lkml.org 
[lkml]   [2007]   [Oct]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
Subject[PATCH 5/6] forcedeth: use NAPI for TX completion

commit a7c00e796597b797ceac3c18e8b85c124196c5ab
Author: Jeff Garzik <jeff@garzik.org>
Date: Tue Oct 16 17:33:19 2007 -0400

[netdrvr] forcedeth: use NAPI for TX completion

A hand-rolled TX poll & work limit system was already in place, so it
was easy to convert the TX path to use NAPI.

This simplifies the code, and enables future improvements and
simplifications.

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

drivers/net/forcedeth.c | 170 +++++++++++++++++++++++++-----------------------
1 file changed, 90 insertions(+), 80 deletions(-)
a7c00e796597b797ceac3c18e8b85c124196c5ab
diff --git a/drivers/net/forcedeth.c b/drivers/net/forcedeth.c
index 32a8893..81fe016 100644
--- a/drivers/net/forcedeth.c
+++ b/drivers/net/forcedeth.c
@@ -660,6 +660,7 @@ struct fe_priv {

struct net_device *dev;
struct napi_struct napi;
+ struct napi_struct tx_napi;

/* General data:
* Locking: spin_lock(&np->lock); */
@@ -725,7 +726,6 @@ struct fe_priv {
union ring_type tx_ring;
u32 tx_flags;
int tx_ring_size;
- int tx_stop;

/* vlan fields */
struct vlan_group *vlangrp;
@@ -1732,10 +1732,7 @@ 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;
}

@@ -1848,10 +1845,7 @@ 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;
}

@@ -1956,14 +1950,15 @@ 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;
- struct ring_desc* orig_get_tx = np->get_tx.orig;
+ int tx_work = 0;

while ((np->get_tx.orig != np->put_tx.orig) &&
- !((flags = le32_to_cpu(np->get_tx.orig->flaglen)) & NV_TX_VALID)) {
+ !((flags = le32_to_cpu(np->get_tx.orig->flaglen)) & NV_TX_VALID) &&
+ (tx_work < limit)) {

dprintk(KERN_DEBUG "%s: nv_tx_done: flags 0x%x.\n",
dev->name, flags);
@@ -2008,22 +2003,25 @@ 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;
+
+ tx_work++;
}
- if (unlikely((np->tx_stop == 1) && (np->get_tx.orig != orig_get_tx))) {
- np->tx_stop = 0;
+
+ if (tx_work)
netif_wake_queue(dev);
- }
+
+ return tx_work;
}

-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;
- struct ring_desc_ex* orig_get_tx = np->get_tx.ex;
+ int tx_work = 0;

while ((np->get_tx.ex != np->put_tx.ex) &&
!((flags = le32_to_cpu(np->get_tx.ex->flaglen)) & NV_TX_VALID) &&
- (limit-- > 0)) {
+ (tx_work < limit)) {

dprintk(KERN_DEBUG "%s: nv_tx_done_optimized: flags 0x%x.\n",
dev->name, flags);
@@ -2043,11 +2041,14 @@ static void nv_tx_done_optimized(struct net_device *dev, int limit)
np->get_tx.ex = np->first_tx.ex;
if (unlikely(np->get_tx_ctx++ == np->last_tx_ctx))
np->get_tx_ctx = np->first_tx_ctx;
+
+ tx_work++;
}
- if (unlikely((np->tx_stop == 1) && (np->get_tx.ex != orig_get_tx))) {
- np->tx_stop = 0;
+
+ if (tx_work)
netif_wake_queue(dev);
- }
+
+ return tx_work;
}

/*
@@ -2120,7 +2121,7 @@ static void nv_tx_timeout(struct net_device *dev)

/* 2) check that the packets were not sent already: */
if (!nv_optimized(np))
- nv_tx_done(dev);
+ nv_tx_done(dev, np->tx_ring_size);
else
nv_tx_done_optimized(dev, np->tx_ring_size);

@@ -2888,7 +2889,7 @@ static irqreturn_t __nv_nic_irq(struct net_device *dev, bool optimized)
{
struct fe_priv *np = netdev_priv(dev);
u8 __iomem *base = np->base;
- u32 events;
+ u32 events, updmask = 0;
int i;

dprintk(KERN_DEBUG "%s: __nv_nic_irq\n", dev->name);
@@ -2905,22 +2906,22 @@ static irqreturn_t __nv_nic_irq(struct net_device *dev, bool optimized)
if (!(events & np->irqmask))
break;

- spin_lock(&np->lock);
- if (optimized)
- nv_tx_done_optimized(dev, TX_WORK_PER_LOOP);
- else
- nv_tx_done(dev);
- spin_unlock(&np->lock);
-
if (events & NVREG_IRQ_RX_ALL) {
netif_rx_schedule(dev, &np->napi);
+ updmask |= NVREG_IRQ_RX_ALL;
+ }
+
+ if (events & NVREG_IRQ_TX_ALL) {
+ netif_rx_schedule(dev, &np->tx_napi);
+ updmask |= NVREG_IRQ_TX_ALL;
+ }

