LiraNuna | hello! congratulations on ubuntu on N8xx! | 08:16 |
---|---|---|
lool | Well it wasn't really us | 08:17 |
LiraNuna | I know - but still a step forward | 08:17 |
lool | Eh | 08:18 |
lool | I wish we'd have some time to provide OMAP2 kernels | 08:18 |
Stskeeps | morning persia | 10:57 |
persia | Stskeeps, Good morning. | 10:57 |
=== mcasadevall is now known as NCommander | ||
lool | ogra: So we need to fix this fis data | 17:00 |
lool | ogra: You have to test everything yourself; fis or fconfig provide *no* safety net (which is another reason they are a bit suckish) | 17:00 |
lool | ogra: Also, you should pick larger partitions; I can tell you which values I used | 17:01 |
lool | RedBoot: addr = 0x00000000, size = 0x00040000, entry = 0x00000000, length = 0x00000000, cksum = 0x00000000 | 17:02 |
lool | RedBoot config: addr = 0x0005f000, size = 0x00001000, entry = 0x00000000, length = 0x00000000, cksum = 0x00000000 | 17:02 |
lool | FIS directory: addr = 0x00040000, size = 0x0001f000, entry = 0x00000000, length = 0x00000000, cksum = 0x00000000 | 17:02 |
lool | kernel: addr = 0x00060000, size = 0x00400000, entry = 0x00100000, length = 0x001dbdf0, cksum = 0x3defe9b2 | 17:02 |
lool | initrd: addr = 0x00460000, size = 0x00940000, entry = 0xffffffff, length = 0x00940000, cksum = 0xd256cea1 | 17:02 |
lool | ogra: ^ | 17:02 |
ogra | fis -d $IMAGENAME -o 0x40000 -s 0x1F000 create "kernel" -f 0x60000 -l 0x500000 -r 0x100000 -e 0x100000 -c $KERNEL | 17:14 |
ogra | fis -d $IMAGENAME -o 0x40000 -s 0x1F000 create "initramfs" -f 0x560000 -l 0x940000 -r 0x1000000 -e 0x1000000 -c $INITRAMFS | 17:14 |
lool | -rw-r--r-- 1 root root 3913573 Mar 31 14:07 initrd.img-2.6.28-11-imx51 | 17:14 |
lool | -rw-r--r-- 1 root root 1949168 Mar 31 14:05 vmlinuz-2.6.28-11-imx51 | 17:14 |
ogra | http://paste.ubuntu.com/141537/ | 17:16 |
lool | ogra: Actually you miss the load addresses from the above dump | 17:48 |
ogra | do i? | 17:48 |
* ogra checks | 17:48 | |
lool | ogra: Yes, the same data which I miss for ramdisk load address in flash kernel ;) | 17:49 |
ogra | ah, well, might be because i copied your code :) | 17:49 |
lool | ogra: Do you know about the fis list -d switch? | 17:49 |
ogra | ogra@osiris:/var/build/babbage/images$ sudo fis -d /dev/mmcblk0 -o 0x40000 list | 17:50 |
ogra | [sudo] password for ogra: | 17:50 |
ogra | RedBoot: addr = 0x00000000, size = 0x00040000, entry = 0x00000000, length = 0x00029f2c, cksum = 0x2f2287ac | 17:50 |
ogra | FIS directory: addr = 0x00040000, size = 0x0001f000, entry = 0x00000000, length = 0x00000000, cksum = 0x00000000 | 17:50 |
ogra | RedBoot config: addr = 0x0005f000, size = 0x00001000, entry = 0x00000000, length = 0x00000000, cksum = 0x00000000 | 17:50 |
ogra | kernel: addr = 0x00060000, size = 0x00500000, entry = 0x00100000, length = 0x001dea50, cksum = 0xdb979b00 | 17:50 |
ogra | initramfs: addr = 0x00560000, size = 0x002e0000, entry = 0x01000000, length = 0x00452133, cksum = 0x1d051f73 | 17:50 |
ogra | that one you mean ? | 17:50 |
ogra | sure i do | 17:50 |
lool | http://paste.ubuntu.com/141560/ | 17:50 |
lool | What you miss is the "Mem addr" data | 17:51 |
ogra | addr ?? | 17:51 |
lool | flash addr is addr, length is size, datalen is length, entry point is entry | 17:51 |
ogra | i dont see a difference between -d and without .d | 17:52 |
ogra | *-d | 17:52 |
lool | ogra: length versus data length | 17:52 |
lool | ogra: length is the size of the partition, data length is the size of the used data within | 17:52 |
ogra | OOOH!! | 17:52 |
ogra | hmm | 17:52 |
lool | We're padding the initramfs which is why data length == length | 17:52 |
ogra | ogra@osiris:/var/build/babbage/images$ sudo fis -d /dev/mmcblk0 -o 0x40000 list -d | 17:52 |
ogra | Extra arguments after 'list' | 17:52 |
lool | == size | 17:52 |
lool | ogra: It's only in redboot | 17:52 |
ogra | sad that it doesnt work here | 17:52 |
ogra | yeah | 17:53 |
lool | ogra: That's why I was saying it needs patching | 17:53 |
ogra | right, but not critical atm | 17:53 |
lool | ogra: dev="/dev/mmcblk0" => in theory you could get that from find_fs /, but not sure it's worth the risk | 17:55 |
ogra | nah | 17:56 |
ogra | if we support different bootmedia it will make sense | 17:56 |
lool | ogra: To retrieve the config offset, you could copy my fis_info() from flash-kernel | 17:58 |
ogra | fis -d /dev/mmcblk0 -o 0x40000 list| sed -n '/config:/s/^.*addr = \(.*\), size = \(.*\), entry.*$/\1 \2/p' | 17:59 |
ogra | its already in my code | 17:59 |
lool | It's not the same at all | 17:59 |
ogra | just cut off the vals i dont need | 17:59 |
lool | You're using .* instead of 0x([0-9a-f]{8}) | 17:59 |
ogra | because i'm fine with the hex values as they are returned | 18:00 |
lool | I'm also matching the start of the line, I think you could as well "RedBoot config: | 18:00 |
lool | ogra: if you want the 0x, move the parenthesis | 18:00 |
ogra | what do you fear ? | 18:00 |
lool | ogra: But addr = \(.*\), entry.* will match e.g. addr = 0x00060000, size = 0x00400000, entry = 0x00100000, | 18:00 |
ogra | ogra@osiris:/var/build/babbage/images$ sudo fis -d /dev/mmcblk0 -o 0x40000 list| sed -n '/RedBoot config:/s/^.*addr = \(.*\), size = \(.*\), entry.*$/\1 \2/p' | 18:01 |
ogra | 0x0005f000 0x00001000 | 18:01 |
lool | ogra: I fear that if I add ram = in the middle it will break | 18:01 |
ogra | it matches exactly what i need | 18:01 |
lool | ogra: Currently, but if I add the RAM address in the middle of the values, it will break your script | 18:01 |
ogra | how would you add ram in the middle without patching heavily | 18:01 |
lool | ogra: I don't need to patch heavily?! it currently spits addr, size, entry, length, and cksum | 18:02 |
lool | Just changing the order of the values will break in ugly ways | 18:02 |
lool | Because .* can match a full new column, while [0-9a-f]{8} can not | 18:03 |
ogra | ok, thats fine, but if you add ram in the middle your fis_info will break in the same way | 18:03 |
lool | http://people.ubuntu.com/~lool/linux-image-2.6.28-11-imx51_2.6.28-11.38arm3_armel.deb | 18:03 |
lool | ogra: Yes my fis_info will break cleanly :) | 18:04 |
ogra | lol | 18:04 |
lool | ogra: Because it will return no output and that is tested | 18:04 |
ogra | ok ok ... | 18:04 |
lool | ogra: Also you could anchor the start of the expression, that will avoid matching "Backup of my RedBoot config:" :-P | 18:04 |
lool | Yeah I'm pushing hard, but it's easy to do solidly, and we have a good example of what happens when you don't verify everything *cough* | 18:05 |
lool | On the fconfig line, you use "${script}" instead of just "$script"; you seem to use $ without {} the rest of the time so... | 18:20 |
lool | ogra: ^ | 18:26 |
ogra | http://paste.ubuntu.com/141583/ | 18:28 |
ogra | lool, ^^ | 18:28 |
ogra | lool, if that code looks ok to you i would upload it | 18:35 |
lool | ogra: at the phone ATM | 18:37 |
ogra | ok | 18:37 |
lool | echo >> "fis_dev=${fis_dev}" >> /target/etc/flash-kernel.conf | 19:16 |
lool | never saw that syntax | 19:16 |
ogra | err | 19:16 |
lool | it doesn't work | 19:16 |
* ogra slaps forhead | 19:16 | |
ogra | indeed | 19:16 |
lool | ogra: Also, don't rely on >> to create the file, I don't think that's POSIX; touch it first or use cat >/etc/flash-kernel.conf <<EOF | 19:16 |
lool | ogra: (parameter expansion works in here docs) | 19:17 |
ogra | well, we can rely on it not being there, i'll just drop one > | 19:17 |
ogra | here docs look horrible, i try to avoid them if its only some lines | 19:17 |
lool | ogra: I fint a here doc more readable than 3 times ">> /target/etc/flash-kernel.conf | 19:18 |
ogra | http://paste.ubuntu.com/141605/ | 19:18 |
lool | but your call | 19:18 |
lool | ok, that works | 19:18 |
lool | ogra: You can drop the unused () pairs in the sed | 19:18 |
lool | You only use \1 | 19:18 |
lool | ogra: I prefer tty0 than tty1 | 19:19 |
lool | tty0 is the active console | 19:19 |
lool | ogra: Hmm what happens if we don't put anything? | 19:20 |
lool | I think we developers will set it to ttymxc0 and users will rely on the default (tty0 or 1 I guess) | 19:20 |
lool | ogra: I'd double quote $fis_dev in the two fis calls, just in case | 19:21 |
ogra | ok, i'll drop console | 19:21 |
lool | ogra: Otherwise that looks fine; thanks! | 19:21 |
ogra | lool, ok, i'll upload after i fixed the indendation | 19:28 |
ogra | somehow my vi setup is messy | 19:28 |
ogra | lool, you dont use a patch system, right ? | 19:28 |
lool | ogra: A patch system for what? | 19:29 |
ogra | flash-kernel | 19:29 |
lool | No | 19:29 |
ogra | great | 19:29 |
* ogra really prefers it that way | 19:29 | |
lool | The archive has the various uploads I did, which map pretty well to new features | 19:29 |
ogra | right ... and debdiff is still the cleanest way | 19:30 |
ogra | damned .. whats wrong with my vi setup | 19:30 |
ogra | ogra@osiris:/var/build/babbage/images$ echo $TERM | 19:31 |
ogra | linux | 19:31 |
ogra | hmm, might be related | 19:31 |
ogra | silly minicom ... forcxes me to set it to linux | 19:31 |
cooloney | lool, for the bug iop32x initrd, i think we need to build cramfs into kernel. | 20:08 |
cooloney | lool, do you want me to provide a patch to merge and then you test the kernel | 20:08 |
NCommander | rwhitby, ! weclome | 20:22 |
lool | debian/config/armel/config.iop32x:CONFIG_CRAMFS=m | 20:45 |
lool | debian/config/armel/config.ixp4xx:CONFIG_CRAMFS=m | 20:45 |
lool | cooloney: it's weird that it would work for NSLU2 | 20:45 |
lool | ogra: Did you have a chance to test the netboot images for NSLU2? | 20:45 |
lool | ogra: Oh wait, you can't remote load them, right? | 20:45 |
suihkulokki | do you really want a initrd or initramfs? | 20:45 |
lool | suihkulokki: For the target device initramfs, for d-i I don't know | 20:46 |
cooloney | lool, yes, it is weird | 20:46 |
cooloney | but for mounting cramfs as initrd, it should be built into the kernel, right? | 20:46 |
suihkulokki | cramfs doesn't matter for initramfs | 20:47 |
lool | What about CONFIG_ROMFS_FS? | 20:50 |
cooloney | is that romfs? | 20:50 |
lool | debian/config/armel/config.iop32x:# CONFIG_ROMFS_FS is not set | 20:50 |
lool | debian/config/armel/config.ixp4xx:# CONFIG_ROMFS_FS is not set | 20:50 |
lool | debian/config/armel/config.versatile:CONFIG_ROMFS_FS=y | 20:50 |
cooloney | yes, that might be the problem | 20:50 |
lool | It could be either, it's possible the netboot images don't work on NSLU2 either as only the flashable image are used | 20:51 |
lool | Please don't turn any config in NSLU2 though, it's really tight on memory ATM | 20:51 |
cooloney | the initrd is romfs or cramfs | 20:51 |
cooloney | ? | 20:51 |
cooloney | lool, need i turn on CRAMFS or ROMFS? | 20:51 |
lool | CONFIG_ROMFS_FS is for fs/romfs and CONFIG_CRAMFS fs/cramfs; saw no mention in initrd | 20:52 |
lool | cooloney: You're the kernel guy! | 20:52 |
cooloney | lool, i just don't the file system type of the initrd, -:)0 | 20:53 |
cooloney | cause the kernel panic said that it tried several file system type to mount that | 20:53 |
cooloney | but it failed | 20:53 |
cooloney | so it might be missing some file system support built in | 20:54 |
lool | cooloney: I really don't know; perhaps you can check the netboot file? | 20:56 |
cooloney | lool, ok, got it. thx | 20:56 |
lool | It seems it's a gnome-session with a different wm and different apps to load | 20:58 |
* lool & | 20:59 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!