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,
108 bool allproc_locked, struct proc **p)
112 PFS_TRACE(("%s (pid: %d, req: %d)",
113 pn->pn_name, pid, td->td_proc->p_pid));
119 proc = allproc_locked ? pfind_locked(pid) : pfind(pid);
122 if (pfs_visible_proc(td, pn, proc)) {
137 pfs_access(struct vop_access_args *va)
139 struct vnode *vn = va->a_vp;
140 struct pfs_vdata *pvd = vn->v_data;
144 PFS_TRACE(("%s", pvd->pvd_pn->pn_name));
147 error = VOP_GETATTR(vn, &vattr, va->a_cred);
150 error = vaccess(vn->v_type, vattr.va_mode, vattr.va_uid,
151 vattr.va_gid, va->a_accmode, va->a_cred, NULL);
156 * Close a file or directory
159 pfs_close(struct vop_close_args *va)
161 struct vnode *vn = va->a_vp;
162 struct pfs_vdata *pvd = vn->v_data;
163 struct pfs_node *pn = pvd->pvd_pn;
167 PFS_TRACE(("%s", pn->pn_name));
168 pfs_assert_not_owned(pn);
171 * Do nothing unless this is the last close and the node has a
172 * last-close handler.
174 if (vrefcnt(vn) > 1 || pn->pn_close == NULL)
177 if (pvd->pvd_pid != NO_PID) {
178 proc = pfind(pvd->pvd_pid);
183 error = pn_close(va->a_td, proc, pn);
192 * Get file attributes
195 pfs_getattr(struct vop_getattr_args *va)
197 struct vnode *vn = va->a_vp;
198 struct pfs_vdata *pvd = vn->v_data;
199 struct pfs_node *pn = pvd->pvd_pn;
200 struct vattr *vap = va->a_vap;
204 PFS_TRACE(("%s", pn->pn_name));
205 pfs_assert_not_owned(pn);
207 if (!pfs_visible(curthread, pn, pvd->pvd_pid, false, &proc))
210 vap->va_type = vn->v_type;
211 vap->va_fileid = pn_fileno(pn, pvd->pvd_pid);
213 vap->va_blocksize = PAGE_SIZE;
214 vap->va_bytes = vap->va_size = 0;
216 vap->va_fsid = vn->v_mount->mnt_stat.f_fsid.val[0];
218 nanotime(&vap->va_ctime);
219 vap->va_atime = vap->va_mtime = vap->va_ctime;
221 switch (pn->pn_type) {
222 case pfstype_procdir:
227 /* compute link count */
233 case pfstype_symlink:
237 printf("shouldn't be here!\n");
243 vap->va_uid = proc->p_ucred->cr_ruid;
244 vap->va_gid = proc->p_ucred->cr_rgid;
250 if (pn->pn_attr != NULL)
251 error = pn_attr(curthread, proc, pn, vap);
263 pfs_ioctl(struct vop_ioctl_args *va)
266 struct pfs_vdata *pvd;
272 vn_lock(vn, LK_SHARED | LK_RETRY);
273 if (vn->v_iflag & VI_DOOMED) {
280 PFS_TRACE(("%s: %lx", pn->pn_name, va->a_command));
281 pfs_assert_not_owned(pn);
283 if (vn->v_type != VREG) {
287 KASSERT_PN_IS_FILE(pn);
289 if (pn->pn_ioctl == NULL) {
295 * This is necessary because process' privileges may
296 * have changed since the open() call.
298 if (!pfs_visible(curthread, pn, pvd->pvd_pid, false, &proc)) {
303 error = pn_ioctl(curthread, proc, pn, va->a_command, va->a_data);
316 pfs_getextattr(struct vop_getextattr_args *va)
318 struct vnode *vn = va->a_vp;
319 struct pfs_vdata *pvd = vn->v_data;
320 struct pfs_node *pn = pvd->pvd_pn;
324 PFS_TRACE(("%s", pn->pn_name));
325 pfs_assert_not_owned(pn);
328 * This is necessary because either process' privileges may
329 * have changed since the open() call.
