Re: [alsa-devel] [PATCH v2 09/11] ASoC: Intel: hdac_hdmi: Set ops to NULL on remove

[Ranjani Sridharan]

On Wed, 2019-06-19 at 10:38 +0200, Amadeusz SÅawiÅski wrote:
> On Tue, 18 Jun 2019 08:58:22 -0700
> Ranjani Sridharan <ranjani.sridharan@xxxxxxxxxxxxxxx> wrote:
> 
> > On Tue, 2019-06-18 at 13:00 +0200, Amadeusz SÅawiÅski wrote:
> > > On Mon, 17 Jun 2019 13:51:42 -0700
> > > Ranjani Sridharan <ranjani.sridharan@xxxxxxxxxxxxxxx> wrote:
> > >   
> > > > On Mon, 2019-06-17 at 13:36 +0200, Amadeusz SÅawiÅski wrote:  
> > > > > When we unload Skylake driver we may end up calling
> > > > > hdac_component_master_unbind(), it uses acomp->audio_ops,
> > > > > which
> > > > > we
> > > > > set
> > > > > in hdmi_codec_probe(), so we need to set it to NULL in
> > > > > hdmi_codec_remove(),
> > > > > otherwise we will dereference no longer existing pointer.    
> > > > 
> > > > Hi Amadeusz,
> > > > 
> > > > It looks like the audio_ops should be deleted
> > > > snd_hdac_acomp_exit().
> > > > Also, this doesnt seem to be the case with when the SOF driver
> > > > is
> > > > removed.
> > > > Could you please give a bit more context on what error you see
> > > > when this happens?  
> > > 
> > > Hi,
> > > 
> > > I get Oops. This is what happens with all other patches in this
> > > series and only this one reverted:
> > > 
> > > root@APL:~# rmmod snd_soc_sst_bxt_rt298
> > > root@APL:~# rmmod snd_soc_hdac_hdmi
> > > root@APL:~# rmmod snd_soc_skl  
> > 
> > Thanks, Amadeusz. I think the order in which the drivers are
> > removed
> > is what's causing the oops in your case. With SOF, the order we
> > remove is
> > 
> > 1. rmmod sof_pci_dev
> > 2. rmmod snd_soc_sst_bxt_rt298
> > 3. rmmod snd_soc_hdac_hdmi
> > 
> 
> Well, there is nothing enforcing the order in which modules can be
> unloaded (and I see no reason to force it), as you can see from
> following excerpt, you can either start unloading from
> snd_soc_sst_bxt_rt298 or snd_soc_skl, and yes if you start from
> snd_soc_skl, there is no problem.
> 
I am good with this patch. I just wanted to understand why we werent
seeing this error with SOF. Sure, there's nothing enforcing the order
in which modules are unloaded  but there must be a logical order for
testing purposes. 

Pierre, can you please comment on it. I vaguely remember discussing
this with you last year.

Thanks,
Ranjani