Index: linux-2.4.20-8/fs/ext3/snap.c
===================================================================
--- linux-2.4.20-8.orig/fs/ext3/snap.c 2003-01-30 18:24:37.000000000 +0800
-+++ linux-2.4.20-8/fs/ext3/snap.c 2004-01-13 00:11:40.000000000 +0800
-@@ -0,0 +1,2645 @@
++++ linux-2.4.20-8/fs/ext3/snap.c 2004-01-16 20:38:50.000000000 +0800
+@@ -0,0 +1,2644 @@
+/* fs/ext3/snap.c
+ *
+ * Copyright (c) 2002 Cluster File Systems, Inc. <info@clusterfs.com>
+#define EXT3_SNAP_INDEX EXT3_XATTR_INDEX_LUSTRE
+
+#ifdef EXT3_SNAP_DEBUG
-+ static long snap_kmem = 0;
+ #define snap_debug(f, a...) \
+ do { \
+ printk (KERN_INFO "SNAP DEBUG: (%s, %d): %s: ", \
+
+ /* we update the new cowed ino list end in memory */
+ SB_LAST_COWED_INO(pri->i_sb) = cpu_to_le32(pri->i_ino);
-+ snap_debug("cowed_inode_list_end %lu, append ino=%d\n",
-+ last_inode->i_ino, le32_to_cpu(snaps->ino[index]));
++ snap_debug("cowed_inode_list_end %lu, append ino=%lu\n",
++ last_inode->i_ino, pri->i_ino);
+exit:
+ if (last_inode)
+ iput(last_inode);
+
+ ext3_copy_meta(handle, dst, src);
+
-+ snap_debug("migrating %ld data blocks from %lu to %lu\n",
-+ blocks, src->i_ino, dst->i_ino);
++ snap_debug("migrating data blocks from %lu to %lu\n", src->i_ino, dst->i_ino);
+ /* Can't check blocks in case of EAs */
+ memcpy(EXT3_I(dst)->i_data, EXT3_I(src)->i_data,
+ sizeof(EXT3_I(src)->i_data));
+ ext3_mark_inode_dirty(handle, src);
+ ext3_mark_inode_dirty(handle, dst);
+
++ truncate_inode_pages(src->i_mapping, 0);
++
+ return SNAP_ERROR(err);
+}
+
+
+static int ext3_copy_block (struct inode *dst, struct inode *src, int blk)
+{
-+ struct buffer_head *bh_dst, *bh_src;
++ struct buffer_head *bh_dst = NULL, *bh_src = NULL;
++ struct page *dst_page = NULL, *src_page = NULL;
+ int err = 0;
+ int journal_data;
+ handle_t *handle = NULL;
+
-+ if (!ext3_bmap(src->i_mapping, blk))
-+ return 0;
-+
++
++ snap_debug("copy blk %d from %lu to %lu \n", blk, src->i_ino, dst->i_ino);
+ /*
+ * ext3_getblk() require handle!=NULL
+ */
+ journal_data = !S_ISREG(src->i_mode);
+
-+ handle = ext3_journal_start(dst, SNAP_COPYBLOCK_TRANS_BLOCKS);
-+ if( !handle )
-+ return -1;
++ if (journal_data) {
++ handle = ext3_journal_start(dst, SNAP_COPYBLOCK_TRANS_BLOCKS);
++ if( !handle )
++ return -1;
+
-+ bh_src = ext3_bread(handle, src, blk, 0, &err);
-+ if (!bh_src) {
-+ snap_err("error for src blk %d, error %d\n", blk, err);
-+ err = -1;
-+ goto exit_stop;
-+ }
-+ bh_dst = ext3_getblk(handle, dst, blk, 1, &err);
-+ if (!bh_dst) {
-+ snap_err("error for dst blk %d, error %d\n", blk, err);
-+ err = -1;
-+ goto exit_rels_src;
-+ }
-+ snap_debug("copy block %lu to %lu (%ld bytes)\n",
-+ bh_src->b_blocknr, bh_dst->b_blocknr,
-+ src->i_sb->s_blocksize);
++ bh_src = ext3_bread(handle, src, blk, 0, &err);
++ if (!bh_src) {
++ snap_err("error for src blk %d, error %d\n", blk, err);
++ goto exit_relese;
++ }
++ bh_dst = ext3_getblk(handle, dst, blk, 1, &err);
++ if (!bh_dst) {
