2 * Copyright (c) 2001 Dag-Erling Coïdan Smørgrav
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_pseudofs.h"
34 #include <sys/param.h>
35 #include <sys/kernel.h>
36 #include <sys/systm.h>
37 #include <sys/ctype.h>
38 #include <sys/dirent.h>
39 #include <sys/fcntl.h>
40 #include <sys/limits.h>
42 #include <sys/malloc.h>
43 #include <sys/mount.h>
44 #include <sys/mutex.h>
45 #include <sys/namei.h>
49 #include <sys/sysctl.h>
50 #include <sys/vnode.h>
52 #include <fs/pseudofs/pseudofs.h>
53 #include <fs/pseudofs/pseudofs_internal.h>
55 #define KASSERT_PN_IS_DIR(pn) \
56 KASSERT((pn)->pn_type == pfstype_root || \
57 (pn)->pn_type == pfstype_dir || \
58 (pn)->pn_type == pfstype_procdir, \
59 ("%s(): VDIR vnode refers to non-directory pfs_node", __func__))
61 #define KASSERT_PN_IS_FILE(pn) \
62 KASSERT((pn)->pn_type == pfstype_file, \
63 ("%s(): VREG vnode refers to non-file pfs_node", __func__))
65 #define KASSERT_PN_IS_LINK(pn) \
66 KASSERT((pn)->pn_type == pfstype_symlink, \
67 ("%s(): VLNK vnode refers to non-link pfs_node", __func__))
70 * Returns the fileno, adjusted for target pid
73 pn_fileno(struct pfs_node *pn, pid_t pid)
76 KASSERT(pn->pn_fileno > 0,
77 ("%s(): no fileno allocated", __func__));
79 return (pn->pn_fileno * NO_PID + pid);
80 return (pn->pn_fileno);
84 * Returns non-zero if given file is visible to given thread.
87 pfs_visible_proc(struct thread *td, struct pfs_node *pn, struct proc *proc)
94 PROC_LOCK_ASSERT(proc, MA_OWNED);
96 visible = ((proc->p_flag & P_WEXIT) == 0);
98 visible = (p_cansee(td, proc) == 0);
99 if (visible && pn->pn_vis != NULL)
100 visible = pn_vis(td, proc, pn);
107 pfs_visible(struct thread *td, struct pfs_node *pn, pid_t pid, struct proc **p)
111 PFS_TRACE(("%s (pid: %d, req: %d)",
112 pn->pn_name, pid, td->td_proc->p_pid));
118 if ((proc = pfind(pid)) == NULL)
120 if (pfs_visible_proc(td, pn, proc)) {
135 pfs_access(struct vop_access_args *va)
137 struct vnode *vn = va->a_vp;
138 struct pfs_vdata *pvd = vn->v_data;
142 PFS_TRACE(("%s", pvd->pvd_pn->pn_name));
145 error = VOP_GETATTR(vn, &vattr, va->a_cred);
148 error = vaccess(vn->v_type, vattr.va_mode, vattr.va_uid,
149 vattr.va_gid, va->a_accmode, va->a_cred, NULL);
154 * Close a file or directory
157 pfs_close(struct vop_close_args *va)
159 struct vnode *vn = va->a_vp;
160 struct pfs_vdata *pvd = vn->v_data;
161 struct pfs_node *pn = pvd->pvd_pn;
165 PFS_TRACE(("%s", pn->pn_name));
166 pfs_assert_not_owned(pn);
169 * Do nothing unless this is the last close and the node has a
170 * last-close handler.
172 if (vrefcnt(vn) > 1 || pn->pn_close == NULL)
175 if (pvd->pvd_pid != NO_PID) {
176 proc = pfind(pvd->pvd_pid);
181 error = pn_close(va->a_td, proc, pn);
190 * Get file attributes
193 pfs_getattr(struct vop_getattr_args *va)
195 struct vnode *vn = va->a_vp;
196 struct pfs_vdata *pvd = vn->v_data;
197 struct pfs_node *pn = pvd->pvd_pn;
198 struct vattr *vap = va->a_vap;
202 PFS_TRACE(("%s", pn->pn_name));
203 pfs_assert_not_owned(pn);
205 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
208 vap->va_type = vn->v_type;
209 vap->va_fileid = pn_fileno(pn, pvd->pvd_pid);
211 vap->va_blocksize = PAGE_SIZE;
212 vap->va_bytes = vap->va_size = 0;
214 vap->va_fsid = vn->v_mount->mnt_stat.f_fsid.val[0];
216 nanotime(&vap->va_ctime);
217 vap->va_atime = vap->va_mtime = vap->va_ctime;
219 switch (pn->pn_type) {
220 case pfstype_procdir:
225 /* compute link count */
231 case pfstype_symlink:
235 printf("shouldn't be here!\n");
241 vap->va_uid = proc->p_ucred->cr_ruid;
242 vap->va_gid = proc->p_ucred->cr_rgid;
248 if (pn->pn_attr != NULL)
249 error = pn_attr(curthread, proc, pn, vap);
261 pfs_ioctl(struct vop_ioctl_args *va)
264 struct pfs_vdata *pvd;
270 vn_lock(vn, LK_SHARED | LK_RETRY);
271 if (vn->v_iflag & VI_DOOMED) {
278 PFS_TRACE(("%s: %lx", pn->pn_name, va->a_command));
279 pfs_assert_not_owned(pn);
281 if (vn->v_type != VREG) {
285 KASSERT_PN_IS_FILE(pn);
287 if (pn->pn_ioctl == NULL) {
293 * This is necessary because process' privileges may
294 * have changed since the open() call.
296 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc)) {
301 error = pn_ioctl(curthread, proc, pn, va->a_command, va->a_data);
314 pfs_getextattr(struct vop_getextattr_args *va)
316 struct vnode *vn = va->a_vp;
317 struct pfs_vdata *pvd = vn->v_data;
318 struct pfs_node *pn = pvd->pvd_pn;
322 PFS_TRACE(("%s", pn->pn_name));
323 pfs_assert_not_owned(pn);
326 * This is necessary because either process' privileges may
327 * have changed since the open() call.
329 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
332 if (pn->pn_getextattr == NULL)
335 error = pn_getextattr(curthread, proc, pn,
336 va->a_attrnamespace, va->a_name, va->a_uio,
337 va->a_size, va->a_cred);
346 * Convert a vnode to its component name
349 pfs_vptocnp(struct vop_vptocnp_args *ap)
351 struct vnode *vp = ap->a_vp;
352 struct vnode **dvp = ap->a_vpp;
353 struct pfs_vdata *pvd = vp->v_data;
354 struct pfs_node *pd = pvd->pvd_pn;
357 char *buf = ap->a_buf;
358 int *buflen = ap->a_buflen;
359 char pidbuf[PFS_NAMELEN];
360 pid_t pid = pvd->pvd_pid;
361 int len, i, error, locked;
368 if (vp->v_type == VDIR && pd->pn_type == pfstype_root) {
373 } else if (vp->v_type == VDIR && pd->pn_type == pfstype_procdir) {
374 len = snprintf(pidbuf, sizeof(pidbuf), "%d", pid);
380 bcopy(pidbuf, buf + i, len);
382 len = strlen(pd->pn_name);
388 bcopy(pd->pn_name, buf + i, len);
395 error = vfs_busy(mp, 0);
400 * vp is held by caller.
402 locked = VOP_ISLOCKED(vp);
405 error = pfs_vncache_alloc(mp, dvp, pn, pid);
407 vn_lock(vp, locked | LK_RETRY);
414 vn_lock(vp, locked | LK_RETRY);
424 * Look up a file or directory
427 pfs_lookup(struct vop_cachedlookup_args *va)
429 struct vnode *vn = va->a_dvp;
430 struct vnode **vpp = va->a_vpp;
431 struct componentname *cnp = va->a_cnp;
432 struct pfs_vdata *pvd = vn->v_data;
433 struct pfs_node *pd = pvd->pvd_pn;
434 struct pfs_node *pn, *pdn = NULL;
436 pid_t pid = pvd->pvd_pid;
438 int error, i, namelen, visible;
440 PFS_TRACE(("%.*s", (int)cnp->cn_namelen, cnp->cn_nameptr));
441 pfs_assert_not_owned(pd);
443 if (vn->v_type != VDIR)
444 PFS_RETURN (ENOTDIR);
445 KASSERT_PN_IS_DIR(pd);
447 error = VOP_ACCESS(vn, VEXEC, cnp->cn_cred, cnp->cn_thread);
452 * Don't support DELETE or RENAME. CREATE is supported so
453 * that O_CREAT will work, but the lookup will still fail if
454 * the file does not exist.
456 if ((cnp->cn_flags & ISLASTCN) &&
457 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
458 PFS_RETURN (EOPNOTSUPP);
460 /* shortcut: check if the name is too long */
461 if (cnp->cn_namelen >= PFS_NAMELEN)
464 /* check that parent directory is visible... */
465 if (!pfs_visible(curthread, pd, pvd->pvd_pid, NULL))
469 namelen = cnp->cn_namelen;
470 pname = cnp->cn_nameptr;
471 if (namelen == 1 && pname[0] == '.') {
481 if (cnp->cn_flags & ISDOTDOT) {
482 if (pd->pn_type == pfstype_root)
484 error = vfs_busy(mp, MBF_NOWAIT);
488 error = vfs_busy(mp, 0);
489 vn_lock(vn, LK_EXCLUSIVE | LK_RETRY);
493 if (vn->v_iflag & VI_DOOMED) {
499 KASSERT(pd->pn_parent != NULL,
500 ("%s(): non-root directory has no parent", __func__));
502 * This one is tricky. Descendents of procdir nodes
503 * inherit their parent's process affinity, but
504 * there's no easy reverse mapping. For simplicity,
505 * we assume that if this node is a procdir, its
506 * parent isn't (which is correct as long as
507 * descendents of procdir nodes are never procdir
510 if (pd->pn_type == pfstype_procdir)
521 for (pn = pd->pn_nodes; pn != NULL; pn = pn->pn_next)
522 if (pn->pn_type == pfstype_procdir)
524 else if (pn->pn_name[namelen] == '\0' &&
525 bcmp(pname, pn->pn_name, namelen) == 0) {
530 /* process dependent node */
531 if ((pn = pdn) != NULL) {
533 for (pid = 0, i = 0; i < namelen && isdigit(pname[i]); ++i)
534 if ((pid = pid * 10 + pname[i] - '0') > PID_MAX)
536 if (i == cnp->cn_namelen) {
547 pfs_assert_not_owned(pd);
548 pfs_assert_not_owned(pn);
549 visible = pfs_visible(curthread, pn, pid, NULL);
555 error = pfs_vncache_alloc(mp, vpp, pn, pid);
559 if (cnp->cn_flags & ISDOTDOT) {
561 vn_lock(vn, LK_EXCLUSIVE | LK_RETRY);
562 if (vn->v_iflag & VI_DOOMED) {
568 if (cnp->cn_flags & MAKEENTRY && !(vn->v_iflag & VI_DOOMED))
569 cache_enter(vn, *vpp, cnp);
572 if (cnp->cn_flags & ISDOTDOT) {
574 vn_lock(vn, LK_EXCLUSIVE | LK_RETRY);
581 * Open a file or directory.
584 pfs_open(struct vop_open_args *va)
586 struct vnode *vn = va->a_vp;
587 struct pfs_vdata *pvd = vn->v_data;
588 struct pfs_node *pn = pvd->pvd_pn;
589 int mode = va->a_mode;
591 PFS_TRACE(("%s (mode 0x%x)", pn->pn_name, mode));
592 pfs_assert_not_owned(pn);
594 /* check if the requested mode is permitted */
595 if (((mode & FREAD) && !(mode & PFS_RD)) ||
596 ((mode & FWRITE) && !(mode & PFS_WR)))
599 /* we don't support locking */
600 if ((mode & O_SHLOCK) || (mode & O_EXLOCK))
601 PFS_RETURN (EOPNOTSUPP);
610 pfs_read(struct vop_read_args *va)
612 struct vnode *vn = va->a_vp;
613 struct pfs_vdata *pvd = vn->v_data;
614 struct pfs_node *pn = pvd->pvd_pn;
615 struct uio *uio = va->a_uio;
617 struct sbuf *sb = NULL;
620 ssize_t buflen, resid;
622 PFS_TRACE(("%s", pn->pn_name));
623 pfs_assert_not_owned(pn);
625 if (vn->v_type != VREG)
627 KASSERT_PN_IS_FILE(pn);
629 if (!(pn->pn_flags & PFS_RD))
632 if (pn->pn_fill == NULL)
636 * This is necessary because either process' privileges may
637 * have changed since the open() call.
639 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
647 locked = VOP_ISLOCKED(vn);
650 if (pn->pn_flags & PFS_RAWRD) {
651 PFS_TRACE(("%zd resid", uio->uio_resid));
652 error = pn_fill(curthread, proc, pn, NULL, uio);
653 PFS_TRACE(("%zd resid", uio->uio_resid));
657 /* beaucoup sanity checks so we don't ask for bogus allocation */
658 if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
659 (offset = uio->uio_offset) != uio->uio_offset ||
660 (resid = uio->uio_resid) != uio->uio_resid ||
661 (buflen = offset + resid) < offset || buflen >= INT_MAX) {
665 if (buflen > MAXPHYS)
668 sb = sbuf_new(sb, NULL, buflen + 1, 0);
674 error = pn_fill(curthread, proc, pn, sb, uio);
682 * XXX: If the buffer overflowed, sbuf_len() will not return
683 * the data length. Then just use the full length because an
684 * overflowed sbuf must be full.
686 if (sbuf_finish(sb) == 0)
687 buflen = sbuf_len(sb);
688 error = uiomove_frombuf(sbuf_data(sb), buflen, uio);
691 vn_lock(vn, locked | LK_RETRY);
699 * Iterate through directory entries
702 pfs_iterate(struct thread *td, struct proc *proc, struct pfs_node *pd,
703 struct pfs_node **pn, struct proc **p)
707 sx_assert(&allproc_lock, SX_SLOCKED);
708 pfs_assert_owned(pd);
713 } else if ((*pn)->pn_type != pfstype_procdir) {
715 *pn = (*pn)->pn_next;
717 if (*pn != NULL && (*pn)->pn_type == pfstype_procdir) {
720 *p = LIST_FIRST(&allproc);
722 *p = LIST_NEXT(*p, p_list);
723 /* out of processes: next node */
725 *pn = (*pn)->pn_next;
734 visible = pfs_visible_proc(td, *pn, *p);
736 } else if (proc != NULL) {
737 visible = pfs_visible_proc(td, *pn, proc);
747 /* Directory entry list */
749 STAILQ_ENTRY(pfsentry) link;
752 STAILQ_HEAD(pfsdirentlist, pfsentry);
755 * Return directory entries.
758 pfs_readdir(struct vop_readdir_args *va)
760 struct vnode *vn = va->a_vp;
761 struct pfs_vdata *pvd = vn->v_data;
762 struct pfs_node *pd = pvd->pvd_pn;
763 pid_t pid = pvd->pvd_pid;
764 struct proc *p, *proc;
767 struct pfsentry *pfsent, *pfsent2;
768 struct pfsdirentlist lst;
774 KASSERT(pd->pn_info == vn->v_mount->mnt_data,
775 ("%s(): pn_info does not match mountpoint", __func__));
776 PFS_TRACE(("%s pid %lu", pd->pn_name, (unsigned long)pid));
777 pfs_assert_not_owned(pd);
779 if (vn->v_type != VDIR)
780 PFS_RETURN (ENOTDIR);
781 KASSERT_PN_IS_DIR(pd);
784 /* only allow reading entire entries */
785 offset = uio->uio_offset;
786 resid = uio->uio_resid;
787 if (offset < 0 || offset % PFS_DELEN != 0 ||
788 (resid && resid < PFS_DELEN))
793 sx_slock(&allproc_lock);
796 /* check if the directory is visible to the caller */
797 if (!pfs_visible(curthread, pd, pid, &proc)) {
798 sx_sunlock(&allproc_lock);
802 KASSERT(pid == NO_PID || proc != NULL,
803 ("%s(): no process for pid %lu", __func__, (unsigned long)pid));
805 /* skip unwanted entries */
806 for (pn = NULL, p = NULL; offset > 0; offset -= PFS_DELEN) {
807 if (pfs_iterate(curthread, proc, pd, &pn, &p) == -1) {
808 /* nothing left... */
812 sx_sunlock(&allproc_lock);
817 /* fill in entries */
818 while (pfs_iterate(curthread, proc, pd, &pn, &p) != -1 &&
819 resid >= PFS_DELEN) {
820 if ((pfsent = malloc(sizeof(struct pfsentry), M_IOV,
821 M_NOWAIT | M_ZERO)) == NULL) {
825 pfsent->entry.d_reclen = PFS_DELEN;
826 pfsent->entry.d_fileno = pn_fileno(pn, pid);
827 /* PFS_DELEN was picked to fit PFS_NAMLEN */
828 for (i = 0; i < PFS_NAMELEN - 1 && pn->pn_name[i] != '\0'; ++i)
829 pfsent->entry.d_name[i] = pn->pn_name[i];
830 pfsent->entry.d_name[i] = 0;
831 pfsent->entry.d_namlen = i;
832 switch (pn->pn_type) {
833 case pfstype_procdir:
835 ("reached procdir node with p == NULL"));
836 pfsent->entry.d_namlen = snprintf(pfsent->entry.d_name,
837 PFS_NAMELEN, "%d", p->p_pid);
843 pfsent->entry.d_type = DT_DIR;
846 pfsent->entry.d_type = DT_REG;
848 case pfstype_symlink:
849 pfsent->entry.d_type = DT_LNK;
852 panic("%s has unexpected node type: %d", pn->pn_name, pn->pn_type);
854 PFS_TRACE(("%s", pfsent->entry.d_name));
855 STAILQ_INSERT_TAIL(&lst, pfsent, link);
862 sx_sunlock(&allproc_lock);
864 STAILQ_FOREACH_SAFE(pfsent, &lst, link, pfsent2) {
866 error = uiomove(&pfsent->entry, PFS_DELEN, uio);
870 PFS_TRACE(("%d bytes", i * PFS_DELEN));
875 * Read a symbolic link
878 pfs_readlink(struct vop_readlink_args *va)
880 struct vnode *vn = va->a_vp;
881 struct pfs_vdata *pvd = vn->v_data;
882 struct pfs_node *pn = pvd->pvd_pn;
883 struct uio *uio = va->a_uio;
884 struct proc *proc = NULL;
889 PFS_TRACE(("%s", pn->pn_name));
890 pfs_assert_not_owned(pn);
892 if (vn->v_type != VLNK)
894 KASSERT_PN_IS_LINK(pn);
896 if (pn->pn_fill == NULL)
899 if (pvd->pvd_pid != NO_PID) {
900 if ((proc = pfind(pvd->pvd_pid)) == NULL)
902 if (proc->p_flag & P_WEXIT) {
910 locked = VOP_ISLOCKED(vn);
913 /* sbuf_new() can't fail with a static buffer */
914 sbuf_new(&sb, buf, sizeof buf, 0);
916 error = pn_fill(curthread, proc, pn, &sb, NULL);
920 vn_lock(vn, locked | LK_RETRY);
928 if (sbuf_finish(&sb) != 0) {
930 PFS_RETURN (ENAMETOOLONG);
933 error = uiomove_frombuf(sbuf_data(&sb), sbuf_len(&sb), uio);
942 pfs_reclaim(struct vop_reclaim_args *va)
944 struct vnode *vn = va->a_vp;
945 struct pfs_vdata *pvd = vn->v_data;
946 struct pfs_node *pn = pvd->pvd_pn;
948 PFS_TRACE(("%s", pn->pn_name));
949 pfs_assert_not_owned(pn);
951 return (pfs_vncache_free(va->a_vp));
958 pfs_setattr(struct vop_setattr_args *va)
960 struct vnode *vn = va->a_vp;
961 struct pfs_vdata *pvd = vn->v_data;
962 struct pfs_node *pn = pvd->pvd_pn;
964 PFS_TRACE(("%s", pn->pn_name));
965 pfs_assert_not_owned(pn);
967 PFS_RETURN (EOPNOTSUPP);
974 pfs_write(struct vop_write_args *va)
976 struct vnode *vn = va->a_vp;
977 struct pfs_vdata *pvd = vn->v_data;
978 struct pfs_node *pn = pvd->pvd_pn;
979 struct uio *uio = va->a_uio;
984 PFS_TRACE(("%s", pn->pn_name));
985 pfs_assert_not_owned(pn);
987 if (vn->v_type != VREG)
989 KASSERT_PN_IS_FILE(pn);
991 if (!(pn->pn_flags & PFS_WR))
994 if (pn->pn_fill == NULL)
998 * This is necessary because either process' privileges may
999 * have changed since the open() call.
1001 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
1008 if (pn->pn_flags & PFS_RAWWR) {
1009 error = pn_fill(curthread, proc, pn, NULL, uio);
1015 sbuf_uionew(&sb, uio, &error);
1022 error = pn_fill(curthread, proc, pn, &sb, uio);
1033 struct vop_vector pfs_vnodeops = {
1034 .vop_default = &default_vnodeops,
1036 .vop_access = pfs_access,
1037 .vop_cachedlookup = pfs_lookup,
1038 .vop_close = pfs_close,
1039 .vop_create = VOP_EOPNOTSUPP,
1040 .vop_getattr = pfs_getattr,
1041 .vop_getextattr = pfs_getextattr,
1042 .vop_ioctl = pfs_ioctl,
1043 .vop_link = VOP_EOPNOTSUPP,
1044 .vop_lookup = vfs_cache_lookup,
1045 .vop_mkdir = VOP_EOPNOTSUPP,
1046 .vop_mknod = VOP_EOPNOTSUPP,
1047 .vop_open = pfs_open,
1048 .vop_read = pfs_read,
1049 .vop_readdir = pfs_readdir,
1050 .vop_readlink = pfs_readlink,
1051 .vop_reclaim = pfs_reclaim,
1052 .vop_remove = VOP_EOPNOTSUPP,
1053 .vop_rename = VOP_EOPNOTSUPP,
1054 .vop_rmdir = VOP_EOPNOTSUPP,
1055 .vop_setattr = pfs_setattr,
1056 .vop_symlink = VOP_EOPNOTSUPP,
1057 .vop_vptocnp = pfs_vptocnp,
1058 .vop_write = pfs_write,
1059 /* XXX I've probably forgotten a few that need VOP_EOPNOTSUPP */