ceph: add dir_layout to inode

Add a ceph_dir_layout to the inode, and calculate dentry hash values based
on the parent directory's specified dir_hash function.  This is needed
because the old default Linux dcache hash function is extremely week and
leads to a poor distribution of files among dir fragments.

Signed-off-by: Sage Weil <sage@newdream.net>

6 files changed, 41 insertions, 4 deletions

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index d902948..562f988 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -1216,6 +1216,26 @@ void ceph_dentry_lru_del(struct dentry *dn)
 	}
 }
 
+/*
+ * Return name hash for a given dentry.  This is dependent on
+ * the parent directory's hash function.
+ */
+unsigned ceph_dentry_hash(struct dentry *dn)
+{
+	struct inode *dir = dn->d_parent->d_inode;
+	struct ceph_inode_info *dci = ceph_inode(dir);
+
+	switch (dci->i_dir_layout.dl_dir_hash) {
+	case 0:	/* for backward compat */
+	case CEPH_STR_HASH_LINUX:
+		return dn->d_name.hash;
+
+	default:
+		return ceph_str_hash(dci->i_dir_layout.dl_dir_hash,
+				     dn->d_name.name, dn->d_name.len);
+	}
+}
+
 const struct file_operations ceph_dir_fops = {
 	.read = ceph_read_dir,
 	.readdir = ceph_readdir,
diff --git a/fs/ceph/export.c b/fs/ceph/export.c
index 2297d94..e410561 100644
--- a/fs/ceph/export.c
+++ b/fs/ceph/export.c
@@ -59,7 +59,7 @@ static int ceph_encode_fh(struct dentry *dentry, u32 *rawfh, int *max_len,
 		dout("encode_fh %p connectable\n", dentry);
 		cfh->ino = ceph_ino(dentry->d_inode);
 		cfh->parent_ino = ceph_ino(parent->d_inode);
-		cfh->parent_name_hash = parent->d_name.hash;
+		cfh->parent_name_hash = ceph_dentry_hash(parent);
 		*max_len = connected_handle_length;
 		type = 2;
 	} else if (*max_len >= handle_length) {
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index bf12865..045283c 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -297,6 +297,8 @@ struct inode *ceph_alloc_inode(struct super_block *sb)
 	ci->i_release_count = 0;
 	ci->i_symlink = NULL;
 
+	memset(&ci->i_dir_layout, 0, sizeof(ci->i_dir_layout));
+
 	ci->i_fragtree = RB_ROOT;
 	mutex_init(&ci->i_fragtree_mutex);
 
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 7f01728..6e08266 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -239,6 +239,7 @@ struct ceph_inode_info {
 	unsigned i_ceph_flags;
 	unsigned long i_release_count;
 
+	struct ceph_dir_layout i_dir_layout;
 	struct ceph_file_layout i_layout;
 	char *i_symlink;
 
@@ -768,6 +769,7 @@ extern void ceph_dentry_lru_add(struct dentry *dn);
 extern void ceph_dentry_lru_touch(struct dentry *dn);
 extern void ceph_dentry_lru_del(struct dentry *dn);
 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
+extern unsigned ceph_dentry_hash(struct dentry *dn);
 
 /*
  * our d_ops vary depending on whether the inode is live,
diff --git a/include/linux/ceph/ceph_fs.h b/include/linux/ceph/ceph_fs.h
index c3c74ae..09dcc0c 100644
--- a/include/linux/ceph/ceph_fs.h
+++ b/include/linux/ceph/ceph_fs.h
@@ -43,6 +43,10 @@
 #define CEPH_FEATURE_NOSRCADDR      (1<<1)
 #define CEPH_FEATURE_MONCLOCKCHECK  (1<<2)
 #define CEPH_FEATURE_FLOCK          (1<<3)
+#define CEPH_FEATURE_SUBSCRIBE2     (1<<4)
+#define CEPH_FEATURE_MONNAMES       (1<<5)
+#define CEPH_FEATURE_RECONNECT_SEQ  (1<<6)
+#define CEPH_FEATURE_DIRLAYOUTHASH  (1<<7)
 
 
 /*
@@ -55,10 +59,10 @@ struct ceph_file_layout {
 	__le32 fl_stripe_count;    /* over this many objects */
 	__le32 fl_object_size;     /* until objects are this big, then move to
 				      new objects */
-	__le32 fl_cas_hash;        /* 0 = none; 1 = sha256 */
+	__le32 fl_cas_hash;        /* UNUSED.  0 = none; 1 = sha256 */
 
 	/* pg -> disk layout */
-	__le32 fl_object_stripe_unit;  /* for per-object parity, if any */
+	__le32 fl_object_stripe_unit;  /* UNUSED.  for per-object parity, if any */
 
 	/* object -> pg layout */
 	__le32 fl_pg_preferred; /* preferred primary for pg (-1 for none) */
@@ -69,6 +73,12 @@ struct ceph_file_layout {
 
 int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
 
+struct ceph_dir_layout {
+	__u8   dl_dir_hash;   /* see ceph_hash.h for ids */
+	__u8   dl_unused1;
+	__u16  dl_unused2;
+	__u32  dl_unused3;
+} __attribute__ ((packed));
 
 /* crypto algorithms */
 #define CEPH_CRYPTO_NONE 0x0
@@ -457,7 +467,7 @@ struct ceph_mds_reply_inode {
 	struct ceph_timespec rctime;
 	struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
 } __attribute__ ((packed));
-/* followed by frag array, then symlink string, then xattr blob */
+/* followed by frag array, symlink string, dir layout, xattr blob */
 
 /* reply_lease follows dname, and reply_inode */
 struct ceph_mds_reply_lease {
diff --git a/net/ceph/ceph_hash.c b/net/ceph/ceph_hash.c
index 815ef88..0a1b53b 100644
--- a/net/ceph/ceph_hash.c
+++ b/net/ceph/ceph_hash.c
@@ -1,5 +1,6 @@
 
 #include <linux/ceph/types.h>
+#include <linux/module.h>
 
 /*
  * Robert Jenkin's hash function.
@@ -104,6 +105,7 @@ unsigned ceph_str_hash(int type, const char *s, unsigned len)
 		return -1;
 	}
 }
+EXPORT_SYMBOL(ceph_str_hash);
 
 const char *ceph_str_hash_name(int type)
 {
@@ -116,3 +118,4 @@ const char *ceph_str_hash_name(int type)
 		return "unknown";
 	}
 }
+EXPORT_SYMBOL(ceph_str_hash_name);