Is there some issues with ESPHome 2026.6.2 OTA rollback detected! Rolled back from partition 'app0' v1.4.5

[cocooma]

Hello Rar,

It seems like the OTA roles back on ESPHome 2026.6.2. ESP32-P4-nano

[11:17:52.591][I][app:158]: ESP32 Chip: ESP32-P4 rev1.3, 2 core(s)
[11:17:52.878][W][safe_mode:094]: OTA rollback detected! Rolled back from partition 'app0'
[11:17:52.878][W][safe_mode:094]: The device reset before the boot was marked successful

I updated it from 2026.5.1

Any idea?

[davidcoulson]

Can you connect to the uart or usb port on the device and get all the logs? It should log why it didn't boot correctly.

[cocooma]

Hello all,

Sorry it took a while to get here to fetch the logs.
My esphome is on 2026.7.0

I've tried to flash the beta firmware v2.0.0-beta.2 and the board gets into a crash loop during the boot.
Then I've reverted it to the v1.4.5 and it is crashing too.

I've flashed the firmware via the usb cable as per instructions with the web flash tool. https://web.esphome.io/

Please see the boot log attached.

Thx ,
Oss
ESP Device-logs.txt

[cocooma]

I can't get the esp32-p4 to work not even if I revert esphome to 2026.5.3. and tigo v1.4.5.
Is there a special flash setting?

[RAR]

Hey! I'll take a look here in a bit. That Beta is pretty outdated, I'll cut a fresh one off the next branch to get that caught up.

[cocooma]

Cool, thx. I have the esp32-p4 on the table so I can test it any time

[RAR]

Thanks for grabbing the logs, Oss — that helped a lot.

Two things: a fresh beta, and what your log is actually telling us.

Fresh beta is up: v2.0.0-beta.3, cut off the next branch so you're no longer on the months-old beta.2. Pin it as your source:

external_components:
  - source:
      type: git
      url: https://github.com/RAR/esphome-tigomonitor
      ref: v2.0.0-beta.3

But heads-up — I don't think a version bump alone fixes what your log shows. The crash is at the framework level, before any of this integration runs:

assert failed: spi_flash_disable_interrupts_caches_and_other_cpu cache_utils.c:127
(esp_task_stack_is_sane_cache_disabled())

It's dying in a tight boot loop before ESPHome prints a single line — so none of the tigo code has run yet. That assert means a task whose stack is in PSRAM did a flash operation (during a flash write the cache is disabled, PSRAM goes unreadable, and the task's own stack disappears). That also lines up with it crashing on both beta.2 and v1.4.5 — it's not the integration version, it's the P4 board/framework config.

A couple of things from your log worth checking:

1. Your partition table is the stock ESPHome default (1.75 MB app slots, no history/tsdb partition), which tells me you're probably not building from one of the P4 board files in boards/. Can you paste your full YAML and your esphome version? That's the missing piece — the crash is config/toolchain-specific, so I need to see how it's put together.

2. One quick diagnostic to try in the meantime — force task stacks out of PSRAM and back into internal RAM:

   esp32:
     framework:
       sdkconfig_options:
         CONFIG_SPIRAM_ALLOW_STACK_EXTERNAL_MEMORY: "n"
         CONFIG_FREERTOS_TASK_CREATE_ALLOW_EXT_MEM: "n"

   If it boots with those set, we've confirmed the root cause and can track down which component is putting a stack in PSRAM.

Also worth confirming: the P4 has no built-in Wi-Fi, so it needs an ESP32-C6/-C5 companion wired up via esp32_hosted:. If your YAML just has a bare wifi: with no hosted block, that's a separate problem we'll want to sort out too — share the YAML and we'll go from there.

[cocooma]

Please see my config
esp32-p4.yaml

[cocooma]

Esphome version is 2026.7.0

[RAR]

That helps — thanks. I can see what's going on now: your config is the old pre-2.0 example-p4.yaml with the ref bumped to beta.3, and there are two separate things to fix.

1. The boot loop — your PSRAM is over-clocked. You've got:

psram:
  mode: hex
  speed: 200MHz

but 200 MHz on the P4 is an experimental PSRAM speed — it only works when experimental IDF features are enabled, which your config doesn't do. So PSRAM comes up unstable, and since a task stack lives in PSRAM, its first flash access trips that esp_task_stack_is_sane_cache_disabled assert before ESPHome even starts — exactly your crash loop.

Two ways to fix; try the first one, it's the simplest test:

# Option A — run PSRAM at a safe speed (no experimental flag needed)
psram:
  mode: hex
  speed: 80MHz

# Option B — keep 200MHz, but enable the experimental gate it requires
esp32:
  framework:
    type: esp-idf
    version: recommended
    advanced:
      enable_idf_experimental_features: yes
psram:
  mode: hex
  speed: 200MHz

Also uncomment version: recommended while you're in there. If Option A boots, we've confirmed the cause.

2. Once it boots, you'll be missing the 2.0 storage bits. The History view (esp_tsdb) needs a flash partition + two managed components that your config doesn't have yet. Add:

esp32:
  flash_size: 16MB
  partitions: partitions/tigo-16mb.csv   # grab from the repo's boards/partitions/
  framework:
    components:
      - name: zakery292/esp_tsdb
        source: https://github.com/RAR/esp_tsdb.git
        ref: tigomonitor          # P4 still needs the fork here
      - joltwallet/littlefs^1.16

Honestly the cleanest path is to start from the current P4 reference config instead of the old example — boards/test-p4-tigomonitor.yaml — and drop in your Wi-Fi/UART pins and panel list. It already has the right esp32:/psram:/partitions:/esp32_hosted: scaffolding for a P4.

I'm also refreshing the two example P4 configs in the repo so they carry the experimental-features flag + storage scaffolding out of the box — that'll spare the next person this exact crash.

Give Option A a shot and paste the boot log if it still loops.

[cocooma]

Used your config template boards/test-p4-tigomonitor.yaml
It loops with the same error :/

Here is the config:
esp32-p4-v2.yaml

Boot log:
ESP Device-logs (1).txt

[cocooma]

Forgot to mention I'm using the esp32-p4-nano board

[cocooma]

https://www.waveshare.com/esp32-p4-nano.htm?srsltid=AfmBOoohVrwolxhFV5t90-sR7H6TjYvIJdEvVGwVHTgN4F-4PQqsqiND

[RAR]

put https://rar.github.io/esphome-tigomonitor/ together this morning to help out with this kinda thing.

Try setting your PSRAM to 80Mhz. I'll update that config to correct that being 200.

[cocooma]

It seems like the esp32-p4 nano only supports
20,100 and 200Mhz for the psram also it can only run in hex mode.
I've tried 20 and 100 none of them worked.

But when I commented the psram: mode: hex speed: 20MHz

section it has booted up, obviously with the performance penalty :/

[RAR]

I have a similar board, let me try it out with your config.

[RAR]

and yeah VARIANT_ESP32P4: (20, 100, 200), my bad totally led you wrong there.

[cocooma]

No worries

[RAR]

OK, I did your exact config on an esp32-p4-wifi6 (https://docs.waveshare.com/ESP32-P4-WIFI6) and it had zero issues at 200Mhz psram. I fear it may be the board itself.

[cocooma]

Let me get another one. Thx for checking

[cocooma]

I got it fixed.

I run on the esphome build environment
esphome clean-all esp32-p4.yaml

and added:
execute_from_psram: true
to
esp32: board: esp32-p4-evboard cpu_frequency: 360MHz flash_size: 16MB partitions: partitions/tigo-16mb.csv framework: type: esp-idf advanced: enable_idf_experimental_features: yes execute_from_psram: true

[cocooma]

Thanks for looking into it

[RAR]

I got it fixed.

I run on the esphome build environment esphome clean-all esp32-p4.yaml

and added: execute_from_psram: true to esp32: board: esp32-p4-evboard cpu_frequency: 360MHz flash_size: 16MB partitions: partitions/tigo-16mb.csv framework: type: esp-idf advanced: enable_idf_experimental_features: yes execute_from_psram: true

Great find, I'll add that in the configs.

[cocooma]

This project is amazing!
Everything works now including the bluetooth sync.
Wow

[RAR]

Glad you're enjoying it! yell with any issues.