[PATCH] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear

[PATCH] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear

[Daniel Golle]

Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
on older MTK_NETSYS_V1 SoCs; OpenWrt users have confirmed the bug on
MT7622 and MT7621 systems.
Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.

Tested on MT7622 (Linksys E8450) and MT7986 (BananaPi BPI-R3).

Suggested-by: Felix Fietkau <nbd@nbd.name>
Fixes: 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear")
Fixes: 33fc42de33278 ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
---
 drivers/net/ethernet/mediatek/mtk_ppe.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_ppe.c b/drivers/net/ethernet/mediatek/mtk_ppe.c
index 887f430734f747..ae00e572390d7b 100644
--- a/drivers/net/ethernet/mediatek/mtk_ppe.c
+++ b/drivers/net/ethernet/mediatek/mtk_ppe.c
@@ -442,7 +442,7 @@ __mtk_foe_entry_clear(struct mtk_ppe *ppe, struct mtk_flow_entry *entry)
 		struct mtk_foe_entry *hwe = mtk_foe_get_entry(ppe, entry->hash);
 
 		hwe->ib1 &= ~MTK_FOE_IB1_STATE;
-		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_UNBIND);
+		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_INVALID);
 		dma_wmb();
 	}
 	entry->hash = 0xffff;
-- 
2.37.3

Re: [PATCH] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear

[Lorenzo Bianconi]

[-- Attachment #1: Type: text/plain, Size: 1650 bytes --]

> Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
> routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
> fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
> on older MTK_NETSYS_V1 SoCs; OpenWrt users have confirmed the bug on
> MT7622 and MT7621 systems.
> Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
> works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.
> 
> Tested on MT7622 (Linksys E8450) and MT7986 (BananaPi BPI-R3).
> 
> Suggested-by: Felix Fietkau <nbd@nbd.name>
> Fixes: 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear")
> Fixes: 33fc42de33278 ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
> Signed-off-by: Daniel Golle <daniel@makrotopia.org>

Acked-by: Lorenzo Bianconi <lorenzo@kernel.org>

> ---
>  drivers/net/ethernet/mediatek/mtk_ppe.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/mediatek/mtk_ppe.c b/drivers/net/ethernet/mediatek/mtk_ppe.c
> index 887f430734f747..ae00e572390d7b 100644
> --- a/drivers/net/ethernet/mediatek/mtk_ppe.c
> +++ b/drivers/net/ethernet/mediatek/mtk_ppe.c
> @@ -442,7 +442,7 @@ __mtk_foe_entry_clear(struct mtk_ppe *ppe, struct mtk_flow_entry *entry)
>  		struct mtk_foe_entry *hwe = mtk_foe_get_entry(ppe, entry->hash);
>  
>  		hwe->ib1 &= ~MTK_FOE_IB1_STATE;
> -		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_UNBIND);
> +		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_INVALID);
>  		dma_wmb();
>  	}
>  	entry->hash = 0xffff;
> -- 
> 2.37.3
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear

[Jakub Kicinski]

On Wed, 28 Sep 2022 02:25:15 +0100 Daniel Golle wrote:
> Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
> routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
> fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
> on older MTK_NETSYS_V1 SoCs; OpenWrt users have confirmed the bug on
> MT7622 and MT7621 systems.
> Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
> works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.
> 
> Tested on MT7622 (Linksys E8450) and MT7986 (BananaPi BPI-R3).
> 
> Suggested-by: Felix Fietkau <nbd@nbd.name>
> Fixes: 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear")
> Fixes: 33fc42de33278 ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
> Signed-off-by: Daniel Golle <daniel@makrotopia.org>

Looks like this patch is generated on top of net-next while it fixes 
a bug in net (judging by mention of 0e80707d94e4c8).
Please rebase on top of net and resent, we'll deal with the conflict.

[PATCH v2] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear

[Daniel Golle]

Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
on older MTK_NETSYS_V1 SoCs, OpenWrt users have confirmed the bug on
MT7622 and MT7621 systems.
Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.

Tested on MT7622 (Linksys E8450) and MT7986 (BananaPi BPI-R3).

Suggested-by: Felix Fietkau <nbd@nbd.name>
Fixes: 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear")
Fixes: 33fc42de33278b ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
---
v2: rebased on top of netdev/net.git;main

 drivers/net/ethernet/mediatek/mtk_ppe.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_ppe.c b/drivers/net/ethernet/mediatek/mtk_ppe.c
index cfe804bc8d2055..148ea636ef9794 100644
--- a/drivers/net/ethernet/mediatek/mtk_ppe.c
+++ b/drivers/net/ethernet/mediatek/mtk_ppe.c
@@ -412,7 +412,7 @@ __mtk_foe_entry_clear(struct mtk_ppe *ppe, struct mtk_flow_entry *entry)
 	if (entry->hash != 0xffff) {
 		ppe->foe_table[entry->hash].ib1 &= ~MTK_FOE_IB1_STATE;
 		ppe->foe_table[entry->hash].ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE,
-							      MTK_FOE_STATE_UNBIND);
+							      MTK_FOE_STATE_INVALID);
 		dma_wmb();
 	}
 	entry->hash = 0xffff;
-- 
2.37.3

Re: [PATCH v2] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (master)
by Jakub Kicinski <kuba@kernel.org>:

On Fri, 30 Sep 2022 01:56:53 +0100 you wrote:
> Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
> routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
> fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
> on older MTK_NETSYS_V1 SoCs, OpenWrt users have confirmed the bug on
> MT7622 and MT7621 systems.
> Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
> works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.
> 
> [...]

Here is the summary with links:
  - [v2] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear
    https://git.kernel.org/netdev/net/c/ae3ed15da588

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

Re: [PATCH v2] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear: manual merge

[Matthieu Baerts]

[-- Attachment #1: Type: text/plain, Size: 1534 bytes --]

Hello,

On 30/09/2022 02:56, Daniel Golle wrote:
> Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
> routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
> fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
> on older MTK_NETSYS_V1 SoCs, OpenWrt users have confirmed the bug on
> MT7622 and MT7621 systems.
> Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
> works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.
> 
> Tested on MT7622 (Linksys E8450) and MT7986 (BananaPi BPI-R3).
> 
> Suggested-by: Felix Fietkau <nbd@nbd.name>
> Fixes: 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear")
> Fixes: 33fc42de33278b ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
> Signed-off-by: Daniel Golle <daniel@makrotopia.org>
> ---
> v2: rebased on top of netdev/net.git;main

FYI and as expected when reading this email thread, we got a small
conflict when merging -net in net-next in the MPTCP tree due to this
patch applied in -net:

  ae3ed15da588 ("net: ethernet: mtk_eth_soc: fix state in
__mtk_foe_entry_clear")

and this one from net-next:

  9d8cb4c096ab ("net: ethernet: mtk_eth_soc: add foe_entry_size to
mtk_eth_soc")

The conflict has been resolved on our side[1] inspired by Daniel's v1.
The resolution we suggest is attached to this email.

Cheers,
Matt

[1] https://github.com/multipath-tcp/mptcp_net-next/commit/7af5fac658ba
-- 
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net

[-- Attachment #2: 7af5fac658ba05d4f5ba19fa7e054aa4ef917128.patch --]
[-- Type: text/x-patch, Size: 786 bytes --]

diff --cc drivers/net/ethernet/mediatek/mtk_ppe.c
index 887f430734f7,148ea636ef97..ae00e572390d
--- a/drivers/net/ethernet/mediatek/mtk_ppe.c
+++ b/drivers/net/ethernet/mediatek/mtk_ppe.c
@@@ -439,10 -410,9 +439,10 @@@ __mtk_foe_entry_clear(struct mtk_ppe *p
  
  	hlist_del_init(&entry->list);
  	if (entry->hash != 0xffff) {
 -		ppe->foe_table[entry->hash].ib1 &= ~MTK_FOE_IB1_STATE;
 -		ppe->foe_table[entry->hash].ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE,
 -							      MTK_FOE_STATE_INVALID);
 +		struct mtk_foe_entry *hwe = mtk_foe_get_entry(ppe, entry->hash);
 +
 +		hwe->ib1 &= ~MTK_FOE_IB1_STATE;
- 		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_UNBIND);
++		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_INVALID);
  		dma_wmb();
  	}
  	entry->hash = 0xffff;

Re: [PATCH v2] net: ethernet: mtk_eth_soc: fix state in __mtk_foe_entry_clear: manual merge

[Daniel Golle]

On Mon, Oct 03, 2022 at 02:59:32PM +0200, Matthieu Baerts wrote:
> Hello,
> 
> On 30/09/2022 02:56, Daniel Golle wrote:
> > Setting ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear
> > routine as done by commit 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc:
> > fix typo in __mtk_foe_entry_clear") breaks flow offloading, at least
> > on older MTK_NETSYS_V1 SoCs, OpenWrt users have confirmed the bug on
> > MT7622 and MT7621 systems.
> > Felix Fietkau suggested to use MTK_FOE_STATE_INVALID instead which
> > works well on both, MTK_NETSYS_V1 and MTK_NETSYS_V2.
> > 
> > Tested on MT7622 (Linksys E8450) and MT7986 (BananaPi BPI-R3).
> > 
> > Suggested-by: Felix Fietkau <nbd@nbd.name>
> > Fixes: 0e80707d94e4c8 ("net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear")
> > Fixes: 33fc42de33278b ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
> > Signed-off-by: Daniel Golle <daniel@makrotopia.org>
> > ---
> > v2: rebased on top of netdev/net.git;main
> 
> FYI and as expected when reading this email thread, we got a small
> conflict when merging -net in net-next in the MPTCP tree due to this
> patch applied in -net:
> 
>   ae3ed15da588 ("net: ethernet: mtk_eth_soc: fix state in
> __mtk_foe_entry_clear")
> 
> and this one from net-next:
> 
>   9d8cb4c096ab ("net: ethernet: mtk_eth_soc: add foe_entry_size to
> mtk_eth_soc")
> 
> The conflict has been resolved on our side[1] inspired by Daniel's v1.
> The resolution we suggest is attached to this email.

conflict resolution: Acked-by: Daniel Golle <daniel@makrotopia.org>

> 
> Cheers,
> Matt
> 
> [1] https://github.com/multipath-tcp/mptcp_net-next/commit/7af5fac658ba
> -- 
> Tessares | Belgium | Hybrid Access Solutions
> www.tessares.net

> diff --cc drivers/net/ethernet/mediatek/mtk_ppe.c
> index 887f430734f7,148ea636ef97..ae00e572390d
> --- a/drivers/net/ethernet/mediatek/mtk_ppe.c
> +++ b/drivers/net/ethernet/mediatek/mtk_ppe.c
> @@@ -439,10 -410,9 +439,10 @@@ __mtk_foe_entry_clear(struct mtk_ppe *p
>   
>   	hlist_del_init(&entry->list);
>   	if (entry->hash != 0xffff) {
>  -		ppe->foe_table[entry->hash].ib1 &= ~MTK_FOE_IB1_STATE;
>  -		ppe->foe_table[entry->hash].ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE,
>  -							      MTK_FOE_STATE_INVALID);
>  +		struct mtk_foe_entry *hwe = mtk_foe_get_entry(ppe, entry->hash);
>  +
>  +		hwe->ib1 &= ~MTK_FOE_IB1_STATE;
> - 		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_UNBIND);
> ++		hwe->ib1 |= FIELD_PREP(MTK_FOE_IB1_STATE, MTK_FOE_STATE_INVALID);
>   		dma_wmb();
>   	}
>   	entry->hash = 0xffff;