Xen devel - Mar 2008 - [Xen-ia64-devel] [Patch][0/2]implement dump_execution_state() on ia64

Hi,

This patch can show Calltraces at BUG_ON() on ia64.
And this supports xm debug-key d on ia64.
This patch needs to modify some common codes for supporting them.

[1/2] x86  side: implement_dump_execution_state.x86.patch
[2/2] ia64 side: implement_dump_execution_state.ia64.patch

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

Best Regards,

Akio Takebe


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@lists.xensource.com
http://lists.xensource.com/xen-ia64-devel

Hi,

This patch is x86  side.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

Best Regards,

Akio Takebe

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@lists.xensource.com
http://lists.xensource.com/xen-ia64-devel

Hi,

This patch is ia64 side.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

Best Regards,

Akio Takebe

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@lists.xensource.com
http://lists.xensource.com/xen-ia64-devel

On 27/3/08 13:05, "Akio Takebe" <takebe_akio@jp.fujitsu.com>
wrote:
> This patch can show Calltraces at BUG_ON() on ia64.
> And this supports xm debug-key d on ia64.
> This patch needs to modify some common codes for supporting them.
Can you carefully explain why the interface changes are required?

 -- Keir



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

Hi, Keir
>On 27/3/08 13:05, "Akio Takebe" <takebe_akio@jp.fujitsu.com>
wrote:
>
>> This patch can show Calltraces at BUG_ON() on ia64.
>> And this supports xm debug-key d on ia64.
>> This patch needs to modify some common codes for supporting them.
>
>Can you carefully explain why the interface changes are required?
>
IA64 needs unwind infomation to show the Calltrace.
To make the unwind infomation, we needs switch stack and pt_regs.
So if possible, I want to use pt_regs at geting a interrupt.

I make many interface change because debug-key use IPI.
But I just want to use pt_regs.

Best Regards,

Akio Takebe


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@lists.xensource.com
http://lists.xensource.com/xen-ia64-devel

On 27/3/08 16:56, "Akio Takebe" <takebe_akio@jp.fujitsu.com>
wrote:
>> Can you carefully explain why the interface changes are required?
>> 
> IA64 needs unwind infomation to show the Calltrace.
> To make the unwind infomation, we needs switch stack and pt_regs.
> So if possible, I want to use pt_regs at geting a interrupt.
> 
> I make many interface change because debug-key use IPI.
> But I just want to use pt_regs.
There is already a function which does what you want (i.e., takes a
cpu_user_regs pointer) -- it''s called show_execution_state().

On x86 we implement dump_execution_state() by forcing an exception (by
executing an invalid opcode) and dumping the calltrace from within the
exception handler. This works because the exception handler has a
cpu_user_regs pointer and hence can call show_execution_state(). We use the
same trick for handling BUG() and ASSERT(). The code concerned is in
arch/x86/traps.c:do_invalid_op().

Can a similar thing be done on ia64?

 -- Keir



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

Hi,
>On 27/3/08 16:56, "Akio Takebe" <takebe_akio@jp.fujitsu.com>
wrote:
>
>>> Can you carefully explain why the interface changes are required?
>>> 
>> IA64 needs unwind infomation to show the Calltrace.
>> To make the unwind infomation, we needs switch stack and pt_regs.
>> So if possible, I want to use pt_regs at geting a interrupt.
>> 
>> I make many interface change because debug-key use IPI.
>> But I just want to use pt_regs.
>
>There is already a function which does what you want (i.e., takes a
>cpu_user_regs pointer) -- it''s called show_execution_state().
>
>On x86 we implement dump_execution_state() by forcing an exception (by
>executing an invalid opcode) and dumping the calltrace from within the
>exception handler. This works because the exception handler has a
>cpu_user_regs pointer and hence can call show_execution_state(). We use the
>same trick for handling BUG() and ASSERT(). The code concerned is in
>arch/x86/traps.c:do_invalid_op().
On x86, does "xm debug-key d" make system hang?
Or it just uses the exception handler, not hangup?

Best Regards,

Akio Takebe


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@lists.xensource.com
http://lists.xensource.com/xen-ia64-devel

On 27/3/08 17:22, "Akio Takebe" <takebe_akio@jp.fujitsu.com>
wrote:
>> There is already a function which does what you want (i.e., takes a
>> cpu_user_regs pointer) -- it''s called show_execution_state().
>> 
>> On x86 we implement dump_execution_state() by forcing an exception (by
>> executing an invalid opcode) and dumping the calltrace from within the
>> exception handler. This works because the exception handler has a
>> cpu_user_regs pointer and hence can call show_execution_state(). We use
the
>> same trick for handling BUG() and ASSERT(). The code concerned is in
>> arch/x86/traps.c:do_invalid_op().
> On x86, does "xm debug-key d" make system hang?
> Or it just uses the exception handler, not hangup?
It works fine. It uses dump_execution_state(), as you have seen when you
modified it, and hence it does indeed force an exception on each CPU in
turn, and the exception handler dumps the cpu_user_regs (also known as
''pt_regs'' in ia64) pointer that it receives.

 -- Keir



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