| Message ID | 20210806082810.9378-1-dafna.hirschfeld@collabora.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] media: mtk-vpu: Ensure alignment of 8 for DTCM buffer | expand |
On Fri, 2021-08-06 at 16:28 +0800, Dafna Hirschfeld wrote: > From: Alexandre Courbot <acourbot@chromium.org> > > 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.") > Signed-off-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> > --- > changes since v1: > 1. change sign-off-by tags > 2. change values to memset > > 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..66276c5a1bc3 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 + len, 0, sizeof(data) - len); > + memcpy(data, buf, len); > + memcpy_toio(send_obj->share_buf, data, round_up(len, 8)); Hi Dafna, Thanks for your patch. In this patch, the data is copied twice even if the 'len' is mulitple of 8. Can we judge the alignment and decide to copy once or twice ? Like below, which may reduce the latency of this function a bit when 'len' is multiple of 8. if (unlikely((len % 8) != 0)) { unsigned char data[SHARE_BUF_SIZE]; memset(data + len, 0, sizeof(data) - len); memcpy(data, buf, len); memcpy_toio(send_obj->share_buf, data, round_up(len, 8)); } else { memcpy_toio(send_obj->share_buf, buf, len); } Regards, Houlong > writel(len, &send_obj->len); > writel(id, &send_obj->id); > > -- > 2.17.1
diff --git a/drivers/media/platform/mtk-vpu/mtk_vpu.c b/drivers/media/platform/mtk-vpu/mtk_vpu.c index ec290dde59cf..66276c5a1bc3 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 + 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);