Date:   Fri, 1 Oct 2021 15:33:30 -0700
From:   Mike Kravetz <mike.kravetz@...cle.com>
To:     Zhenguo Yao <yaozhenguo1@...il.com>
Cc:     mpe@...erman.id.au, benh@...nel.crashing.org, paulus@...ba.org,
        corbet@....net, rppt@...nel.org, akpm@...ux-foundation.org,
        yaozhenguo@...com, willy@...radead.org,
        linux-kernel@...r.kernel.org, linux-doc@...r.kernel.org,
        linux-mm@...ck.org
Subject: Re: [PATCH v7] hugetlbfs: Extend the definition of hugepages
 parameter to support node allocation

On 9/27/21 3:41 AM, Zhenguo Yao wrote:
> We can specify the number of hugepages to allocate at boot. But the
> hugepages is balanced in all nodes at present. In some scenarios,
> we only need hugepages in one node. For example: DPDK needs hugepages
> which are in the same node as NIC. if DPDK needs four hugepages of 1G
> size in node1 and system has 16 numa nodes. We must reserve 64 hugepages
> in kernel cmdline. But, only four hugepages are used. The others should
> be free after boot. If the system memory is low(for example: 64G), it will
> be an impossible task. So, Extending hugepages parameter to support
> specifying hugepages at a specific node.
> For example add following parameter:
> 
> hugepagesz=1G hugepages=0:1,1:3
> 
> It will allocate 1 hugepage in node0 and 3 hugepages in node1.
> 
> Signed-off-by: Zhenguo Yao <yaozhenguo1@...il.com>

Thanks for your continued efforts!  And, thank you Mike R. for your
input.

Just a few very minor comments below.  Everything else looks good.
With suggested updates,

Reviewed-by: Mike Kravetz <mike.kravetz@...cle.com>

> ---
> v6->v7 changes:
> 	- replace nodes_weight(node_states[N_MEMORY] with nr_online_nodes.
> v5->v6 changes:
> 	- Remove v5 codes: using return value to disable node specific alloc.
> 	- Add node_specific_alloc_support weak function to disable node
> 	  specific alloc when arch can't support it.
> 	- Remove useless variable addr in alloc_bootmem_huge_page.
> 	- Add powerpc version of node_specific_alloc_support when
> 	  CONFIG_PPC_BOOK3S_64 is defined.
> v4->v5 changes:
> 	- remove BUG_ON in __alloc_bootmem_huge_page.
> 	- add nid parameter in __alloc_bootmem_huge_page to support
> 	  called in both specific node alloc and normal alloc.
> 	- do normal alloc if architecture can't support node specific alloc.
> 	- return -1 in powerpc version of alloc_bootmem_huge_page when
> 	  nid is not NUMA_NO_NODE.
> v3->v4 changes:
> 	- fix wrong behavior for parameter:
> 	  hugepages=0:1,1:3 default_hugepagesz=1G
> 	- make the change of documentation more reasonable.
> v2->v3 changes:
> 	- Skip gigantic hugepages allocation if hugetlb_cma is enabled.
> 	- Fix wrong behavior for parameter:
> 	  hugepagesz=2M hugepages=2 hugepages=5.
> 	- Update hugetlbpage.rst.
> 	- Fix side effects which v2 brings in.
> 	- add cond_resched in hugetlb_hstate_alloc_pages_onenode.
> v1->v2 changes:
> 	- add checking for max node to avoid array out of bounds.
> 	- fix wrong max_huge_pages after failed allocation.
> 	- fix wrong behavior when parsing parameter: hugepages=0:1,2,3:4.
> 	- return 0 when parsing invalid parameter in hugepages_setup
> ---
>  .../admin-guide/kernel-parameters.txt         |   8 +-
>  Documentation/admin-guide/mm/hugetlbpage.rst  |  12 +-
>  arch/powerpc/mm/hugetlbpage.c                 |   9 +-
>  include/linux/hugetlb.h                       |   6 +-
>  mm/hugetlb.c                                  | 153 +++++++++++++++---
>  5 files changed, 157 insertions(+), 31 deletions(-)
> 
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index 91ba391f9b32..9b3d8791586d 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -1599,9 +1599,11 @@
>  			the number of pages of hugepagesz to be allocated.
>  			If this is the first HugeTLB parameter on the command
>  			line, it specifies the number of pages to allocate for
> -			the default huge page size.  See also
> -			Documentation/admin-guide/mm/hugetlbpage.rst.
> -			Format: <integer>
> +			the default huge page size. If using node format, the
> +			number of pages to allocate per-node can be specified.
> +			See also Documentation/admin-guide/mm/hugetlbpage.rst.
> +			Format: <integer> or (node format)
> +				<node>:<integer>[,<node>:<integer>]
>  
>  	hugepagesz=
>  			[HW] The size of the HugeTLB pages.  This is used in
> diff --git a/Documentation/admin-guide/mm/hugetlbpage.rst b/Documentation/admin-guide/mm/hugetlbpage.rst
> index 8abaeb144e44..d70828c07658 100644
> --- a/Documentation/admin-guide/mm/hugetlbpage.rst
> +++ b/Documentation/admin-guide/mm/hugetlbpage.rst
> @@ -128,7 +128,9 @@ hugepages
>  	implicitly specifies the number of huge pages of default size to
>  	allocate.  If the number of huge pages of default size is implicitly
>  	specified, it can not be overwritten by a hugepagesz,hugepages
> -	parameter pair for the default size.
> +	parameter pair for the default size.  This parameter also has a
> +	node format.  The node format specifies the number of huge pages
> +	to allocate on specific nodes.
>  
>  	For example, on an architecture with 2M default huge page size::
>  
> @@ -138,6 +140,14 @@ hugepages
>  	indicating that the hugepages=512 parameter is ignored.  If a hugepages
>  	parameter is preceded by an invalid hugepagesz parameter, it will
>  	be ignored.
> +
> +	Node format example::
> +
> +		hugepagesz=2M hugepages=0:1,1:2
> +
> +	It will allocate 1 2M hugepage on node0 and 2 2M hugepages on node1.
> +	If the node number is invalid,  the parameter will be ignored.
> +
>  default_hugepagesz
>  	Specify the default huge page size.  This parameter can
>  	only be specified once on the command line.  default_hugepagesz can
> diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
> index 9a75ba078e1b..dd40ce6e7565 100644
> --- a/arch/powerpc/mm/hugetlbpage.c
> +++ b/arch/powerpc/mm/hugetlbpage.c
> @@ -229,17 +229,22 @@ static int __init pseries_alloc_bootmem_huge_page(struct hstate *hstate)
>  	m->hstate = hstate;
>  	return 1;
>  }
> +
> +bool __init node_specific_alloc_support(void)
> +{
> +	return false;
> +}
>  #endif
>  
>  
> -int __init alloc_bootmem_huge_page(struct hstate *h)
> +int __init alloc_bootmem_huge_page(struct hstate *h, int nid)
>  {
>  
>  #ifdef CONFIG_PPC_BOOK3S_64
>  	if (firmware_has_feature(FW_FEATURE_LPAR) && !radix_enabled())
>  		return pseries_alloc_bootmem_huge_page(h);
>  #endif
> -	return __alloc_bootmem_huge_page(h);
> +	return __alloc_bootmem_huge_page(h, nid);
>  }
>  
>  #ifndef CONFIG_PPC_BOOK3S_64
> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> index 1faebe1cd0ed..3504e407567c 100644
> --- a/include/linux/hugetlb.h
> +++ b/include/linux/hugetlb.h
> @@ -605,6 +605,7 @@ struct hstate {
>  	unsigned long nr_overcommit_huge_pages;
>  	struct list_head hugepage_activelist;
>  	struct list_head hugepage_freelists[MAX_NUMNODES];
> +	unsigned int max_huge_pages_node[MAX_NUMNODES];
>  	unsigned int nr_huge_pages_node[MAX_NUMNODES];
>  	unsigned int free_huge_pages_node[MAX_NUMNODES];
>  	unsigned int surplus_huge_pages_node[MAX_NUMNODES];
> @@ -637,8 +638,9 @@ void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
>  				unsigned long address, struct page *page);
>  
>  /* arch callback */
> -int __init __alloc_bootmem_huge_page(struct hstate *h);
> -int __init alloc_bootmem_huge_page(struct hstate *h);
> +int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
> +int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
> +bool __init node_specific_alloc_support(void);

Agree with Mike R. that hugetlb should be in the name.

