* [Ocfs2-devel] [PATCH 3/4] Add group extend for online resize.take 3
2007-12-13 23:40 [Ocfs2-devel] [PATCH 0/4] Add online resize support for ocfs2, take 3 Tao Ma
2007-12-13 23:54 ` [Ocfs2-devel] [PATCH 1/4] Initalize bitmap_cpg of ocfs2_super to be the maximum,take 3 Tao Ma
2007-12-13 23:55 ` [Ocfs2-devel] [PATCH 2/4] Add new ioctl number for OCFS2.take 2 Tao Ma
@ 2007-12-13 23:57 ` Tao Ma
2007-12-17 14:10 ` Mark Fasheh
2007-12-13 23:58 ` [Ocfs2-devel] [PATCH 4/4] Implement "GROUP_ADD" " Tao Ma
3 siblings, 1 reply; 9+ messages in thread
From: Tao Ma @ 2007-12-13 23:57 UTC (permalink / raw)
To: ocfs2-devel
Modification from V2 to V3:
1. Modify the call sequence for ocfs2_commit_trans so that we will always
call it after a successful start_trans.
2 Add some rollback actions to "roll back" the modifications done to the
last group when we fail to change the inode.
3. Add the update of i_size of the inode.
4. Use the new defined ioctl numbers and "int" as the paramter.
5. Add the check of ocfs2_check_group_descriptor() for the last group
before we use it.
6. Some minor fix.
User can do offline resize using tunefs.ocfs2 when a volume isn't
mounted. Now the support for online resize is added into ocfs2.
Please note that the node where online resize goes must already
has the volume mounted. We don't mount it behind the user and the
operation would fail if we find it isn't mounted. As for other
nodes, we don't care whether the volume is mounted or not.
global_bitmap, super block and all the backups will be updated
in the kernel. And if super block or backup's update fails, we
just output some error message in dmesg and continue the work.
The whole process is derived from ext3 and divided into 2 steps:
1. If the last group isn't full, tunefs.ocfs2 will call
OCFS2_IOC_GROUP_EXTEND first to extend it. All the main work is
done in kernel.
2. For every new groups, tunefs.ocfs2 will call OCFS2_IOC_GROUP_ADD
to add them one by one. The new group descriptor is initialized
in userspace, we only check it in the kernel and update the
global_bitap, super blocks etc.
This patch includes the implementation for the 1st step.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/Makefile | 1 +
fs/ocfs2/buffer_head_io.c | 61 +++++++
fs/ocfs2/buffer_head_io.h | 2 +
fs/ocfs2/ioctl.c | 8 +
fs/ocfs2/journal.h | 3 +
fs/ocfs2/ocfs2_fs.h | 2 +
fs/ocfs2/resize.c | 398 +++++++++++++++++++++++++++++++++++++++++++++
fs/ocfs2/resize.h | 31 ++++
fs/ocfs2/suballoc.c | 11 +-
fs/ocfs2/suballoc.h | 8 +
10 files changed, 518 insertions(+), 7 deletions(-)
create mode 100644 fs/ocfs2/resize.c
create mode 100644 fs/ocfs2/resize.h
diff --git a/fs/ocfs2/Makefile b/fs/ocfs2/Makefile
index 9fb8132..4691544 100644
--- a/fs/ocfs2/Makefile
+++ b/fs/ocfs2/Makefile
@@ -21,6 +21,7 @@ ocfs2-objs := \
localalloc.o \
mmap.o \
namei.o \
+ resize.o \
slot_map.o \
suballoc.o \
super.o \
diff --git a/fs/ocfs2/buffer_head_io.c b/fs/ocfs2/buffer_head_io.c
index c903741..6eaa67f 100644
--- a/fs/ocfs2/buffer_head_io.c
+++ b/fs/ocfs2/buffer_head_io.c
@@ -280,3 +280,64 @@ bail:
mlog_exit(status);
return status;
}
+
+/* Check whether the blkno is the super block or one of the backups. */
+static inline void ocfs2_check_super_or_backup(struct super_block *sb,
+ sector_t blkno)
+{
+ int i;
+ u64 backup_blkno;
+
+ if (blkno == OCFS2_SUPER_BLOCK_BLKNO)
+ return;
+
+ for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
+ backup_blkno = ocfs2_backup_super_blkno(sb, i);
+ if (backup_blkno == blkno)
+ return;
+ }
+
+ BUG();
+}
+
+/*
+ * Write super block and bakcups doesn't need to collaborate with journal,
+ * so we don't need to lock ip_io_mutex and inode doesn't need to bea passed
+ * into this function.
