netdev - Re: [PATCH] net: fix a data race when get vlan device

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <5b62ff55-73a4-e10a-89da-b98216c5599b@huawei.com>
Date:   Fri, 16 Apr 2021 10:26:59 +0800
From:   Yunsheng Lin <linyunsheng@...wei.com>
To:     zhudi <zhudi21@...wei.com>, <davem@...emloft.net>,
        <kuba@...nel.org>
CC:     <netdev@...r.kernel.org>, <rose.chen@...wei.com>
Subject: Re: [PATCH] net: fix a data race when get vlan device

dependencyOn 2021/4/15 11:35, zhudi wrote:
> From: Di Zhu <zhudi21@...wei.com>
> 
> We encountered a crash: in the packet receiving process, we got an
> illegal VLAN device address, but the VLAN device address saved in vmcore
> is correct. After checking the code, we found a possible data
> competition:
> CPU 0:                             CPU 1:
>     (RCU read lock)                  (RTNL lock)
>     vlan_do_receive()		       register_vlan_dev()
>       vlan_find_dev()
> 
>         ->__vlan_group_get_device()	 ->vlan_group_prealloc_vid()
> 
> In vlan_group_prealloc_vid(), We need to make sure that kzalloc is
> executed before assigning a value to vlan devices array, otherwise we

As my understanding, there is a dependency between calling kzalloc() and
assigning the address(returned from kzalloc()) to vg->vlan_devices_arrays,
CPU and compiler can see the dependency, why can't it handling the
dependency before adding the smp_wmb()?

See CONTROL DEPENDENCIES section in Documentation/memory-barriers.txt:

However, stores are not speculated.  This means that ordering -is- provided
for load-store control dependencies, as in the following example:

        q = READ_ONCE(a);
        if (q) {
                WRITE_ONCE(b, 1);
        }



> may get a wrong address from the hardware cache on another cpu.
> 
> So fix it by adding memory barrier instruction to ensure the order
> of memory operations.
> 
> Signed-off-by: Di Zhu <zhudi21@...wei.com>
> ---
>  net/8021q/vlan.c | 2 ++
>  net/8021q/vlan.h | 3 +++
>  2 files changed, 5 insertions(+)
> 
> diff --git a/net/8021q/vlan.c b/net/8021q/vlan.c
> index 8b644113715e..4f541e05cd3f 100644
> --- a/net/8021q/vlan.c
> +++ b/net/8021q/vlan.c
> @@ -71,6 +71,8 @@ static int vlan_group_prealloc_vid(struct vlan_group *vg,
>  	if (array == NULL)
>  		return -ENOBUFS;
>  
> +	smp_wmb();
> +
>  	vg->vlan_devices_arrays[pidx][vidx] = array;
>  	return 0;
>  }
> diff --git a/net/8021q/vlan.h b/net/8021q/vlan.h
> index 953405362795..7408fda084d3 100644
> --- a/net/8021q/vlan.h
> +++ b/net/8021q/vlan.h
> @@ -57,6 +57,9 @@ static inline struct net_device *__vlan_group_get_device(struct vlan_group *vg,
>  
>  	array = vg->vlan_devices_arrays[pidx]
>  				       [vlan_id / VLAN_GROUP_ARRAY_PART_LEN];
> +
> +	smp_rmb();
> +
>  	return array ? array[vlan_id % VLAN_GROUP_ARRAY_PART_LEN] : NULL;
>  }
>  
>