----- "Feng LI" <funglee(a)gmail.com> wrote:
Dave,
I regenerate the vmcore dump file to make sure I used the correct
version of dom0 which we have unstripped kernel from critix.
The p2m_mfn pointer is still empty... :( I attached the output we have
.. :(
Too bad... It seems that I am going no where.
&
Thanks for your help.
Kevin,
The p2m_mfn value still needs to be overridden on the command line, so your
latest results are not surprising. But there's one more thing you might
try. Looking at your latest debug output:
xen_kdump_data:
flags: 4 (KDUMP_MFN_LIST)
p2m_mfn: 0 <== incorrect
cr3: 0
last_mfn_read: ffffffff
last_pmd_read: ffffffff
page: 0
accesses: 0
cache_hits: 0
p2m_frames: 0
xen_phys_start: 6f969 <== incorrect
Because of the lack of support of 32-bit dom0's on 64-bit hypervisors,
the p2m_mfn and xen_phys_start values are being loaded as if the
note contained 32-bit values:
Elf64_Nhdr:
n_namesz: 4 ("Xen")
n_descsz: 80
n_type: 1000001 (XEN_ELFNOTE_CRASH_INFO)
00000003 00000000 00000004 00000000
9f7eb409 00000000 9f7ed168 00000000
9f7ed138 00000000 9f7eb3c5 00000000
9f7eb3e2 00000000 00000000 00000000
9f600000 00000000 0006f969 00000000
So in your latest dumpfile, they should be set to:
p2m_mfn: 6f969
xen_phys_start: 9f600000
And because of the bogus p2m_mfn address of 0, you get bogus mfn
values in the dump of mfns shown later on here:
x86_xen_kdump_p2m_create: p2m_mfn: 0
00000000: 7fed9050 7fed904c 7fed9058 7fed9054
00000010: 7fed9060 7fed905c f00016b3 f00016b4
00000020: f000fea5 f000e987 f00016b4 f00016b4
00000030: 98400b7c f00016a7 f000ef57 f00016b4
Those 16 values above should be mfn values.
You can override both the bogus p2m_mfn and xen_phys_start values on
the crash command line like this:
# crash --p2m_mfn 6f969 --xen_phys_start 9f60000 vmlinux vmcore
I'm guessing that you will probably still see bogus data in the list
of mfn values from x86_xen_kdump_p2m_create debug dump above, but it's
worth a shot. But if you do see the correct mfn values there, then
the correct xen_phys_start value will be required after that.
P.S.
I did cat /proc/iomem before I generate the vmcore dump.
it shows
9ead4000-9fb2cfff : System RAM
9f700000-9f8dca7f : Hypervisor code and data
9f899ca0-9f899e93 : Crash note
9f899e94-9f89a027 : Crash note
9f89a088-9f89a21b : Crash note
9f89a27c-9f89a40f : Crash note
is this information helpful to you ?
Not so much...
Dave