On Tue, 2018-02-20 at 17:27 +0100, Rafael J. Wysocki wrote: > On Tue, Feb 20, 2018 at 5:03 PM, Andy Shevchenko > <andriy.shevche...@linux.intel.com> wrote: > > Some platforms might take care of legacy devices on theirs own. Due > > to this, > > export acpi_reduced_hw_init() and put it into struct x86_init_acpi. > > IMO this completely doesn't explain what really happens here. > > You basically want to provide your own versions of > x86_init.timers.timer_init, x86_init.irqs.pre_vector_init and > legacy_pic on some HW-reduced platforms AFAICS, so you make it > possible for the platform to provide its own variant of > acpi_reduced_hw_init().
In this particular case only timers matter, though in general you are right. > I would say something like this: > > "Some ACPI hawdware-reduced platforms need to initialize certain > devices defined by the ACPI hardware specification even though in > principle those devices should not be present in an ACPI > hawdware-reduced platform. To allow that to happen, make it possible > to override the generic x86_init callbacks and provide a custom > legacy_pic value, add a new ->reduced_hw_early_init() callback to > struct x86_init_acpi and make acpi_reduced_hw_init() use it." Thanks for review! I will use your suggestions in the next version. -- Andy Shevchenko <andriy.shevche...@linux.intel.com> Intel Finland Oy
