Re: [PATCH] ARM: spectre-v2: fix smp_processor_id() warning

[Tetsuo Handa]

On 2022/06/23 0:02, Tetsuo Handa wrote:
> On 2022/06/22 23:04, Russell King (Oracle) wrote:
>> Which makes me think... having the loud complaint from the kernel there
>> is actually a good thing, it makes people sit up and notice that
>> something is wrong.
> 
> OK. Then, would you change the code not to emit "BUG:" message, for
> syzkaller stops testing upon encountering "BUG:" string?
> 

Something like this?

 arch/arm/include/asm/system_misc.h | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/arch/arm/include/asm/system_misc.h b/arch/arm/include/asm/system_misc.h
index 98b37340376b..09d5c2262165 100644
--- a/arch/arm/include/asm/system_misc.h
+++ b/arch/arm/include/asm/system_misc.h
@@ -20,10 +20,22 @@ typedef void (*harden_branch_predictor_fn_t)(void);
 DECLARE_PER_CPU(harden_branch_predictor_fn_t, harden_branch_predictor_fn);
 static inline void harden_branch_predictor(void)
 {
-	harden_branch_predictor_fn_t fn = per_cpu(harden_branch_predictor_fn,
-						  smp_processor_id());
+	static DEFINE_RATELIMIT_STATE(predictor_rs, 10 * HZ, 1);
+	const bool is_preemptible = preemptible();
+	harden_branch_predictor_fn_t fn;
+
+	if (unlikely(is_preemptible)) {
+		ratelimit_set_flags(&predictor_rs, RATELIMIT_MSG_ON_RELEASE);
+		if (__ratelimit(&predictor_rs))
+			pr_err("%s[%d] page fault with preemption enabled (exploit attempt?)\n",
+			       current->comm, task_pid_nr(current));
+		preempt_disable_notrace();
+	}
+	fn = per_cpu(harden_branch_predictor_fn, raw_smp_processor_id());
 	if (fn)
 		fn();
+	if (unlikely(is_preemptible))
+		preempt_enable_no_resched_notrace();
 }
 #else
 #define harden_branch_predictor() do { } while (0)
-- 
2.18.4