Hello, having installed FC4 after somnoso good experience with netem on tholder kernels, and deciding thaa clean install is needed, I still gereproduciblSegmentation Faults when change a setting. For a project, I eventually wanto usphpnetemgui for easy configuratio- a wonderful tool, btw. However, for testing purposes, I decided to copy and pastphpnetemgui's settings. Thloss parameter caonly bset in ints, unfortunately, so for now I decided to simply changthaparameter manually until I find and change the right spot ithcode. This is whaI do: This is thoutpufrom the gui when choosing "Display commands only, do noexecute" tc qdisc del dev lo root tc qdisc del dev eth0 root tc qdisc del dev eth1 root tc qdisc del dev eth2 root tc qdisc add dev eth1 roohandl1: htb default 1 tc class add dev eth1 paren1: classid 0:1 htb rat256kbit ceil 256kbit tc qdisc add dev eth2 roohandl1: netem loss 1% tc qdisc add dev eth2 paren1:1 handl10: htb default 1 tc class add dev eth2 paren10: classid 0:1 htb rat2000kbit ceil 2000kbit Theslines arbeing pasted in an ssh. SincI wanonly 0.1% loss, I change the 3rd to last line to the following and executit: tc qdisc changdev eth2 roohandle 1: netem loss 0.1% This reproducibly causes a segmentatiofault. Here's my outpuof unam-a: Linux nete2.6.11-1.1369_FC4 #1 Thu Jun 2 22:55:56 EDT 2005 i686 i686 i386 GNU/Linux And here's tc's "-V"ersion: [root@nete~]# tc -V tc utility, iproute2-ss050314 Any help othis issuwould be greatly appreciated. Thanks iadvance, Uwe Frostephen.earl abt.com Mon Jul 4 03:08:20 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: Segmentatiofaulon tc qdisc change command Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94EF@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I'vhad problems in thpast with using the change command when more than one tc rule is running on an interface. I'm not 100% sure but I think it may be because if you have a qdisc tree already running, you can't change one part of it without disrupting the rest of the tree and causing errors. This could just be because of the way traffic control works or maybe a bug somewhere. If you werto do: tc qdisc add dev eth2 roohandl1: netem loss 1% tc qdisc changdev eth2 roohandle 1: netem loss 0.1% Thawould work ok, buif there are any child qdiscs attached to the root qdisc then it doesn't work. Thonly thing you can do is do a "tc qdisc del dev eth2 root" then reload ththree tc commands for eth2. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of UwE. Bilger Sent: 02 July 2005 19:40 To: netem@xxxxxxxxxxxxxx Subject: Segmentatiofaulon tc qdisc change command Hello, having installed FC4 after somnoso good experience with netem on tholder kernels, and deciding thaa clean install is needed, I still gereproduciblSegmentation Faults when change a setting. For a project, I eventually wanto usphpnetemgui for easy configuratio- a wonderful tool, btw. However, for testing purposes, I decided to copy and pastphpnetemgui's settings. Thloss parameter caonly bset in ints, unfortunately, so for now I decided to simply changthaparameter manually until I find and change the right spot ithcode. This is whaI do: This is thoutpufrom the gui when choosing "Display commands only, do noexecute" tc qdisc del dev lo root tc qdisc del dev eth0 root tc qdisc del dev eth1 root tc qdisc del dev eth2 root tc qdisc add dev eth1 roohandl1: htb default 1 tc class add dev eth1 paren1: classid 0:1 htb rat256kbit ceil 256kbit tc qdisc add dev eth2 roohandl1: netem loss 1% tc qdisc add dev eth2 paren1:1 handl10: htb default 1 tc class add dev eth2 paren10: classid 0:1 htb rat2000kbit ceil 2000kbit Theslines arbeing pasted in an ssh. SincI wanonly 0.1% loss, I change the 3rd to last line to the following and executit: tc qdisc changdev eth2 roohandle 1: netem loss 0.1% This reproducibly causes a segmentatiofault. Here's my outpuof unam-a: Linux nete2.6.11-1.1369_FC4 #1 Thu Jun 2 22:55:56 EDT 2005 i686 i686 i386 GNU/Linux And here's tc's "-V"ersion: [root@nete~]# tc -V tc utility, iproute2-ss050314 Any help othis issuwould be greatly appreciated. Thanks iadvance, Uwe _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Frojuliokriger agmail.com Tue Jul 5 13:18:32 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 12:51:17 2007 Subject: queulength and reordering Message-ID: <682bc30a05070513185d20ceb@xxxxxxxxxxxxxx> Hi, Whilchating with a friend abouthe problems I have trying to replicatan environmenwhere packet reordring, duplication, loss, delay, etc. happens, suddenly comto my mind a strangidea that I would likthayou please comment on. As I havsaid, I havto replicate an environment where packet reordring, duplication, loss, delay, etc. happens. So I usnetem. Good. So I rua server and a clienapp. and they start sending packebetween them. All thpackets are being modified by netem. Good! BuI can'see the reordering, just losses and duplication and delay. Why, why, why I keep asking my self. So this strangidea camup. Why I can't see the reordering? it's becausthskb queue doesn't has enough packets to really do the reordering! So if I add servers and clients, sending packets between theall thtime, the skb queue should have a lot of packets on it and threordering should work. Whado you think abouit? Any suggestion, comment, hint would be welcome. Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Froshemminger aosdl.org Tue Jul 5 14:34:42 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: queulength and reordering In-Reply-To: <682bc30a05070513185d20ceb@xxxxxxxxxxxxxx> References: <682bc30a05070513185d20ceb@xxxxxxxxxxxxxx> Message-ID: <20050705143442.3b0a768f@xxxxxxxxxxxxxxxxx> OTue, 5 Jul 2005 17:18:32 -0300 Julio Kriger <juliokriger@xxxxxxxxx> wrote: > Hi, > Whilchating with a friend abouthe problems I have trying to > replicatan environmenwhere packet reordring, duplication, loss, > delay, etc. happens, suddenly comto my mind a strangidea that I > would likthayou please comment on. > > As I havsaid, I havto replicate an environment where packet > reordring, duplication, loss, delay, etc. happens. So I usnetem. > Good. So I rua server and a clienapp. and they start sending > packebetween them. All thpackets are being modified by netem. > Good! BuI can'see the reordering, just losses and duplication and > delay. Why, why, why I keep asking my self. > So this strangidea camup. Why I can't see the reordering? it's > becausthskb queue doesn't has enough packets to really do the > reordering! So if I add servers and clients, sending packets between > theall thtime, the skb queue should have a lot of packets on it > and threordering should work. > > Whado you think abouit? Any suggestion, comment, hint would be welcome. > Regards, > Julio Yes, you need to havenough delay so reordering can work. Eventually, i plato changnetem to have only it's own private queue and makreordering morintegral. This would mean that