[jcifs] RE: SOLVED: Unverifiable Signature

Michael Kerley michael at enkoo.com
Wed Apr 7 22:31:33 GMT 2004

The signature stuff works great!


> -----Original Message-----
> From: Michael B Allen [mailto:mba2000 at ioplex.com]
> Sent: Tuesday, April 06, 2004 8:47 PM
> To: jcifs at samba.org
> Cc: Michael Kerley; 'Gary Rambo'
> Subject: SOLVED: Unverifiable Signature
> If someone would like to try my double secret pre-release it has the fix
> for the "Unverifiable signature" problem as well as fixes for directory
> timestamp issues, DFS bugs found by Gary, as well as several other SMB
> signature issues.
> However! This code also has some significant refactoring and cleanups.
> NbtSocket has been hacked, NT_STATUS codes are now used, the way MIDs are
> allocated has changed, etc. These kinds of things can change the behavior
> of the client significantly but I haven't tested this stuff at all.
>   http://home.comcast.net/~miallen1/jcifs/jcifs-0.9.0p3.tgz
>   http://home.comcast.net/~miallen1/jcifs/jcifs-0.9.0p3.zip
>   http://home.comcast.net/~miallen1/jcifs/jcifs-0.9.0p3.jar
> I have a few more things on the list (e.g. raw transport) and I'll start
> 0.9.
> Mike
> PS: The "Unverifiable signature" bug turned out to be a padding byte that
> wasn't being properly considered when updating the signing digest.
> Normally this byte is 0x00 which works. But occationally MS servers have
> garbage in this byte. A 0x01 instead of 0x00 is all that's needed for the
> signature to fail. The time dependancy is probably due to how MS servers
> handle buffers. If you wait a half-second you get a 0x01 in the pad byte
> of a ReadAndX response. If you're quick to send the next request it's
> 0x00.

More information about the jcifs mailing list