Re: [PATCH] fs/9p: fix inode nlink accounting

[asmadeus@codewreck.org]

Christian Schoenebeck wrote on Mon, Jan 08, 2024 at 03:55:53PM +0100:
> > I'm good with adding local wrapper functions,

(Agreed having a local wrapper that locks + use these is better than this
current patch -- v2 looks much better, thanks!)

> > I imagine these aren't
> > used in the kernel because for regular file-systems maybe you want the
> > warning that your inode link accounting is wrong.
> > I suppose we could be naughty and not use the kernel functions (which
> > themselves are basically wrappers).
> 
> Well, one half of that code is actually using atomic operations to increment/
> decrement the private counter. Which means to me those kernel functions were
> intended to be called from a concurrent context. So I don't get why the other
> variable is not atomic there. They should be I think.

I think the key difference is inode level vs superblock level -- the
inode is local and holding a lock can be faster if manipulations are
grouped together (x atomic operations are usually slower than a spinlock
and x normal operations), while the sb potentially has contentions and
would be more likely to use atomic...

> I would probably try and send a patch for changing those kernel functions and
> see if people are fine with that. But up to you.

With that said I just checked ext4 and it looks just as racy as we do in
particular the rmdir/unlink case doesn't seem to take any lock, so it's
definitely worth raising the subject on fsdevel!
I'll see how work is busy and ask later today if time allows


-- 
Dominique Martinet | Asmadeus