Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

InterruptIn changes in NCS36510 HAL. #2938

Merged
merged 6 commits into from
Oct 25, 2016

Conversation

radhika-raghavendran
Copy link
Contributor

Notes:

  • Pull requests will not be accepted until the submitter has agreed to the contributer agreement.
  • This is just a template, so feel free to use/remove the unnecessary things

Description

A few sentences describing the overall goals of the pull request's commits.

Status

READY/IN DEVELOPMENT/HOLD

Migrations

If this PR changes any APIs or behaviors, give a short description of what API users should do when this PR is merged.

YES | NO

Related PRs

List related PRs against other branches:

branch PR
other_pr_production link
other_pr_master link

Todos

  • Tests
  • Documentation

Deploy notes

Notes regarding the deployment of this PR. These should note any
required changes in the build environment, tools, compilers, etc.

Steps to test or reproduce

Outline the steps to test or reproduce the PR here.

@0xc0170
Copy link
Contributor

0xc0170 commented Oct 7, 2016

Can you share the test results - how was this tested?

@@ -143,16 +140,17 @@ void gpio_mode(gpio_t *obj, PinMode mode)
void gpio_dir(gpio_t *obj, PinDirection direction)
{
/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@@ -186,12 +186,12 @@ int gpio_read(gpio_t *obj)
int ret;

/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

GpioReg_pt gpioBase;

/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@@ -146,7 +146,10 @@ int gpio_irq_init(gpio_irq_t *obj, PinName pin, gpio_irq_handler handler, uint32
gpioIds[pin] = id;

/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@@ -179,7 +183,10 @@ int gpio_irq_init(gpio_irq_t *obj, PinName pin, gpio_irq_handler handler, uint32
void gpio_irq_free(gpio_irq_t *obj)
{
/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@@ -239,9 +246,12 @@ void gpio_irq_set(gpio_irq_t *obj, gpio_irq_event event, uint32_t enable)
void gpio_irq_enable(gpio_irq_t *obj)
{
/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

/* Enable the GPIO clock */
CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.


default:
break;
case PushPullPullDown:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

obj->GPIOMEMBASE->IRQ_POLARITY_CLEAR = (IO_ALL ^ (obj->pinMask));
}
break;
case IRQ_RISE:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I did run AStyle on this code before submitting a PR. The result is that AStyle shows changes in files but they don't seem to show up in the PR. Is it recommended to format files manually or is running AStyle the preferred method? Please let us know for future PRs.

@sg- sg- added the needs: work label Oct 7, 2016
Copy link
Contributor

@kjbracey kjbracey left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What about the disables of GPIO_CLOCK in other modules like I2C and serial?

@@ -83,6 +83,7 @@
#define CLOCK_ENABLE(a) CLOCKREG->PDIS.WORD &= ~(1 << a)
#define CLOCK_DISABLE(a) CLOCKREG->PDIS.WORD |= (uint32_t)(1 << a)

#define CLOCK_IS_ENABLED(a) ((CLOCKREG->PDIS.WORD >> a) & 1)?0:1
Copy link
Contributor

@kjbracey kjbracey Oct 11, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is dangerous due to lack of enclosing brackets. Indeed, I believe all your uses of it actually fail.

Actually, no, they do work, but rather by accident - the first expression gets boolean inverted before being fed to the ?, so the right answer comes out.

Does it actually matter if you "enable" a clock that is already enabled anyway? Just a write to the peripheral may be as fast as the read to check if it is enabled.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@0xc0170 suggested in a previous pull request that this implementation be changed from calling an enable function repeatedly. Instead check for the clock and enable only if required.
I have added enclosing brackets.

obj->GPIOMEMBASE->IRQ_POLARITY_SET = (obj->pinMask);
obj->GPIOMEMBASE->ANYEDGE_SET = IO_NONE;

obj->GPIOMEMBASE->W_IN |= obj->pinMask;
Copy link
Contributor

@kjbracey kjbracey Oct 11, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Assuming the datasheet I have is correct this is wrong, as is the comment above - you should never be performing read-modify-write on any of these registers.

obj->GPIOMEMBASE->IRQ_EDGE |= obj->pinMask;
obj->GPIOMEMBASE->IRQ_POLARITY_SET |= obj->pinMask;
obj->GPIOMEMBASE->IRQ_ENABLE_SET |= obj->pinMask;
obj->GPIOMEMBASE->ANYEDGE_SET |= IO_NONE;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This presumably should be ANYEDGE_CLEAR = obj->pinMask;

obj->GPIOMEMBASE->W_IN |= obj->pinMask;
obj->GPIOMEMBASE->IRQ_EDGE |= obj->pinMask;
obj->GPIOMEMBASE->IRQ_POLARITY_SET |= obj->pinMask;
obj->GPIOMEMBASE->IRQ_ENABLE_SET |= obj->pinMask;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should interrupts really be enabled initially, before irq_set or irq_enable is called? I'm not sure.

