[dpdk-dev] [dpdk-stable] [PATCH] app/testpmd: fix Segment fault when start fwd

Iremonger, Bernard bernard.iremonger at intel.com
Mon Sep 23 16:18:28 CEST 2019


Hi ShougangX, Ferruh,

> -----Original Message-----
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Ferruh Yigit
> Sent: Monday, September 23, 2019 12:04 PM
> To: Wang, ShougangX <shougangx.wang at intel.com>; dev at dpdk.org
> Cc: Lu, Wenzhuo <wenzhuo.lu at intel.com>; Yang, Qiming
> <qiming.yang at intel.com>; stable at dpdk.org
> Subject: Re: [dpdk-dev] [dpdk-stable] [PATCH] app/testpmd: fix Segment
> fault when start fwd
> 
> On 9/17/2019 10:22 AM, Wang, ShougangX wrote:
> >> Let's assume port already stopped before calling the reset, reset
> >> will cause port to be started.
> >
> > Indeed, automatically start looks strange in this case. User
> > explicitly do the "port start" command should be better.
> >
> >> I am for user explicitly do the stop, reset and start commands,
> >> instead of reset automatically stop and start later.
> >
> > There are two reasons that it is necessary to stop automatically in "port
> reset" command:
> > 1) Even without this patch, user does not need to manually execute
> > "port stop" command before "port reset" too,  because port will be
> stopped in rte_eth_dev_reset() function.
> > But this function does not update the port status flag of testpmd. It
> > makes the port state recorded by testpmd inconsistent with the actual
> > port state. So I add stop processing in
> > reset_port() function to stop port as same as running "port stop"
> command.
> 
> I see 'rte_eth_dev_reset()' API calls the 'rte_eth_dev_stop()' but it isn't
> enough as you said and a testpmd stop is required, instead of doing in the
> function it is easy to detect port is not stopped and return an error in the
> 'reset', it is up to user to stop and do the reset.
> 
> >
> > 2) If let user to stop port manually when he wants to reset port, it is not
> easy to use than before.
> 
> I still prefer 'Explicit is better than implicit.' [1], yes it is two commands now to
> reset the port if it is already started but I don't see this is a problem.
> 
> Anyway this is not a big deal, if testpmd maintainers are OK for it we can
> continue with this...

I am inclined to agree with Ferruh, that it would be better to return an error if the port is not stopped when "port reset" is called rather that silently stopping the port.


> [1] see: The Zen of Python :)
> 
> >
> > Thanks.
> > Shougang
> >
> >
> >> -----Original Message-----
> >> From: Yigit, Ferruh
> >> Sent: Monday, September 16, 2019 11:28 PM
> >> To: Wang, ShougangX <shougangx.wang at intel.com>; dev at dpdk.org
> >> Cc: Lu, Wenzhuo <wenzhuo.lu at intel.com>; Yang, Qiming
> >> <qiming.yang at intel.com>; stable at dpdk.org
> >> Subject: Re: [dpdk-stable] [PATCH] app/testpmd: fix Segment fault
> >> when start fwd
> >>
> >> On 9/16/2019 7:37 AM, Wang, ShougangX wrote:
> >>> Hi Ferruh
> >>>
> >>> Thanks for your reply.
> >>>
> >>>> Not sure if 'reset' command should do more than it says, if there
> >>>> is a requirement that port should be stopped, why not add this
> >>>> condition with an error message so that user can stop the port in
> advance if she wants.
> >>>
> >>> Firstly, port must be stopped before reset. Usually, port is stopped
> >>> by rte_eth_dev_reset(), so testpmd does not prompt user to stop port.
> >>> Although it can stop port, but testpmd does not change its own port
> >>> status
> >> flag to "RTE_PORT_STOPPED" and it will cause "port start" to fail.
> >>> So I add this patch to stop port as same as running "port stop" command.
> >>
> >> Let's assume port already stopped before calling the reset, reset
> >> will cause port to be started.
> >> I am for user explicitly do the stop, reset and start commands,
> >> instead of reset automatically stop and start later.
> >>
> >>>
> >>>> - 'reset_port()' function has a loop inside, 'RTE_ETH_FOREACH_DEV()'
> >>>> but it works on single port, the loop looked unnecessary to me, can
> >>>> you please check and remove the loop if required?
> >>>
> >>> "port reset" supports reset all ports (testpmd > port reset all), so
> >>> this loop is
> >> necessary.
> >>
> >> Got it.
> >>
> >>>
> >>>> - I am not able to see 'reset' has been documented in
> >>>> 'doc/guides/testpmd_app_ug/testpmd_funcs.rst', it may be missing,
> >>>> can you please check and add it if required?
> >>>
> >>> OK. I will add it in the next patch.
> >>>
> >>> Thanks.
> >>> Shougang
> >>>
> >>>
> >>>> -----Original Message-----
> >>>> From: Yigit, Ferruh
> >>>> Sent: Saturday, September 14, 2019 1:35 AM
> >>>> To: Wang, ShougangX <shougangx.wang at intel.com>; dev at dpdk.org
> >>>> Cc: Lu, Wenzhuo <wenzhuo.lu at intel.com>; Yang, Qiming
> >>>> <qiming.yang at intel.com>; stable at dpdk.org
> >>>> Subject: Re: [dpdk-stable] [PATCH] app/testpmd: fix Segment fault
> >>>> when start fwd
> >>>>
> >>>> On 9/6/2019 2:28 AM, Wang ShougangX wrote:
> >>>>> This patch fixed the reset function to avoid crash when user don't
> >>>>> call port reset , port stop and port start functions as sequence.
> >>>>>
> >>>>> Fixes: 97f1e19679 ("app/testpmd: add port reset command")
> >>>>> Cc: stable at dpdk.org
> >>>>>
> >>>>> Signed-off-by: Wang ShougangX <shougangx.wang at intel.com>
> >>>>> ---
> >>>>>  app/test-pmd/testpmd.c | 6 ++++++
> >>>>>  1 file changed, 6 insertions(+)
> >>>>>
> >>>>> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> >>>>> e8e2a39b6..273a7aa02 100644
> >>>>> --- a/app/test-pmd/testpmd.c
> >>>>> +++ b/app/test-pmd/testpmd.c
> >>>>> @@ -2344,6 +2344,9 @@ reset_port(portid_t pid)
> >>>>>  	if (port_id_is_invalid(pid, ENABLED_WARN))
> >>>>>  		return;
> >>>>>
> >>>>> +	printf("Stopping ports...\n");
> >>>>> +	stop_port(pid);
> >>>>> +
> >>>>>  	printf("Resetting ports...\n");
> >>>>>
> >>>>>  	RTE_ETH_FOREACH_DEV(pi) {
> >>>>> @@ -2372,6 +2375,9 @@ reset_port(portid_t pid)
> >>>>>  		}
> >>>>>  	}
> >>>>>
> >>>>> +	printf("Starting ports...\n");
> >>>>> +	start_port(pid);
> >>>>> +
> >>>>>  	printf("Done\n");
> >>>>>  }
> >>>>
> >>>> Hi Shougang,
> >>>>
> >>>> Not sure if 'reset' command should do more than it says, if there
> >>>> is a requirement that port should be stopped, why not add this
> >>>> condition with an error message so that user can stop the port in
> advance if she wants.
> >>>>
> >>>>
> >>>> Btw, a few things related,
> >>>> - 'reset_port()' function has a loop inside, 'RTE_ETH_FOREACH_DEV()'
> >>>> but it works on single port, the loop looked unnecessary to me, can
> >>>> you please check and remove the loop if required?
> >>>>
> >>>> - I am not able to see 'reset' has been documented in
> >>>> 'doc/guides/testpmd_app_ug/testpmd_funcs.rst', it may be missing,
> >>>> can you please check and add it if required?
> >>>>
> >>>> Thanks,
> >>>> ferruh

Regards,

Bernard.




More information about the dev mailing list