TAKE - Fix realtime bug

[Eric Sandeen <sandeen@sgi.com>]

If you created a realtime file & wrote to it with buffered I/O, it 
could wind up on the wrong device due to a check that did not allow
bmaps for block zero.  Block zero on the realtime is perfectly valid,
and only pbm_bn < 0 reflects a special situation such as delalloc.

Also clean up the "weird blockno" check in set_buffer_dirty_uptodate()
so that it won't complain about block zero if it looks like this
might be destined for the realtime device.  And move the whole check 
into #ifdef PAGEBUF_DEBUG for good measure.

Date:  Mon Feb  4 08:31:54 PST 2002
Workarea:  stout.americas.sgi.com:/localhome/eric/2.4.x-xfs/workarea-reallyclean

The following file(s) were checked into:
  bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs


Modid:  2.4.x-xfs:slinx:110920a
linux/fs/xfs/pagebuf/page_buf_io.c - 1.6
	- Allow block zero in __pb_block_prepare_write_async
	  Move block zero check into PAGEBUF_DEBUG, make it a bit smarter.