tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER #20992

FRASTM · 2019-11-26T08:36:30Z

Testing kernel/timer/timer_API with LPTIM on stm32L4R5
the test_timer_duration_period fails w"
test_timer_duration_period: tdata.expire_cnt == 4 is false"

changed z_timer_cycle_get_32() function to have better value returned in nb of HW cycles
changed the calculation for next timeout in case of short delay
source LPTIM with LSE or LSI clock for better accuracy
reduce tick freq per sec and adjust for a better divided value of
LPTIM_CLOCK % CONFIG_SYS_CLOCK_TICKS_PER_SEC

Activation of the LPTIM remains an experimental feature to generate kernel timings.

Fixes #20991

zephyrbot · 2019-11-26T08:38:33Z

All checks are passing now.

Tip: The bot edits this comment instead of posting a new one, so you can check the comment's history to see earlier messages.

FRASTM · 2019-11-26T15:29:16Z

If the irq has not yet been handled but the counter just resets after the roll-over we must add thi roll-ove value in the actual counter position

FRASTM · 2019-11-27T11:00:19Z

Changed the calculation of the nb of hw cycles which is returned by the z_timer_cycle_get_32() function in case of LSE clock.
In case of LSI clock the both methods works.

Remaining work: set a timeout close to the current counter.

FRASTM · 2019-12-02T15:12:58Z

This version passed kernel test on nulceo_l4r5zi board with LPTIM timer running on LSI clock
Device testing on:
nucleo_l4r5zi (None) on /dev/ttyACM0
Adding tasks to the queue...
total complete: 82/ 82 100% skipped: 8, failed: 0
74 of 82 tests passed (100.00%), 0 failed, 8 skipped with 0 warnings in 2287.16 seconds
In total 93 test cases were executed on 1 out of total 213 platforms (0.47%)

FRASTM · 2019-12-06T11:55:48Z

as nucleo_wb55rg board has CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC=32000000
a LPTIM clock source of LSI=32000 gives better accuracy in lptimer calculations

FRASTM · 2019-12-09T16:35:19Z

on nucleo_wb55rg (lptim on LSI) :
test passed with west build -b nucleo_wb55rg tests/kernel/fatal
test passed with west build -b nucleo_wb55rg tests/kernel/sched/schedule_api
test passed with west build -b nucleo_wb55rg tests/kernel/timer/

also passed with nucleo_l4r5zi with lptim sourced by LSI

andyross

Late to this one, sorry. I can't speak to the hardware obviously, but the timer logic all looks correct.

soc/arm/st_stm32/common/Kconfig.defconfig.series

erwango · 2020-05-26T18:18:55Z

drivers/timer/Kconfig.stm32_lptim

-	default 32000 if STM32_LPTIM_CLOCK_LSI
 	default 32768 if STM32_LPTIM_CLOCK_LSE
+	default 32000 if STM32_LPTIM_CLOCK_LSI


Not useful but ok

drivers/timer/Kconfig.stm32_lptim

boards/arm/nucleo_l4r5zi/Kconfig.defconfig

erwango · 2020-05-26T18:21:17Z

@jdascenzio, are you fine with the last changes ?

FRASTM · 2020-05-27T07:17:35Z

change after @erwango 's review

erwango

Few more tweaks and I'm good

erwango · 2020-05-27T09:53:34Z

dts/arm/st/l4/stm32l4.dtsi

@@ -340,7 +340,7 @@
 		};

 		lptim1: timers@40007c00 {
-			compatible = "st,stm32-timers";
+			compatible = "st,stm32-lptim";


Can you fix this on stm32wb as well ?

drivers/timer/Kconfig.stm32_lptim

Autoreload value must be decrement by one Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>

When the tickless kernel isn't used, we don't want to wait for ARROK. This wait can be endless. Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>

In the timeout function, remove the compilation flag and use the macro instead. Signed-off-by: Francois Ramu <francois.ramu@st.com>

The current value of the counter must not be added to the accumulator. It will be added when calling z_timer_cycle_get_32. Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>

set the min and max values of the given ticks from 0 to LPTIM_TIMEBASE which is the full register value In case the timeout is FOREVER, then lptimer is stopped Signed-off-by: Francois Ramu <francois.ramu@st.com>

This change makes the lptimer running with lower tick periods and small tick values Signed-off-by: Francois Ramu <francois.ramu@st.com>

based on PR#25412 Some kernel tests use `CONFIG_TICKLESS_KERNEL=n` with `CONFIG_SYS_CLOCK_TICKS_PER_SEC=1` to detect when a test runs longer than 1 second. These tests break if a tick is announced every time a timeout occurs. Only announce if the measured duration since the last tick is at least the duration of a tick. Signed-off-by: Francois Ramu <francois.ramu@st.com>

This is defining the SYS_CLOCK_TICKS_PER_SEC default value depending on the LPTIM CLOCK frequency in case of LPTIMER, to get a TICK value as a divider of the LPTIM clock source. It gives a better result in formulas when converting ticks to count unit. Signed-off-by: Francois Ramu <francois.ramu@st.com>

