xref: /freebsd/sys/fs/fdescfs/fdesc_vnops.c (revision 262e143bd46171a6415a5b28af260a5efa2a3db8)
1 /*-
2  * Copyright (c) 1992, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software donated to Berkeley by
6  * Jan-Simon Pendry.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
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.
19  *
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
30  * SUCH DAMAGE.
31  *
32  *	@(#)fdesc_vnops.c	8.9 (Berkeley) 1/21/94
33  *
34  * $FreeBSD$
35  */
36 
37 /*
38  * /dev/fd Filesystem
39  */
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/conf.h>
44 #include <sys/dirent.h>
45 #include <sys/filedesc.h>
46 #include <sys/kernel.h>	/* boottime */
47 #include <sys/lock.h>
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>
53 #include <sys/proc.h>
54 #include <sys/stat.h>
55 #include <sys/vnode.h>
56 
57 #include <fs/fdescfs/fdesc.h>
58 
59 #define FDL_WANT	0x01
60 #define FDL_LOCKED	0x02
61 static int fdcache_lock;
62 
63 #define	NFDCACHE 4
64 #define FD_NHASH(ix) \
65 	(&fdhashtbl[(ix) & fdhash])
66 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
67 static u_long fdhash;
68 
69 static vop_getattr_t	fdesc_getattr;
70 static vop_inactive_t	fdesc_inactive;
71 static vop_lookup_t	fdesc_lookup;
72 static vop_open_t	fdesc_open;
73 static vop_readdir_t	fdesc_readdir;
74 static vop_reclaim_t	fdesc_reclaim;
75 static vop_setattr_t	fdesc_setattr;
76 
77 static struct vop_vector fdesc_vnodeops = {
78 	.vop_default =		&default_vnodeops,
79 
80 	.vop_access =		VOP_NULL,
81 	.vop_getattr =		fdesc_getattr,
82 	.vop_inactive =		fdesc_inactive,
83 	.vop_lookup =		fdesc_lookup,
84 	.vop_open =		fdesc_open,
85 	.vop_pathconf =		vop_stdpathconf,
86 	.vop_readdir =		fdesc_readdir,
87 	.vop_reclaim =		fdesc_reclaim,
88 	.vop_setattr =		fdesc_setattr,
89 };
90 
91 /*
92  * Initialise cache headers
93  */
94 int
95 fdesc_init(vfsp)
96 	struct vfsconf *vfsp;
97 {
98 
99 	fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
100 	return (0);
101 }
102 
103 int
104 fdesc_allocvp(ftype, ix, mp, vpp, td)
105 	fdntype ftype;
106 	int ix;
107 	struct mount *mp;
108 	struct vnode **vpp;
109 	struct thread *td;
110 {
111 	struct fdhashhead *fc;
112 	struct fdescnode *fd;
113 	int error = 0;
114 
115 	fc = FD_NHASH(ix);
116 loop:
117 	LIST_FOREACH(fd, fc, fd_hash) {
118 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
119 			if (vget(fd->fd_vnode, 0, td))
120 				goto loop;
121 			*vpp = fd->fd_vnode;
122 			return (error);
123 		}
124 	}
125 
126 	/*
127 	 * otherwise lock the array while we call getnewvnode
128 	 * since that can block.
129 	 */
130 	if (fdcache_lock & FDL_LOCKED) {
131 		fdcache_lock |= FDL_WANT;
132 		(void) tsleep( &fdcache_lock, PINOD, "fdalvp", 0);
133 		goto loop;
134 	}
135 	fdcache_lock |= FDL_LOCKED;
136 
137 	/*
138 	 * Do the MALLOC before the getnewvnode since doing so afterward
139 	 * might cause a bogus v_data pointer to get dereferenced
140 	 * elsewhere if MALLOC should block.
