[DRBD-user] local disk flush failed with status -5

[Emir Mahmutbegovic]

Hallo Lars,

Thanks for this reply, I got this thing working.

Best Regards

Lars Ellenberg wrote:
> On Tue, Mar 18, 2008 at 06:58:18PM +0100, Emir Mahmutbegovic wrote:
>   
>> Hallo,
>>
>> I am running one two node cluster with heartbeat 2.1.3 and drbd. I made 
>> update of drbd (from 8.2.4) to 8.2.5 version and since then i do find this 
>> in dmesg on both nodes (node with primary drbd does not print this out):
>>
>> drbd1: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd0: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd2: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>> drbd1: local disk flush failed with status -5
>>
>> Could someone please tell me what does this error mean ?
>>     
>
> drbd 8.0.11 (and 8.2.5) uses barriers/flushes in certain places
> to be able to work reliably even on io subsystems with volatile
> - but flushable - caches.
>
> whether or not DRBD uses these features is decided by trying once, and
> if that results in "sorry, I don't support that", never try again.
>
> on the way to 2.6.24, kernel.org did change the way how barriers and
> flushes are handled, but forgot to properly pass the "not supported"
> status back, returning a generic "IO error" instead (the -5 above).
>
> this was later fixed again, see below.
> so please use a kernel that contains that fix,
> or a kernel that does still use the old behaviour.
>
> commit cc66b4512cae8df4ed1635483210aabf7690ec27
> Author: Jens Axboe <jens.axboe at oracle.com>
> Date:   Tue Mar 4 11:47:46 2008 +0100
>
>     block: fix blkdev_issue_flush() not detecting and passing EOPNOTSUPP back
>     
>     This is important to eg dm, that tries to decide whether to stop using
>     barriers or not.
>     
>     Tested as working by Anders Henke <anders.henke at 1und1.de>
>     
>     Signed-off-by: Jens Axboe <jens.axboe at oracle.com>
>
> diff --git a/block/blk-barrier.c b/block/blk-barrier.c
> index 6901eed..55c5f1f 100644
> --- a/block/blk-barrier.c
> +++ b/block/blk-barrier.c
> @@ -259,8 +259,11 @@ int blk_do_ordered(struct request_queue *q, struct request **rqp)
>  
>  static void bio_end_empty_barrier(struct bio *bio, int err)
>  {
> -	if (err)
> +	if (err) {
> +		if (err == -EOPNOTSUPP)
> +			set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
>  		clear_bit(BIO_UPTODATE, &bio->bi_flags);
> +	}
>  
>  	complete(bio->bi_private);
>  }
> @@ -309,7 +312,9 @@ int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector)
>  		*error_sector = bio->bi_sector;
>  
>  	ret = 0;
> -	if (!bio_flagged(bio, BIO_UPTODATE))
> +	if (bio_flagged(bio, BIO_EOPNOTSUPP))
> +		ret = -EOPNOTSUPP;
> +	else if (!bio_flagged(bio, BIO_UPTODATE))
>  		ret = -EIO;
>  
>  	bio_put(bio);
>
>