for cases wherratcontrol is desired, you would put netem inside TBF rather thaTBF insidnetem. Also, whesetting up a complex config, I likto validate each step. First, teswith delay. Then teswith loss, then test with duplication. Finally, teswith a mix of all of thabove. Iis hard to know whais correct when there are multiple random possibilities happening aonce. Froshemminger aosdl.org Tue Jul 5 14:37:36 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: Re: NetEon thKNOPPIX In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31589@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31589@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050705143736.30f066d1@xxxxxxxxxxxxxxxxx> OTue, 5 Jul 2005 10:27:55 +0100 <stephen.earl@xxxxxx> wrote: > (Netequestion sento the wrong person) > > -----Original Message----- > From: it_21@xxxxxxxxxxxxxx [mailto:it_21@xxxxxxxxxxxxxx] > Sent: 05 July 2005 00:36 > To: Earl,S,Stephen,XDP65A C > Subject: NetEon thKNOPPIX > > > Mr. StepheHemminger > > Hi. > I'it_21. > > I will questiodirectly becausI seem not to have reached correctly my question to the net > > emailing list. > > > I think thaI will usnetem on KNOPPIX 3.8.1. > > I'll download knoppix and look, buisounds like a problem with the psched_clock configuration. Sincthknoppix kernel is a hardcoded config, it may not be fixable. Frojiayingz aengin.umich.edu Thu Jul 7 10:57:24 2005 From: jiayingz aengin.umich.edu (Jiaying Zhang) Date: Wed Apr 18 12:51:17 2007 Subject: questions obandwidth Message-ID: <003301c5831d$579bdd90$9085d38d@IBM4F8C590F1C2> Hi, I ausing neteto simulate network delay. I found when I = increasthdelay time, the bandwidth between two nodes also drops. Is ipossiblto just increase delay, instead of decreasing capacity? I ausing fedora cor3. The linux kernel is 2.6.12. And thbandwidth is measured with iperf. Thanks a lot! Jiaying -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20050707= /c6686344/attachment.htm Froshemminger aosdl.org Thu Jul 7 12:18:52 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: questions obandwidth In-Reply-To: <003301c5831d$579bdd90$9085d38d@IBM4F8C590F1C2> References: <003301c5831d$579bdd90$9085d38d@IBM4F8C590F1C2> Message-ID: <20050707121852.1bef2f0b@xxxxxxxxxxxxxxxxx> OThu, 7 Jul 2005 13:57:24 -0400 "Jiaying Zhang" <jiayingz@xxxxxxxxxxxxxxx> wrote: > Hi, > > I ausing neteto simulate network delay. I found when I > increasthdelay time, the bandwidth between two nodes also > drops. Is ipossiblto just increase delay, instead of decreasing > capacity? I ausing fedora cor3. The linux kernel is 2.6.12. > And thbandwidth is measured with iperf. Thanks a lot! > > Jiaying TCP needs to btuned as thdelay increases. The default values for TCP maximuwindow sizes artoo small for the delay bandwidth product. Go read somsites like: http://www.psc.edu/networking/projects/tcptune/ http://datatag.web.cern.ch/datatag/howto/tcp.html http://developer.osdl.org/shemminger/LWE2005_TCP.pdf for informatioon increasing thavailable memory. Froji.li3 ahp.com Fri Jul 8 09:52:13 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: Help: NetEpackere-order command halts the computer Message-ID: <628BFCE8B64706469FE4D4852CEC953706C29BC1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Hi, Problesummary: NetEpacket re-ordering command halts the computer WheI run thcommand "# tc qdisc add dev eth0 root netem gap 5 delay 10ms", thmachinhalts almost immediately, and I have to reboot it by pressing thpower button. I tried on thremachines with Fedora Core 3 (2.6.9-1.667smp), and all havthis problem. Other netefunctions, such as "tc qdisc add dev eth0 roonetedelay 100ms" or the command for packeloss, work correctly. Thanks, -Ji -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20050708/672a8f75/attachment.htm Froshemminger aosdl.org Fri Jul 8 10:23:34 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: Help: NetEpackere-order command halts the computer In-Reply-To: <628BFCE8B64706469FE4D4852CEC953706C29BC1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <628BFCE8B64706469FE4D4852CEC953706C29BC1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050708102334.6b45dd3d@xxxxxxxxxxxxxxxxx> OFri, 8 Jul 2005 12:52:13 -0400 "Li, Ji" <ji.li3@xxxxxx> wrote: > Hi, > > Problesummary: NetEpacket re-ordering command halts the computer > > WheI run thcommand "# tc qdisc add dev eth0 root netem gap 5 delay > 10ms", thmachinhalts almost immediately, and I have to reboot it by > pressing thpower button. I tried on thremachines with Fedora Core 3 > (2.6.9-1.667smp), and all havthis problem. Other netefunctions, such > as "tc qdisc add dev eth0 roonetedelay 100ms" or the command for > packeloss, work correctly. > > Thanks, > -Ji Does istill happen with 2.6.12? Froji.li3 ahp.com Fri Jul 8 17:54:42 2005 From: ji.li3 ahp.co(Ji Li) Date: Wed Apr 18 12:51:17 2007 Subject: Help: Teql and NetEcan'work together Message-ID: <50478.151.204.83.142.1120870482.squirrel@xxxxxxxxxxxxxxx> Summary: wheI load neteand teql together, teql doesn't work correctly. (If I load teql only, everything is fine) I loaded both neteand teql. Neteis associated with eth0, and teql is associated with both eth0 and eth1. Butraffic only goes ouof eth1. Attached arthcommands that I used to configure teql and netem (on machin1), and commands to configurteql only (on machien 2), their routing tables and "tc qdisc ls" results. What's wrong with my configuration? (I tried loading teql only withounetem, and teql works as expected: ping traffic goes to both interfaces. ) ========= Machin1 with NetE============ Commands to install NetEand TEQL together: # modprobsch_teql # tc qdisc add dev eth0 roohandl1: netem delay 10ms # tc qdisc add dev eth0 paren1:1 handl10: teql0 # tc qdisc add dev eth1 rooteql0 # ip link sedev teql0 up # ip addr add dev eth0 10.0.0.1/30 # ip addr add dev eth1 10.0.0.5/30 # ip addr add dev teql0 10.0.0.9/30 # echo 0 > /proc/sys/net/ipv4/conf/eth0/rp_filter # echo 0 > /proc/sys/net/ipv4/conf/eth1/rp_filter [root@machine1]# tc qdisc ls qdisc nete1: dev eth0 limi1000 delay 100.0ms qdisc teql0 10: dev eth0 paren1:1 qdisc teql0 8001: dev eth1 qdisc pfifo_fas0: dev teql0 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 Destinatio Gateway Genmask Flags Metric Ref Us Iface 10.0.0.8 0.0.0.0 255.255.255.252 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.252 U 0 0 0 eth0 10.0.0.4 0.0.0.0 255.255.255.252 U 0 0 0 eth1 ============== Machin2 withouNetEm ============== Commands to install TEQL: # modprobsch_teql # tc qdisc add dev eth0 rooteql0 # tc qdisc add dev eth1 rooteql0 # ip link sedev teql0 up # ip addr add dev eth0 10.0.0.2/30 # ip addr add dev eth1 10.0.0.6/30 # ip addr add dev teql0 10.0.0.10/30 # echo 0 > /proc/sys/net/ipv4/conf/eth0/rp_filter # echo 0 > /proc/sys/net/ipv4/conf/eth1/rp_filter [root@machine2]# tc qdisc ls qdisc teql0 8001: dev eth0 qdisc teql0 8002: dev eth1 qdisc pfifo_fas0: dev teql0 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 Destinatio Gateway Genmask Flags Metric Ref Us Iface 10.0.0.8 0.0.0.0 255.255.255.252 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.252 U 0 0 0 eth0 10.0.0.4 0.0.0.0 255.255.255.252 U 0 0 0 eth1 Best, -Ji Froavalon adolphin-3.bg Mon Jul 11 02:13:22 2005 From: avaloadolphin-3.bg (Alexander Kostadinov) Date: Wed Apr 18 12:51:17 2007 Subject: website Message-ID: <Pine.LNX.4.58.0507111205010.30099@xxxxxxxxxxxxxxx> thlasexample in the homepage is broken 30:0 and 1:3 Frostephen.earl abt.com Tue Jul 12 02:45:30 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: website Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3158F@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Iworks finfor me. What exact commands did you use and what was the error message? Also, what kernel and iproute2 versions are you using? -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Alexander Kostadinov Sent: 11 July 2005 10:13 To: netem@xxxxxxxxxxxxxx Subject: website thlasexample in the homepage is broken 30:0 and 1:3 Froji.li3 ahp.com Tue Jul 12 06:45:46 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: enabldelay and loss athe same time? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C5D52A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Hi, I wanto usNetEm to simulate delay and packet loss at the same time, buif I simply do thfollowing, the second command will overwrite the firscommand. # tc qdisc add dev eth0 roonetedelay 100ms # tc qdisc changdev eth0 roonetem loss .1% I guess it's mora tc question rather than NetEquestion, but if any onwould tell mthe correct tc commands, your help is highly appreciated! Thanks, -Ji Froji.li3 ahp.com Tue Jul 12 06:50:42 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: A bug ? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C5D52C@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Hi, I ra # tc qdisc changdev eth0 roonetem gap 5 delay 10ms oa linux bridge, and then ping between two machines connected by the bridge. Thresulshould be "the 5th (10th, 15th, ...) packet will get delayed by 10ms and thothers will pass straighout" according to NetEwebsite, buwhat I observed is the reverse: after 4 packets been delayed for 10ms, thnexpacket went straight out. Is it a bug or I did something wrong? Thanks, -Ji Frostephen.earl abt.com Tue Jul 12 07:01:47 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: enabldelay and loss athe same time? Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31590@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Try: tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: netem loss .1% -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:46 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: enabldelay and loss athe same time? Hi, I wanto usNetEm to simulate delay and packet loss at the same time, buif I simply do thfollowing, the second command will overwrite the firscommand. # tc qdisc add dev eth0 roonetedelay 100ms # tc qdisc changdev eth0 roonetem loss .1% I guess it's mora tc question rather than NetEquestion, but if any onwould tell mthe correct tc commands, your help is highly appreciated! Thanks, -Ji Frojuliokriger agmail.com Tue Jul 12 07:04:12 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 12:51:17 2007 Subject: enabldelay and loss athe same time? In-Reply-To: <628BFCE8B64706469FE4D4852CEC953706C5D52A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <628BFCE8B64706469FE4D4852CEC953706C5D52A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <682bc30a05071207043bbcdf@xxxxxxxxxxxxxx> Hi, You should use: tc qdisc changdev eth0 roonetem delay 100ms loss .1% Regards, Julio O7/12/05, Li, Ji <ji.li3@xxxxxx> wrote: > Hi, > > I wanto usNetEm to simulate delay and packet loss at the same time, > buif I simply do thfollowing, the second command will overwrite the > firscommand. > # tc qdisc add dev eth0 roonetedelay 100ms > # tc qdisc changdev eth0 roonetem loss .1% > > I guess it's mora tc question rather than NetEquestion, but if any > onwould tell mthe correct tc commands, your help is highly > appreciated! > > Thanks, > -Ji > > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx > https://lists.osdl.org/mailman/listinfo/netem > > > -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Frostephen.earl abt.com Tue Jul 12 07:08:52 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: A bug ? Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31591@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I asked this questiobefore, and I gothe following reply: StepheHemminger wrote: I anoa WAN expert, but looking at talks and papers I found in googlsearches and by other references; iseems thathe main source of real-world reordering is either redundant links (bonding) or parallel routers (Jupiter) thaallow successivpackets to go through alternatpaths. Thprobability of reordering can be up to 1% on the Internet backbonand almosall the reorders are simple swaps (ie distance 1). Thoriginal reorder cod(by Catalin) had the problem that it was no good for doing "pseudo real world" simulatiobecausyou effectively lost the delay. Iwas only useful for testing if a protocol handled reordering properly. ----------- Iother words, it's intended and noa bug. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:51 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: A bug ? Hi, I ra # tc qdisc changdev eth0 roonetem gap 5 delay 10ms oa linux bridge, and then ping between two machines connected by the bridge. Thresulshould be "the 5th (10th, 15th, ...) packet will get delayed by 10ms and thothers will pass straighout" according to NetEwebsite, buwhat I observed is the reverse: after 4 packets been delayed for 10ms, thnexpacket went straight out. Is it a bug or I did something wrong? Thanks, -Ji Froji.li3 ahp.com Tue Jul 12 08:30:23 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: enabldelay and loss athe same time? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C2A08A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Thank you all a lot! Both methods work! Best, -Ji -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx] OBehalf Of stephen.earl@xxxxxx Sent: Tuesday, July 12, 2005 10:02 AM To: netem@xxxxxxxxxxxxxx Subject: RE: enabldelay and loss athe same time? Try: tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: netem loss .1% -----Original Message----- From: Julio Kriger [mailto:juliokriger@xxxxxxxxx] Sent: Tuesday, July 12, 2005 10:04 AM To: Li, Ji Cc: netem@xxxxxxxxxxxxxx; jli@xxxxxxx Subject: Re: enabldelay and loss athe same time? Hi, You should use: tc qdisc changdev eth0 roonetem delay 100ms loss .1% Regards, Julio -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:46 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: enabldelay and loss athe same time? Hi, I wanto usNetEm to simulate delay and packet loss at the same time, buif I simply do thfollowing, the second command will overwrite the firscommand. # tc qdisc add dev eth0 roonetedelay 100ms # tc qdisc change dev eth0 rooneteloss .1% I guess it's mora tc question rather than NetEquestion, but if any onwould tell mthe correct tc commands, your help is highly appreciated! Thanks, -Ji _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Froji.li3 ahp.com Tue Jul 12 09:20:27 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: A bug ? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C2A0D2@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Thanks for thinfo! -Ji -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx] OBehalf Of stephen.earl@xxxxxx Sent: Tuesday, July 12, 2005 10:09 AM To: netem@xxxxxxxxxxxxxx Subject: RE: A bug ? I asked this questiobefore, and I gothe following reply: StepheHemminger wrote: I anoa WAN expert, but looking at talks and papers I found in google searches and by other references; iseems thathe main source of real-world reordering is either redundanlinks (bonding) or parallel routers (Jupiter) thaallow successivpackets to go through alternatpaths. Thprobability of reordering can be up to 1% othInternet backbone and almost all the reorders are simple swaps (idistanc1). Thoriginal reorder cod(by Catalin) had the problem that it was no good for doing "pseudo real world" simulatiobecausyou effectively losthdelay. Iwas only useful for testing if a protocol handled reordering properly. ----------- Iother words, it's intended and noa bug. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:51 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: A bug ? Hi, I ran # tc qdisc changdev eth0 roonetem gap 5 delay 10ms on a linux bridge, and theping between two machines connected by thbridge. The resulshould b"the 5th (10th, 15th, ...) packet will get delayed by 10ms and thothers will pass straighout" according to NetEm website, buwhaI observed is the reverse: after 4 packets been delayed for 10ms, thnexpacket went straight out. Is it a bug or I did something wrong? Thanks, -Ji _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Frobgolden asonusnet.com Tue Jul 12 11:18:23 2005 From: bgoldeasonusnet.com (Golden, Bukkiah) Date: Wed Apr 18 12:51:17 2007 Subject: RTP packet, no mortraffic Message-ID: <86C4A3413B3A164C9061737B97D6C2E99F0761@xxxxxxxxxxxxxxxxxxxxxxxx> Hello. I havsetup NETEM between to devices thaare sending eachother RTP voice packets. VOICE1_RTP <----> (eth1)NETEM_LINUX_BOX (eth2) <----> VOICE2_RTP I aattempting to introducsome loss of packets via netem. "tc qdisc add dev eth1 root netem loss 10" This appears to work for ICMP (ping) traffic, I losten percen(or approximately that) of the responses. However, as sooas a run an RTP streaover this interface, no more traffic passes, even the ICMP. "tc qdisc del dev eth1 root" will clear thcondition, and a readdition "tc qdisc add dev eth1 roonetem loss 10" will cause the pings to be again lost according to my expectations. However RTP is whaI really wanto lose. Am I missing something? Do I need to seup a specific rulfor RTP? Thank you, Bukkiah Golden. Frojuliokriger agmail.com Wed Jul 13 08:43:13 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 12:51:17 2007 Subject: reordeing and hungup Message-ID: <682bc30a05071308436af9c19@xxxxxxxxxxxxxx> Hi, I would to sharwith you all something thais happening to me, and any comments arwelcome. I seneteto work on "lo" with a delay of 300ms, a jitter of 100ms, loss of 1%, reordering of 50%. TheI run 3 or mor"ping -f localhost" separately. After all of this, I starworking on a user spacSCTP library. All is doing well until, aa random time (I mean a few minutes later or half ahour or an hour or morlater), the machinhungup, idoesn't give a kernel panic, it just hungup. This is noan error caused by thSCTP library, because it happened sometimes beforI starworking on SCTP. I don't know why this happens, ioccurs 2 o 3 times when I work for 2 or 3 hours. Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Froshemminger aosdl.org Wed Jul 13 18:50:59 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: reordeing and hungup In-Reply-To: <682bc30a05071308436af9c19@xxxxxxxxxxxxxx> References: <682bc30a05071308436af9c19@xxxxxxxxxxxxxx> Message-ID: <42D5C503.1010805@xxxxxxxx> Julio Kriger wrote: >Hi, > I would to sharwith you all something thais happening to me, and >any comments arwelcome. > > I seneteto work on "lo" with a delay of 300ms, a jitter of >100ms, loss of 1%, reordering of 50%. TheI run 3 or mor"ping -f >localhost" separately. After all of this, I starworking on a user >spacSCTP library. All is doing well until, aa random time (I mean >a few minutes later or half ahour or an hour or morlater), the >machinhungup, idoesn't give a kernel panic, it just hungup. This >is noan error caused by thSCTP library, because it happened >sometimes beforI starworking on SCTP. I don't know why this >happens, ioccurs 2 o 3 times when I work for 2 or 3 hours. > > > Regards, >Julio > > Whakernel version. Therwere some bugs with the accounting of the queulength thawould make the qdisc scheduler spin for ever. Thaccounting of queulength would get off in case of loss or duplicatioand nested queudisciplines. Thbugs werfixed in 2.6.12, and 2.6.12 has a check in the qdisc scheduler to avoid sigwrap. Frojuliokriger agmail.com Thu Jul 14 16:17:57 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 12:51:17 2007 Subject: tc problem Message-ID: <682bc30a050714161710c63d52@xxxxxxxxxxxxxx> Hi! I know this is nothright place, but I try the following: tc qdisc add dev lo roohandl1:0 netem delay 300ms 100ms reorder 50% drop 0.1% tc qdisc add dev lo paren1:1 handl10: tbf burst 20480 limit 300 mtu 1514 rat32000bps and get: RTNETLINK answers: Invalid argument Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Frostephen.earl abt.com Fri Jul 15 02:22:46 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: tc problem Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31592@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Whakernel version aryou currently using? For the reorder command to work for the moment you need to have a patched version of iproute2. The patches were posted earlier in the mailing list. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Julio Kriger Sent: 15 July 2005 00:18 To: netem@xxxxxxxx Subject: tc problem Hi! I know this is nothright place, but I try the following: tc qdisc add dev lo roohandl1:0 netem delay 300ms 100ms reorder 50% drop 0.1% tc qdisc add dev lo paren1:1 handl10: tbf burst 20480 limit 300 mtu 1514 rat32000bps and get: RTNETLINK answers: Invalid argument Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Frouwelists au-b.de Sat Jul 2 11:40:21 2005 From: uwelists au-b.d(Uwe E. Bilger) Date: Wed Apr 18 17:37:47 2007 Subject: Segmentatiofaulon tc qdisc change command Message-ID: <EAEJJNOGOBFKMFJLNAOAIEEIFDAA.uwelists@xxxxxx> Hello, having installed FC4 after somnoso good experience with netem on tholder kernels, and deciding thaa clean install is needed, I still gereproduciblSegmentation Faults when change a setting. For a project, I eventually wanto usphpnetemgui for easy configuratio- a wonderful tool, btw. However, for testing purposes, I decided to copy and pastphpnetemgui's settings. Thloss parameter caonly bset in ints, unfortunately, so for now I decided to simply changthaparameter manually until I find and change the right spot ithcode. This is whaI do: This is thoutpufrom the gui when choosing "Display commands only, do noexecute" tc qdisc del dev lo root tc qdisc del dev eth0 root tc qdisc del dev eth1 root tc qdisc del dev eth2 root tc qdisc add dev eth1 roohandl1: htb default 1 tc class add dev eth1 paren1: classid 0:1 htb rat256kbit ceil 256kbit tc qdisc add dev eth2 roohandl1: netem loss 1% tc qdisc add dev eth2 paren1:1 handl10: htb default 1 tc class add dev eth2 paren10: classid 0:1 htb rat2000kbit ceil 2000kbit Theslines arbeing pasted in an ssh. SincI wanonly 0.1% loss, I change the 3rd to last line to the following and executit: tc qdisc changdev eth2 roohandle 1: netem loss 0.1% This reproducibly causes a segmentatiofault. Here's my outpuof unam-a: Linux nete2.6.11-1.1369_FC4 #1 Thu Jun 2 22:55:56 EDT 2005 i686 i686 i386 GNU/Linux And here's tc's "-V"ersion: [root@nete~]# tc -V tc utility, iproute2-ss050314 Any help othis issuwould be greatly appreciated. Thanks iadvance, Uwe Frostephen.earl abt.com Mon Jul 4 03:08:20 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: Segmentatiofaulon tc qdisc change command Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94EF@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I'vhad problems in thpast with using the change command when more than one tc rule is running on an interface. I'm not 100% sure but I think it may be because if you have a qdisc tree already running, you can't change one part of it without disrupting the rest of the tree and causing errors. This could just be because of the way traffic control works or maybe a bug somewhere. If you werto do: tc qdisc add dev eth2 roohandl1: netem loss 1% tc qdisc changdev eth2 roohandle 1: netem loss 0.1% Thawould work ok, buif there are any child qdiscs attached to the root qdisc then it doesn't work. Thonly thing you can do is do a "tc qdisc del dev eth2 root" then reload ththree tc commands for eth2. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of UwE. Bilger Sent: 02 July 2005 19:40 To: netem@xxxxxxxxxxxxxx Subject: Segmentatiofaulon tc qdisc change command Hello, having installed FC4 after somnoso good experience with netem on tholder kernels, and deciding thaa clean install is needed, I still gereproduciblSegmentation Faults when change a setting. For a project, I eventually wanto usphpnetemgui for easy configuratio- a wonderful tool, btw. However, for testing purposes, I decided to copy and pastphpnetemgui's settings. Thloss parameter caonly bset in ints, unfortunately, so for now I decided to simply changthaparameter manually until I find and change the right spot ithcode. This is whaI do: This is thoutpufrom the gui when choosing "Display commands only, do noexecute" tc qdisc del dev lo root tc qdisc del dev eth0 root tc qdisc del dev eth1 root tc qdisc del dev eth2 root tc qdisc add dev eth1 roohandl1: htb default 1 tc class add dev eth1 paren1: classid 0:1 htb rat256kbit ceil 256kbit tc qdisc add dev eth2 roohandl1: netem loss 1% tc qdisc add dev eth2 paren1:1 handl10: htb default 1 tc class add dev eth2 paren10: classid 0:1 htb rat2000kbit ceil 2000kbit Theslines arbeing pasted in an ssh. SincI wanonly 0.1% loss, I change the 3rd to last line to the following and executit: tc qdisc changdev eth2 roohandle 1: netem loss 0.1% This reproducibly causes a segmentatiofault. Here's my outpuof unam-a: Linux nete2.6.11-1.1369_FC4 #1 Thu Jun 2 22:55:56 EDT 2005 i686 i686 i386 GNU/Linux And here's tc's "-V"ersion: [root@nete~]# tc -V tc utility, iproute2-ss050314 Any help othis issuwould be greatly appreciated. Thanks iadvance, Uwe _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Frojuliokriger agmail.com Tue Jul 5 13:18:32 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 17:37:47 2007 Subject: queulength and reordering Message-ID: <682bc30a05070513185d20ceb@xxxxxxxxxxxxxx> Hi, Whilchating with a friend abouthe problems I have trying to replicatan environmenwhere packet reordring, duplication, loss, delay, etc. happens, suddenly comto my mind a strangidea that I would likthayou please comment on. As I havsaid, I havto replicate an environment where packet reordring, duplication, loss, delay, etc. happens. So I usnetem. Good. So I rua server and a clienapp. and they start sending packebetween them. All thpackets are being modified by netem. Good! BuI can'see the reordering, just losses and duplication and delay. Why, why, why I keep asking my self. So this strangidea camup. Why I can't see the reordering? it's becausthskb queue doesn't has enough packets to really do the reordering! So if I add servers and clients, sending packets between theall thtime, the skb queue should have a lot of packets on it and threordering should work. Whado you think abouit? Any suggestion, comment, hint would be welcome. Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Froshemminger aosdl.org Tue Jul 5 14:34:42 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: queulength and reordering In-Reply-To: <682bc30a05070513185d20ceb@xxxxxxxxxxxxxx> References: <682bc30a05070513185d20ceb@xxxxxxxxxxxxxx> Message-ID: <20050705143442.3b0a768f@xxxxxxxxxxxxxxxxx> OTue, 5 Jul 2005 17:18:32 -0300 Julio Kriger <juliokriger@xxxxxxxxx> wrote: > Hi, > Whilchating with a friend abouthe problems I have trying to > replicatan environmenwhere packet reordring, duplication, loss, > delay, etc. happens, suddenly comto my mind a strangidea that I > would likthayou please comment on. > > As I havsaid, I havto replicate an environment where packet > reordring, duplication, loss, delay, etc. happens. So I usnetem. > Good. So I rua server and a clienapp. and they start sending > packebetween them. All thpackets are being modified by netem. > Good! BuI can'see the reordering, just losses and duplication and > delay. Why, why, why I keep asking my self. > So this strangidea camup. Why I can't see the reordering? it's > becausthskb queue doesn't has enough packets to really do the > reordering! So if I add servers and clients, sending packets between > theall thtime, the skb queue should have a lot of packets on it > and threordering should work. > > Whado you think abouit? Any suggestion, comment, hint would be welcome. > Regards, > Julio Yes, you need to havenough delay so reordering can work. Eventually, i plato changnetem to have only it's own private queue and makreordering morintegral. This would mean that for cases wherratcontrol is desired, you would put netem inside TBF rather thaTBF insidnetem. Also, whesetting up a complex config, I likto validate each step. First, teswith delay. Then teswith loss, then test with duplication. Finally, teswith a mix of all of thabove. Iis hard to know whais correct when there are multiple random possibilities happening aonce. Froshemminger aosdl.org Tue Jul 5 14:37:36 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: Re: NetEon thKNOPPIX In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31589@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31589@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050705143736.30f066d1@xxxxxxxxxxxxxxxxx> OTue, 5 Jul 2005 10:27:55 +0100 <stephen.earl@xxxxxx> wrote: > (Netequestion sento the wrong person) > > -----Original Message----- > From: it_21@xxxxxxxxxxxxxx [mailto:it_21@xxxxxxxxxxxxxx] > Sent: 05 July 2005 00:36 > To: Earl,S,Stephen,XDP65A C > Subject: NetEon thKNOPPIX > > > Mr. StepheHemminger > > Hi. > I'it_21. > > I will questiodirectly becausI seem not to have reached correctly my question to the net > > emailing list. > > > I think thaI will usnetem on KNOPPIX 3.8.1. > > I'll download knoppix and look, buisounds like a problem with the psched_clock configuration. Sincthknoppix kernel is a hardcoded config, it may not be fixable. Frojiayingz aengin.umich.edu Thu Jul 7 10:57:24 2005 From: jiayingz aengin.umich.edu (Jiaying Zhang) Date: Wed Apr 18 17:37:47 2007 Subject: questions obandwidth Message-ID: <003301c5831d$579bdd90$9085d38d@IBM4F8C590F1C2> Hi, I ausing neteto simulate network delay. I found when I = increasthdelay time, the bandwidth between two nodes also drops. Is ipossiblto just increase delay, instead of decreasing capacity? I ausing fedora cor3. The linux kernel is 2.6.12. And thbandwidth is measured with iperf. Thanks a lot! Jiaying -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20050707= /c6686344/attachment-0001.htm Froshemminger aosdl.org Thu Jul 7 12:18:52 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: questions obandwidth In-Reply-To: <003301c5831d$579bdd90$9085d38d@IBM4F8C590F1C2> References: <003301c5831d$579bdd90$9085d38d@IBM4F8C590F1C2> Message-ID: <20050707121852.1bef2f0b@xxxxxxxxxxxxxxxxx> OThu, 7 Jul 2005 13:57:24 -0400 "Jiaying Zhang" <jiayingz@xxxxxxxxxxxxxxx> wrote: > Hi, > > I ausing neteto simulate network delay. I found when I > increasthdelay time, the bandwidth between two nodes also > drops. Is ipossiblto just increase delay, instead of decreasing > capacity? I ausing fedora cor3. The linux kernel is 2.6.12. > And thbandwidth is measured with iperf. Thanks a lot! > > Jiaying TCP needs to btuned as thdelay increases. The default values for TCP maximuwindow sizes artoo small for the delay bandwidth product. Go read somsites like: http://www.psc.edu/networking/projects/tcptune/ http://datatag.web.cern.ch/datatag/howto/tcp.html http://developer.osdl.org/shemminger/LWE2005_TCP.pdf for informatioon increasing thavailable memory. Froji.li3 ahp.com Fri Jul 8 09:52:13 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: Help: NetEpackere-order command halts the computer Message-ID: <628BFCE8B64706469FE4D4852CEC953706C29BC1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Hi, Problesummary: NetEpacket re-ordering command halts the computer WheI run thcommand "# tc qdisc add dev eth0 root netem gap 5 delay 10ms", thmachinhalts almost immediately, and I have to reboot it by pressing thpower button. I tried on thremachines with Fedora Core 3 (2.6.9-1.667smp), and all havthis problem. Other netefunctions, such as "tc qdisc add dev eth0 roonetedelay 100ms" or the command for packeloss, work correctly. Thanks, -Ji -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20050708/672a8f75/attachment-0001.htm Froshemminger aosdl.org Fri Jul 8 10:23:34 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: Help: NetEpackere-order command halts the computer In-Reply-To: <628BFCE8B64706469FE4D4852CEC953706C29BC1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <628BFCE8B64706469FE4D4852CEC953706C29BC1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050708102334.6b45dd3d@xxxxxxxxxxxxxxxxx> OFri, 8 Jul 2005 12:52:13 -0400 "Li, Ji" <ji.li3@xxxxxx> wrote: > Hi, > > Problesummary: NetEpacket re-ordering command halts the computer > > WheI run thcommand "# tc qdisc add dev eth0 root netem gap 5 delay > 10ms", thmachinhalts almost immediately, and I have to reboot it by > pressing thpower button. I tried on thremachines with Fedora Core 3 > (2.6.9-1.667smp), and all havthis problem. Other netefunctions, such > as "tc qdisc add dev eth0 roonetedelay 100ms" or the command for > packeloss, work correctly. > > Thanks, > -Ji Does istill happen with 2.6.12? Froji.li3 ahp.com Fri Jul 8 17:54:42 2005 From: ji.li3 ahp.co(Ji Li) Date: Wed Apr 18 17:37:47 2007 Subject: Help: Teql and NetEcan'work together Message-ID: <50478.151.204.83.142.1120870482.squirrel@xxxxxxxxxxxxxxx> Summary: wheI load neteand teql together, teql doesn't work correctly. (If I load teql only, everything is fine) I loaded both neteand teql. Neteis associated with eth0, and teql is associated with both eth0 and eth1. Butraffic only goes ouof eth1. Attached arthcommands that I used to configure teql and netem (on machin1), and commands to configurteql only (on machien 2), their routing tables and "tc qdisc ls" results. What's wrong with my configuration? (I tried loading teql only withounetem, and teql works as expected: ping traffic goes to both interfaces. ) ========= Machin1 with NetE============ Commands to install NetEand TEQL together: # modprobsch_teql # tc qdisc add dev eth0 roohandl1: netem delay 10ms # tc qdisc add dev eth0 paren1:1 handl10: teql0 # tc qdisc add dev eth1 rooteql0 # ip link sedev teql0 up # ip addr add dev eth0 10.0.0.1/30 # ip addr add dev eth1 10.0.0.5/30 # ip addr add dev teql0 10.0.0.9/30 # echo 0 > /proc/sys/net/ipv4/conf/eth0/rp_filter # echo 0 > /proc/sys/net/ipv4/conf/eth1/rp_filter [root@machine1]# tc qdisc ls qdisc nete1: dev eth0 limi1000 delay 100.0ms qdisc teql0 10: dev eth0 paren1:1 qdisc teql0 8001: dev eth1 qdisc pfifo_fas0: dev teql0 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 Destinatio Gateway Genmask Flags Metric Ref Us Iface 10.0.0.8 0.0.0.0 255.255.255.252 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.252 U 0 0 0 eth0 10.0.0.4 0.0.0.0 255.255.255.252 U 0 0 0 eth1 ============== Machin2 withouNetEm ============== Commands to install TEQL: # modprobsch_teql # tc qdisc add dev eth0 rooteql0 # tc qdisc add dev eth1 rooteql0 # ip link sedev teql0 up # ip addr add dev eth0 10.0.0.2/30 # ip addr add dev eth1 10.0.0.6/30 # ip addr add dev teql0 10.0.0.10/30 # echo 0 > /proc/sys/net/ipv4/conf/eth0/rp_filter # echo 0 > /proc/sys/net/ipv4/conf/eth1/rp_filter [root@machine2]# tc qdisc ls qdisc teql0 8001: dev eth0 qdisc teql0 8002: dev eth1 qdisc pfifo_fas0: dev teql0 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 Destinatio Gateway Genmask Flags Metric Ref Us Iface 10.0.0.8 0.0.0.0 255.255.255.252 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.252 U 0 0 0 eth0 10.0.0.4 0.0.0.0 255.255.255.252 U 0 0 0 eth1 Best, -Ji Froavalon adolphin-3.bg Mon Jul 11 02:13:22 2005 From: avaloadolphin-3.bg (Alexander Kostadinov) Date: Wed Apr 18 17:37:47 2007 Subject: website Message-ID: <Pine.LNX.4.58.0507111205010.30099@xxxxxxxxxxxxxxx> thlasexample in the homepage is broken 30:0 and 1:3 Frostephen.earl abt.com Tue Jul 12 02:45:30 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: website Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3158F@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Iworks finfor me. What exact commands did you use and what was the error message? Also, what kernel and iproute2 versions are you using? -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Alexander Kostadinov Sent: 11 July 2005 10:13 To: netem@xxxxxxxxxxxxxx Subject: website thlasexample in the homepage is broken 30:0 and 1:3 Froji.li3 ahp.com Tue Jul 12 06:45:46 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: enabldelay and loss athe same time? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C5D52A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Hi, I wanto usNetEm to simulate delay and packet loss at the same time, buif I simply do thfollowing, the second command will overwrite the firscommand. # tc qdisc add dev eth0 roonetedelay 100ms # tc qdisc changdev eth0 roonetem loss .1% I guess it's mora tc question rather than NetEquestion, but if any onwould tell mthe correct tc commands, your help is highly appreciated! Thanks, -Ji Froji.li3 ahp.com Tue Jul 12 06:50:42 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: A bug ? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C5D52C@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Hi, I ra # tc qdisc changdev eth0 roonetem gap 5 delay 10ms oa linux bridge, and then ping between two machines connected by the bridge. Thresulshould be "the 5th (10th, 15th, ...) packet will get delayed by 10ms and thothers will pass straighout" according to NetEwebsite, buwhat I observed is the reverse: after 4 packets been delayed for 10ms, thnexpacket went straight out. Is it a bug or I did something wrong? Thanks, -Ji Frostephen.earl abt.com Tue Jul 12 07:01:47 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: enabldelay and loss athe same time? Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31590@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Try: tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: netem loss .1% -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:46 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: enabldelay and loss athe same time? Hi, I wanto usNetEm to simulate delay and packet loss at the same time, buif I simply do thfollowing, the second command will overwrite the firscommand. # tc qdisc add dev eth0 roonetedelay 100ms # tc qdisc changdev eth0 roonetem loss .1% I guess it's mora tc question rather than NetEquestion, but if any onwould tell mthe correct tc commands, your help is highly appreciated! Thanks, -Ji Frojuliokriger agmail.com Tue Jul 12 07:04:12 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 17:37:47 2007 Subject: enabldelay and loss athe same time? In-Reply-To: <628BFCE8B64706469FE4D4852CEC953706C5D52A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <628BFCE8B64706469FE4D4852CEC953706C5D52A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <682bc30a05071207043bbcdf@xxxxxxxxxxxxxx> Hi, You should use: tc qdisc changdev eth0 roonetem delay 100ms loss .1% Regards, Julio O7/12/05, Li, Ji <ji.li3@xxxxxx> wrote: > Hi, > > I wanto usNetEm to simulate delay and packet loss at the same time, > buif I simply do thfollowing, the second command will overwrite the > firscommand. > # tc qdisc add dev eth0 roonetedelay 100ms > # tc qdisc changdev eth0 roonetem loss .1% > > I guess it's mora tc question rather than NetEquestion, but if any > onwould tell mthe correct tc commands, your help is highly > appreciated! > > Thanks, > -Ji > > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx > https://lists.osdl.org/mailman/listinfo/netem > > > -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Frostephen.earl abt.com Tue Jul 12 07:08:52 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: A bug ? Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31591@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I asked this questiobefore, and I gothe following reply: StepheHemminger wrote: I anoa WAN expert, but looking at talks and papers I found in googlsearches and by other references; iseems thathe main source of real-world reordering is either redundant links (bonding) or parallel routers (Jupiter) thaallow successivpackets to go through alternatpaths. Thprobability of reordering can be up to 1% on the Internet backbonand almosall the reorders are simple swaps (ie distance 1). Thoriginal reorder cod(by Catalin) had the problem that it was no good for doing "pseudo real world" simulatiobecausyou effectively lost the delay. Iwas only useful for testing if a protocol handled reordering properly. ----------- Iother words, it's intended and noa bug. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:51 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: A bug ? Hi, I ra # tc qdisc changdev eth0 roonetem gap 5 delay 10ms oa linux bridge, and then ping between two machines connected by the bridge. Thresulshould be "the 5th (10th, 15th, ...) packet will get delayed by 10ms and thothers will pass straighout" according to NetEwebsite, buwhat I observed is the reverse: after 4 packets been delayed for 10ms, thnexpacket went straight out. Is it a bug or I did something wrong? Thanks, -Ji Froji.li3 ahp.com Tue Jul 12 08:30:23 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: enabldelay and loss athe same time? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C2A08A@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Thank you all a lot! Both methods work! Best, -Ji -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx] OBehalf Of stephen.earl@xxxxxx Sent: Tuesday, July 12, 2005 10:02 AM To: netem@xxxxxxxxxxxxxx Subject: RE: enabldelay and loss athe same time? Try: tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: netem loss .1% -----Original Message----- From: Julio Kriger [mailto:juliokriger@xxxxxxxxx] Sent: Tuesday, July 12, 2005 10:04 AM To: Li, Ji Cc: netem@xxxxxxxxxxxxxx; jli@xxxxxxx Subject: Re: enabldelay and loss athe same time? Hi, You should use: tc qdisc changdev eth0 roonetem delay 100ms loss .1% Regards, Julio -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:46 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: enabldelay and loss athe same time? Hi, I wanto usNetEm to simulate delay and packet loss at the same time, buif I simply do thfollowing, the second command will overwrite the firscommand. # tc qdisc add dev eth0 roonetedelay 100ms # tc qdisc change dev eth0 rooneteloss .1% I guess it's mora tc question rather than NetEquestion, but if any onwould tell mthe correct tc commands, your help is highly appreciated! Thanks, -Ji _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Froji.li3 ahp.com Tue Jul 12 09:20:27 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: A bug ? Message-ID: <628BFCE8B64706469FE4D4852CEC953706C2A0D2@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Thanks for thinfo! -Ji -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx] OBehalf Of stephen.earl@xxxxxx Sent: Tuesday, July 12, 2005 10:09 AM To: netem@xxxxxxxxxxxxxx Subject: RE: A bug ? I asked this questiobefore, and I gothe following reply: StepheHemminger wrote: I anoa WAN expert, but looking at talks and papers I found in google searches and by other references; iseems thathe main source of real-world reordering is either redundanlinks (bonding) or parallel routers (Jupiter) thaallow successivpackets to go through alternatpaths. Thprobability of reordering can be up to 1% othInternet backbone and almost all the reorders are simple swaps (idistanc1). Thoriginal reorder cod(by Catalin) had the problem that it was no good for doing "pseudo real world" simulatiobecausyou effectively losthdelay. Iwas only useful for testing if a protocol handled reordering properly. ----------- Iother words, it's intended and noa bug. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 12 July 2005 14:51 To: netem@xxxxxxxxxxxxxx Cc: jli@xxxxxxx Subject: A bug ? Hi, I ran # tc qdisc changdev eth0 roonetem gap 5 delay 10ms on a linux bridge, and theping between two machines connected by thbridge. The resulshould b"the 5th (10th, 15th, ...) packet will get delayed by 10ms and thothers will pass straighout" according to NetEm website, buwhaI observed is the reverse: after 4 packets been delayed for 10ms, thnexpacket went straight out. Is it a bug or I did something wrong? Thanks, -Ji _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem Frobgolden asonusnet.com Tue Jul 12 11:18:23 2005 From: bgoldeasonusnet.com (Golden, Bukkiah) Date: Wed Apr 18 17:37:47 2007 Subject: RTP packet, no mortraffic Message-ID: <86C4A3413B3A164C9061737B97D6C2E99F0761@xxxxxxxxxxxxxxxxxxxxxxxx> Hello. I havsetup NETEM between to devices thaare sending eachother RTP voice packets. VOICE1_RTP <----> (eth1)NETEM_LINUX_BOX (eth2) <----> VOICE2_RTP I aattempting to introducsome loss of packets via netem. "tc qdisc add dev eth1 root netem loss 10" This appears to work for ICMP (ping) traffic, I losten percen(or approximately that) of the responses. However, as sooas a run an RTP streaover this interface, no more traffic passes, even the ICMP. "tc qdisc del dev eth1 root" will clear thcondition, and a readdition "tc qdisc add dev eth1 roonetem loss 10" will cause the pings to be again lost according to my expectations. However RTP is whaI really wanto lose. Am I missing something? Do I need to seup a specific rulfor RTP? Thank you, Bukkiah Golden. Frojuliokriger agmail.com Wed Jul 13 08:43:13 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 17:37:47 2007 Subject: reordeing and hungup Message-ID: <682bc30a05071308436af9c19@xxxxxxxxxxxxxx> Hi, I would to sharwith you all something thais happening to me, and any comments arwelcome. I seneteto work on "lo" with a delay of 300ms, a jitter of 100ms, loss of 1%, reordering of 50%. TheI run 3 or mor"ping -f localhost" separately. After all of this, I starworking on a user spacSCTP library. All is doing well until, aa random time (I mean a few minutes later or half ahour or an hour or morlater), the machinhungup, idoesn't give a kernel panic, it just hungup. This is noan error caused by thSCTP library, because it happened sometimes beforI starworking on SCTP. I don't know why this happens, ioccurs 2 o 3 times when I work for 2 or 3 hours. Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Froshemminger aosdl.org Wed Jul 13 18:50:59 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: reordeing and hungup In-Reply-To: <682bc30a05071308436af9c19@xxxxxxxxxxxxxx> References: <682bc30a05071308436af9c19@xxxxxxxxxxxxxx> Message-ID: <42D5C503.1010805@xxxxxxxx> Julio Kriger wrote: >Hi, > I would to sharwith you all something thais happening to me, and >any comments arwelcome. > > I seneteto work on "lo" with a delay of 300ms, a jitter of >100ms, loss of 1%, reordering of 50%. TheI run 3 or mor"ping -f >localhost" separately. After all of this, I starworking on a user >spacSCTP library. All is doing well until, aa random time (I mean >a few minutes later or half ahour or an hour or morlater), the >machinhungup, idoesn't give a kernel panic, it just hungup. This >is noan error caused by thSCTP library, because it happened >sometimes beforI starworking on SCTP. I don't know why this >happens, ioccurs 2 o 3 times when I work for 2 or 3 hours. > > > Regards, >Julio > > Whakernel version. Therwere some bugs with the accounting of the queulength thawould make the qdisc scheduler spin for ever. Thaccounting of queulength would get off in case of loss or duplicatioand nested queudisciplines. Thbugs werfixed in 2.6.12, and 2.6.12 has a check in the qdisc scheduler to avoid sigwrap. Frojuliokriger agmail.com Thu Jul 14 16:17:57 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 17:37:47 2007 Subject: tc problem Message-ID: <682bc30a050714161710c63d52@xxxxxxxxxxxxxx> Hi! I know this is nothright place, but I try the following: tc qdisc add dev lo roohandl1:0 netem delay 300ms 100ms reorder 50% drop 0.1% tc qdisc add dev lo paren1:1 handl10: tbf burst 20480 limit 300 mtu 1514 rat32000bps and get: RTNETLINK answers: Invalid argument Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Frostephen.earl abt.com Fri Jul 15 02:22:46 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: tc problem Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A31592@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Whakernel version aryou currently using? For the reorder command to work for the moment you need to have a patched version of iproute2. The patches were posted earlier in the mailing list. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Julio Kriger Sent: 15 July 2005 00:18 To: netem@xxxxxxxx Subject: tc problem Hi! I know this is nothright place, but I try the following: tc qdisc add dev lo roohandl1:0 netem delay 300ms 100ms reorder 50% drop 0.1% tc qdisc add dev lo paren1:1 handl10: tbf burst 20480 limit 300 mtu 1514 rat32000bps and get: RTNETLINK answers: Invalid argument Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx _______________________________________________ Netemailing list Netem@xxxxxxxxxxxxxx https://lists.osdl.org/mailman/listinfo/netem
