| Message ID | 1386421734-10240-2-git-send-email-alexandre.belloni@free-electrons.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 07/12/2013 14:08, Alexandre Belloni : > When passing a not initialized config parameter, at91_pinconf_get() would return > a bogus value. Fix that by initializing it to zero before using it. > > Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com> > --- > drivers/pinctrl/pinctrl-at91.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/pinctrl/pinctrl-at91.c b/drivers/pinctrl/pinctrl-at91.c > index 6446dc804aa7..b0b78f3468ae 100644 > --- a/drivers/pinctrl/pinctrl-at91.c > +++ b/drivers/pinctrl/pinctrl-at91.c > @@ -722,7 +722,8 @@ static int at91_pinconf_get(struct pinctrl_dev *pctldev, > unsigned pin; > int div; > > - dev_dbg(info->dev, "%s:%d, pin_id=%d, config=0x%lx", __func__, __LINE__, pin_id, *config); > + *config = 0; > + dev_dbg(info->dev, "%s:%d, pin_id=%d", __func__, __LINE__, pin_id); > pio = pin_to_controller(info, pin_to_bank(pin_id)); > pin = pin_id % MAX_NB_GPIO_PER_BANK; Beyond this patch, I must say that I am puzzled by this function. What I read from the prototype documentation and what I see in different implementations is different... Linus, can we have a review of this function because it seems not in line with what is used for u300 (but on the other hand looks like the what is returned by pinctrl-exynos5440.c driver for example). What would be the consequences if we change this function's behavior: I mean use of -EINVAL for pin configuration "available but disabled" as said in include/linux/pinctrl/pinconf.h? Bye,
Hi, On 09/12/2013 09:24, Nicolas Ferre wrote: > On 07/12/2013 14:08, Alexandre Belloni : >> When passing a not initialized config parameter, at91_pinconf_get() >> would return >> a bogus value. Fix that by initializing it to zero before using it. >> >> Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com> >> --- >> drivers/pinctrl/pinctrl-at91.c | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/pinctrl/pinctrl-at91.c >> b/drivers/pinctrl/pinctrl-at91.c >> index 6446dc804aa7..b0b78f3468ae 100644 >> --- a/drivers/pinctrl/pinctrl-at91.c >> +++ b/drivers/pinctrl/pinctrl-at91.c >> @@ -722,7 +722,8 @@ static int at91_pinconf_get(struct pinctrl_dev >> *pctldev, >> unsigned pin; >> int div; >> >> - dev_dbg(info->dev, "%s:%d, pin_id=%d, config=0x%lx", __func__, >> __LINE__, pin_id, *config); >> + *config = 0; >> + dev_dbg(info->dev, "%s:%d, pin_id=%d", __func__, __LINE__, pin_id); >> pio = pin_to_controller(info, pin_to_bank(pin_id)); >> pin = pin_id % MAX_NB_GPIO_PER_BANK; > > Beyond this patch, I must say that I am puzzled by this function. > > What I read from the prototype documentation and what I see in > different implementations is different... > > Linus, can we have a review of this function because it seems not in > line with what is used for u300 (but on the other hand looks like the > what is returned by pinctrl-exynos5440.c driver for example). > > What would be the consequences if we change this function's behavior: > I mean use of -EINVAL for pin configuration "available but disabled" > as said in include/linux/pinctrl/pinconf.h? > From what I understand, it doesn't really matter because that function is never used. I guess the best implementation is the tegra one ;) It is only called in one specific case: - you have ops->is_generic == true (that is only true for a few implmentations) - and you are dumping the pinconf state using debugfs I'm actually wondering if the checks for the ops->pin_config_get are not a bit overkill. We check for that function in: - pinconf_check_ops() - before calling it in pin_config_get_for_pin() which is only used once, in the same path : dump using debugfs and having ops->is_generic == true - in pinconf_pins_show() which is the function calling also in the same path What I would do is: - remove all the checks in pinconf_check_ops() and pinconf_pins_show() so that people are not pressured to implement a function that is simply never used. - modify pin_config_get_for_pin() by removing the dev_err() call and returning -ENOTSUPP instead of -EINVAL (it doesn't change the behaviour but I feel -ENOTSUPP is more appropriate) I have a patch ready but I can't test it as I don't own any of the is_generic platforms.
