If you delete a lot of files and then unmount the filesystem it can take
quite a while to complete the unmount. The last part of the removal of an
inode happens asynchronously from the user thread, the way these were
being processed during unmount meant it could actually get stuck waiting
several seconds for a buffer to become free. This changes the logic to
avoid this.
Date: Fri Jun 8 13:55:16 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:96591a
linux/fs/xfs/xfs_vnodeops.c - 1.504
- Make the mode used on flushing of freed inode controllable - on
unmount
we do not want to be delayed write as we can end up waiting for a
buffer
for several seconds.
linux/fs/xfs/xfs_vfsops.c - 1.318
- change xfs_finish_reclaim call from sync
linux/fs/xfs/xfs_inode.c - 1.320
- Make the mode used on flushing of freed inode controllable - on
unmount
we do not want to be delayed write as we can end up waiting for a
buffer
for several seconds. From xfs_iflush_all use the new parameter to
indicate
we should do async writes not delayed ones.
linux/fs/xfs/xfs_inode.h - 1.150
- Change prototype for xfs_finish_reclaim
|