bpf: Add missing bpf_read_[un]lock_trace() for syscall program
authorYonghong Song <yhs@fb.com>
Mon, 9 Aug 2021 23:51:51 +0000 (16:51 -0700)
committerDaniel Borkmann <daniel@iogearbox.net>
Tue, 10 Aug 2021 08:10:49 +0000 (10:10 +0200)
Commit 79a7f8bdb159d ("bpf: Introduce bpf_sys_bpf() helper and program type.")
added support for syscall program, which is a sleepable program.

But the program run missed bpf_read_lock_trace()/bpf_read_unlock_trace(),
which is needed to ensure proper rcu callback invocations. This patch adds
bpf_read_[un]lock_trace() properly.

Fixes: 79a7f8bdb159d ("bpf: Introduce bpf_sys_bpf() helper and program type.")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210809235151.1663680-1-yhs@fb.com
net/bpf/test_run.c

index 1cc75c8..caa16bf 100644 (file)
@@ -7,6 +7,7 @@
 #include <linux/vmalloc.h>
 #include <linux/etherdevice.h>
 #include <linux/filter.h>
+#include <linux/rcupdate_trace.h>
 #include <linux/sched/signal.h>
 #include <net/bpf_sk_storage.h>
 #include <net/sock.h>
@@ -951,7 +952,10 @@ int bpf_prog_test_run_syscall(struct bpf_prog *prog,
                        goto out;
                }
        }
+
+       rcu_read_lock_trace();
        retval = bpf_prog_run_pin_on_cpu(prog, ctx);
+       rcu_read_unlock_trace();
 
        if (copy_to_user(&uattr->test.retval, &retval, sizeof(u32))) {
                err = -EFAULT;