Re: [PATCH v2 3/9] mm/hugetlb_cgroup: convert set_hugetlb_cgroup*() to folios

[Muchun Song <muchun.song@linux.dev>]

> On Nov 10, 2022, at 08:20, Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote:
> 
> 
> On 11/1/22 11:45 PM, Muchun Song wrote:
>> 
>>> On Nov 2, 2022, at 06:30, Sidhartha Kumar <sidhartha.kumar@oracle.com> wrote:
>>> 
>>> Allows __prep_new_huge_page() to operate on a folio by converting
>>> set_hugetlb_cgroup*() to take in a folio.
>>> 
>>> Signed-off-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>
>>> ---
>>> include/linux/hugetlb_cgroup.h | 12 ++++++------
>>> mm/hugetlb.c                   | 33 +++++++++++++++++++--------------
>>> mm/hugetlb_cgroup.c            | 11 ++++++-----
>>> 3 files changed, 31 insertions(+), 25 deletions(-)
>>> 
>>> diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h
>>> index feb2edafc8b6..a7e3540f7f38 100644
>>> --- a/include/linux/hugetlb_cgroup.h
>>> +++ b/include/linux/hugetlb_cgroup.h
>>> @@ -112,16 +112,16 @@ static inline void __set_hugetlb_cgroup(struct folio *folio,
>>> (unsigned long)h_cg);
>>> }
>>> 
>>> -static inline void set_hugetlb_cgroup(struct page *page,
>>> +static inline void set_hugetlb_cgroup(struct folio *folio,
>>>     struct hugetlb_cgroup *h_cg)
>>> {
>>> - __set_hugetlb_cgroup(page_folio(page), h_cg, false);
>>> + __set_hugetlb_cgroup(folio, h_cg, false);
>>> }
>>> 
>>> -static inline void set_hugetlb_cgroup_rsvd(struct page *page,
>>> +static inline void set_hugetlb_cgroup_rsvd(struct folio *folio,
>>>  struct hugetlb_cgroup *h_cg)
>>> {
>>> - __set_hugetlb_cgroup(page_folio(page), h_cg, true);
>>> + __set_hugetlb_cgroup(folio, h_cg, true);
>>> }
>>> 
>>> static inline bool hugetlb_cgroup_disabled(void)
>>> @@ -199,12 +199,12 @@ hugetlb_cgroup_from_folio_rsvd(struct folio *folio)
>>> return NULL;
>>> }
>>> 
>>> -static inline void set_hugetlb_cgroup(struct page *page,
>>> +static inline void set_hugetlb_cgroup(struct folio *folio,
>>>     struct hugetlb_cgroup *h_cg)
>>> {
>>> }
>>> 
>>> -static inline void set_hugetlb_cgroup_rsvd(struct page *page,
>>> +static inline void set_hugetlb_cgroup_rsvd(struct folio *folio,
>>>  struct hugetlb_cgroup *h_cg)
>>> {
>>> }
>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>> index 27b87dc85c48..a6384fb0b57f 100644
>>> --- a/mm/hugetlb.c
>>> +++ b/mm/hugetlb.c
>>> @@ -1758,19 +1758,21 @@ static void __prep_account_new_huge_page(struct hstate *h, int nid)
>>> h->nr_huge_pages_node[nid]++;
>>> }
>>> 
>>> -static void __prep_new_huge_page(struct hstate *h, struct page *page)
>>> +static void __prep_new_hugetlb_folio(struct hstate *h, struct folio *folio)
>>> {
>>> - hugetlb_vmemmap_optimize(h, page);
>>> - INIT_LIST_HEAD(&page->lru);
>>> - set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
>>> - hugetlb_set_page_subpool(page, NULL);
>>> - set_hugetlb_cgroup(page, NULL);
>>> - set_hugetlb_cgroup_rsvd(page, NULL);
>>> + hugetlb_vmemmap_optimize(h, &folio->page);
>>> + INIT_LIST_HEAD(&folio->lru);
>>> + folio->_folio_dtor = HUGETLB_PAGE_DTOR;
>> Seems like a variant of set_compound_page_dtor() for folio is missing,
>> e.g. set_large_folio_dtor(). It's better to add it in this series.
>> 
>> Thanks.
> 
> Hi Muchun thanks for taking a look,
> 
> Would it be ok to add this functionality in a separate patch series? Some of the earlier patches in this series were 

I'm ok.

Thanks.

> modified by Hugh's series[1] so I'm not sure how a v3 of this series would look now. Let me know which approach you would prefer.
> 
> 
> Thanks,
> 
> Sidhartha Kumar
> 
> [1] https://lore.kernel.org/linux-mm/3818cc9a-9999-d064-d778-9c94c5911e6@google.com/
> 
>>> + hugetlb_set_folio_subpool(folio, NULL);
>>> + set_hugetlb_cgroup(folio, NULL);
>>> + set_hugetlb_cgroup_rsvd(folio, NULL);
>>> }
>>> 
>>> static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
>>> {
>>> - __prep_new_huge_page(h, page);
>>> + struct folio *folio = page_folio(page);
>>> +
>>> + __prep_new_hugetlb_folio(h, folio);
>>> spin_lock_irq(&hugetlb_lock);
>>> __prep_account_new_huge_page(h, nid);
>>> spin_unlock_irq(&hugetlb_lock);
>>> @@ -2731,8 +2733,10 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page,
>>> struct list_head *list)
>>> {
>>> gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
>>> - int nid = page_to_nid(old_page);
>>> + struct folio *old_folio = page_folio(old_page);
>>> + int nid = folio_nid(old_folio);
>>> struct page *new_page;
>>> + struct folio *new_folio;
>>> int ret = 0;
>>> 
>>> /*
>>> @@ -2745,16 +2749,17 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page,
>>> new_page = alloc_buddy_huge_page(h, gfp_mask, nid, NULL, NULL);
>>> if (!new_page)
>>> return -ENOMEM;
>>> - __prep_new_huge_page(h, new_page);
>>> + new_folio = page_folio(new_page);
>>> + __prep_new_hugetlb_folio(h, new_folio);
>>> 
>>> retry:
>>> spin_lock_irq(&hugetlb_lock);
>>> - if (!PageHuge(old_page)) {
>>> + if (!folio_test_hugetlb(old_folio)) {
>>> /*
>>> * Freed from under us. Drop new_page too.
>>> */
>>> goto free_new;
>>> - } else if (page_count(old_page)) {
>>> + } else if (folio_ref_count(old_folio)) {
>>> /*
>>> * Someone has grabbed the page, try to isolate it here.
>>> * Fail with -EBUSY if not possible.
>>> @@ -2763,7 +2768,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page,
>>> ret = isolate_hugetlb(old_page, list);
>>> spin_lock_irq(&hugetlb_lock);
>>> goto free_new;
>>> - } else if (!HPageFreed(old_page)) {
>>> + } else if (!folio_test_hugetlb_freed(old_folio)) {
>>> /*
>>> * Page's refcount is 0 but it has not been enqueued in the
>>> * freelist yet. Race window is small, so we can succeed here if
>>> @@ -2801,7 +2806,7 @@ static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page,
>>> free_new:
>>> spin_unlock_irq(&hugetlb_lock);
>>> /* Page has a zero ref count, but needs a ref to be freed */
>>> - set_page_refcounted(new_page);
>>> + folio_ref_unfreeze(new_folio, 1);
>>> update_and_free_page(h, new_page, false);
>>> 
>>> return ret;
>>> diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
>>> index 600c98560a0f..692b23b5d423 100644
>>> --- a/mm/hugetlb_cgroup.c
>>> +++ b/mm/hugetlb_cgroup.c
>>> @@ -212,7 +212,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
>>> /* Take the pages off the local counter */
>>> page_counter_cancel(counter, nr_pages);
>>> 
>>> - set_hugetlb_cgroup(page, parent);
>>> + set_hugetlb_cgroup(folio, parent);
>>> out:
>>> return;
>>> }
>>> @@ -894,6 +894,7 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
>>> struct hugetlb_cgroup *h_cg_rsvd;
>>> struct hstate *h = page_hstate(oldhpage);
>>> struct folio *old_folio = page_folio(oldhpage);
>>> + struct folio *new_folio = page_folio(newhpage);
>>> 
>>> if (hugetlb_cgroup_disabled())
>>> return;
>>> @@ -901,12 +902,12 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
>>> spin_lock_irq(&hugetlb_lock);
>>> h_cg = hugetlb_cgroup_from_folio(old_folio);
>>> h_cg_rsvd = hugetlb_cgroup_from_folio_rsvd(old_folio);
>>> - set_hugetlb_cgroup(oldhpage, NULL);
>>> - set_hugetlb_cgroup_rsvd(oldhpage, NULL);
>>> + set_hugetlb_cgroup(old_folio, NULL);
>>> + set_hugetlb_cgroup_rsvd(old_folio, NULL);
>>> 
>>> /* move the h_cg details to new cgroup */
>>> - set_hugetlb_cgroup(newhpage, h_cg);
>>> - set_hugetlb_cgroup_rsvd(newhpage, h_cg_rsvd);
>>> + set_hugetlb_cgroup(new_folio, h_cg);
>>> + set_hugetlb_cgroup_rsvd(new_folio, h_cg_rsvd);
>>> list_move(&newhpage->lru, &h->hugepage_activelist);
>>> spin_unlock_irq(&hugetlb_lock);
>>> return;
>>> -- 
>>> 2.31.1
>>> 
>>> 
>>