regression(?) - hauppage quad card tuning

regression(?) - hauppage quad card tuning

[media]

Hi

I need some help determining if there's been a regression for
one of my DVB TV cards. Details below.

I have a Hauppage QuadHD DVB card
which has two multimedia controllers:

 Conexant Systems, Inc. CX23887/8 PCIe Broadcast Audio and Video
   Decoder with 3D Comb (rev 04)

i.e. PCI ID 14f1:8880 (rev 04)

With subvendor:subdevice 0700:0x6a28.

The cards have SiLabs Si2157 tuners.

I had a working system running Debian Bullseye (linux 5.10.197-1).

After I upgraded to Debian Bookworm (running 6.1.52-1)
the tuners in this card no longer get a lock.

I also tried a 'backports' kernel, 6.5.10-1~bpo12+1.
This is based on the upstream version 6.1.66-1.
This kernel also fails to tune and get a lock.

The other symptom, which may or may not be relevant,
is 'i2cdetect -l' returns nothing, though there are
various i2c device detections on dmesg.

There is a second card in the system that tunes and locks on
just fine, with both the 6.1 and 6.5 kernels.
It is a  Philips Semiconductors SAA7164 (rev 81)
marketed as a (Hauppauge WinTV HVR-2200,
i.e. PCI ID 1131:7164, subbvendor:subdevice 0070:8953.

So I can rule out lack of signal; the signal is routed into
the quad card first and then to the working dual-tuner card.

Things I have tried

 - checking the firmware blobs are all present and loading
 - setting debug=6 on the cx23885 module
 - setting tuner_lock_debug=1 on the si2157 module
 - enabling some dynamic debug, e.g.
   echo 'file  si2157.c +p' >/sys/kernel/debug/dynamic_debug/control
 - running w_scan under strace
 - review the kernel commit logs for the cx23885 and si2157 modules

So far I am not seeing any helpful error messages or commits that
could be causing the problem.
Are there some test programs I could run to debug this further?

Any help would be welcome

Re: regression(?) - hauppage quad card tuning

[media]

On Fri, Dec 29, 2023 at 07:44:06PM +1100, media@arewethere.net wrote:
>
> I also tried a 'backports' kernel, 6.5.10-1~bpo12+1.
> This is based on the upstream version 6.1.66-1.
> This kernel also fails to tune and get a lock.

oops, paste error there: this is based on upstream version 6.5.10-1

Any suggestions?

Re: regression(?) - hauppage quad card tuning

[media]

To close this out, in case anyone hits it.

The machine I was working on was remote and I was not able to do try
the ancient solution of turning the box off and turning it off again,
i.e. a proper cold boot. When I was able to try this, it worked.

However I observed an interesting behaviour in the process.
From a warm boot I get this sequence in dmesg

[ 21.283374] si2157 2-0060: Silicon Labs Si2157 successfully attached
[ 21.283374] si2157 2-0060: Silicon Labs Si2157 successfully attached
[ 21.295307] si2157 2-0062: Silicon Labs Si2157 successfully attached
[ 21.295307] si2157 2-0062: Silicon Labs Si2157 successfully attached
[ 22.392453] si2157 7-0060: Silicon Labs Si2157 successfully attached
[ 22.392453] si2157 7-0060: Silicon Labs Si2157 successfully attached
[ 22.405506] si2157 7-0062: Silicon Labs Si2157 successfully attached
[ 22.405506] si2157 7-0062: Silicon Labs Si2157 successfully attached


from a cold boot I get this

[ 21.780082] si2157 2-0060: Silicon Labs Si2157 successfully attached
[ 21.780082] si2157 2-0060: Silicon Labs Si2157 successfully attached
[ 21.791839] si2157 2-0062: Silicon Labs Si2157 successfully attached
[ 21.791839] si2157 2-0062: Silicon Labs Si2157 successfully attached
[ 22.888459] si2157 7-0060: Silicon Labs Si2157 successfully attached
[ 22.888459] si2157 7-0060: Silicon Labs Si2157 successfully attached
[ 22.901286] si2157 7-0062: Silicon Labs Si2157 successfully attached
[ 22.901286] si2157 7-0062: Silicon Labs Si2157 successfully attached
[ 44.152638] si2157 2-0060: found a 'Silicon Labs Si2157-A30 ROM 0x50'
[ 44.152638] si2157 2-0060: found a 'Silicon Labs Si2157-A30 ROM 0x50'
[ 44.152672] si2157 2-0060: firmware: failed to load dvb_driver_si2157_rom50.fw (-2)
[ 44.152842] si2157 2-0060: firmware: failed to load dvb_driver_si2157_rom50.fw (-2)
[ 44.237515] si2157 2-0060: firmware: direct-loading firmware dvb-tuner-si2157-a30-01.fw
[ 44.237662] si2157 2-0060: downloading firmware from file 'dvb-tuner-si2157-a30-01.fw'
[ 44.801622] si2157 7-0060: found a 'Silicon Labs Si2157-A30 ROM 0x50'
[ 44.801622] si2157 7-0060: found a 'Silicon Labs Si2157-A30 ROM 0x50'
[ 44.801684] si2157 7-0060: firmware: failed to load dvb_driver_si2157_rom50.fw (-2)
[ 44.801798] si2157 7-0060: firmware: failed to load dvb_driver_si2157_rom50.fw (-2)
[ 44.802109] si2157 7-0060: downloading firmware from file 'dvb-tuner-si2157-a30-01.fw'
[ 44.805562] si2157 7-0062: found a 'Silicon Labs Si2157-A30 ROM 0x50'
[ 44.805562] si2157 7-0062: found a 'Silicon Labs Si2157-A30 ROM 0x50'
[ 44.805589] si2157 7-0062: firmware: failed to load dvb_driver_si2157_rom50.fw (-2)
[ 44.805697] si2157 7-0062: firmware: failed to load dvb_driver_si2157_rom50.fw (-2)
[ 44.805944] si2157 7-0062: downloading firmware from file 'dvb-tuner-si2157-a30-01.fw'
[ 45.514028] si2157 2-0060: firmware version: 3.0.5
[ 47.444562] si2157 7-0060: firmware version: 3.0.5
[ 47.448329] si2157 7-0062: firmware version: 3.0.5

From this it seems that on a warm boot, we are only calling
si2157_probe(), but this is insufficient for proper operation.

On a cold boot, si2157_find_and_load_firmware() is also called,
which generates the extra output.

Not sure this can be called a bug, but certainly a 'misfeature'.
In any case I don't think there's a regression wrt the 5.10 kernels,
I was probably just lucky with happening to cold-boot the host.