Re: [Cluster-devel] [PATCH 06/32] sched: Add task_struct->faults_disabled_mapping

[Kent Overstreet]

On Fri, May 26, 2023 at 12:25:31AM +0200, Andreas Grünbacher wrote:
> Am Di., 23. Mai 2023 um 18:28 Uhr schrieb Christoph Hellwig <hch@xxxxxxxxxxxxx>:
> > On Tue, May 23, 2023 at 03:34:31PM +0200, Jan Kara wrote:
> > > I've checked the code and AFAICT it is all indeed handled. BTW, I've now
> > > remembered that GFS2 has dealt with the same deadlocks - b01b2d72da25
> > > ("gfs2: Fix mmap + page fault deadlocks for direct I/O") - in a different
> > > way (by prefaulting pages from the iter before grabbing the problematic
> > > lock and then disabling page faults for the iomap_dio_rw() call). I guess
> > > we should somehow unify these schemes so that we don't have two mechanisms
> > > for avoiding exactly the same deadlock. Adding GFS2 guys to CC.
> > >
> > > Also good that you've written a fstest for this, that is definitely a useful
> > > addition, although I suspect GFS2 guys added a test for this not so long
> > > ago when testing their stuff. Maybe they have a pointer handy?
> >
> > generic/708 is the btrfs version of this.
> >
> > But I think all of the file systems that have this deadlock are actually
> > fundamentally broken because they have a mess up locking hierarchy
> > where page faults take the same lock that is held over the the direct I/
> > operation.  And the right thing is to fix this.  I have work in progress
> > for btrfs, and something similar should apply to gfs2, with the added
> > complication that it probably means a revision to their network
> > protocol.
> 
> We do disable page faults, and there can be deadlocks in page fault
> handlers while no page faults are allowed.
> 
> I'm roughly aware of the locking hierarchy that other filesystems use,
> and that's something we want to avoid because of two reasons: (1) it
> would be an incompatible change, and (2) we want to avoid cluster-wide
> locking operations as much as possible because they are very slow.
> 
> These kinds of locking conflicts are so rare in practice that the
> theoretical inefficiency of having to retry the operation doesn't
> matter.

Would you be willing to expand on that? I'm wondering if this would
simplify things for gfs2, but you mention locking heirarchy being an
incompatible change - how does that work?

> 
> > I'm absolutely not in favour to add workarounds for thes kind of locking
> > problems to the core kernel.  I already feel bad for allowing the
> > small workaround in iomap for btrfs, as just fixing the locking back
> > then would have avoid massive ratholing.
> 
> Please let me know when those btrfs changes are in a presentable shape ...

I would also be curious to know what btrfs needs and what the approach
is there.