Re: smsc911x on Gumstix Overo/Tobi doesn't work

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Am 01.04.2012 21:20, schrieb Javier Martinez Canillas:
> On Fri, Mar 30, 2012 at 5:28 PM, Thomas Klute
> <thomas2.klute@xxxxxxxxxxxxxxx> wrote:
>> Hi,
>>
>> I finally had some time to do more tests on this problem. Findings below.
>>
> 
> Great, I guess we are close to find the issue :)
> 
>> Am 20.03.2012 20:47, schrieb Javier Martinez Canillas:
>>> On Tue, Mar 20, 2012 at 3:27 PM, Thomas Klute
>>> <thomas2.klute@xxxxxxxxxxxxxxx> wrote:
>>>> Am 19.03.2012 23:51, schrieb Tony Lindgren:
>>>>> * Thomas Klute <thomas2.klute@xxxxxxxxxxxxxxx> [120319 09:26]:
>>>>>> Am 16.03.2012 20:33, schrieb Tony Lindgren:
>>>>>>> Hi,
>>>>>>>
>>>>>>> * Thomas Klute <thomas2.klute@xxxxxxxxxxxxxxx> [120316 05:08]:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> I have trouble getting the Ethernet port on a Gumstix Overo with Tobi
>>>>>>>> expansion board to work with current kernel versions. With the latest
>>>>>>>> commit from linux-omap git (b8fe1781ec8bed5e086691a827a6ee11facec2aa),
>>>>>>>> the output from loading the smsc911x driver is as follows:
>>>>>>>>
>>>>>>>> du14:~# modprobe smsc911x
>>>>>>>> [  254.843811] smsc911x: Driver version 2008-10-21
>>>>>>>> [  254.854553] smsc911x: Driver version 2008-10-21
>>>>>>>> [  254.859588] _regulator_get: smsc911x.1 supply vdd33a not found, using
>>>>>>>> dummy regulator
>>>>>>>> [  254.868377] _regulator_get: smsc911x.1 supply vddvario not found,
>>>>>>>> using dummy regulator
>>>>>>>>
>>>>>>>> "ip link show" does not show any available Ethernet port.
>>>>>>>
>>>>>>> The first instance one should work the same way as earlier using
>>>>>>> fixed regulator in gpmc-smsc911x.c. Is it not working for you
>>>>>>> somehow? At least it works for me on zoom3.
>>>>>>
>>>>>> The Tobi board has only one Ethernet port.
>>>>>>
>>>>>>>> I know there has been some trouble with changes around smsc911x
>>>>>>>> regulator support and Gumstix Overo in particular. Am I just missing the
>>>>>>>> right regulator in my kernel config or is this a bug? I can test patches
>>>>>>>> in the latter case.
>>>>>>>
>>>>>>> The second smsc911x now needs a regulator. For multiple smsc911x instances,
>>>>>>> we should change things around so no regulator is created if one
>>>>>>> is passed.
>>>>>>>
>>>>>>> Care to test the following patch by passing a fixed regulator
>>>>>>> from board-overo.c?
>>>>>>
>>>>>> After applying the patch the Ethernet port works consistently once I had
>>>>>> done a cold boot (reboot from the unpatched kernel did not work).
>>>>>> Thank you!
>>>>>
>>>>> Hmm but this patch should not change the behaviour for the first smsc911x
>>>>> instance unless you specify a custom regulator.. Did you patch in a
>>>>> custom regulator, or do we have a bug somewhere? Or do you just need to
>>>>> do a cold reset without the patch I posted?
>>>>
>>>> You're right, during further tests I found that the problem lies
>>>> elsewhere: If the Ethernet cable is attached on modprobe, the device
>>>> works as expected, if not, it's not found (with or without the patch).
>>>> This means if I boot with the cable disconnected, the device won't show
>>>> up, but after
>>>>
>>>> # modprobe -r smsc911x
>>>> [attach cable]
>>>> # modprobe smsc911x
>>>>
>>>> it will work. I'd still consider this a bug, but it doesn't seem to be a
>>>> regulator problem.
>>>>
>>>
>>> Hi Thomas,
>>>
>>> I had the same behavior with the smsc911x chip but on an IGEPv2 board.
>>> The problem was when CONFIG_SMSC_PHY=y since the driver for the chip
>>> internal PHY enables an energy detect power-down mode.
>>>
>>> The smsc911x driver probe function tries to software reset the chip
>>> but if the cable is unplugged the energy detect puts the chip in a low
>>> power mode. Since the chip is not in an operational state the reset
>>> fails and hence the driver probe function. If the cable is plugged
>>> then then energy is detected, the chip is in an operational state and
>>> the reset is successful.
>>>
>>> I sent a patch a few months ago to fix this issue. The patch disables
>>> the energy detect power-down mode before reseting the chip and then it
>>> enables again after reset.
>>>
>>> The commit is:
>>>
>>> commit 6386994e03ebbe60338ded3d586308a41e81c0dc
>>> Author: Javier Martinez Canillas <javier@xxxxxxxxxxxx>
>>> Date:   Tue Jan 3 13:36:19 2012 +0000
>>>
>>>     net/smsc911x: Check if PHY is in operational mode before software reset
>>>
>>> When I fix the issue I only guarded against generation 4 chips (i.e:
>>> pdata->generation == 4), but maybe this problem also exists in other
>>> SMSC chips (I didn't know since I only had access to specific
>>> data-sheets).
>>>
>>> Also you can try enabling debug in the driver by setting USE_DEBUG to
>>> 1 in drivers/net/ethernet/smsc/smsc911x.h and also trying disabling
>>> CONFIG_SMSC_PHY, this will use a generic PHY driver that doesn't put
>>> the chip in auto power mode.
>>
>> After looking at the code I set USE_DEBUG to 3 to get as much
>> information as possible and tested with and without the SMSC PHY driver.
>> Results:
>>
>> With the Ethernet cable attached, the device is properly initialized
>> with and without the PHY driver (as before).
>>
>> Without the cable, the smsc911x driver can initialize the card only if
>> the smsc PHY driver had not been loaded previously. Unloading the PHY
>> driver is not enough, even a reboot doesn't help. I have to do a cold
>> boot (or attach the cable).
>>
> 
> This makes sense since is the PHY driver who enables the auto energy
> detect mode.
> 
>> I guess this confirms Javier's guess, but there's one catch: If you take
>> a look at the attached logs (both without cable attached), you'll see
>> that the device is recognized a generation 4, so the Javier's workaround
>> should actually be used. I added a log output in the if
>> (pdata->generation == 4) block to be sure, and indeed it is used.
>>
>> Any hints why the reset still fails?
>>
>> Regards,
>> Thomas
> 
> My guess is that the PHY chip is still in a low power down and not
> being woken up before the MAC chip is tried to be software reseted. I
> didn't find in the SMSC LAN8700 data-sheet how long it takes to wake
> up the chip and probably it depends of the PHY chip (I only tried with
> the 8700) so if I were you I would try increasing the delay time after
> sending the MII command to disable the auto energy detect mode.
> 
> Can you try this patch?

The longer delay did not help, even with mdelay(1000) the reset still
failed. There's one exception: After a cold boot, the device is
discovered correctly without cable and with the smsc PHY driver loaded.

(Sorry about the delay, this problem isn't really an issue for our
project now that we have a reliable workaround...)

> diff --git a/drivers/net/ethernet/smsc/smsc911x.c
> b/drivers/net/ethernet/smsc/smsc911x.c
> index 24d2df0..d08709a 100644
> --- a/drivers/net/ethernet/smsc/smsc911x.c
> +++ b/drivers/net/ethernet/smsc/smsc911x.c
> @@ -1349,7 +1349,7 @@ static int
> smsc911x_phy_disable_energy_detect(struct smsc911x_data *
>                         return rc;
>                 }
> 
> -               mdelay(1);
> +               mdelay(100);
>         }
> 
>         return 0;
> 
> 
> Best regards,
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [Linux Arm (vger)]     [ARM Kernel]     [ARM MSM]     [Linux Tegra]     [Linux WPAN Networking]     [Linux Wireless Networking]     [Maemo Users]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux