Date: Wed, 28 Mar 2018 00:29:51 +0530 From: Reshad Patuck <reshadpatuck1@gmail.com> To: freebsd-net@freebsd.org, "Bjoern A. Zeeb" <bzeeb-lists@lists.zabbadoz.net>, Kristof Provost <kristof@sigsegv.be> Cc: FreeBSD Net <freebsd-net@freebsd.org>,Reshad Patuck <reshad@patuck.net> Subject: Re: [vnet] [epair] epair interface stops working after some time Message-ID: <CEB5C82A-33AA-4F8A-9A80-EC9CBE0300C3@gmail.com> In-Reply-To: <2D15ABDE-0C25-4C97-AEA6-0098459A2795@lists.zabbadoz.net> References: <CADaJeD2LZy=RU0vtqD7%2BdkZkUs0GKW%2B7duGDQkZ19GR-_cS=MQ@mail.gmail.com> <71B1A1BD-6FCF-47BB-9523-CCAAC03799A5@sigsegv.be> <1563563.7DUcjoHYMp@reshadlaptop.patuck.net> <C162AFB2-FF80-4640-BDC8-23B30CC22873@sigsegv.be> <1D6101CD-BCB4-4206-838B-1A75152ACCC4@sigsegv.be> <AB52ED81-F97F-471B-A1BA-F3221152A586@patuck.net> <F382A5B4-6941-43C0-9686-4B108034EBF1@patuck.net> <FDCE9FAA-1289-4E15-9239-1B6FD98B589C@sigsegv.be> <38C78C2B-87D2-4225-8F4B-A5EA48BA5D17@patuck.net> <5803CAA2-DC4A-4E49-B715-6DE472088DDD@sigsegv.be> <9CAB4522-0B0A-42BF-B9A4-BF36AFC60286@patuck.net> <7202AFF2-A314-41FE-BD13-C4C77A95E106@sigsegv.be> <2D15ABDE-0C25-4C97-AEA6-0098459A2795@lists.zabbadoz.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, =E2=80=8B @Kristof: The current value of 'net=2Elink=2Eepair=2Enetisr_maxqlen' is 2100, I will= make it 210=2E Will this require a reboot? or can I just change the sysctl and reload the= epair module? =E2=80=8B @Bjoern: here is the output to 'netstat -Q' ``` # netstat -Q Configuration: Setting Current Limit Thread count 1 1 Default queue limit 256 10240 Dispatch policy direct n/a Threads bound to CPUs disabled n/a =E2=80=8B Protocols: Name Proto QLimit Policy Dispatch Flags ip 1 256 flow default --- igmp 2 256 source default --- rtsock 3 256 source default --- arp 4 256 source default --- ether 5 256 source direct --- ip6 6 256 flow default --- epair 8 2100 cpu default CD- =E2=80=8B Workstreams: WSID CPU Name Len WMark Disp'd HDisp'd QDrops Queued Handled 0 0 ip 0 30 11409267 0 0 13574317 24983409 0 0 igmp 0 0 0 0 0 0 0 0 0 rtsock 0 1 0 0 0 42 42 0 0 arp 0 0 61109751 0 0 0 61109751 0 0 ether 0 0 115098020 0 0 0 1150980= 20 0 0 ip6 0 10 36157577 0 0 4273274 40430846 0 0 epair 0 2100 0 0 210972 303785724 3037857= 24 ``` =E2=80=8B I still have access to a machine in this state, but will need to reset it = to a working state soon=2E =E2=80=8B Please let me know if there is any information you would like me to get fr= om this machine before I reset it=2E =E2=80=8B Best, =E2=80=8B Reshad On 27 March 2018 8:18:29 PM IST, "Bjoern A=2E Zeeb" <bzeeb-lists@lists=2Ez= abbadoz=2Enet> wrote: >On 27 Mar 2018, at 14:40, Kristof Provost wrote: > >> (Re-cc freebsd-net, because this is useful information) >> >> On 27 Mar 2018, at 13:07, Reshad Patuck wrote: >>> The epair crash occurred again today running the epair module code=20 >>> with the added dtrace sdt providers=2E >>> =E2=80=8B >>> Running the same command as last time, 'dtrace -n ::epair\*:' >returns=20 >>> the following: >>> ``` >>> CPU ID FUNCTION:NAME >> =E2=80=A6 >>> 0 66499 epair_transmit_locked:enqueued >>> ``` >> >>> Looks like its filled up a queue somewhere and is dropping=20 >>> connections post that=2E >>> =E2=80=8B >>> The value of the 'error' is 55 I can see both the ifp and m structs=20 >>> but don't know what to look for in them=2E >>> >> That=E2=80=99s useful=2E Error 55 is ENOBUFS, which in IFQ_ENQUEUE() me= ans=20 >> we=E2=80=99re hitting _IF_QFULL()=2E >> There don=E2=80=99t seem to be counters for that drop though, so that m= akes=20 >> it hard to diagnose without these extra probe points=2E >> It also explains why you don=E2=80=99t really see any drop counters=20 >> incrementing=2E >> >> The fact that this queue is full presumably means that the other side > >> is not reading packets off it any more=2E >> That=E2=80=99s supposed to happen in epair_start_locked() (Look for the= =20 >> IFQ_DEQUEUE() calls)=2E >> >> It=E2=80=99s not at all clear to my how, but it looks like the receive = side=20 >> is not doing its work=2E >> >> It looks like the IFQ code is already a fallback for when the netisr=20 >> queue is full=2E >> That code might be broken, or there might be a different issue that=20 >> will just mean you=E2=80=99ll always end up in the same situation,=20 >> regardless of queue size=2E >> >> It=E2=80=99s probably worth trying to play with=20 >> =E2=80=98net=2Eroute=2Enetisr_maxqlen=E2=80=99=2E I=E2=80=99d recommend= *lowering* it, to see=20 >> if the problem happens more frequently that way=2E If it does it=E2=80= =99ll be=20 >> helpful in reproducing and trying to fix this=2E If it doesn=E2=80=99t = the=20 >> full queues is probably a consequence rather than a cause/trigger=2E >> (Of course, once you=E2=80=99ve confirmed that lowering the netisr_maxq= len=20 >> makes the problem more frequent go ahead and increase it=2E) > >netstat -Q will be useful >_______________________________________________ >freebsd-net@freebsd=2Eorg mailing list >https://lists=2Efreebsd=2Eorg/mailman/listinfo/freebsd-net >To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd=2Eorg"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CEB5C82A-33AA-4F8A-9A80-EC9CBE0300C3>