talos-op-linux/drivers/media/platform/coda, branch master

media: coda: avoid hardware lockups with more than 13 capture buffers

2019-12-16T09:09:12+00:00

When decoding using the CODA internal rotator (for example NV12 capture), currently the value vb2_buf.index + CODA_MAX_FRAMEBUFFERS (19) is written into the DEC_PIC_ROT_INDEX register. At least with firmware version 3.1.1 this causes CODA hangups as soon as the register value reaches 32. Instead, always write CODA_MAX_FRAMEBUFFERS. Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: remove redundant platform_get_irq error message

2019-12-16T09:08:29+00:00

Both platform_get_irq_byname() and platform_get_irq() already print an error. Remove the redundant error message. Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: jpeg: add CODA960 JPEG encoder support

2019-12-16T09:07:59+00:00

This patch adds JPEG encoding support for CODA960, handling the JPEG hardware directly. A separate JPEG encoder video device is created due to the separate hardware unit and different supported pixel formats. While the hardware can not change subsampling on the fly, it can encode 4:2:2 subsampled images into JPEGs of the same subsampling. There are two additional tracepoints added to the coda driver that can be used together with the v4l2:v4l2_qbuf and v4l2:v4l2_dqbuf tracepoints to to follow video frames through the mem2mem device when encoding or decoding with the CODA960 JPEG codec: coda:coda_jpeg_run coda:coda_jpeg_done Signed-off-by: Philipp Zabel Tested-by: Richard Leitner [hverkuil-cisco@xs4all.nl: removed unused k and q_data_dst variables] Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: jpeg: add JPEG register definitions for CODA960

2019-12-16T09:04:21+00:00

The CODA960 JPEG codec is controlled directly from the host, there is no support in the BIT processor firmware. This patch adds the necessary register definitions. Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: jpeg: merge Huffman table bits and values

2019-12-16T09:03:51+00:00

The Huffman bits tables are always 16 bytes long, and they are always followed directly by the values tables, both in hardware and in JPEG files. Just merge the two tables. Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: do not skip finish_run if aborting

2019-12-16T09:00:11+00:00

Always call finish_run when the hardware signals completion. This will allow JPEG contexts to clean up even if job_abort was called during the device_run. Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: disable decoder crop selections

2019-11-10T06:27:40+00:00

Disable output side crop selections for the decoder. This fixes the following v4l2-compliance complaint: fail: v4l2-test-formats.cpp(1576): IS_DECODER(node) test Cropping: FAIL Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: disable encoder compose selections

2019-11-10T06:26:11+00:00

Disable capture side compose selections for the encoder. This fixes the following v4l2-compliance complaint: fail: v4l2-test-formats.cpp(1662): IS_ENCODER(node) test Composing: FAIL Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: drop unused irqlock

2019-11-09T08:10:42+00:00

The irqlock spinlock has been unused from the start. Remove it. Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: coda: request to skip kernel mapping for decoded buffers

2019-11-05T11:59:50+00:00

The kernel driver never touches the decoded buffers with the CPU. All accesses are either done by hardware DMA masters or userspace mapping the buffers. This means we don't need a kernel virtual address mapping for those buffers at all. As those buffers are usually quite large, we can save a good deal of kernel vmalloc space by requesting to not have a kernel mapping set up for them. Signed-off-by: Lucas Stach Signed-off-by: Philipp Zabel Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab