msi.c 11.9 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
/*
 * 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>
19
#include <asm/machdep.h>
20
21
22
23
24
25
26
27

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
28
#define RTAS_CHANGE_32MSI_FN	5
29
30
31
32
33
34
35
36
37
38
39
40
41
42

/* 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 {
43
44
		if (func == RTAS_CHANGE_MSI_FN || func == RTAS_CHANGE_MSIX_FN ||
		    func == RTAS_CHANGE_32MSI_FN)
45
46
47
48
49
50
51
52
53
54
55
			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));

56
	/*
57
58
	 * If the RTAS call succeeded, return the number of irqs allocated.
	 * If not, make sure we return a negative error code.
59
	 */
60
61
62
63
	if (rc == 0)
		rc = rtas_ret[0];
	else if (rc > 0)
		rc = -rc;
64

65
66
	pr_debug("rtas_msi: ibm,change_msi(func=%d,num=%d), got %d rc = %d\n",
		 func, num_irqs, rtas_ret[0], rc);
67
68
69
70
71
72
73
74

	return rc;
}

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

75
	pdn = pci_get_pdn(pdev);
76
77
78
	if (!pdn)
		return;

79
80
81
82
83
84
85
86
87
88
89
90
	/*
	 * 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");
		}
	}
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
}

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;

123
		irq_set_msi_desc(entry->irq, NULL);
124
125
126
127
128
129
		irq_dispose_mapping(entry->irq);
	}

	rtas_disable_msi(pdev);
}

130
static int check_req(struct pci_dev *pdev, int nvec, char *prop_name)
131
132
133
{
	struct device_node *dn;
	struct pci_dn *pdn;
134
135
	const __be32 *p;
	u32 req_msi;
136

137
	pdn = pci_get_pdn(pdev);
138
139
140
141
142
	if (!pdn)
		return -ENODEV;

	dn = pdn->node;

143
144
	p = of_get_property(dn, prop_name, NULL);
	if (!p) {
145
		pr_debug("rtas_msi: No %s on %s\n", prop_name, dn->full_name);
146
147
148
		return -ENOENT;
	}

149
150
	req_msi = be32_to_cpup(p);
	if (req_msi < nvec) {
151
		pr_debug("rtas_msi: %s requests < %d MSIs\n", prop_name, nvec);
152

153
		if (req_msi == 0) /* Be paranoid */
154
155
			return -ENOSPC;

156
		return req_msi;
157
158
159
160
161
	}

	return 0;
}

162
163
164
165
166
167
168
169
170
171
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");
}

172
173
174
175
176
/* Quota calculation */

static struct device_node *find_pe_total_msi(struct pci_dev *dev, int *total)
{
	struct device_node *dn;
177
	const __be32 *p;
178
179
180
181
182
183
184

	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);
185
			*total = be32_to_cpup(p);
186
187
188
189
190
191
192
193
194
195
196
197
			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;
198
	struct eeh_dev *edev;
199
200
201
202
203
204
205

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

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

206
207
	/* Get the top level device in the PE */
	edev = of_node_to_eeh_dev(dn);
208
209
	if (edev->pe)
		edev = list_first_entry(&edev->pe->edevs, struct eeh_dev, list);
210
	dn = eeh_dev_to_of_node(edev);
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
	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;
238
	const __be32 *p;
239
240
241
242
243
	u32 class;

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

	p = of_get_property(dn, "class-code", NULL);
244
	class = p ? be32_to_cpup(p) : 0;
245
246
247
248
249
250
251
252
253
254

	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;
255
	const __be32 *p;
256
257
258
259
260
261
262
263
264
265
	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)
266
			req = be32_to_cpup(p);
267
268
269

		p = of_get_property(dn, "ibm,req#msi-x", NULL);
		if (p)
270
			req = max(req, (int)be32_to_cpup(p));
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
	}

	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;
}

339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
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;
}

360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
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);
}

377
static int rtas_setup_msi_irqs(struct pci_dev *pdev, int nvec_in, int type)
378
379
{
	struct pci_dn *pdn;
380
	int hwirq, virq, i, quota, rc;
381
	struct msi_desc *entry;
382
	struct msi_msg msg;
383
	int nvec = nvec_in;
384
	int use_32bit_msi_hack = 0;
385

386
387
388
389
390
391
392
393
394
395
396
397
	if (type == PCI_CAP_ID_MSIX)
		rc = check_req_msix(pdev, nvec);
	else
		rc = check_req_msi(pdev, nvec);

	if (rc)
		return rc;

	quota = msi_quota_for_device(pdev, nvec);

	if (quota && quota < nvec)
		return quota;
398

399
400
401
	if (type == PCI_CAP_ID_MSIX && check_msix_entries(pdev))
		return -EINVAL;

402
403
404
405
406
407
	/*
	 * 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);
408
		quota = msi_quota_for_device(pdev, m);
409
410
411
412
413

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

414
415
	pdn = pci_get_pdn(pdev);

416
417
418
419
420
	/*
	 * 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.
	 */
421
again:
422
	if (type == PCI_CAP_ID_MSI) {
423
		if (pdn->force_32bit_msi) {
424
			rc = rtas_change_msi(pdn, RTAS_CHANGE_32MSI_FN, nvec);
425
426
427
428
429
430
431
432
433
434
435
436
437
438
			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)
439
			rc = rtas_change_msi(pdn, RTAS_CHANGE_MSI_FN, nvec);
440

441
		if (rc < 0) {
442
443
444
			pr_debug("rtas_msi: trying the old firmware call.\n");
			rc = rtas_change_msi(pdn, RTAS_CHANGE_FN, nvec);
		}
445
446
447

		if (use_32bit_msi_hack && rc > 0)
			rtas_hack_32bit_msi_gen2(pdev);
448
449
450
	} else
		rc = rtas_change_msi(pdn, RTAS_CHANGE_MSIX_FN, nvec);

451
	if (rc != nvec) {
452
453
454
455
		if (nvec != nvec_in) {
			nvec = nvec_in;
			goto again;
		}
456
		pr_debug("rtas_msi: rtas_change_msi() failed\n");
457
		return rc;
458
459
460
461
	}

	i = 0;
	list_for_each_entry(entry, &pdev->msi_list, list) {
462
		hwirq = rtas_query_irq_number(pdn, i++);
463
464
		if (hwirq < 0) {
			pr_debug("rtas_msi: error (%d) getting hwirq\n", rc);
465
			return hwirq;
466
467
468
469
470
471
		}

		virq = irq_create_mapping(NULL, hwirq);

		if (virq == NO_IRQ) {
			pr_debug("rtas_msi: Failed mapping hwirq %d\n", hwirq);
472
			return -ENOSPC;
473
474
475
		}

		dev_dbg(&pdev->dev, "rtas_msi: allocated virq %d\n", virq);
476
		irq_set_msi_desc(virq, entry);
477
478
479
480

		/* Read config space back so we can restore after reset */
		read_msi_msg(virq, &msg);
		entry->msg = msg;
481
482
483
484
485
486
487
488
489
490
491
492
493
494
	}

	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 */
495
496
	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");
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
		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;

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

	return 0;
}
526
machine_arch_initcall(pseries, rtas_msi_init);