[PATCH bpf-next 0/1] Fix BPF struct_ops BTF cleanup race condition

[Gregory Bell <grbell@redhat.com>]

This patch fixes a race condition in BPF struct_ops where consecutive
test runs fail due to a timing mismatch between BTF object cleanup and
BTF module removal.

When running BPF struct_ops tests consecutively, the second test fails
because BTF object references take longer to be released while their
associated module is removed from btf_modules list immediately
during module unload. This creates a race condition where the second
test retrieves the cached BTF object but cannot find valid module
information, causing btf_try_get_module() to return NULL.

The fix synchronizes these cleanup operations by waiting for active BTF
references to be released before completing module cleanup, eliminating
the race condition.

This issue is reproducible with any of the following tests: global_map_resize,
pro_epilogue, struct_ops_module and struct_ops_multi

	./test_progs -t global_map_resize; ./test_progs -t global_map_resize

	#134/1   global_map_resize/global_map_resize_bss:OK
	#134/2   global_map_resize/global_map_resize_data:OK
	#134/3   global_map_resize/global_map_resize_invalid:OK
	#134     global_map_resize:OK
	Summary: 1/3 PASSED, 0 SKIPPED, 0 FAILED
	global_map_resize_bss_subtest:PASS:test_global_map_resize__open 0 nsec
	libbpf: map 'test_glo.bss': cannot be resized, last var must be an array
	libbpf: map 'test_glo.bss': failed to adjust resized BTF, clearing BTF key/value info: -EINVAL
	global_map_resize_bss_subtest:PASS:bpf_map__set_value_size 0 nsec
	global_map_resize_bss_subtest:PASS:resize 0 nsec
	global_map_resize_bss_subtest:PASS:percpu_arr_resize 0 nsec
	global_map_resize_bss_subtest:PASS:array_len 0 nsec
	global_map_resize_bss_subtest:PASS:bpf_map__initial_value (ptr) 0 nsec
	global_map_resize_bss_subtest:PASS:bpf_map__initial_value (size) 0 nsec
	libbpf: map 'st_ops_resize': failed to create: -EINVAL
	libbpf: failed to load BPF skeleton 'test_global_map_resize': -EINVAL
	global_map_resize_bss_subtest:FAIL:test_global_map_resize__load unexpected error: -22 (errno 22)
	#134/1   global_map_resize/global_map_resize_bss:FAIL
	global_map_resize_data_subtest:PASS:test_global_map_resize__open 0 nsec
	global_map_resize_data_subtest:PASS:bpf_map__set_value_size 0 nsec
	global_map_resize_data_subtest:PASS:resize 0 nsec
	global_map_resize_data_subtest:PASS:percpu_arr_resize 0 nsec
	global_map_resize_data_subtest:PASS:array_len 0 nsec
	global_map_resize_data_subtest:PASS:bpf_map__initial_value (ptr) 0 nsec
	global_map_resize_data_subtest:PASS:bpf_map__initial_value (size) 0 nsec
	libbpf: map 'st_ops_resize': failed to create: -EINVAL
	libbpf: failed to load BPF skeleton 'test_global_map_resize': -EINVAL
	global_map_resize_data_subtest:FAIL:test_global_map_resize__load unexpected error: -22 (errno 22)
	#134/2   global_map_resize/global_map_resize_data:FAIL
	#134/3   global_map_resize/global_map_resize_invalid:OK
	#134     global_map_resize:FAIL

After the fix, consecutive runs work reliably without requiring delays.

Gregory Bell (1):
  bpf: Fix BTF module cleanup race condition in struct_ops

 kernel/bpf/btf.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

-- 
2.52.0