Hi,
On 8/13/25 12:34 PM, Michael Tokarev wrote:
> On 13.08.2025 13:26, Paolo Abeni wrote:
>
>> +typedef struct NetOffloads {
>> + bool csum;
>> + bool tso4;
>> + bool tso6;
>> + bool ecn;
>> + bool ufo;
>> + bool uso4;
>> + bool uso6;
>> +} NetOffloads;
>
> BTW, can't this be a bitmask (single integer) or a bitfield?
Yes, it could. It has been discussed here:
https://lists.gnu.org/archive/html/qemu-devel/2025-07/msg02998.html
I prefer to avoid the (virtio-define-based) bitmask as personal preference.
> Dunno how well it'll work, but it feels like this construct:
>
> NetOffloads ol = {.csum = cso_state };
>
> and similar should do a large(ish) memzero() which kind
> of defeats the purpose of an offload (for speed).
Note that this is the "control path" any minimal (if measurable at all):
overhead will not impact the actual data transfer.
Cheers,
Paolo
