This smooths the integration of the delayed buffer concept in the main
kernel and removes the need for the toss_page address space method.
XFS now uses no extra address_space methods.
There is a still a hole which is going to affect highmem systems,
but I am pretty convinced I could make ext2 hit it as well, write
a large file, truncate it on one cpu and run sync on the other.
Date: Mon Jun 11 14:21:15 PDT 2001
Workarea: jen.americas.sgi.com:/src/lord/xfs-linux.2.4
The following file(s) were checked into:
bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs
Modid: 2.4.x-xfs:slinx:96695a
linux/mm/filemap.c - 1.76
- remove the call to toss_page - we now use the unmap_buffer call in
try_to_free_buffers to clear the delalloc state.
linux/kernel/ksyms.c - 1.94
- export block_flushpage as pagebuf calls it now
linux/include/linux/fs.h - 1.98
- remove the toss_page method
linux/fs/buffer.c - 1.65
- Add the delay flag to the fields unmap_buffer resets, with this we
can get rid of the toss_pages address space method.
linux/drivers/block/ll_rw_blk.c - 1.68
- Cleanup, use buffer_delay rather than reference BH_Delay directly
linux/fs/xfs/linux/xfs_iops.c - 1.108
- remove the toss_page call
linux/include/linux/page_buf.h - 1.94
- remove pagebuf_toss_pages prototype
linux/fs/pagebuf/page_buf_io.c - 1.83
- Use buffer_delay rather than BH_Delay directly,
simplify _unmark_delalloc as there are less cases now,
remove pagebuf_toss_page,
mark delalloc buffers as mapped so that unmap_buffer clears the state,
use the buffer_delay flag instead of the !buffer_mapped test to
look for buffers which need conversion.
|