Re: NFS on 2.6.0-test9:

From: Voluspa
Date: Mon Nov 03 2003 - 12:12:28 EST

First, pardon for the silly typo : that changed the subject.

On 03 Nov 2003 11:09:54 -0500
Trond Myklebust wrote:

> >>>>> " " == lista2 <Voluspa> writes:
> > Using TCP or DIRECTIO in -test9 makes no difference. Here's the
> > relevant .config-section:
> So where *does* the big jump from 3 minutes to 29 minutes occur? Are

Not a clue. Gave up on NFS when nothing had changed in -test4 and -test5
(the first regression). This weekend I spent compiling to find the
'readahead' values. Am on a PII 400, so takes time to nail specifics.
Can't do another marathon compile until next weekend.

> those numbers on TCP or are they UDP only?

Am rather clueless when it comes to networking. You have to deduce from
the following. Remote is a Pentium 120 with Slackware 7.1.0 (Universal
NFS Server 2.2beta47) kernel 2.2.16. Does it even speak TCP? I compiled
this -test9 with TCP only when I saw the 29 minutes - and no change
afterwards. Looks like I'm using UDP, or?

> What NIC are you using? From the error logs you showed us, it looked
> like it might be a de4x5 driver. Have you tried using the tulip
> driver?

On this -test9 it is an integrated NIC which needs the de4x5 (Compaq
Presario 5640). When trying out Slackware 9.something it was
autodetected as a tulip but didn't work with that driver. The P120 uses
an ISA card driven by a cs89x0 module.

> Are you using ACPI to set up the hardware? If so, does turning it off
> make a difference.

Kernels are compiled without everything in that area. Not even the older
power management is compiled in.

> Does 'netstat -s' offer any other hints?

As I said, clueless about networking. So I'll have to use the weekend
for more info-gathering.

Mats Johannesson
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at