"Phelan, Michael" <[email protected]> writes:

>> -----Original Message-----
>> From: Aaron Conole <[email protected]>
>> Sent: Thursday, September 12, 2024 10:01 PM
>> To: Phelan, Michael <[email protected]>
>> Cc: [email protected]; Chaudron, Eelco <[email protected]>; Ilya
>> Maximets <[email protected]>
>> Subject: Re: [ovs-build] |fail| pw1977882 [ovs-dev, 3/7] dpif: Fix potential
>> NULL pointer access in log_flow_message().
>> 
>> "Phelan, Michael" <[email protected]> writes:
>> 
>> >> -----Original Message-----
>> >> From: Aaron Conole <[email protected]>
>> >> Sent: Tuesday, September 3, 2024 1:15 PM
>> >> To: Phelan, Michael <[email protected]>
>> >> Cc: [email protected]; Chaudron, Eelco <[email protected]>; Ilya
>> >> Maximets <[email protected]>
>> >> Subject: Re: [ovs-build] |fail| pw1977882 [ovs-dev, 3/7] dpif: Fix
>> >> potential NULL pointer access in log_flow_message().
>> >>
>> >> "Phelan, Michael" <[email protected]> writes:
>> >>
>> >> > Hi Aaron,
>> >> > I have looked into the failures on the CI and ran some tests
>> >> > manually and tests 1010, 1015, 1020, 1025, 1030 and 1035 are
>> >> > failing in make check even on the main branch.
>> >> >
>> >> > They fail in this block of calls from ofproto.at:
>> >> > CHECK_FLOW_MONITORING([1.0], [OpenFlow10], [],
>> >> NXST_FLOW_MONITOR)
>> >> > CHECK_FLOW_MONITORING([1.1], [OpenFlow11], [ (OF1.1)],
>> >> > NXST_FLOW_MONITOR) CHECK_FLOW_MONITORING([1.2],
>> >> [OpenFlow12], [
>> >> > (OF1.2)], NXST_FLOW_MONITOR) CHECK_FLOW_MONITORING([1.3],
>> >> > [OpenFlow13], [ (OF1.3)], ONFST_FLOW_MONITOR)
>> >> > CHECK_FLOW_MONITORING([1.4], [OpenFlow14], [ (OF1.4)],
>> >> > OFPST_FLOW_MONITOR) CHECK_FLOW_MONITORING([1.5],
>> >> [OpenFlow15], [
>> >> > (OF1.5)], OFPST_FLOW_MONITOR)
>> >>
>> >> Thanks for narrowing it down.
>> >>
>> >> > The error in the log files is:
>> >> > cat: ovs-ofctl.pid: No such file or directory
>> >>
>> >> Weird.  I wonder if there is some issue with the way the ovs-ofctl
>> >> pidfile is being referenced.  That error is likely from (ofproto.at):
>> >>
>> >>   on_exit 'kill `cat ovs-ofctl.pid`'
>> >>
>> >> Just a guess - could it be the way the autotools are expanding the
>> >> on_exit line and it is trying to evaluate the `cat ovs-ofctl.pid`
>> >> before ofctl has started?  Can you try moving those lines in ofproto.at 
>> >> just
>> to check if that could be it?
>> > Hi Aaron,
>> > Sorry for the delay in replying, I was OOO yesterday.
>> >
>> > I think ovs-ofctl is stopped before these checks are performed which is
>> causing the failure.
>> >
>> > The section above these checks is:
>> > OVS_APP_EXIT_AND_WAIT([ovs-ofctl])
>> > OVS_VSWITCHD_STOP
>> > AT_CLEANUP
>> > ])
>> >
>> > When I moved the CHECK_FLOW_MONITORING commands to before this
>> block, the tests passed.
>> 
>> Hrrm... That seems strange.  Those are part of the define.  I'm surprised 
>> that
>> worked and didn't end in forever recursion.
>> 
>> This might need some deeper looking.  I wonder if moving those simply
>> stopped running the checks.
>
> Sorry, yes looks like they're not checked after being moved.
>
>> 
>> Maybe there are some permissions or resource issues on your build system
>> that are preventing the needed files from being generated?
>
> I don't think they are any permissions restricting these files from being 
> generated.
>
> What resources would you suggest I check?

Is it possible someone can get access to debug it?  There could be some
issue with selinux(?), or with the /var/lib/jenkins/... stuff, or
perhaps an rlimit that is being cascaded to the ovs build system
children?

>> 
>> > Thanks,
>> > Michael.
>> >>
>> >> > Do you have any idea what might be the problem?
>> >> >
>> >> > Thanks,
>> >> > Michael.
>> >> >> -----Original Message-----
>> >> >> From: Phelan, Michael
>> >> >> Sent: Friday, August 30, 2024 4:12 PM
>> >> >> To: Aaron Conole <[email protected]>; [email protected]
>> >> >> Cc: Chaudron, Eelco <[email protected]>
>> >> >> Subject: RE: [ovs-build] |fail| pw1977882 [ovs-dev, 3/7] dpif: Fix
>> >> >> potential NULL pointer access in log_flow_message().
>> >> >>
>> >> >> Hi Aaron,
>> >> >> Yes I've noticed that jobs have been failing quite a lot since
>> >> >> make check was added to the tests.
>> >> >>
>> >> >> I will try to spend some time on this to debug and get back to you.
>> >> >>
>> >> >> Kind regards,
>> >> >> Michael.
>> >> >>
>> >> >> > -----Original Message-----
>> >> >> > From: Aaron Conole <[email protected]>
>> >> >> > Sent: Thursday, August 29, 2024 8:39 PM
>> >> >> > To: [email protected]
>> >> >> > Cc: Chaudron, Eelco <[email protected]>; Phelan, Michael
>> >> >> > <[email protected]>
>> >> >> > Subject: Re: [ovs-build] |fail| pw1977882 [ovs-dev, 3/7] dpif:
>> >> >> > Fix potential NULL pointer access in log_flow_message().
>> >> >> >
>> >> >> >
>> >> >> > Hi Michael,
>> >> >> >
>> >> >> > I'm seeing the most recent jobs failing with this error, and
>> >> >> > it's been there for a while.  Is there someone from Intel side
>> >> >> > who can help debug the Intel CI?
>> >> >> >
>> >> >> > -Aaron
>> >> >> >
>> >> >> > [email protected] writes:
>> >> >> >
>> >> >> > > Test-Label: intel-ovs-compilation
>> >> >> > > Test-Status: fail
>> >> >> > > http://patchwork.ozlabs.org/api/patches/1977882/
>> >> >> > >
>> >> >> > > AVX-512_compilation: failed
>> >> >> > > DPLCS Test: fail
>> >> >> > > DPIF Test: fail
>> >> >> > > MFEX Test: fail
>> >> >> > > Actions Test: fail
>> >> >> > > Errors in DPCLS test:
>> >> >> > > make -j CFLAGS=-g -O0 -march=native make  all-am
>> >> >>>><snip>

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to