Re: [PATCH v6 3/6] perf maps: Add maps__mutate_mapping

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Self-deadlock in the kallsyms symbol loading path.
--

commit 1340dc1c4b3b633e8a032536f0c7948f87b17ce3
Author: Ian Rogers <irogers@google.com>

perf maps: Add maps__mutate_mapping

This commit introduces maps__mutate_mapping to safely update map boundaries
by acquiring a write lock on the maps collection. It applies this new
synchronization to kernel ELF parsing, kallsyms image loading, and dynamic
memory map updates to close concurrent lookup races.

> There is a potential for self deadlock if maps__mutate_mapping is called
> with the lock held, such as with maps__for_each_map but this problem also
> existed with the previous remove and insert approaches.

Is this commit message accurate for all modified paths? Prior to this patch,
dso__load_kernel_sym and dso__load_guest_kernel_sym used map__fixup_start and
map__fixup_end directly without locking.

> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> index fcaeeddbbb6b9..09b93e844887d 100644
> --- a/tools/perf/util/symbol.c
> +++ b/tools/perf/util/symbol.c
[ ... ]
> @@ -2121,10 +2128,11 @@ static int dso__load_kernel_sym(struct dso *dso, struct map *map)
>  	free(kallsyms_allocated_filename);
>  
>  	if (err > 0 && !dso__is_kcore(dso)) {
> +		struct maps *kmaps = map__kmaps(map);
> +
>  		dso__set_binary_type(dso, DSO_BINARY_TYPE__KALLSYMS);
>  		dso__set_long_name(dso, DSO__NAME_KALLSYMS, false);
> -		map__fixup_start(map);
> -		map__fixup_end(map);
> +		maps__mutate_mapping(kmaps, map, map_fixup_cb, NULL);

Can this cause a self-deadlock when resolving kernel symbols by name?

If a tool resolves a symbol upon first load, the call chain appears to hold a
read lock while attempting to acquire a write lock on the exact same maps
collection:

maps__find_symbol_by_name()
  maps__for_each_map()                  <-- acquires read lock
    maps__find_symbol_by_name_cb()
      map__load()
        dso__load_kernel_sym()
          maps__mutate_mapping()        <-- attempts to acquire write lock

Since rw_semaphore cannot be upgraded from read to write without releasing the
read lock first, won't this immediately deadlock on the kmaps lock?

>  	}
>  
>  	return err;
[ ... ]
> @@ -2164,10 +2172,11 @@ static int dso__load_guest_kernel_sym(struct dso *dso, struct map *map)
>  	if (err > 0)
>  		pr_debug("Using %s for symbols\n", kallsyms_filename);
>  	if (err > 0 && !dso__is_kcore(dso)) {
> +		struct maps *kmaps = map__kmaps(map);
> +
>  		dso__set_binary_type(dso, DSO_BINARY_TYPE__GUEST_KALLSYMS);
>  		dso__set_long_name(dso, machine->mmap_name, false);
> -		map__fixup_start(map);
> -		map__fixup_end(map);
> +		maps__mutate_mapping(kmaps, map, map_fixup_cb, NULL);

Does this guest kernel symbol loading path suffer from the exact same
self-deadlock sequence?

>  	}
>  
>  	return err;

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260508082726.2795191-1-irogers@google.com?part=3