| 2015-08-10 21:43:03 |
Chris J Arges |
description |
Problem Description
=============================
crash's sub-command cannot show stack frame of active tasks on ppc64le systems. Please see
https://www.redhat.com/archives/crash-utility/2015-January/msg00033.html
for details.
Contact Information = Ping Tian Han/pthan@cn.ibm.com, Mikhail Afanasiev/afanasie@us.ibm.com
---uname output---
Linux thymelp2.isst.aus.stglabs.ibm.com 3.10.0-221.ael7b.ppc64le #1 SMP Wed Jan 7 09:27:09 EST 2015 ppc64le ppc64le ppc64le GNU/Linux
Machine Type = CHRP IBM,8247-22L lpar
Steps to Reproduce
=============================
1. capturing a vmcore by kdump on ppc64le system
2. issuing crash with this vmcore
3. run 'bt' in crash
4. no stack frame displaied
Userspace tool common name: crash
The userspace tool has the following bit modes: 64-bit
Userspace rpm: crash-7.0.9-2.ael7b.ppc64le
== Comment: #5 - Hari Krishna Bathini <hbathini@in.ibm.com> - 2015-02-06 13:18:18 ==
For active tasks, there are two methods to get backtrace.
Firstly, using pt_note registers. If this fails,
use default stack search method to get backtrace.
In ppc64le, currently both methods seem to fail.
The below patch resolves problem in default stack search method.
https://github.com/crash-utility/crash/commit/dc4ea682a21567dd9d093862ec54eb8529199c05
This effectively resolves this bug.
But the problem with first method is still open.
I couldn't look into the problem with first method yet.
Will try to work on this next week.
Thanks
Hari
== Comment: #6 - Hari Krishna Bathini <hbathini@in.ibm.com> - 2015-02-20 00:08:46 ==
(In reply to comment #5)
> (In reply to comment #4)
> > Any update?
>
> For active tasks, there are two methods to get backtrace.
> Firstly, using pt_note registers. If this fails,
> use default stack search method to get backtrace.
> In ppc64le, currently both methods seem to fail.
> The below patch resolves problem in default stack search method.
> https://github.com/crash-utility/crash/commit/
> dc4ea682a21567dd9d093862ec54eb8529199c05
> This effectively resolves this bug.
>
> But the problem with first method is still open.
> I couldn't look into the problem with first method yet.
> Will try to work on this next week.
>
> Thanks
> Hari
I would suggest, we close this bug as the issue reported is resolved with the below patch
https://github.com/crash-utility/crash/commit/dc4ea682a21567dd9d093862ec54eb8529199c05
We could track the problem with first method offline or in a separate bug if needed.
Thanks
Hari
== Comment: #10 - Breno Henrique Leitao <brenohl@br.ibm.com> - 2015-03-02 15:42:05 ==
Canonical,
Can we move Crash to versin 7.1.0 that already contains this fix?
Thanks.
== Comment: #11 - Hari Krishna Bathini <hbathini@in.ibm.com> - 2015-03-05 23:18:17 ==
(In reply to comment #6)
> (In reply to comment #5)
> > (In reply to comment #4)
> > > Any update?
> >
> > For active tasks, there are two methods to get backtrace.
> > Firstly, using pt_note registers. If this fails,
> > use default stack search method to get backtrace.
> > In ppc64le, currently both methods seem to fail.
> > The below patch resolves problem in default stack search method.
> > https://github.com/crash-utility/crash/commit/
> > dc4ea682a21567dd9d093862ec54eb8529199c05
> > This effectively resolves this bug.
> >
> > But the problem with first method is still open.
> > I couldn't look into the problem with first method yet.
> > Will try to work on this next week.
> >
> > Thanks
> > Hari
>
> I would suggest, we close this bug as the issue reported is resolved with
> the below patch
> https://github.com/crash-utility/crash/commit/
> dc4ea682a21567dd9d093862ec54eb8529199c05
>
> We could track the problem with first method offline or in a separate bug if
> needed.
Further, this is more of an alternative approach for getting backtrace.
So, the problem I mentioned here doesn't have any issues in terms of functionality
with regard to this bug or any other for that matter.
Since, there are no issues in terms of functionality, I am not sure on whether to raise a bug.
It is on my TODO list and I am thinking of chasing this offline..
Thanks
Hari |
SRU Justification
[Impact]
crash's sub-command cannot show stack frame of active tasks on ppc64le systems.
[Test Case]
1. capturing a vmcore by kdump on ppc64le system
2. issuing crash with this vmcore
3. run 'bt' in crash
4. no stack frame displaied
[Fix]
https://github.com/crash-utility/crash/commit/dc4ea682a21567dd9d093862ec54eb8529199c05
[Regression potential]
This is easily testable and only affects ppc64le systems. This has already been backported to vivid without issue.
--
Problem Description
=============================
crash's sub-command cannot show stack frame of active tasks on ppc64le systems. Please see
https://www.redhat.com/archives/crash-utility/2015-January/msg00033.html
for details.
Contact Information = Ping Tian Han/pthan@cn.ibm.com, Mikhail Afanasiev/afanasie@us.ibm.com
---uname output---
Linux thymelp2.isst.aus.stglabs.ibm.com 3.10.0-221.ael7b.ppc64le #1 SMP Wed Jan 7 09:27:09 EST 2015 ppc64le ppc64le ppc64le GNU/Linux
Machine Type = CHRP IBM,8247-22L lpar
Steps to Reproduce
=============================
1. capturing a vmcore by kdump on ppc64le system
2. issuing crash with this vmcore
3. run 'bt' in crash
4. no stack frame displaied
Userspace tool common name: crash
The userspace tool has the following bit modes: 64-bit
Userspace rpm: crash-7.0.9-2.ael7b.ppc64le
== Comment: #5 - Hari Krishna Bathini <hbathini@in.ibm.com> - 2015-02-06 13:18:18 ==
For active tasks, there are two methods to get backtrace.
Firstly, using pt_note registers. If this fails,
use default stack search method to get backtrace.
In ppc64le, currently both methods seem to fail.
The below patch resolves problem in default stack search method.
https://github.com/crash-utility/crash/commit/dc4ea682a21567dd9d093862ec54eb8529199c05
This effectively resolves this bug.
But the problem with first method is still open.
I couldn't look into the problem with first method yet.
Will try to work on this next week.
Thanks
Hari
== Comment: #6 - Hari Krishna Bathini <hbathini@in.ibm.com> - 2015-02-20 00:08:46 ==
(In reply to comment #5)
> (In reply to comment #4)
> > Any update?
>
> For active tasks, there are two methods to get backtrace.
> Firstly, using pt_note registers. If this fails,
> use default stack search method to get backtrace.
> In ppc64le, currently both methods seem to fail.
> The below patch resolves problem in default stack search method.
> https://github.com/crash-utility/crash/commit/
> dc4ea682a21567dd9d093862ec54eb8529199c05
> This effectively resolves this bug.
>
> But the problem with first method is still open.
> I couldn't look into the problem with first method yet.
> Will try to work on this next week.
>
> Thanks
> Hari
I would suggest, we close this bug as the issue reported is resolved with the below patch
https://github.com/crash-utility/crash/commit/dc4ea682a21567dd9d093862ec54eb8529199c05
We could track the problem with first method offline or in a separate bug if needed.
Thanks
Hari
== Comment: #10 - Breno Henrique Leitao <brenohl@br.ibm.com> - 2015-03-02 15:42:05 ==
Canonical,
Can we move Crash to versin 7.1.0 that already contains this fix?
Thanks.
== Comment: #11 - Hari Krishna Bathini <hbathini@in.ibm.com> - 2015-03-05 23:18:17 ==
(In reply to comment #6)
> (In reply to comment #5)
> > (In reply to comment #4)
> > > Any update?
> >
> > For active tasks, there are two methods to get backtrace.
> > Firstly, using pt_note registers. If this fails,
> > use default stack search method to get backtrace.
> > In ppc64le, currently both methods seem to fail.
> > The below patch resolves problem in default stack search method.
> > https://github.com/crash-utility/crash/commit/
> > dc4ea682a21567dd9d093862ec54eb8529199c05
> > This effectively resolves this bug.
> >
> > But the problem with first method is still open.
> > I couldn't look into the problem with first method yet.
> > Will try to work on this next week.
> >
> > Thanks
> > Hari
>
> I would suggest, we close this bug as the issue reported is resolved with
> the below patch
> https://github.com/crash-utility/crash/commit/
> dc4ea682a21567dd9d093862ec54eb8529199c05
>
> We could track the problem with first method offline or in a separate bug if
> needed.
Further, this is more of an alternative approach for getting backtrace.
So, the problem I mentioned here doesn't have any issues in terms of functionality
with regard to this bug or any other for that matter.
Since, there are no issues in terms of functionality, I am not sure on whether to raise a bug.
It is on my TODO list and I am thinking of chasing this offline..
Thanks
Hari |
|
