more randconfig failures

more randconfig failures

[Stefano Stabellini]

Hi all,

These are two recent randconfig build failures reported by gitlab (the
two patches that triggered the CI-loop are two patches to the
MAINTAINERS file -- certainly not the cause of the build issues):

x86 randconfig failure:
https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647

arm randconfig failure:
https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472

Cheers,

Stefano

Re: more randconfig failures

[Andrew Cooper]

On 27/01/2021 21:16, Stefano Stabellini wrote:
> Hi all,
>
> These are two recent randconfig build failures reported by gitlab (the
> two patches that triggered the CI-loop are two patches to the
> MAINTAINERS file -- certainly not the cause of the build issues):
>
> x86 randconfig failure:
> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647

This has been known and a source of intermittent failures for ages.

I already laid out how to fix it, by removing some inappropriate Kconfig
dependencies.  I'll get around to writing some patches when there aren't
more important tasks to do for 4.15.

~Andrew

Re: more randconfig failures

[Julien Grall]

Hi,

On Wed, 27 Jan 2021 at 21:16, Stefano Stabellini <sstabellini@kernel.org> wrote:
>
> Hi all,
>
> These are two recent randconfig build failures reported by gitlab (the
> two patches that triggered the CI-loop are two patches to the
> MAINTAINERS file -- certainly not the cause of the build issues):
>
> x86 randconfig failure:
> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
>
> arm randconfig failure:
> https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472

make[5]: Leaving directory '/builds/xen-project/patchew/xen/xen/common/libfdt'
smmu-v3.c: In function 'acpi_smmu_get_options':
smmu-v3.c:3017:7: error: 'ACPI_IORT_SMMU_V3_CAVIUM_CN99XX' undeclared
(first use in this function)
 3017 |  case ACPI_IORT_SMMU_V3_CAVIUM_CN99XX:
      |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
smmu-v3.c:3017:7: note: each undeclared identifier is reported only
once for each function it appears in
smmu-v3.c:3020:7: error: 'ACPI_IORT_SMMU_V3_HISILICON_HI161X'
undeclared (first use in this function)
 3020 |  case ACPI_IORT_SMMU_V3_HISILICON_HI161X:
      |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  INIT_O  efi-dom0.init.o
smmu-v3.c: In function 'arm_smmu_device_acpi_probe':
smmu-v3.c:3035:36: error: implicit declaration of function
'dev_get_platdata' [-Werror=implicit-function-declaration]
 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
      |                                    ^~~~~~~~~~~~~~~~
smmu-v3.c:3035:36: error: nested extern declaration of
'dev_get_platdata' [-Werror=nested-externs]
smmu-v3.c:3035:10: error: cast to pointer from integer of different
size [-Werror=int-to-pointer-cast]
 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
      |          ^
ld    -EL  -r -o built_in.o memcpy.o memcmp.o memmove.o memset.o
memchr.o clear_page.o bitops.o find_next_bit.o strchr.o strcmp.o
strlen.o strncmp.o strnlen.o strrchr.o
cc1: all warnings being treated as errors

It looks like the SMMUv3 driver has not been built tested with
CONFIG_ACPI=y.  Bertrand, Rahul, can you take a look?

Cheers,

Re: more randconfig failures

[Rahul Singh]

