Skip to contentSkip to conversation

DVFS support for Venus

PATCH Superseded Rajendra Nayak 9 participants 29 messages Jul 23, 2020 – Jul 29, 2020

Subsystems (7)

v4 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v4 cover DVFS support for Venus

v4: Moved code from probe/remove/runtime_suspend into different pm_ops callbacks

v3: Renamed the optional power domain as cx

v2: Fixed up the labels of OPP nodes in patch 4
Included the bindings update patch as part of this series, a resend of https://lore.kernel.org/patchwork/patch/1241077/

These patches add DVFS support for Venus

Patch 1 will need to be picked by Rob.
Patch 2 and 3 will need to be picked by Stan,
Patch 4 and 5 should land via the qcom tree.

Rajendra Nayak (5):
dt-bindings: media: venus: Add an optional power domain for perf
voting
media: venus: core: Fix error handling in probe
media: venus: core: Add support for opp tables/perf voting
arm64: dts: sdm845: Add OPP tables and power-domains for venus
arm64: dts: sc7180: Add OPP tables and power-domains for venus

 .../bindings/media/qcom,sc7180-venus.yaml          |  6 +-
 .../bindings/media/qcom,sdm845-venus-v2.yaml       |  6 +-
 arch/arm64/boot/dts/qcom/sc7180.dtsi               | 35 +++++++-
 arch/arm64/boot/dts/qcom/sdm845.dtsi               | 40 +++++++++-
 drivers/media/platform/qcom/venus/core.c           | 17 ++--
 drivers/media/platform/qcom/venus/core.h           |  5 ++
 drivers/media/platform/qcom/venus/pm_helpers.c     | 92 ++++++++++++++++++++--
 7 files changed, 183 insertions(+), 18 deletions(-)
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#1 lore raw

PATCH v4 1/5 dt-bindings: media: venus: Add an optional power domain for perf voting

Venus needs to vote for the performance state of a power domain (cx) to be able to support DVFS. This 'cx' power domain is controlled by rpmh and is a common power domain (scalable) not specific to venus alone. This is optional in the sense that, leaving this power domain out does not really impact the functionality but just makes the platform a little less power efficient.

---
 Documentation/devicetree/bindings/media/qcom,sc7180-venus.yaml    | 6 +++++-
 Documentation/devicetree/bindings/media/qcom,sdm845-venus-v2.yaml | 6 +++++-
 2 files changed, 10 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/media/qcom,sc7180-venus.yaml +5 -1
index 55f2d67..04e303b 100644
@@ -25,12 +25,16 @@ properties:
     maxItems: 1
 
   power-domains:
-    maxItems: 2
+    minItems: 2
+    maxItems: 3
 
   power-domain-names:
+    minItems: 2
+    maxItems: 3
     items:
       - const: venus
       - const: vcodec0
+      - const: cx
 
   clocks:
     maxItems: 5
Documentation/devicetree/bindings/media/qcom,sdm845-venus-v2.yaml +5 -1
index 157dff8..90013d4 100644
@@ -25,13 +25,17 @@ properties:
     maxItems: 1
 
   power-domains:
-    maxItems: 3
+    minItems: 3
+    maxItems: 4
 
   power-domain-names:
+    minItems: 3
+    maxItems: 4
     items:
       - const: venus
       - const: vcodec0
       - const: vcodec1
+      - const: cx
 
   clocks:
     maxItems: 7
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#5 lore raw

PATCH v4 3/5 media: venus: core: Add support for opp tables/perf voting

In reply to Rajendra Nayak

Add support to add OPP tables and perf voting on the OPP powerdomain. This is needed so venus votes on the corresponding performance state for the OPP powerdomain along with setting the core clock rate.

---
 drivers/media/platform/qcom/venus/core.c       |  2 +
 drivers/media/platform/qcom/venus/core.h       |  5 ++
 drivers/media/platform/qcom/venus/pm_helpers.c | 92 ++++++++++++++++++++++++--
 3 files changed, 92 insertions(+), 7 deletions(-)
drivers/media/platform/qcom/venus/core.c +2 -0
index bfcaba3..a3e98a5 100644
@@ -525,6 +525,7 @@ static const struct venus_resources sdm845_res_v2 = {
 	.vcodec_clks_num = 2,
 	.vcodec_pmdomains = { "venus", "vcodec0", "vcodec1" },
 	.vcodec_pmdomains_num = 3,
+	.opp_pmdomain = (const char *[]) { "cx", NULL },
 	.vcodec_num = 2,
 	.max_load = 3110400,	/* 4096x2160@90 */
 	.hfi_version = HFI_VERSION_4XX,
@@ -570,6 +571,7 @@ static const struct venus_resources sc7180_res = {
 	.vcodec_clks_num = 2,
 	.vcodec_pmdomains = { "venus", "vcodec0" },
 	.vcodec_pmdomains_num = 2,
+	.opp_pmdomain = (const char *[]) { "cx", NULL },
 	.vcodec_num = 1,
 	.hfi_version = HFI_VERSION_4XX,
 	.vmem_id = VIDC_RESOURCE_NONE,
drivers/media/platform/qcom/venus/core.h +5 -0
index 7118612..b0cc544 100644
@@ -62,6 +62,7 @@ struct venus_resources {
 	unsigned int vcodec_clks_num;
 	const char * const vcodec_pmdomains[VIDC_PMDOMAINS_NUM_MAX];
 	unsigned int vcodec_pmdomains_num;
+	const char **opp_pmdomain;
 	unsigned int vcodec_num;
 	enum hfi_version hfi_version;
 	u32 max_load;
@@ -145,8 +146,12 @@ struct venus_core {
 	struct clk *vcodec1_clks[VIDC_VCODEC_CLKS_NUM_MAX];
 	struct icc_path *video_path;
 	struct icc_path *cpucfg_path;
+	struct opp_table *opp_table;
+	bool has_opp_table;
 	struct device_link *pd_dl_venus;
 	struct device *pmdomains[VIDC_PMDOMAINS_NUM_MAX];
+	struct device_link *opp_dl_venus;
+	struct device *opp_pmdomain;
 	struct video_device *vdev_dec;
 	struct video_device *vdev_enc;
 	struct v4l2_device v4l2_dev;
drivers/media/platform/qcom/venus/pm_helpers.c +85 -7

163 lines not on this page · raw message

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#3 lore raw

PATCH v4 2/5 media: venus: core: Fix error handling in probe

In reply to Rajendra Nayak

Post a successful pm_ops->core_get, an error in probe should exit by doing a pm_ops->core_put which seems to be missing. So fix it.

