xref: /kernel/linux/linux-5.10/arch/s390/pci/pci.c (revision 8c2ecf20)
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 */
43static LIST_HEAD(zpci_list);
44static DEFINE_SPINLOCK(zpci_list_lock);
45
46static DECLARE_BITMAP(zpci_domain, ZPCI_DOMAIN_BITMAP_SIZE);
47static 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
53unsigned int s390_pci_no_rid;
54
55static DEFINE_SPINLOCK(zpci_iomap_lock);
56static unsigned long *zpci_iomap_bitmap;
57struct zpci_iomap_entry *zpci_iomap_start;
58EXPORT_SYMBOL_GPL(zpci_iomap_start);
59
60DEFINE_STATIC_KEY_FALSE(have_mio);
61
62static struct kmem_cache *zdev_fmb_cache;
63
64struct zpci_dev *get_zdev_by_fid(u32 fid)
65{
66	struct zpci_dev *tmp, *zdev = NULL;
67
68	spin_lock(&zpci_list_lock);
69	list_for_each_entry(tmp, &zpci_list, entry) {
70		if (tmp->fid == fid) {
71			zdev = tmp;
72			zpci_zdev_get(zdev);
73			break;
74		}
75	}
76	spin_unlock(&zpci_list_lock);
77	return zdev;
78}
79
80void zpci_remove_reserved_devices(void)
81{
82	struct zpci_dev *tmp, *zdev;
83	enum zpci_state state;
84	LIST_HEAD(remove);
85
86	spin_lock(&zpci_list_lock);
87	list_for_each_entry_safe(zdev, tmp, &zpci_list, entry) {
88		if (zdev->state == ZPCI_FN_STATE_STANDBY &&
89		    !clp_get_state(zdev->fid, &state) &&
90		    state == ZPCI_FN_STATE_RESERVED)
91			list_move_tail(&zdev->entry, &remove);
92	}
93	spin_unlock(&zpci_list_lock);
94
95	list_for_each_entry_safe(zdev, tmp, &remove, entry)
96		zpci_device_reserved(zdev);
97}
98
99int pci_domain_nr(struct pci_bus *bus)
100{
101	return ((struct zpci_bus *) bus->sysdata)->domain_nr;
102}
103EXPORT_SYMBOL_GPL(pci_domain_nr);
104
105int pci_proc_domain(struct pci_bus *bus)
106{
107	return pci_domain_nr(bus);
108}
109EXPORT_SYMBOL_GPL(pci_proc_domain);
110
111/* Modify PCI: Register I/O address translation parameters */
112int zpci_register_ioat(struct zpci_dev *zdev, u8 dmaas,
113		       u64 base, u64 limit, u64 iota)
114{
115	u64 req = ZPCI_CREATE_REQ(zdev->fh, dmaas, ZPCI_MOD_FC_REG_IOAT);
116	struct zpci_fib fib = {0};
117	u8 status;
118
119	WARN_ON_ONCE(iota & 0x3fff);
120	fib.pba = base;
121	fib.pal = limit;
122	fib.iota = iota | ZPCI_IOTA_RTTO_FLAG;
123	return zpci_mod_fc(req, &fib, &status) ? -EIO : 0;
124}
125
126/* Modify PCI: Unregister I/O address translation parameters */
127int zpci_unregister_ioat(struct zpci_dev *zdev, u8 dmaas)
128{
129	u64 req = ZPCI_CREATE_REQ(zdev->fh, dmaas, ZPCI_MOD_FC_DEREG_IOAT);
130	struct zpci_fib fib = {0};
131	u8 cc, status;
132
133	cc = zpci_mod_fc(req, &fib, &status);
134	if (cc == 3) /* Function already gone. */
135		cc = 0;
136	return cc ? -EIO : 0;
137}
138
139/* Modify PCI: Set PCI function measurement parameters */
140int zpci_fmb_enable_device(struct zpci_dev *zdev)
141{
142	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_SET_MEASURE);
143	struct zpci_fib fib = {0};
144	u8 cc, status;
145
146	if (zdev->fmb || sizeof(*zdev->fmb) < zdev->fmb_length)
147		return -EINVAL;
148
149	zdev->fmb = kmem_cache_zalloc(zdev_fmb_cache, GFP_KERNEL);
150	if (!zdev->fmb)
151		return -ENOMEM;
152	WARN_ON((u64) zdev->fmb & 0xf);
153
154	/* reset software counters */
155	atomic64_set(&zdev->allocated_pages, 0);
156	atomic64_set(&zdev->mapped_pages, 0);
157	atomic64_set(&zdev->unmapped_pages, 0);
158
159	fib.fmb_addr = virt_to_phys(zdev->fmb);
160	cc = zpci_mod_fc(req, &fib, &status);
161	if (cc) {
162		kmem_cache_free(zdev_fmb_cache, zdev->fmb);
163		zdev->fmb = NULL;
164	}
165	return cc ? -EIO : 0;
166}
167
168/* Modify PCI: Disable PCI function measurement */
169int zpci_fmb_disable_device(struct zpci_dev *zdev)
170{
171	u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_SET_MEASURE);
172	struct zpci_fib fib = {0};
173	u8 cc, status;
174
175	if (!zdev->fmb)
176		return -EINVAL;
177
178	/* Function measurement is disabled if fmb address is zero */
179	cc = zpci_mod_fc(req, &fib, &status);
180	if (cc == 3) /* Function already gone. */
181		cc = 0;
182
183	if (!cc) {
184		kmem_cache_free(zdev_fmb_cache, zdev->fmb);
185		zdev->fmb = NULL;
186	}
187	return cc ? -EIO : 0;
188}
189
190static int zpci_cfg_load(struct zpci_dev *zdev, int offset, u32 *val, u8 len)
191{
192	u64 req = ZPCI_CREATE_REQ(zdev->fh, ZPCI_PCIAS_CFGSPC, len);
193	u64 data;
194	int rc;
195
196	rc = __zpci_load(&data, req, offset);
197	if (!rc) {
198		data = le64_to_cpu((__force __le64) data);
199		data >>= (8 - len) * 8;
200		*val = (u32) data;
201	} else
202		*val = 0xffffffff;
203	return rc;
204}
205
206static int zpci_cfg_store(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 = val;
210	int rc;
211
212	data <<= (8 - len) * 8;
213	data = (__force u64) cpu_to_le64(data);
214	rc = __zpci_store(data, req, offset);
215	return rc;
216}
217
218resource_size_t pcibios_align_resource(void *data, const struct resource *res,
219				       resource_size_t size,
220				       resource_size_t align)
221{
222	return 0;
223}
224
225/* combine single writes by using store-block insn */
226void __iowrite64_copy(void __iomem *to, const void *from, size_t count)
227{
228       zpci_memcpy_toio(to, from, count);
229}
230
231static void __iomem *__ioremap(phys_addr_t addr, size_t size, pgprot_t prot)
232{
233	unsigned long offset, vaddr;
234	struct vm_struct *area;
235	phys_addr_t last_addr;
236
237	last_addr = addr + size - 1;
238	if (!size || last_addr < addr)
239		return NULL;
240
241	if (!static_branch_unlikely(&have_mio))
242		return (void __iomem *) addr;
243
244	offset = addr & ~PAGE_MASK;
245	addr &= PAGE_MASK;
246	size = PAGE_ALIGN(size + offset);
247	area = get_vm_area(size, VM_IOREMAP);
248	if (!area)
249		return NULL;
250
251	vaddr = (unsigned long) area->addr;
252	if (ioremap_page_range(vaddr, vaddr + size, addr, prot)) {
253		free_vm_area(area);
254		return NULL;
255	}
256	return (void __iomem *) ((unsigned long) area->addr + offset);
257}
258
259void __iomem *ioremap_prot(phys_addr_t addr, size_t size, unsigned long prot)
260{
261	return __ioremap(addr, size, __pgprot(prot));
262}
263EXPORT_SYMBOL(ioremap_prot);
264
265void __iomem *ioremap(phys_addr_t addr, size_t size)
266{
267	return __ioremap(addr, size, PAGE_KERNEL);
268}
269EXPORT_SYMBOL(ioremap);
270
271void __iomem *ioremap_wc(phys_addr_t addr, size_t size)
272{
273	return __ioremap(addr, size, pgprot_writecombine(PAGE_KERNEL));
274}
275EXPORT_SYMBOL(ioremap_wc);
276
277void __iomem *ioremap_wt(phys_addr_t addr, size_t size)
278{
279	return __ioremap(addr, size, pgprot_writethrough(PAGE_KERNEL));
280}
281EXPORT_SYMBOL(ioremap_wt);
282
283void iounmap(volatile void __iomem *addr)
284{
285	if (static_branch_likely(&have_mio))
286		vunmap((__force void *) ((unsigned long) addr & PAGE_MASK));
287}
288EXPORT_SYMBOL(iounmap);
289
290/* Create a virtual mapping cookie for a PCI BAR */
291static void __iomem *pci_iomap_range_fh(struct pci_dev *pdev, int bar,
292					unsigned long offset, unsigned long max)
293{
294	struct zpci_dev *zdev =	to_zpci(pdev);
295	int idx;
296
297	idx = zdev->bars[bar].map_idx;
298	spin_lock(&zpci_iomap_lock);
299	/* Detect overrun */
300	WARN_ON(!++zpci_iomap_start[idx].count);
301	zpci_iomap_start[idx].fh = zdev->fh;
302	zpci_iomap_start[idx].bar = bar;
303	spin_unlock(&zpci_iomap_lock);
304
305	return (void __iomem *) ZPCI_ADDR(idx) + offset;
306}
307
308static void __iomem *pci_iomap_range_mio(struct pci_dev *pdev, int bar,
309					 unsigned long offset,
310					 unsigned long max)
311{
312	unsigned long barsize = pci_resource_len(pdev, bar);
313	struct zpci_dev *zdev = to_zpci(pdev);
314	void __iomem *iova;
315
316	iova = ioremap((unsigned long) zdev->bars[bar].mio_wt, barsize);
317	return iova ? iova + offset : iova;
318}
319
320void __iomem *pci_iomap_range(struct pci_dev *pdev, int bar,
321			      unsigned long offset, unsigned long max)
322{
323	if (bar >= PCI_STD_NUM_BARS || !pci_resource_len(pdev, bar))
324		return NULL;
325
326	if (static_branch_likely(&have_mio))
327		return pci_iomap_range_mio(pdev, bar, offset, max);
328	else
329		return pci_iomap_range_fh(pdev, bar, offset, max);
330}
331EXPORT_SYMBOL(pci_iomap_range);
332
333void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long maxlen)
334{
335	return pci_iomap_range(dev, bar, 0, maxlen);
336}
337EXPORT_SYMBOL(pci_iomap);
338
339static void __iomem *pci_iomap_wc_range_mio(struct pci_dev *pdev, int bar,
340					    unsigned long offset, unsigned long max)
341{
342	unsigned long barsize = pci_resource_len(pdev, bar);
343	struct zpci_dev *zdev = to_zpci(pdev);
344	void __iomem *iova;
345
346	iova = ioremap((unsigned long) zdev->bars[bar].mio_wb, barsize);
347	return iova ? iova + offset : iova;
348}
349
350void __iomem *pci_iomap_wc_range(struct pci_dev *pdev, int bar,
351				 unsigned long offset, unsigned long max)
352{
353	if (bar >= PCI_STD_NUM_BARS || !pci_resource_len(pdev, bar))
354		return NULL;
355
356	if (static_branch_likely(&have_mio))
357		return pci_iomap_wc_range_mio(pdev, bar, offset, max);
358	else
359		return pci_iomap_range_fh(pdev, bar, offset, max);
360}
361EXPORT_SYMBOL(pci_iomap_wc_range);
362
363void __iomem *pci_iomap_wc(struct pci_dev *dev, int bar, unsigned long maxlen)
364{
365	return pci_iomap_wc_range(dev, bar, 0, maxlen);
366}
367EXPORT_SYMBOL(pci_iomap_wc);
368
369static void pci_iounmap_fh(struct pci_dev *pdev, void __iomem *addr)
370{
371	unsigned int idx = ZPCI_IDX(addr);
372
373	spin_lock(&zpci_iomap_lock);
374	/* Detect underrun */
375	WARN_ON(!zpci_iomap_start[idx].count);
376	if (!--zpci_iomap_start[idx].count) {
377		zpci_iomap_start[idx].fh = 0;
378		zpci_iomap_start[idx].bar = 0;
379	}
380	spin_unlock(&zpci_iomap_lock);
381}
382
383static void pci_iounmap_mio(struct pci_dev *pdev, void __iomem *addr)
384{
385	iounmap(addr);
386}
387
388void pci_iounmap(struct pci_dev *pdev, void __iomem *addr)
389{
390	if (static_branch_likely(&have_mio))
391		pci_iounmap_mio(pdev, addr);
392	else
393		pci_iounmap_fh(pdev, addr);
394}
395EXPORT_SYMBOL(pci_iounmap);
396
397static int pci_read(struct pci_bus *bus, unsigned int devfn, int where,
398		    int size, u32 *val)
399{
400	struct zpci_dev *zdev = get_zdev_by_bus(bus, devfn);
401
402	return (zdev) ? zpci_cfg_load(zdev, where, val, size) : -ENODEV;
403}
404
405static int pci_write(struct pci_bus *bus, unsigned int devfn, int where,
406		     int size, u32 val)
407{
408	struct zpci_dev *zdev = get_zdev_by_bus(bus, devfn);
409
410	return (zdev) ? zpci_cfg_store(zdev, where, val, size) : -ENODEV;
411}
412
413static struct pci_ops pci_root_ops = {
414	.read = pci_read,
415	.write = pci_write,
416};
417
418static void zpci_map_resources(struct pci_dev *pdev)
419{
420	struct zpci_dev *zdev = to_zpci(pdev);
421	resource_size_t len;
422	int i;
423
424	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
425		len = pci_resource_len(pdev, i);
426		if (!len)
427			continue;
428
429		if (zpci_use_mio(zdev))
430			pdev->resource[i].start =
431				(resource_size_t __force) zdev->bars[i].mio_wt;
432		else
433			pdev->resource[i].start = (resource_size_t __force)
434				pci_iomap_range_fh(pdev, i, 0, 0);
435		pdev->resource[i].end = pdev->resource[i].start + len - 1;
436	}
437
438	zpci_iov_map_resources(pdev);
439}
440
441static void zpci_unmap_resources(struct pci_dev *pdev)
442{
443	struct zpci_dev *zdev = to_zpci(pdev);
444	resource_size_t len;
445	int i;
446
447	if (zpci_use_mio(zdev))
448		return;
449
450	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
451		len = pci_resource_len(pdev, i);
452		if (!len)
453			continue;
454		pci_iounmap_fh(pdev, (void __iomem __force *)
455			       pdev->resource[i].start);
456	}
457}
458
459static int zpci_alloc_iomap(struct zpci_dev *zdev)
460{
461	unsigned long entry;
462
463	spin_lock(&zpci_iomap_lock);
464	entry = find_first_zero_bit(zpci_iomap_bitmap, ZPCI_IOMAP_ENTRIES);
465	if (entry == ZPCI_IOMAP_ENTRIES) {
466		spin_unlock(&zpci_iomap_lock);
467		return -ENOSPC;
468	}
469	set_bit(entry, zpci_iomap_bitmap);
470	spin_unlock(&zpci_iomap_lock);
471	return entry;
472}
473
474static void zpci_free_iomap(struct zpci_dev *zdev, int entry)
475{
476	spin_lock(&zpci_iomap_lock);
477	memset(&zpci_iomap_start[entry], 0, sizeof(struct zpci_iomap_entry));
478	clear_bit(entry, zpci_iomap_bitmap);
479	spin_unlock(&zpci_iomap_lock);
480}
481
482static struct resource *__alloc_res(struct zpci_dev *zdev, unsigned long start,
483				    unsigned long size, unsigned long flags)
484{
485	struct resource *r;
486
487	r = kzalloc(sizeof(*r), GFP_KERNEL);
488	if (!r)
489		return NULL;
490
491	r->start = start;
492	r->end = r->start + size - 1;
493	r->flags = flags;
494	r->name = zdev->res_name;
495
496	if (request_resource(&iomem_resource, r)) {
497		kfree(r);
498		return NULL;
499	}
500	return r;
501}
502
503int zpci_setup_bus_resources(struct zpci_dev *zdev,
504			     struct list_head *resources)
505{
506	unsigned long addr, size, flags;
507	struct resource *res;
508	int i, entry;
509
510	snprintf(zdev->res_name, sizeof(zdev->res_name),
511		 "PCI Bus %04x:%02x", zdev->uid, ZPCI_BUS_NR);
512
513	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
514		if (!zdev->bars[i].size)
515			continue;
516		entry = zpci_alloc_iomap(zdev);
517		if (entry < 0)
518			return entry;
519		zdev->bars[i].map_idx = entry;
520
521		/* only MMIO is supported */
522		flags = IORESOURCE_MEM;
523		if (zdev->bars[i].val & 8)
524			flags |= IORESOURCE_PREFETCH;
525		if (zdev->bars[i].val & 4)
526			flags |= IORESOURCE_MEM_64;
527
528		if (zpci_use_mio(zdev))
529			addr = (unsigned long) zdev->bars[i].mio_wt;
530		else
531			addr = ZPCI_ADDR(entry);
532		size = 1UL << zdev->bars[i].size;
533
534		res = __alloc_res(zdev, addr, size, flags);
535		if (!res) {
536			zpci_free_iomap(zdev, entry);
537			return -ENOMEM;
538		}
539		zdev->bars[i].res = res;
540		pci_add_resource(resources, res);
541	}
542
543	return 0;
544}
545
546static void zpci_cleanup_bus_resources(struct zpci_dev *zdev)
547{
548	int i;
549
550	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
551		if (!zdev->bars[i].size || !zdev->bars[i].res)
552			continue;
553
554		zpci_free_iomap(zdev, zdev->bars[i].map_idx);
555		release_resource(zdev->bars[i].res);
556		kfree(zdev->bars[i].res);
557	}
558}
559
560int pcibios_add_device(struct pci_dev *pdev)
561{
562	struct zpci_dev *zdev = to_zpci(pdev);
563	struct resource *res;
564	int i;
565
566	/* The pdev has a reference to the zdev via its bus */
567	zpci_zdev_get(zdev);
568	if (pdev->is_physfn)
569		pdev->no_vf_scan = 1;
570
571	pdev->dev.groups = zpci_attr_groups;
572	pdev->dev.dma_ops = &s390_pci_dma_ops;
573	zpci_map_resources(pdev);
574
575	for (i = 0; i < PCI_STD_NUM_BARS; i++) {
576		res = &pdev->resource[i];
577		if (res->parent || !res->flags)
578			continue;
579		pci_claim_resource(pdev, i);
580	}
581
582	return 0;
583}
584
585void pcibios_release_device(struct pci_dev *pdev)
586{
587	struct zpci_dev *zdev = to_zpci(pdev);
588
589	zpci_unmap_resources(pdev);
590	zpci_zdev_put(zdev);
591}
592
593int pcibios_enable_device(struct pci_dev *pdev, int mask)
594{
595	struct zpci_dev *zdev = to_zpci(pdev);
596
597	zpci_debug_init_device(zdev, dev_name(&pdev->dev));
598	zpci_fmb_enable_device(zdev);
599
600	return pci_enable_resources(pdev, mask);
601}
602
603void pcibios_disable_device(struct pci_dev *pdev)
604{
605	struct zpci_dev *zdev = to_zpci(pdev);
606
607	zpci_fmb_disable_device(zdev);
608	zpci_debug_exit_device(zdev);
609}
610
611static int __zpci_register_domain(int domain)
612{
613	spin_lock(&zpci_domain_lock);
614	if (test_bit(domain, zpci_domain)) {
615		spin_unlock(&zpci_domain_lock);
616		pr_err("Domain %04x is already assigned\n", domain);
617		return -EEXIST;
618	}
619	set_bit(domain, zpci_domain);
620	spin_unlock(&zpci_domain_lock);
621	return domain;
622}
623
624static int __zpci_alloc_domain(void)
625{
626	int domain;
627
628	spin_lock(&zpci_domain_lock);
629	/*
630	 * We can always auto allocate domains below ZPCI_NR_DEVICES.
631	 * There is either a free domain or we have reached the maximum in
632	 * which case we would have bailed earlier.
633	 */
634	domain = find_first_zero_bit(zpci_domain, ZPCI_NR_DEVICES);
635	set_bit(domain, zpci_domain);
636	spin_unlock(&zpci_domain_lock);
637	return domain;
638}
639
640int zpci_alloc_domain(int domain)
641{
642	if (zpci_unique_uid) {
643		if (domain)
644			return __zpci_register_domain(domain);
645		pr_warn("UID checking was active but no UID is provided: switching to automatic domain allocation\n");
646		update_uid_checking(false);
647	}
648	return __zpci_alloc_domain();
649}
650
651void zpci_free_domain(int domain)
652{
653	spin_lock(&zpci_domain_lock);
654	clear_bit(domain, zpci_domain);
655	spin_unlock(&zpci_domain_lock);
656}
657
658
659int zpci_enable_device(struct zpci_dev *zdev)
660{
661	int rc;
662
663	if (clp_enable_fh(zdev, ZPCI_NR_DMA_SPACES)) {
664		rc = -EIO;
665		goto out;
666	}
667
668	rc = zpci_dma_init_device(zdev);
669	if (rc)
670		goto out_dma;
671
672	zdev->state = ZPCI_FN_STATE_ONLINE;
673	return 0;
674
675out_dma:
676	clp_disable_fh(zdev);
677out:
678	return rc;
679}
680EXPORT_SYMBOL_GPL(zpci_enable_device);
681
682int zpci_disable_device(struct zpci_dev *zdev)
683{
684	zpci_dma_exit_device(zdev);
685	/*
686	 * The zPCI function may already be disabled by the platform, this is
687	 * detected in clp_disable_fh() which becomes a no-op.
688	 */
689	return clp_disable_fh(zdev) ? -EIO : 0;
690}
691EXPORT_SYMBOL_GPL(zpci_disable_device);
692
693/* zpci_remove_device - Removes the given zdev from the PCI core
694 * @zdev: the zdev to be removed from the PCI core
695 * @set_error: if true the device's error state is set to permanent failure
696 *
697 * Sets a zPCI device to a configured but offline state; the zPCI
698 * device is still accessible through its hotplug slot and the zPCI
699 * API but is removed from the common code PCI bus, making it
700 * no longer available to drivers.
701 */
702void zpci_remove_device(struct zpci_dev *zdev, bool set_error)
703{
704	struct zpci_bus *zbus = zdev->zbus;
705	struct pci_dev *pdev;
706
707	if (!zdev->zbus->bus)
708		return;
709
710	pdev = pci_get_slot(zbus->bus, zdev->devfn);
711	if (pdev) {
712		if (set_error)
713			pdev->error_state = pci_channel_io_perm_failure;
714		if (pdev->is_virtfn) {
715			zpci_iov_remove_virtfn(pdev, zdev->vfn);
716			/* balance pci_get_slot */
717			pci_dev_put(pdev);
718			return;
719		}
720		pci_stop_and_remove_bus_device_locked(pdev);
721		/* balance pci_get_slot */
722		pci_dev_put(pdev);
723	}
724}
725
726/**
727 * zpci_create_device() - Create a new zpci_dev and add it to the zbus
728 * @fid: Function ID of the device to be created
729 * @fh: Current Function Handle of the device to be created
730 * @state: Initial state after creation either Standby or Configured
731 *
732 * Creates a new zpci device and adds it to its, possibly newly created, zbus
733 * as well as zpci_list.
734 *
735 * Returns: 0 on success, an error value otherwise
736 */
737int zpci_create_device(u32 fid, u32 fh, enum zpci_state state)
738{
739	struct zpci_dev *zdev;
740	int rc;
741
742	zpci_dbg(3, "add fid:%x, fh:%x, c:%d\n", fid, fh, state);
743	zdev = kzalloc(sizeof(*zdev), GFP_KERNEL);
744	if (!zdev)
745		return -ENOMEM;
746
747	/* FID and Function Handle are the static/dynamic identifiers */
748	zdev->fid = fid;
749	zdev->fh = fh;
750
751	/* Query function properties and update zdev */
752	rc = clp_query_pci_fn(zdev);
753	if (rc)
754		goto error;
755	zdev->state =  state;
756
757	kref_init(&zdev->kref);
758	mutex_init(&zdev->lock);
759
760	rc = zpci_init_iommu(zdev);
761	if (rc)
762		goto error;
763
764	if (zdev->state == ZPCI_FN_STATE_CONFIGURED) {
765		rc = zpci_enable_device(zdev);
766		if (rc)
767			goto error_destroy_iommu;
768	}
769
770	rc = zpci_bus_device_register(zdev, &pci_root_ops);
771	if (rc)
772		goto error_disable;
773
774	spin_lock(&zpci_list_lock);
775	list_add_tail(&zdev->entry, &zpci_list);
776	spin_unlock(&zpci_list_lock);
777
778	return 0;
779
780error_disable:
781	if (zdev->state == ZPCI_FN_STATE_ONLINE)
782		zpci_disable_device(zdev);
783error_destroy_iommu:
784	zpci_destroy_iommu(zdev);
785error:
786	zpci_dbg(0, "add fid:%x, rc:%d\n", fid, rc);
787	kfree(zdev);
788	return rc;
789}
790
791bool zpci_is_device_configured(struct zpci_dev *zdev)
792{
793	enum zpci_state state = zdev->state;
794
795	return state != ZPCI_FN_STATE_RESERVED &&
796		state != ZPCI_FN_STATE_STANDBY;
797}
798
799/**
800 * zpci_device_reserved() - Mark device as resverved
801 * @zdev: the zpci_dev that was reserved
802 *
803 * Handle the case that a given zPCI function was reserved by another system.
804 * After a call to this function the zpci_dev can not be found via
805 * get_zdev_by_fid() anymore but may still be accessible via existing
806 * references though it will not be functional anymore.
807 */
808void zpci_device_reserved(struct zpci_dev *zdev)
809{
810	if (zdev->has_hp_slot)
811		zpci_exit_slot(zdev);
812	/*
813	 * Remove device from zpci_list as it is going away. This also
814	 * makes sure we ignore subsequent zPCI events for this device.
815	 */
816	spin_lock(&zpci_list_lock);
817	list_del(&zdev->entry);
818	spin_unlock(&zpci_list_lock);
819	zdev->state = ZPCI_FN_STATE_RESERVED;
820	zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
821	zpci_zdev_put(zdev);
822}
823
824void zpci_release_device(struct kref *kref)
825{
826	struct zpci_dev *zdev = container_of(kref, struct zpci_dev, kref);
827
828	if (zdev->zbus->bus)
829		zpci_remove_device(zdev, false);
830
831	switch (zdev->state) {
832	case ZPCI_FN_STATE_ONLINE:
833	case ZPCI_FN_STATE_CONFIGURED:
834		zpci_disable_device(zdev);
835		fallthrough;
836	case ZPCI_FN_STATE_STANDBY:
837		if (zdev->has_hp_slot)
838			zpci_exit_slot(zdev);
839		spin_lock(&zpci_list_lock);
840		list_del(&zdev->entry);
841		spin_unlock(&zpci_list_lock);
842		zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
843		fallthrough;
844	case ZPCI_FN_STATE_RESERVED:
845		zpci_cleanup_bus_resources(zdev);
846		zpci_bus_device_unregister(zdev);
847		zpci_destroy_iommu(zdev);
848		fallthrough;
849	default:
850		break;
851	}
852	zpci_dbg(3, "rem fid:%x\n", zdev->fid);
853	kfree(zdev);
854}
855
856int zpci_report_error(struct pci_dev *pdev,
857		      struct zpci_report_error_header *report)
858{
859	struct zpci_dev *zdev = to_zpci(pdev);
860
861	return sclp_pci_report(report, zdev->fh, zdev->fid);
862}
863EXPORT_SYMBOL(zpci_report_error);
864
865static int zpci_mem_init(void)
866{
867	BUILD_BUG_ON(!is_power_of_2(__alignof__(struct zpci_fmb)) ||
868		     __alignof__(struct zpci_fmb) < sizeof(struct zpci_fmb));
869
870	zdev_fmb_cache = kmem_cache_create("PCI_FMB_cache", sizeof(struct zpci_fmb),
871					   __alignof__(struct zpci_fmb), 0, NULL);
872	if (!zdev_fmb_cache)
873		goto error_fmb;
874
875	zpci_iomap_start = kcalloc(ZPCI_IOMAP_ENTRIES,
876				   sizeof(*zpci_iomap_start), GFP_KERNEL);
877	if (!zpci_iomap_start)
878		goto error_iomap;
879
880	zpci_iomap_bitmap = kcalloc(BITS_TO_LONGS(ZPCI_IOMAP_ENTRIES),
881				    sizeof(*zpci_iomap_bitmap), GFP_KERNEL);
882	if (!zpci_iomap_bitmap)
883		goto error_iomap_bitmap;
884
885	if (static_branch_likely(&have_mio))
886		clp_setup_writeback_mio();
887
888	return 0;
889error_iomap_bitmap:
890	kfree(zpci_iomap_start);
891error_iomap:
892	kmem_cache_destroy(zdev_fmb_cache);
893error_fmb:
894	return -ENOMEM;
895}
896
897static void zpci_mem_exit(void)
898{
899	kfree(zpci_iomap_bitmap);
900	kfree(zpci_iomap_start);
901	kmem_cache_destroy(zdev_fmb_cache);
902}
903
904static unsigned int s390_pci_probe __initdata = 1;
905unsigned int s390_pci_force_floating __initdata;
906static unsigned int s390_pci_initialized;
907
908char * __init pcibios_setup(char *str)
909{
910	if (!strcmp(str, "off")) {
911		s390_pci_probe = 0;
912		return NULL;
913	}
914	if (!strcmp(str, "nomio")) {
915		S390_lowcore.machine_flags &= ~MACHINE_FLAG_PCI_MIO;
916		return NULL;
917	}
918	if (!strcmp(str, "force_floating")) {
919		s390_pci_force_floating = 1;
920		return NULL;
921	}
922	if (!strcmp(str, "norid")) {
923		s390_pci_no_rid = 1;
924		return NULL;
925	}
926	return str;
927}
928
929bool zpci_is_enabled(void)
930{
931	return s390_pci_initialized;
932}
933
934static int __init pci_base_init(void)
935{
936	int rc;
937
938	if (!s390_pci_probe)
939		return 0;
940
941	if (!test_facility(69) || !test_facility(71))
942		return 0;
943
944	if (MACHINE_HAS_PCI_MIO) {
945		static_branch_enable(&have_mio);
946		ctl_set_bit(2, 5);
947	}
948
949	rc = zpci_debug_init();
950	if (rc)
951		goto out;
952
953	rc = zpci_mem_init();
954	if (rc)
955		goto out_mem;
956
957	rc = zpci_irq_init();
958	if (rc)
959		goto out_irq;
960
961	rc = zpci_dma_init();
962	if (rc)
963		goto out_dma;
964
965	rc = clp_scan_pci_devices();
966	if (rc)
967		goto out_find;
968
969	s390_pci_initialized = 1;
970	return 0;
971
972out_find:
973	zpci_dma_exit();
974out_dma:
975	zpci_irq_exit();
976out_irq:
977	zpci_mem_exit();
978out_mem:
979	zpci_debug_exit();
980out:
981	return rc;
982}
983subsys_initcall_sync(pci_base_init);
984