Steven Stallion wrote:
> Garrett D'Amore wrote:
>
>> Typically you set up two systems. The first system has your dnet device
>> in it. The second should have a "known good" device. I'd actually
>> prefer your partner was a gigabit device, but the Centaur should work OK
>> too.
>>
>
> Shouldn't be a problem; I was planning on testing with a dedicated bge
> interface on the other host.
>
> Any preference on a switch? At the moment the dnet device is on an SMC
> 10/100, but I also have a couple of different gigabit switches available.
>
>
Reading the code review, it looks pretty good. Two things stood out though:
1) enable_interurpts() used to conditionally enable tx interrutps. Now
its unconditional. Its probably for the best (simpler logic, less
likely to result in stalls, etc.) You might want to note something
somewhere, though, as it potentially has a small performance impact.
2) much worse, your transmit logic only seems to transmit a single
packet linked by mp->b_cont. With GLDv3, you can get multiple packets,
linked together with mp->b_next. You need to have some logic to handle
this in a loop, or somesuch. If you can't find any example code, let me
know.
Item #1 is a small issue. Item #2 needs to be fixed before putback.
Have you had a chance to do any testing yet?
-- Garrett
_______________________________________________
driver-discuss mailing list
driver-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/driver-discuss
