[PATCH] f2fs: avoid unnecessary reads in f2fs_update_meta_block()

[PATCH] f2fs: avoid unnecessary reads in f2fs_update_meta_block()

[Seongjae Jeong]

When the packed SSA feature is enabled, f2fs_update_meta_block()
uses f2fs_get_meta_cache_retry(), which reads the target block from
disk if the cache entry is not uptodate.

However, f2fs_update_meta_block() overwrites the entire filesystem
block with memcpy(), so the existing contents are not needed.
Partial updates to packed SSA blocks are handled separately in
write_sum_block().

Use f2fs_grab_meta_cache() unconditionally to avoid reading blocks
that will be fully overwritten. Also remove the error check, since
f2fs_grab_meta_cache() retries allocation failures internally and
does not return an error pointer.

Signed-off-by: Seongjae Jeong <jsjlee1020@gmail.com>
---
 fs/f2fs/segment.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 794e0d99fc7c..51858f64e43e 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -2844,13 +2844,7 @@ void f2fs_update_meta_block(struct f2fs_sb_info *sbi,
 {
 	struct f2fs_cached_block *entry;
 
-	if (!f2fs_sb_has_packed_ssa(sbi))
-		entry = f2fs_grab_meta_cache(sbi, blk_addr);
-	else
-		entry = f2fs_get_meta_cache_retry(sbi, blk_addr);
-
-	if (IS_ERR(entry))
-		return;
+	entry = f2fs_grab_meta_cache(sbi, blk_addr);
 
 	memcpy(cache_address(entry), src, F2FS_BLKSIZE(sbi));
 	f2fs_mark_cache_dirty(entry);
-- 
2.53.0

Re: [PATCH] f2fs: avoid unnecessary reads in f2fs_update_meta_block()

[Chao Yu]

On 10/2/26 14:48, Seongjae Jeong wrote:
> When the packed SSA feature is enabled, f2fs_update_meta_block()
> uses f2fs_get_meta_cache_retry(), which reads the target block from
> disk if the cache entry is not uptodate.
> 
> However, f2fs_update_meta_block() overwrites the entire filesystem
> block with memcpy(), so the existing contents are not needed.
> Partial updates to packed SSA blocks are handled separately in
> write_sum_block().
> 
> Use f2fs_grab_meta_cache() unconditionally to avoid reading blocks
> that will be fully overwritten. Also remove the error check, since
> f2fs_grab_meta_cache() retries allocation failures internally and
> does not return an error pointer.
> 
> Signed-off-by: Seongjae Jeong <jsjlee1020@gmail.com>
> ---
>   fs/f2fs/segment.c | 8 +-------
>   1 file changed, 1 insertion(+), 7 deletions(-)
> 
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index 794e0d99fc7c..51858f64e43e 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -2844,13 +2844,7 @@ void f2fs_update_meta_block(struct f2fs_sb_info *sbi,
>   {
>   	struct f2fs_cached_block *entry;
>   
> -	if (!f2fs_sb_has_packed_ssa(sbi))
> -		entry = f2fs_grab_meta_cache(sbi, blk_addr);
> -	else
> -		entry = f2fs_get_meta_cache_retry(sbi, blk_addr);
> -
> -	if (IS_ERR(entry))
> -		return;
> +	entry = f2fs_grab_meta_cache(sbi, blk_addr);

Daeho,

 From code side, I feel this is clean.

But, could you double check this cleanup? in case if we handle this
intentionally in original patch for something.

Thanks,

>   
>   	memcpy(cache_address(entry), src, F2FS_BLKSIZE(sbi));
>   	f2fs_mark_cache_dirty(entry);