Active Topics
-
Camera phone competition March 2023 "Endless" (35)
to General by Maemish - 2 days, 17 hrs ago -
TLS 1.2 for N9 is ready (28)
to MeeGo / Harmattan by smartblu9 - 2 days, 19 hrs ago -
Maemo repository (3)
to Community by torsen - 3 days, 10 hrs ago -
Error after installing Leste on sdcard (11)
to Maemo 7 / Leste by Maemish - 4 days, 8 hrs ago -
Introducing ubiboot N9 (multiboot OS loader) (1,393)
to Alternatives by smatkovi - 5 days, 14 hrs ago - more...
|
2010-04-15
, 09:25
|
|
Posts: 540 |
Thanked: 288 times |
Joined on Sep 2009
|
#72
|
Originally Posted by ignarus
No but if the cifs module is compiled for the stock kernel (I mean compiled later, ie not included in the default stock config) then there definitely is version mismatch.
First of all awesome stuff this mobile hotspot!
After installing the mobile hotspot debs, I had some trouble with the cifs module though. Dmesg error on modprobing:
Is this hotspot kernel missing something that was present in the stock kernel?
Edit: The hotspot kernel is configured literally by adding a few options to the stock config as generated by the stock makefile, see hotspotize.sh.
Originally Posted by ignarus
The other kernel should use the same scheme as me and titan use with Provides: -line (for example http://maemo.org/packages/package_in....6.28-maemo24/).
Also wouldn't it be a good idea to remove the dependencies to the custom kernel and modules from the mobilehotspot_0.2.0_all.deb? This way the mobile hotspot can be installed/updated while using some other 'custom' kernels which also support everything needed for doing nat.
Correct dependencies and provides declarations are the base of sane package management...
Last edited by rambo; 2010-04-15 at 09:53.
|
|
2010-04-15
, 11:00
|
|
Posts: 224 |
Thanked: 107 times |
Joined on Aug 2009
|
#73
|
Any progress on getting Infrastructure mode working?
|
|
2010-04-15
, 11:37
|
|
Posts: 540 |
Thanked: 288 times |
Joined on Sep 2009
|
#74
|
Originally Posted by bocaJ
No, userspace NAT is the next priority I think (since wek has a solution for it it "just" needs to be integrated [which needs a major redesign of the whole backend...]).
Any progress on getting Infrastructure mode working?
But if it's important to you feel free to start experimenting. Best way to get features you feel important is to figure out how to implement them and then supply patches...
I spent over a week in the futile effort of trying to get working kernel-space NAT for stock kernel when starting this project, (learned at lot while at it though) wek and qole have spent a lot of time on the userspace NAT solution. No need to be expert outright, willingness to learn and to spent some time at it will go a very long way.
|
|
2010-04-15
, 13:23
|
|
Posts: 4 |
Thanked: 0 times |
Joined on Apr 2010
|
#75
|
Thanks for the reply.
I do not exactly understand why there should be a version mismatch as both kernels were compiled from the same source? But I must admit that I do not know much about how this version checking works exactly.
I'm not sure that I understand what you mean here. What happens if another package which requires a custom kernel config needs to be installed. Let's say a package that needs usbnet supported. Then the maintainer of this package needs to recompile hotspot-kernel with the addition of usbnet? And would the package on which it depends still be called hotspot-kernel?
I don't think you should have a dependency to a certain kernel config. I've never seen it done anyway.
Last edited by ignarus; 2010-04-15 at 13:30.
No but if the cifs module is compiled for the stock kernel (I mean compiled later, ie not included in the default stock config) then there definitely is version mismatch.
The other kernel should use the same scheme as me and titan use with Provides: -line (for example http://maemo.org/packages/package_in....6.28-maemo24/).
I don't think you should have a dependency to a certain kernel config. I've never seen it done anyway.
Last edited by ignarus; 2010-04-15 at 13:30.
|
|
2010-04-15
, 20:09
|
|
Posts: 540 |
Thanked: 288 times |
Joined on Sep 2009
|
#76
|
Originally Posted by ignarus
Because hotspot kernels version string is 2.6.38-hotspot and stock kernels version string is 2.6.38-omap1 (or something like that), titans kernel has again separate version string. This is done for a multitude of good reasons.
I do not exactly understand why there should be a version mismatch as both kernels were compiled from the same source? But I must admit that I do not know much about how this version checking works exactly.
Originally Posted by ignarus
It's not a certain kernel config or even certain kernel package, it's certain kernel feature, and when the software will not work without said feature it must depend on it (or there is no sense packaging it at all).
I'm not sure that I understand what you mean here. What happens if another package which requires a custom kernel config needs to be installed. Let's say a package that needs usbnet supported. Then the maintainer of this package needs to recompile hotspot-kernel with the addition of usbnet? And would the package on which it depends still be called hotspot-kernel?
I don't think you should have a dependency to a certain kernel config. I've never seen it done anyway.
This if why mobilehotspot depends on either hotspot-kernel or kernel-feature-netfilter, now titans kernel happens to provide said feature, hotspot-kernel also provides that feature but the dependency is listed like that to prefer the hotspot-kernel if no kernel providing the netfilter feature is yet installed (of course the hotspot kernel is not in autobuilder yet since I have been lazy and since titans kernel has much more features)
If there is a package that needs the usbnet module then it should specify that dependency in some commonly accepted way (like the kernel-feature-xxx that we coined with titan) and if maintainer of same package happens to maintain a package that provides this module to stock kernel then his package should state that it provides said feature (again the dependency can be stated as his_package | kernel-feature-xxx to prefer his).
He could also compile the usbnet module (since it's one of those that are easy to compile afterwards and the ko insers fine since it does not depend on anything changing in the kernel proper) for the hotspot kernel too and name that package hotspot-kernel-modules-usbnet (or something along that line) in case he feels that people using the hotspot kernel are better served by this than switching to titans (or some other kernel that provides both usbnet and netfilter features).
Getting a new feature enabled in titans kernel requires just informing him why this feature would be usefull (as long as it does not have weird conflicts with other features but generally usefull ones do not)
|
|
2010-04-15
, 20:26
|
|
Posts: 515 |
Thanked: 259 times |
Joined on Jan 2010
|
#77
|
Originally Posted by ignarus
HI,
geohsia: I've heard somewhere that this joiku spot is using a custom kernel module. Maybe it's no longer working because of the change of kernel / module directory?
You could maybe check the link /lib/modules/current to see if it's reset to the original modules dir
Thanks for the reply. I'm right now set at 2.6.28.10power-omap1
I've re-installed joiku many times but still it doesn't work.
What's the difference between the package that pops up on the extras-dev repository and the manual configs.
|
|
2010-04-15
, 20:34
|
|
Posts: 540 |
Thanked: 288 times |
Joined on Sep 2009
|
#78
|
Originally Posted by geohsia
Please move this discussion to either the relevant kernels or joikuspots thread.
I'm right now set at 2.6.28.10power-omap1
AFAIRecall titans kernel has the joikuspot module compiled in but maybe Joikuspot itself is being stupid (hint: trusting a symlink "current" that is provided by a package is not smart, the software could figure out the running kernels version [and by extension exact module path...] with a very simple call...).
|
|
2010-04-15
, 20:45
|
|
Posts: 515 |
Thanked: 259 times |
Joined on Jan 2010
|
#79
|
Originally Posted by rambo
Hi Rambo. The reason I brought this up here was because my system was working fine until I installed this Hotspot software. I didn't think that it would break Joikuspot.
Please move this discussion to either the relevant kernels or joikuspots thread.
AFAIRecall titans kernel has the joikuspot module compiled in but maybe Joikuspot itself is being stupid (hint: trusting a symlink "current" that is provided by a package is not smart, the software could figure out the running kernels version [and by extension exact module path...] with a very simple call...).
I do have a messge in another Joikuspot specific thread, but I'm trying to track backwards to see what I would have to do to restore functionality. Just looking for some help.
|
|
2010-04-16
, 05:34
|
|
Posts: 540 |
Thanked: 288 times |
Joined on Sep 2009
|
#80
|
Originally Posted by geohsia
Joikuspot was working fine with titans kernel but stopped working after installing mobilehotspot ? That would be interesting but I cannot see how the hotspot would mess joikuspot up in any permanent way. The frontend does nothing but set some configuration variables and run the backend script, the backend does insert some kernel modules on start and might not be able to remove all of them on stop (due to interesting interdependencies), said modules might interfere with Joiku (but in that case Joiku has serious issues to begin with) but reboot will always solve that.
Hi Rambo. The reason I brought this up here was because my system was working fine until I installed this Hotspot software. I didn't think that it would break Joikuspot.
Edit: If you want to get rid of titans kernel run "apt-get install --reinstall kernel kernel-flasher && reboot" as root (this will reinstall and flash stock kernel).
Last edited by rambo; 2010-04-16 at 05:37.
You could maybe check the link /lib/modules/current to see if it's reset to the original modules dir