Skip to contentSkip to conversation

Revert "venus: firmware: Correct non-pix start and end addresses"

PATCH In review Javier Martinez Canillas 5 participants 21 messages Feb 7, 2023 – Apr 3, 2023 2 messages not in the archive

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Revert "venus: firmware: Correct non-pix start and end addresses"

This reverts commit a837e5161cfffbb3242cc0eb574f8bf65fd32640, which broke probing of the venus driver, at least on the SC7180 SoC HP X2 Chromebook:

  [   11.455782] qcom-venus aa00000.video-codec: Adding to iommu group 11
  [   11.506980] qcom-venus aa00000.video-codec: non legacy binding
  [   12.143432] qcom-venus aa00000.video-codec: failed to reset venus core
  [   12.156440] qcom-venus: probe of aa00000.video-codec failed with error -110

Matthias Kaehlcke also reported that the same change caused a regression in SC7180 and sc7280, that prevents AOSS from entering sleep mode during system suspend. So let's revert this commit for now to fix both issues.

Fixes a837e5161cff ("venus: firmware: Correct non-pix start and end addresses")

---

 drivers/media/platform/qcom/venus/firmware.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/media/platform/qcom/venus/firmware.c +2 -2
index 142d4c74017c..d59ecf776715 100644
@@ -38,8 +38,8 @@ static void venus_reset_cpu(struct venus_core *core)
 	writel(fw_size, wrapper_base + WRAPPER_FW_END_ADDR);
 	writel(0, wrapper_base + WRAPPER_CPA_START_ADDR);
 	writel(fw_size, wrapper_base + WRAPPER_CPA_END_ADDR);
-	writel(0, wrapper_base + WRAPPER_NONPIX_START_ADDR);
-	writel(0, wrapper_base + WRAPPER_NONPIX_END_ADDR);
+	writel(fw_size, wrapper_base + WRAPPER_NONPIX_START_ADDR);
+	writel(fw_size, wrapper_base + WRAPPER_NONPIX_END_ADDR);
 
 	if (IS_V6(core)) {
 		/* Bring XTSS out of reset */
Signature · 1 line
2.39.1
#1 lore raw

Hi Javier and Matthias,
Can we try the attached patch if that fixes the suspend issue for sc7180 and sc7280 ?

Quoted 59 lines

-----Original Message-----
From: Javier Martinez Canillas <[email protected]>
Sent: Tuesday, February 7, 2023 3:53 PM
To: [email protected]
Cc: Albert Esteve <[email protected]>; [email protected]; Matthias Kaehlcke <[email protected]>; Enric Balletbo i Serra <[email protected]>; Javier Martinez Canillas <[email protected]>; Andy Gross <[email protected]>; Bjorn Andersson <[email protected]>; Konrad Dybcio <[email protected]>; Mauro Carvalho Chehab <[email protected]>; Stanimir Varbanov <[email protected]>; Vikash Garodia (QUIC) <[email protected]>; [email protected]; linux- [email protected]
Subject: [PATCH] Revert "venus: firmware: Correct non-pix start and end addresses"

WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.

This reverts commit a837e5161cfffbb3242cc0eb574f8bf65fd32640, which broke probing of the venus driver, at least on the SC7180 SoC HP X2 Chromebook:

  [   11.455782] qcom-venus aa00000.video-codec: Adding to iommu group 11
  [   11.506980] qcom-venus aa00000.video-codec: non legacy binding
  [   12.143432] qcom-venus aa00000.video-codec: failed to reset venus core
  [   12.156440] qcom-venus: probe of aa00000.video-codec failed with error -

110

Matthias Kaehlcke also reported that the same change caused a regression in SC7180 and sc7280, that prevents AOSS from entering sleep mode during system suspend. So let's revert this commit for now to fix both issues.

Fixes a837e5161cff ("venus: firmware: Correct non-pix start and end

addresses")

---

 drivers/media/platform/qcom/venus/firmware.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/media/platform/qcom/venus/firmware.c +2 -3
index 142d4c74017c..d59ecf776715 100644
@@ -38,8 +38,8 @@ static void venus_reset_cpu(struct venus_core *core)
        writel(fw_size, wrapper_base + WRAPPER_FW_END_ADDR);
        writel(0, wrapper_base + WRAPPER_CPA_START_ADDR);
        writel(fw_size, wrapper_base + WRAPPER_CPA_END_ADDR);
-       writel(0, wrapper_base + WRAPPER_NONPIX_START_ADDR);
-       writel(0, wrapper_base + WRAPPER_NONPIX_END_ADDR);
+       writel(fw_size, wrapper_base + WRAPPER_NONPIX_START_ADDR);
+       writel(fw_size, wrapper_base + WRAPPER_NONPIX_END_ADDR);
 
        if (IS_V6(core)) {
                /* Bring XTSS out of reset */
--
Quoted 1 line

2.39.1

Thanks,
Vikash

From bcce3fd9e60a47cb7063b6a5b66867e19a0d0316 Mon Sep 17 00:00:00 2001
From: Vikash Garodia <[email protected]>
Date: Tue, 7 Feb 2023 17:42:11 +0530
Subject: [PATCH] firmware: video: configure the secure and non secure region

Program video registers about the secure and non secure regions.

---
 drivers/media/platform/qcom/venus/core.c     |  8 ++++++++
 drivers/media/platform/qcom/venus/firmware.c | 13 +++++++++++++
 2 files changed, 21 insertions(+)
drivers/media/platform/qcom/venus/core.c +8 -0
index 990a151..e0a49c9 100644
@@ -762,6 +762,10 @@ static const struct venus_resources sc7180_res = {
 	.vmem_size = 0,
 	.vmem_addr = 0,
 	.dma_mask = 0xe0000000 - 1,
+	.cp_start = 0,
+	.cp_size = 0x70800000,
+	.cp_nonpixel_start = 0x1000000,
+	.cp_nonpixel_size = 0x24800000,
 	.fwname = "qcom/venus-5.4/venus.mdt",
 };
 
@@ -872,6 +876,10 @@ static const struct venus_resources sc7280_res = {
 	.vmem_id = VIDC_RESOURCE_NONE,
 	.vmem_size = 0,
 	.vmem_addr = 0,
+	.cp_start = 0,
+	.cp_size = 0x25800000,
+	.cp_nonpixel_start = 0x1000000,
+	.cp_nonpixel_size = 0x24800000,
 	.dma_mask = 0xe0000000 - 1,
 	.fwname = "qcom/vpu-2.0/venus.mbn",
 };
drivers/media/platform/qcom/venus/firmware.c +13 -0
index 142d4c7..5cc977c 100644
@@ -210,6 +210,7 @@ int venus_boot(struct venus_core *core)
 	struct device *dev = core->dev;
 	const struct venus_resources *res = core->res;
 	const char *fwpath = NULL;
+	void __iomem *wrapper_base;
 	phys_addr_t mem_phys;
 	size_t mem_size;
 	int ret;
@@ -253,6 +254,18 @@ int venus_boot(struct venus_core *core)
 		}
 	}
 
+	if (!core->use_tz && res->cp_size) {
+		if (IS_V6(core))
+			wrapper_base = core->wrapper_tz_base;
+		else
+			wrapper_base = core->wrapper_base;
+
+		writel(res->cp_start, wrapper_base + WRAPPER_CPA_START_ADDR);
+		writel(res->cp_start + res->cp_size, wrapper_base + WRAPPER_CPA_END_ADDR);
+		writel(res->cp_nonpixel_start, wrapper_base + WRAPPER_NONPIX_START_ADDR);
+		writel(res->cp_nonpixel_start + res->cp_nonpixel_size, wrapper_base + WRAPPER_NONPIX_END_ADDR);
+	}
+
 	return 0;
 }
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#2 lore raw

Hi Vikash,

Quoted 2 lines · Vikash Garodia

Hi Javier and Matthias,
Can we try the attached patch if that fixes the suspend issue for sc7180 and sc7280 ?

On my side the patch fixes the issue for sc7280, but not sc7180.

Quoted 63 lines
Quoted 59 lines

-----Original Message-----
From: Javier Martinez Canillas <[email protected]>
Sent: Tuesday, February 7, 2023 3:53 PM
To: [email protected]
Cc: Albert Esteve <[email protected]>; [email protected]; Matthias Kaehlcke <[email protected]>; Enric Balletbo i Serra <[email protected]>; Javier Martinez Canillas <[email protected]>; Andy Gross <[email protected]>; Bjorn Andersson <[email protected]>; Konrad Dybcio <[email protected]>; Mauro Carvalho Chehab <[email protected]>; Stanimir Varbanov <[email protected]>; Vikash Garodia (QUIC) <[email protected]>; [email protected]; linux- [email protected]
Subject: [PATCH] Revert "venus: firmware: Correct non-pix start and end addresses"

WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.

This reverts commit a837e5161cfffbb3242cc0eb574f8bf65fd32640, which broke probing of the venus driver, at least on the SC7180 SoC HP X2 Chromebook:

  [   11.455782] qcom-venus aa00000.video-codec: Adding to iommu group 11
  [   11.506980] qcom-venus aa00000.video-codec: non legacy binding
  [   12.143432] qcom-venus aa00000.video-codec: failed to reset venus core
  [   12.156440] qcom-venus: probe of aa00000.video-codec failed with error -

110

Matthias Kaehlcke also reported that the same change caused a regression in SC7180 and sc7280, that prevents AOSS from entering sleep mode during system suspend. So let's revert this commit for now to fix both issues.

Fixes a837e5161cff ("venus: firmware: Correct non-pix start and end

addresses")

---

 drivers/media/platform/qcom/venus/firmware.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/media/platform/qcom/venus/firmware.c +2 -3
index 142d4c74017c..d59ecf776715 100644
@@ -38,8 +38,8 @@ static void venus_reset_cpu(struct venus_core *core)
        writel(fw_size, wrapper_base + WRAPPER_FW_END_ADDR);
        writel(0, wrapper_base + WRAPPER_CPA_START_ADDR);
        writel(fw_size, wrapper_base + WRAPPER_CPA_END_ADDR);
-       writel(0, wrapper_base + WRAPPER_NONPIX_START_ADDR);
-       writel(0, wrapper_base + WRAPPER_NONPIX_END_ADDR);
+       writel(fw_size, wrapper_base + WRAPPER_NONPIX_START_ADDR);
+       writel(fw_size, wrapper_base + WRAPPER_NONPIX_END_ADDR);
 
        if (IS_V6(core)) {
                /* Bring XTSS out of reset */
--
Quoted 1 line

2.39.1

Thanks,
Vikash

#3 lore raw
Quoted 7 lines · On Tue, Feb 07, 2023 at 05:50:19PM +0000, [email protected] wrote:

Hi Vikash,

Quoted 2 lines · Vikash Garodia

Hi Javier and Matthias,
Can we try the attached patch if that fixes the suspend issue for sc7180 and sc7280 ?

On my side the patch fixes the issue for sc7280, but not sc7180.

Some more info for sc7180:

[   10.313055] qcom-venus aa00000.video-codec: failed to reset venus core
[   10.331454] qcom-venus: probe of aa00000.video-codec failed with error -110

So venus didn't probe successfully. As a result sync_state() of its rpmhpd and interconnects isn't called and they keep running at max speed, which prevents the Always-On subsystem from suspending:

[   30.171148] qcom-rpmhpd 18200000.rsc:power-controller: Consumer 'aa00000.video-codec' did not probe (successfully)
[   30.682950] qnoc-sc7180 9680000.interconnect: Consumer 'aa00000.video-codec' did not probe (successfully)
[   30.701843] qnoc-sc7180 1740000.interconnect: Consumer 'aa00000.video-codec' did not probe (successfully)
[   30.720168] qnoc-sc7180 1638000.interconnect: Consumer 'aa00000.video-codec' did not probe (successfully)
[   30.738478] qnoc-sc7180 1500000.interconnect: Consumer 'aa00000.video-codec' did not probe (successfully)

(these debug logs are not upstream)

Quoted 65 lines
Quoted 63 lines
Quoted 59 lines

-----Original Message-----
From: Javier Martinez Canillas <[email protected]>
Sent: Tuesday, February 7, 2023 3:53 PM
To: [email protected]
Cc: Albert Esteve <[email protected]>; [email protected]; Matthias Kaehlcke <[email protected]>; Enric Balletbo i Serra <[email protected]>; Javier Martinez Canillas <[email protected]>; Andy Gross <[email protected]>; Bjorn Andersson <[email protected]>; Konrad Dybcio <[email protected]>; Mauro Carvalho Chehab <[email protected]>; Stanimir Varbanov <[email protected]>; Vikash Garodia (QUIC) <[email protected]>; [email protected]; linux- [email protected]
Subject: [PATCH] Revert "venus: firmware: Correct non-pix start and end addresses"

WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.

This reverts commit a837e5161cfffbb3242cc0eb574f8bf65fd32640, which broke probing of the venus driver, at least on the SC7180 SoC HP X2 Chromebook:

  [   11.455782] qcom-venus aa00000.video-codec: Adding to iommu group 11
  [   11.506980] qcom-venus aa00000.video-codec: non legacy binding
  [   12.143432] qcom-venus aa00000.video-codec: failed to reset venus core
  [   12.156440] qcom-venus: probe of aa00000.video-codec failed with error -

110

Matthias Kaehlcke also reported that the same change caused a regression in SC7180 and sc7280, that prevents AOSS from entering sleep mode during system suspend. So let's revert this commit for now to fix both issues.

Fixes a837e5161cff ("venus: firmware: Correct non-pix start and end

addresses")

---

 drivers/media/platform/qcom/venus/firmware.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/media/platform/qcom/venus/firmware.c +2 -3
index 142d4c74017c..d59ecf776715 100644
@@ -38,8 +38,8 @@ static void venus_reset_cpu(struct venus_core *core)
        writel(fw_size, wrapper_base + WRAPPER_FW_END_ADDR);
        writel(0, wrapper_base + WRAPPER_CPA_START_ADDR);
        writel(fw_size, wrapper_base + WRAPPER_CPA_END_ADDR);
-       writel(0, wrapper_base + WRAPPER_NONPIX_START_ADDR);
-       writel(0, wrapper_base + WRAPPER_NONPIX_END_ADDR);
+       writel(fw_size, wrapper_base + WRAPPER_NONPIX_START_ADDR);
+       writel(fw_size, wrapper_base + WRAPPER_NONPIX_END_ADDR);
 
        if (IS_V6(core)) {
                /* Bring XTSS out of reset */
--
Quoted 1 line

2.39.1

Thanks,
Vikash

#4 lore raw
In reply to Vikash Garodia

Hello Vikash,

Quoted 3 lines · Vikash Garodia

Hi Javier and Matthias,
Can we try the attached patch if that fixes the suspend issue for sc7180 and sc7280 ?

I tested your attached patch on an SC7180 machine (HP X2 Chromebook) and as Matthias mentioned, it still causes the driver's probe to fail:

[ 2119.063779] qcom-venus aa00000.video-codec: non legacy binding
[ 2119.085695] platform video-firmware.0: Adding to iommu group 11
[ 2119.156302] arm-smmu 15000000.iommu: Unhandled context fault: fsr=0x402, iova=0x000000b0, fsynr=0x61, cbfrsynra=0xc40, cb=7
[ 2119.259382] qcom-venus aa00000.video-codec: failed to reset venus core
[ 2119.267782] platform video-firmware.0: Removing from iommu group 11
[ 2119.275052] qcom-venus: probe of aa00000.video-codec failed with error -110
Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
#5 lore raw
Quoted 10 lines · Javier Martinez Canillas

Hello Vikash,

Quoted 3 lines · Vikash Garodia

Hi Javier and Matthias,
Can we try the attached patch if that fixes the suspend issue for sc7180 and sc7280 ?

I tested your attached patch on an SC7180 machine (HP X2 Chromebook) and as Matthias mentioned, it still causes the driver's probe to fail:

Quoted 7 lines
[ 2119.063779] qcom-venus aa00000.video-codec: non legacy binding
[ 2119.085695] platform video-firmware.0: Adding to iommu group 11
[ 2119.156302] arm-smmu 15000000.iommu: Unhandled context fault: fsr=0x402, iova=0x000000b0, fsynr=0x61, cbfrsynra=0xc40, cb=7
[ 2119.259382] qcom-venus aa00000.video-codec: failed to reset venus core
[ 2119.267782] platform video-firmware.0: Removing from iommu group 11
[ 2119.275052] qcom-venus: probe of aa00000.video-codec failed with error -110

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
#6 lore raw

Hi Javier,

Quoted 42 lines

-----Original Message-----
From: Javier Martinez Canillas <[email protected]>
Sent: Friday, February 10, 2023 1:45 PM
To: Vikash Garodia <[email protected]>; linux- [email protected]; [email protected]
Cc: Albert Esteve <[email protected]>; [email protected]; Enric Balletbo i Serra <[email protected]>; Andy Gross <[email protected]>; Bjorn Andersson <[email protected]>; Konrad Dybcio <[email protected]>; Mauro Carvalho Chehab <[email protected]>; Stanimir Varbanov <[email protected]>; Vikash Garodia (QUIC) <[email protected]>; [email protected]; linux- [email protected]; Fritz Koenig <[email protected]>; Dikshita Agarwal (QUIC) <[email protected]>; Rajeshwar Kurapaty (QUIC) <[email protected]>
Subject: Re: [PATCH] Revert "venus: firmware: Correct non-pix start and end addresses"

WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.

Quoted 5 lines · Javier Martinez Canillas

Hello Vikash,

Quoted 2 lines · Vikash Garodia

Hi Javier and Matthias,
Can we try the attached patch if that fixes the suspend issue for sc7180 and

sc7280 ?

Quoted 5 lines
Quoted 1 line

I tested your attached patch on an SC7180 machine (HP X2 Chromebook) and as Matthias mentioned, it still causes the driver's probe to fail:

Quoted 9 lines
[ 2119.063779] qcom-venus aa00000.video-codec: non legacy binding [

2119.085695] platform video-firmware.0: Adding to iommu group 11 [ 2119.156302] arm-smmu 15000000.iommu: Unhandled context fault:
fsr=0x402, iova=0x000000b0, fsynr=0x61, cbfrsynra=0xc40, cb=7 [ 2119.259382] qcom-venus aa00000.video-codec: failed to reset venus core [ 2119.267782] platform video-firmware.0: Removing from iommu group 11 [ 2119.275052] qcom-venus: probe of aa00000.video-codec failed with error -110

Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Quoted 6 lines

--
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat

#7 lore raw

Hello Vikash,

On 2/10/23 10:22, Vikash Garodia wrote:

[...]

Quoted 10 lines
Quoted 4 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Great, if you are working on a proper fix then that would be much better indeed.

Thanks for the follow-up!

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
#8 lore raw
Quoted 6 lines · Javier Martinez Canillas
Quoted 4 lines · Vikash Garodia
Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

Quoted 6 lines
Quoted 4 lines

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Great, if you are working on a proper fix then that would be much better indeed.

Yeah, that's great, but OTOH: there is almost certainly just one week before 6.2 will be released. Ideally this should be fixed by then. Vikash, do you think that's in the cards? If not: why not revert this now to make sure 6.2 works fine?

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#9 lore raw
Quoted 12 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 11 lines · Javier Martinez Canillas
Quoted 8 lines · Vikash Garodia
Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Quoted 6 lines
Quoted 1 line

Great, if you are working on a proper fix then that would be much better indeed.

Yeah, that's great, but OTOH: there is almost certainly just one week before 6.2 will be released. Ideally this should be fixed by then. Vikash, do you think that's in the cards? If not: why not revert this now to make sure 6.2 works fine?

Hmm, no reply. And we meanwhile have Wednesday and 6.2 is almost certainly going to be out on Sunday. And the problem was called "a quite serious regression" above. So why not quickly fix this with the revert, as proposed earlier?

Vikash? Javier?

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#regzbot ignore-activity

#10 lore raw
Quoted 21 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 12 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 11 lines · Javier Martinez Canillas
Quoted 8 lines · Vikash Garodia
Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Quoted 6 lines
Quoted 1 line

Great, if you are working on a proper fix then that would be much better indeed.

Yeah, that's great, but OTOH: there is almost certainly just one week before 6.2 will be released. Ideally this should be fixed by then. Vikash, do you think that's in the cards? If not: why not revert this now to make sure 6.2 works fine?

Quoted 7 lines

Hmm, no reply. And we meanwhile have Wednesday and 6.2 is almost certainly going to be out on Sunday. And the problem was called "a quite serious regression" above. So why not quickly fix this with the revert, as proposed earlier?

Vikash? Javier?

I agree with you, that we should land this revert and then properly fix the page fault issue in v6.3.

But it's not my call, the v4l2/media folks have to decide that.

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
#11 lore raw
Quoted 24 lines · Javier Martinez Canillas
Quoted 16 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 10 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 9 lines · Javier Martinez Canillas
Quoted 7 lines · Vikash Garodia
Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Quoted 5 lines
Quoted 1 line

Great, if you are working on a proper fix then that would be much better indeed.

Yeah, that's great, but OTOH: there is almost certainly just one week before 6.2 will be released. Ideally this should be fixed by then. Vikash, do you think that's in the cards? If not: why not revert this now to make sure 6.2 works fine?

Quoted 5 lines

Hmm, no reply. And we meanwhile have Wednesday and 6.2 is almost certainly going to be out on Sunday. And the problem was called "a quite serious regression" above. So why not quickly fix this with the revert, as proposed earlier?
Vikash? Javier?

Quoted 4 lines

I agree with you, that we should land this revert and then properly fix the page fault issue in v6.3.

But it's not my call, the v4l2/media folks have to decide that.

In that case: Mauro, what's your opinion here?

Thread starts here:
https://lore.kernel.org/lkml/[email protected]/

Regression report:
https://lore.kernel.org/lkml/Y9LSMap%[email protected]/

Ciao, Thorsten

#12 lore raw
Quoted 30 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 24 lines · Javier Martinez Canillas
Quoted 16 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 10 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 9 lines · Javier Martinez Canillas
Quoted 7 lines · Vikash Garodia
Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we proceed to revert.

Quoted 5 lines
Quoted 1 line

Great, if you are working on a proper fix then that would be much better indeed.

Yeah, that's great, but OTOH: there is almost certainly just one week before 6.2 will be released. Ideally this should be fixed by then. Vikash, do you think that's in the cards? If not: why not revert this now to make sure 6.2 works fine?

Quoted 5 lines

Hmm, no reply. And we meanwhile have Wednesday and 6.2 is almost certainly going to be out on Sunday. And the problem was called "a quite serious regression" above. So why not quickly fix this with the revert, as proposed earlier?
Vikash? Javier?

Quoted 4 lines

I agree with you, that we should land this revert and then properly fix the page fault issue in v6.3.

But it's not my call, the v4l2/media folks have to decide that.

Quoted 7 lines

In that case: Mauro, what's your opinion here?

Thread starts here:
https://lore.kernel.org/lkml/[email protected]/

Regression report:
https://lore.kernel.org/lkml/Y9LSMap%[email protected]/

No reply from Mauro and Linus chose to not apply the revert I pointed him to. That at this point leads to the question:

Vikash, did you or somebody else make any progress to fix this properly?

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#regzbot poke

#13 lore raw

Hi All,

Quoted 60 lines

-----Original Message-----
From: Thorsten Leemhuis <[email protected]>
Sent: Tuesday, February 21, 2023 8:33 PM
To: Vikash Garodia <[email protected]>
Cc: [email protected]; [email protected]; Albert Esteve <[email protected]>; [email protected]; Enric Balletbo i Serra <[email protected]>; Andy Gross <[email protected]>; Bjorn Andersson <[email protected]>; Konrad Dybcio <[email protected]>; Stanimir Varbanov <[email protected]>; Vikash Garodia (QUIC) <[email protected]>; [email protected]; linux- [email protected]; Fritz Koenig <[email protected]>; Dikshita Agarwal (QUIC) <[email protected]>; Rajeshwar Kurapaty (QUIC) <[email protected]>; Javier Martinez Canillas <[email protected]>; Linux regressions mailing list <[email protected]>; Mauro Carvalho Chehab <[email protected]>
Subject: Re: [PATCH] Revert "venus: firmware: Correct non-pix start and end addresses"

WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.

Quoted 7 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 6 lines · Javier Martinez Canillas
Quoted 4 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 3 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 2 lines · Javier Martinez Canillas

On 2/10/23 10:22, Vikash Garodia wrote:

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines

So what should we do about this folks? Since not allowing the driver to probe on at least SC7180 is a quite serious regression, can we revert for now until a proper fix is figured out?

I am able to repro this issue on sc7180 and discussing with firmware team on the cause of reset failure. The original patch was raised for fixing rare SMMU faults during warm boot of video hardware. Hence looking to understand the regressing part before we

proceed to revert.

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

Great, if you are working on a proper fix then that would be much better

indeed.

Quoted 21 lines
Quoted 10 lines
Quoted 9 lines
Quoted 4 lines

Yeah, that's great, but OTOH: there is almost certainly just one week before 6.2 will be released. Ideally this should be fixed by then. Vikash, do you think that's in the cards? If not: why not revert this now to make sure 6.2 works fine?

Hmm, no reply. And we meanwhile have Wednesday and 6.2 is almost certainly going to be out on Sunday. And the problem was called "a quite serious regression" above. So why not quickly fix this with the revert, as proposed earlier?
Vikash? Javier?

Quoted 4 lines

I agree with you, that we should land this revert and then properly fix the page fault issue in v6.3.

But it's not my call, the v4l2/media folks have to decide that.

In that case: Mauro, what's your opinion here?

Thread starts here:
https://lore.kernel.org/lkml/[email protected]
om/

Quoted 7 lines
Quoted 2 lines

Regression report:
https://lore.kernel.org/lkml/Y9LSMap%[email protected]/

No reply from Mauro and Linus chose to not apply the revert I pointed him to. That at this point leads to the question:

Vikash, did you or somebody else make any progress to fix this properly?

We tried with different settings for the registers and arrive at a conclusion that the original configuration was proper. There is no need to explicitly configure the secure non-pixel region when there is no support for the usecase. So, in summary, we are good to have the revert.

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Quoted 7 lines

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#regzbot poke

#14 lore raw

Vikash Garodia <[email protected]> writes:

Hello Vikash,

Quoted 2 lines

Hi All,

[...]

Quoted 11 lines
Quoted 5 lines

No reply from Mauro and Linus chose to not apply the revert I pointed him to. That at this point leads to the question:

Vikash, did you or somebody else make any progress to fix this properly?

We tried with different settings for the registers and arrive at a conclusion that the original configuration was proper. There is no need to explicitly configure the secure non-pixel region when there is no support for the usecase. So, in summary, we are good to have the revert.

Perfect. Thanks a lot for looking at this.

Quoted 3 lines

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Other fix posted is "media: venus: dec: Fix capture formats enumeration order":

https://patchwork.kernel.org/project/linux-media/patch/[email protected]/

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
#15 lore raw
Quoted 23 lines · Javier Martinez Canillas

Vikash Garodia <[email protected]> writes:

Hello Vikash,

Quoted 2 lines

Hi All,

[...]

Quoted 11 lines
Quoted 5 lines

No reply from Mauro and Linus chose to not apply the revert I pointed him to. That at this point leads to the question:

Vikash, did you or somebody else make any progress to fix this properly?

We tried with different settings for the registers and arrive at a conclusion that the original configuration was proper. There is no need to explicitly configure the secure non-pixel region when there is no support for the usecase. So, in summary, we are good to have the revert.

Perfect. Thanks a lot for looking at this.

Quoted 8 lines
Quoted 3 lines

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Other fix posted is "media: venus: dec: Fix capture formats enumeration order":

https://patchwork.kernel.org/project/linux-media/patch/[email protected]/

Vikash,

Could you or someone else from QC please Review/Ack these two patches, since it seems that Stanimir moved on and maybe is not working in this driver anymore?

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

#18 lore raw

Dikshita Agarwal <[email protected]> writes:

Hello Dikshita,

Quoted 32 lines · Dikshita Agarwal
Quoted 22 lines · Javier Martinez Canillas
Quoted 19 lines
Quoted 17 lines · Javier Martinez Canillas

Vikash Garodia<[email protected]> writes:

Hello Vikash,

Quoted 2 lines

Hi All,

[...]

Quoted 9 lines
Quoted 4 lines

No reply from Mauro and Linus chose to not apply the revert I pointed him to. That at this point leads to the question:

Vikash, did you or somebody else make any progress to fix this properly?

We tried with different settings for the registers and arrive at a conclusion that the original configuration was proper. There is no need to explicitly configure the secure non-pixel region when there is no support for the usecase. So, in summary, we are good to have the revert.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Perfect. Thanks a lot for looking at this.

Quoted 3 lines

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Other fix posted is "media: venus: dec: Fix capture formats enumeration order":

https://patchwork.kernel.org/project/linux-media/patch/[email protected]/

Quoted 9 lines

Hi Javier,

Thanks for this patch "media: venus: dec: Fix capture formats enumeration order".

Somehow I can't find it in my mailbox to be able to reply there.

Could you please explain what is the regression you see here?

You can find the thread and explanation of the issue here:

https://lore.kernel.org/lkml/Y+KPW18o%[email protected]/T/

But Stanimir already picked it and sent a PR for v6.3 including it.

Signature · 5 lines
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat
#19 lore raw

Hi Javier, Dikshita, Stan,

the revert wasn't applied to v6.2 series. Can you please apply it and include it for v6.2.10?

Quoted 19 lines · Javier Martinez Canillas
Quoted 19 lines · Dikshita Agarwal
Quoted 10 lines · Javier Martinez Canillas
Quoted 9 lines
Quoted 8 lines · Javier Martinez Canillas
Quoted 3 lines · Vikash Garodia

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Other fix posted is "media: venus: dec: Fix capture formats enumeration order":

https://patchwork.kernel.org/project/linux-media/patch/[email protected]/

Hi Javier,

Thanks for this patch "media: venus: dec: Fix capture formats enumeration order".

Somehow I can't find it in my mailbox to be able to reply there.

Quoted 8 lines
Quoted 2 lines

Could you please explain what is the regression you see here?

You can find the thread and explanation of the issue here:

https://lore.kernel.org/lkml/Y+KPW18o%[email protected]/T/

But Stanimir already picked it and sent a PR for v6.3 including it.

While "media: venus: dec: Fix capture formats enumeration order" may have been applied to v6.3, this still leaves the regression introduced by "venus:
firmware: Correct non-pix start and end addresses". As pointed out by Matthias Kaehlcke, the commit prevents SC7180 and sc7280 AOSS from entering sleep mode during system suspend. This is a serious regression in v6.2 kernel series.

Best regards,
Leonard Lausen

#20 lore raw
Quoted 27 lines · Leonard Lausen

Hi Javier, Dikshita, Stan,

the revert wasn't applied to v6.2 series. Can you please apply it and include it for v6.2.10?

Quoted 22 lines · Javier Martinez Canillas
Quoted 21 lines · Dikshita Agarwal
Quoted 10 lines · Javier Martinez Canillas
Quoted 9 lines
Quoted 8 lines · Javier Martinez Canillas
Quoted 3 lines · Vikash Garodia

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Other fix posted is "media: venus: dec: Fix capture formats enumeration order":

https://patchwork.kernel.org/project/linux-media/patch/[email protected]/

Hi Javier,

Thanks for this patch "media: venus: dec: Fix capture formats enumeration order".

Somehow I can't find it in my mailbox to be able to reply there.

Could you please explain what is the regression you see here?

Quoted 8 lines
Quoted 5 lines

You can find the thread and explanation of the issue here:

https://lore.kernel.org/lkml/Y+KPW18o%[email protected]/T/

But Stanimir already picked it and sent a PR for v6.3 including it.

While "media: venus: dec: Fix capture formats enumeration order" may have been applied to v6.3,

To me it looks like it was submitted[1], but not yet applied even to the media tree[2] -- while guess, maybe due problems mentioned in[3]? Or am I missing something?

[1]
https://lore.kernel.org/all/[email protected]/
[2] https://git.linuxtv.org/media_tree.git/log/?h=fixes
[3]
https://lore.kernel.org/all/[email protected]/

Quoted 4 lines

this still leaves the regression introduced by "venus:
firmware: Correct non-pix start and end addresses". As pointed out by Matthias Kaehlcke, the commit prevents SC7180 and sc7280 AOSS from entering sleep mode during system suspend. This is a serious regression in v6.2 kernel series.

That fix is sitting in the media tree for a while and afaics still hasn't been sent to Linus (which is needed to get this fixed in 6.2.y).

Mauro, could you maybe take care of that?

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#21 lore raw
Quoted 8 lines · April 2, 2023 at 1:02 AM, <[email protected]> wrote:
Quoted 5 lines

this still leaves the regression introduced by "venus:
firmware: Correct non-pix start and end addresses". As pointed out by Matthias Kaehlcke, the commit prevents SC7180 and sc7280 AOSS from entering sleep mode during system suspend. This is a serious regression in v6.2 kernel series.

That fix is sitting in the media tree for a while and afaics still hasn't been sent to Linus (which is needed to get this fixed in 6.2.y). Mauro, could you maybe take care of that?

I see the revert made it to 6.3-rc5 as commit f95b8ea7. Now it just needs to be included for v6.2.10

Thank you
Leonard Lausen

#22 lore raw
In reply to Linux regression tracking (Thorsten Leemhuis)
Quoted 3 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 2 lines · Leonard Lausen

the revert wasn't applied to v6.2 series. Can you please apply it and include it for v6.2.10?

I pointed Linus to this and he merged the revert directly; and it's already queued for the next 6.2.y release:

https://lore.kernel.org/all/CAHk-%3DwhRs_MavKCqtV3%3DK31dq9Z6HzbaG8Uxo-EV%[email protected]/
https://git.kernel.org/torvalds/c/f95b8ea79c47c0ad3d18f45ad538f9970e414d1f
https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/commit/?id=902f9eb696dfdd40e88d99bafa34ea25f1f9e927

Now to the remaining venus regression:

Quoted 36 lines
Quoted 23 lines
Quoted 22 lines · Javier Martinez Canillas
Quoted 21 lines · Dikshita Agarwal
Quoted 10 lines · Javier Martinez Canillas
Quoted 9 lines
Quoted 8 lines · Javier Martinez Canillas
Quoted 3 lines · Vikash Garodia

Stan, could you please help with the revert and a pull request having this revert alongwith other pending changes ?

Other fix posted is "media: venus: dec: Fix capture formats enumeration order":

https://patchwork.kernel.org/project/linux-media/patch/[email protected]/

Hi Javier,

Thanks for this patch "media: venus: dec: Fix capture formats enumeration order".

Somehow I can't find it in my mailbox to be able to reply there.

Could you please explain what is the regression you see here?

Quoted 8 lines
Quoted 5 lines

You can find the thread and explanation of the issue here:

https://lore.kernel.org/lkml/Y+KPW18o%[email protected]/T/

But Stanimir already picked it and sent a PR for v6.3 including it.

While "media: venus: dec: Fix capture formats enumeration order" may have been applied to v6.3,

To me it looks like it was submitted[1], but not yet applied even to the media tree[2] -- while guess, maybe due problems mentioned in[3]? Or am I missing something?

Quoted 5 lines

[1]
https://lore.kernel.org/all/[email protected]/
[2] https://git.linuxtv.org/media_tree.git/log/?h=fixes
[3]
https://lore.kernel.org/all/[email protected]/

I only notice now: from [1] above it looks like that regression fix was applied to a tree that seems to be intended for 6.4. Is that okay for everybody, or should we ask Linus to pick this up as well (unless of course Mauro shows up and forwards the patch, of course)? They fix a regression from 5.19 afaics, so not a fresh problem, but apparently one that bugged a few people recently.

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.