Hello,
i've trying to update the codebase of our current project (under development) from NCS 1.9.x to NCS 2.0.x.
I'm experiencing hard fault crashes when I try to perform operations that imply writing to the internal uC flash from inside a BLE attribute write callback. This never happened with the previous NCS versions (< 2).
In particular, I have a BLE characteristic for receiving commands via BLE, declared as
BT_GATT_CHARACTERISTIC(&cmds_uuid.uuid, BT_GATT_CHRC_WRITE, BT_GATT_PERM_WRITE, NULL, write_cmds_cb, NULL),
Inside the write callback (write_cmds_cb), I noticed that writing something like this:
settings_save_one("dev/conn_int", &val, sizeof(val));
causes a Hard Fault of the processor. The RTT Log is the following
00> [00:01:07.655,181] [1;31m<err> os: ***** USAGE FAULT ***** 00> [00:01:07.655,212] [1;31m<err> os: Illegal load of EXC_RETURN into PC[0m 00> [00:01:07.655,212] [1;31m<err> os: r0/a1: 0x00028dd8 r1/a2: 0x61000000 r2/a3: 0x00000040 00> [00:01:07.655,242] [1;31m<err> os: r3/a4: 0x0003d911 r12/ip: 0x20003864 r14/lr: 0x0003ce37[0m 00> [00:01:07.655,242] [1;31m<err> os: xpsr: 0x20003800[0m 00> [00:01:07.655,242] [1;31m<err> os: Faulting instruction address (r15/pc): 0x00000008[0m 00> [00:01:07.655,273] [1;31m<err> os: >>> ZEPHYR FATAL ERROR 0: CPU exception on CPU 0 00> 00> [00:01:07.655,303] [1;31m<err> os: Current thread: 0x20002790 (logging) 00> 00> [00:01:14.902,313] [1;31m<err> os: Halting system
Interestingly, the log reports the current thread to be "logging"...
If i move the settings_save_one operation out from the BLE callback into the system workqueue, it works without crashing.
As i said, I did not have any issues with the previous NCS versions based on Zephyr 2.x.
Any idea?
Thanks,
marco