- /* Disable furthur receive irq's */
+ if (updmask) {
spin_lock(&np->lock);
- np->irqmask &= ~NVREG_IRQ_RX_ALL;
+ np->irqmask &= ~updmask;

if (np->msi_flags & NV_MSI_X_ENABLED)
- writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask);
+ writel(updmask, base + NvRegIrqMask);
else
writel(np->irqmask, base + NvRegIrqMask);
spin_unlock(&np->lock);
@@ -2998,58 +2999,11 @@ static irqreturn_t nv_nic_irq_optimized(int foo, void *data)
return __nv_nic_irq(dev, true);
}

-static irqreturn_t nv_nic_irq_tx(int foo, void *data)
-{
- struct net_device *dev = 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);
-
- 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;
-
- spin_lock_irqsave(&np->lock, flags);
- nv_tx_done_optimized(dev, TX_WORK_PER_LOOP);
- spin_unlock_irqrestore(&np->lock, flags);
-
- 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);
-
- if (netif_running(dev)) {
- 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;
- }
-
- }
- dprintk(KERN_DEBUG "%s: nv_nic_irq_tx completed\n", dev->name);
-
- return IRQ_RETVAL(i);
-}
-
static int nv_napi_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);
+ u8 __iomem *base = np->base;
unsigned long flags;
int pkts, retcode;

@@ -3089,7 +3043,7 @@ static irqreturn_t nv_nic_irq_rx(int foo, void *data)
{
struct net_device *dev = data;
struct fe_priv *np = netdev_priv(dev);
- u8 __iomem *base = get_hwbase(dev);
+ u8 __iomem *base = np->base;
u32 events;

events = readl(base + NvRegMSIXIrqStatus) & NVREG_IRQ_RX_ALL;
@@ -3104,6 +3058,57 @@ static irqreturn_t nv_nic_irq_rx(int foo, void *data)
return IRQ_HANDLED;
}

+static irqreturn_t nv_nic_irq_tx(int foo, void *data)
+{
+ struct net_device *dev = data;
+ struct fe_priv *np = netdev_priv(dev);
+ u8 __iomem *base = np->base;
+ u32 events;
+
+ events = readl(base + NvRegMSIXIrqStatus) & NVREG_IRQ_TX_ALL;
+ writel(NVREG_IRQ_TX_ALL, base + NvRegMSIXIrqStatus);
+
+ 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;
+}
+
+static int nv_napi_tx_poll(struct napi_struct *napi, int budget)
+{
+ struct fe_priv *np = container_of(napi, struct fe_priv, tx_napi);
+ struct net_device *dev = np->dev;
+ u8 __iomem *base = np->base;
+ unsigned long flags;
+ int pkts;
+
+ spin_lock_irqsave(&np->lock, flags);
+
+ if (nv_optimized(np))
+ pkts = nv_tx_done_optimized(dev, budget);
+ else
+ pkts = nv_tx_done(dev, budget);
+
+ if (pkts < budget) {
+ /* re-enable receive interrupts */
+
+ __netif_rx_complete(dev, napi);
+
+ np->irqmask |= NVREG_IRQ_RX_ALL;
+ if (np->msi_flags & NV_MSI_X_ENABLED)
+ writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask);
+ else
+ writel(np->irqmask, base + NvRegIrqMask);
+ }
+
+ spin_unlock_irqrestore(&np->lock, flags);
+
+ return pkts;
+}
+
static irqreturn_t nv_nic_irq_other(int foo, void *data)
{
struct net_device *dev = data;
@@ -4348,6 +4353,7 @@ static void nv_self_test(struct net_device *dev, struct ethtool_test *test, u64
if (netif_running(dev)) {
netif_stop_queue(dev);

+ napi_disable(&np->tx_napi);
napi_disable(&np->napi);

netif_tx_lock_bh(dev);
@@ -4406,6 +4412,7 @@ static void nv_self_test(struct net_device *dev, struct ethtool_test *test, u64
netif_start_queue(dev);

napi_enable(&np->napi);
+ napi_enable(&np->tx_napi);

nv_enable_hw_interrupts(dev, np->irqmask);
}
@@ -4632,6 +4639,7 @@ static int nv_open(struct net_device *dev)
nv_start_rxtx(dev);
netif_start_queue(dev);
napi_enable(&np->napi);
+ napi_enable(&np->tx_napi);

if (ret) {
netif_carrier_on(dev);
@@ -4662,6 +4670,7 @@ static int nv_close(struct net_device *dev)
spin_lock_irq(&np->lock);
spin_unlock_irq(&np->lock);

+ napi_disable(&np->tx_napi);
napi_disable(&np->napi);
synchronize_irq(dev->irq);

@@ -4876,6 +4885,7 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i
#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, TX_WORK_PER_LOOP);

SET_ETHTOOL_OPS(dev, &ops);
dev->tx_timeout = nv_tx_timeout;
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
\
 
 \ /
  Last update: 2007-10-17 07:59    [from the cache]
©2003-2011 Jasper Spaans