- Pmem - lists.openanolis.cn

Re: [PATCH v2] mm: make cpu-less memory node to movable node only

by Baolin Wang

On 9/1/2022 3:46 PM, Feng Tang wrote: > From 1784d7c7a6911a3d6a93d1847d84e9297c8e4f4c Mon Sep 17 00:00:00 2001 > From: Feng Tang <feng.tang(a)intel.com> > Date: Wed, 31 Aug 2022 15:31:04 +0800 > Subject: [PATCH v2] mm: make cpu-less memory node to movable node only > > When platform has PMEM node, those node are mostly a pure memory node > without any CPU attached. To avoid performance/latency sensitive > kernel allocation on these nodes, make all cpu-less memory nodes > be enabled as node with movable zone only. > > And when users want it to be normal node, appending "cpuless_node_normal" > to kernel cmdline parameters can achive that. > > Signed-off-by: Feng Tang <feng.tang(a)intel.com> > Signed-off-by: Huaisheng Ye <huaisheng.ye(a)intel.com> LGTM. Thanks. Reviewed-by: Baolin Wang <baolin.wang(a)linux.alibaba.com> > --- > Changelog: > > Since v1 > * directly jump to out2 after movable handling (Baolin Wang) > > arch/x86/mm/srat.c | 5 +++++ > drivers/base/node.c | 2 ++ > include/linux/nodemask.h | 1 + > mm/page_alloc.c | 32 ++++++++++++++++++++++++++++++++ > 4 files changed, 40 insertions(+) > > diff --git a/arch/x86/mm/srat.c b/arch/x86/mm/srat.c > index dac07e4f5834..dcb568067d79 100644 > --- a/arch/x86/mm/srat.c > +++ b/arch/x86/mm/srat.c > @@ -60,6 +60,9 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa) > node_set(node, numa_nodes_parsed); > printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u\n", > pxm, apic_id, node); > + > + if (!node_state(node, N_POSSIBLE_CPU)) > + node_set_state(node, N_POSSIBLE_CPU); > } > > /* Callback for Proximity Domain -> LAPIC mapping */ > @@ -101,6 +104,8 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa) > node_set(node, numa_nodes_parsed); > printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%02x -> Node %u\n", > pxm, apic_id, node); > + if (!node_state(node, N_POSSIBLE_CPU)) > + node_set_state(node, N_POSSIBLE_CPU); > } > > int __init x86_acpi_numa_init(void) > diff --git a/drivers/base/node.c b/drivers/base/node.c > index 619fb1e38dd2..39ee52c69609 100644 > --- a/drivers/base/node.c > +++ b/drivers/base/node.c > @@ -990,6 +990,7 @@ static struct node_attr node_state_attr[] = { > #endif > [N_MEMORY] = _NODE_ATTR(has_memory, N_MEMORY), > [N_CPU] = _NODE_ATTR(has_cpu, N_CPU), > + [N_POSSIBLE_CPU] = _NODE_ATTR(has_possible_cpu, N_POSSIBLE_CPU), > }; > > static struct attribute *node_state_attrs[] = { > @@ -1001,6 +1002,7 @@ static struct attribute *node_state_attrs[] = { > #endif > &node_state_attr[N_MEMORY].attr.attr, > &node_state_attr[N_CPU].attr.attr, > + &node_state_attr[N_POSSIBLE_CPU].attr.attr, > NULL > }; > > diff --git a/include/linux/nodemask.h b/include/linux/nodemask.h > index 5a30ad594ccc..ad3100b06938 100644 > --- a/include/linux/nodemask.h > +++ b/include/linux/nodemask.h > @@ -399,6 +399,7 @@ enum node_states { > #endif > N_MEMORY, /* The node has memory(regular, high, movable) */ > N_CPU, /* The node has one or more cpus */ > + N_POSSIBLE_CPU, /* possible nodes that have cpu attached, only used in early boot phase */ > NR_NODE_STATES > }; > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index 8cfce927fb9b..f20976536cfc 100644 > --- a/mm/page_alloc.c > +++ b/mm/page_alloc.c > @@ -294,6 +294,7 @@ static unsigned long required_kernelcore_percent __initdata; > static unsigned long required_movablecore __initdata; > static unsigned long required_movablecore_percent __initdata; > static unsigned long zone_movable_pfn[MAX_NUMNODES] __meminitdata; > +static bool enable_cpuless_memnode_normal_node __initdata; > static bool mirrored_kernelcore __meminitdata; > > /* movable_zone is the "real" zone pages in ZONE_MOVABLE are taken from */ > @@ -7163,6 +7164,26 @@ static void __init find_zone_movable_pfns_for_nodes(void) > required_movablecore = (totalpages * 100 * required_movablecore_percent) / > 10000UL; > > + if (!required_kernelcore && !required_movablecore && > + !nodes_empty(node_states[N_POSSIBLE_CPU]) && > + !enable_cpuless_memnode_normal_node) { > + > + /* Put meory from cpu-less nodes into movable zones */ > + for_each_memblock(memory, r) { > + nid = memblock_get_region_node(r); > + > + if (node_isset(nid, node_states[N_POSSIBLE_CPU])) > + continue; > + > + usable_startpfn = PFN_DOWN(r->base); > + zone_movable_pfn[nid] = zone_movable_pfn[nid] ? > + min(usable_startpfn, zone_movable_pfn[nid]) : > + usable_startpfn; > + } > + > + goto out2; > + } > + > /* > * If movablecore= was specified, calculate what size of > * kernelcore that corresponds so that memory usable for > @@ -7457,8 +7478,19 @@ static int __init cmdline_parse_movablecore(char *p) > &required_movablecore_percent); > } > > +/* > + * cpuless memory nodes will be enabled to movable node by default, > + * add this cmdline to make it be enabled as a normal node > + */ > +static int __init cmdline_parse_cpuless_memnode(char *str) > +{ > + enable_cpuless_memnode_normal_node = true; > + return 0; > +} > + > early_param("kernelcore", cmdline_parse_kernelcore); > early_param("movablecore", cmdline_parse_movablecore); > +early_param("cpuless_node_normal", cmdline_parse_cpuless_memnode); > > #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ >

1 year, 8 months

1
0
0 0

Re: [PATCH] mm: make cpu-less memory node to movable node only

by zhong jiang

On 2022/9/1 9:46 上午, Feng Tang wrote: > Hi Zhong Jiang, > > Thanks for the review! > > On Wed, Aug 31, 2022 at 07:20:21PM +0800, zhong jiang wrote: >> On 2022/8/31 3:31 下午, Huaisheng Ye wrote: >>> From: Feng Tang <feng.tang(a)intel.com> >>> >>> When platform has PMEM node, those node are mostly a pure memory node >>> without any CPU attached. To avoid performance/latency sensitive >>> kernel allocation on these nodes, make all cpu-less memory nodes >>> be enabled as node with movable zone only. >>> >>> And when users want it to be normal node, appending "cpuless_node_normal" >>> to kernel cmdline parameters can achive that. >>> >>> Signed-off-by: Feng Tang <feng.tang(a)intel.com> >>> Signed-off-by: Huaisheng Ye <huaisheng.ye(a)intel.com> >>> --- >>> arch/x86/mm/srat.c | 5 +++++ >>> drivers/base/node.c | 2 ++ >>> include/linux/nodemask.h | 1 + >>> mm/page_alloc.c | 30 ++++++++++++++++++++++++++++++ >>> 4 files changed, 38 insertions(+) >>> >>> diff --git a/arch/x86/mm/srat.c b/arch/x86/mm/srat.c >>> index dac07e4f5834..dcb568067d79 100644 >>> --- a/arch/x86/mm/srat.c >>> +++ b/arch/x86/mm/srat.c >>> @@ -60,6 +60,9 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa) >>> node_set(node, numa_nodes_parsed); >>> printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u\n", >>> pxm, apic_id, node); >>> + >>> + if (!node_state(node, N_POSSIBLE_CPU)) >>> + node_set_state(node, N_POSSIBLE_CPU); >>> } >>> /* Callback for Proximity Domain -> LAPIC mapping */ >>> @@ -101,6 +104,8 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa) >>> node_set(node, numa_nodes_parsed); >>> printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%02x -> Node %u\n", >>> pxm, apic_id, node); >>> + if (!node_state(node, N_POSSIBLE_CPU)) >>> + node_set_state(node, N_POSSIBLE_CPU); >>> } >> IIUC, pmem memory also store in the srat table, the above change seems to >> fail to work because pmem memory has marked as the N_POSSIBLE_CPU. > PMEM node will not be set for N_POSSIBLE_CPU, the upper 2 functions > are only for parsing CPU entry with interrupt controllers. > > IIUC, you are refering the NUMA node info maintained by architectural > code (say arch/x86): node_possible_map / numa_nodes_parsed, where PMEM > node will be set. > > And the node map here N_POSSIBLE_CPU is from another set mainly for > mm/ code, and used later in memory init part. > > This patch has been tested on QEMU and RE7P machines. Thanks for your clarification. > Thanks, > Feng > > >> Am I missing? >> >> Thanks, >>> int __init x86_acpi_numa_init(void) >>> diff --git a/drivers/base/node.c b/drivers/base/node.c >>> index 619fb1e38dd2..39ee52c69609 100644 >>> --- a/drivers/base/node.c >>> +++ b/drivers/base/node.c >>> @@ -990,6 +990,7 @@ static struct node_attr node_state_attr[] = { >>> #endif >>> [N_MEMORY] = _NODE_ATTR(has_memory, N_MEMORY), >>> [N_CPU] = _NODE_ATTR(has_cpu, N_CPU), >>> + [N_POSSIBLE_CPU] = _NODE_ATTR(has_possible_cpu, N_POSSIBLE_CPU), >>> }; >>> static struct attribute *node_state_attrs[] = { >>> @@ -1001,6 +1002,7 @@ static struct attribute *node_state_attrs[] = { >>> #endif >>> &node_state_attr[N_MEMORY].attr.attr, >>> &node_state_attr[N_CPU].attr.attr, >>> + &node_state_attr[N_POSSIBLE_CPU].attr.attr, >>> NULL >>> }; >>> diff --git a/include/linux/nodemask.h b/include/linux/nodemask.h >>> index 5a30ad594ccc..ad3100b06938 100644 >>> --- a/include/linux/nodemask.h >>> +++ b/include/linux/nodemask.h >>> @@ -399,6 +399,7 @@ enum node_states { >>> #endif >>> N_MEMORY, /* The node has memory(regular, high, movable) */ >>> N_CPU, /* The node has one or more cpus */ >>> + N_POSSIBLE_CPU, /* possible nodes that have cpu attached, only used in early boot phase */ >>> NR_NODE_STATES >>> }; >>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c >>> index 8cfce927fb9b..9570ee499c12 100644 >>> --- a/mm/page_alloc.c >>> +++ b/mm/page_alloc.c >>> @@ -294,6 +294,7 @@ static unsigned long required_kernelcore_percent __initdata; >>> static unsigned long required_movablecore __initdata; >>> static unsigned long required_movablecore_percent __initdata; >>> static unsigned long zone_movable_pfn[MAX_NUMNODES] __meminitdata; >>> +static bool enable_cpuless_memnode_normal_node __initdata; >>> static bool mirrored_kernelcore __meminitdata; >>> /* movable_zone is the "real" zone pages in ZONE_MOVABLE are taken from */ >>> @@ -7163,6 +7164,24 @@ static void __init find_zone_movable_pfns_for_nodes(void) >>> required_movablecore = (totalpages * 100 * required_movablecore_percent) / >>> 10000UL; >>> + if (!required_kernelcore && !required_movablecore && >>> + !nodes_empty(node_states[N_POSSIBLE_CPU]) && >>> + !enable_cpuless_memnode_normal_node) { >>> + >>> + /* Put meory from cpu-less nodes into movable zones */ >>> + for_each_memblock(memory, r) { >>> + nid = memblock_get_region_node(r); >>> + >>> + if (node_isset(nid, node_states[N_POSSIBLE_CPU])) >>> + continue; >>> + >>> + usable_startpfn = PFN_DOWN(r->base); >>> + zone_movable_pfn[nid] = zone_movable_pfn[nid] ? >>> + min(usable_startpfn, zone_movable_pfn[nid]) : >>> + usable_startpfn; >>> + } >>> + } >>> + >>> /* >>> * If movablecore= was specified, calculate what size of >>> * kernelcore that corresponds so that memory usable for >>> @@ -7457,8 +7476,19 @@ static int __init cmdline_parse_movablecore(char *p) >>> &required_movablecore_percent); >>> } >>> +/* >>> + * cpuless memory nodes will be enabled to movable node by default, >>> + * add this cmdline to make it be enabled as a normal node >>> + */ >>> +static int __init cmdline_parse_cpuless_memnode(char *str) >>> +{ >>> + enable_cpuless_memnode_normal_node = true; >>> + return 0; >>> +} >>> + >>> early_param("kernelcore", cmdline_parse_kernelcore); >>> early_param("movablecore", cmdline_parse_movablecore); >>> +early_param("cpuless_node_normal", cmdline_parse_cpuless_memnode); >>> #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */

1 year, 8 months

1
0
0 0

[PATCH] mm: make cpu-less memory node to movable node only

by Huaisheng Ye

From: Feng Tang <feng.tang(a)intel.com> When platform has PMEM node, those node are mostly a pure memory node without any CPU attached. To avoid performance/latency sensitive kernel allocation on these nodes, make all cpu-less memory nodes be enabled as node with movable zone only. And when users want it to be normal node, appending "cpuless_node_normal" to kernel cmdline parameters can achive that. Signed-off-by: Feng Tang <feng.tang(a)intel.com> Signed-off-by: Huaisheng Ye <huaisheng.ye(a)intel.com> --- arch/x86/mm/srat.c | 5 +++++ drivers/base/node.c | 2 ++ include/linux/nodemask.h | 1 + mm/page_alloc.c | 30 ++++++++++++++++++++++++++++++ 4 files changed, 38 insertions(+) diff --git a/arch/x86/mm/srat.c b/arch/x86/mm/srat.c index dac07e4f5834..dcb568067d79 100644 --- a/arch/x86/mm/srat.c +++ b/arch/x86/mm/srat.c @@ -60,6 +60,9 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa) node_set(node, numa_nodes_parsed); printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u\n", pxm, apic_id, node); + + if (!node_state(node, N_POSSIBLE_CPU)) + node_set_state(node, N_POSSIBLE_CPU); } /* Callback for Proximity Domain -> LAPIC mapping */ @@ -101,6 +104,8 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa) node_set(node, numa_nodes_parsed); printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%02x -> Node %u\n", pxm, apic_id, node); + if (!node_state(node, N_POSSIBLE_CPU)) + node_set_state(node, N_POSSIBLE_CPU); } int __init x86_acpi_numa_init(void) diff --git a/drivers/base/node.c b/drivers/base/node.c index 619fb1e38dd2..39ee52c69609 100644 --- a/drivers/base/node.c +++ b/drivers/base/node.c @@ -990,6 +990,7 @@ static struct node_attr node_state_attr[] = { #endif [N_MEMORY] = _NODE_ATTR(has_memory, N_MEMORY), [N_CPU] = _NODE_ATTR(has_cpu, N_CPU), + [N_POSSIBLE_CPU] = _NODE_ATTR(has_possible_cpu, N_POSSIBLE_CPU), }; static struct attribute *node_state_attrs[] = { @@ -1001,6 +1002,7 @@ static struct attribute *node_state_attrs[] = { #endif &node_state_attr[N_MEMORY].attr.attr, &node_state_attr[N_CPU].attr.attr, + &node_state_attr[N_POSSIBLE_CPU].attr.attr, NULL }; diff --git a/include/linux/nodemask.h b/include/linux/nodemask.h index 5a30ad594ccc..ad3100b06938 100644 --- a/include/linux/nodemask.h +++ b/include/linux/nodemask.h @@ -399,6 +399,7 @@ enum node_states { #endif N_MEMORY, /* The node has memory(regular, high, movable) */ N_CPU, /* The node has one or more cpus */ + N_POSSIBLE_CPU, /* possible nodes that have cpu attached, only used in early boot phase */ NR_NODE_STATES }; diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 8cfce927fb9b..9570ee499c12 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -294,6 +294,7 @@ static unsigned long required_kernelcore_percent __initdata; static unsigned long required_movablecore __initdata; static unsigned long required_movablecore_percent __initdata; static unsigned long zone_movable_pfn[MAX_NUMNODES] __meminitdata; +static bool enable_cpuless_memnode_normal_node __initdata; static bool mirrored_kernelcore __meminitdata; /* movable_zone is the "real" zone pages in ZONE_MOVABLE are taken from */ @@ -7163,6 +7164,24 @@ static void __init find_zone_movable_pfns_for_nodes(void) required_movablecore = (totalpages * 100 * required_movablecore_percent) / 10000UL; + if (!required_kernelcore && !required_movablecore && + !nodes_empty(node_states[N_POSSIBLE_CPU]) && + !enable_cpuless_memnode_normal_node) { + + /* Put meory from cpu-less nodes into movable zones */ + for_each_memblock(memory, r) { + nid = memblock_get_region_node(r); + + if (node_isset(nid, node_states[N_POSSIBLE_CPU])) + continue; + + usable_startpfn = PFN_DOWN(r->base); + zone_movable_pfn[nid] = zone_movable_pfn[nid] ? + min(usable_startpfn, zone_movable_pfn[nid]) : + usable_startpfn; + } + } + /* * If movablecore= was specified, calculate what size of * kernelcore that corresponds so that memory usable for @@ -7457,8 +7476,19 @@ static int __init cmdline_parse_movablecore(char *p) &required_movablecore_percent); } +/* + * cpuless memory nodes will be enabled to movable node by default, + * add this cmdline to make it be enabled as a normal node + */ +static int __init cmdline_parse_cpuless_memnode(char *str) +{ + enable_cpuless_memnode_normal_node = true; + return 0; +} + early_param("kernelcore", cmdline_parse_kernelcore); early_param("movablecore", cmdline_parse_movablecore); +early_param("cpuless_node_normal", cmdline_parse_cpuless_memnode); #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ -- 2.27.0

1 year, 8 months

3
3
0 0

[PATCH] autonuma: reduce cache footprint when scanning page tables

by Baolin Wang

From: Huang Ying <ying.huang(a)intel.com> ANBZ: #80 commit a818f5363a0eba04bcff986c64c919d3f44b8017 upstream In auto NUMA balancing page table scanning, if the pte_protnone() is true, the PTE needs not to be changed because it's in target state already. So other checking on corresponding struct page is unnecessary too. So, if we check pte_protnone() firstly for each PTE, we can avoid unnecessary struct page accessing, so that reduce the cache footprint of NUMA balancing page table scanning. In the performance test of pmbench memory accessing benchmark with 80:20 read/write ratio and normal access address distribution on a 2 socket Intel server with Optance DC Persistent Memory, perf profiling shows that the autonuma page table scanning time reduces from 1.23% to 0.97% (that is, reduced 21%) with the patch. Link: http://lkml.kernel.org/r/20191101075727.26683-3-ying.huang@intel.com Signed-off-by: "Huang, Ying" <ying.huang(a)intel.com> Acked-by: Mel Gorman <mgorman(a)suse.de> Cc: Michal Hocko <mhocko(a)suse.com> Cc: Rik van Riel <riel(a)redhat.com> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Ingo Molnar <mingo(a)kernel.org> Cc: Dave Hansen <dave.hansen(a)linux.intel.com> Cc: Dan Williams <dan.j.williams(a)intel.com> Cc: Fengguang Wu <fengguang.wu(a)intel.com> Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds(a)linux-foundation.org> Signed-off-by: Baolin Wang <baolin.wang(a)linux.alibaba.com> --- mm/mprotect.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/mm/mprotect.c b/mm/mprotect.c index 6525e96..01681e3 100644 --- a/mm/mprotect.c +++ b/mm/mprotect.c @@ -84,6 +84,10 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd, int nid; bool toptier; + /* Avoid TLB flush if possible */ + if (pte_protnone(oldpte)) + continue; + page = vm_normal_page(vma, addr, oldpte); if (!page || PageKsm(page)) continue; @@ -93,10 +97,6 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd, page_mapcount(page) != 1) continue; - /* Avoid TLB flush if possible */ - if (pte_protnone(oldpte)) - continue; - /* * Don't mess with PTEs if page is already on the node * a single-threaded process is running on. -- 1.8.3.1

2 years, 1 month

2
1
0 0

The commit to reduce TLB shootdown during page promotion

by Huang, Ying

Hi, All, I have a patch as below to reduce TLB shootdown during page promotion. If not yet, you may consider to backport it to Anolis kernel. b99a342d4f11a5455d999b12f5fee42ab6acaf8c Author: Huang Ying <ying.huang(a)intel.com> AuthorDate: Thu Apr 29 22:57:41 2021 -0700 Commit: Linus Torvalds <torvalds(a)linux-foundation.org> CommitDate: Fri Apr 30 11:20:39 2021 -0700 NUMA balancing: reduce TLB flush via delaying mapping on hint page fault Best Regards, Huang, Ying

2 years, 1 month

2
1
0 0

[PATCH] anolis: mm: Add page promotion throttle statistic

by Baolin Wang

ANBZ: #80 Add page promotion throttle statistic, which can be used to check how many cold pages were trying to be prmoted to DRAM, and help to tuning the latency threhold. Signed-off-by: Baolin Wang <baolin.wang(a)linux.alibaba.com> --- include/linux/mmzone.h | 1 + kernel/sched/fair.c | 5 ++++- mm/vmstat.c | 1 + 3 files changed, 6 insertions(+), 1 deletion(-) diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h index d03a536..eecffa7 100644 --- a/include/linux/mmzone.h +++ b/include/linux/mmzone.h @@ -213,6 +213,7 @@ enum node_stat_item { PGPROMOTE_FILE, PGPROMOTE_TRY, /* pages to try to migrate via NUMA balancing */ PGDEMOTED_HOT, + PGPROMOTE_COLD_THROTTLE, #endif NR_VM_NODE_STAT_ITEMS }; diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index c473da4..0a11b03 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -3021,8 +3021,11 @@ bool should_numa_migrate_memory(struct task_struct *p, struct page * page, if (flags & TNF_WRITE) th *= 2; latency = numa_hint_fault_latency(page); - if (latency > th) + if (latency > th) { + mod_node_page_state(pgdat, PGPROMOTE_COLD_THROTTLE, + hpage_nr_pages(page)); return false; + } if (flags & TNF_DEMOTED) mod_node_page_state(pgdat, PGDEMOTED_HOT, diff --git a/mm/vmstat.c b/mm/vmstat.c index 7f2c498..98f2a03 100644 --- a/mm/vmstat.c +++ b/mm/vmstat.c @@ -1193,6 +1193,7 @@ int fragmentation_index(struct zone *zone, unsigned int order) "pgpromote_file", "pgpromote_try", "pgpromote_demoted", + "pgpromote_cold_throttle", #endif /* enum writeback_stat_item counters */ -- 1.8.3.1

2 years, 1 month

2
9
0 0

[PMEM PATCH 0/3] memory tiered: adjust promotion threshold to

by zhongjiang-ali

Mysql benchmark will produce too mnay promotion due to DRAM node is not enough, but promotion frequently will result in performance decrease. hence, we perfer to remote access rather than demote/promote traffic frequently. Huang Ying (3): mm, migrate: use flags parameter for remove_migration_ptes() memory tiering: measure whether demoted pages are hot memory tiering: adjust promotion threshold based on hot pages demoted include/linux/mmzone.h | 3 ++ include/linux/page-flags.h | 9 ++++++ include/linux/page_ext.h | 3 ++ include/linux/rmap.h | 8 ++++- include/linux/sched/numa_balancing.h | 62 ++++++++++++++++++++++++++++++++++++ include/linux/sched/sysctl.h | 3 ++ include/trace/events/mmflags.h | 8 ++++- kernel/sched/fair.c | 27 +++++++++++++--- kernel/sysctl.c | 16 ++++++++++ mm/huge_memory.c | 6 ++-- mm/mempolicy.c | 2 ++ mm/migrate.c | 60 ++++++++++++++++++++++++++++------ mm/vmstat.c | 1 + 13 files changed, 189 insertions(+), 19 deletions(-) -- 1.8.3.1

