maemo.org - Talk

maemo.org - Talk (https://talk.maemo.org/index.php)
-   Alternatives (https://talk.maemo.org/forumdisplay.php?f=36)
-   -   Introducing ubiboot N9 (multiboot OS loader) (https://talk.maemo.org/showthread.php?t=89345)

juiceme 2013-11-17 23:08

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by Eztran (Post 1387121)
Here's one (why I haven't looked at the syslog yet, I can only explain with 'whoops'). Can't tell exactly where it happens as I forgot to get the time, but it was definitely within this (could be line 13 onwards? Certainly after this you can see apps exiting/restarting)

This could propably be around the start of installer (package manager):
Code:

Nov 17 18:13:10 (2013) pkgmgrd[8477]: GLIB MESSAGE default - ASSUME NETWORK 0

This sounds like it could have something to do with the phenomenon:
Code:

Nov 17 18:13:59 (2013) applauncherd[5028]: Daemon: State file not on aegis protected file system, exiting.

If that is related, then the problms are indeed caused by some aegis mismatch.

Eztran 2013-11-17 23:52

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by juiceme (Post 1387164)
This sounds like it could have something to do with the phenomenon:
Code:

Nov 17 18:13:59 (2013) applauncherd[5028]: Daemon: State file not on aegis protected file system, exiting.

Well, that's interesting. Google brings up some seemingly-relevant code.

It looks to me (note: I don't know C++) that after an installation is complete, applauncherd attempts to reload its previous state.

/var/run/applauncherd (which contains the state file, /var/run/applauncherd/saved-state) is checked as to whether it's an aegisfs mountpoint first. Somehow/somewhere, this fails, causing applauncherd to exit.

Well, that would explain why there's only a problem in openmode. What I don't understand is why this problem doesn't happen all the time, and not for all openmode users.

Also possibly useful: aegis_system_mode_t, aegis_crypto_verify_aegisfs

juiceme 2013-11-18 05:46

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by Eztran (Post 1387167)
Well, that's interesting. Google brings up some seemingly-relevant code.

It looks to me (note: I don't know C++) that after an installation is complete, applauncherd attempts to reload its previous state.

/var/run/applauncherd (which contains the state file, /var/run/applauncherd/saved-state) is checked as to whether it's an aegisfs mountpoint first. Somehow/somewhere, this fails, causing applauncherd to exit.

Well, that would explain why there's only a problem in openmode. What I don't understand is why this problem doesn't happen all the time, and not for all openmode users.

Also possibly useful: aegis_system_mode_t, aegis_crypto_verify_aegisfs


Good catch that :D

Now as I check that, it is indeed in the non-closed part of harmattan, distributed as applauncherd_3.0.3+0m8.tar.gz on the PR1.3 cdrom.
What this means, is that it can be fixed :D

jalyst 2013-11-18 06:09

Re: Introducing ubiboot N9 (multiboot OS loader)
 
http://i1109.photobucket.com/albums/...ps9959a696.gif

coderus 2013-11-18 06:24

Re: Introducing ubiboot N9 (multiboot OS loader)
 
@juiceme i think we need to make ubibootctl cli-tool with
--add-kernel=/path/to
--remove-kernel=/path/to
--type=harmattan|nemo|android|sailfish
--refresh
options. this tool should automatically add kernel path to active .conf file and set init options to it or remove kernel from config. then we can easily pack kernels to deb archive and use ubibootctl in postinst and prerm scripts =)

juiceme 2013-11-18 08:54

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by juiceme (Post 1387191)
Good catch that :D

Now as I check that, it is indeed in the non-closed part of harmattan, distributed as applauncherd_3.0.3+0m8.tar.gz on the PR1.3 cdrom.
What this means, is that it can be fixed :D


Allright, I recompiled "/usr/lib/applauncherd/libapplauncherd.so" with the slightest possible modification, just set the debug flags on so that it will not die on the aegis conflict.

Please try the new library (md5sum=238b47681749eb2bf78a819ac6bb42f7) and see if it makes any difference for you :)

Eztran 2013-11-18 09:41

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by juiceme (Post 1387218)
Please try the new library (md5sum=238b47681749eb2bf78a819ac6bb42f7) and see if it makes any difference for you :)

Thanks. Well, no crashes yet. However, there also hasn't been any note of the conflict in the syslog, so I can't tell if the bug just hasn't happened yet - stupid unreliable bugs ;)

Will keep checking after installs etc. Anyone else having this problem, now is probably the time to test! :)

TMavica 2013-11-18 09:43

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by juiceme (Post 1387218)
Allright, I recompiled "/usr/lib/applauncherd/libapplauncherd.so" with the slightest possible modification, just set the debug flags on so that it will not die on the aegis conflict.

Please try the new library (md5sum=238b47681749eb2bf78a819ac6bb42f7) and see if it makes any difference for you :)

just overwrite original one?

juiceme 2013-11-18 10:00

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by TMavica (Post 1387226)
just overwrite original one?

That's what I did. Or actually, to make it safe I first copied the old library off just in case it would not work with the new one.

Code:

~ $ devel-su
Password:

BusyBox v1.20.0.git (MeeGo 3:1.20-0.2+0m8) built-in shell (ash)
Enter 'help' for a list of built-in commands.

~ #
~ # cp /usr/lib/applauncherd/libapplauncherd.so /root/libapplauncherd.so.backup
~ # cp /root/libapplauncherd.so /usr/lib/applauncherd/libapplauncherd.so
~ #


Hurrian 2013-11-18 12:07

Re: Introducing ubiboot N9 (multiboot OS loader)
 
Quote:

Originally Posted by juiceme (Post 1387218)
Allright, I recompiled "/usr/lib/applauncherd/libapplauncherd.so" with the slightest possible modification, just set the debug flags on so that it will not die on the aegis conflict.

Awesome!
And I thought this was a legit Harmattan bug - it was getting annoying!

I wonder how deep Aegis is embedded into the system, and whether we can just stub it out/remove it completely from libs?


All times are GMT. The time now is 12:37.

vBulletin® Version 3.8.8