[linux-dvb] HVR-4000 driver problems - i2c error

[Steven Toth]

Mika Laitio wrote:
>> The ~stoth/hg/s2 has no DVB-T support on the HVR4000 yet. Those 
>> patches will appear very shortly in ~stoth/hg/s2-mfe.
> 
> Thanks,
> 
> I build and installed drivers from s2-mfe branch and now atleast the 
> dvb-t and dvb-s scan were working both for hvr-1300 and hvr-4000.
> With hvr-4000, the scan worked with
> "./scan -a 1 -f 1 "... parameters.
> 
>>>  WIth liplianinis multiproto version the selection between DVB-S and 
>>> DVB-T
>>>  works by using the "options cx88-dvb frontend=1" but I am seeing the 
>>> i2c
>>>  errors described below.
>>>
>>>  Could you have any URL and changeset tag to patch in some repository 
>>> where
>>>  this I2C thing has been fixed?
>>
>> I'm speculating that your issue is the same issue I fixed sometime ago 
>> (2-3 months in the master repo). I'd suggest you wait for the ~ 
>> stoth/hg/s2-mfe patches to appear later tonight and test again.
>>
>> That tree (and ~stoth/hg/s2 for that matter) have the i2c fix I'm 
>> referring to.
> 
> I tried to find the fix by checking from the log your commits for cx88 
> couple of months ago and could not find the one you were referring.
> Then I remembered that in 2.6.25 kernel, similar errors for HVR-1300 
> could be fixed by defining the radio information for the board structure.
> 
> So I tried the same trick also for hvr-4000, and that made dvb-t working 
> without I2C errors also for HVR4000. ( with liplianis version of 
> multiproto tree.)
> 
> --- cx88-cards.c_old    2008-09-25 00:51:18.000000000 +0300
> +++ cx88-cards.c    2008-09-24 10:44:15.000000000 +0300
> @@ -1480,6 +1480,10 @@
>          }},
>          /* fixme: Add radio support */
>          .mpeg           = CX88_MPEG_DVB,
> +        .radio = {
> +            .type   = CX88_RADIO,
> +            .gpio0    = 0xe780,
> +        },
>      },
>      [CX88_BOARD_HAUPPAUGE_HVR4000LITE] = {
>          .name           = "Hauppauge WinTV-HVR4000(Lite) DVB-S/S2",
> 
> In S2 and S2-mfe, the dvb-t worked even without radio data defined for 
> hvr-4000.
> 
> Other non-API related changes that I could found between multiproto and 
> s2 trees for CX88 were pretty small. Is this udelay really needed?
> 
> @@ -2594,8 +2598,9 @@
>          break;
>      case CX88_BOARD_HAUPPAUGE_HVR3000: /* ? */
>      case CX88_BOARD_HAUPPAUGE_HVR4000:
> -        /* Init GPIO for DVB-S/S2/Analog */
> -        cx_write(MO_GP0_IO,core->board.input[0].gpio0);
> +        /* Init GPIO */
> +        cx_write(MO_GP0_IO, core->board.input[0].gpio0);
> +        udelay(1000);
>          break;
> 
>      case CX88_BOARD_PROLINK_PV_8000GT:
> @@ -2939,6 +2944,7 @@
>          cx88_card_list(core, pci);
>      }
> 
> +    memset(&core->board, 0, sizeof(core->board));
>      memcpy(&core->board, &cx88_boards[core->boardnr], 
> sizeof(core->board));
> 
>      info_printk(core, "subsystem: %04x:%04x, board: %s [card=%d,%s]\n",
> 
> Those did not help

Thanks.

I'm going to rebase the s2-mfe tree from master (containing the s2api). 
Darron also has a i2c related fix (which isn't perfect) but does address 
an issue like this.

I'll rebase and push both of those patches into ~stoth/s2-mfe, if you 
can retest after this then that would be useful.

Watch for changes to this tree tonight, it may help.

Regards,

- Steve