[Ocfs2-commits] manish commits r2074 - trunk/fs/ocfs2

svn-commits at oss.oracle.com svn-commits at oss.oracle.com
Tue Mar 29 22:14:40 CST 2005


Author: manish
Signed-off-by: mfasheh
Date: 2005-03-29 22:14:38 -0600 (Tue, 29 Mar 2005)
New Revision: 2074

Removed:
   trunk/fs/ocfs2/ocfs_journal.h
Modified:
   trunk/fs/ocfs2/Makefile
   trunk/fs/ocfs2/alloc.c
   trunk/fs/ocfs2/aops.c
   trunk/fs/ocfs2/buffer_head_io.c
   trunk/fs/ocfs2/dir.c
   trunk/fs/ocfs2/dlmglue.c
   trunk/fs/ocfs2/file.c
   trunk/fs/ocfs2/heartbeat.c
   trunk/fs/ocfs2/inode.c
   trunk/fs/ocfs2/journal.c
   trunk/fs/ocfs2/journal.h
   trunk/fs/ocfs2/localalloc.c
   trunk/fs/ocfs2/namei.c
   trunk/fs/ocfs2/proc.c
   trunk/fs/ocfs2/suballoc.c
   trunk/fs/ocfs2/super.c
   trunk/fs/ocfs2/symlink.c
   trunk/fs/ocfs2/sysfile.c
   trunk/fs/ocfs2/vote.c
Log:
Merge ocfs_journal.h and journal.h

Signed-off-by: mfasheh


Modified: trunk/fs/ocfs2/Makefile
===================================================================
--- trunk/fs/ocfs2/Makefile	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/Makefile	2005-03-30 04:14:38 UTC (rev 2074)
@@ -92,7 +92,6 @@
 	ocfs2.h			\
 	ocfs_log.h		\
 	ocfs_compat.h		\
-	ocfs_journal.h		\
 	buffer_head_io.h	\
 	aio.h			\
 	alloc.h			\

Modified: trunk/fs/ocfs2/alloc.c
===================================================================
--- trunk/fs/ocfs2/alloc.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/alloc.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -43,7 +43,7 @@
 #include "sysfile.h"
 #include "file.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 /* Tracing */

Modified: trunk/fs/ocfs2/aops.c
===================================================================
--- trunk/fs/ocfs2/aops.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/aops.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -37,7 +37,7 @@
 #include "inode.h"
 #include "symlink.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_INODE

Modified: trunk/fs/ocfs2/buffer_head_io.c
===================================================================
--- trunk/fs/ocfs2/buffer_head_io.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/buffer_head_io.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -36,7 +36,7 @@
 
 #include "alloc.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_IO

Modified: trunk/fs/ocfs2/dir.c
===================================================================
--- trunk/fs/ocfs2/dir.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/dir.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -53,7 +53,7 @@
 #include "extent_map.h"
 #include "file.h"
 #include "inode.h"
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "namei.h"
 #include "suballoc.h"
 

Modified: trunk/fs/ocfs2/dlmglue.c
===================================================================
--- trunk/fs/ocfs2/dlmglue.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/dlmglue.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -51,7 +51,7 @@
 #include "slot_map.h"
 #include "vote.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_DLMGLUE

Modified: trunk/fs/ocfs2/file.c
===================================================================
--- trunk/fs/ocfs2/file.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/file.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -49,7 +49,7 @@
 #include "mmap.h"
 #include "suballoc.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_FILE

Modified: trunk/fs/ocfs2/heartbeat.c
===================================================================
--- trunk/fs/ocfs2/heartbeat.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/heartbeat.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -42,7 +42,7 @@
 #include "alloc.h"
 #include "heartbeat.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 /* Tracing */

Modified: trunk/fs/ocfs2/inode.c
===================================================================
--- trunk/fs/ocfs2/inode.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/inode.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -50,7 +50,7 @@
 #include "sysfile.h"
 #include "vote.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_INODE

Modified: trunk/fs/ocfs2/journal.c
===================================================================
--- trunk/fs/ocfs2/journal.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/journal.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -48,7 +48,7 @@
 #include "vote.h"
 #include "sysfile.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_JOURNAL

Modified: trunk/fs/ocfs2/journal.h
===================================================================
--- trunk/fs/ocfs2/journal.h	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/journal.h	2005-03-30 04:14:38 UTC (rev 2074)
@@ -3,9 +3,9 @@
  *
  * journal.h
  *
- * Function prototypes
+ * Defines journalling api and structures.
  *
- * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ * Copyright (C) 2003, 2005 Oracle.  All rights reserved.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public
@@ -26,7 +26,424 @@
 #ifndef OCFS2_JOURNAL_H
 #define OCFS2_JOURNAL_H
 
+#include <linux/fs.h>
+#include <linux/jbd.h>
+
+#define OCFS_CHECKPOINT_INTERVAL        (8 * HZ)
+
+enum ocfs_journal_state {
+	OCFS_JOURNAL_FREE = 0,
+	OCFS_JOURNAL_LOADED,
+	OCFS_JOURNAL_IN_SHUTDOWN,
+};
+
+struct _ocfs_super;
+struct _ocfs2_dinode;
+struct _ocfs_journal_handle;
+
+typedef struct _ocfs_journal ocfs_journal;
+struct _ocfs_journal {
+	enum ocfs_journal_state   j_state;    /* Journals current state   */
+
+	journal_t                 *j_journal; /* The kernels journal type */
+	struct inode              *j_inode;   /* Kernel inode pointing to
+					       * this journal             */
+	struct _ocfs_super        *j_osb;     /* pointer to the super
+					       * block for the node
+					       * we're currently
+					       * running on -- not
+					       * necessarily the super
+					       * block from the node
+					       * which we usually run
+					       * from (recovery,
+					       * etc)                     */
+	struct buffer_head        *j_bh;      /* Journal disk inode block */
+	atomic_t                  j_num_trans; /* Number of transactions 
+					        * currently in the system. */
+	unsigned long             j_trans_id;
+	struct rw_semaphore       j_trans_barrier;
+	wait_queue_head_t         j_checkpointed;
+
+	spinlock_t                j_lock;
+	unsigned int              j_cleanup_orphans;
+	struct list_head          j_la_cleanups;
+	struct work_struct        j_recovery_work;
+};
+
+extern spinlock_t trans_inc_lock;
+
+/* wrap j_trans_id so we never have it equal to zero. */
+static inline unsigned long ocfs_inc_trans_id(ocfs_journal *j)
+{
+	unsigned long old_id;
+	spin_lock(&trans_inc_lock);
+	old_id = j->j_trans_id++;
+	if (!j->j_trans_id)
+		j->j_trans_id = 1;
+	spin_unlock(&trans_inc_lock);
+	return old_id;
+}
+
+static inline void ocfs_set_inode_lock_trans(ocfs_journal *journal,
+					     struct inode *inode)
+{
+	spin_lock(&trans_inc_lock);
+	OCFS_I(inode)->ip_last_trans = journal->j_trans_id;
+	spin_unlock(&trans_inc_lock);
+}
+
+/* Used to figure out whether it's safe to drop a metadata lock on an
+ * inode. Returns true if all the inodes changes have been
+ * checkpointed to disk. You should be holding the spinlock on the
+ * metadata lock while calling this to be sure that nobody can take
+ * the lock and put it on another transaction. */
+static inline int ocfs_inode_fully_checkpointed(struct inode *inode)
+{
+	int ret;
+	ocfs_journal *journal = OCFS2_SB(inode->i_sb)->journal;
+
+	spin_lock(&trans_inc_lock);
+	ret = time_after(journal->j_trans_id, OCFS_I(inode)->ip_last_trans);
+	spin_unlock(&trans_inc_lock);
+	return ret;
+}
+
+/* convenience function to check if an inode is still new (has never
+ * hit disk) Will do you a favor and set created_trans = 0 when you've
+ * been checkpointed.  returns '1' if the inode is still new. */
+static inline int ocfs_inode_is_new(struct inode *inode)
+{
+	int ret;
+
+	/* System files are never "new" as they're written out by
+	 * mkfs. This helps us early during mount, before we have the
+	 * journal open and j_trans_id could be junk. */
+	if (OCFS_I(inode)->ip_flags & OCFS_INODE_SYSTEM_FILE)
+		return 0;
+	spin_lock(&trans_inc_lock);
+	ret = !(time_after(OCFS2_SB(inode->i_sb)->journal->j_trans_id, 
+			   OCFS_I(inode)->ip_created_trans));
+	if (!ret)
+		OCFS_I(inode)->ip_created_trans = 0;
+	spin_unlock(&trans_inc_lock);
+	return ret;
+}
+
+static inline void ocfs_inode_set_new(ocfs_super *osb, 
+				      struct inode *inode)
+{
+	spin_lock(&trans_inc_lock);
+	OCFS_I(inode)->ip_created_trans = osb->journal->j_trans_id;
+	spin_unlock(&trans_inc_lock);
+}
+
+typedef struct _ocfs_journal_lock ocfs_journal_lock;
+struct _ocfs_journal_lock {
+	struct inode     *jl_inode;
+	struct list_head  jl_lock_list;
+};
+
+struct _ocfs_journal_handle {
+	handle_t            *k_handle; /* kernel handle.                */
+	ocfs_journal        *journal;
+	struct _ocfs_super  *osb;      /* what super block we belong to */
+	u32                 flags;     /* see flags below.              */
+	int                 max_buffs; /* Buffs reserved by this handle */
+
+	/* The following two fields are for ocfs_handle_add_lock */
+	int                 num_locks; 
+	struct list_head    locks;     /* A bunch of locks to 
+					* release on commit. This 
+					* should be a list_head */
+
+	struct list_head     inode_list;
+};
+
+#define OCFS_HANDLE_STARTED			1
+/* should we sync-commit this handle? */
+#define OCFS_HANDLE_SYNC			2
+static inline int ocfs_handle_started(ocfs_journal_handle *handle)
+{
+	return handle->flags & OCFS_HANDLE_STARTED;
+}
+
+static inline void ocfs_handle_set_sync(ocfs_journal_handle *handle, int sync)
+{
+	if (sync)
+		handle->flags |= OCFS_HANDLE_SYNC;
+	else
+		handle->flags &= ~OCFS_HANDLE_SYNC;
+}
+
+/* Exported only for the journal struct init code in super.c. Do not call. */
+void ocfs2_complete_recovery(void *data);
+
 int ocfs_commit_thread(void *arg);
 int ocfs_recover_orphans(ocfs_super *osb);
 
+/*
+ *  Journal Control:
+ *  Initialize, Load, Shutdown, Wipe a journal.
+ *  
+ *  ocfs_journal_init     - Initialize journal structures in the OSB.
+ *  ocfs_journal_load     - Load the given journal off disk. Replay it if
+ *                          there's transactions still in there.
+ *  ocfs_journal_shutdown - Shutdown a journal, this will flush all 
+ *                          uncommitted, uncheckpointed transactions.
+ *  ocfs_journal_wipe     - Wipe transactions from a journal. Optionally 
+ *                          zero out each block.
+ *  ocfs_recovery_thread  - Perform recovery on a node. osb is our own osb.
+ *  ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
+ *                          event on.
+ *  ocfs_start_checkpoint - Kick the commit thread to do a checkpoint.
+ */
+int    ocfs_journal_init(ocfs_journal *journal, int *dirty);
+void   ocfs_journal_shutdown(struct _ocfs_super *osb);
+int    ocfs_journal_wipe(ocfs_journal *journal, int full);
+int    ocfs_journal_load(ocfs_journal *journal);
+void   ocfs_recovery_thread(struct _ocfs_super *osb, int node_num);
+int    ocfs2_mark_dead_nodes(ocfs_super *osb);
+
+static inline void ocfs_start_checkpoint(struct _ocfs_super *osb)
+{
+	atomic_set(&osb->needs_checkpoint, 1);
+	wake_up(&osb->checkpoint_event);
+}
+
+static inline void ocfs2_checkpoint_inode(struct inode *inode)
+{
+	ocfs_super *osb = OCFS_SB(inode->i_sb);
+
+	if (!ocfs_inode_fully_checkpointed(inode)) {
+		/* WARNING: This only kicks off a single
+		 * checkpoint. If someone races you and adds more
+		 * metadata to the journal, you won't know, and will
+		 * wind up waiting *alot* longer than necessary. Right
+		 * now we only use this in clear_inode so that's
+		 * OK. */
+		ocfs_start_checkpoint(osb);
+
+		wait_event(osb->journal->j_checkpointed,
+			   ocfs_inode_fully_checkpointed(inode));
+	}
+}
+
+/*
+ *  Transaction Handling:
+ *  Manage the lifetime of a transaction handle.
+ *  
+ *  ocfs_alloc_handle     - Only allocate a handle so we can start putting 
+ *                          cluster locks on it. To actually change blocks, 
+ *                          call ocfs_start_trans with the handle returned 
+ *                          from this function. You may call ocfs_commit_trans
+ *                           at any time in the lifetime of a handle.
+ *  ocfs_start_trans      - Begin a transaction. Give it an upper estimate of 
+ *                          the number of blocks that will be changed during 
+ *                          this handle.
+ *  ocfs_commit_trans     - Complete a handle.
+ *  ocfs_extend_trans     - Extend a handle by nblocks credits. This may 
+ *                          commit the handle to disk in the process, but will
+ *                          not release any locks taken during the transaction.
+ *  ocfs_journal_access   - Notify the handle that we want to journal this 
+ *                          buffer. Will have to call ocfs_journal_dirty once
+ *                          we've actually dirtied it. Type is one of . or .
+ *  ocfs_journal_dirty    - Mark a journalled buffer as having dirty data.
+ *  ocfs_handle_add_lock  - Sometimes we need to delay lock release
+ *                          until after a transaction has been completed. Use
+ *                          ocfs_handle_add_lock to indicate that a lock needs
+ *                          to be released at the end of that handle. Locks 
+ *                          will be released in the order that they are added. 
+ *  ocfs_handle_add_inode - Add a locked inode to a transaction.
+ */
+
+/* You must always start_trans with a number of buffs > 0, but it's
+ * perfectly legal to go through an entire transaction without having
+ * dirtied any buffers. */
+ocfs_journal_handle *ocfs_alloc_handle(ocfs_super *osb);
+ocfs_journal_handle *ocfs_start_trans(struct _ocfs_super *osb, 
+				      ocfs_journal_handle *handle, 
+				      int max_buffs);
+void                 ocfs_commit_trans(ocfs_journal_handle *handle);
+int                  ocfs_extend_trans(ocfs_journal_handle *handle, 
+				       int nblocks);
+
+/*
+ * Create access is for when we get a newly created buffer and we're
+ * not gonna read it off disk, but rather fill it ourselves.  Right
+ * now, we don't do anything special with this (it turns into a write
+ * request), but this is a good placeholder in case we do...
+ *
+ * Write access is for when we read a block off disk and are going to
+ * modify it. This way the journalling layer knows it may need to make
+ * a copy of that block (if it's part of another, uncommitted
+ * transaction) before we do so.
+ */
+#define OCFS_JOURNAL_ACCESS_CREATE 0
+#define OCFS_JOURNAL_ACCESS_WRITE  1
+#define OCFS_JOURNAL_ACCESS_UNDO   2
+
+int                  ocfs_journal_access(ocfs_journal_handle *handle, 
+					 struct inode *inode,
+					 struct buffer_head *bh, 
+					 int type);
+/*
+ * A word about the journal_access/journal_dirty "dance". It is
+ * entirely legal to journal_access a buffer more than once (as long
+ * as the access type is the same -- I'm not sure what will happen if
+ * access type is different but this should never happen anyway) It is
+ * also legal to journal_dirty a buffer more than once. In fact, you
+ * can even journal_access a buffer after you've done a
+ * journal_access/journal_dirty pair. The only thing you cannot do
+ * however, is journal_dirty a buffer which you haven't yet passed to
+ * journal_access at least once.
+ *
+ * That said, 99% of the time this doesn't matter and this is what the
+ * path looks like: 
+ * 
+ *	<read a bh>
+ *	ocfs_journal_access(handle, bh,	OCFS_JOURNAL_ACCESS_WRITE); 
+ *	<modify the bh>
+ * 	ocfs_journal_dirty(handle, bh);
+ */
+int                  ocfs_journal_dirty(ocfs_journal_handle *handle, 
+					struct buffer_head *bh);
+int                  ocfs_handle_add_lock(ocfs_journal_handle *handle,
+			                  struct inode *inode);
+/*
+ * Use this to protect from other processes reading buffer state while
+ * it's in flight.
+ */
+void                 ocfs_handle_add_inode(ocfs_journal_handle *handle, 
+					   struct inode *inode);
+
+/*
+ *  Credit Macros:
+ *  Convenience macros to calculate number of credits needed.
+ *
+ *  For convenience sake, I have a set of macros here which calculate
+ *  the *maximum* number of sectors which will be changed for various
+ *  metadata updates.
+ */
+
+/* simple file updates like chmod, etc. */
+#define OCFS_INODE_UPDATE_CREDITS 1
+
+/* get one bit out of a suballocator: dinode + group descriptor +
+ * prev. group desc. if we relink. */
+#define OCFS_SUBALLOC_ALLOC (3)
+
+/* dinode + group descriptor update. We don't relink on free yet. */
+#define OCFS_SUBALLOC_FREE  (2)
+
+/* data block for new dir/symlink, 2 for bitmap updates (bitmap fe +
+ * bitmap block for the new bit) */
+#define OCFS_DIR_LINK_ADDITIONAL_CREDITS (1 + 2)
+
+/* parent fe, parent block, new file entry, inode alloc fe, inode alloc
+ * group descriptor + mkdir/symlink blocks */
+#define OCFS_MKNOD_CREDITS (3 + OCFS_SUBALLOC_ALLOC                           \
+			    + OCFS_DIR_LINK_ADDITIONAL_CREDITS)
+
+/* local alloc metadata change + main bitmap updates */
+#define OCFS_WINDOW_MOVE_CREDITS (OCFS_INODE_UPDATE_CREDITS                   \
+				  + OCFS_SUBALLOC_ALLOC + OCFS_SUBALLOC_FREE)
+
+/* used when we don't need an allocation change for a dir extend. One
+ * for the dinode, one for the new block. */
+#define OCFS_SIMPLE_DIR_EXTEND_CREDITS (2)
+
+/* file update (nlink, etc) + dir entry block */
+#define OCFS_LINK_CREDITS  (OCFS_INODE_UPDATE_CREDITS + 1)
+
+/* inode + dir inode (if we unlink a dir), + dir entry block + orphan
+ * dir inode link */
+#define OCFS_UNLINK_CREDITS  (2 * OCFS_INODE_UPDATE_CREDITS + 1               \
+			      + OCFS_LINK_CREDITS)
+
+/* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
+ * inode alloc group descriptor */
+#define OCFS_DELETE_INODE_CREDITS (3 * OCFS_INODE_UPDATE_CREDITS + 1 + 1)
+
+/* dinode update, old dir dinode update, new dir dinode update, old
+ * dir dir entry, new dir dir entry, dir entry update for renaming
+ * directory + target unlink */
+#define OCFS_RENAME_CREDITS (3 * OCFS_INODE_UPDATE_CREDITS + 3                \
+			     + OCFS_UNLINK_CREDITS)
+
+static inline int ocfs_calc_extend_credits(struct super_block *sb,
+					   ocfs2_dinode *fe,
+					   u32 bits_wanted)
+{
+	int bitmap_blocks, sysfile_bitmap_blocks, dinode_blocks;
+
+	/* bitmap dinode, group desc. + relinked group. */
+	bitmap_blocks = OCFS_SUBALLOC_ALLOC;
+
+	/* we might need to shift tree depth so lets assume an
+	 * absolute worst case of complete fragmentation.  Even with
+	 * that, we only need one update for the dinode, and then
+	 * however many metadata chunks needed * a remaining suballoc
+	 * alloc. */
+	sysfile_bitmap_blocks = 1 + 
+		(OCFS_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(fe);
+
+	/* this does not include *new* metadata blocks, which are
+	 * accounted for in sysfile_bitmap_blocks. fe +
+	 * prev. last_eb_blk + blocks along edge of tree. 
+	 * calc_symlink_credits passes because we just need 1 
+	 * credit for the dinode there. */
+	dinode_blocks = 1 + 1 + fe->id2.i_list.l_tree_depth;
+
+	return bitmap_blocks + sysfile_bitmap_blocks + dinode_blocks;
+}
+
+static inline int ocfs_calc_symlink_credits(struct super_block *sb)
+{
+	int blocks = OCFS_MKNOD_CREDITS;
+
+	/* links can be longer than one block so we may update many
+	 * within our single allocated extent. */
+	blocks += ocfs2_clusters_to_blocks(sb, 1);
+
+	return blocks;
+}
+
+static inline int ocfs_calc_group_alloc_credits(struct super_block *sb,
+						unsigned int cpg)
+{
+	int blocks;
+	int bitmap_blocks = OCFS_SUBALLOC_ALLOC + 1;
+	/* parent inode update + new block group header + bitmap inode update 
+	   + bitmap blocks affected */
+	blocks = 1 + 1 + 1 + bitmap_blocks;
+	return blocks;
+}
+
+static inline int ocfs_calc_tree_trunc_credits(struct super_block *sb,
+					       unsigned int clusters_to_del,
+					       ocfs2_dinode *fe,
+					       ocfs2_extent_list *last_el)
+{
+ 	/* for file entry + all headers in this pass + update to next leaf */
+	int credits = 1 + fe->id2.i_list.l_tree_depth + 1;
+	int bitmap_blocks, i;
+
+	i = last_el->l_next_free_rec - 1;
+	OCFS_ASSERT(i >= 0);
+
+	/* We may be deleting metadata blocks, so metadata alloc dinode +
+	   one desc. block for each possible delete. */
+	if (fe->id2.i_list.l_tree_depth 
+	    && (last_el->l_next_free_rec == 1)
+	    && ((last_el->l_recs[i].e_clusters - clusters_to_del) == 0))
+		credits += 1 + fe->id2.i_list.l_tree_depth;
+
+	/* bitmap fe + group descriptor */
+	bitmap_blocks = OCFS_SUBALLOC_FREE;
+
+	credits += bitmap_blocks;
+
+	return credits;
+}
+
 #endif /* OCFS2_JOURNAL_H */

Modified: trunk/fs/ocfs2/localalloc.c
===================================================================
--- trunk/fs/ocfs2/localalloc.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/localalloc.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -43,7 +43,7 @@
 #include "suballoc.h"
 #include "sysfile.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 /* The largest cluster size where we even consider using local alloc. */

Modified: trunk/fs/ocfs2/namei.c
===================================================================
--- trunk/fs/ocfs2/namei.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/namei.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -60,7 +60,7 @@
 #include "symlink.h"
 #include "vote.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_NAMEI

Deleted: trunk/fs/ocfs2/ocfs_journal.h
===================================================================
--- trunk/fs/ocfs2/ocfs_journal.h	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/ocfs_journal.h	2005-03-30 04:14:38 UTC (rev 2074)
@@ -1,442 +0,0 @@
-/* -*- mode: c; c-basic-offset: 8; -*-
- * vim: noexpandtab sw=8 ts=8 sts=0:
- *
- * journal.h
- *
- * Defines journalling api and structures.
- *
- * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
- *
- * 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.
- * 
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- * 
- * You should have received a copy of the GNU General Public
- * License along with this program; if not, write to the
- * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 021110-1307, USA.
- */
-
-#ifndef  _OCFSJOURNAL_H_
-#define  _OCFSJOURNAL_H_
-
-#include <linux/fs.h>
-#include <linux/jbd.h>
-
-#define OCFS_CHECKPOINT_INTERVAL        (8 * HZ)
-
-enum ocfs_journal_state {
-	OCFS_JOURNAL_FREE = 0,
-	OCFS_JOURNAL_LOADED,
-	OCFS_JOURNAL_IN_SHUTDOWN,
-};
-
-struct _ocfs_super;
-struct _ocfs2_dinode;
-struct _ocfs_journal_handle;
-
-typedef struct _ocfs_journal ocfs_journal;
-struct _ocfs_journal {
-	enum ocfs_journal_state   j_state;    /* Journals current state   */
-
-	journal_t                 *j_journal; /* The kernels journal type */
-	struct inode              *j_inode;   /* Kernel inode pointing to
-					       * this journal             */
-	struct _ocfs_super        *j_osb;     /* pointer to the super
-					       * block for the node
-					       * we're currently
-					       * running on -- not
-					       * necessarily the super
-					       * block from the node
-					       * which we usually run
-					       * from (recovery,
-					       * etc)                     */
-	struct buffer_head        *j_bh;      /* Journal disk inode block */
-	atomic_t                  j_num_trans; /* Number of transactions 
-					        * currently in the system. */
-	unsigned long             j_trans_id;
-	struct rw_semaphore       j_trans_barrier;
-	wait_queue_head_t         j_checkpointed;
-
-	spinlock_t                j_lock;
-	unsigned int              j_cleanup_orphans;
-	struct list_head          j_la_cleanups;
-	struct work_struct        j_recovery_work;
-};
-
-extern spinlock_t trans_inc_lock;
-
-/* wrap j_trans_id so we never have it equal to zero. */
-static inline unsigned long ocfs_inc_trans_id(ocfs_journal *j)
-{
-	unsigned long old_id;
-	spin_lock(&trans_inc_lock);
-	old_id = j->j_trans_id++;
-	if (!j->j_trans_id)
-		j->j_trans_id = 1;
-	spin_unlock(&trans_inc_lock);
-	return old_id;
-}
-
-static inline void ocfs_set_inode_lock_trans(ocfs_journal *journal,
-					     struct inode *inode)
-{
-	spin_lock(&trans_inc_lock);
-	OCFS_I(inode)->ip_last_trans = journal->j_trans_id;
-	spin_unlock(&trans_inc_lock);
-}
-
-/* Used to figure out whether it's safe to drop a metadata lock on an
- * inode. Returns true if all the inodes changes have been
- * checkpointed to disk. You should be holding the spinlock on the
- * metadata lock while calling this to be sure that nobody can take
- * the lock and put it on another transaction. */
-static inline int ocfs_inode_fully_checkpointed(struct inode *inode)
-{
-	int ret;
-	ocfs_journal *journal = OCFS2_SB(inode->i_sb)->journal;
-
-	spin_lock(&trans_inc_lock);
-	ret = time_after(journal->j_trans_id, OCFS_I(inode)->ip_last_trans);
-	spin_unlock(&trans_inc_lock);
-	return ret;
-}
-
-/* convenience function to check if an inode is still new (has never
- * hit disk) Will do you a favor and set created_trans = 0 when you've
- * been checkpointed.  returns '1' if the inode is still new. */
-static inline int ocfs_inode_is_new(struct inode *inode)
-{
-	int ret;
-
-	/* System files are never "new" as they're written out by
-	 * mkfs. This helps us early during mount, before we have the
-	 * journal open and j_trans_id could be junk. */
-	if (OCFS_I(inode)->ip_flags & OCFS_INODE_SYSTEM_FILE)
-		return 0;
-	spin_lock(&trans_inc_lock);
-	ret = !(time_after(OCFS2_SB(inode->i_sb)->journal->j_trans_id, 
-			   OCFS_I(inode)->ip_created_trans));
-	if (!ret)
-		OCFS_I(inode)->ip_created_trans = 0;
-	spin_unlock(&trans_inc_lock);
-	return ret;
-}
-
-static inline void ocfs_inode_set_new(ocfs_super *osb, 
-				      struct inode *inode)
-{
-	spin_lock(&trans_inc_lock);
-	OCFS_I(inode)->ip_created_trans = osb->journal->j_trans_id;
-	spin_unlock(&trans_inc_lock);
-}
-
-typedef struct _ocfs_journal_lock ocfs_journal_lock;
-struct _ocfs_journal_lock {
-	struct inode     *jl_inode;
-	struct list_head  jl_lock_list;
-};
-
-struct _ocfs_journal_handle {
-	handle_t            *k_handle; /* kernel handle.                */
-	ocfs_journal        *journal;
-	struct _ocfs_super  *osb;      /* what super block we belong to */
-	u32                 flags;     /* see flags below.              */
-	int                 max_buffs; /* Buffs reserved by this handle */
-
-	/* The following two fields are for ocfs_handle_add_lock */
