[Crash-utility] Re: crash seek error

Dave Anderson anderson at redhat.com
Wed Mar 11 17:54:35 UTC 2009


----- "Dave Anderson" <anderson at redhat.com> wrote:

> ----- "Dharmosoth Seetharam" <seetharam_21 at yahoo.com> wrote:

> > dump_header:
> > dh_magic_number: 618f23ed (DUMP_MAGIC_NUMBER)
> > dh_version: 8 (LKCD_DUMP_V8)
> > dh_header_size: 734
> > dh_dump_level: f
> > (DUMP_LEVEL_HEADER|DUMP_LEVEL_KERN|DUMP_LEVEL_USED|DUMP_LEVEL_ALL)
> > dh_page_size: 4096
> > dh_memory_size: 524153
> > dh_memory_start: c0000000
> > dh_memory_end: 618f23ed
> > dh_num_pages: 524153
> > dh_panic_string: Compulsory dump(stat of mkexec was set as 2).
> > dh_time: Tue Mar 10 17:48:00 2009
> > dh_utsname_sysname: Linux
> > dh_utsname_nodename: Assam
> > dh_utsname_release: 2.6.12-5MKEXEC
> > dh_utsname_version: #7 SMP Thu Mar 5 15:25:22 IST 2009
> > dh_utsname_machine: i686
> > dh_utsname_domainname: (none)
> > dh_current_task: efc4a020
> > dh_dump_compress: 0 (DUMP_COMPRESS_NONE)
> > dh_dump_flags: 80000000 ()
> > dh_dump_device: 0
> > unknown page flag in dump: 2de
> > found DUMP_DH_END
> > <readmem: 8015564b, KVADDR, "x86_omit_frame_pointer", 8, (ROE), 7fbbe228>
> > crash: seek error: kernel virtual address: 8015564b type: "x86_omit_frame_point er"
> > <readmem: 804b1210, KVADDR, "xtime", 8, (FOE), 834d234> 
> > crash: seek error: kernel virtual address: 804b1210 type: "xtime"
> > [root at Assam ~]#
> >
> > can you please help me in this.

One other thing to look at...

> > dh_memory_start: c0000000

The failing kernel virtual addresses are 8015564b and 804b1210, so apparently
you're running a kernel configured with a 2G/2G split?  I'm not sure
whether the crash utility even works with that configuration?  Crash does
support the old RHEL4 "hugemem" 4G/4G kernels, but I've never worked with
a 2G/2G kernel.  In any case, it may work by dumb luck -- to be sure, first
try to run crash on the live system.

Anyway, even though the dump header advertises a kernel configured with 
the traditional 3G/1G split (with kernel memory starting at c000000), 
that "dh_memory_start" field is not used by the crash utility.

Dave







More information about the Crash-utility mailing list