[debian] 2.6.28 kernel problems (was: Re: When will kernel 2.6.24 be replaced...)
Paul Fertser wrote: rm -rf staging mkdir -p staging make ARCH=arm modules_install INSTALL_MOD_PATH=staging cd staging tar czf ../modules-whatever.tar.gz . cd .. And then scp your modules-whatever.tar.gz to FR and untar to /. Sorry if i'm annoying you, but now it worked. :) After i read build how you mentioned i found the problem why no modules tar.gz were created. The build script needs a second parameter to create the archive where the parameter can be whatever. I don't know why this check is build into the script but with help of that it worked. Second problem was that i had to copy the .config into the dummy directory and not the linux-2.6 dir to be found. The problem is that the info on the toolchain wiki [1] is a bit wrong concerning kernel compiling. So i added a remark with the right steps on this page. I updated my fso-frameworkd with help of the patch from [2] and copied the kernel and modules to the right place. Than i booted with the new 2.6.28 kernel but it didn't work at the first try. I have a 64MB swap file activated in /etc/fstab which lies on the third vfat partition of my sd-card. And at the point where the message activating swap partitions came it hung and showed an error message after 120sec timeout. So i thought that vfat wasn't working at boot time and i disabled the swap file. Next boot goes through and the freerunner started without any suspicious error messages. After it booted the sd-card was mounted what means that vfat should work. I had the same problem with the swap file a few weeks ago with an older andy kernel and i don't know what the problem is. But after a short test i found a few things that doesn't work anymore. Don't know if they are related to the fso-frameworkd patches or to the new kernel. Openmoko-panel-plugin doesn't start anymore with an error message [3]. The system time was set back to 1 Jan 2000. And the greatest problem i had was wlan. I could access the eth0 device and it shows all available wlan stations but i can't use my wpa2 protected wlan with wpa_supplicant anymore. Iwconfig showed that i should be connected to my wlan but i can't get an ip with dhcp anymore like with my old 2.6.24 kernel. The dhcp client always tries and after 6-7 dhcpdiscover it exits with no dhcp server available. That is the reason why i'm now back to the 2.6.24 kernel where these things work. Maybe the userspace tools need a bit more tweaking before they work together with the new 2.6.28 kernel. Or the configuration for the 2.6.28 kernel have to be modified to work with my system. I removed a few unneeded modules (i.e. Android things) and added a few modules but i don't think that these have to do anything with my problems. :/ I attached my dmesg log. Maybe someone know of any other modifications i can try to get the new kernel running a bit smoother. :) Ciao, Rainer [1] http://wiki.openmoko.org/wiki/Toolchain#Building_Openmoko_Kernel_from_git_repo_using_Toolchain [2] http://trac.freesmartphone.org/attachment/ticket/293/0001-Modify-sys-paths-to-work-with-andy-tracking-kernel.patch [3] Traceback (most recent call last): File /usr/bin/openmoko-panel-plugin, line 52, in module icon = PanelPlugin(dbus_object, notificationAvailable) File /usr/lib/python2.5/site-packages/panelplugin/PanelPlugin.py, line 32, in __init__ self.configReader() File /usr/lib/python2.5/site-packages/panelplugin/PanelPlugin.py, line 46, in configReader self.appendIcon(icon) File /usr/lib/python2.5/site-packages/panelplugin/PanelPlugin.py, line 72, in appendIcon self.icons.append(USBIcon(self.dbus, self, self.notificationAvailable)) File /usr/lib/python2.5/site-packages/panelplugin/USBIcon.py, line 17, in __init__ PowerStateIcon.__init__(self, self.dbus, None, None, UsbHost, self.parentObject, usb, self.notificationAvailable) File /usr/lib/python2.5/site-packages/panelplugin/PowerStateIcon.py, line 19, in __init__ self.powerstate = self.dbusdeviceiface.GetPower() File /var/lib/python-support/python2.5/dbus/proxies.py, line 68, in __call__ return self._proxy_method(*args, **keywords) File /var/lib/python-support/python2.5/dbus/proxies.py, line 140, in __call__ **keywords) File /var/lib/python-support/python2.5/dbus/connection.py, line 622, in call_blocking message, timeout) dbus.exceptions.DBusException: org.freedesktop.DBus.Error.UnknownMethod: Method GetPower with signature on interface org.freesmartphone.Device.PowerControl doesn't exist [0.00] Linux version 2.6.28-GTA02_andy-tracking_b8b36e5ec3db71d5-mokodev (r...@wayne) (gcc version 4.1.2) #1 PREEMPT Wed Jan 7 14:41:32 CET 2009 [0.00] CPU: ARM920T [41129200] revision 0 (ARMv4T), cr=c0007177 [0.00] CPU: VIVT data cache, VIVT instruction cache [0.00] Machine: GTA02 [0.00] Memory policy: ECC disabled, Data cache writeback [0.00] On node 0 totalpages: 32768 [0.00] free_area_init_node: node 0, pgdat c04b98bc, node_mem_map
Re: [debian] 2.6.28 kernel problems
Hi, Fox Mulder quakem...@gmx.net writes: Sorry if i'm annoying you, but now it worked. :) No problem, i'm glad it worked :) After i read build how you mentioned i found the problem why no modules tar.gz were created. The build script needs a second parameter I think it needs a first parameter exactly as described on the wiki. Second problem was that i had to copy the .config into the dummy directory and not the linux-2.6 dir to be found. I have no idea where that dummy directory came from :) I'm afraid you're misinterpreting the build script. It seems to me that instructions on wiki are correct. error messages. After it booted the sd-card was mounted what means that vfat should work. I had the same problem with the swap file a few weeks ago with an older andy kernel and i don't know what the problem is. There was a test report from a user stating that his uSD is unstable unless he lowers sd_clock. It might be a problem here, as well. Openmoko-panel-plugin doesn't start anymore with an error message [3]. That's because you use an old unfixed version of o-p-p that tries to read information about usb host power from /sys nodes that were changed. I think you can fix it yourself with the help of http://wiki.openmoko.org/wiki/Kernel#GTA02_.2Fsys_mappings_for_andy-tracking_vs_stable and suggest to o-p-p developers to be compatible with andy-tracking. And the greatest problem i had was wlan. I could access the eth0 device and it shows all available wlan stations but i can't use my wpa2 protected wlan with wpa_supplicant anymore. Iwconfig showed that i should be connected to my wlan but i can't get an ip with dhcp anymore like with my old 2.6.24 kernel. The dhcp client always tries and after 6-7 dhcpdiscover it exits with no dhcp server available. That clearly looks as a regression. I think you should write in details what problems with wlan you experience along with wpa_supplicant debug output and relevant dmesg to the kernel mailing list, where it would (hopefully) be investigated. For me wlan with wpa (not wpa2) works just fine. HTH -- Be free, use free (http://www.gnu.org/philosophy/free-sw.html) software! mailto:fercer...@gmail.com ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: [debian] 2.6.28 kernel problems
Paul Fertser wrote: Hi, Fox Mulder quakem...@gmx.net writes: Sorry if i'm annoying you, but now it worked. :) No problem, i'm glad it worked :) After i read build how you mentioned i found the problem why no modules tar.gz were created. The build script needs a second parameter I think it needs a first parameter exactly as described on the wiki. Thats right. Only the name dummy can also be anything else like GTA02 or whatever. Second problem was that i had to copy the .config into the dummy directory and not the linux-2.6 dir to be found. I have no idea where that dummy directory came from :) I'm afraid you're misinterpreting the build script. It seems to me that instructions on wiki are correct. I took a closer look at the build script and i'm quite sure that i'm interpreting it correct. Maybe the name dummy as first parameter is not very good. In the wiki it uses dummy as first parameter which is nothing else than the name of the output subdirectory. But the build script also has a check for a second parameter with if [ ! -z $2 ] ; then in line 96. And only if a second parameter is given, no matter what it is, the code between line 97 and 100 for making a modules tar.gz file is executed. But this parameter is not mentioned in the wiki. error messages. After it booted the sd-card was mounted what means that vfat should work. I had the same problem with the swap file a few weeks ago with an older andy kernel and i don't know what the problem is. There was a test report from a user stating that his uSD is unstable unless he lowers sd_clock. It might be a problem here, as well. Ok, i will try if this helps at boot time. Openmoko-panel-plugin doesn't start anymore with an error message [3]. That's because you use an old unfixed version of o-p-p that tries to read information about usb host power from /sys nodes that were changed. I think you can fix it yourself with the help of http://wiki.openmoko.org/wiki/Kernel#GTA02_.2Fsys_mappings_for_andy-tracking_vs_stable and suggest to o-p-p developers to be compatible with andy-tracking. Sounds good. And the greatest problem i had was wlan. I could access the eth0 device and it shows all available wlan stations but i can't use my wpa2 protected wlan with wpa_supplicant anymore. Iwconfig showed that i should be connected to my wlan but i can't get an ip with dhcp anymore like with my old 2.6.24 kernel. The dhcp client always tries and after 6-7 dhcpdiscover it exits with no dhcp server available. That clearly looks as a regression. I think you should write in details what problems with wlan you experience along with wpa_supplicant debug output and relevant dmesg to the kernel mailing list, where it would (hopefully) be investigated. For me wlan with wpa (not wpa2) works just fine. I hope i find much debug information and will mail it in the kernel list. Ciao, Rainer ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: [debian] 2.6.28 kernel problems
