When hugetlb vmemmap optimization was introduced, the overhead of enabling
the option was measured as described in commit 426e5c429d16 [1]. The summary
states that allocating a hugetlb page should be ~2x slower with optimization
and freeing a hugetlb page should be ~2-3x slower. Such overhead was deemed
an acceptable trade off for the memory savings obtained by freeing vmemmap
pages.
It was recently reported that the overhead associated with enabling vmemmap
optimization could be as high as 190x for hugetlb page allocations.
Yes, 190x! Some actual numbers from other environments are:
Bare Metal 8 socket Intel(R) Xeon(R) CPU E7-8895
------------------------------------------------
Unmodified next-20230824, vm.hugetlb_optimize_vmemmap = 0
time echo 500000 > .../hugepages-2048kB/nr_hugepages
real 0m4.119s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m4.477s
Unmodified next-20230824, vm.hugetlb_optimize_vmemmap = 1
time echo 500000 > .../hugepages-2048kB/nr_hugepages
real 0m28.973s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m36.748s
VM with 252 vcpus on host with 2 socket AMD EPYC 7J13 Milan
-----------------------------------------------------------
Unmodified next-20230824, vm.hugetlb_optimize_vmemmap = 0
time echo 524288 > .../hugepages-2048kB/nr_hugepages
real 0m2.463s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m2.931s
Unmodified next-20230824, vm.hugetlb_optimize_vmemmap = 1
time echo 524288 > .../hugepages-2048kB/nr_hugepages
real 2m27.609s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 2m29.924s
In the VM environment, the slowdown of enabling hugetlb vmemmap optimization
resulted in allocation times being 61x slower.
A quick profile showed that the vast majority of this overhead was due to
TLB flushing. Each time we modify the kernel pagetable we need to flush
the TLB. For each hugetlb that is optimized, there could be potentially
two TLB flushes performed. One for the vmemmap pages associated with the
hugetlb page, and potentially another one if the vmemmap pages are mapped
at the PMD level and must be split. The TLB flushes required for the kernel
pagetable, result in a broadcast IPI with each CPU having to flush a range
of pages, or do a global flush if a threshold is exceeded. So, the flush
time increases with the number of CPUs. In addition, in virtual environments
the broadcast IPI can’t be accelerated by hypervisor hardware and leads to
traps that need to wakeup/IPI all vCPUs which is very expensive. Because of
this the slowdown in virtual environments is even worse than bare metal as
the number of vCPUS/CPUs is increased.
The following series attempts to reduce amount of time spent in TLB flushing.
The idea is to batch the vmemmap modification operations for multiple hugetlb
pages. Instead of doing one or two TLB flushes for each page, we do two TLB
flushes for each batch of pages. One flush after splitting pages mapped at
the PMD level, and another after remapping vmemmap associated with all
hugetlb pages. Results of such batching are as follows:
Bare Metal 8 socket Intel(R) Xeon(R) CPU E7-8895
------------------------------------------------
next-20230824 + Batching patches, vm.hugetlb_optimize_vmemmap = 0
time echo 500000 > .../hugepages-2048kB/nr_hugepages
real 0m4.719s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m4.245s
next-20230824 + Batching patches, vm.hugetlb_optimize_vmemmap = 1
time echo 500000 > .../hugepages-2048kB/nr_hugepages
real 0m7.267s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m13.199s
VM with 252 vcpus on host with 2 socket AMD EPYC 7J13 Milan
-----------------------------------------------------------
next-20230824 + Batching patches, vm.hugetlb_optimize_vmemmap = 0
time echo 524288 > .../hugepages-2048kB/nr_hugepages
real 0m2.715s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m3.186s
next-20230824 + Batching patches, vm.hugetlb_optimize_vmemmap = 1
time echo 524288 > .../hugepages-2048kB/nr_hugepages
real 0m4.799s
time echo 0 > .../hugepages-2048kB/nr_hugepages
real 0m5.273s
With batching, results are back in the 2-3x slowdown range.
