X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=fs%2Fjfs%2Fjfs_dmap.c;h=23546c8fd48bd433c67f4233879c494f1c5457e9;hb=97bf2856c6014879bd04983a3e9dfcdac1e7fe85;hp=d86e467c6e421628a2f845668b5da630581bdf7e;hpb=6a77f38946aaee1cd85eeec6cf4229b204c15071;p=linux-2.6.git diff --git a/fs/jfs/jfs_dmap.c b/fs/jfs/jfs_dmap.c index d86e467c6..23546c8fd 100644 --- a/fs/jfs/jfs_dmap.c +++ b/fs/jfs/jfs_dmap.c @@ -3,16 +3,16 @@ * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or + * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. - * + * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See * the GNU General Public License for more details. * * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software + * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ @@ -25,63 +25,33 @@ #include "jfs_metapage.h" #include "jfs_debug.h" -/* - * Debug code for double-checking block map - */ -/* #define _JFS_DEBUG_DMAP 1 */ - -#ifdef _JFS_DEBUG_DMAP -#define DBINITMAP(size,ipbmap,results) \ - DBinitmap(size,ipbmap,results) -#define DBALLOC(dbmap,mapsize,blkno,nblocks) \ - DBAlloc(dbmap,mapsize,blkno,nblocks) -#define DBFREE(dbmap,mapsize,blkno,nblocks) \ - DBFree(dbmap,mapsize,blkno,nblocks) -#define DBALLOCCK(dbmap,mapsize,blkno,nblocks) \ - DBAllocCK(dbmap,mapsize,blkno,nblocks) -#define DBFREECK(dbmap,mapsize,blkno,nblocks) \ - DBFreeCK(dbmap,mapsize,blkno,nblocks) - -static void DBinitmap(s64, struct inode *, u32 **); -static void DBAlloc(uint *, s64, s64, s64); -static void DBFree(uint *, s64, s64, s64); -static void DBAllocCK(uint *, s64, s64, s64); -static void DBFreeCK(uint *, s64, s64, s64); -#else -#define DBINITMAP(size,ipbmap,results) -#define DBALLOC(dbmap, mapsize, blkno, nblocks) -#define DBFREE(dbmap, mapsize, blkno, nblocks) -#define DBALLOCCK(dbmap, mapsize, blkno, nblocks) -#define DBFREECK(dbmap, mapsize, blkno, nblocks) -#endif /* _JFS_DEBUG_DMAP */ - /* * SERIALIZATION of the Block Allocation Map. * * the working state of the block allocation map is accessed in * two directions: - * + * * 1) allocation and free requests that start at the dmap * level and move up through the dmap control pages (i.e. * the vast majority of requests). - * - * 2) allocation requests that start at dmap control page + * + * 2) allocation requests that start at dmap control page * level and work down towards the dmaps. - * - * the serialization scheme used here is as follows. * - * requests which start at the bottom are serialized against each - * other through buffers and each requests holds onto its buffers - * as it works it way up from a single dmap to the required level + * the serialization scheme used here is as follows. + * + * requests which start at the bottom are serialized against each + * other through buffers and each requests holds onto its buffers + * as it works it way up from a single dmap to the required level * of dmap control page. * requests that start at the top are serialized against each other * and request that start from the bottom by the multiple read/single * write inode lock of the bmap inode. requests starting at the top * take this lock in write mode while request starting at the bottom * take the lock in read mode. a single top-down request may proceed - * exclusively while multiple bottoms-up requests may proceed - * simultaneously (under the protection of busy buffers). - * + * exclusively while multiple bottoms-up requests may proceed + * simultaneously (under the protection of busy buffers). + * * in addition to information found in dmaps and dmap control pages, * the working state of the block allocation map also includes read/ * write information maintained in the bmap descriptor (i.e. total @@ -89,14 +59,14 @@ static void DBFreeCK(uint *, s64, s64, s64); * a single exclusive lock (BMAP_LOCK) is used to guard this information * in the face of multiple-bottoms up requests. * (lock ordering: IREAD_LOCK, BMAP_LOCK); - * + * * accesses to the persistent state of the block allocation map (limited * to the persistent bitmaps in dmaps) is guarded by (busy) buffers. */ -#define BMAP_LOCK_INIT(bmp) init_MUTEX(&bmp->db_bmaplock) -#define BMAP_LOCK(bmp) down(&bmp->db_bmaplock) -#define BMAP_UNLOCK(bmp) up(&bmp->db_bmaplock) +#define BMAP_LOCK_INIT(bmp) mutex_init(&bmp->db_bmaplock) +#define BMAP_LOCK(bmp) mutex_lock(&bmp->db_bmaplock) +#define BMAP_UNLOCK(bmp) mutex_unlock(&bmp->db_bmaplock) /* * forward references @@ -104,8 +74,8 @@ static void DBFreeCK(uint *, s64, s64, s64); static void dbAllocBits(struct bmap * bmp, struct dmap * dp, s64 blkno, int nblocks); static void dbSplit(dmtree_t * tp, int leafno, int splitsz, int newval); -static void dbBackSplit(dmtree_t * tp, int leafno); -static void dbJoin(dmtree_t * tp, int leafno, int newval); +static int dbBackSplit(dmtree_t * tp, int leafno); +static int dbJoin(dmtree_t * tp, int leafno, int newval); static void dbAdjTree(dmtree_t * tp, int leafno, int newval); static int dbAdjCtl(struct bmap * bmp, s64 blkno, int newval, int alloc, int level); @@ -128,8 +98,8 @@ static int dbExtend(struct inode *ip, s64 blkno, s64 nblocks, s64 addnblocks); static int dbFindBits(u32 word, int l2nb); static int dbFindCtl(struct bmap * bmp, int l2nb, int level, s64 * blkno); static int dbFindLeaf(dmtree_t * tp, int