platform/x86/intel/pmt: Add PMT Discovery driver

This patch introduces a new driver to enumerate and expose Intel Platform
Monitoring Technology (PMT) capabilities via a simple discovery mechanism.
The PMT Discovery driver parses hardware-provided discovery tables from
Intel Out of Band Management Services Modules (OOBMSM) and extracts feature
information for various providers (such as TPMI, Telemetry, Crash Log,
etc). This unified interface simplifies the process of determining which
manageability and telemetry features are supported by a given platform.

This new feature is described in the Intel Platform Monitoring Technology
3.0 specification, section 6.6 Capability.

Key changes and additions:

New file drivers/platform/x86/intel/pmt/discovery.c:
  – Implements the discovery logic to map the discovery resource, read
    the feature discovery table, and validate feature parameters.
New file drivers/platform/x86/intel/pmt/features.c:
  – Defines feature names, layouts, and associated capability masks.
  – Provides a mapping between raw hardware attributes and sysfs
    representations for easier integration with user-space tools.
New header include/linux/intel_pmt_features.h:
  – Declares constants, masks, and feature identifiers used across the
    PMT framework.
Sysfs integration:
  – Feature attributes are exposed under /sys/class/intel_pmt.
  – Each device is represented by a subfolder within the intel_pmt class,
    named using its DBDF (Domain:Bus:Device.Function), e.g.:
        features-0000:00:03.1
  – Example directory layout for a device:

    /sys/class/intel_pmt/features-0000:00:03.1/
    ├── accelerator_telemetry
    ├── crash_log
    ├── per_core_environment_telemetry
    ├── per_core_performance_telemetry
    ├── per_rmid_energy_telemetry
    ├── per_rmid_perf_telemetry
    ├── tpmi_control
    ├── tracing
    └── uncore_telemetry

By exposing PMT feature details through sysfs and integrating with the
existing PMT class, this driver paves the way for more streamlined
integration of PMT-based manageability and telemetry tools.

Link: https://www.intel.com/content/www/us/en/content-details/710389/intel-platform-monitoring-technology-intel-pmt-external-specification.html
Signed-off-by: David E. Box <david.e.box@linux.intel.com>
Link: https://lore.kernel.org/r/20250703022832.1302928-9-david.e.box@linux.intel.com
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>

1 files changed, 602 insertions, 0 deletions

diff --git a/drivers/platform/x86/intel/pmt/discovery.c b/drivers/platform/x86/intel/pmt/discovery.c
new file mode 100644
index 000000000000..4b4fa3137ad2
--- /dev/null
+++ b/drivers/platform/x86/intel/pmt/discovery.c
@@ -0,0 +1,602 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Intel Platform Monitory Technology Discovery driver
+ *
+ * Copyright (c) 2025, Intel Corporation.
+ * All Rights Reserved.
+ */
+
+#include <linux/auxiliary_bus.h>
+#include <linux/bitfield.h>
+#include <linux/bits.h>
+#include <linux/bug.h>
+#include <linux/cleanup.h>
+#include <linux/container_of.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+#include <linux/kdev_t.h>
+#include <linux/kobject.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/overflow.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/string_choices.h>
+#include <linux/sysfs.h>
+#include <linux/types.h>
+#include <linux/uaccess.h>
+
+#include <linux/intel_pmt_features.h>
+#include <linux/intel_vsec.h>
+
+#include "class.h"
+
+#define MAX_FEATURE_VERSION	0
+#define DT_TBIR			GENMASK(2, 0)
+#define FEAT_ATTR_SIZE(x)	((x) * sizeof(u32))
+#define PMT_GUID_SIZE(x)	((x) * sizeof(u32))
+#define PMT_ACCESS_TYPE_RSVD	0xF
+#define SKIP_FEATURE		1
+
+struct feature_discovery_table {
+	u32	access_type:4;
+	u32	version:8;
+	u32	size:16;
+	u32	reserved:4;
+	u32	id;
+	u32	offset;
+	u32	reserved2;
+};
+
+/* Common feature table header */
+struct feature_header {
+	u32	attr_size:8;
+	u32	num_guids:8;
+	u32	reserved:16;
+};
+
+/* Feature attribute fields */
+struct caps {
+	u32		caps;
+};
+
+struct command {
+	u32		max_stream_size:16;
+	u32		max_command_size:16;
+};
+
+struct watcher {
+	u32		reserved:21;
+	u32		period:11;
+	struct command	command;
+};
+
+struct rmid {
+	u32		num_rmids:16;	/* Number of Resource Monitoring IDs */
+	u32		reserved:16;
+	struct watcher	watcher;
+};
+
+struct feature_table {
+	struct feature_header	header;
+	struct caps		caps;
+	union {
+		struct command command;
+		struct watcher watcher;
+		struct rmid rmid;
+	};
+	u32			*guids;
+};
+
+/* For backreference in struct feature */
+struct pmt_features_priv;
+
+struct feature {
+	struct feature_table		table;
+	struct kobject			kobj;
+	struct pmt_features_priv	*priv;
+	struct list_head		list;
+	const struct attribute_group	*attr_group;
+	enum pmt_feature_id		id;
+};
+
+struct pmt_features_priv {
+	struct device		*parent;
+	struct device		*dev;
+	int			count;
+	u32			mask;
+	struct feature		feature[];
+};
+
+static LIST_HEAD(pmt_feature_list);
+static DEFINE_MUTEX(feature_list_lock);
+
+#define to_pmt_feature(x) container_of(x, struct feature, kobj)
+static void pmt_feature_release(struct kobject *kobj)
+{
+}
+
+static ssize_t caps_show(struct kobject *kobj, struct kobj_attribute *attr,
+			 char *buf)
+{
+	struct feature *feature = to_pmt_feature(kobj);
+	struct pmt_cap **pmt_caps;
+	u32 caps = feature->table.caps.caps;
+	ssize_t ret = 0;
+
+	switch (feature->id) {
+	case FEATURE_PER_CORE_PERF_TELEM:
+		pmt_caps = pmt_caps_pcpt;
+		break;
+	case FEATURE_PER_CORE_ENV_TELEM:
+		pmt_caps = pmt_caps_pcet;
+		break;
+	case FEATURE_PER_RMID_PERF_TELEM:
+		pmt_caps = pmt_caps_rmid_perf;
+		break;
+	case FEATURE_ACCEL_TELEM:
+		pmt_caps = pmt_caps_accel;
+		break;
+	case FEATURE_UNCORE_TELEM:
+		pmt_caps = pmt_caps_uncore;
+		break;
+	case FEATURE_CRASH_LOG:
+		pmt_caps = pmt_caps_crashlog;
+		break;
+	case FEATURE_PETE_LOG:
+		pmt_caps = pmt_caps_pete;
+		break;
+	case FEATURE_TPMI_CTRL:
+		pmt_caps = pmt_caps_tpmi;
+		break;
+	case FEATURE_TRACING:
+		pmt_caps = pmt_caps_tracing;
+		break;
+	case FEATURE_PER_RMID_ENERGY_TELEM:
+		pmt_caps = pmt_caps_rmid_energy;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	while (*pmt_caps) {
+		struct pmt_cap *pmt_cap = *pmt_caps;
+
+		while (pmt_cap->name) {
+			ret += sysfs_emit_at(buf, ret, "%-40s Available: %s\n", pmt_cap->name,
+					     str_yes_no(pmt_cap->mask & caps));
+			pmt_cap++;
+		}
+		pmt_caps++;
+	}
+
+	return ret;
+}
+static struct kobj_attribute caps_attribute = __ATTR_RO(caps);
+
+static struct watcher *get_watcher(struct feature *feature)
+{
+	switch (feature_layout[feature->id]) {
+	case LAYOUT_RMID:
+		return &feature->table.rmid.watcher;
+	case LAYOUT_WATCHER:
+		return &feature->table.watcher;
+	default:
+		return ERR_PTR(-EINVAL);
+	}
+}
+
+static struct command *get_command(struct feature *feature)
+{
+	switch (feature_layout[feature->id]) {
+	case LAYOUT_RMID:
+		return &feature->table.rmid.watcher.command;
+	case LAYOUT_WATCHER:
+		return &feature->table.watcher.command;
+	case LAYOUT_COMMAND:
+		return &feature->table.command;
+	default:
+		return ERR_PTR(-EINVAL);
+	}
+}
+
+static ssize_t num_rmids_show(struct kobject *kobj,
+			      struct kobj_attribute *attr, char *buf)
+{
+	struct feature *feature = to_pmt_feature(kobj);
+
+	return sysfs_emit(buf, "%u\n", feature->table.rmid.num_rmids);
+}
+static struct kobj_attribute num_rmids_attribute = __ATTR_RO(num_rmids);
+
+static ssize_t min_watcher_period_ms_show(struct kobject *kobj,
+					  struct kobj_attribute *attr, char *buf)
+{
+	struct feature *feature = to_pmt_feature(kobj);
+	struct watcher *watcher = get_watcher(feature);
+
+	if (IS_ERR(watcher))
+		return PTR_ERR(watcher);
+
+	return sysfs_emit(buf, "%u\n", watcher->period);
+}
+static struct kobj_attribute min_watcher_period_ms_attribute =
+	__ATTR_RO(min_watcher_period_ms);
+
+static ssize_t max_stream_size_show(struct kobject *kobj,
+				    struct kobj_attribute *attr, char *buf)
+{
+	struct feature *feature = to_pmt_feature(kobj);
+	struct command *command = get_command(feature);
+
+	if (IS_ERR(command))
+		return PTR_ERR(command);
+
+	return sysfs_emit(buf, "%u\n", command->max_stream_size);
+}
+static struct kobj_attribute max_stream_size_attribute =
+	__ATTR_RO(max_stream_size);
+
+static ssize_t max_command_size_show(struct kobject *kobj,
+				     struct kobj_attribute *attr, char *buf)
+{
+	struct feature *feature = to_pmt_feature(kobj);
+	struct command *command = get_command(feature);
+
+	if (IS_ERR(command))
+		return PTR_ERR(command);
+
+	return sysfs_emit(buf, "%u\n", command->max_command_size);
+}
+static struct kobj_attribute max_command_size_attribute =
+	__ATTR_RO(max_command_size);
+
+static ssize_t guids_show(struct kobject *kobj, struct kobj_attribute *attr,
+			  char *buf)
+{
+	struct feature *feature = to_pmt_feature(kobj);
+	int i, count = 0;
+
+	for (i = 0; i < feature->table.header.num_guids; i++)
+		count += sysfs_emit_at(buf, count, "0x%x\n",
+				       feature->table.guids[i]);
+
+	return count;
+}
+static struct kobj_attribute guids_attribute = __ATTR_RO(guids);
+
+static struct attribute *pmt_feature_rmid_attrs[] = {
+	&caps_attribute.attr,
+	&num_rmids_attribute.attr,
+	&min_watcher_period_ms_attribute.attr,
+	&max_stream_size_attribute.attr,
+	&max_command_size_attribute.attr,
+	&guids_attribute.attr,
+	NULL
+};
+ATTRIBUTE_GROUPS(pmt_feature_rmid);
+
+static const struct kobj_type pmt_feature_rmid_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+	.release = pmt_feature_release,
+	.default_groups = pmt_feature_rmid_groups,
+};
+
+static struct attribute *pmt_feature_watcher_attrs[] = {
+	&caps_attribute.attr,
+	&min_watcher_period_ms_attribute.attr,
+	&max_stream_size_attribute.attr,
+	&max_command_size_attribute.attr,
+	&guids_attribute.attr,
+	NULL
+};
+ATTRIBUTE_GROUPS(pmt_feature_watcher);
+
+static const struct kobj_type pmt_feature_watcher_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+	.release = pmt_feature_release,
+	.default_groups = pmt_feature_watcher_groups,
+};
+
+static struct attribute *pmt_feature_command_attrs[] = {
+	&caps_attribute.attr,
+	&max_stream_size_attribute.attr,
+	&max_command_size_attribute.attr,
+	&guids_attribute.attr,
+	NULL
+};
+ATTRIBUTE_GROUPS(pmt_feature_command);
+
+static const struct kobj_type pmt_feature_command_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+	.release = pmt_feature_release,
+	.default_groups = pmt_feature_command_groups,
+};
+
