This fixes some issues between creating new inodes and tearing
down old ones, freezing and thawing a filesystem seemed to be a
good way of exposing this . It is a definite improvement over the
old code, 36 hours continuous stress testing and still going vs
about 10 minutes for the old code. Not convinced this is a totally
closed issue yet though.
Date: Sun Sep 1 05:43:34 PDT 2002
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:126536a
linux/fs/xfs/xfs_vnodeops.c - 1.560
- fix a hole between sync and reclaim, clean up finish_reclaim code,
add comments
linux/fs/xfs/xfs_vfsops.c - 1.377
- pass flush mode into xfs_finish_reclaim instead of a flag
linux/fs/xfs/xfs_iget.c - 1.169
- deal with the case where we race between looking up and tearing down
a linux inode.
linux/fs/xfs/xfs_inode.c - 1.348
- pass flush mode into xfs_finish_reclaim instead of a flag
linux/fs/xfs/linux/xfs_vnode.c - 1.96
- Fail a vn_get on a vnode who's inode is being freed
|