Skip to contentSkip to conversation

Enable OA unit for Gen 8 and 9 in i915 perf

PATCH Superseded Robert Bragg 3 participants 17 messages Mar 23, 2017 – Apr 5, 2017

Subsystems (3)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover Enable OA unit for Gen 8 and 9 in i915 perf

Compared to the last Gen8+ OA series I've been investigating and debugging a number of issues with the configuration of the Flexible EU counters whose state is per-context:

* Removes assumption about the mmio registers having a contiguous range of addresses which wasn't true.
* Ensures that newly allocated contexts (while the OA unit is in use) will have their OA state properly initialized.
* Makes sure to only update render context state.

Instead of attempting to have a general purpose uncore api for dealing with the details of allowing mmio writes to per-context registers this just adds a constrained utility in i915_perf.c itself.

The per-gen initialization/configuration has been reworked to avoid a lot of copy and paste boilerplate.

Addresses the codegen issue Matt noticed with the Broadwell ComputeExtended metric set after checking with the authors of the XML files that the second unconditional MUX config should affectively be appended to the end of all the conditional configs.

(as with the last gen8+ series I sent, it's based on the various gen7 prep patches that have been sent out separately)

These patches can be pulled from my wip/rib/oa-next branch here:

https://github.com/rib/linux

In case anyone wants to take a look at the IGT tests so far they can be found here:

https://github.com/rib/intel-gpu-tools/commits/wip/rib/i915-perf-tests

Regards,
- Robert

Robert Bragg (5):
drm/i915: expose _SLICE_MASK GETPARM drm/i915: expose _SUBSLICE_MASK GETPARM drm/i915: Add 'render basic' Gen8+ OA unit configs drm/i915: Add OA unit support for Gen 8+ drm/i915: Add more OA configs for BDW, CHV, SKL + BXT

 drivers/gpu/drm/i915/Makefile           |    8 +-
 drivers/gpu/drm/i915/i915_drv.c         |   10 +
 drivers/gpu/drm/i915/i915_drv.h         |   31 +-
 drivers/gpu/drm/i915/i915_gem_context.h |    1 +
 drivers/gpu/drm/i915/i915_oa_bdw.c      | 5154 +++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_bdw.h      |   38 +
 drivers/gpu/drm/i915/i915_oa_bxt.c      | 2541 +++++++++++++++
 drivers/gpu/drm/i915/i915_oa_bxt.h      |   38 +
 drivers/gpu/drm/i915/i915_oa_chv.c      | 2730 ++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_chv.h      |   38 +
 drivers/gpu/drm/i915/i915_oa_hsw.c      |   58 +-
 drivers/gpu/drm/i915/i915_oa_sklgt2.c   | 3303 ++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt2.h   |   38 +
 drivers/gpu/drm/i915/i915_oa_sklgt3.c   | 2856 +++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt3.h   |   38 +
 drivers/gpu/drm/i915/i915_oa_sklgt4.c   | 2910 +++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt4.h   |   38 +
 drivers/gpu/drm/i915/i915_perf.c        | 1034 ++++++-
 drivers/gpu/drm/i915/i915_reg.h         |   22 +
 drivers/gpu/drm/i915/intel_lrc.c        |    5 +
 include/uapi/drm/i915_drm.h             |   21 +-
 21 files changed, 20825 insertions(+), 87 deletions(-)
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bdw.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bdw.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bxt.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bxt.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_chv.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_chv.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt2.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt2.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt3.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt3.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt4.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt4.h
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

PATCH v2 1/5 drm/i915: expose _SLICE_MASK GETPARM

Enables userspace to determine the number of slices enabled and also know what specific slices are enabled. This information is required, for example, to be able to analyse some OA counter reports where the counter configuration depends on the HW slice configuration.

---
 drivers/gpu/drm/i915/i915_drv.c | 5 +++++
 include/uapi/drm/i915_drm.h     | 1 +
 2 files changed, 6 insertions(+)
drivers/gpu/drm/i915/i915_drv.c +5 -0
index 6d9944a00b7d..ad12ffc356be 100644
@@ -357,6 +357,11 @@ static int i915_getparam(struct drm_device *dev, void *data,
 		 */
 		value = 1;
 		break;
+	case I915_PARAM_SLICE_MASK:
+		value = INTEL_INFO(dev_priv)->sseu.slice_mask;
+		if (!value)
+			return -ENODEV;
+		break;
 	default:
 		DRM_DEBUG("Unknown parameter %d\n", param->param);
 		return -EINVAL;
include/uapi/drm/i915_drm.h +1 -0
index 3554495bef13..f47fb7f26f36 100644
@@ -392,6 +392,7 @@ typedef struct drm_i915_irq_wait {
 #define I915_PARAM_HAS_POOLED_EU	 38
 #define I915_PARAM_MIN_EU_IN_POOL	 39
 #define I915_PARAM_MMAP_GTT_VERSION	 40
+#define I915_PARAM_SLICE_MASK		 45 /* XXX: rebase before landing */
 
 /* Query whether DRM_I915_GEM_EXECBUFFER2 supports user defined execution
  * priorities and the driver will attempt to execute batches in priority order.
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v2 1/5
#2 lore raw

Re: [PATCH v2 1/5] drm/i915: expose _SLICE_MASK GETPARM

Quoted 6 lines · Robert Bragg

Enables userspace to determine the number of slices enabled and also know what specific slices are enabled. This information is required, for example, to be able to analyse some OA counter reports where the counter configuration depends on the HW slice configuration.

Assuming you remember to update the 45 when landing:

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#3 lore raw

PATCH v2 2/5 drm/i915: expose _SUBSLICE_MASK GETPARM

In reply to Robert Bragg

Assuming a uniform mask across all slices, this enables userspace to determine the specific sub slices enabled. This information is required, for example, to be able to analyse some OA counter reports where the counter configuration depends on the HW sub slice configuration.

---
 drivers/gpu/drm/i915/i915_drv.c | 5 +++++
 include/uapi/drm/i915_drm.h     | 1 +
 2 files changed, 6 insertions(+)
drivers/gpu/drm/i915/i915_drv.c +5 -0
index ad12ffc356be..c39a03c1f78d 100644
@@ -362,6 +362,11 @@ static int i915_getparam(struct drm_device *dev, void *data,
 		if (!value)
 			return -ENODEV;
 		break;
+	case I915_PARAM_SUBSLICE_MASK:
+		value = INTEL_INFO(dev_priv)->sseu.subslice_mask;
+		if (!value)
+			return -ENODEV;
+		break;
 	default:
 		DRM_DEBUG("Unknown parameter %d\n", param->param);
 		return -EINVAL;
include/uapi/drm/i915_drm.h +1 -0
index f47fb7f26f36..e0599e729e68 100644
@@ -393,6 +393,7 @@ typedef struct drm_i915_irq_wait {
 #define I915_PARAM_MIN_EU_IN_POOL	 39
 #define I915_PARAM_MMAP_GTT_VERSION	 40
 #define I915_PARAM_SLICE_MASK		 45 /* XXX: rebase before landing */
+#define I915_PARAM_SUBSLICE_MASK	 46
 
 /* Query whether DRM_I915_GEM_EXECBUFFER2 supports user defined execution
  * priorities and the driver will attempt to execute batches in priority order.
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v2 2/5
#4 lore raw

Re: [PATCH v2 2/5] drm/i915: expose _SUBSLICE_MASK GETPARM

Quoted 6 lines · Robert Bragg

Assuming a uniform mask across all slices, this enables userspace to determine the specific sub slices enabled. This information is required, for example, to be able to analyse some OA counter reports where the counter configuration depends on the HW sub slice configuration.

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#5 lore raw

PATCH v2 3/5 drm/i915: Add 'render basic' Gen8+ OA unit configs

In reply to Robert Bragg

Adds a static OA unit, MUX, B Counter + Flex EU configurations for basic render metrics on Broadwell, Cherryview, Skylake and Broxton. These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml WHITELIST=RenderBasic
---
 drivers/gpu/drm/i915/Makefile         |   8 +-
 drivers/gpu/drm/i915/i915_drv.h       |   2 +
 drivers/gpu/drm/i915/i915_oa_bdw.c    | 380 ++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_bdw.h    |  38 ++++
 drivers/gpu/drm/i915/i915_oa_bxt.c    | 238 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_bxt.h    |  38 ++++
 drivers/gpu/drm/i915/i915_oa_chv.c    | 225 ++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_chv.h    |  38 ++++
 drivers/gpu/drm/i915/i915_oa_sklgt2.c | 228 ++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt2.h |  38 ++++
 drivers/gpu/drm/i915/i915_oa_sklgt3.c | 236 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt3.h |  38 ++++
 drivers/gpu/drm/i915/i915_oa_sklgt4.c | 247 ++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt4.h |  38 ++++
 14 files changed, 1791 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bdw.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bdw.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bxt.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bxt.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_chv.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_chv.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt2.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt2.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt3.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt3.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt4.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt4.h
drivers/gpu/drm/i915/Makefile +7 -1
index 2cf04504e494..41400a138a1e 100644
@@ -127,7 +127,13 @@ i915-y += i915_vgpu.o
 
 # perf code
 i915-y += i915_perf.o \
-	  i915_oa_hsw.o
+	  i915_oa_hsw.o \
+	  i915_oa_bdw.o \
+	  i915_oa_chv.o \
+	  i915_oa_sklgt2.o \
+	  i915_oa_sklgt3.o \
+	  i915_oa_sklgt4.o \
+	  i915_oa_bxt.o
 
 ifeq ($(CONFIG_DRM_I915_GVT),y)
 i915-y += intel_gvt.o
drivers/gpu/drm/i915/i915_drv.h +2 -0
index a7986c0c29ad..c4156a8a5dc0 100644
@@ -2486,6 +2486,8 @@ struct drm_i915_private {
 			int mux_regs_len;
 			const struct i915_oa_reg *b_counter_regs;
 			int b_counter_regs_len;
+			const struct i915_oa_reg *flex_regs;
+			int flex_regs_len;
 
 			struct {
 				struct i915_vma *vma;
drivers/gpu/drm/i915/i915_oa_bdw.c new file +380 -0

380 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_bdw.h new file +38 -0
new file mode 100644
index 000000000000..ea6969aaf91a
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_BDW_H__
+#define __I915_OA_BDW_H__
+
+extern int i915_oa_n_builtin_metric_sets_bdw;
+
+extern int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_bdw(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_bdw(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_bxt.c new file +238 -0

238 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_bxt.h new file +38 -0
new file mode 100644
index 000000000000..9e86662c891d
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_BXT_H__
+#define __I915_OA_BXT_H__
+
+extern int i915_oa_n_builtin_metric_sets_bxt;
+
+extern int i915_oa_select_metric_set_bxt(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_bxt(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_bxt(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_chv.c new file +225 -0

225 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_chv.h new file +38 -0
new file mode 100644
index 000000000000..e494200ff04f
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_CHV_H__
+#define __I915_OA_CHV_H__
+
+extern int i915_oa_n_builtin_metric_sets_chv;
+
+extern int i915_oa_select_metric_set_chv(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_chv(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_chv(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_sklgt2.c new file +228 -0

228 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt2.h new file +38 -0
new file mode 100644
index 000000000000..e99b03f1bf9e
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_SKLGT2_H__
+#define __I915_OA_SKLGT2_H__
+
+extern int i915_oa_n_builtin_metric_sets_sklgt2;
+
+extern int i915_oa_select_metric_set_sklgt2(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_sklgt2(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_sklgt2(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_sklgt3.c new file +236 -0

236 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt3.h new file +38 -0
new file mode 100644
index 000000000000..ff7ec3206bf2
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_SKLGT3_H__
+#define __I915_OA_SKLGT3_H__
+
+extern int i915_oa_n_builtin_metric_sets_sklgt3;
+
+extern int i915_oa_select_metric_set_sklgt3(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_sklgt3(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_sklgt3(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_sklgt4.c new file +247 -0

247 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt4.h new file +38 -0
new file mode 100644
index 000000000000..4faa8c4579cd
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_SKLGT4_H__
+#define __I915_OA_SKLGT4_H__
+
+extern int i915_oa_n_builtin_metric_sets_sklgt4;
+
+extern int i915_oa_select_metric_set_sklgt4(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_sklgt4(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_sklgt4(struct drm_i915_private *dev_priv);
+
+#endif
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v2 3/5
#6 lore raw

Re: [PATCH v2 3/5] drm/i915: Add 'render basic' Gen8+ OA unit configs

Quoted 12 lines · Robert Bragg

Adds a static OA unit, MUX, B Counter + Flex EU configurations for basic render metrics on Broadwell, Cherryview, Skylake and Broxton. These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml WHITELIST=RenderBasic

<SNIP>

Quoted 7 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +7 -0
+
+int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv)
+{
+       dev_priv->perf.oa.mux_regs = NULL;
+       dev_priv->perf.oa.mux_regs_len = 0;
+       dev_priv->perf.oa.b_counter_regs = NULL;
+       dev_priv->perf.oa.b_counter_regs_len = 0;

drivers/gpu/drm/i915/i915_oa_bdw.c line 308

Quoted 10 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +10 -0
+       dev_priv->perf.oa.flex_regs = NULL;
+       dev_priv->perf.oa.flex_regs_len = 0;
+
+       switch (dev_priv->perf.oa.metrics_set) {
+       case METRIC_SET_ID_RENDER_BASIC:
+               dev_priv->perf.oa.mux_regs =
+                       get_render_basic_mux_config(dev_priv,
+                                                   &dev_priv->perf.oa.mux_regs_len);
+               if (!dev_priv->perf.oa.mux_regs) {
+                       DRM_DEBUG_DRIVER("No suitable MUX config for \"RENDER_BASIC\" metric set");

You forgot to update your script ;)

drivers/gpu/drm/i915/i915_oa_bdw.c line 312

Quoted 4 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +4 -0
+
+                       /* EINVAL because *_register_sysfs already checked this
+                        * and so it wouldn't have been advertised so userspace and
+                        * so shouldn't have been requested

s/so userspace/to userspace/ ?

Otherwise assuming the configs are indeed correct:

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#7 lore raw

Re: [PATCH v2 3/5] drm/i915: Add 'render basic' Gen8+ OA unit configs

Quoted 26 lines · Matthew Auld
Quoted 12 lines · Robert Bragg

Adds a static OA unit, MUX, B Counter + Flex EU configurations for basic render metrics on Broadwell, Cherryview, Skylake and Broxton. These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml WHITELIST=RenderBasic

<SNIP>

Quoted 10 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +10 -0
+
+int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv)
+{
+       dev_priv->perf.oa.mux_regs = NULL;
+       dev_priv->perf.oa.mux_regs_len = 0;
+       dev_priv->perf.oa.b_counter_regs = NULL;
+       dev_priv->perf.oa.b_counter_regs_len = 0;
+       dev_priv->perf.oa.flex_regs = NULL;
+       dev_priv->perf.oa.flex_regs_len = 0;
+
Quoted 8 lines

drivers/gpu/drm/i915/i915_oa_bdw.c line 308

Quoted 7 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +7 -0
+       switch (dev_priv->perf.oa.metrics_set) {
+       case METRIC_SET_ID_RENDER_BASIC:
+               dev_priv->perf.oa.mux_regs =
+                       get_render_basic_mux_config(dev_priv,
+                                                   &dev_priv->perf.oa.mux_regs_len);
+               if (!dev_priv->perf.oa.mux_regs) {
+                       DRM_DEBUG_DRIVER("No suitable MUX config for \"RENDER_BASIC\" metric set");

You forgot to update your script ;)

Hmm, that's odd; I didn't:
https://github.com/rib/gputop/commit/95bc05957d488e2004ef2e7b5ba5b33d7dd559dd

So need to double check what's up here.

Quoted 6 lines

drivers/gpu/drm/i915/i915_oa_bdw.c line 312

Quoted 5 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +4 -0
+
+                       /* EINVAL because *_register_sysfs already checked this
+                        * and so it wouldn't have been advertised so userspace and
+                        * so shouldn't have been requested

s/so userspace/to userspace/ ?

Ah yup.

Quoted 3 lines

Otherwise assuming the configs are indeed correct:

Thanks
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#8 lore raw

Re: [PATCH v2 3/5] drm/i915: Add 'render basic' Gen8+ OA unit configs

Quoted 37 lines · Robert Bragg
Quoted 26 lines · Matthew Auld
Quoted 12 lines · Robert Bragg

Adds a static OA unit, MUX, B Counter + Flex EU configurations for basic render metrics on Broadwell, Cherryview, Skylake and Broxton. These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml WHITELIST=RenderBasic

<SNIP>

Quoted 10 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +10 -0
+
+int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv)
+{
+       dev_priv->perf.oa.mux_regs = NULL;
+       dev_priv->perf.oa.mux_regs_len = 0;
+       dev_priv->perf.oa.b_counter_regs = NULL;
+       dev_priv->perf.oa.b_counter_regs_len = 0;
+       dev_priv->perf.oa.flex_regs = NULL;
+       dev_priv->perf.oa.flex_regs_len = 0;
+
Quoted 8 lines

drivers/gpu/drm/i915/i915_oa_bdw.c line 308

Quoted 7 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +7 -0
+       switch (dev_priv->perf.oa.metrics_set) {
+       case METRIC_SET_ID_RENDER_BASIC:
+               dev_priv->perf.oa.mux_regs =
+                       get_render_basic_mux_config(dev_priv,
+                                                   &dev_priv->perf.oa.mux_regs_len);
+               if (!dev_priv->perf.oa.mux_regs) {
+                       DRM_DEBUG_DRIVER("No suitable MUX config for \"RENDER_BASIC\" metric set");

You forgot to update your script ;)

Quoted 4 lines

Hmm, that's odd; I didn't:
https://github.com/rib/gputop/commit/95bc05957d488e2004ef2e7b5ba5b33d7dd559dd

So need to double check what's up here.

*facepalm* - must have been on autopilot, sorry will fix.

Quoted 8 lines
Quoted 7 lines

drivers/gpu/drm/i915/i915_oa_bdw.c line 312

Quoted 6 lines
drivers/gpu/drm/i915/i915_oa_bdw.c new file +4 -0
+
+                       /* EINVAL because *_register_sysfs already checked this
+                        * and so it wouldn't have been advertised so userspace and
+                        * so shouldn't have been requested

s/so userspace/to userspace/ ?

Quoted 7 lines

Ah yup.

Quoted 3 lines

Otherwise assuming the configs are indeed correct:

Thanks

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#9 lore raw

PATCH v2 1/1 drm/i915: Add 'render basic' Gen8+ OA unit configs

In reply to Matthew Auld

Adds a static OA unit, MUX, B Counter + Flex EU configurations for basic render metrics on Broadwell, Cherryview, Skylake and Broxton. These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml WHITELIST=RenderBasic

v2: add newlines to debug messages + fix comment (Matthew Auld)

---
 drivers/gpu/drm/i915/Makefile         |   8 +-
 drivers/gpu/drm/i915/i915_drv.h       |   2 +
 drivers/gpu/drm/i915/i915_oa_bdw.c    | 380 ++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_bdw.h    |  38 ++++
 drivers/gpu/drm/i915/i915_oa_bxt.c    | 238 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_bxt.h    |  38 ++++
 drivers/gpu/drm/i915/i915_oa_chv.c    | 225 ++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_chv.h    |  38 ++++
 drivers/gpu/drm/i915/i915_oa_sklgt2.c | 228 ++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt2.h |  38 ++++
 drivers/gpu/drm/i915/i915_oa_sklgt3.c | 236 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt3.h |  38 ++++
 drivers/gpu/drm/i915/i915_oa_sklgt4.c | 247 ++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_oa_sklgt4.h |  38 ++++
 14 files changed, 1791 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bdw.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bdw.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bxt.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_bxt.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_chv.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_chv.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt2.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt2.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt3.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt3.h
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt4.c
 create mode 100644 drivers/gpu/drm/i915/i915_oa_sklgt4.h
drivers/gpu/drm/i915/Makefile +7 -1
index 2cf04504e494..41400a138a1e 100644
@@ -127,7 +127,13 @@ i915-y += i915_vgpu.o
 
 # perf code
 i915-y += i915_perf.o \
-	  i915_oa_hsw.o
+	  i915_oa_hsw.o \
+	  i915_oa_bdw.o \
+	  i915_oa_chv.o \
+	  i915_oa_sklgt2.o \
+	  i915_oa_sklgt3.o \
+	  i915_oa_sklgt4.o \
+	  i915_oa_bxt.o
 
 ifeq ($(CONFIG_DRM_I915_GVT),y)
 i915-y += intel_gvt.o
drivers/gpu/drm/i915/i915_drv.h +2 -0
index a7986c0c29ad..c4156a8a5dc0 100644
@@ -2486,6 +2486,8 @@ struct drm_i915_private {
 			int mux_regs_len;
 			const struct i915_oa_reg *b_counter_regs;
 			int b_counter_regs_len;
+			const struct i915_oa_reg *flex_regs;
+			int flex_regs_len;
 
 			struct {
 				struct i915_vma *vma;
drivers/gpu/drm/i915/i915_oa_bdw.c new file +380 -0

380 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_bdw.h new file +38 -0
new file mode 100644
index 000000000000..ea6969aaf91a
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_BDW_H__
+#define __I915_OA_BDW_H__
+
+extern int i915_oa_n_builtin_metric_sets_bdw;
+
+extern int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_bdw(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_bdw(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_bxt.c new file +238 -0

238 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_bxt.h new file +38 -0
new file mode 100644
index 000000000000..9e86662c891d
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_BXT_H__
+#define __I915_OA_BXT_H__
+
+extern int i915_oa_n_builtin_metric_sets_bxt;
+
+extern int i915_oa_select_metric_set_bxt(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_bxt(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_bxt(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_chv.c new file +225 -0

225 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_chv.h new file +38 -0
new file mode 100644
index 000000000000..e494200ff04f
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_CHV_H__
+#define __I915_OA_CHV_H__
+
+extern int i915_oa_n_builtin_metric_sets_chv;
+
+extern int i915_oa_select_metric_set_chv(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_chv(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_chv(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_sklgt2.c new file +228 -0

228 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt2.h new file +38 -0
new file mode 100644
index 000000000000..e99b03f1bf9e
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_SKLGT2_H__
+#define __I915_OA_SKLGT2_H__
+
+extern int i915_oa_n_builtin_metric_sets_sklgt2;
+
+extern int i915_oa_select_metric_set_sklgt2(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_sklgt2(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_sklgt2(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_sklgt3.c new file +236 -0

236 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt3.h new file +38 -0
new file mode 100644
index 000000000000..ff7ec3206bf2
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_SKLGT3_H__
+#define __I915_OA_SKLGT3_H__
+
+extern int i915_oa_n_builtin_metric_sets_sklgt3;
+
+extern int i915_oa_select_metric_set_sklgt3(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_sklgt3(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_sklgt3(struct drm_i915_private *dev_priv);
+
+#endif
drivers/gpu/drm/i915/i915_oa_sklgt4.c new file +247 -0

247 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt4.h new file +38 -0
new file mode 100644
index 000000000000..4faa8c4579cd
@@ -0,0 +1,38 @@
+/*
+ * Autogenerated file, DO NOT EDIT manually!
+ *
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __I915_OA_SKLGT4_H__
+#define __I915_OA_SKLGT4_H__
+
+extern int i915_oa_n_builtin_metric_sets_sklgt4;
+
+extern int i915_oa_select_metric_set_sklgt4(struct drm_i915_private *dev_priv);
+
+extern int i915_perf_register_sysfs_sklgt4(struct drm_i915_private *dev_priv);
+
+extern void i915_perf_unregister_sysfs_sklgt4(struct drm_i915_private *dev_priv);
+
+#endif
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#10 lore raw

PATCH v2 4/5 drm/i915: Add OA unit support for Gen 8+

In reply to Robert Bragg

Enables access to OA unit metrics for BDW, CHV, SKL and BXT which all share (more-or-less) the same OA unit design.

Of particular note in comparison to Haswell: some OA unit HW config state has become per-context state and as a consequence it is somewhat more complicated to manage synchronous state changes from the cpu while there's no guarantee of what context (if any) is currently actively running on the gpu.

The periodic sampling frequency which can be particularly useful for system-wide analysis (as opposed to command stream synchronised MI_REPORT_PERF_COUNT commands) is perhaps the most surprising state to have become per-context save and restored (while the OABUFFER destination is still a shared, system-wide resource).

This support for gen8+ takes care to consider a number of timing challenges involved in synchronously updating per-context state primarily by programming all config state from the cpu and updating all current and saved contexts synchronously while the OA unit is still disabled.

The driver intentionally avoids depending on command streamer programming to update OA state considering the lack of synchronization between the automatic loading of OACTXCONTROL state (that includes the periodic sampling state and enable state) on context restore and the parsing of any general purpose BB the driver can control. I.e. this implementation is careful to avoid the possibility of a context restore temporarily enabling any out-of-date periodic sampling state. In addition to the risk of transiently-out-of-date state being loaded automatically; there are also internal HW latencies involved in the loading of MUX configurations which would be difficult to account for from the command streamer (and we only want to enable the unit when once the MUX configuration is complete).

Since the Gen8+ OA unit design no longer supports clock gating the unit off for a single given context (which effectively stopped any progress of counters while any other context was running) and instead supports tagging OA reports with a context ID for filtering on the CPU, it means we can no longer hide the system-wide progress of counters from a non-privileged application only interested in metrics for its own context. Although we could theoretically try and subtract the progress of other contexts before forwarding reports via read() we aren't in a position to filter reports captured via MI_REPORT_PERF_COUNT commands. As a result, for Gen8+, we always require the dev.i915.perf_stream_paranoid to be unset for any access to OA metrics if not root.

---
 drivers/gpu/drm/i915/i915_drv.h         |   29 +-
 drivers/gpu/drm/i915/i915_gem_context.h |    1 +
 drivers/gpu/drm/i915/i915_perf.c        | 1034 ++++++++++++++++++++++++++++---
 drivers/gpu/drm/i915/i915_reg.h         |   22 +
 drivers/gpu/drm/i915/intel_lrc.c        |    5 +
 include/uapi/drm/i915_drm.h             |   19 +-
 6 files changed, 1029 insertions(+), 81 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +26 -3
index c4156a8a5dc0..190e699d5851 100644
@@ -731,6 +731,7 @@ intel_uncore_forcewake_for_reg(struct drm_i915_private *dev_priv,
 			       i915_reg_t reg, unsigned int op);
 
 struct intel_uncore_funcs {
+	int (*wait_for_rcs_busy)(struct drm_i915_private *dev_priv);
 	void (*force_wake_get)(struct drm_i915_private *dev_priv,
 			       enum forcewake_domains domains);
 	void (*force_wake_put)(struct drm_i915_private *dev_priv,
@@ -2084,9 +2085,17 @@ struct i915_oa_ops {
 	void (*init_oa_buffer)(struct drm_i915_private *dev_priv);
 
 	/**
-	 * @enable_metric_set: Applies any MUX configuration to set up the
-	 * Boolean and Custom (B/C) counters that are part of the counter
-	 * reports being sampled. May apply system constraints such as
+	 * @select_metric_set: The auto generated code that checks whether a
+	 * requested OA config is applicable to the system and if so sets up
+	 * the mux, oa and flex eu register config pointers according to the
+	 * current dev_priv->perf.oa.metrics_set.
+	 */
+	int (*select_metric_set)(struct drm_i915_private *dev_priv);
+
+	/**
+	 * @enable_metric_set: Selects and applies any MUX configuration to set
+	 * up the Boolean and Custom (B/C) counters that are part of the
+	 * counter reports being sampled. May apply system constraints such as
 	 * disabling EU clock gating as required.
 	 */
 	int (*enable_metric_set)(struct drm_i915_private *dev_priv);
@@ -2492,6 +2501,7 @@ struct drm_i915_private {
 			struct {
 				struct i915_vma *vma;
 				u8 *vaddr;
+				u32 last_ctx_id;
 				int format;
 				int format_size;
 
@@ -2561,6 +2571,14 @@ struct drm_i915_private {
 			} oa_buffer;
 
 			u32 gen7_latched_oastatus1;
+			u32 ctx_oactxctrl_off;
+			u32 ctx_flexeu0_off;
+
+			/* The RPT_ID/reason field for Gen8+ includes a bit
+			 * to determine if the CTX ID in the report is valid
+			 * but the specific bit differs between Gen 8 and 9
+			 */
+			u32 gen8_valid_ctx_bit;
 
 			struct i915_oa_ops ops;
 			const struct i915_oa_format *oa_formats;
@@ -2871,6 +2889,8 @@ intel_info(const struct drm_i915_private *dev_priv)
 #define IS_KBL_ULX(dev_priv)	(INTEL_DEVID(dev_priv) == 0x590E || \
 				 INTEL_DEVID(dev_priv) == 0x5915 || \
 				 INTEL_DEVID(dev_priv) == 0x591E)
+#define IS_SKL_GT2(dev_priv)	(IS_SKYLAKE(dev_priv) && \
+				 (INTEL_DEVID(dev_priv) & 0x00F0) == 0x0010)
 #define IS_SKL_GT3(dev_priv)	(IS_SKYLAKE(dev_priv) && \
 				 (INTEL_DEVID(dev_priv) & 0x00F0) == 0x0020)
 #define IS_SKL_GT4(dev_priv)	(IS_SKYLAKE(dev_priv) && \
@@ -3622,6 +3642,9 @@ i915_gem_context_lookup_timeline(struct i915_gem_context *ctx,
 
 int i915_perf_open_ioctl(struct drm_device *dev, void *data,
 			 struct drm_file *file);
+void i915_oa_update_reg_state(struct intel_engine_cs *engine,
+			      struct i915_gem_context *ctx,
+			      uint32_t *reg_state);
 
 /* i915_gem_evict.c */
 int __must_check i915_gem_evict_something(struct i915_address_space *vm,
drivers/gpu/drm/i915/i915_gem_context.h +1 -0
index 4af2ab94558b..3f4ce73bea43 100644
@@ -151,6 +151,7 @@ struct i915_gem_context {
 		u64 lrc_desc;
 		int pin_count;
 		bool initialised;
+		atomic_t oa_state_dirty;
 	} engine[I915_NUM_ENGINES];
 
 	/** ring_size: size for allocating the per-engine ring buffer */
drivers/gpu/drm/i915/i915_perf.c +963 -71

1,176 lines not on this page · raw message

drivers/gpu/drm/i915/i915_reg.h +22 -0
index 04c8f69fcc62..0052289ed8ad 100644
@@ -645,6 +645,12 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 
 #define GEN8_OACTXID _MMIO(0x2364)
 
+#define GEN8_OA_DEBUG _MMIO(0x2B04)
+#define  GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS    (1<<5)
+#define  GEN9_OA_DEBUG_INCLUDE_CLK_RATIO	    (1<<6)
+#define  GEN9_OA_DEBUG_DISABLE_GO_1_0_REPORTS	    (1<<2)
+#define  GEN9_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS   (1<<1)
+
 #define GEN8_OACONTROL _MMIO(0x2B00)
 #define  GEN8_OA_REPORT_FORMAT_A12	    (0<<2)
 #define  GEN8_OA_REPORT_FORMAT_A12_B8_C8    (2<<2)
@@ -666,6 +672,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define  GEN7_OABUFFER_STOP_RESUME_ENABLE   (1<<1)
 #define  GEN7_OABUFFER_RESUME		    (1<<0)
 
+#define GEN8_OABUFFER_UDW _MMIO(0x23b4)
 #define GEN8_OABUFFER _MMIO(0x2b14)
 
 #define GEN7_OASTATUS1 _MMIO(0x2364)
@@ -684,7 +691,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define  GEN8_OASTATUS_REPORT_LOST	    (1<<0)
 
 #define GEN8_OAHEADPTR _MMIO(0x2B0C)
+#define GEN8_OAHEADPTR_MASK    0xffffffc0
 #define GEN8_OATAILPTR _MMIO(0x2B10)
+#define GEN8_OATAILPTR_MASK    0xffffffc0
 
 #define OABUFFER_SIZE_128K  (0<<3)
 #define OABUFFER_SIZE_256K  (1<<3)
@@ -697,7 +706,17 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 
 #define OA_MEM_SELECT_GGTT  (1<<0)
 
+/*
+ * Flexible, Aggregate EU Counter Registers.
+ * Note: these aren't contiguous
+ */
 #define EU_PERF_CNTL0	    _MMIO(0xe458)
+#define EU_PERF_CNTL1	    _MMIO(0xe558)
+#define EU_PERF_CNTL2	    _MMIO(0xe658)
+#define EU_PERF_CNTL3	    _MMIO(0xe758)
+#define EU_PERF_CNTL4	    _MMIO(0xe45c)
+#define EU_PERF_CNTL5	    _MMIO(0xe55c)
+#define EU_PERF_CNTL6	    _MMIO(0xe65c)
 
 #define GDT_CHICKEN_BITS    _MMIO(0x9840)
 #define GT_NOA_ENABLE	    0x00000080
@@ -2317,6 +2336,9 @@ enum skl_disp_power_wells {
 #define   GEN8_RC_SEMA_IDLE_MSG_DISABLE	(1 << 12)
 #define   GEN8_FF_DOP_CLOCK_GATE_DISABLE	(1<<10)
 
+#define GEN6_RCS_PWR_FSM _MMIO(0x22ac)
+#define GEN9_RCS_FE_FSM2 _MMIO(0x22a4)
+
 /* Fuse readout registers for GT */
 #define CHV_FUSE_GT			_MMIO(VLV_DISPLAY_BASE + 0x2168)
 #define   CHV_FGT_DISABLE_SS0		(1 << 10)
drivers/gpu/drm/i915/intel_lrc.c +5 -0
index eec1e714f531..e6d9e4197d3d 100644
@@ -337,6 +337,8 @@ static u64 execlists_update_context(struct drm_i915_gem_request *rq)
 	if (ppgtt && !i915_vm_is_48bit(&ppgtt->base))
 		execlists_update_context_pdps(ppgtt, reg_state);
 
+	i915_oa_update_reg_state(rq->engine, rq->ctx, reg_state);
+
 	return ce->lrc_desc;
 }
 
@@ -1878,6 +1880,9 @@ static void execlists_init_reg_state(u32 *regs,
 		regs[CTX_LRI_HEADER_2] = MI_LOAD_REGISTER_IMM(1);
 		CTX_REG(regs, CTX_R_PWR_CLK_STATE, GEN8_R_PWR_CLK_STATE,
 			make_rpcs(dev_priv));
+
+		atomic_set(&ctx->engine[RCS].oa_state_dirty, 1);
+		i915_oa_update_reg_state(engine, ctx, regs);
 	}
 }
 
include/uapi/drm/i915_drm.h +12 -7
index e0599e729e68..03b833849919 100644
@@ -1295,13 +1295,18 @@ struct drm_i915_gem_context_param {
 };
 
 enum drm_i915_oa_format {
-	I915_OA_FORMAT_A13 = 1,
-	I915_OA_FORMAT_A29,
-	I915_OA_FORMAT_A13_B8_C8,
-	I915_OA_FORMAT_B4_C8,
-	I915_OA_FORMAT_A45_B8_C8,
-	I915_OA_FORMAT_B4_C8_A16,
-	I915_OA_FORMAT_C4_B8,
+	I915_OA_FORMAT_A13 = 1,	    /* HSW only */
+	I915_OA_FORMAT_A29,	    /* HSW only */
+	I915_OA_FORMAT_A13_B8_C8,   /* HSW only */
+	I915_OA_FORMAT_B4_C8,	    /* HSW only */
+	I915_OA_FORMAT_A45_B8_C8,   /* HSW only */
+	I915_OA_FORMAT_B4_C8_A16,   /* HSW only */
+	I915_OA_FORMAT_C4_B8,	    /* HSW+ */
+
+	/* Gen8+ */
+	I915_OA_FORMAT_A12,
+	I915_OA_FORMAT_A12_B8_C8,
+	I915_OA_FORMAT_A32u40_A4u32_B8_C8,
 
 	I915_OA_FORMAT_MAX	    /* non-ABI */
 };
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#11 lore raw

Re: [PATCH v2 4/5] drm/i915: Add OA unit support for Gen 8+

Quoted 62 lines · Robert Bragg

Enables access to OA unit metrics for BDW, CHV, SKL and BXT which all share (more-or-less) the same OA unit design.

Of particular note in comparison to Haswell: some OA unit HW config state has become per-context state and as a consequence it is somewhat more complicated to manage synchronous state changes from the cpu while there's no guarantee of what context (if any) is currently actively running on the gpu.

The periodic sampling frequency which can be particularly useful for system-wide analysis (as opposed to command stream synchronised MI_REPORT_PERF_COUNT commands) is perhaps the most surprising state to have become per-context save and restored (while the OABUFFER destination is still a shared, system-wide resource).

This support for gen8+ takes care to consider a number of timing challenges involved in synchronously updating per-context state primarily by programming all config state from the cpu and updating all current and saved contexts synchronously while the OA unit is still disabled.

The driver intentionally avoids depending on command streamer programming to update OA state considering the lack of synchronization between the automatic loading of OACTXCONTROL state (that includes the periodic sampling state and enable state) on context restore and the parsing of any general purpose BB the driver can control. I.e. this implementation is careful to avoid the possibility of a context restore temporarily enabling any out-of-date periodic sampling state. In addition to the risk of transiently-out-of-date state being loaded automatically; there are also internal HW latencies involved in the loading of MUX configurations which would be difficult to account for from the command streamer (and we only want to enable the unit when once the MUX configuration is complete).

Since the Gen8+ OA unit design no longer supports clock gating the unit off for a single given context (which effectively stopped any progress of counters while any other context was running) and instead supports tagging OA reports with a context ID for filtering on the CPU, it means we can no longer hide the system-wide progress of counters from a non-privileged application only interested in metrics for its own context. Although we could theoretically try and subtract the progress of other contexts before forwarding reports via read() we aren't in a position to filter reports captured via MI_REPORT_PERF_COUNT commands. As a result, for Gen8+, we always require the dev.i915.perf_stream_paranoid to be unset for any access to OA metrics if not root.

---
 drivers/gpu/drm/i915/i915_drv.h         |   29 +-
 drivers/gpu/drm/i915/i915_gem_context.h |    1 +
 drivers/gpu/drm/i915/i915_perf.c        | 1034 ++++++++++++++++++++++++++++---
 drivers/gpu/drm/i915/i915_reg.h         |   22 +
 drivers/gpu/drm/i915/intel_lrc.c        |    5 +
 include/uapi/drm/i915_drm.h             |   19 +-
 6 files changed, 1029 insertions(+), 81 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +0 -0
index c4156a8a5dc0..190e699d5851 100644

drivers/gpu/drm/i915/i915_drv.h intel_uncore_forcewake_for_reg(struct drm_i915_private *dev_priv, · line 734

Quoted 4 lines
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -731,6 +731,7 @@ intel_uncore_forcewake_for_reg(struct drm_i915_private *dev_priv,
 			       i915_reg_t reg, unsigned int op);
 
 struct intel_uncore_funcs {
+	int (*wait_for_rcs_busy)(struct drm_i915_private *dev_priv);

No longer needed.

Quoted 33 lines
drivers/gpu/drm/i915/i915_drv.h +12 -3
 	void (*force_wake_get)(struct drm_i915_private *dev_priv,
 			       enum forcewake_domains domains);
 	void (*force_wake_put)(struct drm_i915_private *dev_priv,
@@ -2084,9 +2085,17 @@ struct i915_oa_ops {
 	void (*init_oa_buffer)(struct drm_i915_private *dev_priv);
 
 	/**
-	 * @enable_metric_set: Applies any MUX configuration to set up the
-	 * Boolean and Custom (B/C) counters that are part of the counter
-	 * reports being sampled. May apply system constraints such as
+	 * @select_metric_set: The auto generated code that checks whether a
+	 * requested OA config is applicable to the system and if so sets up
+	 * the mux, oa and flex eu register config pointers according to the
+	 * current dev_priv->perf.oa.metrics_set.
+	 */
+	int (*select_metric_set)(struct drm_i915_private *dev_priv);
+
+	/**
+	 * @enable_metric_set: Selects and applies any MUX configuration to set
+	 * up the Boolean and Custom (B/C) counters that are part of the
+	 * counter reports being sampled. May apply system constraints such as
 	 * disabling EU clock gating as required.
 	 */
 	int (*enable_metric_set)(struct drm_i915_private *dev_priv);
@@ -2492,6 +2501,7 @@ struct drm_i915_private {
 			struct {
 				struct i915_vma *vma;
 				u8 *vaddr;
+				u32 last_ctx_id;
 				int format;
 				int format_size;
 

drivers/gpu/drm/i915/i915_drv.h struct drm_i915_private { · line 2575

Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +2 -0
@@ -2561,6 +2571,14 @@ struct drm_i915_private {
 			} oa_buffer;
 
 			u32 gen7_latched_oastatus1;
+			u32 ctx_oactxctrl_off;
+			u32 ctx_flexeu0_off;

Could we use _offset, for a second I thought were actually turning something off...

Quoted 58 lines
drivers/gpu/drm/i915/i915_drv.h +11 -0
+
+			/* The RPT_ID/reason field for Gen8+ includes a bit
+			 * to determine if the CTX ID in the report is valid
+			 * but the specific bit differs between Gen 8 and 9
+			 */
+			u32 gen8_valid_ctx_bit;
 
 			struct i915_oa_ops ops;
 			const struct i915_oa_format *oa_formats;
@@ -2871,6 +2889,8 @@ intel_info(const struct drm_i915_private *dev_priv)
 #define IS_KBL_ULX(dev_priv)	(INTEL_DEVID(dev_priv) == 0x590E || \
 				 INTEL_DEVID(dev_priv) == 0x5915 || \
 				 INTEL_DEVID(dev_priv) == 0x591E)
+#define IS_SKL_GT2(dev_priv)	(IS_SKYLAKE(dev_priv) && \
+				 (INTEL_DEVID(dev_priv) & 0x00F0) == 0x0010)
 #define IS_SKL_GT3(dev_priv)	(IS_SKYLAKE(dev_priv) && \
 				 (INTEL_DEVID(dev_priv) & 0x00F0) == 0x0020)
 #define IS_SKL_GT4(dev_priv)	(IS_SKYLAKE(dev_priv) && \
@@ -3622,6 +3642,9 @@ i915_gem_context_lookup_timeline(struct i915_gem_context *ctx,
 
 int i915_perf_open_ioctl(struct drm_device *dev, void *data,
 			 struct drm_file *file);
+void i915_oa_update_reg_state(struct intel_engine_cs *engine,
+			      struct i915_gem_context *ctx,
+			      uint32_t *reg_state);
 
 /* i915_gem_evict.c */
 int __must_check i915_gem_evict_something(struct i915_address_space *vm,
drivers/gpu/drm/i915/i915_gem_context.h +1 -0
index 4af2ab94558b..3f4ce73bea43 100644
@@ -151,6 +151,7 @@ struct i915_gem_context {
 		u64 lrc_desc;
 		int pin_count;
 		bool initialised;
+		atomic_t oa_state_dirty;
 	} engine[I915_NUM_ENGINES];
 
 	/** ring_size: size for allocating the per-engine ring buffer */
drivers/gpu/drm/i915/i915_perf.c +6 -0
index 36d07ca68029..dc5f0121e305 100644
@@ -196,6 +196,12 @@
 
 #include "i915_drv.h"
 #include "i915_oa_hsw.h"
+#include "i915_oa_bdw.h"
+#include "i915_oa_chv.h"
+#include "i915_oa_sklgt2.h"
+#include "i915_oa_sklgt3.h"
+#include "i915_oa_sklgt4.h"
+#include "i915_oa_bxt.h"
 
 /* HW requires this to be a power of two, between 128k and 16M, though driver
  * is currently generally designed assuming the largest 16M size is used such

drivers/gpu/drm/i915/i915_perf.c static u32 i915_perf_stream_paranoid = true; · line 286

Quoted 9 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
@@ -272,6 +278,13 @@ static u32 i915_perf_stream_paranoid = true;
 
 #define INVALID_CTX_ID 0xffffffff
 
+/* On Gen8+ automatically triggered OA reports include a 'reason' field... */
+#define OAREPORT_REASON_MASK           0x3f
+#define OAREPORT_REASON_SHIFT          19
+#define OAREPORT_REASON_TIMER          (1<<0)
+#define OAREPORT_REASON_CTX_SWITCH     (1<<3)
+#define OAREPORT_REASON_CLK_RATIO      (1<<5)

Would the expectation be that userspace would peek at the reason field? If so do we not want to throw this stuff in i915_drm.h?

Quoted 60 lines
drivers/gpu/drm/i915/i915_perf.c +46 -0
+
 
 /* For sysctl proc_dointvec_minmax of i915_oa_max_sample_rate
  *
@@ -303,6 +316,13 @@ static struct i915_oa_format hsw_oa_formats[I915_OA_FORMAT_MAX] = {
 	[I915_OA_FORMAT_C4_B8]	    = { 7, 64 },
 };
 
+static struct i915_oa_format gen8_plus_oa_formats[I915_OA_FORMAT_MAX] = {
+	[I915_OA_FORMAT_A12]		    = { 0, 64 },
+	[I915_OA_FORMAT_A12_B8_C8]	    = { 2, 128 },
+	[I915_OA_FORMAT_A32u40_A4u32_B8_C8] = { 5, 256 },
+	[I915_OA_FORMAT_C4_B8]		    = { 7, 64 },
+};
+
 #define SAMPLE_OA_REPORT      (1<<0)
 
 /**
@@ -333,6 +353,122 @@ struct perf_open_properties {
 };
 
 /**
+ * gen8_oa_buffer_check_unlocked - check for data and update tail ptr state
+ * @dev_priv: i915 device instance
+ *
+ * This is either called via fops (for blocking reads in user ctx) or the poll
+ * check hrtimer (atomic ctx) to check the OA buffer tail pointer and check
+ * if there is data available for userspace to read.
+ *
+ * This function is central to providing a workaround for the OA unit tail
+ * pointer having a race with respect to what data is visible to the CPU.
+ * It is responsible for reading tail pointers from the hardware and giving
+ * the pointers time to 'age' before they are made available for reading.
+ * (See description of OA_TAIL_MARGIN_NSEC above for further details.)
+ *
+ * Besides returning true when there is data available to read() this function
+ * also has the side effect of updating the oa_buffer.tails[], .aging_timestamp
+ * and .aged_tail_idx state used for reading.
+ *
+ * Note: It's safe to read OA config state here unlocked, assuming that this is
+ * only called while the stream is enabled, while the global OA configuration
+ * can't be modified.
+ *
+ * Returns: %true if the OA buffer contains data, else %false
+ */
+static bool gen8_oa_buffer_check_unlocked(struct drm_i915_private *dev_priv)
+{
+	int report_size = dev_priv->perf.oa.oa_buffer.format_size;
+	unsigned long flags;
+	unsigned int aged_idx;
+	u32 head, hw_tail, aged_tail, aging_tail;
+	u64 now;
+
+	/* We have to consider the (unlikely) possibility that read() errors
+	 * could result in an OA buffer reset which might reset the head,
+	 * tails[] and aged_tail state.
+	 */
+	spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+	/* NB: The head we observe here might effectively be a little out of

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 403

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+	 * date (between head and tails[aged_idx].offset if there is currently
+	 * a read() in progress.
+	 */
+	head = dev_priv->perf.oa.oa_buffer.head;
+
+	aged_idx = dev_priv->perf.oa.oa_buffer.aged_tail_idx;
+	aged_tail = dev_priv->perf.oa.oa_buffer.tails[aged_idx].offset;
+	aging_tail = dev_priv->perf.oa.oa_buffer.tails[!aged_idx].offset;
+
+	hw_tail = I915_READ(GEN8_OATAILPTR) & GEN8_OATAILPTR_MASK;

You didn't fancy turning this into a vfunc or something, the gen7 oa_buffer_check is identical except for the tail register business?

Quoted 263 lines
drivers/gpu/drm/i915/i915_perf.c +256 -0
+
+	/* The tail pointer increases in 64 byte increments,
+	 * not in report_size steps...
+	 */
+	hw_tail &= ~(report_size - 1);
+
+	now = ktime_get_mono_fast_ns();
+
+	/* Update the aged tail
+	 *
+	 * Flip the tail pointer available for read()s once the aging tail is
+	 * old enough to trust that the corresponding data will be visible to
+	 * the CPU...
+	 *
+	 * Do this before updating the aging pointer in case we may be able to
+	 * immediately start aging a new pointer too (if new data has become
+	 * available) without needing to wait for a later hrtimer callback.
+	 */
+	if (aging_tail != INVALID_TAIL_PTR &&
+	    ((now - dev_priv->perf.oa.oa_buffer.aging_timestamp) >
+	     OA_TAIL_MARGIN_NSEC)) {
+
+		aged_idx ^= 1;
+		dev_priv->perf.oa.oa_buffer.aged_tail_idx = aged_idx;
+
+		aged_tail = aging_tail;
+
+		/* Mark that we need a new pointer to start aging... */
+		dev_priv->perf.oa.oa_buffer.tails[!aged_idx].offset = INVALID_TAIL_PTR;
+		aging_tail = INVALID_TAIL_PTR;
+	}
+
+	/* Update the aging tail
+	 *
+	 * We throttle aging tail updates until we have a new tail that
+	 * represents >= one report more data than is already available for
+	 * reading. This ensures there will be enough data for a successful
+	 * read once this new pointer has aged and ensures we will give the new
+	 * pointer time to age.
+	 */
+	if (aging_tail == INVALID_TAIL_PTR &&
+	    (aged_tail == INVALID_TAIL_PTR ||
+	     OA_TAKEN(hw_tail, aged_tail) >= report_size)) {
+		struct i915_vma *vma = dev_priv->perf.oa.oa_buffer.vma;
+		u32 gtt_offset = i915_ggtt_offset(vma);
+
+		/* Be paranoid and do a bounds check on the pointer read back
+		 * from hardware, just in case some spurious hardware condition
+		 * could put the tail out of bounds...
+		 */
+		if (hw_tail >= gtt_offset &&
+		    hw_tail < (gtt_offset + OA_BUFFER_SIZE)) {
+			dev_priv->perf.oa.oa_buffer.tails[!aged_idx].offset =
+				aging_tail = hw_tail;
+			dev_priv->perf.oa.oa_buffer.aging_timestamp = now;
+		} else {
+			DRM_ERROR("Ignoring spurious out of range OA buffer tail pointer = %u\n",
+				  hw_tail);
+		}
+	}
+
+	spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+	return aged_tail == INVALID_TAIL_PTR ?
+		false : OA_TAKEN(aged_tail, head) >= report_size;
+}
+
+/**
  * gen7_oa_buffer_check_unlocked - check for data and update tail ptr state
  * @dev_priv: i915 device instance
  *
@@ -553,6 +689,284 @@ static int append_oa_sample(struct i915_perf_stream *stream,
  *
  * Returns: 0 on success, negative error code on failure.
  */
+static int gen8_append_oa_reports(struct i915_perf_stream *stream,
+				  char __user *buf,
+				  size_t count,
+				  size_t *offset)
+{
+	struct drm_i915_private *dev_priv = stream->dev_priv;
+	int report_size = dev_priv->perf.oa.oa_buffer.format_size;
+	u8 *oa_buf_base = dev_priv->perf.oa.oa_buffer.vaddr;
+	u32 gtt_offset = i915_ggtt_offset(dev_priv->perf.oa.oa_buffer.vma);
+	u32 mask = (OA_BUFFER_SIZE - 1);
+	size_t start_offset = *offset;
+	unsigned long flags;
+	unsigned int aged_tail_idx;
+	u32 head, tail;
+	u32 taken;
+	int ret = 0;
+
+	if (WARN_ON(!stream->enabled))
+		return -EIO;
+
+	spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+	head = dev_priv->perf.oa.oa_buffer.head;
+	aged_tail_idx = dev_priv->perf.oa.oa_buffer.aged_tail_idx;
+	tail = dev_priv->perf.oa.oa_buffer.tails[aged_tail_idx].offset;
+
+	spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+	/* An invalid tail pointer here means we're still waiting for the poll
+	 * hrtimer callback to give us a pointer
+	 */
+	if (tail == INVALID_TAIL_PTR)
+		return -EAGAIN;
+
+	/* NB: oa_buffer.head/tail include the gtt_offset which we don't want
+	 * while indexing relative to oa_buf_base.
+	 */
+	head -= gtt_offset;
+	tail -= gtt_offset;
+
+	/* An out of bounds or misaligned head or tail pointer implies a driver
+	 * bug since we validate + align the tail pointers we read from the
+	 * hardware and we are in full control of the head pointer which should
+	 * only be incremented by multiples of the report size (notably also
+	 * all a power of two).
+	 */
+	if (WARN_ONCE(head > OA_BUFFER_SIZE || head % report_size ||
+		      tail > OA_BUFFER_SIZE || tail % report_size,
+		      "Inconsistent OA buffer pointers: head = %u, tail = %u\n",
+		      head, tail))
+		return -EIO;
+
+
+	for (/* none */;
+	     (taken = OA_TAKEN(tail, head));
+	     head = (head + report_size) & mask) {
+		u8 *report = oa_buf_base + head;
+		u32 *report32 = (void *)report;
+		u32 ctx_id;
+		u32 reason;
+
+		/* All the report sizes factor neatly into the buffer
+		 * size so we never expect to see a report split
+		 * between the beginning and end of the buffer.
+		 *
+		 * Given the initial alignment check a misalignment
+		 * here would imply a driver bug that would result
+		 * in an overrun.
+		 */
+		if (WARN_ON((OA_BUFFER_SIZE - head) < report_size)) {
+			DRM_ERROR("Spurious OA head ptr: non-integral report offset\n");
+			break;
+		}
+
+		/* The reason field includes flags identifying what
+		 * triggered this specific report (mostly timer
+		 * triggered or e.g. due to a context switch).
+		 *
+		 * This field is never expected to be zero so we can
+		 * check that the report isn't invalid before copying
+		 * it to userspace...
+		 */
+		reason = ((report32[0] >> OAREPORT_REASON_SHIFT) &
+			  OAREPORT_REASON_MASK);
+		if (reason == 0) {
+			if (__ratelimit(&dev_priv->perf.oa.spurious_report_rs))
+				DRM_NOTE("Skipping spurious, invalid OA report\n");
+			continue;
+		}
+
+		/* XXX: Just keep the lower 21 bits for now since I'm not
+		 * entirely sure if the HW touches any of the higher bits in
+		 * this field
+		 */
+		ctx_id = report32[2] & 0x1fffff;
+
+		/* Squash whatever is in the CTX_ID field if it's
+		 * marked as invalid to be sure we avoid
+		 * false-positive, single-context filtering below...
+		 */
+		if (!(report32[0] & dev_priv->perf.oa.gen8_valid_ctx_bit))
+			ctx_id = report32[2] = INVALID_CTX_ID;
+
+		/* NB: For Gen 8 the OA unit no longer supports clock gating
+		 * off for a specific context and the kernel can't securely
+		 * stop the counters from updating as system-wide / global
+		 * values.
+		 *
+		 * Automatic reports now include a context ID so reports can be
+		 * filtered on the cpu but it's not worth trying to
+		 * automatically subtract/hide counter progress for other
+		 * contexts while filtering since we can't stop userspace
+		 * issuing MI_REPORT_PERF_COUNT commands which would still
+		 * provide a side-band view of the real values.
+		 *
+		 * To allow userspace (such as Mesa/GL_INTEL_performance_query)
+		 * to normalize counters for a single filtered context then it
+		 * needs be forwarded bookend context-switch reports so that it
+		 * can track switches in between MI_REPORT_PERF_COUNT commands
+		 * and can itself subtract/ignore the progress of counters
+		 * associated with other contexts. Note that the hardware
+		 * automatically triggers reports when switching to a new
+		 * context which are tagged with the ID of the newly active
+		 * context. To avoid the complexity (and likely fragility) of
+		 * reading ahead while parsing reports to try and minimize
+		 * forwarding redundant context switch reports (i.e. between
+		 * other, unrelated contexts) we simply elect to forward them
+		 * all.
+		 *
+		 * We don't rely solely on the reason field to identify context
+		 * switches since it's not-uncommon for periodic samples to
+		 * identify a switch before any 'context switch' report.
+		 */
+		if (!dev_priv->perf.oa.exclusive_stream->ctx ||
+		    dev_priv->perf.oa.specific_ctx_id == ctx_id ||
+		    (dev_priv->perf.oa.oa_buffer.last_ctx_id ==
+		     dev_priv->perf.oa.specific_ctx_id) ||
+		    reason & OAREPORT_REASON_CTX_SWITCH) {
+
+			/* While filtering for a single context we avoid
+			 * leaking the IDs of other contexts.
+			 */
+			if (dev_priv->perf.oa.exclusive_stream->ctx &&
+			    dev_priv->perf.oa.specific_ctx_id != ctx_id) {
+				report32[2] = INVALID_CTX_ID;
+			}
+
+			ret = append_oa_sample(stream, buf, count, offset,
+					       report);
+			if (ret)
+				break;
+
+			dev_priv->perf.oa.oa_buffer.last_ctx_id = ctx_id;
+		}
+
+		/* The above reason field sanity check is based on
+		 * the assumption that the OA buffer is initially
+		 * zeroed and we reset the field after copying so the
+		 * check is still meaningful once old reports start
+		 * being overwritten.
+		 */
+		report32[0] = 0;
+	}
+
+	if (start_offset != *offset) {
+		spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+		/* We removed the gtt_offset for the copy loop above, indexing
+		 * relative to oa_buf_base so put back here...
+		 */
+		head += gtt_offset;
+
+		I915_WRITE(GEN8_OAHEADPTR, head & GEN8_OAHEADPTR_MASK);
+		dev_priv->perf.oa.oa_buffer.head = head;
+
+		spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+	}
+
+	return ret;
+}
+
+/**
+ * gen8_oa_read - copy status records then buffered OA reports
+ * @stream: An i915-perf stream opened for OA metrics
+ * @buf: destination buffer given by userspace
+ * @count: the number of bytes userspace wants to read
+ * @offset: (inout): the current position for writing into @buf
+ *

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream, · line 889

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+ * Checks OA unit status registers and if necessary appends corresponding
+ * status records for userspace (such as for a buffer full condition) and then
+ * initiate appending any buffered OA reports.
+ *
+ * Updates @offset according to the number of bytes successfully copied into
+ * the userspace buffer.
+ *
+ * NB: some data may be successfully copied to the userspace buffer
+ * even if an error is returned, and this is reflected in the
+ * updated @read_state.

updated @read_state ?

Quoted 256 lines
drivers/gpu/drm/i915/i915_perf.c +187 -27
+ *
+ * Returns: zero on success or a negative error code
+ */
+static int gen8_oa_read(struct i915_perf_stream *stream,
+			char __user *buf,
+			size_t count,
+			size_t *offset)
+{
+	struct drm_i915_private *dev_priv = stream->dev_priv;
+	u32 oastatus;
+	int ret;
+
+	if (WARN_ON(!dev_priv->perf.oa.oa_buffer.vaddr))
+		return -EIO;
+
+	oastatus = I915_READ(GEN8_OASTATUS);
+
+	/* We treat OABUFFER_OVERFLOW as a significant error:
+	 *
+	 * Although theoretically we could handle this more gracefully
+	 * sometimes, some Gens don't correctly suppress certain
+	 * automatically triggered reports in this condition and so we
+	 * have to assume that old reports are now being trampled
+	 * over.
+	 *
+	 * Considering how we don't currently give userspace control
+	 * over the OA buffer size and always configure a large 16MB
+	 * buffer, then a buffer overflow does anyway likely indicate
+	 * that something has gone quite badly wrong.
+	 */
+	if (oastatus & GEN8_OASTATUS_OABUFFER_OVERFLOW) {
+		ret = append_oa_status(stream, buf, count, offset,
+				       DRM_I915_PERF_RECORD_OA_BUFFER_LOST);
+		if (ret)
+			return ret;
+
+		DRM_DEBUG("OA buffer overflow (exponent = %d): force restart\n",
+			  dev_priv->perf.oa.period_exponent);
+
+		dev_priv->perf.oa.ops.oa_disable(dev_priv);
+		dev_priv->perf.oa.ops.oa_enable(dev_priv);
+
+		/* Note: .oa_enable() is expected to re-init the oabuffer
+		 * and reset GEN8_OASTATUS for us
+		 */
+		oastatus = I915_READ(GEN8_OASTATUS);
+	}
+
+	if (oastatus & GEN8_OASTATUS_REPORT_LOST) {
+		ret = append_oa_status(stream, buf, count, offset,
+				       DRM_I915_PERF_RECORD_OA_REPORT_LOST);
+		if (ret == 0) {
+			I915_WRITE(GEN8_OASTATUS,
+				   oastatus & ~GEN8_OASTATUS_REPORT_LOST);
+		}
+	}
+
+	return gen8_append_oa_reports(stream, buf, count, offset);
+}
+
+/**
+ * Copies all buffered OA reports into userspace read() buffer.
+ * @stream: An i915-perf stream opened for OA metrics
+ * @buf: destination buffer given by userspace
+ * @count: the number of bytes userspace wants to read
+ * @offset: (inout): the current position for writing into @buf
+ *
+ * Notably any error condition resulting in a short read (-%ENOSPC or
+ * -%EFAULT) will be returned even though one or more records may
+ * have been successfully copied. In this case it's up to the caller
+ * to decide if the error should be squashed before returning to
+ * userspace.
+ *
+ * Note: reports are consumed from the head, and appended to the
+ * tail, so the tail chases the head?... If you think that's mad
+ * and back-to-front you're not alone, but this follows the
+ * Gen PRM naming convention.
+ *
+ * Returns: 0 on success, negative error code on failure.
+ */
 static int gen7_append_oa_reports(struct i915_perf_stream *stream,
 				  char __user *buf,
 				  size_t count,
@@ -733,7 +1147,8 @@ static int gen7_oa_read(struct i915_perf_stream *stream,
 		if (ret)
 			return ret;
 
-		DRM_DEBUG("OA buffer overflow: force restart\n");
+		DRM_DEBUG("OA buffer overflow (exponent = %d): force restart\n",
+			  dev_priv->perf.oa.period_exponent);
 
 		dev_priv->perf.oa.ops.oa_disable(dev_priv);
 		dev_priv->perf.oa.ops.oa_enable(dev_priv);
@@ -833,33 +1248,37 @@ static int i915_oa_read(struct i915_perf_stream *stream,
 static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
 {
 	struct drm_i915_private *dev_priv = stream->dev_priv;
-	struct intel_engine_cs *engine = dev_priv->engine[RCS];
-	int ret;
 
-	ret = i915_mutex_lock_interruptible(&dev_priv->drm);
-	if (ret)
-		return ret;
+	if (i915.enable_execlists)
+		dev_priv->perf.oa.specific_ctx_id = stream->ctx->hw_id;
+	else {
+		struct intel_engine_cs *engine = dev_priv->engine[RCS];
+		int ret;
 
-	/* As the ID is the gtt offset of the context's vma we pin
-	 * the vma to ensure the ID remains fixed.
-	 *
-	 * NB: implied RCS engine...
-	 */
-	ret = engine->context_pin(engine, stream->ctx);
-	if (ret)
-		goto unlock;
+		ret = i915_mutex_lock_interruptible(&dev_priv->drm);
+		if (ret)
+			return ret;
 
-	/* Explicitly track the ID (instead of calling i915_ggtt_offset()
-	 * on the fly) considering the difference with gen8+ and
-	 * execlists
-	 */
-	dev_priv->perf.oa.specific_ctx_id =
-		i915_ggtt_offset(stream->ctx->engine[engine->id].state);
+		/* As the ID is the gtt offset of the context's vma we pin
+		 * the vma to ensure the ID remains fixed.
+		 */
+		ret = engine->context_pin(engine, stream->ctx);
+		if (ret) {
+			mutex_unlock(&dev_priv->drm.struct_mutex);
+			return ret;
+		}
 
-unlock:
-	mutex_unlock(&dev_priv->drm.struct_mutex);
+		/* Explicitly track the ID (instead of calling
+		 * i915_ggtt_offset() on the fly) considering the difference
+		 * with gen8+ and execlists
+		 */
+		dev_priv->perf.oa.specific_ctx_id =
+			i915_ggtt_offset(stream->ctx->engine[engine->id].state);
 
-	return ret;
+		mutex_unlock(&dev_priv->drm.struct_mutex);
+	}
+
+	return 0;
 }
 
 /**
@@ -874,12 +1293,16 @@ static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
 	struct drm_i915_private *dev_priv = stream->dev_priv;
 	struct intel_engine_cs *engine = dev_priv->engine[RCS];
 
-	mutex_lock(&dev_priv->drm.struct_mutex);
+	if (i915.enable_execlists) {
+		dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID;
+	} else {
+		mutex_lock(&dev_priv->drm.struct_mutex);
 
-	dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID;
-	engine->context_unpin(engine, stream->ctx);
+		dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID;
+		engine->context_unpin(engine, stream->ctx);
 
-	mutex_unlock(&dev_priv->drm.struct_mutex);
+		mutex_unlock(&dev_priv->drm.struct_mutex);
+	}
 }
 
 static void
@@ -964,6 +1387,56 @@ static void gen7_init_oa_buffer(struct drm_i915_private *dev_priv)
 	dev_priv->perf.oa.pollin = false;
 }
 
+static void gen8_init_oa_buffer(struct drm_i915_private *dev_priv)
+{
+	u32 gtt_offset = i915_ggtt_offset(dev_priv->perf.oa.oa_buffer.vma);
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+	I915_WRITE(GEN8_OASTATUS, 0);
+	I915_WRITE(GEN8_OAHEADPTR, gtt_offset);
+	dev_priv->perf.oa.oa_buffer.head = gtt_offset;
+
+	I915_WRITE(GEN8_OABUFFER_UDW, 0);
+
+	/* PRM says:
+	 *
+	 *  "This MMIO must be set before the OATAILPTR
+	 *  register and after the OAHEADPTR register. This is
+	 *  to enable proper functionality of the overflow
+	 *  bit."
+	 */
+	I915_WRITE(GEN8_OABUFFER, gtt_offset |
+		   OABUFFER_SIZE_16M | OA_MEM_SELECT_GGTT);
+	I915_WRITE(GEN8_OATAILPTR, gtt_offset & GEN8_OATAILPTR_MASK);
+
+	/* Mark that we need updated tail pointers to read from... */
+	dev_priv->perf.oa.oa_buffer.tails[0].offset = INVALID_TAIL_PTR;
+	dev_priv->perf.oa.oa_buffer.tails[1].offset = INVALID_TAIL_PTR;
+
+	spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+
+	/* NB: although the OA buffer will initially be allocated
+	 * zeroed via shmfs (and so this memset is redundant when
+	 * first allocating), we may re-init the OA buffer, either
+	 * when re-enabling a stream or in error/reset paths.
+	 *
+	 * The reason we clear the buffer for each re-init is for the
+	 * sanity check in gen8_append_oa_reports() that looks at the
+	 * reason field to make sure it's non-zero which relies on
+	 * the assumption that new reports are being written to zeroed
+	 * memory...
+	 */
+	memset(dev_priv->perf.oa.oa_buffer.vaddr, 0, OA_BUFFER_SIZE);
+
+	/* Maybe make ->pollin per-stream state if we support multiple
+	 * concurrent streams in the future.
+	 */
+	dev_priv->perf.oa.pollin = false;
+}
+
 static int alloc_oa_buffer(struct drm_i915_private *dev_priv)
 {
 	struct drm_i915_gem_object *bo;
@@ -1108,6 +1581,200 @@ static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)
 				      ~GT_NOA_ENABLE));
 }
 
+/*
+ * From Broadwell PRM, 3D-Media-GPGPU -> Register State Context
+ *
+ * MMIO reads or writes to any of the registers listed in the
+ * “Register State Context image” subsections through HOST/IA
+ * MMIO interface for debug purposes must follow the steps below:
+ *
+ * - SW should set the Force Wakeup bit to prevent GT from entering C6.
+ * - Write 0x2050[31:0] = 0x00010001 (disable sequence).
+ * - Disable IDLE messaging in CS (Write 0x2050[31:0] = 0x00010001).
+ * - BDW:  Poll/Wait for register bits of 0x22AC[6:0] turn to 0x30 value.
+ * - SKL+: Poll/Wait for register bits of 0x22A4[6:0] turn to 0x30 value.
+ * - Read/Write to desired MMIO registers.
+ * - Enable IDLE messaging in CS (Write 0x2050[31:0] = 0x00010000).
+ * - Force Wakeup bit should be reset to enable C6 entry.
+ *
+ * XXX: don't nest or overlap calls to this API, it has no ref
+ * counting to track how many entities require the RCS to be
+ * blocked from being idle.
+ */
+static int gen8_begin_ctx_mmio(struct drm_i915_private *dev_priv)

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv) · line 1614

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+{
+	i915_reg_t fsm_reg = dev_priv->info.gen > 8 ?
+		GEN9_RCS_FE_FSM2 : GEN6_RCS_PWR_FSM;
+	int ret = 0;
+
+	/* There's only no active context while idle in execlist mode
+	 * (though we shouldn't be using this in any other case)
+	 */
+	if (WARN_ON(!i915.enable_execlists))
+		return ret;

return -EINVAL ?

Quoted 75 lines
drivers/gpu/drm/i915/i915_perf.c +75 -0
+
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_RENDER);
+
+	I915_WRITE(GEN6_RC_SLEEP_PSMI_CONTROL,
+		   _MASKED_BIT_ENABLE(GEN6_PSMI_SLEEP_MSG_DISABLE));
+
+	/* Note: we don't currently have a good handle on the maximum
+	 * latency for this wake up so while we only need to hold rcs
+	 * busy from process context we at least keep the waiting
+	 * interruptible...
+	 */
+	ret = wait_for((I915_READ(fsm_reg) & 0x3f) == 0x30, 50);
+	if (ret)
+	    intel_uncore_forcewake_put(dev_priv, FORCEWAKE_RENDER);
+
+	return ret;
+}
+
+static void gen8_end_ctx_mmio(struct drm_i915_private *dev_priv)
+{
+	if (WARN_ON(!i915.enable_execlists))
+		return;
+
+	I915_WRITE(GEN6_RC_SLEEP_PSMI_CONTROL,
+		   _MASKED_BIT_DISABLE(GEN6_PSMI_SLEEP_MSG_DISABLE));
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_RENDER);
+}
+
+/* Manages updating the per-context aspects of the OA stream
+ * configuration across all contexts.
+ *
+ * The awkward consideration here is that OACTXCONTROL controls the
+ * exponent for periodic sampling which is primarily used for system
+ * wide profiling where we'd like a consistent sampling period even in
+ * the face of context switches.
+ *
+ * Our approach of updating the register state context (as opposed to
+ * say using a workaround batch buffer) ensures that the hardware
+ * won't automatically reload an out-of-date timer exponent even
+ * transiently before a WA BB could be parsed.
+ *
+ * This function needs to:
+ * - Ensure the currently running context's per-context OA state is
+ *   updated
+ * - Ensure that all existing contexts will have the correct per-context
+ *   OA state if they are scheduled for use.
+ * - Ensure any new contexts will be initialized with the correct
+ *   per-context OA state.
+ *
+ * Note: it's only the RCS/Render context that has any OA state.
+ */
+static int configure_all_contexts(struct drm_i915_private *dev_priv)
+{
+	struct i915_gem_context *ctx;
+	int ret;
+
+	ret = i915_mutex_lock_interruptible(&dev_priv->drm);
+	if (ret)
+		return ret;
+
+	/* Since execlist submission may be happening asynchronously here then
+	 * we first mark existing contexts dirty before we update the current
+	 * context so if any switches happen in the middle we can expect
+	 * that the act of scheduling will have itself ensured a consistent
+	 * OA state update.
+	 */
+	list_for_each_entry(ctx, &dev_priv->context_list, link) {
+		/* The actual update of the register state context will happen
+		 * the next time this logical ring is submitted. (See
+		 * i915_oa_update_reg_state() which hooks into
+		 * execlists_update_context())
+		 */
+		atomic_set(&ctx->engine[RCS].oa_state_dirty, 1);
+	}
+

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv) · line 1699

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+
+	/* Now update the current context.
+	 *
+	 * Note: Using MMIO to update per-context registers requires
+	 * some extra care...
+	 */
+	ret = gen8_begin_ctx_mmio(dev_priv);
+	if (ret) {
+		DRM_ERROR("Failed to bring RCS out of idle to update current ctx OA state");

Missing newline for DRM_ERROR.

Quoted 59 lines
drivers/gpu/drm/i915/i915_perf.c +59 -0
+		return ret;
+	}
+
+	I915_WRITE(GEN8_OACTXCONTROL, ((dev_priv->perf.oa.period_exponent <<
+					GEN8_OA_TIMER_PERIOD_SHIFT) |
+				      (dev_priv->perf.oa.periodic ?
+				       GEN8_OA_TIMER_ENABLE : 0) |
+				      GEN8_OA_COUNTER_RESUME));
+
+	config_oa_regs(dev_priv, dev_priv->perf.oa.flex_regs,
+			dev_priv->perf.oa.flex_regs_len);
+
+	gen8_end_ctx_mmio(dev_priv);
+
+	return 0;
+}
+
+static int gen8_enable_metric_set(struct drm_i915_private *dev_priv)
+{
+	int ret = dev_priv->perf.oa.ops.select_metric_set(dev_priv);
+
+	if (ret)
+		return ret;
+
+	/* We disable slice/unslice clock ratio change reports on SKL since
+	 * they are too noisy. The HW generates a lot of redundant reports
+	 * where the ratio hasn't really changed causing a lot of redundant
+	 * work to processes and increasing the chances we'll hit buffer
+	 * overruns.
+	 *
+	 * Although we don't currently use the 'disable overrun' OABUFFER
+	 * feature it's worth noting that clock ratio reports have to be
+	 * disabled before considering to use that feature since the HW doesn't
+	 * correctly block these reports.
+	 *
+	 * Currently none of the high-level metrics we have depend on knowing
+	 * this ratio to normalize.
+	 *
+	 * Note: This register is not power context saved and restored, but
+	 * that's OK considering that we disable RC6 while the OA unit is
+	 * enabled.
+	 *
+	 * The _INCLUDE_CLK_RATIO bit allows the slice/unslice frequency to
+	 * be read back from automatically triggered reports, as part of the
+	 * RPT_ID field.
+	 */
+	if (IS_SKYLAKE(dev_priv) || IS_BROXTON(dev_priv)) {
+		I915_WRITE(GEN8_OA_DEBUG,
+			   _MASKED_BIT_ENABLE(GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS |
+					      GEN9_OA_DEBUG_INCLUDE_CLK_RATIO));
+	}
+
+	I915_WRITE(GDT_CHICKEN_BITS, 0xA0);
+	config_oa_regs(dev_priv, dev_priv->perf.oa.mux_regs,
+		       dev_priv->perf.oa.mux_regs_len);
+	I915_WRITE(GDT_CHICKEN_BITS, 0x80);
+
+	/* It takes a fairly long time for a new MUX configuration to
+	 * be be applied after these register writes. This delay

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv) · line 1768

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+	 * duration is take from Haswell (derived empirically based on
+	 * the render_basic config) but hopefully it covers the
+	 * maximum configuration latency for Gen8+ too...
+	 */
+	usleep_range(15000, 20000);
+
+	config_oa_regs(dev_priv, dev_priv->perf.oa.b_counter_regs,
+		       dev_priv->perf.oa.b_counter_regs_len);
+
+	configure_all_contexts(dev_priv);

Check the return value here, and bubble up.

Quoted 57 lines
drivers/gpu/drm/i915/i915_perf.c +37 -0
+
+	return 0;
+}
+
+static void gen8_disable_metric_set(struct drm_i915_private *dev_priv)
+{
+	/* NOP */
+}
+
 static void gen7_update_oacontrol_locked(struct drm_i915_private *dev_priv)
 {
 	lockdep_assert_held(&dev_priv->perf.hook_lock);
@@ -1152,6 +1819,29 @@ static void gen7_oa_enable(struct drm_i915_private *dev_priv)
 	spin_unlock_irqrestore(&dev_priv->perf.hook_lock, flags);
 }
 
+static void gen8_oa_enable(struct drm_i915_private *dev_priv)
+{
+	u32 report_format = dev_priv->perf.oa.oa_buffer.format;
+
+	/* Reset buf pointers so we don't forward reports from before now.
+	 *
+	 * Think carefully if considering trying to avoid this, since it
+	 * also ensures status flags and the buffer itself are cleared
+	 * in error paths, and we have checks for invalid reports based
+	 * on the assumption that certain fields are written to zeroed
+	 * memory which this helps maintains.
+	 */
+	gen8_init_oa_buffer(dev_priv);
+
+	/* Note: we don't rely on the hardware to perform single context
+	 * filtering and instead filter on the cpu based on the context-id
+	 * field of reports
+	 */
+	I915_WRITE(GEN8_OACONTROL, (report_format <<
+				    GEN8_OA_REPORT_FORMAT_SHIFT) |
+				   GEN8_OA_COUNTER_ENABLE);
+}
+
 /**
  * i915_oa_stream_enable - handle `I915_PERF_IOCTL_ENABLE` for OA stream
  * @stream: An i915 perf stream opened for OA metrics
@@ -1178,6 +1868,11 @@ static void gen7_oa_disable(struct drm_i915_private *dev_priv)
 	I915_WRITE(GEN7_OACONTROL, 0);
 }
 
+static void gen8_oa_disable(struct drm_i915_private *dev_priv)
+{
+	I915_WRITE(GEN8_OACONTROL, 0);
+}
+
 /**
  * i915_oa_stream_disable - handle `I915_PERF_IOCTL_DISABLE` for OA stream
  * @stream: An i915 perf stream opened for OA metrics
@@ -1336,6 +2031,88 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 	return ret;
 }

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream, · line 2042

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
 
+/* NB: It must always remain pointer safe to run this even if the OA unit
+ * has been disabled.
+ *
+ * It's fine to put out-of-date values into these per-context registers
+ * in the case that the OA unit has been disabled.
+ */
+static void gen8_update_reg_state_unlocked(struct intel_engine_cs *engine,
+					   struct i915_gem_context *ctx,
+					   uint32_t *reg_state)

u32 for consistency, elsewhere also.

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream, · line 2048

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+{
+	struct drm_i915_private *dev_priv = ctx->i915;
+	const struct i915_oa_reg *flex_regs = dev_priv->perf.oa.flex_regs;
+	int n_flex_regs = dev_priv->perf.oa.flex_regs_len;
+	int ctx_oactxctrl = dev_priv->perf.oa.ctx_oactxctrl_off;
+	int ctx_flexeu0 = dev_priv->perf.oa.ctx_flexeu0_off;

Can we also make these u32, for consistency.

Quoted 216 lines
drivers/gpu/drm/i915/i915_perf.c +137 -27
+	/* The MMIO offsets for Flex EU registers aren't contiguous */
+	u32 flex_mmio[] = {
+		i915_mmio_reg_offset(EU_PERF_CNTL0),
+		i915_mmio_reg_offset(EU_PERF_CNTL1),
+		i915_mmio_reg_offset(EU_PERF_CNTL2),
+		i915_mmio_reg_offset(EU_PERF_CNTL3),
+		i915_mmio_reg_offset(EU_PERF_CNTL4),
+		i915_mmio_reg_offset(EU_PERF_CNTL5),
+		i915_mmio_reg_offset(EU_PERF_CNTL6),
+	};
+	int i;
+
+	reg_state[ctx_oactxctrl] = i915_mmio_reg_offset(GEN8_OACTXCONTROL);
+	reg_state[ctx_oactxctrl+1] = (dev_priv->perf.oa.period_exponent <<
+				      GEN8_OA_TIMER_PERIOD_SHIFT) |
+				     (dev_priv->perf.oa.periodic ?
+				      GEN8_OA_TIMER_ENABLE : 0) |
+				     GEN8_OA_COUNTER_RESUME;
+
+	for (i = 0; i < ARRAY_SIZE(flex_mmio); i++) {
+		uint32_t state_offset = ctx_flexeu0 + i * 2;
+		uint32_t mmio = flex_mmio[i];
+
+		/* This arbitrary default will select the 'EU FPU0 Pipeline
+		 * Active' event. In the future it's anticipated that there
+		 * will be an explicit 'No Event' we can select, but not yet...
+		 */
+		uint32_t value = 0;
+		int j;
+
+		for (j = 0; j < n_flex_regs; j++) {
+			if (i915_mmio_reg_offset(flex_regs[j].addr) == mmio) {
+				value = flex_regs[j].value;
+				break;
+			}
+		}
+
+		reg_state[state_offset] = mmio;
+		reg_state[state_offset+1] = value;
+	}
+}
+
+void i915_oa_update_reg_state(struct intel_engine_cs *engine,
+			      struct i915_gem_context *ctx,
+			      uint32_t *reg_state)
+{
+	struct drm_i915_private *dev_priv = engine->i915;
+
+	if (engine->id != RCS)
+		return;
+
+	if (!dev_priv->perf.initialized)
+		return;
+
+	/* XXX: We don't take a lock here and this may run async with
+	 * respect to stream methods. Notably we don't want to block
+	 * context switches by long i915 perf read() operations.
+	 *
+	 * It's expected to always be safe to read the dev_priv->perf
+	 * state needed here, and expected to be benign to redundantly
+	 * update the state if the OA unit has been disabled since
+	 * oa_state_dirty was last set.
+	 */
+	if (atomic_cmpxchg(&ctx->engine[engine->id].oa_state_dirty, 1, 0))
+		gen8_update_reg_state_unlocked(engine, ctx, reg_state);
+}
+
 /**
  * i915_perf_read_locked - &i915_perf_stream_ops->read with error normalisation
  * @stream: An i915 perf stream
@@ -1750,6 +2527,7 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv,
 	struct i915_gem_context *specific_ctx = NULL;
 	struct i915_perf_stream *stream = NULL;
 	unsigned long f_flags = 0;
+	bool privileged_op = true;
 	int stream_fd;
 	int ret;
 
@@ -1767,12 +2545,28 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv,
 		}
 	}
 
+	/* On Haswell the OA unit supports clock gating off for a specific
+	 * context and in this mode there's no visibility of metrics for the
+	 * rest of the system, which we consider acceptable for a
+	 * non-privileged client.
+	 *
+	 * For Gen8+ the OA unit no longer supports clock gating off for a
+	 * specific context and the kernel can't securely stop the counters
+	 * from updating as system-wide / global values. Even though we can
+	 * filter reports based on the included context ID we can't block
+	 * clients from seeing the raw / global counter values via
+	 * MI_REPORT_PERF_COUNT commands and so consider it a privileged op to
+	 * enable the OA unit by default.
+	 */
+	if (IS_HASWELL(dev_priv) && specific_ctx)
+		privileged_op = false;
+
 	/* Similar to perf's kernel.perf_paranoid_cpu sysctl option
 	 * we check a dev.i915.perf_stream_paranoid sysctl option
 	 * to determine if it's ok to access system wide OA counters
 	 * without CAP_SYS_ADMIN privileges.
 	 */
-	if (!specific_ctx &&
+	if (privileged_op &&
 	    i915_perf_stream_paranoid && !capable(CAP_SYS_ADMIN)) {
 		DRM_DEBUG("Insufficient privileges to open system-wide i915 perf stream\n");
 		ret = -EACCES;
@@ -2039,9 +2833,6 @@ int i915_perf_open_ioctl(struct drm_device *dev, void *data,
  */
 void i915_perf_register(struct drm_i915_private *dev_priv)
 {
-	if (!IS_HASWELL(dev_priv))
-		return;
-
 	if (!dev_priv->perf.initialized)
 		return;
 
@@ -2057,11 +2848,38 @@ void i915_perf_register(struct drm_i915_private *dev_priv)
 	if (!dev_priv->perf.metrics_kobj)
 		goto exit;
 
-	if (i915_perf_register_sysfs_hsw(dev_priv)) {
-		kobject_put(dev_priv->perf.metrics_kobj);
-		dev_priv->perf.metrics_kobj = NULL;
+	if (IS_HASWELL(dev_priv)) {
+		if (i915_perf_register_sysfs_hsw(dev_priv))
+			goto sysfs_error;
+	} else if (IS_BROADWELL(dev_priv)) {
+		if (i915_perf_register_sysfs_bdw(dev_priv))
+			goto sysfs_error;
+	} else if (IS_CHERRYVIEW(dev_priv)) {
+		if (i915_perf_register_sysfs_chv(dev_priv))
+			goto sysfs_error;
+	} else if (IS_SKYLAKE(dev_priv)) {
+		if (IS_SKL_GT2(dev_priv)) {
+			if (i915_perf_register_sysfs_sklgt2(dev_priv))
+				goto sysfs_error;
+		} else if (IS_SKL_GT3(dev_priv)) {
+			if (i915_perf_register_sysfs_sklgt3(dev_priv))
+				goto sysfs_error;
+		} else if (IS_SKL_GT4(dev_priv)) {
+			if (i915_perf_register_sysfs_sklgt4(dev_priv))
+				goto sysfs_error;
+		} else
+			goto sysfs_error;
+	} else if (IS_BROXTON(dev_priv)) {
+		if (i915_perf_register_sysfs_bxt(dev_priv))
+			goto sysfs_error;
 	}
 
+	goto exit;
+
+sysfs_error:
+	kobject_put(dev_priv->perf.metrics_kobj);
+	dev_priv->perf.metrics_kobj = NULL;
+
 exit:
 	mutex_unlock(&dev_priv->perf.lock);
 }
@@ -2077,13 +2895,24 @@ void i915_perf_register(struct drm_i915_private *dev_priv)
  */
 void i915_perf_unregister(struct drm_i915_private *dev_priv)
 {
-	if (!IS_HASWELL(dev_priv))
-		return;
-
 	if (!dev_priv->perf.metrics_kobj)
 		return;
 
-	i915_perf_unregister_sysfs_hsw(dev_priv);
+        if (IS_HASWELL(dev_priv))
+                i915_perf_unregister_sysfs_hsw(dev_priv);
+        else if (IS_BROADWELL(dev_priv))
+                i915_perf_unregister_sysfs_bdw(dev_priv);
+        else if (IS_CHERRYVIEW(dev_priv))
+                i915_perf_unregister_sysfs_chv(dev_priv);
+        else if (IS_SKYLAKE(dev_priv)) {
+		if (IS_SKL_GT2(dev_priv))
+			i915_perf_unregister_sysfs_sklgt2(dev_priv);
+		else if (IS_SKL_GT3(dev_priv))
+			i915_perf_unregister_sysfs_sklgt3(dev_priv);
+		else if (IS_SKL_GT4(dev_priv))
+			i915_perf_unregister_sysfs_sklgt4(dev_priv);
+	} else if (IS_BROXTON(dev_priv))
+                i915_perf_unregister_sysfs_bxt(dev_priv);
 
 	kobject_put(dev_priv->perf.metrics_kobj);
 	dev_priv->perf.metrics_kobj = NULL;
@@ -2142,45 +2971,107 @@ static struct ctl_table dev_root[] = {
  */
 void i915_perf_init(struct drm_i915_private *dev_priv)
 {
-	if (!IS_HASWELL(dev_priv))
-		return;
-
-	/* Using the same limiting factors as printk_ratelimit() */
-	ratelimit_state_init(&dev_priv->perf.oa.spurious_report_rs,
-			5 * HZ, 10);
-	/* We use a DRM_NOTE for spurious reports so it would be
-	 * inconsistent to print a warning for throttling.
-	 */
-	ratelimit_set_flags(&dev_priv->perf.oa.spurious_report_rs,
-			RATELIMIT_MSG_ON_RELEASE);
-
-	hrtimer_init(&dev_priv->perf.oa.poll_check_timer,
-		     CLOCK_MONOTONIC, HRTIMER_MODE_REL);
-	dev_priv->perf.oa.poll_check_timer.function = oa_poll_check_timer_cb;
-	init_waitqueue_head(&dev_priv->perf.oa.poll_wq);
+	dev_priv->perf.oa.n_builtin_sets = 0;
+
+	if (IS_HASWELL(dev_priv)) {
+		dev_priv->perf.oa.ops.init_oa_buffer = gen7_init_oa_buffer;
+		dev_priv->perf.oa.ops.enable_metric_set = hsw_enable_metric_set;
+		dev_priv->perf.oa.ops.disable_metric_set = hsw_disable_metric_set;
+		dev_priv->perf.oa.ops.oa_enable = gen7_oa_enable;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 2990

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+		dev_priv->perf.oa.ops.oa_disable = gen7_oa_disable;
+		dev_priv->perf.oa.ops.read = gen7_oa_read;
+		dev_priv->perf.oa.ops.oa_buffer_check =
+			gen7_oa_buffer_check_unlocked;
+
+		dev_priv->perf.oa.oa_formats = hsw_oa_formats;
+
+		dev_priv->perf.oa.n_builtin_sets =
+			i915_oa_n_builtin_metric_sets_hsw;
+	} else if (i915.enable_execlists) {

Maybe a comment for why we don't support legacy mode with gen8+, for the next reader.

Quoted 83 lines
drivers/gpu/drm/i915/i915_perf.c +67 -13
+		if (IS_GEN8(dev_priv)) {
+			dev_priv->perf.oa.ctx_oactxctrl_off = 0x120;
+			dev_priv->perf.oa.ctx_flexeu0_off = 0x2ce;
+			dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<25);
+
+			if (IS_BROADWELL(dev_priv)) {
+				dev_priv->perf.oa.n_builtin_sets =
+					i915_oa_n_builtin_metric_sets_bdw;
+				dev_priv->perf.oa.ops.select_metric_set =
+					i915_oa_select_metric_set_bdw;
+			} else if (IS_CHERRYVIEW(dev_priv)) {
+				dev_priv->perf.oa.n_builtin_sets =
+					i915_oa_n_builtin_metric_sets_chv;
+				dev_priv->perf.oa.ops.select_metric_set =
+					i915_oa_select_metric_set_chv;
+			}
+		} else if (IS_GEN9(dev_priv)) {
+			dev_priv->perf.oa.ctx_oactxctrl_off = 0x128;
+			dev_priv->perf.oa.ctx_flexeu0_off = 0x3de;
+			dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16);
+
+			if (IS_SKL_GT2(dev_priv)) {
+				dev_priv->perf.oa.n_builtin_sets =
+					i915_oa_n_builtin_metric_sets_sklgt2;
+				dev_priv->perf.oa.ops.select_metric_set =
+					i915_oa_select_metric_set_sklgt2;
+			} else if (IS_SKL_GT3(dev_priv)) {
+				dev_priv->perf.oa.n_builtin_sets =
+					i915_oa_n_builtin_metric_sets_sklgt3;
+				dev_priv->perf.oa.ops.select_metric_set =
+					i915_oa_select_metric_set_sklgt3;
+			} else if (IS_SKL_GT4(dev_priv)) {
+				dev_priv->perf.oa.n_builtin_sets =
+					i915_oa_n_builtin_metric_sets_sklgt4;
+				dev_priv->perf.oa.ops.select_metric_set =
+					i915_oa_select_metric_set_sklgt4;
+			} else if (IS_BROXTON(dev_priv)) {
+				dev_priv->perf.oa.n_builtin_sets =
+					i915_oa_n_builtin_metric_sets_bxt;
+				dev_priv->perf.oa.ops.select_metric_set =
+					i915_oa_select_metric_set_bxt;
+			}
+		}
 
-	INIT_LIST_HEAD(&dev_priv->perf.streams);
-	mutex_init(&dev_priv->perf.lock);
-	spin_lock_init(&dev_priv->perf.hook_lock);
-	spin_lock_init(&dev_priv->perf.oa.oa_buffer.ptr_lock);
+		if (dev_priv->perf.oa.n_builtin_sets) {
+			dev_priv->perf.oa.ops.init_oa_buffer = gen8_init_oa_buffer;
+			dev_priv->perf.oa.ops.enable_metric_set =
+				gen8_enable_metric_set;
+			dev_priv->perf.oa.ops.disable_metric_set =
+				gen8_disable_metric_set;
+			dev_priv->perf.oa.ops.oa_enable = gen8_oa_enable;
+			dev_priv->perf.oa.ops.oa_disable = gen8_oa_disable;
+			dev_priv->perf.oa.ops.read = gen8_oa_read;
+			dev_priv->perf.oa.ops.oa_buffer_check =
+				gen8_oa_buffer_check_unlocked;
+
+			dev_priv->perf.oa.oa_formats = gen8_plus_oa_formats;
+		}
+	}
 
-	dev_priv->perf.oa.ops.init_oa_buffer = gen7_init_oa_buffer;
-	dev_priv->perf.oa.ops.enable_metric_set = hsw_enable_metric_set;
-	dev_priv->perf.oa.ops.disable_metric_set = hsw_disable_metric_set;
-	dev_priv->perf.oa.ops.oa_enable = gen7_oa_enable;
-	dev_priv->perf.oa.ops.oa_disable = gen7_oa_disable;
-	dev_priv->perf.oa.ops.read = gen7_oa_read;
-	dev_priv->perf.oa.ops.oa_buffer_check =
-		gen7_oa_buffer_check_unlocked;
+	if (dev_priv->perf.oa.n_builtin_sets) {
+		/* Using the same limiting factors as printk_ratelimit() */
+		ratelimit_state_init(&dev_priv->perf.oa.spurious_report_rs,
+				5 * HZ, 10);
+		/* We use a DRM_NOTE for spurious reports so it would be
+		 * inconsistent to print a warning for throttling.
+		 */
+		ratelimit_set_flags(&dev_priv->perf.oa.spurious_report_rs,
+				RATELIMIT_MSG_ON_RELEASE);
 
-	dev_priv->perf.oa.oa_formats = hsw_oa_formats;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 3068

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +7 -2
+		hrtimer_init(&dev_priv->perf.oa.poll_check_timer,
+				CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+		dev_priv->perf.oa.poll_check_timer.function = oa_poll_check_timer_cb;
+		init_waitqueue_head(&dev_priv->perf.oa.poll_wq);
 
-	dev_priv->perf.oa.n_builtin_sets =
-		i915_oa_n_builtin_metric_sets_hsw;
+		INIT_LIST_HEAD(&dev_priv->perf.streams);
+		mutex_init(&dev_priv->perf.lock);
+		spin_lock_init(&dev_priv->perf.hook_lock);

Not strictly related to this patch, but why do we still need perf.hook_lock?

Quoted 141 lines
drivers/gpu/drm/i915/i915_perf.c +5 -2
+		spin_lock_init(&dev_priv->perf.oa.oa_buffer.ptr_lock);
 
-	dev_priv->perf.sysctl_header = register_sysctl_table(dev_root);
+		dev_priv->perf.sysctl_header = register_sysctl_table(dev_root);
 
-	dev_priv->perf.initialized = true;
+		dev_priv->perf.initialized = true;
+	}
 }
 
 /**
@@ -2200,5 +3091,6 @@ void i915_perf_fini(struct drm_i915_private *dev_priv)
 	unregister_sysctl_table(dev_priv->perf.sysctl_header);
 
 	memset(&dev_priv->perf.oa.ops, 0, sizeof(dev_priv->perf.oa.ops));
+
 	dev_priv->perf.initialized = false;
 }
drivers/gpu/drm/i915/i915_reg.h +22 -0
index 04c8f69fcc62..0052289ed8ad 100644
@@ -645,6 +645,12 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 
 #define GEN8_OACTXID _MMIO(0x2364)
 
+#define GEN8_OA_DEBUG _MMIO(0x2B04)
+#define  GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS    (1<<5)
+#define  GEN9_OA_DEBUG_INCLUDE_CLK_RATIO	    (1<<6)
+#define  GEN9_OA_DEBUG_DISABLE_GO_1_0_REPORTS	    (1<<2)
+#define  GEN9_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS   (1<<1)
+
 #define GEN8_OACONTROL _MMIO(0x2B00)
 #define  GEN8_OA_REPORT_FORMAT_A12	    (0<<2)
 #define  GEN8_OA_REPORT_FORMAT_A12_B8_C8    (2<<2)
@@ -666,6 +672,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define  GEN7_OABUFFER_STOP_RESUME_ENABLE   (1<<1)
 #define  GEN7_OABUFFER_RESUME		    (1<<0)
 
+#define GEN8_OABUFFER_UDW _MMIO(0x23b4)
 #define GEN8_OABUFFER _MMIO(0x2b14)
 
 #define GEN7_OASTATUS1 _MMIO(0x2364)
@@ -684,7 +691,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define  GEN8_OASTATUS_REPORT_LOST	    (1<<0)
 
 #define GEN8_OAHEADPTR _MMIO(0x2B0C)
+#define GEN8_OAHEADPTR_MASK    0xffffffc0
 #define GEN8_OATAILPTR _MMIO(0x2B10)
+#define GEN8_OATAILPTR_MASK    0xffffffc0
 
 #define OABUFFER_SIZE_128K  (0<<3)
 #define OABUFFER_SIZE_256K  (1<<3)
@@ -697,7 +706,17 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 
 #define OA_MEM_SELECT_GGTT  (1<<0)
 
+/*
+ * Flexible, Aggregate EU Counter Registers.
+ * Note: these aren't contiguous
+ */
 #define EU_PERF_CNTL0	    _MMIO(0xe458)
+#define EU_PERF_CNTL1	    _MMIO(0xe558)
+#define EU_PERF_CNTL2	    _MMIO(0xe658)
+#define EU_PERF_CNTL3	    _MMIO(0xe758)
+#define EU_PERF_CNTL4	    _MMIO(0xe45c)
+#define EU_PERF_CNTL5	    _MMIO(0xe55c)
+#define EU_PERF_CNTL6	    _MMIO(0xe65c)
 
 #define GDT_CHICKEN_BITS    _MMIO(0x9840)
 #define GT_NOA_ENABLE	    0x00000080
@@ -2317,6 +2336,9 @@ enum skl_disp_power_wells {
 #define   GEN8_RC_SEMA_IDLE_MSG_DISABLE	(1 << 12)
 #define   GEN8_FF_DOP_CLOCK_GATE_DISABLE	(1<<10)
 
+#define GEN6_RCS_PWR_FSM _MMIO(0x22ac)
+#define GEN9_RCS_FE_FSM2 _MMIO(0x22a4)
+
 /* Fuse readout registers for GT */
 #define CHV_FUSE_GT			_MMIO(VLV_DISPLAY_BASE + 0x2168)
 #define   CHV_FGT_DISABLE_SS0		(1 << 10)
drivers/gpu/drm/i915/intel_lrc.c +5 -0
index eec1e714f531..e6d9e4197d3d 100644
@@ -337,6 +337,8 @@ static u64 execlists_update_context(struct drm_i915_gem_request *rq)
 	if (ppgtt && !i915_vm_is_48bit(&ppgtt->base))
 		execlists_update_context_pdps(ppgtt, reg_state);
 
+	i915_oa_update_reg_state(rq->engine, rq->ctx, reg_state);
+
 	return ce->lrc_desc;
 }
 
@@ -1878,6 +1880,9 @@ static void execlists_init_reg_state(u32 *regs,
 		regs[CTX_LRI_HEADER_2] = MI_LOAD_REGISTER_IMM(1);
 		CTX_REG(regs, CTX_R_PWR_CLK_STATE, GEN8_R_PWR_CLK_STATE,
 			make_rpcs(dev_priv));
+
+		atomic_set(&ctx->engine[RCS].oa_state_dirty, 1);
+		i915_oa_update_reg_state(engine, ctx, regs);
 	}
 }
 
include/uapi/drm/i915_drm.h +12 -7
index e0599e729e68..03b833849919 100644
@@ -1295,13 +1295,18 @@ struct drm_i915_gem_context_param {
 };
 
 enum drm_i915_oa_format {
-	I915_OA_FORMAT_A13 = 1,
-	I915_OA_FORMAT_A29,
-	I915_OA_FORMAT_A13_B8_C8,
-	I915_OA_FORMAT_B4_C8,
-	I915_OA_FORMAT_A45_B8_C8,
-	I915_OA_FORMAT_B4_C8_A16,
-	I915_OA_FORMAT_C4_B8,
+	I915_OA_FORMAT_A13 = 1,	    /* HSW only */
+	I915_OA_FORMAT_A29,	    /* HSW only */
+	I915_OA_FORMAT_A13_B8_C8,   /* HSW only */
+	I915_OA_FORMAT_B4_C8,	    /* HSW only */
+	I915_OA_FORMAT_A45_B8_C8,   /* HSW only */
+	I915_OA_FORMAT_B4_C8_A16,   /* HSW only */
+	I915_OA_FORMAT_C4_B8,	    /* HSW+ */
+
+	/* Gen8+ */
+	I915_OA_FORMAT_A12,
+	I915_OA_FORMAT_A12_B8_C8,
+	I915_OA_FORMAT_A32u40_A4u32_B8_C8,
 
 	I915_OA_FORMAT_MAX	    /* non-ABI */
 };
Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#12 lore raw

Re: [PATCH v2 4/5] drm/i915: Add OA unit support for Gen 8+

Quoted 66 lines · Matthew Auld
Quoted 47 lines · Robert Bragg

Enables access to OA unit metrics for BDW, CHV, SKL and BXT which all share (more-or-less) the same OA unit design.

Of particular note in comparison to Haswell: some OA unit HW config state has become per-context state and as a consequence it is somewhat more complicated to manage synchronous state changes from the cpu while there's no guarantee of what context (if any) is currently actively running on the gpu.

The periodic sampling frequency which can be particularly useful for system-wide analysis (as opposed to command stream synchronised MI_REPORT_PERF_COUNT commands) is perhaps the most surprising state to have become per-context save and restored (while the OABUFFER destination is still a shared, system-wide resource).

This support for gen8+ takes care to consider a number of timing challenges involved in synchronously updating per-context state primarily by programming all config state from the cpu and updating all current and saved contexts synchronously while the OA unit is still disabled.

The driver intentionally avoids depending on command streamer programming to update OA state considering the lack of synchronization between the automatic loading of OACTXCONTROL state (that includes the periodic sampling state and enable state) on context restore and the parsing of any general purpose BB the driver can control. I.e. this implementation is careful to avoid the possibility of a context restore temporarily enabling any out-of-date periodic sampling state. In addition to the risk of transiently-out-of-date state being loaded automatically; there are also internal HW latencies involved in the loading of MUX configurations which would be difficult to account for from the command streamer (and we only want to enable the unit when once the MUX configuration is complete).

Since the Gen8+ OA unit design no longer supports clock gating the unit off for a single given context (which effectively stopped any progress of counters while any other context was running) and instead supports tagging OA reports with a context ID for filtering on the CPU, it means we can no longer hide the system-wide progress of counters from a non-privileged application only interested in metrics for its own context. Although we could theoretically try and subtract the progress of other contexts before forwarding reports via read() we aren't in a position to filter reports captured via MI_REPORT_PERF_COUNT commands. As a result, for Gen8+, we always require the dev.i915.perf_stream_paranoid to be unset for any access to OA metrics if not root.

Quoted 5 lines
---
 drivers/gpu/drm/i915/i915_drv.h         |   29 +-
 drivers/gpu/drm/i915/i915_gem_context.h |    1 +
 drivers/gpu/drm/i915/i915_perf.c        | 1034

++++++++++++++++++++++++++++---

Review

Quoted 6 lines
 drivers/gpu/drm/i915/i915_reg.h         |   22 +
 drivers/gpu/drm/i915/intel_lrc.c        |    5 +
 include/uapi/drm/i915_drm.h             |   19 +-
 6 files changed, 1029 insertions(+), 81 deletions(-)
(quoted diff) +0 -0

b/drivers/gpu/drm/i915/i915_drv.h

drivers/gpu/drm/i915/i915_drv.h intel_uncore_forcewake_for_reg(struct

Quoted 4 lines

index c4156a8a5dc0..190e699d5851 100644

drivers/gpu/drm/i915/i915_drv.h +0 -0
@@ -731,6 +731,7 @@ intel_uncore_forcewake_for_reg(struct

drm_i915_private *dev_priv,

Quoted 6 lines

drivers/gpu/drm/i915/i915_drv.h intel_uncore_forcewake_for_reg(struct drm_i915_private *dev_priv, · line 734

Quoted 4 lines
drivers/gpu/drm/i915/i915_drv.h +1 -0
                             i915_reg_t reg, unsigned int op);
 
 struct intel_uncore_funcs {
+     int (*wait_for_rcs_busy)(struct drm_i915_private *dev_priv);

No longer needed.

Ah, yup, removed now.

Quoted 38 lines

drivers/gpu/drm/i915/i915_drv.h struct i915_oa_ops { · line 2088

Quoted 12 lines
drivers/gpu/drm/i915/i915_drv.h +1 -3
      void (*force_wake_get)(struct drm_i915_private *dev_priv,
                             enum forcewake_domains domains);
      void (*force_wake_put)(struct drm_i915_private *dev_priv,
@@ -2084,9 +2085,17 @@ struct i915_oa_ops {
      void (*init_oa_buffer)(struct drm_i915_private *dev_priv);
 
      /**
-      * @enable_metric_set: Applies any MUX configuration to set up the
-      * Boolean and Custom (B/C) counters that are part of the counter
-      * reports being sampled. May apply system constraints such as
+      * @select_metric_set: The auto generated code that checks whether

a

drivers/gpu/drm/i915/i915_drv.h struct i915_oa_ops { · line 2089

Quoted 1 line
drivers/gpu/drm/i915/i915_drv.h +1 -0
+      * requested OA config is applicable to the system and if so sets

up

drivers/gpu/drm/i915/i915_drv.h struct i915_oa_ops { · line 2090

Quoted 1 line
drivers/gpu/drm/i915/i915_drv.h +1 -0
+      * the mux, oa and flex eu register config pointers according to

the

drivers/gpu/drm/i915/i915_drv.h struct i915_oa_ops {

Quoted 6 lines
drivers/gpu/drm/i915/i915_drv.h +6 -0
+      * current dev_priv->perf.oa.metrics_set.
+      */
+     int (*select_metric_set)(struct drm_i915_private *dev_priv);
+
+     /**
+      * @enable_metric_set: Selects and applies any MUX configuration

to set

drivers/gpu/drm/i915/i915_drv.h struct i915_oa_ops { · line 2098

Quoted 2 lines
drivers/gpu/drm/i915/i915_drv.h +2 -0
+      * up the Boolean and Custom (B/C) counters that are part of the
+      * counter reports being sampled. May apply system constraints

such as

Quoted 11 lines
drivers/gpu/drm/i915/i915_drv.h +1 -0
       * disabling EU clock gating as required.
       */
      int (*enable_metric_set)(struct drm_i915_private *dev_priv);
@@ -2492,6 +2501,7 @@ struct drm_i915_private {
                      struct {
                              struct i915_vma *vma;
                              u8 *vaddr;
+                             u32 last_ctx_id;
                              int format;
                              int format_size;
 
Quoted 9 lines

drivers/gpu/drm/i915/i915_drv.h struct drm_i915_private { · line 2575

Quoted 6 lines
drivers/gpu/drm/i915/i915_drv.h +2 -0
@@ -2561,6 +2571,14 @@ struct drm_i915_private {
                      } oa_buffer;
 
                      u32 gen7_latched_oastatus1;
+                     u32 ctx_oactxctrl_off;
+                     u32 ctx_flexeu0_off;

Could we use _offset, for a second I thought were actually turning something off...

Yeah, sounds better.

Quoted 73 lines

drivers/gpu/drm/i915/i915_drv.h struct drm_i915_private {

Quoted 4 lines
drivers/gpu/drm/i915/i915_drv.h +3 -0
+
+                     /* The RPT_ID/reason field for Gen8+ includes a bit
+                      * to determine if the CTX ID in the report is

valid

drivers/gpu/drm/i915/i915_drv.h intel_info(const struct drm_i915_private *dev_priv) · line 2893

Quoted 12 lines
drivers/gpu/drm/i915/i915_drv.h +5 -0
+                      * but the specific bit differs between Gen 8 and 9
+                      */
+                     u32 gen8_valid_ctx_bit;
 
                      struct i915_oa_ops ops;
                      const struct i915_oa_format *oa_formats;
@@ -2871,6 +2889,8 @@ intel_info(const struct drm_i915_private *dev_priv)
 #define IS_KBL_ULX(dev_priv) (INTEL_DEVID(dev_priv) == 0x590E || \
                               INTEL_DEVID(dev_priv) == 0x5915 || \
                               INTEL_DEVID(dev_priv) == 0x591E)
+#define IS_SKL_GT2(dev_priv) (IS_SKYLAKE(dev_priv) && \
+                              (INTEL_DEVID(dev_priv) & 0x00F0) ==

0x0010)

drivers/gpu/drm/i915/i915_drv.h intel_info(const struct drm_i915_private *dev_priv) · line 2895

Quoted 2 lines
drivers/gpu/drm/i915/i915_drv.h +0 -0
 #define IS_SKL_GT3(dev_priv) (IS_SKYLAKE(dev_priv) && \
                               (INTEL_DEVID(dev_priv) & 0x00F0) ==

0x0020)

drivers/gpu/drm/i915/i915_drv.h i915_gem_context_lookup_timeline(struct

Quoted 2 lines
drivers/gpu/drm/i915/i915_drv.h +0 -0
 #define IS_SKL_GT4(dev_priv) (IS_SKYLAKE(dev_priv) && \
@@ -3622,6 +3642,9 @@ i915_gem_context_lookup_timeline(struct

i915_gem_context *ctx,

drivers/gpu/drm/i915/i915_drv.h i915_gem_context_lookup_timeline(struct i915_gem_context *ctx,

Quoted 9 lines
drivers/gpu/drm/i915/i915_drv.h +3 -0
 int i915_perf_open_ioctl(struct drm_device *dev, void *data,
                       struct drm_file *file);
+void i915_oa_update_reg_state(struct intel_engine_cs *engine,
+                           struct i915_gem_context *ctx,
+                           uint32_t *reg_state);
 
 /* i915_gem_evict.c */
 int __must_check i915_gem_evict_something(struct i915_address_space

*vm,

Review

Quoted 1 line
(quoted diff) +0 -0

b/drivers/gpu/drm/i915/i915_gem_context.h

Review

Quoted 12 lines

index 4af2ab94558b..3f4ce73bea43 100644

drivers/gpu/drm/i915/i915_gem_context.h +1 -0
@@ -151,6 +151,7 @@ struct i915_gem_context {
              u64 lrc_desc;
              int pin_count;
              bool initialised;
+             atomic_t oa_state_dirty;
      } engine[I915_NUM_ENGINES];
 
      /** ring_size: size for allocating the per-engine ring buffer */
(quoted diff) +0 -0

b/drivers/gpu/drm/i915/i915_perf.c

Quoted 5 lines

index 36d07ca68029..dc5f0121e305 100644

drivers/gpu/drm/i915/i915_perf.c +0 -0
@@ -196,6 +196,12 @@
 

drivers/gpu/drm/i915/i915_perf.c line 206

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
 #include "i915_drv.h"
 #include "i915_oa_hsw.h"
+#include "i915_oa_bdw.h"
+#include "i915_oa_chv.h"
+#include "i915_oa_sklgt2.h"
+#include "i915_oa_sklgt3.h"
+#include "i915_oa_sklgt4.h"
+#include "i915_oa_bxt.h"
 
 /* HW requires this to be a power of two, between 128k and 16M, though

driver

drivers/gpu/drm/i915/i915_perf.c line 207

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +0 -0
  * is currently generally designed assuming the largest 16M size is

used such

drivers/gpu/drm/i915/i915_perf.c static u32 i915_perf_stream_paranoid = true; · line 281

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
@@ -272,6 +278,13 @@ static u32 i915_perf_stream_paranoid = true;
 
 #define INVALID_CTX_ID 0xffffffff
 
+/* On Gen8+ automatically triggered OA reports include a 'reason'

field... */

Quoted 8 lines

drivers/gpu/drm/i915/i915_perf.c static u32 i915_perf_stream_paranoid = true; · line 286

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+#define OAREPORT_REASON_MASK           0x3f
+#define OAREPORT_REASON_SHIFT          19
+#define OAREPORT_REASON_TIMER          (1<<0)
+#define OAREPORT_REASON_CTX_SWITCH     (1<<3)
+#define OAREPORT_REASON_CLK_RATIO      (1<<5)

Would the expectation be that userspace would peek at the reason field? If so do we not want to throw this stuff in i915_drm.h?

I consider this part of the hardware interface documented in the PRMs for userspace to reference, not software ABI. Maybe a grey area but it doesn't seem to me that it should be in i915_drm.h.

Quoted 79 lines

drivers/gpu/drm/i915/i915_perf.c static struct i915_oa_format

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
+
 
 /* For sysctl proc_dointvec_minmax of i915_oa_max_sample_rate
  *
@@ -303,6 +316,13 @@ static struct i915_oa_format

hsw_oa_formats[I915_OA_FORMAT_MAX] = {

drivers/gpu/drm/i915/i915_perf.c static struct i915_oa_format hsw_oa_formats[I915_OA_FORMAT_MAX] = {

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
      [I915_OA_FORMAT_C4_B8]      = { 7, 64 },
 };
 
+static struct i915_oa_format gen8_plus_oa_formats[I915_OA_FORMAT_MAX]

= {

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+     [I915_OA_FORMAT_A12]                = { 0, 64 },
+     [I915_OA_FORMAT_A12_B8_C8]          = { 2, 128 },
+     [I915_OA_FORMAT_A32u40_A4u32_B8_C8] = { 5, 256 },
+     [I915_OA_FORMAT_C4_B8]              = { 7, 64 },
+};
+
 #define SAMPLE_OA_REPORT      (1<<0)
 
 /**

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 356

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
@@ -333,6 +353,122 @@ struct perf_open_properties {
 };
 
 /**
+ * gen8_oa_buffer_check_unlocked - check for data and update tail ptr

state

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+ * @dev_priv: i915 device instance
+ *
+ * This is either called via fops (for blocking reads in user ctx) or

the poll

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 360

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * check hrtimer (atomic ctx) to check the OA buffer tail pointer and

check

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+ * if there is data available for userspace to read.
+ *
+ * This function is central to providing a workaround for the OA unit

tail

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 364

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * pointer having a race with respect to what data is visible to the

CPU.

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 365

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * It is responsible for reading tail pointers from the hardware and

giving

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 366

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * the pointers time to 'age' before they are made available for

reading.

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+ * (See description of OA_TAIL_MARGIN_NSEC above for further details.)
+ *
+ * Besides returning true when there is data available to read() this

function

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 370

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * also has the side effect of updating the oa_buffer.tails[],

.aging_timestamp

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+ * and .aged_tail_idx state used for reading.
+ *
+ * Note: It's safe to read OA config state here unlocked, assuming that

this is

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 374

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * only called while the stream is enabled, while the global OA

configuration

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+ * can't be modified.
+ *
+ * Returns: %true if the OA buffer contains data, else %false
+ */
+static bool gen8_oa_buffer_check_unlocked(struct drm_i915_private

*dev_priv)

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 8 lines
drivers/gpu/drm/i915/i915_perf.c +8 -0
+{
+     int report_size = dev_priv->perf.oa.oa_buffer.format_size;
+     unsigned long flags;
+     unsigned int aged_idx;
+     u32 head, hw_tail, aged_tail, aging_tail;
+     u64 now;
+
+     /* We have to consider the (unlikely) possibility that read()

errors

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+      * could result in an OA buffer reset which might reset the head,
+      * tails[] and aged_tail state.
+      */
+     spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+     /* NB: The head we observe here might effectively be a little out

of

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 394

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+      * date (between head and tails[aged_idx].offset if there is

currently

Quoted 12 lines

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 403

Quoted 9 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
+      * a read() in progress.
+      */
+     head = dev_priv->perf.oa.oa_buffer.head;
+
+     aged_idx = dev_priv->perf.oa.oa_buffer.aged_tail_idx;
+     aged_tail = dev_priv->perf.oa.oa_buffer.tails[aged_idx].offset;
+     aging_tail = dev_priv->perf.oa.oa_buffer.tails[!aged_idx].offset;
+
+     hw_tail = I915_READ(GEN8_OATAILPTR) & GEN8_OATAILPTR_MASK;

You didn't fancy turning this into a vfunc or something, the gen7 oa_buffer_check is identical except for the tail register business?

Yeah, it had definitly crossed my mind to do this, and I'd been considering doing this as a follow up, but it's probably worth just doing upfront instead. It's not great to have a copy of this fiddly workaround code.

Okey, just made the change locally and will send out in a bit.

Quoted 309 lines

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 12 lines
drivers/gpu/drm/i915/i915_perf.c +11 -0
+
+     /* The tail pointer increases in 64 byte increments,
+      * not in report_size steps...
+      */
+     hw_tail &= ~(report_size - 1);
+
+     now = ktime_get_mono_fast_ns();
+
+     /* Update the aged tail
+      *
+      * Flip the tail pointer available for read()s once the aging tail

is

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 415

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+      * old enough to trust that the corresponding data will be visible

to

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+      * the CPU...
+      *
+      * Do this before updating the aging pointer in case we may be

able to

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 419

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+      * immediately start aging a new pointer too (if new data has

become

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 456

Quoted 13 lines
drivers/gpu/drm/i915/i915_perf.c +13 -0
+      * available) without needing to wait for a later hrtimer callback.
+      */
+     if (aging_tail != INVALID_TAIL_PTR &&
+         ((now - dev_priv->perf.oa.oa_buffer.aging_timestamp) >
+          OA_TAIL_MARGIN_NSEC)) {
+
+             aged_idx ^= 1;
+             dev_priv->perf.oa.oa_buffer.aged_tail_idx = aged_idx;
+
+             aged_tail = aging_tail;
+
+             /* Mark that we need a new pointer to start aging... */
+             dev_priv->perf.oa.oa_buffer.tails[!aged_idx].offset =

INVALID_TAIL_PTR;

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 9 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
+             aging_tail = INVALID_TAIL_PTR;
+     }
+
+     /* Update the aging tail
+      *
+      * We throttle aging tail updates until we have a new tail that
+      * represents >= one report more data than is already available for
+      * reading. This ensures there will be enough data for a successful
+      * read once this new pointer has aged and ensures we will give

the new

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 9 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
+      * pointer time to age.
+      */
+     if (aging_tail == INVALID_TAIL_PTR &&
+         (aged_tail == INVALID_TAIL_PTR ||
+          OA_TAKEN(hw_tail, aged_tail) >= report_size)) {
+             struct i915_vma *vma = dev_priv->perf.oa.oa_buffer.vma;
+             u32 gtt_offset = i915_ggtt_offset(vma);
+
+             /* Be paranoid and do a bounds check on the pointer read

back

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties { · line 451

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+              * from hardware, just in case some spurious hardware

condition

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+              * could put the tail out of bounds...
+              */
+             if (hw_tail >= gtt_offset &&
+                 hw_tail < (gtt_offset + OA_BUFFER_SIZE)) {
+                     dev_priv->perf.oa.oa_buffer.tails[!aged_idx].offset

=

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +4 -0
+                             aging_tail = hw_tail;
+                     dev_priv->perf.oa.oa_buffer.aging_timestamp = now;
+             } else {
+                     DRM_ERROR("Ignoring spurious out of range OA

buffer tail pointer = %u\n",

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+                               hw_tail);
+             }
+     }
+
+     spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock,

flags);

drivers/gpu/drm/i915/i915_perf.c struct perf_open_properties {

Quoted 7 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+
+     return aged_tail == INVALID_TAIL_PTR ?
+             false : OA_TAKEN(aged_tail, head) >= report_size;
+}
+
+/**
  * gen7_oa_buffer_check_unlocked - check for data and update tail ptr

state

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +0 -0
  * @dev_priv: i915 device instance
  *
@@ -553,6 +689,284 @@ static int append_oa_sample(struct

i915_perf_stream *stream,

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 12 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
  *
  * Returns: 0 on success, negative error code on failure.
  */
+static int gen8_append_oa_reports(struct i915_perf_stream *stream,
+                               char __user *buf,
+                               size_t count,
+                               size_t *offset)
+{
+     struct drm_i915_private *dev_priv = stream->dev_priv;
+     int report_size = dev_priv->perf.oa.oa_buffer.format_size;
+     u8 *oa_buf_base = dev_priv->perf.oa.oa_buffer.vaddr;
+     u32 gtt_offset = i915_ggtt_offset(dev_priv->

perf.oa.oa_buffer.vma);

Quoted 8 lines
drivers/gpu/drm/i915/i915_perf.c +8 -0
+     u32 mask = (OA_BUFFER_SIZE - 1);
+     size_t start_offset = *offset;
+     unsigned long flags;
+     unsigned int aged_tail_idx;
+     u32 head, tail;
+     u32 taken;
+     int ret = 0;
+

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+     if (WARN_ON(!stream->enabled))
+             return -EIO;
+
+     spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+     head = dev_priv->perf.oa.oa_buffer.head;
+     aged_tail_idx = dev_priv->perf.oa.oa_buffer.aged_tail_idx;
+     tail = dev_priv->perf.oa.oa_buffer.tails[aged_tail_idx].offset;
+
+     spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock,

flags);

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 720

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +2 -0
+
+     /* An invalid tail pointer here means we're still waiting for the

poll

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+      * hrtimer callback to give us a pointer
+      */
+     if (tail == INVALID_TAIL_PTR)
+             return -EAGAIN;
+
+     /* NB: oa_buffer.head/tail include the gtt_offset which we don't

want

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+      * while indexing relative to oa_buf_base.
+      */
+     head -= gtt_offset;
+     tail -= gtt_offset;
+
+     /* An out of bounds or misaligned head or tail pointer implies a

driver

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 734

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +2 -0
+      * bug since we validate + align the tail pointers we read from the
+      * hardware and we are in full control of the head pointer which

should

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 735

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+      * only be incremented by multiples of the report size (notably

also

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+      * all a power of two).
+      */
+     if (WARN_ONCE(head > OA_BUFFER_SIZE || head % report_size ||
+                   tail > OA_BUFFER_SIZE || tail % report_size,
+                   "Inconsistent OA buffer pointers: head = %u, tail =

%u\n",

Quoted 12 lines
drivers/gpu/drm/i915/i915_perf.c +12 -0
+                   head, tail))
+             return -EIO;
+
+
+     for (/* none */;
+          (taken = OA_TAKEN(tail, head));
+          head = (head + report_size) & mask) {
+             u8 *report = oa_buf_base + head;
+             u32 *report32 = (void *)report;
+             u32 ctx_id;
+             u32 reason;
+

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+             /* All the report sizes factor neatly into the buffer
+              * size so we never expect to see a report split
+              * between the beginning and end of the buffer.
+              *
+              * Given the initial alignment check a misalignment
+              * here would imply a driver bug that would result
+              * in an overrun.
+              */
+             if (WARN_ON((OA_BUFFER_SIZE - head) < report_size)) {
+                     DRM_ERROR("Spurious OA head ptr: non-integral

report offset\n");

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+                     break;
+             }
+
+             /* The reason field includes flags identifying what
+              * triggered this specific report (mostly timer

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+              * triggered or e.g. due to a context switch).
+              *
+              * This field is never expected to be zero so we can
+              * check that the report isn't invalid before copying
+              * it to userspace...
+              */
+             reason = ((report32[0] >> OAREPORT_REASON_SHIFT) &
+                       OAREPORT_REASON_MASK);
+             if (reason == 0) {
+                     if (__ratelimit(&dev_priv->perf.

oa.spurious_report_rs))

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 778

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+                             DRM_NOTE("Skipping spurious, invalid OA

report\n");

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+                     continue;
+             }
+
+             /* XXX: Just keep the lower 21 bits for now since I'm not
+              * entirely sure if the HW touches any of the higher bits

in

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 12 lines
drivers/gpu/drm/i915/i915_perf.c +12 -0
+              * this field
+              */
+             ctx_id = report32[2] & 0x1fffff;
+
+             /* Squash whatever is in the CTX_ID field if it's
+              * marked as invalid to be sure we avoid
+              * false-positive, single-context filtering below...
+              */
+             if (!(report32[0] & dev_priv->perf.oa.gen8_valid_ctx_bit))
+                     ctx_id = report32[2] = INVALID_CTX_ID;
+
+             /* NB: For Gen 8 the OA unit no longer supports clock

gating

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+              * off for a specific context and the kernel can't securely
+              * stop the counters from updating as system-wide / global
+              * values.
+              *
+              * Automatic reports now include a context ID so reports

can be

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 7 lines
drivers/gpu/drm/i915/i915_perf.c +7 -0
+              * filtered on the cpu but it's not worth trying to
+              * automatically subtract/hide counter progress for other
+              * contexts while filtering since we can't stop userspace
+              * issuing MI_REPORT_PERF_COUNT commands which would still
+              * provide a side-band view of the real values.
+              *
+              * To allow userspace (such as Mesa/GL_INTEL_performance_

query)

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 808

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+              * to normalize counters for a single filtered context

then it

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 809

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+              * needs be forwarded bookend context-switch reports so

that it

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 810

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+              * can track switches in between MI_REPORT_PERF_COUNT

commands

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+              * and can itself subtract/ignore the progress of counters
+              * associated with other contexts. Note that the hardware
+              * automatically triggers reports when switching to a new
+              * context which are tagged with the ID of the newly active
+              * context. To avoid the complexity (and likely fragility)

of

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 817

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +2 -0
+              * reading ahead while parsing reports to try and minimize
+              * forwarding redundant context switch reports (i.e.

between

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 818

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+              * other, unrelated contexts) we simply elect to forward

them

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+              * all.
+              *
+              * We don't rely solely on the reason field to identify

context

Quoted 26 lines
drivers/gpu/drm/i915/i915_perf.c +26 -0
+              * switches since it's not-uncommon for periodic samples to
+              * identify a switch before any 'context switch' report.
+              */
+             if (!dev_priv->perf.oa.exclusive_stream->ctx ||
+                 dev_priv->perf.oa.specific_ctx_id == ctx_id ||
+                 (dev_priv->perf.oa.oa_buffer.last_ctx_id ==
+                  dev_priv->perf.oa.specific_ctx_id) ||
+                 reason & OAREPORT_REASON_CTX_SWITCH) {
+
+                     /* While filtering for a single context we avoid
+                      * leaking the IDs of other contexts.
+                      */
+                     if (dev_priv->perf.oa.exclusive_stream->ctx &&
+                         dev_priv->perf.oa.specific_ctx_id != ctx_id) {
+                             report32[2] = INVALID_CTX_ID;
+                     }
+
+                     ret = append_oa_sample(stream, buf, count, offset,
+                                            report);
+                     if (ret)
+                             break;
+
+                     dev_priv->perf.oa.oa_buffer.last_ctx_id = ctx_id;
+             }
+
+             /* The above reason field sanity check is based on

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+              * the assumption that the OA buffer is initially
+              * zeroed and we reset the field after copying so the
+              * check is still meaningful once old reports start
+              * being overwritten.
+              */
+             report32[0] = 0;
+     }
+
+     if (start_offset != *offset) {
+             spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock,

flags);

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 859

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +2 -0
+
+             /* We removed the gtt_offset for the copy loop above,

indexing

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 8 lines
drivers/gpu/drm/i915/i915_perf.c +8 -0
+              * relative to oa_buf_base so put back here...
+              */
+             head += gtt_offset;
+
+             I915_WRITE(GEN8_OAHEADPTR, head & GEN8_OAHEADPTR_MASK);
+             dev_priv->perf.oa.oa_buffer.head = head;
+
+             spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock,

flags);

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 13 lines
drivers/gpu/drm/i915/i915_perf.c +13 -0
+     }
+
+     return ret;
+}
+
+/**
+ * gen8_oa_read - copy status records then buffered OA reports
+ * @stream: An i915-perf stream opened for OA metrics
+ * @buf: destination buffer given by userspace
+ * @count: the number of bytes userspace wants to read
+ * @offset: (inout): the current position for writing into @buf
+ *
+ * Checks OA unit status registers and if necessary appends

corresponding

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct · line 881

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * status records for userspace (such as for a buffer full condition)

and then

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+ * initiate appending any buffered OA reports.
+ *
+ * Updates @offset according to the number of bytes successfully copied
Quoted 8 lines

into

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream, · line 889

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+ * the userspace buffer.
+ *
+ * NB: some data may be successfully copied to the userspace buffer
+ * even if an error is returned, and this is reflected in the
+ * updated @read_state.

updated @read_state ?

Ah, the api used to return the offset via a @read_state structure, this should be 'updated @offset'

Quoted 275 lines
Quoted 24 lines
drivers/gpu/drm/i915/i915_perf.c +23 -0
+ *
+ * Returns: zero on success or a negative error code
+ */
+static int gen8_oa_read(struct i915_perf_stream *stream,
+                     char __user *buf,
+                     size_t count,
+                     size_t *offset)
+{
+     struct drm_i915_private *dev_priv = stream->dev_priv;
+     u32 oastatus;
+     int ret;
+
+     if (WARN_ON(!dev_priv->perf.oa.oa_buffer.vaddr))
+             return -EIO;
+
+     oastatus = I915_READ(GEN8_OASTATUS);
+
+     /* We treat OABUFFER_OVERFLOW as a significant error:
+      *
+      * Although theoretically we could handle this more gracefully
+      * sometimes, some Gens don't correctly suppress certain
+      * automatically triggered reports in this condition and so we
+      * have to assume that old reports are now being trampled

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+      * over.
+      *
+      * Considering how we don't currently give userspace control
+      * over the OA buffer size and always configure a large 16MB
+      * buffer, then a buffer overflow does anyway likely indicate
+      * that something has gone quite badly wrong.
+      */
+     if (oastatus & GEN8_OASTATUS_OABUFFER_OVERFLOW) {
+             ret = append_oa_status(stream, buf, count, offset,
+                                    DRM_I915_PERF_RECORD_OA_

BUFFER_LOST);

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +4 -0
+             if (ret)
+                     return ret;
+
+             DRM_DEBUG("OA buffer overflow (exponent = %d): force

restart\n",

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +4 -0
+                       dev_priv->perf.oa.period_exponent);
+
+             dev_priv->perf.oa.ops.oa_disable(dev_priv);
+             dev_priv->perf.oa.ops.oa_enable(dev_priv);

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+
+             /* Note: .oa_enable() is expected to re-init the oabuffer
+              * and reset GEN8_OASTATUS for us
+              */
+             oastatus = I915_READ(GEN8_OASTATUS);
+     }
+
+     if (oastatus & GEN8_OASTATUS_REPORT_LOST) {
+             ret = append_oa_status(stream, buf, count, offset,
+                                    DRM_I915_PERF_RECORD_OA_

REPORT_LOST);

drivers/gpu/drm/i915/i915_perf.c static int gen7_oa_read(struct i915_perf_stream

Quoted 33 lines
drivers/gpu/drm/i915/i915_perf.c +29 -0
+             if (ret == 0) {
+                     I915_WRITE(GEN8_OASTATUS,
+                                oastatus & ~GEN8_OASTATUS_REPORT_LOST);
+             }
+     }
+
+     return gen8_append_oa_reports(stream, buf, count, offset);
+}
+
+/**
+ * Copies all buffered OA reports into userspace read() buffer.
+ * @stream: An i915-perf stream opened for OA metrics
+ * @buf: destination buffer given by userspace
+ * @count: the number of bytes userspace wants to read
+ * @offset: (inout): the current position for writing into @buf
+ *
+ * Notably any error condition resulting in a short read (-%ENOSPC or
+ * -%EFAULT) will be returned even though one or more records may
+ * have been successfully copied. In this case it's up to the caller
+ * to decide if the error should be squashed before returning to
+ * userspace.
+ *
+ * Note: reports are consumed from the head, and appended to the
+ * tail, so the tail chases the head?... If you think that's mad
+ * and back-to-front you're not alone, but this follows the
+ * Gen PRM naming convention.
+ *
+ * Returns: 0 on success, negative error code on failure.
+ */
 static int gen7_append_oa_reports(struct i915_perf_stream *stream,
                                char __user *buf,
                                size_t count,
@@ -733,7 +1147,8 @@ static int gen7_oa_read(struct i915_perf_stream

*stream,

drivers/gpu/drm/i915/i915_perf.c static int gen7_oa_read(struct i915_perf_stream *stream,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +1 -1
              if (ret)
                      return ret;
 
-             DRM_DEBUG("OA buffer overflow: force restart\n");
+             DRM_DEBUG("OA buffer overflow (exponent = %d): force

restart\n",

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_read(struct i915_perf_stream

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
+                       dev_priv->perf.oa.period_exponent);
 
              dev_priv->perf.oa.ops.oa_disable(dev_priv);
              dev_priv->perf.oa.ops.oa_enable(dev_priv);
@@ -833,33 +1248,37 @@ static int i915_oa_read(struct i915_perf_stream

*stream,

Quoted 36 lines
drivers/gpu/drm/i915/i915_perf.c +11 -19
 static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
 {
      struct drm_i915_private *dev_priv = stream->dev_priv;
-     struct intel_engine_cs *engine = dev_priv->engine[RCS];
-     int ret;
 
-     ret = i915_mutex_lock_interruptible(&dev_priv->drm);
-     if (ret)
-             return ret;
+     if (i915.enable_execlists)
+             dev_priv->perf.oa.specific_ctx_id = stream->ctx->hw_id;
+     else {
+             struct intel_engine_cs *engine = dev_priv->engine[RCS];
+             int ret;
 
-     /* As the ID is the gtt offset of the context's vma we pin
-      * the vma to ensure the ID remains fixed.
-      *
-      * NB: implied RCS engine...
-      */
-     ret = engine->context_pin(engine, stream->ctx);
-     if (ret)
-             goto unlock;
+             ret = i915_mutex_lock_interruptible(&dev_priv->drm);
+             if (ret)
+                     return ret;
 
-     /* Explicitly track the ID (instead of calling i915_ggtt_offset()
-      * on the fly) considering the difference with gen8+ and
-      * execlists
-      */
-     dev_priv->perf.oa.specific_ctx_id =
-             i915_ggtt_offset(stream->ctx->engine[engine->id].state);
+             /* As the ID is the gtt offset of the context's vma we pin
+              * the vma to ensure the ID remains fixed.
+              */

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_read(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +7 -2
+             ret = engine->context_pin(engine, stream->ctx);
+             if (ret) {
+                     mutex_unlock(&dev_priv->drm.struct_mutex);
+                     return ret;
+             }
 
-unlock:
-     mutex_unlock(&dev_priv->drm.struct_mutex);
+             /* Explicitly track the ID (instead of calling
+              * i915_ggtt_offset() on the fly) considering the

difference

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_read(struct i915_perf_stream *stream,

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +4 -0
+              * with gen8+ and execlists
+              */
+             dev_priv->perf.oa.specific_ctx_id =
+                     i915_ggtt_offset(stream->ctx->

engine[engine->id].state);

drivers/gpu/drm/i915/i915_perf.c static void oa_put_render_ctx_id(struct

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +4 -1
-     return ret;
+             mutex_unlock(&dev_priv->drm.struct_mutex);
+     }
+
+     return 0;
 }
 
 /**
@@ -874,12 +1293,16 @@ static void oa_put_render_ctx_id(struct

i915_perf_stream *stream)

drivers/gpu/drm/i915/i915_perf.c static void gen7_init_oa_buffer(struct

Quoted 21 lines
drivers/gpu/drm/i915/i915_perf.c +8 -4
      struct drm_i915_private *dev_priv = stream->dev_priv;
      struct intel_engine_cs *engine = dev_priv->engine[RCS];
 
-     mutex_lock(&dev_priv->drm.struct_mutex);
+     if (i915.enable_execlists) {
+             dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID;
+     } else {
+             mutex_lock(&dev_priv->drm.struct_mutex);
 
-     dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID;
-     engine->context_unpin(engine, stream->ctx);
+             dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID;
+             engine->context_unpin(engine, stream->ctx);
 
-     mutex_unlock(&dev_priv->drm.struct_mutex);
+             mutex_unlock(&dev_priv->drm.struct_mutex);
+     }
 }
 
 static void
@@ -964,6 +1387,56 @@ static void gen7_init_oa_buffer(struct

drm_i915_private *dev_priv)

drivers/gpu/drm/i915/i915_perf.c static void gen7_init_oa_buffer(struct drm_i915_private *dev_priv)

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
      dev_priv->perf.oa.pollin = false;
 }
 
+static void gen8_init_oa_buffer(struct drm_i915_private *dev_priv)
+{
+     u32 gtt_offset = i915_ggtt_offset(dev_priv->

perf.oa.oa_buffer.vma);

Quoted 16 lines
drivers/gpu/drm/i915/i915_perf.c +16 -0
+     unsigned long flags;
+
+     spin_lock_irqsave(&dev_priv->perf.oa.oa_buffer.ptr_lock, flags);
+
+     I915_WRITE(GEN8_OASTATUS, 0);
+     I915_WRITE(GEN8_OAHEADPTR, gtt_offset);
+     dev_priv->perf.oa.oa_buffer.head = gtt_offset;
+
+     I915_WRITE(GEN8_OABUFFER_UDW, 0);
+
+     /* PRM says:
+      *
+      *  "This MMIO must be set before the OATAILPTR
+      *  register and after the OAHEADPTR register. This is
+      *  to enable proper functionality of the overflow
+      *  bit."

drivers/gpu/drm/i915/i915_perf.c static void gen7_init_oa_buffer(struct drm_i915_private *dev_priv)

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+      */
+     I915_WRITE(GEN8_OABUFFER, gtt_offset |
+                OABUFFER_SIZE_16M | OA_MEM_SELECT_GGTT);
+     I915_WRITE(GEN8_OATAILPTR, gtt_offset & GEN8_OATAILPTR_MASK);
+
+     /* Mark that we need updated tail pointers to read from... */
+     dev_priv->perf.oa.oa_buffer.tails[0].offset = INVALID_TAIL_PTR;
+     dev_priv->perf.oa.oa_buffer.tails[1].offset = INVALID_TAIL_PTR;
+
+     spin_unlock_irqrestore(&dev_priv->perf.oa.oa_buffer.ptr_lock,

flags);

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct

Quoted 25 lines
drivers/gpu/drm/i915/i915_perf.c +21 -0
+
+
+     /* NB: although the OA buffer will initially be allocated
+      * zeroed via shmfs (and so this memset is redundant when
+      * first allocating), we may re-init the OA buffer, either
+      * when re-enabling a stream or in error/reset paths.
+      *
+      * The reason we clear the buffer for each re-init is for the
+      * sanity check in gen8_append_oa_reports() that looks at the
+      * reason field to make sure it's non-zero which relies on
+      * the assumption that new reports are being written to zeroed
+      * memory...
+      */
+     memset(dev_priv->perf.oa.oa_buffer.vaddr, 0, OA_BUFFER_SIZE);
+
+     /* Maybe make ->pollin per-stream state if we support multiple
+      * concurrent streams in the future.
+      */
+     dev_priv->perf.oa.pollin = false;
+}
+
 static int alloc_oa_buffer(struct drm_i915_private *dev_priv)
 {
      struct drm_i915_gem_object *bo;
@@ -1108,6 +1581,200 @@ static void hsw_disable_metric_set(struct

drm_i915_private *dev_priv)

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
                                    ~GT_NOA_ENABLE));
 }
 
+/*

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+ * From Broadwell PRM, 3D-Media-GPGPU -> Register State Context
+ *
+ * MMIO reads or writes to any of the registers listed in the
+ * “Register State Context image” subsections through HOST/IA
+ * MMIO interface for debug purposes must follow the steps below:
+ *
+ * - SW should set the Force Wakeup bit to prevent GT from entering C6.
+ * - Write 0x2050[31:0] = 0x00010001 (disable sequence).
+ * - Disable IDLE messaging in CS (Write 0x2050[31:0] = 0x00010001).
+ * - BDW:  Poll/Wait for register bits of 0x22AC[6:0] turn to 0x30

value.

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct · line 1595

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+ * - SKL+: Poll/Wait for register bits of 0x22A4[6:0] turn to 0x30

value.

Quoted 12 lines
drivers/gpu/drm/i915/i915_perf.c +12 -0
+ * - Read/Write to desired MMIO registers.
+ * - Enable IDLE messaging in CS (Write 0x2050[31:0] = 0x00010000).
+ * - Force Wakeup bit should be reset to enable C6 entry.
+ *
+ * XXX: don't nest or overlap calls to this API, it has no ref
+ * counting to track how many entities require the RCS to be
+ * blocked from being idle.
+ */
+static int gen8_begin_ctx_mmio(struct drm_i915_private *dev_priv)
+{
+     i915_reg_t fsm_reg = dev_priv->info.gen > 8 ?
+             GEN9_RCS_FE_FSM2 : GEN6_RCS_PWR_FSM;
Quoted 9 lines

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv) · line 1614

Quoted 7 lines
drivers/gpu/drm/i915/i915_perf.c +7 -0
+     int ret = 0;
+
+     /* There's only no active context while idle in execlist mode
+      * (though we shouldn't be using this in any other case)
+      */
+     if (WARN_ON(!i915.enable_execlists))
+             return ret;

return -EINVAL ?

Ah yup.

Quoted 83 lines
Quoted 52 lines
drivers/gpu/drm/i915/i915_perf.c +51 -0
+
+     intel_uncore_forcewake_get(dev_priv, FORCEWAKE_RENDER);
+
+     I915_WRITE(GEN6_RC_SLEEP_PSMI_CONTROL,
+                _MASKED_BIT_ENABLE(GEN6_PSMI_SLEEP_MSG_DISABLE));
+
+     /* Note: we don't currently have a good handle on the maximum
+      * latency for this wake up so while we only need to hold rcs
+      * busy from process context we at least keep the waiting
+      * interruptible...
+      */
+     ret = wait_for((I915_READ(fsm_reg) & 0x3f) == 0x30, 50);
+     if (ret)
+         intel_uncore_forcewake_put(dev_priv, FORCEWAKE_RENDER);
+
+     return ret;
+}
+
+static void gen8_end_ctx_mmio(struct drm_i915_private *dev_priv)
+{
+     if (WARN_ON(!i915.enable_execlists))
+             return;
+
+     I915_WRITE(GEN6_RC_SLEEP_PSMI_CONTROL,
+                _MASKED_BIT_DISABLE(GEN6_PSMI_SLEEP_MSG_DISABLE));
+     intel_uncore_forcewake_put(dev_priv, FORCEWAKE_RENDER);
+}
+
+/* Manages updating the per-context aspects of the OA stream
+ * configuration across all contexts.
+ *
+ * The awkward consideration here is that OACTXCONTROL controls the
+ * exponent for periodic sampling which is primarily used for system
+ * wide profiling where we'd like a consistent sampling period even in
+ * the face of context switches.
+ *
+ * Our approach of updating the register state context (as opposed to
+ * say using a workaround batch buffer) ensures that the hardware
+ * won't automatically reload an out-of-date timer exponent even
+ * transiently before a WA BB could be parsed.
+ *
+ * This function needs to:
+ * - Ensure the currently running context's per-context OA state is
+ *   updated
+ * - Ensure that all existing contexts will have the correct per-context
+ *   OA state if they are scheduled for use.
+ * - Ensure any new contexts will be initialized with the correct
+ *   per-context OA state.
+ *
+ * Note: it's only the RCS/Render context that has any OA state.
+ */

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+static int configure_all_contexts(struct drm_i915_private *dev_priv)
+{
+     struct i915_gem_context *ctx;
+     int ret;
+
+     ret = i915_mutex_lock_interruptible(&dev_priv->drm);
+     if (ret)
+             return ret;
+
+     /* Since execlist submission may be happening asynchronously here

then

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct · line 1676

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+      * we first mark existing contexts dirty before we update the

current

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+      * context so if any switches happen in the middle we can expect
+      * that the act of scheduling will have itself ensured a consistent
+      * OA state update.
+      */
+     list_for_each_entry(ctx, &dev_priv->context_list, link) {
+             /* The actual update of the register state context will

happen

Quoted 11 lines
drivers/gpu/drm/i915/i915_perf.c +11 -0
+              * the next time this logical ring is submitted. (See
+              * i915_oa_update_reg_state() which hooks into
+              * execlists_update_context())
+              */
+             atomic_set(&ctx->engine[RCS].oa_state_dirty, 1);
+     }
+
+     mutex_unlock(&dev_priv->drm.struct_mutex);
+
+     /* Now update the current context.
+      *
Quoted 9 lines

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+      * Note: Using MMIO to update per-context registers requires
+      * some extra care...
+      */
+     ret = gen8_begin_ctx_mmio(dev_priv);
+     if (ret) {
+             DRM_ERROR("Failed to bring RCS out of idle to update

current ctx OA state");
Missing newline for DRM_ERROR.

fixed

Quoted 69 lines

drivers/gpu/drm/i915/i915_perf.c static int append_oa_sample(struct i915_perf_stream *stream,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +4 -0
+             return ret;
+     }
+
+     I915_WRITE(GEN8_OACTXCONTROL, ((dev_priv->perf.oa.period_exponent

<<

Quoted 19 lines
drivers/gpu/drm/i915/i915_perf.c +19 -0
+                                     GEN8_OA_TIMER_PERIOD_SHIFT) |
+                                   (dev_priv->perf.oa.periodic ?
+                                    GEN8_OA_TIMER_ENABLE : 0) |
+                                   GEN8_OA_COUNTER_RESUME));
+
+     config_oa_regs(dev_priv, dev_priv->perf.oa.flex_regs,
+                     dev_priv->perf.oa.flex_regs_len);
+
+     gen8_end_ctx_mmio(dev_priv);
+
+     return 0;
+}
+
+static int gen8_enable_metric_set(struct drm_i915_private *dev_priv)
+{
+     int ret = dev_priv->perf.oa.ops.select_metric_set(dev_priv);
+
+     if (ret)
+             return ret;

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+
+     /* We disable slice/unslice clock ratio change reports on SKL since
+      * they are too noisy. The HW generates a lot of redundant reports
+      * where the ratio hasn't really changed causing a lot of redundant
+      * work to processes and increasing the chances we'll hit buffer
+      * overruns.
+      *
+      * Although we don't currently use the 'disable overrun' OABUFFER
+      * feature it's worth noting that clock ratio reports have to be
+      * disabled before considering to use that feature since the HW

doesn't

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +3 -0
+      * correctly block these reports.
+      *
+      * Currently none of the high-level metrics we have depend on

knowing

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 8 lines
drivers/gpu/drm/i915/i915_perf.c +8 -0
+      * this ratio to normalize.
+      *
+      * Note: This register is not power context saved and restored, but
+      * that's OK considering that we disable RC6 while the OA unit is
+      * enabled.
+      *
+      * The _INCLUDE_CLK_RATIO bit allows the slice/unslice frequency to
+      * be read back from automatically triggered reports, as part of

the

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+      * RPT_ID field.
+      */
+     if (IS_SKYLAKE(dev_priv) || IS_BROXTON(dev_priv)) {
+             I915_WRITE(GEN8_OA_DEBUG,
+                        _MASKED_BIT_ENABLE(GEN9_OA_

DEBUG_DISABLE_CLK_RATIO_REPORTS |

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct · line 1749

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+                                           GEN9_OA_DEBUG_INCLUDE_CLK_

RATIO));

Quoted 12 lines
drivers/gpu/drm/i915/i915_perf.c +12 -0
+     }
+
+     I915_WRITE(GDT_CHICKEN_BITS, 0xA0);
+     config_oa_regs(dev_priv, dev_priv->perf.oa.mux_regs,
+                    dev_priv->perf.oa.mux_regs_len);
+     I915_WRITE(GDT_CHICKEN_BITS, 0x80);
+
+     /* It takes a fairly long time for a new MUX configuration to
+      * be be applied after these register writes. This delay
+      * duration is take from Haswell (derived empirically based on
+      * the render_basic config) but hopefully it covers the
+      * maximum configuration latency for Gen8+ too...
Quoted 9 lines

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv) · line 1768

Quoted 7 lines
drivers/gpu/drm/i915/i915_perf.c +7 -0
+      */
+     usleep_range(15000, 20000);
+
+     config_oa_regs(dev_priv, dev_priv->perf.oa.b_counter_regs,
+                    dev_priv->perf.oa.b_counter_regs_len);
+
+     configure_all_contexts(dev_priv);

Check the return value here, and bubble up.

Oops, right, fixed.

Quoted 72 lines

drivers/gpu/drm/i915/i915_perf.c static void hsw_disable_metric_set(struct drm_i915_private *dev_priv)

Quoted 11 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
+
+     return 0;
+}
+
+static void gen8_disable_metric_set(struct drm_i915_private *dev_priv)
+{
+     /* NOP */
+}
+
 static void gen7_update_oacontrol_locked(struct drm_i915_private

*dev_priv)

drivers/gpu/drm/i915/i915_perf.c static void gen7_oa_enable(struct

Quoted 3 lines
drivers/gpu/drm/i915/i915_perf.c +0 -0
 {
      lockdep_assert_held(&dev_priv->perf.hook_lock);
@@ -1152,6 +1819,29 @@ static void gen7_oa_enable(struct

drm_i915_private *dev_priv)

drivers/gpu/drm/i915/i915_perf.c static void gen7_oa_disable(struct

Quoted 30 lines
drivers/gpu/drm/i915/i915_perf.c +23 -0
      spin_unlock_irqrestore(&dev_priv->perf.hook_lock, flags);
 }
 
+static void gen8_oa_enable(struct drm_i915_private *dev_priv)
+{
+     u32 report_format = dev_priv->perf.oa.oa_buffer.format;
+
+     /* Reset buf pointers so we don't forward reports from before now.
+      *
+      * Think carefully if considering trying to avoid this, since it
+      * also ensures status flags and the buffer itself are cleared
+      * in error paths, and we have checks for invalid reports based
+      * on the assumption that certain fields are written to zeroed
+      * memory which this helps maintains.
+      */
+     gen8_init_oa_buffer(dev_priv);
+
+     /* Note: we don't rely on the hardware to perform single context
+      * filtering and instead filter on the cpu based on the context-id
+      * field of reports
+      */
+     I915_WRITE(GEN8_OACONTROL, (report_format <<
+                                 GEN8_OA_REPORT_FORMAT_SHIFT) |
+                                GEN8_OA_COUNTER_ENABLE);
+}
+
 /**
  * i915_oa_stream_enable - handle `I915_PERF_IOCTL_ENABLE` for OA stream
  * @stream: An i915 perf stream opened for OA metrics
@@ -1178,6 +1868,11 @@ static void gen7_oa_disable(struct

drm_i915_private *dev_priv)

drivers/gpu/drm/i915/i915_perf.c static void gen7_oa_disable(struct drm_i915_private *dev_priv)

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
      I915_WRITE(GEN7_OACONTROL, 0);
 }
 
+static void gen8_oa_disable(struct drm_i915_private *dev_priv)
+{
+     I915_WRITE(GEN8_OACONTROL, 0);
+}
+
 /**
  * i915_oa_stream_disable - handle `I915_PERF_IOCTL_DISABLE` for OA

stream

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +0 -0
  * @stream: An i915 perf stream opened for OA metrics
@@ -1336,6 +2031,88 @@ static int i915_oa_stream_init(struct

i915_perf_stream *stream,

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream,

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
      return ret;
 }
 
+/* NB: It must always remain pointer safe to run this even if the OA

unit

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream,

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+ * has been disabled.
+ *
+ * It's fine to put out-of-date values into these per-context registers
+ * in the case that the OA unit has been disabled.
+ */
+static void gen8_update_reg_state_unlocked(struct intel_engine_cs
Quoted 5 lines

*engine,

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct · line 2042

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +2 -0
+                                        struct i915_gem_context *ctx,
+                                        uint32_t *reg_state)

u32 for consistency, elsewhere also.

updated.

Quoted 9 lines

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream, · line 2048

Quoted 7 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+{
+     struct drm_i915_private *dev_priv = ctx->i915;
+     const struct i915_oa_reg *flex_regs = dev_priv->perf.oa.flex_regs;
+     int n_flex_regs = dev_priv->perf.oa.flex_regs_len;
+     int ctx_oactxctrl = dev_priv->perf.oa.ctx_oactxctrl_off;
+     int ctx_flexeu0 = dev_priv->perf.oa.ctx_flexeu0_off;

Can we also make these u32, for consistency.

yup

Quoted 231 lines

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream,

Quoted 13 lines
drivers/gpu/drm/i915/i915_perf.c +13 -0
+     /* The MMIO offsets for Flex EU registers aren't contiguous */
+     u32 flex_mmio[] = {
+             i915_mmio_reg_offset(EU_PERF_CNTL0),
+             i915_mmio_reg_offset(EU_PERF_CNTL1),
+             i915_mmio_reg_offset(EU_PERF_CNTL2),
+             i915_mmio_reg_offset(EU_PERF_CNTL3),
+             i915_mmio_reg_offset(EU_PERF_CNTL4),
+             i915_mmio_reg_offset(EU_PERF_CNTL5),
+             i915_mmio_reg_offset(EU_PERF_CNTL6),
+     };
+     int i;
+
+     reg_state[ctx_oactxctrl] = i915_mmio_reg_offset(GEN8_

OACTXCONTROL);

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream,

Quoted 13 lines
drivers/gpu/drm/i915/i915_perf.c +13 -0
+     reg_state[ctx_oactxctrl+1] = (dev_priv->perf.oa.period_exponent <<
+                                   GEN8_OA_TIMER_PERIOD_SHIFT) |
+                                  (dev_priv->perf.oa.periodic ?
+                                   GEN8_OA_TIMER_ENABLE : 0) |
+                                  GEN8_OA_COUNTER_RESUME;
+
+     for (i = 0; i < ARRAY_SIZE(flex_mmio); i++) {
+             uint32_t state_offset = ctx_flexeu0 + i * 2;
+             uint32_t mmio = flex_mmio[i];
+
+             /* This arbitrary default will select the 'EU FPU0 Pipeline
+              * Active' event. In the future it's anticipated that there
+              * will be an explicit 'No Event' we can select, but not

yet...

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream,

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+              */
+             uint32_t value = 0;
+             int j;
+
+             for (j = 0; j < n_flex_regs; j++) {
+                     if (i915_mmio_reg_offset(flex_regs[j].addr) ==

mmio) {

Quoted 27 lines
drivers/gpu/drm/i915/i915_perf.c +27 -0
+                             value = flex_regs[j].value;
+                             break;
+                     }
+             }
+
+             reg_state[state_offset] = mmio;
+             reg_state[state_offset+1] = value;
+     }
+}
+
+void i915_oa_update_reg_state(struct intel_engine_cs *engine,
+                           struct i915_gem_context *ctx,
+                           uint32_t *reg_state)
+{
+     struct drm_i915_private *dev_priv = engine->i915;
+
+     if (engine->id != RCS)
+             return;
+
+     if (!dev_priv->perf.initialized)
+             return;
+
+     /* XXX: We don't take a lock here and this may run async with
+      * respect to stream methods. Notably we don't want to block
+      * context switches by long i915 perf read() operations.
+      *
+      * It's expected to always be safe to read the dev_priv->perf

drivers/gpu/drm/i915/i915_perf.c static int i915_oa_stream_init(struct i915_perf_stream *stream,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +8 -0
+      * state needed here, and expected to be benign to redundantly
+      * update the state if the OA unit has been disabled since
+      * oa_state_dirty was last set.
+      */
+     if (atomic_cmpxchg(&ctx->engine[engine->id].oa_state_dirty, 1, 0))
+             gen8_update_reg_state_unlocked(engine, ctx, reg_state);
+}
+
 /**
  * i915_perf_read_locked - &i915_perf_stream_ops->read with error

normalisation

drivers/gpu/drm/i915/i915_perf.c i915_perf_open_ioctl_locked(struct

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +0 -0
  * @stream: An i915 perf stream
@@ -1750,6 +2527,7 @@ i915_perf_open_ioctl_locked(struct

drm_i915_private *dev_priv,

drivers/gpu/drm/i915/i915_perf.c i915_perf_open_ioctl_locked(struct

Quoted 8 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
      struct i915_gem_context *specific_ctx = NULL;
      struct i915_perf_stream *stream = NULL;
      unsigned long f_flags = 0;
+     bool privileged_op = true;
      int stream_fd;
      int ret;
 
@@ -1767,12 +2545,28 @@ i915_perf_open_ioctl_locked(struct

drm_i915_private *dev_priv,

drivers/gpu/drm/i915/i915_perf.c i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv,

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +2 -0
              }
      }
 
+     /* On Haswell the OA unit supports clock gating off for a specific
+      * context and in this mode there's no visibility of metrics for

the

drivers/gpu/drm/i915/i915_perf.c i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv,

Quoted 9 lines
drivers/gpu/drm/i915/i915_perf.c +9 -0
+      * rest of the system, which we consider acceptable for a
+      * non-privileged client.
+      *
+      * For Gen8+ the OA unit no longer supports clock gating off for a
+      * specific context and the kernel can't securely stop the counters
+      * from updating as system-wide / global values. Even though we can
+      * filter reports based on the included context ID we can't block
+      * clients from seeing the raw / global counter values via
+      * MI_REPORT_PERF_COUNT commands and so consider it a privileged

op to

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +4 -0
+      * enable the OA unit by default.
+      */
+     if (IS_HASWELL(dev_priv) && specific_ctx)
+             privileged_op = false;

drivers/gpu/drm/i915/i915_perf.c i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv,

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +2 -1
+
      /* Similar to perf's kernel.perf_paranoid_cpu sysctl option
       * we check a dev.i915.perf_stream_paranoid sysctl option
       * to determine if it's ok to access system wide OA counters
       * without CAP_SYS_ADMIN privileges.
       */
-     if (!specific_ctx &&
+     if (privileged_op &&
          i915_perf_stream_paranoid && !capable(CAP_SYS_ADMIN)) {
              DRM_DEBUG("Insufficient privileges to open system-wide

i915 perf stream\n");

drivers/gpu/drm/i915/i915_perf.c int i915_perf_open_ioctl(struct drm_device *dev,

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +0 -0
              ret = -EACCES;
@@ -2039,9 +2833,6 @@ int i915_perf_open_ioctl(struct drm_device *dev,

void *data,

drivers/gpu/drm/i915/i915_perf.c void i915_perf_register(struct drm_i915_private

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +0 -3
  */
 void i915_perf_register(struct drm_i915_private *dev_priv)
 {
-     if (!IS_HASWELL(dev_priv))
-             return;
-
      if (!dev_priv->perf.initialized)
              return;
 
@@ -2057,11 +2848,38 @@ void i915_perf_register(struct drm_i915_private

*dev_priv)

drivers/gpu/drm/i915/i915_perf.c void i915_perf_register(struct drm_i915_private

Quoted 42 lines
drivers/gpu/drm/i915/i915_perf.c +30 -3
      if (!dev_priv->perf.metrics_kobj)
              goto exit;
 
-     if (i915_perf_register_sysfs_hsw(dev_priv)) {
-             kobject_put(dev_priv->perf.metrics_kobj);
-             dev_priv->perf.metrics_kobj = NULL;
+     if (IS_HASWELL(dev_priv)) {
+             if (i915_perf_register_sysfs_hsw(dev_priv))
+                     goto sysfs_error;
+     } else if (IS_BROADWELL(dev_priv)) {
+             if (i915_perf_register_sysfs_bdw(dev_priv))
+                     goto sysfs_error;
+     } else if (IS_CHERRYVIEW(dev_priv)) {
+             if (i915_perf_register_sysfs_chv(dev_priv))
+                     goto sysfs_error;
+     } else if (IS_SKYLAKE(dev_priv)) {
+             if (IS_SKL_GT2(dev_priv)) {
+                     if (i915_perf_register_sysfs_sklgt2(dev_priv))
+                             goto sysfs_error;
+             } else if (IS_SKL_GT3(dev_priv)) {
+                     if (i915_perf_register_sysfs_sklgt3(dev_priv))
+                             goto sysfs_error;
+             } else if (IS_SKL_GT4(dev_priv)) {
+                     if (i915_perf_register_sysfs_sklgt4(dev_priv))
+                             goto sysfs_error;
+             } else
+                     goto sysfs_error;
+     } else if (IS_BROXTON(dev_priv)) {
+             if (i915_perf_register_sysfs_bxt(dev_priv))
+                     goto sysfs_error;
      }
 
+     goto exit;
+
+sysfs_error:
+     kobject_put(dev_priv->perf.metrics_kobj);
+     dev_priv->perf.metrics_kobj = NULL;
+
 exit:
      mutex_unlock(&dev_priv->perf.lock);
 }
@@ -2077,13 +2895,24 @@ void i915_perf_register(struct drm_i915_private

*dev_priv)

Quoted 37 lines
drivers/gpu/drm/i915/i915_perf.c +15 -9
  */
 void i915_perf_unregister(struct drm_i915_private *dev_priv)
 {
-     if (!IS_HASWELL(dev_priv))
-             return;
-
      if (!dev_priv->perf.metrics_kobj)
              return;
 
-     i915_perf_unregister_sysfs_hsw(dev_priv);
+        if (IS_HASWELL(dev_priv))
+                i915_perf_unregister_sysfs_hsw(dev_priv);
+        else if (IS_BROADWELL(dev_priv))
+                i915_perf_unregister_sysfs_bdw(dev_priv);
+        else if (IS_CHERRYVIEW(dev_priv))
+                i915_perf_unregister_sysfs_chv(dev_priv);
+        else if (IS_SKYLAKE(dev_priv)) {
+             if (IS_SKL_GT2(dev_priv))
+                     i915_perf_unregister_sysfs_sklgt2(dev_priv);
+             else if (IS_SKL_GT3(dev_priv))
+                     i915_perf_unregister_sysfs_sklgt3(dev_priv);
+             else if (IS_SKL_GT4(dev_priv))
+                     i915_perf_unregister_sysfs_sklgt4(dev_priv);
+     } else if (IS_BROXTON(dev_priv))
+                i915_perf_unregister_sysfs_bxt(dev_priv);
 
      kobject_put(dev_priv->perf.metrics_kobj);
      dev_priv->perf.metrics_kobj = NULL;
@@ -2142,45 +2971,107 @@ static struct ctl_table dev_root[] = {
  */
 void i915_perf_init(struct drm_i915_private *dev_priv)
 {
-     if (!IS_HASWELL(dev_priv))
-             return;
-
-     /* Using the same limiting factors as printk_ratelimit() */
-     ratelimit_state_init(&dev_priv->perf.oa.spurious_report_rs,

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 2159

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +0 -10
-                     5 * HZ, 10);
-     /* We use a DRM_NOTE for spurious reports so it would be
-      * inconsistent to print a warning for throttling.
-      */
-     ratelimit_set_flags(&dev_priv->perf.oa.spurious_report_rs,
-                     RATELIMIT_MSG_ON_RELEASE);
-
-     hrtimer_init(&dev_priv->perf.oa.poll_check_timer,
-                  CLOCK_MONOTONIC, HRTIMER_MODE_REL);
-     dev_priv->perf.oa.poll_check_timer.function =

oa_poll_check_timer_cb;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +4 -1
-     init_waitqueue_head(&dev_priv->perf.oa.poll_wq);
+     dev_priv->perf.oa.n_builtin_sets = 0;
+
+     if (IS_HASWELL(dev_priv)) {
+             dev_priv->perf.oa.ops.init_oa_buffer =

gen7_init_oa_buffer;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 3037

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+             dev_priv->perf.oa.ops.enable_metric_set =

hsw_enable_metric_set;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 3039

Quoted 1 line
drivers/gpu/drm/i915/i915_perf.c +1 -0
+             dev_priv->perf.oa.ops.disable_metric_set =

hsw_disable_metric_set;

Quoted 14 lines

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 2990

Quoted 11 lines
drivers/gpu/drm/i915/i915_perf.c +11 -0
+             dev_priv->perf.oa.ops.oa_enable = gen7_oa_enable;
+             dev_priv->perf.oa.ops.oa_disable = gen7_oa_disable;
+             dev_priv->perf.oa.ops.read = gen7_oa_read;
+             dev_priv->perf.oa.ops.oa_buffer_check =
+                     gen7_oa_buffer_check_unlocked;
+
+             dev_priv->perf.oa.oa_formats = hsw_oa_formats;
+
+             dev_priv->perf.oa.n_builtin_sets =
+                     i915_oa_n_builtin_metric_sets_hsw;
+     } else if (i915.enable_execlists) {

Maybe a comment for why we don't support legacy mode with gen8+, for the next reader.

It pretty much just came down to simplicity I think. Earlier iterations of the driver did try and support both, but since BDW enables execlist mode by default now it avoided a bunch of complexity to ignore the case with BDW and legacy ringbuffer mode here. I've added a comment.

Quoted 96 lines
Quoted 14 lines
drivers/gpu/drm/i915/i915_perf.c +14 -0
+             if (IS_GEN8(dev_priv)) {
+                     dev_priv->perf.oa.ctx_oactxctrl_off = 0x120;
+                     dev_priv->perf.oa.ctx_flexeu0_off = 0x2ce;
+                     dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<25);
+
+                     if (IS_BROADWELL(dev_priv)) {
+                             dev_priv->perf.oa.n_builtin_sets =
+                                     i915_oa_n_builtin_metric_sets_bdw;
+                             dev_priv->perf.oa.ops.select_metric_set =
+                                     i915_oa_select_metric_set_bdw;
+                     } else if (IS_CHERRYVIEW(dev_priv)) {
+                             dev_priv->perf.oa.n_builtin_sets =
+                                     i915_oa_n_builtin_metric_sets_chv;
+                             dev_priv->perf.oa.ops.select_metric_set =

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +10 -0
+                                     i915_oa_select_metric_set_chv;
+                     }
+             } else if (IS_GEN9(dev_priv)) {
+                     dev_priv->perf.oa.ctx_oactxctrl_off = 0x128;
+                     dev_priv->perf.oa.ctx_flexeu0_off = 0x3de;
+                     dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16);
+
+                     if (IS_SKL_GT2(dev_priv)) {
+                             dev_priv->perf.oa.n_builtin_sets =
+                                     i915_oa_n_builtin_metric_sets_

sklgt2;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+                             dev_priv->perf.oa.ops.select_metric_set =
+                                     i915_oa_select_metric_set_sklgt2;
+                     } else if (IS_SKL_GT3(dev_priv)) {
+                             dev_priv->perf.oa.n_builtin_sets =
+                                     i915_oa_n_builtin_metric_sets_

sklgt3;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+                             dev_priv->perf.oa.ops.select_metric_set =
+                                     i915_oa_select_metric_set_sklgt3;
+                     } else if (IS_SKL_GT4(dev_priv)) {
+                             dev_priv->perf.oa.n_builtin_sets =
+                                     i915_oa_n_builtin_metric_sets_

sklgt4;

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+                             dev_priv->perf.oa.ops.select_metric_set =
+                                     i915_oa_select_metric_set_sklgt4;
+                     } else if (IS_BROXTON(dev_priv)) {
+                             dev_priv->perf.oa.n_builtin_sets =
+                                     i915_oa_n_builtin_metric_sets_bxt;
+                             dev_priv->perf.oa.ops.select_metric_set =

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +5 -4
+                                     i915_oa_select_metric_set_bxt;
+                     }
+             }
 
-     INIT_LIST_HEAD(&dev_priv->perf.streams);
-     mutex_init(&dev_priv->perf.lock);
-     spin_lock_init(&dev_priv->perf.hook_lock);
-     spin_lock_init(&dev_priv->perf.oa.oa_buffer.ptr_lock);
+             if (dev_priv->perf.oa.n_builtin_sets) {
+                     dev_priv->perf.oa.ops.init_oa_buffer =

gen8_init_oa_buffer;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 6 lines
drivers/gpu/drm/i915/i915_perf.c +6 -0
+                     dev_priv->perf.oa.ops.enable_metric_set =
+                             gen8_enable_metric_set;
+                     dev_priv->perf.oa.ops.disable_metric_set =
+                             gen8_disable_metric_set;
+                     dev_priv->perf.oa.ops.oa_enable = gen8_oa_enable;
+                     dev_priv->perf.oa.ops.oa_disable =

gen8_oa_disable;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +5 -0
+                     dev_priv->perf.oa.ops.read = gen8_oa_read;
+                     dev_priv->perf.oa.ops.oa_buffer_check =
+                             gen8_oa_buffer_check_unlocked;
+
+                     dev_priv->perf.oa.oa_formats =

gen8_plus_oa_formats;

Quoted 4 lines
drivers/gpu/drm/i915/i915_perf.c +2 -1
+             }
+     }
 
-     dev_priv->perf.oa.ops.init_oa_buffer = gen7_init_oa_buffer;

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 10 lines
drivers/gpu/drm/i915/i915_perf.c +3 -7
-     dev_priv->perf.oa.ops.enable_metric_set = hsw_enable_metric_set;
-     dev_priv->perf.oa.ops.disable_metric_set = hsw_disable_metric_set;
-     dev_priv->perf.oa.ops.oa_enable = gen7_oa_enable;
-     dev_priv->perf.oa.ops.oa_disable = gen7_oa_disable;
-     dev_priv->perf.oa.ops.read = gen7_oa_read;
-     dev_priv->perf.oa.ops.oa_buffer_check =
-             gen7_oa_buffer_check_unlocked;
+     if (dev_priv->perf.oa.n_builtin_sets) {
+             /* Using the same limiting factors as printk_ratelimit() */
+             ratelimit_state_init(&dev_priv->perf.oa.spurious_report_

rs,

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 11 lines
drivers/gpu/drm/i915/i915_perf.c +9 -1
+                             5 * HZ, 10);
+             /* We use a DRM_NOTE for spurious reports so it would be
+              * inconsistent to print a warning for throttling.
+              */
+             ratelimit_set_flags(&dev_priv->perf.oa.spurious_report_rs,
+                             RATELIMIT_MSG_ON_RELEASE);
 
-     dev_priv->perf.oa.oa_formats = hsw_oa_formats;
+             hrtimer_init(&dev_priv->perf.oa.poll_check_timer,
+                             CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+             dev_priv->perf.oa.poll_check_timer.function =

oa_poll_check_timer_cb;

Quoted 2 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
+             init_waitqueue_head(&dev_priv->perf.oa.poll_wq);
 
Quoted 8 lines

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = { · line 3068

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +3 -2
-     dev_priv->perf.oa.n_builtin_sets =
-             i915_oa_n_builtin_metric_sets_hsw;
+             INIT_LIST_HEAD(&dev_priv->perf.streams);
+             mutex_init(&dev_priv->perf.lock);
+             spin_lock_init(&dev_priv->perf.hook_lock);

Not strictly related to this patch, but why do we still need perf.hook_lock?

Right gen7_update_oacontrol_locked can be folded into gen7_oa_enable and the only usage left is actually redundant since gen7 no longer relies on a hook for notifications of context pinning.

I think in some earlier review for HSW I had been a little reluctant to push to remove this lock early since I was using it in the gen8+ patches at the time, but that's not the case anymore. The i915_oa_update_reg_state() hook is designed to be safe to run without any i915-perf specific locking.

I'll aim to address this in a follow up patch.

Thanks for the review, I'll send out an update in a bit.

Quoted 147 lines

drivers/gpu/drm/i915/i915_perf.c static struct ctl_table dev_root[] = {

Quoted 5 lines
drivers/gpu/drm/i915/i915_perf.c +2 -1
+             spin_lock_init(&dev_priv->perf.oa.oa_buffer.ptr_lock);
 
-     dev_priv->perf.sysctl_header = register_sysctl_table(dev_root);
+             dev_priv->perf.sysctl_header = register_sysctl_table(dev_

root);

drivers/gpu/drm/i915/i915_perf.c void i915_perf_fini(struct drm_i915_private

Quoted 8 lines
drivers/gpu/drm/i915/i915_perf.c +2 -1
-     dev_priv->perf.initialized = true;
+             dev_priv->perf.initialized = true;
+     }
 }
 
 /**
@@ -2200,5 +3091,6 @@ void i915_perf_fini(struct drm_i915_private

*dev_priv)

Review

Quoted 7 lines
drivers/gpu/drm/i915/i915_perf.c +1 -0
      unregister_sysctl_table(dev_priv->perf.sysctl_header);
 
      memset(&dev_priv->perf.oa.ops, 0, sizeof(dev_priv->perf.oa.ops));
+
      dev_priv->perf.initialized = false;
 }
(quoted diff) +0 -0

b/drivers/gpu/drm/i915/i915_reg.h

drivers/gpu/drm/i915/i915_reg.h static inline bool i915_mmio_reg_valid(i915_reg_t

Quoted 4 lines

index 04c8f69fcc62..0052289ed8ad 100644

drivers/gpu/drm/i915/i915_reg.h +0 -0
@@ -645,6 +645,12 @@ static inline bool i915_mmio_reg_valid(i915_reg_t

reg)

drivers/gpu/drm/i915/i915_reg.h static inline bool i915_mmio_reg_valid(i915_reg_t

Quoted 13 lines
drivers/gpu/drm/i915/i915_reg.h +6 -0
 #define GEN8_OACTXID _MMIO(0x2364)
 
+#define GEN8_OA_DEBUG _MMIO(0x2B04)
+#define  GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS    (1<<5)
+#define  GEN9_OA_DEBUG_INCLUDE_CLK_RATIO         (1<<6)
+#define  GEN9_OA_DEBUG_DISABLE_GO_1_0_REPORTS            (1<<2)
+#define  GEN9_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS   (1<<1)
+
 #define GEN8_OACONTROL _MMIO(0x2B00)
 #define  GEN8_OA_REPORT_FORMAT_A12       (0<<2)
 #define  GEN8_OA_REPORT_FORMAT_A12_B8_C8    (2<<2)
@@ -666,6 +672,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t

reg)

drivers/gpu/drm/i915/i915_reg.h static inline bool i915_mmio_reg_valid(i915_reg_t

Quoted 8 lines
drivers/gpu/drm/i915/i915_reg.h +1 -0
 #define  GEN7_OABUFFER_STOP_RESUME_ENABLE   (1<<1)
 #define  GEN7_OABUFFER_RESUME                    (1<<0)
 
+#define GEN8_OABUFFER_UDW _MMIO(0x23b4)
 #define GEN8_OABUFFER _MMIO(0x2b14)
 
 #define GEN7_OASTATUS1 _MMIO(0x2364)
@@ -684,7 +691,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t

reg)

drivers/gpu/drm/i915/i915_reg.h static inline bool i915_mmio_reg_valid(i915_reg_t

Quoted 10 lines
drivers/gpu/drm/i915/i915_reg.h +2 -0
 #define  GEN8_OASTATUS_REPORT_LOST       (1<<0)
 
 #define GEN8_OAHEADPTR _MMIO(0x2B0C)
+#define GEN8_OAHEADPTR_MASK    0xffffffc0
 #define GEN8_OATAILPTR _MMIO(0x2B10)
+#define GEN8_OATAILPTR_MASK    0xffffffc0
 
 #define OABUFFER_SIZE_128K  (0<<3)
 #define OABUFFER_SIZE_256K  (1<<3)
@@ -697,7 +706,17 @@ static inline bool i915_mmio_reg_valid(i915_reg_t

reg)

Review

Quoted 28 lines
drivers/gpu/drm/i915/i915_reg.h +13 -0
 #define OA_MEM_SELECT_GGTT  (1<<0)
 
+/*
+ * Flexible, Aggregate EU Counter Registers.
+ * Note: these aren't contiguous
+ */
 #define EU_PERF_CNTL0            _MMIO(0xe458)
+#define EU_PERF_CNTL1            _MMIO(0xe558)
+#define EU_PERF_CNTL2            _MMIO(0xe658)
+#define EU_PERF_CNTL3            _MMIO(0xe758)
+#define EU_PERF_CNTL4            _MMIO(0xe45c)
+#define EU_PERF_CNTL5            _MMIO(0xe55c)
+#define EU_PERF_CNTL6            _MMIO(0xe65c)
 
 #define GDT_CHICKEN_BITS    _MMIO(0x9840)
 #define GT_NOA_ENABLE            0x00000080
@@ -2317,6 +2336,9 @@ enum skl_disp_power_wells {
 #define   GEN8_RC_SEMA_IDLE_MSG_DISABLE      (1 << 12)
 #define   GEN8_FF_DOP_CLOCK_GATE_DISABLE     (1<<10)
 
+#define GEN6_RCS_PWR_FSM _MMIO(0x22ac)
+#define GEN9_RCS_FE_FSM2 _MMIO(0x22a4)
+
 /* Fuse readout registers for GT */
 #define CHV_FUSE_GT                  _MMIO(VLV_DISPLAY_BASE + 0x2168)
 #define   CHV_FGT_DISABLE_SS0                (1 << 10)
(quoted diff) +0 -0

b/drivers/gpu/drm/i915/intel_lrc.c

drivers/gpu/drm/i915/intel_lrc.c static u64 execlists_update_context(struct

Quoted 4 lines

index eec1e714f531..e6d9e4197d3d 100644

drivers/gpu/drm/i915/intel_lrc.c +0 -0
@@ -337,6 +337,8 @@ static u64 execlists_update_context(struct

drm_i915_gem_request *rq)

Quoted 51 lines
drivers/gpu/drm/i915/intel_lrc.c +5 -0
      if (ppgtt && !i915_vm_is_48bit(&ppgtt->base))
              execlists_update_context_pdps(ppgtt, reg_state);
 
+     i915_oa_update_reg_state(rq->engine, rq->ctx, reg_state);
+
      return ce->lrc_desc;
 }
 
@@ -1878,6 +1880,9 @@ static void execlists_init_reg_state(u32 *regs,
              regs[CTX_LRI_HEADER_2] = MI_LOAD_REGISTER_IMM(1);
              CTX_REG(regs, CTX_R_PWR_CLK_STATE, GEN8_R_PWR_CLK_STATE,
                      make_rpcs(dev_priv));
+
+             atomic_set(&ctx->engine[RCS].oa_state_dirty, 1);
+             i915_oa_update_reg_state(engine, ctx, regs);
      }
 }
 
include/uapi/drm/i915_drm.h +12 -8
index e0599e729e68..03b833849919 100644
@@ -1295,13 +1295,18 @@ struct drm_i915_gem_context_param {
 };
 
 enum drm_i915_oa_format {
-     I915_OA_FORMAT_A13 = 1,
-     I915_OA_FORMAT_A29,
-     I915_OA_FORMAT_A13_B8_C8,
-     I915_OA_FORMAT_B4_C8,
-     I915_OA_FORMAT_A45_B8_C8,
-     I915_OA_FORMAT_B4_C8_A16,
-     I915_OA_FORMAT_C4_B8,
+     I915_OA_FORMAT_A13 = 1,     /* HSW only */
+     I915_OA_FORMAT_A29,         /* HSW only */
+     I915_OA_FORMAT_A13_B8_C8,   /* HSW only */
+     I915_OA_FORMAT_B4_C8,       /* HSW only */
+     I915_OA_FORMAT_A45_B8_C8,   /* HSW only */
+     I915_OA_FORMAT_B4_C8_A16,   /* HSW only */
+     I915_OA_FORMAT_C4_B8,       /* HSW+ */
+
+     /* Gen8+ */
+     I915_OA_FORMAT_A12,
+     I915_OA_FORMAT_A12_B8_C8,
+     I915_OA_FORMAT_A32u40_A4u32_B8_C8,
 
      I915_OA_FORMAT_MAX          /* non-ABI */
 };
--

2.12.0

Quoted 5 lines
Quoted 4 lines

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#13 lore raw

PATCH v2 5/5 drm/i915: Add more OA configs for BDW, CHV, SKL + BXT

In reply to Robert Bragg

These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml
---
 drivers/gpu/drm/i915/i915_oa_bdw.c    | 4856 ++++++++++++++++++++++++++++++++-
 drivers/gpu/drm/i915/i915_oa_bxt.c    | 2323 +++++++++++++++-
 drivers/gpu/drm/i915/i915_oa_chv.c    | 2525 ++++++++++++++++-
 drivers/gpu/drm/i915/i915_oa_hsw.c    |   58 +-
 drivers/gpu/drm/i915/i915_oa_sklgt2.c | 3157 ++++++++++++++++++++-
 drivers/gpu/drm/i915/i915_oa_sklgt3.c | 2702 +++++++++++++++++-
 drivers/gpu/drm/i915/i915_oa_sklgt4.c | 2745 ++++++++++++++++++-
 7 files changed, 18177 insertions(+), 189 deletions(-)
drivers/gpu/drm/i915/i915_oa_bdw.c +4815 -41
index 539484aef6b5..bdf630e9e99c 100644
@@ -31,9 +31,30 @@
 
 enum metric_set_id {
 	METRIC_SET_ID_RENDER_BASIC = 1,
+	METRIC_SET_ID_COMPUTE_BASIC,
+	METRIC_SET_ID_RENDER_PIPE_PROFILE,
+	METRIC_SET_ID_MEMORY_READS,
+	METRIC_SET_ID_MEMORY_WRITES,
+	METRIC_SET_ID_COMPUTE_EXTENDED,
+	METRIC_SET_ID_COMPUTE_L3_CACHE,
+	METRIC_SET_ID_DATA_PORT_READS_COALESCING,
+	METRIC_SET_ID_DATA_PORT_WRITES_COALESCING,
+	METRIC_SET_ID_HDC_AND_SF,
+	METRIC_SET_ID_L3_1,
+	METRIC_SET_ID_L3_2,
+	METRIC_SET_ID_L3_3,
+	METRIC_SET_ID_L3_4,
+	METRIC_SET_ID_RASTERIZER_AND_PIXEL_BACKEND,
+	METRIC_SET_ID_SAMPLER_1,
+	METRIC_SET_ID_SAMPLER_2,
+	METRIC_SET_ID_TDL_1,
+	METRIC_SET_ID_TDL_2,
+	METRIC_SET_ID_COMPUTE_EXTRA,
+	METRIC_SET_ID_VME_PIPE,
+	METRIC_SET_ID_TEST_OA,
 };
 
-int i915_oa_n_builtin_metric_sets_bdw = 1;
+int i915_oa_n_builtin_metric_sets_bdw = 22;
 
 static const struct i915_oa_reg b_counter_config_render_basic[] = {
 	{ _MMIO(0x2710), 0x00000000 },
@@ -290,66 +311,4609 @@ get_render_basic_mux_config(struct drm_i915_private *dev_priv,
 	return NULL;
 }
 
+static const struct i915_oa_reg b_counter_config_compute_basic[] = {
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0x00800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x00800000 },
+	{ _MMIO(0x2740), 0x00000000 },
+};
+
+static const struct i915_oa_reg flex_eu_config_compute_basic[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00000003 },
+	{ _MMIO(0xe658), 0x00002001 },
+	{ _MMIO(0xe758), 0x00778008 },
+	{ _MMIO(0xe45c), 0x00088078 },
+	{ _MMIO(0xe55c), 0x00808708 },
+	{ _MMIO(0xe65c), 0x00a08908 },
+};
+
+static const struct i915_oa_reg mux_config_compute_basic_0_slices_0x01[] = {
+	{ _MMIO(0x9888), 0x105c00e0 },
+	{ _MMIO(0x9888), 0x105800e0 },
+	{ _MMIO(0x9888), 0x103800e0 },
+	{ _MMIO(0x9888), 0x3580001a },
+	{ _MMIO(0x9888), 0x3b800060 },
+	{ _MMIO(0x9888), 0x3d800005 },
+	{ _MMIO(0x9888), 0x065c2100 },
+	{ _MMIO(0x9888), 0x0a5c0041 },
+	{ _MMIO(0x9888), 0x0c5c6600 },
+	{ _MMIO(0x9888), 0x005c6580 },
+	{ _MMIO(0x9888), 0x085c8000 },
+	{ _MMIO(0x9888), 0x0e5c8000 },
+	{ _MMIO(0x9888), 0x00580042 },
+	{ _MMIO(0x9888), 0x08582080 },
+	{ _MMIO(0x9888), 0x0c58004c },
+	{ _MMIO(0x9888), 0x0e582580 },
+	{ _MMIO(0x9888), 0x005b4000 },
+	{ _MMIO(0x9888), 0x185b1000 },
+	{ _MMIO(0x9888), 0x1a5b0104 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1faa00 },
+	{ _MMIO(0x9888), 0x101f02aa },
+	{ _MMIO(0x9888), 0x08380042 },
+	{ _MMIO(0x9888), 0x0a382080 },
+	{ _MMIO(0x9888), 0x0e38404c },
+	{ _MMIO(0x9888), 0x0238404b },
+	{ _MMIO(0x9888), 0x00384000 },
+	{ _MMIO(0x9888), 0x16380000 },
+	{ _MMIO(0x9888), 0x18381145 },
+	{ _MMIO(0x9888), 0x04380000 },
+	{ _MMIO(0x9888), 0x0039a000 },
+	{ _MMIO(0x9888), 0x06398000 },
+	{ _MMIO(0x9888), 0x0839a000 },
+	{ _MMIO(0x9888), 0x0a39a000 },
+	{ _MMIO(0x9888), 0x0c39a000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x02392000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8aaaa0 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x238b02a0 },
+	{ _MMIO(0x9888), 0x258b5550 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x1f850a80 },
+	{ _MMIO(0x9888), 0x2185aaa0 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x03844000 },
+	{ _MMIO(0x9888), 0x17808137 },
+	{ _MMIO(0x9888), 0x1980c147 },
+	{ _MMIO(0x9888), 0x1b80c0e5 },
+	{ _MMIO(0x9888), 0x1d80c0e3 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x13804000 },
+	{ _MMIO(0x9888), 0x15800000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d801000 },
+	{ _MMIO(0x9888), 0x4f800111 },
+	{ _MMIO(0x9888), 0x43800062 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800062 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800062 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f801062 },
+	{ _MMIO(0x9888), 0x41801084 },
+};
+
+static const struct i915_oa_reg mux_config_compute_basic_2_slices_0x02[] = {
+	{ _MMIO(0x9888), 0x10dc00e0 },
+	{ _MMIO(0x9888), 0x10d800e0 },
+	{ _MMIO(0x9888), 0x10b800e0 },
+	{ _MMIO(0x9888), 0x3580001a },
+	{ _MMIO(0x9888), 0x3b800060 },
+	{ _MMIO(0x9888), 0x3d800005 },
+	{ _MMIO(0x9888), 0x06dc2100 },
+	{ _MMIO(0x9888), 0x0adc0041 },
+	{ _MMIO(0x9888), 0x0cdc6600 },
+	{ _MMIO(0x9888), 0x00dc6580 },
+	{ _MMIO(0x9888), 0x08dc8000 },
+	{ _MMIO(0x9888), 0x0edc8000 },
+	{ _MMIO(0x9888), 0x00d80042 },
+	{ _MMIO(0x9888), 0x08d82080 },
+	{ _MMIO(0x9888), 0x0cd8004c },
+	{ _MMIO(0x9888), 0x0ed82580 },
+	{ _MMIO(0x9888), 0x00db4000 },
+	{ _MMIO(0x9888), 0x18db1000 },
+	{ _MMIO(0x9888), 0x1adb0104 },
+	{ _MMIO(0x9888), 0x0c9fa800 },
+	{ _MMIO(0x9888), 0x0e9faa00 },
+	{ _MMIO(0x9888), 0x109f02aa },
+	{ _MMIO(0x9888), 0x08b80042 },
+	{ _MMIO(0x9888), 0x0ab82080 },
+	{ _MMIO(0x9888), 0x0eb8404c },
+	{ _MMIO(0x9888), 0x02b8404b },
+	{ _MMIO(0x9888), 0x00b84000 },
+	{ _MMIO(0x9888), 0x16b80000 },
+	{ _MMIO(0x9888), 0x18b81145 },
+	{ _MMIO(0x9888), 0x04b80000 },
+	{ _MMIO(0x9888), 0x00b9a000 },
+	{ _MMIO(0x9888), 0x06b98000 },
+	{ _MMIO(0x9888), 0x08b9a000 },
+	{ _MMIO(0x9888), 0x0ab9a000 },
+	{ _MMIO(0x9888), 0x0cb9a000 },
+	{ _MMIO(0x9888), 0x0eb9a000 },
+	{ _MMIO(0x9888), 0x02b92000 },
+	{ _MMIO(0x9888), 0x01888000 },
+	{ _MMIO(0x9888), 0x0d88f800 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x238b0540 },
+	{ _MMIO(0x9888), 0x258baaa0 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x018c4000 },
+	{ _MMIO(0x9888), 0x0f8c4000 },
+	{ _MMIO(0x9888), 0x178c2000 },
+	{ _MMIO(0x9888), 0x198c5500 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x018da000 },
+	{ _MMIO(0x9888), 0x078d8000 },
+	{ _MMIO(0x9888), 0x098da000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x038d2000 },
+	{ _MMIO(0x9888), 0x1f850a80 },
+	{ _MMIO(0x9888), 0x2185aaa0 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x03844000 },
+	{ _MMIO(0x9888), 0x17808137 },
+	{ _MMIO(0x9888), 0x1980c147 },
+	{ _MMIO(0x9888), 0x1b80c0e5 },
+	{ _MMIO(0x9888), 0x1d80c0e3 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x13804000 },
+	{ _MMIO(0x9888), 0x15800000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d805000 },
+	{ _MMIO(0x9888), 0x4f800555 },
+	{ _MMIO(0x9888), 0x43800062 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800062 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800062 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800062 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg *
+get_compute_basic_mux_config(struct drm_i915_private *dev_priv,
+			     int *len)
+{
+	if (INTEL_INFO(dev_priv)->sseu.slice_mask & 0x01) {
+		*len = ARRAY_SIZE(mux_config_compute_basic_0_slices_0x01);
+		return mux_config_compute_basic_0_slices_0x01;
+	} else if (INTEL_INFO(dev_priv)->sseu.slice_mask & 0x02) {
+		*len = ARRAY_SIZE(mux_config_compute_basic_2_slices_0x02);
+		return mux_config_compute_basic_2_slices_0x02;
+	}
+
+	*len = 0;
+	return NULL;
+}
+
+static const struct i915_oa_reg b_counter_config_render_pipe_profile[] = {
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007ffea },
+	{ _MMIO(0x2774), 0x00007ffc },
+	{ _MMIO(0x2778), 0x0007affa },
+	{ _MMIO(0x277c), 0x0000f5fd },
+	{ _MMIO(0x2780), 0x00079ffa },
+	{ _MMIO(0x2784), 0x0000f3fb },
+	{ _MMIO(0x2788), 0x0007bf7a },
+	{ _MMIO(0x278c), 0x0000f7e7 },
+	{ _MMIO(0x2790), 0x0007fefa },
+	{ _MMIO(0x2794), 0x0000f7cf },
+	{ _MMIO(0x2798), 0x00077ffa },
+	{ _MMIO(0x279c), 0x0000efdf },
+	{ _MMIO(0x27a0), 0x0006fffa },
+	{ _MMIO(0x27a4), 0x0000cfbf },
+	{ _MMIO(0x27a8), 0x0003fffa },
+	{ _MMIO(0x27ac), 0x00005f7f },
+};
+
+static const struct i915_oa_reg flex_eu_config_render_pipe_profile[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00015014 },
+	{ _MMIO(0xe658), 0x00025024 },
+	{ _MMIO(0xe758), 0x00035034 },
+	{ _MMIO(0xe45c), 0x00045044 },
+	{ _MMIO(0xe55c), 0x00055054 },
+	{ _MMIO(0xe65c), 0x00065064 },
+};
+
+static const struct i915_oa_reg mux_config_render_pipe_profile[] = {
+	{ _MMIO(0x9888), 0x0a1e0000 },
+	{ _MMIO(0x9888), 0x0c1f000f },
+	{ _MMIO(0x9888), 0x10176800 },
+	{ _MMIO(0x9888), 0x1191001f },
+	{ _MMIO(0x9888), 0x0b880320 },
+	{ _MMIO(0x9888), 0x01890c40 },
+	{ _MMIO(0x9888), 0x118a1c00 },
+	{ _MMIO(0x9888), 0x118d7c00 },
+	{ _MMIO(0x9888), 0x118e0020 },
+	{ _MMIO(0x9888), 0x118f4c00 },
+	{ _MMIO(0x9888), 0x11900000 },
+	{ _MMIO(0x9888), 0x13900001 },
+	{ _MMIO(0x9888), 0x065c4000 },
+	{ _MMIO(0x9888), 0x0c3d8000 },
+	{ _MMIO(0x9888), 0x06584000 },
+	{ _MMIO(0x9888), 0x0c5b4000 },
+	{ _MMIO(0x9888), 0x081e0040 },
+	{ _MMIO(0x9888), 0x0e1e0000 },
+	{ _MMIO(0x9888), 0x021f5400 },
+	{ _MMIO(0x9888), 0x001f0000 },
+	{ _MMIO(0x9888), 0x101f0010 },
+	{ _MMIO(0x9888), 0x0e1f0080 },
+	{ _MMIO(0x9888), 0x0c384000 },
+	{ _MMIO(0x9888), 0x06392000 },
+	{ _MMIO(0x9888), 0x0c13c000 },
+	{ _MMIO(0x9888), 0x06164000 },
+	{ _MMIO(0x9888), 0x06170012 },
+	{ _MMIO(0x9888), 0x00170000 },
+	{ _MMIO(0x9888), 0x01910005 },
+	{ _MMIO(0x9888), 0x07880002 },
+	{ _MMIO(0x9888), 0x01880c00 },
+	{ _MMIO(0x9888), 0x0f880000 },
+	{ _MMIO(0x9888), 0x0d880000 },
+	{ _MMIO(0x9888), 0x05880000 },
+	{ _MMIO(0x9888), 0x09890032 },
+	{ _MMIO(0x9888), 0x078a0800 },
+	{ _MMIO(0x9888), 0x0f8a0a00 },
+	{ _MMIO(0x9888), 0x198a4000 },
+	{ _MMIO(0x9888), 0x1b8a2000 },
+	{ _MMIO(0x9888), 0x1d8a0000 },
+	{ _MMIO(0x9888), 0x038a4000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x238b54c0 },
+	{ _MMIO(0x9888), 0x258baa55 },
+	{ _MMIO(0x9888), 0x278b0019 },
+	{ _MMIO(0x9888), 0x198c0100 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x0f8d0015 },
+	{ _MMIO(0x9888), 0x018d1000 },
+	{ _MMIO(0x9888), 0x098d8000 },
+	{ _MMIO(0x9888), 0x0b8df000 },
+	{ _MMIO(0x9888), 0x0d8d3000 },
+	{ _MMIO(0x9888), 0x038de000 },
+	{ _MMIO(0x9888), 0x058d3000 },
+	{ _MMIO(0x9888), 0x0d8e0004 },
+	{ _MMIO(0x9888), 0x058e000c },
+	{ _MMIO(0x9888), 0x098e0000 },
+	{ _MMIO(0x9888), 0x078e0000 },
+	{ _MMIO(0x9888), 0x038e0000 },
+	{ _MMIO(0x9888), 0x0b8f0020 },
+	{ _MMIO(0x9888), 0x198f0c00 },
+	{ _MMIO(0x9888), 0x078f8000 },
+	{ _MMIO(0x9888), 0x098f4000 },
+	{ _MMIO(0x9888), 0x0b900980 },
+	{ _MMIO(0x9888), 0x03900d80 },
+	{ _MMIO(0x9888), 0x01900000 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaaa },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x0784c000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x1780c000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d801111 },
+	{ _MMIO(0x9888), 0x3d800800 },
+	{ _MMIO(0x9888), 0x4f801011 },
+	{ _MMIO(0x9888), 0x43800443 },
+	{ _MMIO(0x9888), 0x51801111 },
+	{ _MMIO(0x9888), 0x45800422 },
+	{ _MMIO(0x9888), 0x53801111 },
+	{ _MMIO(0x9888), 0x47800c60 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800422 },
+	{ _MMIO(0x9888), 0x41800021 },
+};
+
+static const struct i915_oa_reg *
+get_render_pipe_profile_mux_config(struct drm_i915_private *dev_priv,
+				   int *len)
+{
+	*len = ARRAY_SIZE(mux_config_render_pipe_profile);
+	return mux_config_render_pipe_profile;
+}
+
+static const struct i915_oa_reg b_counter_config_memory_reads[] = {
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x274c), 0x86543210 },
+	{ _MMIO(0x2748), 0x86543210 },
+	{ _MMIO(0x2744), 0x00006667 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x275c), 0x86543210 },
+	{ _MMIO(0x2758), 0x86543210 },
+	{ _MMIO(0x2754), 0x00006465 },
+	{ _MMIO(0x2750), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007f81a },
+	{ _MMIO(0x2774), 0x0000fe00 },
+	{ _MMIO(0x2778), 0x0007f82a },
+	{ _MMIO(0x277c), 0x0000fe00 },
+	{ _MMIO(0x2780), 0x0007f872 },
+	{ _MMIO(0x2784), 0x0000fe00 },
+	{ _MMIO(0x2788), 0x0007f8ba },
+	{ _MMIO(0x278c), 0x0000fe00 },
+	{ _MMIO(0x2790), 0x0007f87a },
+	{ _MMIO(0x2794), 0x0000fe00 },
+	{ _MMIO(0x2798), 0x0007f8ea },
+	{ _MMIO(0x279c), 0x0000fe00 },
+	{ _MMIO(0x27a0), 0x0007f8e2 },
+	{ _MMIO(0x27a4), 0x0000fe00 },
+	{ _MMIO(0x27a8), 0x0007f8f2 },
+	{ _MMIO(0x27ac), 0x0000fe00 },
+};
+
+static const struct i915_oa_reg flex_eu_config_memory_reads[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00015014 },
+	{ _MMIO(0xe658), 0x00025024 },
+	{ _MMIO(0xe758), 0x00035034 },
+	{ _MMIO(0xe45c), 0x00045044 },
+	{ _MMIO(0xe55c), 0x00055054 },
+	{ _MMIO(0xe65c), 0x00065064 },
+};
+
+static const struct i915_oa_reg mux_config_memory_reads[] = {
+	{ _MMIO(0x9888), 0x198b0343 },
+	{ _MMIO(0x9888), 0x13845800 },
+	{ _MMIO(0x9888), 0x15840018 },
+	{ _MMIO(0x9888), 0x3580001a },
+	{ _MMIO(0x9888), 0x038b6300 },
+	{ _MMIO(0x9888), 0x058b6b62 },
+	{ _MMIO(0x9888), 0x078b006a },
+	{ _MMIO(0x9888), 0x118b0000 },
+	{ _MMIO(0x9888), 0x238b0000 },
+	{ _MMIO(0x9888), 0x258b0000 },
+	{ _MMIO(0x9888), 0x1f85a080 },
+	{ _MMIO(0x9888), 0x2185aaaa },
+	{ _MMIO(0x9888), 0x2385000a },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x01840018 },
+	{ _MMIO(0x9888), 0x07844c80 },
+	{ _MMIO(0x9888), 0x09840d9a },
+	{ _MMIO(0x9888), 0x0b840e9c },
+	{ _MMIO(0x9888), 0x0d840f9e },
+	{ _MMIO(0x9888), 0x0f840010 },
+	{ _MMIO(0x9888), 0x11840000 },
+	{ _MMIO(0x9888), 0x03848000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x2f8000e5 },
+	{ _MMIO(0x9888), 0x138080e3 },
+	{ _MMIO(0x9888), 0x1580c0e1 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x11804000 },
+	{ _MMIO(0x9888), 0x1780c000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f804000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800800 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800842 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800842 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47801042 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800084 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg *
+get_memory_reads_mux_config(struct drm_i915_private *dev_priv,
+			    int *len)
+{
+	*len = ARRAY_SIZE(mux_config_memory_reads);
+	return mux_config_memory_reads;
+}
+
+static const struct i915_oa_reg b_counter_config_memory_writes[] = {
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x274c), 0x86543210 },
+	{ _MMIO(0x2748), 0x86543210 },
+	{ _MMIO(0x2744), 0x00006667 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x275c), 0x86543210 },
+	{ _MMIO(0x2758), 0x86543210 },
+	{ _MMIO(0x2754), 0x00006465 },
+	{ _MMIO(0x2750), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007f81a },
+	{ _MMIO(0x2774), 0x0000fe00 },
+	{ _MMIO(0x2778), 0x0007f82a },
+	{ _MMIO(0x277c), 0x0000fe00 },
+	{ _MMIO(0x2780), 0x0007f822 },
+	{ _MMIO(0x2784), 0x0000fe00 },
+	{ _MMIO(0x2788), 0x0007f8ba },
+	{ _MMIO(0x278c), 0x0000fe00 },
+	{ _MMIO(0x2790), 0x0007f87a },
+	{ _MMIO(0x2794), 0x0000fe00 },
+	{ _MMIO(0x2798), 0x0007f8ea },
+	{ _MMIO(0x279c), 0x0000fe00 },
+	{ _MMIO(0x27a0), 0x0007f8e2 },
+	{ _MMIO(0x27a4), 0x0000fe00 },
+	{ _MMIO(0x27a8), 0x0007f8f2 },
+	{ _MMIO(0x27ac), 0x0000fe00 },
+};
+
+static const struct i915_oa_reg flex_eu_config_memory_writes[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00015014 },
+	{ _MMIO(0xe658), 0x00025024 },
+	{ _MMIO(0xe758), 0x00035034 },
+	{ _MMIO(0xe45c), 0x00045044 },
+	{ _MMIO(0xe55c), 0x00055054 },
+	{ _MMIO(0xe65c), 0x00065064 },
+};
+
+static const struct i915_oa_reg mux_config_memory_writes[] = {
+	{ _MMIO(0x9888), 0x198b0343 },
+	{ _MMIO(0x9888), 0x13845400 },
+	{ _MMIO(0x9888), 0x3580001a },
+	{ _MMIO(0x9888), 0x3d800805 },
+	{ _MMIO(0x9888), 0x038b6300 },
+	{ _MMIO(0x9888), 0x058b6b62 },
+	{ _MMIO(0x9888), 0x078b006a },
+	{ _MMIO(0x9888), 0x118b0000 },
+	{ _MMIO(0x9888), 0x238b0000 },
+	{ _MMIO(0x9888), 0x258b0000 },
+	{ _MMIO(0x9888), 0x1f85a080 },
+	{ _MMIO(0x9888), 0x2185aaaa },
+	{ _MMIO(0x9888), 0x23850002 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x01840010 },
+	{ _MMIO(0x9888), 0x07844880 },
+	{ _MMIO(0x9888), 0x09840992 },
+	{ _MMIO(0x9888), 0x0b840a94 },
+	{ _MMIO(0x9888), 0x0d840b96 },
+	{ _MMIO(0x9888), 0x11840000 },
+	{ _MMIO(0x9888), 0x03848000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x2d800147 },
+	{ _MMIO(0x9888), 0x2f8000e5 },
+	{ _MMIO(0x9888), 0x138080e3 },
+	{ _MMIO(0x9888), 0x1580c0e1 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x11804000 },
+	{ _MMIO(0x9888), 0x1780c000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f800000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800842 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800842 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47801082 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800084 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg *
+get_memory_writes_mux_config(struct drm_i915_private *dev_priv,
+			     int *len)
+{
+	*len = ARRAY_SIZE(mux_config_memory_writes);
+	return mux_config_memory_writes;
+}
+
+static const struct i915_oa_reg b_counter_config_compute_extended[] = {
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007fc2a },
+	{ _MMIO(0x2774), 0x0000bf00 },
+	{ _MMIO(0x2778), 0x0007fc6a },
+	{ _MMIO(0x277c), 0x0000bf00 },
+	{ _MMIO(0x2780), 0x0007fc92 },
+	{ _MMIO(0x2784), 0x0000bf00 },
+	{ _MMIO(0x2788), 0x0007fca2 },
+	{ _MMIO(0x278c), 0x0000bf00 },
+	{ _MMIO(0x2790), 0x0007fc32 },
+	{ _MMIO(0x2794), 0x0000bf00 },
+	{ _MMIO(0x2798), 0x0007fc9a },
+	{ _MMIO(0x279c), 0x0000bf00 },
+	{ _MMIO(0x27a0), 0x0007fe6a },
+	{ _MMIO(0x27a4), 0x0000bf00 },
+	{ _MMIO(0x27a8), 0x0007fe7a },
+	{ _MMIO(0x27ac), 0x0000bf00 },
+};
+
+static const struct i915_oa_reg flex_eu_config_compute_extended[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00000003 },
+	{ _MMIO(0xe658), 0x00002001 },
+	{ _MMIO(0xe758), 0x00778008 },
+	{ _MMIO(0xe45c), 0x00088078 },
+	{ _MMIO(0xe55c), 0x00808708 },
+	{ _MMIO(0xe65c), 0x00a08908 },
+};
+
+static const struct i915_oa_reg mux_config_compute_extended_0_subslices_0x01[] = {
+	{ _MMIO(0x9888), 0x143d0160 },
+	{ _MMIO(0x9888), 0x163d2800 },
+	{ _MMIO(0x9888), 0x183d0120 },
+	{ _MMIO(0x9888), 0x105800e0 },
+	{ _MMIO(0x9888), 0x005cc000 },
+	{ _MMIO(0x9888), 0x065c8000 },
+	{ _MMIO(0x9888), 0x085cc000 },
+	{ _MMIO(0x9888), 0x0a5cc000 },
+	{ _MMIO(0x9888), 0x0c5cc000 },
+	{ _MMIO(0x9888), 0x0e5cc000 },
+	{ _MMIO(0x9888), 0x025cc000 },
+	{ _MMIO(0x9888), 0x045cc000 },
+	{ _MMIO(0x9888), 0x003d0011 },
+	{ _MMIO(0x9888), 0x063d0900 },
+	{ _MMIO(0x9888), 0x083d0a13 },
+	{ _MMIO(0x9888), 0x0a3d0b15 },
+	{ _MMIO(0x9888), 0x0c3d2317 },
+	{ _MMIO(0x9888), 0x043d21b7 },
+	{ _MMIO(0x9888), 0x103d0000 },
+	{ _MMIO(0x9888), 0x0e3d0000 },
+	{ _MMIO(0x9888), 0x1a3d0000 },
+	{ _MMIO(0x9888), 0x0e5825c1 },
+	{ _MMIO(0x9888), 0x00586100 },
+	{ _MMIO(0x9888), 0x0258204c },
+	{ _MMIO(0x9888), 0x06588000 },
+	{ _MMIO(0x9888), 0x0858c000 },
+	{ _MMIO(0x9888), 0x0a58c000 },
+	{ _MMIO(0x9888), 0x0c58c000 },
+	{ _MMIO(0x9888), 0x0458c000 },
+	{ _MMIO(0x9888), 0x005b4000 },
+	{ _MMIO(0x9888), 0x0e5b4000 },
+	{ _MMIO(0x9888), 0x185b5400 },
+	{ _MMIO(0x9888), 0x1a5b0155 },
+	{ _MMIO(0x9888), 0x025b4000 },
+	{ _MMIO(0x9888), 0x045b4000 },
+	{ _MMIO(0x9888), 0x065b4000 },
+	{ _MMIO(0x9888), 0x085b4000 },
+	{ _MMIO(0x9888), 0x0a5b4000 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1faa2a },
+	{ _MMIO(0x9888), 0x101f02aa },
+	{ _MMIO(0x9888), 0x00384000 },
+	{ _MMIO(0x9888), 0x0e384000 },
+	{ _MMIO(0x9888), 0x16384000 },
+	{ _MMIO(0x9888), 0x18381555 },
+	{ _MMIO(0x9888), 0x02384000 },
+	{ _MMIO(0x9888), 0x04384000 },
+	{ _MMIO(0x9888), 0x06384000 },
+	{ _MMIO(0x9888), 0x08384000 },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x0039a000 },
+	{ _MMIO(0x9888), 0x06398000 },
+	{ _MMIO(0x9888), 0x0839a000 },
+	{ _MMIO(0x9888), 0x0a39a000 },
+	{ _MMIO(0x9888), 0x0c39a000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x0239a000 },
+	{ _MMIO(0x9888), 0x0439a000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8aaaa0 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x238b2aa0 },
+	{ _MMIO(0x9888), 0x258b5551 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa2 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg mux_config_compute_extended_2_subslices_0x02[] = {
+	{ _MMIO(0x9888), 0x105c00e0 },
+	{ _MMIO(0x9888), 0x145b0160 },
+	{ _MMIO(0x9888), 0x165b2800 },
+	{ _MMIO(0x9888), 0x185b0120 },
+	{ _MMIO(0x9888), 0x0e5c25c1 },
+	{ _MMIO(0x9888), 0x005c6100 },
+	{ _MMIO(0x9888), 0x025c204c },
+	{ _MMIO(0x9888), 0x065c8000 },
+	{ _MMIO(0x9888), 0x085cc000 },
+	{ _MMIO(0x9888), 0x0a5cc000 },
+	{ _MMIO(0x9888), 0x0c5cc000 },
+	{ _MMIO(0x9888), 0x045cc000 },
+	{ _MMIO(0x9888), 0x005b0011 },
+	{ _MMIO(0x9888), 0x065b0900 },
+	{ _MMIO(0x9888), 0x085b0a13 },
+	{ _MMIO(0x9888), 0x0a5b0b15 },
+	{ _MMIO(0x9888), 0x0c5b2317 },
+	{ _MMIO(0x9888), 0x045b21b7 },
+	{ _MMIO(0x9888), 0x105b0000 },
+	{ _MMIO(0x9888), 0x0e5b0000 },
+	{ _MMIO(0x9888), 0x1a5b0000 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1faa2a },
+	{ _MMIO(0x9888), 0x101f02aa },
+	{ _MMIO(0x9888), 0x00384000 },
+	{ _MMIO(0x9888), 0x0e384000 },
+	{ _MMIO(0x9888), 0x16384000 },
+	{ _MMIO(0x9888), 0x18381555 },
+	{ _MMIO(0x9888), 0x02384000 },
+	{ _MMIO(0x9888), 0x04384000 },
+	{ _MMIO(0x9888), 0x06384000 },
+	{ _MMIO(0x9888), 0x08384000 },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x0039a000 },
+	{ _MMIO(0x9888), 0x06398000 },
+	{ _MMIO(0x9888), 0x0839a000 },
+	{ _MMIO(0x9888), 0x0a39a000 },
+	{ _MMIO(0x9888), 0x0c39a000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x0239a000 },
+	{ _MMIO(0x9888), 0x0439a000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8aaaa0 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x238b2aa0 },
+	{ _MMIO(0x9888), 0x258b5551 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa2 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg mux_config_compute_extended_4_subslices_0x04[] = {
+	{ _MMIO(0x9888), 0x103800e0 },
+	{ _MMIO(0x9888), 0x143a0160 },
+	{ _MMIO(0x9888), 0x163a2800 },
+	{ _MMIO(0x9888), 0x183a0120 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1faa2a },
+	{ _MMIO(0x9888), 0x101f02aa },
+	{ _MMIO(0x9888), 0x0e38a5c1 },
+	{ _MMIO(0x9888), 0x0038a100 },
+	{ _MMIO(0x9888), 0x0238204c },
+	{ _MMIO(0x9888), 0x16388000 },
+	{ _MMIO(0x9888), 0x183802aa },
+	{ _MMIO(0x9888), 0x04380000 },
+	{ _MMIO(0x9888), 0x06380000 },
+	{ _MMIO(0x9888), 0x08388000 },
+	{ _MMIO(0x9888), 0x0a388000 },
+	{ _MMIO(0x9888), 0x0039a000 },
+	{ _MMIO(0x9888), 0x06398000 },
+	{ _MMIO(0x9888), 0x0839a000 },
+	{ _MMIO(0x9888), 0x0a39a000 },
+	{ _MMIO(0x9888), 0x0c39a000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x0239a000 },
+	{ _MMIO(0x9888), 0x0439a000 },
+	{ _MMIO(0x9888), 0x003a0011 },
+	{ _MMIO(0x9888), 0x063a0900 },
+	{ _MMIO(0x9888), 0x083a0a13 },
+	{ _MMIO(0x9888), 0x0a3a0b15 },
+	{ _MMIO(0x9888), 0x0c3a2317 },
+	{ _MMIO(0x9888), 0x043a21b7 },
+	{ _MMIO(0x9888), 0x103a0000 },
+	{ _MMIO(0x9888), 0x0e3a0000 },
+	{ _MMIO(0x9888), 0x1a3a0000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8aaaa0 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x238b2aa0 },
+	{ _MMIO(0x9888), 0x258b5551 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa2 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg mux_config_compute_extended_1_subslices_0x08[] = {
+	{ _MMIO(0x9888), 0x14bd0160 },
+	{ _MMIO(0x9888), 0x16bd2800 },
+	{ _MMIO(0x9888), 0x18bd0120 },
+	{ _MMIO(0x9888), 0x10d800e0 },
+	{ _MMIO(0x9888), 0x00dcc000 },
+	{ _MMIO(0x9888), 0x06dc8000 },
+	{ _MMIO(0x9888), 0x08dcc000 },
+	{ _MMIO(0x9888), 0x0adcc000 },
+	{ _MMIO(0x9888), 0x0cdcc000 },
+	{ _MMIO(0x9888), 0x0edcc000 },
+	{ _MMIO(0x9888), 0x02dcc000 },
+	{ _MMIO(0x9888), 0x04dcc000 },
+	{ _MMIO(0x9888), 0x00bd0011 },
+	{ _MMIO(0x9888), 0x06bd0900 },
+	{ _MMIO(0x9888), 0x08bd0a13 },
+	{ _MMIO(0x9888), 0x0abd0b15 },
+	{ _MMIO(0x9888), 0x0cbd2317 },
+	{ _MMIO(0x9888), 0x04bd21b7 },
+	{ _MMIO(0x9888), 0x10bd0000 },
+	{ _MMIO(0x9888), 0x0ebd0000 },
+	{ _MMIO(0x9888), 0x1abd0000 },
+	{ _MMIO(0x9888), 0x0ed825c1 },
+	{ _MMIO(0x9888), 0x00d86100 },
+	{ _MMIO(0x9888), 0x02d8204c },
+	{ _MMIO(0x9888), 0x06d88000 },
+	{ _MMIO(0x9888), 0x08d8c000 },
+	{ _MMIO(0x9888), 0x0ad8c000 },
+	{ _MMIO(0x9888), 0x0cd8c000 },
+	{ _MMIO(0x9888), 0x04d8c000 },
+	{ _MMIO(0x9888), 0x00db4000 },
+	{ _MMIO(0x9888), 0x0edb4000 },
+	{ _MMIO(0x9888), 0x18db5400 },
+	{ _MMIO(0x9888), 0x1adb0155 },
+	{ _MMIO(0x9888), 0x02db4000 },
+	{ _MMIO(0x9888), 0x04db4000 },
+	{ _MMIO(0x9888), 0x06db4000 },
+	{ _MMIO(0x9888), 0x08db4000 },
+	{ _MMIO(0x9888), 0x0adb4000 },
+	{ _MMIO(0x9888), 0x0c9fa800 },
+	{ _MMIO(0x9888), 0x0e9faa2a },
+	{ _MMIO(0x9888), 0x109f02aa },
+	{ _MMIO(0x9888), 0x00b84000 },
+	{ _MMIO(0x9888), 0x0eb84000 },
+	{ _MMIO(0x9888), 0x16b84000 },
+	{ _MMIO(0x9888), 0x18b81555 },
+	{ _MMIO(0x9888), 0x02b84000 },
+	{ _MMIO(0x9888), 0x04b84000 },
+	{ _MMIO(0x9888), 0x06b84000 },
+	{ _MMIO(0x9888), 0x08b84000 },
+	{ _MMIO(0x9888), 0x0ab84000 },
+	{ _MMIO(0x9888), 0x00b9a000 },
+	{ _MMIO(0x9888), 0x06b98000 },
+	{ _MMIO(0x9888), 0x08b9a000 },
+	{ _MMIO(0x9888), 0x0ab9a000 },
+	{ _MMIO(0x9888), 0x0cb9a000 },
+	{ _MMIO(0x9888), 0x0eb9a000 },
+	{ _MMIO(0x9888), 0x02b9a000 },
+	{ _MMIO(0x9888), 0x04b9a000 },
+	{ _MMIO(0x9888), 0x01888000 },
+	{ _MMIO(0x9888), 0x0d88f800 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x07888000 },
+	{ _MMIO(0x9888), 0x09888000 },
+	{ _MMIO(0x9888), 0x0b888000 },
+	{ _MMIO(0x9888), 0x238b5540 },
+	{ _MMIO(0x9888), 0x258baaa2 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x018c4000 },
+	{ _MMIO(0x9888), 0x0f8c4000 },
+	{ _MMIO(0x9888), 0x178c2000 },
+	{ _MMIO(0x9888), 0x198c5500 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x078c4000 },
+	{ _MMIO(0x9888), 0x098c4000 },
+	{ _MMIO(0x9888), 0x0b8c4000 },
+	{ _MMIO(0x9888), 0x018da000 },
+	{ _MMIO(0x9888), 0x078d8000 },
+	{ _MMIO(0x9888), 0x098da000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x038da000 },
+	{ _MMIO(0x9888), 0x058da000 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa2 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg mux_config_compute_extended_3_subslices_0x10[] = {
+	{ _MMIO(0x9888), 0x10dc00e0 },
+	{ _MMIO(0x9888), 0x14db0160 },
+	{ _MMIO(0x9888), 0x16db2800 },
+	{ _MMIO(0x9888), 0x18db0120 },
+	{ _MMIO(0x9888), 0x0edc25c1 },
+	{ _MMIO(0x9888), 0x00dc6100 },
+	{ _MMIO(0x9888), 0x02dc204c },
+	{ _MMIO(0x9888), 0x06dc8000 },
+	{ _MMIO(0x9888), 0x08dcc000 },
+	{ _MMIO(0x9888), 0x0adcc000 },
+	{ _MMIO(0x9888), 0x0cdcc000 },
+	{ _MMIO(0x9888), 0x04dcc000 },
+	{ _MMIO(0x9888), 0x00db0011 },
+	{ _MMIO(0x9888), 0x06db0900 },
+	{ _MMIO(0x9888), 0x08db0a13 },
+	{ _MMIO(0x9888), 0x0adb0b15 },
+	{ _MMIO(0x9888), 0x0cdb2317 },
+	{ _MMIO(0x9888), 0x04db21b7 },
+	{ _MMIO(0x9888), 0x10db0000 },
+	{ _MMIO(0x9888), 0x0edb0000 },
+	{ _MMIO(0x9888), 0x1adb0000 },
+	{ _MMIO(0x9888), 0x0c9fa800 },
+	{ _MMIO(0x9888), 0x0e9faa2a },
+	{ _MMIO(0x9888), 0x109f02aa },
+	{ _MMIO(0x9888), 0x00b84000 },
+	{ _MMIO(0x9888), 0x0eb84000 },
+	{ _MMIO(0x9888), 0x16b84000 },
+	{ _MMIO(0x9888), 0x18b81555 },
+	{ _MMIO(0x9888), 0x02b84000 },
+	{ _MMIO(0x9888), 0x04b84000 },
+	{ _MMIO(0x9888), 0x06b84000 },
+	{ _MMIO(0x9888), 0x08b84000 },
+	{ _MMIO(0x9888), 0x0ab84000 },
+	{ _MMIO(0x9888), 0x00b9a000 },
+	{ _MMIO(0x9888), 0x06b98000 },
+	{ _MMIO(0x9888), 0x08b9a000 },
+	{ _MMIO(0x9888), 0x0ab9a000 },
+	{ _MMIO(0x9888), 0x0cb9a000 },
+	{ _MMIO(0x9888), 0x0eb9a000 },
+	{ _MMIO(0x9888), 0x02b9a000 },
+	{ _MMIO(0x9888), 0x04b9a000 },
+	{ _MMIO(0x9888), 0x01888000 },
+	{ _MMIO(0x9888), 0x0d88f800 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x07888000 },
+	{ _MMIO(0x9888), 0x09888000 },
+	{ _MMIO(0x9888), 0x0b888000 },
+	{ _MMIO(0x9888), 0x238b5540 },
+	{ _MMIO(0x9888), 0x258baaa2 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x018c4000 },
+	{ _MMIO(0x9888), 0x0f8c4000 },
+	{ _MMIO(0x9888), 0x178c2000 },
+	{ _MMIO(0x9888), 0x198c5500 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x078c4000 },
+	{ _MMIO(0x9888), 0x098c4000 },
+	{ _MMIO(0x9888), 0x0b8c4000 },
+	{ _MMIO(0x9888), 0x018da000 },
+	{ _MMIO(0x9888), 0x078d8000 },
+	{ _MMIO(0x9888), 0x098da000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x038da000 },
+	{ _MMIO(0x9888), 0x058da000 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa2 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg mux_config_compute_extended_5_subslices_0x20[] = {
+	{ _MMIO(0x9888), 0x10b800e0 },
+	{ _MMIO(0x9888), 0x14ba0160 },
+	{ _MMIO(0x9888), 0x16ba2800 },
+	{ _MMIO(0x9888), 0x18ba0120 },
+	{ _MMIO(0x9888), 0x0c9fa800 },
+	{ _MMIO(0x9888), 0x0e9faa2a },
+	{ _MMIO(0x9888), 0x109f02aa },
+	{ _MMIO(0x9888), 0x0eb8a5c1 },
+	{ _MMIO(0x9888), 0x00b8a100 },
+	{ _MMIO(0x9888), 0x02b8204c },
+	{ _MMIO(0x9888), 0x16b88000 },
+	{ _MMIO(0x9888), 0x18b802aa },
+	{ _MMIO(0x9888), 0x04b80000 },
+	{ _MMIO(0x9888), 0x06b80000 },
+	{ _MMIO(0x9888), 0x08b88000 },
+	{ _MMIO(0x9888), 0x0ab88000 },
+	{ _MMIO(0x9888), 0x00b9a000 },
+	{ _MMIO(0x9888), 0x06b98000 },
+	{ _MMIO(0x9888), 0x08b9a000 },
+	{ _MMIO(0x9888), 0x0ab9a000 },
+	{ _MMIO(0x9888), 0x0cb9a000 },
+	{ _MMIO(0x9888), 0x0eb9a000 },
+	{ _MMIO(0x9888), 0x02b9a000 },
+	{ _MMIO(0x9888), 0x04b9a000 },
+	{ _MMIO(0x9888), 0x00ba0011 },
+	{ _MMIO(0x9888), 0x06ba0900 },
+	{ _MMIO(0x9888), 0x08ba0a13 },
+	{ _MMIO(0x9888), 0x0aba0b15 },
+	{ _MMIO(0x9888), 0x0cba2317 },
+	{ _MMIO(0x9888), 0x04ba21b7 },
+	{ _MMIO(0x9888), 0x10ba0000 },
+	{ _MMIO(0x9888), 0x0eba0000 },
+	{ _MMIO(0x9888), 0x1aba0000 },
+	{ _MMIO(0x9888), 0x01888000 },
+	{ _MMIO(0x9888), 0x0d88f800 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x07888000 },
+	{ _MMIO(0x9888), 0x09888000 },
+	{ _MMIO(0x9888), 0x0b888000 },
+	{ _MMIO(0x9888), 0x238b5540 },
+	{ _MMIO(0x9888), 0x258baaa2 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x018c4000 },
+	{ _MMIO(0x9888), 0x0f8c4000 },
+	{ _MMIO(0x9888), 0x178c2000 },
+	{ _MMIO(0x9888), 0x198c5500 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x078c4000 },
+	{ _MMIO(0x9888), 0x098c4000 },
+	{ _MMIO(0x9888), 0x0b8c4000 },
+	{ _MMIO(0x9888), 0x018da000 },
+	{ _MMIO(0x9888), 0x078d8000 },
+	{ _MMIO(0x9888), 0x098da000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x038da000 },
+	{ _MMIO(0x9888), 0x058da000 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa2 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800000 },
+};
+
+static const struct i915_oa_reg *
+get_compute_extended_mux_config(struct drm_i915_private *dev_priv,
+				int *len)
+{
+	if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x01) {
+		*len = ARRAY_SIZE(mux_config_compute_extended_0_subslices_0x01);
+		return mux_config_compute_extended_0_subslices_0x01;
+	} else if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x02) {
+		*len = ARRAY_SIZE(mux_config_compute_extended_2_subslices_0x02);
+		return mux_config_compute_extended_2_subslices_0x02;
+	} else if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x04) {
+		*len = ARRAY_SIZE(mux_config_compute_extended_4_subslices_0x04);
+		return mux_config_compute_extended_4_subslices_0x04;
+	} else if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x08) {
+		*len = ARRAY_SIZE(mux_config_compute_extended_1_subslices_0x08);
+		return mux_config_compute_extended_1_subslices_0x08;
+	} else if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x10) {
+		*len = ARRAY_SIZE(mux_config_compute_extended_3_subslices_0x10);
+		return mux_config_compute_extended_3_subslices_0x10;
+	} else if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x20) {
+		*len = ARRAY_SIZE(mux_config_compute_extended_5_subslices_0x20);
+		return mux_config_compute_extended_5_subslices_0x20;
+	}
+
+	*len = 0;
+	return NULL;
+}
+
+static const struct i915_oa_reg b_counter_config_compute_l3_cache[] = {
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0x30800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x30800000 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007fffa },
+	{ _MMIO(0x2774), 0x0000fefe },
+	{ _MMIO(0x2778), 0x0007fffa },
+	{ _MMIO(0x277c), 0x0000fefd },
+	{ _MMIO(0x2790), 0x0007fffa },
+	{ _MMIO(0x2794), 0x0000fbef },
+	{ _MMIO(0x2798), 0x0007fffa },
+	{ _MMIO(0x279c), 0x0000fbdf },
+};
+
+static const struct i915_oa_reg flex_eu_config_compute_l3_cache[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00000003 },
+	{ _MMIO(0xe658), 0x00002001 },
+	{ _MMIO(0xe758), 0x00101100 },
+	{ _MMIO(0xe45c), 0x00201200 },
+	{ _MMIO(0xe55c), 0x00301300 },
+	{ _MMIO(0xe65c), 0x00401400 },
+};
+
+static const struct i915_oa_reg mux_config_compute_l3_cache[] = {
+	{ _MMIO(0x9888), 0x143f00b3 },
+	{ _MMIO(0x9888), 0x14bf00b3 },
+	{ _MMIO(0x9888), 0x138303c0 },
+	{ _MMIO(0x9888), 0x3b800060 },
+	{ _MMIO(0x9888), 0x3d800805 },
+	{ _MMIO(0x9888), 0x003f0029 },
+	{ _MMIO(0x9888), 0x063f1400 },
+	{ _MMIO(0x9888), 0x083f1225 },
+	{ _MMIO(0x9888), 0x0e3f1327 },
+	{ _MMIO(0x9888), 0x103f0000 },
+	{ _MMIO(0x9888), 0x005a4000 },
+	{ _MMIO(0x9888), 0x065a8000 },
+	{ _MMIO(0x9888), 0x085ac000 },
+	{ _MMIO(0x9888), 0x0e5ac000 },
+	{ _MMIO(0x9888), 0x001d4000 },
+	{ _MMIO(0x9888), 0x061d8000 },
+	{ _MMIO(0x9888), 0x081dc000 },
+	{ _MMIO(0x9888), 0x0e1dc000 },
+	{ _MMIO(0x9888), 0x0c1f0800 },
+	{ _MMIO(0x9888), 0x0e1f2a00 },
+	{ _MMIO(0x9888), 0x101f0280 },
+	{ _MMIO(0x9888), 0x00391000 },
+	{ _MMIO(0x9888), 0x06394000 },
+	{ _MMIO(0x9888), 0x08395000 },
+	{ _MMIO(0x9888), 0x0e395000 },
+	{ _MMIO(0x9888), 0x0abf1429 },
+	{ _MMIO(0x9888), 0x0cbf1225 },
+	{ _MMIO(0x9888), 0x00bf1380 },
+	{ _MMIO(0x9888), 0x02bf0026 },
+	{ _MMIO(0x9888), 0x10bf0000 },
+	{ _MMIO(0x9888), 0x0adac000 },
+	{ _MMIO(0x9888), 0x0cdac000 },
+	{ _MMIO(0x9888), 0x00da8000 },
+	{ _MMIO(0x9888), 0x02da4000 },
+	{ _MMIO(0x9888), 0x0a9dc000 },
+	{ _MMIO(0x9888), 0x0c9dc000 },
+	{ _MMIO(0x9888), 0x009d8000 },
+	{ _MMIO(0x9888), 0x029d4000 },
+	{ _MMIO(0x9888), 0x0e9f8000 },
+	{ _MMIO(0x9888), 0x109f002a },
+	{ _MMIO(0x9888), 0x0c9fa000 },
+	{ _MMIO(0x9888), 0x0ab95000 },
+	{ _MMIO(0x9888), 0x0cb95000 },
+	{ _MMIO(0x9888), 0x00b94000 },
+	{ _MMIO(0x9888), 0x02b91000 },
+	{ _MMIO(0x9888), 0x0d88c000 },
+	{ _MMIO(0x9888), 0x0f880003 },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8a8020 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x238b0520 },
+	{ _MMIO(0x9888), 0x258ba950 },
+	{ _MMIO(0x9888), 0x278b0016 },
+	{ _MMIO(0x9888), 0x198c5400 },
+	{ _MMIO(0x9888), 0x1b8c0001 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x018d8000 },
+	{ _MMIO(0x9888), 0x038d2000 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaa0 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x03835180 },
+	{ _MMIO(0x9888), 0x05834022 },
+	{ _MMIO(0x9888), 0x11830000 },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x07830000 },
+	{ _MMIO(0x9888), 0x09830000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x07848000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x05844000 },
+	{ _MMIO(0x9888), 0x1b80c137 },
+	{ _MMIO(0x9888), 0x1d80c147 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x17808000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x15804000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d801000 },
+	{ _MMIO(0x9888), 0x4f800111 },
+	{ _MMIO(0x9888), 0x43800842 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800840 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800800 },
+	{ _MMIO(0x9888), 0x418014a2 },
+};
+
+static const struct i915_oa_reg *
+get_compute_l3_cache_mux_config(struct drm_i915_private *dev_priv,
+				int *len)
+{
+	*len = ARRAY_SIZE(mux_config_compute_l3_cache);
+	return mux_config_compute_l3_cache;
+}
+
+static const struct i915_oa_reg b_counter_config_data_port_reads_coalescing[] = {
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x274c), 0xba98ba98 },
+	{ _MMIO(0x2748), 0xba98ba98 },
+	{ _MMIO(0x2744), 0x00003377 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007fff2 },
+	{ _MMIO(0x2774), 0x00007ff0 },
+	{ _MMIO(0x2778), 0x0007ffe2 },
+	{ _MMIO(0x277c), 0x00007ff0 },
+	{ _MMIO(0x2780), 0x0007ffc2 },
+	{ _MMIO(0x2784), 0x00007ff0 },
+	{ _MMIO(0x2788), 0x0007ff82 },
+	{ _MMIO(0x278c), 0x00007ff0 },
+	{ _MMIO(0x2790), 0x0007fffa },
+	{ _MMIO(0x2794), 0x0000bfef },
+	{ _MMIO(0x2798), 0x0007fffa },
+	{ _MMIO(0x279c), 0x0000bfdf },
+	{ _MMIO(0x27a0), 0x0007fffa },
+	{ _MMIO(0x27a4), 0x0000bfbf },
+	{ _MMIO(0x27a8), 0x0007fffa },
+	{ _MMIO(0x27ac), 0x0000bf7f },
+};
+
+static const struct i915_oa_reg flex_eu_config_data_port_reads_coalescing[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00000003 },
+	{ _MMIO(0xe658), 0x00002001 },
+	{ _MMIO(0xe758), 0x00778008 },
+	{ _MMIO(0xe45c), 0x00088078 },
+	{ _MMIO(0xe55c), 0x00808708 },
+	{ _MMIO(0xe65c), 0x00a08908 },
+};
+
+static const struct i915_oa_reg mux_config_data_port_reads_coalescing_0_subslices_0x01[] = {
+	{ _MMIO(0x9888), 0x103d0005 },
+	{ _MMIO(0x9888), 0x163d240b },
+	{ _MMIO(0x9888), 0x1058022f },
+	{ _MMIO(0x9888), 0x185b5520 },
+	{ _MMIO(0x9888), 0x198b0003 },
+	{ _MMIO(0x9888), 0x005cc000 },
+	{ _MMIO(0x9888), 0x065cc000 },
+	{ _MMIO(0x9888), 0x085cc000 },
+	{ _MMIO(0x9888), 0x0a5cc000 },
+	{ _MMIO(0x9888), 0x0c5cc000 },
+	{ _MMIO(0x9888), 0x0e5cc000 },
+	{ _MMIO(0x9888), 0x025c4000 },
+	{ _MMIO(0x9888), 0x045c8000 },
+	{ _MMIO(0x9888), 0x003d0000 },
+	{ _MMIO(0x9888), 0x063d00b0 },
+	{ _MMIO(0x9888), 0x083d0182 },
+	{ _MMIO(0x9888), 0x0a3d10a0 },
+	{ _MMIO(0x9888), 0x0c3d11a2 },
+	{ _MMIO(0x9888), 0x0e3d0000 },
+	{ _MMIO(0x9888), 0x183d0000 },
+	{ _MMIO(0x9888), 0x1a3d0000 },
+	{ _MMIO(0x9888), 0x0e582242 },
+	{ _MMIO(0x9888), 0x00586700 },
+	{ _MMIO(0x9888), 0x0258004f },
+	{ _MMIO(0x9888), 0x0658c000 },
+	{ _MMIO(0x9888), 0x0858c000 },
+	{ _MMIO(0x9888), 0x0a58c000 },
+	{ _MMIO(0x9888), 0x0c58c000 },
+	{ _MMIO(0x9888), 0x045b6300 },
+	{ _MMIO(0x9888), 0x105b0000 },
+	{ _MMIO(0x9888), 0x005b4000 },
+	{ _MMIO(0x9888), 0x0e5b4000 },
+	{ _MMIO(0x9888), 0x1a5b0155 },
+	{ _MMIO(0x9888), 0x025b4000 },
+	{ _MMIO(0x9888), 0x0a5b0000 },
+	{ _MMIO(0x9888), 0x0c5b4000 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1faaa0 },
+	{ _MMIO(0x9888), 0x101f02aa },
+	{ _MMIO(0x9888), 0x00384000 },
+	{ _MMIO(0x9888), 0x0e384000 },
+	{ _MMIO(0x9888), 0x16384000 },
+	{ _MMIO(0x9888), 0x18381555 },
+	{ _MMIO(0x9888), 0x02384000 },
+	{ _MMIO(0x9888), 0x04384000 },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x0c384000 },
+	{ _MMIO(0x9888), 0x0039a000 },
+	{ _MMIO(0x9888), 0x0639a000 },
+	{ _MMIO(0x9888), 0x0839a000 },
+	{ _MMIO(0x9888), 0x0a39a000 },
+	{ _MMIO(0x9888), 0x0c39a000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x02392000 },
+	{ _MMIO(0x9888), 0x04398000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8aaaa0 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x038b6300 },
+	{ _MMIO(0x9888), 0x058b0062 },
+	{ _MMIO(0x9888), 0x118b0000 },
+	{ _MMIO(0x9888), 0x238b02a0 },
+	{ _MMIO(0x9888), 0x258b5555 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaaa },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x0784c000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x1780c000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d801000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800001 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800041 },
+};
+
+static const struct i915_oa_reg *
+get_data_port_reads_coalescing_mux_config(struct drm_i915_private *dev_priv,
+					  int *len)
+{
+	if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x01) {
+		*len = ARRAY_SIZE(mux_config_data_port_reads_coalescing_0_subslices_0x01);
+		return mux_config_data_port_reads_coalescing_0_subslices_0x01;
+	}
+
+	*len = 0;
+	return NULL;
+}
+
+static const struct i915_oa_reg b_counter_config_data_port_writes_coalescing[] = {
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x274c), 0xba98ba98 },
+	{ _MMIO(0x2748), 0xba98ba98 },
+	{ _MMIO(0x2744), 0x00003377 },
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2770), 0x0007ff72 },
+	{ _MMIO(0x2774), 0x0000bfd0 },
+	{ _MMIO(0x2778), 0x0007ff62 },
+	{ _MMIO(0x277c), 0x0000bfd0 },
+	{ _MMIO(0x2780), 0x0007ff42 },
+	{ _MMIO(0x2784), 0x0000bfd0 },
+	{ _MMIO(0x2788), 0x0007ff02 },
+	{ _MMIO(0x278c), 0x0000bfd0 },
+	{ _MMIO(0x2790), 0x0005fff2 },
+	{ _MMIO(0x2794), 0x0000bfd0 },
+	{ _MMIO(0x2798), 0x0005ffe2 },
+	{ _MMIO(0x279c), 0x0000bfd0 },
+	{ _MMIO(0x27a0), 0x0005ffc2 },
+	{ _MMIO(0x27a4), 0x0000bfd0 },
+	{ _MMIO(0x27a8), 0x0005ff82 },
+	{ _MMIO(0x27ac), 0x0000bfd0 },
+};
+
+static const struct i915_oa_reg flex_eu_config_data_port_writes_coalescing[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00000003 },
+	{ _MMIO(0xe658), 0x00002001 },
+	{ _MMIO(0xe758), 0x00778008 },
+	{ _MMIO(0xe45c), 0x00088078 },
+	{ _MMIO(0xe55c), 0x00808708 },
+	{ _MMIO(0xe65c), 0x00a08908 },
+};
+
+static const struct i915_oa_reg mux_config_data_port_writes_coalescing_0_subslices_0x01[] = {
+	{ _MMIO(0x9888), 0x103d0005 },
+	{ _MMIO(0x9888), 0x143d0120 },
+	{ _MMIO(0x9888), 0x163d2400 },
+	{ _MMIO(0x9888), 0x1058022f },
+	{ _MMIO(0x9888), 0x105b0000 },
+	{ _MMIO(0x9888), 0x198b0003 },
+	{ _MMIO(0x9888), 0x005cc000 },
+	{ _MMIO(0x9888), 0x065cc000 },
+	{ _MMIO(0x9888), 0x085cc000 },
+	{ _MMIO(0x9888), 0x0a5cc000 },
+	{ _MMIO(0x9888), 0x0e5cc000 },
+	{ _MMIO(0x9888), 0x025c4000 },
+	{ _MMIO(0x9888), 0x045c8000 },
+	{ _MMIO(0x9888), 0x003d0000 },
+	{ _MMIO(0x9888), 0x063d0094 },
+	{ _MMIO(0x9888), 0x083d0182 },
+	{ _MMIO(0x9888), 0x0a3d1814 },
+	{ _MMIO(0x9888), 0x0e3d0000 },
+	{ _MMIO(0x9888), 0x183d0000 },
+	{ _MMIO(0x9888), 0x1a3d0000 },
+	{ _MMIO(0x9888), 0x0c3d0000 },
+	{ _MMIO(0x9888), 0x0e582242 },
+	{ _MMIO(0x9888), 0x00586700 },
+	{ _MMIO(0x9888), 0x0258004f },
+	{ _MMIO(0x9888), 0x0658c000 },
+	{ _MMIO(0x9888), 0x0858c000 },
+	{ _MMIO(0x9888), 0x0a58c000 },
+	{ _MMIO(0x9888), 0x045b6a80 },
+	{ _MMIO(0x9888), 0x005b4000 },
+	{ _MMIO(0x9888), 0x0e5b4000 },
+	{ _MMIO(0x9888), 0x185b5400 },
+	{ _MMIO(0x9888), 0x1a5b0141 },
+	{ _MMIO(0x9888), 0x025b4000 },
+	{ _MMIO(0x9888), 0x0a5b0000 },
+	{ _MMIO(0x9888), 0x0c5b4000 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1faaa0 },
+	{ _MMIO(0x9888), 0x101f0282 },
+	{ _MMIO(0x9888), 0x00384000 },
+	{ _MMIO(0x9888), 0x0e384000 },
+	{ _MMIO(0x9888), 0x16384000 },
+	{ _MMIO(0x9888), 0x18381415 },
+	{ _MMIO(0x9888), 0x02384000 },
+	{ _MMIO(0x9888), 0x04384000 },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x0c384000 },
+	{ _MMIO(0x9888), 0x0039a000 },
+	{ _MMIO(0x9888), 0x0639a000 },
+	{ _MMIO(0x9888), 0x0839a000 },
+	{ _MMIO(0x9888), 0x0a39a000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x02392000 },
+	{ _MMIO(0x9888), 0x04398000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x198a8000 },
+	{ _MMIO(0x9888), 0x1b8a82a0 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x038b6300 },
+	{ _MMIO(0x9888), 0x058b0062 },
+	{ _MMIO(0x9888), 0x118b0000 },
+	{ _MMIO(0x9888), 0x238b02a0 },
+	{ _MMIO(0x9888), 0x258b1555 },
+	{ _MMIO(0x9888), 0x278b0014 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x21852aaa },
+	{ _MMIO(0x9888), 0x23850028 },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830141 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x0784c000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x1780c000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0xd24), 0x00000000 },
+	{ _MMIO(0x9888), 0x4d801000 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f800001 },
+	{ _MMIO(0x9888), 0x43800000 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800420 },
+	{ _MMIO(0x9888), 0x3f800421 },
+	{ _MMIO(0x9888), 0x41800041 },
+};
+
+static const struct i915_oa_reg *
+get_data_port_writes_coalescing_mux_config(struct drm_i915_private *dev_priv,
+					   int *len)
+{
+	if (INTEL_INFO(dev_priv)->sseu.subslice_mask & 0x01) {
+		*len = ARRAY_SIZE(mux_config_data_port_writes_coalescing_0_subslices_0x01);
+		return mux_config_data_port_writes_coalescing_0_subslices_0x01;
+	}
+
+	*len = 0;
+	return NULL;
+}
+
+static const struct i915_oa_reg b_counter_config_hdc_and_sf[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0x10800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x00800000 },
+	{ _MMIO(0x2770), 0x00000002 },
+	{ _MMIO(0x2774), 0x0000fff7 },
+};
+
+static const struct i915_oa_reg flex_eu_config_hdc_and_sf[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_hdc_and_sf[] = {
+	{ _MMIO(0x9888), 0x105c0232 },
+	{ _MMIO(0x9888), 0x10580232 },
+	{ _MMIO(0x9888), 0x10380232 },
+	{ _MMIO(0x9888), 0x10dc0232 },
+	{ _MMIO(0x9888), 0x10d80232 },
+	{ _MMIO(0x9888), 0x10b80232 },
+	{ _MMIO(0x9888), 0x118e4400 },
+	{ _MMIO(0x9888), 0x025c6080 },
+	{ _MMIO(0x9888), 0x045c004b },
+	{ _MMIO(0x9888), 0x005c8000 },
+	{ _MMIO(0x9888), 0x00582080 },
+	{ _MMIO(0x9888), 0x0258004b },
+	{ _MMIO(0x9888), 0x025b4000 },
+	{ _MMIO(0x9888), 0x045b4000 },
+	{ _MMIO(0x9888), 0x0c1fa000 },
+	{ _MMIO(0x9888), 0x0e1f00aa },
+	{ _MMIO(0x9888), 0x04386080 },
+	{ _MMIO(0x9888), 0x0638404b },
+	{ _MMIO(0x9888), 0x02384000 },
+	{ _MMIO(0x9888), 0x08384000 },
+	{ _MMIO(0x9888), 0x0a380000 },
+	{ _MMIO(0x9888), 0x0c380000 },
+	{ _MMIO(0x9888), 0x00398000 },
+	{ _MMIO(0x9888), 0x0239a000 },
+	{ _MMIO(0x9888), 0x0439a000 },
+	{ _MMIO(0x9888), 0x06392000 },
+	{ _MMIO(0x9888), 0x0cdc25c1 },
+	{ _MMIO(0x9888), 0x0adcc000 },
+	{ _MMIO(0x9888), 0x0ad825c1 },
+	{ _MMIO(0x9888), 0x18db4000 },
+	{ _MMIO(0x9888), 0x1adb0001 },
+	{ _MMIO(0x9888), 0x0e9f8000 },
+	{ _MMIO(0x9888), 0x109f02aa },
+	{ _MMIO(0x9888), 0x0eb825c1 },
+	{ _MMIO(0x9888), 0x18b80154 },
+	{ _MMIO(0x9888), 0x0ab9a000 },
+	{ _MMIO(0x9888), 0x0cb9a000 },
+	{ _MMIO(0x9888), 0x0eb9a000 },
+	{ _MMIO(0x9888), 0x0d88c000 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x258baa05 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x238b2a80 },
+	{ _MMIO(0x9888), 0x198c5400 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x098dc000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x098e05c0 },
+	{ _MMIO(0x9888), 0x058e0000 },
+	{ _MMIO(0x9888), 0x198f0020 },
+	{ _MMIO(0x9888), 0x2185aa0a },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x19835000 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x09848000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x19808000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x51800040 },
+	{ _MMIO(0x9888), 0x43800400 },
+	{ _MMIO(0x9888), 0x45800800 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47800c62 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f801042 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x418014a4 },
+};
+
+static const struct i915_oa_reg *
+get_hdc_and_sf_mux_config(struct drm_i915_private *dev_priv,
+			  int *len)
+{
+	*len = ARRAY_SIZE(mux_config_hdc_and_sf);
+	return mux_config_hdc_and_sf;
+}
+
+static const struct i915_oa_reg b_counter_config_l3_1[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2770), 0x00100070 },
+	{ _MMIO(0x2774), 0x0000fff1 },
+	{ _MMIO(0x2778), 0x00014002 },
+	{ _MMIO(0x277c), 0x0000c3ff },
+	{ _MMIO(0x2780), 0x00010002 },
+	{ _MMIO(0x2784), 0x0000c7ff },
+	{ _MMIO(0x2788), 0x00004002 },
+	{ _MMIO(0x278c), 0x0000d3ff },
+	{ _MMIO(0x2790), 0x00100700 },
+	{ _MMIO(0x2794), 0x0000ff1f },
+	{ _MMIO(0x2798), 0x00001402 },
+	{ _MMIO(0x279c), 0x0000fc3f },
+	{ _MMIO(0x27a0), 0x00001002 },
+	{ _MMIO(0x27a4), 0x0000fc7f },
+	{ _MMIO(0x27a8), 0x00000402 },
+	{ _MMIO(0x27ac), 0x0000fd3f },
+};
+
+static const struct i915_oa_reg flex_eu_config_l3_1[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_l3_1[] = {
+	{ _MMIO(0x9888), 0x10bf03da },
+	{ _MMIO(0x9888), 0x14bf0001 },
+	{ _MMIO(0x9888), 0x12980340 },
+	{ _MMIO(0x9888), 0x12990340 },
+	{ _MMIO(0x9888), 0x0cbf1187 },
+	{ _MMIO(0x9888), 0x0ebf1205 },
+	{ _MMIO(0x9888), 0x00bf0500 },
+	{ _MMIO(0x9888), 0x02bf042b },
+	{ _MMIO(0x9888), 0x04bf002c },
+	{ _MMIO(0x9888), 0x0cdac000 },
+	{ _MMIO(0x9888), 0x0edac000 },
+	{ _MMIO(0x9888), 0x00da8000 },
+	{ _MMIO(0x9888), 0x02dac000 },
+	{ _MMIO(0x9888), 0x04da4000 },
+	{ _MMIO(0x9888), 0x04983400 },
+	{ _MMIO(0x9888), 0x10980000 },
+	{ _MMIO(0x9888), 0x06990034 },
+	{ _MMIO(0x9888), 0x10990000 },
+	{ _MMIO(0x9888), 0x0c9dc000 },
+	{ _MMIO(0x9888), 0x0e9dc000 },
+	{ _MMIO(0x9888), 0x009d8000 },
+	{ _MMIO(0x9888), 0x029dc000 },
+	{ _MMIO(0x9888), 0x049d4000 },
+	{ _MMIO(0x9888), 0x109f02a8 },
+	{ _MMIO(0x9888), 0x0c9fa000 },
+	{ _MMIO(0x9888), 0x0e9f00ba },
+	{ _MMIO(0x9888), 0x0cb88000 },
+	{ _MMIO(0x9888), 0x0cb95000 },
+	{ _MMIO(0x9888), 0x0eb95000 },
+	{ _MMIO(0x9888), 0x00b94000 },
+	{ _MMIO(0x9888), 0x02b95000 },
+	{ _MMIO(0x9888), 0x04b91000 },
+	{ _MMIO(0x9888), 0x06b92000 },
+	{ _MMIO(0x9888), 0x0cba4000 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x07888000 },
+	{ _MMIO(0x9888), 0x09888000 },
+	{ _MMIO(0x9888), 0x0b888000 },
+	{ _MMIO(0x9888), 0x0d880400 },
+	{ _MMIO(0x9888), 0x258b800a },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x238b5500 },
+	{ _MMIO(0x9888), 0x198c4000 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x078c4000 },
+	{ _MMIO(0x9888), 0x098c4000 },
+	{ _MMIO(0x9888), 0x0b8c4000 },
+	{ _MMIO(0x9888), 0x0d8c4000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x018d8000 },
+	{ _MMIO(0x9888), 0x038da000 },
+	{ _MMIO(0x9888), 0x058da000 },
+	{ _MMIO(0x9888), 0x078d2000 },
+	{ _MMIO(0x9888), 0x2185800a },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x1b830154 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x47800000 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x41800060 },
+};
+
+static const struct i915_oa_reg *
+get_l3_1_mux_config(struct drm_i915_private *dev_priv,
+		    int *len)
+{
+	*len = ARRAY_SIZE(mux_config_l3_1);
+	return mux_config_l3_1;
+}
+
+static const struct i915_oa_reg b_counter_config_l3_2[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2770), 0x00100070 },
+	{ _MMIO(0x2774), 0x0000fff1 },
+	{ _MMIO(0x2778), 0x00014002 },
+	{ _MMIO(0x277c), 0x0000c3ff },
+	{ _MMIO(0x2780), 0x00010002 },
+	{ _MMIO(0x2784), 0x0000c7ff },
+	{ _MMIO(0x2788), 0x00004002 },
+	{ _MMIO(0x278c), 0x0000d3ff },
+	{ _MMIO(0x2790), 0x00100700 },
+	{ _MMIO(0x2794), 0x0000ff1f },
+	{ _MMIO(0x2798), 0x00001402 },
+	{ _MMIO(0x279c), 0x0000fc3f },
+	{ _MMIO(0x27a0), 0x00001002 },
+	{ _MMIO(0x27a4), 0x0000fc7f },
+	{ _MMIO(0x27a8), 0x00000402 },
+	{ _MMIO(0x27ac), 0x0000fd3f },
+};
+
+static const struct i915_oa_reg flex_eu_config_l3_2[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_l3_2[] = {
+	{ _MMIO(0x9888), 0x103f03da },
+	{ _MMIO(0x9888), 0x143f0001 },
+	{ _MMIO(0x9888), 0x12180340 },
+	{ _MMIO(0x9888), 0x12190340 },
+	{ _MMIO(0x9888), 0x0c3f1187 },
+	{ _MMIO(0x9888), 0x0e3f1205 },
+	{ _MMIO(0x9888), 0x003f0500 },
+	{ _MMIO(0x9888), 0x023f042b },
+	{ _MMIO(0x9888), 0x043f002c },
+	{ _MMIO(0x9888), 0x0c5ac000 },
+	{ _MMIO(0x9888), 0x0e5ac000 },
+	{ _MMIO(0x9888), 0x005a8000 },
+	{ _MMIO(0x9888), 0x025ac000 },
+	{ _MMIO(0x9888), 0x045a4000 },
+	{ _MMIO(0x9888), 0x04183400 },
+	{ _MMIO(0x9888), 0x10180000 },
+	{ _MMIO(0x9888), 0x06190034 },
+	{ _MMIO(0x9888), 0x10190000 },
+	{ _MMIO(0x9888), 0x0c1dc000 },
+	{ _MMIO(0x9888), 0x0e1dc000 },
+	{ _MMIO(0x9888), 0x001d8000 },
+	{ _MMIO(0x9888), 0x021dc000 },
+	{ _MMIO(0x9888), 0x041d4000 },
+	{ _MMIO(0x9888), 0x101f02a8 },
+	{ _MMIO(0x9888), 0x0c1fa000 },
+	{ _MMIO(0x9888), 0x0e1f00ba },
+	{ _MMIO(0x9888), 0x0c388000 },
+	{ _MMIO(0x9888), 0x0c395000 },
+	{ _MMIO(0x9888), 0x0e395000 },
+	{ _MMIO(0x9888), 0x00394000 },
+	{ _MMIO(0x9888), 0x02395000 },
+	{ _MMIO(0x9888), 0x04391000 },
+	{ _MMIO(0x9888), 0x06392000 },
+	{ _MMIO(0x9888), 0x0c3a4000 },
+	{ _MMIO(0x9888), 0x1b8aa800 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x258b4005 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x238b2a80 },
+	{ _MMIO(0x9888), 0x2185800a },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x1b830154 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x45800000 },
+	{ _MMIO(0x9888), 0x47800000 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f800000 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x41800060 },
+};
+
+static const struct i915_oa_reg *
+get_l3_2_mux_config(struct drm_i915_private *dev_priv,
+		    int *len)
+{
+	*len = ARRAY_SIZE(mux_config_l3_2);
+	return mux_config_l3_2;
+}
+
+static const struct i915_oa_reg b_counter_config_l3_3[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2770), 0x00100070 },
+	{ _MMIO(0x2774), 0x0000fff1 },
+	{ _MMIO(0x2778), 0x00014002 },
+	{ _MMIO(0x277c), 0x0000c3ff },
+	{ _MMIO(0x2780), 0x00010002 },
+	{ _MMIO(0x2784), 0x0000c7ff },
+	{ _MMIO(0x2788), 0x00004002 },
+	{ _MMIO(0x278c), 0x0000d3ff },
+	{ _MMIO(0x2790), 0x00100700 },
+	{ _MMIO(0x2794), 0x0000ff1f },
+	{ _MMIO(0x2798), 0x00001402 },
+	{ _MMIO(0x279c), 0x0000fc3f },
+	{ _MMIO(0x27a0), 0x00001002 },
+	{ _MMIO(0x27a4), 0x0000fc7f },
+	{ _MMIO(0x27a8), 0x00000402 },
+	{ _MMIO(0x27ac), 0x0000fd3f },
+};
+
+static const struct i915_oa_reg flex_eu_config_l3_3[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_l3_3[] = {
+	{ _MMIO(0x9888), 0x121b0340 },
+	{ _MMIO(0x9888), 0x103f0274 },
+	{ _MMIO(0x9888), 0x123f0000 },
+	{ _MMIO(0x9888), 0x129b0340 },
+	{ _MMIO(0x9888), 0x10bf0274 },
+	{ _MMIO(0x9888), 0x12bf0000 },
+	{ _MMIO(0x9888), 0x041b3400 },
+	{ _MMIO(0x9888), 0x101b0000 },
+	{ _MMIO(0x9888), 0x045c8000 },
+	{ _MMIO(0x9888), 0x0a3d4000 },
+	{ _MMIO(0x9888), 0x003f0080 },
+	{ _MMIO(0x9888), 0x023f0793 },
+	{ _MMIO(0x9888), 0x043f0014 },
+	{ _MMIO(0x9888), 0x04588000 },
+	{ _MMIO(0x9888), 0x005a8000 },
+	{ _MMIO(0x9888), 0x025ac000 },
+	{ _MMIO(0x9888), 0x045a4000 },
+	{ _MMIO(0x9888), 0x0a5b4000 },
+	{ _MMIO(0x9888), 0x001d8000 },
+	{ _MMIO(0x9888), 0x021dc000 },
+	{ _MMIO(0x9888), 0x041d4000 },
+	{ _MMIO(0x9888), 0x0c1fa000 },
+	{ _MMIO(0x9888), 0x0e1f002a },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x00394000 },
+	{ _MMIO(0x9888), 0x02395000 },
+	{ _MMIO(0x9888), 0x04399000 },
+	{ _MMIO(0x9888), 0x069b0034 },
+	{ _MMIO(0x9888), 0x109b0000 },
+	{ _MMIO(0x9888), 0x06dc4000 },
+	{ _MMIO(0x9888), 0x0cbd4000 },
+	{ _MMIO(0x9888), 0x0cbf0981 },
+	{ _MMIO(0x9888), 0x0ebf0a0f },
+	{ _MMIO(0x9888), 0x06d84000 },
+	{ _MMIO(0x9888), 0x0cdac000 },
+	{ _MMIO(0x9888), 0x0edac000 },
+	{ _MMIO(0x9888), 0x0cdb4000 },
+	{ _MMIO(0x9888), 0x0c9dc000 },
+	{ _MMIO(0x9888), 0x0e9dc000 },
+	{ _MMIO(0x9888), 0x109f02a8 },
+	{ _MMIO(0x9888), 0x0e9f0080 },
+	{ _MMIO(0x9888), 0x0cb84000 },
+	{ _MMIO(0x9888), 0x0cb95000 },
+	{ _MMIO(0x9888), 0x0eb95000 },
+	{ _MMIO(0x9888), 0x06b92000 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x0d880400 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x258b8009 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x238b2a80 },
+	{ _MMIO(0x9888), 0x198c4000 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x0d8c4000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x078d2000 },
+	{ _MMIO(0x9888), 0x2185800a },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x1b830154 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x45800c00 },
+	{ _MMIO(0x9888), 0x47800c63 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f8014a5 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x41800045 },
+};
+
+static const struct i915_oa_reg *
+get_l3_3_mux_config(struct drm_i915_private *dev_priv,
+		    int *len)
+{
+	*len = ARRAY_SIZE(mux_config_l3_3);
+	return mux_config_l3_3;
+}
+
+static const struct i915_oa_reg b_counter_config_l3_4[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0xf0800000 },
+	{ _MMIO(0x2770), 0x00100070 },
+	{ _MMIO(0x2774), 0x0000fff1 },
+	{ _MMIO(0x2778), 0x00014002 },
+	{ _MMIO(0x277c), 0x0000c3ff },
+	{ _MMIO(0x2780), 0x00010002 },
+	{ _MMIO(0x2784), 0x0000c7ff },
+	{ _MMIO(0x2788), 0x00004002 },
+	{ _MMIO(0x278c), 0x0000d3ff },
+	{ _MMIO(0x2790), 0x00100700 },
+	{ _MMIO(0x2794), 0x0000ff1f },
+	{ _MMIO(0x2798), 0x00001402 },
+	{ _MMIO(0x279c), 0x0000fc3f },
+	{ _MMIO(0x27a0), 0x00001002 },
+	{ _MMIO(0x27a4), 0x0000fc7f },
+	{ _MMIO(0x27a8), 0x00000402 },
+	{ _MMIO(0x27ac), 0x0000fd3f },
+};
+
+static const struct i915_oa_reg flex_eu_config_l3_4[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_l3_4[] = {
+	{ _MMIO(0x9888), 0x121a0340 },
+	{ _MMIO(0x9888), 0x103f0017 },
+	{ _MMIO(0x9888), 0x123f0020 },
+	{ _MMIO(0x9888), 0x129a0340 },
+	{ _MMIO(0x9888), 0x10bf0017 },
+	{ _MMIO(0x9888), 0x12bf0020 },
+	{ _MMIO(0x9888), 0x041a3400 },
+	{ _MMIO(0x9888), 0x101a0000 },
+	{ _MMIO(0x9888), 0x043b8000 },
+	{ _MMIO(0x9888), 0x0a3e0010 },
+	{ _MMIO(0x9888), 0x003f0200 },
+	{ _MMIO(0x9888), 0x023f0113 },
+	{ _MMIO(0x9888), 0x043f0014 },
+	{ _MMIO(0x9888), 0x02592000 },
+	{ _MMIO(0x9888), 0x005a8000 },
+	{ _MMIO(0x9888), 0x025ac000 },
+	{ _MMIO(0x9888), 0x045a4000 },
+	{ _MMIO(0x9888), 0x0a1c8000 },
+	{ _MMIO(0x9888), 0x001d8000 },
+	{ _MMIO(0x9888), 0x021dc000 },
+	{ _MMIO(0x9888), 0x041d4000 },
+	{ _MMIO(0x9888), 0x0a1e8000 },
+	{ _MMIO(0x9888), 0x0c1fa000 },
+	{ _MMIO(0x9888), 0x0e1f001a },
+	{ _MMIO(0x9888), 0x00394000 },
+	{ _MMIO(0x9888), 0x02395000 },
+	{ _MMIO(0x9888), 0x04391000 },
+	{ _MMIO(0x9888), 0x069a0034 },
+	{ _MMIO(0x9888), 0x109a0000 },
+	{ _MMIO(0x9888), 0x06bb4000 },
+	{ _MMIO(0x9888), 0x0abe0040 },
+	{ _MMIO(0x9888), 0x0cbf0984 },
+	{ _MMIO(0x9888), 0x0ebf0a02 },
+	{ _MMIO(0x9888), 0x02d94000 },
+	{ _MMIO(0x9888), 0x0cdac000 },
+	{ _MMIO(0x9888), 0x0edac000 },
+	{ _MMIO(0x9888), 0x0c9c0400 },
+	{ _MMIO(0x9888), 0x0c9dc000 },
+	{ _MMIO(0x9888), 0x0e9dc000 },
+	{ _MMIO(0x9888), 0x0c9e0400 },
+	{ _MMIO(0x9888), 0x109f02a8 },
+	{ _MMIO(0x9888), 0x0e9f0040 },
+	{ _MMIO(0x9888), 0x0cb95000 },
+	{ _MMIO(0x9888), 0x0eb95000 },
+	{ _MMIO(0x9888), 0x0f88000f },
+	{ _MMIO(0x9888), 0x0d880400 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x258b8009 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x238b2a80 },
+	{ _MMIO(0x9888), 0x198c4000 },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x0d8c4000 },
+	{ _MMIO(0x9888), 0x0d8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x078d2000 },
+	{ _MMIO(0x9888), 0x2185800a },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x1b830154 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x45800800 },
+	{ _MMIO(0x9888), 0x47800842 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f801084 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x41800044 },
+};
+
+static const struct i915_oa_reg *
+get_l3_4_mux_config(struct drm_i915_private *dev_priv,
+		    int *len)
+{
+	*len = ARRAY_SIZE(mux_config_l3_4);
+	return mux_config_l3_4;
+}
+
+static const struct i915_oa_reg b_counter_config_rasterizer_and_pixel_backend[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x30800000 },
+	{ _MMIO(0x2770), 0x00006000 },
+	{ _MMIO(0x2774), 0x0000f3ff },
+	{ _MMIO(0x2778), 0x00001800 },
+	{ _MMIO(0x277c), 0x0000fcff },
+	{ _MMIO(0x2780), 0x00000600 },
+	{ _MMIO(0x2784), 0x0000ff3f },
+	{ _MMIO(0x2788), 0x00000180 },
+	{ _MMIO(0x278c), 0x0000ffcf },
+	{ _MMIO(0x2790), 0x00000060 },
+	{ _MMIO(0x2794), 0x0000fff3 },
+	{ _MMIO(0x2798), 0x00000018 },
+	{ _MMIO(0x279c), 0x0000fffc },
+};
+
+static const struct i915_oa_reg flex_eu_config_rasterizer_and_pixel_backend[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_rasterizer_and_pixel_backend[] = {
+	{ _MMIO(0x9888), 0x143b000e },
+	{ _MMIO(0x9888), 0x043c55c0 },
+	{ _MMIO(0x9888), 0x0a1e0280 },
+	{ _MMIO(0x9888), 0x0c1e0408 },
+	{ _MMIO(0x9888), 0x10390000 },
+	{ _MMIO(0x9888), 0x12397a1f },
+	{ _MMIO(0x9888), 0x14bb000e },
+	{ _MMIO(0x9888), 0x04bc5000 },
+	{ _MMIO(0x9888), 0x0a9e0296 },
+	{ _MMIO(0x9888), 0x0c9e0008 },
+	{ _MMIO(0x9888), 0x10b90000 },
+	{ _MMIO(0x9888), 0x12b97a1f },
+	{ _MMIO(0x9888), 0x063b0042 },
+	{ _MMIO(0x9888), 0x103b0000 },
+	{ _MMIO(0x9888), 0x083c0000 },
+	{ _MMIO(0x9888), 0x0a3e0040 },
+	{ _MMIO(0x9888), 0x043f8000 },
+	{ _MMIO(0x9888), 0x02594000 },
+	{ _MMIO(0x9888), 0x045a8000 },
+	{ _MMIO(0x9888), 0x0c1c0400 },
+	{ _MMIO(0x9888), 0x041d8000 },
+	{ _MMIO(0x9888), 0x081e02c0 },
+	{ _MMIO(0x9888), 0x0e1e0000 },
+	{ _MMIO(0x9888), 0x0c1fa800 },
+	{ _MMIO(0x9888), 0x0e1f0260 },
+	{ _MMIO(0x9888), 0x101f0014 },
+	{ _MMIO(0x9888), 0x003905e0 },
+	{ _MMIO(0x9888), 0x06390bc0 },
+	{ _MMIO(0x9888), 0x02390018 },
+	{ _MMIO(0x9888), 0x04394000 },
+	{ _MMIO(0x9888), 0x04bb0042 },
+	{ _MMIO(0x9888), 0x10bb0000 },
+	{ _MMIO(0x9888), 0x02bc05c0 },
+	{ _MMIO(0x9888), 0x08bc0000 },
+	{ _MMIO(0x9888), 0x0abe0004 },
+	{ _MMIO(0x9888), 0x02bf8000 },
+	{ _MMIO(0x9888), 0x02d91000 },
+	{ _MMIO(0x9888), 0x02da8000 },
+	{ _MMIO(0x9888), 0x089c8000 },
+	{ _MMIO(0x9888), 0x029d8000 },
+	{ _MMIO(0x9888), 0x089e8000 },
+	{ _MMIO(0x9888), 0x0e9e0000 },
+	{ _MMIO(0x9888), 0x0e9fa806 },
+	{ _MMIO(0x9888), 0x109f0142 },
+	{ _MMIO(0x9888), 0x08b90617 },
+	{ _MMIO(0x9888), 0x0ab90be0 },
+	{ _MMIO(0x9888), 0x02b94000 },
+	{ _MMIO(0x9888), 0x0d88f000 },
+	{ _MMIO(0x9888), 0x0f88000c },
+	{ _MMIO(0x9888), 0x07888000 },
+	{ _MMIO(0x9888), 0x09888000 },
+	{ _MMIO(0x9888), 0x018a8000 },
+	{ _MMIO(0x9888), 0x0f8a8000 },
+	{ _MMIO(0x9888), 0x1b8a2800 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x238b52a0 },
+	{ _MMIO(0x9888), 0x258b6a95 },
+	{ _MMIO(0x9888), 0x278b0029 },
+	{ _MMIO(0x9888), 0x178c2000 },
+	{ _MMIO(0x9888), 0x198c1500 },
+	{ _MMIO(0x9888), 0x1b8c0014 },
+	{ _MMIO(0x9888), 0x078c4000 },
+	{ _MMIO(0x9888), 0x098c4000 },
+	{ _MMIO(0x9888), 0x098da000 },
+	{ _MMIO(0x9888), 0x0b8da000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x038d8000 },
+	{ _MMIO(0x9888), 0x058d2000 },
+	{ _MMIO(0x9888), 0x1f85aa80 },
+	{ _MMIO(0x9888), 0x2185aaaa },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x01834000 },
+	{ _MMIO(0x9888), 0x0f834000 },
+	{ _MMIO(0x9888), 0x19835400 },
+	{ _MMIO(0x9888), 0x1b830155 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0184c000 },
+	{ _MMIO(0x9888), 0x0784c000 },
+	{ _MMIO(0x9888), 0x0984c000 },
+	{ _MMIO(0x9888), 0x0b84c000 },
+	{ _MMIO(0x9888), 0x0d84c000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x1180c000 },
+	{ _MMIO(0x9888), 0x1780c000 },
+	{ _MMIO(0x9888), 0x1980c000 },
+	{ _MMIO(0x9888), 0x1b80c000 },
+	{ _MMIO(0x9888), 0x1d80c000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x4d800444 },
+	{ _MMIO(0x9888), 0x3d800000 },
+	{ _MMIO(0x9888), 0x4f804000 },
+	{ _MMIO(0x9888), 0x43801080 },
+	{ _MMIO(0x9888), 0x51800000 },
+	{ _MMIO(0x9888), 0x45800084 },
+	{ _MMIO(0x9888), 0x53800044 },
+	{ _MMIO(0x9888), 0x47801080 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x3f800000 },
+	{ _MMIO(0x9888), 0x41800840 },
+};
+
+static const struct i915_oa_reg *
+get_rasterizer_and_pixel_backend_mux_config(struct drm_i915_private *dev_priv,
+					    int *len)
+{
+	*len = ARRAY_SIZE(mux_config_rasterizer_and_pixel_backend);
+	return mux_config_rasterizer_and_pixel_backend;
+}
+
+static const struct i915_oa_reg b_counter_config_sampler_1[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0x70800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x00800000 },
+	{ _MMIO(0x2770), 0x0000c000 },
+	{ _MMIO(0x2774), 0x0000e7ff },
+	{ _MMIO(0x2778), 0x00003000 },
+	{ _MMIO(0x277c), 0x0000f9ff },
+	{ _MMIO(0x2780), 0x00000c00 },
+	{ _MMIO(0x2784), 0x0000fe7f },
+};
+
+static const struct i915_oa_reg flex_eu_config_sampler_1[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_sampler_1[] = {
+	{ _MMIO(0x9888), 0x18921400 },
+	{ _MMIO(0x9888), 0x149500ab },
+	{ _MMIO(0x9888), 0x18b21400 },
+	{ _MMIO(0x9888), 0x14b500ab },
+	{ _MMIO(0x9888), 0x18d21400 },
+	{ _MMIO(0x9888), 0x14d500ab },
+	{ _MMIO(0x9888), 0x0cdc8000 },
+	{ _MMIO(0x9888), 0x0edc4000 },
+	{ _MMIO(0x9888), 0x02dcc000 },
+	{ _MMIO(0x9888), 0x04dcc000 },
+	{ _MMIO(0x9888), 0x1abd00a0 },
+	{ _MMIO(0x9888), 0x0abd8000 },
+	{ _MMIO(0x9888), 0x0cd88000 },
+	{ _MMIO(0x9888), 0x0ed84000 },
+	{ _MMIO(0x9888), 0x04d88000 },
+	{ _MMIO(0x9888), 0x1adb0050 },
+	{ _MMIO(0x9888), 0x04db8000 },
+	{ _MMIO(0x9888), 0x06db8000 },
+	{ _MMIO(0x9888), 0x08db8000 },
+	{ _MMIO(0x9888), 0x0adb4000 },
+	{ _MMIO(0x9888), 0x109f02a0 },
+	{ _MMIO(0x9888), 0x0c9fa000 },
+	{ _MMIO(0x9888), 0x0e9f00aa },
+	{ _MMIO(0x9888), 0x18b82500 },
+	{ _MMIO(0x9888), 0x02b88000 },
+	{ _MMIO(0x9888), 0x04b84000 },
+	{ _MMIO(0x9888), 0x06b84000 },
+	{ _MMIO(0x9888), 0x08b84000 },
+	{ _MMIO(0x9888), 0x0ab84000 },
+	{ _MMIO(0x9888), 0x0cb88000 },
+	{ _MMIO(0x9888), 0x0cb98000 },
+	{ _MMIO(0x9888), 0x0eb9a000 },
+	{ _MMIO(0x9888), 0x00b98000 },
+	{ _MMIO(0x9888), 0x02b9a000 },
+	{ _MMIO(0x9888), 0x04b9a000 },
+	{ _MMIO(0x9888), 0x06b92000 },
+	{ _MMIO(0x9888), 0x1aba0200 },
+	{ _MMIO(0x9888), 0x02ba8000 },
+	{ _MMIO(0x9888), 0x0cba8000 },
+	{ _MMIO(0x9888), 0x04908000 },
+	{ _MMIO(0x9888), 0x04918000 },
+	{ _MMIO(0x9888), 0x04927300 },
+	{ _MMIO(0x9888), 0x10920000 },
+	{ _MMIO(0x9888), 0x1893000a },
+	{ _MMIO(0x9888), 0x0a934000 },
+	{ _MMIO(0x9888), 0x0a946000 },
+	{ _MMIO(0x9888), 0x0c959000 },
+	{ _MMIO(0x9888), 0x0e950098 },
+	{ _MMIO(0x9888), 0x10950000 },
+	{ _MMIO(0x9888), 0x04b04000 },
+	{ _MMIO(0x9888), 0x04b14000 },
+	{ _MMIO(0x9888), 0x04b20073 },
+	{ _MMIO(0x9888), 0x10b20000 },
+	{ _MMIO(0x9888), 0x04b38000 },
+	{ _MMIO(0x9888), 0x06b38000 },
+	{ _MMIO(0x9888), 0x08b34000 },
+	{ _MMIO(0x9888), 0x04b4c000 },
+	{ _MMIO(0x9888), 0x02b59890 },
+	{ _MMIO(0x9888), 0x10b50000 },
+	{ _MMIO(0x9888), 0x06d04000 },
+	{ _MMIO(0x9888), 0x06d14000 },
+	{ _MMIO(0x9888), 0x06d20073 },
+	{ _MMIO(0x9888), 0x10d20000 },
+	{ _MMIO(0x9888), 0x18d30020 },
+	{ _MMIO(0x9888), 0x02d38000 },
+	{ _MMIO(0x9888), 0x0cd34000 },
+	{ _MMIO(0x9888), 0x0ad48000 },
+	{ _MMIO(0x9888), 0x04d42000 },
+	{ _MMIO(0x9888), 0x0ed59000 },
+	{ _MMIO(0x9888), 0x00d59800 },
+	{ _MMIO(0x9888), 0x10d50000 },
+	{ _MMIO(0x9888), 0x0f88000e },
+	{ _MMIO(0x9888), 0x03888000 },
+	{ _MMIO(0x9888), 0x05888000 },
+	{ _MMIO(0x9888), 0x07888000 },
+	{ _MMIO(0x9888), 0x09888000 },
+	{ _MMIO(0x9888), 0x0b888000 },
+	{ _MMIO(0x9888), 0x0d880400 },
+	{ _MMIO(0x9888), 0x278b002a },
+	{ _MMIO(0x9888), 0x238b5500 },
+	{ _MMIO(0x9888), 0x258b000a },
+	{ _MMIO(0x9888), 0x1b8c0015 },
+	{ _MMIO(0x9888), 0x038c4000 },
+	{ _MMIO(0x9888), 0x058c4000 },
+	{ _MMIO(0x9888), 0x078c4000 },
+	{ _MMIO(0x9888), 0x098c4000 },
+	{ _MMIO(0x9888), 0x0b8c4000 },
+	{ _MMIO(0x9888), 0x0d8c4000 },
+	{ _MMIO(0x9888), 0x0d8d8000 },
+	{ _MMIO(0x9888), 0x0f8da000 },
+	{ _MMIO(0x9888), 0x018d8000 },
+	{ _MMIO(0x9888), 0x038da000 },
+	{ _MMIO(0x9888), 0x058da000 },
+	{ _MMIO(0x9888), 0x078d2000 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x2185000a },
+	{ _MMIO(0x9888), 0x1b830150 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0d848000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x1d808000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47801021 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f800c64 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x41800c02 },
+};
+
+static const struct i915_oa_reg *
+get_sampler_1_mux_config(struct drm_i915_private *dev_priv,
+			 int *len)
+{
+	*len = ARRAY_SIZE(mux_config_sampler_1);
+	return mux_config_sampler_1;
+}
+
+static const struct i915_oa_reg b_counter_config_sampler_2[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0x70800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x00800000 },
+	{ _MMIO(0x2770), 0x0000c000 },
+	{ _MMIO(0x2774), 0x0000e7ff },
+	{ _MMIO(0x2778), 0x00003000 },
+	{ _MMIO(0x277c), 0x0000f9ff },
+	{ _MMIO(0x2780), 0x00000c00 },
+	{ _MMIO(0x2784), 0x0000fe7f },
+};
+
+static const struct i915_oa_reg flex_eu_config_sampler_2[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_sampler_2[] = {
+	{ _MMIO(0x9888), 0x18121400 },
+	{ _MMIO(0x9888), 0x141500ab },
+	{ _MMIO(0x9888), 0x18321400 },
+	{ _MMIO(0x9888), 0x143500ab },
+	{ _MMIO(0x9888), 0x18521400 },
+	{ _MMIO(0x9888), 0x145500ab },
+	{ _MMIO(0x9888), 0x0c5c8000 },
+	{ _MMIO(0x9888), 0x0e5c4000 },
+	{ _MMIO(0x9888), 0x025cc000 },
+	{ _MMIO(0x9888), 0x045cc000 },
+	{ _MMIO(0x9888), 0x1a3d00a0 },
+	{ _MMIO(0x9888), 0x0a3d8000 },
+	{ _MMIO(0x9888), 0x0c588000 },
+	{ _MMIO(0x9888), 0x0e584000 },
+	{ _MMIO(0x9888), 0x04588000 },
+	{ _MMIO(0x9888), 0x1a5b0050 },
+	{ _MMIO(0x9888), 0x045b8000 },
+	{ _MMIO(0x9888), 0x065b8000 },
+	{ _MMIO(0x9888), 0x085b8000 },
+	{ _MMIO(0x9888), 0x0a5b4000 },
+	{ _MMIO(0x9888), 0x101f02a0 },
+	{ _MMIO(0x9888), 0x0c1fa000 },
+	{ _MMIO(0x9888), 0x0e1f00aa },
+	{ _MMIO(0x9888), 0x18382500 },
+	{ _MMIO(0x9888), 0x02388000 },
+	{ _MMIO(0x9888), 0x04384000 },
+	{ _MMIO(0x9888), 0x06384000 },
+	{ _MMIO(0x9888), 0x08384000 },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x0c388000 },
+	{ _MMIO(0x9888), 0x0c398000 },
+	{ _MMIO(0x9888), 0x0e39a000 },
+	{ _MMIO(0x9888), 0x00398000 },
+	{ _MMIO(0x9888), 0x0239a000 },
+	{ _MMIO(0x9888), 0x0439a000 },
+	{ _MMIO(0x9888), 0x06392000 },
+	{ _MMIO(0x9888), 0x1a3a0200 },
+	{ _MMIO(0x9888), 0x023a8000 },
+	{ _MMIO(0x9888), 0x0c3a8000 },
+	{ _MMIO(0x9888), 0x04108000 },
+	{ _MMIO(0x9888), 0x04118000 },
+	{ _MMIO(0x9888), 0x04127300 },
+	{ _MMIO(0x9888), 0x10120000 },
+	{ _MMIO(0x9888), 0x1813000a },
+	{ _MMIO(0x9888), 0x0a134000 },
+	{ _MMIO(0x9888), 0x0a146000 },
+	{ _MMIO(0x9888), 0x0c159000 },
+	{ _MMIO(0x9888), 0x0e150098 },
+	{ _MMIO(0x9888), 0x10150000 },
+	{ _MMIO(0x9888), 0x04304000 },
+	{ _MMIO(0x9888), 0x04314000 },
+	{ _MMIO(0x9888), 0x04320073 },
+	{ _MMIO(0x9888), 0x10320000 },
+	{ _MMIO(0x9888), 0x04338000 },
+	{ _MMIO(0x9888), 0x06338000 },
+	{ _MMIO(0x9888), 0x08334000 },
+	{ _MMIO(0x9888), 0x0434c000 },
+	{ _MMIO(0x9888), 0x02359890 },
+	{ _MMIO(0x9888), 0x10350000 },
+	{ _MMIO(0x9888), 0x06504000 },
+	{ _MMIO(0x9888), 0x06514000 },
+	{ _MMIO(0x9888), 0x06520073 },
+	{ _MMIO(0x9888), 0x10520000 },
+	{ _MMIO(0x9888), 0x18530020 },
+	{ _MMIO(0x9888), 0x02538000 },
+	{ _MMIO(0x9888), 0x0c534000 },
+	{ _MMIO(0x9888), 0x0a548000 },
+	{ _MMIO(0x9888), 0x04542000 },
+	{ _MMIO(0x9888), 0x0e559000 },
+	{ _MMIO(0x9888), 0x00559800 },
+	{ _MMIO(0x9888), 0x10550000 },
+	{ _MMIO(0x9888), 0x1b8aa000 },
+	{ _MMIO(0x9888), 0x1d8a0002 },
+	{ _MMIO(0x9888), 0x038a8000 },
+	{ _MMIO(0x9888), 0x058a8000 },
+	{ _MMIO(0x9888), 0x078a8000 },
+	{ _MMIO(0x9888), 0x098a8000 },
+	{ _MMIO(0x9888), 0x0b8a8000 },
+	{ _MMIO(0x9888), 0x0d8a8000 },
+	{ _MMIO(0x9888), 0x278b0015 },
+	{ _MMIO(0x9888), 0x238b2a80 },
+	{ _MMIO(0x9888), 0x258b0005 },
+	{ _MMIO(0x9888), 0x2385002a },
+	{ _MMIO(0x9888), 0x1f85aa00 },
+	{ _MMIO(0x9888), 0x2185000a },
+	{ _MMIO(0x9888), 0x1b830150 },
+	{ _MMIO(0x9888), 0x03834000 },
+	{ _MMIO(0x9888), 0x05834000 },
+	{ _MMIO(0x9888), 0x07834000 },
+	{ _MMIO(0x9888), 0x09834000 },
+	{ _MMIO(0x9888), 0x0b834000 },
+	{ _MMIO(0x9888), 0x0d834000 },
+	{ _MMIO(0x9888), 0x0d848000 },
+	{ _MMIO(0x9888), 0x0f84c000 },
+	{ _MMIO(0x9888), 0x01848000 },
+	{ _MMIO(0x9888), 0x0384c000 },
+	{ _MMIO(0x9888), 0x0584c000 },
+	{ _MMIO(0x9888), 0x07844000 },
+	{ _MMIO(0x9888), 0x1d808000 },
+	{ _MMIO(0x9888), 0x1f80c000 },
+	{ _MMIO(0x9888), 0x11808000 },
+	{ _MMIO(0x9888), 0x1380c000 },
+	{ _MMIO(0x9888), 0x1580c000 },
+	{ _MMIO(0x9888), 0x17804000 },
+	{ _MMIO(0x9888), 0x53800000 },
+	{ _MMIO(0x9888), 0x47801021 },
+	{ _MMIO(0x9888), 0x21800000 },
+	{ _MMIO(0x9888), 0x31800000 },
+	{ _MMIO(0x9888), 0x4d800000 },
+	{ _MMIO(0x9888), 0x3f800c64 },
+	{ _MMIO(0x9888), 0x4f800000 },
+	{ _MMIO(0x9888), 0x41800c02 },
+};
+
+static const struct i915_oa_reg *
+get_sampler_2_mux_config(struct drm_i915_private *dev_priv,
+			 int *len)
+{
+	*len = ARRAY_SIZE(mux_config_sampler_2);
+	return mux_config_sampler_2;
+}
+
+static const struct i915_oa_reg b_counter_config_tdl_1[] = {
+	{ _MMIO(0x2740), 0x00000000 },
+	{ _MMIO(0x2744), 0x00800000 },
+	{ _MMIO(0x2710), 0x00000000 },
+	{ _MMIO(0x2714), 0xf0800000 },
+	{ _MMIO(0x2720), 0x00000000 },
+	{ _MMIO(0x2724), 0x30800000 },
+	{ _MMIO(0x2770), 0x00000002 },
+	{ _MMIO(0x2774), 0x0000fdff },
+	{ _MMIO(0x2778), 0x00000000 },
+	{ _MMIO(0x277c), 0x0000fe7f },
+	{ _MMIO(0x2780), 0x00000002 },
+	{ _MMIO(0x2784), 0x0000ffbf },
+	{ _MMIO(0x2788), 0x00000000 },
+	{ _MMIO(0x278c), 0x0000ffcf },
+	{ _MMIO(0x2790), 0x00000002 },
+	{ _MMIO(0x2794), 0x0000fff7 },
+	{ _MMIO(0x2798), 0x00000000 },
+	{ _MMIO(0x279c), 0x0000fff9 },
+};
+
+static const struct i915_oa_reg flex_eu_config_tdl_1[] = {
+	{ _MMIO(0xe458), 0x00005004 },
+	{ _MMIO(0xe558), 0x00010003 },
+	{ _MMIO(0xe658), 0x00012011 },
+	{ _MMIO(0xe758), 0x00015014 },
+	{ _MMIO(0xe45c), 0x00051050 },
+	{ _MMIO(0xe55c), 0x00053052 },
+	{ _MMIO(0xe65c), 0x00055054 },
+};
+
+static const struct i915_oa_reg mux_config_tdl_1[] = {
+	{ _MMIO(0x9888), 0x16154d60 },
+	{ _MMIO(0x9888), 0x16352e60 },
+	{ _MMIO(0x9888), 0x16554d60 },
+	{ _MMIO(0x9888), 0x16950000 },
+	{ _MMIO(0x9888), 0x16b50000 },
+	{ _MMIO(0x9888), 0x16d50000 },
+	{ _MMIO(0x9888), 0x005c8000 },
+	{ _MMIO(0x9888), 0x045cc000 },
+	{ _MMIO(0x9888), 0x065c4000 },
+	{ _MMIO(0x9888), 0x083d8000 },
+	{ _MMIO(0x9888), 0x0a3d8000 },
+	{ _MMIO(0x9888), 0x0458c000 },
+	{ _MMIO(0x9888), 0x025b8000 },
+	{ _MMIO(0x9888), 0x085b4000 },
+	{ _MMIO(0x9888), 0x0a5b4000 },
+	{ _MMIO(0x9888), 0x0c5b8000 },
+	{ _MMIO(0x9888), 0x0c1fa000 },
+	{ _MMIO(0x9888), 0x0e1f00aa },
+	{ _MMIO(0x9888), 0x02384000 },
+	{ _MMIO(0x9888), 0x04388000 },
+	{ _MMIO(0x9888), 0x06388000 },
+	{ _MMIO(0x9888), 0x08384000 },
+	{ _MMIO(0x9888), 0x0a384000 },
+	{ _MMIO(0x9888), 0x0c384000 },
+	{ _MMIO(0x9888), 0x00398000 },
+	{ _MMIO(0x9888), 0x0239a000 },
+	{ _MMIO(0x9888), 0x0439a000 },
+	{ _MMIO(0x9888), 0x06392000 },
+	{ _MMIO(0x9888), 0x043a8000 },
+	{ _MMIO(0x9888), 0x063a8000 },
+	{ _MMIO(0x9888), 0x08138000 },
+	{ _MMIO(0x9888), 0x0a138000 },
+	{ _MMIO(0x9888), 0x06143000 },
+	{ _MMIO(0x9888), 0x0415cfc7 },
+	{ _MMIO(0x9888), 0x10150000 },
+	{ _MMIO(0x9888), 0x02338000 },
+	{ _MMIO(0x9888), 0x0c338000 },
+	{ _MMIO(0x9888), 0x04342000 },
+	{ _MMIO(0x9888), 0x06344000 },
+	{ _MMIO(0x9888), 0x0035c700 },
+	{ _MMIO(0x9888), 0x063500cf },
+	{ _MMIO(0x9888), 0x10350000 },
+	{ _MMIO(0x9888), 0x04538000 },
+	{ _MMIO(0x9888), 0x06538000 },
+	{ _MMIO(0x9888), 0x0454c000 },
+	{ _MMIO(0x9888), 0x0255cfc7 },
+	{ _MMIO(0x9888), 0x10550000 },
+	{ _MMIO(0x9888), 0x06dc8000 },
+	{ _MMIO(0x9888), 0x08dc4000 },
+	{ _MMIO(0x9888), 0x0cdcc000 },
+	{ _MMIO(0x9888), 0x0edcc000 },
+	{ _MMIO(0x9888), 0x1abd00a8 },
+	{ _MMIO(0x9888), 0x0cd8c000 },
+	{ _MMIO(0x9888), 0x0ed84000 },
+	{ _MMIO(0x9888), 0x0edb8000 },
+	{ _MMIO(0x9888), 0x18db0800 },
+	{ _MMIO(0x9888), 0x1adb0254 },
[… more lines in the raw message]

Lines of this file are not shown. View the raw message.

drivers/gpu/drm/i915/i915_oa_bxt.c +2313 -10

2,323 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_chv.c +2515 -10

2,525 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_hsw.c +53 -5

58 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt2.c +3116 -41

3,157 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt3.c +2661 -41

2,702 lines not on this page · raw message

drivers/gpu/drm/i915/i915_oa_sklgt4.c +2704 -41

2,745 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 6 lines
2.12.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Part of the diff is not shown here. View the raw message.

Reviewed-by on v2 5/5
#14 lore raw

Re: [PATCH v2 5/5] drm/i915: Add more OA configs for BDW, CHV, SKL + BXT

Quoted 11 lines · Robert Bragg

These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml

---

<SNIP>

Quoted 4 lines

int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv) {

drivers/gpu/drm/i915/i915_oa_bdw.c +0 -2
-	dev_priv->perf.oa.mux_regs = NULL;
-	dev_priv->perf.oa.mux_regs_len = 0;

drivers/gpu/drm/i915/i915_oa_bdw.c get_render_basic_mux_config(struct drm_i915_private *dev_priv, · line 3877

Quoted 10 lines
drivers/gpu/drm/i915/i915_oa_bdw.c +6 -4
-	dev_priv->perf.oa.b_counter_regs = NULL;
-	dev_priv->perf.oa.b_counter_regs_len = 0;
-	dev_priv->perf.oa.flex_regs = NULL;
-	dev_priv->perf.oa.flex_regs_len = 0;
+	dev_priv->perf.oa.mux_regs = NULL;
+	dev_priv->perf.oa.mux_regs_len = 0;
+	dev_priv->perf.oa.b_counter_regs = NULL;
+	dev_priv->perf.oa.b_counter_regs_len = 0;
+	dev_priv->perf.oa.flex_regs = NULL;
+	dev_priv->perf.oa.flex_regs_len = 0;

What changed? I can't tell from the diff...

Otherwise assuming you re-spin with the DRM_DEBUG changes:

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#15 lore raw

Re: [PATCH v2 5/5] drm/i915: Add more OA configs for BDW, CHV, SKL + BXT

Quoted 29 lines · Matthew Auld
Quoted 11 lines · Robert Bragg

These are auto generated from an XML description of metric sets, currently maintained in gputop, ref:

https://github.com/rib/gputop

Quoted 2 lines

gputop-data/oa-*.xml
scripts/i915-perf-kernelgen.py

 $ make -C gputop-data -f Makefile.xml

---

<SNIP>

Quoted 4 lines

int i915_oa_select_metric_set_bdw(struct drm_i915_private *dev_priv) {

drivers/gpu/drm/i915/i915_oa_bdw.c +0 -2
-     dev_priv->perf.oa.mux_regs = NULL;
-     dev_priv->perf.oa.mux_regs_len = 0;

drivers/gpu/drm/i915/i915_oa_bdw.c get_render_basic_mux_config(struct drm_i915_private *dev_priv, · line 3877

Quoted 10 lines
drivers/gpu/drm/i915/i915_oa_bdw.c +6 -4
-     dev_priv->perf.oa.b_counter_regs = NULL;
-     dev_priv->perf.oa.b_counter_regs_len = 0;
-     dev_priv->perf.oa.flex_regs = NULL;
-     dev_priv->perf.oa.flex_regs_len = 0;
+     dev_priv->perf.oa.mux_regs = NULL;
+     dev_priv->perf.oa.mux_regs_len = 0;
+     dev_priv->perf.oa.b_counter_regs = NULL;
+     dev_priv->perf.oa.b_counter_regs_len = 0;
+     dev_priv->perf.oa.flex_regs = NULL;
+     dev_priv->perf.oa.flex_regs_len = 0;
Quoted 2 lines

What changed? I can't tell from the diff...

I don't think anything changed in those lines, it's just that the diff uses the start of the function for context and then has to delete these to add the full replacement for the function body which included substantial changes to add the cases for the additional configs.

Quoted 4 lines

Otherwise assuming you re-spin with the DRM_DEBUG changes:

thanks

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#16 lore raw

✗ Fi.CI.BAT: failure for Enable OA unit for Gen 8 and 9 in i915 perf (rev3)

In reply to Robert Bragg
== Series Details ==

== Series Details ==

Series: Enable OA unit for Gen 8 and 9 in i915 perf (rev3)
URL : https://patchwork.freedesktop.org/series/20084/
State : failure

== Summary ==

Series 20084v3 Enable OA unit for Gen 8 and 9 in i915 perf
https://patchwork.freedesktop.org/api/1.0/series/20084/revisions/3/mbox/

Test gem_exec_fence:

        Subgroup await-hang-default:
                pass       -> INCOMPLETE (fi-hsw-4770)
fi-bdw-5557u     total:278  pass:267  dwarn:0   dfail:0   fail:0   skip:11  time: 470s
fi-bdw-gvtdvm    total:278  pass:256  dwarn:8   dfail:0   fail:0   skip:14  time: 467s
fi-bsw-n3050     total:278  pass:239  dwarn:0   dfail:0   fail:0   skip:39  time: 584s
fi-bxt-j4205     total:278  pass:259  dwarn:0   dfail:0   fail:0   skip:19  time: 534s
fi-bxt-t5700     total:278  pass:258  dwarn:0   dfail:0   fail:0   skip:20  time: 568s
fi-byt-j1900     total:278  pass:251  dwarn:0   dfail:0   fail:0   skip:27  time: 509s
fi-hsw-4770      total:48   pass:41   dwarn:0   dfail:0   fail:0   skip:6   time: 0s
fi-hsw-4770r     total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16  time: 432s
fi-ilk-650       total:278  pass:228  dwarn:0   dfail:0   fail:0   skip:50  time: 442s
fi-ivb-3520m     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 510s
fi-ivb-3770      total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 495s
fi-kbl-7500u     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 488s
fi-skl-6260u     total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time: 481s
fi-skl-6700hq    total:278  pass:261  dwarn:0   dfail:0   fail:0   skip:17  time: 606s
fi-skl-6700k     total:278  pass:256  dwarn:4   dfail:0   fail:0   skip:18  time: 492s
fi-skl-6770hq    total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time: 516s
fi-skl-gvtdvm    total:278  pass:265  dwarn:0   dfail:0   fail:0   skip:13  time: 461s
fi-snb-2520m     total:278  pass:250  dwarn:0   dfail:0   fail:0   skip:28  time: 546s
fi-snb-2600      total:278  pass:249  dwarn:0   dfail:0   fail:0   skip:29  time: 426s

fi-byt-n2820 failed to collect. IGT log at Patchwork_4288/fi-byt-n2820/igt.log

fd27e1e4c9b5dc11966b4953432bd6e0510da308 drm-tip: 2017y-03m-24d-08h-39m-20s UTC integration manifest 913d37d drm/i915: Add 'render basic' Gen8+ OA unit configs
9500fa8 drm/i915: expose _SUBSLICE_MASK GETPARM
fd8966b drm/i915: expose _SLICE_MASK GETPARM

== Logs ==

For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_4288/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx