2 * Copyright (c) 2000 Dag-Erling Coïdan Smørgrav
3 * Copyright (c) 1999 Pierre Beyssac
4 * Copyright (c) 1993, 1995 Jan-Simon Pendry
5 * Copyright (c) 1993, 1995
6 * The Regents of the University of California. All rights reserved.
8 * This code is derived from software contributed to Berkeley by
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * @(#)procfs_vnops.c 8.18 (Berkeley) 5/21/95
41 * $FreeBSD: src/sys/i386/linux/linprocfs/linprocfs_vnops.c,v 1.3.2.5 2001/08/12 14:29:19 rwatson Exp $
45 * procfs vnode interface
48 #include <sys/param.h>
49 #include <sys/systm.h>
51 #include <sys/kernel.h>
53 #include <sys/fcntl.h>
56 #include <sys/signalvar.h>
57 #include <sys/vnode.h>
58 #include <sys/mount.h>
59 #include <sys/namei.h>
60 #include <sys/dirent.h>
61 #include <sys/malloc.h>
64 #include <vm/vm_zone.h>
65 #include "linprocfs.h"
66 #include <sys/pioctl.h>
67 #include <sys/spinlock2.h>
69 #include <machine/limits.h>
71 extern struct vnode *procfs_findtextvp (struct proc *);
73 static int linprocfs_access (struct vop_access_args *);
74 static int linprocfs_badop (struct vop_generic_args *);
75 static int linprocfs_bmap (struct vop_bmap_args *);
76 static int linprocfs_close (struct vop_close_args *);
77 static int linprocfs_getattr (struct vop_getattr_args *);
78 static int linprocfs_inactive (struct vop_inactive_args *);
79 static int linprocfs_ioctl (struct vop_ioctl_args *);
80 static int linprocfs_lookup (struct vop_old_lookup_args *);
81 static int linprocfs_open (struct vop_open_args *);
82 static int linprocfs_print (struct vop_print_args *);
83 static int linprocfs_readdir (struct vop_readdir_args *);
84 static int linprocfs_readlink (struct vop_readlink_args *);
85 static int linprocfs_reclaim (struct vop_reclaim_args *);
86 static int linprocfs_setattr (struct vop_setattr_args *);
88 static int linprocfs_readdir_proc(struct vop_readdir_args *);
89 static int linprocfs_readdir_root(struct vop_readdir_args *);
90 static int linprocfs_readdir_net(struct vop_readdir_args *ap);
91 static int linprocfs_readdir_sys(struct vop_readdir_args *ap);
92 static int linprocfs_readdir_syskernel(struct vop_readdir_args *ap);
95 * procfs vnode operations.
97 struct vop_ops linprocfs_vnode_vops = {
98 .vop_default = vop_defaultop,
99 .vop_access = linprocfs_access,
100 .vop_advlock = (void *)linprocfs_badop,
101 .vop_bmap = linprocfs_bmap,
102 .vop_close = linprocfs_close,
103 .vop_old_create = (void *)linprocfs_badop,
104 .vop_getattr = linprocfs_getattr,
105 .vop_inactive = linprocfs_inactive,
106 .vop_old_link = (void *)linprocfs_badop,
107 .vop_old_lookup = linprocfs_lookup,
108 .vop_old_mkdir = (void *)linprocfs_badop,
109 .vop_old_mknod = (void *)linprocfs_badop,
110 .vop_open = linprocfs_open,
111 .vop_pathconf = vop_stdpathconf,
112 .vop_print = linprocfs_print,
113 .vop_read = (void *)linprocfs_rw,
114 .vop_readdir = linprocfs_readdir,
115 .vop_readlink = linprocfs_readlink,
116 .vop_reclaim = linprocfs_reclaim,
117 .vop_old_remove = (void *)linprocfs_badop,
118 .vop_old_rename = (void *)linprocfs_badop,
119 .vop_old_rmdir = (void *)linprocfs_badop,
120 .vop_setattr = linprocfs_setattr,
121 .vop_old_symlink = (void *)linprocfs_badop,
122 .vop_write = (void *)linprocfs_rw,
123 .vop_ioctl = linprocfs_ioctl
127 * This is a list of the valid names in the
128 * process-specific sub-directories. It is
129 * used in linprocfs_lookup and linprocfs_readdir
131 static struct proc_target {
136 int (*pt_valid) (struct proc *p);
138 #define N(s) sizeof(s)-1, s
139 /* name type validp */
140 { DT_DIR, N("."), Pproc, NULL },
141 { DT_DIR, N(".."), Proot, NULL },
142 { DT_REG, N("mem"), Pmem, NULL },
144 { DT_LNK, N("exe"), Pexe, NULL },
145 { DT_LNK, N("cwd"), Pcwd, NULL },
146 { DT_LNK, N("root"), Pprocroot, NULL },
147 { DT_LNK, N("fd"), Pfd, NULL },
149 { DT_REG, N("stat"), Pprocstat, NULL },
150 { DT_REG, N("status"), Pprocstatus, NULL },
151 { DT_REG, N("maps"), Pmaps, NULL },
152 { DT_REG, N("statm"), Pstatm, NULL },
154 { DT_REG, N("cmdline"), Pcmdline, NULL },
155 { DT_REG, N("environ"), Penviron, NULL },
159 static const int nproc_targets = NELEM(proc_targets);
161 static pid_t atopid (const char *, u_int);
164 * set things up for doing i/o on
165 * the pfsnode (vp). (vp) is locked
166 * on entry, and should be left locked
169 * for procfs we don't need to do anything
170 * in particular for i/o. all that is done
171 * is to support exclusive open on process
175 linprocfs_open(struct vop_open_args *ap)
177 struct pfsnode *pfs = VTOPFS(ap->a_vp);
181 p2 = linprocfs_pfind(pfs->pfs_pid);
184 } else if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred)) {
189 switch (pfs->pfs_type) {
191 if (((pfs->pfs_flags & FWRITE) &&
192 (ap->a_mode & O_EXCL)) ||
193 ((pfs->pfs_flags & O_EXCL) &&
194 (ap->a_mode & FWRITE))) {
199 if (p_trespass(ap->a_cred, p2->p_ucred)) {
203 if (ap->a_mode & FWRITE)
204 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
211 error = vop_stdopen(ap);
218 * close the pfsnode (vp) after doing i/o.
219 * (vp) is not locked on entry or exit.
221 * nothing to do for procfs other than undo
222 * any exclusive open flag (see _open above).
225 linprocfs_close(struct vop_close_args *ap)
227 struct pfsnode *pfs = VTOPFS(ap->a_vp);
230 switch (pfs->pfs_type) {
232 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
233 pfs->pfs_flags &= ~(FWRITE|O_EXCL);
235 * If this is the last close, then it checks to see if
236 * the target process has PF_LINGER set in p_pfsflags,
237 * if this is *not* the case, then the process' stop flags
238 * are cleared, and the process is woken up. This is
239 * to help prevent the case where a process has been
240 * told to stop on an event, but then the requesting process
241 * has gone away or forgotten about it.
244 if ((ap->a_vp->v_opencount < 2)
245 && (p = linprocfs_pfind(pfs->pfs_pid))
246 && !(p->p_pfsflags & PF_LINGER)) {
257 return (vop_stdclose(ap));
261 * do an ioctl operation on a pfsnode (vp).
262 * (vp) is not locked on entry or exit.
265 linprocfs_ioctl(struct vop_ioctl_args *ap)
267 struct pfsnode *pfs = VTOPFS(ap->a_vp);
271 struct procfs_status *psp;
274 procp = linprocfs_pfind(pfs->pfs_pid);
278 if (p_trespass(ap->a_cred, procp->p_ucred)) {
283 switch (ap->a_command) {
285 procp->p_stops |= *(unsigned int*)ap->a_data;
288 procp->p_stops &= ~*(unsigned int*)ap->a_data;
292 * NFLAGS is "non-suser_xxx flags" -- currently, only
293 * PFS_ISUGID ("ignore set u/g id");
295 #define NFLAGS (PF_ISUGID)
296 flags = (unsigned char)*(unsigned int*)ap->a_data;
297 if (flags & NFLAGS && (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)))
299 procp->p_pfsflags = flags;
302 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
304 psp = (struct procfs_status *)ap->a_data;
305 psp->state = (procp->p_step == 0);
306 psp->flags = procp->p_pfsflags;
307 psp->events = procp->p_stops;
309 psp->why = procp->p_stype;
310 psp->val = procp->p_xstat;
312 psp->why = psp->val = 0; /* Not defined values */
316 psp = (struct procfs_status *)ap->a_data;
317 spin_lock(&procp->p_spin);
318 if (procp->p_step == 0) {
319 spin_unlock(&procp->p_spin);
320 tsleep_interlock(&procp->p_stype, PCATCH);
321 if (procp->p_stops == 0) {
325 if (procp->p_flags & P_POSTEXIT) {
329 if (procp->p_flags & P_INEXEC) {
333 error = tsleep(&procp->p_stype, PCATCH | PINTERLOCKED,
338 spin_unlock(&procp->p_spin);
340 psp->state = 1; /* It stopped */
341 psp->flags = procp->p_pfsflags;
342 psp->events = procp->p_stops;
343 psp->why = procp->p_stype; /* why it stopped */
344 psp->val = procp->p_xstat; /* any extra info */
346 case PIOCCONT: /* Restart a proc */
347 if (procp->p_step == 0) {
348 error = EINVAL; /* Can only start a stopped process */
351 if ((signo = *(int*)ap->a_data) != 0) {
352 if (signo >= NSIG || signo <= 0) {
356 ksignal(procp, signo);
359 wakeup(&procp->p_step);
373 * do block mapping for pfsnode (vp).
374 * since we don't use the buffer cache
375 * for procfs this function should never
376 * be called. in any case, it's not clear
377 * what part of the kernel ever makes use
378 * of this function. for sanity, this is the
379 * usual no-op bmap, although returning
380 * (EIO) would be a reasonable alternative.
383 linprocfs_bmap(struct vop_bmap_args *ap)
385 if (ap->a_doffsetp != NULL)
386 *ap->a_doffsetp = ap->a_loffset;
387 if (ap->a_runp != NULL)
389 if (ap->a_runb != NULL)
395 * linprocfs_inactive is called when the pfsnode
396 * is vrele'd and the reference count is about
397 * to go to zero. (vp) will be on the vnode free
398 * list, so to get it back vget() must be
401 * (vp) is locked on entry and must remain locked
405 linprocfs_inactive(struct vop_inactive_args *ap)
407 /*struct vnode *vp = ap->a_vp;*/
413 * _reclaim is called when getnewvnode()
414 * wants to make use of an entry on the vnode
415 * free list. at this time the filesystem needs
416 * to free any private data and remove the node
417 * from any private lists.
420 linprocfs_reclaim(struct vop_reclaim_args *ap)
422 return (linprocfs_freevp(ap->a_vp));
426 * _print is used for debugging.
427 * just print a readable description
431 linprocfs_print(struct vop_print_args *ap)
433 struct pfsnode *pfs = VTOPFS(ap->a_vp);
435 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
436 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
441 * generic entry point for unsupported operations
444 linprocfs_badop(struct vop_generic_args *ap __unused)
451 * Invent attributes for pfsnode (vp) and store
453 * Directories lengths are returned as zero since
454 * any real length would require the genuine size
455 * to be computed, and nothing cares anyway.
457 * this is relatively minimal for procfs.
460 linprocfs_getattr(struct vop_getattr_args *ap)
462 struct pfsnode *pfs = VTOPFS(ap->a_vp);
463 struct vattr *vap = ap->a_vap;
468 * First make sure that the process and its credentials
471 switch (pfs->pfs_type) {
478 procp = linprocfs_pfind(pfs->pfs_pid);
479 if (procp == NULL || procp->p_ucred == NULL) {
487 /* start by zeroing out the attributes */
490 /* next do all the common fields */
491 vap->va_type = ap->a_vp->v_type;
492 vap->va_mode = pfs->pfs_mode;
493 vap->va_fileid = pfs->pfs_fileno;
495 vap->va_blocksize = PAGE_SIZE;
496 vap->va_bytes = vap->va_size = 0;
497 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
500 * Make all times be current TOD.
501 * It would be possible to get the process start
502 * time from the p_stat structure, but there's
503 * no "file creation" time stamp anyway, and the
504 * p_stat structure is not addressible if u. gets
505 * swapped out for that process.
507 nanotime(&vap->va_ctime);
508 vap->va_atime = vap->va_mtime = vap->va_ctime;
511 * now do the object specific fields
513 * The size could be set from struct reg, but it's hardly
514 * worth the trouble, and it puts some (potentially) machine
515 * dependent data into this machine-independent code. If it
516 * becomes important then this function should break out into
517 * a per-file stat function in the corresponding .c file.
522 vap->va_uid = procp->p_ucred->cr_uid;
523 vap->va_gid = procp->p_ucred->cr_gid;
526 switch (pfs->pfs_type) {
532 * Set nlink to 1 to tell fts(3) we don't actually know.
537 vap->va_size = vap->va_bytes = DEV_BSIZE;
541 char buf[16]; /* should be enough */
544 vap->va_size = vap->va_bytes =
545 ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
550 vap->va_nlink = nproc_targets;
551 vap->va_size = vap->va_bytes = DEV_BSIZE;
555 char *fullpath, *freepath;
556 error = cache_fullpath(procp, &procp->p_textnch, &fullpath, &freepath, 0);
557 /* error = vn_fullpath(procp, NULL, &fullpath, &freepath); */
559 vap->va_size = strlen(fullpath);
560 kfree(freepath, M_TEMP);
562 vap->va_size = sizeof("unknown") - 1;
565 vap->va_bytes = vap->va_size;
569 char *fullpath, *freepath;
570 error = cache_fullpath(procp, &procp->p_fd->fd_ncdir, &fullpath, &freepath, 0);
572 vap->va_size = strlen(fullpath);
573 kfree(freepath, M_TEMP);
575 vap->va_size = sizeof("unknown") - 1;
578 vap->va_bytes = vap->va_size;
582 struct nchandle *nchp;
583 char *fullpath, *freepath;
584 nchp = jailed(procp->p_ucred) ? &procp->p_fd->fd_njdir : &procp->p_fd->fd_nrdir;
585 error = cache_fullpath(procp, nchp, &fullpath, &freepath, 0);
587 vap->va_size = strlen(fullpath);
588 kfree(freepath, M_TEMP);
590 vap->va_size = sizeof("unknown") - 1;
593 vap->va_bytes = vap->va_size;
597 if (procp == curproc) {
598 vap->va_size = sizeof("/dev/fd") - 1;
601 vap->va_size = sizeof("unknown") - 1;
604 vap->va_bytes = vap->va_size;
620 vap->va_bytes = vap->va_size = 0;
627 * If we denied owner access earlier, then we have to
628 * change the owner to root - otherwise 'ps' and friends
629 * will break even though they are setgid kmem. *SIGH*
631 if (procp->p_flags & P_SUGID)
634 vap->va_uid = procp->p_ucred->cr_uid;
643 vap->va_bytes = vap->va_size = 0;
644 /* uid, gid are already set */
648 panic("linprocfs_getattr");
657 linprocfs_setattr(struct vop_setattr_args *ap)
660 if (ap->a_vap->va_flags != VNOVAL)
664 * just fake out attribute setting
665 * it's not good to generate an error
666 * return, otherwise things like creat()
667 * will fail when they try to set the
668 * file length to 0. worse, this means
669 * that echo $note > /proc/$pid/note will fail.
676 * implement access checking.
678 * something very similar to this code is duplicated
679 * throughout the 4bsd kernel and should be moved
680 * into kern/vfs_subr.c sometime.
682 * actually, the check for super-user is slightly
683 * broken since it will allow read access to write-only
684 * objects. this doesn't cause any particular trouble
685 * but does mean that the i/o entry points need to check
686 * that the operation really does make sense.
689 linprocfs_access(struct vop_access_args *ap)
696 * If you're the super-user,
697 * you always get access.
699 if (ap->a_cred->cr_uid == 0)
703 error = VOP_GETATTR(ap->a_vp, vap);
708 * Access check is based on only one of owner, group, public.
709 * If not owner, then check group. If not a member of the
710 * group, then check public access.
712 if (ap->a_cred->cr_uid != vap->va_uid) {
717 gp = ap->a_cred->cr_groups;
718 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
719 if (vap->va_gid == *gp)
726 if ((vap->va_mode & ap->a_mode) == ap->a_mode)
733 * lookup. this is incredibly complicated in the general case, however
734 * for most pseudo-filesystems very little needs to be done.
737 linprocfs_lookup(struct vop_old_lookup_args *ap)
739 struct componentname *cnp = ap->a_cnp;
740 struct vnode **vpp = ap->a_vpp;
741 struct vnode *dvp = ap->a_dvp;
742 char *pname = cnp->cn_nameptr;
743 struct proc_target *pt;
753 if (cnp->cn_nameiop == NAMEI_DELETE ||
754 cnp->cn_nameiop == NAMEI_RENAME ||
755 cnp->cn_nameiop == NAMEI_CREATE) {
761 if (cnp->cn_namelen == 1 && *pname == '.') {
768 switch (pfs->pfs_type) {
770 if (cnp->cn_flags & CNP_ISDOTDOT) {
771 error = linprocfs_root(dvp->v_mount, vpp);
774 if (CNEQ(cnp, "kernel", 6)) {
775 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Psyskernel);
780 if (cnp->cn_flags & CNP_ISDOTDOT) {
781 error = linprocfs_root(dvp->v_mount, vpp);
784 if (CNEQ(cnp, "dev", 3)) {
785 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pnetdev);
790 if (cnp->cn_flags & CNP_ISDOTDOT) {
791 /* XXX: this is wrong, wrong, wrong. */
792 error = linprocfs_root(dvp->v_mount, vpp);
795 if (CNEQ(cnp, "osrelease", 9)) {
796 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Posrelease);
799 if (CNEQ(cnp, "ostype", 6)) {
800 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Postype);
803 if (CNEQ(cnp, "pid_max", 7)) {
804 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Ppidmax);
807 if (CNEQ(cnp, "version", 7)) {
808 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pversion);
814 if (cnp->cn_flags & CNP_ISDOTDOT)
817 if (CNEQ(cnp, "self", 4)) {
818 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pself);
821 if (CNEQ(cnp, "meminfo", 7)) {
822 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pmeminfo);
825 if (CNEQ(cnp, "cpuinfo", 7)) {
826 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pcpuinfo);
829 if (CNEQ(cnp, "mounts", 6)) {
830 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pmounts);
833 if (CNEQ(cnp, "stat", 4)) {
834 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pstat);
837 if (CNEQ(cnp, "uptime", 6)) {
838 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Puptime);
841 if (CNEQ(cnp, "version", 7)) {
842 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pversion);
845 if (CNEQ(cnp, "loadavg", 7)) {
846 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Ploadavg);
849 if (CNEQ(cnp, "net", 3)) {
850 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pnet);
853 if (CNEQ(cnp, "sys", 3)) {
854 error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Psys);
858 pid = atopid(pname, cnp->cn_namelen);
862 p = linprocfs_pfind(pid);
866 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
869 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
870 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
873 error = linprocfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
877 if (cnp->cn_flags & CNP_ISDOTDOT) {
878 error = linprocfs_root(dvp->v_mount, vpp);
882 p = linprocfs_pfind(pfs->pfs_pid);
886 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
889 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
890 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
893 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
894 if (cnp->cn_namelen == pt->pt_namlen &&
895 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
896 (pt->pt_valid == NULL || (*pt->pt_valid)(p)))
902 error = linprocfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
911 if (cnp->cn_nameiop == NAMEI_LOOKUP)
917 * If no error occured *vpp will hold a referenced locked vnode.
918 * dvp was passed to us locked and *vpp must be returned locked
919 * so if dvp != *vpp and CNP_LOCKPARENT is not set, unlock dvp.
925 if (*vpp != dvp && (cnp->cn_flags & CNP_LOCKPARENT) == 0) {
926 cnp->cn_flags |= CNP_PDIRUNLOCK;
934 * Does this process have a text file?
937 linprocfs_validfile(struct proc *p)
940 return (procfs_findtextvp(p) != NULLVP);
944 * readdir() returns directory entries from pfsnode (vp).
946 * We generate just one directory entry at a time, as it would probably
947 * not pay off to buffer several entries locally to save uiomove calls.
949 * linprocfs_readdir(struct vnode *a_vp, struct uio *a_uio,
950 * struct ucred *a_cred, int *a_eofflag,
951 * int *a_ncookies, off_t **a_cookies)
954 linprocfs_readdir(struct vop_readdir_args *ap)
959 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
962 pfs = VTOPFS(ap->a_vp);
963 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0)
966 switch (pfs->pfs_type) {
969 * This is for the process-specific sub-directories.
970 * all that is needed to is copy out all the entries
971 * from the procent[] table (top of this file).
973 error = linprocfs_readdir_proc(ap);
977 * This is for the root of the procfs filesystem
978 * what is needed is a special entry for "self"
979 * followed by an entry for each process on allproc
981 error = linprocfs_readdir_root(ap);
984 error = linprocfs_readdir_net(ap);
987 error = linprocfs_readdir_sys(ap);
990 error = linprocfs_readdir_syskernel(ap);
1002 linprocfs_readdir_proc(struct vop_readdir_args *ap)
1004 struct pfsnode *pfs;
1005 int error, i, retval;
1007 struct proc_target *pt;
1008 struct uio *uio = ap->a_uio;
1010 pfs = VTOPFS(ap->a_vp);
1011 p = linprocfs_pfind(pfs->pfs_pid);
1014 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) {
1020 i = uio->uio_offset;
1022 for (pt = &proc_targets[i];
1023 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
1024 if (pt->pt_valid && (*pt->pt_valid)(p) == 0)
1027 retval = vop_write_dirent(&error, uio,
1028 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
1029 pt->pt_namlen, pt->pt_name);
1034 uio->uio_offset = i;
1040 struct linprocfs_readdir_root_info {
1049 * Scan the root directory by scanning all process
1051 static int linprocfs_readdir_root_callback(struct proc *p, void *data);
1054 linprocfs_readdir_root(struct vop_readdir_args *ap)
1056 struct linprocfs_readdir_root_info info;
1057 struct uio *uio = ap->a_uio;
1061 info.i = uio->uio_offset;
1064 info.cred = ap->a_cred;
1066 while (info.pcnt < 12) {
1067 res = linprocfs_readdir_root_callback(NULL, &info);
1072 allproc_scan(linprocfs_readdir_root_callback, &info);
1074 uio->uio_offset = info.i;
1079 linprocfs_readdir_root_callback(struct proc *p, void *data)
1081 struct linprocfs_readdir_root_info *info = data;
1083 struct uio *uio = info->uio;
1086 char d_name_pid[20];
1090 switch (info->pcnt) {
1092 d_ino = PROCFS_FILENO(0, Proot);
1098 d_ino = PROCFS_FILENO(0, Proot);
1105 d_ino = PROCFS_FILENO(0, Proot);
1112 d_ino = PROCFS_FILENO(0, Pmeminfo);
1119 d_ino = PROCFS_FILENO(0, Pcpuinfo);
1126 d_ino = PROCFS_FILENO(0, Pstat);
1133 d_ino = PROCFS_FILENO(0, Puptime);
1140 d_ino = PROCFS_FILENO(0, Pversion);
1147 d_ino = PROCFS_FILENO(0, Ploadavg);
1153 d_ino = PROCFS_FILENO(0, Pnet);
1159 d_ino = PROCFS_FILENO(0, Psys);
1165 d_ino = PROCFS_FILENO(0, Pmounts);
1172 d_ino = PROCFS_FILENO(0, Pdevices);
1180 * Ignore processes that aren't in our prison
1182 if (PRISON_CHECK(info->cred, p->p_ucred) == 0)
1186 * Ignore processes that we do not want to be visible.
1188 if (ps_showallprocs == 0 &&
1189 info->cred->cr_uid != 0 &&
1190 info->cred->cr_uid != p->p_ucred->cr_uid) {
1195 * Skip processes we have already read (optimization)
1197 if (info->pcnt < info->i) {
1201 d_ino = PROCFS_FILENO(p->p_pid, Pproc);
1202 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid),
1203 "%ld", (long)p->p_pid);
1204 d_name = d_name_pid;
1210 * Skip processes we have already read
1212 if (info->pcnt < info->i) {
1216 retval = vop_write_dirent(&info->error, info->uio,
1217 d_ino, d_type, d_namlen, d_name);
1219 ++info->pcnt; /* iterate proc candidates scanned */
1220 ++info->i; /* iterate entries written */
1222 if (retval || info->error || uio->uio_resid <= 0)
1228 * Scan the root directory by scanning all process
1230 static int linprocfs_readdir_net_callback(struct proc *p, void *data);
1233 linprocfs_readdir_net(struct vop_readdir_args *ap)
1235 struct linprocfs_readdir_root_info info;
1236 struct uio *uio = ap->a_uio;
1240 info.i = uio->uio_offset;
1243 info.cred = ap->a_cred;
1245 while (info.pcnt < 3) {
1246 res = linprocfs_readdir_net_callback(NULL, &info);
1251 uio->uio_offset = info.i;
1256 linprocfs_readdir_net_callback(struct proc *p, void *data)
1258 struct linprocfs_readdir_root_info *info = data;
1260 struct uio *uio = info->uio;
1266 switch (info->pcnt) {
1268 d_ino = PROCFS_FILENO(0, Pnet);
1274 d_ino = PROCFS_FILENO(0, Proot);
1281 d_ino = PROCFS_FILENO(0, Pnet);
1295 * Skip processes we have already read
1297 if (info->pcnt < info->i) {
1301 retval = vop_write_dirent(&info->error, info->uio,
1302 d_ino, d_type, d_namlen, d_name);
1304 ++info->pcnt; /* iterate proc candidates scanned */
1305 ++info->i; /* iterate entries written */
1307 if (retval || info->error || uio->uio_resid <= 0)
1319 * Scan the root directory by scanning all process
1321 static int linprocfs_readdir_sys_callback(struct proc *p, void *data);
1324 linprocfs_readdir_sys(struct vop_readdir_args *ap)
1326 struct linprocfs_readdir_root_info info;
1327 struct uio *uio = ap->a_uio;
1331 info.i = uio->uio_offset;
1334 info.cred = ap->a_cred;
1336 while (info.pcnt < 3) {
1337 res = linprocfs_readdir_sys_callback(NULL, &info);
1342 uio->uio_offset = info.i;
1347 linprocfs_readdir_sys_callback(struct proc *p, void *data)
1349 struct linprocfs_readdir_root_info *info = data;
1351 struct uio *uio = info->uio;
1357 switch (info->pcnt) {
1359 d_ino = PROCFS_FILENO(0, Psys);
1365 d_ino = PROCFS_FILENO(0, Proot);
1372 d_ino = PROCFS_FILENO(0, Psyskernel);
1386 * Skip processes we have already read
1388 if (info->pcnt < info->i) {
1392 retval = vop_write_dirent(&info->error, info->uio,
1393 d_ino, d_type, d_namlen, d_name);
1395 ++info->pcnt; /* iterate proc candidates scanned */
1396 ++info->i; /* iterate entries written */
1398 if (retval || info->error || uio->uio_resid <= 0)
1408 * Scan the root directory by scanning all process
1410 static int linprocfs_readdir_syskernel_callback(struct proc *p, void *data);
1413 linprocfs_readdir_syskernel(struct vop_readdir_args *ap)
1415 struct linprocfs_readdir_root_info info;
1416 struct uio *uio = ap->a_uio;
1420 info.i = uio->uio_offset;
1423 info.cred = ap->a_cred;
1425 while (info.pcnt < 6) {
1426 res = linprocfs_readdir_syskernel_callback(NULL, &info);
1431 uio->uio_offset = info.i;
1436 linprocfs_readdir_syskernel_callback(struct proc *p, void *data)
1438 struct linprocfs_readdir_root_info *info = data;
1440 struct uio *uio = info->uio;
1446 switch (info->pcnt) {
1448 d_ino = PROCFS_FILENO(0, Psyskernel);
1454 d_ino = PROCFS_FILENO(0, Psys);
1461 d_ino = PROCFS_FILENO(0, Posrelease);
1463 d_name = "osrelease";
1468 d_ino = PROCFS_FILENO(0, Postype);
1475 d_ino = PROCFS_FILENO(0, Pversion);
1482 d_ino = PROCFS_FILENO(0, Ppidmax);
1496 * Skip processes we have already read
1498 if (info->pcnt < info->i) {
1502 retval = vop_write_dirent(&info->error, info->uio,
1503 d_ino, d_type, d_namlen, d_name);
1505 ++info->pcnt; /* iterate proc candidates scanned */
1506 ++info->i; /* iterate entries written */
1508 if (retval || info->error || uio->uio_resid <= 0)
1514 * readlink reads the link of `self' or `exe'
1517 linprocfs_readlink(struct vop_readlink_args *ap)
1519 char buf[16]; /* should be enough */
1521 struct vnode *vp = ap->a_vp;
1522 struct nchandle *nchp;
1523 struct pfsnode *pfs = VTOPFS(vp);
1524 char *fullpath, *freepath;
1530 switch (pfs->pfs_type) {
1532 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pself))
1535 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
1537 error = uiomove(buf, len, ap->a_uio);
1540 * There _should_ be no way for an entire process to disappear
1544 procp = linprocfs_pfind(pfs->pfs_pid);
1545 if (procp == NULL || procp->p_ucred == NULL) {
1546 kprintf("linprocfs_readlink: pid %d disappeared\n",
1548 error = uiomove("unknown", sizeof("unknown") - 1,
1552 error = cache_fullpath(procp, &procp->p_textnch, &fullpath, &freepath, 0);
1554 error = uiomove("unknown", sizeof("unknown") - 1,
1558 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1559 kfree(freepath, M_TEMP);
1562 procp = linprocfs_pfind(pfs->pfs_pid);
1563 if (procp == NULL || procp->p_ucred == NULL) {
1564 kprintf("linprocfs_readlink: pid %d disappeared\n",
1566 error = uiomove("unknown", sizeof("unknown") - 1,
1570 error = cache_fullpath(procp, &procp->p_fd->fd_ncdir,
1571 &fullpath, &freepath, 0);
1573 error = uiomove("unknown", sizeof("unknown") - 1,
1577 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1578 kfree(freepath, M_TEMP);
1581 procp = linprocfs_pfind(pfs->pfs_pid);
1582 if (procp == NULL || procp->p_ucred == NULL) {
1583 kprintf("linprocfs_readlink: pid %d disappeared\n",
1585 error = uiomove("unknown", sizeof("unknown") - 1,
1589 nchp = jailed(procp->p_ucred) ? &procp->p_fd->fd_njdir : &procp->p_fd->fd_nrdir;
1590 error = cache_fullpath(procp, nchp, &fullpath, &freepath, 0);
1592 error = uiomove("unknown", sizeof("unknown") - 1,
1596 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1597 kfree(freepath, M_TEMP);
1600 procp = linprocfs_pfind(pfs->pfs_pid);
1601 if (procp == NULL || procp->p_ucred == NULL) {
1602 kprintf("linprocfs_readlink: pid %d disappeared\n",
1604 error = uiomove("unknown", sizeof("unknown") - 1,
1608 if (procp == curproc) {
1609 error = uiomove("/dev/fd", sizeof("/dev/fd") - 1,
1613 error = uiomove("unknown", sizeof("unknown") - 1,
1629 * convert decimal ascii to pid_t
1632 atopid(const char *b, u_int len)
1638 if (c < '0' || c > '9')
1640 p = 10 * p + (c - '0');