l2nb, int *leafidx); -static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, - int nblocks); +static int dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, + int nblocks); static int dbFreeDmap(struct bmap * bmp, struct dmap * dp, s64 blkno, int nblocks); static int dbMaxBud(u8 * cp); @@ -150,12 +120,12 @@ static int dbGetL2AGSize(s64 nblocks); /* * buddy table * - * table used for determining buddy sizes within characters of + * table used for determining buddy sizes within characters of * dmap bitmap words. the characters themselves serve as indexes * into the table, with the table elements yielding the maximum * binary buddy of free bits within the character. */ -static s8 budtab[256] = { +static const s8 budtab[256] = { 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, @@ -176,7 +146,7 @@ static s8 budtab[256] = { /* - * NAME: dbMount() + * NAME: dbMount() * * FUNCTION: initializate the block allocation map. * @@ -242,7 +212,6 @@ int dbMount(struct inode *ipbmap) JFS_SBI(ipbmap->i_sb)->bmap = bmp; memset(bmp->db_active, 0, sizeof(bmp->db_active)); - DBINITMAP(bmp->db_mapsize, ipbmap, &bmp->db_DBmap); /* * allocate/initialize the bmap lock @@ -254,12 +223,12 @@ int dbMount(struct inode *ipbmap) /* - * NAME: dbUnmount() + * NAME: dbUnmount() * * FUNCTION: terminate the block allocation map in preparation for * file system unmount. * - * the in-core bmap descriptor is written to disk and + * the in-core bmap descriptor is written to disk and * the memory for this descriptor is freed. * * PARAMETERS: @@ -272,7 +241,6 @@ int dbMount(struct inode *ipbmap) int dbUnmount(struct inode *ipbmap, int mounterror) { struct bmap *bmp = JFS_SBI(ipbmap->i_sb)->bmap; - int i; if (!(mounterror || isReadOnly(ipbmap))) dbSync(ipbmap); @@ -282,14 +250,6 @@ int dbUnmount(struct inode *ipbmap, int mounterror) */ truncate_inode_pages(ipbmap->i_mapping, 0); - /* - * Sanity Check - */ - for (i = 0; i < bmp->db_numag; i++) - if (atomic_read(&bmp->db_active[i])) - printk(KERN_ERR "dbUnmount: db_active[%d] = %d\n", - i, atomic_read(&bmp->db_active[i])); - /* free the memory for the in-memory bmap. */ kfree(bmp); @@ -342,10 +302,8 @@ int dbSync(struct inode *ipbmap) /* * write out dirty pages of bmap */ - filemap_fdatawrite(ipbmap->i_mapping); - filemap_fdatawait(ipbmap->i_mapping); + filemap_write_and_wait(ipbmap->i_mapping); - ipbmap->i_state |= I_DIRTY; diWriteSpecial(ipbmap, 0); return (0); @@ -353,7 +311,7 @@ int dbSync(struct inode *ipbmap) /* - * NAME: dbFree() + * NAME: dbFree() * * FUNCTION: free the specified block range from the working block * allocation map. @@ -416,16 +374,13 @@ int dbFree(struct inode *ip, s64 blkno, s64 nblocks) */ nb = min(rem, BPERDMAP - (blkno & (BPERDMAP - 1))); - DBALLOCCK(bmp->db_DBmap, bmp->db_mapsize, blkno, nb); - /* free the blocks. */ if ((rc = dbFreeDmap(bmp, dp, blkno, nb))) { + jfs_error(ip->i_sb, "dbFree: error in block map\n"); release_metapage(mp); IREAD_UNLOCK(ipbmap); return (rc); } - - DBFREE(bmp->db_DBmap, bmp->db_mapsize, blkno, nb); } /* write the last buffer. */ @@ -442,14 +397,14 @@ int dbFree(struct inode *ip, s64 blkno, s64 nblocks) * * FUNCTION: update the allocation state (free or allocate) of the * specified block range in the persistent block allocation map. - * + * * the blocks will be updated in the persistent map one * dmap at a time. * * PARAMETERS: * ipbmap - pointer to in-core inode for the block map. - * free - TRUE if block range is to be freed from the persistent - * map; FALSE if it is to be allocated. + * free - 'true' if block range is to be freed from the persistent + * map; 'false' if it is to be allocated. * blkno - starting block number of the range. * nblocks - number of contiguous blocks in the range. * tblk - transaction block; @@ -471,6 +426,7 @@ dbUpdatePMap(struct inode *ipbmap, struct metapage *mp; struct jfs_log *log; int lsn, difft, diffp; + unsigned long flags; /* the blocks better be within the mapsize. */ if (blkno + nblocks > bmp->db_mapsize) { @@ -504,6 +460,7 @@ dbUpdatePMap(struct inode *ipbmap, 0); if (mp == NULL) return -EIO; + metapage_wait_for_io(mp); } dp = (struct dmap *) mp->data; @@ -518,7 +475,7 @@ dbUpdatePMap(struct inode *ipbmap, /* update the bits of the dmap words. the first and last * words may only have a subset of their bits updated. if * this is the case, we'll work against that word (i.e. - * partial first and/or last) only in a single pass. a + * partial first and/or last) only in a single pass. a * single pass will also be used to update all words that * are to have all their bits updated. */ @@ -575,6 +532,7 @@ dbUpdatePMap(struct inode *ipbmap, lastlblkno = lblkno; + LOGSYNC_LOCK(log, flags); if (mp->lsn != 0) { /* inherit older/smaller lsn */ logdiff(diffp, mp->lsn, log); @@ -582,31 +540,25 @@ dbUpdatePMap(struct inode *ipbmap, mp->lsn = lsn; /* move bp after tblock in logsync list */ - LOGSYNC_LOCK(log); list_move(&mp->synclist, &tblk->synclist); - LOGSYNC_UNLOCK(log); } /* inherit younger/larger clsn */ - LOGSYNC_LOCK(log); logdiff(difft, tblk->clsn, log); logdiff(diffp, mp->clsn, log); if (difft > diffp) mp->clsn = tblk->clsn; - LOGSYNC_UNLOCK(log); } else { mp->log = log; mp->lsn = lsn; /* insert bp after tblock in logsync list */ - LOGSYNC_LOCK(log); - log->count++; list_add(&mp->synclist, &tblk->synclist); mp->clsn = tblk->clsn; - LOGSYNC_UNLOCK(log); } + LOGSYNC_UNLOCK(log, flags); } /* write the last buffer. */ @@ -710,11 +662,11 @@ unlock: * the block allocation policy uses hints and a multi-step * approach. * - * for allocation requests smaller than the number of blocks + * for allocation requests smaller than the number of blocks * per dmap, we first try to allocate the new blocks * immediately following the hint. if these blocks are not * available, we try to allocate blocks near the hint. if - * no blocks near the hint are available, we next try to + * no blocks near the hint are available, we next try to * allocate within the same dmap as contains the hint. * * if no blocks are available in the dmap or the allocation @@ -761,7 +713,7 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) #endif /* _STILL_TO_PORT */ /* get the log2 number of blocks to be allocated. - * if the number of blocks is not a log2 multiple, + * if the number of blocks is not a log2 multiple, * it will be rounded up to the next log2 multiple. */ l2nb = BLKSTOL2(nblocks); @@ -784,10 +736,6 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) IWRITE_LOCK(ipbmap); rc = dbAllocAny(bmp, nblocks, l2nb, results); - if (rc == 0) { - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, *results, - nblocks); - } goto write_unlock; } @@ -845,8 +793,6 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) != -ENOSPC) { if (rc == 0) { *results = blkno; - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, - *results, nblocks); mark_metapage_dirty(mp); } @@ -872,11 +818,8 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) if ((rc = dbAllocNear(bmp, dp, blkno, (int) nblocks, l2nb, results)) != -ENOSPC) { - if (rc == 0) { - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, - *results, nblocks); + if (rc == 0) mark_metapage_dirty(mp); - } release_metapage(mp); goto read_unlock; @@ -887,11 +830,8 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) */ if ((rc = dbAllocDmapLev(bmp, dp, (int) nblocks, l2nb, results)) != -ENOSPC) { - if (rc == 0) { - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, - *results, nblocks); + if (rc == 0) mark_metapage_dirty(mp); - } release_metapage(mp); goto read_unlock; @@ -905,13 +845,9 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) * the same allocation group as the hint. */ IWRITE_LOCK(ipbmap); - if ((rc = dbAllocAG(bmp, agno, nblocks, l2nb, results)) - != -ENOSPC) { - if (rc == 0) - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, - *results, nblocks); + if ((rc = dbAllocAG(bmp, agno, nblocks, l2nb, results)) != -ENOSPC) goto write_unlock; - } + IWRITE_UNLOCK(ipbmap); @@ -927,9 +863,6 @@ int dbAlloc(struct inode *ip, s64 hint, s64 nblocks, s64 * results) */ if ((rc = dbAllocAG(bmp, agno, nblocks, l2nb, results)) == -ENOSPC) rc = dbAllocAny(bmp, nblocks, l2nb, results); - if (rc == 0) { - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, *results, nblocks); - } write_unlock: IWRITE_UNLOCK(ipbmap); @@ -973,7 +906,7 @@ int dbAllocExact(struct inode *ip, s64 blkno, int nblocks) * validate extent request: * * note: defragfs policy: - * max 64 blocks will be moved. + * max 64 blocks will be moved. * allocation request size must be satisfied from a single dmap. */ if (nblocks <= 0 || nblocks > BPERDMAP || blkno >= bmp->db_mapsize) { @@ -1001,10 +934,9 @@ int dbAllocExact(struct inode *ip, s64 blkno, int nblocks) IREAD_UNLOCK(ipbmap); - if (rc == 0) { - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, blkno, nblocks); + if (rc == 0) mark_metapage_dirty(mp); - } + release_metapage(mp); return (rc); @@ -1153,7 +1085,6 @@ static int dbExtend(struct inode *ip, s64 blkno, s64 nblocks, s64 addnblocks) return -EIO; } - DBALLOCCK(bmp->db_DBmap, bmp->db_mapsize, blkno, nblocks); dp = (struct dmap *) mp->data; /* try to allocate the blocks immediately following the @@ -1164,11 +1095,9 @@ static int dbExtend(struct inode *ip, s64 blkno, s64 nblocks, s64 addnblocks) IREAD_UNLOCK(ipbmap); /* were we successful ? */ - if (rc == 0) { - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, extblkno, - addnblocks); + if (rc == 0) write_metapage(mp); - } else + else /* we were not successful */ release_metapage(mp); @@ -1404,7 +1333,7 @@ dbAllocNear(struct bmap * bmp, * or two sub-trees, depending on the allocation group size. * we search the top nodes of these subtrees left to right for * sufficient free space. if sufficient free space is found, - * the subtree is searched to find the leftmost leaf that + * the subtree is searched to find the leftmost leaf that * has free space. once we have made it to the leaf, we * move the search to the next lower level dmap control page * corresponding to this leaf. we continue down the dmap control @@ -1469,7 +1398,7 @@ dbAllocAG(struct bmap * bmp, int agno, s64 nblocks, int l2nb, s64 * results) * that fully describes the allocation group since the allocation * group is already fully described by a dmap. in this case, we * just call dbAllocCtl() to search the dmap tree and allocate the - * required space if available. + * required space if available. * * if the allocation group is completely free, dbAllocCtl() is * also called to allocate the required space. this is done for @@ -1521,7 +1450,7 @@ dbAllocAG(struct bmap * bmp, int agno, s64 nblocks, int l2nb, s64 * results) (1 << (L2LPERCTL - (bmp->db_agheigth << 1))) / bmp->db_agwidth; ti = bmp->db_agstart + bmp->db_agwidth * (agno & (agperlev - 1)); - /* dmap control page trees fan-out by 4 and a single allocation + /* dmap control page trees fan-out by 4 and a single allocation * group may be described by 1 or 2 subtrees within the ag level * dmap control page, depending upon the ag size. examine the ag's * subtrees for sufficient free space, starting with the leftmost @@ -1704,7 +1633,7 @@ static int dbFindCtl(struct bmap * bmp, int l2nb, int level, s64 * blkno) /* starting at the specified dmap control page level and block * number, search down the dmap control levels for the starting - * block number of a dmap page that contains or starts off + * block number of a dmap page that contains or starts off * sufficient free blocks. */ for (lev = level, b = *blkno; lev >= 0; lev--) { @@ -1748,7 +1677,7 @@ static int dbFindCtl(struct bmap * bmp, int l2nb, int level, s64 * blkno) } /* adjust the block number to reflect the location within - * the dmap control page (i.e. the leaf) at which free + * the dmap control page (i.e. the leaf) at which free * space was found. */ b += (((s64) leafidx) << budmin); @@ -1771,12 +1700,12 @@ static int dbFindCtl(struct bmap * bmp, int l2nb, int level, s64 * blkno) * NAME: dbAllocCtl() * * FUNCTION: attempt to allocate a specified number of contiguous - * blocks starting within a specific dmap. - * + * blocks starting within a specific dmap. + * * this routine is called by higher level routines that search * the dmap control pages above the actual dmaps for contiguous * free space. the result of successful searches by these - * routines are the starting block numbers within dmaps, with + * routines are the starting block numbers within dmaps, with * the dmaps themselves containing the desired contiguous free * space or starting a contiguous free space of desired size * that is made up of the blocks of one or more dmaps. these @@ -1943,14 +1872,14 @@ dbAllocCtl(struct bmap * bmp, s64 nblocks, int l2nb, s64 blkno, s64 * results) * * FUNCTION: attempt to allocate a specified number of contiguous blocks * from a specified dmap. - * + * * this routine checks if the contiguous blocks are available. * if so, nblocks of blocks are allocated; otherwise, ENOSPC is * returned. * * PARAMETERS: * mp - pointer to bmap descriptor - * dp - pointer to dmap to attempt to allocate blocks from. + * dp - pointer to dmap to attempt to allocate blocks from. * l2nb - log2 number of contiguous block desired. * nblocks - actual number of contiguous block desired. * results - on successful return, set to the starting block number @@ -1961,7 +1890,7 @@ dbAllocCtl(struct bmap * bmp, s64 nblocks, int l2nb, s64 blkno, s64 * results) * -ENOSPC - insufficient disk resources * -EIO - i/o error * - * serialization: IREAD_LOCK(ipbmap), e.g., from dbAlloc(), or + * serialization: IREAD_LOCK(ipbmap), e.g., from dbAlloc(), or * IWRITE_LOCK(ipbmap), e.g., dbAllocCtl(), held on entry/exit; */ static int @@ -2087,7 +2016,7 @@ static int dbFreeDmap(struct bmap * bmp, struct dmap * dp, s64 blkno, int nblocks) { s8 oldroot; - int rc, word; + int rc = 0, word; /* save the current value of the root (i.e. maximum free string) * of the dmap tree. @@ -2095,15 +2024,15 @@ static int dbFreeDmap(struct bmap * bmp, struct dmap * dp, s64 blkno, oldroot = dp->tree.stree[ROOT]; /* free the specified (blocks) bits */ - dbFreeBits(bmp, dp, blkno, nblocks); + rc = dbFreeBits(bmp, dp, blkno, nblocks); - /* if the root has not changed, done. */ - if (dp->tree.stree[ROOT] == oldroot) - return (0); + /* if error or the root has not changed, done. */ + if (rc || (dp->tree.stree[ROOT] == oldroot)) + return (rc); /* root changed. bubble the change up to the dmap control pages. * if the adjustment of the upper level control pages fails, - * backout the deallocation. + * backout the deallocation. */ if ((rc = dbAdjCtl(bmp, blkno, dp->tree.stree[ROOT], 0, 0))) { word = (blkno & (BPERDMAP - 1)) >> L2DBWORD; @@ -2288,15 +2217,16 @@ static void dbAllocBits(struct bmap * bmp, struct dmap * dp, s64 blkno, * blkno - starting block number of the bits to be freed. * nblocks - number of bits to be freed. * - * RETURN VALUES: none + * RETURN VALUES: 0 for success * * serialization: IREAD_LOCK(ipbmap) or IWRITE_LOCK(ipbmap) held on entry/exit; */ -static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, +static int dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, int nblocks) { int dbitno, word, rembits, nb, nwords, wbitno, nw, agno; dmtree_t *tp = (dmtree_t *) & dp->tree; + int rc = 0; int size; /* determine the bit number and word within the dmap of the @@ -2315,7 +2245,7 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, * words (i.e. partial first and/or last) on an individual basis * (a single pass), freeing the bits of interest by hand and updating * the leaf corresponding to the dmap word. a single pass will be used - * for all dmap words fully contained within the specified range. + * for all dmap words fully contained within the specified range. * within this pass, the bits of all fully contained dmap words will * be