Skip to contentSkip to conversation

net: ftgmac100: Fix AST2600EVB NCSI RX issue

PATCH In review Hongwei Zhang 5 participants 15 messages Dec 15, 2020 – Jan 11, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover net: ftgmac100: Fix AST2600EVB NCSI RX issue

Dear Reviewer,

Fix AST2600 EVB NCSI RX timeout issue by removing FTGMAC100_RXDES0_RX_ERR bit from macro RXDES0_ANY_ERROR.

Hongwei Zhang (1):
net: ftgmac100: Fix AST2600 EVB NCSI RX issue

 drivers/net/ethernet/faraday/ftgmac100.h | 1 -
 1 file changed, 1 deletion(-)
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/1 net: ftgmac100: Fix AST2600 EVB NCSI RX issue

Fix AST2600 EVB NCSI RX timeout issue by removing FTGMAC100_RXDES0_RX_ERR bit from macro RXDES0_ANY_ERROR.

---
 drivers/net/ethernet/faraday/ftgmac100.h | 1 -
 1 file changed, 1 deletion(-)
drivers/net/ethernet/faraday/ftgmac100.h +0 -1
index 63b3e02fab16..734477e6f72f 100644
@@ -252,7 +252,6 @@ struct ftgmac100_rxdes {
 
 /* Errors we care about for dropping packets */
 #define RXDES0_ANY_ERROR		( \
-	FTGMAC100_RXDES0_RX_ERR		| \
 	FTGMAC100_RXDES0_CRC_ERR	| \
 	FTGMAC100_RXDES0_FTL		| \
 	FTGMAC100_RXDES0_RUNT		| \
Signature · 1 line
2.17.1
maintainer
#2 lore raw

Re: [Aspeed,ncsi-rx, v1 1/1] net: ftgmac100: Fix AST2600 EVB NCSI RX issue

Quoted 4 lines · On Tue, 15 Dec 2020 14:23:23 -0500 Hongwei Zhang wrote:

Fix AST2600 EVB NCSI RX timeout issue by removing FTGMAC100_RXDES0_RX_ERR bit from macro RXDES0_ANY_ERROR.

Thanks for the patch. Please repost CCing the netdev mailing list so it can be merged to the networking tree (which I assume is your intent). Please also include a Fixes tag pointing to the commit where the timeout issue started (even if it's the first commit of the driver).

#3 lore raw

Re: [Aspeed,ncsi-rx, v1 1/1] net: ftgmac100: Fix AST2600 EVB NCSI RX issue

In reply to Hongwei Zhang
Quoted 2 lines · Hongwei Zhang

Fix AST2600 EVB NCSI RX timeout issue by removing FTGMAC100_RXDES0_RX_ERR bit from macro RXDES0_ANY_ERROR.

But why? What is wrong with the EVB that this change resolves? Which revision of the EVB?

The change affects all designs using the MAC, not just the AST2600 EVB. Why is this patch an appropriate course of action? Can we not add a quirk targeting the specific board (e.g. a devicetree property)?

Andrew

#4 lore raw

cover net: ftgmac100: Fix AST2600EVB NCSI RX issue

In reply to Hongwei Zhang

Dear Reviewer,

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue. One example is Intel I210 Ethernet controller on AST2600 BMC, with FTGMAC100 driver, it always trigger RXDES0_RX_ERR error, cause NCSI initialization failure, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR fix the issue.

Here are part of the debug logs:
......

[   35.075552] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.080843] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 55
[   35.087141] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.094448] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8 
[   35.101498] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.108205] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   35.287808] i2c i2c-1: new_device: Instantiated device slave-mqueue at 0x12
[   35.428379] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.433624] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 56
[   35.439915] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.447225] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.454273] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.460972] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   35.797825] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.803241] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 57
[   35.809541] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.816848] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.823899] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.830597] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   36.179914] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.185160] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 58
[   36.191454] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   36.198761] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   36.205813] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   36.212513] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   36.593688] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.602937] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 59
[   36.609244] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   36.616558] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   36.623608] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   36.630315] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   37.031524] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[   37.067831] IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready

............

This patch add a configurable flag, FTGMAC100_RXDES0_RX_ERR_CHK, in FTGMAC100
driver, it is YES by default, so keep the orignal define of
RXDES0_ANY_ERROR. If it is needed, user can set the flag to NO to remove the RXDES0_RX_ERR bit, to fix the issue.

Hongwei Zhang (1):
net: ftgmac100: Fix AST2600 EVB NCSI RX issue

 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)
Signature · 1 line
2.17.1
#5 lore raw
Quoted 8 lines · Hongwei Zhang

Dear Reviewer,

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue. One example is Intel I210 Ethernet controller on AST2600 BMC, with FTGMAC100 driver, it always trigger RXDES0_RX_ERR error, cause NCSI initialization failure, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR fix the issue.

I work with a few systems that use the i210 on the 2600. We haven't seen this issue in our testing.

Is there something specific about the setup that you use to trigger this?

Ryan, is this an issue that Aspeed is aware of?

Cheers,

Joel

Quoted 53 lines

Here are part of the debug logs:
......

[   35.075552] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.080843] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 55
[   35.087141] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.094448] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.101498] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.108205] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   35.287808] i2c i2c-1: new_device: Instantiated device slave-mqueue at 0x12
[   35.428379] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.433624] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 56
[   35.439915] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.447225] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.454273] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.460972] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   35.797825] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.803241] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 57
[   35.809541] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.816848] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.823899] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.830597] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   36.179914] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.185160] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 58
[   36.191454] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   36.198761] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   36.205813] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   36.212513] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   36.593688] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.602937] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 59
[   36.609244] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   36.616558] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   36.623608] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   36.630315] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   37.031524] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[   37.067831] IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready

............

This patch add a configurable flag, FTGMAC100_RXDES0_RX_ERR_CHK, in FTGMAC100
driver, it is YES by default, so keep the orignal define of
RXDES0_ANY_ERROR. If it is needed, user can set the flag to NO to remove the RXDES0_RX_ERR bit, to fix the issue.

Hongwei Zhang (1):
net: ftgmac100: Fix AST2600 EVB NCSI RX issue

 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)

--
2.17.1

#6 lore raw

Hi Hongwei,

The NCSI should run on 3.3V RMII. According your log, you enabled NCSI on ftgmac100 at 1e660000 which can only support 1.8V I/O voltage. Did you observe the same error on ftgmac100 at 1e670000 (MAC3) or ftgmac100 at 1e690000 (MAC4)?

--
Dylan

Quoted 109 lines

-----Original Message-----
From: Linux-aspeed [mailto:linux-aspeed-bounces+dylan_hung=aspeedtech.com at lists.ozlabs.org] On Behalf Of Joel Stanley
Sent: 2020?12?22? 10:26 AM
To: Hongwei Zhang <[email protected]>; Ryan Chen <[email protected]>
Cc: linux-aspeed <[email protected]>; netdev <[email protected]>; OpenBMC Maillist <[email protected]>; Linux Kernel Mailing List <[email protected]>; Jakub Kicinski <[email protected]>; David S Miller <[email protected]>
Subject: Re: [Aspeed, ncsi-rx, v1 0/1] net: ftgmac100: Fix AST2600EVB NCSI RX issue

Quoted 8 lines · Hongwei Zhang

Dear Reviewer,

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue. One example is Intel I210 Ethernet controller on AST2600 BMC, with FTGMAC100 driver, it always trigger RXDES0_RX_ERR error, cause NCSI initialization failure, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR fix the issue.

I work with a few systems that use the i210 on the 2600. We haven't seen this issue in our testing.

Is there something specific about the setup that you use to trigger this?

Ryan, is this an issue that Aspeed is aware of?

Cheers,

Joel

Quoted 6 lines

Here are part of the debug logs:
......

[   35.075552] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.080843] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 55
[   35.087141] ftgmac100 1e660000.ethernet eth0: rx_packet_error

RXDES0=0xb0070040

Quoted 1 line
[   35.094448] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000

RXDES2=88f8

Quoted 1 line
[   35.101498] ftgmac100 1e660000.ethernet eth0: rx_packet_error

0xb0070040

Quoted 1 line
[   35.108205] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040:

RX_ERR

Quoted 1 line
[   35.287808] i2c i2c-1: new_device: Instantiated device slave-mqueue at

0x12

Quoted 3 lines
[   35.428379] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.433624] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 56
[   35.439915] ftgmac100 1e660000.ethernet eth0: rx_packet_error

RXDES0=0xb0070040

Quoted 1 line
[   35.447225] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000

RXDES2=88f8

Quoted 1 line
[   35.454273] ftgmac100 1e660000.ethernet eth0: rx_packet_error

0xb0070040

Quoted 1 line
[   35.460972] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040:

RX_ERR

Quoted 3 lines
[   35.797825] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.803241] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 57
[   35.809541] ftgmac100 1e660000.ethernet eth0: rx_packet_error

RXDES0=0xb0070040

Quoted 1 line
[   35.816848] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000

RXDES2=88f8

Quoted 1 line
[   35.823899] ftgmac100 1e660000.ethernet eth0: rx_packet_error

0xb0070040

Quoted 1 line
[   35.830597] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040:

RX_ERR

Quoted 3 lines
[   36.179914] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.185160] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 58
[   36.191454] ftgmac100 1e660000.ethernet eth0: rx_packet_error

RXDES0=0xb0070040

Quoted 1 line
[   36.198761] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000

RXDES2=88f8

Quoted 1 line
[   36.205813] ftgmac100 1e660000.ethernet eth0: rx_packet_error

0xb0070040

Quoted 1 line
[   36.212513] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040:

RX_ERR

Quoted 3 lines
[   36.593688] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.602937] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 59
[   36.609244] ftgmac100 1e660000.ethernet eth0: rx_packet_error

RXDES0=0xb0070040

Quoted 1 line
[   36.616558] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000

RXDES2=88f8

Quoted 1 line
[   36.623608] ftgmac100 1e660000.ethernet eth0: rx_packet_error

0xb0070040

Quoted 1 line
[   36.630315] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040:

RX_ERR

Quoted 19 lines
[   37.031524] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[   37.067831] IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready

............

This patch add a configurable flag, FTGMAC100_RXDES0_RX_ERR_CHK, in FTGMAC100 driver, it is YES by default, so keep the orignal define of RXDES0_ANY_ERROR. If it is needed, user can set the flag to NO to remove the RXDES0_RX_ERR bit, to fix the issue.

Hongwei Zhang (1):
net: ftgmac100: Fix AST2600 EVB NCSI RX issue

 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)

--
2.17.1

#7 lore raw

PATCH 1/1 net: ftgmac100: Fix AST2600 EVB NCSI RX issue

In reply to Hongwei Zhang

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR can fix the issue.

Fixes 7ee2d5b4d4340353 ("ARM: dts: nuvoton: Add Fii Kudo system")
---
 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)
drivers/net/ethernet/faraday/Kconfig +9 -0
index c2677ec0564d..ccd0c30be0db 100644
@@ -38,4 +38,13 @@ config FTGMAC100
 	  from Faraday. It is used on Faraday A369, Andes AG102 and some
 	  other ARM/NDS32 SoC's.
 
+config FTGMAC100_RXDES0_RX_ERR_CHK
+	bool "Include FTGMAC100_RXDES0_RX_ERR in RXDES0_ANY_ERROR"
+	default y
+	depends on FTGMAC100
+	help
+	  Say N here if the NCSI controller on your platform has compatible
+	  issue with FTGMAC100, thus always trigger RXDES0_RX_ERR. Exclude
+	  this bit can fix the issue.
+
 endif # NET_VENDOR_FARADAY
drivers/net/ethernet/faraday/ftgmac100.h +8 -0
index 63b3e02fab16..59e1bd52d261 100644
@@ -251,12 +251,20 @@ struct ftgmac100_rxdes {
 #define FTGMAC100_RXDES0_RXPKT_RDY	(1 << 31)
 
 /* Errors we care about for dropping packets */
+#ifdef CONFIG_FTGMAC100_RXDES0_RX_ERR_CHK
 #define RXDES0_ANY_ERROR		( \
 	FTGMAC100_RXDES0_RX_ERR		| \
 	FTGMAC100_RXDES0_CRC_ERR	| \
 	FTGMAC100_RXDES0_FTL		| \
 	FTGMAC100_RXDES0_RUNT		| \
 	FTGMAC100_RXDES0_RX_ODD_NB)
+#else
+#define RXDES0_ANY_ERROR		( \
+	FTGMAC100_RXDES0_CRC_ERR	| \
+	FTGMAC100_RXDES0_FTL		| \
+	FTGMAC100_RXDES0_RUNT		| \
+	FTGMAC100_RXDES0_RX_ODD_NB)
+#endif
 
 #define FTGMAC100_RXDES1_VLANTAG_CI	0xffff
 #define FTGMAC100_RXDES1_PROT_MASK	(0x3 << 20)
Signature · 1 line
2.17.1
#8 lore raw

cover net: ftgmac100: Fix AST2600EVB NCSI RX issue

In reply to Hongwei Zhang

Hi Andrew,

Quoted 10 lines

From: Andrew Jeffery <[email protected]>

Quoted 2 lines

Fix AST2600 EVB NCSI RX timeout issue by removing FTGMAC100_RXDES0_RX_ERR bit from macro RXDES0_ANY_ERROR.

But why? What is wrong with the EVB that this change resolves? Which revision of the EVB?

The change affects all designs using the MAC, not just the AST2600 EVB. Why is this patch an appropriate course of action? Can we not add a quirk targeting the specific board (e.g. a devicetree property)?

You are correct. I was in a rush and didn't put this patch under whole picture of the driver it affected. I revised the cover letter with more information, also use a configurable flag to isolate the change. please review the new approach.

Quoted 3 lines

Andrew

Thanks,
--Hongwei

#9 lore raw

PATCH 1/1 Answer to initial submission

In reply to Hongwei Zhang

Hi Jakub,

Quoted 10 lines

From: Jakub Kicinski <[email protected]>

Quoted 2 lines

...

Thanks for the patch. Please repost CCing the netdev mailing list so it can be merged to the networking tree (which I assume is your intent).
Please also include a Fixes tag pointing to the commit where the timeout issue started (even if it's the first commit of the driver).

I updated the cc list and cover letter accordingly, also addressed Andrew's question. please review.

Thanks,
--Hongwei

Signature · 1 line
2.17.1
#10 lore raw

v2 cover net: ftgmac100: Fix AST2600EVB NCSI RX issue

In reply to Hongwei Zhang

Dear Reviewer,

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue. One example is Intel I210 Ethernet controller on AST2600 BMC, with FTGMAC100 driver, it always trigger RXDES0_RX_ERR error, cause NCSI initialization failure, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR fix the issue.

Here are part of the debug logs:
......

[   35.075552] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.080843] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 55
[   35.087141] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.094448] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8 
[   35.101498] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.108205] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   35.287808] i2c i2c-1: new_device: Instantiated device slave-mqueue at 0x12
[   35.428379] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.433624] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 56
[   35.439915] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.447225] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.454273] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.460972] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   35.797825] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.803241] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 57
[   35.809541] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   35.816848] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   35.823899] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   35.830597] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   36.179914] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.185160] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 58
[   36.191454] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   36.198761] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   36.205813] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   36.212513] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   36.593688] ftgmac100_hard_start_xmit TXDESO=b000003c
[   36.602937] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 59
[   36.609244] ftgmac100 1e660000.ethernet eth0: rx_packet_error RXDES0=0xb0070040
[   36.616558] ftgmac100_rx_packet RXDES0=b0070040 RXDES1=f0800000 RXDES2=88f8
[   36.623608] ftgmac100 1e660000.ethernet eth0: rx_packet_error 0xb0070040
[   36.630315] ftgmac100 1e660000.ethernet eth0: [ISR] = 0xb0070040: RX_ERR
[   37.031524] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[   37.067831] IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready

............

