Re: [PATCH v2.2] x86/amd_node, platform/x86/amd/hsmp: Have HSMP use SMN through AMD_NODE

[Suma Hegde <Suma.Hegde@amd.com>]

Hi Yazen,


On 12/17/2024 12:03 AM, Yazen Ghannam wrote:
> On Sat, Dec 14, 2024 at 11:05:07AM +0100, Borislav Petkov wrote:
>> On Fri, Dec 13, 2024 at 03:22:06PM +0000, Yazen Ghannam wrote:
>>> The HSMP interface is just an SMN interface with different offsets.
>>>
>>> Define an HSMP wrapper in the SMN code and have the HSMP platform driver
>>> use that rather than a local solution.
>>>
>>> Also, remove the "root" member from AMD_NB, since there are no more
>>> users of it.
>>>
>>> Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
>>> ---
>>>
>>> Notes:
>>>      Link:
>>>      https://lore.kernel.org/20241212172711.1944927-1-yazen.ghannam@amd.com
>>>      
>>>      v2.1-v2.2:
>>>      * Include <linux/build_bug.h> for static_assert()
>>>      
>>>      v2->v2.1:
>>>      * Include static_assert() and comment for sysfs attributes.
>>>      
>>>      v1->v2:
>>>      * Rebase on recent HSMP rework.
>>>
>>>   arch/x86/include/asm/amd_nb.h         |  1 -
>>>   arch/x86/include/asm/amd_node.h       |  3 +++
>>>   arch/x86/kernel/amd_nb.c              |  1 -
>>>   arch/x86/kernel/amd_node.c            |  9 +++++++
>>>   drivers/platform/x86/amd/hsmp/Kconfig |  2 +-
>>>   drivers/platform/x86/amd/hsmp/acpi.c  |  7 +++---
>>>   drivers/platform/x86/amd/hsmp/hsmp.c  |  1 -
>>>   drivers/platform/x86/amd/hsmp/hsmp.h  |  3 ---
>>>   drivers/platform/x86/amd/hsmp/plat.c  | 36 +++++++++------------------
>>>   9 files changed, 29 insertions(+), 34 deletions(-)
>> ld: drivers/platform/x86/amd/hsmp/plat.o: in function `amd_hsmp_pci_rdwr':
>> /home/amd/kernel/linux/drivers/platform/x86/amd/hsmp/plat.c:44: undefined reference to `amd_smn_hsmp_rdwr'
>> make[2]: *** [scripts/Makefile.vmlinux:77: vmlinux] Error 1
>> make[1]: *** [/home/amd/bpetkov/kernel/linux/Makefile:1225: vmlinux] Error 2
>> make[1]: *** Waiting for unfinished jobs....
>> make: *** [Makefile:251: __sub-make] Error 2
>>
>> Config attached.
>>
> I think the fix is this:
>
> diff --git a/drivers/platform/x86/amd/hsmp/Kconfig b/drivers/platform/x86/amd/hsmp/Kconfig
> index d6f7a62d55b5..fc90ef11a8ad 100644
> --- a/drivers/platform/x86/amd/hsmp/Kconfig
> +++ b/drivers/platform/x86/amd/hsmp/Kconfig
> @@ -7,7 +7,6 @@ config AMD_HSMP
>   	tristate
>   
>   menu "AMD HSMP Driver"
> -	depends on AMD_NODE || COMPILE_TEST
>   
>   config AMD_HSMP_ACPI
>   	tristate "AMD HSMP ACPI device driver"
> @@ -29,6 +28,7 @@ config AMD_HSMP_ACPI
>   
>   config AMD_HSMP_PLAT
>   	tristate "AMD HSMP platform device driver"
> +	depends on AMD_NODE
>   	select AMD_HSMP
>   	help
>   	  Host System Management Port (HSMP) interface is a mailbox interface
>
> With all the recent rework, only AMD_HSMP_PLAT has a hard dependency on
> AMD_NODE due to needing the SMN register access interface.
>
> Various HSMP files still pull in <asm/amd_node.h>, but they just need
> defines and an inline function.
>
> It seems COMPILE_TEST is not needed, since this change allows
> AMD_HSMP_ACPI to build without AMD_NODE which seems like the intent.

COMPILE_TEST was added to provide build coverage to HSMP based on Ilpo's 
suggestion.

We can probably add #ifdef CONFIG_AMD_NODE and #else as below to fix 
this instead of removing COMPILE_TEST?

#ifdef CONFIG_AMD_NODE
/* Should only be used by the HSMP driver. */
int __must_check amd_smn_hsmp_rdwr(u16 node, u32 address, u32 *value, 
bool write);
#else
static int amd_smn_hsmp_rdwr(u16 node, u32 address, u32 *value, bool write);
{
         return 0;
}

#endif

May be Ilpo can comment more.

>
> Thoughts from Platform/HSMP folks? I can send another revision if this
> is okay.
>
> Thanks,
> Yazen

Thanks and Regards,

Suma