@@ -40,6 +40,7 @@ config DRM_I915_DEBUG
select DRM_I915_DEBUG_MMIO
select DRM_I915_TRACK_WAKEREF
select DRM_I915_DEBUG_RUNTIME_PM
+ select DRM_I915_DEBUG_WAKEREF
select DRM_I915_SW_FENCE_DEBUG_OBJECTS
select DRM_I915_SELFTEST
select BROKEN # for prototype uAPI
@@ -249,3 +250,17 @@ config DRM_I915_DEBUG_RUNTIME_PM
Recommended for driver developers only.
If in doubt, say "N"
+
+config DRM_I915_DEBUG_WAKEREF
+ bool "Enable extra tracking for wakerefs"
+ depends on DRM_I915
+ default n
+ select STACKDEPOT
+ select STACKTRACE
+ select DRM_I915_TRACK_WAKEREF
+ help
+ Choose this option to turn on extra state checking and usage
+ tracking for the wakerefPM functionality. This may introduce
+ overhead during driver runtime.
+
+ If in doubt, say "N"
@@ -253,6 +253,7 @@ struct i915_execbuffer {
struct intel_gt *gt; /* gt for the execbuf */
struct intel_context *context; /* logical state for the request */
struct i915_gem_context *gem_context; /** caller's context */
+ intel_wakeref_t wakeref;
/** our requests to build */
struct i915_request *requests[MAX_ENGINE_INSTANCE + 1];
@@ -2713,7 +2714,7 @@ eb_select_engine(struct i915_execbuffer *eb)
for_each_child(ce, child)
intel_context_get(child);
- intel_gt_pm_get(ce->engine->gt);
+ eb->wakeref = intel_gt_pm_get(ce->engine->gt);
if (!test_bit(CONTEXT_ALLOC_BIT, &ce->flags)) {
err = intel_context_alloc_state(ce);
@@ -2752,7 +2753,7 @@ eb_select_engine(struct i915_execbuffer *eb)
return err;
err:
- intel_gt_pm_put(ce->engine->gt);
+ intel_gt_pm_put(ce->engine->gt, eb->wakeref);
for_each_child(ce, child)
intel_context_put(child);
intel_context_put(ce);
@@ -2765,7 +2766,7 @@ eb_put_engine(struct i915_execbuffer *eb)
struct intel_context *child;
i915_vm_put(eb->context->vm);
- intel_gt_pm_put(eb->gt);
+ intel_gt_pm_put(eb->context->engine->gt, eb->wakeref);
for_each_child(eb->context, child)
intel_context_put(child);
intel_context_put(eb->context);
@@ -85,6 +85,7 @@ static int cpu_get(struct context *ctx, unsigned long offset, u32 *v)
static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
{
+ intel_wakeref_t wakeref;
struct i915_vma *vma;
u32 __iomem *map;
int err = 0;
@@ -99,7 +100,7 @@ static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
if (IS_ERR(vma))
return PTR_ERR(vma);
- intel_gt_pm_get(vma->vm->gt);
+ wakeref = intel_gt_pm_get(vma->vm->gt);
map = i915_vma_pin_iomap(vma);
i915_vma_unpin(vma);
@@ -112,12 +113,13 @@ static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
i915_vma_unpin_iomap(vma);
out_rpm:
- intel_gt_pm_put(vma->vm->gt);
+ intel_gt_pm_put(vma->vm->gt, wakeref);
return err;
}
static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
{
+ intel_wakeref_t wakeref;
struct i915_vma *vma;
u32 __iomem *map;
int err = 0;
@@ -132,7 +134,7 @@ static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
if (IS_ERR(vma))
return PTR_ERR(vma);
- intel_gt_pm_get(vma->vm->gt);
+ wakeref = intel_gt_pm_get(vma->vm->gt);
map = i915_vma_pin_iomap(vma);
i915_vma_unpin(vma);
@@ -145,7 +147,7 @@ static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
i915_vma_unpin_iomap(vma);
out_rpm:
- intel_gt_pm_put(vma->vm->gt);
+ intel_gt_pm_put(vma->vm->gt, wakeref);
return err;
}
@@ -630,14 +630,14 @@ static bool assert_mmap_offset(struct drm_i915_private *i915,
static void disable_retire_worker(struct drm_i915_private *i915)
{
i915_gem_driver_unregister__shrinker(i915);
- intel_gt_pm_get(to_gt(i915));
+ intel_gt_pm_get_untracked(to_gt(i915));
cancel_delayed_work_sync(&to_gt(i915)->requests.retire_work);
}
static void restore_retire_worker(struct drm_i915_private *i915)
{
igt_flush_test(i915);
- intel_gt_pm_put(to_gt(i915));
+ intel_gt_pm_put_untracked(to_gt(i915));
i915_gem_driver_register__shrinker(i915);
}
@@ -778,6 +778,7 @@ static int igt_mmap_offset_exhaustion(void *arg)
static int gtt_set(struct drm_i915_gem_object *obj)
{
+ intel_wakeref_t wakeref;
struct i915_vma *vma;
void __iomem *map;
int err = 0;
@@ -786,7 +787,7 @@ static int gtt_set(struct drm_i915_gem_object *obj)
if (IS_ERR(vma))
return PTR_ERR(vma);
- intel_gt_pm_get(vma->vm->gt);
+ wakeref = intel_gt_pm_get(vma->vm->gt);
map = i915_vma_pin_iomap(vma);
i915_vma_unpin(vma);
if (IS_ERR(map)) {
@@ -798,12 +799,13 @@ static int gtt_set(struct drm_i915_gem_object *obj)
i915_vma_unpin_iomap(vma);
out:
- intel_gt_pm_put(vma->vm->gt);
+ intel_gt_pm_put(vma->vm->gt, wakeref);
return err;
}
static int gtt_check(struct drm_i915_gem_object *obj)
{
+ intel_wakeref_t wakeref;
struct i915_vma *vma;
void __iomem *map;
int err = 0;
@@ -812,7 +814,7 @@ static int gtt_check(struct drm_i915_gem_object *obj)
if (IS_ERR(vma))
return PTR_ERR(vma);
- intel_gt_pm_get(vma->vm->gt);
+ wakeref = intel_gt_pm_get(vma->vm->gt);
map = i915_vma_pin_iomap(vma);
i915_vma_unpin(vma);
if (IS_ERR(map)) {
@@ -828,7 +830,7 @@ static int gtt_check(struct drm_i915_gem_object *obj)
i915_vma_unpin_iomap(vma);
out:
- intel_gt_pm_put(vma->vm->gt);
+ intel_gt_pm_put(vma->vm->gt, wakeref);
return err;
}
@@ -28,11 +28,14 @@ static void irq_disable(struct intel_breadcrumbs *b)
static void __intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b)
{
+ intel_wakeref_t wakeref;
+
/*
* Since we are waiting on a request, the GPU should be busy
* and should have its own rpm reference.
*/
- if (GEM_WARN_ON(!intel_gt_pm_get_if_awake(b->irq_engine->gt)))
+ wakeref = intel_gt_pm_get_if_awake(b->irq_engine->gt);
+ if (GEM_WARN_ON(!wakeref))
return;
/*
@@ -41,7 +44,7 @@ static void __intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b)
* which we can add a new waiter and avoid the cost of re-enabling
* the irq.
