[linux-dvb] bt8xx docu fixes and other docu fixes; pci subsystem id flakyness

Sigmund Augdal Helberg sigmund at snap.tv
Tue Mar 28 08:23:11 CEST 2006

On Fri, 2006-03-24 at 22:46 +0100, Gunther Mayer wrote:
> Sigmund Augdal Helberg wrote:
> >On Fri, 2006-03-24 at 15:04 +0100, Uwe Bugla wrote:
> >  
> >
> >>>--- Ursprüngliche Nachricht ---
> >>>Von: Edgar Toernig <froese at gmx.de>
> >>>An: linux-dvb at linuxtv.org
> >>>Betreff: Re: [linux-dvb] bt8xx docu fixes and other docu fixes
> >>>Datum: Thu, 23 Mar 2006 23:03:16 +0100
> >>>
> >>>Ralph Metzler wrote:
> >>>      
> >>>
> >>>>No, the hardware is not not working, it is not there.
> >>>>The first Twinhan cards did not have an EEPROM. 
> >>>>        
> >>>>
> >>>Oh.  So how are they supposed to work now when the
> >>>bt878_probe only accepts cards with a know subsystem
> >>>id?
> >>>
> >>>Ciao, ET.
> >>>
> >>>_______________________________________________
> >>>linux-dvb mailing list
> >>>linux-dvb at linuxtv.org
> >>>http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
> >>>
> >>>      
> >>>
> >>My proposal:
> >>If I get 5 test reports here proving me that autodetection works for all
> >>TwinHan cards, I´ll sit down and write another docu fix eliminating that
> >>card specific annotations.
> >>Question: Are the annotations for running multiple cards also unnecessary,
> >>or is that just an assumption?
> >>    
> >>
> >I can confirm that autodetection does not allways work for some twinhan
> >cards. That is for the same card autodetection some times work some
> >  
> >
> Possible root causes:
> a) the card (or eeprom) is flakey; perhaps temperature dependent
> b) the driver leaves the i2c bus on the card in a random state, so 
> sometimes the eeprom does not respond correctly
The cause is almost certainly related to timing and reset during the
first stages of a cold boot. The card has a led and when subsystem id is
missing the led is not lit. This state kept over warm-boots and is only
changed by either a cold boot or but forcing card=0x71 when modprobing
bttv and then load dst and dvb_bt8xx. The led lights up and the card
starts reporting the currect subsystem id once the rdc_8820_reset call
in dst.c:probe() is finnished.
> @Sigmund:
> See if the correct pci subsystem id is reported ("lspci -vn") for
> - multiple "halt" + plug off the power cord for 30 seconds (aka ice-cold 
> reboots), _without the driver ever loaded_
> - multiple "reboot" (aka warm reboots), _without the driver ever loaded_
> - repeat with driver loaded
> At which point does "lspci -vn" stop to report the correct id?
I do not have the time nor will to carry out these tests. I have
allready a good understanding about when this issue happens and how to
deal with it.


> To rule out flakeyness "multiple= 10 times"
> >times not. When autodetection fails and no card= option is given for
> >that card the computer locks up.
> >  
> >
> The lock up is a fatal bug and must be fixed in the driver.

More information about the linux-dvb mailing list