Hi all,
while debugging network outages on a customers hardware I found, that the MDI-X function of the lan8710 phy seemed to cause trouble.
When connecting to almost any kind of 100/1000MBit switch, the link would seem to come up and data where sent out to the network. But all incoming packets got lost somehow. This is quite bad, as the system runs from nfsroot while booting up during development.
When I disabled the auto MDI-X function of the phy the problem went away.
Signed-off-by: Peter Turczak <[email protected]>
---
drivers/net/phy/Kconfig | 10 ++++++++++
drivers/net/phy/smsc.c | 15 +++++++++++++++
include/linux/smscphy.h | 5 +++++
3 files changed, 30 insertions(+), 0 deletions(-)
diff --git a/drivers/net/phy/Kconfig b/drivers/net/phy/Kconfig
index 961f0b2..341f5aa 100644
--- a/drivers/net/phy/Kconfig
+++ b/drivers/net/phy/Kconfig
@@ -60,6 +60,16 @@ config SMSC_PHY
---help---
Currently supports the LAN83C185, LAN8187 and LAN8700 PHYs
+config SMSC_PHY_DISABLE_AUTOX
+ bool "Disable MDI-X upon start"
+ depends on SMSC_PHY
+ ---help---
+ When you experience problems estabishing a stable connection
+ to a network and you have e.g. a LAN8710 ethernet phy
+ this option might help you out.
+
+ In doubt, say N
+
config BROADCOM_PHY
tristate "Drivers for Broadcom PHYs"
---help---
diff --git a/drivers/net/phy/smsc.c b/drivers/net/phy/smsc.c
index 88e3991..651f71e 100644
--- a/drivers/net/phy/smsc.c
+++ b/drivers/net/phy/smsc.c
@@ -24,6 +24,8 @@
#include <linux/netdevice.h>
#include <linux/smscphy.h>
+static struct phy_driver lan8710_driver;
+
static int smsc_phy_config_intr(struct phy_device *phydev)
{
int rc = phy_write (phydev, MII_LAN83C185_IM,
@@ -53,6 +55,19 @@ static int smsc_phy_config_init(struct phy_device *phydev)
if (rc < 0)
return rc;
+#ifdef CONFIG_SMSC_PHY_DISABLE_AUTOX
+ if (phydev->drv == &lan8710_driver) {
+ rc = phy_read(phydev, MII_LAN8710_SCSI);
+ if (rc < 0)
+ return rc;
+ rc = phy_write(phydev, MII_LAN8710_SCSI,
+ rc | MII_LAN8710_SCSI_AMDIXCTRL);
+
+ if (rc < 0)
+ return rc;
+ }
+#endif
+
return smsc_phy_ack_interrupt (phydev);
}
diff --git a/include/linux/smscphy.h b/include/linux/smscphy.h
index ce718cb..4084b64 100644
--- a/include/linux/smscphy.h
+++ b/include/linux/smscphy.h
@@ -22,4 +22,9 @@
#define MII_LAN83C185_EDPWRDOWN (1 << 13) /* EDPWRDOWN */
#define MII_LAN83C185_ENERGYON (1 << 1) /* ENERGYON */
+#define MII_LAN8710_SCSI 27 /* Special Control/Status register */
+
+#define MII_LAN8710_SCSI_AMDIXCTRL (1<<15) /* Flag to disable Auto-MDIX */
+
+
#endif /* __LINUX_SMSCPHY_H__ */
--
1.7.0.4
On Fri, 30 Nov 2012, Peter Turczak wrote:
> while debugging network outages on a customers hardware I found, that
> the MDI-X function of the lan8710 phy seemed to cause trouble. When
> connecting to almost any kind of 100/1000MBit switch, the link would
> seem to come up and data where sent out to the network. But all incoming
> packets got lost somehow. This is quite bad, as the system runs from
> nfsroot while booting up during development.
>
> When I disabled the auto MDI-X function of the phy the problem went away.
>
> Signed-off-by: Peter Turczak <[email protected]>
> ---
> drivers/net/phy/Kconfig | 10 ++++++++++
> drivers/net/phy/smsc.c | 15 +++++++++++++++
> include/linux/smscphy.h | 5 +++++
> 3 files changed, 30 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/net/phy/Kconfig b/drivers/net/phy/Kconfig
> index 961f0b2..341f5aa 100644
> --- a/drivers/net/phy/Kconfig
> +++ b/drivers/net/phy/Kconfig
> @@ -60,6 +60,16 @@ config SMSC_PHY
> ---help---
> Currently supports the LAN83C185, LAN8187 and LAN8700 PHYs
>
> +config SMSC_PHY_DISABLE_AUTOX
> + bool "Disable MDI-X upon start"
> + depends on SMSC_PHY
> + ---help---
> + When you experience problems estabishing a stable connection
> + to a network and you have e.g. a LAN8710 ethernet phy
> + this option might help you out.
> +
> + In doubt, say N
> +
I am not sure whether compile-time option for something like this is
appropriate. Kernel module parameter, perhaps?
Of course it'd be far better if faulty hardware can be autodetected in
runtime.
--
Jiri Kosina
SUSE Labs
From: Jiri Kosina <[email protected]>
Date: Fri, 30 Nov 2012 16:23:30 +0100 (CET)
> Of course it'd be far better if faulty hardware can be autodetected in
> runtime.
That is how this must be handled.
On Fri, 2012-11-30 at 12:23 -0500, David Miller wrote:
> From: Jiri Kosina <[email protected]>
> Date: Fri, 30 Nov 2012 16:23:30 +0100 (CET)
>
> > Of course it'd be far better if faulty hardware can be autodetected in
> > runtime.
>
> That is how this must be handled.
Well, we also have the recent extension to ETHTOOL_SSET for forcing
MDI-X on or off. If it's not possible to detect the faulty hardware
then the driver can implement this.
Ben.
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
From: Ben Hutchings <[email protected]>
Date: Fri, 30 Nov 2012 19:14:40 +0000
> On Fri, 2012-11-30 at 12:23 -0500, David Miller wrote:
>> From: Jiri Kosina <[email protected]>
>> Date: Fri, 30 Nov 2012 16:23:30 +0100 (CET)
>>
>> > Of course it'd be far better if faulty hardware can be autodetected in
>> > runtime.
>>
>> That is how this must be handled.
>
> Well, we also have the recent extension to ETHTOOL_SSET for forcing
> MDI-X on or off. If it's not possible to detect the faulty hardware
> then the driver can implement this.
Agreed.