linux-kernel - Re: [PATCH] hv_netvsc: Make sure out channel is fully opened on send

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20180313123521.5b486da1@xeon-e3>
Date:   Tue, 13 Mar 2018 12:35:21 -0700
From:   Stephen Hemminger <stephen@...workplumber.org>
To:     Mohammed Gamal <mgamal@...hat.com>
Cc:     netdev@...r.kernel.org, sthemmin@...rosoft.com, otubo@...hat.com,
        linux-kernel@...r.kernel.org, devel@...uxdriverproject.org,
        vkuznets@...hat.com, davem@...emloft.net
Subject: Re: [PATCH] hv_netvsc: Make sure out channel is fully opened on
 send

On Tue, 13 Mar 2018 20:06:50 +0100
Mohammed Gamal <mgamal@...hat.com> wrote:

> Dring high network traffic changes to network interface parameters
> such as number of channels or MTU can cause a kernel panic with a NULL
> pointer dereference. This is due to netvsc_device_remove() being
> called and deallocating the channel ring buffers, which can then be
> accessed by netvsc_send_pkt() before they're allocated on calling
> netvsc_device_add()
> 
> The patch fixes this problem by checking the channel state and returning
> ENODEV if not yet opened. We also move the call to hv_ringbuf_avail_percent()
> which may access the uninitialized ring buffer.
> 
> Signed-off-by: Mohammed Gamal <mgamal@...hat.com>
> ---
>  drivers/net/hyperv/netvsc.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
> index 0265d70..44a8358 100644
> --- a/drivers/net/hyperv/netvsc.c
> +++ b/drivers/net/hyperv/netvsc.c
> @@ -757,7 +757,7 @@ static inline int netvsc_send_pkt(
>  	struct netdev_queue *txq = netdev_get_tx_queue(ndev, packet->q_idx);
>  	u64 req_id;
>  	int ret;
> -	u32 ring_avail = hv_ringbuf_avail_percent(&out_channel->outbound);
> +	u32 ring_avail;
>  
>  	nvmsg.hdr.msg_type = NVSP_MSG1_TYPE_SEND_RNDIS_PKT;
>  	if (skb)
> @@ -773,7 +773,7 @@ static inline int netvsc_send_pkt(
>  
>  	req_id = (ulong)skb;
>  
> -	if (out_channel->rescind)
> +	if (out_channel->rescind || out_channel->state != CHANNEL_OPENED_STATE)
>  		return -ENODEV;
>  
>  	if (packet->page_buf_cnt) {
> @@ -791,6 +791,7 @@ static inline int netvsc_send_pkt(
>  				       VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
>  	}
>  
> +	ring_avail = hv_ringbuf_avail_percent(&out_channel->outbound);
>  	if (ret == 0) {
>  		atomic_inc_return(&nvchan->queue_sends);
>  

Thanks for your patch. Yes there are races with the current update
logic. The root cause goes higher up in the flow; the send queues should
be stopped before netvsc_device_remove is called. Solving it where you tried
to is racy and not going to work reliably.

Network patches should go to netdev@...r.kernel.org

You can't move the ring_avail check until after the vmbus_sendpacket because
that will break the flow control logic.

Instead, you should just move the avail_read check until just after the existing rescind
check.

Also, you shouldn't need to check for OPENED_STATE, just rescind is enough.