rename parent directory and v2 lease breaks

[Steve French]

On Fri, Mar 6, 2015 at 4:42 PM, Steve French <smfrench at gmail.com> wrote:

>
>
> On Wed, Mar 4, 2015 at 2:08 AM, Ralph Böhme <rb at sernet.de> wrote:
>
>> On Wed, Mar 04, 2015 at 06:54:26AM +0100, Volker Lendecke wrote:
>> > On Tue, Mar 03, 2015 at 02:41:31PM -0600, Steve French wrote:
>> > > I was testing Samba 4.2 and noticed that renaming the parent directory
>> > > did not trigger a v2 lease break for files opened in that directory
>> > > (it looks like Windows does send one though).
>> > >
>> > > Shouldn't rename of the parent directory trigger a lease break for
>> > > open files in the directory?
>> >
>> > Yes, it should.
>>
>> Hm, I fail to find this in the doc, the only applicable rule I see is:
>>
>> * If Open.File.FileType is DirectoryFile, determine whether Open.File
>>   contains open files ...
>>
>> * If Open.File contains open files, the operation MUST be failed with
>>   STATUS_ACCESS_DENIED
>>
>> in MS-FSA 2.1.5.14.11. The pseudocode algo doesn't mention any lease
>> break triggering. Am I missing anything? Is this missing in the doc?
>>
>>
> The Microsoft test cases are usually pretty good about referencing which
> subsection of the document they apply to, but clearly Windows 8.1
> (as a server, among others) does generate the lease break for the
> children - otherwise reopen of the file to writeout data after intermittent
> failure (or any path based op on the file where data integrity of cached
> data matters) could fail.
>
>
>
I haven't tried strict rename yet (will see if that helps). but did open
open
bug # 11178 to track the problem as it can have data integrity
implications


-- 
Thanks,

Steve