1/*	$NetBSD: machfb.c,v 1.107 2022/09/25 17:52:25 thorpej Exp $	*/
2
3/*
4 * Copyright (c) 2002 Bang Jun-Young
5 * Copyright (c) 2005, 2006, 2007 Michael Lorenz
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. The name of the author may not be used to endorse or promote products
17 *    derived from this software without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 */
30
31/*
32 * Some code is derived from ATI Rage Pro and Derivatives Programmer's Guide.
33 */
34
35#include <sys/cdefs.h>
36__KERNEL_RCSID(0,
37	"$NetBSD: machfb.c,v 1.107 2022/09/25 17:52:25 thorpej Exp $");
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/kernel.h>
42#include <sys/device.h>
43#include <sys/callout.h>
44#include <sys/lwp.h>
45#include <sys/kauth.h>
46
47#include <dev/videomode/videomode.h>
48#include <dev/videomode/edidvar.h>
49
50#include <dev/pci/pcivar.h>
51#include <dev/pci/pcireg.h>
52#include <dev/pci/pcidevs.h>
53#include <dev/pci/pciio.h>
54#include <dev/pci/machfbreg.h>
55
56#include <dev/wscons/wsdisplayvar.h>
57
58#include <dev/wscons/wsconsio.h>
59#include <dev/wsfont/wsfont.h>
60#include <dev/rasops/rasops.h>
61#include <dev/pci/wsdisplay_pci.h>
62
63#include <dev/wscons/wsdisplay_vconsvar.h>
64#include <dev/wscons/wsdisplay_glyphcachevar.h>
65
66#include "opt_wsemul.h"
67#include "opt_machfb.h"
68#include "opt_glyphcache.h"
69
70#ifdef MACHFB_DEBUG
71#define DPRINTF printf
72#else
73#define DPRINTF while (0) printf
74#endif
75
76#define MACH64_REG_SIZE		0x800
77#define MACH64_REG_OFF		0x7ff800
78
79#define	NBARS		3	/* number of Mach64 PCI BARs */
80
81struct vga_bar {
82	bus_addr_t vb_base;
83	bus_size_t vb_size;
84	pcireg_t vb_type;
85	int vb_flags;
86};
87
88struct mach64_softc {
89	device_t sc_dev;
90	pci_chipset_tag_t sc_pc;
91	pcitag_t sc_pcitag;
92
93	struct vga_bar sc_bars[NBARS];
94	struct vga_bar sc_rom;
95
96#define sc_aperbase 	sc_bars[0].vb_base
97#define sc_apersize	sc_bars[0].vb_size
98
99#define sc_iobase	sc_bars[1].vb_base
100#define sc_iosize	sc_bars[1].vb_size
101
102#define sc_regbase	sc_bars[2].vb_base
103#define sc_regsize	sc_bars[2].vb_size
104
105	bus_space_tag_t sc_regt;
106	bus_space_tag_t sc_memt;
107	bus_space_tag_t sc_iot;
108	bus_space_handle_t sc_regh;
109	bus_space_handle_t sc_memh;
110#if 0
111	void *sc_aperture;		/* mapped aperture vaddr */
112	void *sc_registers;		/* mapped registers vaddr */
113#endif
114	uint32_t sc_nbus, sc_ndev, sc_nfunc;
115	size_t memsize;
116	int memtype;
117
118	int sc_mode;
119	int sc_bg;
120	int sc_locked;
121
122	int has_dsp;
123	int bits_per_pixel;
124	int max_x;
125	int max_y;
126	int virt_x;
127	int virt_y;
128	int stride;	/* in pixels */
129	int color_depth;
130
131	int mem_freq;
132	int ramdac_freq;
133	int ref_freq;
134	int vclk_freq;
135
136	int ref_div;
137	int log2_vclk_post_div;
138	int vclk_post_div;
139	int vclk_fb_div;
140	int mclk_post_div;
141	int mclk_fb_div;
142	int sc_clock;	/* which clock to use */
143	int minref, m;
144
145	struct videomode *sc_my_mode;
146	int sc_edid_size;
147	uint8_t sc_edid_data[1024];
148    	struct edid_info sc_ei;
149    	int sc_setmode;
150    	int sc_gen_cntl;
151
152	u_char sc_cmap_red[256];
153	u_char sc_cmap_green[256];
154	u_char sc_cmap_blue[256];
155	int sc_dacw, sc_blanked, sc_console;
156	struct vcons_data vd;
157	struct wsdisplay_accessops sc_accessops;
158	glyphcache sc_gc;
159};
160
161struct mach64_crtcregs {
162	uint32_t h_total_disp;
163	uint32_t h_sync_strt_wid;
164	uint32_t v_total_disp;
165	uint32_t v_sync_strt_wid;
166	uint32_t gen_cntl;
167	uint32_t clock_cntl;
168	uint32_t color_depth;
169	uint32_t dot_clock;
170};
171
172static struct {
173	uint16_t chip_id;
174	uint32_t ramdac_freq;
175} const mach64_info[] = {
176	{ PCI_PRODUCT_ATI_MACH64_GX, 135000 },
177	{ PCI_PRODUCT_ATI_MACH64_CX, 135000 },
178	{ PCI_PRODUCT_ATI_MACH64_CT, 135000 },
179	{ PCI_PRODUCT_ATI_RAGE_PRO_AGP, 230000 },
180	{ PCI_PRODUCT_ATI_RAGE_PRO_AGP1X, 230000 },
181	{ PCI_PRODUCT_ATI_RAGE_PRO_PCI_B, 230000 },
182	{ PCI_PRODUCT_ATI_RAGE_XL_AGP, 230000 },
183	{ PCI_PRODUCT_ATI_RAGE_PRO_PCI_P, 230000 },
184	{ PCI_PRODUCT_ATI_RAGE_PRO_PCI_L, 230000 },
185	{ PCI_PRODUCT_ATI_RAGE_XL_PCI, 230000 },
186	{ PCI_PRODUCT_ATI_RAGE_XL_PCI66, 230000 },
187	{ PCI_PRODUCT_ATI_RAGE_II, 135000 },
188	{ PCI_PRODUCT_ATI_RAGE_IIP, 200000 },
189	{ PCI_PRODUCT_ATI_RAGE_IIC_PCI, 230000 },
190	{ PCI_PRODUCT_ATI_RAGE_IIC_AGP_B, 230000 },
191	{ PCI_PRODUCT_ATI_RAGE_IIC_AGP_P, 230000 },
192#if 0
193	{ PCI_PRODUCT_ATI_RAGE_MOB_M3_PCI, 230000 },
194	{ PCI_PRODUCT_ATI_RAGE_MOB_M3_AGP, 230000 },
195	{ PCI_PRODUCT_ATI_RAGE_MOBILITY, 230000 },
196#endif
197	{ PCI_PRODUCT_ATI_RAGE_L_MOB_M1_PCI, 230000 },
198	{ PCI_PRODUCT_ATI_RAGE_LT_PRO_AGP, 230000 },
199	{ PCI_PRODUCT_ATI_RAGE_LT_PRO, 230000 },
200	{ PCI_PRODUCT_ATI_RAGE_LT, 230000 },
201	{ PCI_PRODUCT_ATI_RAGE_LT_PRO_PCI, 230000 },
202	{ PCI_PRODUCT_ATI_MACH64_VT, 170000 },
203	{ PCI_PRODUCT_ATI_MACH64_VTB, 200000 },
204	{ PCI_PRODUCT_ATI_MACH64_VT4, 230000 }
205};
206
207static int mach64_chip_id, mach64_chip_rev;
208static struct videomode default_mode = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, };
209
210static const char *mach64_gx_memtype_names[] = {
211	"DRAM", "VRAM", "VRAM", "DRAM",
212	"DRAM", "VRAM", "VRAM", "(unknown type)"
213};
214
215static const char *mach64_memtype_names[] = {
216	"(N/A)", "DRAM", "EDO DRAM", "EDO DRAM", "SDRAM", "SGRAM", "WRAM",
217	"(unknown type)"
218};
219
220extern const u_char rasops_cmap[768];
221
222static int	mach64_match(device_t, cfdata_t, void *);
223static void	mach64_attach(device_t, device_t, void *);
224
225CFATTACH_DECL_NEW(machfb, sizeof(struct mach64_softc), mach64_match,
226    mach64_attach, NULL, NULL);
227
228static void	mach64_init(struct mach64_softc *);
229static int	mach64_get_memsize(struct mach64_softc *);
230static int	mach64_get_max_ramdac(struct mach64_softc *);
231static int	mach64_ref_freq(void);
232
233#ifdef MACHFB_DEBUG
234static void	mach64_get_mode(struct mach64_softc *, struct videomode *);
235static void	mach64_print_reg(struct mach64_softc *);
236#endif
237
238static int	mach64_calc_crtcregs(struct mach64_softc *,
239				     struct mach64_crtcregs *,
240				     struct videomode *);
241static void	mach64_set_crtcregs(struct mach64_softc *,
242				    struct mach64_crtcregs *);
243
244static int	mach64_modeswitch(struct mach64_softc *, struct videomode *);
245static void	mach64_set_dsp(struct mach64_softc *);
246static void	mach64_set_pll(struct mach64_softc *, int);
247static void	mach64_reset_engine(struct mach64_softc *);
248static void	mach64_init_engine(struct mach64_softc *);
249#if 0
250static void	mach64_adjust_frame(struct mach64_softc *, int, int);
251#endif
252static void	mach64_init_lut(struct mach64_softc *);
253
254static void	mach64_init_screen(void *, struct vcons_screen *, int, long *);
255static int	mach64_is_console(struct mach64_softc *);
256
257static void	mach64_cursor(void *, int, int, int);
258#if 0
259static int	mach64_mapchar(void *, int, u_int *);
260#endif
261static void	mach64_putchar_mono(void *, int, int, u_int, long);
262static void	mach64_putchar_aa8(void *, int, int, u_int, long);
263static void	mach64_copycols(void *, int, int, int, int);
264static void	mach64_erasecols(void *, int, int, int, long);
265static void	mach64_copyrows(void *, int, int, int);
266static void	mach64_eraserows(void *, int, int, long);
267static void 	mach64_clearscreen(struct mach64_softc *);
268
269static int	mach64_putcmap(struct mach64_softc *, struct wsdisplay_cmap *);
270static int	mach64_getcmap(struct mach64_softc *, struct wsdisplay_cmap *);
271static int	mach64_putpalreg(struct mach64_softc *, uint8_t, uint8_t,
272				 uint8_t, uint8_t);
273static void	mach64_bitblt(void *, int, int, int, int, int, int, int);
274static void	mach64_rectfill(struct mach64_softc *, int, int, int, int, int);
275static void	mach64_setup_mono(struct mach64_softc *, int, int, int, int,
276				  uint32_t, uint32_t);
277static void	mach64_feed_bytes(struct mach64_softc *, int, uint8_t *);
278#if 0
279static void	mach64_showpal(struct mach64_softc *);
280#endif
281
282static void	machfb_blank(struct mach64_softc *, int);
283static int	machfb_drm_print(void *, const char *);
284
285static struct wsscreen_descr mach64_defaultscreen = {
286	"default",
287	80, 30,
288	NULL,
289	8, 16,
290	WSSCREEN_WSCOLORS | WSSCREEN_HILIT | WSSCREEN_UNDERLINE
291	 | WSSCREEN_RESIZE ,
292	NULL
293};
294
295static const struct wsscreen_descr *_mach64_scrlist[] = {
296	&mach64_defaultscreen,
297};
298
299static struct wsscreen_list mach64_screenlist = {
300	__arraycount(_mach64_scrlist),
301	_mach64_scrlist
302};
303
304static int	mach64_ioctl(void *, void *, u_long, void *, int,
305		             struct lwp *);
306static paddr_t	mach64_mmap(void *, void *, off_t, int);
307
308static struct vcons_screen mach64_console_screen;
309
310/*
311 * Inline functions for getting access to register aperture.
312 */
313
314static inline uint32_t
315regr(struct mach64_softc *sc, uint32_t index)
316{
317	return bus_space_read_4(sc->sc_regt, sc->sc_regh, index + 0x400);
318}
319
320static inline uint8_t
321regrb(struct mach64_softc *sc, uint32_t index)
322{
323	return bus_space_read_1(sc->sc_regt, sc->sc_regh, index + 0x400);
324}
325
326static inline void
327regw(struct mach64_softc *sc, uint32_t index, uint32_t data)
328{
329	bus_space_write_4(sc->sc_regt, sc->sc_regh, index + 0x400, data);
330	bus_space_barrier(sc->sc_regt, sc->sc_regh, index, 4,
331	    BUS_SPACE_BARRIER_WRITE);
332}
333
334static inline void
335regws(struct mach64_softc *sc, uint32_t index, uint32_t data)
336{
337	bus_space_write_stream_4(sc->sc_regt, sc->sc_regh, index + 0x400, data);
338	bus_space_barrier(sc->sc_regt, sc->sc_regh, index + 0x400, 4,
339	    BUS_SPACE_BARRIER_WRITE);
340}
341
342static inline void
343regwb(struct mach64_softc *sc, uint32_t index, uint8_t data)
344{
345	bus_space_write_1(sc->sc_regt, sc->sc_regh, index + 0x400, data);
346	bus_space_barrier(sc->sc_regt, sc->sc_regh, index + 0x400, 1,
347	    BUS_SPACE_BARRIER_WRITE);
348}
349
350static inline void
351regwb_pll(struct mach64_softc *sc, uint32_t index, uint8_t data)
352{
353	uint32_t reg;
354
355	reg = regr(sc, CLOCK_CNTL);
356	reg |= PLL_WR_EN;
357	regw(sc, CLOCK_CNTL, reg);
358	reg &= ~(PLL_ADDR | PLL_DATA);
359	reg |= (index & 0x3f) << PLL_ADDR_SHIFT;
360	reg |= data << PLL_DATA_SHIFT;
361	reg |= CLOCK_STROBE;
362	regw(sc, CLOCK_CNTL, reg);
363	reg &= ~PLL_WR_EN;
364	regw(sc, CLOCK_CNTL, reg);
365}
366
367static inline uint8_t
368regrb_pll(struct mach64_softc *sc, uint32_t index)
369{
370
371	regwb(sc, CLOCK_CNTL + 1, index << 2);
372	return regrb(sc, CLOCK_CNTL + 2);
373}
374
375static inline void
376wait_for_fifo(struct mach64_softc *sc, uint8_t v)
377{
378	while ((regr(sc, FIFO_STAT) & 0xffff) > (0x8000 >> v))
379		continue;
380}
381
382static inline void
383wait_for_idle(struct mach64_softc *sc)
384{
385	wait_for_fifo(sc, 16);
386	while ((regr(sc, GUI_STAT) & 1) != 0)
387		continue;
388}
389
390static int
391mach64_match(device_t parent, cfdata_t match, void *aux)
392{
393	struct pci_attach_args *pa = (struct pci_attach_args *)aux;
394	int i;
395
396	if (PCI_CLASS(pa->pa_class) != PCI_CLASS_DISPLAY ||
397	    PCI_SUBCLASS(pa->pa_class) != PCI_SUBCLASS_DISPLAY_VGA)
398		return 0;
399
400	for (i = 0; i < __arraycount(mach64_info); i++)
401		if (PCI_PRODUCT(pa->pa_id) == mach64_info[i].chip_id) {
402			mach64_chip_id = PCI_PRODUCT(pa->pa_id);
403			mach64_chip_rev = PCI_REVISION(pa->pa_class);
404			return 100;
405		}
406
407	return 0;
408}
409
410static void
411mach64_attach(device_t parent, device_t self, void *aux)
412{
413	struct mach64_softc *sc = device_private(self);
414	struct pci_attach_args *pa = aux;
415	struct rasops_info *ri;
416	const char *mptr = NULL;
417	prop_data_t edid_data;
418	const struct videomode *mode = NULL;
419	int bar, id, expected_id;
420	int is_gx;
421	const char **memtype_names;
422	struct wsemuldisplaydev_attach_args aa;
423	long defattr;
424	int width = 1024, height = 768;
425	pcireg_t screg;
426	uint32_t reg;
427	const pcireg_t enables = PCI_COMMAND_MEM_ENABLE;
428	int use_mmio = FALSE;
429
430	sc->sc_dev = self;
431	sc->sc_pc = pa->pa_pc;
432	sc->sc_pcitag = pa->pa_tag;
433	sc->sc_dacw = -1;
434	sc->sc_mode = WSDISPLAYIO_MODE_EMUL;
435	sc->sc_nbus = pa->pa_bus;
436	sc->sc_ndev = pa->pa_device;
437	sc->sc_nfunc = pa->pa_function;
438	sc->sc_locked = 0;
439	sc->sc_iot = pa->pa_iot;
440	sc->sc_accessops.ioctl = mach64_ioctl;
441	sc->sc_accessops.mmap = mach64_mmap;
442	sc->sc_setmode = 0;
443
444	pci_aprint_devinfo(pa, "Graphics processor");
445#ifdef MACHFB_DEBUG
446	printf(prop_dictionary_externalize(device_properties(self)));
447#endif
448
449	/* enable memory access */
450	screg = pci_conf_read(sc->sc_pc, sc->sc_pcitag, PCI_COMMAND_STATUS_REG);
451	if ((screg & enables) != enables) {
452		screg |= enables;
453		pci_conf_write(sc->sc_pc, sc->sc_pcitag,
454		    PCI_COMMAND_STATUS_REG, screg);
455	}
456	for (bar = 0; bar < NBARS; bar++) {
457		reg = PCI_MAPREG_START + (bar * 4);
458		sc->sc_bars[bar].vb_type = pci_mapreg_type(sc->sc_pc,
459		    sc->sc_pcitag, reg);
460		(void)pci_mapreg_info(sc->sc_pc, sc->sc_pcitag, reg,
461		    sc->sc_bars[bar].vb_type, &sc->sc_bars[bar].vb_base,
462		    &sc->sc_bars[bar].vb_size, &sc->sc_bars[bar].vb_flags);
463	}
464	aprint_debug_dev(sc->sc_dev, "aperture size %08x\n",
465	    (uint32_t)sc->sc_apersize);
466
467	sc->sc_rom.vb_type = PCI_MAPREG_TYPE_ROM;
468	pci_mapreg_info(sc->sc_pc, sc->sc_pcitag, PCI_MAPREG_ROM,
469		    sc->sc_rom.vb_type, &sc->sc_rom.vb_base,
470		    &sc->sc_rom.vb_size, &sc->sc_rom.vb_flags);
471	sc->sc_memt = pa->pa_memt;
472
473	/* use MMIO register aperture if available */
474	if ((sc->sc_regbase != 0) && (sc->sc_regbase != 0xffffffff)) {
475		if (pci_mapreg_map(pa, MACH64_BAR_MMIO,  PCI_MAPREG_TYPE_MEM,
476		    0, &sc->sc_regt, &sc->sc_regh, &sc->sc_regbase,
477		    &sc->sc_regsize) == 0) {
478
479			/*
480			 * the MMIO aperture maps both 1KB register blocks, but
481			 * all register offsets are relative to the 2nd one so
482			 * for now fix this up in MACH64_REG_OFF and the access
483			 * functions
484			 */
485			aprint_normal_dev(sc->sc_dev, "using MMIO aperture\n");
486			use_mmio = TRUE;
487		}
488	}
489	if (!use_mmio) {
490		if (bus_space_map(sc->sc_memt, sc->sc_aperbase,
491		    sc->sc_apersize, BUS_SPACE_MAP_LINEAR, &sc->sc_memh)) {
492			panic("%s: failed to map aperture",
493			    device_xname(sc->sc_dev));
494		}
495
496		sc->sc_regt = sc->sc_memt;
497		bus_space_subregion(sc->sc_regt, sc->sc_memh, MACH64_REG_OFF,
498		    MACH64_REG_SIZE, &sc->sc_regh);
499	}
500
501	mach64_init(sc);
502
503	aprint_normal_dev(sc->sc_dev,
504	    "%d MB aperture at 0x%08x, %d KB registers at 0x%08x\n",
505	    (u_int)(sc->sc_apersize / (1024 * 1024)),
506	    (u_int)sc->sc_aperbase, (u_int)(sc->sc_regsize / 1024),
507	    (u_int)sc->sc_regbase);
508
509	printf("%s: %d KB ROM at 0x%08x\n", device_xname(sc->sc_dev),
510	    (int)sc->sc_rom.vb_size >> 10, (uint32_t)sc->sc_rom.vb_base);
511#ifdef MACHFB_DEBUG
512	mach64_get_mode(sc, NULL);
513	mach64_print_reg(sc);
514#endif
515
516	prop_dictionary_get_uint32(device_properties(self), "width", &width);
517	prop_dictionary_get_uint32(device_properties(self), "height", &height);
518
519	default_mode.hdisplay = width;
520	default_mode.vdisplay = height;
521
522	prop_dictionary_get_string(device_properties(sc->sc_dev),
523	    "videomode", &mptr);
524
525	memset(&sc->sc_ei, 0, sizeof(sc->sc_ei));
526	if (mptr == NULL &&
527	    (edid_data = prop_dictionary_get(device_properties(self), "EDID"))
528	    != NULL) {
529
530		sc->sc_edid_size = uimin(1024, prop_data_size(edid_data));
531		memset(sc->sc_edid_data, 0, sizeof(sc->sc_edid_data));
532		memcpy(sc->sc_edid_data, prop_data_value(edid_data),
533		    sc->sc_edid_size);
534
535		edid_parse(sc->sc_edid_data, &sc->sc_ei);
536
537#ifdef MACHFB_DEBUG
538		edid_print(&sc->sc_ei);
539#endif
540	}
541	is_gx = 0;
542	switch(mach64_chip_id) {
543		case PCI_PRODUCT_ATI_MACH64_GX:
544		case PCI_PRODUCT_ATI_MACH64_CX:
545			is_gx = 1;
546			/* FALLTHROUGH */
547		case PCI_PRODUCT_ATI_MACH64_CT:
548			sc->has_dsp = 0;
549			break;
550		case PCI_PRODUCT_ATI_MACH64_VT:
551		case PCI_PRODUCT_ATI_RAGE_II:
552			if((mach64_chip_rev & 0x07) == 0) {
553				sc->has_dsp = 0;
554				break;
555			}
556			/* FALLTHROUGH */
557		default:
558			sc->has_dsp = 1;
559	}
560
561	memtype_names = is_gx ? mach64_gx_memtype_names : mach64_memtype_names;
562
563	sc->memsize = mach64_get_memsize(sc);
564
565	if(is_gx)
566		sc->memtype = (regr(sc, CONFIG_STAT0) >> 3) & 0x07;
567	else
568		sc->memtype = regr(sc, CONFIG_STAT0) & 0x07;
569
570	sc->ref_freq = mach64_ref_freq();
571
572	reg = regr(sc, CLOCK_CNTL);
573	sc->sc_clock = reg & 3;
574	DPRINTF("using clock %d\n", sc->sc_clock);
575
576	DPRINTF("ref_freq: %d\n", sc->ref_freq);
577	sc->ref_div = regrb_pll(sc, PLL_REF_DIV);
578	DPRINTF("ref_div: %d\n", sc->ref_div);
579	sc->mclk_fb_div = regrb_pll(sc, MCLK_FB_DIV);
580	DPRINTF("mclk_fb_div: %d\n", sc->mclk_fb_div);
581	sc->mem_freq = (2 * sc->ref_freq * sc->mclk_fb_div) /
582	    (sc->ref_div * 2);
583	sc->mclk_post_div = (sc->mclk_fb_div * 2 * sc->ref_freq) /
584	    (sc->mem_freq * sc->ref_div);
585	sc->ramdac_freq = mach64_get_max_ramdac(sc);
586	{
587		sc->minref = sc->ramdac_freq / 510;
588		sc->m = sc->ref_freq / sc->minref;
589		DPRINTF("minref: %d m: %d\n", sc->minref, sc->m);
590	}
591	aprint_normal_dev(sc->sc_dev,
592	    "%ld KB %s %d.%d MHz, maximum RAMDAC clock %d MHz\n",
593	    (u_long)sc->memsize,
594	    memtype_names[sc->memtype],
595	    sc->mem_freq / 1000, sc->mem_freq % 1000,
596	    sc->ramdac_freq / 1000);
597
598	id = regr(sc, CONFIG_CHIP_ID) & 0xffff;
599	switch(mach64_chip_id) {
600		case PCI_PRODUCT_ATI_MACH64_GX:
601			expected_id = 0x00d7;
602			break;
603		case PCI_PRODUCT_ATI_MACH64_CX:
604			expected_id = 0x0057;
605			break;
606		default:
607			/* Most chip IDs match their PCI product ID. */
608			expected_id = mach64_chip_id;
609	}
610
611	if (id != expected_id) {
612		aprint_error_dev(sc->sc_dev,
613		    "chip ID mismatch, 0x%x != 0x%x\n", id, expected_id);
614		return;
615	}
616
617	sc->sc_console = mach64_is_console(sc);
618	sc->sc_gen_cntl = regr(sc, CRTC_GEN_CNTL);
619	aprint_debug("gen_cntl: %08x\n", sc->sc_gen_cntl);
620	sc->sc_gen_cntl &= CRTC_CSYNC_EN;
621	aprint_normal_dev(sc->sc_dev, "found composite sync %s\n",
622	    sc->sc_gen_cntl ? "enabled" : "disabled");
623
624#define MODE_IS_VALID(m) ((sc->ramdac_freq >= (m)->dot_clock) && \
625			  ((m)->hdisplay <= 1280))
626
627	/* no mode setting support on ancient chips with external clocks */
628	sc->sc_setmode = 0;
629	if (!is_gx) {
630		/*
631		 * Now pick a mode.
632		 */
633		if ((sc->sc_ei.edid_preferred_mode != NULL)) {
634			struct videomode *m = sc->sc_ei.edid_preferred_mode;
635			if (MODE_IS_VALID(m)) {
636				memcpy(&default_mode, m,
637				    sizeof(struct videomode));
638				sc->sc_setmode = 1;
639			} else {
640				aprint_normal_dev(sc->sc_dev,
641				    "unable to use EDID preferred mode "
642				    "(%d x %d)\n", m->hdisplay, m->vdisplay);
643			}
644		}
645		/*
646		 * if we can't use the preferred mode go look for the
647		 * best one we can support
648		 */
649		if (sc->sc_setmode == 0) {
650			struct videomode *m = sc->sc_ei.edid_modes;
651
652			mode = NULL;
653			sort_modes(sc->sc_ei.edid_modes,
654			    &sc->sc_ei.edid_preferred_mode,
655			    sc->sc_ei.edid_nmodes);
656			for (int n = 0; n < sc->sc_ei.edid_nmodes; n++)
657				if (MODE_IS_VALID(&m[n])) {
658					mode = &m[n];
659					break;
660				}
661			if (mode != NULL) {
662				memcpy(&default_mode, mode,
663				    sizeof(struct videomode));
664				sc->sc_setmode = 1;
665			}
666		}
667	}
668
669	/* make sure my_mode points at something sensible if the above fails */
670	if (default_mode.dot_clock == 0) {
671		sc->sc_setmode = 0;
672		mode = pick_mode_by_ref(width, height, 60);
673		if (mode != NULL) {
674			memcpy(&default_mode, mode, sizeof(default_mode));
675		} else if ((width > 0) && (height > 0)) {
676			default_mode.hdisplay = width;
677			default_mode.vdisplay = height;
678		} else {
679			/*
680			 * if we end up here we're probably dealing with
681			 * uninitialized hardware - try to set 1024x768@60 and
682			 * hope for the best...
683			 */
684			mode = pick_mode_by_ref(1024, 768, 60);
685			if (mode == NULL) return;
686			memcpy(&default_mode, mode, sizeof(default_mode));
687			if (!is_gx) sc->sc_setmode = 1;
688		}
689	}
690
691	sc->sc_my_mode = &default_mode;
692
693	if ((width == sc->sc_my_mode->hdisplay) &&
694	    (height == sc->sc_my_mode->vdisplay))
695		sc->sc_setmode = 0;
696
697	sc->bits_per_pixel = 8;
698	sc->virt_x = sc->sc_my_mode->hdisplay;
699	sc->virt_y = sc->sc_my_mode->vdisplay;
700	sc->stride = (sc->virt_x + 7) & ~7;	/* hw needs multiples of 8 */
701	sc->max_x = sc->virt_x - 1;
702	sc->max_y = (sc->memsize * 1024) /
703	    (sc->stride * (sc->bits_per_pixel / 8)) - 1;
704
705	sc->color_depth = CRTC_PIX_WIDTH_8BPP;
706
707	mach64_init_engine(sc);
708
709	if (sc->sc_setmode)
710		mach64_modeswitch(sc, sc->sc_my_mode);
711
712	aprint_normal_dev(sc->sc_dev,
713	    "initial resolution %dx%d at %d bpp\n",
714	    sc->sc_my_mode->hdisplay, sc->sc_my_mode->vdisplay,
715	    sc->bits_per_pixel);
716
717	wsfont_init();
718
719#ifdef GLYPHCACHE_DEBUG
720	/* shrink the screen so we can see part of the glyph cache */
721	sc->sc_my_mode->vdisplay -= 200;
722#endif
723
724	vcons_init(&sc->vd, sc, &mach64_defaultscreen, &sc->sc_accessops);
725	sc->vd.init_screen = mach64_init_screen;
726	sc->vd.show_screen_cookie = &sc->sc_gc;
727	sc->vd.show_screen_cb = glyphcache_adapt;
728
729	sc->sc_gc.gc_bitblt = mach64_bitblt;
730	sc->sc_gc.gc_blitcookie = sc;
731	sc->sc_gc.gc_rop = MIX_SRC;
732
733	ri = &mach64_console_screen.scr_ri;
734	if (sc->sc_console) {
735
736		vcons_init_screen(&sc->vd, &mach64_console_screen, 1,
737		    &defattr);
738		mach64_console_screen.scr_flags |= VCONS_SCREEN_IS_STATIC;
739
740		mach64_defaultscreen.textops = &ri->ri_ops;
741		mach64_defaultscreen.capabilities = ri->ri_caps;
742		mach64_defaultscreen.nrows = ri->ri_rows;
743		mach64_defaultscreen.ncols = ri->ri_cols;
744		glyphcache_init(&sc->sc_gc, sc->sc_my_mode->vdisplay + 5,
745		    ((sc->memsize * 1024) / sc->stride) -
746		      sc->sc_my_mode->vdisplay - 5,
747		    sc->stride,
748		    ri->ri_font->fontwidth,
749		    ri->ri_font->fontheight,
750		    defattr);
751		wsdisplay_cnattach(&mach64_defaultscreen, ri, 0, 0, defattr);
752	} else {
753		/*
754		 * since we're not the console we can postpone the rest
755		 * until someone actually allocates a screen for us
756		 */
757		if (mach64_console_screen.scr_ri.ri_rows == 0) {
758			/* do some minimal setup to avoid weirdnesses later */
759			vcons_init_screen(&sc->vd, &mach64_console_screen, 1,
760			    &defattr);
761		} else
762			(*ri->ri_ops.allocattr)(ri, 0, 0, 0, &defattr);
763
764		glyphcache_init(&sc->sc_gc, sc->sc_my_mode->vdisplay + 5,
765		    ((sc->memsize * 1024) / sc->stride) -
766		      sc->sc_my_mode->vdisplay - 5,
767		    sc->stride,
768		    ri->ri_font->fontwidth,
769		    ri->ri_font->fontheight,
770		    defattr);
771	}
772
773	sc->sc_bg = mach64_console_screen.scr_ri.ri_devcmap[WS_DEFAULT_BG];
774	mach64_clearscreen(sc);
775	mach64_init_lut(sc);
776
777	if (sc->sc_console)
778		vcons_replay_msgbuf(&mach64_console_screen);
779
780	machfb_blank(sc, 0);	/* unblank the screen */
781
782	aa.console = sc->sc_console;
783	aa.scrdata = &mach64_screenlist;
784	aa.accessops = &sc->sc_accessops;
785	aa.accesscookie = &sc->vd;
786
787	config_found(self, &aa, wsemuldisplaydevprint,
788	    CFARGS(.iattr = "wsemuldisplaydev"));
789#if 0
790	/* XXX
791	 * turns out some firmware doesn't turn these back on when needed
792	 * so we need to turn them off only when mapping vram in
793	 * WSDISPLAYIO_MODE_DUMB would overlap ( unlikely but far from
794	 * impossible )
795	 */
796	if (use_mmio) {
797		/*
798		 * Now that we took over, turn off the aperture registers if we
799		 * don't use them. Can't do this earlier since on some hardware
800		 * we use firmware calls as early console output which may in
801		 * turn try to access these registers.
802		 */
803		reg = regr(sc, BUS_CNTL);
804		aprint_debug_dev(sc->sc_dev, "BUS_CNTL: %08x\n", reg);
805		reg |= BUS_APER_REG_DIS;
806		regw(sc, BUS_CNTL, reg);
807	}
808#endif
809	config_found(self, aux, machfb_drm_print,
810	    CFARGS(.iattr = "drm"));
811}
812
813static int
814machfb_drm_print(void *aux, const char *pnp)
815{
816	if (pnp)
817		aprint_normal("direct rendering for %s", pnp);
818	return (UNSUPP);
819}
820
821static void
822mach64_init_screen(void *cookie, struct vcons_screen *scr, int existing,
823    long *defattr)
824{
825	struct mach64_softc *sc = cookie;
826	struct rasops_info *ri = &scr->scr_ri;
827
828	ri->ri_depth = sc->bits_per_pixel;
829	ri->ri_width = sc->sc_my_mode->hdisplay;
830	ri->ri_height = sc->sc_my_mode->vdisplay;
831	ri->ri_stride = sc->stride;
832	ri->ri_flg = RI_CENTER | RI_FULLCLEAR;
833	if (ri->ri_depth == 8)
834		ri->ri_flg |= RI_8BIT_IS_RGB | RI_ENABLE_ALPHA |
835			      RI_PREFER_ALPHA;
836
837#ifdef VCONS_DRAW_INTR
838	scr->scr_flags |= VCONS_DONT_READ;
839#endif
840	scr->scr_flags |= VCONS_LOADFONT;
841
842	rasops_init(ri, 0, 0);
843	ri->ri_caps = WSSCREEN_WSCOLORS | WSSCREEN_HILIT | WSSCREEN_UNDERLINE |
844		      WSSCREEN_RESIZE;
845	rasops_reconfig(ri, sc->sc_my_mode->vdisplay / ri->ri_font->fontheight,
846		    sc->sc_my_mode->hdisplay / ri->ri_font->fontwidth);
847
848	/* enable acceleration */
849	ri->ri_hw = scr;
850	ri->ri_ops.copyrows = mach64_copyrows;
851	ri->ri_ops.copycols = mach64_copycols;
852	ri->ri_ops.eraserows = mach64_eraserows;
853	ri->ri_ops.erasecols = mach64_erasecols;
854	ri->ri_ops.cursor = mach64_cursor;
855	if (FONT_IS_ALPHA(ri->ri_font)) {
856		ri->ri_ops.putchar = mach64_putchar_aa8;
857	} else
858		ri->ri_ops.putchar = mach64_putchar_mono;
859}
860
861static void
862mach64_init(struct mach64_softc *sc)
863{
864	sc->sc_blanked = 0;
865}
866
867static int
868mach64_get_memsize(struct mach64_softc *sc)
869{
870	int tmp, memsize;
871	int mem_tab[] = {
872		512, 1024, 2048, 4096, 6144, 8192, 12288, 16384
873	};
874	tmp = regr(sc, MEM_CNTL);
875#ifdef DIAGNOSTIC
876	aprint_debug_dev(sc->sc_dev, "memctl %08x\n", tmp);
877#endif
878	if (sc->has_dsp) {
879		tmp &= 0x0000000f;
880		if (tmp < 8)
881			memsize = (tmp + 1) * 512;
882		else if (tmp < 12)
883			memsize = (tmp - 3) * 1024;
884		else
885			memsize = (tmp - 7) * 2048;
886	} else {
887		memsize = mem_tab[tmp & 0x07];
888	}
889
890	return memsize;
891}
892
893static int
894mach64_get_max_ramdac(struct mach64_softc *sc)
895{
896	int i;
897
898	if ((mach64_chip_id == PCI_PRODUCT_ATI_MACH64_VT ||
899	     mach64_chip_id == PCI_PRODUCT_ATI_RAGE_II) &&
900	     (mach64_chip_rev & 0x07))
901		return 170000;
902
903	for (i = 0; i < __arraycount(mach64_info); i++)
904		if (mach64_chip_id == mach64_info[i].chip_id)
905			return mach64_info[i].ramdac_freq;
906
907	if (sc->bits_per_pixel == 8)
908		return 135000;
909	else
910		return 80000;
911}
912
913static int
914mach64_ref_freq(void)
915{
916	/*
917	 * There doesn't seem to be any way to calculate the reference
918	 * frequency from known values
919	 */
920	if ((mach64_chip_id == PCI_PRODUCT_ATI_RAGE_XL_PCI) ||
921	    ((mach64_chip_id >= PCI_PRODUCT_ATI_RAGE_LT_PRO_PCI) &&
922	    (mach64_chip_id <= PCI_PRODUCT_ATI_RAGE_L_MOB_M1_PCI)))
923		return 29498;
924	else
925		return 14318;
926}
927
928#ifdef MACHFB_DEBUG
929static void
930mach64_get_mode(struct mach64_softc *sc, struct videomode *mode)
931{
932	int htotal, hdisplay, hsync_start, hsync_end;
933	int vtotal, vdisplay, vsync_start, vsync_end;
934	int clk_ctl, clock;
935	int ref_freq, ref_div, vclk_post_div, vclk_fb_div;
936	int nhsync, nvsync;
937	int post_div, dot_clock, vrefresh, vrefresh2;
938
939	hdisplay = regr(sc, CRTC_H_TOTAL_DISP);
940	hsync_end = regr(sc, CRTC_H_SYNC_STRT_WID);
941	vdisplay = regr(sc, CRTC_V_TOTAL_DISP);
942	vsync_end = regr(sc, CRTC_V_SYNC_STRT_WID);
943	clk_ctl = regr(sc, CLOCK_CNTL);
944	clock = clk_ctl & 3;
945	ref_div = regrb_pll(sc, PLL_REF_DIV);
946	vclk_post_div = regrb_pll(sc, VCLK_POST_DIV);
947	vclk_fb_div = regrb_pll(sc, VCLK0_FB_DIV + clock);
948	ref_freq = mach64_ref_freq();
949
950	htotal = ((hdisplay & 0x01ff) + 1) << 3;
951	hdisplay = (((hdisplay & 0x1ff0000) >> 16) + 1) << 3;
952	if (hsync_end & CRTC_HSYNC_NEG)
953		nhsync = 1;
954	else
955		nhsync = 0;
956	hsync_start = (((hsync_end & 0xff) + 1) << 3) +
957	    ((hsync_end & 0x700) >> 8);
958	hsync_end = (((hsync_end & 0x1f0000) >> 16) << 3) + hsync_start;
959
960	vtotal = (vdisplay & 0x07ff) + 1;
961	vdisplay = ((vdisplay & 0x7ff0000) >> 16) + 1;
962	if (vsync_end & CRTC_VSYNC_NEG)
963		nvsync = 1;
964	else
965		nvsync = 0;
966	vsync_start = (vsync_end & 0x07ff) + 1;
967	vsync_end = ((vsync_end & 0x1f0000) >> 16) + vsync_start;
968
969	switch ((vclk_post_div >> (clock * 2)) & 3) {
970		case 3:
971			post_div = 8;
972			break;
973		case 2:
974			post_div = 4;
975			break;
976		case 1:
977			post_div = 2;
978			break;
979		default:
980			post_div = 1;
981			break;
982	}
983	dot_clock = (2 * ref_freq * vclk_fb_div) / (ref_div * post_div);
984	vrefresh = (dot_clock * 1000) / (htotal * vtotal);
985	vrefresh2 = ((dot_clock * 1000) - (vrefresh * htotal * vtotal)) * 100 /
986	    (htotal * vtotal);
987
988	aprint_normal_dev(sc->sc_dev, "Video mode:\n");
989	aprint_normal("\t%d" "x%d @ %d.%02dHz "
990	    "(%d %d %d %d %d %d %d %cH %cV)\n",
991	    hdisplay, vdisplay, vrefresh, vrefresh2, dot_clock,
992	    hsync_start, hsync_end, htotal, vsync_start, vsync_end, vtotal,
993	    nhsync == 1 ? '-' : '+', nvsync == 1 ? '-' : '+');
994
995	if (mode != NULL) {
996		mode->dot_clock = dot_clock;
997		mode->htotal = htotal;
998		mode->hdisplay = hdisplay;
999		mode->hsync_start = hsync_start;
1000		mode->hsync_end = hsync_end;
1001		mode->vtotal = vtotal;
1002		mode->vdisplay = vdisplay;
1003		mode->vsync_start = vsync_start;
1004		mode->vsync_end = vsync_end;
1005		mode->flags = 0;
1006		if (nhsync)
1007			mode->flags |= VID_NHSYNC;
1008		if (nvsync)
1009			mode->flags |= VID_NVSYNC;
1010	}
1011}
1012
1013static void
1014mach64_print_reg(struct mach64_softc *sc)
1015{
1016	struct reglist {
1017		int offset;
1018		const char *name;
1019	};
1020	static const struct reglist reglist_tab[] = {
1021		{ 0x0000, "CRTC_H_TOTAL_DISP" },
1022		{ 0x0004, "CRTC_H_SYNC_STRT_WID" },
1023		{ 0x0008, "CRTC_V_TOTAL_DISP" },
1024		{ 0x000C, "CRTC_V_SYNC_STRT_WID" },
1025		{ 0x0010, "CRTC_VLINE_CRNT_VLINE" },
1026		{ 0x0014, "CRTC_OFF_PITCH" },
1027		{ 0x001C, "CRTC_GEN_CNTL" },
1028		{ 0x0090, "CLOCK_CNTL" },
1029		{ 0, NULL }
1030	};
1031	static const struct reglist plllist_tab[] = {
1032		{ 0x02, "PLL_REF_DIV" },
1033		{ 0x03, "PLL_GEN_CNTL" },
1034		{ 0x04, "MCLK_FB_DIV" },
1035		{ 0x05, "PLL_VCLK_CNTL" },
1036		{ 0x06, "VCLK_POST_DIV" },
1037		{ 0x07, "VCLK0_FB_DIV" },
1038		{ 0x08, "VCLK1_FB_DIV" },
1039		{ 0x09, "VCLK2_FB_DIV" },
1040		{ 0x0A, "VCLK3_FB_DIV" },
1041		{ 0x0B, "PLL_XCLK_CNTL" },
1042		{ 0x10, "LVDSPLL_CNTL0" },
1043		{ 0x11, "LVDSPLL_CNTL0" },
1044		{ 0x19, "EXT_VPLL_CNTL" },
1045		{ 0x1A, "EXT_VPLL_REF_DIV" },
1046		{ 0x1B, "EXT_VPLL_FB_DIV" },
1047		{ 0x1C, "EXT_VPLL_MSB" },
1048		{ 0, NULL }
1049	};
1050	const struct reglist *r;
1051
1052	aprint_normal("CRTC registers\n");
1053	for (r = reglist_tab; r->name != NULL; r++)
1054		aprint_normal("0x%04x 0x%08x %s\n", r->offset,
1055		    regr(sc, r->offset), r->name);
1056	aprint_normal("PLL registers\n");
1057	for (r = plllist_tab; r->name != NULL; r++)
1058		aprint_normal("0x%02x 0x%02x %s\n", r->offset,
1059		    regrb_pll(sc, r->offset), r->name);
1060}
1061#endif
1062
1063static int
1064mach64_calc_crtcregs(struct mach64_softc *sc, struct mach64_crtcregs *crtc,
1065    struct videomode *mode)
1066{
1067
1068	if (mode->dot_clock > sc->ramdac_freq)
1069		/* Clock too high. */
1070		return 1;
1071
1072	crtc->h_total_disp = (((mode->hdisplay >> 3) - 1) << 16) |
1073	    ((mode->htotal >> 3) - 1);
1074	crtc->h_sync_strt_wid =
1075	    (((mode->hsync_end - mode->hsync_start) >> 3) << 16) |
1076	    ((mode->hsync_start >> 3) - 1) | ((mode->hsync_start & 7) << 8);
1077
1078	crtc->v_total_disp = ((mode->vdisplay - 1) << 16) |
1079	    (mode->vtotal - 1);
1080	crtc->v_sync_strt_wid =
1081	    ((mode->vsync_end - mode->vsync_start) << 16) |
1082	    (mode->vsync_start - 1);
1083
1084	if (mode->flags & VID_NVSYNC)
1085		crtc->v_sync_strt_wid |= CRTC_VSYNC_NEG;
1086
1087	switch (sc->bits_per_pixel) {
1088	case 8:
1089		crtc->color_depth = CRTC_PIX_WIDTH_8BPP;
1090		break;
1091	case 16:
1092		crtc->color_depth = CRTC_PIX_WIDTH_16BPP;
1093		break;
1094	case 32:
1095		crtc->color_depth = CRTC_PIX_WIDTH_32BPP;
1096		break;
1097	}
1098
1099	crtc->gen_cntl = 0;
1100	if (mode->flags & VID_INTERLACE)
1101		crtc->gen_cntl |= CRTC_INTERLACE_EN;
1102
1103	if (mode->flags & VID_CSYNC)
1104		crtc->gen_cntl |= CRTC_CSYNC_EN;
1105
1106	crtc->dot_clock = mode->dot_clock;
1107
1108	return 0;
1109}
1110
1111static void
1112mach64_set_crtcregs(struct mach64_softc *sc, struct mach64_crtcregs *crtc)
1113{
1114
1115	mach64_set_pll(sc, crtc->dot_clock);
1116
1117	if (sc->has_dsp)
1118		mach64_set_dsp(sc);
1119
1120	DPRINTF("\th total: 0x%08x  h sync: 0x%08x\n",
1121	    crtc->h_total_disp, crtc->h_sync_strt_wid);
1122	DPRINTF("\tv total: 0x%08x  v sync: 0x%08x\n",
1123	    crtc->v_total_disp, crtc->v_sync_strt_wid);
1124	regw(sc, CRTC_H_TOTAL_DISP, crtc->h_total_disp);
1125	regw(sc, CRTC_H_SYNC_STRT_WID, crtc->h_sync_strt_wid);
1126	regw(sc, CRTC_V_TOTAL_DISP, crtc->v_total_disp);
1127	regw(sc, CRTC_V_SYNC_STRT_WID, crtc->v_sync_strt_wid);
1128
1129	regw(sc, CRTC_VLINE_CRNT_VLINE, 0);
1130
1131	regw(sc, CRTC_OFF_PITCH, (sc->stride >> 3) << 22);
1132
1133	regw(sc, CRTC_GEN_CNTL, crtc->gen_cntl | crtc->color_depth |
1134	    sc->sc_gen_cntl | CRTC_EXT_DISP_EN | CRTC_EXT_EN);
1135}
1136
1137static int
1138mach64_modeswitch(struct mach64_softc *sc, struct videomode *mode)
1139{
1140	struct mach64_crtcregs crtc;
1141
1142	memset(&crtc, 0, sizeof crtc);	/* XXX gcc */
1143
1144	if (mach64_calc_crtcregs(sc, &crtc, mode))
1145		return 1;
1146	aprint_debug("crtc dot clock: %d\n", crtc.dot_clock);
1147	if (crtc.dot_clock == 0) {
1148		aprint_error("%s: preposterous dot clock (%d)\n",
1149		    device_xname(sc->sc_dev), crtc.dot_clock);
1150		return 1;
1151	}
1152	mach64_set_crtcregs(sc, &crtc);
1153	return 0;
1154}
1155
1156static void
1157mach64_reset_engine(struct mach64_softc *sc)
1158{
1159
1160	/* Reset engine.*/
1161	regw(sc, GEN_TEST_CNTL, regr(sc, GEN_TEST_CNTL) & ~GUI_ENGINE_ENABLE);
1162
1163	/* Enable engine. */
1164	regw(sc, GEN_TEST_CNTL, regr(sc, GEN_TEST_CNTL) | GUI_ENGINE_ENABLE);
1165
1166	/* Ensure engine is not locked up by clearing any FIFO or
1167	   host errors. */
1168	regw(sc, BUS_CNTL, regr(sc, BUS_CNTL) | BUS_HOST_ERR_ACK |
1169	    BUS_FIFO_ERR_ACK);
1170}
1171
1172static void
1173mach64_init_engine(struct mach64_softc *sc)
1174{
1175	uint32_t pitch_value;
1176
1177	pitch_value = sc->stride;
1178
1179	if (sc->bits_per_pixel == 24)
1180		pitch_value *= 3;
1181
1182	mach64_reset_engine(sc);
1183
1184	wait_for_fifo(sc, 14);
1185
1186	regw(sc, CONTEXT_MASK, 0xffffffff);
1187
1188	regw(sc, DST_OFF_PITCH, (pitch_value >> 3) << 22);
1189
1190	/* make sure the visible area starts where we're going to draw */
1191	regw(sc, CRTC_OFF_PITCH, (sc->stride >> 3) << 22);
1192
1193	regw(sc, DST_Y_X, 0);
1194	regw(sc, DST_HEIGHT, 0);
1195	regw(sc, DST_BRES_ERR, 0);
1196	regw(sc, DST_BRES_INC, 0);
1197	regw(sc, DST_BRES_DEC, 0);
1198
1199	regw(sc, DST_CNTL, DST_LAST_PEL | DST_X_LEFT_TO_RIGHT |
1200	    DST_Y_TOP_TO_BOTTOM);
1201
1202	regw(sc, SRC_OFF_PITCH, (pitch_value >> 3) << 22);
1203
1204	regw(sc, SRC_Y_X, 0);
1205	regw(sc, SRC_HEIGHT1_WIDTH1, 1);
1206	regw(sc, SRC_Y_X_START, 0);
1207	regw(sc, SRC_HEIGHT2_WIDTH2, 1);
1208
1209	regw(sc, SRC_CNTL, SRC_LINE_X_LEFT_TO_RIGHT);
1210
1211	wait_for_fifo(sc, 13);
1212	regw(sc, HOST_CNTL, 0);
1213
1214	regw(sc, PAT_REG0, 0);
1215	regw(sc, PAT_REG1, 0);
1216	regw(sc, PAT_CNTL, 0);
1217
1218	regw(sc, SC_LEFT, 0);
1219	regw(sc, SC_TOP, 0);
1220	regw(sc, SC_BOTTOM, 0x3fff);
1221	regw(sc, SC_RIGHT, pitch_value - 1);
1222
1223	regw(sc, DP_BKGD_CLR, WS_DEFAULT_BG);
1224	regw(sc, DP_FRGD_CLR, WS_DEFAULT_FG);
1225	regw(sc, DP_WRITE_MASK, 0xffffffff);
1226	regw(sc, DP_MIX, (MIX_SRC << 16) | MIX_DST);
1227
1228	regw(sc, DP_SRC, FRGD_SRC_FRGD_CLR);
1229
1230	wait_for_fifo(sc, 3);
1231	regw(sc, CLR_CMP_CLR, 0);
1232	regw(sc, CLR_CMP_MASK, 0xffffffff);
1233	regw(sc, CLR_CMP_CNTL, 0);
1234
1235	wait_for_fifo(sc, 3);
1236	switch (sc->bits_per_pixel) {
1237	case 8:
1238		regw(sc, DP_PIX_WIDTH, HOST_1BPP | SRC_8BPP | DST_8BPP);
1239		regw(sc, DP_CHAIN_MASK, DP_CHAIN_8BPP);
1240		/* We want 8 bit per channel */
1241		regw(sc, DAC_CNTL, regr(sc, DAC_CNTL) | DAC_8BIT_EN);
1242		break;
1243	case 32:
1244		regw(sc, DP_PIX_WIDTH, HOST_1BPP | SRC_32BPP | DST_32BPP);
1245		regw(sc, DP_CHAIN_MASK, DP_CHAIN_32BPP);
1246		regw(sc, DAC_CNTL, regr(sc, DAC_CNTL) | DAC_8BIT_EN);
1247		break;
1248	}
1249	regw(sc, DP_WRITE_MASK, 0xff);
1250
1251	wait_for_fifo(sc, 5);
1252	regw(sc, CRTC_INT_CNTL, regr(sc, CRTC_INT_CNTL) & ~0x20);
1253	regw(sc, GUI_TRAJ_CNTL, DST_X_LEFT_TO_RIGHT | DST_Y_TOP_TO_BOTTOM);
1254
1255	wait_for_idle(sc);
1256}
1257
1258#if 0
1259static void
1260mach64_adjust_frame(struct mach64_softc *sc, int x, int y)
1261{
1262	int offset;
1263
1264	offset = ((x + y * sc->stride) * (sc->bits_per_pixel >> 3)) >> 3;
1265
1266	regw(sc, CRTC_OFF_PITCH, (regr(sc, CRTC_OFF_PITCH) & 0xfff00000) |
1267	     offset);
1268}
1269#endif
1270
1271static void
1272mach64_set_dsp(struct mach64_softc *sc)
1273{
1274	uint32_t fifo_depth, page_size, dsp_precision, dsp_loop_latency;
1275	uint32_t dsp_off, dsp_on, dsp_xclks_per_qw;
1276	uint32_t xclks_per_qw, xclks_per_qw_m, y;
1277	uint32_t fifo_off, fifo_on;
1278
1279	aprint_normal_dev(sc->sc_dev, "initializing the DSP\n");
1280
1281	if (mach64_chip_id == PCI_PRODUCT_ATI_MACH64_VT ||
1282	    mach64_chip_id == PCI_PRODUCT_ATI_RAGE_II ||
1283	    mach64_chip_id == PCI_PRODUCT_ATI_RAGE_IIP ||
1284	    mach64_chip_id == PCI_PRODUCT_ATI_RAGE_IIC_PCI ||
1285	    mach64_chip_id == PCI_PRODUCT_ATI_RAGE_IIC_AGP_B ||
1286	    mach64_chip_id == PCI_PRODUCT_ATI_RAGE_IIC_AGP_P) {
1287		dsp_loop_latency = 0;
1288		fifo_depth = 24;
1289	} else {
1290		dsp_loop_latency = 2;
1291		fifo_depth = 32;
1292	}
1293
1294	dsp_precision = 0;
1295
1296	xclks_per_qw = (sc->mclk_fb_div * sc->vclk_post_div * 64 << 11) /
1297	    (sc->vclk_fb_div * sc->mclk_post_div * sc->bits_per_pixel);
1298
1299	xclks_per_qw_m = (sc->mem_freq * 64 << 4) /
1300		       (sc->vclk_freq * sc->bits_per_pixel);
1301
1302	DPRINTF("xclks_per_qw %d %d\n", xclks_per_qw >> 7, xclks_per_qw_m);
1303	DPRINTF("mem %dkHz v %dkHz\n", sc->mem_freq, sc->vclk_freq);
1304
1305	y = (xclks_per_qw * fifo_depth) >> 11;
1306
1307	while (y) {
1308		y >>= 1;
1309		dsp_precision++;
1310	}
1311	dsp_precision -= 5;
1312	fifo_off = ((xclks_per_qw * (fifo_depth - 1)) >> 5) + (3 << 6);
1313
1314	switch (sc->memtype) {
1315	case DRAM:
1316	case EDO_DRAM:
1317	case PSEUDO_EDO:
1318		if (sc->memsize > 1024) {
1319			page_size = 9;
1320			dsp_loop_latency += 6;
1321		} else {
1322			page_size = 10;
1323			if (sc->memtype == DRAM)
1324				dsp_loop_latency += 8;
1325			else
1326				dsp_loop_latency += 7;
1327		}
1328		break;
1329	case SDRAM:
1330		if (sc->memsize > 1024) {
1331			page_size = 8;
1332			dsp_loop_latency += 8;
1333		} else {
1334			page_size = 10;
1335			dsp_loop_latency += 9;
1336		}
1337		break;
1338	case SGRAM:
1339		page_size = 8;
1340		dsp_loop_latency = 8;
1341		break;
1342	default:
1343		page_size = 10;
1344		dsp_loop_latency += 9;
1345		break;
1346	}
1347
1348	if (xclks_per_qw >= (page_size << 11))
1349		fifo_on = ((2 * page_size + 1) << 6) + (xclks_per_qw >> 5);
1350	else
1351		fifo_on = (3 * page_size + 2) << 6;
1352
1353	dsp_xclks_per_qw = xclks_per_qw >> dsp_precision;
1354	dsp_on = fifo_on >> dsp_precision;
1355	dsp_off = fifo_off >> dsp_precision;
1356
1357	DPRINTF("dsp_xclks_per_qw = %d, dsp_on = %d, dsp_off = %d,\n"
1358	    "dsp_precision = %d, dsp_loop_latency = %d,\n"
1359	    "mclk_fb_div = %d, vclk_fb_div = %d,\n"
1360	    "mclk_post_div = %d, vclk_post_div = %d\n",
1361	    dsp_xclks_per_qw, dsp_on, dsp_off, dsp_precision, dsp_loop_latency,
1362	    sc->mclk_fb_div, sc->vclk_fb_div,
1363	    sc->mclk_post_div, sc->vclk_post_div);
1364	DPRINTF("DSP_ON_OFF %08x\n", regr(sc, DSP_ON_OFF));
1365	DPRINTF("DSP_CONFIG %08x\n", regr(sc, DSP_CONFIG));
1366	regw(sc, DSP_ON_OFF, ((dsp_on << 16) & DSP_ON) | (dsp_off & DSP_OFF));
1367	regw(sc, DSP_CONFIG, ((dsp_precision << 20) & DSP_PRECISION) |
1368	    ((dsp_loop_latency << 16) & DSP_LOOP_LATENCY) |
1369	    (dsp_xclks_per_qw & DSP_XCLKS_PER_QW));
1370	DPRINTF("DSP_ON_OFF %08x\n", regr(sc, DSP_ON_OFF));
1371	DPRINTF("DSP_CONFIG %08x\n", regr(sc, DSP_CONFIG));
1372}
1373
1374static void
1375mach64_set_pll(struct mach64_softc *sc, int clock)
1376{
1377	uint32_t q, clockreg;
1378	int clockshift = sc->sc_clock << 1;
1379	uint8_t reg, vclk_ctl;
1380
1381	q = (clock * sc->ref_div * 100) / (2 * sc->ref_freq);
1382#ifdef MACHFB_DEBUG
1383	printf("q = %d\n", q);
1384#endif
1385	if (q > 25500) {
1386		aprint_error_dev(sc->sc_dev, "Warning: q > 25500\n");
1387		q = 25500;
1388		sc->vclk_post_div = 1;
1389		sc->log2_vclk_post_div = 0;
1390	} else if (q > 12750) {
1391		sc->vclk_post_div = 1;
1392		sc->log2_vclk_post_div = 0;
1393	} else if (q > 6350) {
1394		sc->vclk_post_div = 2;
1395		sc->log2_vclk_post_div = 1;
1396	} else if (q > 3150) {
1397		sc->vclk_post_div = 4;
1398		sc->log2_vclk_post_div = 2;
1399	} else if (q >= 1600) {
1400		sc->vclk_post_div = 8;
1401		sc->log2_vclk_post_div = 3;
1402	} else {
1403		aprint_error_dev(sc->sc_dev, "Warning: q < 1600\n");
1404		sc->vclk_post_div = 8;
1405		sc->log2_vclk_post_div = 3;
1406	}
1407	sc->vclk_fb_div = q * sc->vclk_post_div / 100;
1408	DPRINTF("post_div: %d log2_post_div: %d mclk_div: %d\n",
1409	    sc->vclk_post_div, sc->log2_vclk_post_div, sc->mclk_fb_div);
1410
1411	vclk_ctl = regrb_pll(sc, PLL_VCLK_CNTL);
1412	aprint_debug("vclk_ctl: %02x\n", vclk_ctl);
1413	vclk_ctl |= PLL_VCLK_RESET;
1414	regwb_pll(sc, PLL_VCLK_CNTL, vclk_ctl);
1415
1416	DPRINTF("target: %d output: %d\n", clock,
1417	    (2 * sc->ref_freq * sc->vclk_fb_div) /
1418	    (sc->ref_div * sc->vclk_post_div));
1419
1420	regwb_pll(sc, MCLK_FB_DIV, sc->mclk_fb_div);
1421	reg = regrb_pll(sc, VCLK_POST_DIV);
1422	reg &= ~(3 << clockshift);
1423	reg |= (sc->log2_vclk_post_div << clockshift);
1424	regwb_pll(sc, VCLK_POST_DIV, reg);
1425	regwb_pll(sc, VCLK0_FB_DIV + sc->sc_clock, sc->vclk_fb_div);
1426
1427	vclk_ctl &= ~PLL_VCLK_RESET;
1428	regwb_pll(sc, PLL_VCLK_CNTL, vclk_ctl);
1429
1430	clockreg = regr(sc, CLOCK_CNTL);
1431	clockreg &= ~CLOCK_SEL;
1432	clockreg |= sc->sc_clock | CLOCK_STROBE;
1433	regw(sc, CLOCK_CNTL, clockreg);
1434	sc->vclk_freq = clock;
1435}
1436
1437static void
1438mach64_init_lut(struct mach64_softc *sc)
1439{
1440	uint8_t cmap[768];
1441	int i, idx;
1442
1443	rasops_get_cmap(&mach64_console_screen.scr_ri, cmap, sizeof(cmap));
1444	idx = 0;
1445	for (i = 0; i < 256; i++) {
1446		mach64_putpalreg(sc, i, cmap[idx], cmap[idx + 1],
1447		    cmap[idx + 2]);
1448		idx += 3;
1449	}
1450}
1451
1452static int
1453mach64_putpalreg(struct mach64_softc *sc, uint8_t index, uint8_t r, uint8_t g,
1454    uint8_t b)
1455{
1456	sc->sc_cmap_red[index] = r;
1457	sc->sc_cmap_green[index] = g;
1458	sc->sc_cmap_blue[index] = b;
1459	/*
1460	 * writing the dac index takes a while, in theory we can poll some
1461	 * register to see when it's ready - but we better avoid writing it
1462	 * unnecessarily
1463	 */
1464	if (index != sc->sc_dacw) {
1465		regwb(sc, DAC_MASK, 0xff);
1466		regwb(sc, DAC_WINDEX, index);
1467	}
1468	sc->sc_dacw = index + 1;
1469	regwb(sc, DAC_DATA, r);
1470	regwb(sc, DAC_DATA, g);
1471	regwb(sc, DAC_DATA, b);
1472	return 0;
1473}
1474
1475static int
1476mach64_putcmap(struct mach64_softc *sc, struct wsdisplay_cmap *cm)
1477{
1478	uint index = cm->index;
1479	uint count = cm->count;
1480	int i, error;
1481	uint8_t rbuf[256], gbuf[256], bbuf[256];
1482	uint8_t *r, *g, *b;
1483
1484	if (cm->index >= 256 || cm->count > 256 ||
1485	    (cm->index + cm->count) > 256)
1486		return EINVAL;
1487	error = copyin(cm->red, &rbuf[index], count);
1488	if (error)
1489		return error;
1490	error = copyin(cm->green, &gbuf[index], count);
1491	if (error)
1492		return error;
1493	error = copyin(cm->blue, &bbuf[index], count);
1494	if (error)
1495		return error;
1496
1497	memcpy(&sc->sc_cmap_red[index], &rbuf[index], count);
1498	memcpy(&sc->sc_cmap_green[index], &gbuf[index], count);
1499	memcpy(&sc->sc_cmap_blue[index], &bbuf[index], count);
1500
1501	r = &sc->sc_cmap_red[index];
1502	g = &sc->sc_cmap_green[index];
1503	b = &sc->sc_cmap_blue[index];
1504
1505	for (i = 0; i < count; i++) {
1506		mach64_putpalreg(sc, index, *r, *g, *b);
1507		index++;
1508		r++, g++, b++;
1509	}
1510	return 0;
1511}
1512
1513static int
1514mach64_getcmap(struct mach64_softc *sc, struct wsdisplay_cmap *cm)
1515{
1516	u_int index = cm->index;
1517	u_int count = cm->count;
1518	int error;
1519
1520	if (index >= 255 || count > 256 || index + count > 256)
1521		return EINVAL;
1522
1523	error = copyout(&sc->sc_cmap_red[index],   cm->red,   count);
1524	if (error)
1525		return error;
1526	error = copyout(&sc->sc_cmap_green[index], cm->green, count);
1527	if (error)
1528		return error;
1529	error = copyout(&sc->sc_cmap_blue[index],  cm->blue,  count);
1530	if (error)
1531		return error;
1532
1533	return 0;
1534}
1535
1536static int
1537mach64_is_console(struct mach64_softc *sc)
1538{
1539	bool console = 0;
1540
1541	prop_dictionary_get_bool(device_properties(sc->sc_dev),
1542	    "is_console", &console);
1543	return console;
1544}
1545
1546/*
1547 * wsdisplay_emulops
1548 */
1549
1550static void
1551mach64_cursor(void *cookie, int on, int row, int col)
1552{
1553	struct rasops_info *ri = cookie;
1554	struct vcons_screen *scr = ri->ri_hw;
1555	struct mach64_softc *sc = scr->scr_cookie;
1556	int x, y, wi, he;
1557
1558	wi = ri->ri_font->fontwidth;
1559	he = ri->ri_font->fontheight;
1560
1561	if ((!sc->sc_locked) && (sc->sc_mode == WSDISPLAYIO_MODE_EMUL)) {
1562		x = ri->ri_ccol * wi + ri->ri_xorigin;
1563		y = ri->ri_crow * he + ri->ri_yorigin;
1564		if (ri->ri_flg & RI_CURSOR) {
1565			mach64_bitblt(sc, x, y, x, y, wi, he, MIX_NOT_SRC);
1566			ri->ri_flg &= ~RI_CURSOR;
1567		}
1568		ri->ri_crow = row;
1569		ri->ri_ccol = col;
1570		if (on) {
1571			x = ri->ri_ccol * wi + ri->ri_xorigin;
1572			y = ri->ri_crow * he + ri->ri_yorigin;
1573			mach64_bitblt(sc, x, y, x, y, wi, he, MIX_NOT_SRC);
1574			ri->ri_flg |= RI_CURSOR;
1575		}
1576	} else {
1577		scr->scr_ri.ri_crow = row;
1578		scr->scr_ri.ri_ccol = col;
1579		scr->scr_ri.ri_flg &= ~RI_CURSOR;
1580	}
1581}
1582
1583#if 0
1584static int
1585mach64_mapchar(void *cookie, int uni, u_int *index)
1586{
1587	return 0;
1588}
1589#endif
1590
1591static void
1592mach64_putchar_mono(void *cookie, int row, int col, u_int c, long attr)
1593{
1594	struct rasops_info *ri = cookie;
1595	struct wsdisplay_font *font = PICK_FONT(ri, c);
1596	struct vcons_screen *scr = ri->ri_hw;
1597	struct mach64_softc *sc = scr->scr_cookie;
1598
1599	if (sc->sc_mode == WSDISPLAYIO_MODE_EMUL) {
1600		int fg, bg, uc;
1601		uint8_t *data;
1602		int x, y, wi, he;
1603		wi = font->fontwidth;
1604		he = font->fontheight;
1605
1606		if (!CHAR_IN_FONT(c, font))
1607			return;
1608		bg = ri->ri_devcmap[(attr >> 16) & 0x0f];
1609		fg = ri->ri_devcmap[(attr >> 24) & 0x0f];
1610		x = ri->ri_xorigin + col * wi;
1611		y = ri->ri_yorigin + row * he;
1612		if (c == 0x20) {
1613			mach64_rectfill(sc, x, y, wi, he, bg);
1614		} else {
1615			uc = c - font->firstchar;
1616			data = (uint8_t *)font->data + uc *
1617			    ri->ri_fontscale;
1618
1619			mach64_setup_mono(sc, x, y, wi, he, fg, bg);
1620			mach64_feed_bytes(sc, ri->ri_fontscale, data);
1621		}
1622		if (attr & 1)
1623			mach64_rectfill(sc, x, y + he - 2, wi, 1, fg);
1624	}
1625}
1626
1627static void
1628mach64_putchar_aa8(void *cookie, int row, int col, u_int c, long attr)
1629{
1630	struct rasops_info *ri = cookie;
1631	struct wsdisplay_font *font = PICK_FONT(ri, c);
1632	struct vcons_screen *scr = ri->ri_hw;
1633	struct mach64_softc *sc = scr->scr_cookie;
1634	uint32_t bg, fg, latch = 0, bg8, fg8, pixel;
1635	int i, x, y, wi, he, r, g, b, aval;
1636	int r1, g1, b1, r0, g0, b0, fgo, bgo;
1637	uint8_t *data8;
1638	int rv = 0, cnt = 0;
1639
1640	if (sc->sc_mode != WSDISPLAYIO_MODE_EMUL)
1641		return;
1642
1643	if (!CHAR_IN_FONT(c, font))
1644		return;
1645
1646	wi = font->fontwidth;
1647	he = font->fontheight;
1648	bg = (u_char)ri->ri_devcmap[(attr >> 16) & 0x0f];
1649	fg = (u_char)ri->ri_devcmap[(attr >> 24) & 0x0f];
1650	x = ri->ri_xorigin + col * wi;
1651	y = ri->ri_yorigin + row * he;
1652
1653	if (c == 0x20) {
1654		mach64_rectfill(sc, x, y, wi, he, bg);
1655		if (attr & 1)
1656			mach64_rectfill(sc, x, y + he - 2, wi, 1, fg);
1657		return;
1658	}
1659
1660	rv = glyphcache_try(&sc->sc_gc, c, x, y, attr);
1661	if (rv == GC_OK)
1662		return;
1663
1664	data8 = WSFONT_GLYPH(c, font);
1665
1666	wait_for_fifo(sc, 11);
1667	regw(sc, DP_PIX_WIDTH, DST_8BPP | SRC_8BPP | HOST_8BPP);
1668	regw(sc, DP_SRC, MONO_SRC_ONE | BKGD_SRC_HOST | FRGD_SRC_HOST);
1669	regw(sc, DP_MIX, ((MIX_SRC & 0xffff) << 16) | MIX_SRC);
1670	regw(sc, CLR_CMP_CNTL ,0);	/* no transparency */
1671	regw(sc, SRC_CNTL, SRC_LINE_X_LEFT_TO_RIGHT);
1672	regw(sc, DST_CNTL, DST_Y_TOP_TO_BOTTOM | DST_X_LEFT_TO_RIGHT);
1673	regw(sc, HOST_CNTL, HOST_BYTE_ALIGN);
1674	regw(sc, SRC_Y_X, 0);
1675	regw(sc, SRC_WIDTH1, wi);
1676	regw(sc, DST_Y_X, (x << 16) | y);
1677	regw(sc, DST_HEIGHT_WIDTH, (wi << 16) | he);
1678
1679	/*
1680	 * we need the RGB colours here, so get offsets into rasops_cmap
1681	 */
1682	fgo = ((attr >> 24) & 0xf) * 3;
1683	bgo = ((attr >> 16) & 0xf) * 3;
1684
1685	r0 = rasops_cmap[bgo];
1686	r1 = rasops_cmap[fgo];
1687	g0 = rasops_cmap[bgo + 1];
1688	g1 = rasops_cmap[fgo + 1];
1689	b0 = rasops_cmap[bgo + 2];
1690	b1 = rasops_cmap[fgo + 2];
1691#define R3G3B2(r, g, b) ((r & 0xe0) | ((g >> 3) & 0x1c) | (b >> 6))
1692	bg8 = R3G3B2(r0, g0, b0);
1693	fg8 = R3G3B2(r1, g1, b1);
1694
1695	wait_for_fifo(sc, 10);
1696
1697	for (i = 0; i < ri->ri_fontscale; i++) {
1698		aval = *data8;
1699		if (aval == 0) {
1700			pixel = bg8;
1701		} else if (aval == 255) {
1702			pixel = fg8;
1703		} else {
1704			r = aval * r1 + (255 - aval) * r0;
1705			g = aval * g1 + (255 - aval) * g0;
1706			b = aval * b1 + (255 - aval) * b0;
1707			pixel = ((r & 0xe000) >> 8) |
1708				((g & 0xe000) >> 11) |
1709				((b & 0xc000) >> 14);
1710		}
1711		latch = (latch << 8) | pixel;
1712		/* write in 32bit chunks */
1713		if ((i & 3) == 3) {
1714			regws(sc, HOST_DATA0, latch);
1715			/*
1716			 * not strictly necessary, old data should be shifted
1717			 * out
1718			 */
1719			latch = 0;
1720			cnt++;
1721			if (cnt > 8) {
1722				wait_for_fifo(sc, 10);
1723				cnt = 0;
1724			}
1725		}
1726		data8++;
1727	}
1728	/* if we have pixels left in latch write them out */
1729	if ((i & 3) != 0) {
1730		latch = latch << ((4 - (i & 3)) << 3);
1731		regws(sc, HOST_DATA0, latch);
1732	}
1733
1734	if (rv == GC_ADD) {
1735		glyphcache_add(&sc->sc_gc, c, x, y);
1736	} else 	if (attr & 1) {
1737		mach64_rectfill(sc, x, y + he - 2, wi, 1, fg);
1738	}
1739
1740}
1741
1742static void
1743mach64_copycols(void *cookie, int row, int srccol, int dstcol, int ncols)
1744{
1745	struct rasops_info *ri = cookie;
1746	struct vcons_screen *scr = ri->ri_hw;
1747	struct mach64_softc *sc = scr->scr_cookie;
1748	int32_t xs, xd, y, width, height;
1749
1750	if ((sc->sc_locked == 0) && (sc->sc_mode == WSDISPLAYIO_MODE_EMUL)) {
1751		xs = ri->ri_xorigin + ri->ri_font->fontwidth * srccol;
1752		xd = ri->ri_xorigin + ri->ri_font->fontwidth * dstcol;
1753		y = ri->ri_yorigin + ri->ri_font->fontheight * row;
1754		width = ri->ri_font->fontwidth * ncols;
1755		height = ri->ri_font->fontheight;
1756		mach64_bitblt(sc, xs, y, xd, y, width, height, MIX_SRC);
1757	}
1758}
1759
1760static void
1761mach64_erasecols(void *cookie, int row, int startcol, int ncols, long fillattr)
1762{
1763	struct rasops_info *ri = cookie;
1764	struct vcons_screen *scr = ri->ri_hw;
1765	struct mach64_softc *sc = scr->scr_cookie;
1766	int32_t x, y, width, height, fg, bg, ul;
1767
1768	if ((sc->sc_locked == 0) && (sc->sc_mode == WSDISPLAYIO_MODE_EMUL)) {
1769		x = ri->ri_xorigin + ri->ri_font->fontwidth * startcol;
1770		y = ri->ri_yorigin + ri->ri_font->fontheight * row;
1771		width = ri->ri_font->fontwidth * ncols;
1772		height = ri->ri_font->fontheight;
1773		rasops_unpack_attr(fillattr, &fg, &bg, &ul);
1774
1775		mach64_rectfill(sc, x, y, width, height, ri->ri_devcmap[bg]);
1776	}
1777}
1778
1779static void
1780mach64_copyrows(void *cookie, int srcrow, int dstrow, int nrows)
1781{
1782	struct rasops_info *ri = cookie;
1783	struct vcons_screen *scr = ri->ri_hw;
1784	struct mach64_softc *sc = scr->scr_cookie;
1785	int32_t x, ys, yd, width, height;
1786
1787	if ((sc->sc_locked == 0) && (sc->sc_mode == WSDISPLAYIO_MODE_EMUL)) {
1788		x = ri->ri_xorigin;
1789		ys = ri->ri_yorigin + ri->ri_font->fontheight * srcrow;
1790		yd = ri->ri_yorigin + ri->ri_font->fontheight * dstrow;
1791		width = ri->ri_emuwidth;
1792		height = ri->ri_font->fontheight*nrows;
1793		mach64_bitblt(sc, x, ys, x, yd, width, height, MIX_SRC);
1794	}
1795}
1796
1797static void
1798mach64_eraserows(void *cookie, int row, int nrows, long fillattr)
1799{
1800	struct rasops_info *ri = cookie;
1801	struct vcons_screen *scr = ri->ri_hw;
1802	struct mach64_softc *sc = scr->scr_cookie;
1803	int32_t x, y, width, height, fg, bg, ul;
1804
1805	if ((sc->sc_locked == 0) && (sc->sc_mode == WSDISPLAYIO_MODE_EMUL)) {
1806		if ((row == 0) && (nrows == ri->ri_rows)) {
1807			/* clear full screen */
1808			x = 0;
1809			y = 0;
1810			width = sc->virt_x;
1811			height = sc->virt_y;
1812		} else {
1813			x = ri->ri_xorigin;
1814			y = ri->ri_yorigin + ri->ri_font->fontheight * row;
1815			width = ri->ri_emuwidth;
1816			height = ri->ri_font->fontheight * nrows;
1817		}
1818		rasops_unpack_attr(fillattr, &fg, &bg, &ul);
1819
1820		mach64_rectfill(sc, x, y, width, height, ri->ri_devcmap[bg]);
1821	}
1822}
1823
1824static void
1825mach64_bitblt(void *cookie, int xs, int ys, int xd, int yd, int width,
1826    int height, int rop)
1827{
1828	struct mach64_softc *sc = cookie;
1829	uint32_t dest_ctl = 0;
1830
1831#if 0
1832	wait_for_idle(sc);
1833#else
1834	wait_for_fifo(sc, 10);
1835#endif
1836
1837	regw(sc, DP_PIX_WIDTH, DST_8BPP | SRC_8BPP | HOST_8BPP);
1838	regw(sc, DP_SRC, FRGD_SRC_BLIT);
1839	regw(sc, DP_MIX, (rop & 0xffff) << 16);
1840	regw(sc, CLR_CMP_CNTL, 0);	/* no transparency */
1841	if (yd < ys) {
1842		dest_ctl = DST_Y_TOP_TO_BOTTOM;
1843	} else {
1844		ys += height - 1;
1845		yd += height - 1;
1846		dest_ctl = DST_Y_BOTTOM_TO_TOP;
1847	}
1848	if (xd < xs) {
1849		dest_ctl |= DST_X_LEFT_TO_RIGHT;
1850		regw(sc, SRC_CNTL, SRC_LINE_X_LEFT_TO_RIGHT);
1851	} else {
1852		dest_ctl |= DST_X_RIGHT_TO_LEFT;
1853		xs += width - 1;
1854		xd += width - 1;
1855		regw(sc, SRC_CNTL, SRC_LINE_X_RIGHT_TO_LEFT);
1856	}
1857	regw(sc, DST_CNTL, dest_ctl);
1858
1859	regw(sc, SRC_Y_X, (xs << 16) | ys);
1860	regw(sc, SRC_WIDTH1, width);
1861	regw(sc, DST_Y_X, (xd << 16) | yd);
1862	regw(sc, DST_HEIGHT_WIDTH, (width << 16) | height);
1863}
1864
1865static void
1866mach64_setup_mono(struct mach64_softc *sc, int xd, int yd, int width,
1867     int height, uint32_t fg, uint32_t bg)
1868{
1869	wait_for_idle(sc);
1870	regw(sc, DP_WRITE_MASK, 0xff);	/* XXX only good for 8 bit */
1871	regw(sc, DP_PIX_WIDTH, DST_8BPP | SRC_1BPP | HOST_1BPP);
1872	regw(sc, DP_SRC, MONO_SRC_HOST | BKGD_SRC_BKGD_CLR | FRGD_SRC_FRGD_CLR);
1873	regw(sc, DP_MIX, ((MIX_SRC & 0xffff) << 16) | MIX_SRC);
1874	regw(sc, CLR_CMP_CNTL ,0);	/* no transparency */
1875	regw(sc, SRC_CNTL, SRC_LINE_X_LEFT_TO_RIGHT);
1876	regw(sc, DST_CNTL, DST_Y_TOP_TO_BOTTOM | DST_X_LEFT_TO_RIGHT);
1877	regw(sc, HOST_CNTL, HOST_BYTE_ALIGN);
1878	regw(sc, DP_BKGD_CLR, bg);
1879	regw(sc, DP_FRGD_CLR, fg);
1880	regw(sc, SRC_Y_X, 0);
1881	regw(sc, SRC_WIDTH1, width);
1882	regw(sc, DST_Y_X, (xd << 16) | yd);
1883	regw(sc, DST_HEIGHT_WIDTH, (width << 16) | height);
1884	/* now feed the data into the chip */
1885}
1886
1887static void
1888mach64_feed_bytes(struct mach64_softc *sc, int count, uint8_t *data)
1889{
1890	int i;
1891	uint32_t latch = 0, bork;
1892	int shift = 0;
1893	int reg = 0;
1894
1895	for (i = 0; i < count; i++) {
1896		bork = data[i];
1897		latch |= (bork << shift);
1898		if (shift == 24) {
1899			regw(sc, HOST_DATA0 + reg, latch);
1900			latch = 0;
1901			shift = 0;
1902			reg = (reg + 4) & 0x3c;
1903		} else
1904			shift += 8;
1905	}
1906	if (shift != 0)	/* 24 */
1907		regw(sc, HOST_DATA0 + reg, latch);
1908}
1909
1910
1911static void
1912mach64_rectfill(struct mach64_softc *sc, int x, int y, int width, int height,
1913    int colour)
1914{
1915	wait_for_fifo(sc, 11);
1916	regw(sc, DP_FRGD_CLR, colour);
1917	regw(sc, DP_PIX_WIDTH, DST_8BPP | SRC_8BPP | HOST_8BPP);
1918	regw(sc, DP_SRC, FRGD_SRC_FRGD_CLR);
1919	regw(sc, DP_MIX, MIX_SRC << 16);
1920	regw(sc, CLR_CMP_CNTL, 0);	/* no transparency */
1921	regw(sc, SRC_CNTL, SRC_LINE_X_LEFT_TO_RIGHT);
1922	regw(sc, DST_CNTL, DST_X_LEFT_TO_RIGHT | DST_Y_TOP_TO_BOTTOM);
1923
1924	regw(sc, SRC_Y_X, (x << 16) | y);
1925	regw(sc, SRC_WIDTH1, width);
1926	regw(sc, DST_Y_X, (x << 16) | y);
1927	regw(sc, DST_HEIGHT_WIDTH, (width << 16) | height);
1928}
1929
1930static void
1931mach64_clearscreen(struct mach64_softc *sc)
1932{
1933	mach64_rectfill(sc, 0, 0, sc->virt_x, sc->virt_y, sc->sc_bg);
1934}
1935
1936
1937#if 0
1938static void
1939mach64_showpal(struct mach64_softc *sc)
1940{
1941	int i, x = 0;
1942
1943	for (i = 0; i < 16; i++) {
1944		mach64_rectfill(sc, x, 0, 64, 64, i);
1945		x += 64;
1946	}
1947}
1948#endif
1949
1950/*
1951 * wsdisplay_accessops
1952 */
1953
1954static int
1955mach64_ioctl(void *v, void *vs, u_long cmd, void *data, int flag,
1956	struct lwp *l)
1957{
1958	struct vcons_data *vd = v;
1959	struct mach64_softc *sc = vd->cookie;
1960	struct wsdisplay_fbinfo *wdf;
1961	struct vcons_screen *ms = vd->active;
1962
1963	switch (cmd) {
1964	case WSDISPLAYIO_GTYPE:
1965		*(u_int *)data = WSDISPLAY_TYPE_PCIMISC;
1966		return 0;
1967
1968	case WSDISPLAYIO_LINEBYTES:
1969		*(u_int *)data = sc->stride * sc->bits_per_pixel / 8;
1970		return 0;
1971
1972	case WSDISPLAYIO_GINFO:
1973		wdf = (void *)data;
1974		wdf->height = sc->virt_y;
1975		wdf->width = sc->virt_x;
1976		wdf->depth = sc->bits_per_pixel;
1977		wdf->cmsize = 256;
1978		return 0;
1979
1980	case WSDISPLAYIO_GETCMAP:
1981		return mach64_getcmap(sc,
1982		    (struct wsdisplay_cmap *)data);
1983
1984	case WSDISPLAYIO_PUTCMAP:
1985		return mach64_putcmap(sc,
1986		    (struct wsdisplay_cmap *)data);
1987
1988	/* PCI config read/write passthrough. */
1989	case PCI_IOC_CFGREAD:
1990	case PCI_IOC_CFGWRITE:
1991		return pci_devioctl(sc->sc_pc, sc->sc_pcitag,
1992		    cmd, data, flag, l);
1993
1994	case WSDISPLAYIO_GET_BUSID:
1995		return wsdisplayio_busid_pci(sc->sc_dev, sc->sc_pc,
1996		    sc->sc_pcitag, data);
1997
1998	case WSDISPLAYIO_SMODE: {
1999		int new_mode = *(int*)data;
2000		if (new_mode != sc->sc_mode) {
2001			sc->sc_mode = new_mode;
2002			if ((new_mode == WSDISPLAYIO_MODE_EMUL)
2003			    && (ms != NULL))
2004			{
2005				/* restore initial video mode */
2006				mach64_init(sc);
2007				mach64_init_engine(sc);
2008				mach64_init_lut(sc);
2009				if (sc->sc_setmode)
2010					mach64_modeswitch(sc, sc->sc_my_mode);
2011				mach64_clearscreen(sc);
2012				glyphcache_wipe(&sc->sc_gc);
2013				vcons_redraw_screen(ms);
2014			}
2015		}
2016		}
2017		return 0;
2018	case WSDISPLAYIO_GET_EDID: {
2019		struct wsdisplayio_edid_info *d = data;
2020		return wsdisplayio_get_edid(sc->sc_dev, d);
2021	}
2022
2023	case WSDISPLAYIO_GET_FBINFO: {
2024		struct wsdisplayio_fbinfo *fbi = data;
2025		return wsdisplayio_get_fbinfo(&ms->scr_ri, fbi);
2026	}
2027	}
2028	return EPASSTHROUGH;
2029}
2030
2031static paddr_t
2032mach64_mmap(void *v, void *vs, off_t offset, int prot)
2033{
2034	struct vcons_data *vd = v;
2035	struct mach64_softc *sc = vd->cookie;
2036	paddr_t pa;
2037
2038	if (sc->sc_mode == WSDISPLAYIO_MODE_DUMBFB) {
2039		/*
2040		 *'regular' framebuffer mmap()ing
2041		 */
2042		if (offset < (sc->memsize * 1024)) {
2043			pa = bus_space_mmap(sc->sc_memt, sc->sc_aperbase,
2044			    offset, prot, BUS_SPACE_MAP_LINEAR);
2045			return pa;
2046		}
2047	} else if (sc->sc_mode == WSDISPLAYIO_MODE_MAPPED) {
2048		/*
2049		 * restrict all other mappings to processes with superuser
2050		 * privileges
2051		 */
2052		if (kauth_authorize_machdep(kauth_cred_get(),
2053		    KAUTH_MACHDEP_UNMANAGEDMEM,
2054		    NULL, NULL, NULL, NULL) != 0) {
2055			return -1;
2056		}
2057		if ((offset >= sc->sc_aperbase) &&
2058		    (offset < (sc->sc_aperbase + sc->sc_apersize))) {
2059			pa = bus_space_mmap(sc->sc_memt, offset, 0, prot,
2060			    BUS_SPACE_MAP_LINEAR);
2061			return pa;
2062		}
2063
2064		if ((offset >= sc->sc_regbase) &&
2065		    (offset < (sc->sc_regbase + sc->sc_regsize))) {
2066			pa = bus_space_mmap(sc->sc_regt, offset, 0, prot,
2067			    BUS_SPACE_MAP_LINEAR);
2068			return pa;
2069		}
2070
2071		if ((offset >= sc->sc_rom.vb_base) &&
2072		    (offset < (sc->sc_rom.vb_base + sc->sc_rom.vb_size))) {
2073			pa = bus_space_mmap(sc->sc_memt, offset, 0, prot,
2074			    BUS_SPACE_MAP_LINEAR);
2075			return pa;
2076		}
2077
2078#ifdef PCI_MAGIC_IO_RANGE
2079		if ((offset >= PCI_MAGIC_IO_RANGE) &&
2080		    (offset <= PCI_MAGIC_IO_RANGE + 0x10000)) {
2081		    	return bus_space_mmap(sc->sc_iot,
2082		    	   offset - PCI_MAGIC_IO_RANGE, 0, prot, 0);
2083		}
2084#endif
2085	}
2086	return -1;
2087}
2088
2089#if 0
2090static int
2091mach64_load_font(void *v, void *cookie, struct wsdisplay_font *data)
2092{
2093
2094	return 0;
2095}
2096#endif
2097
2098void
2099machfb_blank(struct mach64_softc *sc, int blank)
2100{
2101	uint32_t reg;
2102
2103#define MACH64_BLANK (CRTC_DISPLAY_DIS | CRTC_HSYNC_DIS | CRTC_VSYNC_DIS)
2104
2105	switch (blank)
2106	{
2107    		case 0:
2108			reg = regr(sc, CRTC_GEN_CNTL);
2109			regw(sc, CRTC_GEN_CNTL, reg & ~(MACH64_BLANK));
2110			sc->sc_blanked = 0;
2111			break;
2112		case 1:
2113			reg = regr(sc, CRTC_GEN_CNTL);
2114			regw(sc, CRTC_GEN_CNTL, reg | (MACH64_BLANK));
2115			sc->sc_blanked = 1;
2116			break;
2117		default:
2118        		break;
2119	}
2120}
2121