[PATCH 16/36] libxfs: stop caching inode structures

[Dave Chinner <david@fromorbit.com>]

From: Christoph Hellwig <hch@infradead.org>

Currently libxfs has a cache for xfs_inode structures.  Unlike in kernelspace
where the inode cache, and the associated page cache for file data is used
for all filesystem operations the libxfs inode cache is only used in few
places:

 - the libxfs init code reads the root and realtime inodes when called from
   xfs_db using a special flag, but these inode structure are never referenced
   again
 - mkfs uses namespace and bmap routines that take the xfs_inode structure
   to create the root and realtime inodes, as well as any additional files
   specified in the proto file
 - the xfs_db attr code uses xfs_inode-based attr routines in the attrset
   and attrget commands
 - phase6 of xfs_repair uses xfs_inode-based routines for rebuilding
   directories and moving files to the lost+found directory.
 - phase7 of xfs_repair uses struct xfs_inode to modify the nlink count
   of inodes.

So except in repair we never ever reuse a cached inode, and even in repair
the logical inode caching doesn't help:

 - in phase 6a we iterate over each inode in the incore inode tree,
   and if it's a directory check/rebuild it
 - phase6b then updates the "." and ".." entries for directories
   that need, which means we require the backing buffers.
 - phase6c moves disconnected inodes to lost_found, which again needs
   the backing buffer to actually do anything.
 - phase7 then only touches inodes for which we need to reset i_nlink,
   which always involves reading, modifying and writing the physical
   inode.
   which always involves modifying the . and .. entries.

Given these facts stop caching the inodes to reduce memory usage
especially in xfs_repair, where this makes a different for large inode
count inodes.  On the upper end this allows repair to complete for
filesystem / amount of memory combinations that previously wouldn't.

With this we probably could increase the memory available to the buffer
cache in xfs_repair, but trying to do so I got a bit lost - the current
formula seems to magic to me to make any sense, and simply doubling the
buffer cache size causes us to run out of memory given that the data cached
in the buffer cache (typically lots of 8k inode buffers and few 4k other
metadata buffers) are much bigger than the inodes cached in the inode
cache.  We probably need a sizing scheme that takes the actual amount
of memory allocated to the buffer cache into account to solve this better.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
 include/libxfs.h      |  5 ---
 libxfs/init.c         |  9 ------
 libxfs/rdwr.c         | 87 ++++++++++++---------------------------------------
 man/man8/xfs_repair.8 |  6 ----
 mkfs/xfs_mkfs.c       |  1 -
 repair/xfs_repair.c   | 14 ++-------
 6 files changed, 23 insertions(+), 99 deletions(-)

diff --git a/include/libxfs.h b/include/libxfs.h
index 3df8c07..e017b32 100644
--- a/include/libxfs.h
+++ b/include/libxfs.h
@@ -264,7 +264,6 @@ typedef struct xfs_perag {
 #define LIBXFS_MOUNT_COMPAT_ATTR	0x0008
 #define LIBXFS_MOUNT_ATTR2		0x0010
 
-#define LIBXFS_IHASHSIZE(sbp)		(1<<10)
 #define LIBXFS_BHASHSIZE(sbp) 		(1<<10)
 
 extern xfs_mount_t	*libxfs_mount (xfs_mount_t *, xfs_sb_t *,
@@ -448,7 +447,6 @@ extern int	libxfs_writebuf_int(xfs_buf_t *, int);
 extern int	libxfs_readbufr(struct xfs_buftarg *, xfs_daddr_t, xfs_buf_t *, int, int);
 
 extern int libxfs_bhash_size;
-extern int libxfs_ihash_size;
 
 #define LIBXFS_BREAD	0x1
 #define LIBXFS_BWRITE	0x2
@@ -648,9 +646,6 @@ extern void	libxfs_trans_ichgtime(struct xfs_trans *,
 extern int	libxfs_iflush_int (xfs_inode_t *, xfs_buf_t *);
 
 /* Inode Cache Interfaces */
-extern struct cache	*libxfs_icache;
-extern struct cache_operations	libxfs_icache_operations;
-extern void	libxfs_icache_purge (void);
 extern int	libxfs_iget (xfs_mount_t *, xfs_trans_t *, xfs_ino_t,
 				uint, xfs_inode_t **, xfs_daddr_t);
 extern void	libxfs_iput (xfs_inode_t *, uint);
diff --git a/libxfs/init.c b/libxfs/init.c
index 33c01f5..9a3cf22 100644
--- a/libxfs/init.c
+++ b/libxfs/init.c
@@ -22,9 +22,6 @@
 
 char *progname = "libxfs";	/* default, changed by each tool */
 
-struct cache *libxfs_icache;	/* global inode cache */
-int libxfs_ihash_size;		/* #buckets in icache */
-
 struct cache *libxfs_bcache;	/* global buffer cache */
 int libxfs_bhash_size;		/* #buckets in bcache */
 
