Re: [PATCH v4 3/6] arm64: hw_breakpoint: Add lockdep_assert_irqs_disabled() on install/uninstall

[Will Deacon]

On Tue, Apr 07, 2026 at 09:29:45AM -0500, Rob Herring (Arm) wrote:
> The breakpoint install/uninstall/restore code depends on interrupts
> being disabled. Make this requirement explicit with a
> lockdep_assert_irqs_disabled() assertion.
> 
> Signed-off-by: Rob Herring (Arm) <robh@xxxxxxxxxx>
> ---
>  arch/arm64/kernel/hw_breakpoint.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/arch/arm64/kernel/hw_breakpoint.c b/arch/arm64/kernel/hw_breakpoint.c
> index bb39bc759810..a9266dc710b4 100644
> --- a/arch/arm64/kernel/hw_breakpoint.c
> +++ b/arch/arm64/kernel/hw_breakpoint.c
> @@ -231,6 +231,8 @@ static int hw_breakpoint_control(struct perf_event *bp,
>  	enum dbg_active_el dbg_el = debug_exception_level(info->ctrl.privilege);
>  	u32 ctrl;
>  
> +	lockdep_assert_irqs_disabled();

This function (hw_breakpoint_control()) is static and only has three
callers:

  1. Via the cpu hotplug CPUHP_AP_PERF_ARM_HW_BREAKPOINT_STARTING notifier
  2. From arch_install_hw_breakpoint()
  3. From arch_uninstall_hw_breakpoint()

So if we're called with irqs enabled, the core code has gone very wrong
and I don't think we should necessarily be checking that in the arch
backend. We also already have a WARN_ON(preemptible()) in
{enable,disable}_debug_monitors() so if you really want to add this then
please can you spell out why you're specifically concerned about the
preemption-disabled but irq-enabled case in the commit message?

Will