Re: [PATCH] AF903X driver update, v1.02

[Malcolm Priestley <tvboxspy@gmail.com>]

On Mon, 2012-03-12 at 18:10 +0100, Gianluca Gennari wrote:
> Il 11/03/2012 16:23, Hans-Frieder Vogt ha scritto:
> > This is an update of the patch "Basic AF9035/AF9033 driver" that I send to the mailing list on 22 Feb.
> > The driver provides support for DVB-T USB 2.0 sticks based on AF9035/AF9033. Currently supported devices:
> > - Terratec T5 Ver.2 (also known as T6), Tuner FC0012
> > - Avermedia Volar HD Nano (A867R), Tuner MxL5007t
> > 
> > Ver 1.02 of the driver includes the following changes compared to the initial version:
> > 
> > - significantly reduced number of mutex calls (only remaining protection in low-level af903x_send_cmd)
> >   this change made some multiply defined function unnecessary (_internal functions and non _internal functions)
> >   maybe this reduction was a bit too agressive, but I didn't get any problems in several days testing 
> > - reduced number of iterations in loop for lock detection (should improve response)
> > - correct errors in initial contribution and add proper entries in dvb-usb-ids.h (thanks to Gianluca Gennari)
> > - removed unnecessary (loading of rc key table) and commented out code
> > - minor cleanup (e.g. af903x_fe_is_locked)
> > 
> 
> Hi Hans,
> I just tested the new version of the driver on my a867 stick (single
> mxl5007 tuner), on Ubuntu 11.10 with a vanilla 3.2.9 kernel and the
> latest media_build tree installed.
> 
> I can confirm the two main issues I found on the previous version are
> completely solved.
> In fact, Kaffeine is much more responsive now, and I can also open the
> window with the signal strength/quality bars without any sluggishness.
> I also unplugged the device from the USB port under several different
> conditions (watching a channel, doing a channel scan and so on) without
> any kernel crash.
> 
> I tested several channels (UHF, VHF, QAM64, QAM16, QPSK, 8MHz, 7MHz,...)
> with no problem: the reception was stable in all configurations. Zapping
> is also much faster with this version. I will run longer tests in the
> coming days.
> 
> The only minor issues that I could find are related to channel scanning.
> A full scan with Kaffeine takes much longer than with other sticks (most
> of the time is spent on empty frequencies).
> Despite the long wait, some frequency is not locked (for example, the
> very first VHF frequency, channel 5, is missed 100% of the times).
> This problem seem to disappear disabling the PID filter (but I need
> further testing to draw any conclusion).

This is a problem with the stream USB urb buffers being too big with the
PID filter on. Scan, w_scan etc only request network information PIDs

Network information PIDs are not transmitted that often so the buffer
takes ages to fill up and scan etc times out.

Extended time out is required.

scan -5
w_scan -F

A solution is to keep the stream USB urb buffer size low. 940(5 * 188)
bytes is the ideal, but there is a concern that the USB bus may not be
able to reach throughput with just 10 buffers on HD channels or when the
user turns the PID filter off. A buffer size of 3948 (21 * 188) seems a
compromise.


Regards


Malcolm












> Also, during a full scan, Kaffeine becomes sluggish again.
> 
> In the end, the driver seem to work fine now, with only a few minor issues.
> 
> Thank you for your effort.
> 
> Regards,
> Gianluca