Iappears thaNetem drops packets when the magnitude of jitter goes beyond thinter-packedelay. For example.... Assumthaour jitter distribution would have a series of packets delayed by (milliseconds) 5, 10, 20, 5, 10, 20 and thapackets A-F arriva10 millisecond intervals... This should resulin thfollowing timings... Pack Arrive Depart A 0 5 B 10 20 C 20 40 D 30 35 E 40 50 F 50 70 So, thpackets should go ouin the order A, B, D, C, E, F However, iseems likthis causes packet C to be dropped. This looks pretty serious, sincthawould mean there is no jitter distribution thacould bused under live conditions without inducing packet loss. Is this really thcasor have I misunderstood something? -Joel Peshkin Froking_turlutor ahotmail.com Thu Jun 2 07:32:33 2005 From: king_turlutor ahotmail.co(Tutu Lord) Date: Wed Apr 18 12:51:17 2007 Subject: beginner with netem Message-ID: <BAY104-F4D075E450C38C5569B68098060@xxxxxxx> hello, i havLinux Mandrak10.1, and i am interested by Netem. I havgo: - nistnet.2.0.12b.tar.gz - iproute2-2.6-ss040701.tar.gz Bui havfew problems: - i don'know compil and install .tar.gz ( i havread the README) - linux don'wansee the two network interfaces (either one or other) this projecis importanfor me. thank luc _________________________________________________________________ MSHotmail : antivirus eantispam gratuits http://www.imagine-msn.com/hotmail/default.aspx?locale=fr-FR Froshemminger aosdl.org Thu Jun 2 11:26:38 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: Jitter problems with Netem?? In-Reply-To: <429E71DA.30001@xxxxxxxxxxx> References: <429E71DA.30001@xxxxxxxxxxx> Message-ID: <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> OWed, 01 Jun 2005 19:41:30 -0700 Joel Peshki<bugreport@xxxxxxxxxxx> wrote: > > Iappears thaNetem drops packets when the magnitude of jitter goes > beyond thinter-packedelay. For example.... > > Assumthaour jitter distribution would have a series of packets > delayed by (milliseconds) 5, 10, 20, 5, 10, 20 and thapackets A-F > arriva10 millisecond intervals... > > This should resulin thfollowing timings... > Pack Arrive Depart > A 0 5 > B 10 20 > C 20 40 > D 30 35 > E 40 50 > F 50 70 > So, thpackets should go ouin the order A, B, D, C, E, F > > However, iseems likthis causes packet C to be dropped. This looks > pretty serious, sincthawould mean there is no jitter distribution > thacould bused under live conditions without inducing packet loss. > > Is this really thcasor have I misunderstood something? > > -Joel Peshkin > First, which kernel version. Thermay bbugs that already have been fixed. Thexpected behaviour would is to preservorder always, but Pack Arrive Delay Depart A 0 5 5 B 10 10 20 C 20 20 40 D 30 5 40 E 40 10 50 F 50 20 70 Whenetegets to looking at packet C, it's time has already expired so ican go righaway. I did have a patch that caused C to go out ahead of B buuser's complained thathey did not expect time based reordering and thabehaviour was onof the things they hated about Nist net. Froshemminger aosdl.org Thu Jun 2 11:35:16 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: beginner with netem In-Reply-To: <BAY104-F4D075E450C38C5569B68098060@xxxxxxx> References: <BAY104-F4D075E450C38C5569B68098060@xxxxxxx> Message-ID: <20050602113516.113df76c@xxxxxxxxxxxxxxxxx> OThu, 02 Jun 2005 16:32:33 +0200 "Tutu Lord" <king_turlutor@xxxxxxxxxxx> wrote: > hello, > > i havLinux Mandrak10.1, and i am interested by Netem. > I havgo: > - nistnet.2.0.12b.tar.gz don'need thafor netem > - iproute2-2.6-ss040701.tar.gz that's probably older thaMandrake. > > > Bui havfew problems: > > - i don'know compil and install .tar.gz ( i havread the README) > - linux don'wansee the two network interfaces (either one or other) > > this projecis importanfor me. > > thank > > luc Firsoff, try iwithout compiling anything. Look at the version of iproute2 installed with Mandrake. tc -V will show version. Also, if /usr/lib/tc exists and contains q_netem.so and a bunch of .disfiles, you almoscertainly have a good enough version. To giva simpltest, just do (as root): tc qdisc add dev eth0 roonetem If isucceeds then you havjust put netem (with no options) ias thqdisc for the normal default ether device. Netem with no options is jusa simplFIFO so it is mostly harmless (1ms delay oearlier kernels pr2.6.12). Building iproute2 is relatively painless. Download latesfrom http://developer.osdl.org/dev/iproute2/download theunpack with tar xzf iproute2-2.6.11-050330.targ.gz to build cd iproute2-2.6.11-050330 make to install (as root) makinstall You do need to bcareful abouwhere the 'tc' command gets put. Somdistributions puit in /sbin some in /usr/sbin. The make install will puiin /usr/sbin, so check and delete old versions. Frobugreporat peshkin.net Thu Jun 2 11:45:05 2005 From: bugreporapeshkin.net (Joel Peshkin) Date: Wed Apr 18 12:51:17 2007 Subject: Jitter problems with Netem?? In-Reply-To: <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> References: <429E71DA.30001@xxxxxxxxxxx> <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> Message-ID: <429F53B1.7000709@xxxxxxxxxxx> StepheHemminger wrote: > <snip> > >First, which kernel version. Thermay bbugs that already have been fixed. >Thexpected behaviour would is to preservorder always, but > >Pack Arrive Delay Depart >A 0 5 5 >B 10 10 20 >C 20 20 40 >D 30 5 40 >E 40 10 50 >F 50 20 70 > >Whenetegets to looking at packet C, it's time has already expired so >ican go righaway. I did have a patch that caused C to go out ahead of B >buuser's complained thathey did not expect time based reordering and >thabehaviour was onof the things they hated about Nist net. > > Warusing 2.6.11 and will happily move to the latest or use a patch. Thosusers musnot have been involved in testing jitter buffer adaptation. Thtime-based reordering is critically important. Is i possiblthayou could make the patch available or, better, make it a selectablbehavior? Thank You, Joel Peshkin Froshemminger aosdl.org Fri Jun 3 10:45:54 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: beginner with netem In-Reply-To: <BAY104-F223BCFC67BFCCBE63DAE0F98070@xxxxxxx> References: <20050602113516.113df76c@xxxxxxxxxxxxxxxxx> <BAY104-F223BCFC67BFCCBE63DAE0F98070@xxxxxxx> Message-ID: <20050603104554.06dcb1bb@xxxxxxxxxxxxxxxxx> O > > > i havLinux Mandrak10.1, and i am interested by Netem. > > > I havgo: > > > - nistnet.2.0.12b.tar.gz > > don'need thafor netem > > > - iproute2-2.6-ss040701.tar.gz > > that's probably older thaMandrake. > > > > > > > > > Bui havfew problems: > > > > > > - i don'know compil and install .tar.gz ( i havread the README) > > > - linux don'wansee the two network interfaces (either one or other) > > > > > > this projecis importanfor me. > > > > > > thank > > > > > > luc > > >Firsoff, try iwithout compiling anything. Look at the version > >of iproute2 installed with Mandrake. > > > > tc -V > >will show version. Also, if /usr/lib/tc exists and contains q_netem.so > >and a bunch of .disfiles, you almoscertainly have a good enough > >version. > > > >To giva simpltest, just do (as root): > > tc qdisc add dev eth0 roonetem > >If isucceeds then you havjust put netem (with no options) > >ias thqdisc for the normal default ether device. Netem with no > >options is jusa simplFIFO so it is mostly harmless > >(1ms delay oearlier kernels pr2.6.12). > > > >Building iproute2 is relatively painless. Download latesfrom > > http://developer.osdl.org/dev/iproute2/download > >theunpack with > > tar xzf iproute2-2.6.11-050330.targ.gz > >to build > > cd iproute2-2.6.11-050330 > > make > >to install (as root) > > makinstall > >You do need to bcareful abouwhere the 'tc' command gets put. > >Somdistributions puit in /sbin some in /usr/sbin. The make install > >will puiin /usr/sbin, so check and delete old versions. > > > Thank for thaswer ! > > my versiois ss060832 bui don't know how delete it (i am beginner with > linux too, mandrake10.1) Looks lika vendor version, noin any of my iproute2 archives. To removold versions and do thinstall do the following: wgehttp://developer.osdl.org/dev/iproute2/iproute2-2.6.11-050330.tar.gz tar xzf iproute2-2.6.11-050330.tar.gz cd iproute2-2.6.11-050330 make # becomroot: sudo -s # or su - r`which tc` # removold tc wherever it is makinstall > i havanother question: jusiproute2 is sufficient for build an wan > simulator ? > Therartwo parts the kernel (which is in 2.4 and 2.6) and thiproute2 command, you need both. Sewebsite: http://developer.osdl.org/shemminger/netem and paper http://developer.osdl.org/shemminger/netem/LCA2005_paper.pdf and talk http://developer.osdl.org/shemminger/netem/LCA2005_netem.pdf Frotzenes agmail.com Mon Jun 6 13:30:06 2005 From: tzenes agmail.co(Sam Rossoff) Date: Wed Apr 18 12:51:17 2007 Subject: Problems with Netem: RTNETLINK answers: Invalid argument Message-ID: <467fbc4e050606133029cd5d38@xxxxxxxxxxxxxx> I recently reformatted onof my computers and installed slackware 10.0 oito act as a router. Now I wanted to use Netem as network emulatiosoftwarto set artifical drop rates and delays to test a veriaty of programs. Now, whei upgraded to thlatest linux kernel (2.6.11) for somreason ididn't come with netem (or i could find it whecompiling my kernel). So I did thlogical thing and downloaded thlatescopy of iproute2 (2.6.11-050330) which comes with netem. Now wheI run: #tc qdisc add dev eth0 roonetehelp i gea nichelp message: Usage: ... nete[ limiPACKETS ] [ delay TIME [ JITTER [CORRELATION]]] ... etc however if i try to usnetem #tc qdisc add dev eth0 roonete i gethfollowing error: RTNETLINK answers: Invalid argument I'vtried adding options: #tc qdisc add dev eth0 roonetedelay 100ms #tc qdisc add dev eth0 roohandl1: netem delay 100ms ... etc to no avail whei misspell an option, i get: #tc qdisc add dev eth0 roonetedealy 100ms Whais "dealy"? followed by thhelp message, so i know neteis installed and working to comextent now tc qdisc works finwith p_fifo and other options ionly seems to bfailing with netem evewhen i try to makit the child of another qdisc it continues to givmthe same error: RTNETLINK answers: Invalid argument I evechecked my config filto see if i enabled CONFIG_IP_MULTIPLE_TABLES and iis indeed =y now this probaleis fairly vexing, so i was wondering if you could help mout? This is pretty much a brand new kernel and filsystem, and thlatest versioof IProute2 so I'not sure what the problem could be Frotzenes agmail.com Mon Jun 6 13:58:43 2005 From: tzenes agmail.co(Sam Rossoff) Date: Wed Apr 18 12:51:17 2007 Subject: Problems with Netem: RTNETLINK answers: Invalid argument In-Reply-To: <467fbc4e050606133029cd5d38@xxxxxxxxxxxxxx> References: <467fbc4e050606133029cd5d38@xxxxxxxxxxxxxx> Message-ID: <467fbc4e050606135813f9045@xxxxxxxxxxxxxx> I recently reformatted onof my computers and installed slackware 10.0 oito act as a router. Now I wanted to use Netem as network emulatiosoftwarto set artifical drop rates and delays to test a veriaty of programs. Now, whei upgraded to thlatest linux kernel (2.6.11) for somreason ididn't come with netem (or i could find it whecompiling my kernel). So I did thlogical thing and downloaded thlatescopy of iproute2 (2.6.11-050330) which comes with netem. Now wheI run: #tc qdisc add dev eth0 roonetehelp i gea nichelp message: Usage: ... nete[ limiPACKETS ] [ delay TIME [ JITTER [CORRELATION]]] ... etc however if i try to usnetem #tc qdisc add dev eth0 roonete i gethfollowing error: RTNETLINK answers: Invalid argument I'vtried adding options: #tc qdisc add dev eth0 roonetedelay 100ms #tc qdisc add dev eth0 roohandl1: netem delay 100ms ... etc to no avail whei misspell an option, i get: #tc qdisc add dev eth0 roonetedealy 100ms Whais "dealy"? followed by thhelp message, so i know neteis installed and working to comextent now tc qdisc works finwith p_fifo and other options ionly seems to bfailing with netem evewhen i try to makit the child of another qdisc it continues to givmthe same error: RTNETLINK answers: Invalid argument I evechecked my config filto see if i enabled CONFIG_IP_MULTIPLE_TABLES and iis indeed =y now this probaleis fairly vexing, so i was wondering if you could help me out? This is pretty much a brand new kernel and filsystem, and thlatest versioof IProute2 so I'not sure what the problem could be ps. I appologizif this duplicates, i'not sure how this list is supposed to work Frostephen.earl abt.com Mon Jun 13 06:18:11 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: Questioon new reordering rule Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3156D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Ithprevious version of netem that just had the simple gap rule, the reordering worked like the following: Exery X packets are delayed by the amount specified in the delay part of the command, and the rest are not delayed. Now with thnew reorder command iworks like this: All packets are delayed by the amount specified in the delay part of the command apart from the packets randomly selected by the reorder part of the command and with a gap of X packets. So beforI would, say, sethe gap to 2 and the delay to 10ms and the packets would have a delay like: '0ms 0ms 10ms 0ms 0ms 10ms 0ms 0ms 10ms' etc. Now if I sedelay to 10ms, reorder to 100% and gap to 2 igoes like this: '10ms 10ms 0ms 10ms 10ms 0ms 10ms 10ms 0ms' etc. Why is this? And is ireally how a real lifnetwork would or could behave? Froshemminger aosdl.org Thu Jun 16 13:50:19 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: Questioon new reordering rule In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3156D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3156D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050616135019.3d6c32f0@xxxxxxxxxxxxxxxxx> OMon, 13 Jun 2005 14:18:11 +0100 <stephen.earl@xxxxxx> wrote: > > Ithprevious version of netem that just had the simple gap rule, the reordering worked like the following: Exery X packets are delayed by the amount specified in the delay part of the command, and the rest are not delayed. > > Now with thnew reorder command iworks like this: All packets are delayed by the amount specified in the delay part of the command apart from the packets randomly selected by the reorder part of the command and with a gap of X packets. > > So beforI would, say, sethe gap to 2 and the delay to 10ms and the packets would have a delay like: '0ms 0ms 10ms 0ms 0ms 10ms 0ms 0ms 10ms' etc. > > Now if I sedelay to 10ms, reorder to 100% and gap to 2 igoes like this: '10ms 10ms 0ms 10ms 10ms 0ms 10ms 10ms 0ms' etc. > > Why is this? And is ireally how a real lifnetwork would or could behave? > They why is simple, becausiseemed like a natural way to map the two parameters of reordering (probability and distance), into existing api with minimal upset. 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. Froshemminger aosdl.org Thu Jun 16 13:54:13 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: Jitter problems with Netem?? In-Reply-To: <429F53B1.7000709@xxxxxxxxxxx> References: <429E71DA.30001@xxxxxxxxxxx> <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> <429F53B1.7000709@xxxxxxxxxxx> Message-ID: <20050616135413.4eafe87b@xxxxxxxxxxxxxxxxx> OThu, 02 Jun 2005 11:45:05 -0700 Joel Peshki<bugreport@xxxxxxxxxxx> wrote: > StepheHemminger wrote: > > > <snip> > > > >First, which kernel version. Thermay bbugs that already have been fixed. > >Thexpected behaviour would is to preservorder always, but > > > >Pack Arrive Delay Depart > >A 0 5 5 > >B 10 10 20 > >C 20 20 40 > >D 30 5 40 > >E 40 10 50 > >F 50 20 70 > > > >Whenetegets to looking at packet C, it's time has already expired so > >ican go righaway. I did have a patch that caused C to go out ahead of B > >buuser's complained thathey did not expect time based reordering and > >thabehaviour was onof the things they hated about Nist net. > > > > > Warusing 2.6.11 and will happily move to the latest or use a patch. > > Thosusers musnot have been involved in testing jitter buffer > adaptation. Thtime-based reordering is critically important. Is i > possiblthayou could make the patch available or, better, make it a > selectablbehavior? > After 2.6.12, I will revisithis. My "besidea so far" is to: * gerid of inner qdisc completely; thmain use now is to pusomething likTBF inside netem, but putting netem inside TBF works as well. * jususa sk_buff queue directly * fix reorder to pupackets back rather than at head of queue * allow timbased reordering as an option (or modulparameter, not sure). * maybadd somform RED support. Froji.li3 ahp.com Thu Jun 23 08:14:43 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: CaNetEbe bound to one interface without affecting the other? Message-ID: <628BFCE8B64706469FE4D4852CEC953706B090CC@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Dear all, I havtwo machines, each with two network cards. I wanto connect two eth0 and two eth1 respectively so thaI havtwo connections between thtwo machines. I havdirectly connected two eth0, but I want to connectwo eth1 in a way thaI can control the properties of this connection. Iseems thaNetEm is a good choice. The question is: can NetEbbound to one interface (eth1 in this case)? Otherwise, I will need a third computer as a gateway or bridgbetween two eth1 and run NetEon it. Is icorrect? Thanks a lot! -Ji Froshemminger aosdl.org Thu Jun 23 08:44:24 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: CaNetEbe bound to one interface without affecting the other? In-Reply-To: <628BFCE8B64706469FE4D4852CEC953706B090CC@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <628BFCE8B64706469FE4D4852CEC953706B090CC@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050623084424.7925993b@xxxxxxxxxxxxxxxxx> OThu, 23 Jun 2005 11:14:43 -0400 "Li, Ji" <ji.li3@xxxxxx> wrote: > Dear all, > > I havtwo machines, each with two network cards. I wanto connect two > eth0 and two eth1 respectively so thaI havtwo connections between > thtwo machines. I havdirectly connected two eth0, but I want to > connectwo eth1 in a way thaI can control the properties of this > connection. Iseems thaNetEm is a good choice. The question is: can > NetEbbound to one interface (eth1 in this case)? Otherwise, I will > need a third computer as a gateway or bridgbetween two eth1 and run > NetEon it. Is icorrect? > Thanks a lot! > > -Ji Yes neteis always bound to an interface. Froneteat smyles.plus.com Thu Jun 23 13:29:30 2005 From: neteasmyles.plus.com (Barry Myles) Date: Wed Apr 18 12:51:17 2007 Subject: phpnetemgui 0.9 released Message-ID: <42BB1BAA.4040008@xxxxxxxxxxxxxxx> Jusreleased version 0.9 of phpnetemgui. This is a seof php web pages thaallow you to control thsetup of netem in a reasonable easy way. New ithis version: - ability to add multiplIP matching rules to an interface. - reorder supporon systems wherit is available. http://www.smyles.plus.com/phpnetemgui/ -- Barry Myles Froji.li3 ahp.com Thu Jun 23 14:43:43 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: Help: can'add netem Message-ID: <628BFCE8B64706469FE4D4852CEC953706B09239@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Dear all, I jusstarted to usNetEm today. When I tried to add netem, I got the following error message: # tc qdisc add dev eth1 roonetedelay 100ms RTNETLINK answers: Filexists Whadoes therror message mean? I suspect it's related to the teql I added before. WhaI did beforthat is to add teql, since I need both teql and netefor my work. Thcommands are: # 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 10.0.0.10/30 # ip addr add dev teql0 10.0.0.14/30 I was wondering if TEQL and NetEconflicwith each other or not. Can eth1 only badded to both teql and netem? Thanks, -Ji Frostephen.earl abt.com Fri Jun 24 04:01:30 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 12:51:17 2007 Subject: Help: can'add netem Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94EA@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Yes, you gethis error becausyou already did the 'tc qdisc add dev eth1 root teql0' command. To use the teql0 and netem qdisc together you can do the following commands instead: tc qdisc add dev eth0 roohandl1: teql0 tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms You caonly add onqdisc at the root level in an interface. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 23 Jun2005 22:44 To: netem@xxxxxxxxxxxxxx Subject: Help: can'add netem Dear all, I jusstarted to usNetEm today. When I tried to add netem, I got the following error message: # tc qdisc add dev eth1 roonetedelay 100ms RTNETLINK answers: Filexists Whadoes therror message mean? I suspect it's related to the teql I added before. WhaI did beforthat is to add teql, since I need both teql and netefor my work. Thcommands are: # 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 10.0.0.10/30 # ip addr add dev teql0 10.0.0.14/30 I was wondering if TEQL and NetEconflicwith each other or not. Can eth1 only badded to both teql and netem? Thanks, -Ji Froji.li3 ahp.com Fri Jun 24 07:27:07 2005 From: ji.li3 ahp.co(Ji Li) Date: Wed Apr 18 12:51:17 2007 Subject: Help: can'add netem Message-ID: <42071.161.114.64.74.1119623227.squirrel@xxxxxxxxxxxxxxx> Thanks you all for thkind help to a newbie! I believI avery close to what I want, but I still got a little problem, hopyou guys would hava look. Therror messagI got is (I need packets to go to teql0 first, and then bforwarded to netem): [root@princeton2 mtcp]# tc qdisc add dev eth0 roohandl1: teql0 [root@princeton2 mtcp]# tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms RTNETLINK answers: Invalid argument I searched for a while, and guess thait's becausI haven't loaded all relevenmodules. Thfollowing is copied from /lib/modules/2.6.9-1.667smp/build/.config . Would you pleasleme know whaI need to do to fix this problem? # # QoS and/or fair queueing # CONFIG_NET_SCHED=y CONFIG_NET_SCH_CLK_JIFFIES=y # CONFIG_NET_SCH_CLK_GETTIMEOFDAY is noset # CONFIG_NET_SCH_CLK_CPU is noset CONFIG_NET_SCH_CBQ=m CONFIG_NET_SCH_HTB=m CONFIG_NET_SCH_HFSC=m CONFIG_NET_SCH_ATM=m CONFIG_NET_SCH_PRIO=m CONFIG_NET_SCH_RED=m CONFIG_NET_SCH_SFQ=m CONFIG_NET_SCH_TEQL=m CONFIG_NET_SCH_TBF=m CONFIG_NET_SCH_GRED=m CONFIG_NET_SCH_DSMARK=m CONFIG_NET_SCH_NETEM=m CONFIG_NET_SCH_INGRESS=m CONFIG_NET_QOS=y CONFIG_NET_ESTIMATOR=y CONFIG_NET_CLS=y CONFIG_NET_CLS_TCINDEX=m CONFIG_NET_CLS_ROUTE4=m CONFIG_NET_CLS_ROUTE=y CONFIG_NET_CLS_FW=m CONFIG_NET_CLS_U32=m CONFIG_CLS_U32_PERF=y CONFIG_NET_CLS_IND=y CONFIG_NET_CLS_RSVP=m CONFIG_NET_CLS_RSVP6=m # CONFIG_NET_CLS_ACT is noset CONFIG_NET_CLS_POLICE=y Best, -Ji > Yes, you gethis error becausyou already did the 'tc qdisc add dev eth1 > rooteql0' command. To usthe teql0 and netem qdisc together you can do > thfollowing commands instead: > > tc qdisc add dev eth0 roohandl1: teql0 > tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms > > You caonly add onqdisc at the root level in an interface. > > -----Original Message----- > From: netem-bounces@xxxxxxxxxxxxxx > [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji > Sent: 23 Jun2005 22:44 > To: netem@xxxxxxxxxxxxxx > Subject: Help: can'add netem > > > Dear all, > > I jusstarted to usNetEm today. When I tried to add netem, I got the > following error message: > > # tc qdisc add dev eth1 roonetedelay 100ms > RTNETLINK answers: Filexists > > Whadoes therror message mean? I suspect it's related to the teql I > added before. WhaI did beforthat is to add teql, since I need both > teql and netefor my work. Thcommands are: > # 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 10.0.0.10/30 > # ip addr add dev teql0 10.0.0.14/30 > > I was wondering if TEQL and NetEconflicwith each other or not. Can > eth1 only badded to both teql and netem? > > > Thanks, > -Ji > > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx > http://lists.osdl.org/mailman/listinfo/netem > Froji.li3 ahp.com Fri Jun 24 09:34:43 2005 From: ji.li3 ahp.co(Ji Li) Date: Wed Apr 18 12:51:17 2007 Subject: Help: can'add netem In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3157B@i2km34-ukdy.domain1.systemho st.net> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3157B@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <55139.161.114.64.75.1119630883.squirrel@xxxxxxxxxxxxxxx> Yes, thescommands work! Thank you! However (I hatthis word!), after I add both eth0 and eth1 to teql0, I ping thteql0 on another machine, all ping traffic goes to eth1, none goes to eth0. Othother machine, I didn't add netem but only added eth0 and eth1 to teql0, and ping thteql0 on thfirst machine, and it works as expected: ping traffic goes to both interfaces. Attached arthe commands I used to configurteql0 on both machines, routing tables and "tc qdisc ls" results. Whacould bthe problem? Maybe the command to add eth1 to teql0 is wrong whertherare hierarchy? SORRY to bother you guys, and your help is highly appreciated! ============== Machin1 with NetE================ commands to install NetEand TEQL together: modprobsch_teql tc qdisc add dev eth0 roohandl1: netem delay 100ms 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 10.0.0.0/31 ip addr add dev eth1 10.0.0.2/31 ip addr add dev teql0 10.0.0.4/31 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 UsIface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 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 10.0.0.1/31 ip addr add dev eth1 10.0.0.3/31 ip addr add dev teql0 10.0.0.5/31 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 UsIface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 U 0 0 0 eth1 > I'vjusfound out that those two commands I gave you don't work for me > either. You havto puthe netem qdisc first: > > tc qdisc add dev eth0 roohandl1: netem delay 100ms > tc qdisc add dev eth0 paren1:1 handl10: teql0 > > Thashould work. > > -----Original Message----- > From: Ji Li [mailto:ji.li3@xxxxxx] > Sent: 24 Jun2005 15:27 > To: Earl,S,Stephen,XDP65A C; nistnet_user@xxxxxxxxx > Cc: netem@xxxxxxxxxxxxxx > Subject: RE: Help: can'add netem > > > Thanks you all for thkind help to a newbie! > > I believI avery close to what I want, but I still got a little > problem, hopyou guys would hava look. > > Therror messagI got is (I need packets to go to teql0 first, and then > bforwarded to netem): > [root@princeton2 mtcp]# tc qdisc add dev eth0 roohandl1: teql0 > [root@princeton2 mtcp]# tc qdisc add dev eth0 paren1:1 handl10: netem > delay 100ms > RTNETLINK answers: Invalid argument > > I searched for a while, and guess thait's becausI haven't loaded all > relevenmodules. Thfollowing is copied from > /lib/modules/2.6.9-1.667smp/build/.config . Would you pleasleme know > whaI need to do to fix this problem? > > # > # QoS and/or fair queueing > # > CONFIG_NET_SCHED=y > CONFIG_NET_SCH_CLK_JIFFIES=y > # CONFIG_NET_SCH_CLK_GETTIMEOFDAY is noset > # CONFIG_NET_SCH_CLK_CPU is noset > CONFIG_NET_SCH_CBQ=m > CONFIG_NET_SCH_HTB=m > CONFIG_NET_SCH_HFSC=m > CONFIG_NET_SCH_ATM=m > CONFIG_NET_SCH_PRIO=m > CONFIG_NET_SCH_RED=m > CONFIG_NET_SCH_SFQ=m > CONFIG_NET_SCH_TEQL=m > CONFIG_NET_SCH_TBF=m > CONFIG_NET_SCH_GRED=m > CONFIG_NET_SCH_DSMARK=m > CONFIG_NET_SCH_NETEM=m > CONFIG_NET_SCH_INGRESS=m > CONFIG_NET_QOS=y > CONFIG_NET_ESTIMATOR=y > CONFIG_NET_CLS=y > CONFIG_NET_CLS_TCINDEX=m > CONFIG_NET_CLS_ROUTE4=m > CONFIG_NET_CLS_ROUTE=y > CONFIG_NET_CLS_FW=m > CONFIG_NET_CLS_U32=m > CONFIG_CLS_U32_PERF=y > CONFIG_NET_CLS_IND=y > CONFIG_NET_CLS_RSVP=m > CONFIG_NET_CLS_RSVP6=m > # CONFIG_NET_CLS_ACT is noset > CONFIG_NET_CLS_POLICE=y > > Best, > -Ji > >> Yes, you gethis error becausyou already did the 'tc qdisc add dev >> eth1 >> rooteql0' command. To usthe teql0 and netem qdisc together you can >> do >> thfollowing commands instead: >> >> tc qdisc add dev eth0 roohandl1: teql0 >> tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms >> >> You caonly add onqdisc at the root level in an interface. >> >> -----Original Message----- >> From: netem-bounces@xxxxxxxxxxxxxx >> [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji >> Sent: 23 Jun2005 22:44 >> To: netem@xxxxxxxxxxxxxx >> Subject: Help: can'add netem >> >> >> Dear all, >> >> I jusstarted to usNetEm today. When I tried to add netem, I got the >> following error message: >> >> # tc qdisc add dev eth1 roonetedelay 100ms >> RTNETLINK answers: Filexists >> >> Whadoes therror message mean? I suspect it's related to the teql I >> added before. WhaI did beforthat is to add teql, since I need both >> teql and netefor my work. Thcommands are: >> # 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 10.0.0.10/30 >> # ip addr add dev teql0 10.0.0.14/30 >> >> I was wondering if TEQL and NetEconflicwith each other or not. Can >> eth1 only badded to both teql and netem? >> >> >> Thanks, >> -Ji >> >> >> >> _______________________________________________ >> Netemailing list >> Netem@xxxxxxxxxxxxxx >> http://lists.osdl.org/mailman/listinfo/netem >> > > > Froji.li3 ahp.com Fri Jun 24 10:05:13 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: Help: can'add netem Message-ID: <628BFCE8B64706469FE4D4852CEC953706B09400@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I havsomother IP addresses assigned to the same machines, which enables mto access thInternet. Those configuration may affect the result. I acleaning up everything to tesagain. Thanks and sorry to spam! Besregards, -Ji -----Original Message----- From: Li, Ji Sent: Friday, Jun24, 2005 12:35 PM To: stephen.earl@xxxxxx Cc: netem@xxxxxxxxxxxxxx Subject: RE: Help: can'add netem Yes, thescommands work! Thank you! However (I hatthis word!), after I add both eth0 and eth1 to teql0, I ping thteql0 on another machine, all ping traffic goes to eth1, none goes to eth0. Othother machine, I didn't add netem but only added eth0 and eth1 to teql0, and ping thteql0 on thfirst machine, and it works as expected: ping traffic goes to both interfaces. Attached are thcommands I used to configurteql0 on both machines, routing tables and "tc qdisc ls" results. Whacould bthe problem? Maybe the command to add eth1 to teql0 is wrong whertherare hierarchy? SORRY to bother you guys, and your help is highly appreciated! ============== Machin1 with NetE================ commands to install NetEand TEQL together: modprobsch_teql tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: teql0 tc qdisc add dev eth1 root teql0 ip link sedev teql0 up ip addr add dev eth0 10.0.0.0/31 ip addr add dev eth1 10.0.0.2/31 ip addr add dev teql0 10.0.0.4/31 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 Use Iface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 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 10.0.0.1/31 ip addr add dev eth1 10.0.0.3/31 ip addr add dev teql0 10.0.0.5/31 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 Use Iface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 U 0 0 0 eth1 > I'vjusfound out that those two commands I gave you don't work for > meither. You havto put the netem qdisc first: > > tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add > dev eth0 paren1:1 handl10: teql0 > > Thashould work. > > -----Original Message----- > From: Ji Li [mailto:ji.li3@xxxxxx] > Sent: 24 Jun2005 15:27 > To: Earl,S,Stephen,XDP65A C; nistnet_user@xxxxxxxxx > Cc: netem@xxxxxxxxxxxxxx > Subject: RE: Help: can'add netem > > > Thanks you all for thkind help to a newbie! > > I believI avery close to what I want, but I still got a little > problem, hopyou guys would hava look. > > Therror messagI got is (I need packets to go to teql0 first, and > thebforwarded to netem): > [root@princeton2 mtcp]# tc qdisc add dev eth0 roohandl1: teql0 > [root@princeton2 mtcp]# tc qdisc add dev eth0 paren1:1 handl10: > netedelay 100ms RTNETLINK answers: Invalid argument > > I searched for a while, and guess thait's becausI haven't loaded > all relevenmodules. Thfollowing is copied from > /lib/modules/2.6.9-1.667smp/build/.config . Would you pleasleme > know whaI need to do to fix this problem? > > # > # QoS and/or fair queueing > # > CONFIG_NET_SCHED=y > CONFIG_NET_SCH_CLK_JIFFIES=y > # CONFIG_NET_SCH_CLK_GETTIMEOFDAY is nose# CONFIG_NET_SCH_CLK_CPU > is noseCONFIG_NET_SCH_CBQ=m CONFIG_NET_SCH_HTB=m > CONFIG_NET_SCH_HFSC=CONFIG_NET_SCH_ATM=CONFIG_NET_SCH_PRIO=m > CONFIG_NET_SCH_RED=CONFIG_NET_SCH_SFQ=CONFIG_NET_SCH_TEQL=m > CONFIG_NET_SCH_TBF=CONFIG_NET_SCH_GRED=CONFIG_NET_SCH_DSMARK=m > CONFIG_NET_SCH_NETEM=CONFIG_NET_SCH_INGRESS=CONFIG_NET_QOS=y > CONFIG_NET_ESTIMATOR=y CONFIG_NET_CLS=y CONFIG_NET_CLS_TCINDEX= > CONFIG_NET_CLS_ROUTE4=CONFIG_NET_CLS_ROUTE=y CONFIG_NET_CLS_FW= > CONFIG_NET_CLS_U32=CONFIG_CLS_U32_PERF=y CONFIG_NET_CLS_IND=y > CONFIG_NET_CLS_RSVP=CONFIG_NET_CLS_RSVP6=# CONFIG_NET_CLS_ACT is > noseCONFIG_NET_CLS_POLICE=y > > Best, > -Ji > >> Yes, you gethis error becausyou already did the 'tc qdisc add dev >> eth1 >> rooteql0' command. To usthe teql0 and netem qdisc together you >> cado thfollowing commands instead: >> >> tc qdisc add dev eth0 roohandl1: teql0 tc qdisc add dev eth0 >> paren1:1 handl10: netem delay 100ms >> >> You caonly add onqdisc at the root level in an interface. >> >> -----Original Message----- >> From: netem-bounces@xxxxxxxxxxxxxx >> [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji >> Sent: 23 Jun2005 22:44 >> To: netem@xxxxxxxxxxxxxx >> Subject: Help: can'add netem >> >> >> Dear all, >> >> I jusstarted to usNetEm today. When I tried to add netem, I got >> thfollowing error message: >> >> # tc qdisc add dev eth1 roonetedelay 100ms RTNETLINK answers: >> Filexists >> >> Whadoes therror message mean? I suspect it's related to the teql >> I added before. WhaI did beforthat is to add teql, since I need >> both teql and netefor my work. Thcommands are: >> # 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 10.0.0.10/30 >> # ip addr add dev teql0 10.0.0.14/30 >> >> I was wondering if TEQL and NetEconflicwith each other or not. >> Can >> eth1 only badded to both teql and netem? >> >> >> Thanks, >> -Ji >> >> >> >> _______________________________________________ >> Netemailing list >> Netem@xxxxxxxxxxxxxx >> http://lists.osdl.org/mailman/listinfo/netem >> > > > Froji.li3 ahp.com Fri Jun 24 11:56:26 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 12:51:17 2007 Subject: How to makneteand teql work together? Message-ID: <628BFCE8B64706469FE4D4852CEC953706B09490@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> (I asked thsamquestion in previous thread, but now the problem is a littldifferent, so I stara new thread. If I spam too much, please lemknow and I'll stop :P ) SORRY to bother you guys, and your help is highly appreciated! Shorstory: when I load neteand teql together, teql doesn't work correctly. (If I load teql only, everything is fine) I load both neteand teql. Neteis associated with eth0, and teql is associated with both eth0 and eth1. Buafter that, traffic only goes ouof eth1. I also tried loading teql only (with netem), then teql works as expected: ping traffic goes to both interfaces. Attached are thcommands thaI used to configure teql and netem (machine 1), and teql only (machie2), their routing tables and "tc qdisc ls" results. What's wrong? ============== 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 Use 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 Use 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 Froamanol aiit.demokritos.gr Wed Jun 29 03:17:25 2005 From: amanol aiit.demokritos.gr (Apostolos Manolitzas) Date: Wed Apr 18 12:51:17 2007 Subject: Interfacfreezes with delay Message-ID: <42C27535.1000105@xxxxxxxxxxxxxxxxx> Hello all, I try to usneteto add delay to my network. I havthfollowing configuration: Sus9.3 Linux cor2.6.11.4-20a-defaul#1 Mon Jun 27 17:19:03 EEST 2005 i686 i686 i386 GNU/Linux Na : iproute2 Relocations: (not relocatable) Versio : 2.6.10 Vendor: SUSE LINUX Products GmbH, Nuernberg, Germany Releas : 4 Build Date: Tu22 Mar 2005 08:49:29 PM EET and I usthfollowing command: tc qdisc add dev eth1 roonetedelay 240ms Thping works finand I get the proper delay. Buafter pushing traffic of 3Mbit/s for 4-5 minutes thinterface freezes, and I havto restarit. I can'figurout what is the reason of it. I havtested with kernel 2.6.8.x and I hadthe same results. Does anyonhavany clue about it? regards, Apostolos Froshemminger aosdl.org Wed Jun 29 14:09:31 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 12:51:17 2007 Subject: Interfacfreezes with delay In-Reply-To: <42C27535.1000105@xxxxxxxxxxxxxxxxx> References: <42C27535.1000105@xxxxxxxxxxxxxxxxx> Message-ID: <20050629140931.3a129086@xxxxxxxxxxxxxxxxx> OWed, 29 Jun 2005 13:17:25 +0300 Apostolos Manolitzas <amanol@xxxxxxxxxxxxxxxxx> wrote: > Hello all, > > I try to usneteto add delay to my network. > I havthfollowing configuration: > Sus9.3 > Linux cor2.6.11.4-20a-defaul#1 Mon Jun 27 17:19:03 EEST 2005 i686 > i686 i386 GNU/Linux > > Na : iproute2 Relocations: (not relocatable) > Versio : 2.6.10 Vendor: SUSE LINUX > Products GmbH, Nuernberg, Germany > Releas : 4 Build Date: Tu22 Mar 2005 > 08:49:29 PM EET > > and I usthfollowing command: > tc qdisc add dev eth1 roonetedelay 240ms > > Thping works finand I get the proper delay. > Buafter pushing traffic of 3Mbit/s for 4-5 minutes thinterface freezes, > and I havto restarit. > I can'figurout what is the reason of it. > I havtested with kernel 2.6.8.x and I hadthe same results. > > Does anyonhavany clue about it? > I assumyou arseeing the network interface stop. Could you look astats with ifconfig and seit is still receiving? Whais thnetwork interface? I have seen problems with older versions of the1000 driver when thtransmit ring gets overlaoded. Frojuliokriger agmail.com Thu Jun 30 06:21:34 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 12:51:17 2007 Subject: neteand sctplib on user space Message-ID: <682bc30a05063006217c2d112e@xxxxxxxxxxxxxx> Hi! Yesterday I testes a user spacSCTP library with netem. I sea delay of 1 second o"lo". Then, started two prograthat send packages to onanother. Igavme the impresion that netem didn't delay the packages. Does neteworking with user spaclibraries? Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Froamanol aiit.demokritos.gr Thu Jun 30 06:38:52 2005 From: amanol aiit.demokritos.gr (Apostolos Manolitzas) Date: Wed Apr 18 12:51:17 2007 Subject: Interfacfreezes with delay In-Reply-To: <20050629140931.3a129086@xxxxxxxxxxxxxxxxx> References: <42C27535.1000105@xxxxxxxxxxxxxxxxx> <20050629140931.3a129086@xxxxxxxxxxxxxxxxx> Message-ID: <42C3F5EC.3080300@xxxxxxxxxxxxxxxxx> StepheHemminger wrote: >OWed, 29 Jun 2005 13:17:25 +0300 >Apostolos Manolitzas <amanol@xxxxxxxxxxxxxxxxx> wrote: > > >>Hello all, >> >>I try to usneteto add delay to my network. >>I havthfollowing configuration: >>Sus9.3 >>Linux cor2.6.11.4-20a-defaul#1 Mon Jun 27 17:19:03 EEST 2005 i686 >>i686 i386 GNU/Linux >> >>Na : iproute2 Relocations: (not relocatable) >>Versio : 2.6.10 Vendor: SUSE LINUX >>Products GmbH, Nuernberg, Germany >>Releas : 4 Build Date: Tu22 Mar 2005 >>08:49:29 PM EET >> >>and I usthfollowing command: >>tc qdisc add dev eth1 roonetedelay 240ms >> >>Thping works finand I get the proper delay. >>Buafter pushing traffic of 3Mbit/s for 4-5 minutes thinterface freezes, >>and I havto restarit. >>I can'figurout what is the reason of it. >>I havtested with kernel 2.6.8.x and I hadthe same results. >> >>Does anyonhavany clue about it? >> >> > >I assumyou arseeing the network interface stop. >Could you look astats with ifconfig and seit is still receiving? > >Whais thnetwork interface? I have seen problems with older versions >of the1000 driver when thtransmit ring gets overlaoded. > > Whewmean freezes, the interfaces can't be accessed with ping or anything else. I havto ifdown eth0 and then ifup eth0 to starworking again. If during "freezing" state, I removthdelay the I/F is up and running again. I checked thstats, and idoesn't receive anything. FurthermorI changed thdelay to 120ms but the interface freezes again. Also thfreezin timis not proportional to the delay. Is therany way to identify why thinterface is self-blocked? Frobugreporat peshkin.net Wed Jun 1 19:41:30 2005 From: bugreporapeshkin.net (Joel Peshkin) Date: Wed Apr 18 17:37:47 2007 Subject: Jitter problems with Netem?? Message-ID: <429E71DA.30001@xxxxxxxxxxx> Iappears thaNetem drops packets when the magnitude of jitter goes beyond thinter-packedelay. For example.... Assumthaour jitter distribution would have a series of packets delayed by (milliseconds) 5, 10, 20, 5, 10, 20 and thapackets A-F arriva10 millisecond intervals... This should resulin thfollowing timings... Pack Arrive Depart A 0 5 B 10 20 C 20 40 D 30 35 E 40 50 F 50 70 So, thpackets should go ouin the order A, B, D, C, E, F However, iseems likthis causes packet C to be dropped. This looks pretty serious, sincthawould mean there is no jitter distribution thacould bused under live conditions without inducing packet loss. Is this really thcasor have I misunderstood something? -Joel Peshkin Froking_turlutor ahotmail.com Thu Jun 2 07:32:33 2005 From: king_turlutor ahotmail.co(Tutu Lord) Date: Wed Apr 18 17:37:47 2007 Subject: beginner with netem Message-ID: <BAY104-F4D075E450C38C5569B68098060@xxxxxxx> hello, i havLinux Mandrak10.1, and i am interested by Netem. I havgo: - nistnet.2.0.12b.tar.gz - iproute2-2.6-ss040701.tar.gz Bui havfew problems: - i don'know compil and install .tar.gz ( i havread the README) - linux don'wansee the two network interfaces (either one or other) this projecis importanfor me. thank luc _________________________________________________________________ MSHotmail : antivirus eantispam gratuits http://www.imagine-msn.com/hotmail/default.aspx?locale=fr-FR Froshemminger aosdl.org Thu Jun 2 11:26:38 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: Jitter problems with Netem?? In-Reply-To: <429E71DA.30001@xxxxxxxxxxx> References: <429E71DA.30001@xxxxxxxxxxx> Message-ID: <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> OWed, 01 Jun 2005 19:41:30 -0700 Joel Peshki<bugreport@xxxxxxxxxxx> wrote: > > Iappears thaNetem drops packets when the magnitude of jitter goes > beyond thinter-packedelay. For example.... > > Assumthaour jitter distribution would have a series of packets > delayed by (milliseconds) 5, 10, 20, 5, 10, 20 and thapackets A-F > arriva10 millisecond intervals... > > This should resulin thfollowing timings... > Pack Arrive Depart > A 0 5 > B 10 20 > C 20 40 > D 30 35 > E 40 50 > F 50 70 > So, thpackets should go ouin the order A, B, D, C, E, F > > However, iseems likthis causes packet C to be dropped. This looks > pretty serious, sincthawould mean there is no jitter distribution > thacould bused under live conditions without inducing packet loss. > > Is this really thcasor have I misunderstood something? > > -Joel Peshkin > First, which kernel version. Thermay bbugs that already have been fixed. Thexpected behaviour would is to preservorder always, but Pack Arrive Delay Depart A 0 5 5 B 10 10 20 C 20 20 40 D 30 5 40 E 40 10 50 F 50 20 70 Whenetegets to looking at packet C, it's time has already expired so ican go righaway. I did have a patch that caused C to go out ahead of B buuser's complained thathey did not expect time based reordering and thabehaviour was onof the things they hated about Nist net. Froshemminger aosdl.org Thu Jun 2 11:35:16 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: beginner with netem In-Reply-To: <BAY104-F4D075E450C38C5569B68098060@xxxxxxx> References: <BAY104-F4D075E450C38C5569B68098060@xxxxxxx> Message-ID: <20050602113516.113df76c@xxxxxxxxxxxxxxxxx> OThu, 02 Jun 2005 16:32:33 +0200 "Tutu Lord" <king_turlutor@xxxxxxxxxxx> wrote: > hello, > > i havLinux Mandrak10.1, and i am interested by Netem. > I havgo: > - nistnet.2.0.12b.tar.gz don'need thafor netem > - iproute2-2.6-ss040701.tar.gz that's probably older thaMandrake. > > > Bui havfew problems: > > - i don'know compil and install .tar.gz ( i havread the README) > - linux don'wansee the two network interfaces (either one or other) > > this projecis importanfor me. > > thank > > luc Firsoff, try iwithout compiling anything. Look at the version of iproute2 installed with Mandrake. tc -V will show version. Also, if /usr/lib/tc exists and contains q_netem.so and a bunch of .disfiles, you almoscertainly have a good enough version. To giva simpltest, just do (as root): tc qdisc add dev eth0 roonetem If isucceeds then you havjust put netem (with no options) ias thqdisc for the normal default ether device. Netem with no options is jusa simplFIFO so it is mostly harmless (1ms delay oearlier kernels pr2.6.12). Building iproute2 is relatively painless. Download latesfrom http://developer.osdl.org/dev/iproute2/download theunpack with tar xzf iproute2-2.6.11-050330.targ.gz to build cd iproute2-2.6.11-050330 make to install (as root) makinstall You do need to bcareful abouwhere the 'tc' command gets put. Somdistributions puit in /sbin some in /usr/sbin. The make install will puiin /usr/sbin, so check and delete old versions. Frobugreporat peshkin.net Thu Jun 2 11:45:05 2005 From: bugreporapeshkin.net (Joel Peshkin) Date: Wed Apr 18 17:37:47 2007 Subject: Jitter problems with Netem?? In-Reply-To: <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> References: <429E71DA.30001@xxxxxxxxxxx> <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> Message-ID: <429F53B1.7000709@xxxxxxxxxxx> StepheHemminger wrote: > <snip> > >First, which kernel version. Thermay bbugs that already have been fixed. >Thexpected behaviour would is to preservorder always, but > >Pack Arrive Delay Depart >A 0 5 5 >B 10 10 20 >C 20 20 40 >D 30 5 40 >E 40 10 50 >F 50 20 70 > >Whenetegets to looking at packet C, it's time has already expired so >ican go righaway. I did have a patch that caused C to go out ahead of B >buuser's complained thathey did not expect time based reordering and >thabehaviour was onof the things they hated about Nist net. > > Warusing 2.6.11 and will happily move to the latest or use a patch. Thosusers musnot have been involved in testing jitter buffer adaptation. Thtime-based reordering is critically important. Is i possiblthayou could make the patch available or, better, make it a selectablbehavior? Thank You, Joel Peshkin Froshemminger aosdl.org Fri Jun 3 10:45:54 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: beginner with netem In-Reply-To: <BAY104-F223BCFC67BFCCBE63DAE0F98070@xxxxxxx> References: <20050602113516.113df76c@xxxxxxxxxxxxxxxxx> <BAY104-F223BCFC67BFCCBE63DAE0F98070@xxxxxxx> Message-ID: <20050603104554.06dcb1bb@xxxxxxxxxxxxxxxxx> O > > > i havLinux Mandrak10.1, and i am interested by Netem. > > > I havgo: > > > - nistnet.2.0.12b.tar.gz > > don'need thafor netem > > > - iproute2-2.6-ss040701.tar.gz > > that's probably older thaMandrake. > > > > > > > > > Bui havfew problems: > > > > > > - i don'know compil and install .tar.gz ( i havread the README) > > > - linux don'wansee the two network interfaces (either one or other) > > > > > > this projecis importanfor me. > > > > > > thank > > > > > > luc > > >Firsoff, try iwithout compiling anything. Look at the version > >of iproute2 installed with Mandrake. > > > > tc -V > >will show version. Also, if /usr/lib/tc exists and contains q_netem.so > >and a bunch of .disfiles, you almoscertainly have a good enough > >version. > > > >To giva simpltest, just do (as root): > > tc qdisc add dev eth0 roonetem > >If isucceeds then you havjust put netem (with no options) > >ias thqdisc for the normal default ether device. Netem with no > >options is jusa simplFIFO so it is mostly harmless > >(1ms delay oearlier kernels pr2.6.12). > > > >Building iproute2 is relatively painless. Download latesfrom > > http://developer.osdl.org/dev/iproute2/download > >theunpack with > > tar xzf iproute2-2.6.11-050330.targ.gz > >to build > > cd iproute2-2.6.11-050330 > > make > >to install (as root) > > makinstall > >You do need to bcareful abouwhere the 'tc' command gets put. > >Somdistributions puit in /sbin some in /usr/sbin. The make install > >will puiin /usr/sbin, so check and delete old versions. > > > Thank for thaswer ! > > my versiois ss060832 bui don't know how delete it (i am beginner with > linux too, mandrake10.1) Looks lika vendor version, noin any of my iproute2 archives. To removold versions and do thinstall do the following: wgehttp://developer.osdl.org/dev/iproute2/iproute2-2.6.11-050330.tar.gz tar xzf iproute2-2.6.11-050330.tar.gz cd iproute2-2.6.11-050330 make # becomroot: sudo -s # or su - r`which tc` # removold tc wherever it is makinstall > i havanother question: jusiproute2 is sufficient for build an wan > simulator ? > Therartwo parts the kernel (which is in 2.4 and 2.6) and thiproute2 command, you need both. Sewebsite: http://developer.osdl.org/shemminger/netem and paper http://developer.osdl.org/shemminger/netem/LCA2005_paper.pdf and talk http://developer.osdl.org/shemminger/netem/LCA2005_netem.pdf Frotzenes agmail.com Mon Jun 6 13:30:06 2005 From: tzenes agmail.co(Sam Rossoff) Date: Wed Apr 18 17:37:47 2007 Subject: Problems with Netem: RTNETLINK answers: Invalid argument Message-ID: <467fbc4e050606133029cd5d38@xxxxxxxxxxxxxx> I recently reformatted onof my computers and installed slackware 10.0 oito act as a router. Now I wanted to use Netem as network emulatiosoftwarto set artifical drop rates and delays to test a veriaty of programs. Now, whei upgraded to thlatest linux kernel (2.6.11) for somreason ididn't come with netem (or i could find it whecompiling my kernel). So I did thlogical thing and downloaded thlatescopy of iproute2 (2.6.11-050330) which comes with netem. Now wheI run: #tc qdisc add dev eth0 roonetehelp i gea nichelp message: Usage: ... nete[ limiPACKETS ] [ delay TIME [ JITTER [CORRELATION]]] ... etc however if i try to usnetem #tc qdisc add dev eth0 roonete i gethfollowing error: RTNETLINK answers: Invalid argument I'vtried adding options: #tc qdisc add dev eth0 roonetedelay 100ms #tc qdisc add dev eth0 roohandl1: netem delay 100ms ... etc to no avail whei misspell an option, i get: #tc qdisc add dev eth0 roonetedealy 100ms Whais "dealy"? followed by thhelp message, so i know neteis installed and working to comextent now tc qdisc works finwith p_fifo and other options ionly seems to bfailing with netem evewhen i try to makit the child of another qdisc it continues to givmthe same error: RTNETLINK answers: Invalid argument I evechecked my config filto see if i enabled CONFIG_IP_MULTIPLE_TABLES and iis indeed =y now this probaleis fairly vexing, so i was wondering if you could help mout? This is pretty much a brand new kernel and filsystem, and thlatest versioof IProute2 so I'not sure what the problem could be Frotzenes agmail.com Mon Jun 6 13:58:43 2005 From: tzenes agmail.co(Sam Rossoff) Date: Wed Apr 18 17:37:47 2007 Subject: Problems with Netem: RTNETLINK answers: Invalid argument In-Reply-To: <467fbc4e050606133029cd5d38@xxxxxxxxxxxxxx> References: <467fbc4e050606133029cd5d38@xxxxxxxxxxxxxx> Message-ID: <467fbc4e050606135813f9045@xxxxxxxxxxxxxx> I recently reformatted onof my computers and installed slackware 10.0 oito act as a router. Now I wanted to use Netem as network emulatiosoftwarto set artifical drop rates and delays to test a veriaty of programs. Now, whei upgraded to thlatest linux kernel (2.6.11) for somreason ididn't come with netem (or i could find it whecompiling my kernel). So I did thlogical thing and downloaded thlatescopy of iproute2 (2.6.11-050330) which comes with netem. Now wheI run: #tc qdisc add dev eth0 roonetehelp i gea nichelp message: Usage: ... nete[ limiPACKETS ] [ delay TIME [ JITTER [CORRELATION]]] ... etc however if i try to usnetem #tc qdisc add dev eth0 roonete i gethfollowing error: RTNETLINK answers: Invalid argument I'vtried adding options: #tc qdisc add dev eth0 roonetedelay 100ms #tc qdisc add dev eth0 roohandl1: netem delay 100ms ... etc to no avail whei misspell an option, i get: #tc qdisc add dev eth0 roonetedealy 100ms Whais "dealy"? followed by thhelp message, so i know neteis installed and working to comextent now tc qdisc works finwith p_fifo and other options ionly seems to bfailing with netem evewhen i try to makit the child of another qdisc it continues to givmthe same error: RTNETLINK answers: Invalid argument I evechecked my config filto see if i enabled CONFIG_IP_MULTIPLE_TABLES and iis indeed =y now this probaleis fairly vexing, so i was wondering if you could help me out? This is pretty much a brand new kernel and filsystem, and thlatest versioof IProute2 so I'not sure what the problem could be ps. I appologizif this duplicates, i'not sure how this list is supposed to work Frostephen.earl abt.com Mon Jun 13 06:18:11 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: Questioon new reordering rule Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3156D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Ithprevious version of netem that just had the simple gap rule, the reordering worked like the following: Exery X packets are delayed by the amount specified in the delay part of the command, and the rest are not delayed. Now with thnew reorder command iworks like this: All packets are delayed by the amount specified in the delay part of the command apart from the packets randomly selected by the reorder part of the command and with a gap of X packets. So beforI would, say, sethe gap to 2 and the delay to 10ms and the packets would have a delay like: '0ms 0ms 10ms 0ms 0ms 10ms 0ms 0ms 10ms' etc. Now if I sedelay to 10ms, reorder to 100% and gap to 2 igoes like this: '10ms 10ms 0ms 10ms 10ms 0ms 10ms 10ms 0ms' etc. Why is this? And is ireally how a real lifnetwork would or could behave? Froshemminger aosdl.org Thu Jun 16 13:50:19 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: Questioon new reordering rule In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3156D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3156D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050616135019.3d6c32f0@xxxxxxxxxxxxxxxxx> OMon, 13 Jun 2005 14:18:11 +0100 <stephen.earl@xxxxxx> wrote: > > Ithprevious version of netem that just had the simple gap rule, the reordering worked like the following: Exery X packets are delayed by the amount specified in the delay part of the command, and the rest are not delayed. > > Now with thnew reorder command iworks like this: All packets are delayed by the amount specified in the delay part of the command apart from the packets randomly selected by the reorder part of the command and with a gap of X packets. > > So beforI would, say, sethe gap to 2 and the delay to 10ms and the packets would have a delay like: '0ms 0ms 10ms 0ms 0ms 10ms 0ms 0ms 10ms' etc. > > Now if I sedelay to 10ms, reorder to 100% and gap to 2 igoes like this: '10ms 10ms 0ms 10ms 10ms 0ms 10ms 10ms 0ms' etc. > > Why is this? And is ireally how a real lifnetwork would or could behave? > They why is simple, becausiseemed like a natural way to map the two parameters of reordering (probability and distance), into existing api with minimal upset. 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. Froshemminger aosdl.org Thu Jun 16 13:54:13 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: Jitter problems with Netem?? In-Reply-To: <429F53B1.7000709@xxxxxxxxxxx> References: <429E71DA.30001@xxxxxxxxxxx> <20050602112638.3225843a@xxxxxxxxxxxxxxxxx> <429F53B1.7000709@xxxxxxxxxxx> Message-ID: <20050616135413.4eafe87b@xxxxxxxxxxxxxxxxx> OThu, 02 Jun 2005 11:45:05 -0700 Joel Peshki<bugreport@xxxxxxxxxxx> wrote: > StepheHemminger wrote: > > > <snip> > > > >First, which kernel version. Thermay bbugs that already have been fixed. > >Thexpected behaviour would is to preservorder always, but > > > >Pack Arrive Delay Depart > >A 0 5 5 > >B 10 10 20 > >C 20 20 40 > >D 30 5 40 > >E 40 10 50 > >F 50 20 70 > > > >Whenetegets to looking at packet C, it's time has already expired so > >ican go righaway. I did have a patch that caused C to go out ahead of B > >buuser's complained thathey did not expect time based reordering and > >thabehaviour was onof the things they hated about Nist net. > > > > > Warusing 2.6.11 and will happily move to the latest or use a patch. > > Thosusers musnot have been involved in testing jitter buffer > adaptation. Thtime-based reordering is critically important. Is i > possiblthayou could make the patch available or, better, make it a > selectablbehavior? > After 2.6.12, I will revisithis. My "besidea so far" is to: * gerid of inner qdisc completely; thmain use now is to pusomething likTBF inside netem, but putting netem inside TBF works as well. * jususa sk_buff queue directly * fix reorder to pupackets back rather than at head of queue * allow timbased reordering as an option (or modulparameter, not sure). * maybadd somform RED support. Froji.li3 ahp.com Thu Jun 23 08:14:43 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: CaNetEbe bound to one interface without affecting the other? Message-ID: <628BFCE8B64706469FE4D4852CEC953706B090CC@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Dear all, I havtwo machines, each with two network cards. I wanto connect two eth0 and two eth1 respectively so thaI havtwo connections between thtwo machines. I havdirectly connected two eth0, but I want to connectwo eth1 in a way thaI can control the properties of this connection. Iseems thaNetEm is a good choice. The question is: can NetEbbound to one interface (eth1 in this case)? Otherwise, I will need a third computer as a gateway or bridgbetween two eth1 and run NetEon it. Is icorrect? Thanks a lot! -Ji Froshemminger aosdl.org Thu Jun 23 08:44:24 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: CaNetEbe bound to one interface without affecting the other? In-Reply-To: <628BFCE8B64706469FE4D4852CEC953706B090CC@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <628BFCE8B64706469FE4D4852CEC953706B090CC@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20050623084424.7925993b@xxxxxxxxxxxxxxxxx> OThu, 23 Jun 2005 11:14:43 -0400 "Li, Ji" <ji.li3@xxxxxx> wrote: > Dear all, > > I havtwo machines, each with two network cards. I wanto connect two > eth0 and two eth1 respectively so thaI havtwo connections between > thtwo machines. I havdirectly connected two eth0, but I want to > connectwo eth1 in a way thaI can control the properties of this > connection. Iseems thaNetEm is a good choice. The question is: can > NetEbbound to one interface (eth1 in this case)? Otherwise, I will > need a third computer as a gateway or bridgbetween two eth1 and run > NetEon it. Is icorrect? > Thanks a lot! > > -Ji Yes neteis always bound to an interface. Froneteat smyles.plus.com Thu Jun 23 13:29:30 2005 From: neteasmyles.plus.com (Barry Myles) Date: Wed Apr 18 17:37:47 2007 Subject: phpnetemgui 0.9 released Message-ID: <42BB1BAA.4040008@xxxxxxxxxxxxxxx> Jusreleased version 0.9 of phpnetemgui. This is a seof php web pages thaallow you to control thsetup of netem in a reasonable easy way. New ithis version: - ability to add multiplIP matching rules to an interface. - reorder supporon systems wherit is available. http://www.smyles.plus.com/phpnetemgui/ -- Barry Myles Froji.li3 ahp.com Thu Jun 23 14:43:43 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: Help: can'add netem Message-ID: <628BFCE8B64706469FE4D4852CEC953706B09239@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Dear all, I jusstarted to usNetEm today. When I tried to add netem, I got the following error message: # tc qdisc add dev eth1 roonetedelay 100ms RTNETLINK answers: Filexists Whadoes therror message mean? I suspect it's related to the teql I added before. WhaI did beforthat is to add teql, since I need both teql and netefor my work. Thcommands are: # 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 10.0.0.10/30 # ip addr add dev teql0 10.0.0.14/30 I was wondering if TEQL and NetEconflicwith each other or not. Can eth1 only badded to both teql and netem? Thanks, -Ji Frostephen.earl abt.com Fri Jun 24 04:01:30 2005 From: stephen.earl abt.co(stephen.earl@xxxxxx) Date: Wed Apr 18 17:37:47 2007 Subject: Help: can'add netem Message-ID: <B1A7C3B14EAAED48B27E84CBE6FFB37307AF94EA@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Yes, you gethis error becausyou already did the 'tc qdisc add dev eth1 root teql0' command. To use the teql0 and netem qdisc together you can do the following commands instead: tc qdisc add dev eth0 roohandl1: teql0 tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms You caonly add onqdisc at the root level in an interface. -----Original Message----- From: netem-bounces@xxxxxxxxxxxxxx [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji Sent: 23 Jun2005 22:44 To: netem@xxxxxxxxxxxxxx Subject: Help: can'add netem Dear all, I jusstarted to usNetEm today. When I tried to add netem, I got the following error message: # tc qdisc add dev eth1 roonetedelay 100ms RTNETLINK answers: Filexists Whadoes therror message mean? I suspect it's related to the teql I added before. WhaI did beforthat is to add teql, since I need both teql and netefor my work. Thcommands are: # 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 10.0.0.10/30 # ip addr add dev teql0 10.0.0.14/30 I was wondering if TEQL and NetEconflicwith each other or not. Can eth1 only badded to both teql and netem? Thanks, -Ji Froji.li3 ahp.com Fri Jun 24 07:27:07 2005 From: ji.li3 ahp.co(Ji Li) Date: Wed Apr 18 17:37:47 2007 Subject: Help: can'add netem Message-ID: <42071.161.114.64.74.1119623227.squirrel@xxxxxxxxxxxxxxx> Thanks you all for thkind help to a newbie! I believI avery close to what I want, but I still got a little problem, hopyou guys would hava look. Therror messagI got is (I need packets to go to teql0 first, and then bforwarded to netem): [root@princeton2 mtcp]# tc qdisc add dev eth0 roohandl1: teql0 [root@princeton2 mtcp]# tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms RTNETLINK answers: Invalid argument I searched for a while, and guess thait's becausI haven't loaded all relevenmodules. Thfollowing is copied from /lib/modules/2.6.9-1.667smp/build/.config . Would you pleasleme know whaI need to do to fix this problem? # # QoS and/or fair queueing # CONFIG_NET_SCHED=y CONFIG_NET_SCH_CLK_JIFFIES=y # CONFIG_NET_SCH_CLK_GETTIMEOFDAY is noset # CONFIG_NET_SCH_CLK_CPU is noset CONFIG_NET_SCH_CBQ=m CONFIG_NET_SCH_HTB=m CONFIG_NET_SCH_HFSC=m CONFIG_NET_SCH_ATM=m CONFIG_NET_SCH_PRIO=m CONFIG_NET_SCH_RED=m CONFIG_NET_SCH_SFQ=m CONFIG_NET_SCH_TEQL=m CONFIG_NET_SCH_TBF=m CONFIG_NET_SCH_GRED=m CONFIG_NET_SCH_DSMARK=m CONFIG_NET_SCH_NETEM=m CONFIG_NET_SCH_INGRESS=m CONFIG_NET_QOS=y CONFIG_NET_ESTIMATOR=y CONFIG_NET_CLS=y CONFIG_NET_CLS_TCINDEX=m CONFIG_NET_CLS_ROUTE4=m CONFIG_NET_CLS_ROUTE=y CONFIG_NET_CLS_FW=m CONFIG_NET_CLS_U32=m CONFIG_CLS_U32_PERF=y CONFIG_NET_CLS_IND=y CONFIG_NET_CLS_RSVP=m CONFIG_NET_CLS_RSVP6=m # CONFIG_NET_CLS_ACT is noset CONFIG_NET_CLS_POLICE=y Best, -Ji > Yes, you gethis error becausyou already did the 'tc qdisc add dev eth1 > rooteql0' command. To usthe teql0 and netem qdisc together you can do > thfollowing commands instead: > > tc qdisc add dev eth0 roohandl1: teql0 > tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms > > You caonly add onqdisc at the root level in an interface. > > -----Original Message----- > From: netem-bounces@xxxxxxxxxxxxxx > [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji > Sent: 23 Jun2005 22:44 > To: netem@xxxxxxxxxxxxxx > Subject: Help: can'add netem > > > Dear all, > > I jusstarted to usNetEm today. When I tried to add netem, I got the > following error message: > > # tc qdisc add dev eth1 roonetedelay 100ms > RTNETLINK answers: Filexists > > Whadoes therror message mean? I suspect it's related to the teql I > added before. WhaI did beforthat is to add teql, since I need both > teql and netefor my work. Thcommands are: > # 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 10.0.0.10/30 > # ip addr add dev teql0 10.0.0.14/30 > > I was wondering if TEQL and NetEconflicwith each other or not. Can > eth1 only badded to both teql and netem? > > > Thanks, > -Ji > > > > _______________________________________________ > Netemailing list > Netem@xxxxxxxxxxxxxx > http://lists.osdl.org/mailman/listinfo/netem > Froji.li3 ahp.com Fri Jun 24 09:34:43 2005 From: ji.li3 ahp.co(Ji Li) Date: Wed Apr 18 17:37:47 2007 Subject: Help: can'add netem In-Reply-To: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3157B@i2km34-ukdy.domain1.systemho st.net> References: <B1A7C3B14EAAED48B27E84CBE6FFB37307A3157B@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <55139.161.114.64.75.1119630883.squirrel@xxxxxxxxxxxxxxx> Yes, thescommands work! Thank you! However (I hatthis word!), after I add both eth0 and eth1 to teql0, I ping thteql0 on another machine, all ping traffic goes to eth1, none goes to eth0. Othother machine, I didn't add netem but only added eth0 and eth1 to teql0, and ping thteql0 on thfirst machine, and it works as expected: ping traffic goes to both interfaces. Attached arthe commands I used to configurteql0 on both machines, routing tables and "tc qdisc ls" results. Whacould bthe problem? Maybe the command to add eth1 to teql0 is wrong whertherare hierarchy? SORRY to bother you guys, and your help is highly appreciated! ============== Machin1 with NetE================ commands to install NetEand TEQL together: modprobsch_teql tc qdisc add dev eth0 roohandl1: netem delay 100ms 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 10.0.0.0/31 ip addr add dev eth1 10.0.0.2/31 ip addr add dev teql0 10.0.0.4/31 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 UsIface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 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 10.0.0.1/31 ip addr add dev eth1 10.0.0.3/31 ip addr add dev teql0 10.0.0.5/31 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 UsIface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 U 0 0 0 eth1 > I'vjusfound out that those two commands I gave you don't work for me > either. You havto puthe netem qdisc first: > > tc qdisc add dev eth0 roohandl1: netem delay 100ms > tc qdisc add dev eth0 paren1:1 handl10: teql0 > > Thashould work. > > -----Original Message----- > From: Ji Li [mailto:ji.li3@xxxxxx] > Sent: 24 Jun2005 15:27 > To: Earl,S,Stephen,XDP65A C; nistnet_user@xxxxxxxxx > Cc: netem@xxxxxxxxxxxxxx > Subject: RE: Help: can'add netem > > > Thanks you all for thkind help to a newbie! > > I believI avery close to what I want, but I still got a little > problem, hopyou guys would hava look. > > Therror messagI got is (I need packets to go to teql0 first, and then > bforwarded to netem): > [root@princeton2 mtcp]# tc qdisc add dev eth0 roohandl1: teql0 > [root@princeton2 mtcp]# tc qdisc add dev eth0 paren1:1 handl10: netem > delay 100ms > RTNETLINK answers: Invalid argument > > I searched for a while, and guess thait's becausI haven't loaded all > relevenmodules. Thfollowing is copied from > /lib/modules/2.6.9-1.667smp/build/.config . Would you pleasleme know > whaI need to do to fix this problem? > > # > # QoS and/or fair queueing > # > CONFIG_NET_SCHED=y > CONFIG_NET_SCH_CLK_JIFFIES=y > # CONFIG_NET_SCH_CLK_GETTIMEOFDAY is noset > # CONFIG_NET_SCH_CLK_CPU is noset > CONFIG_NET_SCH_CBQ=m > CONFIG_NET_SCH_HTB=m > CONFIG_NET_SCH_HFSC=m > CONFIG_NET_SCH_ATM=m > CONFIG_NET_SCH_PRIO=m > CONFIG_NET_SCH_RED=m > CONFIG_NET_SCH_SFQ=m > CONFIG_NET_SCH_TEQL=m > CONFIG_NET_SCH_TBF=m > CONFIG_NET_SCH_GRED=m > CONFIG_NET_SCH_DSMARK=m > CONFIG_NET_SCH_NETEM=m > CONFIG_NET_SCH_INGRESS=m > CONFIG_NET_QOS=y > CONFIG_NET_ESTIMATOR=y > CONFIG_NET_CLS=y > CONFIG_NET_CLS_TCINDEX=m > CONFIG_NET_CLS_ROUTE4=m > CONFIG_NET_CLS_ROUTE=y > CONFIG_NET_CLS_FW=m > CONFIG_NET_CLS_U32=m > CONFIG_CLS_U32_PERF=y > CONFIG_NET_CLS_IND=y > CONFIG_NET_CLS_RSVP=m > CONFIG_NET_CLS_RSVP6=m > # CONFIG_NET_CLS_ACT is noset > CONFIG_NET_CLS_POLICE=y > > Best, > -Ji > >> Yes, you gethis error becausyou already did the 'tc qdisc add dev >> eth1 >> rooteql0' command. To usthe teql0 and netem qdisc together you can >> do >> thfollowing commands instead: >> >> tc qdisc add dev eth0 roohandl1: teql0 >> tc qdisc add dev eth0 paren1:1 handl10: netem delay 100ms >> >> You caonly add onqdisc at the root level in an interface. >> >> -----Original Message----- >> From: netem-bounces@xxxxxxxxxxxxxx >> [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji >> Sent: 23 Jun2005 22:44 >> To: netem@xxxxxxxxxxxxxx >> Subject: Help: can'add netem >> >> >> Dear all, >> >> I jusstarted to usNetEm today. When I tried to add netem, I got the >> following error message: >> >> # tc qdisc add dev eth1 roonetedelay 100ms >> RTNETLINK answers: Filexists >> >> Whadoes therror message mean? I suspect it's related to the teql I >> added before. WhaI did beforthat is to add teql, since I need both >> teql and netefor my work. Thcommands are: >> # 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 10.0.0.10/30 >> # ip addr add dev teql0 10.0.0.14/30 >> >> I was wondering if TEQL and NetEconflicwith each other or not. Can >> eth1 only badded to both teql and netem? >> >> >> Thanks, >> -Ji >> >> >> >> _______________________________________________ >> Netemailing list >> Netem@xxxxxxxxxxxxxx >> http://lists.osdl.org/mailman/listinfo/netem >> > > > Froji.li3 ahp.com Fri Jun 24 10:05:13 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: Help: can'add netem Message-ID: <628BFCE8B64706469FE4D4852CEC953706B09400@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> I havsomother IP addresses assigned to the same machines, which enables mto access thInternet. Those configuration may affect the result. I acleaning up everything to tesagain. Thanks and sorry to spam! Besregards, -Ji -----Original Message----- From: Li, Ji Sent: Friday, Jun24, 2005 12:35 PM To: stephen.earl@xxxxxx Cc: netem@xxxxxxxxxxxxxx Subject: RE: Help: can'add netem Yes, thescommands work! Thank you! However (I hatthis word!), after I add both eth0 and eth1 to teql0, I ping thteql0 on another machine, all ping traffic goes to eth1, none goes to eth0. Othother machine, I didn't add netem but only added eth0 and eth1 to teql0, and ping thteql0 on thfirst machine, and it works as expected: ping traffic goes to both interfaces. Attached are thcommands I used to configurteql0 on both machines, routing tables and "tc qdisc ls" results. Whacould bthe problem? Maybe the command to add eth1 to teql0 is wrong whertherare hierarchy? SORRY to bother you guys, and your help is highly appreciated! ============== Machin1 with NetE================ commands to install NetEand TEQL together: modprobsch_teql tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add dev eth0 paren1:1 handl10: teql0 tc qdisc add dev eth1 root teql0 ip link sedev teql0 up ip addr add dev eth0 10.0.0.0/31 ip addr add dev eth1 10.0.0.2/31 ip addr add dev teql0 10.0.0.4/31 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 Use Iface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 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 10.0.0.1/31 ip addr add dev eth1 10.0.0.3/31 ip addr add dev teql0 10.0.0.5/31 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 Use Iface 10.0.0.4 0.0.0.0 255.255.255.254 U 0 0 0 teql0 10.0.0.0 0.0.0.0 255.255.255.254 U 0 0 0 eth0 10.0.0.2 0.0.0.0 255.255.255.254 U 0 0 0 eth1 > I'vjusfound out that those two commands I gave you don't work for > meither. You havto put the netem qdisc first: > > tc qdisc add dev eth0 roohandl1: netem delay 100ms tc qdisc add > dev eth0 paren1:1 handl10: teql0 > > Thashould work. > > -----Original Message----- > From: Ji Li [mailto:ji.li3@xxxxxx] > Sent: 24 Jun2005 15:27 > To: Earl,S,Stephen,XDP65A C; nistnet_user@xxxxxxxxx > Cc: netem@xxxxxxxxxxxxxx > Subject: RE: Help: can'add netem > > > Thanks you all for thkind help to a newbie! > > I believI avery close to what I want, but I still got a little > problem, hopyou guys would hava look. > > Therror messagI got is (I need packets to go to teql0 first, and > thebforwarded to netem): > [root@princeton2 mtcp]# tc qdisc add dev eth0 roohandl1: teql0 > [root@princeton2 mtcp]# tc qdisc add dev eth0 paren1:1 handl10: > netedelay 100ms RTNETLINK answers: Invalid argument > > I searched for a while, and guess thait's becausI haven't loaded > all relevenmodules. Thfollowing is copied from > /lib/modules/2.6.9-1.667smp/build/.config . Would you pleasleme > know whaI need to do to fix this problem? > > # > # QoS and/or fair queueing > # > CONFIG_NET_SCHED=y > CONFIG_NET_SCH_CLK_JIFFIES=y > # CONFIG_NET_SCH_CLK_GETTIMEOFDAY is nose# CONFIG_NET_SCH_CLK_CPU > is noseCONFIG_NET_SCH_CBQ=m CONFIG_NET_SCH_HTB=m > CONFIG_NET_SCH_HFSC=CONFIG_NET_SCH_ATM=CONFIG_NET_SCH_PRIO=m > CONFIG_NET_SCH_RED=CONFIG_NET_SCH_SFQ=CONFIG_NET_SCH_TEQL=m > CONFIG_NET_SCH_TBF=CONFIG_NET_SCH_GRED=CONFIG_NET_SCH_DSMARK=m > CONFIG_NET_SCH_NETEM=CONFIG_NET_SCH_INGRESS=CONFIG_NET_QOS=y > CONFIG_NET_ESTIMATOR=y CONFIG_NET_CLS=y CONFIG_NET_CLS_TCINDEX= > CONFIG_NET_CLS_ROUTE4=CONFIG_NET_CLS_ROUTE=y CONFIG_NET_CLS_FW= > CONFIG_NET_CLS_U32=CONFIG_CLS_U32_PERF=y CONFIG_NET_CLS_IND=y > CONFIG_NET_CLS_RSVP=CONFIG_NET_CLS_RSVP6=# CONFIG_NET_CLS_ACT is > noseCONFIG_NET_CLS_POLICE=y > > Best, > -Ji > >> Yes, you gethis error becausyou already did the 'tc qdisc add dev >> eth1 >> rooteql0' command. To usthe teql0 and netem qdisc together you >> cado thfollowing commands instead: >> >> tc qdisc add dev eth0 roohandl1: teql0 tc qdisc add dev eth0 >> paren1:1 handl10: netem delay 100ms >> >> You caonly add onqdisc at the root level in an interface. >> >> -----Original Message----- >> From: netem-bounces@xxxxxxxxxxxxxx >> [mailto:netem-bounces@xxxxxxxxxxxxxx]OBehalf Of Li, Ji >> Sent: 23 Jun2005 22:44 >> To: netem@xxxxxxxxxxxxxx >> Subject: Help: can'add netem >> >> >> Dear all, >> >> I jusstarted to usNetEm today. When I tried to add netem, I got >> thfollowing error message: >> >> # tc qdisc add dev eth1 roonetedelay 100ms RTNETLINK answers: >> Filexists >> >> Whadoes therror message mean? I suspect it's related to the teql >> I added before. WhaI did beforthat is to add teql, since I need >> both teql and netefor my work. Thcommands are: >> # 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 10.0.0.10/30 >> # ip addr add dev teql0 10.0.0.14/30 >> >> I was wondering if TEQL and NetEconflicwith each other or not. >> Can >> eth1 only badded to both teql and netem? >> >> >> Thanks, >> -Ji >> >> >> >> _______________________________________________ >> Netemailing list >> Netem@xxxxxxxxxxxxxx >> http://lists.osdl.org/mailman/listinfo/netem >> > > > Froji.li3 ahp.com Fri Jun 24 11:56:26 2005 From: ji.li3 ahp.co(Li, Ji) Date: Wed Apr 18 17:37:47 2007 Subject: How to makneteand teql work together? Message-ID: <628BFCE8B64706469FE4D4852CEC953706B09490@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> (I asked thsamquestion in previous thread, but now the problem is a littldifferent, so I stara new thread. If I spam too much, please lemknow and I'll stop :P ) SORRY to bother you guys, and your help is highly appreciated! Shorstory: when I load neteand teql together, teql doesn't work correctly. (If I load teql only, everything is fine) I load both neteand teql. Neteis associated with eth0, and teql is associated with both eth0 and eth1. Buafter that, traffic only goes ouof eth1. I also tried loading teql only (with netem), then teql works as expected: ping traffic goes to both interfaces. Attached are thcommands thaI used to configure teql and netem (machine 1), and teql only (machie2), their routing tables and "tc qdisc ls" results. What's wrong? ============== 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 Use 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 Use 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 Froamanol aiit.demokritos.gr Wed Jun 29 03:17:25 2005 From: amanol aiit.demokritos.gr (Apostolos Manolitzas) Date: Wed Apr 18 17:37:47 2007 Subject: Interfacfreezes with delay Message-ID: <42C27535.1000105@xxxxxxxxxxxxxxxxx> Hello all, I try to usneteto add delay to my network. I havthfollowing configuration: Sus9.3 Linux cor2.6.11.4-20a-defaul#1 Mon Jun 27 17:19:03 EEST 2005 i686 i686 i386 GNU/Linux Na : iproute2 Relocations: (not relocatable) Versio : 2.6.10 Vendor: SUSE LINUX Products GmbH, Nuernberg, Germany Releas : 4 Build Date: Tu22 Mar 2005 08:49:29 PM EET and I usthfollowing command: tc qdisc add dev eth1 roonetedelay 240ms Thping works finand I get the proper delay. Buafter pushing traffic of 3Mbit/s for 4-5 minutes thinterface freezes, and I havto restarit. I can'figurout what is the reason of it. I havtested with kernel 2.6.8.x and I hadthe same results. Does anyonhavany clue about it? regards, Apostolos Froshemminger aosdl.org Wed Jun 29 14:09:31 2005 From: shemminger aosdl.org (Stephen Hemminger) Date: Wed Apr 18 17:37:47 2007 Subject: Interfacfreezes with delay In-Reply-To: <42C27535.1000105@xxxxxxxxxxxxxxxxx> References: <42C27535.1000105@xxxxxxxxxxxxxxxxx> Message-ID: <20050629140931.3a129086@xxxxxxxxxxxxxxxxx> OWed, 29 Jun 2005 13:17:25 +0300 Apostolos Manolitzas <amanol@xxxxxxxxxxxxxxxxx> wrote: > Hello all, > > I try to usneteto add delay to my network. > I havthfollowing configuration: > Sus9.3 > Linux cor2.6.11.4-20a-defaul#1 Mon Jun 27 17:19:03 EEST 2005 i686 > i686 i386 GNU/Linux > > Na : iproute2 Relocations: (not relocatable) > Versio : 2.6.10 Vendor: SUSE LINUX > Products GmbH, Nuernberg, Germany > Releas : 4 Build Date: Tu22 Mar 2005 > 08:49:29 PM EET > > and I usthfollowing command: > tc qdisc add dev eth1 roonetedelay 240ms > > Thping works finand I get the proper delay. > Buafter pushing traffic of 3Mbit/s for 4-5 minutes thinterface freezes, > and I havto restarit. > I can'figurout what is the reason of it. > I havtested with kernel 2.6.8.x and I hadthe same results. > > Does anyonhavany clue about it? > I assumyou arseeing the network interface stop. Could you look astats with ifconfig and seit is still receiving? Whais thnetwork interface? I have seen problems with older versions of the1000 driver when thtransmit ring gets overlaoded. Frojuliokriger agmail.com Thu Jun 30 06:21:34 2005 From: juliokriger agmail.co(Julio Kriger) Date: Wed Apr 18 17:37:47 2007 Subject: neteand sctplib on user space Message-ID: <682bc30a05063006217c2d112e@xxxxxxxxxxxxxx> Hi! Yesterday I testes a user spacSCTP library with netem. I sea delay of 1 second o"lo". Then, started two prograthat send packages to onanother. Igavme the impresion that netem didn't delay the packages. Does neteworking with user spaclibraries? Regards, Julio -- ---------------------------- Julio Kriger mailto:juliokriger@xxxxxxxxx Froamanol aiit.demokritos.gr Thu Jun 30 06:38:52 2005 From: amanol aiit.demokritos.gr (Apostolos Manolitzas) Date: Wed Apr 18 17:37:47 2007 Subject: Interfacfreezes with delay In-Reply-To: <20050629140931.3a129086@xxxxxxxxxxxxxxxxx> References: <42C27535.1000105@xxxxxxxxxxxxxxxxx> <20050629140931.3a129086@xxxxxxxxxxxxxxxxx> Message-ID: <42C3F5EC.3080300@xxxxxxxxxxxxxxxxx> StepheHemminger wrote: >OWed, 29 Jun 2005 13:17:25 +0300 >Apostolos Manolitzas <amanol@xxxxxxxxxxxxxxxxx> wrote: > > >>Hello all, >> >>I try to usneteto add delay to my network. >>I havthfollowing configuration: >>Sus9.3 >>Linux cor2.6.11.4-20a-defaul#1 Mon Jun 27 17:19:03 EEST 2005 i686 >>i686 i386 GNU/Linux >> >>Na : iproute2 Relocations: (not relocatable) >>Versio : 2.6.10 Vendor: SUSE LINUX >>Products GmbH, Nuernberg, Germany >>Releas : 4 Build Date: Tu22 Mar 2005 >>08:49:29 PM EET >> >>and I usthfollowing command: >>tc qdisc add dev eth1 roonetedelay 240ms >> >>Thping works finand I get the proper delay. >>Buafter pushing traffic of 3Mbit/s for 4-5 minutes thinterface freezes, >>and I havto restarit. >>I can'figurout what is the reason of it. >>I havtested with kernel 2.6.8.x and I hadthe same results. >> >>Does anyonhavany clue about it? >> >> > >I assumyou arseeing the network interface stop. >Could you look astats with ifconfig and seit is still receiving? > >Whais thnetwork interface? I have seen problems with older versions >of the1000 driver when thtransmit ring gets overlaoded. > > Whewmean freezes, the interfaces can't be accessed with ping or anything else. I havto ifdown eth0 and then ifup eth0 to starworking again. If during "freezing" state, I removthdelay the I/F is up and running again. I checked thstats, and idoesn't receive anything. FurthermorI changed thdelay to 120ms but the interface freezes again. Also thfreezin timis not proportional to the delay. Is therany way to identify why thinterface is self-blocked?
