RE: [PATCH v9 6/7] mm: zswap: Support large folios in zswap_store().

[Sridhar, Kanchana P]

> -----Original Message-----
> From: Yosry Ahmed <yosryahmed@xxxxxxxxxx>
> Sent: Monday, September 30, 2024 11:00 PM
> To: Sridhar, Kanchana P <kanchana.p.sridhar@xxxxxxxxx>
> Cc: linux-kernel@xxxxxxxxxxxxxxx; linux-mm@xxxxxxxxx;
> hannes@xxxxxxxxxxx; nphamcs@xxxxxxxxx; chengming.zhou@xxxxxxxxx;
> usamaarif642@xxxxxxxxx; shakeel.butt@xxxxxxxxx; ryan.roberts@xxxxxxx;
> Huang, Ying <ying.huang@xxxxxxxxx>; 21cnbao@xxxxxxxxx; akpm@linux-
> foundation.org; willy@xxxxxxxxxxxxx; Zou, Nanhai <nanhai.zou@xxxxxxxxx>;
> Feghali, Wajdi K <wajdi.k.feghali@xxxxxxxxx>; Gopal, Vinodh
> <vinodh.gopal@xxxxxxxxx>
> Subject: Re: [PATCH v9 6/7] mm: zswap: Support large folios in zswap_store().
> 
> [..]
> > > >  store_failed:
> > > >         zpool_free(entry->pool->zpool, entry->handle);
> > > > -put_pool:
> > > > -       zswap_pool_put(entry->pool);
> > > > -freepage:
> > > > +put_pool_objcg:
> > > > +       zswap_pool_put(pool);
> > > > +       obj_cgroup_put(objcg);
> > >
> > > I think if we reorder the function we can drop these calls, make the
> > > comments positioned a bit better, and centralize the entry
> > > initializations. I am also not a fan of passing a semi-initialized
> > > entry to zswap_compress() to get the pool pointer.
> > >
> > > Does the following diff improve things or did I miss something?
> >
> > We shouldn’t be adding the entry to the xarray before initializing its pool
> > and objcg, right? Please let me know if I am misunderstanding what you're
> > proposing in the diff.
> 
> It should be safe. We already initialize entry->lru after we insert
> the entry in the tree. See the comment above the call to
> zswap_lru_add(). Basically we are protected against concurrent
> stores/loads through the folio lock, and are protected against
> writeback because the entry is not on the LRU yet.

Thanks for the clarification, Yosry. Since this is a change in the entry
initialization wrt the mainline, is it Ok if this is done in a follow-up patch?

Thanks,
Kanchana

> 
> >
> > >
> > > diff --git a/mm/zswap.c b/mm/zswap.c
> > > index b74c8de996468..eac1f846886a6 100644
> > > --- a/mm/zswap.c
> > > +++ b/mm/zswap.c
> > > @@ -881,7 +881,8 @@ static int zswap_cpu_comp_dead(unsigned int
> cpu,
> > > struct hlist_node *node)
> > >         return 0;
> > >  }
> > >
> > > -static bool zswap_compress(struct page *page, struct zswap_entry
> *entry)
> > > +static bool zswap_compress(struct page *page, struct zswap_entry
> *entry,
> > > +                          struct zswap_pool *pool)
> > >  {
> > >         struct crypto_acomp_ctx *acomp_ctx;
> > >         struct scatterlist input, output;
> > > @@ -893,7 +894,7 @@ static bool zswap_compress(struct page *page,
> > > struct zswap_entry *entry)
> > >         gfp_t gfp;
> > >         u8 *dst;
> > >
> > > -       acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
> > > +       acomp_ctx = raw_cpu_ptr(pool->acomp_ctx);
> > >
> > >         mutex_lock(&acomp_ctx->mutex);
> > >
> > > @@ -926,7 +927,7 @@ static bool zswap_compress(struct page *page,
> > > struct zswap_entry *entry)
> > >         if (comp_ret)
> > >                 goto unlock;
> > >
> > > -       zpool = entry->pool->zpool;
> > > +       zpool = pool->zpool;
> > >         gfp = __GFP_NORETRY | __GFP_NOWARN |
> __GFP_KSWAPD_RECLAIM;
> > >         if (zpool_malloc_support_movable(zpool))
> > >                 gfp |= __GFP_HIGHMEM | __GFP_MOVABLE;
> > > @@ -1435,23 +1436,11 @@ static bool zswap_store_page(struct page
> > > *page,
> > >         entry = zswap_entry_cache_alloc(GFP_KERNEL,
> > > folio_nid(page_folio(page)));
> > >         if (!entry) {
> > >                 zswap_reject_kmemcache_fail++;
> > > -               goto reject;
> > > +               return false;
> > >         }
> > >
> > > -       /* zswap_store() already holds a ref on 'objcg' and 'pool' */
> > > -       if (objcg)
> > > -               obj_cgroup_get(objcg);
> > > -       zswap_pool_get(pool);
> > > -
> > > -       /* if entry is successfully added, it keeps the reference */
> > > -       entry->pool = pool;
> > > -
> > > -       if (!zswap_compress(page, entry))
> > > -               goto put_pool_objcg;
> > > -
> > > -       entry->swpentry = page_swap_entry(page);
> > > -       entry->objcg = objcg;
> > > -       entry->referenced = true;
> > > +       if (!zswap_compress(page, entry, pool))
> > > +               goto compress_failed;
> > >
> > >         old = xa_store(tree, swp_offset(entry->swpentry), entry,
> GFP_KERNEL);
> > >         if (xa_is_err(old)) {
> > > @@ -1470,6 +1459,16 @@ static bool zswap_store_page(struct page
> *page,
> > >         if (old)
> > >                 zswap_entry_free(old);
> > >
> > > +       /*
> > > +        * The entry is successfully compressed and stored in the tree, there
> is
> > > +        * no further possibility of failure. Grab refs to the pool and objcg.
> > > +        * These refs will be dropped by zswap_entry_free() when the entry
> is
> > > +        * removed from the tree.
> > > +        */
> > > +       zswap_pool_get(pool);
> > > +       if (objcg)
> > > +               obj_cgroup_get(objcg);
> > > +
> > >         /*
> > >          * We finish initializing the entry while it's already in xarray.
> > >          * This is safe because:
> > > @@ -1480,26 +1479,22 @@ static bool zswap_store_page(struct page
> > > *page,
> > >          *    The publishing order matters to prevent writeback from seeing
> > >          *    an incoherent entry.
> > >          */
> 
> I am referring to the comment here ^
> 
> > > +       entry->pool = pool;
> > > +       entry->swpentry = page_swap_entry(page);
> > > +       entry->objcg = objcg;
> > > +       entry->referenced = true;
> > >         if (entry->length) {
> > >                 *compressed_bytes += entry->length;
> > >                 INIT_LIST_HEAD(&entry->lru);
> > >                 zswap_lru_add(&zswap_list_lru, entry);
> > >         }