Hello Julien,

> On 27 Jan 2021, at 9:25 pm, Julien Grall <julien.grall.oss@gmail.com> wrote:
> 
> Hi,
> 
> On Wed, 27 Jan 2021 at 21:16, Stefano Stabellini <sstabellini@kernel.org> wrote:
>> 
>> Hi all,
>> 
>> These are two recent randconfig build failures reported by gitlab (the
>> two patches that triggered the CI-loop are two patches to the
>> MAINTAINERS file -- certainly not the cause of the build issues):
>> 
>> x86 randconfig failure:
>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
>> 
>> arm randconfig failure:
>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472
> 
> make[5]: Leaving directory '/builds/xen-project/patchew/xen/xen/common/libfdt'
> smmu-v3.c: In function 'acpi_smmu_get_options':
> smmu-v3.c:3017:7: error: 'ACPI_IORT_SMMU_V3_CAVIUM_CN99XX' undeclared
> (first use in this function)
> 3017 |  case ACPI_IORT_SMMU_V3_CAVIUM_CN99XX:
>      |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> smmu-v3.c:3017:7: note: each undeclared identifier is reported only
> once for each function it appears in
> smmu-v3.c:3020:7: error: 'ACPI_IORT_SMMU_V3_HISILICON_HI161X'
> undeclared (first use in this function)
> 3020 |  case ACPI_IORT_SMMU_V3_HISILICON_HI161X:
>      |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>  INIT_O  efi-dom0.init.o
> smmu-v3.c: In function 'arm_smmu_device_acpi_probe':
> smmu-v3.c:3035:36: error: implicit declaration of function
> 'dev_get_platdata' [-Werror=implicit-function-declaration]
> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>      |                                    ^~~~~~~~~~~~~~~~
> smmu-v3.c:3035:36: error: nested extern declaration of
> 'dev_get_platdata' [-Werror=nested-externs]
> smmu-v3.c:3035:10: error: cast to pointer from integer of different
> size [-Werror=int-to-pointer-cast]
> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>      |          ^
> ld    -EL  -r -o built_in.o memcpy.o memcmp.o memmove.o memset.o
> memchr.o clear_page.o bitops.o find_next_bit.o strchr.o strcmp.o
> strlen.o strncmp.o strnlen.o strrchr.o
> cc1: all warnings being treated as errors
> 
> It looks like the SMMUv3 driver has not been built tested with
> CONFIG_ACPI=y.  Bertrand, Rahul, can you take a look?

Sorry my mistake I didn’t enable the ARM_SMMU_V3 with ACPI and tested the compilation.

There are two option to fix this:

1. #undef-ining the CONFIG_ACPI in the smmu-v3.c file and fix the compilation. In this case once we have ACPI IORT table code ported to XEN smmu-v3 code can be used at that time.
2.  Remove the arm_smmu_device_acpi_probe() from the smmu-v3.c now and once we have  ACPI IORT table code ported to XEN we can add it back.

Let me know your view I will send the patch to fix the compilation.

Regards,
Rahul
> 
> Cheers,

Re: more randconfig failures

[Jan Beulich]

On 27.01.2021 22:23, Andrew Cooper wrote:
> On 27/01/2021 21:16, Stefano Stabellini wrote:
>> Hi all,
>>
>> These are two recent randconfig build failures reported by gitlab (the
>> two patches that triggered the CI-loop are two patches to the
>> MAINTAINERS file -- certainly not the cause of the build issues):
>>
>> x86 randconfig failure:
>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
> 
> This has been known and a source of intermittent failures for ages.
> 
> I already laid out how to fix it, by removing some inappropriate Kconfig
> dependencies.  I'll get around to writing some patches when there aren't
> more important tasks to do for 4.15.

Are you sure? I see

ld: ld: /builds/xen-project/patchew/xen/xen/.xen-syms.0: hidden symbol `xenmem_add_to_physmap_one' isn't defined

which rather reminds me of a recent report where DCE doesn't do what we
expect it to do.

Jan

Re: more randconfig failures

[Julien Grall]

On 27/01/2021 23:11, Rahul Singh wrote:
> Hello Julien,

Hi Rahul,

>> On 27 Jan 2021, at 9:25 pm, Julien Grall <julien.grall.oss@gmail.com> wrote:
>>
>> Hi,
>>
>> On Wed, 27 Jan 2021 at 21:16, Stefano Stabellini <sstabellini@kernel.org> wrote:
>>>
>>> Hi all,
>>>
>>> These are two recent randconfig build failures reported by gitlab (the
>>> two patches that triggered the CI-loop are two patches to the
>>> MAINTAINERS file -- certainly not the cause of the build issues):
>>>
>>> x86 randconfig failure:
>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
>>>
>>> arm randconfig failure:
>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472
>>
>> make[5]: Leaving directory '/builds/xen-project/patchew/xen/xen/common/libfdt'
>> smmu-v3.c: In function 'acpi_smmu_get_options':
>> smmu-v3.c:3017:7: error: 'ACPI_IORT_SMMU_V3_CAVIUM_CN99XX' undeclared
>> (first use in this function)
>> 3017 |  case ACPI_IORT_SMMU_V3_CAVIUM_CN99XX:
>>       |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> smmu-v3.c:3017:7: note: each undeclared identifier is reported only
>> once for each function it appears in
>> smmu-v3.c:3020:7: error: 'ACPI_IORT_SMMU_V3_HISILICON_HI161X'
>> undeclared (first use in this function)
>> 3020 |  case ACPI_IORT_SMMU_V3_HISILICON_HI161X:
>>       |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>   INIT_O  efi-dom0.init.o
>> smmu-v3.c: In function 'arm_smmu_device_acpi_probe':
>> smmu-v3.c:3035:36: error: implicit declaration of function
>> 'dev_get_platdata' [-Werror=implicit-function-declaration]
>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>       |                                    ^~~~~~~~~~~~~~~~
>> smmu-v3.c:3035:36: error: nested extern declaration of
>> 'dev_get_platdata' [-Werror=nested-externs]
>> smmu-v3.c:3035:10: error: cast to pointer from integer of different
>> size [-Werror=int-to-pointer-cast]
>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>       |          ^
>> ld    -EL  -r -o built_in.o memcpy.o memcmp.o memmove.o memset.o
>> memchr.o clear_page.o bitops.o find_next_bit.o strchr.o strcmp.o
>> strlen.o strncmp.o strnlen.o strrchr.o
>> cc1: all warnings being treated as errors
>>
>> It looks like the SMMUv3 driver has not been built tested with
>> CONFIG_ACPI=y.  Bertrand, Rahul, can you take a look?
> 
> Sorry my mistake I didn’t enable the ARM_SMMU_V3 with ACPI and tested the compilation.
> 
> There are two option to fix this:
> 
> 1. #undef-ining the CONFIG_ACPI in the smmu-v3.c file and fix the compilation. In this case once we have ACPI IORT table code ported to XEN smmu-v3 code can be used at that time.
> 2.  Remove the arm_smmu_device_acpi_probe() from the smmu-v3.c now and once we have  ACPI IORT table code ported to XEN we can add it back.

How about the following?

3. Only allow the user to build the SMMUv3 when !CONFIG_ACPI.

Cheers,

-- 
Julien Grall

Re: more randconfig failures

[Jan Beulich]

On 28.01.2021 17:34, Julien Grall wrote:
> On 27/01/2021 23:11, Rahul Singh wrote:
>> Hello Julien,
> 
> Hi Rahul,
> 
>>> On 27 Jan 2021, at 9:25 pm, Julien Grall <julien.grall.oss@gmail.com> wrote:
>>>
>>> Hi,
>>>
>>> On Wed, 27 Jan 2021 at 21:16, Stefano Stabellini <sstabellini@kernel.org> wrote:
>>>>
>>>> Hi all,
>>>>
>>>> These are two recent randconfig build failures reported by gitlab (the
>>>> two patches that triggered the CI-loop are two patches to the
>>>> MAINTAINERS file -- certainly not the cause of the build issues):
>>>>
>>>> x86 randconfig failure:
>>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
>>>>
>>>> arm randconfig failure:
>>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472
>>>
>>> make[5]: Leaving directory '/builds/xen-project/patchew/xen/xen/common/libfdt'
>>> smmu-v3.c: In function 'acpi_smmu_get_options':
>>> smmu-v3.c:3017:7: error: 'ACPI_IORT_SMMU_V3_CAVIUM_CN99XX' undeclared
>>> (first use in this function)
>>> 3017 |  case ACPI_IORT_SMMU_V3_CAVIUM_CN99XX:
>>>       |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> smmu-v3.c:3017:7: note: each undeclared identifier is reported only
>>> once for each function it appears in
>>> smmu-v3.c:3020:7: error: 'ACPI_IORT_SMMU_V3_HISILICON_HI161X'
>>> undeclared (first use in this function)
>>> 3020 |  case ACPI_IORT_SMMU_V3_HISILICON_HI161X:
>>>       |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>   INIT_O  efi-dom0.init.o
>>> smmu-v3.c: In function 'arm_smmu_device_acpi_probe':
>>> smmu-v3.c:3035:36: error: implicit declaration of function
>>> 'dev_get_platdata' [-Werror=implicit-function-declaration]
>>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>>       |                                    ^~~~~~~~~~~~~~~~
>>> smmu-v3.c:3035:36: error: nested extern declaration of
>>> 'dev_get_platdata' [-Werror=nested-externs]
>>> smmu-v3.c:3035:10: error: cast to pointer from integer of different
>>> size [-Werror=int-to-pointer-cast]
>>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>>       |          ^
>>> ld    -EL  -r -o built_in.o memcpy.o memcmp.o memmove.o memset.o
>>> memchr.o clear_page.o bitops.o find_next_bit.o strchr.o strcmp.o
>>> strlen.o strncmp.o strnlen.o strrchr.o
>>> cc1: all warnings being treated as errors
>>>
>>> It looks like the SMMUv3 driver has not been built tested with
>>> CONFIG_ACPI=y.  Bertrand, Rahul, can you take a look?
>>
>> Sorry my mistake I didn’t enable the ARM_SMMU_V3 with ACPI and tested the compilation.
>>
>> There are two option to fix this:
>>
>> 1. #undef-ining the CONFIG_ACPI in the smmu-v3.c file and fix the compilation. In this case once we have ACPI IORT table code ported to XEN smmu-v3 code can be used at that time.
>> 2.  Remove the arm_smmu_device_acpi_probe() from the smmu-v3.c now and once we have  ACPI IORT table code ported to XEN we can add it back.
> 
> How about the following?
> 
> 3. Only allow the user to build the SMMUv3 when !CONFIG_ACPI.

And then perhaps as "depends on !ACPI || BROKEN" in Kconfig?

Jan

Re: more randconfig failures

[Rahul Singh]

Hello Julien,

> On 28 Jan 2021, at 4:34 pm, Julien Grall <julien@xen.org> wrote:
> 
> On 27/01/2021 23:11, Rahul Singh wrote:
>> Hello Julien,
> 
> Hi Rahul,
> 
>>> On 27 Jan 2021, at 9:25 pm, Julien Grall <julien.grall.oss@gmail.com> wrote:
>>> 
>>> Hi,
>>> 
>>> On Wed, 27 Jan 2021 at 21:16, Stefano Stabellini <sstabellini@kernel.org> wrote:
>>>> 
>>>> Hi all,
>>>> 
>>>> These are two recent randconfig build failures reported by gitlab (the
>>>> two patches that triggered the CI-loop are two patches to the
>>>> MAINTAINERS file -- certainly not the cause of the build issues):
>>>> 
>>>> x86 randconfig failure:
>>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
>>>> 
>>>> arm randconfig failure:
>>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472
>>> 
>>> make[5]: Leaving directory '/builds/xen-project/patchew/xen/xen/common/libfdt'
>>> smmu-v3.c: In function 'acpi_smmu_get_options':
>>> smmu-v3.c:3017:7: error: 'ACPI_IORT_SMMU_V3_CAVIUM_CN99XX' undeclared
>>> (first use in this function)
>>> 3017 |  case ACPI_IORT_SMMU_V3_CAVIUM_CN99XX:
>>>      |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> smmu-v3.c:3017:7: note: each undeclared identifier is reported only
>>> once for each function it appears in
>>> smmu-v3.c:3020:7: error: 'ACPI_IORT_SMMU_V3_HISILICON_HI161X'
>>> undeclared (first use in this function)
>>> 3020 |  case ACPI_IORT_SMMU_V3_HISILICON_HI161X:
>>>      |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>  INIT_O  efi-dom0.init.o
>>> smmu-v3.c: In function 'arm_smmu_device_acpi_probe':
>>> smmu-v3.c:3035:36: error: implicit declaration of function
>>> 'dev_get_platdata' [-Werror=implicit-function-declaration]
>>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>>      |                                    ^~~~~~~~~~~~~~~~
>>> smmu-v3.c:3035:36: error: nested extern declaration of
>>> 'dev_get_platdata' [-Werror=nested-externs]
>>> smmu-v3.c:3035:10: error: cast to pointer from integer of different
>>> size [-Werror=int-to-pointer-cast]
>>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>>      |          ^
>>> ld    -EL  -r -o built_in.o memcpy.o memcmp.o memmove.o memset.o
>>> memchr.o clear_page.o bitops.o find_next_bit.o strchr.o strcmp.o
>>> strlen.o strncmp.o strnlen.o strrchr.o
>>> cc1: all warnings being treated as errors
>>> 
>>> It looks like the SMMUv3 driver has not been built tested with
>>> CONFIG_ACPI=y.  Bertrand, Rahul, can you take a look?
>> Sorry my mistake I didn’t enable the ARM_SMMU_V3 with ACPI and tested the compilation.
>> There are two option to fix this:
>> 1. #undef-ining the CONFIG_ACPI in the smmu-v3.c file and fix the compilation. In this case once we have ACPI IORT table code ported to XEN smmu-v3 code can be used at that time.
>> 2.  Remove the arm_smmu_device_acpi_probe() from the smmu-v3.c now and once we have  ACPI IORT table code ported to XEN we can add it back.
> 
> How about the following?
> 
> 3. Only allow the user to build the SMMUv3 when !CONFIG_ACPI.
> 
Yes that also works for me I will send the patch to fix.

Regards,
Rahul
> Cheers,
> 
> -- 
> Julien Grall

Re: more randconfig failures

[Julien Grall]

On 28/01/2021 16:46, Jan Beulich wrote:
> On 28.01.2021 17:34, Julien Grall wrote:
>> On 27/01/2021 23:11, Rahul Singh wrote:
>>> Hello Julien,
>>
>> Hi Rahul,
>>
>>>> On 27 Jan 2021, at 9:25 pm, Julien Grall <julien.grall.oss@gmail.com> wrote:
>>>>
>>>> Hi,
>>>>
>>>> On Wed, 27 Jan 2021 at 21:16, Stefano Stabellini <sstabellini@kernel.org> wrote:
>>>>>
>>>>> Hi all,
>>>>>
>>>>> These are two recent randconfig build failures reported by gitlab (the
>>>>> two patches that triggered the CI-loop are two patches to the
>>>>> MAINTAINERS file -- certainly not the cause of the build issues):
>>>>>
>>>>> x86 randconfig failure:
>>>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990347647
>>>>>
>>>>> arm randconfig failure:
>>>>> https://gitlab.com/xen-project/patchew/xen/-/jobs/990335472
>>>>
>>>> make[5]: Leaving directory '/builds/xen-project/patchew/xen/xen/common/libfdt'
>>>> smmu-v3.c: In function 'acpi_smmu_get_options':
>>>> smmu-v3.c:3017:7: error: 'ACPI_IORT_SMMU_V3_CAVIUM_CN99XX' undeclared
>>>> (first use in this function)
>>>> 3017 |  case ACPI_IORT_SMMU_V3_CAVIUM_CN99XX:
>>>>        |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>> smmu-v3.c:3017:7: note: each undeclared identifier is reported only
>>>> once for each function it appears in
>>>> smmu-v3.c:3020:7: error: 'ACPI_IORT_SMMU_V3_HISILICON_HI161X'
>>>> undeclared (first use in this function)
>>>> 3020 |  case ACPI_IORT_SMMU_V3_HISILICON_HI161X:
>>>>        |       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>>    INIT_O  efi-dom0.init.o
>>>> smmu-v3.c: In function 'arm_smmu_device_acpi_probe':
>>>> smmu-v3.c:3035:36: error: implicit declaration of function
>>>> 'dev_get_platdata' [-Werror=implicit-function-declaration]
>>>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>>>        |                                    ^~~~~~~~~~~~~~~~
>>>> smmu-v3.c:3035:36: error: nested extern declaration of
>>>> 'dev_get_platdata' [-Werror=nested-externs]
>>>> smmu-v3.c:3035:10: error: cast to pointer from integer of different
>>>> size [-Werror=int-to-pointer-cast]
>>>> 3035 |  node = *(struct acpi_iort_node **)dev_get_platdata(dev);
>>>>        |          ^
>>>> ld    -EL  -r -o built_in.o memcpy.o memcmp.o memmove.o memset.o
>>>> memchr.o clear_page.o bitops.o find_next_bit.o strchr.o strcmp.o
>>>> strlen.o strncmp.o strnlen.o strrchr.o
>>>> cc1: all warnings being treated as errors
>>>>
>>>> It looks like the SMMUv3 driver has not been built tested with
>>>> CONFIG_ACPI=y.  Bertrand, Rahul, can you take a look?
>>>
>>> Sorry my mistake I didn’t enable the ARM_SMMU_V3 with ACPI and tested the compilation.
>>>
>>> There are two option to fix this:
>>>
>>> 1. #undef-ining the CONFIG_ACPI in the smmu-v3.c file and fix the compilation. In this case once we have ACPI IORT table code ported to XEN smmu-v3 code can be used at that time.
>>> 2.  Remove the arm_smmu_device_acpi_probe() from the smmu-v3.c now and once we have  ACPI IORT table code ported to XEN we can add it back.
>>
>> How about the following?
>>
>> 3. Only allow the user to build the SMMUv3 when !CONFIG_ACPI.
> 
> And then perhaps as "depends on !ACPI || BROKEN" in Kconfig?

I am fine with that.

Cheers,

-- 
Julien Grall