revisiting scullpipe implementation

[elezegarcia@yahoo.com.ar (Ezequiel García)]

Hi everybody,

I don't know if anyone cares about this subject. But here it goes anyway. 
I've been reading through the excellent Linux Device Driver book (third ed.) for the past months. 
I have decided to give it a try to the drivers samples, starting by scull. After some minor updates (mostly related to new mutex interface), I started playing with scull and scullpipe. 

The thing is the scullpipe implementation makes buffer initialization, and sets read and write pointer to zero at device open. Acordingly, it releases the buffer at device close.
This implementation makes this possible:

(first, on one terminal) $ cat /dev/scullpipe0

(second, another) $ dd if=/dev/zero of=/dev/scullpipe0 bs=512 count=1

But not the other way round:

(first, write) $ dd if=/dev/zero of=/dev/scullpipe0 bs=5000 count=5

(then, read) $ cat /dev/scullpipe0

Because when the reader (cat) opens the device the read/write pointers are made equal and buffer looks empty. (Besides, the buffer resets itself on device release; so it does not persist when nobody is using it.)

Questions: Did I misunderstand the driver's idea and this behavior is not buggy, or ... is the code just aimed at learning so it does not matter?

Actually it is pretty easy (and good training for starters) to change it. I had to move some code from device open to device init and from device release to device clean_up.

Maybe someone is reading the book and might be interested.

Greetings,
Ezequiel. 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20110522/b354d766/attachment.html