**** BEGIN LOGGING AT Sun Oct 31 02:59:58 2010
Oct 31 09:47:09 <rabeeh> j #pandraboard
Oct 31 21:41:31 <Majost> Has anyone seen an issue with there plug in which the rootfs is mounted via USB, and when a USB device is connected and fails to init on a different port - the plug tries to reset the entire bus and kills the mounted rootfs?
Oct 31 21:42:19 <Majost> I just upgraded to 2.6.36 this morning -- and I am not sure if this is a kernel regression or if its a hardware limitation
Oct 31 21:53:22 <Tootoot222> Majost: hmm, i've never tried that
Oct 31 21:53:29 <Tootoot222> i assume you have a guruplug plus?
Oct 31 21:56:50 <Majost> yeah
Oct 31 21:57:29 <Tootoot222> what kind of device are you plugging in that fails?
Oct 31 21:57:53 <Majost> It was a laser printer in sleep mode
Oct 31 21:57:54 <Majost> heh
Oct 31 21:58:02 <Tootoot222> lol
Oct 31 21:58:30 <Tootoot222> i actually have / on the nand, but /usr /var and /home on an eSata drive
Oct 31 21:58:36 <Tootoot222> so i couldn't suffer from tha bug
Oct 31 21:58:50 <Majost> yeah, I think I need to do that too
Oct 31 21:59:53 <Majost> I don't mind it resetting the bus -- except for when I am using it for the OS
Oct 31 21:59:54 <Majost> heh
Oct 31 22:01:03 <Majost> http://pastebin.com/7VJhEDet
Oct 31 22:01:12 <Majost> That's what happened
Oct 31 22:01:13 <Majost> heh
Oct 31 22:01:42 <Majost> it seems to have cascaded
Oct 31 22:03:48 <Tootoot222> yeah, it's pretty easy to put your / on the nand, just do a clean shutdown and tar it up, boot it like normal and extract it onto the nand (sudo mount -t ubifs ubi0:rootfs /mnt), partition your drive and edit /etc/fstab
Oct 31 22:05:14 <Tootoot222> untar with tar -xapf root.tar.gz -C /mnt/
Oct 31 22:05:25 <Majost> *nod*
Oct 31 22:05:40 <Majost> I probably can just rsync it hot
Oct 31 22:06:13 <Tootoot222> only if you feel like doing --exclude on /proc, /sys, /home, and half of /dev, but not the other half
Oct 31 22:06:16 <Majost> as long as I exclude /var, /usr, and probably /proc and /dev
Oct 31 22:06:33 <Majost> hrmm
Oct 31 22:06:47 <Tootoot222> there are parts of /dev that are required to boot the system, but other parts that shouldn't be there unless the system is already running
Oct 31 22:06:56 <Majost> ah
Oct 31 22:07:03 <Tootoot222> it's better (imo) just to do a clean shutdown because it brings it to a consistant state
Oct 31 22:07:12 * Majost nods
Oct 31 22:07:36 <Majost> its on a USB stick too, so it probably will be a little easier
Oct 31 22:08:27 <Tootoot222> at a minimum you'll need to put /home on its own partition, and depending on how big the remaining files are, you might need to put /usr
Oct 31 22:08:58 <Tootoot222> also, even though /var probably doesn't contain many (or any) big files, it does contain files that are written to a lot (/var/log anyone?)
Oct 31 22:09:03 <Tootoot222> which will wear out your nand quickly
Oct 31 22:09:21 <Tootoot222> (the nand only surviving for 100.000 writes or something)
Oct 31 22:09:24 <kerio> quickly?
Oct 31 22:09:26 <kerio> wut
Oct 31 22:09:28 <Majost> yeah, I have been building stuff directly on the plug -- so /usr is bloated with -dev packages and whatnot
Oct 31 22:09:29 <kerio> ubifs does wear leveling
Oct 31 22:10:05 <Tootoot222> kerio: iirc each block can only survive 100.000 writes before it breaks
Oct 31 22:10:14 <kerio> yeah
Oct 31 22:10:19 <kerio> and that's why ubifs does wear leveling
Oct 31 22:10:19 <Majost> that seems like a very small number
Oct 31 22:10:29 <Tootoot222> Majost: well there's 2048 blocks
Oct 31 22:10:33 <Majost> I though most nand was like in the 100K+ range
Oct 31 22:10:59 <Tootoot222> 100.000 = 100k
Oct 31 22:11:00 <Tootoot222> =p
Oct 31 22:11:00 <Majost> thought rather
Oct 31 22:11:26 <Majost> oh... I thought that it was one-hundred point zero
Oct 31 22:11:31 <Tootoot222> lol no
Oct 31 22:13:47 <ron> linux can do milliwrites now?
Oct 31 22:14:36 <Tootoot222> milliwrites?
Oct 31 22:14:36 <kerio> yeah, you can write 0.001 of a bit
Oct 31 22:14:42 <Tootoot222> oh
Oct 31 22:14:48 <Tootoot222> wait what?
Oct 31 22:14:55 <Tootoot222> >.<
Oct 31 22:14:57 <kerio> it's an easy way to store more MP3s
Oct 31 22:15:01 <kerio> write them tinier
Oct 31 22:15:07 <kerio> they'll sound weaker though
Oct 31 22:15:20 <kerio> i like my music loud, so i use a byte for each bit
Oct 31 22:15:23 <Majost> how can you write less than a bit?
Oct 31 22:15:32 <Tootoot222> Majost: methinks he's trolling lol
Oct 31 22:15:39 <Majost> heh
Oct 31 22:15:50 <kerio> Majost: Tootoot222 is right in his assumption, you have been trolled quite hard
Oct 31 22:15:53 <kerio> you may want to sit down
Oct 31 22:16:44 <ron> who's trolling?  codec people proudly save things like 0.06 bits all the time!
Oct 31 22:17:14 <ron> entropy is cool like that
Oct 31 22:17:16 <Majost> save != write
Oct 31 22:17:55 <ron> your magnetic disk probably measures its bits to 3 sig figures too ...
Oct 31 22:18:35 <ron> where do you think the NSA gets them back from after you think you erased them ;?
Oct 31 22:21:09 <kerio> ron: go look up "the great zero challenge"
Oct 31 22:21:34 <kerio> if there was a way to write a bit, overwrite it and then read the old one back the HD producers would use it to magically double the capacity of any given disk
Oct 31 22:22:42 <ron> different problems.
Oct 31 22:23:02 <ron> the threshold exists because HD makers need near 100% reliability
Oct 31 22:23:31 <ron> think of how many bits your AM radio can lose before it becomes totally unintelligble
Oct 31 22:23:54 <ron> you wouldn't buy one if they were always that quality, but you can certainly hear what people are saying
Oct 31 22:24:54 <kerio> >radio
Oct 31 22:24:55 <kerio> lol
Oct 31 22:25:31 <ron> well your HD is just a glorified magnetic tape, so the comparison seemed to work :)
Oct 31 22:26:59 <ron> and you only have GBe and USB{2,3}.0 because people were good at splitting bits that other people said were impossible
Oct 31 22:28:53 <ron> (which isn't at all to say that 'data recovery' firms and CAs aren't just total scams :)
Oct 31 22:32:33 <kerio> data recovery firms are just a glorified ddrecover
Oct 31 22:34:40 <ron> even the challenge qualifies it to "people without expensive equipment" -- you won't read it off with the same thing that erased it, otherwise it wouldn't be, well, erased, and your disk wouldn't work anymore
Oct 31 22:35:13 <ron> but I'll bet you can 'overclock' disks to a fair proportion if you have the nous for it ...
Oct 31 22:35:31 <ron> mostly they're probably just too cheap for most people who could to care
Oct 31 22:38:57 <Majost> thats not entirely true... for most data recovery jobs -- they "special equipment" is typically a system which has the ability to bootstrap the disk and act as a local controller for the disk.
Oct 31 22:39:21 <Majost> they basically use the disk as it is for the most part
Oct 31 22:40:19 <Majost> the exception being instances in which a physical head crash has occurred.
Oct 31 22:43:55 <ron> I'd presume people with equipment like that know it works (they payed a lot of money for it after all!) and don't feel the need to prove it to slashdot ;)
Oct 31 22:44:05 <ron> they're probably busy enough already
**** ENDING LOGGING AT Mon Nov 01 02:59:57 2010