From: Darrick J. Wong <djwong@xxxxxxxxxx>
Ideally, we'd put all the metadata inodes in one place if we could, so
that the metadata all stay reasonably close together instead of
spreading out over the disk. Furthermore, if the log is internal we'd
probably prefer to keep the metadata near the log. Therefore, disable
AGI rotoring for metadata inode allocations.
Signed-off-by: Darrick J. Wong <djwong@xxxxxxxxxx>
---
db/iunlink.c | 2 +-
libxfs/xfs_ialloc.c | 56 +++++++++++++++++++++++++++++++++++----------------
libxfs/xfs_ialloc.h | 2 +-
libxfs/xfs_imeta.c | 4 ++--
mkfs/proto.c | 3 +--
repair/phase6.c | 2 +-
6 files changed, 44 insertions(+), 25 deletions(-)
diff --git a/db/iunlink.c b/db/iunlink.c
index c87b98431e5..fd5ed64c9e2 100644
--- a/db/iunlink.c
+++ b/db/iunlink.c
@@ -221,7 +221,7 @@ create_unlinked(
return error;
}
- error = -libxfs_dialloc(&tp, 0, args.mode, &ino);
+ error = -libxfs_dialloc(&tp, args.pip, args.mode, &ino);
if (error) {
dbprintf(_("alloc inode: %s\n"), strerror(error));
goto out_cancel;
diff --git a/libxfs/xfs_ialloc.c b/libxfs/xfs_ialloc.c
index 2c941603986..19543f76994 100644
--- a/libxfs/xfs_ialloc.c
+++ b/libxfs/xfs_ialloc.c
@@ -1794,6 +1794,37 @@ xfs_dialloc_try_ag(
return error;
}
+/*
+ * Pick an AG for the new inode.
+ *
+ * Directories, symlinks, and regular files frequently allocate at least one
+ * block, so factor that potential expansion when we examine whether an AG has
+ * enough space for file creation. Try to keep metadata files all in the same
+ * AG.
+ */
+static inline xfs_agnumber_t
+xfs_dialloc_pick_ag(
+ struct xfs_mount *mp,
+ struct xfs_inode *dp,
+ umode_t mode)
+{
+ xfs_agnumber_t start_agno;
+
+ if (!dp)
+ return 0;
+ if (xfs_is_metadir_inode(dp))
+ return 0;
+
+ if (S_ISDIR(mode))
+ return (atomic_inc_return(&mp->m_agirotor) - 1) % mp->m_maxagi;
+
+ start_agno = XFS_INO_TO_AGNO(mp, dp->i_ino);
+ if (start_agno >= mp->m_maxagi)
+ start_agno = 0;
+
+ return start_agno;
+}
+
/*
* Allocate an on-disk inode.
*
@@ -1805,34 +1836,23 @@ xfs_dialloc_try_ag(
int
xfs_dialloc(
struct xfs_trans **tpp,
- xfs_ino_t parent,
+ struct xfs_inode *dp,
umode_t mode,
xfs_ino_t *new_ino)
{
struct xfs_mount *mp = (*tpp)->t_mountp;
- xfs_agnumber_t agno;
- int error = 0;
- xfs_agnumber_t start_agno;
struct xfs_perag *pag;
struct xfs_ino_geometry *igeo = M_IGEO(mp);
+ xfs_ino_t ino = NULLFSINO;
+ xfs_ino_t parent = dp ? dp->i_ino : 0;
+ xfs_agnumber_t agno;
+ xfs_agnumber_t start_agno;
bool ok_alloc = true;
bool low_space = false;
int flags;
- xfs_ino_t ino = NULLFSINO;
+ int error = 0;
- /*
- * Directories, symlinks, and regular files frequently allocate at least
- * one block, so factor that potential expansion when we examine whether
- * an AG has enough space for file creation.
- */
- if (S_ISDIR(mode))
- start_agno = (atomic_inc_return(&mp->m_agirotor) - 1) %
- mp->m_maxagi;
- else {
- start_agno = XFS_INO_TO_AGNO(mp, parent);
- if (start_agno >= mp->m_maxagi)
- start_agno = 0;
- }
+ start_agno = xfs_dialloc_pick_ag(mp, dp, mode);
/*
* If we have already hit the ceiling of inode blocks then clear
diff --git a/libxfs/xfs_ialloc.h b/libxfs/xfs_ialloc.h
index f1412183bb4..9bfe2d8d84b 100644
--- a/libxfs/xfs_ialloc.h
+++ b/libxfs/xfs_ialloc.h
@@ -37,7 +37,7 @@ xfs_make_iptr(struct xfs_mount *mp, struct xfs_buf *b, int o)
* Allocate an inode on disk. Mode is used to tell whether the new inode will
* need space, and whether it is a directory.
*/
-int xfs_dialloc(struct xfs_trans **tpp, xfs_ino_t parent, umode_t mode,
+int xfs_dialloc(struct xfs_trans **tpp, struct xfs_inode *dp, umode_t mode,
xfs_ino_t *new_ino);
int xfs_difree(struct xfs_trans *tp, struct xfs_perag *pag,
diff --git a/libxfs/xfs_imeta.c b/libxfs/xfs_imeta.c
index b1c5c6ec5e6..2defee9562b 100644
--- a/libxfs/xfs_imeta.c
+++ b/libxfs/xfs_imeta.c
@@ -229,7 +229,7 @@ xfs_imeta_sb_create(
return -EEXIST;
/* Create a new inode and set the sb pointer. */
- error = xfs_dialloc(&upd->tp, 0, mode, &ino);
+ error = xfs_dialloc(&upd->tp, NULL, mode, &ino);
if (error)
return error;
error = xfs_icreate(upd->tp, ino, &args, &upd->ip);
@@ -661,7 +661,7 @@ xfs_imeta_dir_create(
* entry pointing to them, but a directory also the "." entry
* pointing to itself.
*/
- error = xfs_dialloc(&upd->tp, upd->dp->i_ino, mode, &ino);
+ error = xfs_dialloc(&upd->tp, upd->dp, mode, &ino);
if (error)
return error;
error = xfs_icreate(upd->tp, ino, &args, &upd->ip);
diff --git a/mkfs/proto.c b/mkfs/proto.c
index 5e17ea420f4..0103fe54a5d 100644
--- a/mkfs/proto.c
+++ b/mkfs/proto.c
@@ -431,7 +431,6 @@ creatproto(
XFS_ICREATE_ARGS_FORCE_MODE,
};
struct xfs_inode *ip;
- xfs_ino_t parent_ino = dp ? dp->i_ino : 0;
xfs_ino_t ino;
int error;
@@ -442,7 +441,7 @@ creatproto(
* Call the space management code to pick the on-disk inode to be
* allocated.
*/
- error = -libxfs_dialloc(tpp, parent_ino, mode, &ino);
+ error = -libxfs_dialloc(tpp, dp, mode, &ino);
if (error)
return error;
diff --git a/repair/phase6.c b/repair/phase6.c
index 6a3c5e2a37a..fe9a4da62dc 100644
--- a/repair/phase6.c
+++ b/repair/phase6.c
@@ -870,7 +870,7 @@ mk_orphanage(
if (i)
res_failed(i);
- error = -libxfs_dialloc(&tp, mp->m_sb.sb_rootino, mode, &ino);
+ error = -libxfs_dialloc(&tp, du.dp, mode, &ino);
if (error)
do_error(_("%s inode allocation failed %d\n"),
ORPHANAGE, error);
