Re: [PATCH] ALSA: usb-audio: Initialize status1 to fix uninitialized symbol errors

[Takashi Iwai <tiwai@suse.de>]

On Wed, 03 Dec 2025 12:22:27 +0100,
HariKrishna Sagala wrote:
> 
> On Wed, 3 Dec 2025, Takashi Iwai wrote:
> 
> > On Wed, 03 Dec 2025 12:01:35 +0100,
> > HariKrishna Sagala wrote:
> > >
> > > On Wed, 3 Dec 2025, Takashi Iwai wrote:
> > >
> > > > On Wed, 03 Dec 2025 09:33:20 +0100,
> > > > hariconscious@gmail.com wrote:
> > > > >
> > > > > From: HariKrishna Sagala <hariconscious@gmail.com>
> > > > >
> > > > > Initialize 'status1' with a default value to resolve the static analysis
> > > > > smatch reported error "uninitialized symbol 'status1'".
> > > > > The 'status1' variable is used to create a buff using "kmemdup".
> > > > > So, ensure to initialize the value before it is read.
> > > > >
> > > > > Signed-off-by: HariKrishna Sagala <hariconscious@gmail.com>
> > > > > ---
> > > > > This patch fixes the below smatch reported errors.
> > > > > sound/usb/mixer_quirks.c:2462 snd_rme_rate_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2467 snd_rme_rate_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2472 snd_rme_rate_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2495 snd_rme_sync_state_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2501 snd_rme_sync_state_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2522 snd_rme_spdif_if_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2535 snd_rme_spdif_format_get() error: uninitialized symbol 'status1'.
> > > > > sound/usb/mixer_quirks.c:2548 snd_rme_sync_source_get() error: uninitialized symbol 'status1'.
> > > > >
> > > > > The below is the flow of 'status1' it is used before initialization.
> > > > >
> > > > > snd_rme_rate_get -> status1 is uninitialized and passed
> > > > > 	snd_rme_get_status1 -> passed as is
> > > > > 		snd_rme_read_value -> passed as is
> > > > > 			snd_usb_ctl_msg -> created buf from status1 using kmemdup
> > > > > 				usb_control_msg -> sent buf for reading/writing
> > > > >
> > > > > Description of "usb_control_msg", states as
> > > > > " * @data: pointer to the data to send"
> > > > >
> > > > > Later from Usb control request, dst buf is copied to src buf but usb
> > > > > control msg request is made before initialization.
> > > > >
> > > > > Thank you.
> > > > >
> > > > >  sound/usb/mixer_quirks.c | 10 +++++-----
> > > > >  1 file changed, 5 insertions(+), 5 deletions(-)
> > > > >
> > > > > diff --git a/sound/usb/mixer_quirks.c b/sound/usb/mixer_quirks.c
> > > > > index 828af3095b86..06903c5de087 100644
> > > > > --- a/sound/usb/mixer_quirks.c
> > > > > +++ b/sound/usb/mixer_quirks.c
> > > > > @@ -2449,7 +2449,7 @@ static int snd_rme_get_status1(struct snd_kcontrol *kcontrol,
> > > > >  static int snd_rme_rate_get(struct snd_kcontrol *kcontrol,
> > > > >  			    struct snd_ctl_elem_value *ucontrol)
> > > > >  {
> > > > > -	u32 status1;
> > > > > +	u32 status1 = 0;
> > > > >  	u32 rate = 0;
> > > > >  	int idx;
> > > > >  	int err;
> > > > > @@ -2483,7 +2483,7 @@ static int snd_rme_rate_get(struct snd_kcontrol *kcontrol,
> > > > >  static int snd_rme_sync_state_get(struct snd_kcontrol *kcontrol,
> > > > >  				  struct snd_ctl_elem_value *ucontrol)
> > > > >  {
> > > > > -	u32 status1;
> > > > > +	u32 status1 = 0;
> > > > >  	int idx = SND_RME_CLOCK_NOLOCK;
> > > > >  	int err;
> > > > >
> > > > > @@ -2513,7 +2513,7 @@ static int snd_rme_sync_state_get(struct snd_kcontrol *kcontrol,
> > > > >  static int snd_rme_spdif_if_get(struct snd_kcontrol *kcontrol,
> > > > >  				struct snd_ctl_elem_value *ucontrol)
> > > > >  {
> > > > > -	u32 status1;
> > > > > +	u32 status1 = 0;
> > > > >  	int err;
> > > > >
> > > > >  	err = snd_rme_get_status1(kcontrol, &status1);
> > > > > @@ -2526,7 +2526,7 @@ static int snd_rme_spdif_if_get(struct snd_kcontrol *kcontrol,
> > > > >  static int snd_rme_spdif_format_get(struct snd_kcontrol *kcontrol,
> > > > >  				    struct snd_ctl_elem_value *ucontrol)
> > > > >  {
> > > > > -	u32 status1;
> > > > > +	u32 status1 = 0;
> > > > >  	int err;
> > > > >
> > > > >  	err = snd_rme_get_status1(kcontrol, &status1);
> > > > > @@ -2539,7 +2539,7 @@ static int snd_rme_spdif_format_get(struct snd_kcontrol *kcontrol,
> > > > >  static int snd_rme_sync_source_get(struct snd_kcontrol *kcontrol,
> > > > >  				   struct snd_ctl_elem_value *ucontrol)
> > > > >  {
> > > > > -	u32 status1;
> > > > > +	u32 status1 = 0;
> > > > >  	int err;
> > > > >
> > > > >  	err = snd_rme_get_status1(kcontrol, &status1);
> > > > >
> > > >
> > > > The warning itself is rather dubious.  But it'd certainly give a safer
> > > > feeling to cover the uninitialized variables, so it would still make
> > > > some sense.
> > > >
> > > > But, the code change can be improved.  e.g. initialize the value in
> > > > the callee side, instead of callers; then it'll reduce all changes to
> > > > a one-liner.
> > > >
> > > > At the next time, please look at the patterns you changed more closely
> > > > and think again whether it's the best change or not before submission.
> > > > On the second look, often you see things from a different perspective
> > > > :)
> > > >
> > > >
> > > > thanks,
> > > >
> > > > Takashi
> > > >
> > >
> > > Hi Takashi,
> > >
> > > Thank you for the feedback and complement.
> > > Yes, will check the pattern and avoid this going forward.
> > > As all the functions call "snd_rme_get_status1",
> > > initialized here but smatch didn't silence them.
> > >
> > > I prefer initializing a variable in caller side as it owns
> > > and aware of initialized value sent.
> > >
> > > Sorry, if I had misunderstood your suggestion.
> >
> > Well, you seem driven too much by smatch.  Just look at your patch
> > again.
> >
> > You're repeating the same initializations and passing to the same
> > function.  And those are the only callers of snd_rme_get_status1().
> > So, instead of initializing in the caller side, initialize the value
> > in snd_rme_get_status1().  Then the whole your changes will be
> > oneliner like below.
> >
> >
> > Takashi
> >
> > --- a/sound/usb/mixer_quirks.c
> > +++ b/sound/usb/mixer_quirks.c
> > @@ -2548,6 +2548,7 @@ static int snd_rme_get_status1(struct snd_kcontrol *kcontrol,
> >  	CLASS(snd_usb_lock, pm)(chip);
> >  	if (pm.err < 0)
> >  		return pm.err;
> > +	*status1 = 0;
> >  	return snd_rme_read_value(chip, SND_RME_GET_STATUS1, status1);
> >  }
> >
> >
> 
> Hi Takashi,
> 
> Yes, tried that before, smatch didn't silence them.

Then it's an obvious failure of smatch.

Or you might set *status1 = 0 at the beginning of
snd_rme_get_status1() although it makes little sense.

> If it is worth having this change though they are not silenced,
> will raise v2 otherwise will drop this patch.

If the smatch still complains, it's not worth.


Takashi