Microsoft Has More SMB3/CIFS Enhancements For Linux 5.16, Including For Performance

[arQon]

Why is putting SMB3 file server inside the kernel a good idea?

It isn't. But apparently someone wanted it badly enough to just keep arguing until they got their way. I suspect that if weren't in the era of the new "mellower" Linus the idea would have been shot down early on and rightly abandoned, but instead it was allowed to progress to being ready to merge, and by the point he was pretty much stuck with taking it.

SMB3 in its own right isn't THAT terrible an idea - it's not substantially worse than having NFS there, which is the line of argument used to get it added. But once the inevitable DMA extensions come in, we'll have our very own version of a WannaCry-level remote code execution engine. Yay! :/

[lacek]

I have 0 background in this but my 2 cents is....
By making Linux more compatible with Windows it'll help making the switch from Windows to Linux easier..
functionality

Better samba needs to be an in-kernel samba? Surely Samba that works well is important. It is not only about moving. I for example have a Apple "Time Capsule" that is a major pain to use on Linux, as it uses some old SMB protocol, and GUI attempts to mount it end up in errors (while in MacOs and Windows it works out of the box). Somehow I feel that the main problem with that was not that smbd was not in kernel, but it was untested for that use case. Alternatively Apple's implementation of SMB uses some undocumented quirks.

[indepe]

the memory footprint and processor usage are important to consider. You cannot beat an in kernel server when it comes to processor efficiency. And that's important for the embedded world.

Says who? I think this is more or less contradicted by the finding that with IO_URING throughput is increased 10x. (Unless that is somehow a misleading benchmark.)

There is usually a close relationship between throughput and processor effiency.

[NobodyXu]

Sounds like a huge security vulnerability waiting to happen to me. Not involving the OS sounds really dangerous to just have a computer on the network take content from memory. What if the remote computer took your AES decryption keys for example?

I think RDMA will only happen after authentication.
It is just another way to optimize the transfer of data by using zero-copy networking.

Though maybe io-uring in future will also support zero-copy.

[oleid]

I have to? I'd think IO_URING would work for "basic" SMB just as well.

the memory footprint and processor usage are important to consider. You cannot beat an in kernel server when it comes to processor efficiency. And that's important for the embedded world.

[indepe]

Sigh.. Samba features outside of IO_URING. Today there's not Samba built on top of KSMBD. But this is being looked at. So, TODAY... KSMBD, limited when compared to features of Samba.

IO_URING is a underlying technology, not a feature. I'm not talking about features. It's a way to run multiple requests asynchronously, which means they are initiated in userspace but then handled within the kernel. Without having to add code inside the kernel. (A bit of a simplification, but I hope it gets the point across.)

[oleid]

Sounds like a huge security vulnerability waiting to happen to me. Not involving the OS sounds really dangerous to just have a computer on the network take content from memory. What if the remote computer took your AES decryption keys for example?

well, clearly the operating system and the samba server will be involved at some point, i.e. when defining the memory area to be shared.

[kylew77]

I think ksmb supports RDMA:

Remote Direct Memory Access (RDMA) is a technology that allows computers in a network to exchange data in main memory without involving the processor, cache or operating system of either computer. Like locally based Direct Memory Access (DMA), RDMA improves throughput and performance because it frees up resources.

Copied from searchstorage.techtarget.com/definition/Remote-Direct-Memory-Access

Sounds like a huge security vulnerability waiting to happen to me. Not involving the OS sounds really dangerous to just have a computer on the network take content from memory. What if the remote computer took your AES decryption keys for example?

[cjcox]

However it is the result of decisions that were made some time ago, and IO_URING didn't come into being just today. It doesn't mean that it was technically a wise decision to accept the making of an in-kernel implementation in the first place (which then has to be supported forever). And I'm worried that more such decisions will be made.

Sigh.. Samba features outside of IO_URING. Today there's not Samba built on top of KSMBD. But this is being looked at. So, TODAY... KSMBD, limited when compared to features of Samba.

[Chrispynut]

Renews my tears that my ISP supplied router (for FTTH that was installed a few months ago) uses SMB1.
GJ Vodafone, GFJ!