A number of unwritten extent fixes, fix some reference count
and locking issues, also make allocation interface always
go into core xfs via the same route rather than having a
distinct interface for unwritten extents.
I still have some failure cases after this, but wanted
to get this out first. Filesystems without unwritten
extents, or ones which do not use preallocation calls
will be fine.
Date: Fri Apr 18 07:10:22 PDT 2003
Workarea: penguin.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:146255a
linux/fs/xfs/xfsidbg.c - 1.219
- fix up printing of pagebuf flags
linux/fs/xfs/pagebuf/page_buf.c - 1.114
- When flushing metadata for a device, ensure there are no pending
unwritten extent conversions pending for the I/O completion threads.
linux/fs/xfs/pagebuf/page_buf.h - 1.66
- change flag definitions for the allocator interface.
linux/fs/xfs/linux/xfs_aops.c - 1.33
- Move all allocation requst flags to a seperate set of defines
instead of abusing the pagebuf flags. Make unwritten extent
conversion go into the allocator the same way all other
requests do. Fix a reference count issue which caused use
after free and one case where we passed an unlocked buffer
into submit_bh.
linux/fs/xfs/linux/xfs_iomap.c - 1.9
- switch allocator to its own set of control flags, merge the
unwritten extent path into the BMAP call more fully than it
was.
|