2009-02-20 00:02 -!- shapor(~shapor@yzf.shapor.com) has joined #tux3 2009-02-20 00:02 -!- flips(~phillips@phunq.net) has joined #tux3 2009-02-20 00:02 -!- vcgomes(~vcgomes@li17-238.members.linode.com) has joined #tux3 2009-02-20 00:02 -!- vomjom(~vomjom@99-157-248-71.lightspeed.stlsmo.sbcglobal.net) has joined #tux3 2009-02-20 00:37 googlebot indexing the mail archives now 2009-02-20 00:38 my download speed drops by 2/3rds when there is a bot snooping around 2009-02-20 00:45 Tux3 Report is the 3rd most popular lkml message on lkml.org: http://lkml.org/ 2009-02-20 00:46 I suppose that mainly means that people like to read about filesystems 2009-02-20 00:46 probably has something to do with everything in unix is a filesystem 2009-02-20 00:57 oh 2009-02-20 00:57 I found one article 2009-02-20 00:57 http://www.linux-magazin.de/meldung/34306 2009-02-20 00:57 heh 2009-02-20 00:57 those guys know me 2009-02-20 00:57 from when I lived in germany 2009-02-20 00:58 one of the guys I worked with is now an editor for linux magazine 2009-02-20 00:58 oh 2009-02-20 00:58 it says, phillips and hirofumi booted linux on a tux3 filesystem in the last few days 2009-02-20 00:59 you made news in germany :) 2009-02-20 00:59 is it the first time? probably not, because they are also interested in fatfs there 2009-02-20 00:59 :) 2009-02-20 01:00 they tell the story in detail, how the root filesystem was copied 2009-02-20 01:00 that's why I like germans 2009-02-20 01:00 they are interested in every detail 2009-02-20 01:01 i see 2009-02-20 01:02 it says, after a little fiddling phillips was able to run apt-get install 2009-02-20 01:02 and it says, for not the computer ran on tux3 without a problem 2009-02-20 01:03 it seems that google translate is not so bad 2009-02-20 01:03 it tells about the upcoming SCALE talk, and it notes that tux3 has no crash recovery right now 2009-02-20 01:03 :) 2009-02-20 01:04 and it says how we will implement versioning during the review process 2009-02-20 01:04 well, I hope my translation was as good as google's ;) 2009-02-20 01:05 I guees so much good than google :) 2009-02-20 01:29 -!- dagle1(~dagle@host162-104.bornet.net) has joined #tux3 2009-02-20 02:00 flips: what about the /dev directory ? is that fine as well ? 2009-02-20 02:21 bh, /dev is fine, thanks to hirofumi's patch from two days ago 2009-02-20 02:21 this path is not in the repository 2009-02-20 02:21 this patch I mean 2009-02-20 02:49 I tink it nice 2009-02-20 02:49 nice 2009-02-20 02:49 bah 2009-02-20 02:49 garbage on the entry line 2009-02-20 02:49 well, that's a huge step to have it be a live file system 2009-02-20 02:49 hopefully, more developers will come on board because of it 2009-02-20 02:50 it is quite likely 2009-02-20 02:50 yeah, less of an appearance of tux3 being vaporware 2009-02-20 02:50 which is always important trying to attract help 2009-02-20 02:54 this vaporware connected to irc yesterday ;) 2009-02-20 02:55 hm.. every new distri uses tmpfs for /dev 2009-02-20 02:55 nice :0 2009-02-20 02:55 :) 2009-02-20 02:56 the more exposure you get the better the chance to discover and fix bugs. I'm pretty impressed with how low the bug count is at this moment, there aren't that many wierd corner cases hit so far which could be either a good or bad sign :) 2009-02-20 02:57 running it as a root file system will definitely help 2009-02-20 03:32 the bug count has in fact been unusually low 2009-02-20 03:32 this is for a number of reasons, hirofumi is one of the reasons 2009-02-20 03:32 and the method of developing mainly in user space is a big reason 2009-02-20 03:33 also, the structure is fairly simple, where there is complexity is is largely local 2009-02-20 03:34 the unit tests are probably the biggest reason for the low bug count 2009-02-20 06:46 -!- tim_dimm(~timothyhu@cpe-76-168-94-231.socal.res.rr.com) has joined #tux3 2009-02-20 07:37 -!- tim_dimm(~timothyhu@cpe-76-168-94-231.socal.res.rr.com) has joined #tux3 2009-02-20 08:13 -!- RazvanM(~RazvanM@dazzler.isi.jhu.edu) has joined #tux3 2009-02-20 09:18 -!- gaurav(~gaurav@59.95.7.96) has joined #tux3 2009-02-20 09:44 -!- mingming(~mingming@32.97.110.51) has joined #tux3 2009-02-20 11:37 flips: unit testing is critical 2009-02-20 11:38 there's no reason to write a chunk of software like this without it 2009-02-20 12:25 -!- cdk(~chinmay@115.109.14.217) has joined #tux3 2009-02-20 13:43 -!- marcin(~marcin@c-76-23-106-132.hsd1.sc.comcast.net) has joined #tux3 2009-02-20 15:42 it just occurred to me that a variant of advance that always advances to the next block and returns, as opposed to advancing to the next leaf block, would be useful for writing some utilities 2009-02-20 15:42 that is, it would return after doing a single push, if a push is required 2009-02-20 15:42 for symmetry, return after a single pop 2009-02-20 15:44 return a result code: 1 = pushed, but not at leaf; 0 = at leaf; -1 = popped; -2 = finished (something like that) 2009-02-20 15:53 how difficult would it be to write a program that i'd invoke with 'superstat file1' and it would spit out number of all the blocks that it occupies? 2009-02-20 15:54 not difficult 2009-02-20 15:55 easiest is to add it to tux3.c as an additional command 2009-02-20 15:55 i.e. "tux3 marcindump " 2009-02-20 16:28 flips: another test would be to fill the file system near capacity to see how the allocation and speed of goes 2009-02-20 16:28 fitting all of those blocks and stuff would then be harder 2009-02-20 16:32 I already know how it will go, it will suck 2009-02-20 16:32 tux3 has no allocation policy yet 2009-02-20 16:36 yea i know that's why i wanna aproach it right at the beginning 2009-02-20 16:38 right after scale for me 2009-02-20 16:38 -!- tim_dimm(~timothyhu@cpe-76-168-94-231.socal.res.rr.com) has joined #tux3 2009-02-20 16:39 sk8 oclock 2009-02-20 16:39 ok 2009-02-20 17:38 -!- samlh(~sam@67.129.121.145) has joined #tux3 2009-02-20 18:15 -!- samlh(~sam@67.129.121.145) has joined #tux3 2009-02-20 18:53 http://www.experts-exchange.com/Storage/Hard_Drives/Q_24159694.html <- in spite of intel's denial 2009-02-20 19:05 -!- marcin(~marcin@c-76-23-106-132.hsd1.sc.comcast.net) has joined #tux3 2009-02-20 19:09 ACTION hands marcin a trophy for being the 3rd person in history to run tux3 on a real machine 2009-02-20 19:23 -!- tim_dimm(~timothyhu@cpe-76-168-94-231.socal.res.rr.com) has joined #tux3 2009-02-20 19:42 -!- marcin(~marcin@c-76-23-106-132.hsd1.sc.comcast.net) has joined #tux3 2009-02-20 19:49 -!- marcin(~marcin@c-76-23-106-132.hsd1.sc.comcast.net) has joined #tux3 2009-02-20 19:52 hirofumi, there? 2009-02-20 19:54 ok so how would we go about tracing what blew it up? 2009-02-20 19:54 a lockup without any oops, assert or panic can be tricky 2009-02-20 19:54 best is to see how to reproduce it 2009-02-20 19:55 and then we try to reproduce 2009-02-20 19:55 ok if my irc freezes again, that means i was sucessful ;) 2009-02-20 19:55 then characterize the sequence of events and figure out how to narrow it down 2009-02-20 19:55 well, and think about a dedicated test box 2009-02-20 19:55 surely you have one sitting around 2009-02-20 19:55 an old junker 2009-02-20 19:55 i do have a server laying around, real scsi goodness 2009-02-20 19:56 but that thing is LOUD 2009-02-20 19:56 art is hard ;) 2009-02-20 19:56 i only turn it on if i absolutely have to, or just buy bigger speakers ;) 2009-02-20 19:56 let it run in the garage 2009-02-20 19:57 I have one of those by the way 2009-02-20 19:57 well another way to go, is put out the call to your friends for an old windows box that now gathers dust because latest windows can't handle it 2009-02-20 19:58 ok, time to put the upgraded hg back in the ancient vaio 2009-02-20 19:58 everything new and spiffy now 2009-02-20 19:58 oh, first I should verify I can compile the kernel in the chroot 2009-02-20 20:00 i've been reading about all kinds of wacky container mechanisms in linux lately 2009-02-20 20:01 i think it was on ibm linux dev's site 2009-02-20 20:01 chroot is working great for me 2009-02-20 20:02 actually how would you set up a test box for something that blows up kernels? is chroot enough? wouldnt you want like a vm server? 2009-02-20 20:02 apt-get upgraded a sid system that had already fallen into disrepair 3 years go 2009-02-20 20:02 kvm is good 2009-02-20 20:03 hirofumi uses that 2009-02-20 20:03 qemu also 2009-02-20 20:03 or uml 2009-02-20 20:03 actually, somebody posted changes to our root filesystem to support that, and I haven't incorporated them as promised 2009-02-20 20:03 i've been playing with vmserver2.0 lately, poor man's esx but with much better hwd support 2009-02-20 20:04 need to get off my butt 2009-02-20 20:04 hmm, vmserver, an oracle fork of some other vm project? 2009-02-20 20:05 vmware server 2009-02-20 20:05 2 is VERY different from 1 2009-02-20 20:05 ah yes, also good 2009-02-20 20:05 esx? 2009-02-20 20:06 esx is the enterprise vmware 2009-02-20 20:06 full hypervisor, remote machine migration, storage on nas/san, etc 2009-02-20 20:07 uberleet 2009-02-20 20:08 hey, i just remounted the tux3 partition that crashed, and the files are still there 2009-02-20 20:08 no need to remake stuff 2009-02-20 20:09 nope, i take that back 2009-02-20 20:09 i ran du -sh on it and it segfaulted 2009-02-20 20:09 I'd be very surprised 2009-02-20 20:09 if it survived 2009-02-20 20:10 well this is were we start thinking about tux3fsck 2009-02-20 20:10 or tux3 fsck 2009-02-20 20:10 hm, now i cannot unmount it :/ 2009-02-20 20:11 also no surprise 2009-02-20 20:11 reboot 2009-02-20 20:11 and do tux3 mkfs 2009-02-20 20:11 linux is like that, everything has to run to completion or things get borked 2009-02-20 20:12 3 years from now it probably be more tolerant of broken kernel code 2009-02-20 20:12 well 2009-02-20 20:12 another thing you can do is just hide the mount point 2009-02-20 20:12 marcin /mnt $ lsof +D tux3/ 2009-02-20 20:12 lsof: WARNING: can't lstat(tux3/linux-2.6.28.7/block): No child processes 2009-02-20 20:12 there's a umount option for that 2009-02-20 20:12 lazy umount 2009-02-20 20:12 does that mean anything to you? 2009-02-20 20:12 no 2009-02-20 20:12 I'll google 2009-02-20 20:13 lsof +D searches for processes using the path 2009-02-20 20:13 but i've never seen it spit out this warning 2009-02-20 20:15 anyway, umount -l 2009-02-20 20:26 -!- marcin_(~marcin@c-76-23-106-132.hsd1.sc.comcast.net) has joined #tux3 2009-02-20 21:18 -!- samlh(~sam@67.129.121.145) has joined #tux3 2009-02-20 21:40 hi 2009-02-20 22:41 -!- RazvanM(~RazvanM@dazzler.isi.jhu.edu) has joined #tux3