[PATCH] media: mtk-vpu: Ensure alignment of 8 for DTCM buffer

[PATCH] media: mtk-vpu: Ensure alignment of 8 for DTCM buffer

[Dafna Hirschfeld]

From: Enric Balletbo i Serra <enric.balletbo@collabora.com>

When running memcpy_toio:
memcpy_toio(send_obj->share_buf, buf, len);
it was found that errors appear if len is not a multiple of 8:

[58.350841] mtk-mdp 14001000.rdma: processing failed: -22

This patch ensure copy of a multile of 8 size by calling
round_up(len, 8) when copying

Fixes: e6599adfad30 ("media: mtk-vpu: avoid unaligned access to DTCM buffer.")
Reported-by: Alexandre Courbot <acourbot@chromium.org>
Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
---
 drivers/media/platform/mtk-vpu/mtk_vpu.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/media/platform/mtk-vpu/mtk_vpu.c b/drivers/media/platform/mtk-vpu/mtk_vpu.c
index ec290dde59cf..b464d8192119 100644
--- a/drivers/media/platform/mtk-vpu/mtk_vpu.c
+++ b/drivers/media/platform/mtk-vpu/mtk_vpu.c
@@ -316,6 +316,7 @@ int vpu_ipi_send(struct platform_device *pdev,
 {
 	struct mtk_vpu *vpu = platform_get_drvdata(pdev);
 	struct share_obj __iomem *send_obj = vpu->send_buf;
+	unsigned char data[SHARE_BUF_SIZE];
 	unsigned long timeout;
 	int ret = 0;
 
@@ -349,7 +350,9 @@ int vpu_ipi_send(struct platform_device *pdev,
 		}
 	} while (vpu_cfg_readl(vpu, HOST_TO_VPU));
 
-	memcpy_toio(send_obj->share_buf, buf, len);
+	memset(data, 0, sizeof(data));
+	memcpy(data, buf, len);
+	memcpy_toio(send_obj->share_buf, data, round_up(len, 8));
 	writel(len, &send_obj->len);
 	writel(id, &send_obj->id);
 
-- 
2.17.1

Re: [PATCH] media: mtk-vpu: Ensure alignment of 8 for DTCM buffer

[Enric Balletbo Serra]

Hi Dafna,

Thank you for working on this and sending the patch to upstream for
further discussion.

Missatge de Dafna Hirschfeld <dafna.hirschfeld@collabora.com> del dia
dc., 14 de jul. 2021 a les 15:49:
>
> From: Enric Balletbo i Serra <enric.balletbo@collabora.com>
>

Sorry to confuse you, but I'm not the author of this patch, my from is
only here because I shared with you that patch that I really picked
from a discussion that took place here:

  https://chromium-review.googlesource.com/c/chromiumos/third_party/kernel/+/2251840

It'd be really nice to continue the discussion here and set a proper
fix and understand better the root cause and the fix. The patch really
fixes an issue in mainline for mtk-vcodec driver on the mt8173
devices, I can confirm that without it doesn't work properly.

Waiting for your thoughts,
  Enric

> When running memcpy_toio:
> memcpy_toio(send_obj->share_buf, buf, len);
> it was found that errors appear if len is not a multiple of 8:
>
> [58.350841] mtk-mdp 14001000.rdma: processing failed: -22
>
> This patch ensure copy of a multile of 8 size by calling
> round_up(len, 8) when copying
>
> Fixes: e6599adfad30 ("media: mtk-vpu: avoid unaligned access to DTCM buffer.")
> Reported-by: Alexandre Courbot <acourbot@chromium.org>
> Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
> Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
> ---
>  drivers/media/platform/mtk-vpu/mtk_vpu.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/media/platform/mtk-vpu/mtk_vpu.c b/drivers/media/platform/mtk-vpu/mtk_vpu.c
> index ec290dde59cf..b464d8192119 100644
> --- a/drivers/media/platform/mtk-vpu/mtk_vpu.c
> +++ b/drivers/media/platform/mtk-vpu/mtk_vpu.c
> @@ -316,6 +316,7 @@ int vpu_ipi_send(struct platform_device *pdev,
>  {
>         struct mtk_vpu *vpu = platform_get_drvdata(pdev);
>         struct share_obj __iomem *send_obj = vpu->send_buf;
> +       unsigned char data[SHARE_BUF_SIZE];
>         unsigned long timeout;
>         int ret = 0;
>
> @@ -349,7 +350,9 @@ int vpu_ipi_send(struct platform_device *pdev,
>                 }
>         } while (vpu_cfg_readl(vpu, HOST_TO_VPU));
>
> -       memcpy_toio(send_obj->share_buf, buf, len);
> +       memset(data, 0, sizeof(data));
> +       memcpy(data, buf, len);
> +       memcpy_toio(send_obj->share_buf, data, round_up(len, 8));
>         writel(len, &send_obj->len);
>         writel(id, &send_obj->id);
>
> --
> 2.17.1
>
>
> _______________________________________________
> Linux-mediatek mailing list
> Linux-mediatek@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-mediatek

Re: [PATCH] media: mtk-vpu: Ensure alignment of 8 for DTCM buffer

[Hans Verkuil]

On 14/07/2021 15:48, Dafna Hirschfeld wrote:
> From: Enric Balletbo i Serra <enric.balletbo@collabora.com>
> 
> When running memcpy_toio:
> memcpy_toio(send_obj->share_buf, buf, len);
> it was found that errors appear if len is not a multiple of 8:
> 
> [58.350841] mtk-mdp 14001000.rdma: processing failed: -22
> 
> This patch ensure copy of a multile of 8 size by calling
> round_up(len, 8) when copying
> 
> Fixes: e6599adfad30 ("media: mtk-vpu: avoid unaligned access to DTCM buffer.")
> Reported-by: Alexandre Courbot <acourbot@chromium.org>
> Signed-off-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
> Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
> ---
>  drivers/media/platform/mtk-vpu/mtk_vpu.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/media/platform/mtk-vpu/mtk_vpu.c b/drivers/media/platform/mtk-vpu/mtk_vpu.c
> index ec290dde59cf..b464d8192119 100644
> --- a/drivers/media/platform/mtk-vpu/mtk_vpu.c
> +++ b/drivers/media/platform/mtk-vpu/mtk_vpu.c
> @@ -316,6 +316,7 @@ int vpu_ipi_send(struct platform_device *pdev,
>  {
>  	struct mtk_vpu *vpu = platform_get_drvdata(pdev);
>  	struct share_obj __iomem *send_obj = vpu->send_buf;
> +	unsigned char data[SHARE_BUF_SIZE];
>  	unsigned long timeout;
>  	int ret = 0;
>  
> @@ -349,7 +350,9 @@ int vpu_ipi_send(struct platform_device *pdev,
>  		}
>  	} while (vpu_cfg_readl(vpu, HOST_TO_VPU));
>  
> -	memcpy_toio(send_obj->share_buf, buf, len);
> +	memset(data, 0, sizeof(data));

Hmm, how about:

	memset(data + len, 0, sizeof(data) - len);

> +	memcpy(data, buf, len);
> +	memcpy_toio(send_obj->share_buf, data, round_up(len, 8));
>  	writel(len, &send_obj->len);
>  	writel(id, &send_obj->id);
>  
> 

Also note Enric's reply.

Regards,

	Hans