Unverified Commit 02a28d02 authored by openeuler-ci-bot's avatar openeuler-ci-bot Committed by Gitee
Browse files

!15251 Intel: Backport upstream auto-tune per-CPU pageset size patchset 

Merge Pull Request from: @aubrey-intel 
 
There are 10 patches of Auto-tune per-CPU pageset size, 9 original patches:

[1/9] mm, pcp: avoid to drain PCP when process exit
[2/9] cacheinfo: calculate per-CPU data cache size
[3/9] mm, pcp: reduce lock contention for draining high-order pages
[4/9] mm: restrict the pcp batch scale factor to avoid too long latency
[5/9] mm, page_alloc: scale the number of pages that are batch allocated
[6/9] mm: add framework for PCP high auto-tuning
[7/9] mm: tune PCP high automatically
[8/9] mm, pcp: decrease PCP high if free pages < high watermark
[9/9] mm, pcp: reduce detecting time of consecutive high order page freeing

And a coming fix patch,
[1/1] mm and cache_info: remove unnecessary CPU cache info update

The original 9 patches has merged into OLK-6.6, this PR only contains the fix patch. 
 
Link:https://gitee.com/openeuler/kernel/pulls/15251

 

Reviewed-by: default avatarJason Zeng <jason.zeng@intel.com>
Reviewed-by: default avatarKefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: default avatarZhang Peng <zhangpeng362@huawei.com>
parents 2a45fac1 a74ecc06

drivers/base/cacheinfo.c

+44 −6
Original line number Diff line number Diff line
@@ -939,6 +939,37 @@ static int cache_add_dev(unsigned int cpu) 
	return rc;

}



static unsigned int cpu_map_shared_cache(bool online, unsigned int cpu,

					 cpumask_t **map)

{

	struct cacheinfo *llc, *sib_llc;

	unsigned int sibling;



	if (!last_level_cache_is_valid(cpu))

		return 0;



	llc = per_cpu_cacheinfo_idx(cpu, cache_leaves(cpu) - 1);



	if (llc->type != CACHE_TYPE_DATA && llc->type != CACHE_TYPE_UNIFIED)

		return 0;



	if (online) {

		*map = &llc->shared_cpu_map;

		return cpumask_weight(*map);

	}



	/* shared_cpu_map of offlined CPU will be cleared, so use sibling map */

	for_each_cpu(sibling, &llc->shared_cpu_map) {

		if (sibling == cpu || !last_level_cache_is_valid(sibling))

			continue;

		sib_llc = per_cpu_cacheinfo_idx(sibling, cache_leaves(sibling) - 1);

		*map = &sib_llc->shared_cpu_map;

		return cpumask_weight(*map);

	}



	return 0;

}



/*

 * Calculate the size of the per-CPU data cache slice.  This can be

 * used to estimate the size of the data cache slice that can be used
@@ -970,28 +1001,31 @@ static void update_per_cpu_data_slice_size_cpu(unsigned int cpu) 
		ci->per_cpu_data_slice_size = llc->size / nr_shared;

}



static void update_per_cpu_data_slice_size(bool cpu_online, unsigned int cpu)

static void update_per_cpu_data_slice_size(bool cpu_online, unsigned int cpu,

					   cpumask_t *cpu_map)

{

	unsigned int icpu;



	for_each_online_cpu(icpu) {

	for_each_cpu(icpu, cpu_map) {

		if (!cpu_online && icpu == cpu)

			continue;

		update_per_cpu_data_slice_size_cpu(icpu);

		setup_pcp_cacheinfo(icpu);

	}

}



static int cacheinfo_cpu_online(unsigned int cpu)

{

	int rc = detect_cache_attributes(cpu);

	cpumask_t *cpu_map;



	if (rc)

		return rc;

	rc = cache_add_dev(cpu);

	if (rc)

		goto err;

	update_per_cpu_data_slice_size(true, cpu);

	setup_pcp_cacheinfo();

	if (cpu_map_shared_cache(true, cpu, &cpu_map))

		update_per_cpu_data_slice_size(true, cpu, cpu_map);

	return 0;

err:

	free_cache_attributes(cpu);
@@ -1000,12 +1034,16 @@ static int cacheinfo_cpu_online(unsigned int cpu) 


static int cacheinfo_cpu_pre_down(unsigned int cpu)

{

	cpumask_t *cpu_map;

	unsigned int nr_shared;



	nr_shared = cpu_map_shared_cache(false, cpu, &cpu_map);

	if (cpumask_test_and_clear_cpu(cpu, &cache_dev_map))

		cpu_cache_sysfs_exit(cpu);



	free_cache_attributes(cpu);

	update_per_cpu_data_slice_size(false, cpu);

	setup_pcp_cacheinfo();

	if (nr_shared > 1)

		update_per_cpu_data_slice_size(false, cpu, cpu_map);

	return 0;

}

include/linux/gfp.h

+1 −1
Original line number Diff line number Diff line
@@ -341,7 +341,7 @@ void drain_all_pages(struct zone *zone); 
void drain_local_pages(struct zone *zone);



void page_alloc_init_late(void);

void setup_pcp_cacheinfo(void);

void setup_pcp_cacheinfo(unsigned int cpu);



/*

 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what

mm/page_alloc.c

+18 −21
Original line number Diff line number Diff line
@@ -5832,13 +5832,11 @@ static void zone_pcp_update(struct zone *zone, int cpu_online) 
	mutex_unlock(&pcp_batch_high_lock);

}



static void zone_pcp_update_cacheinfo(struct zone *zone)

static void zone_pcp_update_cacheinfo(struct zone *zone, unsigned int cpu)

{

	int cpu;

	struct per_cpu_pages *pcp;

	struct cpu_cacheinfo *cci;



	for_each_online_cpu(cpu) {

	pcp = per_cpu_ptr(zone->per_cpu_pageset, cpu);

	cci = get_cpu_cacheinfo(cpu);

	/*
@@ -5855,14 +5853,13 @@ static void zone_pcp_update_cacheinfo(struct zone *zone) 
		pcp->flags &= ~PCPF_FREE_HIGH_BATCH;

	spin_unlock(&pcp->lock);

}

}



void setup_pcp_cacheinfo(void)

void setup_pcp_cacheinfo(unsigned int cpu)

{

	struct zone *zone;



	for_each_populated_zone(zone)

		zone_pcp_update_cacheinfo(zone);

		zone_pcp_update_cacheinfo(zone, cpu);

}



/*