lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <905d78f8-26a2-4229-8892-38d7cf24e39c@xs4all.nl>
Date:   Wed, 11 Oct 2023 10:04:30 +0200
From:   Hans Verkuil <hverkuil-cisco@...all.nl>
To:     Benjamin Gaignard <benjamin.gaignard@...labora.com>,
        mchehab@...nel.org, tfiga@...omium.org, m.szyprowski@...sung.com,
        ming.qian@....com, ezequiel@...guardiasur.com.ar,
        p.zabel@...gutronix.de, gregkh@...uxfoundation.org,
        nicolas.dufresne@...labora.com
Cc:     linux-media@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org,
        linux-mediatek@...ts.infradead.org, linux-arm-msm@...r.kernel.org,
        linux-rockchip@...ts.infradead.org, linux-staging@...ts.linux.dev,
        kernel@...labora.com
Subject: Re: [PATCH v10 01/54] media: videobuf2: Rework offset 'cookie'
 encoding pattern

Hi Benjamin,

On 03/10/2023 10:06, 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@...labora.com>
> ---
> changes in version 10:
> - Make BUFFER_INDEX_MASK definition more readable.
> - Correct typo.
> 
>  .../media/common/videobuf2/videobuf2-core.c   | 72 +++++++++----------
>  1 file changed, 33 insertions(+), 39 deletions(-)
> 
> diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
> index 27aee92f3eea..5591ac830668 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
> +#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 +363,24 @@ static void __setup_offsets(struct vb2_buffer *vb)
>  	unsigned int plane;
>  	unsigned long off = 0;

This is a rather ugly name. I suggest renaming this to "offset" to avoid
confusion with "on"/"off". See additional comments on this below.

>  
> -	if (vb->index) {
> -		struct vb2_buffer *prev = q->bufs[vb->index - 1];
> -		struct vb2_plane *p = &prev->planes[prev->num_planes - 1];
> -
> -		off = PAGE_ALIGN(p->m.offset + p->length);
> -	}
> +	/*
> +	 * Offsets cookies value have the following constraints:

I'd rephrase this a little bit:

	 * The offset "cookie" value has the following constraints:

> +	 * - a buffer can have up to 8 planes.
> +	 * - v4l2 mem2mem uses bit 30 to distinguish between source and destination buffers.

Related to Sakari's comments: the mem2mem framework uses source and destination
a lot, but vb2 is mostly OUTPUT/CAPTURE. So it is best to mention both:

source -> OUTPUT (aka "source", bit 30 is 0)
destination -> CAPTURE (aka "destination", bit 30 is 1)

> +	 * - 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.
> +	 */
> +	off = vb->index << PLANE_INDEX_SHIFT;
>  
>  	for (plane = 0; plane < vb->num_planes; ++plane) {
> -		vb->planes[plane].m.offset = off;
> +		vb->planes[plane].m.offset = off + (plane << PAGE_SHIFT);
>  
>  		dprintk(q, 3, "buffer %d, plane %d offset 0x%08lx\n",
>  				vb->index, plane, off);
> -
> -		off += vb->planes[plane].length;
> -		off = PAGE_ALIGN(off);
>  	}
>  }
>  
> @@ -2185,13 +2193,12 @@ int vb2_core_streamoff(struct vb2_queue *q, unsigned int type)
>  EXPORT_SYMBOL_GPL(vb2_core_streamoff);
>  
>  /*
> - * __find_plane_by_offset() - find plane associated with the given offset off
> + * __find_plane_by_offset() - find video buffer and plane associated with the given offset off
>   */
>  static int __find_plane_by_offset(struct vb2_queue *q, unsigned long off,

Same here: rename "off" to "offset", that also simplifies the comment above:
"with the given offset off" can now just be "with the given offset".

The same change can be done in vb2_mmap and vb2_get_unmapped_area.

I think it would be best if the off -> offset rename is done in a separate initial
patch.

I know that this isn't really DELETE_BUFS related, but I think it makes the code
a bit easier to follow.

> -			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 +2216,15 @@ static int __find_plane_by_offset(struct vb2_queue *q, unsigned long off,
>  		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 = (off >> PLANE_INDEX_SHIFT) & BUFFER_INDEX_MASK;
> +	*plane = (off >> PAGE_SHIFT) & PLANE_INDEX_MASK;
>  
> -		for (plane = 0; plane < vb->num_planes; ++plane) {
> -			if (vb->planes[plane].m.offset == off) {
> -				*_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 +2304,7 @@ int vb2_mmap(struct vb2_queue *q, struct vm_area_struct *vma)
>  {
>  	unsigned long off = 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 +2333,10 @@ 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, off, &buffer, &plane);
> +	ret = __find_plane_by_offset(q, off, &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 +2364,7 @@ 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 +2378,7 @@ unsigned long vb2_get_unmapped_area(struct vb2_queue *q,
>  {
>  	unsigned long off = pgoff << PAGE_SHIFT;
>  	struct vb2_buffer *vb;
> -	unsigned int buffer, plane;
> +	unsigned int plane;
>  	void *vaddr;
>  	int ret;
>  
> @@ -2392,12 +2388,10 @@ 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, off, &buffer, &plane);
> +	ret = __find_plane_by_offset(q, off, &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;

Regards,

	Hans

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