net: check qdisc_pkt_len_segs_init() return value on ingress
Subsystems (2)
PATCH v2 1/1 net: check qdisc_pkt_len_segs_init() return value on ingress
Commit 7fb4c1967011 ("net: pull headers in qdisc_pkt_len_segs_init()") changed qdisc_pkt_len_segs_init() to return an skb drop reason when it detects malicious GSO packets. The egress path in __dev_queue_xmit() checks this return value and drops bad packets, but the ingress path in sch_handle_ingress() ignores it.
This means malformed GSO packets entering via TC ingress are not dropped and could be redirected to another interface or cause incorrect qdisc accounting.
Check the return value and drop the packet when a bad GSO is detected.
---
v1 -> v2: reorder variable declarations for reverse xmas tree
v1: https://lore.kernel.org/netdev/[email protected]/
net/core/dev.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)
net/core/dev.c +10 -2
@@ -4459,8 +4459,8 @@ sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret, struct net_device *orig_dev, bool *another) { struct bpf_mprog_entry *entry = rcu_dereference_bh(skb->dev->tcx_ingress); -enum skb_drop_reason drop_reason =SKB_DROP_REASON_TC_INGRESS ;struct bpf_net_context __bpf_net_ctx, *bpf_net_ctx; +enum skb_drop_reason drop_reason;int sch_ret;if (!entry)
@@ -4472,7 +4472,15 @@ sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret, *pt_prev =NULL ; } -qdisc_pkt_len_segs_init (skb); + drop_reason =qdisc_pkt_len_segs_init (skb); +if (unlikely (drop_reason)) { +kfree_skb_reason (skb, drop_reason); + *ret =NET_RX_DROP ; +bpf_net_ctx_clear (bpf_net_ctx); +return NULL ; + } + + drop_reason =SKB_DROP_REASON_TC_INGRESS ;tcx_set_ingress (skb, true);if (static_branch_unlikely (&tcx_needed_key)) {
Signature · 1 line
2.53.0
Quoted 9 lines · David Carlier
Commit
7fb4c1967011("net: pull headers inqdisc_pkt_len_segs_init()") changedqdisc_pkt_len_segs_init()to return an skb drop reason when it detects malicious GSO packets. The egress path in__dev_queue_xmit()checks this return value and drops bad packets, but the ingress path insch_handle_ingress()ignores it.This means malformed GSO packets entering via TC ingress are not dropped and could be redirected to another interface or cause incorrect qdisc accounting.
Why we need to do this on both sides (and what's the perf impact)? If TC ingress redirects it to some other device, then don't we hit the same via __dev_queue_xmit() where the 7fb4c1967011 added the qdisc_pkt_len_segs_init()?
Quoted 42 lines
Check the return value and drop the packet when a bad GSO is detected.
Fixes 7fb4c1967011 ("net: pull headers in qdisc_pkt_len_segs_init()")Signed-off-by David Carlier <[email protected]>---
v1 -> v2: reorder variable declarations for reverse xmas tree
v1: https://lore.kernel.org/netdev/[email protected]/net/core/dev.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)net/core/dev.c +10
-2@@ -4459,8 +4459,8 @@ sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret, struct net_device *orig_dev, bool *another) { struct bpf_mprog_entry *entry = rcu_dereference_bh(skb->dev->tcx_ingress); -enum skb_drop_reason drop_reason =SKB_DROP_REASON_TC_INGRESS ;struct bpf_net_context __bpf_net_ctx, *bpf_net_ctx; +enum skb_drop_reason drop_reason;int sch_ret;if (!entry)@@ -4472,7 +4472,15 @@ sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret, *pt_prev =NULL ; } -qdisc_pkt_len_segs_init (skb); + drop_reason =qdisc_pkt_len_segs_init (skb); +if (unlikely (drop_reason)) { +kfree_skb_reason (skb, drop_reason); + *ret =NET_RX_DROP ; +bpf_net_ctx_clear (bpf_net_ctx); +return NULL ; + } + + drop_reason =SKB_DROP_REASON_TC_INGRESS ;tcx_set_ingress (skb, true);if (static_branch_unlikely (&tcx_needed_key)) {
Quoted 13 lines · David Carlier
Commit
7fb4c1967011("net: pull headers inqdisc_pkt_len_segs_init()") changedqdisc_pkt_len_segs_init()to return an skb drop reason when it detects malicious GSO packets. The egress path in__dev_queue_xmit()checks this return value and drops bad packets, but the ingress path insch_handle_ingress()ignores it.This means malformed GSO packets entering via TC ingress are not dropped and could be redirected to another interface or cause incorrect qdisc accounting.
Check the return value and drop the packet when a bad GSO is detected.
Quoted 8 lines
Fixes 7fb4c1967011 ("net: pull headers in qdisc_pkt_len_segs_init()")Signed-off-by David Carlier <[email protected]>---
v1 -> v2: reorder variable declarations for reverse xmas tree
v1: https://lore.kernel.org/netdev/[email protected]/net/core/dev.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)
NACK. This is is not needed.
I will not even bother, this is an obvious LLM based patch.
Hi Eric,
You're right on both counts. An LLM was used for the initial
bug-finding /
pre-analysis on this one, and clearly it wasn't deep enough.
Daniel's
question made me go back and trace the redirect paths properly — the premise was wrong, __dev_queue_xmit() already catches them. I should
have
re-read your 7fb4c1967011 commit message before sending.
Dropping it. Sorry for the noise.
Cheers,
Quoted 27 lines · Eric Dumazet
Quoted 13 lines · David Carlier
Commit
7fb4c1967011("net: pull headers inqdisc_pkt_len_segs_init()") changedqdisc_pkt_len_segs_init()to return an skb drop reason when it detects malicious GSO packets. The egress path in__dev_queue_xmit()checks this return value and drops bad packets, but the ingress path insch_handle_ingress()ignores it.This means malformed GSO packets entering via TC ingress are not dropped and could be redirected to another interface or cause incorrect qdisc accounting.
Check the return value and drop the packet when a bad GSO is detected.
Quoted 8 lines
Fixes 7fb4c1967011 ("net: pull headers in qdisc_pkt_len_segs_init()")Signed-off-by David Carlier <[email protected]>---
v1 -> v2: reorder variable declarations for reverse xmas tree
v1: https://lore.kernel.org/netdev/[email protected]/net/core/dev.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)NACK. This is is not needed.
I will not even bother, this is an obvious LLM based patch.