[RFC][PATCH 00/10] cifs: local caching support using FS-Cache

[RFC][PATCH 00/10] cifs: local caching support using FS-Cache

[Suresh Jayaraman]

This patchset is a first stab at adding persistent, local caching facility for
CIFS using the FS-Cache interface.

The index hierarchy which is mainly used to locate a file object or discard
a certain subset of the files cached, currently has three levels:
	- Server
	- Share 
	- File

The server index object is keyed by hostname of the server. The superblock
index object is keyed by the sharename and the inode object is keyed by the
UniqueId. The cache coherency is ensured by checking the 'LastWriteTime' and
size of file.

To use this, apply this patchset in order, mount the share with rsize=4096 and
try copying a huge file (say few hundred MBs) from mount point to local
filesystem. During the first time, the cache will be initialized. When you copy
the second time, it should read from the local cache.

To reduce the impact of page cache and see the local caching in action
readily, try doing a sync and drop the caches by doing:
	sync; echo 3 > /proc/sys/vm/drop_caches

Known issues
-------------
	- the cache coherency check may not be reliable always as some
	  CIFS servers are known not to update mtime until the filehandle is
	  closed.
	- not all the Servers under all circumstances provide a unique
	  'UniqueId'.

Todo's
-------
	- improvements to avoid potential key collisions
	- address the above known issues

This set is lightly tested and all the bugs seen during my testing have been
fixed. However, this can be considered as an RFC for now.

Any Comments or Suggestions are welcome.

Suresh Jayaraman (10)
  cifs: add kernel config option for CIFS Client caching support
  cifs: guard cifsglob.h against multiple inclusion
  cifs: register CIFS for caching
  cifs: define server-level cache index objects and register them with FS-Cache
  cifs: define superblock-level cache index objects and register them
  cifs: define inode-level cache object and register them
  cifs: FS-Cache page management
  cifs: store pages into local cache
  cifs: read pages from FS-Cache
  cifs: add mount option to enable local caching

 Kconfig      |    9 ++
 Makefile     |    2 
 cache.c      |  251 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 cifs_fs_sb.h |    1 
 cifsfs.c     |   15 +++
 cifsglob.h   |   14 +++
 connect.c    |   16 +++
 file.c       |   51 +++++++++++
 fscache.c    |  244 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 fscache.h    |  135 +++++++++++++++++++++++++++++++
 inode.c      |    4 
 11 files changed, 742 insertions(+)

[RFC][PATCH 01/10] cifs: add kernel config option for CIFS Client caching support

[Suresh Jayaraman]

Add a kernel config option to enable local caching for CIFS.

Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>
---
 fs/cifs/Kconfig |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/fs/cifs/Kconfig b/fs/cifs/Kconfig
index 80f3525..5739fd7 100644
--- a/fs/cifs/Kconfig
+++ b/fs/cifs/Kconfig
@@ -131,6 +131,15 @@ config CIFS_DFS_UPCALL
 	    IP addresses) which is needed for implicit mounts of DFS junction
 	    points. If unsure, say N.
 
+config CIFS_FSCACHE
+	  bool "Provide CIFS client caching support (EXPERIMENTAL)"
+	  depends on EXPERIMENTAL
+	  depends on CIFS=m && FSCACHE || CIFS=y && FSCACHE=y
+	  help
+	    Makes CIFS FS-Cache capable. Say Y here if you want your CIFS data
+	    to be cached locally on disk through the general filesystem cache
+	    manager. If unsure, say N.
+
 config CIFS_EXPERIMENTAL
 	  bool "CIFS Experimental Features (EXPERIMENTAL)"
 	  depends on CIFS && EXPERIMENTAL
-- 
1.6.4.2

[RFC][PATCH 02/10] cifs: guard cifsglob.h against multiple inclusion

[Suresh Jayaraman]

Add conditional compile macros to guard the header file against multiple
inclusion.

Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>
---
 fs/cifs/cifsglob.h |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
index a88479c..6b2c39d 100644
--- a/fs/cifs/cifsglob.h
+++ b/fs/cifs/cifsglob.h
@@ -16,6 +16,9 @@
  *   the GNU Lesser General Public License for more details.
  *
  */
+#ifndef _CIFS_GLOB_H
+#define _CIFS_GLOB_H
+
 #include <linux/in.h>
 #include <linux/in6.h>
 #include <linux/slab.h>
@@ -733,3 +736,5 @@ GLOBAL_EXTERN unsigned int cifs_min_small;  /* min size of small buf pool */
 GLOBAL_EXTERN unsigned int cifs_max_pending; /* MAX requests at once to server*/
 
 extern const struct slow_work_ops cifs_oplock_break_ops;
+
+#endif	/* _CIFS_GLOB_H */
-- 
1.6.4.2

[RFC][PATCH 03/10] cifs: register CIFS for caching

[Suresh Jayaraman]

Define CIFS for FS-Cache and register for caching. Upon registration the
top-level index object cookie will be stuck to the netfs definition by
FS-Cache.

Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
---
 fs/cifs/Makefile  |    2 ++
 fs/cifs/cache.c   |   53 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/cifs/cifsfs.c  |    8 ++++++++
 fs/cifs/fscache.h |   40 ++++++++++++++++++++++++++++++++++++++++
 4 files changed, 103 insertions(+), 0 deletions(-)
 create mode 100644 fs/cifs/cache.c
 create mode 100644 fs/cifs/fscache.h

diff --git a/fs/cifs/Makefile b/fs/cifs/Makefile
index 9948c00..e2de709 100644
--- a/fs/cifs/Makefile
+++ b/fs/cifs/Makefile
@@ -11,3 +11,5 @@ cifs-y := cifsfs.o cifssmb.o cifs_debug.o connect.o dir.o file.o inode.o \
 cifs-$(CONFIG_CIFS_UPCALL) += cifs_spnego.o
 
 cifs-$(CONFIG_CIFS_DFS_UPCALL) += dns_resolve.o cifs_dfs_ref.o
+
+cifs-$(CONFIG_CIFS_FSCACHE) += cache.o
diff --git a/fs/cifs/cache.c b/fs/cifs/cache.c
new file mode 100644
index 0000000..1080b96
--- /dev/null
+++ b/fs/cifs/cache.c
@@ -0,0 +1,53 @@
+/*
+ *   fs/cifs/cache.c - CIFS filesystem cache index structure definitions
+ *
+ *   Copyright (c) 2010 Novell, Inc.
+ *   Authors(s): Suresh Jayaraman (sjayaraman-l3A5Bk7waGM@public.gmane.org>
+ *
+ *   This library is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU Lesser General Public License as published
+ *   by the Free Software Foundation; either version 2.1 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This library is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ *   the GNU Lesser General Public License for more details.
+ *
+ *   You should have received a copy of the GNU Lesser General Public License
+ *   along with this library; if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+
+#include "fscache.h"
+#include "cifsglob.h"
+#include "cifs_debug.h"
+
+/*
+ * CIFS filesystem definition for FS-Cache
+ */
+struct fscache_netfs cifs_fscache_netfs = {
+	.name = "cifs",
+	.version = 0,
+};
+
+/*
+ * Register CIFS for caching with FS-Cache
+ */
+int cifs_fscache_register(void)
+{
+	return fscache_register_netfs(&cifs_fscache_netfs);
+}
+
+/*
+ * Unregister CIFS for caching
+ */
+void cifs_fscache_unregister(void)
+{
+	fscache_unregister_netfs(&cifs_fscache_netfs);
+}
+
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index 484e52b..c2a7aa9 100644
--- a/fs/cifs/cifsfs.c
+++ b/fs/cifs/cifsfs.c
@@ -47,6 +47,7 @@
 #include <linux/key-type.h>
 #include "dns_resolve.h"
 #include "cifs_spnego.h"
+#include "fscache.h"
 #define CIFS_MAGIC_NUMBER 0xFF534D42	/* the first four bytes of SMB PDUs */
 
 int cifsFYI = 0;
@@ -902,6 +903,10 @@ init_cifs(void)
 		cFYI(1, "cifs_max_pending set to max of 256");
 	}
 
+	rc = cifs_fscache_register();
+	if (rc)
+		goto out;
+
 	rc = cifs_init_inodecache();
 	if (rc)
 		goto out_clean_proc;
@@ -949,8 +954,10 @@ init_cifs(void)
 	cifs_destroy_mids();
  out_destroy_inodecache:
 	cifs_destroy_inodecache();
+	cifs_fscache_unregister();
  out_clean_proc:
 	cifs_proc_clean();
+ out:
 	return rc;
 }
 
@@ -959,6 +966,7 @@ exit_cifs(void)
 {
 	cFYI(DBG2, "exit_cifs");
 	cifs_proc_clean();
+	cifs_fscache_unregister();
 #ifdef CONFIG_CIFS_DFS_UPCALL
 	cifs_dfs_release_automount_timer();
 	unregister_key_type(&key_type_dns_resolver);
diff --git a/fs/cifs/fscache.h b/fs/cifs/fscache.h
new file mode 100644
index 0000000..cec9e2b
--- /dev/null
+++ b/fs/cifs/fscache.h
@@ -0,0 +1,40 @@
+/*
+ *   fs/cifs/fscache.h - CIFS filesystem cache interface definitions
+ *
+ *   Copyright (c) 2010 Novell, Inc.
+ *   Authors(s): Suresh Jayaraman (sjayaraman-l3A5Bk7waGM@public.gmane.org>
+ *
+ *   This library is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU Lesser General Public License as published
+ *   by the Free Software Foundation; either version 2.1 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This library is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ *   the GNU Lesser General Public License for more details.
+ *
+ *   You should have received a copy of the GNU Lesser General Public License
+ *   along with this library; if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#ifndef _CIFS_FSCACHE_H
+#define _CIFS_FSCACHE_H
+
+#include <linux/fscache.h>
+#include "cifsglob.h"
+
+#ifdef CONFIG_CIFS_FSCACHE
+
+extern struct fscache_netfs cifs_fscache_netfs;
+
+extern int cifs_fscache_register(void);
+extern void cifs_fscache_unregister(void);
+
+#else /* CONFIG_CIFS_FSCACHE */
+static inline int cifs_fscache_register(void) { return 0; }
+static inline void cifs_fscache_unregister(void) {}
+
+#endif /* CONFIG_CIFS_FSCACHE */
+
+#endif /* _CIFS_FSCACHE_H */
-- 
1.6.4.2

[RFC][PATCH 04/10] cifs: define server-level cache index objects and register them with FS-Cache

[Suresh Jayaraman]

Define server-level cache index objects (as managed by TCP_ServerInfo structs).
Each server object is created in the CIFS top-level index object and is itself
an index into which superblock-level objects are inserted.

Currently, the server objects are keyed by hostname.

Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
---
 fs/cifs/Makefile   |    2 +-
 fs/cifs/cache.c    |   25 +++++++++++++++++++++++++
 fs/cifs/cifsglob.h |    3 +++
 fs/cifs/connect.c  |    4 ++++
 fs/cifs/fscache.c  |   47 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/cifs/fscache.h  |   12 ++++++++++++
 6 files changed, 92 insertions(+), 1 deletion(-)
 create mode 100644 fs/cifs/fscache.c

Index: cifs-2.6/fs/cifs/Makefile
===================================================================
--- cifs-2.6.orig/fs/cifs/Makefile
+++ cifs-2.6/fs/cifs/Makefile
@@ -12,4 +12,4 @@ cifs-$(CONFIG_CIFS_UPCALL) += cifs_spneg
 
 cifs-$(CONFIG_CIFS_DFS_UPCALL) += dns_resolve.o cifs_dfs_ref.o
 
-cifs-$(CONFIG_CIFS_FSCACHE) += cache.o
+cifs-$(CONFIG_CIFS_FSCACHE) += fscache.o cache.o
Index: cifs-2.6/fs/cifs/cache.c
===================================================================
--- cifs-2.6.orig/fs/cifs/cache.c
+++ cifs-2.6/fs/cifs/cache.c
@@ -51,3 +51,28 @@ void cifs_fscache_unregister(void)
 	fscache_unregister_netfs(&cifs_fscache_netfs);
 }
 
+/*
+ * Server object currently keyed by hostname
+ */
+static uint16_t cifs_server_get_key(const void *cookie_netfs_data,
+				   void *buffer, uint16_t maxbuf)
+{
+	const struct TCP_Server_Info *server = cookie_netfs_data;
+	uint16_t len = strnlen(server->hostname, sizeof(server->hostname));
+
+	if (len > maxbuf)
+		return 0;
+
+	memcpy(buffer, server->hostname, len);
+
+	return len;
+}
+
+/*
+ * Server object for FS-Cache
+ */
+const struct fscache_cookie_def cifs_fscache_server_index_def = {
+	.name = "CIFS.server",
+	.type = FSCACHE_COOKIE_TYPE_INDEX,
+	.get_key = cifs_server_get_key,
+};
Index: cifs-2.6/fs/cifs/cifsglob.h
===================================================================
--- cifs-2.6.orig/fs/cifs/cifsglob.h
+++ cifs-2.6/fs/cifs/cifsglob.h
@@ -193,6 +193,9 @@ struct TCP_Server_Info {
 	bool	sec_mskerberos;		/* supports legacy MS Kerberos */
 	bool	sec_kerberosu2u;	/* supports U2U Kerberos */
 	bool	sec_ntlmssp;		/* supports NTLMSSP */
+#ifdef CONFIG_CIFS_FSCACHE
+	struct fscache_cookie   *fscache; /* client index cache cookie */
+#endif
 };
 
 /*
Index: cifs-2.6/fs/cifs/connect.c
===================================================================
--- cifs-2.6.orig/fs/cifs/connect.c
+++ cifs-2.6/fs/cifs/connect.c
@@ -48,6 +48,7 @@
 #include "nterr.h"
 #include "rfc1002pdu.h"
 #include "cn_cifs.h"
+#include "fscache.h"
 
 #define CIFS_PORT 445
 #define RFC1001_PORT 139
@@ -1453,6 +1454,8 @@ cifs_put_tcp_session(struct TCP_Server_I
 		return;
 	}
 
+	cifs_fscache_release_client_cookie(server);
+
 	list_del_init(&server->tcp_ses_list);
 	write_unlock(&cifs_tcp_ses_lock);
 
@@ -1572,6 +1575,7 @@ cifs_get_tcp_session(struct smb_vol *vol
 		goto out_err;
 	}
 
+	cifs_fscache_get_client_cookie(tcp_ses);
 	/* thread spawned, put it on the list */
 	write_lock(&cifs_tcp_ses_lock);
 	list_add(&tcp_ses->tcp_ses_list, &cifs_tcp_ses_list);
Index: cifs-2.6/fs/cifs/fscache.c
===================================================================
--- /dev/null
+++ cifs-2.6/fs/cifs/fscache.c
@@ -0,0 +1,47 @@
+/*
+ *   fs/cifs/fscache.c - CIFS filesystem cache interface
+ *
+ *   Copyright (c) 2010 Novell, Inc.
+ *   Authors(s): Suresh Jayaraman (sjayaraman-l3A5Bk7waGM@public.gmane.org>
+ *
+ *   This library is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU Lesser General Public License as published
+ *   by the Free Software Foundation; either version 2.1 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This library is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ *   the GNU Lesser General Public License for more details.
+ *
+ *   You should have received a copy of the GNU Lesser General Public License
+ *   along with this library; if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/in6.h>
+
+#include "fscache.h"
+#include "cifsglob.h"
+#include "cifs_debug.h"
+
+void cifs_fscache_get_client_cookie(struct TCP_Server_Info *server)
+{
+	server->fscache =
+		fscache_acquire_cookie(cifs_fscache_netfs.primary_index,
+				&cifs_fscache_server_index_def, server);
+	cFYI(1, "CIFS: get client cookie (0x%p/0x%p)\n",
+				server, server->fscache);
+}
+
+void cifs_fscache_release_client_cookie(struct TCP_Server_Info *server)
+{
+	cFYI(1, "CIFS: release client cookie (0x%p/0x%p)\n",
+				server, server->fscache);
+	fscache_relinquish_cookie(server->fscache, 0);
+	server->fscache = NULL;
+}
+
Index: cifs-2.6/fs/cifs/fscache.h
===================================================================
--- cifs-2.6.orig/fs/cifs/fscache.h
+++ cifs-2.6/fs/cifs/fscache.h
@@ -27,14 +27,26 @@
 #ifdef CONFIG_CIFS_FSCACHE
 
 extern struct fscache_netfs cifs_fscache_netfs;
+extern const struct fscache_cookie_def cifs_fscache_server_index_def;
 
 extern int cifs_fscache_register(void);
 extern void cifs_fscache_unregister(void);
 
+/*
+ * fscache.c
+ */
+extern void cifs_fscache_get_client_cookie(struct TCP_Server_Info *);
+extern void cifs_fscache_release_client_cookie(struct TCP_Server_Info *);
+
 #else /* CONFIG_CIFS_FSCACHE */
 static inline int cifs_fscache_register(void) { return 0; }
 static inline void cifs_fscache_unregister(void) {}
 
+static inline void
+cifs_fscache_get_client_cookie(struct TCP_Server_Info *server) {}
+static inline void
+cifs_fscache_get_client_cookie(struct TCP_Server_Info *server); {}
+
 #endif /* CONFIG_CIFS_FSCACHE */
 
 #endif /* _CIFS_FSCACHE_H */

[RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[Suresh Jayaraman]

Define superblock-level cache index objects (managed by cifsTconInfo structs).
Each superblock object is created in a server-level index object and in itself
an index into which inode-level objects are inserted.

Currently, the superblock objects are keyed by sharename.

Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
---
 fs/cifs/cache.c    |   62 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/cifs/cifsglob.h |    3 ++
 fs/cifs/connect.c  |    4 +++
 fs/cifs/fscache.c  |   17 ++++++++++++++
 fs/cifs/fscache.h  |    6 +++++
 5 files changed, 92 insertions(+)

Index: cifs-2.6/fs/cifs/cache.c
===================================================================
--- cifs-2.6.orig/fs/cifs/cache.c
+++ cifs-2.6/fs/cifs/cache.c
@@ -76,3 +76,65 @@ const struct fscache_cookie_def cifs_fsc
 	.type = FSCACHE_COOKIE_TYPE_INDEX,
 	.get_key = cifs_server_get_key,
 };
+
+static char *extract_sharename(const char *treename)
+{
+	const char *src;
+	char *delim, *dst;
+	int len;
+
+	/* skip double chars at the beginning */
+	src = treename + 2;
+
+	/* share name is always preceded by '\\' now */
+	delim = strchr(src, '\\');
+	if (!delim)
+		return ERR_PTR(-EINVAL);
+	delim++;
+	len = strlen(delim);
+
+	/* caller has to free the memory */
+	dst = kstrndup(delim, len, GFP_KERNEL);
+	if (!dst)
+		return ERR_PTR(-ENOMEM);
+
+	return dst;
+}
+
+/*
+ * Superblock object currently keyed by share name
+ */
+static uint16_t cifs_super_get_key(const void *cookie_netfs_data, void *buffer,
+				   uint16_t maxbuf)
+{
+	const struct cifsTconInfo *tcon = cookie_netfs_data;
+	char *sharename;
+	uint16_t len;
+
+	sharename = extract_sharename(tcon->treeName);
+	if (IS_ERR(sharename)) {
+		cFYI(1, "CIFS: couldn't extract sharename\n");
+		sharename = NULL;
+		return 0;
+	}
+
+	len = strlen(sharename);
+	if (len > maxbuf)
+		return 0;
+
+	memcpy(buffer, sharename, len);
+
+	kfree(sharename);
+
+	return len;
+}
+
+/*
+ * Superblock object for FS-Cache
+ */
+const struct fscache_cookie_def cifs_fscache_super_index_def = {
+	.name = "CIFS.super",
+	.type = FSCACHE_COOKIE_TYPE_INDEX,
+	.get_key = cifs_super_get_key,
+};
+
Index: cifs-2.6/fs/cifs/cifsglob.h
===================================================================
--- cifs-2.6.orig/fs/cifs/cifsglob.h
+++ cifs-2.6/fs/cifs/cifsglob.h
@@ -317,6 +317,9 @@ struct cifsTconInfo {
 	bool local_lease:1; /* check leases (only) on local system not remote */
 	bool broken_posix_open; /* e.g. Samba server versions < 3.3.2, 3.2.9 */
 	bool need_reconnect:1; /* connection reset, tid now invalid */
+#ifdef CONFIG_CIFS_FSCACHE
+	struct fscache_cookie *fscache; /* cookie for share */
+#endif
 	/* BB add field for back pointer to sb struct(s)? */
 };
 
Index: cifs-2.6/fs/cifs/connect.c
===================================================================
--- cifs-2.6.orig/fs/cifs/connect.c
+++ cifs-2.6/fs/cifs/connect.c
@@ -1773,6 +1773,8 @@ cifs_put_tcon(struct cifsTconInfo *tcon)
 	list_del_init(&tcon->tcon_list);
 	write_unlock(&cifs_tcp_ses_lock);
 
+	cifs_fscache_release_super_cookie(tcon);
+
 	xid = GetXid();
 	CIFSSMBTDis(xid, tcon);
 	_FreeXid(xid);
@@ -1843,6 +1845,8 @@ cifs_get_tcon(struct cifsSesInfo *ses, s
 	tcon->nocase = volume_info->nocase;
 	tcon->local_lease = volume_info->local_lease;
 
+	cifs_fscache_get_super_cookie(tcon);
+
 	write_lock(&cifs_tcp_ses_lock);
 	list_add(&tcon->tcon_list, &ses->tcon_list);
 	write_unlock(&cifs_tcp_ses_lock);
Index: cifs-2.6/fs/cifs/fscache.c
===================================================================
--- cifs-2.6.orig/fs/cifs/fscache.c
+++ cifs-2.6/fs/cifs/fscache.c
@@ -45,3 +45,20 @@ void cifs_fscache_release_client_cookie(
 	server->fscache = NULL;
 }
 
+void cifs_fscache_get_super_cookie(struct cifsTconInfo *tcon)
+{
+	tcon->fscache =
+		fscache_acquire_cookie(tcon->ses->server->fscache,
+				&cifs_fscache_super_index_def, tcon);
+	cFYI(1, "CIFS: get superblock cookie (0x%p/0x%p)\n",
+				tcon, tcon->fscache);
+}
+
+void cifs_fscache_release_super_cookie(struct cifsTconInfo *tcon)
+{
+	cFYI(1, "CIFS: releasing superblock cookie (0x%p/0x%p)\n",
+			tcon, tcon->fscache);
+	fscache_relinquish_cookie(tcon->fscache, 0);
+	tcon->fscache = NULL;
+}
+
Index: cifs-2.6/fs/cifs/fscache.h
===================================================================
--- cifs-2.6.orig/fs/cifs/fscache.h
+++ cifs-2.6/fs/cifs/fscache.h
@@ -28,6 +28,7 @@
 
 extern struct fscache_netfs cifs_fscache_netfs;
 extern const struct fscache_cookie_def cifs_fscache_server_index_def;
+extern const struct fscache_cookie_def cifs_fscache_super_index_def;
 
 extern int cifs_fscache_register(void);
 extern void cifs_fscache_unregister(void);
@@ -37,6 +38,8 @@ extern void cifs_fscache_unregister(void
  */
 extern void cifs_fscache_get_client_cookie(struct TCP_Server_Info *);
 extern void cifs_fscache_release_client_cookie(struct TCP_Server_Info *);
+extern void cifs_fscache_get_super_cookie(struct cifsTconInfo *);
+extern void cifs_fscache_release_super_cookie(struct cifsTconInfo *);
 
 #else /* CONFIG_CIFS_FSCACHE */
 static inline int cifs_fscache_register(void) { return 0; }
@@ -46,6 +49,9 @@ static inline void
 cifs_fscache_get_client_cookie(struct TCP_Server_Info *server) {}
 static inline void
 cifs_fscache_get_client_cookie(struct TCP_Server_Info *server); {}
+static inline void cifs_fscache_get_super_cookie(struct cifsTconInfo *tcon) {}
+static inline void
+cifs_fscache_release_super_cookie(struct cifsTconInfo *tcon) {}
 
 #endif /* CONFIG_CIFS_FSCACHE */
 

[RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Suresh Jayaraman]

Define inode-level data storage objects (managed by cifsInodeInfo structs).
Each inode-level object is created in a super-block level object and is itself
a data storage object in to which pages from the inode are stored.

The inode object is keyed by UniqueId. The coherency data being used is
LastWriteTime and the file size.

Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>
---
 fs/cifs/cache.c    |   80 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/cifs/cifsfs.c   |    7 ++++
 fs/cifs/cifsglob.h |    3 +
 fs/cifs/file.c     |    6 +++
 fs/cifs/fscache.c  |   68 +++++++++++++++++++++++++++++++++++++++++++++
 fs/cifs/fscache.h  |   12 +++++++
 fs/cifs/inode.c    |    4 ++
 7 files changed, 180 insertions(+)

Index: cifs-2.6/fs/cifs/cache.c
===================================================================
--- cifs-2.6.orig/fs/cifs/cache.c
+++ cifs-2.6/fs/cifs/cache.c
@@ -138,3 +138,83 @@ const struct fscache_cookie_def cifs_fsc
 	.get_key = cifs_super_get_key,
 };
 
+/*
+ * Auxiliary data attached to CIFS inode within the cache
+ */
+struct cifs_fscache_inode_auxdata {
+	struct timespec	last_write_time;
+	loff_t		size;
+};
+
+static uint16_t cifs_fscache_inode_get_key(const void *cookie_netfs_data,
+					   void *buffer, uint16_t maxbuf)
+{
+	const struct cifsInodeInfo *cifsi = cookie_netfs_data;
+	uint16_t keylen;
+
+	/* use the UniqueId as the key */
+	keylen = sizeof(cifsi->uniqueid);
+	if (keylen > maxbuf)
+		keylen = 0;
+	else
+		memcpy(buffer, &cifsi->uniqueid, keylen);
+
+	return keylen;
+}
+
+static void
+cifs_fscache_inode_get_attr(const void *cookie_netfs_data, uint64_t *size)
+{
+	const struct cifsInodeInfo *cifsi = cookie_netfs_data;
+
+	*size = cifsi->vfs_inode.i_size;
+}
+
+static uint16_t
+cifs_fscache_inode_get_aux(const void *cookie_netfs_data, void *buffer,
+			   uint16_t maxbuf)
+{
+	struct cifs_fscache_inode_auxdata auxdata;
+	const struct cifsInodeInfo *cifsi = cookie_netfs_data;
+
+	memset(&auxdata, 0, sizeof(auxdata));
+	auxdata.size = cifsi->vfs_inode.i_size;
+	auxdata.last_write_time = cifsi->vfs_inode.i_ctime;
+
+	if (maxbuf > sizeof(auxdata))
+		maxbuf = sizeof(auxdata);
+
+	memcpy(buffer, &auxdata, maxbuf);
+
+	return maxbuf;
+}
+
+static enum
+fscache_checkaux cifs_fscache_inode_check_aux(void *cookie_netfs_data,
+					      const void *data,
+					      uint16_t datalen)
+{
+	struct cifs_fscache_inode_auxdata auxdata;
+	struct cifsInodeInfo *cifsi = cookie_netfs_data;
+
+	if (datalen != sizeof(auxdata))
+		return FSCACHE_CHECKAUX_OBSOLETE;
+
+	memset(&auxdata, 0, sizeof(auxdata));
+	auxdata.size = cifsi->vfs_inode.i_size;
+	auxdata.last_write_time = cifsi->vfs_inode.i_ctime;
+
+	if (memcmp(data, &auxdata, datalen) != 0)
+		return FSCACHE_CHECKAUX_OBSOLETE;
+
+	return FSCACHE_CHECKAUX_OKAY;
+}
+
+const struct fscache_cookie_def cifs_fscache_inode_object_def = {
+	.name		= "CIFS.uniqueid",
+	.type		= FSCACHE_COOKIE_TYPE_DATAFILE,
+	.get_key	= cifs_fscache_inode_get_key,
+	.get_attr	= cifs_fscache_inode_get_attr,
+	.get_aux	= cifs_fscache_inode_get_aux,
+	.check_aux	= cifs_fscache_inode_check_aux,
+};
Index: cifs-2.6/fs/cifs/cifsfs.c
===================================================================
--- cifs-2.6.orig/fs/cifs/cifsfs.c
+++ cifs-2.6/fs/cifs/cifsfs.c
@@ -330,6 +330,12 @@ cifs_destroy_inode(struct inode *inode)
 }
 
 static void
+cifs_clear_inode(struct inode *inode)
+{
+	cifs_fscache_release_inode_cookie(inode);
+}
+
+static void
 cifs_show_address(struct seq_file *s, struct TCP_Server_Info *server)
 {
 	seq_printf(s, ",addr=");
@@ -490,6 +496,7 @@ static const struct super_operations cif
 	.alloc_inode = cifs_alloc_inode,
 	.destroy_inode = cifs_destroy_inode,
 	.drop_inode	= cifs_drop_inode,
+	.clear_inode	= cifs_clear_inode,
 /*	.delete_inode	= cifs_delete_inode,  */  /* Do not need above
 	function unless later we add lazy close of inodes or unless the
 	kernel forgets to call us with the same number of releases (closes)
Index: cifs-2.6/fs/cifs/cifsglob.h
===================================================================
--- cifs-2.6.orig/fs/cifs/cifsglob.h
+++ cifs-2.6/fs/cifs/cifsglob.h
@@ -407,6 +407,9 @@ struct cifsInodeInfo {
 	bool invalid_mapping:1;		/* pagecache is invalid */
 	u64  server_eof;		/* current file size on server */
 	u64  uniqueid;			/* server inode number */
+#ifdef CONFIG_CIFS_FSCACHE
+	struct fscache_cookie *fscache;
+#endif
 	struct inode vfs_inode;
 };
 
Index: cifs-2.6/fs/cifs/file.c
===================================================================
--- cifs-2.6.orig/fs/cifs/file.c
+++ cifs-2.6/fs/cifs/file.c
@@ -40,6 +40,7 @@
 #include "cifs_unicode.h"
 #include "cifs_debug.h"
 #include "cifs_fs_sb.h"
+#include "fscache.h"
 
 static inline int cifs_convert_flags(unsigned int flags)
 {
@@ -282,6 +283,9 @@ int cifs_open(struct inode *inode, struc
 				CIFSSMBClose(xid, tcon, netfid);
 				rc = -ENOMEM;
 			}
+
+			cifs_fscache_set_inode_cookie(inode, file);
+
 			goto out;
 		} else if ((rc == -EINVAL) || (rc == -EOPNOTSUPP)) {
 			if (tcon->ses->serverNOS)
@@ -373,6 +377,8 @@ int cifs_open(struct inode *inode, struc
 		goto out;
 	}
 
+	cifs_fscache_set_inode_cookie(inode, file);
+
 	if (oplock & CIFS_CREATE_ACTION) {
 		/* time to set mode which we can not set earlier due to
 		   problems creating new read-only files */
Index: cifs-2.6/fs/cifs/fscache.c
===================================================================
--- cifs-2.6.orig/fs/cifs/fscache.c
+++ cifs-2.6/fs/cifs/fscache.c
@@ -62,3 +62,71 @@ void cifs_fscache_release_super_cookie(s
 	tcon->fscache = NULL;
 }
 
+static void cifs_fscache_enable_inode_cookie(struct inode *inode)
+{
+	struct cifsInodeInfo *cifsi = CIFS_I(inode);
+	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
+
+	if (cifsi->fscache)
+		return;
+
+	cifsi->fscache = fscache_acquire_cookie(cifs_sb->tcon->fscache,
+				&cifs_fscache_inode_object_def,
+				cifsi);
+	cFYI(1, "CIFS: got FH cookie (0x%p/0x%p/0x%p)\n",
+			cifs_sb->tcon, cifsi, cifsi->fscache);
+}
+
+void cifs_fscache_release_inode_cookie(struct inode *inode)
+{
+	struct cifsInodeInfo *cifsi = CIFS_I(inode);
+
+	if (cifsi->fscache) {
+		cFYI(1, "CIFS releasing inode cookie (0x%p/0x%p)\n",
+				cifsi, cifsi->fscache);
+		fscache_relinquish_cookie(cifsi->fscache, 0);
+		cifsi->fscache = NULL;
+	}
+}
+
+static void cifs_fscache_disable_inode_cookie(struct inode *inode)
+{
+	struct cifsInodeInfo *cifsi = CIFS_I(inode);
+
+	if (cifsi->fscache) {
+		cFYI(1, "CIFS disabling inode cookie (0x%p/0x%p)\n",
+				cifsi, cifsi->fscache);
+		fscache_relinquish_cookie(cifsi->fscache, 1);
+		cifsi->fscache = NULL;
+	}
+}
+
+void cifs_fscache_set_inode_cookie(struct inode *inode, struct file *filp)
+{
+	/* BB: parallel opens - need locking? */
+	if ((filp->f_flags & O_ACCMODE) != O_RDONLY)
+		cifs_fscache_disable_inode_cookie(inode);
+	else {
+		cifs_fscache_enable_inode_cookie(inode);
+		cFYI(1, "CIFS: fscache inode cookie set\n");
+	}
+}
+
+void cifs_fscache_reset_inode_cookie(struct inode *inode)
+{
+	struct cifsInodeInfo *cifsi = CIFS_I(inode);
+	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
+	struct fscache_cookie *old = cifsi->fscache;
+
+	if (cifsi->fscache) {
+		/* retire the current fscache cache and get a new one */
+		fscache_relinquish_cookie(cifsi->fscache, 1);
+
+		cifsi->fscache = fscache_acquire_cookie(cifs_sb->tcon->fscache,
+					&cifs_fscache_inode_object_def,
+					cifsi);
+		cFYI(1, "CIFS: new cookie (0x%p/0x%p) oldcookie 0x%p\n",
+				cifsi, cifsi->fscache, old);
+	}
+}
+
Index: cifs-2.6/fs/cifs/fscache.h
===================================================================
--- cifs-2.6.orig/fs/cifs/fscache.h
+++ cifs-2.6/fs/cifs/fscache.h
@@ -29,6 +29,8 @@
 extern struct fscache_netfs cifs_fscache_netfs;
 extern const struct fscache_cookie_def cifs_fscache_server_index_def;
 extern const struct fscache_cookie_def cifs_fscache_super_index_def;
+extern const struct fscache_cookie_def cifs_fscache_inode_object_def;
+
 
 extern int cifs_fscache_register(void);
 extern void cifs_fscache_unregister(void);
@@ -41,6 +43,10 @@ extern void cifs_fscache_release_client_
 extern void cifs_fscache_get_super_cookie(struct cifsTconInfo *);
 extern void cifs_fscache_release_super_cookie(struct cifsTconInfo *);
 
+extern void cifs_fscache_release_inode_cookie(struct inode *);
+extern void cifs_fscache_set_inode_cookie(struct inode *, struct file *);
+extern void cifs_fscache_reset_inode_cookie(struct inode *);
+
 #else /* CONFIG_CIFS_FSCACHE */
 static inline int cifs_fscache_register(void) { return 0; }
 static inline void cifs_fscache_unregister(void) {}
@@ -53,6 +59,12 @@ static inline void cifs_fscache_get_supe
 static inline void
 cifs_fscache_release_super_cookie(struct cifsTconInfo *tcon) {}
 
+static inline void cifs_fscache_release_inode_cookie(struct inode *inode) {}
+static inline void cifs_fscache_set_inode_cookie(struct inode *inode,
+			struct file *filp) {}
+static inline void cifs_fscache_reset_inode_cookie(struct inode *inode) {}
+
+
 #endif /* CONFIG_CIFS_FSCACHE */
 
 #endif /* _CIFS_FSCACHE_H */
Index: cifs-2.6/fs/cifs/inode.c
===================================================================
--- cifs-2.6.orig/fs/cifs/inode.c
+++ cifs-2.6/fs/cifs/inode.c
@@ -29,6 +29,7 @@
 #include "cifsproto.h"
 #include "cifs_debug.h"
 #include "cifs_fs_sb.h"
+#include "fscache.h"
 
 
 static void cifs_set_ops(struct inode *inode, const bool is_dfs_referral)
@@ -776,6 +777,8 @@ retry_iget5_locked:
 			inode->i_flags |= S_NOATIME | S_NOCMTIME;
 		if (inode->i_state & I_NEW) {
 			inode->i_ino = hash;
+			/* initialize per-inode cache cookie pointer */
+			CIFS_I(inode)->fscache = NULL;
 			unlock_new_inode(inode);
 		}
 	}
@@ -1568,6 +1571,7 @@ cifs_invalidate_mapping(struct inode *in
 			cifs_i->write_behind_rc = rc;
 	}
 	invalidate_remote_inode(inode);
+	cifs_fscache_reset_inode_cookie(inode);
 }
 
 int cifs_revalidate_file(struct file *filp)

[RFC][PATCH 07/10] cifs: FS-Cache page management

[Suresh Jayaraman]

Takes care of invalidation and release of FS-Cache marked pages and also
invalidation of the FsCache page flag when the inode is removed.

Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>
---
 fs/cifs/cache.c   |   31 +++++++++++++++++++++++++++++++
 fs/cifs/file.c    |   20 ++++++++++++++++++++
 fs/cifs/fscache.c |   26 ++++++++++++++++++++++++++
 fs/cifs/fscache.h |   16 ++++++++++++++++
 4 files changed, 93 insertions(+), 0 deletions(-)

diff --git a/fs/cifs/cache.c b/fs/cifs/cache.c
index b205424..3a733c1 100644
--- a/fs/cifs/cache.c
+++ b/fs/cifs/cache.c
@@ -210,6 +210,36 @@ fscache_checkaux cifs_fscache_inode_check_aux(void *cookie_netfs_data,
 	return FSCACHE_CHECKAUX_OKAY;
 }
 
+static void cifs_fscache_inode_now_uncached(void *cookie_netfs_data)
+{
+	struct cifsInodeInfo *cifsi = cookie_netfs_data;
+	struct pagevec pvec;
+	pgoff_t first;
+	int loop, nr_pages;
+
+	pagevec_init(&pvec, 0);
+	first = 0;
+
+	cFYI(1, "cifs inode 0x%p now uncached\n", cifsi);
+
+	for (;;) {
+		nr_pages = pagevec_lookup(&pvec,
+					  cifsi->vfs_inode.i_mapping, first,
+					  PAGEVEC_SIZE - pagevec_count(&pvec));
+		if (!nr_pages)
+			break;
+
+		for (loop = 0; loop < nr_pages; loop++)
+			ClearPageFsCache(pvec.pages[loop]);
+
+		first = pvec.pages[nr_pages - 1]->index + 1;
+
+		pvec.nr = nr_pages;
+		pagevec_release(&pvec);
+		cond_resched();
+	}
+}
+
 const struct fscache_cookie_def cifs_fscache_inode_object_def = {
 	.name		= "CIFS.uniqueid",
 	.type		= FSCACHE_COOKIE_TYPE_DATAFILE,
@@ -217,4 +247,5 @@ const struct fscache_cookie_def cifs_fscache_inode_object_def = {
 	.get_attr	= cifs_fscache_inode_get_attr,
 	.get_aux	= cifs_fscache_inode_get_aux,
 	.check_aux	= cifs_fscache_inode_check_aux,
+	.now_uncached	= cifs_fscache_inode_now_uncached,
 };
diff --git a/fs/cifs/file.c b/fs/cifs/file.c
index 55ecb55..786ec04 100644
--- a/fs/cifs/file.c
+++ b/fs/cifs/file.c
@@ -2271,6 +2271,22 @@ out:
 	return rc;
 }
 
+static int cifs_release_page(struct page *page, gfp_t gfp)
+{
+	if (PagePrivate(page))
+		return 0;
+
+	return cifs_fscache_release_page(page, gfp);
+}
+
+static void cifs_invalidate_page(struct page *page, unsigned long offset)
+{
+	struct cifsInodeInfo *cifsi = CIFS_I(page->mapping->host);
+
+	if (offset == 0)
+		cifs_fscache_invalidate_page(page, &cifsi->vfs_inode);
+}
+
 static void
 cifs_oplock_break(struct slow_work *work)
 {
@@ -2344,6 +2360,8 @@ const struct address_space_operations cifs_addr_ops = {
 	.write_begin = cifs_write_begin,
 	.write_end = cifs_write_end,
 	.set_page_dirty = __set_page_dirty_nobuffers,
+	.releasepage = cifs_release_page,
+	.invalidatepage = cifs_invalidate_page,
 	/* .sync_page = cifs_sync_page, */
 	/* .direct_IO = */
 };
@@ -2360,6 +2378,8 @@ const struct address_space_operations cifs_addr_ops_smallbuf = {
 	.write_begin = cifs_write_begin,
 	.write_end = cifs_write_end,
 	.set_page_dirty = __set_page_dirty_nobuffers,
+	.releasepage = cifs_release_page,
+	.invalidatepage = cifs_invalidate_page,
 	/* .sync_page = cifs_sync_page, */
 	/* .direct_IO = */
 };
diff --git a/fs/cifs/fscache.c b/fs/cifs/fscache.c
index ddfd355..c09d3b8 100644
--- a/fs/cifs/fscache.c
+++ b/fs/cifs/fscache.c
@@ -130,3 +130,29 @@ void cifs_fscache_reset_inode_cookie(struct inode *inode)
 	}
 }
 
+int cifs_fscache_release_page(struct page *page, gfp_t gfp)
+{
+	if (PageFsCache(page)) {
+		struct inode *inode = page->mapping->host;
+		struct cifsInodeInfo *cifsi = CIFS_I(inode);
+
+		cFYI(1, "CIFS: fscache release page (0x%p/0x%p)\n",
+				cifsi->fscache, page);
+		if (!fscache_maybe_release_page(cifsi->fscache, page, gfp))
+			return 0;
+	}
+
+	return 1;
+}
+
+void __cifs_fscache_invalidate_page(struct page *page, struct inode *inode)
+{
+	struct cifsInodeInfo *cifsi = CIFS_I(inode);
+	struct fscache_cookie *cookie = cifsi->fscache;
+
+	cFYI(1, "CIFS: fscache invalidatepage (0x%p/0x%p/0x%p)\n",
+			cookie, page, cifsi);
+	fscache_wait_on_page_write(cookie, page);
+	fscache_uncache_page(cookie, page);
+}
+
diff --git a/fs/cifs/fscache.h b/fs/cifs/fscache.h
index 836bb02..127cb0a 100644
--- a/fs/cifs/fscache.h
+++ b/fs/cifs/fscache.h
@@ -47,6 +47,16 @@ extern void cifs_fscache_release_inode_cookie(struct inode *);
 extern void cifs_fscache_set_inode_cookie(struct inode *, struct file *);
 extern void cifs_fscache_reset_inode_cookie(struct inode *);
 
+extern void __cifs_fscache_invalidate_page(struct page *, struct inode *);
+extern int cifs_fscache_release_page(struct page *page, gfp_t gfp);
+
+static inline void cifs_fscache_invalidate_page(struct page *page,
+					       struct inode *inode)
+{
+	if (PageFsCache(page))
+		__cifs_fscache_invalidate_page(page, inode);
+}
+
 #else /* CONFIG_CIFS_FSCACHE */
 static inline int cifs_fscache_register(void) { return 0; }
 static inline void cifs_fscache_unregister(void) {}
@@ -63,7 +73,13 @@ static inline void cifs_fscache_release_inode_cookie(struct inode *inode) {}
 static inline void cifs_fscache_set_inode_cookie(struct inode *inode,
 			struct file *filp) {}
 static inline void cifs_fscache_reset_inode_cookie(struct inode *inode) {}
+static inline void cifs_fscache_release_page(struct page *page, gfp_t gfp)
+{
+	return 1; /* May release page */
+}
 
+static inline int cifs_fscache_invalidate_page(struct page *page,
+			struct inode *) {}
 
 #endif /* CONFIG_CIFS_FSCACHE */
 
-- 
1.6.4.2

[RFC][PATCH 08/10] cifs: store pages into local cache

[Suresh Jayaraman]

Store pages from an CIFS inode into the data storage object associated with
that inode.

Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>
---
 fs/cifs/file.c    |    6 ++++++
 fs/cifs/fscache.c |   13 +++++++++++++
 fs/cifs/fscache.h |   11 +++++++++++
 3 files changed, 30 insertions(+), 0 deletions(-)

diff --git a/fs/cifs/file.c b/fs/cifs/file.c
index 786ec04..39c1ce0 100644
--- a/fs/cifs/file.c
+++ b/fs/cifs/file.c
@@ -2060,6 +2060,8 @@ static int cifs_readpages(struct file *file, struct address_space *mapping,
 				   we will hit it on next read */
 
 				/* break; */
+				/* send this page to FS-Cache */
+				cifs_readpage_to_fscache(mapping->host, page);
 			}
 		} else {
 			cFYI(1, "No bytes read (%d) at offset %lld . "
@@ -2117,6 +2119,10 @@ static int cifs_readpage_worker(struct file *file, struct page *page,
 
 	flush_dcache_page(page);
 	SetPageUptodate(page);
+
+	/* send this page to the cache */
+	cifs_readpage_to_fscache(file->f_path.dentry->d_inode, page);
+
 	rc = 0;
 
 io_error:
diff --git a/fs/cifs/fscache.c b/fs/cifs/fscache.c
index c09d3b8..13e47d5 100644
--- a/fs/cifs/fscache.c
+++ b/fs/cifs/fscache.c
@@ -145,6 +145,19 @@ int cifs_fscache_release_page(struct page *page, gfp_t gfp)
 	return 1;
 }
 
+void __cifs_readpage_to_fscache(struct inode *inode, struct page *page)
+{
+	int ret;
+
+	cFYI(1, "CIFS: readpage_to_fscache(fsc: %p, p: %p, i: %p\n",
+			CIFS_I(inode)->fscache, page, inode);
+	ret = fscache_write_page(CIFS_I(inode)->fscache, page, GFP_KERNEL);
+	cFYI(1, "CIFS: fscache_write_page returned %d\n", ret);
+
+	if (ret != 0)
+		fscache_uncache_page(CIFS_I(inode)->fscache, page);
+}
+
 void __cifs_fscache_invalidate_page(struct page *page, struct inode *inode)
 {
 	struct cifsInodeInfo *cifsi = CIFS_I(inode);
diff --git a/fs/cifs/fscache.h b/fs/cifs/fscache.h
index 127cb0a..e34d8ab 100644
--- a/fs/cifs/fscache.h
+++ b/fs/cifs/fscache.h
@@ -50,6 +50,8 @@ extern void cifs_fscache_reset_inode_cookie(struct inode *);
 extern void __cifs_fscache_invalidate_page(struct page *, struct inode *);
 extern int cifs_fscache_release_page(struct page *page, gfp_t gfp);
 
+extern void __cifs_readpage_to_fscache(struct inode *, struct page *);
+
 static inline void cifs_fscache_invalidate_page(struct page *page,
 					       struct inode *inode)
 {
@@ -57,6 +59,13 @@ static inline void cifs_fscache_invalidate_page(struct page *page,
 		__cifs_fscache_invalidate_page(page, inode);
 }
 
+static inline void cifs_readpage_to_fscache(struct inode *inode,
+					    struct page *page)
+{
+	if (PageFsCache(page))
+		__cifs_readpage_to_fscache(inode, page);
+}
+
 #else /* CONFIG_CIFS_FSCACHE */
 static inline int cifs_fscache_register(void) { return 0; }
 static inline void cifs_fscache_unregister(void) {}
@@ -80,6 +89,8 @@ static inline void cifs_fscache_release_page(struct page *page, gfp_t gfp)
 
 static inline int cifs_fscache_invalidate_page(struct page *page,
 			struct inode *) {}
+static inline void cifs_readpage_to_fscache(struct inode *inode,
+			struct page *page) {}
 
 #endif /* CONFIG_CIFS_FSCACHE */
 
-- 
1.6.4.2

[RFC][PATCH 09/10] cifs: read pages from FS-Cache

[Suresh Jayaraman]

Read pages from a FS-Cache data storage object into a CIFS inode.

Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
---
 fs/cifs/file.c    |   19 ++++++++++++++
 fs/cifs/fscache.c |   73 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/cifs/fscache.h |   40 ++++++++++++++++++++++++++++-
 3 files changed, 131 insertions(+), 1 deletions(-)

diff --git a/fs/cifs/file.c b/fs/cifs/file.c
index 39c1ce0..42d2f25 100644
--- a/fs/cifs/file.c
+++ b/fs/cifs/file.c
@@ -1978,6 +1978,16 @@ static int cifs_readpages(struct file *file, struct address_space *mapping,
 	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
 	pTcon = cifs_sb->tcon;
 
+	/*
+	 * Reads as many pages as possible from fscache. Returns -ENOBUFS
+	 * immediately if the cookie is negative
+	 */
+	rc = cifs_readpages_from_fscache(mapping->host, mapping, page_list,
+					 &num_pages);
+	cFYI(1, "CIFS: readpages_from_fscache returned %d\n", rc);
+	if (rc == 0)
+		goto read_complete;
+
 	cFYI(DBG2, "rpages: num pages %d", num_pages);
 	for (i = 0; i < num_pages; ) {
 		unsigned contig_pages;
@@ -2090,6 +2100,7 @@ static int cifs_readpages(struct file *file, struct address_space *mapping,
 		smb_read_data = NULL;
 	}
 
+read_complete:
 	FreeXid(xid);
 	return rc;
 }
@@ -2100,6 +2111,12 @@ static int cifs_readpage_worker(struct file *file, struct page *page,
 	char *read_data;
 	int rc;
 
+	/* Is the page cached? */
+	rc = cifs_readpage_from_fscache(file->f_path.dentry->d_inode, page);
+	cFYI(1, "CIFS: cifs_readpage_from_fscache returned %d\n", rc);
+	if (rc == 0)
+		goto read_complete;
+
 	page_cache_get(page);
 	read_data = kmap(page);
 	/* for reads over a certain size could initiate async read ahead */
@@ -2128,6 +2145,8 @@ static int cifs_readpage_worker(struct file *file, struct page *page,
 io_error:
 	kunmap(page);
 	page_cache_release(page);
+
+read_complete:
 	return rc;
 }
 
diff --git a/fs/cifs/fscache.c b/fs/cifs/fscache.c
index 13e47d5..6813737 100644
--- a/fs/cifs/fscache.c
+++ b/fs/cifs/fscache.c
@@ -145,6 +145,79 @@ int cifs_fscache_release_page(struct page *page, gfp_t gfp)
 	return 1;
 }
 
+static void cifs_readpage_from_fscache_complete(struct page *page, void *ctx,
+						int error)
+{
+	cFYI(1, "CFS: readpage_from_fscache_complete (0x%p/%d)\n",
+			page, error);
+	if (!error)
+		SetPageUptodate(page);
+	unlock_page(page);
+}
+
+/*
+ * Retrieve a page from FS-Cache
+ */
+int __cifs_readpage_from_fscache(struct inode *inode, struct page *page)
+{
+	int ret;
+
+	cFYI(1, "CIFS: readpage_from_fscache(fsc:%p, p:%p, i:0x%p\n",
+			CIFS_I(inode)->fscache, page, inode);
+	ret = fscache_read_or_alloc_page(CIFS_I(inode)->fscache, page,
+					 cifs_readpage_from_fscache_complete,
+					 NULL,
+					 GFP_KERNEL);
+	switch (ret) {
+
+	case 0: /* page found in fscache, read submitted */
+		cFYI(1, "CIFS: readpage_from_fscache: submitted\n");
+		return ret;
+	case -ENOBUFS:	/* page won't be cached */
+	case -ENODATA:	/* page not in cache */
+		cFYI(1, "CIFS: readpage_from_fscache %d\n", ret);
+		return 1;
+
+	default:
+		cFYI(1, "unknown error ret = %d", ret);
+	}
+	return ret;
+}
+
+/*
+ * Retrieve a set of pages from FS-Cache
+ */
+int __cifs_readpages_from_fscache(struct inode *inode,
+				struct address_space *mapping,
+				struct list_head *pages,
+				unsigned *nr_pages)
+{
+	int ret;
+
+	cFYI(1, "CIFS: __cifs_readpages_from_fscache (0x%p/%u/0x%p)\n",
+			CIFS_I(inode)->fscache, *nr_pages, inode);
+	ret = fscache_read_or_alloc_pages(CIFS_I(inode)->fscache, mapping,
+					  pages, nr_pages,
+					  cifs_readpage_from_fscache_complete,
+					  NULL,
+					  mapping_gfp_mask(mapping));
+	switch (ret) {
+	case 0:	/* read submitted to the cache for all pages */
+		cFYI(1, "CIFS: readpages_from_fscache\n");
+		return ret;
+
+	case -ENOBUFS:	/* some pages are not cached and can't be */
+	case -ENODATA:	/* some pages are not cached */
+		cFYI(1, "CIFS: readpages_from_fscache: no page\n");
+		return 1;
+
+	default:
+		cFYI(1, "unknown error ret = %d", ret);
+	}
+
+	return ret;
+}
+
 void __cifs_readpage_to_fscache(struct inode *inode, struct page *page)
 {
 	int ret;
diff --git a/fs/cifs/fscache.h b/fs/cifs/fscache.h
index e34d8ab..03bd3fe 100644
--- a/fs/cifs/fscache.h
+++ b/fs/cifs/fscache.h
@@ -31,7 +31,6 @@ extern const struct fscache_cookie_def cifs_fscache_server_index_def;
 extern const struct fscache_cookie_def cifs_fscache_super_index_def;
 extern const struct fscache_cookie_def cifs_fscache_inode_object_def;
 
-
 extern int cifs_fscache_register(void);
 extern void cifs_fscache_unregister(void);
 
@@ -49,6 +48,11 @@ extern void cifs_fscache_reset_inode_cookie(struct inode *);
 
 extern void __cifs_fscache_invalidate_page(struct page *, struct inode *);
 extern int cifs_fscache_release_page(struct page *page, gfp_t gfp);
+extern int __cifs_readpage_from_fscache(struct inode *, struct page *);
+extern int __cifs_readpages_from_fscache(struct inode *,
+					 struct address_space *,
+					 struct list_head *,
+					 unsigned *);
 
 extern void __cifs_readpage_to_fscache(struct inode *, struct page *);
 
@@ -59,6 +63,26 @@ static inline void cifs_fscache_invalidate_page(struct page *page,
 		__cifs_fscache_invalidate_page(page, inode);
 }
 
+static inline int cifs_readpage_from_fscache(struct inode *inode,
+					     struct page *page)
+{
+	if (CIFS_I(inode)->fscache)
+		return __cifs_readpage_from_fscache(inode, page);
+
+	return -ENOBUFS;
+}
+
+static inline int cifs_readpages_from_fscache(struct inode *inode,
+					      struct address_space *mapping,
+					      struct list_head *pages,
+					      unsigned *nr_pages)
+{
+	if (CIFS_I(inode)->fscache)
+		return __cifs_readpages_from_fscache(inode, mapping, pages,
+						     nr_pages);
+	return -ENOBUFS;
+}
+
 static inline void cifs_readpage_to_fscache(struct inode *inode,
 					    struct page *page)
 {
@@ -89,6 +113,20 @@ static inline void cifs_fscache_release_page(struct page *page, gfp_t gfp)
 
 static inline int cifs_fscache_invalidate_page(struct page *page,
 			struct inode *) {}
+static inline int
+cifs_readpage_from_fscache(struct inode *inode, struct page *page)
+{
+	return -ENOBUFS;
+}
+
+static inline int cifs_readpages_from_fscache(struct inode *inode,
+					      struct address_space *mapping,
+					      struct list_head *pages,
+					      unsigned *nr_pages)
+{
+	return -ENOBUFS;
+}
+
 static inline void cifs_readpage_to_fscache(struct inode *inode,
 			struct page *page) {}
 
-- 
1.6.4.2

[RFC][PATCH 10/10] cifs: add mount option to enable local caching

[Suresh Jayaraman]

Add a mount option 'fsc' to enable local caching on CIFS.

As the cifs-utils (userspace) changes are not done yet, this patch enables
'fsc' by default to assist testing.

Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
---
 fs/cifs/cifs_fs_sb.h |    1 +
 fs/cifs/connect.c    |    8 ++++++++
 2 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/fs/cifs/cifs_fs_sb.h b/fs/cifs/cifs_fs_sb.h
index 246a167..9e77145 100644
--- a/fs/cifs/cifs_fs_sb.h
+++ b/fs/cifs/cifs_fs_sb.h
@@ -35,6 +35,7 @@
 #define CIFS_MOUNT_DYNPERM      0x1000 /* allow in-memory only mode setting   */
 #define CIFS_MOUNT_NOPOSIXBRL   0x2000 /* mandatory not posix byte range lock */
 #define CIFS_MOUNT_NOSSYNC      0x4000 /* don't do slow SMBflush on every sync*/
+#define CIFS_MOUNT_FSCACHE	0x8000 /* local caching enabled */
 
 struct cifs_sb_info {
 	struct cifsTconInfo *tcon;	/* primary mount */
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 4844dbd..6c6ff3c 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -98,6 +98,7 @@ struct smb_vol {
 	bool noblocksnd:1;
 	bool noautotune:1;
 	bool nostrictsync:1; /* do not force expensive SMBflush on every sync */
+	bool fsc:1;	/* enable fscache */
 	unsigned int rsize;
 	unsigned int wsize;
 	bool sockopt_tcp_nodelay:1;
@@ -843,6 +844,9 @@ cifs_parse_mount_options(char *options, const char *devname,
 	/* default to using server inode numbers where available */
 	vol->server_ino = 1;
 
+	/* XXX: default to fsc for testing until mount.cifs pieces are done */
+	vol->fsc = 1;
+
 	if (!options)
 		return 1;
 
@@ -1332,6 +1336,8 @@ cifs_parse_mount_options(char *options, const char *devname,
 			printk(KERN_WARNING "CIFS: Mount option noac not "
 				"supported. Instead set "
 				"/proc/fs/cifs/LookupCacheEnabled to 0\n");
+		} else if (strnicmp(data, "fsc", 3) == 0) {
+			vol->fsc = true;
 		} else
 			printk(KERN_WARNING "CIFS: Unknown mount option %s\n",
 						data);
@@ -2405,6 +2411,8 @@ static void setup_cifs_sb(struct smb_vol *pvolume_info,
 		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_OVERR_GID;
 	if (pvolume_info->dynperm)
 		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_DYNPERM;
+	if (pvolume_info->fsc)
+		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_FSCACHE;
 	if (pvolume_info->direct_io) {
 		cFYI(1, "mounting share using direct i/o");
 		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_DIRECT_IO;
-- 
1.6.4.2

Re: [RFC][PATCH 02/10] cifs: guard cifsglob.h against multiple inclusion

[Jeff Layton]

On Tue, 22 Jun 2010 20:52:50 +0530
Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> Add conditional compile macros to guard the header file against multiple
> inclusion.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
> ---
>  fs/cifs/cifsglob.h |    5 +++++
>  1 files changed, 5 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
> index a88479c..6b2c39d 100644
> --- a/fs/cifs/cifsglob.h
> +++ b/fs/cifs/cifsglob.h
> @@ -16,6 +16,9 @@
>   *   the GNU Lesser General Public License for more details.
>   *
>   */
> +#ifndef _CIFS_GLOB_H
> +#define _CIFS_GLOB_H
> +
>  #include <linux/in.h>
>  #include <linux/in6.h>
>  #include <linux/slab.h>
> @@ -733,3 +736,5 @@ GLOBAL_EXTERN unsigned int cifs_min_small;  /* min size of small buf pool */
>  GLOBAL_EXTERN unsigned int cifs_max_pending; /* MAX requests at once to server*/
>  
>  extern const struct slow_work_ops cifs_oplock_break_ops;
> +
> +#endif	/* _CIFS_GLOB_H */

Strong ACK

Acked-by: Jeff Layton <jlayton-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org>

Re: [RFC][PATCH 04/10] cifs: define server-level cache index objects and register them with FS-Cache

[Jeff Layton]

On Tue, 22 Jun 2010 20:53:18 +0530
Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> Define server-level cache index objects (as managed by TCP_ServerInfo structs).
> Each server object is created in the CIFS top-level index object and is itself
> an index into which superblock-level objects are inserted.
> 
> Currently, the server objects are keyed by hostname.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
> ---
>  fs/cifs/Makefile   |    2 +-
>  fs/cifs/cache.c    |   25 +++++++++++++++++++++++++
>  fs/cifs/cifsglob.h |    3 +++
>  fs/cifs/connect.c  |    4 ++++
>  fs/cifs/fscache.c  |   47 +++++++++++++++++++++++++++++++++++++++++++++++
>  fs/cifs/fscache.h  |   12 ++++++++++++
>  6 files changed, 92 insertions(+), 1 deletion(-)
>  create mode 100644 fs/cifs/fscache.c
> 
> Index: cifs-2.6/fs/cifs/Makefile
> ===================================================================
> --- cifs-2.6.orig/fs/cifs/Makefile
> +++ cifs-2.6/fs/cifs/Makefile
> @@ -12,4 +12,4 @@ cifs-$(CONFIG_CIFS_UPCALL) += cifs_spneg
>  
>  cifs-$(CONFIG_CIFS_DFS_UPCALL) += dns_resolve.o cifs_dfs_ref.o
>  
> -cifs-$(CONFIG_CIFS_FSCACHE) += cache.o
> +cifs-$(CONFIG_CIFS_FSCACHE) += fscache.o cache.o
> Index: cifs-2.6/fs/cifs/cache.c
> ===================================================================
> --- cifs-2.6.orig/fs/cifs/cache.c
> +++ cifs-2.6/fs/cifs/cache.c
> @@ -51,3 +51,28 @@ void cifs_fscache_unregister(void)
>  	fscache_unregister_netfs(&cifs_fscache_netfs);
>  }
>  
> +/*
> + * Server object currently keyed by hostname
> + */
> +static uint16_t cifs_server_get_key(const void *cookie_netfs_data,
> +				   void *buffer, uint16_t maxbuf)
> +{
> +	const struct TCP_Server_Info *server = cookie_netfs_data;
> +	uint16_t len = strnlen(server->hostname, sizeof(server->hostname));
> +

Would a tuple of address/family/port be a better choice here? Imagine I
mount "foo" and then later mount "foor.bar.baz". If they are the same
address and only the UNC differs, then you won't get the benefit of
the cache, right?

> +	if (len > maxbuf)
> +		return 0;
> +
> +	memcpy(buffer, server->hostname, len);
> +
> +	return len;
> +}
> +
> +/*
> + * Server object for FS-Cache
> + */
> +const struct fscache_cookie_def cifs_fscache_server_index_def = {
> +	.name = "CIFS.server",
> +	.type = FSCACHE_COOKIE_TYPE_INDEX,
> +	.get_key = cifs_server_get_key,
> +};
> Index: cifs-2.6/fs/cifs/cifsglob.h
> ===================================================================
> --- cifs-2.6.orig/fs/cifs/cifsglob.h
> +++ cifs-2.6/fs/cifs/cifsglob.h
> @@ -193,6 +193,9 @@ struct TCP_Server_Info {
>  	bool	sec_mskerberos;		/* supports legacy MS Kerberos */
>  	bool	sec_kerberosu2u;	/* supports U2U Kerberos */
>  	bool	sec_ntlmssp;		/* supports NTLMSSP */
> +#ifdef CONFIG_CIFS_FSCACHE
> +	struct fscache_cookie   *fscache; /* client index cache cookie */
> +#endif
>  };
>  
>  /*
> Index: cifs-2.6/fs/cifs/connect.c
> ===================================================================
> --- cifs-2.6.orig/fs/cifs/connect.c
> +++ cifs-2.6/fs/cifs/connect.c
> @@ -48,6 +48,7 @@
>  #include "nterr.h"
>  #include "rfc1002pdu.h"
>  #include "cn_cifs.h"
> +#include "fscache.h"
>  
>  #define CIFS_PORT 445
>  #define RFC1001_PORT 139
> @@ -1453,6 +1454,8 @@ cifs_put_tcp_session(struct TCP_Server_I
>  		return;
>  	}
>  
> +	cifs_fscache_release_client_cookie(server);
> +
>  	list_del_init(&server->tcp_ses_list);
>  	write_unlock(&cifs_tcp_ses_lock);
>  
> @@ -1572,6 +1575,7 @@ cifs_get_tcp_session(struct smb_vol *vol
>  		goto out_err;
>  	}
>  
> +	cifs_fscache_get_client_cookie(tcp_ses);
>  	/* thread spawned, put it on the list */
>  	write_lock(&cifs_tcp_ses_lock);
>  	list_add(&tcp_ses->tcp_ses_list, &cifs_tcp_ses_list);
> Index: cifs-2.6/fs/cifs/fscache.c
> ===================================================================
> --- /dev/null
> +++ cifs-2.6/fs/cifs/fscache.c
> @@ -0,0 +1,47 @@
> +/*
> + *   fs/cifs/fscache.c - CIFS filesystem cache interface
> + *
> + *   Copyright (c) 2010 Novell, Inc.
> + *   Authors(s): Suresh Jayaraman (sjayaraman-l3A5Bk7waGM@public.gmane.org>
> + *
> + *   This library is free software; you can redistribute it and/or modify
> + *   it under the terms of the GNU Lesser General Public License as published
> + *   by the Free Software Foundation; either version 2.1 of the License, or
> + *   (at your option) any later version.
> + *
> + *   This library is distributed in the hope that it will be useful,
> + *   but WITHOUT ANY WARRANTY; without even the implied warranty of
> + *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
> + *   the GNU Lesser General Public License for more details.
> + *
> + *   You should have received a copy of the GNU Lesser General Public License
> + *   along with this library; if not, write to the Free Software
> + *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
> + */
> +#include <linux/init.h>
> +#include <linux/kernel.h>
> +#include <linux/sched.h>
> +#include <linux/mm.h>
> +#include <linux/in6.h>
> +
> +#include "fscache.h"
> +#include "cifsglob.h"
> +#include "cifs_debug.h"
> +
> +void cifs_fscache_get_client_cookie(struct TCP_Server_Info *server)
> +{
> +	server->fscache =
> +		fscache_acquire_cookie(cifs_fscache_netfs.primary_index,
> +				&cifs_fscache_server_index_def, server);
> +	cFYI(1, "CIFS: get client cookie (0x%p/0x%p)\n",
> +				server, server->fscache);
> +}
> +
> +void cifs_fscache_release_client_cookie(struct TCP_Server_Info *server)
> +{
> +	cFYI(1, "CIFS: release client cookie (0x%p/0x%p)\n",
> +				server, server->fscache);
> +	fscache_relinquish_cookie(server->fscache, 0);
> +	server->fscache = NULL;
> +}
> +
> Index: cifs-2.6/fs/cifs/fscache.h
> ===================================================================
> --- cifs-2.6.orig/fs/cifs/fscache.h
> +++ cifs-2.6/fs/cifs/fscache.h
> @@ -27,14 +27,26 @@
>  #ifdef CONFIG_CIFS_FSCACHE
>  
>  extern struct fscache_netfs cifs_fscache_netfs;
> +extern const struct fscache_cookie_def cifs_fscache_server_index_def;
>  
>  extern int cifs_fscache_register(void);
>  extern void cifs_fscache_unregister(void);
>  
> +/*
> + * fscache.c
> + */
> +extern void cifs_fscache_get_client_cookie(struct TCP_Server_Info *);
> +extern void cifs_fscache_release_client_cookie(struct TCP_Server_Info *);
> +
>  #else /* CONFIG_CIFS_FSCACHE */
>  static inline int cifs_fscache_register(void) { return 0; }
>  static inline void cifs_fscache_unregister(void) {}
>  
> +static inline void
> +cifs_fscache_get_client_cookie(struct TCP_Server_Info *server) {}
> +static inline void
> +cifs_fscache_get_client_cookie(struct TCP_Server_Info *server); {}
> +
>  #endif /* CONFIG_CIFS_FSCACHE */
>  
>  #endif /* _CIFS_FSCACHE_H */
> --
> To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 


-- 
Jeff Layton <jlayton-vpEMnDpepFuMZCB2o+C8xQ@public.gmane.org>

Re: [RFC][PATCH 04/10] cifs: define server-level cache index objects and register them with FS-Cache

[Suresh Jayaraman]

On 06/23/2010 03:22 AM, Jeff Layton wrote:
> On Tue, 22 Jun 2010 20:53:18 +0530
> Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:
> 
>> Define server-level cache index objects (as managed by TCP_ServerInfo structs).
>> Each server object is created in the CIFS top-level index object and is itself
>> an index into which superblock-level objects are inserted.
>>
>> Currently, the server objects are keyed by hostname.
>>
>> Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>
>> ---
>>  fs/cifs/Makefile   |    2 +-
>>  fs/cifs/cache.c    |   25 +++++++++++++++++++++++++
>>  fs/cifs/cifsglob.h |    3 +++
>>  fs/cifs/connect.c  |    4 ++++
>>  fs/cifs/fscache.c  |   47 +++++++++++++++++++++++++++++++++++++++++++++++
>>  fs/cifs/fscache.h  |   12 ++++++++++++
>>  6 files changed, 92 insertions(+), 1 deletion(-)
>>  create mode 100644 fs/cifs/fscache.c
>>
>> Index: cifs-2.6/fs/cifs/Makefile
>> ===================================================================
>> --- cifs-2.6.orig/fs/cifs/Makefile
>> +++ cifs-2.6/fs/cifs/Makefile
>> @@ -12,4 +12,4 @@ cifs-$(CONFIG_CIFS_UPCALL) += cifs_spneg
>>  
>>  cifs-$(CONFIG_CIFS_DFS_UPCALL) += dns_resolve.o cifs_dfs_ref.o
>>  
>> -cifs-$(CONFIG_CIFS_FSCACHE) += cache.o
>> +cifs-$(CONFIG_CIFS_FSCACHE) += fscache.o cache.o
>> Index: cifs-2.6/fs/cifs/cache.c
>> ===================================================================
>> --- cifs-2.6.orig/fs/cifs/cache.c
>> +++ cifs-2.6/fs/cifs/cache.c
>> @@ -51,3 +51,28 @@ void cifs_fscache_unregister(void)
>>  	fscache_unregister_netfs(&cifs_fscache_netfs);
>>  }
>>  
>> +/*
>> + * Server object currently keyed by hostname
>> + */
>> +static uint16_t cifs_server_get_key(const void *cookie_netfs_data,
>> +				   void *buffer, uint16_t maxbuf)
>> +{
>> +	const struct TCP_Server_Info *server = cookie_netfs_data;
>> +	uint16_t len = strnlen(server->hostname, sizeof(server->hostname));
>> +
> 
> Would a tuple of address/family/port be a better choice here? Imagine I
> mount "foo" and then later mount "foor.bar.baz". If they are the same
> address and only the UNC differs, then you won't get the benefit of
> the cache, right?
> 

Good point. I'll fix it up when I do a respin.

Thanks,

-- 
Suresh Jayaraman

Re: [RFC][PATCH 03/10] cifs: register CIFS for caching

[David Howells]

Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> +	rc = cifs_fscache_register();
> +	if (rc)
> +		goto out;
> +
>  	rc = cifs_init_inodecache();
>  	if (rc)
>  		goto out_clean_proc;
> @@ -949,8 +954,10 @@ init_cifs(void)
>  	cifs_destroy_mids();
>   out_destroy_inodecache:
>  	cifs_destroy_inodecache();
> +	cifs_fscache_unregister();
>   out_clean_proc:

This is incorrect.  You need to call cifs_fscache_unregister() if
cifs_init_inodecache() fails.

David

Re: [RFC][PATCH 04/10] cifs: define server-level cache index objects and register them with FS-Cache

[David Howells]

Suresh Jayaraman <sjayaraman@suse.de> wrote:

> Define server-level cache index objects (as managed by TCP_ServerInfo
> structs).  Each server object is created in the CIFS top-level index object
> and is itself an index into which superblock-level objects are inserted.
> 
> Currently, the server objects are keyed by hostname.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>

Looks reasonable, apart from the index key.  I agree with Jeff that you
probably want {address,port,family} rather than a hostname.

David

Re: [RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[David Howells]

Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> Define superblock-level cache index objects (managed by cifsTconInfo
> structs).  Each superblock object is created in a server-level index object
> and in itself an index into which inode-level objects are inserted.
> 
> Currently, the superblock objects are keyed by sharename.

Seems reasonable.  Is there any way you can check that the share you are
looking at on a server is the same as the last time you looked?  Can you
validate the root directory of the share in some way?

David

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[David Howells]

Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> Define inode-level data storage objects (managed by cifsInodeInfo structs).
> Each inode-level object is created in a super-block level object and is
> itself a data storage object in to which pages from the inode are stored.
> 
> The inode object is keyed by UniqueId. The coherency data being used is
> LastWriteTime and the file size.

Isn't there a file creation time too?

I take it you don't support caching on files that are open for writing at this
time?

David

Re: [RFC][PATCH 07/10] cifs: FS-Cache page management

[David Howells]

Suresh Jayaraman <sjayaraman@suse.de> wrote:

> Takes care of invalidation and release of FS-Cache marked pages and also
> invalidation of the FsCache page flag when the inode is removed.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>

Acked-by: David Howells <dhowells@redhat.com>

Re: [RFC][PATCH 08/10] cifs: store pages into local cache

[David Howells]

Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> Store pages from an CIFS inode into the data storage object associated with
> that inode.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>

Acked-by: David Howells <dhowells-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>

Re: [RFC][PATCH 09/10] cifs: read pages from FS-Cache

[David Howells]

Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> Read pages from a FS-Cache data storage object into a CIFS inode.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org>

Acked-by: David Howells <dhowells-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>

Re: [RFC][PATCH 10/10] cifs: add mount option to enable local caching

[David Howells]

Suresh Jayaraman <sjayaraman@suse.de> wrote:

> Add a mount option 'fsc' to enable local caching on CIFS.
> 
> As the cifs-utils (userspace) changes are not done yet, this patch enables
> 'fsc' by default to assist testing.
> 
> Signed-off-by: Suresh Jayaraman <sjayaraman@suse.de>

Acked-by: David Howells <dhowells@redhat.com>

(Give or take the debugging bit)

Re: [RFC][PATCH 10/10] cifs: add mount option to enable local caching

[Scott Lovenberg]

On 6/22/2010 11:25 AM, Suresh Jayaraman wrote:
> Add a mount option 'fsc' to enable local caching on CIFS.
>
> As the cifs-utils (userspace) changes are not done yet, this patch enables
> 'fsc' by default to assist testing.
>    
[...]
> @@ -1332,6 +1336,8 @@ cifs_parse_mount_options(char *options, const char *devname,
>   			printk(KERN_WARNING "CIFS: Mount option noac not "
>   				"supported. Instead set "
>   				"/proc/fs/cifs/LookupCacheEnabled to 0\n");
> +		} else if (strnicmp(data, "fsc", 3) == 0) {
> +			vol->fsc = true;
>   		} else
>   			printk(KERN_WARNING "CIFS: Unknown mount option %s\n",
>   						data);
> @@ -2405,6 +2411,8 @@ static void setup_cifs_sb(struct smb_vol *pvolume_info,
>   		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_OVERR_GID;
>   	if (pvolume_info->dynperm)
>   		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_DYNPERM;
> +	if (pvolume_info->fsc)
> +		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_FSCACHE;
>   	if (pvolume_info->direct_io) {
>   		cFYI(1, "mounting share using direct i/o");
>   		cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_DIRECT_IO;
>    
I reworked the CIFS mount option parsing a while back; I'm not sure 
whether that patch was going to be in the 2.6.35 tree or not (the window 
just opened, didn't it?).

Jeff, Steve, can you confirm if that patch is going to be in 2.6.35?

Patch refs: http://patchwork.ozlabs.org/patch/53059/  and 
http://patchwork.ozlabs.org/patch/53674/

Re: [RFC][PATCH 10/10] cifs: add mount option to enable local caching

[Suresh Jayaraman]

On 06/24/2010 12:02 AM, Scott Lovenberg wrote:
> On 6/22/2010 11:25 AM, Suresh Jayaraman wrote:
>> Add a mount option 'fsc' to enable local caching on CIFS.
>>
>> As the cifs-utils (userspace) changes are not done yet, this patch
>> enables
>> 'fsc' by default to assist testing.
>>    
> [...]
>> @@ -1332,6 +1336,8 @@ cifs_parse_mount_options(char *options, const
>> char *devname,
>>               printk(KERN_WARNING "CIFS: Mount option noac not "
>>                   "supported. Instead set "
>>                   "/proc/fs/cifs/LookupCacheEnabled to 0\n");
>> +        } else if (strnicmp(data, "fsc", 3) == 0) {
>> +            vol->fsc = true;
>>           } else
>>               printk(KERN_WARNING "CIFS: Unknown mount option %s\n",
>>                           data);
>> @@ -2405,6 +2411,8 @@ static void setup_cifs_sb(struct smb_vol
>> *pvolume_info,
>>           cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_OVERR_GID;
>>       if (pvolume_info->dynperm)
>>           cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_DYNPERM;
>> +    if (pvolume_info->fsc)
>> +        cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_FSCACHE;
>>       if (pvolume_info->direct_io) {
>>           cFYI(1, "mounting share using direct i/o");
>>           cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_DIRECT_IO;
>>    
> I reworked the CIFS mount option parsing a while back; I'm not sure
> whether that patch was going to be in the 2.6.35 tree or not (the window
> just opened, didn't it?).

Not a problem, I could redo this patch alone when the reworked option
parsing patches get in.

> Jeff, Steve, can you confirm if that patch is going to be in 2.6.35?
> 
> Patch refs: http://patchwork.ozlabs.org/patch/53059/  and
> http://patchwork.ozlabs.org/patch/53674/
> 

Thanks,

-- 
Suresh Jayaraman

Re: [RFC][PATCH 03/10] cifs: register CIFS for caching

[Suresh Jayaraman]

On 06/23/2010 10:21 PM, David Howells wrote:
> Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:
> 
>> +	rc = cifs_fscache_register();
>> +	if (rc)
>> +		goto out;
>> +
>>  	rc = cifs_init_inodecache();
>>  	if (rc)
>>  		goto out_clean_proc;
>> @@ -949,8 +954,10 @@ init_cifs(void)
>>  	cifs_destroy_mids();
>>   out_destroy_inodecache:
>>  	cifs_destroy_inodecache();
>> +	cifs_fscache_unregister();
>>   out_clean_proc:
> 
> This is incorrect.  You need to call cifs_fscache_unregister() if
> cifs_init_inodecache() fails.
> 

Doh! I'll fix it.


-- 
Suresh Jayaraman

Re: [RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[Suresh Jayaraman]

On 06/23/2010 10:28 PM, David Howells wrote:
> Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:
> 
>> Define superblock-level cache index objects (managed by cifsTconInfo
>> structs).  Each superblock object is created in a server-level index object
>> and in itself an index into which inode-level objects are inserted.
>>
>> Currently, the superblock objects are keyed by sharename.
> 
> Seems reasonable.  Is there any way you can check that the share you are
> looking at on a server is the same as the last time you looked?  Can you

Good point.

I thought of using TID (Tree identifier; a unique ID for a resource in
use by client) along with sharename. But, Server is free to reuse them
when the tree connection closes and does not guarantee the same Tid for
a particular resource across tree connections.

Also, considering the UNC name of the resource (//server/share) may not
be a good idea too as the cache will not be used when for e.g. IPaddress
is used to mount.

So, if a server does something like this:
   - export a share 'foo' (original server path: /export/vol1/foo)
   - client mounts and uses it
   - server unexports the share 'foo'
   - server exports 'foo' (original sever path: /export/vol2/foo)

we have a bit of problem..

> validate the root directory of the share in some way?
> 

I don't know if there is a way to do this.

Thanks,


-- 
Suresh Jayaraman

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Suresh Jayaraman]

On 06/23/2010 10:32 PM, David Howells wrote:
> Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:
> 
>> Define inode-level data storage objects (managed by cifsInodeInfo structs).
>> Each inode-level object is created in a super-block level object and is
>> itself a data storage object in to which pages from the inode are stored.
>>
>> The inode object is keyed by UniqueId. The coherency data being used is
>> LastWriteTime and the file size.
> 
> Isn't there a file creation time too?

I think the creation time is currently being ignored as we won't be able
to accomodate in POSIX stat struct.

> I take it you don't support caching on files that are open for writing at this
> time?
> 

Yes.


-- 
Suresh Jayaraman

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[David Howells]

Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:

> I think the creation time is currently being ignored as we won't be able
> to accomodate in POSIX stat struct.

The FS-Cache interface doesn't use the POSIX stat struct, but it could be
really useful to save it and use it for cache coherency inside the kernel.

Out of interest, what does Samba do when it comes to generating a creation time
for UNIX where one does not exist?

David

Re: [RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[David Howells]

Suresh Jayaraman <sjayaraman@suse.de> wrote:

> Also, considering the UNC name of the resource (//server/share) may not
> be a good idea too as the cache will not be used when for e.g. IPaddress
> is used to mount.

You could convert the UNC name to an IP address, and just use that as your
key.

> > validate the root directory of the share in some way?
>
> I don't know if there is a way to do this.

Is there an inode number or something?  Even the creation time might do.

David

Re: [RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[David Howells]

David Howells <dhowells-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:

> > > validate the root directory of the share in some way?
> >
> > I don't know if there is a way to do this.
> 
> Is there an inode number or something?  Even the creation time might do.

Looking in cifspdu.h, there are a number of things that it might be possible
to use.

 (1) FILE_ALL_INFO: CreationTime, IndexNumber, IndexNumber1, FileName
     (assuming this isn't flattened to '\' or something for the root of a
     share.

 (2) FILE_UNIX_BASIC_INFO: DevMajor, DevMinor, UniqueId.

 (3) FILE_INFO_STANDARD: CreationDate, CreationTime.

 (4) FILE_INFO_BASIC: CreationTime.

 (5) FILE_DIRECTORY_INFO: FileIndex, CreationTime, FileName.

 (6) SEARCH_ID_FULL_DIR_INFO: FileIndex, CreationTime, UniqueId, FileName.

 (7) FILE_BOTH_DIRECTORY_INFO: FileIndex, CreationTime, ShortName, FileName.

 (8) OPEN_RSP_EXT: Fid, CreationTime, VolumeGUID, FileId.

You may have to choose different sets of things, depending on what the server
has on offer.  Also, don't forget, if you can't work out whether a share is
coherent or not from the above, you can always use LastWriteTime, ChangeTime
and EndOfFile and just discard the whole subtree if they differ.

David

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Jeff Layton]

On Fri, 25 Jun 2010 13:55:49 +0100
David Howells <dhowells-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:

> Suresh Jayaraman <sjayaraman-l3A5Bk7waGM@public.gmane.org> wrote:
> 
> > I think the creation time is currently being ignored as we won't be able
> > to accomodate in POSIX stat struct.
> 
> The FS-Cache interface doesn't use the POSIX stat struct, but it could be
> really useful to save it and use it for cache coherency inside the kernel.
> 
> Out of interest, what does Samba do when it comes to generating a creation time
> for UNIX where one does not exist?
> 

(cc'ing samba-technical since we're talking about the create time)

Looks like it mostly uses the ctime. IMO, the mtime would be a better
choice since it changes less frequently, but I don't guess that it
matters very much.

I have a few patches that make the cifs_iget code do more stringent
checks. One of those makes it use the create time like an i_generation
field to guard against matching inodes that have the same number but
that have undergone a delete/create cycle. They need a bit more testing
but I'm planning to post them in time for 2.6.36.

Because of how samba generates this number, it could be somewhat
problematic to do this. What may save us though is that Linux<->Samba
mostly uses unix extensions unless someone has specifically disabled
them on either end. The unix extension calls don't generally send any
sort of create time field, so we can't rely on it in those codepaths
anyway.

-- 
Jeff Layton <jlayton-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org>

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[David Howells]

Jeff Layton <jlayton@samba.org> wrote:

> Looks like it mostly uses the ctime. IMO, the mtime would be a better
> choice since it changes less frequently, but I don't guess that it
> matters very much.

I'd've thought mtime changes more frequently since that's altered when data is
written.  ctime is changed when attributes are changed.

Note that Ext4 appears to have a file creation time field in its inode
(struct ext4_inode::i_crtime[_extra]).  Can Samba be made to use that?

David

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Jeff Layton]

On Fri, 25 Jun 2010 22:46:38 +0100
David Howells <dhowells@redhat.com> wrote:

> Jeff Layton <jlayton@samba.org> wrote:
> 
> > Looks like it mostly uses the ctime. IMO, the mtime would be a better
> > choice since it changes less frequently, but I don't guess that it
> > matters very much.
> 
> I'd've thought mtime changes more frequently since that's altered when data is
> written.  ctime is changed when attributes are changed.
> 

IIUC, updating mtime for a write is also an attribute change, and that
affects ctime. According to the stat(2) manpage:

       The field st_ctime is changed by writing or by setting  inode  informa-
       tion (i.e., owner, group, link count, mode, etc.).

> Note that Ext4 appears to have a file creation time field in its inode
> (struct ext4_inode::i_crtime[_extra]).  Can Samba be made to use that?
> 

Is it exposed to userspace in any (standard) way? It would be handy to
have that. While we're wishing...it might also be nice to have a
standard way to get at the i_generation from userspace too.

-- 
Jeff Layton <jlayton@samba.org>

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[David Howells]

Jeff Layton <jlayton-eUNUBHrolfbYtjvyW6yDsg@public.gmane.org> wrote:

> IIUC, updating mtime for a write is also an attribute change, and that
> affects ctime. According to the stat(2) manpage:

You're right.  Okay, ctime is the more frequently changed.

> > Note that Ext4 appears to have a file creation time field in its inode
> > (struct ext4_inode::i_crtime[_extra]).  Can Samba be made to use that?
> 
> Is it exposed to userspace in any (standard) way? It would be handy to
> have that. While we're wishing...it might also be nice to have a
> standard way to get at the i_generation from userspace too.

Not at present, but it's something that could be exported by ioctl() or
getxattr().

David

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Steve French]

On Fri, Jun 25, 2010 at 5:26 PM, Jeff Layton <jlayton@samba.org> wrote:
>
> On Fri, 25 Jun 2010 22:46:38 +0100
> David Howells <dhowells@redhat.com> wrote:
>
> > Jeff Layton <jlayton@samba.org> wrote:
> >
> > > Looks like it mostly uses the ctime. IMO, the mtime would be a better
> > > choice since it changes less frequently, but I don't guess that it
> > > matters very much.
> >
> > I'd've thought mtime changes more frequently since that's altered when data is
> > written.  ctime is changed when attributes are changed.
> >
>
> IIUC, updating mtime for a write is also an attribute change, and that
> affects ctime. According to the stat(2) manpage:
>
>       The field st_ctime is changed by writing or by setting  inode  informa-
>       tion (i.e., owner, group, link count, mode, etc.).
>
> > Note that Ext4 appears to have a file creation time field in its inode
> > (struct ext4_inode::i_crtime[_extra]).  Can Samba be made to use that?
> >
>
> Is it exposed to userspace in any (standard) way? It would be handy to
> have that. While we're wishing...it might also be nice to have a
> standard way to get at the i_generation from userspace too.
>

Yes - I have talked with MingMing and Aneesh about those (NFS may
someday be able to use those too).  An obstacle in the past had been
that samba server stores its own fake creation time in an ndr encoded
xattr which complicates things.

MingMing/Annesh -
Xattr or other way to get at birth time?


--
Thanks,

Steve

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Mingming Cao]

Steve French <smfrench@gmail.com> wrote on 06/25/2010 04:05:30 PM:

> Steve French <smfrench@gmail.com>
> 06/25/2010 04:05 PM
>
> To
>
> Jeff Layton <jlayton@samba.org>, "Aneesh Kumar K.V"
> <aneesh.kumar@linux.vnet.ibm.com>, Mingming Cao/Beaverton/IBM@IBMUS
>
> cc
>
> David Howells <dhowells@redhat.com>, Suresh Jayaraman
> <sjayaraman@suse.de>, linux-cifs@vger.kernel.org, linux-
> fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, samba-
> technical@lists.samba.org, Jeff Layton <jlayton@redhat.com>
>
> Subject
>
> Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and
> register them
>
> On Fri, Jun 25, 2010 at 5:26 PM, Jeff Layton <jlayton@samba.org> wrote:
> >
> > On Fri, 25 Jun 2010 22:46:38 +0100
> > David Howells <dhowells@redhat.com> wrote:
> >
> > > Jeff Layton <jlayton@samba.org> wrote:
> > >
> > > > Looks like it mostly uses the ctime. IMO, the mtime would be a
better
> > > > choice since it changes less frequently, but I don't guess that it
> > > > matters very much.
> > >
> > > I'd've thought mtime changes more frequently since that's
> altered when data is
> > > written.  ctime is changed when attributes are changed.
> > >
> >
> > IIUC, updating mtime for a write is also an attribute change, and that
> > affects ctime. According to the stat(2) manpage:
> >
> >       The field st_ctime is changed by writing or by setting
>  inode  informa-
> >       tion (i.e., owner, group, link count, mode, etc.).
> >
> > > Note that Ext4 appears to have a file creation time field in its
inode
> > > (struct ext4_inode::i_crtime[_extra]).  Can Samba be made to use
that?
> > >
> >
> > Is it exposed to userspace in any (standard) way? It would be handy to
> > have that. While we're wishing...it might also be nice to have a
> > standard way to get at the i_generation from userspace too.
> >
>
> Yes - I have talked with MingMing and Aneesh about those (NFS may
> someday be able to use those too).  An obstacle in the past had been
> that samba server stores its own fake creation time in an ndr encoded
> xattr which complicates things.
>
> MingMing/Annesh -
> Xattr or other way to get at birth time?
>
>

Not yet,
 The ext4 file creation time only accesable from the kernel at the moment.
There were discussion
to make this information avaliable via xattr before, but was rejected,
since most people
agree that making this info avalibele via stat() is more standard. However
modifying stat() would imply
big interface change. thus no action has been taken yet.

> --
> Thanks,
>
> Steve

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Aneesh Kumar K. V]

On Fri, 25 Jun 2010 17:52:24 -0700, Mingming Cao <mcao@us.ibm.com> wrote:
> 
> 
> Steve French <smfrench@gmail.com> wrote on 06/25/2010 04:05:30 PM:
> 
> > Steve French <smfrench@gmail.com>
> > 06/25/2010 04:05 PM
> >
> > To
> >
> > Jeff Layton <jlayton@samba.org>, "Aneesh Kumar K.V"
> > <aneesh.kumar@linux.vnet.ibm.com>, Mingming Cao/Beaverton/IBM@IBMUS
> >
> > cc
> >
> > David Howells <dhowells@redhat.com>, Suresh Jayaraman
> > <sjayaraman@suse.de>, linux-cifs@vger.kernel.org, linux-
> > fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, samba-
> > technical@lists.samba.org, Jeff Layton <jlayton@redhat.com>
> >
> > Subject
> >
> > Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and
> > register them
> >
> > On Fri, Jun 25, 2010 at 5:26 PM, Jeff Layton <jlayton@samba.org> wrote:
> > >
> > > On Fri, 25 Jun 2010 22:46:38 +0100
> > > David Howells <dhowells@redhat.com> wrote:
> > >
> > > > Jeff Layton <jlayton@samba.org> wrote:
> > > >
> > > > > Looks like it mostly uses the ctime. IMO, the mtime would be a
> better
> > > > > choice since it changes less frequently, but I don't guess that it
> > > > > matters very much.
> > > >
> > > > I'd've thought mtime changes more frequently since that's
> > altered when data is
> > > > written.  ctime is changed when attributes are changed.
> > > >
> > >
> > > IIUC, updating mtime for a write is also an attribute change, and that
> > > affects ctime. According to the stat(2) manpage:
> > >
> > >       The field st_ctime is changed by writing or by setting
> >  inode  informa-
> > >       tion (i.e., owner, group, link count, mode, etc.).
> > >
> > > > Note that Ext4 appears to have a file creation time field in its
> inode
> > > > (struct ext4_inode::i_crtime[_extra]).  Can Samba be made to use
> that?
> > > >
> > >
> > > Is it exposed to userspace in any (standard) way? It would be handy to
> > > have that. While we're wishing...it might also be nice to have a
> > > standard way to get at the i_generation from userspace too.
> > >
> >
> > Yes - I have talked with MingMing and Aneesh about those (NFS may
> > someday be able to use those too).  An obstacle in the past had been
> > that samba server stores its own fake creation time in an ndr encoded
> > xattr which complicates things.
> >
> > MingMing/Annesh -
> > Xattr or other way to get at birth time?
> >
> >
> 
> Not yet,
>  The ext4 file creation time only accesable from the kernel at the moment.
> There were discussion
> to make this information avaliable via xattr before, but was rejected,
> since most people
> agree that making this info avalibele via stat() is more standard. However
> modifying stat() would imply
> big interface change. thus no action has been taken yet.

NFS ganesha pNFS also had a requirement for getting i_generation and
inode number in userspace. So may be we should now look at updating
stat or add a variant syscall that include i_generation and create time
in the return value

-aneesh

Re: [RFC][PATCH 06/10] cifs: define inode-level cache object and register them

[Christoph Hellwig]

On Sun, Jun 27, 2010 at 11:47:21PM +0530, Aneesh Kumar K. V wrote:
> NFS ganesha pNFS also had a requirement for getting i_generation and
> inode number in userspace. So may be we should now look at updating
> stat or add a variant syscall that include i_generation and create time
> in the return value

What's missing in knfsd that you feel the sudden urge to move backwards
to a userspace nfsd (one with a horribly crappy codebase, too).

Re: [RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[Suresh Jayaraman]

On 06/25/2010 06:56 PM, David Howells wrote:
> David Howells <dhowells-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
> 
>>>> validate the root directory of the share in some way?
>>>
>>> I don't know if there is a way to do this.
>>
>> Is there an inode number or something?  Even the creation time might do.
> 
> Looking in cifspdu.h, there are a number of things that it might be possible
> to use.
> 
>  (1) FILE_ALL_INFO: CreationTime, IndexNumber, IndexNumber1, FileName
>      (assuming this isn't flattened to '\' or something for the root of a
>      share.
> 
>  (2) FILE_UNIX_BASIC_INFO: DevMajor, DevMinor, UniqueId.
> 
>  (3) FILE_INFO_STANDARD: CreationDate, CreationTime.
> 
>  (4) FILE_INFO_BASIC: CreationTime.
> 
>  (5) FILE_DIRECTORY_INFO: FileIndex, CreationTime, FileName.
> 
>  (6) SEARCH_ID_FULL_DIR_INFO: FileIndex, CreationTime, UniqueId, FileName.
> 
>  (7) FILE_BOTH_DIRECTORY_INFO: FileIndex, CreationTime, ShortName, FileName.
> 
>  (8) OPEN_RSP_EXT: Fid, CreationTime, VolumeGUID, FileId.
> 
> You may have to choose different sets of things, depending on what the server
> has on offer.  Also, don't forget, if you can't work out whether a share is

Did you mean we need to validate differently for different servers?

I just did some testing and it looks like we could rely on CreationTime,
IndexNumber for validating with Windows servers (FileName is relative to
the mapped drive) and UniqueId for validating with Samba servers. I did
not test all possibilities (there could be more).

> coherent or not from the above, you can always use LastWriteTime, ChangeTime
> and EndOfFile and just discard the whole subtree if they differ.
> 

Thanks,

-- 
Suresh Jayaraman

Re: [RFC][PATCH 05/10] cifs: define superblock-level cache index objects and register them

[David Howells]

Suresh Jayaraman <sjayaraman@suse.de> wrote:

> Did you mean we need to validate differently for different servers?

You may need to, yes, as different servers may make different attributes
available.

This isn't too bad.  Each server index record in the cache has freeform
auxiliary data, just as does each file data record.  You could, say, stick a
byte at the front that indicates what you've stored in there.

David