**** BEGIN LOGGING AT Sat Sep 04 02:59:57 2010 Sep 04 11:19:39 ah, there we go Sep 04 15:33:07 hi. I just got my sheeva plug and upgraded to 2.6.35.4 (I need cifs). Now I'm stuck at the newbie problem that the kernel doesn't boot. it hangs after "Uncompressing Linux... done, booting the kernel" Sep 04 15:33:36 it correctly finds the boot image and prints the correct information, but it simply doesn't boot Sep 04 15:34:17 I got the kernel from here: http://sheeva.with-linux.com/sheeva/ Sep 04 15:34:45 kasi: go to the Marvell>> prompt and do: printenv bootcmd; printenv bootargs Sep 04 15:34:53 and paste the result here Sep 04 15:38:10 sorry, macos just crashed. stupid ftdi driver Sep 04 15:38:32 lol Sep 04 15:38:59 any idea what might be the reason it can't boot the kernel? Sep 04 15:39:36 you don't have root= pointed at something valid Sep 04 15:39:47 go to the Marvell>> prompt and do: printenv bootcmd; printenv bootargs Sep 04 15:39:53 and paste the result here Sep 04 15:42:11 Tootoot222: http://paste.lisp.org/display/114227 Sep 04 15:43:56 I have the rootfstype set Sep 04 15:44:47 ah, bootargs_root=root=/dev/nfs rw Sep 04 15:44:50 that can't be right Sep 04 15:45:09 that's not actually run though Sep 04 15:45:26 only bootcmd is run at boot, and bootargs is passed as args to the loaded kernel Sep 04 15:46:21 well, it hangs right after 'booting the kernel' Sep 04 15:47:18 kasi: have you run the sheevaplug_installer_v1.0 on your plug yet? Sep 04 15:47:21 i think not Sep 04 15:50:41 Marvell>> printenv console Sep 04 15:50:42 console=console=ttyS0,115200 mtdparts=nand_mtd:0xc0000@0(uboot)ro,0x1ff00000@0x100000(root) Sep 04 15:50:56 there must be something wrong with my bootcmd string Sep 04 15:52:33 see but those things you're printing arn't being used Sep 04 15:52:58 have you made any changes to your install you'd be unwilling to lose? Sep 04 15:53:10 no Sep 04 15:53:19 I can try setting up tftp Sep 04 15:53:23 and boot a kernel via tftp Sep 04 15:53:45 http://www.plugcomputer.org/plugwiki/index.php/SheevaPlug_Installer Sep 04 15:53:51 that will fix up your uboot env ^ Sep 04 15:54:08 the tarbal it has also fixes all the weird problems the default one comes with Sep 04 15:54:16 (ubuntu) Sep 04 15:54:48 but I'll be stuck with the same problem after I upgraded to 2.6.35.4 Sep 04 15:55:19 no it should fix that Sep 04 15:55:38 and you can take the uImage of 2.6.35.4 and rename it to 'uImage' and put it onto the usb stick instead of the one in the installer Sep 04 15:57:07 but then modules might be missing etc Sep 04 15:57:12 I'll try it though Sep 04 15:57:18 thanks for the pointer and the help Sep 04 16:03:23 the newest version of uboot is much faster than the default one as well. Sep 04 17:42:25 hm. the installer does runme.php also work on macos? Sep 04 17:42:53 does runme.php also work on macos? (ignore the mangled sentence above) Sep 04 17:47:13 it should Sep 04 17:47:21 if it doesn't i rewrote it into bash if you like Sep 04 17:48:18 http://sprunge.us/NWVW Sep 04 17:48:53 you can use it the same way as the php one (sudo ./runme.sh nand), or you can add some of the optional flags (runme.sh -h to see the flags) Sep 04 17:49:56 thanks will adapt it. there're some specific for linux kernel in the script Sep 04 17:53:03 hm. i'm missing the openocd/config directory. does it contain anything specific? Sep 04 17:53:14 I have openocd installed in the system via ports Sep 04 17:54:27 well, it contains all of the configuration for openocd to communicate with the sheevaplug Sep 04 17:56:05 yes, on the current install of openocd, the directory is called openocd/interface Sep 04 17:56:25 $ ls uboot/openocd/config/ Sep 04 17:56:25 board interface target Sep 04 17:57:46 /* rant start */ macos is soo ill equipped for embedded stuff, it's amazing! /* rant stop; need to reboot*/ Sep 04 17:57:57 lol Sep 04 17:58:16 if you have an ubuntu live CD/USB you can boot into that quickly to reflash Sep 04 17:58:30 (ubuntu has all the drivers needed by default) Sep 04 17:58:46 nope. I want to have it setup on that machine, because then I can mess around with it. I also want to setup tftpd later Sep 04 17:58:52 thanks to macports, I have openocd etc Sep 04 18:07:51 well it looks like it's doing something Sep 04 18:07:58 I wasn't aware of openocd Sep 04 18:08:48 seems to be an interesting project, and I hope it works reliably than some of the commercial stuff Sep 04 18:17:21 well, the installer didn't seem to work properly. I have the new uboot, but it didn't recover the device: http://paste.lisp.org/display/114232 Sep 04 18:21:39 well, booting from usb didn't really help. still stuck after the 'done, booting the kernel': http://paste.lisp.org/display/114233 Sep 04 18:23:53 if recover1 wasn't set it means it didn't actually flash the varibles Sep 04 18:24:34 looks like it Sep 04 18:25:33 well, I didn't have a installer/uboot-custom.txt file Sep 04 18:25:34 yeah it looks like it burned the default stuff but not the ones from the nand file Sep 04 18:25:45 run it with nand Sep 04 18:25:54 # ./runme.sh nand Sep 04 18:26:04 # ./runme.sh -e nand Sep 04 18:26:08 would save some time Sep 04 18:26:17 (it doesn't burn uboot, since that's already working) Sep 04 18:44:40 interesting. it's working now Sep 04 18:44:48 with the old stuff in the nand Sep 04 18:45:04 really weird, it seems that openocd did only flash uboot, but not the whole image Sep 04 18:45:26 the trick was that my root is /dev/mtdblock2 instead of /dev/mtdblock1 Sep 04 18:45:37 the settings are stored in a dynamicly loaded part of the nand Sep 04 18:45:38 ah Sep 04 18:52:12 working nicely now with cifs. thanks a lot Tootoot222 Sep 04 18:52:16 =) Sep 04 18:52:36 I think I'll get a couple more of these for the lab and the students Sep 04 18:53:01 it's pretty fun trying to install gentoo on it ;D Sep 04 18:53:07 extra-credit assignment there Sep 04 18:54:16 hm. since it has gigabit ethernet, I could hook up a GigE camera Sep 04 18:55:26 i'm going to turn my guruplug into a router/firewall/gateway with it's sexy dual ethernet Sep 04 18:55:53 it's only 4 iptables rules + dhcp server to turn it into a simple router Sep 04 18:57:39 I didn't know the guruplug. are there other plugs as well? Sep 04 18:57:44 I only saw the sheevaplug Sep 04 18:58:13 those are the only 2 globalscale produces atm (they have plans for one with a hdmi output, but no release date afaik) Sep 04 18:58:29 other people make models too Sep 04 18:58:29 uh Sep 04 18:58:39 how much bigger is the guruplug than the sheevaplug? Sep 04 18:58:43 http://www.ionicsplug.com/ Sep 04 18:59:01 the dual 1GB ethernet would be good (one for the camera, one for network traffic) Sep 04 18:59:12 Physical Dimension Sep 04 18:59:12 95mm (L) x 65mm (W) x 48.5 mm Sep 04 18:59:20 of the guruplug^ Sep 04 18:59:59 Physical Dimension Sep 04 18:59:59 110mm (L) x 69.5mm (W) x 48.5 mm (H) Sep 04 19:00:02 of the sheevaplug ^ Sep 04 19:00:08 wow so the sheevaplug is bigger? Sep 04 19:00:12 well, I'm more interested in the board itself. because I would like to use just the board in a small robot Sep 04 19:00:28 two ethernet connectors will make it bigger Sep 04 19:00:56 the board, yes Sep 04 19:01:14 it's not thicker than 2 ethernet ports stacked Sep 04 19:01:15 http://www.globalscaletechnologies.com/t-guruplugdetails.aspx Sep 04 19:01:20 see the pictures near the bottom ^ Sep 04 19:02:43 who makes the Feroceon 88FR131 processor? Sep 04 19:03:55 marvell Sep 04 19:04:18 hm. can't find a processor spec on their site Sep 04 19:04:22 that's the Marvell Kirkwood Sep 04 19:04:31 you'll have better luck searching kirkwood Sep 04 19:06:02 well, can't find anything