gdbstub: allow killing QEMU via vKill command
PATCH 1/1 gdbstub: allow killing QEMU via vKill command
With multiprocess extensions gdb uses 'vKill' packet instead of 'k' to kill the inferior. Handle 'vKill' the same way 'k' was handled in the presence of single process.
(f|s)ThreadInfo and ThreadExtraInfo")
--- gdbstub.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
gdbstub.c +22 -0
@@ -1383,6 +1383,28 @@ static int gdb_handle_packet(GDBState *s, const char *line_buf)put_packet (s, buf);break ; + }else if (strncmp (p,"Kill;" ,5 ) ==0 ) { +unsigned long pid; + + p +=5 ; + +if (qemu_strtoul (p, &p,16 , &pid)) { +put_packet (s,"E22" ); +break ; + } + process =gdb_get_process (s, pid); + +if (process ==NULL ) { +put_packet (s,"E22" ); +break ; + } +if (s->process_num <=1 ) { +/* Kill the target */ +error_report ("QEMU: Terminated via GDBstub" ); +exit (0 ); + } +/* TODO: handle multiprocess case */ + goto unknown_command; }else { goto unknown_command; }
Signature · 1 line
2.11.0
Hi,
Le 1/29/19 à 9:32 PM, Max Filippov a écrit :
Quoted 30 lines
With multiprocess extensions gdb uses 'vKill' packet instead of 'k' to kill the inferior. Handle 'vKill' the same way 'k' was handled in the presence of single process.
Fixes 7cf48f6752e5 ("gdbstub: add multiprocess support to(f|s)ThreadInfo and ThreadExtraInfo")
Cc Luc Michel <[email protected]>Signed-off-by Max Filippov <[email protected]>--- gdbstub.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)gdbstub.c +9
-0@@ -1383,6 +1383,28 @@ static int gdb_handle_packet(GDBState *s, const char *line_buf)put_packet (s, buf);break ; + }else if (strncmp (p,"Kill;" ,5 ) ==0 ) { +unsigned long pid; + + p +=5 ; + +if (qemu_strtoul (p, &p,16 , &pid)) { +put_packet (s,"E22" ); +break ; + }
gdbstub.c static int gdb_handle_packet(GDBState *s, const char *line_buf) · line 1404
Quoted 10 lines
gdbstub.c +10
-0+ process =gdb_get_process (s, pid); + +if (process ==NULL ) { +put_packet (s,"E22" ); +break ; + } +if (s->process_num <=1 ) { +/* Kill the target */ +error_report ("QEMU: Terminated via GDBstub" ); +exit (0 );
I suggest:
#ifdef CONFIG_USER_ONLY
exit(0);
#else
qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);#endif
instead of exit(0);
Quoted 7 lines
gdbstub.c +3
-0+ } +/* TODO: handle multiprocess case */ + goto unknown_command; }else { goto unknown_command; }
Hi,
Quoted 9 lines · KONRAD Frederic
Hi,
Le 1/29/19 à 9:32 PM, Max Filippov a écrit :
Quoted 6 lines
With multiprocess extensions gdb uses 'vKill' packet instead of 'k' to kill the inferior. Handle 'vKill' the same way 'k' was handled in the presence of single process.
Fixes 7cf48f6752e5 ("gdbstub: add multiprocess support to(f|s)ThreadInfo and ThreadExtraInfo")
Thanks for the patch.
Quoted 37 lines
Quoted 24 lines
Cc Luc Michel <[email protected]>Signed-off-by Max Filippov <[email protected]>---
gdbstub.c| 22 ++++++++++++++++++++++1 file changed, 22 insertions(+)gdbstub.c +9
-0@@ -1383,6 +1383,28 @@ static int gdb_handle_packet(GDBState *s, const char *line_buf) put_packet(s, buf); break; + } else if (strncmp(p, "Kill;", 5) == 0) { + unsigned long pid; + + p += 5; + + if (qemu_strtoul(p, &p, 16, &pid)) { + put_packet(s, "E22"); + break; + }gdbstub.c static int gdb_handle_packet(GDBState *s, const · line 1404
Quoted 10 lines
gdbstub.c +10
-0+ process = gdb_get_process(s, pid); + + if (process == NULL) { + put_packet(s, "E22"); + break; + } + if (s->process_num <= 1) { + /* Kill the target */ + error_report("QEMU: Terminated via GDBstub"); + exit(0);I suggest:
Quoted 4 lines
#ifdef
CONFIG_USER_ONLY
exit(0);
#elseqemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
You are missing a `break;' here.
Regarding the cause, I think it's more an 'host' cause than a 'guest' one.
Quoted 3 lines
#endif
instead of exit(0);
I just tested, this is not ideal in the current shape. Not quitting immediately makes the stub sends a "W00" packet. It's better if we send nothing because GDB interprets that as the connection was closed (which is effectively the case). Otherwise in multiprocess mode, GDB thinks the other processes are still alive (see below).
with exit(0):
(gdb) k
Kill the program being debugged? (y or n) y
Sending packet: $vKill;1#6e...Ack
Remote connection closed
with qemu_system_shutdown_request:
(gdb) k
Kill the program being debugged? (y or n) y
Sending packet: $vKill;1#6e...Ack
Packet received: W00
Packet vKill (kill) is supported
[Inferior 1 (process 1) killed]
So maybe a solution (for another patch) would be to send a 'W' packet for each attached process (using the multiprocess format). Another is to prevent 'W' packets to be sent at all... they are not required in this case.
Quoted 3 lines
gdbstub.c static int gdb_handle_packet(GDBState *s, const · line 1406
Quoted 3 lines
gdbstub.c +2
-0+ } + /* TODO: handle multiprocess case */
About the multiprocess case, I would vote to adopt the same behaviour (i.e. terminate QEMU), whatever the PID is or how many processes are attached. Otherwise we would have different behaviour for the GDB kill command, whether the simulated board exposes multiple processes or not. Moreover, I think considering just one PID for a kill command does not make much sense in our case.
I just tested, GDB is fine with QEMU closing the connection when having multiple processes attached if we do exit(0) directly (i.e. do not send a 'W00' packet).
Thanks.
Luc
Quoted 5 lines
Quoted 5 lines
gdbstub.c +1
-0+ goto unknown_command; } else { goto unknown_command; }
Le 1/30/19 à 3:31 PM, Luc Michel a écrit :
Quoted 73 lines
Hi,
Quoted 9 lines · KONRAD Frederic
Hi,
Le 1/29/19 à 9:32 PM, Max Filippov a écrit :
Quoted 6 lines
With multiprocess extensions gdb uses 'vKill' packet instead of 'k' to kill the inferior. Handle 'vKill' the same way 'k' was handled in the presence of single process.
Fixes 7cf48f6752e5 ("gdbstub: add multiprocess support to(f|s)ThreadInfo and ThreadExtraInfo")
Thanks for the patch.
Quoted 37 lines
Quoted 24 lines
Cc Luc Michel <[email protected]>Signed-off-by Max Filippov <[email protected]>--- gdbstub.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)gdbstub.c +9
-0@@ -1383,6 +1383,28 @@ static int gdb_handle_packet(GDBState *s, const char *line_buf) put_packet(s, buf); break; + } else if (strncmp(p, "Kill;", 5) == 0) { + unsigned long pid; + + p += 5; + + if (qemu_strtoul(p, &p, 16, &pid)) { + put_packet(s, "E22"); + break; + }gdbstub.c static int gdb_handle_packet(GDBState *s, const · line 1404
Quoted 10 lines
gdbstub.c +10
-0+ process = gdb_get_process(s, pid); + + if (process == NULL) { + put_packet(s, "E22"); + break; + } + if (s->process_num <= 1) { + /* Kill the target */ + error_report("QEMU: Terminated via GDBstub"); + exit(0);I suggest:
Quoted 4 lines
#ifdef
CONFIG_USER_ONLY
exit(0);
#else
qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);You are missing a `break;' here.
Regarding the cause, I think it's more an 'host' cause than a 'guest' one.
Quoted 3 lines
#endif
instead of exit(0);
I just tested, this is not ideal in the current shape. Not quitting immediately makes the stub sends a "W00" packet. It's better if we send nothing because GDB interprets that as the connection was closed (which is effectively the case). Otherwise in multiprocess mode, GDB thinks the other processes are still alive (see below).
with exit(0):
(gdb) k
Kill the program being debugged? (y or n) y
Sending packet: $vKill;1#6e...Ack
Remote connection closed
Quoted 7 lines
with
qemu_system_shutdown_request:
(gdb) k
Kill the program being debugged? (y or n) y
Sending packet: $vKill;1#6e...Ack
Packet received: W00
Packet vKill (kill) is supported
[Inferior 1 (process 1) killed]
Ok makes sense. I didn't think of that.
Quoted 15 lines
So maybe a solution (for another patch) would be to send a 'W' packet for each attached process (using the multiprocess format). Another is to prevent 'W' packets to be sent at all... they are not required in this case.
Quoted 3 lines
gdbstub.c static int gdb_handle_packet(GDBState *s, const · line 1406
Quoted 3 lines
gdbstub.c +2
-0+ } + /* TODO: handle multiprocess case */About the multiprocess case, I would vote to adopt the same behaviour (i.e. terminate QEMU), whatever the PID is or how many processes are attached. Otherwise we would have different behaviour for the GDB
killcommand, whether the simulated board exposes multiple processes or not. Moreover, I think considering just one PID for a kill command does not make much sense in our case.
Quoted 3 lines
I just tested, GDB is fine with QEMU closing the connection when having multiple processes attached if we do exit(0) directly (i.e. do not send a 'W00' packet).
Yes I see the same behavior with Linux but it makes GDB crashes with MinGW at least before this patch and the multiprocess one (eg: v3.0.0).
I'll wait for this patch to land and send some trace if I can still reproduce.
Cheers,
Fred
Quoted 10 lines
Thanks.
Luc
Quoted 5 lines
Quoted 5 lines
gdbstub.c +1
-0+ goto unknown_command; } else { goto unknown_command; }