Re: [PATCH v5 2/5] media: venus: vdec: AR50_LITE video core support

[Jorge Ramirez <jorge.ramirez@oss.qualcomm.com>]

On 10/07/25 14:25:33, Dikshita Agarwal wrote:
> 
> 
> On 7/9/2025 12:44 AM, Jorge Ramirez wrote:
> > On 30/06/25 12:17:32, Dikshita Agarwal wrote:
> >>
> >>
> >> On 6/27/2025 8:48 PM, Jorge Ramirez wrote:
> >>> On 27/06/25 18:17:27, Dikshita Agarwal wrote:
> >>>>
> >>>>
> >>>> On 6/26/2025 7:29 PM, Jorge Ramirez-Ortiz wrote:
> >>>>> The AR50_LITE is a streamlined variant of the AR50 video core, designed
> >>>>> for power and cost-efficient platforms.
> >>>>>
> >>>>> It supports hardware-accelerated decoding of H.264, HEVC, and VP9
> >>>>> formats.
> >>>>>
> >>>>> Co-developed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
> >>>>> Signed-off-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
> >>>>> Signed-off-by: Jorge Ramirez-Ortiz <jorge.ramirez@oss.qualcomm.com>
> >>>>> Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
> >>>>> ---
> >>>>>  drivers/media/platform/qcom/venus/core.c      | 11 ++-
> >>>>>  drivers/media/platform/qcom/venus/core.h      | 11 ++-
> >>>>>  drivers/media/platform/qcom/venus/firmware.c  |  8 +-
> >>>>>  drivers/media/platform/qcom/venus/helpers.c   | 80 +++++++++++++++++++
> >>>>>  drivers/media/platform/qcom/venus/helpers.h   |  2 +
> >>>>>  .../media/platform/qcom/venus/hfi_helper.h    | 10 ++-
> >>>>>  drivers/media/platform/qcom/venus/hfi_venus.c | 14 ++--
> >>>>>  .../media/platform/qcom/venus/pm_helpers.c    |  1 +
> >>>>>  drivers/media/platform/qcom/venus/vdec.c      | 15 ++--
> >>>>>  9 files changed, 128 insertions(+), 24 deletions(-)
> >>>>>
> >>>>> diff --git a/drivers/media/platform/qcom/venus/core.c b/drivers/media/platform/qcom/venus/core.c
> >>>>> index d305d74bb152..736ef53d988d 100644
> >>>>> --- a/drivers/media/platform/qcom/venus/core.c
> >>>>> +++ b/drivers/media/platform/qcom/venus/core.c
> >>>>> @@ -254,14 +254,19 @@ static int venus_enumerate_codecs(struct venus_core *core, u32 type)
> >>>>>  
> >>>>>  static void venus_assign_register_offsets(struct venus_core *core)
> >>>>>  {
> >>>>> -	if (IS_IRIS2(core) || IS_IRIS2_1(core)) {
> >>>>> -		core->vbif_base = core->base + VBIF_BASE;
> >>>>> +	if (IS_IRIS2(core) || IS_IRIS2_1(core) || IS_AR50_LITE(core)) {
> >>>>>  		core->cpu_base = core->base + CPU_BASE_V6;
> >>>>>  		core->cpu_cs_base = core->base + CPU_CS_BASE_V6;
> >>>>>  		core->cpu_ic_base = core->base + CPU_IC_BASE_V6;
> >>>>>  		core->wrapper_base = core->base + WRAPPER_BASE_V6;
> >>>>>  		core->wrapper_tz_base = core->base + WRAPPER_TZ_BASE_V6;
> >>>>> -		core->aon_base = core->base + AON_BASE_V6;
> >>>>> +		if (IS_AR50_LITE(core)) {
> >>>>> +			core->vbif_base = NULL;
> >>>>> +			core->aon_base = NULL;
> >>>>> +		} else {
> >>>>> +			core->vbif_base = core->base + VBIF_BASE;
> >>>>> +			core->aon_base = core->base + AON_BASE_V6;
> >>>>> +		}
> >>>>>  	} else {
> >>>>>  		core->vbif_base = core->base + VBIF_BASE;
> >>>>>  		core->cpu_base = core->base + CPU_BASE;
> >>>>> diff --git a/drivers/media/platform/qcom/venus/core.h b/drivers/media/platform/qcom/venus/core.h
> >>>>> index b412e0c5515a..e755a28e919b 100644
> >>>>> --- a/drivers/media/platform/qcom/venus/core.h
> >>>>> +++ b/drivers/media/platform/qcom/venus/core.h
> >>>>> @@ -382,6 +382,7 @@ enum venus_inst_modes {
> >>>>>   * @lock:	instance lock
> >>>>>   * @core:	a reference to the core struct
> >>>>>   * @clk_data:	clock data per core ID
> >>>>> + * @eosbufs:	a lit of EOS buffers
> >>>>>   * @dpbbufs:	a list of decoded picture buffers
> >>>>>   * @internalbufs:	a list of internal bufferes
> >>>>>   * @registeredbufs:	a list of registered capture bufferes
> >>>>> @@ -450,6 +451,7 @@ struct venus_inst {
> >>>>>  	struct mutex lock;
> >>>>>  	struct venus_core *core;
> >>>>>  	struct clock_data clk_data;
> >>>>> +	struct list_head eosbufs;
> >>>>>  	struct list_head dpbbufs;
> >>>>>  	struct list_head internalbufs;
> >>>>>  	struct list_head registeredbufs;
> >>>>> @@ -520,7 +522,14 @@ struct venus_inst {
> >>>>>  #define IS_V1(core)	((core)->res->hfi_version == HFI_VERSION_1XX)
> >>>>>  #define IS_V3(core)	((core)->res->hfi_version == HFI_VERSION_3XX)
> >>>>>  #define IS_V4(core)	((core)->res->hfi_version == HFI_VERSION_4XX)
> >>>>> -#define IS_V6(core)	((core)->res->hfi_version == HFI_VERSION_6XX)
> >>>>> +static inline bool IS_V6(struct venus_core *core)
> >>>>> +{
> >>>>> +	if (WARN_ON_ONCE(!core))
> >>>>> +		return false;
> >>>>> +
> >>>>> +	return core->res->hfi_version == HFI_VERSION_6XX ||
> >>>>> +	       core->res->hfi_version == HFI_VERSION_6XX_LITE;
> >>>>> +}
> >>>> It should be HFI_VERSION_4XX_LITE for AR50_LITE. 4XX represents SC7280 and
> >>>> SDM845 which are AR50.
> >>>
> >>> ah good information - where is this documented? I never found this
> >>> information... I'd appreciate if you could confirm with some document
> >>> for future reference.
> >>>
> >>>>>  
> >>>>>  #define IS_AR50(core)		((core)->res->vpu_version == VPU_VERSION_AR50)
> >>>>>  #define IS_AR50_LITE(core)	((core)->res->vpu_version == VPU_VERSION_AR50_LITE)
> >>>>> diff --git a/drivers/media/platform/qcom/venus/firmware.c b/drivers/media/platform/qcom/venus/firmware.c
> >>>>> index 66a18830e66d..f8dcef0426ac 100644
> >>>>> --- a/drivers/media/platform/qcom/venus/firmware.c
> >>>>> +++ b/drivers/media/platform/qcom/venus/firmware.c
> >>>>> @@ -30,7 +30,7 @@ static void venus_reset_cpu(struct venus_core *core)
> >>>>>  	u32 fw_size = core->fw.mapped_mem_size;
> >>>>>  	void __iomem *wrapper_base;
> >>>>>  
> >>>>> -	if (IS_IRIS2_1(core))
> >>>>> +	if (IS_IRIS2_1(core) || IS_AR50_LITE(core))
> >>>>>  		wrapper_base = core->wrapper_tz_base;
> >>>>>  	else
> >>>>>  		wrapper_base = core->wrapper_base;
> >>>>> @@ -42,7 +42,7 @@ static void venus_reset_cpu(struct venus_core *core)
> >>>>>  	writel(fw_size, wrapper_base + WRAPPER_NONPIX_START_ADDR);
> >>>>>  	writel(fw_size, wrapper_base + WRAPPER_NONPIX_END_ADDR);
> >>>>>  
> >>>>> -	if (IS_IRIS2_1(core)) {
> >>>>> +	if (IS_IRIS2_1(core) || IS_AR50_LITE(core)) {
> >>>>>  		/* Bring XTSS out of reset */
> >>>>>  		writel(0, wrapper_base + WRAPPER_TZ_XTSS_SW_RESET);
> >>>>>  	} else {
> >>>>> @@ -68,7 +68,7 @@ int venus_set_hw_state(struct venus_core *core, bool resume)
> >>>>>  	if (resume) {
> >>>>>  		venus_reset_cpu(core);
> >>>>>  	} else {
> >>>>> -		if (IS_IRIS2_1(core))
> >>>>> +		if (IS_IRIS2_1(core) || IS_AR50_LITE(core))
> >>>>>  			writel(WRAPPER_XTSS_SW_RESET_BIT,
> >>>>>  			       core->wrapper_tz_base + WRAPPER_TZ_XTSS_SW_RESET);
> >>>>>  		else
> >>>>> @@ -181,7 +181,7 @@ static int venus_shutdown_no_tz(struct venus_core *core)
> >>>>>  	void __iomem *wrapper_base = core->wrapper_base;
> >>>>>  	void __iomem *wrapper_tz_base = core->wrapper_tz_base;
> >>>>>  
> >>>>> -	if (IS_IRIS2_1(core)) {
> >>>>> +	if (IS_IRIS2_1(core) || IS_AR50_LITE(core)) {
> >>>>>  		/* Assert the reset to XTSS */
> >>>>>  		reg = readl(wrapper_tz_base + WRAPPER_TZ_XTSS_SW_RESET);
> >>>> No need to handle no-tz case. Pls drop.
> >>>
> >>> ok
> >>>
> >>>>>  		reg |= WRAPPER_XTSS_SW_RESET_BIT;
> >>>>> diff --git a/drivers/media/platform/qcom/venus/helpers.c b/drivers/media/platform/qcom/venus/helpers.c
> >>>>> index 8295542e1a7c..812bec9a05be 100644
> >>>>> --- a/drivers/media/platform/qcom/venus/helpers.c
> >>>>> +++ b/drivers/media/platform/qcom/venus/helpers.c
> >>>>> @@ -230,6 +230,79 @@ int venus_helper_alloc_dpb_bufs(struct venus_inst *inst)
> >>>>>  }
> >>>>>  EXPORT_SYMBOL_GPL(venus_helper_alloc_dpb_bufs);
> >>>>>  
> >>>>> +static void free_eos_buf(struct venus_inst *inst, struct intbuf *buf)
> >>>>> +{
> >>>>> +	list_del_init(&buf->list);
> >>>>> +	dma_free_attrs(inst->core->dev, buf->size, buf->va, buf->da,
> >>>>> +		       buf->attrs);
> >>>>> +	kfree(buf);
> >>>>> +}
> >>>>> +
> >>>>> +int venus_helper_free_eos_bufs(struct venus_inst *inst)
> >>>>> +{
> >>>>> +	struct intbuf *buf, *n;
> >>>>> +
> >>>>> +	list_for_each_entry_safe(buf, n, &inst->eosbufs, list) {
> >>>>> +		free_eos_buf(inst, buf);
> >>>>> +	}
> >>>>> +
> >>>>> +	if (list_empty(&inst->eosbufs))
> >>>>> +		INIT_LIST_HEAD(&inst->eosbufs);
> >>>>> +
> >>>>> +	return 0;
> >>>>> +}
> >>>>> +EXPORT_SYMBOL_GPL(venus_helper_free_eos_bufs);
> >>>>> +
> >>>>> +int venus_helper_alloc_eos_buf(struct venus_inst *inst,
> >>>>> +			       struct hfi_frame_data *data)
> >>>>> +{
> >>>>> +	struct venus_core *core = inst->core;
> >>>>> +	struct device *dev = core->dev;
> >>>>> +	struct intbuf *buf;
> >>>>> +	int ret = 0;
> >>>>> +
> >>>>> +	memset(data, 0, sizeof(*data));
> >>>>> +
> >>>>> +	data->buffer_type = HFI_BUFFER_INPUT;
> >>>>> +	data->flags = HFI_BUFFERFLAG_EOS;
> >>>>> +
> >>>>> +	if (IS_AR50_LITE(inst->core)) {
> >>>>> +		/* We must send valid sizes and addresses */
> >>>>> +		buf = kzalloc(sizeof(*buf), GFP_KERNEL);
> >>>>> +		if (!buf) {
> >>>>> +			ret = -ENOMEM;
> >>>>> +			goto fail;
> >>>>> +		}
> >>>>> +
> >>>>> +		buf->type = HFI_BUFFER_INPUT;
> >>>>> +		buf->size = SZ_4K;
> >>>>> +		buf->attrs = DMA_ATTR_NO_KERNEL_MAPPING;
> >>>>> +		buf->va = dma_alloc_attrs(dev, buf->size, &buf->da, GFP_KERNEL,
> >>>>> +					  buf->attrs);
> >>>>> +		if (!buf->va) {
> >>>>> +			ret = -ENOMEM;
> >>>>> +			goto fail;
> >>>>> +		}
> >>>>> +
> >>>>> +		list_add_tail(&buf->list, &inst->eosbufs);
> >>>>> +
> >>>>> +		data->alloc_len = buf->size;
> >>>>> +		data->device_addr = buf->da;
> >>>>> +
> >>>> why this special handling for eos buffer is needed for AR50_LITE?
> >>>
> >>> this _fix_ was develope through testing: without it there is no EOS and
> >>> errors are reporting upon killing the player
> >>>
> >> Would be better to see why there is no EOS from firmware,
> >> there shouldn't be the need to have a dma allocation for this dummy
> >> buffers, as there is no useful info in the buffer. Having the device
> >> address as 0 or 0xdeadb000 should be enough.
> >>
> > 
> > hi dikshita,
> > 
> > I am still keeping this on v6 as per our internal discussions and
> > because v6 is quite different from v5 so wanted to provide early access
> > to users.
> > 
> > if the firwmare is fixed to address this issue on time, I might revert
> > the EOS page buffer. 
> > 
> I'd prefer to resolve this via correct EOS handling or gain clarity on why
> AR50_LITE requires special treatment, instead of proceeding with new patch
> sets built around this design.
>
Fully agree.

However this patch is the actual proper implementation - it follows the
HFI spec - while the current code upstream is not.

We should revert over time the current implementation to avoid hitting
issues when the firmware stops checking for things like 0xdeadb000.