On Sat, 3 Oct 2009, Miles Nordin wrote:
re> The best I can tell, the comments are changed to indicate
re> fletcher2 is deprecated.
You are saying the ``fix'' was a change in documentation, nothing
else? The default is still fletcher2, and there is no correct
implementation of the Fletcher checksum only the
good-enough-but-broken fletcher4, which is not the default?
It seems that my memory is kind of crappy (like fletcher2).
There were discussions of the fletcher2 issue on the zfs-code list
starting in March and ending in May:
http://mail.opensolaris.org/pipermail/zfs-code/2009-March/thread.html
http://mail.opensolaris.org/pipermail/zfs-code/2009-April/thread.html
http://mail.opensolaris.org/pipermail/zfs-code/2009-May/thread.html
Unless someone has a legal requirement to prove data integrity, it
does not seem like the fletcher2 woes are much for most people to be
worried about. After all, before zfs, this level of validation did
not exist at all. Fletcher2 will still catch most instances of data
corruption.
One thing I did learn from this discussion is that when accessing
uncached memory, the performance of fletcher2 and fletcher4 is roughly
equivalent so there is usually no penalty for enabling fletcher4. It
does seem like there could be some CPU impact for synchronous writes
from fletcher4 since it is more likely that the data is in cache for a
synchronous write.
Bob
--
Bob Friesenhahn
bfrie...@simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
