Re: [PATCH] nommu: remove __GFP_HIGHMEM in vmalloc/vzalloc

[Matthew Wilcox <willy@infradead.org>]

On Mon, May 03, 2021 at 08:44:51PM +0800, Chen Li wrote:
> 
> From mm/nommu.c:
> void *__vmalloc(unsigned long size, gfp_t gfp_mask)
> {
> 	/*
> 	 *  You can't specify __GFP_HIGHMEM with kmalloc() since kmalloc()
> 	 * returns only a logical address.
> 	 */
> 	return kmalloc(size, (gfp_mask | __GFP_COMP) & ~__GFP_HIGHMEM);
> }
> 
> nommu's __vmalloc just uses kmalloc internally and elimitates __GFP_HIGHMEM,
> so it makes no sense to add __GFP_HIGHMEM for nommu's vmalloc/vzalloc.

I think this was originally [1] a copy of vmalloc() from vmalloc.c,
but of course the two have drifted apart over time.  At the time,
vmalloc.c's vmalloc() looked like this:

+void * vmalloc (unsigned long size)
+{
+       return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM, PAGE_KERNEL);
+}

[1] And I do mean originally; this was present in Alan Cox's original
submission of mm/nommu.c in 2002.

This patch makes sense to me, although I don't imagine it makes
much difference.

Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>

> Signed-off-by: Chen Li <chenli@uniontech.com>
> ---
>  mm/nommu.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/nommu.c b/mm/nommu.c
> index 5c9ab799c0e6..339a2f2eb1aa 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -233,7 +233,7 @@ long vwrite(char *buf, char *addr, unsigned long count)
>   */
>  void *vmalloc(unsigned long size)
>  {
> -       return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM);
> +       return __vmalloc(size, GFP_KERNEL);
>  }
>  EXPORT_SYMBOL(vmalloc);
>  
> @@ -251,7 +251,7 @@ EXPORT_SYMBOL(vmalloc);
>   */
>  void *vzalloc(unsigned long size)
>  {
> -	return __vmalloc(size, GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO);
> +	return __vmalloc(size, GFP_KERNEL | __GFP_ZERO);
>  }
>  EXPORT_SYMBOL(vzalloc);
>  
> -- 
> 2.31.1
> 
> 
> 
>