Hi all,
Below I've pasted a piece of e1000_intr() (Interrupt handler for
the e1000 driver). I've a few questions on this.
<<<<<<<<<<<<<<<<<<< snip >>>>>>>>>>>>>>>>>>>>
icr = E1000_READ_REG(&adapter->hw, ICR);
if (icr && netif_rx_schedule_prep(netdev)) {
e1000_irq_disable(adapter);
__netif_rx_schedule(netdev);
<<<<<<<<<<<<<<<<<<<< snip >>>>>>>>>>>>>>>>>>>>>>>>>
I don't have access to e1000 specs, but by going thru the code I assume
that the E1000_READ_REG line will clear the Interrupt Status Register
(Clear On Read).
If that is true, I have a question about what will happen if the e1000
card is sharing interrupt with some other device.
Suppose the e1000 driver is currently inside e1000_poll and we get an
interrupt (due to the other device sharing the interrupt line). Since all
the shared interrupt handlers are called by the kernel, e1000_intr will
also be called. Now we call E1000_READ_REG which will clear the Interrupt
Status Register (assuming that some status was pending to be read). We
will not enter the if() block as netif_rx_schedule_prep() will fail due to
the device already being polled.
The question is.
Won't this status be _lost_ for ever.
If the other device is generating lots of interrupts, we will have lots of
such e1000 device interrupt losses. This might have bad impact on
interrupt latency.
A better thing would be to do
adapter->icr_pending |= icr;
This way the polling routine will come to know abt the interrupt, even
though we have cleared the status register.
Thanx,
tomar
-- You have moved the mouse. Windows must be restarted for the
changes to take effect.
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