}
break;
case IRQ_RISE:
obj->GPIOMEMBASE->IRQ_EDGE |= obj->pinMask;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Again, wrong by the data sheet.

obj->GPIOMEMBASE->IRQ_POLARITY_SET |= obj->pinMask;
} else if (enable == 0) {
/* Disable rising edge */
obj->GPIOMEMBASE->IRQ_POLARITY_SET |= (IO_ALL ^ (obj->pinMask));
Copy link
Contributor

@kjbracey kjbracey Oct 11, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

enable 0 doesn't mean "disable rising edge" - it's a request to disable the interrupts.

I'd suggest two separate tests - one to program IRQ_POLARITY_SET or CLEAR depending on event, and a separate one to program IRQ_ENABLE_SET or IRQ_ENABLE_CLEAR depending on enable.

And again, all of those just being = obj->pinMask, if my register map is correct. (And it seems logical to me).

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have taken care of all the comments above. Would I need to rebase again, since the structure of mbed-os has changed since this PR was generated?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have rebased my fork on the latest mbed-os structure. Let me know if I need to do something more on my side.

Copy link
Contributor

@kjbracey kjbracey left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems better. Does it work for 2 buttons now?

I'm still worried by all the CLOCK_DISABLE(CLOCK_GPIO) in other drivers. Would it be a good idea to remove those?

@@ -82,6 +82,7 @@

#define CLOCK_ENABLE(a) CLOCKREG->PDIS.WORD &= ~(1 << a)
#define CLOCK_DISABLE(a) CLOCKREG->PDIS.WORD |= (uint32_t)(1 << a)
#define CLOCK_IS_ENABLED(a) (((CLOCKREG->PDIS.WORD) >> a) & 1)?0:1
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The brackets need to be around the entire expression. Otherwise something like if (!CLOCK_IS_ENABLED(foo) && something_else) would go horribly wrong.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Taken care and modified.


obj->GPIOMEMBASE->W_IN = (IO_ALL ^ (obj->pinMask));
/* Make the pin as output in order to release it */
obj->GPIOMEMBASE->W_OUT = obj->pinMask;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you really want to make the pin an output? Start driving your input line high or low? Surely not.

Don't you just want to disable the interrupt, but leave the pin as it is?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Taken care and modified.

if (enable == 1) {
/* Enable rising edge */
obj->GPIOMEMBASE->IRQ_POLARITY_SET = (obj->pinMask);

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This if is the same, and so no longer needs to be in the switch.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Rebase and conflict resolution went horribly wrong and original files were chosen during an auto merge process. Hence none of the recent changes in my fork came out during the rebase. This piece of code has been modified in this checkin.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

My comment still applies - you're repeating the same enable code in each case of the switch - the enable code does not depend on the mode, and it doesn't need to be in the switch.

@@ -157,10 +161,8 @@ int gpio_irq_init(gpio_irq_t *obj, PinName pin, gpio_irq_handler handler, uint32
* then change this setting to obj->GPIOMEMBASE->W_IN |= obj->pinMask. All parameter setting needs to change from = to |=
*/
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment is obsolete and wrong, so should be removed. (Although have you actually tested it with two inputs now? That's what we were trying to get working - the two buttons.)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have tested with two buttons as well as the same button triggering two inputs. Both scenarios are working.

CLOCK_ENABLE(CLOCK_GPIO);
if(!CLOCK_IS_ENABLED(CLOCK_GPIO)) {
CLOCK_ENABLE(CLOCK_GPIO);
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd like to ask @0xc0170 whether he really thinks this is worthwhile. Reading the hardware to find out if it's worth writing to the hardware? More code, and probably no faster - usually 1 read instead of 1 write.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The GPIO disable in other drivers was also part of a rebase and automerge gone horribly wrong. I hope it is reflected in my latest checkin and pull request. Please let me know.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd like to ask @0xc0170 whether he really thinks this is worthwhile

It's not, good point !

@bridadan
Copy link
Contributor

/morph test

@mbed-bot
Copy link

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 0

All builds and test passed!

@radhika-raghavendran
Copy link
Contributor Author

Is this merged with the main branch? The label status has not changed over the last couple of days. All the tests have passed in the checklist.

@0xc0170
Copy link
Contributor

0xc0170 commented Oct 18, 2016

Is this merged with the main branch? The label status has not changed over the last couple of days. All the tests have passed in the checklist.

What do you think about checking if gpio is enabled vs just enable the gpio clocks? The question asked above.

It would be useful to provide more details in the commit messages

@radhika-raghavendran
Copy link
Contributor Author

I have reverted GPIO clock enablement code to a previous implementation which was present in older version of the code.
I hope this pull request can be merged with main branch now.

Copy link
Contributor

@kjbracey kjbracey left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Okay, but still a couple of minor concerns.

@@ -178,10 +169,11 @@ int gpio_irq_init(gpio_irq_t *obj, PinName pin, gpio_irq_handler handler, uint32
*/
void gpio_irq_free(gpio_irq_t *obj)
{
/* Enable the GPIO clock */
/* Enable the GPIO clock which may have been switched off by other drivers */
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Really you need to ensure no other driver does turn it off - assuming the clock needs to stay running for the interrupts to work, not just for the register access? (If the clock only needed to be on for the register access, you could disable it at the end of the function).

if (enable == 1) {
/* Enable rising edge */
obj->GPIOMEMBASE->IRQ_POLARITY_SET = (obj->pinMask);

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

My comment still applies - you're repeating the same enable code in each case of the switch - the enable code does not depend on the mode, and it doesn't need to be in the switch.

@radhika-raghavendran
Copy link
Contributor Author

IRQ enable code moved out of switch case. Regarding clock disable of GPIO from other drivers, other code may need modification and testing. Hence CLOCK_ENABLE(CLOCK_GPIO) remains in code for now.

break;

default:
/* No event is set */
break;
}
/* Enable the IRQ based on enable parameter */
if (enable == 1) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I personally would simplify further to just if (enable) {} else {}, on the assumption that it is supposed to be a boolean, even if not declared as that. But this is fine.

@radhika-raghavendran
Copy link
Contributor Author

The IRQ API is void gpio_irq_set(gpio_irq_t *obj, gpio_irq_event event, uint32_t enable)
enable is a 32 bit parameter, where a simple Boolean would have been sufficient. Hence checking against 1 or 0.

@kjbracey
Copy link
Contributor

Hence checking against 1 or 0.

Hard to say whether that's better or worse, as it's not actually documented anywhere (as far as I can see) that the "enable" value is specifically 1.

Generally speaking writing if (x == true) is bad C style, as it looks like it's the same as if (x) or the opposite of if (!x) or of if (x == false), but it isn't.

Specifically writing it like this means you are expecting future expansion, so deliberately want to do nothing if 2 is passed. I personally think if passed 2, it's probably better to enable.

Quick survey of 4 other drivers showed the 3 doing if (enable) else, and one evil driver doing if (enable == 1) else.

But I don't actually really care - this is fine if you prefer it. It seems safe to assume (given all the different current driver behaviour) that extra values will never be used.

@radhika-raghavendran
Copy link
Contributor Author

Done. Check is now against if(enable) else.

@radhika-raghavendran
Copy link
Contributor Author

All changes in place now. All checks have passed.

@sg-
Copy link
Contributor

sg- commented Oct 19, 2016

/morph test

@mbed-bot
Copy link

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 0

All builds and test passed!

@0xc0170 0xc0170 merged commit 2175009 into ARMmbed:master Oct 25, 2016
aisair pushed a commit to aisair/mbed that referenced this pull request Apr 30, 2024
Ports for Upcoming Targets


Fixes and Changes

2966: Add kw24 support ARMmbed/mbed-os#2966
3068: MultiTech mDot - clean up PeripheralPins.c and add new pin names ARMmbed/mbed-os#3068
3089: Kinetis HAL: Remove clock initialization code from serial and ticker  ARMmbed/mbed-os#3089
2943: [NRF5] NVIC_SetVector functionality ARMmbed/mbed-os#2943
2938: InterruptIn changes in NCS36510 HAL. ARMmbed/mbed-os#2938
3108: Fix sleep function for NRF52. ARMmbed/mbed-os#3108
3076: STM32F1: Correct timer master value reading ARMmbed/mbed-os#3076
3085: Add LOWPOWERTIMER capability for NUCLEO_F303ZE ARMmbed/mbed-os#3085
3046: [BEETLE] Update BLE stack on Beetle board ARMmbed/mbed-os#3046
3122: [Silicon Labs] Update of Silicon Labs HAL ARMmbed/mbed-os#3122
3022: OnSemi RAM usage fix ARMmbed/mbed-os#3022
3121: STM32F3: Correct UART4 and UART5 defines when using DEVICE_SERIAL_ASYNCH ARMmbed/mbed-os#3121
3142: Targets- NUMAKER_PFM_NUC47216 remove mbed 2 ARMmbed/mbed-os#3142
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants