xref: /openbmc/linux/drivers/video/fbdev/nvidia/nv_hw.c (revision c796f021)
1  /***************************************************************************\
2 |*                                                                           *|
3 |*       Copyright 1993-2003 NVIDIA, Corporation.  All rights reserved.      *|
4 |*                                                                           *|
5 |*     NOTICE TO USER:   The source code  is copyrighted under  U.S. and     *|
6 |*     international laws.  Users and possessors of this source code are     *|
7 |*     hereby granted a nonexclusive,  royalty-free copyright license to     *|
8 |*     use this code in individual and commercial software.                  *|
9 |*                                                                           *|
10 |*     Any use of this source code must include,  in the user documenta-     *|
11 |*     tion and  internal comments to the code,  notices to the end user     *|
12 |*     as follows:                                                           *|
13 |*                                                                           *|
14 |*       Copyright 1993-2003 NVIDIA, Corporation.  All rights reserved.      *|
15 |*                                                                           *|
16 |*     NVIDIA, CORPORATION MAKES NO REPRESENTATION ABOUT THE SUITABILITY     *|
17 |*     OF  THIS SOURCE  CODE  FOR ANY PURPOSE.  IT IS  PROVIDED  "AS IS"     *|
18 |*     WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.  NVIDIA, CORPOR-     *|
19 |*     ATION DISCLAIMS ALL WARRANTIES  WITH REGARD  TO THIS SOURCE CODE,     *|
20 |*     INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGE-     *|
21 |*     MENT,  AND FITNESS  FOR A PARTICULAR PURPOSE.   IN NO EVENT SHALL     *|
22 |*     NVIDIA, CORPORATION  BE LIABLE FOR ANY SPECIAL,  INDIRECT,  INCI-     *|
23 |*     DENTAL, OR CONSEQUENTIAL DAMAGES,  OR ANY DAMAGES  WHATSOEVER RE-     *|
24 |*     SULTING FROM LOSS OF USE,  DATA OR PROFITS,  WHETHER IN AN ACTION     *|
25 |*     OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION,  ARISING OUT OF     *|
26 |*     OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE.     *|
27 |*                                                                           *|
28 |*     U.S. Government  End  Users.   This source code  is a "commercial     *|
29 |*     item,"  as that  term is  defined at  48 C.F.R. 2.101 (OCT 1995),     *|
30 |*     consisting  of "commercial  computer  software"  and  "commercial     *|
31 |*     computer  software  documentation,"  as such  terms  are  used in     *|
32 |*     48 C.F.R. 12.212 (SEPT 1995)  and is provided to the U.S. Govern-     *|
33 |*     ment only as  a commercial end item.   Consistent with  48 C.F.R.     *|
34 |*     12.212 and  48 C.F.R. 227.7202-1 through  227.7202-4 (JUNE 1995),     *|
35 |*     all U.S. Government End Users  acquire the source code  with only     *|
36 |*     those rights set forth herein.                                        *|
37 |*                                                                           *|
38  \***************************************************************************/
39 
40 /*
41  * GPL Licensing Note - According to Mark Vojkovich, author of the Xorg/
42  * XFree86 'nv' driver, this source code is provided under MIT-style licensing
43  * where the source code is provided "as is" without warranty of any kind.
44  * The only usage restriction is for the copyright notices to be retained
45  * whenever code is used.
46  *
47  * Antonino Daplas <adaplas@pol.net> 2005-03-11
48  */
49 
50 /* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/nv/nv_hw.c,v 1.4 2003/11/03 05:11:25 tsi Exp $ */
51 
52 #include <linux/pci.h>
53 #include "nv_type.h"
54 #include "nv_local.h"
55 #include "nv_proto.h"
56 
57 void NVLockUnlock(struct nvidia_par *par, int Lock)
58 {
59 	u8 cr11;
60 
61 	VGA_WR08(par->PCIO, 0x3D4, 0x1F);
62 	VGA_WR08(par->PCIO, 0x3D5, Lock ? 0x99 : 0x57);
63 
64 	VGA_WR08(par->PCIO, 0x3D4, 0x11);
65 	cr11 = VGA_RD08(par->PCIO, 0x3D5);
66 	if (Lock)
67 		cr11 |= 0x80;
68 	else
69 		cr11 &= ~0x80;
70 	VGA_WR08(par->PCIO, 0x3D5, cr11);
71 }
72 
73 int NVShowHideCursor(struct nvidia_par *par, int ShowHide)
74 {
75 	int cur = par->CurrentState->cursor1;
76 
77 	par->CurrentState->cursor1 = (par->CurrentState->cursor1 & 0xFE) |
78 	    (ShowHide & 0x01);
79 	VGA_WR08(par->PCIO, 0x3D4, 0x31);
80 	VGA_WR08(par->PCIO, 0x3D5, par->CurrentState->cursor1);
81 
82 	if (par->Architecture == NV_ARCH_40)
83 		NV_WR32(par->PRAMDAC, 0x0300, NV_RD32(par->PRAMDAC, 0x0300));
84 
85 	return (cur & 0x01);
86 }
87 
88 /****************************************************************************\
89 *                                                                            *
90 * The video arbitration routines calculate some "magic" numbers.  Fixes      *
91 * the snow seen when accessing the framebuffer without it.                   *
92 * It just works (I hope).                                                    *
93 *                                                                            *
94 \****************************************************************************/
95 
96 typedef struct {
97 	int graphics_lwm;
98 	int video_lwm;
99 	int graphics_burst_size;
100 	int video_burst_size;
101 	int valid;
102 } nv4_fifo_info;
103 
104 typedef struct {
105 	int pclk_khz;
106 	int mclk_khz;
107 	int nvclk_khz;
108 	char mem_page_miss;
109 	char mem_latency;
110 	int memory_width;
111 	char enable_video;
112 	char gr_during_vid;
113 	char pix_bpp;
114 	char mem_aligned;
115 	char enable_mp;
116 } nv4_sim_state;
117 
118 typedef struct {
119 	int graphics_lwm;
120 	int video_lwm;
121 	int graphics_burst_size;
122 	int video_burst_size;
123 	int valid;
124 } nv10_fifo_info;
125 
126 typedef struct {
127 	int pclk_khz;
128 	int mclk_khz;
129 	int nvclk_khz;
130 	char mem_page_miss;
131 	char mem_latency;
132 	u32 memory_type;
133 	int memory_width;
134 	char enable_video;
135 	char gr_during_vid;
136 	char pix_bpp;
137 	char mem_aligned;
138 	char enable_mp;
139 } nv10_sim_state;
140 
141 static void nvGetClocks(struct nvidia_par *par, unsigned int *MClk,
142 			unsigned int *NVClk)
143 {
144 	unsigned int pll, N, M, MB, NB, P;
145 
146 	if (par->Architecture >= NV_ARCH_40) {
147 		pll = NV_RD32(par->PMC, 0x4020);
148 		P = (pll >> 16) & 0x07;
149 		pll = NV_RD32(par->PMC, 0x4024);
150 		M = pll & 0xFF;
151 		N = (pll >> 8) & 0xFF;
152 		if (((par->Chipset & 0xfff0) == 0x0290) ||
153 		    ((par->Chipset & 0xfff0) == 0x0390)) {
154 			MB = 1;
155 			NB = 1;
156 		} else {
157 			MB = (pll >> 16) & 0xFF;
158 			NB = (pll >> 24) & 0xFF;
159 		}
160 		*MClk = ((N * NB * par->CrystalFreqKHz) / (M * MB)) >> P;
161 
162 		pll = NV_RD32(par->PMC, 0x4000);
163 		P = (pll >> 16) & 0x07;
164 		pll = NV_RD32(par->PMC, 0x4004);
165 		M = pll & 0xFF;
166 		N = (pll >> 8) & 0xFF;
167 		MB = (pll >> 16) & 0xFF;
168 		NB = (pll >> 24) & 0xFF;
169 
170 		*NVClk = ((N * NB * par->CrystalFreqKHz) / (M * MB)) >> P;
171 	} else if (par->twoStagePLL) {
172 		pll = NV_RD32(par->PRAMDAC0, 0x0504);
173 		M = pll & 0xFF;
174 		N = (pll >> 8) & 0xFF;
175 		P = (pll >> 16) & 0x0F;
176 		pll = NV_RD32(par->PRAMDAC0, 0x0574);
177 		if (pll & 0x80000000) {
178 			MB = pll & 0xFF;
179 			NB = (pll >> 8) & 0xFF;
180 		} else {
181 			MB = 1;
182 			NB = 1;
183 		}
184 		*MClk = ((N * NB * par->CrystalFreqKHz) / (M * MB)) >> P;
185 
186 		pll = NV_RD32(par->PRAMDAC0, 0x0500);
187 		M = pll & 0xFF;
188 		N = (pll >> 8) & 0xFF;
189 		P = (pll >> 16) & 0x0F;
190 		pll = NV_RD32(par->PRAMDAC0, 0x0570);
191 		if (pll & 0x80000000) {
192 			MB = pll & 0xFF;
193 			NB = (pll >> 8) & 0xFF;
194 		} else {
195 			MB = 1;
196 			NB = 1;
197 		}
198 		*NVClk = ((N * NB * par->CrystalFreqKHz) / (M * MB)) >> P;
199 	} else
200 	    if (((par->Chipset & 0x0ff0) == 0x0300) ||
201 		((par->Chipset & 0x0ff0) == 0x0330)) {
202 		pll = NV_RD32(par->PRAMDAC0, 0x0504);
203 		M = pll & 0x0F;
204 		N = (pll >> 8) & 0xFF;
205 		P = (pll >> 16) & 0x07;
206 		if (pll & 0x00000080) {
207 			MB = (pll >> 4) & 0x07;
208 			NB = (pll >> 19) & 0x1f;
209 		} else {
210 			MB = 1;
211 			NB = 1;
212 		}
213 		*MClk = ((N * NB * par->CrystalFreqKHz) / (M * MB)) >> P;
214 
215 		pll = NV_RD32(par->PRAMDAC0, 0x0500);
216 		M = pll & 0x0F;
217 		N = (pll >> 8) & 0xFF;
218 		P = (pll >> 16) & 0x07;
219 		if (pll & 0x00000080) {
220 			MB = (pll >> 4) & 0x07;
221 			NB = (pll >> 19) & 0x1f;
222 		} else {
223 			MB = 1;
224 			NB = 1;
225 		}
226 		*NVClk = ((N * NB * par->CrystalFreqKHz) / (M * MB)) >> P;
227 	} else {
228 		pll = NV_RD32(par->PRAMDAC0, 0x0504);
229 		M = pll & 0xFF;
230 		N = (pll >> 8) & 0xFF;
231 		P = (pll >> 16) & 0x0F;
232 		*MClk = (N * par->CrystalFreqKHz / M) >> P;
233 
234 		pll = NV_RD32(par->PRAMDAC0, 0x0500);
235 		M = pll & 0xFF;
236 		N = (pll >> 8) & 0xFF;
237 		P = (pll >> 16) & 0x0F;
238 		*NVClk = (N * par->CrystalFreqKHz / M) >> P;
239 	}
240 }
241 
242 static void nv4CalcArbitration(nv4_fifo_info * fifo, nv4_sim_state * arb)
243 {
244 	int data, pagemiss, cas, width, video_enable, bpp;
245 	int nvclks, mclks, pclks, vpagemiss, crtpagemiss, vbs;
246 	int found, mclk_extra, mclk_loop, cbs, m1, p1;
247 	int mclk_freq, pclk_freq, nvclk_freq, mp_enable;
248 	int us_m, us_n, us_p, video_drain_rate, crtc_drain_rate;
249 	int vpm_us, us_video, vlwm, video_fill_us, cpm_us, us_crt, clwm;
250 
251 	fifo->valid = 1;
252 	pclk_freq = arb->pclk_khz;
253 	mclk_freq = arb->mclk_khz;
254 	nvclk_freq = arb->nvclk_khz;
255 	pagemiss = arb->mem_page_miss;
256 	cas = arb->mem_latency;
257 	width = arb->memory_width >> 6;
258 	video_enable = arb->enable_video;
259 	bpp = arb->pix_bpp;
260 	mp_enable = arb->enable_mp;
261 	clwm = 0;
262 	vlwm = 0;
263 	cbs = 128;
264 	pclks = 2;
265 	nvclks = 2;
266 	nvclks += 2;
267 	nvclks += 1;
268 	mclks = 5;
269 	mclks += 3;
270 	mclks += 1;
271 	mclks += cas;
272 	mclks += 1;
273 	mclks += 1;
274 	mclks += 1;
275 	mclks += 1;
276 	mclk_extra = 3;
277 	nvclks += 2;
278 	nvclks += 1;
279 	nvclks += 1;
280 	nvclks += 1;
281 	if (mp_enable)
282 		mclks += 4;
283 	nvclks += 0;
284 	pclks += 0;
285 	found = 0;
286 	vbs = 0;
287 	while (found != 1) {
288 		fifo->valid = 1;
289 		found = 1;
290 		mclk_loop = mclks + mclk_extra;
291 		us_m = mclk_loop * 1000 * 1000 / mclk_freq;
292 		us_n = nvclks * 1000 * 1000 / nvclk_freq;
293 		us_p = nvclks * 1000 * 1000 / pclk_freq;
294 		if (video_enable) {
295 			video_drain_rate = pclk_freq * 2;
296 			crtc_drain_rate = pclk_freq * bpp / 8;
297 			vpagemiss = 2;
298 			vpagemiss += 1;
299 			crtpagemiss = 2;
300 			vpm_us =
301 			    (vpagemiss * pagemiss) * 1000 * 1000 / mclk_freq;
302 			if (nvclk_freq * 2 > mclk_freq * width)
303 				video_fill_us =
304 				    cbs * 1000 * 1000 / 16 / nvclk_freq;
305 			else
306 				video_fill_us =
307 				    cbs * 1000 * 1000 / (8 * width) /
308 				    mclk_freq;
309 			us_video = vpm_us + us_m + us_n + us_p + video_fill_us;
310 			vlwm = us_video * video_drain_rate / (1000 * 1000);
311 			vlwm++;
312 			vbs = 128;
313 			if (vlwm > 128)
314 				vbs = 64;
315 			if (vlwm > (256 - 64))
316 				vbs = 32;
317 			if (nvclk_freq * 2 > mclk_freq * width)
318 				video_fill_us =
319 				    vbs * 1000 * 1000 / 16 / nvclk_freq;
320 			else
321 				video_fill_us =
322 				    vbs * 1000 * 1000 / (8 * width) /
323 				    mclk_freq;
324 			cpm_us =
325 			    crtpagemiss * pagemiss * 1000 * 1000 / mclk_freq;
326 			us_crt =
327 			    us_video + video_fill_us + cpm_us + us_m + us_n +
328 			    us_p;
329 			clwm = us_crt * crtc_drain_rate / (1000 * 1000);
330 			clwm++;
331 		} else {
332 			crtc_drain_rate = pclk_freq * bpp / 8;
333 			crtpagemiss = 2;
334 			crtpagemiss += 1;
335 			cpm_us =
336 			    crtpagemiss * pagemiss * 1000 * 1000 / mclk_freq;
337 			us_crt = cpm_us + us_m + us_n + us_p;
338 			clwm = us_crt * crtc_drain_rate / (1000 * 1000);
339 			clwm++;
340 		}
341 		m1 = clwm + cbs - 512;
342 		p1 = m1 * pclk_freq / mclk_freq;
343 		p1 = p1 * bpp / 8;
344 		if ((p1 < m1) && (m1 > 0)) {
345 			fifo->valid = 0;
346 			found = 0;
347 			if (mclk_extra == 0)
348 				found = 1;
349 			mclk_extra--;
350 		} else if (video_enable) {
351 			if ((clwm > 511) || (vlwm > 255)) {
352 				fifo->valid = 0;
353 				found = 0;
354 				if (mclk_extra == 0)
355 					found = 1;
356 				mclk_extra--;
357 			}
358 		} else {
359 			if (clwm > 519) {
360 				fifo->valid = 0;
361 				found = 0;
362 				if (mclk_extra == 0)
363 					found = 1;
364 				mclk_extra--;
365 			}
366 		}
367 		if (clwm < 384)
368 			clwm = 384;
369 		if (vlwm < 128)
370 			vlwm = 128;
371 		data = (int)(clwm);
372 		fifo->graphics_lwm = data;
373 		fifo->graphics_burst_size = 128;
374 		data = (int)((vlwm + 15));
375 		fifo->video_lwm = data;
376 		fifo->video_burst_size = vbs;
377 	}
378 }
379 
380 static void nv4UpdateArbitrationSettings(unsigned VClk,
381 					 unsigned pixelDepth,
382 					 unsigned *burst,
383 					 unsigned *lwm, struct nvidia_par *par)
384 {
385 	nv4_fifo_info fifo_data;
386 	nv4_sim_state sim_data;
387 	unsigned int MClk, NVClk, cfg1;
388 
389 	nvGetClocks(par, &MClk, &NVClk);
390 
391 	cfg1 = NV_RD32(par->PFB, 0x00000204);
392 	sim_data.pix_bpp = (char)pixelDepth;
393 	sim_data.enable_video = 0;
394 	sim_data.enable_mp = 0;
395 	sim_data.memory_width = (NV_RD32(par->PEXTDEV, 0x0000) & 0x10) ?
396 	    128 : 64;
397 	sim_data.mem_latency = (char)cfg1 & 0x0F;
398 	sim_data.mem_aligned = 1;
399 	sim_data.mem_page_miss =
400 	    (char)(((cfg1 >> 4) & 0x0F) + ((cfg1 >> 31) & 0x01));
401 	sim_data.gr_during_vid = 0;
402 	sim_data.pclk_khz = VClk;
403 	sim_data.mclk_khz = MClk;
404 	sim_data.nvclk_khz = NVClk;
405 	nv4CalcArbitration(&fifo_data, &sim_data);
406 	if (fifo_data.valid) {
407 		int b = fifo_data.graphics_burst_size >> 4;
408 		*burst = 0;
409 		while (b >>= 1)
410 			(*burst)++;
411 		*lwm = fifo_data.graphics_lwm >> 3;
412 	}
413 }
414 
415 static void nv10CalcArbitration(nv10_fifo_info * fifo, nv10_sim_state * arb)
416 {
417 	int data, pagemiss, width, video_enable, bpp;
418 	int nvclks, mclks, pclks, vpagemiss, crtpagemiss;
419 	int nvclk_fill;
420 	int found, mclk_extra, mclk_loop, cbs, m1;
421 	int mclk_freq, pclk_freq, nvclk_freq, mp_enable;
422 	int us_m, us_m_min, us_n, us_p, crtc_drain_rate;
423 	int vus_m;
424 	int vpm_us, us_video, cpm_us, us_crt, clwm;
425 	int clwm_rnd_down;
426 	int m2us, us_pipe_min, p1clk, p2;
427 	int min_mclk_extra;
428 	int us_min_mclk_extra;
429 
430 	fifo->valid = 1;
431 	pclk_freq = arb->pclk_khz;	/* freq in KHz */
432 	mclk_freq = arb->mclk_khz;
433 	nvclk_freq = arb->nvclk_khz;
434 	pagemiss = arb->mem_page_miss;
435 	width = arb->memory_width / 64;
436 	video_enable = arb->enable_video;
437 	bpp = arb->pix_bpp;
438 	mp_enable = arb->enable_mp;
439 	clwm = 0;
440 
441 	cbs = 512;
442 
443 	pclks = 4;	/* lwm detect. */
444 
445 	nvclks = 3;	/* lwm -> sync. */
446 	nvclks += 2;	/* fbi bus cycles (1 req + 1 busy) */
447 	/* 2 edge sync.  may be very close to edge so just put one. */
448 	mclks = 1;
449 	mclks += 1;	/* arb_hp_req */
450 	mclks += 5;	/* ap_hp_req   tiling pipeline */
451 
452 	mclks += 2;	/* tc_req     latency fifo */
453 	mclks += 2;	/* fb_cas_n_  memory request to fbio block */
454 	mclks += 7;	/* sm_d_rdv   data returned from fbio block */
455 
456 	/* fb.rd.d.Put_gc   need to accumulate 256 bits for read */
457 	if (arb->memory_type == 0)
458 		if (arb->memory_width == 64)	/* 64 bit bus */
459 			mclks += 4;
460 		else
461 			mclks += 2;
462 	else if (arb->memory_width == 64)	/* 64 bit bus */
463 		mclks += 2;
464 	else
465 		mclks += 1;
466 
467 	if ((!video_enable) && (arb->memory_width == 128)) {
468 		mclk_extra = (bpp == 32) ? 31 : 42;	/* Margin of error */
469 		min_mclk_extra = 17;
470 	} else {
471 		mclk_extra = (bpp == 32) ? 8 : 4;	/* Margin of error */
472 		/* mclk_extra = 4; *//* Margin of error */
473 		min_mclk_extra = 18;
474 	}
475 
476 	/* 2 edge sync.  may be very close to edge so just put one. */
477 	nvclks += 1;
478 	nvclks += 1;		/* fbi_d_rdv_n */
479 	nvclks += 1;		/* Fbi_d_rdata */
480 	nvclks += 1;		/* crtfifo load */
481 
482 	if (mp_enable)
483 		mclks += 4;	/* Mp can get in with a burst of 8. */
484 	/* Extra clocks determined by heuristics */
485 
486 	nvclks += 0;
487 	pclks += 0;
488 	found = 0;
489 	while (found != 1) {
490 		fifo->valid = 1;
491 		found = 1;
492 		mclk_loop = mclks + mclk_extra;
493 		/* Mclk latency in us */
494 		us_m = mclk_loop * 1000 * 1000 / mclk_freq;
495 		/* Minimum Mclk latency in us */
496 		us_m_min = mclks * 1000 * 1000 / mclk_freq;
497 		us_min_mclk_extra = min_mclk_extra * 1000 * 1000 / mclk_freq;
498 		/* nvclk latency in us */
499 		us_n = nvclks * 1000 * 1000 / nvclk_freq;
500 		/* nvclk latency in us */
501 		us_p = pclks * 1000 * 1000 / pclk_freq;
502 		us_pipe_min = us_m_min + us_n + us_p;
503 
504 		/* Mclk latency in us */
505 		vus_m = mclk_loop * 1000 * 1000 / mclk_freq;
506 
507 		if (video_enable) {
508 			crtc_drain_rate = pclk_freq * bpp / 8;	/* MB/s */
509 
510 			vpagemiss = 1;	/* self generating page miss */
511 			vpagemiss += 1;	/* One higher priority before */
512 
513 			crtpagemiss = 2;	/* self generating page miss */
514 			if (mp_enable)
515 				crtpagemiss += 1;	/* if MA0 conflict */
516 
517 			vpm_us =
518 			    (vpagemiss * pagemiss) * 1000 * 1000 / mclk_freq;
519 
520 			/* Video has separate read return path */
521 			us_video = vpm_us + vus_m;
522 
523 			cpm_us =
524 			    crtpagemiss * pagemiss * 1000 * 1000 / mclk_freq;
525 			/* Wait for video */
526 			us_crt = us_video
527 			    + cpm_us	/* CRT Page miss */
528 			    + us_m + us_n + us_p	/* other latency */
529 			    ;
530 
531 			clwm = us_crt * crtc_drain_rate / (1000 * 1000);
532 			/* fixed point <= float_point - 1.  Fixes that */
533 			clwm++;
534 		} else {
535 		    /* bpp * pclk/8 */
536 			crtc_drain_rate = pclk_freq * bpp / 8;
537 
538 			crtpagemiss = 1;	/* self generating page miss */
539 			crtpagemiss += 1;	/* MA0 page miss */
540 			if (mp_enable)
541 				crtpagemiss += 1;	/* if MA0 conflict */
542 			cpm_us =
543 			    crtpagemiss * pagemiss * 1000 * 1000 / mclk_freq;
544 			us_crt = cpm_us + us_m + us_n + us_p;
545 			clwm = us_crt * crtc_drain_rate / (1000 * 1000);
546 			/* fixed point <= float_point - 1.  Fixes that */
547 			clwm++;
548 
549 			/* Finally, a heuristic check when width == 64 bits */
550 			if (width == 1) {
551 				nvclk_fill = nvclk_freq * 8;
552 				if (crtc_drain_rate * 100 >= nvclk_fill * 102)
553 					/*Large number to fail */
554 					clwm = 0xfff;
555 
556 				else if (crtc_drain_rate * 100 >=
557 					 nvclk_fill * 98) {
558 					clwm = 1024;
559 					cbs = 512;
560 				}
561 			}
562 		}
563 
564 		/*
565 		   Overfill check:
566 		 */
567 
568 		clwm_rnd_down = ((int)clwm / 8) * 8;
569 		if (clwm_rnd_down < clwm)
570 			clwm += 8;
571 
572 		m1 = clwm + cbs - 1024;	/* Amount of overfill */
573 		m2us = us_pipe_min + us_min_mclk_extra;
574 
575 		/* pclk cycles to drain */
576 		p1clk = m2us * pclk_freq / (1000 * 1000);
577 		p2 = p1clk * bpp / 8;	/* bytes drained. */
578 
579 		if ((p2 < m1) && (m1 > 0)) {
580 			fifo->valid = 0;
581 			found = 0;
582 			if (min_mclk_extra == 0) {
583 				if (cbs <= 32) {
584 					/* Can't adjust anymore! */
585 					found = 1;
586 				} else {
587 					/* reduce the burst size */
588 					cbs = cbs / 2;
589 				}
590 			} else {
591 				min_mclk_extra--;
592 			}
593 		} else {
594 			if (clwm > 1023) {	/* Have some margin */
595 				fifo->valid = 0;
596 				found = 0;
597 				if (min_mclk_extra == 0)
598 					/* Can't adjust anymore! */
599 					found = 1;
600 				else
601 					min_mclk_extra--;
602 			}
603 		}
604 
605 		if (clwm < (1024 - cbs + 8))
606 			clwm = 1024 - cbs + 8;
607 		data = (int)(clwm);
608 		/*  printf("CRT LWM: %f bytes, prog: 0x%x, bs: 256\n",
609 		    clwm, data ); */
610 		fifo->graphics_lwm = data;
611 		fifo->graphics_burst_size = cbs;
612 
613 		fifo->video_lwm = 1024;
614 		fifo->video_burst_size = 512;
615 	}
616 }
617 
618 static void nv10UpdateArbitrationSettings(unsigned VClk,
619 					  unsigned pixelDepth,
620 					  unsigned *burst,
621 					  unsigned *lwm,
622 					  struct nvidia_par *par)
623 {
624 	nv10_fifo_info fifo_data;
625 	nv10_sim_state sim_data;
626 	unsigned int MClk, NVClk, cfg1;
627 
628 	nvGetClocks(par, &MClk, &NVClk);
629 
630 	cfg1 = NV_RD32(par->PFB, 0x0204);
631 	sim_data.pix_bpp = (char)pixelDepth;
632 	sim_data.enable_video = 1;
633 	sim_data.enable_mp = 0;
634 	sim_data.memory_type = (NV_RD32(par->PFB, 0x0200) & 0x01) ? 1 : 0;
635 	sim_data.memory_width = (NV_RD32(par->PEXTDEV, 0x0000) & 0x10) ?
636 	    128 : 64;
637 	sim_data.mem_latency = (char)cfg1 & 0x0F;
638 	sim_data.mem_aligned = 1;
639 	sim_data.mem_page_miss =
640 	    (char)(((cfg1 >> 4) & 0x0F) + ((cfg1 >> 31) & 0x01));
641 	sim_data.gr_during_vid = 0;
642 	sim_data.pclk_khz = VClk;
643 	sim_data.mclk_khz = MClk;
644 	sim_data.nvclk_khz = NVClk;
645 	nv10CalcArbitration(&fifo_data, &sim_data);
646 	if (fifo_data.valid) {
647 		int b = fifo_data.graphics_burst_size >> 4;
648 		*burst = 0;
649 		while (b >>= 1)
650 			(*burst)++;
651 		*lwm = fifo_data.graphics_lwm >> 3;
652 	}
653 }
654 
655 static void nv30UpdateArbitrationSettings (
656     struct nvidia_par *par,
657     unsigned int      *burst,
658     unsigned int      *lwm
659 )
660 {
661     unsigned int MClk, NVClk;
662     unsigned int fifo_size, burst_size, graphics_lwm;
663 
664     fifo_size = 2048;
665     burst_size = 512;
666     graphics_lwm = fifo_size - burst_size;
667 
668     nvGetClocks(par, &MClk, &NVClk);
669 
670     *burst = 0;
671     burst_size >>= 5;
672     while(burst_size >>= 1) (*burst)++;
673     *lwm = graphics_lwm >> 3;
674 }
675 
676 static void nForceUpdateArbitrationSettings(unsigned VClk,
677 					    unsigned pixelDepth,
678 					    unsigned *burst,
679 					    unsigned *lwm,
680 					    struct nvidia_par *par)
681 {
682 	nv10_fifo_info fifo_data;
683 	nv10_sim_state sim_data;
684 	unsigned int M, N, P, pll, MClk, NVClk, memctrl;
685 	struct pci_dev *dev;
686 	int domain = pci_domain_nr(par->pci_dev->bus);
687 
688 	if ((par->Chipset & 0x0FF0) == 0x01A0) {
689 		unsigned int uMClkPostDiv;
690 		dev = pci_get_domain_bus_and_slot(domain, 0, 3);
691 		pci_read_config_dword(dev, 0x6C, &uMClkPostDiv);
692 		uMClkPostDiv = (uMClkPostDiv >> 8) & 0xf;
693 
694 		if (!uMClkPostDiv)
695 			uMClkPostDiv = 4;
696 		MClk = 400000 / uMClkPostDiv;
697 	} else {
698 		dev = pci_get_domain_bus_and_slot(domain, 0, 5);
699 		pci_read_config_dword(dev, 0x4c, &MClk);
700 		MClk /= 1000;
701 	}
702 	pci_dev_put(dev);
703 	pll = NV_RD32(par->PRAMDAC0, 0x0500);
704 	M = (pll >> 0) & 0xFF;
705 	N = (pll >> 8) & 0xFF;
706 	P = (pll >> 16) & 0x0F;
707 	NVClk = (N * par->CrystalFreqKHz / M) >> P;
708 	sim_data.pix_bpp = (char)pixelDepth;
709 	sim_data.enable_video = 0;
710 	sim_data.enable_mp = 0;
711 	dev = pci_get_domain_bus_and_slot(domain, 0, 1);
712 	pci_read_config_dword(dev, 0x7C, &sim_data.memory_type);
713 	pci_dev_put(dev);
714 	sim_data.memory_type = (sim_data.memory_type >> 12) & 1;
715 	sim_data.memory_width = 64;
716 
717 	dev = pci_get_domain_bus_and_slot(domain, 0, 3);
718 	pci_read_config_dword(dev, 0, &memctrl);
719 	pci_dev_put(dev);
720 	memctrl >>= 16;
721 
722 	if ((memctrl == 0x1A9) || (memctrl == 0x1AB) || (memctrl == 0x1ED)) {
723 		u32 dimm[3];
724 
725 		dev = pci_get_domain_bus_and_slot(domain, 0, 2);
726 		pci_read_config_dword(dev, 0x40, &dimm[0]);
727 		dimm[0] = (dimm[0] >> 8) & 0x4f;
728 		pci_read_config_dword(dev, 0x44, &dimm[1]);
729 		dimm[1] = (dimm[1] >> 8) & 0x4f;
730 		pci_read_config_dword(dev, 0x48, &dimm[2]);
731 		dimm[2] = (dimm[2] >> 8) & 0x4f;
732 
733 		if ((dimm[0] + dimm[1]) != dimm[2]) {
734 			printk("nvidiafb: your nForce DIMMs are not arranged "
735 			       "in optimal banks!\n");
736 		}
737 		pci_dev_put(dev);
738 	}
739 
740 	sim_data.mem_latency = 3;
741 	sim_data.mem_aligned = 1;
742 	sim_data.mem_page_miss = 10;
743 	sim_data.gr_during_vid = 0;
744 	sim_data.pclk_khz = VClk;
745 	sim_data.mclk_khz = MClk;
746 	sim_data.nvclk_khz = NVClk;
747 	nv10CalcArbitration(&fifo_data, &sim_data);
748 	if (fifo_data.valid) {
749 		int b = fifo_data.graphics_burst_size >> 4;
750 		*burst = 0;
751 		while (b >>= 1)
752 			(*burst)++;
753 		*lwm = fifo_data.graphics_lwm >> 3;
754 	}
755 }
756 
757 /****************************************************************************\
758 *                                                                            *
759 *                          RIVA Mode State Routines                          *
760 *                                                                            *
761 \****************************************************************************/
762 
763 /*
764  * Calculate the Video Clock parameters for the PLL.
765  */
766 static void CalcVClock(int clockIn,
767 		       int *clockOut, u32 * pllOut, struct nvidia_par *par)
768 {
769 	unsigned lowM, highM;
770 	unsigned DeltaNew, DeltaOld;
771 	unsigned VClk, Freq;
772 	unsigned M, N, P;
773 
774 	DeltaOld = 0xFFFFFFFF;
775 
776 	VClk = (unsigned)clockIn;
777 
778 	if (par->CrystalFreqKHz == 13500) {
779 		lowM = 7;
780 		highM = 13;
781 	} else {
782 		lowM = 8;
783 		highM = 14;
784 	}
785 
786 	for (P = 0; P <= 4; P++) {
787 		Freq = VClk << P;
788 		if ((Freq >= 128000) && (Freq <= 350000)) {
789 			for (M = lowM; M <= highM; M++) {
790 				N = ((VClk << P) * M) / par->CrystalFreqKHz;
791 				if (N <= 255) {
792 					Freq =
793 					    ((par->CrystalFreqKHz * N) /
794 					     M) >> P;
795 					if (Freq > VClk)
796 						DeltaNew = Freq - VClk;
797 					else
798 						DeltaNew = VClk - Freq;
799 					if (DeltaNew < DeltaOld) {
800 						*pllOut =
801 						    (P << 16) | (N << 8) | M;
802 						*clockOut = Freq;
803 						DeltaOld = DeltaNew;
804 					}
805 				}
806 			}
807 		}
808 	}
809 }
810 
811 static void CalcVClock2Stage(int clockIn,
812 			     int *clockOut,
813 			     u32 * pllOut,
814 			     u32 * pllBOut, struct nvidia_par *par)
815 {
816 	unsigned DeltaNew, DeltaOld;
817 	unsigned VClk, Freq;
818 	unsigned M, N, P;
819 
820 	DeltaOld = 0xFFFFFFFF;
821 
822 	*pllBOut = 0x80000401;	/* fixed at x4 for now */
823 
824 	VClk = (unsigned)clockIn;
825 
826 	for (P = 0; P <= 6; P++) {
827 		Freq = VClk << P;
828 		if ((Freq >= 400000) && (Freq <= 1000000)) {
829 			for (M = 1; M <= 13; M++) {
830 				N = ((VClk << P) * M) /
831 				    (par->CrystalFreqKHz << 2);
832 				if ((N >= 5) && (N <= 255)) {
833 					Freq =
834 					    (((par->CrystalFreqKHz << 2) * N) /
835 					     M) >> P;
836 					if (Freq > VClk)
837 						DeltaNew = Freq - VClk;
838 					else
839 						DeltaNew = VClk - Freq;
840 					if (DeltaNew < DeltaOld) {
841 						*pllOut =
842 						    (P << 16) | (N << 8) | M;
843 						*clockOut = Freq;
844 						DeltaOld = DeltaNew;
845 					}
846 				}
847 			}
848 		}
849 	}
850 }
851 
852 /*
853  * Calculate extended mode parameters (SVGA) and save in a
854  * mode state structure.
855  */
856 void NVCalcStateExt(struct nvidia_par *par,
857 		    RIVA_HW_STATE * state,
858 		    int bpp,
859 		    int width,
860 		    int hDisplaySize, int height, int dotClock, int flags)
861 {
862 	int pixelDepth, VClk = 0;
863 	/*
864 	 * Save mode parameters.
865 	 */
866 	state->bpp = bpp;	/* this is not bitsPerPixel, it's 8,15,16,32 */
867 	state->width = width;
868 	state->height = height;
869 	/*
870 	 * Extended RIVA registers.
871 	 */
872 	pixelDepth = (bpp + 1) / 8;
873 	if (par->twoStagePLL)
874 		CalcVClock2Stage(dotClock, &VClk, &state->pll, &state->pllB,
875 				 par);
876 	else
877 		CalcVClock(dotClock, &VClk, &state->pll, par);
878 
879 	switch (par->Architecture) {
880 	case NV_ARCH_04:
881 		nv4UpdateArbitrationSettings(VClk,
882 					     pixelDepth * 8,
883 					     &(state->arbitration0),
884 					     &(state->arbitration1), par);
885 		state->cursor0 = 0x00;
886 		state->cursor1 = 0xbC;
887 		if (flags & FB_VMODE_DOUBLE)
888 			state->cursor1 |= 2;
889 		state->cursor2 = 0x00000000;
890 		state->pllsel = 0x10000700;
891 		state->config = 0x00001114;
892 		state->general = bpp == 16 ? 0x00101100 : 0x00100100;
893 		state->repaint1 = hDisplaySize < 1280 ? 0x04 : 0x00;
894 		break;
895 	case NV_ARCH_40:
896 		if (!par->FlatPanel)
897 			state->control = NV_RD32(par->PRAMDAC0, 0x0580) &
898 				0xeffffeff;
899 		fallthrough;
900 	case NV_ARCH_10:
901 	case NV_ARCH_20:
902 	case NV_ARCH_30:
903 	default:
904 		if ((par->Chipset & 0xfff0) == 0x0240 ||
905 		    (par->Chipset & 0xfff0) == 0x03d0) {
906 			state->arbitration0 = 256;
907 			state->arbitration1 = 0x0480;
908 		} else if (((par->Chipset & 0xffff) == 0x01A0) ||
909 		    ((par->Chipset & 0xffff) == 0x01f0)) {
910 			nForceUpdateArbitrationSettings(VClk,
911 							pixelDepth * 8,
912 							&(state->arbitration0),
913 							&(state->arbitration1),
914 							par);
915 		} else if (par->Architecture < NV_ARCH_30) {
916 			nv10UpdateArbitrationSettings(VClk,
917 						      pixelDepth * 8,
918 						      &(state->arbitration0),
919 						      &(state->arbitration1),
920 						      par);
921 		} else {
922 			nv30UpdateArbitrationSettings(par,
923 						      &(state->arbitration0),
924 						      &(state->arbitration1));
925 		}
926 
927 		state->cursor0 = 0x80 | (par->CursorStart >> 17);
928 		state->cursor1 = (par->CursorStart >> 11) << 2;
929 		state->cursor2 = par->CursorStart >> 24;
930 		if (flags & FB_VMODE_DOUBLE)
931 			state->cursor1 |= 2;
932 		state->pllsel = 0x10000700;
933 		state->config = NV_RD32(par->PFB, 0x00000200);
934 		state->general = bpp == 16 ? 0x00101100 : 0x00100100;
935 		state->repaint1 = hDisplaySize < 1280 ? 0x04 : 0x00;
936 		break;
937 	}
938 
939 	if (bpp != 8)		/* DirectColor */
940 		state->general |= 0x00000030;
941 
942 	state->repaint0 = (((width / 8) * pixelDepth) & 0x700) >> 3;
943 	state->pixel = (pixelDepth > 2) ? 3 : pixelDepth;
944 }
945 
946 void NVLoadStateExt(struct nvidia_par *par, RIVA_HW_STATE * state)
947 {
948 	int i, j;
949 
950 	NV_WR32(par->PMC, 0x0140, 0x00000000);
951 	NV_WR32(par->PMC, 0x0200, 0xFFFF00FF);
952 	NV_WR32(par->PMC, 0x0200, 0xFFFFFFFF);
953 
954 	NV_WR32(par->PTIMER, 0x0200 * 4, 0x00000008);
955 	NV_WR32(par->PTIMER, 0x0210 * 4, 0x00000003);
956 	NV_WR32(par->PTIMER, 0x0140 * 4, 0x00000000);
957 	NV_WR32(par->PTIMER, 0x0100 * 4, 0xFFFFFFFF);
958 
959 	if (par->Architecture == NV_ARCH_04) {
960 		if (state)
961 			NV_WR32(par->PFB, 0x0200, state->config);
962 	} else if ((par->Architecture < NV_ARCH_40) ||
963 		   (par->Chipset & 0xfff0) == 0x0040) {
964 		for (i = 0; i < 8; i++) {
965 			NV_WR32(par->PFB, 0x0240 + (i * 0x10), 0);
966 			NV_WR32(par->PFB, 0x0244 + (i * 0x10),
967 				par->FbMapSize - 1);
968 		}
969 	} else {
970 		int regions = 12;
971 
972 		if (((par->Chipset & 0xfff0) == 0x0090) ||
973 		    ((par->Chipset & 0xfff0) == 0x01D0) ||
974 		    ((par->Chipset & 0xfff0) == 0x0290) ||
975 		    ((par->Chipset & 0xfff0) == 0x0390) ||
976 		    ((par->Chipset & 0xfff0) == 0x03D0))
977 			regions = 15;
978 		for(i = 0; i < regions; i++) {
979 			NV_WR32(par->PFB, 0x0600 + (i * 0x10), 0);
980 			NV_WR32(par->PFB, 0x0604 + (i * 0x10),
981 				par->FbMapSize - 1);
982 		}
983 	}
984 
985 	if (par->Architecture >= NV_ARCH_40) {
986 		NV_WR32(par->PRAMIN, 0x0000 * 4, 0x80000010);
987 		NV_WR32(par->PRAMIN, 0x0001 * 4, 0x00101202);
988 		NV_WR32(par->PRAMIN, 0x0002 * 4, 0x80000011);
989 		NV_WR32(par->PRAMIN, 0x0003 * 4, 0x00101204);
990 		NV_WR32(par->PRAMIN, 0x0004 * 4, 0x80000012);
991 		NV_WR32(par->PRAMIN, 0x0005 * 4, 0x00101206);
992 		NV_WR32(par->PRAMIN, 0x0006 * 4, 0x80000013);
993 		NV_WR32(par->PRAMIN, 0x0007 * 4, 0x00101208);
994 		NV_WR32(par->PRAMIN, 0x0008 * 4, 0x80000014);
995 		NV_WR32(par->PRAMIN, 0x0009 * 4, 0x0010120A);
996 		NV_WR32(par->PRAMIN, 0x000A * 4, 0x80000015);
997 		NV_WR32(par->PRAMIN, 0x000B * 4, 0x0010120C);
998 		NV_WR32(par->PRAMIN, 0x000C * 4, 0x80000016);
999 		NV_WR32(par->PRAMIN, 0x000D * 4, 0x0010120E);
1000 		NV_WR32(par->PRAMIN, 0x000E * 4, 0x80000017);
1001 		NV_WR32(par->PRAMIN, 0x000F * 4, 0x00101210);
1002 		NV_WR32(par->PRAMIN, 0x0800 * 4, 0x00003000);
1003 		NV_WR32(par->PRAMIN, 0x0801 * 4, par->FbMapSize - 1);
1004 		NV_WR32(par->PRAMIN, 0x0802 * 4, 0x00000002);
1005 		NV_WR32(par->PRAMIN, 0x0808 * 4, 0x02080062);
1006 		NV_WR32(par->PRAMIN, 0x0809 * 4, 0x00000000);
1007 		NV_WR32(par->PRAMIN, 0x080A * 4, 0x00001200);
1008 		NV_WR32(par->PRAMIN, 0x080B * 4, 0x00001200);
1009 		NV_WR32(par->PRAMIN, 0x080C * 4, 0x00000000);
1010 		NV_WR32(par->PRAMIN, 0x080D * 4, 0x00000000);
1011 		NV_WR32(par->PRAMIN, 0x0810 * 4, 0x02080043);
1012 		NV_WR32(par->PRAMIN, 0x0811 * 4, 0x00000000);
1013 		NV_WR32(par->PRAMIN, 0x0812 * 4, 0x00000000);
1014 		NV_WR32(par->PRAMIN, 0x0813 * 4, 0x00000000);
1015 		NV_WR32(par->PRAMIN, 0x0814 * 4, 0x00000000);
1016 		NV_WR32(par->PRAMIN, 0x0815 * 4, 0x00000000);
1017 		NV_WR32(par->PRAMIN, 0x0818 * 4, 0x02080044);
1018 		NV_WR32(par->PRAMIN, 0x0819 * 4, 0x02000000);
1019 		NV_WR32(par->PRAMIN, 0x081A * 4, 0x00000000);
1020 		NV_WR32(par->PRAMIN, 0x081B * 4, 0x00000000);
1021 		NV_WR32(par->PRAMIN, 0x081C * 4, 0x00000000);
1022 		NV_WR32(par->PRAMIN, 0x081D * 4, 0x00000000);
1023 		NV_WR32(par->PRAMIN, 0x0820 * 4, 0x02080019);
1024 		NV_WR32(par->PRAMIN, 0x0821 * 4, 0x00000000);
1025 		NV_WR32(par->PRAMIN, 0x0822 * 4, 0x00000000);
1026 		NV_WR32(par->PRAMIN, 0x0823 * 4, 0x00000000);
1027 		NV_WR32(par->PRAMIN, 0x0824 * 4, 0x00000000);
1028 		NV_WR32(par->PRAMIN, 0x0825 * 4, 0x00000000);
1029 		NV_WR32(par->PRAMIN, 0x0828 * 4, 0x020A005C);
1030 		NV_WR32(par->PRAMIN, 0x0829 * 4, 0x00000000);
1031 		NV_WR32(par->PRAMIN, 0x082A * 4, 0x00000000);
1032 		NV_WR32(par->PRAMIN, 0x082B * 4, 0x00000000);
1033 		NV_WR32(par->PRAMIN, 0x082C * 4, 0x00000000);
1034 		NV_WR32(par->PRAMIN, 0x082D * 4, 0x00000000);
1035 		NV_WR32(par->PRAMIN, 0x0830 * 4, 0x0208009F);
1036 		NV_WR32(par->PRAMIN, 0x0831 * 4, 0x00000000);
1037 		NV_WR32(par->PRAMIN, 0x0832 * 4, 0x00001200);
1038 		NV_WR32(par->PRAMIN, 0x0833 * 4, 0x00001200);
1039 		NV_WR32(par->PRAMIN, 0x0834 * 4, 0x00000000);
1040 		NV_WR32(par->PRAMIN, 0x0835 * 4, 0x00000000);
1041 		NV_WR32(par->PRAMIN, 0x0838 * 4, 0x0208004A);
1042 		NV_WR32(par->PRAMIN, 0x0839 * 4, 0x02000000);
1043 		NV_WR32(par->PRAMIN, 0x083A * 4, 0x00000000);
1044 		NV_WR32(par->PRAMIN, 0x083B * 4, 0x00000000);
1045 		NV_WR32(par->PRAMIN, 0x083C * 4, 0x00000000);
1046 		NV_WR32(par->PRAMIN, 0x083D * 4, 0x00000000);
1047 		NV_WR32(par->PRAMIN, 0x0840 * 4, 0x02080077);
1048 		NV_WR32(par->PRAMIN, 0x0841 * 4, 0x00000000);
1049 		NV_WR32(par->PRAMIN, 0x0842 * 4, 0x00001200);
1050 		NV_WR32(par->PRAMIN, 0x0843 * 4, 0x00001200);
1051 		NV_WR32(par->PRAMIN, 0x0844 * 4, 0x00000000);
1052 		NV_WR32(par->PRAMIN, 0x0845 * 4, 0x00000000);
1053 		NV_WR32(par->PRAMIN, 0x084C * 4, 0x00003002);
1054 		NV_WR32(par->PRAMIN, 0x084D * 4, 0x00007FFF);
1055 		NV_WR32(par->PRAMIN, 0x084E * 4,
1056 			par->FbUsableSize | 0x00000002);
1057 
1058 #ifdef __BIG_ENDIAN
1059 		NV_WR32(par->PRAMIN, 0x080A * 4,
1060 			NV_RD32(par->PRAMIN, 0x080A * 4) | 0x01000000);
1061 		NV_WR32(par->PRAMIN, 0x0812 * 4,
1062 			NV_RD32(par->PRAMIN, 0x0812 * 4) | 0x01000000);
1063 		NV_WR32(par->PRAMIN, 0x081A * 4,
1064 			NV_RD32(par->PRAMIN, 0x081A * 4) | 0x01000000);
1065 		NV_WR32(par->PRAMIN, 0x0822 * 4,
1066 			NV_RD32(par->PRAMIN, 0x0822 * 4) | 0x01000000);
1067 		NV_WR32(par->PRAMIN, 0x082A * 4,
1068 			NV_RD32(par->PRAMIN, 0x082A * 4) | 0x01000000);
1069 		NV_WR32(par->PRAMIN, 0x0832 * 4,
1070 			NV_RD32(par->PRAMIN, 0x0832 * 4) | 0x01000000);
1071 		NV_WR32(par->PRAMIN, 0x083A * 4,
1072 			NV_RD32(par->PRAMIN, 0x083A * 4) | 0x01000000);
1073 		NV_WR32(par->PRAMIN, 0x0842 * 4,
1074 			NV_RD32(par->PRAMIN, 0x0842 * 4) | 0x01000000);
1075 		NV_WR32(par->PRAMIN, 0x0819 * 4, 0x01000000);
1076 		NV_WR32(par->PRAMIN, 0x0839 * 4, 0x01000000);
1077 #endif
1078 	} else {
1079 		NV_WR32(par->PRAMIN, 0x0000 * 4, 0x80000010);
1080 		NV_WR32(par->PRAMIN, 0x0001 * 4, 0x80011201);
1081 		NV_WR32(par->PRAMIN, 0x0002 * 4, 0x80000011);
1082 		NV_WR32(par->PRAMIN, 0x0003 * 4, 0x80011202);
1083 		NV_WR32(par->PRAMIN, 0x0004 * 4, 0x80000012);
1084 		NV_WR32(par->PRAMIN, 0x0005 * 4, 0x80011203);
1085 		NV_WR32(par->PRAMIN, 0x0006 * 4, 0x80000013);
1086 		NV_WR32(par->PRAMIN, 0x0007 * 4, 0x80011204);
1087 		NV_WR32(par->PRAMIN, 0x0008 * 4, 0x80000014);
1088 		NV_WR32(par->PRAMIN, 0x0009 * 4, 0x80011205);
1089 		NV_WR32(par->PRAMIN, 0x000A * 4, 0x80000015);
1090 		NV_WR32(par->PRAMIN, 0x000B * 4, 0x80011206);
1091 		NV_WR32(par->PRAMIN, 0x000C * 4, 0x80000016);
1092 		NV_WR32(par->PRAMIN, 0x000D * 4, 0x80011207);
1093 		NV_WR32(par->PRAMIN, 0x000E * 4, 0x80000017);
1094 		NV_WR32(par->PRAMIN, 0x000F * 4, 0x80011208);
1095 		NV_WR32(par->PRAMIN, 0x0800 * 4, 0x00003000);
1096 		NV_WR32(par->PRAMIN, 0x0801 * 4, par->FbMapSize - 1);
1097 		NV_WR32(par->PRAMIN, 0x0802 * 4, 0x00000002);
1098 		NV_WR32(par->PRAMIN, 0x0803 * 4, 0x00000002);
1099 		if (par->Architecture >= NV_ARCH_10)
1100 			NV_WR32(par->PRAMIN, 0x0804 * 4, 0x01008062);
1101 		else
1102 			NV_WR32(par->PRAMIN, 0x0804 * 4, 0x01008042);
1103 		NV_WR32(par->PRAMIN, 0x0805 * 4, 0x00000000);
1104 		NV_WR32(par->PRAMIN, 0x0806 * 4, 0x12001200);
1105 		NV_WR32(par->PRAMIN, 0x0807 * 4, 0x00000000);
1106 		NV_WR32(par->PRAMIN, 0x0808 * 4, 0x01008043);
1107 		NV_WR32(par->PRAMIN, 0x0809 * 4, 0x00000000);
1108 		NV_WR32(par->PRAMIN, 0x080A * 4, 0x00000000);
1109 		NV_WR32(par->PRAMIN, 0x080B * 4, 0x00000000);
1110 		NV_WR32(par->PRAMIN, 0x080C * 4, 0x01008044);
1111 		NV_WR32(par->PRAMIN, 0x080D * 4, 0x00000002);
1112 		NV_WR32(par->PRAMIN, 0x080E * 4, 0x00000000);
1113 		NV_WR32(par->PRAMIN, 0x080F * 4, 0x00000000);
1114 		NV_WR32(par->PRAMIN, 0x0810 * 4, 0x01008019);
1115 		NV_WR32(par->PRAMIN, 0x0811 * 4, 0x00000000);
1116 		NV_WR32(par->PRAMIN, 0x0812 * 4, 0x00000000);
1117 		NV_WR32(par->PRAMIN, 0x0813 * 4, 0x00000000);
1118 		NV_WR32(par->PRAMIN, 0x0814 * 4, 0x0100A05C);
1119 		NV_WR32(par->PRAMIN, 0x0815 * 4, 0x00000000);
1120 		NV_WR32(par->PRAMIN, 0x0816 * 4, 0x00000000);
1121 		NV_WR32(par->PRAMIN, 0x0817 * 4, 0x00000000);
1122 		if (par->WaitVSyncPossible)
1123 			NV_WR32(par->PRAMIN, 0x0818 * 4, 0x0100809F);
1124 		else
1125 			NV_WR32(par->PRAMIN, 0x0818 * 4, 0x0100805F);
1126 		NV_WR32(par->PRAMIN, 0x0819 * 4, 0x00000000);
1127 		NV_WR32(par->PRAMIN, 0x081A * 4, 0x12001200);
1128 		NV_WR32(par->PRAMIN, 0x081B * 4, 0x00000000);
1129 		NV_WR32(par->PRAMIN, 0x081C * 4, 0x0100804A);
1130 		NV_WR32(par->PRAMIN, 0x081D * 4, 0x00000002);
1131 		NV_WR32(par->PRAMIN, 0x081E * 4, 0x00000000);
1132 		NV_WR32(par->PRAMIN, 0x081F * 4, 0x00000000);
1133 		NV_WR32(par->PRAMIN, 0x0820 * 4, 0x01018077);
1134 		NV_WR32(par->PRAMIN, 0x0821 * 4, 0x00000000);
1135 		NV_WR32(par->PRAMIN, 0x0822 * 4, 0x12001200);
1136 		NV_WR32(par->PRAMIN, 0x0823 * 4, 0x00000000);
1137 		NV_WR32(par->PRAMIN, 0x0824 * 4, 0x00003002);
1138 		NV_WR32(par->PRAMIN, 0x0825 * 4, 0x00007FFF);
1139 		NV_WR32(par->PRAMIN, 0x0826 * 4,
1140 			par->FbUsableSize | 0x00000002);
1141 		NV_WR32(par->PRAMIN, 0x0827 * 4, 0x00000002);
1142 #ifdef __BIG_ENDIAN
1143 		NV_WR32(par->PRAMIN, 0x0804 * 4,
1144 			NV_RD32(par->PRAMIN, 0x0804 * 4) | 0x00080000);
1145 		NV_WR32(par->PRAMIN, 0x0808 * 4,
1146 			NV_RD32(par->PRAMIN, 0x0808 * 4) | 0x00080000);
1147 		NV_WR32(par->PRAMIN, 0x080C * 4,
1148 			NV_RD32(par->PRAMIN, 0x080C * 4) | 0x00080000);
1149 		NV_WR32(par->PRAMIN, 0x0810 * 4,
1150 			NV_RD32(par->PRAMIN, 0x0810 * 4) | 0x00080000);
1151 		NV_WR32(par->PRAMIN, 0x0814 * 4,
1152 			NV_RD32(par->PRAMIN, 0x0814 * 4) | 0x00080000);
1153 		NV_WR32(par->PRAMIN, 0x0818 * 4,
1154 			NV_RD32(par->PRAMIN, 0x0818 * 4) | 0x00080000);
1155 		NV_WR32(par->PRAMIN, 0x081C * 4,
1156 			NV_RD32(par->PRAMIN, 0x081C * 4) | 0x00080000);
1157 		NV_WR32(par->PRAMIN, 0x0820 * 4,
1158 			NV_RD32(par->PRAMIN, 0x0820 * 4) | 0x00080000);
1159 		NV_WR32(par->PRAMIN, 0x080D * 4, 0x00000001);
1160 		NV_WR32(par->PRAMIN, 0x081D * 4, 0x00000001);
1161 #endif
1162 	}
1163 	if (par->Architecture < NV_ARCH_10) {
1164 		if ((par->Chipset & 0x0fff) == 0x0020) {
1165 			NV_WR32(par->PRAMIN, 0x0824 * 4,
1166 				NV_RD32(par->PRAMIN, 0x0824 * 4) | 0x00020000);
1167 			NV_WR32(par->PRAMIN, 0x0826 * 4,
1168 				NV_RD32(par->PRAMIN,
1169 					0x0826 * 4) + par->FbAddress);
1170 		}
1171 		NV_WR32(par->PGRAPH, 0x0080, 0x000001FF);
1172 		NV_WR32(par->PGRAPH, 0x0080, 0x1230C000);
1173 		NV_WR32(par->PGRAPH, 0x0084, 0x72111101);
1174 		NV_WR32(par->PGRAPH, 0x0088, 0x11D5F071);
1175 		NV_WR32(par->PGRAPH, 0x008C, 0x0004FF31);
1176 		NV_WR32(par->PGRAPH, 0x008C, 0x4004FF31);
1177 		NV_WR32(par->PGRAPH, 0x0140, 0x00000000);
1178 		NV_WR32(par->PGRAPH, 0x0100, 0xFFFFFFFF);
1179 		NV_WR32(par->PGRAPH, 0x0170, 0x10010100);
1180 		NV_WR32(par->PGRAPH, 0x0710, 0xFFFFFFFF);
1181 		NV_WR32(par->PGRAPH, 0x0720, 0x00000001);
1182 		NV_WR32(par->PGRAPH, 0x0810, 0x00000000);
1183 		NV_WR32(par->PGRAPH, 0x0608, 0xFFFFFFFF);
1184 	} else {
1185 		NV_WR32(par->PGRAPH, 0x0080, 0xFFFFFFFF);
1186 		NV_WR32(par->PGRAPH, 0x0080, 0x00000000);
1187 
1188 		NV_WR32(par->PGRAPH, 0x0140, 0x00000000);
1189 		NV_WR32(par->PGRAPH, 0x0100, 0xFFFFFFFF);
1190 		NV_WR32(par->PGRAPH, 0x0144, 0x10010100);
1191 		NV_WR32(par->PGRAPH, 0x0714, 0xFFFFFFFF);
1192 		NV_WR32(par->PGRAPH, 0x0720, 0x00000001);
1193 		NV_WR32(par->PGRAPH, 0x0710,
1194 			NV_RD32(par->PGRAPH, 0x0710) & 0x0007ff00);
1195 		NV_WR32(par->PGRAPH, 0x0710,
1196 			NV_RD32(par->PGRAPH, 0x0710) | 0x00020100);
1197 
1198 		if (par->Architecture == NV_ARCH_10) {
1199 			NV_WR32(par->PGRAPH, 0x0084, 0x00118700);
1200 			NV_WR32(par->PGRAPH, 0x0088, 0x24E00810);
1201 			NV_WR32(par->PGRAPH, 0x008C, 0x55DE0030);
1202 
1203 			for (i = 0; i < 32; i++)
1204 				NV_WR32(&par->PGRAPH[(0x0B00 / 4) + i], 0,
1205 					NV_RD32(&par->PFB[(0x0240 / 4) + i],
1206 						0));
1207 
1208 			NV_WR32(par->PGRAPH, 0x640, 0);
1209 			NV_WR32(par->PGRAPH, 0x644, 0);
1210 			NV_WR32(par->PGRAPH, 0x684, par->FbMapSize - 1);
1211 			NV_WR32(par->PGRAPH, 0x688, par->FbMapSize - 1);
1212 
1213 			NV_WR32(par->PGRAPH, 0x0810, 0x00000000);
1214 			NV_WR32(par->PGRAPH, 0x0608, 0xFFFFFFFF);
1215 		} else {
1216 			if (par->Architecture >= NV_ARCH_40) {
1217 				NV_WR32(par->PGRAPH, 0x0084, 0x401287c0);
1218 				NV_WR32(par->PGRAPH, 0x008C, 0x60de8051);
1219 				NV_WR32(par->PGRAPH, 0x0090, 0x00008000);
1220 				NV_WR32(par->PGRAPH, 0x0610, 0x00be3c5f);
1221 				NV_WR32(par->PGRAPH, 0x0bc4,
1222 					NV_RD32(par->PGRAPH, 0x0bc4) |
1223 					0x00008000);
1224 
1225 				j = NV_RD32(par->REGS, 0x1540) & 0xff;
1226 
1227 				if (j) {
1228 					for (i = 0; !(j & 1); j >>= 1, i++);
1229 					NV_WR32(par->PGRAPH, 0x5000, i);
1230 				}
1231 
1232 				if ((par->Chipset & 0xfff0) == 0x0040) {
1233 					NV_WR32(par->PGRAPH, 0x09b0,
1234 						0x83280fff);
1235 					NV_WR32(par->PGRAPH, 0x09b4,
1236 						0x000000a0);
1237 				} else {
1238 					NV_WR32(par->PGRAPH, 0x0820,
1239 						0x83280eff);
1240 					NV_WR32(par->PGRAPH, 0x0824,
1241 						0x000000a0);
1242 				}
1243 
1244 				switch (par->Chipset & 0xfff0) {
1245 				case 0x0040:
1246 				case 0x0210:
1247 					NV_WR32(par->PGRAPH, 0x09b8,
1248 						0x0078e366);
1249 					NV_WR32(par->PGRAPH, 0x09bc,
1250 						0x0000014c);
1251 					NV_WR32(par->PFB, 0x033C,
1252 						NV_RD32(par->PFB, 0x33C) &
1253 						0xffff7fff);
1254 					break;
1255 				case 0x00C0:
1256 				case 0x0120:
1257 					NV_WR32(par->PGRAPH, 0x0828,
1258 						0x007596ff);
1259 					NV_WR32(par->PGRAPH, 0x082C,
1260 						0x00000108);
1261 					break;
1262 				case 0x0160:
1263 				case 0x01D0:
1264 				case 0x0240:
1265 				case 0x03D0:
1266 					NV_WR32(par->PMC, 0x1700,
1267 						NV_RD32(par->PFB, 0x020C));
1268 					NV_WR32(par->PMC, 0x1704, 0);
1269 					NV_WR32(par->PMC, 0x1708, 0);
1270 					NV_WR32(par->PMC, 0x170C,
1271 						NV_RD32(par->PFB, 0x020C));
1272 					NV_WR32(par->PGRAPH, 0x0860, 0);
1273 					NV_WR32(par->PGRAPH, 0x0864, 0);
1274 					NV_WR32(par->PRAMDAC, 0x0608,
1275 						NV_RD32(par->PRAMDAC,
1276 							0x0608) | 0x00100000);
1277 					break;
1278 				case 0x0140:
1279 					NV_WR32(par->PGRAPH, 0x0828,
1280 						0x0072cb77);
1281 					NV_WR32(par->PGRAPH, 0x082C,
1282 						0x00000108);
1283 					break;
1284 				case 0x0220:
1285 					NV_WR32(par->PGRAPH, 0x0860, 0);
1286 					NV_WR32(par->PGRAPH, 0x0864, 0);
1287 					NV_WR32(par->PRAMDAC, 0x0608,
1288 						NV_RD32(par->PRAMDAC, 0x0608) |
1289 						0x00100000);
1290 					break;
1291 				case 0x0090:
1292 				case 0x0290:
1293 				case 0x0390:
1294 					NV_WR32(par->PRAMDAC, 0x0608,
1295 						NV_RD32(par->PRAMDAC, 0x0608) |
1296 						0x00100000);
1297 					NV_WR32(par->PGRAPH, 0x0828,
1298 						0x07830610);
1299 					NV_WR32(par->PGRAPH, 0x082C,
1300 						0x0000016A);
1301 					break;
1302 				default:
1303 					break;
1304 				}
1305 
1306 				NV_WR32(par->PGRAPH, 0x0b38, 0x2ffff800);
1307 				NV_WR32(par->PGRAPH, 0x0b3c, 0x00006000);
1308 				NV_WR32(par->PGRAPH, 0x032C, 0x01000000);
1309 				NV_WR32(par->PGRAPH, 0x0220, 0x00001200);
1310 			} else if (par->Architecture == NV_ARCH_30) {
1311 				NV_WR32(par->PGRAPH, 0x0084, 0x40108700);
1312 				NV_WR32(par->PGRAPH, 0x0890, 0x00140000);
1313 				NV_WR32(par->PGRAPH, 0x008C, 0xf00e0431);
1314 				NV_WR32(par->PGRAPH, 0x0090, 0x00008000);
1315 				NV_WR32(par->PGRAPH, 0x0610, 0xf04b1f36);
1316 				NV_WR32(par->PGRAPH, 0x0B80, 0x1002d888);
1317 				NV_WR32(par->PGRAPH, 0x0B88, 0x62ff007f);
1318 			} else {
1319 				NV_WR32(par->PGRAPH, 0x0084, 0x00118700);
1320 				NV_WR32(par->PGRAPH, 0x008C, 0xF20E0431);
1321 				NV_WR32(par->PGRAPH, 0x0090, 0x00000000);
1322 				NV_WR32(par->PGRAPH, 0x009C, 0x00000040);
1323 
1324 				if ((par->Chipset & 0x0ff0) >= 0x0250) {
1325 					NV_WR32(par->PGRAPH, 0x0890,
1326 						0x00080000);
1327 					NV_WR32(par->PGRAPH, 0x0610,
1328 						0x304B1FB6);
1329 					NV_WR32(par->PGRAPH, 0x0B80,
1330 						0x18B82880);
1331 					NV_WR32(par->PGRAPH, 0x0B84,
1332 						0x44000000);
1333 					NV_WR32(par->PGRAPH, 0x0098,
1334 						0x40000080);
1335 					NV_WR32(par->PGRAPH, 0x0B88,
1336 						0x000000ff);
1337 				} else {
1338 					NV_WR32(par->PGRAPH, 0x0880,
1339 						0x00080000);
1340 					NV_WR32(par->PGRAPH, 0x0094,
1341 						0x00000005);
1342 					NV_WR32(par->PGRAPH, 0x0B80,
1343 						0x45CAA208);
1344 					NV_WR32(par->PGRAPH, 0x0B84,
1345 						0x24000000);
1346 					NV_WR32(par->PGRAPH, 0x0098,
1347 						0x00000040);
1348 					NV_WR32(par->PGRAPH, 0x0750,
1349 						0x00E00038);
1350 					NV_WR32(par->PGRAPH, 0x0754,
1351 						0x00000030);
1352 					NV_WR32(par->PGRAPH, 0x0750,
1353 						0x00E10038);
1354 					NV_WR32(par->PGRAPH, 0x0754,
1355 						0x00000030);
1356 				}
1357 			}
1358 
1359 			if ((par->Architecture < NV_ARCH_40) ||
1360 			    ((par->Chipset & 0xfff0) == 0x0040)) {
1361 				for (i = 0; i < 32; i++) {
1362 					NV_WR32(par->PGRAPH, 0x0900 + i*4,
1363 						NV_RD32(par->PFB, 0x0240 +i*4));
1364 					NV_WR32(par->PGRAPH, 0x6900 + i*4,
1365 						NV_RD32(par->PFB, 0x0240 +i*4));
1366 				}
1367 			} else {
1368 				if (((par->Chipset & 0xfff0) == 0x0090) ||
1369 				    ((par->Chipset & 0xfff0) == 0x01D0) ||
1370 				    ((par->Chipset & 0xfff0) == 0x0290) ||
1371 				    ((par->Chipset & 0xfff0) == 0x0390) ||
1372 				    ((par->Chipset & 0xfff0) == 0x03D0)) {
1373 					for (i = 0; i < 60; i++) {
1374 						NV_WR32(par->PGRAPH,
1375 							0x0D00 + i*4,
1376 							NV_RD32(par->PFB,
1377 								0x0600 + i*4));
1378 						NV_WR32(par->PGRAPH,
1379 							0x6900 + i*4,
1380 							NV_RD32(par->PFB,
1381 								0x0600 + i*4));
1382 					}
1383 				} else {
1384 					for (i = 0; i < 48; i++) {
1385 						NV_WR32(par->PGRAPH,
1386 							0x0900 + i*4,
1387 							NV_RD32(par->PFB,
1388 								0x0600 + i*4));
1389 						if(((par->Chipset & 0xfff0)
1390 						    != 0x0160) &&
1391 						   ((par->Chipset & 0xfff0)
1392 						    != 0x0220) &&
1393 						   ((par->Chipset & 0xfff0)
1394 						    != 0x240))
1395 							NV_WR32(par->PGRAPH,
1396 								0x6900 + i*4,
1397 								NV_RD32(par->PFB,
1398 									0x0600 + i*4));
1399 					}
1400 				}
1401 			}
1402 
1403 			if (par->Architecture >= NV_ARCH_40) {
1404 				if ((par->Chipset & 0xfff0) == 0x0040) {
1405 					NV_WR32(par->PGRAPH, 0x09A4,
1406 						NV_RD32(par->PFB, 0x0200));
1407 					NV_WR32(par->PGRAPH, 0x09A8,
1408 						NV_RD32(par->PFB, 0x0204));
1409 					NV_WR32(par->PGRAPH, 0x69A4,
1410 						NV_RD32(par->PFB, 0x0200));
1411 					NV_WR32(par->PGRAPH, 0x69A8,
1412 						NV_RD32(par->PFB, 0x0204));
1413 
1414 					NV_WR32(par->PGRAPH, 0x0820, 0);
1415 					NV_WR32(par->PGRAPH, 0x0824, 0);
1416 					NV_WR32(par->PGRAPH, 0x0864,
1417 						par->FbMapSize - 1);
1418 					NV_WR32(par->PGRAPH, 0x0868,
1419 						par->FbMapSize - 1);
1420 				} else {
1421 					if ((par->Chipset & 0xfff0) == 0x0090 ||
1422 					    (par->Chipset & 0xfff0) == 0x01D0 ||
1423 					    (par->Chipset & 0xfff0) == 0x0290 ||
1424 					    (par->Chipset & 0xfff0) == 0x0390) {
1425 						NV_WR32(par->PGRAPH, 0x0DF0,
1426 							NV_RD32(par->PFB, 0x0200));
1427 						NV_WR32(par->PGRAPH, 0x0DF4,
1428 							NV_RD32(par->PFB, 0x0204));
1429 					} else {
1430 						NV_WR32(par->PGRAPH, 0x09F0,
1431 							NV_RD32(par->PFB, 0x0200));
1432 						NV_WR32(par->PGRAPH, 0x09F4,
1433 							NV_RD32(par->PFB, 0x0204));
1434 					}
1435 					NV_WR32(par->PGRAPH, 0x69F0,
1436 						NV_RD32(par->PFB, 0x0200));
1437 					NV_WR32(par->PGRAPH, 0x69F4,
1438 						NV_RD32(par->PFB, 0x0204));
1439 
1440 					NV_WR32(par->PGRAPH, 0x0840, 0);
1441 					NV_WR32(par->PGRAPH, 0x0844, 0);
1442 					NV_WR32(par->PGRAPH, 0x08a0,
1443 						par->FbMapSize - 1);
1444 					NV_WR32(par->PGRAPH, 0x08a4,
1445 						par->FbMapSize - 1);
1446 				}
1447 			} else {
1448 				NV_WR32(par->PGRAPH, 0x09A4,
1449 					NV_RD32(par->PFB, 0x0200));
1450 				NV_WR32(par->PGRAPH, 0x09A8,
1451 					NV_RD32(par->PFB, 0x0204));
1452 				NV_WR32(par->PGRAPH, 0x0750, 0x00EA0000);
1453 				NV_WR32(par->PGRAPH, 0x0754,
1454 					NV_RD32(par->PFB, 0x0200));
1455 				NV_WR32(par->PGRAPH, 0x0750, 0x00EA0004);
1456 				NV_WR32(par->PGRAPH, 0x0754,
1457 					NV_RD32(par->PFB, 0x0204));
1458 
1459 				NV_WR32(par->PGRAPH, 0x0820, 0);
1460 				NV_WR32(par->PGRAPH, 0x0824, 0);
1461 				NV_WR32(par->PGRAPH, 0x0864,
1462 					par->FbMapSize - 1);
1463 				NV_WR32(par->PGRAPH, 0x0868,
1464 					par->FbMapSize - 1);
1465 			}
1466 			NV_WR32(par->PGRAPH, 0x0B20, 0x00000000);
1467 			NV_WR32(par->PGRAPH, 0x0B04, 0xFFFFFFFF);
1468 		}
1469 	}
1470 	NV_WR32(par->PGRAPH, 0x053C, 0);
1471 	NV_WR32(par->PGRAPH, 0x0540, 0);
1472 	NV_WR32(par->PGRAPH, 0x0544, 0x00007FFF);
1473 	NV_WR32(par->PGRAPH, 0x0548, 0x00007FFF);
1474 
1475 	NV_WR32(par->PFIFO, 0x0140 * 4, 0x00000000);
1476 	NV_WR32(par->PFIFO, 0x0141 * 4, 0x00000001);
1477 	NV_WR32(par->PFIFO, 0x0480 * 4, 0x00000000);
1478 	NV_WR32(par->PFIFO, 0x0494 * 4, 0x00000000);
1479 	if (par->Architecture >= NV_ARCH_40)
1480 		NV_WR32(par->PFIFO, 0x0481 * 4, 0x00010000);
1481 	else
1482 		NV_WR32(par->PFIFO, 0x0481 * 4, 0x00000100);
1483 	NV_WR32(par->PFIFO, 0x0490 * 4, 0x00000000);
1484 	NV_WR32(par->PFIFO, 0x0491 * 4, 0x00000000);
1485 	if (par->Architecture >= NV_ARCH_40)
1486 		NV_WR32(par->PFIFO, 0x048B * 4, 0x00001213);
1487 	else
1488 		NV_WR32(par->PFIFO, 0x048B * 4, 0x00001209);
1489 	NV_WR32(par->PFIFO, 0x0400 * 4, 0x00000000);
1490 	NV_WR32(par->PFIFO, 0x0414 * 4, 0x00000000);
1491 	NV_WR32(par->PFIFO, 0x0084 * 4, 0x03000100);
1492 	NV_WR32(par->PFIFO, 0x0085 * 4, 0x00000110);
1493 	NV_WR32(par->PFIFO, 0x0086 * 4, 0x00000112);
1494 	NV_WR32(par->PFIFO, 0x0143 * 4, 0x0000FFFF);
1495 	NV_WR32(par->PFIFO, 0x0496 * 4, 0x0000FFFF);
1496 	NV_WR32(par->PFIFO, 0x0050 * 4, 0x00000000);
1497 	NV_WR32(par->PFIFO, 0x0040 * 4, 0xFFFFFFFF);
1498 	NV_WR32(par->PFIFO, 0x0415 * 4, 0x00000001);
1499 	NV_WR32(par->PFIFO, 0x048C * 4, 0x00000000);
1500 	NV_WR32(par->PFIFO, 0x04A0 * 4, 0x00000000);
1501 #ifdef __BIG_ENDIAN
1502 	NV_WR32(par->PFIFO, 0x0489 * 4, 0x800F0078);
1503 #else
1504 	NV_WR32(par->PFIFO, 0x0489 * 4, 0x000F0078);
1505 #endif
1506 	NV_WR32(par->PFIFO, 0x0488 * 4, 0x00000001);
1507 	NV_WR32(par->PFIFO, 0x0480 * 4, 0x00000001);
1508 	NV_WR32(par->PFIFO, 0x0494 * 4, 0x00000001);
1509 	NV_WR32(par->PFIFO, 0x0495 * 4, 0x00000001);
1510 	NV_WR32(par->PFIFO, 0x0140 * 4, 0x00000001);
1511 
1512     if (!state) {
1513 	    par->CurrentState = NULL;
1514 	    return;
1515     }
1516 
1517 	if (par->Architecture >= NV_ARCH_10) {
1518 		if (par->twoHeads) {
1519 			NV_WR32(par->PCRTC0, 0x0860, state->head);
1520 			NV_WR32(par->PCRTC0, 0x2860, state->head2);
1521 		}
1522 		NV_WR32(par->PRAMDAC, 0x0404, NV_RD32(par->PRAMDAC, 0x0404) |
1523 			(1 << 25));
1524 
1525 		NV_WR32(par->PMC, 0x8704, 1);
1526 		NV_WR32(par->PMC, 0x8140, 0);
1527 		NV_WR32(par->PMC, 0x8920, 0);
1528 		NV_WR32(par->PMC, 0x8924, 0);
1529 		NV_WR32(par->PMC, 0x8908, par->FbMapSize - 1);
1530 		NV_WR32(par->PMC, 0x890C, par->FbMapSize - 1);
1531 		NV_WR32(par->PMC, 0x1588, 0);
1532 
1533 		NV_WR32(par->PCRTC, 0x0810, state->cursorConfig);
1534 		NV_WR32(par->PCRTC, 0x0830, state->displayV - 3);
1535 		NV_WR32(par->PCRTC, 0x0834, state->displayV - 1);
1536 
1537 		if (par->FlatPanel) {
1538 			if ((par->Chipset & 0x0ff0) == 0x0110) {
1539 				NV_WR32(par->PRAMDAC, 0x0528, state->dither);
1540 			} else if (par->twoHeads) {
1541 				NV_WR32(par->PRAMDAC, 0x083C, state->dither);
1542 			}
1543 
1544 			VGA_WR08(par->PCIO, 0x03D4, 0x53);
1545 			VGA_WR08(par->PCIO, 0x03D5, state->timingH);
1546 			VGA_WR08(par->PCIO, 0x03D4, 0x54);
1547 			VGA_WR08(par->PCIO, 0x03D5, state->timingV);
1548 			VGA_WR08(par->PCIO, 0x03D4, 0x21);
1549 			VGA_WR08(par->PCIO, 0x03D5, 0xfa);
1550 		}
1551 
1552 		VGA_WR08(par->PCIO, 0x03D4, 0x41);
1553 		VGA_WR08(par->PCIO, 0x03D5, state->extra);
1554 	}
1555 
1556 	VGA_WR08(par->PCIO, 0x03D4, 0x19);
1557 	VGA_WR08(par->PCIO, 0x03D5, state->repaint0);
1558 	VGA_WR08(par->PCIO, 0x03D4, 0x1A);
1559 	VGA_WR08(par->PCIO, 0x03D5, state->repaint1);
1560 	VGA_WR08(par->PCIO, 0x03D4, 0x25);
1561 	VGA_WR08(par->PCIO, 0x03D5, state->screen);
1562 	VGA_WR08(par->PCIO, 0x03D4, 0x28);
1563 	VGA_WR08(par->PCIO, 0x03D5, state->pixel);
1564 	VGA_WR08(par->PCIO, 0x03D4, 0x2D);
1565 	VGA_WR08(par->PCIO, 0x03D5, state->horiz);
1566 	VGA_WR08(par->PCIO, 0x03D4, 0x1C);
1567 	VGA_WR08(par->PCIO, 0x03D5, state->fifo);
1568 	VGA_WR08(par->PCIO, 0x03D4, 0x1B);
1569 	VGA_WR08(par->PCIO, 0x03D5, state->arbitration0);
1570 	VGA_WR08(par->PCIO, 0x03D4, 0x20);
1571 	VGA_WR08(par->PCIO, 0x03D5, state->arbitration1);
1572 
1573 	if(par->Architecture >= NV_ARCH_30) {
1574 		VGA_WR08(par->PCIO, 0x03D4, 0x47);
1575 		VGA_WR08(par->PCIO, 0x03D5, state->arbitration1 >> 8);
1576 	}
1577 
1578 	VGA_WR08(par->PCIO, 0x03D4, 0x30);
1579 	VGA_WR08(par->PCIO, 0x03D5, state->cursor0);
1580 	VGA_WR08(par->PCIO, 0x03D4, 0x31);
1581 	VGA_WR08(par->PCIO, 0x03D5, state->cursor1);
1582 	VGA_WR08(par->PCIO, 0x03D4, 0x2F);
1583 	VGA_WR08(par->PCIO, 0x03D5, state->cursor2);
1584 	VGA_WR08(par->PCIO, 0x03D4, 0x39);
1585 	VGA_WR08(par->PCIO, 0x03D5, state->interlace);
1586 
1587 	if (!par->FlatPanel) {
1588 		if (par->Architecture >= NV_ARCH_40)
1589 			NV_WR32(par->PRAMDAC0, 0x0580, state->control);
1590 
1591 		NV_WR32(par->PRAMDAC0, 0x050C, state->pllsel);
1592 		NV_WR32(par->PRAMDAC0, 0x0508, state->vpll);
1593 		if (par->twoHeads)
1594 			NV_WR32(par->PRAMDAC0, 0x0520, state->vpll2);
1595 		if (par->twoStagePLL) {
1596 			NV_WR32(par->PRAMDAC0, 0x0578, state->vpllB);
1597 			NV_WR32(par->PRAMDAC0, 0x057C, state->vpll2B);
1598 		}
1599 	} else {
1600 		NV_WR32(par->PRAMDAC, 0x0848, state->scale);
1601 		NV_WR32(par->PRAMDAC, 0x0828, state->crtcSync +
1602 			par->PanelTweak);
1603 	}
1604 
1605 	NV_WR32(par->PRAMDAC, 0x0600, state->general);
1606 
1607 	NV_WR32(par->PCRTC, 0x0140, 0);
1608 	NV_WR32(par->PCRTC, 0x0100, 1);
1609 
1610 	par->CurrentState = state;
1611 }
1612 
1613 void NVUnloadStateExt(struct nvidia_par *par, RIVA_HW_STATE * state) {
1614 	VGA_WR08(par->PCIO, 0x03D4, 0x19);
1615 	state->repaint0 = VGA_RD08(par->PCIO, 0x03D5);
1616 	VGA_WR08(par->PCIO, 0x03D4, 0x1A);
1617 	state->repaint1 = VGA_RD08(par->PCIO, 0x03D5);
1618 	VGA_WR08(par->PCIO, 0x03D4, 0x25);
1619 	state->screen = VGA_RD08(par->PCIO, 0x03D5);
1620 	VGA_WR08(par->PCIO, 0x03D4, 0x28);
1621 	state->pixel = VGA_RD08(par->PCIO, 0x03D5);
1622 	VGA_WR08(par->PCIO, 0x03D4, 0x2D);
1623 	state->horiz = VGA_RD08(par->PCIO, 0x03D5);
1624 	VGA_WR08(par->PCIO, 0x03D4, 0x1C);
1625 	state->fifo         = VGA_RD08(par->PCIO, 0x03D5);
1626 	VGA_WR08(par->PCIO, 0x03D4, 0x1B);
1627 	state->arbitration0 = VGA_RD08(par->PCIO, 0x03D5);
1628 	VGA_WR08(par->PCIO, 0x03D4, 0x20);
1629 	state->arbitration1 = VGA_RD08(par->PCIO, 0x03D5);
1630 
1631 	if(par->Architecture >= NV_ARCH_30) {
1632 		VGA_WR08(par->PCIO, 0x03D4, 0x47);
1633 		state->arbitration1 |= (VGA_RD08(par->PCIO, 0x03D5) & 1) << 8;
1634 	}
1635 
1636 	VGA_WR08(par->PCIO, 0x03D4, 0x30);
1637 	state->cursor0 = VGA_RD08(par->PCIO, 0x03D5);
1638 	VGA_WR08(par->PCIO, 0x03D4, 0x31);
1639 	state->cursor1 = VGA_RD08(par->PCIO, 0x03D5);
1640 	VGA_WR08(par->PCIO, 0x03D4, 0x2F);
1641 	state->cursor2 = VGA_RD08(par->PCIO, 0x03D5);
1642 	VGA_WR08(par->PCIO, 0x03D4, 0x39);
1643 	state->interlace = VGA_RD08(par->PCIO, 0x03D5);
1644 	state->vpll = NV_RD32(par->PRAMDAC0, 0x0508);
1645 	if (par->twoHeads)
1646 		state->vpll2 = NV_RD32(par->PRAMDAC0, 0x0520);
1647 	if (par->twoStagePLL) {
1648 		state->vpllB = NV_RD32(par->PRAMDAC0, 0x0578);
1649 		state->vpll2B = NV_RD32(par->PRAMDAC0, 0x057C);
1650 	}
1651 	state->pllsel = NV_RD32(par->PRAMDAC0, 0x050C);
1652 	state->general = NV_RD32(par->PRAMDAC, 0x0600);
1653 	state->scale = NV_RD32(par->PRAMDAC, 0x0848);
1654 	state->config = NV_RD32(par->PFB, 0x0200);
1655 
1656 	if (par->Architecture >= NV_ARCH_40 && !par->FlatPanel)
1657 		state->control  = NV_RD32(par->PRAMDAC0, 0x0580);
1658 
1659 	if (par->Architecture >= NV_ARCH_10) {
1660 		if (par->twoHeads) {
1661 			state->head = NV_RD32(par->PCRTC0, 0x0860);
1662 			state->head2 = NV_RD32(par->PCRTC0, 0x2860);
1663 			VGA_WR08(par->PCIO, 0x03D4, 0x44);
1664 			state->crtcOwner = VGA_RD08(par->PCIO, 0x03D5);
1665 		}
1666 		VGA_WR08(par->PCIO, 0x03D4, 0x41);
1667 		state->extra = VGA_RD08(par->PCIO, 0x03D5);
1668 		state->cursorConfig = NV_RD32(par->PCRTC, 0x0810);
1669 
1670 		if ((par->Chipset & 0x0ff0) == 0x0110) {
1671 			state->dither = NV_RD32(par->PRAMDAC, 0x0528);
1672 		} else if (par->twoHeads) {
1673 			state->dither = NV_RD32(par->PRAMDAC, 0x083C);
1674 		}
1675 
1676 		if (par->FlatPanel) {
1677 			VGA_WR08(par->PCIO, 0x03D4, 0x53);
1678 			state->timingH = VGA_RD08(par->PCIO, 0x03D5);
1679 			VGA_WR08(par->PCIO, 0x03D4, 0x54);
1680 			state->timingV = VGA_RD08(par->PCIO, 0x03D5);
1681 		}
1682 	}
1683 }
1684 
1685 void NVSetStartAddress(struct nvidia_par *par, u32 start)
1686 {
1687 	NV_WR32(par->PCRTC, 0x800, start);
1688 }
1689