mbox series

[v1,0/3] ACPI: OSL/EC: Use spin locks without disabling interrupts

Message ID 2330034.ElGaqSPkdT@kreacher
Headers show
Series ACPI: OSL/EC: Use spin locks without disabling interrupts | expand

Message

Rafael J. Wysocki Dec. 15, 2023, 11:23 a.m. UTC
Hi Everyone,

After

https://patchwork.kernel.org/project/linux-acpi/patch/5745568.DvuYhMxLoT@kreacher/

the SCI interrupt handler is threaded, so ACPICA spin locks can be used without
disabling interrupts, because they cannot be acquired from a hardirq handler
running on the same CPU as any code already holding the spin lock.  This patch
[1/3] removes disabling interrupts from acpi_os_acquire_lock() and updates
acpi_os_release_lock() accordingly.

Patch [2/2] uses the observation that if acpi_handle_interrupt() can run
in a threaded interrupt handler in the GPE case, it can also run in a threaded
interrupt handler in the dedicated IRQ case, so it makes the EC driver use a
threaded handler for the dedicated EC interrupt.

Patch [3/3], in analogy with patch [1/3], uses the observation that after patch
[2/2] all of the EC code runs in thread context, so it need not disable
interrupts on the local CPU before acquiring a spin lock, so the EC driver is
updated to operate its spin lock without disabling interrupts.

Thanks!

Comments

Rafael J. Wysocki Dec. 28, 2023, 1:18 p.m. UTC | #1
On Fri, Dec 15, 2023 at 12:34 PM Rafael J. Wysocki <rjw@rjwysocki.net> wrote:
>
> Hi Everyone,
>
> After
>
> https://patchwork.kernel.org/project/linux-acpi/patch/5745568.DvuYhMxLoT@kreacher/
>
> the SCI interrupt handler is threaded, so ACPICA spin locks can be used without
> disabling interrupts, because they cannot be acquired from a hardirq handler
> running on the same CPU as any code already holding the spin lock.  This patch
> [1/3] removes disabling interrupts from acpi_os_acquire_lock() and updates
> acpi_os_release_lock() accordingly.
>
> Patch [2/2] uses the observation that if acpi_handle_interrupt() can run
> in a threaded interrupt handler in the GPE case, it can also run in a threaded
> interrupt handler in the dedicated IRQ case, so it makes the EC driver use a
> threaded handler for the dedicated EC interrupt.
>
> Patch [3/3], in analogy with patch [1/3], uses the observation that after patch
> [2/2] all of the EC code runs in thread context, so it need not disable
> interrupts on the local CPU before acquiring a spin lock, so the EC driver is
> updated to operate its spin lock without disabling interrupts.

I don't see any objections to these patches, so queuing them up for 6.8.

Thanks!