Hi Alex,
Thanks for the review!
On 1/23/19 8:10 AM, Alexandre Courbot wrote:
> Sorry for the delayed review! >_<
>
> On Wed, Jan 9, 2019 at 5:46 PM Stanimir Varbanov
> <stanimir.varbanov@xxxxxxxxxx> wrote:
>>
>> By historical reasons we defined firmware memory size to be 6MB even
>> that the firmware size for all supported Venus versions is 5MBs. Correct
>> that by compare the required firmware size returned from mdt loader and
>> the one provided by DT reserved memory region. We proceed further if the
>> required firmware size is smaller than provided by DT memory region.
>>
>> Signed-off-by: Stanimir Varbanov <stanimir.varbanov@xxxxxxxxxx>
>> ---
>> drivers/media/platform/qcom/venus/core.h | 1 +
>> drivers/media/platform/qcom/venus/firmware.c | 54 +++++++++++---------
>> 2 files changed, 31 insertions(+), 24 deletions(-)
>>
>> diff --git a/drivers/media/platform/qcom/venus/core.h b/drivers/media/platform/qcom/venus/core.h
>> index 6382cea29185..79c7e816c706 100644
>> --- a/drivers/media/platform/qcom/venus/core.h
>> +++ b/drivers/media/platform/qcom/venus/core.h
>> @@ -134,6 +134,7 @@ struct venus_core {
>> struct video_firmware {
>> struct device *dev;
>> struct iommu_domain *iommu_domain;
>> + size_t mapped_mem_size;
>> } fw;
>> struct mutex lock;
>> struct list_head instances;
>> diff --git a/drivers/media/platform/qcom/venus/firmware.c b/drivers/media/platform/qcom/venus/firmware.c
>> index c29acfd70c1b..6b509ffd022a 100644
>> --- a/drivers/media/platform/qcom/venus/firmware.c
>> +++ b/drivers/media/platform/qcom/venus/firmware.c
>> @@ -35,14 +35,15 @@
>>
>> static void venus_reset_cpu(struct venus_core *core)
>> {
>> + u32 fw_size = core->fw.mapped_mem_size;
>> void __iomem *base = core->base;
>>
>> writel(0, base + WRAPPER_FW_START_ADDR);
>> - writel(VENUS_FW_MEM_SIZE, base + WRAPPER_FW_END_ADDR);
>> + writel(fw_size, base + WRAPPER_FW_END_ADDR);
>> writel(0, base + WRAPPER_CPA_START_ADDR);
>> - writel(VENUS_FW_MEM_SIZE, base + WRAPPER_CPA_END_ADDR);
>> - writel(VENUS_FW_MEM_SIZE, base + WRAPPER_NONPIX_START_ADDR);
>> - writel(VENUS_FW_MEM_SIZE, base + WRAPPER_NONPIX_END_ADDR);
>> + writel(fw_size, base + WRAPPER_CPA_END_ADDR);
>> + writel(fw_size, base + WRAPPER_NONPIX_START_ADDR);
>> + writel(fw_size, base + WRAPPER_NONPIX_END_ADDR);
>> writel(0x0, base + WRAPPER_CPU_CGC_DIS);
>> writel(0x0, base + WRAPPER_CPU_CLOCK_CONFIG);
>>
>> @@ -74,6 +75,9 @@ static int venus_load_fw(struct venus_core *core, const char *fwname,
>> void *mem_va;
>> int ret;
>>
>> + *mem_phys = 0;
>> + *mem_size = 0;
>> +
>> dev = core->dev;
>> node = of_parse_phandle(dev->of_node, "memory-region", 0);
>> if (!node) {
>> @@ -85,28 +89,30 @@ static int venus_load_fw(struct venus_core *core, const char *fwname,
>> if (ret)
>> return ret;
>>
>> + ret = request_firmware(&mdt, fwname, dev);
>> + if (ret < 0)
>> + return ret;
>> +
>> + fw_size = qcom_mdt_get_size(mdt);
>> + if (fw_size < 0) {
>> + ret = fw_size;
>> + goto err_release_fw;
>> + }
>> +
>> *mem_phys = r.start;
>> *mem_size = resource_size(&r);
>>
>> - if (*mem_size < VENUS_FW_MEM_SIZE)
>> - return -EINVAL;
>> + if (*mem_size < fw_size || fw_size > VENUS_FW_MEM_SIZE) {
>
> Do we still need to check for fw_size > VENUS_FW_MEM_SIZE ? If we
> don't then we can remove the definition of VENUS_FW_MEM_SIZE
> altogether.
I know, it is a bit paranoid but I'd want to avoid if someone set
unreasonable memory size. So I'd like to have some sanitized firmware
region size in the driver.
>
>> + ret = -EINVAL;
>> + goto err_release_fw;
>> + }
>>
>> mem_va = memremap(r.start, *mem_size, MEMREMAP_WC);
>> if (!mem_va) {
>> dev_err(dev, "unable to map memory region: %pa+%zx\n",
>> &r.start, *mem_size);
>> - return -ENOMEM;
>> - }
>> -
>> - ret = request_firmware(&mdt, fwname, dev);
>> - if (ret < 0)
>> - goto err_unmap;
>> -
>> - fw_size = qcom_mdt_get_size(mdt);
>> - if (fw_size < 0) {
>> - ret = fw_size;
>> - release_firmware(mdt);
>> - goto err_unmap;
>> + ret = -ENOMEM;
>> + goto err_release_fw;
>> }
>>
>> if (core->use_tz)
>> @@ -116,10 +122,9 @@ static int venus_load_fw(struct venus_core *core, const char *fwname,
>> ret = qcom_mdt_load_no_init(dev, mdt, fwname, VENUS_PAS_ID,
>> mem_va, *mem_phys, *mem_size, NULL);
>>
>> - release_firmware(mdt);
>> -
>> -err_unmap:
>> memunmap(mem_va);
>> +err_release_fw:
>> + release_firmware(mdt);
>> return ret;
>> }
>>
>> @@ -135,6 +140,7 @@ static int venus_boot_no_tz(struct venus_core *core, phys_addr_t mem_phys,
>> return -EPROBE_DEFER;
>>
>> iommu = core->fw.iommu_domain;
>> + core->fw.mapped_mem_size = mem_size;
>>
>> ret = iommu_map(iommu, VENUS_FW_START_ADDR, mem_phys, mem_size,
>> IOMMU_READ | IOMMU_WRITE | IOMMU_PRIV);
>> @@ -151,7 +157,7 @@ static int venus_boot_no_tz(struct venus_core *core, phys_addr_t mem_phys,
>> static int venus_shutdown_no_tz(struct venus_core *core)
>> {
>> struct iommu_domain *iommu;
>> - size_t unmapped;
>> + size_t unmapped, mapped = core->fw.mapped_mem_size;
>
> mapped should probably be const here.
Ok.
>
> With these minor comments:
>
> Reviewed-by: Alexandre Courbot <acourbot@xxxxxxxxxxxx>
> Tested-by: Alexandre Courbot <acourbot@xxxxxxxxxxxx>
>
> For the 4 patches in this series. I could see the improvement in
> decoder performance introduced by patches 2 and 3, thanks!
>
--
regards,
Stan