331 if (!pfs_visible(curthread, pn, pvd->pvd_pid, false, &proc))
334 if (pn->pn_getextattr == NULL)
337 error = pn_getextattr(curthread, proc, pn,
338 va->a_attrnamespace, va->a_name, va->a_uio,
339 va->a_size, va->a_cred);
348 * Convert a vnode to its component name
351 pfs_vptocnp(struct vop_vptocnp_args *ap)
353 struct vnode *vp = ap->a_vp;
354 struct vnode **dvp = ap->a_vpp;
355 struct pfs_vdata *pvd = vp->v_data;
356 struct pfs_node *pd = pvd->pvd_pn;
359 char *buf = ap->a_buf;
360 int *buflen = ap->a_buflen;
361 char pidbuf[PFS_NAMELEN];
362 pid_t pid = pvd->pvd_pid;
363 int len, i, error, locked;
370 if (vp->v_type == VDIR && pd->pn_type == pfstype_root) {
375 } else if (vp->v_type == VDIR && pd->pn_type == pfstype_procdir) {
376 len = snprintf(pidbuf, sizeof(pidbuf), "%d", pid);
382 bcopy(pidbuf, buf + i, len);
384 len = strlen(pd->pn_name);
390 bcopy(pd->pn_name, buf + i, len);
397 error = vfs_busy(mp, 0);
402 * vp is held by caller.
404 locked = VOP_ISLOCKED(vp);
407 error = pfs_vncache_alloc(mp, dvp, pn, pid);
409 vn_lock(vp, locked | LK_RETRY);
416 vn_lock(vp, locked | LK_RETRY);
426 * Look up a file or directory
429 pfs_lookup(struct vop_cachedlookup_args *va)
431 struct vnode *vn = va->a_dvp;
432 struct vnode **vpp = va->a_vpp;
433 struct componentname *cnp = va->a_cnp;
434 struct pfs_vdata *pvd = vn->v_data;
435 struct pfs_node *pd = pvd->pvd_pn;
436 struct pfs_node *pn, *pdn = NULL;
438 pid_t pid = pvd->pvd_pid;
440 int error, i, namelen, visible;
442 PFS_TRACE(("%.*s", (int)cnp->cn_namelen, cnp->cn_nameptr));
443 pfs_assert_not_owned(pd);
445 if (vn->v_type != VDIR)
446 PFS_RETURN (ENOTDIR);
447 KASSERT_PN_IS_DIR(pd);
449 error = VOP_ACCESS(vn, VEXEC, cnp->cn_cred, cnp->cn_thread);
454 * Don't support DELETE or RENAME. CREATE is supported so
455 * that O_CREAT will work, but the lookup will still fail if
456 * the file does not exist.
458 if ((cnp->cn_flags & ISLASTCN) &&
459 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
460 PFS_RETURN (EOPNOTSUPP);
462 /* shortcut: check if the name is too long */
463 if (cnp->cn_namelen >= PFS_NAMELEN)
466 /* check that parent directory is visible... */
467 if (!pfs_visible(curthread, pd, pvd->pvd_pid, false, NULL))
471 namelen = cnp->cn_namelen;
472 pname = cnp->cn_nameptr;
473 if (namelen == 1 && pname[0] == '.') {
483 if (cnp->cn_flags & ISDOTDOT) {
484 if (pd->pn_type == pfstype_root)
486 error = vfs_busy(mp, MBF_NOWAIT);
490 error = vfs_busy(mp, 0);
491 vn_lock(vn, LK_EXCLUSIVE | LK_RETRY);
495 if (vn->v_iflag & VI_DOOMED) {
501 KASSERT(pd->pn_parent != NULL,
502 ("%s(): non-root directory has no parent", __func__));
504 * This one is tricky. Descendents of procdir nodes
505 * inherit their parent's process affinity, but
506 * there's no easy reverse mapping. For simplicity,
507 * we assume that if this node is a procdir, its
508 * parent isn't (which is correct as long as
509 * descendents of procdir nodes are never procdir
512 if (pd->pn_type == pfstype_procdir)
523 for (pn = pd->pn_nodes; pn != NULL; pn = pn->pn_next)
524 if (pn->pn_type == pfstype_procdir)
526 else if (pn->pn_name[namelen] == '\0' &&
527 bcmp(pname, pn->pn_name, namelen) == 0) {
532 /* process dependent node */
533 if ((pn = pdn) != NULL) {
535 for (pid = 0, i = 0; i < namelen && isdigit(pname[i]); ++i)
536 if ((pid = pid * 10 + pname[i] - '0') > PID_MAX)
538 if (i == cnp->cn_namelen) {
549 pfs_assert_not_owned(pd);
550 pfs_assert_not_owned(pn);
551 visible = pfs_visible(curthread, pn, pid, false, NULL);
557 error = pfs_vncache_alloc(mp, vpp, pn, pid);
561 if (cnp->cn_flags & ISDOTDOT) {
563 vn_lock(vn, LK_EXCLUSIVE | LK_RETRY);
564 if (vn->v_iflag & VI_DOOMED) {
570 if (cnp->cn_flags & MAKEENTRY && !(vn->v_iflag & VI_DOOMED))
571 cache_enter(vn, *vpp, cnp);
574 if (cnp->cn_flags & ISDOTDOT) {
576 vn_lock(vn, LK_EXCLUSIVE | LK_RETRY);
583 * Open a file or directory.
