[PATCH v5 00/10] Add CMDQ secure driver for SVP

[PATCH v5 00/10] Add CMDQ secure driver for SVP

[Shawn Sung]

From: Hsiao Chien Sung <shawn.sung@mediatek.corp-partner.google.com>

For the Secure Video Path (SVP) feature, inculding the memory stored
secure video content, the registers of display HW pipeline and the
HW configure operations are required to execute in the secure world.

So using a CMDQ secure driver to make all display HW registers
configuration secure DRAM access permision settings execute by GCE
secure thread in the secure world.

We are landing this feature on mt8188 and mt8195 currently.
---
Based on 2 series and 1 patch:
[1] Add CMDQ driver support for mt8188
- https://patchwork.kernel.org/project/linux-mediatek/list/?series=810382
[2] Add mediatek,gce-events definition to mediatek,gce-mailbox bindings
- https://patchwork.kernel.org/project/linux-mediatek/list/?series=810938
[3] soc: mediatek: Add register definitions for GCE
- https://patchwork.kernel.org/project/linux-mediatek/patch/20231017064717.21616-2-shawn.sung@mediatek.com/
---

Changes in v5:
1. Sync the local changes

Changes in v4:
1. Rebase on mediatek-drm-next(278640d4d74cd) and fix the conflicts
2. This series is based on 20240307013458.23550-1-jason-jh.lin@mediatek.com

Changes in v3:
1. separate mt8188 driver porting patches to another series
2. separate adding 'mediatek,gce-events' event prop to another series
3. sepatate mailbox helper and controller driver modification to a
   single patch for adding looping thread
4. add kerneldoc for secure mailbox related definition
5. add moving reuseable definition patch before adding secure mailbox
   driver patch
6. adjust redundant logic in mtk-cmdq-sec-mailbox

Changes in v2:
1. adjust dt-binding SW event define patch before the dt-binding patch using it
2. adjust dt-binding patch for secure cmdq driver
3. remove the redundant patches or merge the patches of modification for the same API

CK Hu (1):
  drm/mediatek: Add interface to allocate MediaTek GEM buffer.

Jason-JH.Lin (9):
  dt-bindings: gce: mt8195: Add CMDQ_SYNC_TOKEN_SECURE_THR_EOF event id
  dt-bindings: mailbox: Add mboxes property for CMDQ secure driver
  soc: mediatek: cmdq: Add cmdq_pkt_logic_command to support math
    operation
  soc: mediatek: cmdq: Add cmdq_pkt_write_s_reg_value to support write
    value to reg
  mailbox: mtk-cmdq: Support GCE loop packets in interrupt handler
  soc: mediatek: cmdq: Add cmdq_pkt_finalize_loop for looping cmd with
    irq
  mailbox: mediatek: Move reuseable definition to header for secure
    driver
  mailbox: mediatek: Add CMDQ secure mailbox driver
  mailbox: mediatek: Add secure CMDQ driver support for CMDQ driver

 .../mailbox/mediatek,gce-mailbox.yaml         |   10 +
 drivers/gpu/drm/mediatek/mtk_drm_drv.c        |   16 +-
 drivers/gpu/drm/mediatek/mtk_gem.c            |   40 +
 drivers/gpu/drm/mediatek/mtk_gem.h            |   11 +
 drivers/mailbox/Makefile                      |    2 +-
 drivers/mailbox/mtk-cmdq-mailbox.c            |  108 +-
 drivers/mailbox/mtk-cmdq-sec-mailbox.c        | 1045 +++++++++++++++++
 drivers/mailbox/mtk-cmdq-sec-tee.c            |  165 +++
 drivers/soc/mediatek/mtk-cmdq-helper.c        |   75 ++
 include/dt-bindings/gce/mt8195-gce.h          |    6 +
 include/linux/mailbox/mtk-cmdq-mailbox.h      |   37 +
 .../linux/mailbox/mtk-cmdq-sec-iwc-common.h   |  385 ++++++
 include/linux/mailbox/mtk-cmdq-sec-mailbox.h  |  159 +++
 include/linux/mailbox/mtk-cmdq-sec-tee.h      |  105 ++
 include/linux/soc/mediatek/mtk-cmdq.h         |   61 +
 include/uapi/drm/mediatek_drm.h               |   64 +
 16 files changed, 2257 insertions(+), 32 deletions(-)
 create mode 100644 drivers/mailbox/mtk-cmdq-sec-mailbox.c
 create mode 100644 drivers/mailbox/mtk-cmdq-sec-tee.c
 create mode 100644 include/linux/mailbox/mtk-cmdq-sec-iwc-common.h
 create mode 100644 include/linux/mailbox/mtk-cmdq-sec-mailbox.h
 create mode 100644 include/linux/mailbox/mtk-cmdq-sec-tee.h
 create mode 100644 include/uapi/drm/mediatek_drm.h

--
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 01/10] dt-bindings: gce: mt8195: Add CMDQ_SYNC_TOKEN_SECURE_THR_EOF event id

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

There are 2 kind of GCE event signal:
- The SW token means: a GCE event signal triggered by SW drivers.
e.g. SW driver append a GCE command to set a GCE event after a specific
GCE command. Or SW driver use CPU to write a event id to GCE register to
trigger the GCE event corresponding to that event id.

- The HW event means: a GCE event signal triggered by HW engines.
e.g. When HW OVL fetches all the data in frame buffer, HW OVL will send
a frame done irq and also send a frame done GCE event via HW bus directly.

CMDQ_SYNC_TOKEN_SECURE_THR_EOF is a SW token event that is set in the
end of each cmdq secure pkt. It is used as a secure irq to notify
CMDQ driver in the normal world that GCE secure thread has completed
a secure cmd buffer in thee secure world.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 include/dt-bindings/gce/mt8195-gce.h | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/include/dt-bindings/gce/mt8195-gce.h b/include/dt-bindings/gce/mt8195-gce.h
index dcfb302b8a5bc..b6b3db82c381e 100644
--- a/include/dt-bindings/gce/mt8195-gce.h
+++ b/include/dt-bindings/gce/mt8195-gce.h
@@ -800,6 +800,12 @@
 #define CMDQ_EVENT_WPE_VPP1_WPE_GCE_FRAME_DONE	969
 #define CMDQ_EVENT_WPE_VPP1_WPE_DONE_SYNC_OUT	970
 
+/*
+ * Notify normal CMDQ there are some secure task done,
+ * this token sync with secure world.
+ */
+#define CMDQ_SYNC_TOKEN_SECURE_THR_EOF			980
+
 #define CMDQ_EVENT_DP_TX_VBLANK_FALLING	994
 #define CMDQ_EVENT_DP_TX_VSC_FINISH	995
 
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

Add mboxes to define a GCE loopping thread as a secure irq handler.
This property is only required if CMDQ secure driver is supported.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml
index cef9d76013985..c0d80cc770899 100644
--- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml
+++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml
@@ -49,6 +49,16 @@ properties:
     items:
       - const: gce
 
+  mediatek,gce-events:
+    description:
+      The event id which is mapping to the specific hardware event signal
+      to gce. The event id is defined in the gce header
+      include/dt-bindings/gce/<chip>-gce.h of each chips.
+    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
+
+  mboxes:
+    maxItems: 1
+
 required:
   - compatible
   - "#mbox-cells"
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Rob Herring]

On Wed, 03 Apr 2024 18:25:54 +0800, Shawn Sung wrote:
> From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> 
> Add mboxes to define a GCE loopping thread as a secure irq handler.
> This property is only required if CMDQ secure driver is supported.
> 
> Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> ---
>  .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 

My bot found errors running 'make DT_CHECKER_FLAGS=-m dt_binding_check'
on your patch (DT_CHECKER_FLAGS is new in v5.13):

yamllint warnings/errors:

dtschema/dtc warnings/errors:
/builds/robherring/dt-review-ci/linux/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml:
Unresolvable JSON pointer: 'definitions/uint32-arrayi'

doc reference errors (make refcheckdocs):

See https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20240403102602.32155-3-shawn.sung@mediatek.com

The base for the series is generally the latest rc1. A different dependency
should be noted in *this* patch.

If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure 'yamllint' is installed and dt-schema is up to
date:

pip3 install dtschema --upgrade

Please check and re-submit after running the above command yourself. Note
that DT_SCHEMA_FILES can be set to your schema file to speed up checking
your schema. However, it must be unset to test all examples with your schema.


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Conor Dooley]

[-- Attachment #1.1: Type: text/plain, Size: 1430 bytes --]

On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> 
> Add mboxes to define a GCE loopping thread as a secure irq handler.
> This property is only required if CMDQ secure driver is supported.
> 
> Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> ---
>  .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml
> index cef9d76013985..c0d80cc770899 100644
> --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml
> +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml
> @@ -49,6 +49,16 @@ properties:
>      items:
>        - const: gce
>  
> +  mediatek,gce-events:
> +    description:
> +      The event id which is mapping to the specific hardware event signal
> +      to gce. The event id is defined in the gce header
> +      include/dt-bindings/gce/<chip>-gce.h of each chips.

Missing any info here about when this should be used, hint - you have it
in the commit message.

> +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi

Why is the ID used by the CMDQ service not fixed for each SoC?

Cheers,
Conor

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Jason-JH Lin (林睿祥)]

Hi Conor,

Thanks for the reviews.

On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > 
> > Add mboxes to define a GCE loopping thread as a secure irq handler.
> > This property is only required if CMDQ secure driver is supported.
> > 
> > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> > ---
> >  .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10
> > ++++++++++
> >  1 file changed, 10 insertions(+)
> > 
> > diff --git
> > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > mailbox.yaml
> > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > mailbox.yaml
> > index cef9d76013985..c0d80cc770899 100644
> > --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > mailbox.yaml
> > +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > mailbox.yaml
> > @@ -49,6 +49,16 @@ properties:
> >      items:
> >        - const: gce
> >  
> > +  mediatek,gce-events:
> > +    description:
> > +      The event id which is mapping to the specific hardware event
> > signal
> > +      to gce. The event id is defined in the gce header
> > +      include/dt-bindings/gce/<chip>-gce.h of each chips.
> 
> Missing any info here about when this should be used, hint - you have
> it
> in the commit message.
> 
> > +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
> 
> Why is the ID used by the CMDQ service not fixed for each SoC?
> 
I forgot to sync with Shawn about this:
https://lore.kernel.org/all/20240124011459.12204-1-jason-
jh.lin@mediatek.com

I'll fix it at the next version.

Regards,
Jason-JH.Lin

> Cheers,
> Conor
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Conor Dooley]

[-- Attachment #1.1: Type: text/plain, Size: 2253 bytes --]

On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥) wrote:
> Hi Conor,
> 
> Thanks for the reviews.
> 
> On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > > 
> > > Add mboxes to define a GCE loopping thread as a secure irq handler.
> > > This property is only required if CMDQ secure driver is supported.
> > > 
> > > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > > Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> > > ---
> > >  .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10
> > > ++++++++++
> > >  1 file changed, 10 insertions(+)
> > > 
> > > diff --git
> > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > mailbox.yaml
> > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > mailbox.yaml
> > > index cef9d76013985..c0d80cc770899 100644
> > > --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > mailbox.yaml
> > > +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > mailbox.yaml
> > > @@ -49,6 +49,16 @@ properties:
> > >      items:
> > >        - const: gce
> > >  
> > > +  mediatek,gce-events:
> > > +    description:
> > > +      The event id which is mapping to the specific hardware event
> > > signal
> > > +      to gce. The event id is defined in the gce header
> > > +      include/dt-bindings/gce/<chip>-gce.h of each chips.
> > 
> > Missing any info here about when this should be used, hint - you have
> > it
> > in the commit message.
> > 
> > > +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
> > 
> > Why is the ID used by the CMDQ service not fixed for each SoC?
> > 
> I forgot to sync with Shawn about this:
> https://lore.kernel.org/all/20240124011459.12204-1-jason-
> jh.lin@mediatek.com
> 
> I'll fix it at the next version.

When I say "fixed" I don't mean "this is wrong, please fix it", I mean
"why is the value not static for a particular SoC". This needs to be
explained in the patch (and the description for the event here needs to
explain what the gce-mailbox is reserving an event for).

Thanks,
Conor.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Jason-JH Lin (林睿祥)]

On Thu, 2024-04-04 at 15:52 +0100, Conor Dooley wrote:
> On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥) wrote:
> > Hi Conor,
> > 
> > Thanks for the reviews.
> > 
> > On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> > > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > > > 
> > > > Add mboxes to define a GCE loopping thread as a secure irq
> > > > handler.
> > > > This property is only required if CMDQ secure driver is
> > > > supported.
> > > > 
> > > > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > > > Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> > > > ---
> > > >  .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10
> > > > ++++++++++
> > > >  1 file changed, 10 insertions(+)
> > > > 
> > > > diff --git
> > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > mailbox.yaml
> > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > mailbox.yaml
> > > > index cef9d76013985..c0d80cc770899 100644
> > > > --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > mailbox.yaml
> > > > +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > mailbox.yaml
> > > > @@ -49,6 +49,16 @@ properties:
> > > >      items:
> > > >        - const: gce
> > > >  
> > > > +  mediatek,gce-events:
> > > > +    description:
> > > > +      The event id which is mapping to the specific hardware
> > > > event
> > > > signal
> > > > +      to gce. The event id is defined in the gce header
> > > > +      include/dt-bindings/gce/<chip>-gce.h of each chips.
> > > 
> > > Missing any info here about when this should be used, hint - you
> > > have
> > > it
> > > in the commit message.
> > > 
> > > > +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
> > > 
> > > Why is the ID used by the CMDQ service not fixed for each SoC?
> > > 
> > 
> > I forgot to sync with Shawn about this:
> > https://lore.kernel.org/all/20240124011459.12204-1-jason-
> > jh.lin@mediatek.com
> > 
> > I'll fix it at the next version.
> 
> When I say "fixed" I don't mean "this is wrong, please fix it", I
> mean
> "why is the value not static for a particular SoC". This needs to be
> explained in the patch (and the description for the event here needs
> to
> explain what the gce-mailbox is reserving an event for).
> 
Oh, I see. Thanks for noticing me.

We do want to reserve a static event ID for gce-mailbox to different
SoCs. There are 2 mainly reasons to why we set it in DTS:
1. There are 1024 events IDs for GCE to use to execute instructions in
the specific event happened. These events could be signaled by HW or SW
and their value would be different in different SoC because of HW event
IDs distribution range from 0 to 1023.
If we set a static event ID: 855 for mt8188, it might be conflict the
event ID original set in mt8195.

2. If we defined the event ID in DTS, we might know how many SW or HW
event IDs are used.
If someone wants to use a new event ID for a new feature, they could
find out the used event IDs in DTS easily and avoid the event ID
conflicting.

The reason why we define a event ID is we want to get a SW signal from
secure world. We design a GCE looping thread in gce-mailbox driver to
wait for the GCE execute done event for each cmdq secure packets from
secure world.

Regards,
Jason-JH.Lin

> Thanks,
> Conor.
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Conor Dooley]

[-- Attachment #1.1: Type: text/plain, Size: 4393 bytes --]

On Fri, Apr 05, 2024 at 02:33:14PM +0000, Jason-JH Lin (林睿祥) wrote:
> On Thu, 2024-04-04 at 15:52 +0100, Conor Dooley wrote:
> > On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥) wrote:
> > > Hi Conor,
> > > 
> > > Thanks for the reviews.
> > > 
> > > On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > > > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> > > > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > > > > 
> > > > > Add mboxes to define a GCE loopping thread as a secure irq
> > > > > handler.
> > > > > This property is only required if CMDQ secure driver is
> > > > > supported.
> > > > > 
> > > > > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > > > > Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> > > > > ---
> > > > >  .../bindings/mailbox/mediatek,gce-mailbox.yaml         | 10
> > > > > ++++++++++
> > > > >  1 file changed, 10 insertions(+)
> > > > > 
> > > > > diff --git
> > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > mailbox.yaml
> > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > mailbox.yaml
> > > > > index cef9d76013985..c0d80cc770899 100644
> > > > > --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > mailbox.yaml
> > > > > +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > mailbox.yaml
> > > > > @@ -49,6 +49,16 @@ properties:
> > > > >      items:
> > > > >        - const: gce
> > > > >  
> > > > > +  mediatek,gce-events:
> > > > > +    description:
> > > > > +      The event id which is mapping to the specific hardware
> > > > > event
> > > > > signal
> > > > > +      to gce. The event id is defined in the gce header
> > > > > +      include/dt-bindings/gce/<chip>-gce.h of each chips.
> > > > 
> > > > Missing any info here about when this should be used, hint - you
> > > > have
> > > > it
> > > > in the commit message.
> > > > 
> > > > > +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
> > > > 
> > > > Why is the ID used by the CMDQ service not fixed for each SoC?
> > > > 
> > > 
> > > I forgot to sync with Shawn about this:
> > > https://lore.kernel.org/all/20240124011459.12204-1-jason-
> > > jh.lin@mediatek.com
> > > 
> > > I'll fix it at the next version.
> > 
> > When I say "fixed" I don't mean "this is wrong, please fix it", I
> > mean
> > "why is the value not static for a particular SoC". This needs to be
> > explained in the patch (and the description for the event here needs
> > to
> > explain what the gce-mailbox is reserving an event for).
> > 
> Oh, I see. Thanks for noticing me.
> 
> We do want to reserve a static event ID for gce-mailbox to different
> SoCs. There are 2 mainly reasons to why we set it in DTS:
> 1. There are 1024 events IDs for GCE to use to execute instructions in
> the specific event happened. These events could be signaled by HW or SW
> and their value would be different in different SoC because of HW event
> IDs distribution range from 0 to 1023.
> If we set a static event ID: 855 for mt8188, it might be conflict the
> event ID original set in mt8195.

That's not a problem, we have compatibles for this purpose.

> 2. If we defined the event ID in DTS, we might know how many SW or HW
> event IDs are used.
> If someone wants to use a new event ID for a new feature, they could
> find out the used event IDs in DTS easily and avoid the event ID
> conflicting.

Are the event IDs not documented in the reference manual for the SoC in
question? Or in documentation for the secure world for these devices? A
DTS should not be the authoritive source for this information for
developers.

Additionally, the driver could very easily detect if someone does happen
to put in the reserved ID. That could be generically useful (IOW, check
all of them for re-use) if the ID are to not allowed to be shared.

> The reason why we define a event ID is we want to get a SW signal from
> secure world. We design a GCE looping thread in gce-mailbox driver to
> wait for the GCE execute done event for each cmdq secure packets from
> secure world.

This sort of information needs to be in the commit message, but I don't
think this property is needed at all since it seems to be something
detectable from the compatible.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Jason-JH Lin (林睿祥)]

On Fri, 2024-04-05 at 17:13 +0100, Conor Dooley wrote:
> On Fri, Apr 05, 2024 at 02:33:14PM +0000, Jason-JH Lin (林睿祥) wrote:
> > On Thu, 2024-04-04 at 15:52 +0100, Conor Dooley wrote:
> > > On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥)
> > > wrote:
> > > > Hi Conor,
> > > > 
> > > > Thanks for the reviews.
> > > > 
> > > > On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > > > > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> > > > > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > > > > > 
> > > > > > Add mboxes to define a GCE loopping thread as a secure irq
> > > > > > handler.
> > > > > > This property is only required if CMDQ secure driver is
> > > > > > supported.
> > > > > > 
> > > > > > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > > > > > Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> > > > > > ---
> > > > > >  .../bindings/mailbox/mediatek,gce-mailbox.yaml         |
> > > > > > 10
> > > > > > ++++++++++
> > > > > >  1 file changed, 10 insertions(+)
> > > > > > 
> > > > > > diff --git
> > > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > mailbox.yaml
> > > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > mailbox.yaml
> > > > > > index cef9d76013985..c0d80cc770899 100644
> > > > > > ---
> > > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > mailbox.yaml
> > > > > > +++
> > > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > mailbox.yaml
> > > > > > @@ -49,6 +49,16 @@ properties:
> > > > > >      items:
> > > > > >        - const: gce
> > > > > >  
> > > > > > +  mediatek,gce-events:
> > > > > > +    description:
> > > > > > +      The event id which is mapping to the specific
> > > > > > hardware
> > > > > > event
> > > > > > signal
> > > > > > +      to gce. The event id is defined in the gce header
> > > > > > +      include/dt-bindings/gce/<chip>-gce.h of each chips.
> > > > > 
> > > > > Missing any info here about when this should be used, hint -
> > > > > you
> > > > > have
> > > > > it
> > > > > in the commit message.
> > > > > 
> > > > > > +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
> > > > > 
> > > > > Why is the ID used by the CMDQ service not fixed for each
> > > > > SoC?
> > > > > 
> > > > 
> > > > I forgot to sync with Shawn about this:
> > > > https://lore.kernel.org/all/20240124011459.12204-1-jason-
> > > > jh.lin@mediatek.com
> > > > 
> > > > I'll fix it at the next version.
> > > 
> > > When I say "fixed" I don't mean "this is wrong, please fix it", I
> > > mean
> > > "why is the value not static for a particular SoC". This needs to
> > > be
> > > explained in the patch (and the description for the event here
> > > needs
> > > to
> > > explain what the gce-mailbox is reserving an event for).
> > > 
> > 
> > Oh, I see. Thanks for noticing me.
> > 
> > We do want to reserve a static event ID for gce-mailbox to
> > different
> > SoCs. There are 2 mainly reasons to why we set it in DTS:
> > 1. There are 1024 events IDs for GCE to use to execute instructions
> > in
> > the specific event happened. These events could be signaled by HW
> > or SW
> > and their value would be different in different SoC because of HW
> > event
> > IDs distribution range from 0 to 1023.
> > If we set a static event ID: 855 for mt8188, it might be conflict
> > the
> > event ID original set in mt8195.
> 
> That's not a problem, we have compatibles for this purpose.

I agree that compatibles can do the same things.

> 
> > 2. If we defined the event ID in DTS, we might know how many SW or
> > HW
> > event IDs are used.
> > If someone wants to use a new event ID for a new feature, they
> > could
> > find out the used event IDs in DTS easily and avoid the event ID
> > conflicting.
> 
> Are the event IDs not documented in the reference manual for the SoC
> in
> question? Or in documentation for the secure world for these devices?
> A
> DTS should not be the authoritive source for this information for
> developers.
> 
The event IDs were defined in:
inculde/dt-bindings/mailbox/mediatek,mt8188-gce.h.

> Additionally, the driver could very easily detect if someone does
> happen
> to put in the reserved ID. That could be generically useful (IOW,
> check
> all of them for re-use) if the ID are to not allowed to be shared.
> 
> > The reason why we define a event ID is we want to get a SW signal
> > from
> > secure world. We design a GCE looping thread in gce-mailbox driver
> > to
> > wait for the GCE execute done event for each cmdq secure packets
> > from
> > secure world.
> 
> This sort of information needs to be in the commit message, but I
> don't
> think this property is needed at all since it seems to be something
> detectable from the compatible.

I think put this event ID in driver data and distinguish them by
different compatibles can achieve the same thing.

However, I originally thought that align to the existing way like
MUTEX, CCORR, WDMA in 
https://lore.kernel.org/all/20240124011459.12204-4-jason-jh.lin@mediatek.com
 would be better choice.
I think their usage of gce-events are the same. 

What do you think?

Regards,
Jason-JH.Lin
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Conor Dooley]

[-- Attachment #1.1: Type: text/plain, Size: 7164 bytes --]

On Sat, Apr 06, 2024 at 04:15:51PM +0000, Jason-JH Lin (林睿祥) wrote:
> On Fri, 2024-04-05 at 17:13 +0100, Conor Dooley wrote:
> > On Fri, Apr 05, 2024 at 02:33:14PM +0000, Jason-JH Lin (林睿祥) wrote:
> > > On Thu, 2024-04-04 at 15:52 +0100, Conor Dooley wrote:
> > > > On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥)
> > > > wrote:
> > > > > Hi Conor,
> > > > > 
> > > > > Thanks for the reviews.
> > > > > 
> > > > > On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > > > > > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung wrote:
> > > > > > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > > > > > > 
> > > > > > > Add mboxes to define a GCE loopping thread as a secure irq
> > > > > > > handler.
> > > > > > > This property is only required if CMDQ secure driver is
> > > > > > > supported.
> > > > > > > 
> > > > > > > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > > > > > > Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
> > > > > > > ---
> > > > > > >  .../bindings/mailbox/mediatek,gce-mailbox.yaml         |
> > > > > > > 10
> > > > > > > ++++++++++
> > > > > > >  1 file changed, 10 insertions(+)
> > > > > > > 
> > > > > > > diff --git
> > > > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > > mailbox.yaml
> > > > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > > mailbox.yaml
> > > > > > > index cef9d76013985..c0d80cc770899 100644
> > > > > > > ---
> > > > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > > mailbox.yaml
> > > > > > > +++
> > > > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gce-
> > > > > > > mailbox.yaml
> > > > > > > @@ -49,6 +49,16 @@ properties:
> > > > > > >      items:
> > > > > > >        - const: gce
> > > > > > >  
> > > > > > > +  mediatek,gce-events:
> > > > > > > +    description:
> > > > > > > +      The event id which is mapping to the specific
> > > > > > > hardware
> > > > > > > event
> > > > > > > signal
> > > > > > > +      to gce. The event id is defined in the gce header
> > > > > > > +      include/dt-bindings/gce/<chip>-gce.h of each chips.
> > > > > > 
> > > > > > Missing any info here about when this should be used, hint -
> > > > > > you
> > > > > > have
> > > > > > it
> > > > > > in the commit message.
> > > > > > 
> > > > > > > +    $ref: /schemas/types.yaml#/definitions/uint32-arrayi
> > > > > > 
> > > > > > Why is the ID used by the CMDQ service not fixed for each
> > > > > > SoC?
> > > > > > 
> > > > > 
> > > > > I forgot to sync with Shawn about this:
> > > > > https://lore.kernel.org/all/20240124011459.12204-1-jason-
> > > > > jh.lin@mediatek.com
> > > > > 
> > > > > I'll fix it at the next version.
> > > > 
> > > > When I say "fixed" I don't mean "this is wrong, please fix it", I
> > > > mean
> > > > "why is the value not static for a particular SoC". This needs to
> > > > be
> > > > explained in the patch (and the description for the event here
> > > > needs
> > > > to
> > > > explain what the gce-mailbox is reserving an event for).
> > > > 
> > > 
> > > Oh, I see. Thanks for noticing me.
> > > 
> > > We do want to reserve a static event ID for gce-mailbox to
> > > different
> > > SoCs. There are 2 mainly reasons to why we set it in DTS:
> > > 1. There are 1024 events IDs for GCE to use to execute instructions
> > > in
> > > the specific event happened. These events could be signaled by HW
> > > or SW
> > > and their value would be different in different SoC because of HW
> > > event
> > > IDs distribution range from 0 to 1023.
> > > If we set a static event ID: 855 for mt8188, it might be conflict
> > > the
> > > event ID original set in mt8195.
> > 
> > That's not a problem, we have compatibles for this purpose.
> 
> I agree that compatibles can do the same things.
> 
> > 
> > > 2. If we defined the event ID in DTS, we might know how many SW or
> > > HW
> > > event IDs are used.
> > > If someone wants to use a new event ID for a new feature, they
> > > could
> > > find out the used event IDs in DTS easily and avoid the event ID
> > > conflicting.
> > 
> > Are the event IDs not documented in the reference manual for the SoC
> > in
> > question? Or in documentation for the secure world for these devices?
> > A
> > DTS should not be the authoritive source for this information for
> > developers.
> > 
> The event IDs were defined in:
> inculde/dt-bindings/mailbox/mediatek,mt8188-gce.h.
> 
> > Additionally, the driver could very easily detect if someone does
> > happen
> > to put in the reserved ID. That could be generically useful (IOW,
> > check
> > all of them for re-use) if the ID are to not allowed to be shared.
> > 
> > > The reason why we define a event ID is we want to get a SW signal
> > > from
> > > secure world. We design a GCE looping thread in gce-mailbox driver
> > > to
> > > wait for the GCE execute done event for each cmdq secure packets
> > > from
> > > secure world.
> > 
> > This sort of information needs to be in the commit message, but I
> > don't
> > think this property is needed at all since it seems to be something
> > detectable from the compatible.
> 
> I think put this event ID in driver data and distinguish them by
> different compatibles can achieve the same thing.
> 
> However, I originally thought that align to the existing way like
> MUTEX, CCORR, WDMA in 
> https://lore.kernel.org/all/20240124011459.12204-4-jason-jh.lin@mediatek.com
>  would be better choice.
> I think their usage of gce-events are the same. 
> 
> What do you think?

To me it comes down to whether the IDs are fixed on a particular SoC (in
which case they can be deduced by the compatible) or not. I don't really
see how this is actually a fixed property of the SoC though, if you came
along tomorrow with a "gce-2.0" you could totally end up with different
numbering because (as far as I can tell) this numbering is actually a
property of the os-firmware interface, not actually a property of the
SoC itself. I was expecting you to say "no" when I asked if the IDs were
fixed for a given SoC because changing the firmware /could/ change the IDs.
Although, I think you'd likely not ever want to change them, because
that'd just be an annoying ABI break to deal with.

What I think is that you need to write a property description that
explains what the mailbox is using the gce channel for so that someone
can populate the property correctly. The commit message also needs to
explain why this is not a fixed value for a given SoC.

And yes, as you pointed out earlier in this thread, Shawn needs to
update this to have a reference to the gce-events binding which has a
great description in it of what a gce event is.

> I think their usage of gce-events are the same. 

Also, just because a property got introduced doesn't mean that it is
correct and adding new instances can definitely by required to provide
justification, not just saying "it's used by xyz too".

Cheers,
Conor.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Jason-JH Lin (林睿祥)]

On Tue, 2024-04-09 at 18:52 +0100, Conor Dooley wrote:
> On Sat, Apr 06, 2024 at 04:15:51PM +0000, Jason-JH Lin (林睿祥) wrote:
> > On Fri, 2024-04-05 at 17:13 +0100, Conor Dooley wrote:
> > > On Fri, Apr 05, 2024 at 02:33:14PM +0000, Jason-JH Lin (林睿祥)
> > > wrote:
> > > > On Thu, 2024-04-04 at 15:52 +0100, Conor Dooley wrote:
> > > > > On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥)
> > > > > wrote:
> > > > > > Hi Conor,
> > > > > > 
> > > > > > Thanks for the reviews.
> > > > > > 
> > > > > > On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > > > > > > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung
> > > > > > > wrote:
> > > > > > > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>
> > > > > > > > 
> > > > > > > > Add mboxes to define a GCE loopping thread as a secure
> > > > > > > > irq
> > > > > > > > handler.
> > > > > > > > This property is only required if CMDQ secure driver is
> > > > > > > > supported.
> > > > > > > > 
> > > > > > > > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > > > > > > > Signed-off-by: Hsiao Chien Sung <
> > > > > > > > shawn.sung@mediatek.com>
> > > > > > > > ---
> > > > > > > >  .../bindings/mailbox/mediatek,gce-
> > > > > > > > mailbox.yaml         |
> > > > > > > > 10
> > > > > > > > ++++++++++
> > > > > > > >  1 file changed, 10 insertions(+)
> > > > > > > > 
> > > > > > > > diff --git
> > > > > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gc
> > > > > > > > e-
> > > > > > > > mailbox.yaml
> > > > > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gc
> > > > > > > > e-
> > > > > > > > mailbox.yaml
> > > > > > > > index cef9d76013985..c0d80cc770899 100644
> > > > > > > > ---
> > > > > > > > a/Documentation/devicetree/bindings/mailbox/mediatek,gc
> > > > > > > > e-
> > > > > > > > mailbox.yaml
> > > > > > > > +++
> > > > > > > > b/Documentation/devicetree/bindings/mailbox/mediatek,gc
> > > > > > > > e-
> > > > > > > > mailbox.yaml
> > > > > > > > @@ -49,6 +49,16 @@ properties:
> > > > > > > >      items:
> > > > > > > >        - const: gce
> > > > > > > >  
> > > > > > > > +  mediatek,gce-events:
> > > > > > > > +    description:
> > > > > > > > +      The event id which is mapping to the specific
> > > > > > > > hardware
> > > > > > > > event
> > > > > > > > signal
> > > > > > > > +      to gce. The event id is defined in the gce
> > > > > > > > header
> > > > > > > > +      include/dt-bindings/gce/<chip>-gce.h of each
> > > > > > > > chips.
> > > > > > > 
> > > > > > > Missing any info here about when this should be used,
> > > > > > > hint -
> > > > > > > you
> > > > > > > have
> > > > > > > it
> > > > > > > in the commit message.
> > > > > > > 
> > > > > > > > +    $ref: /schemas/types.yaml#/definitions/uint32-
> > > > > > > > arrayi
> > > > > > > 
> > > > > > > Why is the ID used by the CMDQ service not fixed for each
> > > > > > > SoC?

Did I misunderstand the ID here?
I thought we were talking about event IDs, but it looks like we are
talking about mbox IDs.

> > > > > > > 
> > > > > > 
> > > > > > I forgot to sync with Shawn about this:
> > > > > > https://lore.kernel.org/all/20240124011459.12204-1-jason-
> > > > > > jh.lin@mediatek.com
> > > > > > 
> > > > > > I'll fix it at the next version.
> > > > > 
> > > > > When I say "fixed" I don't mean "this is wrong, please fix
> > > > > it", I
> > > > > mean
> > > > > "why is the value not static for a particular SoC". This
> > > > > needs to
> > > > > be
> > > > > explained in the patch (and the description for the event
> > > > > here
> > > > > needs
> > > > > to
> > > > > explain what the gce-mailbox is reserving an event for).
> > > > > 
> > > > 
> > > > Oh, I see. Thanks for noticing me.
> > > > 
> > > > We do want to reserve a static event ID for gce-mailbox to
> > > > different
> > > > SoCs. There are 2 mainly reasons to why we set it in DTS:
> > > > 1. There are 1024 events IDs for GCE to use to execute
> > > > instructions
> > > > in
> > > > the specific event happened. These events could be signaled by
> > > > HW
> > > > or SW
> > > > and their value would be different in different SoC because of
> > > > HW
> > > > event
> > > > IDs distribution range from 0 to 1023.
> > > > If we set a static event ID: 855 for mt8188, it might be
> > > > conflict
> > > > the
> > > > event ID original set in mt8195.
> > > 
> > > That's not a problem, we have compatibles for this purpose.
> > 
> > I agree that compatibles can do the same things.
> > 
> > > 
> > > > 2. If we defined the event ID in DTS, we might know how many SW
> > > > or
> > > > HW
> > > > event IDs are used.
> > > > If someone wants to use a new event ID for a new feature, they
> > > > could
> > > > find out the used event IDs in DTS easily and avoid the event
> > > > ID
> > > > conflicting.
> > > 
> > > Are the event IDs not documented in the reference manual for the
> > > SoC
> > > in
> > > question? Or in documentation for the secure world for these
> > > devices?
> > > A
> > > DTS should not be the authoritive source for this information for
> > > developers.
> > > 
> > 
> > The event IDs were defined in:
> > inculde/dt-bindings/mailbox/mediatek,mt8188-gce.h.
> > 
> > > Additionally, the driver could very easily detect if someone does
> > > happen
> > > to put in the reserved ID. That could be generically useful (IOW,
> > > check
> > > all of them for re-use) if the ID are to not allowed to be
> > > shared.
> > > 
> > > > The reason why we define a event ID is we want to get a SW
> > > > signal
> > > > from
> > > > secure world. We design a GCE looping thread in gce-mailbox
> > > > driver
> > > > to
> > > > wait for the GCE execute done event for each cmdq secure
> > > > packets
> > > > from
> > > > secure world.
> > > 
> > > This sort of information needs to be in the commit message, but I
> > > don't
> > > think this property is needed at all since it seems to be
> > > something
> > > detectable from the compatible.
> > 
> > I think put this event ID in driver data and distinguish them by
> > different compatibles can achieve the same thing.
> > 
> > However, I originally thought that align to the existing way like
> > MUTEX, CCORR, WDMA in 
> > 
https://lore.kernel.org/all/20240124011459.12204-4-jason-jh.lin@mediatek.com
> >  would be better choice.
> > I think their usage of gce-events are the same. 
> > 
> > What do you think?
> 
> To me it comes down to whether the IDs are fixed on a particular SoC
> (in
> which case they can be deduced by the compatible) or not. I don't
> really
> see how this is actually a fixed property of the SoC though, if you
> came
> along tomorrow with a "gce-2.0" you could totally end up with
> different
> numbering because (as far as I can tell) this numbering is actually a
> property of the os-firmware interface, not actually a property of the
> SoC itself. I was expecting you to say "no" when I asked if the IDs
> were
> fixed for a given SoC because changing the firmware /could/ change
> the IDs.
> Although, I think you'd likely not ever want to change them, because
> that'd just be an annoying ABI break to deal with.
> 

mbox IDs are "not" fixed property of the SoC.


Some of the event IDs(HW event) are fixed of the SoC and some of the
event IDs are not fixed of the SoC.
So we need to change the non-fixed event IDs to make sure they won't
conflict to the fixed event IDs of another SoC.

> What I think is that you need to write a property description that
> explains what the mailbox is using the gce channel for so that
> someone
> can populate the property correctly. The commit message also needs to
> explain why this is not a fixed value for a given SoC.
> 

Totally agree!

> And yes, as you pointed out earlier in this thread, Shawn needs to
> update this to have a reference to the gce-events binding which has a
> great description in it of what a gce event is.
> 
> > I think their usage of gce-events are the same. 
> 
> Also, just because a property got introduced doesn't mean that it is
> correct and adding new instances can definitely by required to
> provide
> justification, not just saying "it's used by xyz too".
> 

Sure, I think we need to add more explanation for what we use a mbox
channel for.

I think the main reason is we want to request a mbox channel for
sending a looping command to GCE thread. The looping command will get 
secure task done event signal every time from secure world and notify 
CMDQ driver to do the same operation as the IRQ handler.

This can reduce the latency of software communication between secure
world and we can also remove the complex logic after every secure task
done in the secure world.

Regards,
Jason-JH.Lin

> Cheers,
> Conor.
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v5 02/10] dt-bindings: mailbox: Add mboxes property for CMDQ secure driver

[Conor Dooley]

[-- Attachment #1.1: Type: text/plain, Size: 2074 bytes --]

On Fri, Apr 12, 2024 at 09:00:53AM +0000, Jason-JH Lin (林睿祥) wrote:
> On Tue, 2024-04-09 at 18:52 +0100, Conor Dooley wrote:
> > On Sat, Apr 06, 2024 at 04:15:51PM +0000, Jason-JH Lin (林睿祥) wrote:
> > > On Fri, 2024-04-05 at 17:13 +0100, Conor Dooley wrote:
> > > > On Fri, Apr 05, 2024 at 02:33:14PM +0000, Jason-JH Lin (林睿祥)
> > > > wrote:
> > > > > On Thu, 2024-04-04 at 15:52 +0100, Conor Dooley wrote:
> > > > > > On Thu, Apr 04, 2024 at 04:31:06AM +0000, Jason-JH Lin (林睿祥)
> > > > > > wrote:
> > > > > > > Hi Conor,
> > > > > > > 
> > > > > > > Thanks for the reviews.
> > > > > > > 
> > > > > > > On Wed, 2024-04-03 at 16:46 +0100, Conor Dooley wrote:
> > > > > > > > On Wed, Apr 03, 2024 at 06:25:54PM +0800, Shawn Sung
> > > > > > > > wrote:
> > > > > > > > > From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

> > > > > > > > > +  mediatek,gce-events:
> > > > > > > > > +    description:
> > > > > > > > > +      The event id which is mapping to the specific
> > > > > > > > > hardware
> > > > > > > > > event
> > > > > > > > > signal
> > > > > > > > > +      to gce. The event id is defined in the gce
> > > > > > > > > header
> > > > > > > > > +      include/dt-bindings/gce/<chip>-gce.h of each
> > > > > > > > > chips.
> > > > > > > > 
> > > > > > > > Missing any info here about when this should be used,
> > > > > > > > hint -
> > > > > > > > you
> > > > > > > > have
> > > > > > > > it
> > > > > > > > in the commit message.
> > > > > > > > 
> > > > > > > > > +    $ref: /schemas/types.yaml#/definitions/uint32-
> > > > > > > > > arrayi
> > > > > > > > 
> > > > > > > > Why is the ID used by the CMDQ service not fixed for each
> > > > > > > > SoC?
> 
> Did I misunderstand the ID here?
> I thought we were talking about event IDs, but it looks like we are
> talking about mbox IDs.

We were talking about the event IDs FWIW.

Just send a new version with some actual explanations added (as we
discussed earlier on this thread) and we should be good here I think.

Cheers,
Conor.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 03/10] soc: mediatek: cmdq: Add cmdq_pkt_logic_command to support math operation

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

Add cmdq_pkt_logic_command to support math operation.

cmdq_pkt_logic_command can append logic command to the CMDQ packet,
ask GCE to execute a arithmetic calculate instruction,
such as add, subtract, multiply, AND, OR and NOT, etc.

Note that all arithmetic instructions are unsigned calculations.
If there are any overflows, GCE will sent the invalid IRQ to notify
CMDQ driver.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/soc/mediatek/mtk-cmdq-helper.c | 39 ++++++++++++++++++++++++
 include/linux/soc/mediatek/mtk-cmdq.h  | 42 ++++++++++++++++++++++++++
 2 files changed, 81 insertions(+)

diff --git a/drivers/soc/mediatek/mtk-cmdq-helper.c b/drivers/soc/mediatek/mtk-cmdq-helper.c
index 8acd8e38283e0..72350138591c3 100644
--- a/drivers/soc/mediatek/mtk-cmdq-helper.c
+++ b/drivers/soc/mediatek/mtk-cmdq-helper.c
@@ -15,9 +15,21 @@
 /* dedicate the last GPR_R15 to assign the register address to be poll */
 #define CMDQ_POLL_ADDR_GPR	(15)
 #define CMDQ_EOC_IRQ_EN		BIT(0)
+#define CMDQ_IMMEDIATE_VALUE	0
 #define CMDQ_REG_TYPE		1
 #define CMDQ_JUMP_RELATIVE	1
 
+#define CMDQ_GET_ARG_B(arg)		(((arg) & GENMASK(31, 16)) >> 16)
+#define CMDQ_GET_ARG_C(arg)		((arg) & GENMASK(15, 0))
+
+#define CMDQ_OPERAND_GET_IDX_VALUE(operand) \
+	({ \
+		struct cmdq_operand *op = operand; \
+		op->reg ? op->idx : op->value; \
+	})
+#define CMDQ_OPERAND_TYPE(operand) \
+	((operand)->reg ? CMDQ_REG_TYPE : CMDQ_IMMEDIATE_VALUE)
+
 struct cmdq_instruction {
 	union {
 		u32 value;
@@ -470,6 +482,33 @@ int cmdq_pkt_poll_addr(struct cmdq_pkt *pkt, dma_addr_t addr, u32 value, u32 mas
 }
 EXPORT_SYMBOL(cmdq_pkt_poll_addr);
 
+int cmdq_pkt_logic_command(struct cmdq_pkt *pkt, u16 result_reg_idx,
+			   struct cmdq_operand *left_operand,
+			   enum cmdq_logic_op s_op,
+			   struct cmdq_operand *right_operand)
+{
+	struct cmdq_instruction inst = { {0} };
+	u32 left_idx_value;
+	u32 right_idx_value;
+
+	if (!left_operand || !right_operand || s_op >= CMDQ_LOGIC_MAX)
+		return -EINVAL;
+
+	left_idx_value = CMDQ_OPERAND_GET_IDX_VALUE(left_operand);
+	right_idx_value = CMDQ_OPERAND_GET_IDX_VALUE(right_operand);
+	inst.op = CMDQ_CODE_LOGIC;
+	inst.dst_t = CMDQ_REG_TYPE;
+	inst.src_t = CMDQ_OPERAND_TYPE(left_operand);
+	inst.arg_c_t = CMDQ_OPERAND_TYPE(right_operand);
+	inst.sop = s_op;
+	inst.reg_dst = right_idx_value;
+	inst.src_reg = left_idx_value;
+	inst.arg_c = result_reg_idx;
+
+	return cmdq_pkt_append_command(pkt, inst);
+}
+EXPORT_SYMBOL(cmdq_pkt_logic_command);
+
 int cmdq_pkt_assign(struct cmdq_pkt *pkt, u16 reg_idx, u32 value)
 {
 	struct cmdq_instruction inst = {};
diff --git a/include/linux/soc/mediatek/mtk-cmdq.h b/include/linux/soc/mediatek/mtk-cmdq.h
index f708bcfebdd8a..1edb391ec604a 100644
--- a/include/linux/soc/mediatek/mtk-cmdq.h
+++ b/include/linux/soc/mediatek/mtk-cmdq.h
@@ -25,6 +25,31 @@
 
 struct cmdq_pkt;
 
+enum cmdq_logic_op {
+	CMDQ_LOGIC_ASSIGN = 0,
+	CMDQ_LOGIC_ADD = 1,
+	CMDQ_LOGIC_SUBTRACT = 2,
+	CMDQ_LOGIC_MULTIPLY = 3,
+	CMDQ_LOGIC_XOR = 8,
+	CMDQ_LOGIC_NOT = 9,
+	CMDQ_LOGIC_OR = 10,
+	CMDQ_LOGIC_AND = 11,
+	CMDQ_LOGIC_LEFT_SHIFT = 12,
+	CMDQ_LOGIC_RIGHT_SHIFT = 13,
+	CMDQ_LOGIC_MAX,
+};
+
+struct cmdq_operand {
+	/* register type */
+	bool reg;
+	union {
+		/* index */
+		u16 idx;
+		/* value */
+		u16 value;
+	};
+};
+
 struct cmdq_client_reg {
 	u8 subsys;
 	u16 offset;
@@ -286,6 +311,23 @@ int cmdq_pkt_poll_mask(struct cmdq_pkt *pkt, u8 subsys,
  */
 int cmdq_pkt_poll_addr(struct cmdq_pkt *pkt, dma_addr_t addr, u32 value, u32 mask);
 
+/**
+ * cmdq_pkt_logic_command() - Append logic command to the CMDQ packet, ask GCE to
+ *		          execute an instruction that store the result of logic operation
+ *		          with left and right operand into result_reg_idx.
+ * @pkt:		the CMDQ packet
+ * @result_reg_idx:	SPR index that store operation result of left_operand and right_operand
+ * @left_operand:	left operand
+ * @s_op:		the logic operator enum
+ * @right_operand:	right operand
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_pkt_logic_command(struct cmdq_pkt *pkt, u16 result_reg_idx,
+			   struct cmdq_operand *left_operand,
+			   enum cmdq_logic_op s_op,
+			   struct cmdq_operand *right_operand);
+
 /**
  * cmdq_pkt_assign() - Append logic assign command to the CMDQ packet, ask GCE
  *		       to execute an instruction that set a constant value into
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 04/10] soc: mediatek: cmdq: Add cmdq_pkt_write_s_reg_value to support write value to reg

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

Add cmdq_pkt_write_s_reg_value to support write a value to a register.

It appends write_s command to the command buffer in a CMDQ packet,
ask GCE to excute a write instruction to write a value to a register
with low 16 bits physical address offset.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/soc/mediatek/mtk-cmdq-helper.c | 13 +++++++++++++
 include/linux/soc/mediatek/mtk-cmdq.h  | 11 +++++++++++
 2 files changed, 24 insertions(+)

diff --git a/drivers/soc/mediatek/mtk-cmdq-helper.c b/drivers/soc/mediatek/mtk-cmdq-helper.c
index 72350138591c3..42a0194198d5b 100644
--- a/drivers/soc/mediatek/mtk-cmdq-helper.c
+++ b/drivers/soc/mediatek/mtk-cmdq-helper.c
@@ -292,6 +292,19 @@ int cmdq_pkt_write_s_value(struct cmdq_pkt *pkt, u8 high_addr_reg_idx,
 }
 EXPORT_SYMBOL(cmdq_pkt_write_s_value);
 
+int cmdq_pkt_write_s_reg_value(struct cmdq_pkt *pkt, u8 high_addr_reg_idx, u32 value)
+{
+	struct cmdq_instruction inst = {};
+
+	inst.op = CMDQ_CODE_WRITE_S;
+	inst.dst_t = CMDQ_REG_TYPE;
+	inst.reg_dst = high_addr_reg_idx;
+	inst.value = value;
+
+	return cmdq_pkt_append_command(pkt, inst);
+}
+EXPORT_SYMBOL(cmdq_pkt_write_s_reg_value);
+
 int cmdq_pkt_write_s_mask_value(struct cmdq_pkt *pkt, u8 high_addr_reg_idx,
 				u16 addr_low, u32 value, u32 mask)
 {
diff --git a/include/linux/soc/mediatek/mtk-cmdq.h b/include/linux/soc/mediatek/mtk-cmdq.h
index 1edb391ec604a..7bd77d40a5720 100644
--- a/include/linux/soc/mediatek/mtk-cmdq.h
+++ b/include/linux/soc/mediatek/mtk-cmdq.h
@@ -207,6 +207,17 @@ int cmdq_pkt_write_s_value(struct cmdq_pkt *pkt, u8 high_addr_reg_idx,
 int cmdq_pkt_write_s_mask_value(struct cmdq_pkt *pkt, u8 high_addr_reg_idx,
 				u16 addr_low, u32 value, u32 mask);
 
+/**
+ * cmdq_pkt_write_s_reg_value() - append write_s command to the CMDQ packet which
+ *				  write value to a register with low address pa
+ * @pkt:	the CMDQ packet
+ * @high_addr_reg_idx:	internal register ID which contains high address of pa
+ * @value:	the specified target value
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_pkt_write_s_reg_value(struct cmdq_pkt *pkt, u8 high_addr_reg_idx, u32 value);
+
 /**
  * cmdq_pkt_mem_move() - append memory move command to the CMDQ packet
  * @pkt:	the CMDQ packet
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 05/10] mailbox: mtk-cmdq: Support GCE loop packets in interrupt handler

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

1. Add a loop flag for CMDQ packet struct.
CMDQ helper will use a loop flag to mark CMDQ packet as lopping command
and make current command buffer jumps to the beginning when GCE executes
to the end of command buffer.

2. Add a looping task handle flow in irq handler.
GCE irq occurs when GCE executes to the end of command(EOC) instruction.
If the CMDQ packet is a loopping command, GCE irq handler can not
delete the CMDQ task and disable the GCE thread.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/mailbox/mtk-cmdq-mailbox.c       | 11 +++++++++++
 include/linux/mailbox/mtk-cmdq-mailbox.h |  1 +
 2 files changed, 12 insertions(+)

diff --git a/drivers/mailbox/mtk-cmdq-mailbox.c b/drivers/mailbox/mtk-cmdq-mailbox.c
index ead2200f39ba0..5906e0343d1fc 100644
--- a/drivers/mailbox/mtk-cmdq-mailbox.c
+++ b/drivers/mailbox/mtk-cmdq-mailbox.c
@@ -267,6 +267,17 @@ static void cmdq_thread_irq_handler(struct cmdq *cmdq,
 
 	curr_pa = readl(thread->base + CMDQ_THR_CURR_ADDR) << cmdq->pdata->shift;
 
+	task = list_first_entry_or_null(&thread->task_busy_list,
+					struct cmdq_task, list_entry);
+	if (task && task->pkt->loop) {
+		struct cmdq_cb_data data;
+
+		data.sta = err;
+		data.pkt = task->pkt;
+		mbox_chan_received_data(task->thread->chan, &data);
+		return;
+	}
+
 	list_for_each_entry_safe(task, tmp, &thread->task_busy_list,
 				 list_entry) {
 		task_end_pa = task->pa_base + task->pkt->cmd_buf_size;
diff --git a/include/linux/mailbox/mtk-cmdq-mailbox.h b/include/linux/mailbox/mtk-cmdq-mailbox.h
index a8f0070c7aa98..f78a08e7c6ede 100644
--- a/include/linux/mailbox/mtk-cmdq-mailbox.h
+++ b/include/linux/mailbox/mtk-cmdq-mailbox.h
@@ -76,6 +76,7 @@ struct cmdq_pkt {
 	size_t			cmd_buf_size; /* command occupied size */
 	size_t			buf_size; /* real buffer size */
 	void			*cl;
+	bool			loop;
 };
 
 u8 cmdq_get_shift_pa(struct mbox_chan *chan);
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 06/10] soc: mediatek: cmdq: Add cmdq_pkt_finalize_loop for looping cmd with irq

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

Add cmdq_pkt_finalize_loop to CMDQ driver.

cmdq_pkt_finalize_loop appends end of command(EOC) instruction and
jump to start of command buffer instruction to make the command
buffer loopable.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/soc/mediatek/mtk-cmdq-helper.c | 23 +++++++++++++++++++++++
 include/linux/soc/mediatek/mtk-cmdq.h  |  8 ++++++++
 2 files changed, 31 insertions(+)

diff --git a/drivers/soc/mediatek/mtk-cmdq-helper.c b/drivers/soc/mediatek/mtk-cmdq-helper.c
index 42a0194198d5b..3e5a9c4592fd0 100644
--- a/drivers/soc/mediatek/mtk-cmdq-helper.c
+++ b/drivers/soc/mediatek/mtk-cmdq-helper.c
@@ -568,6 +568,29 @@ int cmdq_pkt_finalize(struct cmdq_pkt *pkt)
 }
 EXPORT_SYMBOL(cmdq_pkt_finalize);
 
+int cmdq_pkt_finalize_loop(struct cmdq_pkt *pkt)
+{
+	struct cmdq_instruction inst = { {0} };
+	int err;
+
+	/* insert EOC and generate IRQ for each command iteration */
+	inst.op = CMDQ_CODE_EOC;
+	inst.value = CMDQ_EOC_IRQ_EN;
+	err = cmdq_pkt_append_command(pkt, inst);
+	if (err < 0)
+		return err;
+
+	/* JUMP to start of pkt */
+	err = cmdq_pkt_jump(pkt, pkt->pa_base);
+	if (err < 0)
+		return err;
+
+	pkt->loop = true;
+
+	return err;
+}
+EXPORT_SYMBOL(cmdq_pkt_finalize_loop);
+
 int cmdq_pkt_flush_async(struct cmdq_pkt *pkt)
 {
 	int err;
diff --git a/include/linux/soc/mediatek/mtk-cmdq.h b/include/linux/soc/mediatek/mtk-cmdq.h
index 7bd77d40a5720..f986117d930c0 100644
--- a/include/linux/soc/mediatek/mtk-cmdq.h
+++ b/include/linux/soc/mediatek/mtk-cmdq.h
@@ -371,6 +371,14 @@ int cmdq_pkt_jump(struct cmdq_pkt *pkt, dma_addr_t addr);
  */
 int cmdq_pkt_finalize(struct cmdq_pkt *pkt);
 
+/**
+ * cmdq_pkt_finalize_loop() - Append EOC and jump to start command.
+ * @pkt:	the CMDQ packet
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_pkt_finalize_loop(struct cmdq_pkt *pkt);
+
 /**
  * cmdq_pkt_flush_async() - trigger CMDQ to asynchronously execute the CMDQ
  *                          packet and call back at the end of done packet
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 07/10] mailbox: mediatek: Move reuseable definition to header for secure driver

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

To support CMDQ secure driver, move some reuseable definition to header.
- define: e.g. CMDQ_GCE_NUM_MAX, CMDQ_THR_BASE, CMDQ_THR_SIZE.
- struct: e.g. cmdq_thread, cmdq, cmdq_task.
- include: e.g. <linux/clk.h>.

Add "#include <linux/mailbox_controller.h>" for the function that takes
"struct mbox_chan * chan" as a parameter. That may occur a build error
if secure driver header includes the mtk-cmdq-mailbox.h.
- function: e.g. cmdq_get_shift_pa(struct mbox_chan *chan).

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/mailbox/mtk-cmdq-mailbox.c       | 30 ---------------------
 include/linux/mailbox/mtk-cmdq-mailbox.h | 34 ++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 30 deletions(-)

diff --git a/drivers/mailbox/mtk-cmdq-mailbox.c b/drivers/mailbox/mtk-cmdq-mailbox.c
index 5906e0343d1fc..e04302ca6ec03 100644
--- a/drivers/mailbox/mtk-cmdq-mailbox.c
+++ b/drivers/mailbox/mtk-cmdq-mailbox.c
@@ -3,7 +3,6 @@
 // Copyright (c) 2018 MediaTek Inc.
 
 #include <linux/bitops.h>
-#include <linux/clk.h>
 #include <linux/clk-provider.h>
 #include <linux/dma-mapping.h>
 #include <linux/errno.h>
@@ -22,13 +21,10 @@
 
 #define CMDQ_OP_CODE_MASK		(0xff << CMDQ_OP_CODE_SHIFT)
 #define CMDQ_NUM_CMD(t)			(t->cmd_buf_size / CMDQ_INST_SIZE)
-#define CMDQ_GCE_NUM_MAX		(2)
 
 #define CMDQ_CURR_IRQ_STATUS		0x10
 #define CMDQ_SYNC_TOKEN_UPDATE		0x68
 #define CMDQ_THR_SLOT_CYCLES		0x30
-#define CMDQ_THR_BASE			0x100
-#define CMDQ_THR_SIZE			0x80
 #define CMDQ_THR_WARM_RESET		0x00
 #define CMDQ_THR_ENABLE_TASK		0x04
 #define CMDQ_THR_SUSPEND_TASK		0x08
@@ -59,32 +55,6 @@
 #define CMDQ_JUMP_BY_OFFSET		0x10000000
 #define CMDQ_JUMP_BY_PA			0x10000001
 
-struct cmdq_thread {
-	struct mbox_chan	*chan;
-	void __iomem		*base;
-	struct list_head	task_busy_list;
-	u32			priority;
-};
-
-struct cmdq_task {
-	struct cmdq		*cmdq;
-	struct list_head	list_entry;
-	dma_addr_t		pa_base;
-	struct cmdq_thread	*thread;
-	struct cmdq_pkt		*pkt; /* the packet sent from mailbox client */
-};
-
-struct cmdq {
-	struct mbox_controller	mbox;
-	void __iomem		*base;
-	int			irq;
-	u32			irq_mask;
-	const struct gce_plat	*pdata;
-	struct cmdq_thread	*thread;
-	struct clk_bulk_data	clocks[CMDQ_GCE_NUM_MAX];
-	bool			suspended;
-};
-
 struct gce_plat {
 	u32 thread_nr;
 	u8 shift;
diff --git a/include/linux/mailbox/mtk-cmdq-mailbox.h b/include/linux/mailbox/mtk-cmdq-mailbox.h
index f78a08e7c6ede..092dcadea0d8a 100644
--- a/include/linux/mailbox/mtk-cmdq-mailbox.h
+++ b/include/linux/mailbox/mtk-cmdq-mailbox.h
@@ -7,10 +7,17 @@
 #ifndef __MTK_CMDQ_MAILBOX_H__
 #define __MTK_CMDQ_MAILBOX_H__
 
+#include <linux/clk.h>
+#include <linux/mailbox_controller.h>
 #include <linux/platform_device.h>
 #include <linux/slab.h>
 #include <linux/types.h>
 
+#define CMDQ_GCE_NUM_MAX		2
+
+#define CMDQ_THR_BASE			0x100
+#define CMDQ_THR_SIZE			0x80
+
 #define CMDQ_INST_SIZE			8 /* instruction is 64-bit */
 #define CMDQ_SUBSYS_SHIFT		16
 #define CMDQ_OP_CODE_SHIFT		24
@@ -79,6 +86,33 @@ struct cmdq_pkt {
 	bool			loop;
 };
 
+struct cmdq_thread {
+	struct mbox_chan	*chan;
+	void __iomem		*base;
+	struct list_head	task_busy_list;
+	u32			priority;
+	u32			idx;
+};
+
+struct cmdq {
+	struct mbox_controller	mbox;
+	void __iomem		*base;
+	int			irq;
+	u32			irq_mask;
+	const struct gce_plat	*pdata;
+	struct cmdq_thread	*thread;
+	struct clk_bulk_data	clocks[CMDQ_GCE_NUM_MAX];
+	bool			suspended;
+};
+
+struct cmdq_task {
+	struct cmdq		*cmdq;
+	struct list_head	list_entry;
+	dma_addr_t		pa_base;
+	struct cmdq_thread	*thread;
+	struct cmdq_pkt		*pkt; /* the packet sent from mailbox client */
+};
+
 u8 cmdq_get_shift_pa(struct mbox_chan *chan);
 
 #endif /* __MTK_CMDQ_MAILBOX_H__ */
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 08/10] mailbox: mediatek: Add CMDQ secure mailbox driver

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

To support secure video path feature, GCE have to read/write registgers
in the secure world. GCE will enable the secure access permission to the
HW who wants to access the secure content buffer.

Add CMDQ secure mailbox driver to make CMDQ client user is able to
sending their HW settings to the secure world. So that GCE can execute
all instructions to configure HW in the secure world.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/mailbox/Makefile                      |    2 +-
 drivers/mailbox/mtk-cmdq-sec-mailbox.c        | 1045 +++++++++++++++++
 drivers/mailbox/mtk-cmdq-sec-tee.c            |  165 +++
 include/linux/mailbox/mtk-cmdq-mailbox.h      |    2 +
 .../linux/mailbox/mtk-cmdq-sec-iwc-common.h   |  385 ++++++
 include/linux/mailbox/mtk-cmdq-sec-mailbox.h  |  159 +++
 include/linux/mailbox/mtk-cmdq-sec-tee.h      |  105 ++
 7 files changed, 1862 insertions(+), 1 deletion(-)
 create mode 100644 drivers/mailbox/mtk-cmdq-sec-mailbox.c
 create mode 100644 drivers/mailbox/mtk-cmdq-sec-tee.c
 create mode 100644 include/linux/mailbox/mtk-cmdq-sec-iwc-common.h
 create mode 100644 include/linux/mailbox/mtk-cmdq-sec-mailbox.h
 create mode 100644 include/linux/mailbox/mtk-cmdq-sec-tee.h

diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
index 18793e6caa2f1..bd4b4d64cc317 100644
--- a/drivers/mailbox/Makefile
+++ b/drivers/mailbox/Makefile
@@ -51,7 +51,7 @@ obj-$(CONFIG_STM32_IPCC) 	+= stm32-ipcc.o
 
 obj-$(CONFIG_MTK_ADSP_MBOX)	+= mtk-adsp-mailbox.o
 
-obj-$(CONFIG_MTK_CMDQ_MBOX)	+= mtk-cmdq-mailbox.o
+obj-$(CONFIG_MTK_CMDQ_MBOX)	+= mtk-cmdq-mailbox.o mtk-cmdq-sec-mailbox.o mtk-cmdq-sec-tee.o
 
 obj-$(CONFIG_ZYNQMP_IPI_MBOX)	+= zynqmp-ipi-mailbox.o
 
diff --git a/drivers/mailbox/mtk-cmdq-sec-mailbox.c b/drivers/mailbox/mtk-cmdq-sec-mailbox.c
new file mode 100644
index 0000000000000..6c8d5966ac968
--- /dev/null
+++ b/drivers/mailbox/mtk-cmdq-sec-mailbox.c
@@ -0,0 +1,1045 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2019 MediaTek Inc.
+ */
+
+#include <linux/clk.h>
+#include <linux/dma-mapping.h>
+#include <linux/io.h>
+#include <linux/mailbox_controller.h>
+#include <linux/of_platform.h>
+#include <linux/sched/clock.h>
+#include <linux/timer.h>
+
+#include <linux/mailbox/mtk-cmdq-sec-mailbox.h>
+
+#define CMDQ_THR_EXEC_CNT_PA		(0x28)
+
+#define ADDR_METADATA_MAX_COUNT_ORIGIN	(8)
+#define CMDQ_TIMEOUT_DEFAULT		(1000)
+
+struct cmdq_sec_task {
+	struct cmdq_task		task;
+
+	/* secure CMDQ */
+	bool				reset_exec;
+	u32				wait_cookie;
+	u64				engine_flag;
+	s32				scenario;
+	u64				trigger;
+	u64				exec_time;
+	struct work_struct		exec_work;
+};
+
+struct cmdq_sec_thread {
+	struct cmdq_thread		thread;
+
+	/* secure CMDQ */
+	struct device			*dev;
+	u32				idx;
+	struct timer_list		timeout;
+	u32				timeout_ms;
+	struct work_struct		timeout_work;
+	u32				wait_cookie;
+	u32				next_cookie;
+	u32				task_cnt;
+	struct workqueue_struct		*task_exec_wq;
+};
+
+/**
+ * struct cmdq_sec_context - CMDQ secure context structure.
+ * @tgid: tgid of process context.
+ * @state: state of inter-world communicatiom.
+ * @iwc_msg: buffer for inter-world communicatiom message.
+ * @tee_ctx: context structure for tee vendor.
+ *
+ * Note it is not global data, each process has its own cmdq_sec_context.
+ */
+struct cmdq_sec_context {
+	u32				tgid;
+	enum cmdq_iwc_state_enum	state;
+	void				*iwc_msg;
+	struct cmdq_sec_tee_context	tee_ctx;
+};
+
+/**
+ * struct cmdq_sec_shared_mem - shared memory between normal and secure world
+ * @va: virtual address of share memory.
+ * @pa: physical address of share memory.
+ * @size: size of share memory.
+ *
+ */
+struct cmdq_sec_shared_mem {
+	void				*va;
+	dma_addr_t			pa;
+	u32				size;
+};
+
+struct cmdq_sec {
+	struct device			dev;
+	const struct gce_sec_plat	*pdata;
+	void __iomem			*base;
+	phys_addr_t			base_pa;
+	struct cmdq_sec_thread		*sec_thread;
+	struct cmdq_client		*clt;
+	struct cmdq_pkt			clt_pkt;
+
+	atomic_t			path_res;
+	struct cmdq_sec_shared_mem	*shared_mem;
+	struct cmdq_sec_context		*context;
+	struct iwc_cmdq_cancel_task_t	cancel;
+
+	struct workqueue_struct		*timeout_wq;
+	u64				sec_invoke;
+	u64				sec_done;
+
+	bool				notify_run;
+	struct work_struct		irq_notify_work;
+	struct workqueue_struct		*notify_wq;
+	/* mutex for cmdq_sec_thread excuting cmdq_sec_task */
+	struct mutex			exec_lock;
+};
+
+static atomic_t cmdq_path_res = ATOMIC_INIT(0);
+
+static int cmdq_sec_task_submit(struct cmdq_sec *cmdq, struct cmdq_sec_task *sec_task,
+				const u32 iwc_cmd, const u32 thrd_idx, void *data);
+
+int cmdq_sec_insert_backup_cookie(struct cmdq_pkt *pkt)
+{
+	struct cmdq_client *cl = (struct cmdq_client *)pkt->cl;
+	struct cmdq_thread *thread = ((struct mbox_chan *)(cl->chan))->con_priv;
+	struct cmdq_sec_thread *sec_thread = container_of(thread, struct cmdq_sec_thread, thread);
+	struct cmdq_sec *cmdq = container_of(sec_thread->dev, struct cmdq_sec, dev);
+	struct cmdq_operand left, right;
+	dma_addr_t addr;
+
+	if (!cmdq->shared_mem)
+		return -EFAULT;
+
+	dev_dbg(&cmdq->dev, "%s %d: pkt:%p thread:%u gce:%#lx",
+		__func__, __LINE__, pkt, sec_thread->idx,
+		(unsigned long)cmdq->base_pa);
+
+	addr = (u32)(cmdq->base_pa + CMDQ_THR_BASE +
+		CMDQ_THR_SIZE * sec_thread->idx + CMDQ_THR_EXEC_CNT_PA);
+
+	cmdq_pkt_assign(pkt, CMDQ_THR_SPR_IDX1, CMDQ_ADDR_HIGH(addr));
+	cmdq_pkt_read_s(pkt, CMDQ_THR_SPR_IDX1, CMDQ_ADDR_LOW(addr), CMDQ_THR_SPR_IDX1);
+
+	left.reg = true;
+	left.idx = CMDQ_THR_SPR_IDX1;
+	right.reg = false;
+	right.value = 1;
+	cmdq_pkt_logic_command(pkt, CMDQ_THR_SPR_IDX1, &left, CMDQ_LOGIC_ADD, &right);
+
+	addr = cmdq->shared_mem->pa + CMDQ_SEC_SHARED_THR_CNT_OFFSET +
+		sec_thread->idx * sizeof(u32);
+
+	cmdq_pkt_assign(pkt, CMDQ_THR_SPR_IDX2, CMDQ_ADDR_HIGH(addr));
+	cmdq_pkt_write_s(pkt, CMDQ_THR_SPR_IDX2, CMDQ_ADDR_LOW(addr), CMDQ_THR_SPR_IDX1);
+	cmdq_pkt_set_event(pkt, cmdq->pdata->cmdq_event);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_insert_backup_cookie);
+
+static int cmdq_sec_realloc_addr_list(struct cmdq_pkt *pkt, const u32 count)
+{
+	struct cmdq_sec_data *sec_data = (struct cmdq_sec_data *)pkt->sec_data;
+	void *prev = (void *)(unsigned long)sec_data->addr_metadatas, *curr;
+
+	if (count <= sec_data->addr_metadata_max_cnt)
+		return 0;
+
+	curr = kcalloc(count, sizeof(*sec_data), GFP_KERNEL);
+	if (!curr)
+		return -ENOMEM;
+
+	if (count && sec_data->addr_metadatas)
+		memcpy(curr, prev, sizeof(*sec_data) * sec_data->addr_metadata_max_cnt);
+
+	kfree(prev);
+
+	sec_data->addr_metadatas = (uintptr_t)curr;
+	sec_data->addr_metadata_max_cnt = count;
+	return 0;
+}
+
+void cmdq_sec_pkt_free_sec_data(struct cmdq_pkt *pkt)
+{
+	kfree(pkt->sec_data);
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_pkt_free_sec_data);
+
+int cmdq_sec_pkt_alloc_sec_data(struct cmdq_pkt *pkt)
+{
+	struct cmdq_sec_data *sec_data;
+
+	if (pkt->sec_data)
+		return 0;
+
+	sec_data = kzalloc(sizeof(*sec_data), GFP_KERNEL);
+	if (!sec_data)
+		return -ENOMEM;
+
+	pkt->sec_data = (void *)sec_data;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_pkt_alloc_sec_data);
+
+static int cmdq_sec_append_metadata(struct cmdq_pkt *pkt,
+				    const enum cmdq_iwc_addr_metadata_type type,
+				    const u64 base, const u32 offset, const u32 size,
+				    const u32 port)
+{
+	struct cmdq_sec_data *sec_data;
+	struct iwc_cmdq_addr_metadata_t *meta;
+	int idx, max, ret;
+
+	pr_debug("[%s %d] pkt:%p type:%u base:%#llx offset:%#x size:%#x port:%#x",
+		 __func__, __LINE__, pkt, type, base, offset, size, port);
+
+	ret = cmdq_sec_pkt_alloc_sec_data(pkt);
+	if (ret < 0)
+		return ret;
+
+	sec_data = (struct cmdq_sec_data *)pkt->sec_data;
+	idx = sec_data->addr_metadata_cnt;
+	if (idx >= CMDQ_IWC_MAX_ADDR_LIST_LENGTH) {
+		pr_err("idx:%u reach over:%u", idx, CMDQ_IWC_MAX_ADDR_LIST_LENGTH);
+		return -EFAULT;
+	}
+
+	if (!sec_data->addr_metadata_max_cnt)
+		max = ADDR_METADATA_MAX_COUNT_ORIGIN;
+	else if (idx >= sec_data->addr_metadata_max_cnt)
+		max = sec_data->addr_metadata_max_cnt * 2;
+	else
+		max = sec_data->addr_metadata_max_cnt;
+
+	ret = cmdq_sec_realloc_addr_list(pkt, max);
+	if (ret)
+		return ret;
+
+	if (!sec_data->addr_metadatas) {
+		pr_info("addr_metadatas is missing");
+
+		meta = kzalloc(sizeof(*meta), GFP_KERNEL);
+		if (!meta)
+			return -ENOMEM;
+
+		sec_data->addr_metadatas = (uintptr_t)(void *)meta;
+	}
+	meta = (struct iwc_cmdq_addr_metadata_t *)(uintptr_t)sec_data->addr_metadatas;
+
+	meta[idx].instr_idx = pkt->cmd_buf_size / CMDQ_INST_SIZE - 1;
+	meta[idx].type = type;
+	meta[idx].base_handle = base;
+	meta[idx].offset = offset;
+	meta[idx].size = size;
+	meta[idx].port = port;
+	sec_data->addr_metadata_cnt += 1;
+	return 0;
+}
+
+int cmdq_sec_pkt_set_data(struct cmdq_pkt *pkt, const u64 dapc_engine,
+			  const u64 port_sec_engine, const enum cmdq_sec_scenario scenario)
+{
+	struct cmdq_sec_data *sec_data;
+	int ret;
+
+	if (!pkt) {
+		pr_err("invalid pkt:%p", pkt);
+		return -EINVAL;
+	}
+
+	ret = cmdq_sec_pkt_alloc_sec_data(pkt);
+	if (ret < 0)
+		return ret;
+
+	pr_debug("[%s %d] pkt:%p sec_data:%p dapc:%llu port_sec:%llu scen:%u",
+		 __func__, __LINE__, pkt, pkt->sec_data, dapc_engine, port_sec_engine, scenario);
+
+	sec_data = (struct cmdq_sec_data *)pkt->sec_data;
+	sec_data->engs_need_dapc |= dapc_engine;
+	sec_data->engs_need_sec_port |= port_sec_engine;
+	sec_data->scenario = scenario;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_pkt_set_data);
+
+int cmdq_sec_pkt_write(struct cmdq_pkt *pkt, u32 addr, u64 base,
+		       const enum cmdq_iwc_addr_metadata_type type,
+		       const u32 offset, const u32 size, const u32 port)
+{
+	int ret;
+
+	ret = cmdq_pkt_assign(pkt, CMDQ_THR_SPR_IDX0, addr);
+	if (ret)
+		return ret;
+
+	ret = cmdq_pkt_write_s_reg_value(pkt, CMDQ_THR_SPR_IDX0, (u32)base);
+	if (ret)
+		return ret;
+
+	return cmdq_sec_append_metadata(pkt, type, base, offset, size, port);
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_pkt_write);
+
+static u32 cmdq_sec_get_cookie(struct cmdq_sec *cmdq, u32 idx)
+{
+	return *(u32 *)(cmdq->shared_mem->va +
+		CMDQ_SEC_SHARED_THR_CNT_OFFSET + idx * sizeof(u32));
+}
+
+static void cmdq_sec_task_done(struct cmdq_sec_task *sec_task, int sta)
+{
+	struct cmdq_cb_data data;
+
+	data.sta = sta;
+	data.pkt = sec_task->task.pkt;
+
+	pr_debug("%s sec_task:%p pkt:%p err:%d",
+		 __func__, sec_task, sec_task->task.pkt, sta);
+
+	mbox_chan_received_data(sec_task->task.thread->chan, &data);
+
+	list_del_init(&sec_task->task.list_entry);
+	kfree(sec_task);
+}
+
+static bool cmdq_sec_irq_handler(struct cmdq_sec_thread *sec_thread,
+				 const u32 cookie, const int err)
+{
+	struct cmdq_sec_task *sec_task;
+	struct cmdq_task *task, *temp, *cur_task = NULL;
+	struct cmdq_sec *cmdq = container_of(sec_thread->dev, struct cmdq_sec, dev);
+	unsigned long flags;
+	int done;
+
+	spin_lock_irqsave(&sec_thread->thread.chan->lock, flags);
+	if (sec_thread->wait_cookie <= cookie)
+		done = cookie - sec_thread->wait_cookie + 1;
+	else if (sec_thread->wait_cookie == (cookie + 1) % CMDQ_MAX_COOKIE_VALUE)
+		done = 0;
+	else
+		done = CMDQ_MAX_COOKIE_VALUE - sec_thread->wait_cookie + 1 + cookie + 1;
+
+	list_for_each_entry_safe(task, temp, &sec_thread->thread.task_busy_list, list_entry) {
+		if (!done)
+			break;
+
+		sec_task = container_of(task, struct cmdq_sec_task, task);
+		cmdq_sec_task_done(sec_task, err);
+
+		if (sec_thread->task_cnt)
+			sec_thread->task_cnt -= 1;
+
+		done--;
+	}
+
+	cur_task = list_first_entry_or_null(&sec_thread->thread.task_busy_list,
+					    struct cmdq_task, list_entry);
+	if (err && cur_task) {
+		spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+
+		sec_task = container_of(cur_task, struct cmdq_sec_task, task);
+
+		/* for error task, cancel, callback and done */
+		memset(&cmdq->cancel, 0, sizeof(cmdq->cancel));
+		cmdq_sec_task_submit(cmdq, sec_task, CMD_CMDQ_IWC_CANCEL_TASK,
+				     sec_thread->idx, &cmdq->cancel);
+
+		cmdq_sec_task_done(sec_task, err);
+
+		spin_lock_irqsave(&sec_thread->thread.chan->lock, flags);
+
+		task = list_first_entry_or_null(&sec_thread->thread.task_busy_list,
+						struct cmdq_task, list_entry);
+		if (cur_task == task)
+			cmdq_sec_task_done(sec_task, err);
+		else
+			dev_err(&cmdq->dev, "task list changed");
+
+		/*
+		 * error case stop all task for secure,
+		 * since secure tdrv always remove all when cancel
+		 */
+		while (!list_empty(&sec_thread->thread.task_busy_list)) {
+			cur_task = list_first_entry(&sec_thread->thread.task_busy_list,
+						    struct cmdq_task, list_entry);
+
+			sec_task = container_of(cur_task, struct cmdq_sec_task, task);
+			cmdq_sec_task_done(sec_task, -ECONNABORTED);
+		}
+	} else if (err) {
+		dev_dbg(&cmdq->dev, "error but all task done, check notify callback");
+	}
+
+	if (list_empty(&sec_thread->thread.task_busy_list)) {
+		sec_thread->wait_cookie = 0;
+		sec_thread->next_cookie = 0;
+		sec_thread->task_cnt = 0;
+		__raw_writel(0, (void __iomem *)cmdq->shared_mem->va +
+			     CMDQ_SEC_SHARED_THR_CNT_OFFSET +
+			     sec_thread->idx * sizeof(u32));
+		spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+		del_timer(&sec_thread->timeout);
+		return true;
+	}
+
+	sec_thread->wait_cookie = cookie % CMDQ_MAX_COOKIE_VALUE + 1;
+
+	mod_timer(&sec_thread->timeout, jiffies + msecs_to_jiffies(sec_thread->timeout_ms));
+	spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+
+	return false;
+}
+
+static void cmdq_sec_irq_notify_work(struct work_struct *work_item)
+{
+	struct cmdq_sec *cmdq = container_of(work_item, struct cmdq_sec, irq_notify_work);
+	int i;
+
+	mutex_lock(&cmdq->exec_lock);
+
+	for (i = 0; i <= cmdq->pdata->secure_thread_nr; i++) {
+		struct cmdq_sec_thread *sec_thread = &cmdq->sec_thread[i];
+		u32 cookie = cmdq_sec_get_cookie(cmdq, sec_thread->idx);
+
+		if (cookie < sec_thread->wait_cookie || !sec_thread->task_cnt)
+			continue;
+
+		cmdq_sec_irq_handler(sec_thread, cookie, 0);
+	}
+
+	mutex_unlock(&cmdq->exec_lock);
+}
+
+static void cmdq_sec_irq_notify_callback(struct mbox_client *cl, void *mssg)
+{
+	struct cmdq_cb_data *data = (struct cmdq_cb_data *)mssg;
+	struct cmdq_sec *cmdq = container_of(data->pkt, struct cmdq_sec, clt_pkt);
+
+	if (work_pending(&cmdq->irq_notify_work)) {
+		dev_dbg(&cmdq->dev, "%s last notify callback working", __func__);
+		return;
+	}
+
+	queue_work(cmdq->notify_wq, &cmdq->irq_notify_work);
+}
+
+static int cmdq_sec_irq_notify_start(struct cmdq_sec *cmdq)
+{
+	int err;
+	dma_addr_t dma_addr;
+	struct cmdq_thread *thread;
+
+	if (cmdq->notify_run)
+		return 0;
+
+	cmdq->clt = cmdq_mbox_create(cmdq->pdata->mbox->dev, 0);
+	if (!cmdq->clt || IS_ERR(cmdq->clt)) {
+		dev_err(&cmdq->dev, "clt mbox_create failed clt:%p", cmdq->clt);
+		return -EINVAL;
+	}
+	cmdq->clt->client.rx_callback = cmdq_sec_irq_notify_callback;
+
+	cmdq->clt_pkt.va_base = kzalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!cmdq->clt_pkt.va_base) {
+		dev_err(&cmdq->dev, "clt_pkt cmdq_pkt_create failed");
+		return -ENOMEM;
+	}
+
+	cmdq->clt_pkt.buf_size = PAGE_SIZE;
+	cmdq->clt_pkt.cl = (void *)cmdq->clt;
+
+	dma_addr = dma_map_single(cmdq->pdata->mbox->dev, cmdq->clt_pkt.va_base,
+				  cmdq->clt_pkt.buf_size, DMA_TO_DEVICE);
+	if (dma_mapping_error(cmdq->pdata->mbox->dev, dma_addr)) {
+		dev_err(cmdq->pdata->mbox->dev, "dma map failed, size=%lu\n", PAGE_SIZE);
+		kfree(cmdq->clt_pkt.va_base);
+		return -ENOMEM;
+	}
+	cmdq->clt_pkt.pa_base = dma_addr;
+
+	INIT_WORK(&cmdq->irq_notify_work, cmdq_sec_irq_notify_work);
+
+	cmdq_pkt_wfe(&cmdq->clt_pkt, cmdq->pdata->cmdq_event, true);
+	cmdq_pkt_finalize_loop(&cmdq->clt_pkt);
+	dma_sync_single_for_device(cmdq->pdata->mbox->dev,
+				   cmdq->clt_pkt.pa_base,
+				   cmdq->clt_pkt.cmd_buf_size,
+				   DMA_TO_DEVICE);
+	err = mbox_send_message(cmdq->clt->chan, &cmdq->clt_pkt);
+	mbox_client_txdone(cmdq->clt->chan, 0);
+	if (err < 0) {
+		dev_err(&cmdq->dev, "%s failed:%d", __func__, err);
+		dma_unmap_single(cmdq->pdata->mbox->dev, cmdq->clt_pkt.pa_base,
+				 cmdq->clt_pkt.buf_size, DMA_TO_DEVICE);
+		kfree(cmdq->clt_pkt.va_base);
+		cmdq_mbox_destroy(cmdq->clt);
+
+		return err;
+	}
+
+	cmdq->notify_run = true;
+	thread = (struct cmdq_thread *)cmdq->clt->chan->con_priv;
+	dev_dbg(&cmdq->dev, "%s thread:%d success!", __func__, thread->idx);
+
+	return 0;
+}
+
+static int cmdq_sec_session_init(struct cmdq_sec_context *context)
+{
+	int err = 0;
+
+	if (context->state >= IWC_SES_OPENED) {
+		pr_debug("session opened:%u", context->state);
+		return 0;
+	}
+
+	switch (context->state) {
+	case IWC_INIT:
+		err = cmdq_sec_init_context(&context->tee_ctx);
+		if (err)
+			return err;
+		context->state = IWC_CONTEXT_INITED;
+	fallthrough;
+	case IWC_CONTEXT_INITED:
+		if (context->iwc_msg) {
+			pr_err("iwcMessage not NULL:%p", context->iwc_msg);
+			return -EINVAL;
+		}
+
+		err = cmdq_sec_allocate_wsm(&context->tee_ctx, &context->iwc_msg,
+					    sizeof(struct iwc_cmdq_message_t));
+		if (err)
+			return err;
+
+		context->state = IWC_WSM_ALLOCATED;
+	fallthrough;
+	case IWC_WSM_ALLOCATED:
+		err = cmdq_sec_open_session(&context->tee_ctx, context->iwc_msg);
+		if (err)
+			return err;
+
+		context->state = IWC_SES_OPENED;
+	fallthrough;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int cmdq_sec_fill_iwc_msg(struct cmdq_sec_context *context,
+				 struct cmdq_sec_task *sec_task, u32 thrd_idx)
+{
+	struct iwc_cmdq_message_t *iwc_msg = NULL;
+	struct cmdq_sec_data *data = (struct cmdq_sec_data *)sec_task->task.pkt->sec_data;
+	u32 size = 0, *instr;
+
+	iwc_msg = (struct iwc_cmdq_message_t *)context->iwc_msg;
+
+	if (sec_task->task.pkt->cmd_buf_size + 4 * CMDQ_INST_SIZE > CMDQ_TZ_CMD_BLOCK_SIZE) {
+		pr_err("sec_task:%p size:%zu > %u",
+		       sec_task, sec_task->task.pkt->cmd_buf_size, CMDQ_TZ_CMD_BLOCK_SIZE);
+		return -EFAULT;
+	}
+
+	if (thrd_idx == CMDQ_INVALID_THREAD) {
+		iwc_msg->command.cmd_size = 0;
+		iwc_msg->command.metadata.addr_list_length = 0;
+		return -EINVAL;
+	}
+
+	iwc_msg->command.thread = thrd_idx;
+	iwc_msg->command.scenario = sec_task->scenario;
+	iwc_msg->command.eng_flag = sec_task->engine_flag;
+	size = sec_task->task.pkt->cmd_buf_size;
+	memcpy(iwc_msg->command.va_base, sec_task->task.pkt->va_base, size);
+	iwc_msg->command.cmd_size += size;
+
+	instr = &iwc_msg->command.va_base[iwc_msg->command.cmd_size / 4 - 4];
+	/* Remove IRQ_EN in EOC */
+	if (*(u64 *)instr == CMDQ_EOC_CMD)
+		instr[0] = 0;
+	else
+		pr_err("%s %d: find EOC failed: %#x %#x",
+		       __func__, __LINE__, instr[1], instr[0]);
+
+	iwc_msg->command.wait_cookie = sec_task->wait_cookie;
+	iwc_msg->command.reset_exec = sec_task->reset_exec;
+
+	if (data->addr_metadata_cnt) {
+		iwc_msg->command.metadata.addr_list_length = data->addr_metadata_cnt;
+		memcpy(iwc_msg->command.metadata.addr_list,
+		       (u32 *)(unsigned long)data->addr_metadatas,
+		       data->addr_metadata_cnt * sizeof(struct iwc_cmdq_addr_metadata_t));
+	}
+
+	iwc_msg->command.metadata.engs_need_dapc = data->engs_need_dapc;
+	iwc_msg->command.metadata.engs_need_sec_port = data->engs_need_sec_port;
+	iwc_msg->command.normal_task_handle = (unsigned long)sec_task->task.pkt;
+
+	return 0;
+}
+
+static int cmdq_sec_session_send(struct cmdq_sec_context *context,
+				 struct cmdq_sec_task *sec_task, const u32 iwc_cmd,
+				 const u32 thrd_idx, struct cmdq_sec *cmdq)
+{
+	int err = 0;
+	u64 cost;
+	struct iwc_cmdq_message_t *iwc_msg = NULL;
+
+	iwc_msg = (struct iwc_cmdq_message_t *)context->iwc_msg;
+
+	memset(iwc_msg, 0, sizeof(*iwc_msg));
+	iwc_msg->cmd = iwc_cmd;
+	iwc_msg->cmdq_id = cmdq->pdata->hwid;
+	iwc_msg->command.thread = thrd_idx;
+
+	switch (iwc_cmd) {
+	case CMD_CMDQ_IWC_SUBMIT_TASK:
+		err = cmdq_sec_fill_iwc_msg(context, sec_task, thrd_idx);
+		if (err)
+			return err;
+		break;
+	case CMD_CMDQ_IWC_CANCEL_TASK:
+		iwc_msg->cancel_task.wait_cookie = sec_task->wait_cookie;
+		iwc_msg->cancel_task.thread = thrd_idx;
+		break;
+	case CMD_CMDQ_IWC_PATH_RES_ALLOCATE:
