[Bug 10925] non-atomic xattr replacement in btrfs => rsync --read-batch random errors

samba-bugs at samba.org samba-bugs at samba.org
Fri Nov 21 22:13:26 MST 2014


--- Comment #3 from Alexandre Oliva <oliva at gnu.org> ---
Yeah, the btrfs bug is now fixed, so now the rsync --read-batch misbehavior
changed from bad error reports to silent waste of cpu cycles: it doesn't make
sense for rsync to test and set xattrs concurrently on the same files: it
should either test them in the scanner process BEFORE telling the changer
process whether xattr changes need to be applied in the first place, OR it
should check them AFTER the changer has made the changes.  As it stands, the
check is wasteful: it is either ignored altogether, or it may see xattrs before
they are changed and conclude the changes have to be made again.  Either way,
we waste cpu cycles, syscalls, and even disk space, if we modify a
copy-on-write snapshot overwriting xattrs with the values they already have,
unnecessarily touching inodes' ctimes in the process.

You are receiving this mail because:
You are the QA Contact for the bug.

More information about the rsync mailing list