netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH net-next 1/2] net: add a napi variant for RT-well-behaved drivers
@ 2021-05-14 22:24 Jakub Kicinski
  2021-05-14 22:24 ` [PATCH net-next 2/2] nfp: use napi_schedule_irq() Jakub Kicinski
                   ` (3 more replies)
  0 siblings, 4 replies; 12+ messages in thread
From: Jakub Kicinski @ 2021-05-14 22:24 UTC (permalink / raw)
  To: davem, tglx
  Cc: netdev, eric.dumazet, simon.horman, oss-drivers, bigeasy, Jakub Kicinski

Most networking drivers use napi_schedule_irqoff() to schedule
NAPI from hardware IRQ handler. Unfortunately, as explained in
commit 8380c81d5c4f ("net: Treat __napi_schedule_irqoff() as
__napi_schedule() on PREEMPT_RT") the current implementation
is problematic for RT.

The best solution seems to be to mark the irq handler with
IRQF_NO_THREAD, to avoid going through an irq thread just
to schedule NAPI and therefore wake up ksoftirqd.

Since analyzing the 40 callers of napi_schedule_irqoff()
to figure out which handlers are light-weight enough to
warrant IRQF_NO_THREAD seems like a larger effort add
a new helper for drivers which set IRQF_NO_THREAD.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 include/linux/netdevice.h | 21 ++++++++++++++++-----
 net/core/dev.c            | 13 +++++++++++--
 2 files changed, 27 insertions(+), 7 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 5cbc950b34df..457e2e3ef5a5 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -436,7 +436,8 @@ typedef enum rx_handler_result rx_handler_result_t;
 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
 
 void __napi_schedule(struct napi_struct *n);
-void __napi_schedule_irqoff(struct napi_struct *n);
+void __napi_schedule_irqoff(struct napi_struct *n); /* deprecated */
+void __napi_schedule_irq(struct napi_struct *n);
 
 static inline bool napi_disable_pending(struct napi_struct *n)
 {
@@ -463,16 +464,26 @@ static inline void napi_schedule(struct napi_struct *n)
 		__napi_schedule(n);
 }
 
+/* Deprecated, use napi_schedule_irq(). */
+static inline void napi_schedule_irqoff(struct napi_struct *n)
+{
+	if (napi_schedule_prep(n))
+		__napi_schedule_irqoff(n);
+}
+
 /**
- *	napi_schedule_irqoff - schedule NAPI poll
- *	@n: NAPI context
+ * napi_schedule_irq() - schedule NAPI poll from hardware IRQ
+ * @n: NAPI context
  *
  * Variant of napi_schedule(), assuming hard irqs are masked.
+ * Hardware interrupt handler must be marked with IRQF_NO_THREAD
+ * to safely invoke this function on CONFIG_RT=y kernels (unless
+ * it manually masks the interrupts already).
  */
-static inline void napi_schedule_irqoff(struct napi_struct *n)
+static inline void napi_schedule_irq(struct napi_struct *n)
 {
 	if (napi_schedule_prep(n))
-		__napi_schedule_irqoff(n);
+		__napi_schedule_irq(n);
 }
 
 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
diff --git a/net/core/dev.c b/net/core/dev.c
index febb23708184..2e20858b5df6 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6497,20 +6497,29 @@ bool napi_schedule_prep(struct napi_struct *n)
 }
 EXPORT_SYMBOL(napi_schedule_prep);
 
+void __napi_schedule_irq(struct napi_struct *n)
+{
+	____napi_schedule(this_cpu_ptr(&softnet_data), n);
+}
+EXPORT_SYMBOL(__napi_schedule_irq);
+
 /**
  * __napi_schedule_irqoff - schedule for receive
  * @n: entry to schedule
  *
- * Variant of __napi_schedule() assuming hard irqs are masked.
+ * Legacy variant of __napi_schedule() assuming hard irqs are masked.
  *
  * On PREEMPT_RT enabled kernels this maps to __napi_schedule()
  * because the interrupt disabled assumption might not be true
  * due to force-threaded interrupts and spinlock substitution.
+ *
+ * For light weight IRQ handlers prefer use of napi_schedule_irq(),
+ * and marking IRQ handler with IRQF_NO_THREAD.
  */
 void __napi_schedule_irqoff(struct napi_struct *n)
 {
 	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
-		____napi_schedule(this_cpu_ptr(&softnet_data), n);
+		__napi_schedule_irq(n);
 	else
 		__napi_schedule(n);
 }
-- 
2.31.1


^ permalink raw reply	[flat|nested] 12+ messages in thread

end of thread, other threads:[~2021-05-21 14:11 UTC | newest]

Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-05-14 22:24 [PATCH net-next 1/2] net: add a napi variant for RT-well-behaved drivers Jakub Kicinski
2021-05-14 22:24 ` [PATCH net-next 2/2] nfp: use napi_schedule_irq() Jakub Kicinski
2021-05-17  9:48   ` Simon Horman
2021-05-15  0:17 ` [PATCH net-next 1/2] net: add a napi variant for RT-well-behaved drivers Thomas Gleixner
2021-05-15  0:21   ` Jakub Kicinski
2021-05-15  0:29     ` Thomas Gleixner
2021-05-15  9:49 ` Thomas Gleixner
2021-05-15 20:38   ` Jakub Kicinski
2021-05-15 11:07 ` Sebastian Andrzej Siewior
2021-05-15 20:31   ` Jakub Kicinski
2021-05-15 20:53     ` Thomas Gleixner
2021-05-21 14:11       ` Juri Lelli

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).