Commit | Line | Data |
---|---|---|
fade7ad5 BS |
1 | /* |
2 | * Copyright 2010 Red Hat Inc. | |
3 | * | |
4 | * Permission is hereby granted, free of charge, to any person obtaining a | |
5 | * copy of this software and associated documentation files (the "Software"), | |
6 | * to deal in the Software without restriction, including without limitation | |
7 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, | |
8 | * and/or sell copies of the Software, and to permit persons to whom the | |
9 | * Software is furnished to do so, subject to the following conditions: | |
10 | * | |
11 | * The above copyright notice and this permission notice shall be included in | |
12 | * all copies or substantial portions of the Software. | |
13 | * | |
14 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
15 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
16 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
17 | * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR | |
18 | * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, | |
19 | * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR | |
20 | * OTHER DEALINGS IN THE SOFTWARE. | |
21 | * | |
22 | * Authors: Ben Skeggs | |
23 | */ | |
24 | ||
25 | #include "drmP.h" | |
26 | #include "nouveau_drv.h" | |
27 | #include "nouveau_bios.h" | |
28 | #include "nouveau_pm.h" | |
29 | ||
ca94a71f | 30 | static u32 read_clk(struct drm_device *, int, bool); |
cec2a270 | 31 | static u32 read_pll(struct drm_device *, int, u32); |
3b0582d3 BS |
32 | |
33 | static u32 | |
ca94a71f BS |
34 | read_vco(struct drm_device *dev, int clk) |
35 | { | |
36 | u32 sctl = nv_rd32(dev, 0x4120 + (clk * 4)); | |
37 | if ((sctl & 0x00000030) != 0x00000030) | |
cec2a270 BS |
38 | return read_pll(dev, 0x41, 0x00e820); |
39 | return read_pll(dev, 0x42, 0x00e8a0); | |
ca94a71f BS |
40 | } |
41 | ||
42 | static u32 | |
43 | read_clk(struct drm_device *dev, int clk, bool ignore_en) | |
3b0582d3 | 44 | { |
64e740bb | 45 | struct drm_nouveau_private *dev_priv = dev->dev_private; |
3b0582d3 BS |
46 | u32 sctl, sdiv, sclk; |
47 | ||
64e740bb | 48 | /* refclk for the 0xe8xx plls is a fixed frequency */ |
3b0582d3 | 49 | if (clk >= 0x40) |
64e740bb | 50 | return dev_priv->crystal; |
3b0582d3 BS |
51 | |
52 | sctl = nv_rd32(dev, 0x4120 + (clk * 4)); | |
ca94a71f BS |
53 | if (!ignore_en && !(sctl & 0x00000100)) |
54 | return 0; | |
55 | ||
56 | switch (sctl & 0x00003000) { | |
57 | case 0x00000000: | |
64e740bb | 58 | return dev_priv->crystal; |
ca94a71f | 59 | case 0x00002000: |
3b0582d3 BS |
60 | if (sctl & 0x00000040) |
61 | return 108000; | |
62 | return 100000; | |
ca94a71f BS |
63 | case 0x00003000: |
64 | sclk = read_vco(dev, clk); | |
3b0582d3 | 65 | sdiv = ((sctl & 0x003f0000) >> 16) + 2; |
3b0582d3 BS |
66 | return (sclk * 2) / sdiv; |
67 | default: | |
68 | return 0; | |
69 | } | |
70 | } | |
71 | ||
72 | static u32 | |
cec2a270 | 73 | read_pll(struct drm_device *dev, int clk, u32 pll) |
3b0582d3 BS |
74 | { |
75 | u32 ctrl = nv_rd32(dev, pll + 0); | |
93e692dc | 76 | u32 sclk = 0, P = 1, N = 1, M = 1; |