On Mon, Dec 9, 2013 at 9:24 AM, Nicolas Ferre <nicolas.ferre@atmel.com> wrote: >> - dev_dbg(info->dev, "%s:%d, pin_id=%d, config=0x%lx", __func__, >> __LINE__, pin_id, *config); >> + *config = 0; >> + dev_dbg(info->dev, "%s:%d, pin_id=%d", __func__, __LINE__, >> pin_id); >> pio = pin_to_controller(info, pin_to_bank(pin_id)); >> pin = pin_id % MAX_NB_GPIO_PER_BANK; > > > Beyond this patch, I must say that I am puzzled by this function. > > What I read from the prototype documentation and what I see in different > implementations is different... Yeah, we need to fix this mess. > Linus, can we have a review of this function because it seems not in line > with what is used for u300 (but on the other hand looks like the what is > returned by pinctrl-exynos5440.c driver for example). It is supposed to read out one config at the time, if and only if used with the generic pin config. Typically: enum pin_config_param param = (enum pin_config_param) *config; switch (param) { case PIN_CONFIG_BIAS_HIGH_IMPEDANCE: *config = 0; if (biasmode) return 0; else return -EINVAL; break; (...) return -ENOTSUPP; However AT91 is *not* using generic pin config, so the semantics of this call is driver-dependent. In your case the implementation get all the configs at once, which is an efficient shortcut if you don't need to be general and enumerat all possible configs. > What would be the consequences if we change this function's behavior: I mean > use of -EINVAL for pin configuration "available but disabled" as said in > include/linux/pinctrl/pinconf.h? That code will propagate back ... I guess you'd have to test it :-/ Yours, Linus Walleij
On Mon, Dec 9, 2013 at 10:55 AM, Alexandre Belloni <alexandre.belloni@free-electrons.com> wrote: > I'm actually wondering if the checks for the ops->pin_config_get are not > a bit overkill. We check for that function in: > - pinconf_check_ops() > - before calling it in pin_config_get_for_pin() which is only used > once, in the same path : dump using debugfs and having ops->is_generic > == true > - in pinconf_pins_show() which is the function calling also in the same > path > > What I would do is: > - remove all the checks in pinconf_check_ops() and pinconf_pins_show() > so that people are not pressured to implement a function that is simply > never used. > - modify pin_config_get_for_pin() by removing the dev_err() call and > returning -ENOTSUPP instead of -EINVAL (it doesn't change the behaviour > but I feel -ENOTSUPP is more appropriate) > > I have a patch ready but I can't test it as I don't own any of the > is_generic platforms. Mail it out with a [CFT: ] "call for testing" prefix. Yours, Linus Walleij
On Sat, Dec 7, 2013 at 2:08 PM, Alexandre Belloni <alexandre.belloni@free-electrons.com> wrote: > When passing a not initialized config parameter, at91_pinconf_get() would return > a bogus value. Fix that by initializing it to zero before using it. > > Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com> Patch applied. Yours, Linus Walleij
On 12/12/2013 15:40, Linus Walleij wrote: > Mail it out with a [CFT: ] "call for testing" prefix. > I actually already sent it but without the CFT. Do you want me to resend ? Regards,
diff --git a/drivers/pinctrl/pinctrl-at91.c b/drivers/pinctrl/pinctrl-at91.c index 6446dc804aa7..b0b78f3468ae 100644 --- a/drivers/pinctrl/pinctrl-at91.c +++ b/drivers/pinctrl/pinctrl-at91.c @@ -722,7 +722,8 @@ static int at91_pinconf_get(struct pinctrl_dev *pctldev, unsigned pin; int div; - dev_dbg(info->dev, "%s:%d, pin_id=%d, config=0x%lx", __func__, __LINE__, pin_id, *config); + *config = 0; + dev_dbg(info->dev, "%s:%d, pin_id=%d", __func__, __LINE__, pin_id); pio = pin_to_controller(info, pin_to_bank(pin_id)); pin = pin_id % MAX_NB_GPIO_PER_BANK;
When passing a not initialized config parameter, at91_pinconf_get() would return a bogus value. Fix that by initializing it to zero before using it. Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com> --- drivers/pinctrl/pinctrl-at91.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)