on that specific one. the guruplug seems to use the 88F6281 Sep 04 19:07:20 kasi: http://www.marvell.com/products/processors/embedded/kirkwood/ Sep 04 19:07:22 hm. they seem to have no AD/DACs on board. sad Sep 04 19:08:05 AD/DAC? Sep 04 19:08:32 armin76: 88FR131 isn't listed there though Sep 04 19:08:38 analog-digital (and reverse) converters Sep 04 19:08:42 ah Sep 04 19:09:03 Tootoot222: the processor isn't sold alone Sep 04 19:09:37 armin76: yes, but I might want to use the whole board and see what's on the processor Sep 04 19:10:05 armin76: for example just route the GPIOs outside the plug Sep 04 19:22:38 Tootoot222: does the guru server not have a JTAG-miniUSB port built in? Sep 04 19:22:51 no, they moved it to an external box Sep 04 19:23:12 which is good if you want to rebrand & sell the guruplug with your custom software Sep 04 19:23:21 you don't want the user to easily be able to mess with your stuff Sep 04 19:23:52 it also saves globalescale money as they don't have to incorperate 100 jtag parts on 100 plugs, you just buy 1 jtag box for 100 plugs Sep 04 19:24:06 I see Sep 04 19:25:55 kasi: all the current "plugs" use a Marvell 88F6281 Sep 04 19:26:02 for example http://www.globalscaletechnologies.com/t-guruplugdetails.aspx#hw_block Sep 04 19:26:28 tinker-f595: my /proc/cpuinfo says "Feroceon 88FR131" Sep 04 19:26:53 now that is just plain strange Sep 04 19:28:08 however, the sheevaplug schematics also list the 88F6281: http://www.plugcomputer.org/data/docs/tech/SheevaPlug%20DevKit%20Reference%20Design-Rev1.1.pdf Sep 04 19:28:27 mine has the same 88FR131 Sep 04 19:28:46 oh I think I know what is going on Sep 04 19:29:13 that is probably the core identifiier. The SOC part is 88F6281 Sep 04 19:29:35 might be. I have rev 1 (v51) Sep 04 19:29:48 kasi: i think it's v5l Sep 04 19:30:08 but, if it's the 88F6281, that means it should have two SATA interfaces on board as well hidden inside the plug Sep 04 19:30:18 let me look at the hardware manual for the 88F6281 Sep 04 19:30:22 have it on my drive Sep 04 19:30:36 kasi on the SOC yes. Sep 04 19:30:45 but not necessarily on the board Sep 04 19:31:14 well, the pins should be there. the question is whether they're accessible or just terminated right at the chip Sep 04 19:31:49 it also then should have two GigE macs Sep 04 19:31:55 (on the soc) Sep 04 19:32:11 if you want access to both sata ports purchase an OpenRD . Much more working space to mess with the board. Sep 04 19:33:41 hm. does the 88F6281 have an mmu? Sep 04 19:34:20 http://www.globalscaletechnologies.com/t-openrdudetails.aspx#extern Sep 04 19:36:28 88F6281 hardware specs are here ( removing the : in the URL so it does not load into IRC clients that support download) http//www.marvell.com/products/processors/embedded/kirkwood/HW_88F6281_OpenSource.pdf Sep 04 19:36:39 never mind that did not work Sep 04 19:36:49 don't worry, I got it. Sep 04 19:37:16 I was trying to avoid spamming everyone here with a pdf download Sep 04 19:37:50 here is the functional specs http://www.marvell.com/products/processors/embedded/kirkwood/FS_88F6180_9x_6281_OpenSource.pdf Sep 04 19:37:59 it doesn't look like it has an mmu Sep 04 19:40:46 chapter 24.2.3 Sep 04 19:41:08 in the Functional Specification Sep 04 19:42:10 explains the MMU Sep 04 19:42:57 interesting, but the MMU never shows up in the diagrams on the processor schematics Sep 04 19:43:29 it just says memory controller Sep 04 19:45:21 because it is hidden in the cryptographic engine ! Sep 04 19:46:15 oh and looking through Marvell docs it turns out I was correct the 88fr131 refers to the core of the SOC and not the SOC part Sep 04 19:48:56 yes, seems to have an MMU. the kernel config also has it enabled. Sep 04 19:54:26 if you are interested in ARM kernel development then http://lists.infradead.org/mailman/listinfo/linux-arm-kernel