Description of problem:


Version-Release number of selected component (if applicable):
device-mapper-1.01.02-1.0


How reproducible:
100%

Steps to Reproduce:
1. Create a pv, vg and finaly an lv, let's call it "test". then
2. for x in $(seq 1 20); do
      lvcreate -s -L1G -n test$x /dev/vg/test || break
   done

On my local system I get an ENOMEM after about 16 snapshots.
  
Expected results:
Many more snapsots, not limited by the maximum size of a single kmalloc.

Additional info:
We have managed to trace the problem to the use of mempools for the _io_pool and
that adding snspshots requires the pool be grown, which ends up hitting the 
limits that kmalloc() imposes on allocations.

Is this a known limitiation?  Are there any plans to solve/eliminate it?  If
somebody were to take on reworking that area of code/design, do you have any
preferences or designs in how it is done?