Reiterating for the benefit of the folks on linux-xfs:
mkfs.xfs contains code which looks to see if all AGs will start on the
same disk stripe, a non-optimal configuration, and changes the ag size
by +/- 1 data stripe unit to avoid this.
However, in the 1 AG case, this can lead to an ag size > data size,
which ultimately fails in an ASSERT.
This can be demonstrated with:
mkfs.xfs -f -b size=4096 -d file,name=foo,size=10m,sunit=8,swidth=8
This can be avoided by not even going down the path for the 1-ag case, but
a check for (agsize > dblocks) is added as well, to ensure
that this never happens.
Date: Mon Sep 16 13:38:02 PDT 2002
Workarea:
stout.americas.sgi.com:/localhome/src/sandeen/2.4.x-xfs/workarea-alwaysclean
The following file(s) were checked into:
bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs
Modid: xfs-cmds:slinx:127521a
cmd/xfsprogs/VERSION - 1.58
- Bump to 2.3.3
cmd/xfsprogs/doc/CHANGES - 1.83
- Document changes for 2.3.3
cmd/xfsprogs/mkfs/xfs_mkfs.c - 1.36
- Don't do ag resizing based on stripe alignment for single ag fs
Don't ever let agsize > dblocks when doing this resizing.
|