samba - Jun 2018 - works with 4.7.6, fails with 4.8.2: echo > $(mktemp /mnt/XXXXXX)

Hi,

I have a regression test suite that creates this simple share:

[tmp]
    comment = Temp Directory
    guest ok = Yes
    path = /tmp
    read only = No

This is mounted via SMB1:
mount //localhost/tmp -o guest,vers=1.0 /mnt

The client (ubuntu 18.10 dev) is using cifs-utils 2:6.8-2 from debian
(it's a sync). Client and server are the same machine (localhost).

When the server is 2:4.8.2+dfsg-2ubuntu1 (a package I'm preparing,
based on debian's 4.8.2+dfsg-2), this fails:

root at ubuntu:~# mktemp /mnt/test.XXXXXX
/mnt/test.iev5IU
root at ubuntu:~# echo hello > /mnt/test.iev5IU
-su: /mnt/test.iev5IU: Permission denied
root at ubuntu:~# l /mnt/test.iev5IU
-rw-------+ 1 nobody nogroup 0 Jun 25 17:40 /mnt/test.iev5IU
root at ubuntu:~#


When the server is running 2:4.7.6+dfsg~ubuntu-0ubuntu2, it works:
root at ubuntu:~# mktemp /mnt/test.XXXXXX
/mnt/test.yGv3JP
root at ubuntu:~# echo hello > /mnt/test.yGv3JP
root at ubuntu:~# l /mnt/test.yGv3JP
-rw-------+ 1 nobody nogroup 6 Jun 25 17:41 /mnt/test.yGv3JP

The output of getcifsacl is quite different in both cases:

# 4.7.6 server:
root at ubuntu:~# getcifsacl /mnt/test.yGv3JP
REVISION:0x1
CONTROL:0x9004
OWNER:S-1-5-21-3096988004-4058363749-2517819802-501
GROUP:S-1-22-2-65534
ACL:S-1-5-21-3096988004-4058363749-2517819802-501:ALLOWED/0x0/RW
ACL:S-1-22-2-65534:ALLOWED/0x0/
ACL:S-1-1-0:ALLOWED/0x0/


# 4.8.2 server:
root at ubuntu:~# getcifsacl /mnt/test.iev5IU
getxattr error: 13
REVISION:0x0
CONTROL:0x0


The cifs mount in both cases is shown as:
root at ubuntu:~# mount -t cifs
//localhost/tmp on /mnt type cifs
(rw,relatime,vers=1.0,sec=none,cache=strict,domain=,uid=0,noforceuid,gid=0,noforcegid,addr=127.0.0.1,soft,unix,posixpaths,serverino,mapposix,acl,rsize=1048576,wsize=65536,echo_interval=60,actimeo=1)


Is this a bug that was fixed, a bug that was introduced, or neither
and I'm holding it wrong?

Thanks!

I narrowed it a bit further and can see the change happening between
4.8.0 and 4.8.1. It works with 4.8.0, and I get permission denied in
4.8.1. This time I tested with plain upstream tarballs, no
distribution patches.

On Mon, Jun 25, 2018 at 2:50 PM, Andreas Hasenack <andreas at
canonical.com> wrote:
> Hi,
>
> I have a regression test suite that creates this simple share:
>
> [tmp]
>     comment = Temp Directory
>     guest ok = Yes
>     path = /tmp
>     read only = No
>
> This is mounted via SMB1:
> mount //localhost/tmp -o guest,vers=1.0 /mnt
>
> The client (ubuntu 18.10 dev) is using cifs-utils 2:6.8-2 from debian
> (it's a sync). Client and server are the same machine (localhost).
>
> When the server is 2:4.8.2+dfsg-2ubuntu1 (a package I'm preparing,
> based on debian's 4.8.2+dfsg-2), this fails:
>
> root at ubuntu:~# mktemp /mnt/test.XXXXXX
> /mnt/test.iev5IU
> root at ubuntu:~# echo hello > /mnt/test.iev5IU
> -su: /mnt/test.iev5IU: Permission denied
> root at ubuntu:~# l /mnt/test.iev5IU
> -rw-------+ 1 nobody nogroup 0 Jun 25 17:40 /mnt/test.iev5IU
> root at ubuntu:~#
>
>
> When the server is running 2:4.7.6+dfsg~ubuntu-0ubuntu2, it works:
> root at ubuntu:~# mktemp /mnt/test.XXXXXX
> /mnt/test.yGv3JP
> root at ubuntu:~# echo hello > /mnt/test.yGv3JP
> root at ubuntu:~# l /mnt/test.yGv3JP
> -rw-------+ 1 nobody nogroup 6 Jun 25 17:41 /mnt/test.yGv3JP
>
> The output of getcifsacl is quite different in both cases:
>
> # 4.7.6 server:
> root at ubuntu:~# getcifsacl /mnt/test.yGv3JP
> REVISION:0x1
> CONTROL:0x9004
> OWNER:S-1-5-21-3096988004-4058363749-2517819802-501
> GROUP:S-1-22-2-65534
> ACL:S-1-5-21-3096988004-4058363749-2517819802-501:ALLOWED/0x0/RW
> ACL:S-1-22-2-65534:ALLOWED/0x0/
> ACL:S-1-1-0:ALLOWED/0x0/
>
>
> # 4.8.2 server:
> root at ubuntu:~# getcifsacl /mnt/test.iev5IU
> getxattr error: 13
> REVISION:0x0
> CONTROL:0x0
>
>
> The cifs mount in both cases is shown as:
> root at ubuntu:~# mount -t cifs
> //localhost/tmp on /mnt type cifs
>
(rw,relatime,vers=1.0,sec=none,cache=strict,domain=,uid=0,noforceuid,gid=0,noforcegid,addr=127.0.0.1,soft,unix,posixpaths,serverino,mapposix,acl,rsize=1048576,wsize=65536,echo_interval=60,actimeo=1)
>
>
> Is this a bug that was fixed, a bug that was introduced, or neither
> and I'm holding it wrong?
>
> Thanks!