File: | dev/pci/drm/radeon/r600.c |
Warning: | line 1794, column 3 Value stored to 'tmp' is never read |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
1 | /* |
2 | * Copyright 2008 Advanced Micro Devices, Inc. |
3 | * Copyright 2008 Red Hat Inc. |
4 | * Copyright 2009 Jerome Glisse. |
5 | * |
6 | * Permission is hereby granted, free of charge, to any person obtaining a |
7 | * copy of this software and associated documentation files (the "Software"), |
8 | * to deal in the Software without restriction, including without limitation |
9 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
10 | * and/or sell copies of the Software, and to permit persons to whom the |
11 | * Software is furnished to do so, subject to the following conditions: |
12 | * |
13 | * The above copyright notice and this permission notice shall be included in |
14 | * all copies or substantial portions of the Software. |
15 | * |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
19 | * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR |
20 | * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
21 | * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
22 | * OTHER DEALINGS IN THE SOFTWARE. |
23 | * |
24 | * Authors: Dave Airlie |
25 | * Alex Deucher |
26 | * Jerome Glisse |
27 | */ |
28 | |
29 | #include <linux/firmware.h> |
30 | #include <linux/module.h> |
31 | #include <linux/pci.h> |
32 | #include <linux/slab.h> |
33 | #include <linux/seq_file.h> |
34 | |
35 | #include <drm/drm_device.h> |
36 | #include <drm/drm_vblank.h> |
37 | #include <drm/radeon_drm.h> |
38 | |
39 | #include "atom.h" |
40 | #include "avivod.h" |
41 | #include "evergreen.h" |
42 | #include "r600.h" |
43 | #include "r600d.h" |
44 | #include "rv770.h" |
45 | #include "radeon.h" |
46 | #include "radeon_asic.h" |
47 | #include "radeon_audio.h" |
48 | #include "radeon_mode.h" |
49 | #include "radeon_ucode.h" |
50 | |
51 | /* Firmware Names */ |
52 | MODULE_FIRMWARE("radeon/R600_pfp.bin"); |
53 | MODULE_FIRMWARE("radeon/R600_me.bin"); |
54 | MODULE_FIRMWARE("radeon/RV610_pfp.bin"); |
55 | MODULE_FIRMWARE("radeon/RV610_me.bin"); |
56 | MODULE_FIRMWARE("radeon/RV630_pfp.bin"); |
57 | MODULE_FIRMWARE("radeon/RV630_me.bin"); |
58 | MODULE_FIRMWARE("radeon/RV620_pfp.bin"); |
59 | MODULE_FIRMWARE("radeon/RV620_me.bin"); |
60 | MODULE_FIRMWARE("radeon/RV635_pfp.bin"); |
61 | MODULE_FIRMWARE("radeon/RV635_me.bin"); |
62 | MODULE_FIRMWARE("radeon/RV670_pfp.bin"); |
63 | MODULE_FIRMWARE("radeon/RV670_me.bin"); |
64 | MODULE_FIRMWARE("radeon/RS780_pfp.bin"); |
65 | MODULE_FIRMWARE("radeon/RS780_me.bin"); |
66 | MODULE_FIRMWARE("radeon/RV770_pfp.bin"); |
67 | MODULE_FIRMWARE("radeon/RV770_me.bin"); |
68 | MODULE_FIRMWARE("radeon/RV770_smc.bin"); |
69 | MODULE_FIRMWARE("radeon/RV730_pfp.bin"); |
70 | MODULE_FIRMWARE("radeon/RV730_me.bin"); |
71 | MODULE_FIRMWARE("radeon/RV730_smc.bin"); |
72 | MODULE_FIRMWARE("radeon/RV740_smc.bin"); |
73 | MODULE_FIRMWARE("radeon/RV710_pfp.bin"); |
74 | MODULE_FIRMWARE("radeon/RV710_me.bin"); |
75 | MODULE_FIRMWARE("radeon/RV710_smc.bin"); |
76 | MODULE_FIRMWARE("radeon/R600_rlc.bin"); |
77 | MODULE_FIRMWARE("radeon/R700_rlc.bin"); |
78 | MODULE_FIRMWARE("radeon/CEDAR_pfp.bin"); |
79 | MODULE_FIRMWARE("radeon/CEDAR_me.bin"); |
80 | MODULE_FIRMWARE("radeon/CEDAR_rlc.bin"); |
81 | MODULE_FIRMWARE("radeon/CEDAR_smc.bin"); |
82 | MODULE_FIRMWARE("radeon/REDWOOD_pfp.bin"); |
83 | MODULE_FIRMWARE("radeon/REDWOOD_me.bin"); |
84 | MODULE_FIRMWARE("radeon/REDWOOD_rlc.bin"); |
85 | MODULE_FIRMWARE("radeon/REDWOOD_smc.bin"); |
86 | MODULE_FIRMWARE("radeon/JUNIPER_pfp.bin"); |
87 | MODULE_FIRMWARE("radeon/JUNIPER_me.bin"); |
88 | MODULE_FIRMWARE("radeon/JUNIPER_rlc.bin"); |
89 | MODULE_FIRMWARE("radeon/JUNIPER_smc.bin"); |
90 | MODULE_FIRMWARE("radeon/CYPRESS_pfp.bin"); |
91 | MODULE_FIRMWARE("radeon/CYPRESS_me.bin"); |
92 | MODULE_FIRMWARE("radeon/CYPRESS_rlc.bin"); |
93 | MODULE_FIRMWARE("radeon/CYPRESS_smc.bin"); |
94 | MODULE_FIRMWARE("radeon/PALM_pfp.bin"); |
95 | MODULE_FIRMWARE("radeon/PALM_me.bin"); |
96 | MODULE_FIRMWARE("radeon/SUMO_rlc.bin"); |
97 | MODULE_FIRMWARE("radeon/SUMO_pfp.bin"); |
98 | MODULE_FIRMWARE("radeon/SUMO_me.bin"); |
99 | MODULE_FIRMWARE("radeon/SUMO2_pfp.bin"); |
100 | MODULE_FIRMWARE("radeon/SUMO2_me.bin"); |
101 | |
102 | static const u32 crtc_offsets[2] = |
103 | { |
104 | 0, |
105 | AVIVO_D2CRTC_H_TOTAL0x6800 - AVIVO_D1CRTC_H_TOTAL0x6000 |
106 | }; |
107 | |
108 | static void r600_debugfs_mc_info_init(struct radeon_device *rdev); |
109 | |
110 | /* r600,rv610,rv630,rv620,rv635,rv670 */ |
111 | int r600_mc_wait_for_idle(struct radeon_device *rdev); |
112 | static void r600_gpu_init(struct radeon_device *rdev); |
113 | void r600_fini(struct radeon_device *rdev); |
114 | void r600_irq_disable(struct radeon_device *rdev); |
115 | static void r600_pcie_gen2_enable(struct radeon_device *rdev); |
116 | |
117 | /* |
118 | * Indirect registers accessor |
119 | */ |
120 | u32 r600_rcu_rreg(struct radeon_device *rdev, u32 reg) |
121 | { |
122 | unsigned long flags; |
123 | u32 r; |
124 | |
125 | spin_lock_irqsave(&rdev->rcu_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->rcu_idx_lock); } while (0); |
126 | WREG32(R600_RCU_INDEX, ((reg) & 0x1fff))r100_mm_wreg(rdev, (0x0100), (((reg) & 0x1fff)), 0); |
127 | r = RREG32(R600_RCU_DATA)r100_mm_rreg(rdev, (0x0104), 0); |
128 | spin_unlock_irqrestore(&rdev->rcu_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->rcu_idx_lock); } while (0); |
129 | return r; |
130 | } |
131 | |
132 | void r600_rcu_wreg(struct radeon_device *rdev, u32 reg, u32 v) |
133 | { |
134 | unsigned long flags; |
135 | |
136 | spin_lock_irqsave(&rdev->rcu_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->rcu_idx_lock); } while (0); |
137 | WREG32(R600_RCU_INDEX, ((reg) & 0x1fff))r100_mm_wreg(rdev, (0x0100), (((reg) & 0x1fff)), 0); |
138 | WREG32(R600_RCU_DATA, (v))r100_mm_wreg(rdev, (0x0104), ((v)), 0); |
139 | spin_unlock_irqrestore(&rdev->rcu_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->rcu_idx_lock); } while (0); |
140 | } |
141 | |
142 | u32 r600_uvd_ctx_rreg(struct radeon_device *rdev, u32 reg) |
143 | { |
144 | unsigned long flags; |
145 | u32 r; |
146 | |
147 | spin_lock_irqsave(&rdev->uvd_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->uvd_idx_lock); } while (0); |
148 | WREG32(R600_UVD_CTX_INDEX, ((reg) & 0x1ff))r100_mm_wreg(rdev, (0xf4a0), (((reg) & 0x1ff)), 0); |
149 | r = RREG32(R600_UVD_CTX_DATA)r100_mm_rreg(rdev, (0xf4a4), 0); |
150 | spin_unlock_irqrestore(&rdev->uvd_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->uvd_idx_lock); } while (0); |
151 | return r; |
152 | } |
153 | |
154 | void r600_uvd_ctx_wreg(struct radeon_device *rdev, u32 reg, u32 v) |
155 | { |
156 | unsigned long flags; |
157 | |
158 | spin_lock_irqsave(&rdev->uvd_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->uvd_idx_lock); } while (0); |
159 | WREG32(R600_UVD_CTX_INDEX, ((reg) & 0x1ff))r100_mm_wreg(rdev, (0xf4a0), (((reg) & 0x1ff)), 0); |
160 | WREG32(R600_UVD_CTX_DATA, (v))r100_mm_wreg(rdev, (0xf4a4), ((v)), 0); |
161 | spin_unlock_irqrestore(&rdev->uvd_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->uvd_idx_lock); } while (0); |
162 | } |
163 | |
164 | /** |
165 | * r600_get_allowed_info_register - fetch the register for the info ioctl |
166 | * |
167 | * @rdev: radeon_device pointer |
168 | * @reg: register offset in bytes |
169 | * @val: register value |
170 | * |
171 | * Returns 0 for success or -EINVAL for an invalid register |
172 | * |
173 | */ |
174 | int r600_get_allowed_info_register(struct radeon_device *rdev, |
175 | u32 reg, u32 *val) |
176 | { |
177 | switch (reg) { |
178 | case GRBM_STATUS0x8010: |
179 | case GRBM_STATUS20x8014: |
180 | case R_000E50_SRBM_STATUS0x0E50: |
181 | case DMA_STATUS_REG0xd034: |
182 | case UVD_STATUS0xf6bc: |
183 | *val = RREG32(reg)r100_mm_rreg(rdev, (reg), 0); |
184 | return 0; |
185 | default: |
186 | return -EINVAL22; |
187 | } |
188 | } |
189 | |
190 | /** |
191 | * r600_get_xclk - get the xclk |
192 | * |
193 | * @rdev: radeon_device pointer |
194 | * |
195 | * Returns the reference clock used by the gfx engine |
196 | * (r6xx, IGPs, APUs). |
197 | */ |
198 | u32 r600_get_xclk(struct radeon_device *rdev) |
199 | { |
200 | return rdev->clock.spll.reference_freq; |
201 | } |
202 | |
203 | int r600_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) |
204 | { |
205 | unsigned fb_div = 0, ref_div, vclk_div = 0, dclk_div = 0; |
206 | int r; |
207 | |
208 | /* bypass vclk and dclk with bclk */ |
209 | WREG32_P(CG_UPLL_FUNC_CNTL_2,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~(0x01F00000 | 0x3E000000)); tmp_ |= ((((1) << 20) | ( (1) << 25)) & ~(~(0x01F00000 | 0x3E000000))); r100_mm_wreg (rdev, (0x7e4), (tmp_), 0); } while (0) |
210 | VCLK_SRC_SEL(1) | DCLK_SRC_SEL(1),do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~(0x01F00000 | 0x3E000000)); tmp_ |= ((((1) << 20) | ( (1) << 25)) & ~(~(0x01F00000 | 0x3E000000))); r100_mm_wreg (rdev, (0x7e4), (tmp_), 0); } while (0) |
211 | ~(VCLK_SRC_SEL_MASK | DCLK_SRC_SEL_MASK))do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~(0x01F00000 | 0x3E000000)); tmp_ |= ((((1) << 20) | ( (1) << 25)) & ~(~(0x01F00000 | 0x3E000000))); r100_mm_wreg (rdev, (0x7e4), (tmp_), 0); } while (0); |
212 | |
213 | /* assert BYPASS_EN, deassert UPLL_RESET, UPLL_SLEEP and UPLL_CTLREQ */ |
214 | WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_BYPASS_EN_MASK, ~(do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~( 0x00000001 | 0x00000002 | 0x00000008)); tmp_ |= ((0x00000004 ) & ~(~( 0x00000001 | 0x00000002 | 0x00000008))); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0) |
215 | UPLL_RESET_MASK | UPLL_SLEEP_MASK | UPLL_CTLREQ_MASK))do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~( 0x00000001 | 0x00000002 | 0x00000008)); tmp_ |= ((0x00000004 ) & ~(~( 0x00000001 | 0x00000002 | 0x00000008))); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0); |
216 | |
217 | if (rdev->family >= CHIP_RS780) |
218 | WREG32_P(GFX_MACRO_BYPASS_CNTL, UPLL_BYPASS_CNTL,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x30c0), 0); tmp_ &= (~(1 << 1)); tmp_ |= (((1 << 1)) & ~(~(1 << 1))); r100_mm_wreg(rdev, (0x30c0), (tmp_), 0); } while (0) |
219 | ~UPLL_BYPASS_CNTL)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x30c0), 0); tmp_ &= (~(1 << 1)); tmp_ |= (((1 << 1)) & ~(~(1 << 1))); r100_mm_wreg(rdev, (0x30c0), (tmp_), 0); } while (0); |
220 | |
221 | if (!vclk || !dclk) { |
222 | /* keep the Bypass mode, put PLL to sleep */ |
223 | WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_SLEEP_MASK, ~UPLL_SLEEP_MASK)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~0x00000002); tmp_ |= ((0x00000002) & ~(~0x00000002)); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0); |
224 | return 0; |
225 | } |
226 | |
227 | if (rdev->clock.spll.reference_freq == 10000) |
228 | ref_div = 34; |
229 | else |
230 | ref_div = 4; |
231 | |
232 | r = radeon_uvd_calc_upll_dividers(rdev, vclk, dclk, 50000, 160000, |
233 | ref_div + 1, 0xFFF, 2, 30, ~0, |
234 | &fb_div, &vclk_div, &dclk_div); |
235 | if (r) |
236 | return r; |
237 | |
238 | if (rdev->family >= CHIP_RV670 && rdev->family < CHIP_RS780) |
239 | fb_div >>= 1; |
240 | else |
241 | fb_div |= 1; |
242 | |
243 | r = radeon_uvd_send_upll_ctlreq(rdev, CG_UPLL_FUNC_CNTL0x7e0); |
244 | if (r) |
245 | return r; |
246 | |
247 | /* assert PLL_RESET */ |
248 | WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_RESET_MASK, ~UPLL_RESET_MASK)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~0x00000001); tmp_ |= ((0x00000001) & ~(~0x00000001)); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0); |
249 | |
250 | /* For RS780 we have to choose ref clk */ |
251 | if (rdev->family >= CHIP_RS780) |
252 | WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_REFCLK_SRC_SEL_MASK,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~0x20000000); tmp_ |= ((0x20000000) & ~(~0x20000000)); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0) |
253 | ~UPLL_REFCLK_SRC_SEL_MASK)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~0x20000000); tmp_ |= ((0x20000000) & ~(~0x20000000)); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0); |
254 | |
255 | /* set the required fb, ref and post divder values */ |
256 | WREG32_P(CG_UPLL_FUNC_CNTL,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~(0x0000FFF0 | 0x003F0000)); tmp_ |= ((((fb_div) << 4 ) | ((ref_div) << 16)) & ~(~(0x0000FFF0 | 0x003F0000 ))); r100_mm_wreg(rdev, (0x7e0), (tmp_), 0); } while (0) |
257 | UPLL_FB_DIV(fb_div) |do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~(0x0000FFF0 | 0x003F0000)); tmp_ |= ((((fb_div) << 4 ) | ((ref_div) << 16)) & ~(~(0x0000FFF0 | 0x003F0000 ))); r100_mm_wreg(rdev, (0x7e0), (tmp_), 0); } while (0) |
258 | UPLL_REF_DIV(ref_div),do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~(0x0000FFF0 | 0x003F0000)); tmp_ |= ((((fb_div) << 4 ) | ((ref_div) << 16)) & ~(~(0x0000FFF0 | 0x003F0000 ))); r100_mm_wreg(rdev, (0x7e0), (tmp_), 0); } while (0) |
259 | ~(UPLL_FB_DIV_MASK | UPLL_REF_DIV_MASK))do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~(0x0000FFF0 | 0x003F0000)); tmp_ |= ((((fb_div) << 4 ) | ((ref_div) << 16)) & ~(~(0x0000FFF0 | 0x003F0000 ))); r100_mm_wreg(rdev, (0x7e0), (tmp_), 0); } while (0); |
260 | WREG32_P(CG_UPLL_FUNC_CNTL_2,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0) |
261 | UPLL_SW_HILEN(vclk_div >> 1) |do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0) |
262 | UPLL_SW_LOLEN((vclk_div >> 1) + (vclk_div & 1)) |do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0) |
263 | UPLL_SW_HILEN2(dclk_div >> 1) |do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0) |
264 | UPLL_SW_LOLEN2((dclk_div >> 1) + (dclk_div & 1)) |do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0) |
265 | UPLL_DIVEN_MASK | UPLL_DIVEN2_MASK,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0) |
266 | ~UPLL_SW_MASK)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~0x0003FFFF); tmp_ |= ((((vclk_div >> 1) << 0) | (((vclk_div >> 1) + (vclk_div & 1)) << 4) | ( (dclk_div >> 1) << 8) | (((dclk_div >> 1) + (dclk_div & 1)) << 12) | 0x00010000 | 0x00020000) & ~(~0x0003FFFF)); r100_mm_wreg(rdev, (0x7e4), (tmp_), 0); } while (0); |
267 | |
268 | /* give the PLL some time to settle */ |
269 | mdelay(15); |
270 | |
271 | /* deassert PLL_RESET */ |
272 | WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_RESET_MASK)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~0x00000001); tmp_ |= ((0) & ~(~0x00000001)); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0); |
273 | |
274 | mdelay(15); |
275 | |
276 | /* deassert BYPASS EN */ |
277 | WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_BYPASS_EN_MASK)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e0), 0); tmp_ &= (~0x00000004); tmp_ |= ((0) & ~(~0x00000004)); r100_mm_wreg (rdev, (0x7e0), (tmp_), 0); } while (0); |
278 | |
279 | if (rdev->family >= CHIP_RS780) |
280 | WREG32_P(GFX_MACRO_BYPASS_CNTL, 0, ~UPLL_BYPASS_CNTL)do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x30c0), 0); tmp_ &= (~(1 << 1)); tmp_ |= ((0) & ~(~(1 << 1))); r100_mm_wreg (rdev, (0x30c0), (tmp_), 0); } while (0); |
281 | |
282 | r = radeon_uvd_send_upll_ctlreq(rdev, CG_UPLL_FUNC_CNTL0x7e0); |
283 | if (r) |
284 | return r; |
285 | |
286 | /* switch VCLK and DCLK selection */ |
287 | WREG32_P(CG_UPLL_FUNC_CNTL_2,do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~(0x01F00000 | 0x3E000000)); tmp_ |= ((((2) << 20) | ( (2) << 25)) & ~(~(0x01F00000 | 0x3E000000))); r100_mm_wreg (rdev, (0x7e4), (tmp_), 0); } while (0) |
288 | VCLK_SRC_SEL(2) | DCLK_SRC_SEL(2),do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~(0x01F00000 | 0x3E000000)); tmp_ |= ((((2) << 20) | ( (2) << 25)) & ~(~(0x01F00000 | 0x3E000000))); r100_mm_wreg (rdev, (0x7e4), (tmp_), 0); } while (0) |
289 | ~(VCLK_SRC_SEL_MASK | DCLK_SRC_SEL_MASK))do { uint32_t tmp_ = r100_mm_rreg(rdev, (0x7e4), 0); tmp_ &= (~(0x01F00000 | 0x3E000000)); tmp_ |= ((((2) << 20) | ( (2) << 25)) & ~(~(0x01F00000 | 0x3E000000))); r100_mm_wreg (rdev, (0x7e4), (tmp_), 0); } while (0); |
290 | |
291 | mdelay(100); |
292 | |
293 | return 0; |
294 | } |
295 | |
296 | void dce3_program_fmt(struct drm_encoder *encoder) |
297 | { |
298 | struct drm_device *dev = encoder->dev; |
299 | struct radeon_device *rdev = dev->dev_private; |
300 | struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder)({ const __typeof( ((struct radeon_encoder *)0)->base ) *__mptr = (encoder); (struct radeon_encoder *)( (char *)__mptr - __builtin_offsetof (struct radeon_encoder, base) );}); |
301 | struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc)({ const __typeof( ((struct radeon_crtc *)0)->base ) *__mptr = (encoder->crtc); (struct radeon_crtc *)( (char *)__mptr - __builtin_offsetof(struct radeon_crtc, base) );}); |
302 | struct drm_connector *connector = radeon_get_connector_for_encoder(encoder); |
303 | int bpc = 0; |
304 | u32 tmp = 0; |
305 | enum radeon_connector_dither dither = RADEON_FMT_DITHER_DISABLE; |
306 | |
307 | if (connector) { |
308 | struct radeon_connector *radeon_connector = to_radeon_connector(connector)({ const __typeof( ((struct radeon_connector *)0)->base ) * __mptr = (connector); (struct radeon_connector *)( (char *)__mptr - __builtin_offsetof(struct radeon_connector, base) );}); |
309 | bpc = radeon_get_monitor_bpc(connector); |
310 | dither = radeon_connector->dither; |
311 | } |
312 | |
313 | /* LVDS FMT is set up by atom */ |
314 | if (radeon_encoder->devices & ATOM_DEVICE_LCD_SUPPORT((0x1L << 0x00000001 ) | (0x1L << 0x00000005 ))) |
315 | return; |
316 | |
317 | /* not needed for analog */ |
318 | if ((radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC10x15) || |
319 | (radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC20x16)) |
320 | return; |
321 | |
322 | if (bpc == 0) |
323 | return; |
324 | |
325 | switch (bpc) { |
326 | case 6: |
327 | if (dither == RADEON_FMT_DITHER_ENABLE) |
328 | /* XXX sort out optimal dither settings */ |
329 | tmp |= FMT_SPATIAL_DITHER_EN(1 << 8); |
330 | else |
331 | tmp |= FMT_TRUNCATE_EN(1 << 0); |
332 | break; |
333 | case 8: |
334 | if (dither == RADEON_FMT_DITHER_ENABLE) |
335 | /* XXX sort out optimal dither settings */ |
336 | tmp |= (FMT_SPATIAL_DITHER_EN(1 << 8) | FMT_SPATIAL_DITHER_DEPTH(1 << 12)); |
337 | else |
338 | tmp |= (FMT_TRUNCATE_EN(1 << 0) | FMT_TRUNCATE_DEPTH(1 << 4)); |
339 | break; |
340 | case 10: |
341 | default: |
342 | /* not needed */ |
343 | break; |
344 | } |
345 | |
346 | WREG32(FMT_BIT_DEPTH_CONTROL + radeon_crtc->crtc_offset, tmp)r100_mm_wreg(rdev, (0x6710 + radeon_crtc->crtc_offset), (tmp ), 0); |
347 | } |
348 | |
349 | /* get temperature in millidegrees */ |
350 | int rv6xx_get_temp(struct radeon_device *rdev) |
351 | { |
352 | u32 temp = (RREG32(CG_THERMAL_STATUS)r100_mm_rreg(rdev, (0x7F4), 0) & ASIC_T_MASK0x1FF) >> |
353 | ASIC_T_SHIFT0; |
354 | int actual_temp = temp & 0xff; |
355 | |
356 | if (temp & 0x100) |
357 | actual_temp -= 256; |
358 | |
359 | return actual_temp * 1000; |
360 | } |
361 | |
362 | void r600_pm_get_dynpm_state(struct radeon_device *rdev) |
363 | { |
364 | int i; |
365 | |
366 | rdev->pm.dynpm_can_upclock = true1; |
367 | rdev->pm.dynpm_can_downclock = true1; |
368 | |
369 | /* power state array is low to high, default is first */ |
370 | if ((rdev->flags & RADEON_IS_IGP) || (rdev->family == CHIP_R600)) { |
371 | int min_power_state_index = 0; |
372 | |
373 | if (rdev->pm.num_power_states > 2) |
374 | min_power_state_index = 1; |
375 | |
376 | switch (rdev->pm.dynpm_planned_action) { |
377 | case DYNPM_ACTION_MINIMUM: |
378 | rdev->pm.requested_power_state_index = min_power_state_index; |
379 | rdev->pm.requested_clock_mode_index = 0; |
380 | rdev->pm.dynpm_can_downclock = false0; |
381 | break; |
382 | case DYNPM_ACTION_DOWNCLOCK: |
383 | if (rdev->pm.current_power_state_index == min_power_state_index) { |
384 | rdev->pm.requested_power_state_index = rdev->pm.current_power_state_index; |
385 | rdev->pm.dynpm_can_downclock = false0; |
386 | } else { |
387 | if (rdev->pm.active_crtc_count > 1) { |
388 | for (i = 0; i < rdev->pm.num_power_states; i++) { |
389 | if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY(1 << 0)) |
390 | continue; |
391 | else if (i >= rdev->pm.current_power_state_index) { |
392 | rdev->pm.requested_power_state_index = |
393 | rdev->pm.current_power_state_index; |
394 | break; |
395 | } else { |
396 | rdev->pm.requested_power_state_index = i; |
397 | break; |
398 | } |
399 | } |
400 | } else { |
401 | if (rdev->pm.current_power_state_index == 0) |
402 | rdev->pm.requested_power_state_index = |
403 | rdev->pm.num_power_states - 1; |
404 | else |
405 | rdev->pm.requested_power_state_index = |
406 | rdev->pm.current_power_state_index - 1; |
407 | } |
408 | } |
409 | rdev->pm.requested_clock_mode_index = 0; |
410 | /* don't use the power state if crtcs are active and no display flag is set */ |
411 | if ((rdev->pm.active_crtc_count > 0) && |
412 | (rdev->pm.power_state[rdev->pm.requested_power_state_index]. |
413 | clock_info[rdev->pm.requested_clock_mode_index].flags & |
414 | RADEON_PM_MODE_NO_DISPLAY(1 << 0))) { |
415 | rdev->pm.requested_power_state_index++; |
416 | } |
417 | break; |
418 | case DYNPM_ACTION_UPCLOCK: |
419 | if (rdev->pm.current_power_state_index == (rdev->pm.num_power_states - 1)) { |
420 | rdev->pm.requested_power_state_index = rdev->pm.current_power_state_index; |
421 | rdev->pm.dynpm_can_upclock = false0; |
422 | } else { |
423 | if (rdev->pm.active_crtc_count > 1) { |
424 | for (i = (rdev->pm.num_power_states - 1); i >= 0; i--) { |
425 | if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY(1 << 0)) |
426 | continue; |
427 | else if (i <= rdev->pm.current_power_state_index) { |
428 | rdev->pm.requested_power_state_index = |
429 | rdev->pm.current_power_state_index; |
430 | break; |
431 | } else { |
432 | rdev->pm.requested_power_state_index = i; |
433 | break; |
434 | } |
435 | } |
436 | } else |
437 | rdev->pm.requested_power_state_index = |
438 | rdev->pm.current_power_state_index + 1; |
439 | } |
440 | rdev->pm.requested_clock_mode_index = 0; |
441 | break; |
442 | case DYNPM_ACTION_DEFAULT: |
443 | rdev->pm.requested_power_state_index = rdev->pm.default_power_state_index; |
444 | rdev->pm.requested_clock_mode_index = 0; |
445 | rdev->pm.dynpm_can_upclock = false0; |
446 | break; |
447 | case DYNPM_ACTION_NONE: |
448 | default: |
449 | DRM_ERROR("Requested mode for not defined action\n")__drm_err("Requested mode for not defined action\n"); |
450 | return; |
451 | } |
452 | } else { |
453 | /* XXX select a power state based on AC/DC, single/dualhead, etc. */ |
454 | /* for now just select the first power state and switch between clock modes */ |
455 | /* power state array is low to high, default is first (0) */ |
456 | if (rdev->pm.active_crtc_count > 1) { |
457 | rdev->pm.requested_power_state_index = -1; |
458 | /* start at 1 as we don't want the default mode */ |
459 | for (i = 1; i < rdev->pm.num_power_states; i++) { |
460 | if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY(1 << 0)) |
461 | continue; |
462 | else if ((rdev->pm.power_state[i].type == POWER_STATE_TYPE_PERFORMANCE) || |
463 | (rdev->pm.power_state[i].type == POWER_STATE_TYPE_BATTERY)) { |
464 | rdev->pm.requested_power_state_index = i; |
465 | break; |
466 | } |
467 | } |
468 | /* if nothing selected, grab the default state. */ |
469 | if (rdev->pm.requested_power_state_index == -1) |
470 | rdev->pm.requested_power_state_index = 0; |
471 | } else |
472 | rdev->pm.requested_power_state_index = 1; |
473 | |
474 | switch (rdev->pm.dynpm_planned_action) { |
475 | case DYNPM_ACTION_MINIMUM: |
476 | rdev->pm.requested_clock_mode_index = 0; |
477 | rdev->pm.dynpm_can_downclock = false0; |
478 | break; |
479 | case DYNPM_ACTION_DOWNCLOCK: |
480 | if (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index) { |
481 | if (rdev->pm.current_clock_mode_index == 0) { |
482 | rdev->pm.requested_clock_mode_index = 0; |
483 | rdev->pm.dynpm_can_downclock = false0; |
484 | } else |
485 | rdev->pm.requested_clock_mode_index = |
486 | rdev->pm.current_clock_mode_index - 1; |
487 | } else { |
488 | rdev->pm.requested_clock_mode_index = 0; |
489 | rdev->pm.dynpm_can_downclock = false0; |
490 | } |
491 | /* don't use the power state if crtcs are active and no display flag is set */ |
492 | if ((rdev->pm.active_crtc_count > 0) && |
493 | (rdev->pm.power_state[rdev->pm.requested_power_state_index]. |
494 | clock_info[rdev->pm.requested_clock_mode_index].flags & |
495 | RADEON_PM_MODE_NO_DISPLAY(1 << 0))) { |
496 | rdev->pm.requested_clock_mode_index++; |
497 | } |
498 | break; |
499 | case DYNPM_ACTION_UPCLOCK: |
500 | if (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index) { |
501 | if (rdev->pm.current_clock_mode_index == |
502 | (rdev->pm.power_state[rdev->pm.requested_power_state_index].num_clock_modes - 1)) { |
503 | rdev->pm.requested_clock_mode_index = rdev->pm.current_clock_mode_index; |
504 | rdev->pm.dynpm_can_upclock = false0; |
505 | } else |
506 | rdev->pm.requested_clock_mode_index = |
507 | rdev->pm.current_clock_mode_index + 1; |
508 | } else { |
509 | rdev->pm.requested_clock_mode_index = |
510 | rdev->pm.power_state[rdev->pm.requested_power_state_index].num_clock_modes - 1; |
511 | rdev->pm.dynpm_can_upclock = false0; |
512 | } |
513 | break; |
514 | case DYNPM_ACTION_DEFAULT: |
515 | rdev->pm.requested_power_state_index = rdev->pm.default_power_state_index; |
516 | rdev->pm.requested_clock_mode_index = 0; |
517 | rdev->pm.dynpm_can_upclock = false0; |
518 | break; |
519 | case DYNPM_ACTION_NONE: |
520 | default: |
521 | DRM_ERROR("Requested mode for not defined action\n")__drm_err("Requested mode for not defined action\n"); |
522 | return; |
523 | } |
524 | } |
525 | |
526 | DRM_DEBUG_DRIVER("Requested: e: %d m: %d p: %d\n",___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes) |
527 | rdev->pm.power_state[rdev->pm.requested_power_state_index].___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes) |
528 | clock_info[rdev->pm.requested_clock_mode_index].sclk,___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes) |
529 | rdev->pm.power_state[rdev->pm.requested_power_state_index].___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes) |
530 | clock_info[rdev->pm.requested_clock_mode_index].mclk,___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes) |
531 | rdev->pm.power_state[rdev->pm.requested_power_state_index].___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes) |
532 | pcie_lanes)___drm_dbg(((void *)0), DRM_UT_DRIVER, "Requested: e: %d m: %d p: %d\n" , rdev->pm.power_state[rdev->pm.requested_power_state_index ]. clock_info[rdev->pm.requested_clock_mode_index].sclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. clock_info[rdev->pm.requested_clock_mode_index].mclk, rdev ->pm.power_state[rdev->pm.requested_power_state_index]. pcie_lanes); |
533 | } |
534 | |
535 | void rs780_pm_init_profile(struct radeon_device *rdev) |
536 | { |
537 | if (rdev->pm.num_power_states == 2) { |
538 | /* default */ |
539 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
540 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
541 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_cm_idx = 0; |
542 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_cm_idx = 0; |
543 | /* low sh */ |
544 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_ps_idx = 0; |
545 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_ps_idx = 0; |
546 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_cm_idx = 0; |
547 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_cm_idx = 0; |
548 | /* mid sh */ |
549 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_ps_idx = 0; |
550 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_ps_idx = 0; |
551 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_cm_idx = 0; |
552 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_cm_idx = 0; |
553 | /* high sh */ |
554 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_ps_idx = 0; |
555 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_ps_idx = 1; |
556 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_cm_idx = 0; |
557 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_cm_idx = 0; |
558 | /* low mh */ |
559 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_ps_idx = 0; |
560 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_ps_idx = 0; |
561 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_cm_idx = 0; |
562 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_cm_idx = 0; |
563 | /* mid mh */ |
564 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_ps_idx = 0; |
565 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_ps_idx = 0; |
566 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_cm_idx = 0; |
567 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_cm_idx = 0; |
568 | /* high mh */ |
569 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_ps_idx = 0; |
570 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_ps_idx = 1; |
571 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_cm_idx = 0; |
572 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_cm_idx = 0; |
573 | } else if (rdev->pm.num_power_states == 3) { |
574 | /* default */ |
575 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
576 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
577 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_cm_idx = 0; |
578 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_cm_idx = 0; |
579 | /* low sh */ |
580 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_ps_idx = 1; |
581 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_ps_idx = 1; |
582 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_cm_idx = 0; |
583 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_cm_idx = 0; |
584 | /* mid sh */ |
585 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_ps_idx = 1; |
586 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_ps_idx = 1; |
587 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_cm_idx = 0; |
588 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_cm_idx = 0; |
589 | /* high sh */ |
590 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_ps_idx = 1; |
591 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_ps_idx = 2; |
592 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_cm_idx = 0; |
593 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_cm_idx = 0; |
594 | /* low mh */ |
595 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_ps_idx = 1; |
596 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_ps_idx = 1; |
597 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_cm_idx = 0; |
598 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_cm_idx = 0; |
599 | /* mid mh */ |
600 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_ps_idx = 1; |
601 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_ps_idx = 1; |
602 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_cm_idx = 0; |
603 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_cm_idx = 0; |
604 | /* high mh */ |
605 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_ps_idx = 1; |
606 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_ps_idx = 2; |
607 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_cm_idx = 0; |
608 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_cm_idx = 0; |
609 | } else { |
610 | /* default */ |
611 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
612 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
613 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_cm_idx = 0; |
614 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_cm_idx = 0; |
615 | /* low sh */ |
616 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_ps_idx = 2; |
617 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_ps_idx = 2; |
618 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_cm_idx = 0; |
619 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_cm_idx = 0; |
620 | /* mid sh */ |
621 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_ps_idx = 2; |
622 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_ps_idx = 2; |
623 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_cm_idx = 0; |
624 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_cm_idx = 0; |
625 | /* high sh */ |
626 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_ps_idx = 2; |
627 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_ps_idx = 3; |
628 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_cm_idx = 0; |
629 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_cm_idx = 0; |
630 | /* low mh */ |
631 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_ps_idx = 2; |
632 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_ps_idx = 0; |
633 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_cm_idx = 0; |
634 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_cm_idx = 0; |
635 | /* mid mh */ |
636 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_ps_idx = 2; |
637 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_ps_idx = 0; |
638 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_cm_idx = 0; |
639 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_cm_idx = 0; |
640 | /* high mh */ |
641 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_ps_idx = 2; |
642 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_ps_idx = 3; |
643 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_cm_idx = 0; |
644 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_cm_idx = 0; |
645 | } |
646 | } |
647 | |
648 | void r600_pm_init_profile(struct radeon_device *rdev) |
649 | { |
650 | int idx; |
651 | |
652 | if (rdev->family == CHIP_R600) { |
653 | /* XXX */ |
654 | /* default */ |
655 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
656 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
657 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_cm_idx = 0; |
658 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_cm_idx = 0; |
659 | /* low sh */ |
660 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
661 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
662 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_cm_idx = 0; |
663 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_cm_idx = 0; |
664 | /* mid sh */ |
665 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
666 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
667 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_cm_idx = 0; |
668 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_cm_idx = 0; |
669 | /* high sh */ |
670 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
671 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
672 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_cm_idx = 0; |
673 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_cm_idx = 0; |
674 | /* low mh */ |
675 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
676 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
677 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_cm_idx = 0; |
678 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_cm_idx = 0; |
679 | /* mid mh */ |
680 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
681 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
682 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_cm_idx = 0; |
683 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_cm_idx = 0; |
684 | /* high mh */ |
685 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
686 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
687 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_cm_idx = 0; |
688 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_cm_idx = 0; |
689 | } else { |
690 | if (rdev->pm.num_power_states < 4) { |
691 | /* default */ |
692 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
693 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
694 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_cm_idx = 0; |
695 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_cm_idx = 2; |
696 | /* low sh */ |
697 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_ps_idx = 1; |
698 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_ps_idx = 1; |
699 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_cm_idx = 0; |
700 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_cm_idx = 0; |
701 | /* mid sh */ |
702 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_ps_idx = 1; |
703 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_ps_idx = 1; |
704 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_cm_idx = 0; |
705 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_cm_idx = 1; |
706 | /* high sh */ |
707 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_ps_idx = 1; |
708 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_ps_idx = 1; |
709 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_cm_idx = 0; |
710 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_cm_idx = 2; |
711 | /* low mh */ |
712 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_ps_idx = 2; |
713 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_ps_idx = 2; |
714 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_cm_idx = 0; |
715 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_cm_idx = 0; |
716 | /* low mh */ |
717 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_ps_idx = 2; |
718 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_ps_idx = 2; |
719 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_cm_idx = 0; |
720 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_cm_idx = 1; |
721 | /* high mh */ |
722 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_ps_idx = 2; |
723 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_ps_idx = 2; |
724 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_cm_idx = 0; |
725 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_cm_idx = 2; |
726 | } else { |
727 | /* default */ |
728 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_ps_idx = rdev->pm.default_power_state_index; |
729 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_ps_idx = rdev->pm.default_power_state_index; |
730 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_off_cm_idx = 0; |
731 | rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX0].dpms_on_cm_idx = 2; |
732 | /* low sh */ |
733 | if (rdev->flags & RADEON_IS_MOBILITY) |
734 | idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0); |
735 | else |
736 | idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); |
737 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_ps_idx = idx; |
738 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_ps_idx = idx; |
739 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_off_cm_idx = 0; |
740 | rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX1].dpms_on_cm_idx = 0; |
741 | /* mid sh */ |
742 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_ps_idx = idx; |
743 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_ps_idx = idx; |
744 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_off_cm_idx = 0; |
745 | rdev->pm.profiles[PM_PROFILE_MID_SH_IDX2].dpms_on_cm_idx = 1; |
746 | /* high sh */ |
747 | idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0); |
748 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_ps_idx = idx; |
749 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_ps_idx = idx; |
750 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_off_cm_idx = 0; |
751 | rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX3].dpms_on_cm_idx = 2; |
752 | /* low mh */ |
753 | if (rdev->flags & RADEON_IS_MOBILITY) |
754 | idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1); |
755 | else |
756 | idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); |
757 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_ps_idx = idx; |
758 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_ps_idx = idx; |
759 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_off_cm_idx = 0; |
760 | rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX4].dpms_on_cm_idx = 0; |
761 | /* mid mh */ |
762 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_ps_idx = idx; |
763 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_ps_idx = idx; |
764 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_off_cm_idx = 0; |
765 | rdev->pm.profiles[PM_PROFILE_MID_MH_IDX5].dpms_on_cm_idx = 1; |
766 | /* high mh */ |
767 | idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1); |
768 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_ps_idx = idx; |
769 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_ps_idx = idx; |
770 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_off_cm_idx = 0; |
771 | rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX6].dpms_on_cm_idx = 2; |
772 | } |
773 | } |
774 | } |
775 | |
776 | void r600_pm_misc(struct radeon_device *rdev) |
777 | { |
778 | int req_ps_idx = rdev->pm.requested_power_state_index; |
779 | int req_cm_idx = rdev->pm.requested_clock_mode_index; |
780 | struct radeon_power_state *ps = &rdev->pm.power_state[req_ps_idx]; |
781 | struct radeon_voltage *voltage = &ps->clock_info[req_cm_idx].voltage; |
782 | |
783 | if ((voltage->type == VOLTAGE_SW) && voltage->voltage) { |
784 | /* 0xff01 is a flag rather then an actual voltage */ |
785 | if (voltage->voltage == 0xff01) |
786 | return; |
787 | if (voltage->voltage != rdev->pm.current_vddc) { |
788 | radeon_atom_set_voltage(rdev, voltage->voltage, SET_VOLTAGE_TYPE_ASIC_VDDC1); |
789 | rdev->pm.current_vddc = voltage->voltage; |
790 | DRM_DEBUG_DRIVER("Setting: v: %d\n", voltage->voltage)___drm_dbg(((void *)0), DRM_UT_DRIVER, "Setting: v: %d\n", voltage ->voltage); |
791 | } |
792 | } |
793 | } |
794 | |
795 | bool_Bool r600_gui_idle(struct radeon_device *rdev) |
796 | { |
797 | if (RREG32(GRBM_STATUS)r100_mm_rreg(rdev, (0x8010), 0) & GUI_ACTIVE(1<<31)) |
798 | return false0; |
799 | else |
800 | return true1; |
801 | } |
802 | |
803 | /* hpd for digital panel detect/disconnect */ |
804 | bool_Bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd) |
805 | { |
806 | bool_Bool connected = false0; |
807 | |
808 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
809 | switch (hpd) { |
810 | case RADEON_HPD_1: |
811 | if (RREG32(DC_HPD1_INT_STATUS)r100_mm_rreg(rdev, (0x7d00), 0) & DC_HPDx_SENSE(1 << 1)) |
812 | connected = true1; |
813 | break; |
814 | case RADEON_HPD_2: |
815 | if (RREG32(DC_HPD2_INT_STATUS)r100_mm_rreg(rdev, (0x7d0c), 0) & DC_HPDx_SENSE(1 << 1)) |
816 | connected = true1; |
817 | break; |
818 | case RADEON_HPD_3: |
819 | if (RREG32(DC_HPD3_INT_STATUS)r100_mm_rreg(rdev, (0x7d18), 0) & DC_HPDx_SENSE(1 << 1)) |
820 | connected = true1; |
821 | break; |
822 | case RADEON_HPD_4: |
823 | if (RREG32(DC_HPD4_INT_STATUS)r100_mm_rreg(rdev, (0x7d24), 0) & DC_HPDx_SENSE(1 << 1)) |
824 | connected = true1; |
825 | break; |
826 | /* DCE 3.2 */ |
827 | case RADEON_HPD_5: |
828 | if (RREG32(DC_HPD5_INT_STATUS)r100_mm_rreg(rdev, (0x7dc0), 0) & DC_HPDx_SENSE(1 << 1)) |
829 | connected = true1; |
830 | break; |
831 | case RADEON_HPD_6: |
832 | if (RREG32(DC_HPD6_INT_STATUS)r100_mm_rreg(rdev, (0x7df4), 0) & DC_HPDx_SENSE(1 << 1)) |
833 | connected = true1; |
834 | break; |
835 | default: |
836 | break; |
837 | } |
838 | } else { |
839 | switch (hpd) { |
840 | case RADEON_HPD_1: |
841 | if (RREG32(DC_HOT_PLUG_DETECT1_INT_STATUS)r100_mm_rreg(rdev, (0x7d04), 0) & DC_HOT_PLUG_DETECTx_SENSE(1 << 1)) |
842 | connected = true1; |
843 | break; |
844 | case RADEON_HPD_2: |
845 | if (RREG32(DC_HOT_PLUG_DETECT2_INT_STATUS)r100_mm_rreg(rdev, (0x7d14), 0) & DC_HOT_PLUG_DETECTx_SENSE(1 << 1)) |
846 | connected = true1; |
847 | break; |
848 | case RADEON_HPD_3: |
849 | if (RREG32(DC_HOT_PLUG_DETECT3_INT_STATUS)r100_mm_rreg(rdev, (0x7d28), 0) & DC_HOT_PLUG_DETECTx_SENSE(1 << 1)) |
850 | connected = true1; |
851 | break; |
852 | default: |
853 | break; |
854 | } |
855 | } |
856 | return connected; |
857 | } |
858 | |
859 | void r600_hpd_set_polarity(struct radeon_device *rdev, |
860 | enum radeon_hpd_id hpd) |
861 | { |
862 | u32 tmp; |
863 | bool_Bool connected = r600_hpd_sense(rdev, hpd); |
864 | |
865 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
866 | switch (hpd) { |
867 | case RADEON_HPD_1: |
868 | tmp = RREG32(DC_HPD1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d04), 0); |
869 | if (connected) |
870 | tmp &= ~DC_HPDx_INT_POLARITY(1 << 8); |
871 | else |
872 | tmp |= DC_HPDx_INT_POLARITY(1 << 8); |
873 | WREG32(DC_HPD1_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d04), (tmp), 0); |
874 | break; |
875 | case RADEON_HPD_2: |
876 | tmp = RREG32(DC_HPD2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d10), 0); |
877 | if (connected) |
878 | tmp &= ~DC_HPDx_INT_POLARITY(1 << 8); |
879 | else |
880 | tmp |= DC_HPDx_INT_POLARITY(1 << 8); |
881 | WREG32(DC_HPD2_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d10), (tmp), 0); |
882 | break; |
883 | case RADEON_HPD_3: |
884 | tmp = RREG32(DC_HPD3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d1c), 0); |
885 | if (connected) |
886 | tmp &= ~DC_HPDx_INT_POLARITY(1 << 8); |
887 | else |
888 | tmp |= DC_HPDx_INT_POLARITY(1 << 8); |
889 | WREG32(DC_HPD3_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d1c), (tmp), 0); |
890 | break; |
891 | case RADEON_HPD_4: |
892 | tmp = RREG32(DC_HPD4_INT_CONTROL)r100_mm_rreg(rdev, (0x7d28), 0); |
893 | if (connected) |
894 | tmp &= ~DC_HPDx_INT_POLARITY(1 << 8); |
895 | else |
896 | tmp |= DC_HPDx_INT_POLARITY(1 << 8); |
897 | WREG32(DC_HPD4_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d28), (tmp), 0); |
898 | break; |
899 | case RADEON_HPD_5: |
900 | tmp = RREG32(DC_HPD5_INT_CONTROL)r100_mm_rreg(rdev, (0x7dc4), 0); |
901 | if (connected) |
902 | tmp &= ~DC_HPDx_INT_POLARITY(1 << 8); |
903 | else |
904 | tmp |= DC_HPDx_INT_POLARITY(1 << 8); |
905 | WREG32(DC_HPD5_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7dc4), (tmp), 0); |
906 | break; |
907 | /* DCE 3.2 */ |
908 | case RADEON_HPD_6: |
909 | tmp = RREG32(DC_HPD6_INT_CONTROL)r100_mm_rreg(rdev, (0x7df8), 0); |
910 | if (connected) |
911 | tmp &= ~DC_HPDx_INT_POLARITY(1 << 8); |
912 | else |
913 | tmp |= DC_HPDx_INT_POLARITY(1 << 8); |
914 | WREG32(DC_HPD6_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7df8), (tmp), 0); |
915 | break; |
916 | default: |
917 | break; |
918 | } |
919 | } else { |
920 | switch (hpd) { |
921 | case RADEON_HPD_1: |
922 | tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d08), 0); |
923 | if (connected) |
924 | tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
925 | else |
926 | tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
927 | WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d08), (tmp), 0); |
928 | break; |
929 | case RADEON_HPD_2: |
930 | tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d18), 0); |
931 | if (connected) |
932 | tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
933 | else |
934 | tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
935 | WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d18), (tmp), 0); |
936 | break; |
937 | case RADEON_HPD_3: |
938 | tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d2c), 0); |
939 | if (connected) |
940 | tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
941 | else |
942 | tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
943 | WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d2c), (tmp), 0); |
944 | break; |
945 | default: |
946 | break; |
947 | } |
948 | } |
949 | } |
950 | |
951 | void r600_hpd_init(struct radeon_device *rdev) |
952 | { |
953 | struct drm_device *dev = rdev->ddev; |
954 | struct drm_connector *connector; |
955 | unsigned enable = 0; |
956 | |
957 | list_for_each_entry(connector, &dev->mode_config.connector_list, head)for (connector = ({ const __typeof( ((__typeof(*connector) *) 0)->head ) *__mptr = ((&dev->mode_config.connector_list )->next); (__typeof(*connector) *)( (char *)__mptr - __builtin_offsetof (__typeof(*connector), head) );}); &connector->head != (&dev->mode_config.connector_list); connector = ({ const __typeof( ((__typeof(*connector) *)0)->head ) *__mptr = ( connector->head.next); (__typeof(*connector) *)( (char *)__mptr - __builtin_offsetof(__typeof(*connector), head) );})) { |
958 | struct radeon_connector *radeon_connector = to_radeon_connector(connector)({ const __typeof( ((struct radeon_connector *)0)->base ) * __mptr = (connector); (struct radeon_connector *)( (char *)__mptr - __builtin_offsetof(struct radeon_connector, base) );}); |
959 | |
960 | if (connector->connector_type == DRM_MODE_CONNECTOR_eDP14 || |
961 | connector->connector_type == DRM_MODE_CONNECTOR_LVDS7) { |
962 | /* don't try to enable hpd on eDP or LVDS avoid breaking the |
963 | * aux dp channel on imac and help (but not completely fix) |
964 | * https://bugzilla.redhat.com/show_bug.cgi?id=726143 |
965 | */ |
966 | continue; |
967 | } |
968 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
969 | u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4)((0x9c4) << 0) | DC_HPDx_RX_INT_TIMER(0xfa)((0xfa) << 16); |
970 | if (ASIC_IS_DCE32(rdev)((rdev->family >= CHIP_RV730))) |
971 | tmp |= DC_HPDx_EN(1 << 28); |
972 | |
973 | switch (radeon_connector->hpd.hpd) { |
974 | case RADEON_HPD_1: |
975 | WREG32(DC_HPD1_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d08), (tmp), 0); |
976 | break; |
977 | case RADEON_HPD_2: |
978 | WREG32(DC_HPD2_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d14), (tmp), 0); |
979 | break; |
980 | case RADEON_HPD_3: |
981 | WREG32(DC_HPD3_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d20), (tmp), 0); |
982 | break; |
983 | case RADEON_HPD_4: |
984 | WREG32(DC_HPD4_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d2c), (tmp), 0); |
985 | break; |
986 | /* DCE 3.2 */ |
987 | case RADEON_HPD_5: |
988 | WREG32(DC_HPD5_CONTROL, tmp)r100_mm_wreg(rdev, (0x7dc8), (tmp), 0); |
989 | break; |
990 | case RADEON_HPD_6: |
991 | WREG32(DC_HPD6_CONTROL, tmp)r100_mm_wreg(rdev, (0x7dfc), (tmp), 0); |
992 | break; |
993 | default: |
994 | break; |
995 | } |
996 | } else { |
997 | switch (radeon_connector->hpd.hpd) { |
998 | case RADEON_HPD_1: |
999 | WREG32(DC_HOT_PLUG_DETECT1_CONTROL, DC_HOT_PLUG_DETECTx_EN)r100_mm_wreg(rdev, (0x7d00), ((1 << 0)), 0); |
1000 | break; |
1001 | case RADEON_HPD_2: |
1002 | WREG32(DC_HOT_PLUG_DETECT2_CONTROL, DC_HOT_PLUG_DETECTx_EN)r100_mm_wreg(rdev, (0x7d10), ((1 << 0)), 0); |
1003 | break; |
1004 | case RADEON_HPD_3: |
1005 | WREG32(DC_HOT_PLUG_DETECT3_CONTROL, DC_HOT_PLUG_DETECTx_EN)r100_mm_wreg(rdev, (0x7d24), ((1 << 0)), 0); |
1006 | break; |
1007 | default: |
1008 | break; |
1009 | } |
1010 | } |
1011 | if (radeon_connector->hpd.hpd != RADEON_HPD_NONE) |
1012 | enable |= 1 << radeon_connector->hpd.hpd; |
1013 | radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd)(rdev)->asic->hpd.set_polarity((rdev), (radeon_connector ->hpd.hpd)); |
1014 | } |
1015 | radeon_irq_kms_enable_hpd(rdev, enable); |
1016 | } |
1017 | |
1018 | void r600_hpd_fini(struct radeon_device *rdev) |
1019 | { |
1020 | struct drm_device *dev = rdev->ddev; |
1021 | struct drm_connector *connector; |
1022 | unsigned disable = 0; |
1023 | |
1024 | list_for_each_entry(connector, &dev->mode_config.connector_list, head)for (connector = ({ const __typeof( ((__typeof(*connector) *) 0)->head ) *__mptr = ((&dev->mode_config.connector_list )->next); (__typeof(*connector) *)( (char *)__mptr - __builtin_offsetof (__typeof(*connector), head) );}); &connector->head != (&dev->mode_config.connector_list); connector = ({ const __typeof( ((__typeof(*connector) *)0)->head ) *__mptr = ( connector->head.next); (__typeof(*connector) *)( (char *)__mptr - __builtin_offsetof(__typeof(*connector), head) );})) { |
1025 | struct radeon_connector *radeon_connector = to_radeon_connector(connector)({ const __typeof( ((struct radeon_connector *)0)->base ) * __mptr = (connector); (struct radeon_connector *)( (char *)__mptr - __builtin_offsetof(struct radeon_connector, base) );}); |
1026 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
1027 | switch (radeon_connector->hpd.hpd) { |
1028 | case RADEON_HPD_1: |
1029 | WREG32(DC_HPD1_CONTROL, 0)r100_mm_wreg(rdev, (0x7d08), (0), 0); |
1030 | break; |
1031 | case RADEON_HPD_2: |
1032 | WREG32(DC_HPD2_CONTROL, 0)r100_mm_wreg(rdev, (0x7d14), (0), 0); |
1033 | break; |
1034 | case RADEON_HPD_3: |
1035 | WREG32(DC_HPD3_CONTROL, 0)r100_mm_wreg(rdev, (0x7d20), (0), 0); |
1036 | break; |
1037 | case RADEON_HPD_4: |
1038 | WREG32(DC_HPD4_CONTROL, 0)r100_mm_wreg(rdev, (0x7d2c), (0), 0); |
1039 | break; |
1040 | /* DCE 3.2 */ |
1041 | case RADEON_HPD_5: |
1042 | WREG32(DC_HPD5_CONTROL, 0)r100_mm_wreg(rdev, (0x7dc8), (0), 0); |
1043 | break; |
1044 | case RADEON_HPD_6: |
1045 | WREG32(DC_HPD6_CONTROL, 0)r100_mm_wreg(rdev, (0x7dfc), (0), 0); |
1046 | break; |
1047 | default: |
1048 | break; |
1049 | } |
1050 | } else { |
1051 | switch (radeon_connector->hpd.hpd) { |
1052 | case RADEON_HPD_1: |
1053 | WREG32(DC_HOT_PLUG_DETECT1_CONTROL, 0)r100_mm_wreg(rdev, (0x7d00), (0), 0); |
1054 | break; |
1055 | case RADEON_HPD_2: |
1056 | WREG32(DC_HOT_PLUG_DETECT2_CONTROL, 0)r100_mm_wreg(rdev, (0x7d10), (0), 0); |
1057 | break; |
1058 | case RADEON_HPD_3: |
1059 | WREG32(DC_HOT_PLUG_DETECT3_CONTROL, 0)r100_mm_wreg(rdev, (0x7d24), (0), 0); |
1060 | break; |
1061 | default: |
1062 | break; |
1063 | } |
1064 | } |
1065 | if (radeon_connector->hpd.hpd != RADEON_HPD_NONE) |
1066 | disable |= 1 << radeon_connector->hpd.hpd; |
1067 | } |
1068 | radeon_irq_kms_disable_hpd(rdev, disable); |
1069 | } |
1070 | |
1071 | /* |
1072 | * R600 PCIE GART |
1073 | */ |
1074 | void r600_pcie_gart_tlb_flush(struct radeon_device *rdev) |
1075 | { |
1076 | unsigned i; |
1077 | u32 tmp; |
1078 | |
1079 | /* flush hdp cache so updates hit vram */ |
1080 | if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_RV740) && |
1081 | !(rdev->flags & RADEON_IS_AGP)) { |
1082 | void __iomem *ptr = (void *)rdev->gart.ptr; |
1083 | |
1084 | /* r7xx hw bug. write to HDP_DEBUG1 followed by fb read |
1085 | * rather than write to HDP_REG_COHERENCY_FLUSH_CNTL |
1086 | * This seems to cause problems on some AGP cards. Just use the old |
1087 | * method for them. |
1088 | */ |
1089 | WREG32(HDP_DEBUG1, 0)r100_mm_wreg(rdev, (0x2F34), (0), 0); |
1090 | readl((void __iomem *)ptr)ioread32((void *)ptr); |
1091 | } else |
1092 | WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1)r100_mm_wreg(rdev, (0x5480), (0x1), 0); |
1093 | |
1094 | WREG32(VM_CONTEXT0_INVALIDATION_LOW_ADDR, rdev->mc.gtt_start >> 12)r100_mm_wreg(rdev, (0x1490), (rdev->mc.gtt_start >> 12 ), 0); |
1095 | WREG32(VM_CONTEXT0_INVALIDATION_HIGH_ADDR, (rdev->mc.gtt_end - 1) >> 12)r100_mm_wreg(rdev, (0x14B0), ((rdev->mc.gtt_end - 1) >> 12), 0); |
1096 | WREG32(VM_CONTEXT0_REQUEST_RESPONSE, REQUEST_TYPE(1))r100_mm_wreg(rdev, (0x1470), ((((1) & 0xf) << 0)), 0 ); |
1097 | for (i = 0; i < rdev->usec_timeout; i++) { |
1098 | /* read MC_STATUS */ |
1099 | tmp = RREG32(VM_CONTEXT0_REQUEST_RESPONSE)r100_mm_rreg(rdev, (0x1470), 0); |
1100 | tmp = (tmp & RESPONSE_TYPE_MASK0x000000F0) >> RESPONSE_TYPE_SHIFT4; |
1101 | if (tmp == 2) { |
1102 | pr_warn("[drm] r600 flush TLB failed\n")printk("\0014" "[drm] r600 flush TLB failed\n"); |
1103 | return; |
1104 | } |
1105 | if (tmp) { |
1106 | return; |
1107 | } |
1108 | udelay(1); |
1109 | } |
1110 | } |
1111 | |
1112 | int r600_pcie_gart_init(struct radeon_device *rdev) |
1113 | { |
1114 | int r; |
1115 | |
1116 | if (rdev->gart.robj) { |
1117 | WARN(1, "R600 PCIE GART already initialized\n")({ int __ret = !!(1); if (__ret) printf("R600 PCIE GART already initialized\n" ); __builtin_expect(!!(__ret), 0); }); |
1118 | return 0; |
1119 | } |
1120 | /* Initialize common gart structure */ |
1121 | r = radeon_gart_init(rdev); |
1122 | if (r) |
1123 | return r; |
1124 | rdev->gart.table_size = rdev->gart.num_gpu_pages * 8; |
1125 | return radeon_gart_table_vram_alloc(rdev); |
1126 | } |
1127 | |
1128 | static int r600_pcie_gart_enable(struct radeon_device *rdev) |
1129 | { |
1130 | u32 tmp; |
1131 | int r, i; |
1132 | |
1133 | if (rdev->gart.robj == NULL((void *)0)) { |
1134 | dev_err(rdev->dev, "No VRAM object for PCIE GART.\n")printf("drm:pid%d:%s *ERROR* " "No VRAM object for PCIE GART.\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
1135 | return -EINVAL22; |
1136 | } |
1137 | r = radeon_gart_table_vram_pin(rdev); |
1138 | if (r) |
1139 | return r; |
1140 | |
1141 | /* Setup L2 cache */ |
1142 | WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |r100_mm_wreg(rdev, (0x1400), ((1 << 0) | (1 << 1) | (1 << 9) | (((7) & 7) << 13)), 0) |
1143 | ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |r100_mm_wreg(rdev, (0x1400), ((1 << 0) | (1 << 1) | (1 << 9) | (((7) & 7) << 13)), 0) |
1144 | EFFECTIVE_L2_QUEUE_SIZE(7))r100_mm_wreg(rdev, (0x1400), ((1 << 0) | (1 << 1) | (1 << 9) | (((7) & 7) << 13)), 0); |
1145 | WREG32(VM_L2_CNTL2, 0)r100_mm_wreg(rdev, (0x1404), (0), 0); |
1146 | WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1))r100_mm_wreg(rdev, (0x1408), ((((0) & 0x1f) << 0) | (((1) & 0x1f) << 5)), 0); |
1147 | /* Setup TLB control */ |
