On Wed, Sep 25, 2024 at 09:27:03AM +0200, Laurent Vivier wrote:On 25/09/2024 09:01, David Gibson wrote:It probably shouldn't be an ASSERT(); that would indicate hitting it was a bug in past, whereas IIUC here it would be triggered by strange behaviour from the guest or qemu. I think a die() would suffice, though. Of course, just resetting the tap connection or even just dropping that flow would be even better, but given the unlikeliness of the event, I think die() would be fine. -- David Gibson (he or they) | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you, not the other way | around. http://www.ozlabs.org/~dgibsonDo you think an ASSERT() is enough?I don't want to write complicated code only to write the checksum of the tcp header.I agree, but I think we should at least test and bail with an error message if our assumptions about the alignments of the IOVs were given aren't true.