[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210716080013.652969bf@cakuba>
Date: Fri, 16 Jul 2021 08:00:13 +0200
From: Jakub Kicinski <kuba@...nel.org>
To: Guangbin Huang <huangguangbin2@...wei.com>
Cc: <davem@...emloft.net>, <jiri@...dia.com>, <netdev@...r.kernel.org>,
<linux-kernel@...r.kernel.org>, <lipeng321@...wei.com>,
<chenhao288@...ilicon.com>
Subject: Re: [PATCH V2 net-next 1/9] devlink: add documentation for hns3
driver
On Thu, 15 Jul 2021 15:45:02 +0800, Guangbin Huang wrote:
> From: Hao Chen <chenhao288@...ilicon.com>
>
> Add a file to document devlink support for hns3 driver.
>
> Now support devlink param and devlink info.
>
> Signed-off-by: Hao Chen <chenhao288@...ilicon.com>
> Signed-off-by: Guangbin Huang <huangguangbin2@...wei.com>
> +This document describes the devlink features implemented by the ``hns3``
> +device driver.
> +
> +Parameters
> +==========
> +
> +The ``hns3`` driver implements the following driver-specific
> +parameters.
> +
> +.. list-table:: Driver-specific parameters implemented
> + :widths: 10 10 10 70
> +
> + * - Name
> + - Type
> + - Mode
> + - Description
> + * - ``rx_buf_len``
> + - U32
> + - driverinit
> + - Set rx BD buffer size, now only support setting 2048 and 4096.
> +
> + * The feature is used to change the buffer size of each BD of Rx ring
> + between 2KB and 4KB, then do devlink reload operation to take effect.
Does the reload required here differ from the reload performed when the
ring size is changed? You can extend the ethtool API, devlink params
should be used for very vendor specific configuration. Which page
fragment size very much is not.
> + * - ``tx_buf_size``
> + - U32
> + - driverinit
> + - Set tx bounce buf size.
> +
> + * The size is setted for tx bounce feature. Tx bounce buffer feature is
> + used for small size packet or frag. It adds a queue based tx shared
> + bounce buffer to memcpy the small packet when the len of xmitted skb is
> + below tx_copybreak(value to distinguish small size and normal size),
> + and reduce the overhead of dma map and unmap when IOMMU is on.
IMHO setting the tx_copybreak should be configured thru the same API as
the size of the buffer it uses. Hence, again, ethtool.
Powered by blists - more mailing lists