path: root/gdb-rhel-13298-inferior-funcall-bp-condition-2-of-5.patch

From FEDORA_PATCHES Mon Sep 17 00:00:00 2001
From: Andrew Burgess <aburgess@redhat.com>
Date: Fri, 9 Oct 2020 13:27:13 +0200
Subject: gdb-rhel-13298-inferior-funcall-bp-condition-2-of-5.patch

;;gdb: fix b/p conditions with infcalls in multi-threaded inferiors
;;(Andrew Burgess, RHEL-13298)

This commit fixes bug PR 28942, that is, creating a conditional
breakpoint in a multi-threaded inferior, where the breakpoint
condition includes an inferior function call.

Currently, when a user tries to create such a breakpoint, then GDB
will fail with:

  (gdb) break infcall-from-bp-cond-single.c:61 if (return_true ())
  Breakpoint 2 at 0x4011fa: file /tmp/build/gdb/testsuite/../../../src/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.c, line 61.
  (gdb) continue
  Continuing.
  [New Thread 0x7ffff7c5d700 (LWP 2460150)]
  [New Thread 0x7ffff745c700 (LWP 2460151)]
  [New Thread 0x7ffff6c5b700 (LWP 2460152)]
  [New Thread 0x7ffff645a700 (LWP 2460153)]
  [New Thread 0x7ffff5c59700 (LWP 2460154)]
  Error in testing breakpoint condition:
  Couldn't get registers: No such process.
  An error occurred while in a function called from GDB.
  Evaluation of the expression containing the function
  (return_true) will be abandoned.
  When the function is done executing, GDB will silently stop.
  Selected thread is running.
  (gdb)

Or, in some cases, like this:

  (gdb) break infcall-from-bp-cond-simple.c:56 if (is_matching_tid (arg, 1))
  Breakpoint 2 at 0x401194: file /tmp/build/gdb/testsuite/../../../src/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.c, line 56.
  (gdb) continue
  Continuing.
  [New Thread 0x7ffff7c5d700 (LWP 2461106)]
  [New Thread 0x7ffff745c700 (LWP 2461107)]
  ../../src.release/gdb/nat/x86-linux-dregs.c:146: internal-error: x86_linux_update_debug_registers: Assertion `lwp_is_stopped (lwp)' failed.
  A problem internal to GDB has been detected,
  further debugging may prove unreliable.

The precise error depends on the exact thread state; so there's race
conditions depending on which threads have fully started, and which
have not.  But the underlying problem is always the same; when GDB
tries to execute the inferior function call from within the breakpoint
condition, GDB will, incorrectly, try to resume threads that are
already running - GDB doesn't realise that some threads might already
be running.

The solution proposed in this patch requires an additional member
variable thread_info::in_cond_eval.  This flag is set to true (in
breakpoint.c) when GDB is evaluating a breakpoint condition.

In user_visible_resume_ptid (infrun.c), when the in_cond_eval flag is
true, then GDB will only try to resume the current thread, that is,
the thread for which the breakpoint condition is being evaluated.
This solves the problem of GDB trying to resume threads that are
already running.

The next problem is that inferior function calls are assumed to be
synchronous, that is, GDB doesn't expect to start an inferior function
call in thread #1, then receive a stop from thread #2 for some other,
unrelated reason.  To prevent GDB responding to an event from another
thread, we update fetch_inferior_event and do_target_wait in infrun.c,
so that, when an inferior function call (on behalf of a breakpoint
condition) is in progress, we only wait for events from the current
thread (the one evaluating the condition).

In do_target_wait I had to change the inferior_matches lambda
function, which is used to select which inferior to wait on.
Previously the logic was this:

   auto inferior_matches = [&wait_ptid] (inferior *inf)
     {
       return (inf->process_target () != nullptr
               && ptid_t (inf->pid).matches (wait_ptid));
     };

This compares the pid of the inferior against the complete ptid we
want to wait on.  Before this commit wait_ptid was only ever
minus_one_ptid (which is special, and means any process), and so every
inferior would match.

After this commit though wait_ptid might represent a specific thread
in a specific inferior.  If we compare the pid of the inferior to a
specific ptid then these will not match.  The fix is to compare
against the pid extracted from the wait_ptid, not against the complete
wait_ptid itself.

In fetch_inferior_event, after receiving the event, we only want to
stop all the other threads, and call inferior_event_handler with
INF_EXEC_COMPLETE, if we are not evaluating a conditional breakpoint.
If we are, then all the other threads should be left doing whatever
they were before.  The inferior_event_handler call will be performed
once the breakpoint condition has finished being evaluated, and GDB
decides to stop or not.

The final problem that needs solving relates to GDB's commit-resume
mechanism, which allows GDB to collect resume requests into a single
packet in order to reduce traffic to a remote target.

The problem is that the commit-resume mechanism will not send any
resume requests for an inferior if there are already events pending on
the GDB side.

Imagine an inferior with two threads.  Both threads hit a breakpoint,
maybe the same conditional breakpoint.  At this point there are two
pending events, one for each thread.

GDB selects one of the events and spots that this is a conditional
breakpoint, GDB evaluates the condition.

The condition includes an inferior function call, so GDB sets up for
the call and resumes the one thread, the resume request is added to
the commit-resume queue.

When the commit-resume queue is committed GDB sees that there is a
pending event from another thread, and so doesn't send any resume
requests to the actual target, GDB is assuming that when we wait we
will select the event from the other thread.

However, as this is an inferior function call for a condition
evaluation, we will not select the event from the other thread, we
only care about events from the thread that is evaluating the
condition - and the resume for this thread was never sent to the
target.

And so, GDB hangs, waiting for an event from a thread that was never
fully resumed.

To fix this issue I have added the concept of "forcing" the
commit-resume queue.  When enabling commit resume, if the force flag
is true, then any resumes will be committed to the target, even if
there are other threads with pending events.

A note on authorship: this patch was based on some work done by
Natalia Saiapova and Tankut Baris Aktemur from Intel[1].  I have made
some changes to their work in this version.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28942

[1] https://sourceware.org/pipermail/gdb-patches/2020-October/172454.html

Co-authored-by: Natalia Saiapova <natalia.saiapova@intel.com>
Co-authored-by: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Reviewed-By: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Tested-By: Luis Machado <luis.machado@arm.com>
Tested-By: Keith Seitz <keiths@redhat.com>

diff --git a/gdb/breakpoint.c b/gdb/breakpoint.c
--- a/gdb/breakpoint.c
+++ b/gdb/breakpoint.c
@@ -5679,6 +5679,8 @@ bpstat_check_breakpoint_conditions (bpstat *bs, thread_info *thread)
 	{
 	  try
 	    {
+	      scoped_restore reset_in_cond_eval
+		= make_scoped_restore (&thread->control.in_cond_eval, true);
 	      condition_result = breakpoint_cond_eval (cond);
 	    }
 	  catch (const gdb_exception_error &ex)
diff --git a/gdb/gdbthread.h b/gdb/gdbthread.h
--- a/gdb/gdbthread.h
+++ b/gdb/gdbthread.h
@@ -171,6 +171,9 @@ struct thread_control_state
      command.  This is used to decide whether "set scheduler-locking
      step" behaves like "on" or "off".  */
   int stepping_command = 0;
+
+  /* True if the thread is evaluating a BP condition.  */
+  bool in_cond_eval = false;
 };
 
 /* Inferior thread specific part of `struct infcall_suspend_state'.  */
diff --git a/gdb/infcall.c b/gdb/infcall.c
--- a/gdb/infcall.c
+++ b/gdb/infcall.c
@@ -672,6 +672,12 @@ run_inferior_call (std::unique_ptr<call_thread_fsm> sm,
 
       proceed (real_pc, GDB_SIGNAL_0);
 
+      /* Enable commit resume, but pass true for the force flag.  This
+	 ensures any thread we set running in proceed will actually be
+	 committed to the target, even if some other thread in the current
+	 target has a pending event.  */
+      scoped_enable_commit_resumed enable ("infcall", true);
+
       infrun_debug_show_threads ("non-exited threads after proceed for inferior-call",
 				 all_non_exited_threads ());
 
diff --git a/gdb/infrun.c b/gdb/infrun.c
--- a/gdb/infrun.c
+++ b/gdb/infrun.c
@@ -2292,6 +2292,14 @@ user_visible_resume_ptid (int step)
 	 mode.  */
       resume_ptid = inferior_ptid;
     }
+  else if (inferior_ptid != null_ptid
+	   && inferior_thread ()->control.in_cond_eval)
+    {
+      /* The inferior thread is evaluating a BP condition.  Other threads
+	 might be stopped or running and we do not want to change their
+	 state, thus, resume only the current thread.  */
+      resume_ptid = inferior_ptid;
+    }
   else if (!sched_multi && target_supports_multi_process ())
     {
       /* Resume all threads of the current process (and none of other
@@ -3020,12 +3028,24 @@ schedlock_applies (struct thread_info *tp)
 					    execution_direction)));
 }
 
-/* Set process_stratum_target::COMMIT_RESUMED_STATE in all target
-   stacks that have threads executing and don't have threads with
-   pending events.  */
+/* When FORCE_P is false, set process_stratum_target::COMMIT_RESUMED_STATE
+   in all target stacks that have threads executing and don't have threads
+   with pending events.
+
+   When FORCE_P is true, set process_stratum_target::COMMIT_RESUMED_STATE
+   in all target stacks that have threads executing regardless of whether
+   there are pending events or not.
+
+   Passing FORCE_P as false makes sense when GDB is going to wait for
+   events from all threads and will therefore spot the pending events.
+   However, if GDB is only going to wait for events from select threads
+   (i.e. when performing an inferior call) then a pending event on some
+   other thread will not be spotted, and if we fail to commit the resume
+   state for the thread performing the inferior call, then the inferior
+   call will never complete (or even start).  */
 
 static void
-maybe_set_commit_resumed_all_targets ()
+maybe_set_commit_resumed_all_targets (bool force_p)
 {
   scoped_restore_current_thread restore_thread;
 
@@ -3054,7 +3074,7 @@ maybe_set_commit_resumed_all_targets ()
 	 status to report, handle it before requiring the target to
 	 commit its resumed threads: handling the status might lead to
 	 resuming more threads.  */
-      if (proc_target->has_resumed_with_pending_wait_status ())
+      if (!force_p && proc_target->has_resumed_with_pending_wait_status ())
 	{
 	  infrun_debug_printf ("not requesting commit-resumed for target %s, a"
 			       " thread has a pending waitstatus",
@@ -3064,7 +3084,7 @@ maybe_set_commit_resumed_all_targets ()
 
       switch_to_inferior_no_thread (inf);
 
-      if (target_has_pending_events ())
+      if (!force_p && target_has_pending_events ())
 	{
 	  infrun_debug_printf ("not requesting commit-resumed for target %s, "
 			       "target has pending events",
@@ -3157,7 +3177,7 @@ scoped_disable_commit_resumed::reset ()
     {
       /* This is the outermost instance, re-enable
 	 COMMIT_RESUMED_STATE on the targets where it's possible.  */
-      maybe_set_commit_resumed_all_targets ();
+      maybe_set_commit_resumed_all_targets (false);
     }
   else
     {
@@ -3190,7 +3210,7 @@ scoped_disable_commit_resumed::reset_and_commit ()
 /* See infrun.h.  */
 
 scoped_enable_commit_resumed::scoped_enable_commit_resumed
-  (const char *reason)
+  (const char *reason, bool force_p)
   : m_reason (reason),
     m_prev_enable_commit_resumed (enable_commit_resumed)
 {
@@ -3202,7 +3222,7 @@ scoped_enable_commit_resumed::scoped_enable_commit_resumed
 
       /* Re-enable COMMIT_RESUMED_STATE on the targets where it's
 	 possible.  */
-      maybe_set_commit_resumed_all_targets ();
+      maybe_set_commit_resumed_all_targets (force_p);
 
       maybe_call_commit_resumed_all_targets ();
     }
@@ -3957,10 +3977,11 @@ do_target_wait (ptid_t wait_ptid, execution_control_state *ecs,
      polling the rest of the inferior list starting from that one in a
      circular fashion until the whole list is polled once.  */
 
-  auto inferior_matches = [&wait_ptid] (inferior *inf)
+  ptid_t wait_ptid_pid {wait_ptid.pid ()};
+  auto inferior_matches = [&wait_ptid_pid] (inferior *inf)
     {
       return (inf->process_target () != nullptr
-	      && ptid_t (inf->pid).matches (wait_ptid));
+	      && ptid_t (inf->pid).matches (wait_ptid_pid));
     };
 
   /* First see how many matching inferiors we have.  */
@@ -4429,7 +4450,17 @@ fetch_inferior_event ()
        the event.  */
     scoped_disable_commit_resumed disable_commit_resumed ("handling event");
 
-    if (!do_target_wait (minus_one_ptid, &ecs, TARGET_WNOHANG))
+    /* Is the current thread performing an inferior function call as part
+       of a breakpoint condition evaluation?  */
+    bool in_cond_eval = (inferior_ptid != null_ptid
+			 && inferior_thread ()->control.in_cond_eval);
+
+    /* If the thread is in the middle of the condition evaluation, wait for
+       an event from the current thread.  Otherwise, wait for an event from
+       any thread.  */
+    ptid_t waiton_ptid = in_cond_eval ? inferior_ptid : minus_one_ptid;
+
+    if (!do_target_wait (waiton_ptid, &ecs, TARGET_WNOHANG))
       {
 	infrun_debug_printf ("do_target_wait returned no event");
 	disable_commit_resumed.reset_and_commit ();
@@ -4487,7 +4518,12 @@ fetch_inferior_event ()
 	    bool should_notify_stop = true;
 	    bool proceeded = false;
 
-	    stop_all_threads_if_all_stop_mode ();
+	    /* If the thread that stopped just completed an inferior
+	       function call as part of a condition evaluation, then we
+	       don't want to stop all the other threads.  */
+	    if (ecs.event_thread == nullptr
+		|| !ecs.event_thread->control.in_cond_eval)
+	      stop_all_threads_if_all_stop_mode ();
 
 	    clean_up_just_stopped_threads_fsms (&ecs);
 
@@ -4514,7 +4550,7 @@ fetch_inferior_event ()
 		  proceeded = normal_stop ();
 	      }
 
-	    if (!proceeded)
+	    if (!proceeded && !in_cond_eval)
 	      {
 		inferior_event_handler (INF_EXEC_COMPLETE);
 		cmd_done = 1;
diff --git a/gdb/infrun.h b/gdb/infrun.h
--- a/gdb/infrun.h
+++ b/gdb/infrun.h
@@ -406,7 +406,8 @@ extern void maybe_call_commit_resumed_all_targets ();
 
 struct scoped_enable_commit_resumed
 {
-  explicit scoped_enable_commit_resumed (const char *reason);
+  explicit scoped_enable_commit_resumed (const char *reason,
+					 bool force_p = false);
   ~scoped_enable_commit_resumed ();
 
   DISABLE_COPY_AND_ASSIGN (scoped_enable_commit_resumed);
diff --git a/gdb/testsuite/gdb.threads/infcall-from-bp-cond-other-thread-event.c b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-other-thread-event.c
new file mode 100644
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-other-thread-event.c
@@ -0,0 +1,135 @@
+/* Copyright 2022-2024 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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 <unistd.h>
+#include <stdlib.h>
+#include <sched.h>
+
+#define NUM_THREADS 2
+
+pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/* Some global variables to poke, just for something to do.  */
+volatile int global_var_0 = 0;
+volatile int global_var_1 = 0;
+
+/* This flag is updated from GDB.  */
+volatile int raise_signal = 0;
+
+/* Implement the breakpoint condition function.  Release the other thread
+   and try to give the other thread a chance to run.  Then return ANSWER.  */
+int
+condition_core_func (int answer)
+{
+  /* This unlock should release the other thread.  */
+  if (pthread_mutex_unlock (&mutex) != 0)
+    abort ();
+
+  /* And this yield and sleep should (hopefully) give the other thread a
+     chance to run.  This isn't guaranteed of course, but once the other
+     thread does run it should hit a breakpoint, which GDB should
+     (temporarily) ignore, so there's no easy way for us to know the other
+     thread has done what it needs to, thus, yielding and sleeping is the
+     best we can do.  */
+  sched_yield ();
+  sleep (2);
+
+  return answer;
+}
+
+void
+stop_marker ()
+{
+  int a = 100;	/* Final breakpoint here.  */
+}
+
+/* A breakpoint condition function that always returns true.  */
+int
+condition_true_func ()
+{
+  return condition_core_func (1);
+}
+
+/* A breakpoint condition function that always returns false.  */
+int
+condition_false_func ()
+{
+  return condition_core_func (0);
+}
+
+void *
+worker_func (void *arg)
+{
+  volatile int *ptr = 0;
+  int tid = *((int *) arg);
+
+  switch (tid)
+    {
+    case 0:
+      global_var_0 = 11;	/* First thread breakpoint.  */
+      break;
+
+    case 1:
+      if (pthread_mutex_lock (&mutex) != 0)
+	abort ();
+      if (raise_signal)
+	global_var_1 = *ptr;	/* Signal here.  */
+      else
+	global_var_1 = 99;	/* Other thread breakpoint.  */
+      break;
+
+    default:
+      abort ();
+    }
+
+  return NULL;
+}
+
+int
+main ()
+{
+  pthread_t threads[NUM_THREADS];
+  int args[NUM_THREADS];
+
+  /* Set an alarm, just in case the test deadlocks.  */
+  alarm (300);
+
+  /* We want the mutex to start locked.  */
+  if (pthread_mutex_lock (&mutex) != 0)
+    abort ();
+
+  for (int i = 0; i < NUM_THREADS; i++)
+    {
+      args[i] = i;
+      pthread_create (&threads[i], NULL, worker_func, &args[i]);
+    }
+
+  for (int i = 0; i < NUM_THREADS; i++)
+    {
+      void *retval;
+      pthread_join (threads[i], &retval);
+    }
+
+  /* Unlock once we're done, just for cleanliness.  */
+  if (pthread_mutex_unlock (&mutex) != 0)
+    abort ();
+
+  stop_marker ();
+
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.threads/infcall-from-bp-cond-other-thread-event.exp b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-other-thread-event.exp
new file mode 100644
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-other-thread-event.exp
@@ -0,0 +1,174 @@
+# Copyright 2022-2024 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 for conditional breakpoints where the breakpoint condition includes
+# an inferior function call.
+#
+# The tests in this script are testing what happens when an event arrives in
+# another thread while GDB is waiting for the inferior function call (in the
+# breakpoint condition) to finish.
+#
+# The expectation is that GDB will queue events for other threads and wait
+# for the inferior function call to complete, if the condition is true, then
+# the conditional breakpoint should be reported first.  The other thread
+# event should of course, not get lost, and should be reported as soon as
+# the user tries to continue the inferior.
+#
+# If the conditional breakpoint ends up not being taken (the condition is
+# false), then the other thread event should be reported immediately.
+#
+# This script tests what happens when the other thread event is (a) the
+# other thread hitting a breakpoint, and (b) the other thread taking a
+# signal (SIGSEGV in this case).
+
+standard_testfile
+
+if { [build_executable "failed to prepare" ${binfile} "${srcfile}" \
+	  {debug pthreads}] == -1 } {
+    return
+}
+
+set cond_bp_line [gdb_get_line_number "First thread breakpoint"]
+set other_bp_line [gdb_get_line_number "Other thread breakpoint"]
+set final_bp_line [gdb_get_line_number "Final breakpoint here"]
+set signal_line [gdb_get_line_number "Signal here"]
+
+# Start GDB based on TARGET_ASYNC and TARGET_NON_STOP, and then runto main.
+proc start_gdb_and_runto_main { target_async target_non_stop } {
+    save_vars { ::GDBFLAGS } {
+	append ::GDBFLAGS \
+	    " -ex \"maint set target-non-stop $target_non_stop\""
+	append ::GDBFLAGS \
+	    " -ex \"maintenance set target-async ${target_async}\""
+
+	clean_restart ${::binfile}
+    }
+
+    if { ![runto_main] } {
+	return -1
+    }
+
+    return 0
+}
+
+# Run a test of GDB's conditional breakpoints, where the conditions include
+# inferior function calls.  While the inferior function call is executing
+# another thread will hit a breakpoint (when OTHER_THREAD_SIGNAL is false),
+# or receive a signal (when OTHER_THREAD_SIGNAL is true).  GDB should report
+# the conditional breakpoint first (if the condition is true), and then
+# report the second thread event once the inferior is continued again.
+#
+# When STOP_AT_COND is true then the conditional breakpoint will have a
+# condition that evaluates to true (and GDB will stop at the breakpoint),
+# otherwise, the condition will evaluate to false (and GDB will not stop at
+# the breakpoint).
+proc run_condition_test { stop_at_cond other_thread_signal \
+			      target_async target_non_stop } {
+    if { [start_gdb_and_runto_main $target_async \
+	      $target_non_stop] == -1 } {
+	return
+    }
+
+    # Setup the conditional breakpoint.
+    if { $stop_at_cond } {
+	set cond_func "condition_true_func"
+    } else {
+	set cond_func "condition_false_func"
+    }
+    gdb_breakpoint \
+	"${::srcfile}:${::cond_bp_line} if (${cond_func} ())"
+    set cond_bp_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+			"get number for conditional breakpoint"]
+
+    if { $other_thread_signal } {
+	# Arrange for the other thread to raise a signal while GDB is
+	# evaluating the breakpoint condition.
+	gdb_test_no_output "set raise_signal = 1"
+    } else {
+	# And a breakpoint that will be hit by another thread only once the
+	# breakpoint condition starts to be evaluated.
+	gdb_breakpoint "${::srcfile}:${::other_bp_line}"
+	set other_bp_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+			      "get number for other breakpoint"]
+    }
+
+    # A final breakpoint once the test has completed.
+    gdb_breakpoint "${::srcfile}:${::final_bp_line}"
+    set final_bp_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+			  "get number for final breakpoint"]
+
+    if { $stop_at_cond } {
+	# Continue.  The first breakpoint we hit should be the conditional
+	# breakpoint.  The other thread will have hit its breakpoint, but
+	# that will have been deferred until the conditional breakpoint is
+	# reported.
+	gdb_test "continue" \
+	    [multi_line \
+		 "Continuing\\." \
+		 ".*" \
+		 "" \
+		 "Thread ${::decimal} \"\[^\"\r\n\]+\" hit Breakpoint ${cond_bp_num}, worker_func \[^\r\n\]+:${::cond_bp_line}" \
+		 "${::decimal}\\s+\[^\r\n\]+First thread breakpoint\[^\r\n\]+"] \
+	    "hit the conditional breakpoint"
+    }
+
+    if { $other_thread_signal } {
+	# Now continue again, the other thread will now report that it
+	# received a signal.
+	gdb_test "continue" \
+	    [multi_line \
+		 "Continuing\\." \
+		 ".*" \
+		 "Thread ${::decimal} \"\[^\"\r\n\]+\" received signal SIGSEGV, Segmentation fault\\." \
+		 "\\\[Switching to Thread \[^\r\n\]+\\\]" \
+		 "${::hex} in worker_func \[^\r\n\]+:${::signal_line}" \
+		 "${::decimal}\\s+\[^\r\n\]+Signal here\[^\r\n\]+"] \
+	    "received signal in other thread"
+    } else {
+	# Now continue again, the other thread will now report its
+	# breakpoint.
+	gdb_test "continue" \
+	    [multi_line \
+		 "Continuing\\." \
+		 ".*" \
+		 "" \
+		 "Thread ${::decimal} \"\[^\"\r\n\]+\" hit Breakpoint ${other_bp_num}, worker_func \[^\r\n\]+:${::other_bp_line}" \
+		 "${::decimal}\\s+\[^\r\n\]+Other thread breakpoint\[^\r\n\]+"] \
+	    "hit the breakpoint in other thread"
+
+	# Run to the stop marker.
+	gdb_test "continue" \
+	    [multi_line \
+		 "Continuing\\." \
+		 ".*" \
+		 "" \
+		 "Thread ${::decimal} \"\[^\"\r\n\]+\" hit Breakpoint ${final_bp_num}, stop_marker \[^\r\n\]+:${::final_bp_line}" \
+		 "${::decimal}\\s+\[^\r\n\]+Final breakpoint here\[^\r\n\]+"] \
+	    "hit the final breakpoint"
+    }
+
+    gdb_exit
+}
+
+foreach_with_prefix target_async { "on" "off" } {
+    foreach_with_prefix target_non_stop { "on" "off" } {
+	foreach_with_prefix other_thread_signal { true false } {
+	    foreach_with_prefix stop_at_cond { true false } {
+		run_condition_test $stop_at_cond $other_thread_signal \
+		    $target_async $target_non_stop
+	    }
+	}
+    }
+}
diff --git a/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.c b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.c
new file mode 100644
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.c
@@ -0,0 +1,89 @@
+/* Copyright 2022-2024 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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 <unistd.h>
+
+#define NUM_THREADS 3
+
+int
+is_matching_tid (int *tid_ptr, int tid_value)
+{
+  return *tid_ptr == tid_value;
+}
+
+int
+return_true ()
+{
+  return 1;
+}
+
+int
+return_false ()
+{
+  return 0;
+}
+
+int
+function_that_segfaults ()
+{
+  int *p = 0;
+  *p = 1;	/* Segfault happens here.   */
+}
+
+int
+function_with_breakpoint ()
+{
+  return 1;	/* Nested breakpoint.  */
+}
+
+void *
+worker_func (void *arg)
+{
+  int a = 42;	/* Breakpoint here.  */
+}
+
+void
+stop_marker ()
+{
+  int b = 99;	/* Stop marker.  */
+}
+
+int
+main ()
+{
+  pthread_t threads[NUM_THREADS];
+  int args[NUM_THREADS];
+
+  alarm (300);
+
+  for (int i = 0; i < NUM_THREADS; i++)
+    {
+      args[i] = i;
+      pthread_create (&threads[i], NULL, worker_func, &args[i]);
+    }
+
+  for (int i = 0; i < NUM_THREADS; i++)
+    {
+      void *retval;
+      pthread_join (threads[i], &retval);
+    }
+
+  stop_marker ();
+
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.exp b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.exp
new file mode 100644
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-simple.exp
@@ -0,0 +1,235 @@
+# Copyright 2022-2024 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/>.
+
+# Some simple tests of inferior function calls from breakpoint
+# conditions, in multi-threaded inferiors.
+#
+# This test sets up a multi-threaded inferior, and places a breakpoint
+# at a location that many of the threads will reach.  We repeat the
+# test with different conditions, sometimes a single thread should
+# stop at the breakpoint, sometimes multiple threads should stop, and
+# sometimes no threads should stop.
+
+standard_testfile
+
+if { [build_executable "failed to prepare" ${binfile} "${srcfile}" \
+	  {debug pthreads}] == -1 } {
+    return
+}
+
+set cond_bp_line [gdb_get_line_number "Breakpoint here"]
+set stop_bp_line [gdb_get_line_number "Stop marker"]
+set nested_bp_line [gdb_get_line_number "Nested breakpoint"]
+set segv_line [gdb_get_line_number "Segfault happens here"]
+
+# Start GDB based on TARGET_ASYNC and TARGET_NON_STOP, and then runto main.
+proc start_gdb_and_runto_main { target_async target_non_stop } {
+    save_vars { ::GDBFLAGS } {
+	append ::GDBFLAGS \
+	    " -ex \"maint set target-non-stop $target_non_stop\""
+	append ::GDBFLAGS \
+	    " -ex \"maintenance set target-async ${target_async}\""
+
+	clean_restart ${::binfile}
+    }
+
+    if { ![runto_main] } {
+	return -1
+    }
+
+    return 0
+}
+
+# Run a test of GDB's conditional breakpoints, where the conditions include
+# inferior function calls.
+#
+# CONDITION is the expression to be used as the breakpoint condition.
+#
+# N_EXPECTED_HITS is the number of threads that we expect to stop due to
+# CONDITON.
+#
+# MESSAGE is used as a test name prefix.
+proc run_condition_test { message n_expected_hits condition \
+			      target_async target_non_stop } {
+    with_test_prefix $message {
+
+	if { [start_gdb_and_runto_main $target_async \
+		  $target_non_stop] == -1 } {
+	    return
+	}
+
+	# Use this convenience variable to track how often the
+	# breakpoint condition has been evaluated.  This should be
+	# once per thread.
+	gdb_test "set \$n_cond_eval = 0"
+
+	# Setup the conditional breakpoint.
+	gdb_breakpoint \
+	    "${::srcfile}:${::cond_bp_line} if ((++\$n_cond_eval) && (${condition}))"
+
+	# And a breakpoint that we hit when the test is over, this one is
+	# not conditional.  Only the main thread gets here once all the
+	# other threads have finished.
+	gdb_breakpoint "${::srcfile}:${::stop_bp_line}"
+
+	# The number of times we stop at the conditional breakpoint.
+	set n_hit_condition 0
+
+	# Now keep 'continue'-ing GDB until all the threads have finished
+	# and we reach the stop_marker breakpoint.
+	gdb_test_multiple "continue" "spot all breakpoint hits" {
+	    -re " worker_func \[^\r\n\]+${::srcfile}:${::cond_bp_line}\r\n${::decimal}\\s+\[^\r\n\]+Breakpoint here\[^\r\n\]+\r\n${::gdb_prompt} $" {
+		incr n_hit_condition
+		send_gdb "continue\n"
+		exp_continue
+	    }
+
+	    -re " stop_marker \[^\r\n\]+${::srcfile}:${::stop_bp_line}\r\n${::decimal}\\s+\[^\r\n\]+Stop marker\[^\r\n\]+\r\n${::gdb_prompt} $" {
+		pass $gdb_test_name
+	    }
+	}
+
+	gdb_assert { $n_hit_condition == $n_expected_hits } \
+	    "stopped at breakpoint the expected number of times"
+
+	# Ensure the breakpoint condition was evaluated once per thread.
+	gdb_test "print \$n_cond_eval" "= 3" \
+	    "condition was evaluated in each thread"
+    }
+}
+
+# Check that after handling a conditional breakpoint (where the condition
+# includes an inferior call), it is still possible to kill the running
+# inferior, and then restart the inferior.
+#
+# At once point doing this would result in GDB giving an assertion error.
+proc_with_prefix run_kill_and_restart_test { target_async target_non_stop } {
+    # This test relies on the 'start' command, which is not possible with
+    # the plain 'remote' target.
+    if { [target_info gdb_protocol] == "remote" } {
+	return
+    }
+
+    if { [start_gdb_and_runto_main $target_async \
+	      $target_non_stop] == -1 } {
+	return
+    }
+
+    # Setup the conditional breakpoint.
+    gdb_breakpoint \
+	"${::srcfile}:${::cond_bp_line} if (is_matching_tid (arg, 1))"
+    gdb_continue_to_breakpoint "worker_func"
+
+    # Now kill the program being debugged.
+    gdb_test "kill" "" "kill process" \
+	"Kill the program being debugged.*y or n. $" "y"
+
+    # Check we can restart the inferior.  At one point this would trigger an
+    # assertion.
+    gdb_start_cmd
+}
+
+# Create a conditional breakpoint which includes a call to a function that
+# segfaults.  Run GDB and check what happens when the inferior segfaults
+# during the inferior call.
+proc_with_prefix run_bp_cond_segfaults { target_async target_non_stop } {
+    if { [start_gdb_and_runto_main $target_async \
+	      $target_non_stop] == -1 } {
+	return
+    }
+
+    # This test relies on the inferior segfaulting when trying to
+    # access address zero.
+    if { [is_address_zero_readable] } {
+	return
+    }
+
+    # Setup the conditional breakpoint, include a call to
+    # 'function_that_segfaults', which triggers the segfault.
+    gdb_breakpoint \
+	"${::srcfile}:${::cond_bp_line} if (is_matching_tid (arg, 0) && function_that_segfaults ())"
+    set bp_1_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+		      "get number of conditional breakpoint"]
+
+    gdb_test "continue" \
+	[multi_line \
+	     "Continuing\\." \
+	     ".*" \
+	     "Thread ${::decimal} \"infcall-from-bp\" received signal SIGSEGV, Segmentation fault\\." \
+	     "${::hex} in function_that_segfaults \\(\\) at \[^\r\n\]+:${::segv_line}" \
+	     "${::decimal}\\s+\[^\r\n\]+Segfault happens here\[^\r\n\]+" \
+	     "Error in testing condition for breakpoint ${bp_1_num}:" \
+	     "The program being debugged was signaled while in a function called from GDB\\." \
+	     "GDB remains in the frame where the signal was received\\." \
+	     "To change this behavior use \"set unwindonsignal on\"\\." \
+	     "Evaluation of the expression containing the function" \
+	     "\\(function_that_segfaults\\) will be abandoned\\." \
+	     "When the function is done executing, GDB will silently stop\\."]
+}
+
+# Create a conditional breakpoint which includes a call to a function that
+# itself has a breakpoint set within it.  Run GDB and check what happens
+# when GDB hits the nested breakpoint.
+proc_with_prefix run_bp_cond_hits_breakpoint { target_async target_non_stop } {
+    if { [start_gdb_and_runto_main $target_async \
+	      $target_non_stop] == -1 } {
+	return
+    }
+
+    # Setup the conditional breakpoint, include a call to
+    # 'function_with_breakpoint' in which we will shortly place a
+    # breakpoint.
+    gdb_breakpoint \
+	"${::srcfile}:${::cond_bp_line} if (is_matching_tid (arg, 0) && function_with_breakpoint ())"
+    set bp_1_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+		      "get number of conditional breakpoint"]
+
+    gdb_breakpoint "${::srcfile}:${::nested_bp_line}"
+    set bp_2_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+		      "get number of nested breakpoint"]
+
+    gdb_test "continue" \
+	[multi_line \
+	     "Continuing\\." \
+	     ".*" \
+	     "Thread ${::decimal} \"infcall-from-bp\" hit Breakpoint ${bp_2_num}, function_with_breakpoint \\(\\) at \[^\r\n\]+:${::nested_bp_line}" \
+	     "${::decimal}\\s+\[^\r\n\]+Nested breakpoint\[^\r\n\]+" \
+	     "Error in testing condition for breakpoint ${bp_1_num}:" \
+	     "The program being debugged stopped while in a function called from GDB\\." \
+	     "Evaluation of the expression containing the function" \
+	     "\\(function_with_breakpoint\\) will be abandoned\\." \
+	     "When the function is done executing, GDB will silently stop\\."]
+}
+
+foreach_with_prefix target_async { "on" "off" } {
+    foreach_with_prefix target_non_stop { "on" "off" } {
+	run_condition_test "exactly one thread is hit" \
+	    1 "is_matching_tid (arg, 1)" \
+	    $target_async $target_non_stop
+	run_condition_test "exactly two threads are hit" \
+	    2 "(is_matching_tid (arg, 0) || is_matching_tid (arg, 2))" \
+	    $target_async $target_non_stop
+	run_condition_test "all three threads are hit" \
+	    3 "return_true ()" \
+	    $target_async $target_non_stop
+	run_condition_test "no thread is hit" \
+	    0 "return_false ()" \
+	    $target_async $target_non_stop
+
+	run_kill_and_restart_test $target_async $target_non_stop
+	run_bp_cond_segfaults $target_async $target_non_stop
+	run_bp_cond_hits_breakpoint $target_async $target_non_stop
+    }
+}
diff --git a/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.c b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.c
new file mode 100644
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.c
@@ -0,0 +1,139 @@
+/* Copyright 2022-2024 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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 <unistd.h>
+#include <semaphore.h>
+#include <stdlib.h>
+
+#define NUM_THREADS 5
+
+/* Semaphores, used to track when threads have started, and to control
+   when the threads finish.  */
+sem_t startup_semaphore;
+sem_t finish_semaphore;
+
+/* Mutex to control when the first worker thread hit a breakpoint
+   location.  */
+pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/* Global variable to poke, just so threads have something to do.  */
+volatile int global_var = 0;
+
+int
+return_true ()
+{
+  return 1;
+}
+
+int
+return_false ()
+{
+  return 0;
+}
+
+void *
+worker_func (void *arg)
+{
+  int tid = *((int *) arg);
+
+  switch (tid)
+    {
+    case 0:
+      /* Wait for MUTEX to become available, then pass through the
+	 conditional breakpoint location.  */
+      if (pthread_mutex_lock (&mutex) != 0)
+	abort ();
+      global_var = 99;	/* Conditional breakpoint here.  */
+      if (pthread_mutex_unlock (&mutex) != 0)
+	abort ();
+      break;
+
+    default:
+      /* Notify the main thread that the thread has started, then wait for
+	 the main thread to tell us to finish.  */
+      sem_post (&startup_semaphore);
+      if (sem_wait (&finish_semaphore) != 0)
+	abort ();
+      break;
+    }
+}
+
+void
+stop_marker ()
+{
+  global_var = 99;	/* Stop marker.  */
+}
+
+int
+main ()
+{
+  pthread_t threads[NUM_THREADS];
+  int args[NUM_THREADS];
+  void *retval;
+
+  /* An alarm, just in case the thread deadlocks.  */
+  alarm (300);
+
+  /* Semaphore initialization.  */
+  if (sem_init (&startup_semaphore, 0, 0) != 0)
+    abort ();
+  if (sem_init (&finish_semaphore, 0, 0) != 0)
+    abort ();
+
+  /* Lock MUTEX, this prevents the first worker thread from rushing ahead.  */
+  if (pthread_mutex_lock (&mutex) != 0)
+    abort ();
+
+  /* Worker thread creation.  */
+  for (int i = 0; i < NUM_THREADS; i++)
+    {
+      args[i] = i;
+      pthread_create (&threads[i], NULL, worker_func, &args[i]);
+    }
+
+  /* Wait for every thread (other than the first) to tell us it has started
+     up.  */
+  for (int i = 1; i < NUM_THREADS; i++)
+    {
+      if (sem_wait (&startup_semaphore) != 0)
+	abort ();
+    }
+
+  /* Unlock the first thread so it can proceed.  */
+  if (pthread_mutex_unlock (&mutex) != 0)
+    abort ();
+
+  /* Wait for the first thread only.  */
+  pthread_join (threads[0], &retval);
+
+  /* Now post FINISH_SEMAPHORE to allow all the other threads to finish.  */
+  for (int i = 1; i < NUM_THREADS; i++)
+    sem_post (&finish_semaphore);
+
+  /* Now wait for the remaining threads to complete.  */
+  for (int i = 1; i < NUM_THREADS; i++)
+    pthread_join (threads[i], &retval);
+
+  /* Semaphore cleanup.  */
+  sem_destroy (&finish_semaphore);
+  sem_destroy (&startup_semaphore);
+
+  stop_marker ();
+
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.exp b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.exp
new file mode 100644
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/infcall-from-bp-cond-single.exp
@@ -0,0 +1,117 @@
+# Copyright 2022-2024 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/>.
+
+# This test reprocuces bug gdb/28942, performing an inferior function
+# call from a breakpoint condition in a multi-threaded inferior.
+#
+# The important part of this test is that, when the conditional
+# breakpoint is hit, and the condition (which includes an inferior
+# function call) is evaluated, the other threads are running.
+
+standard_testfile
+
+if { [build_executable "failed to prepare" ${binfile} "${srcfile}" \
+	  {debug pthreads}] == -1 } {
+    return
+}
+
+set cond_bp_line [gdb_get_line_number "Conditional breakpoint here"]
+set final_bp_line [gdb_get_line_number "Stop marker"]
+
+# Start GDB based on TARGET_ASYNC and TARGET_NON_STOP, and then runto main.
+proc start_gdb_and_runto_main { target_async target_non_stop } {
+    save_vars { ::GDBFLAGS } {
+	append ::GDBFLAGS \
+	    " -ex \"maint set target-non-stop $target_non_stop\""
+	append ::GDBFLAGS \
+	    " -ex \"maintenance set target-async ${target_async}\""
+
+	clean_restart ${::binfile}
+    }
+
+    if { ![runto_main] } {
+	return -1
+    }
+
+    return 0
+}
+
+# Run a test of GDB's conditional breakpoints, where the conditions include
+# inferior function calls.
+#
+# TARGET_ASYNC and TARGET_NON_STOP are used when starting up GDB.
+#
+# When STOP_AT_COND is true the breakpoint condtion will evaluate to
+# true, and GDB will stop at the breakpoint.  Otherwise, the
+# breakpoint condition will evaluate to false and GDB will not stop at
+# the breakpoint.
+proc run_condition_test { stop_at_cond \
+			      target_async target_non_stop } {
+    if { [start_gdb_and_runto_main $target_async \
+	      $target_non_stop] == -1 } {
+	return
+    }
+
+    # Setup the conditional breakpoint.
+    if { $stop_at_cond } {
+	set cond_func "return_true"
+    } else {
+	set cond_func "return_false"
+    }
+    gdb_breakpoint \
+	"${::srcfile}:${::cond_bp_line} if (${cond_func} ())"
+    set cond_bp_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+			"get number for conditional breakpoint"]
+
+    # And a breakpoint that we hit when the test is over, this one is
+    # not conditional.
+    gdb_breakpoint "${::srcfile}:${::final_bp_line}"
+    set final_bp_num [get_integer_valueof "\$bpnum" "*UNKNOWN*" \
+			  "get number for final breakpoint"]
+
+    if { $stop_at_cond } {
+	# Continue.  The first breakpoint we hit should be the conditional
+	# breakpoint.  The other thread will have hit its breakpoint, but
+	# that will have been deferred until the conditional breakpoint is
+	# reported.
+	gdb_test "continue" \
+	    [multi_line \
+		 "Continuing\\." \
+		 ".*" \
+		 "" \
+		 "Thread ${::decimal} \"\[^\"\r\n\]+\" hit Breakpoint ${cond_bp_num}, worker_func \[^\r\n\]+:${::cond_bp_line}" \
+		 "${::decimal}\\s+\[^\r\n\]+Conditional breakpoint here\[^\r\n\]+"] \
+	    "hit the conditional breakpoint"
+    }
+
+    # Run to the stop marker.
+    gdb_test "continue" \
+	[multi_line \
+	     "Continuing\\." \
+	     ".*" \
+	     "" \
+	     "Thread ${::decimal} \"\[^\"\r\n\]+\" hit Breakpoint ${final_bp_num}, stop_marker \[^\r\n\]+:${::final_bp_line}" \
+	     "${::decimal}\\s+\[^\r\n\]+Stop marker\[^\r\n\]+"] \
+	"hit the final breakpoint"
+}
+
+foreach_with_prefix target_async { "on" "off" } {
+    foreach_with_prefix target_non_stop { "on" "off" } {
+	foreach_with_prefix stop_at_cond { true false } {
+	    run_condition_test $stop_at_cond \
+		$target_async $target_non_stop
+	}
+    }
+}