Re: [PATCH v4 2/6] mm/zswap: reuse dstmem when decompress

[Barry Song <21cnbao@gmail.com>]

On Wed, Dec 27, 2023 at 4:56 AM Chengming Zhou
<zhouchengming@bytedance.com> wrote:
>
> In the !zpool_can_sleep_mapped() case such as zsmalloc, we need to first
> copy the entry->handle memory to a temporary memory, which is allocated
> using kmalloc.
>
> Obviously we can reuse the per-compressor dstmem to avoid allocating
> every time, since it's percpu-compressor and protected in percpu mutex.

what is the benefit of this since we are actually increasing lock contention
by reusing this buffer between multiple compression and decompression
threads?

this mainly affects zsmalloc which can't sleep? do we have performance
data?

and it seems this patch is also negatively affecting z3fold and zbud.c
which actually don't need to allocate a tmp buffer.

>
> Reviewed-by: Nhat Pham <nphamcs@gmail.com>
> Reviewed-by: Yosry Ahmed <yosryahmed@google.com>
> Acked-by: Chris Li <chrisl@kernel.org> (Google)
> Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
> ---
>  mm/zswap.c | 44 ++++++++++++--------------------------------
>  1 file changed, 12 insertions(+), 32 deletions(-)
>
> diff --git a/mm/zswap.c b/mm/zswap.c
> index 976f278aa507..6b872744e962 100644
> --- a/mm/zswap.c
> +++ b/mm/zswap.c
> @@ -1417,19 +1417,13 @@ static int zswap_writeback_entry(struct zswap_entry *entry,
>         struct crypto_acomp_ctx *acomp_ctx;
>         struct zpool *pool = zswap_find_zpool(entry);
>         bool page_was_allocated;
> -       u8 *src, *tmp = NULL;
> +       u8 *src;
>         unsigned int dlen;
>         int ret;
>         struct writeback_control wbc = {
>                 .sync_mode = WB_SYNC_NONE,
>         };
>
> -       if (!zpool_can_sleep_mapped(pool)) {
> -               tmp = kmalloc(PAGE_SIZE, GFP_KERNEL);
> -               if (!tmp)
> -                       return -ENOMEM;
> -       }
> -
>         /* try to allocate swap cache page */
>         mpol = get_task_policy(current);
>         page = __read_swap_cache_async(swpentry, GFP_KERNEL, mpol,
> @@ -1465,15 +1459,15 @@ static int zswap_writeback_entry(struct zswap_entry *entry,
>         /* decompress */
>         acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
>         dlen = PAGE_SIZE;
> +       mutex_lock(acomp_ctx->mutex);
>
>         src = zpool_map_handle(pool, entry->handle, ZPOOL_MM_RO);
>         if (!zpool_can_sleep_mapped(pool)) {
> -               memcpy(tmp, src, entry->length);
> -               src = tmp;
> +               memcpy(acomp_ctx->dstmem, src, entry->length);
> +               src = acomp_ctx->dstmem;
>                 zpool_unmap_handle(pool, entry->handle);
>         }
>
> -       mutex_lock(acomp_ctx->mutex);
>         sg_init_one(&input, src, entry->length);
>         sg_init_table(&output, 1);
>         sg_set_page(&output, page, PAGE_SIZE, 0);
> @@ -1482,9 +1476,7 @@ static int zswap_writeback_entry(struct zswap_entry *entry,
>         dlen = acomp_ctx->req->dlen;
>         mutex_unlock(acomp_ctx->mutex);
>
> -       if (!zpool_can_sleep_mapped(pool))
> -               kfree(tmp);
> -       else
> +       if (zpool_can_sleep_mapped(pool))
>                 zpool_unmap_handle(pool, entry->handle);
>
>         BUG_ON(ret);
> @@ -1508,9 +1500,6 @@ static int zswap_writeback_entry(struct zswap_entry *entry,
>         return ret;
>
>  fail:
> -       if (!zpool_can_sleep_mapped(pool))
> -               kfree(tmp);
> -
>         /*
>          * If we get here because the page is already in swapcache, a
>          * load may be happening concurrently. It is safe and okay to
> @@ -1771,7 +1760,7 @@ bool zswap_load(struct folio *folio)
>         struct zswap_entry *entry;
>         struct scatterlist input, output;
>         struct crypto_acomp_ctx *acomp_ctx;
> -       u8 *src, *dst, *tmp;
> +       u8 *src, *dst;
>         struct zpool *zpool;
>         unsigned int dlen;
>         bool ret;
> @@ -1796,26 +1785,19 @@ bool zswap_load(struct folio *folio)
>         }
>
>         zpool = zswap_find_zpool(entry);
> -       if (!zpool_can_sleep_mapped(zpool)) {
> -               tmp = kmalloc(entry->length, GFP_KERNEL);
> -               if (!tmp) {
> -                       ret = false;
> -                       goto freeentry;
> -               }
> -       }
>
>         /* decompress */
>         dlen = PAGE_SIZE;
> -       src = zpool_map_handle(zpool, entry->handle, ZPOOL_MM_RO);
> +       acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
> +       mutex_lock(acomp_ctx->mutex);
>
> +       src = zpool_map_handle(zpool, entry->handle, ZPOOL_MM_RO);
>         if (!zpool_can_sleep_mapped(zpool)) {
> -               memcpy(tmp, src, entry->length);
> -               src = tmp;
> +               memcpy(acomp_ctx->dstmem, src, entry->length);
> +               src = acomp_ctx->dstmem;
>                 zpool_unmap_handle(zpool, entry->handle);
>         }
>
> -       acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
> -       mutex_lock(acomp_ctx->mutex);
>         sg_init_one(&input, src, entry->length);
>         sg_init_table(&output, 1);
>         sg_set_page(&output, page, PAGE_SIZE, 0);
> @@ -1826,15 +1808,13 @@ bool zswap_load(struct folio *folio)
>
>         if (zpool_can_sleep_mapped(zpool))
>                 zpool_unmap_handle(zpool, entry->handle);
> -       else
> -               kfree(tmp);
>
>         ret = true;
>  stats:
>         count_vm_event(ZSWPIN);
>         if (entry->objcg)
>                 count_objcg_event(entry->objcg, ZSWPIN);
> -freeentry:
> +
>         spin_lock(&tree->lock);
>         if (ret && zswap_exclusive_loads_enabled) {
>                 zswap_invalidate_entry(tree, entry);
>
> --
> b4 0.10.1
>