The Linux Kernel Mailing List
 help / color / mirror / Atom feed
From: Namhyung Kim <namhyung@kernel.org>
To: Ian Rogers <irogers@google.com>
Cc: acme@kernel.org, gmx@google.com, james.clark@linaro.org,
	adrian.hunter@intel.com, jolsa@kernel.org,
	linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
	mingo@redhat.com, peterz@infradead.org
Subject: Re: [PATCH v6 3/6] perf maps: Add maps__mutate_mapping
Date: Mon, 11 May 2026 00:07:26 -0700	[thread overview]
Message-ID: <agGALmRbvBcJS9sM@z2> (raw)
In-Reply-To: <20260508082726.2795191-4-irogers@google.com>

On Fri, May 08, 2026 at 01:27:23AM -0700, Ian Rogers wrote:
> During kernel ELF symbol parsing (dso__process_kernel_symbol), proc
> kallsyms image loading (dso__load_kernel_sym,
> dso__load_guest_kernel_sym), and dynamic kernel memory map alignment
> updates (machine__update_kernel_mmap), the loader directly modifies
> live virtual address boundary keys fields on map objects.  If these
> boundaries are mutated while the map pointer actively resides inside
> the parent maps cache array list (kmaps) outside of any lock closure,
> an unsafe concurrent window is exposed where parallel worker lookup
> threads (e.g., inside perf top) can mistakenly assume the cache
> remains sorted based on stale parameters, executing binary search
> queries (bsearch) across an unsorted range and triggering lookups
> failure.
> 
> Fix this by introducing a thread-safe, atomic transactional framework
> routine maps__mutate_mapping() that explicitly acquires the parent
> maps write semaphore lock, executes an incoming mutation callback
> block to perform the field updates under full lock protection, and
> invalidates the sorted tracking flags prior to releasing the write
> lock. This guarantees absolute atomic synchronization invariants,
> completely closing the concurrent lookup race window. The adjacent
> module alignment pass inside machine__create_kernel_maps() is safely
> preserved as a high-performance lockless pass, as its invocation
> lifecycle bounds remain strictly single-threaded by contract during
> session initialization construction. 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.
> 
> Fixes: 39b12f781271 ("perf tools: Make it possible to read object code from vmlinux")
> Signed-off-by: Ian Rogers <irogers@google.com>
> ---
>  tools/perf/util/machine.c    | 32 +++++++++++++++++-----------
>  tools/perf/util/maps.c       | 26 +++++++++++++++++++++++
>  tools/perf/util/maps.h       |  2 ++
>  tools/perf/util/symbol-elf.c | 41 +++++++++++++++++++++++-------------
>  tools/perf/util/symbol.c     | 17 +++++++++++----
>  5 files changed, 87 insertions(+), 31 deletions(-)
> 
> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> index e76f8c86e62a..8d4452c70cb5 100644
> --- a/tools/perf/util/machine.c
> +++ b/tools/perf/util/machine.c
> @@ -1522,22 +1522,30 @@ static void machine__set_kernel_mmap(struct machine *machine,
>  		map__set_end(machine->vmlinux_map, ~0ULL);
>  }
>  
> -static int machine__update_kernel_mmap(struct machine *machine,
> -				     u64 start, u64 end)
> +struct kernel_mmap_mutation_ctx {
> +	u64 start;
> +	u64 end;
> +};
> +
> +static int kernel_mmap_mutate_cb(struct map *map, void *data)
>  {
> -	struct map *orig, *updated;
> -	int err;
> +	struct kernel_mmap_mutation_ctx *ctx = data;
>  
> -	orig = machine->vmlinux_map;
> -	updated = map__get(orig);
> +	map__set_start(map, ctx->start);
> +	map__set_end(map, ctx->end);
> +	if (ctx->start == 0 && ctx->end == 0)
> +		map__set_end(map, ~0ULL);
> +	return 0;
> +}
>  
> -	machine->vmlinux_map = updated;
> -	maps__remove(machine__kernel_maps(machine), orig);
> -	machine__set_kernel_mmap(machine, start, end);
> -	err = maps__insert(machine__kernel_maps(machine), updated);
> -	map__put(orig);
> +static int machine__update_kernel_mmap(struct machine *machine,
> +				     u64 start, u64 end)
> +{
> +	struct kernel_mmap_mutation_ctx ctx = { .start = start, .end = end };
>  
> -	return err;
> +	return maps__mutate_mapping(machine__kernel_maps(machine),
> +				     machine->vmlinux_map,
> +				     kernel_mmap_mutate_cb, &ctx);
>  }
>  
>  int machine__create_kernel_maps(struct machine *machine)
> diff --git a/tools/perf/util/maps.c b/tools/perf/util/maps.c
> index 81a97ac34077..91345a773aa2 100644
> --- a/tools/perf/util/maps.c
> +++ b/tools/perf/util/maps.c
> @@ -575,6 +575,32 @@ void maps__remove(struct maps *maps, struct map *map)
>  #endif
>  }
>  
> +int maps__mutate_mapping(struct maps *maps, struct map *map,
> +			 int (*mutate_cb)(struct map *map, void *data), void *data)
> +{
> +	int err = 0;
> +
> +	if (maps)
> +		down_write(maps__lock(maps));
> +
> +	err = mutate_cb(map, data);
> +
> +	if (maps) {
> +		RC_CHK_ACCESS(maps)->maps_by_address_sorted = false;
> +		RC_CHK_ACCESS(maps)->maps_by_name_sorted = false;
> +	}
> +
> +	if (maps)
> +		up_write(maps__lock(maps));
> +
> +#ifdef HAVE_LIBDW_SUPPORT
> +	if (maps)
> +		libdw__invalidate_dwfl(maps, maps__libdw_addr_space_dwfl(maps));
> +#endif
> +
> +	return err;
> +}

Could be simplified by checking 'maps' once.  But I'm not sure if
there's a case it doesn't have the maps.

Thanks,
Namhyung

> +
>  bool maps__empty(struct maps *maps)
>  {
>  	bool res;
> diff --git a/tools/perf/util/maps.h b/tools/perf/util/maps.h
> index 20c52084ba9e..de74ccbb8a12 100644
> --- a/tools/perf/util/maps.h
> +++ b/tools/perf/util/maps.h
> @@ -61,6 +61,8 @@ size_t maps__fprintf(struct maps *maps, FILE *fp);
>  
>  int maps__insert(struct maps *maps, struct map *map);
>  void maps__remove(struct maps *maps, struct map *map);
> +int maps__mutate_mapping(struct maps *maps, struct map *map,
> +			 int (*mutate_cb)(struct map *map, void *data), void *data);
>  
>  struct map *maps__find(struct maps *maps, u64 addr);
>  struct symbol *maps__find_symbol(struct maps *maps, u64 addr, struct map **mapp);
> diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c
> index 7afa8a117139..dc4ab58857b3 100644
> --- a/tools/perf/util/symbol-elf.c
> +++ b/tools/perf/util/symbol-elf.c
> @@ -1341,6 +1341,24 @@ static u64 ref_reloc(struct kmap *kmap)
>  void __weak arch__sym_update(struct symbol *s __maybe_unused,
>  		GElf_Sym *sym __maybe_unused) { }
>  
> +struct remap_kernel_ctx {
> +	u64 sh_addr;
> +	u64 sh_size;
> +	u64 sh_offset;
> +	struct kmap *kmap;
> +};
> +
> +static int remap_kernel_cb(struct map *map, void *data)
> +{
> +	struct remap_kernel_ctx *ctx = data;
> +
> +	map__set_start(map, ctx->sh_addr + ref_reloc(ctx->kmap));
> +	map__set_end(map, map__start(map) + ctx->sh_size);
> +	map__set_pgoff(map, ctx->sh_offset);
> +	map__set_mapping_type(map, MAPPING_TYPE__DSO);
> +	return 0;
> +}
> +
>  static int dso__process_kernel_symbol(struct dso *dso, struct map *map,
>  				      GElf_Sym *sym, GElf_Shdr *shdr,
>  				      struct maps *kmaps, struct kmap *kmap,
> @@ -1371,22 +1389,15 @@ static int dso__process_kernel_symbol(struct dso *dso, struct map *map,
>  		 * map to the kernel dso.
>  		 */
>  		if (*remap_kernel && dso__kernel(dso) && !kmodule) {
> +			struct remap_kernel_ctx ctx = {
> +				.sh_addr = shdr->sh_addr,
> +				.sh_size = shdr->sh_size,
> +				.sh_offset = shdr->sh_offset,
> +				.kmap = kmap
> +			};
> +
>  			*remap_kernel = false;
> -			map__set_start(map, shdr->sh_addr + ref_reloc(kmap));
> -			map__set_end(map, map__start(map) + shdr->sh_size);
> -			map__set_pgoff(map, shdr->sh_offset);
> -			map__set_mapping_type(map, MAPPING_TYPE__DSO);
> -			/* Ensure maps are correctly ordered */
> -			if (kmaps) {
> -				int err;
> -				struct map *tmp = map__get(map);
> -
> -				maps__remove(kmaps, map);
> -				err = maps__insert(kmaps, map);
> -				map__put(tmp);
> -				if (err)
> -					return err;
> -			}
> +			maps__mutate_mapping(kmaps, map, remap_kernel_cb, &ctx);
>  		}
>  
>  		/*
> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> index fcaeeddbbb6b..09b93e844887 100644
> --- a/tools/perf/util/symbol.c
> +++ b/tools/perf/util/symbol.c
> @@ -48,6 +48,13 @@
>  #include <symbol/kallsyms.h>
>  #include <sys/utsname.h>
>  
> +static int map_fixup_cb(struct map *map, void *data __maybe_unused)
> +{
> +	map__fixup_start(map);
> +	map__fixup_end(map);
> +	return 0;
> +}
> +
>  static int dso__load_kernel_sym(struct dso *dso, struct map *map);
>  static int dso__load_guest_kernel_sym(struct dso *dso, struct map *map);
>  static bool symbol__is_idle(const char *name);
> @@ -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);
>  	}
>  
>  	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);
>  	}
>  
>  	return err;
> -- 
> 2.54.0.563.g4f69b47b94-goog
> 

