[BUG] 2.6.29-rc7 broke r8169 MAC on Thecus n2100 ARM board

[BUG] 2.6.29-rc7 broke r8169 MAC on Thecus n2100 ARM board

[Mikael Pettersson]

With the 2.6.29-rc7 kernel I'm unable to 'ifup' eth0 on my ARM-based
Thecus n2100 NAS box, which has an RTL8169 NIC (10ec:8169 (rev 10)).
A diff of the kernel logs from 2.6.29-rc6 and -rc7 shows that -rc7
first mangles and then zaps the MAC addresses:

--- dmesg-2.6.29-rc6	2009-03-09 20:27:49.000000000 +0100
+++ dmesg-2.6.29-rc7	2009-03-09 20:25:54.000000000 +0100
@@ -145,11 +145,13 @@
 r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
 r8169 0000:00:01.0: enabling Mem-Wr-Inval
 r8169 0000:00:01.0: no PCI Express capability
-eth0: RTL8169sb/8110sb at 0xe080c200, 00:14:fd:30:11:78, XID 10000000 IRQ 27
+r8169 0000:00:01.0: MAC address found in EEPROM: 00:14:fd:10:00:00
+eth0: RTL8169sb/8110sb at 0xe080c200, 00:00:00:00:00:00, XID 10000000 IRQ 27
 r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
 r8169 0000:00:02.0: enabling Mem-Wr-Inval
 r8169 0000:00:02.0: no PCI Express capability
-eth1: RTL8169sb/8110sb at 0xe0810300, 00:14:fd:30:11:79, XID 10000000 IRQ 30
+r8169 0000:00:02.0: MAC address found in EEPROM: 00:14:fd:10:00:00
+eth1: RTL8169sb/8110sb at 0xe0810300, 00:00:00:00:00:00, XID 10000000 IRQ 30
 Driver 'sd' needs updating - please use bus_type methods
 sata_sil 0000:00:03.0: version 2.4
 sata_sil 0000:00:03.0: Applying R_ERR on DMA activate FIS errata fix
@@ -242,6 +244,4 @@
 EXT3 FS on sda5, internal journal
 EXT3-fs: mounted filesystem with ordered data mode.
 Adding 2008116k swap on /dev/sda3.  Priority:-1 extents:1 across:2008116k 
-r8169: eth0: link up
-r8169: eth0: link up
 warning: `dbus-daemon' uses 32-bit capabilities (legacy support in use)

Reverting 6709fe9a27e43a4931938fe0d7f2cc5edef31386
"r8169: read MAC address from EEPROM on init (2nd attempt)"
fixes the problem on this machine.

/Mikael

Re: [BUG] 2.6.29-rc7 broke r8169 MAC on Thecus n2100 ARM board

[Francois Romieu]

[-- Attachment #1: Type: text/plain, Size: 483 bytes --]

Mikael Pettersson <mikpe@it.uu.se> :
> With the 2.6.29-rc7 kernel I'm unable to 'ifup' eth0 on my ARM-based
> Thecus n2100 NAS box, which has an RTL8169 NIC (10ec:8169 (rev 10)).
> A diff of the kernel logs from 2.6.29-rc6 and -rc7 shows that -rc7
> first mangles and then zaps the MAC addresses:

Can you apply the attached patch and see if it fixes any of these
problems ?

Tim, could you give it a try too ? It applies with "patch -p1" from
the kernel tree directory.

