[AF_RXRPC]: Make the in-kernel AFS filesystem use AF_RXRPC.

Make the in-kernel AFS filesystem use AF_RXRPC instead of the old RxRPC code.
Signed-off-by: NDavid Howells <dhowells@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

fs/Kconfig

@@ -2019,7 +2019,7 @@ config CODA_FS_OLD_API
 config AFS_FS
 	tristate "Andrew File System support (AFS) (EXPERIMENTAL)"
 	depends on INET && EXPERIMENTAL
-	select RXRPC
+	select AF_RXRPC
 	help
 	  If you say Y here, you will get an experimental Andrew File System
 	  driver. It currently only supports unsecured read-only AFS access.
@@ -2028,6 +2028,17 @@ config AFS_FS
 
 	  If unsure, say N.
 
+config AFS_DEBUG
+	bool "AFS dynamic debugging"
+	depends on AFS_FS
+	help
+	  Say Y here to make runtime controllable debugging messages appear.
+
+	  See <file:Documentation/filesystems/afs.txt> for more information.
+
+	  If unsure, say N.
+
+
 config RXRPC
 	tristate
 

fs/afs/Makefile

@@ -10,12 +10,11 @@ kafs-objs := \
 	file.o \
 	fsclient.o \
 	inode.o \
-	kafsasyncd.o \
-	kafstimod.o \
 	main.o \
 	misc.o \
 	mntpt.o \
 	proc.o \
+	rxrpc.o \
 	server.o \
 	super.o \
 	vlclient.o \

fs/afs/types.h → fs/afs/afs.h

-/* AFS types
+/* AFS common types
  *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
+ * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
  * Written by David Howells (dhowells@redhat.com)
  *
  * This program is free software; you can redistribute it and/or
@@ -9,10 +9,10 @@
  * 2 of the License, or (at your option) any later version.
  */
 
-#ifndef AFS_TYPES_H
-#define AFS_TYPES_H
+#ifndef AFS_H
+#define AFS_H
 
-#include <rxrpc/types.h>
+#include <linux/in.h>
 
 typedef unsigned			afs_volid_t;
 typedef unsigned			afs_vnodeid_t;
@@ -31,9 +31,6 @@ typedef enum {
 	AFS_FTYPE_SYMLINK	= 3,
 } afs_file_type_t;
 
-struct afs_cell;
-struct afs_vnode;
-
 /*
  * AFS file identifier
  */
@@ -54,14 +51,13 @@ typedef enum {
 } afs_callback_type_t;
 
 struct afs_callback {
-	struct afs_server	*server;	/* server that made the promise */
 	struct afs_fid		fid;		/* file identifier */
 	unsigned		version;	/* callback version */
 	unsigned		expiry;		/* time at which expires */
 	afs_callback_type_t	type;		/* type of callback */
 };
 
-#define AFSCBMAX 50
+#define AFSCBMAX 50	/* maximum callbacks transferred per bulk op */
 
 /*
  * AFS volume information
@@ -88,7 +84,7 @@ struct afs_file_status {
 	afs_file_type_t		type;		/* file type */
 	unsigned		nlink;		/* link count */
 	size_t			size;		/* file size */
-	afs_dataversion_t	version;	/* current data version */
+	afs_dataversion_t	data_version;	/* current data version */
 	unsigned		author;		/* author ID */
 	unsigned		owner;		/* owner ID */
 	unsigned		caller_access;	/* access rights for authenticated caller */
@@ -106,4 +102,4 @@ struct afs_volsync {
 	time_t			creation;	/* volume creation time */
 };
 
-#endif /* AFS_TYPES_H */
+#endif /* AFS_H */

fs/afs/mount.h → fs/afs/afs_cm.h

-/* mount parameters
+/* AFS Cache Manager definitions
  *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
+ * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
  * Written by David Howells (dhowells@redhat.com)
  *
  * This program is free software; you can redistribute it and/or
@@ -9,15 +9,20 @@
  * 2 of the License, or (at your option) any later version.
  */
 
-#ifndef AFS_MOUNT_H
-#define AFS_MOUNT_H
+#ifndef AFS_CM_H
+#define AFS_CM_H
 
-struct afs_mountdata {
-	const char		*volume;	/* name of volume */
-	const char		*cell;		/* name of cell containing volume */
-	const char		*cache;		/* name of cache block device */
-	size_t			nservers;	/* number of server addresses listed */
-	uint32_t		servers[10];	/* IP addresses of servers in this cell */
+#define AFS_CM_PORT		7001	/* AFS file server port */
+#define CM_SERVICE		1	/* AFS File Service ID */
+
+enum AFS_CM_Operations {
+	CBCallBack		= 204,	/* break callback promises */
+	CBInitCallBackState	= 205,	/* initialise callback state */
+	CBProbe			= 206,	/* probe client */
+	CBGetLock		= 207,	/* get contents of CM lock table */
+	CBGetCE			= 208,	/* get cache file description */
+	CBGetXStatsVersion	= 209,	/* get version of extended statistics */
+	CBGetXStats		= 210,	/* get contents of extended statistics data */
 };
 
-#endif /* AFS_MOUNT_H */
+#endif /* AFS_FS_H */

fs/afs/errors.h → fs/afs/afs_fs.h

-/* AFS abort/error codes
+/* AFS File Service definitions
  *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
+ * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
  * Written by David Howells (dhowells@redhat.com)
  *
  * This program is free software; you can redistribute it and/or
@@ -9,15 +9,22 @@
  * 2 of the License, or (at your option) any later version.
  */
 
-#ifndef AFS_ERRORS_H
-#define AFS_ERRORS_H
+#ifndef AFS_FS_H
+#define AFS_FS_H
 
-#include "types.h"
+#define AFS_FS_PORT		7000	/* AFS file server port */
+#define FS_SERVICE		1	/* AFS File Service ID */
 
-/*
- * file server abort codes
- */
-typedef enum {
+enum AFS_FS_Operations {
+	FSFETCHSTATUS		= 132,	/* AFS Fetch file status */
+	FSFETCHDATA		= 130,	/* AFS Fetch file data */
+	FSGIVEUPCALLBACKS	= 147,	/* AFS Discard callback promises */
+	FSGETVOLUMEINFO		= 148,	/* AFS Get root volume information */
+	FSGETROOTVOLUME		= 151,	/* AFS Get root volume name */
+	FSLOOKUP		= 161	/* AFS lookup file in directory */
+};
+
+enum AFS_FS_Errors {
 	VSALVAGE	= 101,	/* volume needs salvaging */
 	VNOVNODE	= 102,	/* no such file/dir (vnode) */
 	VNOVOL		= 103,	/* no such volume or volume unavailable */
@@ -29,8 +36,6 @@ typedef enum {
 	VOVERQUOTA	= 109,	/* volume's maximum quota exceeded */
 	VBUSY		= 110,	/* volume is temporarily unavailable */
 	VMOVED		= 111,	/* volume moved to new server - ask this FS where */
-} afs_rxfs_abort_t;
-
-extern int afs_abort_to_error(int);
+};
 
-#endif /* AFS_ERRORS_H */
+#endif /* AFS_FS_H */

fs/afs/vlclient.h → fs/afs/afs_vl.h

-/* Volume Location Service client interface
+/* AFS Volume Location Service client interface
  *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
+ * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
  * Written by David Howells (dhowells@redhat.com)
  *
  * This program is free software; you can redistribute it and/or
@@ -9,10 +9,19 @@
  * 2 of the License, or (at your option) any later version.
  */
 
-#ifndef AFS_VLCLIENT_H
-#define AFS_VLCLIENT_H
+#ifndef AFS_VL_H
+#define AFS_VL_H
 
-#include "types.h"
+#include "afs.h"
+
+#define AFS_VL_PORT		7003	/* volume location service port */
+#define VL_SERVICE		52	/* RxRPC service ID for the Volume Location service */
+
+enum AFSVL_Operations {
+	VLGETENTRYBYID		= 503,	/* AFS Get Cache Entry By ID operation ID */
+	VLGETENTRYBYNAME	= 504,	/* AFS Get Cache Entry By Name operation ID */
+	VLPROBE			= 514,	/* AFS Probe Volume Location Service operation ID */
+};
 
 enum AFSVL_Errors {
 	AFSVL_IDEXIST 		= 363520,	/* Volume Id entry exists in vl database */
@@ -40,14 +49,16 @@ enum AFSVL_Errors {
 	AFSVL_BADVOLOPER 	= 363542,	/* Bad volume operation code */
 	AFSVL_BADRELLOCKTYPE 	= 363543,	/* Bad release lock type */
 	AFSVL_RERELEASE 	= 363544,	/* Status report: last release was aborted */
-	AFSVL_BADSERVERFLAG 	= 363545,	/* Invalid replication site server ag */
+	AFSVL_BADSERVERFLAG 	= 363545,	/* Invalid replication site server °ag */
 	AFSVL_PERM 		= 363546,	/* No permission access */
 	AFSVL_NOMEM 		= 363547,	/* malloc/realloc failed to alloc enough memory */
 };
 
-/* maps to "struct vldbentry" in vvl-spec.pdf */
+/*
+ * maps to "struct vldbentry" in vvl-spec.pdf
+ */
 struct afs_vldbentry {
-	char		name[65];		/* name of volume (including NUL char) */
+	char		name[65];		/* name of volume (with NUL char) */
 	afs_voltype_t	type;			/* volume type */
 	unsigned	num_servers;		/* num servers that hold instances of this vol */
 	unsigned	clone_id;		/* cloning ID */
@@ -70,16 +81,4 @@ struct afs_vldbentry {
 	} servers[8];
 };
 
-extern int afs_rxvl_get_entry_by_name(struct afs_server *, const char *,
-				      unsigned, struct afs_cache_vlocation *);
-extern int afs_rxvl_get_entry_by_id(struct afs_server *, afs_volid_t,
-				    afs_voltype_t,
-				    struct afs_cache_vlocation *);
-
-extern int afs_rxvl_get_entry_by_id_async(struct afs_async_op *,
-					  afs_volid_t, afs_voltype_t);
-
-extern int afs_rxvl_get_entry_by_id_async2(struct afs_async_op *,
-					   struct afs_cache_vlocation *);
-
-#endif /* AFS_VLCLIENT_H */
+#endif /* AFS_VL_H */

