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;
Date:   Fri, 24 Jul 2020 09:39:42 +0200
From:   Michal Hocko <mhocko@kernel.org>
To:     Muchun Song <songmuchun@bytedance.com>
Cc:     mike.kravetz@oracle.com, akpm@linux-foundation.org,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        Jianchao Guo <guojianchao@bytedance.com>
Subject: Re: [PATCH] mm/hugetlb: add mempolicy check in the reservation
 routine
Message-ID: <20200724073942.GE4061@dhcp22.suse.cz>
References: <20200723074417.89467-1-songmuchun@bytedance.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200723074417.89467-1-songmuchun@bytedance.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu 23-07-20 15:44:17, Muchun Song wrote:
> In the reservation routine, we only check whether the cpuset meets
> the memory allocation requirements. But we ignore the mempolicy of
> MPOL_BIND case. If someone mmap hugetlb succeeds, but the subsequent
> memory allocation may fail due to mempolicy restrictions and receives
> the SIGBUS signal. This can be reproduced by the follow steps.
> 
>  1) Compile the test case.
>     cd tools/testing/selftests/vm/
>     gcc map_hugetlb.c -o map_hugetlb
> 
>  2) Pre-allocate huge pages. Suppose there are 2 numa nodes in the
>     system. Each node will pre-allocate one huge page.
>     echo 2 > /proc/sys/vm/nr_hugepages
> 
>  3) Run test case(mmap 4MB). We receive the SIGBUS signal.
>     numactl --membind=0 ./map_hugetlb 4

Cpusets and mempolicy interaction has always been a nightmare and
semantic might get really awkward in some cases. In this case I am not
really sure anybody really does soemthing like that but anyway...

[...]

> -static unsigned int cpuset_mems_nr(unsigned int *array)
> +static nodemask_t *mempolicy_current_bind_nodemask(void)
> +{
> +	struct mempolicy *mpol;
> +	nodemask_t *nodemask;
> +
> +	mpol = get_task_policy(current);
> +	if (mpol->mode == MPOL_BIND)
> +		nodemask = &mpol->v.nodes;
> +	else
> +		nodemask = NULL;
> +
> +	return nodemask;
> +}

We already have policy_nodemask which tries to do this. Is there any
reason to not reuse it?
-- 
Michal Hocko
SUSE Labs