Hello Steve and other cifs developers, On 2022-04-07 at 16:25 -0500 Steve French via samba sent off: > If you have particular hot bugs for cifs.ko that you need fixed - please > let me know. back in 2020 we've added the cifs mailing list as qa contact in bugzilla, so that bugzilla bugs get more visible to the developers. At the same point I also went through all the open bug reports of cifs vfs, which had partly been very old. I also closed a bunch of them as they had been fixed a while after they were reported there - but not because they were reported in bugzilla obviously. I was hoping to improve that situation with bug reports getting not enough attention my adding the cifs mailinglist as qa contact, this was my motivation. Just pointing out those bugs that I myself reported in bugzilla.samba.org in 2020, all stayed uncommented till today; except for the "better error message" bug all of them are important for enterprise customers, this is also where they popped up: 14398 cifs vfs should pause if krb5 ticket is not valid 14440 creator owner (S-1-3-0) ACE not honored 14506 cifs mount with missing krb5 key should give better error message 14507 cifs ACL exec permission granted where it should be denied And last but not least a big topic: 14499 expose NT ACLs via system.nfs4_acl EA This idea popped up in the discussion of the 2020 SambaXP discussion after your talk. Having a standardized permission management tool like the nfs4-acl-tools is really something that is important. Also the fact that Linux still has no standardized ACL implementation (NFS4 ACL are the only standardized ones) in the kernel is preventing many enterprise customers to use Linux on client machines. Without that, permission management is such a pain that they usually give up any client installation efforts sooner or later. I think the cifs vfs developers would have the power to convince the responsible kernel developers to add this to the kernel. I can say from my experience at various customer sites very clearly that cifs vfs will *not* be accepted widely in the enterprise world, without generic NFS4 ACLs implemented in the kernel alongside. > SMB3.1.1 is simpler in some ways than NFS (no SunRPC to deal with) and > should be easier to fix bugs in many cases. maybe it is simpler, but for the above mentioned reasons, SMB on the Linux client side is no option for most enterprise environments. They prefer NFS and I understand why. I really wish this would change, this is why I write these lines so bluntly. Björn -- SerNet GmbH - Bahnhofsallee 1b - 37081 Göttingen phone: +495513700000 mailto:contact@xxxxxxxxxx AG Göttingen: HR-B 2816 - https://www.sernet.com Manag. Directors Johannes Loxen and Reinhild Jung data privacy policy https://www.sernet.de/privacy Samba eXPerience 2022 - online edition! from May 31st to June 2nd, 2022 at Zoom sponsored by Google, Microsoft & SerNet more information at https://sambaXP.org
