[Crash-utility] dom0 analysis for IA64

Dave Anderson anderson at redhat.com
Fri May 11 15:01:17 UTC 2007


Akio Takebe wrote:

> Hi, Dave and Oda
>
> Thank you for your work.
>
> >The sample vmcore was taken by fujitsu's "sadump" tool which
> >is a hardware assist tool.
> >Putting the dump button casuse INIT inerrupt and then
> >registers are saved and CPUs stop. The hardware dumps all
> >memory after that. The memory image is converted to ELF
> >vmcore (netdump style) finally.
> >
> >It is hard to add XEN_ELFNOTE_CRASH_INFO notes in this
> >procedure (I think).
> >(I think it is easy to change to converting kdump style.)
> >
> >I will ask to fujitsu's people whether it is available.
> >
> Yes, it is hard. So I cannot change the dump style soon.
> Could you please leave check_netdump_xen()?
> Or should we add a option?
>
> Best Regards,
>
> Akio Takebe

OK, I understand.  This is the first time I've ever heard of
the Fujitsu "sadump" facility.

The function name "check_netdump_xen()" is confusing because
the "real" netdump facility and xen kernels don't have any
relationship.

So I'd prefer to make it easier to understand what's going
on by doing something like this in main():

        } else if (is_netdump(argv[optind], NETDUMP_LOCAL)) {
                if (pc->flags & MEMORY_SOURCES) {
                        error(INFO,
                            "too many dumpfile arguments\n");
                        program_usage(SHORT_FORM);
                }
                pc->flags |= NETDUMP;
                pc->dumpfile = argv[optind];

                if (is_sadump_xen()) {
                        pc->readmem = read_kdump;
                        pc->writemem = write_kdump;
                } else {
                        pc->readmem = read_netdump;
                        pc->writemem = write_netdump;
                }
        }

and then make is_sadump_xen() do this:

/*
 *  Fujitsu dom0/HV sadump-generated dumpfile, which requires
 *  the --p2m_mfn command line argument.
 */
int
is_sadump_xen()
{
        if (xen_kdump_data.p2m_mfn) {
                if (!XEN_CORE_DUMPFILE()) {
                        pc->flags |= XEN_CORE;
                        nd->xen_kdump_data = &xen_kdump_data;
                        nd->xen_kdump_data->last_mfn_read = BADVAL;
                        nd->xen_kdump_data->flags |= KDUMP_MFN_LIST;
                }
                return TRUE;
        }

        return FALSE;
}

Does that seem reasonable?

Dave

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/crash-utility/attachments/20070511/7a867076/attachment.htm>


More information about the Crash-utility mailing list