linux-kernel - Re: [PATCH 2/4] thp: fix MADV_DONTNEED vs. numa balancing race

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: <f105f6a5-bb5e-9480-6b2e-d2d15f631af9@suse.cz>
Date:   Wed, 12 Apr 2017 15:33:35 +0200
From:   Vlastimil Babka <vbabka@...e.cz>
To:     "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
        Andrea Arcangeli <aarcange@...hat.com>,
        Andrew Morton <akpm@...ux-foundation.org>
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/4] thp: fix MADV_DONTNEED vs. numa balancing race

On 03/02/2017 04:10 PM, Kirill A. Shutemov wrote:
> In case prot_numa, we are under down_read(mmap_sem). It's critical
> to not clear pmd intermittently to avoid race with MADV_DONTNEED
> which is also under down_read(mmap_sem):
> 
> 	CPU0:				CPU1:
> 				change_huge_pmd(prot_numa=1)
> 				 pmdp_huge_get_and_clear_notify()
> madvise_dontneed()
>  zap_pmd_range()
>   pmd_trans_huge(*pmd) == 0 (without ptl)
>   // skip the pmd
> 				 set_pmd_at();
> 				 // pmd is re-established
> 
> The race makes MADV_DONTNEED miss the huge pmd and don't clear it
> which may break userspace.
> 
> Found by code analysis, never saw triggered.
> 
> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@...ux.intel.com>
> ---
>  mm/huge_memory.c | 34 +++++++++++++++++++++++++++++++++-
>  1 file changed, 33 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index e7ce73b2b208..bb2b3646bd78 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -1744,7 +1744,39 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
>  	if (prot_numa && pmd_protnone(*pmd))
>  		goto unlock;
>  
> -	entry = pmdp_huge_get_and_clear_notify(mm, addr, pmd);
> +	/*
> +	 * In case prot_numa, we are under down_read(mmap_sem). It's critical
> +	 * to not clear pmd intermittently to avoid race with MADV_DONTNEED
> +	 * which is also under down_read(mmap_sem):
> +	 *
> +	 *	CPU0:				CPU1:
> +	 *				change_huge_pmd(prot_numa=1)
> +	 *				 pmdp_huge_get_and_clear_notify()
> +	 * madvise_dontneed()
> +	 *  zap_pmd_range()
> +	 *   pmd_trans_huge(*pmd) == 0 (without ptl)
> +	 *   // skip the pmd
> +	 *				 set_pmd_at();
> +	 *				 // pmd is re-established
> +	 *
> +	 * The race makes MADV_DONTNEED miss the huge pmd and don't clear it
> +	 * which may break userspace.
> +	 *
> +	 * pmdp_invalidate() is required to make sure we don't miss
> +	 * dirty/young flags set by hardware.
> +	 */
> +	entry = *pmd;
> +	pmdp_invalidate(vma, addr, pmd);
> +
> +	/*
> +	 * Recover dirty/young flags.  It relies on pmdp_invalidate to not
> +	 * corrupt them.
> +	 */

pmdp_invalidate() does:

        pmd_t entry = *pmdp;
        set_pmd_at(vma->vm_mm, address, pmdp, pmd_mknotpresent(entry));

so it's not atomic and if CPU sets dirty or accessed in the middle of
this, they will be lost?

But I don't see how the other invalidate caller
__split_huge_pmd_locked() deals with this either. Andrea, any idea?

Vlastimil

> +	if (pmd_dirty(*pmd))
> +		entry = pmd_mkdirty(entry);
> +	if (pmd_young(*pmd))
> +		entry = pmd_mkyoung(entry);
> +
>  	entry = pmd_modify(entry, newprot);
>  	if (preserve_write)
>  		entry = pmd_mk_savedwrite(entry);
>