From: Kemeng Shi <shikemeng@huaweicloud.com>
To: tytso@mit.edu, adilger.kernel@dilger.ca, ojaswin@linux.ibm.com
Cc: linux-ext4@vger.kernel.org, linux-kernel@vger.kernel.org,
shikemeng@huaweicloud.com
Subject: [PATCH v4 16/19] ext4: call ext4_mb_mark_group_bb in ext4_mb_clear_bb
Date: Sat, 3 Jun 2023 23:03:24 +0800 [thread overview]
Message-ID: <20230603150327.3596033-17-shikemeng@huaweicloud.com> (raw)
In-Reply-To: <20230603150327.3596033-1-shikemeng@huaweicloud.com>
call ext4_mb_mark_group_bb in ext4_mb_clear_bb to remove repeat code
to update block bitmap and group descriptor on disk.
Note: ext4_mb_clear_bb will update buddy and bitmap in two critical sections
instead of update in the same critical section.
Original lock behavior introduced in 7a2fcbf7f857 ("ext4: don't use
blocks freed but not yet committed in buddy cache init") to avoid
race betwwen ext4_mb_free_blocks and ext4_mb_init_cache:
ext4_mb_load_buddy_gfp
ext4_lock_group
mb_clear_bits(bitmap_bh, ...)
mb_free_blocks/ext4_mb_free_metadata
ext4_unlock_group
ext4_mb_unload_buddy
New lock behavior in this patch:
ext4_mb_load_buddy_gfp
ext4_lock_group
mb_clear_bits(bitmap_bh, ...)
ext4_unlock_group
/* no ext4_mb_init_cache for the same group will be called as
ext4_mb_load_buddy_gfp will ensure buddy page is update-to-date */
ext4_lock_group
mb_free_blocks/ext4_mb_free_metadata
ext4_unlock_group
ext4_mb_unload_buddy
As buddy page for group is always update-to-date between
ext4_mb_load_buddy_gfp and ext4_mb_unload_buddy. Then no
ext4_mb_init_cache will be called for the same group concurrentlly when
we update bitmap and buddy page betwwen buddy load and unload.
Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
---
fs/ext4/mballoc.c | 90 ++++++++++++-----------------------------------
1 file changed, 23 insertions(+), 67 deletions(-)
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 46b37f5c9223..e4f1b34448e3 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -6135,19 +6135,21 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
ext4_fsblk_t block, unsigned long count,
int flags)
{
- struct buffer_head *bitmap_bh = NULL;
+ struct ext4_mark_context mc = {
+ .handle = handle,
+ .sb = inode->i_sb,
+ .state = 0,
+ };
struct super_block *sb = inode->i_sb;
- struct ext4_group_desc *gdp;
struct ext4_group_info *grp;
unsigned int overflow;
ext4_grpblk_t bit;
- struct buffer_head *gd_bh;
ext4_group_t block_group;
struct ext4_sb_info *sbi;
struct ext4_buddy e4b;
unsigned int count_clusters;
int err = 0;
- int ret;
+ int mark_flags = 0;
sbi = EXT4_SB(sb);
@@ -6179,18 +6181,6 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
/* The range changed so it's no longer validated */
flags &= ~EXT4_FREE_BLOCKS_VALIDATED;
}
- count_clusters = EXT4_NUM_B2C(sbi, count);
- 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;
- }
- 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)) {
@@ -6200,28 +6190,7 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
goto error_return;
}
- 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;
-
- /*
- * We are about to modify some metadata. Call the journal APIs
- * to unshare ->b_data if a currently-committing transaction is
- * using it
- */
- 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;
-#ifdef AGGRESSIVE_CHECK
- {
- int i;
- for (i = 0; i < count_clusters; i++)
- BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data));
- }
-#endif
+ 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. */
@@ -6230,6 +6199,22 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
if (err)
goto error_return;
+#ifdef AGGRESSIVE_CHECK
+ mark_flags |= EXT4_MB_BITMAP_MARKED_CHECK;
+#endif
+ err = ext4_mb_mark_group_bb(&mc, block_group, bit, count_clusters,
+ mark_flags);
+
+
+ if (err && mc.changed == 0) {
+ ext4_mb_unload_buddy(&e4b);
+ goto error_return;
+ }
+
+#ifdef AGGRESSIVE_CHECK
+ BUG_ON(mc.changed != count_clusters);
+#endif
+
/*
* We need to make sure we don't reuse the freed block until after the
* transaction is committed. We make an exception if the inode is to be
@@ -6252,13 +6237,8 @@ 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);
@@ -6271,23 +6251,11 @@ 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
@@ -6302,26 +6270,14 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
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;
- put_bh(bitmap_bh);
/* The range changed so it's no longer validated */
flags &= ~EXT4_FREE_BLOCKS_VALIDATED;
goto do_more;
}
error_return:
- brelse(bitmap_bh);
ext4_std_error(sb, err);
return;
}
--
2.30.0
next prev parent reply other threads:[~2023-06-03 7:02 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-06-03 15:03 [PATCH v4 00/19] Fixes, cleanups and unit test for mballoc Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 01/19] ext4: fix wrong unit use in ext4_mb_normalize_request Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 02/19] ext4: fix unit mismatch in ext4_mb_new_blocks_simple Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 03/19] ext4: fix wrong unit use in ext4_mb_find_by_goal Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 04/19] ext4: treat stripe in block unit Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 05/19] ext4: add EXT4_MB_HINT_GOAL_ONLY test in ext4_mb_use_preallocated Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 06/19] ext4: remove ext4_block_group and ext4_block_group_offset declaration Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 07/19] ext4: try all groups in ext4_mb_new_blocks_simple Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 08/19] ext4: get block from bh before pass it to ext4_free_blocks_simple in ext4_free_blocks Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 09/19] ext4: remove unsed parameter and unnecessary forward declaration of ext4_mb_new_blocks_simple Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 10/19] ext4: fix wrong unit use in ext4_mb_clear_bb Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 11/19] ext4: fix wrong unit use in ext4_mb_new_blocks Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 12/19] ext4: factor out codes to update block bitmap and group descriptor on disk from ext4_mb_mark_bb Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 13/19] ext4: call ext4_mb_mark_group_bb in ext4_free_blocks_simple Kemeng Shi
2023-06-11 5:05 ` Theodore Ts'o
2023-06-12 2:24 ` Kemeng Shi
2023-06-12 3:49 ` Theodore Ts'o
2023-06-13 1:22 ` Kemeng Shi
2023-06-20 1:50 ` Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 14/19] ext4: extent ext4_mb_mark_group_bb to support allocation under journal Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 15/19] ext4: call ext4_mb_mark_group_bb in ext4_mb_mark_diskspace_used Kemeng Shi
2023-06-03 15:03 ` Kemeng Shi [this message]
2023-06-06 9:39 ` [PATCH v4 16/19] ext4: call ext4_mb_mark_group_bb in ext4_mb_clear_bb Ojaswin Mujoo
2023-06-06 14:16 ` Kemeng Shi
2023-06-08 6:16 ` Ojaswin Mujoo
2023-06-03 15:03 ` [PATCH v4 17/19] ext4: call ext4_mb_mark_group_bb in ext4_group_add_blocks Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 18/19] ext4: add some kunit stub for mballoc kunit test Kemeng Shi
2023-06-03 15:03 ` [PATCH v4 19/19] ext4: add first unit test for ext4_mb_new_blocks_simple in mballoc Kemeng Shi
2023-06-09 3:14 ` [PATCH v4 00/19] Fixes, cleanups and unit test for mballoc Theodore Ts'o
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230603150327.3596033-17-shikemeng@huaweicloud.com \
--to=shikemeng@huaweicloud.com \
--cc=adilger.kernel@dilger.ca \
--cc=linux-ext4@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=ojaswin@linux.ibm.com \
--cc=tytso@mit.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).