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>
56 * Returns the fileno, adjusted for target pid
59 pn_fileno(struct pfs_node *pn, pid_t pid)
62 KASSERT(pn->pn_fileno > 0,
63 ("%s(): no fileno allocated", __func__));
65 return (pn->pn_fileno * NO_PID + pid);
66 return (pn->pn_fileno);
70 * Returns non-zero if given file is visible to given thread.
73 pfs_visible_proc(struct thread *td, struct pfs_node *pn, struct proc *proc)
80 PROC_LOCK_ASSERT(proc, MA_OWNED);
82 visible = ((proc->p_flag & P_WEXIT) == 0);
84 visible = (p_cansee(td, proc) == 0);
85 if (visible && pn->pn_vis != NULL)
86 visible = pn_vis(td, proc, pn);
93 pfs_visible(struct thread *td, struct pfs_node *pn, pid_t pid, struct proc **p)
97 PFS_TRACE(("%s (pid: %d, req: %d)",
98 pn->pn_name, pid, td->td_proc->p_pid));
104 if ((proc = pfind(pid)) == NULL)
106 if (pfs_visible_proc(td, pn, proc)) {
121 pfs_access(struct vop_access_args *va)
123 struct vnode *vn = va->a_vp;
124 struct pfs_vdata *pvd = vn->v_data;
128 PFS_TRACE(("%s", pvd->pvd_pn->pn_name));
131 error = VOP_GETATTR(vn, &vattr, va->a_cred, va->a_td);
134 error = vaccess(vn->v_type, vattr.va_mode, vattr.va_uid,
135 vattr.va_gid, va->a_mode, va->a_cred, NULL);
140 * Close a file or directory
143 pfs_close(struct vop_close_args *va)
145 struct vnode *vn = va->a_vp;
146 struct pfs_vdata *pvd = vn->v_data;
147 struct pfs_node *pn = pvd->pvd_pn;
151 PFS_TRACE(("%s", pn->pn_name));
152 pfs_assert_not_owned(pn);
155 * Do nothing unless this is the last close and the node has a
156 * last-close handler.
158 if (vrefcnt(vn) > 1 || pn->pn_close == NULL)
161 if (pvd->pvd_pid != NO_PID) {
162 proc = pfind(pvd->pvd_pid);
167 error = pn_close(va->a_td, proc, pn);
176 * Get file attributes
179 pfs_getattr(struct vop_getattr_args *va)
181 struct vnode *vn = va->a_vp;
182 struct pfs_vdata *pvd = vn->v_data;
183 struct pfs_node *pn = pvd->pvd_pn;
184 struct vattr *vap = va->a_vap;
188 PFS_TRACE(("%s", pn->pn_name));
189 pfs_assert_not_owned(pn);
191 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
194 vap->va_type = vn->v_type;
195 vap->va_fileid = pn_fileno(pn, pvd->pvd_pid);
197 vap->va_blocksize = PAGE_SIZE;
198 vap->va_bytes = vap->va_size = 0;
200 vap->va_fsid = vn->v_mount->mnt_stat.f_fsid.val[0];
202 nanotime(&vap->va_ctime);
203 vap->va_atime = vap->va_mtime = vap->va_ctime;
205 switch (pn->pn_type) {
206 case pfstype_procdir:
211 /* compute link count */
217 case pfstype_symlink:
221 printf("shouldn't be here!\n");
227 vap->va_uid = proc->p_ucred->cr_ruid;
228 vap->va_gid = proc->p_ucred->cr_rgid;
229 if (pn->pn_attr != NULL)
230 error = pn_attr(va->a_td, proc, pn, vap);
244 pfs_ioctl(struct vop_ioctl_args *va)
246 struct vnode *vn = va->a_vp;
247 struct pfs_vdata *pvd = vn->v_data;
248 struct pfs_node *pn = pvd->pvd_pn;
252 PFS_TRACE(("%s: %lx", pn->pn_name, va->a_command));
253 pfs_assert_not_owned(pn);
255 if (vn->v_type != VREG)
258 if (pn->pn_ioctl == NULL)
262 * This is necessary because process' privileges may
263 * have changed since the open() call.
265 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
268 error = pn_ioctl(curthread, proc, pn, va->a_command, va->a_data);
280 pfs_getextattr(struct vop_getextattr_args *va)
282 struct vnode *vn = va->a_vp;
283 struct pfs_vdata *pvd = vn->v_data;
284 struct pfs_node *pn = pvd->pvd_pn;
288 PFS_TRACE(("%s", pn->pn_name));
289 pfs_assert_not_owned(pn);
292 * This is necessary because either process' privileges may
293 * have changed since the open() call.
295 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
298 if (pn->pn_getextattr == NULL)
301 error = pn_getextattr(curthread, proc, pn,
302 va->a_attrnamespace, va->a_name, va->a_uio,
303 va->a_size, va->a_cred);
313 * Look up a file or directory
316 pfs_lookup(struct vop_cachedlookup_args *va)
318 struct vnode *vn = va->a_dvp;
319 struct vnode **vpp = va->a_vpp;
320 struct componentname *cnp = va->a_cnp;
321 struct pfs_vdata *pvd = vn->v_data;
322 struct pfs_node *pd = pvd->pvd_pn;
323 struct pfs_node *pn, *pdn = NULL;
324 pid_t pid = pvd->pvd_pid;
326 int error, i, namelen, visible;
328 PFS_TRACE(("%.*s", (int)cnp->cn_namelen, cnp->cn_nameptr));
329 pfs_assert_not_owned(pd);
331 if (vn->v_type != VDIR)
332 PFS_RETURN (ENOTDIR);
334 error = VOP_ACCESS(vn, VEXEC, cnp->cn_cred, cnp->cn_thread);
339 * Don't support DELETE or RENAME. CREATE is supported so
340 * that O_CREAT will work, but the lookup will still fail if
341 * the file does not exist.
343 if ((cnp->cn_flags & ISLASTCN) &&
344 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
345 PFS_RETURN (EOPNOTSUPP);
347 /* shortcut: check if the name is too long */
348 if (cnp->cn_namelen >= PFS_NAMELEN)
351 /* check that parent directory is visible... */
352 if (!pfs_visible(curthread, pd, pvd->pvd_pid, NULL))
356 namelen = cnp->cn_namelen;
357 pname = cnp->cn_nameptr;
358 if (namelen == 1 && pname[0] == '.') {
366 if (cnp->cn_flags & ISDOTDOT) {
367 if (pd->pn_type == pfstype_root)
369 VOP_UNLOCK(vn, 0, cnp->cn_thread);
370 KASSERT(pd->pn_parent != NULL,
371 ("%s(): non-root directory has no parent", __func__));
373 * This one is tricky. Descendents of procdir nodes
374 * inherit their parent's process affinity, but
375 * there's no easy reverse mapping. For simplicity,
376 * we assume that if this node is a procdir, its
377 * parent isn't (which is correct as long as
378 * descendents of procdir nodes are never procdir
381 if (pd->pn_type == pfstype_procdir)
392 for (pn = pd->pn_nodes; pn != NULL; pn = pn->pn_next)
393 if (pn->pn_type == pfstype_procdir)
395 else if (pn->pn_name[namelen] == '\0' &&
396 bcmp(pname, pn->pn_name, namelen) == 0) {
401 /* process dependent node */
402 if ((pn = pdn) != NULL) {
404 for (pid = 0, i = 0; i < namelen && isdigit(pname[i]); ++i)
405 if ((pid = pid * 10 + pname[i] - '0') > PID_MAX)
407 if (i == cnp->cn_namelen) {
418 pfs_assert_not_owned(pd);
419 pfs_assert_not_owned(pn);
420 visible = pfs_visible(curthread, pn, pid, NULL);
426 error = pfs_vncache_alloc(vn->v_mount, vpp, pn, pid);
430 if (cnp->cn_flags & ISDOTDOT)
431 vn_lock(vn, LK_EXCLUSIVE|LK_RETRY, cnp->cn_thread);
432 if (cnp->cn_flags & MAKEENTRY)
433 cache_enter(vn, *vpp, cnp);
436 if (cnp->cn_flags & ISDOTDOT)
437 vn_lock(vn, LK_EXCLUSIVE|LK_RETRY, cnp->cn_thread);
442 * Open a file or directory.
445 pfs_open(struct vop_open_args *va)
447 struct vnode *vn = va->a_vp;
448 struct pfs_vdata *pvd = vn->v_data;
449 struct pfs_node *pn = pvd->pvd_pn;
450 int mode = va->a_mode;
452 PFS_TRACE(("%s (mode 0x%x)", pn->pn_name, mode));
453 pfs_assert_not_owned(pn);
455 /* check if the requested mode is permitted */
456 if (((mode & FREAD) && !(mode & PFS_RD)) ||
457 ((mode & FWRITE) && !(mode & PFS_WR)))
460 /* we don't support locking */
461 if ((mode & O_SHLOCK) || (mode & O_EXLOCK))
462 PFS_RETURN (EOPNOTSUPP);
471 pfs_read(struct vop_read_args *va)
473 struct vnode *vn = va->a_vp;
474 struct pfs_vdata *pvd = vn->v_data;
475 struct pfs_node *pn = pvd->pvd_pn;
476 struct uio *uio = va->a_uio;
478 struct sbuf *sb = NULL;
480 unsigned int buflen, offset, resid;
482 PFS_TRACE(("%s", pn->pn_name));
483 pfs_assert_not_owned(pn);
485 if (vn->v_type != VREG)
488 if (!(pn->pn_flags & PFS_RD))
491 if (pn->pn_fill == NULL)
495 * This is necessary because either process' privileges may
496 * have changed since the open() call.
