Userspace mount.cifs -o cifsacl Mechanism
Subsystems (2)
A message that is not in the archive (sent off-list or not yet received): [email protected]
I am trying to grasp the mechanism of cifsacl in mount.cifs.
I fought my way from joining a Debian to AD, using sssd, doing dyn_dns to heed Windows Server's RFC 2136 (found systemd-resolved is incompatible), reusing krb5 ticket for mount.cifs, piping sss idmap to cifs-idmap, now currently stuck at cifsacl.
I am doing this
export KRB5CCNAME=/tmp/krb5cc_10001_xxxxx
mount -t cifs //nas.domain /mnt/drive -o sec=krb5,cruid=10001,cifsacl
Something is wrong with the cifs.upcall that it does not like extra xxxxx unique identifiers being appended after uid. I can blame sssd for doing this.
/etc/cifs-utils/idmap-plugin --> /usr/lib/.../cifs_idmap_sss.so
Once mounted, I could use getcifsacl to look at the DACLs with the resolved name from sss.so But the permission is not enforced through the ACL.
I am seeing OWNER in cifsacl is set as the uid on files.
But all ACLs on Windows AD groups are not translated into ACL. So unless I go do -o uid and gid at mount.cifs, they all remain at root. Is this the current limitations of cifs or I am doing something wrong?
Many thanks.
Lucas Cang.
Have you tried experiments with multiuser mount option to see what SID the different user's files get created as?
Might be a good sssd question how to enable debugging for their plugin
Have you also looked at this:
https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/7/html/windows_integration_guide/smb-sssd
Quoted 22 lines · Household Cang
I am trying to grasp the mechanism of cifsacl in mount.cifs.
I fought my way from joining a Debian to AD, using sssd, doingdyn_dnsto heed Windows Server's RFC 2136 (found systemd-resolved is incompatible), reusing krb5 ticket for mount.cifs, piping sss idmap to cifs-idmap, now currently stuck at cifsacl.I am doing this
export KRB5CCNAME=/tmp/krb5cc_10001_xxxxx
mount -t cifs //nas.domain/mnt/drive-o sec=krb5,cruid=10001,cifsaclSomething is wrong with the cifs.upcall that it does not like extra xxxxx unique identifiers being appended after uid. I can blame sssd for doing this.
/etc/cifs-utils/idmap-plugin-->/usr/lib/.../cifs_idmap_sss.soOnce mounted, I could use getcifsacl to look at the DACLs with the resolved name from sss.so But the permission is not enforced through the ACL.
I am seeing OWNER in cifsacl is set as the uid on files.
But all ACLs on Windows AD groups are not translated into ACL. So unless I go do -o uid and gid at mount.cifs, they all remain at root. Is this the current limitations of cifs or I am doing something wrong?Many thanks.
Lucas Cang.
Signature · 3 lines
Thanks, Steve