Hi Andrew,
On Thu, 2019-05-30 at 14:58 +0200, Andrew Lunn wrote:
> On Thu, May 30, 2019 at 04:54:43PM +0800, Biao Huang wrote:
> > the default value of tx-frames is 25, it's too late when
> > passing tstamp to stack, then the ptp4l will fail:
> >
> > ptp4l -i eth0 -f gPTP.cfg -m
> > ptp4l: selected /dev/ptp0 as PTP clock
> > ptp4l: port 1: INITIALIZING to LISTENING on INITIALIZE
> > ptp4l: port 0: INITIALIZING to LISTENING on INITIALIZE
> > ptp4l: port 1: link up
> > ptp4l: timed out while polling for tx timestamp
> > ptp4l: increasing tx_timestamp_timeout may correct this issue,
> > but it is likely caused by a driver bug
> > ptp4l: port 1: send peer delay response failed
> > ptp4l: port 1: LISTENING to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
> >
> > ptp4l tests pass when changing the tx-frames from 25 to 1 with
> > ethtool -C option.
> > It should be fine to set tx-frames default value to 1, so ptp4l will pass
> > by default.
>
> Hi Biao
>
> What does this do to the number of interrupts? Do we get 25 times more
> interrupts? Have you done any performance tests to see if this causes
> performance regressions?
Yes, it seems tx-frames=25 can reduce interrupts.
But the tx interrupt is handled in napi now, which will disable/enable
tx interrupts at the beginning/ending of napi flow.
Here is the test result on our platform:
tx-frames=1 tx-frames=25
irq number 478514 393750
performance 904Mbits/sec 902Mbits/sec
commands for test:
"cat /proc/interrupts | grep eth0"
"iperf3 -c ipaddress -w 256K -t 60"
Thanks to napi, the interrupts will not grow 25 times more(almost the
same level), and no obvious performance degradation.
Is there anybody can double check the performance with tx-frames = 0 or
25?
>
> Andrew
Thanks.
Biao
