Bjorn,
Thus spake Bjorn Wielens (Uniacke1@xxxxxxxx):
> Hi Zsolt, Nikolay, and all others working on the agrsm-HDA code:
>
> Further investigation reveals that this kernel (2.6.30) uses alsa 1.0.20.
> It is worth noting that this version once again has symbols for reading
> and writing to the HDA codecs in the code
> (see hda_codec.c in sound/pci/hda of the kernel source, lines like
> EXPORT_SYMBOL_HDA(snd_hda_codec_read); )
> These are the interfaces to the HDA card, which were essentially
> re-implemented in the HDA.c file of the driver.
> (You may recall earlier versions of the driver required kernel/alsa
> recompilation and manually adding the export_symbol lines).
>
>
> I've been going through the code and updating the various HDA.c access
> routines to match the alsa 1.0.20 code, but I still get errors. I will
> hopefully be able to trace the exact line causing the problem sometime
> next weekend. (Debugging is quite time-consuming, since the system must
> be rebooted after each panic- otherwise modprobe hangs).
>
> At this point though, I do have the following suggestions for future
> kernel support, if these are feasible:
> 1. Revert to using the symbols supplied by ALSA-- this will reduce
> breakage with newer kernels and updated ALSA code (hopefully).
IMHO this means changing MODULE_LICENSE("Proprietary") to MODULE_LICENSE("GPL")
so that the kbuild would stop complaining about mixing GPL-only code and non-GPL
symbols. I am not sure if there is a license issue here or not, I am an expert
on licenses. Also, some of the older kernels do not export those symbols, as you
point out above. So, we would force the user to recompile those kernels if we
decide to use the exported symbols.
> 2. Get the code to use the hda_codec.h and serial26.c in the kernel tree
> to reduce the need to update these as frequently.
Does the header hda_codec.h have that kind of visibility? I think it may be
internal to the alsa implementation, so one can not just include
<linux/.../hda_codec.h>, but I am not exactly sure about this.
--
WBR, Nikolay Zhuravlev
>
> I will report more as I find out.
>
> Best,
> Bjorn.
>
>
>
> Nikolay Zhuravlev wrote:
> > Bjorn,
> >
> > Same here with the 2.6.30 rc6. The modules compile Ok but wvdialconf
> > segfaulted and messed up the kernel too.
> > My patched "6mdv" sources are more or less the same as "7mdv".
> > dmesg reports 'null pointer dereference', this is not good...
> >
> > <<<<
> > [ 235.790591] * Hello Agere Driver **
> > [ 236.937771] Loading module Agere Modem Controller driver version 2.1.80 (2008
> > -10-31)
> > [ 251.857186] ttyAGS3 at I/O 0x8004 (irq = 22) is a AgereModem
> > [ 251.857413] Loading module Agere Modem Interface driver version 2.1.80.0 (200
> > 8-10-31)
> > [ 271.833830] ==> codecType = 0x32
> > [ 271.833858] BUG: unable to handle kernel NULL pointer dereference at 0000002b
> >
>
>
