Linux-Rockchip Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Tomasz Figa <tfiga@chromium.org>
To: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Cc: mchehab@kernel.org, m.szyprowski@samsung.com, ming.qian@nxp.com,
	ezequiel@vanguardiasur.com.ar, p.zabel@pengutronix.de,
	gregkh@linuxfoundation.org, hverkuil-cisco@xs4all.nl,
	nicolas.dufresne@collabora.com, linux-media@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-mediatek@lists.infradead.org,
	linux-arm-msm@vger.kernel.org,
	linux-rockchip@lists.infradead.org,
	linux-staging@lists.linux.dev, kernel@collabora.com
Subject: Re: [PATCH v14 02/56] media: videobuf2: Rework offset 'cookie' encoding pattern
Date: Wed, 8 Nov 2023 06:23:22 +0000	[thread overview]
Message-ID: <20231108062322.l2k2n2jlpoaa4tws@chromium.org> (raw)
In-Reply-To: <20231031163104.112469-3-benjamin.gaignard@collabora.com>

Hi Benjamin,

On Tue, Oct 31, 2023 at 05:30:10PM +0100, Benjamin Gaignard wrote:
> Change how offset 'cookie' field value is computed to make possible
> to use more buffers.
> The maximum number of buffers depends of PAGE_SHIFT value and can
> go up to 0x7fff when PAGE_SHIFT = 12.
> With this encoding pattern we know the maximum number that a queue
> could store so we can check it at  queue init time.
> It also make easier and faster to find buffer and plane from using
> the offset field.
> Change __find_plane_by_offset() prototype to return the video buffer
> itself rather than it index.
> 
> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
> ---
>  .../media/common/videobuf2/videobuf2-core.c   | 73 +++++++++----------
>  1 file changed, 35 insertions(+), 38 deletions(-)
> 

Thanks a lot for working on this series. Just one comment inline.

> diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
> index a5e57affeb30..09be8e026044 100644
> --- a/drivers/media/common/videobuf2/videobuf2-core.c
> +++ b/drivers/media/common/videobuf2/videobuf2-core.c
> @@ -31,6 +31,11 @@
>  
>  #include <trace/events/vb2.h>
>  
> +#define PLANE_INDEX_SHIFT	(PAGE_SHIFT + 3)
> +#define PLANE_INDEX_MASK	0x7

Could we somehow ensure that it indeed matches VIDEO_MAX_PLANES?

One thing that comes to my mind would be

#define PLANE_INDEX_BITS	3
#define PLANE_INDEX_SHIFT	(PAGE_SHIFT + PLANE_INDEX_BITS)
#define PLANE_INDEX_MASK	BIT_MASK(PLANE_INDEX_BITS)

BUILD_BUG_ON(BIT(PLANE_INDEX_BITS) != VIDEO_MAX_PLANES);

Alternatively we could change the videodev2.h, so that we have

#define VIDEO_MAX_PLANES_BITS	3
#define VIDEO_MAX_PLANES	(1U < VIDEO_MAX_PLANES_BITS)

and then we can just VIDEO_MAX_PLANES_BITS in our definitions in this file
and no checks would be necessary.

