linux-kernel - Re: [PATCH] mm: centralize and fix max map count limit checking

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <aLk_zbXxYxHs1Igk@kernel.org>
Date: Thu, 4 Sep 2025 10:29:17 +0300
From: Mike Rapoport <rppt@...nel.org>
To: Kalesh Singh <kaleshsingh@...gle.com>
Cc: akpm@...ux-foundation.org, minchan@...nel.org,
	lorenzo.stoakes@...cle.com, kernel-team@...roid.com,
	android-mm@...gle.com, David Hildenbrand <david@...hat.com>,
	"Liam R. Howlett" <Liam.Howlett@...cle.com>,
	Vlastimil Babka <vbabka@...e.cz>,
	Suren Baghdasaryan <surenb@...gle.com>,
	Michal Hocko <mhocko@...e.com>, Jann Horn <jannh@...gle.com>,
	Pedro Falcato <pfalcato@...e.de>, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm: centralize and fix max map count limit checking

On Wed, Sep 03, 2025 at 04:24:35PM -0700, Kalesh Singh wrote:
> The check against the max map count (sysctl_max_map_count) was
> open-coded in several places. This led to inconsistent enforcement
> and subtle bugs where the limit could be exceeded.
> 
> For example, some paths would check map_count > sysctl_max_map_count
> before allocating a new VMA and incrementing the count, allowing the
> process to reach sysctl_max_map_count + 1:
> 
>     int do_brk_flags(...)
>     {
>         if (mm->map_count > sysctl_max_map_count)
>             return -ENOMEM;
> 
>         /* We can get here with mm->map_count == sysctl_max_map_count */
> 
>         vma = vm_area_alloc(mm);
>         ...
>         mm->map_count++   /* We've now exceeded the threshold. */
>     }
> 
> To fix this and unify the logic, introduce a new function,
> exceeds_max_map_count(), to consolidate the check. All open-coded
> checks are replaced with calls to this new function, ensuring the
> limit is applied uniformly and correctly.
> 
> To improve encapsulation, sysctl_max_map_count is now static to
> mm/mmap.c. The new helper also adds a rate-limited warning to make
> debugging applications that exhaust their VMA limit easier.
> 
> Cc: Andrew Morton <akpm@...ux-foundation.org>
> Cc: Minchan Kim <minchan@...nel.org>
> Cc: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
> Signed-off-by: Kalesh Singh <kaleshsingh@...gle.com>
> ---
>  include/linux/mm.h | 11 ++++++++++-
>  mm/mmap.c          | 15 ++++++++++++++-
>  mm/mremap.c        |  7 ++++---
>  mm/nommu.c         |  2 +-
>  mm/util.c          |  1 -
>  mm/vma.c           |  6 +++---
>  6 files changed, 32 insertions(+), 10 deletions(-)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 1ae97a0b8ec7..d4e64e6a9814 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -192,7 +192,16 @@ static inline void __mm_zero_struct_page(struct page *page)
>  #define MAPCOUNT_ELF_CORE_MARGIN	(5)
>  #define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
>  
> -extern int sysctl_max_map_count;
> +/**
> + * exceeds_max_map_count - check if a VMA operation would exceed max_map_count
> + * @mm: The memory descriptor for the process.
> + * @new_vmas: The number of new VMAs the operation will create.
> + *
> + * Returns true if the operation would cause the number of VMAs to exceed
> + * the sysctl_max_map_count limit, false otherwise. A rate-limited warning
> + * is logged if the limit is exceeded.

kernel-doc will be unhappy because of missing Return: description. Please
s/Returns/Return:/

> + */
> +extern bool exceeds_max_map_count(struct mm_struct *mm, unsigned int new_vmas);

No need for 'extern' here. And the declaration should go to mm/internal.h

>  extern unsigned long sysctl_user_reserve_kbytes;
>  extern unsigned long sysctl_admin_reserve_kbytes;

-- 
Sincerely yours,
Mike.