| Message ID | 20210425070752.17783-1-songmuchun@bytedance.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Free some vmemmap pages of HugeTLB page | expand |
Thanks! I will take a look at the modifications soon. I applied the patches to Andrew's mmotm-2021-04-21-23-03, ran some tests and got the following warning. We may need to special case that call to __prep_new_huge_page/free_huge_page_vmemmap from alloc_and_dissolve_huge_page as it is holding hugetlb lock with IRQs disabled. Sorry I missed that previously.
On Wed, Apr 28, 2021 at 7:47 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: > > Thanks! I will take a look at the modifications soon. > > I applied the patches to Andrew's mmotm-2021-04-21-23-03, ran some tests and > got the following warning. We may need to special case that call to > __prep_new_huge_page/free_huge_page_vmemmap from alloc_and_dissolve_huge_page > as it is holding hugetlb lock with IRQs disabled. Good catch. Thanks Mike. I will fix it in the next version. How about this: @@ -1618,7 +1617,8 @@ static void __prep_new_huge_page(struct hstate *h, struct page *page) static void prep_new_huge_page(struct hstate *h, struct page *page, int nid) { + free_huge_page_vmemmap(h, page); __prep_new_huge_page(page); spin_lock_irq(&hugetlb_lock); __prep_account_new_huge_page(h, nid); spin_unlock_irq(&hugetlb_lock); @@ -2429,6 +2429,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, if (!new_page) return -ENOMEM; + free_huge_page_vmemmap(h, new_page); retry: spin_lock_irq(&hugetlb_lock); if (!PageHuge(old_page)) { @@ -2489,7 +2490,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, free_new: spin_unlock_irq(&hugetlb_lock); - __free_pages(new_page, huge_page_order(h)); + update_and_free_page(h, new_page, false); return ret; } > > Sorry I missed that previously. > -- > Mike Kravetz > > [ 1521.579890] ------------[ cut here ]------------ > [ 1521.581309] WARNING: CPU: 1 PID: 1046 at kernel/smp.c:884 smp_call_function_many_cond+0x1bb/0x390 > [ 1521.583895] Modules linked in: ip6t_rpfilter ip6t_REJECT nf_reject_ipv6 xt_conntrack ebtable_nat ip6table_nat ip6table_mangle ip6table_raw ip6table_security iptable_nat nf_nat iptable_mangle iptable_raw iptable_security nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 rfkill ebtable_filter ebtables 9p ip6table_filter ip6_tables sunrpc snd_hda_codec_generic snd_hda_intel snd_intel_dspcfg snd_hda_codec snd_hwdep snd_hda_core snd_seq joydev crct10dif_pclmul snd_seq_device crc32_pclmul snd_pcm ghash_clmulni_intel snd_timer 9pnet_virtio snd 9pnet virtio_balloon soundcore i2c_piix4 virtio_net virtio_console net_failover virtio_blk failover 8139too qxl drm_ttm_helper ttm drm_kms_helper drm crc32c_intel serio_raw virtio_pci virtio_pci_modern_dev 8139cp virtio_ring mii ata_generic virtio pata_acpi > [ 1521.598644] CPU: 1 PID: 1046 Comm: bash Not tainted 5.12.0-rc8-mm1+ #2 > [ 1521.599787] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-1.fc33 04/01/2014 > [ 1521.601259] RIP: 0010:smp_call_function_many_cond+0x1bb/0x390 > [ 1521.602232] Code: 87 75 71 01 85 d2 0f 84 c8 fe ff ff 65 8b 05 94 3d e9 7e 85 c0 0f 85 b9 fe ff ff 65 8b 05 f9 3a e8 7e 85 c0 0f 85 aa fe ff ff <0f> 0b e9 a3 fe ff ff 65 8b 05 47 33 e8 7e a9 ff ff ff 7f 0f 85 75 > [ 1521.605167] RSP: 0018:ffffc90001fcb928 EFLAGS: 00010046 > [ 1521.606049] RAX: 0000000000000000 RBX: ffffffff828a85d0 RCX: 0000000000000001 > [ 1521.607103] RDX: 0000000000000001 RSI: 0000000000000000 RDI: 0000000000000001 > [ 1521.608127] RBP: 0000000000000000 R08: 0000000000000000 R09: ffffea0008fa6f88 > [ 1521.609144] R10: 0000000000000001 R11: 0000000000000001 R12: ffff888237d3bfc0 > [ 1521.610112] R13: dead000000000122 R14: dead000000000100 R15: ffffea0007bb8000 > [ 1521.611106] FS: 00007f8a11223740(0000) GS:ffff888237d00000(0000) knlGS:0000000000000000 > [ 1521.612231] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 1521.612952] CR2: 0000555e1d00a430 CR3: 000000019ef5a005 CR4: 0000000000370ee0 > [ 1521.614295] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > [ 1521.615539] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 > [ 1521.616814] Call Trace: > [ 1521.617241] ? flush_tlb_one_kernel+0x20/0x20 > [ 1521.618041] on_each_cpu_cond_mask+0x25/0x30 > [ 1521.618797] flush_tlb_kernel_range+0xa5/0xc0 > [ 1521.619577] vmemmap_remap_free+0x7d/0x150 > [ 1521.620319] ? sparse_remove_section+0x80/0x80 > [ 1521.621120] free_huge_page_vmemmap+0x2f/0x40 > [ 1521.621898] __prep_new_huge_page+0xe/0xd0 > [ 1521.622633] isolate_or_dissolve_huge_page+0x300/0x360 > [ 1521.623559] isolate_migratepages_block+0x4c4/0xe20 > [ 1521.624430] ? verify_cpu+0x100/0x100 > [ 1521.625096] isolate_migratepages_range+0x6b/0xc0 > [ 1521.625936] alloc_contig_range+0x220/0x3d0 > [ 1521.626729] cma_alloc+0x1ae/0x5f0 > [ 1521.627333] alloc_fresh_huge_page+0x67/0x190 > [ 1521.628054] alloc_pool_huge_page+0x72/0xf0 > [ 1521.628769] set_max_huge_pages+0x128/0x2c0 > [ 1521.629540] __nr_hugepages_store_common+0x3d/0xb0 > [ 1521.630457] ? _kstrtoull+0x35/0xd0 > [ 1521.631182] nr_hugepages_store+0x73/0x80 > [ 1521.631903] kernfs_fop_write_iter+0x127/0x1c0 > [ 1521.632698] new_sync_write+0x11f/0x1b0 > [ 1521.633408] vfs_write+0x26f/0x380 > [ 1521.633946] ksys_write+0x68/0xe0 > [ 1521.634444] do_syscall_64+0x40/0x80 > [ 1521.634914] entry_SYSCALL_64_after_hwframe+0x44/0xae > [ 1521.635669] RIP: 0033:0x7f8a11313ff8 > [ 1521.636251] Code: 89 02 48 c7 c0 ff ff ff ff eb b3 0f 1f 80 00 00 00 00 f3 0f 1e fa 48 8d 05 25 77 0d 00 8b 00 85 c0 75 17 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 58 c3 0f 1f 80 00 00 00 00 41 54 49 89 d4 55 > [ 1521.639758] RSP: 002b:00007ffd26f79b18 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 > [ 1521.641118] RAX: ffffffffffffffda RBX: 0000000000000002 RCX: 00007f8a11313ff8 > [ 1521.642425] RDX: 0000000000000002 RSI: 0000555e1cf94960 RDI: 0000000000000001 > [ 1521.643644] RBP: 0000555e1cf94960 R08: 000000000000000a R09: 00007f8a113a5e80 > [ 1521.644904] R10: 000000000000000a R11: 0000000000000246 R12: 00007f8a113e7780 > [ 1521.646177] R13: 0000000000000002 R14: 00007f8a113e2740 R15: 0000000000000002 > [ 1521.647450] irq event stamp: 10006640 > [ 1521.648103] hardirqs last enabled at (10006639): [<ffffffff812ad02b>] bad_range+0x15b/0x180 > [ 1521.649577] hardirqs last disabled at (10006640): [<ffffffff81abcea1>] _raw_spin_lock_irq+0x51/0x60 > [ 1521.651194] softirqs last enabled at (10006630): [<ffffffff810da5e2>] __irq_exit_rcu+0xd2/0x100 > [ 1521.652763] softirqs last disabled at (10006625): [<ffffffff810da5e2>] __irq_exit_rcu+0xd2/0x100 > [ 1521.654251] ---[ end trace 561fa19f90280f2f ]---
On 4/28/21 5:26 AM, Muchun Song wrote: > On Wed, Apr 28, 2021 at 7:47 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: >> >> Thanks! I will take a look at the modifications soon. >> >> I applied the patches to Andrew's mmotm-2021-04-21-23-03, ran some tests and >> got the following warning. We may need to special case that call to >> __prep_new_huge_page/free_huge_page_vmemmap from alloc_and_dissolve_huge_page >> as it is holding hugetlb lock with IRQs disabled. > > Good catch. Thanks Mike. I will fix it in the next version. How about this: > > @@ -1618,7 +1617,8 @@ static void __prep_new_huge_page(struct hstate > *h, struct page *page) > > static void prep_new_huge_page(struct hstate *h, struct page *page, int nid) > { > + free_huge_page_vmemmap(h, page); > __prep_new_huge_page(page); > spin_lock_irq(&hugetlb_lock); > __prep_account_new_huge_page(h, nid); > spin_unlock_irq(&hugetlb_lock); > @@ -2429,6 +2429,7 @@ static int alloc_and_dissolve_huge_page(struct > hstate *h, struct page *old_page, > if (!new_page) > return -ENOMEM; > > + free_huge_page_vmemmap(h, new_page); > retry: > spin_lock_irq(&hugetlb_lock); > if (!PageHuge(old_page)) { > @@ -2489,7 +2490,7 @@ static int alloc_and_dissolve_huge_page(struct > hstate *h, struct page *old_page, > > free_new: > spin_unlock_irq(&hugetlb_lock); > - __free_pages(new_page, huge_page_order(h)); > + update_and_free_page(h, new_page, false); > > return ret; > } > > Another option would be to leave the prep* routines as is and only modify alloc_and_dissolve_huge_page as follows: diff --git a/mm/hugetlb.c b/mm/hugetlb.c index 9c617c19fc18..f8e5013a6b46 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -2420,14 +2420,15 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, /* * Before dissolving the page, we need to allocate a new one for the - * pool to remain stable. Using alloc_buddy_huge_page() allows us to - * not having to deal with prep_new_huge_page() and avoids dealing of any - * counters. This simplifies and let us do the whole thing under the - * lock. + * pool to remain stable. Here, we allocate the page and 'prep' it + * by doing everything but actually updating counters and adding to + * the pool. This simplifies and let us do most of the processing + * under the lock. */ new_page = alloc_buddy_huge_page(h, gfp_mask, nid, NULL, NULL); if (!new_page) return -ENOMEM; + __prep_new_huge_page(h, new_page); retry: spin_lock_irq(&hugetlb_lock); @@ -2473,7 +2474,6 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, * Reference count trick is needed because allocator gives us * referenced page but the pool requires pages with 0 refcount. */ - __prep_new_huge_page(h, new_page); __prep_account_new_huge_page(h, nid); page_ref_dec(new_page); enqueue_huge_page(h, new_page); @@ -2489,7 +2489,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, free_new: spin_unlock_irq(&hugetlb_lock); - __free_pages(new_page, huge_page_order(h)); + update_and_free_page(h, old_page, false); return ret; }
On Thu, Apr 29, 2021 at 10:32 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: > > On 4/28/21 5:26 AM, Muchun Song wrote: > > On Wed, Apr 28, 2021 at 7:47 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: > >> > >> Thanks! I will take a look at the modifications soon. > >> > >> I applied the patches to Andrew's mmotm-2021-04-21-23-03, ran some tests and > >> got the following warning. We may need to special case that call to > >> __prep_new_huge_page/free_huge_page_vmemmap from alloc_and_dissolve_huge_page > >> as it is holding hugetlb lock with IRQs disabled. > > > > Good catch. Thanks Mike. I will fix it in the next version. How about this: > > > > @@ -1618,7 +1617,8 @@ static void __prep_new_huge_page(struct hstate > > *h, struct page *page) > > > > static void prep_new_huge_page(struct hstate *h, struct page *page, int nid) > > { > > + free_huge_page_vmemmap(h, page); > > __prep_new_huge_page(page); > > spin_lock_irq(&hugetlb_lock); > > __prep_account_new_huge_page(h, nid); > > spin_unlock_irq(&hugetlb_lock); > > @@ -2429,6 +2429,7 @@ static int alloc_and_dissolve_huge_page(struct > > hstate *h, struct page *old_page, > > if (!new_page) > > return -ENOMEM; > > > > + free_huge_page_vmemmap(h, new_page); > > retry: > > spin_lock_irq(&hugetlb_lock); > > if (!PageHuge(old_page)) { > > @@ -2489,7 +2490,7 @@ static int alloc_and_dissolve_huge_page(struct > > hstate *h, struct page *old_page, > > > > free_new: > > spin_unlock_irq(&hugetlb_lock); > > - __free_pages(new_page, huge_page_order(h)); > > + update_and_free_page(h, new_page, false); > > > > return ret; > > } > > > > > > Another option would be to leave the prep* routines as is and only > modify alloc_and_dissolve_huge_page as follows: OK. LGTM. I will use this. Thanks Mike. > > diff --git a/mm/hugetlb.c b/mm/hugetlb.c > index 9c617c19fc18..f8e5013a6b46 100644 > --- a/mm/hugetlb.c > +++ b/mm/hugetlb.c > @@ -2420,14 +2420,15 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, > > /* > * Before dissolving the page, we need to allocate a new one for the > - * pool to remain stable. Using alloc_buddy_huge_page() allows us to > - * not having to deal with prep_new_huge_page() and avoids dealing of any > - * counters. This simplifies and let us do the whole thing under the > - * lock. > + * pool to remain stable. Here, we allocate the page and 'prep' it > + * by doing everything but actually updating counters and adding to > + * the pool. This simplifies and let us do most of the processing > + * under the lock. > */ > new_page = alloc_buddy_huge_page(h, gfp_mask, nid, NULL, NULL); > if (!new_page) > return -ENOMEM; > + __prep_new_huge_page(h, new_page); > > retry: > spin_lock_irq(&hugetlb_lock); > @@ -2473,7 +2474,6 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, > * Reference count trick is needed because allocator gives us > * referenced page but the pool requires pages with 0 refcount. > */ > - __prep_new_huge_page(h, new_page); > __prep_account_new_huge_page(h, nid); > page_ref_dec(new_page); > enqueue_huge_page(h, new_page); > @@ -2489,7 +2489,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, > > free_new: > spin_unlock_irq(&hugetlb_lock); > - __free_pages(new_page, huge_page_order(h)); > + update_and_free_page(h, old_page, false); > > return ret; > } > > -- > Mike Kravetz
On 4/28/21 9:02 PM, Muchun Song wrote: > On Thu, Apr 29, 2021 at 10:32 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: >> >> On 4/28/21 5:26 AM, Muchun Song wrote: >>> On Wed, Apr 28, 2021 at 7:47 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: >>>> >>>> Thanks! I will take a look at the modifications soon. >>>> >>>> I applied the patches to Andrew's mmotm-2021-04-21-23-03, ran some tests and >>>> got the following warning. We may need to special case that call to >>>> __prep_new_huge_page/free_huge_page_vmemmap from alloc_and_dissolve_huge_page >>>> as it is holding hugetlb lock with IRQs disabled. >>> >>> Good catch. Thanks Mike. I will fix it in the next version. How about this: >>> >>> @@ -1618,7 +1617,8 @@ static void __prep_new_huge_page(struct hstate >>> *h, struct page *page) >>> >>> static void prep_new_huge_page(struct hstate *h, struct page *page, int nid) >>> { >>> + free_huge_page_vmemmap(h, page); >>> __prep_new_huge_page(page); >>> spin_lock_irq(&hugetlb_lock); >>> __prep_account_new_huge_page(h, nid); >>> spin_unlock_irq(&hugetlb_lock); >>> @@ -2429,6 +2429,7 @@ static int alloc_and_dissolve_huge_page(struct >>> hstate *h, struct page *old_page, >>> if (!new_page) >>> return -ENOMEM; >>> >>> + free_huge_page_vmemmap(h, new_page); >>> retry: >>> spin_lock_irq(&hugetlb_lock); >>> if (!PageHuge(old_page)) { >>> @@ -2489,7 +2490,7 @@ static int alloc_and_dissolve_huge_page(struct >>> hstate *h, struct page *old_page, >>> >>> free_new: >>> spin_unlock_irq(&hugetlb_lock); >>> - __free_pages(new_page, huge_page_order(h)); >>> + update_and_free_page(h, new_page, false); >>> >>> return ret; >>> } >>> >>> >> >> Another option would be to leave the prep* routines as is and only >> modify alloc_and_dissolve_huge_page as follows: > > OK. LGTM. I will use this. Thanks Mike. There are issues with my suggested patch below. I am occasionally hitting the BUG that checks for page ref count being zero at put_page time. Still do not fully understand, but I do not hit the same BUG with your patch above. Please do not use my patch below.
On 4/29/21 3:23 PM, Mike Kravetz wrote: > On 4/28/21 9:02 PM, Muchun Song wrote: >> On Thu, Apr 29, 2021 at 10:32 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: >>> >>> On 4/28/21 5:26 AM, Muchun Song wrote: >>>> On Wed, Apr 28, 2021 at 7:47 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: >>>>> >>>>> Thanks! I will take a look at the modifications soon. >>>>> >>>>> I applied the patches to Andrew's mmotm-2021-04-21-23-03, ran some tests and >>>>> got the following warning. We may need to special case that call to >>>>> __prep_new_huge_page/free_huge_page_vmemmap from alloc_and_dissolve_huge_page >>>>> as it is holding hugetlb lock with IRQs disabled. >>>> >>>> Good catch. Thanks Mike. I will fix it in the next version. How about this: >>>> >>>> @@ -1618,7 +1617,8 @@ static void __prep_new_huge_page(struct hstate >>>> *h, struct page *page) >>>> >>>> static void prep_new_huge_page(struct hstate *h, struct page *page, int nid) >>>> { >>>> + free_huge_page_vmemmap(h, page); >>>> __prep_new_huge_page(page); >>>> spin_lock_irq(&hugetlb_lock); >>>> __prep_account_new_huge_page(h, nid); >>>> spin_unlock_irq(&hugetlb_lock); >>>> @@ -2429,6 +2429,7 @@ static int alloc_and_dissolve_huge_page(struct >>>> hstate *h, struct page *old_page, >>>> if (!new_page) >>>> return -ENOMEM; >>>> >>>> + free_huge_page_vmemmap(h, new_page); >>>> retry: >>>> spin_lock_irq(&hugetlb_lock); >>>> if (!PageHuge(old_page)) { >>>> @@ -2489,7 +2490,7 @@ static int alloc_and_dissolve_huge_page(struct >>>> hstate *h, struct page *old_page, >>>> >>>> free_new: >>>> spin_unlock_irq(&hugetlb_lock); >>>> - __free_pages(new_page, huge_page_order(h)); >>>> + update_and_free_page(h, new_page, false); >>>> >>>> return ret; >>>> } >>>> >>>> >>> >>> Another option would be to leave the prep* routines as is and only >>> modify alloc_and_dissolve_huge_page as follows: >> >> OK. LGTM. I will use this. Thanks Mike. > > There are