Bug#452069: nfs-kernel-server: broken NFS server on Thecus N2100
On Sun, Dec 09, 2007 at 06:48:42PM -0700, Martin Michlmayr wrote:
> * dann frazier <dannf@debian.org> [2007-11-22 10:10]:
> > > dannf, could you please take a look at #452069 and let me know if this
> > > bug is worth addressing in a stable update.
> > Yes, this certainly qualifies - I've updated the usertags to reflect
> > that. Thanks for tracking this one down.
>
> I prepared the patch below which works for me and which according to
> upstream shouldn't break anything.
ok
> On the one hand, he agreed it
> shouldn't impact performance, but on the other hand he mentioned:
>
> | If the driver does not feel as fast as expected, try
> | e9f63f30863fd778a5329e93c7e2208b9bcb5b79 first. No known downside so
> | far.
>
> | If the speed goes brutally down with nfs/smb, give a try to
> | d78ae2dcc2acebb9a1048278f47f762c069db75c
>
> So I'm not quite sure whether to apply the patch or whether we need
> verification that it won't slow down things.
well, slower is certainly better than corruption :)
My suggestion would be to provide a build and see if people measure
any significant changes. netperf might be useful here (but its in
non-free due to a non-commercial usage clause).
> I also asked what other patches would be of interest to backport and
> got:
>
> | Let aside the "align" fixes, the short list below contains some candidates
> | in reverse order:
>
> | 315917d23fdd20a0f4ff99b9228de5840d9d276c
don't have this one
> | 9cb427b6ff0b3e235c518acf5c1fcbbfc95f0ae2
we have this one
> | d03902b8864d7814c938f67befade5a3bba68708 | you should already have those
> | a27993f3d9daca0dffa26577a83822db99c952e2 |
looks like we do, but applied only on arm
> | eb2a021c4710b98081daa797d5a729ac23c240cd
> | 2efa53f373ed811d4860904f5205b8a3b376e253
> | 99f252b097a3bd6280047ba2175b605671da4a23
> | 1371fa6db0bbb8e23f988a641f5ae7361bc629dd
we don't appear to have these
> | It's gross though: there are 99 changes from v2.6.18.8 to current master
> | for the r8169 driver and some registers init changes may have been partially
> | reverted later.
>
> I don't have the time or experience to look into these, so again, I'm
> not quite sure how to proceed. I guess it also depends how close the
> etch 1+1/2 update is.
That's tough to say; I think we should fix this bug in .18
regardless and consider the others on a case-by-case basis.
--
dann frazier
Reply to: