?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Ok. I didn't make 2.4.0 in 2000. Tough. I tried, but we had some
[davej-history.git] / fs / proc / base.c
blobc1a960c5db2ca3b45a2af0f42f265078866d1a9b
1 /*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
14 */
15
16 #include <asm/uaccess.h>
17
18 #include <linux/config.h>
19 #include <linux/errno.h>
20 #include <linux/sched.h>
21 #include <linux/proc_fs.h>
22 #include <linux/stat.h>
23 #include <linux/init.h>
24 #include <linux/file.h>
25 #include <linux/string.h>
26
27 /*
28 * For hysterical raisins we keep the same inumbers as in the old procfs.
29 * Feel free to change the macro below - just keep the range distinct from
30 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
31 * As soon as we'll get a separate superblock we will be able to forget
32 * about magical ranges too.
33 */
34
35 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
36
37 ssize_t proc_pid_read_maps(struct task_struct*,struct file*,char*,size_t,loff_t*);
38 int proc_pid_stat(struct task_struct*,char*);
39 int proc_pid_status(struct task_struct*,char*);
40 int proc_pid_statm(struct task_struct*,char*);
41 int proc_pid_cpu(struct task_struct*,char*);
42
43 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
44 {
45 if (inode->u.proc_i.file) {
46 *mnt = mntget(inode->u.proc_i.file->f_vfsmnt);
47 *dentry = dget(inode->u.proc_i.file->f_dentry);
48 return 0;
49 }
50 return -ENOENT;
51 }
52
53 static int proc_exe_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
54 {
55 struct mm_struct * mm;
56 struct vm_area_struct * vma;
57 int result = -ENOENT;
58 struct task_struct *task = inode->u.proc_i.task;
59
60 task_lock(task);
61 mm = task->mm;
62 if (mm)
63 atomic_inc(&mm->mm_users);
64 task_unlock(task);
65 if (!mm)
66 goto out;
67 down(&mm->mmap_sem);
68 vma = mm->mmap;
69 while (vma) {
70 if ((vma->vm_flags & VM_EXECUTABLE) &&
71 vma->vm_file) {
72 *mnt = mntget(vma->vm_file->f_vfsmnt);
73 *dentry = dget(vma->vm_file->f_dentry);
74 result = 0;
75 break;
76 }
77 vma = vma->vm_next;
78 }
79 up(&mm->mmap_sem);
80 mmput(mm);
81 out:
82 return result;
83 }
84
85 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
86 {
87 struct fs_struct *fs;
88 int result = -ENOENT;
89 task_lock(inode->u.proc_i.task);
90 fs = inode->u.proc_i.task->fs;
91 if(fs)
92 atomic_inc(&fs->count);
93 task_unlock(inode->u.proc_i.task);
94 if (fs) {
95 read_lock(&fs->lock);
96 *mnt = mntget(fs->pwdmnt);
97 *dentry = dget(fs->pwd);
98 read_unlock(&fs->lock);
99 result = 0;
100 put_fs_struct(fs);
101 }
102 return result;
103 }
104
105 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
106 {
107 struct fs_struct *fs;
108 int result = -ENOENT;
109 task_lock(inode->u.proc_i.task);
110 fs = inode->u.proc_i.task->fs;
111 if(fs)
112 atomic_inc(&fs->count);
113 task_unlock(inode->u.proc_i.task);
114 if (fs) {
115 read_lock(&fs->lock);
116 *mnt = mntget(fs->rootmnt);
117 *dentry = dget(fs->root);
118 read_unlock(&fs->lock);
119 result = 0;
120 put_fs_struct(fs);
121 }
122 return result;
123 }
124
125 static int proc_pid_environ(struct task_struct *task, char * buffer)
126 {
127 struct mm_struct *mm;
128 int res = 0;
129 task_lock(task);
130 mm = task->mm;
131 if (mm)
132 atomic_inc(&mm->mm_users);
133 task_unlock(task);
134 if (mm) {
135 int len = mm->env_end - mm->env_start;
136 if (len > PAGE_SIZE)
137 len = PAGE_SIZE;
138 res = access_process_vm(task, mm->env_start, buffer, len, 0);
139 mmput(mm);
140 }
141 return res;
142 }
143
144 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
145 {
146 struct mm_struct *mm;
147 int res = 0;
148 task_lock(task);
149 mm = task->mm;
150 if (mm)
151 atomic_inc(&mm->mm_users);
152 task_unlock(task);
153 if (mm) {
154 int len = mm->arg_end - mm->arg_start;
155 if (len > PAGE_SIZE)
156 len = PAGE_SIZE;
157 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
158 // If the nul at the end of args has been overwritten, then
159 // assume application is using setproctitle(3).
160 if ( res > 0 && buffer[res-1] != '\0' )
161 {
162 len = strnlen( buffer, res );
163 if ( len < res )
164 {
165 res = len;
166 }
167 else
168 {
169 len = mm->env_end - mm->env_start;
170 if (len > PAGE_SIZE - res)
171 len = PAGE_SIZE - res;
172 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
173 res = strnlen( buffer, res );
174 }
175 }
176 mmput(mm);
177 }
178 return res;
179 }
180
181 /************************************************************************/
182 /* Here the fs part begins */
183 /************************************************************************/
184
185 /* permission checks */
186
187 static int standard_permission(struct inode *inode, int mask)
188 {
189 int mode = inode->i_mode;
190
191 if ((mask & S_IWOTH) && IS_RDONLY(inode) &&
192 (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
193 return -EROFS; /* Nobody gets write access to a read-only fs */
194 else if ((mask & S_IWOTH) && IS_IMMUTABLE(inode))
195 return -EACCES; /* Nobody gets write access to an immutable file */
196 else if (current->fsuid == inode->i_uid)
197 mode >>= 6;
198 else if (in_group_p(inode->i_gid))
199 mode >>= 3;
200 if (((mode & mask & S_IRWXO) == mask) || capable(CAP_DAC_OVERRIDE))
201 return 0;
202 /* read and search access */
203 if ((mask == S_IROTH) ||
204 (S_ISDIR(mode) && !(mask & ~(S_IROTH | S_IXOTH))))
205 if (capable(CAP_DAC_READ_SEARCH))
206 return 0;
207 return -EACCES;
208 }
209
210 static int proc_check_root(struct inode *inode)
211 {
212 struct dentry *de, *base, *root;
213 struct vfsmount *our_vfsmnt, *vfsmnt, *mnt;
214 int res = 0;
215
216 if (proc_root_link(inode, &root, &vfsmnt)) /* Ewww... */
217 return -ENOENT;
218 read_lock(&current->fs->lock);
219 our_vfsmnt = mntget(current->fs->rootmnt);
220 base = dget(current->fs->root);
221 read_unlock(&current->fs->lock);
222
223 spin_lock(&dcache_lock);
224 de = root;
225 mnt = vfsmnt;
226
227 while (vfsmnt != our_vfsmnt) {
228 if (vfsmnt == vfsmnt->mnt_parent)
229 goto out;
230 de = vfsmnt->mnt_mountpoint;
231 vfsmnt = vfsmnt->mnt_parent;
232 }
233
234 if (!is_subdir(de, base))
235 goto out;
236 spin_unlock(&dcache_lock);
237
238 exit:
239 dput(base);
240 mntput(our_vfsmnt);
241 dput(root);
242 mntput(mnt);
243 return res;
244 out:
245 spin_unlock(&dcache_lock);
246 res = -EACCES;
247 goto exit;
248 }
249
250 static int proc_permission(struct inode *inode, int mask)
251 {
252 if (standard_permission(inode, mask) != 0)
253 return -EACCES;
254 return proc_check_root(inode);
255 }
256
257 static ssize_t pid_maps_read(struct file * file, char * buf,
258 size_t count, loff_t *ppos)
259 {
260 struct inode * inode = file->f_dentry->d_inode;
261 struct task_struct *task = inode->u.proc_i.task;
262 ssize_t res;
263
264 res = proc_pid_read_maps(task, file, buf, count, ppos);
265 return res;
266 }
267
268 static struct file_operations proc_maps_operations = {
269 read: pid_maps_read,
270 };
271
272 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
273
274 static ssize_t proc_info_read(struct file * file, char * buf,
275 size_t count, loff_t *ppos)
276 {
277 struct inode * inode = file->f_dentry->d_inode;
278 unsigned long page;
279 ssize_t length;
280 ssize_t end;
281 struct task_struct *task = inode->u.proc_i.task;
282
283 if (count > PROC_BLOCK_SIZE)
284 count = PROC_BLOCK_SIZE;
285 if (!(page = __get_free_page(GFP_KERNEL)))
286 return -ENOMEM;
287
288 length = inode->u.proc_i.op.proc_read(task, (char*)page);
289
290 if (length < 0) {
291 free_page(page);
292 return length;
293 }
294 /* Static 4kB (or whatever) block capacity */
295 if (*ppos >= length) {
296 free_page(page);
297 return 0;
298 }
299 if (count + *ppos > length)
300 count = length - *ppos;
301 end = count + *ppos;
302 copy_to_user(buf, (char *) page + *ppos, count);
303 *ppos = end;
304 free_page(page);
305 return count;
306 }
307
308 static struct file_operations proc_info_file_operations = {
309 read: proc_info_read,
310 };
311
312 #define MAY_PTRACE(p) \
313 (p==current||(p->p_pptr==current&&(p->ptrace & PT_PTRACED)&&p->state==TASK_STOPPED))
314
315 static ssize_t mem_read(struct file * file, char * buf,
316 size_t count, loff_t *ppos)
317 {
318 struct task_struct *task = file->f_dentry->d_inode->u.proc_i.task;
319 char *page;
320 unsigned long src = *ppos;
321 int copied = 0;
322
323 if (!MAY_PTRACE(task))
324 return -ESRCH;
325
326 page = (char *)__get_free_page(GFP_USER);
327 if (!page)
328 return -ENOMEM;
329
330 while (count > 0) {
331 int this_len, retval;
332
333 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
334 retval = access_process_vm(task, src, page, this_len, 0);
335 if (!retval) {
336 if (!copied)
337 copied = -EIO;
338 break;
339 }
340 if (copy_to_user(buf, page, retval)) {
341 copied = -EFAULT;
342 break;
343 }
344 copied += retval;
345 src += retval;
346 buf += retval;
347 count -= retval;
348 }
349 *ppos = src;
350 free_page((unsigned long) page);
351 return copied;
352 }
353
354 #define mem_write NULL
355
356 #ifndef mem_write
357 /* This is a security hazard */
358 static ssize_t mem_write(struct file * file, const char * buf,
359 size_t count, loff_t *ppos)
360 {
361 int copied = 0;
362 char *page;
363 struct task_struct *task = file->f_dentry->d_inode->u.proc_i.task;
364 unsigned long dst = *ppos;
365
366 if (!MAY_PTRACE(task))
367 return -ESRCH;
368
369 page = (char *)__get_free_page(GFP_USER);
370 if (!page)
371 return -ENOMEM;
372
373 while (count > 0) {
374 int this_len, retval;
375
376 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
377 if (copy_from_user(page, buf, this_len)) {
378 copied = -EFAULT;
379 break;
380 }
381 retval = access_process_vm(task, dst, page, this_len, 1);
382 if (!retval) {
383 if (!copied)
384 copied = -EIO;
385 break;
386 }
387 copied += retval;
388 buf += retval;
389 dst += retval;
390 count -= retval;
391 }
392 *ppos = dst;
393 free_page((unsigned long) page);
394 return copied;
395 }
396 #endif
397
398 static struct file_operations proc_mem_operations = {
399 read: mem_read,
400 write: mem_write,
401 };
402
403 static struct inode_operations proc_mem_inode_operations = {
404 permission: proc_permission,
405 };
406
407 static int proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
408 {
409 struct inode *inode = dentry->d_inode;
410 int error = -EACCES;
411
412 /* We don't need a base pointer in the /proc filesystem */
413 path_release(nd);
414
415 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
416 goto out;
417 error = proc_check_root(inode);
418 if (error)
419 goto out;
420
421 error = inode->u.proc_i.op.proc_get_link(inode, &nd->dentry, &nd->mnt);
422 nd->last_type = LAST_BIND;
423 out:
424 return error;
425 }
426
427 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
428 char * buffer, int buflen)
429 {
430 struct inode * inode;
431 char * tmp = (char*)__get_free_page(GFP_KERNEL), *path;
432 int len;
433
434 if (!tmp)
435 return -ENOMEM;
436
437 inode = dentry->d_inode;
438 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
439 len = tmp + PAGE_SIZE - 1 - path;
440
441 if (len < buflen)
442 buflen = len;
443 copy_to_user(buffer, path, buflen);
444 free_page((unsigned long)tmp);
445 return buflen;
446 }
447
448 static int proc_pid_readlink(struct dentry * dentry, char * buffer, int buflen)
449 {
450 int error = -EACCES;
451 struct inode *inode = dentry->d_inode;
452 struct dentry *de;
453 struct vfsmount *mnt = NULL;
454
455 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
456 goto out;
457 error = proc_check_root(inode);
458 if (error)
459 goto out;
460
461 error = inode->u.proc_i.op.proc_get_link(inode, &de, &mnt);
462 if (error)
463 goto out;
464
465 error = do_proc_readlink(de, mnt, buffer, buflen);
466 dput(de);
467 mntput(mnt);
468 out:
469 return error;
470 }
471
472 static struct inode_operations proc_pid_link_inode_operations = {
473 readlink: proc_pid_readlink,
474 follow_link: proc_pid_follow_link
475 };
476
477 struct pid_entry {
478 int type;
479 int len;
480 char *name;
481 mode_t mode;
482 };
483
484 enum pid_directory_inos {
485 PROC_PID_INO = 2,
486 PROC_PID_STATUS,
487 PROC_PID_MEM,
488 PROC_PID_CWD,
489 PROC_PID_ROOT,
490 PROC_PID_EXE,
491 PROC_PID_FD,
492 PROC_PID_ENVIRON,
493 PROC_PID_CMDLINE,
494 PROC_PID_STAT,
495 PROC_PID_STATM,
496 PROC_PID_MAPS,
497 PROC_PID_CPU,
498 PROC_PID_FD_DIR = 0x8000, /* 0x8000-0xffff */
499 };
500
501 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
502 static struct pid_entry base_stuff[] = {
503 E(PROC_PID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
504 E(PROC_PID_ENVIRON, "environ", S_IFREG|S_IRUSR),
505 E(PROC_PID_STATUS, "status", S_IFREG|S_IRUGO),
506 E(PROC_PID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
507 E(PROC_PID_STAT, "stat", S_IFREG|S_IRUGO),
508 E(PROC_PID_STATM, "statm", S_IFREG|S_IRUGO),
509 #ifdef CONFIG_SMP
510 E(PROC_PID_CPU, "cpu", S_IFREG|S_IRUGO),
511 #endif
512 E(PROC_PID_MAPS, "maps", S_IFREG|S_IRUGO),
513 E(PROC_PID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
514 E(PROC_PID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
515 E(PROC_PID_ROOT, "root", S_IFLNK|S_IRWXUGO),
516 E(PROC_PID_EXE, "exe", S_IFLNK|S_IRWXUGO),
517 {0,0,NULL,0}
518 };
519 #undef E
520
521 #define NUMBUF 10
522
523 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
524 {
525 struct inode *inode = filp->f_dentry->d_inode;
526 struct task_struct *p = inode->u.proc_i.task;
527 unsigned int fd, pid, ino;
528 int retval;
529 char buf[NUMBUF];
530 struct files_struct * files;
531
532 retval = 0;
533 pid = p->pid;
534
535 fd = filp->f_pos;
536 switch (fd) {
537 case 0:
538 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
539 goto out;
540 filp->f_pos++;
541 case 1:
542 ino = fake_ino(pid, PROC_PID_INO);
543 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
544 goto out;
545 filp->f_pos++;
546 default:
547 task_lock(p);
548 files = p->files;
549 if (files)
550 atomic_inc(&files->count);
551 task_unlock(p);
552 if (!files)
553 goto out;
554 for (fd = filp->f_pos-2;
555 fd < files->max_fds;
556 fd++, filp->f_pos++) {
557 unsigned int i,j;
558
559 if (!fcheck_files(files, fd))
560 continue;
561
562 j = NUMBUF;
563 i = fd;
564 do {
565 j--;
566 buf[j] = '0' + (i % 10);
567 i /= 10;
568 } while (i);
569
570 ino = fake_ino(pid, PROC_PID_FD_DIR + fd);
571 if (filldir(dirent, buf+j, NUMBUF-j, fd+2, ino, DT_LNK) < 0)
572 break;
573 }
574 put_files_struct(files);
575 }
576 out:
577 return retval;
578 }
579
580 static int proc_base_readdir(struct file * filp,
581 void * dirent, filldir_t filldir)
582 {
583 int i;
584 int pid;
585 struct inode *inode = filp->f_dentry->d_inode;
586 struct pid_entry *p;
587
588 pid = inode->u.proc_i.task->pid;
589 if (!inode->u.proc_i.task->p_pptr)
590 return -ENOENT;
591 i = filp->f_pos;
592 switch (i) {
593 case 0:
594 if (filldir(dirent, ".", 1, i, inode->i_ino, DT_DIR) < 0)
595 return 0;
596 i++;
597 filp->f_pos++;
598 /* fall through */
599 case 1:
600 if (filldir(dirent, "..", 2, i, PROC_ROOT_INO, DT_DIR) < 0)
601 return 0;
602 i++;
603 filp->f_pos++;
604 /* fall through */
605 default:
606 i -= 2;
607 if (i>=sizeof(base_stuff)/sizeof(base_stuff[0]))
608 return 1;
609 p = base_stuff + i;
610 while (p->name) {
611 if (filldir(dirent, p->name, p->len, filp->f_pos,
612 fake_ino(pid, p->type), p->mode >> 12) < 0)
613 return 0;
614 filp->f_pos++;
615 p++;
616 }
617 }
618 return 1;
619 }
620
621 /* building an inode */
622
623 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
624 {
625 struct inode * inode;
626
627 /* We need a new inode */
628
629 inode = new_inode(sb);
630 if (!inode)
631 goto out;
632
633 /* Common stuff */
634
635 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
636 inode->i_ino = fake_ino(task->pid, ino);
637
638 inode->u.proc_i.file = NULL;
639 /*
640 * grab the reference to task.
641 */
642 inode->u.proc_i.task = task;
643 get_task_struct(task);
644 if (!task->p_pptr)
645 goto out_unlock;
646
647 inode->i_uid = 0;
648 inode->i_gid = 0;
649 if (ino == PROC_PID_INO || task->dumpable) {
650 inode->i_uid = task->euid;
651 inode->i_gid = task->egid;
652 }
653
654 out:
655 return inode;
656
657 out_unlock:
658 iput(inode);
659 return NULL;
660 }
661
662 /* dentry stuff */
663
664 static int pid_fd_revalidate(struct dentry * dentry, int flags)
665 {
666 return 0;
667 }
668
669 /*
670 * Exceptional case: normally we are not allowed to unhash a busy
671 * directory. In this case, however, we can do it - no aliasing problems
672 * due to the way we treat inodes.
673 */
674 static int pid_base_revalidate(struct dentry * dentry, int flags)
675 {
676 if (dentry->d_inode->u.proc_i.task->p_pptr)
677 return 1;
678 d_drop(dentry);
679 return 0;
680 }
681
682 static int pid_delete_dentry(struct dentry * dentry)
683 {
684 return 1;
685 }
686
687 static struct dentry_operations pid_fd_dentry_operations =
688 {
689 d_revalidate: pid_fd_revalidate,
690 d_delete: pid_delete_dentry,
691 };
692
693 static struct dentry_operations pid_dentry_operations =
694 {
695 d_delete: pid_delete_dentry,
696 };
697
698 static struct dentry_operations pid_base_dentry_operations =
699 {
700 d_revalidate: pid_base_revalidate,
701 d_delete: pid_delete_dentry,
702 };
703
704 /* Lookups */
705 #define MAX_MULBY10 ((~0U-9)/10)
706
707 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry)
708 {
709 unsigned int fd, c;
710 struct task_struct *task = dir->u.proc_i.task;
711 struct file * file;
712 struct files_struct * files;
713 struct inode *inode;
714 const char *name;
715 int len;
716
717 fd = 0;
718 len = dentry->d_name.len;
719 name = dentry->d_name.name;
720 if (len > 1 && *name == '0') goto out;
721 while (len-- > 0) {
722 c = *name - '0';
723 name++;
724 if (c > 9)
725 goto out;
726 if (fd >= MAX_MULBY10)
727 goto out;
728 fd *= 10;
729 fd += c;
730 }
731
732 inode = proc_pid_make_inode(dir->i_sb, task, PROC_PID_FD_DIR+fd);
733 if (!inode)
734 goto out;
735 task_lock(task);
736 files = task->files;
737 if (files)
738 atomic_inc(&files->count);
739 task_unlock(task);
740 if (!files)
741 goto out_unlock;
742 read_lock(&files->file_lock);
743 file = inode->u.proc_i.file = fcheck_files(files, fd);
744 if (!file)
745 goto out_unlock2;
746 get_file(file);
747 read_unlock(&files->file_lock);
748 put_files_struct(files);
749 inode->i_op = &proc_pid_link_inode_operations;
750 inode->i_size = 64;
751 inode->i_mode = S_IFLNK;
752 inode->u.proc_i.op.proc_get_link = proc_fd_link;
753 if (file->f_mode & 1)
754 inode->i_mode |= S_IRUSR | S_IXUSR;
755 if (file->f_mode & 2)
756 inode->i_mode |= S_IWUSR | S_IXUSR;
757 dentry->d_op = &pid_fd_dentry_operations;
758 d_add(dentry, inode);
759 return NULL;
760
761 out_unlock2:
762 put_files_struct(files);
763 read_unlock(&files->file_lock);
764 out_unlock:
765 iput(inode);
766 out:
767 return ERR_PTR(-ENOENT);
768 }
769
770 static struct file_operations proc_fd_operations = {
771 read: generic_read_dir,
772 readdir: proc_readfd,
773 };
774
775 /*
776 * proc directories can do almost nothing..
777 */
778 static struct inode_operations proc_fd_inode_operations = {
779 lookup: proc_lookupfd,
780 permission: proc_permission,
781 };
782
783 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
784 {
785 struct inode *inode;
786 int error;
787 struct task_struct *task = dir->u.proc_i.task;
788 struct pid_entry *p;
789
790 error = -ENOENT;
791 inode = NULL;
792
793 for (p = base_stuff; p->name; p++) {
794 if (p->len != dentry->d_name.len)
795 continue;
796 if (!memcmp(dentry->d_name.name, p->name, p->len))
797 break;
798 }
799 if (!p->name)
800 goto out;
801
802 error = -EINVAL;
803 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
804 if (!inode)
805 goto out;
806
807 inode->i_mode = p->mode;
808 /*
809 * Yes, it does not scale. And it should not. Don't add
810 * new entries into /proc/<pid>/ without very good reasons.
811 */
812 switch(p->type) {
813 case PROC_PID_FD:
814 inode->i_nlink = 2;
815 inode->i_op = &proc_fd_inode_operations;
816 inode->i_fop = &proc_fd_operations;
817 break;
818 case PROC_PID_EXE:
819 inode->i_op = &proc_pid_link_inode_operations;
820 inode->u.proc_i.op.proc_get_link = proc_exe_link;
821 break;
822 case PROC_PID_CWD:
823 inode->i_op = &proc_pid_link_inode_operations;
824 inode->u.proc_i.op.proc_get_link = proc_cwd_link;
825 break;
826 case PROC_PID_ROOT:
827 inode->i_op = &proc_pid_link_inode_operations;
828 inode->u.proc_i.op.proc_get_link = proc_root_link;
829 break;
830 case PROC_PID_ENVIRON:
831 inode->i_fop = &proc_info_file_operations;
832 inode->u.proc_i.op.proc_read = proc_pid_environ;
833 break;
834 case PROC_PID_STATUS:
835 inode->i_fop = &proc_info_file_operations;
836 inode->u.proc_i.op.proc_read = proc_pid_status;
837 break;
838 case PROC_PID_STAT:
839 inode->i_fop = &proc_info_file_operations;
840 inode->u.proc_i.op.proc_read = proc_pid_stat;
841 break;
842 case PROC_PID_CMDLINE:
843 inode->i_fop = &proc_info_file_operations;
844 inode->u.proc_i.op.proc_read = proc_pid_cmdline;
845 break;
846 case PROC_PID_STATM:
847 inode->i_fop = &proc_info_file_operations;
848 inode->u.proc_i.op.proc_read = proc_pid_statm;
849 break;
850 case PROC_PID_MAPS:
851 inode->i_fop = &proc_maps_operations;
852 break;
853 #ifdef CONFIG_SMP
854 case PROC_PID_CPU:
855 inode->i_fop = &proc_info_file_operations;
856 inode->u.proc_i.op.proc_read = proc_pid_cpu;
857 break;
858 #endif
859 case PROC_PID_MEM:
860 inode->i_op = &proc_mem_inode_operations;
861 inode->i_fop = &proc_mem_operations;
862 break;
863 default:
864 printk("procfs: impossible type (%d)",p->type);
865 iput(inode);
866 return ERR_PTR(-EINVAL);
867 }
868 dentry->d_op = &pid_dentry_operations;
869 d_add(dentry, inode);
870 return NULL;
871
872 out:
873 return ERR_PTR(error);
874 }
875
876 static struct file_operations proc_base_operations = {
877 read: generic_read_dir,
878 readdir: proc_base_readdir,
879 };
880
881 static struct inode_operations proc_base_inode_operations = {
882 lookup: proc_base_lookup,
883 };
884
885 /*
886 * /proc/self:
887 */
888 static int proc_self_readlink(struct dentry *dentry, char *buffer, int buflen)
889 {
890 char tmp[30];
891 sprintf(tmp, "%d", current->pid);
892 return vfs_readlink(dentry,buffer,buflen,tmp);
893 }
894
895 static int proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
896 {
897 char tmp[30];
898 sprintf(tmp, "%d", current->pid);
899 return vfs_follow_link(nd,tmp);
900 }
901
902 static struct inode_operations proc_self_inode_operations = {
903 readlink: proc_self_readlink,
904 follow_link: proc_self_follow_link,
905 };
906
907 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry)
908 {
909 unsigned int pid, c;
910 struct task_struct *task;
911 const char *name;
912 struct inode *inode;
913 int len;
914
915 pid = 0;
916 name = dentry->d_name.name;
917 len = dentry->d_name.len;
918 if (len == 4 && !memcmp(name, "self", 4)) {
919 inode = new_inode(dir->i_sb);
920 if (!inode)
921 return ERR_PTR(-ENOMEM);
922 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
923 inode->i_ino = fake_ino(0, PROC_PID_INO);
924 inode->u.proc_i.file = NULL;
925 inode->u.proc_i.task = NULL;
926 inode->i_mode = S_IFLNK|S_IRWXUGO;
927 inode->i_uid = inode->i_gid = 0;
928 inode->i_size = 64;
929 inode->i_op = &proc_self_inode_operations;
930 d_add(dentry, inode);
931 return NULL;
932 }
933 while (len-- > 0) {
934 c = *name - '0';
935 name++;
936 if (c > 9)
937 goto out;
938 if (pid >= MAX_MULBY10)
939 goto out;
940 pid *= 10;
941 pid += c;
942 if (!pid)
943 goto out;
944 }
945
946 read_lock(&tasklist_lock);
947 task = find_task_by_pid(pid);
948 if (task)
949 get_task_struct(task);
950 read_unlock(&tasklist_lock);
951 if (!task)
952 goto out;
953
954 inode = proc_pid_make_inode(dir->i_sb, task, PROC_PID_INO);
955
956 free_task_struct(task);
957
958 if (!inode)
959 goto out;
960 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
961 inode->i_op = &proc_base_inode_operations;
962 inode->i_fop = &proc_base_operations;
963 inode->i_nlink = 3;
964 inode->i_flags|=S_IMMUTABLE;
965
966 dentry->d_op = &pid_base_dentry_operations;
967 d_add(dentry, inode);
968 return NULL;
969 out:
970 return ERR_PTR(-ENOENT);
971 }
972
973 void proc_pid_delete_inode(struct inode *inode)
974 {
975 if (inode->u.proc_i.file)
976 fput(inode->u.proc_i.file);
977 if (inode->u.proc_i.task)
978 free_task_struct(inode->u.proc_i.task);
979 }
980
981 #define PROC_NUMBUF 10
982 #define PROC_MAXPIDS 20
983
984 /*
985 * Get a few pid's to return for filldir - we need to hold the
986 * tasklist lock while doing this, and we must release it before
987 * we actually do the filldir itself, so we use a temp buffer..
988 */
989 static int get_pid_list(int index, unsigned int *pids)
990 {
991 struct task_struct *p;
992 int nr_pids = 0;
993
994 index--;
995 read_lock(&tasklist_lock);
996 for_each_task(p) {
997 int pid = p->pid;
998 if (!pid)
999 continue;
1000 if (--index >= 0)
1001 continue;
1002 pids[nr_pids] = pid;
1003 nr_pids++;
1004 if (nr_pids >= PROC_MAXPIDS)
1005 break;
1006 }
1007 read_unlock(&tasklist_lock);
1008 return nr_pids;
1009 }
1010
1011 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
1012 {
1013 unsigned int pid_array[PROC_MAXPIDS];
1014 char buf[PROC_NUMBUF];
1015 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
1016 unsigned int nr_pids, i;
1017
1018 if (!nr) {
1019 ino_t ino = fake_ino(0,PROC_PID_INO);
1020 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
1021 return 0;
1022 filp->f_pos++;
1023 nr++;
1024 }
1025
1026 nr_pids = get_pid_list(nr, pid_array);
1027
1028 for (i = 0; i < nr_pids; i++) {
1029 int pid = pid_array[i];
1030 ino_t ino = fake_ino(pid,PROC_PID_INO);
1031 unsigned long j = PROC_NUMBUF;
1032
1033 do buf[--j] = '0' + (pid % 10); while (pid/=10);
1034
1035 if (filldir(dirent, buf+j, PROC_NUMBUF-j, filp->f_pos, ino, DT_DIR) < 0)
1036 break;
1037 filp->f_pos++;
1038 }
1039 return 0;
1040 }
Copy of Dave Jones history.git as used by http://archive.org/details/git-history-of-linux