patch standalone_next3_module.patch

fs/buffer.c

@@ -16,8 +16,6 @@
  * Added 32k buffer block sizes - these are required older ARM systems. - RMK
  *
  * async buffer flushing, 1999 Andrea Arcangeli <andrea@suse.de>
- *
- * Tracked buffer read for Next3, Amir Goldstein <amir73il@users.sf.net>, 2008
  */
 
 #include <linux/kernel.h>
@@ -296,123 +294,6 @@ static void free_more_memory(void)
 	}
 }
 
-#ifdef CONFIG_NEXT3_FS_SNAPSHOT_RACE_READ
-/*
- * Tracked read functions.
- * When reading through a next3 snapshot file hole to a block device block,
- * all writes to this block need to wait for completion of the async read.
- * next3_snapshot_readpage() always calls block_read_full_page() to attach
- * a buffer head to the page and be aware of tracked reads.
- * next3_snapshot_get_block() calls start_buffer_tracked_read() to mark both
- * snapshot page buffer and block device page buffer.
- * next3_snapshot_get_block() calls cancel_buffer_tracked_read() if snapshot
- * doesn't need to read through to the block device.
- * block_read_full_page() calls submit_buffer_tracked_read() to submit a
- * tracked async read.
- * end_buffer_async_read() calls end_buffer_tracked_read() to complete the
- * tracked read operation.
- * The only lock needed in all these functions is PageLock on the snapshot page,
- * which is guarantied in readpage() and verified in block_read_full_page().
- * The block device page buffer doesn't need any lock because the operations
- * {get|put}_bh_tracked_reader() are atomic.
- */
-
-/*
- * start buffer tracked read
- * called from inside get_block()
- * get tracked reader ref count on buffer cache entry
- * and set buffer tracked read flag
- */
-int start_buffer_tracked_read(struct buffer_head *bh)
-{
-	struct buffer_head *bdev_bh;
-
-	BUG_ON(buffer_tracked_read(bh));
-	BUG_ON(!buffer_mapped(bh));
-
-	/* grab the buffer cache entry */
-	bdev_bh = __getblk(bh->b_bdev, bh->b_blocknr, bh->b_size);
-	if (!bdev_bh)
-		return -EIO;
-
-	BUG_ON(bdev_bh == bh);
-	set_buffer_tracked_read(bh);
-	get_bh_tracked_reader(bdev_bh);
-	put_bh(bdev_bh);
-	return 0;
-}
-EXPORT_SYMBOL_GPL(start_buffer_tracked_read);
-
-/*
- * cancel buffer tracked read
- * called for tracked read that was started but was not submitted
- * put tracked reader ref count on buffer cache entry
- * and clear buffer tracked read flag
- */
-void cancel_buffer_tracked_read(struct buffer_head *bh)
-{
-	struct buffer_head *bdev_bh;
-
-	BUG_ON(!buffer_tracked_read(bh));
-	BUG_ON(!buffer_mapped(bh));
-
-	/* try to grab the buffer cache entry */
-	bdev_bh = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
-	BUG_ON(!bdev_bh || bdev_bh == bh);
-	clear_buffer_tracked_read(bh);
-	clear_buffer_mapped(bh);
-	put_bh_tracked_reader(bdev_bh);
-	put_bh(bdev_bh);
-}
-EXPORT_SYMBOL_GPL(cancel_buffer_tracked_read);
-
-/*
- * submit buffer tracked read
- * save a reference to buffer cache entry and submit I/O
- */
-static int submit_buffer_tracked_read(struct buffer_head *bh)
-{
-	struct buffer_head *bdev_bh;
-	BUG_ON(!buffer_tracked_read(bh));
-	BUG_ON(!buffer_mapped(bh));
-	/* tracked read doesn't work with multiple buffers per page */
-	BUG_ON(bh->b_this_page != bh);
-
-	/*
-	 * Try to grab the buffer cache entry before submitting async read
-	 * because we cannot call blocking function __find_get_block()
-	 * in interrupt context inside end_buffer_tracked_read().
-	 */
-	bdev_bh = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
-	BUG_ON(!bdev_bh || bdev_bh == bh);
-	/* override page buffers list with reference to buffer cache entry */
-	bh->b_this_page = bdev_bh;
-	submit_bh(READ, bh);
-	return 0;
-}
-
-/*
- * end buffer tracked read
- * complete submitted tracked read
- */
-static void end_buffer_tracked_read(struct buffer_head *bh)
-{
-	struct buffer_head *bdev_bh = bh->b_this_page;
-
-	BUG_ON(!buffer_tracked_read(bh));
-	BUG_ON(!bdev_bh || bdev_bh == bh);
-	bh->b_this_page = bh;
-	/*
-	 * clear the buffer mapping to make sure
-	 * that get_block() will always be called
-	 */
-	clear_buffer_mapped(bh);
-	clear_buffer_tracked_read(bh);
-	put_bh_tracked_reader(bdev_bh);
-	put_bh(bdev_bh);
-}
-
-#endif
 /*
  * I/O completion handler for block_read_full_page() - pages
  * which come unlocked at the end of I/O.
@@ -427,11 +308,6 @@ static void end_buffer_async_read(struct buffer_head *bh, int uptodate)
 
 	BUG_ON(!buffer_async_read(bh));
 
-#ifdef CONFIG_NEXT3_FS_SNAPSHOT_RACE_READ
-	if (buffer_tracked_read(bh))
-		end_buffer_tracked_read(bh);
-
-#endif
 	page = bh->b_page;
 	if (uptodate) {
 		set_buffer_uptodate(bh);
@@ -2335,10 +2211,6 @@ int block_read_full_page(struct page *page, get_block_t *get_block)
 	 */
 	for (i = 0; i < nr; i++) {
 		bh = arr[i];
-#ifdef CONFIG_NEXT3_FS_SNAPSHOT_RACE_READ
-		if (buffer_tracked_read(bh))
-			return submit_buffer_tracked_read(bh);
-#endif
 		if (buffer_uptodate(bh))
 			end_buffer_async_read(bh, 1);
 		else

fs/jbd/journal.c

@@ -1903,6 +1903,11 @@ void journal_put_journal_head(struct journal_head *jh)
 u8 journal_enable_debug __read_mostly;
 EXPORT_SYMBOL(journal_enable_debug);
 
+#ifdef CONFIG_NEXT3_FS_SNAPSHOT_JOURNAL_TRACE
+const u8 journal_handle_size = sizeof(handle_t);
+EXPORT_SYMBOL(journal_handle_size);
+
+#endif
 static struct dentry *jbd_debugfs_dir;
 static struct dentry *jbd_debug;
 

fs/next3/Kconfig

@@ -1,6 +1,6 @@
 config NEXT3_FS
 	tristate "Next3 journalling file system support"
-	select JBD
+	depends on JBD
 	help
 	  This is the journalling version of the Second extended file system
 	  (often called next3), the de facto standard Linux file system
@@ -65,7 +65,7 @@ config NEXT3_FS_XATTR
 config NEXT3_FS_POSIX_ACL
 	bool "Next3 POSIX Access Control Lists"
 	depends on NEXT3_FS_XATTR
-	select FS_POSIX_ACL
+	depends on FS_POSIX_ACL
 	help
 	  Posix Access Control Lists (ACLs) support permissions for users and
 	  groups beyond the owner/group/world scheme.

