xref: /freebsd/sys/fs/fdescfs/fdesc_vnops.c (revision 6af83ee0d2941d18880b6aaa2b4facd1d30c6106)
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 extern struct vop_vector fdesc_vnodeops;
78 
79 /*
80  * Initialise cache headers
81  */
82 int
83 fdesc_init(vfsp)
84 	struct vfsconf *vfsp;
85 {
86 
87 	fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
88 	return (0);
89 }
90 
91 int
92 fdesc_allocvp(ftype, ix, mp, vpp, td)
93 	fdntype ftype;
94 	int ix;
95 	struct mount *mp;
96 	struct vnode **vpp;
97 	struct thread *td;
98 {
99 	struct fdhashhead *fc;
100 	struct fdescnode *fd;
101 	int error = 0;
102 
103 	fc = FD_NHASH(ix);
104 loop:
105 	LIST_FOREACH(fd, fc, fd_hash) {
106 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
107 			if (vget(fd->fd_vnode, 0, td))
108 				goto loop;
109 			*vpp = fd->fd_vnode;
110 			return (error);
111 		}
112 	}
113 
114 	/*
115 	 * otherwise lock the array while we call getnewvnode
116 	 * since that can block.
117 	 */
118 	if (fdcache_lock & FDL_LOCKED) {
119 		fdcache_lock |= FDL_WANT;
120 		(void) tsleep( &fdcache_lock, PINOD, "fdalvp", 0);
121 		goto loop;
122 	}
123 	fdcache_lock |= FDL_LOCKED;
124 
125 	/*
126 	 * Do the MALLOC before the getnewvnode since doing so afterward
127 	 * might cause a bogus v_data pointer to get dereferenced
128 	 * elsewhere if MALLOC should block.
129 	 */
130 	MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
131 
132 	error = getnewvnode("fdesc", mp, &fdesc_vnodeops, vpp);
133 	if (error) {
134 		FREE(fd, M_TEMP);
135 		goto out;
136 	}
137 	(*vpp)->v_data = fd;
138 	fd->fd_vnode = *vpp;
139 	fd->fd_type = ftype;
140 	fd->fd_fd = -1;
141 	fd->fd_ix = ix;
142 	LIST_INSERT_HEAD(fc, fd, fd_hash);
143 
144 out:
145 	fdcache_lock &= ~FDL_LOCKED;
146 
147 	if (fdcache_lock & FDL_WANT) {
148 		fdcache_lock &= ~FDL_WANT;
149 		wakeup( &fdcache_lock);
150 	}
151 
152 	return (error);
153 }
154 
155 /*
156  * vp is the current namei directory
157  * ndp is the name to locate in that directory...
158  */
159 static int
160 fdesc_lookup(ap)
161 	struct vop_lookup_args /* {
162 		struct vnode * a_dvp;
163 		struct vnode ** a_vpp;
164 		struct componentname * a_cnp;
165 	} */ *ap;
166 {
167 	struct vnode **vpp = ap->a_vpp;
168 	struct vnode *dvp = ap->a_dvp;
169 	struct componentname *cnp = ap->a_cnp;
170 	char *pname = cnp->cn_nameptr;
171 	struct thread *td = cnp->cn_thread;
172 	struct file *fp;
173 	int nlen = cnp->cn_namelen;
174 	u_int fd;
175 	int error;
176 	struct vnode *fvp;
177 
178 	if ((cnp->cn_flags & ISLASTCN) &&
179 	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
180 		error = EROFS;
181 		goto bad;
182 	}
183 
184 	VOP_UNLOCK(dvp, 0, td);
185 	if (cnp->cn_namelen == 1 && *pname == '.') {
186 		*vpp = dvp;
187 		VREF(dvp);
188 		vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
189 		return (0);
190 	}
191 
192 	if (VTOFDESC(dvp)->fd_type != Froot) {
193 		error = ENOTDIR;
194 		goto bad;
195 	}
196 
197 	fd = 0;
198 	/* the only time a leading 0 is acceptable is if it's "0" */
199 	if (*pname == '0' && nlen != 1) {
200 		error = ENOENT;
201 		goto bad;
202 	}
203 	while (nlen--) {
204 		if (*pname < '0' || *pname > '9') {
205 			error = ENOENT;
206 			goto bad;
207 		}
208 		fd = 10 * fd + *pname++ - '0';
209 	}
210 
211 	if ((error = fget(td, fd, &fp)) != 0)
212 		goto bad;
213 
214 	error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp, td);
215 	fdrop(fp, td);
216 	if (error)
217 		goto bad;
218 	VTOFDESC(fvp)->fd_fd = fd;
219 	vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY, td);
220 	*vpp = fvp;
221 	return (0);
222 
223 bad:
224 	vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
225 	*vpp = NULL;
226 	return (error);
227 }
228 
229 static int
230 fdesc_open(ap)
231 	struct vop_open_args /* {
232 		struct vnode *a_vp;
233 		int  a_mode;
234 		struct ucred *a_cred;
235 		struct thread *a_td;
236 	} */ *ap;
237 {
238 	struct vnode *vp = ap->a_vp;
239 
240 	if (VTOFDESC(vp)->fd_type == Froot)
241 		return (0);
242 
243 	/*
244 	 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
245 	 * descriptor being sought for duplication. The error return ensures
246 	 * that the vnode for this device will be released by vn_open. Open
247 	 * will detect this special error and take the actions in dupfdopen.