+		if (!cmdq->shared_mem || !cmdq->shared_mem->va) {
+			dev_err(&cmdq->dev, "%s %d: shared_mem is NULL", __func__, __LINE__);
+			return -EFAULT;
+		}
+		iwc_msg->path_resource.size = cmdq->shared_mem->size;
+		iwc_msg->path_resource.share_memoy_pa = cmdq->shared_mem->pa;
+		iwc_msg->path_resource.use_normal_irq = 1;
+		break;
+	default:
+		break;
+	}
+
+	cmdq->sec_invoke = sched_clock();
+	dev_dbg(&cmdq->dev, "%s execute cmdq:%p sec_task:%p command:%u thread:%u cookie:%d",
+		__func__, cmdq, sec_task, iwc_cmd, thrd_idx,
+		sec_task ? sec_task->wait_cookie : -1);
+
+	/* send message */
+	err = cmdq_sec_execute_session(&context->tee_ctx, iwc_cmd, CMDQ_TIMEOUT_DEFAULT);
+
+	cmdq->sec_done = sched_clock();
+	cost = div_u64(cmdq->sec_done - cmdq->sec_invoke, 1000000);
+	if (cost >= CMDQ_TIMEOUT_DEFAULT)
+		dev_err(&cmdq->dev, "%s execute timeout cmdq:%p sec_task:%p cost:%lluus",
+			__func__, cmdq, sec_task, cost);
+	else
+		dev_dbg(&cmdq->dev, "%s execute done cmdq:%p sec_task:%p cost:%lluus",
+			__func__, cmdq, sec_task, cost);
+
+	if (err)
+		return err;
+
+	context->state = IWC_SES_ON_TRANSACTED;
+	return 0;
+}
+
+static int cmdq_sec_session_reply(const u32 iwc_cmd, struct iwc_cmdq_message_t *iwc_msg,
+				  void *data, struct cmdq_sec_task *sec_task)
+{
+	struct iwc_cmdq_cancel_task_t *cancel = data;
+	struct cmdq_sec_data *sec_data = sec_task->task.pkt->sec_data;
+
+	if (iwc_cmd == CMD_CMDQ_IWC_SUBMIT_TASK && iwc_msg->rsp < 0) {
+		/* submit fail case copy status */
+		memcpy(&sec_data->sec_status, &iwc_msg->sec_status,
+		       sizeof(sec_data->sec_status));
+		sec_data->response = iwc_msg->rsp;
+	} else if (iwc_cmd == CMD_CMDQ_IWC_CANCEL_TASK && cancel) {
+		/* cancel case only copy cancel result */
+		memcpy(cancel, &iwc_msg->cancel_task, sizeof(*cancel));
+	}
+
+	return iwc_msg->rsp;
+}
+
+static int cmdq_sec_task_submit(struct cmdq_sec *cmdq, struct cmdq_sec_task *sec_task,
+				const u32 iwc_cmd, const u32 thrd_idx, void *data)
+{
+	struct cmdq_sec_context *context;
+	int err = 0;
+
+	if (!cmdq->context) {
+		context = kzalloc(sizeof(*cmdq->context), GFP_ATOMIC);
+		if (!context)
+			return -ENOMEM;
+
+		cmdq->context = context;
+		cmdq->context->state = IWC_INIT;
+		cmdq->context->tgid = current->tgid;
+	}
+
+	if (cmdq->context->state == IWC_INIT)
+		cmdq_sec_setup_tee_context(&cmdq->context->tee_ctx);
+
+	err = cmdq_sec_session_init(cmdq->context);
+	if (err) {
+		dev_err(&cmdq->dev, "%s %d: cmdq_sec_session_init fail: %d",
+			__func__, __LINE__, err);
+		return err;
+	}
+
+	err = cmdq_sec_irq_notify_start(cmdq);
+	if (err) {
+		dev_err(&cmdq->dev, "%s %d: cmdq_sec_irq_notify_start fail: %d",
+			__func__, __LINE__, err);
+		return err;
+	}
+
+	err = cmdq_sec_session_send(cmdq->context, sec_task, iwc_cmd, thrd_idx, cmdq);
+	if (err) {
+		dev_err(&cmdq->dev, "%s %d: iwc_cmd:%d err:%d sec_task:%p thread:%u gce:%#lx",
+			__func__, __LINE__, iwc_cmd, err, sec_task, thrd_idx,
+			(unsigned long)cmdq->base_pa);
+		return err;
+	}
+
+	err = cmdq_sec_session_reply(iwc_cmd, cmdq->context->iwc_msg, data, sec_task);
+	if (err) {
+		dev_err(&cmdq->dev, "%s %d: cmdq_sec_session_reply fail: %d",
+			__func__, __LINE__, err);
+		return err;
+	}
+
+	return 0;
+}
+
+static void cmdq_sec_task_exec_work(struct work_struct *work_item)
+{
+	struct cmdq_sec_task *sec_task = container_of(work_item,
+						      struct cmdq_sec_task, exec_work);
+	struct cmdq_sec_thread *sec_thread = container_of(sec_task->task.thread,
+							 struct cmdq_sec_thread, thread);
+	struct cmdq_sec *cmdq = container_of(sec_thread->dev, struct cmdq_sec, dev);
+	struct cmdq_sec_data *data;
+	unsigned long flags;
+	int err;
+
+	dev_dbg(&cmdq->dev, "%s gce:%#lx sec_task:%p pkt:%p thread:%u",
+		__func__, (unsigned long)cmdq->base_pa,
+		sec_task, sec_task->task.pkt, sec_thread->idx);
+
+	if (!sec_task->task.pkt->sec_data) {
+		dev_err(&cmdq->dev, "pkt:%p without sec_data", sec_task->task.pkt);
+		return;
+	}
+	data = (struct cmdq_sec_data *)sec_task->task.pkt->sec_data;
+
+	mutex_lock(&cmdq->exec_lock);
+
+	spin_lock_irqsave(&sec_thread->thread.chan->lock, flags);
+	if (!sec_thread->task_cnt) {
+		mod_timer(&sec_thread->timeout, jiffies +
+			  msecs_to_jiffies(sec_thread->timeout_ms));
+		sec_thread->wait_cookie = 1;
+		sec_thread->next_cookie = 1;
+		sec_thread->task_cnt = 0;
+		__raw_writel(0, (void __iomem *)cmdq->shared_mem->va +
+			     CMDQ_SEC_SHARED_THR_CNT_OFFSET + sec_thread->idx * sizeof(u32));
+	}
+
+	sec_task->reset_exec = sec_thread->task_cnt ? false : true;
+	sec_task->wait_cookie = sec_thread->next_cookie;
+	sec_thread->next_cookie = (sec_thread->next_cookie + 1) % CMDQ_MAX_COOKIE_VALUE;
+	list_add_tail(&sec_task->task.list_entry, &sec_thread->thread.task_busy_list);
+	sec_thread->task_cnt += 1;
+	spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+	sec_task->trigger = sched_clock();
+
+	if (!atomic_cmpxchg(&cmdq_path_res, 0, 1)) {
+		err = cmdq_sec_task_submit(cmdq, NULL, CMD_CMDQ_IWC_PATH_RES_ALLOCATE,
+					   CMDQ_INVALID_THREAD, NULL);
+		if (err) {
+			atomic_set(&cmdq_path_res, 0);
+			goto task_end;
+		}
+	}
+
+	if (sec_thread->task_cnt > CMDQ_MAX_TASK_IN_SECURE_THREAD) {
+		dev_err(&cmdq->dev, "task_cnt:%u cannot more than %u sec_task:%p thread:%u",
+			sec_thread->task_cnt, CMDQ_MAX_TASK_IN_SECURE_THREAD,
+			sec_task, sec_thread->idx);
+		err = -EMSGSIZE;
+		goto task_end;
+	}
+
+	err = cmdq_sec_task_submit(cmdq, sec_task, CMD_CMDQ_IWC_SUBMIT_TASK,
+				   sec_thread->idx, NULL);
+	if (err)
+		dev_err(&cmdq->dev, "cmdq_sec_task_submit err:%d sec_task:%p thread:%u",
+			err, sec_task, sec_thread->idx);
+
+task_end:
+	if (err) {
+		struct cmdq_cb_data cb_data;
+
+		cb_data.sta = err;
+		cb_data.pkt = sec_task->task.pkt;
+		mbox_chan_received_data(sec_thread->thread.chan, &cb_data);
+
+		spin_lock_irqsave(&sec_thread->thread.chan->lock, flags);
+		if (!sec_thread->task_cnt)
+			dev_err(&cmdq->dev, "thread:%u task_cnt:%u cannot below zero",
+				sec_thread->idx, sec_thread->task_cnt);
+		else
+			sec_thread->task_cnt -= 1;
+
+		sec_thread->next_cookie = (sec_thread->next_cookie - 1 +
+			CMDQ_MAX_COOKIE_VALUE) % CMDQ_MAX_COOKIE_VALUE;
+		list_del(&sec_task->task.list_entry);
+		dev_dbg(&cmdq->dev, "gce:%#lx err:%d sec_task:%p pkt:%p",
+			(unsigned long)cmdq->base_pa, err, sec_task, sec_task->task.pkt);
+		dev_dbg(&cmdq->dev, "thread:%u task_cnt:%u wait_cookie:%u next_cookie:%u",
+			sec_thread->idx, sec_thread->task_cnt,
+			sec_thread->wait_cookie, sec_thread->next_cookie);
+		spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+
+		kfree(sec_task);
+	}
+
+	mutex_unlock(&cmdq->exec_lock);
+}
+
+static int cmdq_sec_mbox_send_data(struct mbox_chan *chan, void *data)
+{
+	struct cmdq_pkt *pkt = (struct cmdq_pkt *)data;
+	struct cmdq_sec_data *sec_data = (struct cmdq_sec_data *)pkt->sec_data;
+	struct cmdq_thread *thread = (struct cmdq_thread *)chan->con_priv;
+	struct cmdq_sec_thread *sec_thread = container_of(thread, struct cmdq_sec_thread, thread);
+	struct cmdq_sec_task *sec_task;
+
+	if (!sec_data)
+		return -EINVAL;
+
+	sec_task = kzalloc(sizeof(*sec_task), GFP_ATOMIC);
+	if (!sec_task)
+		return -ENOMEM;
+
+	sec_task->task.pkt = pkt;
+	sec_task->task.thread = thread;
+	sec_task->scenario = sec_data->scenario;
+	sec_task->engine_flag = sec_data->engs_need_dapc | sec_data->engs_need_sec_port;
+
+	INIT_WORK(&sec_task->exec_work, cmdq_sec_task_exec_work);
+	queue_work(sec_thread->task_exec_wq, &sec_task->exec_work);
+
+	return 0;
+}
+
+static void cmdq_sec_thread_timeout(struct timer_list *t)
+{
+	struct cmdq_sec_thread *sec_thread = from_timer(sec_thread, t, timeout);
+	struct cmdq_sec *cmdq = container_of(sec_thread->dev, struct cmdq_sec, dev);
+
+	if (!work_pending(&sec_thread->timeout_work))
+		queue_work(cmdq->timeout_wq, &sec_thread->timeout_work);
+}
+
+static void cmdq_sec_task_timeout_work(struct work_struct *work_item)
+{
+	struct cmdq_sec_thread *sec_thread = container_of(work_item,
+							  struct cmdq_sec_thread, timeout_work);
+	struct cmdq_sec *cmdq = container_of(sec_thread->dev, struct cmdq_sec, dev);
+	struct cmdq_task *task;
+	struct cmdq_sec_task *sec_task;
+	unsigned long flags;
+	u64 duration;
+	u32 cookie;
+
+	mutex_lock(&cmdq->exec_lock);
+
+	spin_lock_irqsave(&sec_thread->thread.chan->lock, flags);
+	if (list_empty(&sec_thread->thread.task_busy_list)) {
+		dev_err(&cmdq->dev, "thread:%u task_list is empty", sec_thread->idx);
+		spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+		goto done;
+	}
+
+	task = list_first_entry(&sec_thread->thread.task_busy_list,
+				struct cmdq_task, list_entry);
+	sec_task = container_of(task, struct cmdq_sec_task, task);
+	duration = div_u64(sched_clock() - sec_task->trigger, 1000000);
+	if (duration < sec_thread->timeout_ms) {
+		mod_timer(&sec_thread->timeout, jiffies +
+			  msecs_to_jiffies(sec_thread->timeout_ms - duration));
+		spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+		goto done;
+	}
+
+	cookie = cmdq_sec_get_cookie(cmdq, sec_thread->idx);
+	spin_unlock_irqrestore(&sec_thread->thread.chan->lock, flags);
+
+	dev_err(&cmdq->dev, "%s duration:%llu cookie:%u thread:%u",
+		__func__, duration, cookie, sec_thread->idx);
+	cmdq_sec_irq_handler(sec_thread, cookie, -ETIMEDOUT);
+
+done:
+	mutex_unlock(&cmdq->exec_lock);
+}
+
+static int cmdq_sec_mbox_startup(struct mbox_chan *chan)
+{
+	struct cmdq_thread *thread = (struct cmdq_thread *)chan->con_priv;
+	struct cmdq_sec_thread *sec_thread = container_of(thread,
+							  struct cmdq_sec_thread, thread);
+	char name[20];
+
+	snprintf(name, sizeof(name), "task_exec_wq_%u", sec_thread->idx);
+	sec_thread->task_exec_wq = create_singlethread_workqueue(name);
+
+	return 0;
+}
+
+static int cmdq_sec_mbox_flush(struct mbox_chan *chan, unsigned long timeout)
+{
+	struct cmdq_thread *thread = (struct cmdq_thread *)chan->con_priv;
+	struct cmdq_sec_thread *sec_thread = container_of(thread,
+							  struct cmdq_sec_thread, thread);
+	struct cmdq_sec *cmdq = container_of(sec_thread->dev, struct cmdq_sec, dev);
+	u32 cookie = 0;
+
+	mutex_lock(&cmdq->exec_lock);
+
+	if (list_empty(&thread->task_busy_list)) {
+		mutex_unlock(&cmdq->exec_lock);
+		return 0;
+	}
+
+	cookie = cmdq_sec_get_cookie(cmdq, sec_thread->idx);
+	if (cookie >= sec_thread->wait_cookie && sec_thread->task_cnt > 0)
+		cmdq_sec_irq_handler(sec_thread, cookie, -ECONNABORTED);
+
+	mutex_unlock(&cmdq->exec_lock);
+	return 0;
+}
+
+static void cmdq_sec_mbox_shutdown(struct mbox_chan *chan)
+{
+	cmdq_sec_mbox_flush(chan, 0);
+}
+
+static const struct mbox_chan_ops cmdq_sec_mbox_chan_ops = {
+	.send_data = cmdq_sec_mbox_send_data,
+	.startup = cmdq_sec_mbox_startup,
+	.shutdown = cmdq_sec_mbox_shutdown,
+	.flush = cmdq_sec_mbox_flush,
+};
+
+struct cmdq_sec_mailbox cmdq_sec_mbox = {
+	.ops = &cmdq_sec_mbox_chan_ops,
+};
+
+static int cmdq_sec_probe(struct platform_device *pdev)
+{
+	int i;
+	struct cmdq_sec *cmdq;
+	struct device *dev = &pdev->dev;
+	struct resource *res;
+
+	cmdq = devm_kzalloc(dev, sizeof(*cmdq), GFP_KERNEL);
+	if (!cmdq)
+		return -ENOMEM;
+
+	cmdq->dev = pdev->dev;
+	cmdq->pdata = (struct gce_sec_plat *)pdev->dev.platform_data;
+	if (!cmdq->pdata) {
+		dev_err(dev, "no valid gce platform data!\n");
+		return -EINVAL;
+	}
+
+	cmdq->base = cmdq->pdata->base;
+	res = platform_get_resource(to_platform_device(cmdq->pdata->mbox->dev),
+				    IORESOURCE_MEM, 0);
+	if (IS_ERR(cmdq->base)) {
+		dev_err(dev, "devm_platform_get_and_ioremap_resource failed!\n");
+		return PTR_ERR(cmdq->base);
+	}
+
+	cmdq->base_pa = res->start;
+
+	cmdq->sec_thread = devm_kcalloc(dev, cmdq->pdata->secure_thread_nr,
+					sizeof(*cmdq->sec_thread), GFP_KERNEL);
+	if (!cmdq->sec_thread)
+		return -ENOMEM;
+
+	mutex_init(&cmdq->exec_lock);
+	for (i = 0; i < cmdq->pdata->secure_thread_nr; i++) {
+		u32  idx = i + cmdq->pdata->secure_thread_min;
+
+		cmdq->sec_thread[i].dev = &cmdq->dev;
+		cmdq->sec_thread[i].thread.idx = idx;
+		cmdq->sec_thread[i].idx = cmdq->sec_thread[i].thread.idx;
+		cmdq->sec_thread[i].thread.base = cmdq->base + CMDQ_THR_BASE + CMDQ_THR_SIZE * idx;
+		cmdq->sec_thread[i].timeout_ms = CMDQ_TIMEOUT_DEFAULT;
+		INIT_LIST_HEAD(&cmdq->sec_thread[i].thread.task_busy_list);
+		cmdq->pdata->mbox->chans[idx].con_priv = (void *)&cmdq->sec_thread[i].thread;
+		dev_dbg(dev, "re-assign chans[%d] as secure thread\n", idx);
+		timer_setup(&cmdq->sec_thread[i].timeout, cmdq_sec_thread_timeout, 0);
+		INIT_WORK(&cmdq->sec_thread[i].timeout_work, cmdq_sec_task_timeout_work);
+	}
+
+	cmdq->notify_wq = create_singlethread_workqueue("mtk_cmdq_sec_notify_wq");
+	cmdq->timeout_wq = create_singlethread_workqueue("mtk_cmdq_sec_timeout_wq");
+
+	cmdq->shared_mem = devm_kzalloc(dev, sizeof(*cmdq->shared_mem), GFP_KERNEL);
+	if (!cmdq->shared_mem)
+		return -ENOMEM;
+
+	cmdq->shared_mem->va = dma_alloc_coherent(dev, PAGE_SIZE,
+						  &cmdq->shared_mem->pa, GFP_KERNEL);
+	cmdq->shared_mem->size = PAGE_SIZE;
+
+	platform_set_drvdata(pdev, cmdq);
+
+	return 0;
+}
+
+static int cmdq_sec_remove(struct platform_device *pdev)
+{
+	struct cmdq_sec *cmdq = platform_get_drvdata(pdev);
+
+	if (cmdq->context)
+		cmdq_sec_free_wsm(&cmdq->context->tee_ctx, &cmdq->context->iwc_msg);
+
+	return 0;
+}
+
+static struct platform_driver cmdq_sec_drv = {
+	.probe = cmdq_sec_probe,
+	.remove = cmdq_sec_remove,
+	.driver = {
+		.name = "mtk-cmdq-sec",
+	},
+};
+
+static int __init cmdq_sec_init(void)
+{
+	return platform_driver_register(&cmdq_sec_drv);
+}
+
+static void __exit cmdq_sec_exit(void)
+{
+	platform_driver_unregister(&cmdq_sec_drv);
+}
+
+module_init(cmdq_sec_init);
+module_exit(cmdq_sec_exit);
+
+MODULE_LICENSE("GPL");
diff --git a/drivers/mailbox/mtk-cmdq-sec-tee.c b/drivers/mailbox/mtk-cmdq-sec-tee.c
new file mode 100644
index 0000000000000..e3f5ba6aaaaad
--- /dev/null
+++ b/drivers/mailbox/mtk-cmdq-sec-tee.c
@@ -0,0 +1,165 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2019 MediaTek Inc.
+ */
+
+#include <linux/math64.h>
+#include <linux/sched/clock.h>
+
+#include <linux/mailbox/mtk-cmdq-sec-tee.h>
+#include <linux/soc/mediatek/mtk-cmdq.h>
+
+/* lock to protect atomic secure task execution */
+static DEFINE_MUTEX(cmdq_sec_exec_lock);
+
+void cmdq_sec_setup_tee_context(struct cmdq_sec_tee_context *tee)
+{
+	/* 09010000 0000 0000 0000000000000000 */
+	memset(tee->uuid, 0, sizeof(tee->uuid));
+	tee->uuid[0] = 0x9;
+	tee->uuid[1] = 0x1;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_setup_tee_context);
+
+static int tee_dev_match(struct tee_ioctl_version_data *t, const void *v)
+{
+	if (t->impl_id == TEE_IMPL_ID_OPTEE)
+		return 1;
+
+	return 0;
+}
+
+int cmdq_sec_init_context(struct cmdq_sec_tee_context *tee)
+{
+	tee->tee_context = tee_client_open_context(NULL, tee_dev_match, NULL, NULL);
+	if (!tee->tee_context) {
+		pr_err("[%s][%d] tee_client_open_context failed!", __func__, __LINE__);
+		return -EFAULT;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_init_context);
+
+int cmdq_sec_deinit_context(struct cmdq_sec_tee_context *tee)
+{
+	if (tee && tee->tee_context)
+		tee_client_close_context(tee->tee_context);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_deinit_context);
+
+int cmdq_sec_allocate_wsm(struct cmdq_sec_tee_context *tee, void **wsm_buffer, u32 size)
+{
+	void *buffer;
+
+	if (!wsm_buffer)
+		return -EINVAL;
+
+	if (size == 0)
+		return -EINVAL;
+
+	buffer = kmalloc(size, GFP_KERNEL);
+	if (!buffer)
+		return -ENOMEM;
+
+	tee->shared_mem = tee_shm_register_kernel_buf(tee->tee_context, buffer, size);
+	if (!tee->shared_mem) {
+		kfree(buffer);
+		return -ENOMEM;
+	}
+
+	*wsm_buffer = buffer;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_allocate_wsm);
+
+int cmdq_sec_free_wsm(struct cmdq_sec_tee_context *tee, void **wsm_buffer)
+{
+	if (!wsm_buffer)
+		return -EINVAL;
+
+	tee_shm_put(tee->shared_mem);
+	tee->shared_mem = NULL;
+	kfree(*wsm_buffer);
+	*wsm_buffer = NULL;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_free_wsm);
+
+int cmdq_sec_open_session(struct cmdq_sec_tee_context *tee, void *wsm_buffer)
+{
+	struct tee_ioctl_open_session_arg osarg = {0};
+	struct tee_param params = {0};
+	int ret = 0;
+
+	if (!wsm_buffer)
+		return -EINVAL;
+
+	osarg.num_params = 1;
+	memcpy(osarg.uuid, tee->uuid, sizeof(osarg.uuid));
+	osarg.clnt_login = 0;
+
+	ret = tee_client_open_session(tee->tee_context, &osarg, &params);
+	if (ret)
+		return -EFAULT;
+
+	if (!osarg.ret)
+		tee->session = osarg.session;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_open_session);
+
+int cmdq_sec_close_session(struct cmdq_sec_tee_context *tee)
+{
+	tee_client_close_session(tee->tee_context, tee->session);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_close_session);
+
+int cmdq_sec_execute_session(struct cmdq_sec_tee_context *tee, u32 cmd, s32 timeout_ms)
+{
+	struct tee_ioctl_invoke_arg invoke_arg = {0};
+	struct tee_param params = {0};
+	u64 ts = sched_clock();
+	int ret = 0;
+
+	mutex_lock(&cmdq_sec_exec_lock);
+
+	params.attr = TEE_IOCTL_PARAM_ATTR_TYPE_MEMREF_INOUT;
+	params.u.memref.shm = tee->shared_mem;
+	params.u.memref.shm_offs = 0;
+	params.u.memref.size = tee->shared_mem->size;
+
+	invoke_arg.num_params = 1;
+	invoke_arg.session = tee->session;
+	invoke_arg.func = cmd;
+
+	ret = tee_client_invoke_func(tee->tee_context, &invoke_arg, &params);
+	if (ret) {
+		pr_err("tee_client_invoke_func failed, ret=%d\n", ret);
+		return -EFAULT;
+	}
+
+	ret = invoke_arg.ret;
+
+	mutex_unlock(&cmdq_sec_exec_lock);
+
+	ts = div_u64(sched_clock() - ts, 1000000);
+
+	if (ret != 0)
+		pr_err("[SEC]execute: TEEC_InvokeCommand:%u ret:%d cost:%lluus", cmd, ret, ts);
+	else if (ts > timeout_ms)
+		pr_err("[SEC]execute: TEEC_InvokeCommand:%u ret:%d cost:%lluus", cmd, ret, ts);
+	else
+		pr_debug("[SEC]execute: TEEC_InvokeCommand:%u ret:%d cost:%lluus", cmd, ret, ts);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(cmdq_sec_execute_session);
+
+MODULE_LICENSE("GPL");
diff --git a/include/linux/mailbox/mtk-cmdq-mailbox.h b/include/linux/mailbox/mtk-cmdq-mailbox.h
index 092dcadea0d8a..06fe9edb14009 100644
--- a/include/linux/mailbox/mtk-cmdq-mailbox.h
+++ b/include/linux/mailbox/mtk-cmdq-mailbox.h
@@ -22,6 +22,7 @@
 #define CMDQ_SUBSYS_SHIFT		16
 #define CMDQ_OP_CODE_SHIFT		24
 #define CMDQ_JUMP_PASS			CMDQ_INST_SIZE
+#define CMDQ_EOC_CMD			0x4000000000000001ULL
 
 #define CMDQ_WFE_UPDATE			BIT(31)
 #define CMDQ_WFE_UPDATE_VALUE		BIT(16)
@@ -84,6 +85,7 @@ struct cmdq_pkt {
 	size_t			buf_size; /* real buffer size */
 	void			*cl;
 	bool			loop;
+	void			*sec_data;
 };
 
 struct cmdq_thread {
diff --git a/include/linux/mailbox/mtk-cmdq-sec-iwc-common.h b/include/linux/mailbox/mtk-cmdq-sec-iwc-common.h
new file mode 100644
index 0000000000000..655185ea8a82c
--- /dev/null
+++ b/include/linux/mailbox/mtk-cmdq-sec-iwc-common.h
@@ -0,0 +1,385 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2019 MediaTek Inc.
+ */
+
+#ifndef __CMDQ_SEC_IWC_COMMON_H__
+#define __CMDQ_SEC_IWC_COMMON_H__
+
+/**
+ * CMDQ_SEC_SHARED_THR_CNT_OFFSET - shared memory offset to store thread count.
+ */
+#define CMDQ_SEC_SHARED_THR_CNT_OFFSET		0x100
+
+/**
+ * CMDQ_TZ_CMD_BLOCK_SIZE - total command buffer size copy from normal world to secure world.
+ * maximum 20 pages may be requested for MDP readback.
+ */
+#define CMDQ_TZ_CMD_BLOCK_SIZE		(20 << 12)
+
+/**
+ * CMDQ_IWC_MAX_CMD_LENGTH - max length of u32 array to store commanad buffer.
+ */
+#define CMDQ_IWC_MAX_CMD_LENGTH		(CMDQ_TZ_CMD_BLOCK_SIZE / sizeof(u32))
+
+/**
+ * CMDQ_IWC_MAX_ADDR_LIST_LENGTH - max length of addr metadata list.
+ */
+#define CMDQ_IWC_MAX_ADDR_LIST_LENGTH	(30)
+
+/**
+ * CMDQ_IWC_CLIENT_NAME - length for caller_name in iwc_cmdq_command_t.
+ */
+#define CMDQ_IWC_CLIENT_NAME		(16)
+
+/**
+ * CMDQ_MAX_READBACK_ENG - length for readback_engs in iwc_cmdq_command_t.
+ */
+#define CMDQ_MAX_READBACK_ENG		(8)
+
+/**
+ * CMDQ_SEC_MESSAGE_INST_LEN - length for sec_inst in iwc_cmdq_sec_status_t.
+ */
+#define CMDQ_SEC_MESSAGE_INST_LEN	(8)
+
+/**
+ * CMDQ_SEC_DISPATCH_LEN - length for dispatch in iwc_cmdq_sec_status_t.
+ */
+#define CMDQ_SEC_DISPATCH_LEN		(8)
+
+/*
+ * IWC Command IDs - ID for normal world(TLC or linux kernel) to secure world.
+ */
+#define CMD_CMDQ_IWC_SUBMIT_TASK	(1) /* submit current task */
+#define CMD_CMDQ_IWC_CANCEL_TASK	(3) /* cancel current task */
+#define CMD_CMDQ_IWC_PATH_RES_ALLOCATE	(4) /* create global resource for secure path */
+
+/**
+ * enum cmdq_iwc_addr_metadata_type - address medadata type to be converted in secure world.
+ * @CMDQ_IWC_H_2_PA: secure handle to sec PA.
+ * @CMDQ_IWC_H_2_MVA: secure handle to sec MVA.
+ * @CMDQ_IWC_NMVA_2_MVA: map normal MVA to secure world.
+ * @CMDQ_IWC_PH_2_MVA: session protected handle to sec MVA.
+ *
+ * To tell secure world waht operation to use for converting address in metadata list.
+ */
+enum cmdq_iwc_addr_metadata_type {
+	CMDQ_IWC_H_2_PA		= 0,
+	CMDQ_IWC_H_2_MVA	= 1,
+	CMDQ_IWC_NMVA_2_MVA	= 2,
+	CMDQ_IWC_PH_2_MVA	= 3,
+};
+
+/*
+ * enum cmdq_sec_engine_enum - the flag for HW engines need to be proteced in secure world.
+ * Each enum is a bit in a u64 engine flag variable.
+ */
+enum cmdq_sec_engine_enum {
+	/* MDP */
+	CMDQ_SEC_MDP_RDMA0		= 0,
+	CMDQ_SEC_MDP_RDMA1		= 1,
+	CMDQ_SEC_MDP_WDMA		= 2,
+	CMDQ_SEC_MDP_RDMA2		= 3,
+	CMDQ_SEC_MDP_RDMA3		= 4,
+	CMDQ_SEC_MDP_WROT0		= 5,
+	CMDQ_SEC_MDP_WROT1		= 6,
+	CMDQ_SEC_MDP_WROT2		= 7,
+	CMDQ_SEC_MDP_WROT3		= 8,
+	CMDQ_SEC_MDP_HDR0		= 9,
+	CMDQ_SEC_MDP_HDR1		= 10,
+	CMDQ_SEC_MDP_HDR2		= 11,
+	CMDQ_SEC_MDP_HDR3		= 12,
+	CMDQ_SEC_MDP_AAL0		= 13,
+	CMDQ_SEC_MDP_AAL1		= 14,
+	CMDQ_SEC_MDP_AAL2		= 15,
+	CMDQ_SEC_MDP_AAL3		= 16,
+
+	/* DISP (VDOSYS0) */
+	CMDQ_SEC_DISP_RDMA0		= 17,
+	CMDQ_SEC_DISP_RDMA1		= 18,
+	CMDQ_SEC_DISP_WDMA0		= 19,
+	CMDQ_SEC_DISP_WDMA1		= 20,
+	CMDQ_SEC_DISP_OVL0		= 21,
+	CMDQ_SEC_DISP_OVL1		= 22,
+	CMDQ_SEC_DISP_OVL2		= 23,
+	CMDQ_SEC_DISP_2L_OVL0		= 24,
+	CMDQ_SEC_DISP_2L_OVL1		= 25,
+	CMDQ_SEC_DISP_2L_OVL2		= 26,
+
+	/* DSIP (VDOSYS1) */
+	CMDQ_SEC_VDO1_DISP_RDMA_L0	= 27,
+	CMDQ_SEC_VDO1_DISP_RDMA_L1	= 28,
+	CMDQ_SEC_VDO1_DISP_RDMA_L2	= 29,
+	CMDQ_SEC_VDO1_DISP_RDMA_L3	= 30,
+
+	/* VENC */
+	CMDQ_SEC_VENC_BSDMA		= 31,
+	CMDQ_SEC_VENC_CUR_LUMA		= 32,
+	CMDQ_SEC_VENC_CUR_CHROMA	= 33,
+	CMDQ_SEC_VENC_REF_LUMA		= 34,
+	CMDQ_SEC_VENC_REF_CHROMA	= 35,
+	CMDQ_SEC_VENC_REC		= 36,
+	CMDQ_SEC_VENC_SUB_R_LUMA	= 37,
+	CMDQ_SEC_VENC_SUB_W_LUMA	= 38,
+	CMDQ_SEC_VENC_SV_COMV		= 39,
+	CMDQ_SEC_VENC_RD_COMV		= 40,
+	CMDQ_SEC_VENC_NBM_RDMA		= 41,
+	CMDQ_SEC_VENC_NBM_WDMA		= 42,
+	CMDQ_SEC_VENC_NBM_RDMA_LITE	= 43,
+	CMDQ_SEC_VENC_NBM_WDMA_LITE	= 44,
+	CMDQ_SEC_VENC_FCS_NBM_RDMA	= 45,
+	CMDQ_SEC_VENC_FCS_NBM_WDMA	= 46,
+
+	CMDQ_SEC_MAX_ENG_COUNT
+};
+
+/**
+ * struct iwc_cmdq_addr_metadata_t - metadata structure for converting address of secure buffer.
+ * @instr_idx: index of instruction.
+ * @type: addr metadata type.
+ * @base_handle: secure address handle.
+ * @block_offset: block offset from handle(PA) to current block(plane).
+ * @offset: buffser offset to secure handle.
+ * @size: buffer size.
+ * @port: HW port id (i.e. M4U port id)
+ */
+struct iwc_cmdq_addr_metadata_t {
+	/**
+	 * @instr_idx: update its arg_b value to real PA/MVA in secure world.
+	 */
+	u32 instr_idx;
+
+	/**
+	 * @type: address medadata type to be converted in secure world.
+	 */
+	u32 type;
+
+	/**
+	 * @base_handle:
+	 * @block_offset:
+	 * @offset:
+	 * @size:
+	 * these members are used to store the buffer and offset relationship.
+	 *
+	 *   -------------
+	 *   |     |     |
+	 *   -------------
+	 *   ^     ^  ^  ^
+	 *   A     B  C  D
+	 *
+	 *  A: base_handle
+	 *  B: base_handle + block_offset
+	 *  C: base_handle + block_offset + offset
+	 *  A~B or B~D: size
+	 */
+	u64 base_handle;
+	u32 block_offset;
+	u32 offset;
+	u32 size;
+
+	/**
+	 * @port: [IN]
+	 *
+	 * used to configure M4U port id.
+	 */
+	u32 port;
+};
+
+/**
+ * struct iwc_cmdq_metadata_t - metadata structure for converting a list of secure buffer address.
+ * @addr_list_length: length of metadata address list.
+ * @addr_list: array of metadata address list.
+ * @engs_need_dapc: HW engines need to be protected by dapc.
+ * @engs_need_sec_port: HW engines need to be protected by secure larb port.
+ */
+struct iwc_cmdq_metadata_t {
+	u32 addr_list_length;
+	struct iwc_cmdq_addr_metadata_t addr_list[CMDQ_IWC_MAX_ADDR_LIST_LENGTH];
+	u64 engs_need_dapc;
+	u64 engs_need_sec_port;
+};
+
+/**
+ * enum sec_extension_iwc - extension HW engine flag to be protcted in secure world.
+ * @IWC_MDP_AAL: for MDP AAL engine.
+ * @IWC_MDP_TDSHP: for MDP TDSHP engine.
+ */
+enum sec_extension_iwc {
+	IWC_MDP_AAL = 0,
+	IWC_MDP_TDSHP,
+};
+
+/**
+ * struct readback_engine - readback engine parameters.
+ * @engine: HW engine flag for readback.
+ * @start: start address pa of readback buffer.
+ * @count: u32 size count of readback buffer.
+ * @param: other parameters need in secure world.
+ */
+struct readback_engine {
+	u32 engine;
+	u32 start;
+	u32 count;
+	u32 param;
+};
+
+/**
+ * struct iwc_cmdq_command_t - structure for excuting cmdq task in secure world.
+ * @thread: GCE secure thread index to execute command.
+ * @scenario: scenario to execute command.
+ * @priority: priority of GCE secure thread.
+ * @cmd_size: command size used in command buffer.
+ * @eng_flag: HW engine flag need to enable protection configuration.
+ * @va_base: command buffer
+ * @wait_cookie: index in thread's task list, it should be (nextCookie - 1).
+ * @reset_exec: reset HW thread.
+ * @caller_pid: pid of client module.
+ * @caller_name: name of client module.
+ * @metadata: metadata structure for converting a list of secure buffer address.
+ * @extension: extension HW engine flag to be protcted in secure world.
+ * @readback_pa: readback buffer pa.
+ * @normal_task_handle: handle to reference task in normal world.
+ * @mdp_extension: extension MDP HW engine flag to be protcted in secure world.
+ * @readback_engs: array of readback engines parameters.
+ * @readback_cnt: count of readback_engs array.
+ */
+struct iwc_cmdq_command_t {
+	/* basic execution data */
+	u32 thread;
+	u32 scenario;
+	u32 priority;
+	u32 cmd_size;
+	u64 eng_flag;
+	u32 va_base[CMDQ_IWC_MAX_CMD_LENGTH];
+
+	/* exec order data */
+	u32 wait_cookie;
+	bool reset_exec;
+
+	/* client info */
+	s32 caller_pid;
+	char caller_name[CMDQ_IWC_CLIENT_NAME];
+
+	/* metadata */
+	struct iwc_cmdq_metadata_t metadata;
+
+	/* client extension bits */
+	u64 extension;
+	u64 readback_pa;
+
+	/* debug */
+	u64 normal_task_handle;
+
+	/* SVP HDR */
+	u32 mdp_extension;
+	struct readback_engine readback_engs[CMDQ_MAX_READBACK_ENG];
+	u32 readback_cnt;
+};
+
+/**
+ * struct iwc_cmdq_cancel_task_t - structure for canceling cmdq task in the secure world.
+ * @thread: [IN] GCE secure thread index.
+ * @wait_cookie: [IN] execute count cookie to wait.
+ * @throw_aee: [OUT] AEE has thrown.
+ * @has_reset: [OUT] current secure thread has been reset
+ * @irq_status: [OUT] global secure IRQ flag.
+ * @irq_flag: [OUT] thread IRQ flag.
+ * @err_instr: [OUT] err_instr[0] = instruction low bits, err_instr[1] = instruction high bits.
+ * @reg_value: [OUT] value of error register.
+ * @pc: [OUT] current pc.
+ *
+ * used to allocate share memory from secure world.
+ */
+struct iwc_cmdq_cancel_task_t {
+	s32 thread;
+	u32 wait_cookie;
+	bool throw_aee;
+	bool has_reset;
+	s32 irq_status;
+	s32 irq_flag;
+	u32 err_instr[2];
+	u32 reg_value;
+	u32 pc;
+};
+
+/**
+ * struct iwc_cmdq_path_resource_t - Inter-World Communication resource allocation structure.
+ * @share_memoy_pa: use long long for 64 bit compatible support.
+ * @size: size of share memory.
+ * @use_normal_irq: use normal IRQ in secure world.
+ *
+ * used to allocate share memory from secure world.
+ */
+struct iwc_cmdq_path_resource_t {
+	long long share_memoy_pa;
+	u32 size;
+	bool use_normal_irq;
+};
+
+/**
+ * struct iwc_cmdq_debug_config_t - debug config structure for secure debug log.
+ *
+ * @log_level: log level in secure world.
+ * @enable_profile: enable profile in secure world.
+ */
+struct iwc_cmdq_debug_config_t {
+	s32 log_level;
+	s32 enable_profile;
+};
+
+/**
+ * struct iwc_cmdq_sec_status_t - secure status from secure world.
+ *
+ * @step: the step in secure cmdq TA.
+ * @status: the status in secure cmdq TA.
+ * @args: the status arguments in secure cmdq TA.
+ * @sec_inst: current instruction in secure cmdq TA.
+ * @inst_index: current instruction index in secure cmdq TA.
+ * @dispatch: current HW engine configuring in secure cmdq TA.
+ */
+struct iwc_cmdq_sec_status_t {
+	u32 step;
+	s32 status;
+	u32 args[4];
+	u32 sec_inst[CMDQ_SEC_MESSAGE_INST_LEN];
+	u32 inst_index;
+	char dispatch[CMDQ_SEC_DISPATCH_LEN];
+};
+
+/**
+ * struct iwc_cmdq_message_t - Inter-World Communication message structure.
+ * @cmd: [IN] iwc command id.
+ * @rsp: [OUT] respond from secureworld, 0 for success, < 0 for error.
+ * @command: [IN] structure for excuting cmdq task in secure world.
+ * @cancel_task: [IN] structure for canceling cmdq task in the secure world.
+ * @path_resource: [IN]
+ * @debug: [IN] debug config structure for secure debug log.
+ * @sec_status: [OUT] secure status from secure world.
+ * @cmdq_id: [IN] GCE core id.
+ *
+ * Both Linex kernel and mobicore have their own MMU tables for mapping
+ * world shared memory and physical addresses, so mobicore does not understand
+ * linux virtual address mapping.
+ * If we want to transact a large buffer in TCI/DCI, there are 2 ways (both require 1 copy):
+ * 1. Ue mc_map to map the normal world buffer to WSM and pass secure_virt_addr in TCI/DCI buffer.
+ *    Note that mc_map implies a memcopy to copy the content from normal world to WSM.
+ * 2. Declare a fixed-length array in TCI/DCI struct and its size must be < 1M.
+ */
+struct iwc_cmdq_message_t {
+	union {
+		u32 cmd;
+		s32 rsp;
+	};
+
+	union {
+		struct iwc_cmdq_command_t command;
+		struct iwc_cmdq_cancel_task_t cancel_task;
+		struct iwc_cmdq_path_resource_t path_resource;
+	};
+
+	struct iwc_cmdq_debug_config_t debug;
+	struct iwc_cmdq_sec_status_t sec_status;
+
+	u8 cmdq_id;
+};
+#endif /* __CMDQ_SEC_IWC_COMMON_H__ */
diff --git a/include/linux/mailbox/mtk-cmdq-sec-mailbox.h b/include/linux/mailbox/mtk-cmdq-sec-mailbox.h
new file mode 100644
index 0000000000000..f77f7365bac37
--- /dev/null
+++ b/include/linux/mailbox/mtk-cmdq-sec-mailbox.h
@@ -0,0 +1,159 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2019 MediaTek Inc.
+ */
+
+#ifndef __MTK_CMDQ_SEC_MAILBOX_H__
+#define __MTK_CMDQ_SEC_MAILBOX_H__
+
+#include <linux/kernel.h>
+
+#include <linux/mailbox/mtk-cmdq-mailbox.h>
+#include <linux/mailbox/mtk-cmdq-sec-iwc-common.h>
+#include <linux/mailbox/mtk-cmdq-sec-tee.h>
+#include <linux/soc/mediatek/mtk-cmdq.h>
+
+#define CMDQ_INVALID_THREAD		(-1)
+#define CMDQ_MAX_TASK_IN_SECURE_THREAD	(16)
+
+/**
+ * CMDQ_MAX_COOKIE_VALUE - max value of CMDQ_THR_EXEC_CNT_PA (value starts from 0)
+ */
+#define CMDQ_MAX_COOKIE_VALUE           (0xffff)
+
+/**
+ * enum cmdq_sec_scenario - scenario settings for cmdq TA.
+ * @CMDQ_SEC_SCNR_PRIMARY_DISP: primary display vdo mode enable.
+ * @CMDQ_SEC_SCNR_SUB_DISP: external display vdo mode enable.
+ * @CMDQ_SEC_SCNR_PRIMARY_DISP_DISABLE: primary display vdo mode disable.
+ * @CMDQ_SEC_SCNR_SUB_DISP_DISABLE: external display vdo mode disable.
+ * @CMDQ_SEC_SCNR_MAX: the end of enum.
+ *
+ * These states are used to record the state of IWC message structure.
+ */
+enum cmdq_sec_scenario {
+	CMDQ_SEC_SCNR_PRIMARY_DISP		= 1,
+	CMDQ_SEC_SCNR_SUB_DISP			= 4,
+	CMDQ_SEC_SCNR_PRIMARY_DISP_DISABLE	= 18,
+	CMDQ_SEC_SCNR_SUB_DISP_DISABLE		= 19,
+	CMDQ_SEC_SCNR_MAX,
+};
+
+/**
+ * enum cmdq_iwc_state_enum - state of Inter-world Communication(IWC) message
+ * @IWC_INIT: state of initializing tee context, means tee context has not initialized.
+ * @IWC_CONTEXT_INITED: tee context has initialized.
+ * @IWC_WSM_ALLOCATED: world share memory has allocated.
+ * @IWC_SES_OPENED: session to the tee context has opend.
+ * @IWC_SES_ON_TRANSACTED: session to the tee context has transacted.
+ * @IWC_STATE_MAX: the end of enum.
+ *
+ * These states are used to record the state of IWC message structure.
+ */
+enum cmdq_iwc_state_enum {
+	IWC_INIT,
+	IWC_CONTEXT_INITED,
+	IWC_WSM_ALLOCATED,
+	IWC_SES_OPENED,
+	IWC_SES_ON_TRANSACTED,
+	IWC_STATE_MAX,
+};
+
+/**
+ * struct gce_sec_plat - used to pass platform data from cmdq driver.
+ * @mbox: pointer to mbox controller.
+ * @base: GCE register base va.
+ * @hwid: GCE core id.
+ * @secure_thread_nr: number of secure thread.
+ * @secure_thread_min: min index of secure thread.
+ * @cmdq_event: secure EOF event id.
+ */
+struct gce_sec_plat {
+	struct mbox_controller *mbox;
+	void __iomem *base;
+	u32 hwid;
+	u8 secure_thread_nr;
+	u8 secure_thread_min;
+	u32 cmdq_event;
+};
+
+struct cmdq_sec_mailbox {
+	const struct mbox_chan_ops *ops;
+};
+
+extern struct cmdq_sec_mailbox cmdq_sec_mbox;
+
+/**
+ * struct cmdq_sec_data - used to translate secure buffer PA related instruction
+ * @addr_metadata_cnt: count of element in addr_list.
+ * @addr_metadatas: array of iwc_cmdq_addr_metadata_t.
+ * @addr_metadata_max_cnt: Reserved.
+ * @scenario: scenario config for secure world.
+ * @engs_need_dapc: engine falgs that need to config dapc in secure world.
+ * @engs_need_sec_port: engine falgs that need to config secure larb port in secure world.
+ * @response: return value from secure world.
+ * @sec_status: current iwc message status from secure world.
+ *
+ */
+struct cmdq_sec_data {
+	u32 addr_metadata_cnt;
+	u64 addr_metadatas;
+	u32 addr_metadata_max_cnt;
+	enum cmdq_sec_scenario scenario;
+	u64 engs_need_dapc;
+	u64 engs_need_sec_port;
+	s32 response;
+	struct iwc_cmdq_sec_status_t sec_status;
+};
+
+/**
+ * cmdq_sec_pkt_free_sec_data() - free sec_data for CMDQ packet.
+ * @pkt:	the CMDQ packet.
+ */
+void cmdq_sec_pkt_free_sec_data(struct cmdq_pkt *pkt);
+
+/**
+ * cmdq_sec_pkt_alloc_sec_data() - allocate sec_data for CMDQ packet.
+ * @pkt:	the CMDQ packet.
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_sec_pkt_alloc_sec_data(struct cmdq_pkt *pkt);
+
+/**
+ * cmdq_sec_insert_backup_cookie() - append backup cookie related instructions.
+ * @pkt:	the CMDQ packet.
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_sec_insert_backup_cookie(struct cmdq_pkt *pkt);
+
+/**
+ * cmdq_sec_pkt_set_data() - set secure configuration to sec_data in CDMQ packet.
+ * @pkt:	the CMDQ packet.
+ * @dapc_engine:	the engine flag for dapc protection.
+ * @port_sec_engine:	the engine flag for secure larb prot protection.
+ * @scenario:		the scenario to CMDQ TA.
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_sec_pkt_set_data(struct cmdq_pkt *pkt, const u64 dapc_engine,
+			  const u64 port_sec_engine, const enum cmdq_sec_scenario scenario);
+
+/**
+ * cmdq_sec_pkt_write() - append write secure buffer related instructions.
+ * @pkt:	  the CMDQ packet.
+ * @addr:	the register to be configured.
+ * @base:	the secure handle of secure buffer.
+ * @type:	the address metadata conversion type.
+ * @offset:	the address offset of secure buffer.
+ * @size:	the secure buffer size.
+ * @port:	the HW port id to M4U TA.
+ *
+ * Return: 0 for success; else the error code is returned
+ */
+int cmdq_sec_pkt_write(struct cmdq_pkt *pkt, u32 addr, u64 base,
+		       const enum cmdq_iwc_addr_metadata_type type,
+		       const u32 offset, const u32 size, const u32 port);
+
+#endif /* __MTK_CMDQ_SEC_MAILBOX_H__ */
diff --git a/include/linux/mailbox/mtk-cmdq-sec-tee.h b/include/linux/mailbox/mtk-cmdq-sec-tee.h
new file mode 100644
index 0000000000000..d2c97a137e01f
--- /dev/null
+++ b/include/linux/mailbox/mtk-cmdq-sec-tee.h
@@ -0,0 +1,105 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2019 MediaTek Inc.
+ */
+
+#ifndef __MTK_CMDQ_SEC_TEE_H__
+#define __MTK_CMDQ_SEC_TEE_H__
+
+#include <linux/types.h>
+#include <linux/delay.h>
+#include <linux/tee_drv.h>
+
+/**
+ * struct cmdq_sec_tee_context - context for tee vendor
+ * @uuid: Universally Unique Identifier of secure world.
+ * @tee_context: basic tee context.
+ * @session: session handle.
+ * @shared_mem: shared memory.
+ */
+struct cmdq_sec_tee_context {
+	u8			uuid[TEE_IOCTL_UUID_LEN];
+	struct tee_context	*tee_context;
+	u32			session;
+	struct tee_shm		*shared_mem;
+};
+
+/**
+ * cmdq_sec_setup_tee_context() - setup the uuid for the tee context to communicate with
+ * @tee:	context for tee vendor
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+void cmdq_sec_setup_tee_context(struct cmdq_sec_tee_context *tee);
+
+/**
+ * cmdq_sec_init_context() - initialize the tee context
+ * @tee:	context for tee vendor
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_init_context(struct cmdq_sec_tee_context *tee);
+
+/**
+ * cmdq_sec_deinit_context() - de-initialize the tee context
+ * @tee:	context for tee vendor
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_deinit_context(struct cmdq_sec_tee_context *tee);
+
+/**
+ * cmdq_sec_allocate_wsm() - allocate the world share memory to pass message to tee
+ * @tee:	context for tee vendor
+ * @wsm_buffer:	world share memory buffer with parameters pass to tee
+ * @size:	size to allocate
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_allocate_wsm(struct cmdq_sec_tee_context *tee, void **wsm_buffer, u32 size);
+
+/**
+ * cmdq_sec_free_wsm() - free the world share memory
+ * @tee:	context for tee vendor
+ * @wsm_buffer:	world share memory buffer with parameters pass to tee
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_free_wsm(struct cmdq_sec_tee_context *tee, void **wsm_buffer);
+
+/**
+ * cmdq_sec_open_session() - open session to the tee context
+ * @tee:	context for tee vendor
+ * @wsm_buffer:	world share memory buffer with parameters pass to tee
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_open_session(struct cmdq_sec_tee_context *tee, void *wsm_buffer);
+
+/**
+ * cmdq_sec_close_session() - close session to the tee context
+ * @tee:	context for tee vendor
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_close_session(struct cmdq_sec_tee_context *tee);
+
+/**
+ * cmdq_sec_execute_session() - execute session to the tee context
+ * @tee:	context for tee vendor
+ * @cmd:	tee invoke cmd id
+ * @timeout_ms:	timeout ms to current tee invoke cmd
+ *
+ * Return: 0 for success; else the error code is returned
+ *
+ */
+int cmdq_sec_execute_session(struct cmdq_sec_tee_context *tee, u32 cmd, s32 timeout_ms);
+
+#endif	/* __MTK_CMDQ_SEC_TEE_H__ */
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 09/10] mailbox: mediatek: Add secure CMDQ driver support for CMDQ driver

[Shawn Sung]

From: "Jason-JH.Lin" <jason-jh.lin@mediatek.com>

CMDQ driver will probe a secure CMDQ driver when has_sec flag
in platform data is true and its device node in dts has defined a
event id of CMDQ_SYNC_TOKEN_SEC_EOF.

Secure CMDQ driver support on mt8188 and mt8195 currently.
So add a has_secure flag to their driver data to probe it.

Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.com>
---
 drivers/mailbox/mtk-cmdq-mailbox.c | 69 +++++++++++++++++++++++++++++-
 1 file changed, 68 insertions(+), 1 deletion(-)

diff --git a/drivers/mailbox/mtk-cmdq-mailbox.c b/drivers/mailbox/mtk-cmdq-mailbox.c
index e04302ca6ec03..a51140404d116 100644
--- a/drivers/mailbox/mtk-cmdq-mailbox.c
+++ b/drivers/mailbox/mtk-cmdq-mailbox.c
@@ -15,8 +15,8 @@
 #include <linux/pm_runtime.h>
 #include <linux/mailbox_controller.h>
 #include <linux/mailbox/mtk-cmdq-mailbox.h>
+#include <linux/mailbox/mtk-cmdq-sec-mailbox.h>
 #include <linux/of.h>
-
 #define CMDQ_MBOX_AUTOSUSPEND_DELAY_MS	100
 
 #define CMDQ_OP_CODE_MASK		(0xff << CMDQ_OP_CODE_SHIFT)
@@ -55,11 +55,19 @@
 #define CMDQ_JUMP_BY_OFFSET		0x10000000
 #define CMDQ_JUMP_BY_PA			0x10000001
 
+#define CMDQ_IS_SECURE_THREAD(idx, cmdq) (cmdq->pdata->has_secure && \
+					  idx >= cmdq->pdata->secure_thread_min && \
+					  idx < cmdq->pdata->secure_thread_min + \
+					  cmdq->pdata->secure_thread_nr)
+
 struct gce_plat {
 	u32 thread_nr;
 	u8 shift;
 	bool control_by_sw;
 	bool sw_ddr_en;
+	bool has_secure;
+	u32 secure_thread_nr;
+	u32 secure_thread_min;
 	u32 gce_num;
 };
 
@@ -377,6 +385,13 @@ static int cmdq_mbox_send_data(struct mbox_chan *chan, void *data)
 	if (ret < 0)
 		return ret;
 
+	if (CMDQ_IS_SECURE_THREAD(thread->idx, cmdq)) {
+		ret = cmdq_sec_mbox.ops->send_data(chan, data);
+		pm_runtime_mark_last_busy(cmdq->mbox.dev);
+		pm_runtime_put_autosuspend(cmdq->mbox.dev);
+		return ret;
+	}
+
 	task = kzalloc(sizeof(*task), GFP_ATOMIC);
 	if (!task) {
 		pm_runtime_put_autosuspend(cmdq->mbox.dev);
@@ -436,6 +451,12 @@ static int cmdq_mbox_send_data(struct mbox_chan *chan, void *data)
 
 static int cmdq_mbox_startup(struct mbox_chan *chan)
 {
+	struct cmdq *cmdq = dev_get_drvdata(chan->mbox->dev);
+	struct cmdq_thread *thread = (struct cmdq_thread *)chan->con_priv;
+
+	if (CMDQ_IS_SECURE_THREAD(thread->idx, cmdq))
+		cmdq_sec_mbox.ops->startup(chan);
+
 	return 0;
 }
 
@@ -448,6 +469,13 @@ static void cmdq_mbox_shutdown(struct mbox_chan *chan)
 
 	WARN_ON(pm_runtime_get_sync(cmdq->mbox.dev));
 
+	if (CMDQ_IS_SECURE_THREAD(thread->idx, cmdq)) {
+		cmdq_sec_mbox.ops->shutdown(chan);
+		pm_runtime_mark_last_busy(cmdq->mbox.dev);
+		pm_runtime_put_autosuspend(cmdq->mbox.dev);
+		return;
+	}
+
 	spin_lock_irqsave(&thread->chan->lock, flags);
 	if (list_empty(&thread->task_busy_list))
 		goto done;
@@ -494,6 +522,13 @@ static int cmdq_mbox_flush(struct mbox_chan *chan, unsigned long timeout)
 	if (ret < 0)
 		return ret;
 
+	if (CMDQ_IS_SECURE_THREAD(thread->idx, cmdq)) {
+		cmdq_sec_mbox.ops->flush(chan, timeout);
+		pm_runtime_mark_last_busy(cmdq->mbox.dev);
+		pm_runtime_put_autosuspend(cmdq->mbox.dev);
+		return 0;
+	}
+
 	spin_lock_irqsave(&thread->chan->lock, flags);
 	if (list_empty(&thread->task_busy_list))
 		goto out;
@@ -569,6 +604,7 @@ static int cmdq_probe(struct platform_device *pdev)
 	int alias_id = 0;
 	static const char * const clk_name = "gce";
 	static const char * const clk_names[] = { "gce0", "gce1" };
+	u32 hwid = 0;
 
 	cmdq = devm_kzalloc(dev, sizeof(*cmdq), GFP_KERNEL);
 	if (!cmdq)
@@ -594,6 +630,8 @@ static int cmdq_probe(struct platform_device *pdev)
 		dev, cmdq->base, cmdq->irq);
 
 	if (cmdq->pdata->gce_num > 1) {
+		hwid = of_alias_get_id(dev->of_node, clk_name);
+
 		for_each_child_of_node(phandle->parent, node) {
 			alias_id = of_alias_get_id(node, clk_name);
 			if (alias_id >= 0 && alias_id < cmdq->pdata->gce_num) {
@@ -643,6 +681,29 @@ static int cmdq_probe(struct platform_device *pdev)
 		cmdq->mbox.chans[i].con_priv = (void *)&cmdq->thread[i];
 	}
 
+	if (cmdq->pdata->has_secure) {
+		struct platform_device *cmdq_sec;
+		static struct gce_sec_plat sec_plat = {0};
+
+		if (of_property_read_u32_index(dev->of_node, "mediatek,gce-events", 0,
+					       &sec_plat.cmdq_event) == 0) {
+			sec_plat.mbox = &cmdq->mbox;
+			sec_plat.base = cmdq->base;
+			sec_plat.hwid = hwid;
+			sec_plat.secure_thread_nr = cmdq->pdata->secure_thread_nr;
+			sec_plat.secure_thread_min = cmdq->pdata->secure_thread_min;
+
+			cmdq_sec = platform_device_register_data(dev, "mtk-cmdq-sec",
+								 PLATFORM_DEVID_AUTO,
+								 &sec_plat,
+								 sizeof(sec_plat));
+			if (IS_ERR(cmdq_sec)) {
+				dev_err(dev, "failed to register platform_device mtk-cmdq-sec\n");
+				return PTR_ERR(cmdq_sec);
+			}
+		}
+	}
+
 	err = devm_mbox_controller_register(dev, &cmdq->mbox);
 	if (err < 0) {
 		dev_err(dev, "failed to register mailbox: %d\n", err);
@@ -719,6 +780,9 @@ static const struct gce_plat gce_plat_mt8188 = {
 	.thread_nr = 32,
 	.shift = 3,
 	.control_by_sw = true,
+	.has_secure = true,
+	.secure_thread_nr = 2,
+	.secure_thread_min = 8,
 	.gce_num = 2
 };
 
@@ -733,6 +797,9 @@ static const struct gce_plat gce_plat_mt8195 = {
 	.thread_nr = 24,
 	.shift = 3,
 	.control_by_sw = true,
+	.has_secure = true,
+	.secure_thread_nr = 2,
+	.secure_thread_min = 8,
 	.gce_num = 2
 };
 
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v5 10/10] drm/mediatek: Add interface to allocate MediaTek GEM buffer.

[Shawn Sung]

From: CK Hu <ck.hu@mediatek.com>

Add an interface to allocate MediaTek GEM buffers, allow the IOCTLs
to be used by render nodes.
This patch also sets the RENDER driver feature.

Signed-off-by: CK Hu <ck.hu@mediatek.com>
Signed-off-by: Hsiao Chien Sung <shawn.sung@mediatek.corp-partner.google.com>
---
 drivers/gpu/drm/mediatek/mtk_drm_drv.c | 16 ++++++-
 drivers/gpu/drm/mediatek/mtk_gem.c     | 40 ++++++++++++++++
 drivers/gpu/drm/mediatek/mtk_gem.h     | 11 +++++
 include/uapi/drm/mediatek_drm.h        | 64 ++++++++++++++++++++++++++
 4 files changed, 130 insertions(+), 1 deletion(-)
 create mode 100644 include/uapi/drm/mediatek_drm.h

diff --git a/drivers/gpu/drm/mediatek/mtk_drm_drv.c b/drivers/gpu/drm/mediatek/mtk_drm_drv.c
index b5f605751b0a1..41eed3f89316c 100644
--- a/drivers/gpu/drm/mediatek/mtk_drm_drv.c
+++ b/drivers/gpu/drm/mediatek/mtk_drm_drv.c
@@ -23,6 +23,7 @@
 #include <drm/drm_of.h>
 #include <drm/drm_probe_helper.h>
 #include <drm/drm_vblank.h>
+#include <drm/mediatek_drm.h>
 
 #include "mtk_crtc.h"
 #include "mtk_ddp_comp.h"
@@ -570,6 +571,14 @@ static void mtk_drm_kms_deinit(struct drm_device *drm)
 	component_unbind_all(drm->dev, drm);
 }
 
+static const struct drm_ioctl_desc mtk_ioctls[] = {
+	DRM_IOCTL_DEF_DRV(MTK_GEM_CREATE, mtk_gem_create_ioctl,
+			  DRM_UNLOCKED | DRM_AUTH | DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(MTK_GEM_MAP_OFFSET,
+			  mtk_gem_map_offset_ioctl,
+			  DRM_UNLOCKED | DRM_AUTH | DRM_RENDER_ALLOW),
+};
+
 DEFINE_DRM_GEM_FOPS(mtk_drm_fops);
 
 /*
@@ -585,12 +594,17 @@ static struct drm_gem_object *mtk_gem_prime_import(struct drm_device *dev,
 }
 
 static const struct drm_driver mtk_drm_driver = {
-	.driver_features = DRIVER_MODESET | DRIVER_GEM | DRIVER_ATOMIC,
+	.driver_features = DRIVER_MODESET | DRIVER_GEM | DRIVER_ATOMIC |
+			   DRIVER_RENDER,
 
 	.dumb_create = mtk_gem_dumb_create,
 
 	.gem_prime_import = mtk_gem_prime_import,
 	.gem_prime_import_sg_table = mtk_gem_prime_import_sg_table,
+
+	.ioctls = mtk_ioctls,
+	.num_ioctls = ARRAY_SIZE(mtk_ioctls),
+
 	.fops = &mtk_drm_fops,
 
 	.name = DRIVER_NAME,
diff --git a/drivers/gpu/drm/mediatek/mtk_gem.c b/drivers/gpu/drm/mediatek/mtk_gem.c
index 5a82d7cf3ed0d..e59e0727717b7 100644
--- a/drivers/gpu/drm/mediatek/mtk_gem.c
+++ b/drivers/gpu/drm/mediatek/mtk_gem.c
@@ -4,9 +4,11 @@
  */
 
 #include <linux/dma-buf.h>
+#include <drm/mediatek_drm.h>
 
 #include <drm/drm.h>
 #include <drm/drm_device.h>
+#include <drm/drm_drv.h>
 #include <drm/drm_gem.h>
 #include <drm/drm_gem_dma_helper.h>
 #include <drm/drm_prime.h>
@@ -285,3 +287,41 @@ void mtk_gem_prime_vunmap(struct drm_gem_object *obj, struct iosys_map *map)
 	mtk_gem->kvaddr = NULL;
 	kfree(mtk_gem->pages);
 }
+
+int mtk_gem_map_offset_ioctl(struct drm_device *drm, void *data,
+			     struct drm_file *file_priv)
+{
+	struct drm_mtk_gem_map_off *args = data;
+
+	return drm_gem_dumb_map_offset(file_priv, drm, args->handle,
+				       &args->offset);
+}
+
+int mtk_gem_create_ioctl(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv)
+{
+	struct mtk_gem_obj *mtk_gem;
+	struct drm_mtk_gem_create *args = data;
+	int ret;
+
+	mtk_gem = mtk_gem_create(dev, args->size, false);
+	if (IS_ERR(mtk_gem))
+		return PTR_ERR(mtk_gem);
+
+	/*
+	 * allocate a id of idr table where the obj is registered
+	 * and handle has the id what user can see.
+	 */
+	ret = drm_gem_handle_create(file_priv, &mtk_gem->base, &args->handle);
+	if (ret)
+		goto err_handle_create;
+
+	/* drop reference from allocate - handle holds it now. */
+	drm_gem_object_put(&mtk_gem->base);
+
+	return 0;
+
+err_handle_create:
+	mtk_gem_free_object(&mtk_gem->base);
+	return ret;
+}
diff --git a/drivers/gpu/drm/mediatek/mtk_gem.h b/drivers/gpu/drm/mediatek/mtk_gem.h
index 66e5f154f6980..4d7598220ca8f 100644
--- a/drivers/gpu/drm/mediatek/mtk_gem.h
+++ b/drivers/gpu/drm/mediatek/mtk_gem.h
@@ -45,4 +45,15 @@ struct drm_gem_object *mtk_gem_prime_import_sg_table(struct drm_device *dev,
 int mtk_gem_prime_vmap(struct drm_gem_object *obj, struct iosys_map *map);
 void mtk_gem_prime_vunmap(struct drm_gem_object *obj, struct iosys_map *map);
 
+/*
+ * request gem object creation and buffer allocation as the size
+ * that it is calculated with framebuffer information such as width,
+ * height and bpp.
+ */
+int mtk_gem_create_ioctl(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv);
+
+/* get buffer offset to map to user space. */
+int mtk_gem_map_offset_ioctl(struct drm_device *dev, void *data,
+			     struct drm_file *file_priv);
 #endif
diff --git a/include/uapi/drm/mediatek_drm.h b/include/uapi/drm/mediatek_drm.h
new file mode 100644
index 0000000000000..b0dea00bacbc4
--- /dev/null
+++ b/include/uapi/drm/mediatek_drm.h
@@ -0,0 +1,64 @@
+/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
+/*
+ * Copyright (c) 2015 MediaTek Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _UAPI_MEDIATEK_DRM_H
+#define _UAPI_MEDIATEK_DRM_H
+
+#include <drm/drm.h>
+
+#ifdef __KERNEL__
+#include <linux/types.h>
+#else
+#include <stdint.h>
+#endif
+
+/**
+ * User-desired buffer creation information structure.
+ *
+ * @size: user-desired memory allocation size.
+ *	- this size value would be page-aligned internally.
+ * @flags: user request for setting memory type or cache attributes.
+ * @handle: returned a handle to created gem object.
+ *	- this handle will be set by gem module of kernel side.
+ */
+struct drm_mtk_gem_create {
+	uint64_t size;
+	uint32_t flags;
+	uint32_t handle;
+};
+
+/**
+ * A structure for getting buffer offset.
+ *
+ * @handle: a pointer to gem object created.
+ * @pad: just padding to be 64-bit aligned.
+ * @offset: relatived offset value of the memory region allocated.
+ *     - this value should be set by user.
+ */
+struct drm_mtk_gem_map_off {
+	uint32_t handle;
+	uint32_t pad;
+	uint64_t offset;
+};
+
+#define DRM_MTK_GEM_CREATE		0x00
+#define DRM_MTK_GEM_MAP_OFFSET		0x01
+
+#define DRM_IOCTL_MTK_GEM_CREATE	DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_MTK_GEM_CREATE, struct drm_mtk_gem_create)
+
+#define DRM_IOCTL_MTK_GEM_MAP_OFFSET	DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_MTK_GEM_MAP_OFFSET, struct drm_mtk_gem_map_off)
+
+#endif /* _UAPI_MEDIATEK_DRM_H */
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel