what's happening here is that softdep_pageiodone() is trying to determine
which page of the file was being written by retrieving the information
from the pmap about what physical page was mapped in the i/o descriptor.
unfortunately the pmap has lost this information, so we panic.
the same thing can happen when uvm_aio_aiodone() calls uvm_pageratop(),
so you can see the same panic without softdep as well.

-Chuck


On Thu, Oct 25, 2001 at 10:19:12AM -0400, Allen Briggs wrote:
> On Wed, Oct 24, 2001 at 05:31:58PM +0200, Xavier HUMBERT wrote:
> > I had a panic while building libc :
> > panic: kernel diagnostic assertion "rv" failed: file
> >     "../../../../uvm/uvm_pager_i.h", line 64
> > 
> > Stopped in pid 7 (aiodoned) act cpu_Debugger+0x10   lwz  r0,r1,0x14
> 
> I just ran into this with a kernel w/o NEWPMAP on a different PowerPC
> system.  It looks like this is coming from softdep_pageiodone() when
> it calls uvm_pageratop().  You will not see this panic if SOFTDEP is
> disabled.  I don't know enough about UVM or SOFTDEP to know what might
> be happening here.
> 
> In my case, the panic was preceded by some adw0 DMA errors, so I wasn't
> sure if I had run into a disk error (I never saw these errors before
> enabling softdep yesterday, though).
> 
> I didn't see this panic with NEWPMAP, but I did see a different panic
> that was related to NEWPMAP.
> 
> -allen
> 
> -- 
>  Allen Briggs                     briggs@wasabisystems.com
>  http://www.wasabisystems.com/    Quality NetBSD CDs, Sales, Support, Service
> NetBSD dev. for _your_ Alpha, ARM, M68K, MIPS, PowerPC, SH3, Sparc, x86, etc...