Re: [Xen-devel] xen panics when setting int3 traps

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 11/12/15 01:42, quizy_jones@xxxxxxxxxxx wrote:

I'd like to inject int3 traps to hypercall handlers to be able to capture hypercalls. However,  the dom0/xen would reboot whenever I create a new VM. Followup is the console output when this happens.

<snip>

(XEN) Xen call trace:
(XEN)    [<ffff82d080128132>] _write_unlock+0x22/0x40
(XEN)    [<ffff82d080115b98>] guest_remove_page+0xf8/0x2e0
(XEN)    [<ffff82d080116138>] do_memory_op+0x3b8/0x2750
(XEN)    [<ffff82d08017c29c>] __do_update_va_mapping+0x1bc/0x6c0
(XEN)    [<ffff82d08021aef9>] syscall_enter+0xa9/0xae
(XEN)
(XEN)
(XEN) ****************************************
(XEN) Panic on CPU 0:
(XEN) FATAL TRAP: vector = 6 (invalid opcode)
(XEN) ****************************************
(XEN)
(XEN) Reboot in five seconds...

BTW, I'm using a nested xen 4.4.1 on xen 4.6.0.

How are you injecting the traps?  Unless you explicitly register a debugger with Xen, it will consider any debugging of itself fatal, as you can see.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel