Received-SPF: pass (google.com: domain of linux-kernel+bounces-97356-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Fri, 8 Mar 2024 12:11:41 -0500
From: Daniel Jordan <daniel.m.jordan@oracle.com>
To: Gang Li <gang.li@linux.dev>
Cc: Andrew Morton <akpm@linux-foundation.org>,
        David Hildenbrand <david@redhat.com>,
        David Rientjes <rientjes@google.com>,
        Muchun Song <muchun.song@linux.dev>,
        Tim Chen <tim.c.chen@linux.intel.com>,
        Steffen Klassert <steffen.klassert@secunet.com>,
        Jane Chu <jane.chu@oracle.com>,
        "Paul E . McKenney" <paulmck@kernel.org>,
        Randy Dunlap <rdunlap@infradead.org>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, ligang.bdlg@bytedance.com
Subject: Re: [PATCH v6 7/8] hugetlb: parallelize 2M hugetlb allocation and
 initialization
Message-ID: <icrdkacpdksofftv5jwrwcgojsa7qnby4iuvxsdktuxazivhks@ajcy2shag4nz>
References: <20240222140422.393911-1-gang.li@linux.dev>
 <20240222140422.393911-8-gang.li@linux.dev>
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240222140422.393911-8-gang.li@linux.dev>
Precedence: bulk
MIME-Version: 1.0
X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1
X-MS-Exchange-AntiSpam-MessageData-0: 
	=?us-ascii?Q?Y4s+sBjpgmzKzwbUTKxgzXJOhgpUfMVdcz0+rswinCMSUukSIodqoUi6krIn?=
 =?us-ascii?Q?wFtb3qMa9LZf0cPUn4u1oTA3P5OvanShl9ZcdHpCeTd/EmvI/uEZW7Ley32Z?=
 =?us-ascii?Q?vcvOtCmlK+aKvSxH5/vaQWPv8Mgbcz7zKzkdQSdgeZqeO6JJMlAmvojBNCn3?=
 =?us-ascii?Q?gC6xTbd9swP/cAsedvLpnD+017or1x1w30AWRFNAcDNvZr+hyZYcHrd/YFHj?=
 =?us-ascii?Q?0wtq5eGxgZ1z5ArdviMGxQ1R6XkkjWaO5wneYhn6jkwiOQZjcuS4aB/AzeO6?=
 =?us-ascii?Q?VgQ7qXBl65YxmEyfspuYeuRZ8VutN4XL8XBin/PhnJuOzL4We4wjWs+Q0kmR?=
 =?us-ascii?Q?2hESmTcFiwtTxlJJLD6cOc2skvNH4isCMZAsIBme55rAil9nGpD+jypIYT4F?=
 =?us-ascii?Q?AgmpnqEmjwr4wq4TJ0en4+5/0qf/25cd79xbhcZ2T6FwQwAcfgHq8WD8K4HC?=
 =?us-ascii?Q?nMbbSGvCk2ROYvfpQKrmGOuZx9XnJhlkB+Vsm16ED0bP2e8kaQyiahB/rXek?=
 =?us-ascii?Q?N2PWmIkDR+3KV6uASuCzpUPWgZxnR2HV4DwNhMZZLUgQAvicpX+hT47AGVxN?=
 =?us-ascii?Q?mT5Nt8J7cbaP3bEK7bPeuhpFDUCUV17PW8L7+Veq4DidjA7f0Ec+RnRXMvtx?=
 =?us-ascii?Q?dpY+rKjvDAYHgpxPGOxwldbpjiENFTU8OYxgJYWmxsNruom8f311tlx4rpZ9?=
 =?us-ascii?Q?mfpJueNWx9w5MQTF8Ul1E9FFCo3nYnpGSms83MNbwuQGmicbhIko/L+9j7aW?=
 =?us-ascii?Q?3+U76grMUVV1YfE5LyiBrSboMYliXuFcBeF2YQexH1APoZ/MhfUFBYVqWp4A?=
 =?us-ascii?Q?lAbkFbtu9QuWlulCOionT1q1Jmd8HwTQKr0j8iNKPgelC/BkOksfVli90Fki?=
 =?us-ascii?Q?veqd8JxajA1uWxo1IqbKNOP3QsnZWQJCsK5bz/mEhoVLJodN/zYqr+chdDca?=
 =?us-ascii?Q?pTD0ld6ToKQHv8NKvdsVhoGjcPvT0K5CWAmNbDkRB+Z0ZDLJP0/b0aUudxvC?=
 =?us-ascii?Q?o6Ki55ZL+yBMI9GkuJqry27m4k6uOmxFkntDYo1aaE3NsLRrLLWGerCsdt9y?=
 =?us-ascii?Q?KfzuKf6nYkN4ANEaDdBgBkyXid0g+DZ5W/bJYpdGWAmDs2w6fUb9iXgHrm3v?=
 =?us-ascii?Q?UkoW4HWT+IQUflMBjC7YFdk0fLQS/bET3l+GojIyC0d6ccPSbzvB0og/eYB1?=
 =?us-ascii?Q?wEBv3nxv4zR9zkbpow7ngLIksBX00UKBeyB6aypYAKnLtbvDkW3MOIqH2ANM?=
 =?us-ascii?Q?HEVcNl20Cym01339GK0EAodnsdCMCWON0+HpZJwXoqxLTwRk5M+XSpH/zkSq?=
 =?us-ascii?Q?zlPPnj9bjqFhFs6P/OZmMCaPfccLFvFN6PhUIHmyioZj5qIU92hBmIl7LkgM?=
 =?us-ascii?Q?AO7TPCKygZM0ToMAEXVM4zP8BgW8dOmg7wDWIDCpLTzQpGrCxocMPLwGYmVw?=
 =?us-ascii?Q?nyGUdy7RIrzxAFlxO03itwWVa/BLTdQEVBRwD8tVU8j5DJxOrSJ0ywuOJTxP?=
 =?us-ascii?Q?LHIDk0h+E/fzuiWUKc2DtxDXkyMoDyNGYe1Jqh13kVymbtm8l+04kP7UCpxB?=
 =?us-ascii?Q?Z5j0hxFyYwqrNAVd+3srkjTexIdKJ1m+3ez9bVgxkF1OfxhYR7OMi5/KvTSn?=
 =?us-ascii?Q?bg=3D=3D?=
X-MS-Exchange-AntiSpam-ExternalHop-MessageData-ChunkCount: 1
X-MS-Exchange-AntiSpam-ExternalHop-MessageData-0: 
	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
X-OriginatorOrg: oracle.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 8c555f7c-bfcc-4dd6-4f69-08dc3f92d479
X-MS-Exchange-CrossTenant-AuthSource: SA1PR10MB5711.namprd10.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 08 Mar 2024 17:11:44.4228
 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: 4e2c6054-71cb-48f1-bd6c-3a9705aca71b
X-MS-Exchange-CrossTenant-MailboxType: HOSTED
X-MS-Exchange-CrossTenant-UserPrincipalName: lPAel0EW3Gv+Nrflj1JlEWB8ndIa40EAPVAhJ/u/YkRz9bPPdPFPmQes2ubkHmx1o8U5qX4EZ+Kj2g6Nw0qKe74gOYZnkagvwjuDKC8K8Ws=
X-MS-Exchange-Transport-CrossTenantHeadersStamped: SN7PR10MB6954
X-Proofpoint-Virus-Version: vendor=baseguard
 engine=ICAP:2.0.272,Aquarius:18.0.1011,Hydra:6.0.619,FMLib:17.11.176.26
 definitions=2024-03-08_08,2024-03-06_01,2023-05-22_02
X-Proofpoint-Spam-Details: rule=notspam policy=default score=0 phishscore=0 mlxlogscore=999
 suspectscore=0 adultscore=0 bulkscore=0 malwarescore=0 mlxscore=0
 spamscore=0 classifier=spam adjust=0 reason=mlx scancount=1
 engine=8.12.0-2311290000 definitions=main-2403080136
X-Proofpoint-GUID: ERRER47xv4goKQ7tFJ1YUkyHvgvSjUin
X-Proofpoint-ORIG-GUID: ERRER47xv4goKQ7tFJ1YUkyHvgvSjUin

Hi,

On Thu, Feb 22, 2024 at 10:04:20PM +0800, Gang Li wrote:
> By distributing both the allocation and the initialization tasks across
> multiple threads, the initialization of 2M hugetlb will be faster,
> thereby improving the boot speed.
> 
> Here are some test results:
>       test case        no patch(ms)   patched(ms)   saved
>  ------------------- -------------- ------------- --------
>   256c2T(4 node) 2M           3336          1051   68.52%
>   128c1T(2 node) 2M           1943           716   63.15%

Great improvement, and glad to see the multithreading is useful here.

>  static unsigned long __init hugetlb_pages_alloc_boot(struct hstate *h)
>  {
> -	unsigned long i;
> -	struct folio *folio;
> -	LIST_HEAD(folio_list);
> -	nodemask_t node_alloc_noretry;
> -
> -	/* Bit mask controlling how hard we retry per-node allocations.*/
> -	nodes_clear(node_alloc_noretry);
> +	struct padata_mt_job job = {
> +		.fn_arg		= h,
> +		.align		= 1,
> +		.numa_aware	= true
> +	};
>  
> -	for (i = 0; i < h->max_huge_pages; ++i) {
> -		folio = alloc_pool_huge_folio(h, &node_states[N_MEMORY],
> -						&node_alloc_noretry);
> -		if (!folio)
> -			break;
> -		list_add(&folio->lru, &folio_list);
> -		cond_resched();
> -	}
> +	job.thread_fn	= hugetlb_pages_alloc_boot_node;
> +	job.start	= 0;
> +	job.size	= h->max_huge_pages;
>  
> -	prep_and_add_allocated_folios(h, &folio_list);
> +	/*
> +	 * job.max_threads is twice the num_node_state(N_MEMORY),
> +	 *
> +	 * Tests below indicate that a multiplier of 2 significantly improves
> +	 * performance, and although larger values also provide improvements,
> +	 * the gains are marginal.
> +	 *
> +	 * Therefore, choosing 2 as the multiplier strikes a good balance between
> +	 * enhancing parallel processing capabilities and maintaining efficient
> +	 * resource management.
> +	 *
> +	 * +------------+-------+-------+-------+-------+-------+
> +	 * | multiplier |   1   |   2   |   3   |   4   |   5   |
> +	 * +------------+-------+-------+-------+-------+-------+
> +	 * | 256G 2node | 358ms | 215ms | 157ms | 134ms | 126ms |
> +	 * | 2T   4node | 979ms | 679ms | 543ms | 489ms | 481ms |
> +	 * | 50G  2node | 71ms  | 44ms  | 37ms  | 30ms  | 31ms  |
> +	 * +------------+-------+-------+-------+-------+-------+
> +	 */
> +	job.max_threads	= num_node_state(N_MEMORY) * 2;
> +	job.min_chunk	= h->max_huge_pages / num_node_state(N_MEMORY) / 2;

For a single huge page, we get min_chunk of 0.  padata doesn't
explicitly handle that, but 'align' happens to save us from div by 0
later on.  It's an odd case, something to fix if there were another
version.


Not sure what efficient resource management means here.  Avoiding lock
contention?  The system is waiting on this initialization to start pid
1.  On big systems, most CPUs will be idle, so why not use available
resources to optimize it more?  max_threads could scale with CPU count
rather than a magic multiplier.

With that said, the major gain is already there, so either way,

Acked-by: Daniel Jordan <daniel.m.jordan@oracle.com> # padata