**** BEGIN LOGGING AT Wed Jan 10 03:00:00 2007 Jan 10 11:58:11 when running 'python setup.py install' I get error: invalid command 'install'... any clues? Jan 10 12:32:44 hi there Jan 10 15:51:18 Hi, gyus. Did anybody observe the problemm tcpdump dumping only incomming packets on the ethernet interfaces? Jan 10 16:15:55 yup Jan 10 16:15:59 on uNLSUng Jan 10 16:17:43 And not on any other firmware? Do the other firmwares run in little-endian mode? Jan 10 16:28:03 xl0: switch or hub on your LAN? Jan 10 16:31:45 eno: The "device" is connected directly to my pc with a cross cable. It is actully not an NSLU, but an ixp420 based board with my custom kernel and userspace. I'm now tracking the problem through the kernel. Jan 10 19:19:41 tcpdump works ok on slugos/BE Jan 10 19:19:55 it doesnt work on uNSLUng 5.5 Jan 10 19:20:15 outgoing packets are not captured Jan 10 19:21:05 I have two slugs (openslug/BE and uNSLUng) on the same switch Jan 10 19:21:08 one is ok Jan 10 19:21:13 the other one is not Jan 10 19:22:58 on the uNSLUng: Jan 10 19:23:09 [root@LKG7CAC92 ~]$ tcpdump -V Jan 10 19:23:09 tcpdump version 3.9.5 Jan 10 19:23:09 libpcap version 0.9.5 Jan 10 19:24:15 on slugos/BE - openSLUG 3.10: Jan 10 19:24:18 root@LKG7BE78E:~# tcpdump -h Jan 10 19:24:19 tcpdump version 3.9.2 Jan 10 19:24:19 libpcap version 0.9.2 Jan 10 19:25:45 it would be interesting to see how optware/slugosbe tcpdump works Jan 10 19:28:23 how do I add the optware feeds to the openSLUG Jan 10 19:28:37 or how do I force a package to be installed from the optware Jan 10 19:28:54 I also don't want to mess with my existing config Jan 10 19:29:22 it won't Jan 10 19:29:27 start with http://ipkg.nslu2-linux.org/feeds/optware/slugosbe/cross/unstable/ipkg-opt_0.99.163-9_armeb.ipk Jan 10 19:29:34 extract it Jan 10 19:29:53 then s/stable/unstable/ in /opt/etc/ipkg.conf Jan 10 19:29:57 right :D Jan 10 19:30:15 ipkg update; ipkg install tcpdump Jan 10 19:30:47 actually /opt/bin/ipkg update; /opt/bin/ipkg install tcpdump Jan 10 19:31:32 if you prefer, export PATH=/opt/bin:$PATH Jan 10 19:31:47 I prefer to put the full path Jan 10 19:31:56 like this I will know what I'm using :) Jan 10 19:32:38 i set one bash shell with PATH=/opt/bin:$PATH under screen Jan 10 19:32:53 yup Jan 10 19:32:59 that will do it Jan 10 19:33:35 set the window title as well Jan 10 19:33:45 changing the prompt will help too Jan 10 19:33:53 ahhh, I need to be root :p Jan 10 19:35:13 asterisk14 is in slugosbe feed too Jan 10 19:35:21 I know :) Jan 10 19:35:48 something is wrong :( Jan 10 19:35:56 http://ipkg.nslu2-linux.org/feeds/optware/slugosbe/cross/unstable/ipkg-opt_0.99.163-9_armeb.i Jan 10 19:36:03 sorry Jan 10 19:36:06 ~# ipkg -V 5 install ipkg-opt_0.99.163-9_armeb.ipk Jan 10 19:36:06 Debug install_cmd: ipkg-opt_0.99.163-9_armeb.ipk Jan 10 19:36:06 ipkg: ipkg-opt_0.99.163-9_armeb.ipk: No such file or directory Jan 10 19:36:06 pkg_info_preinstall_check: updating arch priority for each package Jan 10 19:36:06 pkg_info_preinstall_check: update file owner list Jan 10 19:36:13 Getting old from pkg_hash_fetch Jan 10 19:36:13 Getting new from pkg_hash_fetch Jan 10 19:36:13 Versions from pkg_hash_fetch in ipkg_install_by_name Jan 10 19:36:13 Configuring unpacked packages Jan 10 19:36:15 Nothing to be done Jan 10 19:36:17 An error ocurred, return value: 4. Jan 10 19:36:22 Collected errors: Jan 10 19:36:23 Cannot find package ipkg-opt_0.99.163-9_armeb.ipk. Jan 10 19:36:25 Check the spelling or perhaps run 'ipkg update' Jan 10 19:36:27 hash_table[pkg-hash] n_buckets=0 n_elements=4058 max_conflicts=0 n_conflicts=0 Jan 10 19:36:29 hash_table[file-hash] n_buckets=0 n_elements=13060 max_conflicts=0 n_conflicts=0 Jan 10 19:36:31 hash_table[obs-file-hash] n_buckets=0 n_elements=0 max_conflicts=0 n_conflicts=0 Jan 10 19:36:57 hmmm ... I'm runing openslug 3.10 Jan 10 19:36:58 i mean use tar xzvf to extract the content of ipkg-opt.*.ipk Jan 10 19:37:06 oh ... Jan 10 19:37:15 * osas needs a coffee Jan 10 19:37:26 openslug ipkg is in ar format Jan 10 19:37:32 optware ipkg is in tar format Jan 10 19:38:45 so .. I will extract data.tar.gz in / Jan 10 19:38:49 the bootstrap of optware/slugosbe is still manual to have kinda barrier of entry Jan 10 19:38:52 and that will create opt Jan 10 19:39:00 yes Jan 10 19:39:04 got it Jan 10 19:40:13 k, lunch break, bbl Jan 10 19:42:36 it looks like an issue with the version Jan 10 19:43:23 /opt/bin/tcpdump -n displays only outgoing packets Jan 10 19:43:47 /usr/sbin/tcpdump -n displays both incoming and outgoing packets Jan 10 19:44:00 could be optware configure problem as well Jan 10 19:44:21 yup Jan 10 19:44:43 anyway, the result is the same on both platforms Jan 10 19:45:29 monit is crashing on uNSLUng and it is working ok on openSLUG (slugos/BE) Jan 10 19:45:52 if we have openslug bb upgrade, then we know Jan 10 19:46:28 I will need to send the monit bb file to Nail Jan 10 19:46:59 I was running monit for quite some time on openSLUG w/o a problem Jan 10 19:47:25 but on uNSLUng is crashing on first or second timer Jan 10 19:48:09 Hmm, so the problem is not with the kernel? Jan 10 19:48:30 it doesn't look like Jan 10 19:49:30 It's also not with the tcpdump version, I'm currently running 3.9.2/0.9.2, and see only incoming packets, and I've also tryed 3.9.5/0.9.5. Jan 10 19:49:58 eno: the slugos/BE optware is compatible with openSLUG 3.10 (i.e. same toolchain)? Jan 10 19:55:46 Ah, I know what most likely happened. In libpap you can specify the capture method. Some methods support kernel-side packet filtering, and some just dump all the packets to the userspace, which may lead to a segnificant overhead. Maybe the openslug and unslug libpcaps were cmpiled with different capture methods? Could somebody please check this? The configure option is "--with-pcap=TYPE". Jan 10 19:58:21 To me it looks like this is a kernel-side packet filter bug. I wrote a simple libpcap sniffer which attaches to an interface, but sets no capture mask. The kernel is not sending any packets there, the application sleeps in recvfrom(). Jan 10 19:58:36 xl0: I will give it a try Jan 10 19:58:38 one sec Jan 10 19:59:00 the TYPE should be TYPE? Jan 10 20:01:29 It should be "linux" in one case and something