smbtorture test smb2.rename failure
Shilpa K
shilpa.krishnareddy at gmail.com
Fri Feb 14 03:26:34 MST 2014
Hello,
I was running one of the smbtorture tests, "smb2.rename". The following 2
sub tests under it are failing:
test: simple_nodelete)
time: 2014-02-13 23:26:16.812166
Creating base directory
Creating test file
Renaming test file
Cleaning up
time: 2014-02-13 23:26:16.931777
failure: simple_nodelete) [
(../source4/torture/smb2/rename.c:184) Incorrect status NT_STATUS_OK -
should be NT_STATUS_ACCESS_DENIED
]
test: no_share_delete_no_delete_access
time: 2014-02-13 23:26:17.232758
Creating base directory
Opening parent directory
Creating test file
Renaming test file
Cleaning up
time: 2014-02-13 23:26:17.351722
failure: no_share_delete_no_delete_access [
(../source4/torture/smb2/rename.c:761) Incorrect status NT_STATUS_OK -
should be NT_STATUS_SHARING_VIOLATION
]
I think the first failed test "simple_nodelete" can be fixed by following
check in the routine can_rename() in reply.c:
if ((fsp->access_mask & (DELETE_ACCESS|FILE_WRITE_ATTRIBUTES)) ==
DELETE_ACCESS) {
return NT_STATUS_OK;
}
As for the second test, no_share_delete_no_delete_access which failed, I
tried to fix it by checking if the parent directory is opened without
FILE_SHARE_DELETE as per the comment in the smbtorture code: "testing of
rename with no delete access on parent dir... parent dir is opened without
share_delete, so rename must fail". With the fix in place, smbd returns
STATUS_SHARING_VIOLATION if the parent directory is not opened for
FILE_SHARE_DELETE and so the test no_share_delete_no_delete_access() passes.
The side effect of the above fix is, the sub test "msword" which passed
earlier, fails. This is because msword test also opens the parent directory
without FILE_SHARE_DELETE option. I captured network traces of both the
tests and from the traces, I couldn't find any obvious difference between
these 2 tests when opening the parent directory as share access and access
mask seem to be compatible. Could you please let me know if I am missing
something while fixing the problem?
Thanks,
Shilpa
More information about the samba-technical
mailing list