xfs: add an extent to the rmap btree
authorDarrick J. Wong <darrick.wong@oracle.com>
Wed, 3 Aug 2016 01:44:21 +0000 (11:44 +1000)
committerDave Chinner <david@fromorbit.com>
Wed, 3 Aug 2016 01:44:21 +0000 (11:44 +1000)
Originally-From: Dave Chinner <dchinner@redhat.com>

Now all the btree, free space and transaction infrastructure is in
place, we can finally add the code to insert reverse mappings to the
rmap btree. Freeing will be done in a separate patch, so just the
addition operation can be focussed on here.

[darrick: handle owner offsets when adding rmaps]
[dchinner: remove remaining debug printk statements]
[darrick: move unwritten bit to rm_offset]

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
fs/xfs/libxfs/xfs_rmap.c
fs/xfs/xfs_trace.h

index 6638bc725b232a2a457e535fcebd1b44c66754b2..1c0e2a8943a4b6a031ff2dc464ae354066f98ad8 100644 (file)
@@ -31,6 +31,7 @@
 #include "xfs_trans.h"
 #include "xfs_alloc.h"
 #include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
 #include "xfs_trans_space.h"
 #include "xfs_trace.h"
 #include "xfs_error.h"
@@ -158,6 +159,218 @@ out_error:
        return error;
 }
 
+/*
+ * A mergeable rmap must have the same owner and the same values for
+ * the unwritten, attr_fork, and bmbt flags.  The startblock and
+ * offset are checked separately.
+ */
+static bool
+xfs_rmap_is_mergeable(
+       struct xfs_rmap_irec    *irec,
+       uint64_t                owner,
+       unsigned int            flags)
+{
+       if (irec->rm_owner == XFS_RMAP_OWN_NULL)
+               return false;
+       if (irec->rm_owner != owner)
+               return false;
+       if ((flags & XFS_RMAP_UNWRITTEN) ^
+           (irec->rm_flags & XFS_RMAP_UNWRITTEN))
+               return false;
+       if ((flags & XFS_RMAP_ATTR_FORK) ^
+           (irec->rm_flags & XFS_RMAP_ATTR_FORK))
+               return false;
+       if ((flags & XFS_RMAP_BMBT_BLOCK) ^
+           (irec->rm_flags & XFS_RMAP_BMBT_BLOCK))
+               return false;
+       return true;
+}
+
+/*
+ * When we allocate a new block, the first thing we do is add a reference to
+ * the extent in the rmap btree. This takes the form of a [agbno, length,
+ * owner, offset] record.  Flags are encoded in the high bits of the offset
+ * field.
+ */
+STATIC int
+xfs_rmap_map(
+       struct xfs_btree_cur    *cur,
+       xfs_agblock_t           bno,
+       xfs_extlen_t            len,
+       bool                    unwritten,
+       struct xfs_owner_info   *oinfo)
+{
+       struct xfs_mount        *mp = cur->bc_mp;
+       struct xfs_rmap_irec    ltrec;
+       struct xfs_rmap_irec    gtrec;
+       int                     have_gt;
+       int                     have_lt;
+       int                     error = 0;
+       int                     i;
+       uint64_t                owner;
+       uint64_t                offset;
+       unsigned int            flags = 0;
+       bool                    ignore_off;
+
+       xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
+       ASSERT(owner != 0);
+       ignore_off = XFS_RMAP_NON_INODE_OWNER(owner) ||
+                       (flags & XFS_RMAP_BMBT_BLOCK);
+       if (unwritten)
+               flags |= XFS_RMAP_UNWRITTEN;
+       trace_xfs_rmap_map(mp, cur->bc_private.a.agno, bno, len,
+                       unwritten, oinfo);
+
+       /*
+        * For the initial lookup, look for an exact match or the left-adjacent
+        * record for our insertion point. This will also give us the record for
+        * start block contiguity tests.
+        */
+       error = xfs_rmap_lookup_le(cur, bno, len, owner, offset, flags,
+                       &have_lt);
+       if (error)
+               goto out_error;
+       XFS_WANT_CORRUPTED_GOTO(mp, have_lt == 1, out_error);
+
+       error = xfs_rmap_get_rec(cur, &ltrec, &have_lt);
+       if (error)
+               goto out_error;
+       XFS_WANT_CORRUPTED_GOTO(mp, have_lt == 1, out_error);
+       trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
+                       cur->bc_private.a.agno, ltrec.rm_startblock,
+                       ltrec.rm_blockcount, ltrec.rm_owner,
+                       ltrec.rm_offset, ltrec.rm_flags);
+
+       if (!xfs_rmap_is_mergeable(&ltrec, owner, flags))
+               have_lt = 0;
+
+       XFS_WANT_CORRUPTED_GOTO(mp,
+               have_lt == 0 ||
+               ltrec.rm_startblock + ltrec.rm_blockcount <= bno, out_error);
+
+       /*
+        * Increment the cursor to see if we have a right-adjacent record to our
+        * insertion point. This will give us the record for end block
+        * contiguity tests.
+        */
+       error = xfs_btree_increment(cur, 0, &have_gt);
+       if (error)
+               goto out_error;
+       if (have_gt) {
+               error = xfs_rmap_get_rec(cur, &gtrec, &have_gt);
+               if (error)
+                       goto out_error;
+               XFS_WANT_CORRUPTED_GOTO(mp, have_gt == 1, out_error);
+               XFS_WANT_CORRUPTED_GOTO(mp, bno + len <= gtrec.rm_startblock,
+                                       out_error);
+               trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
+                       cur->bc_private.a.agno, gtrec.rm_startblock,
+                       gtrec.rm_blockcount, gtrec.rm_owner,
+                       gtrec.rm_offset, gtrec.rm_flags);
+               if (!xfs_rmap_is_mergeable(&gtrec, owner, flags))
+                       have_gt = 0;
+       }
+
+       /*
+        * Note: cursor currently points one record to the right of ltrec, even
+        * if there is no record in the tree to the right.
+        */
+       if (have_lt &&
+           ltrec.rm_startblock + ltrec.rm_blockcount == bno &&
+           (ignore_off || ltrec.rm_offset + ltrec.rm_blockcount == offset)) {
+               /*
+                * left edge contiguous, merge into left record.
+                *
+                *       ltbno     ltlen
+                * orig:   |ooooooooo|
+                * adding:           |aaaaaaaaa|
+                * result: |rrrrrrrrrrrrrrrrrrr|
+                *                  bno       len
+                */
+               ltrec.rm_blockcount += len;
+               if (have_gt &&
+                   bno + len == gtrec.rm_startblock &&
+                   (ignore_off || offset + len == gtrec.rm_offset) &&
+                   (unsigned long)ltrec.rm_blockcount + len +
+                               gtrec.rm_blockcount <= XFS_RMAP_LEN_MAX) {
+                       /*
+                        * right edge also contiguous, delete right record
+                        * and merge into left record.
+                        *
+                        *       ltbno     ltlen    gtbno     gtlen
+                        * orig:   |ooooooooo|         |ooooooooo|
+                        * adding:           |aaaaaaaaa|
+                        * result: |rrrrrrrrrrrrrrrrrrrrrrrrrrrrr|
+                        */
+                       ltrec.rm_blockcount += gtrec.rm_blockcount;
+                       trace_xfs_rmap_delete(mp, cur->bc_private.a.agno,
+                                       gtrec.rm_startblock,
+                                       gtrec.rm_blockcount,
+                                       gtrec.rm_owner,
+                                       gtrec.rm_offset,
+                                       gtrec.rm_flags);
+                       error = xfs_btree_delete(cur, &i);
+                       if (error)
+                               goto out_error;
+                       XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
+               }
+
+               /* point the cursor back to the left record and update */
+               error = xfs_btree_decrement(cur, 0, &have_gt);
+               if (error)
+                       goto out_error;
+               error = xfs_rmap_update(cur, &ltrec);
+               if (error)
+                       goto out_error;
+       } else if (have_gt &&
+                  bno + len == gtrec.rm_startblock &&
+                  (ignore_off || offset + len == gtrec.rm_offset)) {
+               /*
+                * right edge contiguous, merge into right record.
+                *
+                *                 gtbno     gtlen
+                * Orig:             |ooooooooo|
+                * adding: |aaaaaaaaa|
+                * Result: |rrrrrrrrrrrrrrrrrrr|
+                *        bno       len
+                */
+               gtrec.rm_startblock = bno;
+               gtrec.rm_blockcount += len;
+               if (!ignore_off)
+                       gtrec.rm_offset = offset;
+               error = xfs_rmap_update(cur, &gtrec);
+               if (error)
+                       goto out_error;
+       } else {
+               /*
+                * no contiguous edge with identical owner, insert
+                * new record at current cursor position.
+                */
+               cur->bc_rec.r.rm_startblock = bno;
+               cur->bc_rec.r.rm_blockcount = len;
+               cur->bc_rec.r.rm_owner = owner;
+               cur->bc_rec.r.rm_offset = offset;
+               cur->bc_rec.r.rm_flags = flags;
+               trace_xfs_rmap_insert(mp, cur->bc_private.a.agno, bno, len,
+                       owner, offset, flags);
+               error = xfs_btree_insert(cur, &i);
+               if (error)
+                       goto out_error;
+               XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
+       }
+
+       trace_xfs_rmap_map_done(mp, cur->bc_private.a.agno, bno, len,
+                       unwritten, oinfo);
+out_error:
+       if (error)
+               trace_xfs_rmap_map_error(mp, cur->bc_private.a.agno,
+                               error, _RET_IP_);
+       return error;
+}
+
+/*
+ * Add a reference to an extent in the rmap btree.
+ */
 int
 xfs_rmap_alloc(
        struct xfs_trans        *tp,
@@ -168,19 +381,22 @@ xfs_rmap_alloc(
        struct xfs_owner_info   *oinfo)
 {
        struct xfs_mount        *mp = tp->t_mountp;
-       int                     error = 0;
+       struct xfs_btree_cur    *cur;
+       int                     error;
 
        if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
                return 0;
 
-       trace_xfs_rmap_map(mp, agno, bno, len, false, oinfo);
-       if (1)
+       cur = xfs_rmapbt_init_cursor(mp, tp, agbp, agno);
+       error = xfs_rmap_map(cur, bno, len, false, oinfo);
+       if (error)
                goto out_error;
-       trace_xfs_rmap_map_done(mp, agno, bno, len, false, oinfo);
+
+       xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
        return 0;
 
 out_error:
-       trace_xfs_rmap_map_error(mp, agno, error, _RET_IP_);
+       xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
        return error;
 }
 
index 7498c7780d9992c9dc7baa09b0629104fa1a1f57..e099eda75b68f82b26e6d4c570bd08d8d4892b0b 100644 (file)
@@ -2551,6 +2551,8 @@ DEFINE_RMAPBT_EVENT(xfs_rmap_delete);
 DEFINE_AG_ERROR_EVENT(xfs_rmap_insert_error);
 DEFINE_AG_ERROR_EVENT(xfs_rmap_delete_error);
 DEFINE_AG_ERROR_EVENT(xfs_rmap_update_error);
+DEFINE_RMAPBT_EVENT(xfs_rmap_lookup_le_range_result);
+DEFINE_RMAPBT_EVENT(xfs_rmap_find_right_neighbor_result);
 
 #endif /* _TRACE_XFS_H */