2 years, 2 months

4
5
0 0

[PATCH] memory tiered: Do not use thp in pmem node

by zhongjiang-ali

Currently, Mysql testcase show that a large number of thp are migrated from pmem node to toptier node, it will bring in more pgpromote_demoted and migrated failiure. because pmem node memory is marked as prot_none, it will be migrated by cpu access as soon as possible when it is hot, and it is unnesscessary to migrate thp to dram when dram memory is not enough, which will bring in more demoted and promoted. Hence, the patch forbid the thp to produce in pmem node. the result show about 3% improvements. the relative statistics is as follows. before appling patch: mysql prepare: pgpromote_demoted 908267 pgmigrate_fail_dst_node_fail 428223 pgmigrate_fail_numa_isolate_fail 460480 mysql run: pgpromote_demoted 2901105 pgmigrate_fail_dst_node_fail 5653776 pgmigrate_fail_numa_isolate_fail 5686052 after appling patch: mysql prepare: pgpromote_demoted 839297 pgmigrate_fail_dst_node_fail 36585 pgmigrate_fail_numa_isolate_fail 36585 mysql run: pgpromote_demoted 913828 pgmigrate_fail_dst_node_fail 235863 pgmigrate_fail_numa_isolate_fail 235870 Signed-off-by: zhongjiang-ali <zhongjiang-ali(a)linux.alibaba.com> --- mm/page_alloc.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 8cfce92..4fff3cd 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -461,6 +461,17 @@ static __always_inline int get_pfnblock_migratetype(struct page *page, unsigned return __get_pfnblock_flags_mask(page, pfn, PB_migrate_end, MIGRATETYPE_MASK); } +static inline bool allow_hugepage_allocation(int nid, unsigned int order) +{ + if (node_is_toptier(nid)) + return true; + + if (order != HPAGE_PMD_ORDER) + return true; + + return false; +} + /** * set_pfnblock_flags_mask - Set the requested group of flags for a pageblock_nr_pages block of pages * @page: The page within the block of interest @@ -3689,6 +3700,9 @@ static bool zone_allows_reclaim(struct zone *local_zone, struct zone *zone) } } + if (!allow_hugepage_allocation(zone_to_nid(zone), order)) + continue; + if (no_fallback && nr_online_nodes > 1 && zone != ac->preferred_zoneref->zone) { int local_nid; -- 1.8.3.1

2 years, 2 months

4
16
0 0

[Pmem PATCH] memory tiering: fix promote_success count when thp numa fault work

by zhongjiang-ali

Currently, promote_success just include the normal page count when it is migrated from pmem node to toptier node, but an huge page also can trigger the same operation when thp numa fault work. hence it miss the count in migrate_misplaced_transhuge_page. Signed-off-by: zhongjiang-ali <zhongjiang-ali(a)linux.alibaba.com> --- mm/migrate.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/mm/migrate.c b/mm/migrate.c index e9adaa7..9d6cac9 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -2138,7 +2138,7 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma, if (nr_succeeded) { count_vm_numa_events(NUMA_PAGE_MIGRATE, nr_succeeded); if (!node_is_toptier(page_to_nid(page)) && node_is_toptier(node)) - mod_node_page_state(NODE_DATA(node), PGPROMOTE_SUCCESS, + mod_node_page_state(pgdat, PGPROMOTE_SUCCESS, nr_succeeded); } BUG_ON(!list_empty(&migratepages)); @@ -2264,6 +2264,9 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm, count_vm_events(PGMIGRATE_SUCCESS, HPAGE_PMD_NR); count_vm_numa_events(NUMA_PAGE_MIGRATE, HPAGE_PMD_NR); + if (!node_is_toptier(page_to_nid(page)) && node_is_toptier(node)) + mod_node_page_state(pgdat, PGPROMOTE_SUCCESS, + HPAGE_PMD_NR); mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru, -- 1.8.3.1

2 years, 2 months

3
2
0 0

[PATCH] memory tiering: Fix panic when disabling memory tiering mode in runtime

by Huang Ying

It was reported that the following panic may be triggered when NUMA_BALANCING_MEMORY_TIERING is disabled in runtime. BUG: unable to handle kernel paging request at ffffffff9670bae0 PGD 29a620c067 P4D 29a620c067 PUD 29a620d063 PMD 2f3c2ef063 PTE 800fffd6598f4062 Oops: 0000 [#1] SMP PTI CPU: 14 PID: 518 Comm: systemd-journal Kdump: loaded Tainted: G E 4.19.91-001.ali4000_20210617_6287b9d5de_cbp.alios7.x86_64 #1 Hardware name: Alibaba Cloud Alibaba Cloud ECS, BIOS 8c24b4c 04/01/2014 RIP: 0010:should_numa_migrate_memory+0xc3/0x760 Code: ff 00 00 00 0f 84 89 03 00 00 41 0f b6 b5 f0 05 00 00 39 f1 0f 84 79 03 00 00 c1 f8 08 25 ff 01 00 00 48 8b 04 c5 20 37 16 96 <46> 3b 24 30 74 19 48 83 c4 38 31 c0 5b 5d 41 5c 41 5d 41 5e 41 5f RSP: 0000:ffff9ff78d35bd08 EFLAGS: 00010206 RAX: ffffffff966ec000 RBX: ffffc96b10868980 RCX: 000000000000005c RDX: 05169704c2022014 RSI: 0000000000000006 RDI: ffffc96b10868980 RBP: 0000000000000000 R08: ff80003fffffffff R09: ffff93c63ac01d10 R10: 0000000000000002 R11: 0000000000000000 R12: 0000000000000000 R13: ffff93c63af8c740 R14: 000000000001fae0 R15: 0000000000000001 FS: 00007f89d8602880(0000) GS:ffff93c641b80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffffffff9670bae0 CR3: 0000000f77c12006 CR4: 00000000003606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: mpol_misplaced+0x17e/0x200 do_numa_page+0x1e5/0x288 handle_mm_fault+0x91b/0x960 __do_page_fault+0x26b/0x4a0 do_page_fault+0x32/0x110 ? async_page_fault+0x8/0x30 async_page_fault+0x1e/0x30 RIP: 0033:0x558e2f2e21e0 Code: 8b 83 c8 00 00 00 48 8b 48 70 48 c1 e9 04 48 85 c9 0f 84 9c 01 00 00 31 d2 48 89 e8 48 f7 f1 48 c1 e2 04 48 03 93 d0 00 00 00 <4c> 8b 3a 4 Because page "cpupid" field is used to record the CPU and PID when NUMA_BALANCING_MEMORY_TIERING is disabled, while page "cpupid" field is used to record the NUMA balancing page scan time when NUMA_BALANCING_MEMORY_TIERING is enabled. During disabling NUMA_BALANCING_MEMORY_TIERING in runtime, "cpupid" may become invalid so that panic is triggered. This is fixed via checking the validity of "cpupid" before use it. Signed-off-by: "Huang, Ying" <ying.huang(a)intel.com> Reported-and-tested-by: zhong jiang <zhongjiang-ali(a)linux.alibaba.com> --- include/linux/mm.h | 5 +++++ kernel/sched/fair.c | 20 +++++++++++++++++--- 2 files changed, 22 insertions(+), 3 deletions(-) diff --git a/include/linux/mm.h b/include/linux/mm.h index a9ea778eafe0..327fe01b1ae1 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -1440,6 +1440,11 @@ static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); } +static inline bool check_cpupid(int cpupid) +{ + return cpupid_to_cpu(cpupid) < nr_cpu_ids; +} + #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS static inline int page_cpupid_xchg_last(struct page *page, int cpupid) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 7ca4ab033180..85a01906d796 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -1597,6 +1597,14 @@ bool should_numa_migrate_memory(struct task_struct *p, struct page * page, this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid); last_cpupid = page_cpupid_xchg_last(page, this_cpupid); + /* + * The cpupid may be invalid when NUMA_BALANCING_MEMORY_TIERING + * is disabled dynamically. + */ + if (!(sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING) && + !node_is_toptier(src_nid) && !check_cpupid(last_cpupid)) + return false; + /* * Allow first faults or private faults to migrate immediately early in * the lifetime of a task. The magic number 4 is based on waiting for @@ -2812,9 +2820,15 @@ void task_numa_fault(int last_cpupid, int mem_node, int pages, int flags) if (!p->mm) return; - /* Numa faults statistics are unnecessary for the slow memory node */ - if (sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING && - !node_is_toptier(mem_node)) + /* + * NUMA faults statistics are unnecessary for the slow memory node. + * + * And, the cpupid may be invalid when NUMA_BALANCING_MEMORY_TIERING + * is disabled dynamically. + */ + if (!node_is_toptier(mem_node) && + (sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING || + !check_cpupid(last_cpupid))) return; /* Allocate buffer to track faults on a per-node basis */ -- 2.30.2

2 years, 3 months

1
0
0 0

2024

2023

2022

2021

Pmem