On 2026/01/16 07:35, David Gwynne wrote:
> there's an unstated assumption that the hardware (eg, cpu and ram) a
> hypervisor and the guests run on is ok, and the software is correct.
> 
> if you take a step back, the idea behind the checksums in packets
> is to help identify if a packet has been damaged as it travels
> across a medium like a network cable in an ethernet network.

not relevant to hypervisors in particular, but also checks that it
didn't get damaged during processing (nic/router) or on a bus involved
between nic and cpu (pci, etc)

checksum offload is a trade-off between speeding things up in some
cases and missing potential damage in others - I probably don't have any
of those H8SSL any more (they have issues with some quad PCI-X NICs
with onboard bridges) but I certainly learned this the hard way ;)