Skip to contentSkip to conversation

v4l device in userspace

DISCUSSION Salvatore Sanfilippo 4 participants 5 messages Jun 16, 2006 – Jun 19, 2006

lore
Share & tools
lore
#0 lore raw

Hello, I'm trying to implement a v4l device driver for symbian based smart phones. In theory it is very simple:

I've a little program running in the phone, capturing images from the camera and sending it to the linux box via bluetooth.

In the linux box side, I've a deamon capturing this images (via a bluetooth SP channel), and.... I've to pass the images to a fake v4l device driver that actually gets the images form userspace.

Basically I've to pass by the kernel just for the interface, and not to do real kernel-side work (like to access to the some kind of hardware).

So I've some questions ( thanks in advance for any reply).

1) What's the best way to pass relatively high-band data between the v4l fake driver and userspace? A char device will do the work? ioctl?

2) What about some way to handle ioctl directly from userspace? Given this support I may implement the whole code in userspace. And I guess there are a lot of other real world problems that can be handled in userspace given the ability to handle ioctl from there.

If you think 2) is reasonable I may actually implement some simple form of generic
char driver that just allows userspace programs to handle read/write/ioctl opreations, and then use this to fix
my real issue.

Thank you very much for the help,
and sorry if there is something conceptually wrong in my questions.

Regards,
Salvatore

P.S. please take me in CC as I'm not subscribed to the linux kernel mailing list.

Signature · 3 lines
Salvatore 'antirez' Sanfilippo
We're programmers. Programmers are, in their hearts, architects -- Joel Spolsky
http://www.invece.org
#1 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Salvatore" == Salvatore Sanfilippo <[email protected]> writes:

Salvatore> Hello, I'm trying to implement a v4l device driver for Salvatore> symbian based smart phones. In theory it is very Salvatore> simple:

Salvatore> I've a little program running in the phone, capturing Salvatore> images from the camera and sending it to the linux box Salvatore> via bluetooth.

Salvatore> In the linux box side, I've a deamon capturing this Salvatore> images (via a bluetooth SP channel), and.... I've to Salvatore> pass the images to a fake v4l device driver that Salvatore> actually gets the images form userspace.

Salvatore> Basically I've to pass by the kernel just for the Salvatore> interface, and not to do real kernel-side work (like to Salvatore> access to the some kind of hardware).

Salvatore> So I've some questions ( thanks in advance for any Salvatore> reply).

Salvatore> 1) What's the best way to pass relatively high-band Salvatore> data between the v4l fake driver and userspace? A char Salvatore> device will do the work? ioctl?

Salvatore> 2) What about some way to handle ioctl directly from Salvatore> userspace? Given this support I may implement the whole Salvatore> code in userspace. And I guess there are a lot of Salvatore> other real world problems that can be handled in Salvatore> userspace given the ability to handle ioctl from there.

Salvatore> If you think 2) is reasonable I may actually implement Salvatore> some simple form of generic char driver that just Salvatore> allows userspace programs to handle read/write/ioctl Salvatore> opreations, and then use this to fix my real issue.

Salvatore> Thank you very much for the help, and sorry if there is Salvatore> something conceptually wrong in my questions.

Sounds like you should take a look at the v4l loopback device
(http://www.lavrsen.dk/twiki/bin/view/Motion/VideoFourLinuxLoopbackDevice).
Otherwise it may be better to ask the question on the v4l mailing list
(https://listman.redhat.com/mailman/listinfo/video4linux-list).

Anyway, since you already capture the video, why do you have to pipe it through a v4l device?

Marcus

Signature · 5 lines
/--------------------------------------------------------------------\
| Dr. Marcus O.C. Metzler        |                                   |
| [email protected]            | http://www.metzlerbros.de/        |
\--------------------------------------------------------------------/
 |>>>             Quis custodiet ipsos custodes                 <<<|
#2 lore raw
Quoted 1 line · Marcus Metzler

Sounds like you should take a look at the v4l loopback device

Thanks this may contain useful code indeed.

Quoted 2 lines

Anyway, since you already capture the video, why do you have to pipe it through a v4l device?

In order to make every application using the v4l API working with the phone cam without modifications.

Thanks,
Salvatore

Signature · 3 lines
Salvatore 'antirez' Sanfilippo
We're programmers. Programmers are, in their hearts, architects -- Joel Spolsky
http://www.invece.org
#3 lore raw
In reply to Salvatore Sanfilippo
Quoted 43 lines · Salvatore Sanfilippo

Hello, I'm trying to implement a v4l device driver for symbian based smart phones. In theory it is very simple:

I've a little program running in the phone, capturing images from the camera and sending it to the linux box via bluetooth.

In the linux box side, I've a deamon capturing this images (via a bluetooth SP channel), and.... I've to pass the images to a fake v4l device driver that actually gets the images form userspace.

Basically I've to pass by the kernel just for the interface, and not to do real kernel-side work (like to access to the some kind of hardware).

So I've some questions ( thanks in advance for any reply).

1) What's the best way to pass relatively high-band data between the v4l fake driver and userspace? A char device will do the work? ioctl?

2) What about some way to handle ioctl directly from userspace? Given this support I may implement the whole code in userspace. And I guess there are a lot of other real world problems that can be handled in userspace given the ability to handle ioctl from there.

If you think 2) is reasonable I may actually implement some simple form of generic
char driver that just allows userspace programs to handle read/write/ioctl opreations, and then use this to fix
my real issue.

Thank you very much for the help,
and sorry if there is something conceptually wrong in my questions.

Quoted 6 lines

Regards,
Salvatore

P.S. please take me in CC as I'm not subscribed to the linux kernel mailing list.

I recommend resubmitting this question to a mailing list that actually focuses on this topic:

[email protected]

However, I highly recommend that you subscribe before you post, as the redhat server adds a REPLY-TO header to the video4linux list emails, and has been known to drop cc's:

http://www.redhat.com/mailman/listinfo/video4linux-list

See you there!

-Michael Krufky

Signature · 1 line
Michael Krufky
#4 lore raw
In reply to Salvatore Sanfilippo

Hi!

Quoted 3 lines

I've a little program running in the phone, capturing images from the camera and sending it to the linux box via bluetooth.

Nice!

Quoted 19 lines

Basically I've to pass by the kernel just for the interface, and not to do real kernel-side work (like to access to the some kind of hardware).

So I've some questions ( thanks in advance for any reply).

1) What's the best way to pass relatively high-band data between the v4l fake driver and userspace? A char device will do the work? ioctl?

2) What about some way to handle ioctl directly from userspace? Given this support I may implement the whole code in userspace. And I guess there are a lot of other real world problems that can be handled in userspace given the ability to handle ioctl from there.

Quoted 6 lines

If you think 2) is reasonable I may actually implement some simple form of generic
char driver that just allows userspace programs to handle read/write/ioctl opreations, and then use this to fix
my real issue.

You probably want to do something v4l specigic... but generic userspace driver
able to do read/write/ioctl would be very nice. Lots of devices these days are on usb, and that can be done from userspace, for example.
Pavel

Signature · 1 line
64 bytes from 195.113.31.123: icmp_seq=28 ttl=51 time=448769.1 ms