586 pfs_open(struct vop_open_args *va)
588 struct vnode *vn = va->a_vp;
589 struct pfs_vdata *pvd = vn->v_data;
590 struct pfs_node *pn = pvd->pvd_pn;
591 int mode = va->a_mode;
593 PFS_TRACE(("%s (mode 0x%x)", pn->pn_name, mode));
594 pfs_assert_not_owned(pn);
596 /* check if the requested mode is permitted */
597 if (((mode & FREAD) && !(mode & PFS_RD)) ||
598 ((mode & FWRITE) && !(mode & PFS_WR)))
601 /* we don't support locking */
602 if ((mode & O_SHLOCK) || (mode & O_EXLOCK))
603 PFS_RETURN (EOPNOTSUPP);
612 pfs_read(struct vop_read_args *va)
614 struct vnode *vn = va->a_vp;
615 struct pfs_vdata *pvd = vn->v_data;
616 struct pfs_node *pn = pvd->pvd_pn;
617 struct uio *uio = va->a_uio;
619 struct sbuf *sb = NULL;
623 PFS_TRACE(("%s", pn->pn_name));
624 pfs_assert_not_owned(pn);
626 if (vn->v_type != VREG)
628 KASSERT_PN_IS_FILE(pn);
630 if (!(pn->pn_flags & PFS_RD))
633 if (pn->pn_fill == NULL)
637 * This is necessary because either process' privileges may
638 * have changed since the open() call.
640 if (!pfs_visible(curthread, pn, pvd->pvd_pid, false, &proc))
648 locked = VOP_ISLOCKED(vn);
651 if (pn->pn_flags & PFS_RAWRD) {
652 PFS_TRACE(("%zd resid", uio->uio_resid));
653 error = pn_fill(curthread, proc, pn, NULL, uio);
654 PFS_TRACE(("%zd resid", uio->uio_resid));
658 if (uio->uio_resid < 0 || uio->uio_offset < 0 ||
659 uio->uio_resid > OFF_MAX - uio->uio_offset) {
663 buflen = uio->uio_offset + uio->uio_resid;
664 if (buflen > MAXPHYS)
667 sb = sbuf_new(sb, NULL, buflen + 1, 0);
673 error = pn_fill(curthread, proc, pn, sb, uio);
681 * XXX: If the buffer overflowed, sbuf_len() will not return
682 * the data length. Then just use the full length because an
683 * overflowed sbuf must be full.
685 if (sbuf_finish(sb) == 0)
686 buflen = sbuf_len(sb);
687 error = uiomove_frombuf(sbuf_data(sb), buflen, uio);
690 vn_lock(vn, locked | LK_RETRY);
698 * Iterate through directory entries
701 pfs_iterate(struct thread *td, struct proc *proc, struct pfs_node *pd,
702 struct pfs_node **pn, struct proc **p)
706 sx_assert(&allproc_lock, SX_SLOCKED);
707 pfs_assert_owned(pd);
712 } else if ((*pn)->pn_type != pfstype_procdir) {
714 *pn = (*pn)->pn_next;
716 if (*pn != NULL && (*pn)->pn_type == pfstype_procdir) {
719 *p = LIST_FIRST(&allproc);
721 *p = LIST_NEXT(*p, p_list);
722 /* out of processes: next node */
724 *pn = (*pn)->pn_next;
733 visible = pfs_visible_proc(td, *pn, *p);
735 } else if (proc != NULL) {
736 visible = pfs_visible_proc(td, *pn, proc);
746 /* Directory entry list */
748 STAILQ_ENTRY(pfsentry) link;
751 STAILQ_HEAD(pfsdirentlist, pfsentry);
754 * Return directory entries.
757 pfs_readdir(struct vop_readdir_args *va)
759 struct vnode *vn = va->a_vp;
760 struct pfs_vdata *pvd = vn->v_data;
761 struct pfs_node *pd = pvd->pvd_pn;
762 pid_t pid = pvd->pvd_pid;
763 struct proc *p, *proc;
766 struct pfsentry *pfsent, *pfsent2;
767 struct pfsdirentlist lst;
773 KASSERT(pd->pn_info == vn->v_mount->mnt_data,
774 ("%s(): pn_info does not match mountpoint", __func__));
775 PFS_TRACE(("%s pid %lu", pd->pn_name, (unsigned long)pid));
776 pfs_assert_not_owned(pd);
778 if (vn->v_type != VDIR)
779 PFS_RETURN (ENOTDIR);
780 KASSERT_PN_IS_DIR(pd);
783 /* only allow reading entire entries */
784 offset = uio->uio_offset;
785 resid = uio->uio_resid;
786 if (offset < 0 || offset % PFS_DELEN != 0 ||
787 (resid && resid < PFS_DELEN))
792 sx_slock(&allproc_lock);
795 /* check if the directory is visible to the caller */
796 if (!pfs_visible(curthread, pd, pid, true, &proc)) {
797 sx_sunlock(&allproc_lock);
801 KASSERT(pid == NO_PID || proc != NULL,
802 ("%s(): no process for pid %lu", __func__, (unsigned long)pid));
804 /* skip unwanted entries */
805 for (pn = NULL, p = NULL; offset > 0; offset -= PFS_DELEN) {
806 if (pfs_iterate(curthread, proc, pd, &pn, &p) == -1) {
807 /* nothing left... */
811 sx_sunlock(&allproc_lock);
816 /* fill in entries */
817 while (pfs_iterate(curthread, proc, pd, &pn, &p) != -1 &&
818 resid >= PFS_DELEN) {
819 if ((pfsent = malloc(sizeof(struct pfsentry), M_IOV,
820 M_NOWAIT | M_ZERO)) == NULL) {
824 pfsent->entry.d_reclen = PFS_DELEN;
825 pfsent->entry.d_fileno = pn_fileno(pn, pid);
826 /* PFS_DELEN was picked to fit PFS_NAMLEN */
827 for (i = 0; i < PFS_NAMELEN - 1 && pn->pn_name[i] != '\0'; ++i)
828 pfsent->entry.d_name[i] = pn->pn_name[i];
829 pfsent->entry.d_name[i] = 0;
830 pfsent->entry.d_namlen = i;
831 switch (pn->pn_type) {
832 case pfstype_procdir:
834 ("reached procdir node with p == NULL"));
835 pfsent->entry.d_namlen = snprintf(pfsent->entry.d_name,
836 PFS_NAMELEN, "%d", p->p_pid);
842 pfsent->entry.d_type = DT_DIR;
845 pfsent->entry.d_type = DT_REG;
847 case pfstype_symlink:
848 pfsent->entry.d_type = DT_LNK;
851 panic("%s has unexpected node type: %d", pn->pn_name, pn->pn_type);
853 PFS_TRACE(("%s", pfsent->entry.d_name));
854 STAILQ_INSERT_TAIL(&lst, pfsent, link);
861 sx_sunlock(&allproc_lock);
863 STAILQ_FOREACH_SAFE(pfsent, &lst, link, pfsent2) {
865 error = uiomove(&pfsent->entry, PFS_DELEN, uio);
869 PFS_TRACE(("%d bytes", i * PFS_DELEN));
874 * Read a symbolic link
877 pfs_readlink(struct vop_readlink_args *va)
879 struct vnode *vn = va->a_vp;
880 struct pfs_vdata *pvd = vn->v_data;
881 struct pfs_node *pn = pvd->pvd_pn;
882 struct uio *uio = va->a_uio;
883 struct proc *proc = NULL;
888 PFS_TRACE(("%s", pn->pn_name));
