[Samba] Debug file_new then open_file_ntcreate fails with doesn't exist

[Jeremy Allison]

On Fri, Jul 16, 2021 at 01:44:17PM -0700, Matthew Schumacher via samba wrote:
>On 7/16/21 12:12 PM, Jeremy Allison wrote:
>>>
>>>Anybody see this before or thoughts on work arounds?  I'm running 
>>>the latest stable samba 4.14.6 in AD mode.
>>
>>We haven't seen that, it looks strange. I take it there's no
>>other application running that could race and delete ?
>>
>>Can you check the recently release 4.15.rc1 to see if it's
>>something that got fixed ? There have been *lots* of changes
>>in the pathname area between 4.14.x -> 4.15.rc1.
>
>I isolated it to make sure there wasn't a race and checked with 
>smbstatus to ensure nothing else is using the share.
>
>Looks like the same issue with 4.15rc1  The create was called at 
>12:23:28.127137 and the open at 12:23:28.129007 which is only 2ms 
>apart.
>
>Since the samba process to service both requests is the same, it 
>doesn't seem like it's a samba issue, perhaps the filesystem is slow?  
>The samba server is virtualized in qemu.
>
>Any suggestions on other stuff to look at will be helpful.

Once the openat() called from source3/modules_vfs_default.c:

         result = openat(fsp_get_pathref_fd(dirfsp),
                         smb_fname->base_name,
                         flags,
                         mode);

with O_CREAT returns, the file *must* exist, slow filesystem
or not. Can you do an strace on the smbd process whilst it's
doing this call ?