Re: mm: insure topdown mmap chooses addresses above security minimum

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



* Timothy Pepper <timothy.c.pepper@xxxxxxxxxxxxxxx> wrote:

> A security check is performed on mmap addresses in
> security/security.c:security_mmap_addr().  It uses mmap_min_addr to insure
> mmaps don't get addresses lower than a user configurable guard value
> (/proc/sys/vm/mmap_min_addr).  The arch specific mmap topdown searches
> look for a map candidate address all the way down to a low_limit that is
> currently hard coded as PAGE_SIZE.  Depending on compile time options
> and userspace setting the procfs tunable, the security check's view of
> the minimum allowable address may be something greater than PAGE_SIZE.
> This leaves a gap where get_unmapped_area()'s call to get_area() might
> return an address above PAGE_SIZE, but below mmap_min_addr, and thus
> get_unmapped_area() fails.
> 
> This was seen on x86_64 in the case of a topdown address space and a large
> stack rlimit, with mmap_min_addr having been set to 32k by the distro.
> This left a 28k gap where the get area search intends to place a small
> mmap, but then get_unmapped_area() stumbles at the security check.
> 
> What should have happened is the address search wraps back to a higher
> address, the search continues and perhaps succeeds.  Indeed an mmap of
> a larger size gets a topdown search that does wrap around back up into
> the rlimit stack reserve and succeeds assuming suitable free space.
> But a small mmap fits in the low gap and always fails.  It becomes
> possible to make large mmaps but not small ones.
> 
> When an explicit address hint is given, mm/mmap.c's round_hint_to_min()
> will round up to mmap_min_addr.
> 
> A topdown search's low_limit should similarly consider mmap_min_addr
> instead of just PAGE_SIZE.
> 
> Signed-off-by: Tim Pepper <timothy.c.pepper@xxxxxxxxxxxxxxx>
> Cc: linux-mm@xxxxxxxxx
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: "H. Peter Anvin" <hpa@xxxxxxxxx>
> Cc: x86@xxxxxxxxxx
> Cc: Russell King <linux@xxxxxxxxxxxxxxxx>
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> Cc: Ralf Baechle <ralf@xxxxxxxxxxxxxx>
> Cc: linux-mips@xxxxxxxxxxxxxx
> Cc: Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx>
> Cc: Paul Mackerras <paulus@xxxxxxxxx>
> Cc: linuxppc-dev@xxxxxxxxxxxxxxxx
> Cc: Paul Mundt <lethal@xxxxxxxxxxxx>
> Cc: linux-sh@xxxxxxxxxxxxxxx
> Cc: "David S. Miller" <davem@xxxxxxxxxxxxx>
> Cc: sparclinux@xxxxxxxxxxxxxxx
> --
>  arch/arm/mm/mmap.c               | 3 ++-
>  arch/mips/mm/mmap.c              | 3 ++-
>  arch/powerpc/mm/slice.c          | 3 ++-
>  arch/sh/mm/mmap.c                | 3 ++-
>  arch/sparc/kernel/sys_sparc_64.c | 3 ++-
>  arch/x86/kernel/sys_x86_64.c     | 3 ++-
>  6 files changed, 12 insertions(+), 6 deletions(-)
> 
> +	info.low_limit = max(PAGE_SIZE, PAGE_ALIGN(mmap_min_addr));
>  	info.high_limit = mm->mmap_base;
>  	info.align_mask = do_align ? (PAGE_MASK & (SHMLBA - 1)) : 0;
>  	info.align_offset = pgoff << PAGE_SHIFT;

>  		info.flags = VM_UNMAPPED_AREA_TOPDOWN;
> -		info.low_limit = PAGE_SIZE;
> +		info.low_limit = max(PAGE_SIZE, PAGE_ALIGN(mmap_min_addr));
>  		info.high_limit = mm->mmap_base;
>  		addr = vm_unmapped_area(&info);

> -		info.low_limit = addr;
> +		info.low_limit = max(addr, PAGE_ALIGN(mmap_min_addr));

>  	info.flags = VM_UNMAPPED_AREA_TOPDOWN;
>  	info.length = len;
> -	info.low_limit = PAGE_SIZE;
> +	info.low_limit = max(PAGE_SIZE, PAGE_ALIGN(mmap_min_addr));
>  	info.high_limit = mm->mmap_base;
>  	info.align_mask = do_colour_align ? (PAGE_MASK & shm_align_mask) : 0;
>  	info.align_offset = pgoff << PAGE_SHIFT;

>  	info.flags = VM_UNMAPPED_AREA_TOPDOWN;
>  	info.length = len;
> -	info.low_limit = PAGE_SIZE;
> +	info.low_limit = max(PAGE_SIZE, PAGE_ALIGN(mmap_min_addr));
>  	info.high_limit = mm->mmap_base;
>  	info.align_mask = do_color_align ? (PAGE_MASK & (SHMLBA - 1)) : 0;
>  	info.align_offset = pgoff << PAGE_SHIFT;

>  	info.flags = VM_UNMAPPED_AREA_TOPDOWN;
>  	info.length = len;
> -	info.low_limit = PAGE_SIZE;
> +	info.low_limit = max(PAGE_SIZE, PAGE_ALIGN(mmap_min_addr));
>  	info.high_limit = mm->mmap_base;
>  	info.align_mask = filp ? get_align_mask() : 0;
>  	info.align_offset = pgoff << PAGE_SHIFT;

There appears to be a lot of repetition in these methods - instead of 
changing 6 places it would be more future-proof to first factor out the 
common bits and then to apply the fix to the shared implementation.

Thanks,

	Ingo


[Index of Archives]     [Linux MIPS Home]     [LKML Archive]     [Linux ARM Kernel]     [Linux ARM]     [Linux]     [Git]     [Yosemite News]     [Linux SCSI]     [Linux Hams]

  Powered by Linux