xref: /freebsd/sys/fs/fdescfs/fdesc_vnops.c (revision 1e413cf93298b5b97441a21d9a50fdcd0ee9945e)
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 	/* XXX: vnode should be locked here */
155 	error = insmntque(*vpp, mp); /* XXX: Too early for mpsafe fs */
156 	if (error != 0) {
157 		free(fd, M_TEMP);
158 		*vpp = NULLVP;
159 		goto out;
160 	}
161 	LIST_INSERT_HEAD(fc, fd, fd_hash);
162 
163 out:
164 	fdcache_lock &= ~FDL_LOCKED;
165 
166 	if (fdcache_lock & FDL_WANT) {
167 		fdcache_lock &= ~FDL_WANT;
168 		wakeup( &fdcache_lock);
169 	}
170 
171 	return (error);
172 }
173 
174 /*
175  * vp is the current namei directory
176  * ndp is the name to locate in that directory...
177  */
178 static int
179 fdesc_lookup(ap)
180 	struct vop_lookup_args /* {
181 		struct vnode * a_dvp;
182 		struct vnode ** a_vpp;
183 		struct componentname * a_cnp;
184 	} */ *ap;
185 {
186 	struct vnode **vpp = ap->a_vpp;
187 	struct vnode *dvp = ap->a_dvp;
188 	struct componentname *cnp = ap->a_cnp;
189 	char *pname = cnp->cn_nameptr;
190 	struct thread *td = cnp->cn_thread;
191 	struct file *fp;
192 	int nlen = cnp->cn_namelen;
193 	u_int fd;
194 	int error;
195 	struct vnode *fvp;
196 
197 	if ((cnp->cn_flags & ISLASTCN) &&
198 	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
199 		error = EROFS;
200 		goto bad;
201 	}
202 
203 	if (cnp->cn_namelen == 1 && *pname == '.') {
204 		*vpp = dvp;
205 		VREF(dvp);
206 		return (0);
207 	}
208 
209 	if (VTOFDESC(dvp)->fd_type != Froot) {
210 		error = ENOTDIR;
211 		goto bad;
212 	}
213 
214 	fd = 0;
215 	/* the only time a leading 0 is acceptable is if it's "0" */
216 	if (*pname == '0' && nlen != 1) {
217 		error = ENOENT;
218 		goto bad;
219 	}
220 	while (nlen--) {
221 		if (*pname < '0' || *pname > '9') {
222 			error = ENOENT;
223 			goto bad;
224 		}
225 		fd = 10 * fd + *pname++ - '0';
226 	}
227 
228 	if ((error = fget(td, fd, &fp)) != 0)
229 		goto bad;
230 
231 	error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp, td);
232 	fdrop(fp, td);
233 	if (error)
234 		goto bad;
235 	VTOFDESC(fvp)->fd_fd = fd;
236 	if (fvp != dvp)
237 		vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY);
238 	*vpp = fvp;
239 	return (0);
240 
241 bad:
242 	*vpp = NULL;
243 	return (error);
244 }
245 
246 static int
247 fdesc_open(ap)
248 	struct vop_open_args /* {
249 		struct vnode *a_vp;
250 		int  a_mode;
251 		struct ucred *a_cred;
252 		struct thread *a_td;
253 	} */ *ap;
254 {
255 	struct vnode *vp = ap->a_vp;
256 
257 	if (VTOFDESC(vp)->fd_type == Froot)
258 		return (0);
259 
260 	/*
261 	 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
262 	 * descriptor being sought for duplication. The error return ensures
263 	 * that the vnode for this device will be released by vn_open. Open
264 	 * will detect this special error and take the actions in dupfdopen.
265 	 * Other callers of vn_open or VOP_OPEN will simply report the
266 	 * error.
267 	 */
268 	ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
269 	return (ENODEV);
270 }
271 
272 static int
273 fdesc_getattr(ap)
274 	struct vop_getattr_args /* {
275 		struct vnode *a_vp;
276 		struct vattr *a_vap;
277 		struct ucred *a_cred;
278 		struct thread *a_td;
279 	} */ *ap;
280 {
281 	struct vnode *vp = ap->a_vp;
282 	struct vattr *vap = ap->a_vap;
283 	struct file *fp;
284 	struct stat stb;
285 	u_int fd;
286 	int error = 0;
287 
288 	switch (VTOFDESC(vp)->fd_type) {
289 	case Froot:
290 		VATTR_NULL(vap);
291 
292 		vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
293 		vap->va_type = VDIR;
294 		vap->va_nlink = 2;
295 		vap->va_size = DEV_BSIZE;
296 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
297 		vap->va_uid = 0;
298 		vap->va_gid = 0;
299 		vap->va_blocksize = DEV_BSIZE;
300 		vap->va_atime.tv_sec = boottime.tv_sec;
301 		vap->va_atime.tv_nsec = 0;
302 		vap->va_mtime = vap->va_atime;
303 		vap->va_ctime = vap->va_mtime;
304 		vap->va_gen = 0;
305 		vap->va_flags = 0;
306 		vap->va_rdev = 0;
307 		vap->va_bytes = 0;
308 		break;
309 
310 	case Fdesc:
311 		fd = VTOFDESC(vp)->fd_fd;
312 
313 		if ((error = fget(ap->a_td, fd, &fp)) != 0)
314 			return (error);
315 
316 		bzero(&stb, sizeof(stb));
317 		error = fo_stat(fp, &stb, ap->a_td->td_ucred, ap->a_td);
318 		fdrop(fp, ap->a_td);
319 		if (error == 0) {
320 			VATTR_NULL(vap);
321 			vap->va_type = IFTOVT(stb.st_mode);
322 			vap->va_mode = stb.st_mode;
323 #define FDRX (VREAD|VEXEC)
324 			if (vap->va_type == VDIR)
325 				vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
326 #undef FDRX
327 			vap->va_nlink = 1;
328 			vap->va_flags = 0;
329 			vap->va_bytes = stb.st_blocks * stb.st_blksize;
330 			vap->va_fileid = VTOFDESC(vp)->fd_ix;
331 			vap->va_size = stb.st_size;
332 			vap->va_blocksize = stb.st_blksize;
333 			vap->va_rdev = stb.st_rdev;
334 
335 			/*
336 			 * If no time data is provided, use the current time.
337 			 */
338 			if (stb.st_atimespec.tv_sec == 0 &&
339 			    stb.st_atimespec.tv_nsec == 0)
340 				nanotime(&stb.st_atimespec);
341 
342 			if (stb.st_ctimespec.tv_sec == 0 &&
343 			    stb.st_ctimespec.tv_nsec == 0)
344 				nanotime(&stb.st_ctimespec);
345 
346 			if (stb.st_mtimespec.tv_sec == 0 &&
347 			    stb.st_mtimespec.tv_nsec == 0)
348 				nanotime(&stb.st_mtimespec);
349 
350 			vap->va_atime = stb.st_atimespec;
351 			vap->va_mtime = stb.st_mtimespec;
352 			vap->va_ctime = stb.st_ctimespec;
353 			vap->va_uid = stb.st_uid;
354 			vap->va_gid = stb.st_gid;
355 		}
356 		break;
357 
358 	default:
359 		panic("fdesc_getattr");
360 		break;
361 	}
362 
363 	if (error == 0)
364 		vp->v_type = vap->va_type;
365 	return (error);
366 }
367 
368 static int
369 fdesc_setattr(ap)
370 	struct vop_setattr_args /* {
371 		struct vnode *a_vp;
372 		struct vattr *a_vap;
373 		struct ucred *a_cred;
374 		struct thread *a_td;
375 	} */ *ap;
376 {
377 	struct vattr *vap = ap->a_vap;
378 	struct vnode *vp;
379 	struct mount *mp;
380 	struct file *fp;
381 	unsigned fd;
382 	int error;
383 
384 	/*
385 	 * Can't mess with the root vnode
386 	 */
387 	if (VTOFDESC(ap->a_vp)->fd_type == Froot)
388 		return (EACCES);
389 
390 	fd = VTOFDESC(ap->a_vp)->fd_fd;
391 
392 	/*
393 	 * Allow setattr where there is an underlying vnode.
394 	 */
395 	error = getvnode(ap->a_td->td_proc->p_fd, fd, &fp);
396 	if (error) {
397 		/*
398 		 * getvnode() returns EINVAL if the file descriptor is not
399 		 * backed by a vnode.  Silently drop all changes except
400 		 * chflags(2) in this case.
