UnSolved - M8.5 B5 bugs

[benny_fletch]

I am also having an issue with showfoto, it crashes when I try to open folders on a NFS share (no matter if there is one image or 1000 in the folder). This occurs with both 1.0.0beta6 (8.5) and 1.1.0 (squeeze). When ran from the command line I get an error with kdeinit4 when it crashes. So I am not sure if it is a KDE issue or a digikam/showfoto issue.

Regardless this works without issue on 8.0.x utilizing showfoto 0.9.6 for the CR (as well as every other previous version included with 8.0).

[Davo]

@Jerry
Got exact same errors messages on boot w/ LiveCD (about 10 screens worth ...all seem IRQ-based)
For me it's IRQ 17 ...for you it was IRQ 10

irq event 17: bogus return value f571c108                                       
Pid: 5, comm: events/0 Tainted: P           2.6.32-1-mepis-smp #1               
Call Trace:                                                                     
 [<c106b081>] ? __report_bad_irq+0x2b/0x69                                      
 [<c106b115>] ? note_interrupt+0x56/0x13d                                       
 [<c106b6d8>] ? handle_fasteoi_irq+0x81/0xa2                                    
 [<c1004df3>] ? handle_irq+0x17/0x1b                                            
 [<c1004675>] ? do_IRQ+0x38/0x89                                                
 [<c1003250>] ? common_interrupt+0x30/0x38                                      
 [<c103388f>] ? __do_softirq+0x47/0x16f                                         
 [<c10339e8>] ? do_softirq+0x31/0x3c                                            
 [<c1033b34>] ? irq_exit+0x26/0x68                                              
 [<c101497c>] ? smp_apic_timer_interrupt+0x6c/0x76                              
 [<c1003595>] ? apic_timer_interrupt+0x31/0x38                                  
 [<c123007b>] ? echo_set_canon_col+0x18/0x32                                    
 [<c10200d8>] ? calc_delta_mine+0x4b/0x8a                                       
 [<c102ef0e>] ? release_console_sem+0x14d/0x1a6                                 
 [<c1046165>] ? __atomic_notifier_call_chain+0x1c/0x39                          
 [<c123c5be>] ? console_callback+0xf0/0xf4                                      
 [<c103f818>] ? worker_thread+0x160/0x1e8                                       
 [<c123c4ce>] ? console_callback+0x0/0xf4                                       
 [<c1042756>] ? autoremove_wake_function+0x0/0x2d                               
 [<c103f6b8>] ? worker_thread+0x0/0x1e8                                         
 [<c10424e3>] ? kthread+0x61/0x66                                               
 [<c1042482>] ? kthread+0x0/0x66                                                
 [<c10037a7>] ? kernel_thread_helper+0x7/0x10                                   
handlers:                                                                       
[<c12b5050>] (usb_hcd_irq+0x0/0x71)                                             
[<f841e36e>] (snd_intel8x0_interrupt+0x0/0x1af [snd_intel8x0m])                 
[<f82eb3af>] (io_irq_isr+0x0/0x2c [ndiswrapper])                                
irq event 17: bogus return value f571c108                                       
Pid: 5, comm: events/0 Tainted: P           2.6.32-1-mepis-smp #1               
Call Trace:

$ dmesg | grep 'irq event 17' | wc -l
100
So 100 times a soft kernel panic on IRQ event 17. No crash, just multiple screens of errors.
But ...it booted up just fine afterwards to X, after this slight delay.
Overalll - still a bit of a washed-out plasma panel (changed via SysSettings > Desktop Theme)

Otherwise: GTK Integration-based. Great to see backported GTK libs integrated.
Fixed up the huge Fonts/Toolbars under FF by toggling Fonts under SysSettings
(plus for root apps by # cp .kde/share/config/kdeglobals /root/.kde/share/config/kdeglobals)
and (text under icons etc there ) plus reverted QtCurve to default (overriding Mepis Configs - IMHO just looks better).
Also installed openoffice.org-{gtk/kde} and applied various tweaks

[Jerry3904]

What hardware, and was it up to date with B4 including kernel beforehand (Warren questions)?

[Davo]

Originally Posted by Jerry3904

What hardware, and was it up to date with B4 including kernel beforehand (Warren questions)?

Is this for me?
Hardware: Dell Latitude D610 (that's about as stock as-you-can-get, tried-and-tested and very Linux-friendly)
- a Business laptop with absolutely nothing out of the ordinary Hardware-wise.
Has run Suse, Fedora, Ubuntu, Gentoo & Slackware - never seeing anything out of the ordinary before there under dmesg.
M8.5 - am running both original Alpha fully-updated plus trying all the Beta's individually (including B4), without this dmesg ouput happening.

[JimC]

Although a newer version digikam is in the mepis repos (1.0.0 beta 6, which is dated now since digikam 1.0.0 has been final for a while, and even a newer digikam1.1.0 just went final this month), the only digikam-doc in the repos is a much older version (0.9.3-1), even though it's a recommended package.

If you try to install it, it wants to remove khelpcenter4 and install khelpcenter along with a bunch of other packages (I stopped at that point before letting it continue).

[benny_fletch]

I installed digikam 1.1.0 from the debian squeeze repo w/o any issues in 8.5, with the squeeze repo enabled I see a digikam-docs 0.10.0~svn105645 (not sure if this is the latest revision or not). Also, documentation can be found here:
http://www.digikam.org/drupal/docs

[JimC]

Thanks. I may give that a try and see if it helps things. The main malfunction (for me, anyway) is that I can't even open files raw from some of the newer cameras I've used. But, that may be the associated libraries versus the version of the digiKam itself. The docs being way out of date (trying to remove packages if you try to install them) was just a problem I noticed.

I see a thread about using Squeeze here:

http://mepislovers.org/forums/showthread.php?t=25310

So, I'll go ahead an upgrade to it and see if that helps (or not).

[benny_fletch]

I definitely didn't want to completely upgrade to squeeze, but I enable the repo from time to time to try newer versions of certain apps, I have been pretty lucky it hasn't hosed my system (test partition regardless). I do see a libkdcraw7 4:4.3.4-1+b1 in the squeeze repo, not sure what it tries to install/remove (I am currently in 8.0).

Best method for me is to enable the squeeze repo by simply changing "lenny" to "squeeze" in Synaptic, then under "Preferences -> Distribution" I select "Always prefer the highest version". Once I have installed said app, I simply undo the above steps.

[JimC]

I added the repos and upgraded to newer versions showfoto and digikam using the Force Package option in the menus. So, I've got the 1.1.0 versions of both now. I'm also running the newest version of the kipi-plugins I see (which is out of date), as well as the newest version of libkdcraw7 I see (which is also out of date).

So, the versions of it in Squeeze don't help either (the libraries are just too out of date to open raw files from a number of cameras released last year).

For jpeg images, it's probably fine. But for raw files (and I always shoot raw + jpeg), the versions in Squeeze are going to be useless to me.

[Davo]

@ Jerry
I sort-of figured out the 100s of Kernel Call Trace errors on boot-up.
This stuff:

irq event 17: bogus return value f571c108                                       
Pid: 5, comm: events/0 Tainted: P           2.6.32-1-mepis-smp #1               
Call Trace:                                                                     
 [<c106b081>] ? __report_bad_irq+0x2b/0x69                                      
 [<c106b115>] ? note_interrupt+0x56/0x13d                                       
 [<c106b6d8>] ? handle_fasteoi_irq+0x81/0xa2                                    
 [<c1004df3>] ? handle_irq+0x17/0x1b 
...

It's also reported under Fedora http://www.mail-archive.com/fedora-l.../msg57687.html
Basically, it happens when ndiswrapper is being loaded, which generates spurious IRQ errors
Confirmed, as when I boot Beta5 with nondis cheatcode (also using nonet ), it completely disappears.
However, upon manually loading ndiswrapper (#modprobe -v ndiswrapper) after boot,
immediately results in the 100s of errors appearing under #cat /var/log/messages.

I traced it to here http://linux.derkeiler.com/Mailing-L.../msg06897.html
Basically kernel/irq/spurious.c has this

if (action_ret != IRQ_HANDLED && action_ret != IRQ_NONE) {
printk(KERN_ERR "irq event %d: bogus return value %x\n",
irq, action_ret);

Kernel/irq/handle.c has this (from Thomas Gleixner, Kernel dev)

switch (ret) {
case IRQ_WAKE_THREAD:
/*
* Set result to handled so the spurious check
* does not trigger.
*/
ret = IRQ_HANDLED;

Here IRQ_WAKE_THREAD is mapped to IRQ_HANDLED so that "_report_bad_irq" shouldn't trigger and give spurious errors.

In other words, the note_interrupt Kernel call should be being called after handle_irq (in most cases) ...avoiding the error messages completely
But in our output (probably because ndiswrapper isn't using threaded IRQ) it's happening the other way round.
the "note_interrupt" occurs here first and thus spawns a series of "_report_bad_irq" on irq_wake_thread calls.
...with irq_wake_thread not explicitly defined under spurious.c
It'll then just walk the pstree and print spurious alert sequences for all of the currently active processes, before eventually settling.
Note: Kernel devs almost never respond to Bug threads (or fix stuff), if the kernel is tainted (and ndiswrapper taints it)

Could you maybe pass this on to Warren?
I'm not sure what the best solution would be - but he might need to disable ndis on the LiveCD to avoid this happening.
I'm not sure how many users even need ndiswrapper, but it's probably a small minority nowadays.
Haven't a clue as to why both my updgraded Alpha (and Beta3) LiveCD installs aren't affected though.
So, I'd suggest disabling ndis by default might be the way to go, unless Warren can somehow figure out why it's only occurring now.

PS: Jerry, you also have snd_intel8x0m showing on IRQ 10. If you don't need dial-up just
# nano /etc/modprobe.d/alsa-base-blacklist.conf and uncomment it there to remove if from sharing the IRQ (on next boot)