Re: [PATCH v2 11/11] arm64: debug: remove debug exception registration infrastructure

[Will Deacon <will@kernel.org>]

On Wed, May 28, 2025 at 05:41:43PM +0100, Ada Couprie Diaz wrote:
> On 21/05/2025 10:38, Will Deacon wrote:
> 
> > On Mon, May 12, 2025 at 06:43:26PM +0100, Ada Couprie Diaz wrote:
> > > diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
> > > index 50d7a6a75f45..ce20f46e08cb 100644
> > > --- a/arch/arm64/kernel/traps.c
> > > +++ b/arch/arm64/kernel/traps.c
> > > @@ -1101,29 +1101,5 @@ int ubsan_brk_handler(struct pt_regs *regs, unsigned long esr)
> > >   }
> > >   #endif
> > > -/*
> > > - * Initial handler for AArch64 BRK exceptions
> > > - * This handler only used until debug_traps_init().
> > > - */
> > > -int __init early_brk64(unsigned long addr, unsigned long esr,
> > > -		struct pt_regs *regs)
> > > -{
> > > -#ifdef CONFIG_CFI_CLANG
> > > -	if (esr_is_cfi_brk(esr))
> > > -		return cfi_brk_handler(regs, esr) != DBG_HOOK_HANDLED;
> > > -#endif
> > > -#ifdef CONFIG_KASAN_SW_TAGS
> > > -	if ((esr_brk_comment(esr) & ~KASAN_BRK_MASK) == KASAN_BRK_IMM)
> > > -		return kasan_brk_handler(regs, esr) != DBG_HOOK_HANDLED;
> > > -#endif
> > > -#ifdef CONFIG_UBSAN_TRAP
> > > -	if ((esr_brk_comment(esr) & ~UBSAN_BRK_MASK) == UBSAN_BRK_IMM)
> > > -		return ubsan_brk_handler(regs, esr) != DBG_HOOK_HANDLED;
> > > -#endif
> > > -	return bug_brk_handler(regs, esr) != DBG_HOOK_HANDLED;
> > The early bug handler is now only called if the brk immediate matches,
> > but I couldn't spot any problems with that.
> 
> Correct, as far as I can tell the behaviour is unchanged for a fall-through
> :
> `bug_brk_handler()` would have called `report_bug()`, which would
> not have found a non-BUG BRK and bailed early with BUG_TRAP_TYPE_NONE,
> leading `bug_brk_handler()` to return `DBG_HOOK_ERROR`.
> 
> Now, we would return `DBG_HOOK_ERROR` anyway if it's not an immediate we
> know how to handle. (This also seems to match better with the expectations
> of the comment in `is_valid_bugaddr()`, in `kernel/traps.c` !)
> 
> Happy to have my logic checked or add a mention in a commit message
> (either this one, or patch 02 ?)

It wouldn't hurt to have a note there, but I don't mind either way. I
just highlighted it to check that we agreed.

Will