Re: [PATCH v10 11/12] mm/vmalloc: Hugepage vmalloc mappings

From: Christophe Leroy <christophe.leroy@csgroup.eu>
To: Nicholas Piggin <npiggin@gmail.com>,
	linux-mm@kvack.org, Andrew Morton <akpm@linux-foundation.org>
Cc: linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org,
	linuxppc-dev@lists.ozlabs.org, Zefan Li <lizefan@huawei.com>,
	Jonathan Cameron <Jonathan.Cameron@Huawei.com>,
	Christoph Hellwig <hch@infradead.org>,
	Rick Edgecombe <rick.p.edgecombe@intel.com>,
	Ding Tianhong <dingtianhong@huawei.com>
Subject: Re: [PATCH v10 11/12] mm/vmalloc: Hugepage vmalloc mappings
Date: Mon, 25 Jan 2021 10:14:54 +0100	[thread overview]
Message-ID: <933352bd-dcf3-c483-4d7a-07afe1116cf1@csgroup.eu> (raw)
In-Reply-To: <20210124082230.2118861-12-npiggin@gmail.com>

Le 24/01/2021 à 09:22, Nicholas Piggin a écrit :
> Support huge page vmalloc mappings. Config option HAVE_ARCH_HUGE_VMALLOC
> enables support on architectures that define HAVE_ARCH_HUGE_VMAP and
> supports PMD sized vmap mappings.
> 
> vmalloc will attempt to allocate PMD-sized pages if allocating PMD size
> or larger, and fall back to small pages if that was unsuccessful.
> 
> Architectures must ensure that any arch specific vmalloc allocations
> that require PAGE_SIZE mappings (e.g., module allocations vs strict
> module rwx) use the VM_NOHUGE flag to inhibit larger mappings.
> 
> When hugepage vmalloc mappings are enabled in the next patch, this
> reduces TLB misses by nearly 30x on a `git diff` workload on a 2-node
> POWER9 (59,800 -> 2,100) and reduces CPU cycles by 0.54%.
> 
> This can result in more internal fragmentation and memory overhead for a
> given allocation, an option nohugevmalloc is added to disable at boot.
> 
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
>   arch/Kconfig            |  10 +++
>   include/linux/vmalloc.h |  18 ++++
>   mm/page_alloc.c         |   5 +-
>   mm/vmalloc.c            | 192 ++++++++++++++++++++++++++++++----------
>   4 files changed, 177 insertions(+), 48 deletions(-)
> 

> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 0377e1d059e5..eef61e0f5170 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c

> @@ -2691,15 +2746,18 @@ EXPORT_SYMBOL_GPL(vmap_pfn);
>   #endif /* CONFIG_VMAP_PFN */
>   
>   static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
> -				 pgprot_t prot, int node)
> +				 pgprot_t prot, unsigned int page_shift,
> +				 int node)
>   {
>   	const gfp_t nested_gfp = (gfp_mask & GFP_RECLAIM_MASK) | __GFP_ZERO;
> -	unsigned int nr_pages = get_vm_area_size(area) >> PAGE_SHIFT;
> -	unsigned long array_size;
> -	unsigned int i;
> +	unsigned int page_order = page_shift - PAGE_SHIFT;
> +	unsigned long addr = (unsigned long)area->addr;
> +	unsigned long size = get_vm_area_size(area);
> +	unsigned int nr_small_pages = size >> PAGE_SHIFT;
>   	struct page **pages;
> +	unsigned int i;
>   
> -	array_size = (unsigned long)nr_pages * sizeof(struct page *);
> +	array_size = (unsigned long)nr_small_pages * sizeof(struct page *);

array_size() is a function in include/linux/overflow.h

For some reason, it breaks the build with your series.

>   	gfp_mask |= __GFP_NOWARN;
>   	if (!(gfp_mask & (GFP_DMA | GFP_DMA32)))
>   		gfp_mask |= __GFP_HIGHMEM;