---
 drivers/media/platform/qcom/venus/core.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
drivers/media/platform/qcom/venus/core.c +10 -5
index 203c653..bfcaba3 100644
@@ -224,13 +224,15 @@ static int venus_probe(struct platform_device *pdev)
 
 	ret = dma_set_mask_and_coherent(dev, core->res->dma_mask);
 	if (ret)
-		return ret;
+		goto err_core_put;
 
 	if (!dev->dma_parms) {
 		dev->dma_parms = devm_kzalloc(dev, sizeof(*dev->dma_parms),
 					      GFP_KERNEL);
-		if (!dev->dma_parms)
-			return -ENOMEM;
+		if (!dev->dma_parms) {
+			ret = -ENOMEM;
+			goto err_core_put;
+		}
 	}
 	dma_set_max_seg_size(dev, DMA_BIT_MASK(32));
 
@@ -242,11 +244,11 @@ static int venus_probe(struct platform_device *pdev)
 					IRQF_TRIGGER_HIGH | IRQF_ONESHOT,
 					"venus", core);
 	if (ret)
-		return ret;
+		goto err_core_put;
 
 	ret = hfi_create(core, &venus_core_ops);
 	if (ret)
-		return ret;
+		goto err_core_put;
 
 	pm_runtime_enable(dev);
 
@@ -302,6 +304,9 @@ static int venus_probe(struct platform_device *pdev)
 	pm_runtime_set_suspended(dev);
 	pm_runtime_disable(dev);
 	hfi_destroy(core);
+err_core_put:
+	if (core->pm_ops->core_put)
+		core->pm_ops->core_put(dev);
 	return ret;
 }
 
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#7 lore raw

PATCH v4 4/5 arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
 arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +38 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
 			interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
 			power-domains = <&videocc VENUS_GDSC>,
 					<&videocc VCODEC0_GDSC>,
-					<&videocc VCODEC1_GDSC>;
-			power-domain-names = "venus", "vcodec0", "vcodec1";
+					<&videocc VCODEC1_GDSC>,
+					<&rpmhpd SDM845_CX>;
+			power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+			operating-points-v2 = <&venus_opp_table>;
 			clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
 				 <&videocc VIDEO_CC_VENUS_AHB_CLK>,
 				 <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
 			video-core1 {
 				compatible = "venus-encoder";
 			};
+
+			venus_opp_table: venus-opp-table {
+				compatible = "operating-points-v2";
+
+				opp-100000000 {
+					opp-hz = /bits/ 64 <100000000>;
+					required-opps = <&rpmhpd_opp_min_svs>;
+				};
+
+				opp-200000000 {
+					opp-hz = /bits/ 64 <200000000>;
+					required-opps = <&rpmhpd_opp_low_svs>;
+				};
+
+				opp-320000000 {
+					opp-hz = /bits/ 64 <320000000>;
+					required-opps = <&rpmhpd_opp_svs>;
+				};
+
+				opp-380000000 {
+					opp-hz = /bits/ 64 <380000000>;
+					required-opps = <&rpmhpd_opp_svs_l1>;
+				};
+
+				opp-444000000 {
+					opp-hz = /bits/ 64 <444000000>;
+					required-opps = <&rpmhpd_opp_nom>;
+				};
+
+				opp-533000000 {
+					opp-hz = /bits/ 64 <533000000>;
+					required-opps = <&rpmhpd_opp_turbo>;
+				};
+			};
 		};
 
 		videocc: clock-controller@ab00000 {
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#28 lore raw

PATCH v4 5/5 arm64: dts: sc7180: Add OPP tables and power-domains for venus

In reply to Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain

---
 arch/arm64/boot/dts/qcom/sc7180.dtsi | 35 +++++++++++++++++++++++++++++++++--
 1 file changed, 33 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sc7180.dtsi +33 -2
index 16df08d..cb6137d 100644
@@ -2664,8 +2664,10 @@
 			reg = <0 0x0aa00000 0 0xff000>;
 			interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
 			power-domains = <&videocc VENUS_GDSC>,
-					<&videocc VCODEC0_GDSC>;
-			power-domain-names = "venus", "vcodec0";
+					<&videocc VCODEC0_GDSC>,
+					<&rpmhpd SC7180_CX>;
+			power-domain-names = "venus", "vcodec0", "cx";
+			operating-points-v2 = <&venus_opp_table>;
 			clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
 				 <&videocc VIDEO_CC_VENUS_AHB_CLK>,
 				 <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -2686,6 +2688,35 @@
 			video-encoder {
 				compatible = "venus-encoder";
 			};
+
+			venus_opp_table: venus-opp-table {
+				compatible = "operating-points-v2";
+
+				opp-150000000 {
+					opp-hz = /bits/ 64 <150000000>;
+					required-opps = <&rpmhpd_opp_low_svs>;
+				};
+
+				opp-270000000 {
+					opp-hz = /bits/ 64 <270000000>;
+					required-opps = <&rpmhpd_opp_svs>;
+				};
+
+				opp-340000000 {
+					opp-hz = /bits/ 64 <340000000>;
+					required-opps = <&rpmhpd_opp_svs_l1>;
+				};
+
+				opp-434000000 {
+					opp-hz = /bits/ 64 <434000000>;
+					required-opps = <&rpmhpd_opp_nom>;
+				};
+
+				opp-500000000 {
+					opp-hz = /bits/ 64 <500000000>;
+					required-opps = <&rpmhpd_opp_turbo>;
+				};
+			};
 		};
 
 		videocc: clock-controller@ab00000 {
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
maintainer Reviewed-by on v4 1/5
#2 lore raw

Re: [PATCH v4 1/5] dt-bindings: media: venus: Add an optional power domain for perf voting

In reply to Rajendra Nayak
Quoted 7 lines · Rajendra Nayak

Venus needs to vote for the performance state of a power domain (cx) to be able to support DVFS. This 'cx' power domain is controlled by rpmh and is a common power domain (scalable) not specific to venus alone. This is optional in the sense that, leaving this power domain out does not really impact the functionality but just makes the platform a little less power efficient.

Quoted 6 lines
---
 Documentation/devicetree/bindings/media/qcom,sc7180-venus.yaml    | 6 +++++-
 Documentation/devicetree/bindings/media/qcom,sdm845-venus-v2.yaml | 6 +++++-
 2 files changed, 10 insertions(+), 2 deletions(-)
#8 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Rajendra Nayak

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Quoted 67 lines · Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
 arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +38 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
 			interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
 			power-domains = <&videocc VENUS_GDSC>,
 					<&videocc VCODEC0_GDSC>,
-					<&videocc VCODEC1_GDSC>;
-			power-domain-names = "venus", "vcodec0", "vcodec1";
+					<&videocc VCODEC1_GDSC>,
+					<&rpmhpd SDM845_CX>;
+			power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+			operating-points-v2 = <&venus_opp_table>;
 			clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
 				 <&videocc VIDEO_CC_VENUS_AHB_CLK>,
 				 <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
 			video-core1 {
 				compatible = "venus-encoder";
 			};
+
+			venus_opp_table: venus-opp-table {
+				compatible = "operating-points-v2";
+
+				opp-100000000 {
+					opp-hz = /bits/ 64 <100000000>;
+					required-opps = <&rpmhpd_opp_min_svs>;
+				};
+
+				opp-200000000 {
+					opp-hz = /bits/ 64 <200000000>;
+					required-opps = <&rpmhpd_opp_low_svs>;
+				};
+
+				opp-320000000 {
+					opp-hz = /bits/ 64 <320000000>;
+					required-opps = <&rpmhpd_opp_svs>;
+				};
+
+				opp-380000000 {
+					opp-hz = /bits/ 64 <380000000>;
+					required-opps = <&rpmhpd_opp_svs_l1>;
+				};
+
+				opp-444000000 {
+					opp-hz = /bits/ 64 <444000000>;
+					required-opps = <&rpmhpd_opp_nom>;
+				};
+
+				opp-533000000 {
+					opp-hz = /bits/ 64 <533000000>;
+					required-opps = <&rpmhpd_opp_turbo>;
+				};
+			};
 		};
 
 		videocc: clock-controller@ab00000 {
Signature · 2 lines
regards,
Stan
#9 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Hey Stan,

Quoted 9 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

I was hoping to be able to reproduce this on the 845 mtp (I don't have a db845), but I can;t seem to get venus working on mainline [1] (neither with linux-next, nor with linaro integration)
Do you know if I might be missing some fix?

Quoted 4 lines

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Sure, I'll send a patch for that and see what the rpmh owners have to say.

[1]

[    1.632147] qcom-venus aa00000.video-codec: Adding to iommu group 2
[    1.638920] qcom-venus aa00000.video-codec: non legacy binding
[    1.648313] ------------[ cut here ]------------
[    1.652976] video_cc_venus_ctl_axi_clk status stuck at 'off'
[    1.653068] WARNING: CPU: 7 PID: 1 at drivers/clk/qcom/clk-branch.c:92 clk_b8
[    1.667977] Modules linked in:
[    1.671076] CPU: 7 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc6-00254-gc43551
[    1.678704] Hardware name: Qualcomm Technologies, Inc. SDM845 MTP (DT)
[    1.685294] pstate: 60c00085 (nZCv daIf +PAN +UAO BTYPE=--)
[    1.690911] pc : clk_branch_toggle+0x14c/0x168
[    1.695397] lr : clk_branch_toggle+0x14c/0x168
[    1.699881] sp : ffff80001005b900
[    1.703232] x29: ffff80001005b900 x28: ffffb58586ac2f38
[    1.708594] x27: ffff0000f86c6d48 x26: ffffb58586f09000
[    1.713953] x25: ffffb585867c0bd8 x24: 0000000000000000
[    1.719312] x23: ffffb5858702df28 x22: ffffb58585a3c6a8
[    1.724672] x21: 0000000000000001 x20: ffffb58586f09000
[    1.730031] x19: 0000000000000000 x18: ffffb58586f09948
[    1.735390] x17: 0000000000000001 x16: 0000000000000019
[    1.740749] x15: ffff80009005b5a7 x14: 0000000000000006
[    1.746107] x13: ffff80001005b5b5 x12: ffffb58586f21d68
[    1.751466] x11: 0000000000000000 x10: 0000000005f5e0ff
[    1.756825] x9 : ffff80001005b900 x8 : 2766666f27207461
[    1.762184] x7 : 206b637574732073 x6 : ffffb58587141848
[    1.767543] x5 : 0000000000000000 x4 : 0000000000000000
[    1.772902] x3 : 00000000ffffffff x2 : ffff4a7b76cd8000
[    1.778261] x1 : ad405f90446fcf00 x0 : 0000000000000000
[    1.783624] Call trace:
[    1.786103]  clk_branch_toggle+0x14c/0x168
[    1.790241]  clk_branch2_enable+0x18/0x20
[    1.794306]  clk_core_enable+0x60/0xa8
[    1.798090]  clk_core_enable_lock+0x20/0x40
[    1.802316]  clk_enable+0x14/0x28
[    1.805682]  core_clks_enable+0x94/0xd8
[    1.809562]  core_power_v4+0x48/0x50
[    1.813178]  venus_runtime_resume+0x24/0x40
[    1.817417]  pm_generic_runtime_resume+0x28/0x40
[    1.822079]  __rpm_callback+0xa0/0x138
[    1.825861]  rpm_callback+0x24/0x98
[    1.829390]  rpm_resume+0x32c/0x490
[    1.832917]  __pm_runtime_resume+0x38/0x88
[    1.837051]  venus_probe+0x1f0/0x34c
[    1.840667]  platform_drv_probe+0x4c/0xa8
[    1.844727]  really_probe+0x100/0x388
[    1.848428]  driver_probe_device+0x54/0xb8
[    1.852563]  device_driver_attach+0x6c/0x78
[    1.856790]  __driver_attach+0xb0/0xf0
[    1.860576]  bus_for_each_dev+0x68/0xc8
[    1.864456]  driver_attach+0x20/0x28
[    1.868064]  bus_add_driver+0x148/0x200
[    1.871941]  driver_register+0x60/0x110
[    1.875819]  __platform_driver_register+0x44/0x50
[    1.880576]  qcom_venus_driver_init+0x18/0x20
[    1.884990]  do_one_initcall+0x58/0x1a0
[    1.888878]  kernel_init_freeable+0x1fc/0x28c
[    1.893292]  kernel_init+0x10/0x108
[    1.896821]  ret_from_fork+0x10/0x1c
[    1.900441] ---[ end trace f12a7e5e182f3e4e ]---
[    1.906415] qcom-venus: probe of aa00000.video-codec failed with error -16
Quoted 70 lines · Rajendra Nayak
Quoted 67 lines

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
  arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--
  1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +38 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
  			interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
  			power-domains = <&videocc VENUS_GDSC>,
  					<&videocc VCODEC0_GDSC>,
-					<&videocc VCODEC1_GDSC>;
-			power-domain-names = "venus", "vcodec0", "vcodec1";
+					<&videocc VCODEC1_GDSC>,
+					<&rpmhpd SDM845_CX>;
+			power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+			operating-points-v2 = <&venus_opp_table>;
  			clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
  				 <&videocc VIDEO_CC_VENUS_AHB_CLK>,
  				 <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
  			video-core1 {
  				compatible = "venus-encoder";
  			};
+
+			venus_opp_table: venus-opp-table {
+				compatible = "operating-points-v2";
+
+				opp-100000000 {
+					opp-hz = /bits/ 64 <100000000>;
+					required-opps = <&rpmhpd_opp_min_svs>;
+				};
+
+				opp-200000000 {
+					opp-hz = /bits/ 64 <200000000>;
+					required-opps = <&rpmhpd_opp_low_svs>;
+				};
+
+				opp-320000000 {
+					opp-hz = /bits/ 64 <320000000>;
+					required-opps = <&rpmhpd_opp_svs>;
+				};
+
+				opp-380000000 {
+					opp-hz = /bits/ 64 <380000000>;
+					required-opps = <&rpmhpd_opp_svs_l1>;
+				};
+
+				opp-444000000 {
+					opp-hz = /bits/ 64 <444000000>;
+					required-opps = <&rpmhpd_opp_nom>;
+				};
+
+				opp-533000000 {
+					opp-hz = /bits/ 64 <533000000>;
+					required-opps = <&rpmhpd_opp_turbo>;
+				};
+			};
  		};
 
  		videocc: clock-controller@ab00000 {
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#11 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Stanimir Varbanov

Hi Maulik/Lina,

Quoted 5 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

Quoted 8 lines

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

thanks,
Rajendra

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#10 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Rajendra Nayak

Hi,

Quoted 8 lines · Rajendra Nayak

Hey Stan,

Quoted 5 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

Quoted 8 lines
Quoted 4 lines

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

I was hoping to be able to reproduce this on the 845 mtp (I don't have a db845), but I can;t seem to get venus working on mainline [1] (neither with linux-next, nor with linaro integration)

The driver should at least load without errors on -next and linaro-integration for db845. As to MTP and haven't checked because I don't have such board with me.

I will try to debug dev_pm_opp_set_rate -ERANGE error.

Quoted 7 lines

Do you know if I might be missing some fix?

Quoted 4 lines

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Quoted 8 lines

Sure, I'll send a patch for that and see what the rpmh owners have to say.

[1]

[    1.632147] qcom-venus aa00000.video-codec: Adding to iommu group 2 [    1.638920] qcom-venus aa00000.video-codec: non legacy binding [    1.648313] ------------[ cut here ]------------
[    1.652976] video_cc_venus_ctl_axi_clk status stuck at 'off'

I guess this means that venus_gdsc is not powered on.

Quoted 132 lines

[    1.653068] WARNING: CPU: 7 PID: 1 at drivers/clk/qcom/clk-branch.c:92 clk_b8
[    1.667977] Modules linked in:
[    1.671076] CPU: 7 PID: 1 Comm: swapper/0 Not tainted 5.8.0-rc6-00254-gc43551
[    1.678704] Hardware name: Qualcomm Technologies, Inc. SDM845 MTP (DT) [    1.685294] pstate: 60c00085 (nZCv daIf +PAN +UAO BTYPE=--) [    1.690911] pc : clk_branch_toggle+0x14c/0x168
[    1.695397] lr : clk_branch_toggle+0x14c/0x168
[    1.699881] sp : ffff80001005b900
[    1.703232] x29: ffff80001005b900 x28: ffffb58586ac2f38
[    1.708594] x27: ffff0000f86c6d48 x26: ffffb58586f09000
[    1.713953] x25: ffffb585867c0bd8 x24: 0000000000000000
[    1.719312] x23: ffffb5858702df28 x22: ffffb58585a3c6a8
[    1.724672] x21: 0000000000000001 x20: ffffb58586f09000
[    1.730031] x19: 0000000000000000 x18: ffffb58586f09948
[    1.735390] x17: 0000000000000001 x16: 0000000000000019
[    1.740749] x15: ffff80009005b5a7 x14: 0000000000000006
[    1.746107] x13: ffff80001005b5b5 x12: ffffb58586f21d68
[    1.751466] x11: 0000000000000000 x10: 0000000005f5e0ff
[    1.756825] x9 : ffff80001005b900 x8 : 2766666f27207461
[    1.762184] x7 : 206b637574732073 x6 : ffffb58587141848
[    1.767543] x5 : 0000000000000000 x4 : 0000000000000000
[    1.772902] x3 : 00000000ffffffff x2 : ffff4a7b76cd8000
[    1.778261] x1 : ad405f90446fcf00 x0 : 0000000000000000
[    1.783624] Call trace:
[    1.786103]  clk_branch_toggle+0x14c/0x168
[    1.790241]  clk_branch2_enable+0x18/0x20
[    1.794306]  clk_core_enable+0x60/0xa8
[    1.798090]  clk_core_enable_lock+0x20/0x40
[    1.802316]  clk_enable+0x14/0x28
[    1.805682]  core_clks_enable+0x94/0xd8
[    1.809562]  core_power_v4+0x48/0x50
[    1.813178]  venus_runtime_resume+0x24/0x40
[    1.817417]  pm_generic_runtime_resume+0x28/0x40
[    1.822079]  __rpm_callback+0xa0/0x138
[    1.825861]  rpm_callback+0x24/0x98
[    1.829390]  rpm_resume+0x32c/0x490
[    1.832917]  __pm_runtime_resume+0x38/0x88
[    1.837051]  venus_probe+0x1f0/0x34c
[    1.840667]  platform_drv_probe+0x4c/0xa8
[    1.844727]  really_probe+0x100/0x388
[    1.848428]  driver_probe_device+0x54/0xb8
[    1.852563]  device_driver_attach+0x6c/0x78
[    1.856790]  __driver_attach+0xb0/0xf0
[    1.860576]  bus_for_each_dev+0x68/0xc8
[    1.864456]  driver_attach+0x20/0x28
[    1.868064]  bus_add_driver+0x148/0x200
[    1.871941]  driver_register+0x60/0x110
[    1.875819]  __platform_driver_register+0x44/0x50
[    1.880576]  qcom_venus_driver_init+0x18/0x20
[    1.884990]  do_one_initcall+0x58/0x1a0
[    1.888878]  kernel_init_freeable+0x1fc/0x28c
[    1.893292]  kernel_init+0x10/0x108
[    1.896821]  ret_from_fork+0x10/0x1c
[    1.900441] ---[ end trace f12a7e5e182f3e4e ]---
[    1.906415] qcom-venus: probe of aa00000.video-codec failed with error -16

Quoted 72 lines · Rajendra Nayak
Quoted 69 lines

Add the OPP tables in order to be able to vote on the performance state of
a power-domain.

---
  arch/arm64/boot/dts/qcom/sdm845.dtsi | 40
++++++++++++++++++++++++++++++++++--

  1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +38 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
               interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
               power-domains = <&videocc VENUS_GDSC>,
                       <&videocc VCODEC0_GDSC>,
-                    <&videocc VCODEC1_GDSC>;
-            power-domain-names = "venus", "vcodec0", "vcodec1";
+                    <&videocc VCODEC1_GDSC>,
+                    <&rpmhpd SDM845_CX>;
+            power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+            operating-points-v2 = <&venus_opp_table>;
               clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
                    <&videocc VIDEO_CC_VENUS_AHB_CLK>,
                    <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
               video-core1 {
                   compatible = "venus-encoder";
               };
+
+            venus_opp_table: venus-opp-table {
+                compatible = "operating-points-v2";
+
+                opp-100000000 {
+                    opp-hz = /bits/ 64 <100000000>;
+                    required-opps = <&rpmhpd_opp_min_svs>;
+                };
+
+                opp-200000000 {
+                    opp-hz = /bits/ 64 <200000000>;
+                    required-opps = <&rpmhpd_opp_low_svs>;
+                };
+
+                opp-320000000 {
+                    opp-hz = /bits/ 64 <320000000>;
+                    required-opps = <&rpmhpd_opp_svs>;
+                };
+
+                opp-380000000 {
+                    opp-hz = /bits/ 64 <380000000>;
+                    required-opps = <&rpmhpd_opp_svs_l1>;
+                };
+
+                opp-444000000 {
+                    opp-hz = /bits/ 64 <444000000>;
+                    required-opps = <&rpmhpd_opp_nom>;
+                };
+
+                opp-533000000 {
+                    opp-hz = /bits/ 64 <533000000>;
+                    required-opps = <&rpmhpd_opp_turbo>;
+                };
+            };
           };
             videocc: clock-controller@ab00000 {
 
Signature · 2 lines
regards,
Stan
#21 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Stanimir Varbanov

Hi,

Quoted 67 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Quoted 52 lines · Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
 arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +27 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
 			interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
 			power-domains = <&videocc VENUS_GDSC>,
 					<&videocc VCODEC0_GDSC>,
-					<&videocc VCODEC1_GDSC>;
-			power-domain-names = "venus", "vcodec0", "vcodec1";
+					<&videocc VCODEC1_GDSC>,
+					<&rpmhpd SDM845_CX>;
+			power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+			operating-points-v2 = <&venus_opp_table>;
 			clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
 				 <&videocc VIDEO_CC_VENUS_AHB_CLK>,
 				 <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
 			video-core1 {
 				compatible = "venus-encoder";
 			};
+
+			venus_opp_table: venus-opp-table {
+				compatible = "operating-points-v2";
+
+				opp-100000000 {
+					opp-hz = /bits/ 64 <100000000>;
+					required-opps = <&rpmhpd_opp_min_svs>;
+				};
+
+				opp-200000000 {
+					opp-hz = /bits/ 64 <200000000>;
+					required-opps = <&rpmhpd_opp_low_svs>;
+				};
+
+				opp-320000000 {
+					opp-hz = /bits/ 64 <320000000>;
+					required-opps = <&rpmhpd_opp_svs>;
+				};
+
+				opp-380000000 {
+					opp-hz = /bits/ 64 <380000000>;
+					required-opps = <&rpmhpd_opp_svs_l1>;
+				};
Quoted 8 lines

arch/arm64/boot/dts/qcom/sdm845.dtsi line 3689

Quoted 8 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +8 -0
+
+				opp-444000000 {
+					opp-hz = /bits/ 64 <444000000>;
+					required-opps = <&rpmhpd_opp_nom>;
+				};
+
+				opp-533000000 {
+					opp-hz = /bits/ 64 <533000000>;

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

If I change to opp-hz = /bits/ 64 <533000097> the error disappear.

I guess we have to revisit m/n and/or pre-divider for this freq when the source pll is P_VIDEO_PLL0_OUT_MAIN PLL?

Quoted 8 lines
Quoted 7 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +3 -0
+					required-opps = <&rpmhpd_opp_turbo>;
+				};
+			};
 		};
 
 		videocc: clock-controller@ab00000 {
Signature · 2 lines
regards,
Stan
Acked-by on v4 2/5
#4 lore raw

Re: [PATCH v4 2/5] media: venus: core: Fix error handling in probe

In reply to Rajendra Nayak
Quoted 8 lines · Rajendra Nayak

Post a successful pm_ops->core_get, an error in probe should exit by doing a pm_ops->core_put which seems to be missing. So fix it.

---
 drivers/media/platform/qcom/venus/core.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
Quoted 49 lines
drivers/media/platform/qcom/venus/core.c +10 -5
index 203c653..bfcaba3 100644
@@ -224,13 +224,15 @@ static int venus_probe(struct platform_device *pdev)
 
 	ret = dma_set_mask_and_coherent(dev, core->res->dma_mask);
 	if (ret)
-		return ret;
+		goto err_core_put;
 
 	if (!dev->dma_parms) {
 		dev->dma_parms = devm_kzalloc(dev, sizeof(*dev->dma_parms),
 					      GFP_KERNEL);
-		if (!dev->dma_parms)
-			return -ENOMEM;
+		if (!dev->dma_parms) {
+			ret = -ENOMEM;
+			goto err_core_put;
+		}
 	}
 	dma_set_max_seg_size(dev, DMA_BIT_MASK(32));
 
@@ -242,11 +244,11 @@ static int venus_probe(struct platform_device *pdev)
 					IRQF_TRIGGER_HIGH | IRQF_ONESHOT,
 					"venus", core);
 	if (ret)
-		return ret;
+		goto err_core_put;
 
 	ret = hfi_create(core, &venus_core_ops);
 	if (ret)
-		return ret;
+		goto err_core_put;
 
 	pm_runtime_enable(dev);
 
@@ -302,6 +304,9 @@ static int venus_probe(struct platform_device *pdev)
 	pm_runtime_set_suspended(dev);
 	pm_runtime_disable(dev);
 	hfi_destroy(core);
+err_core_put:
+	if (core->pm_ops->core_put)
+		core->pm_ops->core_put(dev);
 	return ret;
 }
 
Signature · 2 lines
regards,
Stan
#12 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Rajendra Nayak
Quoted 17 lines · Rajendra Nayak

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

--Lina

#13 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Hi Lina,

Quoted 12 lines · Lina Iyer
Quoted 11 lines · Rajendra Nayak

Hi Maulik/Lina,

Quoted 8 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

Quoted 7 lines
Quoted 6 lines
Quoted 5 lines

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?

I haven't gig deeply but on every driver pm_runtime_suspend (after applying Rajendra's patches). And I guess it comes after a call to dev_pm_opp_set_rate(dev, 0).

IMO this is too often.

Quoted 10 lines

Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

--Lina

Signature · 2 lines
regards,
Stan
#14 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Quoted 19 lines · Stanimir Varbanov

Hi Lina,

Quoted 16 lines · Lina Iyer
Quoted 15 lines · Rajendra Nayak

Hi Maulik/Lina,

Quoted 12 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

Quoted 7 lines
Quoted 3 lines
Quoted 2 lines
Quoted 1 line

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?

I haven't gig deeply but on every driver pm_runtime_suspend (after applying Rajendra's patches). And I guess it comes after a call to dev_pm_opp_set_rate(dev, 0).

Or it might be when the driver is switching off opp_pmdomain.

Quoted 14 lines

IMO this is too often.

Quoted 10 lines

Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

--Lina

Signature · 2 lines
regards,
Stan
Acked-by on v4 3/5
#6 lore raw

Re: [PATCH v4 3/5] media: venus: core: Add support for opp tables/perf voting

In reply to Rajendra Nayak

Hi Rajendra,

Quoted 11 lines · Rajendra Nayak

Add support to add OPP tables and perf voting on the OPP powerdomain. This is needed so venus votes on the corresponding performance state for the OPP powerdomain along with setting the core clock rate.

---
 drivers/media/platform/qcom/venus/core.c       |  2 +
 drivers/media/platform/qcom/venus/core.h       |  5 ++
 drivers/media/platform/qcom/venus/pm_helpers.c | 92 ++++++++++++++++++++++++--
 3 files changed, 92 insertions(+), 7 deletions(-)

Once we have a fix in opp-table (patch 4/5) to avoid -ERANGE from dev_pm_opp_set_rate:

<cut>

Signature · 2 lines
regards,
Stan
#22 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Stanimir Varbanov
Quoted 73 lines · Stanimir Varbanov

Hi,

Quoted 70 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Quoted 55 lines · Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
  arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--
  1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +30 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
  			interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
  			power-domains = <&videocc VENUS_GDSC>,
  					<&videocc VCODEC0_GDSC>,
-					<&videocc VCODEC1_GDSC>;
-			power-domain-names = "venus", "vcodec0", "vcodec1";
+					<&videocc VCODEC1_GDSC>,
+					<&rpmhpd SDM845_CX>;
+			power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+			operating-points-v2 = <&venus_opp_table>;
  			clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
  				 <&videocc VIDEO_CC_VENUS_AHB_CLK>,
  				 <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
  			video-core1 {
  				compatible = "venus-encoder";
  			};
+
+			venus_opp_table: venus-opp-table {
+				compatible = "operating-points-v2";
+
+				opp-100000000 {
+					opp-hz = /bits/ 64 <100000000>;
+					required-opps = <&rpmhpd_opp_min_svs>;
+				};
+
+				opp-200000000 {
+					opp-hz = /bits/ 64 <200000000>;
+					required-opps = <&rpmhpd_opp_low_svs>;
+				};
+
+				opp-320000000 {
+					opp-hz = /bits/ 64 <320000000>;
+					required-opps = <&rpmhpd_opp_svs>;
+				};
+
+				opp-380000000 {
+					opp-hz = /bits/ 64 <380000000>;
+					required-opps = <&rpmhpd_opp_svs_l1>;
+				};
+
+				opp-444000000 {
+					opp-hz = /bits/ 64 <444000000>;
Quoted 8 lines
Quoted 5 lines

arch/arm64/boot/dts/qcom/sdm845.dtsi line 3689

Quoted 5 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +5 -0
+					required-opps = <&rpmhpd_opp_nom>;
+				};
+
+				opp-533000000 {
+					opp-hz = /bits/ 64 <533000000>;

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

I still don't quite understand why the videocc driver returns this frequency despite this not being in the freq table.
I would expect a clk_round_rate() when called with 533000097 to return a 533000000.

Taniya, Do you know why?

Quoted 15 lines

If I change to opp-hz = /bits/ 64 <533000097> the error disappear.

I guess we have to revisit m/n and/or pre-divider for this freq when the source pll is P_VIDEO_PLL0_OUT_MAIN PLL?

Quoted 8 lines
Quoted 7 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +3 -0
+					required-opps = <&rpmhpd_opp_turbo>;
+				};
+			};
  		};
 
  		videocc: clock-controller@ab00000 {
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#23 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Quoted 85 lines · Rajendra Nayak
Quoted 73 lines · Stanimir Varbanov

Hi,

Quoted 70 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Quoted 55 lines · Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
  arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--

  1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +30 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
               interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
               power-domains = <&videocc VENUS_GDSC>,
                       <&videocc VCODEC0_GDSC>,
-                    <&videocc VCODEC1_GDSC>;
-            power-domain-names = "venus", "vcodec0", "vcodec1";
+                    <&videocc VCODEC1_GDSC>,
+                    <&rpmhpd SDM845_CX>;
+            power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+            operating-points-v2 = <&venus_opp_table>;
               clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
                    <&videocc VIDEO_CC_VENUS_AHB_CLK>,
                    <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
               video-core1 {
                   compatible = "venus-encoder";
               };
+
+            venus_opp_table: venus-opp-table {
+                compatible = "operating-points-v2";
+
+                opp-100000000 {
+                    opp-hz = /bits/ 64 <100000000>;
+                    required-opps = <&rpmhpd_opp_min_svs>;
+                };
+
+                opp-200000000 {
+                    opp-hz = /bits/ 64 <200000000>;
+                    required-opps = <&rpmhpd_opp_low_svs>;
+                };
+
+                opp-320000000 {
+                    opp-hz = /bits/ 64 <320000000>;
+                    required-opps = <&rpmhpd_opp_svs>;
+                };
+
+                opp-380000000 {
+                    opp-hz = /bits/ 64 <380000000>;
+                    required-opps = <&rpmhpd_opp_svs_l1>;
+                };
+
+                opp-444000000 {
+                    opp-hz = /bits/ 64 <444000000>;
Quoted 8 lines
Quoted 5 lines

arch/arm64/boot/dts/qcom/sdm845.dtsi line 3689

Quoted 5 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +5 -0
+                    required-opps = <&rpmhpd_opp_nom>;
+                };
+
+                opp-533000000 {
+                    opp-hz = /bits/ 64 <533000000>;

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

Quoted 2 lines

I still don't quite understand why the videocc driver returns this frequency despite this not being in the freq table.

Ok, so I see the same issue on sc7180 also. clk_round_rate() does seem to return whats in the freq table, but clk_set_rate() goes ahead and sets it to 533000097. Subsequently when we try to set a different OPP, it fails to find the 'current' OPP entry for 533000097. This sounds like an issue with the OPP framework? Should we not fall back to the highest OPP as the current OPP?

Stephen/Viresh, any thoughts?

Quoted 20 lines

I would expect a clk_round_rate() when called with 533000097 to return a 533000000.

Taniya, Do you know why?

Quoted 14 lines

If I change to opp-hz = /bits/ 64 <533000097> the error disappear.

I guess we have to revisit m/n and/or pre-divider for this freq when the source pll is P_VIDEO_PLL0_OUT_MAIN PLL?

Quoted 7 lines
Quoted 6 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +3 -0
+                    required-opps = <&rpmhpd_opp_turbo>;
+                };
+            };
           };
           videocc: clock-controller@ab00000 {
 
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#24 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Quoted 75 lines · Rajendra Nayak
Quoted 73 lines · Rajendra Nayak
Quoted 70 lines · Stanimir Varbanov

Hi,

Quoted 67 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

Quoted 52 lines · Rajendra Nayak

Add the OPP tables in order to be able to vote on the performance state of a power-domain.

---
  arch/arm64/boot/dts/qcom/sdm845.dtsi | 40 ++++++++++++++++++++++++++++++++++--

  1 file changed, 38 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sdm845.dtsi +27 -2
index e506793..5ca2265 100644
@@ -3631,8 +3631,10 @@
               interrupts = <GIC_SPI 174 IRQ_TYPE_LEVEL_HIGH>;
               power-domains = <&videocc VENUS_GDSC>,
                       <&videocc VCODEC0_GDSC>,
-                    <&videocc VCODEC1_GDSC>;
-            power-domain-names = "venus", "vcodec0", "vcodec1";
+                    <&videocc VCODEC1_GDSC>,
+                    <&rpmhpd SDM845_CX>;
+            power-domain-names = "venus", "vcodec0", "vcodec1", "cx";
+            operating-points-v2 = <&venus_opp_table>;
               clocks = <&videocc VIDEO_CC_VENUS_CTL_CORE_CLK>,
                    <&videocc VIDEO_CC_VENUS_AHB_CLK>,
                    <&videocc VIDEO_CC_VENUS_CTL_AXI_CLK>,
@@ -3654,6 +3656,40 @@
               video-core1 {
                   compatible = "venus-encoder";
               };
+
+            venus_opp_table: venus-opp-table {
+                compatible = "operating-points-v2";
+
+                opp-100000000 {
+                    opp-hz = /bits/ 64 <100000000>;
+                    required-opps = <&rpmhpd_opp_min_svs>;
+                };
+
+                opp-200000000 {
+                    opp-hz = /bits/ 64 <200000000>;
+                    required-opps = <&rpmhpd_opp_low_svs>;
+                };
+
+                opp-320000000 {
+                    opp-hz = /bits/ 64 <320000000>;
+                    required-opps = <&rpmhpd_opp_svs>;
+                };
+
+                opp-380000000 {
+                    opp-hz = /bits/ 64 <380000000>;
+                    required-opps = <&rpmhpd_opp_svs_l1>;
+                };
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

arch/arm64/boot/dts/qcom/sdm845.dtsi line 3689

Quoted 8 lines
arch/arm64/boot/dts/qcom/sdm845.dtsi +8 -0
+
+                opp-444000000 {
+                    opp-hz = /bits/ 64 <444000000>;
+                    required-opps = <&rpmhpd_opp_nom>;
+                };
+
+                opp-533000000 {
+                    opp-hz = /bits/ 64 <533000000>;

Is this the highest OPP in table ?

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

I still don't quite understand why the videocc driver returns this frequency despite this not being in the freq table.

Quoted 7 lines

Ok, so I see the same issue on sc7180 also. clk_round_rate() does seem to return whats in the freq table, but clk_set_rate() goes ahead and sets it to 533000097. Subsequently when we try to set a different OPP, it fails to find the 'current' OPP entry for 533000097. This sounds like an issue with the OPP framework? Should we not fall back to the highest OPP as the current OPP?

Stephen/Viresh, any thoughts?

I think we (in all frameworks generally) try to set a frequency <= target frequency and so there may be a problem if the frequency is larger than highest supported. IOW, you need to fix tables a bit.

Signature · 1 line
viresh
#15 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Lina Iyer

Quoting Lina Iyer (2020-07-24 09:28:25)

Quoted 23 lines · Rajendra Nayak
Quoted 17 lines

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 5 lines
Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

I removed this warning message in my patch posted to the list[1]. If it's a serious problem then I suppose a timeout is more appropriate, on the order of several seconds or so and then a pr_warn() and bail out of the async call with an error.

[1] https://lore.kernel.org/r/[email protected]

#25 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Viresh Kumar

Quoting Viresh Kumar (2020-07-27 08:38:06)

Quoted 17 lines · Rajendra Nayak
Quoted 5 lines · Rajendra Nayak
Quoted 4 lines · Stanimir Varbanov
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

+
+                opp-533000000 {
+                    opp-hz = /bits/ 64 <533000000>;

Is this the highest OPP in table ?

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

I still don't quite understand why the videocc driver returns this frequency despite this not being in the freq table.

Ok, so I see the same issue on sc7180 also. clk_round_rate() does seem to return whats in the freq table, but clk_set_rate() goes ahead and sets it

I'm happy to see clk_round_rate() return the actual rate that would be achieved and not just the rate that is in the frequency tables. Would that fix the problem? It may be that we need to make clk_round_rate() do some more math on qcom platforms and actually figure out what the rate is going to be instead of blindly trust the frequency that has been set in the tables.

Quoted 10 lines
Quoted 5 lines

to 533000097. Subsequently when we try to set a different OPP, it fails to find the 'current' OPP entry for 533000097. This sounds like an issue with the OPP framework? Should we not fall back to the highest OPP as the current OPP?

Stephen/Viresh, any thoughts?

I think we (in all frameworks generally) try to set a frequency <= target frequency and so there may be a problem if the frequency is larger than highest supported. IOW, you need to fix tables a bit.

Rounding is annoying for sure.

#26 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Quoted 16 lines · Stephen Boyd

Quoting Viresh Kumar (2020-07-27 08:38:06)

Quoted 15 lines · Rajendra Nayak
Quoted 5 lines · Rajendra Nayak
Quoted 4 lines · Stanimir Varbanov
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

+
+                opp-533000000 {
+                    opp-hz = /bits/ 64 <533000000>;

Is this the highest OPP in table ?

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

I still don't quite understand why the videocc driver returns this frequency despite this not being in the freq table.

Quoted 6 lines
Quoted 2 lines
Quoted 2 lines

Ok, so I see the same issue on sc7180 also. clk_round_rate() does seem to return whats in the freq table, but clk_set_rate() goes ahead and sets it

I'm happy to see clk_round_rate() return the actual rate that would be achieved and not just the rate that is in the frequency tables. Would that fix the problem?

It would, but only if I also update the OPP table to have 533000097 instead of 533000000 (which I guess is needed anyway)
If this is the actual frequency that's achievable, then perhaps even the clock freq table should have this? 533000097 and not 533000000?
That way clk_round_rate() would return the actual rate that's achieved and we don't need any extra math. Isn't that the reason these freq tables exist anyway.

Quoted 18 lines

It may be that we need to make clk_round_rate() do
some more math on qcom platforms and actually figure out what the rate is going to be instead of blindly trust the frequency that has been set in the tables.

Quoted 10 lines
Quoted 5 lines

to 533000097. Subsequently when we try to set a different OPP, it fails to find the 'current' OPP entry for 533000097. This sounds like an issue with the OPP framework? Should we not fall back to the highest OPP as the current OPP?

Stephen/Viresh, any thoughts?

I think we (in all frameworks generally) try to set a frequency <= target frequency and so there may be a problem if the frequency is larger than highest supported. IOW, you need to fix tables a bit.

Rounding is annoying for sure.

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#16 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Stephen Boyd
Quoted 26 lines · Stephen Boyd

Quoting Lina Iyer (2020-07-24 09:28:25)

Quoted 25 lines · Rajendra Nayak
Quoted 17 lines

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

Quoted 9 lines
Quoted 3 lines

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

I removed this warning message in my patch posted to the list[1]. If it's a serious problem then I suppose a timeout is more appropriate, on the order of several seconds or so and then a pr_warn() and bail out of the async call with an error.

The warning used to capture issues that happen within a second and it helps capture system related issues. Timing out after many seconds overlooks the system issues that generally tend to resolve itself, but nevertheless need to be investigated.

--Lina

Quoted 1 line

[1] https://lore.kernel.org/r/[email protected]

#17 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Quoting Lina Iyer (2020-07-28 09:52:12)

Quoted 36 lines · Stephen Boyd
Quoted 26 lines

Quoting Lina Iyer (2020-07-24 09:28:25)

Quoted 25 lines · Rajendra Nayak
Quoted 17 lines

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

Quoted 9 lines
Quoted 3 lines

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

I removed this warning message in my patch posted to the list[1]. If it's a serious problem then I suppose a timeout is more appropriate, on the order of several seconds or so and then a pr_warn() and bail out of the async call with an error.

Quoted 5 lines

The warning used to capture issues that happen within a second and it helps capture system related issues. Timing out after many seconds overlooks the system issues that generally tend to resolve itself, but nevertheless need to be investigated.

Is it correct to read "system related issues" as performance problems where the thread is spinning forever trying to send a message and it can't? So the problem is mostly that it's an unbounded amount of time before the message is sent to rpmh and this printk helps identify those situations where that is happening?

Otherwise as you say above it's a bad system state where the rpmh processor has gotten into a bad state like a crash? Can we recover from that? Or is the only recovery a reboot of the system? Does the rpmh processor reboot the system if it crashes?

#27 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Rajendra Nayak

Quoting Rajendra Nayak (2020-07-27 21:17:28)

Quoted 25 lines · Stephen Boyd
Quoted 16 lines

Quoting Viresh Kumar (2020-07-27 08:38:06)

Quoted 15 lines · Rajendra Nayak
Quoted 5 lines · Rajendra Nayak
Quoted 4 lines · Stanimir Varbanov
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

+
+                opp-533000000 {
+                    opp-hz = /bits/ 64 <533000000>;

Is this the highest OPP in table ?

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

Actually it comes from videocc, where ftbl_video_cc_venus_clk_src defines 533000000 but the real calculated freq is 533000097.

I still don't quite understand why the videocc driver returns this frequency despite this not being in the freq table.

Quoted 6 lines
Quoted 2 lines
Quoted 2 lines

Ok, so I see the same issue on sc7180 also. clk_round_rate() does seem to return whats in the freq table, but clk_set_rate() goes ahead and sets it

I'm happy to see clk_round_rate() return the actual rate that would be achieved and not just the rate that is in the frequency tables. Would that fix the problem?

Quoted 7 lines

It would, but only if I also update the OPP table to have 533000097 instead of 533000000 (which I guess is needed anyway)
If this is the actual frequency that's achievable, then perhaps even the clock freq table should have this? 533000097 and not 533000000?
That way clk_round_rate() would return the actual rate that's achieved and we don't need any extra math. Isn't that the reason these freq tables exist anyway.

Yes the freq tables are there in the clk driver so we don't have to do a bunch of math. Fixing them to be accurate has been deemed "hard" from what I recall because the tables are generated from some math function that truncates the lower Hertz values.

#18 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Stephen Boyd
Quoted 43 lines · Stephen Boyd

Quoting Lina Iyer (2020-07-28 09:52:12)

Quoted 36 lines · Stephen Boyd
Quoted 26 lines

Quoting Lina Iyer (2020-07-24 09:28:25)

Quoted 25 lines · Rajendra Nayak
Quoted 17 lines

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

Quoted 9 lines
Quoted 3 lines

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

I removed this warning message in my patch posted to the list[1]. If it's a serious problem then I suppose a timeout is more appropriate, on the order of several seconds or so and then a pr_warn() and bail out of the async call with an error.

Quoted 5 lines

The warning used to capture issues that happen within a second and it helps capture system related issues. Timing out after many seconds overlooks the system issues that generally tend to resolve itself, but nevertheless need to be investigated.

Quoted 6 lines

Is it correct to read "system related issues" as performance problems where the thread is spinning forever trying to send a message and it can't? So the problem is mostly that it's an unbounded amount of time before the message is sent to rpmh and this printk helps identify those situations where that is happening?

Yes, but mostly a short period of time like when other processors are in the middle of a restart or resource states changes have taken unusual amounts of time. The system will generally recover from this without crashing in this case. User action is investigation of the situation leading to these messages.

Quoted 4 lines

Otherwise as you say above it's a bad system state where the rpmh processor has gotten into a bad state like a crash? Can we recover from that? Or is the only recovery a reboot of the system? Does the rpmh processor reboot the system if it crashes?

We cannot recover from such a state. The remote processor will reboot if it detects a failure at it's end. If the system entered a bad state, it is possible that RPMH requests start timing out in Linux and remote processor may not detect it. Hence, the timeout in rpmh_write() API. The advised course of action is a restart as there is no way to recover from this state.

--Lina

#19 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

Hi,

Quoted 51 lines · Lina Iyer
Quoted 43 lines · Stephen Boyd

Quoting Lina Iyer (2020-07-28 09:52:12)

Quoted 36 lines · Stephen Boyd
Quoted 26 lines

Quoting Lina Iyer (2020-07-24 09:28:25)

Quoted 25 lines · Rajendra Nayak
Quoted 17 lines

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

Quoted 9 lines
Quoted 3 lines

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

I removed this warning message in my patch posted to the list[1]. If it's a serious problem then I suppose a timeout is more appropriate, on the order of several seconds or so and then a pr_warn() and bail out of the async call with an error.

Quoted 5 lines

The warning used to capture issues that happen within a second and it helps capture system related issues. Timing out after many seconds overlooks the system issues that generally tend to resolve itself, but nevertheless need to be investigated.

Quoted 6 lines

Is it correct to read "system related issues" as performance problems where the thread is spinning forever trying to send a message and it can't? So the problem is mostly that it's an unbounded amount of time before the message is sent to rpmh and this printk helps identify those situations where that is happening?

Quoted 5 lines

Yes, but mostly a short period of time like when other processors are in the middle of a restart or resource states changes have taken unusual amounts of time. The system will generally recover from this without crashing in this case. User action is investigation of the situation leading to these messages.

While I do agree that seeing the "TCS Busy, retrying RPMH message send" message printed a lot was usually a sign that something was wrong in the system (possibly someone was spamming RPMh when they shouldn't be), it still feels like we need to remove it. Specifically, the prints would also sometimes come up in normal usage and always sounded a bit scary. These types of prints always confuse people and lead to log pollution where it's super hard to figure out which of the various things in a log are "expected" and which ones are relevant to whatever issue you're debugging.

Presumably we could either change that from a "info" level to "dbg" level. ...or we could find some other thing to check for that's a better signal of problems.

Quoted 14 lines
Quoted 4 lines

Otherwise as you say above it's a bad system state where the rpmh processor has gotten into a bad state like a crash? Can we recover from that? Or is the only recovery a reboot of the system? Does the rpmh processor reboot the system if it crashes?

We cannot recover from such a state. The remote processor will reboot if it detects a failure at it's end. If the system entered a bad state, it is possible that RPMH requests start timing out in Linux and remote processor may not detect it. Hence, the timeout in rpmh_write() API. The advised course of action is a restart as there is no way to recover from this state.

--Lina

#20 lore raw

Re: [PATCH v4 4/5] arm64: dts: sdm845: Add OPP tables and power-domains for venus

In reply to Lina Iyer
Quoted 50 lines · Lina Iyer
Quoted 43 lines · Stephen Boyd

Quoting Lina Iyer (2020-07-28 09:52:12)

Quoted 36 lines · Stephen Boyd
Quoted 26 lines

Quoting Lina Iyer (2020-07-24 09:28:25)

Quoted 25 lines · Rajendra Nayak
Quoted 17 lines

Hi Maulik/Lina,

Quoted 13 lines · Stanimir Varbanov

Hi Rajendra,

After applying 2,3 and 4/5 patches on linaro-integration v5.8-rc2 I see below messages on db845:

qcom-venus aa00000.video-codec: dev_pm_opp_set_rate: failed to find current OPP for freq 533000097 (-34)

^^^ This one is new.

qcom_rpmh TCS Busy, retrying RPMH message send: addr=0x30000

^^^ and this message is annoying, can we make it pr_debug in rpmh?

How annoyingly often do you see this message?
Usually, this is an indication of bad system state either on remote processors in the SoC or in Linux itself. On a smooth sailing build you should not see this 'warning'.

Quoted 2 lines

Would you be fine with moving this message to a pr_debug? Its currently a pr_info_ratelimited()

Quoted 9 lines
Quoted 3 lines

I would rather not, moving this out of sight will mask a lot serious issues that otherwise bring attention to the developers.

I removed this warning message in my patch posted to the list[1]. If it's a serious problem then I suppose a timeout is more appropriate, on the order of several seconds or so and then a pr_warn() and bail out of the async call with an error.

Quoted 5 lines

The warning used to capture issues that happen within a second and it helps capture system related issues. Timing out after many seconds overlooks the system issues that generally tend to resolve itself, but nevertheless need to be investigated.

Quoted 6 lines

Is it correct to read "system related issues" as performance problems where the thread is spinning forever trying to send a message and it can't? So the problem is mostly that it's an unbounded amount of time before the message is sent to rpmh and this printk helps identify those situations where that is happening?

Quoted 6 lines

Yes, but mostly a short period of time like when other processors are in the middle of a restart or resource states changes have taken unusual amounts of time. The system will generally recover from this without crashing in this case. User action is investigation of the situation leading to these messages.

Given that these messages shows up from time and seemingly is harmless, users such as myself implements the action of ignoring these printouts.

In the cases I do see these messages it seems, as you say, to be related to something happening in the firmware. So it's not something that a user typically could investigate/debug anyways.

As such I do second Doug's request of not printing what looks like error messages unless there is a persistent problem - but provide some means for the few who would find them useful..

Regards,
Bjorn

Quoted 14 lines
Quoted 4 lines

Otherwise as you say above it's a bad system state where the rpmh processor has gotten into a bad state like a crash? Can we recover from that? Or is the only recovery a reboot of the system? Does the rpmh processor reboot the system if it crashes?

We cannot recover from such a state. The remote processor will reboot if it detects a failure at it's end. If the system entered a bad state, it is possible that RPMH requests start timing out in Linux and remote processor may not detect it. Hence, the timeout in rpmh_write() API. The advised course of action is a restart as there is no way to recover from this state.

--Lina