Message-ID: <9368735.gdWEK3B62L@archbook>
Date:   Mon, 23 Aug 2021 16:35:42 +0200
From:   Nicolas Frattaroli <frattaroli.nicolas@...il.com>
To:     Liam Girdwood <lgirdwood@...il.com>,
        Mark Brown <broonie@...nel.org>,
        Jaroslav Kysela <perex@...ex.cz>,
        Takashi Iwai <tiwai@...e.com>,
        Heiko Stuebner <heiko@...ech.de>,
        Philipp Zabel <p.zabel@...gutronix.de>
Cc:     linux-kernel@...r.kernel.org, alsa-devel@...a-project.org,
        linux-arm-kernel@...ts.infradead.org,
        linux-rockchip@...ts.infradead.org
Subject: Re: [PATCH v2 1/4] ASoC: rockchip: add support for i2s-tdm controller

Hi Philipp,

On Montag, 23. August 2021 14:03:25 CEST Philipp Zabel wrote:
> Hi Nicolas,
> 
> On Fri, 2021-08-20 at 20:27 +0200, Nicolas Frattaroli wrote:
> [...]
> 
> > diff --git a/sound/soc/rockchip/rockchip_i2s_tdm.c
> > b/sound/soc/rockchip/rockchip_i2s_tdm.c new file mode 100644
> > index 000000000000..c02b66f3c913
> > --- /dev/null
> > +++ b/sound/soc/rockchip/rockchip_i2s_tdm.c
> > @@ -0,0 +1,1737 @@
> 
> [...]
> 
> > +static void rockchip_snd_xfer_reset_assert(struct rk_i2s_tdm_dev
> > *i2s_tdm,
> > +					   int tx_bank, int tx_offset,
> > +					   int rx_bank, int rx_offset)
> > +{
> > +	void __iomem *cru_reset;
> > +	unsigned long flags;
> > +
> > +	cru_reset = i2s_tdm->cru_base + i2s_tdm->soc_data->softrst_offset;
> > +
> > +	if (tx_bank == rx_bank) {
> > +		writel(BIT(tx_offset) | BIT(rx_offset) |
> > +		       (BIT(tx_offset) << 16) | (BIT(rx_offset) << 16),
> > +		       cru_reset + (tx_bank * 4));
> > +	} else {
> > +		local_irq_save(flags);
> > +		writel(BIT(tx_offset) | (BIT(tx_offset) << 16),
> > +		       cru_reset + (tx_bank * 4));
> > +		writel(BIT(rx_offset) | (BIT(rx_offset) << 16),
> > +		       cru_reset + (rx_bank * 4));
> > +		local_irq_restore(flags);
> > +	}
> > +}
> > +
> > +static void rockchip_snd_xfer_reset_deassert(struct rk_i2s_tdm_dev
> > *i2s_tdm, +					     int tx_bank, int 
tx_offset,
> > +					     int rx_bank, int rx_offset)
> > +{
> > +	void __iomem *cru_reset;
> > +	unsigned long flags;
> > +
> > +	cru_reset = i2s_tdm->cru_base + i2s_tdm->soc_data->softrst_offset;
> > +
> > +	if (tx_bank == rx_bank) {
> > +		writel((BIT(tx_offset) << 16) | (BIT(rx_offset) << 16),
> > +		       cru_reset + (tx_bank * 4));
> > +	} else {
> > +		local_irq_save(flags);
> > +		writel((BIT(tx_offset) << 16),
> > +		       cru_reset + (tx_bank * 4));
> > +		writel((BIT(rx_offset) << 16),
> > +		       cru_reset + (rx_bank * 4));
> > +		local_irq_restore(flags);
> > +	}
> > +}
> > +
> > +/*
> > + * Makes sure that both tx and rx are reset at the same time to sync lrck
> > + * when clk_trcm > 0.
> > + */
> > +static void rockchip_snd_xfer_sync_reset(struct rk_i2s_tdm_dev *i2s_tdm)
> > +{
> > +	int tx_id, rx_id;
> > +	int tx_bank, rx_bank, tx_offset, rx_offset;
> > +
> > +	if (!i2s_tdm->cru_base || !i2s_tdm->soc_data)
> > +		return;
> > +
> > +	tx_id = i2s_tdm->tx_reset_id;
> > +	rx_id = i2s_tdm->rx_reset_id;
> > +	if (tx_id < 0 || rx_id < 0)
> > +		return;
> > +
> > +	tx_bank = tx_id / 16;
> > +	tx_offset = tx_id % 16;
> > +	rx_bank = rx_id / 16;
> > +	rx_offset = rx_id % 16;
> > +	dev_dbg(i2s_tdm->dev,
> > +		"tx_bank: %d, rx_bank: %d, tx_offset: %d, rx_offset: %d\n",
> > +		tx_bank, rx_bank, tx_offset, rx_offset);
> > +
> > +	rockchip_snd_xfer_reset_assert(i2s_tdm, tx_bank, tx_offset,
> > +				       rx_bank, rx_offset);
> > +
> > +	udelay(150);
> > +
> > +	rockchip_snd_xfer_reset_deassert(i2s_tdm, tx_bank, tx_offset,
> > +					 rx_bank, rx_offset);
> > +}
> 
> I'm not too fond of reimplementing half a reset controller in here.
> The reset framework does not support synchronized (de)assertion of
> multiple reset controls, I wonder if this would be useful to add.
> Without having thought about this too hard, I could imagine this as an
> extension to the bulk API, or possibly a call to join multiple reset
> controls into a reset control array.

I agree, the code required for synchronised reset seems to be a good
candidate for a generalised solution elsewhere.
However, I'm not sure if I'm the right candidate to design this API
as my first kernel contribution when the board I'm currently testing
this with doesn't even utilise the synchronized reset.

If I come across an opportunity to test synchronised resets, I'll
definitely look into refactoring this. I'd also be happy to hear of
any other driver which is currently implementing its own synchronised
reset.

[...]
 
> > +
> > +	reset_control_assert(rc);
> > +	udelay(1);
> 
> What is the reason for the different delays in
> rockchip_snd_xfer_sync_reset() and rockchip_snd_reset()?
> 

Simply put: I have no idea. This is what downstream does, and it
appears to work for me. The git history of the downstream kernel
also doesn't tell me anything about why the sync reset is 150
and this one is 1.

I've got no device to test the sync reset with at the moment so
I'm wary of playing with the delay value.

> > +	reset_control_deassert(rc);
> > +}
> 
> [...]
> 
> > +static int rockchip_i2s_tdm_probe(struct platform_device *pdev)
> > +{
> > +	struct device_node *node = pdev->dev.of_node;
> > +	struct device_node *cru_node;
> > +	const struct of_device_id *of_id;
> > +	struct rk_i2s_tdm_dev *i2s_tdm;
> > +	struct resource *res;
> > +	void __iomem *regs;
> > +	int ret;
> > +	int val;
> > +
> > +	i2s_tdm = devm_kzalloc(&pdev->dev, sizeof(*i2s_tdm), GFP_KERNEL);
> > +	if (!i2s_tdm)
> > +		return -ENOMEM;
> > +
> > +	i2s_tdm->dev = &pdev->dev;
> > +
> > +	of_id = of_match_device(rockchip_i2s_tdm_match, &pdev->dev);
> > +	if (!of_id || !of_id->data)
> > +		return -EINVAL;
> > +
> > +	spin_lock_init(&i2s_tdm->lock);
> > +	i2s_tdm->soc_data = (struct rk_i2s_soc_data *)of_id->data;
> > +
> > +	i2s_tdm->frame_width = 64;
> > +	if (!of_property_read_u32(node, "rockchip,frame-width", &val)) {
> > +		if (val >= 32 && (val % 2 == 0) && val <= 512) {
> > +			i2s_tdm->frame_width = val;
> > +		} else {
> > +			dev_err(i2s_tdm->dev, "unsupported frame width: 
'%d'\n",
> > +				val);
> > +			return -EINVAL;
> > +		}
> > +	}
> > +
> > +	i2s_tdm->clk_trcm = TRCM_TXRX;
> > +	if (of_property_read_bool(node, "rockchip,trcm-sync-tx-only"))
> > +		i2s_tdm->clk_trcm = TRCM_TX;
> > +	if (of_property_read_bool(node, "rockchip,trcm-sync-rx-only")) {
> > +		if (i2s_tdm->clk_trcm) {
> > +			dev_err(i2s_tdm->dev, "invalid trcm-sync 
configuration\n");
> > +			return -EINVAL;
> > +		}
> > +		i2s_tdm->clk_trcm = TRCM_RX;
> > +	}
> > +	if (i2s_tdm->clk_trcm != TRCM_TXRX)
> > +		i2s_tdm_dai.symmetric_rate = 1;
> > +
> > +	i2s_tdm->tdm_fsync_half_frame =
> > +		of_property_read_bool(node, "rockchip,tdm-fsync-half-frame");
> > +
> > +	i2s_tdm->grf = syscon_regmap_lookup_by_phandle(node, "rockchip,grf");
> > +	if (IS_ERR(i2s_tdm->grf))
> > +		return dev_err_probe(i2s_tdm->dev, PTR_ERR(i2s_tdm->grf),
> > +				     "Error in rockchip,grf\n");
> > +
> > +	if (i2s_tdm->clk_trcm != TRCM_TXRX) {
> > +		cru_node = of_parse_phandle(node, "rockchip,cru", 0);
> > +		i2s_tdm->cru_base = of_iomap(cru_node, 0);
> 
> This is a bit ugly if there is another driver sitting on the
> rockchip,cru compatible node. Which reset controller driver is backing
> the reset controls below?

I'm not sure if I understand the question (I only just today learned that
reset controls have drivers) but I think the reset it is using in the
Quartz64 dts is backed by rk3568-cru. Let me know if I misunderstood you.

[...]

> > +static int rockchip_i2s_tdm_remove(struct platform_device *pdev)
> > +{
> > +	struct rk_i2s_tdm_dev *i2s_tdm = dev_get_drvdata(&pdev->dev);
> > +
> > +	pm_runtime_disable(&pdev->dev);
> > +	if (!pm_runtime_status_suspended(&pdev->dev))
> > +		i2s_tdm_runtime_suspend(&pdev->dev);
> > +
> > +	if (!IS_ERR(i2s_tdm->mclk_tx))
> > +		clk_prepare_enable(i2s_tdm->mclk_tx);
> > +	if (!IS_ERR(i2s_tdm->mclk_rx))
> > +		clk_prepare_enable(i2s_tdm->mclk_rx);
> 
> Why are we enabling these clocks now?
> 

I wondered that too while I was looking into the pm_runtime stuff,
and decided to just throw these two calls out. They don't seem to
make sense to me, and nothing I tested broke when I removed them.

If left in (and the code works as I hypothesise it works) then
this would simply re-enable clocks we have just disabled, which
would make the number of enable and disable calls unbalanced.
Highly sus, as the kids would say, and completely omits the other
mclk_tx_src etc. clocks. (Though those are forgotten elsewhere in
the code as well, which I have since fixed, along with any of the
review points I don't directly respond to.)

> > +	if (!IS_ERR(i2s_tdm->hclk))
> > +		clk_disable_unprepare(i2s_tdm->hclk);
> > +
> > +	return 0;
> > +}
> 
> regards
> Philipp

Regards,
Nicolas Frattaroli

Powered by blists - more mailing lists