[PATCH] OCFS2: The Second Oracle Cluster Filesystem

The OCFS2 file system module.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>

Documentation/filesystems/00-INDEX

@@ -36,6 +36,8 @@ ntfs.txt
 	- info and mount options for the NTFS filesystem (Windows NT).
 proc.txt
 	- info on Linux's /proc filesystem.
+ocfs2.txt
+	- info and mount options for the OCFS2 clustered filesystem.
 romfs.txt
 	- Description of the ROMFS filesystem.
 smbfs.txt

Documentation/filesystems/ocfs2.txt

+OCFS2 filesystem
+==================
+OCFS2 is a general purpose extent based shared disk cluster file
+system with many similarities to ext3. It supports 64 bit inode
+numbers, and has automatically extending metadata groups which may
+also make it attractive for non-clustered use.
+
+You'll want to install the ocfs2-tools package in order to at least
+get "mount.ocfs2" and "ocfs2_hb_ctl".
+
+Project web page:    http://oss.oracle.com/projects/ocfs2
+Tools web page:      http://oss.oracle.com/projects/ocfs2-tools
+OCFS2 mailing lists: http://oss.oracle.com/projects/ocfs2/mailman/
+
+All code copyright 2005 Oracle except when otherwise noted.
+
+CREDITS:
+Lots of code taken from ext3 and other projects.
+
+Authors in alphabetical order:
+Joel Becker   <joel.becker@oracle.com>
+Zach Brown    <zach.brown@oracle.com>
+Mark Fasheh   <mark.fasheh@oracle.com>
+Kurt Hackel   <kurt.hackel@oracle.com>
+Sunil Mushran <sunil.mushran@oracle.com>
+Manish Singh  <manish.singh@oracle.com>
+
+Caveats
+=======
+Features which OCFS2 does not support yet:
+	- sparse files
+	- extended attributes
+	- shared writeable mmap
+	- loopback is supported, but data written will not
+	  be cluster coherent.
+	- quotas
+	- cluster aware flock
+	- Directory change notification (F_NOTIFY)
+	- Distributed Caching (F_SETLEASE/F_GETLEASE/break_lease)
+	- POSIX ACLs
+	- readpages / writepages (not user visible)
+
+Mount options
+=============
+
+OCFS2 supports the following mount options:
+(*) == default
+
+barrier=1		This enables/disables barriers. barrier=0 disables it,
+			barrier=1 enables it.
+errors=remount-ro(*)	Remount the filesystem read-only on an error.
+errors=panic		Panic and halt the machine if an error occurs.
+intr		(*)	Allow signals to interrupt cluster operations.
+nointr			Do not allow signals to interrupt cluster
+			operations.

MAINTAINERS

@@ -1905,6 +1905,15 @@ M:	ajoshi@shell.unixbox.com
 L:	linux-nvidia@lists.surfsouth.com
 S:	Maintained
 
+ORACLE CLUSTER FILESYSTEM 2 (OCFS2)
+P:	Mark Fasheh
+M:	mark.fasheh@oracle.com
+P:	Kurt Hackel
+M:	kurt.hackel@oracle.com
+L:	ocfs2-devel@oss.oracle.com
+W:	http://oss.oracle.com/projects/ocfs2/
+S:	Supported	
+
 OLYMPIC NETWORK DRIVER
 P:	Peter De Shrijver
 M:	p2@ace.ulyssis.student.kuleuven.ac.be

fs/ocfs2/Makefile

+EXTRA_CFLAGS += -Ifs/ocfs2
+
+EXTRA_CFLAGS += -DCATCH_BH_JBD_RACES
+
+obj-$(CONFIG_OCFS2_FS) += ocfs2.o
+
+ocfs2-objs := \
+	alloc.o 		\
+	aops.o 			\
+	buffer_head_io.o	\
+	dcache.o 		\
+	dir.o 			\
+	dlmglue.o 		\
+	export.o 		\
+	extent_map.o 		\
+	file.o 			\
+	heartbeat.o 		\
+	inode.o 		\
+	journal.o 		\
+	localalloc.o 		\
+	mmap.o 			\
+	namei.o 		\
+	slot_map.o 		\
+	suballoc.o 		\
+	super.o 		\
+	symlink.o 		\
+	sysfile.o 		\
+	uptodate.o		\
+	ver.o 			\
+	vote.o
+
+obj-$(CONFIG_OCFS2_FS) += cluster/
+obj-$(CONFIG_OCFS2_FS) += dlm/

fs/ocfs2/alloc.h

+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * alloc.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 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 OCFS2_ALLOC_H
+#define OCFS2_ALLOC_H
+
+struct ocfs2_alloc_context;
+int ocfs2_insert_extent(struct ocfs2_super *osb,
+			struct ocfs2_journal_handle *handle,
+			struct inode *inode,
+			struct buffer_head *fe_bh,
+			u64 blkno,
+			u32 new_clusters,
+			struct ocfs2_alloc_context *meta_ac);
+int ocfs2_num_free_extents(struct ocfs2_super *osb,
+			   struct inode *inode,
+			   struct ocfs2_dinode *fe);
+/* how many new metadata chunks would an allocation need at maximum? */
+static inline int ocfs2_extend_meta_needed(struct ocfs2_dinode *fe)
+{
+	/*
+	 * Rather than do all the work of determining how much we need
+	 * (involves a ton of reads and locks), just ask for the
+	 * maximal limit.  That's a tree depth shift.  So, one block for
+	 * level of the tree (current l_tree_depth), one block for the
+	 * new tree_depth==0 extent_block, and one block at the new
+	 * top-of-the tree.
+	 */
+	return le16_to_cpu(fe->id2.i_list.l_tree_depth) + 2;
+}
+
+int ocfs2_truncate_log_init(struct ocfs2_super *osb);
+void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
+void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
+				       int cancel);
+int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
+int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
+				      int slot_num,
+				      struct ocfs2_dinode **tl_copy);
+int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
+					 struct ocfs2_dinode *tl_copy);
+
+struct ocfs2_truncate_context {
+	struct inode *tc_ext_alloc_inode;
+	struct buffer_head *tc_ext_alloc_bh;
+	int tc_ext_alloc_locked; /* is it cluster locked? */
+	/* these get destroyed once it's passed to ocfs2_commit_truncate. */
+	struct buffer_head *tc_last_eb_bh;
+};
+
+int ocfs2_prepare_truncate(struct ocfs2_super *osb,
+			   struct inode *inode,
+			   struct buffer_head *fe_bh,
+			   struct ocfs2_truncate_context **tc);
+int ocfs2_commit_truncate(struct ocfs2_super *osb,
+			  struct inode *inode,
+			  struct buffer_head *fe_bh,
+			  struct ocfs2_truncate_context *tc);
+
+#endif /* OCFS2_ALLOC_H */

fs/ocfs2/aops.c

+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2002, 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.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <asm/byteorder.h>
+
+#define MLOG_MASK_PREFIX ML_FILE_IO
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "aops.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "super.h"
+#include "symlink.h"
+
+#include "buffer_head_io.h"
+
+static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
+				   struct buffer_head *bh_result, int create)
+{
+	int err = -EIO;
+	int status;
+	struct ocfs2_dinode *fe = NULL;
+	struct buffer_head *bh = NULL;
+	struct buffer_head *buffer_cache_bh = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	void *kaddr;
+
+	mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
+		   (unsigned long long)iblock, bh_result, create);
+
+	BUG_ON(ocfs2_inode_is_fast_symlink(inode));
+
+	if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
+		mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
+		     (unsigned long long)iblock);
+		goto bail;
+	}
+
+	status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+				  OCFS2_I(inode)->ip_blkno,
+				  &bh, OCFS2_BH_CACHED, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	fe = (struct ocfs2_dinode *) bh->b_data;
+
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		mlog(ML_ERROR, "Invalid dinode #%"MLFu64": signature = %.*s\n",
+		     fe->i_blkno, 7, fe->i_signature);
+		goto bail;
+	}
+
+	if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
+						    le32_to_cpu(fe->i_clusters))) {
+		mlog(ML_ERROR, "block offset is outside the allocated size: "
+		     "%llu\n", (unsigned long long)iblock);
+		goto bail;
+	}
+
+	/* We don't use the page cache to create symlink data, so if
+	 * need be, copy it over from the buffer cache. */
+	if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
+		u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
+			    iblock;
+		buffer_cache_bh = sb_getblk(osb->sb, blkno);
+		if (!buffer_cache_bh) {
+			mlog(ML_ERROR, "couldn't getblock for symlink!\n");
+			goto bail;
+		}
+
+		/* we haven't locked out transactions, so a commit
+		 * could've happened. Since we've got a reference on
+		 * the bh, even if it commits while we're doing the
+		 * copy, the data is still good. */
+		if (buffer_jbd(buffer_cache_bh)
+		    && ocfs2_inode_is_new(inode)) {
+			kaddr = kmap_atomic(bh_result->b_page, KM_USER0);
+			if (!kaddr) {
+				mlog(ML_ERROR, "couldn't kmap!\n");
+				goto bail;
+			}
+			memcpy(kaddr + (bh_result->b_size * iblock),
+			       buffer_cache_bh->b_data,
+			       bh_result->b_size);
+			kunmap_atomic(kaddr, KM_USER0);
+			set_buffer_uptodate(bh_result);
+		}
+		brelse(buffer_cache_bh);
+	}
+
+	map_bh(bh_result, inode->i_sb,
+	       le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);
+
+	err = 0;
+
+bail:
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(err);
+	return err;
+}
+
+static int ocfs2_get_block(struct inode *inode, sector_t iblock,
+			   struct buffer_head *bh_result, int create)
+{
+	int err = 0;
+	u64 p_blkno, past_eof;
+
+	mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
+		   (unsigned long long)iblock, bh_result, create);
+
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
+		mlog(ML_NOTICE, "get_block on system inode 0x%p (%lu)\n",
+		     inode, inode->i_ino);
+
+	if (S_ISLNK(inode->i_mode)) {
+		/* this always does I/O for some reason. */
+		err = ocfs2_symlink_get_block(inode, iblock, bh_result, create);
+		goto bail;
+	}
+
+	/* this can happen if another node truncs after our extend! */
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	if (iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
+					       OCFS2_I(inode)->ip_clusters))
+		err = -EIO;
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+	if (err)
+		goto bail;
+
+	err = ocfs2_extent_map_get_blocks(inode, iblock, 1, &p_blkno,
+					  NULL);
+	if (err) {
+		mlog(ML_ERROR, "Error %d from get_blocks(0x%p, %llu, 1, "
+		     "%"MLFu64", NULL)\n", err, inode,
+		     (unsigned long long)iblock, p_blkno);
+		goto bail;
+	}
+
+	map_bh(bh_result, inode->i_sb, p_blkno);
+
+	if (bh_result->b_blocknr == 0) {
+		err = -EIO;
+		mlog(ML_ERROR, "iblock = %llu p_blkno = %"MLFu64" "
+		     "blkno=(%"MLFu64")\n", (unsigned long long)iblock,
+		     p_blkno, OCFS2_I(inode)->ip_blkno);
+	}
+
+	past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
+	mlog(0, "Inode %lu, past_eof = %"MLFu64"\n", inode->i_ino, past_eof);
+
+	if (create && (iblock >= past_eof))
+		set_buffer_new(bh_result);
+
+bail:
+	if (err < 0)
+		err = -EIO;
+
+	mlog_exit(err);
+	return err;
+}
+
+static int ocfs2_readpage(struct file *file, struct page *page)
+{
+	struct inode *inode = page->mapping->host;
+	loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT;
+	int ret, unlock = 1;
+
+	mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0));
+
+	ret = ocfs2_meta_lock_with_page(inode, NULL, NULL, 0, page);
+	if (ret != 0) {
+		if (ret == AOP_TRUNCATED_PAGE)
+			unlock = 0;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	/*
+	 * i_size might have just been updated as we grabed the meta lock.  We
+	 * might now be discovering a truncate that hit on another node.
+	 * block_read_full_page->get_block freaks out if it is asked to read
+	 * beyond the end of a file, so we check here.  Callers
+	 * (generic_file_read, fault->nopage) are clever enough to check i_size
+	 * and notice that the page they just read isn't needed.
+	 *
+	 * XXX sys_readahead() seems to get that wrong?
+	 */
+	if (start >= i_size_read(inode)) {
+		char *addr = kmap(page);
+		memset(addr, 0, PAGE_SIZE);
+		flush_dcache_page(page);
+		kunmap(page);
+		SetPageUptodate(page);
+		ret = 0;
+		goto out_alloc;
+	}
+
+	ret = ocfs2_data_lock_with_page(inode, 0, page);
+	if (ret != 0) {
+		if (ret == AOP_TRUNCATED_PAGE)
+			unlock = 0;
+		mlog_errno(ret);
+		goto out_alloc;
+	}
+
+	ret = block_read_full_page(page, ocfs2_get_block);
+	unlock = 0;
+
+	ocfs2_data_unlock(inode, 0);
+out_alloc:
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+	ocfs2_meta_unlock(inode, 0);
+out:
+	if (unlock)
+		unlock_page(page);
+	mlog_exit(ret);
+	return ret;
+}
+
+/* Note: Because we don't support holes, our allocation has
+ * already happened (allocation writes zeros to the file data)
+ * so we don't have to worry about ordered writes in
+ * ocfs2_writepage.
+ *
+ * ->writepage is called during the process of invalidating the page cache
+ * during blocked lock processing.  It can't block on any cluster locks
+ * to during block mapping.  It's relying on the fact that the block
+ * mapping can't have disappeared under the dirty pages that it is
+ * being asked to write back.
+ */
+static int ocfs2_writepage(struct page *page, struct writeback_control *wbc)
+{
+	int ret;
+
+	mlog_entry("(0x%p)\n", page);
+
+	ret = block_write_full_page(page, ocfs2_get_block, wbc);
+
+	mlog_exit(ret);
+
+	return ret;
+}
+
+/*
+ * ocfs2_prepare_write() can be an outer-most ocfs2 call when it is called
+ * from loopback.  It must be able to perform its own locking around
+ * ocfs2_get_block().
+ */
+int ocfs2_prepare_write(struct file *file, struct page *page,
+			unsigned from, unsigned to)
+{
+	struct inode *inode = page->mapping->host;
+	int ret;
+
+	mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to);
+
+	ret = ocfs2_meta_lock_with_page(inode, NULL, NULL, 0, page);
+	if (ret != 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ret = block_prepare_write(page, from, to, ocfs2_get_block);
+
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ocfs2_meta_unlock(inode, 0);
+out:
+	mlog_exit(ret);
+	return ret;
+}
+
+/* Taken from ext3. We don't necessarily need the full blown
+ * functionality yet, but IMHO it's better to cut and paste the whole
+ * thing so we can avoid introducing our own bugs (and easily pick up
+ * their fixes when they happen) --Mark */
+static int walk_page_buffers(	handle_t *handle,
+				struct buffer_head *head,
+				unsigned from,
+				unsigned to,
+				int *partial,
+				int (*fn)(	handle_t *handle,
+						struct buffer_head *bh))
+{
+	struct buffer_head *bh;
+	unsigned block_start, block_end;
+	unsigned blocksize = head->b_size;
+	int err, ret = 0;
+	struct buffer_head *next;
+
+	for (	bh = head, block_start = 0;
+		ret == 0 && (bh != head || !block_start);
+	    	block_start = block_end, bh = next)
+	{
+		next = bh->b_this_page;
+		block_end = block_start + blocksize;
+		if (block_end <= from || block_start >= to) {
+			if (partial && !buffer_uptodate(bh))
+				*partial = 1;
+			continue;
+		}
+		err = (*fn)(handle, bh);
+		if (!ret)
+			ret = err;
+	}
+	return ret;
+}
+
+struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode,
+							 struct page *page,
+							 unsigned from,
+							 unsigned to)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_journal_handle *handle = NULL;
+	int ret = 0;
+
+	handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS);
+	if (!handle) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	if (ocfs2_should_order_data(inode)) {
+		ret = walk_page_buffers(handle->k_handle,
+					page_buffers(page),
+					from, to, NULL,
+					ocfs2_journal_dirty_data);
+		if (ret < 0) 
+			mlog_errno(ret);
+	}
+out:
+	if (ret) {
+		if (handle)
+			ocfs2_commit_trans(handle);
+		handle = ERR_PTR(ret);
+	}
+	return handle;
+}
+
+static int ocfs2_commit_write(struct file *file, struct page *page,
+			      unsigned from, unsigned to)
+{
+	int ret, extending = 0, locklevel = 0;
+	loff_t new_i_size;
+	struct buffer_head *di_bh = NULL;
+	struct inode *inode = page->mapping->host;
+	struct ocfs2_journal_handle *handle = NULL;
+
+	mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to);
+
+	/* NOTE: ocfs2_file_aio_write has ensured that it's safe for
+	 * us to sample inode->i_size here without the metadata lock:
+	 *
+	 * 1) We're currently holding the inode alloc lock, so no
+	 *    nodes can change it underneath us.
+	 *
+	 * 2) We've had to take the metadata lock at least once
+	 *    already to check for extending writes, hence insuring
+	 *    that our current copy is also up to date.
+	 */
+	new_i_size = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
+	if (new_i_size > i_size_read(inode)) {
+		extending = 1;
+		locklevel = 1;
+	}
+
+	ret = ocfs2_meta_lock_with_page(inode, NULL, &di_bh, locklevel, page);
+	if (ret != 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_data_lock_with_page(inode, 1, page);
+	if (ret != 0) {
+		mlog_errno(ret);
+		goto out_unlock_meta;
+	}
+
+	if (extending) {
+		handle = ocfs2_start_walk_page_trans(inode, page, from, to);
+		if (IS_ERR(handle)) {
+			ret = PTR_ERR(handle);
+			handle = NULL;
+			goto out_unlock_data;
+		}
+
+		/* Mark our buffer early. We'd rather catch this error up here
+		 * as opposed to after a successful commit_write which would
+		 * require us to set back inode->i_size. */
+		ret = ocfs2_journal_access(handle, inode, di_bh,
+					   OCFS2_JOURNAL_ACCESS_WRITE);