2 * Copyright (c) 1993, 1995 Jan-Simon Pendry
3 * Copyright (c) 1993, 1995
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
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. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)procfs_vnops.c 8.18 (Berkeley) 5/21/95
35 * $FreeBSD: src/sys/miscfs/procfs/procfs_vnops.c,v 1.76.2.7 2002/01/22 17:22:59 nectar Exp $
39 * procfs vnode interface
42 #include <sys/param.h>
43 #include <sys/systm.h>
45 #include <sys/kernel.h>
47 #include <sys/fcntl.h>
50 #include <sys/signalvar.h>
51 #include <sys/vnode.h>
53 #include <sys/mount.h>
54 #include <sys/namei.h>
55 #include <sys/dirent.h>
56 #include <sys/malloc.h>
58 #include <vm/vm_zone.h>
59 #include <vfs/procfs/procfs.h>
60 #include <sys/pioctl.h>
62 #include <sys/spinlock2.h>
64 #include <machine/limits.h>
66 static int procfs_access (struct vop_access_args *);
67 static int procfs_badop (struct vop_generic_args *);
68 static int procfs_bmap (struct vop_bmap_args *);
69 static int procfs_close (struct vop_close_args *);
70 static int procfs_getattr (struct vop_getattr_args *);
71 static int procfs_inactive (struct vop_inactive_args *);
72 static int procfs_ioctl (struct vop_ioctl_args *);
73 static int procfs_lookup (struct vop_old_lookup_args *);
74 static int procfs_open (struct vop_open_args *);
75 static int procfs_print (struct vop_print_args *);
76 static int procfs_readdir (struct vop_readdir_args *);
77 static int procfs_readlink (struct vop_readlink_args *);
78 static int procfs_reclaim (struct vop_reclaim_args *);
79 static int procfs_setattr (struct vop_setattr_args *);
81 static int procfs_readdir_proc(struct vop_readdir_args *);
82 static int procfs_readdir_root(struct vop_readdir_args *);
85 * procfs vnode operations.
87 struct vop_ops procfs_vnode_vops = {
88 .vop_default = vop_defaultop,
89 .vop_access = procfs_access,
90 .vop_advlock = (void *)procfs_badop,
91 .vop_bmap = procfs_bmap,
92 .vop_close = procfs_close,
93 .vop_old_create = (void *)procfs_badop,
94 .vop_getattr = procfs_getattr,
95 .vop_inactive = procfs_inactive,
96 .vop_old_link = (void *)procfs_badop,
97 .vop_old_lookup = procfs_lookup,
98 .vop_old_mkdir = (void *)procfs_badop,
99 .vop_old_mknod = (void *)procfs_badop,
100 .vop_open = procfs_open,
101 .vop_pathconf = vop_stdpathconf,
102 .vop_print = procfs_print,
103 .vop_read = procfs_rw,
104 .vop_readdir = procfs_readdir,
105 .vop_readlink = procfs_readlink,
106 .vop_reclaim = procfs_reclaim,
107 .vop_old_remove = (void *)procfs_badop,
108 .vop_old_rename = (void *)procfs_badop,
109 .vop_old_rmdir = (void *)procfs_badop,
110 .vop_setattr = procfs_setattr,
111 .vop_old_symlink = (void *)procfs_badop,
112 .vop_write = (void *)procfs_rw,
113 .vop_ioctl = procfs_ioctl
118 * This is a list of the valid names in the
119 * process-specific sub-directories. It is
120 * used in procfs_lookup and procfs_readdir
122 static struct proc_target {
127 int (*pt_valid) (struct lwp *p);
129 #define N(s) sizeof(s)-1, s
130 /* name type validp */
131 { DT_DIR, N("."), Pproc, NULL },