fs/afs/cache.c

+/* AFS caching stuff
+ *
+ * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_cell_cache_match(void *target,
+						const void *entry);
+static void afs_cell_cache_update(void *source, void *entry);
+
+struct cachefs_index_def afs_cache_cell_index_def = {
+	.name			= "cell_ix",
+	.data_size		= sizeof(struct afs_cache_cell),
+	.keys[0]		= { CACHEFS_INDEX_KEYS_ASCIIZ, 64 },
+	.match			= afs_cell_cache_match,
+	.update			= afs_cell_cache_update,
+};
+#endif
+
+/*
+ * match a cell record obtained from the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_cell_cache_match(void *target,
+						const void *entry)
+{
+	const struct afs_cache_cell *ccell = entry;
+	struct afs_cell *cell = target;
+
+	_enter("{%s},{%s}", ccell->name, cell->name);
+
+	if (strncmp(ccell->name, cell->name, sizeof(ccell->name)) == 0) {
+		_leave(" = SUCCESS");
+		return CACHEFS_MATCH_SUCCESS;
+	}
+
+	_leave(" = FAILED");
+	return CACHEFS_MATCH_FAILED;
+}
+#endif
+
+/*
+ * update a cell record in the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static void afs_cell_cache_update(void *source, void *entry)
+{
+	struct afs_cache_cell *ccell = entry;
+	struct afs_cell *cell = source;
+
+	_enter("%p,%p", source, entry);
+
+	strncpy(ccell->name, cell->name, sizeof(ccell->name));
+
+	memcpy(ccell->vl_servers,
+	       cell->vl_addrs,
+	       min(sizeof(ccell->vl_servers), sizeof(cell->vl_addrs)));
+
+}
+#endif
+
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_vlocation_cache_match(void *target,
+						     const void *entry);
+static void afs_vlocation_cache_update(void *source, void *entry);
+
+struct cachefs_index_def afs_vlocation_cache_index_def = {
+	.name		= "vldb",
+	.data_size	= sizeof(struct afs_cache_vlocation),
+	.keys[0]	= { CACHEFS_INDEX_KEYS_ASCIIZ, 64 },
+	.match		= afs_vlocation_cache_match,
+	.update		= afs_vlocation_cache_update,
+};
+#endif
+
+/*
+ * match a VLDB record stored in the cache
+ * - may also load target from entry
+ */
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_vlocation_cache_match(void *target,
+						     const void *entry)
+{
+	const struct afs_cache_vlocation *vldb = entry;
+	struct afs_vlocation *vlocation = target;
+
+	_enter("{%s},{%s}", vlocation->vldb.name, vldb->name);
+
+	if (strncmp(vlocation->vldb.name, vldb->name, sizeof(vldb->name)) == 0
+	    ) {
+		if (!vlocation->valid ||
+		    vlocation->vldb.rtime == vldb->rtime
+		    ) {
+			vlocation->vldb = *vldb;
+			vlocation->valid = 1;
+			_leave(" = SUCCESS [c->m]");
+			return CACHEFS_MATCH_SUCCESS;
+		} else if (memcmp(&vlocation->vldb, vldb, sizeof(*vldb)) != 0) {
+			/* delete if VIDs for this name differ */
+			if (memcmp(&vlocation->vldb.vid,
+				   &vldb->vid,
+				   sizeof(vldb->vid)) != 0) {
+				_leave(" = DELETE");
+				return CACHEFS_MATCH_SUCCESS_DELETE;
+			}
+
+			_leave(" = UPDATE");
+			return CACHEFS_MATCH_SUCCESS_UPDATE;
+		} else {
+			_leave(" = SUCCESS");
+			return CACHEFS_MATCH_SUCCESS;
+		}
+	}
+
+	_leave(" = FAILED");
+	return CACHEFS_MATCH_FAILED;
+}
+#endif
+
+/*
+ * update a VLDB record stored in the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static void afs_vlocation_cache_update(void *source, void *entry)
+{
+	struct afs_cache_vlocation *vldb = entry;
+	struct afs_vlocation *vlocation = source;
+
+	_enter("");
+
+	*vldb = vlocation->vldb;
+}
+#endif
+
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_volume_cache_match(void *target,
+						  const void *entry);
+static void afs_volume_cache_update(void *source, void *entry);
+
+struct cachefs_index_def afs_volume_cache_index_def = {
+	.name		= "volume",
+	.data_size	= sizeof(struct afs_cache_vhash),
+	.keys[0]	= { CACHEFS_INDEX_KEYS_BIN, 1 },
+	.keys[1]	= { CACHEFS_INDEX_KEYS_BIN, 1 },
+	.match		= afs_volume_cache_match,
+	.update		= afs_volume_cache_update,
+};
+#endif
+
+/*
+ * match a volume hash record stored in the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_volume_cache_match(void *target,
+						  const void *entry)
+{
+	const struct afs_cache_vhash *vhash = entry;
+	struct afs_volume *volume = target;
+
+	_enter("{%u},{%u}", volume->type, vhash->vtype);
+
+	if (volume->type == vhash->vtype) {
+		_leave(" = SUCCESS");
+		return CACHEFS_MATCH_SUCCESS;
+	}
+
+	_leave(" = FAILED");
+	return CACHEFS_MATCH_FAILED;
+}
+#endif
+
+/*
+ * update a volume hash record stored in the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static void afs_volume_cache_update(void *source, void *entry)
+{
+	struct afs_cache_vhash *vhash = entry;
+	struct afs_volume *volume = source;
+
+	_enter("");
+
+	vhash->vtype = volume->type;
+}
+#endif
+
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_vnode_cache_match(void *target,
+						 const void *entry);
+static void afs_vnode_cache_update(void *source, void *entry);
+
+struct cachefs_index_def afs_vnode_cache_index_def = {
+	.name		= "vnode",
+	.data_size	= sizeof(struct afs_cache_vnode),
+	.keys[0]	= { CACHEFS_INDEX_KEYS_BIN, 4 },
+	.match		= afs_vnode_cache_match,
+	.update		= afs_vnode_cache_update,
+};
+#endif
+
+/*
+ * match a vnode record stored in the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static cachefs_match_val_t afs_vnode_cache_match(void *target,
+						 const void *entry)
+{
+	const struct afs_cache_vnode *cvnode = entry;
+	struct afs_vnode *vnode = target;
+
+	_enter("{%x,%x,%Lx},{%x,%x,%Lx}",
+	       vnode->fid.vnode,
+	       vnode->fid.unique,
+	       vnode->status.version,
+	       cvnode->vnode_id,
+	       cvnode->vnode_unique,
+	       cvnode->data_version);
+
+	if (vnode->fid.vnode != cvnode->vnode_id) {
+		_leave(" = FAILED");
+		return CACHEFS_MATCH_FAILED;
+	}
+
+	if (vnode->fid.unique != cvnode->vnode_unique ||
+	    vnode->status.version != cvnode->data_version) {
+		_leave(" = DELETE");
+		return CACHEFS_MATCH_SUCCESS_DELETE;
+	}
+
+	_leave(" = SUCCESS");
+	return CACHEFS_MATCH_SUCCESS;
+}
+#endif
+
+/*
+ * update a vnode record stored in the cache
+ */
+#ifdef AFS_CACHING_SUPPORT
+static void afs_vnode_cache_update(void *source, void *entry)
+{
+	struct afs_cache_vnode *cvnode = entry;
+	struct afs_vnode *vnode = source;
+
+	_enter("");
+
+	cvnode->vnode_id	= vnode->fid.vnode;
+	cvnode->vnode_unique	= vnode->fid.unique;
+	cvnode->data_version	= vnode->status.version;
+}
+#endif

fs/afs/callback.c

 /*
- * Copyright (c) 2002 Red Hat, Inc. All rights reserved.
+ * Copyright (c) 2002, 2007 Red Hat, Inc. All rights reserved.
  *
  * This software may be freely redistributed under the terms of the
  * GNU General Public License.
@@ -16,83 +16,182 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/init.h>
-#include "server.h"
-#include "vnode.h"
+#include <linux/circ_buf.h>
 #include "internal.h"
-#include "cmservice.h"
+
+unsigned afs_vnode_update_timeout = 10;
+
+#define afs_breakring_space(server) \
+	CIRC_SPACE((server)->cb_break_head, (server)->cb_break_tail,	\
+		   ARRAY_SIZE((server)->cb_break))
+
+//static void afs_callback_updater(struct work_struct *);
+
+static struct workqueue_struct *afs_callback_update_worker;
 
 /*
  * allow the fileserver to request callback state (re-)initialisation
  */
-int SRXAFSCM_InitCallBackState(struct afs_server *server)
+void afs_init_callback_state(struct afs_server *server)
 {
-	struct list_head callbacks;
-
-	_enter("%p", server);
+	struct afs_vnode *vnode;
 
-	INIT_LIST_HEAD(&callbacks);
+	_enter("{%p}", server);
 
-	/* transfer the callback list from the server to a temp holding area */
 	spin_lock(&server->cb_lock);
 
-	list_add(&callbacks, &server->cb_promises);
-	list_del_init(&server->cb_promises);
+	/* kill all the promises on record from this server */
+	while (!RB_EMPTY_ROOT(&server->cb_promises)) {
+		vnode = rb_entry(server->cb_promises.rb_node,
+				 struct afs_vnode, cb_promise);
+		printk("\nUNPROMISE on %p\n", vnode);
+		rb_erase(&vnode->cb_promise, &server->cb_promises);
+		vnode->cb_promised = false;
+	}
 
-	/* munch our way through the list, grabbing the inode, dropping all the
-	 * locks and regetting them in the right order
+	spin_unlock(&server->cb_lock);
+	_leave("");
+}
+
+/*
+ * handle the data invalidation side of a callback being broken
  */
-	while (!list_empty(&callbacks)) {
-		struct afs_vnode *vnode;
-		struct inode *inode;
+void afs_broken_callback_work(struct work_struct *work)
+{
+	struct afs_vnode *vnode =
+		container_of(work, struct afs_vnode, cb_broken_work);
 
-		vnode = list_entry(callbacks.next, struct afs_vnode, cb_link);
-		list_del_init(&vnode->cb_link);
+	_enter("");
 
-		/* try and grab the inode - may fail */
-		inode = igrab(AFS_VNODE_TO_I(vnode));
-		if (inode) {
-			int release = 0;
+	if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
+		return;
 
-			spin_unlock(&server->cb_lock);
-			spin_lock(&vnode->lock);
+	/* we're only interested in dealing with a broken callback on *this*
+	 * vnode and only if no-one else has dealt with it yet */
+	if (!mutex_trylock(&vnode->cb_broken_lock))
+		return; /* someone else is dealing with it */
 
-			if (vnode->cb_server == server) {
-				vnode->cb_server = NULL;
-				afs_kafstimod_del_timer(&vnode->cb_timeout);
-				spin_lock(&afs_cb_hash_lock);
-				list_del_init(&vnode->cb_hash_link);
-				spin_unlock(&afs_cb_hash_lock);
-				release = 1;
+	if (test_bit(AFS_VNODE_CB_BROKEN, &vnode->flags)) {
+		if (afs_vnode_fetch_status(vnode) < 0)
+			goto out;
+
+		if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
+			goto out;
+
+		/* if the vnode's data version number changed then its contents
+		 * are different */
+		if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags)) {
+			_debug("zap data");
+			invalidate_remote_inode(&vnode->vfs_inode);
+		}
 	}
 
-			spin_unlock(&vnode->lock);
+out:
+	mutex_unlock(&vnode->cb_broken_lock);
 
-			iput(inode);
-			afs_put_server(server);
+	/* avoid the potential race whereby the mutex_trylock() in this
+	 * function happens again between the clear_bit() and the
+	 * mutex_unlock() */
+	if (test_bit(AFS_VNODE_CB_BROKEN, &vnode->flags)) {
+		_debug("requeue");
+		queue_work(afs_callback_update_worker, &vnode->cb_broken_work);
+	}
+	_leave("");
+}
+
+/*
+ * actually break a callback
+ */
+static void afs_break_callback(struct afs_server *server,
+			       struct afs_vnode *vnode)
+{
+	_enter("");
+
+	set_bit(AFS_VNODE_CB_BROKEN, &vnode->flags);
+
+	if (vnode->cb_promised) {
+		spin_lock(&vnode->lock);
+
+		_debug("break callback");
 
 		spin_lock(&server->cb_lock);
+		if (vnode->cb_promised) {
+			rb_erase(&vnode->cb_promise, &server->cb_promises);
+			vnode->cb_promised = false;
 		}
+		spin_unlock(&server->cb_lock);
+
+		queue_work(afs_callback_update_worker, &vnode->cb_broken_work);
+		spin_unlock(&vnode->lock);
 	}
+}
 
-	spin_unlock(&server->cb_lock);
+/*
+ * allow the fileserver to explicitly break one callback
+ * - happens when
+ *   - the backing file is changed
+ *   - a lock is released
+ */
+static void afs_break_one_callback(struct afs_server *server,
+				   struct afs_fid *fid)
+{
+	struct afs_vnode *vnode;
+	struct rb_node *p;
 
-	_leave(" = 0");
-	return 0;
+	_debug("find");
+	spin_lock(&server->fs_lock);
+	p = server->fs_vnodes.rb_node;
+	while (p) {
+		vnode = rb_entry(p, struct afs_vnode, server_rb);
+		if (fid->vid < vnode->fid.vid)
+			p = p->rb_left;
+		else if (fid->vid > vnode->fid.vid)
+			p = p->rb_right;
+		else if (fid->vnode < vnode->fid.vnode)
+			p = p->rb_left;
+		else if (fid->vnode > vnode->fid.vnode)
+			p = p->rb_right;
+		else if (fid->unique < vnode->fid.unique)
+			p = p->rb_left;
+		else if (fid->unique > vnode->fid.unique)
+			p = p->rb_right;
+		else
+			goto found;
+	}
+
+	/* not found so we just ignore it (it may have moved to another
+	 * server) */
+not_available:
+	_debug("not avail");
+	spin_unlock(&server->fs_lock);
+	_leave("");
+	return;
+
+found:
+	_debug("found");
+	ASSERTCMP(server, ==, vnode->server);
+
+	if (!igrab(AFS_VNODE_TO_I(vnode)))
+		goto not_available;
+	spin_unlock(&server->fs_lock);
+
+	afs_break_callback(server, vnode);
+	iput(&vnode->vfs_inode);
+	_leave("");
 }
 
 /*
  * allow the fileserver to break callback promises
  */
-int SRXAFSCM_CallBack(struct afs_server *server, size_t count,
+void afs_break_callbacks(struct afs_server *server, size_t count,
 			 struct afs_callback callbacks[])
 {
-	_enter("%p,%u,", server, count);
+	_enter("%p,%zu,", server, count);
 
-	for (; count > 0; callbacks++, count--) {
-		struct afs_vnode *vnode = NULL;
-		struct inode *inode = NULL;
-		int valid = 0;
+	ASSERT(server != NULL);
+	ASSERTCMP(count, <=, AFSCBMAX);
 
+	for (; count > 0; callbacks++, count--) {
 		_debug("- Fid { vl=%08x n=%u u=%u }  CB { v=%u x=%u t=%u }",
 		       callbacks->fid.vid,
 		       callbacks->fid.vnode,
@@ -101,66 +200,244 @@ int SRXAFSCM_CallBack(struct afs_server *server, size_t count,
 		       callbacks->expiry,
 		       callbacks->type
 		       );
+		afs_break_one_callback(server, &callbacks->fid);
+	}
 
-		/* find the inode for this fid */
-		spin_lock(&afs_cb_hash_lock);
+	_leave("");
+	return;
+}
 
-		list_for_each_entry(vnode,
-				    &afs_cb_hash(server, &callbacks->fid),
-				    cb_hash_link) {
-			if (memcmp(&vnode->fid, &callbacks->fid,
-				   sizeof(struct afs_fid)) != 0)
-				continue;
+/*
+ * record the callback for breaking
+ * - the caller must hold server->cb_lock
+ */
+static void afs_do_give_up_callback(struct afs_server *server,
+				    struct afs_vnode *vnode)
+{
+	struct afs_callback *cb;
+
+	_enter("%p,%p", server, vnode);
 
-			/* right vnode, but is it same server? */
-			if (vnode->cb_server != server)
-				break; /* no */
+	cb = &server->cb_break[server->cb_break_head];
+	cb->fid		= vnode->fid;
+	cb->version	= vnode->cb_version;
+	cb->expiry	= vnode->cb_expiry;
+	cb->type	= vnode->cb_type;
+	smp_wmb();
+	server->cb_break_head =
+		(server->cb_break_head + 1) &
+		(ARRAY_SIZE(server->cb_break) - 1);
 
-			/* try and nail the inode down */
-			inode = igrab(AFS_VNODE_TO_I(vnode));
+	/* defer the breaking of callbacks to try and collect as many as
+	 * possible to ship in one operation */
+	switch (atomic_inc_return(&server->cb_break_n)) {
+	case 1 ... AFSCBMAX - 1:
+		queue_delayed_work(afs_callback_update_worker,
+				   &server->cb_break_work, HZ * 2);
+		break;
+	case AFSCBMAX:
+		afs_flush_callback_breaks(server);
+		break;
+	default:
 		break;
 	}
 
-		spin_unlock(&afs_cb_hash_lock);
+	ASSERT(server->cb_promises.rb_node != NULL);
+	rb_erase(&vnode->cb_promise, &server->cb_promises);
+	vnode->cb_promised = false;
+	_leave("");
+}
 
-		if (inode) {
-			/* we've found the record for this vnode */
-			spin_lock(&vnode->lock);
-			if (vnode->cb_server == server) {
-				/* the callback _is_ on the calling server */
-				vnode->cb_server = NULL;
-				valid = 1;
+/*
+ * give up the callback registered for a vnode on the file server when the
+ * inode is being cleared
+ */
+void afs_give_up_callback(struct afs_vnode *vnode)
+{
+	struct afs_server *server = vnode->server;
+
+	DECLARE_WAITQUEUE(myself, current);
 
-				afs_kafstimod_del_timer(&vnode->cb_timeout);
-				vnode->flags |= AFS_VNODE_CHANGED;
+	_enter("%d", vnode->cb_promised);
 
+	_debug("GIVE UP INODE %p", &vnode->vfs_inode);
+
+	if (!vnode->cb_promised) {
+		_leave(" [not promised]");
+		return;
+	}
+
+	ASSERT(server != NULL);
+
+	spin_lock(&server->cb_lock);
+	if (vnode->cb_promised && afs_breakring_space(server) == 0) {
+		add_wait_queue(&server->cb_break_waitq, &myself);
+		for (;;) {
+			set_current_state(TASK_UNINTERRUPTIBLE);
+			if (!vnode->cb_promised ||
+			    afs_breakring_space(server) != 0)
+				break;
+			spin_unlock(&server->cb_lock);
+			schedule();
+			spin_lock(&server->cb_lock);
+		}
+		remove_wait_queue(&server->cb_break_waitq, &myself);
+		__set_current_state(TASK_RUNNING);
+	}
+
+	/* of course, it's always possible for the server to break this vnode's
+	 * callback first... */
+	if (vnode->cb_promised)
+		afs_do_give_up_callback(server, vnode);
+
+	spin_unlock(&server->cb_lock);
+	_leave("");
+}
+
+/*
+ * dispatch a deferred give up callbacks operation
+ */
+void afs_dispatch_give_up_callbacks(struct work_struct *work)
+{
+	struct afs_server *server =
+		container_of(work, struct afs_server, cb_break_work.work);
+
+	_enter("");
+
+	/* tell the fileserver to discard the callback promises it has
+	 * - in the event of ENOMEM or some other error, we just forget that we
+	 *   had callbacks entirely, and the server will call us later to break
+	 *   them
+	 */
+	afs_fs_give_up_callbacks(server, &afs_async_call);
+}
+
+/*
+ * flush the outstanding callback breaks on a server
+ */
+void afs_flush_callback_breaks(struct afs_server *server)
+{
+	cancel_delayed_work(&server->cb_break_work);
+	queue_delayed_work(afs_callback_update_worker,
+			   &server->cb_break_work, 0);
+}
+
+#if 0
+/*
+ * update a bunch of callbacks
+ */
+static void afs_callback_updater(struct work_struct *work)
+{
+	struct afs_server *server;
+	struct afs_vnode *vnode, *xvnode;
+	time_t now;
+	long timeout;
+	int ret;
+
+	server = container_of(work, struct afs_server, updater);
+
+	_enter("");
+
+	now = get_seconds();
+
+	/* find the first vnode to update */
 	spin_lock(&server->cb_lock);
-				list_del_init(&vnode->cb_link);
+	for (;;) {
+		if (RB_EMPTY_ROOT(&server->cb_promises)) {
 			spin_unlock(&server->cb_lock);
+			_leave(" [nothing]");
+			return;
+		}
 
-				spin_lock(&afs_cb_hash_lock);
-				list_del_init(&vnode->cb_hash_link);
-				spin_unlock(&afs_cb_hash_lock);
+		vnode = rb_entry(rb_first(&server->cb_promises),
+				 struct afs_vnode, cb_promise);
+		if (atomic_read(&vnode->usage) > 0)
+			break;
+		rb_erase(&vnode->cb_promise, &server->cb_promises);
+		vnode->cb_promised = false;
 	}
-			spin_unlock(&vnode->lock);
 
-			if (valid) {
-				invalidate_remote_inode(inode);
-				afs_put_server(server);
+	timeout = vnode->update_at - now;
+	if (timeout > 0) {
+		queue_delayed_work(afs_vnode_update_worker,
+				   &afs_vnode_update, timeout * HZ);
+		spin_unlock(&server->cb_lock);
+		_leave(" [nothing]");
+		return;
 	}
-			iput(inode);
+
+	list_del_init(&vnode->update);
+	atomic_inc(&vnode->usage);
+	spin_unlock(&server->cb_lock);
+
+	/* we can now perform the update */
+	_debug("update %s", vnode->vldb.name);
+	vnode->state = AFS_VL_UPDATING;
+	vnode->upd_rej_cnt = 0;
+	vnode->upd_busy_cnt = 0;
+
+	ret = afs_vnode_update_record(vl, &vldb);
+	switch (ret) {
+	case 0:
+		afs_vnode_apply_update(vl, &vldb);
+		vnode->state = AFS_VL_UPDATING;
+		break;
+	case -ENOMEDIUM:
+		vnode->state = AFS_VL_VOLUME_DELETED;
+		break;
+	default:
+		vnode->state = AFS_VL_UNCERTAIN;
+		break;
 	}
+
+	/* and then reschedule */
+	_debug("reschedule");
+	vnode->update_at = get_seconds() + afs_vnode_update_timeout;
+
+	spin_lock(&server->cb_lock);
+
+	if (!list_empty(&server->cb_promises)) {
+		/* next update in 10 minutes, but wait at least 1 second more
+		 * than the newest record already queued so that we don't spam
+		 * the VL server suddenly with lots of requests
+		 */
+		xvnode = list_entry(server->cb_promises.prev,
+				    struct afs_vnode, update);
+		if (vnode->update_at <= xvnode->update_at)
+			vnode->update_at = xvnode->update_at + 1;
+		xvnode = list_entry(server->cb_promises.next,
+				    struct afs_vnode, update);
+		timeout = xvnode->update_at - now;
+		if (timeout < 0)
+			timeout = 0;
+	} else {
+		timeout = afs_vnode_update_timeout;
 	}
 
-	_leave(" = 0");
-	return 0;
+	list_add_tail(&vnode->update, &server->cb_promises);
+
+	_debug("timeout %ld", timeout);
+	queue_delayed_work(afs_vnode_update_worker,
+			   &afs_vnode_update, timeout * HZ);
+	spin_unlock(&server->cb_lock);
+	afs_put_vnode(vl);
+}
+#endif
+
+/*
+ * initialise the callback update process
+ */
+int __init afs_callback_update_init(void)
+{
+	afs_callback_update_worker =
+		create_singlethread_workqueue("kafs_callbackd");
+	return afs_callback_update_worker ? 0 : -ENOMEM;
 }
 
 /*
- * allow the fileserver to see if the cache manager is still alive
+ * shut down the callback update process
  */
-int SRXAFSCM_Probe(struct afs_server *server)
+void __exit afs_callback_update_kill(void)
 {
-	_debug("SRXAFSCM_Probe(%p)\n", server);
-	return 0;
+	destroy_workqueue(afs_callback_update_worker);
 }

fs/afs/cell.c

@@ -11,15 +11,6 @@
 
 #include <linux/module.h>
 #include <linux/slab.h>
-#include <rxrpc/peer.h>
-#include <rxrpc/connection.h>
-#include "volume.h"
-#include "cell.h"
-#include "server.h"
-#include "transport.h"
-#include "vlclient.h"
-#include "kafstimod.h"
-#include "super.h"
 #include "internal.h"
 
 DECLARE_RWSEM(afs_proc_cells_sem);
@@ -28,34 +19,21 @@ LIST_HEAD(afs_proc_cells);
 static struct list_head afs_cells = LIST_HEAD_INIT(afs_cells);
 static DEFINE_RWLOCK(afs_cells_lock);
 static DECLARE_RWSEM(afs_cells_sem); /* add/remove serialisation */
+static DECLARE_WAIT_QUEUE_HEAD(afs_cells_freeable_wq);
 static struct afs_cell *afs_cell_root;
 
-#ifdef AFS_CACHING_SUPPORT
-static cachefs_match_val_t afs_cell_cache_match(void *target,
-						const void *entry);
-static void afs_cell_cache_update(void *source, void *entry);
-
-struct cachefs_index_def afs_cache_cell_index_def = {
-	.name			= "cell_ix",
-	.data_size		= sizeof(struct afs_cache_cell),
-	.keys[0]		= { CACHEFS_INDEX_KEYS_ASCIIZ, 64 },
-	.match			= afs_cell_cache_match,
-	.update			= afs_cell_cache_update,
-};
-#endif
-
 /*
  * create a cell record
  * - "name" is the name of the cell
  * - "vllist" is a colon separated list of IP addresses in "a.b.c.d" format
  */
-int afs_cell_create(const char *name, char *vllist, struct afs_cell **_cell)
+struct afs_cell *afs_cell_create(const char *name, char *vllist)
 {
 	struct afs_cell *cell;
 	char *next;
 	int ret;
 
-	_enter("%s", name);
+	_enter("%s,%s", name, vllist);
 
 	BUG_ON(!name); /* TODO: want to look up "this cell" in the cache */
 
@@ -63,27 +41,24 @@ int afs_cell_create(const char *name, char *vllist, struct afs_cell **_cell)
 	cell = kmalloc(sizeof(struct afs_cell) + strlen(name) + 1, GFP_KERNEL);
 	if (!cell) {
 		_leave(" = -ENOMEM");
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 	}
 
 	down_write(&afs_cells_sem);
 
 	memset(cell, 0, sizeof(struct afs_cell));
-	atomic_set(&cell->usage, 0);
+	atomic_set(&cell->usage, 1);
 
 	INIT_LIST_HEAD(&cell->link);
 
-	rwlock_init(&cell->sv_lock);
-	INIT_LIST_HEAD(&cell->sv_list);
-	INIT_LIST_HEAD(&cell->sv_graveyard);
-	spin_lock_init(&cell->sv_gylock);
+	rwlock_init(&cell->servers_lock);
+	INIT_LIST_HEAD(&cell->servers);
 
 	init_rwsem(&cell->vl_sem);
 	INIT_LIST_HEAD(&cell->vl_list);
-	INIT_LIST_HEAD(&cell->vl_graveyard);
-	spin_lock_init(&cell->vl_gylock);
+	spin_lock_init(&cell->vl_lock);
 
-	strcpy(cell->name,name);
+	strcpy(cell->name, name);
 
 	/* fill in the VL server list from the rest of the string */
 	ret = -EINVAL;
@@ -106,9 +81,9 @@ int afs_cell_create(const char *name, char *vllist, struct afs_cell **_cell)
 		if (cell->vl_naddrs >= AFS_CELL_MAX_ADDRS)
 			break;
 
-	} while(vllist = next, vllist);
+	} while ((vllist = next));
 
-	/* add a proc dir for this cell */
+	/* add a proc directory for this cell */
 	ret = afs_proc_cell_setup(cell);
 	if (ret < 0)
 		goto error;
@@ -129,30 +104,29 @@ int afs_cell_create(const char *name, char *vllist, struct afs_cell **_cell)
 	down_write(&afs_proc_cells_sem);
 	list_add_tail(&cell->proc_link, &afs_proc_cells);
 	up_write(&afs_proc_cells_sem);
-
-	*_cell = cell;
 	up_write(&afs_cells_sem);
 
-	_leave(" = 0 (%p)", cell);
-	return 0;
+	_leave(" = %p", cell);
+	return cell;
 
 badaddr:
-	printk(KERN_ERR "kAFS: bad VL server IP address: '%s'\n", vllist);
+	printk(KERN_ERR "kAFS: bad VL server IP address\n");
 error:
 	up_write(&afs_cells_sem);
 	kfree(cell);
 	_leave(" = %d", ret);
-	return ret;
+	return ERR_PTR(ret);
 }
 
 /*
- * initialise the cell database from module parameters
+ * set the root cell information
+ * - can be called with a module parameter string
+ * - can be called from a write to /proc/fs/afs/rootcell
  */
 int afs_cell_init(char *rootcell)
 {
 	struct afs_cell *old_root, *new_root;
 	char *cp;
-	int ret;
 
 	_enter("");
 
@@ -160,79 +134,60 @@ int afs_cell_init(char *rootcell)
 		/* module is loaded with no parameters, or built statically.
 		 * - in the future we might initialize cell DB here.
 		 */
-		_leave(" = 0 (but no root)");
+		_leave(" = 0 [no root]");
 		return 0;
 	}
 
 	cp = strchr(rootcell, ':');
 	if (!cp) {
 		printk(KERN_ERR "kAFS: no VL server IP addresses specified\n");
-		_leave(" = %d (no colon)", -EINVAL);
+		_leave(" = -EINVAL");
 		return -EINVAL;
 	}
 
 	/* allocate a cell record for the root cell */
 	*cp++ = 0;
-	ret = afs_cell_create(rootcell, cp, &new_root);
-	if (ret < 0) {
-		_leave(" = %d", ret);
-		return ret;
+	new_root = afs_cell_create(rootcell, cp);
+	if (IS_ERR(new_root)) {
+		_leave(" = %ld", PTR_ERR(new_root));
+		return PTR_ERR(new_root);
 	}
 
-	/* as afs_put_cell() takes locks by itself, we have to do
-	 * a little gymnastics to be race-free.
-	 */
-	afs_get_cell(new_root);
-
+	/* install the new cell */
 	write_lock(&afs_cells_lock);
-	while (afs_cell_root) {
 	old_root = afs_cell_root;
-		afs_cell_root = NULL;
-		write_unlock(&afs_cells_lock);
-		afs_put_cell(old_root);
-		write_lock(&afs_cells_lock);
-	}
 	afs_cell_root = new_root;
 	write_unlock(&afs_cells_lock);
+	afs_put_cell(old_root);
 
-	_leave(" = %d", ret);
-	return ret;
+	_leave(" = 0");
+	return 0;
 }
 
 /*
  * lookup a cell record
  */
-int afs_cell_lookup(const char *name, unsigned namesz, struct afs_cell **_cell)
+struct afs_cell *afs_cell_lookup(const char *name, unsigned namesz)
 {
 	struct afs_cell *cell;
-	int ret;
 
 	_enter("\"%*.*s\",", namesz, namesz, name ? name : "");
 
-	*_cell = NULL;
+	down_read(&afs_cells_sem);
+	read_lock(&afs_cells_lock);
 
 	if (name) {
 		/* if the cell was named, look for it in the cell record list */
-		ret = -ENOENT;
-		cell = NULL;
-		read_lock(&afs_cells_lock);
-
 		list_for_each_entry(cell, &afs_cells, link) {
 			if (strncmp(cell->name, name, namesz) == 0) {
 				afs_get_cell(cell);
 				goto found;
 			}
 		}
-		cell = NULL;
+		cell = ERR_PTR(-ENOENT);
 	found:
-
-		read_unlock(&afs_cells_lock);
-
-		if (cell)
-			ret = 0;
+		;
 	} else {
-		read_lock(&afs_cells_lock);
-
 		cell = afs_cell_root;
 		if (!cell) {
 			/* this should not happen unless user tries to mount
@@ -241,37 +196,32 @@ int afs_cell_lookup(const char *name, unsigned namesz, struct afs_cell **_cell)
 			 * ENOENT might be "more appropriate" but they happen
 			 * for other reasons.
 			 */
-			ret = -EDESTADDRREQ;
+			cell = ERR_PTR(-EDESTADDRREQ);
 		} else {
 			afs_get_cell(cell);
-			ret = 0;
 		}
 
-		read_unlock(&afs_cells_lock);
 	}
 
-	*_cell = cell;
-	_leave(" = %d (%p)", ret, cell);
-	return ret;
+	read_unlock(&afs_cells_lock);
+	up_read(&afs_cells_sem);
+	_leave(" = %p", cell);
+	return cell;
 }
 
 /*
  * try and get a cell record
  */
-struct afs_cell *afs_get_cell_maybe(struct afs_cell **_cell)
+struct afs_cell *afs_get_cell_maybe(struct afs_cell *cell)
 {
-	struct afs_cell *cell;
-
 	write_lock(&afs_cells_lock);
 
-	cell = *_cell;
 	if (cell && !list_empty(&cell->link))
 		afs_get_cell(cell);
 	else
 		cell = NULL;
 
 	write_unlock(&afs_cells_lock);
-
 	return cell;
 }
 
@@ -285,8 +235,7 @@ void afs_put_cell(struct afs_cell *cell)
 
 	_enter("%p{%d,%s}", cell, atomic_read(&cell->usage), cell->name);
 
-	/* sanity check */
-	BUG_ON(atomic_read(&cell->usage) <= 0);
+	ASSERTCMP(atomic_read(&cell->usage), >, 0);
 
 	/* to prevent a race, the decrement and the dequeue must be effectively
 	 * atomic */
@@ -298,35 +247,49 @@ void afs_put_cell(struct afs_cell *cell)
 		return;
 	}
 
+	ASSERT(list_empty(&cell->servers));
+	ASSERT(list_empty(&cell->vl_list));
+
 	write_unlock(&afs_cells_lock);
 
-	BUG_ON(!list_empty(&cell->sv_list));
-	BUG_ON(!list_empty(&cell->sv_graveyard));
-	BUG_ON(!list_empty(&cell->vl_list));
-	BUG_ON(!list_empty(&cell->vl_graveyard));
+	wake_up(&afs_cells_freeable_wq);
 
 	_leave(" [unused]");
 }
 
 /*
  * destroy a cell record
+ * - must be called with the afs_cells_sem write-locked
+ * - cell->link should have been broken by the caller
  */
 static void afs_cell_destroy(struct afs_cell *cell)
 {
 	_enter("%p{%d,%s}", cell, atomic_read(&cell->usage), cell->name);
 
-	/* to prevent a race, the decrement and the dequeue must be effectively
-	 * atomic */
-	write_lock(&afs_cells_lock);
+	ASSERTCMP(atomic_read(&cell->usage), >=, 0);
+	ASSERT(list_empty(&cell->link));
 
-	/* sanity check */
-	BUG_ON(atomic_read(&cell->usage) != 0);
+	/* wait for everyone to stop using the cell */
+	if (atomic_read(&cell->usage) > 0) {
+		DECLARE_WAITQUEUE(myself, current);
 
-	list_del_init(&cell->link);
+		_debug("wait for cell %s", cell->name);
+		set_current_state(TASK_UNINTERRUPTIBLE);
+		add_wait_queue(&afs_cells_freeable_wq, &myself);
 
-	write_unlock(&afs_cells_lock);
+		while (atomic_read(&cell->usage) > 0) {
+			schedule();
+			set_current_state(TASK_UNINTERRUPTIBLE);
+		}
 
-	down_write(&afs_cells_sem);
+		remove_wait_queue(&afs_cells_freeable_wq, &myself);
+		set_current_state(TASK_RUNNING);
+	}
+
+	_debug("cell dead");
+	ASSERTCMP(atomic_read(&cell->usage), ==, 0);
+	ASSERT(list_empty(&cell->servers));
+	ASSERT(list_empty(&cell->vl_list));
 
 	afs_proc_cell_remove(cell);
 
@@ -338,101 +301,25 @@ static void afs_cell_destroy(struct afs_cell *cell)
 	cachefs_relinquish_cookie(cell->cache, 0);
 #endif
 
-	up_write(&afs_cells_sem);
-
-	BUG_ON(!list_empty(&cell->sv_list));
-	BUG_ON(!list_empty(&cell->sv_graveyard));
-	BUG_ON(!list_empty(&cell->vl_list));
-	BUG_ON(!list_empty(&cell->vl_graveyard));
-
-	/* finish cleaning up the cell */
 	kfree(cell);
 
 	_leave(" [destroyed]");
 }
 
-/*
- * lookup the server record corresponding to an Rx RPC peer
- */
-int afs_server_find_by_peer(const struct rxrpc_peer *peer,
-			    struct afs_server **_server)
-{
-	struct afs_server *server;
-	struct afs_cell *cell;
-
-	_enter("%p{a=%08x},", peer, ntohl(peer->addr.s_addr));
-
-	/* search the cell list */
-	read_lock(&afs_cells_lock);
-
-	list_for_each_entry(cell, &afs_cells, link) {
-
-		_debug("? cell %s",cell->name);
-
-		write_lock(&cell->sv_lock);
-
-		/* check the active list */
-		list_for_each_entry(server, &cell->sv_list, link) {
-			_debug("?? server %08x", ntohl(server->addr.s_addr));
-
-			if (memcmp(&server->addr, &peer->addr,
-				   sizeof(struct in_addr)) == 0)
-				goto found_server;
-		}
-
-		/* check the inactive list */
-		spin_lock(&cell->sv_gylock);
-		list_for_each_entry(server, &cell->sv_graveyard, link) {
-			_debug("?? dead server %08x",
-			       ntohl(server->addr.s_addr));
-
-			if (memcmp(&server->addr, &peer->addr,
-				   sizeof(struct in_addr)) == 0)
-				goto found_dead_server;
-		}
-		spin_unlock(&cell->sv_gylock);
-
-		write_unlock(&cell->sv_lock);
-	}
-	read_unlock(&afs_cells_lock);
-
-	_leave(" = -ENOENT");
-	return -ENOENT;
-
-	/* we found it in the graveyard - resurrect it */
-found_dead_server:
-	list_move_tail(&server->link, &cell->sv_list);
-	afs_get_server(server);
-	afs_kafstimod_del_timer(&server->timeout);
-	spin_unlock(&cell->sv_gylock);
-	goto success;
-
-	/* we found it - increment its ref count and return it */
-found_server:
-	afs_get_server(server);
-
-success:
-	write_unlock(&cell->sv_lock);
-	read_unlock(&afs_cells_lock);
-
-	*_server = server;
-	_leave(" = 0 (s=%p c=%p)", server, cell);
-	return 0;
-}
-
 /*
  * purge in-memory cell database on module unload or afs_init() failure
  * - the timeout daemon is stopped before calling this
  */
 void afs_cell_purge(void)
 {
-	struct afs_vlocation *vlocation;
 	struct afs_cell *cell;
 
 	_enter("");
 
 	afs_put_cell(afs_cell_root);
 
+	down_write(&afs_cells_sem);
+
 	while (!list_empty(&afs_cells)) {
 		cell = NULL;
 
@@ -451,102 +338,11 @@ void afs_cell_purge(void)
 			_debug("PURGING CELL %s (%d)",
 			       cell->name, atomic_read(&cell->usage));
 
-			BUG_ON(!list_empty(&cell->sv_list));
-			BUG_ON(!list_empty(&cell->vl_list));
-
-			/* purge the cell's VL graveyard list */
-			_debug(" - clearing VL graveyard");
-
-			spin_lock(&cell->vl_gylock);
-
-			while (!list_empty(&cell->vl_graveyard)) {
-				vlocation = list_entry(cell->vl_graveyard.next,
-						       struct afs_vlocation,
-						       link);
-				list_del_init(&vlocation->link);
-
-				afs_kafstimod_del_timer(&vlocation->timeout);
-
-				spin_unlock(&cell->vl_gylock);
-
-				afs_vlocation_do_timeout(vlocation);
-				/* TODO: race if move to use krxtimod instead
-				 * of kafstimod */
-
-				spin_lock(&cell->vl_gylock);
-			}
-
-			spin_unlock(&cell->vl_gylock);
-
-			/* purge the cell's server graveyard list */
-			_debug(" - clearing server graveyard");
-
-			spin_lock(&cell->sv_gylock);
-
-			while (!list_empty(&cell->sv_graveyard)) {
-				struct afs_server *server;
-
-				server = list_entry(cell->sv_graveyard.next,
-						    struct afs_server, link);
-				list_del_init(&server->link);
-
-				afs_kafstimod_del_timer(&server->timeout);
-
-				spin_unlock(&cell->sv_gylock);
-
-				afs_server_do_timeout(server);
-
-				spin_lock(&cell->sv_gylock);
-			}
-
-			spin_unlock(&cell->sv_gylock);
-
 			/* now the cell should be left with no references */
 			afs_cell_destroy(cell);
 		}
 	}
 
+	up_write(&afs_cells_sem);
 	_leave("");
 }
-
-/*
- * match a cell record obtained from the cache
- */
-#ifdef AFS_CACHING_SUPPORT
-static cachefs_match_val_t afs_cell_cache_match(void *target,
-						const void *entry)
-{
-	const struct afs_cache_cell *ccell = entry;
-	struct afs_cell *cell = target;
-
-	_enter("{%s},{%s}", ccell->name, cell->name);
-
-	if (strncmp(ccell->name, cell->name, sizeof(ccell->name)) == 0) {
-		_leave(" = SUCCESS");
-		return CACHEFS_MATCH_SUCCESS;
-	}
-
-	_leave(" = FAILED");
-	return CACHEFS_MATCH_FAILED;
-}
-#endif
-
-/*
- * update a cell record in the cache
- */
-#ifdef AFS_CACHING_SUPPORT
-static void afs_cell_cache_update(void *source, void *entry)
-{
-	struct afs_cache_cell *ccell = entry;
-	struct afs_cell *cell = source;
-
-	_enter("%p,%p", source, entry);
-
-	strncpy(ccell->name, cell->name, sizeof(ccell->name));
-
-	memcpy(ccell->vl_servers,
-	       cell->vl_addrs,
-	       min(sizeof(ccell->vl_servers), sizeof(cell->vl_addrs)));
-
-}
-#endif

fs/afs/cell.h

-/* AFS cell record
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#ifndef AFS_CELL_H
-#define AFS_CELL_H
-
-#include "types.h"
-#include "cache.h"
-
-#define AFS_CELL_MAX_ADDRS 15
-
-extern volatile int afs_cells_being_purged; /* T when cells are being purged by rmmod */
-
-/*
- * entry in the cached cell catalogue
- */
-struct afs_cache_cell {
-	char			name[64];	/* cell name (padded with NULs) */
-	struct in_addr		vl_servers[15];	/* cached cell VL servers */
-};
-
-/*
- * AFS cell record
- */
-struct afs_cell {
-	atomic_t		usage;
-	struct list_head	link;		/* main cell list link */
-	struct list_head	proc_link;	/* /proc cell list link */
-	struct proc_dir_entry	*proc_dir;	/* /proc dir for this cell */
-#ifdef AFS_CACHING_SUPPORT
-	struct cachefs_cookie	*cache;		/* caching cookie */
-#endif
-
-	/* server record management */
-	rwlock_t		sv_lock;	/* active server list lock */
-	struct list_head	sv_list;	/* active server list */
-	struct list_head	sv_graveyard;	/* inactive server list */
-	spinlock_t		sv_gylock;	/* inactive server list lock */
-
-	/* volume location record management */
-	struct rw_semaphore	vl_sem;		/* volume management serialisation semaphore */
-	struct list_head	vl_list;	/* cell's active VL record list */
-	struct list_head	vl_graveyard;	/* cell's inactive VL record list */
-	spinlock_t		vl_gylock;	/* graveyard lock */
-	unsigned short		vl_naddrs;	/* number of VL servers in addr list */
-	unsigned short		vl_curr_svix;	/* current server index */
-	struct in_addr		vl_addrs[AFS_CELL_MAX_ADDRS];	/* cell VL server addresses */
-
-	char			name[0];	/* cell name - must go last */
-};
-
-extern int afs_cell_init(char *);
-extern int afs_cell_create(const char *, char *, struct afs_cell **);
-extern int afs_cell_lookup(const char *, unsigned, struct afs_cell **);
-
-#define afs_get_cell(C) do { atomic_inc(&(C)->usage); } while(0)
-
-extern struct afs_cell *afs_get_cell_maybe(struct afs_cell **);
-extern void afs_put_cell(struct afs_cell *);
-extern void afs_cell_purge(void);
-
-#endif /* AFS_CELL_H */

fs/afs/cmservice.h

-/* AFS Cache Manager Service declarations
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#ifndef AFS_CMSERVICE_H
-#define AFS_CMSERVICE_H
-
-#include <rxrpc/transport.h>
-#include "types.h"
-
-/* cache manager start/stop */
-extern int afscm_start(void);
-extern void afscm_stop(void);
-
-/* cache manager server functions */
-extern int SRXAFSCM_InitCallBackState(struct afs_server *);
-extern int SRXAFSCM_CallBack(struct afs_server *, size_t,
-			     struct afs_callback[]);
-extern int SRXAFSCM_Probe(struct afs_server *);
-
-#endif /* AFS_CMSERVICE_H */

fs/afs/dir.c

@@ -15,11 +15,6 @@
 #include <linux/slab.h>
 #include <linux/fs.h>
 #include <linux/pagemap.h>
-#include <linux/smp_lock.h>
-#include "vnode.h"
-#include "volume.h"
-#include <rxrpc/call.h>
-#include "super.h"
 #include "internal.h"
 
 static struct dentry *afs_dir_lookup(struct inode *dir, struct dentry *dentry,
@@ -127,9 +122,10 @@ static inline void afs_dir_check_page(struct inode *dir, struct page *page)
 	if (qty == 0)
 		goto error;
 
-	if (page->index==0 && qty!=ntohs(dbuf->blocks[0].pagehdr.npages)) {
+	if (page->index == 0 && qty != ntohs(dbuf->blocks[0].pagehdr.npages)) {
 		printk("kAFS: %s(%lu): wrong number of dir blocks %d!=%hu\n",
-		       __FUNCTION__,dir->i_ino,qty,ntohs(dbuf->blocks[0].pagehdr.npages));
+		       __FUNCTION__, dir->i_ino, qty,
+		       ntohs(dbuf->blocks[0].pagehdr.npages));
 		goto error;
 	}
 #endif
@@ -194,6 +190,7 @@ static struct page *afs_dir_get_page(struct inode *dir, unsigned long index)
 
 fail:
 	afs_dir_put_page(page);
+	_leave(" = -EIO");
 	return ERR_PTR(-EIO);
 }
 
@@ -207,7 +204,7 @@ static int afs_dir_open(struct inode *inode, struct file *file)
 	BUILD_BUG_ON(sizeof(union afs_dir_block) != 2048);
 	BUILD_BUG_ON(sizeof(union afs_dirent) != 32);
 
-	if (AFS_FS_I(inode)->flags & AFS_VNODE_DELETED)
+	if (test_bit(AFS_VNODE_DELETED, &AFS_FS_I(inode)->flags))
 		return -ENOENT;
 
 	_leave(" = 0");
@@ -242,7 +239,7 @@ static int afs_dir_iterate_block(unsigned *fpos,
 		/* skip entries marked unused in the bitmap */
 		if (!(block->pagehdr.bitmap[offset / 8] &
 		      (1 << (offset % 8)))) {
-			_debug("ENT[%Zu.%u]: unused\n",
+			_debug("ENT[%Zu.%u]: unused",
 			       blkoff / sizeof(union afs_dir_block), offset);
 			if (offset >= curr)
 				*fpos = blkoff +
@@ -256,7 +253,7 @@ static int afs_dir_iterate_block(unsigned *fpos,
 			       sizeof(*block) -
 			       offset * sizeof(union afs_dirent));
 
-		_debug("ENT[%Zu.%u]: %s %Zu \"%s\"\n",
+		_debug("ENT[%Zu.%u]: %s %Zu \"%s\"",
 		       blkoff / sizeof(union afs_dir_block), offset,
 		       (offset < curr ? "skip" : "fill"),
 		       nlen, dire->u.name);
@@ -266,7 +263,7 @@ static int afs_dir_iterate_block(unsigned *fpos,
 			if (next >= AFS_DIRENT_PER_BLOCK) {
 				_debug("ENT[%Zu.%u]:"
 				       " %u travelled beyond end dir block"
-				       " (len %u/%Zu)\n",
+				       " (len %u/%Zu)",
 				       blkoff / sizeof(union afs_dir_block),
 				       offset, next, tmp, nlen);
 				return -EIO;
@@ -274,13 +271,13 @@ static int afs_dir_iterate_block(unsigned *fpos,
 			if (!(block->pagehdr.bitmap[next / 8] &
 			      (1 << (next % 8)))) {
 				_debug("ENT[%Zu.%u]:"
-				       " %u unmarked extension (len %u/%Zu)\n",
+				       " %u unmarked extension (len %u/%Zu)",
 				       blkoff / sizeof(union afs_dir_block),
 				       offset, next, tmp, nlen);
 				return -EIO;
 			}
 
-			_debug("ENT[%Zu.%u]: ext %u/%Zu\n",
+			_debug("ENT[%Zu.%u]: ext %u/%Zu",
 			       blkoff / sizeof(union afs_dir_block),
 			       next, tmp, nlen);
 			next++;
@@ -311,7 +308,7 @@ static int afs_dir_iterate_block(unsigned *fpos,
 }
 
 /*
- * read an AFS directory
+ * iterate through the data blob that lists the contents of an AFS directory
  */
 static int afs_dir_iterate(struct inode *dir, unsigned *fpos, void *cookie,
 			   filldir_t filldir)
@@ -324,7 +321,7 @@ static int afs_dir_iterate(struct inode *dir, unsigned *fpos, void *cookie,
 
 	_enter("{%lu},%u,,", dir->i_ino, *fpos);
 
-	if (AFS_FS_I(dir)->flags & AFS_VNODE_DELETED) {
+	if (test_bit(AFS_VNODE_DELETED, &AFS_FS_I(dir)->flags)) {
 		_leave(" = -ESTALE");
 		return -ESTALE;
 	}
@@ -381,10 +378,12 @@ static int afs_dir_readdir(struct file *file, void *cookie, filldir_t filldir)
 	unsigned fpos;
 	int ret;
 
-	_enter("{%Ld,{%lu}}", file->f_pos, file->f_path.dentry->d_inode->i_ino);
+	_enter("{%Ld,{%lu}}",
+	       file->f_pos, file->f_path.dentry->d_inode->i_ino);
 
 	fpos = file->f_pos;
-	ret = afs_dir_iterate(file->f_path.dentry->d_inode, &fpos, cookie, filldir);
+	ret = afs_dir_iterate(file->f_path.dentry->d_inode, &fpos,
+			      cookie, filldir);
 	file->f_pos = fpos;
 
 	_leave(" = %d", ret);
@@ -401,9 +400,13 @@ static int afs_dir_lookup_filldir(void *_cookie, const char *name, int nlen,
 {
 	struct afs_dir_lookup_cookie *cookie = _cookie;
 
-	_enter("{%s,%Zu},%s,%u,,%lu,%u",
+	_enter("{%s,%Zu},%s,%u,,%llu,%u",
 	       cookie->name, cookie->nlen, name, nlen, ino, dtype);
 
+	/* insanity checks first */
+	BUILD_BUG_ON(sizeof(union afs_dir_block) != 2048);
+	BUILD_BUG_ON(sizeof(union afs_dirent) != 32);
+
 	if (cookie->nlen != nlen || memcmp(cookie->name, name, nlen) != 0) {
 		_leave(" = 0 [no]");
 		return 0;
@@ -418,34 +421,17 @@ static int afs_dir_lookup_filldir(void *_cookie, const char *name, int nlen,
 }
 
 /*
- * look up an entry in a directory
+ * do a lookup in a directory
  */
-static struct dentry *afs_dir_lookup(struct inode *dir, struct dentry *dentry,
-				     struct nameidata *nd)
+static int afs_do_lookup(struct inode *dir, struct dentry *dentry,
+			 struct afs_fid *fid)
 {
 	struct afs_dir_lookup_cookie cookie;
 	struct afs_super_info *as;
-	struct afs_vnode *vnode;
-	struct inode *inode;
 	unsigned fpos;
 	int ret;
 
-	_enter("{%lu},%p{%s}", dir->i_ino, dentry, dentry->d_name.name);
-
-	/* insanity checks first */
-	BUILD_BUG_ON(sizeof(union afs_dir_block) != 2048);
-	BUILD_BUG_ON(sizeof(union afs_dirent) != 32);
-
-	if (dentry->d_name.len > 255) {
-		_leave(" = -ENAMETOOLONG");
-		return ERR_PTR(-ENAMETOOLONG);
-	}
-
-	vnode = AFS_FS_I(dir);
-	if (vnode->flags & AFS_VNODE_DELETED) {
-		_leave(" = -ESTALE");
-		return ERR_PTR(-ESTALE);
-	}
+	_enter("{%lu},%p{%s},", dir->i_ino, dentry, dentry->d_name.name);
 
 	as = dir->i_sb->s_fs_info;
 
@@ -458,54 +444,130 @@ static struct dentry *afs_dir_lookup(struct inode *dir, struct dentry *dentry,
 	fpos = 0;
 	ret = afs_dir_iterate(dir, &fpos, &cookie, afs_dir_lookup_filldir);
 	if (ret < 0) {
-		_leave(" = %d", ret);
-		return ERR_PTR(ret);
+		_leave(" = %d [iter]", ret);
+		return ret;
 	}
 
 	ret = -ENOENT;
 	if (!cookie.found) {
-		_leave(" = %d", ret);
-		return ERR_PTR(ret);
+		_leave(" = -ENOENT [not found]");
+		return -ENOENT;
 	}
 
-	/* instantiate the dentry */
-	ret = afs_iget(dir->i_sb, &cookie.fid, &inode);
+	*fid = cookie.fid;
+	_leave(" = 0 { vn=%u u=%u }", fid->vnode, fid->unique);
+	return 0;
+}
+
+/*
+ * look up an entry in a directory
+ */
+static struct dentry *afs_dir_lookup(struct inode *dir, struct dentry *dentry,
+				     struct nameidata *nd)
+{
+	struct afs_vnode *vnode;
+	struct afs_fid fid;
+	struct inode *inode;
+	int ret;
+
+	_enter("{%lu},%p{%s}", dir->i_ino, dentry, dentry->d_name.name);
+
+	if (dentry->d_name.len > 255) {
+		_leave(" = -ENAMETOOLONG");
+		return ERR_PTR(-ENAMETOOLONG);
+	}
+
+	vnode = AFS_FS_I(dir);
+	if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
+		_leave(" = -ESTALE");
+		return ERR_PTR(-ESTALE);
+	}
+
+	ret = afs_do_lookup(dir, dentry, &fid);
 	if (ret < 0) {
-		_leave(" = %d", ret);
+		_leave(" = %d [do]", ret);
 		return ERR_PTR(ret);
 	}
 
+	/* instantiate the dentry */
+	inode = afs_iget(dir->i_sb, &fid);
+	if (IS_ERR(inode)) {
+		_leave(" = %ld", PTR_ERR(inode));
+		return ERR_PTR(PTR_ERR(inode));
+	}
+
 	dentry->d_op = &afs_fs_dentry_operations;
-	dentry->d_fsdata = (void *) (unsigned long) vnode->status.version;
 
 	d_add(dentry, inode);
 	_leave(" = 0 { vn=%u u=%u } -> { ino=%lu v=%lu }",
-	       cookie.fid.vnode,
-	       cookie.fid.unique,
+	       fid.vnode,
+	       fid.unique,
 	       dentry->d_inode->i_ino,
 	       dentry->d_inode->i_version);
 
 	return NULL;
 }
 
+/*
+ * propagate changed and modified flags on a directory to all the children of
+ * that directory as they may indicate that the ACL on the dir has changed,
+ * potentially rendering the child inaccessible or that a file has been deleted
+ * or renamed
+ */
+static void afs_propagate_dir_changes(struct dentry *dir)
+{
+	struct dentry *child;
+	bool c, m;
+
+	c = test_bit(AFS_VNODE_CHANGED, &AFS_FS_I(dir->d_inode)->flags);
+	m = test_bit(AFS_VNODE_MODIFIED, &AFS_FS_I(dir->d_inode)->flags);
+
+	_enter("{%d,%d}", c, m);
+
+	spin_lock(&dir->d_lock);
+
+	list_for_each_entry(child, &dir->d_subdirs, d_u.d_child) {
+		if (child->d_inode) {
+			struct afs_vnode *vnode;
+
+			_debug("tag %s", child->d_name.name);
+			vnode = AFS_FS_I(child->d_inode);
+			if (c)
+				set_bit(AFS_VNODE_DIR_CHANGED, &vnode->flags);
+			if (m)
+				set_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags);
+		}
+	}
+
+	spin_unlock(&dir->d_lock);
+}
+
 /*
  * check that a dentry lookup hit has found a valid entry
  * - NOTE! the hit can be a negative hit too, so we can't assume we have an
  *   inode
- * (derived from nfs_lookup_revalidate)
+ * - there are several things we need to check
+ *   - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
+ *     symlink)
+ *   - parent dir metadata changed (security changes)
+ *   - dentry data changed (write, truncate)
+ *   - dentry metadata changed (security changes)
  */
 static int afs_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct afs_dir_lookup_cookie cookie;
+	struct afs_vnode *vnode;
+	struct afs_fid fid;
 	struct dentry *parent;
 	struct inode *inode, *dir;
-	unsigned fpos;
 	int ret;
 
-	_enter("{sb=%p n=%s},", dentry->d_sb, dentry->d_name.name);
+	vnode = AFS_FS_I(dentry->d_inode);
+
+	_enter("{sb=%p n=%s fl=%lx},",
+	       dentry->d_sb, dentry->d_name.name, vnode->flags);
 
 	/* lock down the parent dentry so we can peer at it */
-	parent = dget_parent(dentry->d_parent);
+	parent = dget_parent(dentry);
 
 	dir = parent->d_inode;
 	inode = dentry->d_inode;
@@ -517,81 +579,92 @@ static int afs_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 	/* handle a bad inode */
 	if (is_bad_inode(inode)) {
 		printk("kAFS: afs_d_revalidate: %s/%s has bad inode\n",
-		       dentry->d_parent->d_name.name, dentry->d_name.name);
+		       parent->d_name.name, dentry->d_name.name);
 		goto out_bad;
 	}
 
-	/* force a full look up if the parent directory changed since last the
-	 * server was consulted
-	 * - otherwise this inode must still exist, even if the inode details
-	 *   themselves have changed
-	 */
-	if (AFS_FS_I(dir)->flags & AFS_VNODE_CHANGED)
-		afs_vnode_fetch_status(AFS_FS_I(dir));
-
-	if (AFS_FS_I(dir)->flags & AFS_VNODE_DELETED) {
+	/* check that this dirent still exists if the directory's contents were
+	 * modified */
+	if (test_bit(AFS_VNODE_DELETED, &AFS_FS_I(dir)->flags)) {
 		_debug("%s: parent dir deleted", dentry->d_name.name);
 		goto out_bad;
 	}
 
-	if (AFS_FS_I(inode)->flags & AFS_VNODE_DELETED) {
-		_debug("%s: file already deleted", dentry->d_name.name);
-		goto out_bad;
-	}
-
-	if ((unsigned long) dentry->d_fsdata !=
-	    (unsigned long) AFS_FS_I(dir)->status.version) {
-		_debug("%s: parent changed %lu -> %u",
-		       dentry->d_name.name,
-		       (unsigned long) dentry->d_fsdata,
-		       (unsigned) AFS_FS_I(dir)->status.version);
+	if (test_and_clear_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags)) {
+		/* rm/rmdir/rename may have occurred */
+		_debug("dir modified");
 
 		/* search the directory for this vnode */
-		cookie.name	= dentry->d_name.name;
-		cookie.nlen	= dentry->d_name.len;
-		cookie.fid.vid	= AFS_FS_I(inode)->volume->vid;
-		cookie.found	= 0;
-
-		fpos = 0;
-		ret = afs_dir_iterate(dir, &fpos, &cookie,
-				      afs_dir_lookup_filldir);
+		ret = afs_do_lookup(dir, dentry, &fid);
+		if (ret == -ENOENT) {
+			_debug("%s: dirent not found", dentry->d_name.name);
+			goto not_found;
+		}
 		if (ret < 0) {
 			_debug("failed to iterate dir %s: %d",
 			       parent->d_name.name, ret);
 			goto out_bad;
 		}
 
-		if (!cookie.found) {
-			_debug("%s: dirent not found", dentry->d_name.name);
-			goto not_found;
-		}
-
 		/* if the vnode ID has changed, then the dirent points to a
 		 * different file */
-		if (cookie.fid.vnode != AFS_FS_I(inode)->fid.vnode) {
-			_debug("%s: dirent changed", dentry->d_name.name);
+		if (fid.vnode != vnode->fid.vnode) {
+			_debug("%s: dirent changed [%u != %u]",
+			       dentry->d_name.name, fid.vnode,
+			       vnode->fid.vnode);
 			goto not_found;
 		}
 
 		/* if the vnode ID uniqifier has changed, then the file has
 		 * been deleted */
-		if (cookie.fid.unique != AFS_FS_I(inode)->fid.unique) {
+		if (fid.unique != vnode->fid.unique) {
 			_debug("%s: file deleted (uq %u -> %u I:%lu)",
-			       dentry->d_name.name,
-			       cookie.fid.unique,
-			       AFS_FS_I(inode)->fid.unique,
-			       inode->i_version);
-			spin_lock(&AFS_FS_I(inode)->lock);
-			AFS_FS_I(inode)->flags |= AFS_VNODE_DELETED;
-			spin_unlock(&AFS_FS_I(inode)->lock);
+			       dentry->d_name.name, fid.unique,
+			       vnode->fid.unique, inode->i_version);
+			spin_lock(&vnode->lock);
+			set_bit(AFS_VNODE_DELETED, &vnode->flags);
+			spin_unlock(&vnode->lock);
 			invalidate_remote_inode(inode);
 			goto out_bad;
 		}
+	}
 
-		dentry->d_fsdata =
-			(void *) (unsigned long) AFS_FS_I(dir)->status.version;
+	/* if the directory's metadata were changed then the security may be
+	 * different and we may no longer have access */
+	mutex_lock(&vnode->cb_broken_lock);
+
+	if (test_and_clear_bit(AFS_VNODE_DIR_CHANGED, &vnode->flags) ||
+	    test_bit(AFS_VNODE_CB_BROKEN, &vnode->flags)) {
+		_debug("%s: changed", dentry->d_name.name);
+		set_bit(AFS_VNODE_CB_BROKEN, &vnode->flags);
+		if (afs_vnode_fetch_status(vnode) < 0) {
+			mutex_unlock(&vnode->cb_broken_lock);
+			goto out_bad;
+		}
 	}
 
+	if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
+		_debug("%s: file already deleted", dentry->d_name.name);
+		mutex_unlock(&vnode->cb_broken_lock);
+		goto out_bad;
+	}
+
+	/* if the vnode's data version number changed then its contents are
+	 * different */
+	if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags)) {
+		_debug("zap data");
+		invalidate_remote_inode(inode);
+	}
+
+	if (S_ISDIR(inode->i_mode) &&
+	    (test_bit(AFS_VNODE_CHANGED, &vnode->flags) ||
+	     test_bit(AFS_VNODE_MODIFIED, &vnode->flags)))
+		afs_propagate_dir_changes(dentry);
+
+	clear_bit(AFS_VNODE_CHANGED, &vnode->flags);
+	clear_bit(AFS_VNODE_MODIFIED, &vnode->flags);
+	mutex_unlock(&vnode->cb_broken_lock);
+
 out_valid:
 	dput(parent);
 	_leave(" = 1 [valid]");
@@ -610,12 +683,10 @@ static int afs_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 			goto out_valid;
 	}
 
-	shrink_dcache_parent(dentry);
-
 	_debug("dropping dentry %s/%s",
-	       dentry->d_parent->d_name.name, dentry->d_name.name);
+	       parent->d_name.name, dentry->d_name.name);
+	shrink_dcache_parent(dentry);
 	d_drop(dentry);
-
 	dput(parent);
 
 	_leave(" = 0 [bad]");
@@ -635,10 +706,9 @@ static int afs_d_delete(struct dentry *dentry)
 	if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
 		goto zap;
 
-	if (dentry->d_inode) {
-		if (AFS_FS_I(dentry->d_inode)->flags & AFS_VNODE_DELETED)
+	if (dentry->d_inode &&
+	    test_bit(AFS_VNODE_DELETED, &AFS_FS_I(dentry->d_inode)->flags))
 			goto zap;
-	}
 
 	_leave(" = 0 [keep]");
 	return 0;

fs/afs/file.c

-/* file.c: AFS filesystem file handling
+/* AFS filesystem file handling
  *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
+ * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
  * Written by David Howells (dhowells@redhat.com)
  *
  * This program is free software; you can redistribute it and/or
@@ -15,9 +15,6 @@
 #include <linux/slab.h>
 #include <linux/fs.h>
 #include <linux/pagemap.h>
-#include "volume.h"
-#include "vnode.h"
-#include <rxrpc/call.h>
 #include "internal.h"
 
 #if 0
@@ -80,12 +77,10 @@ static void afs_file_readpage_write_complete(void *cookie_data,
  */
 static int afs_file_readpage(struct file *file, struct page *page)
 {
-	struct afs_rxfs_fetch_descriptor desc;
-#ifdef AFS_CACHING_SUPPORT
-	struct cachefs_page *pageio;
-#endif
 	struct afs_vnode *vnode;
 	struct inode *inode;
+	size_t len;
+	off_t offset;
 	int ret;
 
 	inode = page->mapping->host;
@@ -97,14 +92,10 @@ static int afs_file_readpage(struct file *file, struct page *page)
 	BUG_ON(!PageLocked(page));
 
 	ret = -ESTALE;
-	if (vnode->flags & AFS_VNODE_DELETED)
+	if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
 		goto error;
 
 #ifdef AFS_CACHING_SUPPORT
-	ret = cachefs_page_get_private(page, &pageio, GFP_NOIO);
-	if (ret < 0)
-		goto error;
-
 	/* is it cached? */
 	ret = cachefs_read_or_alloc_page(vnode->cache,
 					 page,
@@ -128,26 +119,19 @@ static int afs_file_readpage(struct file *file, struct page *page)
 	case -ENOBUFS:
 	case -ENODATA:
 	default:
-		desc.fid	= vnode->fid;
-		desc.offset	= page->index << PAGE_CACHE_SHIFT;
-		desc.size	= min((size_t) (inode->i_size - desc.offset),
-				      (size_t) PAGE_SIZE);
-		desc.buffer	= kmap(page);
-
-		clear_page(desc.buffer);
+		offset = page->index << PAGE_CACHE_SHIFT;
+		len = min_t(size_t, i_size_read(inode) - offset, PAGE_SIZE);
 
 		/* read the contents of the file from the server into the
 		 * page */
-		ret = afs_vnode_fetch_data(vnode, &desc);
-		kunmap(page);
+		ret = afs_vnode_fetch_data(vnode, offset, len, page);
 		if (ret < 0) {
-			if (ret==-ENOENT) {
+			if (ret == -ENOENT) {
 				_debug("got NOENT from server"
 				       " - marking file deleted and stale");
-				vnode->flags |= AFS_VNODE_DELETED;
+				set_bit(AFS_VNODE_DELETED, &vnode->flags);
 				ret = -ESTALE;
 			}
-
 #ifdef AFS_CACHING_SUPPORT
 			cachefs_uncache_page(vnode->cache, page);
 #endif
@@ -174,10 +158,9 @@ static int afs_file_readpage(struct file *file, struct page *page)
 	_leave(" = 0");
 	return 0;
 
- error:
+error:
 	SetPageError(page);
 	unlock_page(page);
-
 	_leave(" = %d", ret);
 	return ret;
 }

fs/afs/fsclient.h

-/* AFS File Server client stub declarations
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#ifndef AFS_FSCLIENT_H
-#define AFS_FSCLIENT_H
-
-#include "server.h"
-
-extern int afs_rxfs_get_volume_info(struct afs_server *,
-				    const char *,
-				    struct afs_volume_info *);
-
-extern int afs_rxfs_fetch_file_status(struct afs_server *,
-				      struct afs_vnode *,
-				      struct afs_volsync *);
-
-struct afs_rxfs_fetch_descriptor {
-	struct afs_fid	fid;		/* file ID to fetch */
-	size_t		size;		/* total number of bytes to fetch */
-	off_t		offset;		/* offset in file to start from */
-	void		*buffer;	/* read buffer */
-	size_t		actual;		/* actual size sent back by server */
-};
-
-extern int afs_rxfs_fetch_file_data(struct afs_server *,
-				    struct afs_vnode *,
-				    struct afs_rxfs_fetch_descriptor *,
-				    struct afs_volsync *);
-
-extern int afs_rxfs_give_up_callback(struct afs_server *,
-				     struct afs_vnode *);
-
-/* this doesn't appear to work in OpenAFS server */
-extern int afs_rxfs_lookup(struct afs_server *,
-			   struct afs_vnode *,
-			   const char *,
-			   struct afs_vnode *,
-			   struct afs_volsync *);
-
-/* this is apparently mis-implemented in OpenAFS server */
-extern int afs_rxfs_get_root_volume(struct afs_server *,
-				    char *,
-				    size_t *);
-
-
-#endif /* AFS_FSCLIENT_H */

fs/afs/inode.c

@@ -19,9 +19,6 @@
 #include <linux/slab.h>
 #include <linux/fs.h>
 #include <linux/pagemap.h>
-#include "volume.h"
-#include "vnode.h"
-#include "super.h"
 #include "internal.h"
 
 struct afs_iget_data {
@@ -40,7 +37,7 @@ static int afs_inode_map_status(struct afs_vnode *vnode)
 	       vnode->status.type,
 	       vnode->status.nlink,
 	       vnode->status.size,
-	       vnode->status.version,
+	       vnode->status.data_version,
 	       vnode->status.mode);
 
 	switch (vnode->status.type) {
@@ -78,7 +75,7 @@ static int afs_inode_map_status(struct afs_vnode *vnode)
 	if (vnode->status.type == AFS_FTYPE_SYMLINK) {
 		afs_mntpt_check_symlink(vnode);
 
-		if (vnode->flags & AFS_VNODE_MOUNTPOINT) {
+		if (test_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags)) {
 			inode->i_mode	= S_IFDIR | vnode->status.mode;
 			inode->i_op	= &afs_mntpt_inode_operations;
 			inode->i_fop	= &afs_mntpt_file_operations;
@@ -88,25 +85,6 @@ static int afs_inode_map_status(struct afs_vnode *vnode)
 	return 0;
 }
 
-/*
- * attempt to fetch the status of an inode, coelescing multiple simultaneous
- * fetches
- */
-static int afs_inode_fetch_status(struct inode *inode)
-{
-	struct afs_vnode *vnode;
-	int ret;
-
-	vnode = AFS_FS_I(inode);
-
-	ret = afs_vnode_fetch_status(vnode);
-
-	if (ret == 0)
-		ret = afs_inode_map_status(vnode);
-
-	return ret;
-}
-
 /*
  * iget5() comparator
  */
@@ -137,8 +115,7 @@ static int afs_iget5_set(struct inode *inode, void *opaque)
 /*
  * inode retrieval
  */
-inline int afs_iget(struct super_block *sb, struct afs_fid *fid,
-		    struct inode **_inode)
+inline struct inode *afs_iget(struct super_block *sb, struct afs_fid *fid)
 {
 	struct afs_iget_data data = { .fid = *fid };
 	struct afs_super_info *as;
@@ -155,20 +132,18 @@ inline int afs_iget(struct super_block *sb, struct afs_fid *fid,
 			     &data);
 	if (!inode) {
 		_leave(" = -ENOMEM");
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 	}
 
+	_debug("GOT INODE %p { vl=%x vn=%x, u=%x }",
+	       inode, fid->vid, fid->vnode, fid->unique);
+
 	vnode = AFS_FS_I(inode);
 
 	/* deal with an existing inode */
 	if (!(inode->i_state & I_NEW)) {
-		ret = afs_vnode_fetch_status(vnode);
-		if (ret == 0)
-			*_inode = inode;
-		else
-			iput(inode);
-		_leave(" = %d", ret);
-		return ret;
+		_leave(" = %p", inode);
+		return inode;
 	}
 
 #ifdef AFS_CACHING_SUPPORT
@@ -181,21 +156,19 @@ inline int afs_iget(struct super_block *sb, struct afs_fid *fid,
 #endif
 
 	/* okay... it's a new inode */
-	inode->i_flags |= S_NOATIME;
-	vnode->flags |= AFS_VNODE_CHANGED;
-	ret = afs_inode_fetch_status(inode);
-	if (ret<0)
+	set_bit(AFS_VNODE_CB_BROKEN, &vnode->flags);
+	ret = afs_vnode_fetch_status(vnode);
+	if (ret < 0)
+		goto bad_inode;
+	ret = afs_inode_map_status(vnode);
+	if (ret < 0)
 		goto bad_inode;
 
 	/* success */
+	inode->i_flags |= S_NOATIME;
 	unlock_new_inode(inode);
-
-	*_inode = inode;
-	_leave(" = 0 [CB { v=%u x=%lu t=%u }]",
-	       vnode->cb_version,
-	       vnode->cb_timeout.timo_jif,
-	       vnode->cb_type);
-	return 0;
+	_leave(" = %p [CB { v=%u t=%u }]", inode, vnode->cb_version, vnode->cb_type);
+	return inode;
 
 	/* failure */
 bad_inode:
@@ -204,7 +177,7 @@ inline int afs_iget(struct super_block *sb, struct afs_fid *fid,
 	iput(inode);
 
 	_leave(" = %d [bad]", ret);
-	return ret;
+	return ERR_PTR(ret);
 }
 
 /*
@@ -213,36 +186,13 @@ inline int afs_iget(struct super_block *sb, struct afs_fid *fid,
 int afs_inode_getattr(struct vfsmount *mnt, struct dentry *dentry,
 		      struct kstat *stat)
 {
-	struct afs_vnode *vnode;
 	struct inode *inode;
-	int ret;
 
 	inode = dentry->d_inode;
 
 	_enter("{ ino=%lu v=%lu }", inode->i_ino, inode->i_version);
 
-	vnode = AFS_FS_I(inode);
-
-	ret = afs_inode_fetch_status(inode);
-	if (ret == -ENOENT) {
-		_leave(" = %d [%d %p]",
-		       ret, atomic_read(&dentry->d_count), dentry->d_inode);
-		return ret;
-	} else if (ret < 0) {
-		make_bad_inode(inode);
-		_leave(" = %d", ret);
-		return ret;
-	}
-
-	/* transfer attributes from the inode structure to the stat
-	 * structure */
 	generic_fillattr(inode, stat);
-
-	_leave(" = 0 CB { v=%u x=%u t=%u }",
-	       vnode->cb_version,
-	       vnode->cb_expiry,
-	       vnode->cb_type);
-
 	return 0;
 }
 
@@ -260,12 +210,23 @@ void afs_clear_inode(struct inode *inode)
 	       vnode->fid.vnode,
 	       vnode->cb_version,
 	       vnode->cb_expiry,
-	       vnode->cb_type
-	       );
+	       vnode->cb_type);
+
+	_debug("CLEAR INODE %p", inode);
+
+	ASSERTCMP(inode->i_ino, ==, vnode->fid.vnode);
 
-	BUG_ON(inode->i_ino != vnode->fid.vnode);
+	afs_give_up_callback(vnode);
+
+	if (vnode->server) {
+		spin_lock(&vnode->server->fs_lock);
+		rb_erase(&vnode->server_rb, &vnode->server->fs_vnodes);
+		spin_unlock(&vnode->server->fs_lock);
+		afs_put_server(vnode->server);
+		vnode->server = NULL;
+	}
 
-	afs_vnode_give_up_callback(vnode);
+	ASSERT(!vnode->cb_promised);
 
 #ifdef AFS_CACHING_SUPPORT
 	cachefs_relinquish_cookie(vnode->cache, 0);

fs/afs/kafsasyncd.c

-/* AFS asynchronous operation daemon
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- *
- *
- * The AFS async daemon is used to the following:
- * - probe "dead" servers to see whether they've come back to life yet.
- * - probe "live" servers that we haven't talked to for a while to see if they are better
- *   candidates for serving than what we're currently using
- * - poll volume location servers to keep up to date volume location lists
- */
-
-#include <linux/module.h>
-#include <linux/init.h>
-#include <linux/sched.h>
-#include <linux/completion.h>
-#include <linux/freezer.h>
-#include "cell.h"
-#include "server.h"
-#include "volume.h"
-#include "kafsasyncd.h"
-#include "kafstimod.h"
-#include <rxrpc/call.h>
-#include <asm/errno.h>
-#include "internal.h"
-
-static DECLARE_COMPLETION(kafsasyncd_alive);
-static DECLARE_COMPLETION(kafsasyncd_dead);
-static DECLARE_WAIT_QUEUE_HEAD(kafsasyncd_sleepq);
-static struct task_struct *kafsasyncd_task;
-static int kafsasyncd_die;
-
-static int kafsasyncd(void *arg);
-
-static LIST_HEAD(kafsasyncd_async_attnq);
-static LIST_HEAD(kafsasyncd_async_busyq);
-static DEFINE_SPINLOCK(kafsasyncd_async_lock);
-
-static void kafsasyncd_null_call_attn_func(struct rxrpc_call *call)
-{
-}
-
-static void kafsasyncd_null_call_error_func(struct rxrpc_call *call)
-{
-}
-
-/*
- * start the async daemon
- */
-int afs_kafsasyncd_start(void)
-{
-	int ret;
-
-	ret = kernel_thread(kafsasyncd, NULL, 0);
-	if (ret < 0)
-		return ret;
-
-	wait_for_completion(&kafsasyncd_alive);
-
-	return ret;
-}
-
-/*
- * stop the async daemon
- */
-void afs_kafsasyncd_stop(void)
-{
-	/* get rid of my daemon */
-	kafsasyncd_die = 1;
-	wake_up(&kafsasyncd_sleepq);
-	wait_for_completion(&kafsasyncd_dead);
-}
-
-/*
- * probing daemon
- */
-static int kafsasyncd(void *arg)
-{
-	struct afs_async_op *op;
-	int die;
-
-	DECLARE_WAITQUEUE(myself, current);
-
-	kafsasyncd_task = current;
-
-	printk("kAFS: Started kafsasyncd %d\n", current->pid);
-
-	daemonize("kafsasyncd");
-
-	complete(&kafsasyncd_alive);
-
-	/* loop around looking for things to attend to */
-	do {
-		set_current_state(TASK_INTERRUPTIBLE);
-		add_wait_queue(&kafsasyncd_sleepq, &myself);
-
-		for (;;) {
-			if (!list_empty(&kafsasyncd_async_attnq) ||
-			    signal_pending(current) ||
-			    kafsasyncd_die)
-				break;
-
-			schedule();
-			set_current_state(TASK_INTERRUPTIBLE);
-		}
-
-		remove_wait_queue(&kafsasyncd_sleepq, &myself);
-		set_current_state(TASK_RUNNING);
-
-		try_to_freeze();
-
-		/* discard pending signals */
-		afs_discard_my_signals();
-
-		die = kafsasyncd_die;
-
-		/* deal with the next asynchronous operation requiring
-		 * attention */
-		if (!list_empty(&kafsasyncd_async_attnq)) {
-			struct afs_async_op *op;
-
-			_debug("@@@ Begin Asynchronous Operation");
-
-			op = NULL;
-			spin_lock(&kafsasyncd_async_lock);
-
-			if (!list_empty(&kafsasyncd_async_attnq)) {
-				op = list_entry(kafsasyncd_async_attnq.next,
-						struct afs_async_op, link);
-				list_move_tail(&op->link,
-					      &kafsasyncd_async_busyq);
-			}
-
-			spin_unlock(&kafsasyncd_async_lock);
-
-			_debug("@@@ Operation %p {%p}\n",
-			       op, op ? op->ops : NULL);
-
-			if (op)
-				op->ops->attend(op);
-
-			_debug("@@@ End Asynchronous Operation");
-		}
-
-	} while(!die);
-
-	/* need to kill all outstanding asynchronous operations before
-	 * exiting */
-	kafsasyncd_task = NULL;
-	spin_lock(&kafsasyncd_async_lock);
-
-	/* fold the busy and attention queues together */
-	list_splice_init(&kafsasyncd_async_busyq,
-			 &kafsasyncd_async_attnq);
-
-	/* dequeue kafsasyncd from all their wait queues */
-	list_for_each_entry(op, &kafsasyncd_async_attnq, link) {
-		op->call->app_attn_func = kafsasyncd_null_call_attn_func;
-		op->call->app_error_func = kafsasyncd_null_call_error_func;
-		remove_wait_queue(&op->call->waitq, &op->waiter);
-	}
-
-	spin_unlock(&kafsasyncd_async_lock);
-
-	/* abort all the operations */
-	while (!list_empty(&kafsasyncd_async_attnq)) {
-		op = list_entry(kafsasyncd_async_attnq.next, struct afs_async_op, link);
-		list_del_init(&op->link);
-
-		rxrpc_call_abort(op->call, -EIO);
-		rxrpc_put_call(op->call);
-		op->call = NULL;
-
-		op->ops->discard(op);
-	}
-
-	/* and that's all */
-	_leave("");
-	complete_and_exit(&kafsasyncd_dead, 0);
-}
-
-/*
- * begin an operation
- * - place operation on busy queue
- */
-void afs_kafsasyncd_begin_op(struct afs_async_op *op)
-{
-	_enter("");
-
-	spin_lock(&kafsasyncd_async_lock);
-
-	init_waitqueue_entry(&op->waiter, kafsasyncd_task);
-	add_wait_queue(&op->call->waitq, &op->waiter);
-
-	list_move_tail(&op->link, &kafsasyncd_async_busyq);
-
-	spin_unlock(&kafsasyncd_async_lock);
-
-	_leave("");
-}
-
-/*
- * request attention for an operation
- * - move to attention queue
- */
-void afs_kafsasyncd_attend_op(struct afs_async_op *op)
-{
-	_enter("");
-
-	spin_lock(&kafsasyncd_async_lock);
-
-	list_move_tail(&op->link, &kafsasyncd_async_attnq);
-
-	spin_unlock(&kafsasyncd_async_lock);
-
-	wake_up(&kafsasyncd_sleepq);
-
-	_leave("");
-}
-
-/*
- * terminate an operation
- * - remove from either queue
- */
-void afs_kafsasyncd_terminate_op(struct afs_async_op *op)
-{
-	_enter("");
-
-	spin_lock(&kafsasyncd_async_lock);
-
-	if (!list_empty(&op->link)) {
-		list_del_init(&op->link);
-		remove_wait_queue(&op->call->waitq, &op->waiter);
-	}
-
-	spin_unlock(&kafsasyncd_async_lock);
-
-	wake_up(&kafsasyncd_sleepq);
-
-	_leave("");
-}

fs/afs/kafsasyncd.h

-/* AFS asynchronous operation daemon
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#ifndef AFS_KAFSASYNCD_H
-#define AFS_KAFSASYNCD_H
-
-#include "types.h"
-
-struct afs_async_op;
-
-struct afs_async_op_ops {
-	void (*attend)(struct afs_async_op *);
-	void (*discard)(struct afs_async_op *);
-};
-
-/*
- * asynchronous operation record
- */
-struct afs_async_op {
-	struct list_head		link;
-	struct afs_server		*server;	/* server being contacted */
-	struct rxrpc_call		*call;		/* RxRPC call performing op */
-	wait_queue_t			waiter;		/* wait queue for kafsasyncd */
-	const struct afs_async_op_ops	*ops;		/* operations */
-};
-
-static inline void afs_async_op_init(struct afs_async_op *op,
-				     const struct afs_async_op_ops *ops)
-{
-	INIT_LIST_HEAD(&op->link);
-	op->call = NULL;
-	op->ops = ops;
-}
-
-extern int afs_kafsasyncd_start(void);
-extern void afs_kafsasyncd_stop(void);
-
-extern void afs_kafsasyncd_begin_op(struct afs_async_op *);
-extern void afs_kafsasyncd_attend_op(struct afs_async_op *);
-extern void afs_kafsasyncd_terminate_op(struct afs_async_op *);
-
-#endif /* AFS_KAFSASYNCD_H */

fs/afs/kafstimod.c

-/* AFS timeout daemon
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#include <linux/module.h>
-#include <linux/init.h>
-#include <linux/sched.h>
-#include <linux/completion.h>
-#include <linux/freezer.h>
-#include "cell.h"
-#include "volume.h"
-#include "kafstimod.h"
-#include <asm/errno.h>
-#include "internal.h"
-
-static DECLARE_COMPLETION(kafstimod_alive);
-static DECLARE_COMPLETION(kafstimod_dead);
-static DECLARE_WAIT_QUEUE_HEAD(kafstimod_sleepq);
-static int kafstimod_die;
-
-static LIST_HEAD(kafstimod_list);
-static DEFINE_SPINLOCK(kafstimod_lock);
-
-static int kafstimod(void *arg);
-
-/*
- * start the timeout daemon
- */
-int afs_kafstimod_start(void)
-{
-	int ret;
-
-	ret = kernel_thread(kafstimod, NULL, 0);
-	if (ret < 0)
-		return ret;
-
-	wait_for_completion(&kafstimod_alive);
-
-	return ret;
-}
-
-/*
- * stop the timeout daemon
- */
-void afs_kafstimod_stop(void)
-{
-	/* get rid of my daemon */
-	kafstimod_die = 1;
-	wake_up(&kafstimod_sleepq);
-	wait_for_completion(&kafstimod_dead);
-}
-
-/*
- * timeout processing daemon
- */
-static int kafstimod(void *arg)
-{
-	struct afs_timer *timer;
-
-	DECLARE_WAITQUEUE(myself, current);
-
-	printk("kAFS: Started kafstimod %d\n", current->pid);
-
-	daemonize("kafstimod");
-
-	complete(&kafstimod_alive);
-
-	/* loop around looking for things to attend to */
-loop:
-	set_current_state(TASK_INTERRUPTIBLE);
-	add_wait_queue(&kafstimod_sleepq, &myself);
-
-	for (;;) {
-		unsigned long jif;
-		signed long timeout;
-
-		/* deal with the server being asked to die */
-		if (kafstimod_die) {
-			remove_wait_queue(&kafstimod_sleepq, &myself);
-			_leave("");
-			complete_and_exit(&kafstimod_dead, 0);
-		}
-
-		try_to_freeze();
-
-		/* discard pending signals */
-		afs_discard_my_signals();
-
-		/* work out the time to elapse before the next event */
-		spin_lock(&kafstimod_lock);
-		if (list_empty(&kafstimod_list)) {
-			timeout = MAX_SCHEDULE_TIMEOUT;
-		} else {
-			timer = list_entry(kafstimod_list.next,
-					   struct afs_timer, link);
-			timeout = timer->timo_jif;
-			jif = jiffies;
-
-			if (time_before_eq((unsigned long) timeout, jif))
-				goto immediate;
-			timeout = (long) timeout - (long) jiffies;
-		}
-		spin_unlock(&kafstimod_lock);
-
-		schedule_timeout(timeout);
-
-		set_current_state(TASK_INTERRUPTIBLE);
-	}
-
-	/* the thing on the front of the queue needs processing
-	 * - we come here with the lock held and timer pointing to the expired
-	 *   entry
-	 */
-immediate:
-	remove_wait_queue(&kafstimod_sleepq, &myself);
-	set_current_state(TASK_RUNNING);
-
-	_debug("@@@ Begin Timeout of %p", timer);
-
-	/* dequeue the timer */
-	list_del_init(&timer->link);
-	spin_unlock(&kafstimod_lock);
-
-	/* call the timeout function */
-	timer->ops->timed_out(timer);
-
-	_debug("@@@ End Timeout");
-	goto loop;
-}
-
-/*
- * (re-)queue a timer
- */
-void afs_kafstimod_add_timer(struct afs_timer *timer, unsigned long timeout)
-{
-	struct afs_timer *ptimer;
-	struct list_head *_p;
-
-	_enter("%p,%lu", timer, timeout);
-
-	spin_lock(&kafstimod_lock);
-
-	list_del(&timer->link);
-
-	/* the timer was deferred or reset - put it back in the queue at the
-	 * right place */
-	timer->timo_jif = jiffies + timeout;
-
-	list_for_each(_p, &kafstimod_list) {
-		ptimer = list_entry(_p, struct afs_timer, link);
-		if (time_before(timer->timo_jif, ptimer->timo_jif))
-			break;
-	}
-
-	list_add_tail(&timer->link, _p); /* insert before stopping point */
-
-	spin_unlock(&kafstimod_lock);
-
-	wake_up(&kafstimod_sleepq);
-
-	_leave("");
-}
-
-/*
- * dequeue a timer
- * - returns 0 if the timer was deleted or -ENOENT if it wasn't queued
- */
-int afs_kafstimod_del_timer(struct afs_timer *timer)
-{
-	int ret = 0;
-
-	_enter("%p", timer);
-
-	spin_lock(&kafstimod_lock);
-
-	if (list_empty(&timer->link))
-		ret = -ENOENT;
-	else
-		list_del_init(&timer->link);
-
-	spin_unlock(&kafstimod_lock);
-
-	wake_up(&kafstimod_sleepq);
-
-	_leave(" = %d", ret);
-	return ret;
-}

fs/afs/kafstimod.h

-/* AFS timeout daemon
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#ifndef AFS_KAFSTIMOD_H
-#define AFS_KAFSTIMOD_H
-
-#include "types.h"
-
-struct afs_timer;
-
-struct afs_timer_ops {
-	/* called when the front of the timer queue has timed out */
-	void (*timed_out)(struct afs_timer *);
-};
-
-/*
- * AFS timer/timeout record
- */
-struct afs_timer {
-	struct list_head		link;		/* link in timer queue */
-	unsigned long			timo_jif;	/* timeout time */
-	const struct afs_timer_ops	*ops;		/* timeout expiry function */
-};
-
-static inline void afs_timer_init(struct afs_timer *timer,
-				  const struct afs_timer_ops *ops)
-{
-	INIT_LIST_HEAD(&timer->link);
-	timer->ops = ops;
-}
-
-extern int afs_kafstimod_start(void);
-extern void afs_kafstimod_stop(void);
-extern void afs_kafstimod_add_timer(struct afs_timer *, unsigned long);
-extern int afs_kafstimod_del_timer(struct afs_timer *);
-
-#endif /* AFS_KAFSTIMOD_H */

fs/afs/transport.h

-/* AFS transport management
- *
- * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version
- * 2 of the License, or (at your option) any later version.
- */
-
-#ifndef AFS_TRANSPORT_H
-#define AFS_TRANSPORT_H
-
-#include "types.h"
-#include <rxrpc/transport.h>
-
-/* the cache manager transport endpoint */
-extern struct rxrpc_transport *afs_transport;
-
-#endif /* AFS_TRANSPORT_H */