On 7 August 2015 at 19:06, Grygorii Strashko [off-list ref] wrote:

On 08/07/2015 10:11 AM, Tomeu Vizoso wrote:

quoted

On 6 August 2015 at 22:19, Rob Herring [off-list ref] wrote:

quoted

On Thu, Aug 6, 2015 at 9:11 AM, Tomeu Vizoso [off-list ref] wrote:

quoted

Delay matches of platform devices with OF nodes until late_initcall,
when we are sure that all built-in drivers have been registered already.
This is needed to prevent deferred probes because of some drivers not
having registered yet.

The reason why only platform devices are delayed is that some other
devices are expected to be probed earlier than late_initcall, for
example, the system PNP driver needs to probe its devices in
fs_initcall.

Additionally, only platform devices with OF nodes are delayed because
some machines may depend on oter platform devices being registered at
specific times.

How do we know that these probes occur before the unused clocks and
regulators are turned off? Just getting lucky (as is deferred probe)?
Can we do this one level earlier so we have a level left to do things
after probe.

Those are already late_initcall_sync so I guess we're fine.

I wouldn't be so sure :(
FYI:
http://git.ti.com/ti-linux-kernel/ti-linux-kernel/commit/763d643bbfc0f445c6685c541fcae3c370e4314a

If I understand the situation correctly, this is one more instance of
starting to do some work at some point and hoping that something else
that started before has already finished happening. If that's the
case, how does this series make that worst?

During this development I have found many hacks intended to put some
order, even if not enough care was taken to make sure that the order
was guaranteed. In general I would recommend for moving code into
proper drivers and have them to defer the probe of their devices if
some dependency isn't fulfilled at that moment.

Once that's done and we have a safe and reliable boot, we can avoid
those deferred probes by fulfilling the dependency on-demand as this
series shows.

There was some recent thread about how the disabling of unused clocks
and regulators isn't really safe because after late_initcall_sync more
drivers can be registered from modules. Furthermore, there's async
device probes.

Regards,

Tomeu

--
regards,
-grygorii

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel