From: "Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>
To: Erwan Velu <erwanaliasr1@gmail.com>
Cc: platform-driver-x86@vger.kernel.org,
Naveen Krishna Chatradhi <naveenkrishna.chatradhi@amd.com>,
Carlos Bilbao <carlos.bilbao@kernel.org>,
Hans de Goede <hansg@kernel.org>,
LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 2/4] platform/x86/amd/hsmp: Add /sys/devices/amd_hsmp socket devices
Date: Fri, 16 Oct 2026 14:02:42 +0300 (EEST) [thread overview]
Message-ID: <0ce3cf8c-2e00-835c-caed-c0b099a70c3b@linux.intel.com> (raw)
In-Reply-To: <20261007125908.321186-3-erwanaliasr1@gmail.com>
On Wed, 7 Oct 2026, Erwan Velu wrote:
> From: Erwan Velu <e.velu@criteo.com>
>
> The per-socket attributes live under /sys/devices/platform/AMDI0097:0X
> with the ACPI driver, while the platform driver uses
> /sys/devices/platform/amd_hsmp/socketN. Userspace has to know which
> driver probed to find them.
>
> Add a driver-independent /sys/devices/amd_hsmp/socketN device per
> socket carrying hsmp_sock_groups, and register it from the ACPI driver.
> The AMDI0097:0X attributes are kept, as they are documented ABI.
>
> The socket device is unregistered before remove takes hsmp_sock_rwsem:
> device_unregister() waits for in-flight sysfs reads, which take the
> rwsem for read.
>
> 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/acpi.c | 6 ++
> drivers/platform/x86/amd/hsmp/hsmp.h | 5 ++
> drivers/platform/x86/amd/hsmp/sysfs.c | 112 +++++++++++++++++++++++++-
> 4 files changed, 129 insertions(+), 2 deletions(-)
>
> diff --git a/Documentation/arch/x86/amd_hsmp.rst b/Documentation/arch/x86/amd_hsmp.rst
> index fa1fc240e212..aba4a009bd8d 100644
> --- a/Documentation/arch/x86/amd_hsmp.rst
> +++ b/Documentation/arch/x86/amd_hsmp.rst
> @@ -57,6 +57,12 @@ In-kernel integration:
> HSMP sysfs interface
> ====================
>
> +The ACPI driver exposes its per socket sysfs files in a per socket
> +directory created at /sys/devices/amd_hsmp/socket%d/.
> +
> +For backward compatibility, the files are also available under
> +/sys/devices/platform/AMDI0097:0X/.
> +
> 1. Metrics table binary sysfs
>
> AMD MI300A MCM provides GET_METRICS_TABLE message to retrieve
> @@ -80,7 +86,7 @@ The same is defined in the amd_hsmp.h header.
>
> 2. HSMP telemetry sysfs files
>
> -Following sysfs files are available at /sys/devices/platform/AMDI0097:0X/.
> +Following sysfs files are available at /sys/devices/amd_hsmp/socket%d/.
>
> * c0_residency_input: Percentage of cores in C0 state.
> * prochot_status: Reports 1 if the processor is at thermal threshold value,
> diff --git a/drivers/platform/x86/amd/hsmp/acpi.c b/drivers/platform/x86/amd/hsmp/acpi.c
> index d76979baffa8..468d4b1515db 100644
> --- a/drivers/platform/x86/amd/hsmp/acpi.c
> +++ b/drivers/platform/x86/amd/hsmp/acpi.c
> @@ -441,6 +441,10 @@ static int hsmp_acpi_probe(struct platform_device *pdev)
> dev_dbg(&pdev->dev, "AMD HSMP ACPI misc device registered\n");
> }
>
> + ret = hsmp_sock_dev_register(dev_get_drvdata(&pdev->dev));
> + if (ret)
> + dev_warn(&pdev->dev, "Failed to register HSMP socket device: %d\n", ret);
> +
> return 0;
> }
>
> @@ -448,6 +452,8 @@ static void hsmp_acpi_remove(struct platform_device *pdev)
> {
> struct hsmp_socket *sock = dev_get_drvdata(&pdev->dev);
>
> + hsmp_sock_dev_unregister(sock);
> +
> /*
> * Serialize the kref_put() and any release it triggers against a
> * concurrent probe, and drain the data plane for the whole
> diff --git a/drivers/platform/x86/amd/hsmp/hsmp.h b/drivers/platform/x86/amd/hsmp/hsmp.h
> index 692721e2e6d3..f638edd7768b 100644
> --- a/drivers/platform/x86/amd/hsmp/hsmp.h
> +++ b/drivers/platform/x86/amd/hsmp/hsmp.h
> @@ -28,6 +28,7 @@
>
> #define HSMP_CDEV_NAME "hsmp_cdev"
> #define HSMP_DEVNODE_NAME "hsmp"
> +#define HSMP_ROOT_DEV_NAME "amd_hsmp"
> #define ACPI_HSMP_DEVICE_HID "AMDI0097"
>
> #define DRIVER_VERSION "2.6"
> @@ -47,6 +48,7 @@ struct hsmp_socket {
> /* Size of the region mapped at @metric_tbl_addr, as reported by SMU */
> size_t metric_tbl_size;
> void __iomem *virt_base_addr;
> + struct device *sysfs_dev;
> struct semaphore hsmp_sem;
> /* Serializes HSMP_GET_METRIC_TABLE fill-and-copy for this socket */
> struct mutex metric_read_lock;
> @@ -80,6 +82,9 @@ int hsmp_create_sensor(struct device *dev, u16 sock_ind);
> static inline int hsmp_create_sensor(struct device *dev, u16 sock_ind) { return 0; }
> #endif
> int hsmp_msg_get_nargs(u16 sock_ind, u32 msg_id, u32 *data, u8 num_args);
> +int hsmp_sock_dev_register(struct hsmp_socket *sock);
> +void hsmp_sock_dev_unregister(struct hsmp_socket *sock);
> +
> extern const struct attribute_group *hsmp_sock_groups[];
>
> /*
> diff --git a/drivers/platform/x86/amd/hsmp/sysfs.c b/drivers/platform/x86/amd/hsmp/sysfs.c
> index b1db3b4d77b9..e9e91b640b2a 100644
> --- a/drivers/platform/x86/amd/hsmp/sysfs.c
> +++ b/drivers/platform/x86/amd/hsmp/sysfs.c
> @@ -4,7 +4,8 @@
> * Copyright (c) 2024, AMD.
> * All Rights Reserved.
> *
> - * This file provides the per-socket sysfs attributes of the HSMP interface.
> + * This file provides the per-socket sysfs attributes of the HSMP interface
> + * and the /sys/devices/amd_hsmp/socket%u devices carrying them.
> */
>
> #include <asm/amd/hsmp.h>
> @@ -12,7 +13,12 @@
> #include <linux/array_size.h>
> #include <linux/bitfield.h>
> #include <linux/bits.h>
> +#include <linux/cleanup.h>
> #include <linux/device.h>
> +#include <linux/err.h>
> +#include <linux/lockdep.h>
> +#include <linux/mutex.h>
> +#include <linux/slab.h>
> #include <linux/sysfs.h>
>
> #include "hsmp.h"
> @@ -336,3 +342,107 @@ const struct attribute_group *hsmp_sock_groups[] = {
> NULL
> };
> EXPORT_SYMBOL_NS_GPL(hsmp_sock_groups, "AMD_HSMP");
> +
> +/*
> + * /sys/devices/amd_hsmp is created when the first socket device is registered
> + * and removed with the last one. hsmp_root_lock serializes that against the
> + * per-socket register/unregister calls.
> + */
> +static DEFINE_MUTEX(hsmp_root_lock);
> +static struct device *hsmp_root_dev;
> +static unsigned int hsmp_root_users;
> +
> +static void hsmp_sock_dev_release(struct device *dev)
> +{
> + kfree(dev);
> +}
> +
> +/**
> + * hsmp_sock_dev_register() - Create /sys/devices/amd_hsmp/socket%u for @sock.
> + * @sock: fully probed HSMP socket.
> + *
> + * The device carries hsmp_sock_groups with @sock as its drvdata, so @sock must
> + * stay valid until hsmp_sock_dev_unregister() returns.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int hsmp_sock_dev_register(struct hsmp_socket *sock)
> +{
> + struct device *dev;
> + int ret;
> +
> + guard(mutex)(&hsmp_root_lock);
> +
> + if (!hsmp_root_users) {
> + hsmp_root_dev = root_device_register(HSMP_ROOT_DEV_NAME);
> + if (IS_ERR(hsmp_root_dev)) {
> + ret = PTR_ERR(hsmp_root_dev);
> + hsmp_root_dev = NULL;
> + return ret;
> + }
> + }
> +
> + /*
> + * Built by hand rather than with device_create_with_groups(), which
> + * rejects a NULL class: these devices belong to no class or bus.
> + */
> + dev = kzalloc_obj(*dev);
> + if (!dev) {
> + ret = -ENOMEM;
> + goto err_root;
> + }
> +
> + device_initialize(dev);
> + dev->parent = hsmp_root_dev;
> + dev->groups = hsmp_sock_groups;
> + dev->release = hsmp_sock_dev_release;
> + dev_set_drvdata(dev, sock);
> +
> + ret = dev_set_name(dev, "socket%u", sock->sock_ind);
> + if (!ret)
> + ret = device_add(dev);
> + if (ret) {
Tricky control logic. Preferrably handle errors first.
> + put_device(dev);
> + goto err_root;
> + }
> +
> + sock->sysfs_dev = dev;
> + hsmp_root_users++;
Why reinvent the wheel and not use kref.h? It got immediately complicated
on the rollback path when you tried to do this with your own approach.
> + return 0;
> +
> +err_root:
> + if (!hsmp_root_users) {
> + root_device_unregister(hsmp_root_dev);
> + hsmp_root_dev = NULL;
> + }
> + return ret;
> +}
> +EXPORT_SYMBOL_NS_GPL(hsmp_sock_dev_register, "AMD_HSMP");
> +
> +/**
> + * hsmp_sock_dev_unregister() - Remove /sys/devices/amd_hsmp/socket%u for @sock.
> + * @sock: HSMP socket; a no-op if its device was never registered.
> + *
> + * device_unregister() waits for in-flight show/read handlers, which take
> + * hsmp_sock_rwsem for read. The caller must therefore not hold it, or a
> + * concurrent sysfs read would deadlock the teardown.
> + */
> +void hsmp_sock_dev_unregister(struct hsmp_socket *sock)
> +{
> + lockdep_assert_not_held(&hsmp_sock_rwsem);
> +
> + guard(mutex)(&hsmp_root_lock);
> +
> + if (!sock->sysfs_dev)
> + return;
> +
> + device_unregister(sock->sysfs_dev);
> + sock->sysfs_dev = NULL;
> +
> + if (!--hsmp_root_users) {
> + root_device_unregister(hsmp_root_dev);
> + hsmp_root_dev = NULL;
> + }
> +}
> +EXPORT_SYMBOL_NS_GPL(hsmp_sock_dev_unregister, "AMD_HSMP");
>
--
i.
next prev parent reply other threads:[~2026-10-16 11:02 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-10-07 12:59 [PATCH 0/4] platform/x86/amd/hsmp: Unify per-socket sysfs metrics Erwan Velu
2026-10-07 12:59 ` [PATCH 1/4] platform/x86/amd/hsmp: Move per-socket sysfs attributes to common code Erwan Velu
2026-10-07 12:59 ` [PATCH 2/4] platform/x86/amd/hsmp: Add /sys/devices/amd_hsmp socket devices Erwan Velu
2026-10-16 11:02 ` Ilpo Järvinen [this message]
2026-10-07 12:59 ` [PATCH 3/4] platform/x86/amd/hsmp: Expose socket attributes from the platform driver Erwan Velu
2026-10-07 12:59 ` [PATCH 4/4] platform/x86/amd/hsmp: Add pwr_current_active_freq_limit_source_raw 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=0ce3cf8c-2e00-835c-caed-c0b099a70c3b@linux.intel.com \
--to=ilpo.jarvinen@linux.intel.com \
--cc=carlos.bilbao@kernel.org \
--cc=erwanaliasr1@gmail.com \
--cc=hansg@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=naveenkrishna.chatradhi@amd.com \
--cc=platform-driver-x86@vger.kernel.org \
/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®