On Wed, Jan 31, 2018 at 4:43 PM, Andy Shevchenko <andy.shevchenko@xxxxxxxxx> wrote: > On Mon, Jan 29, 2018 at 5:02 AM, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote: >> On Sun, Jan 28, 2018 at 4:04 PM, Andy Shevchenko >> <andy.shevchenko@xxxxxxxxx> wrote: >>> On Fri, Jan 26, 2018 at 8:21 PM, Juergen Gross <jgross@xxxxxxxx> wrote: >>>> On 26/01/18 19:08, Andy Shevchenko wrote: >>>>> On Thu, Jan 25, 2018 at 4:36 PM, Juergen Gross <jgross@xxxxxxxx> wrote: > >>>>> The problem with weak functions that we can't have more than one >>>>> implementation per kernel while we would like to built several code >>>>> paths. >>>>> >>>>> I have stumbled on the similar stuff and realize that. >>>>> >>>>> Perhaps, one of the solution is to have an additional struct under >>>>> x86_init to alternate ACPI related stuff. >>>> >>>> I think we can go that route when another user of that interface is >>>> appearing. >>> >>> Why not to establish the struct? At least this route I would like to >>> go with [1]. >>> >>> [1]: https://lkml.org/lkml/2018/1/17/834 >> >> Maybe I'm a bit slow today, but care to explain what exactly you mean? > > Instead of declaring function as __weak, establish a new struct for > ACPI related stubs and incorporate it into x86_init. > > That is my proposal. I think I would go this way in my case where I > need to treat differently ACPI HW reduced initialization of legacy > devices. IOW you'd like to have a set of ACPI init callbacks that could be defined by an arch, right?