=== Jack87|Away is now known as Jack87 === panda is now known as Guest54912 === daurnima1or is now known as daurnimator === jussi01_ is now known as jussi === michaelh1 is now known as michaelh1|away [07:51] lilstevie: I bought a TF101 (yaaaay!) Should I update to the latest android firmware before attempting to switch it to debian? [08:19] twb: probably not :p you will end up blowing away everything on the fs and need to redo that stuff anyway [08:21] $coworker said he thought it might have important fixes for the keyboard part or something [08:21] (He bought a 16G one from RUC a few months ago, and followed your stuff to put Ubuntu on it last week.) [08:22] well are you going pure linux, or dualboot [08:22] I don't care about android [08:22] cause things like prime include dock firmware updates [08:22] Yeah, I think he is running prime [08:23] to be honest dock firmware updates really are to benefit android [08:23] I mean MAYBE it'd be useful to not blow away android on day one, but I kinda doubt I'll ever use it [08:23] It's not like I ever used windows or vxworks on my netbooks and routers [08:24] * twb thinks -- I probably shouldn't be drinking either [08:27] heh, I don't have android on my device [08:28] Good man [08:41] Bleh, I forgot I have to get pppd working with this 3G dongle before I'm allowed to play with my transformer [08:41] lol [08:43] I have a Pandaboard rev A3. http://www.omappedia.org/wiki/Prebuilt_ubuntu_binaries only mentions up to A2. What to do? [08:49] soren: it doesn't really matter [08:50] Ok. [08:50] soren: https://wiki.ubuntu.com/ARM/OmapNetbook only has one image [08:51] and the only difference between the A2 and A3 is the CPU has been updated for ES2.2 [08:54] I'm installing the headless image on an SD card right now. I don't have a serial cable handy. I don't suppose that image comes with any sort of networked access (ssh or even telnet)? [08:56] that I don't know I don't have a panda, nor have I used the headless image before [08:58] Ok. === Jack87 is now known as Jack87|Away [09:58] OK, got the 3g doodad working, back to transformer [09:58] * twb digs up pile of napkins from last week when we were talking about this === chrisccoulson_ is now known as chrisccoulson [10:07] OK, so step #1 of flashing it is to find a copy of the nvflash ELF binary that can be trusted, i.e. download it myself from nvidia.com rather than "some forum post" [10:07] Is that achievable? AIUI it's only available bundled in their dev board SDK, but I'm happy to download 200MB of zip file to get a trusted version of that one binary. === jussi01_ is now known as jussi [10:15] twb: it is in the L4T pack [10:15] which isn't that big [10:15] Do you have the exact URL on you? [10:15] Otherwise I'll wade through their search page [10:18] FFS, all the ddg.gg hits are 404d [10:18] http://tegradeveloper.nvidia.com/content/linux-tegra-release-12-alpha-1-released [10:18] ty [10:18] driver package has nvflash binary [10:20] Man, 10MB, why do all these forum weenies keep reposting it [10:21] It's like those bt users who put a bunch of .rar's in another rar [10:21] there is a deb of that driver (not sure it will work on your kernel thrugh) [10:22] https://launchpad.net/~ac100/+archive/ppa [10:23] Well, eventually I'd like to run stock Debian armhf kernel with as few device-specific patches as possible, but for now I was just going to git clone lilstevie's tegralinux one and compile that [10:23] twb: well I begrudgingly package it, if I didn't I would have a bunch of people continually complaining that my pack doesn't work [10:24] twb: which one, my repo has 2 :) [10:24] one is L4T driver compatible but only works with u-boot [10:24] Dunno, I was gonna ask you once I was confident I had nvflash working and metastrap was chugging away building a btrfs debian sid armhf rootfs :P [10:24] Oh, I can replace the bootloader with uboot? [10:24] * twb like uboot [10:25] yes, but u-boot is in its infancy [10:25] Meaning that it isn't production-ready for this device? [10:26] it is missing some stuff :p [10:26] I mean, I'd like to minimize my use of non-free software, but at the same time I don't want to brick it or not be able to e.g. use the screen [10:27] APX is always there [10:27] That's the stage 1 bootloader? [10:27] bootrom [10:27] stage0 [10:27] OK [10:28] It would be nice if I could interactively pick which kernel to boot and whether to pass "single" to it, but I'm assuming that's... nontrivial at the moment. [10:29] well you could do that with the standard bootloader [10:29] Oh, OK. I thought that only had two options "normal boot" and "recovery boot" [10:29] by adding single to the recovery kernel :p [10:29] Yeah, that's plan B [10:30] two kernels plus single vs. no single means four cases not two :P [10:30] if we can get the GPIO driver to co-operate with u-boot it would be trivial [10:30] but at the moment it is non-functional [10:30] OK [10:30] That ac100 PPA has a main and restricted in pool/, but only a main in dists/. [10:32] ogra_: doing 'dhclient -r;dhclient wlan0' seems to add the route, RE: network manager not setting the default route for me in oneiric [10:33] weird [10:33] did you file an NM bug ? [10:33] not yet [10:33] was about to do that, only hit me today to try renew the lease [10:34] renewing* [10:45] twb, go to the package details,. thats a launchpad bug (the package has "restricted" in the control file, LP doesnt really have a concept for restricted packages) [10:46] OK [10:46] was just mentioning it fyi [10:47] yeah, its known :( [10:47] for the P release we will hopefully have the driver in the actual archive :( [10:47] err [10:47] :) [10:48] lilstevie: awesome, that version of nvflash is 1) statically linked; and 2) has --help. Much nicer than "doesn't run at all" old version from that I was looking at last time [10:48] (static linking = works without a 32-bit userland) [10:49] twb: hmm, well the one in my package is from that dump [10:50] the other one floating around is the asus one [11:00] OK, this is weird [11:00] I connect the keyboard (dock) 40-pin connector to my laptop, and lsusb can see two devices on it -- both ASUS vendor USB ID, one is a broadcom bluetooth device [11:02] interesting [11:03] Ah, nm [11:03] the bluetooth is just the onboard bluetooth in my netbook [11:04] Gods, that 40pin cable is flimsy [11:11] OK, why isn't this working? [11:11] http://paste.debian.net/128994/ [11:12] Running nvflash as root seems like the coward's way out; I'd rather just have udev make the tf world-writable or so. [11:15] i think nvflash uses raw device access, you need root [11:15] Surely if I have write access to /dev/foo, that is "raw device access" [11:17] The only related superuser capability I can see is CAP_SYS_RAWIO [11:17] And that looks to be for something else [11:17] well, i havent seen anyone getting nvflash to work without root in the 1.5 years i work with nvflash :) [11:17] if yuo find a way, tell me :) [11:18] OK, I did this [11:18] chgrp -Rh twb /dev/bus/usb/* [11:18] Now I get a different error [11:18] USB device not found255 [11:18] intresting [11:18] Maybe I need to plug the 40pin into the table directly instead of the dock? [11:18] * ogra_ never had to modify anything to flash his ac100 [11:19] btw same error running it as root [11:19] plugging and running nvflash just works [11:19] ogra_: running as root I bet tho [11:19] indeed [11:19] I try to be a bit paranoid :P [11:19] i only use nvflash for new ac100's anyway though [11:19] after the first flashing i can flash from userspace [11:20] yeah openwrt is like that too [11:21] lilstevie: do I need to do something special to put the TF into guest most? [11:21] *mode === apachelogger_ is now known as apachelogger [11:22] hmm, there is some little "asus sync" thing on the tf's screen [11:23] ANd another "usb debugging connected" [11:41] Grmph, I'm not having any joy getting nvflash to see it [11:41] Maybe I should read the docs more than just --help [11:44] Hmm, gentoo's tegra2 page says "To power on the board, you need to press *BOTH* the Force Recovery and the Power on buttons until the LEDs power up. Then execute the following command to flash the bootloader." [11:46] OK, booting while holding the volume down gives "safe mode" [11:46] ok back [11:46] sorry disconnected for a sec [11:47] lilstevie: short version: I'm trying to make "./nvflash --getbct" do something useful [11:47] you should get error 0x4 if you invoke nvflash like that :p [11:47] Hum [11:47] yeah you cant [11:47] the tf is an SBK locked device [11:47] you need to use the --sbk option [11:48] and with that you also need to specify --bl --bct and --config [11:48] lilstevie: first of all, do I need to do anything on the TF to put it into "upgrade mode" or anything? [11:49] vol up plus power on boot will put it in APX [11:49] OK [11:49] And I need to do that? [11:50] need to do which? [11:50] put the tablet in APX? [11:50] Do I need to put it into APX mode to use nvflash [11:50] yes [11:51] OK. [11:51] nvflash is for communicating with apx [11:53] goddammit, now it isn't turning on at all [11:54] ok here we go [11:55] It seems like from off I can't just hold both vol+ and power and have it boot [11:56] Do i get any visual feedback when it goes into APX? [11:58] Aha [11:58] Answer is no, the screen is off, but lsusb suddenly sees an nvidia device [11:59] Progress [11:59] ogra_: you can use nvflash as non-root user provided you have write access to the block device. [12:00] which you usually dont :) [12:00] ogra_: but what this means is that you wrie a udev rule to say something like GROUP=disk [12:00] eeek [12:00] Then instead of running it as root, you run it as a non-root user who is in the disk group [12:00] thats more of a security hole than using root [12:01] Yes, well, you know what I mean. [12:01] the disk group should never be used for users [12:01] If you want use polkit or something to hand it to the user on the local screen or whatever [12:01] * ogra_ -> off for a phone conf [12:01] np [12:01] later ogra_ [12:02] lsusb can't see it anymore after doing that bct (which did finally fail with error 4) [12:02] yep that is cause of SBK [12:02] error 0x4 is when the command is invalid [12:03] Does it quit APX on the first command, or the first bad command? [12:03] which in the case of these SBK locked devices that means the command is incorrectly encrypted [12:04] it shuts down USB communication, only on incorrect bootrom commands [12:04] OK [12:04] but when you use an SBK locked device you need to upload the miniloader (built in to nvflash) and bootloader to get interactive [12:05] at which point commands are no longer encrypted anyway [12:05] OK, the SBK in your forum post (ending in 98) works for me, yay. [12:05] awesome :D [12:06] So now I have a working nvflash I need to learn how to use it. [12:06] First goal is to make a complete dump of all the data on there to begin with, just in case I ever want to put it back [12:06] ok, well best thing is to pass it create, and just configure the flash config [12:06] ok ./nvflash -r --download [12:07] Shouldn't I get the partition table first? :-) I don't know /a priori/ what partition IDs there are [12:08] well you could do that :p [12:08] but why reinvent the wheel [12:08] http://androidroot.mobi/2011/06/13/nvflash-on-asus-transformer/ [12:09] Because I trust a dump I make myself more than "what some guy told me" [12:09] Same reason I wanted to get nvflash direct from nvidia.com [12:09] (No offense intended, I'm just a paranoid sysadmin.) [12:09] nono I mean for flash.cfg and bct [12:10] So presumably you can't just say "nvflash, please fetch flash.cfg and bct from my device, and write it to a file" ? [12:10] no [12:10] what you get is a basic information [12:10] OK, then I fall back to plan B, which is using the prepared version you linked to :-) [12:10] which you then need to work on [12:11] and bct is something that is not really easy to get :p [12:11] on device it is encrypted [12:11] Yeah but symmetrically -- you have the shared secret :-P [12:11] yeah [12:11] :p [12:11] Also this is the 32GB version -- does that matter? [12:12] no [12:12] Phew [12:12] flash.cfg is universal [12:12] Does it matter for bct? [12:12] the 0x808 partition ID fills to end of partition [12:12] OK, awesome [12:12] and no bct is config data for the tegra2 cpu not the emmc controller [12:13] Ooooh right [12:13] and er, partition attribute to end of flash* [12:14] I am used to in uboot sheevaplug where IIRC your "partition table" is basically half a dozen disk offsets and to boot you say "jump to block XXX of the MTD and start executing whatever you find there" [12:14] I thought bct was that stuff [12:14] ah [12:14] afk getting caffeine [12:14] thanks for all your help btw [12:14] on sheeva bct is in the SPI IIRC [12:15] cya [12:18] back [12:18] heh that was quick :p [12:19] got a coffee machine in the office [12:20] ah [12:20] Ah, OK, I got bct mixed up with these .cfg files in your linux-flash-kit.tar.gz [12:22] yeah the cfg files are the partition [12:22] Is flash/default.cfg the flash.cfg that matches a brand new tf? [12:22] yes [12:23] You know, it just occurred to me that if nvflash wants to write