Upgrading to Fedora 18 using Yum - Nouveau causes system freeze?

The documentation for upgrading fedora using yum seems to be much better than before. A very nice script which automates the instructions was terrific. The upgrades of my desktop and the netbook were very smooth. After the upgrade, I installed the Cinnamon desktop as well. It is very nice though I will stick to KDE as I am now comfortable with it.

There was, however, a surprise on the desktop. On booting into Fedora 18, the display vanished and the system hung. The system came up in muti-user mode but not in graphical interface. Gdm was the problem. I switched to kdm. The login screen came and was fine in KDE. However, the system again froze when I tried Cinnamon or Gnome desktops. LXDE worked but the system froze as soon as I started Firefox!

KDE desktop was working with desktop effects disabled. If I enabled them, the desktop froze.

The desktop worked fine with the nomodeset option for booting the kernel. However, it was using the vesa driver. So, the problem is with nouveau driver on "GeForce 6150SE nForce 430" hardware.

Another curious result. The system is working fine with the kernel 3.6.11-1.fc17.x86_64 from fedora 17, which has made it possible to continue with Fedora 18 on the desktop.

Need to reconfirm the problems I faced on the desktop as I did not find a bug report with similar issues.

On Lenovo S10-3, I have so far not faced any issues related to the upgrade.

Upgrading to Fedora 17 - using yum

Since my experience with preupgrade on the desktop was not so great, I decided to take a chance with upgrading my Lenovo S10-3 netbook to fedora 17 using yum.

The additional steps of modifying the filesystem using dracut were not really complex.

Most of the packages had already been downloaded by preupdate, which I had preserved. I copied them into the cache of  yum and executed:

# yum --releasever=17 --disableplugin=presto distro-sync

Unfortunately, after updating about 500 packages, yum seemed to be doing nothing. I killed the command. And ran yum-complete-transaction. It seemed to be using the repositories of Fedora 16 even though fedora-release files had been installed. So, I ran

# yum-complete-transaction --releasever=17

After a pretty long time, it told me that there were problems. I tried again with

# yum-complete-transaction --releasever=17 --skip-broken

An equally long wait and no better result. The problem was that rpm database had entries for both 16 and 17 versions for the packages which were updated.

The following command also did not help (Next time, I need to try with  the -t option to see if it would have tolerated errors).

# package-cleanup --cleandupes

The problem was that packages which had not yet been updated depended on the duplicates. Had I allowed this script to run, it would have deleted most of the packages.

I found a python script I had written when an upgrade from FC7 had failed because of power failure. The script created a file with the list of older duplicates, which could be used as an input to rpm as below:

# rpm --nodeps -e `cat deleteList.txt`

Finally, I cleaned up the failed yum transaction and ran:

# yum --releasever=17 --disableplugin=presto distro-sync

This time, the upgrade finished cleanly. Although it took about the same time and effort as the preupgrade, I enjoyed solving these problems a bit more.

Firefox crashes

As on the desktop, firefox crashed with even a few tabs. It seemed to be a random problem. However, I downloaded Firefox 12 from Mozilla and firefox works fine!

First Experiments with Ubuntu 11.10

The upgrade to Ubuntu 11.10 went through smoothly on Lenovo S10-3 netbook. It looks very similar to 11.04. An application would normally use the full screen as expected. I tried to start a new application and couldn't. No matter where or how hard I pushed the mouse, the dash would not appear.

Finally, I remembered the 'window' key! Having found the workaround, I have not had a chance to experiment with it more as I normally use Arch Linux on the netbook and Fedora on the desktop.

At IIT, Ropar, the default desktops are Ubuntu. So, I decided that I would try to upgrade my LTS desktop to oneiric. I certainly wasn't going to upgrade in steps. The choice was between a fresh installation or an forced update and do what is not recommended - strongly.

The latter is more fun and I hoped to learn something. I changed the sources file to use the oneiric repositories. My installation was fairly simple so I did not expect any significant issues. At worst, some packages would be broken.

Unfortunately, the upgrade failed - too many errors. The last error was a fairly simple error of some conflict in a file in two packages. I made the mistake of rebooting inspite of the broken packages!

The system no longer boots - not even grub menu :(

Well, I will now have to do a fresh installation - and haven't understood what went wrong.

Arch Linux - experience with kernel 3.0

Upgrading to kernel 3.0 on Lenovo S10-3 did have a couple of issues. The first issue was that the wifi was not functioning. This was easily solved by following the instruction on Arch Wiki and blacklisting the conflicting bcma driver.

The other issue was that waking from sleep is not working any more. This is in spite of  adding 'intel_idle.max_cstate=0' to the kernel boot options. Hence, I reverted to kernel 2.6.39 until I have time to examine the problem in more detail and see if  there is a workaround.

Incidentally, I have been facing erratic behaviour with waking from sleep on Lenovo S10-3 with Ubuntu 11.4 with 2.6.38 kernel and Fedora 15 with 2.6.40(!) kernel. I thought that the reason may be that I am using x64 versions of Ubuntu and Fedora whereas on Arch Linux I have installed the i386 version.

Update: A silly mistake in Fedora 15 grub.conf file. I had typed 'state' instead of 'cstate'. Initial tests of sleep work on Fedora 15 with kernel 2.6.38 but not with 2.6.40 (presumably the same as mainline 3.0). I will have to check if the behaviour of waking from sleep on Fedora is consistent as on Arch Linux or erratic as on Ubuntu.

Update 2 -  I found that there is a bug report of the waking from sleep issue on Debian bugs list when upgrading from 2.6 to 3.0.

Enabling the brcm80211 driver in Fedora 15

[Update: As pointed out by Dean Brettle and, subsequently, by AM in the comments, rpmfusion added the kmod-staging package. The steps below may be useful if I ever need to compile a single module again.]

The driver for the wireless on the Lenovo S10-3 is included in the staging directory of the kernel. Fedora 15, unfortunately, did not enable it and wl.kmod or brcm80211.kmod were not available from rpmfusion repository either.

So, the best option was to download the source and enable it. The instructions to download the source of the kernel and prepare the source tree were available at http://fedoraproject.org/wiki/Docs/CustomKernel. The kernel source was now in the rpmbuild/BUILD/kernel-2.6.38.fc15/ directory.

The following steps helped to build only the desired module

1. Modify the .config file to include
   CONFIG_BRCM80211=m
2. Make sure that the following details in Makefile are as per the running kernel, e.g.
   VERSION = 2
   PATCHLEVEL = 6
   SUBLEVEL = 38
   EXTRAVERSION = .6-27.fc15.x86_64
3. Run the command
   $ make drivers/staging/brcm80211
   This step prompted me to enable the additional option -
   CONFIG_BRCM80211_PCI=y
4. $ make M=drivers/staging/brcm80211/
5. $ sudo make modules_install M=drivers/staging/brcm80211/
6. Looking at the OpenSuse's /lib/firmware/brcm, I created two softlinks:
   $ sudo ln -s bcm43xx-0-610-809-0.fw bcm43xx-0.fw
   $ sudo ln -s bcm43xx_hdr-0-610-809-0.fw bcm43xx_hdr-0.fw
7. Add the driver:
   $ sudo modprobe brcm80211

The wifi card was now recognised and working. It took just a few minutes instead of hours for compiling the full kernel - even if it took a few days to understand and reach this point!

Notes:

• If step 2 was not done, modprobe failed to load brcm80211 - invalid module format. The log reported a more meaningful problem - version magic mismatch and the driver was unloaded.

• If step 6 was not done, the driver was loaded but the log reported an error that failed to find firmware /lib/firmware/brcm/bcm43xx-0.fw.

• While searching (unsuccessfully) for an explanation for why steps 3 and 4 were needed, I came across a somewhat different way to compile just one module - http://www.debian-administration.org/articles/640.

OpenSUSE Tumbleweed

I installed the live KDE version of OpenSUSE 11.4 and then added the Tumbleweed repository as documented here on a Lenovo S10-3 netbook. The installation and updates were pretty straightforward. It updated to 2.6.38 kernel. However, the wi-fi was not connecting. I tried to install the broadcom-wl driver and there wasn't one for 2.6.38. I built the non-free driver for 2.6.38 and installed it. It still did not work. Since I could not understand the cause or diagnose it, I removed the hidden network option on the router. The wireless connection worked!

The next step was to realise that there was an open source broadcom driver brcm80211 included in 2.6.38 kernel and the wl driver wasn't needed. The wifi network worked well with the open source driver. I made the network hidden on the router and the network still worked! I am not sure if this workaround is needed every time a new wifi connection is needed - something to be verified.

My preference is for a rolling distribution which distributes delta packages. I still like Arch Linux and the ease with which I could hold back X11 server to an older version for Intel 845/945 graphics cards. It has been successful in demonstrating the power and feasibility of a rolling distribution model. It is great to see OpenSUSE adapt it and OpenSUSE were, probably, the first to offer delta packages. So, I plan to stick to Tumbleweed on the Lenovo Netbook with the KDE Plasma Netbook interface.

I will continue to keep Arch Linux on my mother's computer which has Intel 945 graphics.

My concern is that since OpenSUSE is continuing to work on discrete distributions as well, they may be complicating their environment. I hope they do not drop Tumbleweed in the process.

KDE 4.6 on Arch Linux

The first reaction to KDE 4.6 on Arch Linux was that it looked nicer. Minor display issues in Netbook workspace were gone.

Earlier, it would often display only one favourite application. If one dragged the icon to the left edge, all would show up. Some times the application categories would not show at the start. Searching for an application would make them appear.

The graphics effects would cause the desktop to hang. Now, that is working very well. One nice new feature is that one can close a window on the screen displaying all the windows.

The power button of Lenovo S10-3 would not trigger the power off/suspend menu on KDE Plasma Netbook workspace. This problem is sort of resolved. On first attempt, it takes about half a minute to show the dialog. But if one cancels it and retries, the menu is displayed immediately.

But there is a new 'bug'. Wifi does not connect automatically. It waits for me to click ok on the password. I will try with a fresh KDE session just to confirm if the problem is related to some change in the configuration files.


Update:
The 'bug', where the network manager fails to start the wifi without any user interaction, seems to be that I was not using the KDE Wallet. Once I used the KDE Wallet, network manager starts the wifi without any prompt.

Revisiting Smeegol

My experience with Smeegol had been pretty good except for getting the wifi to work. Now that the wifi on my Lenovo S10-3 was normal, I decided to retry Smeegol. Smeegol 1.0.5 is still based on 11.3 OpenSuse repositories. I could not find one based on the in-progress 11.4.

I could install the wifi driver. However, I had to build it from source as the Packman repository had the binary only for the i586 version. The official Meego boots in under 30 sec. Smeegol took 40 sec. Not a big difference. Unlike Meego, Smeegol uses gdm to launch uxlaunch. The script startmoblin is run using the /usr/share/xsessions/smeegol.desktop.

The directory /etc/gconf/gconf.xml.moblin contains the gconf options for smeegol desktop.

It seems straightforward. So, the question is why is meego desktop not working in Fedora14? It just shows a white screen on Lenovo and on my desktop system. It has been reported as a bug in Fedora's bugzilla. Need to dig and try to narrow down the cause.

Lenovo S10-3 with 2.6.36 - Resuming after Sleep

After installing the 2.6.36 kernel from Arch Linux, I found that the system would not resume after sleep. I spent the last couple of days trying to understand the problem. I installed 2.6.36 kernels from Fedora's Rawhide and from Ubuntu's kernel-ppa.

Fedora kernel had the same problem resuming. However, it worked on Ubuntu. On Ubuntu, I had been using 'intel_idle.max_cstate=0' as a kernel boot option. This was a work around for Ubuntu 10.10's boot problem on S10-3.

Removing the intel_idle option from Ubuntu's 2.6.36 kernel resulted in the S10-3 booting perfectly well. However, it showed the same inability to resume from the suspended state.

Adding that option to Arch Linux kernel resolved the resuming from sleep problem on this platform as well. Needless to say, the same option worked on Fedora kernel as well.

Lenovo Wifi resolved

I decided to test the ideapad-laptop driver from 

git://kernel.ubuntu.com/ikepanhc/ideapad-laptop.git

I had thought that I would just need to compile and install just the driver. I, instead, needed to rebuild the kernel. The version was 2.6.36rc8. I also downloaded and installed the Broadcom wl driver, which, however, does not seem to be recognized - so far.

The nice aspect was that bluetooth was recognized and indicated that it was disabled. I enabled it and the wifi led worked!

I rebooted in the normal kernel and the wifi worked without needing to suspend the system and resume it.

It feels good to be able to resolve this problem without having to install windows - thanks to Ike Panhc and David Woodhouse. 

For more details about the driver: https://patchwork.kernel.org/patch/119373/

Lenovo Wifi behaviour still magical

The behaviour of the wifi on the Lenovo S10-3 continues to puzzle me. It continues to be magical and that feels uncomfortable. The magic is that I need to make the netbook sleep (Fn+F1) and then press the power button. Usually, the wifi will work. Occasionally, I need to repeat these steps.

This worked on Ubuntu netbook edition. It worked on Meego and Arch Linux. But I just haven't been able to get this magic to work on Smeegol from Suse.

There must be a difference. Looking at the logs, I found that if ACPI has an exception on waking from sleep state S3, bluetooth is recognised and wifi starts to work. The following extract from Ubuntu log file is indicative of each os.

Oct 27 14:55:36 anil-laptop kernel: [  690.604909] Initializing CPU#1
Oct 27 14:55:36 anil-laptop kernel: [  690.729043] CPU1 is up
Oct 27 14:55:36 anil-laptop kernel: [  690.729260] ACPI: Waking up from system sleep state S3
Oct 27 14:55:36 anil-laptop kernel: [  696.684074] ACPI Exception: AE_TIME, Returned by Handler for [EmbeddedControl] (20100428/evregion-474)
Oct 27 14:55:36 anil-laptop kernel: [  696.684099] ACPI Error (psparse-0537): Method parse/execution failed [\_SB_.PCI0.LPCB.EC0_.DSSV] (Node f7023f78), AE_TIME
Oct 27 14:55:36 anil-laptop kernel: [  696.684174] ACPI Error (psparse-0537): Method parse/execution failed [\_WAK] (Node f7023030), AE_TIME
Oct 27 14:55:36 anil-laptop kernel: [  696.684370] ACPI Exception: AE_TIME, During Method _WAK (20100428/hwsleep-601)
Oct 27 14:55:36 anil-laptop kernel: [  696.719069] PM: early resume of devices complete after 1.912 msecs

However, on Suse's Smeegol, I just could not get it to show this exception.

On Ubuntu, the system wakes up whether the power button is pressed or sleep hotkey (Fn+F1) is pressed again. However, if the power button is pressed, the ACPI exception will normally show up and wifi works. But if the sleep hotkey is pressed, the exception rarely occurs and wifi does not work.

The next step would be to try the lenovo-laptop driver and hope that the wireless hotkey works and I do not need the above magic any more.

Experimenting with Lenovo S10-3 - Smeegol

I downloaded the x64 version of the iso file. It worked without a hitch and I could install the Suse's version of MeeGo on the S10-3. Unlike the 'official' MeeGo, this version has the full Suse repository available and applications like OpenOffice are already installed.

It uses gdm instead of directly using uxlaunch. The X server in this case is owned by root and not the user as in the case of MeeGo. However, this also implies that it is multiuser unlike MeeGo at present.

It is an attractive and a nice alternative to the Ubuntu Netbook and the KDE netbook workspace.

The wifi issue continues to trouble me. LED is still off and I need to suspend the os (Fn+F1) and bring it back to life before it will recognise the existence of Bluetooth and enable wifi to work. Unfortunately, so far, with Smeegol, it does not recognise Bluetooth in spite of multiple attempts at suspending and resuming the system. I need to dig deeper and try to understand what is happening or supposed to happen upon resumption.

The other irritant on Smeegol so far - if I accidentally press the power button(and I seem to do that often while moving the netbook), a message is displayed that it will power-off after 30 seconds. However, it seems more like 3 seconds!

Figuring out how these buttons actually work along with the configuration files and code relating to them is one of the many pending items on my list of items I want to understand:(

Mouse - Click to Move

I had switched my parents to Fedora 4 years ago when I moved to Delhi area from Goa. One application from Windows98 remained - Freecell. My mother loved to play it and she just could not use drag and drop functionality. She has Parkinson's disease.

Then, after an update, I noticed 'Click to Move' option in AisleRiot. Had I missed it earlier or was it a new feature - I don't know. But after that, even Windows' freecell lost its place on the computer.

The reason I recalled this incident was that I was finding it extremely hard to use drag and drop on Kpat on the Lenovo netbook using the touch pad. Some operations with mouse can be very-very difficult as we age. KDE4.5 has more keyboard short cuts so I found that to play Kpat, I could use them. Not the best of solutions but a workable solution. An option for 'pick and drop' as an alternate to 'drag and drop' would have been very useful.

Sadly, keyboard shortcuts are just that. They are not very useful for older people who may have problem with mouse as well as remembering which key does what in which application.

S10-3 - Unusual wifi problem

I am not sure when exactly it happened or why. I was playing around with the physical switch to turn the wireless off and using the Reliance broadband usb modem. It was frequently disconnecting and speed was low. It probably was a network issue. But I wanted to be sure that it was not a Lenovo netbook or OS problem.

Anyway, when I tried to use wifi, wireless indicator LED was off and connection failed no matter how hard I tried. The LED hasn't been on since.

The symptoms were similar to a problem reported in Ubuntu forum. However, in my case wl driver is used and not ath9k. I tried the same solution and a simpler variation worked.

Bluetooth capability of the netbook is not discovered at boot time. Wifi doesn't work. The wireless indicator LED remains off. However, on Ubuntu, MeeGo and Arch Linux, if the system is suspended and then restarted, Bluetooth device is recognised. But even more important, wifi just works. However, the LED remains off.

I went to the shop from where I bought it - no luck. They fiddled with hardware, reconnected some cables and symptoms remain the same. They wanted Windows to do anything more.

I thought I would at least email the problem to Lenovo support. No luck - the serial number of my system is rejected as being invalid. I hope it is because the sale is not yet in their system. Oh, well, will have to follow-up.

In the process, I found that a driver ideapad-acpi(ideapad-laptop) has been submitted to the mainline kernel. So, the functions for enabling/disabling wireless should work and that may clear up what is definitely a bios bug.

Meanwhile there is one advantage though - no unauthorised user can use of the wifi on my netbook :)