| Message ID | 20210427031342.5333-1-rdunlap@infradead.org (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
| Series | clk: <linux/clk.h>: correct clk_get_parent() documentation | expand |
On Mon, Apr 26, 2021 at 08:13:42PM -0700, Randy Dunlap wrote: > Make the kernel-doc return value agree with both the stub implementation > in <linux/clk.h> and the non-stub function in drivers/clk/clk.c. > > Signed-off-by: Randy Dunlap <rdunlap@infradead.org> > Cc: Russell King <linux@armlinux.org.uk> > Cc: linux-clk@vger.kernel.org > --- > There are several drivers that call clk_get_parent() and use its > return value as though it returns an ERR_PTR(). I am working on a > patch series to fix those. The whole premise of the CLK API is that: - clk values that IS_ERR() returns true are errors - clk values that IS_ERR() returns false are valid and that is that - nice, simple, easy to understand. So the documentation is correct. If clk_get_parent() needs to return an error, it does so by returning an error-pointer, not by returning NULL.
On 4/27/21 2:38 AM, Russell King - ARM Linux admin wrote: > On Mon, Apr 26, 2021 at 08:13:42PM -0700, Randy Dunlap wrote: >> Make the kernel-doc return value agree with both the stub implementation >> in <linux/clk.h> and the non-stub function in drivers/clk/clk.c. >> >> Signed-off-by: Randy Dunlap <rdunlap@infradead.org> >> Cc: Russell King <linux@armlinux.org.uk> >> Cc: linux-clk@vger.kernel.org >> --- >> There are several drivers that call clk_get_parent() and use its >> return value as though it returns an ERR_PTR(). I am working on a >> patch series to fix those. > > The whole premise of the CLK API is that: > > - clk values that IS_ERR() returns true are errors > - clk values that IS_ERR() returns false are valid > > and that is that - nice, simple, easy to understand. So the > documentation is correct. If clk_get_parent() needs to return an > error, it does so by returning an error-pointer, not by returning > NULL. Hm, so the 6 drivers that do this are OK then, if clk_get_parent() returns NULL? parent = clk_get_parent(clk); if (IS_ERR(parent)) { dev_err(cpu_dev, "Cannot get parent clock for CPU0\n"); clk_put(clk); return PTR_ERR(parent); } /* Get parent CPU frequency */ base_frequency = clk_get_rate(parent); boom?
On Tue, Apr 27, 2021 at 10:07:04AM -0700, Randy Dunlap wrote: > On 4/27/21 2:38 AM, Russell King - ARM Linux admin wrote: > > On Mon, Apr 26, 2021 at 08:13:42PM -0700, Randy Dunlap wrote: > >> Make the kernel-doc return value agree with both the stub implementation > >> in <linux/clk.h> and the non-stub function in drivers/clk/clk.c. > >> > >> Signed-off-by: Randy Dunlap <rdunlap@infradead.org> > >> Cc: Russell King <linux@armlinux.org.uk> > >> Cc: linux-clk@vger.kernel.org > >> --- > >> There are several drivers that call clk_get_parent() and use its > >> return value as though it returns an ERR_PTR(). I am working on a > >> patch series to fix those. > > > > The whole premise of the CLK API is that: > > > > - clk values that IS_ERR() returns true are errors > > - clk values that IS_ERR() returns false are valid > > > > and that is that - nice, simple, easy to understand. So the > > documentation is correct. If clk_get_parent() needs to return an > > error, it does so by returning an error-pointer, not by returning > > NULL. > > Hm, so the 6 drivers that do this are OK then, if > clk_get_parent() returns NULL? > > > parent = clk_get_parent(clk); > if (IS_ERR(parent)) { > dev_err(cpu_dev, "Cannot get parent clock for CPU0\n"); > clk_put(clk); > return PTR_ERR(parent); > } > > /* Get parent CPU frequency */ > base_frequency = clk_get_rate(parent); > > boom? As I say, the premise is that only clocks where IS_ERR(clk) is true are an error. Everything else must be treated as an acceptable clock. Sometimes the NULL clock is used for "I don't have a clock", and that must not cause the API to go "boom". In the case of the CCF implementation, clk_get_rate(NULL) will return 0.
On 4/27/21 11:03 AM, Russell King - ARM Linux admin wrote: > On Tue, Apr 27, 2021 at 10:07:04AM -0700, Randy Dunlap wrote: >> On 4/27/21 2:38 AM, Russell King - ARM Linux admin wrote: >>> On Mon, Apr 26, 2021 at 08:13:42PM -0700, Randy Dunlap wrote: >>>> Make the kernel-doc return value agree with both the stub implementation >>>> in <linux/clk.h> and the non-stub function in drivers/clk/clk.c. >>>> >>>> Signed-off-by: Randy Dunlap <rdunlap@infradead.org> >>>> Cc: Russell King <linux@armlinux.org.uk> >>>> Cc: linux-clk@vger.kernel.org >>>> --- >>>> There are several drivers that call clk_get_parent() and use its >>>> return value as though it returns an ERR_PTR(). I am working on a >>>> patch series to fix those. >>> >>> The whole premise of the CLK API is that: >>> >>> - clk values that IS_ERR() returns true are errors >>> - clk values that IS_ERR() returns false are valid >>> >>> and that is that - nice, simple, easy to understand. So the >>> documentation is correct. If clk_get_parent() needs to return an >>> error, it does so by returning an error-pointer, not by returning >>> NULL. >> >> Hm, so the 6 drivers that do this are OK then, if >> clk_get_parent() returns NULL? >> >> >> parent = clk_get_parent(clk); >> if (IS_ERR(parent)) { >> dev_err(cpu_dev, "Cannot get parent clock for CPU0\n"); >> clk_put(clk); >> return PTR_ERR(parent); >> } >> >> /* Get parent CPU frequency */ >> base_frequency = clk_get_rate(parent); >> >> boom? > > As I say, the premise is that only clocks where IS_ERR(clk) is true > are an error. Everything else must be treated as an acceptable clock. > > Sometimes the NULL clock is used for "I don't have a clock", and that > must not cause the API to go "boom". > > In the case of the CCF implementation, clk_get_rate(NULL) will return > 0. > Thank you for the clarification. That is helpful.
--- linux-next-20210426.orig/include/linux/clk.h +++ linux-next-20210426/include/linux/clk.h @@ -744,8 +744,8 @@ int clk_set_parent(struct clk *clk, stru * clk_get_parent - get the parent clock source for this clock * @clk: clock source * - * Returns struct clk corresponding to parent clock source, or - * valid IS_ERR() condition containing errno. + * Return: &struct clk corresponding to parent clock source, or + * %NULL if clk is %NULL. */ struct clk *clk_get_parent(struct clk *clk);
Make the kernel-doc return value agree with both the stub implementation in <linux/clk.h> and the non-stub function in drivers/clk/clk.c. Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Cc: Russell King <linux@armlinux.org.uk> Cc: linux-clk@vger.kernel.org --- There are several drivers that call clk_get_parent() and use its return value as though it returns an ERR_PTR(). I am working on a patch series to fix those. include/linux/clk.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)