Is 3c996b-t gigabit ethernet NIC supported with the
tigon3 driver? The driver in the kernel lacks documentation
as too what it supports, at least I couldn't find any.
On another note, 3com provides GPL drivers for this card,
is there a reason why these cannot be merged into the kernel?
Or is it basically what the tigon3 is? Has anyone tried these
drivers out and found out which performs better? Should I
stick with what 3com provides or go with what is in the kernel
if it works?
Thanks,
Steve
It is just that I was reading this page where they evaluate
gigi cards and they used the drivers from 3com.
http://www.cs.uni.edu/~gray/gig-over-copper/
The 3com drivers are open-source though, which makes
me less hesitent. And the tigon3 drivers lack of documentation
kind of irks me. I would however like to stick with a mainstream kernel
driver cause the support is there if it is needed. I guess I will give
both versions a whirl. From just glancing at the source of both, it seems
that both versions support jumbo frames which is really what I am after.
Steve
Mark Hahn wrote:
>>drivers out and found out which performs better? Should I
>>stick with what 3com provides or go with what is in the kernel
>>if it works?
>>
>>
>
>several good kernel hackers maintain, tune and polish the kernel versions.
>I wouldn't even consider the stuff from the vendor.
>
>
>
>
>
>>>>> "Steve" == kwijibo <[email protected]> writes:
Steve> http://www.cs.uni.edu/~gray/gig-over-copper/
Steve> The 3com drivers are open-source though, which makes me less
Steve> hesitent. And the tigon3 drivers lack of documentation kind of
Steve> irks me. I would however like to stick with a mainstream
Steve> kernel driver cause the support is there if it is needed. I
Steve> guess I will give both versions a whirl. From just glancing at
Steve> the source of both, it seems that both versions support jumbo
Steve> frames which is really what I am after.
Well did you look at the code? It's the infamous Broadcom bcm5700
driver which is probably the worst driver code we have seen in the
Linux community for the last 5 years. Sure you can run it, but don't
come back and complain when you run into trouble. You will be a lot
better off using the tg3 driver, or better yet, getting a NIC thats
less buggy.
Cheers,
Jes
Jes Sorensen wrote:
>>>>>>"Steve" == kwijibo <[email protected]> writes:
>>>>>>
>>>>>>
>
>Steve> http://www.cs.uni.edu/~gray/gig-over-copper/
>
>Steve> The 3com drivers are open-source though, which makes me less
>Steve> hesitent. And the tigon3 drivers lack of documentation kind of
>Steve> irks me. I would however like to stick with a mainstream
>Steve> kernel driver cause the support is there if it is needed. I
>Steve> guess I will give both versions a whirl. From just glancing at
>Steve> the source of both, it seems that both versions support jumbo
>Steve> frames which is really what I am after.
>
>Well did you look at the code? It's the infamous Broadcom bcm5700
>
Just glanced at it.
>driver which is probably the worst driver code we have seen in the
>Linux community for the last 5 years. Sure you can run it, but don't
>come back and complain when you run into trouble. You will be a lot
>better off using the tg3 driver, or better yet, getting a NIC thats
>less buggy.
>
>
Is the NIC hardware buggy or is it just the bcm5700 drivers?
Well that answers the question I was after, I will use the
tg3 kernel drivers. I chose this card cause it had good
performance tests on that web page. The SK9821 beat
all the gigi NICs but it also costs about 4 times as much
as any other.
Steve
>>>>> "Steve" == kwijibo <[email protected]> writes:
Steve> Jes Sorensen wrote:
>> driver which is probably the worst driver code we have seen in the
>> Linux community for the last 5 years. Sure you can run it, but
>> don't come back and complain when you run into trouble. You will be
>> a lot better off using the tg3 driver, or better yet, getting a NIC
>> thats less buggy.
Steve> Is the NIC hardware buggy or is it just the bcm5700 drivers?
The hardware is buggy. The bcm5700 drivers are written in such
terrible C, you can in fact argue whether you'd call it C, that it's
totally impossible to debug the code anyway. Who knows whats hidden in
there.
Jes