drm/xe: Expose PCIe link downgrade attributes

Expose sysfs attributes for PCIe link downgrade capability and status.

v2: Move from debugfs to sysfs (Lucas, Rodrigo, Badal)
    Rework macros and their naming (Rodrigo)
v3: Use sysfs_create_files() (Riana)
    Fix checkpatch warning (Riana)
v4: s/downspeed/downgrade (Lucas, Rodrigo, Riana)
v5: Use PCIe Gen agnostic naming (Rodrigo)
v6: s/pcie_gen/auto_link (Lucas)

Signed-off-by: Raag Jadav <raag.jadav@intel.com>
Reviewed-by: Riana Tauro <riana.tauro@intel.com>
Link: https://lore.kernel.org/r/20250506054835.3395220-3-raag.jadav@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

2 files changed, 96 insertions, 2 deletions

diff --git a/drivers/gpu/drm/xe/xe_device_sysfs.c b/drivers/gpu/drm/xe/xe_device_sysfs.c
index d4c73acea1cf..9628e9a0a0af 100644
--- a/drivers/gpu/drm/xe/xe_device_sysfs.c
+++ b/drivers/gpu/drm/xe/xe_device_sysfs.c
@@ -3,14 +3,16 @@
  * Copyright © 2023 Intel Corporation
  */
 
+#include <linux/device.h>
 #include <linux/kobject.h>
 #include <linux/pci.h>
 #include <linux/sysfs.h>
 
-#include <drm/drm_managed.h>
-
 #include "xe_device.h"
 #include "xe_device_sysfs.h"
+#include "xe_mmio.h"
+#include "xe_pcode_api.h"
+#include "xe_pcode.h"
 #include "xe_pm.h"
 
 /**
@@ -63,12 +65,93 @@ vram_d3cold_threshold_store(struct device *dev, struct device_attribute *attr,
 
 static DEVICE_ATTR_RW(vram_d3cold_threshold);
 
+/**
+ * DOC: PCIe Gen5 Limitations
+ *
+ * Default link speed of discrete GPUs is determined by configuration parameters
+ * stored in their flash memory, which are subject to override through user
+ * initiated firmware updates. It has been observed that devices configured with
+ * PCIe Gen5 as their default link speed can come across link quality issues due
+ * to host or motherboard limitations and may have to auto-downgrade their link
+ * to PCIe Gen4 speed when faced with unstable link at Gen5, which makes
+ * firmware updates rather risky on such setups. It is required to ensure that
+ * the device is capable of auto-downgrading its link to PCIe Gen4 speed before
+ * pushing the firmware image with PCIe Gen5 as default configuration. This can
+ * be done by reading ``auto_link_downgrade_capable`` sysfs entry, which will
+ * denote if the device is capable of auto-downgrading its link to PCIe Gen4
+ * speed with boolean output value of ``0`` or ``1``, meaning `incapable` or
+ * `capable` respectively.
+ *
+ * .. code-block:: shell
+ *
+ *    $ cat /sys/bus/pci/devices/<bdf>/auto_link_downgrade_capable
+ *
+ * Pushing the firmware image with PCIe Gen5 as default configuration on a auto
+ * link downgrade incapable device and facing link instability due to host or
+ * motherboard limitations can result in driver failing to bind to the device,
+ * making further firmware updates impossible with RMA being the only last
+ * resort.
+ *
+ * Link downgrade status of auto link downgrade capable devices is available
+ * through ``auto_link_downgrade_status`` sysfs entry with boolean output value
+ * of ``0`` or ``1``, where ``0`` means no auto-downgrading was required during
+ * link training (which is the optimal scenario) and ``1`` means the device has
+ * auto-downgraded its link to PCIe Gen4 speed due to unstable Gen5 link.
+ *
+ * .. code-block:: shell
+ *
+ *    $ cat /sys/bus/pci/devices/<bdf>/auto_link_downgrade_status
+ */
+
+static ssize_t
+auto_link_downgrade_capable_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct xe_device *xe = pdev_to_xe_device(pdev);
+	u32 cap, val;
+
+	xe_pm_runtime_get(xe);
+	val = xe_mmio_read32(xe_root_tile_mmio(xe), BMG_PCIE_CAP);
+	xe_pm_runtime_put(xe);
+
+	cap = REG_FIELD_GET(LINK_DOWNGRADE, val);
+	return sysfs_emit(buf, "%u\n", cap == DOWNGRADE_CAPABLE ? true : false);
+}
+static DEVICE_ATTR_ADMIN_RO(auto_link_downgrade_capable);
+
+static ssize_t
+auto_link_downgrade_status_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct xe_device *xe = pdev_to_xe_device(pdev);
+	u32 val;
+	int ret;
+
+	xe_pm_runtime_get(xe);
+	ret = xe_pcode_read(xe_device_get_root_tile(xe),
+			    PCODE_MBOX(DGFX_PCODE_STATUS, DGFX_GET_INIT_STATUS, 0),
+			    &val, NULL);
+	xe_pm_runtime_put(xe);
+
+	return ret ?: sysfs_emit(buf, "%u\n", REG_FIELD_GET(DGFX_LINK_DOWNGRADE_STATUS, val));
+}
+static DEVICE_ATTR_ADMIN_RO(auto_link_downgrade_status);
+
+static const struct attribute *auto_link_downgrade_attrs[] = {
+	&dev_attr_auto_link_downgrade_capable.attr,
+	&dev_attr_auto_link_downgrade_status.attr,
+	NULL
+};
+
 static void xe_device_sysfs_fini(void *arg)
 {
 	struct xe_device *xe = arg;
 
 	if (xe->d3cold.capable)
 		sysfs_remove_file(&xe->drm.dev->kobj, &dev_attr_vram_d3cold_threshold.attr);
+
+	if (xe->info.platform == XE_BATTLEMAGE)
+		sysfs_remove_files(&xe->drm.dev->kobj, auto_link_downgrade_attrs);
 }
 
 int xe_device_sysfs_init(struct xe_device *xe)
@@ -82,5 +165,11 @@ int xe_device_sysfs_init(struct xe_device *xe)
 			return ret;
 	}
 
+	if (xe->info.platform == XE_BATTLEMAGE) {
+		ret = sysfs_create_files(&dev->kobj, auto_link_downgrade_attrs);
+		if (ret)
+			return ret;
+	}
+
 	return devm_add_action_or_reset(dev, xe_device_sysfs_fini, xe);
 }
diff --git a/drivers/gpu/drm/xe/xe_pcode_api.h b/drivers/gpu/drm/xe/xe_pcode_api.h
index e622ae17f08d..127d4d26c4cf 100644
--- a/drivers/gpu/drm/xe/xe_pcode_api.h
+++ b/drivers/gpu/drm/xe/xe_pcode_api.h
@@ -34,6 +34,7 @@
 #define   DGFX_PCODE_STATUS		0x7E
 #define     DGFX_GET_INIT_STATUS	0x0
 #define     DGFX_INIT_STATUS_COMPLETE	0x1
+#define     DGFX_LINK_DOWNGRADE_STATUS	REG_BIT(31)
 
 #define   PCODE_POWER_SETUP			0x7C
 #define     POWER_SETUP_SUBCOMMAND_READ_I1	0x4
@@ -66,6 +67,10 @@
 /* Auxiliary info bits */
 #define   AUXINFO_HISTORY_OFFSET	REG_GENMASK(31, 29)
 
+#define BMG_PCIE_CAP			XE_REG(0x138340)
+#define   LINK_DOWNGRADE		REG_GENMASK(1, 0)
+#define     DOWNGRADE_CAPABLE		2
+
 struct pcode_err_decode {
 	int errno;
 	const char *str;