When instrumenting functions, KMSAN obtains the per-task state (mostly pointers to metadata for function arguments and return values) once per function at its beginning, using the `current` pointer. Every time the instrumented function calls another function, this state (`struct kmsan_context_state`) is updated with shadow/origin data of the passed and returned values. When `current` changes in the low-level arch code, instrumented code can not notice that, and will still refer to the old state, possibly corrupting it or using stale data. This may result in false positive reports. To deal with that, we need to apply __no_kmsan_checks to the functions performing context switching - this will result in skipping all KMSAN shadow checks and marking newly created values as initialized, preventing all false positive reports in those functions. False negatives are still possible, but we expect them to be rare and impersistent. Suggested-by: Marco Elver <elver@xxxxxxxxxx> Signed-off-by: Alexander Potapenko <glider@xxxxxxxxxx> Link: https://linux-review.googlesource.com/id/Ib7d4d70946f08128ade207519c1ee405fd812839 --- v2: -- This patch was previously called "kmsan: skip shadow checks in files doing context switches". Per Mark Rutland's suggestion, we now only skip checks in low-level arch-specific code, as context switches in common code should be invisible to KMSAN. We also apply the checks to precisely the functions performing the context switch instead of the whole file. v4: -- Replace KMSAN_ENABLE_CHECKS_process_64.o with __no_kmsan_checks Link: https://linux-review.googlesource.com/id/I45e3ed9c5f66ee79b0409d1673d66ae419029bcb --- arch/x86/kernel/process_64.c | 1 + 1 file changed, 1 insertion(+) diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 1962008fe7437..6b3418bff3261 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -553,6 +553,7 @@ void compat_start_thread(struct pt_regs *regs, u32 new_ip, u32 new_sp, bool x32) * Kprobes not supported here. Set the probe on schedule instead. * Function graph tracer not supported too. */ +__no_kmsan_checks __visible __notrace_funcgraph struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p) { -- 2.37.0.rc0.161.g10f37bed90-goog