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Somebody in the thread at some point said: | I took a closer look at the build script and i'm quite sure that i'm | interpreting it correct. | Maybe the name dummy as first parameter is not very good. In the wiki | it uses dummy as first parameter which is nothing else than the name This has changed in the last week. It was taking me a great deal of time to switch between GTA02 and GTA03 kernel builds, so I changed the build script to use the O= kernel makefile parameter to keep the object files and configs for the builds in different subdirs. http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=aeb4db1b913993a3e06522f13bab79decbd3cdef This reduces the time needed to switch typically to just 10 - 20 seconds or so. It requires the subdir name to be given in the first parameter as described in the changed build script. The optional dummy parameter to signal that you want a module tarball is then the second parameter now. It sounds like the combination of the Wiki and the comments and error help in the build script led to you some kind of hybrid solution, but the idea is you would have a GTA02 subdir with suitable config and build that with ./build GTA02, while having a GTA03 subdir with different config and build that with ./build GTA03. - -Andy -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (GNU/Linux) Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org iEYEARECAAYFAklmOSUACgkQOjLpvpq7dMq5YACZASNu4qpntsx30j0cdxgvBoTw rYoAoILbeIC66Dyqv4u1Gzxa/kO9IOsU =e0hQ -END PGP SIGNATURE- ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: [debian] 2.6.28 kernel problems
Andy Green wrote: Somebody in the thread at some point said: | I took a closer look at the build script and i'm quite sure that i'm | interpreting it correct. | Maybe the name dummy as first parameter is not very good. In the wiki | it uses dummy as first parameter which is nothing else than the name This has changed in the last week. It was taking me a great deal of time to switch between GTA02 and GTA03 kernel builds, so I changed the build script to use the O= kernel makefile parameter to keep the object files and configs for the builds in different subdirs. http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=aeb4db1b913993a3e06522f13bab79decbd3cdef This reduces the time needed to switch typically to just 10 - 20 seconds or so. It requires the subdir name to be given in the first parameter as described in the changed build script. The optional dummy parameter to signal that you want a module tarball is then the second parameter now. It sounds like the combination of the Wiki and the comments and error help in the build script led to you some kind of hybrid solution, but the idea is you would have a GTA02 subdir with suitable config and build that with ./build GTA02, while having a GTA03 subdir with different config and build that with ./build GTA03. Thanks for this information. :) So the right command line would be something like ./build GTA02 dummy to find within the GTA02 directory the modules tar.gz afterwards. But before i first run the build script i have to create the GTA02 directory by hand and copy the .config into it. Else the build script doesn't find any .config and starts the console configuration of the kernel. Ciao, Rainer ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: [debian] 2.6.28 kernel problems
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Somebody in the thread at some point said: | So the right command line would be something like ./build GTA02 dummy | to find within the GTA02 directory the modules tar.gz afterwards. Yes but dummy really is a dummy param, it is just checked if it is there or not, so I usually use ./build GTA02 x for example. To save time, most builds I am not giving that second parameter at all. | But before i first run the build script i have to create the GTA02 | directory by hand and copy the .config into it. Else the build script | doesn't find any .config and starts the console configuration of the kernel. Right, part of the gain from this change is that the configs along with the .o files all live in their little worlds now, so nothing gets overwritten in ./GTA02 if you build ./GTA03. I realize that's not exactly being a problem for most people yet, but it was getting pretty bad here having to copy over the config and wait 10+ minutes every time I had to switch to look at something. - -Andy -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (GNU/Linux) Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org iEYEARECAAYFAklmYjUACgkQOjLpvpq7dMq2vQCeOL4Qg4L2S6I/nDBnPZe371RQ dL0AnjiNLD02U7OTAAZqxU9laXQBkQC0 =IUJK -END PGP SIGNATURE- ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: [debian] 2.6.28 kernel problems
Andy Green a...@openmoko.com writes: Yes but dummy really is a dummy param, it is just checked if it is there or not, so I usually use ./build GTA02 x for example. To save time, most builds I am not giving that second parameter at all. Btw, since I have my compiler in a custom location would be useful if build sourced some local-build-settings file where I could override compiler path? I can modify the real build script but then git complains that I have uncommitted changes. I can also copy the build script to a different name but then I need to constantly keep it in sync with the version in git. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
