#qi-hardware IRC log for Sunday, 2012-06-10

GNUtooDocScrutinizer05, I guess theses laptops aren't cheap00:01
GNUtoothere are 2 rugged laptops supported by coreboot btw00:01
DocScrutinizer05aqctually quite cheap, on fleabay and refurbished00:01
DocScrutinizer05for a CF-29 in top condition00:02
GNUtoopersonally I'm more interested in the Getac P470 or the Roda RK886EX (Rocky III+) which are supported by coreboot, or simply a lenovo x60/t6000:03
DocScrutinizer05T500 typing here00:03
kristianpaulDocScrutinizer05: getting ready for next deluge?00:31
DocScrutinizer05just preparing to finally sniper poettering00:31
kristianpaulsoon you'll read, booting without paying to microsoft a tax is ilegal :)00:35
DocScrutinizer05>>Due to this, many upstream developers have decided to consider the problem of a separate /usr that is not mounted during early boot an outdated question, and started to close bugs regarding these issues as WONTFIX. We certainly cannot blame them, as the benefit of supporting this is questionable and brings a lot of additional work with it.<< BWAHAHAAAHAAAA does this guy really think we'll buy his lame excuse for lazyness and not 00:39
DocScrutinizer05maintaining his system properly?00:39
DocScrutinizer05A)mount /usr *early*! B) move stuff you need before mounting /usr to /[s]bin C) don't use friggin useless stuff like PA in early boot, for the "questionable benefit" of e.g. playing a powerup jingle with the default PA shite00:42
kristianpaulGNUtoo: x60 is really cheap !00:43
kristianpaulanyway .. :-)00:43
Action: kristianpaul argghh not get distracted again :)00:43
GNUtookristianpaul, it's a computer, and the support is close to 100% complete00:43
GNUtooso I guess you just need to :00:44
GNUtoo1) take information on it00:44
GNUtoo2)buy it00:44
GNUtoo3)install coreboot on it + a distro00:44
GNUtooand you're done00:44
GNUtoono need to hack on it00:44
GNUtooI mean on coreboot00:44
kristianpaulgood deal00:45
GNUtooif you can find one in your area it's a good deal00:47
GNUtooelse it become complicated to buy second hand stuff online etc...00:47
kristianpauli can but x40..00:48
qi-bot[commit] Werner Almesberger: components/: generate for connectors CONN_1 to CONN_40X2 (in gencon.lib) (master) http://qi-hw.com/p/kicad-libs/786407001:30
qi-botThe build has FAILED: http://fidelio.qi-hardware.com/~xiangfu/building/Nanonote/Ben/openwrt-xburst.full_system-20120609-0542 01:51
kristianpaulwpwrak: seems your m1 patches not work well http://paste.debian.net/173758/01:59
kristianpaulhttp://paste.debian.net/173760/ :-|02:02
wpwrakthey should work with rtems 5c51ba1333d96e2ada2c374ba22b551d179e668503:51
wpwrakmaybe something has changed upstream since. that was two months ago03:52
kristianpaul1d179e6685, thats what i need it, thanks !03:59
kristianpauloh dear... http://paste.debian.net/173765/ 04:31
Action: kristianpaul trying with 07896ad5d78af2f47e79c6829e3a57718d660e4404:45
wpwrakthe rtems-yaffs i have its cbe7492ee0e5a9bced8267d9c7ab2fd997299fda05:07
virichow does GPL apply to verilog ? 08:53
viricAll verilog synthesized to a single chip has to be GPL?08:53
lindi-viric: what does it mean that "chip is GPL"?08:54
viricAnd it can be connected only to other GPL chips?   A LGPL verilog would allow connecting it to non-GPL chips08:54
viriclindi-: I also don't know. hehe08:55
viricah, I meant "verilog has to be GPL"08:55
viric(synthesized to a single chip)08:55
viricLGPL allows 'linking' to non-GPL pieces, but only if the LGPL pieces can be replaceable. Hence, a chip appart :)08:56
Action: pabs3 didn't think copyright law applied to chips11:21
rohhm. gpl/lgpl.. maybe even agpl...  i should ask the lawyers about that. thats a real good question12:11
viricroh: I found my OOMK hangs relate to reiserfs. Having moved to ext4 it does not hang anymore, under OOMK conditions12:58
viricroh: nevertheless, when I filled the same-size new ext4 fs with the files I had in the reiserfs, not only ext4 took far more disk...12:59
viricbut I had 92% of inode usage.12:59
Action: pabs3 finds http://en.wikipedia.org/wiki/Integrated_circuit_layout_design_protection13:04
kristianpaulwpwrak: it works now thanks a lot !14:22
wpwrakwas a pleasure to help :) and sorry for the inconvenience.14:34
rohviric: inode-count is set on format.14:58
qi-bot[commit] Werner Almesberger: modules/pads-array.fpd: we need loop for pins and for packages, not just one (master) http://qi-hw.com/p/kicad-libs/20c943614:59
qi-bot[commit] Werner Almesberger: components/adxl32x.lib: Analog Devices ADXL321, ... accelerometers (master) http://qi-hw.com/p/kicad-libs/96c7eb314:59
qi-bot[commit] Werner Almesberger: modules/qfn.fpd: add experimental footprint for AD CP-16-5a* MQ_LFCSP_LQ (master) http://qi-hw.com/p/kicad-libs/e01b8d614:59
rohviric: you can easily make it have more inodes.14:59
viricyes, reformatting, I know15:02
viricI used triple inodes (now ~33% usage) on the new format15:02
rohviric: well.. you did that anyhow. usually if you do not run a ntpd on it one doesnt need extra many inodes15:02
viricBut I also noticed that for the very same files, ext4 needs 1,8GB *MORE* than reiserfs (on a 7,8GB disk. That's 23% of the disk)15:02
rohyou are sure you set reseved blocks the same?15:03
viricI've a tar. I unpack it to a reiserfs, or to an ext4. That's the difference in "df" free.15:03
rohand not dialed around on bytes per inode or inode-size?15:03
viricThe default "mkfs.ext4" used 1,6GB more than reiserfs.15:04
viricWith triple inodes (500k vs 1500k) it uses 1,8GB more than reiserfs.15:04
rohwell... i do not have any clue what your distro uses ad defaults15:04
viricmy distro?15:04
viricI run mkfs.ext4 /dev/blabla15:04
viricwhy would the distro matter?15:04
rohthey all patch stuff/package different defaults15:05
viricI'll check the recipe15:05
viric  configureFlags = "--enable-elf-shlibs --disable-libuuid --disable-libblkid --disable-uuidd --disable-fsck";15:06
viricThat's the only detail out of "./configure; make; make install"15:06
rohviric: i have a 116G filesystem here, which has 7684096 inodes. of that 30G are used (4 linux virtual machines) which is 419283 inodes.15:08
rohso inode used count is direct proportional to space used in mbytes (atleast within 1% error)15:09
rohah. no. its 6% inodes used, to 28% diskspace used. sorry. checked the wrong table here.15:09
viricwell, the number of inodes depends a lot on the number of files you have15:10
viricif you store videos, you'll have few inodes :)15:10
virictheir default guess does not match my usage, it seems15:11
viricand for the disk space used... ext4 looks somehow optimised to big files. Small files take a lot, compare to reisersf (23% of the full fs)15:11
rohviric: thats why one can adjust the settings. no heuristics can guess right in advance15:12
viricdo you have any suggestion of settings?15:14
viricI'd be fine with 1500k inodes...15:14
viricBut I want more free space. I've no idea what to touch.15:14
viricroh: I can paste you dumpe2fs15:18
viricroh:  http://sprunge.us/IRUY15:18
rohviric: how big is that fs in total?15:30
rohhalf a million inodes seems low15:30
viricroh: 8GB15:31
viric(7,8GB, well.)15:31
rohthat seems ok.15:31
rohwhat stuff do you put in there that you need so many small files?15:31
viricOS files mostly15:32
viricbut as far as I understand, I can't make ext4 give me more free space for my use case.15:32
rohviric: still weird. try finding out where you have 'many small files'15:33
viricI know where they are... whether they are a lot or not, I can't tell15:33
viricIt's a matter of taste I imagine15:33
rohviric: as shown above i have very few files per 'OS'15:33
rohthe 419283 inodes are 4 complete ubuntu server vms(12.04)15:33
viricwell, it's a development machine; I have all headers, libs, ...15:34
viricnot only runtime15:34
rohah. i see.15:34
rohwell.. then just tell it do use more inodes15:35
viricin any case, I'm loosing 1,8GB that using reiserfs I'd have free15:35
viricA hardcore dev would just fix reiserfs :)15:35
rohi still dont get where those 1.8g should go. i dont have that here15:36
rohwell. yeah. the journal needs to be somewhere, but reiser needs that also, right?15:36
viricroh: I've two loop devices of the same size, same tar unpacked to them. I run 'df', and shazam... 1,8G difference15:36
viricyes, reiser has journal to15:36
virichum maybe I did not pass the hardlinks with tar... does tar pick hardlinks by default?15:37
virichm maybe it's that. I'll resolve the hardlinks15:37
rohviric: reserved space?15:38
viricI should have used --hard-dereference15:38
viricLet's try.15:38
DocScrutinizer05viric: check for stuff hidden under mountpoints19:02
DocScrutinizer05age old prank 19:02
DocScrutinizer05also semantics of "used space" differs a lot, depending if you count and sum up real filelength, or you calculate space taken on storage - incl. sector overhead which is statistically 0.5 sectors/file19:04
DocScrutinizer05plus inodes and whaztnot19:04
viricDocScrutinizer05: I only look at 'df' free space19:05
viricDocScrutinizer05: I mount one fs into ./r, the other into ./o;19:06
DocScrutinizer05viric: ooh, I didn't mean mountpoint where the fs under test got mounted. Hidden files under a mountpoint count for the disk usage of the fs where the mountpoint dir is located, not for the mounted fs19:30
DocScrutinizer05so aiui you mount a fs under ./o or ./r, and probably both have no mounts on them, so are unaffected by any hidden files 19:31
viricin any case I look at 'df'19:34
viricI think that maybe the hardlinks explain the story... one fs has hardlinks, the other not19:35
DocScrutinizer05df probably just looks for free blocks19:35
viricbut I don't know how to convey hardlinks from one side to the other.19:35
virictar --hard-dereference clearly fails19:35
viricin fact when I used --hard-dereference, I had even less free space19:35
viricbecause 'tar' by default respects hard links, and with --hard-reference I made it copy the contents...19:36
DocScrutinizer05yep, sounds right19:36
viricWeird. Then ext4 is really taking 23% of *additional* metadata compared to reiserfs19:36
viric23% of the total filesystem.19:36
viricI've 1,8GB unavailable only because I use ext4 instead of reiserfs.19:36
DocScrutinizer05hard to believe19:36
viricdo you want to test yourself? I could prepare a public tar. :)19:37
DocScrutinizer05well, reiserfs afaik packs files, which means there's no wasted space at and of files (this average half sector)19:37
DocScrutinizer05for a lot of files this can sum up19:38
DocScrutinizer05s/at and/at end/19:38
viricthey have that 'tail' thing, yes19:40
viric400k files19:40
viricDoes somebody know if UML works for anything other than x86?20:26
viric(no arm and no mips, I imagine)20:27
--- Mon Jun 11 201200:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!