diff options
Diffstat (limited to 'drivers/gpu/nvgpu/common/linux/driver_common.c')
-rw-r--r-- | drivers/gpu/nvgpu/common/linux/driver_common.c | 292 |
1 files changed, 292 insertions, 0 deletions
diff --git a/drivers/gpu/nvgpu/common/linux/driver_common.c b/drivers/gpu/nvgpu/common/linux/driver_common.c new file mode 100644 index 00000000..9e567c27 --- /dev/null +++ b/drivers/gpu/nvgpu/common/linux/driver_common.c | |||
@@ -0,0 +1,292 @@ | |||
1 | /* | ||
2 | * Copyright (c) 2016-2017, NVIDIA CORPORATION. All rights reserved. | ||
3 | * | ||
4 | * This program is free software; you can redistribute it and/or modify it | ||
5 | * under the terms and conditions of the GNU General Public License, | ||
6 | * version 2, as published by the Free Software Foundation. | ||
7 | * | ||
8 | * This program is distributed in the hope it will be useful, but WITHOUT | ||
9 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | ||
10 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | ||
11 | * more details. | ||
12 | * | ||
13 | * You should have received a copy of the GNU General Public License | ||
14 | * along with this program. If not, see <http://www.gnu.org/licenses/>. | ||
15 | */ | ||
16 | |||
17 | #include <linux/dma-mapping.h> | ||
18 | #include <uapi/linux/nvgpu.h> | ||
19 | |||
20 | #include <nvgpu/kmem.h> | ||
21 | #include <nvgpu/nvgpu_common.h> | ||
22 | #include <nvgpu/soc.h> | ||
23 | #include <nvgpu/bug.h> | ||
24 | #include <nvgpu/enabled.h> | ||
25 | #include <nvgpu/debug.h> | ||
26 | |||
27 | #include "scale.h" | ||
28 | #include "gk20a/gk20a.h" | ||
29 | #include "platform_gk20a.h" | ||
30 | #include "module.h" | ||
31 | #include "os_linux.h" | ||
32 | #include "sysfs.h" | ||
33 | #include "ioctl.h" | ||
34 | |||
35 | #define EMC3D_DEFAULT_RATIO 750 | ||
36 | |||
37 | static void nvgpu_init_vars(struct gk20a *g) | ||
38 | { | ||
39 | struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g); | ||
40 | struct device *dev = dev_from_gk20a(g); | ||
41 | struct gk20a_platform *platform = dev_get_drvdata(dev); | ||
42 | |||
43 | nvgpu_cond_init(&l->sw_irq_stall_last_handled_wq); | ||
44 | nvgpu_cond_init(&l->sw_irq_nonstall_last_handled_wq); | ||
45 | gk20a_init_gr(g); | ||
46 | |||
47 | init_rwsem(&l->busy_lock); | ||
48 | nvgpu_rwsem_init(&g->deterministic_busy); | ||
49 | |||
50 | nvgpu_spinlock_init(&g->mc_enable_lock); | ||
51 | |||
52 | nvgpu_mutex_init(&platform->railgate_lock); | ||
53 | nvgpu_mutex_init(&g->dbg_sessions_lock); | ||
54 | nvgpu_mutex_init(&g->client_lock); | ||
55 | nvgpu_mutex_init(&g->poweron_lock); | ||
56 | nvgpu_mutex_init(&g->poweroff_lock); | ||
57 | |||
58 | l->regs_saved = l->regs; | ||
59 | l->bar1_saved = l->bar1; | ||
60 | |||
61 | g->emc3d_ratio = EMC3D_DEFAULT_RATIO; | ||
62 | |||
63 | /* Set DMA parameters to allow larger sgt lists */ | ||
64 | dev->dma_parms = &l->dma_parms; | ||
65 | dma_set_max_seg_size(dev, UINT_MAX); | ||
66 | |||
67 | nvgpu_init_list_node(&g->pending_sema_waits); | ||
68 | nvgpu_raw_spinlock_init(&g->pending_sema_waits_lock); | ||
69 | |||
70 | nvgpu_init_list_node(&g->profiler_objects); | ||
71 | |||
72 | nvgpu_init_list_node(&g->boardobj_head); | ||
73 | nvgpu_init_list_node(&g->boardobjgrp_head); | ||
74 | } | ||
75 | |||
76 | static void nvgpu_init_timeout(struct gk20a *g) | ||
77 | { | ||
78 | struct gk20a_platform *platform = dev_get_drvdata(dev_from_gk20a(g)); | ||
79 | |||
80 | g->gr_idle_timeout_default = CONFIG_GK20A_DEFAULT_TIMEOUT; | ||
81 | if (nvgpu_platform_is_silicon(g)) | ||
82 | g->timeouts_enabled = true; | ||
83 | else if (nvgpu_platform_is_fpga(g)) { | ||
84 | g->gr_idle_timeout_default = GK20A_TIMEOUT_FPGA; | ||
85 | g->timeouts_enabled = true; | ||
86 | } | ||
87 | g->ch_wdt_timeout_ms = platform->ch_wdt_timeout_ms; | ||
88 | } | ||
89 | |||
90 | static void nvgpu_init_timeslice(struct gk20a *g) | ||
91 | { | ||
92 | g->runlist_interleave = true; | ||
93 | |||
94 | g->timeslice_low_priority_us = 1300; | ||
95 | g->timeslice_medium_priority_us = 2600; | ||
96 | g->timeslice_high_priority_us = 5200; | ||
97 | |||
98 | g->min_timeslice_us = 1000; | ||
99 | g->max_timeslice_us = 50000; | ||
100 | } | ||
101 | |||
102 | static void nvgpu_init_pm_vars(struct gk20a *g) | ||
103 | { | ||
104 | struct gk20a_platform *platform = dev_get_drvdata(dev_from_gk20a(g)); | ||
105 | |||
106 | /* | ||
107 | * Set up initial power settings. For non-slicon platforms, disable | ||
108 | * power features and for silicon platforms, read from platform data | ||
109 | */ | ||
110 | g->slcg_enabled = | ||
111 | nvgpu_platform_is_silicon(g) ? platform->enable_slcg : false; | ||
112 | g->blcg_enabled = | ||
113 | nvgpu_platform_is_silicon(g) ? platform->enable_blcg : false; | ||
114 | g->elcg_enabled = | ||
115 | nvgpu_platform_is_silicon(g) ? platform->enable_elcg : false; | ||
116 | g->elpg_enabled = | ||
117 | nvgpu_platform_is_silicon(g) ? platform->enable_elpg : false; | ||
118 | g->aelpg_enabled = | ||
119 | nvgpu_platform_is_silicon(g) ? platform->enable_aelpg : false; | ||
120 | g->mscg_enabled = | ||
121 | nvgpu_platform_is_silicon(g) ? platform->enable_mscg : false; | ||
122 | g->can_elpg = | ||
123 | nvgpu_platform_is_silicon(g) ? platform->can_elpg_init : false; | ||
124 | |||
125 | __nvgpu_set_enabled(g, NVGPU_GPU_CAN_ELCG, | ||
126 | nvgpu_platform_is_silicon(g) ? platform->can_elcg : false); | ||
127 | __nvgpu_set_enabled(g, NVGPU_GPU_CAN_SLCG, | ||
128 | nvgpu_platform_is_silicon(g) ? platform->can_slcg : false); | ||
129 | __nvgpu_set_enabled(g, NVGPU_GPU_CAN_BLCG, | ||
130 | nvgpu_platform_is_silicon(g) ? platform->can_blcg : false); | ||
131 | |||
132 | g->default_pri_timeout = platform->default_pri_timeout; | ||
133 | g->aggressive_sync_destroy = platform->aggressive_sync_destroy; | ||
134 | g->aggressive_sync_destroy_thresh = platform->aggressive_sync_destroy_thresh; | ||
