Re: Linux 2.2.14pre18

[Zdenek Kabelac]

Alan Cox wrote:
> 
> Final spit and polish
> 
> 2.2.14pre18
> o       Dave Jones moved                                (Dave Jones)
> o       Fix minor errors in nautilus Alpha code         (Michal Jaegermann)
> o       Remove a debugging printk in SDLA Frame Relay   (Nenad Corbic)
> o       Final IDE clean up                              (Mark Lord)
> o       ISDN update                                     (Karsten Keil)

Hello 

I'm still having few problems with SMP & IDE linux kernel.
However its hard to locate the problem.

Here is the first one (The last checked 2.3.33 kernel by me had the same
problem):

Boot SMP kernel with nosmp option, run hdparm -d 0 /dev/hda; hdparm -t
/dev/hda.
On my smp computer it starts to write messages:  hda: lost interrupt.
(Magic key still works however I couldn't remount hd partitions to
read-only mode,
but I can boot the machine)

Maybe this one is the key for my second problem.
When running in SMP mode, the kernel sometimes hangs up, so that
even SysRq key doesn't work. However I'm unable to locate this bug.
The problem is I haven't found yet any deterministic way how to achieve
this state and sometimes the machine could have even several days
uptime,
but sometimes it locks three times per day - and I can't run
experimental
debugging kernels for days on this machine.
At first I thought it might be nfs or network related, but
over holidays I've seen this lock even on single no network connected
machine
using just two ide drives & 1 CDROM. Lock has always appeared while
copying 
some files.

Is there a way to compile the kernel with some special kind of spinlock
which will prevent this complete deadlock - something like it will allow
to run one processor and will try to log the message about this state
and then will go for reboot, but otherwice the system will be normaly
running ???

bye

-- 
         binary drivers cause sclerosis of the kernel (mark hahn)
  Zdenek Kabelac  http://i.am/kabi/ [e-mail] {debian.org; fi.muni.cz}