Re: [PATCH v4 6/8] hugetlb: batch PMD split for bulk vmemmap dedup

From: Joao Martins
Date: Tue Sep 19 2023 - 11:10:51 EST


On 19/09/2023 09:57, Muchun Song wrote:
>> On Sep 19, 2023, at 16:55, Joao Martins <joao.m.martins@xxxxxxxxxx> wrote:
>> On 19/09/2023 09:41, Muchun Song wrote:
>>>> On Sep 19, 2023, at 16:26, Joao Martins <joao.m.martins@xxxxxxxxxx> wrote:
>>>> On 19/09/2023 07:42, Muchun Song wrote:
>>>>> On 2023/9/19 07:01, Mike Kravetz wrote:
>>>>>> list_for_each_entry(folio, folio_list, lru) {
>>>>>> int ret = __hugetlb_vmemmap_optimize(h, &folio->page,
>>>>>> &vmemmap_pages);
>>>>>
>>>>> This is unlikely to be failed since the page table allocation
>>>>> is moved to the above
>>>>
>>>>> (Note that the head vmemmap page allocation
>>>>> is not mandatory).
>>>>
>>>> Good point that I almost forgot
>>>>
>>>>> So we should handle the error case in the above
>>>>> splitting operation.
>>>>
>>>> But back to the previous discussion in v2... the thinking was that /some/ PMDs
>>>> got split, and say could allow some PTE remapping to occur and free some pages
>>>> back (each page allows 6 more splits worst case). Then the next
>>>> __hugetlb_vmemmap_optimize() will have to split PMD pages again for those
>>>> hugepages that failed the batch PMD split (as we only defer the PTE remap tlb
>>>> flush in this stage).
>>>
>>> Oh, yes. Maybe we could break the above traversal as early as possible
>>> once we enter an ENOMEM?
>>>
>>
>> Sounds good -- no point in keep trying to split if we are failing with OOM.
>>
>> Perhaps a comment in both of these clauses (the early break on split and the OOM
>> handling in batch optimize) could help make this clear.
>
> Make sense.

These are the changes I have so far for this patch based on the discussion so
far. For next one it's at the end:

diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
index e8bc2f7567db..d9c6f2cf698c 100644
--- a/mm/hugetlb_vmemmap.c
+++ b/mm/hugetlb_vmemmap.c
@@ -27,7 +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 bulk operations
+ * @flags: used to modify behavior in vmemmap page table walking
+ * operations.
*/
struct vmemmap_remap_walk {
void (*remap_pte)(pte_t *pte, unsigned long addr,
@@ -36,6 +37,8 @@ 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;
};
@@ -132,7 +135,7 @@ static int vmemmap_pmd_range(pud_t *pud, unsigned long addr,
int ret;

ret = split_vmemmap_huge_pmd(pmd, addr & PMD_MASK,
- walk->flags & VMEMMAP_SPLIT_NO_TLB_FLUSH);
+ !(walk->flags & VMEMMAP_SPLIT_NO_TLB_FLUSH));
if (ret)
return ret;

@@ -677,13 +680,13 @@ void hugetlb_vmemmap_optimize(const struct hstate *h,
struct page *head)
free_vmemmap_page_list(&vmemmap_pages);
}

-static void hugetlb_vmemmap_split(const struct hstate *h, struct page *head)
+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;
+ return 0;

vmemmap_end = vmemmap_start + hugetlb_vmemmap_size(h);
vmemmap_reuse = vmemmap_start;
@@ -693,7 +696,7 @@ static void hugetlb_vmemmap_split(const struct hstate *h,
struct page *head)
* Split PMDs on the vmemmap virtual address range [@vmemmap_start,
* @vmemmap_end]
*/
- vmemmap_remap_split(vmemmap_start, vmemmap_end, vmemmap_reuse);
+ return vmemmap_remap_split(vmemmap_start, vmemmap_end, vmemmap_reuse);
}

void hugetlb_vmemmap_optimize_folios(struct hstate *h, struct list_head
*folio_list)
@@ -701,8 +704,18 @@ void hugetlb_vmemmap_optimize_folios(struct hstate *h,
struct list_head *folio_l
struct folio *folio;
LIST_HEAD(vmemmap_pages);

- list_for_each_entry(folio, folio_list, lru)
- hugetlb_vmemmap_split(h, &folio->page);
+ 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();

For patch 7, I only have commentary added derived from this earlier discussion
above:

diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
index d9c6f2cf698c..f6a1020a4b6a 100644
--- a/mm/hugetlb_vmemmap.c
+++ b/mm/hugetlb_vmemmap.c
@@ -40,6 +40,8 @@ struct vmemmap_remap_walk {

/* Skip the TLB flush when we split the PMD */
#define VMEMMAP_SPLIT_NO_TLB_FLUSH BIT(0)
+/* Skip the TLB flush when we remap the PTE */
#define VMEMMAP_REMAP_NO_TLB_FLUSH BIT(1)
unsigned long flags;
};

@@ -721,19 +739,28 @@ void hugetlb_vmemmap_optimize_folios(struct hstate *h,
struct list_head *folio_l

list_for_each_entry(folio, folio_list, lru) {
int ret = __hugetlb_vmemmap_optimize(h, &folio->page,
&vmemmap_pages,
VMEMMAP_REMAP_NO_TLB_FLUSH);

/*
* Pages to be freed may have been accumulated. If we
* encounter an ENOMEM, free what we have and try again.
+ * This can occur in the case that both spliting fails
+ * halfway and head page allocation also failed. In this
+ * case __hugetlb_vmemmap_optimize() would free memory
+ * allowing more vmemmap remaps to occur.
*/
if (ret == -ENOMEM && !list_empty(&vmemmap_pages)) {