-	int                 num_locks; 
-	struct list_head    locks;     /* A bunch of locks to 
-					* release on commit. This 
-					* should be a list_head */
-
-	struct list_head     inode_list;
-};
-
-#define OCFS_HANDLE_STARTED			1
-/* should we sync-commit this handle? */
-#define OCFS_HANDLE_SYNC			2
-static inline int ocfs_handle_started(ocfs_journal_handle *handle)
-{
-	return handle->flags & OCFS_HANDLE_STARTED;
-}
-
-static inline void ocfs_handle_set_sync(ocfs_journal_handle *handle, int sync)
-{
-	if (sync)
-		handle->flags |= OCFS_HANDLE_SYNC;
-	else
-		handle->flags &= ~OCFS_HANDLE_SYNC;
-}
-
-/* Exported only for the journal struct init code in super.c. Do not call. */
-void ocfs2_complete_recovery(void *data);
-
-/*
- *  Journal Control:
- *  Initialize, Load, Shutdown, Wipe a journal.
- *  
- *  ocfs_journal_init     - Initialize journal structures in the OSB.
- *  ocfs_journal_load     - Load the given journal off disk. Replay it if
- *                          there's transactions still in there.
- *  ocfs_journal_shutdown - Shutdown a journal, this will flush all 
- *                          uncommitted, uncheckpointed transactions.
- *  ocfs_journal_wipe     - Wipe transactions from a journal. Optionally 
- *                          zero out each block.
- *  ocfs_recovery_thread  - Perform recovery on a node. osb is our own osb.
- *  ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
- *                          event on.
- *  ocfs_start_checkpoint - Kick the commit thread to do a checkpoint.
- */
-int    ocfs_journal_init(ocfs_journal *journal, int *dirty);
-void   ocfs_journal_shutdown(struct _ocfs_super *osb);
-int    ocfs_journal_wipe(ocfs_journal *journal, int full);
-int    ocfs_journal_load(ocfs_journal *journal);
-void   ocfs_recovery_thread(struct _ocfs_super *osb, int node_num);
-int    ocfs2_mark_dead_nodes(ocfs_super *osb);
-static inline void ocfs_start_checkpoint(struct _ocfs_super *osb)
-{
-	atomic_set(&osb->needs_checkpoint, 1);
-	wake_up(&osb->checkpoint_event);
-}
-static inline void ocfs2_checkpoint_inode(struct inode *inode)
-{
-	ocfs_super *osb = OCFS_SB(inode->i_sb);
-
-	if (!ocfs_inode_fully_checkpointed(inode)) {
-		/* WARNING: This only kicks off a single
-		 * checkpoint. If someone races you and adds more
-		 * metadata to the journal, you won't know, and will
-		 * wind up waiting *alot* longer than necessary. Right
-		 * now we only use this in clear_inode so that's
-		 * OK. */
-		ocfs_start_checkpoint(osb);
-
-		wait_event(osb->journal->j_checkpointed,
-			   ocfs_inode_fully_checkpointed(inode));
-	}
-}
-
-/*
- *  Transaction Handling:
- *  Manage the lifetime of a transaction handle.
- *  
- *  ocfs_alloc_handle     - Only allocate a handle so we can start putting 
- *                          cluster locks on it. To actually change blocks, 
- *                          call ocfs_start_trans with the handle returned 
- *                          from this function. You may call ocfs_commit_trans
- *                           at any time in the lifetime of a handle.
- *  ocfs_start_trans      - Begin a transaction. Give it an upper estimate of 
- *                          the number of blocks that will be changed during 
- *                          this handle.
- *  ocfs_commit_trans     - Complete a handle.
- *  ocfs_extend_trans     - Extend a handle by nblocks credits. This may 
- *                          commit the handle to disk in the process, but will
- *                          not release any locks taken during the transaction.
- *  ocfs_journal_access   - Notify the handle that we want to journal this 
- *                          buffer. Will have to call ocfs_journal_dirty once
- *                          we've actually dirtied it. Type is one of . or .
- *  ocfs_journal_dirty    - Mark a journalled buffer as having dirty data.
- *  ocfs_handle_add_lock  - Sometimes we need to delay lock release
- *                          until after a transaction has been completed. Use
- *                          ocfs_handle_add_lock to indicate that a lock needs
- *                          to be released at the end of that handle. Locks 
- *                          will be released in the order that they are added. 
- *  ocfs_handle_add_inode - Add a locked inode to a transaction.
- */
-/* You must always start_trans with a number of buffs > 0, but it's
- * perfectly legal to go through an entire transaction without having
- * dirtied any buffers. */
-ocfs_journal_handle *ocfs_alloc_handle(ocfs_super *osb);
-ocfs_journal_handle *ocfs_start_trans(struct _ocfs_super *osb, 
-				      ocfs_journal_handle *handle, 
-				      int max_buffs);
-void                 ocfs_commit_trans(ocfs_journal_handle *handle);
-int                  ocfs_extend_trans(ocfs_journal_handle *handle, 
-				       int nblocks);
-
-/*
- * Create access is for when we get a newly created buffer and we're
- * not gonna read it off disk, but rather fill it ourselves.  Right
- * now, we don't do anything special with this (it turns into a write
- * request), but this is a good placeholder in case we do...
- *
- * Write access is for when we read a block off disk and are going to
- * modify it. This way the journalling layer knows it may need to make
- * a copy of that block (if it's part of another, uncommitted
- * transaction) before we do so.
- */
-#define OCFS_JOURNAL_ACCESS_CREATE 0
-#define OCFS_JOURNAL_ACCESS_WRITE  1
-#define OCFS_JOURNAL_ACCESS_UNDO   2
-int                  ocfs_journal_access(ocfs_journal_handle *handle, 
-					 struct inode *inode,
-					 struct buffer_head *bh, 
-					 int type);
-/*
- * A word about the journal_access/journal_dirty "dance". It is
- * entirely legal to journal_access a buffer more than once (as long
