[ovs-dev] [PATCH v2 3/3] travis: Enable OvS Travis CI for arm

Yanqin Wei Yanqin.Wei at arm.com
Thu Mar 12 09:38:40 UTC 2020


Thanks for the feedback. Replied in line.

> -----Original Message-----
> From: Ilya Maximets <i.maximets at ovn.org>
> Sent: Thursday, March 12, 2020 12:20 AM
> To: Yanqin Wei <Yanqin.Wei at arm.com>; Ilya Maximets
> <i.maximets at ovn.org>; ovs-dev at openvswitch.org; blp at ovn.org; Lance Yang
> <Lance.Yang at arm.com>
> Cc: dwilder at us.ibm.com; Gavin Hu <Gavin.Hu at arm.com>; Ruifeng Wang
> <Ruifeng.Wang at arm.com>; Jieqiang Wang <Jieqiang.Wang at arm.com>;
> Malvika Gupta <Malvika.Gupta at arm.com>; nd <nd at arm.com>
> Subject: Re: [ovs-dev] [PATCH v2 3/3] travis: Enable OvS Travis CI for arm
> 
> On 3/11/20 5:57 AM, Yanqin Wei wrote:
> > Hi Ilya,
> >
> > This patch has been in the review pipeline for some time. It runs stable on
> our internal repo more than two months.
> > Could you give us some suggestion about the next action I can take to speed
> up the merge of this patch?
> 
> Hi.  Sorry for things taking so long.
> I have this patch in my backlog for this or next week.
> 
> The main concern right know is possible significant increase of the checking
> time.  Are you sure that we need all the listed jobs?
> Do you expect some arm64 specific issues on the linking stage?
> I mean, maybe we could reduce number of different combinations of "shared"
> flags.  I had no chance to run this, so I don't know how much time these jobs
> really takes and what is the total time difference.

[Yanqin]  This is latest build report for x86 and arm.  Sparse is disabled here because it has some compiling issue for both x86 and arm today.
X86 and Arm https://travis-ci.com/github/MarcoDWei/ovs/builds/152933388?utm_medium=notification&utm_source=email
Ran for 58 min 3 sec/ Total time 4 hrs 12 min 24 sec
X86 only https://travis-ci.com/github/MarcoDWei/ovs/builds/152942934?utm_medium=notification&utm_source=email
Ran for 38 min 40 sec /Total time 2 hrs 55 min 4 sec

The total time increases around 1hr 17min for SIX new arm jobs. Running time increases around 20 mins. Kernel datapath jobs look most time consuming jobs.
OPTS="--disable-ssl"                                            4 min 31 sec 
KERNEL_LIST="5.0 4.20 4.19 4.18 4.17 4.16" 22 min 39 sec 
KERNEL_LIST="4.15 4.14 4.9 4.4 3.16"            17 min 34 sec 
DPDK=1 OPTS="--enable-shared"                    11 min 17 sec 
DPDK_SHARED=1                                               10 min 35 sec 
DPDK_SHARED=1 OPTS="--enable-shared"   11 min 21 sec

I agree with you to remove some shared tag combination because they have low risk of CPU specific issues in linking stage.
Moreover we could chose a part of kernel version for ARM jobs, which could significantly reduce total time and running time.  The running time is expected to increase around 10 minutes.
OPTS="--disable-ssl"                                            
KERNEL_LIST="5.0 4.20 4.16 4.9 3.16" 
DPDK=1 OPTS="--enable-shared"                    
DPDK_SHARED=1                                               
	
> 
> Best regards, Ilya Maximets.
> 
> 
> >
> > Best Regards,
> > Wei Yanqin
> >
> >> -----Original Message-----
> >> From: Lance Yang <Lance.Yang at arm.com>
> >> Sent: Tuesday, January 21, 2020 9:06 AM
> >> To: Ilya Maximets <i.maximets at ovn.org>; ovs-dev at openvswitch.org
> >> Cc: blp at ovn.org; Yanqin Wei <Yanqin.Wei at arm.com>;
> dwilder at us.ibm.com;
> >> Gavin Hu <Gavin.Hu at arm.com>; Ruifeng Wang <Ruifeng.Wang at arm.com>;
> >> Jieqiang Wang <Jieqiang.Wang at arm.com>; Malvika Gupta
> >> <Malvika.Gupta at arm.com>; nd <nd at arm.com>
> >> Subject: RE: [ovs-dev] [PATCH v2 3/3] travis: Enable OvS Travis CI
> >> for arm
> >>
> >>
> >>> -----Original Message-----
> >>> From: Ilya Maximets <i.maximets at ovn.org>
> >>> Sent: Saturday, December 7, 2019 12:39 AM
> >>> To: Lance Yang (Arm Technology China) <Lance.Yang at arm.com>; ovs-
> >>> dev at openvswitch.org
> >>> Cc: i.maximets at ovn.org; blp at ovn.org; Yanqin Wei (Arm Technology
> >>> China) <Yanqin.Wei at arm.com>; dwilder at us.ibm.com; Gavin Hu (Arm
> >>> Technology
> >>> China) <Gavin.Hu at arm.com>; Ruifeng Wang (Arm Technology China)
> >>> <Ruifeng.Wang at arm.com>; Jieqiang Wang (Arm Technology China)
> >>> <Jieqiang.Wang at arm.com>; Malvika Gupta <Malvika.Gupta at arm.com>;
> nd
> >>> <nd at arm.com>
> >>> Subject: Re: [ovs-dev] [PATCH v2 3/3] travis: Enable OvS Travis CI
> >>> for arm
> >>>
> >>> On 06.12.2019 04:26, Lance Yang wrote:
> >>>> Enable part of travis jobs with gcc compiler for arm64 architecture
> >>>>
> >>>> 1. Add arm jobs into the matrix in .travis.yml configuration file 2.
> >>>> To enable OVS-DPDK jobs, set the build target according to
> >>>> different CPU architectures 3. Temporarily disable sparse checker
> >>>> because of static code checking failure on arm64
> >>>>
> >>>> Successful travis build jobs report:
> >>>> https://travis-ci.org/yzyuestc/ovs/builds/621037339
> >>>>
> >>>> Reviewed-by: Yanqin Wei <Yanqin.Wei at arm.com>
> >>>> Reviewed-by: Jieqiang Wang <Jieqiang.Wang at arm.com>
> >>>> Reviewed-by: Gavin Hu <Gavin.Hu at arm.com>
> >>>> Signed-off-by: Lance Yang <Lance.Yang at arm.com>
> >>>> ---
> >>>
> >>> Compiler crashed while building DPDK:
> >>>
> >>> /home/travis/build/ovsrobot/ovs/dpdk-dir/drivers/net/ixgbe/ixgbe_pf.c:
> >>> In function
> >>> ‘ixgbe_pf_host_configure’:
> >>> /home/travis/build/ovsrobot/ovs/dpdk-dir/drivers/net/ixgbe/ixgbe_pf.c:
> >>> 297:1: internal compiler error: Segmentation fault
> >>>
> >>> https://travis-ci.org/ovsrobot/ovs/jobs/621434216#L1999
> >>>
> >>> This is not good.
> >>> Need to check how frequently this happens.
> >>>
> >>> Best regards, Ilya Maximets.
> >> [Lance]
> >> Hi Ilya,
> >>
> >> After you give us the feedback about the segmentation fault issue, we
> >> keep running the travis CI to observe the frequency. We run those arm
> >> jobs on a regular basis and clear cache on every single build. We
> >> have run the build at least 2 times a day for more than a month.
> >>
> >> The good thing is that we haven't reproduce the segfault issue. Job
> >> reports are available at:
> >> https://travis-ci.org/yzyuestc/ovs/builds/639064033
> >>
> >> Travis CI community also made some adjustment on their side, you can
> >> see their reply here:
> >> https://travis-ci.community/t/segfaults-in-arm64-
> >> environment/5617/13
> >>
> >> I think the segfault issue may occur by chance. At least, it is not a
> >> frequent one, which is unlikely to cause issues to OVS developers.
> >> Could you please check it again?
> >>
> >> Best Regards,
> >> Lance
> >



More information about the dev mailing list