@@ -335,9 +332,6 @@ libxfs_init(libxfs_init_t *a)
 	}
 	if (needcd)
 		chdir(curdir);
-	if (!libxfs_ihash_size)
-		libxfs_ihash_size = LIBXFS_IHASHSIZE(sbp);
-	libxfs_icache = cache_init(libxfs_ihash_size, &libxfs_icache_operations);
 	if (!libxfs_bhash_size)
 		libxfs_bhash_size = LIBXFS_BHASHSIZE(sbp);
 	libxfs_bcache = cache_init(libxfs_bhash_size, &libxfs_bcache_operations);
@@ -799,7 +793,6 @@ libxfs_umount(xfs_mount_t *mp)
 	int			agno;
 
 	libxfs_rtmount_destroy(mp);
-	libxfs_icache_purge();
 	libxfs_bcache_purge();
 
 	for (agno = 0; agno < mp->m_maxagi; agno++) {
@@ -815,7 +808,6 @@ void
 libxfs_destroy(void)
 {
 	manage_zones(1);
-	cache_destroy(libxfs_icache);
 	cache_destroy(libxfs_bcache);
 }
 
@@ -831,7 +823,6 @@ libxfs_report(FILE *fp)
 	time_t t;
 	char *c;
 
-	cache_report(fp, "libxfs_icache", libxfs_icache);
 	cache_report(fp, "libxfs_bcache", libxfs_bcache);
 
 	t = time(NULL);
diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c
index 13dbd23..f507855 100644
--- a/libxfs/rdwr.c
+++ b/libxfs/rdwr.c
@@ -993,26 +993,12 @@ struct cache_operations libxfs_bcache_operations = {
 
 
 /*
- * Inode cache interfaces
+ * Inode cache stubs.
  */
 
 extern kmem_zone_t	*xfs_ili_zone;
 extern kmem_zone_t	*xfs_inode_zone;
 
-static unsigned int
-libxfs_ihash(cache_key_t key, unsigned int hashsize)
-{
-	return ((unsigned int)*(xfs_ino_t *)key) % hashsize;
-}
-
-static int
-libxfs_icompare(struct cache_node *node, cache_key_t key)
-{
-	xfs_inode_t	*ip = (xfs_inode_t *)node;
-
-	return (ip->i_ino == *(xfs_ino_t *)key);
-}
-
 int
 libxfs_iget(xfs_mount_t *mp, xfs_trans_t *tp, xfs_ino_t ino, uint lock_flags,
 		xfs_inode_t **ipp, xfs_daddr_t bno)
@@ -1020,34 +1006,21 @@ libxfs_iget(xfs_mount_t *mp, xfs_trans_t *tp, xfs_ino_t ino, uint lock_flags,
 	xfs_inode_t	*ip;
 	int		error = 0;
 
-	if (cache_node_get(libxfs_icache, &ino, (struct cache_node **)&ip)) {
-#ifdef INO_DEBUG
-		fprintf(stderr, "%s: allocated inode, ino=%llu(%llu), %p\n",
-			__FUNCTION__, (unsigned long long)ino, bno, ip);
-#endif
-		ip->i_ino = ino;
-		ip->i_mount = mp;
-		error = xfs_iread(mp, tp, ip, bno);
-		if (error) {
-			cache_node_purge(libxfs_icache, &ino,
-					(struct cache_node *)ip);
-			ip = NULL;
-		}
-	}
-	*ipp = ip;
-	return error;
-}
+	ip = kmem_zone_zalloc(xfs_inode_zone, 0);
+	if (!ip)
+		return ENOMEM;
 
-void
-libxfs_iput(xfs_inode_t *ip, uint lock_flags)
-{
-	cache_node_put(libxfs_icache, (struct cache_node *)ip);
-}
+	ip->i_ino = ino;
+	ip->i_mount = mp;
+	error = xfs_iread(mp, tp, ip, bno);
+	if (error) {
+		kmem_zone_free(xfs_inode_zone, ip);
+		*ipp = NULL;
+		return error;
+	}
 
-static struct cache_node *
-libxfs_ialloc(cache_key_t key)
-{
-	return kmem_zone_zalloc(xfs_inode_zone, 0);
+	*ipp = ip;
+	return 0;
 }
 
 static void
@@ -1064,32 +1037,12 @@ libxfs_idestroy(xfs_inode_t *ip)
 		libxfs_idestroy_fork(ip, XFS_ATTR_FORK);
 }
 
-static void
-libxfs_irelse(struct cache_node *node)
-{
-	xfs_inode_t	*ip = (xfs_inode_t *)node;
-
-	if (ip != NULL) {
-		if (ip->i_itemp)
-			kmem_zone_free(xfs_ili_zone, ip->i_itemp);
-		ip->i_itemp = NULL;
-		libxfs_idestroy(ip);
-		kmem_zone_free(xfs_inode_zone, ip);
-		ip = NULL;
-	}
-}
-
 void
-libxfs_icache_purge(void)
+libxfs_iput(xfs_inode_t *ip, uint lock_flags)
 {
-	cache_purge(libxfs_icache);
+	if (ip->i_itemp)
+		kmem_zone_free(xfs_ili_zone, ip->i_itemp);
+	ip->i_itemp = NULL;
+	libxfs_idestroy(ip);
+	kmem_zone_free(xfs_inode_zone, ip);
 }
-
-struct cache_operations libxfs_icache_operations = {
-	/* .hash */	libxfs_ihash,
-	/* .alloc */	libxfs_ialloc,
-	/* .flush */	NULL,
-	/* .relse */	libxfs_irelse,
-	/* .compare */	libxfs_icompare,
-	/* .bulkrelse */ NULL
-};
diff --git a/man/man8/xfs_repair.8 b/man/man8/xfs_repair.8
index 96adb29..47436ec 100644
--- a/man/man8/xfs_repair.8
+++ b/man/man8/xfs_repair.8
@@ -130,12 +130,6 @@ The
 supported are:
 .RS 1.0i
 .TP
-.BI ihash= ihashsize
-overrides the default inode cache hash size. The total number of
-inode cache entries are limited to 8 times this amount. The default
-.I ihashsize
-is 1024 (for a total of 8192 entries).
-.TP
 .BI bhash= bhashsize
 overrides the default buffer cache hash size. The total number of
 buffer cache entries are limited to 8 times this amount. The default
diff --git a/mkfs/xfs_mkfs.c b/mkfs/xfs_mkfs.c
index d37e948..3a032c0 100644
--- a/mkfs/xfs_mkfs.c
+++ b/mkfs/xfs_mkfs.c
@@ -2935,7 +2935,6 @@ _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
 	 * Need to drop references to inodes we still hold, first.
 	 */
 	libxfs_rtmount_destroy(mp);
-	libxfs_icache_purge();
 	libxfs_bcache_purge();
 
 	/*
diff --git a/repair/xfs_repair.c b/repair/xfs_repair.c
index 820e7a2..214b7fa 100644
--- a/repair/xfs_repair.c
+++ b/repair/xfs_repair.c
@@ -69,7 +69,6 @@ static char *c_opts[] = {
 };
 
 
-static int	ihash_option_used;
 static int	bhash_option_used;
 static long	max_mem_specified;	/* in megabytes */
 static int	phase2_threads = 32;
@@ -239,13 +238,13 @@ process_args(int argc, char **argv)
 					pre_65_beta = 1;
 					break;
 				case IHASH_SIZE:
-					libxfs_ihash_size = (int)strtol(val, NULL, 0);
-					ihash_option_used = 1;
+					do_warn(
+		_("-o ihash option has been removed and will be ignored\n"));
 					break;
 				case BHASH_SIZE:
 					if (max_mem_specified)
 						do_abort(
-			_("-o bhash option cannot be used with -m option\n"));
+		_("-o bhash option cannot be used with -m option\n"));
 					libxfs_bhash_size = (int)strtol(val, NULL, 0);
 					bhash_option_used = 1;
 					break;
@@ -648,9 +647,7 @@ main(int argc, char **argv)
 		unsigned long	max_mem;
 		struct rlimit	rlim;
 
-		libxfs_icache_purge();
 		libxfs_bcache_purge();
-		cache_destroy(libxfs_icache);
 		cache_destroy(libxfs_bcache);
 
 		mem_used = (mp->m_sb.sb_icount >> (10 - 2)) +
@@ -709,11 +706,6 @@ main(int argc, char **argv)
 			do_log(_("        - block cache size set to %d entries\n"),
 				libxfs_bhash_size * HASH_CACHE_RATIO);
 
-		if (!ihash_option_used)
-			libxfs_ihash_size = libxfs_bhash_size;
-
-		libxfs_icache = cache_init(libxfs_ihash_size,
-						&libxfs_icache_operations);
 		libxfs_bcache = cache_init(libxfs_bhash_size,
 						&libxfs_bcache_operations);
 	}
-- 
1.8.4.rc3

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs