Re: [PATCH] media: mtk-vcodec: set dma max segment size

[Hsin-Yi Wang]

On Wed, Sep 2, 2020 at 10:25 PM Robin Murphy <robin.murphy@xxxxxxx> wrote:
>
> On 2020-08-21 05:14, Hsin-Yi Wang wrote:
> > Set dma max segment size for encoder and decoder driver.
> >
> > Fix following warning with CONFIG_DMA_API_DEBUG_SG=y
> >
> > [   75.147825] ------------[ cut here ]------------
> > [   75.147844] mtk-vcodec-enc 19002000.vcodec: DMA-API: mapping sg segment longer than device claims to support [len=983040] [max=65536]
> > [   75.147876] WARNING: CPU: 2 PID: 4069 at kernel/dma/debug.c:1302 debug_dma_map_sg+0x1a8/0x2c4
> > ...
> > [   75.148139] Call trace:
> > [   75.148149]  debug_dma_map_sg+0x1a8/0x2c4
> > [   75.148165]  vb2_dc_get_userptr+0x228/0x364 [videobuf2_dma_contig]
> > [   75.148178]  __buf_prepare+0x3ac/0x8c0 [videobuf2_common]
> > [   75.148188]  vb2_core_qbuf+0xa4/0x58c [videobuf2_common]
> > [   75.148199]  vb2_qbuf+0x88/0xe4 [videobuf2_v4l2]
> > [   75.148211]  v4l2_m2m_qbuf+0x80/0xf8 [v4l2_mem2mem]
> > [   75.148221]  vidioc_venc_qbuf+0x3c/0x70 [mtk_vcodec_enc]
> > [   75.148234]  v4l_qbuf+0x48/0x58
> > [   75.148242]  __video_do_ioctl+0x200/0x37c
> > [   75.148250]  video_usercopy+0x360/0x834
> > [   75.148259]  video_ioctl2+0x38/0x48
> > [   75.148267]  v4l2_ioctl+0x6c/0x80
> > [   75.148276]  do_video_ioctl+0xefc/0x4b70
> > [   75.148284]  v4l2_compat_ioctl32+0x5c/0xcc
> > [   75.148294]  __arm64_compat_sys_ioctl+0xf4/0x240
> > [   75.148304]  el0_svc_common+0xac/0x198
> > [   75.148312]  el0_svc_compat_handler+0x2c/0x40
> > [   75.148321]  el0_svc_compat+0x8/0x18
> > [   75.148328] irq event stamp: 0
> > [   75.148337] hardirqs last  enabled at (0): [<0000000000000000>]           (null)
> > [   75.148347] hardirqs last disabled at (0): [<ffffff90080e65c0>] copy_process+0x380/0x115c
> > [   75.148356] softirqs last  enabled at (0): [<ffffff90080e65d8>] copy_process+0x398/0x115c
> > [   75.148364] softirqs last disabled at (0): [<0000000000000000>]           (null)
> > [   75.148372] ---[ end trace 588bf529451e3531 ]---
> >
> > Reported-by: Anand K Mistry <amistry@xxxxxxxxxxxx>
> > Signed-off-by: Hsin-Yi Wang <hsinyi@xxxxxxxxxxxx>
> > ---
> >   drivers/media/platform/mtk-vcodec/mtk_vcodec_dec_drv.c | 8 ++++++++
> >   drivers/media/platform/mtk-vcodec/mtk_vcodec_enc_drv.c | 8 ++++++++
> >   2 files changed, 16 insertions(+)
> >
> > diff --git a/drivers/media/platform/mtk-vcodec/mtk_vcodec_dec_drv.c b/drivers/media/platform/mtk-vcodec/mtk_vcodec_dec_drv.c
> > index 97a1b6664c20..3bbd0bac56d6 100644
> > --- a/drivers/media/platform/mtk-vcodec/mtk_vcodec_dec_drv.c
> > +++ b/drivers/media/platform/mtk-vcodec/mtk_vcodec_dec_drv.c
> > @@ -242,6 +242,14 @@ static int mtk_vcodec_probe(struct platform_device *pdev)
> >               mtk_v4l2_err("[VPU] vpu device in not ready");
> >               return -EPROBE_DEFER;
> >       }
> > +     if (!pdev->dev.dma_parms) {
> > +             pdev->dev.dma_parms = devm_kzalloc(&pdev->dev,
> > +                                             sizeof(*pdev->dev.dma_parms),
> > +                                             GFP_KERNEL);
> > +             if (!pdev->dev.dma_parms)
> > +                     return -ENOMEM;
> > +     }
>
> Since 9495b7e92f71, dma_parms is included in the platform_device itself
> and always set, so you can drop these allocation paths.
>
> > +     dma_set_max_seg_size(&pdev->dev, DMA_BIT_MASK(32));
>
> The segment size is simply a size, not a bitmask, so UINT_MAX would be
> more semantically appropriate.
>
> Robin.
>

Thanks. since this patch is already in media tree:
https://git.linuxtv.org/media_tree.git/commit/?id=13483fc2f20f0e2db7ba9c39b095ac7ea46f8de8,
I'll send a patch to modify this.