132 { DT_DIR, N(".."), Proot, NULL },
133 { DT_REG, N("mem"), Pmem, NULL },
134 { DT_REG, N("regs"), Pregs, procfs_validregs },
135 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs },
136 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs },
137 { DT_REG, N("ctl"), Pctl, NULL },
138 { DT_REG, N("status"), Pstatus, NULL },
139 { DT_REG, N("note"), Pnote, NULL },
140 { DT_REG, N("notepg"), Pnotepg, NULL },
141 { DT_REG, N("map"), Pmap, procfs_validmap },
142 { DT_REG, N("etype"), Ptype, procfs_validtype },
143 { DT_REG, N("cmdline"), Pcmdline, NULL },
144 { DT_REG, N("rlimit"), Prlimit, NULL },
145 { DT_LNK, N("file"), Pfile, NULL },
148 static const int nproc_targets = NELEM(proc_targets);
150 static pid_t atopid (const char *, u_int);
153 * set things up for doing i/o on
154 * the pfsnode (vp). (vp) is locked
155 * on entry, and should be left locked
158 * for procfs we don't need to do anything
159 * in particular for i/o. all that is done
160 * is to support exclusive open on process
163 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
167 procfs_open(struct vop_open_args *ap)
169 struct pfsnode *pfs = VTOPFS(ap->a_vp);
170 struct proc *p1, *p2;
173 p2 = pfs_pfind(pfs->pfs_pid);
176 if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred)) {
181 switch (pfs->pfs_type) {
183 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
184 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) {
191 /* Can't trace a process that's currently exec'ing. */
192 if ((p2->p_flags & P_INEXEC) != 0) {
196 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred)) {
201 if (ap->a_mode & FWRITE)
202 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
209 error = vop_stdopen(ap);
216 * close the pfsnode (vp) after doing i/o.
217 * (vp) is not locked on entry or exit.
219 * nothing to do for procfs other than undo
220 * any exclusive open flag (see _open above).
222 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
225 procfs_close(struct vop_close_args *ap)
227 struct pfsnode *pfs = VTOPFS(ap->a_vp);
231 * Make sure the lock is exclusive for opencount tests
233 vn_lock(ap->a_vp, LK_UPGRADE | LK_RETRY);
235 switch (pfs->pfs_type) {
237 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
238 pfs->pfs_flags &= ~(FWRITE|O_EXCL);
240 * v_opencount determines the last real close on the vnode.
242 * If this is the last close, then it checks to see if
243 * the target process has PF_LINGER set in p_pfsflags,
244 * if this is *not* the case, then the process' stop flags
245 * are cleared, and the process is woken up. This is
246 * to help prevent the case where a process has been
247 * told to stop on an event, but then the requesting process
248 * has gone away or forgotten about it.
251 if ((ap->a_vp->v_opencount < 2)
252 && ((p = pfs_pfind(pfs->pfs_pid)) != NULL ||
253 (p = pfs_zpfind(pfs->pfs_pid)) != NULL)
254 && !(p->p_pfsflags & PF_LINGER)) {
255 spin_lock(&p->p_spin);
258 spin_unlock(&p->p_spin);
268 return (vop_stdclose(ap));
272 * do an ioctl operation on a pfsnode (vp).
273 * (vp) is not locked on entry or exit.
276 procfs_ioctl(struct vop_ioctl_args *ap)
278 struct pfsnode *pfs = VTOPFS(ap->a_vp);
283 struct procfs_status *psp;
286 procp = pfs_pfind(pfs->pfs_pid);
295 /* Can't trace a process that's currently exec'ing. */
296 if ((procp->p_flags & P_INEXEC) != 0) {
300 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred)) {
305 switch (ap->a_command) {
307 spin_lock(&procp->p_spin);
308 procp->p_stops |= *(unsigned int*)ap->a_data;
309 spin_unlock(&procp->p_spin);
312 spin_lock(&procp->p_spin);
313 procp->p_stops &= ~*(unsigned int*)ap->a_data;
314 spin_unlock(&procp->p_spin);
318 * NFLAGS is "non-suser_xxx flags" -- currently, only
319 * PFS_ISUGID ("ignore set u/g id");
321 #define NFLAGS (PF_ISUGID)
322 flags = (unsigned char)*(unsigned int*)ap->a_data;
323 if (flags & NFLAGS && (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)))
325 procp->p_pfsflags = flags;
328 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
332 * NOTE: syscall entry deals with stopevents and may run without
335 psp = (struct procfs_status *)ap->a_data;
336 psp->flags = procp->p_pfsflags;
337 psp->events = procp->p_stops;
338 spin_lock(&procp->p_spin);
341 psp->why = procp->p_stype;
342 psp->val = procp->p_xstat;
343 spin_unlock(&procp->p_spin);
346 spin_unlock(&procp->p_spin);
347 psp->why = 0; /* Not defined values */
348 psp->val = 0; /* Not defined values */
353 * NOTE: syscall entry deals with stopevents and may run without
356 psp = (struct procfs_status *)ap->a_data;
357 spin_lock(&procp->p_spin);
358 while (procp->p_step == 0) {
359 tsleep_interlock(&procp->p_stype, PCATCH);
360 spin_unlock(&procp->p_spin);
361 if (procp->p_stops == 0) {
365 if (procp->p_flags & P_POSTEXIT) {
369 if (procp->p_flags & P_INEXEC) {
373 error = tsleep(&procp->p_stype, PCATCH | PINTERLOCKED,
377 spin_lock(&procp->p_spin);
379 spin_unlock(&procp->p_spin);
380 psp->state = 1; /* It stopped */
381 psp->flags = procp->p_pfsflags;
382 psp->events = procp->p_stops;
383 psp->why = procp->p_stype; /* why it stopped */
384 psp->val = procp->p_xstat; /* any extra info */
386 case PIOCCONT: /* Restart a proc */
388 * NOTE: syscall entry deals with stopevents and may run without
389 * the MP lock. However, the caller is presumably interlocked
392 if (procp->p_step == 0) {
393 error = EINVAL; /* Can only start a stopped process */
396 if ((signo = *(int*)ap->a_data) != 0) {
397 if (signo >= NSIG || signo <= 0) {
401 ksignal(procp, signo);
404 wakeup(&procp->p_step);
417 * do block mapping for pfsnode (vp).
418 * since we don't use the buffer cache
419 * for procfs this function should never
420 * be called. in any case, it's not clear
421 * what part of the kernel ever makes use
422 * of this function. for sanity, this is the
423 * usual no-op bmap, although returning
424 * (EIO) would be a reasonable alternative.
426 * XXX mmap assumes buffer cache operation
428 * procfs_bmap(struct vnode *a_vp, off_t a_loffset,
429 * off_t *a_doffsetp, int *a_runp, int *a_runb)
432 procfs_bmap(struct vop_bmap_args *ap)
434 if (ap->a_doffsetp != NULL)
435 *ap->a_doffsetp = ap->a_loffset;
436 if (ap->a_runp != NULL)
438 if (ap->a_runb != NULL)
444 * procfs_inactive is called when the pfsnode
445 * is vrele'd and the reference count goes
446 * to zero. (vp) will be on the vnode free
447 * list, so to get it back vget() must be
450 * (vp) is locked on entry, but must be unlocked on exit.
452 * procfs_inactive(struct vnode *a_vp)
455 procfs_inactive(struct vop_inactive_args *ap)
457 struct pfsnode *pfs = VTOPFS(ap->a_vp);
459 if (pfs->pfs_pid & PFS_DEAD)
465 * _reclaim is called when getnewvnode()
466 * wants to make use of an entry on the vnode
467 * free list. at this time the filesystem needs
468 * to free any private data and remove the node
469 * from any private lists.
471 * procfs_reclaim(struct vnode *a_vp)
474 procfs_reclaim(struct vop_reclaim_args *ap)
476 return (procfs_freevp(ap->a_vp));
480 * _print is used for debugging.
