Dave,

Thanks for your response.  Just in case anyone else hits the same problem.  The more recent kernel built was built with CONFIG_KALLSYMS=y but  CONFIG_KALLSYMS_ALL was not set.  Setting CONFIG_KALLSYMS_ALL=y solved the problem.

-Thomas


> Date: Mon, 1 Nov 2010 15:15:38 -0400
> From: anderson@redhat.com
> To: crash-utility@redhat.com
> Subject: Re: [Crash-utility] core file contents
>
>
> ----- "tom anderson" <xentoma@hotmail.com> wrote:
>
> > I have two different core files generated from two different kernels.
> > I noticed one contains all segments
> >
> > crash> sym -m dpm_kstat
> > ffffffffa00c4000 MODULE START: my_kstat
> > ...
> > ffffffffa00c5290 (t) kstat_numrange_create
> > ffffffffa00c55a0 (r) __func__.41073
> > ...
> > ffffffffa00c5ddb (r) __kstrtab_kstat_numrange_iter_init
> > ffffffffa00c5e60 (d) dprintflevel_kstat
> > ...
> > ffffffffa00c5f40 (d) __this_module
> > ffffffffa00c61a0 (b) kstat_spinlock
> > ...
> > ffffffffa00c6e5e MODULE END: my_kstat
> >
> >
> >
> > while the other contains only text segments. Is there some define in
> > the kernel config or in core file generation which causes this?
> >
> >
> > crash> sym -m dpm_kstat
> > ffffffffa00b9000 MODULE START: dpm_kstat
> > ffffffffa00b9000 (t) kstat_cmd_gettbl
> > ffffffffa00b9930 (t) kstat_exit
> > ffffffffa00b9930 (t) cleanup_module
> > ffffffffa00b9a20 (t) kstat_init
> > ffffffffa00b9a20 (t) init_module
> > ffffffffa00b9bb0 (t) kstat_unregister_provider
> > ffffffffa00b9c90 (t) kstat_register_provider
> > ffffffffa00b9dc0 (t) kstat_numrange_dprintf
> > ffffffffa00b9fa0 (t) kstat_numrange_match
> > ffffffffa00b9fe0 (t) kstat_numrange_iter_init
> > ffffffffa00ba060 (t) kstat_numrange_destroy
> > ffffffffa00ba0a0 (t) kstat_numrange_addentry
> > ffffffffa00ba1e0 (t) kstat_numrange_create
> > ffffffffa00bb5af MODULE END: dpm_kstat
>
> I'm not exactly sure.
>
> Typically if you display the module contents by getting the module
> address from the "mod" command output, and dump it like so:
>
> crash> module <address>
>
> the "num_symtab" field will be roughly equal to the symbol count:
>
> crash> mod | grep ext4
> ffffffffa020e260 ext4 353979 (not loaded) [CONFIG_KALLSYMS]
> crash> module.num_symtab ffffffffa020e260
> num_symtab = 1080,
> crash> sym -m ext4 | wc -l
> 1081
> crash>
>
> When you do the "mod" command on the two different kernels,
> does it show "[CONFIG_KALLSYMS" on the end of each module's
> line? If not, that may have something to do with it.
>
> And if you run the "mod" command to load the debuginfo
> data of that module, it's possible that a subsequent
> "sym -m <module>" command may show additional symbols.
>
> Dave
>
>
>
> --
> Crash-utility mailing list
> Crash-utility@redhat.com
> https://www.redhat.com/mailman/listinfo/crash-utility