xref: /linux/arch/s390/pci/pci.c (revision 9e56ff53b4115875667760445b028357848b4748)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright IBM Corp. 2012
4  *
5  * Author(s):
6  *   Jan Glauber <jang@linux.vnet.ibm.com>
7  *
8  * The System z PCI code is a rewrite from a prototype by
9  * the following people (Kudoz!):
10  *   Alexander Schmidt
11  *   Christoph Raisch
12  *   Hannes Hering
13  *   Hoang-Nam Nguyen
14  *   Jan-Bernd Themann
15  *   Stefan Roscher
16  *   Thomas Klein
17  */
18 
19 #define KMSG_COMPONENT "zpci"
20 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
21 
22 #include <linux/kernel.h>
23 #include <linux/slab.h>
24 #include <linux/err.h>
25 #include <linux/export.h>
26 #include <linux/delay.h>
27 #include <linux/seq_file.h>
28 #include <linux/jump_label.h>
29 #include <linux/pci.h>
30 #include <linux/printk.h>
31 
32 #include <asm/isc.h>
33 #include <asm/airq.h>
34 #include <asm/facility.h>
35 #include <asm/pci_insn.h>
36 #include <asm/pci_clp.h>
37 #include <asm/pci_dma.h>
38 
39 #include "pci_bus.h"
40 #include "pci_iov.h"
41 
42 /* list of all detected zpci devices */
43 static LIST_HEAD(zpci_list);
44 static DEFINE_SPINLOCK(zpci_list_lock);
45 
46 static DECLARE_BITMAP(zpci_domain, ZPCI_DOMAIN_BITMAP_SIZE);
47 static DEFINE_SPINLOCK(zpci_domain_lock);
48 
49 #define ZPCI_IOMAP_ENTRIES						\
50 	min(((unsigned long) ZPCI_NR_DEVICES * PCI_STD_NUM_BARS / 2),	\
51 	    ZPCI_IOMAP_MAX_ENTRIES)
52 
53 unsigned int s390_pci_no_rid;
54 
55 static DEFINE_SPINLOCK(zpci_iomap_lock);
56 static unsigned long *zpci_iomap_bitmap;
57 struct zpci_iomap_entry *zpci_iomap_start;
58 EXPORT_SYMBOL_GPL(zpci_iomap_start);
59 
60 DEFINE_STATIC_KEY_FALSE(have_mio);
61 
62 static struct kmem_cache *zdev_fmb_cache;
63 
64 /* AEN structures that must be preserved over KVM module re-insertion */
65 union zpci_sic_iib *zpci_aipb;
66 EXPORT_SYMBOL_GPL(zpci_aipb);
67 struct airq_iv *zpci_aif_sbv;
68 EXPORT_SYMBOL_GPL(zpci_aif_sbv);
69 
70 struct zpci_dev *get_zdev_by_fid(u32 fid)
71 {
72 	struct zpci_dev *tmp, *zdev = NULL;
73 
74 	spin_lock(&zpci_list_lock);
75 	list_for_each_entry(tmp, &zpci_list, entry) {
76 		if (tmp->fid == fid) {
77 			zdev = tmp;
78 			zpci_zdev_get(zdev);
79 			break;
80 		}
81 	}
82 	spin_unlock(&zpci_list_lock);
83 	return zdev;
84 }
85 
86 void zpci_remove_reserved_devices(void)
87 {
88 	struct zpci_dev *tmp, *zdev;
89 	enum zpci_state state;
90 	LIST_HEAD(remove);
91 
92 	spin_lock(&zpci_list_lock);
93 	list_for_each_entry_safe(zdev, tmp, &zpci_list, entry) {
94 		if (zdev->state == ZPCI_FN_STATE_STANDBY &&
95 		    !clp_get_state(zdev->fid, &state) &&
96 		    state == ZPCI_FN_STATE_RESERVED)
97 			list_move_tail(&zdev->entry, &remove);
98 	}
99 	spin_unlock(&zpci_list_lock);
100 
101 	list_for_each_entry_safe(zdev, tmp, &remove, entry)
102 		zpci_device_reserved(zdev);
103 }
104 
105 int pci_domain_nr(struct pci_bus *bus)
106 {
107 	return ((struct zpci_bus *) bus->sysdata)->domain_nr;
108 }
109 EXPORT_SYMBOL_GPL(pci_domain_nr);
110 
111 int pci_proc_domain(struct pci_bus *bus)
112 {
113 	return pci_domain_nr(bus);
114 }
115 EXPORT_SYMBOL_GPL(pci_proc_domain);
116 
117 /* Modify PCI: Register I/O address translation parameters */
118 int zpci_register_ioat(struct zpci_dev *zdev, u8 dmaas,
119 		       u64 base, u64 limit, u64 iota, u8 *status)
120 {
121 	u64 req = ZPCI_CREATE_REQ(zdev->fh, dmaas, ZPCI_MOD_FC_REG_IOAT);
122 	struct zpci_fib fib = {0};
123 	u8 cc;
124 
125 	WARN_ON_ONCE(iota & 0x3fff);
126 	fib.pba = base;
127 	/* Work around off by one in ISM virt device */
128 	if (zdev->pft == PCI_FUNC_TYPE_ISM && limit > base)
129 		fib.pal = limit + (1 << 12);
130 	else
131 		fib.pal = limit;
132 	fib.iota = iota | ZPCI_IOTA_RTTO_FLAG;
133 	fib.gd = zdev->gisa;
134 	cc = zpci_mod_fc(req, &fib, status);
135 	if (cc)
136 		zpci_dbg(3, "reg ioat fid:%x, cc:%d, status:%d\n", zdev->fid, cc, *status);
137 	return cc;
138 }
139 EXPORT_SYMBOL_GPL(zpci_register_ioat);
140 
141 /* Modify PCI: Unregister I/O address translation parameters */
142 int zpci_unregister_ioat(struct zpci_dev *zdev, u8 dmaas)
143 {
144 	u64 req = ZPCI_CREATE_REQ(zdev->fh, dmaas, ZPCI_MOD_FC_DEREG_IOAT);
145 	struct zpci_fib fib = {0};
146 	u8 cc, status;
147 
148 	fib.gd = zdev->gisa;
149 
150 	cc = zpci_mod_fc(req, &fib, &status);
151 	if (cc)
152 		zpci_dbg(3, "unreg ioat fid:%x, cc:%d, status:%d\n", zdev->fid, cc, status);
153 	return cc;
154 }
155 
156 /* Modify PCI: Set PCI function measurement parameters */
157 int zpci_fmb_enable_device(struct zpci_dev *zdev)
158 {
159 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_SET_MEASURE);
160 	struct zpci_iommu_ctrs *ctrs;
161 	struct zpci_fib fib = {0};
162 	u8 cc, status;
163 
164 	if (zdev->fmb || sizeof(*zdev->fmb) < zdev->fmb_length)
165 		return -EINVAL;
166 
167 	zdev->fmb = kmem_cache_zalloc(zdev_fmb_cache, GFP_KERNEL);
168 	if (!zdev->fmb)
169 		return -ENOMEM;
170 	WARN_ON((u64) zdev->fmb & 0xf);
171 
172 	/* reset software counters */
173 	ctrs = zpci_get_iommu_ctrs(zdev);
174 	if (ctrs) {
175 		atomic64_set(&ctrs->mapped_pages, 0);
176 		atomic64_set(&ctrs->unmapped_pages, 0);
177 		atomic64_set(&ctrs->global_rpcits, 0);
178 		atomic64_set(&ctrs->sync_map_rpcits, 0);
179 		atomic64_set(&ctrs->sync_rpcits, 0);
180 	}
181 
182 
183 	fib.fmb_addr = virt_to_phys(zdev->fmb);
184 	fib.gd = zdev->gisa;
185 	cc = zpci_mod_fc(req, &fib, &status);
186 	if (cc) {
187 		kmem_cache_free(zdev_fmb_cache, zdev->fmb);
188 		zdev->fmb = NULL;
189 	}
190 	return cc ? -EIO : 0;
191 }
192 
193 /* Modify PCI: Disable PCI function measurement */
194 int zpci_fmb_disable_device(struct zpci_dev *zdev)
195 {
196 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_SET_MEASURE);
197 	struct zpci_fib fib = {0};
198 	u8 cc, status;
199 
200 	if (!zdev->fmb)
201 		return -EINVAL;
202 
203 	fib.gd = zdev->gisa;
204 
205 	/* Function measurement is disabled if fmb address is zero */
206 	cc = zpci_mod_fc(req, &fib, &status);
207 	if (cc == 3) /* Function already gone. */
208 		cc = 0;
209 
210 	if (!cc) {
211 		kmem_cache_free(zdev_fmb_cache, zdev->fmb);
212 		zdev->fmb = NULL;
213 	}
214 	return cc ? -EIO : 0;
215 }
216 
217 static int zpci_cfg_load(struct zpci_dev *zdev, int offset, u32 *val, u8 len)
218 {
219 	u64 req = ZPCI_CREATE_REQ(zdev->fh, ZPCI_PCIAS_CFGSPC, len);
220 	u64 data;
221 	int rc;
222 
223 	rc = __zpci_load(&data, req, offset);
224 	if (!rc) {
225 		data = le64_to_cpu((__force __le64) data);
226 		data >>= (8 - len) * 8;
227 		*val = (u32) data;
228 	} else
229 		*val = 0xffffffff;
230 	return rc;
231 }
232 
233 static int zpci_cfg_store(struct zpci_dev *zdev, int offset, u32 val, u8 len)
234 {
235 	u64 req = ZPCI_CREATE_REQ(zdev->fh, ZPCI_PCIAS_CFGSPC, len);
236 	u64 data = val;
237 	int rc;
238 
239 	data <<= (8 - len) * 8;
240 	data = (__force u64) cpu_to_le64(data);
241 	rc = __zpci_store(data, req, offset);
242 	return rc;
243 }
244 
245 resource_size_t pcibios_align_resource(void *data, const struct resource *res,
246 				       resource_size_t size,
247 				       resource_size_t align)
248 {
249 	return 0;
250 }
251 
252 /* combine single writes by using store-block insn */
253 void __iowrite64_copy(void __iomem *to, const void *from, size_t count)
254 {
255        zpci_memcpy_toio(to, from, count);
256 }
257 
258 void __iomem *ioremap_prot(phys_addr_t phys_addr, size_t size,
259 			   unsigned long prot)
260 {
261 	/*
262 	 * When PCI MIO instructions are unavailable the "physical" address
263 	 * encodes a hint for accessing the PCI memory space it represents.
264 	 * Just pass it unchanged such that ioread/iowrite can decode it.
265 	 */
266 	if (!static_branch_unlikely(&have_mio))
267 		return (void __iomem *)phys_addr;
268 
269 	return generic_ioremap_prot(phys_addr, size, __pgprot(prot));
270 }
271 EXPORT_SYMBOL(ioremap_prot);
272 
273 void iounmap(volatile void __iomem *addr)
274 {
275 	if (static_branch_likely(&have_mio))
276 		generic_iounmap(addr);
277 }
278 EXPORT_SYMBOL(iounmap);
279 
280 /* Create a virtual mapping cookie for a PCI BAR */
281 static void __iomem *pci_iomap_range_fh(struct pci_dev *pdev, int bar,
282 					unsigned long offset, unsigned long max)
283 {
284 	struct zpci_dev *zdev =	to_zpci(pdev);
285 	int idx;
286 
287 	idx = zdev->bars[bar].map_idx;
288 	spin_lock(&zpci_iomap_lock);
289 	/* Detect overrun */
290 	WARN_ON(!++zpci_iomap_start[idx].count);
291 	zpci_iomap_start[idx].fh = zdev->fh;
292 	zpci_iomap_start[idx].bar = bar;
293 	spin_unlock(&zpci_iomap_lock);
294 
295 	return (void __iomem *) ZPCI_ADDR(idx) + offset;
296 }
297 
298 static void __iomem *pci_iomap_range_mio(struct pci_dev *pdev, int bar,
299 					 unsigned long offset,
300 					 unsigned long max)
301 {
302 	unsigned long barsize = pci_resource_len(pdev, bar);
303 	struct zpci_dev *zdev = to_zpci(pdev);
304 	void __iomem *iova;
305 
306 	iova = ioremap((unsigned long) zdev->bars[bar].mio_wt, barsize);
307 	return iova ? iova + offset : iova;
308 }
309 
310 void __iomem *pci_iomap_range(struct pci_dev *pdev, int bar,
311 			      unsigned long offset, unsigned long max)
312 {
313 	if (bar >= PCI_STD_NUM_BARS || !pci_resource_len(pdev, bar))
314 		return NULL;
315 
316 	if (static_branch_likely(&have_mio))
317 		return pci_iomap_range_mio(pdev, bar, offset, max);
318 	else
319 		return pci_iomap_range_fh(pdev, bar, offset, max);
320 }
321 EXPORT_SYMBOL(pci_iomap_range);
322 
323 void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long maxlen)
324 {
325 	return pci_iomap_range(dev, bar, 0, maxlen);
326 }
327 EXPORT_SYMBOL(pci_iomap);
328 
329 static void __iomem *pci_iomap_wc_range_mio(struct pci_dev *pdev, int bar,
330 					    unsigned long offset, unsigned long max)
331 {
332 	unsigned long barsize = pci_resource_len(pdev, bar);
333 	struct zpci_dev *zdev = to_zpci(pdev);
334 	void __iomem *iova;
335 
336 	iova = ioremap((unsigned long) zdev->bars[bar].mio_wb, barsize);
337 	return iova ? iova + offset : iova;
338 }
339 
340 void __iomem *pci_iomap_wc_range(struct pci_dev *pdev, int bar,
341 				 unsigned long offset, unsigned long max)
342 {
343 	if (bar >= PCI_STD_NUM_BARS || !pci_resource_len(pdev, bar))
344 		return NULL;
345 
346 	if (static_branch_likely(&have_mio))
347 		return pci_iomap_wc_range_mio(pdev, bar, offset, max);
348 	else
349 		return pci_iomap_range_fh(pdev, bar, offset, max);
350 }
351 EXPORT_SYMBOL(pci_iomap_wc_range);
352 
353 void __iomem *pci_iomap_wc(struct pci_dev *dev, int bar, unsigned long maxlen)
354 {
355 	return pci_iomap_wc_range(dev, bar, 0, maxlen);
356 }
357 EXPORT_SYMBOL(pci_iomap_wc);
358 
359 static void pci_iounmap_fh(struct pci_dev *pdev, void __iomem *addr)
360 {
361 	unsigned int idx = ZPCI_IDX(addr);
362 
363 	spin_lock(&zpci_iomap_lock);
364 	/* Detect underrun */
365 	WARN_ON(!zpci_iomap_start[idx].count);
366 	if (!--zpci_iomap_start[idx].count) {
367 		zpci_iomap_start[idx].fh = 0;
368 		zpci_iomap_start[idx].bar = 0;
369 	}
370 	spin_unlock(&zpci_iomap_lock);
371 }
372 
373 static void pci_iounmap_mio(struct pci_dev *pdev, void __iomem *addr)
374 {
375 	iounmap(addr);
376 }
377 
378 void pci_iounmap(struct pci_dev *pdev, void __iomem *addr)
379 {
380 	if (static_branch_likely(&have_mio))
381 		pci_iounmap_mio(pdev, addr);
382 	else
383 		pci_iounmap_fh(pdev, addr);
384 }
385 EXPORT_SYMBOL(pci_iounmap);
386 
387 static int pci_read(struct pci_bus *bus, unsigned int devfn, int where,
388 		    int size, u32 *val)
389 {
390 	struct zpci_dev *zdev = zdev_from_bus(bus, devfn);
391 
392 	return (zdev) ? zpci_cfg_load(zdev, where, val, size) : -ENODEV;
393 }
394 
395 static int pci_write(struct pci_bus *bus, unsigned int devfn, int where,
396 		     int size, u32 val)
397 {
398 	struct zpci_dev *zdev = zdev_from_bus(bus, devfn);
399 
400 	return (zdev) ? zpci_cfg_store(zdev, where, val, size) : -ENODEV;
401 }
402 
403 static struct pci_ops pci_root_ops = {
404 	.read = pci_read,
405 	.write = pci_write,
406 };
407 
408 static void zpci_map_resources(struct pci_dev *pdev)
409 {
410 	struct zpci_dev *zdev = to_zpci(pdev);
411 	resource_size_t len;
412 	int i;
413 
414 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
415 		len = pci_resource_len(pdev, i);
416 		if (!len)
417 			continue;
418 
419 		if (zpci_use_mio(zdev))
420 			pdev->resource[i].start =
421 				(resource_size_t __force) zdev->bars[i].mio_wt;
422 		else
423 			pdev->resource[i].start = (resource_size_t __force)
424 				pci_iomap_range_fh(pdev, i, 0, 0);
425 		pdev->resource[i].end = pdev->resource[i].start + len - 1;
426 	}
427 
428 	zpci_iov_map_resources(pdev);
429 }
430 
431 static void zpci_unmap_resources(struct pci_dev *pdev)
432 {
433 	struct zpci_dev *zdev = to_zpci(pdev);
434 	resource_size_t len;
435 	int i;
436 
437 	if (zpci_use_mio(zdev))
438 		return;
439 
440 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
441 		len = pci_resource_len(pdev, i);
442 		if (!len)
443 			continue;
444 		pci_iounmap_fh(pdev, (void __iomem __force *)
445 			       pdev->resource[i].start);
446 	}
447 }
448 
449 static int zpci_alloc_iomap(struct zpci_dev *zdev)
450 {
451 	unsigned long entry;
452 
453 	spin_lock(&zpci_iomap_lock);
454 	entry = find_first_zero_bit(zpci_iomap_bitmap, ZPCI_IOMAP_ENTRIES);
455 	if (entry == ZPCI_IOMAP_ENTRIES) {
456 		spin_unlock(&zpci_iomap_lock);
457 		return -ENOSPC;
458 	}
459 	set_bit(entry, zpci_iomap_bitmap);
460 	spin_unlock(&zpci_iomap_lock);
461 	return entry;
462 }
463 
464 static void zpci_free_iomap(struct zpci_dev *zdev, int entry)
465 {
466 	spin_lock(&zpci_iomap_lock);
467 	memset(&zpci_iomap_start[entry], 0, sizeof(struct zpci_iomap_entry));
468 	clear_bit(entry, zpci_iomap_bitmap);
469 	spin_unlock(&zpci_iomap_lock);
470 }
471 
472 static void zpci_do_update_iomap_fh(struct zpci_dev *zdev, u32 fh)
473 {
474 	int bar, idx;
475 
476 	spin_lock(&zpci_iomap_lock);
477 	for (bar = 0; bar < PCI_STD_NUM_BARS; bar++) {
478 		if (!zdev->bars[bar].size)
479 			continue;
480 		idx = zdev->bars[bar].map_idx;
481 		if (!zpci_iomap_start[idx].count)
482 			continue;
483 		WRITE_ONCE(zpci_iomap_start[idx].fh, zdev->fh);
484 	}
485 	spin_unlock(&zpci_iomap_lock);
486 }
487 
488 void zpci_update_fh(struct zpci_dev *zdev, u32 fh)
489 {
490 	if (!fh || zdev->fh == fh)
491 		return;
492 
493 	zdev->fh = fh;
494 	if (zpci_use_mio(zdev))
495 		return;
496 	if (zdev->has_resources && zdev_enabled(zdev))
497 		zpci_do_update_iomap_fh(zdev, fh);
498 }
499 
500 static struct resource *__alloc_res(struct zpci_dev *zdev, unsigned long start,
501 				    unsigned long size, unsigned long flags)
502 {
503 	struct resource *r;
504 
505 	r = kzalloc(sizeof(*r), GFP_KERNEL);
506 	if (!r)
507 		return NULL;
508 
509 	r->start = start;
510 	r->end = r->start + size - 1;
511 	r->flags = flags;
512 	r->name = zdev->res_name;
513 
514 	if (request_resource(&iomem_resource, r)) {
515 		kfree(r);
516 		return NULL;
517 	}
518 	return r;
519 }
520 
521 int zpci_setup_bus_resources(struct zpci_dev *zdev)
522 {
523 	unsigned long addr, size, flags;
524 	struct resource *res;
525 	int i, entry;
526 
527 	snprintf(zdev->res_name, sizeof(zdev->res_name),
528 		 "PCI Bus %04x:%02x", zdev->uid, ZPCI_BUS_NR);
529 
530 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
531 		if (!zdev->bars[i].size)
532 			continue;
533 		entry = zpci_alloc_iomap(zdev);
534 		if (entry < 0)
535 			return entry;
536 		zdev->bars[i].map_idx = entry;
537 
538 		/* only MMIO is supported */
539 		flags = IORESOURCE_MEM;
540 		if (zdev->bars[i].val & 8)
541 			flags |= IORESOURCE_PREFETCH;
542 		if (zdev->bars[i].val & 4)
543 			flags |= IORESOURCE_MEM_64;
544 
545 		if (zpci_use_mio(zdev))
546 			addr = (unsigned long) zdev->bars[i].mio_wt;
547 		else
548 			addr = ZPCI_ADDR(entry);
549 		size = 1UL << zdev->bars[i].size;
550 
551 		res = __alloc_res(zdev, addr, size, flags);
552 		if (!res) {
553 			zpci_free_iomap(zdev, entry);
554 			return -ENOMEM;
555 		}
556 		zdev->bars[i].res = res;
557 	}
558 	zdev->has_resources = 1;
559 
560 	return 0;
561 }
562 
563 static void zpci_cleanup_bus_resources(struct zpci_dev *zdev)
564 {
565 	struct resource *res;
566 	int i;
567 
568 	pci_lock_rescan_remove();
569 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
570 		res = zdev->bars[i].res;
571 		if (!res)
572 			continue;
573 
574 		release_resource(res);
575 		pci_bus_remove_resource(zdev->zbus->bus, res);
576 		zpci_free_iomap(zdev, zdev->bars[i].map_idx);
577 		zdev->bars[i].res = NULL;
578 		kfree(res);
579 	}
580 	zdev->has_resources = 0;
581 	pci_unlock_rescan_remove();
582 }
583 
584 int pcibios_device_add(struct pci_dev *pdev)
585 {
586 	struct zpci_dev *zdev = to_zpci(pdev);
587 	struct resource *res;
588 	int i;
589 
590 	/* The pdev has a reference to the zdev via its bus */
591 	zpci_zdev_get(zdev);
592 	if (pdev->is_physfn)
593 		pdev->no_vf_scan = 1;
594 
595 	pdev->dev.groups = zpci_attr_groups;
596 	zpci_map_resources(pdev);
597 
598 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
599 		res = &pdev->resource[i];
600 		if (res->parent || !res->flags)
601 			continue;
602 		pci_claim_resource(pdev, i);
603 	}
604 
605 	return 0;
606 }
607 
608 void pcibios_release_device(struct pci_dev *pdev)
609 {
610 	struct zpci_dev *zdev = to_zpci(pdev);
611 
612 	zpci_unmap_resources(pdev);
613 	zpci_zdev_put(zdev);
614 }
615 
616 int pcibios_enable_device(struct pci_dev *pdev, int mask)
617 {
618 	struct zpci_dev *zdev = to_zpci(pdev);
619 
620 	zpci_debug_init_device(zdev, dev_name(&pdev->dev));
621 	zpci_fmb_enable_device(zdev);
622 
623 	return pci_enable_resources(pdev, mask);
624 }
625 
626 void pcibios_disable_device(struct pci_dev *pdev)
627 {
628 	struct zpci_dev *zdev = to_zpci(pdev);
629 
630 	zpci_fmb_disable_device(zdev);
631 	zpci_debug_exit_device(zdev);
632 }
633 
634 static int __zpci_register_domain(int domain)
635 {
636 	spin_lock(&zpci_domain_lock);
637 	if (test_bit(domain, zpci_domain)) {
638 		spin_unlock(&zpci_domain_lock);
639 		pr_err("Domain %04x is already assigned\n", domain);
640 		return -EEXIST;
641 	}
642 	set_bit(domain, zpci_domain);
643 	spin_unlock(&zpci_domain_lock);
644 	return domain;
645 }
646 
647 static int __zpci_alloc_domain(void)
648 {
649 	int domain;
650 
651 	spin_lock(&zpci_domain_lock);
652 	/*
653 	 * We can always auto allocate domains below ZPCI_NR_DEVICES.
654 	 * There is either a free domain or we have reached the maximum in
655 	 * which case we would have bailed earlier.
656 	 */
657 	domain = find_first_zero_bit(zpci_domain, ZPCI_NR_DEVICES);
658 	set_bit(domain, zpci_domain);
659 	spin_unlock(&zpci_domain_lock);
660 	return domain;
661 }
662 
663 int zpci_alloc_domain(int domain)
664 {
665 	if (zpci_unique_uid) {
666 		if (domain)
667 			return __zpci_register_domain(domain);
668 		pr_warn("UID checking was active but no UID is provided: switching to automatic domain allocation\n");
669 		update_uid_checking(false);
670 	}
671 	return __zpci_alloc_domain();
672 }
673 
674 void zpci_free_domain(int domain)
675 {
676 	spin_lock(&zpci_domain_lock);
677 	clear_bit(domain, zpci_domain);
678 	spin_unlock(&zpci_domain_lock);
679 }
680 
681 
682 int zpci_enable_device(struct zpci_dev *zdev)
683 {
684 	u32 fh = zdev->fh;
685 	int rc = 0;
686 
687 	if (clp_enable_fh(zdev, &fh, ZPCI_NR_DMA_SPACES))
688 		rc = -EIO;
689 	else
690 		zpci_update_fh(zdev, fh);
691 	return rc;
692 }
693 EXPORT_SYMBOL_GPL(zpci_enable_device);
694 
695 int zpci_disable_device(struct zpci_dev *zdev)
696 {
697 	u32 fh = zdev->fh;
698 	int cc, rc = 0;
699 
700 	cc = clp_disable_fh(zdev, &fh);
701 	if (!cc) {
702 		zpci_update_fh(zdev, fh);
703 	} else if (cc == CLP_RC_SETPCIFN_ALRDY) {
704 		pr_info("Disabling PCI function %08x had no effect as it was already disabled\n",
705 			zdev->fid);
706 		/* Function is already disabled - update handle */
707 		rc = clp_refresh_fh(zdev->fid, &fh);
708 		if (!rc) {
709 			zpci_update_fh(zdev, fh);
710 			rc = -EINVAL;
711 		}
712 	} else {
713 		rc = -EIO;
714 	}
715 	return rc;
716 }
717 EXPORT_SYMBOL_GPL(zpci_disable_device);
718 
719 /**
720  * zpci_hot_reset_device - perform a reset of the given zPCI function
721  * @zdev: the slot which should be reset
722  *
723  * Performs a low level reset of the zPCI function. The reset is low level in
724  * the sense that the zPCI function can be reset without detaching it from the
725  * common PCI subsystem. The reset may be performed while under control of
726  * either DMA or IOMMU APIs in which case the existing DMA/IOMMU translation
727  * table is reinstated at the end of the reset.
728  *
729  * After the reset the functions internal state is reset to an initial state
730  * equivalent to its state during boot when first probing a driver.
731  * Consequently after reset the PCI function requires re-initialization via the
732  * common PCI code including re-enabling IRQs via pci_alloc_irq_vectors()
733  * and enabling the function via e.g.pci_enablde_device_flags().The caller
734  * must guard against concurrent reset attempts.
735  *
736  * In most cases this function should not be called directly but through
737  * pci_reset_function() or pci_reset_bus() which handle the save/restore and
738  * locking.
739  *
740  * Return: 0 on success and an error value otherwise
741  */
742 int zpci_hot_reset_device(struct zpci_dev *zdev)
743 {
744 	u8 status;
745 	int rc;
746 
747 	zpci_dbg(3, "rst fid:%x, fh:%x\n", zdev->fid, zdev->fh);
748 	if (zdev_enabled(zdev)) {
749 		/* Disables device access, DMAs and IRQs (reset state) */
750 		rc = zpci_disable_device(zdev);
751 		/*
752 		 * Due to a z/VM vs LPAR inconsistency in the error state the
753 		 * FH may indicate an enabled device but disable says the
754 		 * device is already disabled don't treat it as an error here.
755 		 */
756 		if (rc == -EINVAL)
757 			rc = 0;
758 		if (rc)
759 			return rc;
760 	}
761 
762 	rc = zpci_enable_device(zdev);
763 	if (rc)
764 		return rc;
765 
766 	if (zdev->dma_table)
767 		rc = zpci_register_ioat(zdev, 0, zdev->start_dma, zdev->end_dma,
768 					virt_to_phys(zdev->dma_table), &status);
769 	if (rc) {
770 		zpci_disable_device(zdev);
771 		return rc;
772 	}
773 
774 	return 0;
775 }
776 
777 /**
778  * zpci_create_device() - Create a new zpci_dev and add it to the zbus
779  * @fid: Function ID of the device to be created
780  * @fh: Current Function Handle of the device to be created
781  * @state: Initial state after creation either Standby or Configured
782  *
783  * Creates a new zpci device and adds it to its, possibly newly created, zbus
784  * as well as zpci_list.
785  *
786  * Returns: the zdev on success or an error pointer otherwise
787  */
788 struct zpci_dev *zpci_create_device(u32 fid, u32 fh, enum zpci_state state)
789 {
790 	struct zpci_dev *zdev;
791 	int rc;
792 
793 	zpci_dbg(1, "add fid:%x, fh:%x, c:%d\n", fid, fh, state);
794 	zdev = kzalloc(sizeof(*zdev), GFP_KERNEL);
795 	if (!zdev)
796 		return ERR_PTR(-ENOMEM);
797 
798 	/* FID and Function Handle are the static/dynamic identifiers */
799 	zdev->fid = fid;
800 	zdev->fh = fh;
801 
802 	/* Query function properties and update zdev */
803 	rc = clp_query_pci_fn(zdev);
804 	if (rc)
805 		goto error;
806 	zdev->state =  state;
807 
808 	kref_init(&zdev->kref);
809 	mutex_init(&zdev->lock);
810 	mutex_init(&zdev->kzdev_lock);
811 
812 	rc = zpci_init_iommu(zdev);
813 	if (rc)
814 		goto error;
815 
816 	rc = zpci_bus_device_register(zdev, &pci_root_ops);
817 	if (rc)
818 		goto error_destroy_iommu;
819 
820 	spin_lock(&zpci_list_lock);
821 	list_add_tail(&zdev->entry, &zpci_list);
822 	spin_unlock(&zpci_list_lock);
823 
824 	return zdev;
825 
826 error_destroy_iommu:
827 	zpci_destroy_iommu(zdev);
828 error:
829 	zpci_dbg(0, "add fid:%x, rc:%d\n", fid, rc);
830 	kfree(zdev);
831 	return ERR_PTR(rc);
832 }
833 
834 bool zpci_is_device_configured(struct zpci_dev *zdev)
835 {
836 	enum zpci_state state = zdev->state;
837 
838 	return state != ZPCI_FN_STATE_RESERVED &&
839 		state != ZPCI_FN_STATE_STANDBY;
840 }
841 
842 /**
843  * zpci_scan_configured_device() - Scan a freshly configured zpci_dev
844  * @zdev: The zpci_dev to be configured
845  * @fh: The general function handle supplied by the platform
846  *
847  * Given a device in the configuration state Configured, enables, scans and
848  * adds it to the common code PCI subsystem if possible. If any failure occurs,
849  * the zpci_dev is left disabled.
850  *
851  * Return: 0 on success, or an error code otherwise
852  */
853 int zpci_scan_configured_device(struct zpci_dev *zdev, u32 fh)
854 {
855 	zpci_update_fh(zdev, fh);
856 	return zpci_bus_scan_device(zdev);
857 }
858 
859 /**
860  * zpci_deconfigure_device() - Deconfigure a zpci_dev
861  * @zdev: The zpci_dev to configure
862  *
863  * Deconfigure a zPCI function that is currently configured and possibly known
864  * to the common code PCI subsystem.
865  * If any failure occurs the device is left as is.
866  *
867  * Return: 0 on success, or an error code otherwise
868  */
869 int zpci_deconfigure_device(struct zpci_dev *zdev)
870 {
871 	int rc;
872 
873 	if (zdev->zbus->bus)
874 		zpci_bus_remove_device(zdev, false);
875 
876 	if (zdev_enabled(zdev)) {
877 		rc = zpci_disable_device(zdev);
878 		if (rc)
879 			return rc;
880 	}
881 
882 	rc = sclp_pci_deconfigure(zdev->fid);
883 	zpci_dbg(3, "deconf fid:%x, rc:%d\n", zdev->fid, rc);
884 	if (rc)
885 		return rc;
886 	zdev->state = ZPCI_FN_STATE_STANDBY;
887 
888 	return 0;
889 }
890 
891 /**
892  * zpci_device_reserved() - Mark device as resverved
893  * @zdev: the zpci_dev that was reserved
894  *
895  * Handle the case that a given zPCI function was reserved by another system.
896  * After a call to this function the zpci_dev can not be found via
897  * get_zdev_by_fid() anymore but may still be accessible via existing
898  * references though it will not be functional anymore.
899  */
900 void zpci_device_reserved(struct zpci_dev *zdev)
901 {
902 	if (zdev->has_hp_slot)
903 		zpci_exit_slot(zdev);
904 	/*
905 	 * Remove device from zpci_list as it is going away. This also
906 	 * makes sure we ignore subsequent zPCI events for this device.
907 	 */
908 	spin_lock(&zpci_list_lock);
909 	list_del(&zdev->entry);
910 	spin_unlock(&zpci_list_lock);
911 	zdev->state = ZPCI_FN_STATE_RESERVED;
912 	zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
913 	zpci_zdev_put(zdev);
914 }
915 
916 void zpci_release_device(struct kref *kref)
917 {
918 	struct zpci_dev *zdev = container_of(kref, struct zpci_dev, kref);
919 	int ret;
920 
921 	if (zdev->zbus->bus)
922 		zpci_bus_remove_device(zdev, false);
923 
924 	if (zdev_enabled(zdev))
925 		zpci_disable_device(zdev);
926 
927 	switch (zdev->state) {
928 	case ZPCI_FN_STATE_CONFIGURED:
929 		ret = sclp_pci_deconfigure(zdev->fid);
930 		zpci_dbg(3, "deconf fid:%x, rc:%d\n", zdev->fid, ret);
931 		fallthrough;
932 	case ZPCI_FN_STATE_STANDBY:
933 		if (zdev->has_hp_slot)
934 			zpci_exit_slot(zdev);
935 		spin_lock(&zpci_list_lock);
936 		list_del(&zdev->entry);
937 		spin_unlock(&zpci_list_lock);
938 		zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
939 		fallthrough;
940 	case ZPCI_FN_STATE_RESERVED:
941 		if (zdev->has_resources)
942 			zpci_cleanup_bus_resources(zdev);
943 		zpci_bus_device_unregister(zdev);
944 		zpci_destroy_iommu(zdev);
945 		fallthrough;
946 	default:
947 		break;
948 	}
949 	zpci_dbg(3, "rem fid:%x\n", zdev->fid);
950 	kfree_rcu(zdev, rcu);
951 }
952 
953 int zpci_report_error(struct pci_dev *pdev,
954 		      struct zpci_report_error_header *report)
955 {
956 	struct zpci_dev *zdev = to_zpci(pdev);
957 
958 	return sclp_pci_report(report, zdev->fh, zdev->fid);
959 }
960 EXPORT_SYMBOL(zpci_report_error);
961 
962 /**
963  * zpci_clear_error_state() - Clears the zPCI error state of the device
964  * @zdev: The zdev for which the zPCI error state should be reset
965  *
966  * Clear the zPCI error state of the device. If clearing the zPCI error state
967  * fails the device is left in the error state. In this case it may make sense
968  * to call zpci_io_perm_failure() on the associated pdev if it exists.
969  *
970  * Returns: 0 on success, -EIO otherwise
971  */
972 int zpci_clear_error_state(struct zpci_dev *zdev)
973 {
974 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_RESET_ERROR);
975 	struct zpci_fib fib = {0};
976 	u8 status;
977 	int cc;
978 
979 	cc = zpci_mod_fc(req, &fib, &status);
980 	if (cc) {
981 		zpci_dbg(3, "ces fid:%x, cc:%d, status:%x\n", zdev->fid, cc, status);
982 		return -EIO;
983 	}
984 
985 	return 0;
986 }
987 
988 /**
989  * zpci_reset_load_store_blocked() - Re-enables L/S from error state
990  * @zdev: The zdev for which to unblock load/store access
991  *
992  * Re-enables load/store access for a PCI function in the error state while
993  * keeping DMA blocked. In this state drivers can poke MMIO space to determine
994  * if error recovery is possible while catching any rogue DMA access from the
995  * device.
996  *
997  * Returns: 0 on success, -EIO otherwise
998  */
999 int zpci_reset_load_store_blocked(struct zpci_dev *zdev)
1000 {
1001 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_RESET_BLOCK);
1002 	struct zpci_fib fib = {0};
1003 	u8 status;
1004 	int cc;
1005 
1006 	cc = zpci_mod_fc(req, &fib, &status);
1007 	if (cc) {
1008 		zpci_dbg(3, "rls fid:%x, cc:%d, status:%x\n", zdev->fid, cc, status);
1009 		return -EIO;
1010 	}
1011 
1012 	return 0;
1013 }
1014 
1015 static int zpci_mem_init(void)
1016 {
1017 	BUILD_BUG_ON(!is_power_of_2(__alignof__(struct zpci_fmb)) ||
1018 		     __alignof__(struct zpci_fmb) < sizeof(struct zpci_fmb));
1019 
1020 	zdev_fmb_cache = kmem_cache_create("PCI_FMB_cache", sizeof(struct zpci_fmb),
1021 					   __alignof__(struct zpci_fmb), 0, NULL);
1022 	if (!zdev_fmb_cache)
1023 		goto error_fmb;
1024 
1025 	zpci_iomap_start = kcalloc(ZPCI_IOMAP_ENTRIES,
1026 				   sizeof(*zpci_iomap_start), GFP_KERNEL);
1027 	if (!zpci_iomap_start)
1028 		goto error_iomap;
1029 
1030 	zpci_iomap_bitmap = kcalloc(BITS_TO_LONGS(ZPCI_IOMAP_ENTRIES),
1031 				    sizeof(*zpci_iomap_bitmap), GFP_KERNEL);
1032 	if (!zpci_iomap_bitmap)
1033 		goto error_iomap_bitmap;
1034 
1035 	if (static_branch_likely(&have_mio))
1036 		clp_setup_writeback_mio();
1037 
1038 	return 0;
1039 error_iomap_bitmap:
1040 	kfree(zpci_iomap_start);
1041 error_iomap:
1042 	kmem_cache_destroy(zdev_fmb_cache);
1043 error_fmb:
1044 	return -ENOMEM;
1045 }
1046 
1047 static void zpci_mem_exit(void)
1048 {
1049 	kfree(zpci_iomap_bitmap);
1050 	kfree(zpci_iomap_start);
1051 	kmem_cache_destroy(zdev_fmb_cache);
1052 }
1053 
1054 static unsigned int s390_pci_probe __initdata = 1;
1055 unsigned int s390_pci_force_floating __initdata;
1056 static unsigned int s390_pci_initialized;
1057 
1058 char * __init pcibios_setup(char *str)
1059 {
1060 	if (!strcmp(str, "off")) {
1061 		s390_pci_probe = 0;
1062 		return NULL;
1063 	}
1064 	if (!strcmp(str, "nomio")) {
1065 		S390_lowcore.machine_flags &= ~MACHINE_FLAG_PCI_MIO;
1066 		return NULL;
1067 	}
1068 	if (!strcmp(str, "force_floating")) {
1069 		s390_pci_force_floating = 1;
1070 		return NULL;
1071 	}
1072 	if (!strcmp(str, "norid")) {
1073 		s390_pci_no_rid = 1;
1074 		return NULL;
1075 	}
1076 	return str;
1077 }
1078 
1079 bool zpci_is_enabled(void)
1080 {
1081 	return s390_pci_initialized;
1082 }
1083 
1084 static int __init pci_base_init(void)
1085 {
1086 	int rc;
1087 
1088 	if (!s390_pci_probe)
1089 		return 0;
1090 
1091 	if (!test_facility(69) || !test_facility(71)) {
1092 		pr_info("PCI is not supported because CPU facilities 69 or 71 are not available\n");
1093 		return 0;
1094 	}
1095 
1096 	if (MACHINE_HAS_PCI_MIO) {
1097 		static_branch_enable(&have_mio);
1098 		system_ctl_set_bit(2, CR2_MIO_ADDRESSING_BIT);
1099 	}
1100 
1101 	rc = zpci_debug_init();
1102 	if (rc)
1103 		goto out;
1104 
1105 	rc = zpci_mem_init();
1106 	if (rc)
1107 		goto out_mem;
1108 
1109 	rc = zpci_irq_init();
1110 	if (rc)
1111 		goto out_irq;
1112 
1113 	rc = clp_scan_pci_devices();
1114 	if (rc)
1115 		goto out_find;
1116 	zpci_bus_scan_busses();
1117 
1118 	s390_pci_initialized = 1;
1119 	return 0;
1120 
1121 out_find:
1122 	zpci_irq_exit();
1123 out_irq:
1124 	zpci_mem_exit();
1125 out_mem:
1126 	zpci_debug_exit();
1127 out:
1128 	return rc;
1129 }
1130 subsys_initcall_sync(pci_base_init);
1131