Re: [Xen-users] Problem with dump-core after restoring machine

[Dawid <murkys@xxxxxxxxx>]

Problem was caused by shared info page in
xc_domain_dumpcore_via_callback() in libxenctrl.so.4.0.0. I'm not sure
whether HVM guest should have shared info page but in this case after
restoring domain, shared_info_page was mapped to nonexistent memory and
this was failing while trying to do dump core.

tools/libxc/xc_core.c:498 (Xen 4.1.2)

    if ( xc_domain_getinfo(xch, domid, 1, &info) != 1 ) {...}
    /* Map the shared info frame */
    live_shinfo = xc_map_foreign_range(xch, domid, PAGE_SIZE,
                                       PROT_READ, info.shared_info_frame);

After creating hvm domain from a scratch (`xm create winxp`)
shared_info_frame was set to something looking like a proper frame
number. But after restoring domain (`xm restore ./winxp.save`)
shared_info_frame was set to 0xFFFFFFFFFFFFFFFF.

And then it was failing at tools/libxc/xc_core.c:785
sts = dump_rtn(xch, args, (char*)live_shinfo, PAGE_SIZE);

Because I'm not quite acquainted with Xen source code and don't have
much time right now, I haven't searched deeper (just commented out this
code and recompiled libxenctrl.so).

Regards,
Dawid

On 04.11.2011 01:02, Dawid wrote:
> Hello, I have problem with dumping core of hvm domain. Problem occurs
> only after previously restoring machine from a save file. There is no
> problem with dumping core after creating domain from a scratch and then
> trying to dump. Also there is no such problem with paravirtualized domains.
>
> Error message I got (tried using both xm and xl):
>
> xc: error: Failed to write buffer (14 = Bad address): Internal error
> libxl: error: libxl.c:479:libxl_domain_core_dump core dumping domain 14
> to ./dump.2: Bad address
> core dump failed (rc=-3)

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users