4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/config.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/string.h>
61 #include <linux/seq_file.h>
62 #include <linux/namei.h>
63 #include <linux/namespace.h>
65 #include <linux/smp_lock.h>
66 #include <linux/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/mount.h>
69 #include <linux/security.h>
70 #include <linux/ptrace.h>
71 #include <linux/seccomp.h>
72 #include <linux/cpuset.h>
73 #include <linux/audit.h>
74 #include <linux/poll.h>
78 * Implementing inode permission operations in /proc is almost
79 * certainly an error. Permission checks need to happen during
80 * each system call not at open time. The reason is that most of
81 * what we wish to check for permissions in /proc varies at runtime.
83 * The classic example of a problem is opening file descriptors
84 * in /proc for a task before it execs a suid executable.
88 * For hysterical raisins we keep the same inumbers as in the old procfs.
89 * Feel free to change the macro below - just keep the range distinct from
90 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
91 * As soon as we'll get a separate superblock we will be able to forget
92 * about magical ranges too.
95 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
97 enum pid_directory_inos {
102 #ifdef CONFIG_SECCOMP
117 PROC_TGID_MOUNTSTATS,
122 #ifdef CONFIG_SCHEDSTATS
125 #ifdef CONFIG_CPUSETS
128 #ifdef CONFIG_SECURITY
130 PROC_TGID_ATTR_CURRENT,
133 PROC_TGID_ATTR_FSCREATE,
134 PROC_TGID_ATTR_KEYCREATE,
136 #ifdef CONFIG_AUDITSYSCALL
140 PROC_TGID_OOM_ADJUST,
144 #ifdef CONFIG_SECCOMP
164 #ifdef CONFIG_SCHEDSTATS
167 #ifdef CONFIG_CPUSETS
170 #ifdef CONFIG_SECURITY
172 PROC_TID_ATTR_CURRENT,
175 PROC_TID_ATTR_FSCREATE,
176 PROC_TID_ATTR_KEYCREATE,
178 #ifdef CONFIG_AUDITSYSCALL
184 /* Add new entries before this */
185 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
188 /* Worst case buffer size needed for holding an integer. */
189 #define PROC_NUMBUF 10
198 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
200 static struct pid_entry tgid_base_stuff[] = {
201 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
202 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
203 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
204 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
205 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
206 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
207 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
208 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
209 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
211 E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
213 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
214 #ifdef CONFIG_SECCOMP
215 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
217 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
218 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
219 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
220 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
221 E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
223 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
225 #ifdef CONFIG_SECURITY
226 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
228 #ifdef CONFIG_KALLSYMS
229 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
231 #ifdef CONFIG_SCHEDSTATS
232 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
234 #ifdef CONFIG_CPUSETS
235 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
237 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
238 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
239 #ifdef CONFIG_AUDITSYSCALL
240 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
244 static struct pid_entry tid_base_stuff[] = {
245 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
246 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
247 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
248 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
249 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
250 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
251 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
252 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
254 E(PROC_TID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
256 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
257 #ifdef CONFIG_SECCOMP
258 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
260 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
261 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
262 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
263 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
265 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
267 #ifdef CONFIG_SECURITY
268 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
270 #ifdef CONFIG_KALLSYMS
271 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
273 #ifdef CONFIG_SCHEDSTATS
274 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
276 #ifdef CONFIG_CPUSETS
277 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
279 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
280 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
281 #ifdef CONFIG_AUDITSYSCALL
282 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
287 #ifdef CONFIG_SECURITY
288 static struct pid_entry tgid_attr_stuff[] = {
289 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
290 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
291 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
292 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
293 E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
296 static struct pid_entry tid_attr_stuff[] = {
297 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
298 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
299 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
300 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
301 E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
308 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
310 struct task_struct *task = get_proc_task(inode);
311 struct files_struct *files = NULL;
313 int fd = proc_fd(inode);
316 files = get_files_struct(task);
317 put_task_struct(task);
321 * We are not taking a ref to the file structure, so we must
324 spin_lock(&files->file_lock);
325 file = fcheck_files(files, fd);
327 *mnt = mntget(file->f_vfsmnt);
328 *dentry = dget(file->f_dentry);
329 spin_unlock(&files->file_lock);
330 put_files_struct(files);
333 spin_unlock(&files->file_lock);
334 put_files_struct(files);
339 static struct fs_struct *get_fs_struct(struct task_struct *task)
341 struct fs_struct *fs;
345 atomic_inc(&fs->count);
350 static int get_nr_threads(struct task_struct *tsk)
352 /* Must be called with the rcu_read_lock held */
356 if (lock_task_sighand(tsk, &flags)) {
357 count = atomic_read(&tsk->signal->count);
358 unlock_task_sighand(tsk, &flags);
363 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
365 struct task_struct *task = get_proc_task(inode);
366 struct fs_struct *fs = NULL;
367 int result = -ENOENT;
370 fs = get_fs_struct(task);
371 put_task_struct(task);
374 read_lock(&fs->lock);
375 *mnt = mntget(fs->pwdmnt);
376 *dentry = dget(fs->pwd);
377 read_unlock(&fs->lock);
384 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
386 struct task_struct *task = get_proc_task(inode);
387 struct fs_struct *fs = NULL;
388 int result = -ENOENT;
391 fs = get_fs_struct(task);
392 put_task_struct(task);
395 read_lock(&fs->lock);
396 *mnt = mntget(fs->rootmnt);
397 *dentry = dget(fs->root);
398 read_unlock(&fs->lock);
405 #define MAY_PTRACE(task) \
406 (task == current || \
407 (task->parent == current && \
408 (task->ptrace & PT_PTRACED) && \
409 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
410 security_ptrace(current,task) == 0))
412 static int proc_pid_environ(struct task_struct *task, char * buffer)
415 struct mm_struct *mm = get_task_mm(task);
417 unsigned int len = mm->env_end - mm->env_start;
420 res = access_process_vm(task, mm->env_start, buffer, len, 0);
421 if (!ptrace_may_attach(task))
428 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
432 struct mm_struct *mm = get_task_mm(task);
436 goto out_mm; /* Shh! No looking before we're done */
438 len = mm->arg_end - mm->arg_start;
443 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
445 // If the nul at the end of args has been overwritten, then
446 // assume application is using setproctitle(3).
447 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
448 len = strnlen(buffer, res);
452 len = mm->env_end - mm->env_start;
453 if (len > PAGE_SIZE - res)
454 len = PAGE_SIZE - res;
455 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
456 res = strnlen(buffer, res);
465 static int proc_pid_auxv(struct task_struct *task, char *buffer)
468 struct mm_struct *mm = get_task_mm(task);
470 unsigned int nwords = 0;
473 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
474 res = nwords * sizeof(mm->saved_auxv[0]);
477 memcpy(buffer, mm->saved_auxv, res);
484 #ifdef CONFIG_KALLSYMS
486 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
487 * Returns the resolved symbol. If that fails, simply return the address.
489 static int proc_pid_wchan(struct task_struct *task, char *buffer)
492 const char *sym_name;
493 unsigned long wchan, size, offset;
494 char namebuf[KSYM_NAME_LEN+1];
496 wchan = get_wchan(task);
498 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
500 return sprintf(buffer, "%s", sym_name);
501 return sprintf(buffer, "%lu", wchan);
503 #endif /* CONFIG_KALLSYMS */
505 #ifdef CONFIG_SCHEDSTATS
507 * Provides /proc/PID/schedstat
509 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
511 return sprintf(buffer, "%lu %lu %lu\n",
512 task->sched_info.cpu_time,
513 task->sched_info.run_delay,
514 task->sched_info.pcnt);
518 /* The badness from the OOM killer */
519 unsigned long badness(struct task_struct *p, unsigned long uptime);
520 static int proc_oom_score(struct task_struct *task, char *buffer)
522 unsigned long points;
523 struct timespec uptime;
525 do_posix_clock_monotonic_gettime(&uptime);
526 points = badness(task, uptime.tv_sec);
527 return sprintf(buffer, "%lu\n", points);
530 /************************************************************************/
531 /* Here the fs part begins */
532 /************************************************************************/
534 /* permission checks */
535 static int proc_fd_access_allowed(struct inode *inode)
537 struct task_struct *task;
539 /* Allow access to a task's file descriptors if it is us or we
540 * may use ptrace attach to the process and find out that
543 task = get_proc_task(inode);
545 allowed = ptrace_may_attach(task);
546 put_task_struct(task);
551 extern struct seq_operations mounts_op;
557 static int mounts_open(struct inode *inode, struct file *file)
559 struct task_struct *task = get_proc_task(inode);
560 struct namespace *namespace = NULL;
561 struct proc_mounts *p;
566 namespace = task->namespace;
568 get_namespace(namespace);
570 put_task_struct(task);
575 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
577 file->private_data = &p->m;
578 ret = seq_open(file, &mounts_op);
580 p->m.private = namespace;
581 p->event = namespace->event;
586 put_namespace(namespace);
591 static int mounts_release(struct inode *inode, struct file *file)
593 struct seq_file *m = file->private_data;
594 struct namespace *namespace = m->private;
595 put_namespace(namespace);
596 return seq_release(inode, file);
599 static unsigned mounts_poll(struct file *file, poll_table *wait)
601 struct proc_mounts *p = file->private_data;
602 struct namespace *ns = p->m.private;
605 poll_wait(file, &ns->poll, wait);
607 spin_lock(&vfsmount_lock);
608 if (p->event != ns->event) {
609 p->event = ns->event;
612 spin_unlock(&vfsmount_lock);
617 static struct file_operations proc_mounts_operations = {
621 .release = mounts_release,
625 extern struct seq_operations mountstats_op;
626 static int mountstats_open(struct inode *inode, struct file *file)
628 int ret = seq_open(file, &mountstats_op);
631 struct seq_file *m = file->private_data;
632 struct namespace *namespace = NULL;
633 struct task_struct *task = get_proc_task(inode);
637 namespace = task->namespace;
639 get_namespace(namespace);
641 put_task_struct(task);
645 m->private = namespace;
647 seq_release(inode, file);
654 static struct file_operations proc_mountstats_operations = {
655 .open = mountstats_open,
658 .release = mounts_release,
661 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
663 static ssize_t proc_info_read(struct file * file, char __user * buf,
664 size_t count, loff_t *ppos)
666 struct inode * inode = file->f_dentry->d_inode;
669 struct task_struct *task = get_proc_task(inode);
675 if (count > PROC_BLOCK_SIZE)
676 count = PROC_BLOCK_SIZE;
679 if (!(page = __get_free_page(GFP_KERNEL)))
682 length = PROC_I(inode)->op.proc_read(task, (char*)page);
685 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
688 put_task_struct(task);
693 static struct file_operations proc_info_file_operations = {
694 .read = proc_info_read,
697 static int mem_open(struct inode* inode, struct file* file)
699 file->private_data = (void*)((long)current->self_exec_id);
703 static ssize_t mem_read(struct file * file, char __user * buf,
704 size_t count, loff_t *ppos)
706 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
708 unsigned long src = *ppos;
710 struct mm_struct *mm;
715 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
719 page = (char *)__get_free_page(GFP_USER);
725 mm = get_task_mm(task);
731 if (file->private_data != (void*)((long)current->self_exec_id))
737 int this_len, retval;
739 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
740 retval = access_process_vm(task, src, page, this_len, 0);
741 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
747 if (copy_to_user(buf, page, retval)) {
762 free_page((unsigned long) page);
764 put_task_struct(task);
769 #define mem_write NULL
772 /* This is a security hazard */
773 static ssize_t mem_write(struct file * file, const char * buf,
774 size_t count, loff_t *ppos)
778 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
779 unsigned long dst = *ppos;
785 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
789 page = (char *)__get_free_page(GFP_USER);
794 int this_len, retval;
796 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
797 if (copy_from_user(page, buf, this_len)) {
801 retval = access_process_vm(task, dst, page, this_len, 1);
813 free_page((unsigned long) page);
815 put_task_struct(task);
821 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
825 file->f_pos = offset;
828 file->f_pos += offset;
833 force_successful_syscall_return();
837 static struct file_operations proc_mem_operations = {
844 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
845 size_t count, loff_t *ppos)
847 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
848 char buffer[PROC_NUMBUF];
851 loff_t __ppos = *ppos;
855 oom_adjust = task->oomkilladj;
856 put_task_struct(task);
858 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
861 if (count > len-__ppos)
863 if (copy_to_user(buf, buffer + __ppos, count))
865 *ppos = __ppos + count;
869 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
870 size_t count, loff_t *ppos)
872 struct task_struct *task;
873 char buffer[PROC_NUMBUF], *end;
876 if (!capable(CAP_SYS_RESOURCE))
878 memset(buffer, 0, sizeof(buffer));
879 if (count > sizeof(buffer) - 1)
880 count = sizeof(buffer) - 1;
881 if (copy_from_user(buffer, buf, count))
883 oom_adjust = simple_strtol(buffer, &end, 0);
884 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
888 task = get_proc_task(file->f_dentry->d_inode);
891 task->oomkilladj = oom_adjust;
892 put_task_struct(task);
893 if (end - buffer == 0)
898 static struct file_operations proc_oom_adjust_operations = {
899 .read = oom_adjust_read,
900 .write = oom_adjust_write,
903 #ifdef CONFIG_AUDITSYSCALL
905 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
906 size_t count, loff_t *ppos)
908 struct inode * inode = file->f_dentry->d_inode;
909 struct task_struct *task = get_proc_task(inode);
911 char tmpbuf[TMPBUFLEN];
915 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
916 audit_get_loginuid(task->audit_context));
917 put_task_struct(task);
918 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
921 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
922 size_t count, loff_t *ppos)
924 struct inode * inode = file->f_dentry->d_inode;
929 if (!capable(CAP_AUDIT_CONTROL))
932 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
935 if (count >= PAGE_SIZE)
936 count = PAGE_SIZE - 1;
939 /* No partial writes. */
942 page = (char*)__get_free_page(GFP_USER);
946 if (copy_from_user(page, buf, count))
950 loginuid = simple_strtoul(page, &tmp, 10);
956 length = audit_set_loginuid(current, loginuid);
957 if (likely(length == 0))
961 free_page((unsigned long) page);
965 static struct file_operations proc_loginuid_operations = {
966 .read = proc_loginuid_read,
967 .write = proc_loginuid_write,
971 #ifdef CONFIG_SECCOMP
972 static ssize_t seccomp_read(struct file *file, char __user *buf,
973 size_t count, loff_t *ppos)
975 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
977 loff_t __ppos = *ppos;
982 /* no need to print the trailing zero, so use only len */
983 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
984 put_task_struct(tsk);
987 if (count > len - __ppos)
988 count = len - __ppos;
989 if (copy_to_user(buf, __buf + __ppos, count))
991 *ppos = __ppos + count;
995 static ssize_t seccomp_write(struct file *file, const char __user *buf,
996 size_t count, loff_t *ppos)
998 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
999 char __buf[20], *end;
1000 unsigned int seccomp_mode;
1007 /* can set it only once to be even more secure */
1009 if (unlikely(tsk->seccomp.mode))
1013 memset(__buf, 0, sizeof(__buf));
1014 count = min(count, sizeof(__buf) - 1);
1015 if (copy_from_user(__buf, buf, count))
1018 seccomp_mode = simple_strtoul(__buf, &end, 0);
1022 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1023 tsk->seccomp.mode = seccomp_mode;
1024 set_tsk_thread_flag(tsk, TIF_SECCOMP);
1028 if (unlikely(!(end - __buf)))
1030 result = end - __buf;
1032 put_task_struct(tsk);
1037 static struct file_operations proc_seccomp_operations = {
1038 .read = seccomp_read,
1039 .write = seccomp_write,
1041 #endif /* CONFIG_SECCOMP */
1043 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1045 struct inode *inode = dentry->d_inode;
1046 int error = -EACCES;
1048 /* We don't need a base pointer in the /proc filesystem */
1051 /* Are we allowed to snoop on the tasks file descriptors? */
1052 if (!proc_fd_access_allowed(inode))
1055 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1056 nd->last_type = LAST_BIND;
1058 return ERR_PTR(error);
1061 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1062 char __user *buffer, int buflen)
1064 struct inode * inode;
1065 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1071 inode = dentry->d_inode;
1072 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1073 len = PTR_ERR(path);
1076 len = tmp + PAGE_SIZE - 1 - path;
1080 if (copy_to_user(buffer, path, len))
1083 free_page((unsigned long)tmp);
1087 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1089 int error = -EACCES;
1090 struct inode *inode = dentry->d_inode;
1092 struct vfsmount *mnt = NULL;
1094 /* Are we allowed to snoop on the tasks file descriptors? */
1095 if (!proc_fd_access_allowed(inode))
1098 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1102 error = do_proc_readlink(de, mnt, buffer, buflen);
1109 static struct inode_operations proc_pid_link_inode_operations = {
1110 .readlink = proc_pid_readlink,
1111 .follow_link = proc_pid_follow_link
1114 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1116 struct dentry *dentry = filp->f_dentry;
1117 struct inode *inode = dentry->d_inode;
1118 struct task_struct *p = get_proc_task(inode);
1119 unsigned int fd, tid, ino;
1121 char buf[PROC_NUMBUF];
1122 struct files_struct * files;
1123 struct fdtable *fdt;
1134 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1138 ino = parent_ino(dentry);
1139 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1143 files = get_files_struct(p);
1147 fdt = files_fdtable(files);
1148 for (fd = filp->f_pos-2;
1150 fd++, filp->f_pos++) {
1153 if (!fcheck_files(files, fd))
1161 buf[j] = '0' + (i % 10);
1165 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1166 if (filldir(dirent, buf+j, PROC_NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1173 put_files_struct(files);
1181 static int proc_pident_readdir(struct file *filp,
1182 void *dirent, filldir_t filldir,
1183 struct pid_entry *ents, unsigned int nents)
1187 struct dentry *dentry = filp->f_dentry;
1188 struct inode *inode = dentry->d_inode;
1189 struct task_struct *task = get_proc_task(inode);
1190 struct pid_entry *p;
1200 put_task_struct(task);
1205 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1211 ino = parent_ino(dentry);
1212 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1225 if (filldir(dirent, p->name, p->len, filp->f_pos,
1226 fake_ino(pid, p->type), p->mode >> 12) < 0)
1238 static int proc_tgid_base_readdir(struct file * filp,
1239 void * dirent, filldir_t filldir)
1241 return proc_pident_readdir(filp,dirent,filldir,
1242 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1245 static int proc_tid_base_readdir(struct file * filp,
1246 void * dirent, filldir_t filldir)
1248 return proc_pident_readdir(filp,dirent,filldir,
1249 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1252 /* building an inode */
1254 static int task_dumpable(struct task_struct *task)
1257 struct mm_struct *mm;
1262 dumpable = mm->dumpable;
1270 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1272 struct inode * inode;
1273 struct proc_inode *ei;
1275 /* We need a new inode */
1277 inode = new_inode(sb);
1283 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1284 inode->i_ino = fake_ino(task->pid, ino);
1287 * grab the reference to task.
1289 ei->pid = get_pid(task->pids[PIDTYPE_PID].pid);
1295 if (task_dumpable(task)) {
1296 inode->i_uid = task->euid;
1297 inode->i_gid = task->egid;
1299 security_task_to_inode(task, inode);
1312 * Exceptional case: normally we are not allowed to unhash a busy
1313 * directory. In this case, however, we can do it - no aliasing problems
1314 * due to the way we treat inodes.
1316 * Rewrite the inode's ownerships here because the owning task may have
1317 * performed a setuid(), etc.
1319 * Before the /proc/pid/status file was created the only way to read
1320 * the effective uid of a /process was to stat /proc/pid. Reading
1321 * /proc/pid/status is slow enough that procps and other packages
1322 * kept stating /proc/pid. To keep the rules in /proc simple I have
1323 * made this apply to all per process world readable and executable
1326 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1328 struct inode *inode = dentry->d_inode;
1329 struct task_struct *task = get_proc_task(inode);
1331 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1332 task_dumpable(task)) {
1333 inode->i_uid = task->euid;
1334 inode->i_gid = task->egid;
1339 security_task_to_inode(task, inode);
1340 put_task_struct(task);
1347 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1349 struct inode *inode = dentry->d_inode;
1350 struct task_struct *task;
1351 generic_fillattr(inode, stat);
1356 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1358 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1359 task_dumpable(task)) {
1360 stat->uid = task->euid;
1361 stat->gid = task->egid;
1368 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1370 struct inode *inode = dentry->d_inode;
1371 struct task_struct *task = get_proc_task(inode);
1372 int fd = proc_fd(inode);
1373 struct files_struct *files;
1376 files = get_files_struct(task);
1379 if (fcheck_files(files, fd)) {
1381 put_files_struct(files);
1382 if (task_dumpable(task)) {
1383 inode->i_uid = task->euid;
1384 inode->i_gid = task->egid;
1389 security_task_to_inode(task, inode);
1390 put_task_struct(task);
1394 put_files_struct(files);
1396 put_task_struct(task);
1402 static int pid_delete_dentry(struct dentry * dentry)
1404 /* Is the task we represent dead?
1405 * If so, then don't put the dentry on the lru list,
1406 * kill it immediately.
1408 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1411 static struct dentry_operations tid_fd_dentry_operations =
1413 .d_revalidate = tid_fd_revalidate,
1414 .d_delete = pid_delete_dentry,
1417 static struct dentry_operations pid_dentry_operations =
1419 .d_revalidate = pid_revalidate,
1420 .d_delete = pid_delete_dentry,
1425 static unsigned name_to_int(struct dentry *dentry)
1427 const char *name = dentry->d_name.name;
1428 int len = dentry->d_name.len;
1431 if (len > 1 && *name == '0')
1434 unsigned c = *name++ - '0';
1437 if (n >= (~0U-9)/10)
1448 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1450 struct task_struct *task = get_proc_task(dir);
1451 unsigned fd = name_to_int(dentry);
1452 struct dentry *result = ERR_PTR(-ENOENT);
1454 struct files_struct * files;
1455 struct inode *inode;
1456 struct proc_inode *ei;
1463 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1468 files = get_files_struct(task);
1471 inode->i_mode = S_IFLNK;
1474 * We are not taking a ref to the file structure, so we must
1477 spin_lock(&files->file_lock);
1478 file = fcheck_files(files, fd);
1481 if (file->f_mode & 1)
1482 inode->i_mode |= S_IRUSR | S_IXUSR;
1483 if (file->f_mode & 2)
1484 inode->i_mode |= S_IWUSR | S_IXUSR;
1485 spin_unlock(&files->file_lock);
1486 put_files_struct(files);
1487 inode->i_op = &proc_pid_link_inode_operations;
1489 ei->op.proc_get_link = proc_fd_link;
1490 dentry->d_op = &tid_fd_dentry_operations;
1491 d_add(dentry, inode);
1492 /* Close the race of the process dying before we return the dentry */
1493 if (tid_fd_revalidate(dentry, NULL))
1496 put_task_struct(task);
1501 spin_unlock(&files->file_lock);
1502 put_files_struct(files);
1508 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1509 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1510 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
1512 static struct file_operations proc_fd_operations = {
1513 .read = generic_read_dir,
1514 .readdir = proc_readfd,
1517 static struct file_operations proc_task_operations = {
1518 .read = generic_read_dir,
1519 .readdir = proc_task_readdir,
1523 * proc directories can do almost nothing..
1525 static struct inode_operations proc_fd_inode_operations = {
1526 .lookup = proc_lookupfd,
1529 static struct inode_operations proc_task_inode_operations = {
1530 .lookup = proc_task_lookup,
1531 .getattr = proc_task_getattr,
1534 #ifdef CONFIG_SECURITY
1535 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1536 size_t count, loff_t *ppos)
1538 struct inode * inode = file->f_dentry->d_inode;
1541 struct task_struct *task = get_proc_task(inode);
1547 if (count > PAGE_SIZE)
1550 if (!(page = __get_free_page(GFP_KERNEL)))
1553 length = security_getprocattr(task,
1554 (char*)file->f_dentry->d_name.name,
1555 (void*)page, count);
1557 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1560 put_task_struct(task);
1565 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1566 size_t count, loff_t *ppos)
1568 struct inode * inode = file->f_dentry->d_inode;
1571 struct task_struct *task = get_proc_task(inode);
1576 if (count > PAGE_SIZE)
1579 /* No partial writes. */
1585 page = (char*)__get_free_page(GFP_USER);
1590 if (copy_from_user(page, buf, count))
1593 length = security_setprocattr(task,
1594 (char*)file->f_dentry->d_name.name,
1595 (void*)page, count);
1597 free_page((unsigned long) page);
1599 put_task_struct(task);
1604 static struct file_operations proc_pid_attr_operations = {
1605 .read = proc_pid_attr_read,
1606 .write = proc_pid_attr_write,
1609 static struct file_operations proc_tid_attr_operations;
1610 static struct inode_operations proc_tid_attr_inode_operations;
1611 static struct file_operations proc_tgid_attr_operations;
1612 static struct inode_operations proc_tgid_attr_inode_operations;
1616 static struct dentry *proc_pident_lookup(struct inode *dir,
1617 struct dentry *dentry,
1618 struct pid_entry *ents)
1620 struct inode *inode;
1621 struct dentry *error;
1622 struct task_struct *task = get_proc_task(dir);
1623 struct pid_entry *p;
1624 struct proc_inode *ei;
1626 error = ERR_PTR(-ENOENT);
1632 for (p = ents; p->name; p++) {
1633 if (p->len != dentry->d_name.len)
1635 if (!memcmp(dentry->d_name.name, p->name, p->len))
1641 error = ERR_PTR(-EINVAL);
1642 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1647 inode->i_mode = p->mode;
1649 * Yes, it does not scale. And it should not. Don't add
1650 * new entries into /proc/<tgid>/ without very good reasons.
1653 case PROC_TGID_TASK:
1655 inode->i_op = &proc_task_inode_operations;
1656 inode->i_fop = &proc_task_operations;
1661 inode->i_op = &proc_fd_inode_operations;
1662 inode->i_fop = &proc_fd_operations;
1666 inode->i_op = &proc_pid_link_inode_operations;
1667 ei->op.proc_get_link = proc_exe_link;
1671 inode->i_op = &proc_pid_link_inode_operations;
1672 ei->op.proc_get_link = proc_cwd_link;
1675 case PROC_TGID_ROOT:
1676 inode->i_op = &proc_pid_link_inode_operations;
1677 ei->op.proc_get_link = proc_root_link;
1679 case PROC_TID_ENVIRON:
1680 case PROC_TGID_ENVIRON:
1681 inode->i_fop = &proc_info_file_operations;
1682 ei->op.proc_read = proc_pid_environ;
1685 case PROC_TGID_AUXV:
1686 inode->i_fop = &proc_info_file_operations;
1687 ei->op.proc_read = proc_pid_auxv;
1689 case PROC_TID_STATUS:
1690 case PROC_TGID_STATUS:
1691 inode->i_fop = &proc_info_file_operations;
1692 ei->op.proc_read = proc_pid_status;
1695 inode->i_fop = &proc_info_file_operations;
1696 ei->op.proc_read = proc_tid_stat;
1698 case PROC_TGID_STAT:
1699 inode->i_fop = &proc_info_file_operations;
1700 ei->op.proc_read = proc_tgid_stat;
1702 case PROC_TID_CMDLINE:
1703 case PROC_TGID_CMDLINE:
1704 inode->i_fop = &proc_info_file_operations;
1705 ei->op.proc_read = proc_pid_cmdline;
1707 case PROC_TID_STATM:
1708 case PROC_TGID_STATM:
1709 inode->i_fop = &proc_info_file_operations;
1710 ei->op.proc_read = proc_pid_statm;
1713 case PROC_TGID_MAPS:
1714 inode->i_fop = &proc_maps_operations;
1717 case PROC_TID_NUMA_MAPS:
1718 case PROC_TGID_NUMA_MAPS:
1719 inode->i_fop = &proc_numa_maps_operations;
1724 inode->i_fop = &proc_mem_operations;
1726 #ifdef CONFIG_SECCOMP
1727 case PROC_TID_SECCOMP:
1728 case PROC_TGID_SECCOMP:
1729 inode->i_fop = &proc_seccomp_operations;
1731 #endif /* CONFIG_SECCOMP */
1732 case PROC_TID_MOUNTS:
1733 case PROC_TGID_MOUNTS:
1734 inode->i_fop = &proc_mounts_operations;
1737 case PROC_TID_SMAPS:
1738 case PROC_TGID_SMAPS:
1739 inode->i_fop = &proc_smaps_operations;
1742 case PROC_TID_MOUNTSTATS:
1743 case PROC_TGID_MOUNTSTATS:
1744 inode->i_fop = &proc_mountstats_operations;
1746 #ifdef CONFIG_SECURITY
1749 inode->i_op = &proc_tid_attr_inode_operations;
1750 inode->i_fop = &proc_tid_attr_operations;
1752 case PROC_TGID_ATTR:
1754 inode->i_op = &proc_tgid_attr_inode_operations;
1755 inode->i_fop = &proc_tgid_attr_operations;
1757 case PROC_TID_ATTR_CURRENT:
1758 case PROC_TGID_ATTR_CURRENT:
1759 case PROC_TID_ATTR_PREV:
1760 case PROC_TGID_ATTR_PREV:
1761 case PROC_TID_ATTR_EXEC:
1762 case PROC_TGID_ATTR_EXEC:
1763 case PROC_TID_ATTR_FSCREATE:
1764 case PROC_TGID_ATTR_FSCREATE:
1765 case PROC_TID_ATTR_KEYCREATE:
1766 case PROC_TGID_ATTR_KEYCREATE:
1767 inode->i_fop = &proc_pid_attr_operations;
1770 #ifdef CONFIG_KALLSYMS
1771 case PROC_TID_WCHAN:
1772 case PROC_TGID_WCHAN:
1773 inode->i_fop = &proc_info_file_operations;
1774 ei->op.proc_read = proc_pid_wchan;
1777 #ifdef CONFIG_SCHEDSTATS
1778 case PROC_TID_SCHEDSTAT:
1779 case PROC_TGID_SCHEDSTAT:
1780 inode->i_fop = &proc_info_file_operations;
1781 ei->op.proc_read = proc_pid_schedstat;
1784 #ifdef CONFIG_CPUSETS
1785 case PROC_TID_CPUSET:
1786 case PROC_TGID_CPUSET:
1787 inode->i_fop = &proc_cpuset_operations;
1790 case PROC_TID_OOM_SCORE:
1791 case PROC_TGID_OOM_SCORE:
1792 inode->i_fop = &proc_info_file_operations;
1793 ei->op.proc_read = proc_oom_score;
1795 case PROC_TID_OOM_ADJUST:
1796 case PROC_TGID_OOM_ADJUST:
1797 inode->i_fop = &proc_oom_adjust_operations;
1799 #ifdef CONFIG_AUDITSYSCALL
1800 case PROC_TID_LOGINUID:
1801 case PROC_TGID_LOGINUID:
1802 inode->i_fop = &proc_loginuid_operations;
1806 printk("procfs: impossible type (%d)",p->type);
1808 error = ERR_PTR(-EINVAL);
1811 dentry->d_op = &pid_dentry_operations;
1812 d_add(dentry, inode);
1813 /* Close the race of the process dying before we return the dentry */
1814 if (pid_revalidate(dentry, NULL))
1817 put_task_struct(task);
1822 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1823 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1826 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1827 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1830 static struct file_operations proc_tgid_base_operations = {
1831 .read = generic_read_dir,
1832 .readdir = proc_tgid_base_readdir,
1835 static struct file_operations proc_tid_base_operations = {
1836 .read = generic_read_dir,
1837 .readdir = proc_tid_base_readdir,
1840 static struct inode_operations proc_tgid_base_inode_operations = {
1841 .lookup = proc_tgid_base_lookup,
1842 .getattr = pid_getattr,
1845 static struct inode_operations proc_tid_base_inode_operations = {
1846 .lookup = proc_tid_base_lookup,
1847 .getattr = pid_getattr,
1850 #ifdef CONFIG_SECURITY
1851 static int proc_tgid_attr_readdir(struct file * filp,
1852 void * dirent, filldir_t filldir)
1854 return proc_pident_readdir(filp,dirent,filldir,
1855 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1858 static int proc_tid_attr_readdir(struct file * filp,
1859 void * dirent, filldir_t filldir)
1861 return proc_pident_readdir(filp,dirent,filldir,
1862 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1865 static struct file_operations proc_tgid_attr_operations = {
1866 .read = generic_read_dir,
1867 .readdir = proc_tgid_attr_readdir,
1870 static struct file_operations proc_tid_attr_operations = {
1871 .read = generic_read_dir,
1872 .readdir = proc_tid_attr_readdir,
1875 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1876 struct dentry *dentry, struct nameidata *nd)
1878 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1881 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1882 struct dentry *dentry, struct nameidata *nd)
1884 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1887 static struct inode_operations proc_tgid_attr_inode_operations = {
1888 .lookup = proc_tgid_attr_lookup,
1889 .getattr = pid_getattr,
1892 static struct inode_operations proc_tid_attr_inode_operations = {
1893 .lookup = proc_tid_attr_lookup,
1894 .getattr = pid_getattr,
1901 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1904 char tmp[PROC_NUMBUF];
1905 sprintf(tmp, "%d", current->tgid);
1906 return vfs_readlink(dentry,buffer,buflen,tmp);
1909 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1911 char tmp[PROC_NUMBUF];
1912 sprintf(tmp, "%d", current->tgid);
1913 return ERR_PTR(vfs_follow_link(nd,tmp));
1916 static struct inode_operations proc_self_inode_operations = {
1917 .readlink = proc_self_readlink,
1918 .follow_link = proc_self_follow_link,
1922 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1924 * @task: task that should be flushed.
1926 * Looks in the dcache for
1928 * /proc/@tgid/task/@pid
1929 * if either directory is present flushes it and all of it'ts children
1932 * It is safe and reasonable to cache /proc entries for a task until
1933 * that task exits. After that they just clog up the dcache with
1934 * useless entries, possibly causing useful dcache entries to be
1935 * flushed instead. This routine is proved to flush those useless
1936 * dcache entries at process exit time.
1938 * NOTE: This routine is just an optimization so it does not guarantee
1939 * that no dcache entries will exist at process exit time it
1940 * just makes it very unlikely that any will persist.
1942 void proc_flush_task(struct task_struct *task)
1944 struct dentry *dentry, *leader, *dir;
1945 char buf[PROC_NUMBUF];
1949 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1950 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1952 shrink_dcache_parent(dentry);
1957 if (thread_group_leader(task))
1961 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1962 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1967 name.len = strlen(name.name);
1968 dir = d_hash_and_lookup(leader, &name);
1970 goto out_put_leader;
1973 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1974 dentry = d_hash_and_lookup(dir, &name);
1976 shrink_dcache_parent(dentry);
1989 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1991 struct dentry *result = ERR_PTR(-ENOENT);
1992 struct task_struct *task;
1993 struct inode *inode;
1994 struct proc_inode *ei;
1997 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1998 inode = new_inode(dir->i_sb);
2000 return ERR_PTR(-ENOMEM);
2002 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2003 inode->i_ino = fake_ino(0, PROC_TGID_INO);
2005 inode->i_mode = S_IFLNK|S_IRWXUGO;
2006 inode->i_uid = inode->i_gid = 0;
2008 inode->i_op = &proc_self_inode_operations;
2009 d_add(dentry, inode);
2012 tgid = name_to_int(dentry);
2017 task = find_task_by_pid(tgid);
2019 get_task_struct(task);
2024 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
2028 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2029 inode->i_op = &proc_tgid_base_inode_operations;
2030 inode->i_fop = &proc_tgid_base_operations;
2031 inode->i_flags|=S_IMMUTABLE;
2032 #ifdef CONFIG_SECURITY
2038 dentry->d_op = &pid_dentry_operations;
2040 d_add(dentry, inode);
2041 /* Close the race of the process dying before we return the dentry */
2042 if (pid_revalidate(dentry, NULL))
2046 put_task_struct(task);
2052 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2054 struct dentry *result = ERR_PTR(-ENOENT);
2055 struct task_struct *task;
2056 struct task_struct *leader = get_proc_task(dir);
2057 struct inode *inode;
2063 tid = name_to_int(dentry);
2068 task = find_task_by_pid(tid);
2070 get_task_struct(task);
2074 if (leader->tgid != task->tgid)
2077 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2082 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2083 inode->i_op = &proc_tid_base_inode_operations;
2084 inode->i_fop = &proc_tid_base_operations;
2085 inode->i_flags|=S_IMMUTABLE;
2086 #ifdef CONFIG_SECURITY
2092 dentry->d_op = &pid_dentry_operations;
2094 d_add(dentry, inode);
2095 /* Close the race of the process dying before we return the dentry */
2096 if (pid_revalidate(dentry, NULL))
2100 put_task_struct(task);
2102 put_task_struct(leader);
2108 * Find the first tgid to return to user space.
2110 * Usually this is just whatever follows &init_task, but if the users
2111 * buffer was too small to hold the full list or there was a seek into
2112 * the middle of the directory we have more work to do.
2114 * In the case of a short read we start with find_task_by_pid.
2116 * In the case of a seek we start with &init_task and walk nr
2119 static struct task_struct *first_tgid(int tgid, unsigned int nr)
2121 struct task_struct *pos;
2124 pos = find_task_by_pid(tgid);
2125 if (pos && thread_group_leader(pos))
2128 /* If nr exceeds the number of processes get out quickly */
2130 if (nr && nr >= nr_processes())
2133 /* If we haven't found our starting place yet start with
2134 * the init_task and walk nr tasks forward.
2136 for (pos = next_task(&init_task); nr > 0; --nr) {
2137 pos = next_task(pos);
2138 if (pos == &init_task) {
2144 get_task_struct(pos);
2151 * Find the next task in the task list.
2152 * Return NULL if we loop or there is any error.
2154 * The reference to the input task_struct is released.
2156 static struct task_struct *next_tgid(struct task_struct *start)
2158 struct task_struct *pos;
2161 if (pid_alive(start))
2162 pos = next_task(start);
2163 if (pid_alive(pos) && (pos != &init_task)) {
2164 get_task_struct(pos);
2170 put_task_struct(start);
2174 /* for the /proc/ directory itself, after non-process stuff has been done */
2175 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2177 char buf[PROC_NUMBUF];
2178 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2179 struct task_struct *task;
2183 ino_t ino = fake_ino(0,PROC_TGID_INO);
2184 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2191 /* f_version caches the tgid value that the last readdir call couldn't
2192 * return. lseek aka telldir automagically resets f_version to 0.
2194 tgid = filp->f_version;
2195 filp->f_version = 0;
2196 for (task = first_tgid(tgid, nr);
2198 task = next_tgid(task), filp->f_pos++) {
2202 len = snprintf(buf, sizeof(buf), "%d", tgid);
2203 ino = fake_ino(tgid, PROC_TGID_INO);
2204 if (filldir(dirent, buf, len, filp->f_pos, ino, DT_DIR) < 0) {
2205 /* returning this tgid failed, save it as the first
2206 * pid for the next readir call */
2207 filp->f_version = tgid;
2208 put_task_struct(task);
2216 * Find the first tid of a thread group to return to user space.
2218 * Usually this is just the thread group leader, but if the users
2219 * buffer was too small or there was a seek into the middle of the
2220 * directory we have more work todo.
2222 * In the case of a short read we start with find_task_by_pid.
2224 * In the case of a seek we start with the leader and walk nr
2227 static struct task_struct *first_tid(struct task_struct *leader,
2230 struct task_struct *pos = NULL;
2233 /* Attempt to start with the pid of a thread */
2234 if (tid && (nr > 0)) {
2235 pos = find_task_by_pid(tid);
2236 if (pos && (pos->group_leader != leader))
2242 /* If nr exceeds the number of threads there is nothing todo */
2244 if (nr >= get_nr_threads(leader))
2248 /* If we haven't found our starting place yet start with the
2249 * leader and walk nr threads forward.
2251 if (!pos && (nr >= 0))
2254 for (; pos && pid_alive(pos); pos = next_thread(pos)) {
2257 get_task_struct(pos);
2267 * Find the next thread in the thread list.
2268 * Return NULL if there is an error or no next thread.
2270 * The reference to the input task_struct is released.
2272 static struct task_struct *next_tid(struct task_struct *start)
2274 struct task_struct *pos;
2277 if (pid_alive(start))
2278 pos = next_thread(start);
2279 if (pid_alive(pos) && (pos != start->group_leader))
2280 get_task_struct(pos);
2284 put_task_struct(start);
2288 /* for the /proc/TGID/task/ directories */
2289 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2291 char buf[PROC_NUMBUF];
2292 struct dentry *dentry = filp->f_dentry;
2293 struct inode *inode = dentry->d_inode;
2294 struct task_struct *leader = get_proc_task(inode);
2295 struct task_struct *task;
2296 int retval = -ENOENT;
2299 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2308 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2313 ino = parent_ino(dentry);
2314 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2320 /* f_version caches the tgid value that the last readdir call couldn't
2321 * return. lseek aka telldir automagically resets f_version to 0.
2323 tid = filp->f_version;
2324 filp->f_version = 0;
2325 for (task = first_tid(leader, tid, pos - 2);
2327 task = next_tid(task), pos++) {
2330 len = snprintf(buf, sizeof(buf), "%d", tid);
2331 ino = fake_ino(tid, PROC_TID_INO);
2332 if (filldir(dirent, buf, len, pos, ino, DT_DIR < 0)) {
2333 /* returning this tgid failed, save it as the first
2334 * pid for the next readir call */
2335 filp->f_version = tid;
2336 put_task_struct(task);
2342 put_task_struct(leader);
2347 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2349 struct inode *inode = dentry->d_inode;
2350 struct task_struct *p = get_proc_task(inode);
2351 generic_fillattr(inode, stat);
2355 stat->nlink += get_nr_threads(p);