Flexible I/O Tester development
 help / color / mirror / Atom feed
* Zoned block device regression test 11
@ 2018-08-22  0:10 Bart Van Assche
  2018-08-22  4:32 ` Michael Kelley (EOSG)
  0 siblings, 1 reply; 11+ messages in thread
From: Bart Van Assche @ 2018-08-22  0:10 UTC (permalink / raw)
  To: mikelley; +Cc: Jens Axboe, fio, Damien Le Moal

Hello Michael,

If I run the fio code from https://github.com/bvanassche/fio against 
Linux kernel v4.18 as follows:

./fio/t/zbd/run-tests-against-zoned-nullb

then all tests except test number 11 pass. If I revert your axmap 
patches however then test 11 passes. I think this means that your axmap 
patches introduced a regression in the axmap code. Can you have a look 
at this?

Thanks,

Bart.


^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: Zoned block device regression test 11
  2018-08-22  0:10 Zoned block device regression test 11 Bart Van Assche
@ 2018-08-22  4:32 ` Michael Kelley (EOSG)
  2018-08-22 15:47   ` Michael Kelley (EOSG)
  0 siblings, 1 reply; 11+ messages in thread
From: Michael Kelley (EOSG) @ 2018-08-22  4:32 UTC (permalink / raw)
  To: Bart Van Assche; +Cc: Jens Axboe, fio@vger.kernel.org, Damien Le Moal

Yes, I'll take a look.

Michael

> -----Original Message-----
> From: Bart Van Assche <bvanassche@acm.org>
> Sent: Tuesday, August 21, 2018 5:11 PM
> To: Michael Kelley (EOSG) <Michael.H.Kelley@microsoft.com>
> Cc: Jens Axboe <axboe@kernel.dk>; fio@vger.kernel.org; Damien Le Moal
> <damien.lemoal@wdc.com>
> Subject: Zoned block device regression test 11
> 
> Hello Michael,
> 
> If I run the fio code from
> https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fbvanassch
> e%2Ffio&amp;data=02%7C01%7CMichael.H.Kelley%40microsoft.com%7C358c12cdc3e345932
> 68b08d607c3b02d%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636704934393729
> 755&amp;sdata=uLMcg3Ep0UVuJ031mKlQt6ZSxYToo2Av6W2p62w%2F7SU%3D&amp;reserved
> =0 against
> Linux kernel v4.18 as follows:
> 
> ./fio/t/zbd/run-tests-against-zoned-nullb
> 
> then all tests except test number 11 pass. If I revert your axmap
> patches however then test 11 passes. I think this means that your axmap
> patches introduced a regression in the axmap code. Can you have a look
> at this?
> 
> Thanks,
> 
> Bart.

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: Zoned block device regression test 11
  2018-08-22  4:32 ` Michael Kelley (EOSG)
@ 2018-08-22 15:47   ` Michael Kelley (EOSG)
  2018-08-22 15:48     ` Jens Axboe
  0 siblings, 1 reply; 11+ messages in thread
From: Michael Kelley (EOSG) @ 2018-08-22 15:47 UTC (permalink / raw)
  To: Bart Van Assche; +Cc: Jens Axboe, fio@vger.kernel.org, Damien Le Moal

Any special configuration needed to run this test?    Any options needs on
command line?  I figured out to enable CONFIG_BLK_DEV_ZONED, but
now I get the following on my 4.18.3 kernel:

root@MHKfio2:~# ./fio/t/zbd/run-tests-against-zoned-nullb
./fio/t/zbd/run-tests-against-zoned-nullb: line 27: ./fio/t/zbd/test-zbd-support: No such file or directory

/sys/kernel/config/nullb/nullb0/* and /dev/nullb0 are present as expected.

Michael

> -----Original Message-----
> From: Michael Kelley (EOSG)
> Sent: Tuesday, August 21, 2018 9:32 PM
> To: Bart Van Assche <bvanassche@acm.org>
> Cc: Jens Axboe <axboe@kernel.dk>; fio@vger.kernel.org; Damien Le Moal
> <damien.lemoal@wdc.com>
> Subject: RE: Zoned block device regression test 11
> 
> Yes, I'll take a look.
> 
> Michael
> 
> > -----Original Message-----
> > From: Bart Van Assche <bvanassche@acm.org>
> > Sent: Tuesday, August 21, 2018 5:11 PM
> > To: Michael Kelley (EOSG) <Michael.H.Kelley@microsoft.com>
> > Cc: Jens Axboe <axboe@kernel.dk>; fio@vger.kernel.org; Damien Le Moal
> > <damien.lemoal@wdc.com>
> > Subject: Zoned block device regression test 11
> >
> > Hello Michael,
> >
> > If I run the fio code from
> > https://github.com/bvanassche/fio against
> > Linux kernel v4.18 as follows:
> >
> > ./fio/t/zbd/run-tests-against-zoned-nullb
> >
> > then all tests except test number 11 pass. If I revert your axmap
> > patches however then test 11 passes. I think this means that your axmap
> > patches introduced a regression in the axmap code. Can you have a look
> > at this?
> >
> > Thanks,
> >
> > Bart.

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: Zoned block device regression test 11
  2018-08-22 15:47   ` Michael Kelley (EOSG)
@ 2018-08-22 15:48     ` Jens Axboe
  2018-08-22 16:54       ` Michael Kelley (EOSG)
  0 siblings, 1 reply; 11+ messages in thread
From: Jens Axboe @ 2018-08-22 15:48 UTC (permalink / raw)
  To: Michael Kelley (EOSG), Bart Van Assche
  Cc: fio@vger.kernel.org, Damien Le Moal

On 8/22/18 9:47 AM, Michael Kelley (EOSG) wrote:
> Any special configuration needed to run this test?    Any options needs on
> command line?  I figured out to enable CONFIG_BLK_DEV_ZONED, but
> now I get the following on my 4.18.3 kernel:
> 
> root@MHKfio2:~# ./fio/t/zbd/run-tests-against-zoned-nullb
> ./fio/t/zbd/run-tests-against-zoned-nullb: line 27: ./fio/t/zbd/test-zbd-support: No such file or directory
> 
> /sys/kernel/config/nullb/nullb0/* and /dev/nullb0 are present as expected.

You need current -git from Linus, the zoned null_blk support was added
after 4.18.

-- 
Jens Axboe



^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: Zoned block device regression test 11
  2018-08-22 15:48     ` Jens Axboe
@ 2018-08-22 16:54       ` Michael Kelley (EOSG)
  2018-08-22 17:04         ` Bart Van Assche
  2018-08-22 17:05         ` Jens Axboe
  0 siblings, 2 replies; 11+ messages in thread
From: Michael Kelley (EOSG) @ 2018-08-22 16:54 UTC (permalink / raw)
  To: Jens Axboe, Bart Van Assche; +Cc: fio@vger.kernel.org, Damien Le Moal

From: Jens Axboe <axboe@kernel.dk> Sent: Wednesday, August 22, 2018 8:49 AM
> 
> On 8/22/18 9:47 AM, Michael Kelley (EOSG) wrote:
> > Any special configuration needed to run this test?    Any options needs on
> > command line?  I figured out to enable CONFIG_BLK_DEV_ZONED, but
> > now I get the following on my 4.18.3 kernel:
> >
> > root@MHKfio2:~# ./fio/t/zbd/run-tests-against-zoned-nullb
> > ./fio/t/zbd/run-tests-against-zoned-nullb: line 27: ./fio/t/zbd/test-zbd-support: No such file or
> directory
> >
> > /sys/kernel/config/nullb/nullb0/* and /dev/nullb0 are present as expected.
> 
> You need current -git from Linus, the zoned null_blk support was added
> after 4.18.

So grab the linux-next tree, or are you referring to something else?  In my
4.18.3 tree, the source file drivers/block/null_blk_zoned.c is there and builds.

Michael

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: Zoned block device regression test 11
  2018-08-22 16:54       ` Michael Kelley (EOSG)
@ 2018-08-22 17:04         ` Bart Van Assche
  2018-08-22 19:04           ` Michael Kelley (EOSG)
  2018-08-22 17:05         ` Jens Axboe
  1 sibling, 1 reply; 11+ messages in thread
From: Bart Van Assche @ 2018-08-22 17:04 UTC (permalink / raw)
  To: Michael Kelley (EOSG), Jens Axboe, Bart Van Assche
  Cc: fio@vger.kernel.org, Damien Le Moal

On 08/22/18 09:54, Michael Kelley (EOSG) wrote:
> From: Jens Axboe <axboe@kernel.dk> Sent: Wednesday, August 22, 2018 8:49 AM
>>
>> On 8/22/18 9:47 AM, Michael Kelley (EOSG) wrote:
>>> Any special configuration needed to run this test?    Any options needs on
>>> command line?  I figured out to enable CONFIG_BLK_DEV_ZONED, but
>>> now I get the following on my 4.18.3 kernel:
>>>
>>> root@MHKfio2:~# ./fio/t/zbd/run-tests-against-zoned-nullb
>>> ./fio/t/zbd/run-tests-against-zoned-nullb: line 27: ./fio/t/zbd/test-zbd-support: No such file or
>> directory
>>>
>>> /sys/kernel/config/nullb/nullb0/* and /dev/nullb0 are present as expected.
>>
>> You need current -git from Linus, the zoned null_blk support was added
>> after 4.18.
> 
> So grab the linux-next tree, or are you referring to something else?  In my
> 4.18.3 tree, the source file drivers/block/null_blk_zoned.c is there and builds.

My advice is to grab Linus' latest tree 
(https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/). 
Linus' tree works fine on my test setups but I have not yet tested 
linux-next. You need the following two commits that are not yet in 
kernel v4.18:
* ca4b2a011948 ("null_blk: add zone support")
* 76f17d8ba1cb ("block: Rename the null_blk_mod kernel module back into 
null_blk")

Bart.



^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: Zoned block device regression test 11
  2018-08-22 16:54       ` Michael Kelley (EOSG)
  2018-08-22 17:04         ` Bart Van Assche
@ 2018-08-22 17:05         ` Jens Axboe
  1 sibling, 0 replies; 11+ messages in thread
From: Jens Axboe @ 2018-08-22 17:05 UTC (permalink / raw)
  To: Michael Kelley (EOSG), Bart Van Assche
  Cc: fio@vger.kernel.org, Damien Le Moal

On 8/22/18 10:54 AM, Michael Kelley (EOSG) wrote:
> From: Jens Axboe <axboe@kernel.dk> Sent: Wednesday, August 22, 2018 8:49 AM
>>
>> On 8/22/18 9:47 AM, Michael Kelley (EOSG) wrote:
>>> Any special configuration needed to run this test?    Any options needs on
>>> command line?  I figured out to enable CONFIG_BLK_DEV_ZONED, but
>>> now I get the following on my 4.18.3 kernel:
>>>
>>> root@MHKfio2:~# ./fio/t/zbd/run-tests-against-zoned-nullb
>>> ./fio/t/zbd/run-tests-against-zoned-nullb: line 27: ./fio/t/zbd/test-zbd-support: No such file or
>> directory
>>>
>>> /sys/kernel/config/nullb/nullb0/* and /dev/nullb0 are present as expected.
>>
>> You need current -git from Linus, the zoned null_blk support was added
>> after 4.18.
> 
> So grab the linux-next tree, or are you referring to something else?  In my
> 4.18.3 tree, the source file drivers/block/null_blk_zoned.c is there and builds.

Just grab Linus's tree, for instance. 4.18.3 does definitely not have support
for zoned null_blk, and null_blk_zoned.c should not exist. It was introduced
by this commit:

commit ca4b2a011948fae4e4d31490107db4926385a983
Author: Matias Bjørling <matias.bjorling@wdc.com>
Date:   Fri Jul 6 19:38:39 2018 +0200

    null_blk: add zone support

which is not in 4.18. So either your sources are messed up, or something else
funky is going on at your end.

-- 
Jens Axboe



^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: Zoned block device regression test 11
  2018-08-22 17:04         ` Bart Van Assche
@ 2018-08-22 19:04           ` Michael Kelley (EOSG)
  2018-08-22 19:54             ` Bart Van Assche
  2018-08-22 19:55             ` Damien Le Moal
  0 siblings, 2 replies; 11+ messages in thread
From: Michael Kelley (EOSG) @ 2018-08-22 19:04 UTC (permalink / raw)
  To: Bart Van Assche, Jens Axboe; +Cc: fio@vger.kernel.org, Damien Le Moal

One more question:  What's the preferred source for the zbc_* commands?
I built from https://github.com/hgst/libzbc, but get this runtime failure:

root@MHKfio2:~# zbc_report_zones /dev/nullb0
Device /dev/nullb0:
    Vendor ID: FAKE HGST HM libzbc
    Emulated zoned block device interface, Host-managed zone model
    2097152 512-bytes sectors
    262144 logical blocks of 4096 B
    262144 physical blocks of 4096 B
    1.074 GB capacity
    Read commands are restricted
    Maximum number of open sequential write required zones: 32
zbc_report_nr_zones at 0, ro 0x00 failed -6

Michael

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: Zoned block device regression test 11
  2018-08-22 19:04           ` Michael Kelley (EOSG)
@ 2018-08-22 19:54             ` Bart Van Assche
  2018-08-23  0:38               ` Michael Kelley (EOSG)
  2018-08-22 19:55             ` Damien Le Moal
  1 sibling, 1 reply; 11+ messages in thread
From: Bart Van Assche @ 2018-08-22 19:54 UTC (permalink / raw)
  To: Michael Kelley (EOSG), Bart Van Assche, Jens Axboe
  Cc: fio@vger.kernel.org, Damien Le Moal

On 08/22/18 12:04, Michael Kelley (EOSG) wrote:
> One more question:  What's the preferred source for the zbc_* commands?
> I built from https://github.com/hgst/libzbc, but get this runtime failure:
> 
> root@MHKfio2:~# zbc_report_zones /dev/nullb0
> Device /dev/nullb0:
>      Vendor ID: FAKE HGST HM libzbc
>      Emulated zoned block device interface, Host-managed zone model
>      2097152 512-bytes sectors
>      262144 logical blocks of 4096 B
>      262144 physical blocks of 4096 B
>      1.074 GB capacity
>      Read commands are restricted
>      Maximum number of open sequential write required zones: 32
> zbc_report_nr_zones at 0, ro 0x00 failed -6

Hello Michael,

Thank you for having started to look quickly into what I had reported. 
After I had reported the behavior change I started looking myself into 
this and I think that I have found the root cause (which is not in the 
axmap code). I will continue working on this and will keep you posted.

Bart.



^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: Zoned block device regression test 11
  2018-08-22 19:04           ` Michael Kelley (EOSG)
  2018-08-22 19:54             ` Bart Van Assche
@ 2018-08-22 19:55             ` Damien Le Moal
  1 sibling, 0 replies; 11+ messages in thread
From: Damien Le Moal @ 2018-08-22 19:55 UTC (permalink / raw)
  To: Michael Kelley (EOSG), Bart Van Assche, Jens Axboe; +Cc: fio@vger.kernel.org

Michael,

On 2018/08/22 12:05, Michael Kelley (EOSG) wrote:
> One more question:  What's the preferred source for the zbc_* commands?
> I built from https://github.com/hgst/libzbc, but get this runtime failure:
> 
> root@MHKfio2:~# zbc_report_zones /dev/nullb0
> Device /dev/nullb0:
>     Vendor ID: FAKE HGST HM libzbc
>     Emulated zoned block device interface, Host-managed zone model

The above line says "emulated" which means that libzbc detected a regular block
device and went into emulation mode. This means that your null_blk setup is not
supporting zoned mode or nullb0 was not created as a zoned device.
You can easily check if the nullb device created is zoned or not: just look into
/sys/block/nullb0/queue/zoned. It should say "host-managed". In your case, the
content is probably "none".

>     2097152 512-bytes sectors
>     262144 logical blocks of 4096 B
>     262144 physical blocks of 4096 B
>     1.074 GB capacity
>     Read commands are restricted
>     Maximum number of open sequential write required zones: 32
> zbc_report_nr_zones at 0, ro 0x00 failed -6
> 
> Michael
> 

On another note, you can also use the sysutils blkzone command to do the same:
blkzone report /dev/nullb0 will five you the zone list.

Best regards.

-- 
Damien Le Moal
Western Digital Research


^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: Zoned block device regression test 11
  2018-08-22 19:54             ` Bart Van Assche
@ 2018-08-23  0:38               ` Michael Kelley (EOSG)
  0 siblings, 0 replies; 11+ messages in thread
From: Michael Kelley (EOSG) @ 2018-08-23  0:38 UTC (permalink / raw)
  To: Bart Van Assche, Jens Axboe; +Cc: fio@vger.kernel.org, Damien Le Moal

From: Bart Van Assche <bvanassche@acm.org>  Sent: Wednesday, August 22, 2018 12:55 PM
>
> Hello Michael,
> 
> Thank you for having started to look quickly into what I had reported.
> After I had reported the behavior change I started looking myself into
> this and I think that I have found the root cause (which is not in the
> axmap code). I will continue working on this and will keep you posted.

I was hoping that might turn out to be the case. :-)   I'm actually out on
vacation the next three work days -- back on Tuesday, August 28th.  But
I'll be watching email in case something comes up that needs my
attention.   And I should do some learning about zoned block devices
and null block devices ....

Michael

^ permalink raw reply	[flat|nested] 11+ messages in thread

end of thread, other threads:[~2018-08-23  4:06 UTC | newest]

Thread overview: 11+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2018-08-22  0:10 Zoned block device regression test 11 Bart Van Assche
2018-08-22  4:32 ` Michael Kelley (EOSG)
2018-08-22 15:47   ` Michael Kelley (EOSG)
2018-08-22 15:48     ` Jens Axboe
2018-08-22 16:54       ` Michael Kelley (EOSG)
2018-08-22 17:04         ` Bart Van Assche
2018-08-22 19:04           ` Michael Kelley (EOSG)
2018-08-22 19:54             ` Bart Van Assche
2018-08-23  0:38               ` Michael Kelley (EOSG)
2018-08-22 19:55             ` Damien Le Moal
2018-08-22 17:05         ` Jens Axboe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox