Sporadic kernel panics on stm32g4 flash erase/writes

[alexanderwachter]

Describe the bug
When writing big data chunks to the flash of an STM32G4 SoC, sporadic kernel panics occur.
It seems that a cache flush is needed after each erase.

To Reproduce
Build mcuboot for the nucleo_g474re and update several times.
Sometimes you will see a panic.

Expected behavior
Flash write and erase should work without panic

Impact
Firmware updates on stm32g4 series do not work

Logs and console output

E: ***** BUS FAULT *****
E:   Precise data bus error
E:   BFAR Address: 0xfab1f1a4
E: r0/a1:  0x0000045d  r1/a2:  0x00000000  r2/a3:  0xfab1f1a4
E: r3/a4:  0x20000208 r12/ip:  0xa0000000 r14/lr:  0x080043e1
E:  xpsr:  0x6100000f
E: Faulting instruction address (r15/pc): 0x08003d6a
E: >>> ZEPHYR FATAL ERROR 0: CPU exception on CPU 0
E: Fault during interrupt handling
E: Current thread: 0x20000208 (unknown)
E: Halting system

[alexanderwachter]

Fix: #32218

[alexanderwachter]

why low @erwango? this affects all stm32g4 users that use the flash driver.

[erwango]

@alexanderwachter This is from zephyr point of view: one series, one driver, not a regression (from what I saw) and not systematic. But doesn't prevent to review quickly and push so that it's getting merged.