UbiquitousOS
/
OpenCloudOS-Kernel
mirror of https://gitee.com/OpenCloudOS/OpenCloudOS-Kernel.git
11
0
Fork 0
Code Issues Projects Releases Wiki Activity

drm/i915/perf: Schedule oa_config after modifying the contexts 

We wish that the scheduler emit the context modification commands prior
to enabling the oa_config, for which we must explicitly inform it of the
ordering constraints. This is especially important as we now wait for
the final oa_config setup to be completed and as this wait may be on a
distinct context to the state modifications, we need that command packet
to be always last in the queue.

We borrow the i915_active for its ability to track multiple timelines
and the last dma_fence on each; a flexible dma_resv. Keeping track of
each dma_fence is important for us so that we can efficiently schedule
the requests and reprioritise as required.

Reported-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200327112212.16046-3-chris@chris-wilson.co.uk
This commit is contained in:
Chris Wilson 2020-03-27 11:22:12 +00:00
parent 229007e02d
commit d7d50f801d
2 changed files with 103 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

156
drivers/gpu/drm/i915/i915_perf.c
View File

@ -1916,10 +1916,11 @@ out:
return i915_vma_get(oa_bo->vma); return i915_vma_get(oa_bo->vma);
} }
static struct i915_request * static int
emit_oa_config(struct i915_perf_stream *stream, emit_oa_config(struct i915_perf_stream *stream,
struct i915_oa_config *oa_config, struct i915_oa_config *oa_config,
struct intel_context *ce) struct intel_context *ce,
struct i915_active *active)
{ {
struct i915_request *rq; struct i915_request *rq;
struct i915_vma *vma; struct i915_vma *vma;
@ -1927,7 +1928,7 @@ emit_oa_config(struct i915_perf_stream *stream,
vma = get_oa_vma(stream, oa_config); vma = get_oa_vma(stream, oa_config);
if (IS_ERR(vma)) if (IS_ERR(vma))
return ERR_CAST(vma); return PTR_ERR(vma);
err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH); err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH);
if (err) if (err)
@ -1941,6 +1942,18 @@ emit_oa_config(struct i915_perf_stream *stream,
goto err_vma_unpin; goto err_vma_unpin;
} }
if (!IS_ERR_OR_NULL(active)) {
/* After all individual context modifications */
err = i915_request_await_active(rq, active,
I915_ACTIVE_AWAIT_ALL);
if (err)
goto err_add_request;
err = i915_active_add_request(active, rq);
if (err)
goto err_add_request;
}
i915_vma_lock(vma); i915_vma_lock(vma);
err = i915_request_await_object(rq, vma->obj, 0); err = i915_request_await_object(rq, vma->obj, 0);
if (!err) if (!err)
@ -1955,14 +1968,13 @@ emit_oa_config(struct i915_perf_stream *stream,
if (err) if (err)
goto err_add_request; goto err_add_request;
i915_request_get(rq);
err_add_request: err_add_request:
i915_request_add(rq); i915_request_add(rq);
err_vma_unpin: err_vma_unpin:
i915_vma_unpin(vma); i915_vma_unpin(vma);
err_vma_put: err_vma_put:
i915_vma_put(vma); i915_vma_put(vma);
return err ? ERR_PTR(err) : rq; return err;
} }
static struct intel_context *oa_context(struct i915_perf_stream *stream) static struct intel_context *oa_context(struct i915_perf_stream *stream)
@ -1970,8 +1982,9 @@ static struct intel_context *oa_context(struct i915_perf_stream *stream)
return stream->pinned_ctx ?: stream->engine->kernel_context; return stream->pinned_ctx ?: stream->engine->kernel_context;
} }
static struct i915_request * static int
hsw_enable_metric_set(struct i915_perf_stream *stream) hsw_enable_metric_set(struct i915_perf_stream *stream,
struct i915_active *active)
{ {
struct intel_uncore *uncore = stream->uncore; struct intel_uncore *uncore = stream->uncore;
@ -1990,7 +2003,9 @@ hsw_enable_metric_set(struct i915_perf_stream *stream)
intel_uncore_rmw(uncore, GEN6_UCGCTL1, intel_uncore_rmw(uncore, GEN6_UCGCTL1,
0, GEN6_CSUNIT_CLOCK_GATE_DISABLE); 0, GEN6_CSUNIT_CLOCK_GATE_DISABLE);
return emit_oa_config(stream, stream->oa_config, oa_context(stream)); return emit_oa_config(stream,
stream->oa_config, oa_context(stream),
active);
} }
static void hsw_disable_metric_set(struct i915_perf_stream *stream) static void hsw_disable_metric_set(struct i915_perf_stream *stream)
@ -2137,8 +2152,10 @@ static int gen8_modify_context(struct intel_context *ce,
return err; return err;
} }
static int gen8_modify_self(struct intel_context *ce, static int
const struct flex *flex, unsigned int count) gen8_modify_self(struct intel_context *ce,
const struct flex *flex, unsigned int count,
struct i915_active *active)
{ {
struct i915_request *rq; struct i915_request *rq;
int err; int err;
@ -2149,8 +2166,17 @@ static int gen8_modify_self(struct intel_context *ce,
if (IS_ERR(rq)) if (IS_ERR(rq))
return PTR_ERR(rq); return PTR_ERR(rq);
err = gen8_load_flex(rq, ce, flex, count); if (!IS_ERR_OR_NULL(active)) {
err = i915_active_add_request(active, rq);
if (err)
goto err_add_request;
}
err = gen8_load_flex(rq, ce, flex, count);
if (err)
goto err_add_request;
err_add_request:
i915_request_add(rq); i915_request_add(rq);
return err; return err;
} }
@ -2184,7 +2210,8 @@ static int gen8_configure_context(struct i915_gem_context *ctx,
return err; return err;
} }
static int gen12_configure_oar_context(struct i915_perf_stream *stream, bool enable) static int gen12_configure_oar_context(struct i915_perf_stream *stream,
struct i915_active *active)
{ {
int err; int err;
struct intel_context *ce = stream->pinned_ctx; struct intel_context *ce = stream->pinned_ctx;
@ -2193,7 +2220,7 @@ static int gen12_configure_oar_context(struct i915_perf_stream *stream, bool ena
{ {
GEN8_OACTXCONTROL, GEN8_OACTXCONTROL,
stream->perf->ctx_oactxctrl_offset + 1, stream->perf->ctx_oactxctrl_offset + 1,
enable ? GEN8_OA_COUNTER_RESUME : 0, active ? GEN8_OA_COUNTER_RESUME : 0,
}, },
}; };
/* Offsets in regs_lri are not used since this configuration is only /* Offsets in regs_lri are not used since this configuration is only
@ -2205,13 +2232,13 @@ static int gen12_configure_oar_context(struct i915_perf_stream *stream, bool ena
GEN12_OAR_OACONTROL, GEN12_OAR_OACONTROL,
GEN12_OAR_OACONTROL_OFFSET + 1, GEN12_OAR_OACONTROL_OFFSET + 1,
(format << GEN12_OAR_OACONTROL_COUNTER_FORMAT_SHIFT) | (format << GEN12_OAR_OACONTROL_COUNTER_FORMAT_SHIFT) |
(enable ? GEN12_OAR_OACONTROL_COUNTER_ENABLE : 0) (active ? GEN12_OAR_OACONTROL_COUNTER_ENABLE : 0)
}, },
{ {
RING_CONTEXT_CONTROL(ce->engine->mmio_base), RING_CONTEXT_CONTROL(ce->engine->mmio_base),
CTX_CONTEXT_CONTROL, CTX_CONTEXT_CONTROL,
_MASKED_FIELD(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE, _MASKED_FIELD(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE,
enable ? active ?
GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE : GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE :
0) 0)
}, },
@ -2228,7 +2255,7 @@ static int gen12_configure_oar_context(struct i915_perf_stream *stream, bool ena
return err; return err;
/* Apply regs_lri using LRI with pinned context */ /* Apply regs_lri using LRI with pinned context */
return gen8_modify_self(ce, regs_lri, ARRAY_SIZE(regs_lri)); return gen8_modify_self(ce, regs_lri, ARRAY_SIZE(regs_lri), active);
} }
/* /*
@ -2256,9 +2283,11 @@ static int gen12_configure_oar_context(struct i915_perf_stream *stream, bool ena
* Note: it's only the RCS/Render context that has any OA state. * Note: it's only the RCS/Render context that has any OA state.
* Note: the first flex register passed must always be R_PWR_CLK_STATE * Note: the first flex register passed must always be R_PWR_CLK_STATE
*/ */
static int oa_configure_all_contexts(struct i915_perf_stream *stream, static int
struct flex *regs, oa_configure_all_contexts(struct i915_perf_stream *stream,
size_t num_regs) struct flex *regs,
size_t num_regs,
struct i915_active *active)
{ {
struct drm_i915_private *i915 = stream->perf->i915; struct drm_i915_private *i915 = stream->perf->i915;
struct intel_engine_cs *engine; struct intel_engine_cs *engine;
@ -2315,7 +2344,7 @@ static int oa_configure_all_contexts(struct i915_perf_stream *stream,
regs[0].value = intel_sseu_make_rpcs(i915, &ce->sseu); regs[0].value = intel_sseu_make_rpcs(i915, &ce->sseu);
err = gen8_modify_self(ce, regs, num_regs); err = gen8_modify_self(ce, regs, num_regs, active);
if (err) if (err)
return err; return err;
} }
@ -2323,8 +2352,10 @@ static int oa_configure_all_contexts(struct i915_perf_stream *stream,
return 0; return 0;
} }
static int gen12_configure_all_contexts(struct i915_perf_stream *stream, static int
const struct i915_oa_config *oa_config) gen12_configure_all_contexts(struct i915_perf_stream *stream,
const struct i915_oa_config *oa_config,
struct i915_active *active)
{ {
struct flex regs[] = { struct flex regs[] = {
{ {
@ -2333,11 +2364,15 @@ static int gen12_configure_all_contexts(struct i915_perf_stream *stream,
}, },
}; };
return oa_configure_all_contexts(stream, regs, ARRAY_SIZE(regs)); return oa_configure_all_contexts(stream,
regs, ARRAY_SIZE(regs),
active);
} }
static int lrc_configure_all_contexts(struct i915_perf_stream *stream, static int
const struct i915_oa_config *oa_config) lrc_configure_all_contexts(struct i915_perf_stream *stream,
const struct i915_oa_config *oa_config,
struct i915_active *active)
{ {
/* The MMIO offsets for Flex EU registers aren't contiguous */ /* The MMIO offsets for Flex EU registers aren't contiguous */
const u32 ctx_flexeu0 = stream->perf->ctx_flexeu0_offset; const u32 ctx_flexeu0 = stream->perf->ctx_flexeu0_offset;
@ -2370,11 +2405,14 @@ static int lrc_configure_all_contexts(struct i915_perf_stream *stream,
for (i = 2; i < ARRAY_SIZE(regs); i++) for (i = 2; i < ARRAY_SIZE(regs); i++)
regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg); regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg);
return oa_configure_all_contexts(stream, regs, ARRAY_SIZE(regs)); return oa_configure_all_contexts(stream,
regs, ARRAY_SIZE(regs),
active);
} }
static struct i915_request * static int
gen8_enable_metric_set(struct i915_perf_stream *stream) gen8_enable_metric_set(struct i915_perf_stream *stream,
struct i915_active *active)
{ {
struct intel_uncore *uncore = stream->uncore; struct intel_uncore *uncore = stream->uncore;
struct i915_oa_config *oa_config = stream->oa_config; struct i915_oa_config *oa_config = stream->oa_config;
@ -2414,11 +2452,13 @@ gen8_enable_metric_set(struct i915_perf_stream *stream)
* to make sure all slices/subslices are ON before writing to NOA * to make sure all slices/subslices are ON before writing to NOA
* registers. * registers.
*/ */
ret = lrc_configure_all_contexts(stream, oa_config); ret = lrc_configure_all_contexts(stream, oa_config, active);
if (ret) if (ret)
return ERR_PTR(ret); return ret;
return emit_oa_config(stream, oa_config, oa_context(stream)); return emit_oa_config(stream,
stream->oa_config, oa_context(stream),
active);
} }
static u32 oag_report_ctx_switches(const struct i915_perf_stream *stream) static u32 oag_report_ctx_switches(const struct i915_perf_stream *stream)
@ -2428,8 +2468,9 @@ static u32 oag_report_ctx_switches(const struct i915_perf_stream *stream)
0 : GEN12_OAG_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS); 0 : GEN12_OAG_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS);
} }
static struct i915_request * static int
gen12_enable_metric_set(struct i915_perf_stream *stream) gen12_enable_metric_set(struct i915_perf_stream *stream,
struct i915_active *active)
{ {
struct intel_uncore *uncore = stream->uncore; struct intel_uncore *uncore = stream->uncore;
struct i915_oa_config *oa_config = stream->oa_config; struct i915_oa_config *oa_config = stream->oa_config;
@ -2458,9 +2499,9 @@ gen12_enable_metric_set(struct i915_perf_stream *stream)
* to make sure all slices/subslices are ON before writing to NOA * to make sure all slices/subslices are ON before writing to NOA
* registers. * registers.
*/ */
ret = gen12_configure_all_contexts(stream, oa_config); ret = gen12_configure_all_contexts(stream, oa_config, active);
if (ret) if (ret)
return ERR_PTR(ret); return ret;
/* /*
* For Gen12, performance counters are context * For Gen12, performance counters are context
@ -2468,12 +2509,14 @@ gen12_enable_metric_set(struct i915_perf_stream *stream)
* requested this. * requested this.
*/ */
if (stream->ctx) { if (stream->ctx) {
ret = gen12_configure_oar_context(stream, true); ret = gen12_configure_oar_context(stream, active);
if (ret) if (ret)
return ERR_PTR(ret); return ret;
} }
return emit_oa_config(stream, oa_config, oa_context(stream)); return emit_oa_config(stream,
stream->oa_config, oa_context(stream),
active);
} }
static void gen8_disable_metric_set(struct i915_perf_stream *stream) static void gen8_disable_metric_set(struct i915_perf_stream *stream)
@ -2481,7 +2524,7 @@ static void gen8_disable_metric_set(struct i915_perf_stream *stream)
struct intel_uncore *uncore = stream->uncore; struct intel_uncore *uncore = stream->uncore;
/* Reset all contexts' slices/subslices configurations. */ /* Reset all contexts' slices/subslices configurations. */
lrc_configure_all_contexts(stream, NULL); lrc_configure_all_contexts(stream, NULL, NULL);
intel_uncore_rmw(uncore, GDT_CHICKEN_BITS, GT_NOA_ENABLE, 0); intel_uncore_rmw(uncore, GDT_CHICKEN_BITS, GT_NOA_ENABLE, 0);
} }
@ -2491,7 +2534,7 @@ static void gen10_disable_metric_set(struct i915_perf_stream *stream)
struct intel_uncore *uncore = stream->uncore; struct intel_uncore *uncore = stream->uncore;
/* Reset all contexts' slices/subslices configurations. */ /* Reset all contexts' slices/subslices configurations. */
lrc_configure_all_contexts(stream, NULL); lrc_configure_all_contexts(stream, NULL, NULL);
/* Make sure we disable noa to save power. */ /* Make sure we disable noa to save power. */
intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0); intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0);
@ -2502,11 +2545,11 @@ static void gen12_disable_metric_set(struct i915_perf_stream *stream)
struct intel_uncore *uncore = stream->uncore; struct intel_uncore *uncore = stream->uncore;
/* Reset all contexts' slices/subslices configurations. */ /* Reset all contexts' slices/subslices configurations. */
gen12_configure_all_contexts(stream, NULL); gen12_configure_all_contexts(stream, NULL, NULL);
/* disable the context save/restore or OAR counters */ /* disable the context save/restore or OAR counters */
if (stream->ctx) if (stream->ctx)
gen12_configure_oar_context(stream, false); gen12_configure_oar_context(stream, NULL);
/* Make sure we disable noa to save power. */ /* Make sure we disable noa to save power. */
intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0); intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0);
@ -2680,16 +2723,19 @@ static const struct i915_perf_stream_ops i915_oa_stream_ops = {
static int i915_perf_stream_enable_sync(struct i915_perf_stream *stream) static int i915_perf_stream_enable_sync(struct i915_perf_stream *stream)
{ {
struct i915_request *rq; struct i915_active *active;
int err;
rq = stream->perf->ops.enable_metric_set(stream); active = i915_active_create();
if (IS_ERR(rq)) if (!active)
return PTR_ERR(rq); return -ENOMEM;
i915_request_wait(rq, 0, MAX_SCHEDULE_TIMEOUT); err = stream->perf->ops.enable_metric_set(stream, active);
i915_request_put(rq); if (err == 0)
__i915_active_wait(active, TASK_UNINTERRUPTIBLE);
return 0; i915_active_put(active);
return err;
} }
static void static void
@ -3171,7 +3217,7 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
return -EINVAL; return -EINVAL;
if (config != stream->oa_config) { if (config != stream->oa_config) {
struct i915_request *rq; int err;
/* /*
* If OA is bound to a specific context, emit the * If OA is bound to a specific context, emit the
@ -3182,13 +3228,11 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
* When set globally, we use a low priority kernel context, * When set globally, we use a low priority kernel context,
* so it will effectively take effect when idle. * so it will effectively take effect when idle.
*/ */
rq = emit_oa_config(stream, config, oa_context(stream)); err = emit_oa_config(stream, config, oa_context(stream), NULL);
if (!IS_ERR(rq)) { if (!err)
config = xchg(&stream->oa_config, config); config = xchg(&stream->oa_config, config);
i915_request_put(rq); else
} else { ret = err;
ret = PTR_ERR(rq);
}
} }
i915_oa_config_put(config); i915_oa_config_put(config);

5
drivers/gpu/drm/i915/i915_perf_types.h
View File

@ -22,6 +22,7 @@
struct drm_i915_private; struct drm_i915_private;
struct file; struct file;
struct i915_active;
struct i915_gem_context; struct i915_gem_context;
struct i915_perf; struct i915_perf;
struct i915_vma; struct i915_vma;
@ -340,8 +341,8 @@ struct i915_oa_ops {
* counter reports being sampled. May apply system constraints such as * counter reports being sampled. May apply system constraints such as
* disabling EU clock gating as required. * disabling EU clock gating as required.
*/ */
struct i915_request * int (*enable_metric_set)(struct i915_perf_stream *stream,
(*enable_metric_set)(struct i915_perf_stream *stream); struct i915_active *active);
/** /**
* @disable_metric_set: Remove system constraints associated with using * @disable_metric_set: Remove system constraints associated with using