On 11/08/15 01:06, David Gwynne wrote:
On 8 Nov 2015, at 8:23 AM, Miod Vallat wrote:
noone has a vr?
You can't expect people to use the crappiest Ethernet chip ever
designed.
theyre in the alix, surely someone has those still.
i thought le(4) was the worst.
Finally tested this patch
On 2015/11/08 10:45, Miod Vallat wrote:
> > i thought le(4) was the worst.
>
> Of course not. le(4) is slow and has some quirks (and the earliest ones
> can't do multicast correctly), but at least they don't collapse into fetal
> position under load and don't need a reset to recover.
I see you to
> i thought le(4) was the worst.
Of course not. le(4) is slow and has some quirks (and the earliest ones
can't do multicast correctly), but at least they don't collapse into fetal
position under load and don't need a reset to recover.
> On 8 Nov 2015, at 8:23 AM, Miod Vallat wrote:
>
>> noone has a vr?
>
> You can't expect people to use the crappiest Ethernet chip ever
> designed.
theyre in the alix, surely someone has those still.
i thought le(4) was the worst.
> noone has a vr?
You can't expect people to use the crappiest Ethernet chip ever
designed.
o requeue the packet so the next call to the start
>>> routine can try fitting it on the ring again.
>>>
>>> this introduces IFQ_PREPEND (cause we currently have IF_PREPEND)
>>> and works on top of both hfsc and priq because i added hfsc_requeue
>>> a while
to IFQ_DEQUEUE will
> > cause unwanted packet loss when encapsulation fails in some cases,
> > such as when the tx ring is already full. to cope, the easiest
> > solution is to requeue the packet so the next call to the start
> > routine can try fitting it on the ring agai
he tx ring is already full. to cope, the easiest
> solution is to requeue the packet so the next call to the start
> routine can try fitting it on the ring again.
>
> this introduces IFQ_PREPEND (cause we currently have IF_PREPEND)
> and works on top of both hfsc and priq because i
call to the start
routine can try fitting it on the ring again.
this introduces IFQ_PREPEND (cause we currently have IF_PREPEND)
and works on top of both hfsc and priq because i added hfsc_requeue
a while back.
this also converts uses of IF_PREPEND in drivers to IFQ_PREPEND.
this improves the