You might want to try mounting the nfs filesystem with an 8k or
	16k block size rather than the (i386) default of 32k.

	mount_nfs -r 8192 remote:/dir /local_dir

On Sat, 7 Jun 2003, Richard Rauch wrote:

> I have two non-critical problems with my ne-using PCMCIA ethernet card.
> The first is one that I've mentioned before; it seems that near the end
> of the last function in the PCMCIA ne driver .c file, the if() test fails
> and {rv} is never set to 0.  This causes the device to fail to probe/attach.
> Setting the {rv} forcefully to 0 solves this problem.  (There's a PR on this
> with the hack pointed out.)  I am curious if anyone can explain:
>
>  (a) What the test is trying to accomplish.
>
>  (b) How difficult it would be to introduce a driver quirk entry to cause
>      my particular PCMCIA card to get around this test without removing the
>      (presumably useful) test for other cards.
>
>
> The second problem is: While effectively removing the test gets rid of the
> problem for me, I tend to get a bunch of
>
>   /netbsd ne0: warning - receiver ring buffer overrun
>
> ...error messages.  These seem to be coming from .../src/dev/ic/dp8390.c
>
> One thing that I notice is that in the ne2000.c file, there's a
> /* XXX really? */ next to the claim that the 88190 chipset has 16K
> memory (this is the type that my card comes up as).  Could it be that my
> card only has 8K memory, hence the overruns?  (No, I haven't yet tried
> just changing that line, but I thought that I'd ask...I assume that there
> is little harm in moving that line to the 8K case.  I guess that I'll go
> build a test kernel after I send this.  (^&)
>
> The overruns cause a certain amount of lost network performance, so it
> would be nice to fix this if possible.  (I mostly notice this in XMMS,
> playing from an NFS-mounted directory of Ogg Vorbis files.  Directory
> scans are slow, and periodically playing files pause for about a second
> when an overrun occurs.)
>
> Or, is this just a "known problem" with this chipset and I'll just have
> to live with overruns or get a new NIC?
>
>
> Well, off to hack the driver and see if an 8K buffer fixes my problem.  (^&
>
> Thanks in advance.
>
>
>

-- 
		David/absolute          -- www.netbsd.org: No hype required --