This bumps read speed for large I/Os (well 16K plus) by using
generic_file_read all the time - pagebuf based read is gone
except in the direct case. Also includes some fixes for direct
write.
Date: Fri Mar 9 07:55:37 PST 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:89419a
linux/mm/filemap.c - 1.71
- remove pagebuf code from do_generic_file_read
linux/include/linux/fs.h - 1.85
- get pagebuf out of the inode ops business.
linux/fs/xfs/linux/xfs_lrw.c - 1.81
- Buffered read calls generic_file_read, fixes for write path and data
flushing.
linux/fs/xfs/linux/xfs_super.c - 1.111
- remove pagebuf_ioinitiate it is a) not used and b) not defined any
more
linux/fs/xfs/linux/xfs_iops.c - 1.96
- Add linvfs methods to pass linvfs bmap function into pagebuf.
linux/fs/xfs/linux/xfs_iops.h - 1.8
- external for linvfs_pb_bmap
linux/include/linux/page_buf.h - 1.80
- changed prototypes
linux/fs/pagebuf/page_buf.c - 1.63
- Remove references to the pagebuf_bmap inode method, it is gone.
linux/fs/pagebuf/page_buf_io.c - 1.66
- Obtain bmap function via function call parameters, do not get it from
the inode ops. The pagebuf read path is now only used by direct I/O.
|