Re: [PATCH v3 1/9] ASoC: sun4i-i2s: Adjust regmap settings

From: Mark Brown <broonie@kernel.org>
To: Chen-Yu Tsai <wens@csie.org>
Cc: Code Kipper <codekipper@gmail.com>,
	Maxime Ripard <maxime.ripard@free-electrons.com>,
	linux-sunxi <linux-sunxi@googlegroups.com>,
	linux-arm-kernel <linux-arm-kernel@lists.infradead.org>,
	Liam Girdwood <lgirdwood@gmail.com>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	Linux-ALSA <alsa-devel@alsa-project.org>,
	"Andrea Venturi (pers)" <be17068@iperbole.bo.it>
Subject: Re: [PATCH v3 1/9] ASoC: sun4i-i2s: Adjust regmap settings
Date: Wed, 9 Jan 2019 18:46:04 +0000	[thread overview]
Message-ID: <20190109184604.GJ10405@sirena.org.uk> (raw)
In-Reply-To: <CAGb2v66-JmONW9AYqoSmifrEj+nstREdsbkcy9ECwnJ47zxARA@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 1698 bytes --]

On Sat, Dec 22, 2018 at 12:44:07AM +0800, Chen-Yu Tsai wrote:
> On Fri, Dec 21, 2018 at 11:21 PM <codekipper@gmail.com> wrote:

> > +       regcache_cache_bypass(i2s->regmap, true);
> >         regmap_update_bits(i2s->regmap, SUN4I_I2S_FIFO_CTRL_REG,
> >                            SUN4I_I2S_FIFO_CTRL_FLUSH_RX,
> >                            SUN4I_I2S_FIFO_CTRL_FLUSH_RX);
> > +       regcache_cache_bypass(i2s->regmap, false);

> IIRC the flush cache bit is self-clearing. So you likely want to mark
> this register as volatile. If it is marked as volatile, then all access
> to that register bypasses the cache, so the regcache_cache_bypass calls
> are unneeded.

Yes, that should be the case.

> However, looking at the code, the write would seem to be ignored if the
> regmap is in the cache_only state. We only set this when the bus clock
> is disabled. Under such a condition, bypassing the cache and forcing a
> write would be unwise, as the system either drops the write, or stalls
> altogether.

Right, access to a cache only register while the device is in cache only
mode is not a great idea - the usual reason we're in cache only mode is
that the device is in a state where I/O isn't going to work.  One thing
that can work for this if you need the register to be cached (but is a
bit gross) is to do a write setting the self clearing bit then another
immediately after resetting it back to the cleared state.  That works OK
for cases where the bit is a strobe and never retains state, though if
the device isn't operational then needing to write to the register might
indicate a bigger picture logic error (or it could be that the register
map mixes random things into one register).

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]