linux-kernel - Re: [PATCH] asm/sections: Check for overflow in memory

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] [day] [month] [year] [list]

Date:   Wed, 18 Dec 2019 15:49:43 +0100
From:   Vincent Whitchurch <vincent.whitchurch@...s.com>
To:     Russell King - ARM Linux admin <linux@...linux.org.uk>
Cc:     "akpm@...ux-foundation.org" <akpm@...ux-foundation.org>,
        "treding@...dia.com" <treding@...dia.com>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        "arnd@...db.de" <arnd@...db.de>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] asm/sections: Check for overflow in memory_contains()

On Tue, Dec 17, 2019 at 11:28:31AM +0100, Russell King - ARM Linux admin wrote:
> On Tue, Dec 17, 2019 at 11:22:38AM +0100, Vincent Whitchurch wrote:
> > ARM uses memory_contains() from its stacktrace code via this function:
> > 
> >  static inline bool in_entry_text(unsigned long addr)
> >  {
> >  	return memory_contains(__entry_text_start, __entry_text_end,
> >  			       (void *)addr, 1);
> >  }
> > 
> > addr is taken from the stack and can be a completely invalid.  If addr
> > is 0xffffffff, there is an overflow in the pointer arithmetic in
> > memory_contains() and in_entry_text() incorrectly returns true.
> > 
> > Fix this by adding an overflow check.  The check is done on unsigned
> > longs to avoid undefined behaviour.
> > 
> > Signed-off-by: Vincent Whitchurch <vincent.whitchurch@...s.com>
> > ---
> >  include/asm-generic/sections.h | 10 +++++++++-
> >  1 file changed, 9 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/asm-generic/sections.h b/include/asm-generic/sections.h
> > index d1779d442aa5..e6e1b381c5df 100644
> > --- a/include/asm-generic/sections.h
> > +++ b/include/asm-generic/sections.h
> > @@ -105,7 +105,15 @@ static inline int arch_is_kernel_initmem_freed(unsigned long addr)
> >  static inline bool memory_contains(void *begin, void *end, void *virt,
> >  				   size_t size)
> >  {
> > -	return virt >= begin && virt + size <= end;
> > +	unsigned long membegin = (unsigned long)begin;
> > +	unsigned long memend = (unsigned long)end;
> > +	unsigned long objbegin = (unsigned long)virt;
> > +	unsigned long objend = objbegin + size;
> > +
> > +	if (objend < objbegin)
> > +		return false;
> > +
> > +	return objbegin >= membegin && objend <= memend;
> 
> Would merely changing to:
> 
> 	return virt >= begin && virt <= end - size;
> 
> be sufficient ?  Is end - size possible to underflow?

Something like this would trigger an underflow and return an incorrect
result with that expression, wouldn't it?

 memory_contains((void *)0x0000, (void *)0x1000, (void *)0x0, 0x1001))

AFAICS no current callers actually send in an object size which is
larger than the size of the memory, but perhaps it's best to be
defensive?