Hotkey-setup still a major problem

My "fix" was actually a fluke -- the system is back to freezing on boot every time when hotkey-setup is loaded. Apparently there is a known major problem with the sony_laptop module in 10.3 - kernel panic, bug 327114. So I had to disable the module again.

Oops -- hotkey-setup not quite fixed yet

Oh, no! Hotkey-setup locked my machine on boot again today :( It re-booted fine, though, which is progress over the previous settings where it locked up every time. Makes me think that the deadlock is with a service of the same priority as hotkey-setup (I have about 10), because there is no guarantee of the order in which they start even with RUN_PARALLEL set to "no". If this happens again, I am going to look into figuring out the dependency.

Emacs keybindings problem solved

I am on a roll today ;-) After some googling, I found a solution to emacs key binding problem. The syntax in Emacs 22 seems to have changed a bit. To get the old "C-x x" and "C-x g" bindings back, you need to add the following to your .emacs:

(global-set-key (kbd "C-x x") 'copy-to-register)
(global-set-key (kbd "C-x g") 'insert-register)

Hotkey-setup freeze and brightness problems solved

These two problems were indeed one. After some more searching around and trying I found out two things: setting RUN_PARALLEL in /etc/sysconfig/boot solves the hotkey-setup freeze problem. Once hotkey-setup is loaded, I can adjust brightness on my machine again, kpowersave allows me to configure it, and my screen dims on AC power. The only annoyance is that with RUN_PARALLEL="no", system boot time is slower.

I am guessing that there is a deadlock with some other service or services at boot. In theory, this could be fixed if I knew the culprit, because then the .depend.start file could me modified to ensure that hotkey-setup is started strictly before or after it. But there are about 15 services in runlevel 5, it's too much hassle to identify one that causes this trouble (that would require testing lots of combinations, and, unless I am very lucky to discover it quickly, many hard reboots). I sent a query to the opensuse mailing list in hopes that someone may have a guess, but other than that I am going to just live with it.

Problem 20: No brightness adjustment on AC power

I thought I have discovered all the issues, but it looks like there is more to come :( The newer version of powersaved thinks that my hardware does not allow for screen brightness adjustment. Thus, when I am on AC power, my screen remains unnaturally bright. But it worked perfectly in 10.0, dimming my screen. The spictl utility says that /dev/sonypi does not exist, and refuses to adjust screen brightness, either. This may be related to the hotkey-setup being disabled - I will try to re-test this first.

Suspend/Sleep update

After a re-test, it looks like "suspend to disk" is working, but "suspend to RAM" isn't - my machine is not on a white list. A small niggle: in the past, after "suspend to disk" when I went to boot up, the system displayed the GRUB menu. So I could boot up my suspended SuSE system, or choose to boot into Windows instead, and boot into SuSE later. But now it goes right into booting Linux, without giving me choice. I would prefer the old behavior to be restored - will investigate later.

Summary of not yet resolved problems

It has been a month, and I thought I should clean up a bit. It looks like many of the original problems were resolved/went away with the kernel update that fixed the kernel uninterruptible sleep problem, but others remain. Here is a summary of not sufficiently resolved issues

• KMail: the original problems (key bindings crashes on startup) were solved, but a new problem appeared. Now every so often KMail will refuse to display a message, showing it as empty. The only way to fix it is to close KMail and re-start again.
  
• Emacs: I removed (common-display-european 1), and this seems to have sped up the startup. But C-x x binding is still not working. I hadn't had time to look for fixes, though I
  
• Very slow YaST. I disabled automatic update on all repositories except for the "online update". This speeds things up, but not quite enough - the startup time is still 5 minutes or more. On the web there are recommendations on running a YaST cache defragmentation, but the bottom line is, it's extra bother, and it is just too slow. I can only hope that someone sees light and comes up with a better solution for 11.0
• iw2200 and NetworkManager: I solved the problem by inserting a line that explicitly loads ipw2200 module on startup into /etc/init.d/network. But this means that I will have to re-do it next time I update the system, unless the problem is solved in 11.0
• Hotkey-setup locking up the system on boot: as far as I know, not solved yet, though I should re-test it after the recent kernel updates
• System freezes on sleep: not re-tested yet
  

Problem 19 with solution: DVD playback

By default OpenSuSE has DVD playback disabled. There is an easy one-click install for KDE here. But my first attempt at using it failed: it requires a yast2-metapackage-handler. For some reason, it was not installed by default. It then refused to install from the web - because I originally installed from the DVD, yast2 insisted on bringing it and all its dependencies from DVD, too, even though I told it to use the web. Eventually I found the DVD and installed from there. Once this was sorted out, the DVD playback installed smoothly, and DVDs play reasonably well, though the image is slightly blurred at times. Interestingly, the metapackage also included a slightly different version of java plugin which fixed my previous java with Firefox problem.

Bad camera connection (almost) solved

Today's batch of SuSe updates brought in a patch to libgphoto2 with Canon driver enhancements. It does not fix the camera issues (problem 15) completely, but it improves the situation a whole lot. I still have to re-try connecting several times, but it works after a minute or two, rather than half an hour as before. This makes Digikam usable again, even if not working perfectly smoothly.

Problem 18: weird acpi behavior

If I boot into the system with the battery almost, but not completely charged (i.e. every evening, because my battery seems to discharge slightly overnight), acpid and/or kpowersave seem to go into overdrive. The status changes every 1-2 seconds from "plugged in" to "unplugged", resulting
in constant annoying beeping. But after about 3 minutes the status settles to "plugged in", and no more changes happen, no matter how long I work. There is a possibility that this is a hardware problem, but it seems strange, since these sudden status changes don't happen any time except immediately after system boot. I didn't notice Windows doing that (though I boot into Windows very rarely), and it only started after I moved from 10.0 to 10.3. Cannot find anything relevant on Google at the moment, so I am stuck with this.

Log update

After looking around some, I decided to try to use syslogd-ng instead of syslogd for logging, since it is supposed to be the new generation logger. The best I can tell, it is now configured to log all messages, including boot messages, to /var/log/messages. This will get tested next time I re-boot the system.

Process hangs/unclean shut-downs update

OpenSuse released a software update today which is supposed to fix my proceeses hung in uninterruptible sleep/unclean shutdowns bug. The YaST updater started downloading it - and, surprise, surprise, got hung in uninterruptible sleep state! I desperately wanted the patch by then ;-) After another unclean shutdown and reboot, it turned out that either Virgin's network or OpenSuse site was overloaded, so the connection was breaking after 1M or so of downloads (I needed to download 20M). After about 2 hours I could finally download and install the patch. Now time will show if it fixed things.

Problems 16 and 17: system freezes after going to sleep, and missing info in the logs

My system went to sleep (suspend to disk, at a guess) because the battery was low, and when it tried to wake up, it froze completely at some point, refusing to respond to keyboard. At that point it was showing me a suse logo with a progress bar, but hard drive activity stopped completely, and nothing but a hard shutdown worked. Suspend to disk worked without any problems whatsoever in 10.0. This might be an instance of my general lockdown/uninterruptible sleep problem, or a new problem, but there is no way to tell. It turns out that messages from boot sequences don't go to /var/log/messages anymore - in fact, they are not saved anywhere at all, the best I can tell. This is a change from my previous configurations, and not something that I did. I guess I will have to figure out what's going on with log files soon, if I am to diagnose any of my issues (hotkey-setup, which is still disabled due to system lockdowns, is another one)

Problem 15: bad connection to the camera

I wanted to download pictures from my Canon S60 using digikam. This worked fine in 10.0 - there would be an occational connection glitch, but most of the time it just worked. This time it took me about an hour, with 10 connect-reconnect-try different buttons attempts to download the pictures. An later attempt wasn't any more successful (I gave up before I could do anything with it). No warning messages in the log - in fact, it faithfully reports every time that the camera has been detected, and then DigiKam says "unable to connect to camera". More investigation to come in the future...

Bootloader problem solved

After re-installing grub via YaST, I could boot XP again. Interestingly, the boot menu listed XP as booting from hdb, when I don't even have a second device. Once I fixed it to be hda1, everything booted. I don't quite know how it got there - I suspect that suse install broke something, though there is a small chance that I did it when I was fixing my boot sector manually. But the latter is again a reflection of a known problem - "repair installed system" does not work in SuSe 10.3 (bug 329702), so I could not fix the menu except by using command-line tools, and it's easier to make a mistake with those.

It's marked as FIXED in Novell's bugzilla, but really it is not. The "fix" is to download and burn an extra cd which uses command-line tools, rather than a real fix to YaST gui.

Problem 14: boot loader broken for WinXP

I just tried to boot Windows XP, and got a message
grub Error 23: Error Parsing Number

I didn't modify anything recently -- apparently some install or update did something to GRUB :( There are some reports on the Internet about this, but no clear solutions. I tried to re-install the bootloader with YaST -- will see if it helps.

Unclean shutdowns update - we've got a little problem here :(

OK, I know what's going on with my unclean shutdowns, but have no way to fix it. It's a kernel bug in suse 10.3. It's bug 336669 Processes randomly enter uninterruptible sleep state. Apparently it affects both the most recent kernel, and earlier kernels, too, so there does not seem to be a way to do anything about it except downgrade to 10.2 or earlier. As someone said on the opensuse mailing list, "this is very uncool". If you are thinking to install 10.3, I suggest that you wait until this is fixed - and I can only hope that I don't fry my hard drive before that happens.

Problem 13: unclean shut-downs

This is a very appropriate problem number ;-) Once every 3-4 shutdowns my KDE refuses to close - it just seems to sit there doing nothing; all windows by KNotes are gone, but other processes such as kio_imap (remained of Kmail) or even Konsole still run in the background. If I switch to a text console and reboot (using either crtl-alt-del or reboot command), everything seems to go well till almost the last minute when it suddenly prints something like
"oops... could not unmount /: device or system busy"

I can't see this in the log because the logger shuts down before that. My computer turns off immediately after, and then replays hundreds of transactions with ReiserFS. Nothing suspicious appears in the log before the logger quits.

I checked my hard drive with smartctl, and it reports that my hardware is OK (of course, it could be wrong). The problem is not easy to reproduce, either, so I am now living with my fingers crossed that it does not kill my hard drive one day.

Wireless/Networkmanager problem solved

As it turned out, with Network Manager my wireless issue Iproblem 12) was worse than I had thought - it absolutely refused to load the ipw2200 module on boot, so there would be no wireless until I run "modprobe ipw2200". I solved it by modifying my /etc/init.d/network, inserting the modprobe command on top as discussed on the opensuse ndiswrapper page, "loading ndiswrapper on boot" section.

Wireless and Network Manager

OK, so Network Manager works fine. But it does not solve my wireless problem - if I boot up with my wireless switch off, there is nothing I can do to turn it on. modprobe ipw2100 does not seem to help. Going into yast, fiddling with settings so that it re-runs all the scripts helps, but that is a hassle. I wonder if this is because I was forced to disable that hotkey-setup module? More investigation to do, I guess.

Problem 12: wireless off on startup means no wireless at all

OK, here we go with a new problem: I just booted a machine with my wireless switch in the "off" position. No wireless interface was created, and turning the switch to "on" did not help. I had to go into YaST to try to change configuration settings, after that it run something (modprobe, maybe?) which caused the my wireless card to be visible. After that the switch worked just fine. It was not a problem in 10.0 - there I could easily boot with my wireless off, and then turn it on later. I do see a message about failing to set up eth1 during boot. Interestingly, when I was running installation, the kernel saw the card even when the switch was off - so it should be detectable. Haven't figured out what's wrong there yet. I wonder if it would be better with Network Manager?

Wireless with KInternet

Just tested wireless with KInternet - this is one thing that seems to be working better than 10.0. I needed to re-run the configuration, but once I did it, it seemed to remember my WEP key just fine, where in 10.0 I had to re-enter it often, for no clear reason, because it seemed to think that network had no encryption, even though everything was set up properly in YaST. I will try the nextwork manager sometime later - I'd like to switch to a newer program if I can.

Logging and konsole problems solved

I have logging again (my problem 4) solved with that change to apparmor configuration), and fonts on my konsole (problem 5) are readable again after I changed sizes and saved the session.

Updating from 10.0 to 10.3

I decided to update from 10.0 to 10.3 directly. I skipped 10.1 because another person in our group had endless issues with software updater crashing. I skipped 10.2 because when I installed it to my Mom's laptop, it gave me a lot of trouble with networking (a PCMCIA card didn't work with an annoying log message "get a better card!"), as well as continued having trouble with the software updater. By my kernel and packages have gotten old, I cannot install the newest Skype, so I decided it was time to upgrade my laptop.

Unfortunately, lots and lots of things don't work. I find it very annoying - on this particular computer, both SuSE 9.3 and OpenSuse 10.0 were very stable and didn't give me any trouble at all. But now it seems that all the improvements are actually making my system less stable, and applications I depend on, like emacs, are changing functionality in ways that make my life more difficult.

Problem1: KMail. Crashed on startup, before I could do anything. After some experimentation, I determined that my cached IMAP account was a problem. I solved this by deleting ~/.kde/share/apps/kmail/dimap but then had to endure a repeat sync. Since I have over 10,000 messages across all my folders, and my internet is currently very slow (shame on Virgin.net, I am changing soon), this took ages. Then all my keybindings were lost. I re-defined several, but Shift-R still refuses to work for "reply to all", for unclear reasons. At least it's not "reply without quote" - that was the new binding which confused me for quite a while.

Problem2: freeze on reboot. It took me 2 tries to even figure out what happens. As soon as hotkey-setup starts on boot, my caps-lock and scroll-lock lights start flashing, and the system freezes completely. This is not a problem in fail-safe mode, but I hate to lose my ACPI. So I temporarily disabled hotkey-setup in YaST until I can find a better solution. I think I can live without the special function keys ;-)

Problem 3: Java not working in Firefox. I have java version 1.5.0_13, and Java plugin is installed. But when I load pages in Firefox with applets that worked on 10.0, nothing happens - the page stays empty. These pages work in Konqueror, so it appears to be a Firefox problem. Unfortunately, there is no Java console for this java build, so I cannot help it. I am putting this off until a better time - or possibly until I decide to move up to 1.6

Problem 4: no logging. Absolutely no messages are appended to /var/log/messages. Apparently this is a known problem if you are running syslog. The default for new installs is syslogd-ng, but since I upgraded from an earlier version, I had syslogd enabled. The fix is to add "k" to the line starting with "syslog" in /ets/apparmor.d/sbin.syslogd. I just put it in, but I haven't tested this yet - I need to reboot first. More info is in bugzilla entry "syslogd can't lock syslogd.pid"

Problem 5: very small fonts in konsole. My fonts looked just fine on 10.0, but now they are tiny and next to impossible to read. Changing the global settings in the KDE control centre does not have any impact on my Konsole. At a guess this is because it is part of a previously saved session. But then why are my fonts suddenly unreadable? I am planning to increase the font manually on each tab (6 of them, sigh) and then try to save the session, see if it works.

Problem6: very slow emacs startup. Loading emacs now seems to take ages. First I had to spend a while cleaning my .emacs file, because it turned out that lazy-font-lock option was removed and it was crashing it. Once I finally found the problem, it now works, but seems to load an incredible number of packages and takes a minute to start (where it used to be a couple of seconds). I guess I need to re-check my .emacs again to see if there is anything else to remove.

Problem7: C-x x binding gone from Emacs 22.1.1. I am very used to using this one, copying text into register and pasting it, sometimes using 2 or 3 registers. But now this binding is gone altogether. I need to investigate if I can bring it back

Problem8: extremely unstable KDE. During this day, I had at least 3 instances where Konqueror, Konsole and the main panel all stopped and refused to respond to keyboard and mouse, so I could not even log out. I could change to a text-based console and reboot, but that was it. No solution yet - hope to see something in system logs after I reboot.

Problem9: crashes during system shutdowns. Twice today when I restarted the system, the process went fine for a while, and then the system rebooted itself before unmounting the hard drives. No data damage so far, but it had to replay a huge number of transactions on my root partition. Again, no way to trace it until logging works - hopefully I will see something after the logger works.

Problem10: random lockups in SuSEConfig. Again, had it freeze 4 times at different scripts: either fontconfig or gtk2.config. The scripts just went into "wait" state and hang on there indefinitely. Killing them and re-running the script after reboot seemed to fix the problem. Again, need logging to find out more

Problem 11: very long startup times for YaST package manager. It seems to want to download lots of stuff because it wants to sync with many online repositories. I am planning to turn off automatic refresh on most of them, but I want to figure out first if there are any where the automatic refresh is essential (e.g. the update repository?)

Problem11a: lots of things are not checked yet. Will my camera work? I had a foiltex package, which depended on tetex, and may need to be updated for livetex. Beegle indexer either does not work, or takes 100% CPU - I didn't use it before, so I need to read the documentation and find out what's going on there. Mike not tested yet, nor is my wireless. After my 10.2 experience with Network Manager (on a different machine, sure) I am reluctant to try it here until I at least fix the logging, and preferably also fix other instabilities. I guess there will be more posts in the future ;-)

What is this blog about?

I finally decided to start posting information about my experience running OpenSuse on a Sony Vaio PCG-V505DP machine. I have been with SuSE since early days (I started with SuSE 6, or maybe even 5 - don't remember by now), and it has been the most stable distribution that I knew of that usually supported my hardware where others didn't. Sadly, it's no longer true - since 10.0, every new version seems to come up with additional problems, and there are more problems with each version, at least on the laptops I use. So I decided to start this, to have a permanent record of problems and solutions.