queue/5.15 kernel build failure

queue/5.15 kernel build failure

[Chuck Lever]

Hi-

Building on RHEL 9.6, I encountered this build failure:

arch/x86/kernel/smp.o: warning: objtool: fred_sysvec_reboot()+0x52:
unreachable instruction
drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
vmw_port_hb_out()+0xbf: stack state mismatch: cfa1=5+16 cfa2=4+8
drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
vmw_port_hb_in()+0xb4: stack state mismatch: cfa1=5+16 cfa2=4+8
drivers/vfio/vfio_iommu_type1.c: In function ‘vfio_pin_pages_remote’:
drivers/vfio/vfio_iommu_type1.c:707:25: error: ISO C90 forbids mixed
declarations and code [-Werror=declaration-after-statement]
  707 |                         long req_pages = min_t(long, npage,
batch->capacity);
      |                         ^~~~
cc1: all warnings being treated as errors
gmake[2]: *** [scripts/Makefile.build:289:
drivers/vfio/vfio_iommu_type1.o] Error 1
gmake[1]: *** [scripts/Makefile.build:552: drivers/vfio] Error 2
gmake[1]: *** Waiting for unfinished jobs....
gmake: *** [Makefile:1926: drivers] Error 2

Appears to be due to:

commit 5c87f3aff907e72fa6759c9dc66eb609dec1815c
Author:     Keith Busch <kbusch@kernel.org>
AuthorDate: Tue Jul 15 11:46:22 2025 -0700
Commit:     Sasha Levin <sashal@kernel.org>
CommitDate: Sun Aug 17 09:30:59 2025 -0400

    vfio/type1: conditional rescheduling while pinning

    [ Upstream commit b1779e4f209c7ff7e32f3c79d69bca4e3a3a68b6 ]

    A large DMA mapping request can loop through dma address pinning for
    many pages. In cases where THP can not be used, the repeated
vmf_insert_pfn can
    be costly, so let the task reschedule as need to prevent CPU stalls.
Failure to

  ...

-- 
Chuck Lever

Re: queue/5.15 kernel build failure

[Sasha Levin]

On Sun, Aug 17, 2025 at 01:11:14PM -0400, Chuck Lever wrote:
>Hi-
>
>Building on RHEL 9.6, I encountered this build failure:
>
>arch/x86/kernel/smp.o: warning: objtool: fred_sysvec_reboot()+0x52:
>unreachable instruction
>drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
>vmw_port_hb_out()+0xbf: stack state mismatch: cfa1=5+16 cfa2=4+8
>drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
>vmw_port_hb_in()+0xb4: stack state mismatch: cfa1=5+16 cfa2=4+8
>drivers/vfio/vfio_iommu_type1.c: In function ‘vfio_pin_pages_remote’:
>drivers/vfio/vfio_iommu_type1.c:707:25: error: ISO C90 forbids mixed
>declarations and code [-Werror=declaration-after-statement]
>  707 |                         long req_pages = min_t(long, npage,
>batch->capacity);
>      |                         ^~~~
>cc1: all warnings being treated as errors
>gmake[2]: *** [scripts/Makefile.build:289:
>drivers/vfio/vfio_iommu_type1.o] Error 1
>gmake[1]: *** [scripts/Makefile.build:552: drivers/vfio] Error 2
>gmake[1]: *** Waiting for unfinished jobs....
>gmake: *** [Makefile:1926: drivers] Error 2
>
>Appears to be due to:
>
>commit 5c87f3aff907e72fa6759c9dc66eb609dec1815c

I've dropped this, thanks for the report.

It's a bit funny - my version of gcc treats it as a warning, and it actually
gives me quite a few mote "mixed decrlataions" warnings in the 5.15
allmodconfig build.

Compilers are hard :)

-- 
Thanks,
Sasha

Re: queue/5.15 kernel build failure

[Chuck Lever]

On 8/17/25 4:02 PM, Sasha Levin wrote:
> On Sun, Aug 17, 2025 at 01:11:14PM -0400, Chuck Lever wrote:
>> Hi-
>>
>> Building on RHEL 9.6, I encountered this build failure:
>>
>> arch/x86/kernel/smp.o: warning: objtool: fred_sysvec_reboot()+0x52:
>> unreachable instruction
>> drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
>> vmw_port_hb_out()+0xbf: stack state mismatch: cfa1=5+16 cfa2=4+8
>> drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
>> vmw_port_hb_in()+0xb4: stack state mismatch: cfa1=5+16 cfa2=4+8
>> drivers/vfio/vfio_iommu_type1.c: In function ‘vfio_pin_pages_remote’:
>> drivers/vfio/vfio_iommu_type1.c:707:25: error: ISO C90 forbids mixed
>> declarations and code [-Werror=declaration-after-statement]
>>  707 |                         long req_pages = min_t(long, npage,
>> batch->capacity);
>>      |                         ^~~~
>> cc1: all warnings being treated as errors
>> gmake[2]: *** [scripts/Makefile.build:289:
>> drivers/vfio/vfio_iommu_type1.o] Error 1
>> gmake[1]: *** [scripts/Makefile.build:552: drivers/vfio] Error 2
>> gmake[1]: *** Waiting for unfinished jobs....
>> gmake: *** [Makefile:1926: drivers] Error 2
>>
>> Appears to be due to:
>>
>> commit 5c87f3aff907e72fa6759c9dc66eb609dec1815c
> 
> I've dropped this, thanks for the report.
> 
> It's a bit funny - my version of gcc treats it as a warning, and it
> actually
> gives me quite a few mote "mixed decrlataions" warnings in the 5.15
> allmodconfig build.
> 
> Compilers are hard :)
> 

Additional context: I copied the RHEL 9.6 /boot/config to do the build.
I think Red Hat likes to keep the "treat warnings as errors" setting
enabled in their builds.


-- 
Chuck Lever

Re: queue/5.15 kernel build failure

[Greg KH]

On Sun, Aug 17, 2025 at 04:05:18PM -0400, Chuck Lever wrote:
> On 8/17/25 4:02 PM, Sasha Levin wrote:
> > On Sun, Aug 17, 2025 at 01:11:14PM -0400, Chuck Lever wrote:
> >> Hi-
> >>
> >> Building on RHEL 9.6, I encountered this build failure:
> >>
> >> arch/x86/kernel/smp.o: warning: objtool: fred_sysvec_reboot()+0x52:
> >> unreachable instruction
> >> drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
> >> vmw_port_hb_out()+0xbf: stack state mismatch: cfa1=5+16 cfa2=4+8
> >> drivers/gpu/drm/vmwgfx/vmwgfx_msg.o: warning: objtool:
> >> vmw_port_hb_in()+0xb4: stack state mismatch: cfa1=5+16 cfa2=4+8
> >> drivers/vfio/vfio_iommu_type1.c: In function ‘vfio_pin_pages_remote’:
> >> drivers/vfio/vfio_iommu_type1.c:707:25: error: ISO C90 forbids mixed
> >> declarations and code [-Werror=declaration-after-statement]
> >>  707 |                         long req_pages = min_t(long, npage,
> >> batch->capacity);
> >>      |                         ^~~~
> >> cc1: all warnings being treated as errors
> >> gmake[2]: *** [scripts/Makefile.build:289:
> >> drivers/vfio/vfio_iommu_type1.o] Error 1
> >> gmake[1]: *** [scripts/Makefile.build:552: drivers/vfio] Error 2
> >> gmake[1]: *** Waiting for unfinished jobs....
> >> gmake: *** [Makefile:1926: drivers] Error 2
> >>
> >> Appears to be due to:
> >>
> >> commit 5c87f3aff907e72fa6759c9dc66eb609dec1815c
> > 
> > I've dropped this, thanks for the report.
> > 
> > It's a bit funny - my version of gcc treats it as a warning, and it
> > actually
> > gives me quite a few mote "mixed decrlataions" warnings in the 5.15
> > allmodconfig build.
> > 
> > Compilers are hard :)
> > 
> 
> Additional context: I copied the RHEL 9.6 /boot/config to do the build.
> I think Red Hat likes to keep the "treat warnings as errors" setting
> enabled in their builds.

It's a good warning to have, 'allmodconfig' enables it, and so my builds
show these as failures too, so you aren't alone.

thanks for pointing it out.

greg k-h