Home | History | Annotate | Line # | Download | only in lfs
lfs_subr.c revision 1.22
      1 /*	$NetBSD: lfs_subr.c,v 1.22 2002/05/14 20:03:54 perseant Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1999, 2000 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Konrad E. Schroder <perseant (at) hhhh.org>.
      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. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *      This product includes software developed by the NetBSD
     21  *      Foundation, Inc. and its contributors.
     22  * 4. Neither the name of The NetBSD Foundation nor the names of its
     23  *    contributors may be used to endorse or promote products derived
     24  *    from this software without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     36  * POSSIBILITY OF SUCH DAMAGE.
     37  */
     38 /*
     39  * Copyright (c) 1991, 1993
     40  *	The Regents of the University of California.  All rights reserved.
     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  *	@(#)lfs_subr.c	8.4 (Berkeley) 5/8/95
     71  */
     72 
     73 #include <sys/cdefs.h>
     74 __KERNEL_RCSID(0, "$NetBSD: lfs_subr.c,v 1.22 2002/05/14 20:03:54 perseant Exp $");
     75 
     76 #include <sys/param.h>
     77 #include <sys/systm.h>
     78 #include <sys/namei.h>
     79 #include <sys/vnode.h>
     80 #include <sys/buf.h>
     81 #include <sys/mount.h>
     82 #include <sys/malloc.h>
     83 #include <sys/proc.h>
     84 
     85 #include <ufs/ufs/inode.h>
     86 #include <ufs/lfs/lfs.h>
     87 #include <ufs/lfs/lfs_extern.h>
     88 
     89 /*
     90  * Return buffer with the contents of block "offset" from the beginning of
     91  * directory "ip".  If "res" is non-zero, fill it in with a pointer to the
     92  * remaining space in the directory.
     93  */
     94 int
     95 lfs_blkatoff(void *v)
     96 {
     97 	struct vop_blkatoff_args /* {
     98 		struct vnode *a_vp;
     99 		off_t a_offset;
    100 		char **a_res;
    101 		struct buf **a_bpp;
    102 		} */ *ap = v;
    103 	struct lfs *fs;
    104 	struct inode *ip;
    105 	struct buf *bp;
    106 	ufs_daddr_t lbn;
    107 	int bsize, error;
    108 
    109 	ip = VTOI(ap->a_vp);
    110 	fs = ip->i_lfs;
    111 	lbn = lblkno(fs, ap->a_offset);
    112 	bsize = blksize(fs, ip, lbn);
    113 
    114 	*ap->a_bpp = NULL;
    115 	if ((error = bread(ap->a_vp, lbn, bsize, NOCRED, &bp)) != 0) {
    116 		brelse(bp);
    117 		return (error);
    118 	}
    119 	if (ap->a_res)
    120 		*ap->a_res = (char *)bp->b_data + blkoff(fs, ap->a_offset);
    121 	*ap->a_bpp = bp;
    122 	return (0);
    123 }
    124 
    125 
    126 /*
    127  * lfs_seglock --
    128  *	Single thread the segment writer.
    129  */
    130 void
    131 lfs_seglock(struct lfs *fs, unsigned long flags)
    132 {
    133 	struct segment *sp;
    134 	int s;
    135 
    136 	if (fs->lfs_seglock) {
    137 		if (fs->lfs_lockpid == curproc->p_pid) {
    138 			++fs->lfs_seglock;
    139 			fs->lfs_sp->seg_flags |= flags;
    140 			return;
    141 		} else while (fs->lfs_seglock)
    142 			(void)tsleep(&fs->lfs_seglock, PRIBIO + 1,
    143 				     "lfs seglock", 0);
    144 	}
    145 
    146 	fs->lfs_seglock = 1;
    147 	fs->lfs_lockpid = curproc->p_pid;
    148 
    149 	sp = fs->lfs_sp = malloc(sizeof(struct segment), M_SEGMENT, M_WAITOK);
    150 	sp->bpp = malloc(((fs->lfs_sumsize - SEGSUM_SIZE(fs)) /
    151 			  sizeof(ufs_daddr_t) + 1) * sizeof(struct buf *),
    152 			 M_SEGMENT, M_WAITOK);
    153 	sp->seg_flags = flags;
    154 	sp->vp = NULL;
    155 	(void) lfs_initseg(fs);
    156 
    157 	/*
    158 	 * Keep a cumulative count of the outstanding I/O operations.  If the
    159 	 * disk drive catches up with us it could go to zero before we finish,
    160 	 * so we artificially increment it by one until we've scheduled all of
    161 	 * the writes we intend to do.
    162 	 */
    163 	s = splbio();
    164 	++fs->lfs_iocount;
    165 	splx(s);
    166 }
    167 
    168 /*
    169  * lfs_segunlock --
    170  *	Single thread the segment writer.
    171  */
    172 void
    173 lfs_segunlock(struct lfs *fs)
    174 {
    175 	struct segment *sp;
    176 	unsigned long sync, ckp;
    177 	int s;
    178 	struct buf *bp;
    179 	struct vnode *vp;
    180 	struct mount *mp;
    181 	extern int lfs_dirvcount;
    182 #ifdef LFS_MALLOC_SUMMARY
    183 	extern int locked_queue_count;
    184 	extern long locked_queue_bytes;
    185 #endif
    186 
    187 	sp = fs->lfs_sp;
    188 
    189 	if (fs->lfs_seglock == 1 && !(sp->seg_flags & SEGM_PROT)) {
    190 
    191 		mp = fs->lfs_ivnode->v_mount;
    192 		/*
    193 		 * Go through and unmark all DIROP vnodes, possibly
    194 		 * calling VOP_INACTIVE (through vrele).  This is
    195 		 * delayed until now in order not to accidentally
    196 		 * write a DIROP node through lfs_flush.
    197 		 */
    198 #ifndef LFS_NO_BACKVP_HACK
    199 	/* BEGIN HACK */
    200 #define	VN_OFFSET	(((caddr_t)&vp->v_mntvnodes.le_next) - (caddr_t)vp)
    201 #define	BACK_VP(VP)	((struct vnode *)(((caddr_t)VP->v_mntvnodes.le_prev) - VN_OFFSET))
    202 #define	BEG_OF_VLIST	((struct vnode *)(((caddr_t)&mp->mnt_vnodelist.lh_first) - VN_OFFSET))
    203 
    204 		/* Find last vnode. */
    205 	loop:	for (vp = mp->mnt_vnodelist.lh_first;
    206 		     vp && vp->v_mntvnodes.le_next != NULL;
    207 		     vp = vp->v_mntvnodes.le_next);
    208 		for (; vp && vp != BEG_OF_VLIST; vp = BACK_VP(vp)) {
    209 #else
    210 	loop:
    211 		 for (vp = mp->mnt_vnodelist.lh_first;
    212 		     vp != NULL;
    213 		     vp = vp->v_mntvnodes.le_next) {
    214 #endif
    215 			if (vp->v_mount != mp) {
    216 				printf("lfs_segunlock: starting over\n");
    217 				goto loop;
    218 			}
    219 			if (vp->v_type == VNON)
    220 				continue;
    221 			if (lfs_vref(vp))
    222 				continue;
    223 			if (VOP_ISLOCKED(vp) &&
    224                             vp->v_lock.lk_lockholder != curproc->p_pid) {
    225 				lfs_vunref(vp);
    226 				continue;
    227 			}
    228 			if ((vp->v_flag & VDIROP) &&
    229 			    !(VTOI(vp)->i_flag & IN_ADIROP)) {
    230 				--lfs_dirvcount;
    231 				vp->v_flag &= ~VDIROP;
    232 				wakeup(&lfs_dirvcount);
    233 				fs->lfs_unlockvp = vp;
    234 				lfs_vunref(vp);
    235 				vrele(vp);
    236 				fs->lfs_unlockvp = NULL;
    237 			} else {
    238 				lfs_vunref(vp);
    239 			}
    240 		}
    241 	}
    242 
    243 	if (fs->lfs_seglock == 1) {
    244 		sync = sp->seg_flags & SEGM_SYNC;
    245 		ckp = sp->seg_flags & SEGM_CKP;
    246 		if (sp->bpp != sp->cbpp) {
    247 			/* Free allocated segment summary */
    248 			fs->lfs_offset -= btofsb(fs, fs->lfs_sumsize);
    249 			bp = *sp->bpp;
    250 #ifdef LFS_MALLOC_SUMMARY
    251 			lfs_freebuf(bp);
    252 #else
    253 			s = splbio();
    254 			bremfree(bp);
    255 			splx(s);
    256 			bp->b_flags |= B_DONE|B_INVAL;
    257 			bp->b_flags &= ~B_DELWRI;
    258 			reassignbuf(bp,bp->b_vp);
    259 			brelse(bp);
    260 #endif
    261 		} else
    262 			printf ("unlock to 0 with no summary");
    263 
    264 		free(sp->bpp, M_SEGMENT);
    265 		sp->bpp = NULL;
    266 		free(sp, M_SEGMENT);
    267 		fs->lfs_sp = NULL;
    268 
    269 		/*
    270 		 * If the I/O count is non-zero, sleep until it reaches zero.
    271 		 * At the moment, the user's process hangs around so we can
    272 		 * sleep.
    273 		 */
    274 		s = splbio();
    275 		if (--fs->lfs_iocount < LFS_THROTTLE)
    276 			wakeup(&fs->lfs_iocount);
    277 		if(fs->lfs_iocount == 0) {
    278 			lfs_countlocked(&locked_queue_count,
    279 					&locked_queue_bytes, "lfs_segunlock");
    280 			wakeup(&locked_queue_count);
    281 			wakeup(&fs->lfs_iocount);
    282 		}
    283 		/*
    284 		 * We let checkpoints happen asynchronously.  That means
    285 		 * that during recovery, we have to roll forward between
    286 		 * the two segments described by the first and second
    287 		 * superblocks to make sure that the checkpoint described
    288 		 * by a superblock completed.
    289 		 */
    290 		while (sync && fs->lfs_iocount)
    291 			(void)tsleep(&fs->lfs_iocount, PRIBIO + 1,
    292 				     "lfs vflush", 0);
    293 		splx(s);
    294 		if (ckp) {
    295 			fs->lfs_nactive = 0;
    296 			/* If we *know* everything's on disk, write both sbs */
    297 			if (sync)
    298 				lfs_writesuper(fs,fs->lfs_sboffs[fs->lfs_activesb]);
    299 			fs->lfs_activesb = 1 - fs->lfs_activesb;
    300 			lfs_writesuper(fs,fs->lfs_sboffs[fs->lfs_activesb]);
    301 		}
    302 		--fs->lfs_seglock;
    303 		fs->lfs_lockpid = 0;
    304 		wakeup(&fs->lfs_seglock);
    305 	} else if (fs->lfs_seglock == 0) {
    306 		panic ("Seglock not held");
    307 	} else {
    308 		--fs->lfs_seglock;
    309 	}
    310 }
    311