ceph: enable/disable dentry complete flags via mount option

Enable/disable use of the dentry dir 'complete' flag via a mount option. This lets the admin control whether ceph uses the dcache to satisfy negative lookups or readdir when it has the entire directory contents in its cache. This is purely a performance optimization; correctness is guaranteed whether it is enabled or not. Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sage Weil <sage@newdream.net>

ceph: enable/disable dentry complete flags via mount option
Enable/disable use of the dentry dir 'complete' flag via a mount option. This lets the admin control whether ceph uses the dcache to satisfy negative lookups or readdir when it has the entire directory contents in its cache. This is purely a performance optimization; correctness is guaranteed whether it is enabled or not. Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sage Weil <sage@newdream.net>
Sage Weil
1 parent 46f72b3492
Showing 4 changed files with 50 additions and 8 deletions Side-by-side Diff
Documentation/filesystems/ceph.txt
fs/ceph/dir.c
fs/ceph/super.c
fs/ceph/super.h
@@ -119,12 +119,20 @@
 	must rely on TCP's error correction to detect data corruption
 	in the data payload.
  
-  noasyncreaddir
-	Disable client's use its local cache to satisfy	readdir
-	requests.  (This does not change correctness; the client uses
-	cached metadata only when a lease or capability ensures it is
-	valid.)
+  dcache
+        Use the dcache contents to perform negative lookups and
+        readdir when the client has the entire directory contents in
+        its cache.  (This does not change correctness; the client uses
+        cached metadata only when a lease or capability ensures it is
+        valid.)
  
+  nodcache
+        Do not use the dcache as above.  This avoids a significant amount of
+        complex code, sacrificing performance without affecting correctness,
+        and is useful for tracking down bugs.
+
+  noasyncreaddir
+	Do not use the dcache as above for readdir.
  
 More Information
 ================
@@ -1094,17 +1094,36 @@
  */
 void ceph_dir_set_complete(struct inode *inode)
 {
-	/* not yet implemented */
+	struct dentry *dentry = d_find_any_alias(inode);
+	
+	if (dentry && ceph_dentry(dentry) &&
+	    ceph_test_mount_opt(ceph_sb_to_client(dentry->d_sb), DCACHE)) {
+		dout(" marking %p (%p) complete\n", inode, dentry);
+		set_bit(CEPH_D_COMPLETE, &ceph_dentry(dentry)->flags);
+	}
+	dput(dentry);
 }
  
 void ceph_dir_clear_complete(struct inode *inode)
 {
-	/* not yet implemented */
+	struct dentry *dentry = d_find_any_alias(inode);
+
+	if (dentry && ceph_dentry(dentry)) {
+		dout(" marking %p (%p) complete\n", inode, dentry);
+		set_bit(CEPH_D_COMPLETE, &ceph_dentry(dentry)->flags);
+	}
+	dput(dentry);
 }
  
 bool ceph_dir_test_complete(struct inode *inode)
 {
-	/* not yet implemented */
+	struct dentry *dentry = d_find_any_alias(inode);
+
+	if (dentry && ceph_dentry(dentry)) {
+		dout(" marking %p (%p) NOT complete\n", inode, dentry);
+		clear_bit(CEPH_D_COMPLETE, &ceph_dentry(dentry)->flags);
+	}
+	dput(dentry);
 	return false;
 }
  
@@ -131,6 +131,8 @@
 	Opt_rbytes,
 	Opt_norbytes,
 	Opt_noasyncreaddir,
+	Opt_dcache,
+	Opt_nodcache,
 	Opt_ino32,
 };
  
@@ -152,6 +154,8 @@
 	{Opt_rbytes, "rbytes"},
 	{Opt_norbytes, "norbytes"},
 	{Opt_noasyncreaddir, "noasyncreaddir"},
+	{Opt_dcache, "dcache"},
+	{Opt_nodcache, "nodcache"},
 	{Opt_ino32, "ino32"},
 	{-1, NULL}
 };
@@ -231,6 +235,12 @@
 	case Opt_noasyncreaddir:
 		fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
 		break;
+	case Opt_dcache:
+		fsopt->flags |= CEPH_MOUNT_OPT_DCACHE;
+		break;
+	case Opt_nodcache:
+		fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
+		break;
 	case Opt_ino32:
 		fsopt->flags |= CEPH_MOUNT_OPT_INO32;
 		break;
@@ -377,6 +387,10 @@
 		seq_puts(m, ",norbytes");
 	if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
 		seq_puts(m, ",noasyncreaddir");
+	if (fsopt->flags & CEPH_MOUNT_OPT_DCACHE)
+		seq_puts(m, ",dcache");
+	else
+		seq_puts(m, ",nodcache");
  
 	if (fsopt->wsize)
 		seq_printf(m, ",wsize=%d", fsopt->wsize);
@@ -28,6 +28,7 @@
 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
+#define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
  
 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES)
...	...	@@ -119,12 +119,20 @@
119	119	must rely on TCP's error correction to detect data corruption
120	120	in the data payload.
121	121
122		- noasyncreaddir
123		- Disable client's use its local cache to satisfy readdir
124		- requests. (This does not change correctness; the client uses
125		- cached metadata only when a lease or capability ensures it is
126		- valid.)
	122	+ dcache
	123	+ Use the dcache contents to perform negative lookups and
	124	+ readdir when the client has the entire directory contents in
	125	+ its cache. (This does not change correctness; the client uses
	126	+ cached metadata only when a lease or capability ensures it is
	127	+ valid.)
127	128
	129	+ nodcache
	130	+ Do not use the dcache as above. This avoids a significant amount of
	131	+ complex code, sacrificing performance without affecting correctness,
	132	+ and is useful for tracking down bugs.
	133	+
	134	+ noasyncreaddir
	135	+ Do not use the dcache as above for readdir.
128	136
129	137	More Information
130	138	================
...	...	@@ -1094,17 +1094,36 @@
1094	1094	*/
1095	1095	void ceph_dir_set_complete(struct inode *inode)
1096	1096	{
1097		- /* not yet implemented */
	1097	+ struct dentry *dentry = d_find_any_alias(inode);
	1098	+
	1099	+ if (dentry && ceph_dentry(dentry) &&
	1100	+ ceph_test_mount_opt(ceph_sb_to_client(dentry->d_sb), DCACHE)) {
	1101	+ dout(" marking %p (%p) complete\n", inode, dentry);
	1102	+ set_bit(CEPH_D_COMPLETE, &ceph_dentry(dentry)->flags);
	1103	+ }
	1104	+ dput(dentry);
1098	1105	}
1099	1106
1100	1107	void ceph_dir_clear_complete(struct inode *inode)
1101	1108	{
1102		- /* not yet implemented */
	1109	+ struct dentry *dentry = d_find_any_alias(inode);
	1110	+
	1111	+ if (dentry && ceph_dentry(dentry)) {
	1112	+ dout(" marking %p (%p) complete\n", inode, dentry);
	1113	+ set_bit(CEPH_D_COMPLETE, &ceph_dentry(dentry)->flags);
	1114	+ }
	1115	+ dput(dentry);
1103	1116	}
1104	1117
1105	1118	bool ceph_dir_test_complete(struct inode *inode)
1106	1119	{
1107		- /* not yet implemented */
	1120	+ struct dentry *dentry = d_find_any_alias(inode);
	1121	+
	1122	+ if (dentry && ceph_dentry(dentry)) {
	1123	+ dout(" marking %p (%p) NOT complete\n", inode, dentry);
	1124	+ clear_bit(CEPH_D_COMPLETE, &ceph_dentry(dentry)->flags);
	1125	+ }
	1126	+ dput(dentry);
1108	1127	return false;
1109	1128	}
1110	1129
...	...	@@ -131,6 +131,8 @@
131	131	Opt_rbytes,
132	132	Opt_norbytes,
133	133	Opt_noasyncreaddir,
	134	+ Opt_dcache,
	135	+ Opt_nodcache,
134	136	Opt_ino32,
135	137	};
136	138
...	...	@@ -152,6 +154,8 @@
152	154	{Opt_rbytes, "rbytes"},
153	155	{Opt_norbytes, "norbytes"},
154	156	{Opt_noasyncreaddir, "noasyncreaddir"},
	157	+ {Opt_dcache, "dcache"},
	158	+ {Opt_nodcache, "nodcache"},
155	159	{Opt_ino32, "ino32"},
156	160	{-1, NULL}
157	161	};
...	...	@@ -231,6 +235,12 @@
231	235	case Opt_noasyncreaddir:
232	236	fsopt->flags \|= CEPH_MOUNT_OPT_NOASYNCREADDIR;
233	237	break;
	238	+ case Opt_dcache:
	239	+ fsopt->flags \|= CEPH_MOUNT_OPT_DCACHE;
	240	+ break;
	241	+ case Opt_nodcache:
	242	+ fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
	243	+ break;
234	244	case Opt_ino32:
235	245	fsopt->flags \|= CEPH_MOUNT_OPT_INO32;
236	246	break;
...	...	@@ -377,6 +387,10 @@
377	387	seq_puts(m, ",norbytes");
378	388	if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
379	389	seq_puts(m, ",noasyncreaddir");
	390	+ if (fsopt->flags & CEPH_MOUNT_OPT_DCACHE)
	391	+ seq_puts(m, ",dcache");
	392	+ else
	393	+ seq_puts(m, ",nodcache");
380	394
381	395	if (fsopt->wsize)
382	396	seq_printf(m, ",wsize=%d", fsopt->wsize);
...	...	@@ -28,6 +28,7 @@
28	28	#define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
29	29	#define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
30	30	#define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
	31	+#define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
31	32
32	33	#define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES)
33	34