On Mon, 18 Oct 2021 12:19:20 +0200
Petr Mladek <pmladek@xxxxxxxx> wrote:
> > -
> > bit = trace_get_context_bit() + start;
> > if (unlikely(val & (1 << bit))) {
> > /*
> > * It could be that preempt_count has not been updated during
> > * a switch between contexts. Allow for a single recursion.
> > */
> > - bit = TRACE_TRANSITION_BIT;
> > + bit = TRACE_CTX_TRANSITION + start;
>
[..]
> Could we please update the comment? I mean to say if it is a race
> or if we trace a function that should not get traced.
What do you think of this change?
diff --git a/include/linux/trace_recursion.h b/include/linux/trace_recursion.h
index 1d8cce02c3fb..24f284eb55a7 100644
--- a/include/linux/trace_recursion.h
+++ b/include/linux/trace_recursion.h
@@ -168,8 +168,12 @@ static __always_inline int trace_test_and_set_recursion(unsigned long ip, unsign
bit = trace_get_context_bit() + start;
if (unlikely(val & (1 << bit))) {
/*
- * It could be that preempt_count has not been updated during
- * a switch between contexts. Allow for a single recursion.
+ * If an interrupt occurs during a trace, and another trace
+ * happens in that interrupt but before the preempt_count is
+ * updated to reflect the new interrupt context, then this
+ * will think a recursion occurred, and the event will be dropped.
+ * Let a single instance happen via the TRANSITION_BIT to
+ * not drop those events.
*/
bit = TRACE_TRANSITION_BIT;
if (val & (1 << bit)) {
-- Steve
