Alessandro Zanni <alessandro.zanni87@xxxxxxxxx> wrote: > In error handling code for "ieee80211_gtk_rekey_add failed", release > the mutex before to return. > > Found with Coccinelle static analisys tool, > script: https://coccinelle.gitlabpages.inria.fr/website/rules/mut.cocci > > Signed-off-by: Alessandro Zanni <alessandro.zanni87@xxxxxxxxx> > --- > drivers/net/wireless/realtek/rtw89/wow.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/net/wireless/realtek/rtw89/wow.c b/drivers/net/wireless/realtek/rtw89/wow.c > index 86e24e07780d..8045acb27cf9 100644 > --- a/drivers/net/wireless/realtek/rtw89/wow.c > +++ b/drivers/net/wireless/realtek/rtw89/wow.c > @@ -624,6 +624,7 @@ static struct ieee80211_key_conf *rtw89_wow_gtk_rekey(struct rtw89_dev *rtwdev, > kfree(rekey_conf); > if (IS_ERR(key)) { > rtw89_err(rtwdev, "ieee80211_gtk_rekey_add failed\n"); > + mutex_unlock(&rtwdev->mutex); > return NULL; > } The driver mutex is held across this function. Only unlock before calling ieee80211_gtk_rekey_add() and lock after that, because the function enters driver again and lock mutex again causing deadlock. Thus this Coccinelle finding is false alarm. NACK this patch.
