kernel/drivers/net/ethernet/mediatek, branch linux-6.2.y

net: ethernet: mtk_eth_soc: drop generic vlan rx offload, only use DSA untagging

2023-05-17T11:58:47Z

[ Upstream commit c6d96df9fa2c1d19525239d4262889cce594ce6c ] Through testing I found out that hardware vlan rx offload support seems to have some hardware issues. At least when using multiple MACs and when receiving tagged packets on the secondary MAC, the hardware can sometimes start to emit wrong tags on the first MAC as well. In order to avoid such issues, drop the feature configuration and use the offload feature only for DSA hardware untagging on MT7621/MT7622 devices where this feature works properly. Fixes: 08666cbb7dd5 ("net: ethernet: mtk_eth_soc: add support for configuring vlan rx offload") Tested-by: Frank Wunderlich Signed-off-by: Felix Fietkau Signed-off-by: Frank Wunderlich Tested-by: Arınç ÜNAL Acked-by: Arınç ÜNAL Link: https://lore.kernel.org/r/20230426172153.8352-1-linux@fw-web.de Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: fix remaining throughput regression

2023-04-13T15:02:38Z

[ Upstream commit e669ce46740a9815953bb4452a6bc5a7fdc21a50 ] Based on further tests, it seems that the QDMA shaper is not able to perform shaping close to the MAC link rate without throughput loss. This cannot be compensated by increasing the shaping rate, so it seems to be an internal limit. Fix the remaining throughput regression by detecting that condition and limiting shaping to ports with lower link speed. This patch intentionally ignores link speed gain from TRGMII, because even on such links, shaping to 1000 Mbit/s incurs some throughput degradation. Fixes: f63959c7eec3 ("net: ethernet: mtk_eth_soc: implement multi-queue support for per-port queues") Tested-By: Frank Wunderlich Reported-by: Frank Wunderlich Signed-off-by: Felix Fietkau Signed-off-by: David S. Miller Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: add missing ppe cache flush when deleting a flow

2023-04-06T10:12:39Z

[ Upstream commit 924531326e2dd4ceabe7240f2b55a88e7d894ec2 ] The cache needs to be flushed to ensure that the hardware stops offloading the flow immediately. Fixes: 33fc42de3327 ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries") Reviewed-by: Simon Horman Signed-off-by: Felix Fietkau Reviewed-by: Leon Romanovsky Link: https://lore.kernel.org/r/20230330120840.52079-3-nbd@nbd.name Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: fix L2 offloading with DSA untag offload

2023-04-06T10:12:39Z

[ Upstream commit 5f36ca1b841fb17a20249fd9fedafc7dc7fdd940 ] Check for skb metadata in order to detect the case where the DSA header is not present. Fixes: 2d7605a72906 ("net: ethernet: mtk_eth_soc: enable hardware DSA untagging") Reviewed-by: Simon Horman Signed-off-by: Felix Fietkau Reviewed-by: Leon Romanovsky Link: https://lore.kernel.org/r/20230330120840.52079-2-nbd@nbd.name Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: fix flow block refcounting logic

2023-04-06T10:12:39Z

[ Upstream commit 8c1cb87c2a5c29da416848451a687473f379611c ] Since we call flow_block_cb_decref on FLOW_BLOCK_UNBIND, we also need to call flow_block_cb_incref for a newly allocated cb. Also fix the accidentally inverted refcount check on unbind. Fixes: 502e84e2382d ("net: ethernet: mtk_eth_soc: add flow offloading support") Reviewed-by: Simon Horman Signed-off-by: Felix Fietkau Reviewed-by: Leon Romanovsky Link: https://lore.kernel.org/r/20230330120840.52079-1-nbd@nbd.name Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: fix tx throughput regression with direct 1G links

2023-04-06T10:12:37Z

[ Upstream commit 07b3af42d8d528374d4f42d688bae86eeb30831a ] Using the QDMA tx scheduler to throttle tx to line speed works fine for switch ports, but apparently caused a regression on non-switch ports. Based on a number of tests, it seems that this throttling can be safely dropped without re-introducing the issues on switch ports that the tx scheduling changes resolved. Link: https://lore.kernel.org/netdev/trinity-92c3826f-c2c8-40af-8339-bc6d0d3ffea4-1678213958520@3c-app-gmx-bs16/ Fixes: f63959c7eec3 ("net: ethernet: mtk_eth_soc: implement multi-queue support for per-port queues") Reported-by: Frank Wunderlich Reported-by: Daniel Golle Tested-by: Daniel Golle Signed-off-by: Felix Fietkau Link: https://lore.kernel.org/r/20230324140404.95745-1-nbd@nbd.name Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: only write values if needed

2023-03-22T12:37:49Z

[ Upstream commit 6e933a804c7db8be64f367f33e63cd7dcc302ebb ] Only restart auto-negotiation and write link timer if actually necessary. This prevents losing the link in case of minor changes. Fixes: 7e538372694b ("net: ethernet: mediatek: Re-add support SGMII") Reviewed-by: Russell King (Oracle) Tested-by: Bjørn Mork Signed-off-by: Daniel Golle Signed-off-by: David S. Miller Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: reset PCS state

