Skip to contentSkip to conversation

r8169: improvements for scheduled task handling

PATCH In review Heiner Kallweit 2 participants 5 messages Mar 22, 2020 – Mar 24, 2020

Subsystems (3)
lore
Share & tools
lore

Heiner Kallweit

author maintainer
#0 lore raw

cover r8169: improvements for scheduled task handling

This series includes some improvements for handling of scheduled tasks.

Heiner Kallweit (3):
r8169: simplify rtl_task
r8169: improve rtl_schedule_task
r8169: improve RTL8168b FIFO overflow workaround

 drivers/net/ethernet/realtek/r8169_main.c | 27 +++++------------------
 1 file changed, 6 insertions(+), 21 deletions(-)
Signature · 1 line
2.25.2
author maintainer
#1 lore raw

PATCH 1/3 r8169: simplify rtl_task

Currently rtl_task() is designed to handle a large number of tasks. However we have just one, so we can remove some overhead.

---
 drivers/net/ethernet/realtek/r8169_main.c | 20 +++-----------------
 1 file changed, 3 insertions(+), 17 deletions(-)
drivers/net/ethernet/realtek/r8169_main.c +3 -17
index d2eef3754..242b14cc7 100644
@@ -4589,31 +4589,17 @@ static irqreturn_t rtl8169_interrupt(int irq, void *dev_instance)
 
 static void rtl_task(struct work_struct *work)
 {
-	static const struct {
-		int bitnr;
-		void (*action)(struct rtl8169_private *);
-	} rtl_work[] = {
-		{ RTL_FLAG_TASK_RESET_PENDING,	rtl_reset_work },
-	};
 	struct rtl8169_private *tp =
 		container_of(work, struct rtl8169_private, wk.work);
-	struct net_device *dev = tp->dev;
-	int i;
 
 	rtl_lock_work(tp);
 
-	if (!netif_running(dev) ||
+	if (!netif_running(tp->dev) ||
 	    !test_bit(RTL_FLAG_TASK_ENABLED, tp->wk.flags))
 		goto out_unlock;
 
-	for (i = 0; i < ARRAY_SIZE(rtl_work); i++) {
-		bool pending;
-
-		pending = test_and_clear_bit(rtl_work[i].bitnr, tp->wk.flags);
-		if (pending)
-			rtl_work[i].action(tp);
-	}
-
+	if (test_and_clear_bit(RTL_FLAG_TASK_RESET_PENDING, tp->wk.flags))
+		rtl_reset_work(tp);
 out_unlock:
 	rtl_unlock_work(tp);
 }
Signature · 1 line
2.25.2
author maintainer
#2 lore raw

PATCH 2/3 r8169: improve rtl_schedule_task

In reply to Heiner Kallweit

The current implementation makes the implicit assumption that if a bit is set, then the work is scheduled already. Remove the need for this implicit assumption and call schedule_work() always. It will check internally whether the work is scheduled already.

---
 drivers/net/ethernet/realtek/r8169_main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/ethernet/realtek/r8169_main.c +2 -2
index 242b14cc7..e9ced0360 100644
@@ -2228,8 +2228,8 @@ u16 rtl8168h_2_get_adc_bias_ioffset(struct rtl8169_private *tp)
 
 static void rtl_schedule_task(struct rtl8169_private *tp, enum rtl_flag flag)
 {
-	if (!test_and_set_bit(flag, tp->wk.flags))
-		schedule_work(&tp->wk.work);
+	set_bit(flag, tp->wk.flags);
+	schedule_work(&tp->wk.work);
 }
 
 static void rtl8169_init_phy(struct rtl8169_private *tp)
Signature · 1 line
2.25.2
author maintainer
#3 lore raw

PATCH 3/3 r8169: improve RTL8168b FIFO overflow workaround

In reply to Heiner Kallweit

So far only the reset bit it set, but the handler executing the reset is not scheduled. Therefore nothing will happen until some other action schedules the handler. Improve this by ensuring that the handler is scheduled.

---
 drivers/net/ethernet/realtek/r8169_main.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/net/ethernet/realtek/r8169_main.c +1 -2
index e9ced0360..b7dc1c112 100644
@@ -4575,8 +4575,7 @@ static irqreturn_t rtl8169_interrupt(int irq, void *dev_instance)
 	if (unlikely(status & RxFIFOOver &&
 	    tp->mac_version == RTL_GIGA_MAC_VER_11)) {
 		netif_stop_queue(tp->dev);
-		/* XXX - Hack alert. See rtl_task(). */
-		set_bit(RTL_FLAG_TASK_RESET_PENDING, tp->wk.flags);
+		rtl_schedule_task(tp, RTL_FLAG_TASK_RESET_PENDING);
 	}
 
 	rtl_irq_disable(tp);
Signature · 1 line
2.25.2