fs/next3/Makefile

@@ -2,11 +2,18 @@
 # Makefile for the linux next3-filesystem routines.
 #
 
+# default configuration for standalone module
+CONFIG_NEXT3_FS?=m
+CONFIG_NEXT3_FS_XATTR?=y
+CONFIG_NEXT3_FS_POSIX_ACL?=$(CONFIG_FS_POSIX_ACL)
+CONFIG_NEXT3_FS_SECURITY?=y
+CONFIG_NEXT3_FS_DEBUG?=y
+
 obj-$(CONFIG_NEXT3_FS) += next3.o
 
 next3-y	:= balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \
 	   ioctl.o namei.o super.o symlink.o hash.o resize.o next3_jbd.o
-next3-y	+= snapshot.o snapshot_ctl.o
+next3-y	+= snapshot.o snapshot_ctl.o buffer.o
 
 next3-$(CONFIG_NEXT3_FS_XATTR)	 += xattr.o xattr_user.o xattr_trusted.o
 next3-$(CONFIG_NEXT3_FS_POSIX_ACL) += acl.o

fs/next3/balloc.c

@@ -1703,7 +1703,7 @@ static int next3_has_free_blocks(struct next3_sb_info *sbi)
 		 * snapshot reserved blocks for COWing to active snapshot
 		 */
 		if (free_blocks < snapshot_r_blocks + 1 &&
-		    !handle->h_cowing) {
+		    !IS_COWING(handle)) {
 			return 0;
 		}
 		/*

fs/next3/inode.c

@@ -1488,9 +1488,9 @@ int next3_get_blocks_handle(handle_t *handle, struct inode *inode,
 	 * operation, because snapshot file has read-only aops and because
 	 * truncate/unlink of snapshot file is not permitted.
 	 */
-	BUG_ON(next3_snapshot_is_active(inode) && !handle->h_cowing);
-	BUG_ON(!next3_snapshot_is_active(inode) && handle->h_cowing);
-	mutex_lock_nested(&ei->truncate_mutex, handle->h_cowing);
+	BUG_ON(next3_snapshot_is_active(inode) && !IS_COWING(handle));
+	BUG_ON(!next3_snapshot_is_active(inode) && IS_COWING(handle));
+	mutex_lock_nested(&ei->truncate_mutex, IS_COWING(handle));
 #else
 	mutex_lock(&ei->truncate_mutex);
 #endif
@@ -2601,7 +2601,7 @@ static int next3_snapshot_get_block(struct inode *inode, sector_t iblock,
 static int next3_snapshot_readpage(struct file *file, struct page *page)
 {
 	/* do read I/O with buffer heads to enable tracked reads */
-	return block_read_full_page(page, next3_snapshot_get_block);
+	return next3_read_full_page(page, next3_snapshot_get_block);
 }
 
 #endif

fs/next3/next3.h

@@ -25,6 +25,19 @@
 #ifdef CONFIG_NEXT3_FS_SNAPSHOT
 #ifndef NEXT3_SUPER_MAGIC
 #define NEXT3_SUPER_MAGIC EXT3_SUPER_MAGIC
+/* if the kernel was not patched, next3 is compiled as standalone module */
+#define CONFIG_NEXT3_FS_STANDALONE
+#endif
+
+#ifdef CONFIG_NEXT3_FS_STANDALONE
+/* configuration options for standalone module */
+#define CONFIG_NEXT3_DEFAULTS_TO_ORDERED
+#define CONFIG_NEXT3_FS_XATTR
+#define CONFIG_NEXT3_FS_SECURITY
+#define CONFIG_NEXT3_FS_DEBUG
+#endif
+#if defined(CONFIG_NEXT3_FS_STANDALONE) && defined(CONFIG_FS_POSIX_ACL)
+#define CONFIG_NEXT3_FS_POSIX_ACL
 #endif
 
 #endif

fs/next3/next3_jbd.c

@@ -95,20 +95,20 @@ int __next3_journal_dirty_metadata(const char *where,
 	if (err)
 		return err;
 
-	if (!handle->h_cowing) {
+	if (!IS_COWING(handle)) {
 		struct journal_head *jh = bh2jh(bh);
 		jbd_lock_bh_state(bh);
-		if (jh->b_modified && !jh->b_user_modified) {
+		if (jh->b_modified == 1) {
 			/*
 			 * buffer_credits was decremented when buffer was
 			 * modified for the first time in the current
 			 * transaction, which may have been during a COW
 			 * operation.  We decrement user_credits and mark
-			 * b_user_modified, on the first time that the buffer
+			 * b_modified = 2, on the first time that the buffer
 			 * is modified not during a COW operation (!h_cowing).
 			 */
-			jh->b_user_modified = 1;
-			handle->h_user_credits--;
+			jh->b_modified = 2;
+			((next3_handle_t *)handle)->h_user_credits--;
 		}
 		jbd_unlock_bh_state(bh);
 	}
@@ -125,7 +125,7 @@ int __next3_journal_release_buffer(const char *where, handle_t *handle,
 {
 	int err = 0;
 
-	if (handle->h_cowing)
+	if (IS_COWING(handle))
 		goto out;
 
 	/*
@@ -142,7 +142,7 @@ int __next3_journal_release_buffer(const char *where, handle_t *handle,
 		snapshot_debug(1, "%s: warning: couldn't extend transaction "
 				"from %s (credits=%d/%d)\n", __func__,
 				where, handle->h_buffer_credits,
-				handle->h_user_credits);
+				((next3_handle_t *)handle)->h_user_credits);
 		err = 0;
 	}
 	next3_journal_trace(SNAP_WARN, where, handle, -1);
@@ -154,9 +154,9 @@ int __next3_journal_release_buffer(const char *where, handle_t *handle,
 #endif
 #ifdef CONFIG_NEXT3_FS_SNAPSHOT_JOURNAL_TRACE
 #ifdef CONFIG_JBD_DEBUG
-static void next3_journal_cow_stats(int n, handle_t *handle)
+static void next3_journal_cow_stats(int n, next3_handle_t *handle)
 {
-	if (!journal_enable_debug)
+	if (!trace_cow_enabled())
 		return;
 	snapshot_debug(n, "COW stats: moved/copied=%d/%d, "
 			 "mapped/bitmap/cached=%d/%d/%d, "
@@ -171,14 +171,14 @@ static void next3_journal_cow_stats(int n, handle_t *handle)
 
 #ifdef CONFIG_NEXT3_FS_DEBUG
 void __next3_journal_trace(int n, const char *fn, const char *caller,
-		handle_t *handle, int nblocks)
+		next3_handle_t *handle, int nblocks)
 {
 	struct super_block *sb = handle->h_transaction->t_journal->j_private;
 	struct inode *active_snapshot = next3_snapshot_has_active(sb);
 	int upper = NEXT3_SNAPSHOT_START_TRANS_BLOCKS(handle->h_base_credits);
 	int lower = NEXT3_SNAPSHOT_TRANS_BLOCKS(handle->h_user_credits);
 	int final = (nblocks == 0 && handle->h_ref == 1 &&
-		     !handle->h_cowing);
+		     !IS_COWING(handle));
 
 	switch (snapshot_enable_debug) {
 	case SNAP_INFO:
@@ -192,7 +192,7 @@ void __next3_journal_trace(int n, const char *fn, const char *caller,
 		 * valid if there is an active snapshot and not during COW
 		 */
 		if (handle->h_buffer_credits < lower &&
-		    active_snapshot && !handle->h_cowing)
+		    active_snapshot && !IS_COWING(handle))
 			break;
 	case SNAP_ERR:
 		/* trace if user credits are too low */
@@ -208,7 +208,7 @@ void __next3_journal_trace(int n, const char *fn, const char *caller,
 		break;
 	}
 
-	snapshot_debug_l(n, handle->h_cowing, "%s(%d): credits=%d, limit=%d/%d,"
+	snapshot_debug_l(n, IS_COWING(handle), "%s(%d): credits=%d, limit=%d/%d,"
 			 " user=%d/%d, ref=%d, caller=%s\n", fn, nblocks,
 			 handle->h_buffer_credits, lower, upper,
 			 handle->h_user_credits, handle->h_base_credits,