is worth following Sep 04 20:04:30 kasi: if it didn't had an mmu it wouldn't be able to run linux Sep 04 20:08:05 armin76: MMU-less devices can run Linux with uclibc. Sep 04 20:08:21 For example, the first few generations of iPods lack MMUs. Sep 04 20:08:48 MostAwesomeDude: yeah, well, uclinux Sep 04 20:09:26 and you are saying the early iPods ran Linux? Sep 04 20:10:08 iPodLinux. Sep 04 20:10:18 ah Sep 04 20:10:41 It's kind of a dead project now, but before Rockbox became available for iPods, it was pretty awesome. Sep 04 20:11:16 never been an iPod fan Sep 04 20:11:30 so have not followed that Sep 04 20:12:00 I got one for a birthday when I was in high school. Sep 04 20:12:22 ha Sep 04 20:12:35 a youngster Sep 04 20:12:56 How does one convince a guruplug running the factory-supplied OS to do an fsck on boot? I'm seeing lots of "end_request: I/O error, dev mtdblock0, sector 216\nuncorrectable error :" which don't please be greatly... /forcefsck does absolutely nothing, however. (Or is there something else I should be doing to this filesystem?) Sep 04 20:12:56 Yeah, well. Sep 04 20:14:49 are you using jffs2 or ubifs? Sep 04 20:14:58 Set the needscheck fs flag Sep 04 20:15:34 which file system are you using ? Sep 04 20:18:03 The output of "mount" is unhelpful (to me least) as to what sort of filesystem it is. It just says "rootfs". (this is literally just the guruplug out of the box running for a couple of months, so whatever it runs when globalscale sends it to you....) Sep 04 20:18:15 themill: cat /proc/mounts Sep 04 20:18:39 ubi0:rootfs / ubifs rw,relatime 0 0 Sep 04 20:20:35 So ubifs... which is supposed to never need an fsck. So what should I be doing about the I/O errors I see? Sep 04 20:22:47 http://www.linux-mtd.infradead.org/doc/ubifs.html Sep 04 20:25:16 * themill isn't sure what solutions he is supposed to be finding in that document Sep 04 20:26:16 http://www.linux-mtd.infradead.org/faq/ubifs.html Sep 04 20:26:48 look for informtion on errors also mounting options Sep 04 20:36:17 themill: what do you see just before the error? Sep 04 20:38:53 nothing of relevance. Previous line in dmesg is "usb-storage: device scan complete" (at the end of the boot sequence). There's often a "Buffer I/O error on device mtdblock0, logical block 31" mixed in there too. The files seem to be intact though -- or at least "debsums" reports no errors. (that said, the e-sata drive attached keeps failing with I/O errors too.... unhappy box ;( Sep 04 20:39:36 Found an email thread indicating that may not be a real error but a problem with udev rules Sep 04 20:39:39 http://markmail.org/message/7b3xay5fbgo63xsr Sep 04 20:39:55 http://paste.debian.net/87818/ Sep 04 20:41:35 hmm... I'm trying to mount the e-SATA disk using UUID which sounds like it would provoke these sorts of meta-data complaints perhaps. Weird though. Sep 04 20:43:40 the gist of the email thread is that the mtd devices need to be in an udev exclude rule for /etc/udev/rules.d/60-persistent-storage.rules Sep 04 20:43:54 read the thread carefully Sep 04 20:44:13 you might even want to send an email to it Sep 04 20:44:30 http://markmail.org/message/7b3xay5fbgo63xsr Sep 04 20:44:56 well it is a nice day here so time to go out and play Sep 04 20:45:37 tinker-f595: thanks for your help. "reading carefully" will be a thing for another day but I'll do so. Enjoy your day! Sep 04 20:46:22 and yes, commenting out the e-SATA drive and its UUID line from /etc/fstab seems to make those errors go away. That's just plain weird. Sep 04 22:33:29 armin76: there's linux without MMU Sep 04 22:47:48 http://opensrc.sec.samsung.com/ Sep 04 22:47:54 seems like it was merged into the master tree Sep 04 23:16:20 sweet, the deskstar is big enough to fit a samba sticker **** ENDING LOGGING AT Sun Sep 05 02:59:57 2010