vblank: add GLX_SGI_video_sync based scheduler

Present extension based scheduler doesn't work well on NVIDIA drivers.
GLX_SGI_video_sync is less accurate, but is rumoured to work. See
[kwin's usage](https://invent.kde.org/plasma/kwin/-/blob/master/src/
backends/x11/standalone/x11_standalone_sgivideosyncvsyncmonitor.cpp)

Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>

src/meson.build

@@ -59,7 +59,7 @@ endif
 
 if get_option('opengl')
 	cflags += ['-DCONFIG_OPENGL', '-DGL_GLEXT_PROTOTYPES']
-	deps += [dependency('gl', required: true), dependency('egl', required: true)]
+	deps += [dependency('gl', required: true), dependency('egl', required: true), dependency('threads', required:true)]
 	srcs += [ 'opengl.c' ]
 endif
 

src/picom.c

@@ -1501,7 +1501,8 @@ static bool redirect_start(session_t *ps) {
 		ps->last_schedule_delay = 0;
 		render_statistics_reset(&ps->render_stats);
 		ps->vblank_scheduler =
-		    vblank_scheduler_new(ps->loop, &ps->c, session_get_target_window(ps));
+		    vblank_scheduler_new(ps->loop, &ps->c, session_get_target_window(ps),
+		                         VBLANK_SCHEDULER_PRESENT);
 		if (!ps->vblank_scheduler) {
 			return false;
 		}

src/vblank.c

@@ -2,12 +2,26 @@
 
 #include <ev.h>
 #include <inttypes.h>
+#include <stdatomic.h>
 #include <string.h>
+#include <time.h>
 #include <xcb/xcb.h>
 #include <xcb/xproto.h>
+#include "config.h"
+
+#ifdef CONFIG_OPENGL
+// Enable sgi_video_sync_vblank_scheduler
+#include <GL/glx.h>
+#include <X11/X.h>
+#include <X11/Xlib-xcb.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include <pthread.h>
+
+#include "backend/gl/glx.h"
+#endif
 
 #include "compiler.h"
-#include "config.h"
 #include "list.h"        // for container_of
 #include "log.h"
 #include "vblank.h"
@@ -48,6 +62,7 @@ struct present_vblank_scheduler {
 };
 
 struct vblank_scheduler_ops {
+	size_t size;
 	void (*init)(struct vblank_scheduler *self);
 	void (*deinit)(struct vblank_scheduler *self);
 	void (*schedule)(struct vblank_scheduler *self);
@@ -57,6 +72,242 @@ struct vblank_scheduler_ops {
 static void
 vblank_scheduler_invoke_callbacks(struct vblank_scheduler *self, struct vblank_event *event);
 
+#ifdef CONFIG_OPENGL
+struct sgi_video_sync_vblank_scheduler {
+	struct vblank_scheduler base;
+
+	// Since glXWaitVideoSyncSGI blocks, we need to run it in a separate thread.
+	// ... and all the thread shenanigans that come with it.
+	_Atomic unsigned int last_msc;
+	_Atomic uint64_t last_ust;
+	ev_async notify;
+	pthread_t sync_thread;
+	bool running, error;
+
+	/// Protects `running`, `error` and `base.vblank_event_requested`
+	pthread_mutex_t vblank_requested_mtx;
+	pthread_cond_t vblank_requested_cnd;
+};
+
+struct sgi_video_sync_thread_args {
+	struct sgi_video_sync_vblank_scheduler *self;
+	int start_status;
+	pthread_mutex_t start_mtx;
+	pthread_cond_t start_cnd;
+};
+
+static bool check_sgi_video_sync_extension(Display *dpy, int screen) {
+	const char *glx_ext = glXQueryExtensionsString(dpy, screen);
+	const char *needle = "GLX_SGI_video_sync";
+	char *found = strstr(glx_ext, needle);
+	if (!found) {
+		return false;
+	}
+	if (found != glx_ext && found[-1] != ' ') {
+		return false;
+	}
+	if (found[strlen(needle)] != ' ' && found[strlen(needle)] != '\0') {
+		return false;
+	}
+
+	glXWaitVideoSyncSGI = (PFNGLXWAITVIDEOSYNCSGIPROC)(void *)glXGetProcAddress(
+	    (const GLubyte *)"glXWaitVideoSyncSGI");
+	if (!glXWaitVideoSyncSGI) {
+		return false;
+	}
+	return true;
+}
+
+static void *sgi_video_sync_thread(void *data) {
+	auto args = (struct sgi_video_sync_thread_args *)data;
+	auto self = args->self;
+	Display *dpy = XOpenDisplay(NULL);
+	int error_code = 0;
+	if (!dpy) {
+		error_code = 1;
+		goto start_failed;
+	}
+	Window root = DefaultRootWindow(dpy), dummy = None;
+	int screen = DefaultScreen(dpy);
+	int ncfg = 0;
+	GLXFBConfig *cfg_ = glXChooseFBConfig(
+	    dpy, screen,
+	    (int[]){GLX_RENDER_TYPE, GLX_RGBA_BIT, GLX_DRAWABLE_TYPE, GLX_WINDOW_BIT, 0},
+	    &ncfg);
+	GLXContext ctx = NULL;
+	GLXDrawable drawable = None;
+
+	if (!cfg_) {
+		error_code = 2;
+		goto start_failed;
+	}
+	GLXFBConfig cfg = cfg_[0];
+	XFree(cfg_);
+
+	XVisualInfo *vi = glXGetVisualFromFBConfig(dpy, cfg);
+	if (!vi) {
+		error_code = 3;
+		goto start_failed;
+	}
+
+	Visual *visual = vi->visual;
+	const int depth = vi->depth;
+	XFree(vi);
+
+	Colormap colormap = XCreateColormap(dpy, root, visual, AllocNone);
+	XSetWindowAttributes attributes;
+	attributes.colormap = colormap;
+
+	dummy = XCreateWindow(dpy, root, 0, 0, 1, 1, 0, depth, InputOutput, visual,
+	                      CWColormap, &attributes);
+	XFreeColormap(dpy, colormap);
+	if (dummy == None) {
+		error_code = 4;
+		goto start_failed;
+	}
+
+	drawable = glXCreateWindow(dpy, cfg, dummy, NULL);
+	if (drawable == None) {
+		error_code = 5;
+		goto start_failed;
+	}
+
+	ctx = glXCreateNewContext(dpy, cfg, GLX_RGBA_TYPE, 0, true);
+	if (ctx == NULL) {
+		error_code = 6;
+		goto start_failed;
+	}
+
+	if (!glXMakeContextCurrent(dpy, drawable, drawable, ctx)) {
+		error_code = 7;
+		goto start_failed;
+	}
+
+	if (!check_sgi_video_sync_extension(dpy, screen)) {
+		error_code = 8;
+		goto start_failed;
+	}
+
+	pthread_mutex_lock(&args->start_mtx);
+	args->start_status = 0;
+	pthread_cond_signal(&args->start_cnd);
+	pthread_mutex_unlock(&args->start_mtx);
+
+	pthread_mutex_lock(&self->vblank_requested_mtx);
+	while (self->running) {
+		if (!self->base.vblank_event_requested) {
+			pthread_cond_wait(&self->vblank_requested_cnd,
+			                  &self->vblank_requested_mtx);
+			continue;
+		}
+		pthread_mutex_unlock(&self->vblank_requested_mtx);
+
+		unsigned int last_msc;
+		glXWaitVideoSyncSGI(1, 0, &last_msc);
+
+		struct timespec now;
+		clock_gettime(CLOCK_MONOTONIC, &now);
+		atomic_store(&self->last_msc, last_msc);
+		atomic_store(&self->last_ust,
+		             (uint64_t)(now.tv_sec * 1000000 + now.tv_nsec / 1000));
+		ev_async_send(self->base.loop, &self->notify);
+		pthread_mutex_lock(&self->vblank_requested_mtx);
+	}
+	pthread_mutex_unlock(&self->vblank_requested_mtx);
+	goto cleanup;
+
+start_failed:
+	pthread_mutex_lock(&args->start_mtx);
+	args->start_status = error_code;
+	pthread_cond_signal(&args->start_cnd);
+	pthread_mutex_unlock(&args->start_mtx);
+
+cleanup:
+	if (dpy) {
+		glXMakeCurrent(dpy, None, NULL);
+		if (ctx) {
+			glXDestroyContext(dpy, ctx);
+		}
+		if (drawable) {
+			glXDestroyWindow(dpy, drawable);
+		}
+		if (dummy) {
+			XDestroyWindow(dpy, dummy);
+		}
+		XCloseDisplay(dpy);
+	}
+	return NULL;
+}
+
+static void sgi_video_sync_scheduler_schedule(struct vblank_scheduler *base) {
+	auto self = (struct sgi_video_sync_vblank_scheduler *)base;
+	log_verbose("Requesting vblank event for msc %d", self->last_msc + 1);
+	pthread_mutex_lock(&self->vblank_requested_mtx);
+	assert(!base->vblank_event_requested);
+	base->vblank_event_requested = true;
+	pthread_cond_signal(&self->vblank_requested_cnd);
+	pthread_mutex_unlock(&self->vblank_requested_mtx);
+}
+
+static void
+sgi_video_sync_scheduler_callback(EV_P attr_unused, ev_async *w, int attr_unused revents) {
+	auto sched = container_of(w, struct sgi_video_sync_vblank_scheduler, notify);
+	auto event = (struct vblank_event){
+	    .msc = atomic_load(&sched->last_msc),
+	    .ust = atomic_load(&sched->last_ust),
+	};
+	sched->base.vblank_event_requested = false;
+	log_verbose("Received vblank event for msc %lu", event.msc);
+	vblank_scheduler_invoke_callbacks(&sched->base, &event);
+}
+
+static void sgi_video_sync_scheduler_init(struct vblank_scheduler *base) {
+	auto self = (struct sgi_video_sync_vblank_scheduler *)base;
+	auto args = (struct sgi_video_sync_thread_args){
+	    .self = self,
+	    .start_status = -1,
+	};
+	pthread_mutex_init(&args.start_mtx, NULL);
+	pthread_cond_init(&args.start_cnd, NULL);
+
+	base->type = VBLANK_SCHEDULER_SGI_VIDEO_SYNC;
+	ev_async_init(&self->notify, sgi_video_sync_scheduler_callback);
+	ev_async_start(base->loop, &self->notify);
+	pthread_mutex_init(&self->vblank_requested_mtx, NULL);
+	pthread_cond_init(&self->vblank_requested_cnd, NULL);
+
+	self->running = true;
+	pthread_create(&self->sync_thread, NULL, sgi_video_sync_thread, &args);
+
+	pthread_mutex_lock(&args.start_mtx);
+	while (args.start_status == -1) {
+		pthread_cond_wait(&args.start_cnd, &args.start_mtx);
+	}
+	if (args.start_status != 0) {
+		log_fatal("Failed to start sgi_video_sync_thread, error code: %d",
+		          args.start_status);
+		abort();
+	}
+	pthread_mutex_destroy(&args.start_mtx);
+	pthread_cond_destroy(&args.start_cnd);
+	log_info("Started sgi_video_sync_thread");
+}
+
+static void sgi_video_sync_scheduler_deinit(struct vblank_scheduler *base) {
+	auto self = (struct sgi_video_sync_vblank_scheduler *)base;
+	ev_async_stop(base->loop, &self->notify);
+	pthread_mutex_lock(&self->vblank_requested_mtx);
+	self->running = false;
+	pthread_cond_signal(&self->vblank_requested_cnd);
+	pthread_mutex_unlock(&self->vblank_requested_mtx);
+
+	pthread_join(self->sync_thread, NULL);
+
+	pthread_mutex_destroy(&self->vblank_requested_mtx);
+	pthread_cond_destroy(&self->vblank_requested_cnd);
+}
+#endif
+
 static void present_vblank_scheduler_schedule(struct vblank_scheduler *base) {
 	auto self = (struct present_vblank_scheduler *)base;
 	log_verbose("Requesting vblank event for window 0x%08x, msc %" PRIu64,
@@ -170,11 +421,22 @@ static bool handle_present_events(struct vblank_scheduler *base) {
 static const struct vblank_scheduler_ops vblank_scheduler_ops[LAST_VBLANK_SCHEDULER] = {
     [VBLANK_SCHEDULER_PRESENT] =
         {
+            .size = sizeof(struct present_vblank_scheduler),
             .init = present_vblank_scheduler_init,
             .deinit = present_vblank_scheduler_deinit,
             .schedule = present_vblank_scheduler_schedule,
             .handle_x_events = handle_present_events,
         },
+#ifdef CONFIG_OPENGL
+    [VBLANK_SCHEDULER_SGI_VIDEO_SYNC] =
+        {
+            .size = sizeof(struct sgi_video_sync_vblank_scheduler),
+            .init = sgi_video_sync_scheduler_init,
+            .deinit = sgi_video_sync_scheduler_deinit,
+            .schedule = sgi_video_sync_scheduler_schedule,
+            .handle_x_events = NULL,
+        },
+#endif
 };
 
 static void vblank_scheduler_schedule_internal(struct vblank_scheduler *self) {
@@ -250,13 +512,22 @@ void vblank_scheduler_free(struct vblank_scheduler *self) {
 	free(self);
 }
 
-struct vblank_scheduler *vblank_scheduler_new(struct ev_loop *loop, struct x_connection *c,
-                                              xcb_window_t target_window) {
-	struct vblank_scheduler *self = calloc(1, sizeof(struct present_vblank_scheduler));
+struct vblank_scheduler *
+vblank_scheduler_new(struct ev_loop *loop, struct x_connection *c,
+                     xcb_window_t target_window, enum vblank_scheduler_type type) {
+	size_t object_size = vblank_scheduler_ops[type].size;
+	auto init_fn = vblank_scheduler_ops[type].init;
+	if (!object_size || !init_fn) {
+		log_error("Unsupported or invalid vblank scheduler type: %d", type);
+		return NULL;
+	}
+
+	assert(object_size >= sizeof(struct vblank_scheduler));
+	struct vblank_scheduler *self = calloc(1, object_size);
 	self->target_window = target_window;
 	self->c = c;
 	self->loop = loop;
-	vblank_scheduler_ops[VBLANK_SCHEDULER_PRESENT].init(self);
+	init_fn(self);
 	return self;
 }
 

src/vblank.h

@@ -8,6 +8,7 @@
 #include <ev.h>
 #include <xcb/xproto.h>
 
+#include "config.h"
 #include "x.h"
 
 /// An object that schedule vblank events.
@@ -38,8 +39,9 @@ typedef enum vblank_callback_action (*vblank_callback_t)(struct vblank_event *ev
 /// is not enough memory.
 bool vblank_scheduler_schedule(struct vblank_scheduler *self, vblank_callback_t cb,
                                void *user_data);
-struct vblank_scheduler *vblank_scheduler_new(struct ev_loop *loop, struct x_connection *c,
-                                              xcb_window_t target_window);
+struct vblank_scheduler *
+vblank_scheduler_new(struct ev_loop *loop, struct x_connection *c,
+                     xcb_window_t target_window, enum vblank_scheduler_type type);
 void vblank_scheduler_free(struct vblank_scheduler *);
 
 bool vblank_handle_x_events(struct vblank_scheduler *self);