xref: /linux/arch/s390/pci/pci.c (revision a90e8608eb0ed93d31ac0feb055f77ce59512542)
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 	fib.pal = limit;
128 	fib.iota = iota | ZPCI_IOTA_RTTO_FLAG;
129 	fib.gd = zdev->gisa;
130 	cc = zpci_mod_fc(req, &fib, status);
131 	if (cc)
132 		zpci_dbg(3, "reg ioat fid:%x, cc:%d, status:%d\n", zdev->fid, cc, *status);
133 	return cc;
134 }
135 EXPORT_SYMBOL_GPL(zpci_register_ioat);
136 
137 /* Modify PCI: Unregister I/O address translation parameters */
138 int zpci_unregister_ioat(struct zpci_dev *zdev, u8 dmaas)
139 {
140 	u64 req = ZPCI_CREATE_REQ(zdev->fh, dmaas, ZPCI_MOD_FC_DEREG_IOAT);
141 	struct zpci_fib fib = {0};
142 	u8 cc, status;
143 
144 	fib.gd = zdev->gisa;
145 
146 	cc = zpci_mod_fc(req, &fib, &status);
147 	if (cc)
148 		zpci_dbg(3, "unreg ioat fid:%x, cc:%d, status:%d\n", zdev->fid, cc, status);
149 	return cc;
150 }
151 
152 /* Modify PCI: Set PCI function measurement parameters */
153 int zpci_fmb_enable_device(struct zpci_dev *zdev)
154 {
155 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_SET_MEASURE);
156 	struct zpci_fib fib = {0};
157 	u8 cc, status;
158 
159 	if (zdev->fmb || sizeof(*zdev->fmb) < zdev->fmb_length)
160 		return -EINVAL;
161 
162 	zdev->fmb = kmem_cache_zalloc(zdev_fmb_cache, GFP_KERNEL);
163 	if (!zdev->fmb)
164 		return -ENOMEM;
165 	WARN_ON((u64) zdev->fmb & 0xf);
166 
167 	/* reset software counters */
168 	atomic64_set(&zdev->allocated_pages, 0);
169 	atomic64_set(&zdev->mapped_pages, 0);
170 	atomic64_set(&zdev->unmapped_pages, 0);
171 
172 	fib.fmb_addr = virt_to_phys(zdev->fmb);
173 	fib.gd = zdev->gisa;
174 	cc = zpci_mod_fc(req, &fib, &status);
175 	if (cc) {
176 		kmem_cache_free(zdev_fmb_cache, zdev->fmb);
177 		zdev->fmb = NULL;
178 	}
179 	return cc ? -EIO : 0;
180 }
181 
182 /* Modify PCI: Disable PCI function measurement */
183 int zpci_fmb_disable_device(struct zpci_dev *zdev)
184 {
185 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_SET_MEASURE);
186 	struct zpci_fib fib = {0};
187 	u8 cc, status;
188 
189 	if (!zdev->fmb)
190 		return -EINVAL;
191 
192 	fib.gd = zdev->gisa;
193 
194 	/* Function measurement is disabled if fmb address is zero */
195 	cc = zpci_mod_fc(req, &fib, &status);
196 	if (cc == 3) /* Function already gone. */
197 		cc = 0;
198 
199 	if (!cc) {
200 		kmem_cache_free(zdev_fmb_cache, zdev->fmb);
201 		zdev->fmb = NULL;
202 	}
203 	return cc ? -EIO : 0;
204 }
205 
206 static int zpci_cfg_load(struct zpci_dev *zdev, int offset, u32 *val, u8 len)
207 {
208 	u64 req = ZPCI_CREATE_REQ(zdev->fh, ZPCI_PCIAS_CFGSPC, len);
209 	u64 data;
210 	int rc;
211 
212 	rc = __zpci_load(&data, req, offset);
213 	if (!rc) {
214 		data = le64_to_cpu((__force __le64) data);
215 		data >>= (8 - len) * 8;
216 		*val = (u32) data;
217 	} else
218 		*val = 0xffffffff;
219 	return rc;
220 }
221 
222 static int zpci_cfg_store(struct zpci_dev *zdev, int offset, u32 val, u8 len)
223 {
224 	u64 req = ZPCI_CREATE_REQ(zdev->fh, ZPCI_PCIAS_CFGSPC, len);
225 	u64 data = val;
226 	int rc;
227 
228 	data <<= (8 - len) * 8;
229 	data = (__force u64) cpu_to_le64(data);
230 	rc = __zpci_store(data, req, offset);
231 	return rc;
232 }
233 
234 resource_size_t pcibios_align_resource(void *data, const struct resource *res,
235 				       resource_size_t size,
236 				       resource_size_t align)
237 {
238 	return 0;
239 }
240 
241 /* combine single writes by using store-block insn */
242 void __iowrite64_copy(void __iomem *to, const void *from, size_t count)
243 {
244        zpci_memcpy_toio(to, from, count);
245 }
246 
247 static void __iomem *__ioremap(phys_addr_t addr, size_t size, pgprot_t prot)
248 {
249 	unsigned long offset, vaddr;
250 	struct vm_struct *area;
251 	phys_addr_t last_addr;
252 
253 	last_addr = addr + size - 1;
254 	if (!size || last_addr < addr)
255 		return NULL;
256 
257 	if (!static_branch_unlikely(&have_mio))
258 		return (void __iomem *) addr;
259 
260 	offset = addr & ~PAGE_MASK;
261 	addr &= PAGE_MASK;
262 	size = PAGE_ALIGN(size + offset);
263 	area = get_vm_area(size, VM_IOREMAP);
264 	if (!area)
265 		return NULL;
266 
267 	vaddr = (unsigned long) area->addr;
268 	if (ioremap_page_range(vaddr, vaddr + size, addr, prot)) {
269 		free_vm_area(area);
270 		return NULL;
271 	}
272 	return (void __iomem *) ((unsigned long) area->addr + offset);
273 }
274 
275 void __iomem *ioremap_prot(phys_addr_t addr, size_t size, unsigned long prot)
276 {
277 	return __ioremap(addr, size, __pgprot(prot));
278 }
279 EXPORT_SYMBOL(ioremap_prot);
280 
281 void __iomem *ioremap(phys_addr_t addr, size_t size)
282 {
283 	return __ioremap(addr, size, PAGE_KERNEL);
284 }
285 EXPORT_SYMBOL(ioremap);
286 
287 void __iomem *ioremap_wc(phys_addr_t addr, size_t size)
288 {
289 	return __ioremap(addr, size, pgprot_writecombine(PAGE_KERNEL));
290 }
291 EXPORT_SYMBOL(ioremap_wc);
292 
293 void __iomem *ioremap_wt(phys_addr_t addr, size_t size)
294 {
295 	return __ioremap(addr, size, pgprot_writethrough(PAGE_KERNEL));
296 }
297 EXPORT_SYMBOL(ioremap_wt);
298 
299 void iounmap(volatile void __iomem *addr)
300 {
301 	if (static_branch_likely(&have_mio))
302 		vunmap((__force void *) ((unsigned long) addr & PAGE_MASK));
303 }
304 EXPORT_SYMBOL(iounmap);
305 
306 /* Create a virtual mapping cookie for a PCI BAR */
307 static void __iomem *pci_iomap_range_fh(struct pci_dev *pdev, int bar,
308 					unsigned long offset, unsigned long max)
309 {
310 	struct zpci_dev *zdev =	to_zpci(pdev);
311 	int idx;
312 
313 	idx = zdev->bars[bar].map_idx;
314 	spin_lock(&zpci_iomap_lock);
315 	/* Detect overrun */
316 	WARN_ON(!++zpci_iomap_start[idx].count);
317 	zpci_iomap_start[idx].fh = zdev->fh;
318 	zpci_iomap_start[idx].bar = bar;
319 	spin_unlock(&zpci_iomap_lock);
320 
321 	return (void __iomem *) ZPCI_ADDR(idx) + offset;
322 }
323 
324 static void __iomem *pci_iomap_range_mio(struct pci_dev *pdev, int bar,
325 					 unsigned long offset,
326 					 unsigned long max)
327 {
328 	unsigned long barsize = pci_resource_len(pdev, bar);
329 	struct zpci_dev *zdev = to_zpci(pdev);
330 	void __iomem *iova;
331 
332 	iova = ioremap((unsigned long) zdev->bars[bar].mio_wt, barsize);
333 	return iova ? iova + offset : iova;
334 }
335 
336 void __iomem *pci_iomap_range(struct pci_dev *pdev, int bar,
337 			      unsigned long offset, unsigned long max)
338 {
339 	if (bar >= PCI_STD_NUM_BARS || !pci_resource_len(pdev, bar))
340 		return NULL;
341 
342 	if (static_branch_likely(&have_mio))
343 		return pci_iomap_range_mio(pdev, bar, offset, max);
344 	else
345 		return pci_iomap_range_fh(pdev, bar, offset, max);
346 }
347 EXPORT_SYMBOL(pci_iomap_range);
348 
349 void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long maxlen)
350 {
351 	return pci_iomap_range(dev, bar, 0, maxlen);
352 }
353 EXPORT_SYMBOL(pci_iomap);
354 
355 static void __iomem *pci_iomap_wc_range_mio(struct pci_dev *pdev, int bar,
356 					    unsigned long offset, unsigned long max)
357 {
358 	unsigned long barsize = pci_resource_len(pdev, bar);
359 	struct zpci_dev *zdev = to_zpci(pdev);
360 	void __iomem *iova;
361 
362 	iova = ioremap((unsigned long) zdev->bars[bar].mio_wb, barsize);
363 	return iova ? iova + offset : iova;
364 }
365 
366 void __iomem *pci_iomap_wc_range(struct pci_dev *pdev, int bar,
367 				 unsigned long offset, unsigned long max)
368 {
369 	if (bar >= PCI_STD_NUM_BARS || !pci_resource_len(pdev, bar))
370 		return NULL;
371 
372 	if (static_branch_likely(&have_mio))
373 		return pci_iomap_wc_range_mio(pdev, bar, offset, max);
374 	else
375 		return pci_iomap_range_fh(pdev, bar, offset, max);
376 }
377 EXPORT_SYMBOL(pci_iomap_wc_range);
378 
379 void __iomem *pci_iomap_wc(struct pci_dev *dev, int bar, unsigned long maxlen)
380 {
381 	return pci_iomap_wc_range(dev, bar, 0, maxlen);
382 }
383 EXPORT_SYMBOL(pci_iomap_wc);
384 
385 static void pci_iounmap_fh(struct pci_dev *pdev, void __iomem *addr)
386 {
387 	unsigned int idx = ZPCI_IDX(addr);
388 
389 	spin_lock(&zpci_iomap_lock);
390 	/* Detect underrun */
391 	WARN_ON(!zpci_iomap_start[idx].count);
392 	if (!--zpci_iomap_start[idx].count) {
393 		zpci_iomap_start[idx].fh = 0;
394 		zpci_iomap_start[idx].bar = 0;
395 	}
396 	spin_unlock(&zpci_iomap_lock);
397 }
398 
399 static void pci_iounmap_mio(struct pci_dev *pdev, void __iomem *addr)
400 {
401 	iounmap(addr);
402 }
403 
404 void pci_iounmap(struct pci_dev *pdev, void __iomem *addr)
405 {
406 	if (static_branch_likely(&have_mio))
407 		pci_iounmap_mio(pdev, addr);
408 	else
409 		pci_iounmap_fh(pdev, addr);
410 }
411 EXPORT_SYMBOL(pci_iounmap);
412 
413 static int pci_read(struct pci_bus *bus, unsigned int devfn, int where,
414 		    int size, u32 *val)
415 {
416 	struct zpci_dev *zdev = zdev_from_bus(bus, devfn);
417 
418 	return (zdev) ? zpci_cfg_load(zdev, where, val, size) : -ENODEV;
419 }
420 
421 static int pci_write(struct pci_bus *bus, unsigned int devfn, int where,
422 		     int size, u32 val)
423 {
424 	struct zpci_dev *zdev = zdev_from_bus(bus, devfn);
425 
426 	return (zdev) ? zpci_cfg_store(zdev, where, val, size) : -ENODEV;
427 }
428 
429 static struct pci_ops pci_root_ops = {
430 	.read = pci_read,
431 	.write = pci_write,
432 };
433 
434 static void zpci_map_resources(struct pci_dev *pdev)
435 {
436 	struct zpci_dev *zdev = to_zpci(pdev);
437 	resource_size_t len;
438 	int i;
439 
440 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
441 		len = pci_resource_len(pdev, i);
442 		if (!len)
443 			continue;
444 
445 		if (zpci_use_mio(zdev))
446 			pdev->resource[i].start =
447 				(resource_size_t __force) zdev->bars[i].mio_wt;
448 		else
449 			pdev->resource[i].start = (resource_size_t __force)
450 				pci_iomap_range_fh(pdev, i, 0, 0);
451 		pdev->resource[i].end = pdev->resource[i].start + len - 1;
452 	}
453 
454 	zpci_iov_map_resources(pdev);
455 }
456 
457 static void zpci_unmap_resources(struct pci_dev *pdev)
458 {
459 	struct zpci_dev *zdev = to_zpci(pdev);
460 	resource_size_t len;
461 	int i;
462 
463 	if (zpci_use_mio(zdev))
464 		return;
465 
466 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
467 		len = pci_resource_len(pdev, i);
468 		if (!len)
469 			continue;
470 		pci_iounmap_fh(pdev, (void __iomem __force *)
471 			       pdev->resource[i].start);
472 	}
473 }
474 
475 static int zpci_alloc_iomap(struct zpci_dev *zdev)
476 {
477 	unsigned long entry;
478 
479 	spin_lock(&zpci_iomap_lock);
480 	entry = find_first_zero_bit(zpci_iomap_bitmap, ZPCI_IOMAP_ENTRIES);
481 	if (entry == ZPCI_IOMAP_ENTRIES) {
482 		spin_unlock(&zpci_iomap_lock);
483 		return -ENOSPC;
484 	}
485 	set_bit(entry, zpci_iomap_bitmap);
486 	spin_unlock(&zpci_iomap_lock);
487 	return entry;
488 }
489 
490 static void zpci_free_iomap(struct zpci_dev *zdev, int entry)
491 {
492 	spin_lock(&zpci_iomap_lock);
493 	memset(&zpci_iomap_start[entry], 0, sizeof(struct zpci_iomap_entry));
494 	clear_bit(entry, zpci_iomap_bitmap);
495 	spin_unlock(&zpci_iomap_lock);
496 }
497 
498 static void zpci_do_update_iomap_fh(struct zpci_dev *zdev, u32 fh)
499 {
500 	int bar, idx;
501 
502 	spin_lock(&zpci_iomap_lock);
503 	for (bar = 0; bar < PCI_STD_NUM_BARS; bar++) {
504 		if (!zdev->bars[bar].size)
505 			continue;
506 		idx = zdev->bars[bar].map_idx;
507 		if (!zpci_iomap_start[idx].count)
508 			continue;
509 		WRITE_ONCE(zpci_iomap_start[idx].fh, zdev->fh);
510 	}
511 	spin_unlock(&zpci_iomap_lock);
512 }
513 
514 void zpci_update_fh(struct zpci_dev *zdev, u32 fh)
515 {
516 	if (!fh || zdev->fh == fh)
517 		return;
518 
519 	zdev->fh = fh;
520 	if (zpci_use_mio(zdev))
521 		return;
522 	if (zdev->has_resources && zdev_enabled(zdev))
523 		zpci_do_update_iomap_fh(zdev, fh);
524 }
525 
526 static struct resource *__alloc_res(struct zpci_dev *zdev, unsigned long start,
527 				    unsigned long size, unsigned long flags)
528 {
529 	struct resource *r;
530 
531 	r = kzalloc(sizeof(*r), GFP_KERNEL);
532 	if (!r)
533 		return NULL;
534 
535 	r->start = start;
536 	r->end = r->start + size - 1;
537 	r->flags = flags;
538 	r->name = zdev->res_name;
539 
540 	if (request_resource(&iomem_resource, r)) {
541 		kfree(r);
542 		return NULL;
543 	}
544 	return r;
545 }
546 
547 int zpci_setup_bus_resources(struct zpci_dev *zdev)
548 {
549 	unsigned long addr, size, flags;
550 	struct resource *res;
551 	int i, entry;
552 
553 	snprintf(zdev->res_name, sizeof(zdev->res_name),
554 		 "PCI Bus %04x:%02x", zdev->uid, ZPCI_BUS_NR);
555 
556 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
557 		if (!zdev->bars[i].size)
558 			continue;
559 		entry = zpci_alloc_iomap(zdev);
560 		if (entry < 0)
561 			return entry;
562 		zdev->bars[i].map_idx = entry;
563 
564 		/* only MMIO is supported */
565 		flags = IORESOURCE_MEM;
566 		if (zdev->bars[i].val & 8)
567 			flags |= IORESOURCE_PREFETCH;
568 		if (zdev->bars[i].val & 4)
569 			flags |= IORESOURCE_MEM_64;
570 
571 		if (zpci_use_mio(zdev))
572 			addr = (unsigned long) zdev->bars[i].mio_wt;
573 		else
574 			addr = ZPCI_ADDR(entry);
575 		size = 1UL << zdev->bars[i].size;
576 
577 		res = __alloc_res(zdev, addr, size, flags);
578 		if (!res) {
579 			zpci_free_iomap(zdev, entry);
580 			return -ENOMEM;
581 		}
582 		zdev->bars[i].res = res;
583 	}
584 	zdev->has_resources = 1;
585 
586 	return 0;
587 }
588 
589 static void zpci_cleanup_bus_resources(struct zpci_dev *zdev)
590 {
591 	struct resource *res;
592 	int i;
593 
594 	pci_lock_rescan_remove();
595 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
596 		res = zdev->bars[i].res;
597 		if (!res)
598 			continue;
599 
600 		release_resource(res);
601 		pci_bus_remove_resource(zdev->zbus->bus, res);
602 		zpci_free_iomap(zdev, zdev->bars[i].map_idx);
603 		zdev->bars[i].res = NULL;
604 		kfree(res);
605 	}
606 	zdev->has_resources = 0;
607 	pci_unlock_rescan_remove();
608 }
609 
610 int pcibios_device_add(struct pci_dev *pdev)
611 {
612 	struct zpci_dev *zdev = to_zpci(pdev);
613 	struct resource *res;
614 	int i;
615 
616 	/* The pdev has a reference to the zdev via its bus */
617 	zpci_zdev_get(zdev);
618 	if (pdev->is_physfn)
619 		pdev->no_vf_scan = 1;
620 
621 	pdev->dev.groups = zpci_attr_groups;
622 	pdev->dev.dma_ops = &s390_pci_dma_ops;
623 	zpci_map_resources(pdev);
624 
625 	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
626 		res = &pdev->resource[i];
627 		if (res->parent || !res->flags)
628 			continue;
629 		pci_claim_resource(pdev, i);
630 	}
631 
632 	return 0;
633 }
634 
635 void pcibios_release_device(struct pci_dev *pdev)
636 {
637 	struct zpci_dev *zdev = to_zpci(pdev);
638 
639 	zpci_unmap_resources(pdev);
640 	zpci_zdev_put(zdev);
641 }
642 
643 int pcibios_enable_device(struct pci_dev *pdev, int mask)
644 {
645 	struct zpci_dev *zdev = to_zpci(pdev);
646 
647 	zpci_debug_init_device(zdev, dev_name(&pdev->dev));
648 	zpci_fmb_enable_device(zdev);
649 
650 	return pci_enable_resources(pdev, mask);
651 }
652 
653 void pcibios_disable_device(struct pci_dev *pdev)
654 {
655 	struct zpci_dev *zdev = to_zpci(pdev);
656 
657 	zpci_fmb_disable_device(zdev);
658 	zpci_debug_exit_device(zdev);
659 }
660 
661 static int __zpci_register_domain(int domain)
662 {
663 	spin_lock(&zpci_domain_lock);
664 	if (test_bit(domain, zpci_domain)) {
665 		spin_unlock(&zpci_domain_lock);
666 		pr_err("Domain %04x is already assigned\n", domain);
667 		return -EEXIST;
668 	}
669 	set_bit(domain, zpci_domain);
670 	spin_unlock(&zpci_domain_lock);
671 	return domain;
672 }
673 
674 static int __zpci_alloc_domain(void)
675 {
676 	int domain;
677 
678 	spin_lock(&zpci_domain_lock);
679 	/*
680 	 * We can always auto allocate domains below ZPCI_NR_DEVICES.
681 	 * There is either a free domain or we have reached the maximum in
682 	 * which case we would have bailed earlier.
683 	 */
684 	domain = find_first_zero_bit(zpci_domain, ZPCI_NR_DEVICES);
685 	set_bit(domain, zpci_domain);
686 	spin_unlock(&zpci_domain_lock);
687 	return domain;
688 }
689 
690 int zpci_alloc_domain(int domain)
691 {
692 	if (zpci_unique_uid) {
693 		if (domain)
694 			return __zpci_register_domain(domain);
695 		pr_warn("UID checking was active but no UID is provided: switching to automatic domain allocation\n");
696 		update_uid_checking(false);
697 	}
698 	return __zpci_alloc_domain();
699 }
700 
701 void zpci_free_domain(int domain)
702 {
703 	spin_lock(&zpci_domain_lock);
704 	clear_bit(domain, zpci_domain);
705 	spin_unlock(&zpci_domain_lock);
706 }
707 
708 
709 int zpci_enable_device(struct zpci_dev *zdev)
710 {
711 	u32 fh = zdev->fh;
712 	int rc = 0;
713 
714 	if (clp_enable_fh(zdev, &fh, ZPCI_NR_DMA_SPACES))
715 		rc = -EIO;
716 	else
717 		zpci_update_fh(zdev, fh);
718 	return rc;
719 }
720 EXPORT_SYMBOL_GPL(zpci_enable_device);
721 
722 int zpci_disable_device(struct zpci_dev *zdev)
723 {
724 	u32 fh = zdev->fh;
725 	int cc, rc = 0;
726 
727 	cc = clp_disable_fh(zdev, &fh);
728 	if (!cc) {
729 		zpci_update_fh(zdev, fh);
730 	} else if (cc == CLP_RC_SETPCIFN_ALRDY) {
731 		pr_info("Disabling PCI function %08x had no effect as it was already disabled\n",
732 			zdev->fid);
733 		/* Function is already disabled - update handle */
734 		rc = clp_refresh_fh(zdev->fid, &fh);
735 		if (!rc) {
736 			zpci_update_fh(zdev, fh);
737 			rc = -EINVAL;
738 		}
739 	} else {
740 		rc = -EIO;
741 	}
742 	return rc;
743 }
744 EXPORT_SYMBOL_GPL(zpci_disable_device);
745 
746 /**
747  * zpci_hot_reset_device - perform a reset of the given zPCI function
748  * @zdev: the slot which should be reset
749  *
750  * Performs a low level reset of the zPCI function. The reset is low level in
751  * the sense that the zPCI function can be reset without detaching it from the
752  * common PCI subsystem. The reset may be performed while under control of
753  * either DMA or IOMMU APIs in which case the existing DMA/IOMMU translation
754  * table is reinstated at the end of the reset.
755  *
756  * After the reset the functions internal state is reset to an initial state
757  * equivalent to its state during boot when first probing a driver.
758  * Consequently after reset the PCI function requires re-initialization via the
759  * common PCI code including re-enabling IRQs via pci_alloc_irq_vectors()
760  * and enabling the function via e.g.pci_enablde_device_flags().The caller
761  * must guard against concurrent reset attempts.
762  *
763  * In most cases this function should not be called directly but through
764  * pci_reset_function() or pci_reset_bus() which handle the save/restore and
765  * locking.
766  *
767  * Return: 0 on success and an error value otherwise
768  */
769 int zpci_hot_reset_device(struct zpci_dev *zdev)
770 {
771 	u8 status;
772 	int rc;
773 
774 	zpci_dbg(3, "rst fid:%x, fh:%x\n", zdev->fid, zdev->fh);
775 	if (zdev_enabled(zdev)) {
776 		/* Disables device access, DMAs and IRQs (reset state) */
777 		rc = zpci_disable_device(zdev);
778 		/*
779 		 * Due to a z/VM vs LPAR inconsistency in the error state the
780 		 * FH may indicate an enabled device but disable says the
781 		 * device is already disabled don't treat it as an error here.
782 		 */
783 		if (rc == -EINVAL)
784 			rc = 0;
785 		if (rc)
786 			return rc;
787 	}
788 
789 	rc = zpci_enable_device(zdev);
790 	if (rc)
791 		return rc;
792 
793 	if (zdev->dma_table)
794 		rc = zpci_register_ioat(zdev, 0, zdev->start_dma, zdev->end_dma,
795 					virt_to_phys(zdev->dma_table), &status);
796 	else
797 		rc = zpci_dma_init_device(zdev);
798 	if (rc) {
799 		zpci_disable_device(zdev);
800 		return rc;
801 	}
802 
803 	return 0;
804 }
805 
806 /**
807  * zpci_create_device() - Create a new zpci_dev and add it to the zbus
808  * @fid: Function ID of the device to be created
809  * @fh: Current Function Handle of the device to be created
810  * @state: Initial state after creation either Standby or Configured
811  *
812  * Creates a new zpci device and adds it to its, possibly newly created, zbus
813  * as well as zpci_list.
814  *
815  * Returns: the zdev on success or an error pointer otherwise
816  */
817 struct zpci_dev *zpci_create_device(u32 fid, u32 fh, enum zpci_state state)
818 {
819 	struct zpci_dev *zdev;
820 	int rc;
821 
822 	zpci_dbg(1, "add fid:%x, fh:%x, c:%d\n", fid, fh, state);
823 	zdev = kzalloc(sizeof(*zdev), GFP_KERNEL);
824 	if (!zdev)
825 		return ERR_PTR(-ENOMEM);
826 
827 	/* FID and Function Handle are the static/dynamic identifiers */
828 	zdev->fid = fid;
829 	zdev->fh = fh;
830 
831 	/* Query function properties and update zdev */
832 	rc = clp_query_pci_fn(zdev);
833 	if (rc)
834 		goto error;
835 	zdev->state =  state;
836 
837 	kref_init(&zdev->kref);
838 	mutex_init(&zdev->lock);
839 	mutex_init(&zdev->kzdev_lock);
840 
841 	rc = zpci_init_iommu(zdev);
842 	if (rc)
843 		goto error;
844 
845 	rc = zpci_bus_device_register(zdev, &pci_root_ops);
846 	if (rc)
847 		goto error_destroy_iommu;
848 
849 	spin_lock(&zpci_list_lock);
850 	list_add_tail(&zdev->entry, &zpci_list);
851 	spin_unlock(&zpci_list_lock);
852 
853 	return zdev;
854 
855 error_destroy_iommu:
856 	zpci_destroy_iommu(zdev);
857 error:
858 	zpci_dbg(0, "add fid:%x, rc:%d\n", fid, rc);
859 	kfree(zdev);
860 	return ERR_PTR(rc);
861 }
862 
863 bool zpci_is_device_configured(struct zpci_dev *zdev)
864 {
865 	enum zpci_state state = zdev->state;
866 
867 	return state != ZPCI_FN_STATE_RESERVED &&
868 		state != ZPCI_FN_STATE_STANDBY;
869 }
870 
871 /**
872  * zpci_scan_configured_device() - Scan a freshly configured zpci_dev
873  * @zdev: The zpci_dev to be configured
874  * @fh: The general function handle supplied by the platform
875  *
876  * Given a device in the configuration state Configured, enables, scans and
877  * adds it to the common code PCI subsystem if possible. If any failure occurs,
878  * the zpci_dev is left disabled.
879  *
880  * Return: 0 on success, or an error code otherwise
881  */
882 int zpci_scan_configured_device(struct zpci_dev *zdev, u32 fh)
883 {
884 	zpci_update_fh(zdev, fh);
885 	return zpci_bus_scan_device(zdev);
886 }
887 
888 /**
889  * zpci_deconfigure_device() - Deconfigure a zpci_dev
890  * @zdev: The zpci_dev to configure
891  *
892  * Deconfigure a zPCI function that is currently configured and possibly known
893  * to the common code PCI subsystem.
894  * If any failure occurs the device is left as is.