2023-03-22T12:37:49Z

[ Upstream commit 611e2dabb4b3243d176739fd6a5a34d007fa3f86 ] Reset the internal PCS state machine when changing interface mode. This prevents confusing the state machine when changing interface modes, e.g. from SGMII to 2500Base-X or vice-versa. Fixes: 7e538372694b ("net: ethernet: mediatek: Re-add support SGMII") Reviewed-by: Russell King (Oracle) Tested-by: Bjørn Mork Signed-off-by: Daniel Golle Signed-off-by: David S. Miller Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: fix RX data corruption issue

2023-03-17T07:57:57Z

[ Upstream commit 193250ace270fecd586dd2d0dfbd9cbd2ade977f ] Fix data corruption issue with SerDes connected PHYs operating at 1.25 Gbps speed where we could previously observe about 30% packet loss while the bad packet counter was increasing. As almost all boards with MediaTek MT7622 or MT7986 use either the MT7531 switch IC operating at 3.125Gbps SerDes rate or single-port PHYs using rate-adaptation to 2500Base-X mode, this issue only got exposed now when we started trying to use SFP modules operating with 1.25 Gbps with the BananaPi R3 board. The fix is to set bit 12 which disables the RX FIFO clear function when setting up MAC MCR, MediaTek SDK did the same change stating: "If without this patch, kernel might receive invalid packets that are corrupted by GMAC."[1] [1]: https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/d8a2975939a12686c4a95c40db21efdc3f821f63 Fixes: 42c03844e93d ("net-next: mediatek: add support for MediaTek MT7622 SoC") Tested-by: Bjørn Mork Signed-off-by: Daniel Golle Reviewed-by: Vladimir Oltean Reviewed-by: Florian Fainelli Link: https://lore.kernel.org/r/138da2735f92c8b6f8578ec2e5a794ee515b665f.1677937317.git.daniel@makrotopia.org Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: ethernet: mtk_eth_soc: fix DSA TX tag hwaccel for switch port 0

2023-02-08T09:11:31Z

Arınç reports that on his MT7621AT Unielec U7621-06 board and MT7623NI Bananapi BPI-R2, packets received by the CPU over mt7530 switch port 0 (of which this driver acts as the DSA master) are not processed correctly by software. More precisely, they arrive without a DSA tag (in packet or in the hwaccel area - skb_metadata_dst()), so DSA cannot demux them towards the switch's interface for port 0. Traffic from other ports receives a skb_metadata_dst() with the correct port and is demuxed properly. Looking at mtk_poll_rx(), it becomes apparent that this driver uses the skb vlan hwaccel area: union { u32 vlan_all; struct { __be16 vlan_proto; __u16 vlan_tci; }; }; as a temporary storage for the VLAN hwaccel tag, or the DSA hwaccel tag. If this is a DSA master it's a DSA hwaccel tag, and finally clears up the skb VLAN hwaccel header. I'm guessing that the problem is the (mis)use of API. skb_vlan_tag_present() looks like this: #define skb_vlan_tag_present(__skb) (!!(__skb)->vlan_all) So if both vlan_proto and vlan_tci are zeroes, skb_vlan_tag_present() returns precisely false. I don't know for sure what is the format of the DSA hwaccel tag, but I surely know that lowermost 3 bits of vlan_proto are 0 when receiving from port 0: unsigned int port = vlan_proto & GENMASK(2, 0); If the RX descriptor has no other bits set to non-zero values in RX_DMA_VTAG, then the call to __vlan_hwaccel_put_tag() will not, in fact, make the subsequent skb_vlan_tag_present() return true, because it's implemented like this: static inline void __vlan_hwaccel_put_tag(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) { skb->vlan_proto = vlan_proto; skb->vlan_tci = vlan_tci; } What we need to do to fix this problem (assuming this is the problem) is to stop using skb->vlan_all as temporary storage for driver affairs, and just create some local variables that serve the same purpose, but hopefully better. Instead of calling skb_vlan_tag_present(), let's look at a boolean has_hwaccel_tag which we set to true when the RX DMA descriptors have something. Disambiguate based on netdev_uses_dsa() whether this is a VLAN or DSA hwaccel tag, and only call __vlan_hwaccel_put_tag() if we're certain it's a VLAN tag. Arınç confirms that the treatment works, so this validates the assumption. Link: https://lore.kernel.org/netdev/704f3a72-fc9e-714a-db54-272e17612637@arinc9.com/ Fixes: 2d7605a72906 ("net: ethernet: mtk_eth_soc: enable hardware DSA untagging") Reported-by: Arınç ÜNAL Tested-by: Arınç ÜNAL Signed-off-by: Vladimir Oltean Reviewed-by: Felix Fietkau Signed-off-by: David S. Miller