On Mon, Jun 22, 2020 at 05:51:16PM +0300, Kalle Valo wrote:
> Dan Carpenter <dan.carpenter@xxxxxxxxxx> writes:
>
> > There is a success path where "ret" isn't initialized where we never
> > have a ATH11K_SPECTRAL_TAG_SCAN_SEARCH and then ret isn't initialized.
> >
> > Fixes: 9d11b7bff950 ("ath11k: add support for spectral scan")
> > Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > ---
> > drivers/net/wireless/ath/ath11k/spectral.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/wireless/ath/ath11k/spectral.c b/drivers/net/wireless/ath/ath11k/spectral.c
> > index 1c5d65bb411f..bfbf905f7507 100644
> > --- a/drivers/net/wireless/ath/ath11k/spectral.c
> > +++ b/drivers/net/wireless/ath/ath11k/spectral.c
> > @@ -677,7 +677,7 @@ static int ath11k_spectral_process_data(struct ath11k *ar,
> > u32 data_len, i;
> > u8 sign, tag;
> > int tlv_len, sample_sz;
> > - int ret;
> > + int ret = 0;
> > bool quit = false;
>
> I try to avoid initialising ret variables so I would like find another
> way. What about doing this (completely untested!) in the end of the
> function:
>
> return 0;
>
> err:
> kfree(fft_sample);
> unlock:
> spin_unlock_bh(&ar->spectral.lock);
> return ret;
I normally avoid it as well... If I were to redo this patch, I would
probably do:
ret = 0;
err:
kfree(fft_sample);
unlock:
spin_unlock_bh(&ar->spectral.lock);
return ret;
Would that be better?
regards,
dan carpenter
