flow_dissector: support FLOW_DISSECTOR_KEY_ETH_ADDRS with BPF
Subsystems (5)
PATCH 1/2 flow_dissector: support FLOW_DISSECTOR_KEY_ETH_ADDRS with BPF
If we have a flow dissector BPF program attached to the namespace, FLOW_DISSECTOR_KEY_ETH_ADDRS won't trigger because we exit early.
Handle FLOW_DISSECTOR_KEY_ETH_ADDRS before BPF and only if we have an skb (used by tc-flower only).
--- net/core/flow_dissector.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-)
net/core/flow_dissector.c +12 -11
@@ -825,6 +825,18 @@ bool __skb_flow_dissect(const struct net *net,else if (skb->sk) net =sock_net (skb->sk); } + +if (dissector_uses_key (flow_dissector, +FLOW_DISSECTOR_KEY_ETH_ADDRS )) { +struct ethhdr *eth =eth_hdr (skb); +struct flow_dissector_key_eth_addrs *key_eth_addrs; + + key_eth_addrs =skb_flow_dissector_target (flow_dissector, +FLOW_DISSECTOR_KEY_ETH_ADDRS , + target_container); +memcpy (key_eth_addrs, ð->h_dest, +sizeof (*key_eth_addrs)); + } }WARN_ON_ONCE (!net);
@@ -860,17 +872,6 @@ bool __skb_flow_dissect(const struct net *net,rcu_read_unlock (); } -if (dissector_uses_key (flow_dissector, -FLOW_DISSECTOR_KEY_ETH_ADDRS )) { -struct ethhdr *eth =eth_hdr (skb); -struct flow_dissector_key_eth_addrs *key_eth_addrs; - - key_eth_addrs =skb_flow_dissector_target (flow_dissector, -FLOW_DISSECTOR_KEY_ETH_ADDRS , - target_container); -memcpy (key_eth_addrs, ð->h_dest,sizeof (*key_eth_addrs)); - } - proto_again: fdret =FLOW_DISSECT_RET_CONTINUE ;
Signature · 1 line
2.21.0.1020.gf2820cf01a-goog
PATCH 2/2 selftests/bpf: test L2 dissection in flow dissector
Make sure that everything that's coming from a pre-defined mac address can be dropped.
--- .../selftests/bpf/test_flow_dissector.sh | 23 +++++++++++++++++++ 1 file changed, 23 insertions(+)
tools/testing/selftests/bpf/test_flow_dissector.sh +23 -0
@@ -112,4 +112,27 @@ tc filter add dev lo parent ffff: protocol ipv6 pref 1337 flower ip_proto \ # Send 10 IPv6/UDP packets from port 10. Filter should not drop any. ./test_flow_dissector -i 6 -f 10 +tc filter del dev lo ingress pref 1337 + +echo "Testing L2..." +ip link set lo address 02:01:03:04:05:06 + +# Drops all packets coming from forged localhost mac +tc filter add dev lo parent ffff: protocol ip pref 1337 flower \ + src_mac 02:01:03:04:05:06 action drop + +# Send packets from any port. Filter should drop all. +./test_flow_dissector -i 4 -f 8 -F + +tc filter del dev lo ingress pref 1337 + +# Drops all packets coming from "random" non-localhost mac +tc filter add dev lo parent ffff: protocol ip pref 1337 flower \ + src_mac 02:01:03:04:05:07 action drop + +# Send packets from any port. Filter should not drop any. +./test_flow_dissector -i 4 -f 8 + +tc filter del dev lo ingress pref 1337 + exit 0
Signature · 1 line
2.21.0.1020.gf2820cf01a-goog
Quoted 3 lines · Stanislav Fomichev
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.
I suppose that this is true for a variety of keys? For instance, also FLOW_DISSECTOR_KEY_IPV4_ADDRS.
We originally intended BPF flow dissection for all paths except tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.
Quoted 15 lines
Handle
FLOW_DISSECTOR_KEY_ETH_ADDRSbefore BPF and only if we have an skb (used by tc-flower only).Fixes d58e468b1112 ("flow_dissector: implements flow dissector BPF hook")Signed-off-by Stanislav Fomichev <[email protected]>--- net/core/flow_dissector.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-)net/core/flow_dissector.c +0
-0
net/core/flow_dissector.c bool __skb_flow_dissect(const struct net *net, · line 831
Quoted 7 lines
net/core/flow_dissector.c +4
-0@@ -825,6 +825,18 @@ bool __skb_flow_dissect(const struct net *net,else if (skb->sk) net =sock_net (skb->sk); } + +if (dissector_uses_key (flow_dissector, +FLOW_DISSECTOR_KEY_ETH_ADDRS )) { +struct ethhdr *eth =eth_hdr (skb);
Here as well as in the original patch: is it safe to just cast to eth_hdr? In the same file, __skb_flow_dissect_gre does test for (encapsulated) protocol first.
