mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Akhil P Oommen <akhilpo@oss.qualcomm.com>
To: Puranam V G Tejaswi <puranam.tejaswi@oss.qualcomm.com>,
	Bjorn Andersson <andersson@kernel.org>,
	Konrad Dybcio <konradybcio@kernel.org>,
	Abel Vesa <abelvesa@kernel.org>,
	Rob Clark <robin.clark@oss.qualcomm.com>,
	Sean Paul <sean@poorly.run>, Dmitry Baryshkov <lumag@kernel.org>,
	Abhinav Kumar <abhinav.kumar@linux.dev>,
	Jessica Zhang <jesszhan0024@gmail.com>,
	Marijn Suijten <marijn.suijten@somainline.org>,
	David Airlie <airlied@gmail.com>, Simona Vetter <simona@ffwll.ch>,
	Maarten Lankhorst <maarten.lankhorst@linux.intel.com>,
	Maxime Ripard <mripard@kernel.org>,
	Thomas Zimmermann <tzimmermann@suse.de>,
	Rob Herring <robh@kernel.org>,
	Krzysztof Kozlowski <krzk+dt@kernel.org>,
	Conor Dooley <conor+dt@kernel.org>
Cc: linux-arm-msm@vger.kernel.org, dri-devel@lists.freedesktop.org,
	freedreno@lists.freedesktop.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 4/6] drm/msm/adreno: Program the SMMU aperture when the secure world can't
Date: Thu, 19 Nov 2026 00:18:03 +0530	[thread overview]
Message-ID: <d42b3653-e4b3-4a0c-814c-02ad511fdb85@oss.qualcomm.com> (raw)
In-Reply-To: <20261118-adreno-scm-fallback-v1-4-e99cdbf3121d@oss.qualcomm.com>



On 11/18/2026 6:38 PM, Puranam V G Tejaswi wrote:
> The CP switches pagetables by writing the GPU SMMU context bank
> registers itself. It reaches them through the CX SMMU aperture, a
> register block in the GPU, which has to point at context bank 0. The
> secure world normally owns this block and programs it when
> adreno_hw_init() calls qcom_scm_set_gpu_smmu_aperture().
> 
> Some platforms have a secure world that does not implement this. Nothing
> programs the aperture on these platforms, so the CP cannot switch
> pagetables.
> 
> If the SCM call is not supported, program the aperture from
> adreno_hw_init() instead. Point the GC aperture at context bank 0 and,
> on GPUs with LPAC, the LPAC aperture at context bank 1, the same way
> the secure world does. Disable the other apertures. The SCM call is
> still used whenever it is supported.
> 
> The values depend on where context bank 0 sits in the GPU register
> space, which differs between GPUs, so they come from a register list in
> the GPU catalog. The aperture region comes from DT. If DT describes the
> region but it cannot be mapped, fail probe. Add the lists for A702,
> whose entry Shikra's GPU also uses, and for A753.
> 
> Signed-off-by: Puranam V G Tejaswi <puranam.tejaswi@oss.qualcomm.com>
> ---
>  drivers/gpu/drm/msm/adreno/a6xx_catalog.c |  9 ++++++++
>  drivers/gpu/drm/msm/adreno/a7xx_catalog.c | 10 +++++++++
>  drivers/gpu/drm/msm/adreno/adreno_gpu.c   | 37 +++++++++++++++++++++++++++++++
>  drivers/gpu/drm/msm/adreno/adreno_gpu.h   | 21 ++++++++++++++++++
>  4 files changed, 77 insertions(+)
> 
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_catalog.c b/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
> index 4ef2bd6e7ad2..fe1902b8de16 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
> @@ -736,6 +736,14 @@ static const struct adreno_reglist a702_hwcg[] = {
>  	{}
>  };
>  
> +/* GPU aperture: GC to CB0, other apertures disabled */
> +static const struct adreno_reglist a702_aperture_reglist_regs[] = {
> +	{ 0x00, 0x000a8000 }, /* GC: CB0 */
> +	{ 0x08, 0x00000000 }, /* CTRL: GC only */

These offsets should be documented in the xml file instead of hardcoded values.
Otherwise LGTM.

-Akhil.
> +};
> +
> +DECLARE_ADRENO_REGPAIR_LIST(a702_aperture_reglist);
> +
>  static const struct adreno_info a6xx_gpus[] = {
>  	{
>  		.chip_ids = ADRENO_CHIP_IDS(0x06010000),
> @@ -1198,6 +1206,7 @@ static const struct adreno_info a6xx_gpus[] = {
>  		.quirks = ADRENO_QUIRK_HAS_HW_APRIV,
>  		.funcs = &a6xx_gmuwrapper_funcs,
>  		.zapfw = "a702_zap.mbn",
> +		.aperture_reglist = &a702_aperture_reglist,
>  		.a6xx = &(const struct a6xx_info) {
>  			.hwcg = a702_hwcg,
>  			.protect = &a650_protect,
> diff --git a/drivers/gpu/drm/msm/adreno/a7xx_catalog.c b/drivers/gpu/drm/msm/adreno/a7xx_catalog.c
> index 9465ffa3dd12..43d681aa3c2e 100644
> --- a/drivers/gpu/drm/msm/adreno/a7xx_catalog.c
> +++ b/drivers/gpu/drm/msm/adreno/a7xx_catalog.c
> @@ -279,6 +279,15 @@ static const struct adreno_reglist_pipe a7xx_dyn_pwrup_reglist_regs[] = {
>  
>  DECLARE_ADRENO_REGLIST_PIPE_LIST(a7xx_dyn_pwrup_reglist);
>  
> +/* GPU aperture: GC to CB0, LPAC to CB1, other apertures disabled */
> +static const struct adreno_reglist a753_aperture_reglist_regs[] = {
> +	{ 0x00, 0x000c0000 }, /* GC: CB0 */
> +	{ 0x04, 0x000c1000 }, /* LPAC: CB1 */
> +	{ 0x08, 0x00000002 }, /* CTRL: GC and LPAC */
> +};
> +
> +DECLARE_ADRENO_REGPAIR_LIST(a753_aperture_reglist);
> +
>  static const struct adreno_info a7xx_gpus[] = {
>  	{
>  		.chip_ids = ADRENO_CHIP_IDS(0x07030001),
> @@ -465,6 +474,7 @@ static const struct adreno_info a7xx_gpus[] = {
>  			  ADRENO_QUIRK_PREEMPTION |
>  			  ADRENO_QUIRK_IFPC,
>  		.funcs = &a7xx_gpu_funcs,
> +		.aperture_reglist = &a753_aperture_reglist,
>  		.a6xx = &(const struct a6xx_info) {
>  			.protect = &a730_protect,
>  			.pwrup_reglist = &a7xx_pwrup_reglist,
> diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> index 54f337475231..c3c84629c5f4 100644
> --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> @@ -675,6 +675,13 @@ int adreno_hw_init(struct msm_gpu *gpu)
>  		ret = qcom_scm_set_gpu_smmu_aperture(0);
>  		if (ret)
>  			DRM_DEV_ERROR(gpu->dev->dev, "unable to set SMMU aperture: %d\n", ret);
> +	} else if (adreno_gpu->aperture) {
> +		const struct adreno_regpair_list *list =
> +			adreno_gpu->info->aperture_reglist;
> +
> +		for (int i = 0; i < list->count; i++)
> +			writel(list->regs[i].value,
> +			       adreno_gpu->aperture + list->regs[i].offset);
>  	}
>  
>  	for (int i = 0; i < gpu->nr_rings; i++) {
> @@ -1193,6 +1200,32 @@ int adreno_read_speedbin(struct device *dev, u32 *speedbin)
>  	return nvmem_cell_read_variable_le_u32(dev, "speed_bin", speedbin);
>  }
>  
> +/*
> + * Map the GPU aperture, so that adreno_hw_init() can program it when
> + * the secure world does not.