481 * just print a readable description
484 * procfs_print(struct vnode *a_vp)
487 procfs_print(struct vop_print_args *ap)
489 struct pfsnode *pfs = VTOPFS(ap->a_vp);
491 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
492 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
497 * generic entry point for unsupported operations
500 procfs_badop(struct vop_generic_args *ap)
506 * Invent attributes for pfsnode (vp) and store
508 * Directories lengths are returned as zero since
509 * any real length would require the genuine size
510 * to be computed, and nothing cares anyway.
512 * this is relatively minimal for procfs.
514 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap)
517 procfs_getattr(struct vop_getattr_args *ap)
519 struct pfsnode *pfs = VTOPFS(ap->a_vp);
520 struct vattr *vap = ap->a_vap;
525 * First make sure that the process and its credentials
528 switch (pfs->pfs_type) {
534 procp = pfs_pfind(pfs->pfs_pid);
535 if (procp == NULL || procp->p_ucred == NULL) {
544 /* start by zeroing out the attributes */
547 /* next do all the common fields */
548 vap->va_type = ap->a_vp->v_type;
549 vap->va_mode = pfs->pfs_mode;
550 vap->va_fileid = pfs->pfs_fileno;
552 vap->va_blocksize = PAGE_SIZE;
553 vap->va_bytes = vap->va_size = 0;
554 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
557 * Make all times be current TOD.
558 * It would be possible to get the process start
559 * time from the p_stat structure, but there's
560 * no "file creation" time stamp anyway, and the
561 * p_stat structure is not addressible if u. gets
562 * swapped out for that process.
564 nanotime(&vap->va_ctime);
565 vap->va_atime = vap->va_mtime = vap->va_ctime;
568 * If the process has exercised some setuid or setgid
569 * privilege, then rip away read/write permission so
570 * that only root can gain access.
572 switch (pfs->pfs_type) {
578 if (procp->p_flags & P_SUGID) {
579 vap->va_mode &= ~((VREAD|VWRITE)|
581 ((VREAD|VWRITE)>>6));
589 * now do the object specific fields
591 * The size could be set from struct reg, but it's hardly
592 * worth the trouble, and it puts some (potentially) machine
593 * dependent data into this machine-independent code. If it
594 * becomes important then this function should break out into
595 * a per-file stat function in the corresponding .c file.
600 if (procp->p_ucred) {
601 vap->va_uid = procp->p_ucred->cr_uid;
602 vap->va_gid = procp->p_ucred->cr_gid;
609 switch (pfs->pfs_type) {
612 * Set nlink to 1 to tell fts(3) we don't actually know.
617 vap->va_size = vap->va_bytes = DEV_BSIZE;
621 char buf[16]; /* should be enough */
625 vap->va_size = ksnprintf(buf, sizeof(buf),
626 "%ld", (long)curproc->p_pid);
627 vap->va_bytes = vap->va_size;
632 vap->va_nlink = nproc_targets;
633 vap->va_size = vap->va_bytes = DEV_BSIZE;
637 char *fullpath, *freepath;
639 if (procp->p_textnch.ncp) {
642 cache_copy(&procp->p_textnch, &nch);
643 error = cache_fullpath(procp, &nch, NULL,
644 &fullpath, &freepath, 0);
651 vap->va_size = strlen(fullpath);
652 kfree(freepath, M_TEMP);
654 vap->va_size = sizeof("unknown") - 1;
657 vap->va_bytes = vap->va_size;
663 * If we denied owner access earlier, then we have to
664 * change the owner to root - otherwise 'ps' and friends
665 * will break even though they are setgid kmem. *SIGH*
667 if (procp->p_flags & P_SUGID)
669 else if (procp->p_ucred)
670 vap->va_uid = procp->p_ucred->cr_uid;
676 vap->va_bytes = vap->va_size = sizeof(struct reg);
680 vap->va_bytes = vap->va_size = sizeof(struct fpreg);
684 vap->va_bytes = vap->va_size = sizeof(struct dbreg);
698 panic("procfs_getattr");
706 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap,
707 * struct ucred *a_cred)
710 procfs_setattr(struct vop_setattr_args *ap)
712 if (ap->a_vap->va_flags != VNOVAL)
716 * just fake out attribute setting
717 * it's not good to generate an error
718 * return, otherwise things like creat()
719 * will fail when they try to set the
720 * file length to 0. worse, this means
721 * that echo $note > /proc/$pid/note will fail.
728 * implement access checking.
730 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred)
733 procfs_access(struct vop_access_args *ap)
738 error = VOP_GETATTR(ap->a_vp, &vattr);
740 error = vop_helper_access(ap, vattr.va_uid, vattr.va_gid,
746 * lookup. this is incredibly complicated in the general case, however
747 * for most pseudo-filesystems very little needs to be done.
749 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
750 * struct componentname *a_cnp)
753 procfs_lookup(struct vop_old_lookup_args *ap)
755 struct componentname *cnp = ap->a_cnp;
756 struct vnode **vpp = ap->a_vpp;
757 struct vnode *dvp = ap->a_dvp;
758 char *pname = cnp->cn_nameptr;
759 /* struct proc *curp = cnp->cn_proc; */
760 struct proc_target *pt;
770 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME)
775 if (cnp->cn_namelen == 1 && *pname == '.') {
782 switch (pfs->pfs_type) {
784 if (cnp->cn_flags & CNP_ISDOTDOT)
787 if (CNEQ(cnp, "curproc", 7)) {
788 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc);
792 pid = atopid(pname, cnp->cn_namelen);
800 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
803 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
804 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
807 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
811 if (cnp->cn_flags & CNP_ISDOTDOT) {
812 error = procfs_root(dvp->v_mount, vpp);
816 p = pfs_pfind(pfs->pfs_pid);
820 lp = FIRST_LWP_IN_PROC(p);
822 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
825 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
826 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
829 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
830 if (cnp->cn_namelen == pt->pt_namlen &&
831 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
832 (pt->pt_valid == NULL || (*pt->pt_valid)(lp)))
837 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
845 if (cnp->cn_nameiop == NAMEI_LOOKUP)
850 * If no error occured *vpp will hold a referenced locked vnode.
851 * dvp was passed to us locked and *vpp must be returned locked.
852 * If *vpp != dvp then we should unlock dvp if (1) this is not the
853 * last component or (2) CNP_LOCKPARENT is not set.
856 if (error == 0 && *vpp != dvp) {
857 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) {
858 cnp->cn_flags |= CNP_PDIRUNLOCK;
867 * Does this process have a text file?
870 procfs_validfile(struct lwp *lp)
872 return (procfs_findtextvp(lp->lwp_proc) != NULLVP);
876 * readdir() returns directory entries from pfsnode (vp).
878 * We generate just one directory entry at a time, as it would probably
879 * not pay off to buffer several entries locally to save uiomove calls.
881 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
882 * int *a_eofflag, int *a_ncookies, off_t **a_cookies)
885 procfs_readdir(struct vop_readdir_args *ap)
890 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
892 error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY | LK_FAILRECLAIM);
895 pfs = VTOPFS(ap->a_vp);
897 switch (pfs->pfs_type) {
900 * this is for the process-specific sub-directories.
901 * all that is needed to is copy out all the entries
902 * from the procent[] table (top of this file).
