGEV_SIGNAL] = "signal",
2299                 [SIGEV_NONE] = "none",
2300                 [SIGEV_THREAD] = "thread",
2301         };
2302 
2303         timer = list_entry((struct list_head *)v, struct k_itimer, list);
2304         notify = timer->it_sigev_notify;
2305 
2306         seq_printf(m, "ID: %d\n", timer->it_id);
2307         seq_printf(m, "signal: %d/%px\n",
2308                    timer->sigq->info.si_signo,
2309                    timer->sigq->info.si_value.sival_ptr);
2310         seq_printf(m, "notify: %s/%s.%d\n",
2311                    nstr[notify & ~SIGEV_THREAD_ID],
2312                    (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2313                    pid_nr_ns(timer->it_pid, tp->ns));
2314         seq_printf(m, "ClockID: %d\n", timer->it_clock);
2315 
2316         return 0;
2317 }
2318 
2319 static const struct seq_operations proc_timers_seq_ops = {
2320         .start  = timers_start,
2321         .next   = timers_next,
2322         .stop   = timers_stop,
2323         .show   = show_timer,
2324 };
2325 
2326 static int proc_timers_open(struct inode *inode, struct file *file)
2327 {
2328         struct timers_private *tp;
2329 
2330         tp = __seq_open_private(file, &proc_timers_seq_ops,
2331                         sizeof(struct timers_private));
2332         if (!tp)
2333                 return -ENOMEM;
2334 
2335         tp->pid = proc_pid(inode);
2336         tp->ns = inode->i_sb->s_fs_info;
2337         return 0;
2338 }
2339 
2340 static const struct file_operations proc_timers_operations = {
2341         .open           = proc_timers_open,
2342         .read           = seq_read,
2343         .llseek         = seq_lseek,
2344         .release        = seq_release_private,
2345 };
2346 #endif
2347 
2348 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2349                                         size_t count, loff_t *offset)
2350 {
2351         struct inode *inode = file_inode(file);
2352         struct task_struct *p;
2353         u64 slack_ns;
2354         int err;
2355 
2356         err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2357         if (err < 0)
2358                 return err;
2359 
2360         p = get_proc_task(inode);
2361         if (!p)
2362                 return -ESRCH;
2363 
2364         if (p != current) {
2365                 if (!capable(CAP_SYS_NICE)) {
2366                         count = -EPERM;
2367                         goto out;
2368                 }
2369 
2370                 err = security_task_setscheduler(p);
2371                 if (err) {
2372                         count = err;
2373                         goto out;
2374                 }
2375         }
2376 
2377         task_lock(p);
2378         if (slack_ns == 0)
2379                 p->timer_slack_ns = p->default_timer_slack_ns;
2380         else
2381                 p->timer_slack_ns = slack_ns;
2382         task_unlock(p);
2383 
2384 out:
2385         put_task_struct(p);
2386 
2387         return count;
2388 }
2389 
2390 static int timerslack_ns_show(struct seq_file *m, void *v)
2391 {
2392         struct inode *inode = m->private;
2393         struct task_struct *p;
2394         int err = 0;
2395 
2396         p = get_proc_task(inode);
2397         if (!p)
2398                 return -ESRCH;
2399 
2400         if (p != current) {
2401 
2402                 if (!capable(CAP_SYS_NICE)) {
2403                         err = -EPERM;
2404                         goto out;
2405                 }
2406                 err = security_task_getscheduler(p);
2407                 if (err)
2408                         goto out;
2409         }
2410 
2411         task_lock(p);
2412         seq_printf(m, "%llu\n", p->timer_slack_ns);
2413         task_unlock(p);
2414 
2415 out:
2416         put_task_struct(p);
2417 
2418         return err;
2419 }
2420 
2421 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2422 {
2423         return single_open(filp, timerslack_ns_show, inode);
2424 }
2425 
2426 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2427         .open           = timerslack_ns_open,
2428         .read           = seq_read,
2429         .write          = timerslack_ns_write,
2430         .llseek         = seq_lseek,
2431         .release        = single_release,
2432 };
2433 
2434 static int proc_pident_instantiate(struct inode *dir,
2435         struct dentry *dentry, struct task_struct *task, const void *ptr)
2436 {
2437         const struct pid_entry *p = ptr;
2438         struct inode *inode;
2439         struct proc_inode *ei;
2440 
2441         inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
2442         if (!inode)
2443                 goto out;
2444 
2445         ei = PROC_I(inode);
2446         if (S_ISDIR(inode->i_mode))
2447                 set_nlink(inode, 2);    /* Use getattr to fix if necessary */
2448         if (p->iop)
2449                 inode->i_op = p->iop;
2450         if (p->fop)
2451                 inode->i_fop = p->fop;
2452         ei->op = p->op;
2453         d_set_d_op(dentry, &pid_dentry_operations);
2454         d_add(dentry, inode);
2455         /* Close the race of the process dying before we return the dentry */
2456         if (pid_revalidate(dentry, 0))
2457                 return 0;
2458 out:
2459         return -ENOENT;
2460 }
2461 
2462 static struct dentry *proc_pident_lookup(struct inode *dir, 
2463                                          struct dentry *dentry,
2464                                          const struct pid_entry *ents,
2465                                          unsigned int nents)
2466 {
2467         int error;
2468         struct task_struct *task = get_proc_task(dir);
2469         const struct pid_entry *p, *last;
2470 
2471         error = -ENOENT;
2472 
2473         if (!task)
2474                 goto out_no_task;
2475 
2476         /*
2477          * Yes, it does not scale. And it should not. Don't add
2478          * new entries into /proc/<tgid>/ without very good reasons.
