Hi Anand I tried applying the above patch on the master branch and after applying the patch I was able to compile vrouter.ko and contrail-vrouter-agent successfully.
Hoewever when I tried ruunning the modified contrail-vrouter-agent with the modified vrouter.ko,the binary was not able to run properly and starting producing alignment errors.Below are the dmesg logs.. vrouter soft reset start [ 324.645142] vrouter soft reset done (0) [ 326.649461] Unhandled fault: alignment fault (0x92000021) at 0x0000ffff87f0003f So I tried running the default contrail-vrouter-agent with the modified (patch applied) vrouter.ko and I was able to run contrail-vrouter-agent properly.However I was getting the same kernel crash while pinging from the VM PFA kernel crash logs .. --Thanks Abhishek Jain On Mon, Dec 28, 2015 at 10:56 AM, Anand H Krishnan <[email protected]> wrote: > Can you try this patch? I am assuming that you are on the mainline branch. > > > Cheers, > > Anand > > > PS: Please recompile agent too after applying the following patch. > > > *diff --git a/include/vr_flow.h b/include/vr_flow.h* > > *index 09853e7..ee00ed1 100644* > > *--- a/include/vr_flow.h* > > *+++ b/include/vr_flow.h* > > @@ -288,10 +288,10 @@ struct vr_flow_queue { > > > > struct vr_dummy_flow_entry { > > struct vr_flow fe_key; > > - uint8_t fe_key_packing; > > - uint16_t fe_tcp_flags; > > unsigned int fe_tcp_seq; > > struct vr_flow_queue *fe_hold_list; > > + uint16_t fe_tcp_flags; > > + uint16_t fe_packing; > > unsigned short fe_action; > > unsigned short fe_flags; > > int fe_rflow; > > @@ -312,10 +312,10 @@ struct vr_dummy_flow_entry { > > /* do not change. any field positions as it might lead to incompatibility > */ > > struct vr_flow_entry { > > struct vr_flow fe_key; > > - uint8_t fe_key_packing; > > - uint16_t fe_tcp_flags; > > unsigned int fe_tcp_seq; > > struct vr_flow_queue *fe_hold_list; > > + uint16_t fe_tcp_flags; > > + uint16_t fe_packing; > > unsigned short fe_action; > > unsigned short fe_flags; > > int fe_rflow; > > > Thanks, > > Anand > > > ------------------------------ > *From:* abhishek jain <[email protected]> > *Sent:* Saturday, December 26, 2015 10:11 AM > *To:* Anand H Krishnan > *Cc:* [email protected]; Rajagopalan Sivaramakrishnan; Prabhjot > Singh Sethi > *Subject:* Re: [opencontrail-dev] kernel crash when ping > > Hi Anand > > Thanks for looking into this. > > I applied some prints in the same function to be more clear on the exact > reason and the exact problem seems to be is with > *&fe->fe_hold_list.* > > func: vr_flush_entry line: 1456 *&fe->fe_hold_list*: *ffff8083010f4c34* > { vfq: ffff80007b339640 > [ 249.477176] Unhandled fault: alignment fault (0x96000021) at > *0xffff8083010f4c34 > * > [ 249.484494] Internal error: : 96000021 [#1] PREEMPT SMP > [ 249.489717] Modules linked in: vrouter(O) > > > What could be the reason for this? > > -Thanks > Abhishek Jain > > > > > On Sat, Dec 26, 2015 at 8:51 AM, Anand H Krishnan <[email protected]> > wrote: > >> Hi Abhishek, >> >> >> We do not have an arm system to analyze the core. Can you please do a >> >> bit more analysis (looks like you already did most of the initial digging >> >> around)? >> >> >> The place where you saw a crash seems highly unlikely. There are a couple >> >> of possibilities. &fe->fe_hold_list is invalid or __sync* stuff doesn't >> work >> >> on arm (guessing). Can you analyze the kernel crash dump? >> >> >> Cheers, >> >> Anand >> >> >> ------------------------------ >> *From:* Dev <[email protected]> on behalf of abhishek >> jain <[email protected]> >> *Sent:* Thursday, December 24, 2015 12:05 AM >> *To:* [email protected]; Rajagopalan Sivaramakrishnan; Prabhjot >> Singh Sethi >> *Subject:* [opencontrail-dev] kernel crash when ping >> >> >> Hi Team >> >> I'm running ARM 64 as a compute node (4.0.4 kernel version) with contrail >> vrouter-agent running on it.The compute node is able to make connection to >> the controller node and I'm able to launch VM on compute node from the >> controller node(Opencontrail) GUI. >> >> The VM on compute node is able obtain IP address from the virtual DHCP >> server running on controller node.However I'm facing kernel crash when I'm >> pinging the virtual GW from the VM. >> >> I'm using the latest opencontrail code.I'm getting kernel crash at >> function vr_flush_entry at below line.. >> >> swapped = __sync_bool_compare_and_swap(&fe->fe_hold_list, vfq, NULL); >> >> Please help me regarding this. >> >> --Thanks >> Abhishek Jain >> > >
kernel-crash-logs
Description: Binary data
_______________________________________________ Dev mailing list [email protected] http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org
