Re: [PATCH 01/11] block: add new BlockDriver handler: bdrv_cancel_in_flight

[Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>]

21.01.2021 01:27, Eric Blake wrote:
> On 11/18/20 12:04 PM, Vladimir Sementsov-Ogievskiy wrote:
>> It will be used to stop retrying NBD requests on mirror cancel.
>>
>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>> ---
>>   include/block/block.h     |  3 +++
>>   include/block/block_int.h |  9 +++++++++
>>   block/io.c                | 11 +++++++++++
>>   3 files changed, 23 insertions(+)
>>
> 
> How does this relate to the recent addition of the QMP yank command?

Hmm. Don't know.. Looking in spec:

# Currently implemented yank instances:
#  - nbd block device:
#    Yanking it will shut down the connection to the nbd server without
#    attempting to reconnect.

Looks like a close thing. But actually, I don't want to stop reconnecting process, but only cancel some requests which we don't want to wait for. After that, the nbd node may successfully reconnect and continue to work.

> 
> 
>> +++ b/include/block/block_int.h
>> @@ -344,6 +344,15 @@ struct BlockDriver {
>>           bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
>>           int64_t *map, BlockDriverState **file);
>>   
>> +    /*
>> +     * This informs the driver that we are not more interested in in-flight
> 
> that we are no longer interested in the result of in-flight requests, so
> 
>> +     * requests results, so don't waste the time if possible.
>> +     *
>> +     * The example usage is to not wait for nbd target nodedreconnect timeout on
>> +     * job-cancel.
> 
> One example usage is to avoid waiting for an nbd target node reconnect
> timeout during job-cancel.
> 
> Reviewed-by: Eric Blake <eblake@redhat.com>
> 


-- 
Best regards,
Vladimir