ESP32-S3 stuck in sha_hal_wait_idle

[kasperl]

Shortly after resuming from deep sleep, we sometimes see ESP32-S3 devices (revision 0.2) getting stuck in:

Code: Select all

sha_hal_wait_idle

when gathering entropy for mbedtls. We're calling:

Code: Select all

mbedtls_entropy_func

from our application code early on and the stack trace looks like this:

Code: Select all

0x420bb40c: sha_hal_wait_idle + 0x8
0x42094cc7: esp_sha_dma + 0x2cf
0x42091f18: mbedtls_sha512_update$part$0 + 0xa4
0x42091f7d: mbedtls_sha512_update + 0x11
0x4209063e: mbedtls_sha512 + 0x22
0x4208cca8: mbedtls_md + 0x8c
0x4208c25b: entropy_update + 0x23
0x4208c38d: entropy_gather_internal + 0x7d
0x4208c448: mbedtls_entropy_func + 0x38
...

Has anyone seen something like this? We've added a task watchdog to capture this information, but we wouldn't expect these calls to ever block.

We can try to investigate more and perhaps create a standalone repro. We are not yet sure if this is reproducible on the ESP32 or other revisions of the ESP32-S3.

[ESP_harshal]

Hi @kasperl,

Let's continue the discussion here: https://esp32.com/viewtopic.php?f=13&t= ... 25#p128625

Thanks!