On 2017年12月20日 16:58, Mark Cave-Ayland wrote:
Series looks good to me.
A small question is that, is this better to keep compute_mcast_idx()?
Thanks
Hi Jason,
I did think about this, however at the very minimum you'd need
big-endian and little-endian variants of compute_mcast_idx(), and then
you see that eepro100 applies a different bitmask/shift which is yet
another variant...
For this reason I moved them all inline to the QEMU driver and that
made it possible to compare the hash calculation directly with the
corresponding Linux driver which found the 3 potential bugs above. So
I think this is a net win (pardon the pun) on all sides :)
ATB,
Ok, applied.
Thanks