135 | g->has_syncpoints = platform->has_syncpoints; | ||
136 | g->ptimer_src_freq = platform->ptimer_src_freq; | ||
137 | g->support_pmu = support_gk20a_pmu(dev_from_gk20a(g)); | ||
138 | g->can_railgate = platform->can_railgate_init; | ||
139 | g->railgate_delay = platform->railgate_delay_init; | ||
140 | __nvgpu_set_enabled(g, NVGPU_PMU_PERFMON, platform->enable_perfmon); | ||
141 | |||
142 | /* set default values to aelpg parameters */ | ||
143 | g->pmu.aelpg_param[0] = APCTRL_SAMPLING_PERIOD_PG_DEFAULT_US; | ||
144 | g->pmu.aelpg_param[1] = APCTRL_MINIMUM_IDLE_FILTER_DEFAULT_US; | ||
145 | g->pmu.aelpg_param[2] = APCTRL_MINIMUM_TARGET_SAVING_DEFAULT_US; | ||
146 | g->pmu.aelpg_param[3] = APCTRL_POWER_BREAKEVEN_DEFAULT_US; | ||
147 | g->pmu.aelpg_param[4] = APCTRL_CYCLES_PER_SAMPLE_MAX_DEFAULT; | ||
148 | |||
149 | __nvgpu_set_enabled(g, NVGPU_SUPPORT_ASPM, !platform->disable_aspm); | ||
150 | } | ||
151 | |||
152 | static void nvgpu_init_vbios_vars(struct gk20a *g) | ||
153 | { | ||
154 | struct gk20a_platform *platform = dev_get_drvdata(dev_from_gk20a(g)); | ||
155 | |||
156 | __nvgpu_set_enabled(g, NVGPU_PMU_RUN_PREOS, platform->run_preos); | ||
157 | g->vbios_min_version = platform->vbios_min_version; | ||
158 | } | ||
159 | |||
160 | static void nvgpu_init_mm_vars(struct gk20a *g) | ||
161 | { | ||
162 | struct gk20a_platform *platform = dev_get_drvdata(dev_from_gk20a(g)); | ||
163 | |||
164 | g->mm.bypass_smmu = platform->bypass_smmu; | ||
165 | g->mm.disable_bigpage = platform->disable_bigpage; | ||
166 | __nvgpu_set_enabled(g, NVGPU_MM_HONORS_APERTURE, | ||
167 | platform->honors_aperture); | ||
168 | __nvgpu_set_enabled(g, NVGPU_MM_UNIFIED_MEMORY, | ||
169 | platform->unified_memory); | ||
170 | __nvgpu_set_enabled(g, NVGPU_MM_UNIFY_ADDRESS_SPACES, | ||
171 | platform->unify_address_spaces); | ||
172 | __nvgpu_set_enabled(g, NVGPU_MM_CE_TSG_REQUIRED, | ||
173 | platform->tsg_required); | ||
174 | |||
175 | nvgpu_mutex_init(&g->mm.tlb_lock); | ||
176 | nvgpu_mutex_init(&g->mm.priv_lock); | ||
177 | } | ||
178 | |||
179 | int nvgpu_probe(struct gk20a *g, | ||
180 | const char *debugfs_symlink, | ||
181 | const char *interface_name, | ||
182 | struct class *class) | ||
183 | { | ||
184 | struct device *dev = dev_from_gk20a(g); | ||
185 | struct gk20a_platform *platform = dev_get_drvdata(dev); | ||
186 | int err = 0; | ||
187 | |||
188 | nvgpu_init_vars(g); | ||
189 | nvgpu_init_timeout(g); | ||
190 | nvgpu_init_timeslice(g); | ||
191 | nvgpu_init_pm_vars(g); | ||
192 | nvgpu_init_vbios_vars(g); | ||
193 | |||
194 | /* Initialize the platform interface. */ | ||
195 | err = platform->probe(dev); | ||
196 | if (err) { | ||
197 | if (err == -EPROBE_DEFER) | ||
198 | nvgpu_info(g, "platform probe failed"); | ||
199 | else | ||
200 | nvgpu_err(g, "platform probe failed"); | ||
201 | return err; | ||
202 | } | ||
203 | |||
204 | /* platform probe can defer do user init only if probe succeeds */ | ||
205 | err = gk20a_user_init(dev, interface_name, class); | ||
206 | if (err) | ||
207 | return err; | ||
208 | |||
209 | |||
210 | /* Initialise scaling */ | ||
211 | if (IS_ENABLED(CONFIG_GK20A_DEVFREQ)) | ||
212 | gk20a_scale_init(dev); | ||
213 | |||
214 | if (platform->late_probe) { | ||
215 | err = platform->late_probe(dev); | ||
216 | if (err) { | ||
217 | nvgpu_err(g, "late probe failed"); | ||
218 | return err; | ||
219 | } | ||
220 | } | ||
221 | |||
222 | nvgpu_init_mm_vars(g); | ||
223 | |||
224 | nvgpu_create_sysfs(dev); | ||
225 | gk20a_debug_init(g, debugfs_symlink); | ||
226 | |||
227 | g->dbg_regops_tmp_buf = nvgpu_kzalloc(g, SZ_4K); | ||
228 | if (!g->dbg_regops_tmp_buf) { | ||
229 | nvgpu_err(g, "couldn't allocate regops tmp buf"); | ||
230 | return -ENOMEM; | ||
231 | } | ||
232 | g->dbg_regops_tmp_buf_ops = | ||
233 | SZ_4K / sizeof(g->dbg_regops_tmp_buf[0]); | ||
234 | |||
235 | g->remove_support = gk20a_remove_support; | ||
236 | |||
237 | nvgpu_ref_init(&g->refcount); | ||
238 | |||
239 | return 0; | ||
240 | } | ||
241 | |||
242 | /** | ||
243 | * cyclic_delta - Returns delta of cyclic integers a and b. | ||
244 | * | ||
245 | * @a - First integer | ||
246 | * @b - Second integer | ||
247 | * | ||
248 | * Note: if a is ahead of b, delta is positive. | ||
249 | */ | ||
250 | static int cyclic_delta(int a, int b) | ||
251 | { | ||
252 | return a - b; | ||
253 | } | ||
254 | |||
255 | /** | ||
256 | * nvgpu_wait_for_deferred_interrupts - Wait for interrupts to complete | ||
257 | * | ||
258 | * @g - The GPU to wait on. | ||
259 | * | ||
260 | * Waits until all interrupt handlers that have been scheduled to run have | ||
261 | * completed. | ||
262 | */ | ||
263 | void nvgpu_wait_for_deferred_interrupts(struct gk20a *g) | ||
264 | { | ||
265 | struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g); | ||
266 | int stall_irq_threshold = atomic_read(&l->hw_irq_stall_count); | ||
267 | int nonstall_irq_threshold = atomic_read(&l->hw_irq_nonstall_count); | ||
268 | |||
269 | /* wait until all stalling irqs are handled */ | ||
270 | NVGPU_COND_WAIT(&l->sw_irq_stall_last_handled_wq, | ||
271 | cyclic_delta(stall_irq_threshold, | ||
272 | atomic_read(&l->sw_irq_stall_last_handled)) | ||
273 | <= 0, 0); | ||
274 | |||
275 | /* wait until all non-stalling irqs are handled */ | ||
276 | NVGPU_COND_WAIT(&l->sw_irq_nonstall_last_handled_wq, | ||
277 | cyclic_delta(nonstall_irq_threshold, | ||
278 | atomic_read(&l->sw_irq_nonstall_last_handled)) | ||
279 | <= 0, 0); | ||
280 | } | ||
281 | |||
282 | static void nvgpu_free_gk20a(struct gk20a *g) | ||
283 | { | ||
284 | struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g); | ||
285 | |||
286 | kfree(l); | ||
287 | } | ||
288 | |||
289 | void nvgpu_init_gk20a(struct gk20a *g) | ||
290 | { | ||
291 | g->free = nvgpu_free_gk20a; | ||
292 | } | ||