Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 27 Sep 2022 21:53:50 +0900
From:   Naoya Horiguchi <naoya.horiguchi@linux.dev>
To:     Hui Tang <tanghui20@huawei.com>
Cc:     naoya.horiguchi@nec.com, linmiaohe@huawei.com,
        akpm@linux-foundation.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] mm/hwpoison: fix build error without
 CONFIG_MEMORY_FAILURE
Message-ID: <20220927125350.GA500539@u2004.lan>
References: <20220927102946.98622-1-tanghui20@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20220927102946.98622-1-tanghui20@huawei.com>
Precedence: bulk

On Tue, Sep 27, 2022 at 06:29:46PM +0800, Hui Tang wrote:
> Building without CONFIG_MEMORY_FAILURE will fail:
> 
> mm/memory-failure.o: In function `action_result':
> memory-failure.c: undefined reference to `memblk_nr_poison_inc'
> mm/memory-failure.o: In function `page_handle_poison':
> memory-failure.c: undefined reference to `memblk_nr_poison_inc'
> mm/memory-failure.o: In function `__get_huge_page_for_hwpoison':
> memory-failure.c: undefined reference to `memblk_nr_poison_inc'
> mm/memory-failure.o: In function `unpoison_memory':
> memory-failure.c: undefined reference to `memblk_nr_poison_sub'
> mm/memory-failure.o: In function `num_poisoned_pages_inc':
> memory-failure.c: undefined reference to `memblk_nr_poison_inc'
> 
> Add CONFIG_MEMORY_FAILURE wrapper for invoking memblk_nr_poison_{inc|sub}.
> 
> Fixes: 69b496f03bb4 ("mm/hwpoison: introduce per-memory_block hwpoison counter")
> Signed-off-by: Hui Tang <tanghui20@huawei.com>

Thank you for the patch.  I have a question.
If you disables CONFIG_MEMORY_FAILURE, mm/memory-failure.c should
not be compiled, so I wonder why you saw the compile error.
Could you share your .config file?

Acutally I saw the similar report a few days ago
https://lore.kernel.org/lkml/20220923095013.1151252-1-michael@walle.cc/
, where the build error happened in aarch64,  so I likely missed some arch
dependency.  I responded to the report by updating the patch by applying
"#if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)"
to the definition of memblk_nr_poison_{inc,sub}.
https://lore.kernel.org/lkml/20220923141204.GA1484969@ik1-406-35019.vs.sakura.ne.jp/

I did not confirm that this fix is really right, because I can't reproduced
the build error in my environment.  So could you check that v5 patch fixes
the build error you're seeing?  (The current version in mm-unstable is v4,
so that should be replaced with v5.)

Thanks,
Naoya Horiguchi

> ---
>  mm/memory-failure.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index 03479895086d..5bb9d2d20234 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -77,14 +77,18 @@ static bool hw_memory_failure __read_mostly = false;
>  void num_poisoned_pages_inc(unsigned long pfn)
>  {
>  	atomic_long_inc(&num_poisoned_pages);
> +#ifdef CONFIG_MEMORY_FAILURE
>  	memblk_nr_poison_inc(pfn);
> +#endif
>  }
>  
>  static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
>  {
>  	atomic_long_sub(i, &num_poisoned_pages);
> +#ifdef CONFIG_MEMORY_FAILURE
>  	if (pfn != -1UL)
>  		memblk_nr_poison_sub(pfn, i);
> +#endif
>  }
>  
>  /*
> -- 
> 2.17.1
> 
> 
>