lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Sat, 25 May 2019 11:09:15 -0700 (PDT)
From:   Hugh Dickins <hughd@...gle.com>
To:     Sebastian Andrzej Siewior <bigeasy@...utronix.de>
cc:     Hugh Dickins <hughd@...gle.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Mike Rapoport <rppt@...ux.ibm.com>,
        Andrea Arcangeli <aarcange@...hat.com>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org, Borislav Petkov <bp@...e.de>,
        Pavel Machek <pavel@....cz>,
        Dave Hansen <dave.hansen@...ux.intel.com>
Subject: Re: [PATCH] mm/gup: continue VM_FAULT_RETRY processing event for
 pre-faults

On Sat, 25 May 2019, Sebastian Andrzej Siewior wrote:
> On 2019-05-24 15:22:51 [-0700], Hugh Dickins wrote:
> > I've now run a couple of hours of load successfully with Mike's patch
> > to GUP, no problem; but whatever the merits of that patch in general,
> > I agree with Andrew that fault_in_pages_writeable() seems altogether
> > more appropriate for copy_fpstate_to_sigframe(), and have now run a
> > couple of hours of load successfully with this instead (rewrite to taste):
> 
> so this patch instead of Mike's GUP patch fixes the issue you observed?

Yes.

> Is this just a taste question or limitation of the function in general?

I'd say it's just a taste question. Though the the fact that your
usage showed up a bug in the get_user_pages_unlocked() implementation,
demanding a fix, does indicate that it's a more fragile and complex
route, better avoided if there's a good simple alternative. If it were
not already on your slowpath, I'd also argue fault_in_pages_writeable()
is a more efficient way to do it.

> 
> I'm asking because it has been suggested and is used in MPX code (in the
> signal path but .mmap) and I'm not aware of any limitation. But as I
> wrote earlier to akpm, if the MM folks suggest to use this instead I am
> happy to switch.

I know nothing of MPX, beyond that Dave Hansen has posted patches to
remove that support entirely, so I'm surprised arch/x86/mm/mpx.c is
still in the tree. But peering at it now, it looks as if it's using
get_user_pages() while holding mmap_sem, whereas you (sensibly enough)
used get_user_pages_unlocked() to handle the mmap_sem for you -
the trouble with that is that since it knows it's in control of
mmap_sem, it feels free to drop it internally, and that takes it
down the path of the premature return when pages NULL that Mike is
fixing. MPX's get_user_pages() is not free to go that way.

> 
> > --- 5.2-rc1/arch/x86/kernel/fpu/signal.c
> > +++ linux/arch/x86/kernel/fpu/signal.c
> > @@ -3,6 +3,7 @@
> >   * FPU signal frame handling routines.
> >   */
> >  
> > +#include <linux/pagemap.h>
> >  #include <linux/compat.h>
> >  #include <linux/cpu.h>
> >  
> > @@ -189,15 +190,7 @@ retry:
> >  	fpregs_unlock();
> >  
> >  	if (ret) {
> > -		int aligned_size;
> > -		int nr_pages;
> > -
> > -		aligned_size = offset_in_page(buf_fx) + fpu_user_xstate_size;
> > -		nr_pages = DIV_ROUND_UP(aligned_size, PAGE_SIZE);
> > -
> > -		ret = get_user_pages_unlocked((unsigned long)buf_fx, nr_pages,
> > -					      NULL, FOLL_WRITE);
> > -		if (ret == nr_pages)
> > +		if (!fault_in_pages_writeable(buf_fx, fpu_user_xstate_size))
> >  			goto retry;
> >  		return -EFAULT;
> >  	}
> > 
> > (I did wonder whether there needs to be an access_ok() check on buf_fx;
> > but if so, then I think it would already have been needed before the
> > earlier copy_fpregs_to_sigframe(); but I didn't get deep enough into
> > that to be sure, nor into whether access_ok() check on buf covers buf_fx.)
> 
> There is an access_ok() at the begin of copy_fpregs_to_sigframe(). The
> memory is allocated from user's stack and there is (later) an
> access_ok() for the whole region (which can be more than the memory used
> by the FPU code).

Yes, but remember I know nothing of this FPU signal code, so I cannot
tell whether an access_ok(buf, size) is good enough to cover the range
of an access_ok(buf_fx, fpu_user_xstate_size).

Your "(later)" worries me a little - I hope you're not writing first
and checking the limits later; but what you're doing may be perfectly
correct, I'm just too far from understanding the details to say; but
raised the matter because (I think) get_user_pages_unlocked() would
entail an access_ok() check where fault_in_pages_writable() would not.

Hugh

Powered by blists - more mailing lists