++ snap_err("error for dst blk %d, error %d\n", blk, err);
++ err = -ENOSPC;
++ goto exit_relese;
++ }
++ snap_debug("copy block %lu to %lu (%ld bytes)\n",
++ bh_src->b_blocknr, bh_dst->b_blocknr,
++ src->i_sb->s_blocksize);
+
-+ if(journal_data){
+ ext3_journal_get_write_access(handle, bh_dst);
-+ }
+
-+ memcpy(bh_dst->b_data, bh_src->b_data, src->i_sb->s_blocksize);
++ memcpy(bh_dst->b_data, bh_src->b_data, src->i_sb->s_blocksize);
+
-+ if(journal_data){
+ ext3_journal_dirty_metadata(handle, bh_dst);
-+ }else{
-+ mark_buffer_dirty(bh_dst);
-+ if (IS_SYNC(src)) {
-+ ll_rw_block (WRITE, 1, &bh_dst);
-+ wait_on_buffer (bh_dst);
++ err = 1;
++exit_relese:
++ if (bh_src) brelse(bh_src);
++ if (bh_dst) brelse(bh_dst);
++ } else {
++ /*copy block of reg file*/
++ unsigned long bytes, index, offset;
++ struct address_space *mapping = src->i_mapping;
++
++ offset = ((blk >> dst->i_sb->s_blocksize_bits) & (PAGE_CACHE_SIZE -1));
++ index = (blk >> dst->i_sb->s_blocksize_bits) >> PAGE_CACHE_SHIFT;
++ bytes = dst->i_sb->s_blocksize;
++
++ src_page = grab_cache_page(src->i_mapping, index);
++ if (!src_page) {
++ snap_err("copy block %d from %lu to %lu ENOMEM \n",
++ blk, src->i_ino, dst->i_ino);
++ err = -ENOMEM;
++ goto exit;
++ }
++ ext3_readpage(NULL, src_page);
++ wait_on_page(src_page);
++
++ kmap(src_page);
++ if (!Page_Uptodate(src_page)) {
++ snap_err("Can not read page index %lu of inode %lu\n",
++ index, src->i_ino);
++ err = -EIO;
++ goto unlock_src_page;
+ }
++ dst_page = grab_cache_page(dst->i_mapping, index);
++ if (!dst_page) {
++ snap_err("copy block %d from %lu to %lu ENOMEM \n",
++ blk, src->i_ino, dst->i_ino);
++ err = -ENOMEM;
++ goto unlock_src_page;
++ }
++ kmap(dst_page);
++
++ err = mapping->a_ops->prepare_write(NULL, dst_page, offset, offset+bytes);
++ if (err)
++ goto unlock_dst_page;
++ memcpy(page_address(dst_page), page_address(src_page), bytes);
++ flush_dcache_page(dst_page);
++
++ err = mapping->a_ops->commit_write(NULL, dst_page, offset, offset+bytes);
++ if (err)
++ goto unlock_dst_page;
++ err = 1;
++unlock_dst_page:
++ kunmap(dst_page);
++ UnlockPage(dst_page);
++ page_cache_release(dst_page);
++unlock_src_page:
++ kunmap(src_page);
++ page_cache_release(src_page);
+ }
-+ brelse(bh_dst);
-+exit_rels_src:
-+ brelse(bh_src);
-+exit_stop:
-+ ext3_journal_stop(handle, dst);
++exit:
++ if (handle)
++ ext3_journal_stop(handle, dst);
+ return err;
+}
+
+{
+ struct inode *ind;
+ handle_t *handle = NULL;
-+ int err;
++ int err = 0;
+ int has_orphan = 0;
+
+ if( pri == pri->i_sb->u.ext3_sb.s_journal_inode ){
+ }
+
+ if (ind && !IS_ERR(ind)) {
-+ snap_err("existing indirect ino %lu for %lu: index %d\n",
-+ ind->i_ino, pri->i_ino, index);
++ snap_debug("existing indirect ino %lu for %lu: index %d\n",
++ ind->i_ino, pri->i_ino, index);
+ err = 0;
+ goto exit;
+ }
+ pri->i_ino, (ulong)le32_to_cpu(snaps->ino[index]), index);
+
+ ind = iget(pri->i_sb, le32_to_cpu (snaps->ino[index]) );
-+ snap_debug("iget ind %lu, ref count = %d\n", ind->i_ino, ind->i_count);
++ snap_debug("iget ind %lu, ref count = %d\n",
++ ind->i_ino, atomic_read(&ind->i_count));
+
+ if ( !ind || IS_ERR(ind) || is_bad_inode(ind) ) {
+ err = -EINVAL;
+ }
+
+ snap_debug("delete indirect ino %lu\n", ind->i_ino);
-+ snap_debug("iput ind %lu, ref count = %d\n", ind->i_ino, ind->i_count);
++ snap_debug("iput ind %lu, ref count = %d\n",
++ ind->i_ino, atomic_read(&ind->i_count));
+ ind->i_nlink = 0;
+ iput (ind);
+
+ }
+}
+
-+#if 1
+static int find_snap_meta_index(
+ struct table_snap_meta_data *snap_meta,
+ char *name)
+ for( i = 0; i < TABLE_ITEM_COUNT; i++){
+ /*compare name Max name Length 15*/
+ if (snap_meta->array[i].name[0]){
-+ if(strcmp(snap_meta->array[i].name, name))
-+ continue;
-+ return i;
++ if(!strncmp(snap_meta->array[i].name, name, strlen(name)))
++ return i;
+ }
+ }
+ return -1; /* can not find */
+ struct inode *inode;
+ struct buffer_head *bh = NULL;
+ struct table_snap_meta_data *s_attr;
-+ unsigned long map_len = 0, index = 0, left_size;
-+ int i, error = 0;
++ unsigned long map_len = 0, left_size;
++ int i, error = 0, index = 0;
+
+ ino = SB_SNAPTABLE_INO(sb);
+ if (ino == 0){
+ s_attr = (struct table_snap_meta_data *)(bh->b_data);
+ index = find_snap_meta_index(s_attr, name);
+ if (index < 0) {
-+ snap_debug("not exit %s meta attr of table ino %llu \n",
-+ name, ip->i_ino);
++ snap_debug("not exit %s meta attr of table ino %lu \n",
++ name, inode->i_ino);
+ error = 0;
+ goto out_iput;
+ }
+ s_attr = (struct table_snap_meta_data *)(bh->b_data);
+ index = find_snap_meta_index(s_attr, name);
+ if (index < 0 && !buf) {
-+ snap_debug("%s meta attr of table ino %llu do not exist\n",
++ snap_debug("%s meta attr of table ino %lu do not exist\n",
+ name, inode->i_ino);
+ error = 0;
++ brelse(bh);
+ goto exit;
+ }
+ if (!buf) {
-+ snap_debug("delete the meta attr %s in the table ino %s",
-+ *name, inode->ino);
++ snap_debug("delete the meta attr %s in the table ino %lu",
++ name, inode->i_ino);
+ /*Here we only delete the entry of the attr
+ *FIXME, should we also delete the block of
+ * this attr
+ s_attr->array[index].len = 0;
+ s_attr->count --;
+ ext3_journal_dirty_metadata(handle, bh);
++ brelse(bh);
+ goto exit;
+ }
+ new_len = (size + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
+ if (index < 0){
+ snap_err("table full of ino %lu \n", inode->i_ino);
+ error = index;
-+ goto exit;
++ brelse(bh);
++ goto exit;
+ }
+ }
+ s_attr->array[index].len = size;
+ journal_dirty_metadata(handle, bh);
++ brelse(bh);
+ /*put this attr to the snap table*/
+ left_size = size;
+ for(i = 0; i < new_len; i++) {
+ s_attr->array[index].start + i, 1, &error);
+ if (!array_bh) {
+ snap_err("inode %lu Can not get the block of attr %s\n",
-+ inode->i_ino, name);
++ inode->i_ino, name);
+ error = -ENOSPC;
++ brelse(array_bh);
+ goto exit;
+ }
+ ext3_journal_get_write_access(handle, array_bh);
+ left_size -= inode->i_sb->s_blocksize;
+ brelse(array_bh);
+ }
-+ ext3_journal_stop(handle, sb->s_root->d_inode);
+exit:
-+ brelse(bh);
++ if (handle)
++ ext3_journal_stop(handle, sb->s_root->d_inode);
+ iput(inode);
+ return error;
+}
-+#else
-+static int ext3_get_meta_attr(struct super_block *sb,
-+ char* name, char* buf,
-+ int *size)
-+{
-+ struct inode *root_inode = sb->s_root->d_inode;
-+ int err = 0;
-+
-+ if (!root_inode)
-+ return -EINVAL;
-+
-+ if (buf)
-+ err = ext3_xattr_get(root_inode, EXT3_SNAP_INDEX, name,
-+ buf, *size);
-+ if (err == -ERANGE || !buf){
-+ /*get the size of the buf*/
-+ *size = ext3_xattr_get(root_inode, EXT3_SNAP_INDEX, name,
-+ NULL, *size);
-+ }
-+
-+ return err;
-+}
-+/*
-+ * set the meta info of the snap system
-+ * argument:
-+ * buf == NULL delete "name" meta attr
-+ * != NULL set "name" meta attr
-+ * return value:
-+ * = 0 ok;
-+ * < 0 error;
-+ */
-+
-+static int ext3_set_meta_attr(struct super_block *sb, char* name,
-+ char* buf, int size)
-+{
-+ struct inode *root_inode = sb->s_root->d_inode;
-+ handle_t *handle = NULL;
-+ int error = 0;
-+
-+
-+ handle = ext3_journal_start(root_inode, EXT3_XATTR_TRANS_BLOCKS);
-+ if(!handle)
-+ return -EINVAL;
-+
-+ error = ext3_xattr_set(handle, root_inode, EXT3_SNAP_INDEX, name, buf, size, 0);
-+
-+ ext3_journal_stop(handle, root_inode);
-+
-+ return error;
-+}
-+#endif
+
+struct snapshot_operations ext3_snap_operations = {
+ ops_version: SNAP_VERSION(2,0,2),
+
+
+
-Index: linux-2.4.20-8/fs/ext3/ioctl.c
-===================================================================
---- linux-2.4.20-8.orig/fs/ext3/ioctl.c 2004-01-05 10:54:00.000000000 +0800
-+++ linux-2.4.20-8/fs/ext3/ioctl.c 2004-01-13 00:10:14.000000000 +0800
-@@ -13,6 +13,12 @@
- #include <linux/ext3_jbd.h>
- #include <linux/sched.h>
- #include <asm/uaccess.h>
-+#include <linux/locks.h>
-+
-+#include <linux/snap.h>
-+extern struct snapshot_operations ext3_snap_operations;
-+extern int ext3_snap_print(struct super_block *sb, int index);
-+extern int ext3_snap_delete(struct super_block *sb, int index);
-
-
- int ext3_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
-@@ -189,6 +195,103 @@
- return ret;
- }
- #endif
-+#if 0
-+ case EXT3_IOC_SNAP_SETFILECOW: {
-+ printk(KERN_INFO "set file cow on dev %x\n",inode->i_dev);
-+
-+ /* clear block cow feature*/
-+ if (EXT3_HAS_COMPAT_FEATURE(inode->i_sb,
-+ EXT3_FEATURE_COMPAT_BLOCKCOW)) {
-+ handle_t *handle = ext3_journal_start(inode, 1);
-+
-+ if (EXT3_HAS_COMPAT_FEATURE(inode->i_sb,
-+ EXT3_FEATURE_COMPAT_SNAPFS)) {
-+ printk(KERN_INFO "can't change cow level while snapfs feature exist");
-+ return -EPERM;
-+ }
-+ lock_super(inode->i_sb);
-+ journal_get_write_access(handle, EXT3_SB(inode->i_sb)->s_sbh);
-+ inode->i_sb->u.ext3_sb.s_es->s_feature_compat &=
-+ cpu_to_le32(~EXT3_FEATURE_COMPAT_BLOCKCOW);
-+ inode->i_sb->s_dirt = 1;
-+ journal_dirty_metadata(handle, EXT3_SB(inode->i_sb)->s_sbh);
-+ unlock_super(inode->i_sb);
-+ ext3_journal_stop(handle, inode);
-+ }
-+ return 0;
-+ }
-+ case EXT3_IOC_CREATE_INDIR: {
-+ struct inode *ind;
-+ printk(KERN_INFO "create indirect on inode %lu\n",inode->i_ino);
-+ ind = ext3_snap_operations.create_indirect(inode, 0, 1, 0);
-+ if (!ind || IS_ERR(ind))
-+ return PTR_ERR(ind);
-+ printk(KERN_INFO "got indirect inode %lu\n",ind->i_ino);
-+ put_user(ind->i_ino,(int *) arg);
-+ iput(ind);
-+ return 0;
-+ }
-+ case EXT3_IOC_GET_INDIR: {
-+ struct inode *ind;
-+ int index = 1;
-+ if (get_user(index, (int *) arg))
-+ return -EFAULT;
-+
-+ printk(KERN_INFO "get indirect on inode %lu, index %d\n",
-+ inode->i_ino, index);
-+ ind = ext3_snap_operations.get_indirect(inode, NULL, index);
-+ if (!ind || IS_ERR(ind)) {
-+ put_user(0,(int *) arg);
-+ return PTR_ERR(ind);
-+ }
-+ printk(KERN_INFO "got indirect inode %lu for index %d\n",
-+ ind->i_ino, index);
-+ put_user(ind->i_ino,(int *) arg);
-+ iput(ind);
-+ return 0;
-+ }
-+ case EXT3_IOC_IS_REDIR: {
-+ int is_redirector = 0;
-+ printk(KERN_INFO "checking if inode %lu is redirector via\n",
-+ inode->i_ino);
-+ is_redirector = ext3_snap_operations.is_redirector(inode);
-+ printk(KERN_INFO "redirector: %s\n",is_redirector ? "yes":"no");
-+ put_user(is_redirector,(int *) arg);
-+
-+ return 0;
-+ }
-+ case EXT3_IOC_RESTORE_INDIR: {
-+ printk(KERN_INFO "restore indirect on inode %lu\n",inode->i_ino);
-+ return ext3_snap_operations.restore_indirect(inode, 1);
-+ }
-+ case EXT3_IOC_SNAP_PRINT: {
-+ int index = 1;
-+ if (get_user(index, (int *) arg))
-+ return -EFAULT;
-+ printk(KERN_INFO "print snap for index %d\n",index);
-+
-+ return ext3_snap_print(inode->i_sb, 1);
-+ }
-+ case EXT3_IOC_SNAP_DELETE: {
-+ int index = 1;
-+ if (get_user(index, (int *) arg))
-+ return -EFAULT;
-+
-+ // XXX: debug code , always set index = 1
-+ if(index !=1) index=1;
-+ printk(KERN_INFO "delete all cowed inode for index %d\n",index);
-+ return ext3_snap_delete(inode->i_sb, index);
-+ }
-+
-+ case EXT3_IOC_DESTROY_INDIR: {
-+ int index = 1;
-+ if (get_user(index, (int *) arg))
-+ index = 1;
-+ printk(KERN_INFO "destroy indirect on ino %lu, index %d\n",
-+ inode->i_ino, index);
-+ return ext3_snap_operations.destroy_indirect(inode, index, NULL);
-+ }
-+#endif
- default:
- return -ENOTTY;
- }
Index: linux-2.4.20-8/fs/ext3/Makefile
===================================================================
--- linux-2.4.20-8.orig/fs/ext3/Makefile 2004-01-05 10:54:03.000000000 +0800
Index: linux-2.4.20-8/fs/ext3/inode.c
===================================================================
--- linux-2.4.20-8.orig/fs/ext3/inode.c 2004-01-05 10:54:03.000000000 +0800
-+++ linux-2.4.20-8/fs/ext3/inode.c 2004-01-13 00:10:14.000000000 +0800
++++ linux-2.4.20-8/fs/ext3/inode.c 2004-01-16 20:39:21.000000000 +0800
@@ -1191,7 +1191,7 @@
* So, if we see any bmap calls here on a modified, data-journaled file,
* take extra steps to flush any blocks which might be in the cache.
{
struct inode *inode = mapping->host;
journal_t *journal;
+@@ -1367,7 +1367,7 @@
+ return ret;
+ }
+
+-static int ext3_readpage(struct file *file, struct page *page)
++int ext3_readpage(struct file *file, struct page *page)
+ {
+ return block_read_full_page(page,ext3_get_block);
+ }
Index: linux-2.4.20-8/fs/ext3/ialloc.c
===================================================================
--- linux-2.4.20-8.orig/fs/ext3/ialloc.c 2004-01-05 10:54:03.000000000 +0800
-+++ linux-2.4.20-8/fs/ext3/ialloc.c 2004-01-13 00:10:14.000000000 +0800
-@@ -78,7 +78,6 @@
- sb->u.ext3_sb.s_inode_bitmap[bitmap_nr] = bh;
- return retval;
- }
--
- /*
- * load_inode_bitmap loads the inode bitmap for a blocks group
- *
-@@ -160,6 +159,13 @@
++++ linux-2.4.20-8/fs/ext3/ialloc.c 2004-01-16 20:48:29.000000000 +0800
+@@ -160,6 +160,13 @@
return retval;
}
Index: linux-2.4.20-8/include/linux/ext3_fs.h
===================================================================
--- linux-2.4.20-8.orig/include/linux/ext3_fs.h 2004-01-05 10:54:03.000000000 +0800
-+++ linux-2.4.20-8/include/linux/ext3_fs.h 2004-01-13 00:10:14.000000000 +0800
++++ linux-2.4.20-8/include/linux/ext3_fs.h 2004-01-16 20:45:09.000000000 +0800
@@ -183,7 +183,13 @@
#define EXT3_INDEX_FL 0x00001000 /* hash-indexed directory */
#define EXT3_IMAGIC_FL 0x00002000 /* AFS directory */
__u8 s_reserved_char_pad;
__u16 s_reserved_word_pad;
- __u32 s_reserved[192]; /* Padding to the end of the block */
-+ /* for snapfs */
+ __u32 s_default_mount_opts;
+ __u32 s_first_meta_bg; /* First metablock group */
+ __u32 s_mkfs_time; /* When the filesystem was created */
++ /* for snapfs */
+ __u32 s_first_cowed_pri_ino; /* For snapfs,the first cowed primary inode */
+ __u32 s_last_cowed_pri_ino; /* last cowed ino in memory */
+ __u32 s_snaptable_ino; /* snaptable ino in memory */
+ __u32 s_last_snap_orphan; /* SnapFS: start of cowing indirect inode */
-+ __u32 s_reserved[185]; /* Padding to the end of the block,originally 204 */
++ __u32 s_reserved[186]; /* Padding to the end of the block,originally 204 */
};
#ifdef __KERNEL__
%diffstat
fs/ext3/Makefile | 2
- fs/ext3/ialloc.c | 8
- fs/ext3/inode.c | 2
- fs/ext3/ioctl.c | 103 +
- fs/ext3/snap.c | 2645 +++++++++++++++++++++++++++++++++++++++++++++
+ fs/ext3/ialloc.c | 7
+ fs/ext3/inode.c | 4
+ fs/ext3/snap.c | 2644 +++++++++++++++++++++++++++++++++++++++++++++
include/linux/ext3_fs.h | 38
include/linux/ext3_fs_sb.h | 7
include/linux/ext3_jbd.h | 27
include/linux/snap.h | 266 ++++
- 9 files changed, 3092 insertions(+), 6 deletions(-)
+ 8 files changed, 2989 insertions(+), 6 deletions(-)