drm/i915: Extract GT ring management

Although the ring management is much smaller compared to the other GT
power management functions, continue the theme of extracting it out of
the huge intel_pm.c for maintenance.

Based on a patch by Chris Wilson.

Signed-off-by: Andi Shyti <andi.shyti@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20191020184139.9145-1-chris@chris-wilson.co.uk
This commit is contained in:
Andi Shyti 2019-10-20 19:41:39 +01:00 committed by Chris Wilson
parent 8814c6d01f
commit 0dc3c562aa
10 changed files with 297 additions and 120 deletions

View file

@ -87,6 +87,7 @@ gt-y += \
gt/intel_gt_pm_irq.o \
gt/intel_gt_requests.o \
gt/intel_hangcheck.o \
gt/intel_llc.o \
gt/intel_lrc.o \
gt/intel_rc6.o \
gt/intel_renderstate.o \

View file

@ -17,6 +17,7 @@
#include "i915_vma.h"
#include "intel_engine_types.h"
#include "intel_llc_types.h"
#include "intel_reset_types.h"
#include "intel_rc6_types.h"
#include "intel_wakeref.h"
@ -79,6 +80,7 @@ struct intel_gt {
*/
intel_wakeref_t awake;
struct intel_llc llc;
struct intel_rc6 rc6;
struct blocking_notifier_head pm_notifications;

View file

@ -0,0 +1,161 @@
/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2019 Intel Corporation
*/
#include <linux/cpufreq.h>
#include "i915_drv.h"
#include "intel_gt.h"
#include "intel_llc.h"
#include "intel_sideband.h"
struct ia_constants {
unsigned int min_gpu_freq;
unsigned int max_gpu_freq;
unsigned int min_ring_freq;
unsigned int max_ia_freq;
};
static struct intel_gt *llc_to_gt(struct intel_llc *llc)
{
return container_of(llc, struct intel_gt, llc);
}
static unsigned int cpu_max_MHz(void)
{
struct cpufreq_policy *policy;
unsigned int max_khz;
policy = cpufreq_cpu_get(0);
if (policy) {
max_khz = policy->cpuinfo.max_freq;
cpufreq_cpu_put(policy);
} else {
/*
* Default to measured freq if none found, PCU will ensure we
* don't go over
*/
max_khz = tsc_khz;
}
return max_khz / 1000;
}
static bool get_ia_constants(struct intel_llc *llc,
struct ia_constants *consts)
{
struct drm_i915_private *i915 = llc_to_gt(llc)->i915;
struct intel_rps *rps = &i915->gt_pm.rps;
if (rps->max_freq <= rps->min_freq)
return false;
consts->max_ia_freq = cpu_max_MHz();
consts->min_ring_freq =
intel_uncore_read(llc_to_gt(llc)->uncore, DCLK) & 0xf;
/* convert DDR frequency from units of 266.6MHz to bandwidth */
consts->min_ring_freq = mult_frac(consts->min_ring_freq, 8, 3);
consts->min_gpu_freq = rps->min_freq;
consts->max_gpu_freq = rps->max_freq;
if (INTEL_GEN(i915) >= 9) {
/* Convert GT frequency to 50 HZ units */
consts->min_gpu_freq /= GEN9_FREQ_SCALER;
consts->max_gpu_freq /= GEN9_FREQ_SCALER;
}
return true;
}
static void calc_ia_freq(struct intel_llc *llc,
unsigned int gpu_freq,
const struct ia_constants *consts,
unsigned int *out_ia_freq,
unsigned int *out_ring_freq)
{
struct drm_i915_private *i915 = llc_to_gt(llc)->i915;
const int diff = consts->max_gpu_freq - gpu_freq;
unsigned int ia_freq = 0, ring_freq = 0;
if (INTEL_GEN(i915) >= 9) {
/*
* ring_freq = 2 * GT. ring_freq is in 100MHz units
* No floor required for ring frequency on SKL.
*/
ring_freq = gpu_freq;
} else if (INTEL_GEN(i915) >= 8) {
/* max(2 * GT, DDR). NB: GT is 50MHz units */
ring_freq = max(consts->min_ring_freq, gpu_freq);
} else if (IS_HASWELL(i915)) {
ring_freq = mult_frac(gpu_freq, 5, 4);
ring_freq = max(consts->min_ring_freq, ring_freq);
/* leave ia_freq as the default, chosen by cpufreq */
} else {
const int min_freq = 15;
const int scale = 180;
/*
* On older processors, there is no separate ring
* clock domain, so in order to boost the bandwidth
* of the ring, we need to upclock the CPU (ia_freq).
*
* For GPU frequencies less than 750MHz,
* just use the lowest ring freq.
*/
if (gpu_freq < min_freq)
ia_freq = 800;
else
ia_freq = consts->max_ia_freq - diff * scale / 2;
ia_freq = DIV_ROUND_CLOSEST(ia_freq, 100);
}
*out_ia_freq = ia_freq;
*out_ring_freq = ring_freq;
}
static void gen6_update_ring_freq(struct intel_llc *llc)
{
struct drm_i915_private *i915 = llc_to_gt(llc)->i915;
struct ia_constants consts;
unsigned int gpu_freq;
if (!get_ia_constants(llc, &consts))
return;
/*
* For each potential GPU frequency, load a ring frequency we'd like
* to use for memory access. We do this by specifying the IA frequency
* the PCU should use as a reference to determine the ring frequency.
*/
for (gpu_freq = consts.max_gpu_freq;
gpu_freq >= consts.min_gpu_freq;
gpu_freq--) {
unsigned int ia_freq, ring_freq;
calc_ia_freq(llc, gpu_freq, &consts, &ia_freq, &ring_freq);
sandybridge_pcode_write(i915,
GEN6_PCODE_WRITE_MIN_FREQ_TABLE,
ia_freq << GEN6_PCODE_FREQ_IA_RATIO_SHIFT |
ring_freq << GEN6_PCODE_FREQ_RING_RATIO_SHIFT |
gpu_freq);
}
}
void intel_llc_enable(struct intel_llc *llc)
{
if (HAS_LLC(llc_to_gt(llc)->i915))
gen6_update_ring_freq(llc);
}
void intel_llc_disable(struct intel_llc *llc)
{
/* Currently there is no HW configuration to be done to disable. */
}
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftest_llc.c"
#endif

View file

@ -0,0 +1,15 @@
/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2019 Intel Corporation
*/
#ifndef INTEL_LLC_H
#define INTEL_LLC_H
struct intel_llc;
void intel_llc_enable(struct intel_llc *llc);
void intel_llc_disable(struct intel_llc *llc);
#endif /* INTEL_LLC_H */

View file

@ -0,0 +1,13 @@
/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2019 Intel Corporation
*/
#ifndef INTEL_LLC_TYPES_H
#define INTEL_LLC_TYPES_H
struct intel_llc {
};
#endif /* INTEL_LLC_TYPES_H */

View file

@ -5,6 +5,8 @@
* Copyright © 2019 Intel Corporation
*/
#include "selftest_llc.h"
static int live_gt_resume(void *arg)
{
struct intel_gt *gt = arg;
@ -32,6 +34,13 @@ static int live_gt_resume(void *arg)
err = -EINVAL;
break;
}
err = st_llc_verify(&gt->llc);
if (err) {
pr_err("llc state not restored upon resume!\n");
intel_gt_set_wedged_on_init(gt);
break;
}
} while (!__igt_timeout(end_time, NULL));
return err;

View file

@ -0,0 +1,77 @@
/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2019 Intel Corporation
*/
#include "intel_pm.h" /* intel_gpu_freq() */
#include "selftest_llc.h"
static int gen6_verify_ring_freq(struct intel_llc *llc)
{
struct drm_i915_private *i915 = llc_to_gt(llc)->i915;
struct ia_constants consts;
intel_wakeref_t wakeref;
unsigned int gpu_freq;
int err = 0;
wakeref = intel_runtime_pm_get(llc_to_gt(llc)->uncore->rpm);
if (!get_ia_constants(llc, &consts)) {
err = -ENODEV;
goto out_rpm;
}
for (gpu_freq = consts.min_gpu_freq;
gpu_freq <= consts.max_gpu_freq;
gpu_freq++) {
unsigned int ia_freq, ring_freq, found;
u32 val;
calc_ia_freq(llc, gpu_freq, &consts, &ia_freq, &ring_freq);
val = gpu_freq;
if (sandybridge_pcode_read(i915,
GEN6_PCODE_READ_MIN_FREQ_TABLE,
&val, NULL)) {
pr_err("Failed to read freq table[%d], range [%d, %d]\n",
gpu_freq, consts.min_gpu_freq, consts.max_gpu_freq);
err = -ENXIO;
break;
}
found = (val >> 0) & 0xff;
if (found != ia_freq) {
pr_err("Min freq table(%d/[%d, %d]):%dMHz did not match expected CPU freq, found %d, expected %d\n",
gpu_freq, consts.min_gpu_freq, consts.max_gpu_freq,
intel_gpu_freq(i915, gpu_freq * (INTEL_GEN(i915) >= 9 ? GEN9_FREQ_SCALER : 1)),
found, ia_freq);
err = -EINVAL;
break;
}
found = (val >> 8) & 0xff;
if (found != ring_freq) {
pr_err("Min freq table(%d/[%d, %d]):%dMHz did not match expected ring freq, found %d, expected %d\n",
gpu_freq, consts.min_gpu_freq, consts.max_gpu_freq,
intel_gpu_freq(i915, gpu_freq * (INTEL_GEN(i915) >= 9 ? GEN9_FREQ_SCALER : 1)),
found, ring_freq);
err = -EINVAL;
break;
}
}
out_rpm:
intel_runtime_pm_put(llc_to_gt(llc)->uncore->rpm, wakeref);
return err;
}
int st_llc_verify(struct intel_llc *llc)
{
int err = 0;
if (HAS_LLC(llc_to_gt(llc)->i915))
err = gen6_verify_ring_freq(llc);
return err;
}

View file

@ -0,0 +1,14 @@
/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2019 Intel Corporation
*/
#ifndef SELFTEST_LLC_H
#define SELFTEST_LLC_H
struct intel_llc;
int st_llc_verify(struct intel_llc *llc);
#endif /* SELFTEST_LLC_H */

View file

@ -605,13 +605,8 @@ struct intel_rps {
struct intel_rps_ei ei;
};
struct intel_llc_pstate {
bool enabled;
};
struct intel_gen6_power_mgmt {
struct intel_rps rps;
struct intel_llc_pstate llc_pstate;
};
/* defined intel_pm.c */

View file

@ -25,7 +25,6 @@
*
*/
#include <linux/cpufreq.h>
#include <linux/module.h>
#include <linux/pm_runtime.h>
@ -38,6 +37,8 @@
#include "display/intel_fbc.h"
#include "display/intel_sprite.h"
#include "gt/intel_llc.h"
#include "i915_drv.h"
#include "i915_irq.h"
#include "i915_trace.h"
@ -7030,93 +7031,6 @@ static void gen6_enable_rps(struct drm_i915_private *dev_priv)
intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL);
}
static void gen6_update_ring_freq(struct drm_i915_private *dev_priv)
{
struct intel_rps *rps = &dev_priv->gt_pm.rps;
const int min_freq = 15;
const int scaling_factor = 180;
unsigned int gpu_freq;
unsigned int max_ia_freq, min_ring_freq;
unsigned int max_gpu_freq, min_gpu_freq;
struct cpufreq_policy *policy;
lockdep_assert_held(&rps->lock);
if (rps->max_freq <= rps->min_freq)
return;
policy = cpufreq_cpu_get(0);
if (policy) {
max_ia_freq = policy->cpuinfo.max_freq;
cpufreq_cpu_put(policy);
} else {
/*
* Default to measured freq if none found, PCU will ensure we
* don't go over
*/
max_ia_freq = tsc_khz;
}
/* Convert from kHz to MHz */
max_ia_freq /= 1000;
min_ring_freq = I915_READ(DCLK) & 0xf;
/* convert DDR frequency from units of 266.6MHz to bandwidth */
min_ring_freq = mult_frac(min_ring_freq, 8, 3);
min_gpu_freq = rps->min_freq;
max_gpu_freq = rps->max_freq;
if (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) {
/* Convert GT frequency to 50 HZ units */
min_gpu_freq /= GEN9_FREQ_SCALER;
max_gpu_freq /= GEN9_FREQ_SCALER;
}
/*
* For each potential GPU frequency, load a ring frequency we'd like
* to use for memory access. We do this by specifying the IA frequency
* the PCU should use as a reference to determine the ring frequency.
*/
for (gpu_freq = max_gpu_freq; gpu_freq >= min_gpu_freq; gpu_freq--) {
const int diff = max_gpu_freq - gpu_freq;
unsigned int ia_freq = 0, ring_freq = 0;
if (IS_GEN9_BC(dev_priv) || INTEL_GEN(dev_priv) >= 10) {
/*
* ring_freq = 2 * GT. ring_freq is in 100MHz units
* No floor required for ring frequency on SKL.
*/
ring_freq = gpu_freq;
} else if (INTEL_GEN(dev_priv) >= 8) {
/* max(2 * GT, DDR). NB: GT is 50MHz units */
ring_freq = max(min_ring_freq, gpu_freq);
} else if (IS_HASWELL(dev_priv)) {
ring_freq = mult_frac(gpu_freq, 5, 4);
ring_freq = max(min_ring_freq, ring_freq);
/* leave ia_freq as the default, chosen by cpufreq */
} else {
/* On older processors, there is no separate ring
* clock domain, so in order to boost the bandwidth
* of the ring, we need to upclock the CPU (ia_freq).
*
* For GPU frequencies less than 750MHz,
* just use the lowest ring freq.
*/
if (gpu_freq < min_freq)
ia_freq = 800;
else
ia_freq = max_ia_freq - ((diff * scaling_factor) / 2);
ia_freq = DIV_ROUND_CLOSEST(ia_freq, 100);
}
sandybridge_pcode_write(dev_priv,
GEN6_PCODE_WRITE_MIN_FREQ_TABLE,
ia_freq << GEN6_PCODE_FREQ_IA_RATIO_SHIFT |
ring_freq << GEN6_PCODE_FREQ_RING_RATIO_SHIFT |
gpu_freq);
}
}
static int cherryview_rps_max_freq(struct drm_i915_private *dev_priv)
{
u32 val, rp0;
@ -7965,18 +7879,6 @@ void intel_sanitize_gt_powersave(struct drm_i915_private *dev_priv)
gen6_reset_rps_interrupts(dev_priv);
}
static inline void intel_disable_llc_pstate(struct drm_i915_private *i915)
{
lockdep_assert_held(&i915->gt_pm.rps.lock);
if (!i915->gt_pm.llc_pstate.enabled)
return;
/* Currently there is no HW configuration to be done to disable. */
i915->gt_pm.llc_pstate.enabled = false;
}
static void intel_disable_rps(struct drm_i915_private *dev_priv)
{
lockdep_assert_held(&dev_priv->gt_pm.rps.lock);
@ -8004,23 +7906,11 @@ void intel_disable_gt_powersave(struct drm_i915_private *dev_priv)
intel_disable_rps(dev_priv);
if (HAS_LLC(dev_priv))
intel_disable_llc_pstate(dev_priv);
intel_llc_disable(&dev_priv->gt.llc);
mutex_unlock(&dev_priv->gt_pm.rps.lock);
}
static inline void intel_enable_llc_pstate(struct drm_i915_private *i915)
{
lockdep_assert_held(&i915->gt_pm.rps.lock);
if (i915->gt_pm.llc_pstate.enabled)
return;
gen6_update_ring_freq(i915);
i915->gt_pm.llc_pstate.enabled = true;
}
static void intel_enable_rps(struct drm_i915_private *dev_priv)
{
struct intel_rps *rps = &dev_priv->gt_pm.rps;
@ -8064,8 +7954,8 @@ void intel_enable_gt_powersave(struct drm_i915_private *dev_priv)
if (HAS_RPS(dev_priv))
intel_enable_rps(dev_priv);
if (HAS_LLC(dev_priv))
intel_enable_llc_pstate(dev_priv);
intel_llc_enable(&dev_priv->gt.llc);
mutex_unlock(&dev_priv->gt_pm.rps.lock);
}