s390/bpf: Fix bpf_arch_text_poke() with new_addr == NULL again

Commit 7ded842b356d ("s390/bpf: Fix bpf_plt pointer arithmetic") has
accidentally removed the critical piece of commit c730fce7c70c
("s390/bpf: Fix bpf_arch_text_poke() with new_addr == NULL"), causing
intermittent kernel panics in e.g. perf's on_switch() prog to reappear.

Restore the fix and add a comment.

Fixes: 7ded842b356d ("s390/bpf: Fix bpf_plt pointer arithmetic")
Cc: stable@vger.kernel.org
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Link: https://lore.kernel.org/r/20250716194524.48109-2-iii@linux.ibm.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

diff --git a/arch/s390/net/bpf_jit_comp.c b/arch/s390/net/bpf_jit_comp.c
index c7f8313ba449716a8f18eafdeb6c77ed3b23f52e..0c9a35782c8337fb35500f72a39d24f959ca0023 100644 (file)
--- a/arch/s390/net/bpf_jit_comp.c
+++ b/arch/s390/net/bpf_jit_comp.c
@@ -566,7 +566,15 @@ static void bpf_jit_plt(struct bpf_plt *plt, void *ret, void *target)
 {
        memcpy(plt, &bpf_plt, sizeof(*plt));
        plt->ret = ret;
-       plt->target = target;
+       /*
+        * (target == NULL) implies that the branch to this PLT entry was
+        * patched and became a no-op. However, some CPU could have jumped
+        * to this PLT entry before patching and may be still executing it.
+        *
+        * Since the intention in this case is to make the PLT entry a no-op,
+        * make the target point to the return label instead of NULL.
+        */
+       plt->target = target ?: ret;
 }
 
 /*