Re: [bug report] blktests srp/002 hang

[Bob Pearson <rpearsonhpe@gmail.com>]

On 10/19/23 14:17, Bart Van Assche wrote:
> On 10/18/23 14:52, Bob Pearson wrote:
>> results are slightly ambiguous but I ran the command here:
>>
>> rpearson:blktests$ sudo use_siw=1 ./check srp/002
>> srp/002 (File I/O on top of multipath concurrently with logout and login (mq)) [failed]time  245.018s  ...
>>      runtime  245.018s  ...  128.110s
>>      --- tests/srp/002.out    2023-02-15 12:07:40.675530344 -0600
>>      +++ /home/rpearson/src/blktests/results/nodev/srp/002.out.bad    2023-10-18 16:36:14.723323257 -0500
>>      @@ -1,2 +1 @@
>>       Configured SRP target driver
>>      -Passed
>> rpearson:blktests$
>>
>> And while it was hung I ran the following:
>>
>> root@rpearson-X570-AORUS-PRO-WIFI: dmsetup ls | while read a b; do dmsetup message $a 0 fail_if_no_path; done
>> device-mapper: message ioctl on mpatha-part1  failed: Invalid argument
>> Command failed.
>> device-mapper: message ioctl on mpatha-part2  failed: Invalid argument
>> Command failed.
>> device-mapper: message ioctl on mpathb-part1  failed: Invalid argument
>> Command failed.
>>
>> mpath[ab]-part[12] are multipath devices (dm-1,2,3) holding the Ubuntu system images and not the devices created
>> by blktests. When this command finished the srp/002 run came back life but did not succeed (see above)
>>
>> The dmesg log is attached
> 
> Hi Bob,
> 
> Thank you for having shared these results. The 'dmsetup message' command
> should only be applied to multipath devices created by the srp/002
> script and not to the multipath devices created by Ubuntu but I'm not
> sure how to do that.
> 
> If the above 'dmsetup message' command resolved the hang, that indicates
> that the root cause of the hang is probably in user space and not in the
> kernel. Did you use the Ubuntu version of multipathd or a self-built
> version of multipathd? I remember that last time I ran the SRP tests on
> an Ubuntu system that I had to replace Ubuntu's multipathd with a self-built binary to make the tests pass.
> 
> Thanks for having shared the dmesg output. I don't think it is possible to derive the root cause from that output, which is unfortunate.
> 
> Thanks,
> 
> Bart.
> 

Bart,

The results from yesterday are from Ubuntu 23.04 which has multipath-tools at version 0.8.8.
Ubuntu 23.10 has version 0.9.4. Github has head of tree at 0.9.6. Do you recall which version
made it work when you were looking at this? It may be less risky for me to upgrade to
Ubuntu 23.10 than try and build multipath-tools from source.

Bob