[v3] Add explicit clock vote and enable power-domain for QCOM-ICE

[PATCH v3 02/12] dt-bindings: crypto: qcom,ice: Require power-domain and iface clk

Posted by Harshal Dev 2 weeks, 6 days ago

Update the DT bindings for inline-crypto engine to require the power-domain
and iface clock for Eliza and Milos.

If the 'clk_ignore_unused' flag is not passed on the kernel command line,
the unused 'iface' clock could be disabled by the kernel before ICE can
probe. This leads to unclocked ICE hardware register accces being observed
during ICE driver probe. On the other hand, If the 'pd_ignore_unused' flag
is not passed on the kernel command line, the unused UFS_PHY_GDSC power
domain could be disabled by the kernel before ICE probes. This results in
a 'stuck' clock issue being observed when ICE attempts to enable the
'core' clock.

Therefore, both the 'iface' clock and the UFS_PHY_GDSC power domain are
mandatory resources for ICE which must be specified in the device tree
node.

Fixes: 618195a7ac3df ("dt-bindings: crypto: qcom,inline-crypto-engine: Document the Eliza ICE")
Fixes: 85faec1e85555 ("dt-bindings: crypto: qcom,inline-crypto-engine: document the Milos ICE")
Signed-off-by: Harshal Dev <harshal.dev@oss.qualcomm.com>
---
 .../bindings/crypto/qcom,inline-crypto-engine.yaml    | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/Documentation/devicetree/bindings/crypto/qcom,inline-crypto-engine.yaml b/Documentation/devicetree/bindings/crypto/qcom,inline-crypto-engine.yaml
index 99c541e7fa8c..ccb6b8dd8e11 100644
--- a/Documentation/devicetree/bindings/crypto/qcom,inline-crypto-engine.yaml
+++ b/Documentation/devicetree/bindings/crypto/qcom,inline-crypto-engine.yaml
@@ -54,6 +54,25 @@ required:
 
 additionalProperties: false
 
+allOf:
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - qcom,eliza-inline-crypto-engine
+              - qcom,milos-inline-crypto-engine
+
+    then:
+      required:
+        - power-domains
+        - clock-names
+      properties:
+        clocks:
+          minItems: 2
+        clock-names:
+          minItems: 2
+
 examples:
   - |
     #include <dt-bindings/clock/qcom,sm8550-gcc.h>

-- 
2.34.1

Re: [PATCH v3 02/12] dt-bindings: crypto: qcom,ice: Require power-domain and iface clk

Posted by Krzysztof Kozlowski 2 weeks, 5 days ago

On Tue, Mar 17, 2026 at 02:50:41PM +0530, Harshal Dev wrote:
> Update the DT bindings for inline-crypto engine to require the power-domain
> and iface clock for Eliza and Milos.
> 

NAK, pointless patch. It makes no sense on its own and it cannot be a
fix for earlier issues, because as a fix standalone is simply WRONG.

Where did you emphasize this is the fix for current RC?

Best regards,
Krzysztof

Re: [PATCH v3 02/12] dt-bindings: crypto: qcom,ice: Require power-domain and iface clk

Posted by Harshal Dev 2 weeks, 5 days ago


On 3/18/2026 12:53 PM, Krzysztof Kozlowski wrote:
> On Tue, Mar 17, 2026 at 02:50:41PM +0530, Harshal Dev wrote:
>> Update the DT bindings for inline-crypto engine to require the power-domain
>> and iface clock for Eliza and Milos.
>>
> 
> NAK, pointless patch. It makes no sense on its own and it cannot be a
> fix for earlier issues, because as a fix standalone is simply WRONG.
> 
> Where did you emphasize this is the fix for current RC?

Ack, I will add explicit emphasis that this is a fix for the current RC in
the commit message after merging this patch into the previous one.

Regards,
Harshal

> 
> Best regards,
> Krzysztof
> 
>

Re: [PATCH v3 02/12] dt-bindings: crypto: qcom,ice: Require power-domain and iface clk

Posted by Dmitry Baryshkov 2 weeks, 6 days ago

On Tue, Mar 17, 2026 at 02:50:41PM +0530, Harshal Dev wrote:
> Update the DT bindings for inline-crypto engine to require the power-domain
> and iface clock for Eliza and Milos.

Again, this mostly duplicates the subject (and your last paragraph).
Either drop it or move it there.

> 
> If the 'clk_ignore_unused' flag is not passed on the kernel command line,
> the unused 'iface' clock could be disabled by the kernel before ICE can
> probe. This leads to unclocked ICE hardware register accces being observed
> during ICE driver probe. On the other hand, If the 'pd_ignore_unused' flag
> is not passed on the kernel command line, the unused UFS_PHY_GDSC power
> domain could be disabled by the kernel before ICE probes. This results in
> a 'stuck' clock issue being observed when ICE attempts to enable the
> 'core' clock.

What's the difference from the previous patch?

> 
> Therefore, both the 'iface' clock and the UFS_PHY_GDSC power domain are
> mandatory resources for ICE which must be specified in the device tree
> node.
> 
> Fixes: 618195a7ac3df ("dt-bindings: crypto: qcom,inline-crypto-engine: Document the Eliza ICE")
> Fixes: 85faec1e85555 ("dt-bindings: crypto: qcom,inline-crypto-engine: document the Milos ICE")
> Signed-off-by: Harshal Dev <harshal.dev@oss.qualcomm.com>
> ---
>  .../bindings/crypto/qcom,inline-crypto-engine.yaml    | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
> 

-- 
With best wishes
Dmitry

Re: [PATCH v3 02/12] dt-bindings: crypto: qcom,ice: Require power-domain and iface clk

Posted by Krzysztof Kozlowski 2 weeks, 5 days ago

On 17/03/2026 16:13, Dmitry Baryshkov wrote:
> On Tue, Mar 17, 2026 at 02:50:41PM +0530, Harshal Dev wrote:
>> Update the DT bindings for inline-crypto engine to require the power-domain
>> and iface clock for Eliza and Milos.
> 
> Again, this mostly duplicates the subject (and your last paragraph).
> Either drop it or move it there.
> 
>>
>> If the 'clk_ignore_unused' flag is not passed on the kernel command line,
>> the unused 'iface' clock could be disabled by the kernel before ICE can
>> probe. This leads to unclocked ICE hardware register accces being observed
>> during ICE driver probe. On the other hand, If the 'pd_ignore_unused' flag
>> is not passed on the kernel command line, the unused UFS_PHY_GDSC power
>> domain could be disabled by the kernel before ICE probes. This results in
>> a 'stuck' clock issue being observed when ICE attempts to enable the
>> 'core' clock.
> 
> What's the difference from the previous patch?

There is no, it's just completely missing the point of development cycle
and ignoring multiple of my feedbacks.

I asked to POST A PATCH targeting current RC. One patch. Current RC, not
next.

How is this supposed to be merged without the first one? How do anyone
here imagine this is supposed to work when applying to the tree?

Best regards,
Krzysztof

Re: [PATCH v3 02/12] dt-bindings: crypto: qcom,ice: Require power-domain and iface clk

Posted by Harshal Dev 2 weeks, 5 days ago


On 3/18/2026 12:57 PM, Krzysztof Kozlowski wrote:
> On 17/03/2026 16:13, Dmitry Baryshkov wrote:
>> On Tue, Mar 17, 2026 at 02:50:41PM +0530, Harshal Dev wrote:
>>> Update the DT bindings for inline-crypto engine to require the power-domain
>>> and iface clock for Eliza and Milos.
>>
>> Again, this mostly duplicates the subject (and your last paragraph).
>> Either drop it or move it there.
>>
>>>
>>> If the 'clk_ignore_unused' flag is not passed on the kernel command line,
>>> the unused 'iface' clock could be disabled by the kernel before ICE can
>>> probe. This leads to unclocked ICE hardware register accces being observed
>>> during ICE driver probe. On the other hand, If the 'pd_ignore_unused' flag
>>> is not passed on the kernel command line, the unused UFS_PHY_GDSC power
>>> domain could be disabled by the kernel before ICE probes. This results in
>>> a 'stuck' clock issue being observed when ICE attempts to enable the
>>> 'core' clock.
>>
>> What's the difference from the previous patch?
> 
> 
> I asked to POST A PATCH targeting current RC. One patch. Current RC, not
> next.

Ack, I will merge the changes in this patch into the previous patch to ensure we
have one consolidated fix for the current RC.

Regards,
Harshal

> 
> 
> Best regards,
> Krzysztof