-- 
Ueimor

[-- Attachment #2: r8169-init-mac-20.patch --]
[-- Type: text/plain, Size: 2081 bytes --]

diff --git a/drivers/net/r8169.c b/drivers/net/r8169.c
index b347340..5d2ae6b 100644
--- a/drivers/net/r8169.c
+++ b/drivers/net/r8169.c
@@ -1833,7 +1833,13 @@ static void rtl_rar_set(struct rtl8169_private *tp, u8 *addr)
 	RTL_W32(MAC4, high);
 	RTL_W8(Cfg9346, Cfg9346_Lock);
 
+	/* PCI commit */
+	RTL_R8(ChipCmd);
+
 	spin_unlock_irq(&tp->lock);
+
+	/* Hack. */
+	msleep_interruptible(1);
 }
 
 static int rtl_set_mac_address(struct net_device *dev, void *p)
@@ -1970,9 +1976,14 @@ static const struct net_device_ops rtl8169_netdev_ops = {
 };
 
 /* Delay between EEPROM clock transitions. Force out buffered PCI writes. */
-#define RTL_EEPROM_DELAY()	RTL_R8(Cfg9346)
 #define RTL_EEPROM_READ_CMD	6
 
+static void rtl_eeprom_delay(void __iomem *ioaddr)
+{
+	RTL_R8(Cfg9346);
+	ndelay(450);
+}
+
 /* read 16bit word stored in EEPROM. EEPROM is addressed by words. */
 static u16 rtl_eeprom_read(void __iomem *ioaddr, int addr)
 {
@@ -1992,7 +2003,7 @@ static u16 rtl_eeprom_read(void __iomem *ioaddr, int addr)
 
 	/* enter programming mode */
 	RTL_W8(Cfg9346, Cfg9346_Program | Cfg9346_EECS);
-	RTL_EEPROM_DELAY();
+	rtl_eeprom_delay(ioaddr);
 
 	/* write command and requested address */
 	while (cmd_len--) {
@@ -2002,29 +2013,29 @@ static u16 rtl_eeprom_read(void __iomem *ioaddr, int addr)
 
 		/* write a bit */
 		RTL_W8(Cfg9346, x);
-		RTL_EEPROM_DELAY();
+		rtl_eeprom_delay(ioaddr);
 
 		/* raise clock */
 		RTL_W8(Cfg9346, x | Cfg9346_EESK);
-		RTL_EEPROM_DELAY();
+		rtl_eeprom_delay(ioaddr);
 	}
 
 	/* lower clock */
 	RTL_W8(Cfg9346, Cfg9346_Program | Cfg9346_EECS);
-	RTL_EEPROM_DELAY();
+	rtl_eeprom_delay(ioaddr);
 
 	/* read back 16bit value */
 	for (i = 16; i > 0; i--) {
 		/* raise clock */
 		RTL_W8(Cfg9346, Cfg9346_Program | Cfg9346_EECS | Cfg9346_EESK);
-		RTL_EEPROM_DELAY();
+		rtl_eeprom_delay(ioaddr);
 
 		result <<= 1;
 		result |= (RTL_R8(Cfg9346) & Cfg9346_EEDO) ? 1 : 0;
 
 		/* lower clock */
 		RTL_W8(Cfg9346, Cfg9346_Program | Cfg9346_EECS);
-		RTL_EEPROM_DELAY();
+		rtl_eeprom_delay(ioaddr);
 	}
 
 	RTL_W8(Cfg9346, Cfg9346_Program);

Re: [BUG] 2.6.29-rc7 broke r8169 MAC on Thecus n2100 ARM board

[Tim Durack]

On Fri, Mar 13, 2009 at 4:15 PM, Francois Romieu <romieu@fr.zoreil.com> wrote:
> Mikael Pettersson <mikpe@it.uu.se> :
>> With the 2.6.29-rc7 kernel I'm unable to 'ifup' eth0 on my ARM-based
>> Thecus n2100 NAS box, which has an RTL8169 NIC (10ec:8169 (rev 10)).
>> A diff of the kernel logs from 2.6.29-rc6 and -rc7 shows that -rc7
>> first mangles and then zaps the MAC addresses:
>
> Can you apply the attached patch and see if it fixes any of these
> problems ?
>
> Tim, could you give it a try too ? It applies with "patch -p1" from
> the kernel tree directory.
>
> --
> Ueimor
>

No luck:

B-25-1NOC-CFG-1:~# modprobe -if r8169
[  120.938729] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[  120.945018] r8169 0000:00:09.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18
[  120.953409] r8169 0000:00:09.0: no PCI Express capability
[  120.960359] r8169 0000:00:09.0: MAC address found in EEPROM:
00:30:18:ab:f0:34
[  120.978441] eth1: RTL8169sc/8110sc at 0xf8154000,
00:00:00:00:f0:34, XID 18000000 IRQ 18
[  120.988468] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[  120.994119] r8169 0000:00:0b.0: PCI INT A -> GSI 19 (level, low) -> IRQ 19
[  121.001139] r8169 0000:00:0b.0: no PCI Express capability
[  121.006922] r8169 0000:00:0b.0: MAC address found in EEPROM:
00:30:18:ab:f0:35
[  121.063615] eth2: RTL8169sc/8110sc at 0xf801e000,
00:00:00:00:f0:35, XID 18000000 IRQ 19
[  121.072495] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[  121.078188] r8169 0000:00:0c.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[  121.085239] r8169 0000:00:0c.0: no PCI Express capability
[  121.091025] r8169 0000:00:0c.0: MAC address found in EEPROM:
00:30:18:ab:f0:36
[  121.147502] eth3: RTL8169sc/8110sc at 0xf8022000,
00:00:00:00:f0:36, XID 18000000 IRQ 16

Tim:>

Re: [BUG] 2.6.29-rc7 broke r8169 MAC on Thecus n2100 ARM board

[Mikael Pettersson]

Tim Durack writes:
 > On Fri, Mar 13, 2009 at 4:15 PM, Francois Romieu <romieu@fr.zoreil.com>wrote:
 > 
 > > Mikael Pettersson <mikpe@it.uu.se> :
 > > > With the 2.6.29-rc7 kernel I'm unable to 'ifup' eth0 on my ARM-based
 > > > Thecus n2100 NAS box, which has an RTL8169 NIC (10ec:8169 (rev 10)).
 > > > A diff of the kernel logs from 2.6.29-rc6 and -rc7 shows that -rc7
 > > > first mangles and then zaps the MAC addresses:
 > >
 > > Can you apply the attached patch and see if it fixes any of these
 > > problems ?
 > >
 > > Tim, could you give it a try too ? It applies with "patch -p1" from
 > > the kernel tree directory.
 > >
 > > --
 > > Ueimor
 > >
 > 
 > No luck:

No joy here either. 2.6.29-rc8 + this patch shows the same misbehaviour
as 2.6.29-rc7 had. Here's a minicom log of an attempted boot:

(kernel messages)

r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
r8169 0000:00:01.0: no PCI Express capability
r8169 0000:00:01.0: MAC address found in EEPROM: 00:14:fd:10:00:00
eth0: RTL8169sb/8110sb at 0xe080c200, 00:00:00:00:00:00, XID 10000000 IRQ 27
r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
r8169 0000:00:02.0: no PCI Express capability
r8169 0000:00:02.0: MAC address found in EEPROM: 00:14:fd:10:00:00
eth1: RTL8169sb/8110sb at 0xe0810300, 00:00:00:00:00:00, XID 10000000 IRQ 30

(init messages)

Bringing up interface eth0:  RTNETLINK answers: Cannot assign requested address
Failed to bring up eth0.

May I suggest that 6709fe9a27e43a4931938fe0d7f2cc5edef31386 is reverted
until this regression is debugged and fixed?

/Mikael

Re: [BUG] 2.6.29-rc7 broke r8169 MAC on Thecus n2100 ARM board

[Francois Romieu]

Mikael Pettersson <mikpe@it.uu.se> :
[...]
> May I suggest that 6709fe9a27e43a4931938fe0d7f2cc5edef31386 is reverted
> until this regression is debugged and fixed?

Yes.

-- 
Ueimor