This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH v3 2/5] btrace: allow recording to be started (and stopped) for running threads
On 01/31/2017 04:06 PM, Metzger, Markus T wrote:
> I was also thinking of changing the error messages to refer to the argument
> thread but that would make them less clear when PTID really refers to the
> selected thread - which should be the case most of the time if not always.
>
> Maybe this is not an issue in practice and the "selected thread" errors are
> understandable enough. What do you think?
I think we may be able to sidestep deciding this. :-)
The only path that leads to btrace_fetch that isn't clearly about
the selected thread is record_btrace_resume_thread, AFAICS.
But in that case, we have:
if (!target_is_non_stop_p ())
{
gdb_assert (ptid_match (inferior_ptid, ptid));
ALL_NON_EXITED_THREADS (tp)
if (ptid_match (tp->ptid, ptid))
{
if (ptid_match (tp->ptid, inferior_ptid))
record_btrace_resume_thread (tp, flag);
else
record_btrace_resume_thread (tp, cflag);
}
}
else
{
ALL_NON_EXITED_THREADS (tp)
if (ptid_match (tp->ptid, ptid))
record_btrace_resume_thread (tp, flag);
}
ALL_NON_EXITED_THREADS guarantees btrace_fetch won't see
exited threads through that path. And then in all-stop,
by definition, all threads are stopped. And in the non-stop
branch, we can be sure that the core won't ever try to resume
a running or exited thread. If it does, and we reach as far
as btrace_fetch, it's an internal error, IMO.
That leaves us with this version of the patch (CL already updated),
which calls validate_registers_access very close to the inferior_ptid
references and the "no thread" checks, making it more obvious these are
about user/selected context.
WDYT?
>From 3927351221c1b49827e82a9e94dda2e2f268f90a Mon Sep 17 00:00:00 2001
From: Markus Metzger <markus.t.metzger@intel.com>
Date: Tue, 31 Jan 2017 16:29:59 +0000
Subject: [PATCH] btrace: allow recording to be started (and stopped) for
running threads
When recording is started for a running thread, GDB was able to start tracing
but then failed to read registers to insert the initial entry for the current
PC. We don't really need that initial entry if we don't know where exactly we
started recording. Skip that step to allow recording to be started while
threads are running.
If we do run into errors, we need to undo the tracing enable to not leak this
thread. The operation did not complete so our caller won't clean up this
thread.
For the BTRACE_FORMAT_PT btrace format, we don't need that initial entry since
it will be recorded in the trace. We can omit the call to btrace_add_pc.
CC: Simon Marchi <simon.marchi@ericsson.com>
gdb/ChangeLog:
2017-01-31 Markus Metzger <markus.t.metzger@intel.com>
* btrace.c (btrace_enable): Do not call btrace_add_pc for
BTRACE_FORMAT_PT or if can_access_registers_ptid returns false.
(btrace_fetch): Assert we can access thread registers.
* record-btrace.c (require_btrace_thread, record_btrace_info):
Call validate_registers_access.
gdb/testsuite/ChangeLog:
2017-01-31 Markus Metzger <markus.t.metzger@intel.com>
* gdb.btrace/enable-running.c: New.
* gdb.btrace/enable-running.exp: New.
---
gdb/btrace.c | 34 ++++++++++--
gdb/record-btrace.c | 2 +
gdb/testsuite/gdb.btrace/enable-running.c | 48 ++++++++++++++++
gdb/testsuite/gdb.btrace/enable-running.exp | 85 +++++++++++++++++++++++++++++
4 files changed, 165 insertions(+), 4 deletions(-)
create mode 100644 gdb/testsuite/gdb.btrace/enable-running.c
create mode 100644 gdb/testsuite/gdb.btrace/enable-running.exp
diff --git a/gdb/btrace.c b/gdb/btrace.c
index d266af7..07f747c 100644
--- a/gdb/btrace.c
+++ b/gdb/btrace.c
@@ -1472,10 +1472,33 @@ btrace_enable (struct thread_info *tp, const struct btrace_config *conf)
tp->btrace.target = target_enable_btrace (tp->ptid, conf);
- /* Add an entry for the current PC so we start tracing from where we
- enabled it. */
- if (tp->btrace.target != NULL)
- btrace_add_pc (tp);
+ /* We're done if we failed to enable tracing. */
+ if (tp->btrace.target == NULL)
+ return;
+
+ /* We need to undo the enable in case of errors. */
+ TRY
+ {
+ /* Add an entry for the current PC so we start tracing from where we
+ enabled it.
+
+ If we can't access TP's registers, TP is most likely running. In this
+ case, we can't really say where tracing was enabled so it should be
+ safe to simply skip this step.
+
+ This is not relevant for BTRACE_FORMAT_PT since the trace will already
+ start at the PC at which tracing was enabled. */
+ if (conf->format != BTRACE_FORMAT_PT
+ && can_access_registers_ptid (tp->ptid))
+ btrace_add_pc (tp);
+ }
+ CATCH (exception, RETURN_MASK_ALL)
+ {
+ btrace_disable (tp);
+
+ throw_exception (exception);
+ }
+ END_CATCH
}
/* See btrace.h. */
@@ -1709,6 +1732,9 @@ btrace_fetch (struct thread_info *tp)
if (btinfo->replay != NULL)
return;
+ /* We shouldn't be called on running or exited threads. */
+ gdb_assert (can_access_registers_ptid (tp->ptid));
+
btrace_data_init (&btrace);
cleanup = make_cleanup_btrace_data (&btrace);
diff --git a/gdb/record-btrace.c b/gdb/record-btrace.c
index 8896241..49f61e6 100644
--- a/gdb/record-btrace.c
+++ b/gdb/record-btrace.c
@@ -116,6 +116,7 @@ require_btrace_thread (void)
tp = find_thread_ptid (inferior_ptid);
if (tp == NULL)
error (_("No thread."));
+ validate_registers_access ();
btrace_fetch (tp);
@@ -416,6 +417,7 @@ record_btrace_info (struct target_ops *self)
tp = find_thread_ptid (inferior_ptid);
if (tp == NULL)
error (_("No thread."));
+ validate_registers_access ();
btinfo = &tp->btrace;
diff --git a/gdb/testsuite/gdb.btrace/enable-running.c b/gdb/testsuite/gdb.btrace/enable-running.c
new file mode 100644
index 0000000..c2edf1e
--- /dev/null
+++ b/gdb/testsuite/gdb.btrace/enable-running.c
@@ -0,0 +1,48 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2017 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 <unistd.h>
+
+#define NTHREADS 3
+
+static void *
+test (void *arg)
+{
+ /* Let's hope this is long enough for GDB to enable tracing and check that
+ everything is working as expected. */
+ sleep (10);
+
+ return arg;
+}
+
+int
+main (void)
+{
+ pthread_t th[NTHREADS];
+ int i;
+
+ for (i = 0; i < NTHREADS; ++i)
+ pthread_create (&th[i], NULL, test, NULL);
+
+ test (NULL); /* bp.1 */
+
+ for (i = 0; i < NTHREADS; ++i)
+ pthread_join (th[i], NULL);
+
+ return 0;
+}
diff --git a/gdb/testsuite/gdb.btrace/enable-running.exp b/gdb/testsuite/gdb.btrace/enable-running.exp
new file mode 100644
index 0000000..fe6aa0b
--- /dev/null
+++ b/gdb/testsuite/gdb.btrace/enable-running.exp
@@ -0,0 +1,85 @@
+# This testcase is part of GDB, the GNU debugger.
+#
+# Copyright 2017 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/>.
+
+if { [skip_btrace_tests] } { return -1 }
+
+standard_testfile
+if {[gdb_compile_pthreads "$srcdir/$subdir/$srcfile" "$binfile" executable {debug}] != "" } {
+ return -1
+}
+
+# We need to enable non-stop mode for the remote case.
+save_vars { GDBFLAGS } {
+ append GDBFLAGS " -ex \"set non-stop on\""
+ clean_restart $testfile
+}
+
+if ![runto_main] {
+ return -1
+}
+
+set bp_1 [gdb_get_line_number "bp.1" $srcfile]
+
+gdb_breakpoint $bp_1
+gdb_continue_to_breakpoint "cont to $bp_1" ".*$bp_1\r\n.*"
+gdb_test "cont&" "Continuing\."
+
+# All threads are running. Let's start recording.
+gdb_test_no_output "record btrace"
+
+proc check_tracing_enabled { thread } {
+ global gdb_prompt
+
+ with_test_prefix "thread $thread" {
+ gdb_test "thread $thread" "(running).*" "is running"
+
+ # We can't read the trace while the thread is running.
+ gdb_test "info record" "Selected thread is running\." \
+ "info record while running"
+
+ # Stop the thread before reading the trace.
+ gdb_test_multiple "interrupt" "interrupt" {
+ -re "interrupt\r\n$gdb_prompt " {
+ pass "interrupt"
+ }
+ }
+ # Wait until the thread actually stopped.
+ gdb_test_multiple "" "stopped" {
+ -re "Thread $thread.*stopped\." {
+ pass "stopped"
+ }
+ }
+ # We will consume the thread's current location as part of the
+ # "info record" output.
+ gdb_test "info record" [multi_line \
+ "Active record target: record-btrace" \
+ "Recording format: .*" \
+ "Recorded \[0-9\]+ instructions \[^\\\r\\\n\]*" \
+ ]
+
+ # Continue the thread again.
+ gdb_test "cont&" "Continuing\."
+ }
+}
+
+# Check that recording was started on each thread.
+foreach thread {1 2 3 4} {
+ check_tracing_enabled $thread
+}
+
+# Stop recording while all threads are running.
+gdb_test "record stop" "Process record is stopped \[^\\\r\\\n\]*"
--
2.5.5