----- "Feng LI" <funglee(a)gmail.com> wrote:
Sorry Dave,
My mistake. Actually, we have to use 64 bit hypervisor... The file I
show to you was created with xen-64 bit and dom0 32 bit. :(
Sorry about my mistakes.
When I tried the crash with --p2m_mfn 000bf969 option.
x86_xen_kdump_p2m_create: p2m_mfn: bf969
00000000: cccccccc cccccccc cccccccc cccccccc
00000010: cccccccc cccccccc cccccccc cccccccc
00000020: cccccccc cccccccc cccccccc cccccccc
00000030: cccccccc cccccccc cccccccc cccccccc
crash32: read error: physical address: cccccccc000 type: "xen kdump p2m mfn list
page"
OK, so we're back to the same starting point, that being that
it's evident that crash does not support 32-bit dom0 vmcores
generated by a 64-bit hypervisor.
The page referenced by the p2m_mfn pointer contains a set of mfn
values that is needed to recreate the kernel's phys_to_machine_mapping[]
array, which is needed to translate the kernel's pseudo-physical pfns to
real machine physical addresses (mfns). Clearly something is out-of-sync
given the contents of the page at bf969000, but I have no idea what that
could be.
In any case, without that essential key, I've pretty much run out of
things to suggest.
Again, I invite the other Xen users/developers on this list to make
any suggestions, or develop support for this configuration. I actually
don't think it would require a major update, but it's not clear to me
what the problem is.
Dave