All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Gianluca Boiano" <morf3089@gmail.com>
Cc: linux-sound@vger.kernel.org, devicetree@vger.kernel.org,
	robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH v4 2/2] ASoC: codecs: add TAS2557 mono smart amplifier driver
Date: Fri, 20 Nov 2026 01:02:04 +0000	[thread overview]
Message-ID: <sashiko-outbox-224281@kernel.org> (raw)
In-Reply-To: <20261120004414.1686884-3-morf3089@gmail.com>

Thank you for your contribution! Sashiko AI review found 4 potential
issues to consider:
- [Medium] ASoC: tas2557: tas2557_power_up_locked() clobbers unmuted
  state on reload
- [Medium] ASoC: tas2557: tas2557_startup_sequence() powers PLL and
  clk-err before ASI
- [Low] ASoC: tas2557: missing offset check for bulk write in
  tas2557_load_block()
- [Low] ASoC: tas2557: tas2557_parse_dt() fails probe with -ENOSYS when
  !CONFIG_OF
--

commit b6c7726849d63f604c14fd11d9de768bf9364c52
Author: Gianluca Boiano <morf3089@gmail.com>

ASoC: codecs: add TAS2557 mono smart amplifier driver

This commit adds an ASoC codec driver for the Texas Instruments TAS2557
mono Class-D smart amplifier, supporting single and dual-chip groups, DSP
firmware loading, and fault recovery.

> diff --git a/sound/soc/codecs/tas2557.c b/sound/soc/codecs/tas2557.c
> new file mode 100644
> index 0000000000000..c49917149c863
> --- /dev/null
> +++ b/sound/soc/codecs/tas2557.c
[ ... ]
> +static const unsigned int tas2557_startup_common_data[] = {
> +	TAS2557_GPI_PIN_REG,      0x15,	/* enable DIN, MCLK, CCI */
> +	/* Power sequencing: Class-D + Boost first, then DSP/PLL */
> +	TAS2557_POWER_CTRL2_REG, 0xA0,		/* Class-D, Boost power up */
> +	TAS2557_POWER_CTRL2_REG, 0xA3,		/* Class-D, Boost, IV-sense power up */
> +	TAS2557_POWER_CTRL1_REG, 0xF8,		/* PLL, DSP, clock dividers power up */
> +	TAS2557_UDELAY, 2000,			/* 2 ms stabilisation */
> +	/* CE1_IC = ASI1; TI's ASI2 build writes the same value */
> +	TAS2557_CLK_ERR_CTRL, 0x2b,		/* enable clock error detection */
> +	TAS2557_DBOOST_CFG_REG, 0x0b,		/* reduce full-band noise */
> +	0xFFFFFFFF, 0xFFFFFFFF
> +};
[ ... ]
> +static int tas2557_startup_sequence(struct tas2557_priv *tas2557)
> +{
> +	const unsigned int *asi = tas2557->asi_port == TAS2557_ASI_PORT1 ?
> +				  tas2557_startup_asi1_data :
> +				  tas2557_startup_asi2_data;
> +	unsigned int dout_reg = tas2557_asi_reg(tas2557, TAS2557_GPIO3_PIN_REG,
> +						TAS2557_GPIO7_PIN_REG);
> +	unsigned int chn;
> +	int ret;
> +
> +	ret = tas2557_load_data_all(tas2557, tas2557_startup_common_data);
> +	if (ret < 0)
> +		return ret;
> +
> +	ret = tas2557_load_data_all(tas2557, asi);
> +	if (ret < 0)
> +		return ret;
> +
> +	/* Only chip 0 drives DOUT of the shared ASI */
> +	for (chn = 1; chn < tas2557->tas_priv->ndev; chn++) {
> +		ret = tas2557_dev_write(tas2557, chn, dout_reg, 0x00);
> +		if (ret < 0)
> +			return ret;
> +	}
> +
> +	return 0;
> +}

