Skip to contentSkip to conversation

pinctrl: mcp23s08: Propagate reset GPIO errors

PATCH In review [email protected] 3 participants 5 messages Oct 12

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 pinctrl: mcp23s08: Propagate reset GPIO errors

From: bui duc phuc <[email protected]>

devm_gpiod_get_optional() may return -EPROBE_DEFER or other errors, but mcp23s08_probe_one() currently ignores them.

Check the return value and propagate errors from the reset GPIO lookup.

Fixes 4e73bfa3ee37 ("pinctrl: mcp23s08: Add optional reset GPIO")

---

Changes in v2:
-return ERR_PTR(-EPROBE_DEFER) instead of uninitialized ret variable

 drivers/pinctrl/pinctrl-mcp23s08.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/pinctrl/pinctrl-mcp23s08.c +3 -0
index cc668f84bf8b..259ac02134c6 100644
@@ -613,6 +613,9 @@ int mcp23s08_probe_one(struct mcp23s08 *mcp, struct device *dev,
 	mcp->chip.owner = THIS_MODULE;
 
 	mcp->reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
+	if (IS_ERR(mcp->reset_gpio))
+		return dev_err_probe(dev, PTR_ERR(mcp->reset_gpio),
+				     "failed to get reset GPIO\n");
 
 	/* verify MCP_IOCON.SEQOP = 0, so sequential reads work,
 	 * and MCP_IOCON.HAEN = 1, so we work with all chips.
Signature · 1 line
2.43.0
#1 lore raw
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [High] pinctrl: mcp23s08: multi-device probe failure due to duplicate reset GPIO request --

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] pinctrl: mcp23s08: multi-device probe failure due to duplicate reset GPIO request --

commit 95196894afec909a218e9bd99fc8884c93869fc7
Author: bui duc phuc <[email protected]>

pinctrl: mcp23s08: Propagate reset GPIO errors

This commit adds a check for the return value of devm_gpiod_get_optional() in mcp23s08_probe_one() to propagate errors instead of ignoring them.

drivers/pinctrl/pinctrl-mcp23s08.c int mcp23s08_probe_one(struct mcp23s08 *mcp, struct device *dev, · line 618

Quoted 11 lines
drivers/pinctrl/pinctrl-mcp23s08.c +3 -0
index cc668f84bf8bc..259ac02134c64 100644
@@ -613,6 +613,9 @@ int mcp23s08_probe_one(struct mcp23s08 *mcp, struct device *dev,
 	mcp->chip.owner = THIS_MODULE;
 
 	mcp->reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
+	if (IS_ERR(mcp->reset_gpio))
+		return dev_err_probe(dev, PTR_ERR(mcp->reset_gpio),
+				     "failed to get reset GPIO\n");

[Severity: High]
Does this cause a regression for multi-chip SPI configurations by aborting the probe due to -EBUSY?

In the SPI driver, multiple physical chips can share a single SPI chip select. This is handled in drivers/pinctrl/pinctrl-mcp23s08_spi.c:mcp23s08_probe() by iterating over spi_present_mask and calling mcp23s08_probe_one() for each:

drivers/pinctrl/pinctrl-mcp23s08_spi.c:mcp23s08_probe() {
	...
	for_each_set_bit(addr, &spi_present_mask, MCP_MAX_DEV_PER_CS) {
		...
		ret = mcp23s08_probe_one(data->mcp[addr], dev, MCP23S08_SPI_BASE | (addr << 1),
					 info->type, -1);
		if (ret < 0)
			return ret;
	}
	...
}

Because all logical chips share the same SPI device, the same struct device *dev pointer is passed to every mcp23s08_probe_one() call.

During the first iteration, the reset GPIO is successfully requested exclusively.

During the second iteration, devm_gpiod_get_optional() is called again on the same dev pointer for the same "reset" connection ID. The GPIO core will attempt to request the same descriptor again, which will return -EBUSY because it is already requested and does not use GPIOD_FLAGS_BIT_NONEXCLUSIVE.

Prior to this patch, the -EBUSY error was ignored, allowing the second and subsequent chips to probe successfully (which was safe since the reset line was already asserted and deasserted by the first chip).

Quoted 3 lines
drivers/pinctrl/pinctrl-mcp23s08.c +0 -0
 	/* verify MCP_IOCON.SEQOP = 0, so sequential reads work,
 	 * and MCP_IOCON.HAEN = 1, so we work with all chips.
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#2 lore raw
Quoted 27 lines

[Severity: High]
Does this cause a regression for multi-chip SPI configurations by aborting the probe due to -EBUSY?

In the SPI driver, multiple physical chips can share a single SPI chip select. This is handled in drivers/pinctrl/pinctrl-mcp23s08_spi.c:mcp23s08_probe() by iterating over spi_present_mask and calling mcp23s08_probe_one() for each:

drivers/pinctrl/pinctrl-mcp23s08_spi.c:mcp23s08_probe() {
        ...
        for_each_set_bit(addr, &spi_present_mask, MCP_MAX_DEV_PER_CS) {
                ...
                ret = mcp23s08_probe_one(data->mcp[addr], dev, MCP23S08_SPI_BASE | (addr << 1),
                                         info->type, -1);
                if (ret < 0)
                        return ret;
        }
        ...
}

Because all logical chips share the same SPI device, the same struct device *dev pointer is passed to every mcp23s08_probe_one() call.

During the first iteration, the reset GPIO is successfully requested exclusively.

Quoted 13 lines

During the second iteration, devm_gpiod_get_optional() is called again on the same dev pointer for the same "reset" connection ID. The GPIO core will attempt to request the same descriptor again, which will return -EBUSY because it is already requested and does not use GPIOD_FLAGS_BIT_NONEXCLUSIVE.

Prior to this patch, the -EBUSY error was ignored, allowing the second and subsequent chips to probe successfully (which was safe since the reset line was already asserted and deasserted by the first chip).

Quoted 3 lines

/* verify MCP_IOCON.SEQOP = 0, so sequential reads work,
* and MCP_IOCON.HAEN = 1, so we work with all chips.

I looked at how the cs35l35 and aic3x drivers handle -EBUSY
when the reset GPIO is shared:
- cs35l35: https://elixir.bootlin.com/linux/v7.3-rc6/source/sound/soc/codecs/cs35l35.c#L1524
- aic3x: https://elixir.bootlin.com/linux/v7.3-rc6/source/sound/soc/codecs/tlv320aic3x.c#L1815

I think handling -EBUSY similarly to cs35l35 may be more appropriate here, preserving the existing behavior for shared reset GPIOs while propagating other errors.

Linus, could you confirm whether this approach would be OK for mcp23s08?

Best regards,
Phuc

maintainer
#3 lore raw
Quoted 2 lines · Bui Duc Phuc

Linus, could you confirm whether this approach would be OK for mcp23s08?

I don't understand Sashiko's fringe use-case rambling here, I think the patch is fine.

Yours,
Linus Walleij

#4 lore raw

Hi Linus,

Thanks for confirming.

Quoted 4 lines

I don't understand Sashiko's fringe use-case rambling here, I think the patch is fine.

I'll keep the patch as proposed.

Best regards,
Phuc