Support Is Being Worked On For Root File-System Support Over SMB Protocol

[oiaohm]

SMB has its own issues for sure, but in the real world its THE filesystem sharing protocol that everyone uses. Well except maybe some universities that still have networks of diskless Sun or HP Apollo workstations that have been installed in the 1980s and never updated since. From a purely practical PoV, I think that Linux should focus primarily on having a rock solid, world class support for SMB, both as client and as server, with feature complete support for root on SMB, authentication, ACLs and all the bells and whistles.

Linux ACLs are different to Windows ACLs. Supporting root on SMB is one thing. Having a SMB server that does posix ACL and other stuff Linux OS exactly want that is a problem in itself.

SMB file-system has more than 1 form. IBM form of SMB starts in 1980s as well. So you do find some really old stuff using SMB1 with Posix ACLs instead of Windows ACLs. Windows ACLs is a latter addition to SMB protocol.

Linux root on SMB to work well I still see requiring some non windows server running SMB. Its a pity that Samba 4 ADS does not support Posix ACLs shares.

[jacob]

Linux ACLs are different to Windows ACLs. Supporting root on SMB is one thing. Having a SMB server that does posix ACL and other stuff Linux OS exactly want that is a problem in itself.

SMB file-system has more than 1 form. IBM form of SMB starts in 1980s as well. So you do find some really old stuff using SMB1 with Posix ACLs instead of Windows ACLs. Windows ACLs is a latter addition to SMB protocol.

Linux root on SMB to work well I still see requiring some non windows server running SMB. Its a pity that Samba 4 ADS does not support Posix ACLs shares.

The ACL issue is a very real one but I would say the onus is on Linux in this case. Linux ACLs are more or less the almost-POSIX ACLs which are simply inferior to the Windows/NFS4 ACLs, that MacOS and FreeBSD support too. So the real answer, in my opinion, is not that Linux can't support SMB well because of its different ACLs, but rather Linux should at long last close the gap, ditch its half-assed ACL support which sucks and therefore is seldom used anyway, and get proper modern ACLs for network shares and local filesystems alike like all other relevant OSes do. That would be the first step.

Of course there are multiple implementations of SMB with varying degrees of compatibility but that is really a non-problem. In practical terms, no-one is going to worry about SMB servers released by IBM in 1980. I don't think that anyone using the term SMB today means anything other than Samba v4 and Windows 2012/2016.

[oiaohm]

The ACL issue is a very real one but I would say the onus is on Linux in this case. Linux ACLs are more or less the almost-POSIX ACLs which are simply inferior to the Windows/NFS4 ACLs, that MacOS and FreeBSD support too.

Really you just got something horrible wrong.

https://forums.freebsd.org/threads/nfsv4-acls-not-being-respected-by-bsd.40982/

MacOS and FreeBSD run NFS4 ACLs in Solaris mode not Windows. Yes a draft that never made it into the NFS4 standard.

There is really only 1 party using Windows ACLs on NFS4.

Lot of was we need as NFS5 this time do the ACLs over with all the need features in fact included.

Linux gets more complex that its using xattrs for LSM modules and depends in places on differences in locking and other logic to windows as well.

Yes BSD and MacOS does at times Posix ACL processing on NFS4 ACL flags.

https://docs.microsoft.com/en-us/windows-server/administration/windows-commands/icacls

There is a major difference. Windows x bit on files also grants read. Linux world x bit on file only grants execute. Basically Windows ACL supported as per Windows equals downgraded security. So your idea that Posix ACL are inferior is wrong. Posix ACL turn out to be finer grain. LSM logic in the Linux kernel is built around that finer grain when they implement MAC. Posix ACL + Linux security descriptors of MAC like selinux is way superior to Windows ACLs. The foundation Linux kernel LSM sit on need a ACL system that is Posix ACL compatible or you under mine them.

I don't think that anyone using the term SMB today means anything other than Samba v4 and Windows 2012/2016.

Samba 4 server can be configured in Samba 3 mode to provide posix acl shares.

[milkylainen]

But -- can it run Crysis?

It sure can.
But if you run it on SMB the aliens will have Pew-pew guns and dance like Steve Ballmer.
If you run it on NFS then the aliens will bitch-slap you to death with Linus-like murder comments.

[Chewi]

Instead of bloating the kernel with this, why don't they just use an initramfs instead? I used Dracut to boot Gentoo from an nfsroot on an ancient 2.6 kernel that I'm stuck with on one ARM box. It was easy to set up and works really well.

[coder]

Man, I dosed off while reading this and had a dream about somebody's CPU falling out of their laptop. Like, it somehow dropped out of the bottom.

I guess it'd make more sense if it were their SSD.

[oiaohm]

Man, I dosed off while reading this and had a dream about somebody's CPU falling out of their laptop. Like, it somehow dropped out of the bottom.

It makes sense and can happen in some laptops(think rugged). There are a few laptops with zero force sockets holding in CPU. Replace CPU and the zero force socket has in fact failed so you close the lever and it has not locked. Some the rugged a independent cpu heat-sink and fan that glues on with heat paste and 4 screws meant to lock it in place and the out of door above that also has a set of 4 screws. To have what you had dream of either you have missed putting/stripped in 8 screws and the zero force socket failed or stripped/missed 8 screws with the laptop getting hit hard enough to break the pins off the cpu or overcome the holding force of the zero force socket. Any one of those 8 screws hold and the cpu will not fall out to the floor.

So that dream makes perfect sense with the right model laptop. Of course its not something you want to have happen to you.

[coder]

It makes sense and can happen in some laptops(think rugged).

Well, I've had two laptops with hatches on the bottom that are how you access the HDD/SSD. So, maybe it sort of tied into the idea of booting diskless systems.

There are a few laptops with zero force sockets holding in CPU.

I had a Pentium M laptop that was socketed and I actually did a successful CPU upgrade. However, the old BIOS only ran the new CPU at 1.7 GHz instead of 2.1 GHz and the GPU wasn't properly supported so I think it fell back on some VBE-based compatibility mode. And the wireless on that laptop never worked in Linux. So, even with a new CPU, SSD (yeah, I found a 2.5" PATA SSD) and maxed-out RAM, it didn't see a whole lot of use before I found a good deal on an open-box "new" Skylake laptop.

Oh, and the procedure was incredibly involved. It probably took at least 2 hours, and that was with instructions posted by a generous soul.

So that dream makes perfect sense with the right model laptop. Of course its not something you want to have happen to you.

In my laptop, the CPU was accessed from the top, by removing the keyboard (and then some).

[oiaohm]

Well, I've had two laptops with hatches on the bottom that are how you access the HDD/SSD. So, maybe it sort of tied into the idea of booting diskless systems.

That is to make end of life job of laptop simple. Hardrives/storage go in one stack and general scrap the other stack.

In my laptop, the CPU was accessed from the top, by removing the keyboard (and then some).

Main reason why CPU is fitted from underside on a proper rugged is the water proof nice keyboard is screwed in against its case from the underside to get its nice solid o-ring seal so getting in by the keyboard is totally not an option as this is hard to get back to restore IP rating.

The video is more modern.

https://www.youtube.com/watch?v=bMbRA5gElOg

The laptop in video is cpu access by underside. When I say 4 screws I should have said at least 4 screws. For the heatsink its somewhere between 4 to 12 for screw/boltdown. For the hatch on the back of case to get at the cpu its somewhere between 4 to 24. The one with 24 you really don't want to mess with there is in fact 12 different bolts with different in length size thread.... that design its out to make you strip something.

To be correct the rugged with the down face cpu changing them is not a major headache other than don't loss track of what bolt/screw/o-ring goes where.

You dream before the more modern intel socket of cpu falling out the bottom of laptop did in fact happen. Yes normally caused by some of the rugged kind of being traps to the nor careful. Note they are being asses for a reason with those 12 different screws out of 24 and the like each screw to get the right water proof seal is meant to have x torque on it. 12 different screws 12 different torque values. Problem is if you put them in the wrong places 100 percent stripped but might feel like its holding something until you turn it over and find out opps as the backs falls off the computer.

Yes some places working on rugged use to have a AFO sticker as in Ass Fell Out and that was laptop where the mounting had been stripped. Some of those had done the dropped CPU heatsink and back-plate/hatch to floor. Normally nothing else wrong other than a lot of effort to install new mountings and removing broken cpu pins from socket. Basically I know it because this is how I got a few really rugged laptops early price of a new cpu and some ram and set of right self tapping screws.

Yes I was typing screws instead of bolts because the cheap hack fixs is just making the stripping process a little worse with screws.

So what you dreamt really could happen to some people. I would class it as having really bad day if the cpu managed to fall out the back of the laptop you are holding and for that to happen someone has major-ally screwed up.

[starshipeleven]

SMB has its own issues for sure, but in the real world its THE filesystem sharing protocol that everyone uses. Well except maybe some universities that still have networks of diskless Sun or HP Apollo workstations that have been installed in the 1980s and never updated since. From a purely practical PoV, I think that Linux should focus primarily on having a rock solid, world class support for SMB, both as client and as server, with feature complete support for root on SMB, authentication, ACLs and all the bells and whistles.

Yes.

Also extend it a bit for its own purposes for the cases you don't need 100% inter-operatibility with Windows.