3b0582d3 BS |
77 | |
78 | if (!(ctrl & 0x00000008)) { | |
93e692dc BS |
79 | if (ctrl & 0x00000001) { |
80 | u32 coef = nv_rd32(dev, pll + 4); | |
81 | M = (coef & 0x000000ff) >> 0; | |
82 | N = (coef & 0x0000ff00) >> 8; | |
83 | P = (coef & 0x003f0000) >> 16; | |
cec2a270 | 84 | |
93e692dc BS |
85 | /* no post-divider on these.. */ |
86 | if ((pll & 0x00ff00) == 0x00e800) | |
87 | P = 1; | |
3b0582d3 | 88 | |
93e692dc BS |
89 | sclk = read_clk(dev, 0x00 + clk, false); |
90 | } | |
3b0582d3 | 91 | } else { |
ca94a71f | 92 | sclk = read_clk(dev, 0x10 + clk, false); |
3b0582d3 BS |
93 | } |
94 | ||
95 | return sclk * N / (M * P); | |
96 | } | |
fade7ad5 | 97 | |
ca94a71f BS |
98 | struct creg { |
99 | u32 clk; | |
100 | u32 pll; | |
fade7ad5 BS |
101 | }; |
102 | ||
215f902e | 103 | static int |
cec2a270 | 104 | calc_clk(struct drm_device *dev, int clk, u32 pll, u32 khz, struct creg *reg) |
215f902e | 105 | { |
ca94a71f BS |
106 | struct pll_lims limits; |
107 | u32 oclk, sclk, sdiv; | |
108 | int P, N, M, diff; | |
109 | int ret; | |
110 | ||
111 | reg->pll = 0; | |
112 | reg->clk = 0; | |
cec2a270 BS |
113 | if (!khz) { |
114 | NV_DEBUG(dev, "no clock for 0x%04x/0x%02x\n", pll, clk); | |
115 | return 0; | |
116 | } | |
ca94a71f BS |
117 | |
118 | switch (khz) { | |
119 | case 27000: | |
120 | reg->clk = 0x00000100; | |
121 | return khz; | |
122 | case 100000: | |
123 | reg->clk = 0x00002100; | |
124 | return khz; | |
125 | case 108000: | |
126 | reg->clk = 0x00002140; | |
127 | return khz; | |
128 | default: | |
129 | sclk = read_vco(dev, clk); | |
130 | sdiv = min((sclk * 2) / (khz - 2999), (u32)65); | |
cec2a270 BS |
131 | /* if the clock has a PLL attached, and we can get a within |
132 | * [-2, 3) MHz of a divider, we'll disable the PLL and use | |
133 | * the divider instead. | |
134 | * | |
135 | * divider can go as low as 2, limited here because NVIDIA | |
136 | * and the VBIOS on my NVA8 seem to prefer using the PLL | |
137 | * for 810MHz - is there a good reason? | |
138 | */ | |
ca94a71f BS |
139 | if (sdiv > 4) { |
140 | oclk = (sclk * 2) / sdiv; | |
141 | diff = khz - oclk; | |
142 | if (!pll || (diff >= -2000 && diff < 3000)) { | |
143 | reg->clk = (((sdiv - 2) << 16) | 0x00003100); | |
144 | return oclk; | |
145 | } | |
146 | } | |
cec2a270 BS |
147 | |
148 | if (!pll) { | |
149 | NV_ERROR(dev, "bad freq %02x: %d %d\n", clk, khz, sclk); | |
150 | return -ERANGE; | |
151 | } | |
152 | ||
ca94a71f | 153 | break; |
215f902e BS |
154 | } |
155 | ||
ca94a71f BS |
156 | ret = get_pll_limits(dev, pll, &limits); |
157 | if (ret) | |
158 | return ret; | |
159 | ||
160 | limits.refclk = read_clk(dev, clk - 0x10, true); | |
161 | if (!limits.refclk) | |
162 | return -EINVAL; | |
163 | ||
164 | ret = nva3_calc_pll(dev, &limits, khz, &N, NULL, &M, &P); | |
165 | if (ret >= 0) { | |
166 | reg->clk = nv_rd32(dev, 0x4120 + (clk * 4)); | |
167 | reg->pll = (P << 16) | (N << 8) | M; | |
168 | } | |
169 | return ret; | |
215f902e BS |
170 | } |
171 | ||
cec2a270 BS |
172 | static void |
173 | prog_pll(struct drm_device *dev, int clk, u32 pll, struct creg *reg) | |
174 | { | |
175 | const u32 src0 = 0x004120 + (clk * 4); | |
176 | const u32 src1 = 0x004160 + (clk * 4); | |
177 | const u32 ctrl = pll + 0; | |
178 | const u32 coef = pll + 4; | |
179 | u32 cntl; | |
180 | ||
181 | if (!reg->clk && !reg->pll) { | |
182 | NV_DEBUG(dev, "no clock for %02x\n", clk); | |
183 | return; | |
184 | } | |
185 | ||
186 | cntl = nv_rd32(dev, ctrl) & 0xfffffff2; | |
187 | if (reg->pll) { | |
188 | nv_mask(dev, src0, 0x00000101, 0x00000101); | |
189 | nv_wr32(dev, coef, reg->pll); | |
190 | nv_wr32(dev, ctrl, cntl | 0x00000015); | |
191 | nv_mask(dev, src1, 0x00000100, 0x00000000); | |
192 | nv_mask(dev, src1, 0x00000001, 0x00000000); | |
193 | } else { | |
194 | nv_mask(dev, src1, 0x003f3141, 0x00000101 | reg->clk); | |
195 | nv_wr32(dev, ctrl, cntl | 0x0000001d); | |
196 | nv_mask(dev, ctrl, 0x00000001, 0x00000000); | |
197 | nv_mask(dev, src0, 0x00000100, 0x00000000); | |
198 | nv_mask(dev, src0, 0x00000001, 0x00000000); | |
199 | } | |
200 | } | |
201 | ||
202 | static void | |
203 | prog_clk(struct drm_device *dev, int clk, struct creg *reg) | |
204 | { | |
205 | if (!reg->clk) { | |
206 | NV_DEBUG(dev, "no clock for %02x\n", clk); | |
207 | return; | |
208 | } | |
209 | ||
210 | nv_mask(dev, 0x004120 + (clk * 4), 0x003f3141, 0x00000101 | reg->clk); | |
211 | } | |
212 | ||
fade7ad5 | 213 | int |
ca94a71f | 214 | nva3_pm_clocks_get(struct drm_device *dev, struct nouveau_pm_level *perflvl) |
fade7ad5 | 215 | { |
cec2a270 BS |
216 | perflvl->core = read_pll(dev, 0x00, 0x4200); |
217 | perflvl->shader = read_pll(dev, 0x01, 0x4220); | |
218 | perflvl->memory = read_pll(dev, 0x02, 0x4000); | |
4fd2847e BS |
219 | perflvl->unka0 = read_clk(dev, 0x20, false); |
220 | perflvl->vdec = read_clk(dev, 0x21, false); | |
9698b9a6 BS |
221 | perflvl->daemon = read_clk(dev, 0x25, false); |
222 | perflvl->copy = perflvl->core; | |
ca94a71f | 223 | return 0; |
fade7ad5 BS |
224 | } |
225 | ||
ca94a71f BS |
226 | struct nva3_pm_state { |
227 | struct creg nclk; | |
228 | struct creg sclk; | |
229 | struct creg mclk; | |
4fd2847e BS |
230 | struct creg vdec; |
231 | struct creg unka0; | |
ca94a71f BS |
232 | }; |
233 | ||
fade7ad5 | 234 | void * |
ca94a71f | 235 | nva3_pm_clocks_pre(struct drm_device *dev, struct nouveau_pm_level *perflvl) |
fade7ad5 | 236 | { |
ca94a71f BS |
237 | struct nva3_pm_state *info; |
238 | int ret; | |
fade7ad5 | 239 | |
ca94a71f BS |
240 | info = kzalloc(sizeof(*info), GFP_KERNEL); |
241 | if (!info) | |
242 | return ERR_PTR(-ENOMEM); | |
243 | ||
cec2a270 | 244 | ret = calc_clk(dev, 0x10, 0x4200, perflvl->core, &info->nclk); |
dac55b58 | 245 | if (ret < 0) |
ca94a71f | 246 | goto out; |
dac55b58 | 247 | |
cec2a270 | 248 | ret = calc_clk(dev, 0x11, 0x4220, perflvl->shader, &info->sclk); |
ca94a71f BS |
249 | if (ret < 0) |
250 | goto out; | |
dac55b58 | 251 | |
cec2a270 | 252 | ret = calc_clk(dev, 0x12, 0x4000, perflvl->memory, &info->mclk); |
ca94a71f BS |
253 | if (ret < 0) |
254 | goto out; | |
dac55b58 | 255 | |
cec2a270 | 256 | ret = calc_clk(dev, 0x20, 0x0000, perflvl->unka0, &info->unka0); |
4fd2847e BS |
257 | if (ret < 0) |
258 | goto out; | |
259 | ||
cec2a270 | 260 | ret = calc_clk(dev, 0x21, 0x0000, perflvl->vdec, &info->vdec); |
4fd2847e BS |
261 | if (ret < 0) |
262 | goto out; | |
263 | ||
ca94a71f BS |
264 | out: |
265 | if (ret < 0) { | |
266 | kfree(info); | |
267 | info = ERR_PTR(ret); | |
fade7ad5 | 268 | } |
ca94a71f BS |
269 | return info; |
270 | } | |
fade7ad5 | 271 | |
d0f67a48 BS |
272 | static bool |
273 | nva3_pm_grcp_idle(void *data) | |
274 | { | |
275 | struct drm_device *dev = data; | |
276 | ||
277 | if (!(nv_rd32(dev, 0x400304) & 0x00000001)) | |
278 | return true; | |
279 | if (nv_rd32(dev, 0x400308) == 0x0050001c) | |
280 | return true; | |
281 | return false; | |
282 | } | |
283 | ||
fade7ad5 | 284 | void |
ca94a71f | 285 | nva3_pm_clocks_set(struct drm_device *dev, void *pre_state) |
fade7ad5 | 286 | { |
d0f67a48 | 287 | struct drm_nouveau_private *dev_priv = dev->dev_private; |
ca94a71f | 288 | struct nva3_pm_state *info = pre_state; |
d0f67a48 BS |
289 | unsigned long flags; |
290 | ||
291 | /* prevent any new grctx switches from starting */ | |
292 | spin_lock_irqsave(&dev_priv->context_switch_lock, flags); | |
293 | nv_wr32(dev, 0x400324, 0x00000000); | |
294 | nv_wr32(dev, 0x400328, 0x0050001c); /* wait flag 0x1c */ | |
295 | /* wait for any pending grctx switches to complete */ | |
296 | if (!nv_wait_cb(dev, nva3_pm_grcp_idle, dev)) { | |
297 | NV_ERROR(dev, "pm: ctxprog didn't go idle\n"); | |
298 | goto cleanup; | |
299 | } | |
300 | /* freeze PFIFO */ | |
301 | nv_mask(dev, 0x002504, 0x00000001, 0x00000001); | |
302 | if (!nv_wait(dev, 0x002504, 0x00000010, 0x00000010)) { | |
303 | NV_ERROR(dev, "pm: fifo didn't go idle\n"); | |
304 | goto cleanup; | |
305 | } | |
ca94a71f | 306 | |
cec2a270 BS |
307 | prog_pll(dev, 0x00, 0x004200, &info->nclk); |
308 | prog_pll(dev, 0x01, 0x004220, &info->sclk); | |
4fd2847e BS |
309 | prog_clk(dev, 0x20, &info->unka0); |
310 | prog_clk(dev, 0x21, &info->vdec); | |
ca94a71f | 311 | |
93e692dc BS |
312 | if (info->mclk.clk || info->mclk.pll) { |
313 | nv_wr32(dev, 0x100210, 0); | |
314 | nv_wr32(dev, 0x1002dc, 1); | |
315 | nv_wr32(dev, 0x004018, 0x00001000); | |
316 | prog_pll(dev, 0x02, 0x004000, &info->mclk); | |
317 | if (nv_rd32(dev, 0x4000) & 0x00000008) | |
318 | nv_wr32(dev, 0x004018, 0x1000d000); | |
319 | else | |
320 | nv_wr32(dev, 0x004018, 0x10005000); | |
321 | nv_wr32(dev, 0x1002dc, 0); | |
322 | nv_wr32(dev, 0x100210, 0x80000000); | |
323 | } | |
ca94a71f | 324 | |
d0f67a48 BS |
325 | cleanup: |
326 | /* unfreeze PFIFO */ | |
327 | nv_mask(dev, 0x002504, 0x00000001, 0x00000000); | |
328 | /* restore ctxprog to normal */ | |
329 | nv_wr32(dev, 0x400324, 0x00000000); | |
330 | nv_wr32(dev, 0x400328, 0x0070009c); /* set flag 0x1c */ | |
331 | /* unblock it if necessary */ | |
332 | if (nv_rd32(dev, 0x400308) == 0x0050001c) | |
333 | nv_mask(dev, 0x400824, 0x10000000, 0x10000000); | |
334 | spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); | |
ca94a71f | 335 | kfree(info); |
fade7ad5 | 336 | } |