IRC log of #maemo-ssu for Monday, 2013-11-25

*** NIN101 has quit IRC00:02
*** thedead1440_ has quit IRC00:07
*** thedead1440_ has joined #maemo-ssu00:07
*** thedead1440_ has quit IRC00:11
*** thedead1440_ has joined #maemo-ssu00:11
*** thedead1440_ has quit IRC00:14
*** thedead1440_ has joined #maemo-ssu00:15
*** thedead1440_ has quit IRC00:18
*** thedead1440_ has joined #maemo-ssu00:18
*** xes has joined #maemo-ssu00:23
*** Pali has quit IRC00:24
*** android_808 has quit IRC00:25
*** M4rtinK has quit IRC01:21
*** FlameReaper has quit IRC02:01
*** zumbi_ has quit IRC02:20
*** zumbi has joined #maemo-ssu02:21
*** LauRoman has quit IRC03:29
*** dos1 has quit IRC03:40
*** xes has quit IRC04:10
*** M13 has joined #maemo-ssu04:53
*** M13 has quit IRC04:55
*** M13 has joined #maemo-ssu04:56
*** nox- has quit IRC04:57
*** amiconn_ has joined #maemo-ssu05:29
*** amiconn has quit IRC05:29
*** amiconn_ is now known as amiconn05:29
*** okias has joined #maemo-ssu07:40
okiasHey guys, did anyone using Charger Manager in linux kernel? (except samsung)07:40
*** LaoLang_cool has joined #maemo-ssu07:41
*** freemangordon_ has joined #maemo-ssu07:54
*** freemangordon has quit IRC07:54
*** okias has quit IRC08:02
DocScrutinizer05huh? charger manager?08:28
*** RST38h has quit IRC08:30
*** Pali has joined #maemo-ssu08:32
*** Pali has quit IRC09:17
*** LauRoman has joined #maemo-ssu09:49
*** scoobertron has quit IRC10:17
*** raccoon_ has quit IRC10:53
*** freemangordon has joined #maemo-ssu11:01
*** freemangordon_ has quit IRC11:01
*** raccoon_ has joined #maemo-ssu11:02
*** sunny_s has joined #maemo-ssu11:27
*** Pali_ has joined #maemo-ssu11:47
*** kolp has joined #maemo-ssu11:48
*** LauRoman has quit IRC11:58
*** Pali_ is now known as Pali12:48
*** Pali has quit IRC12:59
*** Wikiwide has joined #maemo-ssu13:00
*** Wikiwide has quit IRC13:15
*** Pali has joined #maemo-ssu13:23
*** Pali has quit IRC13:58
*** lizardo has joined #maemo-ssu14:16
*** Pali has joined #maemo-ssu14:32
*** Mark__T has joined #maemo-ssu14:41
*** okias has joined #maemo-ssu15:16
*** dos1 has joined #maemo-ssu15:27
*** mkaindl has joined #maemo-ssu15:29
*** sunny_s has quit IRC15:32
*** M13 has quit IRC15:48
*** dhbiker has quit IRC16:27
*** dhbiker has joined #maemo-ssu16:29
*** Mark__T has quit IRC16:49
Palifreemangordon: wl1251 crash in monitor mode with 3.12-rc5 is reproducable17:06
Palinow looking at crash dump and trying to fix it17:07
Palifreemangordon: here is crash dump: http://pastebin.com/5DQdgdKi17:18
Palifreemangordon: I think that in function wl1251_event_process variable wl->vif is NULL and then there is NULL dereference in function ieee80211_beacon_loss17:20
Paliso if this is truth then this is problem in wl1251 driver and not in monitor/packet injection patches17:21
Paliand patch could be simple, just do not call that function if vif is NULL17:21
Palifreemangordon: rigth, it fixing that kernel crash17:31
Palipatch is here: http://pastebin.com/YJWZUtY717:32
Paliwill write later to felipec17:32
*** NishanthMenon has joined #maemo-ssu17:34
*** Pali has quit IRC17:39
keriofreemangordon: omg omg omg updatez omg18:29
freemangordonkerio: hmm?18:29
freemangordontoldya, no more thumb for you :P18:29
kerioi upgraded stuff18:29
freemangordonkerio: what stuff?18:30
freemangordonfrom -devel?18:30
keriono, thumby stuff i think18:30
keriomaybe old stuff tho18:30
freemangordonmaybe there were 2 updates for the last month18:30
freemangordon*maybe,18:30
keriomh, it appears that i can't install the -pr18:32