498 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
506 locked = VOP_ISLOCKED(vn, curthread);
507 VOP_UNLOCK(vn, 0, curthread);
509 if (pn->pn_flags & PFS_RAWRD) {
510 PFS_TRACE(("%lu resid", (unsigned long)uio->uio_resid));
511 error = pn_fill(curthread, proc, pn, NULL, uio);
512 PFS_TRACE(("%lu resid", (unsigned long)uio->uio_resid));
516 /* beaucoup sanity checks so we don't ask for bogus allocation */
517 if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
518 (offset = uio->uio_offset) != uio->uio_offset ||
519 (resid = uio->uio_resid) != uio->uio_resid ||
520 (buflen = offset + resid + 1) < offset || buflen > INT_MAX) {
526 if (buflen > MAXPHYS + 1) {
531 sb = sbuf_new(sb, NULL, buflen, 0);
537 error = pn_fill(curthread, proc, pn, sb, uio);
545 error = uiomove_frombuf(sbuf_data(sb), sbuf_len(sb), uio);
548 vn_lock(vn, locked | LK_RETRY, curthread);
556 * Iterate through directory entries
559 pfs_iterate(struct thread *td, struct proc *proc, struct pfs_node *pd,
560 struct pfs_node **pn, struct proc **p)
564 sx_assert(&allproc_lock, SX_SLOCKED);
565 pfs_assert_owned(pd);
570 } else if ((*pn)->pn_type != pfstype_procdir) {
572 *pn = (*pn)->pn_next;
574 if (*pn != NULL && (*pn)->pn_type == pfstype_procdir) {
577 *p = LIST_FIRST(&allproc);
579 *p = LIST_NEXT(*p, p_list);
580 /* out of processes: next node */
582 *pn = (*pn)->pn_next;
591 visible = pfs_visible_proc(td, *pn, *p);
593 } else if (proc != NULL) {
594 visible = pfs_visible_proc(td, *pn, proc);
605 * Return directory entries.
608 pfs_readdir(struct vop_readdir_args *va)
610 struct vnode *vn = va->a_vp;
611 struct pfs_vdata *pvd = vn->v_data;
612 struct pfs_node *pd = pvd->pvd_pn;
613 pid_t pid = pvd->pvd_pid;
614 struct proc *p, *proc;
616 struct dirent *entry;
622 KASSERT(pd->pn_info == vn->v_mount->mnt_data,
623 ("%s(): pn_info does not match mountpoint", __func__));
624 PFS_TRACE(("%s pid %lu", pd->pn_name, (unsigned long)pid));
625 pfs_assert_not_owned(pd);
627 if (vn->v_type != VDIR)
628 PFS_RETURN (ENOTDIR);
631 /* only allow reading entire entries */
632 offset = uio->uio_offset;
633 resid = uio->uio_resid;
634 if (offset < 0 || offset % PFS_DELEN != 0 ||
635 (resid && resid < PFS_DELEN))
640 /* can't do this while holding the proc lock... */
641 buf = malloc(resid, M_IOV, M_WAITOK | M_ZERO);
642 sx_slock(&allproc_lock);
645 /* check if the directory is visible to the caller */
646 if (!pfs_visible(curthread, pd, pid, &proc)) {
647 sx_sunlock(&allproc_lock);
652 KASSERT(pid == NO_PID || proc != NULL,
653 ("%s(): no process for pid %lu", __func__, (unsigned long)pid));
655 /* skip unwanted entries */
656 for (pn = NULL, p = NULL; offset > 0; offset -= PFS_DELEN) {
657 if (pfs_iterate(curthread, proc, pd, &pn, &p) == -1) {
658 /* nothing left... */
662 sx_sunlock(&allproc_lock);
668 /* fill in entries */
670 while (pfs_iterate(curthread, proc, pd, &pn, &p) != -1 &&
671 resid >= PFS_DELEN) {
672 entry = (struct dirent *)ent;
673 entry->d_reclen = PFS_DELEN;
674 entry->d_fileno = pn_fileno(pn, pid);
675 /* PFS_DELEN was picked to fit PFS_NAMLEN */
676 for (i = 0; i < PFS_NAMELEN - 1 && pn->pn_name[i] != '\0'; ++i)
677 entry->d_name[i] = pn->pn_name[i];
678 entry->d_name[i] = 0;
680 switch (pn->pn_type) {
681 case pfstype_procdir:
683 ("reached procdir node with p == NULL"));
684 entry->d_namlen = snprintf(entry->d_name,
685 PFS_NAMELEN, "%d", p->p_pid);
691 entry->d_type = DT_DIR;
694 entry->d_type = DT_REG;
696 case pfstype_symlink:
697 entry->d_type = DT_LNK;
700 panic("%s has unexpected node type: %d", pn->pn_name, pn->pn_type);
702 PFS_TRACE(("%s", entry->d_name));
710 sx_sunlock(&allproc_lock);
711 PFS_TRACE(("%zd bytes", ent - buf));
712 error = uiomove(buf, ent - buf, uio);
718 * Read a symbolic link
721 pfs_readlink(struct vop_readlink_args *va)
723 struct vnode *vn = va->a_vp;
724 struct pfs_vdata *pvd = vn->v_data;
725 struct pfs_node *pn = pvd->pvd_pn;
726 struct uio *uio = va->a_uio;
727 struct proc *proc = NULL;
732 PFS_TRACE(("%s", pn->pn_name));
733 pfs_assert_not_owned(pn);
735 if (vn->v_type != VLNK)
738 if (pn->pn_fill == NULL)
741 if (pvd->pvd_pid != NO_PID) {
742 if ((proc = pfind(pvd->pvd_pid)) == NULL)
744 if (proc->p_flag & P_WEXIT) {
752 /* sbuf_new() can't fail with a static buffer */
753 sbuf_new(&sb, buf, sizeof buf, 0);
755 error = pn_fill(curthread, proc, pn, &sb, NULL);
766 error = uiomove_frombuf(sbuf_data(&sb), sbuf_len(&sb), uio);
775 pfs_reclaim(struct vop_reclaim_args *va)
777 struct vnode *vn = va->a_vp;
778 struct pfs_vdata *pvd = vn->v_data;
779 struct pfs_node *pn = pvd->pvd_pn;
781 PFS_TRACE(("%s", pn->pn_name));
782 pfs_assert_not_owned(pn);
784 return (pfs_vncache_free(va->a_vp));
791 pfs_setattr(struct vop_setattr_args *va)
793 struct vnode *vn = va->a_vp;
794 struct pfs_vdata *pvd = vn->v_data;
795 struct pfs_node *pn = pvd->pvd_pn;
797 PFS_TRACE(("%s", pn->pn_name));
798 pfs_assert_not_owned(pn);
800 PFS_RETURN (EOPNOTSUPP);
807 pfs_write(struct vop_write_args *va)
809 struct vnode *vn = va->a_vp;
810 struct pfs_vdata *pvd = vn->v_data;
811 struct pfs_node *pn = pvd->pvd_pn;
812 struct uio *uio = va->a_uio;
817 PFS_TRACE(("%s", pn->pn_name));
818 pfs_assert_not_owned(pn);
820 if (vn->v_type != VREG)
822 KASSERT(pn->pn_type != pfstype_file,
823 ("%s(): VREG vnode refers to non-file pfs_node", __func__));
825 if (!(pn->pn_flags & PFS_WR))
828 if (pn->pn_fill == NULL)
832 * This is necessary because either process' privileges may
833 * have changed since the open() call.
835 if (!pfs_visible(curthread, pn, pvd->pvd_pid, &proc))
842 if (pn->pn_flags & PFS_RAWWR) {
844 error = pn_fill(curthread, proc, pn, NULL, uio);
851 sbuf_uionew(&sb, uio, &error);
858 error = pn_fill(curthread, proc, pn, &sb, uio);
869 struct vop_vector pfs_vnodeops = {
870 .vop_default = &default_vnodeops,
872 .vop_access = pfs_access,
873 .vop_cachedlookup = pfs_lookup,
874 .vop_close = pfs_close,
875 .vop_create = VOP_EOPNOTSUPP,
876 .vop_getattr = pfs_getattr,
877 .vop_getextattr = pfs_getextattr,
878 .vop_ioctl = pfs_ioctl,
879 .vop_link = VOP_EOPNOTSUPP,
880 .vop_lookup = vfs_cache_lookup,
881 .vop_mkdir = VOP_EOPNOTSUPP,
882 .vop_mknod = VOP_EOPNOTSUPP,
883 .vop_open = pfs_open,
884 .vop_read = pfs_read,
885 .vop_readdir = pfs_readdir,
886 .vop_readlink = pfs_readlink,
887 .vop_reclaim = pfs_reclaim,
888 .vop_remove = VOP_EOPNOTSUPP,
889 .vop_rename = VOP_EOPNOTSUPP,
890 .vop_rmdir = VOP_EOPNOTSUPP,
891 .vop_setattr = pfs_setattr,
892 .vop_symlink = VOP_EOPNOTSUPP,
893 .vop_write = pfs_write,
894 /* XXX I've probably forgotten a few that need VOP_EOPNOTSUPP */