On Fri Aug 7, 2026 at 8:38 AM CST, Ashok Raj wrote:
> External email: Use caution opening links or attachments
>
>
> On Tue, Aug 04, 2026 at 08:23:15PM +0800, Kai-Heng Feng wrote:
>> Split the Grace CPER processing into a separate decode step and a
>> print step so the parser can be exercised by KUnit without a live
>> ACPI device. Introduce ghes-nvidia.h to hold shared types that the
>> Vera decoder added in the next commit will also reference.
>>
>> Parse the Grace wire payload with __packed structs and
>> get_unaligned_le*() for multi-byte fields, matching the Vera
>> decoder style. Reject nvidia_ghes_grace_reg_pair() calls with a
>> non-zero register count and a NULL grace_regs pointer, and
>> rate-limit Grace decoder diagnostics.
>>
>> Signed-off-by: Kai-Heng Feng <[email protected]>
>
> Hi Kai,
>
>> ---
>> v3:
>> - Use __packed + get_unaligned_le*() for Grace multi-byte fields,
>> including register pairs (Shuai Xue).
>> - Keep __counted_by / struct_size; reject NULL grace_regs when
>> number_regs != 0; document the helper contract.
>> - Rate-limit Grace decoder and notify error messages.
>> v2:
>> - No change.
>> ---
>> MAINTAINERS | 2 +-
>> drivers/acpi/apei/ghes-nvidia.c | 160 ++++++++++++++++++++++----------
>> drivers/acpi/apei/ghes-nvidia.h | 50 ++++++++++
>> 3 files changed, 160 insertions(+), 52 deletions(-)
>> create mode 100644 drivers/acpi/apei/ghes-nvidia.h
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 5114e6db7307..aa9c2ae58ac5 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -19348,7 +19348,7 @@ NVIDIA GHES VENDOR CPER RECORD HANDLER
>> M: Kai-Heng Feng <[email protected]>
>> L: [email protected]
>> S: Maintained
>> -F: drivers/acpi/apei/ghes-nvidia.c
>> +F: drivers/acpi/apei/ghes-nvidia*
>>
>> NVIDIA VRS RTC DRIVER
>> M: Shubhi Garg <[email protected]>
>> diff --git a/drivers/acpi/apei/ghes-nvidia.c
>> b/drivers/acpi/apei/ghes-nvidia.c
>> index 597275d81de8..e7cc912344fd 100644
>> --- a/drivers/acpi/apei/ghes-nvidia.c
>> +++ b/drivers/acpi/apei/ghes-nvidia.c
>> @@ -9,13 +9,18 @@
>> #include <linux/module.h>
>> #include <linux/platform_device.h>
>> #include <linux/types.h>
>> +#include <linux/unaligned.h>
>> #include <linux/uuid.h>
>> #include <acpi/ghes.h>
>
> One naive question, do these CPER decoding for vendor specific pieces
> belong in the kernel? Or you would manage them from user space?
I think so, since pcie-hisi-controller is already doing that. And I
think helpers like devm_ghes_register_vendor_record_notifier() is
designed for this kind of work.
Kai-Heng
>
>
>>
>> -static const guid_t nvidia_sec_guid =
>> +#include <kunit/visibility.h>
>> +#include "ghes-nvidia.h"
>> +
>> +static const guid_t nvidia_grace_sec_guid =
>> GUID_INIT(0x6d5244f2, 0x2712, 0x11ec,
>> 0xbe, 0xa7, 0xcb, 0x3f, 0xdb, 0x95, 0xc7, 0x86);
>>
>> +/* Grace CPER section wire layout (header without flexible register array).
>> */
>> struct cper_sec_nvidia {
>> char signature[16];
>> __le16 error_type;
>> @@ -25,84 +30,137 @@ struct cper_sec_nvidia {
>> u8 number_regs;
>> u8 reserved;
>> __le64 instance_base;
>> - struct {
>> - __le64 addr;
>> - __le64 val;
>> - } regs[] __counted_by(number_regs);
>> -};
>> + struct nvidia_ghes_grace_reg regs[] __counted_by(number_regs);
>> +} __packed;
>> +
>
> [snip]
>
> Cheers,
> Ashok
