**** BEGIN LOGGING AT Wed Feb 17 02:59:58 2010 Feb 17 03:02:26 monnier * r19685 /packages/utils/lvm2/ (Makefile patches/100-readline-link.patch): Fixup TABs; Update to 2.02.60 Feb 17 09:52:04 swalker_, the URL of mii-tool should be http://freshmeat.net/projects/mii-tool/ on http://pastehtml.com/view/5tfitwd.html Feb 17 09:54:55 and for nprobe: http://freshmeat.net/projects/nprobe/ Feb 17 10:36:46 build #11 of gemini is complete: Failure [failed compile_10] Build details are at http://tksite.gotdns.org:8010/builders/gemini/builds/11 Feb 17 10:44:05 wknock is really broken Feb 17 11:07:54 kaloz * r19686 /trunk/toolchain/gcc/ (Config.in Config.version): [toolchain/mips]: use gcc 4.4.3 for mips64 targets Feb 17 11:11:55 kaloz * r19687 /trunk/target/linux/generic-2.6/patches-2.6.32/ (4 files): refresh generic 2.6.32 patches Feb 17 11:32:10 . Feb 17 11:44:26 build #11 of pxcab is complete: Failure [failed compile_3] Build details are at http://tksite.gotdns.org:8010/builders/pxcab/builds/11 Feb 17 12:01:45 hi kernel hackers, i'm trying to port openwrt to a custom arm-based embedded system, but am getting kernel panic on every boot caused by "undefined instruction" and I have no idea how to approach the problem Feb 17 12:02:30 s work and calls init Feb 17 12:03:00 the kernel panic hppens right after the kernel finishes it's work and init takes over Feb 17 12:05:07 probably the kernel tries to execute a CPU instruction which is not implemented on the CPU it runs Feb 17 12:05:30 or maybe it's init Feb 17 12:05:56 may it have something to do with EABI / OABI? Feb 17 12:06:23 i have no idea, in fact i'm not a kernel hacker, i'm just an occasional chatter here Feb 17 12:06:35 support for both is enabled in the kernel Feb 17 12:06:59 but you have built the whole image with kernel in it or you load the kernel via TFTP? Feb 17 12:08:17 qecko: Could you paste the panic message somewhere? Feb 17 12:08:20 the kernel is included in the rootfs (/boot) Feb 17 12:09:20 x10: sure, http://pastebin.com/d693f918f Feb 17 12:10:52 qecko: Looks like the userspace is built for an incompatible arm version, not the kernel. Feb 17 12:11:05 The kernel panics because the init process dies. Feb 17 12:13:10 x10: the only option which affects how the userspace binaries are build is ARCH:=arm in target/linux/.../Makefile isn't it? Feb 17 12:15:21 qecko: The kernel is built correctly, the userspace is not. Feb 17 12:16:39 qecko: There is an option in .config to set the default -march for the userspace. Feb 17 12:16:57 x10: i just did a 'file busybox': busybox: ELF 32-bit LSB executable, ARM, version 1 (SYSV), dynamically linked (uses shared libs), corrupted section header size Feb 17 12:17:03 doesnt look good actually Feb 17 12:17:04 qecko: also take a look at include/target.mk, default flags for arm are defined there Feb 17 12:19:52 qecko: There is a number of arm cpus, and not all of them are compatible. See grep CONFIG_DEFAULT_TARGET_OPTIMIZATION .config Feb 17 12:23:07 CONFIG_DEFAULT_TARGET_OPTIMIZATION="-Os -pipe -march=armv5te -mtune=xscale -funit-at-a-time" i have to find out now wheather armv5te is correct... thank you guys until now Feb 17 12:26:00 qecko: the "corrupted section header size" is normal btw, it happens because of stripping, if you meant that with "doesnt look good" Feb 17 12:59:12 ok, I have now identified the problem Feb 17 12:59:46 I was building for the wrong arm architecture, I need the same flags that gemini uses Feb 17 13:02:05 I set the CFLAGS option in the target/linux/myplatform/Makefile but upon making he build with the default arm flags of arch armv5te and tune=xscale Feb 17 13:02:39 so where is the correct place to insert these? Feb 17 13:44:42 qecko: maybe you should do make target/clean before remaking? Feb 17 14:12:34 Lallino: yes, i did a 'make clean' before Feb 17 14:13:29 nbd * r19688 /trunk/package/mac80211/ (15 files in 2 dirs): mac80211: update to wireless-testing 2010-02-16 Feb 17 14:25:05 ok, i think it compiles with correct flags now... i had to enable the advanced options for developers and set TARGET_OPTIONS=y Feb 17 14:27:19 hi Feb 17 14:27:30 could someone please apply my nfsd nfs V4 patch Feb 17 14:27:36 ping nbd Feb 17 14:27:38 otherwise the CFLAGS option in the Makefile is ignored Feb 17 14:28:16 tripolar_: did you post it on the ml? Feb 17 14:28:21 jow_laptop: yeah Feb 17 14:28:36 now nfs&nfsd is optional Feb 17 14:28:42 now nfs&nfsd V4 is optional Feb 17 14:29:18 about your patch: Feb 17 14:29:26 why have both fs-nfs-common-v4 and fs-nfs-v4? Feb 17 14:29:36 is there anything that uses one, but not the other? Feb 17 14:29:44 none of them works Feb 17 14:29:54 ? Feb 17 14:29:58 so disable or enable them when needed Feb 17 14:30:06 nfs V4 only eats up space Feb 17 14:30:24 and mounting doesnt work when there are no user space tools Feb 17 14:30:52 no, you got my question wrong Feb 17 14:30:57 why have two packages for v4 support? Feb 17 14:31:00 instead of only one Feb 17 14:31:18 because there is also a nfs kernel module and a nfsd kernel module Feb 17 14:32:07 thought this would be a good idea . but yeah they can be merged into on "menuconfig" entry Feb 17 14:32:10 on=one Feb 17 14:32:36 nfs and nfsd share common Feb 17 14:32:56 nbd: maybe someone wants nfs V4 but not nfsd V4 for example Feb 17 14:33:02 or the other way Feb 17 14:33:19 space is limited so i think its a good idea Feb 17 14:33:39 your patch puts auth_rpcgss in both nfs-v4 and nfs-common-v4 Feb 17 14:33:43 that's wrong Feb 17 14:34:12 nbd it was used in common befor so it was also used with nfs[d] Feb 17 14:34:27 why is it wrong? Feb 17 14:34:41 then it was also wrong befor ;) Feb 17 14:34:48 two packages provide the same file, and one depends on the other Feb 17 14:34:57 so there's a conflict on the filesystem Feb 17 14:35:05 ahh okay Feb 17 14:35:22 hmm but i think both need it Feb 17 14:35:31 but if one depends on the other, only one has to provide it :P Feb 17 14:35:43 and no, it wasn't wrong before, as only one package provides it without your patch Feb 17 14:36:07 but it is in common-v4 so this package wont be reinstalled Feb 17 14:36:19 i dont see the problem here Feb 17 14:36:26 both deend on common Feb 17 14:36:31 deend=depend Feb 17 14:36:52 two packages must not provide the same files if installed at the same time Feb 17 14:36:56 it's as simple as that Feb 17 14:37:15 okay and how can this be solved? Feb 17 14:37:45 but nfs V4 cant be installed afterwords because of the =y Kconfig entry Feb 17 14:37:56 the same for nfsd Feb 17 14:38:04 it has to be selected at build time Feb 17 14:38:27 imho just fold everything into common-v4 Feb 17 14:38:51 i had it this way but i then tought it would be a better idea to split it Feb 17 14:38:57 even if it might be possible to use v4 server but only v3 client, i don't think it's worth adding extra complexity for the package selection for that Feb 17 14:39:06 okay Feb 17 14:39:08 do the simple thing first Feb 17 14:39:16 and if somebody needs the more complex thing, we can still consider adding it Feb 17 14:39:22 okay Feb 17 14:39:44 do u modify it or should i do it? Feb 17 14:40:18 please send an updated patch Feb 17 14:40:22 okay Feb 17 14:44:09 nbd: would u add a dep on nfs-common or only a dep on @LINUX_2_6? Feb 17 14:44:20 @LINUX_2_6 is the best idea i think Feb 17 14:44:50 ahh and common Feb 17 14:51:54 nbd: send to ml Feb 17 14:54:01 nbd is it now okay? Feb 17 16:44:31 obsy * r19689 /packages/net/proftpd/Makefile: [package] proftpd: update to 1.3.2d, cleanup Makefile Feb 17 16:48:06 QUIT Feb 17 17:19:04 ping xMff: I've got the remote-gdb script connecting to gdbserver, but it complains about no debug symbols found. Is there an easy way to check the binary? Feb 17 17:21:14 netprince: did you compile it with -g ? Feb 17 17:22:19 lars * r19690 /packages/net/weechat/ (Makefile patches/010-use-pkg-config.patch): Feb 17 17:22:19 [packages] weechat: Use pkg-config instead of gnutls-config Feb 17 17:22:19 We don't ship gnutls-config and configure might end up using the hosts Feb 17 17:22:19 gnutls-config. Feb 17 17:24:11 The iptables INSTALL recommended -ggdb3 -O0 for debugging. I believe I have that added. The binary doubled in size... Feb 17 17:24:47 netprince_: the openwrt calls strip during packaging Feb 17 17:25:07 netprince_: does the client or the server complain about missing symbols? Feb 17 17:25:29 the server complains Feb 17 17:25:40 its there a flag to set to avoid stripping? Feb 17 17:25:43 s/its/is Feb 17 17:26:31 I'm sorry, the remote-gdb complains (thats the client I suppose) Feb 17 17:26:47 yes Feb 17 17:26:56 netprince: the part running on the router? Feb 17 17:27:08 no the local gdb executable Feb 17 17:27:26 netprince_: which binary did you pass to it? Feb 17 17:27:32 the part on the router (gdbserver) seems pretty hapy Feb 17 17:27:36 netprince: then try file /path/to/your/binary in gdb Feb 17 17:28:25 first I tried /sbin/iptables, then I tried iptables-multi from the iptables build directory Feb 17 17:28:27 netprince: it should be somewhere in build_dir/target*/root*/ Feb 17 17:28:56 netprince: use the same as the one you used in the gdbserver command line Feb 17 17:29:20 rtz: ok, will try Feb 17 17:29:34 netprince_: in .config, enable CONFIG_NO_STRIP and disable CONFIG_USE_SSTRIP Feb 17 17:29:42 then rebuild just the iptables package Feb 17 17:29:51 make package/iptables/{clean,compile} Feb 17 17:30:02 ok, will do that too Feb 17 17:30:05 thanks Feb 17 17:32:57 maybe this works too but I'm not sure: make ... RSTRIP=/bin/true Feb 17 17:38:06 yes this works too Feb 17 17:38:11 easier than chaging .config Feb 17 17:38:18 awsome, will do it that way Feb 17 18:27:44 nbd * r19691 /trunk/package/carl9170/ (Makefile patches/120-compile_fix.patch): carl9170: update to version 1.0.1.1 and fix a compile error Feb 17 18:45:22 ping nbd Feb 17 18:48:39 [florian]: ping Feb 17 18:57:26 [florian]: ping Feb 17 18:57:57 [florian]: fixed the reboot problem for the wdt Feb 17 19:11:16 [florian]: fixed the reboot problem for the wdt Feb 17 19:19:36 rtz: good call on the path to iptables Feb 17 19:23:16 netprince_: did you made progress? Feb 17 19:32:32 xMff: yes, both sides are happy now, iptables is bloated like 10x Feb 17 19:33:02 I'm not certain what to do now though Feb 17 19:33:27 netprince: what's your problem? Feb 17 19:33:30 coredump? Feb 17 19:33:55 or segfault or something like that? Feb 17 19:34:01 rtz2: ping Feb 17 19:34:13 rtz: I am hoping to tackle this: https://dev.openwrt.org/ticket/5141 Feb 17 19:36:18 ping nbd Feb 17 19:40:04 netprince_: start with replacing the iptables executable and the multiport userpspace match library Feb 17 19:40:14 netprince_: thne try to trace it Feb 17 19:40:27 netprince: did you check strace? Feb 17 19:40:44 netprince: so you know, if the syscalls are correct or not Feb 17 19:41:03 xMff: ok, I still need to grab the multiport library Feb 17 19:41:25 rtz2: yeah, we didn't see much information from strace Feb 17 19:41:42 netprince_: you can do that in /tmp as well, the iptables extension search path can be overridden with the XTABLES_LIBDIR environment variable Feb 17 19:42:03 multiple entries separated by colon Feb 17 19:42:24 that is helpful, will do that Feb 17 20:04:03 build #14 of octeon is complete: Failure [failed compile_10] Build details are at http://tksite.gotdns.org:8010/builders/octeon/builds/14 Feb 17 20:14:40 tripolar_: i'll take a look at it later Feb 17 20:14:50 btw. you don't have to ping me over and over and over again Feb 17 20:14:59 nbd okay Feb 17 20:15:13 nbd: is the current compat-wireless supposed to work with 2.6.30 kernels? Feb 17 20:17:05 it should Feb 17 20:17:08 tripolar_: what happens when a kernel with all kmods enabled is built? will the nfs4 stuff end up as module or will it increase the kernel image size again? Feb 17 20:17:38 I mean, obviously it will be a module but how much size does it add to the kernel itself Feb 17 20:18:19 xMff the module uses about 500kb more Feb 17 20:18:22 nfsd Feb 17 20:18:35 i used nfsd a long time ago with nfsdv4 Feb 17 20:33:12 xMff: hey question on luci-splash Feb 17 20:54:18 nbd: you still here? Feb 17 20:55:05 nbd: there is a problem int the mac80211 package Feb 17 20:56:02 error: too many arguments to function 'genlmsg_unicast' ? Feb 17 20:56:32 xMff: yes Feb 17 20:57:16 xMff: it changed with 2.6.32 Feb 17 20:57:43 Kaloz: FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF TRAC .10 WORKFLOW Feb 17 20:57:45 xMff: could you create a patch for it? Feb 17 20:57:46 CHANGE IT Feb 17 20:58:41 xMff: it should be enough to leave out the first parameter Feb 17 20:58:58 in line 5670 of nl80211.c Feb 17 21:00:45 need to read into the structure of the compat-wireless package first, wasn't there kernel version specific compat headers or so? Feb 17 21:01:08 and with which kernel version does it fail? Feb 17 21:01:10 xMff: no idea Feb 17 21:01:18 xMff: it fails with .30 Feb 17 21:01:26 xMff: and the change happend with .32 Feb 17 21:01:45 meh Feb 17 21:02:17 question is how long we'll keep < .32 around Feb 17 21:02:32 Weedy: don't blog here Feb 17 21:03:08 xMff: well, I susupect the error will get fixed rather quickly in comat-mac80211 Feb 17 21:03:22 xMff: and the function isn't mentioned anywhere in the package Feb 17 21:05:20 and I would really love to update rdc to .32 :/ Feb 17 21:06:19 xMff: less than two weeks Feb 17 21:08:00 nice Feb 17 21:19:15 Kaloz: wat Feb 17 21:20:20 OutBackDingo: go ahead Feb 17 21:20:24 ping cshore Feb 17 21:21:01 pong xMff Feb 17 21:21:44 cshore: regarding your blank page issue and vlan... this sounds familar somehow, didn't we have had this quite a while ago too on bcm63xx? Feb 17 21:23:15 xMff: yes Feb 17 21:23:43 xMff: you fixed it then Feb 17 21:23:59 actually I never did anything, it fixed itself somehow Feb 17 21:24:06 xMff: oh, ok Feb 17 21:26:34 xMff: this is just a process for auth right, doesnt do any dhcp or stuff right Feb 17 21:27:31 OutBackDingo: yes, it does not authenticate either, it is just for injecting some agreement page - if you need something ... to rely on you're better off with coova or similar I'd say Feb 17 21:28:00 xMff: so just straigh splash page Feb 17 21:28:09 correct Feb 17 21:28:10 got it Feb 17 21:28:39 xMff: do you plan on fixing the mac80211 problem? Feb 17 21:29:18 rtz2: I'll give it a try Feb 17 21:31:41 xMff: should be a simply ifdef thing Feb 17 21:31:58 xMff: there doesn't seem to be any special way to do it Feb 17 22:37:06 jow * r19692 /trunk/package/mac80211/patches/900-linux-2.6.30-compat.patch: [package] mac80211: add compile fix for pre-2.6.32 kernels Feb 17 22:39:47 xMff: thanks Feb 17 22:43:10 jow * r19693 /packages/utils/strace/Makefile: [packages] strace: fix compile with external toolchain by adding kernel include dir to cflags (#6695) Feb 17 22:43:32 np Feb 17 22:54:46 and Feb 17 22:54:49 sorry Feb 17 23:00:50 kaloz * r19694 /trunk/target/linux/sibyte/ (9 files in 2 dirs): [sibyte]: upgrade to 2.6.32.8 Feb 17 23:08:18 ping xMff Feb 17 23:08:40 pong cshore Feb 17 23:08:56 Is it possible to have lucid listen on a port other than 80? Feb 17 23:09:01 yes Feb 17 23:09:52 uci set lucid.http.address=otherport Feb 17 23:10:00 or listenaddr:port Feb 17 23:10:28 great, thanks Feb 17 23:11:42 and then uci commit of course Feb 17 23:11:57 and restart Feb 17 23:11:58 yes Feb 17 23:12:23 xMff: I'm going to try my router with LuCI again: I think the problem may be MTU Feb 17 23:12:26 with vlans Feb 18 00:33:08 thepeople * r19695 /packages/admin/sudo/Makefile: Feb 18 00:33:08 this patch updates sudo to the latest stable version: 1.7.2p3 Feb 18 00:33:08 -Raphael Feb 18 00:42:46 thepeople * r19696 /packages/utils/strace/ (Makefile patches-4.5.16/004-mips64_support.patch): Feb 18 00:42:46 this patches cleans up the Makefile and adds mips64 support for 4.5.16. Feb 18 00:42:46 -Raphael Feb 18 00:46:22 thepeople * r19697 /packages/libs/libupnp/Makefile: Feb 18 00:46:22 this patch fixes libupnp parallel build and cleans the Makefile a Feb 18 00:46:22 little. Feb 18 00:46:22 -Raphael **** ENDING LOGGING AT Thu Feb 18 02:59:57 2010