Revision 1.70, Thu Jun 23 15:22:44 2005 UTC (12 years, 3 months ago) by wkendall
Branch: MAIN
Changes since 1.69: +1 -1 lines

When restoring a filesystem that is nearly full, writes start to
fail with ENOSPC even though the filesystem is not actually full.
This is a result of XFS's delayed allocation.

Change xfsrestore to retry a write that fails with ENOSPC after
issuing a fdatasync call. If it still fails, issue a sync call
and retry once more.