248 	 * Other callers of vn_open or VOP_OPEN will simply report the
249 	 * error.
250 	 */
251 	ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
252 	return (ENODEV);
253 }
254 
255 static int
256 fdesc_getattr(ap)
257 	struct vop_getattr_args /* {
258 		struct vnode *a_vp;
259 		struct vattr *a_vap;
260 		struct ucred *a_cred;
261 		struct thread *a_td;
262 	} */ *ap;
263 {
264 	struct vnode *vp = ap->a_vp;
265 	struct vattr *vap = ap->a_vap;
266 	struct file *fp;
267 	struct stat stb;
268 	u_int fd;
269 	int error = 0;
270 
271 	switch (VTOFDESC(vp)->fd_type) {
272 	case Froot:
273 		VATTR_NULL(vap);
274 
275 		vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
276 		vap->va_type = VDIR;
277 		vap->va_nlink = 2;
278 		vap->va_size = DEV_BSIZE;
279 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
280 		vap->va_uid = 0;
281 		vap->va_gid = 0;
282 		vap->va_blocksize = DEV_BSIZE;
283 		vap->va_atime.tv_sec = boottime.tv_sec;
284 		vap->va_atime.tv_nsec = 0;
285 		vap->va_mtime = vap->va_atime;
286 		vap->va_ctime = vap->va_mtime;
287 		vap->va_gen = 0;
288 		vap->va_flags = 0;
289 		vap->va_rdev = 0;
290 		vap->va_bytes = 0;
291 		break;
292 
293 	case Fdesc:
294 		fd = VTOFDESC(vp)->fd_fd;
295 
296 		if ((error = fget(ap->a_td, fd, &fp)) != 0)
297 			return (error);
298 
299 		bzero(&stb, sizeof(stb));
300 		error = fo_stat(fp, &stb, ap->a_td->td_ucred, ap->a_td);
301 		fdrop(fp, ap->a_td);
302 		if (error == 0) {
303 			VATTR_NULL(vap);
304 			vap->va_type = IFTOVT(stb.st_mode);
305 			vap->va_mode = stb.st_mode;
306 #define FDRX (VREAD|VEXEC)
307 			if (vap->va_type == VDIR)
308 				vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
309 #undef FDRX
310 			vap->va_nlink = 1;
311 			vap->va_flags = 0;
312 			vap->va_bytes = stb.st_blocks * stb.st_blksize;
313 			vap->va_fileid = VTOFDESC(vp)->fd_ix;
314 			vap->va_size = stb.st_size;
315 			vap->va_blocksize = stb.st_blksize;
316 			vap->va_rdev = stb.st_rdev;
317 
318 			/*
319 			 * If no time data is provided, use the current time.
320 			 */
321 			if (stb.st_atimespec.tv_sec == 0 &&
322 			    stb.st_atimespec.tv_nsec == 0)
323 				nanotime(&stb.st_atimespec);
324 
325 			if (stb.st_ctimespec.tv_sec == 0 &&
326 			    stb.st_ctimespec.tv_nsec == 0)
327 				nanotime(&stb.st_ctimespec);
328 
329 			if (stb.st_mtimespec.tv_sec == 0 &&
330 			    stb.st_mtimespec.tv_nsec == 0)
331 				nanotime(&stb.st_mtimespec);
332 
333 			vap->va_atime = stb.st_atimespec;
334 			vap->va_mtime = stb.st_mtimespec;
335 			vap->va_ctime = stb.st_ctimespec;
336 			vap->va_uid = stb.st_uid;
337 			vap->va_gid = stb.st_gid;
338 		}
339 		break;
340 
341 	default:
342 		panic("fdesc_getattr");
343 		break;
344 	}
345 
346 	if (error == 0)
347 		vp->v_type = vap->va_type;
348 	return (error);
349 }
350 
351 static int
352 fdesc_setattr(ap)
353 	struct vop_setattr_args /* {
354 		struct vnode *a_vp;
355 		struct vattr *a_vap;
356 		struct ucred *a_cred;
357 		struct thread *a_td;
358 	} */ *ap;
359 {
360 	struct vattr *vap = ap->a_vap;
361 	struct vnode *vp;
362 	struct mount *mp;
363 	struct file *fp;
364 	unsigned fd;
365 	int error;
366 
367 	/*
368 	 * Can't mess with the root vnode
369 	 */
370 	if (VTOFDESC(ap->a_vp)->fd_type == Froot)
371 		return (EACCES);
372 
373 	fd = VTOFDESC(ap->a_vp)->fd_fd;
374 
375 	/*
376 	 * Allow setattr where there is an underlying vnode.
377 	 */
378 	error = getvnode(ap->a_td->td_proc->p_fd, fd, &fp);
379 	if (error) {
380 		/*
381 		 * getvnode() returns EINVAL if the file descriptor is not
382 		 * backed by a vnode.  Silently drop all changes except
383 		 * chflags(2) in this case.
