Re: [PATCH v4 12/12] mm/swap: reduce swap cache search space

[Kairui Song]

On Wed, May 8, 2024 at 3:27 PM Huang, Ying <ying.huang@xxxxxxxxx> wrote:
>
> Kairui Song <ryncsn@xxxxxxxxx> writes:
>
> > From: Kairui Song <kasong@xxxxxxxxxxx>
> >
> > Currently we use one swap_address_space for every 64M chunk to reduce lock
> > contention, this is like having a set of smaller swap files inside one
> > swap device. But when doing swap cache look up or insert, we are
> > still using the offset of the whole large swap device. This is OK for
> > correctness, as the offset (key) is unique.
> >
> > But Xarray is specially optimized for small indexes, it creates the
> > radix tree levels lazily to be just enough to fit the largest key
> > stored in one Xarray. So we are wasting tree nodes unnecessarily.
> >
> > For 64M chunk it should only take at most 3 levels to contain everything. it should only take at most 3 levels to contain everything.
> > But if we are using the offset from the whole swap device, the offset (key)
> > value will be way beyond 64M, and so will the tree level.
> >
> > Optimize this by using a new helper swap_cache_index to get a swap
> > entry's unique offset in its own 64M swap_address_space.
> >
> > I see a ~1% performance gain in benchmark and actual workload with
> > high memory pressure.
> >
> > Test with `time memhog 128G` inside a 8G memcg using 128G swap (ramdisk
> > with SWP_SYNCHRONOUS_IO dropped, tested 3 times, results are stable. The
> > test result is similar but the improvement is smaller if SWP_SYNCHRONOUS_IO
> > is enabled, as swap out path can never skip swap cache):
> >
> > Before:
> > 6.07user 250.74system 4:17.26elapsed 99%CPU (0avgtext+0avgdata 8373376maxresident)k
> > 0inputs+0outputs (55major+33555018minor)pagefaults 0swaps
> >
> > After (1.8% faster):
> > 6.08user 246.09system 4:12.58elapsed 99%CPU (0avgtext+0avgdata 8373248maxresident)k
> > 0inputs+0outputs (54major+33555027minor)pagefaults 0swaps
> >
> > Similar result with MySQL and sysbench using swap:
> > Before:
> > 94055.61 qps
> >
> > After (0.8% faster):
> > 94834.91 qps
> >
> > Radix tree slab usage is also very slightly lower.
> >
> > Signed-off-by: Kairui Song <kasong@xxxxxxxxxxx>
> > ---
> >  mm/huge_memory.c |  2 +-
> >  mm/memcontrol.c  |  2 +-
> >  mm/mincore.c     |  2 +-
> >  mm/shmem.c       |  2 +-
> >  mm/swap.h        | 15 +++++++++++++++
> >  mm/swap_state.c  | 19 ++++++++++---------
> >  mm/swapfile.c    |  6 +++---
> >  7 files changed, 32 insertions(+), 16 deletions(-)
> >
> > diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> > index d35d526ed48f..45829cc049d2 100644
> > --- a/mm/huge_memory.c
> > +++ b/mm/huge_memory.c
> > @@ -2918,7 +2918,7 @@ static void __split_huge_page(struct page *page, struct list_head *list,
> >       split_page_memcg(head, order, new_order);
> >
> >       if (folio_test_anon(folio) && folio_test_swapcache(folio)) {
> > -             offset = swp_offset(folio->swap);
> > +             offset = swap_cache_index(folio->swap);
> >               swap_cache = swap_address_space(folio->swap);
> >               xa_lock(&swap_cache->i_pages);
> >       }
> > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > index d11536ef59ef..81b005c459cb 100644
> > --- a/mm/memcontrol.c
> > +++ b/mm/memcontrol.c
> > @@ -6165,7 +6165,7 @@ static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
> >        * Because swap_cache_get_folio() updates some statistics counter,
> >        * we call find_get_page() with swapper_space directly.
> >        */
> > -     page = find_get_page(swap_address_space(ent), swp_offset(ent));
> > +     page = find_get_page(swap_address_space(ent), swap_cache_index(ent));
> >       entry->val = ent.val;
> >
> >       return page;
> > diff --git a/mm/mincore.c b/mm/mincore.c
> > index dad3622cc963..e31cf1bde614 100644
> > --- a/mm/mincore.c
> > +++ b/mm/mincore.c
> > @@ -139,7 +139,7 @@ static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
> >                       } else {
> >  #ifdef CONFIG_SWAP
> >                               *vec = mincore_page(swap_address_space(entry),
> > -                                                 swp_offset(entry));
> > +                                                 swap_cache_index(entry));
> >  #else
> >                               WARN_ON(1);
> >                               *vec = 1;
> > diff --git a/mm/shmem.c b/mm/shmem.c
> > index fa2a0ed97507..326315c12feb 100644
> > --- a/mm/shmem.c
> > +++ b/mm/shmem.c
> > @@ -1756,7 +1756,7 @@ static int shmem_replace_folio(struct folio **foliop, gfp_t gfp,
> >
> >       old = *foliop;
> >       entry = old->swap;
> > -     swap_index = swp_offset(entry);
> > +     swap_index = swap_cache_index(entry);
> >       swap_mapping = swap_address_space(entry);
> >
> >       /*
> > diff --git a/mm/swap.h b/mm/swap.h
> > index 82023ab93205..93e3e1b58a7f 100644
> > --- a/mm/swap.h
> > +++ b/mm/swap.h
> > @@ -27,6 +27,7 @@ void __swap_writepage(struct folio *folio, struct writeback_control *wbc);
> >  /* One swap address space for each 64M swap space */
> >  #define SWAP_ADDRESS_SPACE_SHIFT     14
> >  #define SWAP_ADDRESS_SPACE_PAGES     (1 << SWAP_ADDRESS_SPACE_SHIFT)
> > +#define SWAP_ADDRESS_SPACE_MASK              (BIT(SWAP_ADDRESS_SPACE_SHIFT) - 1)
>
> #define SWAP_ADDRESS_SPACE_MASK         (SWAP_ADDRESS_SPACE_PAGES - 1)
> ?
>
> We can use BIT() in SWAP_ADDRESS_SPACE_PAGES definition.
>

I'll just use (SWAP_ADDRESS_SPACE_PAGES - 1) then, I was trying to
make the changes minimal, but prefered the BIT macro, a trivial
change.

> >  extern struct address_space *swapper_spaces[];
> >  #define swap_address_space(entry)                        \
> >       (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
> > @@ -40,6 +41,15 @@ static inline loff_t swap_dev_pos(swp_entry_t entry)
> >       return ((loff_t)swp_offset(entry)) << PAGE_SHIFT;
> >  }
> >
> > +/*
> > + * Return the swap cache index of the swap entry.
> > + */
> > +static inline pgoff_t swap_cache_index(swp_entry_t entry)
> > +{
> > +     BUILD_BUG_ON((SWP_OFFSET_MASK | SWAP_ADDRESS_SPACE_MASK) != SWP_OFFSET_MASK);
> > +     return swp_offset(entry) & SWAP_ADDRESS_SPACE_MASK;
> > +}
> > +
> >  void show_swap_cache_info(void);
> >  bool add_to_swap(struct folio *folio);
> >  void *get_shadow_from_swap_cache(swp_entry_t entry);
> > @@ -86,6 +96,11 @@ static inline struct address_space *swap_address_space(swp_entry_t entry)
> >       return NULL;
> >  }
> >
> > +static inline pgoff_t swap_cache_index(swp_entry_t entry)
> > +{
> > +     return 0;
> > +}
> > +
> >  static inline void show_swap_cache_info(void)
> >  {
> >  }
> > diff --git a/mm/swap_state.c b/mm/swap_state.c
> > index 642c30d8376c..09415d4c7843 100644
> > --- a/mm/swap_state.c
> > +++ b/mm/swap_state.c
> > @@ -72,7 +72,7 @@ void show_swap_cache_info(void)
> >  void *get_shadow_from_swap_cache(swp_entry_t entry)
> >  {
> >       struct address_space *address_space = swap_address_space(entry);
> > -     pgoff_t idx = swp_offset(entry);
> > +     pgoff_t idx = swap_cache_index(entry);
> >       void *shadow;
> >
> >       shadow = xa_load(&address_space->i_pages, idx);
> > @@ -89,7 +89,7 @@ int add_to_swap_cache(struct folio *folio, swp_entry_t entry,
> >                       gfp_t gfp, void **shadowp)
> >  {
> >       struct address_space *address_space = swap_address_space(entry);
> > -     pgoff_t idx = swp_offset(entry);
> > +     pgoff_t idx = swap_cache_index(entry);
> >       XA_STATE_ORDER(xas, &address_space->i_pages, idx, folio_order(folio));
> >       unsigned long i, nr = folio_nr_pages(folio);
> >       void *old;
> > @@ -144,7 +144,7 @@ void __delete_from_swap_cache(struct folio *folio,
> >       struct address_space *address_space = swap_address_space(entry);
> >       int i;
> >       long nr = folio_nr_pages(folio);
> > -     pgoff_t idx = swp_offset(entry);
> > +     pgoff_t idx = swap_cache_index(entry);
> >       XA_STATE(xas, &address_space->i_pages, idx);
> >
> >       xas_set_update(&xas, workingset_update_node);
> > @@ -248,18 +248,19 @@ void delete_from_swap_cache(struct folio *folio)
> >  void clear_shadow_from_swap_cache(int type, unsigned long begin,
> >                               unsigned long end)
> >  {
> > -     unsigned long curr = begin;
> > +     unsigned long curr = begin, offset;
>
> Better to rename "offset" as "index" to avoid confusion?

Good idea.

> >       void *old;
> >
> >       for (;;) {
> > +             offset = curr & SWAP_ADDRESS_SPACE_MASK;
> >               swp_entry_t entry = swp_entry(type, curr);
> >               struct address_space *address_space = swap_address_space(entry);
> > -             XA_STATE(xas, &address_space->i_pages, curr);
> > +             XA_STATE(xas, &address_space->i_pages, offset);
> >
> >               xas_set_update(&xas, workingset_update_node);
> >
> >               xa_lock_irq(&address_space->i_pages);
> > -             xas_for_each(&xas, old, end) {
> > +             xas_for_each(&xas, old, offset + min(end - curr, SWAP_ADDRESS_SPACE_PAGES)) {
>
> Is there a bug in the original code?  It doesn't check SWAP_ADDRESS_SPACE_PAGES.

That's OK, if the (end - curr) goes above SWAP_ADDRESS_SPACE_PAGES, it
means all content in current address_space needs to be purged.
xas_for_each will stop after it iterated all content in the current
address space. This is a bit hackish though.

>
> And should it be changed to
>
>         xas_for_each(&xas, old, min(offset + end - curr, SWAP_ADDRESS_SPACE_PAGES))

It should be equivalent, as described above, but yeah, this looks
cleaner. I'll use your suggested code.

> ?
>
> >                       if (!xa_is_value(old))
> >                               continue;
> >                       xas_store(&xas, NULL);
> > @@ -350,7 +351,7 @@ struct folio *swap_cache_get_folio(swp_entry_t entry,
> >  {
> >       struct folio *folio;
> >
> > -     folio = filemap_get_folio(swap_address_space(entry), swp_offset(entry));
> > +     folio = filemap_get_folio(swap_address_space(entry), swap_cache_index(entry));
> >       if (!IS_ERR(folio)) {
> >               bool vma_ra = swap_use_vma_readahead();
> >               bool readahead;
> > @@ -420,7 +421,7 @@ struct folio *filemap_get_incore_folio(struct address_space *mapping,
> >       si = get_swap_device(swp);
> >       if (!si)
> >               return ERR_PTR(-ENOENT);
> > -     index = swp_offset(swp);
> > +     index = swap_cache_index(swp);
> >       folio = filemap_get_folio(swap_address_space(swp), index);
> >       put_swap_device(si);
> >       return folio;
> > @@ -447,7 +448,7 @@ struct folio *__read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
> >                * that would confuse statistics.
> >                */
> >               folio = filemap_get_folio(swap_address_space(entry),
> > -                                             swp_offset(entry));
> > +                                       swap_cache_index(entry));
> >               if (!IS_ERR(folio))
> >                       goto got_folio;
> >
> > diff --git a/mm/swapfile.c b/mm/swapfile.c
> > index 0b0ae6e8c764..4f0e8b2ac8aa 100644
> > --- a/mm/swapfile.c
> > +++ b/mm/swapfile.c
> > @@ -142,7 +142,7 @@ static int __try_to_reclaim_swap(struct swap_info_struct *si,
> >       struct folio *folio;
> >       int ret = 0;
> >
> > -     folio = filemap_get_folio(swap_address_space(entry), offset);
> > +     folio = filemap_get_folio(swap_address_space(entry), swap_cache_index(entry));
> >       if (IS_ERR(folio))
> >               return 0;
> >       /*
> > @@ -2158,7 +2158,7 @@ static int try_to_unuse(unsigned int type)
> >              (i = find_next_to_unuse(si, i)) != 0) {
> >
> >               entry = swp_entry(type, i);
> > -             folio = filemap_get_folio(swap_address_space(entry), i);
> > +             folio = filemap_get_folio(swap_address_space(entry), swap_cache_index(entry));
> >               if (IS_ERR(folio))
> >                       continue;
> >
> > @@ -3476,7 +3476,7 @@ EXPORT_SYMBOL_GPL(swapcache_mapping);
> >
> >  pgoff_t __folio_swap_cache_index(struct folio *folio)
> >  {
> > -     return swp_offset(folio->swap);
> > +     return swap_cache_index(folio->swap);
> >  }
> >  EXPORT_SYMBOL_GPL(__folio_swap_cache_index);

Thanks for the suggestions!

>
> --
> Best Regards,
> Huang, Ying