Hi,

  On a 4GB SMP box, configured with HIGHMEM support, making a 670G
(obviously using a volume manager) ext2 file system takes 12minutes (over
10minutes of sys time).

  One problem is buffer allocations do not use HIGHMEM, but the
nr_free_buffer_pages() doesn't take this into account causing
balance_dirty_state() to return the wrong state.

  The attached patch fixes the worse part of nr_free_buffer_pages() - with
HIGHMEM it now only counts free+inactive pages in the NORMAL and DMA
zone.  It doesn't fix the inactive_dirty calculation.
  Also, in buffer.c:flush_dirty_buffers(), it is worthwhile to kick the
disk queues if it decides to reschedule.

  With the patch, that 670G filesystem can be created in 5m36secs (under
half the time).

  Patch is against 2.4.1-ac18.

Mark