Re: [PATCH 2/2] platform/x86/amd/hsmp: Add support for variable-sized metrics tables

["Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>]

[-- Attachment #1: Type: text/plain, Size: 2412 bytes --]

On Wed, 1 Apr 2026, Ramalingam, Muthusamy wrote:

> Hi Ilpo,
> 
> Thank you for the suggestions. I will update version v2 shortly to address the
> comments listed below.
> 
> On 31-03-2026 02:13 pm, Ilpo Järvinen wrote:
> > On Thu, 5 Mar 2026, Muthusamy Ramalingam wrote:
> > 
> > > Add support for the new metrics table format introduced in AMD Family 1Ah
> > > Model 50h-5Fh processors with HSMP protocol version 7.
> > > 
> > > Use CPU family/model and protocol versions to map respective
> > > variable-sized
> > > metric table configurations.
> > > The exported hsmp_metric_tbl_read() function provides offset support for
> > > variable-sized tables.
> > 
> > This patch should be split into multiple changes.
> Sure.
> > 
> > > Co-developed-by: Muralidhara M K <muralimk@amd.com>
> > > Signed-off-by: Muralidhara M K <muralimk@amd.com>
> > > Signed-off-by: Muthusamy Ramalingam <muthusamy.ramalingam@amd.com>
> > > ---
> > >   arch/x86/include/uapi/asm/amd_hsmp.h | 86 ++++++++++++++++++++++++++++
> > >   drivers/platform/x86/amd/hsmp/acpi.c |  9 +--
> > >   drivers/platform/x86/amd/hsmp/hsmp.c | 83 ++++++++++++++++++++++-----
> > >   drivers/platform/x86/amd/hsmp/hsmp.h |  3 +-
> > >   drivers/platform/x86/amd/hsmp/plat.c |  3 +-
> > >   5 files changed, 164 insertions(+), 20 deletions(-)
> > > 
> > > --- a/drivers/platform/x86/amd/hsmp/acpi.c
> > > +++ b/drivers/platform/x86/amd/hsmp/acpi.c

> > > @@ -403,8 +424,44 @@ int hsmp_get_tbl_dram_base(u16 sock_ind)
> > >   		dev_err(sock->dev, "Invalid DRAM address for metric table\n");
> > >   		return -ENOMEM;
> > >   	}
> > > -	sock->metric_tbl_addr = devm_ioremap(sock->dev, dram_addr,
> > > -					     sizeof(struct
> > > hsmp_metric_table));
> > > +
> > > +	/* Get metric table version */
> > > +	msg_tbl_ver.sock_ind    = sock_ind;
> > > +	msg_tbl_ver.response_sz =
> > > hsmp_msg_desc_table[HSMP_GET_METRIC_TABLE_VER].response_sz;
> > > +	msg_tbl_ver.msg_id      = HSMP_GET_METRIC_TABLE_VER;
> > > +
> > > +	ret = hsmp_send_message(&msg_tbl_ver);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	table_ver = msg_tbl_ver.args[0];
> > > +
> > > +	hsmp_pdev.hsmp_table_size = 0;
> > 
> > Is this necessary?
>
> Yes, We want this for unsupported family check.

Isn't this initialized to its default value which is 0? So why it needs to 
be explicitly done again?

-- 
 i.