Re: Regression from 2.6.36

[David Rientjes <rientjes@google.com>]

On Wed, 13 Apr 2011, Andrew Morton wrote:

> Azurit reports large increases in system time after 2.6.36 when running
> Apache.  It was bisected down to a892e2d7dcdfa6c76e6 ("vfs: use kmalloc()
> to allocate fdmem if possible").
> 
> That patch caused the vfs to use kmalloc() for very large allocations and
> this is causing excessive work (and presumably excessive reclaim) within
> the page allocator.
> 
> Fix it by falling back to vmalloc() earlier - when the allocation attempt
> would have been considered "costly" by reclaim.
> 
> Reported-by: azurIt <azurit@pobox.sk>
> Cc: Changli Gao <xiaosuo@gmail.com>
> Cc: Americo Wang <xiyou.wangcong@gmail.com>
> Cc: Jiri Slaby <jslaby@suse.cz>
> Cc: Eric Dumazet <eric.dumazet@gmail.com>
> Cc: Mel Gorman <mel@csn.ul.ie>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
>  fs/file.c |   17 ++++++++++-------
>  1 file changed, 10 insertions(+), 7 deletions(-)
> 
> diff -puN fs/file.c~a fs/file.c
> --- a/fs/file.c~a
> +++ a/fs/file.c
> @@ -39,14 +39,17 @@ int sysctl_nr_open_max = 1024 * 1024; /*
>   */
>  static DEFINE_PER_CPU(struct fdtable_defer, fdtable_defer_list);
>  
> -static inline void *alloc_fdmem(unsigned int size)
> +static void *alloc_fdmem(unsigned int size)
>  {
> -	void *data;
> -
> -	data = kmalloc(size, GFP_KERNEL|__GFP_NOWARN);
> -	if (data != NULL)
> -		return data;
> -
> +	/*
> +	 * Very large allocations can stress page reclaim, so fall back to
> +	 * vmalloc() if the allocation size will be considered "large" by the VM.
> +	 */
> +	if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER) {
> +		void *data = kmalloc(size, GFP_KERNEL|__GFP_NOWARN);
> +		if (data != NULL)
> +			return data;
> +	}
>  	return vmalloc(size);
>  }
>  

It's a shame that we can't at least try kmalloc() with sufficiently large 
sizes by doing something like

	gfp_t flags = GFP_NOWAIT | __GFP_NOWARN;

	if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
		flags |= GFP_KERNEL;
	data = kmalloc(size, flags);
	if (data)
		return data;
	return vmalloc(size);

which would at least attempt to use the slab allocator.