[Severity: Medium]
Does tas2557_startup_sequence() power up the PLL and arm clock-error
detection before the ASI clock pins and dividers are enabled?

tas2557_startup_common_data powers up the PLL/DSP, waits 2 ms, and writes
0x2b to TAS2557_CLK_ERR_CTRL while the ASI BCLK/WCLK GPIO muxes and ASI2
dividers in the asi table are still disabled.  In a two-chip group, chip 0
sits with clock-error detection armed and its clock inputs disabled for
over 2 ms while chip 1 runs tas2557_startup_common_data, exceeding the
0.34 ms hysteresis in TAS2557_CLK_ERR_CTRL2 and tripping a clock fault
before chip 0's ASI pins are configured.  Writing TAS2557_CLK_ERR_CTRL in
tas2557_startup_common_data also arms detection when dac_muted is true and
the host has not started the I2S clocks, until tas2557_apply_mute_state()
clears it later.  In addition, tas2557_load_data_all(tas2557, asi)
enables DOUT on every chip before the loop zeroes dout_reg on chn >= 1,
briefly driving the shared DOUT line from both chips at once.

[ ... ]

> +static int tas2557_load_block(struct tas2557_priv *tas2557,
> +			      struct tas2557_block *block, unsigned int chn)
> +{
> +	const u8 *data = block->data;
> +	unsigned int i;
> +	u8 book, page, offset, value;
> +	u16 sleep_time, bulk_len;
> +	int ret;
> +
> +	dev_dbg(tas2557->tas_priv->dev, "chn %u: load block type=0x%x cmds=%u\n",
> +		chn, block->type, block->num_commands);
> +
> +	for (i = 0; i < block->num_commands; ) {
> +		book   = data[i * 4];
> +		page   = data[i * 4 + 1];
> +		offset = data[i * 4 + 2];
> +		value  = data[i * 4 + 3];
> +		i++;
> +
> +		if (offset <= 0x7f) {
> +			ret = tas2557_dev_write(tas2557, chn,
> +						TAS2557_REG(book, page, offset),
> +						value);
> +			if (ret < 0)
> +				return ret;
> +		} else if (offset == 0x81) {
> +			sleep_time = ((u16)book << 8) | page;
> +			msleep(sleep_time);
> +		} else if (offset == 0x85) {
> +			bulk_len = ((u16)book << 8) | page;
> +
> +			if (bulk_len == 0)
> +				return -EINVAL;
> +
> +			if (i >= block->num_commands)
> +				return -EINVAL;
> +			/* Verify bulk payload stays within block->data */
> +			if ((size_t)i * 4 + 3 + bulk_len >
> +			    (size_t)block->num_commands * 4)
> +				return -EINVAL;
> +
> +			book   = data[i * 4];
> +			page   = data[i * 4 + 1];
> +			offset = data[i * 4 + 2];
> +
> +			if (bulk_len > 1)
> +				ret = tas2557_dev_bulk_write(tas2557, chn,
> +							     TAS2557_REG(book, page, offset),
> +							     &data[i * 4 + 3], bulk_len);
> +			else
> +				ret = tas2557_dev_write(tas2557, chn,
> +							TAS2557_REG(book, page, offset),
> +							data[i * 4 + 3]);

[Severity: Low]
Can a bulk write command (0x85) with offset > 0x7f overflow into the next
page in TAS2557_REG(book, page, offset)?

Unlike the single-write branch, the 0x85 branch does not check that the
target register offset is <= 0x7f.  Because TASDEVICE_REG() computes
(book * 256 * 128) + (page * 128) + reg and TASDEVICE_PAGE_REG() reduces
reg modulo 128, an offset >= 0x80 increments the encoded page (or book)
and wraps TASDEVICE_PAGE_REG(reg) to offset - 128, bypassing the
TASDEVICE_PAGE_REG(reg) + len > 128 check in tas2557_dev_bulk_write().

[ ... ]

> +static int tas2557_power_up_locked(struct tas2557_priv *tas2557,
> +				   unsigned int prog_idx)
> +{
> +	int cfg_idx, ret;
> +
> +	lockdep_assert_held(&tas2557->tas_priv->codec_lock);
> +
> +	cfg_idx = tas2557_find_config(tas2557->fw, prog_idx, tas2557->sample_rate);
> +	if (cfg_idx < 0) {
> +		dev_err(tas2557->tas_priv->dev, "no config for program %u at %u Hz\n",
> +			prog_idx, tas2557->sample_rate);
> +		return cfg_idx;
> +	}
> +
> +	if (tas2557->powered || !tas2557->prog_valid ||
> +	    tas2557->devs[0].current_program != prog_idx ||
> +	    tas2557->devs[0].current_config != cfg_idx) {
> +		if (tas2557->powered) {
> +			ret = tas2557_shutdown_sequence(tas2557);
> +			if (ret < 0)
> +				return ret;
> +			/* Callers set powered again on success */
> +			tas2557->powered = false;
> +			tas2557->dac_muted = true;
> +		}
> +
> +		ret = tas2557_set_program(tas2557, prog_idx, cfg_idx);
> +		if (ret < 0) {
> +			dev_err(tas2557->tas_priv->dev,
> +				"firmware program load failed on power-up: %d\n",
> +				ret);
> +			return ret;
> +		}
> +	}
[ ... ]
> +	ret = tas2557_apply_mute_state(tas2557);
> +	if (ret < 0)
> +		goto err_shutdown;

[Severity: Medium]
Does setting tas2557->dac_muted = true when shutting down an
already-powered group prevent tas2557_apply_mute_state() from restoring an
unmuted state?

When tas2557_hw_params() calls tas2557_power_up_locked() for a live
rate_changed reload while the amplifier is powered and unmuted,
tas2557->dac_muted is overwritten with true before tas2557_set_program()
runs.  tas2557_apply_mute_state() then sees dac_muted == true (and
tas2557_hw_params() skips remute when rate_changed is true), leaving the
amplifier muted and clock-error detection disabled.

[ ... ]

> +static int tas2557_parse_dt(struct tas2557_priv *tas2557, struct i2c_client *client)
> +{
[ ... ]
> +	/* Which ASI's pins/dividers/format registers to program */
> +	tas2557->asi_port = TAS2557_ASI_PORT_DEFAULT;
> +	ret = of_property_read_u32(np, "ti,asi-port", &tas2557->asi_port);
> +	if (ret && ret != -EINVAL)
> +		return dev_err_probe(dev, ret, "failed to read ti,asi-port\n");

[Severity: Low]
Does of_property_read_u32() fail probe with -ENOSYS when CONFIG_OF is
disabled?

Earlier in tas2557_parse_dt(), of_property_read_reg() is guarded with
if (np) so non-OF instantiation falls back to client->addr and
client->irq, and SND_SOC_TAS2557 does not depend on OF.  When CONFIG_OF=n,
the of_property_read_u32() stub returns -ENOSYS rather than -EINVAL, which
causes this check to abort probe.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261120004414.1686884-1-morf3089@gmail.com?part=2

  reply	other threads:[~2026-11-20  1:02 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-20  0:43 [PATCH v4 0/2] ASoC: codecs: add Texas Instruments TAS2557 smart amplifier driver Gianluca Boiano
2026-11-20  0:43 ` [PATCH v4 1/2] ASoC: dt-bindings: add Texas Instruments TAS2557 Gianluca Boiano
2026-11-20  0:43 ` [PATCH v4 2/2] ASoC: codecs: add TAS2557 mono smart amplifier driver Gianluca Boiano
2026-11-20  1:02   ` sashiko-bot [this message]
2026-11-20 15:14 ` [PATCH v4 0/2] ASoC: codecs: add Texas Instruments TAS2557 " David Heidelberg

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-224281@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=linux-sound@vger.kernel.org \
    --cc=morf3089@gmail.com \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.