Date:   Sun, 23 Apr 2023 23:56:48 +0100
From:   Lorenzo Stoakes <lstoakes@...il.com>
To:     Dave Chinner <david@...morbit.com>
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        Jason Gunthorpe <jgg@...pe.ca>, Jens Axboe <axboe@...nel.dk>,
        Matthew Wilcox <willy@...radead.org>,
        Dennis Dalessandro <dennis.dalessandro@...nelisnetworks.com>,
        Leon Romanovsky <leon@...nel.org>,
        Christian Benvenuti <benve@...co.com>,
        Nelson Escobar <neescoba@...co.com>,
        Bernard Metzler <bmt@...ich.ibm.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        Ian Rogers <irogers@...gle.com>,
        Adrian Hunter <adrian.hunter@...el.com>,
        Bjorn Topel <bjorn@...nel.org>,
        Magnus Karlsson <magnus.karlsson@...el.com>,
        Maciej Fijalkowski <maciej.fijalkowski@...el.com>,
        Jonathan Lemon <jonathan.lemon@...il.com>,
        "David S . Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        Christian Brauner <brauner@...nel.org>,
        Richard Cochran <richardcochran@...il.com>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jesper Dangaard Brouer <hawk@...nel.org>,
        John Fastabend <john.fastabend@...il.com>,
        linux-fsdevel@...r.kernel.org, linux-perf-users@...r.kernel.org,
        netdev@...r.kernel.org, bpf@...r.kernel.org
Subject: Re: [PATCH] mm/gup: disallow GUP writing to file-backed mappings by
 default

On Mon, Apr 24, 2023 at 08:29:41AM +1000, Dave Chinner wrote:
> On Sat, Apr 22, 2023 at 02:37:05PM +0100, Lorenzo Stoakes wrote:
> > +/*
> > + * Writing to file-backed mappings using GUP is a fundamentally broken operation
> > + * as kernel write access to GUP mappings may not adhere to the semantics
> > + * expected by a file system.
> > + *
> > + * In most instances we disallow this broken behaviour, however there are some
> > + * exceptions to this enforced here.
> > + */
> > +static inline bool can_write_file_mapping(struct vm_area_struct *vma,
> > +					  unsigned long gup_flags)
> > +{
> > +	struct file *file = vma->vm_file;
> > +
> > +	/* If we aren't pinning then no problematic write can occur. */
> > +	if (!(gup_flags & (FOLL_GET | FOLL_PIN)))
> > +		return true;
> > +
> > +	/* Special mappings should pose no problem. */
> > +	if (!file)
> > +		return true;
>
> Ok...
>
> > +
> > +	/* Has the caller explicitly indicated this case is acceptable? */
> > +	if (gup_flags & FOLL_ALLOW_BROKEN_FILE_MAPPING)
> > +		return true;
> > +
> > +	/* shmem and hugetlb mappings do not have problematic semantics. */
> > +	return vma_is_shmem(vma) || is_file_hugepages(file);
> > +}
>
> This looks backwards. We only want the override to occur when the
> target won't otherwise allow it. i.e.  This should be:
>
> 	if (vma_is_shmem(vma))
> 		return true;
> 	if (is_file_hugepages(vma)
> 		return true;
>
> 	/*
> 	 * Issue a warning only if we are allowing a write to a mapping
> 	 * that does not support what we are attempting to do functionality.
> 	 */
> 	if (WARN_ON_ONCE(gup_flags & FOLL_ALLOW_BROKEN_FILE_MAPPING))
> 		return true;
> 	return false;
>
> i.e. we only want the warning to fire when the override is
> triggered - indicating that the caller is actually using a file
> mapping in a broken way, not when it is being used on
> file/filesystem that actually supports file mappings in this way.
>
> >  static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
> >  {
> >  	vm_flags_t vm_flags = vma->vm_flags;
> >  	int write = (gup_flags & FOLL_WRITE);
> >  	int foreign = (gup_flags & FOLL_REMOTE);
> > +	bool vma_anon = vma_is_anonymous(vma);
> >
> >  	if (vm_flags & (VM_IO | VM_PFNMAP))
> >  		return -EFAULT;
> >
> > -	if (gup_flags & FOLL_ANON && !vma_is_anonymous(vma))
> > +	if ((gup_flags & FOLL_ANON) && !vma_anon)
> >  		return -EFAULT;
> >
> >  	if ((gup_flags & FOLL_LONGTERM) && vma_is_fsdax(vma))
> > @@ -978,6 +1008,10 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
> >  		return -EFAULT;
> >
> >  	if (write) {
> > +		if (!vma_anon &&
> > +		    WARN_ON_ONCE(!can_write_file_mapping(vma, gup_flags)))
> > +			return -EFAULT;
>
> Yeah, the warning definitely belongs in the check function when the
> override triggers allow broken behaviour to proceed, not when we
> disallow a write fault because the underlying file/filesystem does
> not support the operation being attempted.

I disagree for two reasons:-

1. There are places in the kernel that rely on this broken behaviour, most
   notably ptrace (and /proc/$pid/mem), but also the other places where you
   can see I've added this flag. I'm not sure spamming warnings for
   ordinary cases would be useful.

2. The purpose of putting a warning here is to catch any case I might have
   missed where broken behaviour is required, but now disallowed, because it
   might actually be hard for a GUP user to track down that this is why the
   GUP is no longer functioning (since all they'll see is an -EFAULT).

This warned upon check should in reality not occur, because it implies the
GUP user is trying to do something broken and is _not_ explicitly telling
GUP that it knows it's doing it and can live with the consequences. And on
that basis, is worthy of a warning so we know we have to go put this flag
in that place (and know it is a source of problematic GUP usage), or fix
the caller.

An example case is placing breakpoints in gdb, without the flag being set
for /proc/$pid/mem this will just fail. Raising a kernel warning when a
user places a breakpoint seems... unhelpful :)

>
> -Dave.
> --
> Dave Chinner
> david@...morbit.com

Powered by blists - more mailing lists