Re: [PATCH] net/mlx4_en: ensure rx_desc updating reaches HW before prod db updating

From: "jianchao.wang" <jianchao.w.wang@oracle.com>
To: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Eric Dumazet <eric.dumazet@gmail.com>,
	Tariq Toukan <tariqt@mellanox.com>,
	junxiao.bi@oracle.com, netdev@vger.kernel.org,
	linux-rdma@vger.kernel.org, linux-kernel@vger.kernel.org,
	Saeed Mahameed <saeedm@mellanox.com>
Subject: Re: [PATCH] net/mlx4_en: ensure rx_desc updating reaches HW before prod db updating
Date: Tue, 23 Jan 2018 11:25:55 +0800	[thread overview]
Message-ID: <f577e24f-e718-9047-8ae3-6ce7fcb9ec18@oracle.com> (raw)
In-Reply-To: <20180122154734.GD14372@ziepe.ca>

Hi Jason

Thanks for your kindly response.

On 01/22/2018 11:47 PM, Jason Gunthorpe wrote:
>>> Yeah, mlx4 NICs in Google fleet receive trillions of packets per
>>> second, and we never noticed an issue.
>>>
>>> Although we are using a slightly different driver, using order-0 pages
>>> and fast pages recycling.
>>>
>>>
>> The driver we use will will set the page reference count to (size of pages)/stride, the
>> pages will be freed by networking stack when the reference become zero, and the order-3
>> pages maybe allocated soon, this give NIC device a chance to corrupt the pages which have
>>  been allocated by others, such as slab.
> But it looks like the wmb() is placed when stuffing new rx descriptors
> into the device - how can it prevent corruption of pages where
> ownership was transfered from device to the host? That sounds more like a
> rmb() is missing someplace to me...
> 
The device may see the prod_db updating before rx_desc updating.
Then it will get stale rx_descs.
These stale rx_descs may contain pages that has been freed to host.

> (Granted the missing wmb() is a bug, but it may not be fully solving this
>  issue??)the wmb() here fix one of the customer's test case.

Thanks
Jianchao