problems in loading capture kernel

Wed Feb 3 16:51:33 EST 2010

cat /proc/iomem | grep -i crash
  02000000-05ffffff : Crash kernel

I reserve memory now at 32M.  I can see the entry in /proc/iomem. But
the problem still persists. Perhaps kexec_load() is not loading into the

available reserved memory somehow.

kexec_load failed: Cannot assign requested address
entry       = 0x92550 flags = 1
nr_segments = 4
segment[0].buf   = 0x8eeb40
segment[0].bufsz = 2e4d
segment[0].mem   = 0x8f000
segment[0].memsz = 3000
segment[1].buf   = 0x8e78d0
segment[1].bufsz = 7100
segment[1].mem   = 0x92000
segment[1].memsz = 9000
segment[2].buf   = 0x7f7d32395e10
segment[2].bufsz = 3159f0
segment[2].mem   = 0x100000
segment[2].memsz = 316000
segment[3].buf   = 0x7f7d32a3d010
segment[3].bufsz = 13b165
segment[3].mem   = 0x7fec4000
segment[3].memsz = 13c000

Thanks
Soumendu

-----Original Message-----
From: Vivek Goyal [mailto:vgoyal at redhat.com] 
Sent: Wednesday, February 03, 2010 4:18 PM
To: Soumendu Satapathy
Cc: Kexec Mailing List
Subject: Re: problems in loading capture kernel

On Wed, Feb 03, 2010 at 03:55:40PM -0500, Soumendu Satapathy wrote:
> cat /proc/iomem
> 
> 00000000-0009b3ff : System RAM
> 0009b400-0009ffff : reserved
> 000e4000-000fffff : reserved
> 00100000-cff6ffff : System RAM
>   00200000-006f0557 : Kernel code
>   006f0558-009b9b3f : Kernel data
>   00be2000-01219ea7 : Kernel bss

[..]
> 100000000-22fffffff : System RAM
> 
> I don't see a crashkernel entry in the above. Perhaps that might be
the
> issue.

Yes.

00be2000-01219ea7 : Kernel bss

You are trying to reserve memory starting at 16MB (0x1000000) and that
area has been claimed by kernel bss reigon. Try reserving memory at
some other address, say 32MB (crashkernel=64M at 32M).

Vivek

> 
> 
> Thanks
> Soumendu
> 
> -----Original Message-----
> From: Vivek Goyal [mailto:vgoyal at redhat.com] 
> Sent: Wednesday, February 03, 2010 3:50 PM
> To: Soumendu Satapathy
> Cc: Kexec Mailing List
> Subject: Re: problems in loading capture kernel
> 
> On Wed, Feb 03, 2010 at 03:41:25PM -0500, Soumendu Satapathy wrote:
> 
> > Oh! I am sorry. That's not an issue, I just did some trace to check
> the
> > errno. The errno = 99 which is invalid address.
> > 
> > 
> > - Have you reserved the space for second kernel?
> > 
> > Yes, I have reserved with the boot parameters. crashkernel=64M at 16M
> > 
> > 
> > - Have you compiled kernel as relocatable one or built a second
kernel
> >   and trying to load vmlinux.
> > CONFIG_KEXEC=y
> > CONFIG_CRASH_DUMP=y
> > CONFIG_PHYSICAL_START=0x1000000
> > CONFIG_RELOCATABLE=y
> > CONFIG_PHYSICAL_ALIGN=0x200000
> > 
> 
> Please do not exclude kexec mailing list from CC. Adding it again.
> 
> Can you paste your `cat /proc/iomem` output and make sure space for
> crash
> kernel has been reserved.
> 
> Thanks
> Vivek
> 
> > I have used the above config option for the capture kernel .
Compiled
> > and build it. And was trying to pre-load it using the following
> command.
> > kexec -p  /boot/vmlinuz-2.6.27_capturekernel03feb
> > --initrd=/boot/initrd-2.6.27.img  --append="root=/dev/nfs
> > nfsroot=192.168.101.3:/FILESYSTEMS/nfs-roots/root-55 ip=dhcp
selinux=0
> > 1  irqpoll  maxcpus=1  reset_devices"
> > 
> > 
> > - What was the last working version or setup of kdump in your
> > environment.
> > 
> > Kdump worked well in my environment with a vanilla kernel (2.6.18).
> The
> > one that I am using now is a Fedora core 2.6.27 and it is somehow
> > failing with it.
> > 
> > 
> > Thanks
> > Soumendu
> > 
> > -----Original Message-----
> > From: Vivek Goyal [mailto:vgoyal at redhat.com] 
> > Sent: Wednesday, February 03, 2010 3:25 PM
> > To: Soumendu Satapathy
> > Cc: Kexec Mailing List
> > Subject: Re: problems in loading capture kernel
> > 
> > On Wed, Feb 03, 2010 at 02:46:53PM -0500, Soumendu Satapathy wrote:
> > > Hi Vivek ,
> > > 
> > >  
> > > 
> > > I am having some problems to load the capture kernel.
> > > 
> > >  
> > > 
> > > The following is the error. I am using a x86_64 box. I am using
> 2.6.27
> > > linux kernel for both the first kernel and the capture kernel. The
> > > capture kenel I have configured with CONFIG_SMP=n.
> > > 
> > > maxcpus=1.
> > > 
> > >  
> > > 
> > > kexec_load failed: Cannot assign requested address  99
> > > 
> > 
> > This looks little odd. Where is this address "99" coming from?
> > 
> > Few things.
> > 
> > - Have you reserved the space for second kernel?
> > - Have you compiled kernel as relocatable one or built a second
kernel
> >   and trying to load vmlinux.
> > - What was the last working version or setup of kdump in your
> > environment.
> > 
> > I would begin at running gdb at kexec to figure out where this
> seemingly
> > bad address "99" coming from.
> > 
> > CCing kexec mailing list, in case somebody has seen similar message.
> > 
> > Thanks
> > Vivek
> > 
> > > entry       = 0x92550 flags = 1
> > > 
> > > nr_segments = 4
> > > 
> > > segment[0].buf   = 0xb75b40
> > > 
> > > segment[0].bufsz = 2e77
> > > 
> > > segment[0].mem   = 0x8f000
> > > 
> > > segment[0].memsz = 3000
> > > 
> > > segment[1].buf   = 0xb6e8d0
> > > 
> > > segment[1].bufsz = 7100
> > > 
> > > segment[1].mem   = 0x92000
> > > 
> > > segment[1].memsz = 9000
> > > 
> > > segment[2].buf   = 0x7f5aab6c4e10
> > > 
> > > segment[2].bufsz = 3159d0
> > > 
> > > segment[2].mem   = 0x100000
> > > 
> > > segment[2].memsz = 316000
> > > 
> > > segment[3].buf   = 0x7f5aabd6c010
> > > 
> > > segment[3].bufsz = 13b153
> > > 
> > > segment[3].mem   = 0x7fec4000
> > > 
> > > segment[3].memsz = 13c000
> > > 
> > >  
> > > 
> > >  
> > > 
> > > Can you please provide some inputs ?
> > > 
> > >  
> > > 
> > > thanks
> > > 
> > > Soumendu S Satapathy
> > > 
> > > Senior Software Developer
> > > 
> > > RELDATA Inc.
> > > 
> > > 1719 Route 10, Suite 209
> > > 
> > > Parsippany, NJ 07054
> > > 
> > > (973) 644-2770  ext. 139  office
> > > 
> > > 732-692-7230 mobile
> > > 
> > > (973) 644-3385 fax
> > > 
> > > soumendu.satapathy at reldata.com
> > > 
> > > www.reldata.com <http://www.reldata.com/> 
> > > 
> > >  
> > > 