>  
>  void __init hugetlb_add_hstate(unsigned order);
>  bool __init arch_hugetlb_valid_size(unsigned long size);
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 95dc7b83381f..ca00676a1bdd 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -66,6 +66,7 @@ static struct hstate * __initdata parsed_hstate;
>  static unsigned long __initdata default_hstate_max_huge_pages;
>  static bool __initdata parsed_valid_hugepagesz = true;
>  static bool __initdata parsed_default_hugepagesz;
> +static unsigned int default_hugepages_in_node[MAX_NUMNODES] __initdata;
>  
>  /*
>   * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
> @@ -2868,33 +2869,41 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  	return ERR_PTR(-ENOSPC);
>  }
>  
> -int alloc_bootmem_huge_page(struct hstate *h)
> +int alloc_bootmem_huge_page(struct hstate *h, int nid)
>  	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
> -int __alloc_bootmem_huge_page(struct hstate *h)
> +int __alloc_bootmem_huge_page(struct hstate *h, int nid)
>  {
>  	struct huge_bootmem_page *m;
>  	int nr_nodes, node;
>  
> +	if (nid >= nr_online_nodes)
> +		return 0;
> +	/* do node specific alloc */
> +	if (nid != NUMA_NO_NODE) {
> +		m = memblock_alloc_try_nid_raw(huge_page_size(h), huge_page_size(h),
> +				0, MEMBLOCK_ALLOC_ACCESSIBLE, nid);
> +		if (m)
> +			goto found;
> +		else
> +			return 0;
> +	}
> +	/* do all node balanced alloc */

I'm sure you saw Smatch does not like this comment.  Perhaps,
	/* allocate from next node when distributing huge pages */

>  	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
> -		void *addr;
> -
> -		addr = memblock_alloc_try_nid_raw(
> +		m = memblock_alloc_try_nid_raw(
>  				huge_page_size(h), huge_page_size(h),
>  				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
> -		if (addr) {
> -			/*
> -			 * Use the beginning of the huge page to store the
> -			 * huge_bootmem_page struct (until gather_bootmem
> -			 * puts them into the mem_map).
> -			 */
> -			m = addr;
> +		/*
> +		 * Use the beginning of the huge page to store the
> +		 * huge_bootmem_page struct (until gather_bootmem
> +		 * puts them into the mem_map).
> +		 */
> +		if (m)
>  			goto found;
> -		}
> +		else
> +			return 0;
>  	}
> -	return 0;
>  
>  found:
> -	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
>  	/* Put them into a private list first because mem_map is not up yet */
>  	INIT_LIST_HEAD(&m->list);
>  	list_add(&m->list, &huge_boot_pages);
> @@ -2934,12 +2943,61 @@ static void __init gather_bootmem_prealloc(void)
>  		cond_resched();
>  	}
>  }
> +static void __init hugetlb_hstate_alloc_pages_onenode(struct hstate *h, int nid)
> +{
> +	unsigned long i;
> +	char buf[32];
> +
> +	for (i = 0; i < h->max_huge_pages_node[nid]; ++i) {
> +		if (hstate_is_gigantic(h)) {
> +			if (!alloc_bootmem_huge_page(h, nid))
> +				break;
> +		} else {
> +			struct page *page;
> +			gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
> +
> +			page = alloc_fresh_huge_page(h, gfp_mask, nid,
> +					&node_states[N_MEMORY], NULL);
> +			if (!page)
> +				break;
> +			put_page(page); /* free it into the hugepage allocator */
> +		}
> +		cond_resched();
> +	}
> +	if (i == h->max_huge_pages_node[nid])
> +		return;
> +
> +	string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
> +	pr_warn("HugeTLB: allocating %u of page size %s failed node%d.  Only allocated %lu hugepages.\n",
> +		h->max_huge_pages_node[nid], buf, nid, i);
> +	h->max_huge_pages -= (h->max_huge_pages_node[nid] - i);
> +	h->max_huge_pages_node[nid] = i;
> +}
>  
>  static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
>  {
>  	unsigned long i;
>  	nodemask_t *node_alloc_noretry;
> +	bool node_specific_alloc = false;
> +
> +	/* skip gigantic hugepages allocation if hugetlb_cma enabled */
> +	if (hstate_is_gigantic(h) && hugetlb_cma_size) {
> +		pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
> +		return;
> +	}
> +
> +	/* do node specific alloc */
> +	for (i = 0; i < nr_online_nodes; i++) {
> +		if (h->max_huge_pages_node[i] > 0) {
> +			hugetlb_hstate_alloc_pages_onenode(h, i);
> +			node_specific_alloc = true;
> +		}
> +	}
>  
> +	if (node_specific_alloc)
> +		return;
> +
> +	/* bellow will do all node balanced alloc */

	   below

>  	if (!hstate_is_gigantic(h)) {
>  		/*
>  		 * Bit mask controlling how hard we retry per-node allocations.
> @@ -2960,11 +3018,7 @@ static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
>  
>  	for (i = 0; i < h->max_huge_pages; ++i) {
>  		if (hstate_is_gigantic(h)) {
> -			if (hugetlb_cma_size) {
> -				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
> -				goto free;
> -			}
> -			if (!alloc_bootmem_huge_page(h))
> +			if (!alloc_bootmem_huge_page(h, NUMA_NO_NODE))
>  				break;
>  		} else if (!alloc_pool_huge_page(h,
>  					 &node_states[N_MEMORY],
> @@ -2980,7 +3034,6 @@ static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
>  			h->max_huge_pages, buf, i);
>  		h->max_huge_pages = i;
>  	}
> -free:
>  	kfree(node_alloc_noretry);
>  }
>  
> @@ -3671,6 +3724,11 @@ static int __init hugetlb_init(void)

>  			default_hstate.max_huge_pages =
>  				default_hstate_max_huge_pages;
> +
> +			for (i = 0; i < nr_online_nodes; i++)
> +				if (default_hugepages_in_node[i] > 0)
> +					default_hstate.max_huge_pages_node[i] =
> +						default_hugepages_in_node[i];

Very minor nit.  Not necessary to change.
The 'if (default_hugepages_in_node[i] > 0)' is not really needed.  I am
not sure if any potential speed up by the check would be noticable.  You
need to 'read' every value in the array.

>  		}
>  	}
>  
> @@ -3731,6 +3789,10 @@ void __init hugetlb_add_hstate(unsigned int order)
>  	parsed_hstate = h;
>  }
>  
> +bool __init __weak node_specific_alloc_support(void)
> +{
> +	return true;
> +}
>  /*
>   * hugepages command line processing
>   * hugepages normally follows a valid hugepagsz or default_hugepagsz
> @@ -3742,6 +3804,10 @@ static int __init hugepages_setup(char *s)
>  {
>  	unsigned long *mhp;
>  	static unsigned long *last_mhp;
> +	int node = NUMA_NO_NODE;
> +	int count;
> +	unsigned long tmp;
> +	char *p = s;
>  
>  	if (!parsed_valid_hugepagesz) {
>  		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
> @@ -3765,8 +3831,40 @@ static int __init hugepages_setup(char *s)
>  		return 0;
>  	}
>  
> -	if (sscanf(s, "%lu", mhp) <= 0)
> -		*mhp = 0;
> +	while (*p) {
> +		count = 0;
> +		if (sscanf(p, "%lu%n", &tmp, &count) != 1)
> +			goto invalid;
> +		/* Parameter is node format */
> +		if (p[count] == ':') {
> +			if (!node_specific_alloc_support()) {
> +				pr_warn("HugeTLB: architecture can't support node specific alloc, ignoring!\n");
> +				return 0;
> +			}
> +			node = tmp;
> +			p += count + 1;
> +			if (node < 0 || node >= nr_online_nodes)
> +				goto invalid;
> +			/* Parse hugepages */
> +			if (sscanf(p, "%lu%n", &tmp, &count) != 1)
> +				goto invalid;
> +			if (!hugetlb_max_hstate)
> +				default_hugepages_in_node[node] = tmp;
> +			else
> +				parsed_hstate->max_huge_pages_node[node] = tmp;
> +			*mhp += tmp;
> +			/* Go to parse next node*/
> +			if (p[count] == ',')
> +				p += count + 1;
> +			else
> +				break;
> +		} else {
> +			if (p != s)
> +				goto invalid;
> +			*mhp = tmp;
> +			break;
> +		}
> +	}
>  
>  	/*
>  	 * Global state is always initialized later in hugetlb_init.
> @@ -3779,6 +3877,10 @@ static int __init hugepages_setup(char *s)
>  	last_mhp = mhp;
>  
>  	return 1;
> +
> +invalid:
> +	pr_warn("HugeTLB: Invalid hugepages parameter %s\n", p);
> +	return 0;
>  }
>  __setup("hugepages=", hugepages_setup);
>  
> @@ -3840,6 +3942,7 @@ __setup("hugepagesz=", hugepagesz_setup);
>  static int __init default_hugepagesz_setup(char *s)
>  {
>  	unsigned long size;
> +	int i;
>  
>  	parsed_valid_hugepagesz = false;
>  	if (parsed_default_hugepagesz) {
> @@ -3868,6 +3971,10 @@ static int __init default_hugepagesz_setup(char *s)
>  	 */
>  	if (default_hstate_max_huge_pages) {
>  		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
> +		for (i = 0; i < nr_online_nodes; i++)
> +			if (default_hugepages_in_node[i] > 0)
> +				default_hstate.max_huge_pages_node[i] =
> +					default_hugepages_in_node[i];

Same minor nit as in hugetlb_init.

>  		if (hstate_is_gigantic(&default_hstate))
>  			hugetlb_hstate_alloc_pages(&default_hstate);
>  		default_hstate_max_huge_pages = 0;
> 

-- 
Mike Kravetz

Powered by blists - more mailing lists