| === Guest74361 is now known as Martyn | ||
| joshzhao | hi | 08:16 |
|---|---|---|
| joshzhao | did ubuntu support s3c6410 ? | 08:17 |
| Stskeeps | think it's armv5 capable | 08:19 |
| joshzhao | yes | 08:39 |
| NCommander | whats the s3c6410 | 09:00 |
| lool | joshzhao: You need to build your own kernel and any integration bits, but the userland should work, it's ARM11 | 09:13 |
| lool | NCommander: 10:11 < lool> NCommander: What are you looking into ATM? Could you take that? | 09:19 |
| lool | 10:12 < lool> NCommander: Also, kexec-tools was finally added to P-a-s, do you think you could try it out? it built on armel | 09:19 |
| lool | NCommander: "that" being d-i | 09:19 |
| NCommander | Been working on that already | 09:20 |
| * NCommander learns how d-i goes together, but I'm making no headway on ARM | 09:20 | |
| NCommander | The last kernel upload made my board completely unreliable, and with it getting files at the speed of dial up .... | 09:20 |
| NCommander | Same thing goes for kexec (that being said, I used the 2.0.0 release of kexec-tools on Babbage, and got nothing, so unless we're carrying patches ...) | 09:21 |
| lool | NCommander: Please try the ubuntu kernels | 09:22 |
| lool | NCommander: If there are bugs, file them | 09:22 |
| lool | NCommander: Perhaps compare the exact same scenario in qemu/i386 and qemu/armel? | 09:23 |
| NCommander | lool, it doesn't work in qemu/armel with our kernel as of the last time I tried it, I'm building a new qemu/armel image now so I can work on d-i, I'll test it once it finishes installing. | 09:24 |
| lool | NCommander: Where's the bug report? | 09:34 |
| lool | NCommander: the same test case passes on i386? | 09:34 |
| NCommander | lool, I dunno if I filed a bug on this in Ubuntu, I did bring it upstream; no ping reply. | 09:35 |
| joshzhao | i don't know what kind of Ubuntu Kernel tree support 6410? | 09:37 |
| joshzhao | s3c6410 is based on arm11 | 09:38 |
| joshzhao | samsung | 09:38 |
| lool | NCommander: Are you waiting for me to ask you to file one or will you file one now? :-) | 09:40 |
| NCommander | lool, I'll file once I rerun the tests, filing bugs based on old data is a bad thing. | 09:40 |
| joshzhao | lool: do you know what kind of Ubuntu Kernel tree support 6410? | 09:42 |
| lool | joshzhao: No | 09:43 |
| lool | joshzhao: The Ubuntu kernel tree is close to the upstream one | 09:43 |
| joshzhao | thanks lool | 09:56 |
| mcasadevall | lool, if you have a few minutes, would you like to process one of the outstanding MIRs on RedBoot or ecosconfig? | 10:01 |
| === Nicke_ is now known as Nicke | ||
| lool | mcasadevall: I'm queueing it up, but don't consider it assigned to me yet; I'm afraid I have a bunch of urgent stuff which is landing on my plate these days | 10:28 |
| mcasadevall | lool, no problem. | 10:28 |
| ogra | amitk, meh, FTBFS on imx51 | 10:30 |
| ogra | Building modules, stage 2. | 10:30 |
| ogra | MODPOST 739 modules | 10:30 |
| ogra | ERROR: "cpufreq_gov_performance" [arch/arm/plat-mxc/cpufreq.ko] undefined! | 10:30 |
| ogra | ERROR: "get_cpu_wp" [arch/arm/plat-mxc/cpufreq.ko] undefined! | 10:30 |
| ogra | ERROR: "dvfs_core_is_active" [arch/arm/plat-mxc/cpufreq.ko] undefined! | 10:30 |
| ogra | ERROR: "cpu_wp_nr" [arch/arm/plat-mxc/cpufreq.ko] undefined! | 10:30 |
| ogra | make[3]: *** [__modpost] Error 1 | 10:30 |
| ogra | make[2]: *** [modules] Error 2 | 10:30 |
| ogra | make[1]: *** [sub-make] Error 2 | 10:30 |
| ogra | amitk, throw cpufreq out, it has no use on the babbage anyway | 10:30 |
| lool | It looks like you need to turn on these configs for the imx51 cpufreq backends | 10:30 |
| lool | Hmm they are | 10:31 |
| ogra | waht for ? | 10:32 |
| amitk | ogra: I know, working on it | 10:32 |
| ogra | ah, good | 10:32 |
| ogra | intresting that it didnt fail in the PPA build | 10:33 |
| amitk | ogra: could you test -11.35 (FINAL). It has AA and aufs, cpufreq solved. This might be our last chance before beta to fix this issue. | 11:03 |
| ogra | amitk, after the meeting | 11:03 |
| amitk | ogra: sure | 11:03 |
| ogra | (i'll try to do i aside the meeting if really urgent though) | 11:03 |
| amitk | naah.. I still have to work on the d-i bits. So anytime in the next 2-3hrs is good | 11:04 |
| ogra | oh, wait, my inner clock is off one hour | 11:04 |
| ogra | i can do it now | 11:04 |
| amitk | did we change time? | 11:04 |
| ogra | no, its 12 UTC | 11:04 |
| * ogra smiles, a working NIC makes everything so much easier | 11:13 | |
| ogra | no more plugging around of USB keys to transfer kernels and initramfs | 11:14 |
| amitk | ogra: make sure you get this http://people.ubuntu.com/~amitk/linux-image-FINAL-2.6.28-11-imx51_2.6.28-11.35_armel.deb | 11:14 |
| ogra | yes, thats what i just installed and copied to my desktop | 11:15 |
| ogra | now i'm doing a dist-upgrade to get new procps etc ... then i'll reboot and xmodem the kernel and initramfs over | 11:15 |
| lool | ogra or amitk: I need /proc/cpuinfo on babbage with CONFIG_NEON=y (probably on the kernels you're running) and with the NEON hwacps merged in 2.6.28-10.33 | 11:16 |
| ogra | amitk, AA oops :( | 11:34 |
| amitk | ogra: that is underinvestigation. just disable in on the cmdline | 11:35 |
| ogra | same | 11:36 |
| ogra | apparmor.enable=0 ... still oopses | 11:36 |
| ogra | grmbl | 11:38 |
| ogra | that costed me my working setup | 11:39 |
| * ogra needs to rebot after upgrade ... | 11:40 | |
| ogra | *reboot | 11:40 |
| amitk | ogra: are you capturing the oops? | 11:46 |
| lool | Would really appreciate a /proc/cpuinfo | 11:46 |
| * ogra cant give a /proc/cpuinfo without getting into the system | 11:46 | |
| ogra | amitk, one sec | 11:46 |
| amitk | lool: I'll have to build a new kernel with NEON enabled again | 11:47 |
| lool | Is there a working babbage kernel I can run right now which has NEON support? | 11:47 |
| lool | is the ubuntu one working for instance? | 11:47 |
| amitk | lool: try the original ubuntu one, or one from http://people.ubuntu.com/~amitk/ | 11:48 |
| ogra | amitk, http://paste.ubuntu.com/133579/ | 11:48 |
| lool | Will either boot without initramfs? | 11:49 |
| ogra | __aa_find_profile it seems | 11:49 |
| amitk | ogra: same oops | 11:49 |
| ogra | lool, yes, but you need rootdelay | 11:49 |
| ogra | and no uuid root= line | 11:49 |
| amitk | ogra: I'm compiling another one with AA turned off by default | 11:56 |
| ogra | thanks | 11:56 |
| === mcasadevall is now known as NCommander | ||
| amitk | ogra: same 'FINAL' .deb with AA compiled in but turned off by default | 12:14 |
| amitk | back from lunch | 12:14 |
| * NCommander sighs | 12:15 | |
| NCommander | THere is a max initrd size it seems on the Babbage | 12:15 |
| NCommander | or something crazy ... | 12:15 |
| ogra | how big did you b'koat yours ? | 12:15 |
| ogra | *bloat | 12:15 |
| ogra | up to 4.5M worked fine here | 12:15 |
| NCommander | 2.5MB, I'm getting incomplete write, which suggests its too large | 12:16 |
| ogra | funny | 12:16 |
| ogra | i'm using a 4.2M one atm | 12:16 |
| ogra | and had a 4.5 one before | 12:16 |
| NCommander | ogra, its I'm an idiot problem :-) | 12:25 |
| === rwhitby` is now known as rwhitby | ||
| ogra | amitk, hmm, lost all my input capabilities again with the latest build | 13:23 |
| ogra | amitk, did you remover the compiled in usb host drivers again ? | 13:23 |
| ogra | *remove | 13:23 |
| ogra | amitk, hrm, and i get compcache by default | 13:24 |
| ogra | amitk, additionally the rest of the usb stack doesnt get loaded either | 13:26 |