401 		 */
402 		if (error == EINVAL) {
403 			if (vap->va_flags != VNOVAL)
404 				error = EOPNOTSUPP;
405 			else
406 				error = 0;
407 		}
408 		return (error);
409 	}
410 	vp = fp->f_vnode;
411 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) {
412 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
413 		error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_td);
414 		VOP_UNLOCK(vp, 0);
415 		vn_finished_write(mp);
416 	}
417 	fdrop(fp, ap->a_td);
418 	return (error);
419 }
420 
421 #define UIO_MX 16
422 
423 static int
424 fdesc_readdir(ap)
425 	struct vop_readdir_args /* {
426 		struct vnode *a_vp;
427 		struct uio *a_uio;
428 		struct ucred *a_cred;
429 		int *a_eofflag;
430 		u_long *a_cookies;
431 		int a_ncookies;
432 	} */ *ap;
433 {
434 	struct uio *uio = ap->a_uio;
435 	struct filedesc *fdp;
436 	struct dirent d;
437 	struct dirent *dp = &d;
438 	int error, i, off, fcnt;
439 
440 	/*
441 	 * We don't allow exporting fdesc mounts, and currently local
442 	 * requests do not need cookies.
443 	 */
444 	if (ap->a_ncookies)
445 		panic("fdesc_readdir: not hungry");
446 
447 	if (VTOFDESC(ap->a_vp)->fd_type != Froot)
448 		panic("fdesc_readdir: not dir");
449 
450 	off = (int)uio->uio_offset;
451 	if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
452 	    uio->uio_resid < UIO_MX)
453 		return (EINVAL);
454 	i = (u_int)off / UIO_MX;
455 	fdp = uio->uio_td->td_proc->p_fd;
456 	error = 0;
457 
458 	fcnt = i - 2;		/* The first two nodes are `.' and `..' */
459 
460 	FILEDESC_SLOCK(fdp);
461 	while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
462 		switch (i) {
463 		case 0:	/* `.' */
464 		case 1: /* `..' */
465 			bzero((caddr_t)dp, UIO_MX);
466 
467 			dp->d_fileno = i + FD_ROOT;
468 			dp->d_namlen = i + 1;
469 			dp->d_reclen = UIO_MX;
470 			bcopy("..", dp->d_name, dp->d_namlen);
471 			dp->d_name[i + 1] = '\0';
472 			dp->d_type = DT_DIR;
473 			break;
474 		default:
475 			if (fdp->fd_ofiles[fcnt] == NULL) {
476 				FILEDESC_SUNLOCK(fdp);
477 				goto done;
478 			}
479 
480 			bzero((caddr_t) dp, UIO_MX);
481 			dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
482 			dp->d_reclen = UIO_MX;
483 			dp->d_type = DT_UNKNOWN;
484 			dp->d_fileno = i + FD_DESC;
485 			break;
486 		}
487 		/*
488 		 * And ship to userland
489 		 */
490 		FILEDESC_SUNLOCK(fdp);
491 		error = uiomove(dp, UIO_MX, uio);
492 		if (error)
493 			goto done;
494 		FILEDESC_SLOCK(fdp);
495 		i++;
496 		fcnt++;
497 	}
498 	FILEDESC_SUNLOCK(fdp);
499 
500 done:
501 	uio->uio_offset = i * UIO_MX;
502 	return (error);
503 }
504 
505 static int
506 fdesc_inactive(ap)
507 	struct vop_inactive_args /* {
508 		struct vnode *a_vp;
509 		struct thread *a_td;
510 	} */ *ap;
511 {
512 	struct vnode *vp = ap->a_vp;
513 
514 	/*
515 	 * Clear out the v_type field to avoid
516 	 * nasty things happening in vgone().
517 	 */
518 	vp->v_type = VNON;
519 	return (0);
520 }
521 
522 static int
523 fdesc_reclaim(ap)
524 	struct vop_reclaim_args /* {
525 		struct vnode *a_vp;
526 	} */ *ap;
527 {
528 	struct vnode *vp = ap->a_vp;
529 	struct fdescnode *fd = VTOFDESC(vp);
530 
531 	LIST_REMOVE(fd, fd_hash);
532 	FREE(vp->v_data, M_TEMP);
533 	vp->v_data = 0;
534 
535 	return (0);
536 }
537