+ */
+int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
+ struct buffer_head *bh)
+{
+ int ret = 0;
+
+ mlog_entry_void();
+
+ BUG_ON(buffer_jbd(bh));
+ ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);
+
+ if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
+ ret = -EROFS;
+ goto out;
+ }
+
+ lock_buffer(bh);
+ set_buffer_uptodate(bh);
+
+ /* remove from dirty list before I/O. */
+ clear_buffer_dirty(bh);
+
+ get_bh(bh); /* for end_buffer_write_sync() */
+ bh->b_end_io = end_buffer_write_sync;
+ submit_bh(WRITE, bh);
+
+ wait_on_buffer(bh);
+
+ if (!buffer_uptodate(bh)) {
+ ret = -EIO;
+ brelse(bh);
+ }
+
+out:
+ mlog_exit(ret);
+ return ret;
+}
diff --git a/fs/ocfs2/buffer_head_io.h b/fs/ocfs2/buffer_head_io.h
index 6cc2093..c2e7861 100644
--- a/fs/ocfs2/buffer_head_io.h
+++ b/fs/ocfs2/buffer_head_io.h
@@ -47,6 +47,8 @@ int ocfs2_read_blocks(struct ocfs2_super *osb,
int flags,
struct inode *inode);
+int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
+ struct buffer_head *bh);
#define OCFS2_BH_CACHED 1
#define OCFS2_BH_READAHEAD 8
diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
index 87dcece..03ee69b 100644
--- a/fs/ocfs2/ioctl.c
+++ b/fs/ocfs2/ioctl.c
@@ -20,6 +20,7 @@
#include "ocfs2_fs.h"
#include "ioctl.h"
+#include "resize.h"
#include <linux/ext2_fs.h>
@@ -115,6 +116,7 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
unsigned int cmd, unsigned long arg)
{
unsigned int flags;
+ int new_clusters;
int status;
struct ocfs2_space_resv sr;
@@ -140,6 +142,11 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
return -EFAULT;
return ocfs2_change_file_space(filp, cmd, &sr);
+ case OCFS2_IOC_GROUP_EXTEND:
+ if (get_user(new_clusters, (int __user *)arg))
+ return -EFAULT;
+
+ return ocfs2_group_extend(inode, new_clusters);
default:
return -ENOTTY;
}
@@ -162,6 +169,7 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
case OCFS2_IOC_RESVSP64:
case OCFS2_IOC_UNRESVSP:
case OCFS2_IOC_UNRESVSP64:
+ case OCFS2_IOC_GROUP_EXTEND:
break;
default:
return -ENOIOCTLCMD;
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index 4b32e09..0ba3a42 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -278,6 +278,9 @@ int ocfs2_journal_dirty_data(handle_t *handle,
/* simple file updates like chmod, etc. */
#define OCFS2_INODE_UPDATE_CREDITS 1
+/* group extend. inode update and last group update. */
+#define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
+
/* get one bit out of a suballocator: dinode + group descriptor +
* prev. group desc. if we relink. */
#define OCFS2_SUBALLOC_ALLOC (3)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 6ef8767..19ac421 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -231,6 +231,8 @@ struct ocfs2_space_resv {
#define OCFS2_IOC_RESVSP64 _IOW ('X', 42, struct ocfs2_space_resv)
#define OCFS2_IOC_UNRESVSP64 _IOW ('X', 43, struct ocfs2_space_resv)
+#define OCFS2_IOC_GROUP_EXTEND _IOW('o', 1, int)
+
/*
* Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
*/
diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c
new file mode 100644
index 0000000..e480639
--- /dev/null
+++ b/fs/ocfs2/resize.c
@@ -0,0 +1,398 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * resize.c
+ *
+ * volume resize.
+ * Inspired by ext3/resize.c.
+ *
+ * Copyright (C) 2007 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/types.h>
+
+#define MLOG_MASK_PREFIX ML_DISK_ALLOC
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "inode.h"
+#include "journal.h"
+#include "super.h"
+#include "sysfile.h"
+#include "uptodate.h"
+
+#include "buffer_head_io.h"
+#include "suballoc.h"
+#include "resize.h"
+
+/*
+ * Check whether there are new backup superblocks exist
+ * in the last group. If there are some, mark them or clear
+ * them in the bitmap.
+ *
+ * Return how many backups we find in the last group.
+ */
+static u16 ocfs2_calc_new_backup_super(struct inode *inode,
+ struct ocfs2_group_desc *gd,
+ int new_clusters,
+ u32 first_new_cluster,
+ u16 cl_cpg,
+ int set)
+{
+ int i;
+ u16 backups = 0;
+ u32 cluster;
+ u64 blkno, gd_blkno, lgd_blkno = le64_to_cpu(gd->bg_blkno);
+
+ for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
+ blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
+ cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
+
+ gd_blkno = ocfs2_which_cluster_group(inode, cluster);
+ if (gd_blkno < lgd_blkno)
+ continue;
+ else if (gd_blkno > lgd_blkno)
+ break;
+
+ if (set)
+ ocfs2_set_bit(cluster % cl_cpg,
+ (unsigned long *)gd->bg_bitmap);
+ else
+ ocfs2_clear_bit(cluster % cl_cpg,
+ (unsigned long *)gd->bg_bitmap);
+ backups++;
+ }
+
+ mlog_exit_void();
+ return backups;
+}
+
+static int ocfs2_update_last_group_and_inode(handle_t *handle,
+ struct inode *bm_inode,
+ struct buffer_head *bm_bh,
+ struct buffer_head *group_bh,
+ u32 first_new_cluster,
+ int new_clusters)
+{
+ int ret = 0;
+ struct ocfs2_super *osb = OCFS2_SB(bm_inode->i_sb);
+ struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bm_bh->b_data;
+ struct ocfs2_chain_list *cl = &fe->id2.i_chain;
+ struct ocfs2_chain_rec *cr;
+ struct ocfs2_group_desc *group;
+ u16 chain, num_bits, backups = 0;
+ u16 cl_bpc = le16_to_cpu(cl->cl_bpc);
+ u16 cl_cpg = le16_to_cpu(cl->cl_cpg);
+
+ mlog_entry("(new_clusters=%d, first_new_cluster = %u)\n",
+ new_clusters, first_new_cluster);
+
+ ret = ocfs2_journal_access(handle, bm_inode, group_bh,
+ OCFS2_JOURNAL_ACCESS_WRITE);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out;
+ }
+
+ group = (struct ocfs2_group_desc *) group_bh->b_data;
+
+ /* update the group first. */
+ num_bits = new_clusters * cl_bpc;
+ le16_add_cpu(&group->bg_bits, num_bits);
+ le16_add_cpu(&group->bg_free_bits_count, num_bits);
+
+ /*
+ * check whether there are some new backup superblocks exist in
+ * this group and update the group bitmap accordingly.
+ */
+ if (OCFS2_HAS_COMPAT_FEATURE(osb->sb,
+ OCFS2_FEATURE_COMPAT_BACKUP_SB)) {
+ backups = ocfs2_calc_new_backup_super(bm_inode,
+ group,
+ new_clusters,
+ first_new_cluster,
+ cl_cpg, 1);
+ le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
+ }
+
+ ret = ocfs2_journal_dirty(handle, group_bh);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_rollback;
+ }
+
+ /* update the inode accordingly. */
+ ret = ocfs2_journal_access(handle, bm_inode, bm_bh,
+ OCFS2_JOURNAL_ACCESS_WRITE);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_rollback;
+ }
+
+ chain = le16_to_cpu(group->bg_chain);
+ cr = (&cl->cl_recs[chain]);
+ le32_add_cpu(&cr->c_total, num_bits);
+ le32_add_cpu(&cr->c_free, num_bits);
+ le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits);
+ le32_add_cpu(&fe->i_clusters, new_clusters);
+
+ if (backups) {
+ le32_add_cpu(&cr->c_free, -1 * backups);
+ le32_add_cpu(&fe->id1.bitmap1.i_used, backups);
+ }
+
+ ocfs2_journal_dirty(handle, bm_bh);
+
+ spin_lock(&OCFS2_I(bm_inode)->ip_lock);
+ OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+ le64_add_cpu(&fe->i_size, new_clusters << osb->s_clustersize_bits);
+ spin_unlock(&OCFS2_I(bm_inode)->ip_lock);
+ i_size_write(bm_inode, le64_to_cpu(fe->i_size));
+
+out_rollback:
+ if (ret < 0) {
+ ocfs2_calc_new_backup_super(bm_inode,
+ group,
+ new_clusters,
+ first_new_cluster,
+ cl_cpg, 0);
+ le16_add_cpu(&group->bg_free_bits_count, backups);
+ le16_add_cpu(&group->bg_bits, -1 * num_bits);
+ le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits);
+ }
+out:
+ mlog_exit(ret);
+ return ret;
+}
+
+static int update_backups(struct inode * inode, u32 clusters, char *data)
+{
+ int i, ret = 0;
+ u32 cluster;
+ u64 blkno;
+ struct buffer_head *backup = NULL;
+ struct ocfs2_dinode *backup_di = NULL;
+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+ /* calculate the real backups we need to update. */
+ for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
+ blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
+ cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
+ if (cluster > clusters)
+ break;
+
+ ret = ocfs2_read_block(osb, blkno, &backup, 0, NULL);
+ if (ret < 0) {
+ mlog_errno(ret);
+ break;
+ }
+
+ memcpy(backup->b_data, data, inode->i_sb->s_blocksize);
+
+ backup_di = (struct ocfs2_dinode *)backup->b_data;
+ backup_di->i_blkno = cpu_to_le64(blkno);
+
+ ret = ocfs2_write_super_or_backup(osb, backup);
+ brelse(backup);
+ backup = NULL;
+ if (ret < 0) {
+ mlog_errno(ret);
+ break;
+ }
+ }
+
+ return ret;
+}
+
+static void ocfs2_update_super_and_backups(struct inode *inode,
+ int new_clusters)
+{
+ int ret;
+ u32 clusters = 0;
+ struct buffer_head *super_bh = NULL;
+ struct ocfs2_dinode *super_di = NULL;
+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+ /*
+ * update the superblock last.
+ * It doesn't matter if the write failed.
+ */
+ ret = ocfs2_read_block(osb, OCFS2_SUPER_BLOCK_BLKNO,
+ &super_bh, 0, NULL);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out;
+ }
+
+ super_di = (struct ocfs2_dinode *)super_bh->b_data;
+ le32_add_cpu(&super_di->i_clusters, new_clusters);
+ clusters = le32_to_cpu(super_di->i_clusters);
+
+ ret = ocfs2_write_super_or_backup(osb, super_bh);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out;
+ }
+
+ if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
+ ret = update_backups(inode, clusters, super_bh->b_data);
+
+out:
+ if (super_bh)
+ brelse(super_bh);
+ if (ret)
+ printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
+ " during fs resize. This condition is not fatal,"
+ " but fsck.ocfs2 should be run to fix it\n",
+ osb->dev_str);
+ return;
+}
+
+/*
+ * Extend the filesystem to the new number of clusters specified. This entry
+ * point is only used to extend the current filesystem to the end of the last
+ * existing group.
+ */
+int ocfs2_group_extend(struct inode * inode, int new_clusters)
+{
+ int ret;
+ handle_t *handle;
+ struct buffer_head *main_bm_bh = NULL;
+ struct buffer_head *group_bh = NULL;
+ struct inode *main_bm_inode = NULL;
+ struct ocfs2_dinode *fe = NULL;
+ struct ocfs2_group_desc *group = NULL;
+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+ u16 cl_bpc;
+ u32 first_new_cluster;
+ u64 lgd_blkno;
+
+ mlog_entry_void();
+
+ if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
+ return -EROFS;
+
+ if (new_clusters < 0)
+ return -EINVAL;
+ else if (new_clusters == 0)
+ return 0;
+
+ main_bm_inode = ocfs2_get_system_file_inode(osb,
+ GLOBAL_BITMAP_SYSTEM_INODE,
+ OCFS2_INVALID_SLOT);
+ if (!main_bm_inode) {
+ ret = -EINVAL;
+ mlog_errno(ret);
+ goto out;
+ }
+
+ mutex_lock(&main_bm_inode->i_mutex);
+
+ ret = ocfs2_meta_lock(main_bm_inode, &main_bm_bh, 1);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_mutex;
+ }
+
+ fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
+
+ if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
+ ocfs2_group_bitmap_size(osb->sb) * 8) {
+ mlog(0, "The disk is too old and small. "
+ "Force to do offline resize.");
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ if (!OCFS2_IS_VALID_DINODE(fe)) {
+ OCFS2_RO_ON_INVALID_DINODE(main_bm_inode->i_sb, fe);
+ ret = -EIO;
+ goto out_unlock;
+ }
+
+ first_new_cluster = le32_to_cpu(fe->i_clusters);
+ lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
+ first_new_cluster - 1);
+
+ ret = ocfs2_read_block(osb, lgd_blkno, &group_bh, OCFS2_BH_CACHED,
+ main_bm_inode);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_unlock;
+ }
+
+ group = (struct ocfs2_group_desc *) group_bh->b_data;
+
+ ret = ocfs2_check_group_descriptor(inode->i_sb, fe, group);
+ if (ret) {
+ mlog_errno(ret);
+ goto out_unlock;
+ }
+
+ cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
+ if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
+ le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ mlog(0, "extend the last group at %llu, new clusters = %d\n",
+ le64_to_cpu(group->bg_blkno), new_clusters);
+
+ handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
+ if (IS_ERR(handle)) {
+ mlog_errno(PTR_ERR(handle));
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ /* update the last group descriptor and inode. */
+ ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
+ main_bm_bh, group_bh,
+ first_new_cluster,
+ new_clusters);
+ if (ret) {
+ mlog_errno(ret);
+ goto out_commit;
+ }
+
+ ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
+
+out_commit:
+ ocfs2_commit_trans(osb, handle);
+out_unlock:
+ if (group_bh)
+ brelse(group_bh);
+
+ if (main_bm_bh)
+ brelse(main_bm_bh);
+
+ ocfs2_meta_unlock(main_bm_inode, 1);
+
+out_mutex:
+ mutex_unlock(&main_bm_inode->i_mutex);
+ iput(main_bm_inode);
+
+out:
+ mlog_exit_void();
+ return ret;
+}
diff --git a/fs/ocfs2/resize.h b/fs/ocfs2/resize.h
new file mode 100644
index 0000000..3acb79a
--- /dev/null
+++ b/fs/ocfs2/resize.h
@@ -0,0 +1,31 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * resize.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2007 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_RESIZE_H
+#define OCFS2_RESIZE_H
+
+int ocfs2_group_extend(struct inode * inode, int new_clusters);
+
+#endif /* OCFS2_RESIZE_H */
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 8f09f52..45060bf 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -101,8 +101,6 @@ static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg
static inline u32 ocfs2_desc_bitmap_to_cluster_off(struct inode *inode,
u64 bg_blkno,
u16 bg_bit_off);
-static inline u64 ocfs2_which_cluster_group(struct inode *inode,
- u32 cluster);
static inline void ocfs2_block_to_cluster_group(struct inode *inode,
u64 data_blkno,
u64 *bg_blkno,
@@ -131,9 +129,9 @@ static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl)
}
/* somewhat more expensive than our other checks, so use sparingly. */
-static int ocfs2_check_group_descriptor(struct super_block *sb,
- struct ocfs2_dinode *di,
- struct ocfs2_group_desc *gd)
+int ocfs2_check_group_descriptor(struct super_block *sb,
+ struct ocfs2_dinode *di,
+ struct ocfs2_group_desc *gd)
{
unsigned int max_bits;
@@ -1443,8 +1441,7 @@ static inline u32 ocfs2_desc_bitmap_to_cluster_off(struct inode *inode,
/* given a cluster offset, calculate which block group it belongs to
* and return that block offset. */
-static inline u64 ocfs2_which_cluster_group(struct inode *inode,
- u32 cluster)
+u64 ocfs2_which_cluster_group(struct inode *inode, u32 cluster)
{
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
u32 group_no;
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index cafe937..8799033 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -147,4 +147,12 @@ static inline int ocfs2_is_cluster_bitmap(struct inode *inode)
int ocfs2_reserve_cluster_bitmap_bits(struct ocfs2_super *osb,
struct ocfs2_alloc_context *ac);
+/* given a cluster offset, calculate which block group it belongs to
+ * and return that block offset. */
+u64 ocfs2_which_cluster_group(struct inode *inode, u32 cluster);
+
+/* somewhat more expensive than our other checks, so use sparingly. */
+int ocfs2_check_group_descriptor(struct super_block *sb,
+ struct ocfs2_dinode *di,
+ struct ocfs2_group_desc *gd);
#endif /* _CHAINALLOC_H_ */
--
gitgui.0.9.0.gd794
^ permalink raw reply related [flat|nested] 9+ messages in thread* [Ocfs2-devel] [PATCH 4/4] Implement "GROUP_ADD" for online resize.take 3
2007-12-13 23:40 [Ocfs2-devel] [PATCH 0/4] Add online resize support for ocfs2, take 3 Tao Ma
` (2 preceding siblings ...)
2007-12-13 23:57 ` [Ocfs2-devel] [PATCH 3/4] Add group extend for online resize.take 3 Tao Ma
@ 2007-12-13 23:58 ` Tao Ma
2007-12-14 16:08 ` Joel Becker
3 siblings, 1 reply; 9+ messages in thread
From: Tao Ma @ 2007-12-13 23:58 UTC (permalink / raw)
To: ocfs2-devel
Modification from V2 to V3:
1. Modify the call sequence for ocfs2_commit_trans so that we will always
call it after a successful start_trans.
2. As we need to update the group's bg_next_group in the kernel, move the group
read in the main function and add the writeout of this group also.
3. Add the update of i_size of the inode.
4. Use the new defined ioctl numbers.
User can do offline resize using tunefs.ocfs2 when a volume isn't
mounted. Now the support for online resize is added into ocfs2.
Please note that the node where online resize goes must already
has the volume mounted. We don't mount it behind the user and the
operation would fail if we find it isn't mounted. As for other
nodes, we don't care whether the volume is mounted or not.
global_bitmap, super block and all the backups will be updated
in the kernel. And if super block or backup's update fails, we
just output some error message in dmesg and continue the work.
The whole process is derived from ext3 and divided into 2 steps:
1. If the last group isn't full, tunefs.ocfs2 will call
OCFS2_IOC_GROUP_EXTEND first to extend it. All the main work is
done in kernel.
2. For every new groups, tunefs.ocfs2 will call OCFS2_IOC_GROUP_ADD
to add them one by one. The new group descriptor is initialized
in userspace, we only check it in the kernel and update the
global_bitap, super blocks etc.
This patch includes the implentation for the 2nd step.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ioctl.c | 9 ++
fs/ocfs2/journal.h | 3 +
fs/ocfs2/ocfs2_fs.h | 12 +++
fs/ocfs2/resize.c | 238 +++++++++++++++++++++++++++++++++++++++++++++++++++
fs/ocfs2/resize.h | 1 +
5 files changed, 263 insertions(+), 0 deletions(-)
diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
index 03ee69b..a4dda60 100644
--- a/fs/ocfs2/ioctl.c
+++ b/fs/ocfs2/ioctl.c
@@ -119,6 +119,7 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
int new_clusters;
int status;
struct ocfs2_space_resv sr;
+ struct ocfs2_new_group_input input;
switch (cmd) {
case OCFS2_IOC_GETFLAGS:
@@ -147,6 +148,12 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
return -EFAULT;
return ocfs2_group_extend(inode, new_clusters);
+ case OCFS2_IOC_GROUP_ADD:
+ case OCFS2_IOC_GROUP_ADD64:
+ if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
+ return -EFAULT;
+
+ return ocfs2_group_add(inode, &input);
default:
return -ENOTTY;
}
@@ -170,6 +177,8 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
case OCFS2_IOC_UNRESVSP:
case OCFS2_IOC_UNRESVSP64:
case OCFS2_IOC_GROUP_EXTEND:
+ case OCFS2_IOC_GROUP_ADD:
+ case OCFS2_IOC_GROUP_ADD64:
break;
default:
return -ENOIOCTLCMD;
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index 0ba3a42..220f3e8 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -281,6 +281,9 @@ int ocfs2_journal_dirty_data(handle_t *handle,
/* group extend. inode update and last group update. */
#define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
+/* group add. inode update and the new group update. */
+#define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
+
/* get one bit out of a suballocator: dinode + group descriptor +
* prev. group desc. if we relink. */
#define OCFS2_SUBALLOC_ALLOC (3)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 19ac421..4255517 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -231,7 +231,19 @@ struct ocfs2_space_resv {
#define OCFS2_IOC_RESVSP64 _IOW ('X', 42, struct ocfs2_space_resv)
#define OCFS2_IOC_UNRESVSP64 _IOW ('X', 43, struct ocfs2_space_resv)
+/* Used to pass group descriptor data when online resize is done */
+struct ocfs2_new_group_input {
+ __u64 group; /* Group descriptor's blkno. */
+ __u32 clusters; /* Total number of clusters in this group */
+ __u32 frees; /* Total free clusters in this group */
+ __u16 chain; /* Chain for this group */
+ __u16 reserved1;
+ __u32 reserved2;
+};
+
#define OCFS2_IOC_GROUP_EXTEND _IOW('o', 1, int)
+#define OCFS2_IOC_GROUP_ADD _IOW('o', 2,struct ocfs2_new_group_input)
+#define OCFS2_IOC_GROUP_ADD64 _IOW('o', 3,struct ocfs2_new_group_input)
/*
* Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c
index e480639..60721ba 100644
--- a/fs/ocfs2/resize.c
+++ b/fs/ocfs2/resize.c
@@ -396,3 +396,241 @@ out:
mlog_exit_void();
return ret;
}
+
+static int ocfs2_check_new_group(struct inode *inode,
+ struct ocfs2_dinode *di,
+ struct ocfs2_new_group_input *input,
+ struct buffer_head *group_bh)
+{
+ int ret;
+ struct ocfs2_group_desc *gd;
+ u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
+ unsigned int max_bits = le16_to_cpu(di->id2.i_chain.cl_cpg) *
+ le16_to_cpu(di->id2.i_chain.cl_bpc);
+
+
+ gd = (struct ocfs2_group_desc *)group_bh->b_data;
+
+ ret = -EIO;
+ if (!OCFS2_IS_VALID_GROUP_DESC(gd))
+ mlog(0, "Group descriptor # %llu isn't valid.\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno));
+ else if (di->i_blkno != gd->bg_parent_dinode)
+ mlog(0, "Group descriptor # %llu has bad parent "
+ "pointer (%llu, expected %llu)\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ (unsigned long long)le64_to_cpu(gd->bg_parent_dinode),
+ (unsigned long long)le64_to_cpu(di->i_blkno));
+ else if (le16_to_cpu(gd->bg_bits) > max_bits)
+ mlog(0, "Group descriptor # %llu has bit count of %u\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ le16_to_cpu(gd->bg_bits));
+ else if (le16_to_cpu(gd->bg_free_bits_count) > le16_to_cpu(gd->bg_bits))
+ mlog(0, "Group descriptor # %llu has bit count %u but "
+ "claims that %u are free\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ le16_to_cpu(gd->bg_bits),
+ le16_to_cpu(gd->bg_free_bits_count));
+ else if (le16_to_cpu(gd->bg_bits) > (8 * le16_to_cpu(gd->bg_size)))
+ mlog(0, "Group descriptor # %llu has bit count %u but "
+ "max bitmap bits of %u\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ le16_to_cpu(gd->bg_bits),
+ 8 * le16_to_cpu(gd->bg_size));
+ else if (le16_to_cpu(gd->bg_chain) != input->chain)
+ mlog(0, "Group descriptor # %llu has bad chain %u "
+ "while input has %u set.\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ le16_to_cpu(gd->bg_chain), input->chain);
+ else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
+ mlog(0, "Group descriptor # %llu has bit count %u but "
+ "input has %u clusters set\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ le16_to_cpu(gd->bg_bits), input->clusters);
+ else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
+ mlog(0, "Group descriptor # %llu has free bit count %u but "
+ "it should have %u set\n",
+ (unsigned long long)le64_to_cpu(gd->bg_blkno),
+ le16_to_cpu(gd->bg_bits),
+ input->frees * cl_bpc);
+ else
+ ret = 0;
+
+ return ret;
+}
+
+static int ocfs2_verify_group_and_input(struct inode *inode,
+ struct ocfs2_dinode *di,
+ struct ocfs2_new_group_input *input,
+ struct buffer_head *group_bh)
+{
+ u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
+ u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
+ u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
+ u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
+ u32 total_clusters = le32_to_cpu(di->i_clusters);
+ int ret = -EINVAL;
+
+ if (cluster < total_clusters)
+ mlog(0, "add a group which is in the current volume.\n");
+ else if (input->chain >= cl_count)
+ mlog(0, "input chain exceeds the limit.\n");
+ else if (next_free != cl_count && next_free != input->chain)
+ mlog(0, "the add group should be in chain %u\n", next_free);
+ else if (total_clusters + input->clusters < total_clusters)
+ mlog(0, "add group's clusters overflow.\n");
+ else if (input->clusters > cl_cpg)
+ mlog(0, "the cluster exceeds the maximum of a group\n");
+ else if (input->frees > input->clusters)
+ mlog(0, "the free cluster exceeds the total clusters\n");
+ else if (total_clusters % cl_cpg != 0)
+ mlog(0, "the last group isn't full. Use group extend first.\n");
+ else if (input->group != ocfs2_which_cluster_group(inode, cluster))
+ mlog(0, "group blkno is invalid\n");
+ else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
+ mlog(0, "group descriptor check failed.\n");
+ else
+ ret = 0;
+
+ return ret;
+}
+
+/* Add a new group descriptor to global_bitmap. */
+int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
+{
+ int ret;
+ handle_t *handle;
+ struct buffer_head *main_bm_bh = NULL;
+ struct inode *main_bm_inode = NULL;
+ struct ocfs2_dinode *fe = NULL;
+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+ struct buffer_head *group_bh = NULL;
+ struct ocfs2_group_desc *group = NULL;
+ struct ocfs2_chain_list *cl;
+ struct ocfs2_chain_rec *cr;
+ u16 cl_bpc;
+
+ mlog_entry_void();
+
+ if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
+ return -EROFS;
+
+ main_bm_inode = ocfs2_get_system_file_inode(osb,
+ GLOBAL_BITMAP_SYSTEM_INODE,
+ OCFS2_INVALID_SLOT);
+ if (!main_bm_inode) {
+ ret = -EINVAL;
+ mlog_errno(ret);
+ goto out;
+ }
+
+ mutex_lock(&main_bm_inode->i_mutex);
+
+ ret = ocfs2_meta_lock(main_bm_inode, &main_bm_bh, 1);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_mutex;
+ }
+
+ fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
+
+ if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
+ ocfs2_group_bitmap_size(osb->sb) * 8) {
+ mlog(0, "The disk is too old and small."
+ " Force to do offline resize.");
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ ret = ocfs2_read_block(osb, input->group, &group_bh, 0, NULL);
+ if (ret < 0)
+ goto out;
+
+ ocfs2_set_new_buffer_uptodate(inode, group_bh);
+
+ ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
+ if (ret) {
+ mlog_errno(ret);
+ goto out_unlock;
+ }
+
+ mlog(0, "Add a new group %llu in chain = %u, length = %u\n",
+ input->group, input->chain, input->clusters);
+
+ handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
+ if (IS_ERR(handle)) {
+ mlog_errno(PTR_ERR(handle));
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
+ cl = &fe->id2.i_chain;
+ cr = &cl->cl_recs[input->chain];
+
+ ret = ocfs2_journal_access(handle, main_bm_inode, group_bh,
+ OCFS2_JOURNAL_ACCESS_WRITE);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_commit;
+ }
+
+ group = (struct ocfs2_group_desc *)group_bh->b_data;
+ group->bg_next_group = cr->c_blkno;
+
+ ret = ocfs2_journal_dirty(handle, group_bh);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_commit;
+ }
+
+ ret = ocfs2_journal_access(handle, main_bm_inode, main_bm_bh,
+ OCFS2_JOURNAL_ACCESS_WRITE);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out_commit;
+ }
+
+ if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
+ le16_add_cpu(&cl->cl_next_free_rec, 1);
+ memset(cr, 0, sizeof(struct ocfs2_chain_rec));
+ }
+
+ cr->c_blkno = le64_to_cpu(input->group);
+ le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
+ le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
+
+ le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
+ le32_add_cpu(&fe->id1.bitmap1.i_used,
+ (input->clusters - input->frees) * cl_bpc);
+ le32_add_cpu(&fe->i_clusters, input->clusters);
+
+ ocfs2_journal_dirty(handle, main_bm_bh);
+
+ spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
+ OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+ le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits);
+ spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
+ i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
+
+ ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
+
+out_commit:
+ ocfs2_commit_trans(osb, handle);
+out_unlock:
+ if (group_bh)
+ brelse(group_bh);
+
+ if (main_bm_bh)
+ brelse(main_bm_bh);
+
+ ocfs2_meta_unlock(main_bm_inode, 1);
+
+out_mutex:
+ mutex_unlock(&main_bm_inode->i_mutex);
+ iput(main_bm_inode);
+
+out:
+ mlog_exit_void();
+ return ret;
+}
diff --git a/fs/ocfs2/resize.h b/fs/ocfs2/resize.h
index 3acb79a..f38841a 100644
--- a/fs/ocfs2/resize.h
+++ b/fs/ocfs2/resize.h
@@ -27,5 +27,6 @@
#define OCFS2_RESIZE_H
int ocfs2_group_extend(struct inode * inode, int new_clusters);
+int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input);
#endif /* OCFS2_RESIZE_H */
--
gitgui.0.9.0.gd794
^ permalink raw reply related [flat|nested] 9+ messages in thread