Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH 2/2] blk-iocost: Use alloc_percpu_gfp() to simplify the
 code
To:     Tejun Heo <tj@kernel.org>
Cc:     axboe@kernel.dk, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <1ba7a38d5a6186b1e71432ef424c23ba1904a365.1607591591.git.baolin.wang@linux.alibaba.com>
 <aa518c5b5c7185e660a1c8515c10d9513fe92132.1607591591.git.baolin.wang@linux.alibaba.com>
 <X9Iv/MlqQI00wZRn@mtj.duckdns.org>
From:   Baolin Wang <baolin.wang@linux.alibaba.com>
Message-ID: <33480f8a-89a3-3ed9-6fd0-95b2944ccbdd@linux.alibaba.com>
Date:   Fri, 11 Dec 2020 15:13:29 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Thunderbird/78.5.1
MIME-Version: 1.0
In-Reply-To: <X9Iv/MlqQI00wZRn@mtj.duckdns.org>
Content-Type: text/plain; charset=gbk; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi Tejun,

> Hello,
> 
> On Thu, Dec 10, 2020 at 06:56:45PM +0800, Baolin Wang wrote:
>> Use alloc_percpu_gfp() with __GFP_ZERO flag, which can remove
>> some explicit initialization code.
> 
> __GFP_ZERO is implicit for percpu allocations and local[64]_t's initial
> states aren't guaranteed to be all zeros on different archs.

Thanks for teaching me this, at least I did not get this from the 
local_ops Documentation before. Just out of curiosity, these local[64]_t 
variables are also allocated from budy allocator ultimately, why they 
can not be initialized to zeros on some ARCHs with __GFP_ZERO? Could you 
elaborate on about this restriction? Thanks.

By the way, seems the kyber-iosched has the same issue, since the 
'struct kyber_cpu_latency' also contains an atomic_t variable.

	kqd->cpu_latency = alloc_percpu_gfp(struct kyber_cpu_latency,
					    GFP_KERNEL | __GFP_ZERO);
	if (!kqd->cpu_latency)
		goto err_kqd;