Very Computer

I'm writing an application to use video4linux. I'm using the ibmcam
driver and with an IBM *.

I capture frames by repeatedly calling:

read(cam->fd, buffer, (cam->vw.width) * (cam->vw.height) *
sizeof(pixel_t) );

Do I need to do something between each call to read()?

I'm dumping each buffer out to a file (I've done ppm and jpg) and
every other frame has garbled data in it. I've slowed down the rate at
which I call read (like once per second), but it doesn't make a
difference.

What am I doing wrong?

BTW, this is on a RedHat 7.1 system.

Thanks!
John