On Wed, Oct 21, 2020 at 11:03:11PM +0800, Hui Wang wrote: > Looks like this will not bring problem on patch_sigmatel.c, NULL and valid > kernel pointer are same for IS_ERR(), they will not make IS_ERR() come true. > Correct, but that is the problem. You can dereference a valid pointer but you can't dereference a NULL. > On 10/21/20 10:21 PM, Hui Wang wrote: > > > > On 10/21/20 8:19 PM, Dan Carpenter wrote: > > > Hello Hui Wang, > > > > > > The patch f4794c6064a8: "ALSA: hda - Don't register a cb func if it > > > is registered already" from Sep 30, 2020, leads to the following > > > static checker warning: > > > > > > sound/pci/hda/patch_sigmatel.c:3075 stac92hd71bxx_fixup_hp_m4() > > > warn: 'jack' can also be NULL > > > > > > sound/pci/hda/patch_sigmatel.c > > > 3069 /* Enable VREF power saving on GPIO1 detect */ > > > 3070 snd_hda_codec_write_cache(codec, codec->core.afg, 0, > > > 3071 AC_VERB_SET_GPIO_UNSOLICITED_RSP_MASK, 0x02); > > > 3072 jack = snd_hda_jack_detect_enable_callback(codec, > > > codec->core.afg, > > > 3073 stac_vref_event); > > > > > > Originally snd_hda_jack_detect_enable_callback() would not return NULL > > > here. > > > > > > 3074 if (!IS_ERR(jack)) > > > 3075 jack->private_data = 0x02; So if "jack" is NULL then it will say it's not an error pointer so it will try to assign jack->private_data = 0x02; and oops. regards, dan carpenter
