Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1489598

[PATCH RT 08/10] x86/preempt-lazy: fixup should_resched()

Show all headers | View raw

4.1.33-rt38-rc1 stable review patch.
If anyone has any objections, please let me know.

------------------

From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

should_resched() returns true if NEED_RESCHED is set and the
preempt_count is 0 _or_ if NEED_RESCHED_LAZY is set ignoring the preempt
counter. Ignoring the preemp counter is wrong. This patch adds this into
account.
While at it, __preempt_count_dec_and_test() ignores preempt_lazy_count
while checking TIF_NEED_RESCHED_LAZY so we this check, too.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
 arch/x86/include/asm/preempt.h | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/preempt.h b/arch/x86/include/asm/preempt.h
index c08949b0314d..eff1b8609f77 100644
--- a/arch/x86/include/asm/preempt.h
+++ b/arch/x86/include/asm/preempt.h
@@ -92,6 +92,8 @@ static __always_inline bool __preempt_count_dec_and_test(void)
 	if (____preempt_count_dec_and_test())
 		return true;
 #ifdef CONFIG_PREEMPT_LAZY
+	if (current_thread_info()->preempt_lazy_count)
+		return false;
 	return test_thread_flag(TIF_NEED_RESCHED_LAZY);
 #else
 	return false;
@@ -104,8 +106,19 @@ static __always_inline bool __preempt_count_dec_and_test(void)
 static __always_inline bool should_resched(int preempt_offset)
 {
 #ifdef CONFIG_PREEMPT_LAZY
-	return unlikely(raw_cpu_read_4(__preempt_count) == preempt_offset ||
-			test_thread_flag(TIF_NEED_RESCHED_LAZY));
+	u32 tmp;
+
+	tmp = raw_cpu_read_4(__preempt_count);
+	if (tmp == preempt_offset)
+		return true;
+
+	/* preempt count == 0 ? */
+	tmp &= ~PREEMPT_NEED_RESCHED;
+	if (tmp)
+		return false;
+	if (current_thread_info()->preempt_lazy_count)
+		return false;
+	return test_thread_flag(TIF_NEED_RESCHED_LAZY);
 #else
 	return unlikely(raw_cpu_read_4(__preempt_count) == preempt_offset);
 #endif
-- 
2.8.1

Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

[PATCH RT 00/10] Linux 4.1.33-rt38-rc1 Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:00 +0200
  [PATCH RT 08/10] x86/preempt-lazy: fixup should_resched() Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:00 +0200
  [PATCH RT 10/10] Linux 4.1.33-rt38-rc1 Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:00 +0200
  [PATCH RT 01/10] timers: wakeup all timer waiters Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:00 +0200
  [PATCH RT 02/10] timers: wakeup all timer waiters without holding the base lock Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:10 +0200
  Re: [PATCH RT 05/10] net: add back the missing serialization in  ip_send_unicast_reply() Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:10 +0200
  [PATCH RT 07/10] fs/dcache: resched/chill only if we make no progress Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:10 +0200
  [PATCH RT 04/10] scsi/fcoe: Fix get_cpu()/put_cpu_light() imbalance in  fcoe_recv_frame() Steven Rostedt <rostedt@goodmis.org> - 2016-09-23 00:10 +0200

csiph-web