RE: [PATCH][next] rtw89: Fix potential dereference of the null pointer sta

[Pkshih]

> -----Original Message-----
> From: kvalo=codeaurora.org@xxxxxxxxxxxxxxxxx <kvalo=codeaurora.org@xxxxxxxxxxxxxxxxx> On
> Behalf Of Kalle Valo
> Sent: Monday, October 18, 2021 8:12 PM
> To: Pkshih <pkshih@xxxxxxxxxxx>
> Cc: Colin King <colin.king@xxxxxxxxxxxxx>; David S . Miller <davem@xxxxxxxxxxxxx>; Jakub
> Kicinski <kuba@xxxxxxxxxx>; linux-wireless@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> kernel-janitors@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH][next] rtw89: Fix potential dereference of the null pointer sta
> 
> Pkshih <pkshih@xxxxxxxxxxx> writes:
> 
> >> -----Original Message-----
> >> From: Colin King <colin.king@xxxxxxxxxxxxx>
> >> Sent: Friday, October 15, 2021 11:46 PM
> >> To: Kalle Valo <kvalo@xxxxxxxxxxxxxx>; David S . Miller <davem@xxxxxxxxxxxxx>; Jakub Kicinski
> >> <kuba@xxxxxxxxxx>; Pkshih <pkshih@xxxxxxxxxxx>; linux-wireless@xxxxxxxxxxxxxxx;
> >> netdev@xxxxxxxxxxxxxxx
> >> Cc: kernel-janitors@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> >> Subject: [PATCH][next] rtw89: Fix potential dereference of the null pointer sta
> >>
> >> From: Colin Ian King <colin.king@xxxxxxxxxxxxx>
> >>
> >> The pointer rtwsta is dereferencing pointer sta before sta is
> >> being null checked, so there is a potential null pointer deference
> >> issue that may occur. Fix this by only assigning rtwsta after sta
> >> has been null checked. Add in a null pointer check on rtwsta before
> >> dereferencing it too.
> >>
> >> Fixes: e3ec7017f6a2 ("rtw89: add Realtek 802.11ax driver")
> >> Addresses-Coverity: ("Dereference before null check")
> >> Signed-off-by: Colin Ian King <colin.king@xxxxxxxxxxxxx>
> >> ---
> >>  drivers/net/wireless/realtek/rtw89/core.c | 9 +++++++--
> >>  1 file changed, 7 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/net/wireless/realtek/rtw89/core.c
> >> b/drivers/net/wireless/realtek/rtw89/core.c
> >> index 06fb6e5b1b37..26f52a25f545 100644
> >> --- a/drivers/net/wireless/realtek/rtw89/core.c
> >> +++ b/drivers/net/wireless/realtek/rtw89/core.c
> >> @@ -1534,9 +1534,14 @@ static bool rtw89_core_txq_agg_wait(struct rtw89_dev *rtwdev,
> >>  {
> >>  	struct rtw89_txq *rtwtxq = (struct rtw89_txq *)txq->drv_priv;
> >>  	struct ieee80211_sta *sta = txq->sta;
> >> -	struct rtw89_sta *rtwsta = (struct rtw89_sta *)sta->drv_priv;
> >
> > 'sta->drv_priv' is only a pointer, we don't really dereference the
> > data right here, so I think this is safe. More, compiler can optimize
> > this instruction that reorder it to the place just right before using.
> > So, it seems like a false alarm.
> >
> >> +	struct rtw89_sta *rtwsta;
> >>
> >> -	if (!sta || rtwsta->max_agg_wait <= 0)
> >> +	if (!sta)
> >> +		return false;
> >> +	rtwsta = (struct rtw89_sta *)sta->drv_priv;
> >> +	if (!rtwsta)
> >> +		return false;
> >> +	if (rtwsta->max_agg_wait <= 0)
> >>  		return false;
> >>
> >>  	if (rtwdev->stats.tx_tfc_lv <= RTW89_TFC_MID)
> >
> > I check the size of object files before/after this patch, and
> > the original one is smaller.
> >
> >    text    data     bss     dec     hex filename
> >   16781    3392       1   20174    4ece core-0.o  // original
> >   16819    3392       1   20212    4ef4 core-1.o  // after this patch
> >
> > Do you think it is worth to apply this patch?
> 
> I think that we should apply the patch. Even though the compiler _may_
> reorder the code, it might choose not to do that.

Understand.

I have another way to fix this coverity warning, like:

@@ -1617,7 +1617,7 @@ static bool rtw89_core_txq_agg_wait(struct rtw89_dev *rtwdev,
 {
        struct rtw89_txq *rtwtxq = (struct rtw89_txq *)txq->drv_priv;
        struct ieee80211_sta *sta = txq->sta;
-       struct rtw89_sta *rtwsta = (struct rtw89_sta *)sta->drv_priv;
+       struct rtw89_sta *rtwsta = sta ? (struct rtw89_sta *)sta->drv_priv : NULL;

        if (!sta || rtwsta->max_agg_wait <= 0)
                return false;

Is this acceptable?
It has a little redundant checking of 'sta', but the code looks clean.

> 
> Another question is that can txq->sta really be null? I didn't check the
> code, but if it should be always set when the null check is not needed.
> 

It says

* struct ieee80211_txq - Software intermediate tx queue
* @sta: station table entry, %NULL for per-vif queue

So, we need to check if 'sta' is NULL.

--
Ping-Ke