mirrors/wlroots
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/wlroots/wlroots.git synced 2025-11-13 13:29:44 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

backend/drm: implement the output layer API

Browse source 
Delegate all of the work to libliftoff.
This commit is contained in:
Simon Ser 2021-08-10 00:33:48 +02:00
parent e45ddda7bc
commit 2a3861024e
3 changed files with 134 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

57
backend/drm/drm.c
View file

@ -423,6 +423,11 @@ static bool drm_crtc_commit(struct wlr_drm_connector *conn,
if (crtc->cursor != NULL) {
drm_fb_move(&crtc->cursor->queued_fb, &crtc->cursor->pending_fb);
}
struct wlr_drm_layer *layer;
wl_list_for_each(layer, &state->base->layers, base.pending.link) {
drm_fb_move(&layer->queued_fb, &layer->pending_fb);
}
} else {
drm_fb_clear(&crtc->primary->pending_fb);
// The set_cursor() hook is a bit special: it's not really synchronized
@ -431,6 +436,11 @@ static bool drm_crtc_commit(struct wlr_drm_connector *conn,
// risk ending up in a state where we don't have a cursor FB but
// wlr_drm_connector.cursor_enabled is true.
// TODO: fix our output interface to avoid this issue.
struct wlr_drm_layer *layer;
wl_list_for_each(layer, &state->base->layers, base.pending.link) {
drm_fb_clear(&layer->pending_fb);
}
}
return ok;
}
@ -545,6 +555,23 @@ static bool drm_connector_set_pending_fb(struct wlr_drm_connector *conn,
return true;
}
static void drm_connector_set_pending_layer_fbs(struct wlr_drm_connector *conn,
const struct wlr_output_state *state) {
struct wlr_drm_backend *drm = conn->backend;
struct wlr_drm_layer *layer;
wl_list_for_each(layer, &state->layers, base.pending.link) {
if (!(layer->base.pending.committed & WLR_OUTPUT_LAYER_STATE_BUFFER)) {
continue;
}
struct wlr_buffer *buffer = layer->base.pending.buffer;
drm_fb_clear(&layer->pending_fb);
if (!drm_fb_import(&layer->pending_fb, drm, buffer, NULL)) {
wlr_drm_conn_log(conn, WLR_DEBUG, "Failed to import layer buffer");
}
}
}
static bool drm_connector_alloc_crtc(struct wlr_drm_connector *conn);
static bool drm_connector_test(struct wlr_output *output) {
@ -602,6 +629,8 @@ static bool drm_connector_test(struct wlr_output *output) {
return true;
}
drm_connector_set_pending_layer_fbs(conn, pending.base);
if (output->pending.committed & WLR_OUTPUT_STATE_BUFFER) {
if (!drm_connector_set_pending_fb(conn, pending.base)) {
return false;
@ -659,6 +688,8 @@ bool drm_connector_commit_state(struct wlr_drm_connector *conn,
}
}
drm_connector_set_pending_layer_fbs(conn, pending.base);
if (pending.base->committed & WLR_OUTPUT_STATE_BUFFER) {
if (!drm_connector_set_pending_fb(conn, pending.base)) {
return false;
@ -1036,6 +1067,25 @@ static const struct wlr_drm_format_set *drm_connector_get_primary_formats(
return &conn->crtc->primary->formats;
}
static struct wlr_output_layer *drm_connector_create_layer(
struct wlr_output *wlr_output) {
struct wlr_drm_layer *layer = calloc(1, sizeof(*layer));
if (layer == NULL) {
return NULL;
}
wlr_output_layer_init(&layer->base, wlr_output);
return &layer->base;
}
static void drm_connector_destroy_layer(struct wlr_output_layer *wlr_layer) {
struct wlr_drm_layer *layer = (struct wlr_drm_layer *)wlr_layer;
drm_fb_clear(&layer->pending_fb);
drm_fb_clear(&layer->queued_fb);
drm_fb_clear(&layer->current_fb);
liftoff_layer_destroy(layer->liftoff);
free(layer);
}
static const struct wlr_output_impl output_impl = {
.set_cursor = drm_connector_set_cursor,
.move_cursor = drm_connector_move_cursor,
@ -1046,6 +1096,8 @@ static const struct wlr_output_impl output_impl = {
.get_cursor_formats = drm_connector_get_cursor_formats,
.get_cursor_size = drm_connector_get_cursor_size,
.get_primary_formats = drm_connector_get_primary_formats,
.create_layer = drm_connector_create_layer,
.destroy_layer = drm_connector_destroy_layer,
};
bool wlr_output_is_drm(struct wlr_output *output) {
@ -1543,6 +1595,11 @@ static void handle_page_flip(int fd, unsigned seq,
&conn->crtc->cursor->queued_fb);
}
struct wlr_drm_layer *layer;
wl_list_for_each(layer, &conn->output.layers, base.current.link) {
drm_fb_move(&layer->current_fb, &layer->queued_fb);
}
uint32_t present_flags = WLR_OUTPUT_PRESENT_VSYNC |
WLR_OUTPUT_PRESENT_HW_CLOCK | WLR_OUTPUT_PRESENT_HW_COMPLETION;
/* Don't report ZERO_COPY in multi-gpu situations, because we had to copy