Being disgusted with SLT, I obtained a Dialog WiMax connection. Getting the connection was like pulling teeth and service disruptions were worse than with SLT and customer support just as bad. Dialog Telekom, once the premier ICT company in Sri Lanka and the largest company on the Colombo Stock Exchange is now in the red. Dialog Broadband is largely responsible for that. Their services to descended to farcial levels and I gave up on them and switched to Lanka Bell.

Lankabell broadband (which is also WiMax like Dialog) is worse than the other two. Hardly a week passes by without a disruption. Usually it’s hours before service is restored. Sometimes it takes them more than a full day to get online again. At other times speed drops to a crawl.

In hind sight SLT ADSL was never that bad. Though it was slow and customer support was terrible, when they had service disruptions, they were pretty quick to fix it. That’s probably because they have a larger customer base and get lot of complaints. Dialog WiMax has only about 5000 users while LankaBell has only about a 1000. Looking at the poor standards they maintain it’s not a suprise.

I already had version 6.1 from a previous manual installation. It works fine even on F10. Since I have tweaked all the settings to make the IDE behave exactly the way I want it to, I have no use for another 6.1 installation. More so since I also happen to have a 6.5 nightly build.

My next approach was to download the IDE+JDK bundle from sun. Normally I don’t use that bundle but on this occasion the netbeans mirrors turned out to be painfully slow and anyway I had to update my JDK so I went ahead with it. Unfortunately that IDE doesn’t start up. Bails out with :

Gtk-Message: Failed to load module “canberra-gtk-module”: libcanberra-gtk-module.so: cannot open shared object file: No such file or directory

Incidentally this same error pops up when I try to launch the zoiper softphone. There is a libcanberra-gtk2-0.10-2.fc10.i386.rpm installed. And that RPM is supposed to provide /usr/lib64/gtk-2.0/modules/libcanberra-gtk-module.so so I suspect that zoiper and netbeans are looking for a 32 bit version.

Zoiper; I can manage with out. Therea re other softphones that are nearly as good but can’t say the same about netbeans. It’s better than most other IDEs that I have encountered so I had no choice but to install the 32bit version of libcanberra along with it’s dependencies and that seems to have solved it.

The first problem was that the computer wouldn’t start up. Grub complained about a wrong partition (error 22 if I remember correctly). I solved this by putting the DVD back into the drive and booting in rescue mode and the editing grub.conf - Even then,  the plymouth splash screen didn’t appear (it worked perfectly on my VirtualBox and Qemu guests). Solved it by adding  ‘vga=0×0318 rhgb quite’ to grub.conf

On boot up, I cannot login to sugar at all. This is not any different from Fedora 9 but I didn’t worry too much about it then and I am not worrying now. KDE does start up but the menus are missing. I was a long time KDE user but gave up on it after seeing the new menu, so that doesn’t worry me either. Gnome works ok.

The next problem demanding a solution was Atrpms - the version number of the preview release is 9.93 not 10 there isn’t a directory named 9.93 on the atrpm servers. Resulting in the following:

Loading mirror speeds from cached hostfile
* atrpms: dl.atrpms.net
* rawhide: mirror.yandex.ru
http://dl.atrpms.net/f9.93-x86_64/atrpms/stable/repodata/repomd.xml: [Errno 14] HTTP Error 404: Not Found
Trying other mirror.
Error: Cannot retrieve repository metadata (repomd.xml) for repository: atrpms. Please verify its path and try again

This was solved by editing the following line in the yum.conf file and hard coding the version number (expect more trouble when version 11 comes)

baseurl=http://dl.atrpms.net/f$releasever-$basearch/atrpms/stable

When trying to login through VNC the following message pops up and gnome wouldn’t display anything more than a blank screen

Killing bonobo-activation-server did allow me to login

Most of the issues that I had with the preview are still there with the gold release as well. KDE still starts up without the menus but with that fedora “Could not start ksmserver.” Check your installation message which no one seems to have a solution for. Sugar doesn’t start at all. The screen just flickers and returns to the login screen. Lot’s of other things are still broken.

Even as I was typing this, I had a crash and the update introduced new problems like messing up my grub and causing a boot failure. It just went grub grub grub grub eternally without going to next step and I had to put it right by booting in rescue mode. Even then I didn’t get the plymouth boot screen. Well at least power saving is back.

Will make another post soon and share what ever solutions that I found.

I didn’t know that dd had a ’sparse’ option and have often had to wait until it created images for Qemu and encrypted file systems. All you need to do is to make use of the ’seek’ option in dd.

Last but not least, I didn’t know that losetup (and mount) had a skip option. That means you can mount virtual partitions inside Qemu images using losetup.

So I learnt three important things in a day and I thought I was an old dog. I am slightly embarrassed really. All these three things were on a post by Karsten M. Self. in a Xen Mailing list.

The last discovery (losetup offset option) is specially interesting.  You can make encrypted file systems even more secure by adding a random number of junk bytes at the start of the image. That would mean the ‘file‘ command cannot be used to find out what kind of a file you have and anyone trying to crack it will not only have to deal with the encryption algorithm but also try to guess that random number you have chosen for the padding.

Lots of other geeks are religious though. They would swear by Java, PHP or some poor souls who need to be rescued even swear by primitive things like C# or .NET (windows really does sucks (No that has nothing to do with religion , it’s a scientific fact)).

So this tag, religion is about such arguments like ‘PHP is better than Java‘ and ‘perl is better than python’ (anyone want to compare Fortran and Python?).

Well what’s really better, is it PHP or Java? well it’s a matter of horses for courses (but PHP has recently lost the plot a bit with an API that is way too big, but then Java is showing it’s age too).
While still on the subject of religion and religious wars you might ask, Is Fedora better than Debian?  Yes Fedora is miles ahead of Debian (but even so, Debian is much better than Windows)

The installation itself was straight forward, I just used ‘rpm -i VirtualBox-2.0.4_38406_fedora9-1.x86_64.rpm’ but once installed the application failed to start up. No error message was produced. When that happens the best thing to do is to try executing it from the command line. The the following error was observed:VirtualBox:

supR3HardenedMainGetTrustedMain: dlopen(”/usr/lib/virtualbox/VirtualBox.so”,) failed: /usr/lib/virtualbox/VirtualBox.so: undefined symbol: _ZN10QStatusBar9showEventEP10QShowEvent

Googling revealed that this is something that happens due to the kernel module not being loaded - which is due to an incompatibility with the kernel version on that machine. The module can be built with  ‘/etc/init.d/vboxdrv setup’ but it complained about not finding the kernel sources even though both the kernel header and the kernel devel packages are installed. Further investigations revealed that the version numbers for both the kernel-devel nd kernel-header packages are different from the kernel iteself. Apparently the last kernel update had gone haywire. No matter it can be easily rectified with a yum update.

After updating , you need to do a reboot and I tried runnig VirtualBox again. Now I recieved a less cryptic error message:

WARNING: The vboxdrv kernel module is not loaded. Either there is no module
available for the current kernel (2.6.27.5-37.fc9.x86_64) or it failed to
load. Please recompile the kernel module and install it by

sudo /etc/init.d/vboxdrv setup

One wonders why this more detailed message wasn’t produced the first time around. Recompiling the kernel made no difference anyway. The _ZN10QStatusBar9showEventEP10QShowEvent error still pops up.

There are few posts in the web on this topic. Some of the suggestions are just laughable, they either blame Fedora (even though Debian and Ubuntu users had also complained) or the kernel version (2.6.25). Well mine was 2.6.26 before the kernel update now it’s 2.6.27. Besides I have Virtual Box running on my desktop which has the exact same version of Fedora/Linux Kernel and Glibc as the new installation.

What I did find was a post about a similar error which blamed Qt for the problem. Looking closely at the error message, it does look like a Qt function. So lets update Qt (yes the version of Qt in rgw two computers happen to be different too)

Updating Qt didn’t work. So I thought to remove the 64 bit VirtualBox rpm and to replace it with the 32bit version. After that the previous error went away, it’s place is now taken up by a new one.

VirtualBox: supR3HardenedMainGetTrustedMain: dlopen(”/usr/lib/virtualbox/VirtualBox.so”,) failed: /usr/lib/virtualbox/VirtualBox.so: undefined symbol: _ZN7QRegionpLERK5QRect

Looking closely at this, I am more than convinced that I am barking up the right tree. QRect is surely a Qt function. I even thought to uninstall the 64 bit version and use the 32 bit version of VirtualBox in it’s place because I felt there may be an incompatibility due to the fact that this processor does not have the VMX flag.

I did manage to solve this by first doing ‘yum clean all’ followed by updating both qt and qt-devel.

]]> http://www.raditha.com/blog/archives/002146.html/feed http://www.raditha.com/blog/archives/002144.html http://www.raditha.com/blog/archives/002144.html#comments Wed, 26 Nov 2008 15:51:35 +0000 admin http://www.raditha.com/blog/?p=2144 Things go haywire with VirtualBox when the host system wakes up from sleep.

Message from syslogd@localhost at Nov 17 15:10:33 …
kernel: Code: c0 0f 30 58 8e d8 58 8e c0 5b 41 5c 41 5d 41 5e 41 5f b8 5b f0 ff ff e9 e9 fe ff ff cc cc cc cc b8 ff ff ff ff 48 21 c7 48 31 c0 <0f> 79 fe 73 06 b8 5f f0 ff ff c3 75 05 b8 60 f0 ff ff c3 cc cc

There is a detailed dump in the syslog

Nov 19 12:22:32 localhost kernel: Code: 57 f3 0f c7 34 24 73 07 b8 5e f0 ff ff eb 07 75 05 b8 5d f0 ff ff 48 83 c4 08 c3 cc cc cc cc 0f 01 c4 c3 cc cc cc cc 48 31 c0 57 <66> 0f c7 34 24 73 05 b8 5f f0 ff ff 48 83 c4 08 c3 cc cc cc cc
Nov 19 12:22:32 localhost kernel: RIP  [_end+518288498/2113624756] :vboxdrv:g_abExecMemory+0x6e9e/0x180000
Nov 19 12:22:32 localhost kernel:  RSP <ffff81000c495cf8>
Nov 19 12:22:32 localhost kernel: ---[ end trace acf3d78972eadbc4 ]---
Nov 19 12:23:58 localhost kernel: invalid opcode: 0000 [3] SMP
Nov 19 12:23:58 localhost kernel: CPU 0
Nov 19 12:23:58 localhost kernel: Modules linked in: nf_nat_ftp iptable_nat nf_nat nf_conntrack_ftp ppp_synctty ppp_async crc_ccitt ppp_generic slhc vboxdrv ib_iser rdma_cm ib_cm iw_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp libiscsi scsi_transport_iscsi coretemp w83627ehf hwmon_vid hwmon fuse sunrpc ipt_REJECT xt_tcpudp nf_conntrack_ipv4 xt_state nf_conntrack iptable_filter ip_tables ip6_tables x_tables xfs reiserfs dm_mirror dm_log dm_multipath dm_mod snd_hda_intel snd_seq_dummy i915 snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc snd_hwdep snd iTCO_wdt iTCO_vendor_support atl2 8139too drm sr_mod cdrom ppdev pcspkr i2c_algo_bit parport_pc parport skge mii soundcore i2c_i801 i2c_core sg floppy ata_generic ata_piix pata_acpi libata sd_mod scsi_mod ext3 jbd mbcache uhci_hcd ohci_hcd ehci_hcd [last unloaded: kvm]
Nov 19 12:23:58 localhost kernel: Pid: 5456, comm: VirtualBox Tainted: G      D   2.6.26.5-45.fc9.x86_64 #1
Nov 19 12:23:58 localhost kernel: RIP: 0010:[_end+518288498/2113624756]  [_end+518288498/2113624756] :vboxdrv:g_abExecMemory+0x6e9e/0x180000
Nov 19 12:23:58 localhost kernel: RSP: 0018:ffff81001d055cf8  EFLAGS: 00010046
Nov 19 12:23:58 localhost kernel: RAX: 0000000000000000 RBX: 00000000fffffffe RCX: ffffc2000153c000
Nov 19 12:23:58 localhost kernel: RDX: 0000000000000007 RSI: 000000000000000d RDI: 0000000010a26000
Nov 19 12:23:58 localhost kernel: RBP: ffff81001d055d28 R08: ffff810026db7810 R09: 0000000000000000
Nov 19 12:23:58 localhost kernel: R10: 0000000000000000 R11: ffffffffa0493e30 R12: ffffc2000153c000
Nov 19 12:23:58 localhost kernel: R13: ffff810026dd2b90 R14: 00000000c0305689 R15: 0000000040ed0e50
Nov 19 12:23:58 localhost kernel: FS:  0000000040ed1950(0063) GS:ffffffff81417000(0000) knlGS:0000000000000000
Nov 19 12:23:58 localhost kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Nov 19 12:23:58 localhost kernel: CR2: 00007fb7dc586560 CR3: 000000006716e000 CR4: 00000000000026e0
Nov 19 12:23:58 localhost kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Nov 19 12:23:58 localhost kernel: DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Nov 19 12:23:58 localhost kernel: Process VirtualBox (pid: 5456, threadinfo ffff81001d054000, task ffff810027d1da80)
Nov 19 12:23:58 localhost kernel: Stack:  0000000010a26000 ffffffffa0493e6e ffff81001d055d38 ffffffffa0497ac2
Nov 19 12:23:58 localhost kernel:  0000000000000287 ffffc2000153c000 ffff81001d055d58 ffffffffa0497aeb
Nov 19 12:23:58 localhost kernel:  ffff81001d055de8 00000000fffffff9 ffff810026dd2b90 ffff810026dd2b90
Nov 19 12:23:58 localhost kernel: Call Trace:
Nov 19 12:23:58 localhost kernel:  [_end+518297890/2113624756] ? :vboxdrv:g_abExecMemory+0x934e/0x180000
Nov 19 12:23:58 localhost kernel:  [_end+518313334/2113624756] ? :vboxdrv:g_abExecMemory+0xcfa2/0x180000
Nov 19 12:23:58 localhost kernel:  [_end+518313375/2113624756] :vboxdrv:g_abExecMemory+0xcfcb/0x180000
Nov 19 12:23:58 localhost kernel:  [_end+518314060/2113624756] :vboxdrv:g_abExecMemory+0xd278/0x180000
Nov 19 12:23:58 localhost kernel:  [_end+518240346/2113624756] :vboxdrv:supdrvIOCtl+0xe73/0x12d6
Nov 19 12:23:58 localhost kernel:  [_end+518248073/2113624756] ? :vboxdrv:rtMemAlloc+0xa5/0xdd
Nov 19 12:23:58 localhost kernel:  [_end+518225179/2113624756] :vboxdrv:VBoxDrvLinuxIOCtl+0x11d/0x19c
Nov 19 12:23:58 localhost kernel:  [hrtick_set+139/252] ? hrtick_set+0x8b/0xfc
Nov 19 12:23:58 localhost kernel:  [vfs_ioctl+42/120] vfs_ioctl+0x2a/0x78
Nov 19 12:23:58 localhost kernel:  [do_vfs_ioctl+583/609] do_vfs_ioctl+0x247/0x261
Nov 19 12:23:58 localhost kernel:  [sys_ioctl+85/119] sys_ioctl+0x55/0x77
Nov 19 12:23:58 localhost kernel:  [sys_write+96/111] ? sys_write+0x60/0x6f
Nov 19 12:23:58 localhost kernel:  [system_call_after_swapgs+138/143] system_call_after_swapgs+0x8a/0x8f
Nov 19 12:23:58 localhost kernel:
Nov 19 12:23:58 localhost kernel:
Nov 19 12:23:58 localhost kernel: Code: 57 f3 0f c7 34 24 73 07 b8 5e f0 ff ff eb 07 75 05 b8 5d f0 ff ff 48 83 c4 08 c3 cc cc cc cc 0f 01 c4 c3 cc cc cc cc 48 31 c0 57 <66> 0f c7 34 24 73 05 b8 5f f0 ff ff 48 83 c4 08 c3 cc cc cc cc
Nov 19 12:23:58 localhost kernel: RIP  [_end+518288498/2113624756] :vboxdrv:g_abExecMemory+0x6e9e/0x180000
Nov 19 12:23:58 localhost kernel:  RSP <ffff81001d055cf8>
Nov 19 12:23:58 localhost kernel: ---[ end trace acf3d78972eadbc4 ]---

What causes VirtualBox to go mad is usmb - the userspace samba file system. My guest images reside on another computer (my file server). I mount the folder using usmb and then start them up with VirtualBox this way I get a huge speed advantage because the hard drive on my desktop isn’t being spun into the ground. I can afford to do this because I have gigabit ethernet.

When the machine goes to sleep usmb sometimes disconnects. It is acceptable for VirtualBox to gag when that happens but what is not acceptable is for VirtualBox to force me to reboot to fix the error. Of course I can continue to work on the computer doing other tasks but I cannot run another virtual machine till I reboot.
Restarting the VirtualBox services does not work.

raditha@localhost:~$ /etc/rc.d/init.d/vboxdrv restart
Stopping VirtualBox kernel module                          [FAILED]
(Cannot unload module vboxdrv)

The kernel absolutely refuses to unload the vboxdrv module. I tried rmmod, i tried modprobe -r, I tried fuser -v /dev/vbox to see which process might be locking it (fuser didn’t say). modprobe -r ran through the night without removing the kernel.

I looked through the process table with a fine tooth comb to find any process that might be even remotely associated with it no luck. The VirtualBox kernel module is hanging on grimly like a barnacle on a ships’ keel. This behaviour is the same on F9 and F10

To be fair though, The speed really isn’t that slow. It feels a lot faster for things like reading emails etc. I reckon they are using QoS to slow down larger uploads and downloads because they now have only very limited bandwidth what with their sub marine link being out of service. The question is how long does this last?

[root@localhost Packages]# ping google.com
PING google.com (209.85.171.99) 56(84) bytes of data.
From 10.12.2.114 icmp_seq=10 Destination Host Unreachable
From 10.12.2.114 icmp_seq=14 Destination Host Unreachable
^C
— google.com ping statistics —
20 packets transmitted, 0 received, +2 errors, 100% packet loss, time 24068ms

[root@localhost Packages]# traceroute google.com
traceroute to google.com (72.14.207.99), 30 hops max, 60 byte packets
1  10.12.0.17 (10.12.0.17)  36.826 ms  41.930 ms  46.927 ms
2  10.12.2.114 (10.12.2.114)  51.688 ms  57.394 ms  57.390 ms
3  * * *
4  * * *
5  * * *
6  * * *
7  * * *
8  * * *
9  * * *

Please note that the names are being resolved because I have a caching name server setup locally. Less said about Lanka Bell DNS the better.

Later we are informed that their submarine link has been damaged. Surely they have a backup? Surely they have satellite links? Well right now I think I am being routed via satellites. Yes service was restored a few hours ago (if you could call it that). It took about two minutes for my wordpress editor page to load.

PING 67.131.250.72 (67.131.250.72) 56(84) bytes of data.
64 bytes from 67.131.250.72: icmp_seq=18 ttl=46 time=1210 ms
64 bytes from 67.131.250.72: icmp_seq=19 ttl=46 time=1208 ms
64 bytes from 67.131.250.72: icmp_seq=20 ttl=46 time=1202 ms
64 bytes from 67.131.250.72: icmp_seq=22 ttl=46 time=1230 ms
^C
— 67.131.250.72 ping statistics —
24 packets transmitted, 4 received, 83% packet loss, time 23138ms

I settled on usmb after looking at a lot of alternatives including fusesmb. Unlike fusesmb , usmb does not have an RPM for 64bit systems and I had to compile it by hand (all you need to do is type make and make install). You do need to have the headers for samba and fuse present in your system. You can install them as follows:

yum install libsmbclient-devel

yum install fuse-devel

Then you need to create a ~/.usmb.conf file and then change it’s permissions (600) so that only the owner can read the file.

After doing all that, I am still having no luck. usmb does not produce any output when you execute the command (which is supposed to mean it was successful , yet the remote folder isn’t mounted). There is nothing in the log file.

It was then that I just thought I would do ls /mnt/radmedia/ (/mnt/radmedia/ being the mount point I had chosen). Lo and behold, there is a directory listing with the content of my home folder on my media server. Yet df does not show anything being mounted at that mount point! oh well, I can live with that.

At this point I went back to fusesmb. As with usmb, no error message appears and df actually shows the remote folder as being mounted (the disk usage percentage is also correct). Yet you cannot list the contents of the directory or try to read any of the files on it.

If that great mathematician Euler was alive today, he would just say use both.

This time I was working with a Qemu image that was on a remote machine. An image that was accessed through a samba share. I was playing with some firewall rules while the virtual machine was running and it appeared to get stuck. When I tried to shutdown the VM and relaunch it Qemu told me that it ‘could not open disk image’  and ls said:

ls: cannot open directory .: Transport endpoint is not connected

Then I thought to mount the samba share again with usmb. That’s when the error popped up.

fuse: bad mount point `/mnt/radmedia': Transport endpoint is not connected
*** glibc detected *** usmb: double free or corruption (!prev): 0x00000000008ea1e0 ***
======= Backtrace: =========
/lib64/libc.so.6[0x3737a78228]
/lib64/libc.so.6(cfree+0x76)[0x3737a7a866]
usmb[0x403925]
usmb[0x402904]
usmb[0x402b78]
/lib64/libc.so.6(__libc_start_main+0xfa)[0x3737a1e32a]
usmb[0x401d99]

Immidiately I realized this might be caused by an internal conflict and thought of unmounting the samba share (even though the system told me it’s no longer mounted). fusermount -u /mnt/radmedia was all it took to solve this.  (/mnt/radmedia being my mount point). Now the smb share has been remounted and the virtual machine is running once again.

Twice the machine crashed when it went to sleep. If the computer sleeps, it doesn’t wake up again. The screen gets garbled and there is no response to mouse and keyboard entry. Interestingly it can still be pinged. I have temporarily disabled power management to work around this. Power management didn’t work perfectly for me in Fedora 9 either but this is just crazy. That means I can’t keep the computer running all the time as I usually do but it’s not running all the time anyway - crashing all the time is a more apt description.

The other couple of times the machine crashed it was for no apparent reason at all. On one occasion I was just browsing. There was nothing running that could even remotely be thought of as being demanding of system resources. Yet the machine froze up it would respond to the mouse though. At least you could move the mouse pointer around on but clicking had no effect. During this crash I could actually SSH into this computer from my Macbook. I reckoned this is just X that’s crashing and tried to enter runlevel 3. Though the init 3 command was accepted and no error produced the screen didn’t even change.