1148 | tmp = ENABLE_L1_TLB(1 << 0) | ENABLE_L1_FRAGMENT_PROCESSING(1 << 1) | |
1149 | SYSTEM_ACCESS_MODE_NOT_IN_SYS(3 << 6) | |
1150 | EFFECTIVE_L1_TLB_SIZE(5)(((5) & 7) << 12) | EFFECTIVE_L1_QUEUE_SIZE(5)(((5) & 7) << 15) | |
1151 | ENABLE_WAIT_L2_QUERY(1 << 11); |
1152 | WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp)r100_mm_wreg(rdev, (0x2200), (tmp), 0); |
1153 | WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp)r100_mm_wreg(rdev, (0x2214), (tmp), 0); |
1154 | WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING)r100_mm_wreg(rdev, (0x2204), (tmp | (1 << 2)), 0); |
1155 | WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp)r100_mm_wreg(rdev, (0x2218), (tmp), 0); |
1156 | WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp)r100_mm_wreg(rdev, (0x219C), (tmp), 0); |
1157 | WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp)r100_mm_wreg(rdev, (0x21A4), (tmp), 0); |
1158 | WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp)r100_mm_wreg(rdev, (0x21A0), (tmp), 0); |
1159 | WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp)r100_mm_wreg(rdev, (0x21A8), (tmp), 0); |
1160 | WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp)r100_mm_wreg(rdev, (0x21FC), (tmp), 0); |
1161 | WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp)r100_mm_wreg(rdev, (0x2210), (tmp), 0); |
1162 | WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp)r100_mm_wreg(rdev, (0x2208), (tmp), 0); |
1163 | WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp)r100_mm_wreg(rdev, (0x221C), (tmp), 0); |
1164 | WREG32(MC_VM_L1_TLB_MCB_RD_UVD_CNTL, tmp)r100_mm_wreg(rdev, (0x2124), (tmp), 0); |
1165 | WREG32(MC_VM_L1_TLB_MCB_WR_UVD_CNTL, tmp)r100_mm_wreg(rdev, (0x212c), (tmp), 0); |
1166 | WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE)r100_mm_wreg(rdev, (0x220C), (tmp | (1 << 10)), 0); |
1167 | WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE)r100_mm_wreg(rdev, (0x2220), (tmp | (1 << 10)), 0); |
1168 | WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12)r100_mm_wreg(rdev, (0x1594), (rdev->mc.gtt_start >> 12 ), 0); |
1169 | WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12)r100_mm_wreg(rdev, (0x15B4), (rdev->mc.gtt_end >> 12 ), 0); |
1170 | WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12)r100_mm_wreg(rdev, (0x1574), (rdev->gart.table_addr >> 12), 0); |
1171 | WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |r100_mm_wreg(rdev, (0x1410), ((1 << 0) | (((0) & 3) << 1) | (1 << 4)), 0) |
1172 | RANGE_PROTECTION_FAULT_ENABLE_DEFAULT)r100_mm_wreg(rdev, (0x1410), ((1 << 0) | (((0) & 3) << 1) | (1 << 4)), 0); |
1173 | WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,r100_mm_wreg(rdev, (0x1554), ((u32)(rdev->dummy_page.addr >> 12)), 0) |
1174 | (u32)(rdev->dummy_page.addr >> 12))r100_mm_wreg(rdev, (0x1554), ((u32)(rdev->dummy_page.addr >> 12)), 0); |
1175 | for (i = 1; i < 7; i++) |
1176 | WREG32(VM_CONTEXT0_CNTL + (i * 4), 0)r100_mm_wreg(rdev, (0x1410 + (i * 4)), (0), 0); |
1177 | |
1178 | r600_pcie_gart_tlb_flush(rdev); |
1179 | DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",printk("\0016" "[" "drm" "] " "PCIE GART of %uM enabled (table at 0x%016llX).\n" , (unsigned)(rdev->mc.gtt_size >> 20), (unsigned long long)rdev->gart.table_addr) |
1180 | (unsigned)(rdev->mc.gtt_size >> 20),printk("\0016" "[" "drm" "] " "PCIE GART of %uM enabled (table at 0x%016llX).\n" , (unsigned)(rdev->mc.gtt_size >> 20), (unsigned long long)rdev->gart.table_addr) |
1181 | (unsigned long long)rdev->gart.table_addr)printk("\0016" "[" "drm" "] " "PCIE GART of %uM enabled (table at 0x%016llX).\n" , (unsigned)(rdev->mc.gtt_size >> 20), (unsigned long long)rdev->gart.table_addr); |
1182 | rdev->gart.ready = true1; |
1183 | return 0; |
1184 | } |
1185 | |
1186 | static void r600_pcie_gart_disable(struct radeon_device *rdev) |
1187 | { |
1188 | u32 tmp; |
1189 | int i; |
1190 | |
1191 | /* Disable all tables */ |
1192 | for (i = 0; i < 7; i++) |
1193 | WREG32(VM_CONTEXT0_CNTL + (i * 4), 0)r100_mm_wreg(rdev, (0x1410 + (i * 4)), (0), 0); |
1194 | |
1195 | /* Disable L2 cache */ |
1196 | WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING |r100_mm_wreg(rdev, (0x1400), ((1 << 1) | (((7) & 7) << 13)), 0) |
1197 | EFFECTIVE_L2_QUEUE_SIZE(7))r100_mm_wreg(rdev, (0x1400), ((1 << 1) | (((7) & 7) << 13)), 0); |
1198 | WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1))r100_mm_wreg(rdev, (0x1408), ((((0) & 0x1f) << 0) | (((1) & 0x1f) << 5)), 0); |
1199 | /* Setup L1 TLB control */ |
1200 | tmp = EFFECTIVE_L1_TLB_SIZE(5)(((5) & 7) << 12) | EFFECTIVE_L1_QUEUE_SIZE(5)(((5) & 7) << 15) | |
1201 | ENABLE_WAIT_L2_QUERY(1 << 11); |
1202 | WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp)r100_mm_wreg(rdev, (0x219C), (tmp), 0); |
1203 | WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp)r100_mm_wreg(rdev, (0x21A4), (tmp), 0); |
1204 | WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp)r100_mm_wreg(rdev, (0x21A0), (tmp), 0); |
1205 | WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp)r100_mm_wreg(rdev, (0x21A8), (tmp), 0); |
1206 | WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp)r100_mm_wreg(rdev, (0x21FC), (tmp), 0); |
1207 | WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp)r100_mm_wreg(rdev, (0x2210), (tmp), 0); |
1208 | WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp)r100_mm_wreg(rdev, (0x2208), (tmp), 0); |
1209 | WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp)r100_mm_wreg(rdev, (0x221C), (tmp), 0); |
1210 | WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp)r100_mm_wreg(rdev, (0x220C), (tmp), 0); |
1211 | WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp)r100_mm_wreg(rdev, (0x2220), (tmp), 0); |
1212 | WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp)r100_mm_wreg(rdev, (0x2200), (tmp), 0); |
1213 | WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp)r100_mm_wreg(rdev, (0x2214), (tmp), 0); |
1214 | WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp)r100_mm_wreg(rdev, (0x2204), (tmp), 0); |
1215 | WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp)r100_mm_wreg(rdev, (0x2218), (tmp), 0); |
1216 | WREG32(MC_VM_L1_TLB_MCB_RD_UVD_CNTL, tmp)r100_mm_wreg(rdev, (0x2124), (tmp), 0); |
1217 | WREG32(MC_VM_L1_TLB_MCB_WR_UVD_CNTL, tmp)r100_mm_wreg(rdev, (0x212c), (tmp), 0); |
1218 | radeon_gart_table_vram_unpin(rdev); |
1219 | } |
1220 | |
1221 | static void r600_pcie_gart_fini(struct radeon_device *rdev) |
1222 | { |
1223 | radeon_gart_fini(rdev); |
1224 | r600_pcie_gart_disable(rdev); |
1225 | radeon_gart_table_vram_free(rdev); |
1226 | } |
1227 | |
1228 | static void r600_agp_enable(struct radeon_device *rdev) |
1229 | { |
1230 | u32 tmp; |
1231 | int i; |
1232 | |
1233 | /* Setup L2 cache */ |
1234 | WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |r100_mm_wreg(rdev, (0x1400), ((1 << 0) | (1 << 1) | (1 << 9) | (((7) & 7) << 13)), 0) |
1235 | ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |r100_mm_wreg(rdev, (0x1400), ((1 << 0) | (1 << 1) | (1 << 9) | (((7) & 7) << 13)), 0) |
1236 | EFFECTIVE_L2_QUEUE_SIZE(7))r100_mm_wreg(rdev, (0x1400), ((1 << 0) | (1 << 1) | (1 << 9) | (((7) & 7) << 13)), 0); |
1237 | WREG32(VM_L2_CNTL2, 0)r100_mm_wreg(rdev, (0x1404), (0), 0); |
1238 | WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1))r100_mm_wreg(rdev, (0x1408), ((((0) & 0x1f) << 0) | (((1) & 0x1f) << 5)), 0); |
1239 | /* Setup TLB control */ |
1240 | tmp = ENABLE_L1_TLB(1 << 0) | ENABLE_L1_FRAGMENT_PROCESSING(1 << 1) | |
1241 | SYSTEM_ACCESS_MODE_NOT_IN_SYS(3 << 6) | |
1242 | EFFECTIVE_L1_TLB_SIZE(5)(((5) & 7) << 12) | EFFECTIVE_L1_QUEUE_SIZE(5)(((5) & 7) << 15) | |
1243 | ENABLE_WAIT_L2_QUERY(1 << 11); |
1244 | WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp)r100_mm_wreg(rdev, (0x2200), (tmp), 0); |
1245 | WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp)r100_mm_wreg(rdev, (0x2214), (tmp), 0); |
1246 | WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING)r100_mm_wreg(rdev, (0x2204), (tmp | (1 << 2)), 0); |
1247 | WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp)r100_mm_wreg(rdev, (0x2218), (tmp), 0); |
1248 | WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp)r100_mm_wreg(rdev, (0x219C), (tmp), 0); |
1249 | WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp)r100_mm_wreg(rdev, (0x21A4), (tmp), 0); |
1250 | WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp)r100_mm_wreg(rdev, (0x21A0), (tmp), 0); |
1251 | WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp)r100_mm_wreg(rdev, (0x21A8), (tmp), 0); |
1252 | WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp)r100_mm_wreg(rdev, (0x21FC), (tmp), 0); |
1253 | WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp)r100_mm_wreg(rdev, (0x2210), (tmp), 0); |
1254 | WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp)r100_mm_wreg(rdev, (0x2208), (tmp), 0); |
1255 | WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp)r100_mm_wreg(rdev, (0x221C), (tmp), 0); |
1256 | WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE)r100_mm_wreg(rdev, (0x220C), (tmp | (1 << 10)), 0); |
1257 | WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE)r100_mm_wreg(rdev, (0x2220), (tmp | (1 << 10)), 0); |
1258 | for (i = 0; i < 7; i++) |
1259 | WREG32(VM_CONTEXT0_CNTL + (i * 4), 0)r100_mm_wreg(rdev, (0x1410 + (i * 4)), (0), 0); |
1260 | } |
1261 | |
1262 | int r600_mc_wait_for_idle(struct radeon_device *rdev) |
1263 | { |
1264 | unsigned i; |
1265 | u32 tmp; |
1266 | |
1267 | for (i = 0; i < rdev->usec_timeout; i++) { |
1268 | /* read MC_STATUS */ |
1269 | tmp = RREG32(R_000E50_SRBM_STATUS)r100_mm_rreg(rdev, (0x0E50), 0) & 0x3F00; |
1270 | if (!tmp) |
1271 | return 0; |
1272 | udelay(1); |
1273 | } |
1274 | return -1; |
1275 | } |
1276 | |
1277 | uint32_t rs780_mc_rreg(struct radeon_device *rdev, uint32_t reg) |
1278 | { |
1279 | unsigned long flags; |
1280 | uint32_t r; |
1281 | |
1282 | spin_lock_irqsave(&rdev->mc_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->mc_idx_lock); } while (0); |
1283 | WREG32(R_0028F8_MC_INDEX, S_0028F8_MC_IND_ADDR(reg))r100_mm_wreg(rdev, (0x28F8), ((((reg) & 0x1FF) << 0 )), 0); |
1284 | r = RREG32(R_0028FC_MC_DATA)r100_mm_rreg(rdev, (0x28FC), 0); |
1285 | WREG32(R_0028F8_MC_INDEX, ~C_0028F8_MC_IND_ADDR)r100_mm_wreg(rdev, (0x28F8), (~0xFFFFFE00), 0); |
1286 | spin_unlock_irqrestore(&rdev->mc_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->mc_idx_lock); } while (0); |
1287 | return r; |
1288 | } |
1289 | |
1290 | void rs780_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v) |
1291 | { |
1292 | unsigned long flags; |
1293 | |
1294 | spin_lock_irqsave(&rdev->mc_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->mc_idx_lock); } while (0); |
1295 | WREG32(R_0028F8_MC_INDEX, S_0028F8_MC_IND_ADDR(reg) |r100_mm_wreg(rdev, (0x28F8), ((((reg) & 0x1FF) << 0 ) | (((1) & 0x1) << 9)), 0) |
1296 | S_0028F8_MC_IND_WR_EN(1))r100_mm_wreg(rdev, (0x28F8), ((((reg) & 0x1FF) << 0 ) | (((1) & 0x1) << 9)), 0); |
1297 | WREG32(R_0028FC_MC_DATA, v)r100_mm_wreg(rdev, (0x28FC), (v), 0); |
1298 | WREG32(R_0028F8_MC_INDEX, 0x7F)r100_mm_wreg(rdev, (0x28F8), (0x7F), 0); |
1299 | spin_unlock_irqrestore(&rdev->mc_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->mc_idx_lock); } while (0); |
1300 | } |
1301 | |
1302 | static void r600_mc_program(struct radeon_device *rdev) |
1303 | { |
1304 | struct rv515_mc_save save; |
1305 | u32 tmp; |
1306 | int i, j; |
1307 | |
1308 | /* Initialize HDP */ |
1309 | for (i = 0, j = 0; i < 32; i++, j += 0x18) { |
1310 | WREG32((0x2c14 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c14 + j)), (0x00000000), 0); |
1311 | WREG32((0x2c18 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c18 + j)), (0x00000000), 0); |
1312 | WREG32((0x2c1c + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c1c + j)), (0x00000000), 0); |
1313 | WREG32((0x2c20 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c20 + j)), (0x00000000), 0); |
1314 | WREG32((0x2c24 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c24 + j)), (0x00000000), 0); |
1315 | } |
1316 | WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0)r100_mm_wreg(rdev, (0x54A0), (0), 0); |
1317 | |
1318 | rv515_mc_stop(rdev, &save); |
1319 | if (r600_mc_wait_for_idle(rdev)) { |
1320 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n")printf("drm:pid%d:%s *WARNING* " "Wait for MC idle timedout !\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
1321 | } |
1322 | /* Lockout access through VGA aperture (doesn't exist before R600) */ |
1323 | WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE)r100_mm_wreg(rdev, (0x328), ((1 << 4)), 0); |
1324 | /* Update configuration */ |
1325 | if (rdev->flags & RADEON_IS_AGP) { |
1326 | if (rdev->mc.vram_start < rdev->mc.gtt_start) { |
1327 | /* VRAM before AGP */ |
1328 | WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,r100_mm_wreg(rdev, (0x2190), (rdev->mc.vram_start >> 12), 0) |
1329 | rdev->mc.vram_start >> 12)r100_mm_wreg(rdev, (0x2190), (rdev->mc.vram_start >> 12), 0); |
1330 | WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,r100_mm_wreg(rdev, (0x2194), (rdev->mc.gtt_end >> 12 ), 0) |
1331 | rdev->mc.gtt_end >> 12)r100_mm_wreg(rdev, (0x2194), (rdev->mc.gtt_end >> 12 ), 0); |
1332 | } else { |
1333 | /* VRAM after AGP */ |
1334 | WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,r100_mm_wreg(rdev, (0x2190), (rdev->mc.gtt_start >> 12 ), 0) |
1335 | rdev->mc.gtt_start >> 12)r100_mm_wreg(rdev, (0x2190), (rdev->mc.gtt_start >> 12 ), 0); |
1336 | WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,r100_mm_wreg(rdev, (0x2194), (rdev->mc.vram_end >> 12 ), 0) |
1337 | rdev->mc.vram_end >> 12)r100_mm_wreg(rdev, (0x2194), (rdev->mc.vram_end >> 12 ), 0); |
1338 | } |
1339 | } else { |
1340 | WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.vram_start >> 12)r100_mm_wreg(rdev, (0x2190), (rdev->mc.vram_start >> 12), 0); |
1341 | WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, rdev->mc.vram_end >> 12)r100_mm_wreg(rdev, (0x2194), (rdev->mc.vram_end >> 12 ), 0); |
1342 | } |
1343 | WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, rdev->vram_scratch.gpu_addr >> 12)r100_mm_wreg(rdev, (0x2198), (rdev->vram_scratch.gpu_addr >> 12), 0); |
1344 | tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16; |
1345 | tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF); |
1346 | WREG32(MC_VM_FB_LOCATION, tmp)r100_mm_wreg(rdev, (0x2180), (tmp), 0); |
1347 | WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8))r100_mm_wreg(rdev, (0x2C04), ((rdev->mc.vram_start >> 8)), 0); |
1348 | WREG32(HDP_NONSURFACE_INFO, (2 << 7))r100_mm_wreg(rdev, (0x2C08), ((2 << 7)), 0); |
1349 | WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF)r100_mm_wreg(rdev, (0x2C0C), (0x3FFFFFFF), 0); |
1350 | if (rdev->flags & RADEON_IS_AGP) { |
1351 | WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 22)r100_mm_wreg(rdev, (0x2184), (rdev->mc.gtt_end >> 22 ), 0); |
1352 | WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 22)r100_mm_wreg(rdev, (0x2188), (rdev->mc.gtt_start >> 22 ), 0); |
1353 | WREG32(MC_VM_AGP_BASE, rdev->mc.agp_base >> 22)r100_mm_wreg(rdev, (0x218C), (rdev->mc.agp_base >> 22 ), 0); |
1354 | } else { |
1355 | WREG32(MC_VM_AGP_BASE, 0)r100_mm_wreg(rdev, (0x218C), (0), 0); |
1356 | WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF)r100_mm_wreg(rdev, (0x2184), (0x0FFFFFFF), 0); |
1357 | WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF)r100_mm_wreg(rdev, (0x2188), (0x0FFFFFFF), 0); |
1358 | } |
1359 | if (r600_mc_wait_for_idle(rdev)) { |
1360 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n")printf("drm:pid%d:%s *WARNING* " "Wait for MC idle timedout !\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
1361 | } |
1362 | rv515_mc_resume(rdev, &save); |
1363 | /* we need to own VRAM, so turn off the VGA renderer here |
1364 | * to stop it overwriting our objects */ |
1365 | rv515_vga_render_disable(rdev); |
1366 | } |
1367 | |
1368 | /** |
1369 | * r600_vram_gtt_location - try to find VRAM & GTT location |
1370 | * @rdev: radeon device structure holding all necessary informations |
1371 | * @mc: memory controller structure holding memory informations |
1372 | * |
1373 | * Function will place try to place VRAM at same place as in CPU (PCI) |
1374 | * address space as some GPU seems to have issue when we reprogram at |
1375 | * different address space. |
1376 | * |
1377 | * If there is not enough space to fit the unvisible VRAM after the |
1378 | * aperture then we limit the VRAM size to the aperture. |
1379 | * |
1380 | * If we are using AGP then place VRAM adjacent to AGP aperture are we need |
1381 | * them to be in one from GPU point of view so that we can program GPU to |
1382 | * catch access outside them (weird GPU policy see ??). |
1383 | * |
1384 | * This function will never fails, worst case are limiting VRAM or GTT. |
1385 | * |
1386 | * Note: GTT start, end, size should be initialized before calling this |
1387 | * function on AGP platform. |
1388 | */ |
1389 | static void r600_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc) |
1390 | { |
1391 | u64 size_bf, size_af; |
1392 | |
1393 | if (mc->mc_vram_size > 0xE0000000) { |
1394 | /* leave room for at least 512M GTT */ |
1395 | dev_warn(rdev->dev, "limiting VRAM\n")printf("drm:pid%d:%s *WARNING* " "limiting VRAM\n", ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof (struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p-> ps_pid, __func__); |
1396 | mc->real_vram_size = 0xE0000000; |
1397 | mc->mc_vram_size = 0xE0000000; |
1398 | } |
1399 | if (rdev->flags & RADEON_IS_AGP) { |
1400 | size_bf = mc->gtt_start; |
1401 | size_af = mc->mc_mask - mc->gtt_end; |
1402 | if (size_bf > size_af) { |
1403 | if (mc->mc_vram_size > size_bf) { |
1404 | dev_warn(rdev->dev, "limiting VRAM\n")printf("drm:pid%d:%s *WARNING* " "limiting VRAM\n", ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof (struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p-> ps_pid, __func__); |
1405 | mc->real_vram_size = size_bf; |
1406 | mc->mc_vram_size = size_bf; |
1407 | } |
1408 | mc->vram_start = mc->gtt_start - mc->mc_vram_size; |
1409 | } else { |
1410 | if (mc->mc_vram_size > size_af) { |
1411 | dev_warn(rdev->dev, "limiting VRAM\n")printf("drm:pid%d:%s *WARNING* " "limiting VRAM\n", ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof (struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p-> ps_pid, __func__); |
1412 | mc->real_vram_size = size_af; |
1413 | mc->mc_vram_size = size_af; |
1414 | } |
1415 | mc->vram_start = mc->gtt_end + 1; |
1416 | } |
1417 | mc->vram_end = mc->vram_start + mc->mc_vram_size - 1; |
1418 | dev_info(rdev->dev, "VRAM: %lluM 0x%08llX - 0x%08llX (%lluM used)\n",do { } while(0) |
1419 | mc->mc_vram_size >> 20, mc->vram_start,do { } while(0) |
1420 | mc->vram_end, mc->real_vram_size >> 20)do { } while(0); |
1421 | } else { |
1422 | u64 base = 0; |
1423 | if (rdev->flags & RADEON_IS_IGP) { |
1424 | base = RREG32(MC_VM_FB_LOCATION)r100_mm_rreg(rdev, (0x2180), 0) & 0xFFFF; |
1425 | base <<= 24; |
1426 | } |
1427 | radeon_vram_location(rdev, &rdev->mc, base); |
1428 | rdev->mc.gtt_base_align = 0; |
1429 | radeon_gtt_location(rdev, mc); |
1430 | } |
1431 | } |
1432 | |
1433 | static int r600_mc_init(struct radeon_device *rdev) |
1434 | { |
1435 | u32 tmp; |
1436 | int chansize, numchan; |
1437 | uint32_t h_addr, l_addr; |
1438 | unsigned long long k8_addr; |
1439 | |
1440 | /* Get VRAM informations */ |
1441 | rdev->mc.vram_is_ddr = true1; |
1442 | tmp = RREG32(RAMCFG)r100_mm_rreg(rdev, (0x2408), 0); |
1443 | if (tmp & CHANSIZE_OVERRIDE(1 << 10)) { |
1444 | chansize = 16; |
1445 | } else if (tmp & CHANSIZE_MASK0x00000080) { |
1446 | chansize = 64; |
1447 | } else { |
1448 | chansize = 32; |
1449 | } |
1450 | tmp = RREG32(CHMAP)r100_mm_rreg(rdev, (0x2004), 0); |
1451 | switch ((tmp & NOOFCHAN_MASK0x00003000) >> NOOFCHAN_SHIFT12) { |
1452 | case 0: |
1453 | default: |
1454 | numchan = 1; |
1455 | break; |
1456 | case 1: |
1457 | numchan = 2; |
1458 | break; |
1459 | case 2: |
1460 | numchan = 4; |
1461 | break; |
1462 | case 3: |
1463 | numchan = 8; |
1464 | break; |
1465 | } |
1466 | rdev->mc.vram_width = numchan * chansize; |
1467 | /* Could aper size report 0 ? */ |
1468 | rdev->mc.aper_base = rdev->fb_aper_offset; |
1469 | rdev->mc.aper_size = rdev->fb_aper_size; |
1470 | /* Setup GPU memory space */ |
1471 | rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE)r100_mm_rreg(rdev, (0x5428), 0); |
1472 | rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE)r100_mm_rreg(rdev, (0x5428), 0); |
1473 | rdev->mc.visible_vram_size = rdev->mc.aper_size; |
1474 | r600_vram_gtt_location(rdev, &rdev->mc); |
1475 | |
1476 | if (rdev->flags & RADEON_IS_IGP) { |
1477 | rs690_pm_info(rdev); |
1478 | rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev); |
1479 | |
1480 | if (rdev->family == CHIP_RS780 || rdev->family == CHIP_RS880) { |
1481 | /* Use K8 direct mapping for fast fb access. */ |
1482 | rdev->fastfb_working = false0; |
1483 | h_addr = G_000012_K8_ADDR_EXT(RREG32_MC(R_000012_MC_MISC_UMA_CNTL))(((rdev->mc_rreg(rdev, (0x12))) >> 0) & 0xFF); |
1484 | l_addr = RREG32_MC(R_000011_K8_FB_LOCATION)rdev->mc_rreg(rdev, (0x11)); |
1485 | k8_addr = ((unsigned long long)h_addr) << 32 | l_addr; |
1486 | #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE) |
1487 | if (k8_addr + rdev->mc.visible_vram_size < 0x100000000ULL) |
1488 | #endif |
1489 | { |
1490 | /* FastFB shall be used with UMA memory. Here it is simply disabled when sideport |
1491 | * memory is present. |
1492 | */ |
1493 | if (rdev->mc.igp_sideport_enabled == false0 && radeon_fastfb == 1) { |
1494 | DRM_INFO("Direct mapping: aper base at 0x%llx, replaced by direct mapping base 0x%llx.\n",printk("\0016" "[" "drm" "] " "Direct mapping: aper base at 0x%llx, replaced by direct mapping base 0x%llx.\n" , (unsigned long long)rdev->mc.aper_base, k8_addr) |
1495 | (unsigned long long)rdev->mc.aper_base, k8_addr)printk("\0016" "[" "drm" "] " "Direct mapping: aper base at 0x%llx, replaced by direct mapping base 0x%llx.\n" , (unsigned long long)rdev->mc.aper_base, k8_addr); |
1496 | rdev->mc.aper_base = (resource_size_t)k8_addr; |
1497 | rdev->fastfb_working = true1; |
1498 | } |
1499 | } |
1500 | } |
1501 | } |
1502 | |
1503 | radeon_update_bandwidth_info(rdev); |
1504 | return 0; |
1505 | } |
1506 | |
1507 | int r600_vram_scratch_init(struct radeon_device *rdev) |
1508 | { |
1509 | int r; |
1510 | |
1511 | if (rdev->vram_scratch.robj == NULL((void *)0)) { |
1512 | r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE4096, |
1513 | PAGE_SIZE(1 << 12), true1, RADEON_GEM_DOMAIN_VRAM0x4, |
1514 | 0, NULL((void *)0), NULL((void *)0), &rdev->vram_scratch.robj); |
1515 | if (r) { |
1516 | return r; |
1517 | } |
1518 | } |
1519 | |
1520 | r = radeon_bo_reserve(rdev->vram_scratch.robj, false0); |
1521 | if (unlikely(r != 0)__builtin_expect(!!(r != 0), 0)) |
1522 | return r; |
1523 | r = radeon_bo_pin(rdev->vram_scratch.robj, |
1524 | RADEON_GEM_DOMAIN_VRAM0x4, &rdev->vram_scratch.gpu_addr); |
1525 | if (r) { |
1526 | radeon_bo_unreserve(rdev->vram_scratch.robj); |
1527 | return r; |
1528 | } |
1529 | r = radeon_bo_kmap(rdev->vram_scratch.robj, |
1530 | (void **)&rdev->vram_scratch.ptr); |
1531 | if (r) |
1532 | radeon_bo_unpin(rdev->vram_scratch.robj); |
1533 | radeon_bo_unreserve(rdev->vram_scratch.robj); |
1534 | |
1535 | return r; |
1536 | } |
1537 | |
1538 | void r600_vram_scratch_fini(struct radeon_device *rdev) |
1539 | { |
1540 | int r; |
1541 | |
1542 | if (rdev->vram_scratch.robj == NULL((void *)0)) { |
1543 | return; |
1544 | } |
1545 | r = radeon_bo_reserve(rdev->vram_scratch.robj, false0); |
1546 | if (likely(r == 0)__builtin_expect(!!(r == 0), 1)) { |
1547 | radeon_bo_kunmap(rdev->vram_scratch.robj); |
1548 | radeon_bo_unpin(rdev->vram_scratch.robj); |
1549 | radeon_bo_unreserve(rdev->vram_scratch.robj); |
1550 | } |
1551 | radeon_bo_unref(&rdev->vram_scratch.robj); |
1552 | } |
1553 | |
1554 | void r600_set_bios_scratch_engine_hung(struct radeon_device *rdev, bool_Bool hung) |
1555 | { |
1556 | u32 tmp = RREG32(R600_BIOS_3_SCRATCH)r100_mm_rreg(rdev, (0x1730), 0); |
1557 | |
1558 | if (hung) |
1559 | tmp |= ATOM_S3_ASIC_GUI_ENGINE_HUNG0x20000000L; |
1560 | else |
1561 | tmp &= ~ATOM_S3_ASIC_GUI_ENGINE_HUNG0x20000000L; |
1562 | |
1563 | WREG32(R600_BIOS_3_SCRATCH, tmp)r100_mm_wreg(rdev, (0x1730), (tmp), 0); |
1564 | } |
1565 | |
1566 | static void r600_print_gpu_status_regs(struct radeon_device *rdev) |
1567 | { |
1568 | dev_info(rdev->dev, " R_008010_GRBM_STATUS = 0x%08X\n",do { } while(0) |
1569 | RREG32(R_008010_GRBM_STATUS))do { } while(0); |
1570 | dev_info(rdev->dev, " R_008014_GRBM_STATUS2 = 0x%08X\n",do { } while(0) |
1571 | RREG32(R_008014_GRBM_STATUS2))do { } while(0); |
1572 | dev_info(rdev->dev, " R_000E50_SRBM_STATUS = 0x%08X\n",do { } while(0) |
1573 | RREG32(R_000E50_SRBM_STATUS))do { } while(0); |
1574 | dev_info(rdev->dev, " R_008674_CP_STALLED_STAT1 = 0x%08X\n",do { } while(0) |
1575 | RREG32(CP_STALLED_STAT1))do { } while(0); |
1576 | dev_info(rdev->dev, " R_008678_CP_STALLED_STAT2 = 0x%08X\n",do { } while(0) |
1577 | RREG32(CP_STALLED_STAT2))do { } while(0); |
1578 | dev_info(rdev->dev, " R_00867C_CP_BUSY_STAT = 0x%08X\n",do { } while(0) |
1579 | RREG32(CP_BUSY_STAT))do { } while(0); |
1580 | dev_info(rdev->dev, " R_008680_CP_STAT = 0x%08X\n",do { } while(0) |
1581 | RREG32(CP_STAT))do { } while(0); |
1582 | dev_info(rdev->dev, " R_00D034_DMA_STATUS_REG = 0x%08X\n",do { } while(0) |
1583 | RREG32(DMA_STATUS_REG))do { } while(0); |
1584 | } |
1585 | |
1586 | static bool_Bool r600_is_display_hung(struct radeon_device *rdev) |
1587 | { |
1588 | u32 crtc_hung = 0; |
1589 | u32 crtc_status[2]; |
1590 | u32 i, j, tmp; |
1591 | |
1592 | for (i = 0; i < rdev->num_crtc; i++) { |
1593 | if (RREG32(AVIVO_D1CRTC_CONTROL + crtc_offsets[i])r100_mm_rreg(rdev, (0x6080 + crtc_offsets[i]), 0) & AVIVO_CRTC_EN(1 << 0)) { |
1594 | crtc_status[i] = RREG32(AVIVO_D1CRTC_STATUS_HV_COUNT + crtc_offsets[i])r100_mm_rreg(rdev, (0x60ac + crtc_offsets[i]), 0); |
1595 | crtc_hung |= (1 << i); |
1596 | } |
1597 | } |
1598 | |
1599 | for (j = 0; j < 10; j++) { |
1600 | for (i = 0; i < rdev->num_crtc; i++) { |
1601 | if (crtc_hung & (1 << i)) { |
1602 | tmp = RREG32(AVIVO_D1CRTC_STATUS_HV_COUNT + crtc_offsets[i])r100_mm_rreg(rdev, (0x60ac + crtc_offsets[i]), 0); |
1603 | if (tmp != crtc_status[i]) |
1604 | crtc_hung &= ~(1 << i); |
1605 | } |
1606 | } |
1607 | if (crtc_hung == 0) |
1608 | return false0; |
1609 | udelay(100); |
1610 | } |
1611 | |
1612 | return true1; |
1613 | } |
1614 | |
1615 | u32 r600_gpu_check_soft_reset(struct radeon_device *rdev) |
1616 | { |
1617 | u32 reset_mask = 0; |
1618 | u32 tmp; |
1619 | |
1620 | /* GRBM_STATUS */ |
1621 | tmp = RREG32(R_008010_GRBM_STATUS)r100_mm_rreg(rdev, (0x8010), 0); |
1622 | if (rdev->family >= CHIP_RV770) { |
1623 | if (G_008010_PA_BUSY(tmp)(((tmp) >> 25) & 1) | G_008010_SC_BUSY(tmp)(((tmp) >> 24) & 1) | |
1624 | G_008010_SH_BUSY(tmp)(((tmp) >> 21) & 1) | G_008010_SX_BUSY(tmp)(((tmp) >> 20) & 1) | |
1625 | G_008010_TA_BUSY(tmp)(((tmp) >> 14) & 1) | G_008010_VGT_BUSY(tmp)(((tmp) >> 17) & 1) | |
1626 | G_008010_DB03_BUSY(tmp)(((tmp) >> 26) & 1) | G_008010_CB03_BUSY(tmp)(((tmp) >> 30) & 1) | |
1627 | G_008010_SPI03_BUSY(tmp)(((tmp) >> 22) & 1) | G_008010_VGT_BUSY_NO_DMA(tmp)(((tmp) >> 16) & 1)) |
1628 | reset_mask |= RADEON_RESET_GFX(1 << 0); |
1629 | } else { |
1630 | if (G_008010_PA_BUSY(tmp)(((tmp) >> 25) & 1) | G_008010_SC_BUSY(tmp)(((tmp) >> 24) & 1) | |
1631 | G_008010_SH_BUSY(tmp)(((tmp) >> 21) & 1) | G_008010_SX_BUSY(tmp)(((tmp) >> 20) & 1) | |
1632 | G_008010_TA03_BUSY(tmp)(((tmp) >> 18) & 1) | G_008010_VGT_BUSY(tmp)(((tmp) >> 17) & 1) | |
1633 | G_008010_DB03_BUSY(tmp)(((tmp) >> 26) & 1) | G_008010_CB03_BUSY(tmp)(((tmp) >> 30) & 1) | |
1634 | G_008010_SPI03_BUSY(tmp)(((tmp) >> 22) & 1) | G_008010_VGT_BUSY_NO_DMA(tmp)(((tmp) >> 16) & 1)) |
1635 | reset_mask |= RADEON_RESET_GFX(1 << 0); |
1636 | } |
1637 | |
1638 | if (G_008010_CF_RQ_PENDING(tmp)(((tmp) >> 7) & 1) | G_008010_PF_RQ_PENDING(tmp)(((tmp) >> 8) & 1) | |
1639 | G_008010_CP_BUSY(tmp)(((tmp) >> 29) & 1) | G_008010_CP_COHERENCY_BUSY(tmp)(((tmp) >> 28) & 1)) |
1640 | reset_mask |= RADEON_RESET_CP(1 << 3); |
1641 | |
1642 | if (G_008010_GRBM_EE_BUSY(tmp)(((tmp) >> 10) & 1)) |
1643 | reset_mask |= RADEON_RESET_GRBM(1 << 4) | RADEON_RESET_GFX(1 << 0) | RADEON_RESET_CP(1 << 3); |
1644 | |
1645 | /* DMA_STATUS_REG */ |
1646 | tmp = RREG32(DMA_STATUS_REG)r100_mm_rreg(rdev, (0xd034), 0); |
1647 | if (!(tmp & DMA_IDLE(1 << 0))) |
