On 2012-11-14 09:43, Marc Kleine-Budde wrote:
Handle incoming events (rx or tx-complete) until:
a) number of handled events == budget
or
b) no more events pending.
while (work_done < budget && interrupts_pending()) {
work_done += handle_rx(budget - work_done);
work_done += handle_tx(budget - work_done);
}
That could starve handle_tx completely though under high rx pressure, but I can
prevent that by making sure that half of the budget is held back in the first
call to handle_rx.
Then, if you have handled less events then budget:
1) call napi_complete()
then
2) enable interrupts.
if (work_done < budget) {
napi_complete();
enable_interrupts();
}
Then, return number of handled events:
return work_done;
Any additional remarks on the following implementation of the poll function?
static int grcan_poll(struct napi_struct *napi, int budget)
{
struct grcan_priv *priv = container_of(napi, struct grcan_priv, napi);
struct net_device *dev = priv->dev;
struct grcan_registers __iomem *regs = priv->regs;
unsigned long flags;
int work_done = 0;
int reserved = budget / 2;
while (work_done < budget) {
int old_work_done = work_done;
/* Prevent grcan_transmit_catch_up from starving by reserving
* part of the budget in the first iteration when calling
* grcan_receive.
*/
work_done += grcan_receive(dev, budget - reserved - work_done);
reserved = 0;
/* Catch up echo skb according to same budget, as
* grcan_transmit_catch_up can trigger echo frames being
* received.
*/
work_done += grcan_transmit_catch_up(dev, budget - work_done);
/* Break out if nothing was done */
if (work_done == old_work_done)
break;
}
if (work_done < budget) {
napi_complete(napi);
/* Guarantee no interference with a running reset that otherwise
* could turn off interrupts.
*/
spin_lock_irqsave(&priv->lock, flags);
/* Enable tx and rx interrupts again. No need to check
* priv->closing as napi_disable in grcan_close is waiting for
* scheduled napi calls to finish.
*/
grcan_set_bits(®s->imr, GRCAN_IRQ_TX | GRCAN_IRQ_RX);
spin_unlock_irqrestore(&priv->lock, flags);
}
return work_done;
}
Cheers,
Andreas
_______________________________________________
devicetree-discuss mailing list
[email protected]
https://lists.ozlabs.org/listinfo/devicetree-discuss
