On Mon, Oct 09, 2023 at 08:29:49PM +0800, Hanjun Guo wrote: > On 2023/10/7 8:11, D Scott Phillips wrote: > > Sudeep Holla <sudeep.holla@xxxxxxx> writes: > > > > > Move all of the ARM-specific initialization into one function namely > > > acpi_arm_init(), so it is not necessary to modify/update bus.c every > > > time a new piece of it is added. > > > > > > Cc: Lorenzo Pieralisi <lpieralisi@xxxxxxxxxx> > > > Cc: Rafael J. Wysocki <rafael@xxxxxxxxxx> > > > Suggested-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx> > > > Reviewed-by: Robin Murphy <robin.murphy@xxxxxxx> > > > Reviewed-by: Hanjun Guo <guohanjun@xxxxxxxxxx> > > > Link: https://lore.kernel.org/r/CAJZ5v0iBZRZmV_oU+VurqxnVMbFN_ttqrL=cLh0sUH+=u0PYsw@xxxxxxxxxxxxxx > > > Signed-off-by: Sudeep Holla <sudeep.holla@xxxxxxx> > > > --- > > > drivers/acpi/arm64/Makefile | 2 +- > > > drivers/acpi/arm64/agdi.c | 2 +- > > > drivers/acpi/arm64/apmt.c | 2 +- > > > drivers/acpi/arm64/init.c | 13 +++++++++++++ > > > drivers/acpi/arm64/init.h | 6 ++++++ > > > drivers/acpi/arm64/iort.c | 1 + > > > drivers/acpi/bus.c | 7 +------ > > > include/linux/acpi.h | 6 ++++++ > > > include/linux/acpi_agdi.h | 13 ------------- > > > include/linux/acpi_apmt.h | 19 ------------------- > > > include/linux/acpi_iort.h | 2 -- > > > 11 files changed, 30 insertions(+), 43 deletions(-) > > > create mode 100644 drivers/acpi/arm64/init.c > > > create mode 100644 drivers/acpi/arm64/init.h > > > delete mode 100644 include/linux/acpi_agdi.h > > > delete mode 100644 include/linux/acpi_apmt.h [...] > > > @@ -1408,7 +1405,7 @@ static int __init acpi_init(void) > > > acpi_init_ffh(); > > > > > > pci_mmcfg_late_init(); > > > - acpi_iort_init(); > > > + acpi_arm_init(); > > > acpi_viot_early_init(); > > > acpi_hest_init(); > > > acpi_ghes_init(); > > > @@ -1420,8 +1417,6 @@ static int __init acpi_init(void) > > > acpi_debugger_init(); > > > acpi_setup_sb_notify_handler(); > > > acpi_viot_init(); > > > - acpi_agdi_init(); > > > - acpi_apmt_init(); > > > > Hi Sudeep, this moves acpi_agdi_init() before acpi_ghes_init(). > > sdei initialization currently happens from ghes_init, so agdi devices > > using SDEI can no longer probe: > > > > | [ 0.515864] sdei: Failed to create event 1073741825: -5 > > | [ 0.515866] agdi agdi.0: Failed to register for SDEI event 1073741825 > > | [ 0.515867] agdi: probe of agdi.0 failed with error -5 > > | ... > > | [ 0.516022] sdei: SDEIv1.0 (0x0) detected in firmware. > Sorry for that. > How about just move acpi_arm_init() to the place of after > acpi_ghes_init()? > Yes that could work but I am not comfortable with such play around with the ordering. I don't have better alternative as well :( so I am OK with the proposal. We had all sorts of issues with initcalls in the past but I understand this one is not as bad as that, I am just cautious. > I checked the functions in acpi_arm_init(), there are no other > dependencies except sdei_init(). > Fair enough. Anyways, for now Hanjun's suggestion should work. Can either one of you write a patch and post ? Or you use want me to do that ? -- Regards, Sudeep
