[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <36956515-35f9-467e-9102-d6bff55f28b7@lucifer.local>
Date: Thu, 1 May 2025 11:37:15 +0100
From: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
To: Jann Horn <jannh@...gle.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
David Hildenbrand <david@...hat.com>,
"Liam R . Howlett" <Liam.Howlett@...cle.com>,
Vlastimil Babka <vbabka@...e.cz>, Mike Rapoport <rppt@...nel.org>,
Pedro Falcato <pfalcato@...e.de>, linux-fsdevel@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-mm@...ck.org,
Alexander Viro <viro@...iv.linux.org.uk>,
Christian Brauner <brauner@...nel.org>, Jan Kara <jack@...e.cz>
Subject: Re: [RFC PATCH 1/3] mm: introduce new .mmap_proto() f_op callback
On Wed, Apr 30, 2025 at 11:44:15PM +0200, Jann Horn wrote:
> On Wed, Apr 30, 2025 at 9:54 PM Lorenzo Stoakes
> <lorenzo.stoakes@...cle.com> wrote:
> > Provide a means by which drivers can specify which fields of those
> > permitted to be changed should be altered to prior to mmap()'ing a
> > range (which may either result from a merge or from mapping an entirely new
> > VMA).
> >
> > Doing so is substantially safer than the existing .mmap() calback which
> > provides unrestricted access to the part-constructed VMA and permits
> > drivers and file systems to do 'creative' things which makes it hard to
> > reason about the state of the VMA after the function returns.
> >
> > The existing .mmap() callback's freedom has caused a great deal of issues,
> > especially in error handling, as unwinding the mmap() state has proven to
> > be non-trivial and caused significant issues in the past, for instance
> > those addressed in commit 5de195060b2e ("mm: resolve faulty mmap_region()
> > error path behaviour").
> >
> > It also necessitates a second attempt at merge once the .mmap() callback
> > has completed, which has caused issues in the past, is awkward, adds
> > overhead and is difficult to reason about.
> >
> > The .mmap_proto() callback eliminates this requirement, as we can update
> > fields prior to even attempting the first merge. It is safer, as we heavily
> > restrict what can actually be modified, and being invoked very early in the
> > mmap() process, error handling can be performed safely with very little
> > unwinding of state required.
>
> I wonder if this requires adjustments to the existing users of
> call_mmap() that use call_mmap() for forwarding mmap operations to
> some kind of backing file. In particular fuse_passthrough_mmap(),
> which I think can operate on fairly arbitrary user-supplied backing
> files (for context, I think fuse_backing_open() allows root to just
> provide an fd to be used as backing file).
Yeah the fact these exist is just another example of us being far, far, far
too permissive on this stuff imo.
I mean it's useful ofc, but the fact you have multiple layers of being able
to do _anything_ isn't great...
>
> I guess the easiest approach would be to add bailouts to those if an
> ->mmap_proto handler exists for now, and revisit this if we ever want
> to use ->mmap_proto for more normal types of files?
Yeah good point, luckily we abstract to call_mmap(), will have that bail
out in that case, thanks!
I think by implication we shouldn't allow .mmap_proto() and .mmap() to
co-exist, rather in future we can add additional callbacks as needed (see
discussion with David).
Will respin accordingly... :)
Thanks for taking a look, much appreciated to both you and David! :)
Cheers, Lorenzo
Powered by blists - more mailing lists