regulator-core has_full_constraints set too late for dt using boards ?

Hi All,

While looking into making regulator_get_optional not print an error when no regulator is found, I've noticed that for dt enabled boards, has_full_constraints is not set until regulator_init_complete() runs, which is a late_initcall.

This means that it gets set after it has already been checked from calls like regulator_get which are likely done by init functions running earlier.

This seems wrong ...

Regards,

Hans

Please look at the current code, that's not how this works any more.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131211/55f674c2/attachment.sig>

Hi,

More current then 3.13-rc3 ? Because that is what I've been looking at.

I could be completely wrong of course. Either way could you give a few hints as to how my interpretation of how this works is wrong ?

Thanks & Regards,

Hans

Yes, -next. It'll be going to Linus probably this week.

It's correct for the code you've been looking at but has since been
changed. Please look at the changes in -next, you should generally
check -next for the current state of things.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131212/b8104725/attachment-0001.sig>

Hi,

Ah yes, I see, thanks for clarifying this.

Note my patch to not do a dev_err when no regulator is found for regulator_get_optional still is needed in -next.

Thanks & Regards,

Hans

It's sitting in my review queue, it's not very well described so it's
not clear that you've selected the correct criteria and I need to find
time to actually look at it properly.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131212/c792fad9/attachment.sig>

Hi,

I did not want to add a third boolean parameter to indicate _regulator_get() was being called from regulator_get_optional.

Since regulator_get_optional is the only one calling _regulator_get() with both allow_dummy and exclusive set to false I've added the following condition to printing the error:

	if (!(!allow_dummy && !exclusive))

Simplified to:

	if (allow_dummy || exclusive)

Which means that it will get printed in all cases it would get printed previously, except when called from regulator_get_optional.

Alternatively a third "optional" boolean parameter could get added to _regulator_get() and the test for printing the error could become:

if (!optional)

Let me know if you would prefer doing things that way and I'll respin the patch.

Thanks & Regards,

Hans

Put some of this analysis in the commit message and perhaps a comment in
the code please...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131212/a3e93479/attachment.sig>