Home | History | Annotate | Line # | Download | only in genfs
layer_subr.c revision 1.7.2.2
      1 /*	$NetBSD: layer_subr.c,v 1.7.2.2 2001/11/14 19:17:09 nathanw Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1999 National Aeronautics & Space Administration
      5  * All rights reserved.
      6  *
      7  * This software was written by William Studenmund of the
      8  * Numerical Aerospace Simulation Facility, NASA Ames Research Center.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  * 3. Neither the name of the National Aeronautics & Space Administration
     19  *    nor the names of its contributors may be used to endorse or promote
     20  *    products derived from this software without specific prior written
     21  *    permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE NATIONAL AERONAUTICS & SPACE ADMINISTRATION
     24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE ADMINISTRATION OR CONTRIB-
     27  * UTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
     28  * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     33  * POSSIBILITY OF SUCH DAMAGE.
     34  */
     35 /*
     36  * Copyright (c) 1992, 1993
     37  *	The Regents of the University of California.  All rights reserved.
     38  *
     39  * This code is derived from software donated to Berkeley by
     40  * Jan-Simon Pendry.
     41  *
     42  * Redistribution and use in source and binary forms, with or without
     43  * modification, are permitted provided that the following conditions
     44  * are met:
     45  * 1. Redistributions of source code must retain the above copyright
     46  *    notice, this list of conditions and the following disclaimer.
     47  * 2. Redistributions in binary form must reproduce the above copyright
     48  *    notice, this list of conditions and the following disclaimer in the
     49  *    documentation and/or other materials provided with the distribution.
     50  * 3. All advertising materials mentioning features or use of this software
     51  *    must display the following acknowledgement:
     52  *	This product includes software developed by the University of
     53  *	California, Berkeley and its contributors.
     54  * 4. Neither the name of the University nor the names of its contributors
     55  *    may be used to endorse or promote products derived from this software
     56  *    without specific prior written permission.
     57  *
     58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     68  * SUCH DAMAGE.
     69  *
     70  *	from: Id: lofs_subr.c,v 1.11 1992/05/30 10:05:43 jsp Exp
     71  *	@(#)null_subr.c	8.7 (Berkeley) 5/14/95
     72  */
     73 
     74 #include <sys/cdefs.h>
     75 __KERNEL_RCSID(0, "$NetBSD: layer_subr.c,v 1.7.2.2 2001/11/14 19:17:09 nathanw Exp $");
     76 
     77 #include <sys/param.h>
     78 #include <sys/systm.h>
     79 #include <sys/proc.h>
     80 #include <sys/time.h>
     81 #include <sys/types.h>
     82 #include <sys/vnode.h>
     83 #include <sys/mount.h>
     84 #include <sys/namei.h>
     85 #include <sys/malloc.h>
     86 #include <miscfs/specfs/specdev.h>
     87 #include <miscfs/genfs/layer.h>
     88 #include <miscfs/genfs/layer_extern.h>
     89 
     90 #define	NLAYERNODECACHE 16
     91 
     92 /*
     93  * layer cache:
     94  * Each cache entry holds a reference to the lower vnode
     95  * along with a pointer to the alias vnode.  When an
     96  * entry is added the lower vnode is VREF'd.  When the
     97  * alias is removed the lower vnode is vrele'd.
     98  */
     99 
    100 /*
    101  * Initialise cache headers
    102  */
    103 void
    104 layerfs_init()
    105 {
    106 #ifdef LAYERFS_DIAGNOSTIC
    107 	printf("layerfs_init\n");		/* printed during system boot */
    108 #endif
    109 }
    110 
    111 /*
    112  * Free global resources of layerfs.
    113  */
    114 void
    115 layerfs_done()
    116 {
    117 #ifdef LAYERFS_DIAGNOSTIC
    118 	printf("layerfs_done\n");		/* printed on layerfs detach */
    119 #endif
    120 }
    121 
    122 /*
    123  * Return a locked, VREF'ed alias for lower vnode if already exists, else 0.
    124  */
    125 struct vnode *
    126 layer_node_find(mp, lowervp)
    127 	struct mount *mp;
    128 	struct vnode *lowervp;
    129 {
    130 	struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
    131 	struct layer_node_hashhead *hd;
    132 	struct layer_node *a;
    133 	struct vnode *vp;
    134 
    135 	/*
    136 	 * Find hash base, and then search the (two-way) linked
    137 	 * list looking for a layer_node structure which is referencing
    138 	 * the lower vnode.  If found, the increment the layer_node
    139 	 * reference count (but NOT the lower vnode's VREF counter)
    140 	 * and return the vnode locked.
    141 	 */
    142 	hd = LAYER_NHASH(lmp, lowervp);
    143 loop:
    144 	simple_lock(&lmp->layerm_hashlock);
    145 	for (a = hd->lh_first; a != 0; a = a->layer_hash.le_next) {
    146 		if (a->layer_lowervp == lowervp && LAYERTOV(a)->v_mount == mp) {
    147 			vp = LAYERTOV(a);
    148 			simple_unlock(&lmp->layerm_hashlock);
    149 			/*
    150 			 * We must be careful here as the fact the lower
    151 			 * vnode is locked will imply vp is locked unless
    152 			 * someone has decided to start vclean'ing either
    153 			 * vp or lowervp.
    154 			 *
    155 			 * So we try for an exclusive, recursive lock
    156 			 * on the upper vnode. If it fails, vcleaning
    157 			 * is in progress (so when we try again, we'll
    158 			 * fail). If it succeeds, we now have double
    159 			 * locked the bottom node. So we do an explicit
    160 			 * VOP_UNLOCK on it to keep the counts right. Note
    161 			 * that we will end up with the upper node and
    162 			 * the lower node locked once.
    163 			 */
    164 			if (vget(vp, LK_EXCLUSIVE | LK_CANRECURSE)) {
    165 				printf ("layer_node_find: vget failed.\n");
    166 				goto loop;
    167 			};
    168 			VOP_UNLOCK(lowervp, 0);
    169 			return (vp);
    170 		}
    171 	}
    172 
    173 	simple_unlock(&lmp->layerm_hashlock);
    174 	return NULL;
    175 }
    176 
    177 
    178 /*
    179  * Make a new layer_node node.
    180  * Vp is the alias vnode, lowervp is the lower vnode.
    181  * Maintain a reference to lowervp.
    182  */
    183 int
    184 layer_node_alloc(mp, lowervp, vpp)
    185 	struct mount *mp;
    186 	struct vnode *lowervp;
    187 	struct vnode **vpp;
    188 {
    189 	struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
    190 	struct layer_node_hashhead *hd;
    191 	struct layer_node *xp;
    192 	struct vnode *vp, *nvp;
    193 	int error;
    194 	extern int (**dead_vnodeop_p) __P((void *));
    195 
    196 	if ((error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p,
    197 			&vp)) != 0)
    198 		return (error);
    199 	vp->v_type = lowervp->v_type;
    200 	vp->v_flag |= VLAYER;
    201 
    202 	MALLOC(xp, struct layer_node *, lmp->layerm_size, M_TEMP, M_WAITOK);
    203 	if (vp->v_type == VBLK || vp->v_type == VCHR) {
    204 		MALLOC(vp->v_specinfo, struct specinfo *,
    205 		    sizeof(struct specinfo), M_VNODE, M_WAITOK);
    206 		vp->v_hashchain = NULL;
    207 		vp->v_rdev = lowervp->v_rdev;
    208 	}
    209 
    210 	vp->v_data = xp;
    211 	xp->layer_vnode = vp;
    212 	xp->layer_lowervp = lowervp;
    213 	xp->layer_flags = 0;
    214 	/*
    215 	 * Before we insert our new node onto the hash chains,
    216 	 * check to see if someone else has beaten us to it.
    217 	 * (We could have slept in MALLOC.)
    218 	 */
    219 	if ((nvp = layer_node_find(mp, lowervp)) != NULL) {
    220 		*vpp = nvp;
    221 
    222 		/* free the substructures we've allocated. */
    223 		FREE(xp, M_TEMP);
    224 		if (vp->v_type == VBLK || vp->v_type == VCHR)
    225 			FREE(vp->v_specinfo, M_VNODE);
    226 
    227 		vp->v_type = VBAD;		/* node is discarded */
    228 		vp->v_op = dead_vnodeop_p;	/* so ops will still work */
    229 		vrele(vp);			/* get rid of it. */
    230 		return (0);
    231 	}
    232 
    233 	simple_lock(&lmp->layerm_hashlock);
    234 
    235 	/*
    236 	 * Now lock the new node. We rely on the fact that we were passed
    237 	 * a locked vnode. If the lower node is exporting a struct lock
    238 	 * (v_vnlock != NULL) then we just set the upper v_vnlock to the
    239 	 * lower one, and both are now locked. If the lower node is exporting
    240 	 * NULL, then we copy that up and manually lock the upper node.
    241 	 *
    242 	 * LAYERFS_UPPERLOCK already has the test, so we use it after copying
    243 	 * up the v_vnlock from below.
    244 	 */
    245 
    246 	vp->v_vnlock = lowervp->v_vnlock;
    247 	LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error);
    248 
    249 	if (error) {
    250 		/*
    251 		 * How did we get a locking error? The node just came off
    252 		 * of the free list, and we're the only routine which
    253 		 * knows it's there...
    254 		 */
    255 		vp->v_vnlock = &vp->v_lock;
    256 		*vpp = NULL;
    257 
    258 		/* free the substructures we've allocated. */
    259 		FREE(xp, M_TEMP);
    260 		if (vp->v_type == VBLK || vp->v_type == VCHR)
    261 			FREE(vp->v_specinfo, M_VNODE);
    262 
    263 		vp->v_type = VBAD;		/* node is discarded */
    264 		vp->v_op = dead_vnodeop_p;	/* so ops will still work */
    265 		vrele(vp);			/* get rid of it. */
    266 		return (error);
    267 	}
    268 	/*
    269 	 * NetBSD used to do an inlined checkalias here. We do not, as
    270 	 * we never flag device nodes as being aliased. The lowervp
    271 	 * node will, when appropriate, be flaged as an alias.
    272 	 */
    273 
    274 	*vpp = vp;
    275 	VREF(lowervp);	/* Take into account reference held in layer_node */
    276 	hd = LAYER_NHASH(lmp, lowervp);
    277 	LIST_INSERT_HEAD(hd, xp, layer_hash);
    278 	uvm_vnp_setsize(vp, 0);
    279 	simple_unlock(&lmp->layerm_hashlock);
    280 	return (0);
    281 }
    282 
    283 
    284 /*
    285  * Try to find an existing layer_node vnode refering
    286  * to it, otherwise make a new layer_node vnode which
    287  * contains a reference to the lower vnode.
    288  *
    289  * >>> we assume that the lower node is already locked upon entry, so we
    290  * propagate the lock state to upper node <<
    291  */
    292 int
    293 layer_node_create(mp, lowervp, newvpp)
    294 	struct mount *mp;
    295 	struct vnode *lowervp;
    296 	struct vnode **newvpp;
    297 {
    298 	struct vnode *aliasvp;
    299 	struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
    300 
    301 	if ((aliasvp = layer_node_find(mp, lowervp)) != NULL) {
    302 		/*
    303 		 * layer_node_find has taken another reference
    304 		 * to the alias vnode and moved the lock holding to
    305 		 * aliasvp
    306 		 */
    307 #ifdef LAYERFS_DIAGNOSTIC
    308 		vprint("layer_node_create: exists", aliasvp);
    309 #endif
    310 	} else {
    311 		int error;
    312 
    313 		/*
    314 		 * Get new vnode.
    315 		 */
    316 #ifdef LAYERFS_DIAGNOSTIC
    317 		printf("layer_node_create: create new alias vnode\n");
    318 #endif
    319 
    320 		/*
    321 		 * Make new vnode reference the layer_node.
    322 		 */
    323 		if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0)
    324 			return error;
    325 
    326 		/*
    327 		 * aliasvp is already VREF'd by getnewvnode()
    328 		 */
    329 	}
    330 
    331 	/*
    332 	 * Now that we have VREF'd the upper vnode, release the reference
    333 	 * to the lower node. The existance of the layer_node retains one
    334 	 * reference to the lower node.
    335 	 */
    336 	vrele(lowervp);
    337 
    338 #ifdef DIAGNOSTIC
    339 	if (lowervp->v_usecount < 1) {
    340 		/* Should never happen... */
    341 		vprint("layer_node_create: alias", aliasvp);
    342 		vprint("layer_node_create: lower", lowervp);
    343 		panic("layer_node_create: lower has 0 usecount.");
    344 	};
    345 #endif
    346 
    347 #ifdef LAYERFS_DIAGNOSTIC
    348 	vprint("layer_node_create: alias", aliasvp);
    349 #endif
    350 	*newvpp = aliasvp;
    351 	return (0);
    352 }
    353 
    354 struct vnode *
    355 layer_checkvp(vp, fil, lno)
    356 	struct vnode *vp;
    357 	char *fil;
    358 	int lno;
    359 {
    360 	struct layer_node *a = VTOLAYER(vp);
    361 #ifdef notyet
    362 	/*
    363 	 * Can't do this check because vop_reclaim runs
    364 	 * with a funny vop vector.
    365 	 *
    366 	 * WRS - no it doesnt...
    367 	 */
    368 	if (vp->v_op != layer_vnodeop_p) {
    369 		printf ("layer_checkvp: on non-layer-node\n");
    370 #ifdef notyet
    371 		while (layer_checkvp_barrier) /*WAIT*/ ;
    372 #endif
    373 		panic("layer_checkvp");
    374 	};
    375 #endif
    376 	if (a->layer_lowervp == NULL) {
    377 		/* Should never happen */
    378 		int i; u_long *p;
    379 		printf("vp = %p, ZERO ptr\n", vp);
    380 		for (p = (u_long *) a, i = 0; i < 8; i++)
    381 			printf(" %lx", p[i]);
    382 		printf("\n");
    383 		/* wait for debugger */
    384 		panic("layer_checkvp");
    385 	}
    386 	if (a->layer_lowervp->v_usecount < 1) {
    387 		int i; u_long *p;
    388 		printf("vp = %p, unref'ed lowervp\n", vp);
    389 		for (p = (u_long *) a, i = 0; i < 8; i++)
    390 			printf(" %lx", p[i]);
    391 		printf("\n");
    392 		/* wait for debugger */
    393 		panic ("layer with unref'ed lowervp");
    394 	};
    395 #ifdef notnow
    396 	printf("layer %p/%d -> %p/%d [%s, %d]\n",
    397 	        LAYERTOV(a), LAYERTOV(a)->v_usecount,
    398 		a->layer_lowervp, a->layer_lowervp->v_usecount,
    399 		fil, lno);
    400 #endif
    401 	return a->layer_lowervp;
    402 }
    403