linux-kernel - Re: [PATCH v3] /dev/zero: try to align PMD

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] [day] [month] [year] [list]

Message-ID: <5dd6063a-4b67-4345-bd89-32085bb5923f@lucifer.local>
Date: Thu, 31 Jul 2025 11:20:52 +0100
From: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
To: Zhang Qilong <zhangqilong3@...wei.com>
Cc: arnd@...db.de, gregkh@...uxfoundation.org, akpm@...ux-foundation.org,
        linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        wangkefeng.wang@...wei.com, sunnanyong@...wei.com,
        David Hildenbrand <david@...hat.com>
Subject: Re: [PATCH v3] /dev/zero: try to align PMD_SIZE for private mapping

On Thu, Jul 31, 2025 at 04:36:55PM +0800, Zhang Qilong wrote:
> Attempt to map aligned to huge page size for private mapping which
> could achieve performance gains, the mprot_tw4m in libMicro average
> execution time on arm64:
>   - Test case:        mprot_tw4m
>   - Before the patch:   22 us
>   - After the patch:    17 us
>
> If THP config is not set, we fall back to system page size mappings.
>
> Acked-by: David Hildenbrand <david@...hat.com>
> Signed-off-by: Zhang Qilong <zhangqilong3@...wei.com>

LGTM, so:

Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>

I did some rudimentary testing on this with a simple MAP_PRIVATE thing
which caused no issues and I observed a 20 MB mapping getting aligned and
thus benefitting from 10 x huge PMD mappings vs. mainline benefitting from
9 due to misalignment.

So feel free to add:

Tested-by: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>

Cheers, Lorenzo

> ---
> v3:
> - collect Acked-by
> - factor out the #ifdef CONFIG_MMU in get_unmapped_area_zero(), per Lorenzo
> - explicitly use #ifdef CONFIG_TRANSPARENT_HUGEPAGE, per Lorenzo and Matthew
>
> v2:
> - add comments on code suggested by Lorenzo
> - use IS_ENABLED to check THP config
>
>  drivers/char/mem.c | 21 +++++++++++++++++----
>  1 file changed, 17 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/char/mem.c b/drivers/char/mem.c
> index 48839958b0b1..b7c4dbe4d2c1 100644
> --- a/drivers/char/mem.c
> +++ b/drivers/char/mem.c
> @@ -510,31 +510,44 @@ static int mmap_zero(struct file *file, struct vm_area_struct *vma)
>  		return shmem_zero_setup(vma);
>  	vma_set_anonymous(vma);
>  	return 0;
>  }
>
> +#ifndef CONFIG_MMU

OK, the convention in this file is to invert this check so this is fine.

> +static unsigned long get_unmapped_area_zero(struct file *file,
> +				unsigned long addr, unsigned long len,
> +				unsigned long pgoff, unsigned long flags)
> +{
> +	return -ENOSYS;
> +}
> +#else
>  static unsigned long get_unmapped_area_zero(struct file *file,
>  				unsigned long addr, unsigned long len,
>  				unsigned long pgoff, unsigned long flags)
>  {
> -#ifdef CONFIG_MMU
>  	if (flags & MAP_SHARED) {
>  		/*
>  		 * mmap_zero() will call shmem_zero_setup() to create a file,
>  		 * so use shmem's get_unmapped_area in case it can be huge;
>  		 * and pass NULL for file as in mmap.c's get_unmapped_area(),
>  		 * so as not to confuse shmem with our handle on "/dev/zero".
>  		 */
>  		return shmem_get_unmapped_area(NULL, addr, len, pgoff, flags);
>  	}
>
> -	/* Otherwise flags & MAP_PRIVATE: with no shmem object beneath it */
> -	return mm_get_unmapped_area(current->mm, file, addr, len, pgoff, flags);
> +	/*
> +	 * Otherwise flags & MAP_PRIVATE: with no shmem object beneath it,
> +	 * attempt to map aligned to huge page size if possible, otherwise we
> +	 * fall back to system page size mappings.
> +	 */
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +	return thp_get_unmapped_area(file, addr, len, pgoff, flags);
>  #else
> -	return -ENOSYS;
> +	return mm_get_unmapped_area(current->mm, file, addr, len, pgoff, flags);
>  #endif
>  }
> +#endif

Nit, but can we add a /* CONFIG_MMU */ here please since we have a bunch of
ifdef's and it's noisy.

>
>  static ssize_t write_full(struct file *file, const char __user *buf,
>  			  size_t count, loff_t *ppos)
>  {
>  	return -ENOSPC;
> --
> 2.43.0
>