Quoted 7 lines · Willem de Bruijn
Quoted 3 lines · Stanislav Fomichev
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.I suppose that this is true for a variety of keys? For instance, also
FLOW_DISSECTOR_KEY_IPV4_ADDRS.
I though the intent was to support most of the basic stuff (eth/ip/tcp/udp) without any esoteric protocols. Not sure about FLOW_DISSECTOR_KEY_IPV4_ADDRS, looks like we support that (except FLOW_DISSECTOR_KEY_TIPC part).
Quoted 4 lines
We originally intended BPF flow dissection for all paths except
tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.
Sorry, didn't get what you meant by the real fix.
Don't care about tc_flower? Just support a minimal set of features needed by selftests?
Quoted 10 lines
Quoted 10 lines
Handle
FLOW_DISSECTOR_KEY_ETH_ADDRSbefore BPF and only if we have an skb (used by tc-flower only).Fixes d58e468b1112 ("flow_dissector: implements flow dissector BPF hook")Signed-off-by Stanislav Fomichev <[email protected]>--- net/core/flow_dissector.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-)
Quoted 16 lines
net/core/flow_dissector.c bool __skb_flow_dissect(const struct net *net, · line 831
Quoted 12 lines
net/core/flow_dissector.c +4
-0@@ -825,6 +825,18 @@ bool __skb_flow_dissect(const struct net *net,else if (skb->sk) net =sock_net (skb->sk); } + +if (dissector_uses_key (flow_dissector, +FLOW_DISSECTOR_KEY_ETH_ADDRS )) { +struct ethhdr *eth =eth_hdr (skb);Here as well as in the original patch: is it safe to just cast to
eth_hdr? In the same file,__skb_flow_dissect_gredoes test for (encapsulated) protocol first.
Good question, I guess the assumption here is that FLOW_DISSECTOR_KEY_ETH_ADDRS is only used by tc_flower and the appropriate checks should be there as well.
It's probably better to check skb->proto here though.
Quoted 4 lines · Stanislav Fomichev
Quoted 2 lines · Willem de Bruijn
Quoted 1 line · Stanislav Fomichev
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.I suppose that this is true for a variety of keys? For instance, also
FLOW_DISSECTOR_KEY_IPV4_ADDRS.I though the intent was to support most of the basic stuff (eth/ip/tcp/udp) without any esoteric protocols.
Indeed. But this applies both to protocols and the feature set. Both are more limited.
Quoted 2 lines
Not sure about
FLOW_DISSECTOR_KEY_IPV4_ADDRS,
looks like we support that (exceptFLOW_DISSECTOR_KEY_TIPCpart).
Ah, I chose a bad example then.
Quoted 8 lines
Quoted 4 lines
We originally intended BPF flow dissection for all paths except
tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.Sorry, didn't get what you meant by the real fix.
Don't care abouttc_flower? Just support a minimal set of features needed by selftests?
I do mean exclude BPF flow dissector (only) for tc_flower, as we cannot guarantee that the BPF program can fully implement the requested feature.
Quoted 28 lines
Quoted 11 lines
Quoted 11 lines
Handle
FLOW_DISSECTOR_KEY_ETH_ADDRSbefore BPF and only if we have an skb (used by tc-flower only).Fixes d58e468b1112 ("flow_dissector: implements flow dissector BPF hook")Signed-off-by Stanislav Fomichev <[email protected]>--- net/core/flow_dissector.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-)Quoted 16 lines
net/core/flow_dissector.c bool __skb_flow_dissect(const struct net *net, · line 831
Quoted 12 lines
net/core/flow_dissector.c +4
-0@@ -825,6 +825,18 @@ bool __skb_flow_dissect(const struct net *net,else if (skb->sk) net =sock_net (skb->sk); } + +if (dissector_uses_key (flow_dissector, +FLOW_DISSECTOR_KEY_ETH_ADDRS )) { +struct ethhdr *eth =eth_hdr (skb);Here as well as in the original patch: is it safe to just cast to
eth_hdr? In the same file,__skb_flow_dissect_gredoes test for (encapsulated) protocol first.
Quoted 4 lines
Good question, I guess the assumption here is that
FLOW_DISSECTOR_KEY_ETH_ADDRSis only used bytc_flowerand the appropriate checks should be there as well.
It's probably better to checkskb->protohere though.
Right, as a mistaken or malicious admin can request it on a non Ethernet device and read garbage.
Quoted 28 lines · Willem de Bruijn
Quoted 4 lines · Stanislav Fomichev
Quoted 2 lines · Willem de Bruijn
Quoted 1 line · Stanislav Fomichev
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.I suppose that this is true for a variety of keys? For instance, also
FLOW_DISSECTOR_KEY_IPV4_ADDRS.I though the intent was to support most of the basic stuff (eth/ip/tcp/udp) without any esoteric protocols.
Indeed. But this applies both to protocols and the feature set. Both are more limited.
Quoted 2 lines
Not sure about
FLOW_DISSECTOR_KEY_IPV4_ADDRS,
looks like we support that (exceptFLOW_DISSECTOR_KEY_TIPCpart).Ah, I chose a bad example then.
Quoted 4 lines
Quoted 4 lines
We originally intended BPF flow dissection for all paths except
tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.
Quoted 7 lines
Quoted 3 lines
Sorry, didn't get what you meant by the real fix.
Don't care abouttc_flower? Just support a minimal set of features needed by selftests?I do mean exclude BPF flow dissector (only) for
tc_flower, as we cannot guarantee that the BPF program can fully implement the requested feature.
Though, the user inserting the BPF flow dissector is the same as the user inserting the flower program, the (per netns) admin. So arguably is aware of the constraints incurred by BPF flow dissection. And else can still detect when a feature is not supported from the (lack of) output, as in this case of Ethernet address. I don't think we want to mix BPF and non-BPF flow dissection though. That subverts the safety argument of switching to BPF for flow dissection.
Quoted 38 lines · Willem de Bruijn
Quoted 28 lines · Willem de Bruijn
Quoted 4 lines · Stanislav Fomichev
Quoted 2 lines · Willem de Bruijn
Quoted 1 line · Stanislav Fomichev
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.I suppose that this is true for a variety of keys? For instance, also
FLOW_DISSECTOR_KEY_IPV4_ADDRS.I though the intent was to support most of the basic stuff (eth/ip/tcp/udp) without any esoteric protocols.
Indeed. But this applies both to protocols and the feature set. Both are more limited.
Quoted 2 lines
Not sure about
FLOW_DISSECTOR_KEY_IPV4_ADDRS,
looks like we support that (exceptFLOW_DISSECTOR_KEY_TIPCpart).Ah, I chose a bad example then.
Quoted 4 lines
Quoted 4 lines
We originally intended BPF flow dissection for all paths except
tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.Quoted 7 lines
Quoted 3 lines
Sorry, didn't get what you meant by the real fix.
Don't care abouttc_flower? Just support a minimal set of features needed by selftests?I do mean exclude BPF flow dissector (only) for
tc_flower, as we cannot guarantee that the BPF program can fully implement the requested feature.
Quoted 7 lines
Though, the user inserting the BPF flow dissector is the same as the user inserting the flower program, the (per netns) admin. So arguably is aware of the constraints incurred by BPF flow dissection. And else can still detect when a feature is not supported from the (lack of) output, as in this case of Ethernet address. I don't think we want to mix BPF and non-BPF flow dissection though. That subverts the safety argument of switching to BPF for flow dissection.
Yes, we cannot completely avoid tc_flower because we use it to do the end-to-end testing. That's why I was trying to make sure "basic" stuff works (it might feel confusing that tc_flower {src,dst}_mac stop working with a bpf program installed).
TBH, I'd not call this particular piece of code that exports src/dst addresses a dissection. At this point, it's a well-formed skb with a proper l2 header and we just copy the addresses out. It's probably part of the reason the original patch didn't include any skb->protocol checks.
Quoted 51 lines · Willem de Bruijn
Quoted 38 lines
Quoted 28 lines · Willem de Bruijn
Quoted 4 lines · Stanislav Fomichev
Quoted 2 lines · Willem de Bruijn
Quoted 1 line · Stanislav Fomichev
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.I suppose that this is true for a variety of keys? For instance, also
FLOW_DISSECTOR_KEY_IPV4_ADDRS.I though the intent was to support most of the basic stuff (eth/ip/tcp/udp) without any esoteric protocols.
Indeed. But this applies both to protocols and the feature set. Both are more limited.
Quoted 2 lines
Not sure about
FLOW_DISSECTOR_KEY_IPV4_ADDRS,
looks like we support that (exceptFLOW_DISSECTOR_KEY_TIPCpart).Ah, I chose a bad example then.
Quoted 4 lines
Quoted 4 lines
We originally intended BPF flow dissection for all paths except
tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.Quoted 7 lines
Quoted 3 lines
Sorry, didn't get what you meant by the real fix.
Don't care abouttc_flower? Just support a minimal set of features needed by selftests?I do mean exclude BPF flow dissector (only) for
tc_flower, as we cannot guarantee that the BPF program can fully implement the requested feature.Quoted 7 lines
Though, the user inserting the BPF flow dissector is the same as the user inserting the flower program, the (per netns) admin. So arguably is aware of the constraints incurred by BPF flow dissection. And else can still detect when a feature is not supported from the (lack of) output, as in this case of Ethernet address. I don't think we want to mix BPF and non-BPF flow dissection though. That subverts the safety argument of switching to BPF for flow dissection.
Yes, we cannot completely avoid
tc_flowerbecause we use it to do the end-to-end testing. That's why I was trying to make sure "basic" stuff works (it might feel confusing thattc_flower{src,dst}_mac stop working with a bpf program installed).
Quoted 5 lines
TBH, I'd not call this particular piece of code that exports src/dst addresses a dissection. At this point, it's a well-formed skb with a proper l2 header and we just copy the addresses out. It's probably part of the reason the original patch didn't include any
skb->protocolchecks.
On the other hand, we can probably follow a simple rule:
if it's not exported via bpf_flow_keys (and src/dsc mac is not), tc_flower is not supported as well.
From: Stanislav Fomichev <[email protected]>
Date: Mon, May 13, 2019 at 7:21 PM
To: Stanislav Fomichev
Cc: Willem de Bruijn <[email protected]>, Network Development, bpf, David Miller, Alexei Starovoitov, Daniel Borkmann, Willem de Bruijn <[email protected]>, Petar Penkov
Quoted 51 lines
Quoted 51 lines · Willem de Bruijn
Quoted 38 lines
Quoted 28 lines · Willem de Bruijn
Quoted 4 lines · Stanislav Fomichev
Quoted 2 lines · Willem de Bruijn
Quoted 1 line · Stanislav Fomichev
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
If we have a flow dissector BPF program attached to the namespace,
FLOW_DISSECTOR_KEY_ETH_ADDRSwon't trigger because we exit early.I suppose that this is true for a variety of keys? For instance, also
FLOW_DISSECTOR_KEY_IPV4_ADDRS.I though the intent was to support most of the basic stuff (eth/ip/tcp/udp) without any esoteric protocols.
Indeed. But this applies both to protocols and the feature set. Both are more limited.
Quoted 2 lines
Not sure about
FLOW_DISSECTOR_KEY_IPV4_ADDRS,
looks like we support that (exceptFLOW_DISSECTOR_KEY_TIPCpart).Ah, I chose a bad example then.
Quoted 4 lines
Quoted 4 lines
We originally intended BPF flow dissection for all paths except
tc_flower. As that catches all the vulnerable cases on the ingress path on the one hand and it is infeasible to support all the flower features, now and future. I think that is the real fix.Quoted 7 lines
Quoted 3 lines
Sorry, didn't get what you meant by the real fix.
Don't care abouttc_flower? Just support a minimal set of features needed by selftests?I do mean exclude BPF flow dissector (only) for
tc_flower, as we cannot guarantee that the BPF program can fully implement the requested feature.Quoted 7 lines
Though, the user inserting the BPF flow dissector is the same as the user inserting the flower program, the (per netns) admin. So arguably is aware of the constraints incurred by BPF flow dissection. And else can still detect when a feature is not supported from the (lack of) output, as in this case of Ethernet address. I don't think we want to mix BPF and non-BPF flow dissection though. That subverts the safety argument of switching to BPF for flow dissection.
Yes, we cannot completely avoid
tc_flowerbecause we use it to do the end-to-end testing. That's why I was trying to make sure "basic" stuff works (it might feel confusing thattc_flower{src,dst}_mac stop working with a bpf program installed).
Quoted 5 lines
Quoted 5 lines
TBH, I'd not call this particular piece of code that exports src/dst addresses a dissection. At this point, it's a well-formed skb with a proper l2 header and we just copy the addresses out. It's probably part of the reason the original patch didn't include any
skb->protocolchecks.
But it is not guaranteed to be an Ethernet link layer device. Making this a good example of why when moving to BPF for safety we should not keep any C dissection code in the path at all.
Quoted 3 lines
On the other hand, we can probably follow a simple rule:
if it's not exported viabpf_flow_keys(and src/dsc mac is not),tc_floweris not supported as well.
Agreed. I was using that as point of reference just now, too.