Skip to contentSkip to conversation

nft_set_pipapo: Fix AVX2 MAC address match, add test

PATCH In review Stefano Brivio 4 participants 7 messages Nov 27, 2021 – Dec 8, 2021

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

cover nft_set_pipapo: Fix AVX2 MAC address match, add test

Patch 1/2 fixes the issue reported by Nikita where a MAC address wouldn't match if given as first field of a set, and patch 2/2 adds the corresponding test.

Stefano Brivio (2):
nft_set_pipapo: Fix bucket load in AVX2 lookup routine for six 8-bit
groups
selftests: netfilter: Add correctness test for mac,net set type

 net/netfilter/nft_set_pipapo_avx2.c           |  2 +-
 .../selftests/netfilter/nft_concat_range.sh   | 24 ++++++++++++++++---
 2 files changed, 22 insertions(+), 4 deletions(-)
Signature · 1 line
2.30.2
#1 lore raw

PATCH 1/2 nft_set_pipapo: Fix bucket load in AVX2 lookup routine for six 8-bit groups

The sixth byte of packet data has to be looked up in the sixth group, not in the seventh one, even if we load the bucket data into ymm6 (and not ymm5, for convenience of tracking stalls).

Without this fix, matching on a MAC address as first field of a set, if 8-bit groups are selected (due to a small set size) would fail, that is, the given MAC address would never match.

Cc <[email protected]> # 5.6.x
Fixes 7400b063969b ("nft_set_pipapo: Introduce AVX2-based lookup implementation")
---
 net/netfilter/nft_set_pipapo_avx2.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/netfilter/nft_set_pipapo_avx2.c +1 -1
index e517663e0cd1..6f4116e72958 100644
@@ -886,7 +886,7 @@ static int nft_pipapo_avx2_lookup_8b_6(unsigned long *map, unsigned long *fill,
 			NFT_PIPAPO_AVX2_BUCKET_LOAD8(4,  lt, 4, pkt[4], bsize);
 
 			NFT_PIPAPO_AVX2_AND(5, 0, 1);
-			NFT_PIPAPO_AVX2_BUCKET_LOAD8(6,  lt, 6, pkt[5], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(6,  lt, 5, pkt[5], bsize);
 			NFT_PIPAPO_AVX2_AND(7, 2, 3);
 
 			/* Stall */
Signature · 1 line
2.30.2
Tested-by on v1 1/2
#2 lore raw

Re: [PATCH nf 1/2] nft_set_pipapo: Fix bucket load in AVX2 lookup routine for six 8-bit groups

Quoted 4 lines

The sixth byte of packet data has to be looked up in the sixth group, not in the seventh one, even if we load the bucket data into ymm6 (and not ymm5, for convenience of tracking stalls).

Quoted 8 lines

Without this fix, matching on a MAC address as first field of a set, if 8-bit groups are selected (due to a small set size) would fail, that is, the given MAC address would never match.

Cc <[email protected]> # 5.6.x
Fixes 7400b063969b ("nft_set_pipapo: Introduce AVX2-based lookup implementation")

Tried it. The issue is indeed fixed.

#3 lore raw

PATCH 2/2 selftests: netfilter: Add correctness test for mac,net set type

In reply to Stefano Brivio

The existing net,mac test didn't cover the issue recently reported by Nikita Yushchenko, where MAC addresses wouldn't match if given as first field of a concatenated set with AVX2 and 8-bit groups, because there's a different code path covering the lookup of six 8-bit groups (MAC addresses) if that's the first field.

Add a similar mac,net test, with MAC address and IPv4 address swapped in the set specification.

---
 .../selftests/netfilter/nft_concat_range.sh   | 24 ++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)
tools/testing/selftests/netfilter/nft_concat_range.sh +21 -3
index 5a4938d6dcf2..ed61f6cab60f 100755
@@ -23,8 +23,8 @@ TESTS="reported_issues correctness concurrency timeout"
 
 # Set types, defined by TYPE_ variables below
 TYPES="net_port port_net net6_port port_proto net6_port_mac net6_port_mac_proto
-       net_port_net net_mac net_mac_icmp net6_mac_icmp net6_port_net6_port
-       net_port_mac_proto_net"
+       net_port_net net_mac mac_net net_mac_icmp net6_mac_icmp
+       net6_port_net6_port net_port_mac_proto_net"
 
 # Reported bugs, also described by TYPE_ variables below
 BUGS="flush_remove_add"
@@ -277,6 +277,23 @@ perf_entries	1000
 perf_proto	ipv4
 "
 
+TYPE_mac_net="
+display		mac,net
+type_spec	ether_addr . ipv4_addr
+chain_spec	ether saddr . ip saddr
+dst		 
+src		mac addr4
+start		1
+count		5
+src_delta	2000
+tools		sendip nc bash
+proto		udp
+
+race_repeat	0
+
+perf_duration	0
+"
+
 TYPE_net_mac_icmp="
 display		net,mac - ICMP
 type_spec	ipv4_addr . ether_addr
@@ -984,7 +1001,8 @@ format() {
 		fi
 	done
 	for f in ${src}; do
-		__expr="${__expr} . "
+		[ "${__expr}" != "{ " ] && __expr="${__expr} . "
+
 		__start="$(eval format_"${f}" "${srcstart}")"
 		__end="$(eval format_"${f}" "${srcend}")"
 
Signature · 1 line
2.30.2
maintainer
#4 lore raw

Re: [PATCH nf 2/2] selftests: netfilter: Add correctness test for mac,net set type

Quoted 6 lines · Stefano Brivio

The existing net,mac test didn't cover the issue recently reported by Nikita Yushchenko, where MAC addresses wouldn't match if given as first field of a concatenated set with AVX2 and 8-bit groups, because there's a different code path covering the lookup of six 8-bit groups (MAC addresses) if that's the first field.

Quoted 7 lines

Add a similar mac,net test, with MAC address and IPv4 address swapped in the set specification.

---
 .../selftests/netfilter/nft_concat_range.sh   | 24 ++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)

<formletter>

This is not the correct way to submit patches for inclusion in the stable kernel tree. Please read:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for how to do this properly.

</formletter>

#5 lore raw

Re: [PATCH nf 2/2] selftests: netfilter: Add correctness test for mac,net set type

Hi Greg,

Quoted 15 lines · On Sat, 27 Nov 2021 12:22:39 +0100 Greg KH <[email protected]> wrote:
Quoted 6 lines · Stefano Brivio

The existing net,mac test didn't cover the issue recently reported by Nikita Yushchenko, where MAC addresses wouldn't match if given as first field of a concatenated set with AVX2 and 8-bit groups, because there's a different code path covering the lookup of six 8-bit groups (MAC addresses) if that's the first field.

Quoted 7 lines

Add a similar mac,net test, with MAC address and IPv4 address swapped in the set specification.

---
 .../selftests/netfilter/nft_concat_range.sh   | 24 ++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)  
Quoted 8 lines

<formletter>

This is not the correct way to submit patches for inclusion in the stable kernel tree. Please read:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for how to do this properly.

</formletter>

This patch (2/2) is not intended for the stable kernel tree, only patch 1/2 is. I Cc'ed [email protected] on the whole series for context.

Signature · 1 line
Stefano
#6 lore raw
In reply to Stefano Brivio
Quoted 3 lines · Stefano Brivio

Patch 1/2 fixes the issue reported by Nikita where a MAC address wouldn't match if given as first field of a set, and patch 2/2 adds the corresponding test.

Series applied, thanks