aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorPedro Alves <palves@redhat.com>2014-07-09 17:52:58 +0100
committerPedro Alves <palves@redhat.com>2014-07-09 17:52:58 +0100
commit1fe2833b6dd03602ba86aa334e81466ea9abe66a (patch)
tree8520c25af7784886891634b1b86c9e93d47edc77
parent0318afefeff53becc1e3e738793ecdb6653629f4 (diff)
downloadfsf-binutils-gdb-upstream/gdb-7.8-branch.zip
fsf-binutils-gdb-upstream/gdb-7.8-branch.tar.gz
fsf-binutils-gdb-upstream/gdb-7.8-branch.tar.bz2
Fix "attach" command vs user input raceupstream/gdb-7.8-branch
On async targets, a synchronous attach is done like this: #1 - target_attach is called (PTRACE_ATTACH is issued) #2 - a continuation is installed #3 - we go back to the event loop #4 - target reports stop (SIGSTOP), event loop wakes up, and attach continuation is called #5 - among other things, the continuation calls target_terminal_inferior, which removes stdin from the event loop Note that in #3, GDB is still processing user input. If the user is fast enough, e.g., with something like: echo -e "attach PID\nset xxx=1" | gdb ... then the "set" command is processed before the attach completes. We get worse behavior even, if input is a tty and therefore readline/editing is enabled, with e.g.,: (gdb) attach PID\nset xxx=1 we then crash readline/gdb, with: Attaching to program: attach-wait-input, process 14537 readline: readline_callback_read_char() called with no handler! Aborted $ Fix this by calling target_terminal_inferior before #3 above. The test covers both scenarios by running with editing/readline forced to both on and off. gdb/ 2014-07-09 Pedro Alves <palves@redhat.com> * infcmd.c (attach_command_post_wait): Don't call target_terminal_inferior here. (attach_command): Call it here instead. gdb/testsuite/ 2014-07-09 Pedro Alves <palves@redhat.com> * gdb.base/attach-wait-input.exp: New file. * gdb.base/attach-wait-input.c: New file.
-rw-r--r--gdb/ChangeLog6
-rw-r--r--gdb/infcmd.c19
-rw-r--r--gdb/testsuite/ChangeLog5
-rw-r--r--gdb/testsuite/gdb.base/attach-wait-input.c40
-rw-r--r--gdb/testsuite/gdb.base/attach-wait-input.exp119
5 files changed, 186 insertions, 3 deletions
diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index d4304e6..1bd19fc 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,9 @@
+2014-07-09 Pedro Alves <palves@redhat.com>
+
+ * infcmd.c (attach_command_post_wait): Don't call
+ target_terminal_inferior here.
+ (attach_command): Call it here instead.
+
2014-07-07 Pedro Alves <palves@redhat.com>
PR gdb/17096
diff --git a/gdb/infcmd.c b/gdb/infcmd.c
index df4fd40..a712b6f 100644
--- a/gdb/infcmd.c
+++ b/gdb/infcmd.c
@@ -2381,9 +2381,6 @@ attach_command_post_wait (char *args, int from_tty, int async_exec)
post_create_inferior (&current_target, from_tty);
- /* Install inferior's terminal modes. */
- target_terminal_inferior ();
-
if (async_exec)
{
/* The user requested an `attach&', so be sure to leave threads
@@ -2499,6 +2496,22 @@ attach_command (char *args, int from_tty)
based on what modes we are starting it with. */
target_terminal_init ();
+ /* Install inferior's terminal modes. This may look like a no-op,
+ as we've just saved them above, however, this does more than
+ restore terminal settings:
+
+ - installs a SIGINT handler that forwards SIGINT to the inferior.
+ Otherwise a Ctrl-C pressed just while waiting for the initial
+ stop would end up as a spurious Quit.
+
+ - removes stdin from the event loop, which we need if attaching
+ in the foreground, otherwise on targets that report an initial
+ stop on attach (which are most) we'd process input/commands
+ while we're in the event loop waiting for that stop. That is,
+ before the attach continuation runs and the command is really
+ finished. */
+ target_terminal_inferior ();
+
/* Set up execution context to know that we should return from
wait_for_inferior as soon as the target reports a stop. */
init_wait_for_inferior ();
diff --git a/gdb/testsuite/ChangeLog b/gdb/testsuite/ChangeLog
index 7667969..66dcc95 100644
--- a/gdb/testsuite/ChangeLog
+++ b/gdb/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2014-07-09 Pedro Alves <palves@redhat.com>
+
+ * gdb.base/attach-wait-input.exp: New file.
+ * gdb.base/attach-wait-input.c: New file.
+
2014-06-23 Siva Chandra Reddy <sivachandra@google.com>
* gdb.python/py-xmethods.exp: Use "progspace" instead of the
diff --git a/gdb/testsuite/gdb.base/attach-wait-input.c b/gdb/testsuite/gdb.base/attach-wait-input.c
new file mode 100644
index 0000000..bddf2e1
--- /dev/null
+++ b/gdb/testsuite/gdb.base/attach-wait-input.c
@@ -0,0 +1,40 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2014 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 <sys/types.h>
+#include <unistd.h>
+
+volatile int should_exit = 0;
+int mypid;
+
+static void
+setup_done (void)
+{
+}
+
+int
+main (void)
+{
+ unsigned int counter = 1;
+
+ mypid = getpid ();
+ setup_done ();
+
+ for (counter = 0; !should_exit && counter < 100; counter++)
+ sleep (1);
+ return 0;
+}
diff --git a/gdb/testsuite/gdb.base/attach-wait-input.exp b/gdb/testsuite/gdb.base/attach-wait-input.exp
new file mode 100644
index 0000000..d6d572e
--- /dev/null
+++ b/gdb/testsuite/gdb.base/attach-wait-input.exp
@@ -0,0 +1,119 @@
+# Copyright 2014 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/>. */
+
+# Verify that GDB waits for the "attach" command to finish before
+# processing the following command.
+#
+# GDB used to have a race where on async targets, in the small window
+# between the attach request and the initial stop for the attach, GDB
+# was still processing user input.
+#
+# The issue was originally detected with:
+#
+# echo -e "attach PID\nset xxx=1" | gdb
+#
+# In that scenario, stdin is not a tty, which disables readline.
+# Explicitly turning off editing exercises the same code path, and is
+# simpler to do, so we test with both editing on and off.
+
+# The test uses the "attach" command.
+if [target_info exists use_gdb_stub] {
+ return
+}
+
+standard_testfile
+
+if {[build_executable "failed to build" $testfile $srcfile debug]} {
+ return -1
+}
+
+# Start the program running, and return its PID, ready for attaching.
+
+proc start_program {binfile} {
+ global gdb_prompt
+ global decimal
+
+ clean_restart $binfile
+
+ if ![runto setup_done] then {
+ fail "Can't run to setup_done"
+ return 0
+ }
+
+ # Get the PID of the test process.
+ set testpid ""
+ set test "get inferior process ID"
+ gdb_test_multiple "p mypid" $test {
+ -re " = ($decimal)\r\n$gdb_prompt $" {
+ set testpid $expect_out(1,string)
+ pass $test
+ }
+ }
+
+ gdb_test "detach" "Detaching from program: .*"
+
+ if {$testpid == ""} {
+ return
+ }
+
+ return $testpid
+}
+
+# Do test proper. EDITING indicates whether "set editing" is on or
+# off.
+
+proc test { editing } {
+ global gdb_prompt
+ global binfile
+ global decimal
+
+ with_test_prefix "editing $editing" {
+
+ set testpid [start_program $binfile]
+ if {$testpid == ""} {
+ return
+ }
+
+ # Enable/disable readline.
+ gdb_test_no_output "set editing $editing"
+
+ # Send both commands at once.
+ send_gdb "attach $testpid\nprint should_exit = 1\n"
+
+ # Use gdb_expect directly instead of gdb_test_multiple to
+ # avoid races with the double prompt.
+ set test "attach and print"
+ gdb_expect {
+ -re "Attaching to program.*process $testpid\r\n.*$gdb_prompt.*$decimal = 1\r\n$gdb_prompt $" {
+ pass "$test"
+ }
+ timeout {
+ fail "$test (timeout)"
+ }
+ }
+
+ # As we've used attach, on quit, we'll detach from the
+ # program. Explicitly kill it in case we failed above.
+ gdb_test "kill" \
+ "" \
+ "after attach, exit" \
+ "Kill the program being debugged.*y or n. $" \
+ "y"
+ }
+}
+
+foreach editing {"on" "off"} {
+ test $editing
+}