From: Muchun Song <hidden> Date: 2021-03-08 10:31:03
Hi everyone,
This patch series will free some vmemmap pages(struct page structures)
associated with each HugeTLB page when preallocated to save memory.
In order to reduce the difficulty of the first version of code review.
From this version, we disable PMD/huge page mapping of vmemmap if this
feature was enabled. This acutely eliminates a bunch of the complex code
doing page table manipulation. When this patch series is solid, we cam add
the code of vmemmap page table manipulation in the future.
The struct page structures (page structs) are used to describe a physical
page frame. By default, there is an one-to-one mapping from a page frame to
it's corresponding page struct.
The HugeTLB pages consist of multiple base page size pages and is supported
by many architectures. See hugetlbpage.rst in the Documentation directory
for more details. On the x86 architecture, HugeTLB pages of size 2MB and 1GB
are currently supported. Since the base page size on x86 is 4KB, a 2MB
HugeTLB page consists of 512 base pages and a 1GB HugeTLB page consists of
4096 base pages. For each base page, there is a corresponding page struct.
Within the HugeTLB subsystem, only the first 4 page structs are used to
contain unique information about a HugeTLB page. HUGETLB_CGROUP_MIN_ORDER
provides this upper limit. The only 'useful' information in the remaining
page structs is the compound_head field, and this field is the same for all
tail pages.
By removing redundant page structs for HugeTLB pages, memory can returned to
the buddy allocator for other uses.
When the system boot up, every 2M HugeTLB has 512 struct page structs which
size is 8 pages(sizeof(struct page) * 512 / PAGE_SIZE).
HugeTLB struct pages(8 pages) page frame(8 pages)
+-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+
| | | 0 | -------------> | 0 |
| | +-----------+ +-----------+
| | | 1 | -------------> | 1 |
| | +-----------+ +-----------+
| | | 2 | -------------> | 2 |
| | +-----------+ +-----------+
| | | 3 | -------------> | 3 |
| | +-----------+ +-----------+
| | | 4 | -------------> | 4 |
| 2MB | +-----------+ +-----------+
| | | 5 | -------------> | 5 |
| | +-----------+ +-----------+
| | | 6 | -------------> | 6 |
| | +-----------+ +-----------+
| | | 7 | -------------> | 7 |
| | +-----------+ +-----------+
| |
| |
| |
+-----------+
The value of page->compound_head is the same for all tail pages. The first
page of page structs (page 0) associated with the HugeTLB page contains the 4
page structs necessary to describe the HugeTLB. The only use of the remaining
pages of page structs (page 1 to page 7) is to point to page->compound_head.
Therefore, we can remap pages 2 to 7 to page 1. Only 2 pages of page structs
will be used for each HugeTLB page. This will allow us to free the remaining
6 pages to the buddy allocator.
Here is how things look after remapping.
HugeTLB struct pages(8 pages) page frame(8 pages)
+-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+
| | | 0 | -------------> | 0 |
| | +-----------+ +-----------+
| | | 1 | -------------> | 1 |
| | +-----------+ +-----------+
| | | 2 | ----------------^ ^ ^ ^ ^ ^
| | +-----------+ | | | | |
| | | 3 | ------------------+ | | | |
| | +-----------+ | | | |
| | | 4 | --------------------+ | | |
| 2MB | +-----------+ | | |
| | | 5 | ----------------------+ | |
| | +-----------+ | |
| | | 6 | ------------------------+ |
| | +-----------+ |
| | | 7 | --------------------------+
| | +-----------+
| |
| |
| |
+-----------+
When a HugeTLB is freed to the buddy system, we should allocate 6 pages for
vmemmap pages and restore the previous mapping relationship.
Apart from 2MB HugeTLB page, we also have 1GB HugeTLB page. It is similar
to the 2MB HugeTLB page. We also can use this approach to free the vmemmap
pages.
In this case, for the 1GB HugeTLB page, we can save 4094 pages. This is a
very substantial gain. On our server, run some SPDK/QEMU applications which
will use 1024GB HugeTLB page. With this feature enabled, we can save ~16GB
(1G hugepage)/~12GB (2MB hugepage) memory.
Because there are vmemmap page tables reconstruction on the freeing/allocating
path, it increases some overhead. Here are some overhead analysis.
1) Allocating 10240 2MB HugeTLB pages.
a) With this patch series applied:
# time echo 10240 > /proc/sys/vm/nr_hugepages
real 0m0.166s
user 0m0.000s
sys 0m0.166s
# bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; }
kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs -
@start[tid]); delete(@start[tid]); }'
Attaching 2 probes...
@latency:
[8K, 16K) 5476 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[16K, 32K) 4760 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ |
[32K, 64K) 4 | |
b) Without this patch series:
# time echo 10240 > /proc/sys/vm/nr_hugepages
real 0m0.067s
user 0m0.000s
sys 0m0.067s
# bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; }
kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs -
@start[tid]); delete(@start[tid]); }'
Attaching 2 probes...
@latency:
[4K, 8K) 10147 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[8K, 16K) 93 | |
Summarize: this feature is about ~2x slower than before.
2) Freeing 10240 2MB HugeTLB pages.
a) With this patch series applied:
# time echo 0 > /proc/sys/vm/nr_hugepages
real 0m0.213s
user 0m0.000s
sys 0m0.213s
# bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; }
kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs -
@start[tid]); delete(@start[tid]); }'
Attaching 2 probes...
@latency:
[8K, 16K) 6 | |
[16K, 32K) 10227 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[32K, 64K) 7 | |
b) Without this patch series:
# time echo 0 > /proc/sys/vm/nr_hugepages
real 0m0.081s
user 0m0.000s
sys 0m0.081s
# bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; }
kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs -
@start[tid]); delete(@start[tid]); }'
Attaching 2 probes...
@latency:
[4K, 8K) 6805 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[8K, 16K) 3427 |@@@@@@@@@@@@@@@@@@@@@@@@@@ |
[16K, 32K) 8 | |
Summarize: The overhead of __free_hugepage is about ~2-3x slower than before.
Although the overhead has increased, the overhead is not significant. Like Mike
said, "However, remember that the majority of use cases create HugeTLB pages at
or shortly after boot time and add them to the pool. So, additional overhead is
at pool creation time. There is no change to 'normal run time' operations of
getting a page from or returning a page to the pool (think page fault/unmap)".
Despite the overhead and in addition to the memory gains from this series. The
following data is obtained by Joao Martins. Very thanks to his effort.
There's an additional benefit which is page (un)pinners will see an improvement
and Joao presumes because there are fewer memmap pages and thus the tail/head
pages are staying in cache more often.
Out of the box Joao saw (when comparing linux-next against linux-next + this series)
with gup_test and pinning a 16G HugeTLB file (with 1G pages):
get_user_pages(): ~32k -> ~9k
unpin_user_pages(): ~75k -> ~70k
Usually any tight loop fetching compound_head(), or reading tail pages data (e.g.
compound_head) benefit a lot. There's some unpinning inefficiencies Joao was
fixing[0], but with that in added it shows even more:
unpin_user_pages(): ~27k -> ~3.8k
[0] https://lore.kernel.org/linux-mm/20210204202500.26474-1-joao.m.martins@oracle.com/
Todo:
- Free all of the tail vmemmap pages
Now for the 2MB HugrTLB page, we only free 6 vmemmap pages. we really can
free 7 vmemmap pages. In this case, we can see 8 of the 512 struct page
structures has beed set PG_head flag. If we can adjust compound_head()
slightly and make compound_head() return the real head struct page when
the parameter is the tail struct page but with PG_head flag set.
In order to make the code evolution route clearer. This feature can can be
a separate patch after this patchset is solid.
- Support for other architectures (e.g. aarch64).
- Enable PMD/huge page mapping of vmemmap even if this feature was enabled.
Changelog in v17 -> v18:
- Add complete copyright to bootmem_info.c (Suggested by Balbir).
- Fix some issues (in patch #4) suggested by Mike.
Thanks to Balbir and Mike's review. Also thanks to Chen Huang and
Bodeddula Balasubramaniam's test.
Changelog in v16 -> v17:
- Fix issues suggested by Mike and Oscar.
- Update commit log suggested by Michal.
Thanks to Mike, David H and Michal's suggestions and review.
Changelog in v15 -> v16:
- Use GFP_KERNEL to allocate vmemmap pages.
Thanks to Mike, David H and Michal's suggestions.
Changelog in v14 -> v15:
- Fix some issues suggested by Oscar. Thanks to Oscar.
- Add numbers which Joao Martins tested to cover letter. Thanks to his effort.
Changelog in v13 -> v14:
- Refuse to free the HugeTLB page when the system is under memory pressure.
- Use GFP_ATOMIC to allocate vmemmap pages instead of GFP_KERNEL.
- Rebase to linux-next 20210202.
- Fix and add some comments for vmemmap_remap_free().
Thanks to Oscar, Mike, David H and David R's suggestions and review.
Changelog in v12 -> v13:
- Remove VM_WARN_ON_PAGE macro.
- Add more comments in vmemmap_pte_range() and vmemmap_remap_free().
Thanks to Oscar and Mike's suggestions and review.
Changelog in v11 -> v12:
- Move VM_WARN_ON_PAGE to a separate patch.
- Call __free_hugepage() with hugetlb_lock (See patch #5.) to serialize
with dissolve_free_huge_page(). It is to prepare for patch #9.
- Introduce PageHugeInflight. See patch #9.
Changelog in v10 -> v11:
- Fix compiler error when !CONFIG_HUGETLB_PAGE_FREE_VMEMMAP.
- Rework some comments and commit changes.
- Rework vmemmap_remap_free() to 3 parameters.
Thanks to Oscar and Mike's suggestions and review.
Changelog in v9 -> v10:
- Fix a bug in patch #11. Thanks to Oscar for pointing that out.
- Rework some commit log or comments. Thanks Mike and Oscar for the suggestions.
- Drop VMEMMAP_TAIL_PAGE_REUSE in the patch #3.
Thank you very much Mike and Oscar for reviewing the code.
Changelog in v8 -> v9:
- Rework some code. Very thanks to Oscar.
- Put all the non-hugetlb vmemmap functions under sparsemem-vmemmap.c.
Changelog in v7 -> v8:
- Adjust the order of patches.
Very thanks to David and Oscar. Your suggestions are very valuable.
Changelog in v6 -> v7:
- Rebase to linux-next 20201130
- Do not use basepage mapping for vmemmap when this feature is disabled.
- Rework some patchs.
[PATCH v6 08/16] mm/hugetlb: Free the vmemmap pages associated with each hugetlb page
[PATCH v6 10/16] mm/hugetlb: Allocate the vmemmap pages associated with each hugetlb page
Thanks to Oscar and Barry.
Changelog in v5 -> v6:
- Disable PMD/huge page mapping of vmemmap if this feature was enabled.
- Simplify the first version code.
Changelog in v4 -> v5:
- Rework somme comments and code in the [PATCH v4 04/21] and [PATCH v4 05/21].
Thanks to Mike and Oscar's suggestions.
Changelog in v3 -> v4:
- Move all the vmemmap functions to hugetlb_vmemmap.c.
- Make the CONFIG_HUGETLB_PAGE_FREE_VMEMMAP default to y, if we want to
disable this feature, we should disable it by a boot/kernel command line.
- Remove vmemmap_pgtable_{init, deposit, withdraw}() helper functions.
- Initialize page table lock for vmemmap through core_initcall mechanism.
Thanks for Mike and Oscar's suggestions.
Changelog in v2 -> v3:
- Rename some helps function name. Thanks Mike.
- Rework some code. Thanks Mike and Oscar.
- Remap the tail vmemmap page with PAGE_KERNEL_RO instead of PAGE_KERNEL.
Thanks Matthew.
- Add some overhead analysis in the cover letter.
- Use vmemap pmd table lock instead of a hugetlb specific global lock.
Changelog in v1 -> v2:
- Fix do not call dissolve_compound_page in alloc_huge_page_vmemmap().
- Fix some typo and code style problems.
- Remove unused handle_vmemmap_fault().
- Merge some commits to one commit suggested by Mike.
Muchun Song (9):
mm: memory_hotplug: factor out bootmem core functions to
bootmem_info.c
mm: hugetlb: introduce a new config HUGETLB_PAGE_FREE_VMEMMAP
mm: hugetlb: free the vmemmap pages associated with each HugeTLB page
mm: hugetlb: alloc the vmemmap pages associated with each HugeTLB page
mm: hugetlb: set the PageHWPoison to the raw error page
mm: hugetlb: add a kernel parameter hugetlb_free_vmemmap
mm: hugetlb: introduce nr_free_vmemmap_pages in the struct hstate
mm: hugetlb: gather discrete indexes of tail page
mm: hugetlb: optimize the code with the help of the compiler
Documentation/admin-guide/kernel-parameters.txt | 14 ++
Documentation/admin-guide/mm/hugetlbpage.rst | 11 +
arch/x86/mm/init_64.c | 13 +-
fs/Kconfig | 6 +
include/linux/bootmem_info.h | 65 ++++++
include/linux/hugetlb.h | 47 +++-
include/linux/hugetlb_cgroup.h | 19 +-
include/linux/memory_hotplug.h | 27 ---
include/linux/mm.h | 5 +
mm/Makefile | 2 +
mm/bootmem_info.c | 127 ++++++++++
mm/hugetlb.c | 176 +++++++++++---
mm/hugetlb_vmemmap.c | 293 ++++++++++++++++++++++++
mm/hugetlb_vmemmap.h | 51 +++++
mm/memory_hotplug.c | 116 ----------
mm/sparse-vmemmap.c | 280 ++++++++++++++++++++++
mm/sparse.c | 1 +
17 files changed, 1065 insertions(+), 188 deletions(-)
create mode 100644 include/linux/bootmem_info.h
create mode 100644 mm/bootmem_info.c
create mode 100644 mm/hugetlb_vmemmap.c
create mode 100644 mm/hugetlb_vmemmap.h
--
2.11.0
From: Muchun Song <hidden> Date: 2021-03-08 10:31:36
Move bootmem info registration common API to individual bootmem_info.c.
And we will use {get,put}_page_bootmem() to initialize the page for the
vmemmap pages or free the vmemmap pages to buddy in the later patch.
So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code
movement without any functional change.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
arch/x86/mm/init_64.c | 3 +-
include/linux/bootmem_info.h | 40 +++++++++++++
include/linux/memory_hotplug.h | 27 ---------
mm/Makefile | 1 +
mm/bootmem_info.c | 127 +++++++++++++++++++++++++++++++++++++++++
mm/memory_hotplug.c | 116 -------------------------------------
mm/sparse.c | 1 +
7 files changed, 171 insertions(+), 144 deletions(-)
create mode 100644 include/linux/bootmem_info.h
create mode 100644 mm/bootmem_info.c
@@ -1571,7 +1572,7 @@ int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,returnerr;}-#if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) && defined(CONFIG_HAVE_BOOTMEM_INFO_NODE)+#ifdef CONFIG_HAVE_BOOTMEM_INFO_NODEvoidregister_page_bootmem_memmap(unsignedlongsection_nr,structpage*start_page,unsignedlongnr_pages){
@@ -18,18 +18,6 @@ struct vmem_altmap;#ifdef CONFIG_MEMORY_HOTPLUGstructpage*pfn_to_online_page(unsignedlongpfn);-/*-*Typesforfreebootmemstoredinpage->lru.next.Thesehavetobein-*somerandomrangeinunsignedlongspacefordebuggingpurposes.-*/-enum{-MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE=12,-SECTION_INFO=MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE,-MIX_SECTION_INFO,-NODE_INFO,-MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE=NODE_INFO,-};-/* Types for control the zone type of onlined and offlined memory */enum{/* Offline the memory. */
From: Muchun Song <hidden> Date: 2021-03-08 10:31:36
Every HugeTLB has more than one struct page structure. We __know__ that
we only use the first 4(HUGETLB_CGROUP_MIN_ORDER) struct page structures
to store metadata associated with each HugeTLB.
There are a lot of struct page structures associated with each HugeTLB
page. For tail pages, the value of compound_head is the same. So we can
reuse first page of tail page structures. We map the virtual addresses
of the remaining pages of tail page structures to the first tail page
struct, and then free these page frames. Therefore, we need to reserve
two pages as vmemmap areas.
When we allocate a HugeTLB page from the buddy, we can free some vmemmap
pages associated with each HugeTLB page. It is more appropriate to do it
in the prep_new_huge_page().
The free_vmemmap_pages_per_hpage(), which indicates how many vmemmap
pages associated with a HugeTLB page can be freed, returns zero for
now, which means the feature is disabled. We will enable it once all
the infrastructure is there.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
include/linux/bootmem_info.h | 27 +++++-
include/linux/mm.h | 3 +
mm/Makefile | 1 +
mm/hugetlb.c | 3 +
mm/hugetlb_vmemmap.c | 219 +++++++++++++++++++++++++++++++++++++++++++
mm/hugetlb_vmemmap.h | 20 ++++
mm/sparse-vmemmap.c | 207 ++++++++++++++++++++++++++++++++++++++++
7 files changed, 479 insertions(+), 1 deletion(-)
create mode 100644 mm/hugetlb_vmemmap.c
create mode 100644 mm/hugetlb_vmemmap.h
@@ -27,8 +27,215 @@#include<linux/spinlock.h>#include<linux/vmalloc.h>#include<linux/sched.h>+#include<linux/pgtable.h>+#include<linux/bootmem_info.h>+#include<asm/dma.h>#include<asm/pgalloc.h>+#include<asm/tlbflush.h>++/**+*vmemmap_remap_walk-walkvmemmappagetable+*+*@remap_pte:calledforeachlowest-levelentry(PTE).+*@reuse_page:thepagewhichisreusedforthetailvmemmappages.+*@reuse_addr:thevirtualaddressofthe@reuse_pagepage.+*@vmemmap_pages:thelistheadofthevmemmappagesthatcanbefreed.+*/+structvmemmap_remap_walk{+void(*remap_pte)(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk);+structpage*reuse_page;+unsignedlongreuse_addr;+structlist_head*vmemmap_pages;+};++staticvoidvmemmap_pte_range(pmd_t*pmd,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pte_t*pte;++pte=pte_offset_kernel(pmd,addr);++/*+*Thereuse_pageisfound'first'intablewalkbeforewestart+*remapping(whichiscalling@walk->remap_pte).+*/+if(!walk->reuse_page){+BUG_ON(pte_none(*pte));+BUG_ON(walk->reuse_addr!=addr);++walk->reuse_page=pte_page(*pte++);+/*+*Becausethereuseaddressispartoftherangethatweare+*walking,skipthereuseaddressrange.+*/+addr+=PAGE_SIZE;+}++for(;addr!=end;addr+=PAGE_SIZE,pte++){+BUG_ON(pte_none(*pte));++walk->remap_pte(pte,addr,walk);+}+}++staticvoidvmemmap_pmd_range(pud_t*pud,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pmd_t*pmd;+unsignedlongnext;++pmd=pmd_offset(pud,addr);+do{+BUG_ON(pmd_none(*pmd)||pmd_leaf(*pmd));++next=pmd_addr_end(addr,end);+vmemmap_pte_range(pmd,addr,next,walk);+}while(pmd++,addr=next,addr!=end);+}++staticvoidvmemmap_pud_range(p4d_t*p4d,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pud_t*pud;+unsignedlongnext;++pud=pud_offset(p4d,addr);+do{+BUG_ON(pud_none(*pud));++next=pud_addr_end(addr,end);+vmemmap_pmd_range(pud,addr,next,walk);+}while(pud++,addr=next,addr!=end);+}++staticvoidvmemmap_p4d_range(pgd_t*pgd,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+p4d_t*p4d;+unsignedlongnext;++p4d=p4d_offset(pgd,addr);+do{+BUG_ON(p4d_none(*p4d));++next=p4d_addr_end(addr,end);+vmemmap_pud_range(p4d,addr,next,walk);+}while(p4d++,addr=next,addr!=end);+}++staticvoidvmemmap_remap_range(unsignedlongstart,unsignedlongend,+structvmemmap_remap_walk*walk)+{+unsignedlongaddr=start;+unsignedlongnext;+pgd_t*pgd;++VM_BUG_ON(!IS_ALIGNED(start,PAGE_SIZE));+VM_BUG_ON(!IS_ALIGNED(end,PAGE_SIZE));++pgd=pgd_offset_k(addr);+do{+BUG_ON(pgd_none(*pgd));++next=pgd_addr_end(addr,end);+vmemmap_p4d_range(pgd,addr,next,walk);+}while(pgd++,addr=next,addr!=end);++/*+*Weonlychangethemappingofthevmemmapvirtualaddressrange+*[@start+PAGE_SIZE,end),soweonlyneedtoflushtheTLBwhich+*belongstotherange.+*/+flush_tlb_kernel_range(start+PAGE_SIZE,end);+}++/*+*Freeavmemmappage.Avmemmappagecanbeallocatedfromthememblock+*allocatororbuddyallocator.IfthePG_reservedflagisset,itmeans+*thatitallocatedfromthememblockallocator,justfreeitviathe+*free_bootmem_page().Otherwise,use__free_page().+*/+staticinlinevoidfree_vmemmap_page(structpage*page)+{+if(PageReserved(page))+free_bootmem_page(page);+else+__free_page(page);+}++/* Free a list of the vmemmap pages */+staticvoidfree_vmemmap_page_list(structlist_head*list)+{+structpage*page,*next;++list_for_each_entry_safe(page,next,list,lru){+list_del(&page->lru);+free_vmemmap_page(page);+}+}++staticvoidvmemmap_remap_pte(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk)+{+/*+*Remapthetailpagesasread-onlytocatchillegalwriteoperation+*tothetailpages.+*/+pgprot_tpgprot=PAGE_KERNEL_RO;+pte_tentry=mk_pte(walk->reuse_page,pgprot);+structpage*page=pte_page(*pte);++list_add(&page->lru,walk->vmemmap_pages);+set_pte_at(&init_mm,addr,pte,entry);+}++/**+*vmemmap_remap_free-remapthevmemmapvirtualaddressrange[@start,@end)+*tothepagewhich@reuseismappedto,thenfreevmemmap+*whichtherangearemappedto.+*@start:startaddressofthevmemmapvirtualaddressrangethatwewant+*toremap.+*@end:endaddressofthevmemmapvirtualaddressrangethatwewantto+*remap.+*@reuse:reuseaddress.+*+*Note:Thisfunctiondependsonvmemmapbeingbasepagemapped.Pleasemake+*surethatwedisablePMDmappingofvmemmappageswhencallingthisfunction.+*/+voidvmemmap_remap_free(unsignedlongstart,unsignedlongend,+unsignedlongreuse)+{+LIST_HEAD(vmemmap_pages);+structvmemmap_remap_walkwalk={+.remap_pte=vmemmap_remap_pte,+.reuse_addr=reuse,+.vmemmap_pages=&vmemmap_pages,+};++/*+*Inordertomakeremappingroutinemostefficientforthehugepages,+*theroutineofvmemmappagetablewalkinghasthefollowingrules+*(seemoredetailsfromthevmemmap_pte_range()):+*+*-Therange[@start,@end)andtherange[@reuse,@reuse+PAGE_SIZE)+*shouldbecontinuous.+*-The@reuseaddressispartoftherange[@reuse,@end)thatweare+*walkingwhichispassedtovmemmap_remap_range().+*-The@reuseaddressisthefirstinthecompleterange.+*+*Soweneedtomakesurethat@startand@reusemeettheaboverules.+*/+BUG_ON(start-reuse!=PAGE_SIZE);++vmemmap_remap_range(reuse,end,&walk);+free_vmemmap_page_list(&vmemmap_pages);+}/**Allocateablockofmemorytobeusedtobackthevirtualmemorymap
From: Muchun Song <hidden> Date: 2021-03-08 10:31:37
The option HUGETLB_PAGE_FREE_VMEMMAP allows for the freeing of
some vmemmap pages associated with pre-allocated HugeTLB pages.
For example, on X86_64 6 vmemmap pages of size 4KB each can be
saved for each 2MB HugeTLB page. 4094 vmemmap pages of size 4KB
each can be saved for each 1GB HugeTLB page.
When a HugeTLB page is allocated or freed, the vmemmap array
representing the range associated with the page will need to be
remapped. When a page is allocated, vmemmap pages are freed
after remapping. When a page is freed, previously discarded
vmemmap pages must be allocated before remapping.
The config option is introduced early so that supporting code
can be written to depend on the option. The initial version of
the code only provides support for x86-64.
Like other code which frees vmemmap, this config option depends on
HAVE_BOOTMEM_INFO_NODE. The routine register_page_bootmem_info() is
used to register bootmem info. Therefore, make sure
register_page_bootmem_info is enabled if HUGETLB_PAGE_FREE_VMEMMAP
is defined.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
---
arch/x86/mm/init_64.c | 2 +-
fs/Kconfig | 6 ++++++
2 files changed, 7 insertions(+), 1 deletion(-)
From: Muchun Song <hidden> Date: 2021-03-08 10:32:08
Because we reuse the first tail vmemmap page frame and remap it
with read-only, we cannot set the PageHWPosion on some tail pages.
So we can use the head[4].private (There are at least 128 struct
page structures associated with the optimized HugeTLB page, so
using head[4].private is safe) to record the real error page index
and set the raw error page PageHWPoison later.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
mm/hugetlb.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++++++++------
1 file changed, 72 insertions(+), 8 deletions(-)
From: Muchun Song <hidden> Date: 2021-03-08 10:32:08
When we free a HugeTLB page to the buddy allocator, we need to allocate
the vmemmap pages associated with it. However, we may not be able to
allocate the vmemmap pages when the system is under memory pressure. In
this case, we just refuse to free the HugeTLB page. This changes behavior
in some corner cases as listed below:
1) Failing to free a huge page triggered by the user (decrease nr_pages).
User needs to try again later.
2) Failing to free a surplus huge page when freed by the application.
Try again later when freeing a huge page next time.
3) Failing to dissolve a free huge page on ZONE_MOVABLE via
offline_pages().
This can happen when we have plenty of ZONE_MOVABLE memory, but
not enough kernel memory to allocate vmemmmap pages. We may even
be able to migrate huge page contents, but will not be able to
dissolve the source huge page. This will prevent an offline
operation and is unfortunate as memory offlining is expected to
succeed on movable zones. Users that depend on memory hotplug
to succeed for movable zones should carefully consider whether the
memory savings gained from this feature are worth the risk of
possibly not being able to offline memory in certain situations.
4) Failing to dissolve a huge page on CMA/ZONE_MOVABLE via
alloc_contig_range() - once we have that handling in place. Mainly
affects CMA and virtio-mem.
Similar to 3). virito-mem will handle migration errors gracefully.
CMA might be able to fallback on other free areas within the CMA
region.
Vmemmap pages are allocated from the page freeing context. In order for
those allocations to be not disruptive (e.g. trigger oom killer)
__GFP_NORETRY is used. hugetlb_lock is dropped for the allocation
because a non sleeping allocation would be too fragile and it could fail
too easily under memory pressure. GFP_ATOMIC or other modes to access
memory reserves is not used because we want to prevent consuming
reserves under heavy hugetlb freeing.
Signed-off-by: Muchun Song <redacted>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
Documentation/admin-guide/mm/hugetlbpage.rst | 8 +++
include/linux/mm.h | 2 +
mm/hugetlb.c | 92 +++++++++++++++++++++-------
mm/hugetlb_vmemmap.c | 32 ++++++----
mm/hugetlb_vmemmap.h | 23 +++++++
mm/sparse-vmemmap.c | 75 ++++++++++++++++++++++-
6 files changed, 197 insertions(+), 35 deletions(-)
@@ -60,6 +60,10 @@ HugePages_Surp the pool above the value in ``/proc/sys/vm/nr_hugepages``. The maximum number of surplus huge pages is controlled by``/proc/sys/vm/nr_overcommit_hugepages``.+ Note: When the feature of freeing unused vmemmap pages associated+ with each hugetlb page is enabled, the number of surplus huge pages+ may be temporarily larger than the maximum number of surplus huge+ pages when the system is under memory pressure. Hugepagesize is the default hugepage size (in Kb). Hugetlb
@@ -80,6 +84,10 @@ returned to the huge page pool when freed by a task. A user with root privileges can dynamically allocate more or free some persistent huge pages by increasing or decreasing the value of ``nr_hugepages``.+Note: When the feature of freeing unused vmemmap pages associated with each+hugetlb page is enabled, we can fail to free the huge pages triggered by+the user when ths system is under memory pressure. Please try again later.+ Pages that are used as huge pages are reserved inside the kernel and cannot be used for other purposes. Huge pages cannot be swapped out under memory pressure.
@@ -1404,9 +1443,9 @@ static void __free_huge_page(struct page *page)}elseif(h->surplus_huge_pages_node[nid]){/* remove the page from active list */list_del(&page->lru);-update_and_free_page(h,page);h->surplus_huge_pages--;h->surplus_huge_pages_node[nid]--;+update_and_free_page(h,page);}else{arch_clear_hugepage_flags(page);enqueue_huge_page(h,page);
@@ -237,6 +238,78 @@ void vmemmap_remap_free(unsigned long start, unsigned long end,free_vmemmap_page_list(&vmemmap_pages);}+staticvoidvmemmap_restore_pte(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk)+{+pgprot_tpgprot=PAGE_KERNEL;+structpage*page;+void*to;++BUG_ON(pte_page(*pte)!=walk->reuse_page);++page=list_first_entry(walk->vmemmap_pages,structpage,lru);+list_del(&page->lru);+to=page_to_virt(page);+copy_page(to,(void*)walk->reuse_addr);++set_pte_at(&init_mm,addr,pte,mk_pte(page,pgprot));+}++staticintalloc_vmemmap_page_list(unsignedlongstart,unsignedlongend,+gfp_tgfp_mask,structlist_head*list)+{+unsignedlongnr_pages=(end-start)>>PAGE_SHIFT;+intnid=page_to_nid((structpage*)start);+structpage*page,*next;++while(nr_pages--){+page=alloc_pages_node(nid,gfp_mask,0);+if(!page)+gotoout;+list_add_tail(&page->lru,list);+}++return0;+out:+list_for_each_entry_safe(page,next,list,lru)+__free_pages(page,0);+return-ENOMEM;+}++/**+*vmemmap_remap_alloc-remapthevmemmapvirtualaddressrange[@start,end)+*tothepagewhichisfromthe@vmemmap_pages+*respectively.+*@start:startaddressofthevmemmapvirtualaddressrangethatwewant+*toremap.+*@end:endaddressofthevmemmapvirtualaddressrangethatwewantto+*remap.+*@reuse:reuseaddress.+*@gpf_mask:GFPflagforallocatingvmemmappages.+*/+intvmemmap_remap_alloc(unsignedlongstart,unsignedlongend,+unsignedlongreuse,gfp_tgfp_mask)+{+LIST_HEAD(vmemmap_pages);+structvmemmap_remap_walkwalk={+.remap_pte=vmemmap_restore_pte,+.reuse_addr=reuse,+.vmemmap_pages=&vmemmap_pages,+};++/* See the comment in the vmemmap_remap_free(). */+BUG_ON(start-reuse!=PAGE_SIZE);++might_sleep_if(gfpflags_allow_blocking(gfp_mask));++if(alloc_vmemmap_page_list(start,end,gfp_mask,&vmemmap_pages))+return-ENOMEM;++vmemmap_remap_range(reuse,end,&walk);++return0;+}+/**Allocateablockofmemorytobeusedtobackthevirtualmemorymap*ortobackthepagetablesthatareusedtocreatethemapping.
From: Muchun Song <hidden> Date: 2021-03-08 10:32:09
Add a kernel parameter hugetlb_free_vmemmap to enable the feature of
freeing unused vmemmap pages associated with each hugetlb page on boot.
We disables PMD mapping of vmemmap pages for x86-64 arch when this
feature is enabled. Because vmemmap_remap_free() depends on vmemmap
being base page mapped.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Barry Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
Documentation/admin-guide/kernel-parameters.txt | 14 ++++++++++++++
Documentation/admin-guide/mm/hugetlbpage.rst | 3 +++
arch/x86/mm/init_64.c | 8 ++++++--
include/linux/hugetlb.h | 19 +++++++++++++++++++
mm/hugetlb_vmemmap.c | 24 ++++++++++++++++++++++++
5 files changed, 66 insertions(+), 2 deletions(-)
@@ -1557,6 +1557,20 @@ Documentation/admin-guide/mm/hugetlbpage.rst. Format: size[KMG]+ hugetlb_free_vmemmap=+ [KNL] When CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is set,+ this controls freeing unused vmemmap pages associated+ with each HugeTLB page. When this option is enabled,+ we disable PMD/huge page mapping of vmemmap pages which+ increase page table pages. So if a user/sysadmin only+ uses a small number of HugeTLB pages (as a percentage+ of system memory), they could end up using more memory+ with hugetlb_free_vmemmap on as opposed to off.+ Format: { on | off (default) }++ on: enable the feature+ off: disable the feature+ hung_task_panic= [KNL] Should the hung task detector generate panics. Format: 0 | 1
@@ -153,6 +153,9 @@ default_hugepagesz will all result in 256 2M huge pages being allocated. Valid default huge page size is architecture dependent.+hugetlb_free_vmemmap+ When CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is set, this enables freeing+ unused vmemmap pages associated with each HugeTLB page. When multiple huge page sizes are supported, ``/proc/sys/vm/nr_hugepages`` indicates the current number of pre-allocated huge pages of the default size.
@@ -1557,7 +1558,8 @@ int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,{interr;-if(end-start<PAGES_PER_SECTION*sizeof(structpage))+if((is_hugetlb_free_vmemmap_enabled()&&!altmap)||+end-start<PAGES_PER_SECTION*sizeof(structpage))err=vmemmap_populate_basepages(start,end,node,NULL);elseif(boot_cpu_has(X86_FEATURE_PSE))err=vmemmap_populate_hugepages(start,end,node,altmap);
@@ -1585,6 +1587,8 @@ void register_page_bootmem_memmap(unsigned long section_nr,pmd_t*pmd;unsignedintnr_pmd_pages;structpage*page;+boolbase_mapping=!boot_cpu_has(X86_FEATURE_PSE)||+is_hugetlb_free_vmemmap_enabled();for(;addr<end;addr=next){pte_t*pte=NULL;
@@ -1610,7 +1614,7 @@ void register_page_bootmem_memmap(unsigned long section_nr,}get_page_bootmem(section_nr,pud_page(*pud),MIX_SECTION_INFO);-if(!boot_cpu_has(X86_FEATURE_PSE)){+if(base_mapping){next=(addr+PAGE_SIZE)&PAGE_MASK;pmd=pmd_offset(pud,addr);if(pmd_none(*pmd))
From: Muchun Song <hidden> Date: 2021-03-08 10:32:40
All the infrastructure is ready, so we introduce nr_free_vmemmap_pages
field in the hstate to indicate how many vmemmap pages associated with
a HugeTLB page that can be freed to buddy allocator. And initialize it
in the hugetlb_vmemmap_init(). This patch is actual enablement of the
feature.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
include/linux/hugetlb.h | 3 +++
mm/hugetlb.c | 1 +
mm/hugetlb_vmemmap.c | 25 +++++++++++++++++++++++++
mm/hugetlb_vmemmap.h | 10 ++++++----
4 files changed, 35 insertions(+), 4 deletions(-)
From: Muchun Song <hidden> Date: 2021-03-08 10:32:40
For HugeTLB page, there are more metadata to save in the struct page.
But the head struct page cannot meet our needs, so we have to abuse
other tail struct page to store the metadata. In order to avoid
conflicts caused by subsequent use of more tail struct pages, we can
gather these discrete indexes of tail struct page. In this case, it
will be easier to add a new tail page index later.
There are only (RESERVE_VMEMMAP_SIZE / sizeof(struct page)) struct
page structs that can be used when CONFIG_HUGETLB_PAGE_FREE_VMEMMAP,
so add a BUILD_BUG_ON to catch invalid usage of the tail struct page.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
include/linux/hugetlb.h | 24 ++++++++++++++++++++++--
include/linux/hugetlb_cgroup.h | 19 +++++++++++--------
mm/hugetlb.c | 6 +++---
mm/hugetlb_vmemmap.c | 8 ++++++++
4 files changed, 44 insertions(+), 13 deletions(-)
From: Muchun Song <hidden> Date: 2021-03-08 10:33:11
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
include/linux/hugetlb.h | 3 ++-
mm/hugetlb_vmemmap.c | 7 +++++++
mm/hugetlb_vmemmap.h | 6 ++++++
3 files changed, 15 insertions(+), 1 deletion(-)
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 14:15:33
[I am sorry for a late review]
On Mon 08-03-21 18:27:59, Muchun Song wrote:
Move bootmem info registration common API to individual bootmem_info.c.
And we will use {get,put}_page_bootmem() to initialize the page for the
vmemmap pages or free the vmemmap pages to buddy in the later patch.
So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code
movement without any functional change.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Separation from memory_hotplug.c is definitely a right step. I am
wondering about the config dependency though
[...]
I would have expected this would depend on CONFIG_SPARSE.
BOOTMEM_INFO_NODE is really an odd thing to depend on here. There is
some functionality which requires the node info but that can be gated
specifically. Or what is the thinking behind?
This doesn't matter right now because it seems that the *_page_bootmem
is only used by x86 outside of the memory hotplug.
Other than that looks good to me.
--
Michal Hocko
SUSE Labs
From: Oscar Salvador <osalvador@suse.de> Date: 2021-03-10 14:21:59
On Mon, Mar 08, 2021 at 06:28:02PM +0800, Muchun Song wrote:
When we free a HugeTLB page to the buddy allocator, we need to allocate
the vmemmap pages associated with it. However, we may not be able to
allocate the vmemmap pages when the system is under memory pressure. In
this case, we just refuse to free the HugeTLB page. This changes behavior
in some corner cases as listed below:
1) Failing to free a huge page triggered by the user (decrease nr_pages).
User needs to try again later.
2) Failing to free a surplus huge page when freed by the application.
Try again later when freeing a huge page next time.
3) Failing to dissolve a free huge page on ZONE_MOVABLE via
offline_pages().
This can happen when we have plenty of ZONE_MOVABLE memory, but
not enough kernel memory to allocate vmemmmap pages. We may even
be able to migrate huge page contents, but will not be able to
dissolve the source huge page. This will prevent an offline
operation and is unfortunate as memory offlining is expected to
succeed on movable zones. Users that depend on memory hotplug
to succeed for movable zones should carefully consider whether the
memory savings gained from this feature are worth the risk of
possibly not being able to offline memory in certain situations.
This is nice to have it here, but a normal user won't dig in the kernel to
figure this out, so my question is: Do we have this documented somewhere under
Documentation/?
If not, could we document it there? It is nice to warn about this things were
sysadmins can find them.
4) Failing to dissolve a huge page on CMA/ZONE_MOVABLE via
alloc_contig_range() - once we have that handling in place. Mainly
affects CMA and virtio-mem.
Similar to 3). virito-mem will handle migration errors gracefully.
CMA might be able to fallback on other free areas within the CMA
region.
Vmemmap pages are allocated from the page freeing context. In order for
those allocations to be not disruptive (e.g. trigger oom killer)
__GFP_NORETRY is used. hugetlb_lock is dropped for the allocation
because a non sleeping allocation would be too fragile and it could fail
too easily under memory pressure. GFP_ATOMIC or other modes to access
memory reserves is not used because we want to prevent consuming
reserves under heavy hugetlb freeing.
Signed-off-by: Muchun Song <redacted>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Sorry for jumping in late.
It looks good to me:
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Minor request above and below:
[...]
Could we place a brief comment about what we expect to return here?
-static inline unsigned long free_vmemmap_pages_size_per_hpage(struct hstate *h)
+int alloc_huge_page_vmemmap(struct hstate *h, struct page *head)
{
- return (unsigned long)free_vmemmap_pages_per_hpage(h) << PAGE_SHIFT;
+ unsigned long vmemmap_addr = (unsigned long)head;
+ unsigned long vmemmap_end, vmemmap_reuse;
+
+ if (!free_vmemmap_pages_per_hpage(h))
+ return 0;
+
+ vmemmap_addr += RESERVE_VMEMMAP_SIZE;
+ vmemmap_end = vmemmap_addr + free_vmemmap_pages_size_per_hpage(h);
+ vmemmap_reuse = vmemmap_addr - PAGE_SIZE;
+ /*
+ * The pages which the vmemmap virtual address range [@vmemmap_addr,
+ * @vmemmap_end) are mapped to are freed to the buddy allocator, and
+ * the range is mapped to the page which @vmemmap_reuse is mapped to.
+ * When a HugeTLB page is freed to the buddy allocator, previously
+ * discarded vmemmap pages must be allocated and remapping.
+ */
+ return vmemmap_remap_alloc(vmemmap_addr, vmemmap_end, vmemmap_reuse,
+ GFP_KERNEL | __GFP_NORETRY | __GFP_THISNODE);
}
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 14:33:15
On Mon 08-03-21 18:28:01, Muchun Song wrote:
Every HugeTLB has more than one struct page structure. We __know__ that
we only use the first 4(HUGETLB_CGROUP_MIN_ORDER) struct page structures
to store metadata associated with each HugeTLB.
I think it would be great to make this explicit somewhere around the
code which uses those struct pages.
There are a lot of struct page structures associated with each HugeTLB
page. For tail pages, the value of compound_head is the same. So we can
reuse first page of tail page structures. We map the virtual addresses
of the remaining pages of tail page structures to the first tail page
struct, and then free these page frames. Therefore, we need to reserve
two pages as vmemmap areas.
When we allocate a HugeTLB page from the buddy, we can free some vmemmap
pages associated with each HugeTLB page. It is more appropriate to do it
in the prep_new_huge_page().
The free_vmemmap_pages_per_hpage(), which indicates how many vmemmap
pages associated with a HugeTLB page can be freed, returns zero for
now, which means the feature is disabled. We will enable it once all
the infrastructure is there.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
I do not see any issues here. I just want to point out that the amount
of *BUG_ON is quite high to my taste. Most of them seem to be added just
in case if something goes wrong or should never happen. These are
usually bad reasons to add them IMHO. I would just drop those unless
there is a very good reason to keep them around.
I really appreciate how you made a high level design documentation to
the source code directly. Talking about struct pages backing struct
pages (vmemmap) is usually a good recipe for headache but those diagrams
make it easy to follow the reasoning.
Anyway
Acked-by: michal Hocko <mhocko@suse.com>
@@ -27,8 +27,215 @@#include<linux/spinlock.h>#include<linux/vmalloc.h>#include<linux/sched.h>+#include<linux/pgtable.h>+#include<linux/bootmem_info.h>+#include<asm/dma.h>#include<asm/pgalloc.h>+#include<asm/tlbflush.h>++/**+*vmemmap_remap_walk-walkvmemmappagetable+*+*@remap_pte:calledforeachlowest-levelentry(PTE).+*@reuse_page:thepagewhichisreusedforthetailvmemmappages.+*@reuse_addr:thevirtualaddressofthe@reuse_pagepage.+*@vmemmap_pages:thelistheadofthevmemmappagesthatcanbefreed.+*/+structvmemmap_remap_walk{+void(*remap_pte)(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk);+structpage*reuse_page;+unsignedlongreuse_addr;+structlist_head*vmemmap_pages;+};++staticvoidvmemmap_pte_range(pmd_t*pmd,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pte_t*pte;++pte=pte_offset_kernel(pmd,addr);++/*+*Thereuse_pageisfound'first'intablewalkbeforewestart+*remapping(whichiscalling@walk->remap_pte).+*/+if(!walk->reuse_page){+BUG_ON(pte_none(*pte));+BUG_ON(walk->reuse_addr!=addr);++walk->reuse_page=pte_page(*pte++);+/*+*Becausethereuseaddressispartoftherangethatweare+*walking,skipthereuseaddressrange.+*/+addr+=PAGE_SIZE;+}++for(;addr!=end;addr+=PAGE_SIZE,pte++){+BUG_ON(pte_none(*pte));++walk->remap_pte(pte,addr,walk);+}+}++staticvoidvmemmap_pmd_range(pud_t*pud,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pmd_t*pmd;+unsignedlongnext;++pmd=pmd_offset(pud,addr);+do{+BUG_ON(pmd_none(*pmd)||pmd_leaf(*pmd));++next=pmd_addr_end(addr,end);+vmemmap_pte_range(pmd,addr,next,walk);+}while(pmd++,addr=next,addr!=end);+}++staticvoidvmemmap_pud_range(p4d_t*p4d,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pud_t*pud;+unsignedlongnext;++pud=pud_offset(p4d,addr);+do{+BUG_ON(pud_none(*pud));++next=pud_addr_end(addr,end);+vmemmap_pmd_range(pud,addr,next,walk);+}while(pud++,addr=next,addr!=end);+}++staticvoidvmemmap_p4d_range(pgd_t*pgd,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+p4d_t*p4d;+unsignedlongnext;++p4d=p4d_offset(pgd,addr);+do{+BUG_ON(p4d_none(*p4d));++next=p4d_addr_end(addr,end);+vmemmap_pud_range(p4d,addr,next,walk);+}while(p4d++,addr=next,addr!=end);+}++staticvoidvmemmap_remap_range(unsignedlongstart,unsignedlongend,+structvmemmap_remap_walk*walk)+{+unsignedlongaddr=start;+unsignedlongnext;+pgd_t*pgd;++VM_BUG_ON(!IS_ALIGNED(start,PAGE_SIZE));+VM_BUG_ON(!IS_ALIGNED(end,PAGE_SIZE));++pgd=pgd_offset_k(addr);+do{+BUG_ON(pgd_none(*pgd));++next=pgd_addr_end(addr,end);+vmemmap_p4d_range(pgd,addr,next,walk);+}while(pgd++,addr=next,addr!=end);++/*+*Weonlychangethemappingofthevmemmapvirtualaddressrange+*[@start+PAGE_SIZE,end),soweonlyneedtoflushtheTLBwhich+*belongstotherange.+*/+flush_tlb_kernel_range(start+PAGE_SIZE,end);+}++/*+*Freeavmemmappage.Avmemmappagecanbeallocatedfromthememblock+*allocatororbuddyallocator.IfthePG_reservedflagisset,itmeans+*thatitallocatedfromthememblockallocator,justfreeitviathe+*free_bootmem_page().Otherwise,use__free_page().+*/+staticinlinevoidfree_vmemmap_page(structpage*page)+{+if(PageReserved(page))+free_bootmem_page(page);+else+__free_page(page);+}++/* Free a list of the vmemmap pages */+staticvoidfree_vmemmap_page_list(structlist_head*list)+{+structpage*page,*next;++list_for_each_entry_safe(page,next,list,lru){+list_del(&page->lru);+free_vmemmap_page(page);+}+}++staticvoidvmemmap_remap_pte(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk)+{+/*+*Remapthetailpagesasread-onlytocatchillegalwriteoperation+*tothetailpages.+*/+pgprot_tpgprot=PAGE_KERNEL_RO;+pte_tentry=mk_pte(walk->reuse_page,pgprot);+structpage*page=pte_page(*pte);++list_add(&page->lru,walk->vmemmap_pages);+set_pte_at(&init_mm,addr,pte,entry);+}++/**+*vmemmap_remap_free-remapthevmemmapvirtualaddressrange[@start,@end)+*tothepagewhich@reuseismappedto,thenfreevmemmap+*whichtherangearemappedto.+*@start:startaddressofthevmemmapvirtualaddressrangethatwewant+*toremap.+*@end:endaddressofthevmemmapvirtualaddressrangethatwewantto+*remap.+*@reuse:reuseaddress.+*+*Note:Thisfunctiondependsonvmemmapbeingbasepagemapped.Pleasemake+*surethatwedisablePMDmappingofvmemmappageswhencallingthisfunction.+*/+voidvmemmap_remap_free(unsignedlongstart,unsignedlongend,+unsignedlongreuse)+{+LIST_HEAD(vmemmap_pages);+structvmemmap_remap_walkwalk={+.remap_pte=vmemmap_remap_pte,+.reuse_addr=reuse,+.vmemmap_pages=&vmemmap_pages,+};++/*+*Inordertomakeremappingroutinemostefficientforthehugepages,+*theroutineofvmemmappagetablewalkinghasthefollowingrules+*(seemoredetailsfromthevmemmap_pte_range()):+*+*-Therange[@start,@end)andtherange[@reuse,@reuse+PAGE_SIZE)+*shouldbecontinuous.+*-The@reuseaddressispartoftherange[@reuse,@end)thatweare+*walkingwhichispassedtovmemmap_remap_range().+*-The@reuseaddressisthefirstinthecompleterange.+*+*Soweneedtomakesurethat@startand@reusemeettheaboverules.+*/+BUG_ON(start-reuse!=PAGE_SIZE);++vmemmap_remap_range(reuse,end,&walk);+free_vmemmap_page_list(&vmemmap_pages);+}/**Allocateablockofmemorytobeusedtobackthevirtualmemorymap
+ set_page_refcounted(page);
+ set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
+
+ /*
+ * If the vmemmap pages associated with the HugeTLB page can be
+ * optimized or the page is gigantic, we might block in
+ * alloc_huge_page_vmemmap() or free_gigantic_page(). In both
+ * cases, drop the hugetlb_lock.
+ */
+ if (free_vmemmap_pages_per_hpage(h) || hstate_is_gigantic(h))
+ spin_unlock(&hugetlb_lock);
+
+ if (alloc_huge_page_vmemmap(h, page)) {
+ spin_lock(&hugetlb_lock);
+ INIT_LIST_HEAD(&page->lru);
+ set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
+ h->nr_huge_pages++;
+ h->nr_huge_pages_node[nid]++;
+
+ /*
+ * If we cannot allocate vmemmap pages, just refuse to free the
+ * page and put the page back on the hugetlb free list and treat
+ * as a surplus page.
+ */
+ h->surplus_huge_pages++;
+ h->surplus_huge_pages_node[nid]++;
+
+ /*
+ * The refcount can possibly be increased by memory-failure or
+ * soft_offline handlers.
This comment could be more helpful. I believe you want to say this
/*
* HWpoisoning code can increment the reference
* count here. If there is a race then bail out
* the holder of the additional reference count will
* free up the page with put_page.
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
[...]
This is quite ugly and confusing. update_and_free_page is careful to do
the proper counters accounting and now you just override it partially.
Why cannot we rely on update_and_free_page do the right thing?
--
Michal Hocko
SUSE Labs
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 15:28:56
On Mon 08-03-21 18:28:03, Muchun Song wrote:
Because we reuse the first tail vmemmap page frame and remap it
with read-only, we cannot set the PageHWPosion on some tail pages.
So we can use the head[4].private (There are at least 128 struct
page structures associated with the optimized HugeTLB page, so
using head[4].private is safe) to record the real error page index
and set the raw error page PageHWPoison later.
Can we have more poisoned tail pages? Also who does consume that index
and set the HWPoison on the proper tail page?
quoted hunk
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
mm/hugetlb.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++++++++------
1 file changed, 72 insertions(+), 8 deletions(-)
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 15:38:07
On Mon 08-03-21 18:28:04, Muchun Song wrote:
quoted hunk
Add a kernel parameter hugetlb_free_vmemmap to enable the feature of
freeing unused vmemmap pages associated with each hugetlb page on boot.
We disables PMD mapping of vmemmap pages for x86-64 arch when this
feature is enabled. Because vmemmap_remap_free() depends on vmemmap
being base page mapped.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Barry Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
Documentation/admin-guide/kernel-parameters.txt | 14 ++++++++++++++
Documentation/admin-guide/mm/hugetlbpage.rst | 3 +++
arch/x86/mm/init_64.c | 8 ++++++--
include/linux/hugetlb.h | 19 +++++++++++++++++++
mm/hugetlb_vmemmap.c | 24 ++++++++++++++++++++++++
5 files changed, 66 insertions(+), 2 deletions(-)
@@ -1557,6 +1557,20 @@ Documentation/admin-guide/mm/hugetlbpage.rst. Format: size[KMG]+ hugetlb_free_vmemmap=+ [KNL] When CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is set,+ this controls freeing unused vmemmap pages associated+ with each HugeTLB page. When this option is enabled,+ we disable PMD/huge page mapping of vmemmap pages which+ increase page table pages. So if a user/sysadmin only+ uses a small number of HugeTLB pages (as a percentage+ of system memory), they could end up using more memory+ with hugetlb_free_vmemmap on as opposed to off.+ Format: { on | off (default) }
Please note this is an admin guide and for those this seems overly low
level. I would use something like the following
[KNL] Reguires CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
enabled.
Allows heavy hugetlb users to free up some more
memory (6 * PAGE_SIZE for each 2MB hugetlb
page).
This feauture is not free though. Large page
tables are not use to back vmemmap pages which
can lead to a performance degradation for some
workloads. Also there will be memory allocation
required when hugetlb pages are freed from the
pool which can lead to corner cases under heavy
memory pressure.
+
+ on: enable the feature
+ off: disable the feature
+
hung_task_panic=
[KNL] Should the hung task detector generate panics.
Format: 0 | 1
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 15:40:49
On Mon 08-03-21 18:28:06, Muchun Song wrote:
For HugeTLB page, there are more metadata to save in the struct page.
But the head struct page cannot meet our needs, so we have to abuse
other tail struct page to store the metadata. In order to avoid
conflicts caused by subsequent use of more tail struct pages, we can
gather these discrete indexes of tail struct page. In this case, it
will be easier to add a new tail page index later.
There are only (RESERVE_VMEMMAP_SIZE / sizeof(struct page)) struct
page structs that can be used when CONFIG_HUGETLB_PAGE_FREE_VMEMMAP,
so add a BUILD_BUG_ON to catch invalid usage of the tail struct page.
OK, so this is what I have asked in an earlier patch. Good. I would
reorder and make this patch prior to the one relying on the fact though.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 15:42:31
On Mon 08-03-21 18:28:07, Muchun Song wrote:
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
I am confused. Don't you check for this in early_hugetlb_free_vmemmap_param already?
Why do we need any runtime checks?
quoted hunk
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
include/linux/hugetlb.h | 3 ++-
mm/hugetlb_vmemmap.c | 7 +++++++
mm/hugetlb_vmemmap.h | 6 ++++++
3 files changed, 15 insertions(+), 1 deletion(-)
From: Randy Dunlap <rdunlap@infradead.org> Date: 2021-03-10 17:20:27
On 3/10/21 7:37 AM, Michal Hocko wrote:
On Mon 08-03-21 18:28:04, Muchun Song wrote:
quoted
Add a kernel parameter hugetlb_free_vmemmap to enable the feature of
freeing unused vmemmap pages associated with each hugetlb page on boot.
We disables PMD mapping of vmemmap pages for x86-64 arch when this
feature is enabled. Because vmemmap_remap_free() depends on vmemmap
being base page mapped.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Barry Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
Documentation/admin-guide/kernel-parameters.txt | 14 ++++++++++++++
Documentation/admin-guide/mm/hugetlbpage.rst | 3 +++
arch/x86/mm/init_64.c | 8 ++++++--
include/linux/hugetlb.h | 19 +++++++++++++++++++
mm/hugetlb_vmemmap.c | 24 ++++++++++++++++++++++++
5 files changed, 66 insertions(+), 2 deletions(-)
@@ -1557,6 +1557,20 @@ Documentation/admin-guide/mm/hugetlbpage.rst. Format: size[KMG]+ hugetlb_free_vmemmap=+ [KNL] When CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is set,+ this controls freeing unused vmemmap pages associated+ with each HugeTLB page. When this option is enabled,+ we disable PMD/huge page mapping of vmemmap pages which+ increase page table pages. So if a user/sysadmin only+ uses a small number of HugeTLB pages (as a percentage+ of system memory), they could end up using more memory+ with hugetlb_free_vmemmap on as opposed to off.+ Format: { on | off (default) }
Please note this is an admin guide and for those this seems overly low
level. I would use something like the following
[KNL] Reguires CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
enabled.
Allows heavy hugetlb users to free up some more
memory (6 * PAGE_SIZE for each 2MB hugetlb
page).
This feauture is not free though. Large page
tables are not use to back vmemmap pages which
are not used
can lead to a performance degradation for some
workloads. Also there will be memory allocation
required when hugetlb pages are freed from the
pool which can lead to corner cases under heavy
memory pressure.
quoted
+
+ on: enable the feature
+ off: disable the feature
+
hung_task_panic=
[KNL] Should the hung task detector generate panics.
Format: 0 | 1
From: Mike Kravetz <hidden> Date: 2021-03-10 18:58:25
On 3/10/21 7:19 AM, Michal Hocko wrote:
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I'll put together a separate patch where we can discuss the merits of
making the change from !in_task to in_atomic, and what work remains in
this put_page area.
--
Mike Kravetz
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-10 21:11:59
On Wed 10-03-21 10:56:08, Mike Kravetz wrote:
On 3/10/21 7:19 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I haven't tried that but PREEMPT_COUNT should be selectable even without
any change to the preemption model (e.g. !PREEMPT).
--
Michal Hocko
SUSE Labs
From: "Paul E. McKenney" <paulmck@kernel.org> Date: 2021-03-10 21:49:49
On Wed, Mar 10, 2021 at 10:11:22PM +0100, Michal Hocko wrote:
On Wed 10-03-21 10:56:08, Mike Kravetz wrote:
quoted
On 3/10/21 7:19 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I haven't tried that but PREEMPT_COUNT should be selectable even without
any change to the preemption model (e.g. !PREEMPT).
It works reliably for me, for example as in the diff below. So,
as Michal says, you should be able to add "select PREEMPT_COUNT" to
whatever Kconfig option you need to.
Thanx, Paul
From: Mike Kravetz <hidden> Date: 2021-03-10 22:12:17
On 3/10/21 1:49 PM, Paul E. McKenney wrote:
On Wed, Mar 10, 2021 at 10:11:22PM +0100, Michal Hocko wrote:
quoted
On Wed 10-03-21 10:56:08, Mike Kravetz wrote:
quoted
On 3/10/21 7:19 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I haven't tried that but PREEMPT_COUNT should be selectable even without
any change to the preemption model (e.g. !PREEMPT).
It works reliably for me, for example as in the diff below. So,
as Michal says, you should be able to add "select PREEMPT_COUNT" to
whatever Kconfig option you need to.
Thanks Paul.
I may have been misreading Michal's suggestion of "make HUGETLB depend on
PREEMPT_COUNT". We could "select PREEMPT_COUNT" if HUGETLB is enabled.
However, since HUGETLB is enabled in most configs, then this would
result in PREEMPT_COUNT also being enabled in most configs. I honestly
do not know how much this will cost us? I assume that if it was free or
really cheap it would already be always on?
--
Mike Kravetz
From: "Paul E. McKenney" <paulmck@kernel.org> Date: 2021-03-10 23:29:59
On Wed, Mar 10, 2021 at 02:10:12PM -0800, Mike Kravetz wrote:
On 3/10/21 1:49 PM, Paul E. McKenney wrote:
quoted
On Wed, Mar 10, 2021 at 10:11:22PM +0100, Michal Hocko wrote:
quoted
On Wed 10-03-21 10:56:08, Mike Kravetz wrote:
quoted
On 3/10/21 7:19 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I haven't tried that but PREEMPT_COUNT should be selectable even without
any change to the preemption model (e.g. !PREEMPT).
It works reliably for me, for example as in the diff below. So,
as Michal says, you should be able to add "select PREEMPT_COUNT" to
whatever Kconfig option you need to.
Thanks Paul.
I may have been misreading Michal's suggestion of "make HUGETLB depend on
PREEMPT_COUNT". We could "select PREEMPT_COUNT" if HUGETLB is enabled.
However, since HUGETLB is enabled in most configs, then this would
result in PREEMPT_COUNT also being enabled in most configs. I honestly
do not know how much this will cost us? I assume that if it was free or
really cheap it would already be always on?
There are a -lot- of configs out there, so are you sure that HUGETLB is
really enabled in most of them? ;-)
More seriously, I was going by earlier emails in this and related threads
plus Michal's "PREEMPT_COUNT should be selectable". But there are other
situations that would like PREEMPT_COUNT. And to your point, some who
would rather PREEMPT_COUNT not be universally enabled. I haven't seen
any performance or kernel-size numbers from any of them, however.
Thanx, Paul
From: Muchun Song <hidden> Date: 2021-03-11 02:59:58
On Wed, Mar 10, 2021 at 10:14 PM Michal Hocko [off-list ref] wrote:
[I am sorry for a late review]
Thanks for your review.
On Mon 08-03-21 18:27:59, Muchun Song wrote:
quoted
Move bootmem info registration common API to individual bootmem_info.c.
And we will use {get,put}_page_bootmem() to initialize the page for the
vmemmap pages or free the vmemmap pages to buddy in the later patch.
So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code
movement without any functional change.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Separation from memory_hotplug.c is definitely a right step. I am
wondering about the config dependency though
[...]
I would have expected this would depend on CONFIG_SPARSE.
BOOTMEM_INFO_NODE is really an odd thing to depend on here. There is
some functionality which requires the node info but that can be gated
specifically. Or what is the thinking behind?
At first my idea was to free vmemmap pages through the bootmem
interface. My first instinct is to rely on BOOTMEM_INFO_NODE.
It makes sense to me to depend on CONFIG_SPARSE. I will
update this in the next version.
Thanks.
This doesn't matter right now because it seems that the *_page_bootmem
is only used by x86 outside of the memory hotplug.
Other than that looks good to me.
--
Michal Hocko
SUSE Labs
From: Muchun Song <hidden> Date: 2021-03-11 03:37:13
On Wed, Mar 10, 2021 at 10:32 PM Michal Hocko [off-list ref] wrote:
On Mon 08-03-21 18:28:01, Muchun Song wrote:
quoted
Every HugeTLB has more than one struct page structure. We __know__ that
we only use the first 4(HUGETLB_CGROUP_MIN_ORDER) struct page structures
to store metadata associated with each HugeTLB.
I think it would be great to make this explicit somewhere around the
code which uses those struct pages.
OK. I will make patch #8 prior to this one. Thanks.
quoted
There are a lot of struct page structures associated with each HugeTLB
page. For tail pages, the value of compound_head is the same. So we can
reuse first page of tail page structures. We map the virtual addresses
of the remaining pages of tail page structures to the first tail page
struct, and then free these page frames. Therefore, we need to reserve
two pages as vmemmap areas.
When we allocate a HugeTLB page from the buddy, we can free some vmemmap
pages associated with each HugeTLB page. It is more appropriate to do it
in the prep_new_huge_page().
The free_vmemmap_pages_per_hpage(), which indicates how many vmemmap
pages associated with a HugeTLB page can be freed, returns zero for
now, which means the feature is disabled. We will enable it once all
the infrastructure is there.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
I do not see any issues here. I just want to point out that the amount
of *BUG_ON is quite high to my taste. Most of them seem to be added just
in case if something goes wrong or should never happen. These are
usually bad reasons to add them IMHO. I would just drop those unless
there is a very good reason to keep them around.
OK. I will drop the useless *BUG_ON.
I really appreciate how you made a high level design documentation to
the source code directly. Talking about struct pages backing struct
pages (vmemmap) is usually a good recipe for headache but those diagrams
make it easy to follow the reasoning.
Anyway
Acked-by: michal Hocko <mhocko@suse.com>
@@ -27,8 +27,215 @@#include<linux/spinlock.h>#include<linux/vmalloc.h>#include<linux/sched.h>+#include<linux/pgtable.h>+#include<linux/bootmem_info.h>+#include<asm/dma.h>#include<asm/pgalloc.h>+#include<asm/tlbflush.h>++/**+*vmemmap_remap_walk-walkvmemmappagetable+*+*@remap_pte:calledforeachlowest-levelentry(PTE).+*@reuse_page:thepagewhichisreusedforthetailvmemmappages.+*@reuse_addr:thevirtualaddressofthe@reuse_pagepage.+*@vmemmap_pages:thelistheadofthevmemmappagesthatcanbefreed.+*/+structvmemmap_remap_walk{+void(*remap_pte)(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk);+structpage*reuse_page;+unsignedlongreuse_addr;+structlist_head*vmemmap_pages;+};++staticvoidvmemmap_pte_range(pmd_t*pmd,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pte_t*pte;++pte=pte_offset_kernel(pmd,addr);++/*+*Thereuse_pageisfound'first'intablewalkbeforewestart+*remapping(whichiscalling@walk->remap_pte).+*/+if(!walk->reuse_page){+BUG_ON(pte_none(*pte));+BUG_ON(walk->reuse_addr!=addr);++walk->reuse_page=pte_page(*pte++);+/*+*Becausethereuseaddressispartoftherangethatweare+*walking,skipthereuseaddressrange.+*/+addr+=PAGE_SIZE;+}++for(;addr!=end;addr+=PAGE_SIZE,pte++){+BUG_ON(pte_none(*pte));++walk->remap_pte(pte,addr,walk);+}+}++staticvoidvmemmap_pmd_range(pud_t*pud,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pmd_t*pmd;+unsignedlongnext;++pmd=pmd_offset(pud,addr);+do{+BUG_ON(pmd_none(*pmd)||pmd_leaf(*pmd));++next=pmd_addr_end(addr,end);+vmemmap_pte_range(pmd,addr,next,walk);+}while(pmd++,addr=next,addr!=end);+}++staticvoidvmemmap_pud_range(p4d_t*p4d,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+pud_t*pud;+unsignedlongnext;++pud=pud_offset(p4d,addr);+do{+BUG_ON(pud_none(*pud));++next=pud_addr_end(addr,end);+vmemmap_pmd_range(pud,addr,next,walk);+}while(pud++,addr=next,addr!=end);+}++staticvoidvmemmap_p4d_range(pgd_t*pgd,unsignedlongaddr,+unsignedlongend,+structvmemmap_remap_walk*walk)+{+p4d_t*p4d;+unsignedlongnext;++p4d=p4d_offset(pgd,addr);+do{+BUG_ON(p4d_none(*p4d));++next=p4d_addr_end(addr,end);+vmemmap_pud_range(p4d,addr,next,walk);+}while(p4d++,addr=next,addr!=end);+}++staticvoidvmemmap_remap_range(unsignedlongstart,unsignedlongend,+structvmemmap_remap_walk*walk)+{+unsignedlongaddr=start;+unsignedlongnext;+pgd_t*pgd;++VM_BUG_ON(!IS_ALIGNED(start,PAGE_SIZE));+VM_BUG_ON(!IS_ALIGNED(end,PAGE_SIZE));++pgd=pgd_offset_k(addr);+do{+BUG_ON(pgd_none(*pgd));++next=pgd_addr_end(addr,end);+vmemmap_p4d_range(pgd,addr,next,walk);+}while(pgd++,addr=next,addr!=end);++/*+*Weonlychangethemappingofthevmemmapvirtualaddressrange+*[@start+PAGE_SIZE,end),soweonlyneedtoflushtheTLBwhich+*belongstotherange.+*/+flush_tlb_kernel_range(start+PAGE_SIZE,end);+}++/*+*Freeavmemmappage.Avmemmappagecanbeallocatedfromthememblock+*allocatororbuddyallocator.IfthePG_reservedflagisset,itmeans+*thatitallocatedfromthememblockallocator,justfreeitviathe+*free_bootmem_page().Otherwise,use__free_page().+*/+staticinlinevoidfree_vmemmap_page(structpage*page)+{+if(PageReserved(page))+free_bootmem_page(page);+else+__free_page(page);+}++/* Free a list of the vmemmap pages */+staticvoidfree_vmemmap_page_list(structlist_head*list)+{+structpage*page,*next;++list_for_each_entry_safe(page,next,list,lru){+list_del(&page->lru);+free_vmemmap_page(page);+}+}++staticvoidvmemmap_remap_pte(pte_t*pte,unsignedlongaddr,+structvmemmap_remap_walk*walk)+{+/*+*Remapthetailpagesasread-onlytocatchillegalwriteoperation+*tothetailpages.+*/+pgprot_tpgprot=PAGE_KERNEL_RO;+pte_tentry=mk_pte(walk->reuse_page,pgprot);+structpage*page=pte_page(*pte);++list_add(&page->lru,walk->vmemmap_pages);+set_pte_at(&init_mm,addr,pte,entry);+}++/**+*vmemmap_remap_free-remapthevmemmapvirtualaddressrange[@start,@end)+*tothepagewhich@reuseismappedto,thenfreevmemmap+*whichtherangearemappedto.+*@start:startaddressofthevmemmapvirtualaddressrangethatwewant+*toremap.+*@end:endaddressofthevmemmapvirtualaddressrangethatwewantto+*remap.+*@reuse:reuseaddress.+*+*Note:Thisfunctiondependsonvmemmapbeingbasepagemapped.Pleasemake+*surethatwedisablePMDmappingofvmemmappageswhencallingthisfunction.+*/+voidvmemmap_remap_free(unsignedlongstart,unsignedlongend,+unsignedlongreuse)+{+LIST_HEAD(vmemmap_pages);+structvmemmap_remap_walkwalk={+.remap_pte=vmemmap_remap_pte,+.reuse_addr=reuse,+.vmemmap_pages=&vmemmap_pages,+};++/*+*Inordertomakeremappingroutinemostefficientforthehugepages,+*theroutineofvmemmappagetablewalkinghasthefollowingrules+*(seemoredetailsfromthevmemmap_pte_range()):+*+*-Therange[@start,@end)andtherange[@reuse,@reuse+PAGE_SIZE)+*shouldbecontinuous.+*-The@reuseaddressispartoftherange[@reuse,@end)thatweare+*walkingwhichispassedtovmemmap_remap_range().+*-The@reuseaddressisthefirstinthecompleterange.+*+*Soweneedtomakesurethat@startand@reusemeettheaboverules.+*/+BUG_ON(start-reuse!=PAGE_SIZE);++vmemmap_remap_range(reuse,end,&walk);+free_vmemmap_page_list(&vmemmap_pages);+}/**Allocateablockofmemorytobeusedtobackthevirtualmemorymap--
From: Muchun Song <hidden> Date: 2021-03-11 04:15:26
On Wed, Mar 10, 2021 at 10:21 PM Oscar Salvador [off-list ref] wrote:
On Mon, Mar 08, 2021 at 06:28:02PM +0800, Muchun Song wrote:
quoted
When we free a HugeTLB page to the buddy allocator, we need to allocate
the vmemmap pages associated with it. However, we may not be able to
allocate the vmemmap pages when the system is under memory pressure. In
this case, we just refuse to free the HugeTLB page. This changes behavior
in some corner cases as listed below:
1) Failing to free a huge page triggered by the user (decrease nr_pages).
User needs to try again later.
2) Failing to free a surplus huge page when freed by the application.
Try again later when freeing a huge page next time.
3) Failing to dissolve a free huge page on ZONE_MOVABLE via
offline_pages().
This can happen when we have plenty of ZONE_MOVABLE memory, but
not enough kernel memory to allocate vmemmmap pages. We may even
be able to migrate huge page contents, but will not be able to
dissolve the source huge page. This will prevent an offline
operation and is unfortunate as memory offlining is expected to
succeed on movable zones. Users that depend on memory hotplug
to succeed for movable zones should carefully consider whether the
memory savings gained from this feature are worth the risk of
possibly not being able to offline memory in certain situations.
This is nice to have it here, but a normal user won't dig in the kernel to
figure this out, so my question is: Do we have this documented somewhere under
Documentation/?
If not, could we document it there? It is nice to warn about this things were
sysadmins can find them.
Make sense. I will do this.
quoted
4) Failing to dissolve a huge page on CMA/ZONE_MOVABLE via
alloc_contig_range() - once we have that handling in place. Mainly
affects CMA and virtio-mem.
Similar to 3). virito-mem will handle migration errors gracefully.
CMA might be able to fallback on other free areas within the CMA
region.
Vmemmap pages are allocated from the page freeing context. In order for
those allocations to be not disruptive (e.g. trigger oom killer)
__GFP_NORETRY is used. hugetlb_lock is dropped for the allocation
because a non sleeping allocation would be too fragile and it could fail
too easily under memory pressure. GFP_ATOMIC or other modes to access
memory reserves is not used because we want to prevent consuming
reserves under heavy hugetlb freeing.
Signed-off-by: Muchun Song <redacted>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Sorry for jumping in late.
It looks good to me:
Reviewed-by: Oscar Salvador <osalvador@suse.de>
[...]
Could we place a brief comment about what we expect to return here?
OK. Will do.
quoted
-static inline unsigned long free_vmemmap_pages_size_per_hpage(struct hstate *h)
+int alloc_huge_page_vmemmap(struct hstate *h, struct page *head)
{
- return (unsigned long)free_vmemmap_pages_per_hpage(h) << PAGE_SHIFT;
+ unsigned long vmemmap_addr = (unsigned long)head;
+ unsigned long vmemmap_end, vmemmap_reuse;
+
+ if (!free_vmemmap_pages_per_hpage(h))
+ return 0;
+
+ vmemmap_addr += RESERVE_VMEMMAP_SIZE;
+ vmemmap_end = vmemmap_addr + free_vmemmap_pages_size_per_hpage(h);
+ vmemmap_reuse = vmemmap_addr - PAGE_SIZE;
+ /*
+ * The pages which the vmemmap virtual address range [@vmemmap_addr,
+ * @vmemmap_end) are mapped to are freed to the buddy allocator, and
+ * the range is mapped to the page which @vmemmap_reuse is mapped to.
+ * When a HugeTLB page is freed to the buddy allocator, previously
+ * discarded vmemmap pages must be allocated and remapping.
+ */
+ return vmemmap_remap_alloc(vmemmap_addr, vmemmap_end, vmemmap_reuse,
+ GFP_KERNEL | __GFP_NORETRY | __GFP_THISNODE);
}
+ set_page_refcounted(page);
+ set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
+
+ /*
+ * If the vmemmap pages associated with the HugeTLB page can be
+ * optimized or the page is gigantic, we might block in
+ * alloc_huge_page_vmemmap() or free_gigantic_page(). In both
+ * cases, drop the hugetlb_lock.
+ */
+ if (free_vmemmap_pages_per_hpage(h) || hstate_is_gigantic(h))
+ spin_unlock(&hugetlb_lock);
+
+ if (alloc_huge_page_vmemmap(h, page)) {
+ spin_lock(&hugetlb_lock);
+ INIT_LIST_HEAD(&page->lru);
+ set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
+ h->nr_huge_pages++;
+ h->nr_huge_pages_node[nid]++;
+
+ /*
+ * If we cannot allocate vmemmap pages, just refuse to free the
+ * page and put the page back on the hugetlb free list and treat
+ * as a surplus page.
+ */
+ h->surplus_huge_pages++;
+ h->surplus_huge_pages_node[nid]++;
+
+ /*
+ * The refcount can possibly be increased by memory-failure or
+ * soft_offline handlers.
This comment could be more helpful. I believe you want to say this
/*
* HWpoisoning code can increment the reference
* count here. If there is a race then bail out
* the holder of the additional reference count will
* free up the page with put_page.
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
[...]
This is quite ugly and confusing. update_and_free_page is careful to do
the proper counters accounting and now you just override it partially.
Why cannot we rely on update_and_free_page do the right thing?
Dissolving path is special here. Since update_and_free_page failed,
the number of surplus pages was incremented. Surplus pages are
the number of pages greater than max_huge_pages. Since we are
incrementing max_huge_pages, we should decrement (undo) the
addition to surplus_huge_pages and surplus_huge_pages_node[nid].
From: Muchun Song <hidden> Date: 2021-03-11 06:35:28
On Wed, Mar 10, 2021 at 11:28 PM Michal Hocko [off-list ref] wrote:
On Mon 08-03-21 18:28:03, Muchun Song wrote:
quoted
Because we reuse the first tail vmemmap page frame and remap it
with read-only, we cannot set the PageHWPosion on some tail pages.
So we can use the head[4].private (There are at least 128 struct
page structures associated with the optimized HugeTLB page, so
using head[4].private is safe) to record the real error page index
and set the raw error page PageHWPoison later.
Can we have more poisoned tail pages? Also who does consume that index
and set the HWPoison on the proper tail page?
Good point. I look at the routine of memory failure closely.
If we do not clear the HWPoison of the head page, we cannot
poison another tail page.
So we should not set the destructor of the huge page from
HUGETLB_PAGE_DTOR to NULL_COMPOUND_DTOR
before calling alloc_huge_page_vmemmap(). In this case,
the below check of PageHuge() always returns true.
I need to fix this in the previous patch.
memory_failure()
if (PageHuge(page))
memory_failure_hugetlb()
head = compound_head(page)
if (TestSetPageHWPoison(head))
return
Thanks.
quoted
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
mm/hugetlb.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++++++++------
1 file changed, 72 insertions(+), 8 deletions(-)
From: Muchun Song <hidden> Date: 2021-03-11 06:37:36
On Wed, Mar 10, 2021 at 11:37 PM Michal Hocko [off-list ref] wrote:
On Mon 08-03-21 18:28:04, Muchun Song wrote:
quoted
Add a kernel parameter hugetlb_free_vmemmap to enable the feature of
freeing unused vmemmap pages associated with each hugetlb page on boot.
We disables PMD mapping of vmemmap pages for x86-64 arch when this
feature is enabled. Because vmemmap_remap_free() depends on vmemmap
being base page mapped.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Barry Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
Documentation/admin-guide/kernel-parameters.txt | 14 ++++++++++++++
Documentation/admin-guide/mm/hugetlbpage.rst | 3 +++
arch/x86/mm/init_64.c | 8 ++++++--
include/linux/hugetlb.h | 19 +++++++++++++++++++
mm/hugetlb_vmemmap.c | 24 ++++++++++++++++++++++++
5 files changed, 66 insertions(+), 2 deletions(-)
@@ -1557,6 +1557,20 @@ Documentation/admin-guide/mm/hugetlbpage.rst. Format: size[KMG]+ hugetlb_free_vmemmap=+ [KNL] When CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is set,+ this controls freeing unused vmemmap pages associated+ with each HugeTLB page. When this option is enabled,+ we disable PMD/huge page mapping of vmemmap pages which+ increase page table pages. So if a user/sysadmin only+ uses a small number of HugeTLB pages (as a percentage+ of system memory), they could end up using more memory+ with hugetlb_free_vmemmap on as opposed to off.+ Format: { on | off (default) }
Please note this is an admin guide and for those this seems overly low
OK.
level. I would use something like the following
[KNL] Reguires CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
enabled.
Allows heavy hugetlb users to free up some more
memory (6 * PAGE_SIZE for each 2MB hugetlb
page).
This feauture is not free though. Large page
tables are not use to back vmemmap pages which
can lead to a performance degradation for some
workloads. Also there will be memory allocation
required when hugetlb pages are freed from the
pool which can lead to corner cases under heavy
memory pressure.
Very thanks. I will update this.
quoted
+
+ on: enable the feature
+ off: disable the feature
+
hung_task_panic=
[KNL] Should the hung task detector generate panics.
Format: 0 | 1
From: Muchun Song <hidden> Date: 2021-03-11 06:38:08
On Thu, Mar 11, 2021 at 1:16 AM Randy Dunlap [off-list ref] wrote:
On 3/10/21 7:37 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:04, Muchun Song wrote:
quoted
Add a kernel parameter hugetlb_free_vmemmap to enable the feature of
freeing unused vmemmap pages associated with each hugetlb page on boot.
We disables PMD mapping of vmemmap pages for x86-64 arch when this
feature is enabled. Because vmemmap_remap_free() depends on vmemmap
being base page mapped.
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Barry Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
Documentation/admin-guide/kernel-parameters.txt | 14 ++++++++++++++
Documentation/admin-guide/mm/hugetlbpage.rst | 3 +++
arch/x86/mm/init_64.c | 8 ++++++--
include/linux/hugetlb.h | 19 +++++++++++++++++++
mm/hugetlb_vmemmap.c | 24 ++++++++++++++++++++++++
5 files changed, 66 insertions(+), 2 deletions(-)
@@ -1557,6 +1557,20 @@ Documentation/admin-guide/mm/hugetlbpage.rst. Format: size[KMG]+ hugetlb_free_vmemmap=+ [KNL] When CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is set,+ this controls freeing unused vmemmap pages associated+ with each HugeTLB page. When this option is enabled,+ we disable PMD/huge page mapping of vmemmap pages which+ increase page table pages. So if a user/sysadmin only+ uses a small number of HugeTLB pages (as a percentage+ of system memory), they could end up using more memory+ with hugetlb_free_vmemmap on as opposed to off.+ Format: { on | off (default) }
Please note this is an admin guide and for those this seems overly low
level. I would use something like the following
[KNL] Reguires CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
enabled.
Allows heavy hugetlb users to free up some more
memory (6 * PAGE_SIZE for each 2MB hugetlb
page).
This feauture is not free though. Large page
tables are not use to back vmemmap pages which
are not used
Thanks.
quoted
can lead to a performance degradation for some
workloads. Also there will be memory allocation
required when hugetlb pages are freed from the
pool which can lead to corner cases under heavy
memory pressure.
quoted
+
+ on: enable the feature
+ off: disable the feature
+
hung_task_panic=
[KNL] Should the hung task detector generate panics.
Format: 0 | 1
From: Muchun Song <hidden> Date: 2021-03-11 07:35:19
On Wed, Mar 10, 2021 at 11:41 PM Michal Hocko [off-list ref] wrote:
On Mon 08-03-21 18:28:07, Muchun Song wrote:
quoted
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
I am confused. Don't you check for this in early_hugetlb_free_vmemmap_param already?
Right.
Why do we need any runtime checks?
If the size of the struct page is not power of 2, compiler can think
is_hugetlb_free_vmemmap_enabled() always return false. So
the code snippet of this user can be optimized away.
E.g.
if (is_hugetlb_free_vmemmap_enabled())
/* do something */
The compiler can drop "/* do something */" directly, because
it knows is_hugetlb_free_vmemmap_enabled() always returns
false.
Thanks.
quoted
Signed-off-by: Muchun Song <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
---
include/linux/hugetlb.h | 3 ++-
mm/hugetlb_vmemmap.c | 7 +++++++
mm/hugetlb_vmemmap.h | 6 ++++++
3 files changed, 15 insertions(+), 1 deletion(-)
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 08:41:14
On Wed 10-03-21 15:28:51, Paul E. McKenney wrote:
On Wed, Mar 10, 2021 at 02:10:12PM -0800, Mike Kravetz wrote:
quoted
On 3/10/21 1:49 PM, Paul E. McKenney wrote:
quoted
On Wed, Mar 10, 2021 at 10:11:22PM +0100, Michal Hocko wrote:
quoted
On Wed 10-03-21 10:56:08, Mike Kravetz wrote:
quoted
On 3/10/21 7:19 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I haven't tried that but PREEMPT_COUNT should be selectable even without
any change to the preemption model (e.g. !PREEMPT).
It works reliably for me, for example as in the diff below. So,
as Michal says, you should be able to add "select PREEMPT_COUNT" to
whatever Kconfig option you need to.
Thanks Paul.
I may have been misreading Michal's suggestion of "make HUGETLB depend on
PREEMPT_COUNT". We could "select PREEMPT_COUNT" if HUGETLB is enabled.
However, since HUGETLB is enabled in most configs, then this would
result in PREEMPT_COUNT also being enabled in most configs. I honestly
do not know how much this will cost us? I assume that if it was free or
really cheap it would already be always on?
There are a -lot- of configs out there, so are you sure that HUGETLB is
really enabled in most of them? ;-)
It certainly is enabled for all distribution kernels and many are
!PREEMPT so I believe this is what Mike was concerned about.
More seriously, I was going by earlier emails in this and related threads
plus Michal's "PREEMPT_COUNT should be selectable". But there are other
situations that would like PREEMPT_COUNT. And to your point, some who
would rather PREEMPT_COUNT not be universally enabled. I haven't seen
any performance or kernel-size numbers from any of them, however.
Yeah per cpu preempt counting shouldn't be noticeable but I have to
confess I haven't benchmarked it.
--
Michal Hocko
SUSE Labs
From: Muchun Song <hidden> Date: 2021-03-11 08:47:07
On Thu, Mar 11, 2021 at 10:58 AM Muchun Song [off-list ref] wrote:
On Wed, Mar 10, 2021 at 10:14 PM Michal Hocko [off-list ref] wrote:
quoted
[I am sorry for a late review]
Thanks for your review.
quoted
On Mon 08-03-21 18:27:59, Muchun Song wrote:
quoted
Move bootmem info registration common API to individual bootmem_info.c.
And we will use {get,put}_page_bootmem() to initialize the page for the
vmemmap pages or free the vmemmap pages to buddy in the later patch.
So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code
movement without any functional change.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Separation from memory_hotplug.c is definitely a right step. I am
wondering about the config dependency though
[...]
I would have expected this would depend on CONFIG_SPARSE.
BOOTMEM_INFO_NODE is really an odd thing to depend on here. There is
some functionality which requires the node info but that can be gated
specifically. Or what is the thinking behind?
I have tried this. And I find that it is better to depend on
BOOTMEM_INFO_NODE instead of SPARSEMEM.
If we enable SPARSEMEM but disable HAVE_BOOTMEM_INFO_NODE,
the bootmem_info.c also is compiled. Actually, we do not
need those functions on other architectures. And these
functions are also related to bootmem info. So it may be
more reasonable to depend on BOOTMEM_INFO_NODE.
Just my thoughts.
Thanks.
At first my idea was to free vmemmap pages through the bootmem
interface. My first instinct is to rely on BOOTMEM_INFO_NODE.
It makes sense to me to depend on CONFIG_SPARSE. I will
update this in the next version.
Thanks.
quoted
This doesn't matter right now because it seems that the *_page_bootmem
is only used by x86 outside of the memory hotplug.
Other than that looks good to me.
--
Michal Hocko
SUSE Labs
This is quite ugly and confusing. update_and_free_page is careful to do
the proper counters accounting and now you just override it partially.
Why cannot we rely on update_and_free_page do the right thing?
Dissolving path is special here. Since update_and_free_page failed,
the number of surplus pages was incremented. Surplus pages are
the number of pages greater than max_huge_pages. Since we are
incrementing max_huge_pages, we should decrement (undo) the
addition to surplus_huge_pages and surplus_huge_pages_node[nid].
Can we make dissolve_free_huge_page less special or tell
update_and_free_page to not account against dissolve_free_huge_page?
--
Michal Hocko
SUSE Labs
This is quite ugly and confusing. update_and_free_page is careful to do
the proper counters accounting and now you just override it partially.
Why cannot we rely on update_and_free_page do the right thing?
Dissolving path is special here. Since update_and_free_page failed,
the number of surplus pages was incremented. Surplus pages are
the number of pages greater than max_huge_pages. Since we are
incrementing max_huge_pages, we should decrement (undo) the
addition to surplus_huge_pages and surplus_huge_pages_node[nid].
Can we make dissolve_free_huge_page less special or tell
update_and_free_page to not account against dissolve_free_huge_page?
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 08:50:54
On Thu 11-03-21 14:34:04, Muchun Song wrote:
On Wed, Mar 10, 2021 at 11:28 PM Michal Hocko [off-list ref] wrote:
quoted
On Mon 08-03-21 18:28:03, Muchun Song wrote:
quoted
Because we reuse the first tail vmemmap page frame and remap it
with read-only, we cannot set the PageHWPosion on some tail pages.
So we can use the head[4].private (There are at least 128 struct
page structures associated with the optimized HugeTLB page, so
using head[4].private is safe) to record the real error page index
and set the raw error page PageHWPoison later.
Can we have more poisoned tail pages? Also who does consume that index
and set the HWPoison on the proper tail page?
Good point. I look at the routine of memory failure closely.
If we do not clear the HWPoison of the head page, we cannot
poison another tail page.
So we should not set the destructor of the huge page from
HUGETLB_PAGE_DTOR to NULL_COMPOUND_DTOR
before calling alloc_huge_page_vmemmap(). In this case,
the below check of PageHuge() always returns true.
I need to fix this in the previous patch.
memory_failure()
if (PageHuge(page))
memory_failure_hugetlb()
head = compound_head(page)
if (TestSetPageHWPoison(head))
return
I have to say that I am not fully familiar with hwpoisoning code
(especially after recent changes) but IIRC it does rely on hugetlb page
dissolving. With the new code this operation can fail which is a new
situation. Unless I am misunderstanding this can lead to a lost memory
failure operation on other tail pages.
Anyway the above answers the question why a single slot is sufficient so
it would be great to mention that in a changelog along with the caveat
that some pages might miss their poisoning.
--
Michal Hocko
SUSE Labs
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 08:54:36
On Thu 11-03-21 16:45:51, Muchun Song wrote:
On Thu, Mar 11, 2021 at 10:58 AM Muchun Song [off-list ref] wrote:
quoted
On Wed, Mar 10, 2021 at 10:14 PM Michal Hocko [off-list ref] wrote:
quoted
[I am sorry for a late review]
Thanks for your review.
quoted
On Mon 08-03-21 18:27:59, Muchun Song wrote:
quoted
Move bootmem info registration common API to individual bootmem_info.c.
And we will use {get,put}_page_bootmem() to initialize the page for the
vmemmap pages or free the vmemmap pages to buddy in the later patch.
So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code
movement without any functional change.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Separation from memory_hotplug.c is definitely a right step. I am
wondering about the config dependency though
[...]
I would have expected this would depend on CONFIG_SPARSE.
BOOTMEM_INFO_NODE is really an odd thing to depend on here. There is
some functionality which requires the node info but that can be gated
specifically. Or what is the thinking behind?
I have tried this. And I find that it is better to depend on
BOOTMEM_INFO_NODE instead of SPARSEMEM.
If we enable SPARSEMEM but disable HAVE_BOOTMEM_INFO_NODE,
the bootmem_info.c also is compiled. Actually, we do not
need those functions on other architectures. And these
functions are also related to bootmem info. So it may be
more reasonable to depend on BOOTMEM_INFO_NODE.
Just my thoughts.
If BOOTMEM_INFO_NODE is disbabled then bootmem_info.c would be
effectivelly only {get,put}_page_bootmem, no?
--
Michal Hocko
SUSE Labs
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 08:55:41
On Thu 11-03-21 15:33:20, Muchun Song wrote:
On Wed, Mar 10, 2021 at 11:41 PM Michal Hocko [off-list ref] wrote:
quoted
On Mon 08-03-21 18:28:07, Muchun Song wrote:
quoted
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
I am confused. Don't you check for this in early_hugetlb_free_vmemmap_param already?
Right.
quoted
Why do we need any runtime checks?
If the size of the struct page is not power of 2, compiler can think
is_hugetlb_free_vmemmap_enabled() always return false. So
the code snippet of this user can be optimized away.
E.g.
if (is_hugetlb_free_vmemmap_enabled())
/* do something */
The compiler can drop "/* do something */" directly, because
it knows is_hugetlb_free_vmemmap_enabled() always returns
false.
OK, so this is a micro-optimization to generate a better code?
Is this measurable to warrant more code?
--
Michal Hocko
SUSE Labs
From: Muchun Song <hidden> Date: 2021-03-11 09:06:54
On Thu, Mar 11, 2021 at 4:53 PM Michal Hocko [off-list ref] wrote:
On Thu 11-03-21 16:45:51, Muchun Song wrote:
quoted
On Thu, Mar 11, 2021 at 10:58 AM Muchun Song [off-list ref] wrote:
quoted
On Wed, Mar 10, 2021 at 10:14 PM Michal Hocko [off-list ref] wrote:
quoted
[I am sorry for a late review]
Thanks for your review.
quoted
On Mon 08-03-21 18:27:59, Muchun Song wrote:
quoted
Move bootmem info registration common API to individual bootmem_info.c.
And we will use {get,put}_page_bootmem() to initialize the page for the
vmemmap pages or free the vmemmap pages to buddy in the later patch.
So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code
movement without any functional change.
Signed-off-by: Muchun Song <redacted>
Acked-by: Mike Kravetz <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Chen Huang <redacted>
Tested-by: Bodeddula Balasubramaniam <redacted>
Separation from memory_hotplug.c is definitely a right step. I am
wondering about the config dependency though
[...]
I would have expected this would depend on CONFIG_SPARSE.
BOOTMEM_INFO_NODE is really an odd thing to depend on here. There is
some functionality which requires the node info but that can be gated
specifically. Or what is the thinking behind?
I have tried this. And I find that it is better to depend on
BOOTMEM_INFO_NODE instead of SPARSEMEM.
If we enable SPARSEMEM but disable HAVE_BOOTMEM_INFO_NODE,
the bootmem_info.c also is compiled. Actually, we do not
need those functions on other architectures. And these
functions are also related to bootmem info. So it may be
more reasonable to depend on BOOTMEM_INFO_NODE.
Just my thoughts.
If BOOTMEM_INFO_NODE is disbabled then bootmem_info.c would be
effectivelly only {get,put}_page_bootmem, no?
{get,put}_page_bootmem also would be effective. I found that
get_page_bootmem is only used in the scope of the
CONFIG_BOOTMEM_INFO_NODE. So I move them
to the bootmem_info.c.
Thanks.
From: Muchun Song <hidden> Date: 2021-03-11 09:10:06
On Thu, Mar 11, 2021 at 4:55 PM Michal Hocko [off-list ref] wrote:
On Thu 11-03-21 15:33:20, Muchun Song wrote:
quoted
On Wed, Mar 10, 2021 at 11:41 PM Michal Hocko [off-list ref] wrote:
quoted
On Mon 08-03-21 18:28:07, Muchun Song wrote:
quoted
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
I am confused. Don't you check for this in early_hugetlb_free_vmemmap_param already?
Right.
quoted
Why do we need any runtime checks?
If the size of the struct page is not power of 2, compiler can think
is_hugetlb_free_vmemmap_enabled() always return false. So
the code snippet of this user can be optimized away.
E.g.
if (is_hugetlb_free_vmemmap_enabled())
/* do something */
The compiler can drop "/* do something */" directly, because
it knows is_hugetlb_free_vmemmap_enabled() always returns
false.
OK, so this is a micro-optimization to generate a better code?
Right.
Is this measurable to warrant more code?
I have disassembled the code to confirm this behavior.
I know this is not the hot path. But it actually can decrease
the code size.
Thanks.
From: Muchun Song <hidden> Date: 2021-03-11 09:14:54
On Thu, Mar 11, 2021 at 4:50 PM Michal Hocko [off-list ref] wrote:
On Thu 11-03-21 14:34:04, Muchun Song wrote:
quoted
On Wed, Mar 10, 2021 at 11:28 PM Michal Hocko [off-list ref] wrote:
quoted
On Mon 08-03-21 18:28:03, Muchun Song wrote:
quoted
Because we reuse the first tail vmemmap page frame and remap it
with read-only, we cannot set the PageHWPosion on some tail pages.
So we can use the head[4].private (There are at least 128 struct
page structures associated with the optimized HugeTLB page, so
using head[4].private is safe) to record the real error page index
and set the raw error page PageHWPoison later.
Can we have more poisoned tail pages? Also who does consume that index
and set the HWPoison on the proper tail page?
Good point. I look at the routine of memory failure closely.
If we do not clear the HWPoison of the head page, we cannot
poison another tail page.
So we should not set the destructor of the huge page from
HUGETLB_PAGE_DTOR to NULL_COMPOUND_DTOR
before calling alloc_huge_page_vmemmap(). In this case,
the below check of PageHuge() always returns true.
I need to fix this in the previous patch.
memory_failure()
if (PageHuge(page))
memory_failure_hugetlb()
head = compound_head(page)
if (TestSetPageHWPoison(head))
return
I have to say that I am not fully familiar with hwpoisoning code
(especially after recent changes) but IIRC it does rely on hugetlb page
dissolving. With the new code this operation can fail which is a new
situation. Unless I am misunderstanding this can lead to a lost memory
failure operation on other tail pages.
Anyway the above answers the question why a single slot is sufficient so
it would be great to mention that in a changelog along with the caveat
that some pages might miss their poisoning.
OK. I will update the changelog. Thanks for your suggestions.
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 09:40:03
On Thu 11-03-21 17:08:34, Muchun Song wrote:
On Thu, Mar 11, 2021 at 4:55 PM Michal Hocko [off-list ref] wrote:
quoted
On Thu 11-03-21 15:33:20, Muchun Song wrote:
quoted
On Wed, Mar 10, 2021 at 11:41 PM Michal Hocko [off-list ref] wrote:
quoted
On Mon 08-03-21 18:28:07, Muchun Song wrote:
quoted
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
I am confused. Don't you check for this in early_hugetlb_free_vmemmap_param already?
Right.
quoted
Why do we need any runtime checks?
If the size of the struct page is not power of 2, compiler can think
is_hugetlb_free_vmemmap_enabled() always return false. So
the code snippet of this user can be optimized away.
E.g.
if (is_hugetlb_free_vmemmap_enabled())
/* do something */
The compiler can drop "/* do something */" directly, because
it knows is_hugetlb_free_vmemmap_enabled() always returns
false.
OK, so this is a micro-optimization to generate a better code?
Right.
quoted
Is this measurable to warrant more code?
I have disassembled the code to confirm this behavior.
I know this is not the hot path. But it actually can decrease
the code size.
struct page which is not power of 2 is not a common case. Are you sure
it makes sense to micro optimize for an outliar. If you really want to
microptimize then do that for a common case - the feature being
disabled - via static key.
--
Michal Hocko
SUSE Labs
From: Muchun Song <hidden> Date: 2021-03-11 10:01:30
On Thu, Mar 11, 2021 at 5:39 PM Michal Hocko [off-list ref] wrote:
On Thu 11-03-21 17:08:34, Muchun Song wrote:
quoted
On Thu, Mar 11, 2021 at 4:55 PM Michal Hocko [off-list ref] wrote:
quoted
On Thu 11-03-21 15:33:20, Muchun Song wrote:
quoted
On Wed, Mar 10, 2021 at 11:41 PM Michal Hocko [off-list ref] wrote:
quoted
On Mon 08-03-21 18:28:07, Muchun Song wrote:
quoted
When the "struct page size" crosses page boundaries we cannot
make use of this feature. Let free_vmemmap_pages_per_hpage()
return zero if that is the case, most of the functions can be
optimized away.
I am confused. Don't you check for this in early_hugetlb_free_vmemmap_param already?
Right.
quoted
Why do we need any runtime checks?
If the size of the struct page is not power of 2, compiler can think
is_hugetlb_free_vmemmap_enabled() always return false. So
the code snippet of this user can be optimized away.
E.g.
if (is_hugetlb_free_vmemmap_enabled())
/* do something */
The compiler can drop "/* do something */" directly, because
it knows is_hugetlb_free_vmemmap_enabled() always returns
false.
OK, so this is a micro-optimization to generate a better code?
Right.
quoted
Is this measurable to warrant more code?
I have disassembled the code to confirm this behavior.
I know this is not the hot path. But it actually can decrease
the code size.
struct page which is not power of 2 is not a common case.
I know this is not a common case. But the check of
is_power_of_2(sizeof(struct page)) does not bring extra
runtime overhead. It just tells the compiler to optimize code
as much as possible.
Are you sure
it makes sense to micro optimize for an outliar. If you really want to
microptimize then do that for a common case - the feature being
disabled - via static key.
We cannot optimize the code size (vmlinux) even if we use a static
key when the size is not power of 2.
Sorry. I am confused why you disagree with this change.
It does not bring any disadvantages.
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 12:17:46
On Thu 11-03-21 18:00:09, Muchun Song wrote:
[...]
Sorry. I am confused why you disagree with this change.
It does not bring any disadvantages.
Because it is adding a code which is not really necessary and which will
have to be maintained. Think of future changes which would need to grow
more of these. Hugetlb code paths shouldn't really think about size of
the struct page.
--
Michal Hocko
SUSE Labs
From: Michal Hocko <mhocko@suse.com> Date: 2021-03-11 12:18:50
On Thu 11-03-21 09:40:57, Michal Hocko wrote:
On Wed 10-03-21 15:28:51, Paul E. McKenney wrote:
quoted
On Wed, Mar 10, 2021 at 02:10:12PM -0800, Mike Kravetz wrote:
quoted
On 3/10/21 1:49 PM, Paul E. McKenney wrote:
quoted
On Wed, Mar 10, 2021 at 10:11:22PM +0100, Michal Hocko wrote:
quoted
On Wed 10-03-21 10:56:08, Mike Kravetz wrote:
quoted
On 3/10/21 7:19 AM, Michal Hocko wrote:
quoted
On Mon 08-03-21 18:28:02, Muchun Song wrote:
[...]
quoted
@@ -1447,7 +1486,7 @@ void free_huge_page(struct page *page) /* * Defer freeing if in non-task context to avoid hugetlb_lock deadlock. */- if (!in_task()) {+ if (in_atomic()) {
As I've said elsewhere in_atomic doesn't work for CONFIG_PREEMPT_COUNT=n.
We need this change for other reasons and so it would be better to pull
it out into a separate patch which also makes HUGETLB depend on
PREEMPT_COUNT.
Yes, the issue of calling put_page for hugetlb pages from any context
still needs work. IMO, that is outside the scope of this series. We
already have code in this path which blocks/sleeps.
Making HUGETLB depend on PREEMPT_COUNT is too restrictive. IIUC,
PREEMPT_COUNT will only be enabled if we enable:
PREEMPT "Preemptible Kernel (Low-Latency Desktop)"
PREEMPT_RT "Fully Preemptible Kernel (Real-Time)"
or, other 'debug' options. These are not enabled in 'more common'
kernels. Of course, we do not want to disable HUGETLB in common
configurations.
I haven't tried that but PREEMPT_COUNT should be selectable even without
any change to the preemption model (e.g. !PREEMPT).
It works reliably for me, for example as in the diff below. So,
as Michal says, you should be able to add "select PREEMPT_COUNT" to
whatever Kconfig option you need to.
Thanks Paul.
I may have been misreading Michal's suggestion of "make HUGETLB depend on
PREEMPT_COUNT". We could "select PREEMPT_COUNT" if HUGETLB is enabled.
However, since HUGETLB is enabled in most configs, then this would
result in PREEMPT_COUNT also being enabled in most configs. I honestly
do not know how much this will cost us? I assume that if it was free or
really cheap it would already be always on?
There are a -lot- of configs out there, so are you sure that HUGETLB is
really enabled in most of them? ;-)
It certainly is enabled for all distribution kernels and many are
!PREEMPT so I believe this is what Mike was concerned about.
quoted
More seriously, I was going by earlier emails in this and related threads
plus Michal's "PREEMPT_COUNT should be selectable". But there are other
situations that would like PREEMPT_COUNT. And to your point, some who
would rather PREEMPT_COUNT not be universally enabled. I haven't seen
any performance or kernel-size numbers from any of them, however.
Yeah per cpu preempt counting shouldn't be noticeable but I have to
confess I haven't benchmarked it.
From: Muchun Song <hidden> Date: 2021-03-11 13:01:31
On Thu, Mar 11, 2021 at 8:16 PM Michal Hocko [off-list ref] wrote:
On Thu 11-03-21 18:00:09, Muchun Song wrote:
[...]
quoted
Sorry. I am confused why you disagree with this change.
It does not bring any disadvantages.
Because it is adding a code which is not really necessary and which will
have to be maintained. Think of future changes which would need to grow
more of these. Hugetlb code paths shouldn't really think about size of
the struct page.
From: Oscar Salvador <osalvador@suse.de> Date: 2021-03-11 13:46:26
On Thu, Mar 11, 2021 at 01:16:37PM +0100, Michal Hocko wrote:
On Thu 11-03-21 18:00:09, Muchun Song wrote:
[...]
quoted
Sorry. I am confused why you disagree with this change.
It does not bring any disadvantages.
Because it is adding a code which is not really necessary and which will
have to be maintained. Think of future changes which would need to grow
more of these. Hugetlb code paths shouldn't really think about size of
the struct page.
I have to confess that when I looked at the patch I found it nice in the way that
wipes out almost all clode dealing with vmemmap when sizeof(struct page) != power_of_2,
and I was convinced by the fact that only two places required the change.
So all in all it did not look like much churn, and not __that__ hard to maintain.
But I did not think in the case where this trick needs to be spread in more places
if the code changes over time.
So I agree that although it gets rid of a lot of code, it would seldomly pay off as
not many configuration out there are running on !power_of_2, and hugetlb is already
tricky enough.
--
Oscar Salvador
SUSE L3
The proper fix for free_huge_page independent of this series would
involve:
- Make hugetlb_lock and subpool lock irq safe
- Hand off freeing to a workque if the freeing could sleep
Today, the only time we can sleep in free_huge_page is for gigantic
pages allocated via cma. I 'think' the concern about undesirable
user visible side effects in this case is minimal as freeing/allocating
1G pages is not something that is going to happen at a high frequency.
My thinking could be wrong?
Of more concern, is the introduction of this series. If this feature
is enabled, then ALL free_huge_page requests must be sent to a workqueue.
Any ideas on how to address this?
--
Mike Kravetz
The proper fix for free_huge_page independent of this series would
involve:
- Make hugetlb_lock and subpool lock irq safe
- Hand off freeing to a workque if the freeing could sleep
Today, the only time we can sleep in free_huge_page is for gigantic
pages allocated via cma. I 'think' the concern about undesirable
user visible side effects in this case is minimal as freeing/allocating
1G pages is not something that is going to happen at a high frequency.
My thinking could be wrong?
Of more concern, is the introduction of this series. If this feature
is enabled, then ALL free_huge_page requests must be sent to a workqueue.
Any ideas on how to address this?
Thinking about this more ...
A call to free_huge_page has two distinct outcomes
1) Page is freed back to the original allocator: buddy or cma
2) Page is put on hugetlb free list
We can only possibly sleep in the first case 1. In addition, freeing a
page back to the original allocator involves these steps:
1) Removing page from hugetlb lists
2) Updating hugetlb counts: nr_hugepages, surplus
3) Updating page fields
4) Allocate vmemmap pages if needed as in this series
5) Calling free routine of original allocator
If hugetlb_lock is irq safe, we can perform the first 3 steps under that
lock without issue. We would then use a workqueue to perform the last
two steps. Since we are updating hugetlb user visible data under the
lock, there should be no delays. Of course, giving those pages back to
the original allocator could still be delayed, and a user may notice
that. Not sure if that would be acceptable? I think Muchun had a
similar setup just for vmemmmap allocation in an early version of this
series.
This would also require changes to where accounting is done in
dissolve_free_huge_page and update_and_free_page as mentioned elsewhere.
P.S. We could further optimize to check for the possibility of sleeping
(cma or vmemmap) and only send to workqueue in those cases.
--
Mike Kravetz
The proper fix for free_huge_page independent of this series would
involve:
- Make hugetlb_lock and subpool lock irq safe
- Hand off freeing to a workque if the freeing could sleep
Today, the only time we can sleep in free_huge_page is for gigantic
pages allocated via cma. I 'think' the concern about undesirable
user visible side effects in this case is minimal as freeing/allocating
1G pages is not something that is going to happen at a high frequency.
My thinking could be wrong?
Of more concern, is the introduction of this series. If this feature
is enabled, then ALL free_huge_page requests must be sent to a workqueue.
Any ideas on how to address this?
Thinking about this more ...
A call to free_huge_page has two distinct outcomes
1) Page is freed back to the original allocator: buddy or cma
2) Page is put on hugetlb free list
We can only possibly sleep in the first case 1. In addition, freeing a
page back to the original allocator involves these steps:
1) Removing page from hugetlb lists
2) Updating hugetlb counts: nr_hugepages, surplus
3) Updating page fields
4) Allocate vmemmap pages if needed as in this series
5) Calling free routine of original allocator
If hugetlb_lock is irq safe, we can perform the first 3 steps under that
lock without issue. We would then use a workqueue to perform the last
two steps. Since we are updating hugetlb user visible data under the
lock, there should be no delays. Of course, giving those pages back to
the original allocator could still be delayed, and a user may notice
that. Not sure if that would be acceptable?
Well, having many in-flight huge pages can certainly be visible. Say you
are freeing hundreds of huge pages and your echo n > nr_hugepages will
return just for you to find out that the memory hasn't been freed and
therefore cannot be reused for another use - recently there was somebody
mentioning their usecase to free up huge pages to prevent OOM for
example. I do expect more people doing something like that.
Now, nr_hugepages can be handled by blocking on the same WQ until all
pre-existing items are processed. Maybe we will need to have a more
generic API to achieve the same for in kernel users but let's wait for
those requests.
I think Muchun had a
similar setup just for vmemmmap allocation in an early version of this
series.
This would also require changes to where accounting is done in
dissolve_free_huge_page and update_and_free_page as mentioned elsewhere.
Normalizing dissolve_free_huge_page is definitely a good idea. It is
really tricky how it sticks out and does half of the job of
update_and_free_page.
That being said, if it is possible to have a fully consistent h state
before handing over to WQ for sleeping operation then we should be all
fine. I am slightly worried about potential tricky situations where the
sleeping operation fails because that would require that page to be
added back to the pool again. As said above we would need some sort of
sync with in-flight operations before returning to the userspace.
--
Michal Hocko
SUSE Labs
The proper fix for free_huge_page independent of this series would
involve:
- Make hugetlb_lock and subpool lock irq safe
- Hand off freeing to a workque if the freeing could sleep
Today, the only time we can sleep in free_huge_page is for gigantic
pages allocated via cma. I 'think' the concern about undesirable
user visible side effects in this case is minimal as freeing/allocating
1G pages is not something that is going to happen at a high frequency.
My thinking could be wrong?
Of more concern, is the introduction of this series. If this feature
is enabled, then ALL free_huge_page requests must be sent to a workqueue.
Any ideas on how to address this?
Thinking about this more ...
A call to free_huge_page has two distinct outcomes
1) Page is freed back to the original allocator: buddy or cma
2) Page is put on hugetlb free list
We can only possibly sleep in the first case 1. In addition, freeing a
page back to the original allocator involves these steps:
1) Removing page from hugetlb lists
2) Updating hugetlb counts: nr_hugepages, surplus
3) Updating page fields
4) Allocate vmemmap pages if needed as in this series
5) Calling free routine of original allocator
If hugetlb_lock is irq safe, we can perform the first 3 steps under that
lock without issue. We would then use a workqueue to perform the last
two steps. Since we are updating hugetlb user visible data under the
lock, there should be no delays. Of course, giving those pages back to
the original allocator could still be delayed, and a user may notice
that. Not sure if that would be acceptable?
Well, having many in-flight huge pages can certainly be visible. Say you
are freeing hundreds of huge pages and your echo n > nr_hugepages will
return just for you to find out that the memory hasn't been freed and
therefore cannot be reused for another use - recently there was somebody
mentioning their usecase to free up huge pages to prevent OOM for
example. I do expect more people doing something like that.
Now, nr_hugepages can be handled by blocking on the same WQ until all
pre-existing items are processed. Maybe we will need to have a more
generic API to achieve the same for in kernel users but let's wait for
those requests.
quoted
I think Muchun had a
similar setup just for vmemmmap allocation in an early version of this
series.
This would also require changes to where accounting is done in
dissolve_free_huge_page and update_and_free_page as mentioned elsewhere.
Normalizing dissolve_free_huge_page is definitely a good idea. It is
really tricky how it sticks out and does half of the job of
update_and_free_page.
That being said, if it is possible to have a fully consistent h state
before handing over to WQ for sleeping operation then we should be all
fine. I am slightly worried about potential tricky situations where the
sleeping operation fails because that would require that page to be
added back to the pool again. As said above we would need some sort of
sync with in-flight operations before returning to the userspace.
Those sysfs interfaces to allocate/free huge pages will need to be
reworked. One thing that is totally unacceptable with hugetlb_lock
being irq safe, are the calls to cond_resched_lock(&hugetlb_lock).
We will need to significantly reduce lock hold time in these situations.
I have some ideas on how this might work, but it is going to require
some a good deal of code restructuring and will take some time.
--
Mike Kravetz