Skip to contentSkip to conversation

perf tools: port UI from GTK2 to GTK4

PATCH Superseded Matt Turner 4 participants 11 messages Sep 6 – Sep 8

Subsystems (2)

v7 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

v7 cover perf tools: port UI from GTK2 to GTK4

GTK2 is long dead upstream and increasingly hard to keep building on current distros. This series ports perf's GTK-based report browser to GTK4 and fixes it up so it's actually loadable at runtime after the port.

Patch 1 does the mechanical port (build system, widget API changes), including the leftover-GTK2-call and signal-handling fixes that were a separate patch 3 in v4. Patch 2 fixes a runtime issue found after the port that prevented the browser from loading. Patch 3 fixes two stack buffer overflows in the hierarchy view that predate the port.

---
Changes in v7:
- Add patch 3: fix two stack buffer overflows in the hierarchy view (perf_gtk__show_hierarchy()'s unbounded strcat() into a 512-byte buffer, and an advance_hpp() size_t underflow in perf_gtk__add_hierarchy_entries()). Both predate the GTK4 port
- Link to v6: https://lore.kernel.org/r/[email protected]

Changes in v6:
- Explicitly include <string.h> in annotate.c (strcpy()) and <stdarg.h>/<stdio.h> in hists.c (va_list, snprintf()) instead of relying on transitive includes, which isn't guaranteed on musl
- Link to v5: https://lore.kernel.org/r/[email protected]

Changes in v5:
- Fold v4's patch 3 into patch 1, since sigprocmask() only blocks delivery to the calling thread: it did nothing to stop the signal handler from running concurrently on another thread and doesn't make the handler's GSList calls async-signal-safe either
- Defer perf_gtk__exit() on SIGINT/SIGQUIT/SIGTERM to a GLib source via g_unix_signal_add() instead of running it straight out of a real signal handler, so it always runs on the main-loop thread, serialized with perf_gtk__error()'s updates to perf_gtk__error_loops, instead of racing them from arbitrary signal-handler context
- Keep a real handler for SIGSEGV/SIGFPE, since those are synchronous faults with no "later" to defer to, but pare it down to reporting and reraising the default disposition: there's no safe way to run
GTK/GLib code from the faulting context itself
- Link to v4: https://lore.kernel.org/r/[email protected]

Changes in v4:
- Fix gtk_widget_show()/gtk_widget_hide() calls left over from the GTK2
port: both were removed in GTK 4. Replace with gtk_widget_set_visible(), adding a small wrapper for the info-bar "response" signal callback
- Block SIGSEGV/SIGFPE/SIGINT/SIGQUIT/SIGTERM around the perf_gtk__error_loops list updates in perf_gtk__error(): the signal handler (perf_gtk__signal() -> perf_gtk__exit() -> perf_gtk__quit_error_dialog()) walks that same list and could fire mid-update, corrupting it
- Link to v3: https://lore.kernel.org/r/[email protected]

Changes in v3:
- Fix GMainLoop leak if perf_gtk__error() is called re-entrantly: track active loops in a list instead of a single global pointer
- Add explicit <stdarg.h>/<stdio.h> includes instead of relying on transitive inclusion, which musl doesn't guarantee
- Drop the gtk4-infobar feature check and HAVE_GTK_INFO_BAR_SUPPORT:
GtkInfoBar has been unconditionally available since well before GTK 4, and the check was failing outright due to its deprecation warning
- Link to v2: https://lore.kernel.org/r/[email protected]

Changes in v2:
- Fix error dialog's nested GMainLoop hanging if the parent window closes or a signal arrives while the dialog is open (quit from "destroy", not just "response")
- Fix build with GTK_INFO_BAR_SUPPORT: gtk_info_bar_get_content_area() is gone in GTK 4, use gtk_info_bar_add_child() instead
- Fix use-after-free in the progress dialog on manual close
- Fix reuse of an exhausted va_list in the vasprintf() failure path
- Link to v1: https://lore.kernel.org/r/[email protected]

---
Matt Turner (3):
tools: port perf ui from GTK 2 to GTK 4
perf tools: make the GTK4 report browser actually loadable at runtime perf tools gtk: fix two hierarchy-view stack buffer overflows

 tools/build/Makefile.feature                     |  4 +-
 tools/build/feature/Makefile                     | 10 +--
 tools/build/feature/test-gtk2-infobar.c          | 12 ---
 tools/build/feature/{test-gtk2.c => test-gtk4.c} |  4 +-
 tools/perf/Documentation/perf-report.txt         |  2 +-
 tools/perf/Makefile                              |  2 +-
 tools/perf/Makefile.config                       | 27 +++----
 tools/perf/Makefile.perf                         |  6 +-
 tools/perf/builtin-annotate.c                    |  8 +-
 tools/perf/builtin-report.c                      |  8 +-
 tools/perf/scripts/install-build-deps.sh         |  4 +-
 tools/perf/tests/make                            |  4 +-
 tools/perf/ui/gtk/annotate.c                     | 36 ++++-----
 tools/perf/ui/gtk/browser.c                      | 96 ++++++++++++++++++-----
 tools/perf/ui/gtk/gtk.h                          | 16 ++--
 tools/perf/ui/gtk/hists.c                        | 66 ++++++++--------
 tools/perf/ui/gtk/progress.c                     | 40 +++++++---
 tools/perf/ui/gtk/setup.c                        |  5 +-
 tools/perf/ui/gtk/util.c                         | 97 ++++++++++++++----------
 tools/perf/ui/setup.c                            |  2 +-
 tools/perf/util/annotate.c                       | 11 +++
 tools/perf/util/annotate.h                       | 12 +--
 22 files changed, 275 insertions(+), 197 deletions(-)

---
base-commit: 9f0346dcbea363787186c94ef94dd01aaa215afa
change-id: 20260906-perf-gtk2-555ca04bb652

Best regards,

Signature · 1 line
Matt Turner <[email protected]>
#1 lore raw

PATCH v7 1/3 tools: port perf ui from GTK 2 to GTK 4

Port straight to GTK 4 rather than GTK 3, since GTK 4 is where new development happens and GTK 3 is old itself now.

GTK 4 drops GtkContainer, GdkScreen, and the gtk_main()/ gtk_dialog_run() family perf's GTK UI relied on. Containers get per-widget setters (gtk_box_append() and friends), monitor geometry comes from GdkMonitor instead of GdkScreen, and the main and error-dialog loops become explicit GMainLoops quit from the "close-request" and "response" signals. Widgets are visible by default now, so gtk_widget_show_all()/set_no_show_all() go away, and the remaining gtk_widget_show()/gtk_widget_hide() calls become gtk_widget_set_visible() (with a small wrapper where "response" needs to pass gtk_widget_hide() as a callback, since it no longer exists as a plain function).

gtk_ui_progress__finish() skips destroying a progress dialog that was never created, since gtk_window_destroy() asserts on NULL where the old widget destroy tolerated it. Two spots the GTK 2 to GTK 3 port had missed (builtin-annotate.c, ui/gtk/setup.c still using HAVE_GTK2_SUPPORT and gtk_main_quit()) are fixed to match.

Runtime fallout from the new signal-driven loops: the error dialog's nested loop hung if the parent window closed (GTK_DIALOG_DESTROY_WITH_PARENT destroys without emitting "response"); gtk_info_bar_get_content_area() is gone, breaking GTK_INFO_BAR_SUPPORT; the progress dialog's static widget pointers dangled after a manual close; perf_gtk__error() and the warning functions reused an exhausted va_list when vasprintf() failed.

