It's definitely not very good, though. There is still a major memory leak,
and it can leak diskspace too (fixed by the fsck that invariably triggers
at the next reboot, but still a problem).
> A nasty bug triggered by removing
> busy files also has disappeared.
Yes, that bug was due to the way we removed the dentry even though we
really shouldn't have (the inode associated with the dentry was still in
use when we severed the connection). Thanks to the half-working garbage
collection, the new dcache could do this the right way.
> One minor problem I'm having -- after I mount and umount a volume,
> there's still an inode held on the volume, so I can't mount it again.
> Is there a call to invalidate the dcache for a device?
Not yet. That is going to be easy once I get the garbage collection to be
correct, right now it's "getting close", but it's not there yet..
I'll make another pre-patch that gets some more cases correct (notably
when it comes to file deletion), but it still won't fix the obvious
problems with unmounting.
Linus