[PATCH v5 0/4] ARM: vfp: Introduce vfp_lock() for VFP locking.

[Russell King (Oracle)]

On Fri, Jun 28, 2024 at 03:46:15PM +0200, Sebastian Andrzej Siewior wrote:
> There was a bug report on PREEMPT_RT which made me look into VFP locking
> on ARM. The usage of local_bh_disable() to ensure exclusive access to
> the VFP unit is not working on PREEMPT_RT because the softirq context is
> preemptible.
> 
> This series introduces vfp_lock() which does the right thing.

I disagre with the term "lock" here - it isn't about any kind of
locking, it's about:

1. preventing the thread moving onto a different CPU
2. preventing kernel-mode usage of VFP while a VFP fault is being
   handled (and thus changing the VFP state.)

The proble with "lock" is that it makes it look like this code is
single-threaded which it isn't. Many CPUs can be processing a VFP
fault at the same time - remembering that each CPU has their own
VFP unit.

Maybe name it vfp_state_hold() ... vfp_state_release() instead?
Hold as in "don't let anything interfere with the state we are
going to be accessing".

I'm open to better names if anyone can think of any!

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!