I tried killing off all processes that looked even remotely like being associated with xorg or Gnome to no avail. The OS refused to respond to anything. Even Windows, inspite of all it’s problems isn’t really this bad. I am willing to concede that this might be because of faulty installation media (I do have Fedora 10 running without any problems on my File Server). I didn’t test the media before installation (who does?) the installer usually bails out when it finds corrupt RPMs (I have seen it happen). I doubt if it’s faulty hardware either. This machine is pretty new.

I have been using linux for around 10 years now and I was a lot happier 10 or even 5 years ago than I am now. Each new version is full of so many problems to solve. It’s as if Red hat wants to continuously challenge us.  it’s no wonder that linux has struggled to gain market share. How can ordinary users even hope to have a chance if old hands like me struggle with it?

Sure I will be able to find solutions for most things with google but there are problems I will have to handle on my own. Not that I can’t do it but I don’t want to do it. Like everyone else I have more important things to do than fix bugs in other people’s programs.

Why does Red Hat insist on shipping products like PackageKit, Upstart and Pulse Audio? If they just ship that crap on the CD it would be fine but what happens is that good solid stable software are pulled out and they are replaced with the half baked stuff. I don’t think any two versions of Fedora had the same sound system - and sound is one of the many things that no longer works for me since upgrading to F10.

Connection Interrupted
The connection to the server was reset while the page was loading.
The network link was interrupted while negotiating a connection. Please try again.

At first I thought this was an SSL error but it wasn’t anything like that. I found the solution in an old mailing list entry made by one Cris Nolan: http://lists.samba.org/archive/samba/2002-August/049448.html

SWAT still relies on xinetd, been a long time since I ran into a daemon that does so. Once xinetd was editing usmb could be started easily and mostly it works fine. However if it’s left ideling and the computer goes to sleep the connection is dropped. You can easily unmount and mount it again and most things are uneffected but there is an exception. VirtualBox goes crazy when that happens. That’s because my virtual machines actualy reside on the remote folder. If the machine goes to sleep VirtualBox just dies off and even completely refuses to start up again. That leaves me wondering if I should go back to ftpfs.

It was only after I added the ‘vga=0×318′ flag (as mentioned at http://fedoraproject.org/wiki/Releases/10/Alpha/ReleaseNotes#Boot_up
and disabled bootchart that the splash screen appeared.

The new boot process is supposed to be faster but it doesn’t feel like it in fact it feels slightly slower. Without boot chart I will need to revert to a stopwatch to time it. The fact that this installation is the result of a live upgrade rather than a proper upgrade from a DVD might be the cause of it (some of the init scripts etc that the installer might have changed might not have been changed by yum)

What does this mean? has google become a social network. Will spammers now be able to vote legitimate search results out of existence? Will companies with a large number of employees be able to get the number 1 rank simply by asking all their employees to vote for their pages?

Only time will tell.

The Gnome menu bar disappears when ever an applications is started. It’s place is taken up by that applications own menu. The running applications title bar also goes on a trip. You cannot resize or drag it. So the only way to get the menu back is to close what’s running. It’s almost like having a single tasking operating system.

I later found that metacity –replace can get your menu back but only temporarily. It disappears if you logout or reboot. Then I found that updating metacity manually solves it permanently.

Firefox doesn’t start up at all. When trying to execute from the command line, the following error is reported:

could not find compatible GRE between version 1.9b5 and 1.9b5

No surprise that it can’t find anything between 1.9b5 and 1.9b5 because there is nothing in between (assuming that version numbers are real numbers and do not have imaginary components in them of course). This issue did go away with the Firefox upgrade. Yet when FF started I was forced to check for Firefox extension updates and to install them because I just couldn’t say no! (the buttons were hidden due to the above metacity issue)

One other minor matter was the bootup process stopping at runlevel 3 - the init update had modified the inittab file to set the runlevel to 3. To get the the plymouth splash screen I had to disable bootchart.

Jigdo scores over torrents because it can skip any RPMs that are common to the new version and the old. Of course it would be more civic minded to use a torrent. In my case, I will be installing Fedora on two machines of my own and I will be sharing with quite a few others as well. Mostly these are are people forced to suffer with 512Kb DSL links. Even without using torrents, I am doing my bit to spread the love.

The reason that I opted for a live update instead of jigdo is because only 7 of the 2800+ files shipped on the DVD are shared between 9 and 10. That means the whole DVD has to be downloaded and even at 2Mbps that would take a bit too long for my liking. The reason that the two versions have so few files in common might be due to the fact that the digital certificates used to sign the RPMs has changed.

Live updates isn’t easy but it’s more fun. If you first try it out using a virtual machine, there is no risk of a half cooked upgrade or any data loss. But after doing the live update, I decided to download the ISO and do it the proper way after all. The RPMs downloaded for the live update can be used by Jigdo (it did make use of about 200 of them). That still left more than 2500 files to be downloaded and it was painstakingly slow. Though I have had a spot of bother with my broadband connection of late, this time the slow down is probably because the mirrors are under a heavy strain.

The initial download was for the 32 bit version - I need it to use the Sun Java Plugin - they still haven’t got around to creating a 64 bit version for Linux. How lame is that?. The 32 bit fedora will only run as a guest. When that jigsawed download was reassembled it was time to get the 64 bit version. This time around Jigdo found 1200 rpms from the earlier download that it liked.

Our own project is to get the Rad Inks site reorganized. It’s been a few years since the last redesign and it’s about time we did it. I thought a complete outsider could do a better job by providing a new perspective. So the project was announced and with in 48 hours had attracted more than 20 bids. Most of those bids were crazy. It’s almost like people have bots running to place bids for any project that gets listed. Its obvious that many hadn’t even fully read the requirement (which was rather brief  and eminently readable).

All those bids were promptly rejected now we are communicating with those that are left over to try to find the good ones from the not so reliable or skilled providers. The trouble is it’s darned hard to tell the difference.

After downloading about 10 Megabytes of meta info from various repositories, YUM will tell you that there is 34Kb download available! it installs easily and you can mount a remote FTP folder just as easily. By remote I don’t actually mean really remote. I mean a machine with in my LAN. You have to be crazy to use FTP over the internet.

curlftpfs radmedia /mnt/radmedia

Will mount a folder on my media server at /mnt/radmedia in my desktop - but it only mounts the annoymous FTP folder. telling it to use a specific username and a path doesn’t work and the error message is ‘couldn’t resolve hostname’ - which shouldn’t be. With SCP and Rsync on the other hand you can specify the username and path easily. Example raditha@radmedia:/home/raditha/

Then experimenting further, I found that you can in fact specify a username and a password in the command line. Not quite sure if that’s a good idea though. If by some chance someone does hack into one of your machine (hardly likely to happen with linux, but why take a chance?) your second computer will also be compromised because your password is saved in the .bash_history file.

This is different from the risk of using FTP. Your username and password will be compromised only if someone manages to sniff during the time that you are actuallyl using FTP. If the client is single threaded the window is smaller because the username and password will (usually be sent only once). In contrast if the information can be found in the .bash_history file, it can be read from there at any time.

I learned that there are three popular products, Ocrad, Tesseract ang GOCR. All three can be installed with yum. The first one I tried was GOCR. It produced the following output:

Iease instaII the buiId and header fiIes for your current Linux XerneI
he current XerneI version is Z.6.Z5-14.fc9.i686
robIems were found which wouId prevent the Guest Additions from instaIIing.
Iease correct these probIems and try again.

Not very accurate is it? (please see the post on Qemu and VirtualBox to for the actual error message). But I am willing to forgive gocr because that white text on a black background is hard enough for even for humans to read and understand. That console font isn’t the easiest font to read either.  The next suspect Ocrad simply refused to have anything at all to do with the image.

ocrad: bad magic number - not a pbm, pgm or ppm file.

It didn’t fare any better when I saved the file as a pbm (earlier it was a png). It produced exactly nothing. yes nothing, the same thing that you are left with after taxes. So gocr inspite of it’s innacuracies is actually doing a pretty decent job.

Tesseract didn’t like the png image and it liked the pbm even less. It wanted a tif file as input. After a lot of huffing and puffing it produced a single byte file as it’s output.

When trying something like this,  you should make sure that the keepcache setting in yum.conf is switched on. Otherwise if something goes wrong you will find yourself having to download the hundreds of megabytes of files all over again.

In my case just typing ‘yum upgrade’ resulted in a long list of dependency matches and a total suggested download of 1.1 Gb. That’s a lot smaller than downloading the whole DVD but I thought I would upgrade just the essentials. In this case the essentials was just xorg.conf (because I want to be able to use the seamless mode in VirtualBox).

Once the deps were resolved for upgrading xorg, 232 packages need to be downloaded and their total size is 374MB. I wasn’t excepting to smooth sailing and it wasn’t long before I ran into rough weather.

Transaction Check Error:
file /usr/lib/xorg/modules/drivers/r128_drv.so from install of xorg-x11-drv-r128-6.8.0-1.fc10.i386 conflicts with file from package xorg-x11-drv-ati-6.8.0-12.fc9.i386
file /usr/share/man/man4/r128.4.gz from install of xorg-x11-drv-r128-6.8.0-1.fc10.i386 conflicts with file from package xorg-x11-drv-ati-6.8.0-12.fc9.i386
file /etc/pki/tls/certs/ca-bundle.crt from install of ca-certificates-2008-7.noarch conflicts with file from package openssl-0.9.8g-6.fc9.i686
file /usr/lib/xorg/modules/drivers/mach64_drv.so from install of xorg-x11-drv-mach64-6.8.0-1.fc10.i386 conflicts with file from package xorg-x11-drv-ati-6.8.0-12.fc9.i386

Usually when faced with problems like this, my approach is to simply remove the offending rpms (keeping my fingers crossed of course). I was reluctant to remove the openssl RPM is more than likely to break yum. So I just removed the others and asked yum to have another go. Obviously this is where using the keepcache option comes in handy. Real handy.

When I ran yum again it wanted to download another 26MB. In for a penny , in for a pound. Even then the update still failed because of the conflict with openssl. I reluctantly decided to update openssl before continuing. There are countless programs and libraries linked to openssl the chances are very high  that ‘yum update openssl’ will lead to  vicious cycle of dependency resolution with the end result of a total download that is as big as what you would see with just plain ‘yum upgrade’

Fortunately that didn’t happen. Only three packages were updated as a result of the ‘yum update openssl’ command. Then in once again typed ‘yum groupupdate “X Window System”‘ and this time the transactions started without any time being spent on downloads.

During the transaction another error appeared on the screen (but yum didn’t bail out because of it)

/sbin/new-kernel-pkg: line 287: /sbin/depmod: Permission denied
/sbin/mkinitrd: line 1677: /sbin/depmod: Permission denied
“/sbin/depmod -a 2.6.27.5-109.fc10.i686″ failed.
mkinitrd failed

So I thought to run /sbin/depmod -a 2.6.27.5-109.fc10.i686 manually and to keep my fingers crossed and hope there wouldn’t be any boot up problems. Further on , yum did run into what appears to be a more serious error.

But that occurred in the cleanup stage. There were 443 items in the transaction (including installs, updates and removes). The above error occurred after the 443rd item. So now I am not sure if the system has been succesfully updated (the error can be ignore) or the update has failed. Well I found out that the update had failed when I tried to reboot - the new kernel didn’t show up in the grub boot menu.

]]> http://www.raditha.com/blog/archives/002121.html/feed http://www.raditha.com/blog/archives/002112.html http://www.raditha.com/blog/archives/002112.html#comments Mon, 17 Nov 2008 10:36:18 +0000 admin http://www.raditha.com/blog/?p=2112 While trying out VirtualBox I wanted to use the existing Qemu images to save me the trouble of having to carry out new guest OS installations. Sadly, qmeu images will not work out of the box with VirtualBox. Though it’s not possible to directly convert them to VirtualBox (vdi) images with Qemu, can convert them to VMDK.

VMDK is the format used by VMware. It is compatible with VirtualBox as well as Qemu. So you have the option of using either VirtualBox or Qemu to start up a virtual machine. Starting up VirtualBox however, wasn’t smooth sailing all the way. First, i was forced to remove the kvm_intel and kvm kernel modules because of this error:

VirtualBox can’t operate in VMX root mode. Please disable the KVM kernel extension, recompile your kernel and reboot.
VBox status code: -4011 (VERR_VMX_IN_VMX_ROOT_MODE).

Even after the modules were removed (rmmod), the boot up process didn’t run to completion. My default runlevel is 5 - but X wasn’t started. It was because the VGA card emulated by Qemu is very different from the one emulated VirtualBox. I reckoned I will have to install the ‘VirtualBox guest additions’ to have the proper vga driver. That also choked with an error.

Please install the build and header files for your current Linux Kernel
The current Kernel version is 2.6.Z5-14.fc9.i686
Problems were found which would prevent the Guest Additions from installing.
Please correct these problems and try again.

After installing the kernel-devel package I managed to install the add-ins but my xserver still refused to start. So I just deleted the xorg.conf file and hoped the system would get by  with the default - it did. And finally the graphical login screen appeared - with an atrociously high resolution (fixed through System->Preferences).

I have yet to suceeded in enabling the ’seamless mode’. Apparently the xorg version shipped with Fedora 9 isn’t fully supported. Now to try an upgrade to Fedora 10.

One aspect in which VirtualBox scores over Qemu is speed it appears to be slightly faster. The boot time is 39 seconds whereas Qemu takes 44 seconds.