Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH v3] mm/page_alloc: Further fix __alloc_pages_bulk() return
 value
To:     Chuck Lever <chuck.lever@oracle.com>, mgorman@techsingularity.net
Cc:     linux-nfs@vger.kernel.org, linux-mm@kvack.org, brouer@redhat.com
References: <162497449506.16614.7781489905877008435.stgit@klimt.1015granger.net>
From:   Jesper Dangaard Brouer <jbrouer@redhat.com>
Message-ID: <be9186d9-1e8e-4d99-ab0f-84c0518025c5@redhat.com>
Date:   Tue, 29 Jun 2021 18:01:12 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <162497449506.16614.7781489905877008435.stgit@klimt.1015granger.net>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Content-Language: en-US
Precedence: bulk

On 29/06/2021 15.48, Chuck Lever wrote:

> The call site in __page_pool_alloc_pages_slow() also seems to be
> confused on this matter. It should be attended to by someone who
> is familiar with that code.

I don't think we need a fix for __page_pool_alloc_pages_slow(), as the 
array is guaranteed to be empty.

But a fix would look like this:

diff --git a/net/core/page_pool.c b/net/core/page_pool.c
index c137ce308c27..1b04538a3da3 100644
--- a/net/core/page_pool.c
+++ b/net/core/page_pool.c
@@ -245,22 +245,23 @@ static struct page 
*__page_pool_alloc_pages_slow(struct page_pool *pool,
         if (unlikely(pp_order))
                 return __page_pool_alloc_page_order(pool, gfp);

         /* Unnecessary as alloc cache is empty, but guarantees zero 
count */
-       if (unlikely(pool->alloc.count > 0))
+       if (unlikely(pool->alloc.count > 0))   // ^^^^^^^^^^^^^^^^^^^^^^
                 return pool->alloc.cache[--pool->alloc.count];

         /* Mark empty alloc.cache slots "empty" for 
alloc_pages_bulk_array */
         memset(&pool->alloc.cache, 0, sizeof(void *) * bulk);

+       /* bulk API ret value also count existing pages, but array is 
empty */
         nr_pages = alloc_pages_bulk_array(gfp, bulk, pool->alloc.cache);
         if (unlikely(!nr_pages))
                 return NULL;

         /* Pages have been filled into alloc.cache array, but count is 
zero and
          * page element have not been (possibly) DMA mapped.
          */
-       for (i = 0; i < nr_pages; i++) {
+       for (i = pool->alloc.count; i < nr_pages; i++) {
                 page = pool->alloc.cache[i];
                 if ((pp_flags & PP_FLAG_DMA_MAP) &&
                     unlikely(!page_pool_dma_map(pool, page))) {
                         put_page(page);