Re: [PATCH v2] riscv: probes: reject kprobes inside LR/SC sequences

From: Nam Cao <namcao@linutronix.de>
To: Xiaofeng Yuan <yuanxiaofeng@eswincomputing.com>,
	Paul Walmsley <pjw@kernel.org>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>
Cc: linux-riscv@lists.infradead.org, yuanxiaofeng@eswincomputing.com
Subject: Re: [PATCH v2] riscv: probes: reject kprobes inside LR/SC sequences
Date: Tue, 25 Aug 2026 09:02:07 +0200	[thread overview]
Message-ID: <87v78y7keo.fsf@yellow.woof> (raw)
In-Reply-To: <20260822105303.2043-1-yuanxiaofeng@eswincomputing.com>

Xiaofeng Yuan <yuanxiaofeng@eswincomputing.com> writes:
> +/*
> + * A trap taken in the middle of an LR/SC sequence clears the load
> + * reservation, so an SC following the probed instruction would always
> + * fail and the enclosing retry loop would re-enter the breakpoint.
> + * Reject probes inside such a sequence.
> + *
> + * A constrained LR/SC loop (Zalrsc) must be contained within a 64-byte
> + * contiguous region of memory and comprise at most 16 instructions.
> + * Both bounds hold regardless of the C extension, so scanning back
> + * 64 bytes from the probed instruction always covers the enclosing
> + * sequence.
> + */
> +#define MAX_ATOMIC_CONTEXT_SIZE	64

Instead of 64 bytes, can we just do 16 instructions?

And instead of finding the start address to scan and then do a forward
scan, doing a backward scan of 16 instructions is probably cleaner and
more efficient?

Nam

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv