kexec/kdump not working in ubuntu 16.04

Thank you for taking the time to report this bug and helping to make Ubuntu better. It seems that your bug report is not filed about a specific source package though, rather it is just filed against Ubuntu in general. It is important that bug reports be filed about source packages so that people interested in the package can find the bugs about it. You can find some hints about determining what package your bug might be about at https://wiki.ubuntu.com/Bugs/FindRightPackage. You might also ask for help in the #ubuntu-bugs irc channel on Freenode.

To change the source package that this bug is filed about visit https://bugs.launchpad.net/ubuntu/+bug/1546260/+editstatus and add the package name in the text box next to the word Package.

[This is an automated message. I apologize if it reached you inappropriately; please just reply to this message indicating so.]

------- Comment From <email address hidden> 2016-02-17 13:30 EDT-------
verify_sha256_digest() function call is returning 1 in the below snippet
on ubuntu 16.04, which is leading to hang
---
printf("I'm in purgatory\n");
setup_arch();
if (verify_sha256_digest()) {
for(;;) {
/* loop forever */
}
}
---

The problem seems to be with the compiler. kexec binary in ubuntu 16.04 is compiled
with gcc v5.2.1 [GCC: (Ubuntu 5.2.1-27ubuntu1) 5.2.1 20151129 to be precise].
On using a kexec binary compiled with gcc v4.9.2, we don't hit the problem.
Not sure what changed between v4.9.2 and v5..1 which makes the purgatory unhappy.

While someone from compiler team looks at this, you could use a kexec binary build with
gcc v4.9.2 as a workaround for this problem

Thanks
Hari

------- Comment From <email address hidden> 2016-02-19 09:22 EDT-------
Please provide materials that will allow the compiler team to reproduce the bug, including preprocessed source and compilation flags. Detailed instructions for GCC bug reporting may be found at https://gcc.gnu.org/bugs/.

------- Comment From <email address hidden> 2016-02-22 12:46 EDT-------
Hi William

Purgatory code checks for the sanity of the code before proceeding [with verify_sha256_digest() call]
This call is failing when compiled with gcc verison 5.3.1-8ubuntu2 (available in Ubuntu 16.04) while
it works just fine when compiled with gcc version 4.9.2-10ubuntu12 available in ubuntu 15.04.

Couple of observations that may be relevant:

1. The below warning is observed when compiled with gcc version 5.3.1-8ubuntu2

purgatory/arch/ppc64/console-ppc64.c: In function 'putchar':
purgatory/arch/ppc64/console-ppc64.c:42:2: warning: '*((void *)&buff+8)' may be used uninitialized in this function [-Wmaybe-uninitialized]
plpar_hcall_norets(H_PUT_TERM_CHAR, 0, 1,

2. In the failing environment "printf("I'm in purgatory\n")" also yields no output on console for pseries.
This usually would succeed if the call plpar_hcall_norets() (hypervisor call) goes through.

Also, things work fine with 5.3.1-8ubuntu2 compiler if -O0 is used instead of -Os.
Attaching the preprocessed code for purgatory and also the compile options. Please let me know,
if I need to add more material.

Thanks
Hari

------- Comment (attachment only) From <email address hidden> 2016-02-22 12:48 EDT-------

------- Comment (attachment only) From <email address hidden> 2016-02-22 12:49 EDT-------

------- Comment From <email address hidden> 2016-02-22 14:12 EDT-------
*** Bug 136794 has been marked as a duplicate of this bug. ***

------- Comment From <email address hidden> 2016-02-22 15:51 EDT-------
Hari, thanks, very helpful! There is still at least one bit missing for me to build the whole thing -- could you please attach ppc64_asm.h? This is included by two of the .S files.

------- Comment on attachment From <email address hidden> 2016-02-23 00:22 EDT-------

(In reply to comment #27)
> Hari, thanks, very helpful! There is still at least one bit missing for me
> to build the whole thing -- could you please attach ppc64_asm.h? This is
> included by two of the .S files.

Sorry for missing that out.

Thanks
Hari

------- Comment From <email address hidden> 2016-02-24 19:04 EDT-------
One thing I noticed may be relevant since we're talking about -Os code, which means out-of-line register save and restore functions. gcc-5 moves some code before the register save call, while gcc-4.9 doesn't seem to. It is important that the call to functions like _savegpr0_14 *not* go through stubs (for example to reach a destination more than 33M away) that might destroy regs. The save/restore functions have non-standard calling conventions, so you can't use stubs that destroy *any* of the volatile regs. In particular, r0, r11 and r12 are not available for use by a stub.

------- Comment on attachment From <email address hidden> 2016-02-25 07:44 EDT-------

This is not a compiler issue after all. See
http://lists.infradead.org/pipermail/kexec/2016-February/015440.html and
http://lists.infradead.org/pipermail/kexec/2016-February/015441.html

These two plus this patch should fix the problem

------- Comment (attachment only) From <email address hidden> 2016-02-25 07:46 EDT-------

------- Comment (attachment only) From <email address hidden> 2016-02-25 07:47 EDT-------

------- Comment on attachment From <email address hidden> 2016-02-25 22:00 EDT-------

The toc alignment patch didn't apply cleanly after Anton's patches. This tarfile contains a rebased patchset including Anton's patches.

------- Comment From <email address hidden> 2016-02-26 04:50 EDT-------
> kexec load operation failed with this patch on an ubuntu lpar (ppc64le).

Yes, please drop "Use ld to provide ppc64 register save/restore functions". This ld feature is broken. :-(

------- Comment (attachment only) From <email address hidden> 2016-02-22 12:49 EDT-------

------- Comment From <email address hidden> 2016-03-09 15:25 EDT-------
*** Bug 138700 has been marked as a duplicate of this bug. ***

------- Comment From <email address hidden> 2016-03-09 15:28 EDT-------
*** Bug 138700 has been marked as a duplicate of this bug. ***

------- Comment From <email address hidden> 2016-03-13 09:48 EDT-------
*** Bug 138700 has been marked as a duplicate of this bug. ***

------- Comment From <email address hidden> 2016-03-15 10:11 EDT-------
The bug is fixed by the below patch-set posted by Anton & Alan

http://lists.infradead.org/pipermail/kexec/2016-February/015446.html
http://lists.infradead.org/pipermail/kexec/2016-February/015447.html
http://lists.infradead.org/pipermail/kexec/2016-February/015448.html

------- Comment From <email address hidden> 2016-03-21 13:03 EDT-------
Moving previous comment to external - Canonical please consider this patch.
(In reply to comment #56)
> Hi Hari
>
> able to apply the patches cleanly on kexec-tools-2.0.11 I think
> canonical can apply on kexec-tools-2.0.11 .
>
> Regards
> Praveen

------- Comment From <email address hidden> 2016-03-24 01:46 EDT-------
(In reply to comment #51)
> The bug is fixed by the below patch-set posted by Anton & Alan
>

...now accepted upstream with commit ids

> http://lists.infradead.org/pipermail/kexec/2016-February/015446.html

4a2ae3a39c64dc43e9d094be9541253234ff4822

> http://lists.infradead.org/pipermail/kexec/2016-February/015447.html

1e423dc297d10eb7ff25c829d2856ef12fc81d77

> http://lists.infradead.org/pipermail/kexec/2016-February/015448.html

3debb8cf3272216119cb2e59a4963ce3c18fe8e3

Thanks
Hari

------- Comment (attachment only) From <email address hidden> 2016-02-22 12:48 EDT-------

------- Comment (attachment only) From <email address hidden> 2016-02-22 12:49 EDT-------

------- Comment on attachment From <email address hidden> 2016-02-23 00:22 EDT-------

(In reply to comment #27)
> Hari, thanks, very helpful! There is still at least one bit missing for me
> to build the whole thing -- could you please attach ppc64_asm.h? This is
> included by two of the .S files.

Sorry for missing that out.

Thanks
Hari

------- Comment on attachment From <email address hidden> 2016-02-25 22:00 EDT-------

The toc alignment patch didn't apply cleanly after Anton's patches. This tarfile contains a rebased patchset including Anton's patches.

Good news! The patches were just accepted earlier today:

http://git.kernel.org/cgit/utils/kernel/kexec/kexec-tools.git/commit/?id=4a2ae3a39c64dc43e9d094be9541253234ff4822
http://git.kernel.org/cgit/utils/kernel/kexec/kexec-tools.git/commit/?id=1e423dc297d10eb7ff25c829d2856ef12fc81d77
http://git.kernel.org/cgit/utils/kernel/kexec/kexec-tools.git/commit/?id=3debb8cf3272216119cb2e59a4963ce3c18fe8e3

Hello,

I have a new test package ready for testing if you are able to verify. I included the three upstream commits. I haven't identified the changes made to the purgatory/arch/ppc64/Makefile file (part of the rebase tarfile) so those are not included.

Since you are already using the PPA that contains kdump-tools 1:1.5.9-4~lp1536904, I can make the test package available there so you can test without having to add a new PPA.

You should be aware though that you should no longer use that PPA. The modifications made to kexec-tools are now present in the public archive. in the 1:1.5.9-5 package.

------- Comment From <email address hidden> 2016-03-28 07:54 EDT-------
(In reply to comment #68)
> Hello,
>
> I have a new test package ready for testing if you are able to verify. I
> included the three upstream commits. I haven't identified the changes made
> to the purgatory/arch/ppc64/Makefile file (part of the rebase tarfile) so
> those are not included.
>
> Since you are already using the PPA that contains kdump-tools
> 1:1.5.9-4~lp1536904, I can make the test package available there so you can
> test without having to add a new PPA.
>
> You should be aware though that you should no longer use that PPA. The
> modifications made to kexec-tools are now present in the public archive. in
> the 1:1.5.9-5 package.

Hi louis ,

need one help i got confused with above statement , we have a bug raised for broken link for kdump initrd and vmlinux there we got PPA build of kdump 1:1.5.9-4~lp1536904 and that got resolved , is all three patches (kexec) you are pushing in same PPA repo or it is part of official build 1:1.5.9-5 package.

Regards
Praveen

Hello,

I am sorry that my previous comment was unclear and a bit misleading.

The kdump package in the PPA (kdump 1:1.5.9-4~lp1536904) fixes https://bugs.launchpad.net/ubuntu/+source/makedumpfile/+bug/1536904 as you said. This is now in the official archive so you should no longer be using the PPA. Please confirm that you will be using the official archive so I can proceed to delete this PPA.

Now the kexec-tools fixes have been uploaded to a different PPA : ppa:louis-bouchard/kexec-tools-test

Please verify that the new package do indeed fixes your issue. If this is the case, I will see that it gets uploaded to the archive.

Kind regards,

...Louis

Hey Louis, thanks for the package.

> Now the kexec-tools fixes have been uploaded to a different PPA : ppa:louis-bouchard/kexec-tools-test

It seems that this PPA is not enabled to build for ppc64el architecture. Would you enable that, please?

Meanwhile, i can test with building from source.

Anyway, I just got the source package from ppa:louis-bouchard/kexec-tools-test and built it on ppc64el and it seems to be working fine.

The dump files are being generated:

$ find
.
./kexec_cmd
./201603291535
./201603291535/dump.201603291535
./201603291535/dmesg.201603291535
./201603291717
./201603291717/dump.201603291717
./201603291717/dmesg.201603291717

I think we are good to have it on the main archive

------- Comment (attachment only) From <email address hidden> 2016-03-30 13:06 EDT-------

------- Comment From <email address hidden> 2016-04-04 00:57 EDT-------
Hi, Canonical. What is the ETA for this to be released?

Hello,

Sorry for the missing build and missing answers, I was not subscribed to the bug.

Following your tests, I have uploaded the fixed version to the archive which is waiting for approval since we are under beta freeze.

Since this is bugfixing, it should be approved. I'll let you know once it hits the archive.

Kind regards,

...Louis

This bug was fixed in the package kexec-tools - 1:2.0.10-1ubuntu2

---------------
kexec-tools (1:2.0.10-1ubuntu2) xenial; urgency=medium

  * [PowerPC64] Fix failure in purgatory when compiled with gcc5
    Application of upstream fixes so kexec-tools work with gcc5 on PowerPC64

    commit 4a2ae3a39c64dc43e9d094be9541253234ff4822,
    1e423dc297d10eb7ff25c829d2856ef12fc81d77,
    3debb8cf3272216119cb2e59a4963ce3c18fe8e3
    [lp: #1546260]

 -- Louis Bouchard <email address hidden> Tue, 05 Apr 2016 11:14:18 +0200

------- Comment From <email address hidden> 2016-04-05 09:42 EDT-------
*** Bug 137124 has been marked as a duplicate of this bug. ***

------- Comment From <email address hidden> 2016-04-11 00:44 EDT-------
Hi

Verified this bug on friday build (8april-2016) and seem me problem got fixed .

Regards
Praveen

------- Comment From <email address hidden> 2016-04-14 19:46 EDT-------
*** Bug 136793 has been marked as a duplicate of this bug. ***