Re: [PATCH 02/14] phy: allwinner: phy-sun6i-mipi-dphy: Support D-PHY Rx mode for MIPI CSI-2

From: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
To: Maxime Ripard <maxime@cerno.tech>
Cc: linux-media@vger.kernel.org, devicetree@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org, devel@driverdev.osuosl.org,
	linux-sunxi@googlegroups.com,
	Mauro Carvalho Chehab <mchehab@kernel.org>,
	Rob Herring <robh+dt@kernel.org>, Chen-Yu Tsai <wens@csie.org>,
	Yong Deng <yong.deng@magewell.com>,
	Kishon Vijay Abraham I <kishon@ti.com>,
	Vinod Koul <vkoul@kernel.org>,
	Helen Koike <helen.koike@collabora.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Philipp Zabel <p.zabel@pengutronix.de>,
	Hans Verkuil <hans.verkuil@cisco.com>,
	Sakari Ailus <sakari.ailus@linux.intel.com>,
	Thomas Petazzoni <thomas.petazzoni@bootlin.com>,
	Hans Verkuil <hverkuil@xs4all.nl>,
	kevin.lhopital@hotmail.com
Subject: Re: [PATCH 02/14] phy: allwinner: phy-sun6i-mipi-dphy: Support D-PHY Rx mode for MIPI CSI-2
Date: Wed, 4 Nov 2020 11:53:34 +0100	[thread overview]
Message-ID: <20201104105334.GE285779@aptenodytes> (raw)
In-Reply-To: <20201027182811.j6372vdmls5yvhri@gilmour.lan>

[-- Attachment #1: Type: text/plain, Size: 6439 bytes --]

Hi,

On Tue 27 Oct 20, 19:28, Maxime Ripard wrote:
> 
> Hi,
> 
> On Tue, Oct 27, 2020 at 10:23:26AM +0100, Paul Kocialkowski wrote:
> > On Mon 26 Oct 20, 16:38, Maxime Ripard wrote:
> > > On Fri, Oct 23, 2020 at 07:45:34PM +0200, Paul Kocialkowski wrote:
> > > > The Allwinner A31 D-PHY supports both Rx and Tx modes. While the latter
> > > > is already supported and used for MIPI DSI this adds support for the
> > > > former, to be used with MIPI CSI-2.
> > > > 
> > > > This implementation is inspired by the Allwinner BSP implementation.
> > > 
> > > Mentionning which BSP you took this from would be helpful
> > 
> > Sure! It's from the Github repo linked from https://linux-sunxi.org/V3s.
> > Would you like that I mention this URL explicitly or would it be enough to
> > mention "Allwinner's V3s Linux SDK" as they seem to call it?
> 
> Yeah, that would be great
> > > > +static int sun6i_dphy_rx_power_on(struct sun6i_dphy *dphy)
> > > > +{
> > > > +	/* Physical clock rate is actually half of symbol rate with DDR. */
> > > > +	unsigned long mipi_symbol_rate = dphy->config.hs_clk_rate;
> > > > +	unsigned long dphy_clk_rate;
> > > > +	unsigned int rx_dly;
> > > > +	unsigned int lprst_dly;
> > > > +	u32 value;
> > > > +
> > > > +	dphy_clk_rate = clk_get_rate(dphy->mod_clk);
> > > > +	if (!dphy_clk_rate)
> > > > +		return -1;
> > > 
> > > Returning -1 is weird here?
> > 
> > What do you think would be a more appropriate error code to return?
> > It looks like some other drivers return -EINVAL when that happens (but many
> > don't do the check).
> 
> Yeah, EINVAL at least is better than ENOPERM 
> 
> > > > +
> > > > +	/* Hardcoded timing parameters from the Allwinner BSP. */
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_RX_TIME0_REG,
> > > > +		     SUN6I_DPHY_RX_TIME0_HS_RX_SYNC(255) |
> > > > +		     SUN6I_DPHY_RX_TIME0_HS_RX_CLK_MISS(255) |
> > > > +		     SUN6I_DPHY_RX_TIME0_LP_RX(255));
> > > > +
> > > > +	/*
> > > > +	 * Formula from the Allwinner BSP, with hardcoded coefficients
> > > > +	 * (probably internal divider/multiplier).
> > > > +	 */
> > > > +	rx_dly = 8 * (unsigned int)(dphy_clk_rate / (mipi_symbol_rate / 8));
> > > > +
> > > > +	/*
> > > > +	 * The Allwinner BSP has an alternative formula for LP_RX_ULPS_WP:
> > > > +	 * lp_ulps_wp_cnt = lp_ulps_wp_ms * lp_clk / 1000
> > > > +	 * but does not use it and hardcodes 255 instead.
> > > > +	 */
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_RX_TIME1_REG,
> > > > +		     SUN6I_DPHY_RX_TIME1_RX_DLY(rx_dly) |
> > > > +		     SUN6I_DPHY_RX_TIME1_LP_RX_ULPS_WP(255));
> > > > +
> > > > +	/* HS_RX_ANA0 value is hardcoded in the Allwinner BSP. */
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_RX_TIME2_REG,
> > > > +		     SUN6I_DPHY_RX_TIME2_HS_RX_ANA0(4));
> > > > +
> > > > +	/*
> > > > +	 * Formula from the Allwinner BSP, with hardcoded coefficients
> > > > +	 * (probably internal divider/multiplier).
> > > > +	 */
> > > > +	lprst_dly = 4 * (unsigned int)(dphy_clk_rate / (mipi_symbol_rate / 2));
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_RX_TIME3_REG,
> > > > +		     SUN6I_DPHY_RX_TIME3_LPRST_DLY(lprst_dly));
> > > > +
> > > > +	/* Analog parameters are hardcoded in the Allwinner BSP. */
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_ANA0_REG,
> > > > +		     SUN6I_DPHY_ANA0_REG_PWS |
> > > > +		     SUN6I_DPHY_ANA0_REG_SLV(7) |
> > > > +		     SUN6I_DPHY_ANA0_REG_SFB(2));
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_ANA1_REG,
> > > > +		     SUN6I_DPHY_ANA1_REG_SVTT(4));
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_ANA4_REG,
> > > > +		     SUN6I_DPHY_ANA4_REG_DMPLVC |
> > > > +		     SUN6I_DPHY_ANA4_REG_DMPLVD(1));
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_ANA2_REG,
> > > > +		     SUN6I_DPHY_ANA2_REG_ENIB);
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_ANA3_REG,
> > > > +		     SUN6I_DPHY_ANA3_EN_LDOR |
> > > > +		     SUN6I_DPHY_ANA3_EN_LDOC |
> > > > +		     SUN6I_DPHY_ANA3_EN_LDOD);
> > > > +
> > > > +	/*
> > > > +	 * Delay comes from the Allwinner BSP, likely for internal regulator
> > > > +	 * ramp-up.
> > > > +	 */
> > > > +	udelay(3);
> > > > +
> > > > +	value = SUN6I_DPHY_RX_CTL_EN_DBC | SUN6I_DPHY_RX_CTL_RX_CLK_FORCE;
> > > > +
> > > > +	/*
> > > > +	 * Rx data lane force-enable bits are used as regular RX enable by the
> > > > +	 * Allwinner BSP.
> > > > +	 */
> > > > +	if (dphy->config.lanes >= 1)
> > > > +		value |= SUN6I_DPHY_RX_CTL_RX_D0_FORCE;
> > > > +	if (dphy->config.lanes >= 2)
> > > > +		value |= SUN6I_DPHY_RX_CTL_RX_D1_FORCE;
> > > > +	if (dphy->config.lanes >= 3)
> > > > +		value |= SUN6I_DPHY_RX_CTL_RX_D2_FORCE;
> > > > +	if (dphy->config.lanes == 4)
> > > > +		value |= SUN6I_DPHY_RX_CTL_RX_D3_FORCE;
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_RX_CTL_REG, value);
> > > > +
> > > > +	regmap_write(dphy->regs, SUN6I_DPHY_GCTL_REG,
> > > > +		     SUN6I_DPHY_GCTL_LANE_NUM(dphy->config.lanes) |
> > > > +		     SUN6I_DPHY_GCTL_EN);
> > > > +
> > > > +	return 0;
> > > > +}
> > > > +
> > > > +static int sun6i_dphy_power_on(struct phy *phy)
> > > > +{
> > > > +	struct sun6i_dphy *dphy = phy_get_drvdata(phy);
> > > > +
> > > > +	switch (dphy->submode) {
> > > > +	case PHY_MIPI_DPHY_SUBMODE_TX:
> > > > +		return sun6i_dphy_tx_power_on(dphy);
> > > > +	case PHY_MIPI_DPHY_SUBMODE_RX:
> > > > +		return sun6i_dphy_rx_power_on(dphy);
> > > > +	default:
> > > > +		return -EINVAL;
> > > > +	}
> > > > +}
> > > > +
> > > 
> > > Can one call power_on before set_mode?
> > 
> > I didn't find anything indicating this is illegal. What would happen here is
> > that the D-PHY would be configured to PHY_MIPI_DPHY_SUBMODE_TX (submode == 0)
> > at power-on if set_mode is not called before.
> > 
> > I think it's fair to expect that it's too late to change the mode once the PHY
> > was powered on. Maybe we should return -EBUSY on set_mode when power on was
> > already requested?
> 
> Or maybe we can just clarify it in the framework/function documentation

Agreed, I'll add a patch in that direction. I would also be tempted to check on
phy->power_count to return -EBUSY in phy_set_mode_ext so that the behavior is
enforced.

What do you think?

Cheers,

Paul

-- 
Paul Kocialkowski, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com

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