2 * Copyright (c) 1992, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software donated to Berkeley by
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)fdesc_vnops.c 8.9 (Berkeley) 1/21/94
41 #include <sys/param.h>
42 #include <sys/systm.h>
44 #include <sys/dirent.h>
45 #include <sys/filedesc.h>
46 #include <sys/kernel.h> /* boottime */
48 #include <sys/mutex.h>
49 #include <sys/malloc.h>
50 #include <sys/file.h> /* Must come after sys/malloc.h */
51 #include <sys/mount.h>
52 #include <sys/namei.h>
55 #include <sys/vnode.h>
57 #include <fs/fdescfs/fdesc.h>
60 #define FD_NHASH(ix) \
61 (&fdhashtbl[(ix) & fdhash])
62 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
65 struct mtx fdesc_hashmtx;
67 static vop_getattr_t fdesc_getattr;
68 static vop_lookup_t fdesc_lookup;
69 static vop_open_t fdesc_open;
70 static vop_readdir_t fdesc_readdir;
71 static vop_reclaim_t fdesc_reclaim;
72 static vop_setattr_t fdesc_setattr;
74 static struct vop_vector fdesc_vnodeops = {
75 .vop_default = &default_vnodeops,
77 .vop_access = VOP_NULL,
78 .vop_getattr = fdesc_getattr,
79 .vop_lookup = fdesc_lookup,
80 .vop_open = fdesc_open,
81 .vop_pathconf = vop_stdpathconf,
82 .vop_readdir = fdesc_readdir,
83 .vop_reclaim = fdesc_reclaim,
84 .vop_setattr = fdesc_setattr,
87 static void fdesc_insmntque_dtr(struct vnode *, void *);
88 static void fdesc_remove_entry(struct fdescnode *);
91 * Initialise cache headers
98 mtx_init(&fdesc_hashmtx, "fdescfs_hash", NULL, MTX_DEF);
99 fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
104 * Uninit ready for unload.
108 struct vfsconf *vfsp;
111 hashdestroy(fdhashtbl, M_CACHE, fdhash);
112 mtx_destroy(&fdesc_hashmtx);
117 * If allocating vnode fails, call this.
120 fdesc_insmntque_dtr(struct vnode *vp, void *arg)
128 * Remove an entry from the hash if it exists.
131 fdesc_remove_entry(struct fdescnode *fd)
133 struct fdhashhead *fc;
134 struct fdescnode *fd2;
136 fc = FD_NHASH(fd->fd_ix);
137 mtx_lock(&fdesc_hashmtx);
138 LIST_FOREACH(fd2, fc, fd_hash) {
140 LIST_REMOVE(fd, fd_hash);
144 mtx_unlock(&fdesc_hashmtx);
148 fdesc_allocvp(ftype, fd_fd, ix, mp, vpp, td)
156 struct fdescmount *fmp;
157 struct fdhashhead *fc;
158 struct fdescnode *fd, *fd2;
159 struct vnode *vp, *vp2;
164 mtx_lock(&fdesc_hashmtx);
166 * If a forced unmount is progressing, we need to drop it. The flags are
167 * protected by the hashmtx.
169 fmp = (struct fdescmount *)mp->mnt_data;
170 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) {
171 mtx_unlock(&fdesc_hashmtx);
175 LIST_FOREACH(fd, fc, fd_hash) {
176 if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
177 /* Get reference to vnode in case it's being free'd */
180 mtx_unlock(&fdesc_hashmtx);
181 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td))
187 mtx_unlock(&fdesc_hashmtx);
189 MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
191 error = getnewvnode("fdescfs", mp, &fdesc_vnodeops, &vp);
196 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
202 error = insmntque1(vp, mp, fdesc_insmntque_dtr, NULL);
208 /* Make sure that someone didn't beat us when inserting the vnode. */
209 mtx_lock(&fdesc_hashmtx);
211 * If a forced unmount is progressing, we need to drop it. The flags are
212 * protected by the hashmtx.
214 fmp = (struct fdescmount *)mp->mnt_data;
215 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) {
216 mtx_unlock(&fdesc_hashmtx);
223 LIST_FOREACH(fd2, fc, fd_hash) {
224 if (fd2->fd_ix == ix && fd2->fd_vnode->v_mount == mp) {
225 /* Get reference to vnode in case it's being free'd */
228 mtx_unlock(&fdesc_hashmtx);
229 error = vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK, td);
230 /* Someone beat us, dec use count and wait for reclaim */
233 /* If we didn't get it, return no vnode. */
241 /* If we came here, we can insert it safely. */
242 LIST_INSERT_HEAD(fc, fd, fd_hash);
243 mtx_unlock(&fdesc_hashmtx);
249 * vp is the current namei directory
250 * ndp is the name to locate in that directory...
254 struct vop_lookup_args /* {
255 struct vnode * a_dvp;
256 struct vnode ** a_vpp;
257 struct componentname * a_cnp;
260 struct vnode **vpp = ap->a_vpp;
261 struct vnode *dvp = ap->a_dvp;
262 struct componentname *cnp = ap->a_cnp;
263 char *pname = cnp->cn_nameptr;
264 struct thread *td = cnp->cn_thread;
266 int nlen = cnp->cn_namelen;
271 if ((cnp->cn_flags & ISLASTCN) &&
272 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
277 if (cnp->cn_namelen == 1 && *pname == '.') {
283 if (VTOFDESC(dvp)->fd_type != Froot) {
289 /* the only time a leading 0 is acceptable is if it's "0" */
290 if (*pname == '0' && nlen != 1) {
295 if (*pname < '0' || *pname > '9') {
299 fd = 10 * fd + *pname++ - '0';
302 if ((error = fget(td, fd, &fp)) != 0)
305 /* Check if we're looking up ourselves. */
306 if (VTOFDESC(dvp)->fd_ix == FD_DESC + fd) {
308 * In case we're holding the last reference to the file, the dvp
309 * will be re-acquired.
312 VOP_UNLOCK(dvp, 0, td);
315 /* Re-aquire the lock afterwards. */
316 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE, td);
321 * Unlock our root node (dvp) when doing this, since we might
322 * deadlock since the vnode might be locked by another thread
323 * and the root vnode lock will be obtained afterwards (in case
324 * we're looking up the fd of the root vnode), which will be the
325 * opposite lock order. Vhold the root vnode first so we don't
329 VOP_UNLOCK(dvp, 0, td);
330 error = fdesc_allocvp(Fdesc, fd, FD_DESC + fd, dvp->v_mount,
334 * The root vnode must be locked last to prevent deadlock condition.
336 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE, td);
352 struct vop_open_args /* {
355 struct ucred *a_cred;
359 struct vnode *vp = ap->a_vp;
361 if (VTOFDESC(vp)->fd_type == Froot)
365 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
366 * descriptor being sought for duplication. The error return ensures
367 * that the vnode for this device will be released by vn_open. Open
368 * will detect this special error and take the actions in dupfdopen.
369 * Other callers of vn_open or VOP_OPEN will simply report the
372 ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd; /* XXX */
378 struct vop_getattr_args /* {
381 struct ucred *a_cred;
385 struct vnode *vp = ap->a_vp;
386 struct vattr *vap = ap->a_vap;
392 switch (VTOFDESC(vp)->fd_type) {
394 vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
397 vap->va_size = DEV_BSIZE;
398 vap->va_fileid = VTOFDESC(vp)->fd_ix;
401 vap->va_blocksize = DEV_BSIZE;
402 vap->va_atime.tv_sec = boottime.tv_sec;
403 vap->va_atime.tv_nsec = 0;
404 vap->va_mtime = vap->va_atime;
405 vap->va_ctime = vap->va_mtime;
408 vap->va_rdev = NODEV;
414 fd = VTOFDESC(vp)->fd_fd;
416 if ((error = fget(ap->a_td, fd, &fp)) != 0)
419 bzero(&stb, sizeof(stb));
420 error = fo_stat(fp, &stb, ap->a_td->td_ucred, ap->a_td);
423 vap->va_type = IFTOVT(stb.st_mode);
424 vap->va_mode = stb.st_mode;
425 #define FDRX (VREAD|VEXEC)
426 if (vap->va_type == VDIR)
427 vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
431 vap->va_bytes = stb.st_blocks * stb.st_blksize;
432 vap->va_fileid = VTOFDESC(vp)->fd_ix;
433 vap->va_size = stb.st_size;
434 vap->va_blocksize = stb.st_blksize;
435 vap->va_rdev = stb.st_rdev;
438 * If no time data is provided, use the current time.
440 if (stb.st_atimespec.tv_sec == 0 &&
441 stb.st_atimespec.tv_nsec == 0)
442 nanotime(&stb.st_atimespec);
444 if (stb.st_ctimespec.tv_sec == 0 &&
445 stb.st_ctimespec.tv_nsec == 0)
446 nanotime(&stb.st_ctimespec);
448 if (stb.st_mtimespec.tv_sec == 0 &&
449 stb.st_mtimespec.tv_nsec == 0)
450 nanotime(&stb.st_mtimespec);
452 vap->va_atime = stb.st_atimespec;
453 vap->va_mtime = stb.st_mtimespec;
454 vap->va_ctime = stb.st_ctimespec;
455 vap->va_uid = stb.st_uid;
456 vap->va_gid = stb.st_gid;
463 panic("fdesc_getattr");
468 vp->v_type = vap->va_type;
474 struct vop_setattr_args /* {
477 struct ucred *a_cred;
481 struct vattr *vap = ap->a_vap;
489 * Can't mess with the root vnode
491 if (VTOFDESC(ap->a_vp)->fd_type == Froot)
494 fd = VTOFDESC(ap->a_vp)->fd_fd;
497 * Allow setattr where there is an underlying vnode.
499 error = getvnode(ap->a_td->td_proc->p_fd, fd, &fp);
502 * getvnode() returns EINVAL if the file descriptor is not
503 * backed by a vnode. Silently drop all changes except
504 * chflags(2) in this case.
506 if (error == EINVAL) {
507 if (vap->va_flags != VNOVAL)
515 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) {
516 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
517 error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_td);
518 VOP_UNLOCK(vp, 0, ap->a_td);
519 vn_finished_write(mp);
529 struct vop_readdir_args /* {
532 struct ucred *a_cred;
538 struct uio *uio = ap->a_uio;
539 struct filedesc *fdp;
541 struct dirent *dp = &d;
542 int error, i, off, fcnt;
545 * We don't allow exporting fdesc mounts, and currently local
546 * requests do not need cookies.
549 panic("fdesc_readdir: not hungry");
551 if (VTOFDESC(ap->a_vp)->fd_type != Froot)
552 panic("fdesc_readdir: not dir");
554 off = (int)uio->uio_offset;
555 if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
556 uio->uio_resid < UIO_MX)
558 i = (u_int)off / UIO_MX;
559 fdp = uio->uio_td->td_proc->p_fd;
562 fcnt = i - 2; /* The first two nodes are `.' and `..' */
565 while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
569 bzero((caddr_t)dp, UIO_MX);
571 dp->d_fileno = i + FD_ROOT;
572 dp->d_namlen = i + 1;
573 dp->d_reclen = UIO_MX;
574 bcopy("..", dp->d_name, dp->d_namlen);
575 dp->d_name[i + 1] = '\0';
579 if (fdp->fd_ofiles[fcnt] == NULL) {
580 FILEDESC_SUNLOCK(fdp);
584 bzero((caddr_t) dp, UIO_MX);
585 dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
586 dp->d_reclen = UIO_MX;
587 dp->d_type = DT_UNKNOWN;
588 dp->d_fileno = i + FD_DESC;
592 * And ship to userland
594 FILEDESC_SUNLOCK(fdp);
595 error = uiomove(dp, UIO_MX, uio);
602 FILEDESC_SUNLOCK(fdp);
605 uio->uio_offset = i * UIO_MX;
611 struct vop_reclaim_args /* {
616 struct fdescnode *fd;
620 fdesc_remove_entry(fd);
621 FREE(vp->v_data, M_TEMP);