Re: [PATCH v2 2/2] media: imagination: Add E5010 JPEG Encoder driver

From: Devarsh Thakkar
Date: Tue Aug 08 2023 - 15:15:13 EST


Hi Krzysztof,

Thanks for the review.

On 27/07/23 17:43, Krzysztof Kozlowski wrote:
> On 27/07/2023 13:25, Devarsh Thakkar wrote:
>> This adds support for stateful V4L2 M2M based driver
>> for Imagination E5010 JPEG Encoder [1] which supports baseline
>> encoding with two different quantization tables and compression
>> ratio as demanded.
>>
>> Support for both contigous and non-contigous YUV420 and YUV422
>> semiplanar formats is added along with alignment restrictions
>> as required by the hardware.
>>
>> System and runtime PM hooks are added in the driver along with v4l2
>> crop and selection API support.
>>
>> Minimum resolution supported is 64x64 and
>> Maximum resolution supported is 8192x8192.
>>
>
>
> ...
>
>> +
>> +static int e5010_release(struct file *file)
>> +{
>> + struct e5010_dev *dev = video_drvdata(file);
>> + struct e5010_context *ctx = file->private_data;
>> +
>> + dprintk(dev, 1, "Releasing instance: 0x%p, m2m_ctx: 0x%p\n", ctx, ctx->fh.m2m_ctx);
>
> Why do you print pointers? Looks like code is buggy and you still keep
> debugging it.
>

This are dynamic debug prints to debug/analyze v4l2 drivers, It's a general
mechanism used in v4l2 drivers for debugging as Nicholas mentioned.

>> + mutex_lock(&dev->mutex);
>> + v4l2_ctrl_handler_free(&ctx->ctrl_handler);
>> + v4l2_m2m_ctx_release(ctx->fh.m2m_ctx);
>> + v4l2_fh_del(&ctx->fh);
>> + v4l2_fh_exit(&ctx->fh);
>> + kfree(ctx);
>> + mutex_unlock(&dev->mutex);
>> +
>> + return 0;
>> +}
>> +
>
> ...
>
>> +static int e5010_init_device(struct e5010_dev *dev)
>> +{
>> + int ret = 0;
>> +
>> + /*TODO: Set MMU in bypass mode until support for the same is added in driver*/
>> + e5010_hw_bypass_mmu(dev->mmu_base, 1);
>> +
>> + if (e5010_hw_enable_auto_clock_gating(dev->jasper_base, 1))
>> + dev_warn(dev->dev, "Failed to enable auto clock gating\n");
>> +
>> + if (e5010_hw_enable_manual_clock_gating(dev->jasper_base, 0))
>> + dev_warn(dev->dev, "Failed to disable manual clock gating\n");
>> +
>> + if (e5010_hw_enable_crc_check(dev->jasper_base, 0))
>> + dev_warn(dev->dev, "Failed to disable CRC check\n");
>> +
>> + if (e5010_hw_enable_output_address_error_irq(dev->jasper_base, 1))
>> + dev_err(dev->dev, "Failed to enable Output Address Error interrupts\n");
>> +
>> + ret = e5010_hw_set_input_source_to_memory(dev->jasper_base, 1);
>> + if (ret) {
>> + dev_err(dev->dev, "Failed to set input source to memory\n");
>> + goto fail;
>
> retturn ret;
>
>> + }
>> +
>> + ret = e5010_hw_enable_picture_done_irq(dev->jasper_base, 1);
>> + if (ret)
>> + dev_err(dev->dev, "Failed to enable Picture Done interrupts\n");
>> +fail:
>> + return ret;
>> +}
>> +
>> +static int e5010_probe(struct platform_device *pdev)
>> +{
>> + const struct of_device_id *of_dev_id;
>> + struct e5010_dev *dev;
>
> dev is struct device, so call it differently.
>

Agreed, will change the name.

>> + struct resource *res;
>> + int irq, ret = 0;
>> +
>> + of_dev_id = of_match_device(e5010_of_match, &pdev->dev);
>> + if (!of_dev_id) {
>> + dev_err(&pdev->dev, "%s: Unable to match device\n", __func__);
>
> I don't think this can happen.

Ok, will remove of_match_device altogether as not needed.

>
>> + return -ENODEV;
>> + }
>> +
>> + ret = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
>> + if (ret) {
>> + dev_err(&pdev->dev, "32-bit consistent DMA enable failed\n");
>> + return ret;
>> + }
>> +
>> + dev = devm_kzalloc(&pdev->dev, sizeof(*dev), GFP_KERNEL);
>> + if (!dev)
>> + return -ENOMEM;
>> +
>> + platform_set_drvdata(pdev, dev);
>> +
>> + dev->dev = &pdev->dev;
>> +
>> + mutex_init(&dev->mutex);
>> + spin_lock_init(&dev->hw_lock);
>> +
>> + dev->vdev = &e5010_videodev;
>> + dev->vdev->v4l2_dev = &dev->v4l2_dev;
>> + dev->vdev->lock = &dev->mutex;
>> + dev->vdev->queue = NULL;
>> + dev->vdev->prio = NULL;
>> + dev->vdev->dev_parent = NULL;
>> + dev->vdev->minor = -1;
>> +
>> + ret = v4l2_device_register(&pdev->dev, &dev->v4l2_dev);
>> + if (ret) {
>> + dev_err(dev->dev, "Failed to register v4l2 device\n");
>> + return ret;
>
> return dev_err_probe
>

Ok, will check on this.

>> + }
>> +
>> + dev->m2m_dev = v4l2_m2m_init(&e5010_m2m_ops);
>> + if (!dev->m2m_dev) {
>
> This cannot happen. Read the function.

Agreed will use IS_ERR here.
>
>> + dev_err(dev->dev, "Failed to initialize m2m device\n");
>> + ret = -ENOMEM;
>> + goto fail_after_v4l2_register;
>> + }
>> +
>> + video_set_drvdata(dev->vdev, dev);
>> +
>> + ret = video_register_device(dev->vdev, VFL_TYPE_VIDEO, 0);
>> + if (ret) {
>> + dev_err(dev->dev, "Failed to register video device\n");
>> + ret = -ENOMEM;
>
> Why?
>

Will propagate the return value instead.

>> + goto fail_after_v4l2_register;
>> + }
>> +
>> + dev_info(dev->dev, "Device registered as /dev/video%d\n",
>> + dev->vdev->num);
>> +
>> + res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "regjasper");
>> + if (!res) {
>> + dev_err(dev->dev, "Missing 'regjasper' resources area\n");
>> + ret = -ENOMEM;
>> + goto fail_after_video_register_device;
>> + }
>> + dev->jasper_base = devm_ioremap_resource(&pdev->dev, res);
>
> Use helper function to combine two calls into one.

Ok will try using devm_platform_ioremap_resource_byname().

>
>> + if (!dev->jasper_base) {
>> + ret = -ENOMEM;
>
> This shouldn't be ENOMEM
>

Ok, will return the error.

>> + goto fail_after_video_register_device;
>> + }
>> +
>> + res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "regmmu");
>> + if (!res) {
>> + dev_err(dev->dev, "Missing 'regmmu' resources area\n");
>> + ret = -ENOMEM;
>> + goto fail_after_video_register_device;
>> + }
>> + dev->mmu_base = devm_ioremap_resource(&pdev->dev, res);
>> + if (!dev->mmu_base) {
>> + ret = -ENOMEM;
>
> The same.
>

Ok, will return the error.
>> + goto fail_after_video_register_device;
>> + }
>> +
>> + dev->last_context_run = NULL;
>> +
>> + irq = platform_get_irq(pdev, 0);
>> + ret = devm_request_irq(dev->dev, irq, e5010_irq, 0,
>> + E5010_MODULE_NAME, dev);
>> + if (ret) {
>> + dev_err(dev->dev, "Failed to register IRQ %d\n", irq);
>> + goto fail_after_video_register_device;
>> + }
>> +
>> + dev->clk = devm_clk_get(&pdev->dev, "core_clk");
>> + if (IS_ERR(dev->clk)) {
>> + dev_err(dev->dev, "failed to get clock\n");
>> + ret = PTR_ERR(dev->clk);
>
> ret = dev_err_probe
>

Ok, will use above helper.

>> + goto fail_after_video_register_device;
>> + }
>> +
>> + pm_runtime_enable(dev->dev);
>> +
>> + return 0;
>> +
>> +fail_after_video_register_device:
>> + v4l2_m2m_release(dev->m2m_dev);
>> +fail_after_v4l2_register:
>> + v4l2_device_unregister(&dev->v4l2_dev);
>> + return ret;
>> +}
>> +
>> +static int e5010_remove(struct platform_device *pdev)
>> +{
>> + struct e5010_dev *dev = platform_get_drvdata(pdev);
>> +
>> + pm_runtime_disable(dev->dev);
>> + video_unregister_device(dev->vdev);
>> + v4l2_m2m_release(dev->m2m_dev);
>> + v4l2_device_unregister(&dev->v4l2_dev);
>> +
>> + return 0;
>> +}
>
> ...
>
>> +#define MAX_PLANES 2
>> +#define HEADER_SIZE 0x025E
>> +#define MIN_DIMENSION 64
>> +#define MAX_DIMENSION 8192
>> +#define DEFAULT_WIDTH 640
>> +#define DEFAULT_HEIGHT 480
>> +#define E5010_MODULE_NAME "e5010"
>> +
>> +/* JPEG marker definitions */
>> +#define START_OF_IMAGE 0xFFD8
>> +#define SOF_BASELINE_DCT 0xFFC0
>> +#define END_OF_IMAGE 0xFFD9
>> +#define START_OF_SCAN 0xFFDA
>> +
>> +/* Definitions for the huffman table specification in the Marker segment */
>> +#define DHT_MARKER 0xFFC4
>> +#define LH_DC 0x001F
>> +#define LH_AC 0x00B5
>> +
>> +/* Definitions for the quantization table specification in the Marker segment */
>> +#define DQT_MARKER 0xFFDB
>> +#define ACMAX 0x03FF
>> +#define DCMAX 0x07FF
>> +
>> +/* Length and precision of the quantization table parameters */
>> +#define LQPQ 0x00430
>> +#define QMAX 255
>> +
>> +/* Misc JPEG header definitions */
>> +#define UC_NUM_COMP 3
>> +#define PRECISION 8
>> +#define HORZ_SAMPLING_FACTOR (2 << 4)
>> +#define VERT_SAMPLING_FACTOR_422 1
>> +#define VERT_SAMPLING_FACTOR_420 2
>> +#define COMPONENTS_IN_SCAN 3
>> +#define PELS_IN_BLOCK 64
>> +
>> +/* Used for Qp table generation */
>> +#define LUMINOSITY 10
>> +#define CONTRAST 1
>> +#define INCREASE 2
>> +#define QP_TABLE_SIZE (8 * 8)
>> +#define QP_TABLE_FIELD_OFFSET 0x04
>> +
>> +/*
>> + * vb2 queue structure
>> + * contains queue data information
>> + *
>> + * @fmt: format info
>> + * @width: frame width
>> + * @height: frame height
>> + * @bytesperline: bytes per line in memory
>> + * @size_image: image size in memory
>> + */
>> +struct e5010_q_data {
>> + struct e5010_fmt *fmt;
>> + u32 width;
>> + u32 height;
>> + u32 width_adjusted;
>> + u32 height_adjusted;
>> + u32 sizeimage[MAX_PLANES];
>> + u32 bytesperline[MAX_PLANES];
>> + bool format_set;
>> + bool streaming;
>> + u32 sequence;
>> + struct v4l2_rect crop;
>
> Unexpected indentation.

Agreed, will correct.

>
>> +};
>> +
>> +/*
>> + * Driver device structure
>> + * Holds all memory handles and global parameters
>> + * Shared by all instances
>> + */
>> +struct e5010_dev {
>> + struct device *dev;
>> + struct v4l2_device v4l2_dev;
>> + struct v4l2_m2m_dev *m2m_dev;
>> + struct video_device *vdev;
>> + void __iomem *jasper_base;
>> + void __iomem *mmu_base;
>> + struct clk *clk;
>
> Please keep style consistent.

Agreed, will correct.

Best Regards
Devarsh
>
>> + struct e5010_context *last_context_run;
>> + /* Protect access to device data */
>> + struct mutex mutex;
>> + /* Protect access to hardware*/
>> + spinlock_t hw_lock;
>> +};
>> +
>
>
> Best regards,
> Krzysztof
>