> + */
> +static int adreno_gpu_aperture_init(struct platform_device *pdev,
> +				    struct adreno_gpu *adreno_gpu)
> +{
> +	struct resource *res;
> +	void __iomem *base;
> +
> +	if (!adreno_gpu->info->aperture_reglist)
> +		return 0;
> +
> +	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "aperture");
> +	if (!res)
> +		return 0;
> +
> +	base = devm_ioremap_resource(&pdev->dev, res);
> +	if (IS_ERR(base))
> +		return PTR_ERR(base);
> +
> +	adreno_gpu->aperture = base;
> +
> +	return 0;
> +}
> +
>  int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
>  		struct adreno_gpu *adreno_gpu,
>  		const struct adreno_gpu_funcs *funcs, int nr_rings)
> @@ -1238,6 +1271,10 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
>  
>  	adreno_gpu_config.ioname = "kgsl_3d0_reg_memory";
>  
> +	ret = adreno_gpu_aperture_init(pdev, adreno_gpu);
> +	if (ret)
> +		return ret;
> +
>  	adreno_gpu_config.nr_rings = nr_rings;
>  
>  	ret = adreno_get_pwrlevels(dev, gpu);
> diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
> index 2911d440ab17..021802c6830c 100644
> --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
> +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
> @@ -88,6 +88,19 @@ struct adreno_reglist {
>  	u32 value;
>  };
>  
> +struct adreno_regpair_list {
> +	/** @regs: List of register offset and value pairs **/
> +	const struct adreno_reglist *regs;
> +	/** @count: Number of registers in the list **/
> +	u32 count;
> +};
> +
> +#define DECLARE_ADRENO_REGPAIR_LIST(name)	\
> +static const struct adreno_regpair_list name = {		\
> +	.regs = name ## _regs,				\
> +	.count = ARRAY_SIZE(name ## _regs),		\
> +}
> +
>  /* Reglist with pipe information */
>  struct adreno_reglist_pipe {
>  	u32 offset;
> @@ -129,6 +142,11 @@ struct adreno_info {
>  	 */
>  	struct adreno_speedbin *speedbins;
>  	u64 preempt_record_size;
> +	/**
> +	 * @aperture_reglist: Optional GPU aperture register values. Used if
> +	 * the secure world does not program the aperture.
> +	 */
> +	const struct adreno_regpair_list *aperture_reglist;
>  };
>  
>  #define ADRENO_CHIP_IDS(tbl...) (uint32_t[]) { tbl, 0 }
> @@ -250,6 +268,9 @@ struct adreno_gpu {
>  
>  	bool has_ray_tracing;
>  
> +	/* GPU aperture, if the driver may need to program it */
> +	void __iomem *aperture;
> +
>  	u64 uche_trap_base;
>  };
>  #define to_adreno_gpu(x) container_of(x, struct adreno_gpu, base)
> 


  reply	other threads:[~2026-11-18 18:48 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18 13:08 [PATCH 0/6] drm/msm/adreno: Handle unavailable GPU SCM calls Puranam V G Tejaswi
2026-11-18 13:08 ` [PATCH 1/6] firmware: qcom: scm: Add qcom_scm_gpu_init_regs_is_available() Puranam V G Tejaswi
2026-11-18 18:42   ` Akhil P Oommen
2026-11-18 13:08 ` [PATCH 2/6] drm/msm/a6xx: Don't fail GMU resume if the GPU init regs SCM call is absent Puranam V G Tejaswi
2026-11-18 18:44   ` Akhil P Oommen
2026-11-18 13:08 ` [PATCH 3/6] dt-bindings: display/msm: gpu: Document the CX SMMU aperture region Puranam V G Tejaswi
2026-11-18 18:45   ` Akhil P Oommen
2026-11-18 13:08 ` [PATCH 4/6] drm/msm/adreno: Program the SMMU aperture when the secure world can't Puranam V G Tejaswi
2026-11-18 18:48   ` Akhil P Oommen [this message]
2026-11-18 13:08 ` [PATCH 5/6] arm64: dts: qcom: nord: Describe the GPU CX SMMU aperture Puranam V G Tejaswi
2026-11-18 18:48   ` Akhil P Oommen
2026-11-18 13:08 ` [PATCH 6/6] arm64: dts: qcom: shikra: " Puranam V G Tejaswi
2026-11-18 18:50   ` Akhil P Oommen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d42b3653-e4b3-4a0c-814c-02ad511fdb85@oss.qualcomm.com \
    --to=akhilpo@oss.qualcomm.com \
    --cc=abelvesa@kernel.org \
    --cc=abhinav.kumar@linux.dev \
    --cc=airlied@gmail.com \
    --cc=andersson@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=freedreno@lists.freedesktop.org \
    --cc=jesszhan0024@gmail.com \
    --cc=konradybcio@kernel.org \
    --cc=krzk+dt@kernel.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lumag@kernel.org \
    --cc=maarten.lankhorst@linux.intel.com \
    --cc=marijn.suijten@somainline.org \
    --cc=mripard@kernel.org \
    --cc=puranam.tejaswi@oss.qualcomm.com \
    --cc=robh@kernel.org \
    --cc=robin.clark@oss.qualcomm.com \
    --cc=sean@poorly.run \
    --cc=simona@ffwll.ch \
    --cc=tzimmermann@suse.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®