The error loop is tracked in a list instead of a single pointer, since perf_gtk__error() can be called re-entrantly (the dialog isn't modal) and a lone global leaked the outer loop when that happened. The list is only ever touched from the main thread: perf_gtk__error() updates it while handling a dialog, and SIGINT/SIGQUIT/SIGTERM are deferred to a GLib source via g_unix_signal_add() rather than calling perf_gtk__exit() straight out of a real signal handler, so quitting on those signals is serialized with the list update instead of racing it from signal-handler context. SIGSEGV/SIGFPE keep a real handler, since they're synchronous faults with no "later" to defer to, but it's pared down to reporting and reraising the default disposition (perf_gtk__fatal_signal()): there's no safe way to run GTK/GLib code from the faulting context. stdarg.h, stdio.h, and string.h are now included explicitly where used (util.c, hists.c, annotate.c) rather than relying on transitive includes, which musl doesn't guarantee.

The gtk4-infobar feature check is dropped: GtkInfoBar has existed unconditionally since GTK 3.10, so the check can only ever pass, and it was failing outright here anyway since gtk_info_bar_new() is deprecated and the check treats deprecation warnings as errors. HAVE_GTK_INFO_BAR_SUPPORT and its statusbar-only fallback go away; the info bar is now built unconditionally.

---
 tools/build/Makefile.feature                     |  4 +-
 tools/build/feature/Makefile                     | 10 +--
 tools/build/feature/test-gtk2-infobar.c          | 12 ---
 tools/build/feature/{test-gtk2.c => test-gtk4.c} |  4 +-
 tools/perf/Documentation/perf-report.txt         |  2 +-
 tools/perf/Makefile                              |  2 +-
 tools/perf/Makefile.config                       | 22 +++---
 tools/perf/Makefile.perf                         |  6 +-
 tools/perf/builtin-annotate.c                    |  8 +-
 tools/perf/builtin-report.c                      |  8 +-
 tools/perf/scripts/install-build-deps.sh         |  4 +-
 tools/perf/tests/make                            |  4 +-
 tools/perf/ui/gtk/annotate.c                     | 36 ++++-----
 tools/perf/ui/gtk/browser.c                      | 96 ++++++++++++++++++-----
 tools/perf/ui/gtk/gtk.h                          | 16 ++--
 tools/perf/ui/gtk/hists.c                        | 43 +++++------
 tools/perf/ui/gtk/progress.c                     | 40 +++++++---
 tools/perf/ui/gtk/setup.c                        |  5 +-
 tools/perf/ui/gtk/util.c                         | 97 ++++++++++++++----------
 tools/perf/ui/setup.c                            |  2 +-
 20 files changed, 239 insertions(+), 182 deletions(-)
tools/build/Makefile.feature +2 -2
index 99eb0ea09537..4ec95c35a5c1 100644
@@ -113,8 +113,7 @@ FEATURE_TESTS_EXTRA :=                  \
          compile-x32                    \
          cplus-demangle                 \
          cxa-demangle                   \
-         gtk2                           \
-         gtk2-infobar                   \
+         gtk4                           \
          hello                          \
          babeltrace2-ctf-writer         \
          libcapstone                    \
@@ -143,6 +142,7 @@ endif
 FEATURE_DISPLAY ?=              \
          libdw                  \
          glibc                  \
+         gtk4                   \
          libelf                 \
          libnuma                \
          numa_num_possible_cpus \
tools/build/feature/Makefile +3 -7
index 7d165018116a..01c48e6ef021 100644
@@ -11,8 +11,7 @@ FILES=                                          \
          test-eventfd.bin                       \
          test-fortify-source.bin                \
          test-glibc.bin                         \
-         test-gtk2.bin                          \
-         test-gtk2-infobar.bin                  \
+         test-gtk4.bin                          \
          test-hello.bin                         \
          test-libbfd.bin                        \
 	 test-libbfd-threadsafe.bin      	\
@@ -240,11 +239,8 @@ $(OUTPUT)test-libcpupower.bin:
 $(OUTPUT)test-libtracefs.bin:
 	 $(BUILD) $(shell $(PKG_CONFIG) --cflags libtracefs 2>/dev/null) -ltracefs
 
-$(OUTPUT)test-gtk2.bin:
-	$(BUILD) $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null) -Wno-deprecated-declarations
-
-$(OUTPUT)test-gtk2-infobar.bin:
-	$(BUILD) $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null)
+$(OUTPUT)test-gtk4.bin:
+	$(BUILD) $(shell $(PKG_CONFIG) --libs --cflags gtk4 2>/dev/null)
 
 grep-libs  = $(filter -l%,$(1))
 strip-libs = $(filter-out -l%,$(1))
tools/build/feature/test-gtk2-infobar.c deleted +0 -12
deleted file mode 100644
index b1b716dd5733..000000000000
@@ -1,12 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0
-#pragma GCC diagnostic ignored "-Wstrict-prototypes"
-#include <gtk/gtk.h>
-#pragma GCC diagnostic error "-Wstrict-prototypes"
-
-int main(int argc, char *argv[])
-{
-	gtk_init(&argc, &argv);
-	gtk_info_bar_new();
-
-	return 0;
-}
tools/build/feature/test-gtk4.c renamed from tools/build/feature/test-gtk2.c +2 -2
similarity index 76%
rename from tools/build/feature/test-gtk2.c
rename to tools/build/feature/test-gtk4.c
index 2aaf4bfc2055..b9520e7408b9 100644
@@ -3,9 +3,9 @@
 #include <gtk/gtk.h>
 #pragma GCC diagnostic error "-Wstrict-prototypes"
 
-int main(int argc, char *argv[])
+int main(void)
 {
-	gtk_init(&argc, &argv);
+	gtk_init();
 
         return 0;
 }
tools/perf/Documentation/perf-report.txt +1 -1
index 22f87eaa3279..7af9b3f81c06 100644
@@ -351,7 +351,7 @@ OPTIONS
 	requires a tty, if one is not present, as when piping to other
 	commands, the stdio interface is used.
 
---gtk:: Use the GTK2 interface.
+--gtk:: Use the GTK4 interface.
 
 -k::
 --vmlinux=<file>::
tools/perf/Makefile +1 -1
index 5b713837eede..56014106479a 100644
@@ -106,7 +106,7 @@ clean:
 # make -C tools/perf -f tests/make
 #
 build-test:
-	@$(MAKE) SHUF=1 -f tests/make REUSE_FEATURES_DUMP=1 MK=Makefile SET_PARALLEL=1 --no-print-directory tarpkg make_static make_with_gtk2 out
+	@$(MAKE) SHUF=1 -f tests/make REUSE_FEATURES_DUMP=1 MK=Makefile SET_PARALLEL=1 --no-print-directory tarpkg make_static make_with_gtk4 out
 
 build-test-tarball:
 	@$(MAKE) -f tests/make REUSE_FEATURES_DUMP=1 MK=Makefile SET_PARALLEL=1 --no-print-directory out
tools/perf/Makefile.config +9 -13
index 4d5993da9f94..3e59e2b7eaec 100644
@@ -774,20 +774,16 @@ ifndef NO_SLANG
   endif
 endif
 
-ifdef GTK2
-  FLAGS_GTK2=$(CFLAGS) $(LDFLAGS) $(EXTLIBS) $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null)
-  $(call feature_check,gtk2)
-  ifneq ($(feature-gtk2), 1)
-    $(warning GTK2 not found, disables GTK2 support. Please install gtk2-devel or libgtk2.0-dev)
-    NO_GTK2 := 1
+ifdef GTK4
+  FLAGS_GTK4=$(CFLAGS) $(LDFLAGS) $(EXTLIBS) $(shell $(PKG_CONFIG) --libs --cflags gtk4 2>/dev/null)
+  $(call feature_check,gtk4)
+  ifneq ($(feature-gtk4), 1)
+    $(warning GTK4 not found, disables GTK4 support. Please install gtk4-devel or libgtk-4-dev)
+    NO_GTK4 := 1
   else
-    $(call feature_check,gtk2-infobar)
-    ifeq ($(feature-gtk2-infobar), 1)
-      GTK_CFLAGS := -DHAVE_GTK_INFO_BAR_SUPPORT
-    endif
-    CFLAGS += -DHAVE_GTK2_SUPPORT
-    GTK_CFLAGS += $(shell $(PKG_CONFIG) --cflags gtk+-2.0 2>/dev/null)
-    GTK_LIBS := $(shell $(PKG_CONFIG) --libs gtk+-2.0 2>/dev/null)
+    CFLAGS += -DHAVE_GTK4_SUPPORT
+    GTK_CFLAGS += $(shell $(PKG_CONFIG) --cflags gtk4 2>/dev/null)
+    GTK_LIBS := $(shell $(PKG_CONFIG) --libs gtk4 2>/dev/null)
     EXTLIBS += -ldl
   endif
 endif
tools/perf/Makefile.perf +3 -3
index 29cfd44c427f..2438b40eaaec 100644
@@ -49,7 +49,7 @@ include ../scripts/utilities.mak
 #
 # Define NO_SLANG if you do not want TUI support.
 #
-# Define GTK2 if you want GTK+ GUI support.
+# Define GTK4 if you want GTK+ GUI support.
 #
 # Define NO_DEMANGLE if you do not want C++ symbol demangling.
 #
@@ -473,7 +473,7 @@ ifneq ($(OUTPUT),)
   CFLAGS += -I$(OUTPUT)
 endif
 
-ifdef GTK2
+ifdef GTK4
   ALL_PROGRAMS += $(OUTPUT)libperf-gtk.so
   GTK_IN := $(OUTPUT)gtk-in.o
 endif
@@ -811,7 +811,7 @@ check: prepare
 
 ### Installation rules
 
-ifdef GTK2
+ifdef GTK4
 install-gtk: $(OUTPUT)libperf-gtk.so
 	$(call QUIET_INSTALL, 'GTK UI') \
 		$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(libdir_SQ)'; \
tools/perf/builtin-annotate.c +4 -4
index 69cb72b2082a..15163e081a8c 100644
@@ -52,7 +52,7 @@ struct perf_annotate {
 	bool	   use_tui;
 #endif
 	bool	   use_stdio, use_stdio2;
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 	bool	   use_gtk;
 #endif
 	bool	   skip_missing;
@@ -712,7 +712,7 @@ int cmd_annotate(int argc, const char **argv)
 	OPT_BOOLEAN('q', "quiet", &quiet, "do now show any warnings or messages"),
 	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
 		    "dump raw trace in ASCII"),
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 	OPT_BOOLEAN(0, "gtk", &annotate.use_gtk, "Use the GTK interface"),
 #endif
 #ifdef HAVE_SLANG_SUPPORT
@@ -828,7 +828,7 @@ int cmd_annotate(int argc, const char **argv)
 	if (annotate_check_args() < 0)
 		return -EINVAL;
 
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 	if (symbol_conf.show_nr_samples && annotate.use_gtk) {
 		pr_err("--show-nr-samples is not available in --gtk mode at this time\n");
 		return ret;
@@ -898,7 +898,7 @@ int cmd_annotate(int argc, const char **argv)
 	else if (annotate.use_tui)
 		use_browser = 1;
 #endif
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 	else if (annotate.use_gtk)
 		use_browser = 2;
 #endif
tools/perf/builtin-report.c +4 -4
index 60d1f166629e..d14384c58466 100644
@@ -82,7 +82,7 @@ struct report {
 #ifdef HAVE_SLANG_SUPPORT
 	bool			use_tui;
 #endif
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 	bool			use_gtk;
 #endif
 	bool			use_stdio;
@@ -1359,8 +1359,8 @@ int cmd_report(int argc, const char **argv)
 #ifdef HAVE_SLANG_SUPPORT
 	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
 #endif
-#ifdef HAVE_GTK2_SUPPORT
-	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
+#ifdef HAVE_GTK4_SUPPORT
+	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK4 interface"),
 #endif
 	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
 		    "Use the stdio interface"),
@@ -1710,7 +1710,7 @@ int cmd_report(int argc, const char **argv)
 	else if (report.use_tui)
 		use_browser = 1;
 #endif
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 	else if (report.use_gtk)
 		use_browser = 2;
 #endif
tools/perf/scripts/install-build-deps.sh +2 -2
index d003e7fab2be..a601a5260c17 100755
@@ -199,8 +199,8 @@ fedora_pkg_for() {
 	# opt-in features, which a default build does not enable: the libbfd
 	# disassembler family (libbfd, libbfd-threadsafe, libbfd-liberty,
 	# disassembler-*, cplus-demangle), only linked on BUILD_NONDISTRO
-	# builds and deprecated in favor of capstone, GTK2, LIBPERL and
-	# LIBUNWIND support (ifdef GTK2 / ifdef LIBPERL / LIBUNWIND=1),
+	# builds and deprecated in favor of capstone, GTK4, LIBPERL and
+	# LIBUNWIND support (ifdef GTK4 / ifdef LIBPERL / LIBUNWIND=1),
 	# and CoreSight (ifdef CORESIGHT), are deliberately not mapped.
 	# libaio is not mapped either: its
 	# test uses the POSIX AIO API (aio.h, aio_*, -lrt), provided by
tools/perf/tests/make +2 -2
index d2c2f526e1db..202ab5501916 100644
@@ -95,7 +95,7 @@ make_no_babeltrace2 := NO_BABELTRACE2=1
 make_with_coresight := CORESIGHT=1
 make_no_sdt	    := NO_SDT=1
 make_no_libpfm4     := NO_LIBPFM4=1
-make_with_gtk2      := GTK2=1
+make_with_gtk4      := GTK4=1
 make_refcnt_check   := EXTRA_CFLAGS="-DREFCNT_CHECKING=1"
 make_tags           := tags
 make_cscope         := cscope
@@ -318,7 +318,7 @@ $(run):
 	$(call test,$@) && \
 	rm -rf $@ $$TMP_DEST || (cat $@ ; false)
 
-make_with_gtk2:
+make_with_gtk4:
 	$(call clean)
 	@TMP_DEST=$$(mktemp -d); \
 	cmd="cd $(PERF) && $(MAKE_F) $($@) $(PARALLEL_OPT) $(O_OPT) DESTDIR=$$TMP_DEST"; \
tools/perf/ui/gtk/annotate.c +15 -21
index 8920e298420a..094dfa045b77 100644
@@ -11,6 +11,7 @@
 #include "ui/helpline.h"
 #include <inttypes.h>
 #include <signal.h>
+#include <string.h>
 
 enum {
 	ANN_COL__PERCENT,
@@ -161,7 +162,7 @@ static int perf_gtk__annotate_symbol(GtkWidget *window, struct map_symbol *ms,
 			gtk_list_store_set(store, &iter, ANN_COL__LINE, s, -1);
 	}
 
-	gtk_container_add(GTK_CONTAINER(window), view);
+	gtk_scrolled_window_set_child(GTK_SCROLLED_WINDOW(window), view);
 
 	list_for_each_entry_safe(pos, n, &notes->src->source, al.node) {
 		list_del_init(&pos->al.node);
@@ -205,40 +206,34 @@ static int symbol__gtk_annotate(struct map_symbol *ms, struct evsel *evsel,
 		GtkWidget *infobar;
 		GtkWidget *statbar;
 
-		signal(SIGSEGV, perf_gtk__signal);
-		signal(SIGFPE,  perf_gtk__signal);
-		signal(SIGINT,  perf_gtk__signal);
-		signal(SIGQUIT, perf_gtk__signal);
-		signal(SIGTERM, perf_gtk__signal);
+		signal(SIGSEGV, perf_gtk__fatal_signal);
+		signal(SIGFPE,  perf_gtk__fatal_signal);
+		perf_gtk__install_quit_signals();
 
-		window = gtk_window_new(GTK_WINDOW_TOPLEVEL);
+		window = gtk_window_new();
 		gtk_window_set_title(GTK_WINDOW(window), "perf annotate");
 
-		g_signal_connect(window, "delete_event", gtk_main_quit, NULL);
-
 		pgctx = perf_gtk__activate_context(window);
 		if (!pgctx)
 			return -1;
 
-		vbox = gtk_vbox_new(FALSE, 0);
+		vbox = gtk_box_new(GTK_ORIENTATION_VERTICAL, 0);
 		notebook = gtk_notebook_new();
 		pgctx->notebook = notebook;
 
-		gtk_box_pack_start(GTK_BOX(vbox), notebook, TRUE, TRUE, 0);
+		gtk_widget_set_vexpand(notebook, TRUE);
+		gtk_box_append(GTK_BOX(vbox), notebook);
 
 		infobar = perf_gtk__setup_info_bar();
-		if (infobar) {
-			gtk_box_pack_start(GTK_BOX(vbox), infobar,
-					   FALSE, FALSE, 0);
-		}
+		gtk_box_append(GTK_BOX(vbox), infobar);
 
 		statbar = perf_gtk__setup_statusbar();
-		gtk_box_pack_start(GTK_BOX(vbox), statbar, FALSE, FALSE, 0);
+		gtk_box_append(GTK_BOX(vbox), statbar);
 
-		gtk_container_add(GTK_CONTAINER(window), vbox);
+		gtk_window_set_child(GTK_WINDOW(window), vbox);
 	}
 
-	scrolled_window = gtk_scrolled_window_new(NULL, NULL);
+	scrolled_window = gtk_scrolled_window_new();
 	tab_label = gtk_label_new(sym->name);
 
 	gtk_scrolled_window_set_policy(GTK_SCROLLED_WINDOW(scrolled_window),
@@ -267,12 +262,11 @@ void perf_gtk__show_annotations(void)
 		return;
 
 	window = pgctx->main_window;
-	gtk_widget_show_all(window);
 
 	perf_gtk__resize_window(window);
-	gtk_window_set_position(GTK_WINDOW(window), GTK_WIN_POS_CENTER);
+	gtk_widget_set_visible(window, TRUE);
 
-	gtk_main();
+	perf_gtk__run_main_loop(window);
 
 	perf_gtk__deactivate_context(&pgctx);
 }
tools/perf/ui/gtk/browser.c +79 -17

144 lines not on this page · raw message

tools/perf/ui/gtk/gtk.h +5 -11
index a2b497f03fd6..6b37fd9943c9 100644
@@ -13,10 +13,8 @@ struct perf_gtk_context {
 	GtkWidget *main_window;
 	GtkWidget *notebook;
 
-#ifdef HAVE_GTK_INFO_BAR_SUPPORT
 	GtkWidget *info_bar;
 	GtkWidget *message_label;
-#endif
 	GtkWidget *statbar;
 	guint statbar_ctx_id;
 };
@@ -38,19 +36,15 @@ void perf_gtk__init_helpline(void);
 void gtk_ui_progress__init(void);
 void perf_gtk__init_hpp(void);
 
-void perf_gtk__signal(int sig);
+void perf_gtk__install_quit_signals(void);
+void perf_gtk__fatal_signal(int sig);
 void perf_gtk__resize_window(GtkWidget *window);
+void perf_gtk__run_main_loop(GtkWidget *window);
+void perf_gtk__quit_main_loop(void);
+void perf_gtk__quit_error_dialog(void);
 const char *perf_gtk__get_percent_color(double percent);
 GtkWidget *perf_gtk__setup_statusbar(void);
-
-#ifdef HAVE_GTK_INFO_BAR_SUPPORT
 GtkWidget *perf_gtk__setup_info_bar(void);
-#else
-static inline GtkWidget *perf_gtk__setup_info_bar(void)
-{
-	return NULL;
-}
-#endif
 
 struct evsel;
 struct evlist;
tools/perf/ui/gtk/hists.c +17 -26

103 lines not on this page · raw message

tools/perf/ui/gtk/progress.c +28 -12
index eea6fcde518a..770f9251b54b 100644
@@ -1,49 +1,65 @@
 // SPDX-License-Identifier: GPL-2.0
 #include <inttypes.h>
+#include <stdio.h>
 
 #include "gtk.h"
 #include "../progress.h"
+#include <linux/compiler.h>
 
 static GtkWidget *dialog;
 static GtkWidget *progress;
 
+static void gtk_ui_progress__destroyed(GtkWidget *widget __maybe_unused,
+					gpointer data __maybe_unused)
+{
+	dialog = NULL;
+	progress = NULL;
+}
+
 static void gtk_ui_progress__update(struct ui_progress *p)
 {
 	double fraction = p->total ? 1.0 * p->curr / p->total : 0.0;
 	char buf[1024];
 
 	if (dialog == NULL) {
-		GtkWidget *vbox = gtk_vbox_new(TRUE, 5);
+		GtkWidget *vbox = gtk_box_new(GTK_ORIENTATION_VERTICAL, 5);
 		GtkWidget *label = gtk_label_new(p->title);
 
-		dialog = gtk_window_new(GTK_WINDOW_TOPLEVEL);
+		dialog = gtk_window_new();
 		progress = gtk_progress_bar_new();
 
-		gtk_box_pack_start(GTK_BOX(vbox), label, TRUE, FALSE, 3);
-		gtk_box_pack_start(GTK_BOX(vbox), progress, TRUE, TRUE, 3);
+		gtk_widget_set_vexpand(label, TRUE);
+		gtk_box_append(GTK_BOX(vbox), label);
+		gtk_widget_set_vexpand(progress, TRUE);
+		gtk_box_append(GTK_BOX(vbox), progress);
 
-		gtk_container_add(GTK_CONTAINER(dialog), vbox);
+		gtk_window_set_child(GTK_WINDOW(dialog), vbox);
+
+		g_signal_connect(dialog, "destroy",
+				 G_CALLBACK(gtk_ui_progress__destroyed), NULL);
 
 		gtk_window_set_title(GTK_WINDOW(dialog), "perf");
-		gtk_window_resize(GTK_WINDOW(dialog), 300, 80);
-		gtk_window_set_position(GTK_WINDOW(dialog), GTK_WIN_POS_CENTER);
+		gtk_window_set_default_size(GTK_WINDOW(dialog), 300, 80);
 
-		gtk_widget_show_all(dialog);
+		gtk_widget_set_visible(dialog, TRUE);
 	}
 
 	gtk_progress_bar_set_fraction(GTK_PROGRESS_BAR(progress), fraction);
 	snprintf(buf, sizeof(buf), "%"PRIu64" / %"PRIu64, p->curr, p->total);
 	gtk_progress_bar_set_text(GTK_PROGRESS_BAR(progress), buf);
 
-	/* we didn't call gtk_main yet, so do it manually */
-	while (gtk_events_pending())
-		gtk_main_iteration();
+	/* we didn't start a main loop yet, so pump events manually */
+	while (g_main_context_pending(NULL))
+		g_main_context_iteration(NULL, FALSE);
 }
 
 static void gtk_ui_progress__finish(void)
 {
+	if (dialog == NULL)
+		return;
+
 	/* this will also destroy all of its children */
-	gtk_widget_destroy(dialog);
+	gtk_window_destroy(GTK_WINDOW(dialog));
 
 	dialog = NULL;
 }
tools/perf/ui/gtk/setup.c +3 -2
index f5eee4d66873..9b44f3719747 100644
@@ -12,7 +12,7 @@ int perf_gtk__init(void)
 	gtk_ui_progress__init();
 	perf_gtk__init_hpp();
 
-	return gtk_init_check(NULL, NULL) ? 0 : -1;
+	return gtk_init_check() ? 0 : -1;
 }
 
 void perf_gtk__exit(bool wait_for_ok __maybe_unused)
@@ -20,5 +20,6 @@ void perf_gtk__exit(bool wait_for_ok __maybe_unused)
 	if (!perf_gtk__is_active_context(pgctx))
 		return;
 	perf_error__unregister(&perf_gtk_eops);
-	gtk_main_quit();
+	perf_gtk__quit_error_dialog();
+	perf_gtk__quit_main_loop();
 }
tools/perf/ui/gtk/util.c +58 -39

152 lines not on this page · raw message

tools/perf/ui/setup.c +1 -1
index ff800047e697..d887346c7a63 100644
@@ -14,7 +14,7 @@ int use_browser = -1;
 
 #define PERF_GTK_DSO "libperf-gtk.so"
 
-#ifdef HAVE_GTK2_SUPPORT
+#ifdef HAVE_GTK4_SUPPORT
 
 static int setup_gtk_browser(void)
 {
Signature · 1 line
2.54.0
#2 lore raw

Re: [PATCH v7 1/3] tools: port perf ui from GTK 2 to GTK 4

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Connecting to the removed 'destroy' signal in GTK4 causes main loop hangs and…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Connecting to the removed 'destroy' signal in GTK4 causes main loop hangs and use-after-free/double-free bugs. --

commit fc78f437c6f22aa072d608cee2f291c6178f6853
Author: Matt Turner <[email protected]>

tools: port perf ui from GTK 2 to GTK 4

Port straight to GTK 4 rather than GTK 3, since GTK 4 is where new development happens and GTK 3 is old itself now.

Quoted 8 lines
tools/perf/ui/gtk/progress.c +1 -0
index eea6fcde518a8..770f9251b54b7 100644
@@ -1,49 +1,65 @@
 // SPDX-License-Identifier: GPL-2.0
 #include <inttypes.h>
+#include <stdio.h>

[ ... ]

Quoted 23 lines
tools/perf/ui/gtk/progress.c +9 -4
+static void gtk_ui_progress__destroyed(GtkWidget *widget __maybe_unused,
+					gpointer data __maybe_unused)
+{
+	dialog = NULL;
+	progress = NULL;
+}
+
 static void gtk_ui_progress__update(struct ui_progress *p)
 {
 	double fraction = p->total ? 1.0 * p->curr / p->total : 0.0;
 	char buf[1024];
 
 	if (dialog == NULL) {
-		GtkWidget *vbox = gtk_vbox_new(TRUE, 5);
+		GtkWidget *vbox = gtk_box_new(GTK_ORIENTATION_VERTICAL, 5);
 		GtkWidget *label = gtk_label_new(p->title);
 
-		dialog = gtk_window_new(GTK_WINDOW_TOPLEVEL);
+		dialog = gtk_window_new();
 		progress = gtk_progress_bar_new();
 
-		gtk_box_pack_start(GTK_BOX(vbox), label, TRUE, FALSE, 3);
-		gtk_box_pack_start(GTK_BOX(vbox), progress, TRUE, TRUE, 3);

tools/perf/ui/gtk/progress.c

Quoted 10 lines
tools/perf/ui/gtk/progress.c +8 -1
+		gtk_widget_set_vexpand(label, TRUE);
+		gtk_box_append(GTK_BOX(vbox), label);
+		gtk_widget_set_vexpand(progress, TRUE);
+		gtk_box_append(GTK_BOX(vbox), progress);
+
+		gtk_window_set_child(GTK_WINDOW(dialog), vbox);
 
-		gtk_container_add(GTK_CONTAINER(dialog), vbox);
+		g_signal_connect(dialog, "destroy",
+				 G_CALLBACK(gtk_ui_progress__destroyed), NULL);

[Severity: Medium]
In gtk_ui_progress__update(), will this signal connection fail at runtime? The destroy signal was removed from GtkWidget in GTK 4. If it fails silently, gtk_ui_progress__destroyed() would not be called to nullify the pointer when the user manually closes the progress window. This could result in a use-after-free or double-free crash when gtk_ui_progress__finish() is subsequently called on the already-freed pointer.

[ ... ]

Quoted 5 lines
tools/perf/ui/gtk/util.c +0 -0
index c47f5c387838c..5823ff0d7c2a7 100644
@@ -28,86 +30,103 @@ int perf_gtk__deactivate_context(struct perf_gtk_context **ctx)

[ ... ]

tools/perf/ui/gtk/util.c int perf_gtk__deactivate_context(struct perf_gtk_context **ctx) · line 93

Quoted 12 lines
tools/perf/ui/gtk/util.c +12 -0
+	/*
+	 * "response" only fires when a button is clicked; DESTROY_WITH_PARENT
+	 * destroys the dialog directly without it. Quit from "destroy"
+	 * instead, which fires either way, so the nested loop below can't
+	 * outlive the dialog and hang.
+	 */
+	loop = g_main_loop_new(NULL, FALSE);
+	perf_gtk__error_loops = g_slist_prepend(perf_gtk__error_loops, loop);
+	g_signal_connect(dialog, "response",
+			 G_CALLBACK(perf_gtk__dialog_response), NULL);
+	g_signal_connect_swapped(dialog, "destroy",
+				 G_CALLBACK(g_main_loop_quit), loop);

[Severity: Medium]
In perf_gtk__error(), since the destroy signal no longer exists in GTK 4, could this leave the nested main loop running indefinitely and hanging the application? Because the destroy signal is not emitted when the dialog is closed (even via the confirmation button), g_main_loop_quit() will never be called.

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#3 lore raw

Re: [PATCH v7 1/3] tools: port perf ui from GTK 2 to GTK 4

Quoted 54 lines · On Sun, Sep 6, 2026 at 4:25 PM <[email protected]> wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Connecting to the removed 'destroy' signal in GTK4 causes main loop hangs and use-after-free/double-free bugs. --

commit fc78f437c6f22aa072d608cee2f291c6178f6853
Author: Matt Turner <[email protected]>

tools: port perf ui from GTK 2 to GTK 4

Port straight to GTK 4 rather than GTK 3, since GTK 4 is where new development happens and GTK 3 is old itself now.

Quoted 8 lines
tools/perf/ui/gtk/progress.c +1 -0
index eea6fcde518a8..770f9251b54b7 100644
@@ -1,49 +1,65 @@
 // SPDX-License-Identifier: GPL-2.0
 #include <inttypes.h>
+#include <stdio.h>

[ ... ]

Quoted 30 lines
tools/perf/ui/gtk/progress.c +15 -4
+static void gtk_ui_progress__destroyed(GtkWidget *widget __maybe_unused,
+                                     gpointer data __maybe_unused)
+{
+     dialog = NULL;
+     progress = NULL;
+}
+
 static void gtk_ui_progress__update(struct ui_progress *p)
 {
      double fraction = p->total ? 1.0 * p->curr / p->total : 0.0;
      char buf[1024];
 
      if (dialog == NULL) {
-             GtkWidget *vbox = gtk_vbox_new(TRUE, 5);
+             GtkWidget *vbox = gtk_box_new(GTK_ORIENTATION_VERTICAL, 5);
              GtkWidget *label = gtk_label_new(p->title);
 
-             dialog = gtk_window_new(GTK_WINDOW_TOPLEVEL);
+             dialog = gtk_window_new();
              progress = gtk_progress_bar_new();
 
-             gtk_box_pack_start(GTK_BOX(vbox), label, TRUE, FALSE, 3);
-             gtk_box_pack_start(GTK_BOX(vbox), progress, TRUE, TRUE, 3);
+             gtk_widget_set_vexpand(label, TRUE);
+             gtk_box_append(GTK_BOX(vbox), label);
+             gtk_widget_set_vexpand(progress, TRUE);
+             gtk_box_append(GTK_BOX(vbox), progress);
+
+             gtk_window_set_child(GTK_WINDOW(dialog), vbox);
Quoted 11 lines

tools/perf/ui/gtk/progress.c line 39

Quoted 3 lines
tools/perf/ui/gtk/progress.c +2 -1
-             gtk_container_add(GTK_CONTAINER(dialog), vbox);
+             g_signal_connect(dialog, "destroy",
+                              G_CALLBACK(gtk_ui_progress__destroyed), NULL);

[Severity: Medium]
In gtk_ui_progress__update(), will this signal connection fail at runtime? The destroy signal was removed from GtkWidget in GTK 4. If it fails silently, gtk_ui_progress__destroyed() would not be called to nullify the pointer when the user manually closes the progress window. This could result in a use-after-free or double-free crash when gtk_ui_progress__finish() is subsequently called on the already-freed pointer.

Incorrect. GTK4 did not remove the destroy signal from GtkWidget, only the gtk_widget_destroy() function was removed. The signal is still registered in gtk/gtkwidget.c (g_signal_new(I_("destroy"), ...)) and emitted from gtk_widget_dispose(). Verified against upstream GTK source, and confirmed live via GObject.signal_list_names(Gtk.Widget) on GTK 4.20.4: destroy is present. g_signal_connect(dialog, "destroy", ...) connects fine and fires. No change needed.

Quoted 23 lines

[ ... ]

Quoted 5 lines
tools/perf/ui/gtk/util.c +0 -0
index c47f5c387838c..5823ff0d7c2a7 100644
@@ -28,86 +30,103 @@ int perf_gtk__deactivate_context(struct perf_gtk_context **ctx)

[ ... ]

tools/perf/ui/gtk/util.c int perf_gtk__deactivate_context(struct perf_gtk_context **ctx) · line 93

Quoted 12 lines
tools/perf/ui/gtk/util.c +12 -0
+     /*
+      * "response" only fires when a button is clicked; DESTROY_WITH_PARENT
+      * destroys the dialog directly without it. Quit from "destroy"
+      * instead, which fires either way, so the nested loop below can't
+      * outlive the dialog and hang.
+      */
+     loop = g_main_loop_new(NULL, FALSE);
+     perf_gtk__error_loops = g_slist_prepend(perf_gtk__error_loops, loop);
+     g_signal_connect(dialog, "response",
+                      G_CALLBACK(perf_gtk__dialog_response), NULL);
+     g_signal_connect_swapped(dialog, "destroy",
+                              G_CALLBACK(g_main_loop_quit), loop);
Quoted 6 lines

[Severity: Medium]
In perf_gtk__error(), since the destroy signal no longer exists in GTK 4, could this leave the nested main loop running indefinitely and hanging the application? Because the destroy signal is not emitted when the dialog is closed (even via the confirmation button), g_main_loop_quit() will never be called.

Also incorrect, same misunderstanding of the destroy signal. When the user closes the dialog via the window controls, gtk_window_close() calls gtk_window_emit_close_request(). The default close_request handler returns FALSE (no hide_on_close set), so control falls through to gtk_window_destroy(), which drops GTK's internal toplevel ref and finalizes the widget, emitting destroy. That is exactly the signal perf_gtk__error() connects to stop the nested loop. Confirmed by reading gtk_window_close() and gtk_window_destroy() in gtkwindow.c, and gtk_widget_dispose() in gtkwidget.c, in current GTK upstream. No hang. No change needed.

#4 lore raw

PATCH v7 2/3 perf tools: make the GTK4 report browser actually loadable at runtime

In reply to Matt Turner

perf report --gtk dlopen()s libperf-gtk.so, which expects to resolve symbols back against the running perf binary (callchain_param, symbol_conf, evsel__name, and friends live in perf, not the plugin). Two things broke that after the GTK 4 port:

perf never passed -rdynamic, so none of its symbols were in its dynamic symbol table for a dlopen()ed plugin to find. Add -rdynamic to LDFLAGS when GTK4 support is enabled.

annotated_source__hist_entry() was a static inline in annotate.h, so ui/gtk/annotate.c calling it pulled hashmap__find()'s expansion, hashmap_find(), into libperf-gtk.so as an undefined symbol. The only hashmap_find perf links against normally is libbpf's internal one (tools/lib/bpf/hashmap.c), built with -fvisibility=hidden, so it can never be exported to a dlopen()ed plugin regardless of LDFLAGS. Move annotated_source__hist_entry() into annotate.c as an ordinary exported function, so the plugin depends on it the same way it already depends on evsel__group_desc() and friends.

With both fixes, a default 'make GTK4=1' build (libbpf statically linked) can dlopen() libperf-gtk.so and open the report browser without NO_LIBBPF=1 or manual LDFLAGS. Verified with perf report --gtk against real perf.data on a GTK4 desktop.

---
 tools/perf/Makefile.config |  5 +++++
 tools/perf/util/annotate.c | 11 +++++++++++
 tools/perf/util/annotate.h | 12 ++----------
 3 files changed, 18 insertions(+), 10 deletions(-)
tools/perf/Makefile.config +5 -0
index 3e59e2b7eaec..4ee7393a39f9 100644
@@ -785,6 +785,11 @@ ifdef GTK4
     GTK_CFLAGS += $(shell $(PKG_CONFIG) --cflags gtk4 2>/dev/null)
     GTK_LIBS := $(shell $(PKG_CONFIG) --libs gtk4 2>/dev/null)
     EXTLIBS += -ldl
+    # libperf-gtk.so is dlopen()ed at runtime and calls back into
+    # symbols defined in the perf binary itself (callchain_param,
+    # symbol_conf, evsel__name, ...): perf needs to export those
+    # dynamically for the plugin to resolve them.
+    LDFLAGS += -rdynamic
   endif
 endif
 
tools/perf/util/annotate.c +11 -0
index df70e95a8470..123b6d5fcea7 100644
@@ -145,6 +145,17 @@ static int annotated_source__alloc_histograms(struct annotated_source *src,
 	return src->histograms ? 0 : -1;
 }
 
+struct sym_hist_entry *
+annotated_source__hist_entry(struct annotated_source *src, const struct evsel *evsel, u64 offset)
+{
+	struct sym_hist_entry *entry;
+	long key = offset << 16 | evsel->core.idx;
+
+	if (!hashmap__find(src->samples, key, &entry))
+		return NULL;
+	return entry;
+}
+
 void symbol__annotate_zero_histograms(struct symbol *sym)
 {
 	struct annotation *notes = symbol__annotation(sym);
tools/perf/util/annotate.h +2 -10
index fa08d09b80f7..40038a3779d4 100644
@@ -406,16 +406,8 @@ static inline struct sym_hist *annotation__histogram(struct annotation *notes,
 	return annotated_source__histogram(notes->src, evsel);
 }
 
-static inline struct sym_hist_entry *
-annotated_source__hist_entry(struct annotated_source *src, const struct evsel *evsel, u64 offset)
-{
-	struct sym_hist_entry *entry;
-	long key = offset << 16 | evsel->core.idx;
-
-	if (!hashmap__find(src->samples, key, &entry))
-		return NULL;
-	return entry;
-}
+struct sym_hist_entry *
+annotated_source__hist_entry(struct annotated_source *src, const struct evsel *evsel, u64 offset);
 
 static inline struct annotation *symbol__annotation(struct symbol *sym)
 {
Signature · 1 line
2.54.0
#5 lore raw

Re: [PATCH v7 2/3] perf tools: make the GTK4 report browser actually loadable at runtime

Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 19 lines

perf report --gtk dlopen()s libperf-gtk.so, which expects to resolve symbols back against the running perf binary (callchain_param, symbol_conf, evsel__name, and friends live in perf, not the plugin). Two things broke that after the GTK 4 port:

perf never passed -rdynamic, so none of its symbols were in its dynamic symbol table for a dlopen()ed plugin to find. Add -rdynamic to LDFLAGS when GTK4 support is enabled.

annotated_source__hist_entry() was a static inline in annotate.h, so ui/gtk/annotate.c calling it pulled hashmap__find()'s expansion, hashmap_find(), into libperf-gtk.so as an undefined symbol. The only hashmap_find perf links against normally is libbpf's internal one (tools/lib/bpf/hashmap.c), built with -fvisibility=hidden, so it can never be exported to a dlopen()ed plugin regardless of LDFLAGS. Move annotated_source__hist_entry() into annotate.c as an ordinary exported function, so the plugin depends on it the same way it already depends on evsel__group_desc() and friends.

Quoted 6 lines

With both fixes, a default 'make GTK4=1' build (libbpf statically linked) can dlopen() libperf-gtk.so and open the report browser without NO_LIBBPF=1 or manual LDFLAGS. Verified with perf report --gtk against real perf.data on a GTK4 desktop.

Sashiko has reviewed this patch and found no issues. It looks great!

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
#6 lore raw

PATCH v7 3/3 perf tools gtk: fix two hierarchy-view stack buffer overflows

In reply to Matt Turner

perf_gtk__show_hierarchy() builds a merged column header for the hierarchy view with unbounded strcat() calls into a 512-byte stack buffer. The pieces being appended come from tracepoint field names and sort-key headers in perf.data, so a file with enough dynamic sort keys or long enough field names overflows the buffer.

perf_gtk__add_hierarchy_entries() has a related bug in the loop that formats each entry's value columns. fmt->entry()/fmt->color() return via scnprintf(), so ret is clamped to at most hpp->size - 1, but advance_hpp(hpp, ret + 2) doesn't clamp: when ret hits that maximum, ret + 2 exceeds hpp->size by one, and hpp->size (size_t) underflows to roughly SIZE_MAX. The next iteration's fmt->entry() then writes into the caller's stack buffer using that bogus size, a second overflow.

Fix the header builder by tracking the write offset and using scnprintf() for each append, same pattern already used elsewhere in this file. Fix the entry loop by clamping the amount passed to advance_hpp() to what's actually left in the buffer.

Both bugs predate the perf GTK UI's move to GTK 4; neither function is touched by that port.

---
 tools/perf/ui/gtk/hists.c | 23 ++++++++++++++++++-----
 1 file changed, 18 insertions(+), 5 deletions(-)
tools/perf/ui/gtk/hists.c +18 -5
index 716dcf02bd0e..342d4c3fecb0 100644
@@ -449,7 +449,7 @@ static void perf_gtk__add_hierarchy_entries(struct hists *hists,
 		bf = hpp->buf;
 		size = hpp->size;
 		perf_hpp_list__for_each_format(he->hpp_list, fmt) {
-			int ret;
+			int ret, inc;
 
 			if (fmt->color)
 				ret = fmt->color(fmt, hpp, he);
@@ -457,7 +457,18 @@ static void perf_gtk__add_hierarchy_entries(struct hists *hists,
 				ret = fmt->entry(fmt, hpp, he);
 
 			snprintf(hpp->buf + ret, hpp->size - ret, "  ");
-			advance_hpp(hpp, ret + 2);
+			/*
+			 * ret can be as large as hpp->size - 1, so ret + 2
+			 * can exceed hpp->size. advance_hpp() doesn't clamp,
+			 * so passing that through would underflow the
+			 * size_t hpp->size and let a later fmt->entry() in
+			 * this loop write past the end of the caller's
+			 * stack buffer.
+			 */
+			inc = ret + 2;
+			if (inc > (int)hpp->size)
+				inc = hpp->size;
+			advance_hpp(hpp, inc);
 		}
 
 		gtk_tree_store_set(store, &iter, col_idx, strim(bf), -1);
@@ -505,6 +516,7 @@ static void perf_gtk__show_hierarchy(GtkWidget *window, struct hists *hists,
 	GtkWidget *view;
 	int col_idx;
 	int nr_cols = 0;
+	int ret;
 	char s[512];
 	char buf[512];
 	bool first_node, first_col;
@@ -541,9 +553,10 @@ static void perf_gtk__show_hierarchy(GtkWidget *window, struct hists *hists,
 	/* construct merged column header since sort keys share single column */
 	buf[0] = '\0';
 	first_node = true;
+	ret = 0;
 	list_for_each_entry_continue(fmt_node, &hists->hpp_formats, list) {
 		if (!first_node)
-			strcat(buf, " / ");
+			ret += scnprintf(buf + ret, sizeof(buf) - ret, " / ");
 		first_node = false;
 
 		first_col = true;
@@ -552,11 +565,11 @@ static void perf_gtk__show_hierarchy(GtkWidget *window, struct hists *hists,
 				continue;
 
 			if (!first_col)
-				strcat(buf, "+");
+				ret += scnprintf(buf + ret, sizeof(buf) - ret, "+");
 			first_col = false;
 
 			fmt->header(fmt, &hpp, hists, 0, NULL);
-			strcat(buf, strim(hpp.buf));
+			ret += scnprintf(buf + ret, sizeof(buf) - ret, "%s", strim(hpp.buf));
 		}
 	}
 
