diff mbox series

[net-next,1/1] netfilter: flowtable: Make sure dst_cache is valid before using it

Message ID 20210411081334.1994938-1-roid@nvidia.com
State New
Headers show
Series [net-next,1/1] netfilter: flowtable: Make sure dst_cache is valid before using it | expand

Commit Message

Roi Dayan April 11, 2021, 8:13 a.m. UTC
It could be dst_cache was not set so check it's not null before using
it.

Fixes: 8b9229d15877 ("netfilter: flowtable: dst_check() from garbage collector path")
Signed-off-by: Roi Dayan <roid@nvidia.com>
---
 net/netfilter/nf_flow_table_core.c | 2 ++
 1 file changed, 2 insertions(+)

Comments

Roi Dayan April 12, 2021, 8:26 a.m. UTC | #1
On 2021-04-11 1:58 PM, Pablo Neira Ayuso wrote:
> Hi Roi,

> 

> On Sun, Apr 11, 2021 at 11:13:34AM +0300, Roi Dayan wrote:

>> It could be dst_cache was not set so check it's not null before using

>> it.

> 

> Could you give a try to this fix?

> 

> net/sched/act_ct.c leaves the xmit_type as FLOW_OFFLOAD_XMIT_UNSPEC

> since it does not cache a route.

> 

> Thanks.

> 


what do you mean? FLOW_OFFLOAD_XMIT_UNSPEC doesn't exists so default 0
is set.

do you suggest adding that enum option as 0?

this is the current xmit_type enum

enum flow_offload_xmit_type {
        FLOW_OFFLOAD_XMIT_NEIGH         = 0,
        FLOW_OFFLOAD_XMIT_XFRM,
        FLOW_OFFLOAD_XMIT_DIRECT,
};



>> Fixes: 8b9229d15877 ("netfilter: flowtable: dst_check() from garbage collector path")

>> Signed-off-by: Roi Dayan <roid@nvidia.com>

>> ---

>>   net/netfilter/nf_flow_table_core.c | 2 ++

>>   1 file changed, 2 insertions(+)

>>

>> diff --git a/net/netfilter/nf_flow_table_core.c b/net/netfilter/nf_flow_table_core.c

>> index 76573bae6664..e426077aaed1 100644

>> --- a/net/netfilter/nf_flow_table_core.c

>> +++ b/net/netfilter/nf_flow_table_core.c

>> @@ -410,6 +410,8 @@ static bool flow_offload_stale_dst(struct flow_offload_tuple *tuple)

>>   	if (tuple->xmit_type == FLOW_OFFLOAD_XMIT_NEIGH ||

>>   	    tuple->xmit_type == FLOW_OFFLOAD_XMIT_XFRM) {

>>   		dst = tuple->dst_cache;

>> +		if (!dst)

>> +			return false;

>>   		if (!dst_check(dst, tuple->dst_cookie))

>>   			return true;

>>   	}

>> -- 

>> 2.26.2

>>
Pablo Neira Ayuso April 12, 2021, 11:42 a.m. UTC | #2
On Mon, Apr 12, 2021 at 11:26:35AM +0300, Roi Dayan wrote:
> 

> 

> On 2021-04-11 1:58 PM, Pablo Neira Ayuso wrote:

> > Hi Roi,

> > 

> > On Sun, Apr 11, 2021 at 11:13:34AM +0300, Roi Dayan wrote:

> > > It could be dst_cache was not set so check it's not null before using

> > > it.

> > 

> > Could you give a try to this fix?

> > 

> > net/sched/act_ct.c leaves the xmit_type as FLOW_OFFLOAD_XMIT_UNSPEC

> > since it does not cache a route.

> > 

> > Thanks.

> > 

> 

> what do you mean? FLOW_OFFLOAD_XMIT_UNSPEC doesn't exists so default 0

> is set.

> 

> do you suggest adding that enum option as 0?


Yes. This could be FLOW_OFFLOAD_XMIT_TC instead if you prefer.

enum flow_offload_xmit_type {
        FLOW_OFFLOAD_XMIT_TC        = 0,
        FLOW_OFFLOAD_XMIT_NEIGH,
        FLOW_OFFLOAD_XMIT_XFRM,
        FLOW_OFFLOAD_XMIT_DIRECT,
};

so there is no need to check for no route in the
FLOW_OFFLOAD_XMIT_NEIGH case (it's assumed this type always has a
route).
Roi Dayan April 13, 2021, 7:58 a.m. UTC | #3
On 2021-04-12 2:42 PM, Pablo Neira Ayuso wrote:
> On Mon, Apr 12, 2021 at 11:26:35AM +0300, Roi Dayan wrote:

>>

>>

>> On 2021-04-11 1:58 PM, Pablo Neira Ayuso wrote:

>>> Hi Roi,

>>>

>>> On Sun, Apr 11, 2021 at 11:13:34AM +0300, Roi Dayan wrote:

>>>> It could be dst_cache was not set so check it's not null before using

>>>> it.

>>>

>>> Could you give a try to this fix?

>>>

>>> net/sched/act_ct.c leaves the xmit_type as FLOW_OFFLOAD_XMIT_UNSPEC

>>> since it does not cache a route.

>>>

>>> Thanks.

>>>

>>

>> what do you mean? FLOW_OFFLOAD_XMIT_UNSPEC doesn't exists so default 0

>> is set.

>>

>> do you suggest adding that enum option as 0?

> 

> Yes. This could be FLOW_OFFLOAD_XMIT_TC instead if you prefer.

> 

> enum flow_offload_xmit_type {

>          FLOW_OFFLOAD_XMIT_TC        = 0,

>          FLOW_OFFLOAD_XMIT_NEIGH,

>          FLOW_OFFLOAD_XMIT_XFRM,

>          FLOW_OFFLOAD_XMIT_DIRECT,

> };

> 

> so there is no need to check for no route in the

> FLOW_OFFLOAD_XMIT_NEIGH case (it's assumed this type always has a

> route).

> 


thanks Pablo. were not sure I wanted to touch the enum.
I prefer unspec actually as you suggested initially.
it works fine by adding the enum.

i'll submit v2 with this suggestion.
diff mbox series

Patch

diff --git a/net/netfilter/nf_flow_table_core.c b/net/netfilter/nf_flow_table_core.c
index 76573bae6664..e426077aaed1 100644
--- a/net/netfilter/nf_flow_table_core.c
+++ b/net/netfilter/nf_flow_table_core.c
@@ -410,6 +410,8 @@  static bool flow_offload_stale_dst(struct flow_offload_tuple *tuple)
 	if (tuple->xmit_type == FLOW_OFFLOAD_XMIT_NEIGH ||
 	    tuple->xmit_type == FLOW_OFFLOAD_XMIT_XFRM) {
 		dst = tuple->dst_cache;
+		if (!dst)
+			return false;
 		if (!dst_check(dst, tuple->dst_cookie))
 			return true;
 	}