On Wed, 20 Mar 2002, Manuel Bouyer wrote:
> On Wed, Mar 20, 2002 at 01:37:21PM +0000, Stephen Borrill wrote:
> > There isn't much information. :-(
> > 
> > panic: pmap_alloc_pvpage
> > stopped in httpd at _debugger +0x4: leave
> 
> OK, so there is a limit which is too low somewhere.
> Now to find which one.
> 
> I suspect it's the kmem_map which is too small pmap_alloc_pvpage() does
> a call to uvm_km_kmemalloc() from kmem_map. Maybe try to increase
> NKMEMCLUSTERS, but it's just a guess. By default it's 6 * 1024 * 1024 / CLBYTES

OK, I put this up to 24 * 1024 * 1024 / CLBYTES (pity you can't put a
formula in your kernel config file which means a lot of digging around to
find a value for CLBYTES).

The system has been in heavy use for a couple of weeks as a web proxy
under conditions which were killing it daily. I've been told that it's not
crashed since, so thanks for your guess Manuel!

> A vm guru can probably say more about this.

But didn't...

Is this still the same in 1.5 and/or -current? The panic message is not at
all enlightening...

-- 
Stephen