Signature · 1 line
2.54.0
#7 lore raw

Re: [PATCH v7 3/3] perf tools gtk: fix two hierarchy-view stack buffer overflows

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Buffer state (hpp->buf and hpp->size) is permanently consumed across sibling nodes in the hierarchy, leading to buffer exhaustion, missing text output, and an out-of-bounds stack read (info leak). --

commit 4426046073bc5d6534ae6439f098dee3dfc79459
Author: Matt Turner <[email protected]>

perf tools gtk: fix two hierarchy-view stack buffer overflows

This commit fixes two stack buffer overflows in the perf GTK hierarchy view. It replaces unbounded strcat calls with scnprintf in header building and clamps the buffer advance in perf_gtk__add_hierarchy_entries.

Quoted 23 lines
tools/perf/ui/gtk/hists.c +6 -2
index 716dcf02bd0e8..342d4c3fecb01 100644
@@ -449,7 +449,7 @@ static void perf_gtk__add_hierarchy_entries(struct hists *hists,
 		bf = hpp->buf;
 		size = hpp->size;
 		perf_hpp_list__for_each_format(he->hpp_list, fmt) {
-			int ret;
+			int ret, inc;
 
 			if (fmt->color)
 				ret = fmt->color(fmt, hpp, he);
@@ -457,7 +457,18 @@ static void perf_gtk__add_hierarchy_entries(struct hists *hists,
 				ret = fmt->entry(fmt, hpp, he);
 
 			snprintf(hpp->buf + ret, hpp->size - ret, "  ");
-			advance_hpp(hpp, ret + 2);
+			/*
+			 * ret can be as large as hpp->size - 1, so ret + 2
+			 * can exceed hpp->size. advance_hpp() doesn't clamp,
+			 * so passing that through would underflow the
+			 * size_t hpp->size and let a later fmt->entry() in

tools/perf/ui/gtk/hists.c static void perf_gtk__add_hierarchy_entries(struct hists *hists, · line 474

Quoted 10 lines
tools/perf/ui/gtk/hists.c +7 -0
+			 * this loop write past the end of the caller's
+			 * stack buffer.
+			 */
+			inc = ret + 2;
+			if (inc > (int)hpp->size)
+				inc = hpp->size;
+			advance_hpp(hpp, inc);
 		}
 
 		gtk_tree_store_set(store, &iter, col_idx, strim(bf), -1);

[Severity: High]
This is a pre-existing issue, but does this logic permanently consume the buffer for all subsequent sibling nodes in the hierarchy?

While this patch successfully fixes the underflow by clamping inc, the buffer state (hpp->buf and hpp->size) is never restored at the end of the outer loop for the next sibling iteration.

Because the state is saved at the start of formatting (bf = hpp->buf;) but not restored across the outer rb-tree traversal loop, the 512-byte buffer is permanently consumed as the tree is traversed.

When the buffer is completely exhausted (hpp->size hits 0), bf points past the end of the original 512-byte stack buffer. Doesn't this mean the subsequent strim(bf) call will perform an out-of-bounds read on the stack, leaking adjacent uninitialized memory into the UI?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
#8 lore raw
In reply to Matt Turner
Quoted 12 lines · Matt Turner

GTK2 is long dead upstream and increasingly hard to keep building on current distros. This series ports perf's GTK-based report browser to GTK4 and fixes it up so it's actually loadable at runtime after the port.

Patch 1 does the mechanical port (build system, widget API changes), including the leftover-GTK2-call and signal-handling fixes that were a separate patch 3 in v4. Patch 2 fixes a runtime issue found after the port that prevented the browser from loading. Patch 3 fixes two stack buffer overflows in the hierarchy view that predate the port.

Quoted 3 lines

---
Changes in v7:

Matt, please stop the spam.
Wait for humans to reply before respinning.
7 revisions in one day is way too much.

#9 lore raw
Quoted 14 lines · Alexei Starovoitov
Quoted 12 lines · Matt Turner

GTK2 is long dead upstream and increasingly hard to keep building on current distros. This series ports perf's GTK-based report browser to GTK4 and fixes it up so it's actually loadable at runtime after the port.

Patch 1 does the mechanical port (build system, widget API changes), including the leftover-GTK2-call and signal-handling fixes that were a separate patch 3 in v4. Patch 2 fixes a runtime issue found after the port that prevented the browser from loading. Patch 3 fixes two stack buffer overflows in the hierarchy view that predate the port.

Quoted 7 lines
Quoted 3 lines

---
Changes in v7:

Matt, please stop the spam.
Wait for humans to reply before respinning.
7 revisions in one day is way too much.

My apologies. I figured it would actually save humans time if I went ahead and addressed all the sashiko-generated feedback. Dunno.

Is there a way to get feedback from sashiko without mailing the public list?

reviewer
#10 lore raw
Quoted 21 lines · Matt Turner
Quoted 18 lines · Alexei Starovoitov
Quoted 15 lines · Matt Turner

GTK2 is long dead upstream and increasingly hard to keep building on current distros. This series ports perf's GTK-based report browser to GTK4 and fixes it up so it's actually loadable at runtime after the port.

Patch 1 does the mechanical port (build system, widget API changes), including the leftover-GTK2-call and signal-handling fixes that were a separate patch 3 in v4. Patch 2 fixes a runtime issue found after the port that prevented the browser from loading. Patch 3 fixes two stack buffer overflows in the hierarchy view that predate the port.

---
Changes in v7:

Quoted 8 lines
Quoted 3 lines

Matt, please stop the spam.
Wait for humans to reply before respinning.
7 revisions in one day is way too much.

My apologies. I figured it would actually save humans time if I went ahead and addressed all the sashiko-generated feedback. Dunno.

Is there a way to get feedback from sashiko without mailing the public list?

TL;DR: drop bpf from the to/cc list when sending patches. This work is most interesting to linux-perf-users where multiple versions of a patch series within a short period of time are common.

I appreciate having the latest patches to review. If Sashiko provides feedback, the change goes on my 'to be reviewed later' pile, which hinders getting code/fixes landed.
It is possible to configure Sashiko to 'embargo' its feedback for a set number of hours. Networking does this, and it can be frustrating when you must wait for a patch to receive review before landing it. It also slows down developers.

Sashiko isn't adopted on all mailing lists, and while it is a widely praised, others don't want the email churn. In a year or so, when the models are better, I imagine the churn will decrease as:
1) If AI is used to create the code then it should be better able to predict Sashiko-like issues,
2) Sashiko itself can report all problems, suggest changes, etc. rather than its current drip drip of issues.

Thanks for the contribution!
Ian