- * as the access type is the same -- I'm not sure what will happen if
- * access type is different but this should never happen anyway) It is
- * also legal to journal_dirty a buffer more than once. In fact, you
- * can even journal_access a buffer after you've done a
- * journal_access/journal_dirty pair. The only thing you cannot do
- * however, is journal_dirty a buffer which you haven't yet passed to
- * journal_access at least once.
- *
- * That said, 99% of the time this doesn't matter and this is what the
- * path looks like: 
- * 
- *	<read a bh>
- *	ocfs_journal_access(handle, bh,	OCFS_JOURNAL_ACCESS_WRITE); 
- *	<modify the bh>
- * 	ocfs_journal_dirty(handle, bh);
- */
-int                  ocfs_journal_dirty(ocfs_journal_handle *handle, 
-					struct buffer_head *bh);
-int                  ocfs_handle_add_lock(ocfs_journal_handle *handle,
-			                  struct inode *inode);
-/*
- * Use this to protect from other processes reading buffer state while
- * it's in flight.
- */
-void                 ocfs_handle_add_inode(ocfs_journal_handle *handle, 
-					   struct inode *inode);
-
-/*
- *  Credit Macros:
- *  Convenience macros to calculate number of credits needed.
- *
- *  For convenience sake, I have a set of macros here which calculate
- *  the *maximum* number of sectors which will be changed for various
- *  metadata updates.
- */
-
-/* simple file updates like chmod, etc. */
-#define OCFS_INODE_UPDATE_CREDITS 1
-
-/* get one bit out of a suballocator: dinode + group descriptor +
- * prev. group desc. if we relink. */
-#define OCFS_SUBALLOC_ALLOC (3)
-
-/* dinode + group descriptor update. We don't relink on free yet. */
-#define OCFS_SUBALLOC_FREE  (2)
-
-/* data block for new dir/symlink, 2 for bitmap updates (bitmap fe +
- * bitmap block for the new bit) */
-#define OCFS_DIR_LINK_ADDITIONAL_CREDITS (1 + 2)
-
-/* parent fe, parent block, new file entry, inode alloc fe, inode alloc
- * group descriptor + mkdir/symlink blocks */
-#define OCFS_MKNOD_CREDITS (3 + OCFS_SUBALLOC_ALLOC                           \
-			    + OCFS_DIR_LINK_ADDITIONAL_CREDITS)
-
-/* local alloc metadata change + main bitmap updates */
-#define OCFS_WINDOW_MOVE_CREDITS (OCFS_INODE_UPDATE_CREDITS                   \
-				  + OCFS_SUBALLOC_ALLOC + OCFS_SUBALLOC_FREE)
-
-/* used when we don't need an allocation change for a dir extend. One
- * for the dinode, one for the new block. */
-#define OCFS_SIMPLE_DIR_EXTEND_CREDITS (2)
-
-/* file update (nlink, etc) + dir entry block */
-#define OCFS_LINK_CREDITS  (OCFS_INODE_UPDATE_CREDITS + 1)
-
-/* inode + dir inode (if we unlink a dir), + dir entry block + orphan
- * dir inode link */
-#define OCFS_UNLINK_CREDITS  (2 * OCFS_INODE_UPDATE_CREDITS + 1               \
-			      + OCFS_LINK_CREDITS)
-
-/* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
- * inode alloc group descriptor */
-#define OCFS_DELETE_INODE_CREDITS (3 * OCFS_INODE_UPDATE_CREDITS + 1 + 1)
-
-/* dinode update, old dir dinode update, new dir dinode update, old
- * dir dir entry, new dir dir entry, dir entry update for renaming
- * directory + target unlink */
-#define OCFS_RENAME_CREDITS (3 * OCFS_INODE_UPDATE_CREDITS + 3                \
-			     + OCFS_UNLINK_CREDITS)
-
-static inline int ocfs_calc_extend_credits(struct super_block *sb,
-					   ocfs2_dinode *fe,
-					   u32 bits_wanted)
-{
-	int bitmap_blocks, sysfile_bitmap_blocks, dinode_blocks;
-
-	/* bitmap dinode, group desc. + relinked group. */
-	bitmap_blocks = OCFS_SUBALLOC_ALLOC;
-
-	/* we might need to shift tree depth so lets assume an
-	 * absolute worst case of complete fragmentation.  Even with
-	 * that, we only need one update for the dinode, and then
-	 * however many metadata chunks needed * a remaining suballoc
-	 * alloc. */
-	sysfile_bitmap_blocks = 1 + 
-		(OCFS_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(fe);
-
-	/* this does not include *new* metadata blocks, which are
-	 * accounted for in sysfile_bitmap_blocks. fe +
-	 * prev. last_eb_blk + blocks along edge of tree. 
-	 * calc_symlink_credits passes because we just need 1 
-	 * credit for the dinode there. */
-	dinode_blocks = 1 + 1 + fe->id2.i_list.l_tree_depth;
-
-	return bitmap_blocks + sysfile_bitmap_blocks + dinode_blocks;
-}
-
-static inline int ocfs_calc_symlink_credits(struct super_block *sb)
-{
-	int blocks = OCFS_MKNOD_CREDITS;
-
-	/* links can be longer than one block so we may update many
-	 * within our single allocated extent. */
-	blocks += ocfs2_clusters_to_blocks(sb, 1);
-
-	return blocks;
-}
-
-static inline int ocfs_calc_group_alloc_credits(struct super_block *sb,
-						unsigned int cpg)
-{
-	int blocks;
-	int bitmap_blocks = OCFS_SUBALLOC_ALLOC + 1;
-	/* parent inode update + new block group header + bitmap inode update 
-	   + bitmap blocks affected */
-	blocks = 1 + 1 + 1 + bitmap_blocks;
-	return blocks;
-}
-
-static inline int ocfs_calc_tree_trunc_credits(struct super_block *sb,
-					       unsigned int clusters_to_del,
-					       ocfs2_dinode *fe,
-					       ocfs2_extent_list *last_el)
-{
- 	/* for file entry + all headers in this pass + update to next leaf */
-	int credits = 1 + fe->id2.i_list.l_tree_depth + 1;
-	int bitmap_blocks, i;
-
-	i = last_el->l_next_free_rec - 1;
-	OCFS_ASSERT(i >= 0);
-
-	/* We may be deleting metadata blocks, so metadata alloc dinode +
-	   one desc. block for each possible delete. */
-	if (fe->id2.i_list.l_tree_depth 
-	    && (last_el->l_next_free_rec == 1)
-	    && ((last_el->l_recs[i].e_clusters - clusters_to_del) == 0))
-		credits += 1 + fe->id2.i_list.l_tree_depth;
-
-	/* bitmap fe + group descriptor */
-	bitmap_blocks = OCFS_SUBALLOC_FREE;
-
-	credits += bitmap_blocks;
-
-	return credits;
-}
-
-#endif /* _OCFSJOURNAL_H_ */

Modified: trunk/fs/ocfs2/proc.c
===================================================================
--- trunk/fs/ocfs2/proc.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/proc.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -40,7 +40,7 @@
 #include "proc.h"
 #include "alloc.h"
 #include "heartbeat.h"
-#include "ocfs_journal.h"
+#include "journal.h"
 
 
 /* Tracing */

Modified: trunk/fs/ocfs2/suballoc.c
===================================================================
--- trunk/fs/ocfs2/suballoc.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/suballoc.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -41,7 +41,7 @@
 #include "suballoc.h"
 #include "sysfile.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_CHAINALLOC

Modified: trunk/fs/ocfs2/super.c
===================================================================
--- trunk/fs/ocfs2/super.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/super.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -66,7 +66,7 @@
 #include "ver.h"
 #include "vote.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT  OCFS_DEBUG_CONTEXT_SUPER

Modified: trunk/fs/ocfs2/symlink.c
===================================================================
--- trunk/fs/ocfs2/symlink.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/symlink.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -49,7 +49,7 @@
 #include "inode.h"
 #include "symlink.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_SYMLINK

Modified: trunk/fs/ocfs2/sysfile.c
===================================================================
--- trunk/fs/ocfs2/sysfile.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/sysfile.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -39,7 +39,7 @@
 #include "inode.h"
 #include "sysfile.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 /* Tracing */

Modified: trunk/fs/ocfs2/vote.c
===================================================================
--- trunk/fs/ocfs2/vote.c	2005-03-30 00:09:23 UTC (rev 2073)
+++ trunk/fs/ocfs2/vote.c	2005-03-30 04:14:38 UTC (rev 2074)
@@ -49,7 +49,7 @@
 #include "slot_map.h"
 #include "vote.h"
 
-#include "ocfs_journal.h"
+#include "journal.h"
 #include "buffer_head_io.h"
 
 #define OCFS_DEBUG_CONTEXT    OCFS_DEBUG_CONTEXT_VOTE



More information about the Ocfs2-commits mailing list