Subject: [PATCH] net: gro: fix double aggregation of flush-marked skbs
Subsystems (4)
PATCH v3 1/1 Subject: [PATCH] net: gro: fix double aggregation of flush-marked skbs
The new skb_gro_receive_list() function is missing a critical safety check present in the legacy skb_gro_receive() path. Specifically, it does not validate NAPI_GRO_CB(skb)->flush before allowing packet aggregation.
This allows already-GRO'd packets with existing frag_list to be re-aggregated into a new GRO session, corrupting the frag_list chain structure. When skb_segment() attempts to unpack these malformed packets, it encounters invalid state and triggers a kernel panic.
Scenario (Tethering/Device forwarding):
1. Driver: Generated aggregated packet P1 via LRO with frag_list
2. Dev A: Receives aggregated fraglist packet and flush flag set
3. Dev A: Re-enters GRO, skb_gro_receive_list() is called
4. Missing flush check allows re-aggregation despite flush flag
5. Frag_list chain becomes corrupted (loops or dangling refs)
6. Dev B: TX path calls skb_segment(), crashes on corrupted frag_list
Root cause in skb_segment():
The check at line ~4891:
if (hsize <= 0 && i >= nfrags && skb_headlen(list_skb) &&
(skb_headlen(list_skb) == len || sg)) { When frag_list is corrupted by double aggregation, when list_skb is a NULL pointer from skb->next, skb_headlen(list_skb) dereference NULL/corrupted pointers occurs.
Call Trace: skb_headlen(NULL skb) skb_segment tcp_gso_segment tcp4_gso_segment inet_gso_segment skb_mac_gso_segment __skb_gso_segment skb_gso_segment validate_xmit_skb validate_xmit_skb_list sch_direct_xmit qdisc_restart __qdisc_run qdisc_run net_tx_action
Fix: Add NAPI_GRO_CB(skb)->flush validation to the early-return check in skb_gro_receive_list(), matching the defensive programming pattern of skb_gro_receive().
--- net/core/gro.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
net/core/gro.c +2 -1
@@ -229,7 +229,8 @@ int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb)int skb_gro_receive_list (struct sk_buff *p,struct sk_buff *skb) { -if (unlikely (p->len + skb->len >=65536 )) +if (unlikely (p->len + skb->len >=65536 || +NAPI_GRO_CB (skb)->flush))return -E2BIG ;if (!pskb_may_pull (skb,skb_gro_offset (skb))) {
Signature · 1 line
2.45.2
Paolo Abeni
maintainerNote: the patch subject is quite uncorrected
Quoted 3 lines · Shiming Cheng
The new
skb_gro_receive_list()function is missing a critical safety check present in the legacyskb_gro_receive()path. Specifically, it does not validate NAPI_GRO_CB(skb)->flush before allowing packet aggregation.
skb_gro_receive_list() is not very "new" and definitely skb_gro_receive() is not legacy.
Quoted 5 lines
This allows already-GRO'd packets with existing
frag_listto be re-aggregated into a new GRO session, corrupting thefrag_listchain structure. Whenskb_segment()attempts to unpack these malformed packets, it encounters invalid state and triggers a kernel panic.
Quoted 7 lines
Scenario (Tethering/Device forwarding):
1. Driver: Generated aggregated packet P1 via LRO withfrag_list
2. Dev A: Receives aggregated fraglist packet and flush flag set
3. Dev A: Re-enters GRO,skb_gro_receive_list()is called
4. Missing flush check allows re-aggregation despite flush flag
5.Frag_listchain becomes corrupted (loops or dangling refs)
6. Dev B: TX path callsskb_segment(), crashes on corruptedfrag_list
I can't parse the above. Is this something that can happen with in-tree drivers or do you need OoT module to trigger it? In any case please clarify the actual order and the involved driver. Possibly a stack strace leading to the critical aggregation could help.
Quoted 5 lines
Fix: Add NAPI_GRO_CB(skb)->flush validation to the early-return check in
skb_gro_receive_list(), matching the defensive programming pattern ofskb_gro_receive().Fixes 8928756d53d5 ("net: add fraglist GRO/GSO support")
The fix tag is wrong, should be:
/P
Quoted 10 lines · Paolo Abeni
Note: the patch subject is quite uncorrected
Quoted 6 lines · Shiming Cheng
The new
skb_gro_receive_list()function is missing a critical safety check
present in the legacyskb_gro_receive()path. Specifically, it does not
validate NAPI_GRO_CB(skb)->flush before allowing packet aggregation.
Quoted 3 lines
skb_gro_receive_list()is not very "new" and definitelyskb_gro_receive()is not legacy.
The wording here may need to be adjusted. I'm referring to the chronological order/which one came first.
Updated:
The skb_gro_receive_list() function is missing a critical safety check that exists in the skb_gro_receive() implementation. Specifically, it does not validate NAPI_GRO_CB(skb)->flush before allowing packet aggregation
Quoted 16 lines
Quoted 7 lines
This allows already-GRO'd packets with existing
frag_listto be re-aggregated into a new GRO session, corrupting thefrag_listchain
structure. Whenskb_segment()attempts to unpack these malformed packets,
it encounters invalid state and triggers a kernel panic.Quoted 8 lines
Scenario (Tethering/Device forwarding):
1. Driver: Generated aggregated packet P1 via LRO withfrag_list
2. Dev A: Receives aggregated fraglist packet and flush flag set
3. Dev A: Re-enters GRO,skb_gro_receive_list()is called
4. Missing flush check allows re-aggregation despite flush flag
5.Frag_listchain becomes corrupted (loops or dangling refs)
6. Dev B: TX path callsskb_segment(), crashes on corruptedfrag_list
Quoted 6 lines
I can't parse the above. Is this something that can happen with in- tree
drivers or do you need OoT module to trigger it? In any case please clarify the actual order and the involved driver. Possibly a stack strace leading to the critical aggregation could help.
We are hitting a GRO/LRO-related failure in a tethering scenario.
On the RX path, the driver performs an LRO-style aggregation before handing packets to the stack. When nfrags exceeds 17, additional packets are no longer appended to the frags array, but are attached through skb_shared_info(skb)->frag_list. After that, the driver still passes the skb into napi_gro_receive(), so the same traffic goes through a second aggregation stage in GRO.
In our tethering case, NAPI_GRO_CB(skb)->is_flist = !sk, so is_flist becomes true, and the skb follows the SKB_GSO_FRAGLIST path, eventually reaching skb_gro_receive_list(). The issue is that some later skbs may already carry their own frag_list as a result of the first aggregation done by the driver. When GRO links those skbs again into a new frag_list chain, the resulting skb layout becomes more complex than expected and eventually triggers the kernel exception.
Actual skb relationships when the issue occurs is as follows. A->frag_list = BB->next = CC->frag_list = D
In the observed layout, A already links B -> C through frag_list, while C itself still carries its own frag_list -> D. In other words, when GRO continues chaining skbs in skb_gro_receive_list(), the later skb is no longer a simple standalone packet, but an skb that already carries shared_info->frag_list from the driver-side LRO stage. This creates a nested frag_list layout and eventually triggers the kernel exception in our case.
Quoted 6 lines
Quoted 6 lines
Fix: Add NAPI_GRO_CB(skb)->flush validation to the early-return check in
skb_gro_receive_list(), matching the defensive programming pattern ofskb_gro_receive().
Quoted 6 lines
Quoted 1 line
Fixes 8928756d53d5 ("net: add fraglist GRO/GSO support")The fix tag is wrong, should be:
Fixes 3a1296a38d0c ('net: Support GRO/GSO fraglist chaining.')
I will update it in the next patch.
Quoted 2 lines
/P