ext3-uninit-2.6.9.patch

非常经典的一个分布式系统

 
 got:
-	ino += group * EXT3_INODES_PER_GROUP(sb) + 1;
-	if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
-		ext3_error (sb, "ext3_new_inode",
-			    "reserved inode or inode > inodes count - "
-			    "block_group = %d, inode=%lu", group, ino);
+	ino++;
+	if ((group == 0 && ino < EXT3_FIRST_INO(sb)) ||
+	    ino > EXT3_INODES_PER_GROUP(sb)) {
+		ext3_error(sb, __FUNCTION__,
+			   "reserved inode or inode > inodes count - "
+			   "block_group = %d, inode=%lu", group,
+			   ino + group * EXT3_INODES_PER_GROUP(sb));
 		err = -EIO;
 		goto fail;
 	}
@@ -578,13 +640,65 @@ got:
 	BUFFER_TRACE(bh2, "get_write_access");
 	err = ext3_journal_get_write_access(handle, bh2);
 	if (err) goto fail;
+
+	/* We may have to initialize the block bitmap if it isn't already */
+	if (EXT3_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM) &&
+	    gdp->bg_flags & cpu_to_le16(EXT3_BG_BLOCK_UNINIT)) {
+		struct buffer_head *block_bh = read_block_bitmap(sb, group);
+
+		BUFFER_TRACE(block_bh, "get block bitmap access");
+		err = ext3_journal_get_write_access(handle, block_bh);
+		if (err) {
+			brelse(block_bh);
+			goto fail;
+		}
+
+		free = 0;
+		spin_lock(sb_bgl_lock(sbi, group));
+		/* recheck and clear flag under lock if we still need to */
+		if (gdp->bg_flags & cpu_to_le16(EXT3_BG_BLOCK_UNINIT)) {
+			gdp->bg_flags &= cpu_to_le16(~EXT3_BG_BLOCK_UNINIT);
+			free = ext3_free_blocks_after_init(sb, group, gdp);
+			gdp->bg_free_blocks_count = cpu_to_le16(free);
+		}
+		spin_unlock(sb_bgl_lock(sbi, group));
+
+		/* Don't need to dirty bitmap block if we didn't change it */
+		if (free) {
+			BUFFER_TRACE(block_bh, "dirty block bitmap");
+			err = ext3_journal_dirty_metadata(handle, block_bh);
+		}
+
+		brelse(block_bh);
+		if (err)
+			goto fail;
+	}
+
 	spin_lock(sb_bgl_lock(sbi, group));
+	/* If we didn't allocate from within the initialized part of the inode
+	 * table then we need to initialize up to this inode. */
+	if (EXT3_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
+		if (gdp->bg_flags & cpu_to_le16(EXT3_BG_INODE_UNINIT)) {
+			gdp->bg_flags &= cpu_to_le16(~EXT3_BG_INODE_UNINIT);
+			free = 0;
+		} else {
+			free = EXT3_INODES_PER_GROUP(sb) -
+				le16_to_cpu(gdp->bg_itable_unused);
+		}
+
+		if (ino > free) {
+			gdp->bg_itable_unused =
+				cpu_to_le16(EXT3_INODES_PER_GROUP(sb) - ino);
+		}
+	}
+
 	gdp->bg_free_inodes_count =
 		cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1);
 	if (S_ISDIR(mode)) {
 		gdp->bg_used_dirs_count =
 			cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1);
 	}
+	gdp->bg_checksum = ext3_group_desc_csum(sbi, group, gdp);
 	spin_unlock(sb_bgl_lock(sbi, group));
 	BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
 	err = ext3_journal_dirty_metadata(handle, bh2);
@@ -606,7 +720,7 @@ got:
 		inode->i_gid = current->fsgid;
 	inode->i_mode = mode;
 
-	inode->i_ino = ino;
+	inode->i_ino = ino + group * EXT3_INODES_PER_GROUP(sb);
 	/* This is the optimal IO size (for stat), not the fs block size */
 	inode->i_blksize = PAGE_SIZE;
 	inode->i_blocks = 0;
Index: linux-2.6.9-full/fs/ext3/mballoc.c
===================================================================
--- linux-2.6.9-full.orig/fs/ext3/mballoc.c	2007-03-28 15:42:45.000000000 +0400
+++ linux-2.6.9-full/fs/ext3/mballoc.c	2007-03-28 18:36:16.000000000 +0400
@@ -36,6 +36,8 @@
 #include <linux/seq_file.h>
 #include <linux/version.h>
 
+#include "group.h"
+
 /*
  * MUSTDO:
  *   - test ext3_ext_search_left() and ext3_ext_search_right()
@@ -323,6 +325,7 @@ struct ext3_group_info {
 	unsigned long	bb_state;
 	unsigned long 	bb_tid;
 	struct ext3_free_metadata *bb_md_cur;
+	struct ext3_group_desc *bb_gdp;
 	unsigned short	bb_first_free;
 	unsigned short	bb_free;
 	unsigned short	bb_fragments;
@@ -928,10 +931,7 @@ static int ext3_mb_init_cache(struct pag
 		if (first_group + i >= EXT3_SB(sb)->s_groups_count)
 			break;
 
-		err = -EIO;
-		desc = ext3_get_group_desc(sb, first_group + i, NULL);
-		if (desc == NULL)
-			goto out;
+		desc = EXT3_GROUP_INFO(sb, first_group + i)->bb_gdp;
 
 		err = -ENOMEM;
 		bh[i] = sb_getblk(sb, le32_to_cpu(desc->bg_block_bitmap));
@@ -946,7 +946,12 @@ static int ext3_mb_init_cache(struct pag
 			unlock_buffer(bh[i]);
 			continue;
 		}
-
+		if (desc->bg_flags & cpu_to_le16(EXT3_BG_BLOCK_UNINIT)) {
+			ext3_init_block_bitmap(sb, bh[i], first_group + i,desc);
+			set_buffer_uptodate(bh[i]);
+			unlock_buffer(bh[i]);
+			continue;
+		}
 		get_bh(bh[i]);
 		bh[i]->b_end_io = end_buffer_read_sync;
 		submit_bh(READ, bh[i]);
@@ -1703,6 +1708,10 @@ static int ext3_mb_good_group(struct ext
 	switch (cr) {
 		case 0:
 			BUG_ON(ac->ac_2order == 0);
+			/* If this group is uninitialized, skip it initially */
+			if (grp->bb_gdp->bg_flags &
+			    cpu_to_le16(EXT3_BG_BLOCK_UNINIT))
+				return 0;
 			bits = ac->ac_sb->s_blocksize_bits + 1;
 			for (i = ac->ac_2order; i <= bits; i++)
 				if (grp->bb_counters[i] > 0)
@@ -1796,7 +1805,9 @@ repeat:
 			}
 
 			ac->ac_groups_scanned++;
-			if (cr == 0)
+			if (cr == 0 || (e3b.bd_info->bb_gdp->bg_flags &
+					cpu_to_le16(EXT3_BG_BLOCK_UNINIT) &&
+					ac->ac_2order != 0))
 				ext3_mb_simple_scan_group(ac, &e3b);
 			else if (cr == 1 && ac->ac_g_ex.fe_len == sbi->s_stripe)
 				ext3_mb_scan_aligned(ac, &e3b);
@@ -2267,12 +2278,13 @@ int ext3_mb_init_backend(struct super_bl
 			i--;
 			goto err_freebuddy;
 		}
+		memset(meta_group_info[j], 0, len);
 		desc = ext3_get_group_desc(sb, i, NULL);
+		meta_group_info[j]->bb_gdp = desc;
 		if (desc == NULL) {
 			printk(KERN_ERR"EXT3-fs: can't read descriptor %u\n",i);
 			goto err_freebuddy;
 		}
-		memset(meta_group_info[j], 0, len);
 		set_bit(EXT3_GROUP_INFO_NEED_INIT_BIT,
 			&meta_group_info[j]->bb_state);
 
@@ -2936,9 +2948,17 @@ int ext3_mb_mark_diskspace_used(struct e
 	mb_set_bits(bitmap_bh->b_data, ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len);
 
 	spin_lock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group));
+	if (gdp->bg_flags & cpu_to_le16(EXT3_BG_BLOCK_UNINIT)) {
+		gdp->bg_flags &= cpu_to_le16(~EXT3_BG_BLOCK_UNINIT);
+		gdp->bg_free_blocks_count =
+			cpu_to_le16(ext3_free_blocks_after_init(sb,
+							    ac->ac_b_ex.fe_group,
+							    gdp));
+	}
 	gdp->bg_free_blocks_count =
 		cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count)
 				- ac->ac_b_ex.fe_len);
+	gdp->bg_checksum = ext3_group_desc_csum(sbi, ac->ac_b_ex.fe_group, gdp);
 	spin_unlock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group));
 	percpu_counter_mod(&sbi->s_freeblocks_counter, - ac->ac_b_ex.fe_len);
 
@@ -4303,6 +4323,7 @@ do_more:
 	spin_lock(sb_bgl_lock(sbi, block_group));
 	gdp->bg_free_blocks_count =
 		cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) + count);
+	gdp->bg_checksum = ext3_group_desc_csum(sbi, block_group, gdp);
 	spin_unlock(sb_bgl_lock(sbi, block_group));
 	percpu_counter_mod(&sbi->s_freeblocks_counter, count);
 
Index: linux-2.6.9-full/fs/ext3/balloc.c
===================================================================
--- linux-2.6.9-full.orig/fs/ext3/balloc.c	2007-03-28 15:45:41.000000000 +0400
+++ linux-2.6.9-full/fs/ext3/balloc.c	2007-03-28 18:36:16.000000000 +0400
@@ -20,6 +20,7 @@
 #include <linux/quotaops.h>
 #include <linux/buffer_head.h>
 
+#include "group.h"
 /*
  * balloc.c contains the blocks allocation and deallocation routines
  */
@@ -73,6 +74,75 @@ struct ext3_group_desc * ext3_get_group_
 	return gdp + desc;
 }
 
+/* Initializes an uninitialized block bitmap if given, and returns the
+ * number of blocks free in the group. */
+unsigned ext3_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
+				int block_group, struct ext3_group_desc *gdp)
+{
+	unsigned long start;
+	int bit, bit_max;
+	unsigned free_blocks;
+	struct ext3_sb_info *sbi = EXT3_SB(sb);
+
+	if (bh) {
+		J_ASSERT_BH(bh, buffer_locked(bh));
+
+		/* If checksum is bad mark all blocks use to prevent allocation,
+		 * essentially implementing a per-group read-only flag. */
+		if (!ext3_group_desc_csum_verify(sbi, block_group, gdp)) {
+			ext3_error(sb, __FUNCTION__,
+				   "Checksum bad for group %u\n", block_group);
+			gdp->bg_free_blocks_count = 0;
+			gdp->bg_free_inodes_count = 0;
+			gdp->bg_itable_unused = 0;
+			memset(bh->b_data, 0xff, sb->s_blocksize);
+			return 0;
+		}
+		memset(bh->b_data, 0, sb->s_blocksize);
+	}
+
+	/* Check for superblock and gdt backups in this group */
+	bit_max = ext3_bg_has_super(sb, block_group);
+
+	if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
+	    block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
+			  sbi->s_desc_per_block) {
+		if (bit_max) {
+			bit_max += ext3_bg_num_gdb(sb, block_group);
+			bit_max +=le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
+		}
+	} else { /* For META_BG_BLOCK_GROUPS */
+		int group_rel = (block_group -
+				 le32_to_cpu(sbi->s_es->s_first_meta_bg)) %
+				EXT3_DESC_PER_BLOCK(sb);
+		if (group_rel == 0 || group_rel == 1 ||
+		    (group_rel == EXT3_DESC_PER_BLOCK(sb) - 1))
+			bit_max += 1;
+	}
+
+	/* Last and first groups are always initialized */
+	free_blocks = EXT3_BLOCKS_PER_GROUP(sb) - bit_max;
+
+	if (bh) {
+		for (bit = 0; bit < bit_max; bit++)
+			ext3_set_bit(bit, bh->b_data);
+
+		start = block_group * EXT3_BLOCKS_PER_GROUP(sb) +
+			le32_to_cpu(sbi->s_es->s_first_data_block);
+
+		/* Set bits for block and inode bitmaps, and inode table */
+		ext3_set_bit(le32_to_cpu(gdp->bg_block_bitmap) - start,
+			     bh->b_data);
+		ext3_set_bit(le32_to_cpu(gdp->bg_inode_bitmap) - start,
+			     bh->b_data);
+		for (bit = le32_to_cpu(gdp->bg_inode_table) - start,
+		     bit_max = bit + sbi->s_itb_per_group; bit < bit_max; bit++)
+			ext3_set_bit(bit, bh->b_data);
+	}
+
+	return free_blocks - sbi->s_itb_per_group - 2;
+}
+
 /*
  * Read the bitmap for a given block_group, reading into the specified 
  * slot in the superblock's bitmap cache.
@@ -88,7 +158,19 @@ read_block_bitmap(struct super_block *sb
 	desc = ext3_get_group_desc (sb, block_group, NULL);
 	if (!desc)
 		goto error_out;
-	bh = sb_bread(sb, le32_to_cpu(desc->bg_block_bitmap));
+	if (desc->bg_flags & cpu_to_le16(EXT3_BG_BLOCK_UNINIT)) {
+		bh = sb_getblk(sb, le32_to_cpu(desc->bg_block_bitmap));
+		if (!buffer_uptodate(bh)) {
+			lock_buffer(bh);
+			if (!buffer_uptodate(bh)) {
+				ext3_init_block_bitmap(sb, bh,block_group,desc);
+				set_buffer_uptodate(bh);
+			}
+			unlock_buffer(bh);
+		}
+	} else {
+		bh = sb_bread(sb, le32_to_cpu(desc->bg_block_bitmap));
+	}
 	if (!bh)
 		ext3_error (sb, "read_block_bitmap",
 			    "Cannot read block bitmap - "
@@ -429,6 +511,7 @@ do_more:
 	gdp->bg_free_blocks_count =
 		cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) +
 			*pdquot_freed_blocks);
+	gdp->bg_checksum = ext3_group_desc_csum(sbi, block_group, gdp);
 	spin_unlock(sb_bgl_lock(sbi, block_group));
 	percpu_counter_mod(&sbi->s_freeblocks_counter, count);
 
@@ -1330,8 +1413,11 @@ allocated:
 			ret_block, goal_hits, goal_attempts);
 
 	spin_lock(sb_bgl_lock(sbi, group_no));
+	if (gdp->bg_flags & cpu_to_le16(EXT3_BG_BLOCK_UNINIT))
+		gdp->bg_flags &= cpu_to_le16(~EXT3_BG_BLOCK_UNINIT);
 	gdp->bg_free_blocks_count =
 			cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) - 1);
+	gdp->bg_checksum = ext3_group_desc_csum(sbi, group_no, gdp);
 	spin_unlock(sb_bgl_lock(sbi, group_no));
 	percpu_counter_mod(&sbi->s_freeblocks_counter, -1);