[PATCH] ext4: fix warning when turn on dioread_nolock and inline_data

[PATCH] ext4: fix warning when turn on dioread_nolock and inline_data

[yangerkun]

mkfs.ext4 -O inline_data /dev/vdb
mount -o dioread_nolock /dev/vdb /mnt
echo "some inline data..." >> /mnt/test-file
echo "some inline data..." >> /mnt/test-file
sync

With upon script, system will trigger
"WARN_ON(!io_end->handle && sbi->s_journal)" since the wrong order
between rsv_blocks calculate and destroy inline data for dealloc.

Signed-off-by: yangerkun <yangerkun@huawei.com>
---
 fs/ext4/inode.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index c7f77c6..3f2a366 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2769,15 +2769,6 @@ static int ext4_writepages(struct address_space *mapping,
 		goto out_writepages;
 	}
 
-	if (ext4_should_dioread_nolock(inode)) {
-		/*
-		 * We may need to convert up to one extent per block in
-		 * the page and we may dirty the inode.
-		 */
-		rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
-						PAGE_SIZE >> inode->i_blkbits);
-	}
-
 	/*
 	 * If we have inline data and arrive here, it means that
 	 * we will soon create the block for the 1st page, so
@@ -2796,6 +2787,15 @@ static int ext4_writepages(struct address_space *mapping,
 		ext4_journal_stop(handle);
 	}
 
+	if (ext4_should_dioread_nolock(inode)) {
+		/*
+		 * We may need to convert up to one extent per block in
+		 * the page and we may dirty the inode.
+		 */
+		rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
+						PAGE_SIZE >> inode->i_blkbits);
+	}
+
 	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
 		range_whole = 1;
 
-- 
2.7.4

Re: [PATCH] ext4: fix warning when turn on dioread_nolock and inline_data

[Jan Kara]

On Thu 04-07-19 20:50:32, yangerkun wrote:
> mkfs.ext4 -O inline_data /dev/vdb
> mount -o dioread_nolock /dev/vdb /mnt
> echo "some inline data..." >> /mnt/test-file
> echo "some inline data..." >> /mnt/test-file
> sync
> 
> With upon script, system will trigger
> "WARN_ON(!io_end->handle && sbi->s_journal)" since the wrong order
> between rsv_blocks calculate and destroy inline data for dealloc.

Thanks for the patch! Good catch! I'd just rephrase the last paragraph as:

The above script will trigger "WARN_ON(!io_end->handle && sbi->s_journal)"
because ext4_should_dioread_nolock() returns false for a file with inline
data. Move the check to a place after we have already removed the inline
data and prepared inode to write normal pages.

Otherwise the patch looks good to me so feel free to add:

Reviewed-by: Jan Kara <jack@suse.cz>

								Honza

> 
> Signed-off-by: yangerkun <yangerkun@huawei.com>
> ---
>  fs/ext4/inode.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
> 
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index c7f77c6..3f2a366 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -2769,15 +2769,6 @@ static int ext4_writepages(struct address_space *mapping,
>  		goto out_writepages;
>  	}
>  
> -	if (ext4_should_dioread_nolock(inode)) {
> -		/*
> -		 * We may need to convert up to one extent per block in
> -		 * the page and we may dirty the inode.
> -		 */
> -		rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
> -						PAGE_SIZE >> inode->i_blkbits);
> -	}
> -
>  	/*
>  	 * If we have inline data and arrive here, it means that
>  	 * we will soon create the block for the 1st page, so
> @@ -2796,6 +2787,15 @@ static int ext4_writepages(struct address_space *mapping,
>  		ext4_journal_stop(handle);
>  	}
>  
> +	if (ext4_should_dioread_nolock(inode)) {
> +		/*
> +		 * We may need to convert up to one extent per block in
> +		 * the page and we may dirty the inode.
> +		 */
> +		rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
> +						PAGE_SIZE >> inode->i_blkbits);
> +	}
> +
>  	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
>  		range_whole = 1;
>  
> -- 
> 2.7.4
> 
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH] ext4: fix warning when turn on dioread_nolock and inline_data

[yangerkun]

Thanks a lot! Will resend the patch soon!

On 2019/7/4 22:55, Jan Kara wrote:
> On Thu 04-07-19 20:50:32, yangerkun wrote:
>> mkfs.ext4 -O inline_data /dev/vdb
>> mount -o dioread_nolock /dev/vdb /mnt
>> echo "some inline data..." >> /mnt/test-file
>> echo "some inline data..." >> /mnt/test-file
>> sync
>>
>> With upon script, system will trigger
>> "WARN_ON(!io_end->handle && sbi->s_journal)" since the wrong order
>> between rsv_blocks calculate and destroy inline data for dealloc.
> Thanks for the patch! Good catch! I'd just rephrase the last paragraph as:
>
> The above script will trigger "WARN_ON(!io_end->handle && sbi->s_journal)"
> because ext4_should_dioread_nolock() returns false for a file with inline
> data. Move the check to a place after we have already removed the inline
> data and prepared inode to write normal pages.
>
> Otherwise the patch looks good to me so feel free to add:
>
> Reviewed-by: Jan Kara <jack@suse.cz>
>
> 								Honza
>
>> Signed-off-by: yangerkun <yangerkun@huawei.com>
>> ---
>>   fs/ext4/inode.c | 18 +++++++++---------
>>   1 file changed, 9 insertions(+), 9 deletions(-)
>>
>> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
>> index c7f77c6..3f2a366 100644
>> --- a/fs/ext4/inode.c
>> +++ b/fs/ext4/inode.c
>> @@ -2769,15 +2769,6 @@ static int ext4_writepages(struct address_space *mapping,
>>   		goto out_writepages;
>>   	}
>>   
>> -	if (ext4_should_dioread_nolock(inode)) {
>> -		/*
>> -		 * We may need to convert up to one extent per block in
>> -		 * the page and we may dirty the inode.
>> -		 */
>> -		rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
>> -						PAGE_SIZE >> inode->i_blkbits);
>> -	}
>> -
>>   	/*
>>   	 * If we have inline data and arrive here, it means that
>>   	 * we will soon create the block for the 1st page, so
>> @@ -2796,6 +2787,15 @@ static int ext4_writepages(struct address_space *mapping,
>>   		ext4_journal_stop(handle);
>>   	}
>>   
>> +	if (ext4_should_dioread_nolock(inode)) {
>> +		/*
>> +		 * We may need to convert up to one extent per block in
>> +		 * the page and we may dirty the inode.
>> +		 */
>> +		rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
>> +						PAGE_SIZE >> inode->i_blkbits);
>> +	}
>> +
>>   	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
>>   		range_whole = 1;
>>   
>> -- 
>> 2.7.4
>>
>>