384 		 */
385 		if (error == EINVAL) {
386 			if (vap->va_flags != VNOVAL)
387 				error = EOPNOTSUPP;
388 			else
389 				error = 0;
390 		}
391 		return (error);
392 	}
393 	vp = fp->f_vnode;
394 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
395 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) {
396 		error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_td);
397 		vn_finished_write(mp);
398 	}
399 	VOP_UNLOCK(vp, 0, ap->a_td);
400 	fdrop(fp, ap->a_td);
401 	return (error);
402 }
403 
404 #define UIO_MX 16
405 
406 static int
407 fdesc_readdir(ap)
408 	struct vop_readdir_args /* {
409 		struct vnode *a_vp;
410 		struct uio *a_uio;
411 		struct ucred *a_cred;
412 		int *a_eofflag;
413 		u_long *a_cookies;
414 		int a_ncookies;
415 	} */ *ap;
416 {
417 	struct uio *uio = ap->a_uio;
418 	struct filedesc *fdp;
419 	struct dirent d;
420 	struct dirent *dp = &d;
421 	int error, i, off, fcnt;
422 
423 	/*
424 	 * We don't allow exporting fdesc mounts, and currently local
425 	 * requests do not need cookies.
426 	 */
427 	if (ap->a_ncookies)
428 		panic("fdesc_readdir: not hungry");
429 
430 	if (VTOFDESC(ap->a_vp)->fd_type != Froot)
431 		panic("fdesc_readdir: not dir");
432 
433 	off = (int)uio->uio_offset;
434 	if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
435 	    uio->uio_resid < UIO_MX)
436 		return (EINVAL);
437 	i = (u_int)off / UIO_MX;
438 	fdp = uio->uio_td->td_proc->p_fd;
439 	error = 0;
440 
441 	fcnt = i - 2;		/* The first two nodes are `.' and `..' */
442 
443 	FILEDESC_LOCK_FAST(fdp);
444 	while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
445 		switch (i) {
446 		case 0:	/* `.' */
447 		case 1: /* `..' */
448 			bzero((caddr_t)dp, UIO_MX);
449 
450 			dp->d_fileno = i + FD_ROOT;
451 			dp->d_namlen = i + 1;
452 			dp->d_reclen = UIO_MX;
453 			bcopy("..", dp->d_name, dp->d_namlen);
454 			dp->d_name[i + 1] = '\0';
455 			dp->d_type = DT_DIR;
456 			break;
457 		default:
458 			if (fdp->fd_ofiles[fcnt] == NULL) {
459 				FILEDESC_UNLOCK_FAST(fdp);
460 				goto done;
461 			}
462 
463 			bzero((caddr_t) dp, UIO_MX);
464 			dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
465 			dp->d_reclen = UIO_MX;
466 			dp->d_type = DT_UNKNOWN;
467 			dp->d_fileno = i + FD_DESC;
468 			break;
469 		}
470 		/*
471 		 * And ship to userland
472 		 */
473 		FILEDESC_UNLOCK_FAST(fdp);
474 		error = uiomove(dp, UIO_MX, uio);
475 		if (error)
476 			goto done;
477 		FILEDESC_LOCK_FAST(fdp);
478 		i++;
479 		fcnt++;
480 	}
481 	FILEDESC_UNLOCK_FAST(fdp);
482 
483 done:
484 	uio->uio_offset = i * UIO_MX;
485 	return (error);
486 }
487 
488 static int
489 fdesc_inactive(ap)
490 	struct vop_inactive_args /* {
491 		struct vnode *a_vp;
492 		struct thread *a_td;
493 	} */ *ap;
494 {
495 	struct vnode *vp = ap->a_vp;
496 
497 	/*
498 	 * Clear out the v_type field to avoid
499 	 * nasty things happening in vgone().
500 	 */
501 	VOP_UNLOCK(vp, 0, ap->a_td);
502 	vp->v_type = VNON;
503 	return (0);
504 }
505 
506 static int
507 fdesc_reclaim(ap)
508 	struct vop_reclaim_args /* {
509 		struct vnode *a_vp;
510 	} */ *ap;
511 {
512 	struct vnode *vp = ap->a_vp;
513 	struct fdescnode *fd = VTOFDESC(vp);
514 
515 	LIST_REMOVE(fd, fd_hash);
516 	FREE(vp->v_data, M_TEMP);
517 	vp->v_data = 0;
518 
519 	return (0);
520 }
521 
522 static struct vop_vector fdesc_vnodeops = {
523 	.vop_default =		&default_vnodeops,
524 
525 	.vop_access =		VOP_NULL,
526 	.vop_getattr =		fdesc_getattr,
527 	.vop_inactive =		fdesc_inactive,
528 	.vop_lookup =		fdesc_lookup,
529 	.vop_open =		fdesc_open,
530 	.vop_pathconf =		vop_stdpathconf,
531 	.vop_readdir =		fdesc_readdir,
532 	.vop_reclaim =		fdesc_reclaim,
533 	.vop_setattr =		fdesc_setattr,
534 };
535