mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Erwan Velu <erwanaliasr1@gmail.com>
To: "Naveen Krishna Chatradhi" <naveenkrishna.chatradhi@amd.com>,
	"Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>,
	"Hans de Goede" <hansg@kernel.org>
Cc: Carlos Bilbao <carlos.bilbao@kernel.org>,
	Muralidhara M K <muralidhara.mk@amd.com>,
	Suma Hegde <suma.hegde@amd.com>,
	Muthusamy Ramalingam <muthusamy.ramalingam@amd.com>,
	platform-driver-x86@vger.kernel.org,
	linux-kernel@vger.kernel.org, Erwan Velu <e.velu@criteo.com>
Subject: [PATCH v2 3/5] platform/x86/amd/hsmp: Expose socket attributes from the platform driver
Date: Mon, 16 Nov 2026 16:51:45 +0100	[thread overview]
Message-ID: <20261116155147.769630-4-erwanaliasr1@gmail.com> (raw)
In-Reply-To: <20261116155147.769630-1-erwanaliasr1@gmail.com>

From: Erwan Velu <e.velu@criteo.com>

The platform driver creates bare socketN attribute groups on the
amd_hsmp platform device, which only hold metrics_bin. On systems
without the HSMP ACPI object, attributes such as
pwr_current_active_freq_limit are therefore missing, although the
driver can send the same HSMP messages as the ACPI one.

Register the amd_hsmp class socketN devices from the platform driver,
as children of the amd_hsmp platform device, and drop the socketN
attribute groups they replace. The socket devices carry
hsmp_sock_groups, metrics_bin included, so all the per-socket
attributes are now available under /sys/class/amd_hsmp/socketN/.

As class devices, they are placed by the driver core in an amd_hsmp
glue directory below their parent, i.e. in
/sys/devices/platform/amd_hsmp/amd_hsmp/socketN/. To keep the existing
/sys/devices/platform/amd_hsmp/socketN/metrics_bin path working, add a
socketN symlink from the platform device to each socket device.

As the socket devices and their symlinks are now the only way
metrics_bin is exposed, failing to create one fails the probe. They are
torn down by a devres action added after hsmp_pltdrv_release(), so they
are gone before the socket array they reference is released.

Systems running HSMP protocol version 7 or later with the platform
driver (Family 1Ah model 0x00-0x0F without the HSMP ACPI object) now
see metrics_bin, as with the ACPI driver: reads return -EOPNOTSUPP and
point userspace to HSMP_IOCTL_GET_TELEMETRY_DATA, as the version 7
table does not fit in a sysfs binary attribute. It used to be hidden
there.

On a single socket Zen4 (Family 19h) server probed by the platform
driver, this gives:

  $ ls -l /sys/devices/platform/amd_hsmp/
  drwxr-xr-x amd_hsmp
  ...
  lrwxrwxrwx socket0 -> amd_hsmp/socket0
  $ ls -l /sys/class/amd_hsmp/
  socket0 -> ../../devices/platform/amd_hsmp/amd_hsmp/socket0
  $ cat /sys/class/amd_hsmp/socket0/pwr_current_active_freq_limit
  3100

This system runs HSMP protocol version 5, so metrics_bin is not
visible, as before this change.

Assisted-by: claude-opus-5-5
Signed-off-by: Erwan Velu <e.velu@criteo.com>
---
 Documentation/arch/x86/amd_hsmp.rst  |   8 +-
 drivers/platform/x86/amd/hsmp/plat.c | 155 +++++++++++----------------
 2 files changed, 67 insertions(+), 96 deletions(-)

diff --git a/Documentation/arch/x86/amd_hsmp.rst b/Documentation/arch/x86/amd_hsmp.rst
index f145ea12af99..c4459a9dddc6 100644
--- a/Documentation/arch/x86/amd_hsmp.rst
+++ b/Documentation/arch/x86/amd_hsmp.rst
@@ -75,14 +75,18 @@ e.g. /sys/devices/platform/AMDI0097:00/amd_hsmp/socket0/. For backward
 compatibility, the files are also available directly under
 /sys/devices/platform/AMDI0097:0X/.
 
+With the plat driver, the socket devices are children of the amd_hsmp
+platform device, i.e. /sys/devices/platform/amd_hsmp/amd_hsmp/socket%d/.
+For backward compatibility, /sys/devices/platform/amd_hsmp/socket%d is a
+symlink to it.
+
 1. Metrics table binary sysfs
 
 AMD MI300A MCM provides GET_METRICS_TABLE message to retrieve
 most of the system management information from SMU in one go.
 
 The metrics table is made available as hexadecimal sysfs binary file
-under per socket sysfs directory created at
-/sys/devices/platform/amd_hsmp/socket%d/metrics_bin
+/sys/class/amd_hsmp/socket%d/metrics_bin
 
 Note: lseek() is not supported as entire metrics table is read.
 
diff --git a/drivers/platform/x86/amd/hsmp/plat.c b/drivers/platform/x86/amd/hsmp/plat.c
index c3dc7c9f45ac..edcfe82aab2a 100644
--- a/drivers/platform/x86/amd/hsmp/plat.c
+++ b/drivers/platform/x86/amd/hsmp/plat.c
@@ -12,7 +12,6 @@
 #include <asm/amd/hsmp.h>
 
 #include <linux/acpi.h>
-#include <linux/build_bug.h>
 #include <linux/cleanup.h>
 #include <linux/device.h>
 #include <linux/dev_printk.h>
@@ -48,98 +47,6 @@ static int amd_hsmp_pci_rdwr(struct hsmp_socket *sock, u32 offset,
 	return amd_smn_hsmp_rdwr(sock->sock_ind, sock->mbinfo.base_addr + offset, value, write);
 }
 
-static ssize_t hsmp_metric_tbl_plat_read(struct file *filp, struct kobject *kobj,
-					 const struct bin_attribute *bin_attr, char *buf,
-					 loff_t off, size_t count)
-{
-	struct hsmp_socket *sock;
-	u16 sock_ind;
-
-	sock_ind = (uintptr_t)bin_attr->private;
-	if (sock_ind >= hsmp_pdev->num_sockets)
-		return -EINVAL;
-
-	sock = &hsmp_pdev->sock[sock_ind];
-
-	return hsmp_metric_tbl_read(sock, buf, count);
-}
-
-static umode_t hsmp_is_sock_attr_visible(struct kobject *kobj,
-					 const struct bin_attribute *battr, int id)
-{
-	u16 sock_ind;
-
-	sock_ind = (uintptr_t)battr->private;
-
-	if (id == 0 && sock_ind >= hsmp_pdev->num_sockets)
-		return SYSFS_GROUP_INVISIBLE;
-
-	if (hsmp_pdev->proto_ver == HSMP_PROTO_VER6)
-		return battr->attr.mode;
-
-	return 0;
-}
-
-/*
- * AMD supports maximum of 8 sockets in a system.
- * Static array of 8 + 1(for NULL) elements is created below
- * to create sysfs groups for sockets.
- * is_bin_visible function is used to show / hide the necessary groups.
- *
- * Validate the maximum number against MAX_AMD_NUM_NODES. If this changes,
- * then the attributes and groups below must be adjusted.
- */
-static_assert(MAX_AMD_NUM_NODES == 8);
-
-#define HSMP_BIN_ATTR(index, _list)					\
-static const struct bin_attribute attr##index = {			\
-	.attr = { .name = HSMP_METRICS_TABLE_NAME, .mode = 0444},	\
-	.private = (void *)index,					\
-	.read = hsmp_metric_tbl_plat_read,				\
-	.size = sizeof(struct hsmp_metric_table),			\
-};									\
-static const struct bin_attribute _list[] = {				\
-	&attr##index,							\
-	NULL								\
-}
-
-HSMP_BIN_ATTR(0, *sock0_attr_list);
-HSMP_BIN_ATTR(1, *sock1_attr_list);
-HSMP_BIN_ATTR(2, *sock2_attr_list);
-HSMP_BIN_ATTR(3, *sock3_attr_list);
-HSMP_BIN_ATTR(4, *sock4_attr_list);
-HSMP_BIN_ATTR(5, *sock5_attr_list);
-HSMP_BIN_ATTR(6, *sock6_attr_list);
-HSMP_BIN_ATTR(7, *sock7_attr_list);
-
-#define HSMP_BIN_ATTR_GRP(index, _list, _name)			\
-static const struct attribute_group sock##index##_attr_grp = {	\
-	.bin_attrs = _list,					\
-	.is_bin_visible = hsmp_is_sock_attr_visible,		\
-	.name = #_name,						\
-}
-
-HSMP_BIN_ATTR_GRP(0, sock0_attr_list, socket0);
-HSMP_BIN_ATTR_GRP(1, sock1_attr_list, socket1);
-HSMP_BIN_ATTR_GRP(2, sock2_attr_list, socket2);
-HSMP_BIN_ATTR_GRP(3, sock3_attr_list, socket3);
-HSMP_BIN_ATTR_GRP(4, sock4_attr_list, socket4);
-HSMP_BIN_ATTR_GRP(5, sock5_attr_list, socket5);
-HSMP_BIN_ATTR_GRP(6, sock6_attr_list, socket6);
-HSMP_BIN_ATTR_GRP(7, sock7_attr_list, socket7);
-
-static const struct attribute_group *hsmp_groups[] = {
-	&sock0_attr_grp,
-	&sock1_attr_grp,
-	&sock2_attr_grp,
-	&sock3_attr_grp,
-	&sock4_attr_grp,
-	&sock5_attr_grp,
-	&sock6_attr_grp,
-	&sock7_attr_grp,
-	NULL
-};
-
 static inline bool is_f1a_m0h(void)
 {
 	if (boot_cpu_data.x86 == 0x1A && boot_cpu_data.x86_model <= 0x0F)
@@ -222,9 +129,49 @@ static void hsmp_pltdrv_release(void *data)
 	hsmp_pdev->sock = NULL;
 }
 
+/*
+ * The amd_hsmp class devices live in an amd_hsmp glue directory below the
+ * platform device, i.e. /sys/devices/platform/amd_hsmp/amd_hsmp/socket%u.
+ * /sys/devices/platform/amd_hsmp/socket%u used to be a plain attribute group
+ * holding metrics_bin, so keep that path working with a symlink to the socket
+ * device.
+ */
+static int hsmp_pltdrv_sock_dev_register(struct device *dev, struct hsmp_socket *sock)
+{
+	int ret;
+
+	ret = hsmp_sock_dev_register(dev, sock);
+	if (ret)
+		return ret;
+
+	ret = sysfs_create_link(&dev->kobj, &sock->sysfs_dev->kobj,
+				dev_name(sock->sysfs_dev));
+	if (ret)
+		hsmp_sock_dev_unregister(sock);
+
+	return ret;
+}
+
+static void hsmp_pltdrv_sock_devs_release(void *data)
+{
+	struct device *dev = data;
+	struct hsmp_socket *sock;
+	u16 i;
+
+	for (i = 0; i < hsmp_pdev->num_sockets; i++) {
+		sock = &hsmp_pdev->sock[i];
+		if (!sock->sysfs_dev)
+			continue;
+
+		sysfs_remove_link(&dev->kobj, dev_name(sock->sysfs_dev));
+		hsmp_sock_dev_unregister(sock);
+	}
+}
+
 static int hsmp_pltdrv_probe(struct platform_device *pdev)
 {
 	int ret;
+	u16 i;
 
 	hsmp_pdev->sock = devm_kzalloc_objs(&pdev->dev, *hsmp_pdev->sock,
 					    hsmp_pdev->num_sockets);
@@ -252,6 +199,27 @@ static int hsmp_pltdrv_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	/*
+	 * Added after hsmp_pltdrv_release() so that devres removes the socket
+	 * devices before the socket array they reference is released.
+	 */
+	ret = devm_add_action_or_reset(&pdev->dev, hsmp_pltdrv_sock_devs_release, &pdev->dev);
+	if (ret)
+		return ret;
+
+	/*
+	 * The socket devices and their symlinks are the only way this driver
+	 * exposes metrics_bin, so failing to create one fails the probe.
+	 */
+	for (i = 0; i < hsmp_pdev->num_sockets; i++) {
+		ret = hsmp_pltdrv_sock_dev_register(&pdev->dev, &hsmp_pdev->sock[i]);
+		if (ret) {
+			dev_err(&pdev->dev, "Failed to register HSMP socket%u device: %d\n",
+				i, ret);
+			return ret;
+		}
+	}
+
 	ret = hsmp_misc_register(&pdev->dev);
 	if (ret) {
 		dev_err(&pdev->dev, "Failed to register misc device\n");
@@ -272,7 +240,6 @@ static struct platform_driver amd_hsmp_driver = {
 	.remove		= hsmp_pltdrv_remove,
 	.driver		= {
 		.name	= DRIVER_NAME,
-		.dev_groups = hsmp_groups,
 	},
 };
 
-- 
2.53.0


  parent reply	other threads:[~2026-11-16 15:52 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-16 15:51 [PATCH v2 0/5] platform/x86/amd/hsmp: Unify per-socket sysfs metrics Erwan Velu
2026-11-16 15:51 ` [PATCH v2 1/5] platform/x86/amd/hsmp: Move per-socket sysfs attributes to common code Erwan Velu
2026-11-16 15:51 ` [PATCH v2 2/5] platform/x86/amd/hsmp: Add amd_hsmp class with per-socket devices Erwan Velu
2026-11-16 15:51 ` Erwan Velu [this message]
2026-11-16 15:51 ` [PATCH v2 4/5] platform/x86/amd/hsmp: Add pwr_current_active_freq_limit_source_raw Erwan Velu
2026-11-16 15:51 ` [PATCH v2 5/5] Documentation/ABI: Add sysfs-class-amd_hsmp Erwan Velu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261116155147.769630-4-erwanaliasr1@gmail.com \
    --to=erwanaliasr1@gmail.com \
    --cc=carlos.bilbao@kernel.org \
    --cc=e.velu@criteo.com \
    --cc=hansg@kernel.org \
    --cc=ilpo.jarvinen@linux.intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=muralidhara.mk@amd.com \
    --cc=muthusamy.ramalingam@amd.com \
    --cc=naveenkrishna.chatradhi@amd.com \
    --cc=platform-driver-x86@vger.kernel.org \
    --cc=suma.hegde@amd.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®