FRASTM · 2020-05-28T12:38:25Z

rebase

aurel32

Beside the minor nitpick, all looks good. I tested it on a Nucleo L432Kc board and I confirm the issue is fixed by this PR.

aurel32 · 2020-05-28T14:22:24Z

drivers/timer/Kconfig.stm32_lptim

-	bool "STM32 Low Power Timer"
-	depends on (SOC_SERIES_STM32L4X || SOC_SERIES_STM32WBX)
+	bool "STM32 Low Power Timer [EXPERIMENTAL]"
+	depends on SOC_SERIES_STM32L4X && DEVICE_POWER_MANAGEMENT


Minor nitpick: I don't fully see the relation of why the LPTIM support needs to depend on DEVICE_POWER_MANAGEMENT.

My goal is to link lptim to low power mode.
This let kernel timer unchanged for most of the usecase.
The lptim is able to preserve the kernel tick counter during low power sleep mode : only in this case is the lptim configured to keep the kernel timeout.
Refering to power: stm32l4: support of the Low Power Mode #19026

Understood. In that case shouldn't it be SYS_POWER_MANAGEMENT instead of DEVICE_POWER_MANAGEMENT?

aurel32 · 2020-05-28T14:42:19Z

Another comment: when setting SYS_CLOCK_TICKS_PER_SEC=32768 (instead of 4096) with a LSI OSC, test_timer_remaining fails.

This seems to be a math issue in k_busy_wait and setting both CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC and SYS_CLOCK_TICKS_PER_SEC to 32768 (like done on the nRF52 series) just makes the issue to disappear.

carlescufi · 2020-05-28T15:15:08Z

@erwango please re-review

Adds a new LPTIM binding for stm32 soc, based on the timer binding. This will makes a specific filter on dt_compat_enabled("stm32,lptim") Signed-off-by: Francois Ramu <francois.ramu@st.com>

Activation of the LPTIMER is valid for SLEEP MODE only The choice of the lptim clock source is STM32_LPTIM_CLOCK set the LSE in first position to have as default value Signed-off-by: Francois Ramu <francois.ramu@st.com>

For nucleo_l4r5zi board, the LPTIM clock source is defined as LSE as default. Signed-off-by: Francois Ramu <francois.ramu@st.com>

FRASTM · 2020-05-29T09:28:24Z

include lptim in the stm32wbx series

FRASTM requested a review from erwango as a code owner November 26, 2019 08:36

zephyrbot added the area: Timer Timer label Nov 26, 2019

erwango added the DNM This PR should not be merged (Do Not Merge) label Nov 26, 2019

FRASTM changed the title ~~tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER~~ WIP: tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER Nov 26, 2019

erwango changed the title ~~WIP: tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER~~ [WIP] tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER Nov 26, 2019

FRASTM force-pushed the issue-20991 branch 2 times, most recently from 7881a98 to 3c3c0f9 Compare November 26, 2019 15:26

FRASTM force-pushed the issue-20991 branch 3 times, most recently from a4cce57 to fa3afb0 Compare November 27, 2019 10:56

FRASTM force-pushed the issue-20991 branch 2 times, most recently from 97ec815 to 5f12eaa Compare December 2, 2019 15:11

FRASTM force-pushed the issue-20991 branch 3 times, most recently from 44bac7b to 79681dd Compare December 6, 2019 11:38

zephyrbot added the area: Boards label Dec 6, 2019

FRASTM force-pushed the issue-20991 branch 3 times, most recently from 1a82e0c to e62e5ff Compare December 9, 2019 16:26

FRASTM force-pushed the issue-20991 branch from e62e5ff to 043666a Compare December 9, 2019 17:22

erwango added the platform: STM32 ST Micro STM32 label Jan 15, 2020

erwango added this to the v2.2.0 milestone Jan 21, 2020

FRASTM force-pushed the issue-20991 branch from 043666a to 223bf42 Compare February 10, 2020 09:43

FRASTM requested a review from avisconti as a code owner February 10, 2020 09:43

andyross approved these changes May 26, 2020

View reviewed changes

erwango requested changes May 26, 2020

View reviewed changes

FRASTM force-pushed the issue-20991 branch from 3fe808c to 0e8fad2 Compare May 27, 2020 07:16

erwango requested changes May 27, 2020

View reviewed changes

jdascenzio and others added 8 commits May 27, 2020 16:37

driver: timer: stm32_lptim: fix autoreload value

a0635bf

Autoreload value must be decrement by one Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>

driver: timer: stm32_lptim: fix hang when no tickless

f20172a

When the tickless kernel isn't used, we don't want to wait for ARROK. This wait can be endless. Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>

driver: timer: stm32_lptim: tickless mode without a compilation flag

38354f2

In the timeout function, remove the compilation flag and use the macro instead. Signed-off-by: Francois Ramu <francois.ramu@st.com>

driver: timer: fix accumulated counter increment

45b68a8

The current value of the counter must not be added to the accumulator. It will be added when calling z_timer_cycle_get_32. Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>

driver: timer: stm32_lptim set_timeout limit values

ae5516d

set the min and max values of the given ticks from 0 to LPTIM_TIMEBASE which is the full register value In case the timeout is FOREVER, then lptimer is stopped Signed-off-by: Francois Ramu <francois.ramu@st.com>

driver: timer: stm32_lptim control the timer duration

26dcacf

This change makes the lptimer running with lower tick periods and small tick values Signed-off-by: Francois Ramu <francois.ramu@st.com>

FRASTM mentioned this pull request May 28, 2020

power: stm32l4: support of the Low Power Mode #19026

Merged

FRASTM force-pushed the issue-20991 branch from 0e8fad2 to 7d89fa8 Compare May 28, 2020 12:38

aurel32 approved these changes May 28, 2020

View reviewed changes

FRASTM force-pushed the issue-20991 branch from 7d89fa8 to 6fd1e6b Compare May 29, 2020 09:24

FRASTM added 3 commits May 29, 2020 11:27

dts: stm32: timer define a new DT binding for LPTIM

bb807d2

Adds a new LPTIM binding for stm32 soc, based on the timer binding. This will makes a specific filter on dt_compat_enabled("stm32,lptim") Signed-off-by: Francois Ramu <francois.ramu@st.com>

drivers: timer: lptim is [EXPERIMENTAL] for stm32 soc series only

2291816

Activation of the LPTIMER is valid for SLEEP MODE only The choice of the lptim clock source is STM32_LPTIM_CLOCK set the LSE in first position to have as default value Signed-off-by: Francois Ramu <francois.ramu@st.com>

boards: arm: stm32l4r5 unset the LPTIM clock source if enabled

d5de1c6

For nucleo_l4r5zi board, the LPTIM clock source is defined as LSE as default. Signed-off-by: Francois Ramu <francois.ramu@st.com>

FRASTM force-pushed the issue-20991 branch from 6fd1e6b to d5de1c6 Compare May 29, 2020 09:27

erwango approved these changes May 29, 2020

View reviewed changes

carlescufi merged commit a089e36 into zephyrproject-rtos:master May 29, 2020

erwango mentioned this pull request May 11, 2022

drivers: timer: stm32 LPTIM input clock with a prescaler #44266

Merged

FRASTM deleted the issue-20991 branch August 16, 2022 14:26

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER #20992

tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER #20992

FRASTM commented Nov 26, 2019 •

edited

Loading

zephyrbot commented Nov 26, 2019 •

edited

Loading

FRASTM commented Nov 26, 2019

FRASTM commented Nov 27, 2019

FRASTM commented Dec 2, 2019

FRASTM commented Dec 6, 2019

FRASTM commented Dec 9, 2019 •

edited

Loading

andyross left a comment

erwango May 26, 2020

erwango commented May 26, 2020

FRASTM commented May 27, 2020

erwango left a comment

erwango May 27, 2020

FRASTM May 29, 2020

FRASTM commented May 28, 2020

aurel32 left a comment

aurel32 May 28, 2020

FRASTM May 29, 2020

aurel32 May 29, 2020

aurel32 commented May 28, 2020

carlescufi commented May 28, 2020

FRASTM commented May 29, 2020

tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER #20992

tests: kernel: timer_api fails with CONFIG_STM32_LPTIM_TIMER #20992

Conversation

FRASTM commented Nov 26, 2019 • edited Loading

zephyrbot commented Nov 26, 2019 • edited Loading

FRASTM commented Nov 26, 2019

FRASTM commented Nov 27, 2019

FRASTM commented Dec 2, 2019

FRASTM commented Dec 6, 2019

FRASTM commented Dec 9, 2019 • edited Loading

andyross left a comment

Choose a reason for hiding this comment

erwango May 26, 2020

Choose a reason for hiding this comment

erwango commented May 26, 2020

FRASTM commented May 27, 2020

erwango left a comment

Choose a reason for hiding this comment

erwango May 27, 2020

Choose a reason for hiding this comment

FRASTM May 29, 2020

Choose a reason for hiding this comment

FRASTM commented May 28, 2020

aurel32 left a comment

Choose a reason for hiding this comment

aurel32 May 28, 2020

Choose a reason for hiding this comment

FRASTM May 29, 2020

Choose a reason for hiding this comment

aurel32 May 29, 2020

Choose a reason for hiding this comment

aurel32 commented May 28, 2020

carlescufi commented May 28, 2020

FRASTM commented May 29, 2020

FRASTM commented Nov 26, 2019 •

edited

Loading

zephyrbot commented Nov 26, 2019 •

edited

Loading

FRASTM commented Dec 9, 2019 •

edited

Loading