[ovs-dev] [PATCH] ovn-controller: Fix crash when sending GARP when openflow disconnection.

Ben Pfaff blp at ovn.org
Thu Feb 15 00:41:06 UTC 2018


On Mon, Feb 12, 2018 at 04:17:47PM +0800, Guoshuai Li wrote:
> This is call stack:
> Program received signal SIGABRT, Aborted.
> 1  0x00007ffff6a4f8e8 in __GI_abort () at abort.c:90
> 2  0x00000000004765d6 in ofputil_protocol_to_ofp_version (protocol=<optimized out>) at lib/ofp-util.c:769
> 3  0x000000000047c19e in ofputil_encode_packet_out (po=po at entry=0x7fffffffa0e0, protocol=<optimized out>) at lib/ofp-util.c:7060
> 4  0x0000000000410870 in send_garp (garp=0x83cfe0, current_time=current_time at entry=1200375400) at ovn/controller/pinctrl.c:1738
> 5  0x000000000041430f in send_garp_run (active_tunnels=<optimized out>, local_datapaths=0x7fffffffc0a0, chassis_index=<optimized out>, chassis=0x8194d0, br_int=<optimized out>, ctx=0x7fffffffc080) at ovn/controller/pinctrl.c:2069
> 
> Signed-off-by: Guoshuai Li <ligs at dtdream.com>

Thanks for fixing the bug!

I think we should not log a warning about this.  It is pretty much
unavoidable that sometimes the controller will be disconnected from the
switch, especially at controller startup.  WARN level log messages
should be reserved for situations that are likely to indicate a problem.
It might make sense to log something at INFO level, but the code for
maintaining OpenFlow connections already logs plenty of INFO messages
about connecting and disconnecting, so I don't think we really need
another one here.

Like Mark, I am not sure that the test is necessary.

Will you submit a v2?

Thanks,

Ben.


More information about the dev mailing list