Add missing commit for the VM load heuristic and page allocation rate
[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 $
72a65e78 40 * $DragonFly: src/sys/vfs/procfs/procfs_vnops.c,v 1.27 2005/11/20 02:23:59 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>
61#include <machine/reg.h>
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
RG
68static int procfs_access (struct vop_access_args *);
69static int procfs_badop (void);
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
984263bc
MD
86/*
87 * This is a list of the valid names in the
88 * process-specific sub-directories. It is
89 * used in procfs_lookup and procfs_readdir
90 */
91static struct proc_target {
92 u_char pt_type;
93 u_char pt_namlen;
94 char *pt_name;
95 pfstype pt_pfstype;
a6ee311a 96 int (*pt_valid) (struct proc *p);
984263bc
MD
97} proc_targets[] = {
98#define N(s) sizeof(s)-1, s
99 /* name type validp */
100 { DT_DIR, N("."), Pproc, NULL },
101 { DT_DIR, N(".."), Proot, NULL },
102 { DT_REG, N("mem"), Pmem, NULL },
103 { DT_REG, N("regs"), Pregs, procfs_validregs },
104 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs },
105 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs },
106 { DT_REG, N("ctl"), Pctl, NULL },
107 { DT_REG, N("status"), Pstatus, NULL },
108 { DT_REG, N("note"), Pnote, NULL },
109 { DT_REG, N("notepg"), Pnotepg, NULL },
110 { DT_REG, N("map"), Pmap, procfs_validmap },
111 { DT_REG, N("etype"), Ptype, procfs_validtype },
112 { DT_REG, N("cmdline"), Pcmdline, NULL },
113 { DT_REG, N("rlimit"), Prlimit, NULL },
114 { DT_LNK, N("file"), Pfile, NULL },
115#undef N
116};
117static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]);
118
a6ee311a 119static pid_t atopid (const char *, u_int);
984263bc
MD
120
121/*
122 * set things up for doing i/o on
123 * the pfsnode (vp). (vp) is locked
124 * on entry, and should be left locked
125 * on exit.
126 *
127 * for procfs we don't need to do anything
128 * in particular for i/o. all that is done
129 * is to support exclusive open on process
130 * memory images.
ac424f9b
CP
131 *
132 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
133 * struct thread *a_td)
984263bc
MD
134 */
135static int
ac424f9b 136procfs_open(struct vop_open_args *ap)
984263bc
MD
137{
138 struct pfsnode *pfs = VTOPFS(ap->a_vp);
139 struct proc *p1, *p2;
140
141 p2 = PFIND(pfs->pfs_pid);
142 if (p2 == NULL)
143 return (ENOENT);
41c20dac 144 if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred))
984263bc
MD
145 return (ENOENT);
146
147 switch (pfs->pfs_type) {
148 case Pmem:
149 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
150 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE)))
151 return (EBUSY);
152
dadab5e9
MD
153 p1 = ap->a_td->td_proc;
154 KKASSERT(p1);
984263bc
MD
155 /* Can't trace a process that's currently exec'ing. */
156 if ((p2->p_flag & P_INEXEC) != 0)
157 return EAGAIN;
41c20dac 158 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred))
984263bc
MD
159 return (EPERM);
160
161 if (ap->a_mode & FWRITE)
162 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
163
164 return (0);
165
166 default:
167 break;
168 }
169
170 return (0);
171}
172
173/*
174 * close the pfsnode (vp) after doing i/o.
175 * (vp) is not locked on entry or exit.
176 *
177 * nothing to do for procfs other than undo
178 * any exclusive open flag (see _open above).
ac424f9b
CP
179 *
180 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
181 * struct thread *a_td)
984263bc
MD
182 */
183static int
ac424f9b 184procfs_close(struct vop_close_args *ap)
984263bc
MD
185{
186 struct pfsnode *pfs = VTOPFS(ap->a_vp);
187 struct proc *p;
188
189 switch (pfs->pfs_type) {
190 case Pmem:
191 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
192 pfs->pfs_flags &= ~(FWRITE|O_EXCL);
193 /*
194 * This rather complicated-looking code is trying to
195 * determine if this was the last close on this particular
196 * vnode. While one would expect v_usecount to be 1 at
197 * that point, it seems that (according to John Dyson)
198 * the VM system will bump up the usecount. So: if the
199 * usecount is 2, and VOBJBUF is set, then this is really
200 * the last close. Otherwise, if the usecount is < 2
201 * then it is definitely the last close.
202 * If this is the last close, then it checks to see if
203 * the target process has PF_LINGER set in p_pfsflags,
204 * if this is *not* the case, then the process' stop flags
205 * are cleared, and the process is woken up. This is
206 * to help prevent the case where a process has been
207 * told to stop on an event, but then the requesting process
208 * has gone away or forgotten about it.
209 */
210 if ((ap->a_vp->v_usecount < 2)
211 && (p = pfind(pfs->pfs_pid))
212 && !(p->p_pfsflags & PF_LINGER)) {
213 p->p_stops = 0;
214 p->p_step = 0;
215 wakeup(&p->p_step);
216 }
217 break;
218 default:
219 break;
220 }
221
222 return (0);
223}
224
225/*
226 * do an ioctl operation on a pfsnode (vp).
227 * (vp) is not locked on entry or exit.
228 */
229static int
ac424f9b 230procfs_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
231{
232 struct pfsnode *pfs = VTOPFS(ap->a_vp);
dadab5e9
MD
233 struct proc *procp;
234 struct proc *p;
984263bc
MD
235 int error;
236 int signo;
237 struct procfs_status *psp;
238 unsigned char flags;
239
984263bc 240 procp = pfind(pfs->pfs_pid);
dadab5e9 241 if (procp == NULL)
984263bc 242 return ENOTTY;
dadab5e9
MD
243 p = ap->a_td->td_proc;
244 if (p == NULL)
245 return EINVAL;
984263bc
MD
246
247 /* Can't trace a process that's currently exec'ing. */
248 if ((procp->p_flag & P_INEXEC) != 0)
249 return EAGAIN;
41c20dac 250 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred))
984263bc
MD
251 return EPERM;
252
253 switch (ap->a_command) {
254 case PIOCBIS:
255 procp->p_stops |= *(unsigned int*)ap->a_data;
256 break;
257 case PIOCBIC:
258 procp->p_stops &= ~*(unsigned int*)ap->a_data;
259 break;
260 case PIOCSFL:
261 /*
262 * NFLAGS is "non-suser_xxx flags" -- currently, only
263 * PFS_ISUGID ("ignore set u/g id");
264 */
265#define NFLAGS (PF_ISUGID)
266 flags = (unsigned char)*(unsigned int*)ap->a_data;
dadab5e9 267 if (flags & NFLAGS && (error = suser_cred(ap->a_cred, 0)))
984263bc
MD
268 return error;
269 procp->p_pfsflags = flags;
270 break;
271 case PIOCGFL:
272 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
273 break;
274 case PIOCSTATUS:
275 psp = (struct procfs_status *)ap->a_data;
276 psp->state = (procp->p_step == 0);
277 psp->flags = procp->p_pfsflags;
278 psp->events = procp->p_stops;
279 if (procp->p_step) {
280 psp->why = procp->p_stype;
281 psp->val = procp->p_xstat;
282 } else {
283 psp->why = psp->val = 0; /* Not defined values */
284 }
285 break;
286 case PIOCWAIT:
287 psp = (struct procfs_status *)ap->a_data;
288 if (procp->p_step == 0) {
377d4740 289 error = tsleep(&procp->p_stype, PCATCH, "piocwait", 0);
984263bc
MD
290 if (error)
291 return error;
292 }
293 psp->state = 1; /* It stopped */
294 psp->flags = procp->p_pfsflags;
295 psp->events = procp->p_stops;
296 psp->why = procp->p_stype; /* why it stopped */
297 psp->val = procp->p_xstat; /* any extra info */
298 break;
299 case PIOCCONT: /* Restart a proc */
300 if (procp->p_step == 0)
301 return EINVAL; /* Can only start a stopped process */
302 if ((signo = *(int*)ap->a_data) != 0) {
303 if (signo >= NSIG || signo <= 0)
304 return EINVAL;
305 psignal(procp, signo);
306 }
307 procp->p_step = 0;
308 wakeup(&procp->p_step);
309 break;
310 default:
311 return (ENOTTY);
312 }
313 return 0;
314}
315
316/*
317 * do block mapping for pfsnode (vp).
318 * since we don't use the buffer cache
319 * for procfs this function should never
320 * be called. in any case, it's not clear
321 * what part of the kernel ever makes use
322 * of this function. for sanity, this is the
323 * usual no-op bmap, although returning
324 * (EIO) would be a reasonable alternative.
ac424f9b
CP
325 *
326 * procfs_bmap(struct vnode *a_vp, daddr_t a_bn, struct vnode **a_vpp,
327 * daddr_t *a_bnp, int *a_runp)
984263bc
MD
328 */
329static int
ac424f9b 330procfs_bmap(struct vop_bmap_args *ap)
984263bc 331{
984263bc
MD
332 if (ap->a_vpp != NULL)
333 *ap->a_vpp = ap->a_vp;
334 if (ap->a_bnp != NULL)
335 *ap->a_bnp = ap->a_bn;
336 if (ap->a_runp != NULL)
337 *ap->a_runp = 0;
338 return (0);
339}
340
341/*
342 * procfs_inactive is called when the pfsnode
343 * is vrele'd and the reference count goes
344 * to zero. (vp) will be on the vnode free
345 * list, so to get it back vget() must be
346 * used.
347 *
348 * (vp) is locked on entry, but must be unlocked on exit.
ac424f9b
CP
349 *
350 * procfs_inactive(struct vnode *a_vp, struct thread *a_td)
984263bc
MD
351 */
352static int
ac424f9b 353procfs_inactive(struct vop_inactive_args *ap)
984263bc 354{
5fd012e0 355 /*struct vnode *vp = ap->a_vp;*/
984263bc
MD
356
357 return (0);
358}
359
360/*
361 * _reclaim is called when getnewvnode()
362 * wants to make use of an entry on the vnode
363 * free list. at this time the filesystem needs
364 * to free any private data and remove the node
365 * from any private lists.
ac424f9b
CP
366 *
367 * procfs_reclaim(struct vnode *a_vp)
984263bc
MD
368 */
369static int
ac424f9b 370procfs_reclaim(struct vop_reclaim_args *ap)
984263bc 371{
984263bc
MD
372 return (procfs_freevp(ap->a_vp));
373}
374
375/*
376 * _print is used for debugging.
377 * just print a readable description
378 * of (vp).
ac424f9b
CP
379 *
380 * procfs_print(struct vnode *a_vp)
984263bc
MD
381 */
382static int
ac424f9b 383procfs_print(struct vop_print_args *ap)
984263bc
MD
384{
385 struct pfsnode *pfs = VTOPFS(ap->a_vp);
386
387 printf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
388 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
389 return (0);
390}
391
392/*
393 * generic entry point for unsupported operations
394 */
395static int
ac424f9b 396procfs_badop(void)
984263bc 397{
984263bc
MD
398 return (EIO);
399}
400
401/*
402 * Invent attributes for pfsnode (vp) and store
403 * them in (vap).
404 * Directories lengths are returned as zero since
405 * any real length would require the genuine size
406 * to be computed, and nothing cares anyway.
407 *
408 * this is relatively minimal for procfs.
ac424f9b
CP
409 *
410 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap,
411 * struct ucred *a_cred, struct thread *a_td)
984263bc
MD
412 */
413static int
ac424f9b 414procfs_getattr(struct vop_getattr_args *ap)
984263bc
MD
415{
416 struct pfsnode *pfs = VTOPFS(ap->a_vp);
417 struct vattr *vap = ap->a_vap;
418 struct proc *procp;
419 int error;
420
421 /*
422 * First make sure that the process and its credentials
423 * still exist.
424 */
425 switch (pfs->pfs_type) {
426 case Proot:
427 case Pcurproc:
428 procp = 0;
429 break;
430
431 default:
432 procp = PFIND(pfs->pfs_pid);
41c20dac 433 if (procp == NULL || procp->p_ucred == NULL)
984263bc
MD
434 return (ENOENT);
435 }
436
437 error = 0;
438
439 /* start by zeroing out the attributes */
440 VATTR_NULL(vap);
441
442 /* next do all the common fields */
443 vap->va_type = ap->a_vp->v_type;
444 vap->va_mode = pfs->pfs_mode;
445 vap->va_fileid = pfs->pfs_fileno;
446 vap->va_flags = 0;
447 vap->va_blocksize = PAGE_SIZE;
448 vap->va_bytes = vap->va_size = 0;
449 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
450
451 /*
452 * Make all times be current TOD.
453 * It would be possible to get the process start
454 * time from the p_stat structure, but there's
455 * no "file creation" time stamp anyway, and the
456 * p_stat structure is not addressible if u. gets
457 * swapped out for that process.
458 */
459 nanotime(&vap->va_ctime);
460 vap->va_atime = vap->va_mtime = vap->va_ctime;
461
462 /*
463 * If the process has exercised some setuid or setgid
464 * privilege, then rip away read/write permission so
465 * that only root can gain access.
466 */
467 switch (pfs->pfs_type) {
468 case Pctl:
469 case Pregs:
470 case Pfpregs:
471 case Pdbregs:
472 case Pmem:
473 if (procp->p_flag & P_SUGID)
474 vap->va_mode &= ~((VREAD|VWRITE)|
475 ((VREAD|VWRITE)>>3)|
476 ((VREAD|VWRITE)>>6));
477 break;
478 default:
479 break;
480 }
481
482 /*
483 * now do the object specific fields
484 *
485 * The size could be set from struct reg, but it's hardly
486 * worth the trouble, and it puts some (potentially) machine
487 * dependent data into this machine-independent code. If it
488 * becomes important then this function should break out into
489 * a per-file stat function in the corresponding .c file.
490 */
491
492 vap->va_nlink = 1;
493 if (procp) {
494 vap->va_uid = procp->p_ucred->cr_uid;
495 vap->va_gid = procp->p_ucred->cr_gid;
496 }
497
498 switch (pfs->pfs_type) {
499 case Proot:
500 /*
501 * Set nlink to 1 to tell fts(3) we don't actually know.
502 */
503 vap->va_nlink = 1;
504 vap->va_uid = 0;
505 vap->va_gid = 0;
506 vap->va_size = vap->va_bytes = DEV_BSIZE;
507 break;
508
509 case Pcurproc: {
510 char buf[16]; /* should be enough */
511 vap->va_uid = 0;
512 vap->va_gid = 0;
513 vap->va_size = vap->va_bytes =
514 snprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
515 break;
516 }
517
518 case Pproc:
519 vap->va_nlink = nproc_targets;
520 vap->va_size = vap->va_bytes = DEV_BSIZE;
521 break;
522
523 case Pfile: {
524 char *fullpath, *freepath;
541044dc 525 error = vn_fullpath(procp, NULL, &fullpath, &freepath);
984263bc
MD
526 if (error == 0) {
527 vap->va_size = strlen(fullpath);
528 free(freepath, M_TEMP);
529 } else {
530 vap->va_size = sizeof("unknown") - 1;
531 error = 0;
532 }
533 vap->va_bytes = vap->va_size;
534 break;
535 }
536
537 case Pmem:
538 /*
539 * If we denied owner access earlier, then we have to
540 * change the owner to root - otherwise 'ps' and friends
541 * will break even though they are setgid kmem. *SIGH*
542 */
543 if (procp->p_flag & P_SUGID)
544 vap->va_uid = 0;
545 else
546 vap->va_uid = procp->p_ucred->cr_uid;
547 break;
548
549 case Pregs:
550 vap->va_bytes = vap->va_size = sizeof(struct reg);
551 break;
552
553 case Pfpregs:
554 vap->va_bytes = vap->va_size = sizeof(struct fpreg);
555 break;
556
557 case Pdbregs:
558 vap->va_bytes = vap->va_size = sizeof(struct dbreg);
559 break;
560
561 case Ptype:
562 case Pmap:
563 case Pctl:
564 case Pstatus:
565 case Pnote:
566 case Pnotepg:
567 case Pcmdline:
568 case Prlimit:
569 break;
570
571 default:
572 panic("procfs_getattr");
573 }
574
575 return (error);
576}
577
ac424f9b
CP
578/*
579 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap,
580 * struct ucred *a_cred, struct thread *a_td)
581 */
984263bc 582static int
ac424f9b 583procfs_setattr(struct vop_setattr_args *ap)
984263bc 584{
984263bc
MD
585 if (ap->a_vap->va_flags != VNOVAL)
586 return (EOPNOTSUPP);
587
588 /*
589 * just fake out attribute setting
590 * it's not good to generate an error
591 * return, otherwise things like creat()
592 * will fail when they try to set the
593 * file length to 0. worse, this means
594 * that echo $note > /proc/$pid/note will fail.
595 */
596
597 return (0);
598}
599
600/*
601 * implement access checking.
602 *
603 * something very similar to this code is duplicated
604 * throughout the 4bsd kernel and should be moved
605 * into kern/vfs_subr.c sometime.
606 *
607 * actually, the check for super-user is slightly
608 * broken since it will allow read access to write-only
609 * objects. this doesn't cause any particular trouble
610 * but does mean that the i/o entry points need to check
611 * that the operation really does make sense.
ac424f9b
CP
612 *
613 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
614 * struct thread *a_td)
984263bc
MD
615 */
616static int
ac424f9b 617procfs_access(struct vop_access_args *ap)
984263bc
MD
618{
619 struct vattr *vap;
620 struct vattr vattr;
621 int error;
622
623 /*
624 * If you're the super-user,
625 * you always get access.
626 */
627 if (ap->a_cred->cr_uid == 0)
628 return (0);
629
630 vap = &vattr;
3b568787 631 error = VOP_GETATTR(ap->a_vp, vap, ap->a_td);
984263bc
MD
632 if (error)
633 return (error);
634
635 /*
636 * Access check is based on only one of owner, group, public.
637 * If not owner, then check group. If not a member of the
638 * group, then check public access.
639 */
640 if (ap->a_cred->cr_uid != vap->va_uid) {
641 gid_t *gp;
642 int i;
643
644 ap->a_mode >>= 3;
645 gp = ap->a_cred->cr_groups;
646 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
647 if (vap->va_gid == *gp)
648 goto found;
649 ap->a_mode >>= 3;
650found:
651 ;
652 }
653
654 if ((vap->va_mode & ap->a_mode) == ap->a_mode)
655 return (0);
656
657 return (EACCES);
658}
659
660/*
3446c007
MD
661 * lookup. this is incredibly complicated in the general case, however
662 * for most pseudo-filesystems very little needs to be done.
ac424f9b
CP
663 *
664 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
665 * struct componentname *a_cnp)
984263bc
MD
666 */
667static int
e62afb5f 668procfs_lookup(struct vop_old_lookup_args *ap)
984263bc
MD
669{
670 struct componentname *cnp = ap->a_cnp;
671 struct vnode **vpp = ap->a_vpp;
672 struct vnode *dvp = ap->a_dvp;
673 char *pname = cnp->cn_nameptr;
674 /* struct proc *curp = cnp->cn_proc; */
675 struct proc_target *pt;
676 pid_t pid;
677 struct pfsnode *pfs;
678 struct proc *p;
679 int i;
3446c007 680 int error;
984263bc
MD
681
682 *vpp = NULL;
683
2b69e610 684 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME)
984263bc
MD
685 return (EROFS);
686
3446c007 687 error = 0;
984263bc
MD
688 if (cnp->cn_namelen == 1 && *pname == '.') {
689 *vpp = dvp;
3446c007
MD
690 vref(*vpp);
691 goto out;
984263bc
MD
692 }
693
694 pfs = VTOPFS(dvp);
695 switch (pfs->pfs_type) {
696 case Proot:
2b69e610 697 if (cnp->cn_flags & CNP_ISDOTDOT)
984263bc
MD
698 return (EIO);
699
3446c007
MD
700 if (CNEQ(cnp, "curproc", 7)) {
701 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc);
702 goto out;
703 }
984263bc
MD
704
705 pid = atopid(pname, cnp->cn_namelen);
706 if (pid == NO_PID)
707 break;
708
709 p = PFIND(pid);
710 if (p == NULL)
711 break;
712
dcab9ef0
JS
713 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
714 break;
715
28c57d20
HP
716 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
717 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
718 break;
719
3446c007
MD
720 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
721 goto out;
984263bc
MD
722
723 case Pproc:
3446c007
MD
724 if (cnp->cn_flags & CNP_ISDOTDOT) {
725 error = procfs_root(dvp->v_mount, vpp);
726 goto out;
727 }
984263bc
MD
728
729 p = PFIND(pfs->pfs_pid);
730 if (p == NULL)
731 break;
732
dcab9ef0
JS
733 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
734 break;
735
28c57d20
HP
736 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
737 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
738 break;
739
984263bc
MD
740 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
741 if (cnp->cn_namelen == pt->pt_namlen &&
742 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
743 (pt->pt_valid == NULL || (*pt->pt_valid)(p)))
744 goto found;
745 }
746 break;
747 found:
3446c007
MD
748 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
749 pt->pt_pfstype);
750 goto out;
984263bc
MD
751
752 default:
3446c007
MD
753 error = ENOTDIR;
754 goto out;
984263bc 755 }
3446c007
MD
756 if (cnp->cn_nameiop == NAMEI_LOOKUP)
757 error = ENOENT;
758 else
759 error = EROFS;
760 /*
761 * If no error occured *vpp will hold a referenced locked vnode.
aa723195
MD
762 * dvp was passed to us locked and *vpp must be returned locked.
763 * If *vpp != dvp then we should unlock dvp if (1) this is not the
764 * last component or (2) CNP_LOCKPARENT is not set.
3446c007
MD
765 */
766out:
aa723195
MD
767 if (error == 0 && *vpp != dvp) {
768 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) {
769 cnp->cn_flags |= CNP_PDIRUNLOCK;
770 VOP_UNLOCK(dvp, 0, cnp->cn_td);
3446c007
MD
771 }
772 }
773 return (error);
984263bc
MD
774}
775
776/*
777 * Does this process have a text file?
778 */
4d4fb029 779int
ac424f9b 780procfs_validfile(struct proc *p)
984263bc 781{
984263bc
MD
782 return (procfs_findtextvp(p) != NULLVP);
783}
784
785/*
786 * readdir() returns directory entries from pfsnode (vp).
787 *
788 * We generate just one directory entry at a time, as it would probably
789 * not pay off to buffer several entries locally to save uiomove calls.
ac424f9b
CP
790 *
791 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
792 * int *a_eofflag, int *a_ncookies, u_long **a_cookies)
984263bc
MD
793 */
794static int
ac424f9b 795procfs_readdir(struct vop_readdir_args *ap)
984263bc 796{
984263bc 797 struct pfsnode *pfs;
dcab9ef0 798 int error;
984263bc 799
dcab9ef0 800 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
984263bc
MD
801 return (EINVAL);
802
dcab9ef0 803 pfs = VTOPFS(ap->a_vp);
984263bc
MD
804
805 switch (pfs->pfs_type) {
806 /*
807 * this is for the process-specific sub-directories.
808 * all that is needed to is copy out all the entries
809 * from the procent[] table (top of this file).
810 */
dcab9ef0
JS
811 case Pproc:
812 error = procfs_readdir_proc(ap);
813 break;
984263bc
MD
814
815 /*
816 * this is for the root of the procfs filesystem
817 * what is needed is a special entry for "curproc"
818 * followed by an entry for each process on allproc
984263bc
MD
819 */
820
dcab9ef0
JS
821 case Proot:
822 error = procfs_readdir_root(ap);
984263bc
MD
823 break;
824
984263bc
MD
825 default:
826 error = ENOTDIR;
827 break;
828 }
829
984263bc
MD
830 return (error);
831}
832
dcab9ef0
JS
833static int
834procfs_readdir_proc(struct vop_readdir_args *ap)
835{
836 struct pfsnode *pfs;
837 int error, i, retval;
838 struct proc *p;
839 struct proc_target *pt;
840 struct uio *uio = ap->a_uio;
841
842 pfs = VTOPFS(ap->a_vp);
843 p = PFIND(pfs->pfs_pid);
844 if (p == NULL)
845 return(0);
846 if (!PRISON_CHECK(ap->a_cred, p->p_ucred))
847 return(0);
848
849 error = 0;
850 i = uio->uio_offset;
851
852 for (pt = &proc_targets[i];
853 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
854 if (pt->pt_valid && (*pt->pt_valid)(p) == 0)
855 continue;
856
857 retval = vop_write_dirent(&error, uio,
858 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
859 pt->pt_namlen, pt->pt_name);
860 if (retval)
861 break;
862 }
863
864 uio->uio_offset = i;
865
866 return(0);
867}
868
869static int
870procfs_readdir_root(struct vop_readdir_args *ap)
871{
872 int error, i, pcnt, retval;
873 struct uio *uio = ap->a_uio;
874 volatile struct proc *p = LIST_FIRST(&allproc);
875 ino_t d_ino;
876 const char *d_name;
877 char d_name_pid[20];
878 size_t d_namlen;
879 uint8_t d_type;
880
881 error = 0;
882 pcnt = 0;
883 i = uio->uio_offset;
884
885 for (; p && uio->uio_resid > 0 && !error; i++, pcnt++) {
886 switch (i) {
887 case 0: /* `.' */
888 d_ino = PROCFS_FILENO(0, Proot);
889 d_name = ".";
890 d_namlen = 1;
891 d_type = DT_DIR;
892 break;
893 case 1: /* `..' */
894 d_ino = PROCFS_FILENO(0, Proot);
895 d_name = "..";
896 d_namlen = 2;
897 d_type = DT_DIR;
898 break;
899
900 case 2:
901 d_ino = PROCFS_FILENO(0, Pcurproc);
902 d_namlen = 7;
903 d_name = "curproc";
904 d_type = DT_LNK;
905 break;
906
907
908 default:
909 while (pcnt < i) {
910 p = LIST_NEXT(p, p_list);
911 if (!p)
912 goto done;
913 if (!PRISON_CHECK(ap->a_cred, p->p_ucred))
914 continue;
915 pcnt++;
916 }
917 while (!PRISON_CHECK(ap->a_cred, p->p_ucred)) {
918 p = LIST_NEXT(p, p_list);
919 if (!p)
920 goto done;
921 }
922 if (ps_showallprocs == 0 &&
923 ap->a_cred->cr_uid != 0 &&
924 ap->a_cred->cr_uid != p->p_ucred->cr_uid) {
925 p = LIST_NEXT(p, p_list);
926 if (!p)
927 goto done;
928 continue;
929 }
930
931 d_ino = PROCFS_FILENO(p->p_pid, Pproc);
932 d_namlen = snprintf(d_name_pid, sizeof(d_name_pid),
933 "%ld", (long)p->p_pid);
934 d_name = d_name_pid;
935 d_type = DT_DIR;
936 p = LIST_NEXT(p, p_list);
937 break;
938 }
939
72a65e78
MD
940 if (p)
941 PHOLD(p);
dcab9ef0
JS
942 retval = vop_write_dirent(&error, uio,
943 d_ino, d_type, d_namlen, d_name);
72a65e78
MD
944 if (p)
945 PRELE(p);
dcab9ef0
JS
946 if (retval)
947 break;
948 }
949done:
950 uio->uio_offset = i;
951 return(0);
952}
953
984263bc
MD
954/*
955 * readlink reads the link of `curproc' or `file'
956 */
957static int
ac424f9b 958procfs_readlink(struct vop_readlink_args *ap)
984263bc
MD
959{
960 char buf[16]; /* should be enough */
961 struct proc *procp;
962 struct vnode *vp = ap->a_vp;
963 struct pfsnode *pfs = VTOPFS(vp);
964 char *fullpath, *freepath;
965 int error, len;
966
967 switch (pfs->pfs_type) {
968 case Pcurproc:
969 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
970 return (EINVAL);
971
972 len = snprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
973
974 return (uiomove(buf, len, ap->a_uio));
975 /*
976 * There _should_ be no way for an entire process to disappear
977 * from under us...
978 */
979 case Pfile:
980 procp = PFIND(pfs->pfs_pid);
41c20dac 981 if (procp == NULL || procp->p_ucred == NULL) {
984263bc
MD
982 printf("procfs_readlink: pid %d disappeared\n",
983 pfs->pfs_pid);
984 return (uiomove("unknown", sizeof("unknown") - 1,
985 ap->a_uio));
986 }
541044dc 987 error = vn_fullpath(procp, NULL, &fullpath, &freepath);
984263bc
MD
988 if (error != 0)
989 return (uiomove("unknown", sizeof("unknown") - 1,
990 ap->a_uio));
991 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
992 free(freepath, M_TEMP);
993 return (error);
994 default:
995 return (EINVAL);
996 }
997}
998
999/*
1000 * convert decimal ascii to pid_t
1001 */
1002static pid_t
ac424f9b 1003atopid(const char *b, u_int len)
984263bc
MD
1004{
1005 pid_t p = 0;
1006
1007 while (len--) {
1008 char c = *b++;
1009 if (c < '0' || c > '9')
1010 return (NO_PID);
1011 p = 10 * p + (c - '0');
1012 if (p > PID_MAX)
1013 return (NO_PID);
1014 }
1015
1016 return (p);
1017}
1018
1019/*
1020 * procfs vnode operations.
1021 */
0961aa92 1022struct vnodeopv_entry_desc procfs_vnodeop_entries[] = {
2d3e977e 1023 { &vop_default_desc, vop_defaultop },
625ddaba
JS
1024 { &vop_access_desc, (vnodeopv_entry_t) procfs_access },
1025 { &vop_advlock_desc, (vnodeopv_entry_t) procfs_badop },
1026 { &vop_bmap_desc, (vnodeopv_entry_t) procfs_bmap },
1027 { &vop_close_desc, (vnodeopv_entry_t) procfs_close },
e62afb5f 1028 { &vop_old_create_desc, (vnodeopv_entry_t) procfs_badop },
625ddaba
JS
1029 { &vop_getattr_desc, (vnodeopv_entry_t) procfs_getattr },
1030 { &vop_inactive_desc, (vnodeopv_entry_t) procfs_inactive },
e62afb5f
MD
1031 { &vop_old_link_desc, (vnodeopv_entry_t) procfs_badop },
1032 { &vop_old_lookup_desc, (vnodeopv_entry_t) procfs_lookup },
1033 { &vop_old_mkdir_desc, (vnodeopv_entry_t) procfs_badop },
1034 { &vop_old_mknod_desc, (vnodeopv_entry_t) procfs_badop },
625ddaba
JS
1035 { &vop_open_desc, (vnodeopv_entry_t) procfs_open },
1036 { &vop_pathconf_desc, (vnodeopv_entry_t) vop_stdpathconf },
1037 { &vop_print_desc, (vnodeopv_entry_t) procfs_print },
1038 { &vop_read_desc, (vnodeopv_entry_t) procfs_rw },
1039 { &vop_readdir_desc, (vnodeopv_entry_t) procfs_readdir },
1040 { &vop_readlink_desc, (vnodeopv_entry_t) procfs_readlink },
1041 { &vop_reclaim_desc, (vnodeopv_entry_t) procfs_reclaim },
e62afb5f
MD
1042 { &vop_old_remove_desc, (vnodeopv_entry_t) procfs_badop },
1043 { &vop_old_rename_desc, (vnodeopv_entry_t) procfs_badop },
1044 { &vop_old_rmdir_desc, (vnodeopv_entry_t) procfs_badop },
625ddaba 1045 { &vop_setattr_desc, (vnodeopv_entry_t) procfs_setattr },
e62afb5f 1046 { &vop_old_symlink_desc, (vnodeopv_entry_t) procfs_badop },
625ddaba
JS
1047 { &vop_write_desc, (vnodeopv_entry_t) procfs_rw },
1048 { &vop_ioctl_desc, (vnodeopv_entry_t) procfs_ioctl },
984263bc
MD
1049 { NULL, NULL }
1050};
984263bc 1051