| Message ID | 20240328081648.13193-1-amishin@t-argos.ru (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | octeontx2-af: Add array index check | expand |
Hi, > In rvu_map_cgx_lmac_pf() the 'iter', which is used as an array index, can > reach value (up to 14) that exceed the size (MAX_LMAC_COUNT = 8) of the > array. > Fix this bug by adding 'iter' value check. > > Found by Linux Verification Center (linuxtesting.org) with SVACE. > Since it is a fix, please add "net" to the subject. Thanks, Hariprasad k > Fixes: 91c6945ea1f9 ("octeontx2-af: cn10k: Add RPM MAC support") > Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru> > --- > drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c > b/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c > index 72e060cf6b61..e9bf9231b018 100644 > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c > @@ -160,6 +160,8 @@ static int rvu_map_cgx_lmac_pf(struct rvu *rvu) > continue; > lmac_bmap = cgx_get_lmac_bmap(rvu_cgx_pdata(cgx, rvu)); > for_each_set_bit(iter, &lmac_bmap, rvu->hw->lmac_per_cgx) > { > + if (iter >= MAX_LMAC_COUNT) > + continue; > lmac = cgx_get_lmacid(rvu_cgx_pdata(cgx, rvu), > iter); > rvu->pf2cgxlmac_map[pf] = cgxlmac_id_to_bmap(cgx, > lmac); > -- > 2.30.2
On Thu, 28 Mar 2024 11:16:48 +0300 Aleksandr Mishin wrote: > In rvu_map_cgx_lmac_pf() the 'iter', which is used as an array index, can reach > value (up to 14) that exceed the size (MAX_LMAC_COUNT = 8) of the array. > Fix this bug by adding 'iter' value check. I'm guessing you got the 14 from: hw->lmac_per_cgx = (nix_const >> 8) & 0xFULL; Seems more reasonable to cap the size at that point than every use afterwards.
On Thu, Mar 28, 2024 at 11:28:18AM -0700, Jakub Kicinski wrote: > On Thu, 28 Mar 2024 11:16:48 +0300 Aleksandr Mishin wrote: > > In rvu_map_cgx_lmac_pf() the 'iter', which is used as an array index, can reach > > value (up to 14) that exceed the size (MAX_LMAC_COUNT = 8) of the array. > > Fix this bug by adding 'iter' value check. > > I'm guessing you got the 14 from: > > hw->lmac_per_cgx = (nix_const >> 8) & 0xFULL; > > Seems more reasonable to cap the size at that point than every use > afterwards. FWIIW, I didn't find any other locations where this overflow might occur, but I do agree that this approach makes sense. I also notice that rvu_map_cgx_lmac_pf() has the following check near the top. I think the lmac_per_cgx portion can be dropped, though that could be a follow-up. if (cgx_cnt_max > 0xF || rvu->hw->lmac_per_cgx > 0xF) return -EINVAL;
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c index 72e060cf6b61..e9bf9231b018 100644 --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c @@ -160,6 +160,8 @@ static int rvu_map_cgx_lmac_pf(struct rvu *rvu) continue; lmac_bmap = cgx_get_lmac_bmap(rvu_cgx_pdata(cgx, rvu)); for_each_set_bit(iter, &lmac_bmap, rvu->hw->lmac_per_cgx) { + if (iter >= MAX_LMAC_COUNT) + continue; lmac = cgx_get_lmacid(rvu_cgx_pdata(cgx, rvu), iter); rvu->pf2cgxlmac_map[pf] = cgxlmac_id_to_bmap(cgx, lmac);
In rvu_map_cgx_lmac_pf() the 'iter', which is used as an array index, can reach value (up to 14) that exceed the size (MAX_LMAC_COUNT = 8) of the array. Fix this bug by adding 'iter' value check. Found by Linux Verification Center (linuxtesting.org) with SVACE. Fixes: 91c6945ea1f9 ("octeontx2-af: cn10k: Add RPM MAC support") Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru> --- drivers/net/ethernet/marvell/octeontx2/af/rvu_cgx.c | 2 ++ 1 file changed, 2 insertions(+)