Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Pedro Alves <palves@redhat.com>
To: gdb-patches@sourceware.org
Subject: [PATCH 08/30] Fix signal handler/event-loop races
Date: Fri, 18 Mar 2016 19:25:00 -0000	[thread overview]
Message-ID: <1458328714-4938-9-git-send-email-palves@redhat.com> (raw)
In-Reply-To: <1458328714-4938-1-git-send-email-palves@redhat.com>

GDB's core signal handling suffers from a classical signal handler /
mainline code race:

  int
  gdb_do_one_event (void)
  {
  ...
  /* First let's see if there are any asynchronous signal handlers
       that are ready.  These would be the result of invoking any of the
       signal handlers.  */
    if (invoke_async_signal_handlers ())
      return 1;
  ...
    /* Block waiting for a new event.  (...).  */

    if (gdb_wait_for_event (1) < 0)
      return -1;
  ...
  }

If a signal is delivered while gdb is blocked in the poll/select
inside gdb_wait_for_event, then the select/poll breaks with EINTR,
we'll loop back around and call invoke_async_signal_handlers.

However, if the signal handler runs between
invoke_async_signal_handlers and gdb_wait_for_event,
gdb_wait_for_event will block, until the next unrelated event...

The fix is to a struct serial_event, and register it in the set of
files that select/poll in gdb_wait_for_event waits on.  The signal
handlers that defer work to invoke_async_signal_handlers call
mark_async_signal_handler, which is adjusted to also set the new
serial event in addition to setting a flag, and is thus now is
garanteed to immediately unblock the next gdb_select/poll call, up
until invoke_async_signal_handlers is called and the event is cleared.

gdb/ChangeLog:
yyyy-mm-dd  Pedro Alves  <palves@redhat.com>

	* event-loop.c: Include "ser-event.h".
	(async_signal_handlers_serial_event): New global.
	(async_signals_handler, initialize_async_signal_handlers): New
	functions.
	(mark_async_signal_handler): Set
	async_signal_handlers_serial_event.
	(invoke_async_signal_handlers): Clear
	async_signal_handlers_serial_event.
	* event-top.c (async_init_signals): Call
	initialize_async_signal_handlers.
---
 gdb/event-loop.c | 32 +++++++++++++++++++++++++++++++-
 gdb/event-loop.h |  2 ++
 gdb/event-top.c  |  2 ++
 3 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/gdb/event-loop.c b/gdb/event-loop.c
index 0e1cb2b..052d535 100644
--- a/gdb/event-loop.c
+++ b/gdb/event-loop.c
@@ -21,6 +21,7 @@
 #include "event-loop.h"
 #include "event-top.h"
 #include "queue.h"
+#include "ser-event.h"
 
 #ifdef HAVE_POLL
 #if defined (HAVE_POLL_H)
@@ -262,6 +263,28 @@ static int update_wait_timeout (void);
 static int poll_timers (void);
 \f
 
+/* This event is signalled whenever an asynchronous handler needs to
+   defer an action to the event loop.  */
+static struct serial_event *async_signal_handlers_serial_event;
+
+/* Callback registered with ASYNC_SIGNAL_HANDLERS_SERIAL_EVENT.  */
+
+static void
+async_signals_handler (int error, gdb_client_data client_data)
+{
+  /* Do nothing.  Handlers are run by invoke_async_signal_handlers
+     from instead.  */
+}
+
+void
+initialize_async_signal_handlers (void)
+{
+  async_signal_handlers_serial_event = make_serial_event ();
+
+  add_file_handler (serial_event_fd (async_signal_handlers_serial_event),
+		    async_signals_handler, NULL);
+}
+
 /* Process one high level event.  If nothing is ready at this time,
    wait for something to happen (via gdb_wait_for_event), then process
    it.  Returns >0 if something was done otherwise returns <0 (this
@@ -905,6 +928,7 @@ void
 mark_async_signal_handler (async_signal_handler * async_handler_ptr)
 {
   async_handler_ptr->ready = 1;
+  serial_event_set (async_signal_handlers_serial_event);
 }
 
 /* See event-loop.h.  */
@@ -925,13 +949,19 @@ async_signal_handler_is_marked (async_signal_handler *async_handler_ptr)
 
 /* Call all the handlers that are ready.  Returns true if any was
    indeed ready.  */
+
 static int
 invoke_async_signal_handlers (void)
 {
   async_signal_handler *async_handler_ptr;
   int any_ready = 0;
 
-  /* Invoke ready handlers.  */
+  /* We're going to handle all pending signals, so no need to wake up
+     the event loop again the next time around.  Note this must be
+     cleared _before_ calling the callbacks, to avoid races.  */
+  serial_event_clear (async_signal_handlers_serial_event);
+
+  /* Invoke all ready handlers.  */
 
   while (1)
     {
diff --git a/gdb/event-loop.h b/gdb/event-loop.h
index 1b765e1..155dafa 100644
--- a/gdb/event-loop.h
+++ b/gdb/event-loop.h
@@ -143,3 +143,5 @@ extern void mark_async_event_handler (struct async_event_handler *handler);
 /* Mark the handler (ASYNC_HANDLER_PTR) as NOT ready.  */
 
 extern void clear_async_event_handler (struct async_event_handler *handler);
+
+extern void initialize_async_signal_handlers (void);
diff --git a/gdb/event-top.c b/gdb/event-top.c
index 769485f..9fff2be 100644
--- a/gdb/event-top.c
+++ b/gdb/event-top.c
@@ -747,6 +747,8 @@ gdb_readline_no_editing_callback (gdb_client_data client_data)
 void
 async_init_signals (void)
 {
+  initialize_async_signal_handlers ();
+
   signal (SIGINT, handle_sigint);
   sigint_token =
     create_async_signal_handler (async_request_quit, NULL);
-- 
2.5.0


  parent reply	other threads:[~2016-03-18 19:25 UTC|newest]

Thread overview: 41+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-03-18 19:19 [PATCH 00/30] Stop throwing exceptions from signal handlers Pedro Alves
2016-03-18 19:18 ` [PATCH 01/30] Don't rely on immediate_quit in command_line_input Pedro Alves
2016-03-18 19:18 ` [PATCH 18/30] Fix inconsistent handling of EINTR in ser-*.c backends Pedro Alves
2016-03-18 19:19 ` [PATCH 14/30] Don't call clear_quit_flag in captured_main Pedro Alves
2016-03-18 19:19 ` [PATCH 19/30] ada-lang.c: Introduce type_as_string and use it Pedro Alves
2016-03-18 19:19 ` [PATCH 30/30] Eliminate target_check_pending_interrupt Pedro Alves
2016-03-18 19:19 ` [PATCH 26/30] Use target_terminal_ours_for_output in MI Pedro Alves
2016-03-18 19:19 ` [PATCH 27/30] TUI: GC tui_target_has_run Pedro Alves
2016-03-18 19:19 ` [PATCH 03/30] TUI: check whether in secondary prompt instead of immediate_quit Pedro Alves
2016-03-18 19:19 ` [PATCH 28/30] target remote: Don't rely on immediate_quit (introduce quit handlers) Pedro Alves
2016-03-18 19:19 ` [PATCH 20/30] Use target_terminal_ours_for_output in cp-support.c Pedro Alves
2016-03-18 19:19 ` [PATCH 02/30] Inline command_loop in read_command_line Pedro Alves
2016-03-18 19:19 ` [PATCH 22/30] Use target_terminal_ours_for_output in infcmd.c Pedro Alves
2016-03-18 19:19 ` [PATCH 29/30] Eliminate immediate_quit Pedro Alves
2016-03-18 19:24 ` [PATCH 25/30] Do target_terminal_ours in query & friends instead of in all callers Pedro Alves
2016-03-18 19:24 ` [PATCH 23/30] Use target_terminal_ours_for_output in warning/internal_error Pedro Alves
2016-03-18 19:24 ` [PATCH 17/30] Pass Ctrl-C to the target in target_terminal_inferior Pedro Alves
2016-03-18 19:25 ` [PATCH 04/30] Don't set immediate_quit in prompt_for_continue Pedro Alves
2016-03-18 19:25 ` [PATCH 12/30] Don't call clear_quit_flag in command_handler Pedro Alves
2016-03-18 19:25 ` [PATCH 13/30] Don't call clear_quit_flag in prepare_to_throw_exception Pedro Alves
2016-03-18 19:25 ` Pedro Alves [this message]
2016-03-18 19:26 ` [PATCH 10/30] Make Python use a struct serial event Pedro Alves
2016-03-18 19:26 ` [PATCH 24/30] Add missing cleanups to defaulted_query and prompt_for_continue Pedro Alves
2016-03-18 19:26 ` [PATCH 07/30] Introduce a serial interface for select'able events Pedro Alves
2016-03-18 19:27 ` [PATCH 15/30] Eliminate clear_quit_flag Pedro Alves
2016-03-18 19:27 ` [PATCH 06/30] Remove unused struct serial::name field Pedro Alves
2016-03-18 19:27 ` [PATCH 21/30] Use target_terminal_ours_for_output in exceptions.c Pedro Alves
2016-03-18 19:28 ` [PATCH 11/30] Don't call clear_quit_flag after check_quit_flag Pedro Alves
2016-03-18 19:28 ` [PATCH 05/30] Stop remote-fileio.c from throwing from SIGINT handler Pedro Alves
2016-03-18 19:37 ` [PATCH 09/30] Introduce interruptible_select Pedro Alves
2016-03-21 17:59   ` Simon Marchi
2016-03-21 18:33     ` Pedro Alves
2016-03-21 19:48   ` Simon Marchi
2016-03-21 19:49     ` Pedro Alves
2016-03-18 19:37 ` [PATCH 16/30] Decouple target_interrupt from all-stop/non-stop modes Pedro Alves
2016-03-21 18:21   ` Simon Marchi
2016-03-21 18:24     ` Pedro Alves
2016-03-21 19:52 ` [PATCH 00/30] Stop throwing exceptions from signal handlers Simon Marchi
2016-03-31 14:43   ` Pedro Alves
2016-03-31 18:44     ` Pedro Alves
2016-04-12 16:15     ` Pedro Alves

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=1458328714-4938-9-git-send-email-palves@redhat.com \
    --to=palves@redhat.com \
    --cc=gdb-patches@sourceware.org \
    /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