Design change for oplock/open code?

[Christian Ambach]

On 04/26/2013 11:46 PM, Jeremy Allison wrote:
> I made 2 changes. One was to add a comment in
> include/vfs.h saying the VFS interface number
> wasn't yet being changed as we're not releasing
> and noting the fields within the files_struct
> that got deleted. The second was to up the debug
> level on an error return from dbwrap_record_watch_recv()
> from 1 to 5 as it can commonly get NT_STATUS_IO_TIMEOUT if
> there really is a sharing violation.

That adjustment made it a bit harder for me to debug the issues
I was experiencing when running bench.nbench and clustering=yes.

I was not able to see the NT_STATUS_ACCESS_DENIED at that spot with
default loglevel, but I finally found the cause and I am proposing the
attached patch to make it work again in clustering setups.
Otherwise, bench.nbench will stall after a short while and smbd will
spin in 100 %CPU until nbench hangs up because it cannot establish the
message channel and so keeps retrying the open forever.

Cheers,
Christian

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-s3-lib-ctdb_conn-make-sure-we-are-root-before-connec.patch
Type: text/x-patch
Size: 1392 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20130516/bb137517/attachment.bin>