Re: [gentoo-mips] shmem

[Kumba <kumba@gentoo.org>]

spbecker wrote:
> 
> Fix the n32 multithreaded kernel bug and we'll talk (see 
> http://beerandrocks.net:8080/~spbecker/oops/).  Until then, you are just 
> complaining about problems we can't help you with.  Furthermore, you started 
> this thread with complaining about your shmem problem, and only additional 
> questioning from Redhatter revealed that it was a n32 specific problem (as so 
> many are when you are playing with unsupported stuff like this).  From now on, 
> you need to specify what sort of userland, as well as what kernel version(s) 
> you reproduced it with before we can even begin to help you.

David: If you want to get a head start tackling this, try this idea (it's what 
I'm going to do when I get my lazy self back into gear).  Start with a 2.6.10 
mips-kernel to use n32.  This will give you a somewhat usable kernel with a 
somewhat usable n32 userland.  The annoying pthreads bug doesn't appear in this 
kernel.

Now, starting from 2.6.12-rc1 (the next kernel rev checked into lmo cvs, they 
skipped 2.6.11), do a checkout of _each day_ of linux-mips CVS, build a kernel, 
and try it.  Keep repeating until you find the first day the pthreads bug geoman 
mentions first appears.  Once we know this day, we can take a look at all the 
CVS commits made that day and likely find the culprit patch that h0rked us. 
Once found, maybe we can get some attention drawn to it on the linux-mips list 
and thus get it fixed.

As for the shmem thing, good luck, heh.  We've been trying to figure that one 
out for months.  hdparm is a program that will generate it every time.  strace 
too.  Maybe one of those programs can be used as a test case to narrow down and 
squish the shmem bug as well.



--Kumba

-- 
Gentoo/MIPS Team Lead
Gentoo Foundation Board of Trustees

"Such is oft the course of deeds that move the wheels of the world: small hands 
do them because they must, while the eyes of the great are elsewhere."  --Elrond
-- 
gentoo-mips@gentoo.org mailing list