to a file instead of stdout, I won't have space to write it on my netbook [12:23] I was planning more like dd if=/dev/sda1 | gzip >sda1.orig.gz [12:23] nvflash normally writes out to stdout [12:23] Cool [12:23] oh wait you mean for the backups? [12:24] Yeah [12:24] e.g. --getbct wants me to provide --bct-file or so [12:24] the partition with UDA is the emmc [12:24] OK [12:24] well, the "emmc to android" [12:25] the rest of your backup isn't that big [12:25] OK [12:26] don't need to back up PT, USP [12:26] PT is generated at nvflash --create [12:26] USP is the "staging" partition that blobs are written to from android [12:27] for OTA updates [12:27] Is android using ext3 or 4? [12:27] ext4 [12:27] Because default.cfg says 3, but the licenses page in android on the device, says ext4 [12:27] OK [12:28] it is ext3 for nvflash [12:28] because nvflash doesn't know or understand the difference [12:28] Yeah, I guess nvflash doesn't distinguish. [12:28] Wish I knew what it actually did different for "ext3" vs "basic" [12:28] What lives in MSC? [12:29] tags [12:29] as for ext3 vs basic [12:29] it writes it differently [12:29] basic just does a 1:1 raw write [12:29] but MSC I guess really does not need to be backed up either [12:30] Whereas ext3 only writes active blocks? [12:30] something like that yeah :) [12:30] Yeah, cool. [12:30] still takes a full image, but only writes the ones that have content [12:30] Even if I completely hose the UDA partition, I can still drop into APX and upload a new one, right? [12:30] Because I plan to use btrfs :-) [12:30] even easier [12:30] :p [12:30] UDA is "User DAra" [12:31] haha [12:31] so like in android, drop to CWM and do a factory reset [12:31] :) [12:31] "honolabaru data-san" [12:32] CWM is the rescue partition? [12:32] Like, the rescue mode uses a completelt separate root= filesystem? [12:32] CWM is Clockwork Recovery Mod [12:32] android recovery doesn't use a root= perse [12:32] per se* [12:33] the initrd is / [12:33] and things like /system get mounted atop it [12:33] android doesn't pivot_root [12:33] Christ, that's... unnervin [12:34] I mean I work with some elaborate initramfs's, but nothing like android or splashtop [12:35] well android has its base set of tools in the initrd as well as the init info [12:35] but the android system itself gets mounted into /system [12:35] I certainly don't consider it "normal" [12:36] Silly embedded devs :P [12:36] heh [12:37] the bootimg is really just a zImage squished up with an initrd, + a custom header with some information for the bootloader [12:38] Yeah, that's how devicevm does it for their splashtop images, too [12:38] not all android devices follow that though [12:38] samsung don't [12:38] When you compile the kernel you can even say "my ramdisk is that dir over there, tack it onto the end of the zimage" [12:39] they use an initramfs compiled when you compile the kernel [12:39] using the proper zImage packing method [12:39] because they use a partition called params [12:39] samsung use a u-boot based bootloader though [12:40] So: ASUS are doing it wrong, film at 11 [12:40] param.lfs is an j4fs filesystem with a param.blk which acts as nvram [12:40] At least it's not using a damn graphical EFI BIOS like these new x86-64 motherboards I got last week [12:40] heh [12:40] I'm used to EFI tbh [12:41] all my computers have EFI [12:41] well AppleEFI [12:41] "Sorry, we only support huge yelling bouncing icons, not text. This way is TEH FUTURE" [12:41] I am referring to MSI ClickBIOS [12:41] ew [12:41] need something like rEFIt [12:42] Last time I touched Apple it was still using OpenFirmware, which I still looooove [12:42] OF was win [12:42] I miss OF [12:42] And POWER beats x86-64 [12:42] hands down [12:43] thats why my file server is an xbox360 :p [12:43] I'm not happy about ARM's royalty model either, but they've pretty much won the embedded space [12:43] using the JTAG hack ofc [12:43] ARM may have a shoddy royalty model, but their processors are win in the embedded space [12:43] Yep [12:44] I couldn't imagine going back to MIPS or the likes [12:44] Although I think they should probably ditch thumb and jazelle and stuff and just bake one ISA into any given core [12:44] thumb does have its benefits [12:44] lilstevie: Forth is best for microcontrollers tho [12:44] REPL FTW [12:45] You said MSC is for tags -- what are tags? [12:45] well like bootloader args [12:46] kinda what I was trying to say, but a bit more complex [12:46] they also serve as tags for telling recovery to do certain actions [12:47] most common use though is a file called recovery tells the bootloader to boot straight into recovery [12:49] Oh, OK [12:50] So kinda like a combination of nvram, syslinux.cfg and /forcefsck [12:50] kinda' [12:50] just done very wrong [12:50] Heh [12:51] I hate the bootloader [12:51] One day [12:51] One day we will have coreboot on the PROM [12:51] I'd love to see UEFI :p [12:51] And it will not load seabios or refit, it will load the openfirmware implementation that sun dumped into coreboot just before oracle ate them. [12:52] heh [12:52] EFI... saying "DOS 6 like UI!" as if that's a good thing [12:52] :p [12:52] And it is, man, it even has commands like "dir" [12:53] And you run "if" at the prompt and it says "error: only works in batch scripts" [12:54] heh [12:54] Ruh roh [12:54] ./nvflash --read UDA /dev/stdout --sbk 0x1682CCD8 0x8A1A43EA 0xA532EEB6 0xECFE1D98 | file - # <-- hung [12:56] ok have you already uploaded the bl and stuff? [12:56] Nope [12:56] also by ID I mean the numerical ID [12:56] you need to put it into bootloader update mode [12:57] apx wont listen to many commands when an SBK is set [12:57] Do I do that with ./nvflash --bl --sbk ...; and then it's in bootloader update mode until next reboot? [12:58] Ah, I need a bootloader.bin? [12:59] yes [12:59] and you also need a bct [12:59] I notice the bootloader.bin you supply doesn't match any of the ones in tegra-linux-12.alpha.1.0 [13:00] no [13:00] it is an asus one [13:00] Hum. Can I d/l that from foo.asus.com? :-) [13:01] yes [13:01] in the dlupdate you need to extract a file called "blob" [13:01] Got the URL handy? [13:01] no sorry [13:06] Is it the stuff labelled "firmware", like 200MB dl file? [13:07] yep [13:08] OK [13:10] Damn page needs js, and isn't working even in my fallback js-capable browser :-/ [13:10] :/ [13:17] OK, reverse-engineered their js [13:18] heh [13:21] It's EeePAD/TF101/UpdateLauncher_US_epaduser8659.zip, just need the hostname... [13:22] Which is in http://support.asus.com/js/Download.js [13:23] Bam: http://dlcdnet.asus.com/pub/ASUS/EeePAD/TF101/UpdateLauncher_US_epaduser8659.zip [13:24] * twb closes Xorg again [13:37] lilstevie: FYI, the md5sum of blob also doesn't match yours [13:37] ? [13:38] oh right you are doing an md5 of the entire blob [13:38] :p [13:38] * twb sighs [13:38] Lemem guess another yak to shave [13:38] that blob is 4 or 5 partitions worth od data [13:38] of* [13:38] Oh, stupid me, the blob is the whole thing [13:38] https://github.com/AndroidRoot/BlobTools [13:38] not the whole thing [13:39] Gotcha [13:39] but it is the partitions EBT(bootloader) PT(tegraparts pt) SOS(recovery kernel) LNX(normal boot kernel) at minimum [13:40] also there are multiple versions of the bootloader [13:40] so md5 may not match [13:40] there are at least 7 versions [13:41] I might as well use the latest one from that blob tho, right? [13:41] but only 3 are visually noticeable [13:41] yeah, well, doesn't really matter :p [13:41] only thing with the newest is that to boot from root=/dev/mmcblk1p* you need to add a rootwait [13:42] Hum [13:42] that is the microsd [13:42] Oh, no, I only care about booting from eMMC [13:43] yeah :p [13:43] Except if I completely brick it, in which case SD isn't gonna work either [13:45] you can't completely brick :) [13:45] APX is always there [13:45] k [13:46] Apparently OMAP on the ROM has enough smarts to boot off a FAT16 SD card, no matter what [13:46] Which is pretty nice, even if you have to carefully align the CHS and shite [13:46] *Apparently on OMAP boards the ROM [13:47] tegra2 does too [13:47] well, [13:49] tegra2 can change boot devices [13:50] but in our case the reason it drops back to APX is because the tegra2 only has the eMMC configured as a boot device [13:51] OK, so which blobunpack result is bootloader.bin ? [13:52] blob.EBT [13:52] Ha, so it's not an AmigaOS bitmap font :P [13:52] they are named for the partition code [13:53] so EBT refers to EBT in the flash.cfg [13:53] Nod [13:57] OK, so now I have a bootloader.bin that I can trace back to asus.com, and a bct from "some guy", and I need to load these in using -bl [14:07] twb: not really "some guy" :p the guy who gave the tf root, and who got the sbk for nvflash in the first place :p [14:07]