Re: kernel 2.6.23 CFS problem?

2007-10-26 Thread Wang, Baojun 
On Friday 26 October 2007 17:05:42, you wrote：
> On Fri, 2007-10-26 at 15:32 +0800, Wang, Baojun wrote:
> > > Perhaps he has any tasks stuck in 'D' state?  If so, SysRq-t output for
> > > these would be a good thing to collect.  What leads him to suspect the
> > > process scheduler, etc.  More details are needed.
> >
> > The task is S+ state, please see in the attachment tarball, SysRq-t
> > output has also been attached, I wish it would help. I doubt it's the CFS
> > problem because it's first introduced into the 2.6.23 kernel.
>
> (yeah, but there were a truckload of other changes as well)
>
> Looking at your data, _my_ first suspect would be the 2.6.22->2.6.23
> futex changes.  I'd revert all of the changes to kernel/futex.c and
> kernel/futex_compat.c, and see if the problem went away.  If it didn't,
> I'd then do a full git bisect to nail it down.
>
>   -Mike

Thanks very much, also my glibc (2.6.1) is built under 2.6.23 (but not 
linux-headers, which is 2.6.22). but I've tried right now (install ELDK4.1) 
on another x86_64 machine, without any problem (x86_64, Core2Duo E6600, 4G 
RAM)

Wang
-- 
Wang, Baojun                                        Lanzhou University
Distributed & Embedded System Lab              http://dslab.lzu.edu.cn
School of Information Science and Engeneering       [EMAIL PROTECTED]
Tianshui South Road 222. Lanzhou 73                     .P.R.China
Tel:+86-931-8912025                                Fax:+86-931-8912022


signature.asc
Description: This is a digitally signed message part.

Re: kernel 2.6.23 CFS problem?

2007-10-26 Thread Wang, Baojun 
On Friday 26 October 2007 14:47:02, you wrote：
> On Fri, 2007-10-26 at 12:45 +0800, WANG Cong wrote:
> > On Fri, Oct 26, 2007 at 10:57:45AM +0800, Wang, Baojun wrote:
> > >hi, list
> > >
> > >  I've upgraded my kernel from 2.6.22.9 to 2.6.23 when it was out, After
> > > that I can't install ELDK 4.1 anymore (The one I installed was
> > > crashed), it always stopped at preparing install package XXX (or YYY
> > > sometimes), I've waited for a very long time(more than 1 hour), but it
> > > still the same. After I switch back to kernel 2.6.22.9, everything is
> > > fine. Is there somebody have the some problem like me?
> >
> > What's your XXX or YYY?
In the attachment
> > What error messages did you get?
No error message, but the installation just "freeze", I've waited for more 
than 1 hour, but still no progress, with 2.6.22.9 it was fine.
> > Why don't you try to update your glibc or something like that?
glibc is 2.6.1
> > I am afraid it's off-topic here.
>
> Not necessarily so if kernel version indeed makes the difference between
> hung install and successful completion.
>
> Perhaps he has any tasks stuck in 'D' state?  If so, SysRq-t output for
> these would be a good thing to collect.  What leads him to suspect the
> process scheduler, etc.  More details are needed.
The task is S+ state, please see in the attachment tarball, SysRq-t output has 
also been attached, I wish it would help. I doubt it's the CFS problem 
because it's first introduced into the 2.6.23 kernel.
> (I went looking for this package, but there are several, and they're
> huge)
>
>   -Mike

Wang

-- 
Wang, Baojun                                        Lanzhou University
Distributed & Embedded System Lab              http://dslab.lzu.edu.cn
School of Information Science and Engeneering       [EMAIL PROTECTED]
Tianshui South Road 222. Lanzhou 73                     .P.R.China
Tel:+86-931-8912025                                Fax:+86-931-8912022


report.tar.bz2
Description: application/tbz


signature.asc
Description: This is a digitally signed message part.

kernel 2.6.23 CFS problem?

2007-10-25 Thread Wang, Baojun 
hi, list

  I've upgraded my kernel from 2.6.22.9 to 2.6.23 when it was out, After that 
I can't install ELDK 4.1 anymore (The one I installed was crashed), it always 
stopped at preparing install package XXX (or YYY sometimes), I've waited for 
a very long time(more than 1 hour), but it still the same. After I switch 
back to kernel 2.6.22.9, everything is fine. Is there somebody have the some 
problem like me?

  I've tried on two machines:

1. AMD DURON 1.6G, 512M RAM, gentoo 2007.0
2. INTEL Pentium 3 667M, 192M RAM, gentoo 2007.0

  Regards,
Wang
-- 
Wang, Baojun                                        Lanzhou University
Distributed & Embedded System Lab              http://dslab.lzu.edu.cn
School of Information Science and Engeneering       [EMAIL PROTECTED]
Tianshui South Road 222. Lanzhou 73                     .P.R.China
Tel:+86-931-8912025                                Fax:+86-931-8912022


config-2.6.23-k7.gz
Description: GNU Zip compressed data


signature.asc
Description: This is a digitally signed message part.

PROBLEM: kernel 2.6.20.6 build failed for ppc board chestnut(ibm ppc 750GX/FX)

2007-04-15 Thread Wang, Baojun 
PROBLEM: linux kernel 2.6.20.6 build failed for ppc board chestnut(ibm ppc 
750GX/FX)

STEPS I followed:
1) make ARCH=ppc CROSS_COMPILE=ppc_4xx- chestnut_defconfig # toolchain from 
eldk 4.1

2) make ARCH=ppc CROSS_COMPILE=ppc_4xx- V=1

...

make -f scripts/Makefile.build obj=arch/ppc/platforms
  
ppc_4xx-gcc -m32 -Wp,-MD,arch/ppc/platforms/.chestnut.o.d  -nostdinc -isystem 
/opt/eldk/usr/bin/../lib/gcc/powerpc-linux/4.0.0/include -D__KERNEL__ -Iinclude 
 -include 
include/linux/autoconf.h -Iarch/ppc -Iarch/ppc/include -Wall -Wundef 
-Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -O2 
-Iarch/ppc -msoft-float -pipe -ffixed-r2 -mmultiple  -mno-altivec -mstring 
-Wa,-maltivec -fomit-frame-pointer  -Wdeclaration-after-statement 
-Wno-pointer-sign-D"KBUILD_STR(s)=#s" 
-D"KBUILD_BASENAME=KBUILD_STR(chestnut)"  
-D"KBUILD_MODNAME=KBUILD_STR(chestnut)" -c -o 
arch/ppc/platforms/chestnut.o arch/ppc/platforms/chestnut.c
arch/ppc/platforms/chestnut.c: In function 'chestnut_setup_mtd':
arch/ppc/platforms/chestnut.c:435: error: 'physmap_map' undeclared (first use 
in this function)
arch/ppc/platforms/chestnut.c:435: error: (Each undeclared identifier is 
reported only once
arch/ppc/platforms/chestnut.c:435: error: for each function it appears in.)
make[1]: *** [arch/ppc/platforms/chestnut.o] Error 1
make: *** [arch/ppc/platforms] Error 2


This PROBLEM should also affect these boards:
find -name '*.[chS]' -exec grep physmap_map {} \; -print
physmap_map.size = CHESTNUT_32BIT_SIZE;
./arch/ppc/platforms/chestnut.c
physmap_map.size = size;
./arch/ppc/platforms/ev64360.c
physmap_map.size = size;
./arch/ppc/platforms/katana.c

I've veryfied 2.6.15, 2.6.19.2 2.6.20.1 2.6.20.6

ONLY 2.6.15 can build successfully. sounds like the problem is caused by the 
interface change of mtd.


this brute force patch sould solve the problem:

diff -Nru /tmp/linux-2.6.20.6/arch/ppc/platforms/chestnut.c \  
linux-2.6.20.6/arch/ppc/platforms/chestnut.c

--- /tmp/linux-2.6.20.6/arch/ppc/platforms/chestnut.c   2007-04-07 
04:02:48.0 +0800
+++ linux-2.6.20.6/arch/ppc/platforms/chestnut.c2007-04-13 
17:09:03.0 +0800
@@ -432,7 +432,9 @@
ptbl.name = "User FS";
ptbl.size = CHESTNUT_32BIT_SIZE;

-   physmap_map.size = CHESTNUT_32BIT_SIZE;
+   // physmap_map.size = CHESTNUT_32BIT_SIZE;
+   physmap_configure(CHESTNUT_32BIT_BASE, CHESTNUT_32BIT_SIZE, 
CONFIG_MTD_PHYSMAP_BANKWIDTH, NULL);
+
    physmap_set_partitions(&ptbl, 1);
return 0;
 }

-- 
Wang, BaojunLanzhou University
Distributed & Embedded System Lab  http://dslab.lzu.edu.cn
School of Information Science and Engeneering[EMAIL PROTECTED]
Tianshui South Road 222. Lanzhou 73 .P.R.China
Tel:+86-931-8912025Fax:+86-931-8912022


pgp8XeK8j0iAy.pgp
Description: PGP signature