On Thu Dec 4, 2025 at 8:39 PM CET, Mark Brown wrote:
s/enable GPIO/enabled GPIOs/ ? (in Subject)
In order to simplify ownership rules for enable GPIOs supplied by drivers
regulator_register() always takes ownership of them, even if it ends up
failing for some other reason. We therefore should not free the GPIO if
registration fails but just let the core worry about things.
I built a 6.18 kernel with this patch added and that fixed the problem.
Thanks :-) Feel free to add:
Tested-by: Diederik de Haas <redacted>
Cheers,
Diederik
quoted hunk
Fixes: 636f4618b1cd (regulator: fixed: fix GPIO descriptor leak on register failure)
Reported-by: Diederik de Haas <redacted>
Closes: https://lore.kernel.org/r/DEPEYUF5BRGY.UKFBWRRE8HNP@cknow-tech.com (local)
Signed-off-by: Mark Brown <broonie@kernel.org>
---
drivers/regulator/fixed.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)
diff --git a/drivers/regulator/fixed.c b/drivers/regulator/fixed.c
index a2d16e9abfb5..254c0a8a4555 100644
--- a/drivers/regulator/fixed.c
+++ b/drivers/regulator/fixed.c
@@ -330,13 +330,10 @@ static int reg_fixed_voltage_probe(struct platform_device *pdev)
drvdata->dev = devm_regulator_register(&pdev->dev, &drvdata->desc,
&cfg);
- if (IS_ERR(drvdata->dev)) {
- ret = dev_err_probe(&pdev->dev, PTR_ERR(drvdata->dev),
- "Failed to register regulator: %ld\n",
- PTR_ERR(drvdata->dev));
- gpiod_put(cfg.ena_gpiod);
- return ret;
- }
+ if (IS_ERR(drvdata->dev))
+ return dev_err_probe(&pdev->dev, PTR_ERR(drvdata->dev),
+ "Failed to register regulator: %ld\n",
+ PTR_ERR(drvdata->dev));
platform_set_drvdata(pdev, drvdata);
---
base-commit: 81d431130ae1af4e64030f6a956ee9137e6fc1b0
change-id: 20251204-regulator-fixed-fix-gpiod-leak-b1f50fb8c388
Best regards,
--
Mark Brown [off-list ref]