Ubuntu
linux-source-2.6.20 package

Missing frontend for avermedia dvb-t 771 (yet working for another 771 card)

Bug #99120 reported by Richard Brooklyn on 2007-03-30

Affects		Status	Importance	Assigned to	Milestone
	linux-source-2.6.20 (Ubuntu)	Invalid	Low	Unassigned

Bug Description

Binary package hint: linux-image-2.6.20-13-generic

Hi,

I have two identical (as far as I know) AverMedia 771 DVB-T cards installed on this system. This system does PVR duties with mythtv, amongst other things.

Running Gentoo, and later Dapper, I was able to use both DVB cards (I think: I'm certain it was okay in Gentoo, not as certain for Dapper, as it only ran for about a week and I didn't test this) without issues. However, since re-formatting and installing Fiesty, the system only seems to work correctly with one of them. It seems to complain about a 'missing frontend' for one of the cards. Here is the relevant dmesg parts:

[ 25.789445] DVB: registering new adapter (bttv0).
[ 25.842741] DVB: registering frontend 0 (Zarlink MT352 DVB-T)...
[ 25.842825] DVB: registering new adapter (bttv1).
[ 25.843263] mt352_read_register: readreg error (reg=127, ret==-5)
[ 25.843315] dvb-bt8xx: A frontend driver was not found for device 109e/0878 subsystem 1461/0771

I have no idea if that mt352 error is relevant, or what it means.

relevant lspci -vnn output:
03:01.0 Multimedia video controller [0400]: Brooktree Corporation Bt878 Video Capture [109e:036e] (rev 11)
        Subsystem: Avermedia Technologies Inc AverMedia AVerTV DVB-T 771 [1461:0771]
        Flags: bus master, medium devsel, latency 64, IRQ 17
        Memory at d0003000 (32-bit, prefetchable) [size=4K]
        Capabilities: [44] Vital Product Data
        Capabilities: [4c] Power Management version 2

03:01.1 Multimedia controller [0480]: Brooktree Corporation Bt878 Audio Capture [109e:0878] (rev 11)
        Subsystem: Avermedia Technologies Inc AverMedia AVerTV DVB-T 771 [1461:0771]
        Flags: bus master, medium devsel, latency 64, IRQ 17
        Memory at d0002000 (32-bit, prefetchable) [size=4K]
        Capabilities: [44] Vital Product Data
        Capabilities: [4c] Power Management version 2

03:02.0 Multimedia video controller [0400]: Brooktree Corporation Bt878 Video Capture [109e:036e] (rev 11)
        Subsystem: Avermedia Technologies Inc AverMedia AVerTV DVB-T 771 [1461:0771]
        Flags: bus master, medium devsel, latency 64, IRQ 18
        Memory at d0001000 (32-bit, prefetchable) [size=4K]
        Capabilities: [44] Vital Product Data
        Capabilities: [4c] Power Management version 2

03:02.1 Multimedia controller [0480]: Brooktree Corporation Bt878 Audio Capture [109e:0878] (rev 11)
        Subsystem: Avermedia Technologies Inc AverMedia AVerTV DVB-T 771 [1461:0771]
        Flags: bus master, medium devsel, latency 64, IRQ 18
        Memory at d0000000 (32-bit, prefetchable) [size=4K]
        Capabilities: [44] Vital Product Data
        Capabilities: [4c] Power Management version 2

This is using the amd64 version of the distro. Everything is up-to-date as of 30th of March 2007.

Revision history for this message

Richard Brooklyn (rbrooklyn) wrote on 2007-04-02:

I get the exact same error message in Mandrive One LiveCD 2007.1

This this is either a hardware issue (in which case, I apologise for wasting time and bandwidth) or a kernel issue.

Interestingly, the mt352_read_register error message appears in the same place (when I modprobe the dvb-bt8xx module), so it's probably significant. I wish I knew what that error code meant.

I shall do more digging later.

Revision history for this message

Phillip Lougher (phillip-lougher) wrote on 2007-04-02:

All DVB-T cards need a front end tuner (FET) to tune to the correct terrestrial multiplex, and then to convert the signal into something that can be de-multiplexed to obtain the desired audio/video MPEG streams.

Unfortunately, there are a number of these FETs, all with different programming interfaces. The dmesg error log shows that the MT352 FET driver has been loaded by the DVB software. It appears that the MT352 driver code when probing the FET on your board has failed to read it. Unsurprisingly, because the MT352 software can't read the chip, it doesn't recognise it as a MT352 FET.

There are a number of possible reasons for this. Because you say it only happens on the second card, it is probably safe to rule out obvious causes such as not having an MT352, or bugs in the chip reading code. The most probable cause is code which assumes only one card is present.

To go any further with this bug, we need to know some more information.

* Can you confirm it only happens on the second card?
* What kernel version does the Mandrive One LiveCD 2007.1 use?
* Can you supply the full dmesg output (dmesg > dmesg,txt and attach it please).

What would be really useful is finding a kernel where both cards definately worked. You mentioned it worked on Gentoo ? What kernel was that using?

Thanks

Changed in linux-source-2.6.20:
assignee:	nobody → phillip-lougher
importance:	Undecided → Low
status:	Unconfirmed → Confirmed
status:	Confirmed → Needs Info

Revision history for this message

Richard Brooklyn (rbrooklyn) wrote on 2007-04-02:

full dmesg output Edit (26.9 KiB, text/plain)

Hello Phillip,

Thank you for your lengthy reply. It's nice to see this bug is being taken seriously, despite my amateurish attempts at reporting it.

Attached is the full dmesg report, as requested.

I cannot confirm right now that it only happens on the second card. I'll have a play with removing the hardware tomorrow and get back to you on that one.

Mandriva uses kernel 2.6.17 according to it's distrowatch page. I shall confirm this tomorrow by booting the livecd again.

The Gentoo system used kernel 2.6.18 or 2.6.19, I'm honestly not sure.

I did find this, which is concerning me. Maybe this is a hardware issue after all: http://www.linuxtv.org/pipermail/linux-dvb/2006-January/007910.html

Is there a way of forcing the driver to see the board as having a specific frontend? I'm certain the cards are identical, and if the driver can see past this error, I guess it might work?

Revision history for this message

Richard Brooklyn (rbrooklyn) wrote on 2007-04-08:

Hi,

Sorry to be taking so long with getting back to you. Work has left me with very little free time or energy recently.

To confirm. Mandriva's kernel version shows as 2.6.17-13mdv in uname.

I did some digging, with surprising results:

When card one is installed on it's own: No issues.

When card two is installed on it's own: No issues. Registers same zarlink frontend as listed in original bug report.

When card one and two are installed, but are in each other's PCI slots: No issues... Mythtv picks up both cards just fine and I'm watching one program whilst recording another as I write this.

I'm slightly freaked out by this. Maybe the second card was just loose the whole time? Would that even make sense? My assumption would be that it wouldn't work at all under those conditions.

Please close this bug. Hopefully it will help anyone with the same issue in the future.

Thank you for your time, Phillip.

Richard Brooklyn (rbrooklyn) on 2007-04-10

Changed in linux-source-2.6.20:
status:	Needs Info → Rejected

Revision history for this message

Phillip Lougher (phillip-lougher) wrote on 2007-04-10:

Thanks for your detailed follow-up information, and I unfortunately recognise and appreciate your comment about work leaving little free time and energy :)

It does seem odd both cards are working in each other's PCI slots. As you said, it may have been a loose card, or perhaps there is a subtle race-condition in the driver. As the driver is now working I'll close this bug. If it occurs again, please re-open the bug, Thanks.