904 error = procfs_readdir_proc(ap);
908 * this is for the root of the procfs filesystem
909 * what is needed is a special entry for "curproc"
910 * followed by an entry for each process on allproc
912 error = procfs_readdir_root(ap);
924 procfs_readdir_proc(struct vop_readdir_args *ap)
927 int error, i, retval;
930 struct proc_target *pt;
931 struct uio *uio = ap->a_uio;
933 pfs = VTOPFS(ap->a_vp);
934 p = pfs_pfind(pfs->pfs_pid);
937 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) {
941 /* XXX lwp, not MPSAFE */
942 lp = FIRST_LWP_IN_PROC(p);
945 i = (int)uio->uio_offset;
951 for (pt = &proc_targets[i];
952 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
953 if (pt->pt_valid && (*pt->pt_valid)(lp) == 0)
956 retval = vop_write_dirent(&error, uio,
957 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
958 pt->pt_namlen, pt->pt_name);
963 uio->uio_offset = (off_t)i;
970 struct procfs_readdir_root_info {
978 static int procfs_readdir_root_callback(struct proc *p, void *data);
981 procfs_readdir_root(struct vop_readdir_args *ap)
983 struct procfs_readdir_root_info info;
984 struct uio *uio = ap->a_uio;
989 info.i = (int)uio->uio_offset;
996 info.cred = ap->a_cred;
997 while (info.pcnt < 3) {
998 res = procfs_readdir_root_callback(NULL, &info);
1003 allproc_scan(procfs_readdir_root_callback, &info);
1004 uio->uio_offset = (off_t)info.i;
1006 return (info.error);
1010 procfs_readdir_root_callback(struct proc *p, void *data)
1012 struct procfs_readdir_root_info *info = data;
1017 char d_name_pid[20];
1023 if (uio->uio_resid <= 0 || info->error)
1026 switch (info->pcnt) {
1028 d_ino = PROCFS_FILENO(0, Proot);
1034 d_ino = PROCFS_FILENO(0, Proot);
1041 d_ino = PROCFS_FILENO(0, Pcurproc);
1049 if (!PRISON_CHECK(info->cred, p->p_ucred))
1051 if (ps_showallprocs == 0 &&
1052 info->cred->cr_uid != 0 &&
1053 info->cred->cr_uid != p->p_ucred->cr_uid) {
1058 * Skip entries we have already returned (optimization)
1060 if (info->pcnt < info->i) {
1065 d_ino = PROCFS_FILENO(p->p_pid, Pproc);
1066 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid),
1067 "%ld", (long)p->p_pid);
1068 d_name = d_name_pid;
1074 * Skip entries we have already returned (optimization)
1076 if (info->pcnt < info->i) {
1081 retval = vop_write_dirent(&info->error, uio,
1082 d_ino, d_type, d_namlen, d_name);
1091 * readlink reads the link of `curproc' or `file'
1094 procfs_readlink(struct vop_readlink_args *ap)
1096 char buf[16]; /* should be enough */
1098 struct vnode *vp = ap->a_vp;
1099 struct pfsnode *pfs = VTOPFS(vp);
1100 char *fullpath, *freepath;
1103 switch (pfs->pfs_type) {
1105 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
1108 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
1110 return (uiomove(buf, len, ap->a_uio));
1113 * procfs's directory topology is somewhat asynchronous from
1114 * reality so it is possible for pid requests to race exiting
1115 * processes. In this situation, bit 31 is set in
1116 * pfs->pfs_pid which guarantees that pfs_pfind() will return
1119 * It is also possible to catch a process in the middle of
1120 * an exit sequence so various fields might wind up being
1121 * NULL that are not normally NULL.
1123 procp = pfs_pfind(pfs->pfs_pid);
1124 if (procp == NULL || procp->p_ucred == NULL) {
1126 return (uiomove("unknown", sizeof("unknown") - 1,
1129 if (procp->p_textnch.ncp) {
1130 struct nchandle nch;
1132 cache_copy(&procp->p_textnch, &nch);
1133 error = cache_fullpath(procp, &nch, NULL,
1134 &fullpath, &freepath, 0);
1142 return (uiomove("unknown", sizeof("unknown") - 1,
1145 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1146 kfree(freepath, M_TEMP);
1155 * convert decimal ascii to pid_t
1158 atopid(const char *b, u_int len)
1164 if (c < '0' || c > '9')
1166 p = 10 * p + (c - '0');