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:   Mon, 5 Dec 2022 10:24:20 +0800
From:   Baoquan He <bhe@redhat.com>
To:     Dennis Zhou <dennis@kernel.org>
Cc:     Wupeng Ma <mawupeng1@huawei.com>, akpm@linux-foundation.org,
        tj@kernel.org, cl@linux.com, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH -next 1/1] percpu: cleanup invalid assignment to err in
 pcpu_alloc
Message-ID: <Y41WVP5vgKBpdMe1@MiWiFi-R3L-srv>
References: <20221204031430.662169-1-mawupeng1@huawei.com>
 <Y4yOMmnrT772wFSL@MiWiFi-R3L-srv>
 <Y407wDMKq5ibE9sc@fedora>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y407wDMKq5ibE9sc@fedora>
Precedence: bulk

On 12/04/22 at 04:30pm, Dennis Zhou wrote:
> Hi Baoquan and Wupeng,
> 
> On Sun, Dec 04, 2022 at 08:11:23PM +0800, Baoquan He wrote:
> > On 12/04/22 at 11:14am, Wupeng Ma wrote:
> > > From: Ma Wupeng <mawupeng1@huawei.com>
> > > 
> > > Assignment to err if is_atomic is true will never be used since warn
> > > message can only be shown if is_atomic is false after label fail. So drop
> > > it.
> > > 
> > > Signed-off-by: Ma Wupeng <mawupeng1@huawei.com>
> > > ---
> > >  mm/percpu.c | 4 +---
> > >  1 file changed, 1 insertion(+), 3 deletions(-)
> > > 
> > > diff --git a/mm/percpu.c b/mm/percpu.c
> > > index acd78da0493b..df86d79325b2 100644
> > > --- a/mm/percpu.c
> > > +++ b/mm/percpu.c
> > > @@ -1817,10 +1817,8 @@ static void __percpu *pcpu_alloc(size_t size, size_t align, bool reserved,
> > >  
> > >  	spin_unlock_irqrestore(&pcpu_lock, flags);
> > >  
> > > -	if (is_atomic) {
> > > -		err = "atomic alloc failed, no space left";
> > > +	if (is_atomic)
> > >  		goto fail;
> > > -	}
> > 
> > This is good catch. But I think Dennis may not like this way because he
> > added the message intentionally in commit 11df02bf9bc1 ("percpu: resolve
> > err may not be initialized in pcpu_alloc").
> > 
> 
> You're right Baoquan haha. I agree with Christoph as well we should
> surface atomic.
> 
> Though I don't think below is quite right either. We should likely have
> a separate warn_limit for atomic and I need to think about dump_stack()
> if there are any requirements there.

Yeah, sounds reasonable. I didn't think it over. 

> 
> 
> > Can we change the conditional checking in fail part as below?
> > 
> > diff --git a/mm/percpu.c b/mm/percpu.c
> > index 27697b2429c2..0ac55500fad9 100644
> > --- a/mm/percpu.c
> > +++ b/mm/percpu.c
> > @@ -1897,7 +1897,7 @@ static void __percpu *pcpu_alloc(size_t size, size_t align, bool reserved,
> >  fail:
> >  	trace_percpu_alloc_percpu_fail(reserved, is_atomic, size, align);
> >  
> > -	if (!is_atomic && do_warn && warn_limit) {
> > +	if (do_warn && warn_limit) {
> >  		pr_warn("allocation failed, size=%zu align=%zu atomic=%d, %s\n",
> >  			size, align, is_atomic, err);
> >  		dump_stack();
> > 
>