BASE-NTDENY1

[Jeremy Allison]

On Sat, Mar 25, 2006 at 12:48:03PM -0800, Jeremy Allison wrote:
> It appears to be a timing related issue.
> I ran the samba4 smbtorture base-ntdeny1
> against W2K3 SP2 and it failed (in a vmware
> session) with exactly the same locking
> mix fail as Samba3 does. I ran it again
> using the same seed and it succeeded !?!
> 
> Now the vmware "cache" is hot so to speak
> (the first run was with a recently restored
> session) it doesn't fail again.
> 
> Seems to be a timing related problem with
> the test.... hmmmm.

Yep - reproduced it against W2K3 again.

Here's the error against W2K3:

-------------------------------------------------------------------------------
bin/smbtorture //192.168.199.129/temp  -Uadministrator%X BASE-NTDENY1
Using seed 1143319791
Running BASE-NTDENY1
Starting 4 clients
4 clients started
starting ntdenytest1 client 0
format: server correct
starting ntdenytest1 client 1
format: server correct
starting ntdenytest1 client 2
format: server correct
starting ntdenytest1 client 3
format: server correct
testing 10 entries on \ntdeny_1.dll
testing 10 entries on \ntdeny_2.dll
testing 10 entries on \ntdeny_3.dll
testing 10 entries on \ntdeny_0.dll
    2/   10
S_R | S_W | S_D      W_DATA | R_ATTR | W_ATTR | EXEC
S_R | S_W | S_D      R_ATTR | W_ATTR | W_EAS  | EXEC                                           -    R  NT_STATUS_OK/NT_STATUS_OKfinshed ntdenytest (1 failures)
finshed ntdenytest (0 failures)
finshed ntdenytest (0 failures)
finshed ntdenytest (0 failures)

BASE-NTDENY1 took 10.0733 secs
-------------------------------------------------------------------------------

And here's the error against Samba3 (again it's not reproducible,
sometimes we pass too).

-------------------------------------------------------------------------------
bin/smbtorture //127.0.0.1/tmp -Ujeremy%X BASE-NTDENY1
Using seed 1143319955
Running BASE-NTDENY1
Starting 4 clients
4 clients started
starting ntdenytest1 client 0
format: server correct
starting ntdenytest1 client 2
starting ntdenytest1 client 1
format: server correct
starting ntdenytest1 client 3
format: server correct
format: server correct
testing 10 entries on \ntdeny_2.dll
testing 10 entries on \ntdeny_3.dll
testing 10 entries on \ntdeny_1.dll
testing 10 entries on \ntdeny_0.dll
    8/   10
S_R | S_D            W_ATTR | W_EAS  | A_DATA
S_W                  W_ATTR | EXEC                                                             -    R  NT_STATUS_OK/NT_STATUS_OK
finshed ntdenytest (1 failures)
    8/   10
S_R | S_W | S_D      W_DATA | W_ATTR | R_EAS
S_W | S_D            W_DATA | R_ATTR | W_ATTR | W_EAS  | A_DATA | EXEC                         W   RW  NT_STATUS_OK/NT_STATUS_OK
finshed ntdenytest (1 failures)
finshed ntdenytest (0 failures)
finshed ntdenytest (0 failures)

BASE-NTDENY1 took 7.64749 secs
-------------------------------------------------------------------------------

Note that the difference : "-    R" is identical in both servers
between predicted and observed. I've also seen the "W   RW" logged
against a W2K3 server too.

There's an error in the test I need to understand better.

Jeremy.