Re: [PATCH] lmedm04 2.06 conversion to dvb-usb-v2 version 2

[Antti Palosaari <crope@iki.fi>]

On 08/06/2012 11:21 PM, Malcolm Priestley wrote:
> Conversion of lmedm04 to dvb-usb-v2
>
> Functional changes m88rs2000 tuner now uses all callbacks.
> TODO migrate other tuners to the callbacks.
>
> This patch is applied on top of [BUG] Re: dvb_usb_lmedm04 crash Kernel (rs2000)
> http://patchwork.linuxtv.org/patch/13584/
>
>
> Signed-off-by: Malcolm Priestley <tvboxspy@gmail.com>

Could you try to make this driver more generic?

You use some internals of dvb-usb directly and most likely those are 
without a reason. For example data streaming, lme2510_kill_urb() kills 
directly urbs allocated and submitted by dvb-usb. Guess that driver is 
broken just after someone changes dvb-usb streaming code.

lme2510_usb_talk() could be replaced by generic dvb_usbv2_generic_rw().

What is function of lme2510_int_read() ? I see you use own low level URB 
routines for here too. It starts "polling", reads remote, tuner, demod, 
etc, and updates state. You would better to implement I2C-adapter 
correctly and then start Kernel work-queue, which reads same information 
using I2C-adapter. Or you could even abuse remote controller polling 
function provided by dvb-usb.

lme2510_get_stream_config() enables pid-filter again over the dvb-usb, 
but I can live with it because there is no dynamic configuration for 
that. Anyhow, is that really needed?

I can live with the pid-filter "abuse", but killing stream URBs on 
behalf of DVB-USB is something I don't like to see. If you have very 
good explanation and I cannot fix DVB USB to meet it I could consider 
that kind of hack. And it should be documented clearly adding necessary 
comments to code.

Re-implementing that driver to use 100% DVB-USB services will reduce 
around 50% of code or more.

regards
Antti


-- 
http://palosaari.fi/