different in the other. Jan 10 20:02:01 The "linux" type relyes on the kernel-side filter. Jan 10 20:05:38 building now Jan 10 20:08:11 Or, hmm, I may be wrong. There seems to be no other capture method working on linux... Jan 10 20:10:26 same thing Jan 10 20:10:31 i didn't help Jan 10 20:10:40 s/i/it/ Jan 10 20:10:40 osas meant: it didn't help Jan 10 20:14:44 gtk-window-decorator wouldn't start with compiz. how do i get window decoration with compiz? Jan 10 20:15:30 osas: Is there a thing like an apt repository I could use to get the sources used to build the openslug packags? Jan 10 20:15:50 oops. wrong channel :/ Jan 10 20:16:45 osas: Btw, package from which version worked for you? Jan 10 20:17:07 check the optware packages on the nslu web site Jan 10 20:17:14 the MasterMakeFile Jan 10 20:17:22 sorry, I need to run now Jan 10 20:18:31 xl0: http://www.nslu2-linux.org/wiki/Development/MasterMakefile Jan 10 21:08:36 osas: Are you sure about the versions? Openslug-3.10 did not have tcpdump/libpcap 3.9.2/0.9.2, and unslug-5.5 has no 3.9.2/0.9.2... Jan 10 21:09:05 I've checked both packages, they all are built with --with-pcap=linux ... Jan 10 22:28:40 osas: optware/slugosbe uses exactly the same toolchain as openslug Jan 10 22:29:01 I know that Jan 10 22:29:19 my question is: is it the same as 3.10? Jan 10 22:36:06 (23:54:56) (HoStyle) is there somebody that make a lamp on it? Jan 10 22:36:06 (23:57:16) (HoStyle) i would know only if it is slow Jan 10 22:36:06 (23:58:00) (HoStyle) i will use it for a home website only Jan 10 22:36:41 as i said... Jan 10 22:37:00 :) Jan 10 22:37:12 (00:00:31) (joshin_) HoStyle: It works fine for that. Just take the limited memory into account. Also, this sort of question belongs in #nslu2-general. Jan 10 22:37:12 (00:01:17) (caplink811-log) HoStyle, yes, lamp is "slow" on unslung-5.5, as also on opendebianslug is (but slow is relative) Jan 10 22:37:23 just try it, depends on yor personal feelings,,, Jan 10 22:37:28 ok updated ;) Jan 10 22:37:39 s/yor/your/ Jan 10 22:37:40 caplink811-log meant: just try it, depends on your personal feelings,,, Jan 10 22:38:33 one log, and one bot... mmhh.. the cyborg are talking o.O Jan 10 22:38:47 sure ;) Jan 10 22:38:58 k i follow the log Jan 10 22:39:10 me too, all it's to try Jan 10 22:39:16 and mod... Jan 10 22:40:06 but, assuming, that we live in the same tz, it's time for me to heading to bed now.. Jan 10 22:40:36 gn, john boy Jan 10 22:40:43 gn, tnx for all Jan 10 22:40:50 np ;) Jan 10 22:40:57 .o/ Jan 10 22:41:13 hey, wtf is tz? Jan 10 22:41:35 timezone, in my case it is cet +1 Jan 10 22:41:50 assuming, as your's is Jan 10 22:42:22 nice Jan 10 22:42:29 im tired too Jan 10 22:42:39 or in other words, in berlin, de it is 00:12 Jan 10 22:42:51 ita 12.11 Jan 10 22:43:07 we have less minutes XD Jan 10 22:43:10 a little time shift ;) Jan 10 22:43:28 gn8 Jan 10 22:43:37 s'ya Jan 10 22:47:32 osas: Could you please send me the strace output from the working tcpdump? I've found where the problem may come from. Most likely we don't see the packets because the ixp ethernet driver uses it's own queue discipline, and I see no way how a different tcpdump version could work. Jan 10 22:48:06 one sec Jan 10 22:49:54 osas: Btw, what's the source for the ixp drivers? Still intel.com? Jan 10 22:50:03 yup Jan 10 22:50:15 openSLUG (slugos/BE) 3.10 Jan 10 22:51:05 installing strace Jan 10 22:53:55 xl0: what do you want to see? Jan 10 22:54:44 osas: The full strace log, ideally from both tcpdump versions. Jan 10 22:54:54 one sec Jan 10 22:56:06 the good one: http://pastebin.ca/312497 Jan 10 22:56:08 HoStyle: BTW, I run my personal webserver on this very slug and it also serves as a nice irssi proxy too. Jan 10 22:56:39 Heck, I'm even running Gentoo on it (bootstrapped on OpenSlug) Jan 10 22:56:51 But I'm nuts. Jan 10 22:57:54 xl0: root@LKG7BE78E:~# tcpdump -h Jan 10 22:57:54 tcpdump version 3.9.2 Jan 10 22:57:54 libpcap version 0.9.2 Jan 10 22:59:27 . Jan 10 22:59:37 eth0 is the ixp ethernet interface? It's called ixp0 in the default driver. Jan 10 22:59:42 the bad one: http://pastebin.ca/312500 Jan 10 23:00:03 yup Jan 10 23:00:20 root@LKG7BE78E:~# /opt/bin/tcpdump -h Jan 10 23:00:20 tcpdump version 3.9.5 Jan 10 23:00:20 libpcap version 0.9.5 Jan 10 23:00:40 you have both posts now Jan 10 23:03:58 osas: does the bad one return from recvfrom on outgoing packets? Jan 10 23:04:29 there are no packets Jan 10 23:04:58 do you want an strace with packets? Jan 10 23:05:45 Would be cool, cause I still have no clue what's going on... Jan 10 23:07:43 one sec Jan 10 23:07:57 I will start dumping and I will send a ping Jan 10 23:08:09 like this, there will be only a few packets Jan 10 23:11:06 xl0: sorry for misleading you, both versions of tcpdump are working Jan 10 23:11:15 I think I need more coffee Jan 10 23:11:52 xl0: openslug version? Jan 10 23:12:12 osas: They are both working one one kernel version and not working on an other. Jan 10 23:12:35 yup Jan 10 23:12:45 openslug 3.10 Jan 10 23:12:54 the openslug is ok Jan 10 23:12:59 unslung is not Jan 10 23:13:06 unslung 5.5 Jan 10 23:14:13 good sleuthing guys Jan 10 23:14:45 :p Jan 10 23:14:48 This is because the unslug linux kernel was compiled with CONFIG_IXP400_ETH_QDISC_ENABLED, and openslug without this option, I suppose. The bug is in the driver abusing the linux packet scheduler. Jan 10 23:15:03 eno: I don't owl a slug. Jan 10 23:15:06 own Jan 10 23:15:28 Ok, let's make a patch now. Jan 10 23:15:58 mwester_: you probably want to know this ^^^^^^^^^^^^ Jan 10 23:17:41 the kernel on unslung is the one that comes from linksys. is that right? Jan 10 23:21:26 osas: i think so, with some bug fix and userland firmware customization i believe Jan 10 23:21:58 I see Jan 10 23:22:09 well ... I'm not a kernel hacker Jan 10 23:22:52 me either Jan 10 23:26:23 I will play with asterisk on openslug these days Jan 10 23:26:35 it is already instaled Jan 10 23:41:41 There is actualy no bug, the option description says: Jan 10 23:41:42 IMPORTANT: This option may NOT work with some debug tools or Jan 10 23:41:42 some kernel features. Jan 11 02:17:31 Has somedoy allready installed the berckley db v4? **** ENDING LOGGING AT Thu Jan 11 02:59:58 2007