marked as free in a single shot and the leaves will be updated. a * single leaf may describe the free space of multiple dmap words, @@ -2337,7 +2267,7 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, */ if (nb < DBWORD) { /* free (zero) the appropriate bits within this - * dmap word. + * dmap word. */ dp->wmap[word] &= cpu_to_le32(~(ONES << (DBWORD - nb) @@ -2345,8 +2275,10 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, /* update the leaf for this dmap word. */ - dbJoin(tp, word, - dbMaxBud((u8 *) & dp->wmap[word])); + rc = dbJoin(tp, word, + dbMaxBud((u8 *) & dp->wmap[word])); + if (rc) + return rc; word += 1; } else { @@ -2377,7 +2309,9 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, /* update the leaf. */ - dbJoin(tp, word, size); + rc = dbJoin(tp, word, size); + if (rc) + return rc; /* get the number of dmap words handled. */ @@ -2393,7 +2327,7 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, BMAP_LOCK(bmp); - /* update the free count for the allocation group and + /* update the free count for the allocation group and * map. */ agno = blkno >> bmp->db_agl2size; @@ -2424,6 +2358,8 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, } BMAP_UNLOCK(bmp); + + return 0; } @@ -2442,7 +2378,7 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, * or deallocation resulted in the root change. this range * is respresented by a single leaf of the current dmapctl * and the leaf will be updated with this value, possibly - * causing a binary buddy system within the leaves to be + * causing a binary buddy system within the leaves to be * split or joined. the update may also cause the dmapctl's * dmtree to be updated. * @@ -2458,7 +2394,7 @@ static void dbFreeBits(struct bmap * bmp, struct dmap * dp, s64 blkno, * requires the dmap control page to be adjusted. * newval - the new value of the lower level dmap or dmap control * page root. - * alloc - TRUE if adjustment is due to an allocation. + * alloc - 'true' if adjustment is due to an allocation. * level - current level of dmap control page (i.e. L0, L1, L2) to * be adjusted. * @@ -2526,12 +2462,16 @@ dbAdjCtl(struct bmap * bmp, s64 blkno, int newval, int alloc, int level) * that it is at the front of a binary buddy system. */ if (oldval == NOFREE) { - dbBackSplit((dmtree_t *) dcp, leafno); + rc = dbBackSplit((dmtree_t *) dcp, leafno); + if (rc) + return rc; oldval = dcp->stree[ti]; } dbSplit((dmtree_t *) dcp, leafno, dcp->budmin, newval); } else { - dbJoin((dmtree_t *) dcp, leafno, newval); + rc = dbJoin((dmtree_t *) dcp, leafno, newval); + if (rc) + return rc; } /* check if the root of the current dmap control page changed due @@ -2650,7 +2590,7 @@ static void dbSplit(dmtree_t * tp, int leafno, int splitsz, int newval) } } - /* adjust the dmap tree to reflect the specified leaf's new + /* adjust the dmap tree to reflect the specified leaf's new * value. */ dbAdjTree(tp, leafno, newval); @@ -2684,7 +2624,7 @@ static void dbSplit(dmtree_t * tp, int leafno, int splitsz, int newval) * * serialization: IREAD_LOCK(ipbmap) or IWRITE_LOCK(ipbmap) held on entry/exit; */ -static void dbBackSplit(dmtree_t * tp, int leafno) +static int dbBackSplit(dmtree_t * tp, int leafno) { int budsz, bud, w, bsz, size; int cursz; @@ -2698,7 +2638,7 @@ static void dbBackSplit(dmtree_t * tp, int leafno) /* the back split is accomplished by iteratively finding the leaf * that starts the buddy system that contains the specified leaf and * splitting that system in two. this iteration continues until - * the specified leaf becomes the start of a buddy system. + * the specified leaf becomes the start of a buddy system. * * determine maximum possible l2 size for the specified leaf. */ @@ -2719,7 +2659,10 @@ static void dbBackSplit(dmtree_t * tp, int leafno) */ for (w = leafno, bsz = budsz;; bsz <<= 1, w = (w < bud) ? w : bud) { - assert(bsz < le32_to_cpu(tp->dmt_nleafs)); + if (bsz >= le32_to_cpu(tp->dmt_nleafs)) { + jfs_err("JFS: block map error in dbBackSplit"); + return -EIO; + } /* determine the buddy. */ @@ -2738,7 +2681,11 @@ static void dbBackSplit(dmtree_t * tp, int leafno) } } - assert(leaf[leafno] == size); + if (leaf[leafno] != size) { + jfs_err("JFS: wrong leaf value in dbBackSplit"); + return -EIO; + } + return 0; } @@ -2756,7 +2703,7 @@ static void dbBackSplit(dmtree_t * tp, int leafno) * * RETURN VALUES: none */ -static void dbJoin(dmtree_t * tp, int leafno, int newval) +static int dbJoin(dmtree_t * tp, int leafno, int newval) { int budsz, buddy; s8 *leaf; @@ -2796,7 +2743,9 @@ static void dbJoin(dmtree_t * tp, int leafno, int newval) if (newval > leaf[buddy]) break; - assert(newval == leaf[buddy]); + /* It shouldn't be less */ + if (newval < leaf[buddy]) + return -EIO; /* check which (leafno or buddy) is the left buddy. * the left buddy gets to claim the blocks resulting @@ -2828,6 +2777,8 @@ static void dbJoin(dmtree_t * tp, int leafno, int newval) /* update the leaf value. */ dbAdjTree(tp, leafno, newval); + + return 0; } @@ -2902,7 +2853,7 @@ static void dbAdjTree(dmtree_t * tp, int leafno, int newval) * NAME: dbFindLeaf() * * FUNCTION: search a dmtree_t for sufficient free blocks, returning - * the index of a leaf describing the free blocks if + * the index of a leaf describing the free blocks if * sufficient free blocks are found. * * the search starts at the top of the dmtree_t tree and @@ -2918,7 +2869,7 @@ static void dbAdjTree(dmtree_t * tp, int leafno, int newval) * * RETURN VALUES: * 0 - success - * -ENOSPC - insufficient free blocks. + * -ENOSPC - insufficient free blocks. */ static int dbFindLeaf(dmtree_t * tp, int l2nb, int *leafidx) { @@ -3108,7 +3059,7 @@ static int cntlz(u32 value) * RETURN VALUES: * log2 number of blocks */ -int blkstol2(s64 nb) +static int blkstol2(s64 nb) { int l2nb; s64 mask; /* meant to be signed */ @@ -3139,7 +3090,7 @@ int blkstol2(s64 nb) /* - * NAME: dbAllocBottomUp() + * NAME: dbAllocBottomUp() * * FUNCTION: alloc the specified block range from the working block * allocation map. @@ -3194,16 +3145,12 @@ int dbAllocBottomUp(struct inode *ip, s64 blkno, s64 nblocks) */ nb = min(rem, BPERDMAP - (blkno & (BPERDMAP - 1))); - DBFREECK(bmp->db_DBmap, bmp->db_mapsize, blkno, nb); - /* allocate the blocks. */ if ((rc = dbAllocDmapBU(bmp, dp, blkno, nb))) { release_metapage(mp); IREAD_UNLOCK(ipbmap); return (rc); } - - DBALLOC(bmp->db_DBmap, bmp->db_mapsize, blkno, nb); } /* write the last buffer. */ @@ -3294,7 +3241,7 @@ static int dbAllocDmapBU(struct bmap * bmp, struct dmap * dp, s64 blkno, BMAP_LOCK(bmp); /* if this allocation group is completely free, - * update the highest active allocation group number + * update the highest active allocation group number * if this allocation group is the new max. */ agno = blkno >> bmp->db_agl2size; @@ -3326,7 +3273,7 @@ static int dbAllocDmapBU(struct bmap * bmp, struct dmap * dp, s64 blkno, * NAME: dbExtendFS() * * FUNCTION: extend bmap from blkno for nblocks; - * dbExtendFS() updates bmap ready for dbAllocBottomUp(); + * dbExtendFS() updates bmap ready for dbAllocBottomUp(); * * L2 * | @@ -3337,13 +3284,13 @@ static int dbAllocDmapBU(struct bmap * bmp, struct dmap * dp, s64 blkno, * d0,...,dn d0,...,dn d0,...,dn d0,...,dn d0,...,dn d0,.,dm; * L2L1L0d0,...,dnL0d0,...,dnL0d0,...,dnL1L0d0,...,dnL0d0,...,dnL0d0,..dm * - * <---old---><----------------------------extend-----------------------> + * <---old---><----------------------------extend-----------------------> */ int dbExtendFS(struct inode *ipbmap, s64 blkno, s64 nblocks) { struct jfs_sb_info *sbi = JFS_SBI(ipbmap->i_sb); int nbperpage = sbi->nbperpage; - int i, i0 = TRUE, j, j0 = TRUE, k, n; + int i, i0 = true, j, j0 = true, k, n; s64 newsize; s64 p; struct metapage *mp, *l2mp, *l1mp = NULL, *l0mp = NULL; @@ -3383,7 +3330,7 @@ int dbExtendFS(struct inode *ipbmap, s64 blkno, s64 nblocks) bmp->db_numag += ((u32) newsize % (u32) bmp->db_agsize) ? 1 : 0; /* - * reconfigure db_agfree[] + * reconfigure db_agfree[] * from old AG configuration to new AG configuration; * * coalesce contiguous k (newAGSize/oldAGSize) AGs; @@ -3451,7 +3398,7 @@ int dbExtendFS(struct inode *ipbmap, s64 blkno, s64 nblocks) j = (blkno & (MAXL1SIZE - 1)) >> L2MAXL0SIZE; l1leaf = l1dcp->stree + CTLLEAFIND + j; p = BLKTOL0(blkno, sbi->l2nbperpage); - j0 = FALSE; + j0 = false; } else { /* assign/init L1 page */ l1mp = get_metapage(ipbmap, p, PSIZE, 0); @@ -3485,7 +3432,7 @@ int dbExtendFS(struct inode *ipbmap, s64 blkno, s64 nblocks) l0leaf = l0dcp->stree + CTLLEAFIND + i; p = BLKTODMAP(blkno, sbi->l2nbperpage); - i0 = FALSE; + i0 = false; } else { /* assign/init L0 page */ l0mp = get_metapage(ipbmap, p, PSIZE, 0); @@ -3544,7 +3491,7 @@ int dbExtendFS(struct inode *ipbmap, s64 blkno, s64 nblocks) } /* for each dmap in a L0 */ /* - * build current L0 page from its leaves, and + * build current L0 page from its leaves, and * initialize corresponding parent L1 leaf */ *l1leaf = dbInitDmapCtl(l0dcp, 0, ++i); @@ -3568,7 +3515,7 @@ int dbExtendFS(struct inode *ipbmap, s64 blkno, s64 nblocks) } /* for each L0 in a L1 */ /* - * build current L1 page from its leaves, and + * build current L1 page from its leaves, and * initialize corresponding parent L2 leaf */ *l2leaf = dbInitDmapCtl(l1dcp, 1, ++j); @@ -3623,7 +3570,7 @@ void dbFinalizeBmap(struct inode *ipbmap) * finalize bmap control page */ //finalize: - /* + /* * compute db_agpref: preferred ag to allocate from * (the leftmost ag with average free space in it); */ @@ -3667,9 +3614,9 @@ void dbFinalizeBmap(struct inode *ipbmap) /* * compute db_aglevel, db_agheigth, db_width, db_agstart: - * an ag is covered in aglevel dmapctl summary tree, - * at agheight level height (from leaf) with agwidth number of nodes - * each, which starts at agstart index node of the smmary tree node + * an ag is covered in aglevel dmapctl summary tree, + * at agheight level height (from leaf) with agwidth number of nodes + * each, which starts at agstart index node of the smmary tree node * array; */ bmp->db_aglevel = BMAPSZTOLEV(bmp->db_agsize); @@ -3688,13 +3635,13 @@ void dbFinalizeBmap(struct inode *ipbmap) /* * NAME: dbInitDmap()/ujfs_idmap_page() - * + * * FUNCTION: initialize working/persistent bitmap of the dmap page * for the specified number of blocks: - * + * * at entry, the bitmaps had been initialized as free (ZEROS); - * The number of blocks will only account for the actually - * existing blocks. Blocks which don't actually exist in + * The number of blocks will only account for the actually + * existing blocks. Blocks which don't actually exist in * the aggregate will be marked as allocated (ONES); * * PARAMETERS: @@ -3730,7 +3677,7 @@ static int dbInitDmap(struct dmap * dp, s64 Blkno, int nblocks) /* * free the bits corresponding to the block range (ZEROS): - * note: not all bits of the first and last words may be contained + * note: not all bits of the first and last words may be contained * within the block range. */ for (r = nblocks; r > 0; r -= nb, blkno += nb) { @@ -3762,7 +3709,7 @@ static int dbInitDmap(struct dmap * dp, s64 Blkno, int nblocks) } /* - * mark bits following the range to be freed (non-existing + * mark bits following the range to be freed (non-existing * blocks) as allocated (ONES) */ @@ -3794,11 +3741,11 @@ static int dbInitDmap(struct dmap * dp, s64 Blkno, int nblocks) /* * NAME: dbInitDmapTree()/ujfs_complete_dmap() - * + * * FUNCTION: initialize summary tree of the specified dmap: * * at entry, bitmap of the dmap has been initialized; - * + * * PARAMETERS: * dp - dmap to complete * blkno - starting block number for this dmap @@ -3822,7 +3769,7 @@ static int dbInitDmapTree(struct dmap * dp) /* init each leaf from corresponding wmap word: * note: leaf is set to NOFREE(-1) if all blocks of corresponding - * bitmap word are allocated. + * bitmap word are allocated. */ cp = tp->stree + le32_to_cpu(tp->leafidx); for (i = 0; i < LPERDMAP; i++) @@ -3835,10 +3782,10 @@ static int dbInitDmapTree(struct dmap * dp) /* * NAME: dbInitTree()/ujfs_adjtree() - * + * * FUNCTION: initialize binary buddy summary tree of a dmap or dmapctl. * - * at entry, the leaves of the tree has been initialized + * at entry, the leaves of the tree has been initialized * from corresponding bitmap word or root of summary tree * of the child control page; * configure binary buddy system at the leaf level, then @@ -3866,15 +3813,15 @@ static int dbInitTree(struct dmaptree * dtp) /* * configure the leaf levevl into binary buddy system * - * Try to combine buddies starting with a buddy size of 1 - * (i.e. two leaves). At a buddy size of 1 two buddy leaves - * can be combined if both buddies have a maximum free of l2min; - * the combination will result in the left-most buddy leaf having - * a maximum free of l2min+1. - * After processing all buddies for a given size, process buddies - * at the next higher buddy size (i.e. current size * 2) and - * the next maximum free (current free + 1). - * This continues until the maximum possible buddy combination + * Try to combine buddies starting with a buddy size of 1 + * (i.e. two leaves). At a buddy size of 1 two buddy leaves + * can be combined if both buddies have a maximum free of l2min; + * the combination will result in the left-most buddy leaf having + * a maximum free of l2min+1. + * After processing all buddies for a given size, process buddies + * at the next higher buddy size (i.e. current size * 2) and + * the next maximum free (current free + 1). + * This continues until the maximum possible buddy combination * yields maximum free. */ for (l2free = dtp->budmin, bsize = 1; l2free < l2max; @@ -3898,10 +3845,10 @@ static int dbInitTree(struct dmaptree * dtp) * bubble summary information of leaves up the tree. * * Starting at the leaf node level, the four nodes described by - * the higher level parent node are compared for a maximum free and - * this maximum becomes the value of the parent node. - * when all lower level nodes are processed in this fashion then - * move up to the next level (parent becomes a lower level node) and + * the higher level parent node are compared for a maximum free and + * this maximum becomes the value of the parent node. + * when all lower level nodes are processed in this fashion then + * move up to the next level (parent becomes a lower level node) and * continue the process for that level. */ for (child = le32_to_cpu(dtp->leafidx), @@ -3910,7 +3857,7 @@ static int dbInitTree(struct dmaptree * dtp) /* get index of 1st node of parent level */ parent = (child - 1) >> 2; - /* set the value of the parent node as the maximum + /* set the value of the parent node as the maximum * of the four nodes of the current level. */ for (i = 0, cp = tp + child, cp1 = tp + parent; @@ -3938,8 +3885,8 @@ static int dbInitDmapCtl(struct dmapctl * dcp, int level, int i) dcp->budmin = L2BPERDMAP + L2LPERCTL * level; /* - * initialize the leaves of current level that were not covered - * by the specified input block range (i.e. the leaves have no + * initialize the leaves of current level that were not covered + * by the specified input block range (i.e. the leaves have no * low level dmapctl or dmap). */ cp = &dcp->stree[CTLLEAFIND + i]; @@ -3953,9 +3900,9 @@ static int dbInitDmapCtl(struct dmapctl * dcp, int level, int i) /* * NAME: dbGetL2AGSize()/ujfs_getagl2size() - * + * * FUNCTION: Determine log2(allocation group size) from aggregate size - * + * * PARAMETERS: * nblocks - Number of blocks in aggregate * @@ -3988,8 +3935,8 @@ static int dbGetL2AGSize(s64 nblocks) /* * NAME: dbMapFileSizeToMapSize() - * - * FUNCTION: compute number of blocks the block allocation map file + * + * FUNCTION: compute number of blocks the block allocation map file * can cover from the map file size; * * RETURNS: Number of blocks which can be covered by this block map file; @@ -4021,7 +3968,7 @@ s64 dbMapFileSizeToMapSize(struct inode * ipbmap) npages = nblocks >> JFS_SBI(sb)->l2nbperpage; level = BMAPPGTOLEV(npages); - /* At each level, accumulate the number of dmap pages covered by + /* At each level, accumulate the number of dmap pages covered by * the number of full child levels below it; * repeat for the last incomplete child level. */ @@ -4043,230 +3990,10 @@ s64 dbMapFileSizeToMapSize(struct inode * ipbmap) npages--; } - /* convert the number of dmaps into the number of blocks + /* convert the number of dmaps into the number of blocks * which can be covered by the dmaps; */ nblocks = ndmaps << L2BPERDMAP; return (nblocks); } - - -#ifdef _JFS_DEBUG_DMAP -/* - * DBinitmap() - */ -static void DBinitmap(s64 size, struct inode *ipbmap, u32 ** results) -{ - int npages; - u32 *dbmap, *d; - int n; - s64 lblkno, cur_block; - struct dmap *dp; - struct metapage *mp; - - npages = size / 32768; - npages += (size % 32768) ? 1 : 0; - - dbmap = (u32 *) xmalloc(npages * 4096, L2PSIZE, kernel_heap); - if (dbmap == NULL) - BUG(); /* Not robust since this is only unused debug code */ - - for (n = 0, d = dbmap; n < npages; n++, d += 1024) - bzero(d, 4096); - - /* Need to initialize from disk map pages - */ - for (d = dbmap, cur_block = 0; cur_block < size; - cur_block += BPERDMAP, d += LPERDMAP) { - lblkno = BLKTODMAP(cur_block, - JFS_SBI(ipbmap->i_sb)->bmap-> - db_l2nbperpage); - mp = read_metapage(ipbmap, lblkno, PSIZE, 0); - if (mp == NULL) { - jfs_error(ipbmap->i_sb, - "DBinitmap: could not read disk map page"); - continue; - } - dp = (struct dmap *) mp->data; - - for (n = 0; n < LPERDMAP; n++) - d[n] = le32_to_cpu(dp->wmap[n]); - - release_metapage(mp); - } - - *results = dbmap; -} - - -/* - * DBAlloc() - */ -void DBAlloc(uint * dbmap, s64 mapsize, s64 blkno, s64 nblocks) -{ - int word, nb, bitno; - u32 mask; - - assert(blkno > 0 && blkno < mapsize); - assert(nblocks > 0 && nblocks <= mapsize); - - assert(blkno + nblocks <= mapsize); - - dbmap += (blkno / 32); - while (nblocks > 0) { - bitno = blkno & (32 - 1); - nb = min(nblocks, 32 - bitno); - - mask = (0xffffffff << (32 - nb) >> bitno); - assert((mask & *dbmap) == 0); - *dbmap |= mask; - - dbmap++; - blkno += nb; - nblocks -= nb; - } -} - - -/* - * DBFree() - */ -static void DBFree(uint * dbmap, s64 mapsize, s64 blkno, s64 nblocks) -{ - int word, nb, bitno; - u32 mask; - - assert(blkno > 0 && blkno < mapsize); - assert(nblocks > 0 && nblocks <= mapsize); - - assert(blkno + nblocks <= mapsize); - - dbmap += (blkno / 32); - while (nblocks > 0) { - bitno = blkno & (32 - 1); - nb = min(nblocks, 32 - bitno); - - mask = (0xffffffff << (32 - nb) >> bitno); - assert((mask & *dbmap) == mask); - *dbmap &= ~mask; - - dbmap++; - blkno += nb; - nblocks -= nb; - } -} - - -/* - * DBAllocCK() - */ -static void DBAllocCK(uint * dbmap, s64 mapsize, s64 blkno, s64 nblocks) -{ - int word, nb, bitno; - u32 mask; - - assert(blkno > 0 && blkno < mapsize); - assert(nblocks > 0 && nblocks <= mapsize); - - assert(blkno + nblocks <= mapsize); - - dbmap += (blkno / 32); - while (nblocks > 0) { - bitno = blkno & (32 - 1); - nb = min(nblocks, 32 - bitno); - - mask = (0xffffffff << (32 - nb) >> bitno); - assert((mask & *dbmap) == mask); - - dbmap++; - blkno += nb; - nblocks -= nb; - } -} - - -/* - * DBFreeCK() - */ -static void DBFreeCK(uint * dbmap, s64 mapsize, s64 blkno, s64 nblocks) -{ - int word, nb, bitno; - u32 mask; - - assert(blkno > 0 && blkno < mapsize); - assert(nblocks > 0 && nblocks <= mapsize); - - assert(blkno + nblocks <= mapsize); - - dbmap += (blkno / 32); - while (nblocks > 0) { - bitno = blkno & (32 - 1); - nb = min(nblocks, 32 - bitno); - - mask = (0xffffffff << (32 - nb) >> bitno); - assert((mask & *dbmap) == 0); - - dbmap++; - blkno += nb; - nblocks -= nb; - } -} - - -/* - * dbPrtMap() - */ -static void dbPrtMap(struct bmap * bmp) -{ - printk(" mapsize: %d%d\n", bmp->db_mapsize); - printk(" nfree: %d%d\n", bmp->db_nfree); - printk(" numag: %d\n", bmp->db_numag); - printk(" agsize: %d%d\n", bmp->db_agsize); - printk(" agl2size: %d\n", bmp->db_agl2size); - printk(" agwidth: %d\n", bmp->db_agwidth); - printk(" agstart: %d\n", bmp->db_agstart); - printk(" agheigth: %d\n", bmp->db_agheigth); - printk(" aglevel: %d\n", bmp->db_aglevel); - printk(" maxlevel: %d\n", bmp->db_maxlevel); - printk(" maxag: %d\n", bmp->db_maxag); - printk(" agpref: %d\n", bmp->db_agpref); - printk(" l2nbppg: %d\n", bmp->db_l2nbperpage); -} - - -/* - * dbPrtCtl() - */ -static void dbPrtCtl(struct dmapctl * dcp) -{ - int i, j, n; - - printk(" height: %08x\n", le32_to_cpu(dcp->height)); - printk(" leafidx: %08x\n", le32_to_cpu(dcp->leafidx)); - printk(" budmin: %08x\n", dcp->budmin); - printk(" nleafs: %08x\n", le32_to_cpu(dcp->nleafs)); - printk(" l2nleafs: %08x\n", le32_to_cpu(dcp->l2nleafs)); - - printk("\n Tree:\n"); - for (i = 0; i < CTLLEAFIND; i += 8) { - n = min(8, CTLLEAFIND - i); - - for (j = 0; j < n; j++) - printf(" [%03x]: %02x", i + j, - (char) dcp->stree[i + j]); - printf("\n"); - } - - printk("\n Tree Leaves:\n"); - for (i = 0; i < LPERCTL; i += 8) { - n = min(8, LPERCTL - i); - - for (j = 0; j < n; j++) - printf(" [%03x]: %02x", - i + j, - (char) dcp->stree[i + j + CTLLEAFIND]); - printf("\n"); - } -} -#endif /* _JFS_DEBUG_DMAP */