1648 | reset_mask |= RADEON_RESET_DMA(1 << 2); |
1649 | |
1650 | /* SRBM_STATUS */ |
1651 | tmp = RREG32(R_000E50_SRBM_STATUS)r100_mm_rreg(rdev, (0x0E50), 0); |
1652 | if (G_000E50_RLC_RQ_PENDING(tmp)(((tmp) >> 3) & 1) | G_000E50_RLC_BUSY(tmp)(((tmp) >> 15) & 1)) |
1653 | reset_mask |= RADEON_RESET_RLC(1 << 6); |
1654 | |
1655 | if (G_000E50_IH_BUSY(tmp)(((tmp) >> 17) & 1)) |
1656 | reset_mask |= RADEON_RESET_IH(1 << 8); |
1657 | |
1658 | if (G_000E50_SEM_BUSY(tmp)(((tmp) >> 14) & 1)) |
1659 | reset_mask |= RADEON_RESET_SEM(1 << 7); |
1660 | |
1661 | if (G_000E50_GRBM_RQ_PENDING(tmp)(((tmp) >> 5) & 1)) |
1662 | reset_mask |= RADEON_RESET_GRBM(1 << 4); |
1663 | |
1664 | if (G_000E50_VMC_BUSY(tmp)(((tmp) >> 8) & 1)) |
1665 | reset_mask |= RADEON_RESET_VMC(1 << 9); |
1666 | |
1667 | if (G_000E50_MCB_BUSY(tmp)(((tmp) >> 9) & 1) | G_000E50_MCDZ_BUSY(tmp)(((tmp) >> 10) & 1) | |
1668 | G_000E50_MCDY_BUSY(tmp)(((tmp) >> 11) & 1) | G_000E50_MCDX_BUSY(tmp)(((tmp) >> 12) & 1) | |
1669 | G_000E50_MCDW_BUSY(tmp)(((tmp) >> 13) & 1)) |
1670 | reset_mask |= RADEON_RESET_MC(1 << 10); |
1671 | |
1672 | if (r600_is_display_hung(rdev)) |
1673 | reset_mask |= RADEON_RESET_DISPLAY(1 << 11); |
1674 | |
1675 | /* Skip MC reset as it's mostly likely not hung, just busy */ |
1676 | if (reset_mask & RADEON_RESET_MC(1 << 10)) { |
1677 | DRM_DEBUG("MC busy: 0x%08X, clearing.\n", reset_mask)___drm_dbg(((void *)0), DRM_UT_CORE, "MC busy: 0x%08X, clearing.\n" , reset_mask); |
1678 | reset_mask &= ~RADEON_RESET_MC(1 << 10); |
1679 | } |
1680 | |
1681 | return reset_mask; |
1682 | } |
1683 | |
1684 | static void r600_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask) |
1685 | { |
1686 | struct rv515_mc_save save; |
1687 | u32 grbm_soft_reset = 0, srbm_soft_reset = 0; |
1688 | u32 tmp; |
1689 | |
1690 | if (reset_mask == 0) |
1691 | return; |
1692 | |
1693 | dev_info(rdev->dev, "GPU softreset: 0x%08X\n", reset_mask)do { } while(0); |
1694 | |
1695 | r600_print_gpu_status_regs(rdev); |
1696 | |
1697 | /* Disable CP parsing/prefetching */ |
1698 | if (rdev->family >= CHIP_RV770) |
1699 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1) | S_0086D8_CP_PFP_HALT(1))r100_mm_wreg(rdev, (0x86D8), ((((1) & 1)<<28) | ((( 1) & 1)<<26)), 0); |
1700 | else |
1701 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1))r100_mm_wreg(rdev, (0x86D8), ((((1) & 1)<<28)), 0); |
1702 | |
1703 | /* disable the RLC */ |
1704 | WREG32(RLC_CNTL, 0)r100_mm_wreg(rdev, (0x3f00), (0), 0); |
1705 | |
1706 | if (reset_mask & RADEON_RESET_DMA(1 << 2)) { |
1707 | /* Disable DMA */ |
1708 | tmp = RREG32(DMA_RB_CNTL)r100_mm_rreg(rdev, (0xd000), 0); |
1709 | tmp &= ~DMA_RB_ENABLE(1 << 0); |
1710 | WREG32(DMA_RB_CNTL, tmp)r100_mm_wreg(rdev, (0xd000), (tmp), 0); |
1711 | } |
1712 | |
1713 | mdelay(50); |
1714 | |
1715 | rv515_mc_stop(rdev, &save); |
1716 | if (r600_mc_wait_for_idle(rdev)) { |
1717 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n")printf("drm:pid%d:%s *WARNING* " "Wait for MC idle timedout !\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
1718 | } |
1719 | |
1720 | if (reset_mask & (RADEON_RESET_GFX(1 << 0) | RADEON_RESET_COMPUTE(1 << 1))) { |
1721 | if (rdev->family >= CHIP_RV770) |
1722 | grbm_soft_reset |= S_008020_SOFT_RESET_DB(1)(((1) & 1) << 3) | |
1723 | S_008020_SOFT_RESET_CB(1)(((1) & 1) << 1) | |
1724 | S_008020_SOFT_RESET_PA(1)(((1) & 1) << 5) | |
1725 | S_008020_SOFT_RESET_SC(1)(((1) & 1) << 6) | |
1726 | S_008020_SOFT_RESET_SPI(1)(((1) & 1) << 8) | |
1727 | S_008020_SOFT_RESET_SX(1)(((1) & 1) << 10) | |
1728 | S_008020_SOFT_RESET_SH(1)(((1) & 1) << 9) | |
1729 | S_008020_SOFT_RESET_TC(1)(((1) & 1) << 11) | |
1730 | S_008020_SOFT_RESET_TA(1)(((1) & 1) << 12) | |
1731 | S_008020_SOFT_RESET_VC(1)(((1) & 1) << 13) | |
1732 | S_008020_SOFT_RESET_VGT(1)(((1) & 1) << 14); |
1733 | else |
1734 | grbm_soft_reset |= S_008020_SOFT_RESET_CR(1)(((1) & 1) << 2) | |
1735 | S_008020_SOFT_RESET_DB(1)(((1) & 1) << 3) | |
1736 | S_008020_SOFT_RESET_CB(1)(((1) & 1) << 1) | |
1737 | S_008020_SOFT_RESET_PA(1)(((1) & 1) << 5) | |
1738 | S_008020_SOFT_RESET_SC(1)(((1) & 1) << 6) | |
1739 | S_008020_SOFT_RESET_SMX(1)(((1) & 1) << 7) | |
1740 | S_008020_SOFT_RESET_SPI(1)(((1) & 1) << 8) | |
1741 | S_008020_SOFT_RESET_SX(1)(((1) & 1) << 10) | |
1742 | S_008020_SOFT_RESET_SH(1)(((1) & 1) << 9) | |
1743 | S_008020_SOFT_RESET_TC(1)(((1) & 1) << 11) | |
1744 | S_008020_SOFT_RESET_TA(1)(((1) & 1) << 12) | |
1745 | S_008020_SOFT_RESET_VC(1)(((1) & 1) << 13) | |
1746 | S_008020_SOFT_RESET_VGT(1)(((1) & 1) << 14); |
1747 | } |
1748 | |
1749 | if (reset_mask & RADEON_RESET_CP(1 << 3)) { |
1750 | grbm_soft_reset |= S_008020_SOFT_RESET_CP(1)(((1) & 1) << 0) | |
1751 | S_008020_SOFT_RESET_VGT(1)(((1) & 1) << 14); |
1752 | |
1753 | srbm_soft_reset |= S_000E60_SOFT_RESET_GRBM(1)(((1) & 1) << 8); |
1754 | } |
1755 | |
1756 | if (reset_mask & RADEON_RESET_DMA(1 << 2)) { |
1757 | if (rdev->family >= CHIP_RV770) |
1758 | srbm_soft_reset |= RV770_SOFT_RESET_DMA(1 << 20); |
1759 | else |
1760 | srbm_soft_reset |= SOFT_RESET_DMA(1 << 12); |
1761 | } |
1762 | |
1763 | if (reset_mask & RADEON_RESET_RLC(1 << 6)) |
1764 | srbm_soft_reset |= S_000E60_SOFT_RESET_RLC(1)(((1) & 1) << 13); |
1765 | |
1766 | if (reset_mask & RADEON_RESET_SEM(1 << 7)) |
1767 | srbm_soft_reset |= S_000E60_SOFT_RESET_SEM(1)(((1) & 1) << 15); |
1768 | |
1769 | if (reset_mask & RADEON_RESET_IH(1 << 8)) |
1770 | srbm_soft_reset |= S_000E60_SOFT_RESET_IH(1)(((1) & 1) << 10); |
1771 | |
1772 | if (reset_mask & RADEON_RESET_GRBM(1 << 4)) |
1773 | srbm_soft_reset |= S_000E60_SOFT_RESET_GRBM(1)(((1) & 1) << 8); |
1774 | |
1775 | if (!(rdev->flags & RADEON_IS_IGP)) { |
1776 | if (reset_mask & RADEON_RESET_MC(1 << 10)) |
1777 | srbm_soft_reset |= S_000E60_SOFT_RESET_MC(1)(((1) & 1) << 11); |
1778 | } |
1779 | |
1780 | if (reset_mask & RADEON_RESET_VMC(1 << 9)) |
1781 | srbm_soft_reset |= S_000E60_SOFT_RESET_VMC(1)(((1) & 1) << 17); |
1782 | |
1783 | if (grbm_soft_reset) { |
1784 | tmp = RREG32(R_008020_GRBM_SOFT_RESET)r100_mm_rreg(rdev, (0x8020), 0); |
1785 | tmp |= grbm_soft_reset; |
1786 | dev_info(rdev->dev, "R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp)do { } while(0); |
1787 | WREG32(R_008020_GRBM_SOFT_RESET, tmp)r100_mm_wreg(rdev, (0x8020), (tmp), 0); |
1788 | tmp = RREG32(R_008020_GRBM_SOFT_RESET)r100_mm_rreg(rdev, (0x8020), 0); |
1789 | |
1790 | udelay(50); |
1791 | |
1792 | tmp &= ~grbm_soft_reset; |
1793 | WREG32(R_008020_GRBM_SOFT_RESET, tmp)r100_mm_wreg(rdev, (0x8020), (tmp), 0); |
1794 | tmp = RREG32(R_008020_GRBM_SOFT_RESET)r100_mm_rreg(rdev, (0x8020), 0); |
Value stored to 'tmp' is never read | |
1795 | } |
1796 | |
1797 | if (srbm_soft_reset) { |
1798 | tmp = RREG32(SRBM_SOFT_RESET)r100_mm_rreg(rdev, (0xe60), 0); |
1799 | tmp |= srbm_soft_reset; |
1800 | dev_info(rdev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp)do { } while(0); |
1801 | WREG32(SRBM_SOFT_RESET, tmp)r100_mm_wreg(rdev, (0xe60), (tmp), 0); |
1802 | tmp = RREG32(SRBM_SOFT_RESET)r100_mm_rreg(rdev, (0xe60), 0); |
1803 | |
1804 | udelay(50); |
1805 | |
1806 | tmp &= ~srbm_soft_reset; |
1807 | WREG32(SRBM_SOFT_RESET, tmp)r100_mm_wreg(rdev, (0xe60), (tmp), 0); |
1808 | tmp = RREG32(SRBM_SOFT_RESET)r100_mm_rreg(rdev, (0xe60), 0); |
1809 | } |
1810 | |
1811 | /* Wait a little for things to settle down */ |
1812 | mdelay(1); |
1813 | |
1814 | rv515_mc_resume(rdev, &save); |
1815 | udelay(50); |
1816 | |
1817 | r600_print_gpu_status_regs(rdev); |
1818 | } |
1819 | |
1820 | static void r600_gpu_pci_config_reset(struct radeon_device *rdev) |
1821 | { |
1822 | struct rv515_mc_save save; |
1823 | u32 tmp, i; |
1824 | |
1825 | dev_info(rdev->dev, "GPU pci config reset\n")do { } while(0); |
1826 | |
1827 | /* disable dpm? */ |
1828 | |
1829 | /* Disable CP parsing/prefetching */ |
1830 | if (rdev->family >= CHIP_RV770) |
1831 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1) | S_0086D8_CP_PFP_HALT(1))r100_mm_wreg(rdev, (0x86D8), ((((1) & 1)<<28) | ((( 1) & 1)<<26)), 0); |
1832 | else |
1833 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1))r100_mm_wreg(rdev, (0x86D8), ((((1) & 1)<<28)), 0); |
1834 | |
1835 | /* disable the RLC */ |
1836 | WREG32(RLC_CNTL, 0)r100_mm_wreg(rdev, (0x3f00), (0), 0); |
1837 | |
1838 | /* Disable DMA */ |
1839 | tmp = RREG32(DMA_RB_CNTL)r100_mm_rreg(rdev, (0xd000), 0); |
1840 | tmp &= ~DMA_RB_ENABLE(1 << 0); |
1841 | WREG32(DMA_RB_CNTL, tmp)r100_mm_wreg(rdev, (0xd000), (tmp), 0); |
1842 | |
1843 | mdelay(50); |
1844 | |
1845 | /* set mclk/sclk to bypass */ |
1846 | if (rdev->family >= CHIP_RV770) |
1847 | rv770_set_clk_bypass_mode(rdev); |
1848 | /* disable BM */ |
1849 | pci_clear_master(rdev->pdev); |
1850 | /* disable mem access */ |
1851 | rv515_mc_stop(rdev, &save); |
1852 | if (r600_mc_wait_for_idle(rdev)) { |
1853 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n")printf("drm:pid%d:%s *WARNING* " "Wait for MC idle timedout !\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
1854 | } |
1855 | |
1856 | /* BIF reset workaround. Not sure if this is needed on 6xx */ |
1857 | tmp = RREG32(BUS_CNTL)r100_mm_rreg(rdev, (0x5420), 0); |
1858 | tmp |= VGA_COHE_SPEC_TIMER_DIS(1 << 9); |
1859 | WREG32(BUS_CNTL, tmp)r100_mm_wreg(rdev, (0x5420), (tmp), 0); |
1860 | |
1861 | tmp = RREG32(BIF_SCRATCH0)r100_mm_rreg(rdev, (0x5438), 0); |
1862 | |
1863 | /* reset */ |
1864 | radeon_pci_config_reset(rdev); |
1865 | mdelay(1); |
1866 | |
1867 | /* BIF reset workaround. Not sure if this is needed on 6xx */ |
1868 | tmp = SOFT_RESET_BIF(1 << 1); |
1869 | WREG32(SRBM_SOFT_RESET, tmp)r100_mm_wreg(rdev, (0xe60), (tmp), 0); |
1870 | mdelay(1); |
1871 | WREG32(SRBM_SOFT_RESET, 0)r100_mm_wreg(rdev, (0xe60), (0), 0); |
1872 | |
1873 | /* wait for asic to come out of reset */ |
1874 | for (i = 0; i < rdev->usec_timeout; i++) { |
1875 | if (RREG32(CONFIG_MEMSIZE)r100_mm_rreg(rdev, (0x5428), 0) != 0xffffffff) |
1876 | break; |
1877 | udelay(1); |
1878 | } |
1879 | } |
1880 | |
1881 | int r600_asic_reset(struct radeon_device *rdev, bool_Bool hard) |
1882 | { |
1883 | u32 reset_mask; |
1884 | |
1885 | if (hard) { |
1886 | r600_gpu_pci_config_reset(rdev); |
1887 | return 0; |
1888 | } |
1889 | |
1890 | reset_mask = r600_gpu_check_soft_reset(rdev); |
1891 | |
1892 | if (reset_mask) |
1893 | r600_set_bios_scratch_engine_hung(rdev, true1); |
1894 | |
1895 | /* try soft reset */ |
1896 | r600_gpu_soft_reset(rdev, reset_mask); |
1897 | |
1898 | reset_mask = r600_gpu_check_soft_reset(rdev); |
1899 | |
1900 | /* try pci config reset */ |
1901 | if (reset_mask && radeon_hard_reset) |
1902 | r600_gpu_pci_config_reset(rdev); |
1903 | |
1904 | reset_mask = r600_gpu_check_soft_reset(rdev); |
1905 | |
1906 | if (!reset_mask) |
1907 | r600_set_bios_scratch_engine_hung(rdev, false0); |
1908 | |
1909 | return 0; |
1910 | } |
1911 | |
1912 | /** |
1913 | * r600_gfx_is_lockup - Check if the GFX engine is locked up |
1914 | * |
1915 | * @rdev: radeon_device pointer |
1916 | * @ring: radeon_ring structure holding ring information |
1917 | * |
1918 | * Check if the GFX engine is locked up. |
1919 | * Returns true if the engine appears to be locked up, false if not. |
1920 | */ |
1921 | bool_Bool r600_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) |
1922 | { |
1923 | u32 reset_mask = r600_gpu_check_soft_reset(rdev); |
1924 | |
1925 | if (!(reset_mask & (RADEON_RESET_GFX(1 << 0) | |
1926 | RADEON_RESET_COMPUTE(1 << 1) | |
1927 | RADEON_RESET_CP(1 << 3)))) { |
1928 | radeon_ring_lockup_update(rdev, ring); |
1929 | return false0; |
1930 | } |
1931 | return radeon_ring_test_lockup(rdev, ring); |
1932 | } |
1933 | |
1934 | u32 r6xx_remap_render_backend(struct radeon_device *rdev, |
1935 | u32 tiling_pipe_num, |
1936 | u32 max_rb_num, |
1937 | u32 total_max_rb_num, |
1938 | u32 disabled_rb_mask) |
1939 | { |
1940 | u32 rendering_pipe_num, rb_num_width, req_rb_num; |
1941 | u32 pipe_rb_ratio, pipe_rb_remain, tmp; |
1942 | u32 data = 0, mask = 1 << (max_rb_num - 1); |
1943 | unsigned i, j; |
1944 | |
1945 | /* mask out the RBs that don't exist on that asic */ |
1946 | tmp = disabled_rb_mask | ((0xff << max_rb_num) & 0xff); |
1947 | /* make sure at least one RB is available */ |
1948 | if ((tmp & 0xff) != 0xff) |
1949 | disabled_rb_mask = tmp; |
1950 | |
1951 | rendering_pipe_num = 1 << tiling_pipe_num; |
1952 | req_rb_num = total_max_rb_num - r600_count_pipe_bits(disabled_rb_mask); |
1953 | BUG_ON(rendering_pipe_num < req_rb_num)((!(rendering_pipe_num < req_rb_num)) ? (void)0 : __assert ("diagnostic ", "/usr/src/sys/dev/pci/drm/radeon/r600.c", 1953 , "!(rendering_pipe_num < req_rb_num)")); |
1954 | |
1955 | pipe_rb_ratio = rendering_pipe_num / req_rb_num; |
1956 | pipe_rb_remain = rendering_pipe_num - pipe_rb_ratio * req_rb_num; |
1957 | |
1958 | if (rdev->family <= CHIP_RV740) { |
1959 | /* r6xx/r7xx */ |
1960 | rb_num_width = 2; |
1961 | } else { |
1962 | /* eg+ */ |
1963 | rb_num_width = 4; |
1964 | } |
1965 | |
1966 | for (i = 0; i < max_rb_num; i++) { |
1967 | if (!(mask & disabled_rb_mask)) { |
1968 | for (j = 0; j < pipe_rb_ratio; j++) { |
1969 | data <<= rb_num_width; |
1970 | data |= max_rb_num - i - 1; |
1971 | } |
1972 | if (pipe_rb_remain) { |
1973 | data <<= rb_num_width; |
1974 | data |= max_rb_num - i - 1; |
1975 | pipe_rb_remain--; |
1976 | } |
1977 | } |
1978 | mask >>= 1; |
1979 | } |
1980 | |
1981 | return data; |
1982 | } |
1983 | |
1984 | int r600_count_pipe_bits(uint32_t val) |
1985 | { |
1986 | return hweight32(val); |
1987 | } |
1988 | |
1989 | static void r600_gpu_init(struct radeon_device *rdev) |
1990 | { |
1991 | u32 tiling_config; |
1992 | u32 ramcfg; |
1993 | u32 cc_gc_shader_pipe_config; |
1994 | u32 tmp; |
1995 | int i, j; |
1996 | u32 sq_config; |
1997 | u32 sq_gpr_resource_mgmt_1 = 0; |
1998 | u32 sq_gpr_resource_mgmt_2 = 0; |
1999 | u32 sq_thread_resource_mgmt = 0; |
2000 | u32 sq_stack_resource_mgmt_1 = 0; |
2001 | u32 sq_stack_resource_mgmt_2 = 0; |
2002 | u32 disabled_rb_mask; |
2003 | |
2004 | rdev->config.r600.tiling_group_size = 256; |
2005 | switch (rdev->family) { |
2006 | case CHIP_R600: |
2007 | rdev->config.r600.max_pipes = 4; |
2008 | rdev->config.r600.max_tile_pipes = 8; |
2009 | rdev->config.r600.max_simds = 4; |
2010 | rdev->config.r600.max_backends = 4; |
2011 | rdev->config.r600.max_gprs = 256; |
2012 | rdev->config.r600.max_threads = 192; |
2013 | rdev->config.r600.max_stack_entries = 256; |
2014 | rdev->config.r600.max_hw_contexts = 8; |
2015 | rdev->config.r600.max_gs_threads = 16; |
2016 | rdev->config.r600.sx_max_export_size = 128; |
2017 | rdev->config.r600.sx_max_export_pos_size = 16; |
2018 | rdev->config.r600.sx_max_export_smx_size = 128; |
2019 | rdev->config.r600.sq_num_cf_insts = 2; |
2020 | break; |
2021 | case CHIP_RV630: |
2022 | case CHIP_RV635: |
2023 | rdev->config.r600.max_pipes = 2; |
2024 | rdev->config.r600.max_tile_pipes = 2; |
2025 | rdev->config.r600.max_simds = 3; |
2026 | rdev->config.r600.max_backends = 1; |
2027 | rdev->config.r600.max_gprs = 128; |
2028 | rdev->config.r600.max_threads = 192; |
2029 | rdev->config.r600.max_stack_entries = 128; |
2030 | rdev->config.r600.max_hw_contexts = 8; |
2031 | rdev->config.r600.max_gs_threads = 4; |
2032 | rdev->config.r600.sx_max_export_size = 128; |
2033 | rdev->config.r600.sx_max_export_pos_size = 16; |
2034 | rdev->config.r600.sx_max_export_smx_size = 128; |
2035 | rdev->config.r600.sq_num_cf_insts = 2; |
2036 | break; |
2037 | case CHIP_RV610: |
2038 | case CHIP_RV620: |
2039 | case CHIP_RS780: |
2040 | case CHIP_RS880: |
2041 | rdev->config.r600.max_pipes = 1; |
2042 | rdev->config.r600.max_tile_pipes = 1; |
2043 | rdev->config.r600.max_simds = 2; |
2044 | rdev->config.r600.max_backends = 1; |
2045 | rdev->config.r600.max_gprs = 128; |
2046 | rdev->config.r600.max_threads = 192; |
2047 | rdev->config.r600.max_stack_entries = 128; |
2048 | rdev->config.r600.max_hw_contexts = 4; |
2049 | rdev->config.r600.max_gs_threads = 4; |
2050 | rdev->config.r600.sx_max_export_size = 128; |
2051 | rdev->config.r600.sx_max_export_pos_size = 16; |
2052 | rdev->config.r600.sx_max_export_smx_size = 128; |
2053 | rdev->config.r600.sq_num_cf_insts = 1; |
2054 | break; |
2055 | case CHIP_RV670: |
2056 | rdev->config.r600.max_pipes = 4; |
2057 | rdev->config.r600.max_tile_pipes = 4; |
2058 | rdev->config.r600.max_simds = 4; |
2059 | rdev->config.r600.max_backends = 4; |
2060 | rdev->config.r600.max_gprs = 192; |
2061 | rdev->config.r600.max_threads = 192; |
2062 | rdev->config.r600.max_stack_entries = 256; |
2063 | rdev->config.r600.max_hw_contexts = 8; |
2064 | rdev->config.r600.max_gs_threads = 16; |
2065 | rdev->config.r600.sx_max_export_size = 128; |
2066 | rdev->config.r600.sx_max_export_pos_size = 16; |
2067 | rdev->config.r600.sx_max_export_smx_size = 128; |
2068 | rdev->config.r600.sq_num_cf_insts = 2; |
2069 | break; |
2070 | default: |
2071 | break; |
2072 | } |
2073 | |
2074 | /* Initialize HDP */ |
2075 | for (i = 0, j = 0; i < 32; i++, j += 0x18) { |
2076 | WREG32((0x2c14 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c14 + j)), (0x00000000), 0); |
2077 | WREG32((0x2c18 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c18 + j)), (0x00000000), 0); |
2078 | WREG32((0x2c1c + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c1c + j)), (0x00000000), 0); |
2079 | WREG32((0x2c20 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c20 + j)), (0x00000000), 0); |
2080 | WREG32((0x2c24 + j), 0x00000000)r100_mm_wreg(rdev, ((0x2c24 + j)), (0x00000000), 0); |
2081 | } |
2082 | |
2083 | WREG32(GRBM_CNTL, GRBM_READ_TIMEOUT(0xff))r100_mm_wreg(rdev, (0x8000), (((0xff) << 0)), 0); |
2084 | |
2085 | /* Setup tiling */ |
2086 | tiling_config = 0; |
2087 | ramcfg = RREG32(RAMCFG)r100_mm_rreg(rdev, (0x2408), 0); |
2088 | switch (rdev->config.r600.max_tile_pipes) { |
2089 | case 1: |
2090 | tiling_config |= PIPE_TILING(0)((0) << 1); |
2091 | break; |
2092 | case 2: |
2093 | tiling_config |= PIPE_TILING(1)((1) << 1); |
2094 | break; |
2095 | case 4: |
2096 | tiling_config |= PIPE_TILING(2)((2) << 1); |
2097 | break; |
2098 | case 8: |
2099 | tiling_config |= PIPE_TILING(3)((3) << 1); |
2100 | break; |
2101 | default: |
2102 | break; |
2103 | } |
2104 | rdev->config.r600.tiling_npipes = rdev->config.r600.max_tile_pipes; |
2105 | rdev->config.r600.tiling_nbanks = 4 << ((ramcfg & NOOFBANK_MASK0x00000001) >> NOOFBANK_SHIFT0); |
2106 | tiling_config |= BANK_TILING((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT)(((ramcfg & 0x00000001) >> 0) << 4); |
2107 | tiling_config |= GROUP_SIZE((ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT)(((ramcfg & 0x00000100) >> 8) << 6); |
2108 | |
2109 | tmp = (ramcfg & NOOFROWS_MASK0x0000001C) >> NOOFROWS_SHIFT2; |
2110 | if (tmp > 3) { |
2111 | tiling_config |= ROW_TILING(3)((3) << 8); |
2112 | tiling_config |= SAMPLE_SPLIT(3)((3) << 14); |
2113 | } else { |
2114 | tiling_config |= ROW_TILING(tmp)((tmp) << 8); |
2115 | tiling_config |= SAMPLE_SPLIT(tmp)((tmp) << 14); |
2116 | } |
2117 | tiling_config |= BANK_SWAPS(1)((1) << 11); |
2118 | |
2119 | cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG)r100_mm_rreg(rdev, (0x8950), 0) & 0x00ffff00; |
2120 | tmp = rdev->config.r600.max_simds - |
2121 | r600_count_pipe_bits((cc_gc_shader_pipe_config >> 16) & R6XX_MAX_SIMDS_MASK0xff); |
2122 | rdev->config.r600.active_simds = tmp; |
2123 | |
2124 | disabled_rb_mask = (RREG32(CC_RB_BACKEND_DISABLE)r100_mm_rreg(rdev, (0x98F4), 0) >> 16) & R6XX_MAX_BACKENDS_MASK0xff; |
2125 | tmp = 0; |
2126 | for (i = 0; i < rdev->config.r600.max_backends; i++) |
2127 | tmp |= (1 << i); |
2128 | /* if all the backends are disabled, fix it up here */ |
2129 | if ((disabled_rb_mask & tmp) == tmp) { |
2130 | for (i = 0; i < rdev->config.r600.max_backends; i++) |
2131 | disabled_rb_mask &= ~(1 << i); |
2132 | } |
2133 | tmp = (tiling_config & PIPE_TILING__MASK0x0000000e) >> PIPE_TILING__SHIFT1; |
2134 | tmp = r6xx_remap_render_backend(rdev, tmp, rdev->config.r600.max_backends, |
2135 | R6XX_MAX_BACKENDS8, disabled_rb_mask); |
2136 | tiling_config |= tmp << 16; |
2137 | rdev->config.r600.backend_map = tmp; |
2138 | |
2139 | rdev->config.r600.tile_config = tiling_config; |
2140 | WREG32(GB_TILING_CONFIG, tiling_config)r100_mm_wreg(rdev, (0x98F0), (tiling_config), 0); |
2141 | WREG32(DCP_TILING_CONFIG, tiling_config & 0xffff)r100_mm_wreg(rdev, (0x6CA0), (tiling_config & 0xffff), 0); |
2142 | WREG32(HDP_TILING_CONFIG, tiling_config & 0xffff)r100_mm_wreg(rdev, (0x2F3C), (tiling_config & 0xffff), 0); |
2143 | WREG32(DMA_TILING_CONFIG, tiling_config & 0xffff)r100_mm_wreg(rdev, (0x3ec4), (tiling_config & 0xffff), 0); |
2144 | |
2145 | tmp = R6XX_MAX_PIPES8 - r600_count_pipe_bits((cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK0x0000FF00) >> 8); |
2146 | WREG32(VGT_OUT_DEALLOC_CNTL, (tmp * 4) & DEALLOC_DIST_MASK)r100_mm_wreg(rdev, (0x28C5C), ((tmp * 4) & 0x0000007F), 0 ); |
2147 | WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, ((tmp * 4) - 2) & VTX_REUSE_DEPTH_MASK)r100_mm_wreg(rdev, (0x28C58), (((tmp * 4) - 2) & 0x000000FF ), 0); |
2148 | |
2149 | /* Setup some CP states */ |
2150 | WREG32(CP_QUEUE_THRESHOLDS, (ROQ_IB1_START(0x16) | ROQ_IB2_START(0x2b)))r100_mm_wreg(rdev, (0x8760), ((((0x16) << 0) | ((0x2b) << 8))), 0); |
2151 | WREG32(CP_MEQ_THRESHOLDS, (MEQ_END(0x40) | ROQ_END(0x40)))r100_mm_wreg(rdev, (0x8764), ((((0x40) << 16) | ((0x40) << 24))), 0); |
2152 | |
2153 | WREG32(TA_CNTL_AUX, (DISABLE_CUBE_ANISO | SYNC_GRADIENT |r100_mm_wreg(rdev, (0x9508), (((1 << 1) | (1 << 24 ) | (1 << 25) | (1 << 26))), 0) |
2154 | SYNC_WALKER | SYNC_ALIGNER))r100_mm_wreg(rdev, (0x9508), (((1 << 1) | (1 << 24 ) | (1 << 25) | (1 << 26))), 0); |
2155 | /* Setup various GPU states */ |
2156 | if (rdev->family == CHIP_RV670) |
2157 | WREG32(ARB_GDEC_RD_CNTL, 0x00000021)r100_mm_wreg(rdev, (0x246C), (0x00000021), 0); |
2158 | |
2159 | tmp = RREG32(SX_DEBUG_1)r100_mm_rreg(rdev, (0x9054), 0); |
2160 | tmp |= SMX_EVENT_RELEASE(1 << 0); |
2161 | if ((rdev->family > CHIP_R600)) |
2162 | tmp |= ENABLE_NEW_SMX_ADDRESS(1 << 16); |
2163 | WREG32(SX_DEBUG_1, tmp)r100_mm_wreg(rdev, (0x9054), (tmp), 0); |
2164 | |
2165 | if (((rdev->family) == CHIP_R600) || |
2166 | ((rdev->family) == CHIP_RV630) || |
2167 | ((rdev->family) == CHIP_RV610) || |
2168 | ((rdev->family) == CHIP_RV620) || |
2169 | ((rdev->family) == CHIP_RS780) || |
2170 | ((rdev->family) == CHIP_RS880)) { |
2171 | WREG32(DB_DEBUG, PREZ_MUST_WAIT_FOR_POSTZ_DONE)r100_mm_wreg(rdev, (0x9830), ((1 << 31)), 0); |
2172 | } else { |
2173 | WREG32(DB_DEBUG, 0)r100_mm_wreg(rdev, (0x9830), (0), 0); |
2174 | } |
2175 | WREG32(DB_WATERMARKS, (DEPTH_FREE(4) | DEPTH_CACHELINE_FREE(16) |r100_mm_wreg(rdev, (0x9838), ((((4) << 0) | ((16) << 20) | ((16) << 5) | ((4) << 15))), 0) |
2176 | DEPTH_FLUSH(16) | DEPTH_PENDING_FREE(4)))r100_mm_wreg(rdev, (0x9838), ((((4) << 0) | ((16) << 20) | ((16) << 5) | ((4) << 15))), 0); |
2177 | |
2178 | WREG32(PA_SC_MULTI_CHIP_CNTL, 0)r100_mm_wreg(rdev, (0x8B20), (0), 0); |
2179 | WREG32(VGT_NUM_INSTANCES, 0)r100_mm_wreg(rdev, (0x8974), (0), 0); |
2180 | |
2181 | WREG32(SPI_CONFIG_CNTL, GPR_WRITE_PRIORITY(0))r100_mm_wreg(rdev, (0x9100), (((0) << 0)), 0); |
2182 | WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(0))r100_mm_wreg(rdev, (0x913C), (((0) << 0)), 0); |
2183 | |
2184 | tmp = RREG32(SQ_MS_FIFO_SIZES)r100_mm_rreg(rdev, (0x8CF0), 0); |
2185 | if (((rdev->family) == CHIP_RV610) || |
2186 | ((rdev->family) == CHIP_RV620) || |
2187 | ((rdev->family) == CHIP_RS780) || |
2188 | ((rdev->family) == CHIP_RS880)) { |
2189 | tmp = (CACHE_FIFO_SIZE(0xa)((0xa) << 0) | |
2190 | FETCH_FIFO_HIWATER(0xa)((0xa) << 8) | |
2191 | DONE_FIFO_HIWATER(0xe0)((0xe0) << 16) | |
2192 | ALU_UPDATE_FIFO_HIWATER(0x8)((0x8) << 24)); |
2193 | } else if (((rdev->family) == CHIP_R600) || |
2194 | ((rdev->family) == CHIP_RV630)) { |
2195 | tmp &= ~DONE_FIFO_HIWATER(0xff)((0xff) << 16); |
2196 | tmp |= DONE_FIFO_HIWATER(0x4)((0x4) << 16); |
2197 | } |
2198 | WREG32(SQ_MS_FIFO_SIZES, tmp)r100_mm_wreg(rdev, (0x8CF0), (tmp), 0); |
2199 | |
2200 | /* SQ_CONFIG, SQ_GPR_RESOURCE_MGMT, SQ_THREAD_RESOURCE_MGMT, SQ_STACK_RESOURCE_MGMT |
2201 | * should be adjusted as needed by the 2D/3D drivers. This just sets default values |
2202 | */ |
2203 | sq_config = RREG32(SQ_CONFIG)r100_mm_rreg(rdev, (0x8c00), 0); |
2204 | sq_config &= ~(PS_PRIO(3)((3) << 24) | |
2205 | VS_PRIO(3)((3) << 26) | |
2206 | GS_PRIO(3)((3) << 28) | |
2207 | ES_PRIO(3)((3) << 30)); |
2208 | sq_config |= (DX9_CONSTS(1 << 2) | |
2209 | VC_ENABLE(1 << 0) | |
2210 | PS_PRIO(0)((0) << 24) | |
2211 | VS_PRIO(1)((1) << 26) | |
2212 | GS_PRIO(2)((2) << 28) | |
2213 | ES_PRIO(3)((3) << 30)); |
2214 | |
2215 | if ((rdev->family) == CHIP_R600) { |
2216 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(124)((124) << 0) | |
2217 | NUM_VS_GPRS(124)((124) << 16) | |
2218 | NUM_CLAUSE_TEMP_GPRS(4)((4) << 28)); |
2219 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(0)((0) << 0) | |
2220 | NUM_ES_GPRS(0)((0) << 16)); |
2221 | sq_thread_resource_mgmt = (NUM_PS_THREADS(136)((136) << 0) | |
2222 | NUM_VS_THREADS(48)((48) << 8) | |
2223 | NUM_GS_THREADS(4)((4) << 16) | |
2224 | NUM_ES_THREADS(4)((4) << 24)); |
2225 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(128)((128) << 0) | |
2226 | NUM_VS_STACK_ENTRIES(128)((128) << 16)); |
2227 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(0)((0) << 0) | |
2228 | NUM_ES_STACK_ENTRIES(0)((0) << 16)); |
2229 | } else if (((rdev->family) == CHIP_RV610) || |
2230 | ((rdev->family) == CHIP_RV620) || |
2231 | ((rdev->family) == CHIP_RS780) || |
2232 | ((rdev->family) == CHIP_RS880)) { |
2233 | /* no vertex cache */ |
2234 | sq_config &= ~VC_ENABLE(1 << 0); |
2235 | |
2236 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44)((44) << 0) | |
2237 | NUM_VS_GPRS(44)((44) << 16) | |
2238 | NUM_CLAUSE_TEMP_GPRS(2)((2) << 28)); |
2239 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(17)((17) << 0) | |
2240 | NUM_ES_GPRS(17)((17) << 16)); |
2241 | sq_thread_resource_mgmt = (NUM_PS_THREADS(79)((79) << 0) | |
2242 | NUM_VS_THREADS(78)((78) << 8) | |
2243 | NUM_GS_THREADS(4)((4) << 16) | |
2244 | NUM_ES_THREADS(31)((31) << 24)); |
2245 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(40)((40) << 0) | |
2246 | NUM_VS_STACK_ENTRIES(40)((40) << 16)); |
2247 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(32)((32) << 0) | |
2248 | NUM_ES_STACK_ENTRIES(16)((16) << 16)); |
2249 | } else if (((rdev->family) == CHIP_RV630) || |
2250 | ((rdev->family) == CHIP_RV635)) { |
2251 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44)((44) << 0) | |
2252 | NUM_VS_GPRS(44)((44) << 16) | |
2253 | NUM_CLAUSE_TEMP_GPRS(2)((2) << 28)); |
2254 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(18)((18) << 0) | |
2255 | NUM_ES_GPRS(18)((18) << 16)); |
2256 | sq_thread_resource_mgmt = (NUM_PS_THREADS(79)((79) << 0) | |
2257 | NUM_VS_THREADS(78)((78) << 8) | |
2258 | NUM_GS_THREADS(4)((4) << 16) | |
2259 | NUM_ES_THREADS(31)((31) << 24)); |
2260 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(40)((40) << 0) | |
2261 | NUM_VS_STACK_ENTRIES(40)((40) << 16)); |
2262 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(32)((32) << 0) | |
2263 | NUM_ES_STACK_ENTRIES(16)((16) << 16)); |
2264 | } else if ((rdev->family) == CHIP_RV670) { |
2265 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44)((44) << 0) | |
2266 | NUM_VS_GPRS(44)((44) << 16) | |
2267 | NUM_CLAUSE_TEMP_GPRS(2)((2) << 28)); |
2268 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(17)((17) << 0) | |
2269 | NUM_ES_GPRS(17)((17) << 16)); |
2270 | sq_thread_resource_mgmt = (NUM_PS_THREADS(79)((79) << 0) | |
2271 | NUM_VS_THREADS(78)((78) << 8) | |
2272 | NUM_GS_THREADS(4)((4) << 16) | |
2273 | NUM_ES_THREADS(31)((31) << 24)); |
2274 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(64)((64) << 0) | |
2275 | NUM_VS_STACK_ENTRIES(64)((64) << 16)); |
2276 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(64)((64) << 0) | |
2277 | NUM_ES_STACK_ENTRIES(64)((64) << 16)); |
2278 | } |
2279 | |
2280 | WREG32(SQ_CONFIG, sq_config)r100_mm_wreg(rdev, (0x8c00), (sq_config), 0); |
2281 | WREG32(SQ_GPR_RESOURCE_MGMT_1, sq_gpr_resource_mgmt_1)r100_mm_wreg(rdev, (0x8c04), (sq_gpr_resource_mgmt_1), 0); |
2282 | WREG32(SQ_GPR_RESOURCE_MGMT_2, sq_gpr_resource_mgmt_2)r100_mm_wreg(rdev, (0x8c08), (sq_gpr_resource_mgmt_2), 0); |
2283 | WREG32(SQ_THREAD_RESOURCE_MGMT, sq_thread_resource_mgmt)r100_mm_wreg(rdev, (0x8c0c), (sq_thread_resource_mgmt), 0); |
2284 | WREG32(SQ_STACK_RESOURCE_MGMT_1, sq_stack_resource_mgmt_1)r100_mm_wreg(rdev, (0x8c10), (sq_stack_resource_mgmt_1), 0); |
2285 | WREG32(SQ_STACK_RESOURCE_MGMT_2, sq_stack_resource_mgmt_2)r100_mm_wreg(rdev, (0x8c14), (sq_stack_resource_mgmt_2), 0); |
2286 | |
2287 | if (((rdev->family) == CHIP_RV610) || |
2288 | ((rdev->family) == CHIP_RV620) || |
2289 | ((rdev->family) == CHIP_RS780) || |
2290 | ((rdev->family) == CHIP_RS880)) { |
2291 | WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(TC_ONLY))r100_mm_wreg(rdev, (0x88C4), (((1)<<0)), 0); |
2292 | } else { |
2293 | WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC))r100_mm_wreg(rdev, (0x88C4), (((2)<<0)), 0); |
2294 | } |
2295 | |
2296 | /* More default values. 2D/3D driver should adjust as needed */ |
2297 | WREG32(PA_SC_AA_SAMPLE_LOCS_2S, (S0_X(0xc) | S0_Y(0x4) |r100_mm_wreg(rdev, (0x8B40), ((((0xc) << 0) | ((0x4) << 4) | ((0x4) << 8) | ((0xc) << 12))), 0) |
2298 | S1_X(0x4) | S1_Y(0xc)))r100_mm_wreg(rdev, (0x8B40), ((((0xc) << 0) | ((0x4) << 4) | ((0x4) << 8) | ((0xc) << 12))), 0); |
2299 | WREG32(PA_SC_AA_SAMPLE_LOCS_4S, (S0_X(0xe) | S0_Y(0xe) |r100_mm_wreg(rdev, (0x8B44), ((((0xe) << 0) | ((0xe) << 4) | ((0x2) << 8) | ((0x2) << 12) | ((0xa) << 16) | ((0x6) << 20) | ((0x6) << 24) | ((0xa) << 28))), 0) |
2300 | S1_X(0x2) | S1_Y(0x2) |r100_mm_wreg(rdev, (0x8B44), ((((0xe) << 0) | ((0xe) << 4) | ((0x2) << 8) | ((0x2) << 12) | ((0xa) << 16) | ((0x6) << 20) | ((0x6) << 24) | ((0xa) << 28))), 0) |
2301 | S2_X(0xa) | S2_Y(0x6) |r100_mm_wreg(rdev, (0x8B44), ((((0xe) << 0) | ((0xe) << 4) | ((0x2) << 8) | ((0x2) << 12) | ((0xa) << 16) | ((0x6) << 20) | ((0x6) << 24) | ((0xa) << 28))), 0) |
2302 | S3_X(0x6) | S3_Y(0xa)))r100_mm_wreg(rdev, (0x8B44), ((((0xe) << 0) | ((0xe) << 4) | ((0x2) << 8) | ((0x2) << 12) | ((0xa) << 16) | ((0x6) << 20) | ((0x6) << 24) | ((0xa) << 28))), 0); |
2303 | WREG32(PA_SC_AA_SAMPLE_LOCS_8S_WD0, (S0_X(0xe) | S0_Y(0xb) |r100_mm_wreg(rdev, (0x8B48), ((((0xe) << 0) | ((0xb) << 4) | ((0x4) << 8) | ((0xc) << 12) | ((0x1) << 16) | ((0x6) << 20) | ((0xa) << 24) | ((0xe) << 28))), 0) |
2304 | S1_X(0x4) | S1_Y(0xc) |r100_mm_wreg(rdev, (0x8B48), ((((0xe) << 0) | ((0xb) << 4) | ((0x4) << 8) | ((0xc) << 12) | ((0x1) << 16) | ((0x6) << 20) | ((0xa) << 24) | ((0xe) << 28))), 0) |
2305 | S2_X(0x1) | S2_Y(0x6) |r100_mm_wreg(rdev, (0x8B48), ((((0xe) << 0) | ((0xb) << 4) | ((0x4) << 8) | ((0xc) << 12) | ((0x1) << 16) | ((0x6) << 20) | ((0xa) << 24) | ((0xe) << 28))), 0) |
2306 | S3_X(0xa) | S3_Y(0xe)))r100_mm_wreg(rdev, (0x8B48), ((((0xe) << 0) | ((0xb) << 4) | ((0x4) << 8) | ((0xc) << 12) | ((0x1) << 16) | ((0x6) << 20) | ((0xa) << 24) | ((0xe) << 28))), 0); |
2307 | WREG32(PA_SC_AA_SAMPLE_LOCS_8S_WD1, (S4_X(0x6) | S4_Y(0x1) |r100_mm_wreg(rdev, (0x8B4C), ((((0x6) << 0) | ((0x1) << 4) | ((0x0) << 8) | ((0x0) << 12) | ((0xb) << 16) | ((0x4) << 20) | ((0x7) << 24) | ((0x8) << 28))), 0) |
2308 | S5_X(0x0) | S5_Y(0x0) |r100_mm_wreg(rdev, (0x8B4C), ((((0x6) << 0) | ((0x1) << 4) | ((0x0) << 8) | ((0x0) << 12) | ((0xb) << 16) | ((0x4) << 20) | ((0x7) << 24) | ((0x8) << 28))), 0) |
2309 | S6_X(0xb) | S6_Y(0x4) |r100_mm_wreg(rdev, (0x8B4C), ((((0x6) << 0) | ((0x1) << 4) | ((0x0) << 8) | ((0x0) << 12) | ((0xb) << 16) | ((0x4) << 20) | ((0x7) << 24) | ((0x8) << 28))), 0) |
2310 | S7_X(0x7) | S7_Y(0x8)))r100_mm_wreg(rdev, (0x8B4C), ((((0x6) << 0) | ((0x1) << 4) | ((0x0) << 8) | ((0x0) << 12) | ((0xb) << 16) | ((0x4) << 20) | ((0x7) << 24) | ((0x8) << 28))), 0); |
2311 | |
2312 | WREG32(VGT_STRMOUT_EN, 0)r100_mm_wreg(rdev, (0x28AB0), (0), 0); |
2313 | tmp = rdev->config.r600.max_pipes * 16; |
2314 | switch (rdev->family) { |
2315 | case CHIP_RV610: |
2316 | case CHIP_RV620: |
2317 | case CHIP_RS780: |
2318 | case CHIP_RS880: |
2319 | tmp += 32; |
2320 | break; |
2321 | case CHIP_RV670: |
2322 | tmp += 128; |
2323 | break; |
2324 | default: |
2325 | break; |
2326 | } |
2327 | if (tmp > 256) { |
2328 | tmp = 256; |
2329 | } |
2330 | WREG32(VGT_ES_PER_GS, 128)r100_mm_wreg(rdev, (0x88CC), (128), 0); |
2331 | WREG32(VGT_GS_PER_ES, tmp)r100_mm_wreg(rdev, (0x88C8), (tmp), 0); |
2332 | WREG32(VGT_GS_PER_VS, 2)r100_mm_wreg(rdev, (0x88E8), (2), 0); |
2333 | WREG32(VGT_GS_VERTEX_REUSE, 16)r100_mm_wreg(rdev, (0x88D4), (16), 0); |
2334 | |
2335 | /* more default values. 2D/3D driver should adjust as needed */ |
2336 | WREG32(PA_SC_LINE_STIPPLE_STATE, 0)r100_mm_wreg(rdev, (0x8B10), (0), 0); |
2337 | WREG32(VGT_STRMOUT_EN, 0)r100_mm_wreg(rdev, (0x28AB0), (0), 0); |
2338 | WREG32(SX_MISC, 0)r100_mm_wreg(rdev, (0x28350), (0), 0); |
2339 | WREG32(PA_SC_MODE_CNTL, 0)r100_mm_wreg(rdev, (0x28A4C), (0), 0); |
2340 | WREG32(PA_SC_AA_CONFIG, 0)r100_mm_wreg(rdev, (0x28C04), (0), 0); |
2341 | WREG32(PA_SC_LINE_STIPPLE, 0)r100_mm_wreg(rdev, (0x28A0C), (0), 0); |
2342 | WREG32(SPI_INPUT_Z, 0)r100_mm_wreg(rdev, (0x286D8), (0), 0); |
2343 | WREG32(SPI_PS_IN_CONTROL_0, NUM_INTERP(2))r100_mm_wreg(rdev, (0x286CC), (((2)<<0)), 0); |
2344 | WREG32(CB_COLOR7_FRAG, 0)r100_mm_wreg(rdev, (0x280FC), (0), 0); |
2345 | |
2346 | /* Clear render buffer base addresses */ |
2347 | WREG32(CB_COLOR0_BASE, 0)r100_mm_wreg(rdev, (0x28040), (0), 0); |
2348 | WREG32(CB_COLOR1_BASE, 0)r100_mm_wreg(rdev, (0x28044), (0), 0); |
2349 | WREG32(CB_COLOR2_BASE, 0)r100_mm_wreg(rdev, (0x28048), (0), 0); |
2350 | WREG32(CB_COLOR3_BASE, 0)r100_mm_wreg(rdev, (0x2804C), (0), 0); |
2351 | WREG32(CB_COLOR4_BASE, 0)r100_mm_wreg(rdev, (0x28050), (0), 0); |
2352 | WREG32(CB_COLOR5_BASE, 0)r100_mm_wreg(rdev, (0x28054), (0), 0); |
2353 | WREG32(CB_COLOR6_BASE, 0)r100_mm_wreg(rdev, (0x28058), (0), 0); |
2354 | WREG32(CB_COLOR7_BASE, 0)r100_mm_wreg(rdev, (0x2805C), (0), 0); |
2355 | WREG32(CB_COLOR7_FRAG, 0)r100_mm_wreg(rdev, (0x280FC), (0), 0); |
2356 | |
2357 | switch (rdev->family) { |
2358 | case CHIP_RV610: |
2359 | case CHIP_RV620: |
2360 | case CHIP_RS780: |
2361 | case CHIP_RS880: |
2362 | tmp = TC_L2_SIZE(8)((8)<<5); |
2363 | break; |
2364 | case CHIP_RV630: |
2365 | case CHIP_RV635: |
2366 | tmp = TC_L2_SIZE(4)((4)<<5); |
2367 | break; |
2368 | case CHIP_R600: |
2369 | tmp = TC_L2_SIZE(0)((0)<<5) | L2_DISABLE_LATE_HIT(1<<9); |
2370 | break; |
2371 | default: |
2372 | tmp = TC_L2_SIZE(0)((0)<<5); |
2373 | break; |
2374 | } |
2375 | WREG32(TC_CNTL, tmp)r100_mm_wreg(rdev, (0x9608), (tmp), 0); |
2376 | |
2377 | tmp = RREG32(HDP_HOST_PATH_CNTL)r100_mm_rreg(rdev, (0x2C00), 0); |
2378 | WREG32(HDP_HOST_PATH_CNTL, tmp)r100_mm_wreg(rdev, (0x2C00), (tmp), 0); |
2379 | |
2380 | tmp = RREG32(ARB_POP)r100_mm_rreg(rdev, (0x2418), 0); |
2381 | tmp |= ENABLE_TC128(1 << 30); |
2382 | WREG32(ARB_POP, tmp)r100_mm_wreg(rdev, (0x2418), (tmp), 0); |
2383 | |
2384 | WREG32(PA_SC_MULTI_CHIP_CNTL, 0)r100_mm_wreg(rdev, (0x8B20), (0), 0); |
2385 | WREG32(PA_CL_ENHANCE, (CLIP_VTX_REORDER_ENA |r100_mm_wreg(rdev, (0x8A14), (((1 << 0) | ((3) << 1))), 0) |
2386 | NUM_CLIP_SEQ(3)))r100_mm_wreg(rdev, (0x8A14), (((1 << 0) | ((3) << 1))), 0); |
2387 | WREG32(PA_SC_ENHANCE, FORCE_EOV_MAX_CLK_CNT(4095))r100_mm_wreg(rdev, (0x8BF0), (((4095) << 0)), 0); |
2388 | WREG32(VC_ENHANCE, 0)r100_mm_wreg(rdev, (0x9714), (0), 0); |
2389 | } |
2390 | |
2391 | |
2392 | /* |
2393 | * Indirect registers accessor |
2394 | */ |
2395 | u32 r600_pciep_rreg(struct radeon_device *rdev, u32 reg) |
2396 | { |
2397 | unsigned long flags; |
2398 | u32 r; |
2399 | |
2400 | spin_lock_irqsave(&rdev->pciep_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->pciep_idx_lock); } while (0); |
2401 | WREG32(PCIE_PORT_INDEX, ((reg) & 0xff))r100_mm_wreg(rdev, (0x0038), (((reg) & 0xff)), 0); |
2402 | (void)RREG32(PCIE_PORT_INDEX)r100_mm_rreg(rdev, (0x0038), 0); |
2403 | r = RREG32(PCIE_PORT_DATA)r100_mm_rreg(rdev, (0x003C), 0); |
2404 | spin_unlock_irqrestore(&rdev->pciep_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->pciep_idx_lock); } while (0); |
2405 | return r; |
2406 | } |
2407 | |
2408 | void r600_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v) |
2409 | { |
2410 | unsigned long flags; |
2411 | |
2412 | spin_lock_irqsave(&rdev->pciep_idx_lock, flags)do { flags = 0; mtx_enter(&rdev->pciep_idx_lock); } while (0); |
2413 | WREG32(PCIE_PORT_INDEX, ((reg) & 0xff))r100_mm_wreg(rdev, (0x0038), (((reg) & 0xff)), 0); |
2414 | (void)RREG32(PCIE_PORT_INDEX)r100_mm_rreg(rdev, (0x0038), 0); |
2415 | WREG32(PCIE_PORT_DATA, (v))r100_mm_wreg(rdev, (0x003C), ((v)), 0); |
2416 | (void)RREG32(PCIE_PORT_DATA)r100_mm_rreg(rdev, (0x003C), 0); |
2417 | spin_unlock_irqrestore(&rdev->pciep_idx_lock, flags)do { (void)(flags); mtx_leave(&rdev->pciep_idx_lock); } while (0); |
2418 | } |
2419 | |
2420 | /* |
2421 | * CP & Ring |
2422 | */ |
2423 | void r600_cp_stop(struct radeon_device *rdev) |
2424 | { |
2425 | if (rdev->asic->copy.copy_ring_index == RADEON_RING_TYPE_GFX_INDEX0) |
2426 | radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); |
2427 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1))r100_mm_wreg(rdev, (0x86D8), ((((1) & 1)<<28)), 0); |
2428 | WREG32(SCRATCH_UMSK, 0)r100_mm_wreg(rdev, (0x8540), (0), 0); |
2429 | rdev->ring[RADEON_RING_TYPE_GFX_INDEX0].ready = false0; |
2430 | } |
2431 | |
2432 | int r600_init_microcode(struct radeon_device *rdev) |
2433 | { |
2434 | const char *chip_name; |
2435 | const char *rlc_chip_name; |
2436 | const char *smc_chip_name = "RV770"; |
2437 | size_t pfp_req_size, me_req_size, rlc_req_size, smc_req_size = 0; |
2438 | char fw_name[30]; |
2439 | int err; |
2440 | |
2441 | DRM_DEBUG("\n")___drm_dbg(((void *)0), DRM_UT_CORE, "\n"); |
2442 | |
2443 | switch (rdev->family) { |
2444 | case CHIP_R600: |
2445 | chip_name = "R600"; |
2446 | rlc_chip_name = "R600"; |
2447 | break; |
2448 | case CHIP_RV610: |
2449 | chip_name = "RV610"; |
2450 | rlc_chip_name = "R600"; |
2451 | break; |
2452 | case CHIP_RV630: |
2453 | chip_name = "RV630"; |
2454 | rlc_chip_name = "R600"; |
2455 | break; |
2456 | case CHIP_RV620: |
2457 | chip_name = "RV620"; |
2458 | rlc_chip_name = "R600"; |
2459 | break; |
2460 | case CHIP_RV635: |
2461 | chip_name = "RV635"; |
2462 | rlc_chip_name = "R600"; |
2463 | break; |
2464 | case CHIP_RV670: |
2465 | chip_name = "RV670"; |
2466 | rlc_chip_name = "R600"; |
2467 | break; |
2468 | case CHIP_RS780: |
2469 | case CHIP_RS880: |
2470 | chip_name = "RS780"; |
2471 | rlc_chip_name = "R600"; |
2472 | break; |
2473 | case CHIP_RV770: |
2474 | chip_name = "RV770"; |
2475 | rlc_chip_name = "R700"; |
2476 | smc_chip_name = "RV770"; |
2477 | smc_req_size = roundup2(RV770_SMC_UCODE_SIZE, 4)(((0x410d) + ((4) - 1)) & (~((__typeof(0x410d))(4) - 1))); |
2478 | break; |
2479 | case CHIP_RV730: |
2480 | chip_name = "RV730"; |
2481 | rlc_chip_name = "R700"; |
2482 | smc_chip_name = "RV730"; |
2483 | smc_req_size = roundup2(RV730_SMC_UCODE_SIZE, 4)(((0x412c) + ((4) - 1)) & (~((__typeof(0x412c))(4) - 1))); |
2484 | break; |
2485 | case CHIP_RV710: |
2486 | chip_name = "RV710"; |
2487 | rlc_chip_name = "R700"; |
2488 | smc_chip_name = "RV710"; |
2489 | smc_req_size = roundup2(RV710_SMC_UCODE_SIZE, 4)(((0x3f1f) + ((4) - 1)) & (~((__typeof(0x3f1f))(4) - 1))); |
2490 | break; |
2491 | case CHIP_RV740: |
2492 | chip_name = "RV730"; |
2493 | rlc_chip_name = "R700"; |
2494 | smc_chip_name = "RV740"; |
2495 | smc_req_size = roundup2(RV740_SMC_UCODE_SIZE, 4)(((0x41c5) + ((4) - 1)) & (~((__typeof(0x41c5))(4) - 1))); |
2496 | break; |
2497 | case CHIP_CEDAR: |
2498 | chip_name = "CEDAR"; |
2499 | rlc_chip_name = "CEDAR"; |
2500 | smc_chip_name = "CEDAR"; |
2501 | smc_req_size = roundup2(CEDAR_SMC_UCODE_SIZE, 4)(((0x5d50) + ((4) - 1)) & (~((__typeof(0x5d50))(4) - 1))); |
2502 | break; |
2503 | case CHIP_REDWOOD: |
2504 | chip_name = "REDWOOD"; |
2505 | rlc_chip_name = "REDWOOD"; |
2506 | smc_chip_name = "REDWOOD"; |
2507 | smc_req_size = roundup2(REDWOOD_SMC_UCODE_SIZE, 4)(((0x5f0a) + ((4) - 1)) & (~((__typeof(0x5f0a))(4) - 1))); |
2508 | break; |
2509 | case CHIP_JUNIPER: |
2510 | chip_name = "JUNIPER"; |
2511 | rlc_chip_name = "JUNIPER"; |
2512 | smc_chip_name = "JUNIPER"; |
2513 | smc_req_size = roundup2(JUNIPER_SMC_UCODE_SIZE, 4)(((0x5f1f) + ((4) - 1)) & (~((__typeof(0x5f1f))(4) - 1))); |
2514 | break; |
2515 | case CHIP_CYPRESS: |
2516 | case CHIP_HEMLOCK: |
2517 | chip_name = "CYPRESS"; |
2518 | rlc_chip_name = "CYPRESS"; |
2519 | smc_chip_name = "CYPRESS"; |
2520 | smc_req_size = roundup2(CYPRESS_SMC_UCODE_SIZE, 4)(((0x61f7) + ((4) - 1)) & (~((__typeof(0x61f7))(4) - 1))); |
2521 | break; |
2522 | case CHIP_PALM: |
2523 | chip_name = "PALM"; |
2524 | rlc_chip_name = "SUMO"; |
2525 | break; |
2526 | case CHIP_SUMO: |
2527 | chip_name = "SUMO"; |
2528 | rlc_chip_name = "SUMO"; |
2529 | break; |
2530 | case CHIP_SUMO2: |
2531 | chip_name = "SUMO2"; |
2532 | rlc_chip_name = "SUMO"; |
2533 | break; |
2534 | default: BUG()do { panic("BUG at %s:%d", "/usr/src/sys/dev/pci/drm/radeon/r600.c" , 2534); } while (0); |
2535 | } |
2536 | |
2537 | if (rdev->family >= CHIP_CEDAR) { |
2538 | pfp_req_size = EVERGREEN_PFP_UCODE_SIZE1120 * 4; |
2539 | me_req_size = EVERGREEN_PM4_UCODE_SIZE1376 * 4; |
2540 | rlc_req_size = EVERGREEN_RLC_UCODE_SIZE768 * 4; |
2541 | } else if (rdev->family >= CHIP_RV770) { |
2542 | pfp_req_size = R700_PFP_UCODE_SIZE848 * 4; |
2543 | me_req_size = R700_PM4_UCODE_SIZE1360 * 4; |
2544 | rlc_req_size = R700_RLC_UCODE_SIZE1024 * 4; |
2545 | } else { |
2546 | pfp_req_size = R600_PFP_UCODE_SIZE576 * 4; |
2547 | me_req_size = R600_PM4_UCODE_SIZE1792 * 12; |
2548 | rlc_req_size = R600_RLC_UCODE_SIZE768 * 4; |
2549 | } |
2550 | |
2551 | DRM_INFO("Loading %s Microcode\n", chip_name)printk("\0016" "[" "drm" "] " "Loading %s Microcode\n", chip_name ); |
2552 | |
2553 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name); |
2554 | err = request_firmware(&rdev->pfp_fw, fw_name, rdev->dev); |
2555 | if (err) |
2556 | goto out; |
2557 | if (rdev->pfp_fw->size != pfp_req_size) { |
2558 | pr_err("r600_cp: Bogus length %zu in firmware \"%s\"\n",printk("\0013" "r600_cp: Bogus length %zu in firmware \"%s\"\n" , rdev->pfp_fw->size, fw_name) |
2559 | rdev->pfp_fw->size, fw_name)printk("\0013" "r600_cp: Bogus length %zu in firmware \"%s\"\n" , rdev->pfp_fw->size, fw_name); |
2560 | err = -EINVAL22; |
2561 | goto out; |
2562 | } |
2563 | |
2564 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name); |
2565 | err = request_firmware(&rdev->me_fw, fw_name, rdev->dev); |
2566 | if (err) |
2567 | goto out; |
2568 | if (rdev->me_fw->size != me_req_size) { |
2569 | pr_err("r600_cp: Bogus length %zu in firmware \"%s\"\n",printk("\0013" "r600_cp: Bogus length %zu in firmware \"%s\"\n" , rdev->me_fw->size, fw_name) |
2570 | rdev->me_fw->size, fw_name)printk("\0013" "r600_cp: Bogus length %zu in firmware \"%s\"\n" , rdev->me_fw->size, fw_name); |
2571 | err = -EINVAL22; |
2572 | goto out; |
2573 | } |
2574 | |
2575 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", rlc_chip_name); |
2576 | err = request_firmware(&rdev->rlc_fw, fw_name, rdev->dev); |
2577 | if (err) |
2578 | goto out; |
2579 | if (rdev->rlc_fw->size != rlc_req_size) { |
2580 | pr_err("r600_rlc: Bogus length %zu in firmware \"%s\"\n",printk("\0013" "r600_rlc: Bogus length %zu in firmware \"%s\"\n" , rdev->rlc_fw->size, fw_name) |
2581 | rdev->rlc_fw->size, fw_name)printk("\0013" "r600_rlc: Bogus length %zu in firmware \"%s\"\n" , rdev->rlc_fw->size, fw_name); |
2582 | err = -EINVAL22; |
2583 | goto out; |
2584 | } |
2585 | |
2586 | if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_HEMLOCK)) { |
2587 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_smc.bin", smc_chip_name); |
2588 | err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev); |
2589 | if (err) { |
2590 | pr_err("smc: error loading firmware \"%s\"\n", fw_name)printk("\0013" "smc: error loading firmware \"%s\"\n", fw_name ); |
2591 | release_firmware(rdev->smc_fw); |
2592 | rdev->smc_fw = NULL((void *)0); |
2593 | err = 0; |
2594 | } else if (rdev->smc_fw->size != smc_req_size) { |
2595 | pr_err("smc: Bogus length %zu in firmware \"%s\"\n",printk("\0013" "smc: Bogus length %zu in firmware \"%s\"\n", rdev ->smc_fw->size, fw_name) |
2596 | rdev->smc_fw->size, fw_name)printk("\0013" "smc: Bogus length %zu in firmware \"%s\"\n", rdev ->smc_fw->size, fw_name); |
2597 | err = -EINVAL22; |
2598 | } |
2599 | } |
2600 | |
2601 | out: |
2602 | if (err) { |
2603 | if (err != -EINVAL22) |
2604 | pr_err("r600_cp: Failed to load firmware \"%s\"\n",printk("\0013" "r600_cp: Failed to load firmware \"%s\"\n", fw_name ) |
2605 | fw_name)printk("\0013" "r600_cp: Failed to load firmware \"%s\"\n", fw_name ); |
2606 | release_firmware(rdev->pfp_fw); |
2607 | rdev->pfp_fw = NULL((void *)0); |
2608 | release_firmware(rdev->me_fw); |
2609 | rdev->me_fw = NULL((void *)0); |
2610 | release_firmware(rdev->rlc_fw); |
2611 | rdev->rlc_fw = NULL((void *)0); |
2612 | release_firmware(rdev->smc_fw); |
2613 | rdev->smc_fw = NULL((void *)0); |
2614 | } |
2615 | return err; |
2616 | } |
2617 | |
2618 | u32 r600_gfx_get_rptr(struct radeon_device *rdev, |
2619 | struct radeon_ring *ring) |
2620 | { |
2621 | u32 rptr; |
2622 | |
2623 | if (rdev->wb.enabled) |
2624 | rptr = rdev->wb.wb[ring->rptr_offs/4]; |
2625 | else |
2626 | rptr = RREG32(R600_CP_RB_RPTR)r100_mm_rreg(rdev, (0x8700), 0); |
2627 | |
2628 | return rptr; |
2629 | } |
2630 | |
2631 | u32 r600_gfx_get_wptr(struct radeon_device *rdev, |
2632 | struct radeon_ring *ring) |
2633 | { |
2634 | return RREG32(R600_CP_RB_WPTR)r100_mm_rreg(rdev, (0xc114), 0); |
2635 | } |
2636 | |
2637 | void r600_gfx_set_wptr(struct radeon_device *rdev, |
2638 | struct radeon_ring *ring) |
2639 | { |
2640 | WREG32(R600_CP_RB_WPTR, ring->wptr)r100_mm_wreg(rdev, (0xc114), (ring->wptr), 0); |
2641 | (void)RREG32(R600_CP_RB_WPTR)r100_mm_rreg(rdev, (0xc114), 0); |
2642 | } |
2643 | |
2644 | static int r600_cp_load_microcode(struct radeon_device *rdev) |
2645 | { |
2646 | const __be32 *fw_data; |
2647 | int i; |
2648 | |
2649 | if (!rdev->me_fw || !rdev->pfp_fw) |
2650 | return -EINVAL22; |
2651 | |
2652 | r600_cp_stop(rdev); |
2653 | |
2654 | WREG32(CP_RB_CNTL,r100_mm_wreg(rdev, (0xC104), (ifdef __BIG_ENDIAN (2 << 16 ) |endif (1 << 27) | ((15) << 8) | ((3) << 0 )), 0) |
2655 | #ifdef __BIG_ENDIANr100_mm_wreg(rdev, (0xC104), (ifdef __BIG_ENDIAN (2 << 16 ) |endif (1 << 27) | ((15) << 8) | ((3) << 0 )), 0) |
2656 | BUF_SWAP_32BIT |r100_mm_wreg(rdev, (0xC104), (ifdef __BIG_ENDIAN (2 << 16 ) |endif (1 << 27) | ((15) << 8) | ((3) << 0 )), 0) |
2657 | #endifr100_mm_wreg(rdev, (0xC104), (ifdef __BIG_ENDIAN (2 << 16 ) |endif (1 << 27) | ((15) << 8) | ((3) << 0 )), 0) |
2658 | RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3))r100_mm_wreg(rdev, (0xC104), (ifdef __BIG_ENDIAN (2 << 16 ) |endif (1 << 27) | ((15) << 8) | ((3) << 0 )), 0); |
2659 | |
2660 | /* Reset cp */ |
2661 | WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP)r100_mm_wreg(rdev, (0x8020), ((1<<0)), 0); |
2662 | RREG32(GRBM_SOFT_RESET)r100_mm_rreg(rdev, (0x8020), 0); |
2663 | mdelay(15); |
2664 | WREG32(GRBM_SOFT_RESET, 0)r100_mm_wreg(rdev, (0x8020), (0), 0); |
2665 | |
2666 | WREG32(CP_ME_RAM_WADDR, 0)r100_mm_wreg(rdev, (0xC15C), (0), 0); |
2667 | |
2668 | fw_data = (const __be32 *)rdev->me_fw->data; |
2669 | WREG32(CP_ME_RAM_WADDR, 0)r100_mm_wreg(rdev, (0xC15C), (0), 0); |
2670 | for (i = 0; i < R600_PM4_UCODE_SIZE1792 * 3; i++) |
2671 | WREG32(CP_ME_RAM_DATA,r100_mm_wreg(rdev, (0xC160), ((__uint32_t)(__builtin_constant_p (*(__uint32_t *)(fw_data++)) ? (__uint32_t)(((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff) << 24 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff00) << 8 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff0000) >> 8 | ((__uint32_t)(*( __uint32_t *)(fw_data++)) & 0xff000000) >> 24) : __swap32md (*(__uint32_t *)(fw_data++)))), 0) |
2672 | be32_to_cpup(fw_data++))r100_mm_wreg(rdev, (0xC160), ((__uint32_t)(__builtin_constant_p (*(__uint32_t *)(fw_data++)) ? (__uint32_t)(((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff) << 24 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff00) << 8 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff0000) >> 8 | ((__uint32_t)(*( __uint32_t *)(fw_data++)) & 0xff000000) >> 24) : __swap32md (*(__uint32_t *)(fw_data++)))), 0); |
2673 | |
2674 | fw_data = (const __be32 *)rdev->pfp_fw->data; |
2675 | WREG32(CP_PFP_UCODE_ADDR, 0)r100_mm_wreg(rdev, (0xC150), (0), 0); |
2676 | for (i = 0; i < R600_PFP_UCODE_SIZE576; i++) |
2677 | WREG32(CP_PFP_UCODE_DATA,r100_mm_wreg(rdev, (0xC154), ((__uint32_t)(__builtin_constant_p (*(__uint32_t *)(fw_data++)) ? (__uint32_t)(((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff) << 24 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff00) << 8 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff0000) >> 8 | ((__uint32_t)(*( __uint32_t *)(fw_data++)) & 0xff000000) >> 24) : __swap32md (*(__uint32_t *)(fw_data++)))), 0) |
2678 | be32_to_cpup(fw_data++))r100_mm_wreg(rdev, (0xC154), ((__uint32_t)(__builtin_constant_p (*(__uint32_t *)(fw_data++)) ? (__uint32_t)(((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff) << 24 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff00) << 8 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff0000) >> 8 | ((__uint32_t)(*( __uint32_t *)(fw_data++)) & 0xff000000) >> 24) : __swap32md (*(__uint32_t *)(fw_data++)))), 0); |
2679 | |
2680 | WREG32(CP_PFP_UCODE_ADDR, 0)r100_mm_wreg(rdev, (0xC150), (0), 0); |
2681 | WREG32(CP_ME_RAM_WADDR, 0)r100_mm_wreg(rdev, (0xC15C), (0), 0); |
2682 | WREG32(CP_ME_RAM_RADDR, 0)r100_mm_wreg(rdev, (0xC158), (0), 0); |
2683 | return 0; |
2684 | } |
2685 | |
2686 | int r600_cp_start(struct radeon_device *rdev) |
2687 | { |
2688 | struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX0]; |
2689 | int r; |
2690 | uint32_t cp_me; |
2691 | |
2692 | r = radeon_ring_lock(rdev, ring, 7); |
2693 | if (r) { |
2694 | DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r)__drm_err("radeon: cp failed to lock ring (%d).\n", r); |
2695 | return r; |
2696 | } |
2697 | radeon_ring_write(ring, PACKET3(PACKET3_ME_INITIALIZE, 5)((3 << 30) | (((0x44) & 0xFF) << 8) | ((5) & 0x3FFF) << 16)); |
2698 | radeon_ring_write(ring, 0x1); |
2699 | if (rdev->family >= CHIP_RV770) { |
2700 | radeon_ring_write(ring, 0x0); |
2701 | radeon_ring_write(ring, rdev->config.rv770.max_hw_contexts - 1); |
2702 | } else { |
2703 | radeon_ring_write(ring, 0x3); |
2704 | radeon_ring_write(ring, rdev->config.r600.max_hw_contexts - 1); |
2705 | } |
2706 | radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1)((1) << 16)); |
2707 | radeon_ring_write(ring, 0); |
2708 | radeon_ring_write(ring, 0); |
2709 | radeon_ring_unlock_commit(rdev, ring, false0); |
2710 | |
2711 | cp_me = 0xff; |
2712 | WREG32(R_0086D8_CP_ME_CNTL, cp_me)r100_mm_wreg(rdev, (0x86D8), (cp_me), 0); |
2713 | return 0; |
2714 | } |
2715 | |
2716 | int r600_cp_resume(struct radeon_device *rdev) |
2717 | { |
2718 | struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX0]; |
2719 | u32 tmp; |
2720 | u32 rb_bufsz; |
2721 | int r; |
2722 | |
2723 | /* Reset cp */ |
2724 | WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP)r100_mm_wreg(rdev, (0x8020), ((1<<0)), 0); |
2725 | RREG32(GRBM_SOFT_RESET)r100_mm_rreg(rdev, (0x8020), 0); |
2726 | mdelay(15); |
2727 | WREG32(GRBM_SOFT_RESET, 0)r100_mm_wreg(rdev, (0x8020), (0), 0); |
2728 | |
2729 | /* Set ring buffer size */ |
2730 | rb_bufsz = order_base_2(ring->ring_size / 8)drm_order(ring->ring_size / 8); |
2731 | tmp = (order_base_2(RADEON_GPU_PAGE_SIZE/8)drm_order(4096/8) << 8) | rb_bufsz; |
2732 | #ifdef __BIG_ENDIAN |
2733 | tmp |= BUF_SWAP_32BIT(2 << 16); |
2734 | #endif |
2735 | WREG32(CP_RB_CNTL, tmp)r100_mm_wreg(rdev, (0xC104), (tmp), 0); |
2736 | WREG32(CP_SEM_WAIT_TIMER, 0x0)r100_mm_wreg(rdev, (0x85BC), (0x0), 0); |
2737 | |
2738 | /* Set the write pointer delay */ |
2739 | WREG32(CP_RB_WPTR_DELAY, 0)r100_mm_wreg(rdev, (0x8704), (0), 0); |
2740 | |
2741 | /* Initialize the ring buffer's read and write pointers */ |
2742 | WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA)r100_mm_wreg(rdev, (0xC104), (tmp | (1 << 31)), 0); |
2743 | WREG32(CP_RB_RPTR_WR, 0)r100_mm_wreg(rdev, (0xC108), (0), 0); |
2744 | ring->wptr = 0; |
2745 | WREG32(CP_RB_WPTR, ring->wptr)r100_mm_wreg(rdev, (0xC114), (ring->wptr), 0); |
2746 | |
2747 | /* set the wb address whether it's enabled or not */ |
2748 | WREG32(CP_RB_RPTR_ADDR,r100_mm_wreg(rdev, (0xC10C), (((rdev->wb.gpu_addr + 1024) & 0xFFFFFFFC)), 0) |
2749 | ((rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC))r100_mm_wreg(rdev, (0xC10C), (((rdev->wb.gpu_addr + 1024) & 0xFFFFFFFC)), 0); |
2750 | WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF)r100_mm_wreg(rdev, (0xC110), (((u32)(((rdev->wb.gpu_addr + 1024) >> 16) >> 16)) & 0xFF), 0); |
2751 | WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF)r100_mm_wreg(rdev, (0x8544), (((rdev->wb.gpu_addr + 0) >> 8) & 0xFFFFFFFF), 0); |
2752 | |
2753 | if (rdev->wb.enabled) |
2754 | WREG32(SCRATCH_UMSK, 0xff)r100_mm_wreg(rdev, (0x8540), (0xff), 0); |
2755 | else { |
2756 | tmp |= RB_NO_UPDATE(1 << 27); |
2757 | WREG32(SCRATCH_UMSK, 0)r100_mm_wreg(rdev, (0x8540), (0), 0); |
2758 | } |
2759 | |
2760 | mdelay(1); |
2761 | WREG32(CP_RB_CNTL, tmp)r100_mm_wreg(rdev, (0xC104), (tmp), 0); |
2762 | |
2763 | WREG32(CP_RB_BASE, ring->gpu_addr >> 8)r100_mm_wreg(rdev, (0xC100), (ring->gpu_addr >> 8), 0 ); |
2764 | WREG32(CP_DEBUG, (1 << 27) | (1 << 28))r100_mm_wreg(rdev, (0xC1FC), ((1 << 27) | (1 << 28 )), 0); |
2765 | |
2766 | r600_cp_start(rdev); |
2767 | ring->ready = true1; |
2768 | r = radeon_ring_test(rdev, RADEON_RING_TYPE_GFX_INDEX, ring)(rdev)->asic->ring[(0)]->ring_test((rdev), (ring)); |
2769 | if (r) { |
2770 | ring->ready = false0; |
2771 | return r; |
2772 | } |
2773 | |
2774 | if (rdev->asic->copy.copy_ring_index == RADEON_RING_TYPE_GFX_INDEX0) |
2775 | radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); |
2776 | |
2777 | return 0; |
2778 | } |
2779 | |
2780 | void r600_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsigned ring_size) |
2781 | { |
2782 | u32 rb_bufsz; |
2783 | int r; |
2784 | |
2785 | /* Align ring size */ |
2786 | rb_bufsz = order_base_2(ring_size / 8)drm_order(ring_size / 8); |
2787 | ring_size = (1 << (rb_bufsz + 1)) * 4; |
2788 | ring->ring_size = ring_size; |
2789 | ring->align_mask = 16 - 1; |
2790 | |
2791 | if (radeon_ring_supports_scratch_reg(rdev, ring)) { |
2792 | r = radeon_scratch_get(rdev, &ring->rptr_save_reg); |
2793 | if (r) { |
2794 | DRM_ERROR("failed to get scratch reg for rptr save (%d).\n", r)__drm_err("failed to get scratch reg for rptr save (%d).\n", r ); |
2795 | ring->rptr_save_reg = 0; |
2796 | } |
2797 | } |
2798 | } |
2799 | |
2800 | void r600_cp_fini(struct radeon_device *rdev) |
2801 | { |
2802 | struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX0]; |
2803 | r600_cp_stop(rdev); |
2804 | radeon_ring_fini(rdev, ring); |
2805 | radeon_scratch_free(rdev, ring->rptr_save_reg); |
2806 | } |
2807 | |
2808 | /* |
2809 | * GPU scratch registers helpers function. |
2810 | */ |
2811 | void r600_scratch_init(struct radeon_device *rdev) |
2812 | { |
2813 | int i; |
2814 | |
2815 | rdev->scratch.num_reg = 7; |
2816 | rdev->scratch.reg_base = SCRATCH_REG00x8500; |
2817 | for (i = 0; i < rdev->scratch.num_reg; i++) { |
2818 | rdev->scratch.free[i] = true1; |
2819 | rdev->scratch.reg[i] = rdev->scratch.reg_base + (i * 4); |
2820 | } |
2821 | } |
2822 | |
2823 | int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) |
2824 | { |
2825 | uint32_t scratch; |
2826 | uint32_t tmp = 0; |
2827 | unsigned i; |
2828 | int r; |
2829 | |
2830 | r = radeon_scratch_get(rdev, &scratch); |
2831 | if (r) { |
2832 | DRM_ERROR("radeon: cp failed to get scratch reg (%d).\n", r)__drm_err("radeon: cp failed to get scratch reg (%d).\n", r); |
2833 | return r; |
2834 | } |
2835 | WREG32(scratch, 0xCAFEDEAD)r100_mm_wreg(rdev, (scratch), (0xCAFEDEAD), 0); |
2836 | r = radeon_ring_lock(rdev, ring, 3); |
2837 | if (r) { |
2838 | DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", ring->idx, r)__drm_err("radeon: cp failed to lock ring %d (%d).\n", ring-> idx, r); |
2839 | radeon_scratch_free(rdev, scratch); |
2840 | return r; |
2841 | } |
2842 | radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
2843 | radeon_ring_write(ring, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2)); |
2844 | radeon_ring_write(ring, 0xDEADBEEF); |
2845 | radeon_ring_unlock_commit(rdev, ring, false0); |
2846 | for (i = 0; i < rdev->usec_timeout; i++) { |
2847 | tmp = RREG32(scratch)r100_mm_rreg(rdev, (scratch), 0); |
2848 | if (tmp == 0xDEADBEEF) |
2849 | break; |
2850 | udelay(1); |
2851 | } |
2852 | if (i < rdev->usec_timeout) { |
2853 | DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i)printk("\0016" "[" "drm" "] " "ring test on %d succeeded in %d usecs\n" , ring->idx, i); |
2854 | } else { |
2855 | DRM_ERROR("radeon: ring %d test failed (scratch(0x%04X)=0x%08X)\n",__drm_err("radeon: ring %d test failed (scratch(0x%04X)=0x%08X)\n" , ring->idx, scratch, tmp) |
2856 | ring->idx, scratch, tmp)__drm_err("radeon: ring %d test failed (scratch(0x%04X)=0x%08X)\n" , ring->idx, scratch, tmp); |
2857 | r = -EINVAL22; |
2858 | } |
2859 | radeon_scratch_free(rdev, scratch); |
2860 | return r; |
2861 | } |
2862 | |
2863 | /* |
2864 | * CP fences/semaphores |
2865 | */ |
2866 | |
2867 | void r600_fence_ring_emit(struct radeon_device *rdev, |
2868 | struct radeon_fence *fence) |
2869 | { |
2870 | struct radeon_ring *ring = &rdev->ring[fence->ring]; |
2871 | u32 cp_coher_cntl = PACKET3_TC_ACTION_ENA(1 << 23) | PACKET3_VC_ACTION_ENA(1 << 24) | |
2872 | PACKET3_SH_ACTION_ENA(1 << 27); |
2873 | |
2874 | if (rdev->family >= CHIP_RV770) |
2875 | cp_coher_cntl |= PACKET3_FULL_CACHE_ENA(1 << 20); |
2876 | |
2877 | if (rdev->wb.use_event) { |
2878 | u64 addr = rdev->fence_drv[fence->ring].gpu_addr; |
2879 | /* flush read cache over gart */ |
2880 | radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)((3 << 30) | (((0x43) & 0xFF) << 8) | ((3) & 0x3FFF) << 16)); |
2881 | radeon_ring_write(ring, cp_coher_cntl); |
2882 | radeon_ring_write(ring, 0xFFFFFFFF); |
2883 | radeon_ring_write(ring, 0); |
2884 | radeon_ring_write(ring, 10); /* poll interval */ |
2885 | /* EVENT_WRITE_EOP - flush caches, send int */ |
2886 | radeon_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)((3 << 30) | (((0x47) & 0xFF) << 8) | ((4) & 0x3FFF) << 16)); |
2887 | radeon_ring_write(ring, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT_TS)(((0x14 << 0)) << 0) | EVENT_INDEX(5)((5) << 8)); |
2888 | radeon_ring_write(ring, lower_32_bits(addr)((u32)(addr))); |
2889 | radeon_ring_write(ring, (upper_32_bits(addr)((u32)(((addr) >> 16) >> 16)) & 0xff) | DATA_SEL(1)((1) << 29) | INT_SEL(2)((2) << 24)); |
2890 | radeon_ring_write(ring, fence->seq); |
2891 | radeon_ring_write(ring, 0); |
2892 | } else { |
2893 | /* flush read cache over gart */ |
2894 | radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)((3 << 30) | (((0x43) & 0xFF) << 8) | ((3) & 0x3FFF) << 16)); |
2895 | radeon_ring_write(ring, cp_coher_cntl); |
2896 | radeon_ring_write(ring, 0xFFFFFFFF); |
2897 | radeon_ring_write(ring, 0); |
2898 | radeon_ring_write(ring, 10); /* poll interval */ |
2899 | radeon_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE, 0)((3 << 30) | (((0x46) & 0xFF) << 8) | ((0) & 0x3FFF) << 16)); |
2900 | radeon_ring_write(ring, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT)(((0x16 << 0)) << 0) | EVENT_INDEX(0)((0) << 8)); |
2901 | /* wait for 3D idle clean */ |
2902 | radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
2903 | radeon_ring_write(ring, (WAIT_UNTIL0x8040 - PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2); |
2904 | radeon_ring_write(ring, WAIT_3D_IDLE_bit(1 << 15) | WAIT_3D_IDLECLEAN_bit(1 << 17)); |
2905 | /* Emit fence sequence & fire IRQ */ |
2906 | radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
2907 | radeon_ring_write(ring, ((rdev->fence_drv[fence->ring].scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2)); |
2908 | radeon_ring_write(ring, fence->seq); |
2909 | /* CP_INTERRUPT packet 3 no longer exists, use packet 0 */ |
2910 | radeon_ring_write(ring, PACKET0(CP_INT_STATUS, 0)((0 << 30) | (((0xc128) >> 2) & 0xFFFF) | ((0 ) & 0x3FFF) << 16)); |
2911 | radeon_ring_write(ring, RB_INT_STAT(1 << 31)); |
2912 | } |
2913 | } |
2914 | |
2915 | /** |
2916 | * r600_semaphore_ring_emit - emit a semaphore on the CP ring |
2917 | * |
2918 | * @rdev: radeon_device pointer |
2919 | * @ring: radeon ring buffer object |
2920 | * @semaphore: radeon semaphore object |
2921 | * @emit_wait: Is this a sempahore wait? |
2922 | * |
2923 | * Emits a semaphore signal/wait packet to the CP ring and prevents the PFP |
2924 | * from running ahead of semaphore waits. |
2925 | */ |
2926 | bool_Bool r600_semaphore_ring_emit(struct radeon_device *rdev, |
2927 | struct radeon_ring *ring, |
2928 | struct radeon_semaphore *semaphore, |
2929 | bool_Bool emit_wait) |
2930 | { |
2931 | uint64_t addr = semaphore->gpu_addr; |
2932 | unsigned sel = emit_wait ? PACKET3_SEM_SEL_WAIT(0x7 << 29) : PACKET3_SEM_SEL_SIGNAL(0x6 << 29); |
2933 | |
2934 | if (rdev->family < CHIP_CAYMAN) |
2935 | sel |= PACKET3_SEM_WAIT_ON_SIGNAL(0x1 << 12); |
2936 | |
2937 | radeon_ring_write(ring, PACKET3(PACKET3_MEM_SEMAPHORE, 1)((3 << 30) | (((0x39) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
2938 | radeon_ring_write(ring, lower_32_bits(addr)((u32)(addr))); |
2939 | radeon_ring_write(ring, (upper_32_bits(addr)((u32)(((addr) >> 16) >> 16)) & 0xff) | sel); |
2940 | |
2941 | /* PFP_SYNC_ME packet only exists on 7xx+, only enable it on eg+ */ |
2942 | if (emit_wait && (rdev->family >= CHIP_CEDAR)) { |
2943 | /* Prevent the PFP from running ahead of the semaphore wait */ |
2944 | radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0)((3 << 30) | (((0x42) & 0xFF) << 8) | ((0) & 0x3FFF) << 16)); |
2945 | radeon_ring_write(ring, 0x0); |
2946 | } |
2947 | |
2948 | return true1; |
2949 | } |
2950 | |
2951 | /** |
2952 | * r600_copy_cpdma - copy pages using the CP DMA engine |
2953 | * |
2954 | * @rdev: radeon_device pointer |
2955 | * @src_offset: src GPU address |
2956 | * @dst_offset: dst GPU address |
2957 | * @num_gpu_pages: number of GPU pages to xfer |
2958 | * @resv: DMA reservation object to manage fences |
2959 | * |
2960 | * Copy GPU paging using the CP DMA engine (r6xx+). |
2961 | * Used by the radeon ttm implementation to move pages if |
2962 | * registered as the asic copy callback. |
2963 | */ |
2964 | struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev, |
2965 | uint64_t src_offset, uint64_t dst_offset, |
2966 | unsigned num_gpu_pages, |
2967 | struct dma_resv *resv) |
2968 | { |
2969 | struct radeon_fence *fence; |
2970 | struct radeon_sync sync; |
2971 | int ring_index = rdev->asic->copy.blit_ring_index; |
2972 | struct radeon_ring *ring = &rdev->ring[ring_index]; |
2973 | u32 size_in_bytes, cur_size_in_bytes, tmp; |
2974 | int i, num_loops; |
2975 | int r = 0; |
2976 | |
2977 | radeon_sync_create(&sync); |
2978 | |
2979 | size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT12); |
2980 | num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff)(((size_in_bytes) + ((0x1fffff) - 1)) / (0x1fffff)); |
2981 | r = radeon_ring_lock(rdev, ring, num_loops * 6 + 24); |
2982 | if (r) { |
2983 | DRM_ERROR("radeon: moving bo (%d).\n", r)__drm_err("radeon: moving bo (%d).\n", r); |
2984 | radeon_sync_free(rdev, &sync, NULL((void *)0)); |
2985 | return ERR_PTR(r); |
2986 | } |
2987 | |
2988 | radeon_sync_resv(rdev, &sync, resv, false0); |
2989 | radeon_sync_rings(rdev, &sync, ring->idx); |
2990 | |
2991 | radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
2992 | radeon_ring_write(ring, (WAIT_UNTIL0x8040 - PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2); |
2993 | radeon_ring_write(ring, WAIT_3D_IDLE_bit(1 << 15)); |
2994 | for (i = 0; i < num_loops; i++) { |
2995 | cur_size_in_bytes = size_in_bytes; |
2996 | if (cur_size_in_bytes > 0x1fffff) |
2997 | cur_size_in_bytes = 0x1fffff; |
2998 | size_in_bytes -= cur_size_in_bytes; |
2999 | tmp = upper_32_bits(src_offset)((u32)(((src_offset) >> 16) >> 16)) & 0xff; |
3000 | if (size_in_bytes == 0) |
3001 | tmp |= PACKET3_CP_DMA_CP_SYNC(1 << 31); |
3002 | radeon_ring_write(ring, PACKET3(PACKET3_CP_DMA, 4)((3 << 30) | (((0x41) & 0xFF) << 8) | ((4) & 0x3FFF) << 16)); |
3003 | radeon_ring_write(ring, lower_32_bits(src_offset)((u32)(src_offset))); |
3004 | radeon_ring_write(ring, tmp); |
3005 | radeon_ring_write(ring, lower_32_bits(dst_offset)((u32)(dst_offset))); |
3006 | radeon_ring_write(ring, upper_32_bits(dst_offset)((u32)(((dst_offset) >> 16) >> 16)) & 0xff); |
3007 | radeon_ring_write(ring, cur_size_in_bytes); |
3008 | src_offset += cur_size_in_bytes; |
3009 | dst_offset += cur_size_in_bytes; |
3010 | } |
3011 | radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
3012 | radeon_ring_write(ring, (WAIT_UNTIL0x8040 - PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2); |
3013 | radeon_ring_write(ring, WAIT_CP_DMA_IDLE_bit(1 << 8)); |
3014 | |
3015 | r = radeon_fence_emit(rdev, &fence, ring->idx); |
3016 | if (r) { |
3017 | radeon_ring_unlock_undo(rdev, ring); |
3018 | radeon_sync_free(rdev, &sync, NULL((void *)0)); |
3019 | return ERR_PTR(r); |
3020 | } |
3021 | |
3022 | radeon_ring_unlock_commit(rdev, ring, false0); |
3023 | radeon_sync_free(rdev, &sync, fence); |
3024 | |
3025 | return fence; |
3026 | } |
3027 | |
3028 | int r600_set_surface_reg(struct radeon_device *rdev, int reg, |
3029 | uint32_t tiling_flags, uint32_t pitch, |
3030 | uint32_t offset, uint32_t obj_size) |
3031 | { |
3032 | /* FIXME: implement */ |
3033 | return 0; |
3034 | } |
3035 | |
3036 | void r600_clear_surface_reg(struct radeon_device *rdev, int reg) |
3037 | { |
3038 | /* FIXME: implement */ |
3039 | } |
3040 | |
3041 | static void r600_uvd_init(struct radeon_device *rdev) |
3042 | { |
3043 | int r; |
3044 | |
3045 | if (!rdev->has_uvd) |
3046 | return; |
3047 | |
3048 | r = radeon_uvd_init(rdev); |
3049 | if (r) { |
3050 | dev_err(rdev->dev, "failed UVD (%d) init.\n", r)printf("drm:pid%d:%s *ERROR* " "failed UVD (%d) init.\n", ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci ) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci; })->ci_curproc->p_p->ps_pid, __func__ , r); |
3051 | /* |
3052 | * At this point rdev->uvd.vcpu_bo is NULL which trickles down |
3053 | * to early fails uvd_v1_0_resume() and thus nothing happens |
3054 | * there. So it is pointless to try to go through that code |
3055 | * hence why we disable uvd here. |
3056 | */ |
3057 | rdev->has_uvd = false0; |
3058 | return; |
3059 | } |
3060 | rdev->ring[R600_RING_TYPE_UVD_INDEX5].ring_obj = NULL((void *)0); |
3061 | r600_ring_init(rdev, &rdev->ring[R600_RING_TYPE_UVD_INDEX5], 4096); |
3062 | } |
3063 | |
3064 | static void r600_uvd_start(struct radeon_device *rdev) |
3065 | { |
3066 | int r; |
3067 | |
3068 | if (!rdev->has_uvd) |
3069 | return; |
3070 | |
3071 | r = uvd_v1_0_resume(rdev); |
3072 | if (r) { |
3073 | dev_err(rdev->dev, "failed UVD resume (%d).\n", r)printf("drm:pid%d:%s *ERROR* " "failed UVD resume (%d).\n", ( {struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , r); |
3074 | goto error; |
3075 | } |
3076 | r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX5); |
3077 | if (r) { |
3078 | dev_err(rdev->dev, "failed initializing UVD fences (%d).\n", r)printf("drm:pid%d:%s *ERROR* " "failed initializing UVD fences (%d).\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , r); |
3079 | goto error; |
3080 | } |
3081 | return; |
3082 | |
3083 | error: |
3084 | rdev->ring[R600_RING_TYPE_UVD_INDEX5].ring_size = 0; |
3085 | } |
3086 | |
3087 | static void r600_uvd_resume(struct radeon_device *rdev) |
3088 | { |
3089 | struct radeon_ring *ring; |
3090 | int r; |
3091 | |
3092 | if (!rdev->has_uvd || !rdev->ring[R600_RING_TYPE_UVD_INDEX5].ring_size) |
3093 | return; |
3094 | |
3095 | ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX5]; |
3096 | r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)((0 << 30) | (((0xeffc) >> 2) & 0xFFFF) | ((0 ) & 0x3FFF) << 16)); |
3097 | if (r) { |
3098 | dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r)printf("drm:pid%d:%s *ERROR* " "failed initializing UVD ring (%d).\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , r); |
3099 | return; |
3100 | } |
3101 | r = uvd_v1_0_init(rdev); |
3102 | if (r) { |
3103 | dev_err(rdev->dev, "failed initializing UVD (%d).\n", r)printf("drm:pid%d:%s *ERROR* " "failed initializing UVD (%d).\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , r); |
3104 | return; |
3105 | } |
3106 | } |
3107 | |
3108 | static int r600_startup(struct radeon_device *rdev) |
3109 | { |
3110 | struct radeon_ring *ring; |
3111 | int r; |
3112 | |
3113 | /* enable pcie gen2 link */ |
3114 | r600_pcie_gen2_enable(rdev); |
3115 | |
3116 | /* scratch needs to be initialized before MC */ |
3117 | r = r600_vram_scratch_init(rdev); |
3118 | if (r) |
3119 | return r; |
3120 | |
3121 | r600_mc_program(rdev); |
3122 | |
3123 | if (rdev->flags & RADEON_IS_AGP) { |
3124 | r600_agp_enable(rdev); |
3125 | } else { |
3126 | r = r600_pcie_gart_enable(rdev); |
3127 | if (r) |
3128 | return r; |
3129 | } |
3130 | r600_gpu_init(rdev); |
3131 | |
3132 | /* allocate wb buffer */ |
3133 | r = radeon_wb_init(rdev); |
3134 | if (r) |
3135 | return r; |
3136 | |
3137 | r = radeon_fence_driver_start_ring(rdev, RADEON_RING_TYPE_GFX_INDEX0); |
3138 | if (r) { |
3139 | dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r)printf("drm:pid%d:%s *ERROR* " "failed initializing CP fences (%d).\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , r); |
3140 | return r; |
3141 | } |
3142 | |
3143 | r600_uvd_start(rdev); |
3144 | |
3145 | /* Enable IRQ */ |
3146 | if (!rdev->irq.installed) { |
3147 | r = radeon_irq_kms_init(rdev); |
3148 | if (r) |
3149 | return r; |
3150 | } |
3151 | |
3152 | r = r600_irq_init(rdev); |
3153 | if (r) { |
3154 | DRM_ERROR("radeon: IH init failed (%d).\n", r)__drm_err("radeon: IH init failed (%d).\n", r); |
3155 | radeon_irq_kms_fini(rdev); |
3156 | return r; |
3157 | } |
3158 | r600_irq_set(rdev); |
3159 | |
3160 | ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX0]; |
3161 | r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET1024, |
3162 | RADEON_CP_PACKET20x80000000); |
3163 | if (r) |
3164 | return r; |
3165 | |
3166 | r = r600_cp_load_microcode(rdev); |
3167 | if (r) |
3168 | return r; |
3169 | r = r600_cp_resume(rdev); |
3170 | if (r) |
3171 | return r; |
3172 | |
3173 | r600_uvd_resume(rdev); |
3174 | |
3175 | r = radeon_ib_pool_init(rdev); |
3176 | if (r) { |
3177 | dev_err(rdev->dev, "IB initialization failed (%d).\n", r)printf("drm:pid%d:%s *ERROR* " "IB initialization failed (%d).\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , r); |
3178 | return r; |
3179 | } |
3180 | |
3181 | r = radeon_audio_init(rdev); |
3182 | if (r) { |
3183 | DRM_ERROR("radeon: audio init failed\n")__drm_err("radeon: audio init failed\n"); |
3184 | return r; |
3185 | } |
3186 | |
3187 | return 0; |
3188 | } |
3189 | |
3190 | void r600_vga_set_state(struct radeon_device *rdev, bool_Bool state) |
3191 | { |
3192 | uint32_t temp; |
3193 | |
3194 | temp = RREG32(CONFIG_CNTL)r100_mm_rreg(rdev, (0x5424), 0); |
3195 | if (!state) { |
3196 | temp &= ~(1<<0); |
3197 | temp |= (1<<1); |
3198 | } else { |
3199 | temp &= ~(1<<1); |
3200 | } |
3201 | WREG32(CONFIG_CNTL, temp)r100_mm_wreg(rdev, (0x5424), (temp), 0); |
3202 | } |
3203 | |
3204 | int r600_resume(struct radeon_device *rdev) |
3205 | { |
3206 | int r; |
3207 | |
3208 | /* Do not reset GPU before posting, on r600 hw unlike on r500 hw, |
3209 | * posting will perform necessary task to bring back GPU into good |
3210 | * shape. |
3211 | */ |
3212 | /* post card */ |
3213 | atom_asic_init(rdev->mode_info.atom_context); |
3214 | |
3215 | if (rdev->pm.pm_method == PM_METHOD_DPM) |
3216 | radeon_pm_resume(rdev); |
3217 | |
3218 | rdev->accel_working = true1; |
3219 | r = r600_startup(rdev); |
3220 | if (r) { |
3221 | DRM_ERROR("r600 startup failed on resume\n")__drm_err("r600 startup failed on resume\n"); |
3222 | rdev->accel_working = false0; |
3223 | return r; |
3224 | } |
3225 | |
3226 | return r; |
3227 | } |
3228 | |
3229 | int r600_suspend(struct radeon_device *rdev) |
3230 | { |
3231 | radeon_pm_suspend(rdev); |
3232 | radeon_audio_fini(rdev); |
3233 | r600_cp_stop(rdev); |
3234 | if (rdev->has_uvd) { |
3235 | radeon_uvd_suspend(rdev); |
3236 | uvd_v1_0_fini(rdev); |
3237 | } |
3238 | r600_irq_suspend(rdev); |
3239 | radeon_wb_disable(rdev); |
3240 | r600_pcie_gart_disable(rdev); |
3241 | |
3242 | return 0; |
3243 | } |
3244 | |
3245 | /* Plan is to move initialization in that function and use |
3246 | * helper function so that radeon_device_init pretty much |
3247 | * do nothing more than calling asic specific function. This |
3248 | * should also allow to remove a bunch of callback function |
3249 | * like vram_info. |
3250 | */ |
3251 | int r600_init(struct radeon_device *rdev) |
3252 | { |
3253 | int r; |
3254 | |
3255 | r600_debugfs_mc_info_init(rdev); |
3256 | /* Read BIOS */ |
3257 | if (!radeon_get_bios(rdev)) { |
3258 | if (ASIC_IS_AVIVO(rdev)((rdev->family >= CHIP_RS600))) |
3259 | return -EINVAL22; |
3260 | } |
3261 | /* Must be an ATOMBIOS */ |
3262 | if (!rdev->is_atom_bios) { |
3263 | dev_err(rdev->dev, "Expecting atombios for R600 GPU\n")printf("drm:pid%d:%s *ERROR* " "Expecting atombios for R600 GPU\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
3264 | return -EINVAL22; |
3265 | } |
3266 | r = radeon_atombios_init(rdev); |
3267 | if (r) |
3268 | return r; |
3269 | /* Post card if necessary */ |
3270 | if (!radeon_card_posted(rdev)) { |
3271 | if (!rdev->bios) { |
3272 | dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n")printf("drm:pid%d:%s *ERROR* " "Card not posted and no BIOS - ignoring\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
3273 | return -EINVAL22; |
3274 | } |
3275 | DRM_INFO("GPU not posted. posting now...\n")printk("\0016" "[" "drm" "] " "GPU not posted. posting now...\n" ); |
3276 | atom_asic_init(rdev->mode_info.atom_context); |
3277 | } |
3278 | /* Initialize scratch registers */ |
3279 | r600_scratch_init(rdev); |
3280 | /* Initialize surface registers */ |
3281 | radeon_surface_init(rdev); |
3282 | /* Initialize clocks */ |
3283 | radeon_get_clock_info(rdev->ddev); |
3284 | /* Fence driver */ |
3285 | radeon_fence_driver_init(rdev); |
3286 | if (rdev->flags & RADEON_IS_AGP) { |
3287 | r = radeon_agp_init(rdev); |
3288 | if (r) |
3289 | radeon_agp_disable(rdev); |
3290 | } |
3291 | r = r600_mc_init(rdev); |
3292 | if (r) |
3293 | return r; |
3294 | /* Memory manager */ |
3295 | r = radeon_bo_init(rdev); |
3296 | if (r) |
3297 | return r; |
3298 | |
3299 | if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { |
3300 | r = r600_init_microcode(rdev); |
3301 | if (r) { |
3302 | DRM_ERROR("Failed to load firmware!\n")__drm_err("Failed to load firmware!\n"); |
3303 | return r; |
3304 | } |
3305 | } |
3306 | |
3307 | /* Initialize power management */ |
3308 | radeon_pm_init(rdev); |
3309 | |
3310 | rdev->ring[RADEON_RING_TYPE_GFX_INDEX0].ring_obj = NULL((void *)0); |
3311 | r600_ring_init(rdev, &rdev->ring[RADEON_RING_TYPE_GFX_INDEX0], 1024 * 1024); |
3312 | |
3313 | r600_uvd_init(rdev); |
3314 | |
3315 | rdev->ih.ring_obj = NULL((void *)0); |
3316 | r600_ih_ring_init(rdev, 64 * 1024); |
3317 | |
3318 | r = r600_pcie_gart_init(rdev); |
3319 | if (r) |
3320 | return r; |
3321 | |
3322 | rdev->accel_working = true1; |
3323 | r = r600_startup(rdev); |
3324 | if (r) { |
3325 | dev_err(rdev->dev, "disabling GPU acceleration\n")printf("drm:pid%d:%s *ERROR* " "disabling GPU acceleration\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__); |
3326 | r600_cp_fini(rdev); |
3327 | r600_irq_fini(rdev); |
3328 | radeon_wb_fini(rdev); |
3329 | radeon_ib_pool_fini(rdev); |
3330 | radeon_irq_kms_fini(rdev); |
3331 | r600_pcie_gart_fini(rdev); |
3332 | rdev->accel_working = false0; |
3333 | } |
3334 | |
3335 | return 0; |
3336 | } |
3337 | |
3338 | void r600_fini(struct radeon_device *rdev) |
3339 | { |
3340 | radeon_pm_fini(rdev); |
3341 | radeon_audio_fini(rdev); |
3342 | r600_cp_fini(rdev); |
3343 | r600_irq_fini(rdev); |
3344 | if (rdev->has_uvd) { |
3345 | uvd_v1_0_fini(rdev); |
3346 | radeon_uvd_fini(rdev); |
3347 | } |
3348 | radeon_wb_fini(rdev); |
3349 | radeon_ib_pool_fini(rdev); |
3350 | radeon_irq_kms_fini(rdev); |
3351 | r600_pcie_gart_fini(rdev); |
3352 | r600_vram_scratch_fini(rdev); |
3353 | radeon_agp_fini(rdev); |
3354 | radeon_gem_fini(rdev); |
3355 | radeon_fence_driver_fini(rdev); |
3356 | radeon_bo_fini(rdev); |
3357 | radeon_atombios_fini(rdev); |
3358 | kfree(rdev->bios); |
3359 | rdev->bios = NULL((void *)0); |
3360 | } |
3361 | |
3362 | |
3363 | /* |
3364 | * CS stuff |
3365 | */ |
3366 | void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) |
3367 | { |
3368 | struct radeon_ring *ring = &rdev->ring[ib->ring]; |
3369 | u32 next_rptr; |
3370 | |
3371 | if (ring->rptr_save_reg) { |
3372 | next_rptr = ring->wptr + 3 + 4; |
3373 | radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16)); |
3374 | radeon_ring_write(ring, ((ring->rptr_save_reg - |
3375 | PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2)); |
3376 | radeon_ring_write(ring, next_rptr); |
3377 | } else if (rdev->wb.enabled) { |
3378 | next_rptr = ring->wptr + 5 + 4; |
3379 | radeon_ring_write(ring, PACKET3(PACKET3_MEM_WRITE, 3)((3 << 30) | (((0x3D) & 0xFF) << 8) | ((3) & 0x3FFF) << 16)); |
3380 | radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); |
3381 | radeon_ring_write(ring, (upper_32_bits(ring->next_rptr_gpu_addr)((u32)(((ring->next_rptr_gpu_addr) >> 16) >> 16 )) & 0xff) | (1 << 18)); |
3382 | radeon_ring_write(ring, next_rptr); |
3383 | radeon_ring_write(ring, 0); |
3384 | } |
3385 | |
3386 | radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)((3 << 30) | (((0x32) & 0xFF) << 8) | ((2) & 0x3FFF) << 16)); |
3387 | radeon_ring_write(ring, |
3388 | #ifdef __BIG_ENDIAN |
3389 | (2 << 0) | |
3390 | #endif |
3391 | (ib->gpu_addr & 0xFFFFFFFC)); |
3392 | radeon_ring_write(ring, upper_32_bits(ib->gpu_addr)((u32)(((ib->gpu_addr) >> 16) >> 16)) & 0xFF); |
3393 | radeon_ring_write(ring, ib->length_dw); |
3394 | } |
3395 | |
3396 | int r600_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) |
3397 | { |
3398 | struct radeon_ib ib; |
3399 | uint32_t scratch; |
3400 | uint32_t tmp = 0; |
3401 | unsigned i; |
3402 | int r; |
3403 | |
3404 | r = radeon_scratch_get(rdev, &scratch); |
3405 | if (r) { |
3406 | DRM_ERROR("radeon: failed to get scratch reg (%d).\n", r)__drm_err("radeon: failed to get scratch reg (%d).\n", r); |
3407 | return r; |
3408 | } |
3409 | WREG32(scratch, 0xCAFEDEAD)r100_mm_wreg(rdev, (scratch), (0xCAFEDEAD), 0); |
3410 | r = radeon_ib_get(rdev, ring->idx, &ib, NULL((void *)0), 256); |
3411 | if (r) { |
3412 | DRM_ERROR("radeon: failed to get ib (%d).\n", r)__drm_err("radeon: failed to get ib (%d).\n", r); |
3413 | goto free_scratch; |
3414 | } |
3415 | ib.ptr[0] = PACKET3(PACKET3_SET_CONFIG_REG, 1)((3 << 30) | (((0x68) & 0xFF) << 8) | ((1) & 0x3FFF) << 16); |
3416 | ib.ptr[1] = ((scratch - PACKET3_SET_CONFIG_REG_OFFSET0x00008000) >> 2); |
3417 | ib.ptr[2] = 0xDEADBEEF; |
3418 | ib.length_dw = 3; |
3419 | r = radeon_ib_schedule(rdev, &ib, NULL((void *)0), false0); |
3420 | if (r) { |
3421 | DRM_ERROR("radeon: failed to schedule ib (%d).\n", r)__drm_err("radeon: failed to schedule ib (%d).\n", r); |
3422 | goto free_ib; |
3423 | } |
3424 | r = radeon_fence_wait_timeout(ib.fence, false0, usecs_to_jiffies((((uint64_t)(1000000)) * hz / 1000000) |
3425 | RADEON_USEC_IB_TEST_TIMEOUT)(((uint64_t)(1000000)) * hz / 1000000)); |
3426 | if (r < 0) { |
3427 | DRM_ERROR("radeon: fence wait failed (%d).\n", r)__drm_err("radeon: fence wait failed (%d).\n", r); |
3428 | goto free_ib; |
3429 | } else if (r == 0) { |
3430 | DRM_ERROR("radeon: fence wait timed out.\n")__drm_err("radeon: fence wait timed out.\n"); |
3431 | r = -ETIMEDOUT60; |
3432 | goto free_ib; |
3433 | } |
3434 | r = 0; |
3435 | for (i = 0; i < rdev->usec_timeout; i++) { |
3436 | tmp = RREG32(scratch)r100_mm_rreg(rdev, (scratch), 0); |
3437 | if (tmp == 0xDEADBEEF) |
3438 | break; |
3439 | udelay(1); |
3440 | } |
3441 | if (i < rdev->usec_timeout) { |
3442 | DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i)printk("\0016" "[" "drm" "] " "ib test on ring %d succeeded in %u usecs\n" , ib.fence->ring, i); |
3443 | } else { |
3444 | DRM_ERROR("radeon: ib test failed (scratch(0x%04X)=0x%08X)\n",__drm_err("radeon: ib test failed (scratch(0x%04X)=0x%08X)\n" , scratch, tmp) |
3445 | scratch, tmp)__drm_err("radeon: ib test failed (scratch(0x%04X)=0x%08X)\n" , scratch, tmp); |
3446 | r = -EINVAL22; |
3447 | } |
3448 | free_ib: |
3449 | radeon_ib_free(rdev, &ib); |
3450 | free_scratch: |
3451 | radeon_scratch_free(rdev, scratch); |
3452 | return r; |
3453 | } |
3454 | |
3455 | /* |
3456 | * Interrupts |
3457 | * |
3458 | * Interrupts use a ring buffer on r6xx/r7xx hardware. It works pretty |
3459 | * the same as the CP ring buffer, but in reverse. Rather than the CPU |
3460 | * writing to the ring and the GPU consuming, the GPU writes to the ring |
3461 | * and host consumes. As the host irq handler processes interrupts, it |
3462 | * increments the rptr. When the rptr catches up with the wptr, all the |
3463 | * current interrupts have been processed. |
3464 | */ |
3465 | |
3466 | void r600_ih_ring_init(struct radeon_device *rdev, unsigned ring_size) |
3467 | { |
3468 | u32 rb_bufsz; |
3469 | |
3470 | /* Align ring size */ |
3471 | rb_bufsz = order_base_2(ring_size / 4)drm_order(ring_size / 4); |
3472 | ring_size = (1 << rb_bufsz) * 4; |
3473 | rdev->ih.ring_size = ring_size; |
3474 | rdev->ih.ptr_mask = rdev->ih.ring_size - 1; |
3475 | rdev->ih.rptr = 0; |
3476 | } |
3477 | |
3478 | int r600_ih_ring_alloc(struct radeon_device *rdev) |
3479 | { |
3480 | int r; |
3481 | |
3482 | /* Allocate ring buffer */ |
3483 | if (rdev->ih.ring_obj == NULL((void *)0)) { |
3484 | r = radeon_bo_create(rdev, rdev->ih.ring_size, |
3485 | PAGE_SIZE(1 << 12), true1, |
3486 | RADEON_GEM_DOMAIN_GTT0x2, 0, |
3487 | NULL((void *)0), NULL((void *)0), &rdev->ih.ring_obj); |
3488 | if (r) { |
3489 | DRM_ERROR("radeon: failed to create ih ring buffer (%d).\n", r)__drm_err("radeon: failed to create ih ring buffer (%d).\n", r ); |
3490 | return r; |
3491 | } |
3492 | r = radeon_bo_reserve(rdev->ih.ring_obj, false0); |
3493 | if (unlikely(r != 0)__builtin_expect(!!(r != 0), 0)) |
3494 | return r; |
3495 | r = radeon_bo_pin(rdev->ih.ring_obj, |
3496 | RADEON_GEM_DOMAIN_GTT0x2, |
3497 | &rdev->ih.gpu_addr); |
3498 | if (r) { |
3499 | radeon_bo_unreserve(rdev->ih.ring_obj); |
3500 | DRM_ERROR("radeon: failed to pin ih ring buffer (%d).\n", r)__drm_err("radeon: failed to pin ih ring buffer (%d).\n", r); |
3501 | return r; |
3502 | } |
3503 | r = radeon_bo_kmap(rdev->ih.ring_obj, |
3504 | (void **)&rdev->ih.ring); |
3505 | radeon_bo_unreserve(rdev->ih.ring_obj); |
3506 | if (r) { |
3507 | DRM_ERROR("radeon: failed to map ih ring buffer (%d).\n", r)__drm_err("radeon: failed to map ih ring buffer (%d).\n", r); |
3508 | return r; |
3509 | } |
3510 | } |
3511 | return 0; |
3512 | } |
3513 | |
3514 | void r600_ih_ring_fini(struct radeon_device *rdev) |
3515 | { |
3516 | int r; |
3517 | if (rdev->ih.ring_obj) { |
3518 | r = radeon_bo_reserve(rdev->ih.ring_obj, false0); |
3519 | if (likely(r == 0)__builtin_expect(!!(r == 0), 1)) { |
3520 | radeon_bo_kunmap(rdev->ih.ring_obj); |
3521 | radeon_bo_unpin(rdev->ih.ring_obj); |
3522 | radeon_bo_unreserve(rdev->ih.ring_obj); |
3523 | } |
3524 | radeon_bo_unref(&rdev->ih.ring_obj); |
3525 | rdev->ih.ring = NULL((void *)0); |
3526 | rdev->ih.ring_obj = NULL((void *)0); |
3527 | } |
3528 | } |
3529 | |
3530 | void r600_rlc_stop(struct radeon_device *rdev) |
3531 | { |
3532 | |
3533 | if ((rdev->family >= CHIP_RV770) && |
3534 | (rdev->family <= CHIP_RV740)) { |
3535 | /* r7xx asics need to soft reset RLC before halting */ |
3536 | WREG32(SRBM_SOFT_RESET, SOFT_RESET_RLC)r100_mm_wreg(rdev, (0xe60), ((1 << 13)), 0); |
3537 | RREG32(SRBM_SOFT_RESET)r100_mm_rreg(rdev, (0xe60), 0); |
3538 | mdelay(15); |
3539 | WREG32(SRBM_SOFT_RESET, 0)r100_mm_wreg(rdev, (0xe60), (0), 0); |
3540 | RREG32(SRBM_SOFT_RESET)r100_mm_rreg(rdev, (0xe60), 0); |
3541 | } |
3542 | |
3543 | WREG32(RLC_CNTL, 0)r100_mm_wreg(rdev, (0x3f00), (0), 0); |
3544 | } |
3545 | |
3546 | static void r600_rlc_start(struct radeon_device *rdev) |
3547 | { |
3548 | WREG32(RLC_CNTL, RLC_ENABLE)r100_mm_wreg(rdev, (0x3f00), ((1 << 0)), 0); |
3549 | } |
3550 | |
3551 | static int r600_rlc_resume(struct radeon_device *rdev) |
3552 | { |
3553 | u32 i; |
3554 | const __be32 *fw_data; |
3555 | |
3556 | if (!rdev->rlc_fw) |
3557 | return -EINVAL22; |
3558 | |
3559 | r600_rlc_stop(rdev); |
3560 | |
3561 | WREG32(RLC_HB_CNTL, 0)r100_mm_wreg(rdev, (0x3f0c), (0), 0); |
3562 | |
3563 | WREG32(RLC_HB_BASE, 0)r100_mm_wreg(rdev, (0x3f10), (0), 0); |
3564 | WREG32(RLC_HB_RPTR, 0)r100_mm_wreg(rdev, (0x3f20), (0), 0); |
3565 | WREG32(RLC_HB_WPTR, 0)r100_mm_wreg(rdev, (0x3f1c), (0), 0); |
3566 | WREG32(RLC_HB_WPTR_LSB_ADDR, 0)r100_mm_wreg(rdev, (0x3f14), (0), 0); |
3567 | WREG32(RLC_HB_WPTR_MSB_ADDR, 0)r100_mm_wreg(rdev, (0x3f18), (0), 0); |
3568 | WREG32(RLC_MC_CNTL, 0)r100_mm_wreg(rdev, (0x3f44), (0), 0); |
3569 | WREG32(RLC_UCODE_CNTL, 0)r100_mm_wreg(rdev, (0x3f48), (0), 0); |
3570 | |
3571 | fw_data = (const __be32 *)rdev->rlc_fw->data; |
3572 | if (rdev->family >= CHIP_RV770) { |
3573 | for (i = 0; i < R700_RLC_UCODE_SIZE1024; i++) { |
3574 | WREG32(RLC_UCODE_ADDR, i)r100_mm_wreg(rdev, (0x3f2c), (i), 0); |
3575 | WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++))r100_mm_wreg(rdev, (0x3f30), ((__uint32_t)(__builtin_constant_p (*(__uint32_t *)(fw_data++)) ? (__uint32_t)(((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff) << 24 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff00) << 8 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff0000) >> 8 | ((__uint32_t)(*( __uint32_t *)(fw_data++)) & 0xff000000) >> 24) : __swap32md (*(__uint32_t *)(fw_data++)))), 0); |
3576 | } |
3577 | } else { |
3578 | for (i = 0; i < R600_RLC_UCODE_SIZE768; i++) { |
3579 | WREG32(RLC_UCODE_ADDR, i)r100_mm_wreg(rdev, (0x3f2c), (i), 0); |
3580 | WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++))r100_mm_wreg(rdev, (0x3f30), ((__uint32_t)(__builtin_constant_p (*(__uint32_t *)(fw_data++)) ? (__uint32_t)(((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff) << 24 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff00) << 8 | ((__uint32_t)(*(__uint32_t *)(fw_data++)) & 0xff0000) >> 8 | ((__uint32_t)(*( __uint32_t *)(fw_data++)) & 0xff000000) >> 24) : __swap32md (*(__uint32_t *)(fw_data++)))), 0); |
3581 | } |
3582 | } |
3583 | WREG32(RLC_UCODE_ADDR, 0)r100_mm_wreg(rdev, (0x3f2c), (0), 0); |
3584 | |
3585 | r600_rlc_start(rdev); |
3586 | |
3587 | return 0; |
3588 | } |
3589 | |
3590 | static void r600_enable_interrupts(struct radeon_device *rdev) |
3591 | { |
3592 | u32 ih_cntl = RREG32(IH_CNTL)r100_mm_rreg(rdev, (0x3e18), 0); |
3593 | u32 ih_rb_cntl = RREG32(IH_RB_CNTL)r100_mm_rreg(rdev, (0x3e00), 0); |
3594 | |
3595 | ih_cntl |= ENABLE_INTR(1 << 0); |
3596 | ih_rb_cntl |= IH_RB_ENABLE(1 << 0); |
3597 | WREG32(IH_CNTL, ih_cntl)r100_mm_wreg(rdev, (0x3e18), (ih_cntl), 0); |
3598 | WREG32(IH_RB_CNTL, ih_rb_cntl)r100_mm_wreg(rdev, (0x3e00), (ih_rb_cntl), 0); |
3599 | rdev->ih.enabled = true1; |
3600 | } |
3601 | |
3602 | void r600_disable_interrupts(struct radeon_device *rdev) |
3603 | { |
3604 | u32 ih_rb_cntl = RREG32(IH_RB_CNTL)r100_mm_rreg(rdev, (0x3e00), 0); |
3605 | u32 ih_cntl = RREG32(IH_CNTL)r100_mm_rreg(rdev, (0x3e18), 0); |
3606 | |
3607 | ih_rb_cntl &= ~IH_RB_ENABLE(1 << 0); |
3608 | ih_cntl &= ~ENABLE_INTR(1 << 0); |
3609 | WREG32(IH_RB_CNTL, ih_rb_cntl)r100_mm_wreg(rdev, (0x3e00), (ih_rb_cntl), 0); |
3610 | WREG32(IH_CNTL, ih_cntl)r100_mm_wreg(rdev, (0x3e18), (ih_cntl), 0); |
3611 | /* set rptr, wptr to 0 */ |
3612 | WREG32(IH_RB_RPTR, 0)r100_mm_wreg(rdev, (0x3e08), (0), 0); |
3613 | WREG32(IH_RB_WPTR, 0)r100_mm_wreg(rdev, (0x3e0c), (0), 0); |
3614 | rdev->ih.enabled = false0; |
3615 | rdev->ih.rptr = 0; |
3616 | } |
3617 | |
3618 | static void r600_disable_interrupt_state(struct radeon_device *rdev) |
3619 | { |
3620 | u32 tmp; |
3621 | |
3622 | WREG32(CP_INT_CNTL, CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE)r100_mm_wreg(rdev, (0xc124), ((1 << 19) | (1 << 20 )), 0); |
3623 | tmp = RREG32(DMA_CNTL)r100_mm_rreg(rdev, (0xd02c), 0) & ~TRAP_ENABLE(1 << 0); |
3624 | WREG32(DMA_CNTL, tmp)r100_mm_wreg(rdev, (0xd02c), (tmp), 0); |
3625 | WREG32(GRBM_INT_CNTL, 0)r100_mm_wreg(rdev, (0x8060), (0), 0); |
3626 | WREG32(DxMODE_INT_MASK, 0)r100_mm_wreg(rdev, (0x6540), (0), 0); |
3627 | WREG32(D1GRPH_INTERRUPT_CONTROL, 0)r100_mm_wreg(rdev, (0x615c), (0), 0); |
3628 | WREG32(D2GRPH_INTERRUPT_CONTROL, 0)r100_mm_wreg(rdev, (0x695c), (0), 0); |
3629 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3630 | WREG32(DCE3_DACA_AUTODETECT_INT_CONTROL, 0)r100_mm_wreg(rdev, (0x7038), (0), 0); |
3631 | WREG32(DCE3_DACB_AUTODETECT_INT_CONTROL, 0)r100_mm_wreg(rdev, (0x7138), (0), 0); |
3632 | tmp = RREG32(DC_HPD1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d04), 0) & DC_HPDx_INT_POLARITY(1 << 8); |
3633 | WREG32(DC_HPD1_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d04), (tmp), 0); |
3634 | tmp = RREG32(DC_HPD2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d10), 0) & DC_HPDx_INT_POLARITY(1 << 8); |
3635 | WREG32(DC_HPD2_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d10), (tmp), 0); |
3636 | tmp = RREG32(DC_HPD3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d1c), 0) & DC_HPDx_INT_POLARITY(1 << 8); |
3637 | WREG32(DC_HPD3_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d1c), (tmp), 0); |
3638 | tmp = RREG32(DC_HPD4_INT_CONTROL)r100_mm_rreg(rdev, (0x7d28), 0) & DC_HPDx_INT_POLARITY(1 << 8); |
3639 | WREG32(DC_HPD4_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d28), (tmp), 0); |
3640 | if (ASIC_IS_DCE32(rdev)((rdev->family >= CHIP_RV730))) { |
3641 | tmp = RREG32(DC_HPD5_INT_CONTROL)r100_mm_rreg(rdev, (0x7dc4), 0) & DC_HPDx_INT_POLARITY(1 << 8); |
3642 | WREG32(DC_HPD5_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7dc4), (tmp), 0); |
3643 | tmp = RREG32(DC_HPD6_INT_CONTROL)r100_mm_rreg(rdev, (0x7df8), 0) & DC_HPDx_INT_POLARITY(1 << 8); |
3644 | WREG32(DC_HPD6_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7df8), (tmp), 0); |
3645 | tmp = RREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET0)r100_mm_rreg(rdev, (0x7604 + (0x7400 - 0x7400)), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3646 | WREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET0, tmp)r100_mm_wreg(rdev, (0x7604 + (0x7400 - 0x7400)), (tmp), 0); |
3647 | tmp = RREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET1)r100_mm_rreg(rdev, (0x7604 + (0x7800 - 0x7400)), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3648 | WREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET1, tmp)r100_mm_wreg(rdev, (0x7604 + (0x7800 - 0x7400)), (tmp), 0); |
3649 | } else { |
3650 | tmp = RREG32(HDMI0_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7408), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3651 | WREG32(HDMI0_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7408), (tmp), 0); |
3652 | tmp = RREG32(DCE3_HDMI1_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7808), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3653 | WREG32(DCE3_HDMI1_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7808), (tmp), 0); |
3654 | } |
3655 | } else { |
3656 | WREG32(DACA_AUTODETECT_INT_CONTROL, 0)r100_mm_wreg(rdev, (0x7838), (0), 0); |
3657 | WREG32(DACB_AUTODETECT_INT_CONTROL, 0)r100_mm_wreg(rdev, (0x7a38), (0), 0); |
3658 | tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d08), 0) & DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
3659 | WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d08), (tmp), 0); |
3660 | tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d18), 0) & DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
3661 | WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d18), (tmp), 0); |
3662 | tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d2c), 0) & DC_HOT_PLUG_DETECTx_INT_POLARITY(1 << 8); |
3663 | WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d2c), (tmp), 0); |
3664 | tmp = RREG32(HDMI0_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7408), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3665 | WREG32(HDMI0_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7408), (tmp), 0); |
3666 | tmp = RREG32(HDMI1_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7708), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3667 | WREG32(HDMI1_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7708), (tmp), 0); |
3668 | } |
3669 | } |
3670 | |
3671 | int r600_irq_init(struct radeon_device *rdev) |
3672 | { |
3673 | int ret = 0; |
3674 | int rb_bufsz; |
3675 | u32 interrupt_cntl, ih_cntl, ih_rb_cntl; |
3676 | |
3677 | /* allocate ring */ |
3678 | ret = r600_ih_ring_alloc(rdev); |
3679 | if (ret) |
3680 | return ret; |
3681 | |
3682 | /* disable irqs */ |
3683 | r600_disable_interrupts(rdev); |
3684 | |
3685 | /* init rlc */ |
3686 | if (rdev->family >= CHIP_CEDAR) |
3687 | ret = evergreen_rlc_resume(rdev); |
3688 | else |
3689 | ret = r600_rlc_resume(rdev); |
3690 | if (ret) { |
3691 | r600_ih_ring_fini(rdev); |
3692 | return ret; |
3693 | } |
3694 | |
3695 | /* setup interrupt control */ |
3696 | /* set dummy read address to dummy page address */ |
3697 | WREG32(INTERRUPT_CNTL2, rdev->dummy_page.addr >> 8)r100_mm_wreg(rdev, (0x546c), (rdev->dummy_page.addr >> 8), 0); |
3698 | interrupt_cntl = RREG32(INTERRUPT_CNTL)r100_mm_rreg(rdev, (0x5468), 0); |
3699 | /* IH_DUMMY_RD_OVERRIDE=0 - dummy read disabled with msi, enabled without msi |
3700 | * IH_DUMMY_RD_OVERRIDE=1 - dummy read controlled by IH_DUMMY_RD_EN |
3701 | */ |
3702 | interrupt_cntl &= ~IH_DUMMY_RD_OVERRIDE(1 << 0); |
3703 | /* IH_REQ_NONSNOOP_EN=1 if ring is in non-cacheable memory, e.g., vram */ |
3704 | interrupt_cntl &= ~IH_REQ_NONSNOOP_EN(1 << 3); |
3705 | WREG32(INTERRUPT_CNTL, interrupt_cntl)r100_mm_wreg(rdev, (0x5468), (interrupt_cntl), 0); |
3706 | |
3707 | WREG32(IH_RB_BASE, rdev->ih.gpu_addr >> 8)r100_mm_wreg(rdev, (0x3e04), (rdev->ih.gpu_addr >> 8 ), 0); |
3708 | rb_bufsz = order_base_2(rdev->ih.ring_size / 4)drm_order(rdev->ih.ring_size / 4); |
3709 | |
3710 | ih_rb_cntl = (IH_WPTR_OVERFLOW_ENABLE(1 << 16) | |
3711 | IH_WPTR_OVERFLOW_CLEAR(1 << 31) | |
3712 | (rb_bufsz << 1)); |
3713 | |
3714 | if (rdev->wb.enabled) |
3715 | ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE(1 << 8); |
3716 | |
3717 | /* set the writeback address whether it's enabled or not */ |
3718 | WREG32(IH_RB_WPTR_ADDR_LO, (rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFFFFFFFC)r100_mm_wreg(rdev, (0x3e14), ((rdev->wb.gpu_addr + 2048) & 0xFFFFFFFC), 0); |
3719 | WREG32(IH_RB_WPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFF)r100_mm_wreg(rdev, (0x3e10), (((u32)(((rdev->wb.gpu_addr + 2048) >> 16) >> 16)) & 0xFF), 0); |
3720 | |
3721 | WREG32(IH_RB_CNTL, ih_rb_cntl)r100_mm_wreg(rdev, (0x3e00), (ih_rb_cntl), 0); |
3722 | |
3723 | /* set rptr, wptr to 0 */ |
3724 | WREG32(IH_RB_RPTR, 0)r100_mm_wreg(rdev, (0x3e08), (0), 0); |
3725 | WREG32(IH_RB_WPTR, 0)r100_mm_wreg(rdev, (0x3e0c), (0), 0); |
3726 | |
3727 | /* Default settings for IH_CNTL (disabled at first) */ |
3728 | ih_cntl = MC_WRREQ_CREDIT(0x10)((0x10) << 15) | MC_WR_CLEAN_CNT(0x10)((0x10) << 20); |
3729 | /* RPTR_REARM only works if msi's are enabled */ |
3730 | if (rdev->msi_enabled) |
3731 | ih_cntl |= RPTR_REARM(1 << 4); |
3732 | WREG32(IH_CNTL, ih_cntl)r100_mm_wreg(rdev, (0x3e18), (ih_cntl), 0); |
3733 | |
3734 | /* force the active interrupt state to all disabled */ |
3735 | if (rdev->family >= CHIP_CEDAR) |
3736 | evergreen_disable_interrupt_state(rdev); |
3737 | else |
3738 | r600_disable_interrupt_state(rdev); |
3739 | |
3740 | /* at this point everything should be setup correctly to enable master */ |
3741 | pci_set_master(rdev->pdev); |
3742 | |
3743 | /* enable irqs */ |
3744 | r600_enable_interrupts(rdev); |
3745 | |
3746 | return ret; |
3747 | } |
3748 | |
3749 | void r600_irq_suspend(struct radeon_device *rdev) |
3750 | { |
3751 | r600_irq_disable(rdev); |
3752 | r600_rlc_stop(rdev); |
3753 | } |
3754 | |
3755 | void r600_irq_fini(struct radeon_device *rdev) |
3756 | { |
3757 | r600_irq_suspend(rdev); |
3758 | r600_ih_ring_fini(rdev); |
3759 | } |
3760 | |
3761 | int r600_irq_set(struct radeon_device *rdev) |
3762 | { |
3763 | u32 cp_int_cntl = CNTX_BUSY_INT_ENABLE(1 << 19) | CNTX_EMPTY_INT_ENABLE(1 << 20); |
3764 | u32 mode_int = 0; |
3765 | u32 hpd1, hpd2, hpd3, hpd4 = 0, hpd5 = 0, hpd6 = 0; |
3766 | u32 grbm_int_cntl = 0; |
3767 | u32 hdmi0, hdmi1; |
3768 | u32 dma_cntl; |
3769 | u32 thermal_int = 0; |
3770 | |
3771 | if (!rdev->irq.installed) { |
3772 | WARN(1, "Can't enable IRQ/MSI because no handler is installed\n")({ int __ret = !!(1); if (__ret) printf("Can't enable IRQ/MSI because no handler is installed\n" ); __builtin_expect(!!(__ret), 0); }); |
3773 | return -EINVAL22; |
3774 | } |
3775 | /* don't enable anything if the ih is disabled */ |
3776 | if (!rdev->ih.enabled) { |
3777 | r600_disable_interrupts(rdev); |
3778 | /* force the active interrupt state to all disabled */ |
3779 | r600_disable_interrupt_state(rdev); |
3780 | return 0; |
3781 | } |
3782 | |
3783 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3784 | hpd1 = RREG32(DC_HPD1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d04), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3785 | hpd2 = RREG32(DC_HPD2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d10), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3786 | hpd3 = RREG32(DC_HPD3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d1c), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3787 | hpd4 = RREG32(DC_HPD4_INT_CONTROL)r100_mm_rreg(rdev, (0x7d28), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3788 | if (ASIC_IS_DCE32(rdev)((rdev->family >= CHIP_RV730))) { |
3789 | hpd5 = RREG32(DC_HPD5_INT_CONTROL)r100_mm_rreg(rdev, (0x7dc4), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3790 | hpd6 = RREG32(DC_HPD6_INT_CONTROL)r100_mm_rreg(rdev, (0x7df8), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3791 | hdmi0 = RREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET0)r100_mm_rreg(rdev, (0x7604 + (0x7400 - 0x7400)), 0) & ~AFMT_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3792 | hdmi1 = RREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET1)r100_mm_rreg(rdev, (0x7604 + (0x7800 - 0x7400)), 0) & ~AFMT_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3793 | } else { |
3794 | hdmi0 = RREG32(HDMI0_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7408), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3795 | hdmi1 = RREG32(DCE3_HDMI1_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7808), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3796 | } |
3797 | } else { |
3798 | hpd1 = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d08), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3799 | hpd2 = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d18), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3800 | hpd3 = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d2c), 0) & ~DC_HPDx_INT_EN(1 << 16); |
3801 | hdmi0 = RREG32(HDMI0_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7408), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3802 | hdmi1 = RREG32(HDMI1_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7708), 0) & ~HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3803 | } |
3804 | |
3805 | dma_cntl = RREG32(DMA_CNTL)r100_mm_rreg(rdev, (0xd02c), 0) & ~TRAP_ENABLE(1 << 0); |
3806 | |
3807 | if ((rdev->family > CHIP_R600) && (rdev->family < CHIP_RV770)) { |
3808 | thermal_int = RREG32(CG_THERMAL_INT)r100_mm_rreg(rdev, (0x7F8), 0) & |
3809 | ~(THERM_INT_MASK_HIGH(1 << 24) | THERM_INT_MASK_LOW(1 << 25)); |
3810 | } else if (rdev->family >= CHIP_RV770) { |
3811 | thermal_int = RREG32(RV770_CG_THERMAL_INT)r100_mm_rreg(rdev, (0x734), 0) & |
3812 | ~(THERM_INT_MASK_HIGH(1 << 24) | THERM_INT_MASK_LOW(1 << 25)); |
3813 | } |
3814 | if (rdev->irq.dpm_thermal) { |
3815 | DRM_DEBUG("dpm thermal\n")___drm_dbg(((void *)0), DRM_UT_CORE, "dpm thermal\n"); |
3816 | thermal_int |= THERM_INT_MASK_HIGH(1 << 24) | THERM_INT_MASK_LOW(1 << 25); |
3817 | } |
3818 | |
3819 | if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])({ typeof(*(&rdev->irq.ring_int[0])) __tmp = *(volatile typeof(*(&rdev->irq.ring_int[0])) *)&(*(&rdev ->irq.ring_int[0])); membar_datadep_consumer(); __tmp; })) { |
3820 | DRM_DEBUG("r600_irq_set: sw int\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: sw int\n" ); |
3821 | cp_int_cntl |= RB_INT_ENABLE(1 << 31); |
3822 | cp_int_cntl |= TIME_STAMP_INT_ENABLE(1 << 26); |
3823 | } |
3824 | |
3825 | if (atomic_read(&rdev->irq.ring_int[R600_RING_TYPE_DMA_INDEX])({ typeof(*(&rdev->irq.ring_int[3])) __tmp = *(volatile typeof(*(&rdev->irq.ring_int[3])) *)&(*(&rdev ->irq.ring_int[3])); membar_datadep_consumer(); __tmp; })) { |
3826 | DRM_DEBUG("r600_irq_set: sw int dma\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: sw int dma\n" ); |
3827 | dma_cntl |= TRAP_ENABLE(1 << 0); |
3828 | } |
3829 | |
3830 | if (rdev->irq.crtc_vblank_int[0] || |
3831 | atomic_read(&rdev->irq.pflip[0])({ typeof(*(&rdev->irq.pflip[0])) __tmp = *(volatile typeof (*(&rdev->irq.pflip[0])) *)&(*(&rdev->irq.pflip [0])); membar_datadep_consumer(); __tmp; })) { |
3832 | DRM_DEBUG("r600_irq_set: vblank 0\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: vblank 0\n" ); |
3833 | mode_int |= D1MODE_VBLANK_INT_MASK(1 << 0); |
3834 | } |
3835 | if (rdev->irq.crtc_vblank_int[1] || |
3836 | atomic_read(&rdev->irq.pflip[1])({ typeof(*(&rdev->irq.pflip[1])) __tmp = *(volatile typeof (*(&rdev->irq.pflip[1])) *)&(*(&rdev->irq.pflip [1])); membar_datadep_consumer(); __tmp; })) { |
3837 | DRM_DEBUG("r600_irq_set: vblank 1\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: vblank 1\n" ); |
3838 | mode_int |= D2MODE_VBLANK_INT_MASK(1 << 8); |
3839 | } |
3840 | if (rdev->irq.hpd[0]) { |
3841 | DRM_DEBUG("r600_irq_set: hpd 1\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hpd 1\n"); |
3842 | hpd1 |= DC_HPDx_INT_EN(1 << 16); |
3843 | } |
3844 | if (rdev->irq.hpd[1]) { |
3845 | DRM_DEBUG("r600_irq_set: hpd 2\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hpd 2\n"); |
3846 | hpd2 |= DC_HPDx_INT_EN(1 << 16); |
3847 | } |
3848 | if (rdev->irq.hpd[2]) { |
3849 | DRM_DEBUG("r600_irq_set: hpd 3\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hpd 3\n"); |
3850 | hpd3 |= DC_HPDx_INT_EN(1 << 16); |
3851 | } |
3852 | if (rdev->irq.hpd[3]) { |
3853 | DRM_DEBUG("r600_irq_set: hpd 4\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hpd 4\n"); |
3854 | hpd4 |= DC_HPDx_INT_EN(1 << 16); |
3855 | } |
3856 | if (rdev->irq.hpd[4]) { |
3857 | DRM_DEBUG("r600_irq_set: hpd 5\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hpd 5\n"); |
3858 | hpd5 |= DC_HPDx_INT_EN(1 << 16); |
3859 | } |
3860 | if (rdev->irq.hpd[5]) { |
3861 | DRM_DEBUG("r600_irq_set: hpd 6\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hpd 6\n"); |
3862 | hpd6 |= DC_HPDx_INT_EN(1 << 16); |
3863 | } |
3864 | if (rdev->irq.afmt[0]) { |
3865 | DRM_DEBUG("r600_irq_set: hdmi 0\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hdmi 0\n" ); |
3866 | hdmi0 |= HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3867 | } |
3868 | if (rdev->irq.afmt[1]) { |
3869 | DRM_DEBUG("r600_irq_set: hdmi 0\n")___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_set: hdmi 0\n" ); |
3870 | hdmi1 |= HDMI0_AZ_FORMAT_WTRIG_MASK(1 << 28); |
3871 | } |
3872 | |
3873 | WREG32(CP_INT_CNTL, cp_int_cntl)r100_mm_wreg(rdev, (0xc124), (cp_int_cntl), 0); |
3874 | WREG32(DMA_CNTL, dma_cntl)r100_mm_wreg(rdev, (0xd02c), (dma_cntl), 0); |
3875 | WREG32(DxMODE_INT_MASK, mode_int)r100_mm_wreg(rdev, (0x6540), (mode_int), 0); |
3876 | WREG32(D1GRPH_INTERRUPT_CONTROL, DxGRPH_PFLIP_INT_MASK)r100_mm_wreg(rdev, (0x615c), ((1 << 0)), 0); |
3877 | WREG32(D2GRPH_INTERRUPT_CONTROL, DxGRPH_PFLIP_INT_MASK)r100_mm_wreg(rdev, (0x695c), ((1 << 0)), 0); |
3878 | WREG32(GRBM_INT_CNTL, grbm_int_cntl)r100_mm_wreg(rdev, (0x8060), (grbm_int_cntl), 0); |
3879 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3880 | WREG32(DC_HPD1_INT_CONTROL, hpd1)r100_mm_wreg(rdev, (0x7d04), (hpd1), 0); |
3881 | WREG32(DC_HPD2_INT_CONTROL, hpd2)r100_mm_wreg(rdev, (0x7d10), (hpd2), 0); |
3882 | WREG32(DC_HPD3_INT_CONTROL, hpd3)r100_mm_wreg(rdev, (0x7d1c), (hpd3), 0); |
3883 | WREG32(DC_HPD4_INT_CONTROL, hpd4)r100_mm_wreg(rdev, (0x7d28), (hpd4), 0); |
3884 | if (ASIC_IS_DCE32(rdev)((rdev->family >= CHIP_RV730))) { |
3885 | WREG32(DC_HPD5_INT_CONTROL, hpd5)r100_mm_wreg(rdev, (0x7dc4), (hpd5), 0); |
3886 | WREG32(DC_HPD6_INT_CONTROL, hpd6)r100_mm_wreg(rdev, (0x7df8), (hpd6), 0); |
3887 | WREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET0, hdmi0)r100_mm_wreg(rdev, (0x7604 + (0x7400 - 0x7400)), (hdmi0), 0); |
3888 | WREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET1, hdmi1)r100_mm_wreg(rdev, (0x7604 + (0x7800 - 0x7400)), (hdmi1), 0); |
3889 | } else { |
3890 | WREG32(HDMI0_AUDIO_PACKET_CONTROL, hdmi0)r100_mm_wreg(rdev, (0x7408), (hdmi0), 0); |
3891 | WREG32(DCE3_HDMI1_AUDIO_PACKET_CONTROL, hdmi1)r100_mm_wreg(rdev, (0x7808), (hdmi1), 0); |
3892 | } |
3893 | } else { |
3894 | WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, hpd1)r100_mm_wreg(rdev, (0x7d08), (hpd1), 0); |
3895 | WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, hpd2)r100_mm_wreg(rdev, (0x7d18), (hpd2), 0); |
3896 | WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, hpd3)r100_mm_wreg(rdev, (0x7d2c), (hpd3), 0); |
3897 | WREG32(HDMI0_AUDIO_PACKET_CONTROL, hdmi0)r100_mm_wreg(rdev, (0x7408), (hdmi0), 0); |
3898 | WREG32(HDMI1_AUDIO_PACKET_CONTROL, hdmi1)r100_mm_wreg(rdev, (0x7708), (hdmi1), 0); |
3899 | } |
3900 | if ((rdev->family > CHIP_R600) && (rdev->family < CHIP_RV770)) { |
3901 | WREG32(CG_THERMAL_INT, thermal_int)r100_mm_wreg(rdev, (0x7F8), (thermal_int), 0); |
3902 | } else if (rdev->family >= CHIP_RV770) { |
3903 | WREG32(RV770_CG_THERMAL_INT, thermal_int)r100_mm_wreg(rdev, (0x734), (thermal_int), 0); |
3904 | } |
3905 | |
3906 | /* posting read */ |
3907 | RREG32(R_000E50_SRBM_STATUS)r100_mm_rreg(rdev, (0x0E50), 0); |
3908 | |
3909 | return 0; |
3910 | } |
3911 | |
3912 | static void r600_irq_ack(struct radeon_device *rdev) |
3913 | { |
3914 | u32 tmp; |
3915 | |
3916 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3917 | rdev->irq.stat_regs.r600.disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS)r100_mm_rreg(rdev, (0x7ddc), 0); |
3918 | rdev->irq.stat_regs.r600.disp_int_cont = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE)r100_mm_rreg(rdev, (0x7de8), 0); |
3919 | rdev->irq.stat_regs.r600.disp_int_cont2 = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE2)r100_mm_rreg(rdev, (0x7dec), 0); |
3920 | if (ASIC_IS_DCE32(rdev)((rdev->family >= CHIP_RV730))) { |
3921 | rdev->irq.stat_regs.r600.hdmi0_status = RREG32(AFMT_STATUS + DCE3_HDMI_OFFSET0)r100_mm_rreg(rdev, (0x7600 + (0x7400 - 0x7400)), 0); |
3922 | rdev->irq.stat_regs.r600.hdmi1_status = RREG32(AFMT_STATUS + DCE3_HDMI_OFFSET1)r100_mm_rreg(rdev, (0x7600 + (0x7800 - 0x7400)), 0); |
3923 | } else { |
3924 | rdev->irq.stat_regs.r600.hdmi0_status = RREG32(HDMI0_STATUS)r100_mm_rreg(rdev, (0x7404), 0); |
3925 | rdev->irq.stat_regs.r600.hdmi1_status = RREG32(DCE3_HDMI1_STATUS)r100_mm_rreg(rdev, (0x7804), 0); |
3926 | } |
3927 | } else { |
3928 | rdev->irq.stat_regs.r600.disp_int = RREG32(DISP_INTERRUPT_STATUS)r100_mm_rreg(rdev, (0x7edc), 0); |
3929 | rdev->irq.stat_regs.r600.disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE)r100_mm_rreg(rdev, (0x7ee8), 0); |
3930 | rdev->irq.stat_regs.r600.disp_int_cont2 = 0; |
3931 | rdev->irq.stat_regs.r600.hdmi0_status = RREG32(HDMI0_STATUS)r100_mm_rreg(rdev, (0x7404), 0); |
3932 | rdev->irq.stat_regs.r600.hdmi1_status = RREG32(HDMI1_STATUS)r100_mm_rreg(rdev, (0x7704), 0); |
3933 | } |
3934 | rdev->irq.stat_regs.r600.d1grph_int = RREG32(D1GRPH_INTERRUPT_STATUS)r100_mm_rreg(rdev, (0x6158), 0); |
3935 | rdev->irq.stat_regs.r600.d2grph_int = RREG32(D2GRPH_INTERRUPT_STATUS)r100_mm_rreg(rdev, (0x6958), 0); |
3936 | |
3937 | if (rdev->irq.stat_regs.r600.d1grph_int & DxGRPH_PFLIP_INT_OCCURRED(1 << 0)) |
3938 | WREG32(D1GRPH_INTERRUPT_STATUS, DxGRPH_PFLIP_INT_CLEAR)r100_mm_wreg(rdev, (0x6158), ((1 << 8)), 0); |
3939 | if (rdev->irq.stat_regs.r600.d2grph_int & DxGRPH_PFLIP_INT_OCCURRED(1 << 0)) |
3940 | WREG32(D2GRPH_INTERRUPT_STATUS, DxGRPH_PFLIP_INT_CLEAR)r100_mm_wreg(rdev, (0x6958), ((1 << 8)), 0); |
3941 | if (rdev->irq.stat_regs.r600.disp_int & LB_D1_VBLANK_INTERRUPT(1 << 4)) |
3942 | WREG32(D1MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK)r100_mm_wreg(rdev, (0x6534), ((1 << 4)), 0); |
3943 | if (rdev->irq.stat_regs.r600.disp_int & LB_D1_VLINE_INTERRUPT(1 << 2)) |
3944 | WREG32(D1MODE_VLINE_STATUS, DxMODE_VLINE_ACK)r100_mm_wreg(rdev, (0x653c), ((1 << 4)), 0); |
3945 | if (rdev->irq.stat_regs.r600.disp_int & LB_D2_VBLANK_INTERRUPT(1 << 5)) |
3946 | WREG32(D2MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK)r100_mm_wreg(rdev, (0x6d34), ((1 << 4)), 0); |
3947 | if (rdev->irq.stat_regs.r600.disp_int & LB_D2_VLINE_INTERRUPT(1 << 3)) |
3948 | WREG32(D2MODE_VLINE_STATUS, DxMODE_VLINE_ACK)r100_mm_wreg(rdev, (0x6d3c), ((1 << 4)), 0); |
3949 | if (rdev->irq.stat_regs.r600.disp_int & DC_HPD1_INTERRUPT(1 << 18)) { |
3950 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3951 | tmp = RREG32(DC_HPD1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d04), 0); |
3952 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3953 | WREG32(DC_HPD1_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d04), (tmp), 0); |
3954 | } else { |
3955 | tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL)r100_mm_rreg(rdev, (0x7d08), 0); |
3956 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3957 | WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d08), (tmp), 0); |
3958 | } |
3959 | } |
3960 | if (rdev->irq.stat_regs.r600.disp_int & DC_HPD2_INTERRUPT(1 << 19)) { |
3961 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3962 | tmp = RREG32(DC_HPD2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d10), 0); |
3963 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3964 | WREG32(DC_HPD2_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d10), (tmp), 0); |
3965 | } else { |
3966 | tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL)r100_mm_rreg(rdev, (0x7d18), 0); |
3967 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3968 | WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d18), (tmp), 0); |
3969 | } |
3970 | } |
3971 | if (rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD3_INTERRUPT(1 << 28)) { |
3972 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
3973 | tmp = RREG32(DC_HPD3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d1c), 0); |
3974 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3975 | WREG32(DC_HPD3_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d1c), (tmp), 0); |
3976 | } else { |
3977 | tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL)r100_mm_rreg(rdev, (0x7d2c), 0); |
3978 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3979 | WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d2c), (tmp), 0); |
3980 | } |
3981 | } |
3982 | if (rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD4_INTERRUPT(1 << 14)) { |
3983 | tmp = RREG32(DC_HPD4_INT_CONTROL)r100_mm_rreg(rdev, (0x7d28), 0); |
3984 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3985 | WREG32(DC_HPD4_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7d28), (tmp), 0); |
3986 | } |
3987 | if (ASIC_IS_DCE32(rdev)((rdev->family >= CHIP_RV730))) { |
3988 | if (rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD5_INTERRUPT(1 << 19)) { |
3989 | tmp = RREG32(DC_HPD5_INT_CONTROL)r100_mm_rreg(rdev, (0x7dc4), 0); |
3990 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3991 | WREG32(DC_HPD5_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7dc4), (tmp), 0); |
3992 | } |
3993 | if (rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD6_INTERRUPT(1 << 21)) { |
3994 | tmp = RREG32(DC_HPD6_INT_CONTROL)r100_mm_rreg(rdev, (0x7df8), 0); |
3995 | tmp |= DC_HPDx_INT_ACK(1 << 0); |
3996 | WREG32(DC_HPD6_INT_CONTROL, tmp)r100_mm_wreg(rdev, (0x7df8), (tmp), 0); |
3997 | } |
3998 | if (rdev->irq.stat_regs.r600.hdmi0_status & AFMT_AZ_FORMAT_WTRIG(1 << 28)) { |
3999 | tmp = RREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET0)r100_mm_rreg(rdev, (0x7604 + (0x7400 - 0x7400)), 0); |
4000 | tmp |= AFMT_AZ_FORMAT_WTRIG_ACK(1 << 29); |
4001 | WREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET0, tmp)r100_mm_wreg(rdev, (0x7604 + (0x7400 - 0x7400)), (tmp), 0); |
4002 | } |
4003 | if (rdev->irq.stat_regs.r600.hdmi1_status & AFMT_AZ_FORMAT_WTRIG(1 << 28)) { |
4004 | tmp = RREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET1)r100_mm_rreg(rdev, (0x7604 + (0x7800 - 0x7400)), 0); |
4005 | tmp |= AFMT_AZ_FORMAT_WTRIG_ACK(1 << 29); |
4006 | WREG32(AFMT_AUDIO_PACKET_CONTROL + DCE3_HDMI_OFFSET1, tmp)r100_mm_wreg(rdev, (0x7604 + (0x7800 - 0x7400)), (tmp), 0); |
4007 | } |
4008 | } else { |
4009 | if (rdev->irq.stat_regs.r600.hdmi0_status & HDMI0_AZ_FORMAT_WTRIG(1 << 28)) { |
4010 | tmp = RREG32(HDMI0_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7408), 0); |
4011 | tmp |= HDMI0_AZ_FORMAT_WTRIG_ACK(1 << 29); |
4012 | WREG32(HDMI0_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7408), (tmp), 0); |
4013 | } |
4014 | if (rdev->irq.stat_regs.r600.hdmi1_status & HDMI0_AZ_FORMAT_WTRIG(1 << 28)) { |
4015 | if (ASIC_IS_DCE3(rdev)((rdev->family >= CHIP_RV620))) { |
4016 | tmp = RREG32(DCE3_HDMI1_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7808), 0); |
4017 | tmp |= HDMI0_AZ_FORMAT_WTRIG_ACK(1 << 29); |
4018 | WREG32(DCE3_HDMI1_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7808), (tmp), 0); |
4019 | } else { |
4020 | tmp = RREG32(HDMI1_AUDIO_PACKET_CONTROL)r100_mm_rreg(rdev, (0x7708), 0); |
4021 | tmp |= HDMI0_AZ_FORMAT_WTRIG_ACK(1 << 29); |
4022 | WREG32(HDMI1_AUDIO_PACKET_CONTROL, tmp)r100_mm_wreg(rdev, (0x7708), (tmp), 0); |
4023 | } |
4024 | } |
4025 | } |
4026 | } |
4027 | |
4028 | void r600_irq_disable(struct radeon_device *rdev) |
4029 | { |
4030 | r600_disable_interrupts(rdev); |
4031 | /* Wait and acknowledge irq */ |
4032 | mdelay(1); |
4033 | r600_irq_ack(rdev); |
4034 | r600_disable_interrupt_state(rdev); |
4035 | } |
4036 | |
4037 | static u32 r600_get_ih_wptr(struct radeon_device *rdev) |
4038 | { |
4039 | u32 wptr, tmp; |
4040 | |
4041 | if (rdev->wb.enabled) |
4042 | wptr = le32_to_cpu(rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4])((__uint32_t)(rdev->wb.wb[2048/4])); |
4043 | else |
4044 | wptr = RREG32(IH_RB_WPTR)r100_mm_rreg(rdev, (0x3e0c), 0); |
4045 | |
4046 | if (wptr & RB_OVERFLOW(1 << 0)) { |
4047 | wptr &= ~RB_OVERFLOW(1 << 0); |
4048 | /* When a ring buffer overflow happen start parsing interrupt |
4049 | * from the last not overwritten vector (wptr + 16). Hopefully |
4050 | * this should allow us to catchup. |
4051 | */ |
4052 | dev_warn(rdev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",printf("drm:pid%d:%s *WARNING* " "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , wptr, rdev ->ih.rptr, (wptr + 16) & rdev->ih.ptr_mask) |
4053 | wptr, rdev->ih.rptr, (wptr + 16) & rdev->ih.ptr_mask)printf("drm:pid%d:%s *WARNING* " "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n" , ({struct cpu_info *__ci; asm volatile("movq %%gs:%P1,%0" : "=r" (__ci) :"n" (__builtin_offsetof(struct cpu_info, ci_self))); __ci;})->ci_curproc->p_p->ps_pid, __func__ , wptr, rdev ->ih.rptr, (wptr + 16) & rdev->ih.ptr_mask); |
4054 | rdev->ih.rptr = (wptr + 16) & rdev->ih.ptr_mask; |
4055 | tmp = RREG32(IH_RB_CNTL)r100_mm_rreg(rdev, (0x3e00), 0); |
4056 | tmp |= IH_WPTR_OVERFLOW_CLEAR(1 << 31); |
4057 | WREG32(IH_RB_CNTL, tmp)r100_mm_wreg(rdev, (0x3e00), (tmp), 0); |
4058 | } |
4059 | return (wptr & rdev->ih.ptr_mask); |
4060 | } |
4061 | |
4062 | /* r600 IV Ring |
4063 | * Each IV ring entry is 128 bits: |
4064 | * [7:0] - interrupt source id |
4065 | * [31:8] - reserved |
4066 | * [59:32] - interrupt source data |
4067 | * [127:60] - reserved |
4068 | * |
4069 | * The basic interrupt vector entries |
4070 | * are decoded as follows: |
4071 | * src_id src_data description |
4072 | * 1 0 D1 Vblank |
4073 | * 1 1 D1 Vline |
4074 | * 5 0 D2 Vblank |
4075 | * 5 1 D2 Vline |
4076 | * 19 0 FP Hot plug detection A |
4077 | * 19 1 FP Hot plug detection B |
4078 | * 19 2 DAC A auto-detection |
4079 | * 19 3 DAC B auto-detection |
4080 | * 21 4 HDMI block A |
4081 | * 21 5 HDMI block B |
4082 | * 176 - CP_INT RB |
4083 | * 177 - CP_INT IB1 |
4084 | * 178 - CP_INT IB2 |
4085 | * 181 - EOP Interrupt |
4086 | * 233 - GUI Idle |
4087 | * |
4088 | * Note, these are based on r600 and may need to be |
4089 | * adjusted or added to on newer asics |
4090 | */ |
4091 | |
4092 | int r600_irq_process(struct radeon_device *rdev) |
4093 | { |
4094 | u32 wptr; |
4095 | u32 rptr; |
4096 | u32 src_id, src_data; |
4097 | u32 ring_index; |
4098 | bool_Bool queue_hotplug = false0; |
4099 | bool_Bool queue_hdmi = false0; |
4100 | bool_Bool queue_thermal = false0; |
4101 | |
4102 | if (!rdev->ih.enabled || rdev->shutdown) |
4103 | return IRQ_NONE; |
4104 | |
4105 | /* No MSIs, need a dummy read to flush PCI DMAs */ |
4106 | if (!rdev->msi_enabled) |
4107 | RREG32(IH_RB_WPTR)r100_mm_rreg(rdev, (0x3e0c), 0); |
4108 | |
4109 | wptr = r600_get_ih_wptr(rdev); |
4110 | |
4111 | if (wptr == rdev->ih.rptr) |
4112 | return IRQ_NONE; |
4113 | restart_ih: |
4114 | /* is somebody else already processing irqs? */ |
4115 | if (atomic_xchg(&rdev->ih.lock, 1)) |
4116 | return IRQ_NONE; |
4117 | |
4118 | rptr = rdev->ih.rptr; |
4119 | DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr)___drm_dbg(((void *)0), DRM_UT_CORE, "r600_irq_process start: rptr %d, wptr %d\n" , rptr, wptr); |
4120 | |
4121 | /* Order reading of wptr vs. reading of IH ring data */ |
4122 | rmb()do { __asm volatile("lfence" ::: "memory"); } while (0); |
4123 | |
4124 | /* display interrupts */ |
4125 | r600_irq_ack(rdev); |
4126 | |
4127 | while (rptr != wptr) { |
4128 | /* wptr/rptr are in bytes! */ |
4129 | ring_index = rptr / 4; |
4130 | src_id = le32_to_cpu(rdev->ih.ring[ring_index])((__uint32_t)(rdev->ih.ring[ring_index])) & 0xff; |
4131 | src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1])((__uint32_t)(rdev->ih.ring[ring_index + 1])) & 0xfffffff; |
4132 | |
4133 | switch (src_id) { |
4134 | case 1: /* D1 vblank/vline */ |
4135 | switch (src_data) { |
4136 | case 0: /* D1 vblank */ |
4137 | if (!(rdev->irq.stat_regs.r600.disp_int & LB_D1_VBLANK_INTERRUPT(1 << 4))) |
4138 | DRM_DEBUG("IH: D1 vblank - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D1 vblank - IH event w/o asserted irq bit?\n" ); |
4139 | |
4140 | if (rdev->irq.crtc_vblank_int[0]) { |
4141 | drm_handle_vblank(rdev->ddev, 0); |
4142 | rdev->pm.vblank_sync = true1; |
4143 | wake_up(&rdev->irq.vblank_queue); |
4144 | } |
4145 | if (atomic_read(&rdev->irq.pflip[0])({ typeof(*(&rdev->irq.pflip[0])) __tmp = *(volatile typeof (*(&rdev->irq.pflip[0])) *)&(*(&rdev->irq.pflip [0])); membar_datadep_consumer(); __tmp; })) |
4146 | radeon_crtc_handle_vblank(rdev, 0); |
4147 | rdev->irq.stat_regs.r600.disp_int &= ~LB_D1_VBLANK_INTERRUPT(1 << 4); |
4148 | DRM_DEBUG("IH: D1 vblank\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D1 vblank\n"); |
4149 | |
4150 | break; |
4151 | case 1: /* D1 vline */ |
4152 | if (!(rdev->irq.stat_regs.r600.disp_int & LB_D1_VLINE_INTERRUPT(1 << 2))) |
4153 | DRM_DEBUG("IH: D1 vline - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D1 vline - IH event w/o asserted irq bit?\n" ); |
4154 | |
4155 | rdev->irq.stat_regs.r600.disp_int &= ~LB_D1_VLINE_INTERRUPT(1 << 2); |
4156 | DRM_DEBUG("IH: D1 vline\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D1 vline\n"); |
4157 | |
4158 | break; |
4159 | default: |
4160 | DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data)___drm_dbg(((void *)0), DRM_UT_CORE, "Unhandled interrupt: %d %d\n" , src_id, src_data); |
4161 | break; |
4162 | } |
4163 | break; |
4164 | case 5: /* D2 vblank/vline */ |
4165 | switch (src_data) { |
4166 | case 0: /* D2 vblank */ |
4167 | if (!(rdev->irq.stat_regs.r600.disp_int & LB_D2_VBLANK_INTERRUPT(1 << 5))) |
4168 | DRM_DEBUG("IH: D2 vblank - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D2 vblank - IH event w/o asserted irq bit?\n" ); |
4169 | |
4170 | if (rdev->irq.crtc_vblank_int[1]) { |
4171 | drm_handle_vblank(rdev->ddev, 1); |
4172 | rdev->pm.vblank_sync = true1; |
4173 | wake_up(&rdev->irq.vblank_queue); |
4174 | } |
4175 | if (atomic_read(&rdev->irq.pflip[1])({ typeof(*(&rdev->irq.pflip[1])) __tmp = *(volatile typeof (*(&rdev->irq.pflip[1])) *)&(*(&rdev->irq.pflip [1])); membar_datadep_consumer(); __tmp; })) |
4176 | radeon_crtc_handle_vblank(rdev, 1); |
4177 | rdev->irq.stat_regs.r600.disp_int &= ~LB_D2_VBLANK_INTERRUPT(1 << 5); |
4178 | DRM_DEBUG("IH: D2 vblank\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D2 vblank\n"); |
4179 | |
4180 | break; |
4181 | case 1: /* D1 vline */ |
4182 | if (!(rdev->irq.stat_regs.r600.disp_int & LB_D2_VLINE_INTERRUPT(1 << 3))) |
4183 | DRM_DEBUG("IH: D2 vline - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D2 vline - IH event w/o asserted irq bit?\n" ); |
4184 | |
4185 | rdev->irq.stat_regs.r600.disp_int &= ~LB_D2_VLINE_INTERRUPT(1 << 3); |
4186 | DRM_DEBUG("IH: D2 vline\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D2 vline\n"); |
4187 | |
4188 | break; |
4189 | default: |
4190 | DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data)___drm_dbg(((void *)0), DRM_UT_CORE, "Unhandled interrupt: %d %d\n" , src_id, src_data); |
4191 | break; |
4192 | } |
4193 | break; |
4194 | case 9: /* D1 pflip */ |
4195 | DRM_DEBUG("IH: D1 flip\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D1 flip\n"); |
4196 | if (radeon_use_pflipirq > 0) |
4197 | radeon_crtc_handle_flip(rdev, 0); |
4198 | break; |
4199 | case 11: /* D2 pflip */ |
4200 | DRM_DEBUG("IH: D2 flip\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: D2 flip\n"); |
4201 | if (radeon_use_pflipirq > 0) |
4202 | radeon_crtc_handle_flip(rdev, 1); |
4203 | break; |
4204 | case 19: /* HPD/DAC hotplug */ |
4205 | switch (src_data) { |
4206 | case 0: |
4207 | if (!(rdev->irq.stat_regs.r600.disp_int & DC_HPD1_INTERRUPT(1 << 18))) |
4208 | DRM_DEBUG("IH: HPD1 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD1 - IH event w/o asserted irq bit?\n" ); |
4209 | |
4210 | rdev->irq.stat_regs.r600.disp_int &= ~DC_HPD1_INTERRUPT(1 << 18); |
4211 | queue_hotplug = true1; |
4212 | DRM_DEBUG("IH: HPD1\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD1\n"); |
4213 | break; |
4214 | case 1: |
4215 | if (!(rdev->irq.stat_regs.r600.disp_int & DC_HPD2_INTERRUPT(1 << 19))) |
4216 | DRM_DEBUG("IH: HPD2 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD2 - IH event w/o asserted irq bit?\n" ); |
4217 | |
4218 | rdev->irq.stat_regs.r600.disp_int &= ~DC_HPD2_INTERRUPT(1 << 19); |
4219 | queue_hotplug = true1; |
4220 | DRM_DEBUG("IH: HPD2\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD2\n"); |
4221 | break; |
4222 | case 4: |
4223 | if (!(rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD3_INTERRUPT(1 << 28))) |
4224 | DRM_DEBUG("IH: HPD3 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD3 - IH event w/o asserted irq bit?\n" ); |
4225 | |
4226 | rdev->irq.stat_regs.r600.disp_int_cont &= ~DC_HPD3_INTERRUPT(1 << 28); |
4227 | queue_hotplug = true1; |
4228 | DRM_DEBUG("IH: HPD3\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD3\n"); |
4229 | break; |
4230 | case 5: |
4231 | if (!(rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD4_INTERRUPT(1 << 14))) |
4232 | DRM_DEBUG("IH: HPD4 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD4 - IH event w/o asserted irq bit?\n" ); |
4233 | |
4234 | rdev->irq.stat_regs.r600.disp_int_cont &= ~DC_HPD4_INTERRUPT(1 << 14); |
4235 | queue_hotplug = true1; |
4236 | DRM_DEBUG("IH: HPD4\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD4\n"); |
4237 | break; |
4238 | case 10: |
4239 | if (!(rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD5_INTERRUPT(1 << 19))) |
4240 | DRM_DEBUG("IH: HPD5 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD5 - IH event w/o asserted irq bit?\n" ); |
4241 | |
4242 | rdev->irq.stat_regs.r600.disp_int_cont2 &= ~DC_HPD5_INTERRUPT(1 << 19); |
4243 | queue_hotplug = true1; |
4244 | DRM_DEBUG("IH: HPD5\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD5\n"); |
4245 | break; |
4246 | case 12: |
4247 | if (!(rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD6_INTERRUPT(1 << 21))) |
4248 | DRM_DEBUG("IH: HPD6 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD6 - IH event w/o asserted irq bit?\n" ); |
4249 | |
4250 | rdev->irq.stat_regs.r600.disp_int_cont2 &= ~DC_HPD6_INTERRUPT(1 << 21); |
4251 | queue_hotplug = true1; |
4252 | DRM_DEBUG("IH: HPD6\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HPD6\n"); |
4253 | |
4254 | break; |
4255 | default: |
4256 | DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data)___drm_dbg(((void *)0), DRM_UT_CORE, "Unhandled interrupt: %d %d\n" , src_id, src_data); |
4257 | break; |
4258 | } |
4259 | break; |
4260 | case 21: /* hdmi */ |
4261 | switch (src_data) { |
4262 | case 4: |
4263 | if (!(rdev->irq.stat_regs.r600.hdmi0_status & HDMI0_AZ_FORMAT_WTRIG(1 << 28))) |
4264 | DRM_DEBUG("IH: HDMI0 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HDMI0 - IH event w/o asserted irq bit?\n" ); |
4265 | |
4266 | rdev->irq.stat_regs.r600.hdmi0_status &= ~HDMI0_AZ_FORMAT_WTRIG(1 << 28); |
4267 | queue_hdmi = true1; |
4268 | DRM_DEBUG("IH: HDMI0\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HDMI0\n"); |
4269 | |
4270 | break; |
4271 | case 5: |
4272 | if (!(rdev->irq.stat_regs.r600.hdmi1_status & HDMI0_AZ_FORMAT_WTRIG(1 << 28))) |
4273 | DRM_DEBUG("IH: HDMI1 - IH event w/o asserted irq bit?\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HDMI1 - IH event w/o asserted irq bit?\n" ); |
4274 | |
4275 | rdev->irq.stat_regs.r600.hdmi1_status &= ~HDMI0_AZ_FORMAT_WTRIG(1 << 28); |
4276 | queue_hdmi = true1; |
4277 | DRM_DEBUG("IH: HDMI1\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: HDMI1\n"); |
4278 | |
4279 | break; |
4280 | default: |
4281 | DRM_ERROR("Unhandled interrupt: %d %d\n", src_id, src_data)__drm_err("Unhandled interrupt: %d %d\n", src_id, src_data); |
4282 | break; |
4283 | } |
4284 | break; |
4285 | case 124: /* UVD */ |
4286 | DRM_DEBUG("IH: UVD int: 0x%08x\n", src_data)___drm_dbg(((void *)0), DRM_UT_CORE, "IH: UVD int: 0x%08x\n", src_data); |
4287 | radeon_fence_process(rdev, R600_RING_TYPE_UVD_INDEX5); |
4288 | break; |
4289 | case 176: /* CP_INT in ring buffer */ |
4290 | case 177: /* CP_INT in IB1 */ |
4291 | case 178: /* CP_INT in IB2 */ |
4292 | DRM_DEBUG("IH: CP int: 0x%08x\n", src_data)___drm_dbg(((void *)0), DRM_UT_CORE, "IH: CP int: 0x%08x\n", src_data ); |
4293 | radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX0); |
4294 | break; |
4295 | case 181: /* CP EOP event */ |
4296 | DRM_DEBUG("IH: CP EOP\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: CP EOP\n"); |
4297 | radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX0); |
4298 | break; |
4299 | case 224: /* DMA trap event */ |
4300 | DRM_DEBUG("IH: DMA trap\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: DMA trap\n"); |
4301 | radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX3); |
4302 | break; |
4303 | case 230: /* thermal low to high */ |
4304 | DRM_DEBUG("IH: thermal low to high\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: thermal low to high\n" ); |
4305 | rdev->pm.dpm.thermal.high_to_low = false0; |
4306 | queue_thermal = true1; |
4307 | break; |
4308 | case 231: /* thermal high to low */ |
4309 | DRM_DEBUG("IH: thermal high to low\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: thermal high to low\n" ); |
4310 | rdev->pm.dpm.thermal.high_to_low = true1; |
4311 | queue_thermal = true1; |
4312 | break; |
4313 | case 233: /* GUI IDLE */ |
4314 | DRM_DEBUG("IH: GUI idle\n")___drm_dbg(((void *)0), DRM_UT_CORE, "IH: GUI idle\n"); |
4315 | break; |
4316 | default: |
4317 | DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data)___drm_dbg(((void *)0), DRM_UT_CORE, "Unhandled interrupt: %d %d\n" , src_id, src_data); |
4318 | break; |
4319 | } |
4320 | |
4321 | /* wptr/rptr are in bytes! */ |
4322 | rptr += 16; |
4323 | rptr &= rdev->ih.ptr_mask; |
4324 | WREG32(IH_RB_RPTR, rptr)r100_mm_wreg(rdev, (0x3e08), (rptr), 0); |
4325 | } |
4326 | if (queue_hotplug) |
4327 | schedule_delayed_work(&rdev->hotplug_work, 0); |
4328 | if (queue_hdmi) |
4329 | schedule_work(&rdev->audio_work); |
4330 | if (queue_thermal && rdev->pm.dpm_enabled) |
4331 | schedule_work(&rdev->pm.dpm.thermal.work); |
4332 | rdev->ih.rptr = rptr; |
4333 | atomic_set(&rdev->ih.lock, 0)({ typeof(*(&rdev->ih.lock)) __tmp = ((0)); *(volatile typeof(*(&rdev->ih.lock)) *)&(*(&rdev->ih. lock)) = __tmp; __tmp; }); |
4334 | |
4335 | /* make sure wptr hasn't changed while processing */ |
4336 | wptr = r600_get_ih_wptr(rdev); |
4337 | if (wptr != rptr) |
4338 | goto restart_ih; |
4339 | |
4340 | return IRQ_HANDLED; |
4341 | } |
4342 | |
4343 | /* |
4344 | * Debugfs info |
4345 | */ |
4346 | #if defined(CONFIG_DEBUG_FS) |
4347 | |
4348 | static int r600_debugfs_mc_info_show(struct seq_file *m, void *unused) |
4349 | { |
4350 | struct radeon_device *rdev = (struct radeon_device *)m->private; |
4351 | |
4352 | DREG32_SYS(m, rdev, R_000E50_SRBM_STATUS)seq_printf((m), "R_000E50_SRBM_STATUS" " : 0x%08X\n", r100_mm_rreg ((rdev), (0x0E50), 0)); |
4353 | DREG32_SYS(m, rdev, VM_L2_STATUS)seq_printf((m), "VM_L2_STATUS" " : 0x%08X\n", r100_mm_rreg((rdev ), (0x140C), 0)); |
4354 | return 0; |
4355 | } |
4356 | |
4357 | DEFINE_SHOW_ATTRIBUTE(r600_debugfs_mc_info); |
4358 | #endif |
4359 | |
4360 | static void r600_debugfs_mc_info_init(struct radeon_device *rdev) |
4361 | { |
4362 | #if defined(CONFIG_DEBUG_FS) |
4363 | struct dentry *root = rdev->ddev->primary->debugfs_root; |
4364 | |
4365 | debugfs_create_file("r600_mc_info", 0444, root, rdev,ERR_PTR(-78) |
4366 | &r600_debugfs_mc_info_fops)ERR_PTR(-78); |
4367 | |
4368 | #endif |
4369 | } |
4370 | |
4371 | /** |
4372 | * r600_mmio_hdp_flush - flush Host Data Path cache via MMIO |
4373 | * @rdev: radeon device structure |
4374 | * |
4375 | * Some R6XX/R7XX don't seem to take into account HDP flushes performed |
4376 | * through the ring buffer. This leads to corruption in rendering, see |
4377 | * http://bugzilla.kernel.org/show_bug.cgi?id=15186 . To avoid this, we |
4378 | * directly perform the HDP flush by writing the register through MMIO. |
4379 | */ |
4380 | void r600_mmio_hdp_flush(struct radeon_device *rdev) |
4381 | { |
4382 | /* r7xx hw bug. write to HDP_DEBUG1 followed by fb read |
4383 | * rather than write to HDP_REG_COHERENCY_FLUSH_CNTL. |
4384 | * This seems to cause problems on some AGP cards. Just use the old |
4385 | * method for them. |
4386 | */ |
4387 | if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_RV740) && |
4388 | rdev->vram_scratch.ptr && !(rdev->flags & RADEON_IS_AGP)) { |
4389 | void __iomem *ptr = (void *)rdev->vram_scratch.ptr; |
4390 | |
4391 | WREG32(HDP_DEBUG1, 0)r100_mm_wreg(rdev, (0x2F34), (0), 0); |
4392 | readl((void __iomem *)ptr)ioread32((void *)ptr); |
4393 | } else |
4394 | WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1)r100_mm_wreg(rdev, (0x5480), (0x1), 0); |
4395 | } |
4396 | |
4397 | void r600_set_pcie_lanes(struct radeon_device *rdev, int lanes) |
4398 | { |
4399 | u32 link_width_cntl, mask; |
4400 | |
4401 | if (rdev->flags & RADEON_IS_IGP) |
4402 | return; |
4403 | |
4404 | if (!(rdev->flags & RADEON_IS_PCIE)) |
4405 | return; |
4406 | |
4407 | /* x2 cards have a special sequence */ |
4408 | if (ASIC_IS_X2(rdev)((rdev->pdev->device == 0x9441) || (rdev->pdev->device == 0x9443) || (rdev->pdev->device == 0x944B) || (rdev-> pdev->device == 0x9506) || (rdev->pdev->device == 0x9509 ) || (rdev->pdev->device == 0x950F) || (rdev->pdev-> device == 0x689C) || (rdev->pdev->device == 0x689D))) |
4409 | return; |
4410 | |
4411 | radeon_gui_idle(rdev)(rdev)->asic->gui_idle((rdev)); |
4412 | |
4413 | switch (lanes) { |
4414 | case 0: |
4415 | mask = RADEON_PCIE_LC_LINK_WIDTH_X00; |
4416 | break; |
4417 | case 1: |
4418 | mask = RADEON_PCIE_LC_LINK_WIDTH_X11; |
4419 | break; |
4420 | case 2: |
4421 | mask = RADEON_PCIE_LC_LINK_WIDTH_X22; |
4422 | break; |
4423 | case 4: |
4424 | mask = RADEON_PCIE_LC_LINK_WIDTH_X43; |
4425 | break; |
4426 | case 8: |
4427 | mask = RADEON_PCIE_LC_LINK_WIDTH_X84; |
4428 | break; |
4429 | case 12: |
4430 | /* not actually supported */ |
4431 | mask = RADEON_PCIE_LC_LINK_WIDTH_X125; |
4432 | break; |
4433 | case 16: |
4434 | mask = RADEON_PCIE_LC_LINK_WIDTH_X166; |
4435 | break; |
4436 | default: |
4437 | DRM_ERROR("invalid pcie lane request: %d\n", lanes)__drm_err("invalid pcie lane request: %d\n", lanes); |
4438 | return; |
4439 | } |
4440 | |
4441 | link_width_cntl = RREG32_PCIE_PORT(RADEON_PCIE_LC_LINK_WIDTH_CNTL)rdev->pciep_rreg(rdev, (0xa2)); |
4442 | link_width_cntl &= ~RADEON_PCIE_LC_LINK_WIDTH_MASK0x7; |
4443 | link_width_cntl |= mask << RADEON_PCIE_LC_LINK_WIDTH_SHIFT0; |
4444 | link_width_cntl |= (RADEON_PCIE_LC_RECONFIG_NOW(1 << 8) | |
4445 | R600_PCIE_LC_RECONFIG_ARC_MISSING_ESCAPE(1 << 7)); |
4446 | |
4447 | WREG32_PCIE_PORT(RADEON_PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl)rdev->pciep_wreg(rdev, (0xa2), (link_width_cntl)); |
4448 | } |
4449 | |
4450 | int r600_get_pcie_lanes(struct radeon_device *rdev) |
4451 | { |
4452 | u32 link_width_cntl; |
4453 | |
4454 | if (rdev->flags & RADEON_IS_IGP) |
4455 | return 0; |
4456 | |
4457 | if (!(rdev->flags & RADEON_IS_PCIE)) |
4458 | return 0; |
4459 | |
4460 | /* x2 cards have a special sequence */ |
4461 | if (ASIC_IS_X2(rdev)((rdev->pdev->device == 0x9441) || (rdev->pdev->device == 0x9443) || (rdev->pdev->device == 0x944B) || (rdev-> pdev->device == 0x9506) || (rdev->pdev->device == 0x9509 ) || (rdev->pdev->device == 0x950F) || (rdev->pdev-> device == 0x689C) || (rdev->pdev->device == 0x689D))) |
4462 | return 0; |
4463 | |
4464 | radeon_gui_idle(rdev)(rdev)->asic->gui_idle((rdev)); |
4465 | |
4466 | link_width_cntl = RREG32_PCIE_PORT(RADEON_PCIE_LC_LINK_WIDTH_CNTL)rdev->pciep_rreg(rdev, (0xa2)); |
4467 | |
4468 | switch ((link_width_cntl & RADEON_PCIE_LC_LINK_WIDTH_RD_MASK0x70) >> RADEON_PCIE_LC_LINK_WIDTH_RD_SHIFT4) { |
4469 | case RADEON_PCIE_LC_LINK_WIDTH_X11: |
4470 | return 1; |
4471 | case RADEON_PCIE_LC_LINK_WIDTH_X22: |
4472 | return 2; |
4473 | case RADEON_PCIE_LC_LINK_WIDTH_X43: |
4474 | return 4; |
4475 | case RADEON_PCIE_LC_LINK_WIDTH_X84: |
4476 | return 8; |
4477 | case RADEON_PCIE_LC_LINK_WIDTH_X125: |
4478 | /* not actually supported */ |
4479 | return 12; |
4480 | case RADEON_PCIE_LC_LINK_WIDTH_X00: |
4481 | case RADEON_PCIE_LC_LINK_WIDTH_X166: |
4482 | default: |
4483 | return 16; |
4484 | } |
4485 | } |
4486 | |
4487 | static void r600_pcie_gen2_enable(struct radeon_device *rdev) |
4488 | { |
4489 | u32 link_width_cntl, lanes, speed_cntl, training_cntl, tmp; |
4490 | u16 link_cntl2; |
4491 | enum pci_bus_speed max_bus_speed; |
4492 | |
4493 | if (radeon_pcie_gen2 == 0) |
4494 | return; |
4495 | |
4496 | if (rdev->flags & RADEON_IS_IGP) |
4497 | return; |
4498 | |
4499 | if (!(rdev->flags & RADEON_IS_PCIE)) |
4500 | return; |
4501 | |
4502 | /* x2 cards have a special sequence */ |
4503 | if (ASIC_IS_X2(rdev)((rdev->pdev->device == 0x9441) || (rdev->pdev->device == 0x9443) || (rdev->pdev->device == 0x944B) || (rdev-> pdev->device == 0x9506) || (rdev->pdev->device == 0x9509 ) || (rdev->pdev->device == 0x950F) || (rdev->pdev-> device == 0x689C) || (rdev->pdev->device == 0x689D))) |
4504 | return; |
4505 | |
4506 | /* only RV6xx+ chips are supported */ |
4507 | if (rdev->family <= CHIP_R600) |
4508 | return; |
4509 | |
4510 | max_bus_speed = pcie_get_speed_cap(rdev->pdev->bus->self); |
4511 | if ((max_bus_speed != PCIE_SPEED_5_0GT) && |
4512 | (max_bus_speed != PCIE_SPEED_8_0GT)) |
4513 | return; |
4514 | |
4515 | speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL)rdev->pciep_rreg(rdev, (0xa4)); |
4516 | if (speed_cntl & LC_CURRENT_DATA_RATE(1 << 11)) { |
4517 | DRM_INFO("PCIE gen 2 link speeds already enabled\n")printk("\0016" "[" "drm" "] " "PCIE gen 2 link speeds already enabled\n" ); |
4518 | return; |
4519 | } |
4520 | |
4521 | DRM_INFO("enabling PCIE gen 2 link speeds, disable with radeon.pcie_gen2=0\n")printk("\0016" "[" "drm" "] " "enabling PCIE gen 2 link speeds, disable with radeon.pcie_gen2=0\n" ); |
4522 | |
4523 | /* 55 nm r6xx asics */ |
4524 | if ((rdev->family == CHIP_RV670) || |
4525 | (rdev->family == CHIP_RV620) || |
4526 | (rdev->family == CHIP_RV635)) { |
4527 | /* advertise upconfig capability */ |
4528 | link_width_cntl = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL)rdev->pciep_rreg(rdev, (0xa2)); |
4529 | link_width_cntl &= ~LC_UPCONFIGURE_DIS(1 << 13); |
4530 | WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl)rdev->pciep_wreg(rdev, (0xa2), (link_width_cntl)); |
4531 | link_width_cntl = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL)rdev->pciep_rreg(rdev, (0xa2)); |
4532 | if (link_width_cntl & LC_RENEGOTIATION_SUPPORT(1 << 9)) { |
4533 | lanes = (link_width_cntl & LC_LINK_WIDTH_RD_MASK0x70) >> LC_LINK_WIDTH_RD_SHIFT4; |
4534 | link_width_cntl &= ~(LC_LINK_WIDTH_MASK0x7 | |
4535 | LC_RECONFIG_ARC_MISSING_ESCAPE(1 << 7)); |
4536 | link_width_cntl |= lanes | LC_RECONFIG_NOW(1 << 8) | LC_RENEGOTIATE_EN(1 << 10); |
4537 | WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl)rdev->pciep_wreg(rdev, (0xa2), (link_width_cntl)); |
4538 | } else { |
4539 | link_width_cntl |= LC_UPCONFIGURE_DIS(1 << 13); |
4540 | WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl)rdev->pciep_wreg(rdev, (0xa2), (link_width_cntl)); |
4541 | } |
4542 | } |
4543 | |
4544 | speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL)rdev->pciep_rreg(rdev, (0xa4)); |
4545 | if ((speed_cntl & LC_OTHER_SIDE_EVER_SENT_GEN2(1 << 23)) && |
4546 | (speed_cntl & LC_OTHER_SIDE_SUPPORTS_GEN2(1 << 24))) { |
4547 | |
4548 | /* 55 nm r6xx asics */ |
4549 | if ((rdev->family == CHIP_RV670) || |
4550 | (rdev->family == CHIP_RV620) || |
4551 | (rdev->family == CHIP_RV635)) { |
4552 | WREG32(MM_CFGREGS_CNTL, 0x8)r100_mm_wreg(rdev, (0x544c), (0x8), 0); |
4553 | link_cntl2 = RREG32(0x4088)r100_mm_rreg(rdev, (0x4088), 0); |
4554 | WREG32(MM_CFGREGS_CNTL, 0)r100_mm_wreg(rdev, (0x544c), (0), 0); |
4555 | /* not supported yet */ |
4556 | if (link_cntl2 & SELECTABLE_DEEMPHASIS(1 << 6)) |
4557 | return; |
4558 | } |
4559 | |
4560 | speed_cntl &= ~LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK(0x3 << 8); |
4561 | speed_cntl |= (0x3 << LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT3); |
4562 | speed_cntl &= ~LC_VOLTAGE_TIMER_SEL_MASK(0xf << 14); |
4563 | speed_cntl &= ~LC_FORCE_DIS_HW_SPEED_CHANGE(1 << 6); |
4564 | speed_cntl |= LC_FORCE_EN_HW_SPEED_CHANGE(1 << 5); |
4565 | WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl)rdev->pciep_wreg(rdev, (0xa4), (speed_cntl)); |
4566 | |
4567 | tmp = RREG32(0x541c)r100_mm_rreg(rdev, (0x541c), 0); |
4568 | WREG32(0x541c, tmp | 0x8)r100_mm_wreg(rdev, (0x541c), (tmp | 0x8), 0); |
4569 | WREG32(MM_CFGREGS_CNTL, MM_WR_TO_CFG_EN)r100_mm_wreg(rdev, (0x544c), ((1 << 3)), 0); |
4570 | link_cntl2 = RREG16(0x4088)ioread16((rdev->rmmio) + (0x4088)); |
4571 | link_cntl2 &= ~TARGET_LINK_SPEED_MASK(0xf << 0); |
4572 | link_cntl2 |= 0x2; |
4573 | WREG16(0x4088, link_cntl2)iowrite16(link_cntl2, (rdev->rmmio) + (0x4088)); |
4574 | WREG32(MM_CFGREGS_CNTL, 0)r100_mm_wreg(rdev, (0x544c), (0), 0); |
4575 | |
4576 | if ((rdev->family == CHIP_RV670) || |
4577 | (rdev->family == CHIP_RV620) || |
4578 | (rdev->family == CHIP_RV635)) { |
4579 | training_cntl = RREG32_PCIE_PORT(PCIE_LC_TRAINING_CNTL)rdev->pciep_rreg(rdev, (0xa1)); |
4580 | training_cntl &= ~LC_POINT_7_PLUS_EN(1 << 6); |
4581 | WREG32_PCIE_PORT(PCIE_LC_TRAINING_CNTL, training_cntl)rdev->pciep_wreg(rdev, (0xa1), (training_cntl)); |
4582 | } else { |
4583 | speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL)rdev->pciep_rreg(rdev, (0xa4)); |
4584 | speed_cntl &= ~LC_TARGET_LINK_SPEED_OVERRIDE_EN(1 << 1); |
4585 | WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl)rdev->pciep_wreg(rdev, (0xa4), (speed_cntl)); |
4586 | } |
4587 | |
4588 | speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL)rdev->pciep_rreg(rdev, (0xa4)); |
4589 | speed_cntl |= LC_GEN2_EN_STRAP(1 << 0); |
4590 | WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl)rdev->pciep_wreg(rdev, (0xa4), (speed_cntl)); |
4591 | |
4592 | } else { |
4593 | link_width_cntl = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL)rdev->pciep_rreg(rdev, (0xa2)); |
4594 | /* XXX: only disable it if gen1 bridge vendor == 0x111d or 0x1106 */ |
4595 | if (1) |
4596 | link_width_cntl |= LC_UPCONFIGURE_DIS(1 << 13); |
4597 | else |
4598 | link_width_cntl &= ~LC_UPCONFIGURE_DIS(1 << 13); |
4599 | WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl)rdev->pciep_wreg(rdev, (0xa2), (link_width_cntl)); |
4600 | } |
4601 | } |
4602 | |
4603 | /** |
4604 | * r600_get_gpu_clock_counter - return GPU clock counter snapshot |
4605 | * |
4606 | * @rdev: radeon_device pointer |
4607 | * |
4608 | * Fetches a GPU clock counter snapshot (R6xx-cayman). |
4609 | * Returns the 64 bit clock counter snapshot. |
4610 | */ |
4611 | uint64_t r600_get_gpu_clock_counter(struct radeon_device *rdev) |
4612 | { |
4613 | uint64_t clock; |
4614 | |
4615 | mutex_lock(&rdev->gpu_clock_mutex)rw_enter_write(&rdev->gpu_clock_mutex); |
4616 | WREG32(RLC_CAPTURE_GPU_CLOCK_COUNT, 1)r100_mm_wreg(rdev, (0x3f40), (1), 0); |
4617 | clock = (uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_LSB)r100_mm_rreg(rdev, (0x3f38), 0) | |
4618 | ((uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_MSB)r100_mm_rreg(rdev, (0x3f3c), 0) << 32ULL); |
4619 | mutex_unlock(&rdev->gpu_clock_mutex)rw_exit_write(&rdev->gpu_clock_mutex); |
4620 | return clock; |
4621 | } |