aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc/platforms/cell/spufs/run.c
blob: 57626600b1a4b5c7cb3d46d17ef214e5b5807d1d (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
#define DEBUG

#include <linux/wait.h>
#include <linux/ptrace.h>

#include <asm/spu.h>
#include <asm/spu_priv1.h>
#include <asm/io.h>
#include <asm/unistd.h>

#include "spufs.h"

/* interrupt-level stop callback function. */
void spufs_stop_callback(struct spu *spu)
{
	struct spu_context *ctx = spu->ctx;

	wake_up_all(&ctx->stop_wq);
}

static inline int spu_stopped(struct spu_context *ctx, u32 * stat)
{
	struct spu *spu;
	u64 pte_fault;

	*stat = ctx->ops->status_read(ctx);
	if (ctx->state != SPU_STATE_RUNNABLE)
		return 1;
	spu = ctx->spu;
	pte_fault = spu->dsisr &
	    (MFC_DSISR_PTE_NOT_FOUND | MFC_DSISR_ACCESS_DENIED);
	return (!(*stat & 0x1) || pte_fault || spu->class_0_pending) ? 1 : 0;
}

static int spu_setup_isolated(struct spu_context *ctx)
{
	int ret;
	u64 __iomem *mfc_cntl;
	u64 sr1;
	u32 status;
	unsigned long timeout;
	const u32 status_loading = SPU_STATUS_RUNNING
		| SPU_STATUS_ISOLATED_STATE | SPU_STATUS_ISOLATED_LOAD_STATUS;

	ret = -ENODEV;
	if (!isolated_loader)
		goto out;

	/*
	 * We need to exclude userspace access to the context.
	 *
	 * To protect against memory access we invalidate all ptes
	 * and make sure the pagefault handlers block on the mutex.
	 */
	spu_unmap_mappings(ctx);

	mfc_cntl = &ctx->spu->priv2->mfc_control_RW;

	/* purge the MFC DMA queue to ensure no spurious accesses before we
	 * enter kernel mode */
	timeout = jiffies + HZ;
	out_be64(mfc_cntl, MFC_CNTL_PURGE_DMA_REQUEST);
	while ((in_be64(mfc_cntl) & MFC_CNTL_PURGE_DMA_STATUS_MASK)
			!= MFC_CNTL_PURGE_DMA_COMPLETE) {
		if (time_after(jiffies, timeout)) {
			printk(KERN_ERR "%s: timeout flushing MFC DMA queue\n",
					__FUNCTION__);
			ret = -EIO;
			goto out;
		}
		cond_resched();
	}

	/* put the SPE in kernel mode to allow access to the loader */
	sr1 = spu_mfc_sr1_get(ctx->spu);
	sr1 &= ~MFC_STATE1_PROBLEM_STATE_MASK;
	spu_mfc_sr1_set(ctx->spu, sr1);

	/* start the loader */
	ctx->ops->signal1_write(ctx, (unsigned long)isolated_loader >> 32);
	ctx->ops->signal2_write(ctx,
			(unsigned long)isolated_loader & 0xffffffff);

	ctx->ops->runcntl_write(ctx,
			SPU_RUNCNTL_RUNNABLE | SPU_RUNCNTL_ISOLATE);

	ret = 0;
	timeout = jiffies + HZ;
	while (((status = ctx->ops->status_read(ctx)) & status_loading) ==
				status_loading) {
		if (time_after(jiffies, timeout)) {
			printk(KERN_ERR "%s: timeout waiting for loader\n",
					__FUNCTION__);
			ret = -EIO;
			goto out_drop_priv;
		}
		cond_resched();
	}

	if (!(status & SPU_STATUS_RUNNING)) {
		/* If isolated LOAD has failed: run SPU, we will get a stop-and
		 * signal later. */
		pr_debug("%s: isolated LOAD failed\n", __FUNCTION__);
		ctx->ops->runcntl_write(ctx, SPU_RUNCNTL_RUNNABLE);
		ret = -EACCES;
		goto out_drop_priv;
	}

	if (!(status & SPU_STATUS_ISOLATED_STATE)) {
		/* This isn't allowed by the CBEA, but check anyway */
		pr_debug("%s: SPU fell out of isolated mode?\n", __FUNCTION__);
		ctx->ops->runcntl_write(ctx, SPU_RUNCNTL_STOP);
		ret = -EINVAL;
		goto out_drop_priv;
	}

out_drop_priv:
	/* Finished accessing the loader. Drop kernel mode */
	sr1 |= MFC_STATE1_PROBLEM_STATE_MASK;
	spu_mfc_sr1_set(ctx->spu, sr1);

out:
	return ret;
}

static int spu_run_init(struct spu_context *ctx, u32 * npc)
{
	if (ctx->flags & SPU_CREATE_ISOLATE) {
		unsigned long runcntl;

		if (!(ctx->ops->status_read(ctx) & SPU_STATUS_ISOLATED_STATE)) {
			int ret = spu_setup_isolated(ctx);
			if (ret)
				return ret;
		}

		/* if userspace has set the runcntrl register (eg, to issue an
		 * isolated exit), we need to re-set it here */
		runcntl = ctx->ops->runcntl_read(ctx) &
			(SPU_RUNCNTL_RUNNABLE | SPU_RUNCNTL_ISOLATE);
		if (runcntl == 0)
			runcntl = SPU_RUNCNTL_RUNNABLE;
		ctx->ops->runcntl_write(ctx, runcntl);
	} else {
		spu_start_tick(ctx);
		ctx->ops->npc_write(ctx, *npc);
		ctx->ops->runcntl_write(ctx, SPU_RUNCNTL_RUNNABLE);
	}

	return 0;
}

static int spu_run_fini(struct spu_context *ctx, u32 * npc,
			       u32 * status)
{
	int ret = 0;

	spu_stop_tick(ctx);
	*status = ctx->ops->status_read(ctx);
	*npc = ctx->ops->npc_read(ctx);
	spu_release(ctx);

	if (signal_pending(current))
		ret = -ERESTARTSYS;

	return ret;
}

static int spu_reacquire_runnable(struct spu_context *ctx, u32 *npc,
				         u32 *status)
{
	int ret;

	ret = spu_run_fini(ctx, npc, status);
	if (ret)
		return ret;

	if (*status & (SPU_STATUS_STOPPED_BY_STOP | SPU_STATUS_STOPPED_BY_HALT))
		return *status;

	ret = spu_acquire_runnable(ctx, 0);
	if (ret)
		return ret;

	ret = spu_run_init(ctx, npc);
	if (ret) {
		spu_release(ctx);
		return ret;
	}
	return 0;
}

/*
 * SPU syscall restarting is tricky because we violate the basic
 * assumption that the signal handler is running on the interrupted
 * thread. Here instead, the handler runs on PowerPC user space code,
 * while the syscall was called from the SPU.
 * This means we can only do a very rough approximation of POSIX
 * signal semantics.
 */
int spu_handle_restartsys(struct spu_context *ctx, long *spu_ret,
			  unsigned int *npc)
{
	int ret;

	switch (*spu_ret) {
	case -ERESTARTSYS:
	case -ERESTARTNOINTR:
		/*
		 * Enter the regular syscall restarting for
		 * sys_spu_run, then restart the SPU syscall
		 * callback.
		 */
		*npc -= 8;
		ret = -ERESTARTSYS;
		break;
	case -ERESTARTNOHAND:
	case -ERESTART_RESTARTBLOCK:
		/*
		 * Restart block is too hard for now, just return -EINTR
		 * to the SPU.
		 * ERESTARTNOHAND comes from sys_pause, we also return
		 * -EINTR from there.
		 * Assume that we need to be restarted ourselves though.
		 */
		*spu_ret = -EINTR;
		ret = -ERESTARTSYS;
		break;
	default:
		printk(KERN_WARNING "%s: unexpected return code %ld\n",
			__FUNCTION__, *spu_ret);
		ret = 0;
	}
	return ret;
}

int spu_process_callback(struct spu_context *ctx)
{
	struct spu_syscall_block s;
	u32 ls_pointer, npc;
	void __iomem *ls;
	long spu_ret;
	int ret;

	/* get syscall block from local store */
	npc = ctx->ops->npc_read(ctx) & ~3;
	ls = (void __iomem *)ctx->ops->get_ls(ctx);
	ls_pointer = in_be32(ls + npc);
	if (ls_pointer > (LS_SIZE - sizeof(s)))
		return -EFAULT;
	memcpy_fromio(&s, ls + ls_pointer, sizeof(s));

	/* do actual syscall without pinning the spu */
	ret = 0;
	spu_ret = -ENOSYS;
	npc += 4;

	if (s.nr_ret < __NR_syscalls) {
		spu_release(ctx);
		/* do actual system call from here */
		spu_ret = spu_sys_callback(&s);
		if (spu_ret <= -ERESTARTSYS) {
			ret = spu_handle_restartsys(ctx, &spu_ret, &npc);
		}
		spu_acquire(ctx);
		if (ret == -ERESTARTSYS)
			return ret;
	}

	/* write result, jump over indirect pointer */
	memcpy_toio(ls + ls_pointer, &spu_ret, sizeof(spu_ret));
	ctx->ops->npc_write(ctx, npc);
	ctx->ops->runcntl_write(ctx, SPU_RUNCNTL_RUNNABLE);
	return ret;
}

static inline int spu_process_events(struct spu_context *ctx)
{
	struct spu *spu = ctx->spu;
	int ret = 0;

	if (spu->class_0_pending)
		ret = spu_irq_class_0_bottom(spu);
	if (!ret && signal_pending(current))
		ret = -ERESTARTSYS;
	return ret;
}

long spufs_run_spu(struct file *file, struct spu_context *ctx,
		   u32 *npc, u32 *event)
{
	int ret;
	u32 status;

	if (mutex_lock_interruptible(&ctx->run_mutex))
		return -ERESTARTSYS;

	ctx->ops->master_start(ctx);
	ctx->event_return = 0;

	ret = spu_acquire_runnable(ctx, 0);
	if (ret)
		return ret;

	ret = spu_run_init(ctx, npc);
	if (ret) {
		spu_release(ctx);
		goto out;
	}

	do {
		ret = spufs_wait(ctx->stop_wq, spu_stopped(ctx, &status));
		if (unlikely(ret))
			break;
		if ((status & SPU_STATUS_STOPPED_BY_STOP) &&
		    (status >> SPU_STOP_STATUS_SHIFT == 0x2104)) {
			ret = spu_process_callback(ctx);
			if (ret)
				break;
			status &= ~SPU_STATUS_STOPPED_BY_STOP;
		}
		ret = spufs_handle_class1(ctx);
		if (ret)
			break;

		if (unlikely(ctx->state != SPU_STATE_RUNNABLE)) {
			ret = spu_reacquire_runnable(ctx, npc, &status);
			if (ret) {
				spu_stop_tick(ctx);
				goto out2;
			}
			continue;
		}
		ret = spu_process_events(ctx);

	} while (!ret && !(status & (SPU_STATUS_STOPPED_BY_STOP |
				      SPU_STATUS_STOPPED_BY_HALT)));

	ctx->ops->master_stop(ctx);
	ret = spu_run_fini(ctx, npc, &status);
	spu_yield(ctx);

out2:
	if ((ret == 0) ||
	    ((ret == -ERESTARTSYS) &&
	     ((status & SPU_STATUS_STOPPED_BY_HALT) ||
	      ((status & SPU_STATUS_STOPPED_BY_STOP) &&
	       (status >> SPU_STOP_STATUS_SHIFT != 0x2104)))))
		ret = status;

	if ((status & SPU_STATUS_STOPPED_BY_STOP)
	    && (status >> SPU_STOP_STATUS_SHIFT) == 0x3fff) {
		force_sig(SIGTRAP, current);
		ret = -ERESTARTSYS;
	}

out:
	*event = ctx->event_return;
	mutex_unlock(&ctx->run_mutex);
	return ret;
}