Kernel - Fix numerous procfs/ptrace issues
[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 $
40 */
41
42/*
43 * procfs vnode interface
44 */
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/time.h>
49#include <sys/kernel.h>
50#include <sys/lock.h>
51#include <sys/fcntl.h>
52#include <sys/proc.h>
895c1f85 53#include <sys/priv.h>
984263bc
MD
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
e193508c
MD
66#include <sys/spinlock2.h>
67
dcab9ef0
JS
68#include <machine/limits.h>
69
a6ee311a 70static int procfs_access (struct vop_access_args *);
66a1ddf5 71static int procfs_badop (struct vop_generic_args *);
a6ee311a
RG
72static int procfs_bmap (struct vop_bmap_args *);
73static int procfs_close (struct vop_close_args *);
74static int procfs_getattr (struct vop_getattr_args *);
75static int procfs_inactive (struct vop_inactive_args *);
76static int procfs_ioctl (struct vop_ioctl_args *);
e62afb5f 77static int procfs_lookup (struct vop_old_lookup_args *);
a6ee311a
RG
78static int procfs_open (struct vop_open_args *);
79static int procfs_print (struct vop_print_args *);
80static int procfs_readdir (struct vop_readdir_args *);
81static int procfs_readlink (struct vop_readlink_args *);
82static int procfs_reclaim (struct vop_reclaim_args *);
83static int procfs_setattr (struct vop_setattr_args *);
984263bc 84
dcab9ef0
JS
85static int procfs_readdir_proc(struct vop_readdir_args *);
86static int procfs_readdir_root(struct vop_readdir_args *);
87
984263bc 88/*
66a1ddf5
MD
89 * procfs vnode operations.
90 */
91struct vop_ops procfs_vnode_vops = {
92 .vop_default = vop_defaultop,
93 .vop_access = procfs_access,
94 .vop_advlock = (void *)procfs_badop,
95 .vop_bmap = procfs_bmap,
96 .vop_close = procfs_close,
97 .vop_old_create = (void *)procfs_badop,
98 .vop_getattr = procfs_getattr,
99 .vop_inactive = procfs_inactive,
100 .vop_old_link = (void *)procfs_badop,
101 .vop_old_lookup = procfs_lookup,
102 .vop_old_mkdir = (void *)procfs_badop,
103 .vop_old_mknod = (void *)procfs_badop,
104 .vop_open = procfs_open,
105 .vop_pathconf = vop_stdpathconf,
106 .vop_print = procfs_print,
107 .vop_read = procfs_rw,
108 .vop_readdir = procfs_readdir,
109 .vop_readlink = procfs_readlink,
110 .vop_reclaim = procfs_reclaim,
111 .vop_old_remove = (void *)procfs_badop,
112 .vop_old_rename = (void *)procfs_badop,
113 .vop_old_rmdir = (void *)procfs_badop,
114 .vop_setattr = procfs_setattr,
115 .vop_old_symlink = (void *)procfs_badop,
116 .vop_write = (void *)procfs_rw,
117 .vop_ioctl = procfs_ioctl
118};
119
120
121/*
984263bc
MD
122 * This is a list of the valid names in the
123 * process-specific sub-directories. It is
124 * used in procfs_lookup and procfs_readdir
125 */
126static struct proc_target {
127 u_char pt_type;
128 u_char pt_namlen;
129 char *pt_name;
130 pfstype pt_pfstype;
c7e98b2f 131 int (*pt_valid) (struct lwp *p);
984263bc
MD
132} proc_targets[] = {
133#define N(s) sizeof(s)-1, s
134 /* name type validp */
135 { DT_DIR, N("."), Pproc, NULL },
136 { DT_DIR, N(".."), Proot, NULL },
137 { DT_REG, N("mem"), Pmem, NULL },
138 { DT_REG, N("regs"), Pregs, procfs_validregs },
139 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs },
140 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs },
141 { DT_REG, N("ctl"), Pctl, NULL },
142 { DT_REG, N("status"), Pstatus, NULL },
143 { DT_REG, N("note"), Pnote, NULL },
144 { DT_REG, N("notepg"), Pnotepg, NULL },
145 { DT_REG, N("map"), Pmap, procfs_validmap },
146 { DT_REG, N("etype"), Ptype, procfs_validtype },
147 { DT_REG, N("cmdline"), Pcmdline, NULL },
148 { DT_REG, N("rlimit"), Prlimit, NULL },
149 { DT_LNK, N("file"), Pfile, NULL },
150#undef N
151};
c157ff7a 152static const int nproc_targets = NELEM(proc_targets);
984263bc 153
a6ee311a 154static pid_t atopid (const char *, u_int);
984263bc
MD
155
156/*
157 * set things up for doing i/o on
158 * the pfsnode (vp). (vp) is locked
159 * on entry, and should be left locked
160 * on exit.
161 *
162 * for procfs we don't need to do anything
163 * in particular for i/o. all that is done
164 * is to support exclusive open on process
165 * memory images.
ac424f9b 166 *
b478fdce
SW
167 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
168 * struct file *a_fp)
984263bc
MD
169 */
170static int
ac424f9b 171procfs_open(struct vop_open_args *ap)
984263bc
MD
172{
173 struct pfsnode *pfs = VTOPFS(ap->a_vp);
174 struct proc *p1, *p2;
58c2553a 175 int error;
984263bc 176
198c0ff2 177 p2 = pfs_pfind(pfs->pfs_pid);
984263bc
MD
178 if (p2 == NULL)
179 return (ENOENT);
58c2553a
MD
180 if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred)) {
181 error = ENOENT;
182 goto done;
183 }
984263bc
MD
184
185 switch (pfs->pfs_type) {
186 case Pmem:
187 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
58c2553a
MD
188 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) {
189 error = EBUSY;
190 goto done;
191 }
984263bc 192
87de5057 193 p1 = curproc;
dadab5e9 194 KKASSERT(p1);
984263bc 195 /* Can't trace a process that's currently exec'ing. */
4643740a 196 if ((p2->p_flags & P_INEXEC) != 0) {
58c2553a
MD
197 error = EAGAIN;
198 goto done;
199 }
200 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred)) {
201 error = EPERM;
202 goto done;
203 }
984263bc
MD
204
205 if (ap->a_mode & FWRITE)
206 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
207
8ddc6004 208 break;
984263bc
MD
209
210 default:
211 break;
212 }
58c2553a
MD
213 error = vop_stdopen(ap);
214done:
f44c73be 215 pfs_pdone(p2);
58c2553a 216 return error;
984263bc
MD
217}
218
219/*
220 * close the pfsnode (vp) after doing i/o.
221 * (vp) is not locked on entry or exit.
222 *
223 * nothing to do for procfs other than undo
224 * any exclusive open flag (see _open above).
ac424f9b 225 *
87de5057 226 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
984263bc
MD
227 */
228static int
ac424f9b 229procfs_close(struct vop_close_args *ap)
984263bc
MD
230{
231 struct pfsnode *pfs = VTOPFS(ap->a_vp);
232 struct proc *p;
233
234 switch (pfs->pfs_type) {
235 case Pmem:
236 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
237 pfs->pfs_flags &= ~(FWRITE|O_EXCL);
238 /*
3c37c940
MD
239 * v_opencount determines the last real close on the vnode.
240 *
984263bc
MD
241 * If this is the last close, then it checks to see if
242 * the target process has PF_LINGER set in p_pfsflags,
243 * if this is *not* the case, then the process' stop flags
244 * are cleared, and the process is woken up. This is
245 * to help prevent the case where a process has been
246 * told to stop on an event, but then the requesting process
247 * has gone away or forgotten about it.
248 */
58c2553a 249 p = NULL;
3c37c940 250 if ((ap->a_vp->v_opencount < 2)
b2a6ad87
MD
251 && ((p = pfs_pfind(pfs->pfs_pid)) != NULL ||
252 (p = pfs_zpfind(pfs->pfs_pid)) != NULL)
984263bc 253 && !(p->p_pfsflags & PF_LINGER)) {
287a8577 254 spin_lock(&p->p_spin);
984263bc
MD
255 p->p_stops = 0;
256 p->p_step = 0;
287a8577 257 spin_unlock(&p->p_spin);
b2a6ad87 258 wakeup(&p->p_stype);
984263bc 259 }
f44c73be 260 pfs_pdone(p);
984263bc
MD
261 break;
262 default:
263 break;
264 }
265
8ddc6004 266 return (vop_stdclose(ap));
984263bc
MD
267}
268
269/*
270 * do an ioctl operation on a pfsnode (vp).
271 * (vp) is not locked on entry or exit.
272 */
273static int
ac424f9b 274procfs_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
275{
276 struct pfsnode *pfs = VTOPFS(ap->a_vp);
dadab5e9
MD
277 struct proc *procp;
278 struct proc *p;
984263bc
MD
279 int error;
280 int signo;
281 struct procfs_status *psp;
282 unsigned char flags;
283
aa7ecbbc 284 procp = pfs_pfind(pfs->pfs_pid);
dadab5e9 285 if (procp == NULL)
984263bc 286 return ENOTTY;
87de5057 287 p = curproc;
58c2553a
MD
288 if (p == NULL) {
289 error = EINVAL;
290 goto done;
291 }
984263bc
MD
292
293 /* Can't trace a process that's currently exec'ing. */
4643740a 294 if ((procp->p_flags & P_INEXEC) != 0) {
58c2553a
MD
295 error = EAGAIN;
296 goto done;
297 }
298 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred)) {
299 error = EPERM;
300 goto done;
301 }
984263bc
MD
302
303 switch (ap->a_command) {
304 case PIOCBIS:
305 procp->p_stops |= *(unsigned int*)ap->a_data;
306 break;
307 case PIOCBIC:
308 procp->p_stops &= ~*(unsigned int*)ap->a_data;
309 break;
310 case PIOCSFL:
311 /*
312 * NFLAGS is "non-suser_xxx flags" -- currently, only
313 * PFS_ISUGID ("ignore set u/g id");
314 */
315#define NFLAGS (PF_ISUGID)
316 flags = (unsigned char)*(unsigned int*)ap->a_data;
895c1f85 317 if (flags & NFLAGS && (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)))
58c2553a 318 goto done;
984263bc
MD
319 procp->p_pfsflags = flags;
320 break;
321 case PIOCGFL:
322 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
323 break;
324 case PIOCSTATUS:
e193508c
MD
325 /*
326 * NOTE: syscall entry deals with stopevents and may run without
327 * the MP lock.
328 */
984263bc 329 psp = (struct procfs_status *)ap->a_data;
984263bc
MD
330 psp->flags = procp->p_pfsflags;
331 psp->events = procp->p_stops;
287a8577 332 spin_lock(&procp->p_spin);
984263bc 333 if (procp->p_step) {
e193508c 334 psp->state = 0;
984263bc
MD
335 psp->why = procp->p_stype;
336 psp->val = procp->p_xstat;
287a8577 337 spin_unlock(&procp->p_spin);
984263bc 338 } else {
e193508c 339 psp->state = 1;
287a8577 340 spin_unlock(&procp->p_spin);
e193508c
MD
341 psp->why = 0; /* Not defined values */
342 psp->val = 0; /* Not defined values */
984263bc
MD
343 }
344 break;
345 case PIOCWAIT:
e193508c
MD
346 /*
347 * NOTE: syscall entry deals with stopevents and may run without
348 * the MP lock.
349 */
984263bc 350 psp = (struct procfs_status *)ap->a_data;
287a8577 351 spin_lock(&procp->p_spin);
e193508c
MD
352 while (procp->p_step == 0) {
353 tsleep_interlock(&procp->p_stype, PCATCH);
287a8577 354 spin_unlock(&procp->p_spin);
b2a6ad87
MD
355 if (procp->p_stops == 0) {
356 error = EINVAL;
357 goto done;
358 }
359 if (procp->p_flags & P_POSTEXIT) {
360 error = EINVAL;
361 goto done;
362 }
363 if (procp->p_flags & P_INEXEC) {
364 error = EAGAIN;
365 goto done;
366 }
367 error = tsleep(&procp->p_stype, PCATCH | PINTERLOCKED,
368 "piocwait", 0);
984263bc 369 if (error)
58c2553a 370 goto done;
287a8577 371 spin_lock(&procp->p_spin);
984263bc 372 }
287a8577 373 spin_unlock(&procp->p_spin);
984263bc
MD
374 psp->state = 1; /* It stopped */
375 psp->flags = procp->p_pfsflags;
376 psp->events = procp->p_stops;
377 psp->why = procp->p_stype; /* why it stopped */
378 psp->val = procp->p_xstat; /* any extra info */
379 break;
380 case PIOCCONT: /* Restart a proc */
e193508c
MD
381 /*
382 * NOTE: syscall entry deals with stopevents and may run without
383 * the MP lock. However, the caller is presumably interlocked
384 * by having waited.
385 */
58c2553a
MD
386 if (procp->p_step == 0) {
387 error = EINVAL; /* Can only start a stopped process */
388 goto done;
389 }
984263bc 390 if ((signo = *(int*)ap->a_data) != 0) {
58c2553a
MD
391 if (signo >= NSIG || signo <= 0) {
392 error = EINVAL;
393 goto done;
394 }
84204577 395 ksignal(procp, signo);
984263bc
MD
396 }
397 procp->p_step = 0;
398 wakeup(&procp->p_step);
399 break;
400 default:
58c2553a
MD
401 error = ENOTTY;
402 goto done;
984263bc 403 }
58c2553a
MD
404 error = 0;
405done:
f44c73be 406 pfs_pdone(procp);
984263bc
MD
407 return 0;
408}
409
410/*
411 * do block mapping for pfsnode (vp).
412 * since we don't use the buffer cache
413 * for procfs this function should never
414 * be called. in any case, it's not clear
415 * what part of the kernel ever makes use
416 * of this function. for sanity, this is the
417 * usual no-op bmap, although returning
418 * (EIO) would be a reasonable alternative.
ac424f9b 419 *
08daea96
MD
420 * XXX mmap assumes buffer cache operation
421 *
422 * procfs_bmap(struct vnode *a_vp, off_t a_loffset,
423 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
424 */
425static int
ac424f9b 426procfs_bmap(struct vop_bmap_args *ap)
984263bc 427{
54078292
MD
428 if (ap->a_doffsetp != NULL)
429 *ap->a_doffsetp = ap->a_loffset;
984263bc
MD
430 if (ap->a_runp != NULL)
431 *ap->a_runp = 0;
54078292
MD
432 if (ap->a_runb != NULL)
433 *ap->a_runb = 0;
984263bc
MD
434 return (0);
435}
436
437/*
438 * procfs_inactive is called when the pfsnode
439 * is vrele'd and the reference count goes
440 * to zero. (vp) will be on the vnode free
441 * list, so to get it back vget() must be
442 * used.
443 *
444 * (vp) is locked on entry, but must be unlocked on exit.
ac424f9b 445 *
b478fdce 446 * procfs_inactive(struct vnode *a_vp)
984263bc
MD
447 */
448static int
ac424f9b 449procfs_inactive(struct vop_inactive_args *ap)
984263bc 450{
e193508c 451 struct pfsnode *pfs = VTOPFS(ap->a_vp);
984263bc 452
e193508c
MD
453 if (pfs->pfs_pid & PFS_DEAD)
454 vrecycle(ap->a_vp);
984263bc
MD
455 return (0);
456}
457
458/*
459 * _reclaim is called when getnewvnode()
460 * wants to make use of an entry on the vnode
461 * free list. at this time the filesystem needs
462 * to free any private data and remove the node
463 * from any private lists.
ac424f9b
CP
464 *
465 * procfs_reclaim(struct vnode *a_vp)
984263bc
MD
466 */
467static int
ac424f9b 468procfs_reclaim(struct vop_reclaim_args *ap)
984263bc 469{
984263bc
MD
470 return (procfs_freevp(ap->a_vp));
471}
472
473/*
474 * _print is used for debugging.
475 * just print a readable description
476 * of (vp).
ac424f9b
CP
477 *
478 * procfs_print(struct vnode *a_vp)
984263bc
MD
479 */
480static int
ac424f9b 481procfs_print(struct vop_print_args *ap)
984263bc
MD
482{
483 struct pfsnode *pfs = VTOPFS(ap->a_vp);
484
086c1d7e 485 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
984263bc
MD
486 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
487 return (0);
488}
489
490/*
491 * generic entry point for unsupported operations
492 */
493static int
66a1ddf5 494procfs_badop(struct vop_generic_args *ap)
984263bc 495{
984263bc
MD
496 return (EIO);
497}
498
499/*
500 * Invent attributes for pfsnode (vp) and store
501 * them in (vap).
502 * Directories lengths are returned as zero since
503 * any real length would require the genuine size
504 * to be computed, and nothing cares anyway.
505 *
506 * this is relatively minimal for procfs.
ac424f9b 507 *
b478fdce 508 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap)
984263bc
MD
509 */
510static int
ac424f9b 511procfs_getattr(struct vop_getattr_args *ap)
984263bc
MD
512{
513 struct pfsnode *pfs = VTOPFS(ap->a_vp);
514 struct vattr *vap = ap->a_vap;
515 struct proc *procp;
516 int error;
517
518 /*
519 * First make sure that the process and its credentials
520 * still exist.
521 */
522 switch (pfs->pfs_type) {
523 case Proot:
524 case Pcurproc:
58c2553a 525 procp = NULL;
984263bc 526 break;
984263bc 527 default:
198c0ff2 528 procp = pfs_pfind(pfs->pfs_pid);
58c2553a
MD
529 if (procp == NULL || procp->p_ucred == NULL) {
530 error = ENOENT;
531 goto done;
532 }
f44c73be 533 break;
984263bc
MD
534 }
535
536 error = 0;
537
538 /* start by zeroing out the attributes */
539 VATTR_NULL(vap);
540
541 /* next do all the common fields */
542 vap->va_type = ap->a_vp->v_type;
543 vap->va_mode = pfs->pfs_mode;
544 vap->va_fileid = pfs->pfs_fileno;
545 vap->va_flags = 0;
546 vap->va_blocksize = PAGE_SIZE;
547 vap->va_bytes = vap->va_size = 0;
548 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
549
550 /*
551 * Make all times be current TOD.
552 * It would be possible to get the process start
553 * time from the p_stat structure, but there's
554 * no "file creation" time stamp anyway, and the
555 * p_stat structure is not addressible if u. gets
556 * swapped out for that process.
557 */
558 nanotime(&vap->va_ctime);
559 vap->va_atime = vap->va_mtime = vap->va_ctime;
560
561 /*
562 * If the process has exercised some setuid or setgid
563 * privilege, then rip away read/write permission so
564 * that only root can gain access.
565 */
566 switch (pfs->pfs_type) {
567 case Pctl:
568 case Pregs:
569 case Pfpregs:
570 case Pdbregs:
571 case Pmem:
f44c73be 572 if (procp->p_flags & P_SUGID) {
984263bc
MD
573 vap->va_mode &= ~((VREAD|VWRITE)|
574 ((VREAD|VWRITE)>>3)|
575 ((VREAD|VWRITE)>>6));
f44c73be 576 }
984263bc
MD
577 break;
578 default:
579 break;
580 }
581
582 /*
583 * now do the object specific fields
584 *
585 * The size could be set from struct reg, but it's hardly
586 * worth the trouble, and it puts some (potentially) machine
587 * dependent data into this machine-independent code. If it
588 * becomes important then this function should break out into
589 * a per-file stat function in the corresponding .c file.
590 */
591
592 vap->va_nlink = 1;
593 if (procp) {
f44c73be
MD
594 if (procp->p_ucred) {
595 vap->va_uid = procp->p_ucred->cr_uid;
596 vap->va_gid = procp->p_ucred->cr_gid;
597 } else {
598 vap->va_uid = -1;
599 vap->va_gid = -1;
600 }
984263bc
MD
601 }
602
603 switch (pfs->pfs_type) {
604 case Proot:
605 /*
606 * Set nlink to 1 to tell fts(3) we don't actually know.
607 */
608 vap->va_nlink = 1;
609 vap->va_uid = 0;
610 vap->va_gid = 0;
611 vap->va_size = vap->va_bytes = DEV_BSIZE;
612 break;
613
614 case Pcurproc: {
615 char buf[16]; /* should be enough */
f44c73be 616
984263bc
MD
617 vap->va_uid = 0;
618 vap->va_gid = 0;
f44c73be
MD
619 vap->va_size = ksnprintf(buf, sizeof(buf),
620 "%ld", (long)curproc->p_pid);
621 vap->va_bytes = vap->va_size;
984263bc
MD
622 break;
623 }
624
625 case Pproc:
626 vap->va_nlink = nproc_targets;
627 vap->va_size = vap->va_bytes = DEV_BSIZE;
628 break;
629
630 case Pfile: {
631 char *fullpath, *freepath;
f44c73be 632
07cdb1d2
MD
633 if (procp->p_textnch.ncp) {
634 struct nchandle nch;
635
636 cache_copy(&procp->p_textnch, &nch);
637 error = cache_fullpath(procp, &nch,
638 &fullpath, &freepath, 0);
639 cache_drop(&nch);
640 } else {
f44c73be 641 error = EINVAL;
07cdb1d2 642 }
f44c73be 643
984263bc
MD
644 if (error == 0) {
645 vap->va_size = strlen(fullpath);
efda3bd0 646 kfree(freepath, M_TEMP);
984263bc
MD
647 } else {
648 vap->va_size = sizeof("unknown") - 1;
649 error = 0;
650 }
651 vap->va_bytes = vap->va_size;
652 break;
653 }
654
655 case Pmem:
656 /*
657 * If we denied owner access earlier, then we have to
658 * change the owner to root - otherwise 'ps' and friends
659 * will break even though they are setgid kmem. *SIGH*
660 */
4643740a 661 if (procp->p_flags & P_SUGID)
984263bc 662 vap->va_uid = 0;
f44c73be 663 else if (procp->p_ucred)
984263bc 664 vap->va_uid = procp->p_ucred->cr_uid;
f44c73be
MD
665 else
666 vap->va_uid = -1;
984263bc
MD
667 break;
668
669 case Pregs:
670 vap->va_bytes = vap->va_size = sizeof(struct reg);
671 break;
672
673 case Pfpregs:
674 vap->va_bytes = vap->va_size = sizeof(struct fpreg);
675 break;
676
677 case Pdbregs:
678 vap->va_bytes = vap->va_size = sizeof(struct dbreg);
679 break;
680
681 case Ptype:
682 case Pmap:
683 case Pctl:
684 case Pstatus:
685 case Pnote:
686 case Pnotepg:
687 case Pcmdline:
688 case Prlimit:
689 break;
690
691 default:
692 panic("procfs_getattr");
693 }
58c2553a 694done:
f44c73be 695 pfs_pdone(procp);
984263bc
MD
696 return (error);
697}
698
ac424f9b
CP
699/*
700 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap,
b478fdce 701 * struct ucred *a_cred)
ac424f9b 702 */
984263bc 703static int
ac424f9b 704procfs_setattr(struct vop_setattr_args *ap)
984263bc 705{
984263bc
MD
706 if (ap->a_vap->va_flags != VNOVAL)
707 return (EOPNOTSUPP);
708
709 /*
710 * just fake out attribute setting
711 * it's not good to generate an error
712 * return, otherwise things like creat()
713 * will fail when they try to set the
714 * file length to 0. worse, this means
715 * that echo $note > /proc/$pid/note will fail.
716 */
717
718 return (0);
719}
720
721/*
722 * implement access checking.
723 *
b478fdce 724 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred)
984263bc
MD
725 */
726static int
ac424f9b 727procfs_access(struct vop_access_args *ap)
984263bc 728{
984263bc
MD
729 struct vattr vattr;
730 int error;
731
52cdca9a
NT
732 error = VOP_GETATTR(ap->a_vp, &vattr);
733 if (!error)
734 error = vop_helper_access(ap, vattr.va_uid, vattr.va_gid,
735 vattr.va_mode, 0);
736 return (error);
984263bc
MD
737}
738
739/*
3446c007
MD
740 * lookup. this is incredibly complicated in the general case, however
741 * for most pseudo-filesystems very little needs to be done.
ac424f9b
CP
742 *
743 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
744 * struct componentname *a_cnp)
984263bc
MD
745 */
746static int
e62afb5f 747procfs_lookup(struct vop_old_lookup_args *ap)
984263bc
MD
748{
749 struct componentname *cnp = ap->a_cnp;
750 struct vnode **vpp = ap->a_vpp;
751 struct vnode *dvp = ap->a_dvp;
752 char *pname = cnp->cn_nameptr;
753 /* struct proc *curp = cnp->cn_proc; */
754 struct proc_target *pt;
755 pid_t pid;
756 struct pfsnode *pfs;
757 struct proc *p;
c7e98b2f 758 struct lwp *lp;
984263bc 759 int i;
3446c007 760 int error;
984263bc
MD
761
762 *vpp = NULL;
763
2b69e610 764 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME)
984263bc
MD
765 return (EROFS);
766
58c2553a 767 p = NULL;
3446c007 768 error = 0;
984263bc
MD
769 if (cnp->cn_namelen == 1 && *pname == '.') {
770 *vpp = dvp;
3446c007
MD
771 vref(*vpp);
772 goto out;
984263bc
MD
773 }
774
775 pfs = VTOPFS(dvp);
776 switch (pfs->pfs_type) {
777 case Proot:
2b69e610 778 if (cnp->cn_flags & CNP_ISDOTDOT)
984263bc
MD
779 return (EIO);
780
3446c007
MD
781 if (CNEQ(cnp, "curproc", 7)) {
782 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc);
783 goto out;
784 }
984263bc
MD
785
786 pid = atopid(pname, cnp->cn_namelen);
787 if (pid == NO_PID)
788 break;
789
198c0ff2 790 p = pfs_pfind(pid);
984263bc
MD
791 if (p == NULL)
792 break;
793
dcab9ef0
JS
794 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
795 break;
796
28c57d20
HP
797 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
798 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
799 break;
800
3446c007
MD
801 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
802 goto out;
984263bc
MD
803
804 case Pproc:
3446c007
MD
805 if (cnp->cn_flags & CNP_ISDOTDOT) {
806 error = procfs_root(dvp->v_mount, vpp);
807 goto out;
808 }
984263bc 809
198c0ff2 810 p = pfs_pfind(pfs->pfs_pid);
984263bc
MD
811 if (p == NULL)
812 break;
c7e98b2f
SS
813 /* XXX lwp */
814 lp = FIRST_LWP_IN_PROC(p);
984263bc 815
dcab9ef0
JS
816 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
817 break;
818
28c57d20
HP
819 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
820 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
821 break;
822
984263bc
MD
823 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
824 if (cnp->cn_namelen == pt->pt_namlen &&
825 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
c7e98b2f 826 (pt->pt_valid == NULL || (*pt->pt_valid)(lp)))
984263bc
MD
827 goto found;
828 }
829 break;
830 found:
3446c007
MD
831 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
832 pt->pt_pfstype);
833 goto out;
984263bc
MD
834
835 default:
3446c007
MD
836 error = ENOTDIR;
837 goto out;
984263bc 838 }
3446c007
MD
839 if (cnp->cn_nameiop == NAMEI_LOOKUP)
840 error = ENOENT;
841 else
842 error = EROFS;
843 /*
844 * If no error occured *vpp will hold a referenced locked vnode.
aa723195
MD
845 * dvp was passed to us locked and *vpp must be returned locked.
846 * If *vpp != dvp then we should unlock dvp if (1) this is not the
847 * last component or (2) CNP_LOCKPARENT is not set.
3446c007
MD
848 */
849out:
aa723195
MD
850 if (error == 0 && *vpp != dvp) {
851 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) {
852 cnp->cn_flags |= CNP_PDIRUNLOCK;
a11aaa81 853 vn_unlock(dvp);
3446c007
MD
854 }
855 }
f44c73be 856 pfs_pdone(p);
3446c007 857 return (error);
984263bc
MD
858}
859
860/*
861 * Does this process have a text file?
862 */
4d4fb029 863int
c7e98b2f 864procfs_validfile(struct lwp *lp)
984263bc 865{
c7e98b2f 866 return (procfs_findtextvp(lp->lwp_proc) != NULLVP);
984263bc
MD
867}
868
869/*
870 * readdir() returns directory entries from pfsnode (vp).
871 *
872 * We generate just one directory entry at a time, as it would probably
873 * not pay off to buffer several entries locally to save uiomove calls.
ac424f9b
CP
874 *
875 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
84009d92 876 * int *a_eofflag, int *a_ncookies, off_t **a_cookies)
984263bc
MD
877 */
878static int
ac424f9b 879procfs_readdir(struct vop_readdir_args *ap)
984263bc 880{
984263bc 881 struct pfsnode *pfs;
dcab9ef0 882 int error;
984263bc 883
dcab9ef0 884 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
984263bc 885 return (EINVAL);
885ecb13
MD
886 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0)
887 return (error);
dcab9ef0 888 pfs = VTOPFS(ap->a_vp);
984263bc
MD
889
890 switch (pfs->pfs_type) {
dcab9ef0 891 case Pproc:
885ecb13
MD
892 /*
893 * this is for the process-specific sub-directories.
894 * all that is needed to is copy out all the entries
895 * from the procent[] table (top of this file).
896 */
dcab9ef0
JS
897 error = procfs_readdir_proc(ap);
898 break;
dcab9ef0 899 case Proot:
885ecb13
MD
900 /*
901 * this is for the root of the procfs filesystem
902 * what is needed is a special entry for "curproc"
903 * followed by an entry for each process on allproc
904 */
dcab9ef0 905 error = procfs_readdir_root(ap);
984263bc 906 break;
984263bc
MD
907 default:
908 error = ENOTDIR;
909 break;
910 }
911
885ecb13 912 vn_unlock(ap->a_vp);
984263bc
MD
913 return (error);
914}
915
dcab9ef0
JS
916static int
917procfs_readdir_proc(struct vop_readdir_args *ap)
918{
919 struct pfsnode *pfs;
920 int error, i, retval;
921 struct proc *p;
c7e98b2f 922 struct lwp *lp;
dcab9ef0
JS
923 struct proc_target *pt;
924 struct uio *uio = ap->a_uio;
925
926 pfs = VTOPFS(ap->a_vp);
198c0ff2 927 p = pfs_pfind(pfs->pfs_pid);
dcab9ef0
JS
928 if (p == NULL)
929 return(0);
58c2553a
MD
930 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) {
931 error = 0;
932 goto done;
933 }
934 /* XXX lwp, not MPSAFE */
c7e98b2f 935 lp = FIRST_LWP_IN_PROC(p);
dcab9ef0
JS
936
937 error = 0;
fb0466c9 938 i = (int)uio->uio_offset;
58c2553a
MD
939 if (i < 0) {
940 error = EINVAL;
941 goto done;
942 }
dcab9ef0
JS
943
944 for (pt = &proc_targets[i];
945 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
c7e98b2f 946 if (pt->pt_valid && (*pt->pt_valid)(lp) == 0)
dcab9ef0
JS
947 continue;
948
949 retval = vop_write_dirent(&error, uio,
950 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
951 pt->pt_namlen, pt->pt_name);
952 if (retval)
953 break;
954 }
955
fb0466c9 956 uio->uio_offset = (off_t)i;
58c2553a
MD
957 error = 0;
958done:
f44c73be 959 pfs_pdone(p);
58c2553a 960 return error;
dcab9ef0
JS
961}
962
51e64ff2
MD
963struct procfs_readdir_root_info {
964 int error;
965 int i;
966 int pcnt;
967 struct uio *uio;
968 struct ucred *cred;
969};
970
971static int procfs_readdir_root_callback(struct proc *p, void *data);
972
dcab9ef0
JS
973static int
974procfs_readdir_root(struct vop_readdir_args *ap)
975{
51e64ff2 976 struct procfs_readdir_root_info info;
dcab9ef0 977 struct uio *uio = ap->a_uio;
51e64ff2
MD
978 int res;
979
980 info.error = 0;
fb0466c9
MD
981 info.i = (int)uio->uio_offset;
982
983 if (info.i < 0)
984 return (EINVAL);
985
51e64ff2
MD
986 info.pcnt = 0;
987 info.uio = uio;
988 info.cred = ap->a_cred;
51e64ff2
MD
989 while (info.pcnt < 3) {
990 res = procfs_readdir_root_callback(NULL, &info);
991 if (res < 0)
992 break;
993 }
994 if (res >= 0)
995 allproc_scan(procfs_readdir_root_callback, &info);
fb0466c9 996 uio->uio_offset = (off_t)info.i;
51e64ff2
MD
997
998 return (info.error);
999}
1000
1001static int
1002procfs_readdir_root_callback(struct proc *p, void *data)
1003{
1004 struct procfs_readdir_root_info *info = data;
1005 struct uio *uio;
1006 int retval;
dcab9ef0
JS
1007 ino_t d_ino;
1008 const char *d_name;
1009 char d_name_pid[20];
1010 size_t d_namlen;
1011 uint8_t d_type;
1012
51e64ff2 1013 uio = info->uio;
dcab9ef0 1014
51e64ff2
MD
1015 if (uio->uio_resid <= 0 || info->error)
1016 return(-1);
dcab9ef0 1017
51e64ff2
MD
1018 switch (info->pcnt) {
1019 case 0: /* `.' */
1020 d_ino = PROCFS_FILENO(0, Proot);
1021 d_name = ".";
1022 d_namlen = 1;
1023 d_type = DT_DIR;
1024 break;
1025 case 1: /* `..' */
1026 d_ino = PROCFS_FILENO(0, Proot);
1027 d_name = "..";
1028 d_namlen = 2;
1029 d_type = DT_DIR;
1030 break;
dcab9ef0 1031
51e64ff2
MD
1032 case 2:
1033 d_ino = PROCFS_FILENO(0, Pcurproc);
1034 d_namlen = 7;
1035 d_name = "curproc";
1036 d_type = DT_LNK;
1037 break;
dcab9ef0 1038
51e64ff2
MD
1039
1040 default:
1041 if (!PRISON_CHECK(info->cred, p->p_ucred))
1042 return(0);
1043 if (ps_showallprocs == 0 &&
1044 info->cred->cr_uid != 0 &&
1045 info->cred->cr_uid != p->p_ucred->cr_uid) {
1046 return(0);
dcab9ef0
JS
1047 }
1048
51e64ff2
MD
1049 /*
1050 * Skip entries we have already returned (optimization)
1051 */
1052 if (info->pcnt < info->i) {
1053 ++info->pcnt;
1054 return(0);
1055 }
1056
1057 d_ino = PROCFS_FILENO(p->p_pid, Pproc);
f8c7a42d 1058 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid),
51e64ff2
MD
1059 "%ld", (long)p->p_pid);
1060 d_name = d_name_pid;
1061 d_type = DT_DIR;
1062 break;
dcab9ef0 1063 }
51e64ff2
MD
1064
1065 /*
1066 * Skip entries we have already returned (optimization)
1067 */
1068 if (info->pcnt < info->i) {
1069 ++info->pcnt;
1070 return(0);
1071 }
1072
1073 retval = vop_write_dirent(&info->error, uio,
1074 d_ino, d_type, d_namlen, d_name);
1075 if (retval)
1076 return(-1);
1077 ++info->pcnt;
1078 ++info->i;
dcab9ef0
JS
1079 return(0);
1080}
1081
984263bc
MD
1082/*
1083 * readlink reads the link of `curproc' or `file'
1084 */
1085static int
ac424f9b 1086procfs_readlink(struct vop_readlink_args *ap)
984263bc
MD
1087{
1088 char buf[16]; /* should be enough */
1089 struct proc *procp;
1090 struct vnode *vp = ap->a_vp;
1091 struct pfsnode *pfs = VTOPFS(vp);
1092 char *fullpath, *freepath;
1093 int error, len;
1094
1095 switch (pfs->pfs_type) {
1096 case Pcurproc:
1097 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
1098 return (EINVAL);
1099
f8c7a42d 1100 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
984263bc
MD
1101
1102 return (uiomove(buf, len, ap->a_uio));
1103 /*
1104 * There _should_ be no way for an entire process to disappear
1105 * from under us...
1106 */
1107 case Pfile:
198c0ff2 1108 procp = pfs_pfind(pfs->pfs_pid);
41c20dac 1109 if (procp == NULL || procp->p_ucred == NULL) {
086c1d7e 1110 kprintf("procfs_readlink: pid %d disappeared\n",
984263bc 1111 pfs->pfs_pid);
f44c73be 1112 pfs_pdone(procp);
984263bc 1113 return (uiomove("unknown", sizeof("unknown") - 1,
f44c73be 1114 ap->a_uio));
984263bc 1115 }
07cdb1d2
MD
1116 if (procp->p_textnch.ncp) {
1117 struct nchandle nch;
1118
1119 cache_copy(&procp->p_textnch, &nch);
1120 error = cache_fullpath(procp, &nch,
f44c73be 1121 &fullpath, &freepath, 0);
07cdb1d2
MD
1122 cache_drop(&nch);
1123 } else {
f44c73be 1124 error = EINVAL;
07cdb1d2 1125 }
f44c73be 1126
58c2553a 1127 if (error != 0) {
f44c73be 1128 pfs_pdone(procp);
984263bc 1129 return (uiomove("unknown", sizeof("unknown") - 1,
f44c73be 1130 ap->a_uio));
58c2553a 1131 }
984263bc 1132 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
efda3bd0 1133 kfree(freepath, M_TEMP);
f44c73be 1134 pfs_pdone(procp);
984263bc
MD
1135 return (error);
1136 default:
1137 return (EINVAL);
1138 }
1139}
1140
1141/*
1142 * convert decimal ascii to pid_t
1143 */
1144static pid_t
ac424f9b 1145atopid(const char *b, u_int len)
984263bc
MD
1146{
1147 pid_t p = 0;
1148
1149 while (len--) {
1150 char c = *b++;
1151 if (c < '0' || c > '9')
1152 return (NO_PID);
1153 p = 10 * p + (c - '0');
1154 if (p > PID_MAX)
1155 return (NO_PID);
1156 }
1157
1158 return (p);
1159}
1160