xfs: remove xfs_btree_cur private firstblock field
authorBrian Foster <bfoster@redhat.com>
Thu, 12 Jul 2018 05:26:29 +0000 (22:26 -0700)
committerDarrick J. Wong <darrick.wong@oracle.com>
Thu, 12 Jul 2018 05:26:29 +0000 (22:26 -0700)
The bmbt cursor private structure has a firstblock field that is
used to maintain locking order on bmbt allocations. The field holds
an actual firstblock value (as opposed to a pointer), so it is
initialized on cursor creation, updated on allocation and then the
value is transferred back to the source before the cursor is
destroyed.

This value is always transferred from and back to the ->t_firstblock
field. Since xfs_btree_cur already carries a reference to the
transaction, we can remove this field from xfs_btree_cur and the
associated copying. The bmbt allocations will update the value in
the transaction directly.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
fs/xfs/libxfs/xfs_bmap.c
fs/xfs/libxfs/xfs_bmap_btree.c
fs/xfs/libxfs/xfs_btree.h

index 183450f..8a1e689 100644 (file)
@@ -688,7 +688,6 @@ xfs_bmap_extents_to_btree(
         * Need a cursor.  Can't allocate until bb_level is filled in.
         */
        cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
-       cur->bc_private.b.firstblock = tp->t_firstblock;
        cur->bc_private.b.flags = wasdel ? XFS_BTCUR_BPRV_WASDEL : 0;
        /*
         * Convert to a btree with two levels, one record in root.
@@ -732,7 +731,7 @@ xfs_bmap_extents_to_btree(
         */
        ASSERT(tp->t_firstblock == NULLFSBLOCK ||
               args.agno >= XFS_FSB_TO_AGNO(mp, tp->t_firstblock));
-       tp->t_firstblock = cur->bc_private.b.firstblock = args.fsbno;
+       tp->t_firstblock = args.fsbno;
        cur->bc_private.b.allocated++;
        ip->i_d.di_nblocks++;
        xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, 1L);
@@ -925,7 +924,6 @@ xfs_bmap_add_attrfork_btree(
                *flags |= XFS_ILOG_DBROOT;
        else {
                cur = xfs_bmbt_init_cursor(mp, tp, ip, XFS_DATA_FORK);
-               cur->bc_private.b.firstblock = tp->t_firstblock;
                error = xfs_bmbt_lookup_first(cur, &stat);
                if (error)
                        goto error0;
@@ -937,7 +935,6 @@ xfs_bmap_add_attrfork_btree(
                        xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
                        return -ENOSPC;
                }
-               tp->t_firstblock = cur->bc_private.b.firstblock;
                cur->bc_private.b.allocated = 0;
                xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
        }
@@ -4058,14 +4055,10 @@ xfs_bmapi_allocate(
        if (error)
                return error;
 
-       if (bma->cur)
-               bma->cur->bc_private.b.firstblock = bma->tp->t_firstblock;
        if (bma->blkno == NULLFSBLOCK)
                return 0;
-       if ((ifp->if_flags & XFS_IFBROOT) && !bma->cur) {
+       if ((ifp->if_flags & XFS_IFBROOT) && !bma->cur)
                bma->cur = xfs_bmbt_init_cursor(mp, bma->tp, bma->ip, whichfork);
-               bma->cur->bc_private.b.firstblock = bma->tp->t_firstblock;
-       }
        /*
         * Bump the number of extents we've allocated
         * in this call.
@@ -4151,7 +4144,6 @@ xfs_bmapi_convert_unwritten(
        if ((ifp->if_flags & XFS_IFBROOT) && !bma->cur) {
                bma->cur = xfs_bmbt_init_cursor(bma->ip->i_mount, bma->tp,
                                        bma->ip, whichfork);
-               bma->cur->bc_private.b.firstblock = bma->tp->t_firstblock;
        }
        mval->br_state = (mval->br_state == XFS_EXT_UNWRITTEN)
                                ? XFS_EXT_NORM : XFS_EXT_UNWRITTEN;
@@ -4458,13 +4450,6 @@ error0:
                xfs_trans_log_inode(tp, ip, bma.logflags);
 
        if (bma.cur) {
-               if (!error) {
-                       ASSERT(tp->t_firstblock == NULLFSBLOCK ||
-                              XFS_FSB_TO_AGNO(mp, tp->t_firstblock) <=
-                              XFS_FSB_TO_AGNO(mp,
-                                      bma.cur->bc_private.b.firstblock));
-                       tp->t_firstblock = bma.cur->bc_private.b.firstblock;
-               }
                xfs_btree_del_cursor(bma.cur,
                        error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
        }
@@ -4528,7 +4513,6 @@ xfs_bmapi_remap(
 
        if (ifp->if_flags & XFS_IFBROOT) {
                cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
-               cur->bc_private.b.firstblock = tp->t_firstblock;
                cur->bc_private.b.flags = 0;
        }
 
@@ -5189,7 +5173,6 @@ __xfs_bunmapi(
        if (ifp->if_flags & XFS_IFBROOT) {
                ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE);
                cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
-               cur->bc_private.b.firstblock = tp->t_firstblock;
                cur->bc_private.b.flags = 0;
        } else
                cur = NULL;
@@ -5457,10 +5440,8 @@ error0:
        if (logflags)
                xfs_trans_log_inode(tp, ip, logflags);
        if (cur) {
-               if (!error) {
-                       tp->t_firstblock = cur->bc_private.b.firstblock;
+               if (!error)
                        cur->bc_private.b.allocated = 0;
-               }
                xfs_btree_del_cursor(cur,
                        error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
        }
@@ -5676,7 +5657,6 @@ xfs_bmap_collapse_extents(
 
        if (ifp->if_flags & XFS_IFBROOT) {
                cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
-               cur->bc_private.b.firstblock = tp->t_firstblock;
                cur->bc_private.b.flags = 0;
        }
 
@@ -5796,7 +5776,6 @@ xfs_bmap_insert_extents(
 
        if (ifp->if_flags & XFS_IFBROOT) {
                cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
-               cur->bc_private.b.firstblock = tp->t_firstblock;
                cur->bc_private.b.flags = 0;
        }
 
@@ -5918,7 +5897,6 @@ xfs_bmap_split_extent_at(
 
        if (ifp->if_flags & XFS_IFBROOT) {
                cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
-               cur->bc_private.b.firstblock = tp->t_firstblock;
                cur->bc_private.b.flags = 0;
                error = xfs_bmbt_lookup_eq(cur, &got, &i);
                if (error)
index e8b01af..8a9b98b 100644 (file)
@@ -175,7 +175,6 @@ xfs_bmbt_dup_cursor(
         * Copy the firstblock, dfops, and flags values,
         * since init cursor doesn't get them.
         */
-       new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
        new->bc_private.b.flags = cur->bc_private.b.flags;
 
        return new;
@@ -186,11 +185,11 @@ xfs_bmbt_update_cursor(
        struct xfs_btree_cur    *src,
        struct xfs_btree_cur    *dst)
 {
-       ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
+       ASSERT((dst->bc_tp->t_firstblock != NULLFSBLOCK) ||
               (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
 
        dst->bc_private.b.allocated += src->bc_private.b.allocated;
-       dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
+       dst->bc_tp->t_firstblock = src->bc_tp->t_firstblock;
 
        src->bc_private.b.allocated = 0;
 }
@@ -208,7 +207,7 @@ xfs_bmbt_alloc_block(
        memset(&args, 0, sizeof(args));
        args.tp = cur->bc_tp;
        args.mp = cur->bc_mp;
-       args.fsbno = cur->bc_private.b.firstblock;
+       args.fsbno = cur->bc_tp->t_firstblock;
        args.firstblock = args.fsbno;
        xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_private.b.ip->i_ino,
                        cur->bc_private.b.whichfork);
@@ -263,7 +262,7 @@ xfs_bmbt_alloc_block(
        }
 
        ASSERT(args.len == 1);
-       cur->bc_private.b.firstblock = args.fsbno;
+       cur->bc_tp->t_firstblock = args.fsbno;
        cur->bc_private.b.allocated++;
        cur->bc_private.b.ip->i_d.di_nblocks++;
        xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
@@ -562,7 +561,6 @@ xfs_bmbt_init_cursor(
 
        cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
        cur->bc_private.b.ip = ip;
-       cur->bc_private.b.firstblock = NULLFSBLOCK;
        cur->bc_private.b.allocated = 0;
        cur->bc_private.b.flags = 0;
        cur->bc_private.b.whichfork = whichfork;
index b986a8f..503615f 100644 (file)
@@ -214,7 +214,6 @@ typedef struct xfs_btree_cur
                } a;
                struct {                        /* needed for BMAP */
                        struct xfs_inode *ip;   /* pointer to our inode */
-                       xfs_fsblock_t   firstblock;     /* 1st blk allocated */
                        int             allocated;      /* count of alloced */
                        short           forksize;       /* fork's inode space */
                        char            whichfork;      /* data or attr fork */