[NET]: Make NAPI polling independent of struct net_device objects.
Several devices have multiple independant RX queues per net
device, and some have a single interrupt doorbell for several
queues.
In either case, it's easier to support layouts like that if the
structure representing the poll is independant from the net
device itself.
The signature of the ->poll() call back goes from:
int foo_poll(struct net_device *dev, int *budget)
to
int foo_poll(struct napi_struct *napi, int budget)
The caller is returned the number of RX packets processed (or
the number of "NAPI credits" consumed if you want to get
abstract). The callee no longer messes around bumping
dev->quota, *budget, etc. because that is all handled in the
caller upon return.
The napi_struct is to be embedded in the device driver private data
structures.
Furthermore, it is the driver's responsibility to disable all NAPI
instances in it's ->stop() device close handler. Since the
napi_struct is privatized into the driver's private data structures,
only the driver knows how to get at all of the napi_struct instances
it may have per-device.
With lots of help and suggestions from Rusty Russell, Roland Dreier,
Michael Chan, Jeff Garzik, and Jamal Hadi Salim.
Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra,
Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan.
[ Ported to current tree and all drivers converted. Integrated
Stephen's follow-on kerneldoc additions, and restored poll_list
handling to the old style to fix mutual exclusion issues. -DaveM ]
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/tsi108_eth.c b/drivers/net/tsi108_eth.c
index 1aabc91..b3069ee 100644
--- a/drivers/net/tsi108_eth.c
+++ b/drivers/net/tsi108_eth.c
@@ -79,6 +79,9 @@
void __iomem *regs; /* Base of normal regs */
void __iomem *phyregs; /* Base of register bank used for PHY access */
+ struct net_device *dev;
+ struct napi_struct napi;
+
unsigned int phy; /* Index of PHY for this interface */
unsigned int irq_num;
unsigned int id;
@@ -837,13 +840,13 @@
return done;
}
-static int tsi108_poll(struct net_device *dev, int *budget)
+static int tsi108_poll(struct napi_struct *napi, int budget)
{
- struct tsi108_prv_data *data = netdev_priv(dev);
+ struct tsi108_prv_data *data = container_of(napi, struct tsi108_prv_data, napi);
+ struct net_device *dev = data->dev;
u32 estat = TSI_READ(TSI108_EC_RXESTAT);
u32 intstat = TSI_READ(TSI108_EC_INTSTAT);
- int total_budget = min(*budget, dev->quota);
- int num_received = 0, num_filled = 0, budget_used;
+ int num_received = 0, num_filled = 0;
intstat &= TSI108_INT_RXQUEUE0 | TSI108_INT_RXTHRESH |
TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR | TSI108_INT_RXWAIT;
@@ -852,7 +855,7 @@
TSI_WRITE(TSI108_EC_INTSTAT, intstat);
if (data->rxpending || (estat & TSI108_EC_RXESTAT_Q0_DESCINT))
- num_received = tsi108_complete_rx(dev, total_budget);
+ num_received = tsi108_complete_rx(dev, budget);
/* This should normally fill no more slots than the number of
* packets received in tsi108_complete_rx(). The exception
@@ -867,7 +870,7 @@
*/
if (data->rxfree < TSI108_RXRING_LEN)
- num_filled = tsi108_refill_rx(dev, total_budget * 2);
+ num_filled = tsi108_refill_rx(dev, budget * 2);
if (intstat & TSI108_INT_RXERROR) {
u32 err = TSI_READ(TSI108_EC_RXERR);
@@ -890,14 +893,9 @@
spin_unlock_irq(&data->misclock);
}
- budget_used = max(num_received, num_filled / 2);
-
- *budget -= budget_used;
- dev->quota -= budget_used;
-
- if (budget_used != total_budget) {
+ if (num_received < budget) {
data->rxpending = 0;
- netif_rx_complete(dev);
+ netif_rx_complete(dev, napi);
TSI_WRITE(TSI108_EC_INTMASK,
TSI_READ(TSI108_EC_INTMASK)
@@ -906,14 +904,11 @@
TSI108_INT_RXOVERRUN |
TSI108_INT_RXERROR |
TSI108_INT_RXWAIT));
-
- /* IRQs are level-triggered, so no need to re-check */
- return 0;
} else {
data->rxpending = 1;
}
- return 1;
+ return num_received;
}
static void tsi108_rx_int(struct net_device *dev)
@@ -931,7 +926,7 @@
* from tsi108_check_rxring().
*/
- if (netif_rx_schedule_prep(dev)) {
+ if (netif_rx_schedule_prep(dev, &data->napi)) {
/* Mask, rather than ack, the receive interrupts. The ack
* will happen in tsi108_poll().
*/
@@ -942,7 +937,7 @@
| TSI108_INT_RXTHRESH |
TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR |
TSI108_INT_RXWAIT);
- __netif_rx_schedule(dev);
+ __netif_rx_schedule(dev, &data->napi);
} else {
if (!netif_running(dev)) {
/* This can happen if an interrupt occurs while the
@@ -1401,6 +1396,8 @@
TSI_WRITE(TSI108_EC_TXQ_PTRLOW, data->txdma);
tsi108_init_phy(dev);
+ napi_enable(&data->napi);
+
setup_timer(&data->timer, tsi108_timed_checker, (unsigned long)dev);
mod_timer(&data->timer, jiffies + 1);
@@ -1425,6 +1422,7 @@
struct tsi108_prv_data *data = netdev_priv(dev);
netif_stop_queue(dev);
+ napi_disable(&data->napi);
del_timer_sync(&data->timer);
@@ -1562,6 +1560,7 @@
printk("tsi108_eth%d: probe...\n", pdev->id);
data = netdev_priv(dev);
+ data->dev = dev;
pr_debug("tsi108_eth%d:regs:phyresgs:phy:irq_num=0x%x:0x%x:0x%x:0x%x\n",
pdev->id, einfo->regs, einfo->phyregs,
@@ -1597,9 +1596,8 @@
dev->set_mac_address = tsi108_set_mac;
dev->set_multicast_list = tsi108_set_rx_mode;
dev->get_stats = tsi108_get_stats;
- dev->poll = tsi108_poll;
+ netif_napi_add(dev, &data->napi, tsi108_poll, 64);
dev->do_ioctl = tsi108_do_ioctl;
- dev->weight = 64; /* 64 is more suitable for GigE interface - klai */
/* Apparently, the Linux networking code won't use scatter-gather
* if the hardware doesn't do checksums. However, it's faster