[Bug gdb/28549] New: ARM/Cortex-M: improper stack unwinding when the target is in lockup state

["tarek.bouchkati at gmail dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=28549

            Bug ID: 28549
           Summary: ARM/Cortex-M: improper stack unwinding when the target
                    is in lockup state
           Product: gdb
           Version: 9.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gdb
          Assignee: unassigned at sourceware dot org
          Reporter: tarek.bouchkati at gmail dot com
  Target Milestone: ---

=== environment ====================================================
$ gdb-multiarch -v
GNU gdb (Ubuntu 9.2-0ubuntu1~20.04) 9.2

using STM32H723ZG MCU, containing a Cortex-M7 MCU
using OpenOCD as gdbserver

=== how to reproduce ===============================================
start debug session with empty flash
$ openocd -f interface/stlink-dap.cfg -f target/stm32h7x.cfg

since the flash is empty, the core is in lockup state,
ie: PC = 0xfffffffe, LR = 0xffffffff

now using gdb-multiarch, do:
(gdb) tar ext :3333

we get the following error in openocd:
Error: Failed to read memory at 0xfffffffe

=== analysis =======================================================
doing the same operation with higher verbosity, here is OpenOCD log:
Debug: gdb_server.c:384 gdb_log_incoming_packet(): received packet: mfffffffe,2
Debug: gdb_server.c:1495 gdb_read_memory_packet(): addr: 0x00000000fffffffe,
len: 0x00000002
Debug: target.c:2466 target_read_buffer(): reading buffer of 2 byte at
0xfffffffe
Debug: stlink_usb.c:1112 stlink_usb_error_check(): STLINK_SWD_AP_WDATA_ERROR
Error: arm_adi_v5.c:561 mem_ap_read(): Failed to read memory at 0xfffffffe

This means that gdb is requesting to read this address, most probably to unwind
the stack

=== expected behavior ==============================================
do not unwind the stack when PC value is "magic" = exception address

-- 
You are receiving this mail because:
You are on the CC list for the bug.