[syzkaller] WARNING in __skb_flow_dissect
#494
Comments
|
this looks like unrelated from mptcp. nf_send_reset() allocates a new skb - with NULL dev and sk - and calls ip_local_out(). The NF hooked there calls nft_trace_init(), which in turns ask for skb_get_hash(skb), and the latter is deemed to splat with when both skb->dev and skb->sk are NULL. |
|
@pabeni: thank you for having checked that! I don't see similar reports on lore after a quick search. I guess it would be better to forward this to netfilter ML before closing this, no? @cpaasch do you mind forwarding this to |
intel-lab-lkp
pushed a commit
to intel-lab-lkp/linux
that referenced
this issue
Jun 4, 2024
skb_get_hash() triggers a (harmless) warn when neither skb->sk or skb->dev is set. In case of nf-generated tcp reset, both sk and dev are NULL: WARNING: .. net/core/flow_dissector.c:1104 [..] skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline] skb_get_hash include/linux/skbuff.h:1578 [inline] nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320 nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268 nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23 nf_hook_slow+0x57/0x160 net/netfilter/core.c:626 __ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118 ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127 nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308 nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30 [..] Fixes: d0e13a1 ("flow_dissector: lookup netns by skb->sk if skb->dev is NULL") Reported-by: Christoph Paasch <[email protected]> Suggested-by: Paolo Abeni <[email protected]> Closes: multipath-tcp/mptcp_net-next#494 Signed-off-by: Florian Westphal <[email protected]>
|
Floriant sent a fix for this: https://lore.kernel.org/netfilter-devel/[email protected]/T/#u |
|
@pabeni : Thank you for the link! I'm closing this ticket, because it will likely not be automatically closed as this patch will be added in another tree, not committed by one of us. |
cpaasch
pushed a commit
that referenced
this issue
Jun 5, 2024
skb_get_hash() triggers a (harmless) warn when neither skb->sk or skb->dev is set. In case of nf-generated tcp reset, both sk and dev are NULL: WARNING: .. net/core/flow_dissector.c:1104 [..] skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline] skb_get_hash include/linux/skbuff.h:1578 [inline] nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320 nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268 nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23 nf_hook_slow+0x57/0x160 net/netfilter/core.c:626 __ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118 ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127 nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308 nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30 [..] Fixes: d0e13a1 ("flow_dissector: lookup netns by skb->sk if skb->dev is NULL") Reported-by: Christoph Paasch <[email protected]> Suggested-by: Paolo Abeni <[email protected]> Closes: #494 Signed-off-by: Florian Westphal <[email protected]>
|
Still happening: |
|
@cpaasch : Just to be sure, that's with @fw-strlen's patch from your syzkaller_export branch, right? |
|
Yes |
intel-lab-lkp
pushed a commit
to intel-lab-lkp/linux
that referenced
this issue
Jun 7, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1 # calls skb_get_hash
tcp dport 42 reject with tcp reset # emits skb with NULL skb dev/sk
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Signed-off-by: Florian Westphal <[email protected]>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 9, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 10, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 11, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
cpaasch
pushed a commit
that referenced
this issue
Jun 12, 2024
skb_get_hash() triggers a (harmless) warn when neither skb->sk or skb->dev is set. In case of nf-generated tcp reset, both sk and dev are NULL: WARNING: .. net/core/flow_dissector.c:1104 [..] skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline] skb_get_hash include/linux/skbuff.h:1578 [inline] nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320 nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268 nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23 nf_hook_slow+0x57/0x160 net/netfilter/core.c:626 __ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118 ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127 nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308 nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30 [..] Fixes: d0e13a1 ("flow_dissector: lookup netns by skb->sk if skb->dev is NULL") Reported-by: Christoph Paasch <[email protected]> Suggested-by: Paolo Abeni <[email protected]> Closes: #494 Signed-off-by: Florian Westphal <[email protected]>
cpaasch
pushed a commit
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: #494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
kuba-moo
pushed a commit
to linux-netdev/testing
that referenced
this issue
Jun 12, 2024
Years ago flow dissector gained ability to delegate flow dissection
to a bpf program, scoped per netns.
Unfortunately, skb_get_hash() only gets an sk_buff argument instead
of both net+skb. This means the flow dissector needs to obtain the
netns pointer from somewhere else.
The netns is derived from skb->dev, and if that is not available, from
skb->sk. If neither is set, we hit a (benign) WARN_ON_ONCE().
Trying both dev and sk covers most cases, but not all, as recently
reported by Christoph Paasch.
In case of nf-generated tcp reset, both sk and dev are NULL:
WARNING: .. net/core/flow_dissector.c:1104
skb_flow_dissect_flow_keys include/linux/skbuff.h:1536 [inline]
skb_get_hash include/linux/skbuff.h:1578 [inline]
nft_trace_init+0x7d/0x120 net/netfilter/nf_tables_trace.c:320
nft_do_chain+0xb26/0xb90 net/netfilter/nf_tables_core.c:268
nft_do_chain_ipv4+0x7a/0xa0 net/netfilter/nft_chain_filter.c:23
nf_hook_slow+0x57/0x160 net/netfilter/core.c:626
__ip_local_out+0x21d/0x260 net/ipv4/ip_output.c:118
ip_local_out+0x26/0x1e0 net/ipv4/ip_output.c:127
nf_send_reset+0x58c/0x700 net/ipv4/netfilter/nf_reject_ipv4.c:308
nft_reject_ipv4_eval+0x53/0x90 net/ipv4/netfilter/nft_reject_ipv4.c:30
[..]
syzkaller did something like this:
table inet filter {
chain input {
type filter hook input priority filter; policy accept;
meta nftrace set 1
tcp dport 42 reject with tcp reset
}
chain output {
type filter hook output priority filter; policy accept;
# empty chain is enough
}
}
... then sends a tcp packet to port 42.
Initial attempt to simply set skb->dev from nf_reject_ipv4 doesn't cover
all cases: skbs generated via ipv4 igmp_send_report trigger similar splat.
Moreover, Pablo Neira found that nft_hash.c uses __skb_get_hash_symmetric()
which would trigger same warn splat for such skbs.
Lets allow callers to pass the current netns explicitly.
The nf_trace infrastructure is adjusted to use the new helper.
__skb_get_hash_symmetric is handled in the next patch.
Reported-by: Christoph Paasch <[email protected]>
Closes: multipath-tcp/mptcp_net-next#494
Reviewed-by: Eric Dumazet <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Signed-off-by: Florian Westphal <[email protected]>
Signed-off-by: NipaLocal <nipa@local>
|
@cpaasch : do you still have this issues? It looks like Florian's patch has been applied: https://lore.kernel.org/netdev/[email protected]/T/ Note that there might be a follow-up patch silencing a |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
syzkaller-id: 8f53d0a72131ff6aa6de74e1f3d77e5d400e1e84
HEAD: f60fb4a
Trace:
Happened on K5
No reproducer
The text was updated successfully, but these errors were encountered: