Re: 2.6.0 NFS-server low to 0 performance

[bill davidsen]

In article <Pine.LNX.4.44.0401071431520.479-100000@xxxxxxxxxxxxx>,
Guennadi Liakhovetski  <g.liakhovetski@xxxxxx> wrote:
| On Tue, 6 Jan 2004, Guennadi Liakhovetski wrote:
| 
| > server with 2.6.0 kernel:
| >
| > fast:2.6.0-test11	2m21s (*)
| > fast:2.4.20		16.5s
| > SA1100:2.4		never finishes (*)
| > PXA:2.4.21-rmk1-pxa1	as above
| > PXA:2.6.0-rmk1-pxa	as above
| >
| > server: 2.4.21
| >
| > fast:2.6.0-test11	6s
| > fast:2.4.20		5s
| > SA1100:2.4.19-rmk7	3.22s
| > PXA:2.4.21-rmk1-pxa1	7s
| > PXA:2.6.0-rmk2-pxa	1) 50s (**)
| > (***)			2) 27s (**)
| 
| s/fast/PC2/
| 
| Further, I tried the old 3c59x card - same problems persist. Also tried
| PC2 as the server - same. nfs-utils version 1.0.6 (Debian Sarge). I sent a
| copy of the yesterday's email + new details to nfs@xxxxxxxxxxxxxxxxxxxxx,
| netdev@xxxxxxxxxxx, linux-net@xxxxxxxxxxxxxxxx
| 
| Strange, that nobody is seeing this problem, but it looks pretty bad here.
| Unless I missed some necessary update somewhere? The only one that seemed
| relevant - nfs-utils on the server(s) from Documentation/Changes I
| checked.

I'm sure you checked this, but does mii-tool show that you have
negotiated the proper connection to the hub or switch? I found that my
3cXXX and eepro100 cards were negotiating half duplex with the switches
and cable modems, causing the throughput to go forth and conjugate the
verb "to suck" until I fixed it.
-- 
bill davidsen <davidsen@xxxxxxx>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/