895  *
896  * Return: 0 on success, or an error code otherwise
897  */
898 int zpci_deconfigure_device(struct zpci_dev *zdev)
899 {
900 	int rc;
901 
902 	if (zdev->zbus->bus)
903 		zpci_bus_remove_device(zdev, false);
904 
905 	if (zdev->dma_table) {
906 		rc = zpci_dma_exit_device(zdev);
907 		if (rc)
908 			return rc;
909 	}
910 	if (zdev_enabled(zdev)) {
911 		rc = zpci_disable_device(zdev);
912 		if (rc)
913 			return rc;
914 	}
915 
916 	rc = sclp_pci_deconfigure(zdev->fid);
917 	zpci_dbg(3, "deconf fid:%x, rc:%d\n", zdev->fid, rc);
918 	if (rc)
919 		return rc;
920 	zdev->state = ZPCI_FN_STATE_STANDBY;
921 
922 	return 0;
923 }
924 
925 /**
926  * zpci_device_reserved() - Mark device as resverved
927  * @zdev: the zpci_dev that was reserved
928  *
929  * Handle the case that a given zPCI function was reserved by another system.
930  * After a call to this function the zpci_dev can not be found via
931  * get_zdev_by_fid() anymore but may still be accessible via existing
932  * references though it will not be functional anymore.
933  */
934 void zpci_device_reserved(struct zpci_dev *zdev)
935 {
936 	if (zdev->has_hp_slot)
937 		zpci_exit_slot(zdev);
938 	/*
939 	 * Remove device from zpci_list as it is going away. This also
940 	 * makes sure we ignore subsequent zPCI events for this device.
941 	 */
942 	spin_lock(&zpci_list_lock);
943 	list_del(&zdev->entry);
944 	spin_unlock(&zpci_list_lock);
945 	zdev->state = ZPCI_FN_STATE_RESERVED;
946 	zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
947 	zpci_zdev_put(zdev);
948 }
949 
950 void zpci_release_device(struct kref *kref)
951 {
952 	struct zpci_dev *zdev = container_of(kref, struct zpci_dev, kref);
953 	int ret;
954 
955 	if (zdev->zbus->bus)
956 		zpci_bus_remove_device(zdev, false);
957 
958 	if (zdev->dma_table)
959 		zpci_dma_exit_device(zdev);
960 	if (zdev_enabled(zdev))
961 		zpci_disable_device(zdev);
962 
963 	switch (zdev->state) {
964 	case ZPCI_FN_STATE_CONFIGURED:
965 		ret = sclp_pci_deconfigure(zdev->fid);
966 		zpci_dbg(3, "deconf fid:%x, rc:%d\n", zdev->fid, ret);
967 		fallthrough;
968 	case ZPCI_FN_STATE_STANDBY:
969 		if (zdev->has_hp_slot)
970 			zpci_exit_slot(zdev);
971 		spin_lock(&zpci_list_lock);
972 		list_del(&zdev->entry);
973 		spin_unlock(&zpci_list_lock);
974 		zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
975 		fallthrough;
976 	case ZPCI_FN_STATE_RESERVED:
977 		if (zdev->has_resources)
978 			zpci_cleanup_bus_resources(zdev);
979 		zpci_bus_device_unregister(zdev);
980 		zpci_destroy_iommu(zdev);
981 		fallthrough;
982 	default:
983 		break;
984 	}
985 	zpci_dbg(3, "rem fid:%x\n", zdev->fid);
986 	kfree_rcu(zdev, rcu);
987 }
988 
989 int zpci_report_error(struct pci_dev *pdev,
990 		      struct zpci_report_error_header *report)
991 {
992 	struct zpci_dev *zdev = to_zpci(pdev);
993 
994 	return sclp_pci_report(report, zdev->fh, zdev->fid);
995 }
996 EXPORT_SYMBOL(zpci_report_error);
997 
998 /**
999  * zpci_clear_error_state() - Clears the zPCI error state of the device
1000  * @zdev: The zdev for which the zPCI error state should be reset
1001  *
1002  * Clear the zPCI error state of the device. If clearing the zPCI error state
1003  * fails the device is left in the error state. In this case it may make sense
1004  * to call zpci_io_perm_failure() on the associated pdev if it exists.
1005  *
1006  * Returns: 0 on success, -EIO otherwise
1007  */
1008 int zpci_clear_error_state(struct zpci_dev *zdev)
1009 {
1010 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_RESET_ERROR);
1011 	struct zpci_fib fib = {0};
1012 	u8 status;
1013 	int cc;
1014 
1015 	cc = zpci_mod_fc(req, &fib, &status);
1016 	if (cc) {
1017 		zpci_dbg(3, "ces fid:%x, cc:%d, status:%x\n", zdev->fid, cc, status);
1018 		return -EIO;
1019 	}
1020 
1021 	return 0;
1022 }
1023 
1024 /**
1025  * zpci_reset_load_store_blocked() - Re-enables L/S from error state
1026  * @zdev: The zdev for which to unblock load/store access
1027  *
1028  * Re-enables load/store access for a PCI function in the error state while
1029  * keeping DMA blocked. In this state drivers can poke MMIO space to determine
1030  * if error recovery is possible while catching any rogue DMA access from the
1031  * device.