+static struct attribute *pmt_feature_guids_attrs[] = {
+	&caps_attribute.attr,
+	&guids_attribute.attr,
+	NULL
+};
+ATTRIBUTE_GROUPS(pmt_feature_guids);
+
+static const struct kobj_type pmt_feature_guids_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+	.release = pmt_feature_release,
+	.default_groups = pmt_feature_guids_groups,
+};
+
+static int
+pmt_feature_get_disc_table(struct pmt_features_priv *priv,
+			   struct resource *disc_res,
+			   struct feature_discovery_table *disc_tbl)
+{
+	void __iomem *disc_base;
+
+	disc_base = devm_ioremap_resource(priv->dev, disc_res);
+	if (IS_ERR(disc_base))
+		return PTR_ERR(disc_base);
+
+	memcpy_fromio(disc_tbl, disc_base, sizeof(*disc_tbl));
+
+	devm_iounmap(priv->dev, disc_base);
+
+	if (priv->mask & BIT(disc_tbl->id))
+		return dev_err_probe(priv->dev, -EINVAL, "Duplicate feature: %s\n",
+				     pmt_feature_names[disc_tbl->id]);
+
+	/*
+	 * Some devices may expose non-functioning entries that are
+	 * reserved for future use. They have zero size. Do not fail
+	 * probe for these. Just ignore them.
+	 */
+	if (disc_tbl->size == 0 || disc_tbl->access_type == PMT_ACCESS_TYPE_RSVD)
+		return SKIP_FEATURE;
+
+	if (disc_tbl->version > MAX_FEATURE_VERSION)
+		return SKIP_FEATURE;
+
+	if (!pmt_feature_id_is_valid(disc_tbl->id))
+		return SKIP_FEATURE;
+
+	priv->mask |= BIT(disc_tbl->id);
+
+	return 0;
+}
+
+static int
+pmt_feature_get_feature_table(struct pmt_features_priv *priv,
+			      struct feature *feature,
+			      struct feature_discovery_table *disc_tbl,
+			      struct resource *disc_res)
+{
+	struct feature_table *feat_tbl = &feature->table;
+	struct feature_header *header;
+	struct resource res = {};
+	resource_size_t res_size;
+	void __iomem *feat_base, *feat_offset;
+	void *tbl_offset;
+	size_t size;
+	u32 *guids;
+	u8 tbir;
+
+	tbir = FIELD_GET(DT_TBIR, disc_tbl->offset);
+
+	switch (disc_tbl->access_type) {
+	case ACCESS_LOCAL:
+		if (tbir)
+			return dev_err_probe(priv->dev, -EINVAL,
+				"Unsupported BAR index %u for access type %u\n",
+				tbir, disc_tbl->access_type);
+
+
+		/*
+		 * For access_type LOCAL, the base address is as follows:
+		 * base address = end of discovery region + base offset + 1
+		 */
+		res = DEFINE_RES_MEM(disc_res->end + disc_tbl->offset + 1,
+				     disc_tbl->size * sizeof(u32));
+		break;
+
+	default:
+		return dev_err_probe(priv->dev, -EINVAL, "Unrecognized access_type %u\n",
+				     disc_tbl->access_type);
+	}
+
+	feature->id = disc_tbl->id;
+
+	/* Get the feature table */
+	feat_base = devm_ioremap_resource(priv->dev, &res);
+	if (IS_ERR(feat_base))
+		return PTR_ERR(feat_base);
+
+	feat_offset = feat_base;
+	tbl_offset = feat_tbl;
+
+	/* Get the header */
+	header = &feat_tbl->header;
+	memcpy_fromio(header, feat_offset, sizeof(*header));
+
+	/* Validate fields fit within mapped resource */
+	size = sizeof(*header) + FEAT_ATTR_SIZE(header->attr_size) +
+	       PMT_GUID_SIZE(header->num_guids);
+	res_size = resource_size(&res);
+	if (WARN(size > res_size, "Bad table size %ld > %pa", size, &res_size))
+		return -EINVAL;
+
+	/* Get the feature attributes, including capability fields */
+	tbl_offset += sizeof(*header);
+	feat_offset += sizeof(*header);
+
+	memcpy_fromio(tbl_offset, feat_offset, FEAT_ATTR_SIZE(header->attr_size));
+
+	/* Finally, get the guids */
+	guids = devm_kmalloc(priv->dev, PMT_GUID_SIZE(header->num_guids), GFP_KERNEL);
+	if (!guids)
+		return -ENOMEM;
+
+	feat_offset += FEAT_ATTR_SIZE(header->attr_size);
+
+	memcpy_fromio(guids, feat_offset, PMT_GUID_SIZE(header->num_guids));
+
+	feat_tbl->guids = guids;
+
+	devm_iounmap(priv->dev, feat_base);
+
+	return 0;
+}
+
+static void pmt_features_add_feat(struct feature *feature)
+{
+	guard(mutex)(&feature_list_lock);
+	list_add(&feature->list, &pmt_feature_list);
+}
+
+static void pmt_features_remove_feat(struct feature *feature)
+{
+	guard(mutex)(&feature_list_lock);
+	list_del(&feature->list);
+}
+
+/* Get the discovery table and use it to get the feature table */
+static int pmt_features_discovery(struct pmt_features_priv *priv,
+				  struct feature *feature,
+				  struct intel_vsec_device *ivdev,
+				  int idx)
+{
+	struct feature_discovery_table disc_tbl = {}; /* Avoid false warning */
+	struct resource *disc_res = &ivdev->resource[idx];
+	const struct kobj_type *ktype;
+	int ret;
+
+	ret = pmt_feature_get_disc_table(priv, disc_res, &disc_tbl);
+	if (ret)
+		return ret;
+
+	ret = pmt_feature_get_feature_table(priv, feature, &disc_tbl, disc_res);
+	if (ret)
+		return ret;
+
+	switch (feature_layout[feature->id]) {
+	case LAYOUT_RMID:
+		ktype = &pmt_feature_rmid_ktype;
+		feature->attr_group = &pmt_feature_rmid_group;
+		break;
+	case LAYOUT_WATCHER:
+		ktype = &pmt_feature_watcher_ktype;
+		feature->attr_group = &pmt_feature_watcher_group;
+		break;
+	case LAYOUT_COMMAND:
+		ktype = &pmt_feature_command_ktype;
+		feature->attr_group = &pmt_feature_command_group;
+		break;
+	case LAYOUT_CAPS_ONLY:
+		ktype = &pmt_feature_guids_ktype;
+		feature->attr_group = &pmt_feature_guids_group;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	ret = kobject_init_and_add(&feature->kobj, ktype, &priv->dev->kobj,
+				   pmt_feature_names[feature->id]);
+	if (ret)
+		return ret;
+
+	kobject_uevent(&feature->kobj, KOBJ_ADD);
+	pmt_features_add_feat(feature);
+
+	return 0;
+}
+
+static void pmt_features_remove(struct auxiliary_device *auxdev)
+{
+	struct pmt_features_priv *priv = auxiliary_get_drvdata(auxdev);
+	int i;
+
+	for (i = 0; i < priv->count; i++) {
+		struct feature *feature = &priv->feature[i];
+
+		pmt_features_remove_feat(feature);
+		sysfs_remove_group(&feature->kobj, feature->attr_group);
+		kobject_put(&feature->kobj);
+	}
+
+	device_unregister(priv->dev);
+}
+
+static int pmt_features_probe(struct auxiliary_device *auxdev, const struct auxiliary_device_id *id)
+{
+	struct intel_vsec_device *ivdev = auxdev_to_ivdev(auxdev);
+	struct pmt_features_priv *priv;
+	size_t size;
+	int ret, i;
+
+	size = struct_size(priv, feature, ivdev->num_resources);
+	priv = devm_kzalloc(&auxdev->dev, size, GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
+	priv->parent = &ivdev->pcidev->dev;
+	auxiliary_set_drvdata(auxdev, priv);
+
+	priv->dev = device_create(&intel_pmt_class, &auxdev->dev, MKDEV(0, 0), priv,
+				  "%s-%s", "features", dev_name(priv->parent));
+	if (IS_ERR(priv->dev))
+		return dev_err_probe(priv->dev, PTR_ERR(priv->dev),
+				     "Could not create %s-%s device node\n",
+				     "features", dev_name(priv->dev));
+
+	/* Initialize each feature */
+	for (i = 0; i < ivdev->num_resources; i++) {
+		struct feature *feature = &priv->feature[priv->count];
+
+		ret = pmt_features_discovery(priv, feature, ivdev, i);
+		if (ret == SKIP_FEATURE)
+			continue;
+		if (ret != 0)
+			goto abort_probe;
+
+		feature->priv = priv;
+		priv->count++;
+	}
+
+	return 0;
+
+abort_probe:
+	/*
+	 * Only fully initialized features are tracked in priv->count, which is
+	 * incremented only after a feature is completely set up (i.e., after
+	 * discovery and sysfs registration). If feature initialization fails,
+	 * the failing feature's state is local and does not require rollback.
+	 *
+	 * Therefore, on error, we can safely call the driver's remove() routine
+	 * pmt_features_remove() to clean up only those features that were
+	 * fully initialized and counted. All other resources are device-managed
+	 * and will be cleaned up automatically during device_unregister().
+	 */
+	pmt_features_remove(auxdev);
+
+	return ret;
+}
+
+static const struct auxiliary_device_id pmt_features_id_table[] = {
+	{ .name = "intel_vsec.discovery" },
+	{}
+};
+MODULE_DEVICE_TABLE(auxiliary, pmt_features_id_table);
+
+static struct auxiliary_driver pmt_features_aux_driver = {
+	.id_table	= pmt_features_id_table,
+	.remove		= pmt_features_remove,
+	.probe		= pmt_features_probe,
+};
+module_auxiliary_driver(pmt_features_aux_driver);
+
+MODULE_AUTHOR("David E. Box <david.e.box@linux.intel.com>");
+MODULE_DESCRIPTION("Intel PMT Discovery driver");
+MODULE_LICENSE("GPL");
+MODULE_IMPORT_NS("INTEL_PMT");