Re: [Xen-devel] xen-blkfront: weird behavior of "iostat" after VM live-migrate which xen-blkfront module has indirect descriptors

["Ouyang Zhaowei (Charles)" <ouyangzhaowei@xxxxxxxxxx>]

On 2015.1.23 19:15, Roger Pau Monnà wrote:
> Hello,
> 
> El 23/01/15 a les 8.59, Ouyang Zhaowei (Charles) ha escrit:
>> Hi Roger,
>>
>> We are testing the indirect feature of xen-blkfront module these days.
>> And we found that, after VM live-migrate a couple of times, the "%util" of 
>> iostat keeps being 100%, and there are several requests stock in "avgqu-sz".
>> We have checked some later version of Linux, and it happens on Ubuntu 14.04, 
>> Ubuntu 14.10 and RHEL 7.0.
>>
>> The iostat shows like below:
>>
>> avg-cpu:  %user   %nice %system %iowait  %steal   %idle
>>            0.00    0.00    0.00    0.00    0.00  100.00
>>
>> Device:         rrqm/s   wrqm/s     r/s     w/s    rkB/s    wkB/s avgrq-sz 
>> avgqu-sz   await r_await w_await  svctm  %util
>> xvda              0.00     0.00    0.00    0.00     0.00     0.00     0.00   
>>   4.00    0.00    0.00    0.00   0.00 100.00
>> dm-0              0.00     0.00    0.00    0.00     0.00     0.00     0.00   
>>   0.00    0.00    0.00    0.00   0.00   0.00
>> dm-1              0.00     0.00    0.00    0.00     0.00     0.00     0.00   
>>   0.00    0.00    0.00    0.00   0.00   0.00
>>
>> Could you tell us that why is this happening, is this a bug?
> 
> It is a bug indeed, thanks for reporting it. The problem seems to be 
> that blk_put_request (which is used to discard the old requests before 
> requeuing them) doesn't update the queue statistics. The following 
> patch solves the problem for me, could you try it and report back?
> 
> ---
> commit bb4317c051ca81a2906edb7ccc505cbd6d1d80c7
> Author: Roger Pau Monne <roger.pau@xxxxxxxxxx>
> Date:   Fri Jan 23 12:10:51 2015 +0100
> 
>     xen-blkfront: fix accounting of reqs when migrating
>     
>     Current migration code uses blk_put_request in order to finish a request
>     before requeuing it. This function doesn't update the statistics of the
>     queue, which completely screws accounting. Use blk_end_request_all instead
>     which properly updates the statistics of the queue.
>     
>     Signed-off-by: Roger Pau Monnà <roger.pau@xxxxxxxxxx>
> 
> diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c
> index 5ac312f..aac41c1 100644
> --- a/drivers/block/xen-blkfront.c
> +++ b/drivers/block/xen-blkfront.c
> @@ -1493,7 +1493,7 @@ static int blkif_recover(struct blkfront_info *info)
>               merge_bio.tail = copy[i].request->biotail;
>               bio_list_merge(&bio_list, &merge_bio);
>               copy[i].request->bio = NULL;
> -             blk_put_request(copy[i].request);
> +             blk_end_request_all(copy[i].request, 0);
>       }
>  
>       kfree(copy);
> @@ -1516,7 +1516,7 @@ static int blkif_recover(struct blkfront_info *info)
>               req->bio = NULL;
>               if (req->cmd_flags & (REQ_FLUSH | REQ_FUA))
>                       pr_alert("diskcache flush request found!\n");
> -             __blk_put_request(info->rq, req);
> +             __blk_end_request_all(req, 0);
>       }
>       spin_unlock_irq(&info->io_lock);
>  
> 

Hi Roger,

Thanks for answering this question. Sure, I'll try this patch and test VM 
migrating, so far it seems this patch has solved this bug (after 10 times 
migrate).
I'll keep testing it for more times and will let you know if it's OK.

Regards,
Ouyang Zhaowei


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel