panic: pmap_enter: invalid page directory pdir=0xc1e063, va=0xffc00000

[Frode Nordahl]

On 4. des. 2005, at 23.44, Frode Nordahl wrote:

> Hello,
>
> After almost 6 months of problem-free operation one of my NFS  
> servers has suddenly started to panic and do a automatic reboot  
> regularly, about every 24 hours.
>
> This is a dual Xeon (with UP kernel for the time being) SE7501HG2  
> with 4 GB RAM and a Intel SRCU42X RAID controller (AMI MegaRAID).
>
> I ran 5.4-RELEASE-p2 on it, but had no kernel.debug for that, so I  
> installed 5.4-RELEASE-p8 on it to get some information from the  
> crashes. (See below). I have also installed a UP kernel to rule out  
> any SMP related problems.
>
> I found no PRs referencing this exact problem, and no mailinglist  
> postings addressing this exact problem (well, found one thread  
> where someone provoked this to happen by tweaking the wrong numbers).
>
>
> My gut feeling tells me that this is a tuning problem, but I really  
> don't know where to start, so any pointers would be great!
>
> By habit I have this in my loader.conf: vm.kmem_size_max=419430400

[snip]

It turns out this was caused by a erratic backup script.

It runs full and incremental backups on two FireWire disks,  
automatically mounting the second one when the first is full.

For some reason the second got full as well and the script ended up  
remounting the device over 340 times (!) trying to squeze some more  
bits out of the unwilling device :-)

I will set up a test case on a crashbox, rule out that 6.0 has this  
problem or file a complete pr with test scripts etc.


Frode Nordahl
frode at nordahl.net