Re: [PATCH v2 13/48] kmsan: add KMSAN runtime core
From: Alexander Potapenko
Date: Mon Apr 04 2022 - 10:44:55 EST
> +depot_stack_handle_t kmsan_save_stack_with_flags(gfp_t flags,
> + unsigned int extra)
> +{
> + unsigned long entries[KMSAN_STACK_DEPTH];
> + unsigned int nr_entries;
> +
> + nr_entries = stack_trace_save(entries, KMSAN_STACK_DEPTH, 0);
> + nr_entries = filter_irq_stacks(entries, nr_entries);
This is redundant, __stack_depot_save() below already calls filter_irq_stacks().
> +
> + if (depth >= MAX_CHAIN_DEPTH) {
> + static atomic_long_t kmsan_skipped_origins;
> + long skipped = atomic_long_inc_return(&kmsan_skipped_origins);
> +
> + if (skipped % NUM_SKIPPED_TO_WARN == 0) {
> + pr_warn("not chained %ld origins\n", skipped);
> + dump_stack();
> + kmsan_print_origin(id);
> + }
> + return id;
> + }
> + depth++;
> + extra_bits = kmsan_extra_bits(depth, uaf);
> +
> + entries[0] = KMSAN_CHAIN_MAGIC_ORIGIN;
> + entries[1] = kmsan_save_stack_with_flags(GFP_ATOMIC, 0);
> + entries[2] = id;
> + return __stack_depot_save(entries, ARRAY_SIZE(entries), extra_bits,
> + GFP_ATOMIC, true);
@entries is initialized in non-instrumented code, so passing it to
filter_irq_stacks() etc. will result in false positives, unless we
explicitly unpoison it.
(right now KMSAN does not instrument kernel/stacktrace.c, but it
probably should)