*/
- WRITE_ONCE(b->irq_armed, true);
+ WRITE_ONCE(b->irq_armed, wakeref);
/* Requests may have completed before we could enable the interrupt. */
if (!b->irq_enabled++ && b->irq_enable(b))
@@ -61,12 +64,14 @@ static void intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b)
static void __intel_breadcrumbs_disarm_irq(struct intel_breadcrumbs *b)
{
+ intel_wakeref_t wakeref = b->irq_armed;
+
GEM_BUG_ON(!b->irq_enabled);
if (!--b->irq_enabled)
b->irq_disable(b);
- WRITE_ONCE(b->irq_armed, false);
- intel_gt_pm_put_async(b->irq_engine->gt);
+ WRITE_ONCE(b->irq_armed, 0);
+ intel_gt_pm_put_async(b->irq_engine->gt, wakeref);
}
static void intel_breadcrumbs_disarm_irq(struct intel_breadcrumbs *b)
@@ -13,6 +13,7 @@
#include <linux/types.h>
#include "intel_engine_types.h"
+#include "intel_wakeref.h"
/*
* Rather than have every client wait upon all user interrupts,
@@ -43,7 +44,7 @@ struct intel_breadcrumbs {
spinlock_t irq_lock; /* protects the interrupt from hardirq context */
struct irq_work irq_work; /* for use from inside irq_lock */
unsigned int irq_enabled;
- bool irq_armed;
+ intel_wakeref_t irq_armed;
/* Not all breadcrumbs are attached to physical HW */
intel_engine_mask_t engine_mask;
@@ -63,7 +63,7 @@ static int __engine_unpark(struct intel_wakeref *wf)
ENGINE_TRACE(engine, "\n");
- intel_gt_pm_get(engine->gt);
+ engine->wakeref_track = intel_gt_pm_get(engine->gt);
/* Discard stale context state from across idling */
ce = engine->kernel_context;
@@ -276,7 +276,7 @@ static int __engine_park(struct intel_wakeref *wf)
engine->park(engine);
/* While gt calls i915_vma_parked(), we have to break the lock cycle */
- intel_gt_pm_put_async(engine->gt);
+ intel_gt_pm_put_async(engine->gt, engine->wakeref_track);
return 0;
}
@@ -431,7 +431,9 @@ struct intel_engine_cs {
unsigned long serial;
unsigned long wakeref_serial;
+ intel_wakeref_t wakeref_track;
struct intel_wakeref wakeref;
+
struct file *default_state;
struct {
@@ -630,7 +630,7 @@ static void __execlists_schedule_out(struct i915_request * const rq,
execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT);
if (engine->fw_domain && !--engine->fw_active)
intel_uncore_forcewake_put(engine->uncore, engine->fw_domain);
- intel_gt_pm_put_async(engine->gt);
+ intel_gt_pm_put_async_untracked(engine->gt);
/*
* If this is part of a virtual engine, its next request may
@@ -49,19 +49,20 @@ static void mtl_media_idle(struct intel_gt *gt)
static void user_forcewake(struct intel_gt *gt, bool suspend)
{
int count = atomic_read(>->user_wakeref);
+ intel_wakeref_t wakeref;
/* Inside suspend/resume so single threaded, no races to worry about. */
if (likely(!count))
return;
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
if (suspend) {
GEM_BUG_ON(count > atomic_read(>->wakeref.count));
atomic_sub(count, >->wakeref.count);
} else {
atomic_add(count, >->wakeref.count);
}
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
}
static void runtime_begin(struct intel_gt *gt)
@@ -248,6 +249,7 @@ int intel_gt_resume(struct intel_gt *gt)
{
struct intel_engine_cs *engine;
enum intel_engine_id id;
+ intel_wakeref_t wakeref;
int err;
err = intel_gt_has_unrecoverable_error(gt);
@@ -264,7 +266,7 @@ int intel_gt_resume(struct intel_gt *gt)
*/
gt_sanitize(gt, true);
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL);
intel_rc6_sanitize(>->rc6);
@@ -307,7 +309,7 @@ int intel_gt_resume(struct intel_gt *gt)
out_fw:
intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
return err;
err_wedged:
@@ -16,19 +16,28 @@ static inline bool intel_gt_pm_is_awake(const struct intel_gt *gt)
return intel_wakeref_is_active(>->wakeref);
}
-static inline void intel_gt_pm_get(struct intel_gt *gt)
+static inline void intel_gt_pm_get_untracked(struct intel_gt *gt)
{
intel_wakeref_get(>->wakeref);
}
+static inline intel_wakeref_t intel_gt_pm_get(struct intel_gt *gt)
+{
+ intel_gt_pm_get_untracked(gt);
+ return intel_wakeref_track(>->wakeref);
+}
+
static inline void __intel_gt_pm_get(struct intel_gt *gt)
{
__intel_wakeref_get(>->wakeref);
}
-static inline bool intel_gt_pm_get_if_awake(struct intel_gt *gt)
+static inline intel_wakeref_t intel_gt_pm_get_if_awake(struct intel_gt *gt)
{
- return intel_wakeref_get_if_active(>->wakeref);
+ if (!intel_wakeref_get_if_active(>->wakeref))
+ return 0;
+
+ return intel_wakeref_track(>->wakeref);
}
static inline void intel_gt_pm_might_get(struct intel_gt *gt)
@@ -36,12 +45,18 @@ static inline void intel_gt_pm_might_get(struct intel_gt *gt)
intel_wakeref_might_get(>->wakeref);
}
-static inline void intel_gt_pm_put(struct intel_gt *gt)
+static inline void intel_gt_pm_put_untracked(struct intel_gt *gt)
{
intel_wakeref_put(>->wakeref);
}
-static inline void intel_gt_pm_put_async(struct intel_gt *gt)
+static inline void intel_gt_pm_put(struct intel_gt *gt, intel_wakeref_t handle)
+{
+ intel_wakeref_untrack(>->wakeref, handle);
+ intel_gt_pm_put_untracked(gt);
+}
+
+static inline void intel_gt_pm_put_async_untracked(struct intel_gt *gt)
{
intel_wakeref_put_async(>->wakeref);
}
@@ -51,9 +66,14 @@ static inline void intel_gt_pm_might_put(struct intel_gt *gt)
intel_wakeref_might_put(>->wakeref);
}
-#define with_intel_gt_pm(gt, tmp) \
- for (tmp = 1, intel_gt_pm_get(gt); tmp; \
- intel_gt_pm_put(gt), tmp = 0)
+static inline void intel_gt_pm_put_async(struct intel_gt *gt, intel_wakeref_t handle)
+{
+ intel_wakeref_untrack(>->wakeref, handle);
+ intel_gt_pm_put_async_untracked(gt);
+}
+
+#define with_intel_gt_pm(gt, wf) \
+ for (wf = intel_gt_pm_get(gt); wf; intel_gt_pm_put(gt, wf), wf = 0)
/**
* with_intel_gt_pm_if_awake - if GT is PM awake, get a reference to prevent
@@ -64,7 +84,7 @@ static inline void intel_gt_pm_might_put(struct intel_gt *gt)
* @wf: pointer to a temporary wakeref.
*/
#define with_intel_gt_pm_if_awake(gt, wf) \
- for (wf = intel_gt_pm_get_if_awake(gt); wf; intel_gt_pm_put_async(gt), wf = 0)
+ for (wf = intel_gt_pm_get_if_awake(gt); wf; intel_gt_pm_put_async(gt, wf), wf = 0)
static inline int intel_gt_pm_wait_for_idle(struct intel_gt *gt)
{
@@ -27,7 +27,7 @@
void intel_gt_pm_debugfs_forcewake_user_open(struct intel_gt *gt)
{
atomic_inc(>->user_wakeref);
- intel_gt_pm_get(gt);
+ intel_gt_pm_get_untracked(gt);
if (GRAPHICS_VER(gt->i915) >= 6)
intel_uncore_forcewake_user_get(gt->uncore);
}
@@ -36,7 +36,7 @@ void intel_gt_pm_debugfs_forcewake_user_release(struct intel_gt *gt)
{
if (GRAPHICS_VER(gt->i915) >= 6)
intel_uncore_forcewake_user_put(gt->uncore);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put_untracked(gt);
atomic_dec(>->user_wakeref);
}
@@ -21,20 +21,22 @@ static int cmp_u32(const void *A, const void *B)
return *a - *b;
}
-static void perf_begin(struct intel_gt *gt)
+static intel_wakeref_t perf_begin(struct intel_gt *gt)
{
- intel_gt_pm_get(gt);
+ intel_wakeref_t wakeref = intel_gt_pm_get(gt);
/* Boost gpufreq to max [waitboost] and keep it fixed */
atomic_inc(>->rps.num_waiters);
schedule_work(>->rps.work);
flush_work(>->rps.work);
+
+ return wakeref;
}
-static int perf_end(struct intel_gt *gt)
+static int perf_end(struct intel_gt *gt, intel_wakeref_t wakeref)
{
atomic_dec(>->rps.num_waiters);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
return igt_flush_test(gt->i915);
}
@@ -133,12 +135,13 @@ static int perf_mi_bb_start(void *arg)
struct intel_gt *gt = arg;
struct intel_engine_cs *engine;
enum intel_engine_id id;
+ intel_wakeref_t wakeref;
int err = 0;
if (GRAPHICS_VER(gt->i915) < 4) /* Any CS_TIMESTAMP? */
return 0;
- perf_begin(gt);
+ wakeref = perf_begin(gt);
for_each_engine(engine, gt, id) {
struct intel_context *ce = engine->kernel_context;
struct i915_vma *batch;
@@ -207,7 +210,7 @@ static int perf_mi_bb_start(void *arg)
pr_info("%s: MI_BB_START cycles: %u\n",
engine->name, trifilter(cycles));
}
- if (perf_end(gt))
+ if (perf_end(gt, wakeref))
err = -EIO;
return err;
@@ -260,12 +263,13 @@ static int perf_mi_noop(void *arg)
struct intel_gt *gt = arg;
struct intel_engine_cs *engine;
enum intel_engine_id id;
+ intel_wakeref_t wakeref;
int err = 0;
if (GRAPHICS_VER(gt->i915) < 4) /* Any CS_TIMESTAMP? */
return 0;
- perf_begin(gt);
+ wakeref = perf_begin(gt);
for_each_engine(engine, gt, id) {
struct intel_context *ce = engine->kernel_context;
struct i915_vma *base, *nop;
@@ -364,7 +368,7 @@ static int perf_mi_noop(void *arg)
pr_info("%s: 16K MI_NOOP cycles: %u\n",
engine->name, trifilter(cycles));
}
- if (perf_end(gt))
+ if (perf_end(gt, wakeref))
err = -EIO;
return err;
@@ -81,6 +81,7 @@ static int live_gt_clocks(void *arg)
struct intel_gt *gt = arg;
struct intel_engine_cs *engine;
enum intel_engine_id id;
+ intel_wakeref_t wakeref;
int err = 0;
if (!gt->clock_frequency) { /* unknown */
@@ -91,7 +92,7 @@ static int live_gt_clocks(void *arg)
if (GRAPHICS_VER(gt->i915) < 4) /* Any CS_TIMESTAMP? */
return 0;
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL);
for_each_engine(engine, gt, id) {
@@ -128,7 +129,7 @@ static int live_gt_clocks(void *arg)
}
intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
return err;
}
@@ -257,11 +257,12 @@ static int igt_atomic_reset(void *arg)
{
struct intel_gt *gt = arg;
const typeof(*igt_atomic_phases) *p;
+ intel_wakeref_t wakeref;
int err = 0;
/* Check that the resets are usable from atomic context */
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
igt_global_reset_lock(gt);
/* Flush any requests before we get started and check basics */
@@ -292,7 +293,7 @@ static int igt_atomic_reset(void *arg)
unlock:
igt_global_reset_unlock(gt);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
return err;
}
@@ -303,6 +304,7 @@ static int igt_atomic_engine_reset(void *arg)
const typeof(*igt_atomic_phases) *p;
struct intel_engine_cs *engine;
enum intel_engine_id id;
+ intel_wakeref_t wakeref;
int err = 0;
/* Check that the resets are usable from atomic context */
@@ -313,7 +315,7 @@ static int igt_atomic_engine_reset(void *arg)
if (intel_uc_uses_guc_submission(>->uc))
return 0;
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
igt_global_reset_lock(gt);
/* Flush any requests before we get started and check basics */
@@ -361,7 +363,7 @@ static int igt_atomic_engine_reset(void *arg)
out_unlock:
igt_global_reset_unlock(gt);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
return err;
}
@@ -223,6 +223,7 @@ int live_rps_clock_interval(void *arg)
struct intel_engine_cs *engine;
enum intel_engine_id id;
struct igt_spinner spin;
+ intel_wakeref_t wakeref;
int err = 0;
if (!intel_rps_is_enabled(rps) || GRAPHICS_VER(gt->i915) < 6)
@@ -235,7 +236,7 @@ int live_rps_clock_interval(void *arg)
saved_work = rps->work.func;
rps->work.func = dummy_rps_work;
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
intel_rps_disable(>->rps);
intel_gt_check_clock_frequency(gt);
@@ -354,7 +355,7 @@ int live_rps_clock_interval(void *arg)
}
intel_rps_enable(>->rps);
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
igt_spinner_fini(&spin);
@@ -375,6 +376,7 @@ int live_rps_control(void *arg)
struct intel_engine_cs *engine;
enum intel_engine_id id;
struct igt_spinner spin;
+ intel_wakeref_t wakeref;
int err = 0;
/*
@@ -397,7 +399,7 @@ int live_rps_control(void *arg)
saved_work = rps->work.func;
rps->work.func = dummy_rps_work;
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
for_each_engine(engine, gt, id) {
struct i915_request *rq;
ktime_t min_dt, max_dt;
@@ -487,7 +489,7 @@ int live_rps_control(void *arg)
break;
}
}
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
igt_spinner_fini(&spin);
@@ -1022,6 +1024,7 @@ int live_rps_interrupt(void *arg)
struct intel_engine_cs *engine;
enum intel_engine_id id;
struct igt_spinner spin;
+ intel_wakeref_t wakeref;
u32 pm_events;
int err = 0;
@@ -1032,9 +1035,9 @@ int live_rps_interrupt(void *arg)
if (!intel_rps_has_interrupts(rps) || GRAPHICS_VER(gt->i915) < 6)
return 0;
- intel_gt_pm_get(gt);
- pm_events = rps->pm_events;
- intel_gt_pm_put(gt);
+ pm_events = 0;
+ with_intel_gt_pm(gt, wakeref)
+ pm_events = rps->pm_events;
if (!pm_events) {
pr_err("No RPS PM events registered, but RPS is enabled?\n");
return -ENODEV;
@@ -241,6 +241,7 @@ static int run_test(struct intel_gt *gt, int test_type)
struct intel_rps *rps = >->rps;
struct intel_engine_cs *engine;
enum intel_engine_id id;
+ intel_wakeref_t wakeref;
struct igt_spinner spin;
u32 slpc_min_freq, slpc_max_freq;
int err = 0;
@@ -278,7 +279,7 @@ static int run_test(struct intel_gt *gt, int test_type)
}
intel_gt_pm_wait_for_idle(gt);
- intel_gt_pm_get(gt);
+ wakeref = intel_gt_pm_get(gt);
for_each_engine(engine, gt, id) {
struct i915_request *rq;
u32 max_act_freq;
@@ -365,7 +366,7 @@ static int run_test(struct intel_gt *gt, int test_type)
if (igt_flush_test(gt->i915))
err = -EIO;
- intel_gt_pm_put(gt);
+ intel_gt_pm_put(gt, wakeref);
igt_spinner_fini(&spin);
intel_gt_pm_wait_for_idle(gt);
@@ -1106,7 +1106,7 @@ static void scrub_guc_desc_for_outstanding_g2h(struct intel_guc *guc)
if (deregister)
guc_signal_context_fence(ce);
if (destroyed) {
- intel_gt_pm_put_async(guc_to_gt(guc));
+ intel_gt_pm_put_async_untracked(guc_to_gt(guc));
release_guc_id(guc, ce);
__guc_context_destroy(ce);
}
@@ -1302,6 +1302,7 @@ static ktime_t guc_engine_busyness(struct intel_engine_cs *engine, ktime_t *now)
unsigned long flags;
u32 reset_count;
bool in_reset;
+ intel_wakeref_t wakeref;
spin_lock_irqsave(&guc->timestamp.lock, flags);
@@ -1324,7 +1325,7 @@ static ktime_t guc_engine_busyness(struct intel_engine_cs *engine, ktime_t *now)
* start_gt_clk is derived from GuC state. To get a consistent
* view of activity, we query the GuC state only if gt is awake.
*/
- if (!in_reset && intel_gt_pm_get_if_awake(gt)) {
+ if (!in_reset && (wakeref = intel_gt_pm_get_if_awake(gt))) {
stats_saved = *stats;
gt_stamp_saved = guc->timestamp.gt_stamp;
/*
@@ -1333,7 +1334,7 @@ static ktime_t guc_engine_busyness(struct intel_engine_cs *engine, ktime_t *now)
*/
guc_update_engine_gt_clks(engine);
guc_update_pm_timestamp(guc, now);
- intel_gt_pm_put_async(gt);
+ intel_gt_pm_put_async(gt, wakeref);
if (i915_reset_count(gpu_error) != reset_count) {
*stats = stats_saved;
guc->timestamp.gt_stamp = gt_stamp_saved;
@@ -4545,7 +4546,7 @@ int intel_guc_deregister_done_process_msg(struct intel_guc *guc,
intel_context_put(ce);
} else if (context_destroyed(ce)) {
/* Context has been destroyed */
- intel_gt_pm_put_async(guc_to_gt(guc));
+ intel_gt_pm_put_async_untracked(guc_to_gt(guc));
release_guc_id(guc, ce);
__guc_context_destroy(ce);
}
@@ -168,19 +168,19 @@ static u64 get_rc6(struct intel_gt *gt)
{
struct drm_i915_private *i915 = gt->i915;
struct i915_pmu *pmu = &i915->pmu;
+ intel_wakeref_t wakeref;
unsigned long flags;
- bool awake = false;
u64 val;
- if (intel_gt_pm_get_if_awake(gt)) {
+ wakeref = intel_gt_pm_get_if_awake(gt);
+ if (wakeref) {
val = __get_rc6(gt);
- intel_gt_pm_put_async(gt);
- awake = true;
+ intel_gt_pm_put_async(gt, wakeref);
}
spin_lock_irqsave(&pmu->lock, flags);
- if (awake) {
+ if (wakeref) {
pmu->sample[__I915_SAMPLE_RC6].cur = val;
} else {
/*
@@ -373,12 +373,14 @@ frequency_sample(struct intel_gt *gt, unsigned int period_ns)
struct drm_i915_private *i915 = gt->i915;
struct i915_pmu *pmu = &i915->pmu;
struct intel_rps *rps = >->rps;
+ intel_wakeref_t wakeref;
if (!frequency_sampling_enabled(pmu))
return;
/* Report 0/0 (actual/requested) frequency while parked. */
- if (!intel_gt_pm_get_if_awake(gt))
+ wakeref = intel_gt_pm_get_if_awake(gt);
+ if (!wakeref)
return;
if (pmu->enable & config_mask(I915_PMU_ACTUAL_FREQUENCY)) {
@@ -409,7 +411,7 @@ frequency_sample(struct intel_gt *gt, unsigned int period_ns)
period_ns / 1000);
}
- intel_gt_pm_put_async(gt);
+ intel_gt_pm_put_async(gt, wakeref);
}
static enum hrtimer_restart i915_sample(struct hrtimer *hrtimer)
@@ -108,6 +108,10 @@ void __intel_wakeref_init(struct intel_wakeref *wf,
INIT_DELAYED_WORK(&wf->work, __intel_wakeref_put_work);
lockdep_init_map(&wf->work.work.lockdep_map,
"wakeref.work", &key->work, 0);
+
+#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_WAKEREF)
+ intel_wakeref_tracker_init(&wf->debug);
+#endif
}
int intel_wakeref_wait_for_idle(struct intel_wakeref *wf)
@@ -43,6 +43,10 @@ struct intel_wakeref {
const struct intel_wakeref_ops *ops;
struct delayed_work work;
+
+#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_WAKEREF)
+ struct intel_wakeref_tracker debug;
+#endif
};
struct intel_wakeref_lockclass {
@@ -261,6 +265,44 @@ __intel_wakeref_defer_park(struct intel_wakeref *wf)
*/
int intel_wakeref_wait_for_idle(struct intel_wakeref *wf);
+#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_WAKEREF)
+
+static inline intel_wakeref_t intel_wakeref_track(struct intel_wakeref *wf)
+{
+ return intel_wakeref_tracker_add(&wf->debug);
+}
+
+static inline void intel_wakeref_untrack(struct intel_wakeref *wf,
+ intel_wakeref_t handle)
+{
+ intel_wakeref_tracker_remove(&wf->debug, handle);
+}
+
+static inline void intel_wakeref_show(struct intel_wakeref *wf,
+ struct drm_printer *p)
+{
+ intel_wakeref_tracker_show(&wf->debug, p);
+}
+
+#else
+
+static inline intel_wakeref_t intel_wakeref_track(struct intel_wakeref *wf)
+{
+ return -1;
+}
+
+static inline void intel_wakeref_untrack(struct intel_wakeref *wf,
+ intel_wakeref_t handle)
+{
+}
+
+static inline void intel_wakeref_show(struct intel_wakeref *wf,
+ struct drm_printer *p)
+{
+}
+
+#endif
+
struct intel_wakeref_auto {
struct intel_runtime_pm *rpm;
struct timer_list timer;