]> asedeno.scripts.mit.edu Git - linux.git/commitdiff
fpga: dfl: fme: add thermal management support
authorWu Hao <hao.wu@intel.com>
Mon, 14 Oct 2019 05:42:02 +0000 (13:42 +0800)
committerMoritz Fischer <mdf@kernel.org>
Thu, 17 Oct 2019 02:18:27 +0000 (19:18 -0700)
This patch adds support to thermal management private feature for DFL
FPGA Management Engine (FME). This private feature driver registers
a hwmon for thermal/temperature monitoring (hwmon temp1_input).
If hardware automatic throttling is supported by this hardware, then
driver also exposes sysfs interfaces under hwmon for thresholds
(temp1_max/ crit/ emergency), threshold alarms (temp1_max_alarm/
temp1_crit_alarm) and throttling policy (temp1_max_policy).

Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Russ Weight <russell.h.weight@intel.com>
Signed-off-by: Xu Yilun <yilun.xu@intel.com>
Signed-off-by: Wu Hao <hao.wu@intel.com>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Moritz Fischer <mdf@kernel.org>
Signed-off-by: Moritz Fischer <mdf@kernel.org>
Documentation/ABI/testing/sysfs-platform-dfl-fme
drivers/fpga/Kconfig
drivers/fpga/dfl-fme-main.c

index 72634d3ae4f4ebb14e576691ccc2b79879b910ce..8eb6d03e06e7bf0ea1b41db2409c5e5e656a7968 100644 (file)
@@ -106,3 +106,67 @@ KernelVersion:  5.4
 Contact:       Wu Hao <hao.wu@intel.com>
 Description:   Read-only. Read this file to get the second error detected by
                hardware.
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/name
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-Only. Read this file to get the name of hwmon device, it
+               supports values:
+                   'dfl_fme_thermal' - thermal hwmon device name
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_input
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-Only. It returns FPGA device temperature in millidegrees
+               Celsius.
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_max
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-Only. It returns hardware threshold1 temperature in
+               millidegrees Celsius. If temperature rises at or above this
+               threshold, hardware starts 50% or 90% throttling (see
+               'temp1_max_policy').
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_crit
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-Only. It returns hardware threshold2 temperature in
+               millidegrees Celsius. If temperature rises at or above this
+               threshold, hardware starts 100% throttling.
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_emergency
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-Only. It returns hardware trip threshold temperature in
+               millidegrees Celsius. If temperature rises at or above this
+               threshold, a fatal event will be triggered to board management
+               controller (BMC) to shutdown FPGA.
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_max_alarm
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-only. It returns 1 if temperature is currently at or above
+               hardware threshold1 (see 'temp1_max'), otherwise 0.
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_crit_alarm
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-only. It returns 1 if temperature is currently at or above
+               hardware threshold2 (see 'temp1_crit'), otherwise 0.
+
+What:          /sys/bus/platform/devices/dfl-fme.0/hwmon/hwmonX/temp1_max_policy
+Date:          October 2019
+KernelVersion: 5.5
+Contact:       Wu Hao <hao.wu@intel.com>
+Description:   Read-Only. Read this file to get the policy of hardware threshold1
+               (see 'temp1_max'). It only supports two values (policies):
+                   0 - AP2 state (90% throttling)
+                   1 - AP1 state (50% throttling)
index 73c779e920edc2254e74f2a83312a9416d1351b6..72380e1d31c7fa1d5052fc9b20775c4cec2cf608 100644 (file)
@@ -156,7 +156,7 @@ config FPGA_DFL
 
 config FPGA_DFL_FME
        tristate "FPGA DFL FME Driver"
-       depends on FPGA_DFL
+       depends on FPGA_DFL && HWMON
        help
          The FPGA Management Engine (FME) is a feature device implemented
          under Device Feature List (DFL) framework. Select this option to
index 4d78e182878f97acc2f4f2b90a626089e3c04c17..752d71c1748eeae0ce0168b4e826802b93631bf4 100644 (file)
@@ -14,6 +14,8 @@
  *   Henry Mitchel <henry.mitchel@intel.com>
  */
 
+#include <linux/hwmon.h>
+#include <linux/hwmon-sysfs.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/uaccess.h>
@@ -181,6 +183,178 @@ static const struct dfl_feature_ops fme_hdr_ops = {
        .ioctl = fme_hdr_ioctl,
 };
 
+#define FME_THERM_THRESHOLD    0x8
+#define TEMP_THRESHOLD1                GENMASK_ULL(6, 0)
+#define TEMP_THRESHOLD1_EN     BIT_ULL(7)
+#define TEMP_THRESHOLD2                GENMASK_ULL(14, 8)
+#define TEMP_THRESHOLD2_EN     BIT_ULL(15)
+#define TRIP_THRESHOLD         GENMASK_ULL(30, 24)
+#define TEMP_THRESHOLD1_STATUS BIT_ULL(32)             /* threshold1 reached */
+#define TEMP_THRESHOLD2_STATUS BIT_ULL(33)             /* threshold2 reached */
+/* threshold1 policy: 0 - AP2 (90% throttle) / 1 - AP1 (50% throttle) */
+#define TEMP_THRESHOLD1_POLICY BIT_ULL(44)
+
+#define FME_THERM_RDSENSOR_FMT1        0x10
+#define FPGA_TEMPERATURE       GENMASK_ULL(6, 0)
+
+#define FME_THERM_CAP          0x20
+#define THERM_NO_THROTTLE      BIT_ULL(0)
+
+#define MD_PRE_DEG
+
+static bool fme_thermal_throttle_support(void __iomem *base)
+{
+       u64 v = readq(base + FME_THERM_CAP);
+
+       return FIELD_GET(THERM_NO_THROTTLE, v) ? false : true;
+}
+
+static umode_t thermal_hwmon_attrs_visible(const void *drvdata,
+                                          enum hwmon_sensor_types type,
+                                          u32 attr, int channel)
+{
+       const struct dfl_feature *feature = drvdata;
+
+       /* temperature is always supported, and check hardware cap for others */
+       if (attr == hwmon_temp_input)
+               return 0444;
+
+       return fme_thermal_throttle_support(feature->ioaddr) ? 0444 : 0;
+}
+
+static int thermal_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
+                             u32 attr, int channel, long *val)
+{
+       struct dfl_feature *feature = dev_get_drvdata(dev);
+       u64 v;
+
+       switch (attr) {
+       case hwmon_temp_input:
+               v = readq(feature->ioaddr + FME_THERM_RDSENSOR_FMT1);
+               *val = (long)(FIELD_GET(FPGA_TEMPERATURE, v) * 1000);
+               break;
+       case hwmon_temp_max:
+               v = readq(feature->ioaddr + FME_THERM_THRESHOLD);
+               *val = (long)(FIELD_GET(TEMP_THRESHOLD1, v) * 1000);
+               break;
+       case hwmon_temp_crit:
+               v = readq(feature->ioaddr + FME_THERM_THRESHOLD);
+               *val = (long)(FIELD_GET(TEMP_THRESHOLD2, v) * 1000);
+               break;
+       case hwmon_temp_emergency:
+               v = readq(feature->ioaddr + FME_THERM_THRESHOLD);
+               *val = (long)(FIELD_GET(TRIP_THRESHOLD, v) * 1000);
+               break;
+       case hwmon_temp_max_alarm:
+               v = readq(feature->ioaddr + FME_THERM_THRESHOLD);
+               *val = (long)FIELD_GET(TEMP_THRESHOLD1_STATUS, v);
+               break;
+       case hwmon_temp_crit_alarm:
+               v = readq(feature->ioaddr + FME_THERM_THRESHOLD);
+               *val = (long)FIELD_GET(TEMP_THRESHOLD2_STATUS, v);
+               break;
+       default:
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+static const struct hwmon_ops thermal_hwmon_ops = {
+       .is_visible = thermal_hwmon_attrs_visible,
+       .read = thermal_hwmon_read,
+};
+
+static const struct hwmon_channel_info *thermal_hwmon_info[] = {
+       HWMON_CHANNEL_INFO(temp, HWMON_T_INPUT | HWMON_T_EMERGENCY |
+                                HWMON_T_MAX   | HWMON_T_MAX_ALARM |
+                                HWMON_T_CRIT  | HWMON_T_CRIT_ALARM),
+       NULL
+};
+
+static const struct hwmon_chip_info thermal_hwmon_chip_info = {
+       .ops = &thermal_hwmon_ops,
+       .info = thermal_hwmon_info,
+};
+
+static ssize_t temp1_max_policy_show(struct device *dev,
+                                    struct device_attribute *attr, char *buf)
+{
+       struct dfl_feature *feature = dev_get_drvdata(dev);
+       u64 v;
+
+       v = readq(feature->ioaddr + FME_THERM_THRESHOLD);
+
+       return sprintf(buf, "%u\n",
+                      (unsigned int)FIELD_GET(TEMP_THRESHOLD1_POLICY, v));
+}
+
+static DEVICE_ATTR_RO(temp1_max_policy);
+
+static struct attribute *thermal_extra_attrs[] = {
+       &dev_attr_temp1_max_policy.attr,
+       NULL,
+};
+
+static umode_t thermal_extra_attrs_visible(struct kobject *kobj,
+                                          struct attribute *attr, int index)
+{
+       struct device *dev = kobj_to_dev(kobj);
+       struct dfl_feature *feature = dev_get_drvdata(dev);
+
+       return fme_thermal_throttle_support(feature->ioaddr) ? attr->mode : 0;
+}
+
+static const struct attribute_group thermal_extra_group = {
+       .attrs          = thermal_extra_attrs,
+       .is_visible     = thermal_extra_attrs_visible,
+};
+__ATTRIBUTE_GROUPS(thermal_extra);
+
+static int fme_thermal_mgmt_init(struct platform_device *pdev,
+                                struct dfl_feature *feature)
+{
+       struct device *hwmon;
+
+       /*
+        * create hwmon to allow userspace monitoring temperature and other
+        * threshold information.
+        *
+        * temp1_input      -> FPGA device temperature
+        * temp1_max        -> hardware threshold 1 -> 50% or 90% throttling
+        * temp1_crit       -> hardware threshold 2 -> 100% throttling
+        * temp1_emergency  -> hardware trip_threshold to shutdown FPGA
+        * temp1_max_alarm  -> hardware threshold 1 alarm
+        * temp1_crit_alarm -> hardware threshold 2 alarm
+        *
+        * create device specific sysfs interfaces, e.g. read temp1_max_policy
+        * to understand the actual hardware throttling action (50% vs 90%).
+        *
+        * If hardware doesn't support automatic throttling per thresholds,
+        * then all above sysfs interfaces are not visible except temp1_input
+        * for temperature.
+        */
+       hwmon = devm_hwmon_device_register_with_info(&pdev->dev,
+                                                    "dfl_fme_thermal", feature,
+                                                    &thermal_hwmon_chip_info,
+                                                    thermal_extra_groups);
+       if (IS_ERR(hwmon)) {
+               dev_err(&pdev->dev, "Fail to register thermal hwmon\n");
+               return PTR_ERR(hwmon);
+       }
+
+       return 0;
+}
+
+static const struct dfl_feature_id fme_thermal_mgmt_id_table[] = {
+       {.id = FME_FEATURE_ID_THERMAL_MGMT,},
+       {0,}
+};
+
+static const struct dfl_feature_ops fme_thermal_mgmt_ops = {
+       .init = fme_thermal_mgmt_init,
+};
+
 static struct dfl_feature_driver fme_feature_drvs[] = {
        {
                .id_table = fme_hdr_id_table,
@@ -194,6 +368,10 @@ static struct dfl_feature_driver fme_feature_drvs[] = {
                .id_table = fme_global_err_id_table,
                .ops = &fme_global_err_ops,
        },
+       {
+               .id_table = fme_thermal_mgmt_id_table,
+               .ops = &fme_thermal_mgmt_ops,
+       },
        {
                .ops = NULL,
        },