On Thursday, November 08, 2012 10:20:42 PM Mika Westerberg wrote: > On Thu, Nov 08, 2012 at 01:46:24AM +0100, Rafael J. Wysocki wrote: > > On Wednesday, November 07, 2012 03:05:48 PM Mika Westerberg wrote: > > > On Wed, Nov 07, 2012 at 12:14:31PM +0100, Rafael J. Wysocki wrote: > > > > > So is the idea now that the ACPI core parses the resources and passes them > > > > > forward via struct acpi_device? I'm just wondering how to proceed with > > > > > these I2C and SPI enumeration patches. > > > > > > > > Well, we definitely don't want to duplicate what drivers/pnp/pnpacpi/rsparser.c > > > > does, so the idea is to move the code from there to the core in such a way that > > > > both the SPI/I2C patches and the PNP layer can use it. > > > > > > Ok. > > > > > > > I'll have some prototype code ready shortly, hopefully, and I'll post it > > > > in that form for comments (and so that you know what to expect). > > > > > > Sounds good. Thanks! > > > > There you go. > > > > I haven't even try to compile it, so most likely it breaks things left, right > > and in between, but I hope it shows the idea. > > Thanks Rafael! > > I'll try this tomorrow (we had problems with the HW today so I wasn't able > to do any testing). > > I'll convert the SPI and I2C enumeration patches to use this method. OK, but I still need to move parsers of interrupt resources from rsparser.c to resource.c. :-) > > It does a couple of things at the same time, so it should be split into a few > > simpler patches. First, it moves some code from drivers/pnp/pnpacpi/rsparser.c > > to a new file drivers/acpi/resource.c and makes pnpacpi use functions from > > there. Second, it changes acpi_platform.c to use those functions too. > > Finally, it adds a list of ACPI resources to struct acpi_device and > > makes acpi_platform.c use that list intead of evaluating _CRS and parsing its > > output with acpi_walk_resources(). > > > > While changing acpi_platform.c I noticed that we had a bug in there, because > > GSIs were registered for the struct acpi_device object, although they should be > > registered for the struct platform_device one created by that code. I didn't > > try to fix that in the patch below, but it generally needs fixing. > > Good point. > > I wonder if the acpi_register_gsi() wants a device that is registered to > the Linux device framework? No, it doesn't, as far as I can tell. At least the pnpacpi code adds devices after registering GSIs for them. Also the existing implementations of acpi_register_gsi() don't require that. > At least with the SPI and I2C we generally > don't have such until we call i2c_new_device() or spi_add_device() and they > are getting passed the IRQ number which should be translated to the Linux > IRQ before that... That's correct. It looks like we could follow the hpet code and pass NULL as the dev argument to acpi_register_gsi() from there, as apparently the dev argument is only used to special-case PCI devices in mp_config_acpi_gsi(). Thanks, Rafael -- I speak only for myself. Rafael J. Wysocki, Intel Open Source Technology Center. -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html