Re: Aw: Re: [PATCH net v3 4/5] net: ethernet: mtk_eth_soc: drop generic vlan rx offload, only use DSA untagging
From: Felix Fietkau
Date: Fri Dec 30 2022 - 09:58:55 EST
On 30.12.22 14:56, Frank Wunderlich wrote:
Hi
thanks for fast response
Gesendet: Freitag, 30. Dezember 2022 um 13:56 Uhr
Von: "Felix Fietkau" <nbd@xxxxxxxx>
An: "Frank Wunderlich" <frank-w@xxxxxxxxxxxxxxx>
Cc: netdev@xxxxxxxxxxxxxxx, "John Crispin" <john@xxxxxxxxxxx>, "Sean Wang" <sean.wang@xxxxxxxxxxxx>, "Mark Lee" <Mark-MC.Lee@xxxxxxxxxxxx>, "Lorenzo Bianconi" <lorenzo@xxxxxxxxxx>, "David S. Miller" <davem@xxxxxxxxxxxxx>, "Eric Dumazet" <edumazet@xxxxxxxxxx>, "Jakub Kicinski" <kuba@xxxxxxxxxx>, "Paolo Abeni" <pabeni@xxxxxxxxxx>, "Matthias Brugger" <matthias.bgg@xxxxxxxxx>, "Russell King" <linux@xxxxxxxxxxxxxxx>, linux-arm-kernel@xxxxxxxxxxxxxxxxxxx, linux-mediatek@xxxxxxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx
Betreff: Re: Aw: [PATCH net v3 4/5] net: ethernet: mtk_eth_soc: drop generic vlan rx offload, only use DSA untagging
On 30.12.22 12:46, Frank Wunderlich wrote:
> Hi,
>
> v2 or v3 seems to break vlan on mt7986 over eth0 (mt7531 switch). v1 was working on next from end of November. But my rebased tree with v1 on 6.2-rc1 has same issue, so something after next 2711 was added which break vlan over mt7531.
>
> Directly over eth1 it works (was not working before).
>
> if i made no mistake there is still something wrong.
>
> btw. mt7622/r64 can also use second gmac (over vlan aware bridge with aux-port of switch to wan-port) it is only not default in mainline. But maybe this should not be used as decision for dropping "dsa-tag" (wrongly vlan-tag).
>
> regards Frank
Thanks for reporting.
Please try this patch on top of the series:
---
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -3218,10 +3218,8 @@ static int mtk_open(struct net_device *dev)
phylink_start(mac->phylink);
netif_tx_start_all_queues(dev);
- if (MTK_HAS_CAPS(eth->soc->caps, MTK_NETSYS_V2))
- return 0;
-
- if (mtk_uses_dsa(dev) && !eth->prog) {
+ if (!MTK_HAS_CAPS(eth->soc->caps, MTK_NETSYS_V2) &&
+ mtk_uses_dsa(dev) && !eth->prog) {
for (i = 0; i < ARRAY_SIZE(eth->dsa_meta); i++) {
struct metadata_dst *md_dst = eth->dsa_meta[i];
@@ -3244,10 +3242,6 @@ static int mtk_open(struct net_device *dev)
val &= ~MTK_CDMP_STAG_EN;
mtk_w32(eth, val, MTK_CDMP_IG_CTRL);
- val = mtk_r32(eth, MTK_CDMQ_IG_CTRL);
- val &= ~MTK_CDMQ_STAG_EN;
- mtk_w32(eth, val, MTK_CDMQ_IG_CTRL);
-
mtk_w32(eth, 0, MTK_CDMP_EG_CTRL);
}
seems not helping...this is how i test it:
ip link set eth1 up
ip link add link eth1 name vlan500 type vlan id 500
ip link add link wan name vlan600 type vlan id 600
ip addr add 192.168.50.1/24 dev vlan500
ip addr add 192.168.60.1/24 dev vlan600
ip link set vlan500 up
ip link set wan up
ip link set vlan600 up
#do this on the other side:
#netif=enp3s0
#sudo ip link add link $netif name vlan500 type vlan id 500
#sudo ip link add link $netif name vlan600 type vlan id 600
#sudo ip link set vlan500 up
#sudo ip link set vlan600 up
#sudo ip addr add 192.168.50.2/24 dev vlan500
#sudo ip addr add 192.168.60.2/24 dev vlan600
verified all used ports on my switch are in trunk-mode with vlan-membership of these 2 vlan.
booted 6.1 and there is vlan on dsa-port broken too, so either my test-setup is broken or code...but wonder why 6.1 is broken too...
with tcp-dump on my laptop i see that some packets came in for both vlan, but they seem not valid, arp only for vlan 500 (eth1 on r3).
$ sudo tcpdump -i enp3s0 -nn -e vlan
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on enp3s0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
14:49:09.548909 e4:b9:7a:f7:c4:8b > 33:33:00:00:83:84, ethertype 802.1Q (0x8100), length 524: vlan 500, p 0, ethertype IPv6 (0x86dd), fe80::e6b9:7aff:fef7:c48b.34177 > ff12::8384.21027: UDP, length 458
14:49:09.548929 e4:b9:7a:f7:c4:8b > 33:33:00:00:83:84, ethertype 802.1Q (0x8100), length 524: vlan 600, p 0, ethertype IPv6 (0x86dd), fe80::e6b9:7aff:fef7:c48b.34177 > ff12::8384.21027: UDP, length 458
14:49:09.549470 e4:b9:7a:f7:c4:8b > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 504: vlan 500, p 0, ethertype IPv4 (0x0800), 192.168.50.2.33050 > 192.168.50.255.21027: UDP, length 458
14:49:09.549522 e4:b9:7a:f7:c4:8b > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 504: vlan 600, p 0, ethertype IPv4 (0x0800), 192.168.60.2.33050 > 192.168.60.255.21027: UDP, length 458
14:49:26.324503 92:65:f3:ec:b0:19 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 60: vlan 500, p 0, ethertype ARP (0x0806), Request who-has 192.168.50.2 tell 192.168.50.1, length 42
14:49:26.324525 e4:b9:7a:f7:c4:8b > 92:65:f3:ec:b0:19, ethertype 802.1Q (0x8100), length 46: vlan 500, p 0, ethertype ARP (0x0806), Reply 192.168.50.2 is-at e4:b9:7a:f7:c4:8b, length 28
14:49:26.325091 92:65:f3:ec:b0:19 > e4:b9:7a:f7:c4:8b, ethertype 802.1Q (0x8100), length 102: vlan 500, p 0, ethertype IPv4 (0x0800), 192.168.50.1 > 192.168.50.2: ICMP echo request, id 44246, seq 1, length 64
14:49:26.325158 e4:b9:7a:f7:c4:8b > 92:65:f3:ec:b0:19, ethertype 802.1Q (0x8100), length 102: vlan 500, p 0, ethertype IPv4 (0x0800), 192.168.50.2 > 192.168.50.1: ICMP echo reply, id 44246, seq 1, length 64
on r3 i see these packets going out (so far it looks good):
root@bpi-r3:~# ping 192.168.60.2
PING 192.168.60.2 (192.168.60.2) 56(84) bytes of data.
13:30:29.782003 08:22:33:44:55:77 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 46: vlan 600, p 0, ethertype ARP (0x0806),
Request who-has 192.168.60.2 tell 192.168.60.1, length 28
13:30:30.788175 08:22:33:44:55:77 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 46: vlan 600, p 0, ethertype ARP (0x0806),
Request who-has 192.168.60.2 tell 192.168.60.1, length 28
13:30:31.828181 08:22:33:44:55:77 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 46: vlan 600, p 0, ethertype ARP (0x0806),
Request who-has 192.168.60.2 tell 192.168.60.1, length 28
From 192.168.60.1 icmp_seq=1 Destination Host Unreachable
13:30:32.868205 08:22:33:44:55:77 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 46: vlan 600, p 0, ethertype ARP (0x0806),
Request who-has 192.168.60.2 tell 192.168.60.1, length 28
13:30:33.908171 08:22:33:44:55:77 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 46: vlan 600, p 0, ethertype ARP (0x0806),
Request who-has 192.168.60.2 tell 192.168.60.1, length 28
HTH, maybe daniel or anyone other can confirm this
Does this help?
---
--- a/net/dsa/tag_mtk.c
+++ b/net/dsa/tag_mtk.c
@@ -25,6 +25,14 @@ static struct sk_buff *mtk_tag_xmit(stru
u8 xmit_tpid;
u8 *mtk_tag;
+ /* The Ethernet switch we are interfaced with needs packets to be at
+ * least 64 bytes (including FCS) otherwise their padding might be
+ * corrupted. With tags enabled, we need to make sure that packets are
+ * at least 68 bytes (including FCS and tag).
+ */
+ if (__skb_put_padto(skb, ETH_ZLEN + MTK_HDR_LEN, false))
+ return NULL;
+
/* Build the special tag after the MAC Source Address. If VLAN header
* is present, it's required that VLAN header and special tag is
* being combined. Only in this way we can allow the switch can parse