| Message ID | 20230911151548.672485-1-u.kleine-koenig@pengutronix.de (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | abaf59c470a7c9c59bda8da3517ec2ff0513f5a6 |
| Headers | show |
| Series | clk: qcom: cbf-msm8996: Convert to platform remove callback returning void | expand |
Quoting Uwe Kleine-König (2023-09-11 08:15:48) > The .remove() callback for a platform driver returns an int which makes > many driver authors wrongly assume it's possible to do error handling by > returning an error code. However the value returned is ignored (apart > from emitting a warning) and this typically results in resource leaks. > To improve here there is a quest to make the remove callback return > void. In the first step of this quest all drivers are converted to > .remove_new() which already returns void. Eventually after all drivers > are converted, .remove_new() is renamed to .remove(). > > qcom_msm8996_cbf_icc_remove() returned zero unconditionally. After > changing this function to return void instead, the driver can be > converted trivially to use .remove_new(). > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > --- Do you want to take this? Otherwise, I can apply it to fixes. > drivers/clk/qcom/clk-cbf-8996.c | 12 +++++------- > 1 file changed, 5 insertions(+), 7 deletions(-) > > diff --git a/drivers/clk/qcom/clk-cbf-8996.c b/drivers/clk/qcom/clk-cbf-8996.c > index 53f205a3f183..fe24b4abeab4 100644 > --- a/drivers/clk/qcom/clk-cbf-8996.c > +++ b/drivers/clk/qcom/clk-cbf-8996.c > @@ -250,13 +250,11 @@ static int qcom_msm8996_cbf_icc_register(struct platform_device *pdev, struct cl > return 0; > } > > -static int qcom_msm8996_cbf_icc_remove(struct platform_device *pdev) > +static void qcom_msm8996_cbf_icc_remove(struct platform_device *pdev) > { > struct icc_provider *provider = platform_get_drvdata(pdev); > > icc_clk_unregister(provider); > - > - return 0; > } > #define qcom_msm8996_cbf_icc_sync_state icc_sync_state > #else > @@ -266,7 +264,7 @@ static int qcom_msm8996_cbf_icc_register(struct platform_device *pdev, struct c > > return 0; > } > -#define qcom_msm8996_cbf_icc_remove(pdev) (0) > +#define qcom_msm8996_cbf_icc_remove(pdev) { } It would be better if this was a static inline function. > #define qcom_msm8996_cbf_icc_sync_state NULL > #endif >
Hello Stephen, On Mon, Sep 11, 2023 at 01:02:53PM -0700, Stephen Boyd wrote: > Quoting Uwe Kleine-König (2023-09-11 08:15:48) > > The .remove() callback for a platform driver returns an int which makes > > many driver authors wrongly assume it's possible to do error handling by > > returning an error code. However the value returned is ignored (apart > > from emitting a warning) and this typically results in resource leaks. > > To improve here there is a quest to make the remove callback return > > void. In the first step of this quest all drivers are converted to > > .remove_new() which already returns void. Eventually after all drivers > > are converted, .remove_new() is renamed to .remove(). > > > > qcom_msm8996_cbf_icc_remove() returned zero unconditionally. After > > changing this function to return void instead, the driver can be > > converted trivially to use .remove_new(). > > > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > --- > > Do you want to take this? Otherwise, I can apply it to fixes. if "you" == "Uwe Kleine-König": Please take it via your tree. There is still much to do before the next synchronous step, so there is no urge. If the patch goes in during the next merge window that's fine, too. > > @@ -266,7 +264,7 @@ static int qcom_msm8996_cbf_icc_register(struct platform_device *pdev, struct c > > > > return 0; > > } > > -#define qcom_msm8996_cbf_icc_remove(pdev) (0) > > +#define qcom_msm8996_cbf_icc_remove(pdev) { } > > It would be better if this was a static inline function. Ack, but this applies to the state before my patch, too, and I think this should be addressed separately. That's currently not in my focus, so if someone else wants to address this, you're welcome. Best regards Uwe
