From: Hou Tao <houtao1@xxxxxxxxxx>
To support sleepable uprobe bpf program, the freeing of trace program
array chains a RCU-tasks-trace grace period with a normal RCU grace
period. But considering in the current implementation of RCU-tasks-trace
that one RCU-tasks-trace grace period implies one normal RCU grace
period, so it is not need for such chaining and it is safe to free the
array in the callback of call_rcu_tasks_trace().
Signed-off-by: Hou Tao <houtao1@xxxxxxxxxx>
---
kernel/bpf/core.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 711fd293b6de..f943620b55b0 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -2247,12 +2247,15 @@ void bpf_prog_array_free(struct bpf_prog_array *progs)
kfree_rcu(progs, rcu);
}
+/* Now RCU Tasks grace period implies RCU grace period, so no need to call
+ * kfree_rcu(), just call kfree() directly.
+ */
static void __bpf_prog_array_free_sleepable_cb(struct rcu_head *rcu)
{
struct bpf_prog_array *progs;
progs = container_of(rcu, struct bpf_prog_array, rcu);
- kfree_rcu(progs, rcu);
+ kfree(progs);
}
void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs)
--
2.29.2
