[PATCH v7 06/12] ext4: Separate block bitmap and buddy bitmap freeing in ext4_mb_clear_bb()

[Kemeng Shi]

This patch separates block bitmap and buddy bitmap freeing in order to
update block bitmap with ext4_mb_mark_context in following patch.

Separated freeing is safe with concurrent allocation as long as:
1. Firstly allocate block in buddy bitmap, and then in block bitmap.
2. Firstly free block in block bitmap, and then buddy bitmap.
Then freed block will only be available to allocation when both buddy
bitmap and block bitmap are updated by freeing.
Allocation obeys rule 1 already, just do sperated freeing with rule 2.

Separated freeing has no race with generate_buddy as:
Once ext4_mb_load_buddy_gfp is executed successfully, the update-to-date
buddy page can be found in sbi->s_buddy_cache and no more buddy
initialization of the buddy page will be executed concurrently until
buddy page is unloaded. As we always do free in "load buddy, free,
unload buddy" sequence, separated freeing has no race with generate_buddy.

Signed-off-by: Kemeng Shi <shikemeng@xxxxxxxxxxxxxxx>
---
 fs/ext4/mballoc.c | 98 +++++++++++++++++++++++------------------------
 1 file changed, 49 insertions(+), 49 deletions(-)

diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 87e5922997fb..8c1127aa7f59 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -6423,7 +6423,7 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 		ext4_error(sb, "Freeing blocks in system zone - "
 			   "Block = %llu, count = %lu", block, count);
 		/* err = 0. ext4_std_error should be a no op */
-		goto error_return;
+		goto error_out;
 	}
 	flags |= EXT4_FREE_BLOCKS_VALIDATED;
 
@@ -6447,31 +6447,39 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 		flags &= ~EXT4_FREE_BLOCKS_VALIDATED;
 	}
 	count_clusters = EXT4_NUM_B2C(sbi, count);
+	trace_ext4_mballoc_free(sb, inode, block_group, bit, count_clusters);
+
+	/* __GFP_NOFAIL: retry infinitely, ignore TIF_MEMDIE and memcg limit. */
+	err = ext4_mb_load_buddy_gfp(sb, block_group, &e4b,
+				     GFP_NOFS|__GFP_NOFAIL);
+	if (err)
+		goto error_out;
+
+	if (!(flags & EXT4_FREE_BLOCKS_VALIDATED) &&
+	    !ext4_inode_block_valid(inode, block, count)) {
+		ext4_error(sb, "Freeing blocks in system zone - "
+			   "Block = %llu, count = %lu", block, count);
+		/* err = 0. ext4_std_error should be a no op */
+		goto error_clean;
+	}
+
 	bitmap_bh = ext4_read_block_bitmap(sb, block_group);
 	if (IS_ERR(bitmap_bh)) {
 		err = PTR_ERR(bitmap_bh);
 		bitmap_bh = NULL;
-		goto error_return;
+		goto error_clean;
 	}
 	gdp = ext4_get_group_desc(sb, block_group, &gd_bh);
 	if (!gdp) {
 		err = -EIO;
-		goto error_return;
-	}
-
-	if (!(flags & EXT4_FREE_BLOCKS_VALIDATED) &&
-	    !ext4_inode_block_valid(inode, block, count)) {
-		ext4_error(sb, "Freeing blocks in system zone - "
-			   "Block = %llu, count = %lu", block, count);
-		/* err = 0. ext4_std_error should be a no op */
-		goto error_return;
+		goto error_clean;
 	}
 
 	BUFFER_TRACE(bitmap_bh, "getting write access");
 	err = ext4_journal_get_write_access(handle, sb, bitmap_bh,
 					    EXT4_JTR_NONE);
 	if (err)
-		goto error_return;
+		goto error_clean;
 
 	/*
 	 * We are about to modify some metadata.  Call the journal APIs
@@ -6481,7 +6489,7 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 	BUFFER_TRACE(gd_bh, "get_write_access");
 	err = ext4_journal_get_write_access(handle, sb, gd_bh, EXT4_JTR_NONE);
 	if (err)
-		goto error_return;
+		goto error_clean;
 #ifdef AGGRESSIVE_CHECK
 	{
 		int i;
@@ -6489,13 +6497,30 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 			BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data));
 	}
 #endif
-	trace_ext4_mballoc_free(sb, inode, block_group, bit, count_clusters);
+	ext4_lock_group(sb, block_group);
+	mb_clear_bits(bitmap_bh->b_data, bit, count_clusters);
+	ret = ext4_free_group_clusters(sb, gdp) + count_clusters;
+	ext4_free_group_clusters_set(sb, gdp, ret);
+	ext4_block_bitmap_csum_set(sb, gdp, bitmap_bh);
+	ext4_group_desc_csum_set(sb, block_group, gdp);
+	ext4_unlock_group(sb, block_group);
 
-	/* __GFP_NOFAIL: retry infinitely, ignore TIF_MEMDIE and memcg limit. */
-	err = ext4_mb_load_buddy_gfp(sb, block_group, &e4b,
-				     GFP_NOFS|__GFP_NOFAIL);
-	if (err)
-		goto error_return;
+	if (sbi->s_log_groups_per_flex) {
+		ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
+		atomic64_add(count_clusters,
+			     &sbi_array_rcu_deref(sbi, s_flex_groups,
+						  flex_group)->free_clusters);
+	}
+
+	/* We dirtied the bitmap block */
+	BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
+	err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
+
+	/* And the group descriptor block */
+	BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
+	ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
+	if (!err)
+		err = ret;
 
 	/*
 	 * We need to make sure we don't reuse the freed block until after the
@@ -6519,13 +6544,8 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 		new_entry->efd_tid = handle->h_transaction->t_tid;
 
 		ext4_lock_group(sb, block_group);
-		mb_clear_bits(bitmap_bh->b_data, bit, count_clusters);
 		ext4_mb_free_metadata(handle, &e4b, new_entry);
 	} else {
-		/* need to update group_info->bb_free and bitmap
-		 * with group lock held. generate_buddy look at
-		 * them with group lock_held
-		 */
 		if (test_opt(sb, DISCARD)) {
 			err = ext4_issue_discard(sb, block_group, bit,
 						 count_clusters, NULL);
@@ -6538,23 +6558,11 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 			EXT4_MB_GRP_CLEAR_TRIMMED(e4b.bd_info);
 
 		ext4_lock_group(sb, block_group);
-		mb_clear_bits(bitmap_bh->b_data, bit, count_clusters);
 		mb_free_blocks(inode, &e4b, bit, count_clusters);
 	}
 
-	ret = ext4_free_group_clusters(sb, gdp) + count_clusters;
-	ext4_free_group_clusters_set(sb, gdp, ret);
-	ext4_block_bitmap_csum_set(sb, gdp, bitmap_bh);
-	ext4_group_desc_csum_set(sb, block_group, gdp);
 	ext4_unlock_group(sb, block_group);
 
-	if (sbi->s_log_groups_per_flex) {
-		ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
-		atomic64_add(count_clusters,
-			     &sbi_array_rcu_deref(sbi, s_flex_groups,
-						  flex_group)->free_clusters);
-	}
-
 	/*
 	 * on a bigalloc file system, defer the s_freeclusters_counter
 	 * update to the caller (ext4_remove_space and friends) so they
@@ -6567,28 +6575,20 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
 				   count_clusters);
 	}
 
-	ext4_mb_unload_buddy(&e4b);
-
-	/* We dirtied the bitmap block */
-	BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
-	err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
-
-	/* And the group descriptor block */
-	BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
-	ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
-	if (!err)
-		err = ret;
-
 	if (overflow && !err) {
 		block += count;
 		count = overflow;
+		ext4_mb_unload_buddy(&e4b);
 		put_bh(bitmap_bh);
 		/* The range changed so it's no longer validated */
 		flags &= ~EXT4_FREE_BLOCKS_VALIDATED;
 		goto do_more;
 	}
-error_return:
+
+error_clean:
+	ext4_mb_unload_buddy(&e4b);
 	brelse(bitmap_bh);
+error_out:
 	ext4_std_error(sb, err);
 }
 
-- 
2.30.0