1225584078 Q * dna Quit: Verlassend
1225584299 J * larsivi ~larsivi@9.80-202-30.nextgentel.com
1225584341 Q * duckx Remote host closed the connection
1225584626 J * duckx ~Duck@81.57.39.234
1225585125 M * Guy- Bertl_zZ: sorry, the patch doesn't appear to fix the problem for me: http://pastebin.be/14657
1225585475 M * daniel_hozac Guy-: try it with http://people.linux-vserver.org/~dhozac/p/k/delta-mutex-fix02.diff too.
1225585528 M * Guy- aye aye
1225585645 M * Guy- does the 'invalid extended attribute' message signify some sort of corruption, or just that the named attribute didn't exist?
1225585659 M * daniel_hozac i've never seen it before...
1225585708 M * Guy- fs/namei.c: In function 'cow_break_link':
1225585708 M * Guy- fs/namei.c:2813: error: invalid type argument of '->' (have 'struct nameidata')
1225585718 M * Guy- this is with the second patch
1225585729 M * daniel_hozac heh, whoops.
1225585769 M * daniel_hozac revert it and use http://people.linux-vserver.org/~dhozac/p/k/delta-mutex-fix03.diff instead.
1225585786 M * daniel_hozac (or change it manually)
1225585812 M * Guy- right
1225585921 M * Guy- OK, it compiles now
1225586055 J * friendly ~friendly@ppp121-44-194-11.lns10.mel4.internode.on.net
1225586076 Q * friendly Remote host closed the connection
1225586095 J * friendly ~friendly@ppp121-44-194-11.lns10.mel4.internode.on.net
1225586113 M * Guy- rebooting
1225586411 M * Guy- OK, it doesn't hang; however, chmod now says chmod: changing permissions of `mount': Too many links
1225586428 M * Guy- fchmodat(AT_FDCWD, "mount", 04755)      = -1 EMLINK (Too many links)
1225586445 M * daniel_hozac what do you get in dmesg (with debugging enabled)?
1225586454 M * Guy- let me paste it
1225586535 M * Guy- http://pastebin.be/14658
1225586706 M * daniel_hozac hmm.
1225586708 M * daniel_hozac EINVAL...
1225586809 M * daniel_hozac what filesystem was this on? jfs?
1225586812 M * Guy- yes
1225586931 M * daniel_hozac can you touch "mount`echo -e '\251'`"?
1225587031 Q * larsivi Quit: Konversation terminated!
1225587080 M * Guy- yes
1225587093 P * LuckyLuke goodbye
1225587097 M * Guy- -rw-r--r-- 1 root root     0 Nov  2 01:51 mount\251
1225587116 M * Guy- I'm not sure this is what we wanted though
1225587178 M * Guy- yes, echo -e just prints \251, verbatim
1225587193 M * daniel_hozac try \xA9 then.
1225587196 M * Guy- I'll retry with /bin/echo
1225587216 M * Guy- touch mount$(/bin/echo -e '\251')
1225587216 M * Guy- touch: setting times of `mount\251': Invalid argument
1225587236 M * Guy- could be a problem with iocharset=utf8?
1225587236 M * daniel_hozac setting times? but the file was created?
1225587241 M * Guy- no
1225587273 M * daniel_hozac okay, good.
1225587273 M * Guy- open("mount©", O_WRONLY|O_CREAT|O_NOCTTY|O_NONBLOCK, 0666) = -1 EINVAL (Invalid argument)
1225587273 M * daniel_hozac yeah, that seems likely.
1225587273 M * daniel_hozac right, that's the same problem the COW code is having.
1225587303 M * Guy- I can live without iocharset=utf8
1225587356 M * daniel_hozac this would be easy to fix though.
1225587397 M * Guy- for the case of utf8, probably, but would that work with all iocharsets?
1225587458 M * daniel_hozac anything <128 should be fine for all of them, i hope.
1225587477 M * Guy- yeah, I guess :)
1225587490 M * Guy- what happens when the original filename is already very long?
1225587496 M * Guy- you just replace the last character then?
1225587518 M * daniel_hozac no, we're unconditionally adding a character.
1225587632 M * Guy- isn't that a highly unlikely accident waiting to happen? :)
1225587636 M * daniel_hozac you could try to change line 2767 in fs/namei.c from \251 to \176, and the \240 in line 2796 to \150.
1225587682 M * daniel_hozac honestly, if somebody has filenames with more than 256 characters, i think it's perfectly fine that COW fails :-)
1225587728 M * Guy- I already removed the iocharset argument, and chmod works now
1225587768 M * Guy- wouldn't 176 and 150 still be invalid in utf8?
1225587829 M * daniel_hozac nah, \y is octal.
1225587835 M * Guy- ah right
1225587876 M * Guy- btw, I don't get this 'jfs is unicode' thing; to me, the intuitive way would be for the filesystem to not care about the contents of filenames
1225587888 M * Guy- what's the advantage of jfs's unicodeness?
1225587925 M * Guy- or should I rather ask IBM? :)
1225587965 M * Guy- if the 'iocharset' setting were part of the state space of each process, I would understand it
1225588000 M * Guy- jfs could present unicode filenames in whatever encoding the application preferred and translate new filenames into unicode based on that
1225588039 M * Guy- but since it is global (well, per-fs), two applications with different locale settings will still have issues with certain filenames, won't they?
1225588060 M * daniel_hozac yep.
1225588076 M * daniel_hozac i don't know what they're trying to achieve.
1225588090 M * Guy- that's a relief
1225588100 M * Guy- I was afraid I was missing something obvious...
1225588124 M * daniel_hozac well, it's possible i'm missing the obvious too...
1225588151 M * Guy- I can always find consolation in the fact that I didn't miss it alone :)
1225588166 M * daniel_hozac hehe
1225588185 M * Guy- anyway, thanks for the patches, you and Bertl
1225588209 M * Guy- will there be a 2.2.0.8 with this fix in it?
1225588350 M * daniel_hozac i guess so.
1225588634 M * Guy- OK, thanks again and good night
1225588708 M * daniel_hozac you're welcome!
1225589445 J * FireEgl Proteus@2001:470:e056:1:4::
1225589639 Q * friendly Ping timeout: 480 seconds
1225601729 Q * xdr Quit: leaving
1225601742 J * xdr ~xdr@99-173-96-87.cust.blixtvik.se
1225607399 J * ntrs_ ~ntrs@77.29.17.42
1225617226 J * ntrs__ ~ntrs@77.29.13.116
1225617654 Q * ntrs_ Ping timeout: 480 seconds
1225618447 Q * pmenier Quit: Konversation terminated!
1225619874 Q * hparker Quit: Read error: 104 (Peer reset by connection)
1225619939 J * bonbons ~bonbons@2001:960:7ab:0:2c0:9fff:fe2d:39d
1225620079 Q * ntrs__ Ping timeout: 480 seconds
1225621577 J * larsivi ~larsivi@9.80-202-30.nextgentel.com
1225622019 Q * micah Ping timeout: 480 seconds
1225622763 J * pmenier ~pmenier@ACaen-152-1-42-100.w83-115.abo.wanadoo.fr
1225623161 J * micah ~micah@micah.riseup.net
1225624555 Q * larsivi Remote host closed the connection
1225626814 J * doener_ ~doener@i577B8DDD.versanet.de
1225626918 Q * doener Ping timeout: 480 seconds
1225633003 J * larsivi ~larsivi@9.80-202-30.nextgentel.com
1225633605 N * Bertl_zZ Bertl
1225633611 M * Bertl morning folks!
1225634287 M * pmenier morning Bertl
1225634325 M * pmenier i'm running 2.6.27.4-vs2.3.0.35.8 sine yesterday: all work fine
1225634330 M * pmenier since*
1225634345 M * Bertl excellent! tx for the feedback!
1225634384 M * pmenier it's a basical config : all mounts are ext3 and i'm just running 6 vservers...
1225634411 J * chigital ~chigital@tmo-100-146.customers.d1-online.com
1225634437 Q * chigital 
1225634449 M * Bertl x86 or x86_64?
1225634473 M * pmenier x86 SMP
1225641623 J * TimG ~root@tor-irc.dnsbl.oftc.net
1225644890 J * ntrs__ ~ntrs@77.29.16.227
1225645713 J * nou ~chaton@2001:6f8:328:bbc:6666:6667::
1225648966 J * LuckyLuke ~lucky@host148-81-static.72-81-b.business.telecomitalia.it
1225648996 M * LuckyLuke hello there
1225649132 M * LuckyLuke due to a bug in dietlibc's mmap() on HPPA when using LARGEFILE, I had to patch util-vserver's Makefile to link check-unixfile and ht2ext against glibc (which doesn't suffer the mmap() problem). I'm no vserver expert but as far as I know everything is working without problems. Do anyone know a reason why those apps need dietlibc? if not, is it possible to apply the patch upstream?
1225649145 M * LuckyLuke this is what I'm using: http://bugs.gentoo.org/attachment.cgi?id=157487
1225649188 M * LuckyLuke (away
1225649194 A * LuckyLuke sorry :)
1225649244 M * Bertl don`t use glibc
1225649253 M * Bertl try to fix dietlibc instead
1225649286 M * Bertl the problem with glibc is, that it harms security 
1225649291 M * LuckyLuke I tried to contact dietlibc upstream but got no reply so far. I'm looking for the mailinglist right now
1225649312 M * Bertl entering the guest, might load modules (glibc resolver for example) from the guest
1225649369 M * LuckyLuke Bertl: I could be wrong here, but as far as I know util-vserver as a whole won't compile without glibc. Some parts will still use it. And those two apps I modified are only run in the host and never inside a guest context
1225649404 M * Bertl ah, so you only compiled some parts against glibc?
1225649412 M * LuckyLuke only those two files
1225649422 M * Bertl if they are not used on an enter or similar commands, it should be fine
1225649422 M * LuckyLuke see the diff at the url I pasted above
1225649442 M * Bertl I`m sure, daniel_hozac will look into that when he gets around
1225649511 M * LuckyLuke also, daniel_hozac proposed me some little changes in dietlibc but they didn't fix the problem in the first place, so right now there isn't any known fix. Maybe fefe will know what to do, maybe not?
1225649592 M * Bertl do you know what the `problem` is?
1225649670 M * LuckyLuke yes, if you call dietlibc's mmap() with largefile support turned on (can't remember the exact flag name), it will always fail with ENODEV on hppa.
1225649726 M * LuckyLuke h2ext and check-unixfile are run on guest images when you use 'vserver GUEST create' so they have a very good reason to use largefile, otherwise they wouldn't work with guest images greater than 4GiB
1225649757 M * LuckyLuke on the other hand, glibc's mmap() will work just fine under the same conditions
1225649784 M * LuckyLuke I'm using this configuration since summer and haven't found any problem so far
1225649793 M * Bertl so we can assume that the kernels mmpa is fine
1225649801 M * LuckyLuke yep
1225649809 M * Bertl just the dietlibc implementation is broken (which is not that unusual)
1225649825 M * Bertl we already fixed a number of dietlibc bugs ...
1225649826 M * LuckyLuke maybe it's just a one-liner stupid thing but I have 0 (or less ;) experience at this level
1225649847 M * Bertl can you compile a binary for me which uses the mmap with and without the largefile?
1225649869 M * Bertl (with dietlibc, and making sure that the one fails)
1225649923 M * LuckyLuke you don't need it: it is already covered by dietlibc's tests. I already created a tgz complete with makefile. Please look at http://bugs.gentoo.org/show_bug.cgi?id=227793
1225649954 M * LuckyLuke I'm not getting much (read: any) support from the gentoo vps devs :(
1225649979 M * Bertl well, I`d like to look at the asm code :)
1225650011 M * LuckyLuke sure. is the small test case ok or you need the actual vserver utilities?
1225650039 M * Bertl simple minimal test, can even be from the dietlibc test suite is fine
1225650057 M * LuckyLuke for what it's worth, I already tested this with both 32bit and 64bit hppa kernels, no change
1225650060 M * LuckyLuke ok. gimme a minute
1225650209 M * LuckyLuke tgz ready, where do I send it?
1225650225 N * pmenier pmenier_off
1225650236 M * Bertl put it somewhere I can download it, if possible
1225650317 M * LuckyLuke kay
1225650565 Q * cga Quit: WeeChat 0.2.6
1225653143 J * Mojo1978 ~Mojo1978@ip-88-152-103-205.unitymediagroup.de
1225654144 Q * xdr Ping timeout: 480 seconds
1225659253 J * xdr ~xdr@gote2.84.cust.blixtvik.net
1225659448 M * Bertl off to bed now .. have a good one everyone!
1225659453 N * Bertl Bertl_zZ
1225660839 Q * xdr Ping timeout: 480 seconds
1225661567 Q * micah Remote host closed the connection
1225662863 J * ntrs_ ~ntrs@77.29.15.62
1225663293 Q * ntrs__ Ping timeout: 480 seconds
1225663449 J * oconnect ~micah@micah.riseup.net
1225663918 Q * bonbons Quit: Leaving
1225664218 J * xdr ~xdr@167-173-96-87.cust.blixtvik.se
1225665253 J * Aiken ~Aiken@ppp118-208-96-59.lns3.bne4.internode.on.net
1225666035 Q * Mojo1978 Read error: Connection reset by peer
1225666366 Q * kiorky Remote host closed the connection
1225666398 J * kiorky ~kiorky@cryptelium.net
1225666708 J * Aiken_ ~Aiken@ppp118-208-115-231.lns4.bne4.internode.on.net
1225667038 Q * ntrs_ Ping timeout: 480 seconds
1225667049 Q * Aiken Ping timeout: 480 seconds
1225668098 J * docelic ~docelic@78.134.192.108
1225669547 Q * docelic Quit: http://www.spinlocksolutions.com/