On Tue, Jan 15, 2019 at 11:08 AM Martin Fuzzey martin.fuzzey@flowbird.group wrote:
Hi Adam,
On 13/01/2019 15:00, Adam Ford wrote:
On Thu, Oct 4, 2018 at 12:55 PM Martin Fuzzey martin.fuzzey@flowbird.group wrote:
Configure the phy regulator if defined by the "phy-supply" DT phandle.
Signed-off-by: Martin Fuzzey martin.fuzzey@flowbird.group
This patch seems to break the Ethernet on my board, but I think I have a possible solution (see below)
Ah you are right my bad.
I guess this means that you already had the phy-supply property in your DT but unused up till now.
I haven't had DM_REGULATOR turned on until recently which is how I found this bug.
I have a board that uses a fixed regulator driven by a GPIO. It's neither always-on, nor it is enabled on boot and it doesn't have a specified current setting. With DM_REGULATOR set, regulator_autoset fails and FEC doesn't come up. Looking at a bunch of other drivers, and how they enable their respective regulators, they're using regulator_set_enable instead of autoset.
Indeed, on my board I am setting the voltage and do have regulator-boot-on;,
which is why I didn't run into this issue.
Incidentally it looks like if you need to set the voltage you *need* regulator-boot-on, which means the regulator will always be enabled too.
There seems to be no way to say "please set this voltage but only switch the regulator on if a driver requests it".
Is there a reason we couldn't use
ret = regulator_set_enable(priv->phy_supply, true);That looks good and tested ok on my board too.
good.
Since my patch has now been merged it needs a follow up patch to replace regulator_autoset() with regulator_set_enable().
I agree.
I'll send a fix patch soon unless you want to do it as the finder of the problem.
I have a patch ready to go. I can just push it. I'll CC you on the e-mail and you can reply as tested-by or reviewed-by or whatever, if you're open to that.
adam
Regards,
Martin