----- Original Message -----
> > After switching to using
dd->sub_header_kdump->phys_base, do you think we can
> > just leave QEMU_MEM_DUMP_COMPRESSED to be dealt by the DISKDUMP_DUMPFILE
> > section, and add another equivalent one for QEMU_MEM_DUMP_ELF &&
!VMCOREINFO?
>
> I would keep the QEMU_MEM_DUMP_NO_VMCOREINFO() section where it is, because it
> still potentially calls your new kdump_phys_base().
Just to be sure, apart from changing [kdump|diskdump]_phys_base to properly
return whatever value they have in their respective phys_base fields, is there
any other change necessary to this section, or is the patch good as it is?
Thanks,
Sergio.
In this odd-ball case, I would prefer that x86_64_virt_phys_base() always
gets called, so how about this:
if (QEMU_MEM_DUMP_NO_VMCOREINFO()) {
if ((KDUMP_DUMPFILE() && kdump_phys_base(&phys_base)) ||
(DISKDUMP_DUMPFILE() && diskdump_phys_base(&phys_base)))
machdep->machspec->phys_base = phys_base;
if (!x86_64_virt_phys_base())
error(WARNING,
"cannot determine physical base address:"
" defaulting to %lx\n\n",
machdep->machspec->phys_base);
return;
}
Dave