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;
MIME-Version: 1.0
References: <20221123065319.56756-1-zhengyongjun3@huawei.com> <CAK-6q+gWK3g-g+6ox240hEmVQu2xjq4ZfxGdO9AHoqKG-5Myfw@mail.gmail.com>
In-Reply-To: <CAK-6q+gWK3g-g+6ox240hEmVQu2xjq4ZfxGdO9AHoqKG-5Myfw@mail.gmail.com>
From:   Alexander Aring <aahringo@redhat.com>
Date:   Wed, 23 Nov 2022 08:49:54 -0500
Message-ID: <CAK-6q+gb074RyyfVnbkeQWiY7FDQf0j-jwJ0LfmM8ojafwy7-w@mail.gmail.com>
Subject: Re: [Cluster-devel] [PATCH -next] fs: dlm: Fix build error without DEBUG_LOCK_ALLOC
To:     Zheng Yongjun <zhengyongjun3@huawei.com>
Cc:     cluster-devel@redhat.com, linux-kernel@vger.kernel.org,
        zhengbin13@huawei.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi,

On Wed, Nov 23, 2022 at 8:39 AM Alexander Aring <aahringo@redhat.com> wrote:
>
> Hi,
>
> On Wed, Nov 23, 2022 at 8:37 AM Zheng Yongjun <zhengyongjun3@huawei.com> wrote:
> >
> > If CONFIG_DLM is y, DEBUG_LOCK_ALLOC is n, building fails:
> >
> > fs/dlm/lowcomms.o: In function `lowcomms_queue_rwork':
> > lowcomms.c:(.text+0x2ac): undefined reference to `lockdep_is_held'
> > lowcomms.c:(.text+0x308): undefined reference to `lockdep_is_held'
> > fs/dlm/lowcomms.o: In function `lowcomms_queue_swork':
> > lowcomms.c:(.text+0x3e0): undefined reference to `lockdep_is_held'
> > fs/dlm/lowcomms.o: In function `restore_callbacks':
> > lowcomms.c:(.text+0xd8c): undefined reference to `lockdep_is_held'
> > lowcomms.c:(.text+0xdcc): undefined reference to `lockdep_is_held'
> >
> > Make DLM select DEBUG_LOCK_ALLOC to fix this.
> >
>
> Thanks, there is a different approach already in dlm/next and I think
> we don't want to have dlm selecting a debug option.

btw. I am curious why lockdep defines a prototype and not an
implementation for it if LOCKDEP is not set. In my opinion there
should be some no-op functions for it that the compiler puts it out if
it's not set. Then we don't end in #idefs inside the implementation...

- Alex