[Samba] samba md5sum
Volker.Lendecke at SerNet.DE
Tue Nov 17 16:14:49 MST 2009
CC'ing the list, this might be of larger interest (Hi,
Steven & Jeff :-))
On Tue, Nov 17, 2009 at 11:52:33PM +0100, Günter Kukkukk wrote:
> Am Dienstag, 17. November 2009 schrieb Volker Lendecke:
> > On Mon, Nov 16, 2009 at 08:57:54PM -0500, Mag Gam wrote:
> > > how can I check what settings the client is using? I can always check
> > > the /etc/mtab but if there a way to see what type of locking mechanism
> > > my client and server are using?
> > cat /proc/mounts
> > BTW, on another front I've got reports that you can get data
> > corruption with cifsfs/Samba without using the directio
> > mount option. I am investigating this with very high
> > priority right now, but mounting with directio might be
> > something you want to try.
> > Volker
> Hi Volker,
> can you drop me a url/link where that cifs vfs "data corruption" and
> "directio" related problem is discussed?
It's a bug by a customer of mine. BTW, it just was reported
as https://bugzilla.samba.org/show_bug.cgi?id=6898 also.
It's funny how things come together. I had stared at
logfiles the whole day today, and then this very nice little
bug getting it to a very good point popped up in my inbox.
It's a bit subtle: cifsfs does not really handle O_APPEND
right, I'm not sure the fix I proposed in the bug is the
last word. You could very well argue that this is a client
bug. Samba just does what the client asks for: O_APPEND. At
this moment the offsets in the write&x calls have no effect
at all. But cifs still sends whatever it has in the page
cache: Look at frame 36 in the sniff from bug 6898, there we
get 6 bytes at offset 0, containing the whole thing again.
With O_APPEND smbd has no chance but to duplicate that
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 197 bytes
Desc: Digital signature
More information about the samba-technical