Bug: openocd 0.10.0-6 Ubuntu dies while debugging with -rtos auto

[kfessel]

Description

$_TARGETNAME configure -rtos auto

seemingly leads to maldetection of a different RTOS with Ubuntu upstream openocd-0.10.0-6 (Ubuntu)

there is no maldetection but a failure in named openocd that happens when the execution is stopped
for some reason the thread update will be called even though no RTOS was detected, which leads to openocd segfaulting.

this is reproducable with openocd-0.10.0-6 on Ubuntu (may also Debian since it uses the same Package)

gdb> attach <openocd pid>

provides following backtrace

Thread 1 "openocd" received signal SIGSEGV, Segmentation fault. uCOS_III_update_thread_offsets (rtos=0x564ae6dc72e0) at src/rtos/uCOS-III.c:193 193 src/rtos/uCOS-III.c: No such file or directory.
after that openocd dies

i patched the pm_set_lowest to do nothing to avoid missing clock beeing the reason

Steps to reproduce the issue

build and flash default example (does happen with other applications to)

then start debugging

examples/default$ make BOARD=nucleo-f767zi debug
<>/RIOT/dist/tools/openocd/openocd.sh debug <>/RIOT/examples/default/bin/nucleo-f767zi/default.elf
### Starting Debugging ###
Open On-Chip Debugger 0.10.0
Licensed under GNU GPL v2
For bug reports, read
        http://openocd.org/doc/doxygen/bugs.html
User : 13 3 command.c:544 command_print(): debug_level: 3
Debug: 14 4 command.c:143 script_debug(): command - ocd_command ocd_command type ocd_log_output /tmp/openocd.log
Debug: 15 4 command.c:143 script_debug(): command - log_output ocd_log_output /tmp/openocd.log
Reading symbols from <>/RIOT/examples/default/bin/nucleo-f767zi/default.elf...
Remote debugging using :3333
warning: Overlapping regions in memory map: ignoring
0x080011fa in idle_thread (arg=<optimised out>) at <>/RIOT/core/init.c:61
61              pm_set_lowest();
(gdb) c
Continuing.
^CRemote connection closed
(gdb)

to debug this get openocd dbg symbols and do ein parrallel

sudo gdb
[sudo] password : 
GNU gdb (Ubuntu 8.3-0ubuntu1) 8.3
Copyright (C) 2019 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Type "show copying" and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>.
Find the GDB manual and other documentation resources online at:
    <http://www.gnu.org/software/gdb/documentation/>.

For help, type "help".
Type "apropos word" to search for commands related to "word".
(gdb) attach 32425
Attaching to process 32425
[New LWP 32426]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
0x00007fac7e5e729b in __GI___select (nfds=nfds@entry=15, readfds=readfds@entry=0x7ffc75292420, writefds=writefds@entry=0x0, exceptfds=exceptfds@entry=0x0, timeout=timeout@entry=0x7ffc75292410) at ../sysdeps/unix/sysv/linux/select.c:41
41      ../sysdeps/unix/sysv/linux/select.c: No such file or directory.
(gdb) c
Continuing.

Thread 1 "openocd" received signal SIGSEGV, Segmentation fault.
uCOS_III_update_thread_offsets (rtos=0x564ae6dc72e0) at src/rtos/uCOS-III.c:193
193     src/rtos/uCOS-III.c: No such file or directory.
(gdb) Quit
(gdb) quit
A debugging session is active.

        Inferior 1 [process 32425] will be detached.

Quit anyway? (y or n) y
Detaching from program: /usr/bin/openocd, process 32425
[Inferior 1 (process 32425) detached]

Expected results

not dieing openocd while debuging

Actual results

openocd dies

Versions

Ubuntu openocd-0.10.0-6
riot master

#4058

[kfessel]

i just saw i forgot the bt of the open ocd crash:

#0  uCOS_III_update_thread_offsets (rtos=0x55587d6c0100) at src/rtos/uCOS-III.c:193
#1  uCOS_III_update_threads (rtos=0x55587d6c0100) at src/rtos/uCOS-III.c:320
#2  0x000055587b70617f in rtos_update_threads (target=target@entry=0x55587d6cde30) at src/rtos/rtos.c:546
#3  0x000055587b6fe1a0 in gdb_signal_reply (connection=0x55587d6cf940, target=0x55587d6cde30) at src/server/gdb_server.c:884
#4  gdb_frontend_halted (target=0x55587d6cde30, connection=0x55587d6cf940) at src/server/gdb_server.c:884
#5  0x000055587b6fe965 in gdb_frontend_halted (connection=<optimised out>, target=<optimised out>) at src/server/gdb_server.c:900
#6  gdb_target_callback_event_handler (event=TARGET_EVENT_GDB_HALT, priv=<optimised out>, target=<optimised out>) at src/server/gdb_server.c:900
#7  gdb_target_callback_event_handler (target=<optimised out>, event=<optimised out>, priv=<optimised out>) at src/server/gdb_server.c:888
#8  0x000055587b6f316a in target_call_event_callbacks (target=0x55587d6cde30, event=event@entry=TARGET_EVENT_GDB_HALT) at src/target/target.c:1523
#9  0x000055587b6f31cf in target_call_event_callbacks (target=target@entry=0x55587d6cde30, event=event@entry=TARGET_EVENT_HALTED) at src/target/target.c:1513
#10 0x000055587b7a220d in adapter_poll (target=0x55587d6cde30) at src/target/hla_target.c:467
#11 0x000055587b6f31fe in target_poll (target=0x55587d6cde30) at src/target/target.c:531
#12 0x000055587b6f43d8 in handle_target (priv=<optimised out>) at src/target/target.c:2579
#13 handle_target (priv=<optimised out>) at src/target/target.c:2502
#14 0x000055587b6eba79 in target_call_timer_callback (now=0x7ffcc49eca80, cb=0x55587d6d9bb0) at src/target/target.c:1612
#15 target_call_timer_callbacks_check_time (checktime=checktime@entry=1) at src/target/target.c:1612
#16 0x000055587b6f02ae in target_call_timer_callbacks_check_time (checktime=1) at src/target/target.c:1623
#17 target_call_timer_callbacks () at src/target/target.c:1623
#18 0x000055587b702828 in server_loop (command_context=command_context@entry=0x55587d66d2a0) at src/server/server.c:444
#19 0x000055587b6ad88b in openocd_thread (cmd_ctx=0x55587d66d2a0, argv=<optimised out>, argc=23) at src/openocd.c:304
#20 openocd_main (argc=23, argv=<optimised out>) at src/openocd.c:341
#21 0x00007f38618811e3 in __libc_start_main (main=0x55587b6ad130 <main>, argc=23, argv=0x7ffcc49eccd8, init=<optimised out>, fini=<optimised out>, rtld_fini=<optimised out>, stack_end=0x7ffcc49eccc8) at ../csu/libc-start.c:308
#22 0x000055587b6ad23e in _start () at src/rtos/rtos.c:245

[kfessel]

This is an openocd bug:
within the RTOS detection it may happen that there is no RTOS detected but there is still a structure allocated for it so the other RTOS function of openocd may not detect the missing RTOS fuctionality

before (malfuctioning)

    openocd-0.10.0/src/rtos/rtos.c: line 337:
} else if (strncmp(packet, "qSymbol", 7) == 0) {
	if (rtos_qsymbol(connection, packet, packet_size) == 1) {
		if (target->rtos_auto_detect == true) {
			target->rtos_auto_detect = false;
			target->rtos->type->create(target);
		}
		target->rtos->type->update_threads(target->rtos);
	}
	return ERROR_OK;
} else

after (working just not knowing how to handle RIOT)

  openocd-0.10.0/src/rtos/rtos.c: line 337:
} else if (strncmp(packet, "qSymbol", 7) == 0) {
	if (rtos_qsymbol(connection, packet, packet_size) == 1) {
		if (target->rtos_auto_detect == true) {
			target->rtos_auto_detect = false;
			target->rtos->type->create(target);
		}
		target->rtos->type->update_threads(target->rtos);
	}else
                  os_free(target);
	return ERROR_OK;
} else

[kfessel]

wrong button

[benemorius]

There's a branch of openocd here that has RTOS support for Riot cooked in. It doesn't get updated all the time but it's a fairly recent build.

[kfessel]

I just also did a forward-port of the daniel-k RIOT for openocd patches

[kfessel]

The openocd with the forward ported daniel-k RIOT RTOS patches is available here https://github.com/kfessel/openocd
i also made the thread priority visible

[aabadie]

The openocd with the forward ported daniel-k RIOT RTOS patches is available here

Any plans to provide this work upstream ?

[kfessel]

@aabadie: atm i have no intention to get this upstream

But something happened there: http://openocd.zylin.com/#/c/4256/ seems like @benpicco got something into openocds current master i think he will have tested with current RIOT

[maribu]

The latest stable OpenOCD works fine and there is little we can do on the RIOT side anyway, so IMO this can be closed.