Re: [PATCH] OMAP3EVM:FIX: Reset the SMSC911x Ethernet controller in board_init

[Kevin Hilman <khilman@ti.com>]

"Hiremath, Vaibhav" <hvaibhav@ti.com> writes:

>> >
>> > With addition of HWMOD support to GPIO, the Ethernet controller
>> > goes undetected for OMAP35xEVM. So explicitely assert the reset signal
>> to
>> > Ethernet controller smsc911x -
>> >
>> > 	- GPIO7 (>=RevG version of EVM's)
>> > 	- GPIO64 (<=RevD version of EVM's)
>> >
>> > I have tested this patch on RevG version of EVM with ES3.1 Si.
>> > This patch is based on intial version from Charulatha V.
>> >
>> > Signed-off-by: Vaibhav Hiremath <hvaibhav@ti.com>
>> 
>> This didn't apply cleanly to l-o master, with or without your previous
>> patches which touch the EVM board file.
>> 
> [Hiremath, Vaibhav] may be order in which you apply the patches is different, but should have been trivial to fix, right? 
> I would suggest you to use newer version.

The order of applying doesn't matter.  Applying the original patch alone
didn't apply, applying it after the other didn't apply.   V2 doesn't
apply cleanly either.

Please be sure it applies cleanly to:

commit 5b698d68c1a9ebeae76a0e4ca4dbfdb10357143d
Author: Thomas Weber <weber@corscience.de>
Date:   Thu Jan 20 14:12:11 2011 +0000

    omap: omap-mcbsp: Fix building after replacement
    
    Fix building omap-mcbsp after replacing ARCH_OMAP24x0 in
    commit 8a9c1aa6a4caa7db1c2fca4b47168af9077e0f95.
 

If your series of EVM patches are interdependent, I suggest sending them
as a series, so it's clear that they are dependent on each other.
Also, please Cc linux-arm-kernel on OMAP patches intended for upstream.

This one (network GPIO) should be separated out though, and go into
.38-rc (as well as -stable) since omap3evm is has not been bootable
since 2.6.37.

>> > ---
>> > NOTE: I have not been able to test it on older version of EVM's.
>> 
>> After manually applying,
>> 
>> Tested-by: Kevin Hilman <khilman@ti.com>
>> 
>> I tested on my rev D board with DHCP + nfs rootfs and it's working well.
>> 
> [Hiremath, Vaibhav] Thanks a lot for validating it.
>
>> While testing though, I also noticed that the smsc driver is dumping
>> some warnings (below) while trying to get the MAC address, resulting in
>> not using the actual MAC but generating a random one.
>> 
>> This isn't related to your patch, since it also happens with l-o master,
>> but was wondering if you saw the same thing?
>> 
> [Hiremath, Vaibhav] Yes I did see this message, and this is coming from 
>
> #ifdef CONFIG_DEBUG_SPINLOCK
> #define SMSC_ASSERT_MAC_LOCK(pdata) \
>                 WARN_ON(!spin_is_locked(&pdata->mac_lock))
>
> And the root-cause is, in __init function we are calling smsc911x_read_mac_address->smsc911x_mac_read without holding mac_lock.
>
> I will submit the patch to net mailing list for this.

Thanks, please Cc linux-omap as well.

Kevin