mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH bpf-next v2] libbpf: Fix stale raw data after btf__relocate()
@ 2026-11-03  8:18 Jiayuan Chen
  2026-11-03 18:10 ` patchwork-bot+netdevbpf
  0 siblings, 1 reply; 2+ messages in thread
From: Jiayuan Chen @ 2026-11-03  8:18 UTC (permalink / raw)
  To: bpf
  Cc: Jiayuan Chen, 洛,
	Alexei Starovoitov, Daniel Borkmann, Andrii Nakryiko,
	Eduard Zingerman, Kumar Kartikeya Dwivedi, Martin KaFai Lau,
	Song Liu, Yonghong Song, Jiri Olsa, Emil Tsalapatis,
	Ihor Solodrai, Alan Maguire, linux-kernel

btf_relocate() rewrites type ids in place but doesn't drop the cached
raw_data, so btf__raw_data() after btf__relocate() returns the old
bytes with pre-relocation ids.

Only modifiable BTF keeps raw_data as a cache. For read-only BTF, e.g.
module BTF relocated in btf_parse_elf(), types_data points into
raw_data and is already rewritten in place, so it must not be freed.
Make btf_invalidate_raw_data() tell the two apart and call it from
btf__relocate(). Also recompute named_start_id since start_id changes
after relocation, like the kernel does.

Fixes: 19e00c897d50 ("libbpf: Split BTF relocation")
Reported-by: 洛 <wanghai8081@gmail.com>
Closes: https://lore.kernel.org/bpf/455f7257-25b9-4070-8313-42af8005b948@linux.dev/T/#t
Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
---
v2:
 - Free raw_data only for modifiable BTF and call btf_invalidate_raw_data()
   from btf__relocate() instead of btf_ensure_modifiable(), per Alexei
 - Recompute named_start_id after relocation
 - Drop the selftest since it is contrived, as Alexei pointed out
 - Target bpf-next
v1: https://lore.kernel.org/bpf/20261101155615.324501-1-jiayuan.chen@linux.dev/
---
 tools/lib/bpf/btf.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/tools/lib/bpf/btf.c b/tools/lib/bpf/btf.c
index d2ff6f5d6d5b..d67f78d76750 100644
--- a/tools/lib/bpf/btf.c
+++ b/tools/lib/bpf/btf.c
@@ -2033,7 +2033,8 @@ struct btf *btf__load_from_kernel_by_id(__u32 id)
 
 static void btf_invalidate_raw_data(struct btf *btf)
 {
-	if (btf->raw_data)
+	/* types_data and strs_data of read-only BTF point into raw_data */
+	if (btf->raw_data && btf_is_modifiable(btf))
 		btf_free_raw_data(btf);
 	if (btf->raw_data_swapped) {
 		free(btf->raw_data_swapped);
@@ -2099,11 +2100,11 @@ static int btf_ensure_modifiable(struct btf *btf)
 	if (!btf->base_btf && btf->hdr.str_len == 1)
 		btf->strs_deduped = true;
 
+	btf->modifiable = true;
+
 	/* invalidate raw_data representation */
 	btf_invalidate_raw_data(btf);
 
-	btf->modifiable = true;
-
 	return 0;
 
 err_out:
@@ -6642,8 +6643,12 @@ int btf__relocate(struct btf *btf, const struct btf *base_btf)
 {
 	int err = btf_relocate(btf, base_btf, NULL, NULL);
 
-	if (!err)
+	/* types are rewritten in place, some of them even on error */
+	btf_invalidate_raw_data(btf);
+	if (!err) {
 		btf->owns_base = false;
+		btf_check_sorted(btf);
+	}
 	return libbpf_err(err);
 }
 
-- 
2.43.0


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-11-03 18:10 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-03  8:18 [PATCH bpf-next v2] libbpf: Fix stale raw data after btf__relocate() Jiayuan Chen
2026-11-03 18:10 ` patchwork-bot+netdevbpf

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®