* [PATCH v2 1/3] platform/x86/intel-uncore-freq: Uncore frequency control via TPMI
2023-04-18 17:13 [PATCH v2 0/3] Uncore frequency scaling using TPMI Srinivas Pandruvada
@ 2023-04-18 17:13 ` Srinivas Pandruvada
2023-04-20 11:25 ` Ilpo Järvinen
2023-04-18 17:13 ` [PATCH v2 2/3] platform/x86/intel-uncore-freq: Support for cluster level controls Srinivas Pandruvada
` (2 subsequent siblings)
3 siblings, 1 reply; 7+ messages in thread
From: Srinivas Pandruvada @ 2023-04-18 17:13 UTC (permalink / raw)
To: hdegoede, markgross
Cc: platform-driver-x86, linux-kernel, Srinivas Pandruvada,
Zhang Rui, Wendy Wang
Implement support of uncore frequency control via TPMI (Topology Aware
Register and PM Capsule Interface). This driver provides the similar
functionality as the current uncore frequency driver using MSRs.
The hardware interface to read/write is basically substitution of MSR
0x620 and 0x621. There are specific MMIO offset and bits to get/set
minimum and maximum uncore ratio, similar to MSRs.
The scope of the uncore MSRs is package/die. But new generation of CPUs
have more granular control at a cluster level. Each package/die can have
multiple power domains, which further can have multiple clusters. The
TPMI interface allows control at cluster level.
The primary use case for uncore sysfs is to set maximum and minimum
uncore frequency to reduce power consumption or latency. The current
uncore sysfs control is per package/die. This is enough for the majority
of users as workload will move to different power domains as it moves
between different CPUs.
The current uncore sysfs provides controls at package/die level. When
user sets maximum/minimum limits, the driver sets the same limits to
each cluster.
Here number of power domains = number of resources in this aux device.
There are offsets and bits to discover number of clusters and offset for
each cluster level controls.
The TPMI documentation can be downloaded from:
https://github.com/intel/tpmi_power_management
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>
---
v2
- Changed mmio to u8* (Hans)
- Not setting pd_info->uncore_base to NULL (Hans)
- Handling failure of devm_kcalloc() (Hans)
- Merged init/remove to probe/remove functions (Rui)
- Log when platform is NULL (Rui)
.../x86/intel/uncore-frequency/Kconfig | 4 +
.../x86/intel/uncore-frequency/Makefile | 2 +
.../uncore-frequency/uncore-frequency-tpmi.c | 338 ++++++++++++++++++
3 files changed, 344 insertions(+)
create mode 100644 drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
diff --git a/drivers/platform/x86/intel/uncore-frequency/Kconfig b/drivers/platform/x86/intel/uncore-frequency/Kconfig
index 21b209124916..a56d55056927 100644
--- a/drivers/platform/x86/intel/uncore-frequency/Kconfig
+++ b/drivers/platform/x86/intel/uncore-frequency/Kconfig
@@ -6,9 +6,13 @@
menu "Intel Uncore Frequency Control"
depends on X86_64 || COMPILE_TEST
+config INTEL_UNCORE_FREQ_CONTROL_TPMI
+ tristate
+
config INTEL_UNCORE_FREQ_CONTROL
tristate "Intel Uncore frequency control driver"
depends on X86_64
+ select INTEL_UNCORE_FREQ_CONTROL_TPMI if INTEL_TPMI
help
This driver allows control of Uncore frequency limits on
supported server platforms.
diff --git a/drivers/platform/x86/intel/uncore-frequency/Makefile b/drivers/platform/x86/intel/uncore-frequency/Makefile
index e0f7968e8285..08ff57492b28 100644
--- a/drivers/platform/x86/intel/uncore-frequency/Makefile
+++ b/drivers/platform/x86/intel/uncore-frequency/Makefile
@@ -7,3 +7,5 @@ obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL) += intel-uncore-frequency.o
intel-uncore-frequency-y := uncore-frequency.o
obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL) += intel-uncore-frequency-common.o
intel-uncore-frequency-common-y := uncore-frequency-common.o
+obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL_TPMI) += intel-uncore-frequency-tpmi.o
+intel-uncore-frequency-tpmi-y := uncore-frequency-tpmi.o
diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
new file mode 100644
index 000000000000..5e454e9dd4a7
--- /dev/null
+++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
@@ -0,0 +1,338 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * uncore-frquency-tpmi: Uncore frequency scaling using TPMI
+ *
+ * Copyright (c) 2023, Intel Corporation.
+ * All Rights Reserved.
+ *
+ * The hardware interface to read/write is basically substitution of
+ * MSR 0x620 and 0x621.
+ * There are specific MMIO offset and bits to get/set minimum and
+ * maximum uncore ratio, similar to MSRs.
+ * The scope of the uncore MSRs was package scope. But TPMI allows
+ * new gen CPUs to have multiple uncore controls at uncore-cluster
+ * level. Each package can have multiple power domains which further
+ * can have multiple clusters.
+ * Here number of power domains = number of resources in this aux
+ * device. There are offsets and bits to discover number of clusters
+ * and offset for each cluster level controls.
+ *
+ */
+
+#include <linux/auxiliary_bus.h>
+#include <linux/bitfield.h>
+#include <linux/bits.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/intel_tpmi.h>
+
+#include "uncore-frequency-common.h"
+
+#define UNCORE_HEADER_VERSION 1
+#define UNCORE_HEADER_INDEX 0
+#define UNCORE_FABRIC_CLUSTER_OFFSET 8
+
+/* status + control + adv_ctl1 + adv_ctl2 */
+#define UNCORE_FABRIC_CLUSTER_SIZE (4 * 8)
+
+#define UNCORE_STATUS_INDEX 0
+#define UNCORE_CONTROL_INDEX 8
+
+#define UNCORE_FREQ_KHZ_MULTIPLIER 100000
+
+struct tpmi_uncore_struct;
+
+/* Information for each cluster */
+struct tpmi_uncore_cluster_info {
+ u8 __iomem *cluster_base;
+ struct uncore_data uncore_data;
+ struct tpmi_uncore_struct *uncore_root;
+};
+
+/* Information for each power domain */
+struct tpmi_uncore_power_domain_info {
+ u8 __iomem *uncore_base;
+ int ufs_header_ver;
+ int cluster_count;
+ struct tpmi_uncore_cluster_info *cluster_infos;
+};
+
+/* Information for all power domains in a package */
+struct tpmi_uncore_struct {
+ int power_domain_count;
+ struct tpmi_uncore_power_domain_info *pd_info;
+ struct tpmi_uncore_cluster_info root_cluster;
+};
+
+#define UNCORE_GENMASK_MIN_RATIO GENMASK_ULL(21, 15)
+#define UNCORE_GENMASK_MAX_RATIO GENMASK_ULL(14, 8)
+
+/* Helper function to read MMIO offset for max/min control frequency */
+static void read_control_freq(struct tpmi_uncore_cluster_info *cluster_info,
+ unsigned int *min, unsigned int *max)
+{
+ u64 control;
+
+ control = readq(cluster_info->cluster_base + UNCORE_CONTROL_INDEX);
+ *max = FIELD_GET(UNCORE_GENMASK_MAX_RATIO, control) * UNCORE_FREQ_KHZ_MULTIPLIER;
+ *min = FIELD_GET(UNCORE_GENMASK_MIN_RATIO, control) * UNCORE_FREQ_KHZ_MULTIPLIER;
+}
+
+#define UNCORE_MAX_RATIO 0x7F
+
+/* Callback for sysfs read for max/min frequencies. Called under mutex locks */
+static int uncore_read_control_freq(struct uncore_data *data, unsigned int *min,
+ unsigned int *max)
+{
+ struct tpmi_uncore_cluster_info *cluster_info;
+ struct tpmi_uncore_struct *uncore_root;
+ int i, _min = 0, _max = 0;
+
+ cluster_info = container_of(data, struct tpmi_uncore_cluster_info, uncore_data);
+ uncore_root = cluster_info->uncore_root;
+
+ *min = UNCORE_MAX_RATIO * UNCORE_FREQ_KHZ_MULTIPLIER;
+ *max = 0;
+
+ /*
+ * Get the max/min by looking at each cluster. Get the lowest
+ * min and highest max.
+ */
+ for (i = 0; i < uncore_root->power_domain_count; ++i) {
+ int j;
+
+ for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j) {
+ read_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
+ &_min, &_max);
+ if (*min > _min)
+ *min = _min;
+ if (*max < _max)
+ *max = _max;
+ }
+ }
+
+ return 0;
+}
+
+/* Helper function to write MMIO offset for max/min control frequency */
+static void write_control_freq(struct tpmi_uncore_cluster_info *cluster_info, unsigned int input,
+ unsigned int min_max)
+{
+ u64 control;
+
+ control = readq(cluster_info->cluster_base + UNCORE_CONTROL_INDEX);
+
+ if (min_max) {
+ control &= ~UNCORE_GENMASK_MAX_RATIO;
+ control |= FIELD_PREP(UNCORE_GENMASK_MAX_RATIO, input);
+ } else {
+ control &= ~UNCORE_GENMASK_MIN_RATIO;
+ control |= FIELD_PREP(UNCORE_GENMASK_MIN_RATIO, input);
+ }
+
+ writeq(control, (cluster_info->cluster_base + UNCORE_CONTROL_INDEX));
+}
+
+/* Callback for sysfs write for max/min frequencies. Called under mutex locks */
+static int uncore_write_control_freq(struct uncore_data *data, unsigned int input,
+ unsigned int min_max)
+{
+ struct tpmi_uncore_cluster_info *cluster_info;
+ struct tpmi_uncore_struct *uncore_root;
+ int i;
+
+ input /= UNCORE_FREQ_KHZ_MULTIPLIER;
+ if (!input || input > UNCORE_MAX_RATIO)
+ return -EINVAL;
+
+ cluster_info = container_of(data, struct tpmi_uncore_cluster_info, uncore_data);
+ uncore_root = cluster_info->uncore_root;
+
+ /* Update each cluster in a package */
+ for (i = 0; i < uncore_root->power_domain_count; ++i) {
+ int j;
+
+ for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j)
+ write_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
+ input, min_max);
+ }
+
+ return 0;
+}
+
+/* Callback for sysfs read for the current uncore frequency. Called under mutex locks */
+static int uncore_read_freq(struct uncore_data *data, unsigned int *freq)
+{
+ return -ENODATA;
+}
+
+#define UNCORE_GENMASK_VERSION GENMASK_ULL(7, 0)
+#define UNCORE_LOCAL_FABRIC_CLUSTER_ID_MASK GENMASK_ULL(15, 8)
+#define UNCORE_CLUSTER_OFF_MASK GENMASK_ULL(7, 0)
+#define UNCORE_MAX_CLUSTER_PER_DOMAIN 8
+
+static int uncore_probe(struct auxiliary_device *auxdev, const struct auxiliary_device_id *id)
+{
+ struct intel_tpmi_plat_info *plat_info;
+ struct tpmi_uncore_struct *tpmi_uncore;
+ int ret, i, pkg = 0;
+ int num_resources;
+
+ /* Get number of power domains, which is equal to number of resources */
+ num_resources = tpmi_get_resource_count(auxdev);
+ if (!num_resources)
+ return -EINVAL;
+
+ /* Register callbacks to uncore core */
+ ret = uncore_freq_common_init(uncore_read_control_freq, uncore_write_control_freq,
+ uncore_read_freq);
+ if (ret)
+ return ret;
+
+ /* Allocate uncore instance per package */
+ tpmi_uncore = devm_kzalloc(&auxdev->dev, sizeof(*tpmi_uncore), GFP_KERNEL);
+ if (!tpmi_uncore) {
+ ret = -ENOMEM;
+ goto err_rem_common;
+ }
+
+ /* Allocate memory for all power domains in a package */
+ tpmi_uncore->pd_info = devm_kcalloc(&auxdev->dev, num_resources,
+ sizeof(*tpmi_uncore->pd_info),
+ GFP_KERNEL);
+ if (!tpmi_uncore->pd_info) {
+ ret = -ENOMEM;
+ goto err_rem_common;
+ }
+
+ tpmi_uncore->power_domain_count = num_resources;
+
+ /* Get the package ID from the TPMI core */
+ plat_info = tpmi_get_platform_data(auxdev);
+ if (plat_info)
+ pkg = plat_info->package_id;
+ else
+ dev_info(&auxdev->dev, "Platform information is NULL\n");
+
+ for (i = 0; i < num_resources; ++i) {
+ struct tpmi_uncore_power_domain_info *pd_info;
+ struct resource *res;
+ u64 cluster_offset;
+ u8 cluster_mask;
+ int mask, j;
+ u64 header;
+
+ res = tpmi_get_resource_at_index(auxdev, i);
+ if (!res)
+ continue;
+
+ pd_info = &tpmi_uncore->pd_info[i];
+
+ pd_info->uncore_base = devm_ioremap_resource(&auxdev->dev, res);
+ if (IS_ERR(pd_info->uncore_base)) {
+ ret = PTR_ERR(pd_info->uncore_base);
+ goto err_rem_common;
+ }
+
+ /* Check for version and skip this resource if there is mismatch */
+ header = readq(pd_info->uncore_base);
+ pd_info->ufs_header_ver = header & UNCORE_GENMASK_VERSION;
+ if (pd_info->ufs_header_ver != UNCORE_HEADER_VERSION) {
+ dev_info(&auxdev->dev, "Uncore: Unsupported version:%d\n",
+ pd_info->ufs_header_ver);
+ continue;
+ }
+
+ /* Get Cluster ID Mask */
+ cluster_mask = FIELD_GET(UNCORE_LOCAL_FABRIC_CLUSTER_ID_MASK, header);
+ if (!cluster_mask) {
+ dev_info(&auxdev->dev, "Uncore: Invalid cluster mask:%x\n", cluster_mask);
+ continue;
+ }
+
+ /* Find out number of clusters in this resource */
+ mask = 0x01;
+ for (j = 0; j < UNCORE_MAX_CLUSTER_PER_DOMAIN; ++j) {
+ if (cluster_mask & mask)
+ pd_info->cluster_count++;
+ mask <<= 1;
+ }
+
+ pd_info->cluster_infos = devm_kcalloc(&auxdev->dev, pd_info->cluster_count,
+ sizeof(struct tpmi_uncore_cluster_info),
+ GFP_KERNEL);
+ if (!pd_info->cluster_infos) {
+ ret = -ENOMEM;
+ goto err_rem_common;
+ }
+ /*
+ * Each byte in the register point to status and control
+ * registers belonging to cluster id 0-8.
+ */
+ cluster_offset = readq(pd_info->uncore_base +
+ UNCORE_FABRIC_CLUSTER_OFFSET);
+
+ for (j = 0; j < pd_info->cluster_count; ++j) {
+ struct tpmi_uncore_cluster_info *cluster_info;
+
+ /* Get the offset for this cluster */
+ mask = (cluster_offset & UNCORE_CLUSTER_OFF_MASK);
+ /* Offset in QWORD, so change to bytes */
+ mask <<= 3;
+
+ cluster_info = &pd_info->cluster_infos[j];
+
+ cluster_info->cluster_base = pd_info->uncore_base + mask;
+
+ cluster_info->uncore_data.package_id = pkg;
+ /* There are no dies like Cascade Lake */
+ cluster_info->uncore_data.die_id = 0;
+
+ /* Point to next cluster offset */
+ cluster_offset >>= UNCORE_MAX_CLUSTER_PER_DOMAIN;
+ }
+ }
+
+ auxiliary_set_drvdata(auxdev, tpmi_uncore);
+
+ tpmi_uncore->root_cluster.uncore_root = tpmi_uncore;
+ tpmi_uncore->root_cluster.uncore_data.package_id = pkg;
+ ret = uncore_freq_add_entry(&tpmi_uncore->root_cluster.uncore_data, 0);
+ if (ret)
+ goto err_rem_common;
+
+ return 0;
+
+err_rem_common:
+ uncore_freq_common_exit();
+
+ return ret;
+}
+
+static void uncore_remove(struct auxiliary_device *auxdev)
+{
+ struct tpmi_uncore_struct *tpmi_uncore = auxiliary_get_drvdata(auxdev);
+
+ uncore_freq_remove_die_entry(&tpmi_uncore->root_cluster.uncore_data);
+
+ uncore_freq_common_exit();
+}
+
+static const struct auxiliary_device_id intel_uncore_id_table[] = {
+ { .name = "intel_vsec.tpmi-uncore" },
+ {}
+};
+MODULE_DEVICE_TABLE(auxiliary, intel_uncore_id_table);
+
+static struct auxiliary_driver intel_uncore_aux_driver = {
+ .id_table = intel_uncore_id_table,
+ .remove = uncore_remove,
+ .probe = uncore_probe,
+};
+
+module_auxiliary_driver(intel_uncore_aux_driver);
+
+MODULE_IMPORT_NS(INTEL_TPMI);
+MODULE_IMPORT_NS(INTEL_UNCORE_FREQUENCY);
+MODULE_DESCRIPTION("Intel TPMI UFS Driver");
+MODULE_LICENSE("GPL");
--
2.38.1
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH v2 1/3] platform/x86/intel-uncore-freq: Uncore frequency control via TPMI
2023-04-18 17:13 ` [PATCH v2 1/3] platform/x86/intel-uncore-freq: Uncore frequency control via TPMI Srinivas Pandruvada
@ 2023-04-20 11:25 ` Ilpo Järvinen
2023-04-20 22:06 ` srinivas pandruvada
0 siblings, 1 reply; 7+ messages in thread
From: Ilpo Järvinen @ 2023-04-20 11:25 UTC (permalink / raw)
To: Srinivas Pandruvada
Cc: hdegoede, markgross, platform-driver-x86, linux-kernel,
Zhang Rui, Wendy Wang
On Tue, 18 Apr 2023, Srinivas Pandruvada wrote:
> Implement support of uncore frequency control via TPMI (Topology Aware
> Register and PM Capsule Interface). This driver provides the similar
> functionality as the current uncore frequency driver using MSRs.
>
> The hardware interface to read/write is basically substitution of MSR
> 0x620 and 0x621. There are specific MMIO offset and bits to get/set
> minimum and maximum uncore ratio, similar to MSRs.
>
> The scope of the uncore MSRs is package/die. But new generation of CPUs
> have more granular control at a cluster level. Each package/die can have
> multiple power domains, which further can have multiple clusters. The
> TPMI interface allows control at cluster level.
>
> The primary use case for uncore sysfs is to set maximum and minimum
> uncore frequency to reduce power consumption or latency. The current
> uncore sysfs control is per package/die. This is enough for the majority
> of users as workload will move to different power domains as it moves
> between different CPUs.
>
> The current uncore sysfs provides controls at package/die level. When
> user sets maximum/minimum limits, the driver sets the same limits to
> each cluster.
>
> Here number of power domains = number of resources in this aux device.
> There are offsets and bits to discover number of clusters and offset for
> each cluster level controls.
