Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Pedro Alves <pedro@palves.net>
To: gdb-patches@sourceware.org
Cc: Tom Tromey <tom@tromey.com>
Subject: [pushed 28/28] Windows gdb: Avoid writing debug registers if watchpoint hit pending
Date: Fri, 24 Apr 2026 21:45:08 +0100	[thread overview]
Message-ID: <20260424204508.1343973-29-pedro@palves.net> (raw)
In-Reply-To: <20260424204508.1343973-1-pedro@palves.net>

Several watchpoint-related testcases, such as
gdb.threads/watchthreads.exp for example, when tested with the backend
in non-stop mode, exposed an interesting detail of the Windows debug
API that wasn't considered before.  The symptom observed is spurious
SIGTRAPs, like:

  Thread 1 "watchthreads" received signal SIGTRAP, Trace/breakpoint trap.
  0x00000001004010b1 in main () at .../src/gdb/testsuite/gdb.threads/watchthreads.c:48
  48              args[i] = 1; usleep (1); /* Init value.  */

After a good amount of staring at logs and headscratching, I realized
the problem:

 #0 - It all starts with the fact that multiple threads can hit an
      event at the same time.  Say, a watchpoint for thread A, and a
      breakpoint for thread B.

 #1 - Say, WaitForDebugEvent reports the breakpoint hit for thread B
      first, then GDB for some reason decides to update debug
      registers, and continue.  Updating debug registers means writing
      the debug registers to _all_ threads, with SetThreadContext.

 #2 - WaitForDebugEvent reports the watchpoint hit for thread A.
      Watchpoint hits are reported as EXCEPTION_SINGLE_STEP.

 #3 - windows-nat checks the Dr6 debug register to check if the step
      was a watchpoint or hardware breakpoint stop, and finds that Dr6
      is completely cleared.  So windows-nat reports a plain SIGTRAP
      (given EXCEPTION_SINGLE_STEP) to the core.

 #4 - Thread A was not supposed to be stepping, so infrun reports the
      SIGTRAP to the user as a random signal.

The strange part is #3 above.  Why was Dr6 cleared?

Turns out that (at least in Windows 10 & 11), writing to _any_ debug
register has the side effect of clearing Dr6, even if you write the
same values the registers already had, back to the registers.

I confirmed it clearly by adding this hack to GDB:

  if (th->context.ContextFlags == 0)
    {
      th->context.ContextFlags = CONTEXT_DEBUGGER_DR;

      /* Get current values of debug registers.  */
      CHECK (GetThreadContext (th->h, &th->context));

      DEBUG_EVENTS ("For 0x%x (once),  Dr6=0x%llx", th->tid, th->context.Dr6);

      /* Write debug registers back to thread, same values,
	 and re-read them.  */
      CHECK (SetThreadContext (th->h, &th->context));
      CHECK (GetThreadContext (th->h, &th->context));

      DEBUG_EVENTS ("For 0x%x (twice), Dr6=0x%llx", th->tid, th->context.Dr6);
    }

Which showed Dr6=0 after the write + re-read:

  [windows events] fill_thread_context: For 0x6a0 (once),  Dr6=0xffff0ff1
  [windows events] fill_thread_context: For 0x6a0 (twice), Dr6=0x0

This commit fixes the issue by detecting that a thread has a pending
watchpoint hit to report (Dr6 has interesting bits set), and if so,
avoid modifying any debug register.  Instead, let the pending
watchpoint hit be reported by WaitForDebugEvent.  If infrun did want
to modify watchpoints, it will still be done when the thread is
eventually re-resumed after the pending watchpoint hit is reported.
(infrun knows how to gracefully handle the case of a watchpoint hit
for a watchpoint that has since been deleted.)

Move the fill_thread_context method from windows_nat_target to
windows_per_inferior so it can be used by gdbserver too.

Approved-By: Tom Tromey <tom@tromey.com>
Change-Id: I21a3daa9e34eecfa054f0fea706e5ab40aabe70a
commit-id:a28f8d4e
---
 gdb/aarch64-windows-nat.c |  7 ++---
 gdb/nat/windows-nat.h     |  6 ++++
 gdb/windows-nat.c         |  2 +-
 gdb/windows-nat.h         |  3 --
 gdb/x86-windows-nat.c     | 64 ++++++++++++++++++++++++++++++++-------
 gdbserver/win32-low.cc    |  8 +++++
 gdbserver/win32-low.h     |  2 ++
 7 files changed, 73 insertions(+), 19 deletions(-)

diff --git a/gdb/aarch64-windows-nat.c b/gdb/aarch64-windows-nat.c
index 871531bb93a..2b082805708 100644
--- a/gdb/aarch64-windows-nat.c
+++ b/gdb/aarch64-windows-nat.c
@@ -30,6 +30,7 @@ struct aarch64_windows_per_inferior : public windows_per_inferior
 {
   aarch64_debug_reg_state dr_state;
 
+  void fill_thread_context (windows_thread_info *th) override;
   void invalidate_thread_context (windows_thread_info *th) override;
 };
 
@@ -42,8 +43,6 @@ struct aarch64_windows_nat_target final
   void initialize_windows_arch (bool attaching) override;
   void cleanup_windows_arch () override;
 
-  void fill_thread_context (windows_thread_info *th) override;
-
   void thread_context_continue (windows_thread_info *th, int killed) override;
   void thread_context_step (windows_thread_info *th) override;
 
@@ -176,10 +175,10 @@ aarch64_windows_nat_target::cleanup_windows_arch ()
   aarch64_remove_debug_reg_state (inferior_ptid.pid ());
 }
 
-/* See windows-nat.h.  */
+/* See nat/windows-nat.h.  */
 
 void
-aarch64_windows_nat_target::fill_thread_context (windows_thread_info *th)
+aarch64_windows_per_inferior::fill_thread_context (windows_thread_info *th)
 {
   CONTEXT *context = &th->context;
 
diff --git a/gdb/nat/windows-nat.h b/gdb/nat/windows-nat.h
index 5258ce52a0f..45d186f771a 100644
--- a/gdb/nat/windows-nat.h
+++ b/gdb/nat/windows-nat.h
@@ -174,6 +174,12 @@ struct windows_process_info
      This function must be supplied by the embedding application.  */
   virtual windows_thread_info *find_thread (ptid_t ptid) = 0;
 
+  /* Fill in the thread's CONTEXT/WOW64_CONTEXT, if it wasn't filled
+     in yet.
+
+     This function must be supplied by the embedding application.  */
+  virtual void fill_thread_context (windows_thread_info *th) = 0;
+
   /* Handle OUTPUT_DEBUG_STRING_EVENT from child process.  Updates
      OURSTATUS and returns the thread id if this represents a thread
      change (this is specific to Cygwin), otherwise 0.
diff --git a/gdb/windows-nat.c b/gdb/windows-nat.c
index fba1ae6c40b..17166fdcfad 100644
--- a/gdb/windows-nat.c
+++ b/gdb/windows-nat.c
@@ -386,7 +386,7 @@ windows_nat_target::fetch_registers (struct regcache *regcache, int r)
   if (th == NULL)
     return;
 
-  fill_thread_context (th);
+  windows_process->fill_thread_context (th);
 
   if (r < 0)
     for (r = 0; r < gdbarch_num_regs (regcache->arch()); r++)
diff --git a/gdb/windows-nat.h b/gdb/windows-nat.h
index e5084631e5d..235844a7c95 100644
--- a/gdb/windows-nat.h
+++ b/gdb/windows-nat.h
@@ -226,9 +226,6 @@ struct windows_nat_target : public inf_child_target
   /* Cleanup arch-specific data after inferior exit.  */
   virtual void cleanup_windows_arch () = 0;
 
-  /* Reload the thread context.  */
-  virtual void fill_thread_context (windows_thread_info *th) = 0;
-
   /* Prepare the thread context for continuing.  */
   virtual void thread_context_continue (windows_thread_info *th,
 					int killed) = 0;
diff --git a/gdb/x86-windows-nat.c b/gdb/x86-windows-nat.c
index 5c09efce32f..5a4876b7fbe 100644
--- a/gdb/x86-windows-nat.c
+++ b/gdb/x86-windows-nat.c
@@ -48,6 +48,7 @@ struct x86_windows_per_inferior : public windows_per_inferior
      a segment register or not.  */
   segment_register_p_ftype *segment_register_p = nullptr;
 
+  void fill_thread_context (windows_thread_info *th) override;
   void invalidate_thread_context (windows_thread_info *th) override;
 };
 
@@ -56,8 +57,6 @@ struct x86_windows_nat_target final : public x86_nat_target<windows_nat_target>
   void initialize_windows_arch (bool attaching) override;
   void cleanup_windows_arch () override;
 
-  void fill_thread_context (windows_thread_info *th) override;
-
   void thread_context_continue (windows_thread_info *th, int killed) override;
   void thread_context_step (windows_thread_info *th) override;
 
@@ -102,10 +101,10 @@ x86_windows_nat_target::cleanup_windows_arch ()
   x86_cleanup_dregs ();
 }
 
-/* See windows-nat.h.  */
+/* See nat/windows-nat.h.  */
 
 void
-x86_windows_nat_target::fill_thread_context (windows_thread_info *th)
+x86_windows_per_inferior::fill_thread_context (windows_thread_info *th)
 {
   x86_windows_process.with_context (th, [&] (auto *context)
     {
@@ -141,13 +140,56 @@ x86_windows_nat_target::thread_context_continue (windows_thread_info *th,
 
       if (th->debug_registers_changed)
 	{
-	  context->ContextFlags |= WindowsContext<decltype(context)>::debug;
-	  context->Dr0 = state->dr_mirror[0];
-	  context->Dr1 = state->dr_mirror[1];
-	  context->Dr2 = state->dr_mirror[2];
-	  context->Dr3 = state->dr_mirror[3];
-	  context->Dr6 = DR6_CLEAR_VALUE;
-	  context->Dr7 = state->dr_control_mirror;
+	  windows_process->fill_thread_context (th);
+
+	  gdb_assert ((context->ContextFlags & CONTEXT_DEBUG_REGISTERS) != 0);
+
+	  /* Check whether the thread has Dr6 set indicating a
+	     watchpoint hit, and we haven't seen the watchpoint event
+	     yet (reported as
+	     EXCEPTION_SINGLE_STEP/STATUS_WX86_SINGLE_STEP).  In that
+	     case, don't change the debug registers.  Changing debug
+	     registers, even if to the same values, makes the kernel
+	     clear Dr6.  The result would be we would lose the
+	     unreported watchpoint hit.  */
+	  if ((context->Dr6 & ~DR6_CLEAR_VALUE) != 0)
+	    {
+	      if (th->last_event.dwDebugEventCode == EXCEPTION_DEBUG_EVENT
+		  && (th->last_event.u.Exception.ExceptionRecord.ExceptionCode
+		      == EXCEPTION_SINGLE_STEP
+		      || (th->last_event.u.Exception.ExceptionRecord.ExceptionCode
+			  == STATUS_WX86_SINGLE_STEP)))
+		{
+		  DEBUG_EVENTS ("0x%x already reported watchpoint", th->tid);
+		}
+	      else
+		{
+		  DEBUG_EVENTS ("0x%x last reported something else (0x%x)",
+				th->tid,
+				th->last_event.dwDebugEventCode);
+
+		  /* Don't touch debug registers.  Let the pending
+		     watchpoint event be reported instead.  We will
+		     update the debug registers later when the thread
+		     is re-resumed by the core after the watchpoint
+		     event.  */
+		  context->ContextFlags &= ~CONTEXT_DEBUG_REGISTERS;
+		}
+	    }
+	  else
+	    DEBUG_EVENTS ("0x%x has no dr6 set", th->tid);
+
+	  if ((context->ContextFlags & CONTEXT_DEBUG_REGISTERS) != 0)
+	    {
+	      DEBUG_EVENTS ("0x%x changing dregs", th->tid);
+	      context->Dr0 = state->dr_mirror[0];
+	      context->Dr1 = state->dr_mirror[1];
+	      context->Dr2 = state->dr_mirror[2];
+	      context->Dr3 = state->dr_mirror[3];
+	      context->Dr6 = DR6_CLEAR_VALUE;
+	      context->Dr7 = state->dr_control_mirror;
+	    }
+
 	  th->debug_registers_changed = false;
 	}
 
diff --git a/gdbserver/win32-low.cc b/gdbserver/win32-low.cc
index e712033898c..826d21867ae 100644
--- a/gdbserver/win32-low.cc
+++ b/gdbserver/win32-low.cc
@@ -118,6 +118,14 @@ win32_require_context (windows_thread_info *th)
 
 /* See nat/windows-nat.h.  */
 
+void
+gdbserver_windows_process::fill_thread_context (windows_thread_info *th)
+{
+  win32_require_context (th);
+}
+
+/* See nat/windows-nat.h.  */
+
 windows_thread_info *
 gdbserver_windows_process::find_thread (ptid_t ptid)
 {
diff --git a/gdbserver/win32-low.h b/gdbserver/win32-low.h
index 01a904043eb..ff680492756 100644
--- a/gdbserver/win32-low.h
+++ b/gdbserver/win32-low.h
@@ -185,6 +185,8 @@ struct gdbserver_windows_process : public windows_nat::windows_process_info
   void handle_unload_dll (const DEBUG_EVENT &current_event) override;
   bool handle_access_violation (const EXCEPTION_RECORD *rec) override;
 
+  void fill_thread_context (windows_nat::windows_thread_info *th) override;
+
   int attaching = 0;
 
   /* A status that hasn't been reported to the core yet, and so
-- 
2.53.0


      parent reply	other threads:[~2026-04-24 20:50 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-04-24 20:44 [pushed 00/28] Windows non-stop prep Pedro Alves
2026-04-24 20:44 ` [pushed 01/28] Windows gdb+gdbserver: New find_thread, replaces thread_rec(DONT_INVALIDATE_CONTEXT) Pedro Alves
2026-04-24 20:44 ` [pushed 02/28] Windows gdb: handle_output_debug_string return type Pedro Alves
2026-04-24 20:44 ` [pushed 03/28] Windows gdb: Eliminate reload_context Pedro Alves
2026-04-24 20:44 ` [pushed 04/28] Windows gdb+gdbserver: Eliminate thread_rec(INVALIDATE_CONTEXT) calls Pedro Alves
2026-04-24 20:44 ` [pushed 05/28] Windows gdb+gdbserver: Eliminate DONT_SUSPEND Pedro Alves
2026-04-24 20:44 ` [pushed 06/28] Windows gdb+gdbserver: Eliminate windows_process_info::thread_rec Pedro Alves
2026-04-24 20:44 ` [pushed 07/28] Windows gdb: Simplify windows_nat_target::wait Pedro Alves
2026-04-26 12:52   ` Hannes Domani
2026-04-27 12:09     ` [pushed/ob] Restore windows_nat_target::wait is_sw_breakpoint change Pedro Alves
2026-04-24 20:44 ` [pushed 08/28] Windows gdb+gdbserver: Move suspending thread to when returning event Pedro Alves
2026-04-24 20:44 ` [pushed 09/28] Windows gdb: Introduce continue_last_debug_event_main_thread Pedro Alves
2026-04-24 20:44 ` [pushed 10/28] Windows gdb: Introduce windows_continue_flags Pedro Alves
2026-04-24 20:44 ` [pushed 11/28] Windows gdb: Factor code out of windows_nat_target::windows_continue Pedro Alves
2026-04-24 20:44 ` [pushed 12/28] Windows gdb: Pending stop and current_event Pedro Alves
2026-04-24 20:44 ` [pushed 13/28] Windows gdb+gdbserver: Elim desired_stop_thread_id / rework pending_stops Pedro Alves
2026-04-24 20:44 ` [pushed 14/28] Windows gdb+gdbserver: Introduce get_last_debug_event_ptid Pedro Alves
2026-04-24 20:44 ` [pushed 15/28] Windows gdb: Can't pass signal to thread other than last stopped thread Pedro Alves
2026-04-24 20:44 ` [pushed 16/28] Windows gdbserver: Fix scheduler-locking Pedro Alves
2026-04-24 20:44 ` [pushed 17/28] Windows gdb: Enable "set scheduler-locking on" Pedro Alves
2026-04-27 12:57   ` [pushed] Mention Windows scheduler-locking in NEWS (Re: [pushed 17/28] Windows gdb: Enable "set scheduler-locking on") Pedro Alves
2026-04-24 20:44 ` [pushed 18/28] Windows gdbserver: Eliminate soft-interrupt mechanism Pedro Alves
2026-04-24 20:44 ` [pushed 19/28] Windows gdb+gdbserver: Make current_event per-thread state Pedro Alves
2026-04-24 20:45 ` [pushed 20/28] Windows gdb+gdbserver: Make last_sig " Pedro Alves
2026-04-24 20:45 ` [pushed 21/28] Windows gdb+gdbserver: Make siginfo_er " Pedro Alves
2026-04-24 20:45 ` [pushed 22/28] Add backpointer from windows_thread_info to windows_process_info Pedro Alves
2026-04-24 20:45 ` [pushed 23/28] Windows gdb+gdbserver: Share $_siginfo reading code Pedro Alves
2026-04-24 20:45 ` [pushed 24/28] Windows gdb+gdbserver: Eliminate struct pending_stop Pedro Alves
2026-04-24 20:45 ` [pushed 25/28] Windows gdb: Change serial_event management Pedro Alves
2026-04-24 20:45 ` [pushed 26/28] Windows gdb: cygwin_set_dr => windows_set_dr, etc Pedro Alves
2026-04-24 20:45 ` [pushed 27/28] Windows gdb: Move debug macros to windows-nat.h Pedro Alves
2026-04-24 20:45 ` Pedro Alves [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260424204508.1343973-29-pedro@palves.net \
    --to=pedro@palves.net \
    --cc=gdb-patches@sourceware.org \
    --cc=tom@tromey.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox