mirror of
https://github.com/hyprwm/wlroots-hyprland.git
synced 2024-12-28 20:39:50 +01:00
8da9d9679e
This is a type which manages gbm_surfaces and imported dmabufs in the same place, and makes the lifetime management between the two shared. It should lead to easier to understand code, and fewer special cases. This also contains a fair bit of refactoring to start using this new type. Co-authored-by: Simon Ser <contact@emersion.fr>
368 lines
12 KiB
C
368 lines
12 KiB
C
#include <assert.h>
|
|
#include <drm_fourcc.h>
|
|
#include <drm_mode.h>
|
|
#include <drm.h>
|
|
#include <gbm.h>
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
#include <wlr/util/log.h>
|
|
#include "backend/drm/util.h"
|
|
|
|
int32_t calculate_refresh_rate(const drmModeModeInfo *mode) {
|
|
int32_t refresh = (mode->clock * 1000000LL / mode->htotal +
|
|
mode->vtotal / 2) / mode->vtotal;
|
|
|
|
if (mode->flags & DRM_MODE_FLAG_INTERLACE) {
|
|
refresh *= 2;
|
|
}
|
|
|
|
if (mode->flags & DRM_MODE_FLAG_DBLSCAN) {
|
|
refresh /= 2;
|
|
}
|
|
|
|
if (mode->vscan > 1) {
|
|
refresh /= mode->vscan;
|
|
}
|
|
|
|
return refresh;
|
|
}
|
|
|
|
// Constructed from http://edid.tv/manufacturer
|
|
static const char *get_manufacturer(uint16_t id) {
|
|
#define ID(a, b, c) ((a & 0x1f) << 10) | ((b & 0x1f) << 5) | (c & 0x1f)
|
|
switch (id) {
|
|
case ID('A', 'A', 'A'): return "Avolites Ltd";
|
|
case ID('A', 'C', 'I'): return "Ancor Communications Inc";
|
|
case ID('A', 'C', 'R'): return "Acer Technologies";
|
|
case ID('A', 'D', 'A'): return "Addi-Data GmbH";
|
|
case ID('A', 'P', 'P'): return "Apple Computer Inc";
|
|
case ID('A', 'S', 'K'): return "Ask A/S";
|
|
case ID('A', 'V', 'T'): return "Avtek (Electronics) Pty Ltd";
|
|
case ID('B', 'N', 'O'): return "Bang & Olufsen";
|
|
case ID('B', 'N', 'Q'): return "BenQ Corporation";
|
|
case ID('C', 'M', 'N'): return "Chimei Innolux Corporation";
|
|
case ID('C', 'M', 'O'): return "Chi Mei Optoelectronics corp.";
|
|
case ID('C', 'R', 'O'): return "Extraordinary Technologies PTY Limited";
|
|
case ID('D', 'E', 'L'): return "Dell Inc.";
|
|
case ID('D', 'G', 'C'): return "Data General Corporation";
|
|
case ID('D', 'O', 'N'): return "DENON, Ltd.";
|
|
case ID('E', 'N', 'C'): return "Eizo Nanao Corporation";
|
|
case ID('E', 'P', 'H'): return "Epiphan Systems Inc.";
|
|
case ID('E', 'X', 'P'): return "Data Export Corporation";
|
|
case ID('F', 'N', 'I'): return "Funai Electric Co., Ltd.";
|
|
case ID('F', 'U', 'S'): return "Fujitsu Siemens Computers GmbH";
|
|
case ID('G', 'S', 'M'): return "Goldstar Company Ltd";
|
|
case ID('H', 'I', 'Q'): return "Kaohsiung Opto Electronics Americas, Inc.";
|
|
case ID('H', 'S', 'D'): return "HannStar Display Corp";
|
|
case ID('H', 'T', 'C'): return "Hitachi Ltd";
|
|
case ID('H', 'W', 'P'): return "Hewlett Packard";
|
|
case ID('I', 'N', 'T'): return "Interphase Corporation";
|
|
case ID('I', 'N', 'X'): return "Communications Supply Corporation (A division of WESCO)";
|
|
case ID('I', 'T', 'E'): return "Integrated Tech Express Inc";
|
|
case ID('I', 'V', 'M'): return "Iiyama North America";
|
|
case ID('L', 'E', 'N'): return "Lenovo Group Limited";
|
|
case ID('M', 'A', 'X'): return "Rogen Tech Distribution Inc";
|
|
case ID('M', 'E', 'G'): return "Abeam Tech Ltd";
|
|
case ID('M', 'E', 'I'): return "Panasonic Industry Company";
|
|
case ID('M', 'T', 'C'): return "Mars-Tech Corporation";
|
|
case ID('M', 'T', 'X'): return "Matrox";
|
|
case ID('N', 'E', 'C'): return "NEC Corporation";
|
|
case ID('N', 'E', 'X'): return "Nexgen Mediatech Inc.";
|
|
case ID('O', 'N', 'K'): return "ONKYO Corporation";
|
|
case ID('O', 'R', 'N'): return "ORION ELECTRIC CO., LTD.";
|
|
case ID('O', 'T', 'M'): return "Optoma Corporation";
|
|
case ID('O', 'V', 'R'): return "Oculus VR, Inc.";
|
|
case ID('P', 'H', 'L'): return "Philips Consumer Electronics Company";
|
|
case ID('P', 'I', 'O'): return "Pioneer Electronic Corporation";
|
|
case ID('P', 'N', 'R'): return "Planar Systems, Inc.";
|
|
case ID('Q', 'D', 'S'): return "Quanta Display Inc.";
|
|
case ID('R', 'A', 'T'): return "Rent-A-Tech";
|
|
case ID('R', 'E', 'N'): return "Renesas Technology Corp.";
|
|
case ID('S', 'A', 'M'): return "Samsung Electric Company";
|
|
case ID('S', 'A', 'N'): return "Sanyo Electric Co., Ltd.";
|
|
case ID('S', 'E', 'C'): return "Seiko Epson Corporation";
|
|
case ID('S', 'H', 'P'): return "Sharp Corporation";
|
|
case ID('S', 'I', 'I'): return "Silicon Image, Inc.";
|
|
case ID('S', 'N', 'Y'): return "Sony";
|
|
case ID('S', 'T', 'D'): return "STD Computer Inc";
|
|
case ID('S', 'V', 'S'): return "SVSI";
|
|
case ID('S', 'Y', 'N'): return "Synaptics Inc";
|
|
case ID('T', 'C', 'L'): return "Technical Concepts Ltd";
|
|
case ID('T', 'O', 'P'): return "Orion Communications Co., Ltd.";
|
|
case ID('T', 'S', 'B'): return "Toshiba America Info Systems Inc";
|
|
case ID('T', 'S', 'T'): return "Transtream Inc";
|
|
case ID('U', 'N', 'K'): return "Unknown";
|
|
case ID('V', 'E', 'S'): return "Vestel Elektronik Sanayi ve Ticaret A. S.";
|
|
case ID('V', 'I', 'T'): return "Visitech AS";
|
|
case ID('V', 'I', 'Z'): return "VIZIO, Inc";
|
|
case ID('V', 'S', 'C'): return "ViewSonic Corporation";
|
|
case ID('Y', 'M', 'H'): return "Yamaha Corporation";
|
|
default: return "Unknown";
|
|
}
|
|
#undef ID
|
|
}
|
|
|
|
/* See https://en.wikipedia.org/wiki/Extended_Display_Identification_Data for layout of EDID data.
|
|
* We don't parse the EDID properly. We just expect to receive valid data.
|
|
*/
|
|
void parse_edid(struct wlr_output *restrict output, size_t len, const uint8_t *data) {
|
|
if (!data || len < 128) {
|
|
snprintf(output->make, sizeof(output->make), "<Unknown>");
|
|
snprintf(output->model, sizeof(output->model), "<Unknown>");
|
|
return;
|
|
}
|
|
|
|
uint16_t id = (data[8] << 8) | data[9];
|
|
snprintf(output->make, sizeof(output->make), "%s", get_manufacturer(id));
|
|
|
|
uint16_t model = data[10] | (data[11] << 8);
|
|
snprintf(output->model, sizeof(output->model), "0x%04X", model);
|
|
|
|
uint32_t serial = data[12] | (data[13] << 8) | (data[14] << 8) | (data[15] << 8);
|
|
snprintf(output->serial, sizeof(output->serial), "0x%08X", serial);
|
|
|
|
for (size_t i = 72; i <= 108; i += 18) {
|
|
uint16_t flag = (data[i] << 8) | data[i + 1];
|
|
if (flag == 0 && data[i + 3] == 0xFC) {
|
|
sprintf(output->model, "%.13s", &data[i + 5]);
|
|
|
|
// Monitor names are terminated by newline if they're too short
|
|
char *nl = strchr(output->model, '\n');
|
|
if (nl) {
|
|
*nl = '\0';
|
|
}
|
|
} else if (flag == 0 && data[i + 3] == 0xFF) {
|
|
sprintf(output->serial, "%.13s", &data[i + 5]);
|
|
|
|
// Monitor serial numbers are terminated by newline if they're too
|
|
// short
|
|
char *nl = strchr(output->serial, '\n');
|
|
if (nl) {
|
|
*nl = '\0';
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
const char *conn_get_name(uint32_t type_id) {
|
|
switch (type_id) {
|
|
case DRM_MODE_CONNECTOR_Unknown: return "Unknown";
|
|
case DRM_MODE_CONNECTOR_VGA: return "VGA";
|
|
case DRM_MODE_CONNECTOR_DVII: return "DVI-I";
|
|
case DRM_MODE_CONNECTOR_DVID: return "DVI-D";
|
|
case DRM_MODE_CONNECTOR_DVIA: return "DVI-A";
|
|
case DRM_MODE_CONNECTOR_Composite: return "Composite";
|
|
case DRM_MODE_CONNECTOR_SVIDEO: return "SVIDEO";
|
|
case DRM_MODE_CONNECTOR_LVDS: return "LVDS";
|
|
case DRM_MODE_CONNECTOR_Component: return "Component";
|
|
case DRM_MODE_CONNECTOR_9PinDIN: return "DIN";
|
|
case DRM_MODE_CONNECTOR_DisplayPort: return "DP";
|
|
case DRM_MODE_CONNECTOR_HDMIA: return "HDMI-A";
|
|
case DRM_MODE_CONNECTOR_HDMIB: return "HDMI-B";
|
|
case DRM_MODE_CONNECTOR_TV: return "TV";
|
|
case DRM_MODE_CONNECTOR_eDP: return "eDP";
|
|
case DRM_MODE_CONNECTOR_VIRTUAL: return "Virtual";
|
|
case DRM_MODE_CONNECTOR_DSI: return "DSI";
|
|
#ifdef DRM_MODE_CONNECTOR_DPI
|
|
case DRM_MODE_CONNECTOR_DPI: return "DPI";
|
|
#endif
|
|
default: return "Unknown";
|
|
}
|
|
}
|
|
|
|
static void free_fb(struct gbm_bo *bo, void *data) {
|
|
uint32_t id = (uintptr_t)data;
|
|
|
|
if (id) {
|
|
struct gbm_device *gbm = gbm_bo_get_device(bo);
|
|
drmModeRmFB(gbm_device_get_fd(gbm), id);
|
|
}
|
|
}
|
|
|
|
uint32_t get_fb_for_bo(struct gbm_bo *bo, bool with_modifiers) {
|
|
uint32_t id = (uintptr_t)gbm_bo_get_user_data(bo);
|
|
if (id) {
|
|
return id;
|
|
}
|
|
|
|
struct gbm_device *gbm = gbm_bo_get_device(bo);
|
|
|
|
int fd = gbm_device_get_fd(gbm);
|
|
uint32_t width = gbm_bo_get_width(bo);
|
|
uint32_t height = gbm_bo_get_height(bo);
|
|
uint32_t format = gbm_bo_get_format(bo);
|
|
|
|
uint32_t handles[4] = {0};
|
|
uint32_t strides[4] = {0};
|
|
uint32_t offsets[4] = {0};
|
|
uint64_t modifiers[4] = {0};
|
|
for (int i = 0; i < gbm_bo_get_plane_count(bo); i++) {
|
|
handles[i] = gbm_bo_get_handle_for_plane(bo, i).u32;
|
|
strides[i] = gbm_bo_get_stride_for_plane(bo, i);
|
|
offsets[i] = gbm_bo_get_offset(bo, i);
|
|
// KMS requires all BO planes to have the same modifier
|
|
modifiers[i] = gbm_bo_get_modifier(bo);
|
|
}
|
|
|
|
if (with_modifiers && gbm_bo_get_modifier(bo) != DRM_FORMAT_MOD_INVALID) {
|
|
if (drmModeAddFB2WithModifiers(fd, width, height, format, handles,
|
|
strides, offsets, modifiers, &id, DRM_MODE_FB_MODIFIERS)) {
|
|
wlr_log_errno(WLR_ERROR, "Unable to add DRM framebuffer");
|
|
}
|
|
} else {
|
|
if (drmModeAddFB2(fd, width, height, format, handles, strides,
|
|
offsets, &id, 0)) {
|
|
wlr_log_errno(WLR_ERROR, "Unable to add DRM framebuffer");
|
|
}
|
|
}
|
|
|
|
gbm_bo_set_user_data(bo, (void *)(uintptr_t)id, free_fb);
|
|
|
|
return id;
|
|
}
|
|
|
|
static inline bool is_taken(size_t n, const uint32_t arr[static n], uint32_t key) {
|
|
for (size_t i = 0; i < n; ++i) {
|
|
if (arr[i] == key) {
|
|
return true;
|
|
}
|
|
}
|
|
return false;
|
|
}
|
|
|
|
/*
|
|
* Store all of the non-recursive state in a struct, so we aren't literally
|
|
* passing 12 arguments to a function.
|
|
*/
|
|
struct match_state {
|
|
const size_t num_objs;
|
|
const uint32_t *restrict objs;
|
|
const size_t num_res;
|
|
size_t score;
|
|
size_t replaced;
|
|
uint32_t *restrict res;
|
|
uint32_t *restrict best;
|
|
const uint32_t *restrict orig;
|
|
bool exit_early;
|
|
};
|
|
|
|
/*
|
|
* skips: The number of SKIP elements encountered so far.
|
|
* score: The number of resources we've matched so far.
|
|
* replaced: The number of changes from the original solution.
|
|
* i: The index of the current element.
|
|
*
|
|
* This tries to match a solution as close to st->orig as it can.
|
|
*
|
|
* Returns whether we've set a new best element with this solution.
|
|
*/
|
|
static bool match_obj_(struct match_state *st, size_t skips, size_t score, size_t replaced, size_t i) {
|
|
// Finished
|
|
if (i >= st->num_res) {
|
|
if (score > st->score ||
|
|
(score == st->score && replaced < st->replaced)) {
|
|
st->score = score;
|
|
st->replaced = replaced;
|
|
memcpy(st->best, st->res, sizeof(st->best[0]) * st->num_res);
|
|
|
|
st->exit_early = (st->score == st->num_res - skips
|
|
|| st->score == st->num_objs)
|
|
&& st->replaced == 0;
|
|
|
|
return true;
|
|
} else {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
if (st->orig[i] == SKIP) {
|
|
st->res[i] = SKIP;
|
|
return match_obj_(st, skips + 1, score, replaced, i + 1);
|
|
}
|
|
|
|
bool has_best = false;
|
|
|
|
/*
|
|
* Attempt to use the current solution first, to try and avoid
|
|
* recalculating everything
|
|
*/
|
|
if (st->orig[i] != UNMATCHED && !is_taken(i, st->res, st->orig[i])) {
|
|
st->res[i] = st->orig[i];
|
|
size_t obj_score = st->objs[st->res[i]] != 0 ? 1 : 0;
|
|
if (match_obj_(st, skips, score + obj_score, replaced, i + 1)) {
|
|
has_best = true;
|
|
}
|
|
}
|
|
if (st->orig[i] == UNMATCHED) {
|
|
st->res[i] = UNMATCHED;
|
|
if (match_obj_(st, skips, score, replaced, i + 1)) {
|
|
has_best = true;
|
|
}
|
|
}
|
|
if (st->exit_early) {
|
|
return true;
|
|
}
|
|
|
|
if (st->orig[i] != UNMATCHED) {
|
|
++replaced;
|
|
}
|
|
|
|
for (size_t candidate = 0; candidate < st->num_objs; ++candidate) {
|
|
// We tried this earlier
|
|
if (candidate == st->orig[i]) {
|
|
continue;
|
|
}
|
|
|
|
// Not compatible
|
|
if (!(st->objs[candidate] & (1 << i))) {
|
|
continue;
|
|
}
|
|
|
|
// Already taken
|
|
if (is_taken(i, st->res, candidate)) {
|
|
continue;
|
|
}
|
|
|
|
st->res[i] = candidate;
|
|
size_t obj_score = st->objs[candidate] != 0 ? 1 : 0;
|
|
if (match_obj_(st, skips, score + obj_score, replaced, i + 1)) {
|
|
has_best = true;
|
|
}
|
|
|
|
if (st->exit_early) {
|
|
return true;
|
|
}
|
|
}
|
|
|
|
if (has_best) {
|
|
return true;
|
|
}
|
|
|
|
// Maybe this resource can't be matched
|
|
st->res[i] = UNMATCHED;
|
|
return match_obj_(st, skips, score, replaced, i + 1);
|
|
}
|
|
|
|
size_t match_obj(size_t num_objs, const uint32_t objs[static restrict num_objs],
|
|
size_t num_res, const uint32_t res[static restrict num_res],
|
|
uint32_t out[static restrict num_res]) {
|
|
uint32_t solution[num_res];
|
|
for (size_t i = 0; i < num_res; ++i) {
|
|
solution[i] = UNMATCHED;
|
|
}
|
|
|
|
struct match_state st = {
|
|
.num_objs = num_objs,
|
|
.num_res = num_res,
|
|
.score = 0,
|
|
.replaced = SIZE_MAX,
|
|
.objs = objs,
|
|
.res = solution,
|
|
.best = out,
|
|
.orig = res,
|
|
.exit_early = false,
|
|
};
|
|
|
|
match_obj_(&st, 0, 0, 0, 0);
|
|
return st.score;
|
|
}
|