netfilter: nf_tables: fix oops when using gotos
Subsystems (2)
Pablo Neira Ayuso
maintainerPATCH 1/1 netfilter: nf_tables: fix oops when using gotos
Fix an oops if you use the "goto" operation, you will end up in non-base chain with no counters and no default policy. While at it, uninlined nft_chain_stats() since it doesn't make sense to optimize a debugging facility that is rarely used.
---
I noticed this when testing the compat layer.
@Patrick: When working on this I noticed I accidentally has skipped this patch: http://www.spinics.net/lists/netfilter-devel/msg29828.html
I have included this change in this fix. Sorry for that.
net/netfilter/nf_tables_core.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
net/netfilter/nf_tables_core.c +8 -3
@@ -67,18 +67,24 @@ struct nft_jumpstack {int rulenum; }; -static inline void +static int nft_chain_stats (const struct nft_chain *this,const struct nft_pktinfo *pkt,struct nft_jumpstack *jumpstack,unsigned int stackptr) {struct nft_stats __percpu *stats;const struct nft_chain *chain = stackptr ? jumpstack[0 ].chain : this; +/* You can reach this through goto chain */ +if (!(chain->flags &NFT_BASE_CHAIN )) +return NF_ACCEPT ; +rcu_read_lock_bh (); stats =rcu_dereference (nft_base_chain (chain)->stats);__this_cpu_inc (stats->pkts);__this_cpu_add (stats->bytes, pkt->skb->len);rcu_read_unlock_bh (); + +return nft_base_chain (chain)->policy; }enum nft_trace {
@@ -209,12 +215,11 @@ next_rule: rulenum = jumpstack[stackptr].rulenum; goto next_rule; } -nft_chain_stats (chain, pkt, jumpstack, stackptr);if (unlikely (pkt->skb->nf_trace))nft_trace_packet (pkt, chain, ++rulenum,NFT_TRACE_POLICY ); -return nft_base_chain (chain)->policy; +return nft_chain_stats (chain, pkt, jumpstack, stackptr); }EXPORT_SYMBOL_GPL (nft_do_chain);
Signature · 1 line
1.7.10.4
Pablo Neira Ayuso
maintainerQuoted 5 lines · Pablo Neira Ayuso
Fix an oops if you use the "goto" operation, you will end up in non-base chain with no counters and no default policy. While at it, uninlined
nft_chain_stats()since it doesn't make sense to optimize a debugging facility that is rarely used.
Quoted 7 lines
Signed-off-by Pablo Neira Ayuso <[email protected]>---
I noticed this when testing the compat layer.@Patrick: When working on this I noticed I accidentally has skipped this patch: http://www.spinics.net/lists/netfilter-devel/msg29828.html
I have included this change in this fix. Sorry for that.
Oops, unlined the wrong function...
Pablo Neira Ayuso
maintainerQuoted 10 lines · Pablo Neira Ayuso
Quoted 9 lines · Pablo Neira Ayuso
Fix an oops if you use the "goto" operation, you will end up in non-base chain with no counters and no default policy. While at it, uninlined
nft_chain_stats()since it doesn't make sense to optimize a debugging facility that is rarely used.Signed-off-by Pablo Neira Ayuso <[email protected]>---
I noticed this when testing the compat layer.
Quoted 5 lines
Quoted 3 lines
@Patrick: When working on this I noticed I accidentally has skipped this patch: http://www.spinics.net/lists/netfilter-devel/msg29828.html
I have included this change in this fix. Sorry for that.Oops, unlined the wrong function...
I'm going to push this leftover fix to nf:
http://patchwork.ozlabs.org/patch/308930/
Will resend a v2 to fix the goto issue.
Quoted 15 lines · Pablo Neira Ayuso
Quoted 14 lines · Pablo Neira Ayuso
Quoted 12 lines · Pablo Neira Ayuso
Fix an oops if you use the "goto" operation, you will end up in non-base chain with no counters and no default policy. While at it, uninlined
nft_chain_stats()since it doesn't make sense to optimize a debugging facility that is rarely used.Signed-off-by Pablo Neira Ayuso <[email protected]>---
I noticed this when testing the compat layer.@Patrick: When working on this I noticed I accidentally has skipped this patch: http://www.spinics.net/lists/netfilter-devel/msg29828.html
I have included this change in this fix. Sorry for that.
Quoted 7 lines
Quoted 1 line
Oops, unlined the wrong function...
I'm going to push this leftover fix to nf:
http://patchwork.ozlabs.org/patch/308930/
Will resend a v2 to fix the goto issue.
Thanks! I was also considering moving this to jump labels and enable the label on nft meta set trace. Not sure if its worth it though, will do some benchmarking at some point ...