Re: ARM kernel snapshots - 2.6.29-rc7
On Mon, Mar 09, 2009 at 09:23:00PM +0100, Martin Michlmayr wrote:
> > > This didn't happen when you were running 2.6.26 on the QNAP, right?
> > yes with the original kernel I had no problems, but the performance was
> > very bad (~1.5MB/s writing).
> Well, if the network connection didn't hang with 2.6.26 but it hangs
> with 2.6.29-rc7 then this sounds like a regression, which is quite
> likely given that the network driver in case saw major changes in
> 2.6.27, 2.6.28 and 2.6.29.
ACK. There have been 100 commits to mv643xx_eth between 2.6.26 and
2.6.29-rc7, which constitute a pretty much complete rewrite of the
driver. It _should_ be possible to bisect this problem (in the git
bisect sense), but it's hard to say what could be the issue without
some more data.
> Lennert Buytenhek, who maintains the mv643xx_eth network driver,
> thinks it would be easier to debug the problem if you had a serial
> console. Do you think you can could make one?
- Are you able to test 2.6.27 or 2.6.28?
- Does your machine panic, or do userland tasks keep running (albeit
without network access)? E.g. if you install a cron job that echos
the current date to a file every minute (and calls sync), does that
file keep being updated after the network becomes inaccessible?
- Are there any oopses in your system log after a reboot?