Re: [PATCH 11/65] media: Replace file->private_data access with custom functions

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Hans,

On Wed, Aug 06, 2025 at 10:16:37AM +0200, Hans Verkuil wrote:
> On 02/08/2025 11:22, Jacopo Mondi wrote:
> > From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> > 
> > Accessing file->private_data manually to retrieve the v4l2_fh pointer is
> > error-prone, as the field is a void * and will happily cast implicitly
> > to any pointer type.
> > 
> > Replace all remaining locations that read the v4l2_fh pointer directly
> > from file->private_data and cast it to driver-specific file handle
> > structures with driver-specific functions that use file_to_v4l2_fh() and
> > perform the same cast.
> > 
> > No functional change is intended, this only paves the way to remove
> > direct accesses to file->private_data and make V4L2 drivers safer.
> > Other accesses to the field will be addressed separately.
> > 
> > Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> > Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> > ---
> >  drivers/media/pci/ivtv/ivtv-driver.h               |  5 ++++
> >  drivers/media/pci/ivtv/ivtv-fileops.c              | 10 +++----
> >  drivers/media/pci/ivtv/ivtv-ioctl.c                |  8 +++---
> >  drivers/media/platform/allegro-dvt/allegro-core.c  |  7 ++++-
> >  drivers/media/platform/amlogic/meson-ge2d/ge2d.c   |  8 ++++--
> >  .../media/platform/chips-media/coda/coda-common.c  |  7 ++++-
> >  .../platform/chips-media/wave5/wave5-helper.c      |  2 +-
> >  .../media/platform/chips-media/wave5/wave5-vpu.h   |  5 ++++
> >  drivers/media/platform/m2m-deinterlace.c           |  7 ++++-
> >  .../media/platform/mediatek/jpeg/mtk_jpeg_core.c   |  7 ++++-
> >  drivers/media/platform/mediatek/mdp/mtk_mdp_m2m.c  |  7 ++++-
> >  .../media/platform/mediatek/mdp3/mtk-mdp3-m2m.c    |  7 ++++-
> >  .../mediatek/vcodec/decoder/mtk_vcodec_dec_drv.c   |  2 +-
> >  .../mediatek/vcodec/decoder/mtk_vcodec_dec_drv.h   |  5 ++++
> >  .../mediatek/vcodec/encoder/mtk_vcodec_enc_drv.c   |  2 +-
> >  .../mediatek/vcodec/encoder/mtk_vcodec_enc_drv.h   |  5 ++++
> >  drivers/media/platform/nxp/imx-jpeg/mxc-jpeg.c     |  7 ++++-
> >  drivers/media/platform/nxp/imx8-isi/imx8-isi-m2m.c |  7 ++++-
> >  drivers/media/platform/nxp/mx2_emmaprp.c           |  7 ++++-
> >  drivers/media/platform/renesas/rcar_fdp1.c         |  7 ++++-
> >  drivers/media/platform/renesas/rcar_jpu.c          |  7 ++++-
> >  drivers/media/platform/rockchip/rga/rga.c          |  3 +--
> >  drivers/media/platform/rockchip/rga/rga.h          |  5 ++++
> >  drivers/media/platform/rockchip/rkvdec/rkvdec.c    |  2 +-
> >  drivers/media/platform/rockchip/rkvdec/rkvdec.h    |  5 ++++
> >  .../media/platform/samsung/exynos-gsc/gsc-core.h   |  6 +++++
> >  .../media/platform/samsung/exynos-gsc/gsc-m2m.c    |  6 ++---
> >  .../media/platform/samsung/exynos4-is/fimc-core.h  |  5 ++++
> >  .../media/platform/samsung/exynos4-is/fimc-m2m.c   |  2 +-
> >  drivers/media/platform/samsung/s5p-g2d/g2d.c       |  7 +++--
> >  .../media/platform/samsung/s5p-jpeg/jpeg-core.c    |  9 +++++--
> >  drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c   |  6 ++---
> >  .../platform/samsung/s5p-mfc/s5p_mfc_common.h      |  6 +++++
> >  drivers/media/platform/st/sti/bdisp/bdisp-v4l2.c   |  7 ++++-
> >  drivers/media/platform/st/sti/delta/delta-v4l2.c   | 26 +++++++++++-------
> >  drivers/media/platform/st/sti/hva/hva-v4l2.c       | 31 ++++++++++++----------
> >  drivers/media/platform/st/sti/hva/hva.h            |  2 --
> >  drivers/media/platform/st/stm32/dma2d/dma2d.c      |  7 +++--
> >  drivers/media/platform/sunxi/sun8i-di/sun8i-di.c   |  3 +--
> >  .../platform/sunxi/sun8i-rotate/sun8i_rotate.c     |  3 +--
> >  drivers/media/platform/ti/omap3isp/ispvideo.c      |  4 +--
> >  drivers/media/platform/ti/omap3isp/ispvideo.h      |  6 +++++
> >  drivers/media/platform/verisilicon/hantro.h        |  5 ++++
> >  drivers/media/platform/verisilicon/hantro_drv.c    |  3 +--
> >  drivers/staging/media/imx/imx-media-csc-scaler.c   |  7 ++++-
> >  drivers/staging/media/meson/vdec/vdec.c            | 24 ++++++-----------
> >  drivers/staging/media/meson/vdec/vdec.h            |  5 ++++
> >  drivers/staging/media/sunxi/cedrus/cedrus.c        |  3 +--
> >  drivers/staging/media/sunxi/cedrus/cedrus.h        |  5 ++++
> >  drivers/staging/media/sunxi/cedrus/cedrus_video.c  |  5 ----
> >  50 files changed, 237 insertions(+), 100 deletions(-)
> > 
> > diff --git a/drivers/media/pci/ivtv/ivtv-driver.h b/drivers/media/pci/ivtv/ivtv-driver.h
> > index a6ffa99e16bc64a5b7d3e48c1ab32b49a7989242..cad548b28e360ecfe2bcb9fcb5d12cd8823c3727 100644
> > --- a/drivers/media/pci/ivtv/ivtv-driver.h
> > +++ b/drivers/media/pci/ivtv/ivtv-driver.h
> > @@ -388,6 +388,11 @@ static inline struct ivtv_open_id *fh2id(struct v4l2_fh *fh)
> >  	return container_of(fh, struct ivtv_open_id, fh);
> >  }
> >  
> > +static inline struct ivtv_open_id *file2id(struct file *filp)
> > +{
> > +	return container_of(file_to_v4l2_fh(filp), struct ivtv_open_id, fh);
> 
> Why not write:
> 
> 	return fh2id(file_to_v4l2_fh(filp));
> 
> Same for all other drivers that do this. I prefer to have the contained_of()
> in just one place.

Because fh2id gets removed in "[PATCH 57/65] media: ivtv: Access v4l2_fh
from file". I can use it in this patch and drop it later, would you
prefer that ?

> > +}
> > +
> >  struct yuv_frame_info
> >  {
> >  	u32 update;
> 
> <snip>
> 
> > diff --git a/drivers/media/platform/allegro-dvt/allegro-core.c b/drivers/media/platform/allegro-dvt/allegro-core.c
> > index 1f134e08923a528cc676f825da68951c97ac2f25..74977f3ae4844022c04de877f31b4fc6aaac0749 100644
> > --- a/drivers/media/platform/allegro-dvt/allegro-core.c
> > +++ b/drivers/media/platform/allegro-dvt/allegro-core.c
> > @@ -302,6 +302,11 @@ struct allegro_channel {
> >  	unsigned int error;
> >  };
> >  
> > +static inline struct allegro_channel *file_to_channel(struct file *filp)
> > +{
> > +	return container_of(file_to_v4l2_fh(filp), struct allegro_channel, fh);
> > +}
> > +
> >  static inline int
> >  allegro_channel_get_i_frame_qp(struct allegro_channel *channel)
> >  {
> > @@ -3229,7 +3234,7 @@ static int allegro_open(struct file *file)
> >  
> >  static int allegro_release(struct file *file)
> >  {
> > -	struct allegro_channel *channel = fh_to_channel(file->private_data);
> > +	struct allegro_channel *channel = file_to_channel(file);
> 
> So a file_to_channel inline function was added, but it is used in just one
> place.
> 
> I would prefer to just drop the inline function and instead write:
> 
> 	struct allegro_channel *channel = fh_to_channel(file_to_v4l2_fh(file));
> 
> If this is needed in two or more places, then the extra inline makes sense,
> but it is a fairly common pattern that it is only needed in the release function.
> 
> Adding a new inline just for that seems overkill to me.

file_to_channel() gets used in more places in "[PATCH 29/65] media:
allegro: Access v4l2_fh from file", where fh_to_channel() is dropped.
I'd rather keep it in this patch instead of having to modify the
allegro_release() function in patch 29/65.

> >  
> >  	v4l2_m2m_ctx_release(channel->fh.m2m_ctx);
> >  

-- 
Regards,

Laurent Pinchart