blob: 5e35e0921f86a22a7a468a115c059f2399ec1a67 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
|
From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Date: Wed, 3 Feb 2021 11:05:25 -0500
Subject: [PATCH 3/4] tracing: Use in_serving_softirq() to deduct softirq
status.
Origin: https://www.kernel.org/pub/linux/kernel/projects/rt/5.10/older/patches-5.10.17-rt32.tar.xz
PREEMPT_RT does not report "serving softirq" because the tracing core
looks at the preemption counter while PREEMPT_RT does not update it
while processing softirqs in order to remain preemptible. The
information is stored somewhere else.
The in_serving_softirq() macro and the SOFTIRQ_OFFSET define are still
working but not on the preempt-counter.
Use in_serving_softirq() macro which works on PREEMPT_RT. On !PREEMPT_RT
the compiler (gcc-10 / clang-11) is smart enough to optimize the
in_serving_softirq() related read of the preemption counter away.
The only difference I noticed by using in_serving_softirq() on
!PREEMPT_RT is that gcc-10 implemented tracing_gen_ctx_flags() as
reading FLAG, jmp _tracing_gen_ctx_flags(). Without in_serving_softirq()
it inlined _tracing_gen_ctx_flags() into tracing_gen_ctx_flags().
Link: https://lkml.kernel.org/r/20210125194511.3924915-4-bigeasy@linutronix.de
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
kernel/trace/trace.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -2589,8 +2589,7 @@ unsigned int tracing_gen_ctx_irq_test(un
trace_flags |= TRACE_FLAG_NMI;
if (pc & HARDIRQ_MASK)
trace_flags |= TRACE_FLAG_HARDIRQ;
-
- if (pc & SOFTIRQ_OFFSET)
+ if (in_serving_softirq())
trace_flags |= TRACE_FLAG_SOFTIRQ;
if (tif_need_resched())
|