2479          */
2480         last = &ents[nents];
2481         for (p = ents; p < last; p++) {
2482                 if (p->len != dentry->d_name.len)
2483                         continue;
2484                 if (!memcmp(dentry->d_name.name, p->name, p->len))
2485                         break;
2486         }
2487         if (p >= last)
2488                 goto out;
2489 
2490         error = proc_pident_instantiate(dir, dentry, task, p);
2491 out:
2492         put_task_struct(task);
2493 out_no_task:
2494         return ERR_PTR(error);
2495 }
2496 
2497 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2498                 const struct pid_entry *ents, unsigned int nents)
2499 {
2500         struct task_struct *task = get_proc_task(file_inode(file));
2501         const struct pid_entry *p;
2502 
2503         if (!task)
2504                 return -ENOENT;
2505 
2506         if (!dir_emit_dots(file, ctx))
2507                 goto out;
2508 
2509         if (ctx->pos >= nents + 2)
2510                 goto out;
2511 
2512         for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2513                 if (!proc_fill_cache(file, ctx, p->name, p->len,
2514                                 proc_pident_instantiate, task, p))
2515                         break;
2516                 ctx->pos++;
2517         }
2518 out:
2519         put_task_struct(task);
2520         return 0;
2521 }
2522 
2523 #ifdef CONFIG_SECURITY
2524 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2525                                   size_t count, loff_t *ppos)
2526 {
2527         struct inode * inode = file_inode(file);
2528         char *p = NULL;
2529         ssize_t length;
2530         struct task_struct *task = get_proc_task(inode);
2531 
2532         if (!task)
2533                 return -ESRCH;
2534 
2535         length = security_getprocattr(task,
2536                                       (char*)file->f_path.dentry->d_name.name,
2537                                       &p);
2538         put_task_struct(task);
2539         if (length > 0)
2540                 length = simple_read_from_buffer(buf, count, ppos, p, length);
2541         kfree(p);
2542         return length;
2543 }
2544 
2545 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2546                                    size_t count, loff_t *ppos)
2547 {
2548         struct inode * inode = file_inode(file);
2549         void *page;
2550         ssize_t length;
2551         struct task_struct *task = get_proc_task(inode);
2552 
2553         length = -ESRCH;
2554         if (!task)
2555                 goto out_no_task;
2556 
2557         /* A task may only write its own attributes. */
2558         length = -EACCES;
2559         if (current != task)
2560                 goto out;
2561 
2562         if (count > PAGE_SIZE)
2563                 count = PAGE_SIZE;
2564 
2565         /* No partial writes. */
2566         length = -EINVAL;
2567         if (*ppos != 0)
2568                 goto out;
2569 
2570         page = memdup_user(buf, count);
2571         if (IS_ERR(page)) {
2572                 length = PTR_ERR(page);
2573                 goto out;
2574         }
2575 
2576         /* Guard against adverse ptrace interaction */
2577         length = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2578         if (length < 0)
2579                 goto out_free;
2580 
2581         length = security_setprocattr(file->f_path.dentry->d_name.name,
2582                                       page, count);
2583         mutex_unlock(&current->signal->cred_guard_mutex);
2584 out_free:
2585         kfree(page);
2586 out:
2587         put_task_struct(task);
2588 out_no_task:
2589         return length;
2590 }
2591 
2592 static const struct file_operations proc_pid_attr_operations = {
2593         .read           = proc_pid_attr_read,
2594         .write          = proc_pid_attr_write,
2595         .llseek         = generic_file_llseek,
2596 };
2597 
2598 static const struct pid_entry attr_dir_stuff[] = {
2599         REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2600         REG("prev",       S_IRUGO,         proc_pid_attr_operations),
2601         REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2602         REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2603         REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2604         REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2605 };
2606 
2607 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2608 {
2609         return proc_pident_readdir(file, ctx, 
2610                                    attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2611 }
2612 
2613 static const struct file_operations proc_attr_dir_operations = {
2614         .read           = generic_read_dir,
2615         .iterate_shared = proc_attr_dir_readdir,
2616         .llseek         = generic_file_llseek,
2617 };
2618 
2619 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2620                                 struct dentry *dentry, unsigned int flags)
2621 {
2622         return proc_pident_lookup(dir, dentry,
2623                                   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2624 }
2625 
2626 static const struct inode_operations proc_attr_dir_inode_operations = {
2627         .lookup         = proc_attr_dir_lookup,
2628         .getattr        = pid_getattr,
2629         .setattr        = proc_setattr,
2630 };
2631 
2632 #endif
2633 
2634 #ifdef CONFIG_ELF_CORE
2635 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2636                                          size_t count, loff_t *ppos)
2637 {
2638         struct task_struct *task = get_proc_task(file_inode(file));
2639         struct mm_struct *mm;
2640         char buffer[PROC_NUMBUF];
2641         size_t len;
2642         int ret;
2643 
2644         if (!task)
2645                 return -ESRCH;
2646 
2647         ret = 0;
2648         mm = get_task_mm(task);
2649         if (mm) {
2650                 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2651                                ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2652                                 MMF_DUMP_FILTER_SHIFT));
2653                 mmput(mm);
2654                 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2655         }
2656 
2657         put_task_struct(task);
2658 
2659         return ret;
2660 }
2661 
2662 static ssize_t proc_coredump_filter_write(struct file *file,
2663                                           const char __user *buf,
2664                                           size_t count,
2665                                           loff_t *ppos)
2666 {
2667         struct task_struct *task;
2668         struct mm_struct *mm;
2669         unsigned int val;
2670         int ret;
2671         int i;
2672         unsigned long mask;
2673 
2674         ret = kstrtouint_from_user(buf, count, 0, &val);
2675         if (ret < 0)
2676                 return ret;
2677 
2678         ret = -ESRCH;
2679         task = get_proc_task(file_inode(file));
2680         if (!task)
2681                 goto out_no_task;
2682 
2683         mm = get_task_mm(task);
2684         if (!mm)
2685                 goto out_no_mm;
2686         ret = 0;
2687 
2688         for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2689                 if (val & mask)
2690                         set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2691                 else
2692                         clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2693         }
2694 
2695         mmput(mm);
2696  out_no_mm:
2697         put_task_struct(task);
2698  out_no_task:
2699         if (ret < 0)
2700                 return ret;
2701         return count;
2702 }
2703 
2704 static const struct file_operations proc_coredump_filter_operations = {
2705         .read           = proc_coredump_filter_read,
2706         .write          = proc_coredump_filter_write,
2707         .llseek         = generic_file_llseek,
2708 };
2709 #endif
2710 
2711 #ifdef CONFIG_TASK_IO_ACCOUNTING
2712 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2713 {
2714         struct task_io_accounting acct = task->ioac;
2715         unsigned long flags;
2716         int result;
2717 
2718         result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2719         if (result)
2720                 return result;
2721 
2722         if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2723                 result = -EACCES;
2724                 goto out_unlock;
2725         }
2726 
2727         if (whole && lock_task_sighand(task, &flags)) {
2728                 struct task_struct *t = task;
2729 
2730                 task_io_accounting_add(&acct, &task->signal->ioac);
2731                 while_each_thread(task, t)
2732                         task_io_accounting_add(&acct, &t->ioac);
2733 
2734                 unlock_task_sighand(task, &flags);
2735         }
2736         seq_printf(m,
2737                    "rchar: %llu\n"
2738                    "wchar: %llu\n"
2739                    "syscr: %llu\n"
2740                    "syscw: %llu\n"
2741                    "read_bytes: %llu\n"
2742                    "write_bytes: %llu\n"
2743                    "cancelled_write_bytes: %llu\n",
2744                    (unsigned long long)acct.rchar,
2745                    (unsigned long long)acct.wchar,
2746                    (unsigned long long)acct.syscr,
2747                    (unsigned long long)acct.syscw,
2748                    (unsigned long long)acct.read_bytes,
2749                    (unsigned long long)acct.write_bytes,
2750                    (unsigned long long)acct.cancelled_write_bytes);
2751         result = 0;
2752 
2753 out_unlock:
2754         mutex_unlock(&task->signal->cred_guard_mutex);
2755         return result;
2756 }
2757 
2758 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2759                                   struct pid *pid, struct task_struct *task)
2760 {
2761         return do_io_accounting(task, m, 0);
2762 }
2763 
2764 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2765                                    struct pid *pid, struct task_struct *task)
2766 {
2767         return do_io_accounting(task, m, 1);
2768 }
2769 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2770 
2771 #ifdef CONFIG_USER_NS
2772 static int proc_id_map_open(struct inode *inode, struct file *file,
2773         const struct seq_operations *seq_ops)
2774 {
2775         struct user_namespace *ns = NULL;
2776         struct task_struct *task;
2777         struct seq_file *seq;
2778         int ret = -EINVAL;
2779 
2780         task = get_proc_task(inode);
2781         if (task) {
2782                 rcu_read_lock();
2783                 ns = get_user_ns(task_cred_xxx(task, user_ns));
2784                 rcu_read_unlock();
2785                 put_task_struct(task);
2786         }
2787         if (!ns)
2788                 goto err;
2789 
2790         ret = seq_open(file, seq_ops);
2791         if (ret)
2792                 goto err_put_ns;
2793 
2794         seq = file->private_data;
2795         seq->private = ns;
2796 
2797         return 0;
2798 err_put_ns:
2799         put_user_ns(ns);
2800 err:
2801         return ret;
2802 }
2803 
2804 static int proc_id_map_release(struct inode *inode, struct file *file)
2805 {
2806         struct seq_file *seq = file->private_data;
2807         struct user_namespace *ns = seq->private;
2808         put_user_ns(ns);
2809         return seq_release(inode, file);
2810 }
2811 
2812 static int proc_uid_map_open(struct inode *inode, struct file *file)
2813 {
2814         return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2815 }
2816 
2817 static int proc_gid_map_open(struct inode *inode, struct file *file)
2818 {
2819         return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2820 }
2821 
2822 static int proc_projid_map_open(struct inode *inode, struct file *file)
2823 {
2824         return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2825 }
2826 
2827 static const struct file_operations proc_uid_map_operations = {
2828         .open           = proc_uid_map_open,
2829         .write          = proc_uid_map_write,
2830         .read           = seq_read,
2831         .llseek         = seq_lseek,
2832         .release        = proc_id_map_release,
2833 };
2834 
2835 static const struct file_operations proc_gid_map_operations = {
2836         .open           = proc_gid_map_open,
2837         .write          = proc_gid_map_write,
2838         .read           = seq_read,
2839         .llseek         = seq_lseek,
2840         .release        = proc_id_map_release,
2841 };
2842 
2843 static const struct file_operations proc_projid_map_operations = {
2844         .open           = proc_projid_map_open,
2845         .write          = proc_projid_map_write,
2846         .read           = seq_read,
2847         .llseek         = seq_lseek,
2848         .release        = proc_id_map_release,
2849 };
2850 
2851 static int proc_setgroups_open(struct inode *inode, struct file *file)
2852 {
2853         struct user_namespace *ns = NULL;
2854         struct task_struct *task;
2855         int ret;
2856 
2857         ret = -ESRCH;
2858         task = get_proc_task(inode);
2859         if (task) {
2860                 rcu_read_lock();
2861                 ns = get_user_ns(task_cred_xxx(task, user_ns));
2862                 rcu_read_unlock();
2863                 put_task_struct(task);
2864         }
2865         if (!ns)
2866                 goto err;
2867 
2868         if (file->f_mode & FMODE_WRITE) {
2869                 ret = -EACCES;
2870                 if (!ns_capable(ns, CAP_SYS_ADMIN))
2871                         goto err_put_ns;
2872         }
2873 
2874         ret = single_open(file, &proc_setgroups_show, ns);
2875         if (ret)
2876                 goto err_put_ns;
2877 
2878         return 0;
2879 err_put_ns:
2880         put_user_ns(ns);
2881 err:
2882         return ret;
2883 }
2884 
2885 static int proc_setgroups_release(struct inode *inode, struct file *file)
2886 {
2887         struct seq_file *seq = file->private_data;
2888         struct user_namespace *ns = seq->private;
2889         int ret = single_release(inode, file);
2890         put_user_ns(ns);
2891         return ret;
2892 }
2893 
2894 static const struct file_operations proc_setgroups_operations = {
2895         .open           = proc_setgroups_open,
2896         .write          = proc_setgroups_write,
2897         .read           = seq_read,
2898         .llseek         = seq_lseek,
2899         .release        = proc_setgroups_release,
2900 };
2901 #endif /* CONFIG_USER_NS */
2902 
2903 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2904                                 struct pid *pid, struct task_struct *task)
2905 {
2906         int err = lock_trace(task);
2907         if (!err) {
2908                 seq_printf(m, "%08x\n", task->personality);
2909                 unlock_trace(task);
2910         }
2911         return err;
2912 }
2913 
2914 #ifdef CONFIG_LIVEPATCH
2915 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2916                                 struct pid *pid, struct task_struct *task)
2917 {
2918         seq_printf(m, "%d\n", task->patch_state);
2919         return 0;
2920 }
2921 #endif /* CONFIG_LIVEPATCH */
2922 
2923 /*
2924  * Thread groups
2925  */
2926 static const struct file_operations proc_task_operations;
2927 static const struct inode_operations proc_task_inode_operations;
2928 
2929 static const struct pid_entry tgid_base_stuff[] = {
2930         DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2931         DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2932         DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2933         DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2934         DIR("ns",         S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2935 #ifdef CONFIG_NET
2936         DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2937 #endif
2938         REG("environ",    S_IRUSR, proc_environ_operations),
2939         REG("auxv",       S_IRUSR, proc_auxv_operations),
2940         ONE("status",     S_IRUGO, proc_pid_status),
2941         ONE("personality", S_IRUSR, proc_pid_personality),
2942         ONE("limits",     S_IRUGO, proc_pid_limits),
2943 #ifdef CONFIG_SCHED_DEBUG
2944         REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2945 #endif
2946 #ifdef CONFIG_SCHED_AUTOGROUP
2947         REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2948 #endif
2949         REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2950 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2951         ONE("syscall",    S_IRUSR, proc_pid_syscall),
2952 #endif
2953         REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
2954         ONE("stat",       S_IRUGO, proc_tgid_stat),
2955         ONE("statm",      S_IRUGO, proc_pid_statm),
2956         REG("maps",       S_IRUGO, proc_pid_maps_operations),
2957 #ifdef CONFIG_NUMA
2958         REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
2959 #endif
2960         REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2961         LNK("cwd",        proc_cwd_link),
2962         LNK("root",       proc_root_link),
2963         LNK("exe",        proc_exe_link),
2964         REG("mounts",     S_IRUGO, proc_mounts_operations),
2965         REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2966         REG("mountstats", S_IRUSR, proc_mountstats_operations),
2967 #ifdef CONFIG_PROC_PAGE_MONITOR
2968         REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2969         REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
2970         REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
2971         REG("pagemap",    S_IRUSR, proc_pagemap_operations),
2972 #endif
2973 #ifdef CONFIG_SECURITY
2974         DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2975 #endif
2976 #ifdef CONFIG_KALLSYMS
2977         ONE("wchan",      S_IRUGO, proc_pid_wchan),
2978 #endif
2979 #ifdef CONFIG_STACKTRACE
2980         ONE("stack",      S_IRUSR, proc_pid_stack),
2981 #endif
2982 #ifdef CONFIG_SCHED_INFO
2983         ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
2984 #endif
2985 #ifdef CONFIG_LATENCYTOP
2986         REG("latency",  S_IRUGO, proc_lstats_operations),
2987 #endif
2988 #ifdef CONFIG_PROC_PID_CPUSET
2989         ONE("cpuset",     S_IRUGO, proc_cpuset_show),
2990 #endif
2991 #ifdef CONFIG_CGROUPS
2992         ONE("cgroup",  S_IRUGO, proc_cgroup_show),
2993 #endif
2994         ONE("oom_score",  S_IRUGO, proc_oom_score),
2995         REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2996         REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2997 #ifdef CONFIG_AUDITSYSCALL
2998         REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2999         REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3000 #endif
3001 #ifdef CONFIG_FAULT_INJECTION
3002         REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3003         REG("fail-nth", 0644, proc_fail_nth_operations),
3004 #endif
3005 #ifdef CONFIG_ELF_CORE
3006         REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3007 #endif
3008 #ifdef CONFIG_TASK_IO_ACCOUNTING
3009         ONE("io",       S_IRUSR, proc_tgid_io_accounting),
3010 #endif
3011 #ifdef CONFIG_HARDWALL
3012         ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
3013 #endif
3014 #ifdef CONFIG_USER_NS
3015         REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3016         REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3017         REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3018         REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3019 #endif
3020 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3021         REG("timers",     S_IRUGO, proc_timers_operations),
3022 #endif
3023         REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3024 #ifdef CONFIG_LIVEPATCH
3025         ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3026 #endif
3027 };
3028 
3029 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3030 {
3031         return proc_pident_readdir(file, ctx,
3032                                    tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3033 }
3034 
3035 static const struct file_operations proc_tgid_base_operations = {
3036         .read           = generic_read_dir,
3037         .iterate_shared = proc_tgid_base_readdir,
3038         .llseek         = generic_file_llseek,
3039 };
3040 
3041 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3042 {
3043         return proc_pident_lookup(dir, dentry,
3044                                   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3045 }
3046 
3047 static const struct inode_operations proc_tgid_base_inode_operations = {
3048         .lookup         = proc_tgid_base_lookup,
3049         .getattr        = pid_getattr,
3050         .setattr        = proc_setattr,
3051         .permission     = proc_pid_permission,
3052 };
3053 
3054 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3055 {
3056         struct dentry *dentry, *leader, *dir;
3057         char buf[10 + 1];
3058         struct qstr name;
3059 
3060         name.name = buf;
3061         name.len = snprintf(buf, sizeof(buf), "%u", pid);
3062         /* no ->d_hash() rejects on procfs */
3063         dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3064         if (dentry) {
3065                 d_invalidate(dentry);
3066                 dput(dentry);
3067         }
3068 
3069         if (pid == tgid)
3070                 return;
3071 
3072         name.name = buf;
3073         name.len = snprintf(buf, sizeof(buf), "%u", tgid);
3074         leader = d_hash_and_lookup(mnt->mnt_root, &name);
3075         if (!leader)
3076                 goto out;
3077 
3078         name.name = "task";
3079         name.len = strlen(name.name);
3080         dir = d_hash_and_lookup(leader, &name);
3081         if (!dir)
3082                 goto out_put_leader;
3083 
3084         name.name = buf;
3085         name.len = snprintf(buf, sizeof(buf), "%u", pid);
3086         dentry = d_hash_and_lookup(dir, &name);
3087         if (dentry) {
3088                 d_invalidate(dentry);
3089                 dput(dentry);
3090         }
3091 
3092         dput(dir);
3093 out_put_leader:
3094         dput(leader);
3095 out:
3096         return;
3097 }
3098 
3099 /**
3100  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
3101  * @task: task that should be flushed.
3102  *
3103  * When flushing dentries from proc, one needs to flush them from global
3104  * proc (proc_mnt) and from all the namespaces' procs this task was seen
3105  * in. This call is supposed to do all of this job.
3106  *
3107  * Looks in the dcache for
3108  * /proc/@pid
3109  * /proc/@tgid/task/@pid
3110  * if either directory is present flushes it and all of it'ts children
3111  * from the dcache.
3112  *
3113  * It is safe and reasonable to cache /proc entries for a task until
3114  * that task exits.  After that they just clog up the dcache with
3115  * useless entries, possibly causing useful dcache entries to be
3116  * flushed instead.  This routine is proved to flush those useless
3117  * dcache entries at process exit time.
3118  *
3119  * NOTE: This routine is just an optimization so it does not guarantee
3120  *       that no dcache entries will exist at process exit time it
3121  *       just makes it very unlikely that any will persist.