>
> The TPMI documentation can be downloaded from:
> https://github.com/intel/tpmi_power_management
>
> Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
> Reviewed-by: Zhang Rui <rui.zhang@intel.com>
> Tested-by: Wendy Wang <wendy.wang@intel.com>
> ---
> v2
> - Changed mmio to u8* (Hans)
> - Not setting pd_info->uncore_base to NULL (Hans)
> - Handling failure of devm_kcalloc() (Hans)
> - Merged init/remove to probe/remove functions (Rui)
> - Log when platform is NULL (Rui)
>
> .../x86/intel/uncore-frequency/Kconfig | 4 +
> .../x86/intel/uncore-frequency/Makefile | 2 +
> .../uncore-frequency/uncore-frequency-tpmi.c | 338 ++++++++++++++++++
> 3 files changed, 344 insertions(+)
> create mode 100644 drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
>
> diff --git a/drivers/platform/x86/intel/uncore-frequency/Kconfig b/drivers/platform/x86/intel/uncore-frequency/Kconfig
> index 21b209124916..a56d55056927 100644
> --- a/drivers/platform/x86/intel/uncore-frequency/Kconfig
> +++ b/drivers/platform/x86/intel/uncore-frequency/Kconfig
> @@ -6,9 +6,13 @@
> menu "Intel Uncore Frequency Control"
> depends on X86_64 || COMPILE_TEST
>
> +config INTEL_UNCORE_FREQ_CONTROL_TPMI
> + tristate
> +
> config INTEL_UNCORE_FREQ_CONTROL
> tristate "Intel Uncore frequency control driver"
> depends on X86_64
> + select INTEL_UNCORE_FREQ_CONTROL_TPMI if INTEL_TPMI
> help
> This driver allows control of Uncore frequency limits on
> supported server platforms.
> diff --git a/drivers/platform/x86/intel/uncore-frequency/Makefile b/drivers/platform/x86/intel/uncore-frequency/Makefile
> index e0f7968e8285..08ff57492b28 100644
> --- a/drivers/platform/x86/intel/uncore-frequency/Makefile
> +++ b/drivers/platform/x86/intel/uncore-frequency/Makefile
> @@ -7,3 +7,5 @@ obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL) += intel-uncore-frequency.o
> intel-uncore-frequency-y := uncore-frequency.o
> obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL) += intel-uncore-frequency-common.o
> intel-uncore-frequency-common-y := uncore-frequency-common.o
> +obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL_TPMI) += intel-uncore-frequency-tpmi.o
> +intel-uncore-frequency-tpmi-y := uncore-frequency-tpmi.o
> diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
> new file mode 100644
> index 000000000000..5e454e9dd4a7
> --- /dev/null
> +++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
> @@ -0,0 +1,338 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * uncore-frquency-tpmi: Uncore frequency scaling using TPMI
> + *
> + * Copyright (c) 2023, Intel Corporation.
> + * All Rights Reserved.
> + *
> + * The hardware interface to read/write is basically substitution of
> + * MSR 0x620 and 0x621.
> + * There are specific MMIO offset and bits to get/set minimum and
> + * maximum uncore ratio, similar to MSRs.
> + * The scope of the uncore MSRs was package scope. But TPMI allows
> + * new gen CPUs to have multiple uncore controls at uncore-cluster
> + * level. Each package can have multiple power domains which further
> + * can have multiple clusters.
> + * Here number of power domains = number of resources in this aux
> + * device. There are offsets and bits to discover number of clusters
> + * and offset for each cluster level controls.
> + *
> + */
> +
> +#include <linux/auxiliary_bus.h>
> +#include <linux/bitfield.h>
> +#include <linux/bits.h>
> +#include <linux/io.h>
> +#include <linux/module.h>
> +#include <linux/intel_tpmi.h>
> +
> +#include "uncore-frequency-common.h"
> +
> +#define UNCORE_HEADER_VERSION 1
> +#define UNCORE_HEADER_INDEX 0
> +#define UNCORE_FABRIC_CLUSTER_OFFSET 8
> +
> +/* status + control + adv_ctl1 + adv_ctl2 */
> +#define UNCORE_FABRIC_CLUSTER_SIZE (4 * 8)
> +
> +#define UNCORE_STATUS_INDEX 0
> +#define UNCORE_CONTROL_INDEX 8
> +
> +#define UNCORE_FREQ_KHZ_MULTIPLIER 100000
> +
> +struct tpmi_uncore_struct;
> +
> +/* Information for each cluster */
> +struct tpmi_uncore_cluster_info {
> + u8 __iomem *cluster_base;
> + struct uncore_data uncore_data;
> + struct tpmi_uncore_struct *uncore_root;
> +};
> +
> +/* Information for each power domain */
> +struct tpmi_uncore_power_domain_info {
> + u8 __iomem *uncore_base;
> + int ufs_header_ver;
> + int cluster_count;
> + struct tpmi_uncore_cluster_info *cluster_infos;
> +};
> +
> +/* Information for all power domains in a package */
> +struct tpmi_uncore_struct {
> + int power_domain_count;
> + struct tpmi_uncore_power_domain_info *pd_info;
> + struct tpmi_uncore_cluster_info root_cluster;
> +};
> +
> +#define UNCORE_GENMASK_MIN_RATIO GENMASK_ULL(21, 15)
> +#define UNCORE_GENMASK_MAX_RATIO GENMASK_ULL(14, 8)
> +
> +/* Helper function to read MMIO offset for max/min control frequency */
> +static void read_control_freq(struct tpmi_uncore_cluster_info *cluster_info,
> + unsigned int *min, unsigned int *max)
> +{
> + u64 control;
> +
> + control = readq(cluster_info->cluster_base + UNCORE_CONTROL_INDEX);
> + *max = FIELD_GET(UNCORE_GENMASK_MAX_RATIO, control) * UNCORE_FREQ_KHZ_MULTIPLIER;
> + *min = FIELD_GET(UNCORE_GENMASK_MIN_RATIO, control) * UNCORE_FREQ_KHZ_MULTIPLIER;
> +}
> +
> +#define UNCORE_MAX_RATIO 0x7F
FIELD_MAX(UNCORE_GENMASK_MAX_RATIO) ?
> +
> +/* Callback for sysfs read for max/min frequencies. Called under mutex locks */
> +static int uncore_read_control_freq(struct uncore_data *data, unsigned int *min,
> + unsigned int *max)
> +{
> + struct tpmi_uncore_cluster_info *cluster_info;
> + struct tpmi_uncore_struct *uncore_root;
> + int i, _min = 0, _max = 0;
> +
> + cluster_info = container_of(data, struct tpmi_uncore_cluster_info, uncore_data);
> + uncore_root = cluster_info->uncore_root;
> +
> + *min = UNCORE_MAX_RATIO * UNCORE_FREQ_KHZ_MULTIPLIER;
> + *max = 0;
> +
> + /*
> + * Get the max/min by looking at each cluster. Get the lowest
> + * min and highest max.
> + */
> + for (i = 0; i < uncore_root->power_domain_count; ++i) {
> + int j;
> +
> + for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j) {
> + read_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
> + &_min, &_max);
> + if (*min > _min)
> + *min = _min;
> + if (*max < _max)
> + *max = _max;
> + }
> + }
> +
> + return 0;
> +}
> +
> +/* Helper function to write MMIO offset for max/min control frequency */
> +static void write_control_freq(struct tpmi_uncore_cluster_info *cluster_info, unsigned int input,
> + unsigned int min_max)
> +{
> + u64 control;
> +
> + control = readq(cluster_info->cluster_base + UNCORE_CONTROL_INDEX);
> +
> + if (min_max) {
> + control &= ~UNCORE_GENMASK_MAX_RATIO;
> + control |= FIELD_PREP(UNCORE_GENMASK_MAX_RATIO, input);
> + } else {
> + control &= ~UNCORE_GENMASK_MIN_RATIO;
> + control |= FIELD_PREP(UNCORE_GENMASK_MIN_RATIO, input);
> + }
> +
> + writeq(control, (cluster_info->cluster_base + UNCORE_CONTROL_INDEX));
> +}
> +
> +/* Callback for sysfs write for max/min frequencies. Called under mutex locks */
> +static int uncore_write_control_freq(struct uncore_data *data, unsigned int input,
> + unsigned int min_max)
> +{
> + struct tpmi_uncore_cluster_info *cluster_info;
> + struct tpmi_uncore_struct *uncore_root;
> + int i;
> +
> + input /= UNCORE_FREQ_KHZ_MULTIPLIER;
> + if (!input || input > UNCORE_MAX_RATIO)
> + return -EINVAL;
> +
> + cluster_info = container_of(data, struct tpmi_uncore_cluster_info, uncore_data);
> + uncore_root = cluster_info->uncore_root;
> +
> + /* Update each cluster in a package */
> + for (i = 0; i < uncore_root->power_domain_count; ++i) {
> + int j;
> +
> + for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j)
> + write_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
> + input, min_max);
> + }
> +
> + return 0;
> +}
> +
> +/* Callback for sysfs read for the current uncore frequency. Called under mutex locks */
> +static int uncore_read_freq(struct uncore_data *data, unsigned int *freq)
> +{
> + return -ENODATA;
> +}
> +
> +#define UNCORE_GENMASK_VERSION GENMASK_ULL(7, 0)
GENMASK makes the name little bit confusing (what is "uncore genmask" ??).
Either drop it entirely (my preference) or add _MASK to the end instead.
> +#define UNCORE_LOCAL_FABRIC_CLUSTER_ID_MASK GENMASK_ULL(15, 8)
> +#define UNCORE_CLUSTER_OFF_MASK GENMASK_ULL(7, 0)
> +#define UNCORE_MAX_CLUSTER_PER_DOMAIN 8
> +
> +static int uncore_probe(struct auxiliary_device *auxdev, const struct auxiliary_device_id *id)
> +{
> + struct intel_tpmi_plat_info *plat_info;
> + struct tpmi_uncore_struct *tpmi_uncore;
> + int ret, i, pkg = 0;
> + int num_resources;
> +
> + /* Get number of power domains, which is equal to number of resources */
> + num_resources = tpmi_get_resource_count(auxdev);
> + if (!num_resources)
> + return -EINVAL;
> +
> + /* Register callbacks to uncore core */
> + ret = uncore_freq_common_init(uncore_read_control_freq, uncore_write_control_freq,
> + uncore_read_freq);
> + if (ret)
> + return ret;
> +
> + /* Allocate uncore instance per package */
> + tpmi_uncore = devm_kzalloc(&auxdev->dev, sizeof(*tpmi_uncore), GFP_KERNEL);
> + if (!tpmi_uncore) {
> + ret = -ENOMEM;
> + goto err_rem_common;
> + }
> +
> + /* Allocate memory for all power domains in a package */
> + tpmi_uncore->pd_info = devm_kcalloc(&auxdev->dev, num_resources,
> + sizeof(*tpmi_uncore->pd_info),
> + GFP_KERNEL);
> + if (!tpmi_uncore->pd_info) {
> + ret = -ENOMEM;
> + goto err_rem_common;
> + }
> +
> + tpmi_uncore->power_domain_count = num_resources;
> +
> + /* Get the package ID from the TPMI core */
> + plat_info = tpmi_get_platform_data(auxdev);
> + if (plat_info)
> + pkg = plat_info->package_id;
> + else
> + dev_info(&auxdev->dev, "Platform information is NULL\n");
> +
> + for (i = 0; i < num_resources; ++i) {
> + struct tpmi_uncore_power_domain_info *pd_info;
> + struct resource *res;
> + u64 cluster_offset;
> + u8 cluster_mask;
> + int mask, j;
> + u64 header;
> +
> + res = tpmi_get_resource_at_index(auxdev, i);
> + if (!res)
> + continue;
> +
> + pd_info = &tpmi_uncore->pd_info[i];
> +
> + pd_info->uncore_base = devm_ioremap_resource(&auxdev->dev, res);
> + if (IS_ERR(pd_info->uncore_base)) {
> + ret = PTR_ERR(pd_info->uncore_base);
> + goto err_rem_common;
> + }
> +
> + /* Check for version and skip this resource if there is mismatch */
> + header = readq(pd_info->uncore_base);
> + pd_info->ufs_header_ver = header & UNCORE_GENMASK_VERSION;
> + if (pd_info->ufs_header_ver != UNCORE_HEADER_VERSION) {
> + dev_info(&auxdev->dev, "Uncore: Unsupported version:%d\n",
> + pd_info->ufs_header_ver);
> + continue;
> + }
> +
> + /* Get Cluster ID Mask */
> + cluster_mask = FIELD_GET(UNCORE_LOCAL_FABRIC_CLUSTER_ID_MASK, header);
> + if (!cluster_mask) {
> + dev_info(&auxdev->dev, "Uncore: Invalid cluster mask:%x\n", cluster_mask);
> + continue;
> + }
> +
> + /* Find out number of clusters in this resource */
> + mask = 0x01;
> + for (j = 0; j < UNCORE_MAX_CLUSTER_PER_DOMAIN; ++j) {
> + if (cluster_mask & mask)
> + pd_info->cluster_count++;
> + mask <<= 1;
> + }
pd_info->cluster_count = hweight8(cluster_mask);
?
--
i.
> +
> + pd_info->cluster_infos = devm_kcalloc(&auxdev->dev, pd_info->cluster_count,
> + sizeof(struct tpmi_uncore_cluster_info),
> + GFP_KERNEL);
> + if (!pd_info->cluster_infos) {
> + ret = -ENOMEM;
> + goto err_rem_common;
> + }
> + /*
> + * Each byte in the register point to status and control
> + * registers belonging to cluster id 0-8.
> + */
> + cluster_offset = readq(pd_info->uncore_base +
> + UNCORE_FABRIC_CLUSTER_OFFSET);
> +
> + for (j = 0; j < pd_info->cluster_count; ++j) {
> + struct tpmi_uncore_cluster_info *cluster_info;
> +
> + /* Get the offset for this cluster */
> + mask = (cluster_offset & UNCORE_CLUSTER_OFF_MASK);
> + /* Offset in QWORD, so change to bytes */
> + mask <<= 3;
> +
> + cluster_info = &pd_info->cluster_infos[j];
> +
> + cluster_info->cluster_base = pd_info->uncore_base + mask;
> +
> + cluster_info->uncore_data.package_id = pkg;
> + /* There are no dies like Cascade Lake */
> + cluster_info->uncore_data.die_id = 0;
> +
> + /* Point to next cluster offset */
> + cluster_offset >>= UNCORE_MAX_CLUSTER_PER_DOMAIN;
> + }
> + }
> +
> + auxiliary_set_drvdata(auxdev, tpmi_uncore);
> +
> + tpmi_uncore->root_cluster.uncore_root = tpmi_uncore;
> + tpmi_uncore->root_cluster.uncore_data.package_id = pkg;
> + ret = uncore_freq_add_entry(&tpmi_uncore->root_cluster.uncore_data, 0);
> + if (ret)
> + goto err_rem_common;
> +
> + return 0;
> +
> +err_rem_common:
> + uncore_freq_common_exit();
> +
> + return ret;
> +}
> +
> +static void uncore_remove(struct auxiliary_device *auxdev)
> +{
> + struct tpmi_uncore_struct *tpmi_uncore = auxiliary_get_drvdata(auxdev);
> +
> + uncore_freq_remove_die_entry(&tpmi_uncore->root_cluster.uncore_data);
> +
> + uncore_freq_common_exit();
> +}
> +
> +static const struct auxiliary_device_id intel_uncore_id_table[] = {
> + { .name = "intel_vsec.tpmi-uncore" },
> + {}
> +};
> +MODULE_DEVICE_TABLE(auxiliary, intel_uncore_id_table);
> +
> +static struct auxiliary_driver intel_uncore_aux_driver = {
> + .id_table = intel_uncore_id_table,
> + .remove = uncore_remove,
> + .probe = uncore_probe,
> +};
> +
> +module_auxiliary_driver(intel_uncore_aux_driver);
> +
> +MODULE_IMPORT_NS(INTEL_TPMI);
> +MODULE_IMPORT_NS(INTEL_UNCORE_FREQUENCY);
> +MODULE_DESCRIPTION("Intel TPMI UFS Driver");
> +MODULE_LICENSE("GPL");
>
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH v2 1/3] platform/x86/intel-uncore-freq: Uncore frequency control via TPMI
2023-04-20 11:25 ` Ilpo Järvinen
@ 2023-04-20 22:06 ` srinivas pandruvada
0 siblings, 0 replies; 7+ messages in thread
From: srinivas pandruvada @ 2023-04-20 22:06 UTC (permalink / raw)
To: Ilpo Järvinen
Cc: hdegoede, markgross, platform-driver-x86, linux-kernel,
Zhang Rui, Wendy Wang
On Thu, 2023-04-20 at 14:25 +0300, Ilpo Järvinen wrote:
> On Tue, 18 Apr 2023, Srinivas Pandruvada wrote:
>
> > Implement support of uncore frequency control via TPMI (Topology
> > Aware
> > Register and PM Capsule Interface). This driver provides the
> > similar
> > functionality as the current uncore frequency driver using MSRs.
> >
> > The hardware interface to read/write is basically substitution of
> > MSR
> > 0x620 and 0x621. There are specific MMIO offset and bits to get/set
> > minimum and maximum uncore ratio, similar to MSRs.
> >
> > The scope of the uncore MSRs is package/die. But new generation of
> > CPUs
> > have more granular control at a cluster level. Each package/die can
> > have
> > multiple power domains, which further can have multiple clusters.
> > The
> > TPMI interface allows control at cluster level.
> >
> > The primary use case for uncore sysfs is to set maximum and minimum
> > uncore frequency to reduce power consumption or latency. The
> > current
> > uncore sysfs control is per package/die. This is enough for the
> > majority
> > of users as workload will move to different power domains as it
> > moves
> > between different CPUs.
> >
> > The current uncore sysfs provides controls at package/die level.
> > When
> > user sets maximum/minimum limits, the driver sets the same limits
> > to
> > each cluster.
> >
> > Here number of power domains = number of resources in this aux
> > device.
> > There are offsets and bits to discover number of clusters and
> > offset for
> > each cluster level controls.
> >
> > The TPMI documentation can be downloaded from:
> > https://github.com/intel/tpmi_power_management
> >
> > Signed-off-by: Srinivas Pandruvada
> > <srinivas.pandruvada@linux.intel.com>
> > Reviewed-by: Zhang Rui <rui.zhang@intel.com>
> > Tested-by: Wendy Wang <wendy.wang@intel.com>
> > ---
Submitted update with the suggested changes here.
Thanks,
Srinivas
> > v2
> > - Changed mmio to u8* (Hans)
> > - Not setting pd_info->uncore_base to NULL (Hans)
> > - Handling failure of devm_kcalloc() (Hans)
> > - Merged init/remove to probe/remove functions (Rui)
> > - Log when platform is NULL (Rui)
> >
> > .../x86/intel/uncore-frequency/Kconfig | 4 +
> > .../x86/intel/uncore-frequency/Makefile | 2 +
> > .../uncore-frequency/uncore-frequency-tpmi.c | 338
> > ++++++++++++++++++
> > 3 files changed, 344 insertions(+)
> > create mode 100644 drivers/platform/x86/intel/uncore-
> > frequency/uncore-frequency-tpmi.c
> >
> > diff --git a/drivers/platform/x86/intel/uncore-frequency/Kconfig
> > b/drivers/platform/x86/intel/uncore-frequency/Kconfig
> > index 21b209124916..a56d55056927 100644
> > --- a/drivers/platform/x86/intel/uncore-frequency/Kconfig
> > +++ b/drivers/platform/x86/intel/uncore-frequency/Kconfig
> > @@ -6,9 +6,13 @@
> > menu "Intel Uncore Frequency Control"
> > depends on X86_64 || COMPILE_TEST
> >
> > +config INTEL_UNCORE_FREQ_CONTROL_TPMI
> > + tristate
> > +
> > config INTEL_UNCORE_FREQ_CONTROL
> > tristate "Intel Uncore frequency control driver"
> > depends on X86_64
> > + select INTEL_UNCORE_FREQ_CONTROL_TPMI if INTEL_TPMI
> > help
> > This driver allows control of Uncore frequency limits on
> > supported server platforms.
> > diff --git a/drivers/platform/x86/intel/uncore-frequency/Makefile
> > b/drivers/platform/x86/intel/uncore-frequency/Makefile
> > index e0f7968e8285..08ff57492b28 100644
> > --- a/drivers/platform/x86/intel/uncore-frequency/Makefile
> > +++ b/drivers/platform/x86/intel/uncore-frequency/Makefile
> > @@ -7,3 +7,5 @@ obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL) += intel-
> > uncore-frequency.o
> > intel-uncore-frequency-y := uncore-frequency.o
> > obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL) += intel-uncore-
> > frequency-common.o
> > intel-uncore-frequency-common-y := uncore-
> > frequency-common.o
> > +obj-$(CONFIG_INTEL_UNCORE_FREQ_CONTROL_TPMI) += intel-uncore-
> > frequency-tpmi.o
> > +intel-uncore-frequency-tpmi-y := uncore-frequency-tpmi.o
> > diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-
> > frequency-tpmi.c b/drivers/platform/x86/intel/uncore-
> > frequency/uncore-frequency-tpmi.c
> > new file mode 100644
> > index 000000000000..5e454e9dd4a7
> > --- /dev/null
> > +++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-
> > tpmi.c
> > @@ -0,0 +1,338 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * uncore-frquency-tpmi: Uncore frequency scaling using TPMI
> > + *
> > + * Copyright (c) 2023, Intel Corporation.
> > + * All Rights Reserved.
> > + *
> > + * The hardware interface to read/write is basically substitution
> > of
> > + * MSR 0x620 and 0x621.
> > + * There are specific MMIO offset and bits to get/set minimum and
> > + * maximum uncore ratio, similar to MSRs.
> > + * The scope of the uncore MSRs was package scope. But TPMI allows
> > + * new gen CPUs to have multiple uncore controls at uncore-cluster
> > + * level. Each package can have multiple power domains which
> > further
> > + * can have multiple clusters.
> > + * Here number of power domains = number of resources in this aux
> > + * device. There are offsets and bits to discover number of
> > clusters
> > + * and offset for each cluster level controls.
> > + *
> > + */
> > +
> > +#include <linux/auxiliary_bus.h>
> > +#include <linux/bitfield.h>
> > +#include <linux/bits.h>
> > +#include <linux/io.h>
> > +#include <linux/module.h>
> > +#include <linux/intel_tpmi.h>
> > +
> > +#include "uncore-frequency-common.h"
> > +
> > +#define UNCORE_HEADER_VERSION 1
> > +#define UNCORE_HEADER_INDEX 0
> > +#define UNCORE_FABRIC_CLUSTER_OFFSET 8
> > +
> > +/* status + control + adv_ctl1 + adv_ctl2 */
> > +#define UNCORE_FABRIC_CLUSTER_SIZE (4 * 8)
> > +
> > +#define UNCORE_STATUS_INDEX 0
> > +#define UNCORE_CONTROL_INDEX 8
> > +
> > +#define UNCORE_FREQ_KHZ_MULTIPLIER 100000
> > +
> > +struct tpmi_uncore_struct;
> > +
> > +/* Information for each cluster */
> > +struct tpmi_uncore_cluster_info {
> > + u8 __iomem *cluster_base;
> > + struct uncore_data uncore_data;
> > + struct tpmi_uncore_struct *uncore_root;
> > +};
> > +
> > +/* Information for each power domain */
> > +struct tpmi_uncore_power_domain_info {
> > + u8 __iomem *uncore_base;
> > + int ufs_header_ver;
> > + int cluster_count;
> > + struct tpmi_uncore_cluster_info *cluster_infos;
> > +};
> > +
> > +/* Information for all power domains in a package */
> > +struct tpmi_uncore_struct {
> > + int power_domain_count;
> > + struct tpmi_uncore_power_domain_info *pd_info;
> > + struct tpmi_uncore_cluster_info root_cluster;
> > +};
> > +
> > +#define UNCORE_GENMASK_MIN_RATIO GENMASK_ULL(21, 15)
> > +#define UNCORE_GENMASK_MAX_RATIO GENMASK_ULL(14, 8)
> > +
> > +/* Helper function to read MMIO offset for max/min control
> > frequency */
> > +static void read_control_freq(struct tpmi_uncore_cluster_info
> > *cluster_info,
> > + unsigned int *min, unsigned int *max)
> > +{
> > + u64 control;
> > +
> > + control = readq(cluster_info->cluster_base +
> > UNCORE_CONTROL_INDEX);
> > + *max = FIELD_GET(UNCORE_GENMASK_MAX_RATIO, control) *
> > UNCORE_FREQ_KHZ_MULTIPLIER;
> > + *min = FIELD_GET(UNCORE_GENMASK_MIN_RATIO, control) *
> > UNCORE_FREQ_KHZ_MULTIPLIER;
> > +}
> > +
> > +#define UNCORE_MAX_RATIO 0x7F
>
> FIELD_MAX(UNCORE_GENMASK_MAX_RATIO) ?
>
> > +
> > +/* Callback for sysfs read for max/min frequencies. Called under
> > mutex locks */
> > +static int uncore_read_control_freq(struct uncore_data *data,
> > unsigned int *min,
> > + unsigned int *max)
> > +{
> > + struct tpmi_uncore_cluster_info *cluster_info;
> > + struct tpmi_uncore_struct *uncore_root;
> > + int i, _min = 0, _max = 0;
> > +
> > + cluster_info = container_of(data, struct
> > tpmi_uncore_cluster_info, uncore_data);
> > + uncore_root = cluster_info->uncore_root;
> > +
> > + *min = UNCORE_MAX_RATIO * UNCORE_FREQ_KHZ_MULTIPLIER;
> > + *max = 0;
> > +
> > + /*
> > + * Get the max/min by looking at each cluster. Get the
> > lowest
> > + * min and highest max.
> > + */
> > + for (i = 0; i < uncore_root->power_domain_count; ++i) {
> > + int j;
> > +
> > + for (j = 0; j < uncore_root-
> > >pd_info[i].cluster_count; ++j) {
> > + read_control_freq(&uncore_root-
> > >pd_info[i].cluster_infos[j],
> > + &_min, &_max);
> > + if (*min > _min)
> > + *min = _min;
> > + if (*max < _max)
> > + *max = _max;
> > + }
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +/* Helper function to write MMIO offset for max/min control
> > frequency */
> > +static void write_control_freq(struct tpmi_uncore_cluster_info
> > *cluster_info, unsigned int input,
> > + unsigned int min_max)
> > +{
> > + u64 control;
> > +
> > + control = readq(cluster_info->cluster_base +
> > UNCORE_CONTROL_INDEX);
> > +
> > + if (min_max) {
> > + control &= ~UNCORE_GENMASK_MAX_RATIO;
> > + control |= FIELD_PREP(UNCORE_GENMASK_MAX_RATIO,
> > input);
> > + } else {
> > + control &= ~UNCORE_GENMASK_MIN_RATIO;
> > + control |= FIELD_PREP(UNCORE_GENMASK_MIN_RATIO,
> > input);
> > + }
> > +
> > + writeq(control, (cluster_info->cluster_base +
> > UNCORE_CONTROL_INDEX));
> > +}
> > +
> > +/* Callback for sysfs write for max/min frequencies. Called under
> > mutex locks */
> > +static int uncore_write_control_freq(struct uncore_data *data,
> > unsigned int input,
> > + unsigned int min_max)
> > +{
> > + struct tpmi_uncore_cluster_info *cluster_info;
> > + struct tpmi_uncore_struct *uncore_root;
> > + int i;
> > +
> > + input /= UNCORE_FREQ_KHZ_MULTIPLIER;
> > + if (!input || input > UNCORE_MAX_RATIO)
> > + return -EINVAL;
> > +
> > + cluster_info = container_of(data, struct
> > tpmi_uncore_cluster_info, uncore_data);
> > + uncore_root = cluster_info->uncore_root;
> > +
> > + /* Update each cluster in a package */
> > + for (i = 0; i < uncore_root->power_domain_count; ++i) {
> > + int j;
> > +
> > + for (j = 0; j < uncore_root-
> > >pd_info[i].cluster_count; ++j)
> > + write_control_freq(&uncore_root-
> > >pd_info[i].cluster_infos[j],
> > + input, min_max);
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +/* Callback for sysfs read for the current uncore frequency.
> > Called under mutex locks */
> > +static int uncore_read_freq(struct uncore_data *data, unsigned int
> > *freq)
> > +{
> > + return -ENODATA;
> > +}
> > +
> > +#define UNCORE_GENMASK_VERSION GENMASK_ULL(7, 0)
>
> GENMASK makes the name little bit confusing (what is "uncore genmask"
> ??).
> Either drop it entirely (my preference) or add _MASK to the end
> instead.
>
> > +#define UNCORE_LOCAL_FABRIC_CLUSTER_ID_MASK GENMASK_ULL(15, 8)
> > +#define
> > UNCORE_CLUSTER_OFF_MASK GENMASK_ULL(7, 0)
> > +#define UNCORE_MAX_CLUSTER_PER_DOMAIN 8
> > +
> > +static int uncore_probe(struct auxiliary_device *auxdev, const
> > struct auxiliary_device_id *id)
> > +{
> > + struct intel_tpmi_plat_info *plat_info;
> > + struct tpmi_uncore_struct *tpmi_uncore;
> > + int ret, i, pkg = 0;
> > + int num_resources;
> > +
> > + /* Get number of power domains, which is equal to number of
> > resources */
> > + num_resources = tpmi_get_resource_count(auxdev);
> > + if (!num_resources)
> > + return -EINVAL;
> > +
> > + /* Register callbacks to uncore core */
> > + ret = uncore_freq_common_init(uncore_read_control_freq,
> > uncore_write_control_freq,
> > + uncore_read_freq);
> > + if (ret)
> > + return ret;
> > +
> > + /* Allocate uncore instance per package */
> > + tpmi_uncore = devm_kzalloc(&auxdev->dev,
> > sizeof(*tpmi_uncore), GFP_KERNEL);
> > + if (!tpmi_uncore) {
> > + ret = -ENOMEM;
> > + goto err_rem_common;
> > + }
> > +
> > + /* Allocate memory for all power domains in a package */
> > + tpmi_uncore->pd_info = devm_kcalloc(&auxdev->dev,
> > num_resources,
> > + sizeof(*tpmi_uncore-
> > >pd_info),
> > + GFP_KERNEL);
> > + if (!tpmi_uncore->pd_info) {
> > + ret = -ENOMEM;
> > + goto err_rem_common;
> > + }
> > +
> > + tpmi_uncore->power_domain_count = num_resources;
> > +
> > + /* Get the package ID from the TPMI core */
> > + plat_info = tpmi_get_platform_data(auxdev);
> > + if (plat_info)
> > + pkg = plat_info->package_id;
> > + else
> > + dev_info(&auxdev->dev, "Platform information is
> > NULL\n");
> > +
> > + for (i = 0; i < num_resources; ++i) {
> > + struct tpmi_uncore_power_domain_info *pd_info;
> > + struct resource *res;
> > + u64 cluster_offset;
> > + u8 cluster_mask;
> > + int mask, j;
> > + u64 header;
> > +
> > + res = tpmi_get_resource_at_index(auxdev, i);
> > + if (!res)
> > + continue;
> > +
> > + pd_info = &tpmi_uncore->pd_info[i];
> > +
> > + pd_info->uncore_base =
> > devm_ioremap_resource(&auxdev->dev, res);
> > + if (IS_ERR(pd_info->uncore_base)) {
> > + ret = PTR_ERR(pd_info->uncore_base);
> > + goto err_rem_common;
> > + }
> > +
> > + /* Check for version and skip this resource if
> > there is mismatch */
> > + header = readq(pd_info->uncore_base);
> > + pd_info->ufs_header_ver = header &
> > UNCORE_GENMASK_VERSION;
> > + if (pd_info->ufs_header_ver !=
> > UNCORE_HEADER_VERSION) {
> > + dev_info(&auxdev->dev, "Uncore: Unsupported
> > version:%d\n",
> > + pd_info->ufs_header_ver);
> > + continue;
> > + }
> > +
> > + /* Get Cluster ID Mask */
> > + cluster_mask =
> > FIELD_GET(UNCORE_LOCAL_FABRIC_CLUSTER_ID_MASK, header);
> > + if (!cluster_mask) {
> > + dev_info(&auxdev->dev, "Uncore: Invalid
> > cluster mask:%x\n", cluster_mask);
> > + continue;
> > + }
> > +
> > + /* Find out number of clusters in this resource */
> > + mask = 0x01;
> > + for (j = 0; j < UNCORE_MAX_CLUSTER_PER_DOMAIN; ++j)
> > {
> > + if (cluster_mask & mask)
> > + pd_info->cluster_count++;
> > + mask <<= 1;
> > + }
>
> pd_info->cluster_count = hweight8(cluster_mask);
>
> ?
>
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH v2 2/3] platform/x86/intel-uncore-freq: Support for cluster level controls
2023-04-18 17:13 [PATCH v2 0/3] Uncore frequency scaling using TPMI Srinivas Pandruvada
2023-04-18 17:13 ` [PATCH v2 1/3] platform/x86/intel-uncore-freq: Uncore frequency control via TPMI Srinivas Pandruvada
@ 2023-04-18 17:13 ` Srinivas Pandruvada
2023-04-18 17:13 ` [PATCH v2 3/3] platform/x86/intel-uncore-freq: tpmi: Provide cluster level control Srinivas Pandruvada
2023-05-09 9:19 ` [PATCH v2 0/3] Uncore frequency scaling using TPMI Hans de Goede
3 siblings, 0 replies; 7+ messages in thread
From: Srinivas Pandruvada @ 2023-04-18 17:13 UTC (permalink / raw)
To: hdegoede, markgross
Cc: platform-driver-x86, linux-kernel, Srinivas Pandruvada,
Zhang Rui, Wendy Wang
An SoC can contain multiple power domains with individual or collection
of mesh partitions. This partition is called fabric cluster.
Certain type of meshes will need to run at the same frequency, they will
be placed in the same fabric cluster. Benefit of fabric cluster is that
it offers a scalable mechanism to deal with partitioned fabrics in a SoC.
The current sysfs interface supports control at package and die level.
This interface is not enough to support more granular control at
fabric cluster level.
SoCs with the support of TPMI (Topology Aware Register and PM Capsule
Interface), can have multiple power domains. Each power domain can
contain one or more fabric clusters.
To support such granular controls, enhance uncore common to optionally
create new directories to provide controls at fabric cluster level. It
is also important to have flexibility to change granularity for future
version of SoCs. If the directory name contains scope like:
"package_*_die_*_power_domain_*_cluster_*", then this is not expandable.
The cpufreq policies also have different scopes. There the scope of the
policy (affected_cpus) specified by attributes inside each policy.
So, follow the same model for uncore frequency scaling sysfs as:
"sys/devices/system/cpu/cpufreq/policy*"
Allow client drivers to optionally support granular control for each
fabric cluster. Here, the directory name will be "uncore" suffixed with
an unique instance number. For example: uncore00, uncore01 etc.
Attributes in the directory identify package id, power domain and
fabric cluster id. This interface is expandable even if some new level
of granularity is introduced. A new sysfs attribute can identify new
level.
For compatibility with the existing sysfs and provide easy way to set
limits for each fabric cluster in the package/die, the existing control
at package/die levels are still provided. For majority of users, this is
an easy approach.
For example: On a single package/die system, with three power domains
and one fabric cluster per power domain:
$tree -L 2 /sys/devices/system/cpu/intel_uncore_frequency/
/sys/devices/system/cpu/intel_uncore_frequency/
├── package_00_die_00
│ ├── current_freq_khz
│ ├── initial_max_freq_khz
│ ├── initial_min_freq_khz
│ ├── max_freq_khz
│ └── min_freq_khz
├── uncore00
│ ├── current_freq_khz
│ ├── domain_id
│ ├── fabric_cluster_id
│ ├── initial_max_freq_khz
│ ├── initial_min_freq_khz
│ ├── max_freq_khz
│ ├── min_freq_khz
│ └── package_id
├── uncore01
│ ├── current_freq_khz
│ ├── domain_id
│ ├── fabric_cluster_id
│ ├── initial_max_freq_khz
│ ├── initial_min_freq_khz
│ ├── max_freq_khz
│ ├── min_freq_khz
│ └── package_id
└── uncore02
├── current_freq_khz
├── domain_id
├── fabric_cluster_id
├── initial_max_freq_khz
├── initial_min_freq_khz
├── max_freq_khz
├── min_freq_khz
└── package_id
The attribute for cluster id is "fabric_cluster_id" instead of just
"cluster_id" is to avoid confusion with usage of term clusters in
other part of the Linux kernel.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>
---
New patch with this series.
.../pm/intel_uncore_frequency_scaling.rst | 57 ++++++++++++++++++-
.../uncore-frequency-common.c | 51 ++++++++++++++++-
.../uncore-frequency-common.h | 16 +++++-
.../intel/uncore-frequency/uncore-frequency.c | 1 +
4 files changed, 121 insertions(+), 4 deletions(-)
diff --git a/Documentation/admin-guide/pm/intel_uncore_frequency_scaling.rst b/Documentation/admin-guide/pm/intel_uncore_frequency_scaling.rst
index 09169d935835..5ab3440e6cee 100644
--- a/Documentation/admin-guide/pm/intel_uncore_frequency_scaling.rst
+++ b/Documentation/admin-guide/pm/intel_uncore_frequency_scaling.rst
@@ -5,7 +5,7 @@
Intel Uncore Frequency Scaling
==============================
-:Copyright: |copy| 2022 Intel Corporation
+:Copyright: |copy| 2022-2023 Intel Corporation
:Author: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
@@ -58,3 +58,58 @@ Each package_*_die_* contains the following attributes:
``current_freq_khz``
This attribute is used to get the current uncore frequency.
+
+SoCs with TPMI (Topology Aware Register and PM Capsule Interface)
+-----------------------------------------------------------------
+
+An SoC can contain multiple power domains with individual or collection
+of mesh partitions. This partition is called fabric cluster.
+
+Certain type of meshes will need to run at the same frequency, they will
+be placed in the same fabric cluster. Benefit of fabric cluster is that it
+offers a scalable mechanism to deal with partitioned fabrics in a SoC.
+
+The current sysfs interface supports controls at package and die level.
+This interface is not enough to support more granular control at
+fabric cluster level.
+
+SoCs with the support of TPMI (Topology Aware Register and PM Capsule
+Interface), can have multiple power domains. Each power domain can
+contain one or more fabric clusters.
+
+To represent controls at fabric cluster level in addition to the
+controls at package and die level (like systems without TPMI
+support), sysfs is enhanced. This granular interface is presented in the
+sysfs with directories names prefixed with "uncore". For example:
+uncore00, uncore01 etc.
+
+The scope of control is specified by attributes "package_id", "domain_id"
+and "fabric_cluster_id" in the directory.
+
+Attributes in each directory:
+
+``domain_id``
+ This attribute is used to get the power domain id of this instance.
+
+``fabric_cluster_id``
+ This attribute is used to get the fabric cluster id of this instance.
+
+``package_id``
+ This attribute is used to get the package id of this instance.
+
+The other attributes are same as presented at package_*_die_* level.
+
+In most of current use cases, the "max_freq_khz" and "min_freq_khz"
+is updated at "package_*_die_*" level. This model will be still supported
+with the following approach:
+
+When user uses controls at "package_*_die_*" level, then every fabric
+cluster is affected in that package and die. For example: user changes
+"max_freq_khz" in the package_00_die_00, then "max_freq_khz" for uncore*
+directory with the same package id will be updated. In this case user can
+still update "max_freq_khz" at each uncore* level, which is more restrictive.
+Similarly, user can update "min_freq_khz" at "package_*_die_*" level
+to apply at each uncore* level.
+
+Support for "current_freq_khz" is available only at each fabric cluster
+level (i.e., in uncore* directory).
diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.c b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.c
index fa8f14c925ec..b86e65a8ffdc 100644
--- a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.c
+++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.c
@@ -16,11 +16,34 @@ static struct kobject *uncore_root_kobj;
/* uncore instance count */
static int uncore_instance_count;
+static DEFINE_IDA(intel_uncore_ida);
+
/* callbacks for actual HW read/write */
static int (*uncore_read)(struct uncore_data *data, unsigned int *min, unsigned int *max);
static int (*uncore_write)(struct uncore_data *data, unsigned int input, unsigned int min_max);
static int (*uncore_read_freq)(struct uncore_data *data, unsigned int *freq);
+static ssize_t show_domain_id(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct uncore_data *data = container_of(attr, struct uncore_data, domain_id_dev_attr);
+
+ return sprintf(buf, "%u\n", data->domain_id);
+}
+
+static ssize_t show_fabric_cluster_id(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct uncore_data *data = container_of(attr, struct uncore_data, fabric_cluster_id_dev_attr);
+
+ return sprintf(buf, "%u\n", data->cluster_id);
+}
+
+static ssize_t show_package_id(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct uncore_data *data = container_of(attr, struct uncore_data, package_id_dev_attr);
+
+ return sprintf(buf, "%u\n", data->package_id);
+}
+
static ssize_t show_min_max_freq_khz(struct uncore_data *data,
char *buf, int min_max)
{
@@ -161,6 +184,15 @@ static int create_attr_group(struct uncore_data *data, char *name)
init_attribute_ro(initial_max_freq_khz);
init_attribute_root_ro(current_freq_khz);
+ if (data->domain_id != UNCORE_DOMAIN_ID_INVALID) {
+ init_attribute_root_ro(domain_id);
+ data->uncore_attrs[index++] = &data->domain_id_dev_attr.attr;
+ init_attribute_root_ro(fabric_cluster_id);
+ data->uncore_attrs[index++] = &data->fabric_cluster_id_dev_attr.attr;
+ init_attribute_root_ro(package_id);
+ data->uncore_attrs[index++] = &data->package_id_dev_attr.attr;
+ }
+
data->uncore_attrs[index++] = &data->max_freq_khz_dev_attr.attr;
data->uncore_attrs[index++] = &data->min_freq_khz_dev_attr.attr;
data->uncore_attrs[index++] = &data->initial_min_freq_khz_dev_attr.attr;
@@ -191,12 +223,24 @@ int uncore_freq_add_entry(struct uncore_data *data, int cpu)
goto uncore_unlock;
}
- sprintf(data->name, "package_%02d_die_%02d", data->package_id, data->die_id);
+ if (data->domain_id != UNCORE_DOMAIN_ID_INVALID) {
+ ret = ida_alloc(&intel_uncore_ida, GFP_KERNEL);
+ if (ret < 0)
+ goto uncore_unlock;
+
+ data->instance_id = ret;
+ sprintf(data->name, "uncore%02d", ret);
+ } else {
+ sprintf(data->name, "package_%02d_die_%02d", data->package_id, data->die_id);
+ }
uncore_read(data, &data->initial_min_freq_khz, &data->initial_max_freq_khz);
ret = create_attr_group(data, data->name);
- if (!ret) {
+ if (ret) {
+ if (data->domain_id != UNCORE_DOMAIN_ID_INVALID)
+ ida_free(&intel_uncore_ida, data->instance_id);
+ } else {
data->control_cpu = cpu;
data->valid = true;
}
@@ -214,6 +258,9 @@ void uncore_freq_remove_die_entry(struct uncore_data *data)
delete_attr_group(data, data->name);
data->control_cpu = -1;
data->valid = false;
+ if (data->domain_id != UNCORE_DOMAIN_ID_INVALID)
+ ida_free(&intel_uncore_ida, data->instance_id);
+
mutex_unlock(&uncore_lock);
}
EXPORT_SYMBOL_NS_GPL(uncore_freq_remove_die_entry, INTEL_UNCORE_FREQUENCY);
diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.h b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.h
index f5dcfa2fb285..7afb69977c7e 100644
--- a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.h
+++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-common.h
@@ -21,6 +21,9 @@
* @valid: Mark the data valid/invalid
* @package_id: Package id for this instance
* @die_id: Die id for this instance
+ * @domain_id: Power domain id for this instance
+ * @cluster_id: cluster id in a domain
+ * @instance_id: Unique instance id to append to directory name
* @name: Sysfs entry name for this instance
* @uncore_attr_group: Attribute group storage
* @max_freq_khz_dev_attr: Storage for device attribute max_freq_khz
@@ -28,6 +31,9 @@
* @initial_max_freq_khz_dev_attr: Storage for device attribute initial_max_freq_khz
* @initial_min_freq_khz_dev_attr: Storage for device attribute initial_min_freq_khz
* @current_freq_khz_dev_attr: Storage for device attribute current_freq_khz
+ * @domain_id_dev_attr: Storage for device attribute domain_id
+ * @fabric_cluster_id_dev_attr: Storage for device attribute fabric_cluster_id
+ * @package_id_dev_attr: Storage for device attribute package_id
* @uncore_attrs: Attribute storage for group creation
*
* This structure is used to encapsulate all data related to uncore sysfs
@@ -41,6 +47,9 @@ struct uncore_data {
bool valid;
int package_id;
int die_id;
+ int domain_id;
+ int cluster_id;
+ int instance_id;
char name[32];
struct attribute_group uncore_attr_group;
@@ -49,9 +58,14 @@ struct uncore_data {
struct device_attribute initial_max_freq_khz_dev_attr;
struct device_attribute initial_min_freq_khz_dev_attr;
struct device_attribute current_freq_khz_dev_attr;
- struct attribute *uncore_attrs[6];
+ struct device_attribute domain_id_dev_attr;
+ struct device_attribute fabric_cluster_id_dev_attr;
+ struct device_attribute package_id_dev_attr;
+ struct attribute *uncore_attrs[9];
};
+#define UNCORE_DOMAIN_ID_INVALID -1
+
int uncore_freq_common_init(int (*read_control_freq)(struct uncore_data *data, unsigned int *min, unsigned int *max),
int (*write_control_freq)(struct uncore_data *data, unsigned int input, unsigned int min_max),
int (*uncore_read_freq)(struct uncore_data *data, unsigned int *freq));
diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency.c b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency.c
index 00ac7e381441..0ea13c5fbba8 100644
--- a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency.c
+++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency.c
@@ -136,6 +136,7 @@ static int uncore_event_cpu_online(unsigned int cpu)
data->package_id = topology_physical_package_id(cpu);
data->die_id = topology_die_id(cpu);
+ data->domain_id = UNCORE_DOMAIN_ID_INVALID;
return uncore_freq_add_entry(data, cpu);
}
--
2.38.1
^ permalink raw reply [flat|nested] 7+ messages in thread* [PATCH v2 3/3] platform/x86/intel-uncore-freq: tpmi: Provide cluster level control
2023-04-18 17:13 [PATCH v2 0/3] Uncore frequency scaling using TPMI Srinivas Pandruvada
2023-04-18 17:13 ` [PATCH v2 1/3] platform/x86/intel-uncore-freq: Uncore frequency control via TPMI Srinivas Pandruvada
2023-04-18 17:13 ` [PATCH v2 2/3] platform/x86/intel-uncore-freq: Support for cluster level controls Srinivas Pandruvada
@ 2023-04-18 17:13 ` Srinivas Pandruvada
2023-05-09 9:19 ` [PATCH v2 0/3] Uncore frequency scaling using TPMI Hans de Goede
3 siblings, 0 replies; 7+ messages in thread
From: Srinivas Pandruvada @ 2023-04-18 17:13 UTC (permalink / raw)
To: hdegoede, markgross
Cc: platform-driver-x86, linux-kernel, Srinivas Pandruvada,
Zhang Rui, Wendy Wang
The new generation of CPUs have granular control at a cluster level.
Each package/die can have multiple power domains, which further can
have multiple fabric clusters. The TPMI interface allows control at
fabric cluster level.
Use the updated uncore sysfs feature to expose controls at cluster
level. At each cluster level there is a control for maximum and minimum
uncore frequency. Also present current uncore frequency at a cluster
level.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>
---
New patch with this series.
.../uncore-frequency/uncore-frequency-tpmi.c | 136 ++++++++++++++----
1 file changed, 108 insertions(+), 28 deletions(-)
diff --git a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
index 5e454e9dd4a7..b7f7d2a7f42c 100644
--- a/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
+++ b/drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
@@ -44,6 +44,7 @@ struct tpmi_uncore_struct;
/* Information for each cluster */
struct tpmi_uncore_cluster_info {
+ bool root_domain;
u8 __iomem *cluster_base;
struct uncore_data uncore_data;
struct tpmi_uncore_struct *uncore_root;
@@ -60,12 +61,15 @@ struct tpmi_uncore_power_domain_info {
/* Information for all power domains in a package */
struct tpmi_uncore_struct {
int power_domain_count;
+ int max_ratio;
+ int min_ratio;
struct tpmi_uncore_power_domain_info *pd_info;
struct tpmi_uncore_cluster_info root_cluster;
};
#define UNCORE_GENMASK_MIN_RATIO GENMASK_ULL(21, 15)
#define UNCORE_GENMASK_MAX_RATIO GENMASK_ULL(14, 8)
+#define UNCORE_GENMASK_CURRENT_RATIO GENMASK_ULL(6, 0)
/* Helper function to read MMIO offset for max/min control frequency */
static void read_control_freq(struct tpmi_uncore_cluster_info *cluster_info,
@@ -85,32 +89,37 @@ static int uncore_read_control_freq(struct uncore_data *data, unsigned int *min,
unsigned int *max)
{
struct tpmi_uncore_cluster_info *cluster_info;
- struct tpmi_uncore_struct *uncore_root;
- int i, _min = 0, _max = 0;
cluster_info = container_of(data, struct tpmi_uncore_cluster_info, uncore_data);
- uncore_root = cluster_info->uncore_root;
- *min = UNCORE_MAX_RATIO * UNCORE_FREQ_KHZ_MULTIPLIER;
- *max = 0;
+ if (cluster_info->root_domain) {
+ struct tpmi_uncore_struct *uncore_root = cluster_info->uncore_root;
+ int i, _min = 0, _max = 0;
- /*
- * Get the max/min by looking at each cluster. Get the lowest
- * min and highest max.
- */
- for (i = 0; i < uncore_root->power_domain_count; ++i) {
- int j;
+ *min = UNCORE_MAX_RATIO * UNCORE_FREQ_KHZ_MULTIPLIER;
+ *max = 0;
- for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j) {
- read_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
- &_min, &_max);
- if (*min > _min)
- *min = _min;
- if (*max < _max)
- *max = _max;
+ /*
+ * Get the max/min by looking at each cluster. Get the lowest
+ * min and highest max.
+ */
+ for (i = 0; i < uncore_root->power_domain_count; ++i) {
+ int j;
+
+ for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j) {
+ read_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
+ &_min, &_max);
+ if (*min > _min)
+ *min = _min;
+ if (*max < _max)
+ *max = _max;
+ }
}
+ return 0;
}
+ read_control_freq(cluster_info, min, max);
+
return 0;
}
@@ -139,7 +148,6 @@ static int uncore_write_control_freq(struct uncore_data *data, unsigned int inpu
{
struct tpmi_uncore_cluster_info *cluster_info;
struct tpmi_uncore_struct *uncore_root;
- int i;
input /= UNCORE_FREQ_KHZ_MULTIPLIER;
if (!input || input > UNCORE_MAX_RATIO)
@@ -149,21 +157,72 @@ static int uncore_write_control_freq(struct uncore_data *data, unsigned int inpu
uncore_root = cluster_info->uncore_root;
/* Update each cluster in a package */
- for (i = 0; i < uncore_root->power_domain_count; ++i) {
- int j;
+ if (cluster_info->root_domain) {
+ struct tpmi_uncore_struct *uncore_root = cluster_info->uncore_root;
+ int i;
+
+ for (i = 0; i < uncore_root->power_domain_count; ++i) {
+ int j;
+
+ for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j)
+ write_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
+ input, min_max);
+ }
- for (j = 0; j < uncore_root->pd_info[i].cluster_count; ++j)
- write_control_freq(&uncore_root->pd_info[i].cluster_infos[j],
- input, min_max);
+ if (min_max)
+ uncore_root->max_ratio = input;
+ else
+ uncore_root->min_ratio = input;
+
+ return 0;
}
+ if (min_max && uncore_root->max_ratio && uncore_root->max_ratio < input)
+ return -EINVAL;
+
+ if (!min_max && uncore_root->min_ratio && uncore_root->min_ratio > input)
+ return -EINVAL;
+
+ write_control_freq(cluster_info, input, min_max);
+
return 0;
}
/* Callback for sysfs read for the current uncore frequency. Called under mutex locks */
static int uncore_read_freq(struct uncore_data *data, unsigned int *freq)
{
- return -ENODATA;
+ struct tpmi_uncore_cluster_info *cluster_info;
+ u64 status;
+
+ cluster_info = container_of(data, struct tpmi_uncore_cluster_info, uncore_data);
+ if (cluster_info->root_domain)
+ return -ENODATA;
+
+ status = readq((u8 __iomem *)cluster_info->cluster_base + UNCORE_STATUS_INDEX);
+ *freq = FIELD_GET(UNCORE_GENMASK_CURRENT_RATIO, status) * UNCORE_FREQ_KHZ_MULTIPLIER;
+
+ return 0;
+}
+
+static void remove_cluster_entries(struct tpmi_uncore_struct *tpmi_uncore)
+{
+ int i;
+
+ for (i = 0; i < tpmi_uncore->power_domain_count; ++i) {
+ struct tpmi_uncore_power_domain_info *pd_info;
+ int j;
+
+ pd_info = &tpmi_uncore->pd_info[i];
+ if (!pd_info->uncore_base)
+ continue;
+
+ for (j = 0; j < pd_info->cluster_count; ++j) {
+ struct tpmi_uncore_cluster_info *cluster_info;
+
+ cluster_info = &pd_info->cluster_infos[j];
+ uncore_freq_remove_die_entry(&cluster_info->uncore_data);
+ }
+ }
}
#define UNCORE_GENMASK_VERSION GENMASK_ULL(7, 0)
@@ -231,7 +290,13 @@ static int uncore_probe(struct auxiliary_device *auxdev, const struct auxiliary_
pd_info->uncore_base = devm_ioremap_resource(&auxdev->dev, res);
if (IS_ERR(pd_info->uncore_base)) {
ret = PTR_ERR(pd_info->uncore_base);
- goto err_rem_common;
+ /*
+ * Set to NULL so that clean up can still remove other
+ * entries already created if any by
+ * remove_cluster_entries()
+ */
+ pd_info->uncore_base = NULL;
+ goto remove_clusters;
}
/* Check for version and skip this resource if there is mismatch */
@@ -263,7 +328,7 @@ static int uncore_probe(struct auxiliary_device *auxdev, const struct auxiliary_
GFP_KERNEL);
if (!pd_info->cluster_infos) {
ret = -ENOMEM;
- goto err_rem_common;
+ goto remove_clusters;
}
/*
* Each byte in the register point to status and control
@@ -287,7 +352,16 @@ static int uncore_probe(struct auxiliary_device *auxdev, const struct auxiliary_
cluster_info->uncore_data.package_id = pkg;
/* There are no dies like Cascade Lake */
cluster_info->uncore_data.die_id = 0;
+ cluster_info->uncore_data.domain_id = i;
+ cluster_info->uncore_data.cluster_id = j;
+
+ cluster_info->uncore_root = tpmi_uncore;
+ ret = uncore_freq_add_entry(&cluster_info->uncore_data, 0);
+ if (ret) {
+ cluster_info->cluster_base = NULL;
+ goto remove_clusters;
+ }
/* Point to next cluster offset */
cluster_offset >>= UNCORE_MAX_CLUSTER_PER_DOMAIN;
}
@@ -295,14 +369,19 @@ static int uncore_probe(struct auxiliary_device *auxdev, const struct auxiliary_
auxiliary_set_drvdata(auxdev, tpmi_uncore);
+ tpmi_uncore->root_cluster.root_domain = true;
tpmi_uncore->root_cluster.uncore_root = tpmi_uncore;
+
tpmi_uncore->root_cluster.uncore_data.package_id = pkg;
+ tpmi_uncore->root_cluster.uncore_data.domain_id = UNCORE_DOMAIN_ID_INVALID;
ret = uncore_freq_add_entry(&tpmi_uncore->root_cluster.uncore_data, 0);
if (ret)
- goto err_rem_common;
+ goto remove_clusters;
return 0;
+remove_clusters:
+ remove_cluster_entries(tpmi_uncore);
err_rem_common:
uncore_freq_common_exit();
@@ -314,6 +393,7 @@ static void uncore_remove(struct auxiliary_device *auxdev)
struct tpmi_uncore_struct *tpmi_uncore = auxiliary_get_drvdata(auxdev);
uncore_freq_remove_die_entry(&tpmi_uncore->root_cluster.uncore_data);
+ remove_cluster_entries(tpmi_uncore);
uncore_freq_common_exit();
}
--
2.38.1
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH v2 0/3] Uncore frequency scaling using TPMI
2023-04-18 17:13 [PATCH v2 0/3] Uncore frequency scaling using TPMI Srinivas Pandruvada
` (2 preceding siblings ...)
2023-04-18 17:13 ` [PATCH v2 3/3] platform/x86/intel-uncore-freq: tpmi: Provide cluster level control Srinivas Pandruvada
@ 2023-05-09 9:19 ` Hans de Goede
3 siblings, 0 replies; 7+ messages in thread
From: Hans de Goede @ 2023-05-09 9:19 UTC (permalink / raw)
To: Srinivas Pandruvada, markgross; +Cc: platform-driver-x86, linux-kernel
Hi,
On 4/18/23 19:13, Srinivas Pandruvada wrote:
> Implement support of uncore frequency control via TPMI (Topology Aware
> Register and PM Capsule Interface). This driver provides the similar
> functionality as the current uncore frequency driver using MSRs.
>
> Previously I posted only 1/3 and got reviewed and addressed with
> this submission. The other two patches were links in the submission.
>
> But it is better that all three patches considered together.
> I know this is close to merge window for 6.4, so it is fine to differ.
>
> Srinivas Pandruvada (3):
> platform/x86/intel-uncore-freq: Uncore frequency control via TPMI
> platform/x86/intel-uncore-freq: Support for cluster level controls
> platform/x86/intel-uncore-freq: tpmi: Provide cluster level control
Thank you for your patch-series, I've applied the series
(with the updated 1/3) to my review-hans branch:
https://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86.git/log/?h=review-hans
Note it will show up in my review-hans branch once I've pushed my
local branch there, which might take a while.
Once I've run some tests on this branch the patches there will be
added to the platform-drivers-x86/for-next branch and eventually
will be included in the pdx86 pull-request to Linus for the next
merge-window.
Regards,
Hans
>
> .../pm/intel_uncore_frequency_scaling.rst | 57 ++-
> .../x86/intel/uncore-frequency/Kconfig | 4 +
> .../x86/intel/uncore-frequency/Makefile | 2 +
> .../uncore-frequency-common.c | 51 ++-
> .../uncore-frequency-common.h | 16 +-
> .../uncore-frequency/uncore-frequency-tpmi.c | 418 ++++++++++++++++++
> .../intel/uncore-frequency/uncore-frequency.c | 1 +
> 7 files changed, 545 insertions(+), 4 deletions(-)
> create mode 100644 drivers/platform/x86/intel/uncore-frequency/uncore-frequency-tpmi.c
>
^ permalink raw reply [flat|nested] 7+ messages in thread