I attached an NTFS SSD disk that had been used by Windows. I ran a virtual machine on that disk, and the virtual machine kept periodically (like once per 10 seconds) locking up and mount.ntfs's CPU usage periodically soared.
Of course, when I had run the virtual machine on the same NTFS SSD under Windows, it did not have such problems. I copied the virtual machine to a BTRFS (where Manjaro was running) disk, and those problems are gone.
Is this normal? Or did I do something wrong? Kernel 5.15.
That is normal so far. Keep in mind: NTFS-3G runs on fuse layer in the user space and not in the kernel space. It would be the same as running a 3rdparty driver for ext4/btrfs on windows, which also run in user space, while ntfs runs in kernel space.
Be aware, that even though this driver is in the kernel, there was some discussion on the kernel mailing list on what to do about it, seems to be mainly unmaintained at this point, even though it’s new. It has only seen a couple of small bugfixes since it was added, so the discussion suggested that it might get removed from the kernel again, because of the lack of maintenance.
it is not enabled by default even if you have 5.15 kernel running. to enable it by default, the recommended way remains by setting default values in udisk; NTFS - ArchWiki
efficiency wise it is better than the default ntfs-3g fuse driver, however it still requires quiet some work. been using it since 5.15.5 and had no issues, except when i had sudden loss of power which caused some form of NTFS:MFT corruption which could only be corrected by chkdsk /f in windows. and like @Strit mentioned development is dormant as soon as it was introduced into the kernel sparking debate as what to do with it.
Lol, patches made with github is never gonna get into the kernel. Linus has said that he despise the way Github does commits, because there’s so much “filler” in it.
Also:
If they can’t even get a set of PGP keys set up, there’s not much hope for maintaining a driver upstream.
Even I have a couple of commits in the upstream kernel git tree. Sure I don’t maintain a driver, but still. If I can get something upstream, so could they.