Patchwork phylib: Fix Freescale TBI PHY detection

login
register
mail settings
Submitter Anton Vorontsov
Date Jan. 13, 2009, 4:05 p.m.
Message ID <20090113160513.GA22083@oksana.dev.rtsoft.ru>
Download mbox | patch
Permalink /patch/18255/
State Accepted, archived
Delegated to: Kumar Gala
Headers show

Comments

Anton Vorontsov - Jan. 13, 2009, 4:05 p.m.
Freescale on-chip TBI PHYs reports PHY ID as 0x0, but as of

commit 3ee82383f0098a2e13acc8cf1be8e47512f41e5a
Author: Giulio Benetti <giulio.benetti@micronovasrl.com>
Date:   Thu Nov 13 21:53:13 2008 +0000

    phy: fix phy address bug

    PHYID returns 0xffff and not 0xffffffff when not found and in some
    case(at91sam9263) 0x0. Maybe this patch could be useful.

phy_device.c treats PHY ID == 0x0 as bogus IDs, and that results in
gianfar driver failure to see the TBI PHYs. This code snippet triggers:

	if (!priv->tbiphy) {
		printk(KERN_WARNING "SGMII mode requires that the device "
				"tree specify a tbi-handle\n");
		return;
	}

Although tbi-handle is specified in the device tree.

Btw, technically PHY ID == 0x0 is a valid ID (if we ever see a PHY
manufactured by Xerox :-).

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
---

There is one thing I don't actually understand though...

Andy, were you testing the TBI support on a hardware where PHY ID
!= 0x0 or maybe your TBI PHY support patch (commit b31a1d8b41513b,
dated Tue Dec 16 15:29:15 2008) was based on a bit outdated kernel
version? Because according to the git timestamps, the TBI support
was not working since the submission.

Just in case, the hardware I'm seeing the PHY ID == 0x0 is
MPC8378E-MDS.

Thanks,

 drivers/net/phy/phy_device.c |    9 ---------
 1 files changed, 0 insertions(+), 9 deletions(-)
Anton Vorontsov - Jan. 14, 2009, 3:03 p.m.
On Tue, Jan 13, 2009 at 07:05:13PM +0300, Anton Vorontsov wrote:
> Freescale on-chip TBI PHYs reports PHY ID as 0x0, but as of
> 
> commit 3ee82383f0098a2e13acc8cf1be8e47512f41e5a
> Author: Giulio Benetti <giulio.benetti@micronovasrl.com>
> Date:   Thu Nov 13 21:53:13 2008 +0000
> 
>     phy: fix phy address bug
> 
>     PHYID returns 0xffff and not 0xffffffff when not found and in some
>     case(at91sam9263) 0x0. Maybe this patch could be useful.
> 
> phy_device.c treats PHY ID == 0x0 as bogus IDs, and that results in
> gianfar driver failure to see the TBI PHYs. This code snippet triggers:
> 
> 	if (!priv->tbiphy) {
> 		printk(KERN_WARNING "SGMII mode requires that the device "
> 				"tree specify a tbi-handle\n");
> 		return;
> 	}
> 
> Although tbi-handle is specified in the device tree.
> 
> Btw, technically PHY ID == 0x0 is a valid ID (if we ever see a PHY
> manufactured by Xerox :-).
> 
> Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
> ---
> 
> There is one thing I don't actually understand though...
> 
> Andy, were you testing the TBI support on a hardware where PHY ID
> != 0x0 or maybe your TBI PHY support patch (commit b31a1d8b41513b,
> dated Tue Dec 16 15:29:15 2008) was based on a bit outdated kernel
> version? Because according to the git timestamps, the TBI support
> was not working since the submission.
> 
> Just in case, the hardware I'm seeing the PHY ID == 0x0 is
> MPC8378E-MDS.

I think I got it. Probably the TBI support patch was based on the
powerpc.git next, and the commit that broke the TBI support
was in the net-next-2.6 tree.

That explains why nobody noticed the issue.