1032  *
1033  * Returns: 0 on success, -EIO otherwise
1034  */
1035 int zpci_reset_load_store_blocked(struct zpci_dev *zdev)
1036 {
1037 	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_RESET_BLOCK);
1038 	struct zpci_fib fib = {0};
1039 	u8 status;
1040 	int cc;
1041 
1042 	cc = zpci_mod_fc(req, &fib, &status);
1043 	if (cc) {
1044 		zpci_dbg(3, "rls fid:%x, cc:%d, status:%x\n", zdev->fid, cc, status);
1045 		return -EIO;
1046 	}
1047 
1048 	return 0;
1049 }
1050 
1051 static int zpci_mem_init(void)
1052 {
1053 	BUILD_BUG_ON(!is_power_of_2(__alignof__(struct zpci_fmb)) ||
1054 		     __alignof__(struct zpci_fmb) < sizeof(struct zpci_fmb));
1055 
1056 	zdev_fmb_cache = kmem_cache_create("PCI_FMB_cache", sizeof(struct zpci_fmb),
1057 					   __alignof__(struct zpci_fmb), 0, NULL);
1058 	if (!zdev_fmb_cache)
1059 		goto error_fmb;
1060 
1061 	zpci_iomap_start = kcalloc(ZPCI_IOMAP_ENTRIES,
1062 				   sizeof(*zpci_iomap_start), GFP_KERNEL);
1063 	if (!zpci_iomap_start)
1064 		goto error_iomap;
1065 
1066 	zpci_iomap_bitmap = kcalloc(BITS_TO_LONGS(ZPCI_IOMAP_ENTRIES),
1067 				    sizeof(*zpci_iomap_bitmap), GFP_KERNEL);
1068 	if (!zpci_iomap_bitmap)
1069 		goto error_iomap_bitmap;
1070 
1071 	if (static_branch_likely(&have_mio))
1072 		clp_setup_writeback_mio();
1073 
1074 	return 0;
1075 error_iomap_bitmap:
1076 	kfree(zpci_iomap_start);
1077 error_iomap:
1078 	kmem_cache_destroy(zdev_fmb_cache);
1079 error_fmb:
1080 	return -ENOMEM;
1081 }
1082 
1083 static void zpci_mem_exit(void)
1084 {
1085 	kfree(zpci_iomap_bitmap);
1086 	kfree(zpci_iomap_start);
1087 	kmem_cache_destroy(zdev_fmb_cache);
1088 }
1089 
1090 static unsigned int s390_pci_probe __initdata = 1;
1091 unsigned int s390_pci_force_floating __initdata;
1092 static unsigned int s390_pci_initialized;
1093 
1094 char * __init pcibios_setup(char *str)
1095 {
1096 	if (!strcmp(str, "off")) {
1097 		s390_pci_probe = 0;
1098 		return NULL;
1099 	}
1100 	if (!strcmp(str, "nomio")) {
1101 		S390_lowcore.machine_flags &= ~MACHINE_FLAG_PCI_MIO;
1102 		return NULL;
1103 	}
1104 	if (!strcmp(str, "force_floating")) {
1105 		s390_pci_force_floating = 1;
1106 		return NULL;
1107 	}
1108 	if (!strcmp(str, "norid")) {
1109 		s390_pci_no_rid = 1;
1110 		return NULL;
1111 	}
1112 	return str;
1113 }
1114 
1115 bool zpci_is_enabled(void)
1116 {
1117 	return s390_pci_initialized;
1118 }
1119 
1120 static int __init pci_base_init(void)
1121 {
1122 	int rc;
1123 
1124 	if (!s390_pci_probe)
1125 		return 0;
1126 
1127 	if (!test_facility(69) || !test_facility(71)) {
1128 		pr_info("PCI is not supported because CPU facilities 69 or 71 are not available\n");
1129 		return 0;
1130 	}
1131 
1132 	if (MACHINE_HAS_PCI_MIO) {
1133 		static_branch_enable(&have_mio);
1134 		ctl_set_bit(2, 5);
1135 	}
1136 
1137 	rc = zpci_debug_init();
1138 	if (rc)
1139 		goto out;
1140 
1141 	rc = zpci_mem_init();
1142 	if (rc)
1143 		goto out_mem;
1144 
1145 	rc = zpci_irq_init();
1146 	if (rc)
1147 		goto out_irq;
1148 
1149 	rc = zpci_dma_init();
1150 	if (rc)
1151 		goto out_dma;
1152 
1153 	rc = clp_scan_pci_devices();
1154 	if (rc)
1155 		goto out_find;
1156 	zpci_bus_scan_busses();
1157 
1158 	s390_pci_initialized = 1;
1159 	return 0;
1160 
1161 out_find:
1162 	zpci_dma_exit();
1163 out_dma:
1164 	zpci_irq_exit();
1165 out_irq:
1166 	zpci_mem_exit();
1167 out_mem:
1168 	zpci_debug_exit();
1169 out:
1170 	return rc;
1171 }
1172 subsys_initcall_sync(pci_base_init);
1173