889 pfs_assert_not_owned(pn);
891 if (vn->v_type != VLNK)
893 KASSERT_PN_IS_LINK(pn);
895 if (pn->pn_fill == NULL)
898 if (pvd->pvd_pid != NO_PID) {
899 if ((proc = pfind(pvd->pvd_pid)) == NULL)
901 if (proc->p_flag & P_WEXIT) {
909 locked = VOP_ISLOCKED(vn);
912 /* sbuf_new() can't fail with a static buffer */
913 sbuf_new(&sb, buf, sizeof buf, 0);
915 error = pn_fill(curthread, proc, pn, &sb, NULL);
919 vn_lock(vn, locked | LK_RETRY);
927 if (sbuf_finish(&sb) != 0) {
929 PFS_RETURN (ENAMETOOLONG);
932 error = uiomove_frombuf(sbuf_data(&sb), sbuf_len(&sb), uio);
941 pfs_reclaim(struct vop_reclaim_args *va)
943 struct vnode *vn = va->a_vp;
944 struct pfs_vdata *pvd = vn->v_data;
945 struct pfs_node *pn = pvd->pvd_pn;
947 PFS_TRACE(("%s", pn->pn_name));
948 pfs_assert_not_owned(pn);
950 return (pfs_vncache_free(va->a_vp));
957 pfs_setattr(struct vop_setattr_args *va)
959 struct vnode *vn = va->a_vp;
960 struct pfs_vdata *pvd = vn->v_data;
961 struct pfs_node *pn = pvd->pvd_pn;
963 PFS_TRACE(("%s", pn->pn_name));
964 pfs_assert_not_owned(pn);
966 PFS_RETURN (EOPNOTSUPP);
973 pfs_write(struct vop_write_args *va)
975 struct vnode *vn = va->a_vp;
976 struct pfs_vdata *pvd = vn->v_data;
977 struct pfs_node *pn = pvd->pvd_pn;
978 struct uio *uio = va->a_uio;
983 PFS_TRACE(("%s", pn->pn_name));
984 pfs_assert_not_owned(pn);
986 if (vn->v_type != VREG)
988 KASSERT_PN_IS_FILE(pn);
990 if (!(pn->pn_flags & PFS_WR))
993 if (pn->pn_fill == NULL)
997 * This is necessary because either process' privileges may
998 * have changed since the open() call.
1000 if (!pfs_visible(curthread, pn, pvd->pvd_pid, false, &proc))
1007 if (pn->pn_flags & PFS_RAWWR) {
1008 error = pn_fill(curthread, proc, pn, NULL, uio);
1014 sbuf_uionew(&sb, uio, &error);
1021 error = pn_fill(curthread, proc, pn, &sb, uio);
1032 struct vop_vector pfs_vnodeops = {
1033 .vop_default = &default_vnodeops,
1035 .vop_access = pfs_access,
1036 .vop_cachedlookup = pfs_lookup,
1037 .vop_close = pfs_close,
1038 .vop_create = VOP_EOPNOTSUPP,
1039 .vop_getattr = pfs_getattr,
1040 .vop_getextattr = pfs_getextattr,
1041 .vop_ioctl = pfs_ioctl,
1042 .vop_link = VOP_EOPNOTSUPP,
1043 .vop_lookup = vfs_cache_lookup,
1044 .vop_mkdir = VOP_EOPNOTSUPP,
1045 .vop_mknod = VOP_EOPNOTSUPP,
1046 .vop_open = pfs_open,
1047 .vop_read = pfs_read,
1048 .vop_readdir = pfs_readdir,
1049 .vop_readlink = pfs_readlink,
1050 .vop_reclaim = pfs_reclaim,
1051 .vop_remove = VOP_EOPNOTSUPP,
1052 .vop_rename = VOP_EOPNOTSUPP,
1053 .vop_rmdir = VOP_EOPNOTSUPP,
1054 .vop_setattr = pfs_setattr,
1055 .vop_symlink = VOP_EOPNOTSUPP,
1056 .vop_vptocnp = pfs_vptocnp,
1057 .vop_write = pfs_write,
1058 /* XXX I've probably forgotten a few that need VOP_EOPNOTSUPP */