linux-kernel - Re: [PATCH] mm: implement "memory.oops_if_bad

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: <df9c25ac-c290-4c65-ac70-729cddf2aee4@suse.cz>
Date: Thu, 10 Jul 2025 09:35:41 +0200
From: Vlastimil Babka <vbabka@...e.cz>
To: Alexey Dobriyan <adobriyan@...il.com>, akpm@...ux-foundation.org
Cc: linux-kernel@...r.kernel.org, linux-mm@...ck.org,
 David Hildenbrand <david@...hat.com>,
 Lorenzo Stoakes <lorenzo.stoakes@...cle.com>,
 "Liam R. Howlett" <Liam.Howlett@...cle.com>, Mike Rapoport
 <rppt@...nel.org>, Suren Baghdasaryan <surenb@...gle.com>,
 Michal Hocko <mhocko@...e.com>
Subject: Re: [PATCH] mm: implement "memory.oops_if_bad_pte=1" boot option

On 7/9/25 20:10, Alexey Dobriyan wrote:
> Implement
> 
> 	memory.oops_if_bad_pte=1
> 
> boot option which oopses the machine instead of dreadful
> 
> 	BUG: Bad page map in process
> 
> message.
> 
> This is intended
> for people who want to panic at the slightest provocation and
> for people who ruled out hardware problems which in turn means that
> delaying vmcore collection is counter-productive.
> 
> Linux doesn't (never?) panicked on PTE corruption and even implemented
> ratelimited version of the message meaning it can go for minutes and
> even hours without anyone noticing which is exactly the opposite of what
> should be done to facilitate debugging.
> 
> Not enabled by default.
> 
> Not advertised.
> 
> Signed-off-by: Alexey Dobriyan <adobriyan@...il.com>

Could we just reuse the existing panic_on_oops? Would anyone want to panic
in this particular without the others, or vice versa?

> ---
> 
>  mm/memory.c                                 | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 
> diff --git a/mm/memory.c b/mm/memory.c
> index b0cda5aab398..90b92b312802 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -42,6 +42,7 @@
>  #include <linux/kernel_stat.h>
>  #include <linux/mm.h>
>  #include <linux/mm_inline.h>
> +#include <linux/moduleparam.h>
>  #include <linux/sched/mm.h>
>  #include <linux/sched/numa_balancing.h>
>  #include <linux/sched/task.h>
> @@ -480,6 +481,13 @@ static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
>  			add_mm_counter(mm, i, rss[i]);
>  }
>  
> +/*
> + * Oops instead of printing "Bad page map in process" message and
> + * trying to continue.
> + */
> +static bool oops_if_bad_pte __ro_after_init = false;
> +module_param(oops_if_bad_pte, bool, 0444);
> +
>  /*
>   * This function is called to print an error when a bad pte
>   * is found. For example, we might have a PFN-mapped pte in
> @@ -490,6 +498,13 @@ static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
>  static void print_bad_pte(struct vm_area_struct *vma, unsigned long addr,
>  			  pte_t pte, struct page *page)
>  {
> +	/*
> +	 * This line is a formality to collect vmcore ASAP. Real bug
> +	 * (hardware or software) happened earlier, current registers and
> +	 * backtrace aren't interesting.
> +	 */
> +	BUG_ON(oops_if_bad_pte);
> +
>  	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
>  	p4d_t *p4d = p4d_offset(pgd, addr);
>  	pud_t *pud = pud_offset(p4d, addr);