diff options
author | Jonathan Corbet <corbet-v4l@lwn.net> | 2006-11-04 07:25:53 -0500 |
---|---|---|
committer | Mauro Carvalho Chehab <mchehab@infradead.org> | 2006-12-10 05:51:13 -0500 |
commit | d905b382d797a213e15868cbf3204f50ed52e30b (patch) | |
tree | 9e328aa8e273bcc0656cd0bb4a51020ea10ef008 /Documentation/video4linux | |
parent | 9c4dfadbde3cfa78b92c28597125b8c41d36ffd0 (diff) |
V4L/DVB (4797): Marvell 88ALP01 "cafe" driver
A driver for the Marvell M88ALP01 "CAFE" CMOS integrated camera
controller. This driver has been renamed "cafe_ccic" since my previous
patch set.
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Diffstat (limited to 'Documentation/video4linux')
-rw-r--r-- | Documentation/video4linux/cafe_ccic | 54 |
1 files changed, 54 insertions, 0 deletions
diff --git a/Documentation/video4linux/cafe_ccic b/Documentation/video4linux/cafe_ccic new file mode 100644 index 000000000000..88821022a5de --- /dev/null +++ b/Documentation/video4linux/cafe_ccic | |||
@@ -0,0 +1,54 @@ | |||
1 | "cafe_ccic" is a driver for the Marvell 88ALP01 "cafe" CMOS camera | ||
2 | controller. This is the controller found in first-generation OLPC systems, | ||
3 | and this driver was written with support from the OLPC project. | ||
4 | |||
5 | Current status: the core driver works. It can generate data in YUV422, | ||
6 | RGB565, and RGB444 formats. (Anybody looking at the code will see RGB32 as | ||
7 | well, but that is a debugging aid which will be removed shortly). VGA and | ||
8 | QVGA modes work; CIF is there but the colors remain funky. Only the OV7670 | ||
9 | sensor is known to work with this controller at this time. | ||
10 | |||
11 | To try it out: either of these commands will work: | ||
12 | |||
13 | mplayer tv:// -tv driver=v4l2:width=640:height=480 -nosound | ||
14 | mplayer tv:// -tv driver=v4l2:width=640:height=480:outfmt=bgr16 -nosound | ||
15 | |||
16 | The "xawtv" utility also works; gqcam does not, for unknown reasons. | ||
17 | |||
18 | There are a few load-time options, most of which can be changed after | ||
19 | loading via sysfs as well: | ||
20 | |||
21 | - alloc_bufs_at_load: Normally, the driver will not allocate any DMA | ||
22 | buffers until the time comes to transfer data. If this option is set, | ||
23 | then worst-case-sized buffers will be allocated at module load time. | ||
24 | This option nails down the memory for the life of the module, but | ||
25 | perhaps decreases the chances of an allocation failure later on. | ||
26 | |||
27 | - dma_buf_size: The size of DMA buffers to allocate. Note that this | ||
28 | option is only consulted for load-time allocation; when buffers are | ||
29 | allocated at run time, they will be sized appropriately for the current | ||
30 | camera settings. | ||
31 | |||
32 | - n_dma_bufs: The controller can cycle through either two or three DMA | ||
33 | buffers. Normally, the driver tries to use three buffers; on faster | ||
34 | systems, however, it will work well with only two. | ||
35 | |||
36 | - min_buffers: The minimum number of streaming I/O buffers that the driver | ||
37 | will consent to work with. Default is one, but, on slower systems, | ||
38 | better behavior with mplayer can be achieved by setting to a higher | ||
39 | value (like six). | ||
40 | |||
41 | - max_buffers: The maximum number of streaming I/O buffers; default is | ||
42 | ten. That number was carefully picked out of a hat and should not be | ||
43 | assumed to actually mean much of anything. | ||
44 | |||
45 | - flip: If this boolean parameter is set, the sensor will be instructed to | ||
46 | invert the video image. Whether it makes sense is determined by how | ||
47 | your particular camera is mounted. | ||
48 | |||
49 | Work is ongoing with this driver, stay tuned. | ||
50 | |||
51 | jon | ||
52 | |||
53 | Jonathan Corbet | ||
54 | corbet@lwn.net | ||