Add IFF_MONITOR support.
[dragonfly.git] / sys / vfs / procfs / procfs_vnops.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1993, 1995 Jan-Simon Pendry
3 * Copyright (c) 1993, 1995
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * Jan-Simon Pendry.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)procfs_vnops.c 8.18 (Berkeley) 5/21/95
38 *
39 * $FreeBSD: src/sys/miscfs/procfs/procfs_vnops.c,v 1.76.2.7 2002/01/22 17:22:59 nectar Exp $
527fddf7 40 * $DragonFly: src/sys/vfs/procfs/procfs_vnops.c,v 1.39 2006/11/07 20:48:15 dillon Exp $
984263bc
MD
41 */
42
43/*
44 * procfs vnode interface
45 */
46
47#include <sys/param.h>
48#include <sys/systm.h>
49#include <sys/time.h>
50#include <sys/kernel.h>
51#include <sys/lock.h>
52#include <sys/fcntl.h>
53#include <sys/proc.h>
54#include <sys/signalvar.h>
55#include <sys/vnode.h>
56#include <sys/uio.h>
57#include <sys/mount.h>
58#include <sys/namei.h>
59#include <sys/dirent.h>
60#include <sys/malloc.h>
527fddf7 61#include <sys/reg.h>
984263bc 62#include <vm/vm_zone.h>
1f2de5d4 63#include <vfs/procfs/procfs.h>
984263bc
MD
64#include <sys/pioctl.h>
65
dcab9ef0
JS
66#include <machine/limits.h>
67
a6ee311a 68static int procfs_access (struct vop_access_args *);
66a1ddf5 69static int procfs_badop (struct vop_generic_args *);
a6ee311a
RG
70static int procfs_bmap (struct vop_bmap_args *);
71static int procfs_close (struct vop_close_args *);
72static int procfs_getattr (struct vop_getattr_args *);
73static int procfs_inactive (struct vop_inactive_args *);
74static int procfs_ioctl (struct vop_ioctl_args *);
e62afb5f 75static int procfs_lookup (struct vop_old_lookup_args *);
a6ee311a
RG
76static int procfs_open (struct vop_open_args *);
77static int procfs_print (struct vop_print_args *);
78static int procfs_readdir (struct vop_readdir_args *);
79static int procfs_readlink (struct vop_readlink_args *);
80static int procfs_reclaim (struct vop_reclaim_args *);
81static int procfs_setattr (struct vop_setattr_args *);
984263bc 82
dcab9ef0
JS
83static int procfs_readdir_proc(struct vop_readdir_args *);
84static int procfs_readdir_root(struct vop_readdir_args *);
85
66a1ddf5
MD
86/*
87 * procfs vnode operations.
88 */
89struct vop_ops procfs_vnode_vops = {
90 .vop_default = vop_defaultop,
91 .vop_access = procfs_access,
92 .vop_advlock = (void *)procfs_badop,
93 .vop_bmap = procfs_bmap,
94 .vop_close = procfs_close,
95 .vop_old_create = (void *)procfs_badop,
96 .vop_getattr = procfs_getattr,
97 .vop_inactive = procfs_inactive,
98 .vop_old_link = (void *)procfs_badop,
99 .vop_old_lookup = procfs_lookup,
100 .vop_old_mkdir = (void *)procfs_badop,
101 .vop_old_mknod = (void *)procfs_badop,
102 .vop_open = procfs_open,
103 .vop_pathconf = vop_stdpathconf,
104 .vop_print = procfs_print,
105 .vop_read = procfs_rw,
106 .vop_readdir = procfs_readdir,
107 .vop_readlink = procfs_readlink,
108 .vop_reclaim = procfs_reclaim,
109 .vop_old_remove = (void *)procfs_badop,
110 .vop_old_rename = (void *)procfs_badop,
111 .vop_old_rmdir = (void *)procfs_badop,
112 .vop_setattr = procfs_setattr,
113 .vop_old_symlink = (void *)procfs_badop,
114 .vop_write = (void *)procfs_rw,
115 .vop_ioctl = procfs_ioctl
116};
117
118
984263bc
MD
119/*
120 * This is a list of the valid names in the
121 * process-specific sub-directories. It is
122 * used in procfs_lookup and procfs_readdir
123 */
124static struct proc_target {
125 u_char pt_type;
126 u_char pt_namlen;
127 char *pt_name;
128 pfstype pt_pfstype;
a6ee311a 129 int (*pt_valid) (struct proc *p);
984263bc
MD
130} proc_targets[] = {
131#define N(s) sizeof(s)-1, s
132 /* name type validp */
133 { DT_DIR, N("."), Pproc, NULL },
134 { DT_DIR, N(".."), Proot, NULL },
135 { DT_REG, N("mem"), Pmem, NULL },
136 { DT_REG, N("regs"), Pregs, procfs_validregs },
137 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs },
138 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs },
139 { DT_REG, N("ctl"), Pctl, NULL },
140 { DT_REG, N("status"), Pstatus, NULL },
141 { DT_REG, N("note"), Pnote, NULL },
142 { DT_REG, N("notepg"), Pnotepg, NULL },
143 { DT_REG, N("map"), Pmap, procfs_validmap },
144 { DT_REG, N("etype"), Ptype, procfs_validtype },
145 { DT_REG, N("cmdline"), Pcmdline, NULL },
146 { DT_REG, N("rlimit"), Prlimit, NULL },
147 { DT_LNK, N("file"), Pfile, NULL },
148#undef N
149};
150static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]);
151
a6ee311a 152static pid_t atopid (const char *, u_int);
984263bc
MD
153
154/*
155 * set things up for doing i/o on
156 * the pfsnode (vp). (vp) is locked
157 * on entry, and should be left locked
158 * on exit.
159 *
160 * for procfs we don't need to do anything
161 * in particular for i/o. all that is done
162 * is to support exclusive open on process
163 * memory images.
ac424f9b 164 *
87de5057 165 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred)
984263bc
MD
166 */
167static int
ac424f9b 168procfs_open(struct vop_open_args *ap)
984263bc
MD
169{
170 struct pfsnode *pfs = VTOPFS(ap->a_vp);
171 struct proc *p1, *p2;
172
173 p2 = PFIND(pfs->pfs_pid);
174 if (p2 == NULL)
175 return (ENOENT);
41c20dac 176 if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred))
984263bc
MD
177 return (ENOENT);
178
179 switch (pfs->pfs_type) {
180 case Pmem:
181 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
182 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE)))
183 return (EBUSY);
184
87de5057 185 p1 = curproc;
dadab5e9 186 KKASSERT(p1);
984263bc
MD
187 /* Can't trace a process that's currently exec'ing. */
188 if ((p2->p_flag & P_INEXEC) != 0)
189 return EAGAIN;
41c20dac 190 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred))
984263bc
MD
191 return (EPERM);
192
193 if (ap->a_mode & FWRITE)
194 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
195
8ddc6004 196 break;
984263bc
MD
197
198 default:
199 break;
200 }
201
8ddc6004 202 return (vop_stdopen(ap));
984263bc
MD
203}
204
205/*
206 * close the pfsnode (vp) after doing i/o.
207 * (vp) is not locked on entry or exit.
208 *
209 * nothing to do for procfs other than undo
210 * any exclusive open flag (see _open above).
ac424f9b 211 *
87de5057 212 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
984263bc
MD
213 */
214static int
ac424f9b 215procfs_close(struct vop_close_args *ap)
984263bc
MD
216{
217 struct pfsnode *pfs = VTOPFS(ap->a_vp);
218 struct proc *p;
219
220 switch (pfs->pfs_type) {
221 case Pmem:
222 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
223 pfs->pfs_flags &= ~(FWRITE|O_EXCL);
224 /*
225 * This rather complicated-looking code is trying to
226 * determine if this was the last close on this particular
227 * vnode. While one would expect v_usecount to be 1 at
228 * that point, it seems that (according to John Dyson)
229 * the VM system will bump up the usecount. So: if the
230 * usecount is 2, and VOBJBUF is set, then this is really
231 * the last close. Otherwise, if the usecount is < 2
232 * then it is definitely the last close.
233 * If this is the last close, then it checks to see if
234 * the target process has PF_LINGER set in p_pfsflags,
235 * if this is *not* the case, then the process' stop flags
236 * are cleared, and the process is woken up. This is
237 * to help prevent the case where a process has been
238 * told to stop on an event, but then the requesting process
239 * has gone away or forgotten about it.
240 */
241 if ((ap->a_vp->v_usecount < 2)
242 && (p = pfind(pfs->pfs_pid))
243 && !(p->p_pfsflags & PF_LINGER)) {
244 p->p_stops = 0;
245 p->p_step = 0;
246 wakeup(&p->p_step);
247 }
248 break;
249 default:
250 break;
251 }
252
8ddc6004 253 return (vop_stdclose(ap));
984263bc
MD
254}
255
256/*
257 * do an ioctl operation on a pfsnode (vp).
258 * (vp) is not locked on entry or exit.
259 */
260static int
ac424f9b 261procfs_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
262{
263 struct pfsnode *pfs = VTOPFS(ap->a_vp);
dadab5e9
MD
264 struct proc *procp;
265 struct proc *p;
984263bc
MD
266 int error;
267 int signo;
268 struct procfs_status *psp;
269 unsigned char flags;
270
984263bc 271 procp = pfind(pfs->pfs_pid);
dadab5e9 272 if (procp == NULL)
984263bc 273 return ENOTTY;
87de5057 274 p = curproc;
dadab5e9
MD
275 if (p == NULL)
276 return EINVAL;
984263bc
MD
277
278 /* Can't trace a process that's currently exec'ing. */
279 if ((procp->p_flag & P_INEXEC) != 0)
280 return EAGAIN;
41c20dac 281 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred))
984263bc
MD
282 return EPERM;
283
284 switch (ap->a_command) {
285 case PIOCBIS:
286 procp->p_stops |= *(unsigned int*)ap->a_data;
287 break;
288 case PIOCBIC:
289 procp->p_stops &= ~*(unsigned int*)ap->a_data;
290 break;
291 case PIOCSFL:
292 /*
293 * NFLAGS is "non-suser_xxx flags" -- currently, only
294 * PFS_ISUGID ("ignore set u/g id");
295 */
296#define NFLAGS (PF_ISUGID)
297 flags = (unsigned char)*(unsigned int*)ap->a_data;
dadab5e9 298 if (flags & NFLAGS && (error = suser_cred(ap->a_cred, 0)))
984263bc
MD
299 return error;
300 procp->p_pfsflags = flags;
301 break;
302 case PIOCGFL:
303 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
304 break;
305 case PIOCSTATUS:
306 psp = (struct procfs_status *)ap->a_data;
307 psp->state = (procp->p_step == 0);
308 psp->flags = procp->p_pfsflags;
309 psp->events = procp->p_stops;
310 if (procp->p_step) {
311 psp->why = procp->p_stype;
312 psp->val = procp->p_xstat;
313 } else {
314 psp->why = psp->val = 0; /* Not defined values */
315 }
316 break;
317 case PIOCWAIT:
318 psp = (struct procfs_status *)ap->a_data;
319 if (procp->p_step == 0) {
377d4740 320 error = tsleep(&procp->p_stype, PCATCH, "piocwait", 0);
984263bc
MD
321 if (error)
322 return error;
323 }
324 psp->state = 1; /* It stopped */
325 psp->flags = procp->p_pfsflags;
326 psp->events = procp->p_stops;
327 psp->why = procp->p_stype; /* why it stopped */
328 psp->val = procp->p_xstat; /* any extra info */
329 break;
330 case PIOCCONT: /* Restart a proc */
331 if (procp->p_step == 0)
332 return EINVAL; /* Can only start a stopped process */
333 if ((signo = *(int*)ap->a_data) != 0) {
334 if (signo >= NSIG || signo <= 0)
335 return EINVAL;
84204577 336 ksignal(procp, signo);
984263bc
MD
337 }
338 procp->p_step = 0;
339 wakeup(&procp->p_step);
340 break;
341 default:
342 return (ENOTTY);
343 }
344 return 0;
345}
346
347/*
348 * do block mapping for pfsnode (vp).
349 * since we don't use the buffer cache
350 * for procfs this function should never
351 * be called. in any case, it's not clear
352 * what part of the kernel ever makes use
353 * of this function. for sanity, this is the
354 * usual no-op bmap, although returning
355 * (EIO) would be a reasonable alternative.
ac424f9b 356 *
54078292
MD
357 * procfs_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
358 * off_t *a_doffsetp, int *a_runp)
984263bc
MD
359 */
360static int
ac424f9b 361procfs_bmap(struct vop_bmap_args *ap)
984263bc 362{
984263bc
MD
363 if (ap->a_vpp != NULL)
364 *ap->a_vpp = ap->a_vp;
54078292
MD
365 if (ap->a_doffsetp != NULL)
366 *ap->a_doffsetp = ap->a_loffset;
984263bc
MD
367 if (ap->a_runp != NULL)
368 *ap->a_runp = 0;
54078292
MD
369 if (ap->a_runb != NULL)
370 *ap->a_runb = 0;
984263bc
MD
371 return (0);
372}
373
374/*
375 * procfs_inactive is called when the pfsnode
376 * is vrele'd and the reference count goes
377 * to zero. (vp) will be on the vnode free
378 * list, so to get it back vget() must be
379 * used.
380 *
381 * (vp) is locked on entry, but must be unlocked on exit.
ac424f9b
CP
382 *
383 * procfs_inactive(struct vnode *a_vp, struct thread *a_td)
984263bc
MD
384 */
385static int
ac424f9b 386procfs_inactive(struct vop_inactive_args *ap)
984263bc 387{
5fd012e0 388 /*struct vnode *vp = ap->a_vp;*/
984263bc
MD
389
390 return (0);
391}
392
393/*
394 * _reclaim is called when getnewvnode()
395 * wants to make use of an entry on the vnode
396 * free list. at this time the filesystem needs
397 * to free any private data and remove the node
398 * from any private lists.
ac424f9b
CP
399 *
400 * procfs_reclaim(struct vnode *a_vp)
984263bc
MD
401 */
402static int
ac424f9b 403procfs_reclaim(struct vop_reclaim_args *ap)
984263bc 404{
984263bc
MD
405 return (procfs_freevp(ap->a_vp));
406}
407
408/*
409 * _print is used for debugging.
410 * just print a readable description
411 * of (vp).
ac424f9b
CP
412 *
413 * procfs_print(struct vnode *a_vp)
984263bc
MD
414 */
415static int
ac424f9b 416procfs_print(struct vop_print_args *ap)
984263bc
MD
417{
418 struct pfsnode *pfs = VTOPFS(ap->a_vp);
419
420 printf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
421 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
422 return (0);
423}
424
425/*
426 * generic entry point for unsupported operations
427 */
428static int
66a1ddf5 429procfs_badop(struct vop_generic_args *ap)
984263bc 430{
984263bc
MD
431 return (EIO);
432}
433
434/*
435 * Invent attributes for pfsnode (vp) and store
436 * them in (vap).
437 * Directories lengths are returned as zero since
438 * any real length would require the genuine size
439 * to be computed, and nothing cares anyway.
440 *
441 * this is relatively minimal for procfs.
ac424f9b
CP
442 *
443 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap,
444 * struct ucred *a_cred, struct thread *a_td)
984263bc
MD
445 */
446static int
ac424f9b 447procfs_getattr(struct vop_getattr_args *ap)
984263bc
MD
448{
449 struct pfsnode *pfs = VTOPFS(ap->a_vp);
450 struct vattr *vap = ap->a_vap;
451 struct proc *procp;
452 int error;
453
454 /*
455 * First make sure that the process and its credentials
456 * still exist.
457 */
458 switch (pfs->pfs_type) {
459 case Proot:
460 case Pcurproc:
461 procp = 0;
462 break;
463
464 default:
465 procp = PFIND(pfs->pfs_pid);
41c20dac 466 if (procp == NULL || procp->p_ucred == NULL)
984263bc
MD
467 return (ENOENT);
468 }
469
470 error = 0;
471
472 /* start by zeroing out the attributes */
473 VATTR_NULL(vap);
474
475 /* next do all the common fields */
476 vap->va_type = ap->a_vp->v_type;
477 vap->va_mode = pfs->pfs_mode;
478 vap->va_fileid = pfs->pfs_fileno;
479 vap->va_flags = 0;
480 vap->va_blocksize = PAGE_SIZE;
481 vap->va_bytes = vap->va_size = 0;
482 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
483
484 /*
485 * Make all times be current TOD.
486 * It would be possible to get the process start
487 * time from the p_stat structure, but there's
488 * no "file creation" time stamp anyway, and the
489 * p_stat structure is not addressible if u. gets
490 * swapped out for that process.
491 */
492 nanotime(&vap->va_ctime);
493 vap->va_atime = vap->va_mtime = vap->va_ctime;
494
495 /*
496 * If the process has exercised some setuid or setgid
497 * privilege, then rip away read/write permission so
498 * that only root can gain access.
499 */
500 switch (pfs->pfs_type) {
501 case Pctl:
502 case Pregs:
503 case Pfpregs:
504 case Pdbregs:
505 case Pmem:
506 if (procp->p_flag & P_SUGID)
507 vap->va_mode &= ~((VREAD|VWRITE)|
508 ((VREAD|VWRITE)>>3)|
509 ((VREAD|VWRITE)>>6));
510 break;
511 default:
512 break;
513 }
514
515 /*
516 * now do the object specific fields
517 *
518 * The size could be set from struct reg, but it's hardly
519 * worth the trouble, and it puts some (potentially) machine
520 * dependent data into this machine-independent code. If it
521 * becomes important then this function should break out into
522 * a per-file stat function in the corresponding .c file.
523 */
524
525 vap->va_nlink = 1;
526 if (procp) {
527 vap->va_uid = procp->p_ucred->cr_uid;
528 vap->va_gid = procp->p_ucred->cr_gid;
529 }
530
531 switch (pfs->pfs_type) {
532 case Proot:
533 /*
534 * Set nlink to 1 to tell fts(3) we don't actually know.
535 */
536 vap->va_nlink = 1;
537 vap->va_uid = 0;
538 vap->va_gid = 0;
539 vap->va_size = vap->va_bytes = DEV_BSIZE;
540 break;
541
542 case Pcurproc: {
543 char buf[16]; /* should be enough */
544 vap->va_uid = 0;
545 vap->va_gid = 0;
546 vap->va_size = vap->va_bytes =
547 snprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
548 break;
549 }
550
551 case Pproc:
552 vap->va_nlink = nproc_targets;
553 vap->va_size = vap->va_bytes = DEV_BSIZE;
554 break;
555
556 case Pfile: {
557 char *fullpath, *freepath;
541044dc 558 error = vn_fullpath(procp, NULL, &fullpath, &freepath);
984263bc
MD
559 if (error == 0) {
560 vap->va_size = strlen(fullpath);
efda3bd0 561 kfree(freepath, M_TEMP);
984263bc
MD
562 } else {
563 vap->va_size = sizeof("unknown") - 1;
564 error = 0;
565 }
566 vap->va_bytes = vap->va_size;
567 break;
568 }
569
570 case Pmem:
571 /*
572 * If we denied owner access earlier, then we have to
573 * change the owner to root - otherwise 'ps' and friends
574 * will break even though they are setgid kmem. *SIGH*
575 */
576 if (procp->p_flag & P_SUGID)
577 vap->va_uid = 0;
578 else
579 vap->va_uid = procp->p_ucred->cr_uid;
580 break;
581
582 case Pregs:
583 vap->va_bytes = vap->va_size = sizeof(struct reg);
584 break;
585
586 case Pfpregs:
587 vap->va_bytes = vap->va_size = sizeof(struct fpreg);
588 break;
589
590 case Pdbregs:
591 vap->va_bytes = vap->va_size = sizeof(struct dbreg);
592 break;
593
594 case Ptype:
595 case Pmap:
596 case Pctl:
597 case Pstatus:
598 case Pnote:
599 case Pnotepg:
600 case Pcmdline:
601 case Prlimit:
602 break;
603
604 default:
605 panic("procfs_getattr");
606 }
607
608 return (error);
609}
610
ac424f9b
CP
611/*
612 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap,
613 * struct ucred *a_cred, struct thread *a_td)
614 */
984263bc 615static int
ac424f9b 616procfs_setattr(struct vop_setattr_args *ap)
984263bc 617{
984263bc
MD
618 if (ap->a_vap->va_flags != VNOVAL)
619 return (EOPNOTSUPP);
620
621 /*
622 * just fake out attribute setting
623 * it's not good to generate an error
624 * return, otherwise things like creat()
625 * will fail when they try to set the
626 * file length to 0. worse, this means
627 * that echo $note > /proc/$pid/note will fail.
628 */
629
630 return (0);
631}
632
633/*
634 * implement access checking.
635 *
636 * something very similar to this code is duplicated
637 * throughout the 4bsd kernel and should be moved
638 * into kern/vfs_subr.c sometime.
639 *
640 * actually, the check for super-user is slightly
641 * broken since it will allow read access to write-only
642 * objects. this doesn't cause any particular trouble
643 * but does mean that the i/o entry points need to check
644 * that the operation really does make sense.
ac424f9b
CP
645 *
646 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
647 * struct thread *a_td)
984263bc
MD
648 */
649static int
ac424f9b 650procfs_access(struct vop_access_args *ap)
984263bc
MD
651{
652 struct vattr *vap;
653 struct vattr vattr;
654 int error;
655
656 /*
657 * If you're the super-user,
658 * you always get access.
659 */
660 if (ap->a_cred->cr_uid == 0)
661 return (0);
662
663 vap = &vattr;
87de5057 664 error = VOP_GETATTR(ap->a_vp, vap);
984263bc
MD
665 if (error)
666 return (error);
667
668 /*
669 * Access check is based on only one of owner, group, public.
670 * If not owner, then check group. If not a member of the
671 * group, then check public access.
672 */
673 if (ap->a_cred->cr_uid != vap->va_uid) {
674 gid_t *gp;
675 int i;
676
677 ap->a_mode >>= 3;
678 gp = ap->a_cred->cr_groups;
679 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
680 if (vap->va_gid == *gp)
681 goto found;
682 ap->a_mode >>= 3;
683found:
684 ;
685 }
686
687 if ((vap->va_mode & ap->a_mode) == ap->a_mode)
688 return (0);
689
690 return (EACCES);
691}
692
693/*
3446c007
MD
694 * lookup. this is incredibly complicated in the general case, however
695 * for most pseudo-filesystems very little needs to be done.
ac424f9b
CP
696 *
697 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
698 * struct componentname *a_cnp)
984263bc
MD
699 */
700static int
e62afb5f 701procfs_lookup(struct vop_old_lookup_args *ap)
984263bc
MD
702{
703 struct componentname *cnp = ap->a_cnp;
704 struct vnode **vpp = ap->a_vpp;
705 struct vnode *dvp = ap->a_dvp;
706 char *pname = cnp->cn_nameptr;
707 /* struct proc *curp = cnp->cn_proc; */
708 struct proc_target *pt;
709 pid_t pid;
710 struct pfsnode *pfs;
711 struct proc *p;
712 int i;
3446c007 713 int error;
984263bc
MD
714
715 *vpp = NULL;
716
2b69e610 717 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME)
984263bc
MD
718 return (EROFS);
719
3446c007 720 error = 0;
984263bc
MD
721 if (cnp->cn_namelen == 1 && *pname == '.') {
722 *vpp = dvp;
3446c007
MD
723 vref(*vpp);
724 goto out;
984263bc
MD
725 }
726
727 pfs = VTOPFS(dvp);
728 switch (pfs->pfs_type) {
729 case Proot:
2b69e610 730 if (cnp->cn_flags & CNP_ISDOTDOT)
984263bc
MD
731 return (EIO);
732
3446c007
MD
733 if (CNEQ(cnp, "curproc", 7)) {
734 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc);
735 goto out;
736 }
984263bc
MD
737
738 pid = atopid(pname, cnp->cn_namelen);
739 if (pid == NO_PID)
740 break;
741
742 p = PFIND(pid);
743 if (p == NULL)
744 break;
745
dcab9ef0
JS
746 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
747 break;
748
28c57d20
HP
749 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
750 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
751 break;
752
3446c007
MD
753 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
754 goto out;
984263bc
MD
755
756 case Pproc:
3446c007
MD
757 if (cnp->cn_flags & CNP_ISDOTDOT) {
758 error = procfs_root(dvp->v_mount, vpp);
759 goto out;
760 }
984263bc
MD
761
762 p = PFIND(pfs->pfs_pid);
763 if (p == NULL)
764 break;
765
dcab9ef0
JS
766 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
767 break;
768
28c57d20
HP
769 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
770 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
771 break;
772
984263bc
MD
773 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
774 if (cnp->cn_namelen == pt->pt_namlen &&
775 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
776 (pt->pt_valid == NULL || (*pt->pt_valid)(p)))
777 goto found;
778 }
779 break;
780 found:
3446c007
MD
781 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
782 pt->pt_pfstype);
783 goto out;
984263bc
MD
784
785 default:
3446c007
MD
786 error = ENOTDIR;
787 goto out;
984263bc 788 }
3446c007
MD
789 if (cnp->cn_nameiop == NAMEI_LOOKUP)
790 error = ENOENT;
791 else
792 error = EROFS;
793 /*
794 * If no error occured *vpp will hold a referenced locked vnode.
aa723195
MD
795 * dvp was passed to us locked and *vpp must be returned locked.
796 * If *vpp != dvp then we should unlock dvp if (1) this is not the
797 * last component or (2) CNP_LOCKPARENT is not set.
3446c007
MD
798 */
799out:
aa723195
MD
800 if (error == 0 && *vpp != dvp) {
801 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) {
802 cnp->cn_flags |= CNP_PDIRUNLOCK;
a11aaa81 803 vn_unlock(dvp);
3446c007
MD
804 }
805 }
806 return (error);
984263bc
MD
807}
808
809/*
810 * Does this process have a text file?
811 */
4d4fb029 812int
ac424f9b 813procfs_validfile(struct proc *p)
984263bc 814{
984263bc
MD
815 return (procfs_findtextvp(p) != NULLVP);
816}
817
818/*
819 * readdir() returns directory entries from pfsnode (vp).
820 *
821 * We generate just one directory entry at a time, as it would probably
822 * not pay off to buffer several entries locally to save uiomove calls.
ac424f9b
CP
823 *
824 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
825 * int *a_eofflag, int *a_ncookies, u_long **a_cookies)
984263bc
MD
826 */
827static int
ac424f9b 828procfs_readdir(struct vop_readdir_args *ap)
984263bc 829{
984263bc 830 struct pfsnode *pfs;
dcab9ef0 831 int error;
984263bc 832
dcab9ef0 833 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
984263bc 834 return (EINVAL);
885ecb13
MD
835 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0)
836 return (error);
dcab9ef0 837 pfs = VTOPFS(ap->a_vp);
984263bc
MD
838
839 switch (pfs->pfs_type) {
dcab9ef0 840 case Pproc:
885ecb13
MD
841 /*
842 * this is for the process-specific sub-directories.
843 * all that is needed to is copy out all the entries
844 * from the procent[] table (top of this file).
845 */
dcab9ef0
JS
846 error = procfs_readdir_proc(ap);
847 break;
dcab9ef0 848 case Proot:
885ecb13
MD
849 /*
850 * this is for the root of the procfs filesystem
851 * what is needed is a special entry for "curproc"
852 * followed by an entry for each process on allproc
853 */
dcab9ef0 854 error = procfs_readdir_root(ap);
984263bc 855 break;
984263bc
MD
856 default:
857 error = ENOTDIR;
858 break;
859 }
860
885ecb13 861 vn_unlock(ap->a_vp);
984263bc
MD
862 return (error);
863}
864
dcab9ef0
JS
865static int
866procfs_readdir_proc(struct vop_readdir_args *ap)
867{
868 struct pfsnode *pfs;
869 int error, i, retval;
870 struct proc *p;
871 struct proc_target *pt;
872 struct uio *uio = ap->a_uio;
873
874 pfs = VTOPFS(ap->a_vp);
875 p = PFIND(pfs->pfs_pid);
876 if (p == NULL)
877 return(0);
878 if (!PRISON_CHECK(ap->a_cred, p->p_ucred))
879 return(0);
880
881 error = 0;
fb0466c9
MD
882 i = (int)uio->uio_offset;
883 if (i < 0)
884 return (EINVAL);
dcab9ef0
JS
885
886 for (pt = &proc_targets[i];
887 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
888 if (pt->pt_valid && (*pt->pt_valid)(p) == 0)
889 continue;
890
891 retval = vop_write_dirent(&error, uio,
892 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
893 pt->pt_namlen, pt->pt_name);
894 if (retval)
895 break;
896 }
897
fb0466c9 898 uio->uio_offset = (off_t)i;
dcab9ef0
JS
899
900 return(0);
901}
902
51e64ff2
MD
903struct procfs_readdir_root_info {
904 int error;
905 int i;
906 int pcnt;
907 struct uio *uio;
908 struct ucred *cred;
909};
910
911static int procfs_readdir_root_callback(struct proc *p, void *data);
912
dcab9ef0
JS
913static int
914procfs_readdir_root(struct vop_readdir_args *ap)
915{
51e64ff2 916 struct procfs_readdir_root_info info;
dcab9ef0 917 struct uio *uio = ap->a_uio;
51e64ff2
MD
918 int res;
919
920 info.error = 0;
fb0466c9
MD
921 info.i = (int)uio->uio_offset;
922
923 if (info.i < 0)
924 return (EINVAL);
925
51e64ff2
MD
926 info.pcnt = 0;
927 info.uio = uio;
928 info.cred = ap->a_cred;
51e64ff2
MD
929 while (info.pcnt < 3) {
930 res = procfs_readdir_root_callback(NULL, &info);
931 if (res < 0)
932 break;
933 }
934 if (res >= 0)
935 allproc_scan(procfs_readdir_root_callback, &info);
fb0466c9 936 uio->uio_offset = (off_t)info.i;
51e64ff2
MD
937
938 return (info.error);
939}
940
941static int
942procfs_readdir_root_callback(struct proc *p, void *data)
943{
944 struct procfs_readdir_root_info *info = data;
945 struct uio *uio;
946 int retval;
dcab9ef0
JS
947 ino_t d_ino;
948 const char *d_name;
949 char d_name_pid[20];
950 size_t d_namlen;
951 uint8_t d_type;
952
51e64ff2 953 uio = info->uio;
dcab9ef0 954
51e64ff2
MD
955 if (uio->uio_resid <= 0 || info->error)
956 return(-1);
dcab9ef0 957
51e64ff2
MD
958 switch (info->pcnt) {
959 case 0: /* `.' */
960 d_ino = PROCFS_FILENO(0, Proot);
961 d_name = ".";
962 d_namlen = 1;
963 d_type = DT_DIR;
964 break;
965 case 1: /* `..' */
966 d_ino = PROCFS_FILENO(0, Proot);
967 d_name = "..";
968 d_namlen = 2;
969 d_type = DT_DIR;
970 break;
dcab9ef0 971
51e64ff2
MD
972 case 2:
973 d_ino = PROCFS_FILENO(0, Pcurproc);
974 d_namlen = 7;
975 d_name = "curproc";
976 d_type = DT_LNK;
977 break;
dcab9ef0 978
51e64ff2
MD
979
980 default:
981 if (!PRISON_CHECK(info->cred, p->p_ucred))
982 return(0);
983 if (ps_showallprocs == 0 &&
984 info->cred->cr_uid != 0 &&
985 info->cred->cr_uid != p->p_ucred->cr_uid) {
986 return(0);
dcab9ef0
JS
987 }
988
51e64ff2
MD
989 /*
990 * Skip entries we have already returned (optimization)
991 */
992 if (info->pcnt < info->i) {
993 ++info->pcnt;
994 return(0);
995 }
996
997 d_ino = PROCFS_FILENO(p->p_pid, Pproc);
998 d_namlen = snprintf(d_name_pid, sizeof(d_name_pid),
999 "%ld", (long)p->p_pid);
1000 d_name = d_name_pid;
1001 d_type = DT_DIR;
1002 break;
dcab9ef0 1003 }
51e64ff2
MD
1004
1005 /*
1006 * Skip entries we have already returned (optimization)
1007 */
1008 if (info->pcnt < info->i) {
1009 ++info->pcnt;
1010 return(0);
1011 }
1012
1013 retval = vop_write_dirent(&info->error, uio,
1014 d_ino, d_type, d_namlen, d_name);
1015 if (retval)
1016 return(-1);
1017 ++info->pcnt;
1018 ++info->i;
dcab9ef0
JS
1019 return(0);
1020}
1021
984263bc
MD
1022/*
1023 * readlink reads the link of `curproc' or `file'
1024 */
1025static int
ac424f9b 1026procfs_readlink(struct vop_readlink_args *ap)
984263bc
MD
1027{
1028 char buf[16]; /* should be enough */
1029 struct proc *procp;
1030 struct vnode *vp = ap->a_vp;
1031 struct pfsnode *pfs = VTOPFS(vp);
1032 char *fullpath, *freepath;
1033 int error, len;
1034
1035 switch (pfs->pfs_type) {
1036 case Pcurproc:
1037 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
1038 return (EINVAL);
1039
1040 len = snprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
1041
1042 return (uiomove(buf, len, ap->a_uio));
1043 /*
1044 * There _should_ be no way for an entire process to disappear
1045 * from under us...
1046 */
1047 case Pfile:
1048 procp = PFIND(pfs->pfs_pid);
41c20dac 1049 if (procp == NULL || procp->p_ucred == NULL) {
984263bc
MD
1050 printf("procfs_readlink: pid %d disappeared\n",
1051 pfs->pfs_pid);
1052 return (uiomove("unknown", sizeof("unknown") - 1,
1053 ap->a_uio));
1054 }
541044dc 1055 error = vn_fullpath(procp, NULL, &fullpath, &freepath);
984263bc
MD
1056 if (error != 0)
1057 return (uiomove("unknown", sizeof("unknown") - 1,
1058 ap->a_uio));
1059 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
efda3bd0 1060 kfree(freepath, M_TEMP);
984263bc
MD
1061 return (error);
1062 default:
1063 return (EINVAL);
1064 }
1065}
1066
1067/*
1068 * convert decimal ascii to pid_t
1069 */
1070static pid_t
ac424f9b 1071atopid(const char *b, u_int len)
984263bc
MD
1072{
1073 pid_t p = 0;
1074
1075 while (len--) {
1076 char c = *b++;
1077 if (c < '0' || c > '9')
1078 return (NO_PID);
1079 p = 10 * p + (c - '0');
1080 if (p > PID_MAX)
1081 return (NO_PID);
1082 }
1083
1084 return (p);
1085}
1086