Best regards,
Tomasz

> +#define MAX_BUFFER_INDEX	BIT_MASK(30 - PLANE_INDEX_SHIFT)
> +#define BUFFER_INDEX_MASK	(MAX_BUFFER_INDEX - 1)
> +
>  static int debug;
>  module_param(debug, int, 0644);
>  
> @@ -358,21 +363,27 @@ static void __setup_offsets(struct vb2_buffer *vb)
>  	unsigned int plane;
>  	unsigned long offset = 0;
>  
> -	if (vb->index) {
> -		struct vb2_buffer *prev = q->bufs[vb->index - 1];
> -		struct vb2_plane *p = &prev->planes[prev->num_planes - 1];
> -
> -		offset = PAGE_ALIGN(p->m.offset + p->length);
> -	}
> +	/*
> +	 * The offset "cookie" value has the following constraints:
> +	 * - a buffer can have up to 8 planes.
> +	 * - v4l2 mem2mem uses bit 30 to distinguish between
> +	 *   OUTPUT (aka "source", bit 30 is 0) and
> +	 *   CAPTURE (aka "destination", bit 30 is 1) buffers.
> +	 * - must be page aligned
> +	 * That led to this bit mapping when PAGE_SHIFT = 12:
> +	 * |30                |29        15|14       12|11 0|
> +	 * |DST_QUEUE_OFF_BASE|buffer index|plane index| 0  |
> +	 * where there are 15 bits to store the buffer index.
> +	 * Depending on PAGE_SHIFT value we can have fewer bits
> +	 * to store the buffer index.
> +	 */
> +	offset = vb->index << PLANE_INDEX_SHIFT;
>  
>  	for (plane = 0; plane < vb->num_planes; ++plane) {
> -		vb->planes[plane].m.offset = offset;
> +		vb->planes[plane].m.offset = offset + (plane << PAGE_SHIFT);
>  
>  		dprintk(q, 3, "buffer %d, plane %d offset 0x%08lx\n",
>  				vb->index, plane, offset);
> -
> -		offset += vb->planes[plane].length;
> -		offset = PAGE_ALIGN(offset);
>  	}
>  }
>  
> @@ -2188,10 +2199,9 @@ EXPORT_SYMBOL_GPL(vb2_core_streamoff);
>   * __find_plane_by_offset() - find plane associated with the given offset
>   */
>  static int __find_plane_by_offset(struct vb2_queue *q, unsigned long offset,
> -			unsigned int *_buffer, unsigned int *_plane)
> +			struct vb2_buffer **vb, unsigned int *plane)
>  {
> -	struct vb2_buffer *vb;
> -	unsigned int buffer, plane;
> +	unsigned int buffer;
>  
>  	/*
>  	 * Sanity checks to ensure the lock is held, MEMORY_MMAP is
> @@ -2209,24 +2219,15 @@ static int __find_plane_by_offset(struct vb2_queue *q, unsigned long offset,
>  		return -EBUSY;
>  	}
>  
> -	/*
> -	 * Go over all buffers and their planes, comparing the given offset
> -	 * with an offset assigned to each plane. If a match is found,
> -	 * return its buffer and plane numbers.
> -	 */
> -	for (buffer = 0; buffer < q->num_buffers; ++buffer) {
> -		vb = q->bufs[buffer];
> +	/* Get buffer and plane from the offset */
> +	buffer = (offset >> PLANE_INDEX_SHIFT) & BUFFER_INDEX_MASK;
> +	*plane = (offset >> PAGE_SHIFT) & PLANE_INDEX_MASK;
>  
> -		for (plane = 0; plane < vb->num_planes; ++plane) {
> -			if (vb->planes[plane].m.offset == offset) {
> -				*_buffer = buffer;
> -				*_plane = plane;
> -				return 0;
> -			}
> -		}
> -	}
> +	if (buffer >= q->num_buffers || *plane >= q->bufs[buffer]->num_planes)
> +		return -EINVAL;
>  
> -	return -EINVAL;
> +	*vb = q->bufs[buffer];
> +	return 0;
>  }
>  
>  int vb2_core_expbuf(struct vb2_queue *q, int *fd, unsigned int type,
> @@ -2306,7 +2307,7 @@ int vb2_mmap(struct vb2_queue *q, struct vm_area_struct *vma)
>  {
>  	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
>  	struct vb2_buffer *vb;
> -	unsigned int buffer = 0, plane = 0;
> +	unsigned int plane = 0;
>  	int ret;
>  	unsigned long length;
>  
> @@ -2335,12 +2336,10 @@ int vb2_mmap(struct vb2_queue *q, struct vm_area_struct *vma)
>  	 * Find the plane corresponding to the offset passed by userspace. This
>  	 * will return an error if not MEMORY_MMAP or file I/O is in progress.
>  	 */
> -	ret = __find_plane_by_offset(q, offset, &buffer, &plane);
> +	ret = __find_plane_by_offset(q, offset, &vb, &plane);
>  	if (ret)
>  		goto unlock;
>  
> -	vb = q->bufs[buffer];
> -
>  	/*
>  	 * MMAP requires page_aligned buffers.
>  	 * The buffer length was page_aligned at __vb2_buf_mem_alloc(),
> @@ -2368,7 +2367,7 @@ int vb2_mmap(struct vb2_queue *q, struct vm_area_struct *vma)
>  	if (ret)
>  		return ret;
>  
> -	dprintk(q, 3, "buffer %d, plane %d successfully mapped\n", buffer, plane);
> +	dprintk(q, 3, "buffer %u, plane %d successfully mapped\n", vb->index, plane);
>  	return 0;
>  }
>  EXPORT_SYMBOL_GPL(vb2_mmap);
> @@ -2382,7 +2381,7 @@ unsigned long vb2_get_unmapped_area(struct vb2_queue *q,
>  {
>  	unsigned long offset = pgoff << PAGE_SHIFT;
>  	struct vb2_buffer *vb;
> -	unsigned int buffer, plane;
> +	unsigned int plane;
>  	void *vaddr;
>  	int ret;
>  
> @@ -2392,12 +2391,10 @@ unsigned long vb2_get_unmapped_area(struct vb2_queue *q,
>  	 * Find the plane corresponding to the offset passed by userspace. This
>  	 * will return an error if not MEMORY_MMAP or file I/O is in progress.
>  	 */
> -	ret = __find_plane_by_offset(q, offset, &buffer, &plane);
> +	ret = __find_plane_by_offset(q, offset, &vb, &plane);
>  	if (ret)
>  		goto unlock;
>  
> -	vb = q->bufs[buffer];
> -
>  	vaddr = vb2_plane_vaddr(vb, plane);
>  	mutex_unlock(&q->mmap_lock);
>  	return vaddr ? (unsigned long)vaddr : -EINVAL;
> -- 
> 2.39.2
> 

_______________________________________________
Linux-rockchip mailing list
Linux-rockchip@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-rockchip

  reply	other threads:[~2023-11-08  6:23 UTC|newest]

Thread overview: 133+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-31 16:30 [PATCH v14 00/56] Add DELETE_BUF ioctl Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 01/56] media: videobuf2: Rename offset parameter Benjamin Gaignard
2023-11-08 15:11   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 02/56] media: videobuf2: Rework offset 'cookie' encoding pattern Benjamin Gaignard
2023-11-08  6:23   ` Tomasz Figa [this message]
2023-11-08 10:14     ` Benjamin Gaignard
2023-11-08 10:31       ` Hans Verkuil
2023-10-31 16:30 ` [PATCH v14 03/56] media: videobuf2: Stop spamming kernel log with all queue counter Benjamin Gaignard
2023-11-08 15:26   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 04/56] media: videobuf2: Use vb2_buffer instead of index Benjamin Gaignard
2023-11-08  8:08   ` Tomasz Figa
2023-11-08 15:52   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 05/56] media: videobuf2: Access vb2_queue bufs array through helper functions Benjamin Gaignard
2023-11-08  8:50   ` Tomasz Figa
2023-11-08 10:24     ` Benjamin Gaignard
2023-11-09  4:27       ` Tomasz Figa
2023-10-31 16:30 ` [PATCH v14 06/56] media: videobuf2: Remove duplicated index vs q->num_buffers check Benjamin Gaignard
2023-11-08 16:46   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 07/56] media: videobuf2: Add helper to get queue number of buffers Benjamin Gaignard
2023-11-08 16:50   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 08/56] media: videobuf2: Use vb2_get_num_buffers() helper Benjamin Gaignard
2023-11-08  9:42   ` Tomasz Figa
2023-11-08 13:22     ` Benjamin Gaignard
2023-11-09  4:36       ` Tomasz Figa
2023-10-31 16:30 ` [PATCH v14 09/56] media: amphion: Use vb2_get_buffer() instead of directly access to buffers array Benjamin Gaignard
2023-11-08 16:55   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 10/56] media: amphion: Stop direct calls to queue num_buffers field Benjamin Gaignard
2023-11-08 16:56   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 11/56] media: mediatek: jpeg: Use vb2_get_buffer() instead of directly access to buffers array Benjamin Gaignard
2023-11-08 17:00   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 12/56] media: mediatek: vdec: Remove useless loop Benjamin Gaignard
2023-11-09  9:15   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 13/56] media: mediatek: vcodec: Stop direct calls to queue num_buffers field Benjamin Gaignard
2023-11-09  9:21   ` Andrzej Pietrasiewicz
2023-11-09  9:23   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 14/56] media: sti: hva: Use vb2_get_buffer() instead of directly access to buffers array Benjamin Gaignard
2023-11-09  9:29   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 15/56] media: visl: " Benjamin Gaignard
2023-11-09  9:38   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 16/56] media: atomisp: " Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 17/56] media: atomisp: Stop direct calls to queue num_buffers field Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 18/56] media: dvb-core: Use vb2_get_buffer() instead of directly access to buffers array Benjamin Gaignard
2023-11-09  9:46   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 19/56] media: dvb-core: Do not initialize twice queue num_buffer field Benjamin Gaignard
2023-11-09  9:50   ` Andrzej Pietrasiewicz
2023-11-09 10:14     ` Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 20/56] media: dvb-frontends: rtl2832: Stop direct calls to queue num_buffers field Benjamin Gaignard
2023-11-09  9:56   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 21/56] media: pci: dt3155: Remove useless check Benjamin Gaignard
2023-11-09 10:03   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 22/56] media: pci: tw686x: Stop direct calls to queue num_buffers field Benjamin Gaignard
2023-11-09 10:05   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 23/56] media: pci: cx18: " Benjamin Gaignard
2023-11-09 10:07   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 24/56] media: pci: netup_unidvb: " Benjamin Gaignard
2023-11-09 10:09   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 25/56] media: pci: tw68: " Benjamin Gaignard
2023-11-09 10:13   ` Andrzej Pietrasiewicz
2023-11-09 10:23     ` Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 26/56] media: i2c: video-i2c: " Benjamin Gaignard
2023-11-09 10:18   ` Andrzej Pietrasiewicz
2023-11-09 10:26     ` Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 27/56] media: coda: " Benjamin Gaignard
2023-11-09 10:22   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 28/56] media: nxp: " Benjamin Gaignard
2023-11-09 10:26   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 29/56] media: verisilicon: " Benjamin Gaignard
2023-11-09 10:28   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 30/56] media: test-drivers: " Benjamin Gaignard
2023-11-09 11:10   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 31/56] media: imx: " Benjamin Gaignard
2023-11-09 11:12   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 32/56] media: meson: vdec: " Benjamin Gaignard
2023-11-09 11:17   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 33/56] touchscreen: sur40: " Benjamin Gaignard
2023-11-09 11:19   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 34/56] sample: v4l: " Benjamin Gaignard
2023-11-09 11:19   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 35/56] media: cedrus: " Benjamin Gaignard
2023-11-09 11:27   ` Andrzej Pietrasiewicz
2023-11-09 14:14     ` Paul Kocialkowski
2023-11-09 15:48       ` Andrzej Pietrasiewicz
2023-11-09 15:54         ` Benjamin Gaignard
2023-11-09 16:29           ` Paul Kocialkowski
2023-10-31 16:30 ` [PATCH v14 36/56] media: nuvoton: " Benjamin Gaignard
2023-11-09 11:29   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 37/56] media: renesas: " Benjamin Gaignard
2023-11-09 11:30   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 38/56] media: ti: " Benjamin Gaignard
2023-11-09 11:32   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 39/56] media: usb: airspy: " Benjamin Gaignard
2023-11-09 11:33   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 40/56] media: usb: cx231xx: " Benjamin Gaignard
2023-11-09 11:35   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 41/56] media: usb: hackrf: " Benjamin Gaignard
2023-11-09 11:36   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 42/56] media: usb: usbtv: " Benjamin Gaignard
2023-11-09 11:37   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 43/56] media: videobuf2: Be more flexible on the number of queue stored buffers Benjamin Gaignard
2023-11-02  8:17   ` Hans Verkuil
2023-11-06 14:29     ` Benjamin Gaignard
2023-10-31 16:30 ` [PATCH v14 44/56] media: core: Report the maximum possible number of buffers for the queue Benjamin Gaignard
2023-11-09 12:28   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 45/56] media: test-drivers: vivid: Increase max supported buffers for capture queues Benjamin Gaignard
2023-11-09 12:31   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 46/56] media: test-drivers: vicodec: Increase max supported capture queue buffers Benjamin Gaignard
2023-11-09 12:32   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 47/56] media: verisilicon: Refactor postprocessor to store more buffers Benjamin Gaignard
2023-11-09 13:58   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 48/56] media: verisilicon: Store chroma and motion vectors offset Benjamin Gaignard
2023-11-09 14:12   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 49/56] media: verisilicon: g2: Use common helpers to compute chroma and mv offsets Benjamin Gaignard
2023-11-09 14:25   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 50/56] media: verisilicon: vp9: Allow to change resolution while streaming Benjamin Gaignard
2023-11-09 14:59   ` Andrzej Pietrasiewicz
2023-10-31 16:30 ` [PATCH v14 51/56] media: core: Rework how create_buf index returned value is computed Benjamin Gaignard
2023-10-31 16:31 ` [PATCH v14 52/56] media: core: Add bitmap manage bufs array entries Benjamin Gaignard
2023-11-08 10:44   ` Tomasz Figa
2023-11-08 15:30     ` Benjamin Gaignard
2023-11-09  7:28       ` Tomasz Figa
2023-10-31 16:31 ` [PATCH v14 53/56] media: core: Free range of buffers Benjamin Gaignard
2023-11-09  9:10   ` Tomasz Figa
2023-10-31 16:31 ` [PATCH v14 54/56] media: v4l2: Add DELETE_BUFS ioctl Benjamin Gaignard
2023-11-09  9:30   ` Tomasz Figa
2023-11-09 10:10     ` Benjamin Gaignard
2023-11-10  8:58       ` Tomasz Figa
2023-11-10  9:14         ` Hans Verkuil
2023-11-10  9:19           ` Tomasz Figa
2023-10-31 16:31 ` [PATCH v14 55/56] media: v4l2: Add mem2mem helpers for " Benjamin Gaignard
2023-10-31 16:31 ` [PATCH v14 56/56] media: test-drivers: Use helper " Benjamin Gaignard
2023-11-09  9:43   ` Tomasz Figa
2023-11-09  9:46     ` Benjamin Gaignard
2023-11-09  9:48       ` Tomasz Figa

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20231108062322.l2k2n2jlpoaa4tws@chromium.org \
    --to=tfiga@chromium.org \
    --cc=benjamin.gaignard@collabora.com \
    --cc=ezequiel@vanguardiasur.com.ar \
    --cc=gregkh@linuxfoundation.org \
    --cc=hverkuil-cisco@xs4all.nl \
    --cc=kernel@collabora.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-media@vger.kernel.org \
    --cc=linux-mediatek@lists.infradead.org \
    --cc=linux-rockchip@lists.infradead.org \
    --cc=linux-staging@lists.linux.dev \
    --cc=m.szyprowski@samsung.com \
    --cc=mchehab@kernel.org \
    --cc=ming.qian@nxp.com \
    --cc=nicolas.dufresne@collabora.com \
    --cc=p.zabel@pengutronix.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox