]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - drivers/gpu/host1x/hw/channel_hw.c
27101c04a8272668988ce5be66dfc584068f8a60
[thirdparty/kernel/stable.git] / drivers / gpu / host1x / hw / channel_hw.c
1 /*
2 * Tegra host1x Channel
3 *
4 * Copyright (c) 2010-2013, NVIDIA Corporation.
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program. If not, see <http://www.gnu.org/licenses/>.
17 */
18
19 #include <linux/host1x.h>
20 #include <linux/iommu.h>
21 #include <linux/slab.h>
22
23 #include <trace/events/host1x.h>
24
25 #include "../channel.h"
26 #include "../dev.h"
27 #include "../intr.h"
28 #include "../job.h"
29
30 #define TRACE_MAX_LENGTH 128U
31
32 static void trace_write_gather(struct host1x_cdma *cdma, struct host1x_bo *bo,
33 u32 offset, u32 words)
34 {
35 struct device *dev = cdma_to_channel(cdma)->dev;
36 void *mem = NULL;
37
38 if (host1x_debug_trace_cmdbuf)
39 mem = host1x_bo_mmap(bo);
40
41 if (mem) {
42 u32 i;
43 /*
44 * Write in batches of 128 as there seems to be a limit
45 * of how much you can output to ftrace at once.
46 */
47 for (i = 0; i < words; i += TRACE_MAX_LENGTH) {
48 u32 num_words = min(words - i, TRACE_MAX_LENGTH);
49
50 offset += i * sizeof(u32);
51
52 trace_host1x_cdma_push_gather(dev_name(dev), bo,
53 num_words, offset,
54 mem);
55 }
56
57 host1x_bo_munmap(bo, mem);
58 }
59 }
60
61 static void submit_gathers(struct host1x_job *job)
62 {
63 struct host1x_cdma *cdma = &job->channel->cdma;
64 #if HOST1X_HW < 6
65 struct device *dev = job->channel->dev;
66 #endif
67 unsigned int i;
68
69 for (i = 0; i < job->num_gathers; i++) {
70 struct host1x_job_gather *g = &job->gathers[i];
71 dma_addr_t addr = g->base + g->offset;
72 u32 op2, op3;
73
74 op2 = lower_32_bits(addr);
75 op3 = upper_32_bits(addr);
76
77 trace_write_gather(cdma, g->bo, g->offset, g->words);
78
79 if (op3 != 0) {
80 #if HOST1X_HW >= 6
81 u32 op1 = host1x_opcode_gather_wide(g->words);
82 u32 op4 = HOST1X_OPCODE_NOP;
83
84 host1x_cdma_push_wide(cdma, op1, op2, op3, op4);
85 #else
86 dev_err(dev, "invalid gather for push buffer %pad\n",
87 &addr);
88 continue;
89 #endif
90 } else {
91 u32 op1 = host1x_opcode_gather(g->words);
92
93 host1x_cdma_push(cdma, op1, op2);
94 }
95 }
96 }
97
98 static inline void synchronize_syncpt_base(struct host1x_job *job)
99 {
100 struct host1x *host = dev_get_drvdata(job->channel->dev->parent);
101 struct host1x_syncpt *sp = host->syncpt + job->syncpt_id;
102 unsigned int id;
103 u32 value;
104
105 value = host1x_syncpt_read_max(sp);
106 id = sp->base->id;
107
108 host1x_cdma_push(&job->channel->cdma,
109 host1x_opcode_setclass(HOST1X_CLASS_HOST1X,
110 HOST1X_UCLASS_LOAD_SYNCPT_BASE, 1),
111 HOST1X_UCLASS_LOAD_SYNCPT_BASE_BASE_INDX_F(id) |
112 HOST1X_UCLASS_LOAD_SYNCPT_BASE_VALUE_F(value));
113 }
114
115 static void host1x_channel_set_streamid(struct host1x_channel *channel)
116 {
117 #if HOST1X_HW >= 6
118 struct iommu_fwspec *spec = dev_iommu_fwspec_get(channel->dev->parent);
119 u32 sid = spec ? spec->ids[0] & 0xffff : 0x7f;
120
121 host1x_ch_writel(channel, sid, HOST1X_CHANNEL_SMMU_STREAMID);
122 #endif
123 }
124
125 static int channel_submit(struct host1x_job *job)
126 {
127 struct host1x_channel *ch = job->channel;
128 struct host1x_syncpt *sp;
129 u32 user_syncpt_incrs = job->syncpt_incrs;
130 u32 prev_max = 0;
131 u32 syncval;
132 int err;
133 struct host1x_waitlist *completed_waiter = NULL;
134 struct host1x *host = dev_get_drvdata(ch->dev->parent);
135
136 sp = host->syncpt + job->syncpt_id;
137 trace_host1x_channel_submit(dev_name(ch->dev),
138 job->num_gathers, job->num_relocs,
139 job->syncpt_id, job->syncpt_incrs);
140
141 /* before error checks, return current max */
142 prev_max = job->syncpt_end = host1x_syncpt_read_max(sp);
143
144 /* get submit lock */
145 err = mutex_lock_interruptible(&ch->submitlock);
146 if (err)
147 goto error;
148
149 completed_waiter = kzalloc(sizeof(*completed_waiter), GFP_KERNEL);
150 if (!completed_waiter) {
151 mutex_unlock(&ch->submitlock);
152 err = -ENOMEM;
153 goto error;
154 }
155
156 host1x_channel_set_streamid(ch);
157
158 /* begin a CDMA submit */
159 err = host1x_cdma_begin(&ch->cdma, job);
160 if (err) {
161 mutex_unlock(&ch->submitlock);
162 goto error;
163 }
164
165 if (job->serialize) {
166 /*
167 * Force serialization by inserting a host wait for the
168 * previous job to finish before this one can commence.
169 */
170 host1x_cdma_push(&ch->cdma,
171 host1x_opcode_setclass(HOST1X_CLASS_HOST1X,
172 host1x_uclass_wait_syncpt_r(), 1),
173 host1x_class_host_wait_syncpt(job->syncpt_id,
174 host1x_syncpt_read_max(sp)));
175 }
176
177 /* Synchronize base register to allow using it for relative waiting */
178 if (sp->base)
179 synchronize_syncpt_base(job);
180
181 syncval = host1x_syncpt_incr_max(sp, user_syncpt_incrs);
182
183 host1x_hw_syncpt_assign_to_channel(host, sp, ch);
184
185 job->syncpt_end = syncval;
186
187 /* add a setclass for modules that require it */
188 if (job->class)
189 host1x_cdma_push(&ch->cdma,
190 host1x_opcode_setclass(job->class, 0, 0),
191 HOST1X_OPCODE_NOP);
192
193 submit_gathers(job);
194
195 /* end CDMA submit & stash pinned hMems into sync queue */
196 host1x_cdma_end(&ch->cdma, job);
197
198 trace_host1x_channel_submitted(dev_name(ch->dev), prev_max, syncval);
199
200 /* schedule a submit complete interrupt */
201 err = host1x_intr_add_action(host, sp, syncval,
202 HOST1X_INTR_ACTION_SUBMIT_COMPLETE, ch,
203 completed_waiter, NULL);
204 completed_waiter = NULL;
205 WARN(err, "Failed to set submit complete interrupt");
206
207 mutex_unlock(&ch->submitlock);
208
209 return 0;
210
211 error:
212 kfree(completed_waiter);
213 return err;
214 }
215
216 static void enable_gather_filter(struct host1x *host,
217 struct host1x_channel *ch)
218 {
219 #if HOST1X_HW >= 6
220 u32 val;
221
222 if (!host->hv_regs)
223 return;
224
225 val = host1x_hypervisor_readl(
226 host, HOST1X_HV_CH_KERNEL_FILTER_GBUFFER(ch->id / 32));
227 val |= BIT(ch->id % 32);
228 host1x_hypervisor_writel(
229 host, val, HOST1X_HV_CH_KERNEL_FILTER_GBUFFER(ch->id / 32));
230 #elif HOST1X_HW >= 4
231 host1x_ch_writel(ch,
232 HOST1X_CHANNEL_CHANNELCTRL_KERNEL_FILTER_GBUFFER(1),
233 HOST1X_CHANNEL_CHANNELCTRL);
234 #endif
235 }
236
237 static int host1x_channel_init(struct host1x_channel *ch, struct host1x *dev,
238 unsigned int index)
239 {
240 #if HOST1X_HW < 6
241 ch->regs = dev->regs + index * 0x4000;
242 #else
243 ch->regs = dev->regs + index * 0x100;
244 #endif
245 enable_gather_filter(dev, ch);
246 return 0;
247 }
248
249 static const struct host1x_channel_ops host1x_channel_ops = {
250 .init = host1x_channel_init,
251 .submit = channel_submit,
252 };