Re: mkstemp fails on cifs linux-2.6.31-rc1 / samba-3.3.6

On Wed, Jul 1, 2009 at 8:44 AM, Wilhelm Meier<[email protected]> wrote:
> Am Mittwoch, 1. Juli 2009 15:17:20 schrieb Shirish Pargaonkar:
>> On Wed, Jul 1, 2009 at 7:13 AM, Jeff Layton<[email protected]> wrote:
>> > On Wed, 2009-07-01 at 06:59 -0500, Shirish Pargaonkar wrote:
>> >> On Wed, Jul 1, 2009 at 6:21 AM, Jeff Layton<[email protected]> wrote:
>> >> > On Wed, 2009-07-01 at 06:01 -0500, Shirish Pargaonkar wrote:
>> >> >> I think the problem is, why EEXIST since file does not exist.  The
>> >> >> looping is I guess
>> >> >> because mktemp then attempts to create another file and receives
>> >> >> EEXIST and it goes on forever.
>> >> >> So the basic issues is why EEXIST since mktemp is attempting a file
>> >> >> which does not
>> >> >> exist on the server.
>> >> >
>> >> > The problem is that the lookup is returning a positive dentry since it
>> >> > just created the file. This makes the VFS turn around and return
>> >> > -EEXIST (see the second O_EXCL check in do_filp_open). So we do have
>> >> > to skip the create on lookup for O_EXCL or it won't work.
>> >> >
>> >> > I think what's needed is something like this patch. It seems to fix
>> >> > the testcase for me. That said, this is not adequately regression
>> >> > tested and should not be committed until it is.
>> >> >
>> >> > On that note, I'd like to reiterate my earlier sentiment that all of
>> >> > this create-on-lookup code was committed prematurely and should be
>> >> > backed out until it's more fully baked.
>> >> >
>> >> > Just my USD$.02...
>> >> >
>> >> > --
>> >> > Jeff Layton <[email protected]>
>> >>
>> >> meant to send this to everybody but ended up sending only to Jeff,
>> >> so here it is again...
>> >>
>> >> I think that is why it is not possible to exclusive create within
>> >> lookup intent code.
>> >> So, IMHO, we should put the check back in the cifs_lookup.
>> >> That is what NFS does, and cifs should do the same
>> >
>> > Right, but do we really need to do the lookup in that case? It seems
>> > like the old check didn't optimize it away (but maybe I'm remembering
>> > incorrectly).
>> >
>> > --
>> > Jeff Layton <[email protected]>
>>
>> With the check, we are back to what was before lookup intent for
>> exclusive create.  I think for exclusive create, a lookup can result in
>> not create if the file exists, so for a command like mktemp, it
>> probably does not help, there will be a lookup and there will be a create.
>
> Well, I don't know nothing about the internals of cifs, but in this case
> lookup and create must be atomic together ...
>
>> _______________________________________________
>> linux-cifs-client mailing list
>> [email protected]
>> https://lists.samba.org/mailman/listinfo/linux-cifs-client
>
> --
> Wilhelm
>

Wilhelm,

I do not know how to maintain atomicity in an operation like exclusive creates
for network file systems, I mean, between a lookup and create from
client, there is a chance
that file gets created on the server.
With cifs_posix_open in lookup, we can gurantee that atomicity but
with the current
vfs implementation, that is not possible (calling cifs_posix_open
within cifs_lookup for
atomic operations like exclusive create) on a Samba server but not on
a Windows server.