Proper deepsleep as proposed upstream

[annejan]

Want to be ahead of the curve micropython#85

[MrSurly]

Would appreciate any feedback or bug reports.

[annejan]

Sure thing!
It seems to work well from my experience sofar 😀

[pacmac]

Sorry to be so dumb, but how can I merge this pull request into my local version so that I can test it ?

[annejan]

I used the pretty invasive method of pulling from @MrSurly's repo-branch with the command
git pull git@github.com:MrSurly/micropython-esp32.git dev-deepsleep
You can do this in a separate branch if you want . .

[pacmac]

@annejan - Thanks that worked, I know the basics, but struggle to understand github sometimes.

Just doing a quick test, this is the output, when the timer expires, I presume I need to check the reset reason, but why the "flash read err, 1000" ??

>>> import machine as m
>>> m.deepsleep(sleep_ms = 3000)
ets Jun  8 2016 00:22:57

rst:0x5 (DEEPSLEEP_RESET),boot:0x17 (SPI_FAST_FLASH_BOOT)
flash read err, 1000
Falling back to built-in command interpreter.
OK

[annejan]

I have not had that happening myself.
Could be you need to make erase and make deploy since it 'sounds' like a storage issue.

In our branch I've added a helper for reading the reset reason . .

STATIC mp_obj_t esp_rtc_get_reset_reason_(mp_obj_t cpu) {
  uint8_t cpu_id = mp_obj_get_int(cpu);
  if (cpu_id > 1) {
    return mp_obj_new_int(0);
  }
  uint8_t val = rtc_get_reset_reason(cpu_id);
  return mp_obj_new_int(val);
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(esp_rtc_get_reset_reason_obj,
                                 esp_rtc_get_reset_reason_);

And exporting the symbols like . .

{MP_ROM_QSTR(MP_QSTR_rtc_get_reset_reason), MP_ROM_PTR(&esp_rtc_get_reset_reason_obj)},

I use it like this . .

if esp.rtc_get_reset_reason(0) == 16:
    print('full reset')
else:
    print('wake from sleep')

http://esp32.info/docs/esp_idf/html/db/dc1/group__rtc__apis.html#ga826b88529aef76135636c06d408513aa

[pacmac]

Thanks, yes I did erase before I flashed it, also calling machine.reset() results in lots of this:

Guru Meditation Error of type IllegalInstruction occurred on core  0. Exception was unhandled.
... Lots & lots more of this..
Guru Meditation Error of type IllegalInstruction occurred on core  0. Exception was unhandled.
Guru Meditation Error of type IllegalInstruction occurred on core  0. Exception was unhandled.
Guru Meditation Error of type IllegalInstruction occurred on core  0. Exception was unhandled.
Guru Meditation Error of type IllegalInstruction occurred on core  0. Exception was unhandled.
Guru Meditation Error of type IllegalI�ets Jun  8 2016 00:22:57

rst:0x10 (RTCWDT_RTC_RESET),boot:0x17 (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0x00
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:2
load:0x3fff0008,len:8
load:0x3fff0010,len:3424
load:0x40078000,len:9460
ho 0 tail 12 room 4
load:0x40080000,len:252
entry 0x40080034

[MrSurly]

In our branch I've added a helper for reading the reset reason . .

@annejan: That's a good idea (and I'm going to steal it), but I'm going to roll it into machine.reset_cause() so it matches mainline MP.

[annejan]

Sounds like I should move it there as well!

Thanks :)

[pacmac]

On the 8266 it's done like this:

if machine.reset_cause() == machine.DEEPSLEEP_RESET:
    print('woke from a deep sleep')
else:
    print('power on or hard reset')

[MrSurly]

I'm working on a full PR for ESP32 upstream. I'll be done today, if you can wait.

…

On Tue, Jun 13, 2017 at 7:49 AM, Anne Jan Brouwer ***@***.***> wrote: Sounds like I should move it there as well! Thanks :) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#4 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAYHCLa8oWY7Auut6ffGRv9GNjxnZ7DNks5sDqGCgaJpZM4NyfxI> .

[MrSurly]

I'm working on a full PR for ESP32 upstream. I'll be done today, if you
can wait.

Ok, this is working, but I'm seeing the same issue as @pacmac, with the multiple illegal instruction errors. This results in an eventual WDT reset, so any call to machine.reset() will look like a WDT reset instead of ... something else.

Here: https://github.com/MrSurly/micropython-esp32/tree/dev-reset-cause

[pacmac]

OK, seems this is a known issue, ( espressif/esp-idf#113) seems to be caused by a slow wakeup of the flash device on the esp32-wroom, there is a work-around built into the IDF: CONFIG_ESP32_DEEP_SLEEP_WAKEUP_DELAY.

>>> import machine as m
>>> m.deepsleep(sleep_ms = 3000)
ets Jun  8 2016 00:22:57

rst:0x5 (DEEPSLEEP_RESET),boot:0x17 (SPI_FAST_FLASH_BOOT)
flash read err, 1000
Falling back to built-in command interpreter.
OK

Although this config setting appears in sdconfig.h, there does not appear to be any code that is looking for this config setting.

As I know next to nothing about the C language, how can I add this check to the code as per the example given in the IDF examples:

#if CONFIG_ESP32_DEEP_SLEEP_WAKEUP_DELAY > 0
    // ROM code has not started yet, so we need to set delay factor
    // used by ets_delay_us first.
    ets_update_cpu_frequency_rom(ets_get_detected_xtal_freq() / 1000000);
    // This delay is configured in menuconfig, it can be used to give
    // the flash chip some time to become ready.
    ets_delay_us(CONFIG_ESP32_DEEP_SLEEP_WAKEUP_DELAY);
#endif

[MrSurly]

Although this config setting appears in sdconfig.h, there does not appear to be any code that is looking for this config setting.

It's used in components/esp32/deep_sleep.c, in the IDF itself.

Simply setting the value should do it.

[pacmac]

Do I set the value directly in sdkconfig.h ?

[MrSurly]

Do I set the value directly in sdkconfig.h ?

Here's what I would do: Use make menuconfig in an empty IDF project, change only the option that might set this flag, then take whatever flags are actually changed, and apply to sdkconfig.h, noting that the format is slightly different from the actual sdkconfig file from the project.