net: validate untrusted gso packets without csum offload
Subsystems (4)
PATCH 1/1 net: validate untrusted gso packets without csum offload
From: Willem de Bruijn <[email protected]>
Syzkaller again found a path to a kernel crash through bad gso input. By building an excessively large packet to cause an skb field to wrap.
If VIRTIO_NET_HDR_F_NEEDS_CSUM was set this would have been dropped in skb_partial_csum_set.
GSO packets that do not set checksum offload are suspicious and rare. Most callers of virtio_net_hdr_to_skb already pass them to skb_probe_transport_header.
Move that test forward, change it to detect parse failure and drop packets on failure as those cleary are not one of the legitimate VIRTIO_NET_HDR_GSO types.
---
This captures a variety of bad gso packets, but to tighten further:
- drop SKB_GSO_DODGY packets with ipip/sit/.. , which cannot be legal. by ipip_gso_segment wrappers around inet_gso_segment expands on 121d57af308d ("gso: validate gso_type in GSO handlers")
- limit the number of ipv6 exthdrs allowed from dodgy sources. not sure where to draw the line. but not at 64K ;)
- validate the network and transport protocol returned in skb_probe_transport_header against the VIRTIO_NET_HDR_GSO type
- probe all dodgy GSO packets, also those that set checksum offload. this will have a performance impact, discussed previously in
http://patchwork.ozlabs.org/patch/861874/
but it would have blocked this latest bug as well
All but the last one seem pretty uncontroversial to me. If no one objects I plan to send those to net-next.
--- include/linux/skbuff.h | 2 +- include/linux/virtio_net.h | 9 +++++++++ 2 files changed, 10 insertions(+), 1 deletion(-)
include/linux/skbuff.h +1 -1
@@ -2434,7 +2434,7 @@ static inline void skb_probe_transport_header(struct sk_buff *skb,if (skb_flow_dissect_flow_keys_basic (skb, &keys,NULL ,0 ,0 ,0 ,0 ))skb_set_transport_header (skb, keys.control.thoff); -else +else if (offset_hint >=0 )skb_set_transport_header (skb, offset_hint); }
include/linux/virtio_net.h +9 -0
@@ -57,6 +57,15 @@ static inline int virtio_net_hdr_to_skb(struct sk_buff *skb,if (!skb_partial_csum_set (skb, start, off))return -EINVAL ; + }else { +/* gso packets without NEEDS_CSUM do not set transport_offset. +* probe and drop if does not match one of the above types. +*/ +if (gso_type) { +skb_probe_transport_header (skb,-1 ); +if (!skb_transport_header_was_set (skb)) +return -EINVAL ; + } }if (hdr->gso_type !=VIRTIO_NET_HDR_GSO_NONE ) {
Signature · 1 line
2.21.0.rc0.258.g878e2cd30e-goog
Quoted 12 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
Syzkaller again found a path to a kernel crash through bad gso input. By building an excessively large packet to cause an skb field to wrap.
If
VIRTIO_NET_HDR_F_NEEDS_CSUMwas set this would have been dropped inskb_partial_csum_set.GSO packets that do not set checksum offload are suspicious and rare. Most callers of
virtio_net_hdr_to_skbalready pass them toskb_probe_transport_header.
Quoted 8 lines
Move that test forward, change it to detect parse failure and drop packets on failure as those cleary are not one of the legitimate
VIRTIO_NET_HDR_GSOtypes.Fixes bfd5f4a3d605 ("packet: Add GSO/csum offload support.")Fixes f43798c27684 ("tun: Allow GSO using virtio_net_hdr")Reported-by syzbot <[email protected]>Signed-off-by Willem de Bruijn <[email protected]>
David Miller
maintainerFrom: Willem de Bruijn <[email protected]>
Date: Fri, 15 Feb 2019 12:15:47 -0500
Quoted 12 lines
From: Willem de Bruijn <[email protected]>
Syzkaller again found a path to a kernel crash through bad gso input. By building an excessively large packet to cause an skb field to wrap.
If
VIRTIO_NET_HDR_F_NEEDS_CSUMwas set this would have been dropped inskb_partial_csum_set.GSO packets that do not set checksum offload are suspicious and rare. Most callers of
virtio_net_hdr_to_skbalready pass them toskb_probe_transport_header.
Quoted 8 lines
Move that test forward, change it to detect parse failure and drop packets on failure as those cleary are not one of the legitimate
VIRTIO_NET_HDR_GSOtypes.Fixes bfd5f4a3d605 ("packet: Add GSO/csum offload support.")Fixes f43798c27684 ("tun: Allow GSO using virtio_net_hdr")Reported-by syzbot <[email protected]>Signed-off-by Willem de Bruijn <[email protected]>
Applied and queued up for -stable, thanks Willem.
Quoted 12 lines
This captures a variety of bad gso packets, but to tighten further:
- drop
SKB_GSO_DODGYpackets with ipip/sit/.. , which cannot be legal. byipip_gso_segmentwrappers aroundinet_gso_segmentexpands on121d57af308d("gso: validategso_typein GSO handlers")- limit the number of ipv6 exthdrs allowed from dodgy sources. not sure where to draw the line. but not at 64K ;)
- validate the network and transport protocol returned in
skb_probe_transport_headeragainst theVIRTIO_NET_HDR_GSOtype
Quoted 7 lines
- probe all dodgy GSO packets, also those that set checksum offload. this will have a performance impact, discussed previously in
http://patchwork.ozlabs.org/patch/861874/
but it would have blocked this latest bug as wellAll but the last one seem pretty uncontroversial to me. If no one objects I plan to send those to net-next.
No objections from me.
Quoted 13 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
Syzkaller again found a path to a kernel crash through bad gso input. By building an excessively large packet to cause an skb field to wrap.
If
VIRTIO_NET_HDR_F_NEEDS_CSUMwas set this would have been dropped inskb_partial_csum_set.GSO packets that do not set checksum offload are suspicious and rare. Most callers of
virtio_net_hdr_to_skbalready pass them toskb_probe_transport_header.
Quoted 8 lines
Move that test forward, change it to detect parse failure and drop packets on failure as those cleary are not one of the legitimate
VIRTIO_NET_HDR_GSOtypes.Fixes bfd5f4a3d605 ("packet: Add GSO/csum offload support.")Fixes f43798c27684 ("tun: Allow GSO using virtio_net_hdr")Reported-by syzbot <[email protected]>Signed-off-by Willem de Bruijn <[email protected]>
This causes false positive drops on virtio-net and tun for these packets with gso without csum_off. And on pf_packet with proto 0.
This happens because skb->protocol is set in these callers after the call to virtio_net_hdr_to_skb. And the flow dissector relies on this to start dissection, not the link layer header (if present). Moving this logic forward is too much churn for net, especially since eth_type_header pulls the header, requiring additional changes to adjust csum_start.
virtio_net_hdr_set_proto() aims to fix this by deriving skb->protocol from the gso_type. But unfortunately for UDP it unconditionally selects ipv4, which will cause drops for UDP over ipv6.
For net I plan to just ignore the error for these callers that do not set skb->protocol.
- if (!skb_transport_header_was_set(skb))
+ if (!skb_transport_header_was_set(skb) && skb->protocol)
Possibly with an extension of tpacket_set_protocol to also cover packet_snd, so that that cannot evade it on purpose. Other callers can wait till net-next.
Quoted 11 lines
---
This captures a variety of bad gso packets, but to tighten further:
- drop
SKB_GSO_DODGYpackets with ipip/sit/.. , which cannot be legal. byipip_gso_segmentwrappers aroundinet_gso_segmentexpands on121d57af308d("gso: validategso_typein GSO handlers")- limit the number of ipv6 exthdrs allowed from dodgy sources. not sure where to draw the line. but not at 64K ;)
This already exists, in the form of skb_flow_dissect_allowed
Quoted 52 lines
- validate the network and transport protocol returned in
skb_probe_transport_headeragainst theVIRTIO_NET_HDR_GSOtype- probe all dodgy GSO packets, also those that set checksum offload. this will have a performance impact, discussed previously in
http://patchwork.ozlabs.org/patch/861874/
but it would have blocked this latest bug as wellAll but the last one seem pretty uncontroversial to me. If no one objects I plan to send those to net-next.
--- include/linux/skbuff.h | 2 +- include/linux/virtio_net.h | 9 +++++++++ 2 files changed, 10 insertions(+), 1 deletion(-)include/linux/skbuff.h +1
-1@@ -2434,7 +2434,7 @@ static inline void skb_probe_transport_header(struct sk_buff *skb,if (skb_flow_dissect_flow_keys_basic (skb, &keys,NULL ,0 ,0 ,0 ,0 ))skb_set_transport_header (skb, keys.control.thoff); -else +else if (offset_hint >=0 )skb_set_transport_header (skb, offset_hint); }include/linux/virtio_net.h +9
-1@@ -57,6 +57,15 @@ static inline int virtio_net_hdr_to_skb(struct sk_buff *skb,if (!skb_partial_csum_set (skb, start, off))return -EINVAL ; + }else { +/* gso packets without NEEDS_CSUM do not set transport_offset. +* probe and drop if does not match one of the above types. +*/ +if (gso_type) { +skb_probe_transport_header (skb,-1 ); +if (!skb_transport_header_was_set (skb)) +return -EINVAL ; + } }if (hdr->gso_type !=VIRTIO_NET_HDR_GSO_NONE ) { --2.21.0.rc0.258.g878e2cd30e-goog
Quoted 35 lines · Willem de Bruijn
Quoted 13 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
Syzkaller again found a path to a kernel crash through bad gso input. By building an excessively large packet to cause an skb field to wrap.
If
VIRTIO_NET_HDR_F_NEEDS_CSUMwas set this would have been dropped inskb_partial_csum_set.GSO packets that do not set checksum offload are suspicious and rare. Most callers of
virtio_net_hdr_to_skbalready pass them toskb_probe_transport_header.Quoted 8 lines
Move that test forward, change it to detect parse failure and drop packets on failure as those cleary are not one of the legitimate
VIRTIO_NET_HDR_GSOtypes.Fixes bfd5f4a3d605 ("packet: Add GSO/csum offload support.")Fixes f43798c27684 ("tun: Allow GSO using virtio_net_hdr")Reported-by syzbot <[email protected]>Signed-off-by Willem de Bruijn <[email protected]>This causes false positive drops on virtio-net and tun for these packets with gso without
csum_off. And onpf_packetwith proto 0.This happens because
skb->protocolis set in these callers after the call tovirtio_net_hdr_to_skb. And the flow dissector relies on this to start dissection, not the link layer header (if present). Moving this logic forward is too much churn for net, especially sinceeth_type_headerpulls the header, requiring additional changes to adjustcsum_start.
Quoted 3 lines
virtio_net_hdr_set_proto()aims to fix this by derivingskb->protocolfrom thegso_type. But unfortunately for UDP it unconditionally selects ipv4, which will cause drops for UDP over ipv6.
Suggested fix at http://patchwork.ozlabs.org/patch/1044429/