| ogra | amitk, and i dont see an indication of a kernel event happening | 13:26 |
| ogra | aha, loading ehci-hcd helps | 13:27 |
| amitk | ogra: I haven't touched anything else, you can check the config file | 13:27 |
| ogra | weird | 13:27 |
| ogra | my rootfs doesnt mount, ehci-hcd should be compiled in | 13:28 |
| ogra | there is no trace of usb-storage in my initramfs | 13:29 |
| amitk | ogra: it is | 13:29 |
| ogra | (initramfs) ls /lib/modules/2.6.28-11-imx51/kernel/drivers/usb/ | 13:29 |
| ogra | host | 13:29 |
| ogra | (initramfs) ls /lib/modules/2.6.28-11-imx51/kernel/drivers/usb/host/ | 13:29 |
| ogra | ehci-hcd.ko | 13:29 |
| ogra | thats all i have | 13:29 |
| ogra | (initramfs) uname -a | 13:29 |
| ogra | Linux (none) 2.6.28-11-imx51 #35 Thu Mar 19 12:38:40 EET 2009 armv7l unknown | 13:29 |
| * amitk checks again | 13:30 | |
| amitk | ogra: shoot me, you are right. EHCI_HCD is a module | 13:31 |
| ogra | (initramfs) cat conf/initramfs.conf |grep MODULES | 13:31 |
| ogra | # MODULES: [ most | netboot | dep | list ] | 13:31 |
| ogra | MODULES=most | 13:31 |
| ogra | so usb-storage should be there actually | 13:32 |
| ogra | hmm | 13:32 |
| amitk | ogra: storage is compiled in, so is USB_HID (for keyboard), but i left EHCI as a module for some reason | 13:33 |
| ogra | ah | 13:33 |
| ogra | that expleins it | 13:33 |
| ogra | *explains | 13:33 |
| amitk | that config file is swimming in front of me now | 13:33 |
| ogra | ok | 13:33 |
| ogra | i dont get why it uses compcache though | 13:33 |
| amitk | ogra: because it is a module | 13:34 |
| amitk | do you want it removed? | 13:34 |
| ogra | it should only be included if there are less than 256M ram | 13:34 |
| ogra | no, keep it | 13:34 |
| ogra | but the script shouldnt fire | 13:34 |
| ogra | (initramfs) cat /proc/meminfo | 13:35 |
| ogra | MemTotal: 482808 kB | 13:35 |
| ogra | there is definately more than 256M | 13:35 |
| amitk | It is only a module | 13:35 |
| ogra | argh | 13:36 |
| ogra | TOTAL_RAM=$( grep MemTotal /proc/meminfo |tr -d ': [A-Z][a-z]') | 13:36 |
| ogra | # Do not use compcache on the liveCD if we have more than 512M | 13:36 |
| ogra | if [ "${BOOT}" = "casper" ]; then | 13:36 |
| ogra | if [ "${TOTAL_RAM}" -gt 524288 ]; then | 13:36 |
| ogra | exit 0 | 13:36 |
| ogra | fi | 13:36 |
| ogra | fi | 13:36 |
| ogra | thats bad | 13:36 |
| amitk | hehe | 13:37 |
| ogra | i mean it works, but its pointless to have on babbage | 13:39 |
| lool | The distro kernel doesn't boot for me | 13:47 |
| lool | Uncompressing Linux............................................................. | 13:47 |
| lool | And nothing more | 13:47 |
| ogra | amitk, oh, does compcache get autoloaded if its there ? (it should never load alone) | 13:48 |
| lool | Can someone hand me a recent kernel build with the NEON hwcaps patch? | 13:48 |
| ogra | http://people.ubuntu.com/~amitk/linux-image-NO-AA-AUFS-USB-COMPILED-2.6.28-10-imx51_2.6.28-10.32_armel.deb not sure about neon | 13:49 |
| ogra | but thats definately a working one | 13:49 |
| lool | Thanks | 13:49 |
| amitk | lool: grab the .deb from my p.u.c and grep the config. | 13:49 |
| lool | amitk: It's not a config | 13:49 |
| amitk | lool: you want the patch AND the CONFIG_NEON configured in, right? | 13:50 |
| lool | amitk: Correct | 13:50 |
| lool | But CONFIG_NEON has been set since forever | 13:50 |
| * ogra wonders how long forever is for an unfinished kernel :) | 13:51 | |
| ogra | if [ "${BOOT}" = "casper" ]; then | 13:52 |
| ogra | if [ "${TOTAL_RAM}" -gt 524288 ]; then | 13:52 |
| ogra | exit 0 | 13:52 |
| ogra | fi | 13:52 |
| ogra | fi | 13:52 |
| ogra | modprobe -q --ignore-install compcache compcache_size_kbytes="$(($(sed -nre 's/^MemTotal:\s*([0-9]+) kB$/\1/p' /proc/meminfo) * 25 / 100))" | 13:52 |
| ogra | ARGH | 13:52 |
| ogra | so if i boot with boot=casper and the system has 512M it will exit ... if the system has no 512M *and* i dont boot with casper compcache will load in *any* case | 13:53 |
| ogra | thats so wrong ... | 13:53 |
| ogra | i guess that needs an else condition ... | 13:54 |
| ogra | oh, my, what was i smoking when i wrote that code | 13:54 |
| * ogra takes a break ... thats so embrassing | 13:55 | |
| amitk | good stuff? :) | 13:55 |
| amitk | ogra: new 'FINAL' uploaded. Hopefully the real FINAL :-/ | 13:56 |
| lool | The above kernel doesn't work for me either | 14:02 |
| lool | Does it need an initramfs to output anything? | 14:02 |
| lool | GRR | 14:07 |
| lool | screen doesn't like too long commands | 14:07 |
| lool | Ok, the kernel loaded over ymodem at least says "done, booting the kernel." but nothing more | 14:15 |
| lool | ogra: is there serial console in the kernel you mentionned as working? | 14:15 |
| lool | ogra: I definitely get zero output from the kernel with the .deb you pointed me at, only up to "done, booting the kernel.", even with console=ttymxc0,115200 console=tty0 on serial and VGA output | 14:23 |
| lool | And the same thing with the FINAL .deb above | 14:26 |
| lool | I'm booting with RedBoot> exec -c "console=ttymxc0,115200 console=tty0 root=/dev/mmcblk0p2 ro debug noinitrd rootdelay=2" | 14:29 |
| lool | I confirmed that the way I upload kernels works as if I reload zImage_DVI it works | 14:31 |
| lool | BTW I'm now updating the SD card directly | 14:32 |
| lool | With dd and fis | 14:32 |
| lool | amitk: Can you confirm the above kernels will work *without* initrd? | 14:35 |
| amitk | lool: dunno. I have an initramfs | 14:39 |
| lool | amitk: Could you show me fis list and/or a boot log so that I can use the proper fis create command here? | 14:40 |
| FlimFlamMan | hello. does anyone know when we might see usable Ubuntu netbooks on ARM? | 14:40 |
| lool | I also need the cmdline with initramfs, so far I only have without | 14:40 |
| lool | FlimFlamMan: No final date yet | 14:41 |
| FlimFlamMan | lool: thanks. what's the best place to keep track of how things are progressing, and what the shape of the final product will be? will the "experience" basically be the same as on x86? | 14:42 |
| lool | FlimFlamMan: The images Ubuntu will release for arm devices match images which exist for i386 already | 14:43 |
| lool | FlimFlamMan: Following Ubuntu news channels should be enough to get the news | 14:44 |
| FlimFlamMan | lool: Thanks for the information. (I'm trying to judge whether i should hold off on a netbook purchase or buy an interim unit.) | 14:47 |
| ogra | lool, yes, works fine for me | 14:47 |
| lool | ogra: without initrd? | 14:49 |
| lool | ogra: I'm using dpkg -x, replacing the kernel with the file in boot/vmlinuz-foo exactly in the same way as zImage_DVI, and it doesn't boot; while zImage_DVI works | 14:49 |
| lool | I'm really confused | 14:49 |
| ogra | i use dpkg -x on my laptop, boot into redboot serial, do xmodem, load the kernel and run a similar exec line to yours | 14:51 |
| lool | ogra: what exact line are you currently running? | 14:51 |
| ogra | might it be the dd/fis combo you use ? | 14:51 |
| lool | it could be | 14:51 |
| lool | I'm not using the default sizes for the fis partitions | 14:51 |
| ogra | currently i'm using the lates kernel *with* initramfs ... | 14:52 |
| ogra | (initramfs) cat /proc/cmdline | 14:52 |
| ogra | console=ttymxc0,115200 root=UUID=ae90832f-ba0d-4164-b710-0402041ab8ed | 14:52 |
| lool | ogra: Ah *with* initramfs | 14:52 |
| ogra | right | 14:52 |
| lool | I suspect that's the problem | 14:52 |
| ogra | but it works fine without initramfs using yours | 14:52 |
| ogra | no | 14:52 |
| amitk | ogra: I put the new kernel on p.u.c | 14:53 |
| ogra | amitk, thanks, trying | 14:53 |
| ogra | lool, drop the "noinitrd" | 14:53 |
| ogra | and the debug | 14:53 |
| lool | amitk: http://people.ubuntu.com/~amitk/linux-image-FINAL-2.6.28-11-imx51_2.6.28-11.35_armel.deb ? | 14:53 |
| amitk | yes | 14:53 |
| ogra | exec -c "console=ttymxc0,115200 console=tty0 root=/dev/mmcblk0p2 rootdelay=2" should work | 14:53 |
| ogra | (for me exec -c "console=ttymxc0,115200 console=tty0 root=/dev/sdb1 rootdelay=2" definately does if amitk doesnt play with USB modularization ;) ) | 14:54 |
| amitk | ogra: hopefully this time I don't have my head up you-know-where :) | 14:55 |
| ogra | heh | 14:55 |
| lool | ogra: no difference, it stops after "done, booting the kernel." | 14:55 |
| ogra | strange | 14:55 |
| ogra | works fine here | 14:55 |
| lool | ogra: Can you show your fis commands? | 14:55 |
| lool | Or your xmodem load rather | 14:55 |
| lool | RedBoot says before booting: | 14:55 |
| lool | entry=0x90008000, target=0x90008000 | 14:55 |
| lool | Using base address 0x00100000 and length 0x002144c0 | 14:55 |
| ogra | load -m xmodem -b 0x100000 -r | 14:56 |
| ogra | fis create kernel | 14:56 |
| ogra | thats what o do | 14:56 |
| ogra | RedBoot> e -r 0x1000000 -s 4020573 -c "console=ttymxc0,115200 console=tty1 root=UUID=ae90832f-ba0d-4164-b710-0402041ab8ed quiet" | 14:56 |
| ogra | entry=0x90008000, target=0x90008000 | 14:56 |
| ogra | Using base address 0x00100000 and length 0x0021020c | 14:56 |
| lool | ogra: Do you have fis list or boot output? | 14:56 |
| ogra | RedBoot> fis list | 14:57 |
| ogra | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 14:57 |
| ogra | Name FLASH addr Mem addr Length Entry point | 14:57 |
| ogra | RedBoot 0x00000000 0x00000000 0x00040000 0x00000000 | 14:57 |
| ogra | FIS directory 0x00040000 0x00040000 0x0001F000 0x00000000 | 14:57 |
| ogra | RedBoot config 0x0005F000 0x0005F000 0x00001000 0x00000000 | 14:57 |
| ogra | initramfs 0x00060000 0x01000000 0x00400000 0x01000000 | 14:57 |
| ogra | kernel 0x00460000 0x00100000 0x00220000 0x00100000 | 14:57 |
| ogra | RedBoot> | 14:57 |
| lool | ogra: and fis list -d? | 14:59 |
| ogra | RedBoot> fis list -d | 14:59 |
| ogra | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 14:59 |
| ogra | Name FLASH addr Mem addr Datalen Entry point | 14:59 |
| ogra | RedBoot 0x00000000 0x00000000 0x00000000 0x00000000 | 14:59 |
| ogra | FIS directory 0x00040000 0x00040000 0x00000000 0x00000000 | 14:59 |
| ogra | RedBoot config 0x0005F000 0x0005F000 0x00000000 0x00000000 | 14:59 |
| ogra | initramfs 0x00060000 0x01000000 0x003D595D 0x01000000 | 14:59 |
| ogra | kernel 0x00460000 0x00100000 0x0021020C 0x00100000 | 14:59 |
| lool | My datalen is 0x002144C0 | 15:00 |
| ogra | lool, i'm about to xmodem again, if you need anything else, say it now :) | 15:00 |
| lool | ogra: I don't, unless you can be tempted to try my way of updating with fis | 15:01 |
| ogra | i do, but only after i have a working initramfs again | 15:01 |
| ogra | which takes a bit | 15:01 |
| lool | ogra: How do you create the initramfs? by installing the package? | 15:01 |
| ogra | right | 15:01 |
| lool | ogra: Hwo do you load it? | 15:01 |
| ogra | installing the package and scp'ing to my laptop | 15:01 |
| ogra | then i reboot and xmodem it over serial | 15:01 |
| lool | load -m xmodem... wihch args | 15:02 |
| ogra | load -m xmodem -b 0x1000000 -r | 15:02 |
| ogra | fis create initramfs | 15:02 |
| ogra | load -m xmodem -b 0x100000 -r | 15:02 |
| ogra | fis create kernel | 15:02 |
| ogra | exec -r 0x1000000 -s *size* -c *command line* | 15:02 |
| ogra | where *size* is the output of ls -l | 15:03 |
| ogra | for the initramfs file | 15:03 |
| ogra | sillyness of the week, its not in hex | 15:03 |
| ogra | while everything else is | 15:03 |
| lool | I really don't get why it doesn't work for me | 15:04 |
| ogra | try serial | 15:04 |
| lool | ogra: Can you show me the messages before loading linux on the next boot? | 15:04 |
| lool | ogra: I did | 15:04 |
| ogra | check if that changes it | 15:04 |
| ogra | i will | 15:04 |
| lool | Will try serial again | 15:04 |
| ogra | xmodem running now ... takes a while | 15:04 |
| ogra | lool, http://paste.ubuntu.com/133688/ | 15:10 |
| ogra | its intresting how noisy it is even though i use quiet | 15:11 |
| amitk | ogra: no explosions yet? | 15:12 |
| ogra | not yet and i was even brave and just re-used the old initramfs from the former build | 15:12 |
| ogra | rolling a fresh one now to make sure | 15:12 |
| lool | ogra: same thing with xmodem | 15:12 |
| ogra | weird | 15:13 |
| lool | http://paste.ubuntu.com/133690/ | 15:13 |
| lool | But without initramfs | 15:13 |
| lool | It's depressing | 15:13 |
| ogra | hmm, i'm not sure the -b 0x100000 is right without initramfs | 15:13 |
| ogra | look on the wiki | 15:14 |
| lool | It's the same I used in the past and that I use for zImage_DVI | 15:15 |
| ogra | hmm | 15:15 |
| lool | wiki says RedBoot> load -m ymodem -b 0x01000000 -r though | 15:16 |
| ogra | shouldnt differ | 15:16 |
| ogra | oh, wait | 15:16 |
| lool | 0x01000000 != 0x100000 | 15:16 |
| ogra | 0x01000000 is the address used for the initramfs | 15:16 |
| ogra | which needs to be loaded first | 15:17 |
| lool | ogra: Yes, but the wiki mentions that for "an image" | 15:17 |
| lool | I think it's wrong though | 15:17 |
| ogra | so its likely that 0x01000000 gets copied into ram first | 15:17 |
| lool | Anyway, it works fine with other kernels with 0x100000 and you use 0x100000 for kernel as well | 15:17 |
| lool | Oh I know | 15:17 |
| ogra | yes | 15:18 |
| * lool bangs head | 15:18 | |
| lool | PFF | 15:18 |
| lool | ogra: RedBoot version | 15:18 |
| lool | *again* | 15:18 |
| ogra | oh | 15:18 |
| ogra | indeed i use NCommanders | 15:18 |
| lool | All makes sense now | 15:18 |
| lool | ogra: You need 2.6.28 redboot for 2.6.28 kernel and I was running 2.6.26 with its redboot still | 15:18 |
| ogra | indeed | 15:18 |
| lool | Two times I kill almost a day because of redboot version | 15:18 |
| ogra | you said zImage_DVI | 15:18 |
| * ogra killed a day because he didnt have lrzsz installed on the host | 15:19 | |
| ogra | *that* was silly | 15:19 |
| ogra | having a wrong redboot version is rather in the middle of sillyness in taht light imho ... not on the edge | 15:20 |
| amitk | ogra: you're not complaining yet. I am getting worried... | 15:20 |
| ogra | amitk, xmodem*ing initramfs atm | 15:20 |
| ogra | let me reboot | 15:20 |
| ogra | and then hammer the system a bit | 15:20 |
| ogra | amitk, cross your fingers, booting now | 15:27 |
| * amitk crosses fingers, hands, legs | 15:27 | |
| ogra | boot looks fine for a start | 15:27 |
| ogra | gdm is up ... | 15:28 |
| ogra | oh, usb errors in demsg | 15:29 |
| ogra | *dmesg | 15:29 |
| ogra | [42949494.010000] usb-storage: -- transfer complete | 15:29 |
| ogra | [42949494.010000] usb-storage: Bulk command transfer result=0 | 15:29 |
| ogra | [42949494.010000] usb-storage: usb_stor_bulk_transfer_sglist: xfer 4096 bytes, 1 entries | 15:29 |
| ogra | [42949494.010000] usb-storage: Status code 0; transferred 4096/4096 | 15:29 |
| ogra | [42949494.010000] usb-storage: -- transfer complete | 15:29 |
| ogra | [42949494.010000] usb-storage: Bulk data transfer result 0x0 | 15:29 |
| ogra | [42949494.010000] usb-storage: Attempting to get CSW... | 15:29 |
| ogra | [42949494.010000] usb-storage: usb_stor_bulk_transfer_buf: xfer 13 bytes | 15:29 |
| ogra | [42949494.010000] usb-storage: Status code 0; transferred 13/13 | 15:29 |
| ogra | [42949494.010000] usb-storage: -- transfer complete | 15:29 |
| ogra | [42949494.010000] usb-storage: Bulk status result = 0 | 15:29 |
| ogra | [42949494.010000] usb-storage: Bulk Status S 0x53425355 T 0x176e R 0 Stat 0x0 | 15:29 |
| ogra | [42949494.010000] usb-storage: scsi cmd done, result=0x0 | 15:29 |
| ogra | [42949494.010000] usb-storage: *** thread sleeping. | 15:29 |
| ogra | over and over | 15:29 |
| amitk | sounds like issues with their usb driver | 15:29 |
| ogra | sounds like NCommanders issue | 15:30 |
| amitk | mcasadevall: could you file a bug regarding your usb errors and tag it arm? | 15:31 |
| amitk | ogra: does it prevent us from going further? | 15:31 |
| ogra | let me see if it survives | 15:31 |
| mcasadevall | amitk, sure, I think its a issue with my network adapter, since the USB storage seems to be okish | 15:31 |
| ogra | i seem to be able to run a dist-upgrade getting 30M | 15:32 |
| ogra | tail -f /var/log/messages doesnt show any probs atm | 15:32 |
| ogra | hmm, now i see a bunch of usb-storage related messages | 15:33 |
| giedz | hi...what is default window manager in ubuntu-arm port? | 15:33 |
| ogra | giedz, any WM you like ... the architecture has noting to do with the CPU | 15:33 |
| ogra | s/architecture/desktop/ | 15:33 |
| giedz | ok but is there any prefered? | 15:34 |
| ogra | amitk, i see a ton of messages, but the package installation seems to work regardless | 15:34 |
| ogra | amitk, will capture it for you once the dist-upgrade finished | 15:34 |
| Martyn | afternoon ogra | 15:34 |
| ogra | hey Martyn | 15:34 |
| Martyn | I'm going to have to wait for your drop :) I have a usb network adapter, but no way to use it until there are some kernel modules to load. *chuckle* | 15:35 |
| ogra | Martyn, http://people.ubuntu.com/~amitk/linux-image-FINAL-2.6.28-11-imx51_2.6.28-11.35_armel.deb | 15:35 |
| ogra | that seems to be relatively good (just testing it here) | 15:36 |
| ogra | ARGH | 15:41 |
| * ogra wasnt aware DPMS works on babbage ... | 15:41 | |
| ogra | i just thought my X crashed | 15:41 |
| ogra | :P | 15:41 |
| Martyn | Heh | 15:42 |
| Martyn | DPMS works great in babbage | 15:43 |
| Martyn | I've come to a blank console :) | 15:43 |
| ogra | apparently | 15:43 |
| Martyn | Now .. if we could only coax DVI to work :) | 15:44 |
| ogra | not without a public driver | 15:44 |
| ogra | * Starting AppArmor | 15:45 |
| ogra | * Loading AppArmor module... | 15:45 |
| ogra | ...fail! | 15:45 |
| ogra | pffft | 15:45 |
| amitk | good | 15:45 |
| ogra | the exclamation mark is really overreaction | 15:45 |
| * ogra grabs the source and adds ...fail! (omg, OMG !!!!) to make it look more scary :P | 15:46 | |
| ogra | amitk, still not died, but /var/log/messages looks really worrying | 15:49 |
| ogra | amitk, http://paste.ubuntu.com/133715/ | 15:51 |
| ogra | amitk, and /var/log/messages grew from 10 to 408k since i started the dist-upgrade, its completely filled with these messages | 15:52 |
| amitk | ogra: I'm not too worried. It seems to be restricted to their storage driver. Fixable at a later date. | 15:53 |
| ogra | apart from the fact that we will try to do installs to a USB device with beta i tend to agree | 15:54 |
| lool | ogra: So yeah, it was definitely redboot and only that | 15:56 |
| lool | I'm booting into my system just fine now | 15:56 |
| ogra | cool | 15:57 |
| lool | With VGA and all | 15:57 |
| ogra | i just had a reboot notification after dist-upgrade :) | 15:57 |
| ogra | so that works as well :) | 15:57 |
| ogra | do you have a binary for fis ? | 15:57 |
| ogra | so i can try to update my SD from the running system | 15:57 |
| Martyn | ogra : I've got errors loading the modules, I'm afraid | 15:59 |
| Martyn | ogra : 'invalid module format' | 15:59 |
| lool | ogra: Sure | 16:00 |
| lool | ogra: I'm mostly running it from my desktop though | 16:00 |
| ogra | then your initramfs isnt in sync with your kernel | 16:00 |
| lool | ogra: Let me write a flash-kernel alike script | 16:00 |
| ogra | lool, oh, so no armel build ? | 16:00 |
| Martyn | oh snap | 16:00 |
| lool | ogra: I have one as well, but am not using it much now | 16:00 |
| Martyn | sure, that makes sense | 16:01 |
| lool | ogra: http://people.ubuntu.com/~lool/fis-armel | 16:01 |
| lool | ogra: I updated redboot and kernel with fis on my desktop to boot into 2.6.28 | 16:01 |
| lool | Works fine | 16:02 |
| ogra | but no initramfs yet, right ? | 16:02 |
| lool | No | 16:02 |
| lool | ogra: Needed a working system first | 16:02 |
| lool | ogra: But that's *easy* ;-) | 16:02 |
| ogra | indeed, like me | 16:02 |
| amitk | ogra: would making the various filesystems (ext2, 3, fat, vfat) as modules cause your problems? | 16:02 |
| lool | ogra: Will write a flash-kernel alike script now | 16:02 |
| lool | Unless you like to | 16:02 |
| amitk | ogra: from perspective of updating an image | 16:03 |
| ogra | amitk, once lool has the script ready he just talks about it wonmt | 16:03 |
| ogra | lool, no, feel free | 16:03 |
| * amitk will leave them compiled-in for a little more then | 16:03 | |
| ogra | i'll go on poking on my image creation tool | 16:03 |
| lool | amitk: Why make them modules? | 16:03 |
| ogra | amitk, though arent ext2/3 in the normal ubuntu kernels as well now ? | 16:03 |
| lool | amitk: it's going ot be harder to boot without initramfs, and I don't think it's what we do on i386 | 16:04 |
| ogra | i thought Keybuk had pulled them in for boot speedup | 16:04 |
| lool | Yes | 16:04 |
| ogra | right, our kernel should be as close to the distro kernel as possible | 16:04 |
| ogra | oh "system has a crash report detected" ... | 16:04 |
| ogra | seems apport works too | 16:05 |
| ogra | lool, do you also have the prob that the vga picture is slightly shifted to the top on the screen ? | 16:06 |
| ogra | it swallows the upper third of the top panel here and i cant get it moved down | 16:06 |
| lool | ogra: My LCD autoadjust | 16:06 |
| lool | ogra: I'm running console ATM | 16:06 |
| ogra | (under X) | 16:06 |
| ogra | ah | 16:07 |
| ogra | console is fine here | 16:07 |
| Martyn | ogra : Are you running from flash, or USB? | 16:07 |
| Martyn | SD or USB I should say... | 16:07 |
| ogra | its just the desktop ... about 5-6px shifted to the top | 16:07 |
| ogra | Martyn, usb atm | 16:07 |
| lool | console is fine here too | 16:07 |
| lool | I'm using SD | 16:07 |
| ogra | both should work fine | 16:08 |
| ogra | mcasadevall, so you just copied the iso to a USB key ? | 16:22 |
| mcasadevall | ogra, pretty much. cdrom-detector didn't find it, but I manually mounted it | 16:23 |
| ogra | what means "pretty much" ? | 16:23 |
| mcasadevall | But its stuck ATM ... I dunno, I think the USB stack on my board leaves something to be desired w.r.t. to stability | 16:23 |
| ogra | mount usb key, loop mount iso ... cp iso/* usb-key/ ?? | 16:24 |
| ogra | or what exactly did you do ? | 16:24 |
| mcasadevall | No, I did that | 16:24 |
| mcasadevall | But once in the installer environment | 16:24 |
| mcasadevall | mount /dev/sda1 /cdrom | 16:24 |
| mcasadevall | "Load Installer Compontents from CD" | 16:24 |
| mcasadevall | │ This partitioner doesn't have information about the default type of │ | 16:24 |
| mcasadevall | ┌│ the partition tables on your architecture. Please send an e-mail │ | 16:24 |
| mcasadevall | ││ message to debian-boot@lists.debian.org with information. │ | 16:24 |
| mcasadevall | heh :-) | 16:24 |
| mcasadevall | so far so good | 16:24 |
| ogra | how did you get *in the installer environment* | 16:24 |
| ogra | would be nice to have an alternate build you can just dd to SD | 16:25 |
| ogra | i.e. with the right boot mechanism in place | 16:25 |
| mcasadevall | ogra, oh, loaded the RAMdisk from RedBoot | 16:25 |
| ogra | right | 16:25 |
| ogra | thats what i mean, we should have an alternate image that has redboot, kernel and initramfs ready to dd it to SD | 16:26 |
| mcasadevall | That's the idea | 16:26 |
| mcasadevall | Its easy to do extactly that. | 16:26 |
| ogra | so you just need to grab the iso/img file | 16:26 |
| mcasadevall | But I wanted to see if the damn thing would work :-) | 16:26 |
| mcasadevall | partman is going to need some work | 16:27 |
| ogra | talk to cjwatson for that | 16:28 |
| ogra | we'll likely need that for ubiquity too | 16:29 |
| mcasadevall | damn it | 16:29 |
| mcasadevall | "Failed to determine toe codename of the release" | 16:29 |
| mcasadevall | So there are a few bugs to work out. | 16:29 |
| mcasadevall | Actually | 16:30 |
| mcasadevall | We got an anonyingly large issue | 16:30 |
| mcasadevall | "Unknown armel subarchitecture: unknown" | 16:30 |
| mcasadevall | Obviously thats something in the filesystem MIA | 16:30 |
| mcasadevall | ogra, so libdebian-installer requires re-education :-) | 16:36 |
| ogra | right the subarch thing is somethng i talked with colin about before but didnt file a bug yet and didnt approach him further about | 16:40 |
| amitk | ogra: btw, are you using aufs now or unionfs? | 16:40 |
| ogra | amitk, heh, neither, i'll test that soon | 16:41 |
| ogra | my usb disk is ext3 | 16:41 |
| Martyn | ogra : Ah! So you did a full install... | 16:41 |
| ogra | (sorry, i'm in a short 1:1 meeting atm) | 16:41 |
| Martyn | I'm building a rootfs right now | 16:41 |
| Martyn | ogra : no worries. | 16:42 |
| ogra | Martyn, right i built a rootfs using my script and untarred it to a usb HD | 16:42 |
| Martyn | lool : Do you have a matching initrd to go with that kernel? | 16:42 |
| ogra | and use an SD to boot that | 16:42 |
| Martyn | I'm doing much the same, but using a usb enclosure with a 7200 rpm drive. | 16:43 |
| ogra | me too | 16:44 |
| ogra | :) | 16:44 |
| lool | Martyn: will soon | 16:44 |
| ogra | a kernel compile takes less than 1h with all modules :) | 16:44 |
| ogra | on that fast disk | 16:44 |
| Martyn | *nod* | 16:46 |
| Martyn | I tried a 10k RPM drive, but the enclosure was just too frigging loud | 16:46 |
| Martyn | It sounded like a small airplane about to take off | 16:46 |
| ogra | heh | 16:46 |
| Martyn | So I switch it out for a WD eco-drive/green drive. 1Tb, 7200rpm, 32mb cache | 16:46 |
| Martyn | most of the time it sits idle, and that's perfect | 16:47 |
| ogra | heh, i have the same setup, but a maxctor drive | 16:47 |
| ogra | *maxtor | 16:47 |
| Martyn | ogra : Hey, I just rebuilt the initrd, and still have the module format error. *grump* | 16:47 |
| ogra | did you transfer it to the SD ? | 16:47 |
| Martyn | yep | 16:47 |
| ogra | weird | 16:48 |
| Martyn | Can I borrow yours and compare? | 16:48 |
| Martyn | I must be doing something wrong | 16:48 |
| * ogra looks for an envelope to send his | 16:48 | |
| ogra | :P | 16:48 |
| Martyn | *groan* | 16:48 |
| ogra | oh, you mean the initramfs | 16:48 |
| Martyn | No, not the SD .. the initramfs | 16:48 |
| ogra | i thought my SD | 16:48 |
| Martyn | yep :) | 16:48 |
| Martyn | and if I wanted to borrow your sd, all you'd have to do is dd of=myfile anyway :) | 16:49 |
| mcasadevall | ogra, so I'm pushing my branch to LP now, and I'll have to start looking at libdebian-installer ;.; | 16:49 |
| Martyn | mcasadevall : It doesn't fail gracefully eh? unknown subarchitecture causes a crit fail? | 16:49 |
| ogra | mcasadevall, make sure do coordinate with #ubuntu-installer | 16:50 |
| mcasadevall | ogra, I am. I'm popping back and forth between the two channels. | 16:50 |
| ogra | Martyn, http://people.ubuntu.com/~ogra/arm/babbage/initrd.img-2.6.28-11-imx51 and http://people.ubuntu.com/~ogra/arm/babbage/vmlinuz-2.6.28-11-imx51 | 16:50 |
| * ogra goes for a break | 16:51 | |
| Martyn | ogra : Got 'em | 16:52 |
| ogra | amitk, aufs looks ok so far, i need to wait for tomorrows build to actually test it since i need the updated procps in the squashfs | 17:49 |
| amitk | ogra: ok. I've pushed the changes to git and build-tested. rtg has done the same. So we might have a good kernel tomorrow. | 17:50 |
| ogra | yay | 17:51 |
| ogra | and with the fixed procps even a live image that can be used | 17:51 |
| amitk | ogra: is imx51 live image created automatically now? | 17:53 |
| ogra | not yet, no | 17:53 |
| ogra | still some missing bits and pieces, i have the skeleton work done to rolla partitioned image and copy the livefs in | 17:53 |
| ogra | tomorrow evening i sould have a working script | 17:54 |
| ogra | creating one manually on top of my old image is easy though | 17:54 |
| ogra | just copy the latest livecd image content into the second partition, pull vmlinuz and initramfs from http://people.ubuntu.com/~ogra/arm/babbage/ and replace them via serial | 17:55 |
| amitk | ogra: would be nice to have an installable image | 17:56 |
| amitk | eventually | 17:56 |
| ogra | yeah, indeed | 17:56 |
| ogra | that might only happen post beta though | 17:56 |
| ogra | we dont know how ubiquity behaves yet | 17:57 |
| ogra | and its a bit tricky since you will need an usb SD cardreader to install to, or we need to trash the install media | 17:57 |
| Martyn | even if we don't have an installer, the rootfs builder provides a system that is just about the same when finished | 18:15 |
| Martyn | So short term, you end up with a working desktop | 18:15 |
| Martyn | (well, working-ish) | 18:16 |
| ogra | indeed, but release target is an installable image | 18:16 |
| ogra | preferably a live image | 18:16 |
| Martyn | abolutely! | 18:16 |
| Martyn | I'll take what I can get though. | 18:16 |
| Martyn | Thank you for the initrd + kernel .. that helped a lot | 18:16 |
| ogra | great | 18:16 |
| lool | root@babbage:~# flash-kernel | 18:35 |
| lool | Flashing kernel... done. | 18:35 |
| lool | Flashing initramfs... done. | 18:35 |
| lool | Ok, system still boots without initramfs | 18:36 |
| lool | Now let's see with a proper boot script | 18:36 |
| Martyn | lool : Did you just compile in all the drivers needed to get to console? | 18:38 |
| lool | Martyn: I used the .debs provided above | 18:38 |
| Martyn | mm | 18:40 |
| lool | ogra: What do you put on your initrd aware command line? | 18:43 |
| lool | exec line | 18:43 |
| lool | I mean in the -c part | 18:43 |
| lool | Ah wait, you mentionned cmdline | 18:43 |
| lool | 15:52 < ogra> console=ttymxc0,115200 root=UUID=ae90832f-ba0d-4164-b710-0402041ab8ed | 18:43 |
| lool | nothing | 18:43 |
| ogra | e -r 0x1000000 -s 3927720 -c "console=ttymxc0,115200 console=tty1 boot=casper LIVEMEDIA=/dev/mmcblk0p1" | 18:44 |
| lool | RedBoot> fis load initrd | 18:44 |
| ogra | for the live image | 18:44 |
| lool | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 18:44 |
| lool | Not a loadable image - try using -b ADDRESS option | 18:44 |
| lool | Hmm | 18:44 |
| Martyn | fis load initramfs | 18:44 |
| lool | ogra: did you manage to put it in fis? | 18:44 |
| lool | Martyn: it is called initrd | 18:44 |
| lool | in fis list | 18:44 |
| Martyn | ah | 18:44 |
| lool | Martyn: Probably I used wrong params, happy to hear what I should use | 18:44 |
| lool | RedBoot> fis list | 18:44 |
| lool | Name FLASH addr Mem addr Length Entry point | 18:44 |
| lool | initrd 0x00460000 0xFFFFFFFF 0x00940000 0x00100000 | 18:45 |
| lool | RedBoot> fis list -d | 18:45 |
| lool | Name FLASH addr Mem addr Datalen Entry point | 18:45 |
| lool | initrd 0x00460000 0xFFFFFFFF 0x002A8EE7 0x00100000 | 18:45 |
| ogra | lool, do you use 0x1000000 as base address for it when dumping it into place ? | 18:46 |
| lool | ogra: Ah no, I see what's wrong now | 18:47 |
| lool | mem addr and entry point are reversed | 18:47 |
| lool | I should have entry point == 0xffffff and not mem addr | 18:47 |
| ogra | yeah | 18:48 |
| ogra | initramfs 0x00060000 0x01000000 0x00400000 0x01000000 | 18:48 |
| ogra | thats what i have | 18:48 |
| Martyn | what's the RedBoot command to load in something from serial, so you can flash it? | 18:50 |
| ogra | load -m xmodem -b 0x1000000 -r | 18:51 |
| ogra | fis create initramfs | 18:51 |
| lool | Unfortunately, I can't get the ram address / load address from the fis command | 18:51 |
| ogra | load -m xmodem -b 0x100000 -r | 18:51 |
| lool | I have hardcoded it for now | 18:51 |
| lool | Need to extend the command | 18:51 |
| ogra | fis create kernel | 18:51 |
| lool | Ok, fis load initrd works now | 18:51 |
| Martyn | ah! xmodem :) | 18:51 |
| Martyn | thanks | 18:51 |
| lool | Now let's get the initrd to be picked up | 18:51 |
| ogra | thats tricky | 18:52 |
| ogra | since you need the size in the exec cmd | 18:52 |
| ogra | so with each initramfs update we have toi touch the cmdline | 18:52 |
| lool | ogra: Well I think we shouldn't need the size in the exec command | 18:52 |
| lool | I'd rather *not* touch the cmdline that'd be awful :-/ | 18:53 |
| ogra | you have to | 18:53 |
| ogra | it wont boot without | 18:53 |
| lool | ogra: my thecus doesn't need that and uses redboot | 18:53 |
| ogra | we can use padding and fill up initrd with zeroes | 18:53 |
| ogra | nopt the FSL redboot | 18:53 |
| lool | I can do that | 18:53 |
| lool | (padding) | 18:53 |
| ogra | with padding the size can always be the sanme and we can use a fixed parameter in exec | 18:54 |
| lool | ogra: even with padding, it doesn't set the address on the cmdline in the exec command | 18:54 |
| lool | exec -c "console=ttyS0,115200 root=/dev/ram0 initrd=0xa0f00000,42M mem" | 18:54 |
| lool | is all there is | 18:54 |
| lool | Sorry, that's cut | 18:54 |
| lool | exec -c "console=ttyS0,115200 root=/dev/ram0 initrd=0xa0f00000,42M mem=128M@0xa0000000" | 18:54 |
| lool | That's a full line from the boot_script_data | 18:54 |
| ogra | right, but with debian kernel and initramfs | 18:55 |
| lool | Well with any | 18:55 |
| * ogra tries to boot without -s parameter | 18:55 | |
| ogra | [42949379.640000] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(2,0) | 18:56 |
| ogra | nope | 18:56 |
| lool | ogra: It worked better for me | 18:57 |
| lool | ogra: I can actually boot with exec -r 0x1000000 -c ... but it prints an error during boot and ignores initramfs | 18:57 |
| ogra | heh | 18:57 |
| lool | let me switch console to serial to copy paste it | 18:57 |
| lool | RAMDISK: Compressed image found at block 0 | 18:58 |
| lool | RAMDISK: incomplete write (-28 != 32768) 4194304 | 18:58 |
| ogra | right | 18:58 |
| Martyn | oop | 18:59 |
| Martyn | that's bad | 18:59 |
| ogra | thats what you get if exec doesnt get the size handed over | 18:59 |
| lool | You think that's solved by padding? | 18:59 |
| ogra | padding to 5000000 | 18:59 |
| ogra | and using exec -r 0x1000000 -s 5000000 | 18:59 |
| ogra | i *assume* that fixes it | 18:59 |
| Martyn | well, at least then the boot script will be consistent | 18:59 |
| Martyn | I'll try it. | 18:59 |
| lool | ogra: I'd liket to avoid the size part | 19:00 |
| ogra | yu wont | 19:00 |
| Martyn | I'm uploading a ramdisk now, so it's just a little extra work to overlay the initrd onto a blank file made with dd if=/dev/zero of=initrd.template bs=1k count=5000 | 19:00 |
| ogra | else we need to fix the redboot we have now | 19:00 |
| lool | According to the redboot manual, -s is mandatory | 19:01 |
| Martyn | I wish the RedBoot supported zmodem. (I switched to ymodem) | 19:01 |
| lool | Too bad, I wish redboot would be clever just like for fis create after a load | 19:02 |
| Martyn | what happens if you lie to redboot? | 19:02 |
| Martyn | just tell it -s is 5000000 | 19:02 |
| Martyn | even if the ramdisk is smaller... | 19:02 |
| ogra | right and pad | 19:02 |
| Martyn | without padding | 19:02 |
| Martyn | all it will do is load garbage bytes | 19:02 |
| ogra | oh, that *might* work | 19:02 |
| * ogra tries | 19:03 | |
| Martyn | <-- lazy | 19:03 |
| * Martyn so badly wants to try to bump the serial port up to 230k ... it takes forever to load things at 115k | 19:03 | |
| ogra | [42949378.620000] RAMDISK: Compressed image found at block 0 | 19:04 |
| ogra | [42949378.900000] RAMDISK: incomplete write (-28 != 32768) 4194304 | 19:04 |
| ogra | :( | 19:04 |
| ogra | i guess we need the padding | 19:04 |
| ogra | blocksize seems to be 32768 | 19:10 |
| ogra | so 524288 should be a proper number for padding | 19:11 |
| ogra | (16 blocks) | 19:11 |
| Martyn | is it really that bad to have to put in the length of the image when doing fis create? | 19:13 |
| ogra | fis create can take a lenght ? | 19:15 |
| ogra | oh, right | 19:15 |
| ogra | does that make -s not mandatory anymore ? | 19:15 |
| Martyn | good question | 19:16 |
| Martyn | I haven't tried it | 19:16 |
| Martyn | it take ~4-5m for me to load the initrd via ymodem | 19:17 |
| Martyn | I really wish the onboard NIC was working :) | 19:17 |
| ogra | yeah | 19:17 |
| Martyn | then I could load via network, and bim-boom-bam :) | 19:17 |
| Martyn | almost done with the upload | 19:18 |
| Martyn | fis create is broken | 19:24 |
| Martyn | it's not setting the length, I think | 19:24 |
| * ogra tries something | 19:26 | |
| Martyn | [42949379.550000] Please append a correct "root=" boot option; here are the available partitions: | 19:33 |
| Martyn | [42949379.570000] b300 1956352 mmcblk0 driver: mmcblk | 19:33 |
| Martyn | [42949379.580000] b301 714892 mmcblk0p1 | 19:33 |
| Martyn | heh, so much for mounting a root filesystem from the mmc | 19:33 |
| ogra | RedBoot> fis create initramfs | 19:34 |
| ogra | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 19:34 |
| ogra | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 19:34 |
| ogra | Invalid FLASH image size/length combination | 19:34 |
| ogra | so much for padded initramfs | 19:34 |
| * ogra tries a smaller one | 19:36 | |
| Martyn | btw .. I booted just fine now | 19:36 |
| Martyn | i'm stuck in the initramfs | 19:36 |
| Martyn | but I did get a boot :) | 19:36 |
| ogra | with the padded intiramfs ? | 19:37 |
| Martyn | no padding | 19:37 |
| ogra | but ? | 19:37 |
| ogra | what did you do ? | 19:37 |
| Martyn | oh, nm | 19:39 |
| Martyn | I forgot to write the fconfig | 19:39 |
| lool | So I tried with padding and failed | 19:39 |
| Martyn | it reverted to using your commandline | 19:39 |
| ogra | me too | 19:39 |
| lool | But hex addresses are support for -s just fine | 19:39 |
| ogra | how big did you make it ? | 19:39 |
| lool | I checked the source code and -s is required | 19:39 |
| lool | ogra: 0x00940000 | 19:39 |
| ogra | 4980736 bytes | 19:39 |
| ogra | is what i'm trying atm | 19:39 |
| Martyn | okay, so after the upload, every time, you have to check the size of the initrd then | 19:39 |
| ogra | using the pad script from d-i | 19:39 |
| lool | ogra: I used the same padding as for N2100 | 19:40 |
| ogra | Martyn, right, and thats what we try to get around atm | 19:40 |
| lool | Now searching for kernel stuff | 19:40 |
| ogra | i tried 5013504 before but seems redboot finds that to big | 19:41 |
| Martyn | [42949379.550000] Please append a correct "root=" boot option; here are the available partitions: | 19:42 |
| Martyn | [42949379.570000] b300 1956352 mmcblk0 driver: mmcblk | 19:42 |
| Martyn | [42949379.580000] b301 714892 mmcblk0p1 | 19:42 |
| lool | Oh perhaps my ramdisk is too big | 19:42 |
| Martyn | oops, wrong one | 19:42 |
| Martyn | [42949378.130000] RAMDISK: Compressed image found at block 0 | 19:42 |
| Martyn | [42949378.180000] RAMDISK: ran out of compressed data | 19:42 |
| Martyn | [42949378.190000] invalid compressed format (err=1) | 19:42 |
| ogra | Martyn, right, thats what you get if you give a to big size | 19:42 |
| Martyn | crap | 19:42 |
| ogra | (bigger than the actual initramfs) | 19:43 |
| lool | I'm trying to locate the ramdisk decompression code, the ATAG parsing one can't help us | 19:43 |
| Martyn | The size I gave was only 0x003e000 | 19:43 |
| lool | init/do_mounts_rd.c | 19:43 |
| ogra | so we're attacking from two sides now :) | 19:43 |
| lool | So >-------printk(KERN_ERR "RAMDISK: incomplete write (%d != %d) %ld\n", | 19:44 |
| lool | >------- written, outcnt, bytes_out); | 19:44 |
| lool | written = sys_write(crd_outfd, window, outcnt); | 19:45 |
| lool | => -28 is an error | 19:45 |
| ogra | yes | 19:45 |
| lool | ENOSPC | 19:45 |
| lool | window = kmalloc(WSIZE, GFP_KERNEL); | 19:46 |
| ogra | RedBoot> fis create initramfs | 19:47 |
| ogra | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 19:47 |
| ogra | ... Read from 0x1fee0000-0x1feff000 at 0x00040000: . | 19:47 |
| ogra | Invalid FLASH image size/length combination | 19:47 |
| ogra | GRRR ! | 19:47 |
| Martyn | I got the same | 19:47 |
| lool | WSIZE is 0x8000 | 19:47 |
| lool | out_fd = sys_open("/dev/ram", O_RDWR, 0); | 19:48 |
| lool | Eh PPC and S390 have a spinner when loading the ramdisk | 19:48 |
| Martyn | yep | 19:49 |
| Martyn | it's just so unfair :) | 19:49 |
| Martyn | I say we make a cooler, better progress loading bar, and show 'em | 19:49 |
| Martyn | -snicker- | 19:49 |
| Martyn | ( We hacked the cobalt kernels, back in 2.4, to output to a parallel LCD panel, and show a graphical loading bar .. it was silly as all hell ) | 19:50 |
| ogra | lool, do we really want to get stuck on that now ? | 19:50 |
| Martyn | ogra : no :) | 19:50 |
| ogra | lets just rewrite fconfig | 19:50 |
| Martyn | -laugh- | 19:50 |
| ogra | i know you would like to get around it, but thats the way that waorks atm | 19:51 |
| ogra | *works | 19:51 |
| Martyn | ogra : It's not really worth it, is it? We're not going to be using redboot forever, right? | 19:51 |
| Martyn | just until u-boot gets ported and mature. | 19:51 |
| ogra | hopefully not ... | 19:51 |
| Martyn | ugh | 19:51 |
| ogra | we have to run fconfig anyway to get the UUID in | 19:52 |
| Martyn | I can't get the initrd to load | 19:52 |
| Martyn | I tried specifying the -precise- -s length | 19:52 |
| Martyn | okay, something's funny | 19:52 |
| ogra | make sure to load it first | 19:52 |
| Martyn | I did | 19:54 |
| Martyn | loads fine | 19:54 |
| Martyn | but when the kernel boots, it's not there | 19:54 |
| Martyn | or at least, I'm getting a vfs error | 19:55 |
| ogra | whats your current exec line ? | 19:55 |
| Martyn | e -r 0x1000000 -s 4001525 -c "console=ttymxc0,115200 console=tty1 root=/dev/sda1 text" | 19:56 |
| ogra | 4001525 ? | 19:56 |
| ogra | right, thats proper | 19:57 |
| ogra | root=/dev/sda1 might be wrong though | 19:57 |
| ogra | try sdb1 | 19:57 |
| Martyn | what would be taking up /dev/sda? the media card is mmcblk0p1 | 19:57 |
| ogra | and eventually use the UUID | 19:57 |
| ogra | nothing, sdX are no guaranteed names anymore | 19:57 |
| ogra | use UUID | 19:58 |
| ogra | root=UUID=<your uuid> | 19:58 |
| Martyn | Which means I'll need to write a disklabel with the UUID :) | 19:58 |
| ogra | no | 19:58 |
| ogra | you have a uuid already | 19:58 |
| ogra | its created if you format the partition | 19:58 |
| ogra | at the initramfs prompt: ls -l /dev/disk/by-uuid/ | 19:59 |
| lool | ogra: Updating the config is really painful | 20:00 |
| ogra | lool, we have to do it anyway | 20:00 |
| lool | ogra: Why so? | 20:00 |
| Martyn | got it | 20:00 |
| lool | mxc_ipu mxc_ipu: VSyncPre occurred before DI1 disable | 20:00 |
| ogra | to get the device uuid into the append string after install | 20:00 |
| ogra | lool, thats your display being switched on and off | 20:01 |
| lool | ogra: indeed | 20:01 |
| ogra | i would propose we concentrate on a better way to write fconfig | 20:01 |
| lool | ogra: Ok; it's enough for me for today though | 20:02 |
| ogra | instead of trying to hack the current working boot methor | 20:02 |
| ogra | *d | 20:02 |
| ogra | lool, same here | 20:02 |
| lool | ogra: I don't think it's hack if we get it to work sanely | 20:02 |
| lool | I consider changing the size on each upgrade dangerous | 20:02 |
| ogra | i doubt we can | 20:02 |
| ogra | we will only end up with a hardcoded size or something i suspect | 20:02 |
| lool | I don't think it's as bad as it could kill the flash, but it's risky | 20:02 |
| ogra | which isnt better | 20:02 |
| lool | ogra: i don't mind a hardcoded size | 20:03 |
| lool | It's not pretty, but it's not risky on upgrades | 20:03 |
| ogra | i would neither if it was easy ... i.e. through padding | 20:03 |
| ogra | but that apparently doesnt work | 20:03 |
| ogra | so we would have to dive deeply into redboot | 20:04 |
| ogra | and hack it up | 20:04 |
| ogra | i honestly prefer to rather find a sane way to update fconfig | 20:04 |
| lool | ogra: I dived into redboot *already* | 20:04 |
| ogra | ok | 20:04 |
| lool | We can't do anything with stock redboot | 20:04 |
| lool | Unless we change it | 20:04 |
| lool | I'm looking at the kernel now | 20:04 |
| lool | My hope is that I can find a magic byte which stops the decompression | 20:05 |
| Martyn | -s is very, very nonoptional | 20:05 |
| lool | Another thing which would be worth trying is whether we can limit the initrd size at another level | 20:05 |
| Martyn | booted to console, went with busybox rather than ubuntu rootfs | 20:05 |
| lool | Martyn: I checked the redboot sources and it's not optional | 20:05 |
| ogra | we could fill up with zeros *before* compressing | 20:05 |
| lool | packages/hal/arm/arch/current/src/redboot_linux_exec.c | 20:05 |
| lool | search for ramdisk_size | 20:05 |
| Martyn | ogra : They would be compressed away | 20:06 |
| lool | And that's why I can assure you it supports hex, and hex worked for me | 20:06 |
| ogra | lool, nonoptional == not optional ;) | 20:06 |
| Martyn | hex works for me just fine as well | 20:06 |
| ogra | sure | 20:06 |
| ogra | but what does hex gain us here | 20:06 |
| Martyn | nothing really | 20:06 |
| ogra | right | 20:06 |
| lool | ogra: I know, I said it earlier as well; I'm just repeating it because you seem to be doing the same research | 20:06 |
| ogra | no, i researched padding | 20:07 |
| Martyn | I'm surprised, though, that it doesn't simply scan or set the length of the ramdisk image somewhere | 20:07 |
| Martyn | rather than forcing the user to set it manually. There must be a reason... | 20:07 |
| lool | ogra: hex> just replying to the fact that it was said to be broken, and it's not | 20:07 |
| Martyn | fis has all the smarts to load it, and stuff the value somewhere... | 20:07 |
| ogra | the thing is, if we fill the cramfs with zeros before zipping it it might probably work | 20:07 |
| lool | Martyn: Yes, that's why I checked as well; but it has not | 20:07 |
| ogra | not sure, but possible | 20:07 |
| lool | We could implement it, but I don't want to rely on our redboot to be present | 20:07 |
| lool | ogra: I did that already | 20:08 |
| Martyn | Well, then we have to write the config each time, and there's no getting around it | 20:08 |
| ogra | lool, padding during compression ? | 20:08 |
| lool | ogra: I created a 0x00940000 file with intiramfs contents + zeroes, checksummed that | 20:08 |
| Martyn | every time the initrd and kernels are updated, the fconfig boot script likewise has to be updated | 20:08 |
| Martyn | UNLESS | 20:08 |
| Martyn | you want to do the two-kernel tango like they do in beowulf | 20:08 |
| Martyn | load a kernel, that chainloads the new kernel and initrd from disk | 20:08 |
| lool | ogra: Ah no, I didn't understand what you meant; I don't see how you'll get exactly the good size though | 20:09 |
| ogra | lool, right, but you likely padded zeros to the end of the already compressed file | 20:09 |
| ogra | lots of math ... | 20:09 |
| ogra | you only need to know the two values ... before and after | 20:09 |
| ogra | then pad cramfs to a certain size ... after gzip you should have the right final size | 20:09 |
| lool | I think it might be enough to pass the initrd size to the kernel to stop decompression at EOG | 20:09 |
| lool | EOF | 20:09 |
| ogra | in the -c command you mean ? | 20:10 |
| lool | Yes | 20:10 |
| ogra | but that still means we need to run fconfig | 20:10 |
| lool | No, I think it works with a zero padded initrd if we pass its size | 20:11 |
| Martyn | I don't think you can get away from that one | 20:11 |
| ogra | no matter where you have to add the exact size in exec ... you have to submit it | 20:11 |
| lool | ogra: it's just a cut size | 20:13 |
| lool | That's my N2100 exec line: | 20:13 |
| lool | exec -c "root=/dev/ram0 rw rootfstype=ext2 initrd=0x01000000,10M noirqdebug mem=32M@0x00000000 console=ttyS0,115200n8" 0x01d00000 | 20:13 |
| lool | The 10M is just to stop the unzip I think | 20:13 |
| ogra | right, but exec still doesnt know what to do | 20:17 |
| lool | ? | 20:17 |
| ogra | the -s tells it to load initramfs, no ? | 20:18 |
| lool | Yes | 20:19 |
| ogra | so how do you tell exec what to do ? | 20:19 |
| lool | Oh I still need the -s for bababge | 20:19 |
| lool | I was pointing at initrd=0x01000000,10M | 20:19 |
| ogra | right | 20:20 |
| Martyn | ogra, lool's command line works! | 20:21 |
| Martyn | I just did it | 20:21 |
| Martyn | no -f, no -s | 20:21 |
| Martyn | just -c | 20:21 |
| lool | Err 0x01d00000 is in the initrd | 20:21 |
| lool | How come | 20:21 |
| lool | Martyn: With the 0x01d00000? | 20:21 |
| Martyn | no, without it | 20:21 |
| lool | Ok | 20:21 |
| Martyn | my line was -> | 20:21 |
| lool | Martyn: For me it works, but the initrd isn't run | 20:22 |
| Martyn | e -c "console=ttymxc0,115200 initrd=0x1000000,10M console=tty1 root=/dev/sdb1 text" | 20:22 |
| Martyn | hmmm | 20:23 |
| Martyn | [ 0.000000] Memory policy: ECC disabled, Data cache writeback | 20:23 |
| Martyn | [ 0.000000] INITRD: 0x01000000+0x00a00000 extends beyond physical memory - disabling initrd | 20:23 |
| Martyn | Interesting, I get the same message | 20:23 |
| lool | I think you need to pass mem | 20:23 |
| Martyn | what's the appended 0x01d00000? | 20:23 |
| lool | It's the kernel entry point for N2100 | 20:24 |
| lool | 0x00a00000 us lel start | 20:24 |
| lool | is mem start | 20:25 |
| Martyn | setting the mem= parameter had no impact | 20:26 |
| lool | It hangs for me if I set mem | 20:27 |
| ogra | here too | 20:27 |
| ogra | it doesnt uncompress | 20:28 |
| Martyn | I did mem=512M, and skipped the @ | 20:28 |
| ogra | though i assume we overwrite something | 20:28 |
| Martyn | is initrd=0x01000000 REALLY what we mean? | 20:28 |
| Martyn | that's where fis put it on the flash, but that's not where it got reloaded in ram, right? | 20:28 |
| ogra | right | 20:29 |
| lool | In the case of my thecus it matches | 20:29 |
| ogra | anyway, i'm really exhausted | 20:30 |
| Martyn | me too | 20:30 |
| lool | You can't set mem alone without setting the start address | 20:30 |
| lool | start = memparse(*p, p); | 20:30 |
| lool | if (**p == ',') { | 20:30 |
| lool | size = memparse((*p) + 1, p); | 20:30 |
| lool | phys_initrd_start = start; | 20:30 |
| lool | phys_initrd_size = size; | 20:30 |
| Martyn | lool : Um, I do that all the time on the x86 platform... | 20:30 |
| Martyn | heh | 20:30 |
| ogra | but not to define initrd size | 20:31 |
| lool | Martyn: that's arm specific though | 20:31 |
| lool | linux/arch/arm/mm/init.c | 20:32 |
| Martyn | http://clug.open2space.com/node/10 | 20:32 |
| Martyn | oops, sorry | 20:33 |
| * ogra calls it a day | 20:35 | |
| ogra | my brain hurts and i dont see us getting anywhere by poking in the dark | 20:35 |
| Martyn | nope. I'll look at redboot tonight, now that I've unpacked the fsl-redboot source | 20:36 |
| * ogra thinks its better to attack that after a relaxing night of sleep | 20:36 | |
| Martyn | I'm a few hours behind you, so I'm fresher | 20:36 |
| Martyn | it's 15:30 here | 20:36 |
| ogra | note that we ise redboot-mxc from the archive | 20:36 |
| ogra | *use | 20:36 |
| Martyn | oh! | 20:36 |
| Martyn | is it downloadable? | 20:36 |
| ogra | (which actually is the same) | 20:37 |
| ogra | indeed | 20:37 |
| Martyn | URI? | 20:37 |
| ogra | its in universe atm | 20:37 |
| ogra | apt-get source redboot-mxc | 20:37 |
| Martyn | ports, universe? | 20:37 |
| Martyn | gotcha | 20:37 |
| ogra | err | 20:37 |
| ogra | redboot-imx | 20:37 |
| ogra | sorry, to tired | 20:38 |
| Martyn | no worries | 20:38 |
| Martyn | downloading, and I'll take a read in a bit | 20:38 |
| ogra | you need to build it on the babbage though | 20:38 |
| Martyn | I have a working-ish boot now, I can probably do it | 20:38 |
| Martyn | but really, I should start poking through the source and understanding the exec code | 20:39 |
| Martyn | and the load code | 20:39 |
| * ogra wonders what packages/redboot/current/src/fs/e2fs.c is :P | 20:44 | |
| lool | ogra: I told you rebdoot supported filesystems! :) | 20:46 |
| ogra | lets look at that tomorrow :) | 20:47 |
| lool | ogra: Well I don't think our version has it | 20:48 |
| lool | Unless it's a new drop, could be | 20:48 |
| lool | Ah ours has e2fs but not fat | 20:48 |
| ogra | right | 20:49 |
| * lool & | 20:50 | |
| ogra | i just pulled michales packge and found packages/redboot/current/src/fs/e2fs.c | 20:50 |
| ogra | sleep tight lool | 20:50 |
| * ogra is off as well now ... finally | 20:50 | |
| lool | http://people.ubuntu.com/~lool/flash-kernel | 20:51 |
| lool | has the padding ATM | 20:51 |
| ogra | bah, you break my naming scheme | 20:54 |
| ogra | error "Cannot find FIS partition 'initrd'" | 20:55 |
| lool | ogra: what did you use? | 22:15 |
| lool | We could support multiple names easily in this place | 22:15 |
| === Guest74361 is now known as Martyn | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!