linux-kernel - Re: [RESEND PATCH] x86: add phys addr validity check for /dev/mem mmap

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:	Wed, 2 Oct 2013 20:48:03 +0200
From:	Frantisek Hrbata <fhrbata@...hat.com>
To:	"H. Peter Anvin" <hpa@...or.com>
Cc:	linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	tglx@...utronix.de, mingo@...hat.com, x86@...nel.org,
	oleg@...hat.com, kamaleshb@...ibm.com, hechjie@...ibm.com,
	akpm@...ux-foundation.org, dave.hansen@...el.com
Subject: Re: [RESEND PATCH] x86: add phys addr validity check for /dev/mem
 mmap

On Wed, Oct 02, 2013 at 11:36:09AM -0700, H. Peter Anvin wrote:
> On 10/02/2013 11:31 AM, Frantisek Hrbata wrote:
> > On Wed, Oct 02, 2013 at 10:46:35AM -0700, H. Peter Anvin wrote:
> >> On 10/02/2013 09:05 AM, Frantisek Hrbata wrote:
> >>> +
> >>> +int valid_phys_addr_range(phys_addr_t addr, size_t count)
> >>> +{
> >>> +	return addr + count <= __pa(high_memory);
> >>> +}
> >>> +
> >>> +int valid_mmap_phys_addr_range(unsigned long pfn, size_t count)
> >>> +{
> >>> +	resource_size_t addr = (pfn << PAGE_SHIFT) + count;
> >>> +	return phys_addr_valid(addr);
> >>> +}
> >>>
> >>
> >> The latter has overflow problems.
> > 
> > Could you please specify what overflow problems do you mean?
> 
> Consider if pfn + count overflows and wraps around, or if (pfn <<
> PAGE_SHIFT) pushes bits out to the left.

Ok, maybe I'm missing something, but isn't this handled in
do_mmap_pgoff?

	/* offset overflow? */
	if ((pgoff + (len >> PAGE_SHIFT)) < pgoff)
		return -EOVERFLOW;

Anyway I can take a closer look and if this can really happen I can fix it.
		       
> 
> >> The former I realize matches the current /dev/mem, but it is still just
> >> plain wrong in multiple ways.
> > 
> > I guess that you are talking about /dev/mem implementation generelly, because
> > this patch is exactly the same as the first one. All I'm trying to do here is to
> > fix this simple problem, which was reported by a customer, using IMHO the least
> > invasive way. Anyway is there any description what is wrong with /dev/mem
> > implementation? Maybe I can try to take a look.
> > 
> 
> The bottom line is that read/write to /dev/mem should be able to access
> the same memory that we can mmap().  Having two different tests is
> ridiculous.

Ok, I can try to look into this. I just want to point out that some other archs
like arm are doing it the same way. I simply replaced the generic check functions
in drivers/char/mem.c with x86 specific ones.

Thanks

> 
> 	-hpa
> 
> 

-- 
Frantisek Hrbata
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/