Got it, the kmem_map is marked VM_MAP_INTRSAFE in kern_malloc.....
Sorry for the noise.

Lars Heidieker wrote:

> Hi,
>
> there something about the pool allocator I wasn't able to figure out yet.
> The Question is, how is it managed that the
> me = pool_get(&uvm_map_entry_kmem_pool, pflags);
> in uvm_map works if it might end up there again, after calling
> pool_allocator_alloc  ---> pool_page_alloc --> uvm_km_kmemalloc --> 
> uvm_map
> and then a need for a new map entry.
> This would result in recursion, but I guess I am just missing some 
> detail.
>
> Lars
>
>