On 10/3/2019 6:08 AM, Ben Pfaff wrote:
On Thu, Oct 03, 2019 at 08:35:42AM -0700, Greg Rose wrote:
Two commits recently fixed an issue with setting the corrrect l3 and l4
flow information when conntrack reassembles packet fragments.
c98f776 datapath: Clear the L4 portion of the key for "later" fragments
2609173 datapath: Properly set L4 keys on "later" IP fragments
This test checks for regressions that might break this feature. It
counts on the fact that when the bug is present the udp src port
will not be correct. It will either be zero or else some other
garbage value. So the test feeds some fragments through for
reassembly and then checks to make sure that the udp srce port
is actually the correct value of 5001.
Tested by reverting the above commits and observing that the test
then fails.
Signed-off-by: Greg Rose <[email protected]>
This patch doesn't apply, with:
error: corrupt patch at line 43
I think that it's probably because this has very long lines. Email
doesn't do well with lines over about 1000 bytes long. I know of two
ways to avoid the problem: use pull requests instead of email, or
shorten the lines, e.g. by splicing with \ or dnl. The latter might be
a little better because it means that patches that later update or
correct the tests will pass through email OK.
Will you try one of these?
Hmm, yes. git send-email complained about that. OK, let me try
reformatting the patch so it works
better with email.
- Greg
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