keriowhy can't those dependencies be met?18:32
keriooh18:32
kerioOH18:32
freemangordonkerio: hmm?18:32
keriofreemangordon: apparently you're hosting libmaemosec0 0.2.218:32
keriothere's now a 0.2.318:32
keriobut i've given priority to your repository18:33
keriocuz of -thumb18:33
freemangordonkerio: make sure to refresh18:33
kerio...yes18:33
freemangordonthe latest is 0.2.318:33
kerioi'm using HAM18:33
keriohonest18:33
freemangordonand -thumb depends on it18:33
freemangordonalso, make sure you have enabled cssu-testing catalog18:34
kerioto be fair, pinning isn't really supported18:34
keriono, i've got all the repos18:34
keriobut i've pinned community-thumb to 600 priority18:34
kerioinstead of 50018:34
keriocould you remove non-thumby packages? :318:35
freemangordonkerio: hmm, I don;t thik there is libmaemosec0 in the thumb repo. could be wrong as well18:35
keriodefinetely wrong18:35
freemangordonkerio: no, I can't18:36
kerio._.18:36
freemangordonjust apt-get install whatever is broken on your side18:36
kerioyou're also hosting osso-systemui-alarm 0.3.3+0cssu018:36
kerioi have to install manual versions18:36
freemangordonme? no, it is merlin1991 :P18:36
kerioand i'm also not sure if it's complete18:36
kerioremoving the priority will also make me upgrade from the thumby packages to the stuff in -devel18:37
kerioand i don't know if i want to18:37
freemangordondon't keep -devel ebnabled18:37
kerio#yolo18:37
kerio...the fuck, my n900 just... rebooted?18:38
kerioi wasn't expecting that18:38
freemangordonwhat?18:38
freemangordonwhat did you do?18:38
keriomh, upgraded hildon-desktop18:38
kerioi think it's cuz of that18:38
*** NIN101 has joined #maemo-ssu18:39
freemangordonhmm, shouldn't happen18:39
kerioi don't have syslog enabled sadly18:39
freemangordonpray for the device to boot fine18:39
keriomh18:39
kerio/proc/bootreason says sw_rst18:39
kerionot sure what happened, really18:39
freemangordonweird18:40
kerioit's seriously weird18:40
kerioto be fair i was also using g_ether18:40
keriobut still18:40
freemangordonhmm, sometimes it happens when you have executable pages re-read from the storage, and in the meantime the storage contents have changed18:40
keriowhy would they change tho?18:41
freemangordonkerio: because you've overwritten it with apt-get ;)18:41
keriowhat?18:42
freemangordonnew executable overwrote the old one18:42
kerioshouldn't it count as an open file though?18:42
freemangordonyes18:43
freemangordonbut I guess there might be some bug in the kernel18:43
freemangordonthe same happens if you copy some .so by hand18:43
freemangordonyou got a segfault18:44
freemangordonin the process using that library18:44
keriothat's awful! D:18:44
kerioi am shocked and appalled! D:18:44
kerioalso i think my face got stuck D:18:45
freemangordonyes, that is why you should upgrade by using HAM ;)18:45
freemangordonkerio: slap it, that will fix the things :D18:45
keriothere *is* quite a lot of stuff in -devel18:51
kerioooh i didn't know you thumbified dosfstools18:51
freemangordone2fsprogs should be too18:51
freemangordoniirc18:51
freemangordonor maybe only on my device18:52
freemangordonyep, only on my device, will include them in the next update18:53
*** arcean has joined #maemo-ssu19:07
DocScrutinizer05bug in kernel ??? :-O19:20
DocScrutinizer05WTF?!19:20
DocScrutinizer05that must be a pretty nasty bug19:21
DocScrutinizer05that would affect about EVERYTHING19:21
DocScrutinizer05fs MUST NOT unlink a file that has a handle open to it19:22
DocScrutinizer05~211919:22
infobotThe key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED",  "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.19:22
*** BCMM has joined #maemo-ssu19:49
*** okias has quit IRC19:50
*** okias has joined #maemo-ssu19:51
*** miha_ has joined #maemo-ssu20:25
*** miha_ is now known as M1320:26
*** M13 has quit IRC20:32
*** Pali has joined #maemo-ssu20:43
PaliDocScrutinizer05: this is another problem: application X has mmaped file Y. And by default replacing file means: truncate(0); write(new_file_data);21:00
Paliso if Y is library in that X, then application crash immeditately if exeucting that code21:01
Palicorrect way is to rm (unlink) file Y and then create new with that name of Y21:01
Paliand dpkg doing this21:01
Palicp (from busybox) not21:01
Palinot bug in kernel, but feature in userspace21:02
*** mkaindl has left #maemo-ssu21:02
Paliyou cannot unlink from FS itself if is opened (you can just unlink it from directory structure), but you can erase it or rewrite it21:03
Paliand truncate to zero size too...21:03
DocScrutinizer05Pali: by default "replacing" a file means to create a new file under same name and unlink the old file's inode entry21:06
Palinot by scp/ssh or busybox cp21:06
Pali(and maybe also not by gnu cp)21:06
DocScrutinizer05what you mean is overwriting an existing file content, not replacing the file21:07
Paliyes, but this cp doing21:07
Paliso this is reason for segfaults21:07
DocScrutinizer05no way21:07
DocScrutinizer05a cat $originafile >$newfile would do that21:09
Paliscp doing it too!21:09
PaliI tested it21:09
DocScrutinizer05a cp $originalfile $newfile should always unlink old file for all I know21:09
Paliit is busybox!!!21:09
DocScrutinizer05meh, fuck busybox21:10
DocScrutinizer05busybox is evidently buggy like hell21:12
DocScrutinizer05if we would want to sanitize CSSU, we shoukd kick busybox for everything but initscripts21:12
DocScrutinizer05while for initscripts we should keep original messybox with all the original flaws and bugs, since something might (and evidently did) need them21:13
DocScrutinizer05strace cp x y21:19
DocScrutinizer05stat("y", 0x7fffca3aba00)               = -1 ENOENT (No such file or directory)21:19
DocScrutinizer05stat("x", {st_mode=S_IFREG|0644, st_size=29, ...}) = 021:19
DocScrutinizer05stat("y", 0x7fffca3ab780)               = -1 ENOENT (No such file or directory)21:19
DocScrutinizer05open("x", O_RDONLY)                     = 321:19
DocScrutinizer05fstat(3, {st_mode=S_IFREG|0644, st_size=29, ...}) = 021:19
DocScrutinizer05open("y", O_WRONLY|O_CREAT|O_EXCL, 0644) = 421:19
DocScrutinizer05fstat(4, {st_mode=S_IFREG|0644, st_size=0, ...}) = 021:19
DocScrutinizer05read(3, "Mo 25. Nov 20:17:28 CET 2013\n", 32768) = 2921:19
DocScrutinizer05write(4, "Mo 25. Nov 20:17:28 CET 2013\n", 29) = 2921:19
DocScrutinizer05.21:20
DocScrutinizer05strace cp x y21:20
DocScrutinizer05stat("y", {st_mode=S_IFREG|0644, st_size=29, ...}) = 021:20
DocScrutinizer05stat("x", {st_mode=S_IFREG|0644, st_size=29, ...}) = 021:20
DocScrutinizer05stat("y", {st_mode=S_IFREG|0644, st_size=29, ...}) = 021:20
DocScrutinizer05open("x", O_RDONLY)                     = 321:20
DocScrutinizer05fstat(3, {st_mode=S_IFREG|0644, st_size=29, ...}) = 021:20
DocScrutinizer05open("y", O_WRONLY|O_TRUNC)             = 421:20
DocScrutinizer05fstat(4, {st_mode=S_IFREG|0644, st_size=0, ...}) = 021:20
DocScrutinizer05read(3, "Mo 25. Nov 20:17:28 CET 2013\n", 32768) = 2921:20
DocScrutinizer05write(4, "Mo 25. Nov 20:17:28 CET 2013\n", 29) = 2921:20
DocScrutinizer05jr@saturn:~> cp --version21:21
DocScrutinizer05cp (GNU coreutils) 8.1621:21
DocScrutinizer05Copyright © 2012 Free Software Foundation, Inc.21:21
DocScrutinizer05so yes, you're right regarding O_TRUNC21:24
DocScrutinizer05I stand corrected21:24
*** BCMM has quit IRC21:31
DocScrutinizer05while a mv x y does21:36
DocScrutinizer05access("y", W_OK)                       = 021:36
DocScrutinizer05rename("x", "y")                        = 021:36
DocScrutinizer05I'd expect proper install tools like apt or dpkg to do an unlink() befoer the open()21:39
DocScrutinizer05and scripts to do a rm $newfilename before any cp $original $newfilename21:41
freemangordonDocScrutinizer05: keep in mind those files (libs and bins) are mmaped21:41
DocScrutinizer05err yes, exactly because of that21:42
freemangordonyou can't move anonymous pages21:42
freemangordonwell, the kernel can't21:42
DocScrutinizer05hmm?21:42
freemangordonfrom the kernel POV, those are anonymous mapped pages, iirc21:43
freemangordonand those are unmovable21:43
DocScrutinizer05well, then everything absolutely fine, no cp $original $new will have any effect on them21:43
DocScrutinizer05but you got it wrng21:44
DocScrutinizer05when a process has a .so opened and I want to update that .so, then I need to first rm old.so, so it actually becomes an anonymous file21:45
DocScrutinizer05only THEN I may cp $new old.so21:45
freemangordonsure, but cp is not supposed to rm21:45
freemangordoniiuc21:45
DocScrutinizer05since that way the process keeps a handle to an unaltered mmaped file with old .so content21:45
DocScrutinizer05see what I wrote above21:46
freemangordonotherwise I agree that the correct way to replace an opened .so is to rm it first21:46
DocScrutinizer05since cp is not doing rm, I said [2013-11-25 20:41:08] <DocScrutinizer05> and scripts to do a rm $newfilename before any cp $original $newfilename21:47
DocScrutinizer05or, even better: you do a cp $new tmp; mv tmp $original21:48
DocScrutinizer05since mv does a rename() which implies a unlink()21:49
DocScrutinizer05(unless you mv across volume boundaries)21:49
DocScrutinizer05(in which case IIRC mv actually does a cp + rm SOURCE)21:50
DocScrutinizer05anyway would be interesting what apt and dpkg do21:51
Palidpkg calling unlink for sure22:06
Paliotherwise we would see couple of crashed apps on all debian systems22:06
Pali(when you are upgrading glibc)22:06
Palidpkg/apt also using libc.so22:07
Palievery non static elf using libc.so22:07
*** M4rtinK has joined #maemo-ssu22:08
DocScrutinizer05:nod:22:09
DocScrutinizer05Pali: thanks22:10
*** kolp has quit IRC22:20
*** kolp_ has joined #maemo-ssu22:20
*** lizardo has quit IRC22:57
*** trx has quit IRC23:38
*** LauRoman has joined #maemo-ssu23:39
*** NIN101 has quit IRC23:43
*** trx has joined #maemo-ssu23:43
*** trx has joined #maemo-ssu23:43
*** arcean has quit IRC23:48
kerioPali: to be fair, programs that map a file should handle truncation of that file23:49
kerioit's an unspoken agreement that files can change from behind your nose23:50
Palikerio: tell this to developers of dynamic linker (ld-linux.so)23:50
Palild-linux.so does not doing any that stuff because of speed23:51
*** amizraa has quit IRC23:52
*** amizraa has joined #maemo-ssu23:53
Palihaving all those fds for inotify only for handling truncate is 1) vaste of file descriptors 2) vaste of kernel operations and handling events23:53
*** NishanthMenon has quit IRC23:54
kerio*waste23:54
*** nox- has joined #maemo-ssu23:55
*** tom___ has joined #maemo-ssu23:57

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!