On Oct 3, 2009, at 12:22 PM, Miles Nordin wrote:
"re" == Richard Elling <richard.ell...@gmail.com> writes:
re> If I was to refer to Fletcher's algorithm, I would use
re> Fletcher. When I am referring to the ZFS checksum setting of
re> "fletcher2" I will continue to use "fletcher2"
haha okay, so to clarify, when reading a Richard Elling post:
fletcher2 = ZFS's broken attempt to implement a 32-bit Fletcher
checksum
Fletcher = hypothetical correct implementation of a Fletcher checksum
In that case, for clarity I think I'll have to use the word ``broken''
a lot more often.
How does the fix included S10u8 and snv_114 work?
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?
Also, there is no way to use a non-broken checksum on the ZIL?
The ZIL is a slightly different beast. If there is a checksum mismatch
while processing the log, it signals the effective end of the log. This
is why log entries are self-checksummed. In other words, if you reach
garbage, then you've reached the end of the log. The probability of
the garbage having both a valid fletcher2 checksum at the proper
offset and having the proper sequence number and having the right
log chain link and having the right block size is considerably lower
than the weakness of fletcher2.
Unfortunately, the ZIL is also latency sensitive, so the performance
case gets stronger while the additional error checking already
boosts the dependability case.
-- richard
doesn't sound fixed to me. At least there is some transparency,
though, and a partial workaround.
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