3122  */
3123 
3124 void proc_flush_task(struct task_struct *task)
3125 {
3126         int i;
3127         struct pid *pid, *tgid;
3128         struct upid *upid;
3129 
3130         pid = task_pid(task);
3131         tgid = task_tgid(task);
3132 
3133         for (i = 0; i <= pid->level; i++) {
3134                 upid = &pid->numbers[i];
3135                 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3136                                         tgid->numbers[i].nr);
3137         }
3138 }
3139 
3140 static int proc_pid_instantiate(struct inode *dir,
3141                                    struct dentry * dentry,
3142                                    struct task_struct *task, const void *ptr)
3143 {
3144         struct inode *inode;
3145 
3146         inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3147         if (!inode)
3148                 goto out;
3149 
3150         inode->i_op = &proc_tgid_base_inode_operations;
3151         inode->i_fop = &proc_tgid_base_operations;
3152         inode->i_flags|=S_IMMUTABLE;
3153 
3154         set_nlink(inode, nlink_tgid);
3155 
3156         d_set_d_op(dentry, &pid_dentry_operations);
3157 
3158         d_add(dentry, inode);
3159         /* Close the race of the process dying before we return the dentry */
3160         if (pid_revalidate(dentry, 0))
3161                 return 0;
3162 out:
3163         return -ENOENT;
3164 }
3165 
3166 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3167 {
3168         int result = -ENOENT;
3169         struct task_struct *task;
3170         unsigned tgid;
3171         struct pid_namespace *ns;
3172 
3173         tgid = name_to_int(&dentry->d_name);
3174         if (tgid == ~0U)
3175                 goto out;
3176 
3177         ns = dentry->d_sb->s_fs_info;
3178         rcu_read_lock();
3179         task = find_task_by_pid_ns(tgid, ns);
3180         if (task)
3181                 get_task_struct(task);
3182         rcu_read_unlock();
3183         if (!task)
3184                 goto out;
3185 
3186         result = proc_pid_instantiate(dir, dentry, task, NULL);
3187         put_task_struct(task);
3188 out:
3189         return ERR_PTR(result);
3190 }
3191 
3192 /*
3193  * Find the first task with tgid >= tgid
3194  *
3195  */
3196 struct tgid_iter {
3197         unsigned int tgid;
3198         struct task_struct *task;
3199 };
3200 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3201 {
3202         struct pid *pid;
3203 
3204         if (iter.task)
3205                 put_task_struct(iter.task);
3206         rcu_read_lock();
3207 retry:
3208         iter.task = NULL;
3209         pid = find_ge_pid(iter.tgid, ns);
3210         if (pid) {
3211                 iter.tgid = pid_nr_ns(pid, ns);
3212                 iter.task = pid_task(pid, PIDTYPE_PID);
3213                 /* What we to know is if the pid we have find is the
3214                  * pid of a thread_group_leader.  Testing for task
3215                  * being a thread_group_leader is the obvious thing
3216                  * todo but there is a window when it fails, due to
3217                  * the pid transfer logic in de_thread.
3218                  *
3219                  * So we perform the straight forward test of seeing
3220                  * if the pid we have found is the pid of a thread
3221                  * group leader, and don't worry if the task we have
3222                  * found doesn't happen to be a thread group leader.
3223                  * As we don't care in the case of readdir.
3224                  */
3225                 if (!iter.task || !has_group_leader_pid(iter.task)) {
3226                         iter.tgid += 1;
3227                         goto retry;
3228                 }
3229                 get_task_struct(iter.task);
3230         }
3231         rcu_read_unlock();
3232         return iter;
3233 }
3234 
3235 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3236 
3237 /* for the /proc/ directory itself, after non-process stuff has been done */
3238 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3239 {
3240         struct tgid_iter iter;
3241         struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3242         loff_t pos = ctx->pos;
3243 
3244         if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3245                 return 0;
3246 
3247         if (pos == TGID_OFFSET - 2) {
3248                 struct inode *inode = d_inode(ns->proc_self);
3249                 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3250                         return 0;
3251                 ctx->pos = pos = pos + 1;
3252         }
3253         if (pos == TGID_OFFSET - 1) {
3254                 struct inode *inode = d_inode(ns->proc_thread_self);
3255                 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3256                         return 0;
3257                 ctx->pos = pos = pos + 1;
3258         }
3259         iter.tgid = pos - TGID_OFFSET;
3260         iter.task = NULL;
3261         for (iter = next_tgid(ns, iter);
3262              iter.task;
3263              iter.tgid += 1, iter = next_tgid(ns, iter)) {
3264                 char name[10 + 1];
3265                 int len;
3266 
3267                 cond_resched();
3268                 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
3269                         continue;
3270 
3271                 len = snprintf(name, sizeof(name), "%u", iter.tgid);
3272                 ctx->pos = iter.tgid + TGID_OFFSET;
3273                 if (!proc_fill_cache(file, ctx, name, len,
3274                                      proc_pid_instantiate, iter.task, NULL)) {
3275                         put_task_struct(iter.task);
3276                         return 0;
3277                 }
3278         }
3279         ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3280         return 0;
3281 }
3282 
3283 /*
3284  * proc_tid_comm_permission is a special permission function exclusively
3285  * used for the node /proc/<pid>/task/<tid>/comm.
3286  * It bypasses generic permission checks in the case where a task of the same
3287  * task group attempts to access the node.
3288  * The rationale behind this is that glibc and bionic access this node for
3289  * cross thread naming (pthread_set/getname_np(!self)). However, if
3290  * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3291  * which locks out the cross thread naming implementation.
3292  * This function makes sure that the node is always accessible for members of
3293  * same thread group.
3294  */
3295 static int proc_tid_comm_permission(struct inode *inode, int mask)
3296 {
3297         bool is_same_tgroup;
3298         struct task_struct *task;
3299 
3300         task = get_proc_task(inode);
3301         if (!task)
3302                 return -ESRCH;
3303         is_same_tgroup = same_thread_group(current, task);
3304         put_task_struct(task);
3305 
3306         if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3307                 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3308                  * read or written by the members of the corresponding
3309                  * thread group.
3310                  */
3311                 return 0;
3312         }
3313 
3314         return generic_permission(inode, mask);
3315 }
3316 
3317 static const struct inode_operations proc_tid_comm_inode_operations = {
3318                 .permission = proc_tid_comm_permission,
3319 };
3320 
3321 /*
3322  * Tasks
3323  */
3324 static const struct pid_entry tid_base_stuff[] = {
3325         DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3326         DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3327         DIR("ns",        S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3328 #ifdef CONFIG_NET
3329         DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3330 #endif
3331         REG("environ",   S_IRUSR, proc_environ_operations),
3332         REG("auxv",      S_IRUSR, proc_auxv_operations),
3333         ONE("status",    S_IRUGO, proc_pid_status),
3334         ONE("personality", S_IRUSR, proc_pid_personality),
3335         ONE("limits",    S_IRUGO, proc_pid_limits),
3336 #ifdef CONFIG_SCHED_DEBUG
3337         REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3338 #endif
3339         NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3340                          &proc_tid_comm_inode_operations,
3341                          &proc_pid_set_comm_operations, {}),
3342 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3343         ONE("syscall",   S_IRUSR, proc_pid_syscall),
3344 #endif
3345         REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3346         ONE("stat",      S_IRUGO, proc_tid_stat),
3347         ONE("statm",     S_IRUGO, proc_pid_statm),
3348         REG("maps",      S_IRUGO, proc_tid_maps_operations),
3349 #ifdef CONFIG_PROC_CHILDREN
3350         REG("children",  S_IRUGO, proc_tid_children_operations),
3351 #endif
3352 #ifdef CONFIG_NUMA
3353         REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3354 #endif
3355         REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3356         LNK("cwd",       proc_cwd_link),
3357         LNK("root",      proc_root_link),
3358         LNK("exe",       proc_exe_link),
3359         REG("mounts",    S_IRUGO, proc_mounts_operations),
3360         REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3361 #ifdef CONFIG_PROC_PAGE_MONITOR
3362         REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3363         REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
3364         REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3365         REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3366 #endif
3367 #ifdef CONFIG_SECURITY
3368         DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3369 #endif
3370 #ifdef CONFIG_KALLSYMS
3371         ONE("wchan",     S_IRUGO, proc_pid_wchan),
3372 #endif
3373 #ifdef CONFIG_STACKTRACE
3374         ONE("stack",      S_IRUSR, proc_pid_stack),
3375 #endif
3376 #ifdef CONFIG_SCHED_INFO
3377         ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3378 #endif
3379 #ifdef CONFIG_LATENCYTOP
3380         REG("latency",  S_IRUGO, proc_lstats_operations),
3381 #endif
3382 #ifdef CONFIG_PROC_PID_CPUSET
3383         ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3384 #endif
3385 #ifdef CONFIG_CGROUPS
3386         ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3387 #endif
3388         ONE("oom_score", S_IRUGO, proc_oom_score),
3389         REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3390         REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3391 #ifdef CONFIG_AUDITSYSCALL
3392         REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3393         REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3394 #endif
3395 #ifdef CONFIG_FAULT_INJECTION
3396         REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3397         REG("fail-nth", 0644, proc_fail_nth_operations),
3398 #endif
3399 #ifdef CONFIG_TASK_IO_ACCOUNTING
3400         ONE("io",       S_IRUSR, proc_tid_io_accounting),
3401 #endif
3402 #ifdef CONFIG_HARDWALL
3403         ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
3404 #endif
3405 #ifdef CONFIG_USER_NS
3406         REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3407         REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3408         REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3409         REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3410 #endif
3411 #ifdef CONFIG_LIVEPATCH
3412         ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3413 #endif
3414 };
3415 
3416 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3417 {
3418         return proc_pident_readdir(file, ctx,
3419                                    tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3420 }
3421 
3422 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3423 {
3424         return proc_pident_lookup(dir, dentry,
3425                                   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3426 }
3427 
3428 static const struct file_operations proc_tid_base_operations = {
3429         .read           = generic_read_dir,
3430         .iterate_shared = proc_tid_base_readdir,
3431         .llseek         = generic_file_llseek,
3432 };
3433 
3434 static const struct inode_operations proc_tid_base_inode_operations = {
3435         .lookup         = proc_tid_base_lookup,
3436         .getattr        = pid_getattr,
3437         .setattr        = proc_setattr,
3438 };
3439 
3440 static int proc_task_instantiate(struct inode *dir,
3441         struct dentry *dentry, struct task_struct *task, const void *ptr)
3442 {
3443         struct inode *inode;
3444         inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3445 
3446         if (!inode)
3447                 goto out;
3448         inode->i_op = &proc_tid_base_inode_operations;
3449         inode->i_fop = &proc_tid_base_operations;
3450         inode->i_flags|=S_IMMUTABLE;
3451 
3452         set_nlink(inode, nlink_tid);
3453 
3454         d_set_d_op(dentry, &pid_dentry_operations);
3455 
3456         d_add(dentry, inode);
3457         /* Close the race of the process dying before we return the dentry */
3458         if (pid_revalidate(dentry, 0))
3459                 return 0;
3460 out:
3461         return -ENOENT;
3462 }
3463 
3464 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3465 {
3466         int result = -ENOENT;
3467         struct task_struct *task;
3468         struct task_struct *leader = get_proc_task(dir);
3469         unsigned tid;
3470         struct pid_namespace *ns;
3471 
3472         if (!leader)
3473                 goto out_no_task;
3474 
3475         tid = name_to_int(&dentry->d_name);
3476         if (tid == ~0U)
3477                 goto out;
3478 
3479         ns = dentry->d_sb->s_fs_info;
3480         rcu_read_lock();
3481         task = find_task_by_pid_ns(tid, ns);
3482         if (task)
3483                 get_task_struct(task);
3484         rcu_read_unlock();
3485         if (!task)
3486                 goto out;
3487         if (!same_thread_group(leader, task))
3488                 goto out_drop_task;
3489 
3490         result = proc_task_instantiate(dir, dentry, task, NULL);
3491 out_drop_task:
3492         put_task_struct(task);
3493 out:
3494         put_task_struct(leader);
3495 out_no_task:
3496         return ERR_PTR(result);
3497 }
3498 
3499 /*
3500  * Find the first tid of a thread group to return to user space.
3501  *
3502  * Usually this is just the thread group leader, but if the users
3503  * buffer was too small or there was a seek into the middle of the
3504  * directory we have more work todo.
3505  *
3506  * In the case of a short read we start with find_task_by_pid.
3507  *
3508  * In the case of a seek we start with the leader and walk nr
3509  * threads past it.
3510  */
3511 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3512                                         struct pid_namespace *ns)
3513 {
3514         struct task_struct *pos, *task;
3515         unsigned long nr = f_pos;
3516 
3517         if (nr != f_pos)        /* 32bit overflow? */
3518                 return NULL;
3519 
3520         rcu_read_lock();
3521         task = pid_task(pid, PIDTYPE_PID);
3522         if (!task)
3523                 goto fail;
3524 
3525         /* Attempt to start with the tid of a thread */
3526         if (tid && nr) {
3527                 pos = find_task_by_pid_ns(tid, ns);
3528                 if (pos && same_thread_group(pos, task))
3529                         goto found;
3530         }
3531 
3532         /* If nr exceeds the number of threads there is nothing todo */
3533         if (nr >= get_nr_threads(task))
3534                 goto fail;
3535 
3536         /* If we haven't found our starting place yet start
3537          * with the leader and walk nr threads forward.
3538          */
3539         pos = task = task->group_leader;
3540         do {
3541                 if (!nr--)
3542                         goto found;
3543         } while_each_thread(task, pos);
3544 fail:
3545         pos = NULL;
3546         goto out;
3547 found:
3548         get_task_struct(pos);
3549 out:
3550         rcu_read_unlock();
3551         return pos;
3552 }
3553 
3554 /*
3555  * Find the next thread in the thread list.
3556  * Return NULL if there is an error or no next thread.
3557  *
3558  * The reference to the input task_struct is released.
3559  */
3560 static struct task_struct *next_tid(struct task_struct *start)
3561 {
3562         struct task_struct *pos = NULL;
3563         rcu_read_lock();
3564         if (pid_alive(start)) {
3565                 pos = next_thread(start);
3566                 if (thread_group_leader(pos))
3567                         pos = NULL;
3568                 else
3569                         get_task_struct(pos);
3570         }
3571         rcu_read_unlock();
3572         put_task_struct(start);
3573         return pos;
3574 }
3575 
3576 /* for the /proc/TGID/task/ directories */
3577 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3578 {
3579         struct inode *inode = file_inode(file);
3580         struct task_struct *task;
3581         struct pid_namespace *ns;
3582         int tid;
3583 
3584         if (proc_inode_is_dead(inode))
3585                 return -ENOENT;
3586 
3587         if (!dir_emit_dots(file, ctx))
3588                 return 0;
3589 
3590         /* f_version caches the tgid value that the last readdir call couldn't
3591          * return. lseek aka telldir automagically resets f_version to 0.
3592          */
3593         ns = inode->i_sb->s_fs_info;
3594         tid = (int)file->f_version;
3595         file->f_version = 0;
3596         for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3597              task;
3598              task = next_tid(task), ctx->pos++) {
3599                 char name[10 + 1];
3600                 int len;
3601                 tid = task_pid_nr_ns(task, ns);
3602                 len = snprintf(name, sizeof(name), "%u", tid);
3603                 if (!proc_fill_cache(file, ctx, name, len,
3604                                 proc_task_instantiate, task, NULL)) {
3605                         /* returning this tgid failed, save it as the first
3606                          * pid for the next readir call */
3607                         file->f_version = (u64)tid;
3608                         put_task_struct(task);
3609                         break;
3610                 }
3611         }
3612 
3613         return 0;
3614 }
3615 
3616 static int proc_task_getattr(const struct path *path, struct kstat *stat,
3617                              u32 request_mask, unsigned int query_flags)
3618 {
3619         struct inode *inode = d_inode(path->dentry);
3620         struct task_struct *p = get_proc_task(inode);
3621         generic_fillattr(inode, stat);
3622 
3623         if (p) {
3624                 stat->nlink += get_nr_threads(p);
3625                 put_task_struct(p);
3626         }
3627 
3628         return 0;
3629 }
3630 
3631 static const struct inode_operations proc_task_inode_operations = {
3632         .lookup         = proc_task_lookup,
3633         .getattr        = proc_task_getattr,
3634         .setattr        = proc_setattr,
3635         .permission     = proc_pid_permission,
3636 };
3637 
3638 static const struct file_operations proc_task_operations = {
3639         .read           = generic_read_dir,
3640         .iterate_shared = proc_task_readdir,
3641         .llseek         = generic_file_llseek,
3642 };
3643 
3644 void __init set_proc_pid_nlink(void)
3645 {
3646         nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3647         nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3648 }
3649 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp