On 11/9/2022 6:37 AM, Dixit, Ashutosh wrote:
On Thu, 03 Nov 2022 05:37:23 -0700, Sujaritha Sundaresan wrote:
Hi Suja,
Adding the rapl_pl1_freq_mhz sysfs attribute.
Signed-off-by: Sujaritha Sundaresan <sujaritha.sundaresan@xxxxxxxxx>
Cc: Ashutosh Dixit <ashutosh.dixit@xxxxxxxxx>
---
drivers/gpu/drm/i915/gt/intel_gt_sysfs_pm.c | 20 ++++++++++
drivers/gpu/drm/i915/gt/intel_rps.c | 44 +++++++++++++++++++++
drivers/gpu/drm/i915/gt/intel_rps.h | 3 ++
drivers/gpu/drm/i915/i915_reg.h | 4 ++
4 files changed, 71 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_sysfs_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_sysfs_pm.c
index 904160952369..e7f00ec252f8 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_sysfs_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt_sysfs_pm.c
@@ -496,6 +496,17 @@ static DEVICE_ATTR_RO(vlv_rpe_freq_mhz);
static const struct attribute * const gen6_rps_attrs[] = GEN6_RPS_ATTR;
static const struct attribute * const gen6_gt_attrs[] = GEN6_GT_ATTR;
+static ssize_t rapl_pl1_freq_mhz_show(struct device *dev,
+ struct device_attribute *attr,
+ char *buff)
+{
+ struct intel_gt *gt = intel_gt_sysfs_get_drvdata(dev, attr->attr.name);
+ u32 rapl_pl1 = intel_rps_read_rapl_pl1_frequency(>->rps);
+
+ return sysfs_emit(buff, "%u\n", rapl_pl1);
+}
+
+
static ssize_t punit_req_freq_mhz_show(struct device *dev,
struct device_attribute *attr,
char *buff)
@@ -534,6 +545,7 @@ struct intel_gt_bool_throttle_attr attr_##sysfs_func__ = { \
.mask = mask__, \
}
+static DEVICE_ATTR_RO(rapl_pl1_freq_mhz);
static DEVICE_ATTR_RO(punit_req_freq_mhz);
Is this patch against old code? Since this is now INTEL_GT_ATTR_RO. Yes the
build failed. So rapl_pl1_freq_mhz will need to follow punit_req_freq_mhz.
Okay yes looks like I might not have grabbed the latest tree.
static INTEL_GT_RPS_BOOL_ATTR_RO(throttle_reason_status, GT0_PERF_LIMIT_REASONS_MASK);
static INTEL_GT_RPS_BOOL_ATTR_RO(throttle_reason_pl1, POWER_LIMIT_1_MASK);
@@ -790,12 +802,20 @@ void intel_gt_sysfs_pm_init(struct intel_gt *gt, struct kobject *kobj)
if (!is_object_gt(kobj))
return;
+ ret = sysfs_create_file(kobj, &dev_attr_rapl_pl1_freq_mhz.attr);
The convention here is to create sysfs files only for platforms on which a
feature (in this case RAPL PL1 freq) is supported.
Also are we sure this is only available on MTL and XEHPSDV and not on DG2?
Since generally a feature appears first on a platform and then is available
for all successive products. If it's available on DG2 too then we can use
something like:
if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 50))
See GRAPHICS_VER_FULL for various platforms in i915_pci.c.
I will check again for DG2.
+ if (ret)
+ drm_warn(>->i915->drm,
+ "failed to create gt%u rapl_pl1_freq_mhz sysfs(%pe)",
+ gt->info.id, ERR_PTR(ret));
+
+
ret = sysfs_create_file(kobj, &dev_attr_punit_req_freq_mhz.attr);
if (ret)
drm_warn(>->i915->drm,
"failed to create gt%u punit_req_freq_mhz sysfs (%pe)",
gt->info.id, ERR_PTR(ret));
+
Remove empty line.
if (i915_mmio_reg_valid(intel_gt_perf_limit_reasons_reg(gt))) {
ret = sysfs_create_files(kobj, throttle_reason_attrs);
if (ret)
diff --git a/drivers/gpu/drm/i915/gt/intel_rps.c b/drivers/gpu/drm/i915/gt/intel_rps.c
index 17b40b625e31..0e89b941e3be 100644
--- a/drivers/gpu/drm/i915/gt/intel_rps.c
+++ b/drivers/gpu/drm/i915/gt/intel_rps.c
@@ -9,6 +9,7 @@
#include "i915_drv.h"
#include "i915_irq.h"
+#include "i915_reg.h"
Not needed, see below.
#include "intel_breadcrumbs.h"
#include "intel_gt.h"
#include "intel_gt_clock_utils.h"
@@ -2422,6 +2423,49 @@ bool rps_read_mask_mmio(struct intel_rps *rps,
return rps_read_mmio(rps, reg32) & mask;
}
+u32 intel_rps_read_rapl_pl1(struct intel_rps *rps)
+{
+ struct drm_i915_private *i915 = rps_to_i915(rps);
+ i915_reg_t rgadr;
+ u32 rapl_pl1;
+
+ if (IS_METEORLAKE(i915)) {
+ rgadr = MTL_RAPL_PL1_FREQ_LIMIT;
+ } else if (IS_XEHPSDV(i915)) {
+ rgadr = XEHPSDV_RAPL_PL1_FREQ_LIMIT;
+ } else {
+ MISSING_CASE(GRAPHICS_VER(i915));
+ rgadr = INVALID_MMIO_REG;
No need for this, the sysfs file will only be visible for platforms on
which this is supported so this will never be hit.
+ }
+
+ if (!i915_mmio_reg_valid(rgadr))
+ rapl_pl1 = 0;
No need for this either.
+ else
+ rapl_pl1 = rps_read_mmio(rps, rgadr);
+
+ return rapl_pl1;
+}
+
+u32 intel_rps_get_rapl(struct intel_rps *rps, u32 rapl_pl1)
+{
+ struct drm_i915_private *i915 = rps_to_i915(rps);
+ u32 rapl = 0;
+
+ if (IS_METEORLAKE(i915) || IS_XEHPSDV(i915))
+ rapl = rapl_pl1 & RAPL_PL1_FREQ_LIMIT_MASK;
+ else
+ MISSING_CASE(GRAPHICS_VER(i915));
No need for this either.
+
+ return rapl;
+}
+
+u32 intel_rps_read_rapl_pl1_frequency(struct intel_rps *rps)
+{
+ u32 rapl_freq = intel_rps_get_rapl(rps, intel_rps_read_rapl_pl1(rps));
+
+ return (rapl_freq >> 8) * GT_FREQUENCY_MULTIPLIER;
Use REG_FIELD_GET to extract the freq.
There doesn't seem any need to have 3 functions here, let's combine them
into a single function and use REG_FIELD_GET.
Got it. Will fix this
+}
+
/* External interface for intel_ips.ko */
static struct drm_i915_private __rcu *ips_mchdev;
diff --git a/drivers/gpu/drm/i915/gt/intel_rps.h b/drivers/gpu/drm/i915/gt/intel_rps.h
index 4509dfdc52e0..4adc6aaedba0 100644
--- a/drivers/gpu/drm/i915/gt/intel_rps.h
+++ b/drivers/gpu/drm/i915/gt/intel_rps.h
@@ -34,6 +34,7 @@ void intel_rps_mark_interactive(struct intel_rps *rps, bool interactive);
int intel_gpu_freq(struct intel_rps *rps, int val);
int intel_freq_opcode(struct intel_rps *rps, int val);
u32 intel_rps_get_cagf(struct intel_rps *rps, u32 rpstat1);
+u32 intel_rps_get_rapl(struct intel_rps *rps, u32 rapl_pl1);
u32 intel_rps_read_actual_frequency(struct intel_rps *rps);
u32 intel_rps_get_requested_frequency(struct intel_rps *rps);
u32 intel_rps_get_min_frequency(struct intel_rps *rps);
@@ -47,6 +48,8 @@ u32 intel_rps_get_rp1_frequency(struct intel_rps *rps);
u32 intel_rps_get_rpn_frequency(struct intel_rps *rps);
u32 intel_rps_read_punit_req(struct intel_rps *rps);
u32 intel_rps_read_punit_req_frequency(struct intel_rps *rps);
+u32 intel_rps_read_rapl_pl1(struct intel_rps *rps);
+u32 intel_rps_read_rapl_pl1_frequency(struct intel_rps *rps);
These should be static and not in the .h file. Anyway I think as pointed
out above these 3 functions should probably be combined into just one.
void gen6_rps_get_freq_caps(struct intel_rps *rps, struct intel_rps_freq_caps *caps);
void intel_rps_raise_unslice(struct intel_rps *rps);
void intel_rps_lower_unslice(struct intel_rps *rps);
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 5003a5ffbc6a..68ff98e27b8d 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -1818,6 +1818,10 @@
#define GT0_PERF_LIMIT_REASONS_LOG_MASK REG_GENMASK(31, 16)
#define MTL_MEDIA_PERF_LIMIT_REASONS _MMIO(0x138030)
+#define XEHPSDV_RAPL_PL1_FREQ_LIMIT _MMIO(0x250070)
+#define MTL_RAPL_PL1_FREQ_LIMIT _MMIO(0x281070)
+#define RAPL_PL1_FREQ_LIMIT_MASK 0xffff
Use REG_GENMASK(15, 0) here.
Also:
* The registers now need to be sorted by offset. Let's move these #defines
to gt/intel_gt_regs.h
* RAPL_PL1_FREQ_LIMIT_MASK will need to be indented a couple of spaces as
done for other registers
Sure will move it to that file.
#define CHV_CLK_CTL1 _MMIO(0x101100)
#define VLV_CLK_CTL2 _MMIO(0x101104)
#define CLK_CTL2_CZCOUNT_30NS_SHIFT 28
--
2.34.1
Thanks.
--
Ashutosh
Thank you for the review
-Suja
