msi.c 12.1 KB
Newer Older
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
/*
 * Copyright 2006 Jake Moilanen <moilanen@austin.ibm.com>, IBM Corp.
 * Copyright 2006-2007 Michael Ellerman, IBM Corp.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; version 2 of the
 * License.
 *
 */

#include <linux/device.h>
#include <linux/irq.h>
#include <linux/msi.h>

#include <asm/rtas.h>
#include <asm/hw_irq.h>
#include <asm/ppc-pci.h>

static int query_token, change_token;

#define RTAS_QUERY_FN		0
#define RTAS_CHANGE_FN		1
#define RTAS_RESET_FN		2
#define RTAS_CHANGE_MSI_FN	3
#define RTAS_CHANGE_MSIX_FN	4
27
#define RTAS_CHANGE_32MSI_FN	5
28
29
30
31
32
33
34
35
36
37
38
39
40
41

/* RTAS Helpers */

static int rtas_change_msi(struct pci_dn *pdn, u32 func, u32 num_irqs)
{
	u32 addr, seq_num, rtas_ret[3];
	unsigned long buid;
	int rc;

	addr = rtas_config_addr(pdn->busno, pdn->devfn, 0);
	buid = pdn->phb->buid;

	seq_num = 1;
	do {
42
43
		if (func == RTAS_CHANGE_MSI_FN || func == RTAS_CHANGE_MSIX_FN ||
		    func == RTAS_CHANGE_32MSI_FN)
44
45
46
47
48
49
50
51
52
53
54
			rc = rtas_call(change_token, 6, 4, rtas_ret, addr,
					BUID_HI(buid), BUID_LO(buid),
					func, num_irqs, seq_num);
		else
			rc = rtas_call(change_token, 6, 3, rtas_ret, addr,
					BUID_HI(buid), BUID_LO(buid),
					func, num_irqs, seq_num);

		seq_num = rtas_ret[1];
	} while (rtas_busy_delay(rc));

55
	/*
56
57
	 * If the RTAS call succeeded, return the number of irqs allocated.
	 * If not, make sure we return a negative error code.
58
	 */
59
60
61
62
	if (rc == 0)
		rc = rtas_ret[0];
	else if (rc > 0)
		rc = -rc;
63

64
65
	pr_debug("rtas_msi: ibm,change_msi(func=%d,num=%d), got %d rc = %d\n",
		 func, num_irqs, rtas_ret[0], rc);
66
67
68
69
70
71
72
73

	return rc;
}

static void rtas_disable_msi(struct pci_dev *pdev)
{
	struct pci_dn *pdn;

74
	pdn = pci_get_pdn(pdev);
75
76
77
	if (!pdn)
		return;

78
79
80
81
82
83
84
85
86
87
88
89
	/*
	 * disabling MSI with the explicit interface also disables MSI-X
	 */
	if (rtas_change_msi(pdn, RTAS_CHANGE_MSI_FN, 0) != 0) {
		/* 
		 * may have failed because explicit interface is not
		 * present
		 */
		if (rtas_change_msi(pdn, RTAS_CHANGE_FN, 0) != 0) {
			pr_debug("rtas_msi: Setting MSIs to 0 failed!\n");
		}
	}
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
}

static int rtas_query_irq_number(struct pci_dn *pdn, int offset)
{
	u32 addr, rtas_ret[2];
	unsigned long buid;
	int rc;

	addr = rtas_config_addr(pdn->busno, pdn->devfn, 0);
	buid = pdn->phb->buid;

	do {
		rc = rtas_call(query_token, 4, 3, rtas_ret, addr,
			       BUID_HI(buid), BUID_LO(buid), offset);
	} while (rtas_busy_delay(rc));

	if (rc) {
		pr_debug("rtas_msi: error (%d) querying source number\n", rc);
		return rc;
	}

	return rtas_ret[0];
}

static void rtas_teardown_msi_irqs(struct pci_dev *pdev)
{
	struct msi_desc *entry;

	list_for_each_entry(entry, &pdev->msi_list, list) {
		if (entry->irq == NO_IRQ)
			continue;

122
		irq_set_msi_desc(entry->irq, NULL);
123
124
125
126
127
128
		irq_dispose_mapping(entry->irq);
	}

	rtas_disable_msi(pdev);
}

129
static int check_req(struct pci_dev *pdev, int nvec, char *prop_name)
130
131
132
{
	struct device_node *dn;
	struct pci_dn *pdn;
133
134
	const __be32 *p;
	u32 req_msi;
135

136
	pdn = pci_get_pdn(pdev);
137
138
139
140
141
	if (!pdn)
		return -ENODEV;

	dn = pdn->node;

142
143
	p = of_get_property(dn, prop_name, NULL);
	if (!p) {
144
		pr_debug("rtas_msi: No %s on %s\n", prop_name, dn->full_name);
145
146
147
		return -ENOENT;
	}

148
149
	req_msi = be32_to_cpup(p);
	if (req_msi < nvec) {
150
		pr_debug("rtas_msi: %s requests < %d MSIs\n", prop_name, nvec);
151

152
		if (req_msi == 0) /* Be paranoid */
153
154
			return -ENOSPC;

155
		return req_msi;
156
157
158
159
160
	}

	return 0;
}

161
162
163
164
165
166
167
168
169
170
static int check_req_msi(struct pci_dev *pdev, int nvec)
{
	return check_req(pdev, nvec, "ibm,req#msi");
}

static int check_req_msix(struct pci_dev *pdev, int nvec)
{
	return check_req(pdev, nvec, "ibm,req#msi-x");
}

171
172
173
174
175
/* Quota calculation */

static struct device_node *find_pe_total_msi(struct pci_dev *dev, int *total)
{
	struct device_node *dn;
176
	const __be32 *p;
177
178
179
180
181
182
183

	dn = of_node_get(pci_device_to_OF_node(dev));
	while (dn) {
		p = of_get_property(dn, "ibm,pe-total-#msi", NULL);
		if (p) {
			pr_debug("rtas_msi: found prop on dn %s\n",
				dn->full_name);
184
			*total = be32_to_cpup(p);
185
186
187
188
189
190
191
192
193
194
195
196
			return dn;
		}

		dn = of_get_next_parent(dn);
	}

	return NULL;
}

static struct device_node *find_pe_dn(struct pci_dev *dev, int *total)
{
	struct device_node *dn;
197
	struct eeh_dev *edev;
198
199
200
201
202
203
204

	/* Found our PE and assume 8 at that point. */

	dn = pci_device_to_OF_node(dev);
	if (!dn)
		return NULL;

205
206
	/* Get the top level device in the PE */
	edev = of_node_to_eeh_dev(dn);
207
208
	if (edev->pe)
		edev = list_first_entry(&edev->pe->edevs, struct eeh_dev, list);
209
	dn = eeh_dev_to_of_node(edev);
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
	if (!dn)
		return NULL;

	/* We actually want the parent */
	dn = of_get_parent(dn);
	if (!dn)
		return NULL;

	/* Hardcode of 8 for old firmwares */
	*total = 8;
	pr_debug("rtas_msi: using PE dn %s\n", dn->full_name);

	return dn;
}

struct msi_counts {
	struct device_node *requestor;
	int num_devices;
	int request;
	int quota;
	int spare;
	int over_quota;
};

static void *count_non_bridge_devices(struct device_node *dn, void *data)
{
	struct msi_counts *counts = data;
237
	const __be32 *p;
238
239
240
241
242
	u32 class;

	pr_debug("rtas_msi: counting %s\n", dn->full_name);

	p = of_get_property(dn, "class-code", NULL);
243
	class = p ? be32_to_cpup(p) : 0;
244
245
246
247
248
249
250
251
252
253

	if ((class >> 8) != PCI_CLASS_BRIDGE_PCI)
		counts->num_devices++;

	return NULL;
}

static void *count_spare_msis(struct device_node *dn, void *data)
{
	struct msi_counts *counts = data;
254
	const __be32 *p;
255
256
257
258
259
260
261
262
263
264
	int req;

	if (dn == counts->requestor)
		req = counts->request;
	else {
		/* We don't know if a driver will try to use MSI or MSI-X,
		 * so we just have to punt and use the larger of the two. */
		req = 0;
		p = of_get_property(dn, "ibm,req#msi", NULL);
		if (p)
265
			req = be32_to_cpup(p);
266
267
268

		p = of_get_property(dn, "ibm,req#msi-x", NULL);
		if (p)
269
			req = max(req, (int)be32_to_cpup(p));
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
	}

	if (req < counts->quota)
		counts->spare += counts->quota - req;
	else if (req > counts->quota)
		counts->over_quota++;

	return NULL;
}

static int msi_quota_for_device(struct pci_dev *dev, int request)
{
	struct device_node *pe_dn;
	struct msi_counts counts;
	int total;

	pr_debug("rtas_msi: calc quota for %s, request %d\n", pci_name(dev),
		  request);

	pe_dn = find_pe_total_msi(dev, &total);
	if (!pe_dn)
		pe_dn = find_pe_dn(dev, &total);

	if (!pe_dn) {
		pr_err("rtas_msi: couldn't find PE for %s\n", pci_name(dev));
		goto out;
	}

	pr_debug("rtas_msi: found PE %s\n", pe_dn->full_name);

	memset(&counts, 0, sizeof(struct msi_counts));

	/* Work out how many devices we have below this PE */
	traverse_pci_devices(pe_dn, count_non_bridge_devices, &counts);

	if (counts.num_devices == 0) {
		pr_err("rtas_msi: found 0 devices under PE for %s\n",
			pci_name(dev));
		goto out;
	}

	counts.quota = total / counts.num_devices;
	if (request <= counts.quota)
		goto out;

	/* else, we have some more calculating to do */
	counts.requestor = pci_device_to_OF_node(dev);
	counts.request = request;
	traverse_pci_devices(pe_dn, count_spare_msis, &counts);

	/* If the quota isn't an integer multiple of the total, we can
	 * use the remainder as spare MSIs for anyone that wants them. */
	counts.spare += total % counts.num_devices;

	/* Divide any spare by the number of over-quota requestors */
	if (counts.over_quota)
		counts.quota += counts.spare / counts.over_quota;

	/* And finally clamp the request to the possibly adjusted quota */
	request = min(counts.quota, request);

	pr_debug("rtas_msi: request clamped to quota %d\n", request);
out:
	of_node_put(pe_dn);

	return request;
}

338
339
static int rtas_msi_check_device(struct pci_dev *pdev, int nvec, int type)
{
340
341
	int quota, rc;

342
	if (type == PCI_CAP_ID_MSIX)
343
344
345
346
347
348
		rc = check_req_msix(pdev, nvec);
	else
		rc = check_req_msi(pdev, nvec);

	if (rc)
		return rc;
349

350
351
352
353
354
355
	quota = msi_quota_for_device(pdev, nvec);

	if (quota && quota < nvec)
		return quota;

	return 0;
356
357
}

358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
static int check_msix_entries(struct pci_dev *pdev)
{
	struct msi_desc *entry;
	int expected;

	/* There's no way for us to express to firmware that we want
	 * a discontiguous, or non-zero based, range of MSI-X entries.
	 * So we must reject such requests. */

	expected = 0;
	list_for_each_entry(entry, &pdev->msi_list, list) {
		if (entry->msi_attrib.entry_nr != expected) {
			pr_debug("rtas_msi: bad MSI-X entries.\n");
			return -EINVAL;
		}
		expected++;
	}

	return 0;
}

379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
static void rtas_hack_32bit_msi_gen2(struct pci_dev *pdev)
{
	u32 addr_hi, addr_lo;

	/*
	 * We should only get in here for IODA1 configs. This is based on the
	 * fact that we using RTAS for MSIs, we don't have the 32 bit MSI RTAS
	 * support, and we are in a PCIe Gen2 slot.
	 */
	dev_info(&pdev->dev,
		 "rtas_msi: No 32 bit MSI firmware support, forcing 32 bit MSI\n");
	pci_read_config_dword(pdev, pdev->msi_cap + PCI_MSI_ADDRESS_HI, &addr_hi);
	addr_lo = 0xffff0000 | ((addr_hi >> (48 - 32)) << 4);
	pci_write_config_dword(pdev, pdev->msi_cap + PCI_MSI_ADDRESS_LO, addr_lo);
	pci_write_config_dword(pdev, pdev->msi_cap + PCI_MSI_ADDRESS_HI, 0);
}

396
static int rtas_setup_msi_irqs(struct pci_dev *pdev, int nvec_in, int type)
397
398
399
400
{
	struct pci_dn *pdn;
	int hwirq, virq, i, rc;
	struct msi_desc *entry;
401
	struct msi_msg msg;
402
	int nvec = nvec_in;
403
	int use_32bit_msi_hack = 0;
404

405
	pdn = pci_get_pdn(pdev);
406
407
408
	if (!pdn)
		return -ENODEV;

409
410
411
	if (type == PCI_CAP_ID_MSIX && check_msix_entries(pdev))
		return -EINVAL;

412
413
414
415
416
417
418
419
420
421
422
423
	/*
	 * Firmware currently refuse any non power of two allocation
	 * so we round up if the quota will allow it.
	 */
	if (type == PCI_CAP_ID_MSIX) {
		int m = roundup_pow_of_two(nvec);
		int quota = msi_quota_for_device(pdev, m);

		if (quota >= m)
			nvec = m;
	}

424
425
426
427
428
	/*
	 * Try the new more explicit firmware interface, if that fails fall
	 * back to the old interface. The old interface is known to never
	 * return MSI-Xs.
	 */
429
again:
430
	if (type == PCI_CAP_ID_MSI) {
431
		if (pdn->force_32bit_msi) {
432
			rc = rtas_change_msi(pdn, RTAS_CHANGE_32MSI_FN, nvec);
433
434
435
436
437
438
439
440
441
442
443
444
445
446
			if (rc < 0) {
				/*
				 * We only want to run the 32 bit MSI hack below if
				 * the max bus speed is Gen2 speed
				 */
				if (pdev->bus->max_bus_speed != PCIE_SPEED_5_0GT)
					return rc;

				use_32bit_msi_hack = 1;
			}
		} else
			rc = -1;

		if (rc < 0)
447
			rc = rtas_change_msi(pdn, RTAS_CHANGE_MSI_FN, nvec);
448

449
		if (rc < 0) {
450
451
452
			pr_debug("rtas_msi: trying the old firmware call.\n");
			rc = rtas_change_msi(pdn, RTAS_CHANGE_FN, nvec);
		}
453
454
455

		if (use_32bit_msi_hack && rc > 0)
			rtas_hack_32bit_msi_gen2(pdev);
456
457
458
	} else
		rc = rtas_change_msi(pdn, RTAS_CHANGE_MSIX_FN, nvec);

459
	if (rc != nvec) {
460
461
462
463
		if (nvec != nvec_in) {
			nvec = nvec_in;
			goto again;
		}
464
		pr_debug("rtas_msi: rtas_change_msi() failed\n");
465
		return rc;
466
467
468
469
	}

	i = 0;
	list_for_each_entry(entry, &pdev->msi_list, list) {
470
		hwirq = rtas_query_irq_number(pdn, i++);
471
472
		if (hwirq < 0) {
			pr_debug("rtas_msi: error (%d) getting hwirq\n", rc);
473
			return hwirq;
474
475
476
477
478
479
		}

		virq = irq_create_mapping(NULL, hwirq);

		if (virq == NO_IRQ) {
			pr_debug("rtas_msi: Failed mapping hwirq %d\n", hwirq);
480
			return -ENOSPC;
481
482
483
		}

		dev_dbg(&pdev->dev, "rtas_msi: allocated virq %d\n", virq);
484
		irq_set_msi_desc(virq, entry);
485
486
487
488

		/* Read config space back so we can restore after reset */
		read_msi_msg(virq, &msg);
		entry->msg = msg;
489
490
491
492
493
494
495
496
497
498
499
500
501
502
	}

	return 0;
}

static void rtas_msi_pci_irq_fixup(struct pci_dev *pdev)
{
	/* No LSI -> leave MSIs (if any) configured */
	if (pdev->irq == NO_IRQ) {
		dev_dbg(&pdev->dev, "rtas_msi: no LSI, nothing to do.\n");
		return;
	}

	/* No MSI -> MSIs can't have been assigned by fw, leave LSI */
503
504
	if (check_req_msi(pdev, 1) && check_req_msix(pdev, 1)) {
		dev_dbg(&pdev->dev, "rtas_msi: no req#msi/x, nothing to do.\n");
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
		return;
	}

	dev_dbg(&pdev->dev, "rtas_msi: disabling existing MSI.\n");
	rtas_disable_msi(pdev);
}

static int rtas_msi_init(void)
{
	query_token  = rtas_token("ibm,query-interrupt-source-number");
	change_token = rtas_token("ibm,change-msi");

	if ((query_token == RTAS_UNKNOWN_SERVICE) ||
			(change_token == RTAS_UNKNOWN_SERVICE)) {
		pr_debug("rtas_msi: no RTAS tokens, no MSI support.\n");
		return -1;
	}

	pr_debug("rtas_msi: Registering RTAS MSI callbacks.\n");

	WARN_ON(ppc_md.setup_msi_irqs);
	ppc_md.setup_msi_irqs = rtas_setup_msi_irqs;
	ppc_md.teardown_msi_irqs = rtas_teardown_msi_irqs;
	ppc_md.msi_check_device = rtas_msi_check_device;

	WARN_ON(ppc_md.pci_irq_fixup);
	ppc_md.pci_irq_fixup = rtas_msi_pci_irq_fixup;

	return 0;
}
arch_initcall(rtas_msi_init);
536