xref: /titanic_50/usr/src/uts/common/fs/dev/sdev_vfsops.c (revision d51e90740114c60620c0febffd4d3ce6e280a107)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 /*
29  * This is the /dev (hence, the sdev_ prefix) filesystem.
30  */
31 
32 #include <sys/types.h>
33 #include <sys/param.h>
34 #include <sys/sysmacros.h>
35 #include <sys/systm.h>
36 #include <sys/kmem.h>
37 #include <sys/time.h>
38 #include <sys/pathname.h>
39 #include <sys/vfs.h>
40 #include <sys/vnode.h>
41 #include <sys/file.h>
42 #include <sys/stat.h>
43 #include <sys/uio.h>
44 #include <sys/stat.h>
45 #include <sys/errno.h>
46 #include <sys/cmn_err.h>
47 #include <sys/cred.h>
48 #include <sys/statvfs.h>
49 #include <sys/policy.h>
50 #include <sys/mount.h>
51 #include <sys/debug.h>
52 #include <sys/modctl.h>
53 #include <sys/mkdev.h>
54 #include <fs/fs_subr.h>
55 #include <sys/fs/sdev_impl.h>
56 #include <sys/fs/sdev_node.h>
57 #include <sys/fs/snode.h>
58 #include <sys/fs/dv_node.h>
59 #include <sys/sunndi.h>
60 #include <sys/mntent.h>
61 
62 /*
63  * /dev vfs operations.
64  */
65 
66 /*
67  * globals
68  */
69 struct sdev_data *sdev_origins; /* mount info for origins under /dev */
70 kmutex_t sdev_lock; /* used for mount/unmount/rename synchronization */
71 
72 /*
73  * static
74  */
75 static major_t devmajor;	/* the fictitious major we live on */
76 static major_t devminor;	/* the fictitious minor of this instance */
77 static struct sdev_data *sdev_mntinfo = NULL;	/* linked list of instances */
78 
79 /* LINTED E_STATIC_UNUSED */		/* useful for debugging */
80 static struct vnode *sdev_stale_attrvp; /* stale root attrvp after remount */
81 
82 static int sdev_mount(struct vfs *, struct vnode *, struct mounta *,
83     struct cred *);
84 static int sdev_unmount(struct vfs *, int, struct cred *);
85 static int sdev_root(struct vfs *, struct vnode **);
86 static int sdev_statvfs(struct vfs *, struct statvfs64 *);
87 static void sdev_insert_mntinfo(struct sdev_data *);
88 static int devinit(int, char *);
89 
90 static vfsdef_t sdev_vfssw = {
91 	VFSDEF_VERSION,
92 	"dev",		/* type name string */
93 	devinit,	/* init routine */
94 	VSW_CANREMOUNT,	/* flags */
95 	NULL		/* mount options table prototype */
96 };
97 
98 
99 /*
100  * Module linkage information
101  */
102 static struct modlfs modlfs = {
103 	&mod_fsops, "/dev filesystem %I%", &sdev_vfssw
104 };
105 
106 static struct modlinkage modlinkage = {
107 	MODREV_1, (void *)&modlfs, NULL
108 };
109 
110 int
111 _init(void)
112 {
113 	int e;
114 
115 	mutex_init(&sdev_lock, NULL, MUTEX_DEFAULT, NULL);
116 	sdev_node_cache_init();
117 	sdev_devfsadm_lockinit();
118 	if ((e = mod_install(&modlinkage)) != 0) {
119 		sdev_devfsadm_lockdestroy();
120 		sdev_node_cache_fini();
121 		mutex_destroy(&sdev_lock);
122 		return (e);
123 	}
124 	return (0);
125 }
126 
127 /*
128  * dev module remained loaded for the global /dev instance
129  */
130 int
131 _fini(void)
132 {
133 	return (EBUSY);
134 }
135 
136 int
137 _info(struct modinfo *modinfop)
138 {
139 	return (mod_info(&modlinkage, modinfop));
140 }
141 
142 /*ARGSUSED*/
143 static int
144 devinit(int fstype, char *name)
145 {
146 	static const fs_operation_def_t dev_vfsops_tbl[] = {
147 		VFSNAME_MOUNT, sdev_mount,	/* mount file system */
148 		VFSNAME_UNMOUNT, sdev_unmount,	/* unmount file system */
149 		VFSNAME_ROOT, sdev_root,	/* get root vnode */
150 		VFSNAME_STATVFS, sdev_statvfs,	/* get file system statistics */
151 		NULL, NULL
152 	};
153 
154 	int	error;
155 	extern major_t getudev(void);
156 
157 	devtype = fstype;
158 
159 	error = vfs_setfsops(fstype, dev_vfsops_tbl, NULL);
160 	if (error != 0) {
161 		cmn_err(CE_WARN, "devinit: bad vfs ops tbl");
162 		return (error);
163 	}
164 
165 	error = vn_make_ops("dev", sdev_vnodeops_tbl, &sdev_vnodeops);
166 	if (error != 0) {
167 		(void) vfs_freevfsops_by_type(fstype);
168 		cmn_err(CE_WARN, "devinit: bad vnode ops tbl");
169 		return (error);
170 	}
171 
172 	if ((devmajor = getudev()) == (major_t)-1) {
173 		cmn_err(CE_WARN, "%s: can't get unique dev", sdev_vfssw.name);
174 		return (1);
175 	}
176 
177 	/* initialize negative cache */
178 	sdev_ncache_init();
179 
180 	return (0);
181 }
182 
183 /*
184  * Both mount point and backing store directory name are
185  * passed in from userland
186  */
187 static int
188 sdev_mount(struct vfs *vfsp, struct vnode *mvp, struct mounta *uap,
189     struct cred *cr)
190 {
191 	struct sdev_data *sdev_data;
192 	struct vnode *avp;
193 	struct sdev_node *dv;
194 	struct sdev_mountargs *args = NULL;
195 	int	error = 0;
196 	dev_t	devdev;
197 
198 	/*
199 	 * security check
200 	 */
201 	if ((secpolicy_fs_mount(cr, mvp, vfsp) != 0) ||
202 	    (secpolicy_sys_devices(cr) != 0))
203 		return (EPERM);
204 
205 	/*
206 	 * Sanity check the mount point
207 	 */
208 	if (mvp->v_type != VDIR)
209 		return (ENOTDIR);
210 
211 	/*
212 	 * Sanity Check for overlay mount.
213 	 */
214 	mutex_enter(&mvp->v_lock);
215 	if ((uap->flags & MS_OVERLAY) == 0 &&
216 	    (uap->flags & MS_REMOUNT) == 0 &&
217 	    (mvp->v_count > 1 || (mvp->v_flag & VROOT))) {
218 		mutex_exit(&mvp->v_lock);
219 		return (EBUSY);
220 	}
221 	mutex_exit(&mvp->v_lock);
222 
223 	args = kmem_zalloc(sizeof (*args), KM_SLEEP);
224 
225 	if ((uap->flags & MS_DATA) &&
226 	    (uap->datalen != 0 && uap->dataptr != NULL)) {
227 		/* copy in the arguments */
228 		if (error = sdev_copyin_mountargs(uap, args))
229 			goto cleanup;
230 	}
231 
232 	/*
233 	 * Sanity check the backing store
234 	 */
235 	if (args->sdev_attrdir) {
236 		/* user supplied an attribute store */
237 		if (error = lookupname((char *)(uintptr_t)args->sdev_attrdir,
238 		    UIO_USERSPACE, FOLLOW, NULLVPP, &avp)) {
239 			cmn_err(CE_NOTE, "/dev fs: lookup on attribute "
240 			    "directory %s failed",
241 			    (char *)(uintptr_t)args->sdev_attrdir);
242 			goto cleanup;
243 		}
244 
245 		if (avp->v_type != VDIR) {
246 			VN_RELE(avp);
247 			error = ENOTDIR;
248 			goto cleanup;
249 		}
250 	} else {
251 		/* use mountp as the attribute store */
252 		avp = mvp;
253 		VN_HOLD(avp);
254 	}
255 
256 	mutex_enter(&sdev_lock);
257 
258 	/*
259 	 * handling installation
260 	 */
261 	if (uap->flags & MS_REMOUNT) {
262 		sdev_data = (struct sdev_data *)vfsp->vfs_data;
263 		ASSERT(sdev_data);
264 
265 		dv = sdev_data->sdev_root;
266 		ASSERT(dv == dv->sdev_dotdot);
267 
268 		/*
269 		 * mark all existing sdev_nodes (except root node) stale
270 		 */
271 		sdev_stale(dv);
272 
273 		/* Reset previous mountargs */
274 		if (sdev_data->sdev_mountargs) {
275 			kmem_free(sdev_data->sdev_mountargs,
276 			    sizeof (struct sdev_mountargs));
277 		}
278 		sdev_data->sdev_mountargs = args;
279 		args = NULL;		/* so it won't be freed below */
280 
281 		sdev_stale_attrvp = dv->sdev_attrvp;
282 		dv->sdev_attrvp = avp;
283 		vfsp->vfs_mtime = ddi_get_time();
284 
285 		mutex_exit(&sdev_lock);
286 		goto cleanup;				/* we're done */
287 	}
288 
289 	/*
290 	 * Create and initialize the vfs-private data.
291 	 */
292 	devdev = makedevice(devmajor, devminor);
293 	while (vfs_devismounted(devdev)) {
294 		devminor = (devminor + 1) & MAXMIN32;
295 
296 		/*
297 		 * All the minor numbers are used up.
298 		 */
299 		if (devminor == 0) {
300 			mutex_exit(&sdev_lock);
301 			VN_RELE(avp);
302 			error = ENODEV;
303 			goto cleanup;
304 		}
305 
306 		devdev = makedevice(devmajor, devminor);
307 	}
308 
309 	dv = sdev_mkroot(vfsp, devdev, mvp, avp, cr);
310 	sdev_data = kmem_zalloc(sizeof (struct sdev_data), KM_SLEEP);
311 	vfsp->vfs_dev = devdev;
312 	vfsp->vfs_data = (caddr_t)sdev_data;
313 	vfsp->vfs_fstype = devtype;
314 	vfsp->vfs_bsize = DEV_BSIZE;
315 	vfsp->vfs_mtime = ddi_get_time();
316 	vfs_make_fsid(&vfsp->vfs_fsid, vfsp->vfs_dev, devtype);
317 
318 	ASSERT(dv == dv->sdev_dotdot);
319 
320 	sdev_data->sdev_vfsp = vfsp;
321 	sdev_data->sdev_root = dv;
322 	sdev_data->sdev_mountargs = args;
323 
324 	/* get acl flavor from attribute dir */
325 	if (VOP_PATHCONF(avp, _PC_ACL_ENABLED, &sdev_data->sdev_acl_flavor,
326 	    kcred) != 0 || sdev_data->sdev_acl_flavor == 0)
327 		sdev_data->sdev_acl_flavor = _ACL_ACLENT_ENABLED;
328 
329 	args = NULL;			/* so it won't be freed below */
330 	sdev_insert_mntinfo(sdev_data);
331 	mutex_exit(&sdev_lock);
332 
333 	if (!SDEV_IS_GLOBAL(dv)) {
334 		ASSERT(sdev_origins);
335 		dv->sdev_flags &= ~SDEV_GLOBAL;
336 		dv->sdev_origin = sdev_origins->sdev_root;
337 	} else {
338 		sdev_ncache_setup();
339 	}
340 
341 	sdev_update_timestamps(dv->sdev_attrvp,
342 		cr, AT_CTIME|AT_MTIME|AT_ATIME);
343 
344 cleanup:
345 	if (args)
346 		kmem_free(args, sizeof (*args));
347 	return (error);
348 }
349 
350 /*
351  * unmounting the non-global /dev instances, e.g. when deleting a Kevlar zone.
352  */
353 static int
354 sdev_unmount(struct vfs *vfsp, int flag, struct cred *cr)
355 {
356 	struct sdev_node *dv;
357 	int error;
358 	struct sdev_data *sdev_data, *prev, *next;
359 
360 	/*
361 	 * enforce the security policies
362 	 */
363 	if ((secpolicy_fs_unmount(cr, vfsp) != 0) ||
364 	    (secpolicy_sys_devices(cr) != 0))
365 		return (EPERM);
366 
367 	if (flag & MS_FORCE)
368 		return (ENOTSUP);
369 
370 	mutex_enter(&sdev_lock);
371 	dv = VFSTOSDEVFS(vfsp)->sdev_root;
372 	ASSERT(dv == dv->sdev_dotdot);
373 	if (SDEVTOV(dv)->v_count > 1) {
374 		mutex_exit(&sdev_lock);
375 		return (EBUSY);
376 	}
377 
378 	/*
379 	 * global instance remains mounted
380 	 */
381 	if (SDEV_IS_GLOBAL(dv)) {
382 		mutex_exit(&sdev_lock);
383 		return (EBUSY);
384 	}
385 	mutex_exit(&sdev_lock);
386 
387 	/* verify the v_count */
388 	if ((error = sdev_cleandir(dv, NULL, 0)) != 0) {
389 		return (error);
390 	}
391 	ASSERT(SDEVTOV(dv)->v_count == 1);
392 
393 	/* release hold on root node and destroy it */
394 	SDEV_RELE(dv);
395 	dv->sdev_nlink -= 2;
396 	sdev_nodedestroy(dv, 0);
397 
398 	sdev_data = (struct sdev_data *)vfsp->vfs_data;
399 	vfsp->vfs_data = (caddr_t)0;
400 
401 	/*
402 	 * XXX separate it into sdev_delete_mntinfo() if useful
403 	 */
404 	mutex_enter(&sdev_lock);
405 	prev = sdev_data->sdev_prev;
406 	next = sdev_data->sdev_next;
407 	if (prev)
408 		prev->sdev_next = next;
409 	else
410 		sdev_mntinfo = next;
411 	if (next)
412 		next->sdev_prev = prev;
413 	mutex_exit(&sdev_lock);
414 
415 	if (sdev_data->sdev_mountargs) {
416 		kmem_free(sdev_data->sdev_mountargs,
417 		    sizeof (struct sdev_mountargs));
418 	}
419 	kmem_free(sdev_data, sizeof (struct sdev_data));
420 	return (0);
421 }
422 
423 /*
424  * return root vnode for given vfs
425  */
426 static int
427 sdev_root(struct vfs *vfsp, struct vnode **vpp)
428 {
429 	*vpp = SDEVTOV(VFSTOSDEVFS(vfsp)->sdev_root);
430 	VN_HOLD(*vpp);
431 	return (0);
432 }
433 
434 /*
435  * return 'generic superblock' information to userland.
436  *
437  * not much that we can usefully admit to here
438  */
439 static int
440 sdev_statvfs(struct vfs *vfsp, struct statvfs64 *sbp)
441 {
442 	dev32_t d32;
443 
444 	bzero(sbp, sizeof (*sbp));
445 	sbp->f_frsize = sbp->f_bsize = vfsp->vfs_bsize;
446 	sbp->f_files = kmem_cache_stat(sdev_node_cache, "alloc");
447 
448 	/* no illusions that free/avail files is relevant to dev */
449 	sbp->f_ffree = 0;
450 	sbp->f_favail = 0;
451 
452 	/* no illusions that blocks are relevant to devfs */
453 	sbp->f_bfree = 0;
454 	sbp->f_bavail = 0;
455 	sbp->f_blocks = 0;
456 
457 	(void) cmpldev(&d32, vfsp->vfs_dev);
458 	sbp->f_fsid = d32;
459 	(void) strcpy(sbp->f_basetype, vfssw[devtype].vsw_name);
460 	sbp->f_flag = vf_to_stf(vfsp->vfs_flag);
461 	sbp->f_namemax = MAXNAMELEN - 1;
462 	(void) strcpy(sbp->f_fstr, "dev");
463 
464 	return (0);
465 }
466 
467 int
468 sdev_module_register(char *mod_name, struct devname_ops *dev_ops)
469 {
470 	struct devname_nsmap *map = NULL;
471 
472 	if (strcmp(mod_name, DEVNAME_NSCONFIG) == 0) {
473 		devname_ns_ops = dev_ops;
474 		return (0);
475 	}
476 
477 	map = sdev_get_nsmap_by_module(mod_name);
478 	if (map == NULL)
479 		return (EFAULT);
480 
481 	rw_enter(&map->dir_lock, RW_WRITER);
482 	map->dir_ops = dev_ops;
483 	rw_exit(&map->dir_lock);
484 	return (0);
485 }
486 
487 static void
488 sdev_insert_mntinfo(struct sdev_data *data)
489 {
490 	ASSERT(mutex_owned(&sdev_lock));
491 	data->sdev_next = sdev_mntinfo;
492 	data->sdev_prev = NULL;
493 	if (sdev_mntinfo) {
494 		sdev_mntinfo->sdev_prev = data;
495 	} else {
496 		sdev_origins = data;
497 	}
498 	sdev_mntinfo = data;
499 }
500 
501 struct sdev_data *
502 sdev_find_mntinfo(char *mntpt)
503 {
504 	struct sdev_data *mntinfo;
505 
506 	mutex_enter(&sdev_lock);
507 	mntinfo = sdev_mntinfo;
508 	while (mntinfo) {
509 		if (strcmp(mntpt, mntinfo->sdev_root->sdev_name) == 0) {
510 			SDEVTOV(mntinfo->sdev_root)->v_count++;
511 			break;
512 		}
513 		mntinfo = mntinfo->sdev_next;
514 	}
515 	mutex_exit(&sdev_lock);
516 	return (mntinfo);
517 }
518 
519 void
520 sdev_mntinfo_rele(struct sdev_data *mntinfo)
521 {
522 	mutex_enter(&sdev_lock);
523 	SDEVTOV(mntinfo->sdev_root)->v_count--;
524 	mutex_exit(&sdev_lock);
525 }
526