>  drivers/net/phy/phy_device.c |    9 ---------
>  1 files changed, 0 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
> index e354601..0a06e4f 100644
> --- a/drivers/net/phy/phy_device.c
> +++ b/drivers/net/phy/phy_device.c
> @@ -231,15 +231,6 @@ struct phy_device * get_phy_device(struct mii_bus *bus, int addr)
>  	if ((phy_id & 0x1fffffff) == 0x1fffffff)
>  		return NULL;
>  
> -	/*
> -	 * Broken hardware is sometimes missing the pull-up resistor on the
> -	 * MDIO line, which results in reads to non-existent devices returning
> -	 * 0 rather than 0xffff. Catch this here and treat 0 as a non-existent
> -	 * device as well.
> -	 */
> -	if (phy_id == 0)
> -		return NULL;
> -
>  	dev = phy_device_create(bus, addr, phy_id);
>  
>  	return dev;
> -- 
> 1.5.6.5
Andy Fleming - Jan. 14, 2009, 6:20 p.m.
On Jan 14, 2009, at 9:03 AM, Anton Vorontsov wrote:
>>
>> There is one thing I don't actually understand though...
>>
>> Andy, were you testing the TBI support on a hardware where PHY ID
>> != 0x0 or maybe your TBI PHY support patch (commit b31a1d8b41513b,
>> dated Tue Dec 16 15:29:15 2008) was based on a bit outdated kernel
>> version? Because according to the git timestamps, the TBI support
>> was not working since the submission.
>>
>> Just in case, the hardware I'm seeing the PHY ID == 0x0 is
>> MPC8378E-MDS.
>
> I think I got it. Probably the TBI support patch was based on the
> powerpc.git next, and the commit that broke the TBI support
> was in the net-next-2.6 tree.
>
> That explains why nobody noticed the issue.


Yeah, I dropped the ball.  I saw the patch go in, thought that might  
break something, but I didn't find time to look into it.  Thanks for  
finding and reverting this bug.

Acked-by: Andy Fleming <afleming@freescale.com>
David Miller - Jan. 14, 2009, 10:38 p.m.
From: Andy Fleming <afleming@freescale.com>
Date: Wed, 14 Jan 2009 12:20:35 -0600

> On Jan 14, 2009, at 9:03 AM, Anton Vorontsov wrote:
> >>
> >> There is one thing I don't actually understand though...
> >>
> >> Andy, were you testing the TBI support on a hardware where PHY ID
> >> != 0x0 or maybe your TBI PHY support patch (commit b31a1d8b41513b,
> >> dated Tue Dec 16 15:29:15 2008) was based on a bit outdated kernel
> >> version? Because according to the git timestamps, the TBI support
> >> was not working since the submission.
> >>
> >> Just in case, the hardware I'm seeing the PHY ID == 0x0 is
> >> MPC8378E-MDS.
> >
> > I think I got it. Probably the TBI support patch was based on the
> > powerpc.git next, and the commit that broke the TBI support
> > was in the net-next-2.6 tree.
> >
> > That explains why nobody noticed the issue.
> 
> 
> Yeah, I dropped the ball.  I saw the patch go in, thought that might break something, but I didn't find time to look into it.  Thanks for finding and reverting this bug.
> 
> Acked-by: Andy Fleming <afleming@freescale.com>

Patch applied, thanks everyone.

I was worried when I applied the patch causing this, that some
device would in fact trigger that specific test.  Turns out
my worries were warranted :)

Patch

diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
index e354601..0a06e4f 100644
--- a/drivers/net/phy/phy_device.c
+++ b/drivers/net/phy/phy_device.c
@@ -231,15 +231,6 @@  struct phy_device * get_phy_device(struct mii_bus *bus, int addr)
 	if ((phy_id & 0x1fffffff) == 0x1fffffff)
 		return NULL;
 
-	/*
-	 * Broken hardware is sometimes missing the pull-up resistor on the
-	 * MDIO line, which results in reads to non-existent devices returning
-	 * 0 rather than 0xffff. Catch this here and treat 0 as a non-existent
-	 * device as well.
-	 */
-	if (phy_id == 0)
-		return NULL;
-
 	dev = phy_device_create(bus, addr, phy_id);
 
 	return dev;