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);
347 * Convert a vnode to its component name
350 pfs_vptocnp(struct vop_vptocnp_args *ap)
352 struct vnode *vp = ap->a_vp;
353 struct vnode **dvp = ap->a_vpp;
354 struct pfs_vdata *pvd = vp->v_data;
355 struct pfs_node *pd = pvd->pvd_pn;
358 char *buf = ap->a_buf;
359 int *buflen = ap->a_buflen;
360 char pidbuf[PFS_NAMELEN];
361 pid_t pid = pvd->pvd_pid;
362 int len, i, error, locked;
369 if (vp->v_type == VDIR && pd->pn_type == pfstype_root) {
374 } else if (vp->v_type == VDIR && pd->pn_type == pfstype_procdir) {
375 len = snprintf(pidbuf, sizeof(pidbuf), "%d", pid);
381 bcopy(pidbuf, buf + i, len);
383 len = strlen(pd->pn_name);
389 bcopy(pd->pn_name, buf + i, len);
396 error = vfs_busy(mp, 0);
401 * vp is held by caller.
403 locked = VOP_ISLOCKED(vp);
406 error = pfs_vncache_alloc(mp, dvp, pn, pid);
408 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;
437 pid_t pid = pvd->pvd_pid;
439 int error, i, namelen, visible;
441 PFS_TRACE(("%.*s", (int)cnp->cn_namelen, cnp->cn_nameptr));
442 pfs_assert_not_owned(pd);
444 if (vn->v_type != VDIR)
445 PFS_RETURN (ENOTDIR);
446 KASSERT_PN_IS_DIR(pd);
448 error = VOP_ACCESS(vn, VEXEC, cnp->cn_cred, cnp->cn_thread);
453 * Don't support DELETE or RENAME. CREATE is supported so
454 * that O_CREAT will work, but the lookup will still fail if
455 * the file does not exist.
457 if ((cnp->cn_flags & ISLASTCN) &&
458 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
459 PFS_RETURN (EOPNOTSUPP);
461 /* shortcut: check if the name is too long */
462 if (cnp->cn_namelen >= PFS_NAMELEN)
465 /* check that parent directory is visible... */
466 if (!pfs_visible(curthread, pd, pvd->pvd_pid, NULL))
470 namelen = cnp->cn_namelen;
471 pname = cnp->cn_nameptr;
472 if (namelen == 1 && pname[0] == '.') {
480 if (cnp->cn_flags & ISDOTDOT) {
481 if (pd->pn_type == pfstype_root)
484 KASSERT(pd->pn_parent != NULL,
485 ("%s(): non-root directory has no parent", __func__));
487 * This one is tricky. Descendents of procdir nodes
488 * inherit their parent's process affinity, but
489 * there's no easy reverse mapping. For simplicity,
490 * we assume that if this node is a procdir, its
491 * parent isn't (which is correct as long as
492 * descendents of procdir nodes are never procdir
495 if (pd->pn_type == pfstype_procdir)
506 for (pn = pd->pn_nodes; pn != NULL; pn = pn->pn_next)
507 if (pn->pn_type == pfstype_procdir)
509 else if (pn->pn_name[namelen] == '\0' &&
510 bcmp(pname, pn->pn_name, namelen) == 0) {
515 /* process dependent node */
516 if ((pn = pdn) != NULL) {
518 for (pid = 0, i = 0; i < namelen && isdigit(pname[i]); ++i)
519 if ((pid = pid * 10 + pname[i] - '0') > PID_MAX)
521 if (i == cnp->cn_namelen) {
532 pfs_assert_not_owned(pd);
533 pfs_assert_not_owned(pn);
534 visible = pfs_visible(curthread, pn, pid, NULL);
540 error = pfs_vncache_alloc(vn->v_mount, vpp, pn, pid);
544 if (cnp->cn_flags & ISDOTDOT)
545 vn_lock(vn, LK_EXCLUSIVE|LK_RETRY);
546 if (cnp->cn_flags & MAKEENTRY)
547 cache_enter(vn, *vpp, cnp);
550 if (cnp->cn_flags & ISDOTDOT)
551 vn_lock(vn, LK_EXCLUSIVE|LK_RETRY);
556 * Open a file or directory.
559 pfs_open(struct vop_open_args *va)
561 struct vnode *vn = va->a_vp;
562 struct pfs_vdata *pvd = vn->v_data;
563 struct pfs_node *pn = pvd->pvd_pn;
564 int mode = va->a_mode;
566 PFS_TRACE(("%s (mode 0x%x)", pn->pn_name, mode));
567 pfs_assert_not_owned(pn);
569 /* check if the requested mode is permitted */
570 if (((mode & FREAD) && !(mode & PFS_RD)) ||
571 ((mode & FWRITE) && !(mode & PFS_WR)))
574 /* we don't support locking */
575 if ((mode & O_SHLOCK) || (mode & O_EXLOCK))
576 PFS_RETURN (EOPNOTSUPP);
585 pfs_read(struct vop_read_args *va)
587 struct vnode *vn = va->a_vp;
588 struct pfs_vdata *pvd = vn->v_data;
589 struct pfs_node *pn = pvd->pvd_pn;
590 struct uio *uio = va->a_uio;
592 struct sbuf *sb = NULL;
594 unsigned int buflen, offset, resid;
596 PFS_TRACE(("%s", pn->pn_name));
597 pfs_assert_not_owned(pn);
599 if (vn->v_type != VREG)
601 KASSERT_PN_IS_FILE(pn);
603 if (!(pn->pn_flags & PFS_RD))
606 if (pn->pn_fill == NULL)
610 * This is necessary because either process' privileges may
611 * have changed since the open() call.
613 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
621 locked = VOP_ISLOCKED(vn);
624 if (pn->pn_flags & PFS_RAWRD) {
625 PFS_TRACE(("%zd resid", uio->uio_resid));
626 error = pn_fill(curthread, proc, pn, NULL, uio);
627 PFS_TRACE(("%zd resid", uio->uio_resid));
631 /* beaucoup sanity checks so we don't ask for bogus allocation */
632 if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
633 (offset = uio->uio_offset) != uio->uio_offset ||
634 (resid = uio->uio_resid) != uio->uio_resid ||
635 (buflen = offset + resid + 1) < offset || buflen > INT_MAX) {
641 if (buflen > MAXPHYS + 1) {
646 sb = sbuf_new(sb, NULL, buflen, 0);
652 error = pn_fill(curthread, proc, pn, sb, uio);
660 error = uiomove_frombuf(sbuf_data(sb), sbuf_len(sb), uio);
663 vn_lock(vn, locked | LK_RETRY);
671 * Iterate through directory entries
674 pfs_iterate(struct thread *td, struct proc *proc, struct pfs_node *pd,
675 struct pfs_node **pn, struct proc **p)
679 sx_assert(&allproc_lock, SX_SLOCKED);
680 pfs_assert_owned(pd);
685 } else if ((*pn)->pn_type != pfstype_procdir) {
687 *pn = (*pn)->pn_next;
689 if (*pn != NULL && (*pn)->pn_type == pfstype_procdir) {
692 *p = LIST_FIRST(&allproc);
694 *p = LIST_NEXT(*p, p_list);
695 /* out of processes: next node */
697 *pn = (*pn)->pn_next;
706 visible = pfs_visible_proc(td, *pn, *p);
708 } else if (proc != NULL) {
709 visible = pfs_visible_proc(td, *pn, proc);
720 * Return directory entries.
723 pfs_readdir(struct vop_readdir_args *va)
725 struct vnode *vn = va->a_vp;
726 struct pfs_vdata *pvd = vn->v_data;
727 struct pfs_node *pd = pvd->pvd_pn;
728 pid_t pid = pvd->pvd_pid;
729 struct proc *p, *proc;
731 struct dirent *entry;
737 KASSERT(pd->pn_info == vn->v_mount->mnt_data,
738 ("%s(): pn_info does not match mountpoint", __func__));
739 PFS_TRACE(("%s pid %lu", pd->pn_name, (unsigned long)pid));
740 pfs_assert_not_owned(pd);
742 if (vn->v_type != VDIR)
743 PFS_RETURN (ENOTDIR);
744 KASSERT_PN_IS_DIR(pd);
747 /* only allow reading entire entries */
748 offset = uio->uio_offset;
749 resid = uio->uio_resid;
750 if (offset < 0 || offset % PFS_DELEN != 0 ||
751 (resid && resid < PFS_DELEN))
756 /* can't do this while holding the proc lock... */
757 buf = malloc(resid, M_IOV, M_WAITOK | M_ZERO);
758 sx_slock(&allproc_lock);
761 /* check if the directory is visible to the caller */
762 if (!pfs_visible(curthread, pd, pid, &proc)) {
763 sx_sunlock(&allproc_lock);
768 KASSERT(pid == NO_PID || proc != NULL,
769 ("%s(): no process for pid %lu", __func__, (unsigned long)pid));
771 /* skip unwanted entries */
772 for (pn = NULL, p = NULL; offset > 0; offset -= PFS_DELEN) {
773 if (pfs_iterate(curthread, proc, pd, &pn, &p) == -1) {
774 /* nothing left... */
778 sx_sunlock(&allproc_lock);
784 /* fill in entries */
786 while (pfs_iterate(curthread, proc, pd, &pn, &p) != -1 &&
787 resid >= PFS_DELEN) {
788 entry = (struct dirent *)ent;
789 entry->d_reclen = PFS_DELEN;
790 entry->d_fileno = pn_fileno(pn, pid);
791 /* PFS_DELEN was picked to fit PFS_NAMLEN */
792 for (i = 0; i < PFS_NAMELEN - 1 && pn->pn_name[i] != '\0'; ++i)
793 entry->d_name[i] = pn->pn_name[i];
794 entry->d_name[i] = 0;
796 switch (pn->pn_type) {
797 case pfstype_procdir:
799 ("reached procdir node with p == NULL"));
800 entry->d_namlen = snprintf(entry->d_name,
801 PFS_NAMELEN, "%d", p->p_pid);
807 entry->d_type = DT_DIR;
810 entry->d_type = DT_REG;
812 case pfstype_symlink:
813 entry->d_type = DT_LNK;
816 panic("%s has unexpected node type: %d", pn->pn_name, pn->pn_type);
818 PFS_TRACE(("%s", entry->d_name));
826 sx_sunlock(&allproc_lock);
827 PFS_TRACE(("%zd bytes", ent - buf));
828 error = uiomove(buf, ent - buf, uio);
834 * Read a symbolic link
837 pfs_readlink(struct vop_readlink_args *va)
839 struct vnode *vn = va->a_vp;
840 struct pfs_vdata *pvd = vn->v_data;
841 struct pfs_node *pn = pvd->pvd_pn;
842 struct uio *uio = va->a_uio;
843 struct proc *proc = NULL;
848 PFS_TRACE(("%s", pn->pn_name));
849 pfs_assert_not_owned(pn);
851 if (vn->v_type != VLNK)
853 KASSERT_PN_IS_LINK(pn);
855 if (pn->pn_fill == NULL)
858 if (pvd->pvd_pid != NO_PID) {
859 if ((proc = pfind(pvd->pvd_pid)) == NULL)
861 if (proc->p_flag & P_WEXIT) {
869 locked = VOP_ISLOCKED(vn);
872 /* sbuf_new() can't fail with a static buffer */
873 sbuf_new(&sb, buf, sizeof buf, 0);
875 error = pn_fill(curthread, proc, pn, &sb, NULL);
879 vn_lock(vn, locked | LK_RETRY);
888 error = uiomove_frombuf(sbuf_data(&sb), sbuf_len(&sb), uio);
897 pfs_reclaim(struct vop_reclaim_args *va)
899 struct vnode *vn = va->a_vp;
900 struct pfs_vdata *pvd = vn->v_data;
901 struct pfs_node *pn = pvd->pvd_pn;
903 PFS_TRACE(("%s", pn->pn_name));
904 pfs_assert_not_owned(pn);
906 return (pfs_vncache_free(va->a_vp));
913 pfs_setattr(struct vop_setattr_args *va)
915 struct vnode *vn = va->a_vp;
916 struct pfs_vdata *pvd = vn->v_data;
917 struct pfs_node *pn = pvd->pvd_pn;
919 PFS_TRACE(("%s", pn->pn_name));
920 pfs_assert_not_owned(pn);
922 PFS_RETURN (EOPNOTSUPP);
929 pfs_write(struct vop_write_args *va)
931 struct vnode *vn = va->a_vp;
932 struct pfs_vdata *pvd = vn->v_data;
933 struct pfs_node *pn = pvd->pvd_pn;
934 struct uio *uio = va->a_uio;
939 PFS_TRACE(("%s", pn->pn_name));
940 pfs_assert_not_owned(pn);
942 if (vn->v_type != VREG)
944 KASSERT_PN_IS_FILE(pn);
946 if (!(pn->pn_flags & PFS_WR))
949 if (pn->pn_fill == NULL)
953 * This is necessary because either process' privileges may
954 * have changed since the open() call.
956 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
963 if (pn->pn_flags & PFS_RAWWR) {
964 error = pn_fill(curthread, proc, pn, NULL, uio);
970 sbuf_uionew(&sb, uio, &error);
977 error = pn_fill(curthread, proc, pn, &sb, uio);
988 struct vop_vector pfs_vnodeops = {
989 .vop_default = &default_vnodeops,
991 .vop_access = pfs_access,
992 .vop_cachedlookup = pfs_lookup,
993 .vop_close = pfs_close,
994 .vop_create = VOP_EOPNOTSUPP,
995 .vop_getattr = pfs_getattr,
996 .vop_getextattr = pfs_getextattr,
997 .vop_ioctl = pfs_ioctl,
998 .vop_link = VOP_EOPNOTSUPP,
999 .vop_lookup = vfs_cache_lookup,
1000 .vop_mkdir = VOP_EOPNOTSUPP,
1001 .vop_mknod = VOP_EOPNOTSUPP,
1002 .vop_open = pfs_open,
1003 .vop_read = pfs_read,
1004 .vop_readdir = pfs_readdir,
1005 .vop_readlink = pfs_readlink,
1006 .vop_reclaim = pfs_reclaim,
1007 .vop_remove = VOP_EOPNOTSUPP,
1008 .vop_rename = VOP_EOPNOTSUPP,
1009 .vop_rmdir = VOP_EOPNOTSUPP,
1010 .vop_setattr = pfs_setattr,
1011 .vop_symlink = VOP_EOPNOTSUPP,
1012 .vop_vptocnp = pfs_vptocnp,
1013 .vop_write = pfs_write,
1014 /* XXX I've probably forgotten a few that need VOP_EOPNOTSUPP */