* [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing
@ 2026-09-22 3:09 Jie Gan
2026-10-13 12:38 ` Konrad Dybcio
0 siblings, 1 reply; 5+ messages in thread
From: Jie Gan @ 2026-09-22 3:09 UTC (permalink / raw)
To: Bjorn Andersson, Konrad Dybcio, Abel Vesa, Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Jie Gan
Cc: Konrad Dybcio, linux-arm-msm, devicetree, linux-kernel, Jie Gan
The etm0-etm3 nodes are declared with "arm,coresight-etm4x-sysreg"
but do not implement self-hosted system-register trace access on
this SoC: ID_AA64DFR0_EL1.TraceVer reads as 0 on the affected CPUs,
so etm4_init_sysreg_access() correctly reports the access mechanism
as unavailable and the ETMv4 driver fails to probe with -EINVAL.
The CPU ETMs have a memory-mapped aperture, so switch each node to
the standard "arm,coresight-etm4x", "arm,primecell" compatible with
its MMIO reg range.
Fixes: eed738c6dd0c ("arm64: dts: qcom: shikra: Add coresight nodes")
Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
---
Signed-off-by: Jie Gan <jiegan@qti.qualcomm.com>
---
arch/arm64/boot/dts/qcom/shikra.dtsi | 140 ++++++++++++++++++-----------------
1 file changed, 72 insertions(+), 68 deletions(-)
diff --git a/arch/arm64/boot/dts/qcom/shikra.dtsi b/arch/arm64/boot/dts/qcom/shikra.dtsi
index c1f06f692f66..62bacf749241 100644
--- a/arch/arm64/boot/dts/qcom/shikra.dtsi
+++ b/arch/arm64/boot/dts/qcom/shikra.dtsi
@@ -164,74 +164,6 @@ eud_in: endpoint {
};
};
- etm0 {
- compatible = "arm,coresight-etm4x-sysreg";
-
- clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
- clock-names = "apb_pclk";
- cpu = <&cpu0>;
- qcom,skip-power-up;
-
- out-ports {
- port {
- etm0_out: endpoint {
- remote-endpoint = <&funnel_cpuss0_in0>;
- };
- };
- };
- };
-
- etm1 {
- compatible = "arm,coresight-etm4x-sysreg";
-
- clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
- clock-names = "apb_pclk";
- cpu = <&cpu1>;
- qcom,skip-power-up;
-
- out-ports {
- port {
- etm1_out: endpoint {
- remote-endpoint = <&funnel_cpuss0_in1>;
- };
- };
- };
- };
-
- etm2 {
- compatible = "arm,coresight-etm4x-sysreg";
-
- clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
- clock-names = "apb_pclk";
- cpu = <&cpu2>;
- qcom,skip-power-up;
-
- out-ports {
- port {
- etm2_out: endpoint {
- remote-endpoint = <&funnel_cpuss0_in2>;
- };
- };
- };
- };
-
- etm3 {
- compatible = "arm,coresight-etm4x-sysreg";
-
- clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
- clock-names = "apb_pclk";
- cpu = <&cpu3>;
- qcom,skip-power-up;
-
- out-ports {
- port {
- etm3_out: endpoint {
- remote-endpoint = <&funnel_cpuss0_in3>;
- };
- };
- };
- };
-
firmware {
scm {
compatible = "qcom,scm-shikra", "qcom,scm";
@@ -3218,6 +3150,24 @@ cti@9020000 {
label = "cti_apss_pe0";
};
+ etm@9040000 {
+ compatible = "arm,coresight-etm4x", "arm,primecell";
+ reg = <0x0 0x09040000 0x0 0x1000>;
+
+ clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
+ clock-names = "apb_pclk";
+ cpu = <&cpu0>;
+ qcom,skip-power-up;
+
+ out-ports {
+ port {
+ etm0_out: endpoint {
+ remote-endpoint = <&funnel_cpuss0_in0>;
+ };
+ };
+ };
+ };
+
cti@90e0000 {
compatible = "arm,coresight-cti", "arm,primecell";
reg = <0x0 0x090e0000 0x0 0x1000>;
@@ -3236,6 +3186,24 @@ cti@9120000 {
label = "cti_apss_pe1";
};
+ etm@9140000 {
+ compatible = "arm,coresight-etm4x", "arm,primecell";
+ reg = <0x0 0x09140000 0x0 0x1000>;
+
+ clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
+ clock-names = "apb_pclk";
+ cpu = <&cpu1>;
+ qcom,skip-power-up;
+
+ out-ports {
+ port {
+ etm1_out: endpoint {
+ remote-endpoint = <&funnel_cpuss0_in1>;
+ };
+ };
+ };
+ };
+
cti@9220000 {
compatible = "arm,coresight-cti", "arm,primecell";
reg = <0x0 0x09220000 0x0 0x1000>;
@@ -3245,6 +3213,24 @@ cti@9220000 {
label = "cti_apss_pe2";
};
+ etm@9240000 {
+ compatible = "arm,coresight-etm4x", "arm,primecell";
+ reg = <0x0 0x09240000 0x0 0x1000>;
+
+ clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
+ clock-names = "apb_pclk";
+ cpu = <&cpu2>;
+ qcom,skip-power-up;
+
+ out-ports {
+ port {
+ etm2_out: endpoint {
+ remote-endpoint = <&funnel_cpuss0_in2>;
+ };
+ };
+ };
+ };
+
cti@9320000 {
compatible = "arm,coresight-cti", "arm,primecell";
reg = <0x0 0x09320000 0x0 0x1000>;
@@ -3254,6 +3240,24 @@ cti@9320000 {
label = "cti_apss_pe3";
};
+ etm@9340000 {
+ compatible = "arm,coresight-etm4x", "arm,primecell";
+ reg = <0x0 0x09340000 0x0 0x1000>;
+
+ clocks = <&rpmcc RPM_SMD_QDSS_CLK>;
+ clock-names = "apb_pclk";
+ cpu = <&cpu3>;
+ qcom,skip-power-up;
+
+ out-ports {
+ port {
+ etm3_out: endpoint {
+ remote-endpoint = <&funnel_cpuss0_in3>;
+ };
+ };
+ };
+ };
+
funnel@9800000 {
compatible = "arm,coresight-dynamic-funnel", "arm,primecell";
reg = <0x0 0x09800000 0x0 0x1000>;
---
base-commit: 5c4d4169604b335c38bbc79bc1fc03042981fc6f
change-id: 20260922-fix-etm-probe-issue-on-shikra-5a624f583975
Best regards,
--
Jie Gan <jie.gan@oss.qualcomm.com>
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing
2026-09-22 3:09 [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing Jie Gan
@ 2026-10-13 12:38 ` Konrad Dybcio
2026-10-14 8:19 ` Jie Gan
0 siblings, 1 reply; 5+ messages in thread
From: Konrad Dybcio @ 2026-10-13 12:38 UTC (permalink / raw)
To: Jie Gan, Bjorn Andersson, Konrad Dybcio, Abel Vesa, Rob Herring,
Krzysztof Kozlowski, Conor Dooley
Cc: linux-arm-msm, devicetree, linux-kernel, Jie Gan
On 9/22/26 5:09 AM, Jie Gan wrote:
> The etm0-etm3 nodes are declared with "arm,coresight-etm4x-sysreg"
> but do not implement self-hosted system-register trace access on
> this SoC: ID_AA64DFR0_EL1.TraceVer reads as 0 on the affected CPUs,
> so etm4_init_sysreg_access() correctly reports the access mechanism
> as unavailable and the ETMv4 driver fails to probe with -EINVAL.
>
> The CPU ETMs have a memory-mapped aperture, so switch each node to
> the standard "arm,coresight-etm4x", "arm,primecell" compatible with
> its MMIO reg range.
>
> Fixes: eed738c6dd0c ("arm64: dts: qcom: shikra: Add coresight nodes")
> Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
> ---
Do we still need qcom,skip-power-up on new SoCs *with* MMIO access?
GPT has concerns regarding whether we need to add
arm,coresight-loses-context-with-cpu
otherwise lgtm
Konrad
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing
2026-10-13 12:38 ` Konrad Dybcio
@ 2026-10-14 8:19 ` Jie Gan
2026-10-14 8:31 ` Konrad Dybcio
0 siblings, 1 reply; 5+ messages in thread
From: Jie Gan @ 2026-10-14 8:19 UTC (permalink / raw)
To: Konrad Dybcio, Bjorn Andersson, Konrad Dybcio, Abel Vesa,
Rob Herring, Krzysztof Kozlowski, Conor Dooley
Cc: linux-arm-msm, devicetree, linux-kernel, Jie Gan
On 10/13/2026 8:38 PM, Konrad Dybcio wrote:
> On 9/22/26 5:09 AM, Jie Gan wrote:
>> The etm0-etm3 nodes are declared with "arm,coresight-etm4x-sysreg"
>> but do not implement self-hosted system-register trace access on
>> this SoC: ID_AA64DFR0_EL1.TraceVer reads as 0 on the affected CPUs,
>> so etm4_init_sysreg_access() correctly reports the access mechanism
>> as unavailable and the ETMv4 driver fails to probe with -EINVAL.
>>
>> The CPU ETMs have a memory-mapped aperture, so switch each node to
>> the standard "arm,coresight-etm4x", "arm,primecell" compatible with
>> its MMIO reg range.
>>
>> Fixes: eed738c6dd0c ("arm64: dts: qcom: shikra: Add coresight nodes")
>> Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
>> ---
>
> Do we still need qcom,skip-power-up on new SoCs *with* MMIO access?
>
Yes, we need qcom,skip-power-up for all ETM device with or w/o MMIO access.
> GPT has concerns regarding whether we need to add
> arm,coresight-loses-context-with-cpu>
This logic to be handled by HYP on Shikra SoC, so I didnt enable it in
kernel.
This property related to etm register save and restore during pm_notify.
Thanks,
Jie
> otherwise lgtm
>
> Konrad
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing
2026-10-14 8:19 ` Jie Gan
@ 2026-10-14 8:31 ` Konrad Dybcio
2026-10-14 8:33 ` Jie Gan
0 siblings, 1 reply; 5+ messages in thread
From: Konrad Dybcio @ 2026-10-14 8:31 UTC (permalink / raw)
To: Jie Gan, Bjorn Andersson, Konrad Dybcio, Abel Vesa, Rob Herring,
Krzysztof Kozlowski, Conor Dooley
Cc: linux-arm-msm, devicetree, linux-kernel, Jie Gan
On 10/14/26 10:19 AM, Jie Gan wrote:
>
>
> On 10/13/2026 8:38 PM, Konrad Dybcio wrote:
>> On 9/22/26 5:09 AM, Jie Gan wrote:
>>> The etm0-etm3 nodes are declared with "arm,coresight-etm4x-sysreg"
>>> but do not implement self-hosted system-register trace access on
>>> this SoC: ID_AA64DFR0_EL1.TraceVer reads as 0 on the affected CPUs,
>>> so etm4_init_sysreg_access() correctly reports the access mechanism
>>> as unavailable and the ETMv4 driver fails to probe with -EINVAL.
>>>
>>> The CPU ETMs have a memory-mapped aperture, so switch each node to
>>> the standard "arm,coresight-etm4x", "arm,primecell" compatible with
>>> its MMIO reg range.
>>>
>>> Fixes: eed738c6dd0c ("arm64: dts: qcom: shikra: Add coresight nodes")
>>> Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
>>> ---
>>
>> Do we still need qcom,skip-power-up on new SoCs *with* MMIO access?
>>
>
> Yes, we need qcom,skip-power-up for all ETM device with or w/o MMIO access.
>
>> GPT has concerns regarding whether we need to add
>> arm,coresight-loses-context-with-cpu>
>
> This logic to be handled by HYP on Shikra SoC, so I didnt enable it in kernel.
https://lore.kernel.org/linux-arm-msm/20261005-shikra-el2-v3-0-618f7343695c@oss.qualcomm.com/
We're going to, in parallel, ship Shikra without the Qualcomm
hypervisor so please keep that in mind. Should the submitters of the
non-Gunyah overlay add these properties there as they respin? Does
that concern all Coresight components?
Konrad
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing
2026-10-14 8:31 ` Konrad Dybcio
@ 2026-10-14 8:33 ` Jie Gan
0 siblings, 0 replies; 5+ messages in thread
From: Jie Gan @ 2026-10-14 8:33 UTC (permalink / raw)
To: Konrad Dybcio, Bjorn Andersson, Konrad Dybcio, Abel Vesa,
Rob Herring, Krzysztof Kozlowski, Conor Dooley
Cc: linux-arm-msm, devicetree, linux-kernel, Jie Gan
On 10/14/2026 4:31 PM, Konrad Dybcio wrote:
> On 10/14/26 10:19 AM, Jie Gan wrote:
>>
>>
>> On 10/13/2026 8:38 PM, Konrad Dybcio wrote:
>>> On 9/22/26 5:09 AM, Jie Gan wrote:
>>>> The etm0-etm3 nodes are declared with "arm,coresight-etm4x-sysreg"
>>>> but do not implement self-hosted system-register trace access on
>>>> this SoC: ID_AA64DFR0_EL1.TraceVer reads as 0 on the affected CPUs,
>>>> so etm4_init_sysreg_access() correctly reports the access mechanism
>>>> as unavailable and the ETMv4 driver fails to probe with -EINVAL.
>>>>
>>>> The CPU ETMs have a memory-mapped aperture, so switch each node to
>>>> the standard "arm,coresight-etm4x", "arm,primecell" compatible with
>>>> its MMIO reg range.
>>>>
>>>> Fixes: eed738c6dd0c ("arm64: dts: qcom: shikra: Add coresight nodes")
>>>> Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
>>>> ---
>>>
>>> Do we still need qcom,skip-power-up on new SoCs *with* MMIO access?
>>>
>>
>> Yes, we need qcom,skip-power-up for all ETM device with or w/o MMIO access.
>>
>>> GPT has concerns regarding whether we need to add
>>> arm,coresight-loses-context-with-cpu>
>>
>> This logic to be handled by HYP on Shikra SoC, so I didnt enable it in kernel.
>
> https://lore.kernel.org/linux-arm-msm/20261005-shikra-el2-v3-0-618f7343695c@oss.qualcomm.com/
>
> We're going to, in parallel, ship Shikra without the Qualcomm
> hypervisor so please keep that in mind. Should the submitters of the
> non-Gunyah overlay add these properties there as they respin? Does
> that concern all Coresight components?
Making sense.
I will add the property.
Thanks,
Jie
>
> Konrad
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2026-10-14 8:33 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-09-22 3:09 [PATCH] arm64: dts: qcom: shikra: Switch CPU ETMs to MMIO-based probing Jie Gan
2026-10-13 12:38 ` Konrad Dybcio
2026-10-14 8:19 ` Jie Gan
2026-10-14 8:31 ` Konrad Dybcio
2026-10-14 8:33 ` Jie Gan
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®