[PATCH 19/25] Don't resume new threads if scheduler-locking is in effect
Pedro Alves
pedro@palves.net
Mon Jun 20 22:54:13 GMT 2022
If scheduler-locking is in effect, like e.g., with "set
scheduler-locking on", and you step over a function that spawns a new
thread, the new thread is allowed to run free, at least until some
event is hit, at which point, whether the new thread is re-resumed
depends on a number of seemingly random factors. E.g., if the target
is all-stop, and the parent thread hits a breakpoint, and gdb decides
the breakpoint isn't interesting to report to the user, then the
parent thread is resumed, but the new thread is left stopped.
I think that letting the new threads run with scheduler-locking
enabled is a defect. This commit fixes that, making use of the new
clone events on Linux, and of target_thread_events() on targets where
new threads have no connection to the thread that spawned them.
Testcase and documentation changes included.
Change-Id: Ie12140138b37534b7fc1d904da34f0f174aa11ce
---
gdb/NEWS | 3 +
gdb/doc/gdb.texinfo | 3 +
gdb/infrun.c | 41 +++++++++---
.../gdb.threads/schedlock-new-thread.c | 46 ++++++++++++++
.../gdb.threads/schedlock-new-thread.exp | 63 +++++++++++++++++++
5 files changed, 148 insertions(+), 8 deletions(-)
create mode 100644 gdb/testsuite/gdb.threads/schedlock-new-thread.c
create mode 100644 gdb/testsuite/gdb.threads/schedlock-new-thread.exp
diff --git a/gdb/NEWS b/gdb/NEWS
index 5576c355b7a..2e842cb00f9 100644
--- a/gdb/NEWS
+++ b/gdb/NEWS
@@ -3,6 +3,9 @@
*** Changes since GDB 12
+* If supported by the target, when scheduler-locking is in effect, new
+ threads created by the resumed thread are held stopped.
+
* "info breakpoints" now displays enabled breakpoint locations of
disabled breakpoints as in the "y-" state. For example:
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index b3e995f166b..775c4b1347b 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -6932,6 +6932,9 @@ current thread away from the thread that you are debugging. The
@code{replay} mode behaves like @code{off} in record mode and like
@code{on} in replay mode.
+If supported by the target, when scheduler-locking is in effect, new
+threads created by the resumed thread are held stopped.
+
@item show scheduler-locking
Display the current scheduler locking mode.
@end table
diff --git a/gdb/infrun.c b/gdb/infrun.c
index c7d5acf2326..6c273fd8cf2 100644
--- a/gdb/infrun.c
+++ b/gdb/infrun.c
@@ -103,6 +103,8 @@ static bool start_step_over (void);
static bool step_over_info_valid_p (void);
+static bool schedlock_applies (struct thread_info *tp);
+
/* Asynchronous signal handler registered as event loop source for
when we have pending events ready to be passed to the core. */
static struct async_event_handler *infrun_async_inferior_event_token;
@@ -1823,7 +1825,13 @@ static void
update_thread_events_after_step_over (thread_info *event_thread,
const target_waitstatus &event_status)
{
- if (target_supports_set_thread_options (0))
+ if (schedlock_applies (event_thread))
+ {
+ /* If scheduler-locking applies, continue reporting
+ thread-created/thread-cloned events. */
+ return;
+ }
+ else if (target_supports_set_thread_options (0))
{
/* We can control per-thread options. Disable events for the
event thread, unless the thread is gone. */
@@ -2395,9 +2403,14 @@ do_target_resume (ptid_t resume_ptid, bool step, enum gdb_signal sig)
to start stopped. We need to release the displaced stepping
buffer if the stepped thread exits, so we also enable
thread-exit events.
+
+ - If scheduler-locking applies, threads that the current thread
+ spawns should remain halted. It's not strictly necessary to
+ enable thread-exit events in this case, but it doesn't hurt.
*/
if (step_over_info_valid_p ()
- || displaced_step_in_progress_thread (tp))
+ || displaced_step_in_progress_thread (tp)
+ || schedlock_applies (tp))
{
gdb_thread_options options = GDB_TO_CLONE | GDB_TO_EXIT;
if (target_supports_set_thread_options (options))
@@ -2405,6 +2418,13 @@ do_target_resume (ptid_t resume_ptid, bool step, enum gdb_signal sig)
else
target_thread_events (true);
}
+ else
+ {
+ if (target_supports_set_thread_options (0))
+ target_set_thread_options (tp, 0);
+ else if (!displaced_step_in_progress_any_thread ())
+ target_thread_events (false);
+ }
infrun_debug_printf ("resume_ptid=%s, step=%d, sig=%s",
resume_ptid.to_string ().c_str (),
@@ -5995,16 +6015,21 @@ handle_inferior_event (struct execution_control_state *ecs)
parent->set_running (false);
/* If resuming the child, mark it running. */
- if (ecs->ws.kind () == TARGET_WAITKIND_THREAD_CLONED
- || (follow_child || (!detach_fork && (non_stop || sched_multi))))
+ if ((ecs->ws.kind () == TARGET_WAITKIND_THREAD_CLONED
+ && !schedlock_applies (ecs->event_thread))
+ || (ecs->ws.kind () != TARGET_WAITKIND_THREAD_CLONED
+ && (follow_child
+ || (!detach_fork && (non_stop || sched_multi)))))
child->set_running (true);
/* In non-stop mode, also resume the other branch. */
if ((ecs->ws.kind () == TARGET_WAITKIND_THREAD_CLONED
- && target_is_non_stop_p ())
- || (!detach_fork && (non_stop
- || (sched_multi
- && target_is_non_stop_p ()))))
+ && target_is_non_stop_p ()
+ && !schedlock_applies (ecs->event_thread))
+ || (ecs->ws.kind () != TARGET_WAITKIND_THREAD_CLONED
+ && (!detach_fork && (non_stop
+ || (sched_multi
+ && target_is_non_stop_p ())))))
{
if (follow_child)
switch_to_thread (parent);
diff --git a/gdb/testsuite/gdb.threads/schedlock-new-thread.c b/gdb/testsuite/gdb.threads/schedlock-new-thread.c
new file mode 100644
index 00000000000..67a2ef61a7b
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/schedlock-new-thread.c
@@ -0,0 +1,46 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2021-2022 Free Software Foundation, Inc.
+
+ This program is free software; you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <pthread.h>
+#include <assert.h>
+#include <unistd.h>
+
+static void *
+thread_func (void *arg)
+{
+ while (1)
+ sleep (1);
+}
+
+int
+main (void)
+{
+ pthread_t thread;
+ int ret;
+
+ ret = pthread_create (&thread, NULL, thread_func, NULL); /* set break 1 here */
+ assert (ret == 0);
+
+ /* When testing with schedlock enabled, the new thread won't run, so
+ we can't join it, as that would hang forever. Instead, sleep for
+ a bit, enough that if the spawned thread is scheduled, it hits
+ the thread_func breakpoint before the main thread reaches the
+ "return 0" line below. */
+ sleep (3);
+
+ return 0; /* set break 2 here */
+}
diff --git a/gdb/testsuite/gdb.threads/schedlock-new-thread.exp b/gdb/testsuite/gdb.threads/schedlock-new-thread.exp
new file mode 100644
index 00000000000..8952cb7531c
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/schedlock-new-thread.exp
@@ -0,0 +1,63 @@
+# Copyright 2021-2022 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+# Test continuing over a thread spawn with scheduler-locking on.
+
+standard_testfile .c
+
+set syscalls_src $srcdir/lib/my-syscalls.S
+
+if { [build_executable "failed to prepare" $testfile \
+ [list $srcfile $syscalls_src] {debug pthreads}] == -1 } {
+ return
+}
+
+proc test {non-stop schedlock} {
+ save_vars ::GDBFLAGS {
+ append ::GDBFLAGS " -ex \"set non-stop ${non-stop}\""
+ clean_restart $::binfile
+ }
+
+ set linenum1 [gdb_get_line_number "set break 1 here"]
+
+ if { ![runto $::srcfile:$linenum1] } {
+ return
+ }
+
+ delete_breakpoints
+
+ set linenum2 [gdb_get_line_number "set break 2 here"]
+ gdb_breakpoint $linenum2
+
+ gdb_breakpoint "thread_func"
+
+ gdb_test_no_output "set scheduler-locking $schedlock"
+
+ if {$schedlock} {
+ gdb_test "continue" \
+ "return 0.*set break 2 here .*" \
+ "continue does not stop in new thread"
+ } else {
+ gdb_test "continue" \
+ "thread_func .*" \
+ "continue stops in new thread"
+ }
+}
+
+foreach_with_prefix non-stop {off on} {
+ foreach_with_prefix schedlock {off on} {
+ test ${non-stop} ${schedlock}
+ }
+}
--
2.36.0
More information about the Gdb-patches
mailing list