issues with my suggested patch below. I am occasionally > hitting the BUG that checks for page ref count being zero at put_page > time. Still do not fully understand, but I do not hit the same BUG > with your patch above. Please do not use my patch below. > Ah! The issue is pretty obvious. > @@ -2489,7 +2489,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page, > > free_new: > spin_unlock_irq(&hugetlb_lock); > - __free_pages(new_page, huge_page_order(h)); > + update_and_free_page(h, old_page, false); That should of course be: update_and_free_page(h, new_page, false); > > return ret; > } >
Hi, Since Mike's patches (make hugetlb put_page safe for all calling contexts[1]) applied into the next-20210412. We can move forward on this patch series now. 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[2], but with that in added it shows even more: unpin_user_pages(): ~27k -> ~3.8k [1] https://lore.kernel.org/linux-mm/20210409205254.242291-1-mike.kravetz@oracle.com/ [2] 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 v20 -> v21: - Collect Reviewed-by tags. - Rebase to next-20210421 which is on the top of Oscar's changes. - Use size_to_hstate() directly in free_hpage_workfn() suggested by Mike. - Add a comment above alloc_huge_page_vmemmap() in dissolve_free_huge_page() suggested by Mike. - Trim mhp_supports_memmap_on_memory() suggested by Oscar. Thanks to Mike and Oscar's suggestions. Changelog in v19 -> v20: - Rebase to next-20210412. - Introduce workqueue to defer freeing HugeTLB pages. - Remove all tags (Reviewed-by ot Tested-by) from patch 6. - Disable memmap_on_memory when hugetlb_free_vmemmap enabled (patch 8). Changelog in v18 -> v19: - Fix compiler error on sparc arch. Thanks Stephen. - Make patch "gather discrete indexes of tail page" prior to "free the vmemmap pages associated with each HugeTLB page". - Remove some BUG_ON from patch #4. - Update patch #6 changelog. - Update Documentation/admin-guide/mm/memory-hotplug.rst. - Drop the patch of "optimize the code with the help of the compiler". - Update Documentation/admin-guide/kernel-parameters.txt in patch #7. - Trim update_and_free_page. Thanks to Michal, Oscar and Mike's review and suggestions. 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: gather discrete indexes of tail page mm: hugetlb: free the vmemmap pages associated with each HugeTLB page mm: hugetlb: defer freeing of HugeTLB pages mm: hugetlb: alloc the vmemmap pages associated with each HugeTLB page mm: hugetlb: add a kernel parameter hugetlb_free_vmemmap mm: memory_hotplug: disable memmap_on_memory when hugetlb_free_vmemmap enabled mm: hugetlb: introduce nr_free_vmemmap_pages in the struct hstate Documentation/admin-guide/kernel-parameters.txt | 25 ++ Documentation/admin-guide/mm/hugetlbpage.rst | 11 + Documentation/admin-guide/mm/memory-hotplug.rst | 13 ++ arch/sparc/mm/init_64.c | 1 + arch/x86/mm/init_64.c | 13 +- drivers/acpi/acpi_memhotplug.c | 1 + fs/Kconfig | 5 + include/linux/bootmem_info.h | 66 ++++++ include/linux/hugetlb.h | 46 +++- 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 | 184 +++++++++++++-- mm/hugetlb_vmemmap.c | 297 ++++++++++++++++++++++++ mm/hugetlb_vmemmap.h | 45 ++++ mm/memory_hotplug.c | 117 +--------- mm/sparse-vmemmap.c | 267 +++++++++++++++++++++ mm/sparse.c | 1 + 20 files changed, 1093 insertions(+), 179 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