Dear Volumionauts,
@gkkpch - Thanks for raising this it’s a very helpful clue.
Lets properly investigate the reported performance drop in NAS scanning under Bookworm Alpha compared to Buster, especially when using CIFS mounts. Based on your observations, here’s what we have so far:
Baseline Comparison Request
We’d like to ask Volumionauts to help confirm:
Can you please run the following test and share your findings?
-
Install v0.063 (Bookworm, previous kernel)
-
Run a full library scan from your NAS
-
Then, repeat the same on v0.066 (current kernel)
-
Let us know:
- Number of tracks detected per refresh
- Any visible delay or bottleneck
- What
vers=
setting you are using (if any)
Initial Observation from Your Case
You found that:
- Mounting without
vers=
caused a downgrade in performance, with the mount defaulting tovers=3.11
- Explicitly setting
vers=1.0
restored the expected scan speed
This confirms a possible regression or increased overhead when using SMB 3.x defaults with kernel 6.12.x in combination with your NAS.
Why This Matters
Kernel 6.12 (and Bookworm) has updated CIFS behavior. When left to auto-negotiate, it often defaults to vers=3.11
, which introduces:
- Encrypted sessions
- Multi-channel negotiation
- Increased metadata overhead
These may significantly affect scan performance on:
- Low-powered NAS systems
- Atom-class SBCs
- Older networking stacks
What Would Help
Try mounts with:
vers=3.0
vers=2.1
vers=1.0
Let’s find the boundary where speed begins to degrade.
Thanks again for putting this on our radar.
Kind Regards,