[PATCH bpf-next] selftests/bpf: Fix uprobe consumer test (again)

[Jiri Olsa <jolsa@kernel.org>]

The new uprobe changes bring bit some new behaviour that we need
to reflect in the consumer test.

There's special case when we have one of the existing uretprobes removed
and at the same time we're adding the other uretprobe. In this case we get
hit on the new uretprobe consumer only if there was already another uprobe
existing so the uprobe object stayed valid for uprobe return instance.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 .../selftests/bpf/prog_tests/uprobe_multi_test.c    | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
index 619b31cd24a1..545b91385749 100644
--- a/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
+++ b/tools/testing/selftests/bpf/prog_tests/uprobe_multi_test.c
@@ -873,10 +873,21 @@ static int consumer_test(struct uprobe_multi_consumers *skel,
 			 * which means one of the 'return' uprobes was alive when probe was hit:
 			 *
 			 *   idxs: 2/3 uprobe return in 'installed' mask
+			 *
+			 * There's special case when we have one of the existing uretprobes removed
+			 * and at the same time we're adding the other uretprobe. In this case we get
+			 * hit on the new uretprobe consumer only if there was already another uprobe
+			 * existing so the uprobe object stayed valid for uprobe return instance.
 			 */
 			unsigned long had_uretprobes  = before & 0b1100; /* is uretprobe installed */
+			unsigned long b = before >> 2, a = after >> 2;
+			bool hit = true;
+
+			/* Match for following a/b cases: 01/10 10/01 */
+			if ((a ^ b) == 0b11)
+				hit = before & 0b11;
 
-			if (had_uretprobes && test_bit(idx, after))
+			if (hit && had_uretprobes && test_bit(idx, after))
 				val++;
 			fmt = "idx 2/3: uretprobe";
 		}
-- 
2.47.0