[ovs-dev] [megaflow masklist fix V2] datapath: Fix a kernel crash caused by corrupted mask list.

Jesse Gross jesse at nicira.com
Sat Jun 22 01:10:27 UTC 2013


Do we need any of the tests for tbl->mask_list at this point if we
always allocate a list head?

On Fri, Jun 21, 2013 at 5:52 PM, Andy Zhou <azhou at nicira.com> wrote:
> O.K.  How about the attached incremental patch?
>
>
> On Fri, Jun 21, 2013 at 4:37 PM, Jesse Gross <jesse at nicira.com> wrote:
>>
>> On Fri, Jun 21, 2013 at 4:18 PM, Andy Zhou <azhou at nicira.com> wrote:
>> > When flow table is copied, the mask list from the old table
>> > is not properly copied into the new table. The corrupted mask
>> > list in the new table will lead to kernel crash. This patch
>> > fixes this bug.
>> >
>> > Bug #18110
>> > Reported-by: Justin Pettit <jpettit at nicira.com>
>>
>> Can we just allocate the mask list head when we create v1 of the
>> table? I think that would eliminate a lot of conditions and also allow
>> us to return -ENOMEM up front rather than silent failing later on.
>
>



More information about the dev mailing list