141 	 */
142 	MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
143 
144 	error = getnewvnode("fdesc", mp, &fdesc_vnodeops, vpp);
145 	if (error) {
146 		FREE(fd, M_TEMP);
147 		goto out;
148 	}
149 	(*vpp)->v_data = fd;
150 	fd->fd_vnode = *vpp;
151 	fd->fd_type = ftype;
152 	fd->fd_fd = -1;
153 	fd->fd_ix = ix;
154 	LIST_INSERT_HEAD(fc, fd, fd_hash);
155 
156 out:
157 	fdcache_lock &= ~FDL_LOCKED;
158 
159 	if (fdcache_lock & FDL_WANT) {
160 		fdcache_lock &= ~FDL_WANT;
161 		wakeup( &fdcache_lock);
162 	}
163 
164 	return (error);
165 }
166 
167 /*
168  * vp is the current namei directory
169  * ndp is the name to locate in that directory...
170  */
171 static int
172 fdesc_lookup(ap)
173 	struct vop_lookup_args /* {
174 		struct vnode * a_dvp;
175 		struct vnode ** a_vpp;
176 		struct componentname * a_cnp;
177 	} */ *ap;
178 {
179 	struct vnode **vpp = ap->a_vpp;
180 	struct vnode *dvp = ap->a_dvp;
181 	struct componentname *cnp = ap->a_cnp;
182 	char *pname = cnp->cn_nameptr;
183 	struct thread *td = cnp->cn_thread;
184 	struct file *fp;
185 	int nlen = cnp->cn_namelen;
186 	u_int fd;
187 	int error;
188 	struct vnode *fvp;
189 
190 	if ((cnp->cn_flags & ISLASTCN) &&
191 	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
192 		error = EROFS;
193 		goto bad;
194 	}
195 
196 	VOP_UNLOCK(dvp, 0, td);
197 	if (cnp->cn_namelen == 1 && *pname == '.') {
198 		*vpp = dvp;
199 		VREF(dvp);
200 		vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
201 		return (0);
202 	}
203 
204 	if (VTOFDESC(dvp)->fd_type != Froot) {
205 		error = ENOTDIR;
206 		goto bad;
207 	}
208 
209 	fd = 0;
210 	/* the only time a leading 0 is acceptable is if it's "0" */
211 	if (*pname == '0' && nlen != 1) {
212 		error = ENOENT;
213 		goto bad;
214 	}
215 	while (nlen--) {
216 		if (*pname < '0' || *pname > '9') {
217 			error = ENOENT;
218 			goto bad;
219 		}
220 		fd = 10 * fd + *pname++ - '0';
221 	}
222 
223 	if ((error = fget(td, fd, &fp)) != 0)
224 		goto bad;
225 
226 	error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp, td);
227 	fdrop(fp, td);
228 	if (error)
229 		goto bad;
230 	VTOFDESC(fvp)->fd_fd = fd;
231 	vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY, td);
232 	*vpp = fvp;
233 	return (0);
234 
235 bad:
236 	vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
237 	*vpp = NULL;
238 	return (error);
239 }
240 
241 static int
242 fdesc_open(ap)
243 	struct vop_open_args /* {
244 		struct vnode *a_vp;
245 		int  a_mode;
246 		struct ucred *a_cred;
247 		struct thread *a_td;
248 	} */ *ap;
249 {
250 	struct vnode *vp = ap->a_vp;
251 
252 	if (VTOFDESC(vp)->fd_type == Froot)
253 		return (0);
254 
255 	/*
256 	 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
257 	 * descriptor being sought for duplication. The error return ensures
258 	 * that the vnode for this device will be released by vn_open. Open
259 	 * will detect this special error and take the actions in dupfdopen.
260 	 * Other callers of vn_open or VOP_OPEN will simply report the
261 	 * error.
262 	 */
263 	ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
264 	return (ENODEV);
265 }
266 
267 static int
268 fdesc_getattr(ap)
269 	struct vop_getattr_args /* {
270 		struct vnode *a_vp;
271 		struct vattr *a_vap;
272 		struct ucred *a_cred;
273 		struct thread *a_td;
274 	} */ *ap;
275 {
276 	struct vnode *vp = ap->a_vp;
277 	struct vattr *vap = ap->a_vap;
278 	struct file *fp;
279 	struct stat stb;
280 	u_int fd;
281 	int error = 0;
282 
283 	switch (VTOFDESC(vp)->fd_type) {
284 	case Froot:
285 		VATTR_NULL(vap);
286 
287 		vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
288 		vap->va_type = VDIR;
289 		vap->va_nlink = 2;
290 		vap->va_size = DEV_BSIZE;
291 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
292 		vap->va_uid = 0;
293 		vap->va_gid = 0;
294 		vap->va_blocksize = DEV_BSIZE;
295 		vap->va_atime.tv_sec = boottime.tv_sec;
296 		vap->va_atime.tv_nsec = 0;
297 		vap->va_mtime = vap->va_atime;
298 		vap->va_ctime = vap->va_mtime;
299 		vap->va_gen = 0;
300 		vap->va_flags = 0;
301 		vap->va_rdev = 0;
302 		vap->va_bytes = 0;
303 		break;
304 
305 	case Fdesc:
306 		fd = VTOFDESC(vp)->fd_fd;
307 
308 		if ((error = fget(ap->a_td, fd, &fp)) != 0)
309 			return (error);
310 
311 		bzero(&stb, sizeof(stb));
312 		error = fo_stat(fp, &stb, ap->a_td->td_ucred, ap->a_td);
313 		fdrop(fp, ap->a_td);
314 		if (error == 0) {
315 			VATTR_NULL(vap);
316 			vap->va_type = IFTOVT(stb.st_mode);
317 			vap->va_mode = stb.st_mode;
318 #define FDRX (VREAD|VEXEC)
319 			if (vap->va_type == VDIR)
320 				vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
321 #undef FDRX
322 			vap->va_nlink = 1;
323 			vap->va_flags = 0;
324 			vap->va_bytes = stb.st_blocks * stb.st_blksize;
325 			vap->va_fileid = VTOFDESC(vp)->fd_ix;
326 			vap->va_size = stb.st_size;
327 			vap->va_blocksize = stb.st_blksize;
328 			vap->va_rdev = stb.st_rdev;
329 
330 			/*
331 			 * If no time data is provided, use the current time.
332 			 */
333 			if (stb.st_atimespec.tv_sec == 0 &&
334 			    stb.st_atimespec.tv_nsec == 0)
335 				nanotime(&stb.st_atimespec);
336 
337 			if (stb.st_ctimespec.tv_sec == 0 &&
338 			    stb.st_ctimespec.tv_nsec == 0)
339 				nanotime(&stb.st_ctimespec);
340 
341 			if (stb.st_mtimespec.tv_sec == 0 &&
342 			    stb.st_mtimespec.tv_nsec == 0)
343 				nanotime(&stb.st_mtimespec);
344 
345 			vap->va_atime = stb.st_atimespec;
346 			vap->va_mtime = stb.st_mtimespec;
347 			vap->va_ctime = stb.st_ctimespec;
348 			vap->va_uid = stb.st_uid;
349 			vap->va_gid = stb.st_gid;
350 		}
351 		break;
352 
353 	default:
354 		panic("fdesc_getattr");
355 		break;
356 	}
357 
358 	if (error == 0)
359 		vp->v_type = vap->va_type;
360 	return (error);
361 }
362 
363 static int
364 fdesc_setattr(ap)
365 	struct vop_setattr_args /* {
366 		struct vnode *a_vp;
367 		struct vattr *a_vap;
368 		struct ucred *a_cred;
369 		struct thread *a_td;
370 	} */ *ap;
371 {
372 	struct vattr *vap = ap->a_vap;
373 	struct vnode *vp;
374 	struct mount *mp;
375 	struct file *fp;
376 	unsigned fd;
377 	int error;
378 
379 	/*
380 	 * Can't mess with the root vnode
381 	 */
382 	if (VTOFDESC(ap->a_vp)->fd_type == Froot)
383 		return (EACCES);
384 
385 	fd = VTOFDESC(ap->a_vp)->fd_fd;
386 
387 	/*
388 	 * Allow setattr where there is an underlying vnode.
389 	 */
390 	error = getvnode(ap->a_td->td_proc->p_fd, fd, &fp);
391 	if (error) {
392 		/*
393 		 * getvnode() returns EINVAL if the file descriptor is not
394 		 * backed by a vnode.  Silently drop all changes except
395 		 * chflags(2) in this case.
396 		 */
397 		if (error == EINVAL) {
398 			if (vap->va_flags != VNOVAL)
399 				error = EOPNOTSUPP;
400 			else
401 				error = 0;
402 		}
403 		return (error);
404 	}
405 	vp = fp->f_vnode;
406 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
407 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) {
408 		error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_td);
409 		vn_finished_write(mp);
410 	}
411 	VOP_UNLOCK(vp, 0, ap->a_td);
412 	fdrop(fp, ap->a_td);
413 	return (error);
414 }
415 
416 #define UIO_MX 16
417 
418 static int
419 fdesc_readdir(ap)
420 	struct vop_readdir_args /* {
421 		struct vnode *a_vp;
422 		struct uio *a_uio;
423 		struct ucred *a_cred;
424 		int *a_eofflag;
425 		u_long *a_cookies;
426 		int a_ncookies;
427 	} */ *ap;
428 {
429 	struct uio *uio = ap->a_uio;
430 	struct filedesc *fdp;
431 	struct dirent d;
432 	struct dirent *dp = &d;
433 	int error, i, off, fcnt;
434 
435 	/*
436 	 * We don't allow exporting fdesc mounts, and currently local
437 	 * requests do not need cookies.
438 	 */
439 	if (ap->a_ncookies)
440 		panic("fdesc_readdir: not hungry");
441 
442 	if (VTOFDESC(ap->a_vp)->fd_type != Froot)
443 		panic("fdesc_readdir: not dir");
444 
445 	off = (int)uio->uio_offset;
446 	if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
447 	    uio->uio_resid < UIO_MX)
448 		return (EINVAL);
449 	i = (u_int)off / UIO_MX;
450 	fdp = uio->uio_td->td_proc->p_fd;
451 	error = 0;
452 
453 	fcnt = i - 2;		/* The first two nodes are `.' and `..' */
454 
455 	FILEDESC_LOCK_FAST(fdp);
456 	while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
457 		switch (i) {
458 		case 0:	/* `.' */
459 		case 1: /* `..' */
460 			bzero((caddr_t)dp, UIO_MX);
461 
462 			dp->d_fileno = i + FD_ROOT;
463 			dp->d_namlen = i + 1;
464 			dp->d_reclen = UIO_MX;
465 			bcopy("..", dp->d_name, dp->d_namlen);
466 			dp->d_name[i + 1] = '\0';
467 			dp->d_type = DT_DIR;
468 			break;
469 		default:
470 			if (fdp->fd_ofiles[fcnt] == NULL) {
471 				FILEDESC_UNLOCK_FAST(fdp);
472 				goto done;
473 			}
474 
475 			bzero((caddr_t) dp, UIO_MX);
476 			dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
477 			dp->d_reclen = UIO_MX;
478 			dp->d_type = DT_UNKNOWN;
479 			dp->d_fileno = i + FD_DESC;
480 			break;
481 		}
482 		/*
483 		 * And ship to userland
484 		 */
485 		FILEDESC_UNLOCK_FAST(fdp);
486 		error = uiomove(dp, UIO_MX, uio);
487 		if (error)
488 			goto done;
489 		FILEDESC_LOCK_FAST(fdp);
490 		i++;
491 		fcnt++;
492 	}
493 	FILEDESC_UNLOCK_FAST(fdp);
494 
495 done:
496 	uio->uio_offset = i * UIO_MX;
497 	return (error);
498 }
499 
500 static int
501 fdesc_inactive(ap)
502 	struct vop_inactive_args /* {
503 		struct vnode *a_vp;
504 		struct thread *a_td;
505 	} */ *ap;
506 {
507 	struct vnode *vp = ap->a_vp;
508 
509 	/*
510 	 * Clear out the v_type field to avoid
511 	 * nasty things happening in vgone().
512 	 */
513 	vp->v_type = VNON;
514 	return (0);
515 }
516 
517 static int
518 fdesc_reclaim(ap)
519 	struct vop_reclaim_args /* {
520 		struct vnode *a_vp;
521 	} */ *ap;
522 {
523 	struct vnode *vp = ap->a_vp;
524 	struct fdescnode *fd = VTOFDESC(vp);
525 
526 	LIST_REMOVE(fd, fd_hash);
527 	FREE(vp->v_data, M_TEMP);
528 	vp->v_data = 0;
529 
530 	return (0);
531 }
532