> For now, let me give a quick explanation of the basics of videobuf.
>
> ---
>
(part 2)
As you know, the original author of videobuf is Gerd. At the changes I
did, I've tried to preserve, as much as possible, the code outside
videobuf without changes(*).
(*) This is also true for the binary code that videobuf-core +
videobuf_dma_sg executes: It is almost the same as the original
videobuf, but separated into two separate modules, with distinct
functions.
a) The first one is about the inherited videobuf_buffer class at each
driver.
The approach used on videobuf is somehow different from what other parts
of the Linux Kernel does. For this to work, the first part of a
videobuf_buffer inherited code should do:
cx23885_buffer {
struct videbuf_buffer;
...
}
Otherwise, the videobuf code will fail.
IMO, it would be better to use container_of as you suggested, but this
would mean rewrite more code at the drivers, and do more tests.
b) The destructor used to free videobuf_buffer memory is just kfree. So,
all memory for the entire class should be allocated with just one
kmalloc. The memory model is something like:
struct derivated_class {
struct cx23885_buffer {
struct videobuf_buffer;
// cx23885 own data
}
// dma s/g own data
};
So, videobuf_pci_malloc do something like:
buf=kmalloc (sizeof(struct derivated_class), GFP_KERNEL);
To free a videobuf, you just need to do:
free (buf);
--
Cheers,
Mauro
_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
