[Crash-utility] Xen Dumping - Expectations

[Dave Anderson]

Joshua Giles wrote:

> Not quite sure if this or the kdump list is appropriate for this
> discussion, but I will start here...
>

Why kdump?

>
> I've been playing with xen dumping on x86_64 and x86 (RHEL5 20061006.2);
> The following is a simple crash session on x86_64 (using "xm dump-core
> -L"):

Interesting.  It kind of looks like there's something different about the
corefile contents when using "xm dump-core" as opposed to forcing
a "real" crash, i.e., such as when using sysrq-c?

>
>
> `crash /usr/lib/debug/lib/modules/2.6.18-1.2714.el5xen/vmlinux /xen/dump/Live_Dump`
>
> "WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
> WARNING: possibly bogus exception frame
>       KERNEL: /usr/lib/debug/lib/modules/2.6.18-1.2714.el5xen/vmlinux
>     DUMPFILE: /xen/dump/Live_Dump
>         CPUS: 1
>         DATE: Mon Oct 16 11:46:02 2006
>       UPTIME: 00:07:12
> LOAD AVERAGE: 0.00, 0.02, 0.00
>        TASKS: 67
>     NODENAME: dhcp78-103.lab.boston.redhat.com
>      RELEASE: 2.6.18-1.2714.el5xen
>      VERSION: #1 SMP Mon Oct 2 17:26:27 EDT 2006
>      MACHINE: x86_64  (3200 Mhz)
>       MEMORY: 264 MB
>        PANIC: ""
>          PID: 0
>      COMMAND: "swapper"
>         TASK: ffffffff804c5ae0  [THREAD_INFO: ffffffff805de000]
>          CPU: 0
>        STATE: TASK_RUNNING (ACTIVE)
>      WARNING: panic task not found
>
> crash>crash> bt
> crash>
> crash> ffffff805dff68] raw_safe_halt at ffffffff8026c989
> crash> ffffff805dff88] xen_idle at ffffffff8026a04d
> crash> ffffff805dff98] cpu_idle at ffffffff80248e4d"
>
> *WARNINGS, WARNINGS, WARNINGS (no such problem like this for i386.)
> *The output of bt doesn't actually work as expected.  You have to hit
> carriage return a couple times before you see anything.
>
> #####i386#############################################################
> crash /usr/lib/debug/lib/modules/2.6.18-1.2714.el5xen/vmlinux /xen/dump/live_dump
> This GDB was configured as "i686-pc-linux-gnu"...
>
>       KERNEL: /usr/lib/debug/lib/modules/2.6.18-1.2714.el5xen/vmlinux
>     DUMPFILE: /xen/dump/live_dump
>         CPUS: 1
>         DATE: Mon Oct 16 11:47:28 2006
>       UPTIME: 00:08:30
> LOAD AVERAGE: 0.00, 0.04, 0.03
>        TASKS: 68
>     NODENAME: dhcp78-115.lab.boston.redhat.com
>      RELEASE: 2.6.18-1.2714.el5xen
>      VERSION: #1 SMP Mon Oct 2 17:40:51 EDT 2006
>      MACHINE: i686  (3200 Mhz)
>       MEMORY: 264 MB
>        PANIC: ""
>          PID: 0
>      COMMAND: "swapper"
>         TASK: c0674a60  [THREAD_INFO: c07a3000]
>          CPU: 0
>        STATE: TASK_RUNNING (ACTIVE)
>      WARNING: panic task not found
>
> crash> bt
>
> crash>
> crash> 7a3fa8] hypercall_page at c04013a7
> crash> 7a3fb8] xen_idle at c0402bdb
> crash> 7a3fc0] cpu_idle at c0402d02
> crash>
>
> *The output of bt doesn't actually work as expected.  You have to hit
> carriage return a couple times before you see anything and the addresses
> are partially visible...
> *panic task not found?
>
> Are these known problems with crash and/or xen?

No -- but it's never been tested with this new "xm dump-core" facility.

> Is xen dumping
> supported on x86, x86_64, ppc, ia64?

x86 and x86_64 only -- ia64 is still TBD.

> Can anyone point me to docs that
> talk about xen dumping (e.g. internal/external wiki?)

None that I'm aware of...

The best thing that you can do is come over to my office,
and we'll get to the bottom of this.  In the meantime, it would
be interesting to know whether the behavior above is the
same when you:

1. log into the domU
2. echo c > /proc/sysrq-trigger

Anyway, when you get the time, please stop by.

Thanks,
  Dave