This series is based on mm-unstable (September 24)
Changes v4 -> v5:
- patch 3 comment style updated, unnecessary INIT_LIST_HEAD
- patch 4 updated hugetlb_vmemmap_restore_folios to pass back number of
restored folios in non-error case. In addition, routine passes back
list of folios with vmemmmap. Naming more consistent.
- patch 5 remover over optimization and added Muchun RB
- patch 6 break and early return in ENOMEM case. Updated comments.
Added Muchun RB.
- patch 7 Updated comments about splitting failure. Added Muchun RB.
- patch 8 Made comments consistent.
Changes v3 -> v4:
- Rebased on mm-unstable and dropped requisite patches.
- patch 2 updated to take bootmem vmemmap initialization into account
- patch 3 more changes for bootmem hugetlb pages. added routine
prep_and_add_bootmem_folios.
- patch 5 in hugetlb_vmemmap_optimize_folios on ENOMEM check for
list_empty before freeing and retry. This is more important in
subsequent patch where we flush_tlb_all after ENOMEM.
Changes v2 -> v3:
- patch 5 was part of an earlier series that was not picked up. It is
included here as it helps with batching optimizations.
- patch 6 hugetlb_vmemmap_restore_folios is changed from type void to
returning an error code as well as an additional output parameter providing
the number folios for which vmemmap was actually restored. The caller can
then be more intelligent about processing the list.
- patch 9 eliminate local list in vmemmap_restore_pte. The routine
hugetlb_vmemmap_optimize_folios checks for ENOMEM and frees accumulated
vmemmap pages while processing the list.
- patch 10 introduce flags field to struct vmemmap_remap_walk and
VMEMMAP_SPLIT_NO_TLB_FLUSH for not flushing during pass to split PMDs.
- patch 11 rename flag VMEMMAP_REMAP_NO_TLB_FLUSH and pass in from callers.
Changes v1 -> v2:
- patch 5 now takes into account the requirement that only compound
pages with hugetlb flag set can be passed to vmemmmap routines. This
involved separating the 'prep' of hugetlb pages even further. The code
dealing with bootmem allocations was also modified so that batching is
possible. Adding a 'batch' of hugetlb pages to their respective free
lists is now done in one lock cycle.
- patch 7 added description of routine hugetlb_vmemmap_restore_folios
(Muchun).
- patch 8 rename bulk_pages to vmemmap_pages and let caller be responsible
for freeing (Muchun)
- patch 9 use 'walk->remap_pte' to determine if a split only operation
is being performed (Muchun). Removed unused variable and
hugetlb_optimize_vmemmap_key (Muchun).
- Patch 10 pass 'flags variable' instead of bool to indicate behavior and
allow for future expansion (Muchun). Single flag VMEMMAP_NO_TLB_FLUSH.
Provide detailed comment about the need to keep old and new vmemmap pages
in sync (Muchun).
- Patch 11 pass flag variable as in patch 10 (Muchun).
Joao Martins (2):
hugetlb: batch PMD split for bulk vmemmap dedup
hugetlb: batch TLB flushes when freeing vmemmap
Mike Kravetz (6):
hugetlb: optimize update_and_free_pages_bulk to avoid lock cycles
hugetlb: restructure pool allocations
hugetlb: perform vmemmap optimization on a list of pages
hugetlb: perform vmemmap restoration on a list of pages
hugetlb: batch freeing of vmemmap pages
hugetlb: batch TLB flushes when restoring vmemmap
mm/hugetlb.c | 300 ++++++++++++++++++++++++++++++++++++-------
mm/hugetlb_vmemmap.c | 273 +++++++++++++++++++++++++++++++++------
mm/hugetlb_vmemmap.h | 15 +++
3 files changed, 505 insertions(+), 83 deletions(-)
--
2.41.0
From: Joao Martins <[email protected]>
In an effort to minimize amount of TLB flushes, batch all PMD splits
belonging to a range of pages in order to perform only 1 (global) TLB
flush.