  parent reply	other threads:[~2026-05-11  7:07 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20260504072937.2103453-1-irogers@google.com>
2026-05-06  0:45 ` [PATCH v5 0/5] perf tools: Add inject --aslr feature and prerequisite robustness fixes Ian Rogers
2026-05-06  0:45   ` [PATCH v5 1/5] perf sched: Add missing mmap2 handler in timehist Ian Rogers
2026-05-06 13:22     ` Arnaldo Carvalho de Melo
2026-05-06 16:16       ` Ian Rogers
2026-05-06  0:45   ` [PATCH v5 2/5] perf tool: Fix missing schedstat delegates and dont_split_sample_group in delegate_tool Ian Rogers
2026-05-06  0:45   ` [PATCH v5 3/5] perf symbols: Fix map removal sequence inside dso__process_kernel_symbol() Ian Rogers
2026-05-06  0:45   ` [PATCH v5 4/5] perf inject/aslr: Add aslr tool to remap/obfuscate virtual addresses Ian Rogers
2026-05-06 18:52     ` Namhyung Kim
2026-05-06 20:01       ` Ian Rogers
2026-05-06  0:45   ` [PATCH v5 5/5] perf test: Add inject ASLR test Ian Rogers
2026-05-07 15:58     ` James Clark
2026-05-07 16:17       ` Ian Rogers
2026-05-08 10:42         ` James Clark
2026-05-08 10:49           ` James Clark
2026-05-08  8:27   ` [PATCH v6 0/6] perf tools: Add inject --aslr feature and prerequisite robustness fixes Ian Rogers
2026-05-08  8:27     ` [PATCH v6 1/6] perf sched: Add missing mmap2 handler in timehist Ian Rogers
2026-05-08  8:27     ` [PATCH v6 2/6] perf tool: Missing delegate_tool schedstat delegates and dont_split_sample_group Ian Rogers
2026-05-08  8:27     ` [PATCH v6 3/6] perf maps: Add maps__mutate_mapping Ian Rogers
2026-05-08 10:57       ` James Clark
2026-05-11  7:07       ` Namhyung Kim [this message]
2026-05-08  8:27     ` [PATCH v6 4/6] perf inject/aslr: Add aslr tool to remap/obfuscate virtual addresses Ian Rogers
2026-05-11  7:32       ` Namhyung Kim
2026-05-08  8:27     ` [PATCH v6 5/6] perf test: Add inject ASLR test Ian Rogers
2026-05-08 13:29       ` James Clark
2026-05-08 14:29         ` James Clark
2026-05-11  7:34       ` Namhyung Kim
2026-05-08  8:27     ` [PATCH v6 6/6] perf aslr: Strip sample registers Ian Rogers

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=agGALmRbvBcJS9sM@z2 \
    --to=namhyung@kernel.org \
    --cc=acme@kernel.org \
    --cc=adrian.hunter@intel.com \
    --cc=gmx@google.com \
    --cc=irogers@google.com \
    --cc=james.clark@linaro.org \
    --cc=jolsa@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=peterz@infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox