00:41 | ChrisG_ | u there biker? |
00:46 | cbxbiker61 | yep |
00:50 | ChrisG_ | am i good with a generic 3.5.7 kernel, or should i be splicing in patches from somewhere? |
00:52 | cbxbiker61 | all of my patches are in http://www.xilka.com/sheeva/kernel/3/3.5/3.5.7/source |
00:52 | rmull | ChrisG_: I use the one in https://github.com/vDorst/linux |
00:52 | ChrisG_ | decisions decisions :-) |
00:53 | rmull | vDorst is called "Thirsty" on the forums and IRC |
00:53 | rmull | He's been busy lately so he hasn't had a chance to update the stuff on github, but it's pretty good |
00:53 | rmull | Meanwhile cbxbiker61 does a good job maintaining xilka so it's up to you |
00:54 | ChrisG_ | thanks. glad to find good help here |
00:55 | ChrisG_ | will the generic kernel work, or am i wasting my time with this compile? |
00:55 | rmull | I haven't bothered trying it, you're probably better off with one that has patches integrated |
00:55 | rmull | but I'm sure your feedback is valuable |
00:55 | rmull | If you do end up trying it |
00:55 | ChrisG_ | i tried reading through the instructions on https://github.com/vDorst/linux and got lost |
00:56 | cbxbiker61 | it may work, but may not, that skb patch may be critical with dual nics |
00:57 | rmull | ChrisG_: git clone git://github.com/vDorst/linux.git |
00:57 | rmull | That will fetch the kernel sources |
00:58 | ChrisG_ | how do i merge the patches into my source? |
00:58 | ChrisG_ | i have that already in my home dir |
00:58 | cbxbiker61 | if you are using my patches, use patch -p1 -i patch.patch |
00:59 | ChrisG_ | should i pick one or the other? |
01:00 | cbxbiker61 | not sure if vDorst has the skb patch, if not...apply it and go from there |
01:00 | ChrisG_ | so, i have the generic kernel under /usr/src/linux |
01:01 | cbxbiker61 | from generic you can apply all of the patches in my patch dir |
01:01 | ChrisG_ | just download them to my home dir and run the patch commands? |
01:02 | cbxbiker61 | local patches35cubox='linux-3.5-CuBox.patch linux-3.5-kirkwood-orion_uart0_init.patch linux-3.5-ZSMalloc-ZCache-arm.patch 201-bump-galcore-to-3184.patch 202-bump-galcore-to-3184-1.patch linux-3.0-SDIO-micro-AP.patch linux-3.0.4-LIBERTAS-WEXT_PRIV.patch 0003-Initial-defconfig.patch sheeva-3-leds.patch 0002-8688-SDIO-micro-AP-suppo-3.2.patch linux-3.5.7-remove-skb-recycling.patch' |
01:02 | cbxbiker61 | that's the order i apply them |
01:09 | ChrisG_ | why is this patch so huge? linux-3.5-CuBox.patch |
02:04 | cbxbiker61 | linux-3.5-CuBox.patch rolls in all of the out-of-band patches for CuBox, which at this point are quite extensive |
03:58 | ChrisG__ | gah... i think i may pull all the hair i have left out :-) |
03:58 | dotarray | not going well for you ChrisG__? |
03:59 | ChrisG__ | after a good 4 hours of recompile, i got the the very end and it was about to write out the new kernel |
03:59 | ChrisG__ | ankd got "ld: no machine record defined" |
04:00 | ChrisG__ | in the mean time was trying to set up a cross compile environment on my centos box, but i fail boated that too :-) |
04:01 | dotarray | ahahahaha... i mean... |
04:01 | dotarray | awwwww :) |
04:01 | ChrisG__ | :-) |
04:03 | ChrisG__ | over the years i guess i got a little too comfortable not needing the recompile the kernel... don't think i have done it since... hmmm... 2007? |
04:04 | ChrisG__ | prob freeBSD 4.x |
04:04 | ChrisG__ | i will not be defeated :-) |
04:08 | ChrisG_ | ever seen that error on compile? |
04:09 | ChrisG_ | google not being helpful |
04:09 | ChrisG_ | mostly hits on people compiling on specific andriod phones |
04:15 | ChrisG_ | any experience compiling kernels dotarray? |
04:16 | dotarray | unfortunately not yet :( |
04:16 | ChrisG_ | if you are ever looking for fun :-) |
04:16 | dotarray | oh yeah, i bet :P |
04:17 | ChrisG_ | @cbxbiker61... any chance you can walk me through this? |
04:31 | ChrisG_ | try #3 |
04:35 | ChrisG_ | using source from here --> git clone https://github.com/vDorst/linux.git |
04:35 | ChrisG_ | and config from here -- http://www.xilka.com/sheeva/3/3.5/3.5.7/release/2/cubox-3.5.7.config |
04:36 | ChrisG_ | added CONFIG_BRIDGE=y to .config |
04:37 | ChrisG_ | ran 'make oldconfig' and walked through the settings |
04:37 | ChrisG_ | this time there was a config option for machine type |
04:38 | ChrisG_ | * Marvell Dove Implementations |
04:38 | ChrisG_ | * |
04:38 | ChrisG_ | Marvell DB-MV88AP510 Development Board (MACH_DOVE_DB) [N/y/?] n |
04:38 | ChrisG_ | CompuLab CM-A510 Board (MACH_CM_A510) [N/y/?] n |
04:38 | ChrisG_ | SolidRun CuBox (MACH_CUBOX) [N/y/?] (NEW) y |
04:38 | ChrisG_ | guess i'll know in another 4 hours :-) |
04:42 | ChrisG_ | @dotarray - so, what are you using your CuBox for? |
06:06 | dotarray | http://shouldihangoutmylaundry.com/ |
06:07 | ChrisG_ | :-) |
06:07 | ChrisG_ | i guess if you want it frozen |
06:07 | dotarray | :D |
06:08 | dotarray | 34C / 94F here :) |
06:08 | ChrisG_ | nice |
06:08 | ChrisG_ | where is there? |
06:08 | dotarray | adelaide, south australia :) |
07:06 | ChrisG_ | bah, just shoot me |
07:10 | ChrisG_ | LD init/built-in.o |
07:10 | ChrisG_ | arch/arm/kernel/built-in.o: In function `iwmmxt_do': |
07:10 | ChrisG_ | :(.text+0x6848): undefined reference to `iwmmxt_task_release' |
07:10 | ChrisG_ | :(.text+0x6860): undefined reference to `iwmmxt_task_switch' |
07:10 | ChrisG_ | make: *** [vmlinux] Error 1 |
07:11 | ChrisG_ | i give up for tonight anyway.... |
18:49 | dung | greetings |
18:54 | dung | ChrisG_, i got such error when a driver/feature configured as module references to a _built-in_ (iwmmxt in your case) |
19:05 | ChrisG_ | hmm... interesting |
19:05 | ChrisG_ | i was finally able to get it to compile, and then the kernel hun on boot |
19:05 | ChrisG_ | hung |
19:05 | dung | i'm just guessing you could a) compile the driver into the kernel, or b) patch arch/arm/kernel/built-in.c? |
19:06 | ChrisG_ | i'm starting over :-) |
19:07 | ChrisG_ | going to compile from rabeeh's tree this time |
19:07 | dung | less experimental :) |
19:08 | ChrisG_ | hmmm, all i really want to do is add BRIDGE support to the existing kernel |
19:08 | ChrisG_ | it was compiled as NO |
19:08 | ChrisG_ | should i add it as compiled in, or add as a module? |
19:09 | dung | before it's loaded as module all the time, compile it in |
19:10 | ChrisG_ | ok |
19:11 | ChrisG_ | before i compile i will search through the config and see how it is loaded |
19:16 | dung | i can't suggest my config as a start - it doesn't support swapping nor package filtering.. |
19:17 | dung | but all the built-ins |
19:17 | dung | into CuBox.. i believe |
19:22 | ChrisG_ | unfortunately the installer package does not give a starting config |
19:22 | ChrisG_ | not that i can find anyway |
19:22 | ChrisG_ | but i was able to build what looks like a reasonable on with make cubox_defconfig and modify from there to enable 802.1d bridging |
19:23 | ChrisG_ | it takes about 4 hours to build :-) |
19:24 | ChrisG_ | i'm working on getting an environment together on my CentOS box to cross compile |
19:24 | ChrisG_ | but i want to see it work native before i throw in that variable |
19:26 | bencoh | hmm |
19:26 | bencoh | why not running for cross directly ? |
19:37 | dung | i meant believing i had compiled CuBoxs hardware drivers directly into _the kernel_, you have to enable modular |
19:39 | dung | the experimental thing is mainline source from kernel.org and DT blob ? |
19:39 | dung | and a lot of variants - oh my! |
19:44 | dung | but.. if the device tree doesn't tell galcore it's less important for distros generic kernels |
19:44 | dung | am i right? |
19:46 | dung | i will look into the DT some times |
19:47 | dung | most ARMs bootloaders are U-Boot |
19:47 | dung | all? |
19:47 | dung | non-PCs? |
19:49 | dung | i'm also interested in the evolution of coreboot |
19:51 | dung | well.. U-Boot is BIOS/bootloader on non-PCs ? |
19:51 | dung | wow |
19:52 | ChrisG_ | @bencoh, sorry, got distracted |
19:52 | ChrisG_ | what do you mean running for cross directly? |
19:53 | bencoh | "but i want to see it work native before i throw in that variable" you were talking about native build, weren't you ? |
19:54 | ChrisG_ | yeah |
19:55 | ChrisG_ | i'm compiling on the cubox |
19:55 | ChrisG_ | slooooowly :-) |
19:55 | bencoh | then crossbuilding the kernel would be quicker, and shouldn't be much different |
19:56 | ChrisG_ | i agree |
19:56 | ChrisG_ | "much different" :-) |
19:56 | bencoh | (make ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- uImage) |
19:56 | ChrisG_ | i have not had a successful build native yet |
19:56 | bencoh | there is no point in trying a native build first for the kernel |
19:56 | bencoh | (since it has no external deps whatsoever) |
19:57 | ChrisG_ | good point |
19:58 | dung | yes - compiling the kernel is easier is that point!# |
19:59 | ChrisG_ | ok, i will give it a try on a more powerful box |
20:02 | dung | kernels don't have depencies *g* |
20:05 | dung | naja |
20:15 | lubiana | lal, there are quite a few germans here |
20:25 | dung | nei |
20:26 | dung | +-0 .at |
20:26 | dung | ch it fr |
20:26 | dung | ... |
20:28 | dung | apropos firebrick.. is a named color |
20:28 | dung | ..right after brown |
20:32 | dung | seagreen is between black and white |
20:33 | dung | that's true *g* |
20:36 | dung | a) materials reflect light, b) rgb on screen |
20:39 | punkt | dung: what do you mean with coreboot for the cubox? |
20:40 | punkt | dung: did you take a look on barebox, its an more comfortable bootloader with lots of examples already |
20:52 | dung | i dont know bb. punkt, coreboot is a bootloader for PCs |
20:53 | dung | s/PC-BIOS/PC/ |
20:56 | dung | s/PC/ix86/ |
20:57 | dung | Coreboot and U-Boot are BIOS bootloaders? |
20:59 | dung | since a few years i want to change such a chip and 'flash' BIOS on i386 |
21:04 | dung | should be possible to flash a lot of PC mainboards |
21:05 | dung | it seems U-Boot is the way for all the other |
21:06 | dung | after BIOS/bootloader - a kernel |
21:07 | dung | with drivers |
21:08 | dung | let's call everything beyond that libraries |
21:09 | dung | or BIOS/bootloader firmware |
21:09 | dung | all that is still software |
21:16 | dung | till i do it, PC has another meaning |
22:10 | ChrisG_ | success! |
22:11 | ChrisG_ | thanks to all for moral support and help :-) |
23:06 | dung | so bridging uses the iwmmxt :) |
23:07 | ChrisG_ | don't think so, not sure... i don't think that was related to my adding the bridge option |
23:07 | ChrisG_ | i think it was my poor choice to take the kernel source from one builder, and a starting config from another :-) |
23:07 | dung | ok |
23:08 | ChrisG_ | but i now have a very small and handy passive sniffer :-) |
23:08 | dung | aha |
23:09 | dung | nice |
23:09 | dung | very handy |
23:10 | dung | how do you use it? |
23:11 | ChrisG_ | i built a passive tap, that has a pass through port (in --> out) and two listening ports |
23:11 | ChrisG_ | one port listens to the send data |
23:11 | ChrisG_ | the other listens to recieve |
23:11 | ChrisG_ | which is why i need to bridge |
23:11 | ChrisG_ | eth1 on send, eth2 on rcv |
23:11 | dung | and you put all this into a db? |
23:12 | ChrisG_ | bridge them together and you see two way traff |
23:12 | dung | and let e.g. snort walk through? |
23:12 | ChrisG_ | right now, working on two applications |
23:12 | ChrisG_ | snort/barnyard |
23:12 | ChrisG_ | and YAF |
23:12 | ChrisG_ | yaf will give me netflow data |
23:12 | ChrisG_ | reporting with snorby |
23:13 | ChrisG_ | snort/barnyard will be local on the cubox |
23:13 | ChrisG_ | mysql will be on another linux box |
23:13 | dung | have fun :) |
23:13 | ChrisG_ | as well as the web reporting |
23:13 | ChrisG_ | :-) |
23:13 | ChrisG_ | making progress... |
23:13 | ChrisG_ | next challenge is to get the bridge config to load correctly on startup |
23:14 | ChrisG_ | works fine from the command line, but i have something borked in /etc/network/interfaces |
23:20 | dung | you'll want a dual nic board |
23:20 | ChrisG_ | i have 3 NICs in my cubox |
23:20 | ChrisG_ | the built in, and 2 USB |
23:20 | ChrisG_ | it's perfect :-) |
23:21 | dung | i think it's cheap to add more ethernet |
23:21 | ChrisG_ | to a cubox? |
23:22 | ChrisG_ | where are you going to add them? |
23:22 | dung | soldering *g* ? |
23:22 | ChrisG_ | riiiight |
23:22 | ChrisG_ | :-) |
23:23 | dung | software side it gives a lot of possibilities |
23:24 | dung | i dont't know yet why 10MB/s LAN feels smoother than 40MB/s USB |
23:24 | dung | i might have missed some point |
23:25 | ChrisG_ | cause you pass through software and CPU when you go through USB |
23:25 | ChrisG_ | and hardware NICs have off-loading capabilities |
23:26 | ChrisG_ | i'm working with a low bandwidth circuit |
23:26 | ChrisG_ | less than 20M |
23:26 | ChrisG_ | so should be find |
23:26 | ChrisG_ | i'm a little worried about loosing traffic under load though |
23:26 | ChrisG_ | will be able to test soon |