On 17 Mar 2010, at 23:09, Guennadi Liakhovetski <g.liakhovetski@xxxxxx> wrote: > On Wed, 17 Mar 2010, Mark Brown wrote: > >> On 17 Mar 2010, at 21:06, Guennadi Liakhovetski <g.liakhovetski@xxxxxx >> > wrote: >> >>> If instantiation of a card failed, we have to remove it from the >>> card >>> list. This fixes an Oops on Migo-R, triggering, when after a failed >>> firmware load attempt the driver modules are removed and re-inserted >>> again. >> >> This looks like the wrong fix - if the card failed to instantiate >> because the >> component drivers weren't ready then it ought to hang around so it >> can be >> instantiated once they appear. Presumably some part of the >> instantiation >> attempt is failing and not being unwound properly? > > If snd_soc_instantiate_card() fails, when called from > snd_soc_register_card(), the card currently stays on the list, but the > instantiated flag is not set. If you then unload drivers and > unregister > the soc-audio platform device, snd_soc_unregister_card() is called, > but > since instantiated is not set, nothing is done and the card stays on > the > list. What's the proper unwind way in such a case then? The card should be removed on unregistration either way. > > Thanks > Guennadi > >> >>> Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@xxxxxx> >>> --- >>> >>> Should also be good for stable. >>> >>> diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c >>> index c8b0556..307eea0 100644 >>> --- a/sound/soc/soc-core.c >>> +++ b/sound/soc/soc-core.c >>> @@ -2361,6 +2361,8 @@ static int snd_soc_register_card(struct >>> snd_soc_card >>> *card) >>> mutex_lock(&client_mutex); >>> list_add(&card->list, &card_list); >>> snd_soc_instantiate_cards(); >>> + if (!card->instantiated) >>> + list_del(&card->list); >>> mutex_unlock(&client_mutex); >>> >>> dev_dbg(card->dev, "Registered card '%s'\n", card->name); > > --- > Guennadi Liakhovetski, Ph.D. > Freelance Open-Source Software Developer > http://www.open-technology.de/ _______________________________________________ Alsa-devel mailing list Alsa-devel@xxxxxxxxxxxxxxxx http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
