[Gnewsense-dev] Re: Linux 2.6.31.6 for Loongson

[Daniel Clark]

This may be useful in more tuning of the gnewsense for yeeloong kernel
options (looks like rmh (aka nyu) are already on the cc, but thought
others on this list may be interested).

In general if you are interested in yeeloong development you probably
want to be on http://groups.google.com/group/loongson-dev

Subject: Re: [Urgent] About the latest kernel for loongson
From: Wu Zhangjin <address@hidden>
To: loongson-dev <address@hidden>
Cc: Zhang Le <address@hidden>, Liu Shiwei <address@hidden>,
Michael Heide <address@hidden>, Robert Millan
<address@hidden>, Tatu Kilappa <address@hidden>, trilkk
<address@hidden>, Ralf Baechle <address@hidden>
In-Reply-To: <address@hidden>
References: <address@hidden>
         <address@hidden>

Hi, forks

New testing report comes from Tatu Kilappa(thanks goes to him), He has
tried CONFIG_SPARSEMEM instead of the CONFIG_FLATMEM, most of the
problems went away.

And as what I knew, CONFIG_FLATMEM will also bring us with the other
troubles, for example, it will make the SUSPEND(STR) fail. and The
linux-mips maintainer(Ralf) told me it also has some other bad
influences. So, I have enabled CONFIG_SPARSEMEM as the default option in
the defconfig:

in 2.6.30/31

$ ls arch/mips/configs/*loong*
arch/mips/configs/fuloong2e_defconfig
arch/mips/configs/fuloong2f_unas_defconfig
arch/mips/configs/fuloong2f_defconfig
arch/mips/configs/yeeloong2f_defconfig

in 2.6.32/33...

arch/mips/configs/fuloong2e_defconfig
arch/mips/configs/lemote2f_defconfig (for all 2F machines)

And please do remember to use SPARSEMEM instead, thanks!

And some other NOTICES here:

1. please do not try PAGE_SIZE_4KB, it not work on Loongson Family
machines.

2. you are recommended to use CONFIG_64BIT instead of CONFIG_32BIT, it
will not only bring us with some performance, and also, will eliminate
some potential problems.

3. and please do remember to use the binutils patch from Lemote

http://sourceware.org/ml/binutils/2009-11/msg00387.html

Anyway, I will try to send a patch to Ralf to disable the problematic
options for Loongson.

Thanks very much!

Best Regards,
        Wu Zhangjin

On Sun, 2009-11-29 at 22:10 +0800, Wu Zhangjin wrote:
> On Sun, 2009-11-29 at 19:39 +0800, Wu Zhangjin wrote:
> > Hi, All
> >
> > At least three users have reported the problem of NFS & CIFS in 2.6.30
> > and 2.6.31 for loongson: when they copy a large number of files from the
> > other machines to loongson machines(FuLoong Mini PC and U-NAS), the
> > files will be broken.
> >
> > I'm trying to reproduce the problem in my own FuLoong box, but not
> > encounter the problem currently, I can not ensure which part have
> > introduced this bug. Michael have helped to test the old loongson
> > patchset(from 2.6.27) on 2.6.31, the problem is also there, so, I guess,
> > there may be some problem is the mainline kernel or the relative network
> > driver.
> >
> > Before the bug is found and fixed, to avoid breaking your files, please
> > stop using the latest kernel in your FuLoong Mini PC and U-Nas system,
> > but use the old 2.6.27 from http://dev.lemote.com/code/linux_loongson
> > instead. thanks!
> >
>
> If really want to use it, the temp solution is:
>
> $ ethtool -K eth0 rx off
>
> or put the above line into /etc/rc.local.
>
> This solution is from Liu Shiwei, thanks goes to him ;)
>
> > And if you have interest, welcome to test the latest kernel on your
> > machines, and help to reproduce the problem. any feedbacks are welcome!
> > thanks!
> >
> > Best Regards,
> >     Wu Zhangjin
>