[PATCH 3/5] forcedeth: process TX completions using NAPI
From: Jeff Garzik
Date: Sat Oct 06 2007 - 11:14:56 EST
commit 57cbfacc00d69be2ba02b65d1021442273b76263
Author: Jeff Garzik <jeff@xxxxxxxxxx>
Date: Fri Oct 5 23:25:56 2007 -0400
[netdrvr] forcedeth: process TX completions using NAPI
Signed-off-by: Jeff Garzik <jgarzik@xxxxxxxxxx>
drivers/net/forcedeth.c | 143 +++++++++++++++++++++++++++---------------------
1 file changed, 83 insertions(+), 60 deletions(-)
57cbfacc00d69be2ba02b65d1021442273b76263
diff --git a/drivers/net/forcedeth.c b/drivers/net/forcedeth.c
index 1d1a5f5..1c236e6 100644
--- a/drivers/net/forcedeth.c
+++ b/drivers/net/forcedeth.c
@@ -744,6 +744,7 @@ struct fe_priv {
struct net_device *dev;
struct napi_struct napi;
+ struct napi_struct tx_napi;
/* General data:
* Locking: spin_lock(&np->lock); */
@@ -810,7 +811,6 @@ struct fe_priv {
union ring_type tx_ring;
u32 tx_flags;
int tx_ring_size;
- int tx_stop;
/* vlan fields */
struct vlan_group *vlangrp;
@@ -1763,10 +1763,7 @@ static int nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
empty_slots = nv_get_empty_tx_slots(np);
if (unlikely(empty_slots <= entries)) {
- spin_lock_irq(&np->lock);
netif_stop_queue(dev);
- np->tx_stop = 1;
- spin_unlock_irq(&np->lock);
return NETDEV_TX_BUSY;
}
@@ -1879,10 +1876,7 @@ static int nv_start_xmit_optimized(struct sk_buff *skb, struct net_device *dev)
empty_slots = nv_get_empty_tx_slots(np);
if (unlikely(empty_slots <= entries)) {
- spin_lock_irq(&np->lock);
netif_stop_queue(dev);
- np->tx_stop = 1;
- spin_unlock_irq(&np->lock);
return NETDEV_TX_BUSY;
}
@@ -1987,12 +1981,16 @@ static int nv_start_xmit_optimized(struct sk_buff *skb, struct net_device *dev)
*
* Caller must own np->lock.
*/
-static void nv_tx_done(struct net_device *dev)
+static int nv_tx_done(struct net_device *dev, int limit)
{
struct fe_priv *np = netdev_priv(dev);
u32 flags;
+ int orig_limit = limit;
struct ring_desc* orig_get_tx = np->get_tx.orig;
+ if (unlikely(limit < 1))
+ return 0;
+
while ((np->get_tx.orig != np->put_tx.orig) &&
!((flags = le32_to_cpu(np->get_tx.orig->flaglen)) & NV_TX_VALID)) {
@@ -2039,19 +2037,28 @@ static void nv_tx_done(struct net_device *dev)
np->get_tx.orig = np->first_tx.orig;
if (unlikely(np->get_tx_ctx++ == np->last_tx_ctx))
np->get_tx_ctx = np->first_tx_ctx;
+
+ limit--;
+ if (limit == 0)
+ break;
}
- if (unlikely((np->tx_stop == 1) && (np->get_tx.orig != orig_get_tx))) {
- np->tx_stop = 0;
+
+ if (np->get_tx.orig != orig_get_tx)
netif_wake_queue(dev);
- }
+
+ return orig_limit - limit;
}
-static void nv_tx_done_optimized(struct net_device *dev, int limit)
+static int nv_tx_done_optimized(struct net_device *dev, int limit)
{
struct fe_priv *np = netdev_priv(dev);
u32 flags;
+ int orig_limit = limit;
struct ring_desc_ex* orig_get_tx = np->get_tx.ex;
+ if (unlikely(limit < 1))
+ return 0;
+
while ((np->get_tx.ex != np->put_tx.ex) &&
!((flags = le32_to_cpu(np->get_tx.ex->flaglen)) & NV_TX_VALID) &&
(limit-- > 0)) {
@@ -2075,10 +2082,11 @@ static void nv_tx_done_optimized(struct net_device *dev, int limit)
if (unlikely(np->get_tx_ctx++ == np->last_tx_ctx))
np->get_tx_ctx = np->first_tx_ctx;
}
- if (unlikely((np->tx_stop == 1) && (np->get_tx.ex != orig_get_tx))) {
- np->tx_stop = 0;
+
+ if (np->get_tx.ex != orig_get_tx)
netif_wake_queue(dev);
- }
+
+ return orig_limit - limit;
}
/*
@@ -2149,9 +2157,9 @@ static void nv_tx_timeout(struct net_device *dev)
/* 1) stop tx engine */
nv_stop_tx(dev);
- /* 2) check that the packets were not sent already: */
+ /* 2) process all pending tx completions */
if (np->desc_ver == DESC_VER_1 || np->desc_ver == DESC_VER_2)
- nv_tx_done(dev);
+ nv_tx_done(dev, np->tx_ring_size);
else
nv_tx_done_optimized(dev, np->tx_ring_size);
@@ -2923,6 +2931,7 @@ static irqreturn_t __nv_nic_irq(struct net_device *dev, bool optimized)
u8 __iomem *base = get_hwbase(dev);
u32 events;
int handled = 0;
+ u32 upd_mask = 0;
dprintk(KERN_DEBUG "%s: nv_nic_irq%s\n", dev->name,
optimized ? "_optimized" : "");
@@ -2942,19 +2951,25 @@ static irqreturn_t __nv_nic_irq(struct net_device *dev, bool optimized)
if (!(events & np->irqmask))
goto out;
- if (optimized)
- nv_tx_done_optimized(dev, TX_WORK_PER_LOOP);
- else
- nv_tx_done(dev);
-
if (events & NVREG_IRQ_RX_ALL) {
netif_rx_schedule(dev, &np->napi);
/* Disable furthur receive irq's */
- np->irqmask &= ~NVREG_IRQ_RX_ALL;
+ upd_mask |= NVREG_IRQ_RX_ALL;
+ }
+
+ if (events & NVREG_IRQ_TX_ALL) {
+ netif_rx_schedule(dev, &np->tx_napi);
+
+ /* Disable furthur xmit irq's */
+ upd_mask |= NVREG_IRQ_TX_ALL;
+ }
+
+ if (upd_mask) {
+ np->irqmask &= ~upd_mask;
if (np->msi_flags & NV_MSI_X_ENABLED)
- writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask);
+ writel(upd_mask, base + NvRegIrqMask);
else
writel(np->irqmask, base + NvRegIrqMask);
}
@@ -3019,7 +3034,7 @@ static irqreturn_t nv_nic_irq_optimized(int foo, void *data)
static irqreturn_t nv_nic_irq_other(int foo, void *data)
{
- struct net_device *dev = (struct net_device *) data;
+ struct net_device *dev = data;
return __nv_nic_irq(dev, true);
}
@@ -3029,45 +3044,48 @@ static irqreturn_t nv_nic_irq_tx(int foo, void *data)
struct fe_priv *np = netdev_priv(dev);
u8 __iomem *base = get_hwbase(dev);
u32 events;
- int i;
- unsigned long flags;
- dprintk(KERN_DEBUG "%s: nv_nic_irq_tx\n", dev->name);
+ events = readl(base + NvRegMSIXIrqStatus) & NVREG_IRQ_TX_ALL;
+ writel(NVREG_IRQ_TX_ALL, base + NvRegMSIXIrqStatus);
- for (i=0; ; i++) {
- events = readl(base + NvRegMSIXIrqStatus) & NVREG_IRQ_TX_ALL;
- writel(NVREG_IRQ_TX_ALL, base + NvRegMSIXIrqStatus);
- dprintk(KERN_DEBUG "%s: tx irq: %08x\n", dev->name, events);
- if (!(events & np->irqmask))
- break;
+ if (events) {
+ netif_rx_schedule(dev, &np->tx_napi);
+ /* disable receive interrupts on the nic */
+ writel(NVREG_IRQ_TX_ALL, base + NvRegIrqMask);
+ pci_push(base);
+ }
+ return IRQ_HANDLED;
+}
- spin_lock_irqsave(&np->lock, flags);
- nv_tx_done_optimized(dev, TX_WORK_PER_LOOP);
- spin_unlock_irqrestore(&np->lock, flags);
+static int nv_napi_tx_poll(struct napi_struct *napi, int budget)
+{
+ struct fe_priv *np = container_of(napi, struct fe_priv, napi);
+ struct net_device *dev = np->dev;
+ u8 __iomem *base = get_hwbase(dev);
+ unsigned long flags;
+ int pkts;
- if (unlikely(events & (NVREG_IRQ_TX_ERR))) {
- dprintk(KERN_DEBUG "%s: received irq with events 0x%x. Probably TX fail.\n",
- dev->name, events);
- }
- if (unlikely(i > max_interrupt_work)) {
- spin_lock_irqsave(&np->lock, flags);
- /* disable interrupts on the nic */
- writel(NVREG_IRQ_TX_ALL, base + NvRegIrqMask);
- pci_push(base);
+ spin_lock_irqsave(&np->lock, flags);
- if (!np->in_shutdown) {
- np->nic_poll_irq |= NVREG_IRQ_TX_ALL;
- mod_timer(&np->nic_poll, jiffies + POLL_WAIT);
- }
- spin_unlock_irqrestore(&np->lock, flags);
- printk(KERN_DEBUG "%s: too many iterations (%d) in nv_nic_irq_tx.\n", dev->name, i);
- break;
- }
+ if (np->desc_ver == DESC_VER_1 || np->desc_ver == DESC_VER_2)
+ pkts = nv_tx_done(dev, budget);
+ else
+ pkts = nv_tx_done_optimized(dev, budget);
+ if (pkts < budget) {
+ /* re-enable receive interrupts */
+ __netif_rx_complete(dev, napi);
+
+ np->irqmask |= NVREG_IRQ_TX_ALL;
+ if (np->msi_flags & NV_MSI_X_ENABLED)
+ writel(NVREG_IRQ_TX_ALL, base + NvRegIrqMask);
+ else
+ writel(np->irqmask, base + NvRegIrqMask);
}
- dprintk(KERN_DEBUG "%s: nv_nic_irq_tx completed\n", dev->name);
- return IRQ_RETVAL(i);
+ spin_unlock_irqrestore(&np->lock, flags);
+
+ return pkts;
}
static int nv_napi_poll(struct napi_struct *napi, int budget)
@@ -4316,7 +4334,8 @@ static void nv_self_test(struct net_device *dev, struct ethtool_test *test, u64
if (test->flags & ETH_TEST_FL_OFFLINE) {
if (netif_running(dev)) {
- netif_stop_queue(dev);
+ netif_tx_disable(dev);
+ napi_disable(&np->tx_napi);
napi_disable(&np->napi);
netif_tx_lock_bh(dev);
spin_lock_irq(&np->lock);
@@ -4375,8 +4394,9 @@ static void nv_self_test(struct net_device *dev, struct ethtool_test *test, u64
/* restart rx engine */
nv_start_rx(dev);
nv_start_tx(dev);
- netif_start_queue(dev);
napi_enable(&np->napi);
+ napi_enable(&np->tx_napi);
+ netif_start_queue(dev);
nv_enable_hw_interrupts(dev, np->irqmask);
}
}
@@ -4603,8 +4623,9 @@ static int nv_open(struct net_device *dev)
ret = nv_update_linkspeed(dev);
nv_start_rx(dev);
nv_start_tx(dev);
- netif_start_queue(dev);
napi_enable(&np->napi);
+ napi_enable(&np->tx_napi);
+ netif_start_queue(dev);
if (ret) {
netif_carrier_on(dev);
@@ -4635,14 +4656,15 @@ static int nv_close(struct net_device *dev)
spin_lock_irq(&np->lock);
np->in_shutdown = 1;
spin_unlock_irq(&np->lock);
+ netif_stop_queue(dev);
napi_disable(&np->napi);
+ napi_disable(&np->tx_napi);
synchronize_irq(dev->irq);
del_timer_sync(&np->oom_kick);
del_timer_sync(&np->nic_poll);
del_timer_sync(&np->stats_poll);
- netif_stop_queue(dev);
spin_lock_irq(&np->lock);
nv_stop_tx(dev);
nv_stop_rx(dev);
@@ -4850,6 +4872,7 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i
dev->poll_controller = nv_poll_controller;
#endif
netif_napi_add(dev, &np->napi, nv_napi_poll, RX_WORK_PER_LOOP);
+ netif_napi_add(dev, &np->tx_napi, nv_napi_tx_poll, RX_WORK_PER_LOOP);
SET_ETHTOOL_OPS(dev, &ops);
dev->tx_timeout = nv_tx_timeout;
dev->watchdog_timeo = NV_WATCHDOG_TIMEO;
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/