Re: [PATCH] random32: Use rcuidle variant for tracepoint

[Eric Dumazet]

On Fri, Aug 21, 2020 at 1:59 AM <peterz@xxxxxxxxxxxxx> wrote:
>
> On Fri, Aug 21, 2020 at 08:30:43AM +0200, Marco Elver wrote:
> > With KCSAN enabled, prandom_u32() may be called from any context,
> > including idle CPUs.
> >
> > Therefore, switch to using trace_prandom_u32_rcuidle(), to avoid various
> > issues due to recursion and lockdep warnings when KCSAN and tracing is
> > enabled.
>
> At some point we're going to have to introduce noinstr to idle as well.
> But until that time this should indeed cure things.

I do not understand what the issue is.  This _rcuidle() is kind of opaque ;)

Would this alternative patch work, or is it something more fundamental ?

Thanks !

diff --git a/lib/random32.c b/lib/random32.c
index 932345323af092a93fc2690b0ebbf4f7485ae4f3..17af2d1631e5ab6e02ad1e9288af7e007bed6d5f
100644
--- a/lib/random32.c
+++ b/lib/random32.c
@@ -83,9 +83,10 @@ u32 prandom_u32(void)
        u32 res;

        res = prandom_u32_state(state);
-       trace_prandom_u32(res);
        put_cpu_var(net_rand_state);

+       trace_prandom_u32(res);
+
        return res;
 }
 EXPORT_SYMBOL(prandom_u32);