[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 26 Aug 2020 08:59:06 +0800
From: Bard liao <yung-chuan.liao@...ux.intel.com>
To: Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>,
alsa-devel@...a-project.org
Cc: linux-kernel@...r.kernel.org, tiwai@...e.de, broonie@...nel.org,
vkoul@...nel.org, gregkh@...uxfoundation.org, jank@...ence.com,
srinivas.kandagatla@...aro.org, slawomir.blauciak@...el.com,
Rander Wang <rander.wang@...ux.intel.com>,
Ranjani Sridharan <ranjani.sridharan@...ux.intel.com>,
Hui Wang <hui.wang@...onical.com>,
Guennadi Liakhovetski <guennadi.liakhovetski@...ux.intel.com>,
Kai Vehmanen <kai.vehmanen@...ux.intel.com>,
"Rafael J. Wysocki" <rafael@...nel.org>
Subject: Re: [PATCH 4/4] regmap: sdw: add support for SoundWire 1.2 MBQ
On 8/26/2020 1:16 AM, Pierre-Louis Bossart wrote:
> The SoundWire 1.1 specification only allowed for reads and writes of
> bytes. The SoundWire 1.2 specification adds a new capability to
> transfer "Multi-Byte Quantities" (MBQ) across the bus. The transfers
> still happens one-byte-at-a-time, but the update is atomic.
>
> For example when writing a 16-bit volume, the first byte transferred
> is only taken into account when the second byte is successfully
> transferred.
>
> The mechanism is symmetrical for read and writes:
> - On a read, the address of the last byte to be read is modified by
> setting the MBQ bit
> - On a write, the address of all but the last byte to be written are
> modified by setting the MBQ bit. The address for the last byte relies
> on the MBQ bit being cleared.
>
> The current definitions for MBQ-based controls in the SDCA draft
> standard are limited to 16 bits for volumes, so for now this is the
> only supported format. An update will be provided if and when support
> for 24-bit and 32-bit values is specified by the SDCA standard.
>
> One possible objection is that this code could have been handled with
> regmap-sdw.c. However this is a new spec addition not handled by every
> SoundWire 1.1 and non-SDCA device, so there's no reason to load code
> that will never be used.
>
> Also in practice it's extremely unlikely that CONFIG_REGMAP would not
> be selected with CONFIG_REGMAP_MBQ selected. However there's no
> functional dependency between the two modules so they can be selected
> separately.
>
> Reviewed-by: Rander Wang <rander.wang@...ux.intel.com>
> Reviewed-by: Guennadi Liakhovetski <guennadi.liakhovetski@...ux.intel.com>
> Reviewed-by: Kai Vehmanen <kai.vehmanen@...ux.intel.com>
> Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>
Acked-by: Bard Liao <yung-chuan.liao@...ux.intel.com>
Powered by blists - more mailing lists