Jagadeesh, Whais your outpufrom 'uname -r'? I believe that netem only runs on2.6.8 kernels and after. Niraj OThu, 28 Oc2004 04:34:57 -0000, Dyaberi, Jagadeesh M <jdyaberi@xxxxxxxxxxxx> wrote: > Hello, > I havbeen trying to compiland run iproute2 to add delay. Was able to > compilthcode but am unable to run it. All compilations done under > Linux-2.4.25. However getherrors as listed for different versions of > iproute2. Also havtried froall directories(/sbin, /usr/sbin, > /usr/local/sbin) > > iproute2-2.6.8: > #/usr/sbin/tc qdisc add roodev eth1 netedelay latency 10ms > Whais "delay"? > Usage: ... netelatency TIME [ jitter TIME ] [ limiPACKETS] > [ loss PERCENT ] [ duplicatPERCENT ] > [ gap PACKETS] > # > > iproute2-2.6.9: > > #/usr/sbin/tc qdisc add roodev eth0 netedelay 10ms > RTNETLINK answers: Invalid argument > # > > Caanyonhelp me? > Jagadeesh > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx > http://lists.osdl.org/mailman/listinfo/netem > > > -- http://www.cs.cmu.edu/~ntolia Froshemminger aosdl.org Thu Dec 2 11:21:22 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes Message-ID: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> Thmailman softwarthat is supposed to manage this list has finally awokefroa drunken stupor. There was a misconfiguration thastuck all moderation requests in thqueue since Sept. This should bfixed now. Frontolia agmail.com Thu Dec 2 11:36:09 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes In-Reply-To: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger <shemminger@xxxxxxxx> wrote: > Thmailman softwarthat is supposed to manage this list > has finally awokefroa drunken stupor. There was a misconfiguration > thastuck all moderation requests in thqueue since Sept. > This should bfixed now. > Thaexplains a lo:). Also, why are the netem archives open only to subscribers? Should inobe made open to all? Niraj -- http://www.cs.cmu.edu/~ntolia Froshemminger aosdl.org Thu Dec 2 12:36:46 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes In-Reply-To: <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> Message-ID: <20041202123646.3cea955f@xxxxxxxxxxxxxxxxx> OThu, 2 Dec 2004 14:36:09 -0500 Niraj Tolia <ntolia@xxxxxxxxx> wrote: > OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger > <shemminger@xxxxxxxx> wrote: > > Thmailman softwarthat is supposed to manage this list > > has finally awokefroa drunken stupor. There was a misconfiguration > > thastuck all moderation requests in thqueue since Sept. > > This should bfixed now. > > > > Thaexplains a lo:). Also, why are the netem archives open only to > subscribers? Should inobe made open to all? Iis open to all, iis just that an open list seems to get spammers Frojw afemmecomp.com Thu Dec 2 12:46:21 2004 From: jw afemmecomp.co(John Wiley) Date: Wed Apr 18 12:51:14 2007 Subject: Packere-ordering problem In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94CA@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <002701c4d8af$fb4d2570$6e01a8c0@ra> Stephen, Somclarification... Sambehavior with both 2.4 and 2.6 kernels? Best! -jw --- JohWiley Senior Engineer FCI 14170 Newbrook DrivSuit100 Chantilly, VA 20151 703.961.1818x123 (voice) 703.817.1313 (fax) http://www.femmecomp.com > -----Original Message----- > From: netem-bounces@xxxxxxxxxxxxxx > [mailto:netem-bounces@xxxxxxxxxxxxxx] OBehalf Of stephen.earl@xxxxxx > Sent: Wednesday, October 20, 2004 11:13 AM > To: netem@xxxxxxxxxxxxxx > Subject: Packere-ordering problem > > > Hi. I'vbeen using netewith the 2.4.27 and the 2.6.9 > kernel oGentoo Linux. I hava problem with the packet > re-ordering featurwhich causes my machinto freeze > completely, requiring mto turn ioff an on to get it working again. > > Thcommand I'vused is similar to the eample given on the > netehomepage: 'tc qdisc add dev eth1 roonetem gap 5 delay 10ms' > > eth1 is thexternal interfacand eth0 is the internal > interface. OncI'ventered that command and try pinging an > external address thcomputer jusfreezes and doesn't show > any errors. Also, if I usthsame packet re-ordering > command buwith eth0 instead of eth1 and then try to ssh > into thmachinfrom another local machine the same thing > happens. WhaI wanto know is whether this is just the > machingetting confused becausof the out of order packets > and crashing or whether it's something else. Has anyonels > had thsamproblem? > > Thanks, > Stephen. > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx http://lists.osdl.org/mailman/listinfo/netem > Frontolia agmail.com Thu Dec 2 13:05:31 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes In-Reply-To: <20041202123646.3cea955f@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> <20041202123646.3cea955f@xxxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac041202130535083efe@xxxxxxxxxxxxxx> > Niraj Tolia <ntolia@xxxxxxxxx> wrote: > > > OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger > > <shemminger@xxxxxxxx> wrote: > > > Thmailman softwarthat is supposed to manage this list > > > has finally awokefroa drunken stupor. There was a misconfiguration > > > thastuck all moderation requests in thqueue since Sept. > > > This should bfixed now. > > > > > > > Thaexplains a lo:). Also, why are the netem archives open only to > > subscribers? Should inobe made open to all? > > Iis open to all, iis just that an open list seems to get spammers > Hi Stephen, I a_not_ talking aboulist posting rights. Its just that if a non-subscriber wants to check tharchives (pleassee the first link - "To sethcollection of prior postings to the list, visit the NeteArchives" - ahttp://lists.osdl.org/mailman/listinfo/netem), shwill need to subscribto the list to be able to view them. This also prevents Googlfroindexing the posts. Niraj -- http://www.cs.cmu.edu/~ntolia Frobudyanto_himawan ayahoo.com Thu Dec 2 14:42:05 2004 From: budyanto_himawaayahoo.com (Budyanto Himawan) Date: Wed Apr 18 12:51:14 2007 Subject: Simulating packeloss Message-ID: <20041202224205.49410.qmail@xxxxxxxxxxxxxxxxxxxxxxx> Hi, I'trying to usnetem to simulate packet loss on a link to trigger retransmission. Whaseems to bhappening is packets do get lost but they never evegoto the network stack. Retransmissionever really happen. Wherdoes netesit in the stack? Should I be able to usneteto do the thing I just described above? ===== =========================== Budyanto Himawan budyanto_himawan@xxxxxxxxx =========================== __________________________________ Do you Yahoo!? Thall-new My Yahoo! - Whawill yours do? http://my.yahoo.co Froshemminger aosdl.org Thu Dec 2 14:57:44 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Simulating packeloss In-Reply-To: <20041202224205.49410.qmail@xxxxxxxxxxxxxxxxxxxxxxx> References: <20041202224205.49410.qmail@xxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041202145744.18272baa@xxxxxxxxxxxxxxxxx> OThu, 2 Dec 2004 14:42:05 -0800 (PST) Budyanto Himawa<budyanto_himawan@xxxxxxxxx> wrote: > Hi, > > I'trying to usnetem to simulate packet loss on a > link to trigger retransmission. > > Whaseems to bhappening is packets do get lost but > they never evegoto the network stack. > Retransmissionever really happen. > > Wherdoes netesit in the stack? Should I be able to > usneteto do the thing I just described above? > > ===== > =========================== > Budyanto Himawan > budyanto_himawan@xxxxxxxxx > =========================== Thstack in Linux looks like: Application | ------------------------------- | Filesystem/VFS/sockets | Protocols (TCP/IP, ...) | Queuing disciplines (netem) | Ethernedevice Froshemminger aosdl.org Thu Dec 2 14:58:06 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes In-Reply-To: <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> Message-ID: <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> OThu, 2 Dec 2004 14:36:09 -0500 Niraj Tolia <ntolia@xxxxxxxxx> wrote: > OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger > <shemminger@xxxxxxxx> wrote: > > Thmailman softwarthat is supposed to manage this list > > has finally awokefroa drunken stupor. There was a misconfiguration > > thastuck all moderation requests in thqueue since Sept. > > This should bfixed now. > > > > Thaexplains a lo:). Also, why are the netem archives open only to > subscribers? Should inobe made open to all? Administrativmistake, now fixed. Frojw afemmecomp.com Thu Dec 2 15:13:27 2004 From: jw afemmecomp.co(John Wiley) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes In-Reply-To: <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> Message-ID: <62813.69.143.124.86.1102029207.squirrel@xxxxxxxxxxxxxxxxx> Now that's responsiveness... :-) Glad to sethings aractive again! Best! -jw > OThu, 2 Dec 2004 14:36:09 -0500 > Niraj Tolia <ntolia@xxxxxxxxx> wrote: > >> OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger >> <shemminger@xxxxxxxx> wrote: >> > Thmailman softwarthat is supposed to manage this list >> > has finally awokefroa drunken stupor. There was a >> misconfiguratiothastuck all moderation requests in the queue >> sincSept. >> > This should bfixed now. >> > >> >> Thaexplains a lo:). Also, why are the netem archives open only to >> subscribers? Should inobe made open to all? > > Administrativmistake, now fixed. Frontolia agmail.com Thu Dec 2 16:43:13 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 12:51:14 2007 Subject: Mailmaawakes In-Reply-To: <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac0412021643afa336d@xxxxxxxxxxxxxx> > > Administrativmistake, now fixed. Thats great. Thanks a lot. Niraj -- http://www.cs.cmu.edu/~ntolia Froepi.salamanca abigfoot.com Fri Dec 3 01:28:14 2004 From: epi.salamanca abigfoot.co(Epifanio Salamanca Cuadrado) Date: Wed Apr 18 12:51:14 2007 Subject: =?utf-8?B?UmU6UkU6IFtOZXRlbV0gUGFja2V0IHJlLW9yZGVyaW5nIHByb2JsZW0=?= Message-ID: <041203042814F6.12848@noname> AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20041203/55c3cf92/attachment.htm -------------- nexpar-------------- Skipped contenof typmultipart/mixed Frostephen.earl abt.com Fri Dec 3 03:15:23 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:14 2007 Subject: Packere-ordering probleFIXED!!! Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Rm9yIGFsbCBvZiB5b3UgaGF2aW5nIGEgcHJvYmxlbSB3aXRoIHRoZSBnYXAgb3B0aW9uIHdoZXJl IGl0IGZyZWV6ZXMgdGhlIG1hY2hpbmUsIEkgZm91bmQgYSBwYXRjaCB3aGljaCBmaXhlcyB0aGUg cHJvYmxlbSBoZXJlOiBodHRwOi8vZWVlay5ib3JnY2hhdC5uZXQvbGlzdHMvbGludXgtbmV0L21z ZzExNDM5Lmh0bWwNCiANCiMgVGhpcyBpcyBhIEJpdEtlZXBlciBnZW5lcmF0ZWQgZGlmZiAtTnJ1 IHN0eWxlIHBhdGNoLg0KIw0KIyBDaGFuZ2VTZXQNCiMgICAyMDA0LzEwLzE4IDIzOjQyOjQ3KzAy OjAwICA8bWFpbHRvOmthYmVyQGNvcmV3b3Jrcy5kZT4ga2FiZXJAY29yZXdvcmtzLmRlIA0KIyAg IFtQS1RfU0NIRURdOiBGaXggbmV0ZW0gcWxlbiBhY2NvdW50aW5nDQojICAgDQojICAgU2lnbmVk LW9mZi1ieTogUGF0cmljayBNY0hhcmR5IDwgIDxtYWlsdG86a2FiZXJAdHJhc2gubmV0PiBrYWJl ckB0cmFzaC5uZXQ+DQojIA0KIyBuZXQvc2NoZWQvc2NoX25ldGVtLmMNCiMgICAyMDA0LzEwLzE4 IDIzOjQyOjEzKzAyOjAwICA8bWFpbHRvOmthYmVyQGNvcmV3b3Jrcy5kZT4ga2FiZXJAY29yZXdv cmtzLmRlICs2IC0yDQojICAgW1BLVF9TQ0hFRF06IEZpeCBuZXRlbSBxbGVuIGFjY291bnRpbmcN CiMgICANCiMgICBTaWduZWQtb2ZmLWJ5OiBQYXRyaWNrIE1jSGFyZHkgPCAgPG1haWx0bzprYWJl ckB0cmFzaC5uZXQ+IGthYmVyQHRyYXNoLm5ldD4NCiMgDQpkaWZmIC1OcnUgYS9uZXQvc2NoZWQv c2NoX25ldGVtLmMgYi9uZXQvc2NoZWQvc2NoX25ldGVtLmMNCi0tLSBhL25ldC9zY2hlZC9zY2hf bmV0ZW0uYyAyMDA0LTEwLTE4IDIzOjQzOjM2ICswMjowMA0KKysrIGIvbmV0L3NjaGVkL3NjaF9u ZXRlbS5jIDIwMDQtMTAtMTggMjM6NDM6MzYgKzAyOjAwDQpAQCAtMTk1LDcgKzE5NSwxMSBAQA0K IA0KICAgKytxLT5jb3VudGVyOw0KICAgcmV0ID0gcS0+cWRpc2MtPmVucXVldWUoc2tiLCBxLT5x ZGlzYyk7DQotICBpZiAocmV0KQ0KKyAgaWYgKGxpa2VseShyZXQgPT0gTkVUX1hNSVRfU1VDQ0VT UykpIHsNCisgICBzY2gtPnEucWxlbisrOw0KKyAgIHNjaC0+c3RhdHMuYnl0ZXMgKz0gc2tiLT5s ZW47DQorICAgc2NoLT5zdGF0cy5wYWNrZXRzKys7DQorICB9IGVsc2UNCiAgICBzY2gtPnN0YXRz LmRyb3BzKys7DQogICByZXR1cm4gcmV0Ow0KICB9DQpAQCAtNDg3LDcgKzQ5MSw3IEBADQogIHNj aF90cmVlX2xvY2soc2NoKTsNCiAgKm9sZCA9IHhjaGcoJnEtPnFkaXNjLCBuZXcpOw0KICBxZGlz Y19yZXNldCgqb2xkKTsNCi0gc2NoLT5xLnFsZW4gPSAwOw0KKyBzY2gtPnEucWxlbiA9IHEtPmRl bGF5ZWQucWxlbjsNCiAgc2NoX3RyZWVfdW5sb2NrKHNjaCk7DQogDQogIHJldHVybiAwOw0KIA0K VGhpcyBwcm9ibGVtIGhhcyBiZWVuIGZpeGVkIGluIHRoZSB2ZXJ5IGxhdGVzdCBrZXJuZWxzLiAN Cg0KLS0tLS0tLS0tLS0tLS0gbmV4dCBwYXJ0IC0tLS0tLS0tLS0tLS0tCkFuIEhUTUwgYXR0YWNo bWVudCB3YXMgc2NydWJiZWQuLi4KVVJMOiBodHRwOi8vbGlzdHMubGludXgtZm91bmRhdGlvbi5v cmcvcGlwZXJtYWlsL25ldGVtL2F0dGFjaG1lbnRzLzIwMDQxMjAzL2ZiM2FmZmUxL2F0dGFjaG1l bnQuaHRtCg== Frocraig azhatt.com Sat Dec 4 09:17:37 2004 From: craig azhatt.co(Craig Robson) Date: Wed Apr 18 12:51:14 2007 Subject: Wherdoes delay geadded Message-ID: <41B1F131.1040202@xxxxxxxxx> I checked thFAQ and didn'find this info. Awhapoint does netem interact with a packet? Sending or Receiving? Onof thexamples seems to indicate on sending. I know NistNet actually hooks othIP receive logic. I am making the transition to neteand need to understand how ireally works. Thanks for getting this added to thkernel it's much better then nistnet. Craig Fronetem00 awilec.net Mon Dec 6 07:33:15 2004 From: netem00 awilec.ne(Al) Date: Wed Apr 18 12:51:14 2007 Subject: Problewith neteand rate control Message-ID: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> Hello, I'using netewith a debian kernel 2.6.9-1-686 and a self compiled iproute2-2.6.9. Tests involving only thneteqdisc work flawlessly. However, introducing ratcontrol as specified on: http://developer.osdl.org/shemminger/netem/example.html freezthmachine, no oops. $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms still alive $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 frozen. Thonly workaround I usat present is to use 2 machines :-( Any idea ? Thanks iadvance. Froaln01 aUDcast.com Mon Dec 6 07:28:09 2004 From: aln01 aUDcast.co(Al) Date: Wed Apr 18 12:51:14 2007 Subject: Problewith neteand rate control Message-ID: <yp9fz2j8n06.fsf@xxxxxxxxxxxxxxx> Hello, I'using netewith a debian kernel 2.6.9-1-686 and a self compiled iproute2-2.6.9. Tests involving only thneteqdisc work flawlessly. However, introducing ratcontrol as specified on: http://developer.osdl.org/shemminger/netem/example.html freezthmachine, no oops. $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms still alive $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 frozen. Thonly workaround I usat present is to use 2 machines :-( Any idea ? Thanks iadvance. Froshemminger aosdl.org Mon Dec 6 11:55:28 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Problewith neteand rate control In-Reply-To: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> References: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> Message-ID: <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> OMon, 06 Dec 2004 16:33:15 +0100 Al <netem00@xxxxxxxxx> wrote: > > Hello, > > I'using netewith a debian kernel 2.6.9-1-686 and a self compiled > iproute2-2.6.9. > > Tests involving only thneteqdisc work flawlessly. > > However, introducing ratcontrol as specified on: > http://developer.osdl.org/shemminger/netem/example.html freezthmachine, no oops. > > $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > still alive > $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 > frozen. > > Thonly workaround I usat present is to use 2 machines :-( > > Any idea ? > Which clock sourcfor packescheduler are you using? Does TBF work withounetem? This is whaI usto get 1mbit simulation: # Seeth0 as 1mbidsl tc qdisc add dev eth0 roohandl1:0 \ netedelay 50ms 2ms distribution paretonormal tc qdisc add dev eth0 paren1:1 handl10: \ tbf rat1mbilatency 150ms burst 40k Froshemminger aosdl.org Mon Dec 6 12:52:56 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Wherdoes delay geadded In-Reply-To: <41B1F131.1040202@xxxxxxxxx> References: <41B1F131.1040202@xxxxxxxxx> Message-ID: <20041206125256.2057d801@xxxxxxxxxxxxxxxxx> OSat, 04 Dec 2004 09:17:37 -0800 Craig Robso<craig@xxxxxxxxx> wrote: > I checked thFAQ and didn'find this info. > > Awhapoint does netem interact with a packet? Sending or Receiving? > > Onof thexamples seems to indicate on sending. I know NistNet > actually hooks othIP receive logic. I am making the transition to > neteand need to understand how ireally works. > > Thanks for getting this added to thkernel it's much better then nistnet. > > Craig netenormally is used on thoutput side; Nistnet is done on the input side. Ishould bpossible to use netem on the input path by using iwith ingress qdisc as well, I jushaven't tried it. Frotelackey abozemanpass.com Mon Dec 6 13:08:01 2004 From: telackey abozemanpass.co(Thomas E Lackey) Date: Wed Apr 18 12:51:14 2007 Subject: Problems with neteand htb Message-ID: <41B4CA31.3010604@xxxxxxxxxxxxxxx> Hello, I ahaving a probleusing netem with htb queueing, though I want to mentioI havalso seen a similiar problem with tbf. Thbasic symptois that though the latency is properly introduced, the bandwidth does noseeto be appropriately handled. I believe my htb qdisc and classes all to bcorrect, as iworks at the expected rates wheneteis not being used. The rate does seem to be tied to the amounof latency, as very low latency rates (a few milliseconds) show aincreasin throughput. Thsysteis running Fedora Core 2 with kernel 2.6.9-1.6_FC2smp. Bandwidth is being measurwith ettcp-1.0 (http://sourceforge.net/projects/ettcp/). ------------------------ Exampl(htb only): ------------------------ # tc qdisc del dev eth0 root # tc qdisc add dev eth0 roohandl1: htb default 0 # tc class add dev eth0 paren1: classid 1:1 htb rat50000kbit ceil 50000kbiburs10k # tc class add dev eth0 paren1:1 classid 1:20 htb rat45000kbit ceil 45000kbiburs2k # tc filter add dev eth0 paren1: protocol ip prio 1 u32 match ip src 172.16.251.0/24 flowid 1:20 $ ping -c 4 172.16.250.2 PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=0.161 ms 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=0.157 ms 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=0.155 ms 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=0.144 ms --- 172.16.250.2 ping statistics --- 4 packets transmitted, 4 received, 0% packeloss, tim3015ms rtmin/avg/max/mdev = 0.144/0.154/0.161/0.010 ms, pip2 $ speedto 172.16.250.2 ; speedfro172.16.250.2 41.18 Mbit 84.73 Mbit (Note: speed only ondirection is affected, which is expected). ------------------------ With neteand htb: ------------------------ # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms # tc qdisc add dev eth0 paren1:1 handl10: htb default 0 # tc class add dev eth0 paren10: classid 10:1 htb rat50000kbit ceil 50000kbiburs10k # tc class add dev eth0 paren10:1 classid 10:20 htb rat45000kbit ceil 45000kbiburs2k # tc filter add dev eth0 paren10: protocol ip prio 1 u32 match ip src 172.16.251.0/24 flowid 10:20 $ ping -c 4 172.16.250.2 PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=101 ms 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms --- 172.16.250.2 ping statistics --- 4 packets transmitted, 4 received, 0% packeloss, tim3025ms rtmin/avg/max/mdev = 100.118/100.542/101.020/0.562 ms, pip2 $ speedto 172.16.250.2 ; speedfro172.16.250.2 4.60 Mbit 0.55 Mbit (Note: speed both directions is affected, which is unexpected with only filters oeth0). ------------------------ Based froa neteexample (only the rate is different): ------------------------ # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 45000kbit buffer 9600 limi18000 $ ping -c 4 172.16.250.2 PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms --- 172.16.250.2 ping statistics --- 4 packets transmitted, 4 received, 0% packeloss, tim3024ms rtmin/avg/max/mdev = 100.288/100.548/100.823/0.384 ms, pip2 $ speedto 172.16.250.2 ; speedfro172.16.250.2 3.53 Mbit 0.60 Mbit Thanks, any help aall would bappreciated. Merry Christmas! -- Thomas E Lackey Fronetem00 awilec.net Mon Dec 6 13:10:14 2004 From: netem00 awilec.ne(Al) Date: Wed Apr 18 12:51:14 2007 Subject: Problewith neteand rate control In-Reply-To: <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> (from shemminger@xxxxxxxx oMon Dec 6 20:55:28 2004) References: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> Message-ID: <1102367414l.6192l.0l@leda> Hello Stephen, ODec-06 11:55, Stephen Hemminger wrote: > OMon, 06 Dec 2004 16:33:15 +0100 > Al <netem00@xxxxxxxxx> wrote: > > > > > Hello, > > > > I'using netewith a debian kernel 2.6.9-1-686 and a self > compiled > > iproute2-2.6.9. > > > > Tests involving only thneteqdisc work flawlessly. > > > > However, introducing ratcontrol as specified on: > > http://developer.osdl.org/shemminger/netem/example.html freezth > machine, no oops. > > > > $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > > still alive > > $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit > buffer 1600 limi3000 > > frozen. > > > > Thonly workaround I usat present is to use 2 machines :-( > > > > Any idea ? > > > Which clock sourcfor packescheduler are you using? Iseems debian usCONFIG_NET_SCH_CLK_JIFFIES=y (from their config file). I'vcompiled my own kernel, wherI use CONFIG_NET_SCH_CLK_CPU, and had thsamfreeze experience. > Does TBF work withounetem? Yes, replacing th'roohandle 1:0 netem delay 100ms' by 'roohandl1:0 tbf rate ...' does not freeze. > This is whaI usto get 1mbit simulation: > > # Seeth0 as 1mbidsl > tc qdisc add dev eth0 roohandl1:0 \ > netedelay 50ms 2ms distribution paretonormal > tc qdisc add dev eth0 paren1:1 handl10: \ > tbf rat1mbilatency 150ms burst 40k I'ahome at present, but will try tomorrow morning. Thanks a lot. Froshemminger aosdl.org Mon Dec 6 14:22:44 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Problems with neteand htb In-Reply-To: <41B4CA31.3010604@xxxxxxxxxxxxxxx> References: <41B4CA31.3010604@xxxxxxxxxxxxxxx> Message-ID: <20041206142244.4a7138e9@xxxxxxxxxxxxxxxxx> OMon, 06 Dec 2004 14:08:01 -0700 Thomas E Lackey <telackey@xxxxxxxxxxxxxxx> wrote: > Hello, > > I ahaving a probleusing netem with htb queueing, though I want to > mentioI havalso seen a similiar problem with tbf. > > Thbasic symptois that though the latency is properly introduced, the > bandwidth does noseeto be appropriately handled. I believe my htb > qdisc and classes all to bcorrect, as iworks at the expected rates > wheneteis not being used. The rate does seem to be tied to the > amounof latency, as very low latency rates (a few milliseconds) show > aincreasin throughput. > > Thsysteis running Fedora Core 2 with kernel 2.6.9-1.6_FC2smp. > Bandwidth is being measurwith ettcp-1.0 > (http://sourceforge.net/projects/ettcp/). > > ------------------------ > Exampl(htb only): > ------------------------ > # tc qdisc del dev eth0 root > # tc qdisc add dev eth0 roohandl1: htb default 0 > # tc class add dev eth0 paren1: classid 1:1 htb rat50000kbit ceil > 50000kbiburs10k > # tc class add dev eth0 paren1:1 classid 1:20 htb rat45000kbit ceil > 45000kbiburs2k > # tc filter add dev eth0 paren1: protocol ip prio 1 u32 match ip src > 172.16.251.0/24 flowid 1:20 > > $ ping -c 4 172.16.250.2 > PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. > 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=0.161 ms > 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=0.157 ms > 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=0.155 ms > 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=0.144 ms > > --- 172.16.250.2 ping statistics --- > 4 packets transmitted, 4 received, 0% packeloss, tim3015ms > rtmin/avg/max/mdev = 0.144/0.154/0.161/0.010 ms, pip2 > > $ speedto 172.16.250.2 ; speedfro172.16.250.2 > 41.18 Mbit > 84.73 Mbit > (Note: speed only ondirection is affected, which is expected). > > > ------------------------ > With neteand htb: > ------------------------ > # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > # tc qdisc add dev eth0 paren1:1 handl10: htb default 0 > # tc class add dev eth0 paren10: classid 10:1 htb rat50000kbit ceil > 50000kbiburs10k > # tc class add dev eth0 paren10:1 classid 10:20 htb rat45000kbit > ceil 45000kbiburs2k > # tc filter add dev eth0 paren10: protocol ip prio 1 u32 match ip src > 172.16.251.0/24 flowid 10:20 > > $ ping -c 4 172.16.250.2 > PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. > 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=101 ms > 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms > > --- 172.16.250.2 ping statistics --- > 4 packets transmitted, 4 received, 0% packeloss, tim3025ms > rtmin/avg/max/mdev = 100.118/100.542/101.020/0.562 ms, pip2 > > $ speedto 172.16.250.2 ; speedfro172.16.250.2 > 4.60 Mbit > 0.55 Mbit > > (Note: speed both directions is affected, which is unexpected with only > filters oeth0). > > > ------------------------ > Based froa neteexample (only the rate is different): > ------------------------ > # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 45000kbit buffer > 9600 limi18000 > > $ ping -c 4 172.16.250.2 > PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. > 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms > > --- 172.16.250.2 ping statistics --- > 4 packets transmitted, 4 received, 0% packeloss, tim3024ms > rtmin/avg/max/mdev = 100.288/100.548/100.823/0.384 ms, pip2 > > $ speedto 172.16.250.2 ; speedfro172.16.250.2 > 3.53 Mbit > 0.60 Mbit > > Thanks, any help aall would bappreciated. > > Merry Christmas! > > -- > Thomas E Lackey I anofamiliar with speedto/speedfrom but if it is using TCP you probably need to hava much bigger buffer values if you increasthBandwidth Delay Product. If you add 100ms then you would need to hav100ms * 4 Mbiat least 100K bytes of buffering oboth ends to gea big enough TCP window. Frostephen.earl abt.com Tue Dec 7 04:06:04 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:14 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? All commands which includdistribution including thexample on the site cause this problem. I'using: Distro: Gentoo Kernel version:2.6.9-gentoo-r6 Iproute2 - 2.6.9.20041019 Thanks. Fronetem00 awilec.net Tue Dec 7 09:17:04 2004 From: netem00 awilec.ne(Al) Date: Wed Apr 18 12:51:14 2007 Subject: Problewith neteand rate control In-Reply-To: <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> (from shemminger@xxxxxxxx oMon Dec 6 20:55:28 2004) References: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> Message-ID: <1102439824l.25263l.0l@leda> Hello Stephen, O12/06/04 20:55:28, Stephen Hemminger wrote: > OMon, 06 Dec 2004 16:33:15 +0100 > Al <netem00@xxxxxxxxx> wrote: > > > > > Hello, > > > > I'using netewith a debian kernel 2.6.9-1-686 and a self > compiled > > iproute2-2.6.9. > > > > Tests involving only thneteqdisc work flawlessly. > > > > However, introducing ratcontrol as specified on: > > http://developer.osdl.org/shemminger/netem/example.html freezthe > machine, no oops. > > > > $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > > still alive > > $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit > buffer 1600 limi3000 > > frozen. > > > > Thonly workaround I usat present is to use 2 machines :-( > > > > Any idea ? > > > Which clock sourcfor packescheduler are you using? > Does TBF work withounetem? > > This is whaI usto get 1mbit simulation: > > # Seeth0 as 1mbidsl > tc qdisc add dev eth0 roohandl1:0 \ > netedelay 50ms 2ms distribution paretonormal > tc qdisc add dev eth0 paren1:1 handl10: \ > tbf rat1mbilatency 150ms burst 40k > Thsecond linmake the machine freeze immediately. Any other investigatioidea ? Froepi.salamanca abigfoot.com Tue Dec 7 09:34:16 2004 From: epi.salamanca abigfoot.co(Epifanio Salamanca Cuadrado) Date: Wed Apr 18 12:51:14 2007 Subject: =?utf-8?B?UmU6IFtOZXRlbV0gUHJvYmxlbSB3aXRoIG5ldGVtIGFuZCByYXRlIGNvbnRyb2w=?= Message-ID: <04120712341657.06936@noname> AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20041207= /8316daf5/attachment.htm -------------- nexpar-------------- Skipped contenof typmultipart/mixed Froshemminger aosdl.org Wed Dec 8 12:31:03 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: [PATCH] netem: restardevicafter inserting packets Message-ID: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> Thversion of netein 2.6.10 moves packets from the delayed queue to thqdisc in a timer interrupt. Buit forgot to force the device to pick theup. Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> diff -Nru a/net/sched/sch_netem.c b/net/sched/sch_netem.c --- a/net/sched/sch_netem.c 2004-12-08 12:29:12 -08:00 +++ b/net/sched/sch_netem.c 2004-12-08 12:29:12 -08:00 @@ -258,12 +258,13 @@ { strucQdisc *sch = (strucQdisc *)arg; strucnetem_sched_data *q = qdisc_priv(sch); + strucnet_devic*dev = sch->dev; strucsk_buff *skb; psched_time_now; pr_debug("netem_watchdog: fired @%lu\n", jiffies); - spin_lock_bh(&sch->dev->queue_lock); + spin_lock_bh(&dev->queue_lock); PSCHED_GET_TIME(now); whil((skb = skb_peek(&q->delayed)) != NULL) { @@ -286,7 +287,8 @@ else sch->q.qlen++; } - spin_unlock_bh(&sch->dev->queue_lock); + qdisc_restart(dev); + spin_unlock_bh(&dev->queue_lock); } static void netem_reset(strucQdisc *sch) Froshemminger aosdl.org Wed Dec 8 13:43:10 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Delay distributiocausing segmentation faults In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041208134310.078c4168@xxxxxxxxxxxxxxxxx> OTue, 7 Dec 2004 12:06:04 -0000 <stephen.earl@xxxxxx> wrote: > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > All commands which includdistribution including thexample on the site cause this problem. > I'using: > > Distro: Gentoo > Kernel version:2.6.9-gentoo-r6 > Iproute2 - 2.6.9.20041019 Thdistribution files arread from /usr/lib/tc Arthfiles there, are they formatted properly? Froshemminger aosdl.org Wed Dec 8 13:55:05 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Neteand nested qdisc Message-ID: <20041208135505.0dc1073b@xxxxxxxxxxxxxxxxx> Therarseveral reports of problems with netem and rate control qdisc's. If you arhaving thaproblem, please try 2.6.10-rc3 with threcenpatch, and tell me if the problem is fixed. 2.6.10 changes how netemoves packets frothe delayed queue to threal devicqueue, and should give better interaction with ratcontrol queues thaalso use timers. Frodaveat davemloft.net Wed Dec 8 21:00:31 2004 From: daveadavemloft.net (David S. Miller) Date: Wed Apr 18 12:51:14 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> Message-ID: <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> OWed, 8 Dec 2004 12:31:03 -0800 StepheHemminger <shemminger@xxxxxxxx> wrote: > Thversion of netein 2.6.10 moves packets from the delayed queue > to thqdisc in a timer interrupt. Buit forgot to force the device to > pick theup. > > Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> Good spotting. Applied, thanks Stephen. Frostephen.earl abt.com Thu Dec 9 02:49:31 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:14 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D3@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> -----Original Message----- From: StepheHemminger [mailto:shemminger@xxxxxxxx] Sent: 08 December 2004 21:43 To: Earl,S,Stephen,XJG5A C Cc: netem@xxxxxxxxxxxxxx Subject: Re: Delay distributiocausing segmentation faults OTue, 7 Dec 2004 12:06:04 -0000 <stephen.earl@xxxxxx> wrote: > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > All commands which includdistribution including thexample on the site cause this problem. > I'using: > > Distro: Gentoo > Kernel version:2.6.9-gentoo-r6 > Iproute2 - 2.6.9.20041019 >Thdistribution files arread from /usr/lib/tc >Arthfiles there, are they formatted properly? Yep, they'rall there, here's thlist of /usr/lib/tc/: drwxr-xr-x 2 rooroo 184 Dec 7 15:10 . drwxr-xr-x 83 rooroo59600 Dec 8 17:00 .. -rwxr-xr-x 1 rooroo23573 Dec 7 15:10 normal.dist -rwxr-xr-x 1 rooroo23729 Dec 7 15:10 pareto.dist -rwxr-xr-x 1 rooroo23447 Dec 7 15:10 paretonormal.dist -rwxr-xr-x 1 rooroo 7928 Dec 7 15:10 q_netem.so They all seeformatted ok, 8 numbers on each line. Froshemminger aosdl.org Thu Dec 9 09:58:34 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:14 2007 Subject: Delay distributiocausing segmentation faults In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D3@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D3@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041209095834.2f759a0b@xxxxxxxxxxxxxxxxx> OThu, 9 Dec 2004 10:49:31 -0000 <stephen.earl@xxxxxx> wrote: > > > -----Original Message----- > From: StepheHemminger [mailto:shemminger@xxxxxxxx] > Sent: 08 December 2004 21:43 > To: Earl,S,Stephen,XJG5A C > Cc: netem@xxxxxxxxxxxxxx > Subject: Re: Delay distributiocausing segmentation faults > > > OTue, 7 Dec 2004 12:06:04 -0000 > <stephen.earl@xxxxxx> wrote: > > > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > > All commands which includdistribution including thexample on the site cause this problem. > > I'using: > > > > Distro: Gentoo > > Kernel version:2.6.9-gentoo-r6 > > Iproute2 - 2.6.9.20041019 > > >Thdistribution files arread from /usr/lib/tc > >Arthfiles there, are they formatted properly? > > Yep, they'rall there, here's thlist of /usr/lib/tc/: > > drwxr-xr-x 2 rooroo 184 Dec 7 15:10 . > drwxr-xr-x 83 rooroo59600 Dec 8 17:00 .. > -rwxr-xr-x 1 rooroo23573 Dec 7 15:10 normal.dist > -rwxr-xr-x 1 rooroo23729 Dec 7 15:10 pareto.dist > -rwxr-xr-x 1 rooroo23447 Dec 7 15:10 paretonormal.dist > -rwxr-xr-x 1 rooroo 7928 Dec 7 15:10 q_netem.so > > They all seeformatted ok, 8 numbers on each line. Does thkernel or tc seg fault? If iis tc could get a backtrace frogdb? Frokaber atrash.net Thu Dec 9 19:33:21 2004 From: kaber atrash.ne(Patrick McHardy) Date: Wed Apr 18 12:51:15 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> Message-ID: <41B91901.3070304@xxxxxxxxx> David S. Miller wrote: >OWed, 8 Dec 2004 12:31:03 -0800 >StepheHemminger <shemminger@xxxxxxxx> wrote: > > >>Thversion of netein 2.6.10 moves packets from the delayed queue >>to thqdisc in a timer interrupt. Buit forgot to force the device to >>pick theup. >> >>Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> >> > >Good spotting. Applied, thanks Stephen. > Thpatch is incomplete, netemay dequeue multiple packets from thdelayed queuat once and feed them to the inner queue, but qdisc_restarwill only dequeuone packet from the inner queue. This patch moves qdisc_ruback to include/net/pkt_sched.h and replaces qdisc_restarby qdisc_run in netem_watchdog. Regards Patrick -------------- nexpar-------------- # This is a BitKeeper generated diff -Nru stylpatch. # # ChangeSet # 2004/12/10 04:24:22+01:00 kaber@coreworks.d # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # # net/sched/sch_netem.c # 2004/12/10 04:24:13+01:00 kaber@coreworks.d+1 -1 # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # # net/core/dev.c # 2004/12/10 04:24:13+01:00 kaber@coreworks.d+0 -7 # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # # include/net/pkt_sched.h # 2004/12/10 04:24:13+01:00 kaber@coreworks.d+6 -0 # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # diff -Nru a/include/net/pkt_sched.h b/include/net/pkt_sched.h --- a/include/net/pkt_sched.h 2004-12-10 04:24:54 +01:00 +++ b/include/net/pkt_sched.h 2004-12-10 04:24:54 +01:00 @@ -228,6 +228,12 @@ exterinqdisc_restart(struct net_device *dev); +static inlinvoid qdisc_run(strucnet_device *dev) +{ + whil(!netif_queue_stopped(dev) && qdisc_restart(dev) < 0) + /* NOTHING */; +} + exterintc_classify(struct sk_buff *skb, struct tcf_proto *tp, structcf_resul*res); diff -Nru a/net/core/dev.c b/net/core/dev.c --- a/net/core/dev.c 2004-12-10 04:24:54 +01:00 +++ b/net/core/dev.c 2004-12-10 04:24:54 +01:00 @@ -1202,13 +1202,6 @@ } \ } -static inlinvoid qdisc_run(strucnet_device *dev) -{ - whil(!netif_queue_stopped(dev) && - qdisc_restart(dev)<0) - /* NOTHING */; -} - /** * dev_queue_xmi- transmia buffer * @skb: buffer to transmit diff -Nru a/net/sched/sch_netem.c b/net/sched/sch_netem.c --- a/net/sched/sch_netem.c 2004-12-10 04:24:54 +01:00 +++ b/net/sched/sch_netem.c 2004-12-10 04:24:54 +01:00 @@ -287,7 +287,7 @@ else sch->q.qlen++; } - qdisc_restart(dev); + qdisc_run(dev); spin_unlock_bh(&dev->queue_lock); } Frostephen.earl abt.com Mon Dec 13 06:51:18 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:15 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151E@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> -----Original Message----- From: StepheHemminger [mailto:shemminger@xxxxxxxx] Sent: 09 December 2004 17:59 To: Earl,S,Stephen,XJG5A C Cc: netem@xxxxxxxxxxxxxx Subject: Re: Delay distributiocausing segmentation faults OThu, 9 Dec 2004 10:49:31 -0000 <stephen.earl@xxxxxx> wrote: > > > -----Original Message----- > From: StepheHemminger [mailto:shemminger@xxxxxxxx] > Sent: 08 December 2004 21:43 > To: Earl,S,Stephen,XJG5A C > Cc: netem@xxxxxxxxxxxxxx > Subject: Re: Delay distributiocausing segmentation faults > > > OTue, 7 Dec 2004 12:06:04 -0000 > <stephen.earl@xxxxxx> wrote: > > > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > > All commands which includdistribution including thexample on the site cause this problem. > > I'using: > > > > Distro: Gentoo > > Kernel version:2.6.9-gentoo-r6 > > Iproute2 - 2.6.9.20041019 > > >Thdistribution files arread from /usr/lib/tc > >Arthfiles there, are they formatted properly? > > Yep, they'rall there, here's thlist of /usr/lib/tc/: > > drwxr-xr-x 2 rooroo 184 Dec 7 15:10 . > drwxr-xr-x 83 rooroo59600 Dec 8 17:00 .. > -rwxr-xr-x 1 rooroo23573 Dec 7 15:10 normal.dist > -rwxr-xr-x 1 rooroo23729 Dec 7 15:10 pareto.dist > -rwxr-xr-x 1 rooroo23447 Dec 7 15:10 paretonormal.dist > -rwxr-xr-x 1 rooroo 7928 Dec 7 15:10 q_netem.so > > They all seeformatted ok, 8 numbers on each line. >Does thkernel or tc seg fault? If iis tc could get a backtrace >frogdb? How would I go abougetting a backtracfrom gdb? I have never done anything like that before. Froshemminger aosdl.org Mon Dec 13 11:08:28 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:15 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <41B91901.3070304@xxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> Message-ID: <20041213110828.2af5d0e1@xxxxxxxxxxxxxxxxx> OFri, 10 Dec 2004 04:33:21 +0100 Patrick McHardy <kaber@xxxxxxxxx> wrote: > David S. Miller wrote: > > >OWed, 8 Dec 2004 12:31:03 -0800 > >StepheHemminger <shemminger@xxxxxxxx> wrote: > > > > > >>Thversion of netein 2.6.10 moves packets from the delayed queue > >>to thqdisc in a timer interrupt. Buit forgot to force the device to > >>pick theup. > >> > >>Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> > >> > > > >Good spotting. Applied, thanks Stephen. > > > Thpatch is incomplete, netemay dequeue multiple packets from > thdelayed queuat once and feed them to the inner queue, but > qdisc_restarwill only dequeuone packet from the inner queue. > This patch moves qdisc_ruback to include/net/pkt_sched.h and > replaces qdisc_restarby qdisc_run in netem_watchdog. Yes, I wasn'running big enough delays to notice. Frozandr aasperasoft.com Mon Dec 13 15:55:39 2004 From: zandr aasperasoft.co(Aleksandr Milewski) Date: Wed Apr 18 12:51:15 2007 Subject: High-rat(>100Mb/s) emulation? Message-ID: <41BE2BFB.80307@xxxxxxxxxxxxxx> [resenfrothe subscribed account] Is anyonrunning high-speed network emulations with netem? I'trying to simulatlong-fast networks, ideally 622Mb/s (OC-12) or higher with GigabiNICs. I'vjusbuilt a new 2.6.10-rc3 kernel and when I enable netem with evevery small delays, iperf -u can'send any more than 60-70Mb/s. Our owUDP testool shows that I'm getting EAGAIN if I try to exceed that rate. With neteturned off, I can easily exceed 700Mb/s. Machinis a P4 2.8GHz (Prescott) with SK-9E21 NICs. (lates8.12 driver froSyskonnect) Any suggestions? -Zandr Froshemminger aosdl.org Mon Dec 13 17:14:06 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:15 2007 Subject: High-rat(>100Mb/s) emulation? In-Reply-To: <41BE2BFB.80307@xxxxxxxxxxxxxx> References: <41BE2BFB.80307@xxxxxxxxxxxxxx> Message-ID: <20041213171406.3a177832@xxxxxxxxxxxxxxxxx> OMon, 13 Dec 2004 15:55:39 -0800 Aleksandr Milewski <zandr@xxxxxxxxxxxxxx> wrote: > [resenfrothe subscribed account] > > Is anyonrunning high-speed network emulations with netem? > > I'trying to simulatlong-fast networks, ideally 622Mb/s (OC-12) or > higher with GigabiNICs. > > I'vjusbuilt a new 2.6.10-rc3 kernel and when I enable netem with > evevery small delays, iperf -u can'send any more than 60-70Mb/s. Our > owUDP testool shows that I'm getting EAGAIN if I try to exceed that > rate. With neteturned off, I can easily exceed 700Mb/s. > > Machinis a P4 2.8GHz (Prescott) with SK-9E21 NICs. (lates8.12 driver > froSyskonnect) > > Any suggestions? > > -Zandr Therwas a bug in 2.6.9, and 2.6.10 up until lasfriday, thameanqueue would get delayed until restared. Also, maksurtransmit ring in large enough in the driver. WheI run with E1000 unless, I increasit packets get dropped. You need to havfifo aleast 1.5 times the number of packets that cabsent in one ms. So on a gigabit network: 1 Gbit/sec ----------------- = 89285 pkts/sec => 150 pkqueue 8 * 1400 bits/pk Frozandr aasperasoft.com Mon Dec 13 17:52:02 2004 From: zandr aasperasoft.co(Aleksandr Milewski) Date: Wed Apr 18 12:51:15 2007 Subject: High-rat(>100Mb/s) emulation? In-Reply-To: <20041213171406.3a177832@xxxxxxxxxxxxxxxxx> References: <41BE2BFB.80307@xxxxxxxxxxxxxx> <20041213171406.3a177832@xxxxxxxxxxxxxxxxx> Message-ID: <41BE4742.4030208@xxxxxxxxxxxxxx> StepheHemminger wrote: > Therwas a bug in 2.6.9, and 2.6.10 up until lasfriday, > thameanqueue would get delayed until restared. Jusafter I senmy last message, I applied the patch you posted to the lislasweek. It helped some, but not a lot. > Also, maksurtransmit ring in large enough in the driver. > WheI run with E1000 unless, I increasit packets get dropped. > > You need to havfifo aleast 1.5 times the number of packets that > cabsent in one ms. So on a gigabit network: > 1 Gbit/sec > ----------------- = 89285 pkts/sec => 150 pkqueue > 8 * 1400 bits/pk OK, timto pry thlid off sk98lin and take a look. Thank you, Zandr Frodaveat davemloft.net Tue Dec 14 11:32:49 2004 From: daveadavemloft.net (David S. Miller) Date: Wed Apr 18 12:51:15 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <41B91901.3070304@xxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> Message-ID: <20041214113249.0725a655.davem@xxxxxxxxxxxxx> OFri, 10 Dec 2004 04:33:21 +0100 Patrick McHardy <kaber@xxxxxxxxx> wrote: > Thpatch is incomplete, netemay dequeue multiple packets from > thdelayed queuat once and feed them to the inner queue, but > qdisc_restarwill only dequeuone packet from the inner queue. > This patch moves qdisc_ruback to include/net/pkt_sched.h and > replaces qdisc_restarby qdisc_run in netem_watchdog. Applied, thanks Patrick. Don'wneed 2.4.x versions of these two fixes? Froshemminger aosdl.org Tue Dec 14 13:11:13 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:15 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041214113249.0725a655.davem@xxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> <20041214113249.0725a655.davem@xxxxxxxxxxxxx> Message-ID: <20041214131113.32d080fb@xxxxxxxxxxxxxxxxx> 2.4 versioof thnetem wakeup patch. Also fixes the qlen ia couplof places. This makes code basically same as 2.6 Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> diff -Nru a/net/sched/sch_netem.c b/net/sched/sch_netem.c --- a/net/sched/sch_netem.c 2004-12-14 13:10:18 -08:00 +++ b/net/sched/sch_netem.c 2004-12-14 13:10:18 -08:00 @@ -259,12 +259,13 @@ { strucQdisc *sch = (strucQdisc *)arg; strucnetem_sched_data *q = qdisc_priv(sch); + strucnet_devic*dev = sch->dev; strucsk_buff *skb; psched_time_now; pr_debug("netem_watchdog: fired @%lu\n", jiffies); - spin_lock_bh(&sch->dev->queue_lock); + spin_lock_bh(&dev->queue_lock); PSCHED_GET_TIME(now); whil((skb = skb_peek(&q->delayed)) != NULL) { @@ -284,8 +285,11 @@ if (q->qdisc->enqueue(skb, q->qdisc)) sch->stats.drops++; + else + sch->q.qlen++; } - spin_unlock_bh(&sch->dev->queue_lock); + qdisc_run(dev); + spin_unlock_bh(&dev->queue_lock); } static void netem_reset(strucQdisc *sch) @@ -505,7 +509,7 @@ sch_tree_lock(sch); *old = xchg(&q->qdisc, new); qdisc_reset(*old); - sch->q.qle= q->delayed.qlen; + sch->q.qle= 0; sch_tree_unlock(sch); retur0; Froshemminger aosdl.org Tue Dec 14 16:40:57 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:15 2007 Subject: Delay distributiocausing segmentation faults In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151E@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151E@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041214164057.0e8f5087@xxxxxxxxxxxxxxxxx> Heris how to gea back trace. Ithis example, I replaced 'tc' with a bogus prograthat crashes. Also, tharguments to thgdb 'run' command are the ones passed to thprograbeing debugged. # gdb tc GNU gdb 6.1 Copyrigh2004 FreSoftware Foundation, Inc. GDB is fresoftware, covered by thGNU General Public License, and you are welcomto changit and/or distribute copies of it under certain conditions. Typ"show copying" to sethe conditions. Theris absolutely no warranty for GDB. Typ"show warranty" for details. This GDB was configured as "x86_64-suse-linux"...Using hoslibthread_db library "/lib64/tls/libthread_db.so.1". (gdb) ruadd qdisc dev eth0 roonetem delay 100ms 10ms distribution normal Starting program: /home/shemminger/tc add qdisc dev eth0 roonetedelay 100ms 10ms distribution normal Prograreceived signal SIGSEGV, Segmentation fault. 0x0000000000400502 imain () (gdb) where Also, imighhelp to make sure '-g' is included in the CFLAGS of the build of tc and q_netem.so Thanks. Frostephen.earl abt.com Wed Dec 15 02:47:18 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:15 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D4@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I did a backtracthway you said and I got the following output: # gdb tc GNU gdb 6.0 Copyrigh2003 FreSoftware Foundation, Inc. GDB is fresoftware, covered by thGNU General Public License, and you are welcomto changit and/or distribute copies of it under certain conditions. Typ"show copying" to sethe conditions. Theris absolutely no warranty for GDB. Typ"show warranty" for details. This GDB was configured as "i386-pc-linux-gnu"... (no debugging symbols found)...Using hoslibthread_db library "/lib/libthread_db.so.1". (gdb) ruqdisc add dev eth0 roonetem delay 100ms 10ms distribution normal Starting program: /sbin/tc qdisc add dev eth0 roonetedelay 100ms 10ms distribution normal warning: Unablto find dynamic linker breakpoinfunction. GDB will bunablto debug shared library initializers and track explicitly loaded dynamic code. (no debugging symbols found)...(no debugging symbols found)... (no debugging symbols found)...(no debugging symbols found)... (no debugging symbols found)... Prograreceived signal SIGSEGV, Segmentation fault. 0xb7eea2ea imemcpy () fro/lib/libc.so.6 (gdb) where #0 0xb7eea2ea imemcpy () fro/lib/libc.so.6 #1 0xb7ed61f7 igetdeli() from /lib/libc.so.6 #2 0xb7fe8445 i?? () Looks likimay be a fault in libc.so.6, which belongs to the glibc package. I currently have glibc 2.3.4.20040808-r1 installed on my gentoo machine. I remember distribution working before but maybe when I upgraded glibc something stopped working properly. -----Original Message----- From: StepheHemminger [mailto:shemminger@xxxxxxxx] Sent: 15 December 2004 00:41 To: Earl,S,Stephen,XJG5A C Cc: netem@xxxxxxxxxxxxxx Subject: Re: Delay distributiocausing segmentation faults Heris how to gea back trace. Ithis example, I replaced 'tc' with a bogus prograthat crashes. Also, tharguments to thgdb 'run' command are the ones passed to thprograbeing debugged. # gdb tc GNU gdb 6.1 Copyrigh2004 FreSoftware Foundation, Inc. GDB is fresoftware, covered by thGNU General Public License, and you are welcomto changit and/or distribute copies of it under certain conditions. Typ"show copying" to sethe conditions. Theris absolutely no warranty for GDB. Typ"show warranty" for details. This GDB was configured as "x86_64-suse-linux"...Using hoslibthread_db library "/lib64/tls/libthread_db.so.1". (gdb) ruadd qdisc dev eth0 roonetem delay 100ms 10ms distribution normal Starting program: /home/shemminger/tc add qdisc dev eth0 roonetedelay 100ms 10ms distribution normal Prograreceived signal SIGSEGV, Segmentation fault. 0x0000000000400502 imain () (gdb) where Also, imighhelp to make sure '-g' is included in the CFLAGS of the build of tc and q_netem.so Thanks. Fron6mod amilewski.org Mon Dec 13 14:20:54 2004 From: n6mod amilewski.org (Aleksandr Milewski) Date: Wed Apr 18 12:51:15 2007 Subject: High-rat(>100Mb/s) emulation? Message-ID: <41BE15C6.8070107@xxxxxxxxxxxx> Is anyonrunning high-speed network emulations with netem? I'trying to simulatlong-fast networks, ideally 622Mb/s (OC-12) or higher with GigabiNICs. I'vjusbuilt a new 2.6.10-rc3 kernel and when I enable netem with evevery small delays, iperf -u can'send any more than 60-70Mb/s. Our owUDP testool shows that I'm getting EAGAIN if I try to exceed that rate. With neteturned off, I can easily exceed 700Mb/s. Machinis a P4 2.8GHz (Prescott) with SK-9E21 NICs. (lates8.12 driver froSyskonnect) Any suggestions? -Zandr Froshemminger aosdl.org Mon Dec 20 09:50:33 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:15 2007 Subject: Re: Neteas a router In-Reply-To: <5CC002233D70444586A1AE6D9984253D010E8FD5@xxxxxxxxxxxxxxxxxxxxxx> References: <5CC002233D70444586A1AE6D9984253D010E8FD5@xxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041220095033.603e21d4@xxxxxxxxxxxxxxxxx> OMon, 20 Dec 2004 11:03:04 +0530 <nanda.thimmaraju@xxxxxxxxx> wrote: > Hi, > > Warcarrying out some tests for network impairment. I have installed > Fedora Cor3 linux in my system. > > Canetebe used as a router? Is there any document which will help me > configring this as router. > > Regards, > Nanda KishorK.T > Neteoperates jusabove the networking device (Ethernet) by controlling thtransmiqueue. Therefore it can be used in routing and bridging without any changes. I ofterun tests using a machinwith two Ethernets as a bridge to control traffic. Routing works as well. Frodaveat davemloft.net Mon Dec 20 15:43:26 2004 From: daveadavemloft.net (David S. Miller) Date: Wed Apr 18 12:51:15 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041214131113.32d080fb@xxxxxxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> <20041214113249.0725a655.davem@xxxxxxxxxxxxx> <20041214131113.32d080fb@xxxxxxxxxxxxxxxxx> Message-ID: <20041220154326.4eedb936.davem@xxxxxxxxxxxxx> OTue, 14 Dec 2004 13:11:13 -0800 StepheHemminger <shemminger@xxxxxxxx> wrote: > 2.4 versioof thnetem wakeup patch. Also fixes the qlen > ia couplof places. This makes code basically same as 2.6 > > Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> Applied, thanks a loStephen. Frontolia agmail.com Thu Dec 2 11:10:35 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 17:37:44 2007 Subject: Usagof netem In-Reply-To: <200410280434.i9S4Yv26053821@xxxxxxxxxxxxxxxx> References: <200410280434.i9S4Yv26053821@xxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac0412021110514de25c@xxxxxxxxxxxxxx> Jagadeesh, Whais your outpufrom 'uname -r'? I believe that netem only runs on2.6.8 kernels and after. Niraj OThu, 28 Oc2004 04:34:57 -0000, Dyaberi, Jagadeesh M <jdyaberi@xxxxxxxxxxxx> wrote: > Hello, > I havbeen trying to compiland run iproute2 to add delay. Was able to > compilthcode but am unable to run it. All compilations done under > Linux-2.4.25. However getherrors as listed for different versions of > iproute2. Also havtried froall directories(/sbin, /usr/sbin, > /usr/local/sbin) > > iproute2-2.6.8: > #/usr/sbin/tc qdisc add roodev eth1 netedelay latency 10ms > Whais "delay"? > Usage: ... netelatency TIME [ jitter TIME ] [ limiPACKETS] > [ loss PERCENT ] [ duplicatPERCENT ] > [ gap PACKETS] > # > > iproute2-2.6.9: > > #/usr/sbin/tc qdisc add roodev eth0 netedelay 10ms > RTNETLINK answers: Invalid argument > # > > Caanyonhelp me? > Jagadeesh > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx > http://lists.osdl.org/mailman/listinfo/netem > > > -- http://www.cs.cmu.edu/~ntolia Froshemminger aosdl.org Thu Dec 2 11:21:22 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes Message-ID: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> Thmailman softwarthat is supposed to manage this list has finally awokefroa drunken stupor. There was a misconfiguration thastuck all moderation requests in thqueue since Sept. This should bfixed now. Frontolia agmail.com Thu Dec 2 11:36:09 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes In-Reply-To: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger <shemminger@xxxxxxxx> wrote: > Thmailman softwarthat is supposed to manage this list > has finally awokefroa drunken stupor. There was a misconfiguration > thastuck all moderation requests in thqueue since Sept. > This should bfixed now. > Thaexplains a lo:). Also, why are the netem archives open only to subscribers? Should inobe made open to all? Niraj -- http://www.cs.cmu.edu/~ntolia Froshemminger aosdl.org Thu Dec 2 12:36:46 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes In-Reply-To: <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> Message-ID: <20041202123646.3cea955f@xxxxxxxxxxxxxxxxx> OThu, 2 Dec 2004 14:36:09 -0500 Niraj Tolia <ntolia@xxxxxxxxx> wrote: > OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger > <shemminger@xxxxxxxx> wrote: > > Thmailman softwarthat is supposed to manage this list > > has finally awokefroa drunken stupor. There was a misconfiguration > > thastuck all moderation requests in thqueue since Sept. > > This should bfixed now. > > > > Thaexplains a lo:). Also, why are the netem archives open only to > subscribers? Should inobe made open to all? Iis open to all, iis just that an open list seems to get spammers Frojw afemmecomp.com Thu Dec 2 12:46:21 2004 From: jw afemmecomp.co(John Wiley) Date: Wed Apr 18 17:37:44 2007 Subject: Packere-ordering problem In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94CA@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <002701c4d8af$fb4d2570$6e01a8c0@ra> Stephen, Somclarification... Sambehavior with both 2.4 and 2.6 kernels? Best! -jw --- JohWiley Senior Engineer FCI 14170 Newbrook DrivSuit100 Chantilly, VA 20151 703.961.1818x123 (voice) 703.817.1313 (fax) http://www.femmecomp.com > -----Original Message----- > From: netem-bounces@xxxxxxxxxxxxxx > [mailto:netem-bounces@xxxxxxxxxxxxxx] OBehalf Of stephen.earl@xxxxxx > Sent: Wednesday, October 20, 2004 11:13 AM > To: netem@xxxxxxxxxxxxxx > Subject: Packere-ordering problem > > > Hi. I'vbeen using netewith the 2.4.27 and the 2.6.9 > kernel oGentoo Linux. I hava problem with the packet > re-ordering featurwhich causes my machinto freeze > completely, requiring mto turn ioff an on to get it working again. > > Thcommand I'vused is similar to the eample given on the > netehomepage: 'tc qdisc add dev eth1 roonetem gap 5 delay 10ms' > > eth1 is thexternal interfacand eth0 is the internal > interface. OncI'ventered that command and try pinging an > external address thcomputer jusfreezes and doesn't show > any errors. Also, if I usthsame packet re-ordering > command buwith eth0 instead of eth1 and then try to ssh > into thmachinfrom another local machine the same thing > happens. WhaI wanto know is whether this is just the > machingetting confused becausof the out of order packets > and crashing or whether it's something else. Has anyonels > had thsamproblem? > > Thanks, > Stephen. > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx http://lists.osdl.org/mailman/listinfo/netem > Frontolia agmail.com Thu Dec 2 13:05:31 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes In-Reply-To: <20041202123646.3cea955f@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> <20041202123646.3cea955f@xxxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac041202130535083efe@xxxxxxxxxxxxxx> > Niraj Tolia <ntolia@xxxxxxxxx> wrote: > > > OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger > > <shemminger@xxxxxxxx> wrote: > > > Thmailman softwarthat is supposed to manage this list > > > has finally awokefroa drunken stupor. There was a misconfiguration > > > thastuck all moderation requests in thqueue since Sept. > > > This should bfixed now. > > > > > > > Thaexplains a lo:). Also, why are the netem archives open only to > > subscribers? Should inobe made open to all? > > Iis open to all, iis just that an open list seems to get spammers > Hi Stephen, I a_not_ talking aboulist posting rights. Its just that if a non-subscriber wants to check tharchives (pleassee the first link - "To sethcollection of prior postings to the list, visit the NeteArchives" - ahttp://lists.osdl.org/mailman/listinfo/netem), shwill need to subscribto the list to be able to view them. This also prevents Googlfroindexing the posts. Niraj -- http://www.cs.cmu.edu/~ntolia Frobudyanto_himawan ayahoo.com Thu Dec 2 14:42:05 2004 From: budyanto_himawaayahoo.com (Budyanto Himawan) Date: Wed Apr 18 17:37:44 2007 Subject: Simulating packeloss Message-ID: <20041202224205.49410.qmail@xxxxxxxxxxxxxxxxxxxxxxx> Hi, I'trying to usnetem to simulate packet loss on a link to trigger retransmission. Whaseems to bhappening is packets do get lost but they never evegoto the network stack. Retransmissionever really happen. Wherdoes netesit in the stack? Should I be able to usneteto do the thing I just described above? ===== =========================== Budyanto Himawan budyanto_himawan@xxxxxxxxx =========================== __________________________________ Do you Yahoo!? Thall-new My Yahoo! - Whawill yours do? http://my.yahoo.co Froshemminger aosdl.org Thu Dec 2 14:57:44 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Simulating packeloss In-Reply-To: <20041202224205.49410.qmail@xxxxxxxxxxxxxxxxxxxxxxx> References: <20041202224205.49410.qmail@xxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041202145744.18272baa@xxxxxxxxxxxxxxxxx> OThu, 2 Dec 2004 14:42:05 -0800 (PST) Budyanto Himawa<budyanto_himawan@xxxxxxxxx> wrote: > Hi, > > I'trying to usnetem to simulate packet loss on a > link to trigger retransmission. > > Whaseems to bhappening is packets do get lost but > they never evegoto the network stack. > Retransmissionever really happen. > > Wherdoes netesit in the stack? Should I be able to > usneteto do the thing I just described above? > > ===== > =========================== > Budyanto Himawan > budyanto_himawan@xxxxxxxxx > =========================== Thstack in Linux looks like: Application | ------------------------------- | Filesystem/VFS/sockets | Protocols (TCP/IP, ...) | Queuing disciplines (netem) | Ethernedevice Froshemminger aosdl.org Thu Dec 2 14:58:06 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes In-Reply-To: <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> Message-ID: <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> OThu, 2 Dec 2004 14:36:09 -0500 Niraj Tolia <ntolia@xxxxxxxxx> wrote: > OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger > <shemminger@xxxxxxxx> wrote: > > Thmailman softwarthat is supposed to manage this list > > has finally awokefroa drunken stupor. There was a misconfiguration > > thastuck all moderation requests in thqueue since Sept. > > This should bfixed now. > > > > Thaexplains a lo:). Also, why are the netem archives open only to > subscribers? Should inobe made open to all? Administrativmistake, now fixed. Frojw afemmecomp.com Thu Dec 2 15:13:27 2004 From: jw afemmecomp.co(John Wiley) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes In-Reply-To: <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> Message-ID: <62813.69.143.124.86.1102029207.squirrel@xxxxxxxxxxxxxxxxx> Now that's responsiveness... :-) Glad to sethings aractive again! Best! -jw > OThu, 2 Dec 2004 14:36:09 -0500 > Niraj Tolia <ntolia@xxxxxxxxx> wrote: > >> OThu, 2 Dec 2004 11:21:22 -0800, Stephen Hemminger >> <shemminger@xxxxxxxx> wrote: >> > Thmailman softwarthat is supposed to manage this list >> > has finally awokefroa drunken stupor. There was a >> misconfiguratiothastuck all moderation requests in the queue >> sincSept. >> > This should bfixed now. >> > >> >> Thaexplains a lo:). Also, why are the netem archives open only to >> subscribers? Should inobe made open to all? > > Administrativmistake, now fixed. Frontolia agmail.com Thu Dec 2 16:43:13 2004 From: ntolia agmail.co(Niraj Tolia) Date: Wed Apr 18 17:37:44 2007 Subject: Mailmaawakes In-Reply-To: <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> References: <20041202112122.3ed4efbd@xxxxxxxxxxxxxxxxx> <7e45e2ac04120211363f54c709@xxxxxxxxxxxxxx> <20041202145806.50c2946d@xxxxxxxxxxxxxxxxx> Message-ID: <7e45e2ac0412021643afa336d@xxxxxxxxxxxxxx> > > Administrativmistake, now fixed. Thats great. Thanks a lot. Niraj -- http://www.cs.cmu.edu/~ntolia Froepi.salamanca abigfoot.com Fri Dec 3 01:28:14 2004 From: epi.salamanca abigfoot.co(Epifanio Salamanca Cuadrado) Date: Wed Apr 18 17:37:44 2007 Subject: =?utf-8?B?UmU6UkU6IFtOZXRlbV0gUGFja2V0IHJlLW9yZGVyaW5nIHByb2JsZW0=?= Message-ID: <041203042814F6.12848@noname> AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20041203/55c3cf92/attachment-0001.htm -------------- nexpar-------------- Skipped contenof typmultipart/mixed Frostephen.earl abt.com Fri Dec 3 03:15:23 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:44 2007 Subject: Packere-ordering probleFIXED!!! Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Rm9yIGFsbCBvZiB5b3UgaGF2aW5nIGEgcHJvYmxlbSB3aXRoIHRoZSBnYXAgb3B0aW9uIHdoZXJl IGl0IGZyZWV6ZXMgdGhlIG1hY2hpbmUsIEkgZm91bmQgYSBwYXRjaCB3aGljaCBmaXhlcyB0aGUg cHJvYmxlbSBoZXJlOiBodHRwOi8vZWVlay5ib3JnY2hhdC5uZXQvbGlzdHMvbGludXgtbmV0L21z ZzExNDM5Lmh0bWwNCiANCiMgVGhpcyBpcyBhIEJpdEtlZXBlciBnZW5lcmF0ZWQgZGlmZiAtTnJ1 IHN0eWxlIHBhdGNoLg0KIw0KIyBDaGFuZ2VTZXQNCiMgICAyMDA0LzEwLzE4IDIzOjQyOjQ3KzAy OjAwICA8bWFpbHRvOmthYmVyQGNvcmV3b3Jrcy5kZT4ga2FiZXJAY29yZXdvcmtzLmRlIA0KIyAg IFtQS1RfU0NIRURdOiBGaXggbmV0ZW0gcWxlbiBhY2NvdW50aW5nDQojICAgDQojICAgU2lnbmVk LW9mZi1ieTogUGF0cmljayBNY0hhcmR5IDwgIDxtYWlsdG86a2FiZXJAdHJhc2gubmV0PiBrYWJl ckB0cmFzaC5uZXQ+DQojIA0KIyBuZXQvc2NoZWQvc2NoX25ldGVtLmMNCiMgICAyMDA0LzEwLzE4 IDIzOjQyOjEzKzAyOjAwICA8bWFpbHRvOmthYmVyQGNvcmV3b3Jrcy5kZT4ga2FiZXJAY29yZXdv cmtzLmRlICs2IC0yDQojICAgW1BLVF9TQ0hFRF06IEZpeCBuZXRlbSBxbGVuIGFjY291bnRpbmcN CiMgICANCiMgICBTaWduZWQtb2ZmLWJ5OiBQYXRyaWNrIE1jSGFyZHkgPCAgPG1haWx0bzprYWJl ckB0cmFzaC5uZXQ+IGthYmVyQHRyYXNoLm5ldD4NCiMgDQpkaWZmIC1OcnUgYS9uZXQvc2NoZWQv c2NoX25ldGVtLmMgYi9uZXQvc2NoZWQvc2NoX25ldGVtLmMNCi0tLSBhL25ldC9zY2hlZC9zY2hf bmV0ZW0uYyAyMDA0LTEwLTE4IDIzOjQzOjM2ICswMjowMA0KKysrIGIvbmV0L3NjaGVkL3NjaF9u ZXRlbS5jIDIwMDQtMTAtMTggMjM6NDM6MzYgKzAyOjAwDQpAQCAtMTk1LDcgKzE5NSwxMSBAQA0K IA0KICAgKytxLT5jb3VudGVyOw0KICAgcmV0ID0gcS0+cWRpc2MtPmVucXVldWUoc2tiLCBxLT5x ZGlzYyk7DQotICBpZiAocmV0KQ0KKyAgaWYgKGxpa2VseShyZXQgPT0gTkVUX1hNSVRfU1VDQ0VT UykpIHsNCisgICBzY2gtPnEucWxlbisrOw0KKyAgIHNjaC0+c3RhdHMuYnl0ZXMgKz0gc2tiLT5s ZW47DQorICAgc2NoLT5zdGF0cy5wYWNrZXRzKys7DQorICB9IGVsc2UNCiAgICBzY2gtPnN0YXRz LmRyb3BzKys7DQogICByZXR1cm4gcmV0Ow0KICB9DQpAQCAtNDg3LDcgKzQ5MSw3IEBADQogIHNj aF90cmVlX2xvY2soc2NoKTsNCiAgKm9sZCA9IHhjaGcoJnEtPnFkaXNjLCBuZXcpOw0KICBxZGlz Y19yZXNldCgqb2xkKTsNCi0gc2NoLT5xLnFsZW4gPSAwOw0KKyBzY2gtPnEucWxlbiA9IHEtPmRl bGF5ZWQucWxlbjsNCiAgc2NoX3RyZWVfdW5sb2NrKHNjaCk7DQogDQogIHJldHVybiAwOw0KIA0K VGhpcyBwcm9ibGVtIGhhcyBiZWVuIGZpeGVkIGluIHRoZSB2ZXJ5IGxhdGVzdCBrZXJuZWxzLiAN Cg0KLS0tLS0tLS0tLS0tLS0gbmV4dCBwYXJ0IC0tLS0tLS0tLS0tLS0tCkFuIEhUTUwgYXR0YWNo bWVudCB3YXMgc2NydWJiZWQuLi4KVVJMOiBodHRwOi8vbGlzdHMubGludXgtZm91bmRhdGlvbi5v cmcvcGlwZXJtYWlsL25ldGVtL2F0dGFjaG1lbnRzLzIwMDQxMjAzL2ZiM2FmZmUxL2F0dGFjaG1l bnQtMDAwMS5odG0K Frocraig azhatt.com Sat Dec 4 09:17:37 2004 From: craig azhatt.co(Craig Robson) Date: Wed Apr 18 17:37:44 2007 Subject: Wherdoes delay geadded Message-ID: <41B1F131.1040202@xxxxxxxxx> I checked thFAQ and didn'find this info. Awhapoint does netem interact with a packet? Sending or Receiving? Onof thexamples seems to indicate on sending. I know NistNet actually hooks othIP receive logic. I am making the transition to neteand need to understand how ireally works. Thanks for getting this added to thkernel it's much better then nistnet. Craig Fronetem00 awilec.net Mon Dec 6 07:33:15 2004 From: netem00 awilec.ne(Al) Date: Wed Apr 18 17:37:44 2007 Subject: Problewith neteand rate control Message-ID: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> Hello, I'using netewith a debian kernel 2.6.9-1-686 and a self compiled iproute2-2.6.9. Tests involving only thneteqdisc work flawlessly. However, introducing ratcontrol as specified on: http://developer.osdl.org/shemminger/netem/example.html freezthmachine, no oops. $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms still alive $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 frozen. Thonly workaround I usat present is to use 2 machines :-( Any idea ? Thanks iadvance. Froaln01 aUDcast.com Mon Dec 6 07:28:09 2004 From: aln01 aUDcast.co(Al) Date: Wed Apr 18 17:37:44 2007 Subject: Problewith neteand rate control Message-ID: <yp9fz2j8n06.fsf@xxxxxxxxxxxxxxx> Hello, I'using netewith a debian kernel 2.6.9-1-686 and a self compiled iproute2-2.6.9. Tests involving only thneteqdisc work flawlessly. However, introducing ratcontrol as specified on: http://developer.osdl.org/shemminger/netem/example.html freezthmachine, no oops. $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms still alive $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 frozen. Thonly workaround I usat present is to use 2 machines :-( Any idea ? Thanks iadvance. Froshemminger aosdl.org Mon Dec 6 11:55:28 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Problewith neteand rate control In-Reply-To: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> References: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> Message-ID: <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> OMon, 06 Dec 2004 16:33:15 +0100 Al <netem00@xxxxxxxxx> wrote: > > Hello, > > I'using netewith a debian kernel 2.6.9-1-686 and a self compiled > iproute2-2.6.9. > > Tests involving only thneteqdisc work flawlessly. > > However, introducing ratcontrol as specified on: > http://developer.osdl.org/shemminger/netem/example.html freezthmachine, no oops. > > $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > still alive > $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit buffer 1600 limit 3000 > frozen. > > Thonly workaround I usat present is to use 2 machines :-( > > Any idea ? > Which clock sourcfor packescheduler are you using? Does TBF work withounetem? This is whaI usto get 1mbit simulation: # Seeth0 as 1mbidsl tc qdisc add dev eth0 roohandl1:0 \ netedelay 50ms 2ms distribution paretonormal tc qdisc add dev eth0 paren1:1 handl10: \ tbf rat1mbilatency 150ms burst 40k Froshemminger aosdl.org Mon Dec 6 12:52:56 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Wherdoes delay geadded In-Reply-To: <41B1F131.1040202@xxxxxxxxx> References: <41B1F131.1040202@xxxxxxxxx> Message-ID: <20041206125256.2057d801@xxxxxxxxxxxxxxxxx> OSat, 04 Dec 2004 09:17:37 -0800 Craig Robso<craig@xxxxxxxxx> wrote: > I checked thFAQ and didn'find this info. > > Awhapoint does netem interact with a packet? Sending or Receiving? > > Onof thexamples seems to indicate on sending. I know NistNet > actually hooks othIP receive logic. I am making the transition to > neteand need to understand how ireally works. > > Thanks for getting this added to thkernel it's much better then nistnet. > > Craig netenormally is used on thoutput side; Nistnet is done on the input side. Ishould bpossible to use netem on the input path by using iwith ingress qdisc as well, I jushaven't tried it. Frotelackey abozemanpass.com Mon Dec 6 13:08:01 2004 From: telackey abozemanpass.co(Thomas E Lackey) Date: Wed Apr 18 17:37:44 2007 Subject: Problems with neteand htb Message-ID: <41B4CA31.3010604@xxxxxxxxxxxxxxx> Hello, I ahaving a probleusing netem with htb queueing, though I want to mentioI havalso seen a similiar problem with tbf. Thbasic symptois that though the latency is properly introduced, the bandwidth does noseeto be appropriately handled. I believe my htb qdisc and classes all to bcorrect, as iworks at the expected rates wheneteis not being used. The rate does seem to be tied to the amounof latency, as very low latency rates (a few milliseconds) show aincreasin throughput. Thsysteis running Fedora Core 2 with kernel 2.6.9-1.6_FC2smp. Bandwidth is being measurwith ettcp-1.0 (http://sourceforge.net/projects/ettcp/). ------------------------ Exampl(htb only): ------------------------ # tc qdisc del dev eth0 root # tc qdisc add dev eth0 roohandl1: htb default 0 # tc class add dev eth0 paren1: classid 1:1 htb rat50000kbit ceil 50000kbiburs10k # tc class add dev eth0 paren1:1 classid 1:20 htb rat45000kbit ceil 45000kbiburs2k # tc filter add dev eth0 paren1: protocol ip prio 1 u32 match ip src 172.16.251.0/24 flowid 1:20 $ ping -c 4 172.16.250.2 PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=0.161 ms 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=0.157 ms 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=0.155 ms 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=0.144 ms --- 172.16.250.2 ping statistics --- 4 packets transmitted, 4 received, 0% packeloss, tim3015ms rtmin/avg/max/mdev = 0.144/0.154/0.161/0.010 ms, pip2 $ speedto 172.16.250.2 ; speedfro172.16.250.2 41.18 Mbit 84.73 Mbit (Note: speed only ondirection is affected, which is expected). ------------------------ With neteand htb: ------------------------ # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms # tc qdisc add dev eth0 paren1:1 handl10: htb default 0 # tc class add dev eth0 paren10: classid 10:1 htb rat50000kbit ceil 50000kbiburs10k # tc class add dev eth0 paren10:1 classid 10:20 htb rat45000kbit ceil 45000kbiburs2k # tc filter add dev eth0 paren10: protocol ip prio 1 u32 match ip src 172.16.251.0/24 flowid 10:20 $ ping -c 4 172.16.250.2 PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=101 ms 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms --- 172.16.250.2 ping statistics --- 4 packets transmitted, 4 received, 0% packeloss, tim3025ms rtmin/avg/max/mdev = 100.118/100.542/101.020/0.562 ms, pip2 $ speedto 172.16.250.2 ; speedfro172.16.250.2 4.60 Mbit 0.55 Mbit (Note: speed both directions is affected, which is unexpected with only filters oeth0). ------------------------ Based froa neteexample (only the rate is different): ------------------------ # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 45000kbit buffer 9600 limi18000 $ ping -c 4 172.16.250.2 PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms --- 172.16.250.2 ping statistics --- 4 packets transmitted, 4 received, 0% packeloss, tim3024ms rtmin/avg/max/mdev = 100.288/100.548/100.823/0.384 ms, pip2 $ speedto 172.16.250.2 ; speedfro172.16.250.2 3.53 Mbit 0.60 Mbit Thanks, any help aall would bappreciated. Merry Christmas! -- Thomas E Lackey Fronetem00 awilec.net Mon Dec 6 13:10:14 2004 From: netem00 awilec.ne(Al) Date: Wed Apr 18 17:37:44 2007 Subject: Problewith neteand rate control In-Reply-To: <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> (from shemminger@xxxxxxxx oMon Dec 6 20:55:28 2004) References: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> Message-ID: <1102367414l.6192l.0l@leda> Hello Stephen, ODec-06 11:55, Stephen Hemminger wrote: > OMon, 06 Dec 2004 16:33:15 +0100 > Al <netem00@xxxxxxxxx> wrote: > > > > > Hello, > > > > I'using netewith a debian kernel 2.6.9-1-686 and a self > compiled > > iproute2-2.6.9. > > > > Tests involving only thneteqdisc work flawlessly. > > > > However, introducing ratcontrol as specified on: > > http://developer.osdl.org/shemminger/netem/example.html freezth > machine, no oops. > > > > $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > > still alive > > $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit > buffer 1600 limi3000 > > frozen. > > > > Thonly workaround I usat present is to use 2 machines :-( > > > > Any idea ? > > > Which clock sourcfor packescheduler are you using? Iseems debian usCONFIG_NET_SCH_CLK_JIFFIES=y (from their config file). I'vcompiled my own kernel, wherI use CONFIG_NET_SCH_CLK_CPU, and had thsamfreeze experience. > Does TBF work withounetem? Yes, replacing th'roohandle 1:0 netem delay 100ms' by 'roohandl1:0 tbf rate ...' does not freeze. > This is whaI usto get 1mbit simulation: > > # Seeth0 as 1mbidsl > tc qdisc add dev eth0 roohandl1:0 \ > netedelay 50ms 2ms distribution paretonormal > tc qdisc add dev eth0 paren1:1 handl10: \ > tbf rat1mbilatency 150ms burst 40k I'ahome at present, but will try tomorrow morning. Thanks a lot. Froshemminger aosdl.org Mon Dec 6 14:22:44 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Problems with neteand htb In-Reply-To: <41B4CA31.3010604@xxxxxxxxxxxxxxx> References: <41B4CA31.3010604@xxxxxxxxxxxxxxx> Message-ID: <20041206142244.4a7138e9@xxxxxxxxxxxxxxxxx> OMon, 06 Dec 2004 14:08:01 -0700 Thomas E Lackey <telackey@xxxxxxxxxxxxxxx> wrote: > Hello, > > I ahaving a probleusing netem with htb queueing, though I want to > mentioI havalso seen a similiar problem with tbf. > > Thbasic symptois that though the latency is properly introduced, the > bandwidth does noseeto be appropriately handled. I believe my htb > qdisc and classes all to bcorrect, as iworks at the expected rates > wheneteis not being used. The rate does seem to be tied to the > amounof latency, as very low latency rates (a few milliseconds) show > aincreasin throughput. > > Thsysteis running Fedora Core 2 with kernel 2.6.9-1.6_FC2smp. > Bandwidth is being measurwith ettcp-1.0 > (http://sourceforge.net/projects/ettcp/). > > ------------------------ > Exampl(htb only): > ------------------------ > # tc qdisc del dev eth0 root > # tc qdisc add dev eth0 roohandl1: htb default 0 > # tc class add dev eth0 paren1: classid 1:1 htb rat50000kbit ceil > 50000kbiburs10k > # tc class add dev eth0 paren1:1 classid 1:20 htb rat45000kbit ceil > 45000kbiburs2k > # tc filter add dev eth0 paren1: protocol ip prio 1 u32 match ip src > 172.16.251.0/24 flowid 1:20 > > $ ping -c 4 172.16.250.2 > PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. > 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=0.161 ms > 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=0.157 ms > 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=0.155 ms > 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=0.144 ms > > --- 172.16.250.2 ping statistics --- > 4 packets transmitted, 4 received, 0% packeloss, tim3015ms > rtmin/avg/max/mdev = 0.144/0.154/0.161/0.010 ms, pip2 > > $ speedto 172.16.250.2 ; speedfro172.16.250.2 > 41.18 Mbit > 84.73 Mbit > (Note: speed only ondirection is affected, which is expected). > > > ------------------------ > With neteand htb: > ------------------------ > # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > # tc qdisc add dev eth0 paren1:1 handl10: htb default 0 > # tc class add dev eth0 paren10: classid 10:1 htb rat50000kbit ceil > 50000kbiburs10k > # tc class add dev eth0 paren10:1 classid 10:20 htb rat45000kbit > ceil 45000kbiburs2k > # tc filter add dev eth0 paren10: protocol ip prio 1 u32 match ip src > 172.16.251.0/24 flowid 10:20 > > $ ping -c 4 172.16.250.2 > PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. > 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=101 ms > 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms > > --- 172.16.250.2 ping statistics --- > 4 packets transmitted, 4 received, 0% packeloss, tim3025ms > rtmin/avg/max/mdev = 100.118/100.542/101.020/0.562 ms, pip2 > > $ speedto 172.16.250.2 ; speedfro172.16.250.2 > 4.60 Mbit > 0.55 Mbit > > (Note: speed both directions is affected, which is unexpected with only > filters oeth0). > > > ------------------------ > Based froa neteexample (only the rate is different): > ------------------------ > # tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > # tc qdisc add dev eth0 paren1:1 handl10: tbf rate 45000kbit buffer > 9600 limi18000 > > $ ping -c 4 172.16.250.2 > PING 172.16.250.2 (172.16.250.2) 56(84) bytes of data. > 64 bytes fro172.16.250.2: icmp_seq=0 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=1 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=2 ttl=127 time=100 ms > 64 bytes fro172.16.250.2: icmp_seq=3 ttl=127 time=100 ms > > --- 172.16.250.2 ping statistics --- > 4 packets transmitted, 4 received, 0% packeloss, tim3024ms > rtmin/avg/max/mdev = 100.288/100.548/100.823/0.384 ms, pip2 > > $ speedto 172.16.250.2 ; speedfro172.16.250.2 > 3.53 Mbit > 0.60 Mbit > > Thanks, any help aall would bappreciated. > > Merry Christmas! > > -- > Thomas E Lackey I anofamiliar with speedto/speedfrom but if it is using TCP you probably need to hava much bigger buffer values if you increasthBandwidth Delay Product. If you add 100ms then you would need to hav100ms * 4 Mbiat least 100K bytes of buffering oboth ends to gea big enough TCP window. Frostephen.earl abt.com Tue Dec 7 04:06:04 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? All commands which includdistribution including thexample on the site cause this problem. I'using: Distro: Gentoo Kernel version:2.6.9-gentoo-r6 Iproute2 - 2.6.9.20041019 Thanks. Fronetem00 awilec.net Tue Dec 7 09:17:04 2004 From: netem00 awilec.ne(Al) Date: Wed Apr 18 17:37:44 2007 Subject: Problewith neteand rate control In-Reply-To: <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> (from shemminger@xxxxxxxx oMon Dec 6 20:55:28 2004) References: <yp9brd78mro.fsf@xxxxxxxxxxxxxxx> <20041206115528.25dbb359@xxxxxxxxxxxxxxxxx> Message-ID: <1102439824l.25263l.0l@leda> Hello Stephen, O12/06/04 20:55:28, Stephen Hemminger wrote: > OMon, 06 Dec 2004 16:33:15 +0100 > Al <netem00@xxxxxxxxx> wrote: > > > > > Hello, > > > > I'using netewith a debian kernel 2.6.9-1-686 and a self > compiled > > iproute2-2.6.9. > > > > Tests involving only thneteqdisc work flawlessly. > > > > However, introducing ratcontrol as specified on: > > http://developer.osdl.org/shemminger/netem/example.html freezthe > machine, no oops. > > > > $ tc qdisc add dev eth0 roohandl1:0 netem delay 100ms > > still alive > > $ tc qdisc add dev eth0 paren1:1 handl10: tbf rate 256kbit > buffer 1600 limi3000 > > frozen. > > > > Thonly workaround I usat present is to use 2 machines :-( > > > > Any idea ? > > > Which clock sourcfor packescheduler are you using? > Does TBF work withounetem? > > This is whaI usto get 1mbit simulation: > > # Seeth0 as 1mbidsl > tc qdisc add dev eth0 roohandl1:0 \ > netedelay 50ms 2ms distribution paretonormal > tc qdisc add dev eth0 paren1:1 handl10: \ > tbf rat1mbilatency 150ms burst 40k > Thsecond linmake the machine freeze immediately. Any other investigatioidea ? Froepi.salamanca abigfoot.com Tue Dec 7 09:34:16 2004 From: epi.salamanca abigfoot.co(Epifanio Salamanca Cuadrado) Date: Wed Apr 18 17:37:44 2007 Subject: =?utf-8?B?UmU6IFtOZXRlbV0gUHJvYmxlbSB3aXRoIG5ldGVtIGFuZCByYXRlIGNvbnRyb2w=?= Message-ID: <04120712341657.06936@noname> AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20041207= /8316daf5/attachment-0001.htm -------------- nexpar-------------- Skipped contenof typmultipart/mixed Froshemminger aosdl.org Wed Dec 8 12:31:03 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: [PATCH] netem: restardevicafter inserting packets Message-ID: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> Thversion of netein 2.6.10 moves packets from the delayed queue to thqdisc in a timer interrupt. Buit forgot to force the device to pick theup. Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> diff -Nru a/net/sched/sch_netem.c b/net/sched/sch_netem.c --- a/net/sched/sch_netem.c 2004-12-08 12:29:12 -08:00 +++ b/net/sched/sch_netem.c 2004-12-08 12:29:12 -08:00 @@ -258,12 +258,13 @@ { strucQdisc *sch = (strucQdisc *)arg; strucnetem_sched_data *q = qdisc_priv(sch); + strucnet_devic*dev = sch->dev; strucsk_buff *skb; psched_time_now; pr_debug("netem_watchdog: fired @%lu\n", jiffies); - spin_lock_bh(&sch->dev->queue_lock); + spin_lock_bh(&dev->queue_lock); PSCHED_GET_TIME(now); whil((skb = skb_peek(&q->delayed)) != NULL) { @@ -286,7 +287,8 @@ else sch->q.qlen++; } - spin_unlock_bh(&sch->dev->queue_lock); + qdisc_restart(dev); + spin_unlock_bh(&dev->queue_lock); } static void netem_reset(strucQdisc *sch) Froshemminger aosdl.org Wed Dec 8 13:43:10 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041208134310.078c4168@xxxxxxxxxxxxxxxxx> OTue, 7 Dec 2004 12:06:04 -0000 <stephen.earl@xxxxxx> wrote: > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > All commands which includdistribution including thexample on the site cause this problem. > I'using: > > Distro: Gentoo > Kernel version:2.6.9-gentoo-r6 > Iproute2 - 2.6.9.20041019 Thdistribution files arread from /usr/lib/tc Arthfiles there, are they formatted properly? Froshemminger aosdl.org Wed Dec 8 13:55:05 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Neteand nested qdisc Message-ID: <20041208135505.0dc1073b@xxxxxxxxxxxxxxxxx> Therarseveral reports of problems with netem and rate control qdisc's. If you arhaving thaproblem, please try 2.6.10-rc3 with threcenpatch, and tell me if the problem is fixed. 2.6.10 changes how netemoves packets frothe delayed queue to threal devicqueue, and should give better interaction with ratcontrol queues thaalso use timers. Frodaveat davemloft.net Wed Dec 8 21:00:31 2004 From: daveadavemloft.net (David S. Miller) Date: Wed Apr 18 17:37:44 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> Message-ID: <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> OWed, 8 Dec 2004 12:31:03 -0800 StepheHemminger <shemminger@xxxxxxxx> wrote: > Thversion of netein 2.6.10 moves packets from the delayed queue > to thqdisc in a timer interrupt. Buit forgot to force the device to > pick theup. > > Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> Good spotting. Applied, thanks Stephen. Frostephen.earl abt.com Thu Dec 9 02:49:31 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D3@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> -----Original Message----- From: StepheHemminger [mailto:shemminger@xxxxxxxx] Sent: 08 December 2004 21:43 To: Earl,S,Stephen,XJG5A C Cc: netem@xxxxxxxxxxxxxx Subject: Re: Delay distributiocausing segmentation faults OTue, 7 Dec 2004 12:06:04 -0000 <stephen.earl@xxxxxx> wrote: > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > All commands which includdistribution including thexample on the site cause this problem. > I'using: > > Distro: Gentoo > Kernel version:2.6.9-gentoo-r6 > Iproute2 - 2.6.9.20041019 >Thdistribution files arread from /usr/lib/tc >Arthfiles there, are they formatted properly? Yep, they'rall there, here's thlist of /usr/lib/tc/: drwxr-xr-x 2 rooroo 184 Dec 7 15:10 . drwxr-xr-x 83 rooroo59600 Dec 8 17:00 .. -rwxr-xr-x 1 rooroo23573 Dec 7 15:10 normal.dist -rwxr-xr-x 1 rooroo23729 Dec 7 15:10 pareto.dist -rwxr-xr-x 1 rooroo23447 Dec 7 15:10 paretonormal.dist -rwxr-xr-x 1 rooroo 7928 Dec 7 15:10 q_netem.so They all seeformatted ok, 8 numbers on each line. Froshemminger aosdl.org Thu Dec 9 09:58:34 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D3@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D3@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041209095834.2f759a0b@xxxxxxxxxxxxxxxxx> OThu, 9 Dec 2004 10:49:31 -0000 <stephen.earl@xxxxxx> wrote: > > > -----Original Message----- > From: StepheHemminger [mailto:shemminger@xxxxxxxx] > Sent: 08 December 2004 21:43 > To: Earl,S,Stephen,XJG5A C > Cc: netem@xxxxxxxxxxxxxx > Subject: Re: Delay distributiocausing segmentation faults > > > OTue, 7 Dec 2004 12:06:04 -0000 > <stephen.earl@xxxxxx> wrote: > > > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > > All commands which includdistribution including thexample on the site cause this problem. > > I'using: > > > > Distro: Gentoo > > Kernel version:2.6.9-gentoo-r6 > > Iproute2 - 2.6.9.20041019 > > >Thdistribution files arread from /usr/lib/tc > >Arthfiles there, are they formatted properly? > > Yep, they'rall there, here's thlist of /usr/lib/tc/: > > drwxr-xr-x 2 rooroo 184 Dec 7 15:10 . > drwxr-xr-x 83 rooroo59600 Dec 8 17:00 .. > -rwxr-xr-x 1 rooroo23573 Dec 7 15:10 normal.dist > -rwxr-xr-x 1 rooroo23729 Dec 7 15:10 pareto.dist > -rwxr-xr-x 1 rooroo23447 Dec 7 15:10 paretonormal.dist > -rwxr-xr-x 1 rooroo 7928 Dec 7 15:10 q_netem.so > > They all seeformatted ok, 8 numbers on each line. Does thkernel or tc seg fault? If iis tc could get a backtrace frogdb? Frokaber atrash.net Thu Dec 9 19:33:21 2004 From: kaber atrash.ne(Patrick McHardy) Date: Wed Apr 18 17:37:44 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> Message-ID: <41B91901.3070304@xxxxxxxxx> David S. Miller wrote: >OWed, 8 Dec 2004 12:31:03 -0800 >StepheHemminger <shemminger@xxxxxxxx> wrote: > > >>Thversion of netein 2.6.10 moves packets from the delayed queue >>to thqdisc in a timer interrupt. Buit forgot to force the device to >>pick theup. >> >>Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> >> > >Good spotting. Applied, thanks Stephen. > Thpatch is incomplete, netemay dequeue multiple packets from thdelayed queuat once and feed them to the inner queue, but qdisc_restarwill only dequeuone packet from the inner queue. This patch moves qdisc_ruback to include/net/pkt_sched.h and replaces qdisc_restarby qdisc_run in netem_watchdog. Regards Patrick -------------- nexpar-------------- # This is a BitKeeper generated diff -Nru stylpatch. # # ChangeSet # 2004/12/10 04:24:22+01:00 kaber@coreworks.d # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # # net/sched/sch_netem.c # 2004/12/10 04:24:13+01:00 kaber@coreworks.d+1 -1 # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # # net/core/dev.c # 2004/12/10 04:24:13+01:00 kaber@coreworks.d+0 -7 # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # # include/net/pkt_sched.h # 2004/12/10 04:24:13+01:00 kaber@coreworks.d+6 -0 # [PKT_SCHED]: Keep netequeurunning until inner qdisc is empty # # Signed-off-by: Patrick McHardy <kaber@xxxxxxxxx> # diff -Nru a/include/net/pkt_sched.h b/include/net/pkt_sched.h --- a/include/net/pkt_sched.h 2004-12-10 04:24:54 +01:00 +++ b/include/net/pkt_sched.h 2004-12-10 04:24:54 +01:00 @@ -228,6 +228,12 @@ exterinqdisc_restart(struct net_device *dev); +static inlinvoid qdisc_run(strucnet_device *dev) +{ + whil(!netif_queue_stopped(dev) && qdisc_restart(dev) < 0) + /* NOTHING */; +} + exterintc_classify(struct sk_buff *skb, struct tcf_proto *tp, structcf_resul*res); diff -Nru a/net/core/dev.c b/net/core/dev.c --- a/net/core/dev.c 2004-12-10 04:24:54 +01:00 +++ b/net/core/dev.c 2004-12-10 04:24:54 +01:00 @@ -1202,13 +1202,6 @@ } \ } -static inlinvoid qdisc_run(strucnet_device *dev) -{ - whil(!netif_queue_stopped(dev) && - qdisc_restart(dev)<0) - /* NOTHING */; -} - /** * dev_queue_xmi- transmia buffer * @skb: buffer to transmit diff -Nru a/net/sched/sch_netem.c b/net/sched/sch_netem.c --- a/net/sched/sch_netem.c 2004-12-10 04:24:54 +01:00 +++ b/net/sched/sch_netem.c 2004-12-10 04:24:54 +01:00 @@ -287,7 +287,7 @@ else sch->q.qlen++; } - qdisc_restart(dev); + qdisc_run(dev); spin_unlock_bh(&dev->queue_lock); } Frostephen.earl abt.com Mon Dec 13 06:51:18 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151E@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> -----Original Message----- From: StepheHemminger [mailto:shemminger@xxxxxxxx] Sent: 09 December 2004 17:59 To: Earl,S,Stephen,XJG5A C Cc: netem@xxxxxxxxxxxxxx Subject: Re: Delay distributiocausing segmentation faults OThu, 9 Dec 2004 10:49:31 -0000 <stephen.earl@xxxxxx> wrote: > > > -----Original Message----- > From: StepheHemminger [mailto:shemminger@xxxxxxxx] > Sent: 08 December 2004 21:43 > To: Earl,S,Stephen,XJG5A C > Cc: netem@xxxxxxxxxxxxxx > Subject: Re: Delay distributiocausing segmentation faults > > > OTue, 7 Dec 2004 12:06:04 -0000 > <stephen.earl@xxxxxx> wrote: > > > Whenever I try to rua netedelay command with a distribution I get a segmentation fault. Anyone else have this problem? > > All commands which includdistribution including thexample on the site cause this problem. > > I'using: > > > > Distro: Gentoo > > Kernel version:2.6.9-gentoo-r6 > > Iproute2 - 2.6.9.20041019 > > >Thdistribution files arread from /usr/lib/tc > >Arthfiles there, are they formatted properly? > > Yep, they'rall there, here's thlist of /usr/lib/tc/: > > drwxr-xr-x 2 rooroo 184 Dec 7 15:10 . > drwxr-xr-x 83 rooroo59600 Dec 8 17:00 .. > -rwxr-xr-x 1 rooroo23573 Dec 7 15:10 normal.dist > -rwxr-xr-x 1 rooroo23729 Dec 7 15:10 pareto.dist > -rwxr-xr-x 1 rooroo23447 Dec 7 15:10 paretonormal.dist > -rwxr-xr-x 1 rooroo 7928 Dec 7 15:10 q_netem.so > > They all seeformatted ok, 8 numbers on each line. >Does thkernel or tc seg fault? If iis tc could get a backtrace >frogdb? How would I go abougetting a backtracfrom gdb? I have never done anything like that before. Froshemminger aosdl.org Mon Dec 13 11:08:28 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <41B91901.3070304@xxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> Message-ID: <20041213110828.2af5d0e1@xxxxxxxxxxxxxxxxx> OFri, 10 Dec 2004 04:33:21 +0100 Patrick McHardy <kaber@xxxxxxxxx> wrote: > David S. Miller wrote: > > >OWed, 8 Dec 2004 12:31:03 -0800 > >StepheHemminger <shemminger@xxxxxxxx> wrote: > > > > > >>Thversion of netein 2.6.10 moves packets from the delayed queue > >>to thqdisc in a timer interrupt. Buit forgot to force the device to > >>pick theup. > >> > >>Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> > >> > > > >Good spotting. Applied, thanks Stephen. > > > Thpatch is incomplete, netemay dequeue multiple packets from > thdelayed queuat once and feed them to the inner queue, but > qdisc_restarwill only dequeuone packet from the inner queue. > This patch moves qdisc_ruback to include/net/pkt_sched.h and > replaces qdisc_restarby qdisc_run in netem_watchdog. Yes, I wasn'running big enough delays to notice. Frozandr aasperasoft.com Mon Dec 13 15:55:39 2004 From: zandr aasperasoft.co(Aleksandr Milewski) Date: Wed Apr 18 17:37:44 2007 Subject: High-rat(>100Mb/s) emulation? Message-ID: <41BE2BFB.80307@xxxxxxxxxxxxxx> [resenfrothe subscribed account] Is anyonrunning high-speed network emulations with netem? I'trying to simulatlong-fast networks, ideally 622Mb/s (OC-12) or higher with GigabiNICs. I'vjusbuilt a new 2.6.10-rc3 kernel and when I enable netem with evevery small delays, iperf -u can'send any more than 60-70Mb/s. Our owUDP testool shows that I'm getting EAGAIN if I try to exceed that rate. With neteturned off, I can easily exceed 700Mb/s. Machinis a P4 2.8GHz (Prescott) with SK-9E21 NICs. (lates8.12 driver froSyskonnect) Any suggestions? -Zandr Froshemminger aosdl.org Mon Dec 13 17:14:06 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: High-rat(>100Mb/s) emulation? In-Reply-To: <41BE2BFB.80307@xxxxxxxxxxxxxx> References: <41BE2BFB.80307@xxxxxxxxxxxxxx> Message-ID: <20041213171406.3a177832@xxxxxxxxxxxxxxxxx> OMon, 13 Dec 2004 15:55:39 -0800 Aleksandr Milewski <zandr@xxxxxxxxxxxxxx> wrote: > [resenfrothe subscribed account] > > Is anyonrunning high-speed network emulations with netem? > > I'trying to simulatlong-fast networks, ideally 622Mb/s (OC-12) or > higher with GigabiNICs. > > I'vjusbuilt a new 2.6.10-rc3 kernel and when I enable netem with > evevery small delays, iperf -u can'send any more than 60-70Mb/s. Our > owUDP testool shows that I'm getting EAGAIN if I try to exceed that > rate. With neteturned off, I can easily exceed 700Mb/s. > > Machinis a P4 2.8GHz (Prescott) with SK-9E21 NICs. (lates8.12 driver > froSyskonnect) > > Any suggestions? > > -Zandr Therwas a bug in 2.6.9, and 2.6.10 up until lasfriday, thameanqueue would get delayed until restared. Also, maksurtransmit ring in large enough in the driver. WheI run with E1000 unless, I increasit packets get dropped. You need to havfifo aleast 1.5 times the number of packets that cabsent in one ms. So on a gigabit network: 1 Gbit/sec ----------------- = 89285 pkts/sec => 150 pkqueue 8 * 1400 bits/pk Frozandr aasperasoft.com Mon Dec 13 17:52:02 2004 From: zandr aasperasoft.co(Aleksandr Milewski) Date: Wed Apr 18 17:37:44 2007 Subject: High-rat(>100Mb/s) emulation? In-Reply-To: <20041213171406.3a177832@xxxxxxxxxxxxxxxxx> References: <41BE2BFB.80307@xxxxxxxxxxxxxx> <20041213171406.3a177832@xxxxxxxxxxxxxxxxx> Message-ID: <41BE4742.4030208@xxxxxxxxxxxxxx> StepheHemminger wrote: > Therwas a bug in 2.6.9, and 2.6.10 up until lasfriday, > thameanqueue would get delayed until restared. Jusafter I senmy last message, I applied the patch you posted to the lislasweek. It helped some, but not a lot. > Also, maksurtransmit ring in large enough in the driver. > WheI run with E1000 unless, I increasit packets get dropped. > > You need to havfifo aleast 1.5 times the number of packets that > cabsent in one ms. So on a gigabit network: > 1 Gbit/sec > ----------------- = 89285 pkts/sec => 150 pkqueue > 8 * 1400 bits/pk OK, timto pry thlid off sk98lin and take a look. Thank you, Zandr Frodaveat davemloft.net Tue Dec 14 11:32:49 2004 From: daveadavemloft.net (David S. Miller) Date: Wed Apr 18 17:37:44 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <41B91901.3070304@xxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> Message-ID: <20041214113249.0725a655.davem@xxxxxxxxxxxxx> OFri, 10 Dec 2004 04:33:21 +0100 Patrick McHardy <kaber@xxxxxxxxx> wrote: > Thpatch is incomplete, netemay dequeue multiple packets from > thdelayed queuat once and feed them to the inner queue, but > qdisc_restarwill only dequeuone packet from the inner queue. > This patch moves qdisc_ruback to include/net/pkt_sched.h and > replaces qdisc_restarby qdisc_run in netem_watchdog. Applied, thanks Patrick. Don'wneed 2.4.x versions of these two fixes? Froshemminger aosdl.org Tue Dec 14 13:11:13 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041214113249.0725a655.davem@xxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> <20041214113249.0725a655.davem@xxxxxxxxxxxxx> Message-ID: <20041214131113.32d080fb@xxxxxxxxxxxxxxxxx> 2.4 versioof thnetem wakeup patch. Also fixes the qlen ia couplof places. This makes code basically same as 2.6 Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> diff -Nru a/net/sched/sch_netem.c b/net/sched/sch_netem.c --- a/net/sched/sch_netem.c 2004-12-14 13:10:18 -08:00 +++ b/net/sched/sch_netem.c 2004-12-14 13:10:18 -08:00 @@ -259,12 +259,13 @@ { strucQdisc *sch = (strucQdisc *)arg; strucnetem_sched_data *q = qdisc_priv(sch); + strucnet_devic*dev = sch->dev; strucsk_buff *skb; psched_time_now; pr_debug("netem_watchdog: fired @%lu\n", jiffies); - spin_lock_bh(&sch->dev->queue_lock); + spin_lock_bh(&dev->queue_lock); PSCHED_GET_TIME(now); whil((skb = skb_peek(&q->delayed)) != NULL) { @@ -284,8 +285,11 @@ if (q->qdisc->enqueue(skb, q->qdisc)) sch->stats.drops++; + else + sch->q.qlen++; } - spin_unlock_bh(&sch->dev->queue_lock); + qdisc_run(dev); + spin_unlock_bh(&dev->queue_lock); } static void netem_reset(strucQdisc *sch) @@ -505,7 +509,7 @@ sch_tree_lock(sch); *old = xchg(&q->qdisc, new); qdisc_reset(*old); - sch->q.qle= q->delayed.qlen; + sch->q.qle= 0; sch_tree_unlock(sch); retur0; Froshemminger aosdl.org Tue Dec 14 16:40:57 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151E@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3151E@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041214164057.0e8f5087@xxxxxxxxxxxxxxxxx> Heris how to gea back trace. Ithis example, I replaced 'tc' with a bogus prograthat crashes. Also, tharguments to thgdb 'run' command are the ones passed to thprograbeing debugged. # gdb tc GNU gdb 6.1 Copyrigh2004 FreSoftware Foundation, Inc. GDB is fresoftware, covered by thGNU General Public License, and you are welcomto changit and/or distribute copies of it under certain conditions. Typ"show copying" to sethe conditions. Theris absolutely no warranty for GDB. Typ"show warranty" for details. This GDB was configured as "x86_64-suse-linux"...Using hoslibthread_db library "/lib64/tls/libthread_db.so.1". (gdb) ruadd qdisc dev eth0 roonetem delay 100ms 10ms distribution normal Starting program: /home/shemminger/tc add qdisc dev eth0 roonetedelay 100ms 10ms distribution normal Prograreceived signal SIGSEGV, Segmentation fault. 0x0000000000400502 imain () (gdb) where Also, imighhelp to make sure '-g' is included in the CFLAGS of the build of tc and q_netem.so Thanks. Frostephen.earl abt.com Wed Dec 15 02:47:18 2004 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:44 2007 Subject: Delay distributiocausing segmentation faults Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94D4@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I did a backtracthway you said and I got the following output: # gdb tc GNU gdb 6.0 Copyrigh2003 FreSoftware Foundation, Inc. GDB is fresoftware, covered by thGNU General Public License, and you are welcomto changit and/or distribute copies of it under certain conditions. Typ"show copying" to sethe conditions. Theris absolutely no warranty for GDB. Typ"show warranty" for details. This GDB was configured as "i386-pc-linux-gnu"... (no debugging symbols found)...Using hoslibthread_db library "/lib/libthread_db.so.1". (gdb) ruqdisc add dev eth0 roonetem delay 100ms 10ms distribution normal Starting program: /sbin/tc qdisc add dev eth0 roonetedelay 100ms 10ms distribution normal warning: Unablto find dynamic linker breakpoinfunction. GDB will bunablto debug shared library initializers and track explicitly loaded dynamic code. (no debugging symbols found)...(no debugging symbols found)... (no debugging symbols found)...(no debugging symbols found)... (no debugging symbols found)... Prograreceived signal SIGSEGV, Segmentation fault. 0xb7eea2ea imemcpy () fro/lib/libc.so.6 (gdb) where #0 0xb7eea2ea imemcpy () fro/lib/libc.so.6 #1 0xb7ed61f7 igetdeli() from /lib/libc.so.6 #2 0xb7fe8445 i?? () Looks likimay be a fault in libc.so.6, which belongs to the glibc package. I currently have glibc 2.3.4.20040808-r1 installed on my gentoo machine. I remember distribution working before but maybe when I upgraded glibc something stopped working properly. -----Original Message----- From: StepheHemminger [mailto:shemminger@xxxxxxxx] Sent: 15 December 2004 00:41 To: Earl,S,Stephen,XJG5A C Cc: netem@xxxxxxxxxxxxxx Subject: Re: Delay distributiocausing segmentation faults Heris how to gea back trace. Ithis example, I replaced 'tc' with a bogus prograthat crashes. Also, tharguments to thgdb 'run' command are the ones passed to thprograbeing debugged. # gdb tc GNU gdb 6.1 Copyrigh2004 FreSoftware Foundation, Inc. GDB is fresoftware, covered by thGNU General Public License, and you are welcomto changit and/or distribute copies of it under certain conditions. Typ"show copying" to sethe conditions. Theris absolutely no warranty for GDB. Typ"show warranty" for details. This GDB was configured as "x86_64-suse-linux"...Using hoslibthread_db library "/lib64/tls/libthread_db.so.1". (gdb) ruadd qdisc dev eth0 roonetem delay 100ms 10ms distribution normal Starting program: /home/shemminger/tc add qdisc dev eth0 roonetedelay 100ms 10ms distribution normal Prograreceived signal SIGSEGV, Segmentation fault. 0x0000000000400502 imain () (gdb) where Also, imighhelp to make sure '-g' is included in the CFLAGS of the build of tc and q_netem.so Thanks. Fron6mod amilewski.org Mon Dec 13 14:20:54 2004 From: n6mod amilewski.org (Aleksandr Milewski) Date: Wed Apr 18 17:37:44 2007 Subject: High-rat(>100Mb/s) emulation? Message-ID: <41BE15C6.8070107@xxxxxxxxxxxx> Is anyonrunning high-speed network emulations with netem? I'trying to simulatlong-fast networks, ideally 622Mb/s (OC-12) or higher with GigabiNICs. I'vjusbuilt a new 2.6.10-rc3 kernel and when I enable netem with evevery small delays, iperf -u can'send any more than 60-70Mb/s. Our owUDP testool shows that I'm getting EAGAIN if I try to exceed that rate. With neteturned off, I can easily exceed 700Mb/s. Machinis a P4 2.8GHz (Prescott) with SK-9E21 NICs. (lates8.12 driver froSyskonnect) Any suggestions? -Zandr Froshemminger aosdl.org Mon Dec 20 09:50:33 2004 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:44 2007 Subject: Re: Neteas a router In-Reply-To: <5CC002233D70444586A1AE6D9984253D010E8FD5@xxxxxxxxxxxxxxxxxxxxxx> References: <5CC002233D70444586A1AE6D9984253D010E8FD5@xxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20041220095033.603e21d4@xxxxxxxxxxxxxxxxx> OMon, 20 Dec 2004 11:03:04 +0530 <nanda.thimmaraju@xxxxxxxxx> wrote: > Hi, > > Warcarrying out some tests for network impairment. I have installed > Fedora Cor3 linux in my system. > > Canetebe used as a router? Is there any document which will help me > configring this as router. > > Regards, > Nanda KishorK.T > Neteoperates jusabove the networking device (Ethernet) by controlling thtransmiqueue. Therefore it can be used in routing and bridging without any changes. I ofterun tests using a machinwith two Ethernets as a bridge to control traffic. Routing works as well. Frodaveat davemloft.net Mon Dec 20 15:43:26 2004 From: daveadavemloft.net (David S. Miller) Date: Wed Apr 18 17:37:44 2007 Subject: Re: [PATCH] netem: restardevicafter inserting packets In-Reply-To: <20041214131113.32d080fb@xxxxxxxxxxxxxxxxx> References: <20041208123103.4cc6b005@xxxxxxxxxxxxxxxxx> <20041208210031.63f0963f.davem@xxxxxxxxxxxxx> <41B91901.3070304@xxxxxxxxx> <20041214113249.0725a655.davem@xxxxxxxxxxxxx> <20041214131113.32d080fb@xxxxxxxxxxxxxxxxx> Message-ID: <20041220154326.4eedb936.davem@xxxxxxxxxxxxx> OTue, 14 Dec 2004 13:11:13 -0800 StepheHemminger <shemminger@xxxxxxxx> wrote: > 2.4 versioof thnetem wakeup patch. Also fixes the qlen > ia couplof places. This makes code basically same as 2.6 > > Signed-off-by: StepheHemminger <shemminger@xxxxxxxx> Applied, thanks a loStephen.