This patch add a configurable flag, FTGMAC100_RXDES0_RX_ERR_CHK, in FTGMAC100
driver, it is YES by default, so keep the orignal define of
RXDES0_ANY_ERROR. If it is needed, user can set the flag to NO to remove the RXDES0_RX_ERR bit, to fix the issue.

Hongwei Zhang (1):
net: ftgmac100: Fix AST2600 EVB NCSI RX issue

 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)
Signature · 1 line
2.17.1
#11 lore raw

PATCH v2 1/1 net: ftgmac100: Fix AST2600 EVB NCSI RX issue

In reply to Hongwei Zhang

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR can fix the issue.

Fixes 7ee2d5b4d4340353 ("ARM: dts: nuvoton: Add Fii Kudo system")
---
 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)
drivers/net/ethernet/faraday/Kconfig +9 -0
index c2677ec0564d..ccd0c30be0db 100644
@@ -38,4 +38,13 @@ config FTGMAC100
 	  from Faraday. It is used on Faraday A369, Andes AG102 and some
 	  other ARM/NDS32 SoC's.
 
+config FTGMAC100_RXDES0_RX_ERR_CHK
+	bool "Include FTGMAC100_RXDES0_RX_ERR in RXDES0_ANY_ERROR"
+	default y
+	depends on FTGMAC100
+	help
+	  Say N here if the NCSI controller on your platform has compatible
+	  issue with FTGMAC100, thus always trigger RXDES0_RX_ERR. Exclude
+	  this bit can fix the issue.
+
 endif # NET_VENDOR_FARADAY
drivers/net/ethernet/faraday/ftgmac100.h +8 -0
index 63b3e02fab16..59e1bd52d261 100644
@@ -251,12 +251,20 @@ struct ftgmac100_rxdes {
 #define FTGMAC100_RXDES0_RXPKT_RDY	(1 << 31)
 
 /* Errors we care about for dropping packets */
+#ifdef CONFIG_FTGMAC100_RXDES0_RX_ERR_CHK
 #define RXDES0_ANY_ERROR		( \
 	FTGMAC100_RXDES0_RX_ERR		| \
 	FTGMAC100_RXDES0_CRC_ERR	| \
 	FTGMAC100_RXDES0_FTL		| \
 	FTGMAC100_RXDES0_RUNT		| \
 	FTGMAC100_RXDES0_RX_ODD_NB)
+#else
+#define RXDES0_ANY_ERROR		( \
+	FTGMAC100_RXDES0_CRC_ERR	| \
+	FTGMAC100_RXDES0_FTL		| \
+	FTGMAC100_RXDES0_RUNT		| \
+	FTGMAC100_RXDES0_RX_ODD_NB)
+#endif
 
 #define FTGMAC100_RXDES1_VLANTAG_CI	0xffff
 #define FTGMAC100_RXDES1_PROT_MASK	(0x3 << 20)
Signature · 1 line
2.17.1
maintainer
#12 lore raw

Re: [Aspeed,ncsi-rx, v2 1/1] net: ftgmac100: Fix AST2600 EVB NCSI RX issue

Quoted 1 line · On Mon, 21 Dec 2020 14:40:26 -0500 Hongwei Zhang wrote:

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few

When you say NCSI Ethernet controller here you mean the main system NIC, right? The MAC on the NCSI side is FTGMAC100, correct?

In that case I'm not sure how user is supposed to control this setting at build time. The system NIC is often pluggable on the PCIe bus, and can be changed at will.

Quoted 4 lines

controllers have compatible issue, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR can fix the issue.

Fixes 7ee2d5b4d4340353 ("ARM: dts: nuvoton: Add Fii Kudo system")

Please fix the commit hash, this hash does not exist upstream:

Commit: 8711d4ef64fa ("net: ftgmac100: Fix AST2600 EVB NCSI RX issue")
Fixes tag: Fixes: 7ee2d5b4d4340353 ("ARM: dts: nuvoton: Add Fii Kudo system") Has these problem(s):
- Target SHA1 does not exist

Quoted 49 lines
---
 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)
drivers/net/ethernet/faraday/Kconfig +9 -0
index c2677ec0564d..ccd0c30be0db 100644
@@ -38,4 +38,13 @@ config FTGMAC100
 	  from Faraday. It is used on Faraday A369, Andes AG102 and some
 	  other ARM/NDS32 SoC's.
 
+config FTGMAC100_RXDES0_RX_ERR_CHK
+	bool "Include FTGMAC100_RXDES0_RX_ERR in RXDES0_ANY_ERROR"
+	default y
+	depends on FTGMAC100
+	help
+	  Say N here if the NCSI controller on your platform has compatible
+	  issue with FTGMAC100, thus always trigger RXDES0_RX_ERR. Exclude
+	  this bit can fix the issue.
+
 endif # NET_VENDOR_FARADAY
drivers/net/ethernet/faraday/ftgmac100.h +8 -0
index 63b3e02fab16..59e1bd52d261 100644
@@ -251,12 +251,20 @@ struct ftgmac100_rxdes {
 #define FTGMAC100_RXDES0_RXPKT_RDY	(1 << 31)
 
 /* Errors we care about for dropping packets */
+#ifdef CONFIG_FTGMAC100_RXDES0_RX_ERR_CHK
 #define RXDES0_ANY_ERROR		( \
 	FTGMAC100_RXDES0_RX_ERR		| \
 	FTGMAC100_RXDES0_CRC_ERR	| \
 	FTGMAC100_RXDES0_FTL		| \
 	FTGMAC100_RXDES0_RUNT		| \
 	FTGMAC100_RXDES0_RX_ODD_NB)
+#else
+#define RXDES0_ANY_ERROR		( \
+	FTGMAC100_RXDES0_CRC_ERR	| \
+	FTGMAC100_RXDES0_FTL		| \
+	FTGMAC100_RXDES0_RUNT		| \
+	FTGMAC100_RXDES0_RX_ODD_NB)
+#endif
 
 #define FTGMAC100_RXDES1_VLANTAG_CI	0xffff
 #define FTGMAC100_RXDES1_PROT_MASK	(0x3 << 20)
#13 lore raw

cover net: ftgmac100: Fix AST2600EVB NCSI RX issue

In reply to Hongwei Zhang
Quoted 7 lines

Hi Hongwei,

The NCSI should run on 3.3V RMII. According your log, you enabled NCSI on ftgmac100 at 1e660000 which can only support 1.8V I/O voltage.
Did you observe the same error on ftgmac100 at 1e670000 (MAC3) or ftgmac100 at 1e690000 (MAC4)?

Hi Dylan,

Thanks for your review and input, you're correct, this issue is not observed on AST2600 MAC4 (ftgmac100 at 1e690000).

Though this issue is caused by using NCSI incompatible MAC ftgmac100 at 1e660000, we thought this patch is still having value, by providing an extra option to user to be able to use ftgmac100 at 1e660000 for NCSI, and this is also true for AST2500.

--Hongwei

Quoted 28 lines
Quoted 28 lines

-----Original Message-----
From: Linux-aspeed [mailto:linux-aspeed-bounces+dylan_hung=aspeedtech.com at lists.ozlabs.or g]
On Behalf Of Joel Stanley
Sent: 2020?12?22? 10:26 AM
To: Hongwei Zhang <[email protected]>; Ryan Chen <[email protected]>
Cc: linux-aspeed <[email protected]>; netdev <[email protected]>; OpenBMC Maillist <[email protected]>; Linux Kernel Mailing List <[email protected]>; Jakub Kicinski <[email protected]>; David S Miller <[email protected]>
Subject: Re: [Aspeed, ncsi-rx, v1 0/1] net: ftgmac100: Fix AST2600EVB NCSI RX issue

Quoted 8 lines · Hongwei Zhang

Dear Reviewer,

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few controllers have compatible issue. One example is Intel I210 Ethernet controller on AST2600 BMC, with FTGMAC100 driver, it always trigger RXDES0_RX_ERR error, cause NCSI initialization failure, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR fix the issue.

I work with a few systems that use the i210 on the 2600. We haven't seen this issue in our testing.

Quoted 7 lines
Quoted 7 lines

Is there something specific about the setup that you use to trigger this?

Ryan, is this an issue that Aspeed is aware of?

Cheers,

Joel

Hello Joel,

Thanks for your review, please see my response to Dylan, he pointed out the root cause of the issue.

-- Hongwei

Quoted 27 lines
Quoted 26 lines
Quoted 7 lines

Here are part of the debug logs:
......

[   35.075552] ftgmac100_hard_start_xmit TXDESO=b000003c
[   35.080843] ftgmac100 1e660000.ethernet eth0: tx_complete_packet 55
[   35.087141] ftgmac100 1e660000.ethernet eth0: rx_packet_error

RXDES0=0xb0070040

Quoted 18 lines
[   37.067831] IPv6: ADDRCONF(NETDEV_UP): eth1: link is not ready

............

This patch add a configurable flag, FTGMAC100_RXDES0_RX_ERR_CHK, in FTGMAC100 driver, it is YES by default, so keep the orignal define of RXDES0_ANY_ERROR. If it is needed, user can set the flag to NO to remove the RXDES0_RX_ERR bit, to fix the issue.

Hongwei Zhang (1):
net: ftgmac100: Fix AST2600 EVB NCSI RX issue

 drivers/net/ethernet/faraday/Kconfig     | 9 +++++++++
 drivers/net/ethernet/faraday/ftgmac100.h | 8 ++++++++
 2 files changed, 17 insertions(+)

--
2.17.1

#14 lore raw

PATCH v2 1/1 net: ftgmac100: Fix AST2600EVB NCSI RX issue

In reply to Hongwei Zhang
Quoted 11 lines

From: Jakub Kicinski <[email protected]>
Sent: Monday, December 21, 2020 5:10 PM
To: Hongwei Zhang

Quoted 1 line · On Mon, 21 Dec 2020 14:40:26 -0500 Hongwei Zhang wrote:

When FTGMAC100 driver is used on other NCSI Ethernet controllers, few

When you say NCSI Ethernet controller here you mean the main system NIC, right? The MAC on the NCSI side is FTGMAC100, correct?

Hi Jakub,

The use case for us is the MAC is configured as NCSI, so it provides network
access for both BMC and Host, the Ethernet controller driver is from BMC kernel
side.

please see my response to Dylan in another thread, <[email protected]>, he points out the root cause of the issue.

Quoted 10 lines

In that case I'm not sure how user is supposed to control this setting at build time. The system NIC is often pluggable on the PCIe bus, and can be changed at will.

Quoted 4 lines

controllers have compatible issue, removing FTGMAC100_RXDES0_RX_ERR bit from RXDES0_ANY_ERROR can fix the issue.

Fixes 7ee2d5b4d4340353 ("ARM: dts: nuvoton: Add Fii Kudo system")

Please fix the commit hash, this hash does not exist upstream:

will do.

Thanks
--Hongwei

Quoted 5 lines

Commit: 8711d4ef64fa ("net: ftgmac100: Fix AST2600 EVB NCSI RX issue")
Fixes tag: Fixes: 7ee2d5b4d4340353 ("ARM: dts: nuvoton: Add Fii Kudo system") Has these problem(s):
- Target SHA1 does not exist