| Message ID | 20221215164140.821796-1-arnd@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | regulator: tps65219: fix Wextra warning | expand |
On Thu, Dec 15, 2022 at 05:41:28PM +0100, Arnd Bergmann wrote: > - tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); > - if (rdev < 0) { > + error = tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); > + if (error) { > dev_err(tps->dev, "Failed to get rdev for %s\n", > irq_type->regulator_name); > - return -EINVAL; > + return error; This will shut up the warning but is leaving the use of the uninitialised rdev (which I'm kind of disappointed the static checkers didn't pick up on). rdev needs to be passed by reference into the function, or set from the return value.
On Thu, Dec 15, 2022, at 17:48, Mark Brown wrote: > On Thu, Dec 15, 2022 at 05:41:28PM +0100, Arnd Bergmann wrote: > >> - tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); >> - if (rdev < 0) { >> + error = tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); >> + if (error) { >> dev_err(tps->dev, "Failed to get rdev for %s\n", >> irq_type->regulator_name); >> - return -EINVAL; >> + return error; > > This will shut up the warning but is leaving the use of the > uninitialised rdev (which I'm kind of disappointed the static checkers > didn't pick up on). rdev needs to be passed by reference into the > function, or set from the return value. Right, I didn't look far enough to see what the function is actually trying to do here, and that it completely fails to do that. I see that the bug was introduced between the first [1] and second []2] version of the driver, but don't see why. I'll leave it up to Jerome to address the problem, he's still in the middle of posting the rest of the series that has not yet been merged, so it makes sense for him to test it all together. Arnd [1] https://lore.kernel.org/lkml/20220719091742.3221-9-jneanne@baylibre.com/ [2] https://lore.kernel.org/lkml/20220726103355.17684-10-jneanne@baylibre.com/
On Fri, Dec 16, 2022, at 11:18, Arnd Bergmann wrote: > On Thu, Dec 15, 2022, at 17:48, Mark Brown wrote: >> On Thu, Dec 15, 2022 at 05:41:28PM +0100, Arnd Bergmann wrote: >> >>> - tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); >>> - if (rdev < 0) { >>> + error = tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); >>> + if (error) { >>> dev_err(tps->dev, "Failed to get rdev for %s\n", >>> irq_type->regulator_name); >>> - return -EINVAL; >>> + return error; >> >> This will shut up the warning but is leaving the use of the >> uninitialised rdev (which I'm kind of disappointed the static checkers >> didn't pick up on). rdev needs to be passed by reference into the >> function, or set from the return value. > > Right, I didn't look far enough to see what the function is > actually trying to do here, and that it completely fails to > do that. > > I see that the bug was introduced between the first [1] and > second []2] version of the driver, but don't see why. I'll > leave it up to Jerome to address the problem, he's still > in the middle of posting the rest of the series that has > not yet been merged, so it makes sense for him to test it > all together. > > Arnd > > [1] https://lore.kernel.org/lkml/20220719091742.3221-9-jneanne@baylibre.com/ > [2] https://lore.kernel.org/lkml/20220726103355.17684-10-jneanne@baylibre.com/ It looks like you merged another workaround from Randy Dunlap now as commit 2bbba115c3c9 ("regulator: tps65219: use IS_ERR() to detect an error pointer"), but I think that one is just as wrong as the one I submitted: the 'rdev' variable still remains uninitialized, and checking its value after it has already been used is not helpful. Arnd
On Thu, Jan 19, 2023 at 09:07:38AM +0100, Arnd Bergmann wrote: > It looks like you merged another workaround from Randy Dunlap now as > commit 2bbba115c3c9 ("regulator: tps65219: use IS_ERR() to detect an error > pointer"), but I think that one is just as wrong as the one I submitted: > the 'rdev' variable still remains uninitialized, and checking its value > after it has already been used is not helpful. Right, that's just changing the way the result is parsed, it's nothing to do with making sure things are initialised. It's just a coccinelle style thing.
diff --git a/drivers/regulator/tps65219-regulator.c b/drivers/regulator/tps65219-regulator.c index c484c943e467..fea3998334b1 100644 --- a/drivers/regulator/tps65219-regulator.c +++ b/drivers/regulator/tps65219-regulator.c @@ -366,11 +366,11 @@ static int tps65219_regulator_probe(struct platform_device *pdev) irq_data[i].dev = tps->dev; irq_data[i].type = irq_type; - tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); - if (rdev < 0) { + error = tps65219_get_rdev_by_name(irq_type->regulator_name, rdevtbl, rdev); + if (error) { dev_err(tps->dev, "Failed to get rdev for %s\n", irq_type->regulator_name); - return -EINVAL; + return error; } irq_data[i].rdev = rdev;