vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 1234 bytes --]

tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
head:   3906fe9bb7f1a2c8667ae54e967dc8690824f4ea
commit: 7bb39313cd6239e7eb95198950a02b4ad2a08316 x86/mce: Make mce_timed_out() identify holdout CPUs
date:   10 months ago
config: x86_64-buildonly-randconfig-r006-20211022 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
reproduce (this is a W=1 build):
        # https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7bb39313cd6239e7eb95198950a02b4ad2a08316
        git remote add linus https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
        git fetch --no-tags linus master
        git checkout 7bb39313cd6239e7eb95198950a02b4ad2a08316
        # save the attached .config to linux build tree
        mkdir build_dir
        make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All warnings (new ones prefixed by >>):

>> vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 37142 bytes --]

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

Hi,

On Tue, Oct 26, 2021 at 04:16:21PM +0800, kernel test robot wrote:
> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> head:   3906fe9bb7f1a2c8667ae54e967dc8690824f4ea
> commit: 7bb39313cd6239e7eb95198950a02b4ad2a08316 x86/mce: Make mce_timed_out() identify holdout CPUs
> date:   10 months ago
> config: x86_64-buildonly-randconfig-r006-20211022 (attached as .config)
> compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
> reproduce (this is a W=1 build):
>         # https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7bb39313cd6239e7eb95198950a02b4ad2a08316
>         git remote add linus https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
>         git fetch --no-tags linus master
>         git checkout 7bb39313cd6239e7eb95198950a02b4ad2a08316
>         # save the attached .config to linux build tree
>         mkdir build_dir
>         make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash
> 
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp@intel.com>
> 
> All warnings (new ones prefixed by >>):
> 
>>> vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

if you have the technical capability to ignore all noinstr warnings
coming from the mce code, please do so.

I have a patchset to address that but more important work preempts me
constantly from cleaning it up and sending it out properly. ;-\

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer, HRB 36809, AG Nürnberg

Re: [kbuild-all] Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Chen, Rong A]

On 10/26/2021 4:23 PM, Borislav Petkov wrote:
> Hi,
> 
> On Tue, Oct 26, 2021 at 04:16:21PM +0800, kernel test robot wrote:
>> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
>> head:   3906fe9bb7f1a2c8667ae54e967dc8690824f4ea
>> commit: 7bb39313cd6239e7eb95198950a02b4ad2a08316 x86/mce: Make mce_timed_out() identify holdout CPUs
>> date:   10 months ago
>> config: x86_64-buildonly-randconfig-r006-20211022 (attached as .config)
>> compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
>> reproduce (this is a W=1 build):
>>          # https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7bb39313cd6239e7eb95198950a02b4ad2a08316
>>          git remote add linus https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
>>          git fetch --no-tags linus master
>>          git checkout 7bb39313cd6239e7eb95198950a02b4ad2a08316
>>          # save the attached .config to linux build tree
>>          mkdir build_dir
>>          make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash
>>
>> If you fix the issue, kindly add following tag as appropriate
>> Reported-by: kernel test robot <lkp@intel.com>
>>
>> All warnings (new ones prefixed by >>):
>>
>>>> vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section
> 
> if you have the technical capability to ignore all noinstr warnings
> coming from the mce code, please do so.
> 
> I have a patchset to address that but more important work preempts me
> constantly from cleaning it up and sending it out properly. ;-\
> 

Hi Borislav,

Thanks for the feedback, We have ignored these warnings from mce.

Best Regards,
Rong Chen

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Tue, 26 Oct 2021 10:23:30 +0200 Borislav Petkov wrote:
> On Tue, Oct 26, 2021 at 04:16:21PM +0800, kernel test robot wrote:
> > tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> > head:   3906fe9bb7f1a2c8667ae54e967dc8690824f4ea
> > commit: 7bb39313cd6239e7eb95198950a02b4ad2a08316 x86/mce: Make mce_timed_out() identify holdout CPUs
> > date:   10 months ago
> > config: x86_64-buildonly-randconfig-r006-20211022 (attached as .config)
> > compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
> > reproduce (this is a W=1 build):
> >         # https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7bb39313cd6239e7eb95198950a02b4ad2a08316
> >         git remote add linus https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> >         git fetch --no-tags linus master
> >         git checkout 7bb39313cd6239e7eb95198950a02b4ad2a08316
> >         # save the attached .config to linux build tree
> >         mkdir build_dir
> >         make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash
> > 
> > If you fix the issue, kindly add following tag as appropriate
> > Reported-by: kernel test robot <lkp@intel.com>
> > 
> > All warnings (new ones prefixed by >>):
> 
> if you have the technical capability to ignore all noinstr warnings
> coming from the mce code, please do so.
> 
> I have a patchset to address that but more important work preempts me
> constantly from cleaning it up and sending it out properly. ;-\

Any progress? Anything I can help with?

I was hoping that 5.17 would fix this but I just merged Linus's tree
with netdev today and the warnings are still there :(

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

On Wed, Jan 19, 2022 at 10:22:50AM -0800, Jakub Kicinski wrote:
> I was hoping that 5.17 would fix this but I just merged Linus's tree
> with netdev today and the warnings are still there :(

Hmm, weird. The set which is supposed to address this got merged last
week:

7e740ae63504 ("Merge tag 'ras_core_for_v5.17_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip")

How exactly can I reproduce what you're seeing?

I need to merge which commit into which branch and then use which
.config to reproduce?

Thx.

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Ivo Totev, HRB 36809, AG Nürnberg

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Wed, 19 Jan 2022 19:34:24 +0100 Borislav Petkov wrote:
> On Wed, Jan 19, 2022 at 10:22:50AM -0800, Jakub Kicinski wrote:
> > I was hoping that 5.17 would fix this but I just merged Linus's tree
> > with netdev today and the warnings are still there :(  
> 
> Hmm, weird. The set which is supposed to address this got merged last
> week:
> 
> 7e740ae63504 ("Merge tag 'ras_core_for_v5.17_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip")
> 
> How exactly can I reproduce what you're seeing?
> 
> I need to merge which commit into which branch and then use which
> .config to reproduce?

allmodconfig 

Merge: 1d1df41c5a33 ff9fc0a31d85

1d1df41c5a33 is linux/master 
ff9fc0a31d85 is
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git master

I see:

vmlinux.o: warning: objtool: mce_start()+0x5e: call to __kasan_check_write() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.isra.1() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_read_aux()+0x8a: call to mca_msr_reg() leaves .noinstr.text section
vmlinux.o: warning: objtool: do_machine_check()+0x196: call to mce_no_way_out() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_severity_amd.constprop.1()+0xd1: call to mce_severity_amd_smca.isra.0() leaves .noinstr.text section

Let me try again with a clean check out.

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Wed, 19 Jan 2022 10:40:17 -0800 Jakub Kicinski wrote:
> On Wed, 19 Jan 2022 19:34:24 +0100 Borislav Petkov wrote:
> > Hmm, weird. The set which is supposed to address this got merged last
> > week:
> > 
> > 7e740ae63504 ("Merge tag 'ras_core_for_v5.17_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip")
> > 
> > How exactly can I reproduce what you're seeing?
> > 
> > I need to merge which commit into which branch and then use which
> > .config to reproduce?  
> 
> allmodconfig 
> 
> Merge: 1d1df41c5a33 ff9fc0a31d85
> 
> 1d1df41c5a33 is linux/master 
> ff9fc0a31d85 is
> git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git master
> 
> I see:
> 
> vmlinux.o: warning: objtool: mce_start()+0x5e: call to __kasan_check_write() leaves .noinstr.text section
> vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.isra.1() leaves .noinstr.text section
> vmlinux.o: warning: objtool: mce_read_aux()+0x8a: call to mca_msr_reg() leaves .noinstr.text section
> vmlinux.o: warning: objtool: do_machine_check()+0x196: call to mce_no_way_out() leaves .noinstr.text section
> vmlinux.o: warning: objtool: mce_severity_amd.constprop.1()+0xd1: call to mce_severity_amd_smca.isra.0() leaves .noinstr.text section
> 
> Let me try again with a clean check out.

Indeed, still present with a fresh tree.

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

On Wed, Jan 19, 2022 at 11:15:23AM -0800, Jakub Kicinski wrote:
> Indeed, still present with a fresh tree.

Doesn't fire here with

$ gcc --version
gcc (SUSE Linux) 7.4.1 20190905 [gcc-7-branch revision 275407]

which compiler are you using?

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Ivo Totev, HRB 36809, AG Nürnberg

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Wed, 19 Jan 2022 21:30:15 +0100 Borislav Petkov wrote:
> On Wed, Jan 19, 2022 at 11:15:23AM -0800, Jakub Kicinski wrote:
> > Indeed, still present with a fresh tree.  
> 
> Doesn't fire here with
> 
> $ gcc --version
> gcc (SUSE Linux) 7.4.1 20190905 [gcc-7-branch revision 275407]
> 
> which compiler are you using?

gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-3)

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

On Wed, Jan 19, 2022 at 12:37:14PM -0800, Jakub Kicinski wrote:
> gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-3)

I have

$ gcc-8 --version
gcc-8 (SUSE Linux) 8.2.1 20180831 [gcc-8-branch revision 264010]

but I can't trigger anything with it either.

So I had a report about a randconfig from the 0day robot guys which
doesn't look like allmodconfig but which triggered some of the issues
you're seeing and I have fixes for that one which we wanna queue after
-rc1.

You could try them to see if they fix your warnings too. First take the
2 here:

https://lore.kernel.org/r/20220113155357.4706-1-bp@alien8.de

and then this one ontop:

https://lore.kernel.org/r/YeWsopFagTZ4W8WA@zn.tnic

Thx.
 
-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Ivo Totev, HRB 36809, AG Nürnberg

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Wed, 19 Jan 2022 22:16:16 +0100 Borislav Petkov wrote:
> On Wed, Jan 19, 2022 at 12:37:14PM -0800, Jakub Kicinski wrote:
> > gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-3)  
> 
> I have
> 
> $ gcc-8 --version
> gcc-8 (SUSE Linux) 8.2.1 20180831 [gcc-8-branch revision 264010]
> 
> but I can't trigger anything with it either.
> 
> So I had a report about a randconfig from the 0day robot guys which
> doesn't look like allmodconfig but which triggered some of the issues
> you're seeing and I have fixes for that one which we wanna queue after
> -rc1.
> 
> You could try them to see if they fix your warnings too. First take the
> 2 here:
> 
> https://lore.kernel.org/r/20220113155357.4706-1-bp@alien8.de
> 
> and then this one ontop:
> 
> https://lore.kernel.org/r/YeWsopFagTZ4W8WA@zn.tnic
> 
> Thx.

That does fix one of the warnings, it seems.

Before:

vmlinux.o: warning: objtool: mce_start()+0x5e: call to __kasan_check_write() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.isra.1() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_read_aux()+0x8a: call to mca_msr_reg() leaves .noinstr.text section
vmlinux.o: warning: objtool: do_machine_check()+0x196: call to mce_no_way_out() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_severity_amd.constprop.1()+0xd1: call to mce_severity_amd_smca.isra.0() leaves .noinstr.text section

After:

vmlinux.o: warning: objtool: mce_start()+0x5e: call to __kasan_check_write() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.isra.1() leaves .noinstr.text section
vmlinux.o: warning: objtool: do_machine_check()+0xa32: call to mce_clear_state() leaves .noinstr.text section
vmlinux.o: warning: objtool: mce_severity_amd.constprop.1()+0xd1: call to mce_severity_amd_smca.isra.0() leaves .noinstr.text section


At a glance mce_read_aux() is gone and do_machine_check() complains
about a different call.

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

Just some off-the-top-of-my-head thoughts:

On Wed, Jan 19, 2022 at 02:57:27PM -0800, Jakub Kicinski wrote:
> vmlinux.o: warning: objtool: mce_start()+0x5e: call to __kasan_check_write() leaves .noinstr.text section

Disable KASAN for the whole of MCE, I dunno.

> vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.isra.1() leaves .noinstr.text section

v8086_mode() is a oneliner which does "return 0;" on 64-bit. I can't
fathom how your crazy compiler decided not to inline it. Unless there's
instrumentation gunk blowing this function up.

> vmlinux.o: warning: objtool: do_machine_check()+0xa32: call to mce_clear_state() leaves .noinstr.text section

I guess we can inline that - it is simple and has a single call site.
The compiler should have inlined it but for some reason I doesn't want
to.

> vmlinux.o: warning: objtool: mce_severity_amd.constprop.1()+0xd1: call to mce_severity_amd_smca.isra.0() leaves .noinstr.text section

Your funny compiler doesn't inline that one either even if it is called
only once, just like the above.

So, I'd like to reproduce that here and play with it a bit. What iso
can I install in a guest here so that I can get pretty close to your
environment?

Thx.

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Ivo Totev, HRB 36809, AG Nürnberg

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Thu, 20 Jan 2022 00:17:23 +0100 Borislav Petkov wrote:
> So, I'd like to reproduce that here and play with it a bit. What iso
> can I install in a guest here so that I can get pretty close to your
> environment?

It's CentOS Stream 8. It may have some extra backports done by my $corp
but I don't think the compiler is backported.

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

On Wed, Jan 19, 2022 at 03:32:18PM -0800, Jakub Kicinski wrote:
> On Thu, 20 Jan 2022 00:17:23 +0100 Borislav Petkov wrote:
> > So, I'd like to reproduce that here and play with it a bit. What iso
> > can I install in a guest here so that I can get pretty close to your
> > environment?
> 
> It's CentOS Stream 8. It may have some extra backports done by my $corp
> but I don't think the compiler is backported.

Ok, while trying to boot an externally supplied kernel with qemu -kernel
and with a guest install of centos8, I managed to trigger the issues on
the test box.

I've pushed a branch here for you to test:

https://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git/log/?h=rc1%2bmce-noinstr

Lemme know how that goes.

Thx.

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Ivo Totev, HRB 36809, AG Nürnberg

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Jakub Kicinski]

On Mon, 31 Jan 2022 13:06:45 +0100 Borislav Petkov wrote:
> Ok, while trying to boot an externally supplied kernel with qemu -kernel
> and with a guest install of centos8, I managed to trigger the issues on
> the test box.
> 
> I've pushed a branch here for you to test:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git/log/?h=rc1%2bmce-noinstr
> 
> Lemme know how that goes.

CLEAN! All the warnings are gone, thank you :)

Re: vmlinux.o: warning: objtool: do_machine_check()+0x5fa: call to memset() leaves .noinstr.text section

[Borislav Petkov]

On Mon, Jan 31, 2022 at 08:56:11AM -0800, Jakub Kicinski wrote:
> CLEAN! All the warnings are gone, thank you :)

Thanks for testing. I'll run some more randconfig build tests and send
them out properly.

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Ivo Totev, HRB 36809, AG Nürnberg