Add a flags field to the walker and pass whether it's a bulk allocation
or just a single page to decide to remap. First value
(VMEMMAP_SPLIT_NO_TLB_FLUSH) designates the request to not do the TLB
flush when we split the PMD.
Rebased and updated by Mike Kravetz
Signed-off-by: Joao Martins <[email protected]>
Signed-off-by: Mike Kravetz <[email protected]>
Reviewed-by: Muchun Song <[email protected]>
---
mm/hugetlb_vmemmap.c | 92 ++++++++++++++++++++++++++++++++++++++++++--
1 file changed, 88 insertions(+), 4 deletions(-)
diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
index 4ac521e596db..10739e4285d5 100644
--- a/mm/hugetlb_vmemmap.c
+++ b/mm/hugetlb_vmemmap.c
@@ -27,6 +27,8 @@
* @reuse_addr: the virtual address of the @reuse_page page.
* @vmemmap_pages: the list head of the vmemmap pages that can be freed
* or is mapped from.
+ * @flags: used to modify behavior in vmemmap page table walking
+ * operations.
*/
struct vmemmap_remap_walk {
void (*remap_pte)(pte_t *pte, unsigned long addr,
@@ -35,9 +37,13 @@ struct vmemmap_remap_walk {
struct page *reuse_page;
unsigned long reuse_addr;
struct list_head *vmemmap_pages;
+
+/* Skip the TLB flush when we split the PMD */
+#define VMEMMAP_SPLIT_NO_TLB_FLUSH BIT(0)
+ unsigned long flags;
};
-static int split_vmemmap_huge_pmd(pmd_t *pmd, unsigned long start)
+static int split_vmemmap_huge_pmd(pmd_t *pmd, unsigned long start, bool flush)
{
pmd_t __pmd;
int i;
@@ -80,7 +86,8 @@ static int split_vmemmap_huge_pmd(pmd_t *pmd, unsigned long start)
/* Make pte visible before pmd. See comment in pmd_install(). */
smp_wmb();
pmd_populate_kernel(&init_mm, pmd, pgtable);
- flush_tlb_kernel_range(start, start + PMD_SIZE);
+ if (flush)
+ flush_tlb_kernel_range(start, start + PMD_SIZE);
} else {
pte_free_kernel(&init_mm, pgtable);
}
@@ -127,11 +134,20 @@ static int vmemmap_pmd_range(pud_t *pud, unsigned long addr,
do {
int ret;
- ret = split_vmemmap_huge_pmd(pmd, addr & PMD_MASK);
+ ret = split_vmemmap_huge_pmd(pmd, addr & PMD_MASK,
+ !(walk->flags & VMEMMAP_SPLIT_NO_TLB_FLUSH));
if (ret)
return ret;
next = pmd_addr_end(addr, end);
+
+ /*
+ * We are only splitting, not remapping the hugetlb vmemmap
+ * pages.
+ */
+ if (!walk->remap_pte)
+ continue;
+
vmemmap_pte_range(pmd, addr, next, walk);
} while (pmd++, addr = next, addr != end);
@@ -198,7 +214,8 @@ static int vmemmap_remap_range(unsigned long start, unsigned long end,
return ret;
} while (pgd++, addr = next, addr != end);
- flush_tlb_kernel_range(start, end);
+ if (walk->remap_pte)
+ flush_tlb_kernel_range(start, end);
return 0;
}
@@ -297,6 +314,36 @@ static void vmemmap_restore_pte(pte_t *pte, unsigned long addr,
set_pte_at(&init_mm, addr, pte, mk_pte(page, pgprot));
}
+/**
+ * vmemmap_remap_split - split the vmemmap virtual address range [@start, @end)
+ * backing PMDs of the directmap into PTEs
+ * @start: start address of the vmemmap virtual address range that we want
+ * to remap.
+ * @end: end address of the vmemmap virtual address range that we want to
+ * remap.
+ * @reuse: reuse address.
+ *
+ * Return: %0 on success, negative error code otherwise.
+ */
+static int vmemmap_remap_split(unsigned long start, unsigned long end,
+ unsigned long reuse)
+{
+ int ret;
+ struct vmemmap_remap_walk walk = {
+ .remap_pte = NULL,
+ .flags = VMEMMAP_SPLIT_NO_TLB_FLUSH,
+ };
+
+ /* See the comment in the vmemmap_remap_free(). */
+ BUG_ON(start - reuse != PAGE_SIZE);
+
+ mmap_read_lock(&init_mm);
+ ret = vmemmap_remap_range(reuse, end, &walk);
+ mmap_read_unlock(&init_mm);
+
+ return ret;
+}
+
/**
* vmemmap_remap_free - remap the vmemmap virtual address range [@start, @end)
* to the page which @reuse is mapped to, then free vmemmap
@@ -320,6 +367,7 @@ static int vmemmap_remap_free(unsigned long start, unsigned long end,
.remap_pte = vmemmap_remap_pte,
.reuse_addr = reuse,
.vmemmap_pages = vmemmap_pages,
+ .flags = 0,
};
int nid = page_to_nid((struct page *)reuse);
gfp_t gfp_mask = GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN;
@@ -368,6 +416,7 @@ static int vmemmap_remap_free(unsigned long start, unsigned long end,
.remap_pte = vmemmap_restore_pte,
.reuse_addr = reuse,
.vmemmap_pages = vmemmap_pages,
+ .flags = 0,
};
vmemmap_remap_range(reuse, end, &walk);
@@ -419,6 +468,7 @@ static int vmemmap_remap_alloc(unsigned long start, unsigned long end,
.remap_pte = vmemmap_restore_pte,
.reuse_addr = reuse,
.vmemmap_pages = &vmemmap_pages,
+ .flags = 0,
};
/* See the comment in the vmemmap_remap_free(). */
@@ -628,11 +678,45 @@ void hugetlb_vmemmap_optimize(const struct hstate *h, struct page *head)
free_vmemmap_page_list(&vmemmap_pages);
}
+static int hugetlb_vmemmap_split(const struct hstate *h, struct page *head)
+{
+ unsigned long vmemmap_start = (unsigned long)head, vmemmap_end;
+ unsigned long vmemmap_reuse;
+
+ if (!vmemmap_should_optimize(h, head))
+ return 0;
+
+ vmemmap_end = vmemmap_start + hugetlb_vmemmap_size(h);
+ vmemmap_reuse = vmemmap_start;
+ vmemmap_start += HUGETLB_VMEMMAP_RESERVE_SIZE;
+
+ /*
+ * Split PMDs on the vmemmap virtual address range [@vmemmap_start,
+ * @vmemmap_end]
+ */
+ return vmemmap_remap_split(vmemmap_start, vmemmap_end, vmemmap_reuse);
+}
+
void hugetlb_vmemmap_optimize_folios(struct hstate *h, struct list_head *folio_list)
{
struct folio *folio;
LIST_HEAD(vmemmap_pages);
+ list_for_each_entry(folio, folio_list, lru) {
+ int ret = hugetlb_vmemmap_split(h, &folio->page);
+
+ /*
+ * Spliting the PMD requires allocating a page, thus lets fail
+ * early once we encounter the first OOM. No point in retrying
+ * as it can be dynamically done on remap with the memory
+ * we get back from the vmemmap deduplication.
+ */
+ if (ret == -ENOMEM)
+ break;
+ }
+
+ flush_tlb_all();
+
list_for_each_entry(folio, folio_list, lru) {
int ret = __hugetlb_vmemmap_optimize(h, &folio->page,
&vmemmap_pages);
--
2.41.0