On Tue, Sep 12, 2023 at 08:53:43AM +0200, Uwe Kleine-König wrote: > Hello Stephen, > > On Mon, Sep 11, 2023 at 01:02:53PM -0700, Stephen Boyd wrote: > > Quoting Uwe Kleine-König (2023-09-11 08:15:48) > > > The .remove() callback for a platform driver returns an int which makes > > > many driver authors wrongly assume it's possible to do error handling by > > > returning an error code. However the value returned is ignored (apart > > > from emitting a warning) and this typically results in resource leaks. > > > To improve here there is a quest to make the remove callback return > > > void. In the first step of this quest all drivers are converted to > > > .remove_new() which already returns void. Eventually after all drivers > > > are converted, .remove_new() is renamed to .remove(). > > > > > > qcom_msm8996_cbf_icc_remove() returned zero unconditionally. After > > > changing this function to return void instead, the driver can be > > > converted trivially to use .remove_new(). > > > > > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > > --- > > > > Do you want to take this? Otherwise, I can apply it to fixes. > > if "you" == "Uwe Kleine-König": Please take it via your tree. There is > still much to do before the next synchronous step, so there is no urge. > If the patch goes in during the next merge window that's fine, too. > @Stephen, should I just pick this in the Qcom tree for 6.7 then? Regards, Bjorn
Quoting Bjorn Andersson (2023-09-20 10:29:58) > On Tue, Sep 12, 2023 at 08:53:43AM +0200, Uwe Kleine-König wrote: > > Hello Stephen, > > > > On Mon, Sep 11, 2023 at 01:02:53PM -0700, Stephen Boyd wrote: > > > Quoting Uwe Kleine-König (2023-09-11 08:15:48) > > > > The .remove() callback for a platform driver returns an int which makes > > > > many driver authors wrongly assume it's possible to do error handling by > > > > returning an error code. However the value returned is ignored (apart > > > > from emitting a warning) and this typically results in resource leaks. > > > > To improve here there is a quest to make the remove callback return > > > > void. In the first step of this quest all drivers are converted to > > > > .remove_new() which already returns void. Eventually after all drivers > > > > are converted, .remove_new() is renamed to .remove(). > > > > > > > > qcom_msm8996_cbf_icc_remove() returned zero unconditionally. After > > > > changing this function to return void instead, the driver can be > > > > converted trivially to use .remove_new(). > > > > > > > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > > > --- > > > > > > Do you want to take this? Otherwise, I can apply it to fixes. > > > > if "you" == "Uwe Kleine-König": Please take it via your tree. There is > > still much to do before the next synchronous step, so there is no urge. > > If the patch goes in during the next merge window that's fine, too. > > > > @Stephen, should I just pick this in the Qcom tree for 6.7 then? > Yes please.
Hello, On Mon, Oct 09, 2023 at 08:49:27PM -0700, Stephen Boyd wrote: > Quoting Bjorn Andersson (2023-09-20 10:29:58) > > On Tue, Sep 12, 2023 at 08:53:43AM +0200, Uwe Kleine-König wrote: > > > On Mon, Sep 11, 2023 at 01:02:53PM -0700, Stephen Boyd wrote: > > > > Quoting Uwe Kleine-König (2023-09-11 08:15:48) > > > > > The .remove() callback for a platform driver returns an int which makes > > > > > many driver authors wrongly assume it's possible to do error handling by > > > > > returning an error code. However the value returned is ignored (apart > > > > > from emitting a warning) and this typically results in resource leaks. > > > > > To improve here there is a quest to make the remove callback return > > > > > void. In the first step of this quest all drivers are converted to > > > > > .remove_new() which already returns void. Eventually after all drivers > > > > > are converted, .remove_new() is renamed to .remove(). > > > > > > > > > > qcom_msm8996_cbf_icc_remove() returned zero unconditionally. After > > > > > changing this function to return void instead, the driver can be > > > > > converted trivially to use .remove_new(). > > > > > > > > > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > > > > --- > > > > > > > > Do you want to take this? Otherwise, I can apply it to fixes. > > > > > > if "you" == "Uwe Kleine-König": Please take it via your tree. There is > > > still much to do before the next synchronous step, so there is no urge. > > > If the patch goes in during the next merge window that's fine, too. > > > > > > > @Stephen, should I just pick this in the Qcom tree for 6.7 then? > > Yes please. This patch isn't in next yet. Is it still scheduled to go in for 6.7-rc1? Best regards Uwe
On Mon, 11 Sep 2023 17:15:48 +0200, Uwe Kleine-König wrote: > The .remove() callback for a platform driver returns an int which makes > many driver authors wrongly assume it's possible to do error handling by > returning an error code. However the value returned is ignored (apart > from emitting a warning) and this typically results in resource leaks. > To improve here there is a quest to make the remove callback return > void. In the first step of this quest all drivers are converted to > .remove_new() which already returns void. Eventually after all drivers > are converted, .remove_new() is renamed to .remove(). > > [...] Applied, thanks! [1/1] clk: qcom: cbf-msm8996: Convert to platform remove callback returning void commit: abaf59c470a7c9c59bda8da3517ec2ff0513f5a6 Best regards,
diff --git a/drivers/clk/qcom/clk-cbf-8996.c b/drivers/clk/qcom/clk-cbf-8996.c index 53f205a3f183..fe24b4abeab4 100644 --- a/drivers/clk/qcom/clk-cbf-8996.c +++ b/drivers/clk/qcom/clk-cbf-8996.c @@ -250,13 +250,11 @@ static int qcom_msm8996_cbf_icc_register(struct platform_device *pdev, struct cl return 0; } -static int qcom_msm8996_cbf_icc_remove(struct platform_device *pdev) +static void qcom_msm8996_cbf_icc_remove(struct platform_device *pdev) { struct icc_provider *provider = platform_get_drvdata(pdev); icc_clk_unregister(provider); - - return 0; } #define qcom_msm8996_cbf_icc_sync_state icc_sync_state #else @@ -266,7 +264,7 @@ static int qcom_msm8996_cbf_icc_register(struct platform_device *pdev, struct c return 0; } -#define qcom_msm8996_cbf_icc_remove(pdev) (0) +#define qcom_msm8996_cbf_icc_remove(pdev) { } #define qcom_msm8996_cbf_icc_sync_state NULL #endif @@ -340,9 +338,9 @@ static int qcom_msm8996_cbf_probe(struct platform_device *pdev) return qcom_msm8996_cbf_icc_register(pdev, &cbf_mux.clkr.hw); } -static int qcom_msm8996_cbf_remove(struct platform_device *pdev) +static void qcom_msm8996_cbf_remove(struct platform_device *pdev) { - return qcom_msm8996_cbf_icc_remove(pdev); + qcom_msm8996_cbf_icc_remove(pdev); } static const struct of_device_id qcom_msm8996_cbf_match_table[] = { @@ -354,7 +352,7 @@ MODULE_DEVICE_TABLE(of, qcom_msm8996_cbf_match_table); static struct platform_driver qcom_msm8996_cbf_driver = { .probe = qcom_msm8996_cbf_probe, - .remove = qcom_msm8996_cbf_remove, + .remove_new = qcom_msm8996_cbf_remove, .driver = { .name = "qcom-msm8996-cbf", .of_match_table = qcom_msm8996_cbf_match_table,
The .remove() callback for a platform driver returns an int which makes many driver authors wrongly assume it's possible to do error handling by returning an error code. However the value returned is ignored (apart from emitting a warning) and this typically results in resource leaks. To improve here there is a quest to make the remove callback return void. In the first step of this quest all drivers are converted to .remove_new() which already returns void. Eventually after all drivers are converted, .remove_new() is renamed to .remove(). qcom_msm8996_cbf_icc_remove() returned zero unconditionally. After changing this function to return void instead, the driver can be converted trivially to use .remove_new(). Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> --- drivers/clk/qcom/clk-cbf-8996.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) base-commit: 0bb80ecc33a8fb5a682236443c1e740d5c917d1d