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]
Message-ID: <aRXDbk9ajOhH7ReX@kernel.org>
Date: Thu, 13 Nov 2025 13:39:26 +0200
From: Mike Rapoport <rppt@...nel.org>
To: Mehdi Ben Hadj Khelifa <mehdi.benhadjkhelifa@...il.com>
Cc: akpm@...ux-foundation.org, peterx@...hat.com, david@...hat.com,
	lorenzo.stoakes@...cle.com, Liam.Howlett@...cle.com, vbabka@...e.cz,
	surenb@...gle.com, mhocko@...e.com, shuah@...nel.org,
	linux-mm@...ck.org, linux-kselftest@...r.kernel.org,
	linux-kernel@...r.kernel.org, skhan@...uxfoundation.org,
	david.hunter.linux@...il.com, khalid@...nel.org,
	linux-kernel-mentees@...ts.linuxfoundation.org
Subject: Re: [PATCH] selftests/mm/uffd: remove static address usage in
 shmem_allocate_area()

On Tue, Nov 11, 2025 at 09:54:27PM +0100, Mehdi Ben Hadj Khelifa wrote:
> The current shmem_allocate_area() implementation uses a hardcoded virtual
> base address(BASE_PMD_ADDR) as a hint for mmap() when creating shmem-backed
> test areas. This approach is fragile and may fail on systems with ASLR or
> different virtual memory layouts, where the chosen address is unavailable.
> 
> Replace the static base address with a dynamically reserved address range
> obtained via mmap(NULL, ..., PROT_NONE). The memfd-backed areas and their
> alias are then mapped into that reserved region using MAP_FIXED, preserving
> the original layout and aliasing semantics while avoiding collisions with
> unrelated mappings.
> 
> This change improves robustness and portability of the test suite without
> altering its behavior or coverage.
> 
> Signed-off-by: Mehdi Ben Hadj Khelifa <mehdi.benhadjkhelifa@...il.com>
> ---
> Testing:
> A diff between running the mm selftests on 6.18-rc5 from before and after
> the change show no regression on x86_64 architecture with 32GB DDR5 RAM.
>  tools/testing/selftests/mm/uffd-common.c | 25 +++++++++++++++---------
>  1 file changed, 16 insertions(+), 9 deletions(-)
> 
> diff --git a/tools/testing/selftests/mm/uffd-common.c b/tools/testing/selftests/mm/uffd-common.c
> index 994fe8c03923..492b21c960bb 100644
> --- a/tools/testing/selftests/mm/uffd-common.c
> +++ b/tools/testing/selftests/mm/uffd-common.c
> @@ -6,11 +6,11 @@
>   */
>  
>  #include "uffd-common.h"
> +#include "asm-generic/mman-common.h"

Please drop this.
There's already include <sys/mman.h> via uffd-common.h/vm_util.h.

>  
>  uffd_test_ops_t *uffd_test_ops;
>  uffd_test_case_ops_t *uffd_test_case_ops;
>  
> -#define BASE_PMD_ADDR ((void *)(1UL << 30))
>  
>  /* pthread_mutex_t starts at page offset 0 */
>  pthread_mutex_t *area_mutex(char *area, unsigned long nr, uffd_global_test_opts_t *gopts)
> @@ -142,30 +142,37 @@ static int shmem_allocate_area(uffd_global_test_opts_t *gopts, void **alloc_area
>  	unsigned long offset = is_src ? 0 : bytes;
>  	char *p = NULL, *p_alias = NULL;
>  	int mem_fd = uffd_mem_fd_create(bytes * 2, false);
> +	size_t region_size = bytes * 2 + hpage_size;
>  
> -	/* TODO: clean this up.  Use a static addr is ugly */
> -	p = BASE_PMD_ADDR;
> -	if (!is_src)
> -		/* src map + alias + interleaved hpages */
> -		p += 2 * (bytes + hpage_size);
> +	void *reserve = mmap(NULL, region_size, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS,
> +			-1, 0);
> +	if (reserve == MAP_FAILED) {
> +		close(mem_fd);
> +		return -errno;
> +	}
> +
> +	p = (char *)reserve;

No need for casting here.

>  	p_alias = p;
>  	p_alias += bytes;
>  	p_alias += hpage_size;  /* Prevent src/dst VMA merge */
>  
> -	*alloc_area = mmap(p, bytes, PROT_READ | PROT_WRITE, MAP_SHARED,
> +	*alloc_area = mmap(p, bytes, PROT_READ | PROT_WRITE, MAP_FIXED | MAP_SHARED,
>  			   mem_fd, offset);
>  	if (*alloc_area == MAP_FAILED) {
> +		munmap(reserve, region_size);

I think it'll be more readable to put munmap() after setting *alloc_area to
NULL.

>  		*alloc_area = NULL;
> +		close(mem_fd);
>  		return -errno;
>  	}
>  	if (*alloc_area != p)
>  		err("mmap of memfd failed at %p", p);
>  
> -	area_alias = mmap(p_alias, bytes, PROT_READ | PROT_WRITE, MAP_SHARED,
> +	area_alias = mmap(p_alias, bytes, PROT_READ | PROT_WRITE, MAP_FIXED | MAP_SHARED,
>  			  mem_fd, offset);
>  	if (area_alias == MAP_FAILED) {
> -		munmap(*alloc_area, bytes);
> +		munmap(reserve, region_size);

Here as well.

>  		*alloc_area = NULL;
> +		close(mem_fd);
>  		return -errno;
>  	}
>  	if (area_alias != p_alias)
> -- 
> 2.51.2
> 

-- 
Sincerely yours,
Mike.

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