[mirror_ubuntu-artful-kernel.git] / arch / s390 / pci / pci_dma.c

/*
 * Copyright IBM Corp. 2012
 *
 * Author(s):
 *   Jan Glauber <jang@linux.vnet.ibm.com>
 */

#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/export.h>
#include <linux/iommu-helper.h>
#include <linux/dma-mapping.h>
#include <linux/vmalloc.h>
#include <linux/pci.h>
#include <asm/pci_dma.h>

static struct kmem_cache *dma_region_table_cache;
static struct kmem_cache *dma_page_table_cache;
static int s390_iommu_strict;

static int zpci_refresh_global(struct zpci_dev *zdev)
{
	return zpci_refresh_trans((u64) zdev->fh << 32, zdev->start_dma,
				  zdev->iommu_pages * PAGE_SIZE);
}

unsigned long *dma_alloc_cpu_table(void)
{
	unsigned long *table, *entry;

	table = kmem_cache_alloc(dma_region_table_cache, GFP_ATOMIC);
	if (!table)
		return NULL;

	for (entry = table; entry < table + ZPCI_TABLE_ENTRIES; entry++)
		*entry = ZPCI_TABLE_INVALID;
	return table;
}

static void dma_free_cpu_table(void *table)
{
	kmem_cache_free(dma_region_table_cache, table);
}

static unsigned long *dma_alloc_page_table(void)
{
	unsigned long *table, *entry;

	table = kmem_cache_alloc(dma_page_table_cache, GFP_ATOMIC);
	if (!table)
		return NULL;

	for (entry = table; entry < table + ZPCI_PT_ENTRIES; entry++)
		*entry = ZPCI_PTE_INVALID;
	return table;
}

static void dma_free_page_table(void *table)
{
	kmem_cache_free(dma_page_table_cache, table);
}

static unsigned long *dma_get_seg_table_origin(unsigned long *entry)
{
	unsigned long *sto;

	if (reg_entry_isvalid(*entry))
		sto = get_rt_sto(*entry);
	else {
		sto = dma_alloc_cpu_table();
		if (!sto)
			return NULL;

		set_rt_sto(entry, sto);
		validate_rt_entry(entry);
		entry_clr_protected(entry);
	}
	return sto;
}

static unsigned long *dma_get_page_table_origin(unsigned long *entry)
{
	unsigned long *pto;

	if (reg_entry_isvalid(*entry))
		pto = get_st_pto(*entry);
	else {
		pto = dma_alloc_page_table();
		if (!pto)
			return NULL;
		set_st_pto(entry, pto);
		validate_st_entry(entry);
		entry_clr_protected(entry);
	}
	return pto;
}

unsigned long *dma_walk_cpu_trans(unsigned long *rto, dma_addr_t dma_addr)
{
	unsigned long *sto, *pto;
	unsigned int rtx, sx, px;

	rtx = calc_rtx(dma_addr);
	sto = dma_get_seg_table_origin(&rto[rtx]);
	if (!sto)
		return NULL;

	sx = calc_sx(dma_addr);
	pto = dma_get_page_table_origin(&sto[sx]);
	if (!pto)
		return NULL;

	px = calc_px(dma_addr);
	return &pto[px];
}

void dma_update_cpu_trans(unsigned long *entry, void *page_addr, int flags)
{
	if (flags & ZPCI_PTE_INVALID) {
		invalidate_pt_entry(entry);
	} else {
		set_pt_pfaa(entry, page_addr);
		validate_pt_entry(entry);
	}

	if (flags & ZPCI_TABLE_PROTECTED)
		entry_set_protected(entry);
	else
		entry_clr_protected(entry);
}

static int dma_update_trans(struct zpci_dev *zdev, unsigned long pa,
			    dma_addr_t dma_addr, size_t size, int flags)
{
	unsigned int nr_pages = PAGE_ALIGN(size) >> PAGE_SHIFT;
	u8 *page_addr = (u8 *) (pa & PAGE_MASK);
	dma_addr_t start_dma_addr = dma_addr;
	unsigned long irq_flags;
	unsigned long *entry;
	int i, rc = 0;

	if (!nr_pages)
		return -EINVAL;

	spin_lock_irqsave(&zdev->dma_table_lock, irq_flags);
	if (!zdev->dma_table) {
		rc = -EINVAL;
		goto no_refresh;
	}

	for (i = 0; i < nr_pages; i++) {
		entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr);
		if (!entry) {
			rc = -ENOMEM;
			goto undo_cpu_trans;
		}
		dma_update_cpu_trans(entry, page_addr, flags);
		page_addr += PAGE_SIZE;
		dma_addr += PAGE_SIZE;
	}

	/*
	 * With zdev->tlb_refresh == 0, rpcit is not required to establish new
	 * translations when previously invalid translation-table entries are
	 * validated. With lazy unmap, it also is skipped for previously valid
	 * entries, but a global rpcit is then required before any address can
	 * be re-used, i.e. after each iommu bitmap wrap-around.
	 */
	if (!zdev->tlb_refresh &&
			(!s390_iommu_strict ||
			((flags & ZPCI_PTE_VALID_MASK) == ZPCI_PTE_VALID)))
		goto no_refresh;

	rc = zpci_refresh_trans((u64) zdev->fh << 32, start_dma_addr,
				nr_pages * PAGE_SIZE);
undo_cpu_trans:
	if (rc && ((flags & ZPCI_PTE_VALID_MASK) == ZPCI_PTE_VALID)) {
		flags = ZPCI_PTE_INVALID;
		while (i-- > 0) {
			page_addr -= PAGE_SIZE;
			dma_addr -= PAGE_SIZE;
			entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr);
			if (!entry)
				break;
			dma_update_cpu_trans(entry, page_addr, flags);
		}
	}

no_refresh:
	spin_unlock_irqrestore(&zdev->dma_table_lock, irq_flags);
	return rc;
}

void dma_free_seg_table(unsigned long entry)
{
	unsigned long *sto = get_rt_sto(entry);
	int sx;

	for (sx = 0; sx < ZPCI_TABLE_ENTRIES; sx++)
		if (reg_entry_isvalid(sto[sx]))
			dma_free_page_table(get_st_pto(sto[sx]));

	dma_free_cpu_table(sto);
}

void dma_cleanup_tables(unsigned long *table)
{
	int rtx;

	if (!table)
		return;

	for (rtx = 0; rtx < ZPCI_TABLE_ENTRIES; rtx++)
		if (reg_entry_isvalid(table[rtx]))
			dma_free_seg_table(table[rtx]);

	dma_free_cpu_table(table);
}

static unsigned long __dma_alloc_iommu(struct device *dev,
				       unsigned long start, int size)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
	unsigned long boundary_size;

	boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
			      PAGE_SIZE) >> PAGE_SHIFT;
	return iommu_area_alloc(zdev->iommu_bitmap, zdev->iommu_pages,
				start, size, zdev->start_dma >> PAGE_SHIFT,
				boundary_size, 0);
}

static unsigned long dma_alloc_iommu(struct device *dev, int size)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
	unsigned long offset, flags;
	int wrap = 0;

	spin_lock_irqsave(&zdev->iommu_bitmap_lock, flags);
	offset = __dma_alloc_iommu(dev, zdev->next_bit, size);
	if (offset == -1) {
		/* wrap-around */
		offset = __dma_alloc_iommu(dev, 0, size);
		wrap = 1;
	}

	if (offset != -1) {
		zdev->next_bit = offset + size;
		if (!zdev->tlb_refresh && !s390_iommu_strict && wrap)
			/* global flush after wrap-around with lazy unmap */
			zpci_refresh_global(zdev);
	}
	spin_unlock_irqrestore(&zdev->iommu_bitmap_lock, flags);
	return offset;
}

static void dma_free_iommu(struct device *dev, unsigned long offset, int size)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
	unsigned long flags;

	spin_lock_irqsave(&zdev->iommu_bitmap_lock, flags);
	if (!zdev->iommu_bitmap)
		goto out;
	bitmap_clear(zdev->iommu_bitmap, offset, size);
	/*
	 * Lazy flush for unmap: need to move next_bit to avoid address re-use
	 * until wrap-around.
	 */
	if (!s390_iommu_strict && offset >= zdev->next_bit)
		zdev->next_bit = offset + size;
out:
	spin_unlock_irqrestore(&zdev->iommu_bitmap_lock, flags);
}

static inline void zpci_err_dma(unsigned long rc, unsigned long addr)
{
	struct {
		unsigned long rc;
		unsigned long addr;
	} __packed data = {rc, addr};

	zpci_err_hex(&data, sizeof(data));
}

static dma_addr_t s390_dma_map_pages(struct device *dev, struct page *page,
				     unsigned long offset, size_t size,
				     enum dma_data_direction direction,
				     unsigned long attrs)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
	unsigned long nr_pages, iommu_page_index;
	unsigned long pa = page_to_phys(page) + offset;
	int flags = ZPCI_PTE_VALID;
	dma_addr_t dma_addr;
	int ret;

	/* This rounds up number of pages based on size and offset */
	nr_pages = iommu_num_pages(pa, size, PAGE_SIZE);
	iommu_page_index = dma_alloc_iommu(dev, nr_pages);
	if (iommu_page_index == -1) {
		ret = -ENOSPC;
		goto out_err;
	}

	/* Use rounded up size */
	size = nr_pages * PAGE_SIZE;

	dma_addr = zdev->start_dma + iommu_page_index * PAGE_SIZE;
	if (dma_addr + size > zdev->end_dma) {
		ret = -ERANGE;
		goto out_free;
	}

	if (direction == DMA_NONE || direction == DMA_TO_DEVICE)
		flags |= ZPCI_TABLE_PROTECTED;

	ret = dma_update_trans(zdev, pa, dma_addr, size, flags);
	if (ret)
		goto out_free;

	atomic64_add(nr_pages, &zdev->mapped_pages);
	return dma_addr + (offset & ~PAGE_MASK);

out_free:
	dma_free_iommu(dev, iommu_page_index, nr_pages);
out_err:
	zpci_err("map error:\n");
	zpci_err_dma(ret, pa);
	return DMA_ERROR_CODE;
}

static void s390_dma_unmap_pages(struct device *dev, dma_addr_t dma_addr,
				 size_t size, enum dma_data_direction direction,
				 unsigned long attrs)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
	unsigned long iommu_page_index;
	int npages, ret;

	npages = iommu_num_pages(dma_addr, size, PAGE_SIZE);
	dma_addr = dma_addr & PAGE_MASK;
	ret = dma_update_trans(zdev, 0, dma_addr, npages * PAGE_SIZE,
			       ZPCI_PTE_INVALID);
	if (ret) {
		zpci_err("unmap error:\n");
		zpci_err_dma(ret, dma_addr);
		return;
	}

	atomic64_add(npages, &zdev->unmapped_pages);
	iommu_page_index = (dma_addr - zdev->start_dma) >> PAGE_SHIFT;
	dma_free_iommu(dev, iommu_page_index, npages);
}

static void *s390_dma_alloc(struct device *dev, size_t size,
			    dma_addr_t *dma_handle, gfp_t flag,
			    unsigned long attrs)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
	struct page *page;
	unsigned long pa;
	dma_addr_t map;

	size = PAGE_ALIGN(size);
	page = alloc_pages(flag, get_order(size));
	if (!page)
		return NULL;

	pa = page_to_phys(page);
	memset((void *) pa, 0, size);

	map = s390_dma_map_pages(dev, page, 0, size, DMA_BIDIRECTIONAL, 0);
	if (dma_mapping_error(dev, map)) {
		free_pages(pa, get_order(size));
		return NULL;
	}

	atomic64_add(size / PAGE_SIZE, &zdev->allocated_pages);
	if (dma_handle)
		*dma_handle = map;
	return (void *) pa;
}

static void s390_dma_free(struct device *dev, size_t size,
			  void *pa, dma_addr_t dma_handle,
			  unsigned long attrs)
{
	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));

	size = PAGE_ALIGN(size);
	atomic64_sub(size / PAGE_SIZE, &zdev->allocated_pages);
	s390_dma_unmap_pages(dev, dma_handle, size, DMA_BIDIRECTIONAL, 0);
	free_pages((unsigned long) pa, get_order(size));
}

static int s390_dma_map_sg(struct device *dev, struct scatterlist *sg,
			   int nr_elements, enum dma_data_direction dir,
			   unsigned long attrs)
{
	int mapped_elements = 0;
	struct scatterlist *s;
	int i;

	for_each_sg(sg, s, nr_elements, i) {
		struct page *page = sg_page(s);
		s->dma_address = s390_dma_map_pages(dev, page, s->offset,
						    s->length, dir, 0);
		if (!dma_mapping_error(dev, s->dma_address)) {
			s->dma_length = s->length;
			mapped_elements++;
		} else
			goto unmap;
	}
out:
	return mapped_elements;

unmap:
	for_each_sg(sg, s, mapped_elements, i) {
		if (s->dma_address)
			s390_dma_unmap_pages(dev, s->dma_address, s->dma_length,
					     dir, 0);
		s->dma_address = 0;
		s->dma_length = 0;
	}
	mapped_elements = 0;
	goto out;
}

static void s390_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
			      int nr_elements, enum dma_data_direction dir,
			      unsigned long attrs)
{
	struct scatterlist *s;
	int i;

	for_each_sg(sg, s, nr_elements, i) {
		s390_dma_unmap_pages(dev, s->dma_address, s->dma_length, dir,
				     0);
		s->dma_address = 0;
		s->dma_length = 0;
	}
}

int zpci_dma_init_device(struct zpci_dev *zdev)
{
	int rc;

	/*
	 * At this point, if the device is part of an IOMMU domain, this would
	 * be a strong hint towards a bug in the IOMMU API (common) code and/or
	 * simultaneous access via IOMMU and DMA API. So let's issue a warning.
	 */
	WARN_ON(zdev->s390_domain);

	spin_lock_init(&zdev->iommu_bitmap_lock);
	spin_lock_init(&zdev->dma_table_lock);

	zdev->dma_table = dma_alloc_cpu_table();
	if (!zdev->dma_table) {
		rc = -ENOMEM;
		goto out;
	}

	/*
	 * Restrict the iommu bitmap size to the minimum of the following:
	 * - main memory size
	 * - 3-level pagetable address limit minus start_dma offset
	 * - DMA address range allowed by the hardware (clp query pci fn)
	 *
	 * Also set zdev->end_dma to the actual end address of the usable
	 * range, instead of the theoretical maximum as reported by hardware.
	 */
	zdev->start_dma = PAGE_ALIGN(zdev->start_dma);
	zdev->iommu_size = min3((u64) high_memory,
				ZPCI_TABLE_SIZE_RT - zdev->start_dma,
				zdev->end_dma - zdev->start_dma + 1);
	zdev->end_dma = zdev->start_dma + zdev->iommu_size - 1;
	zdev->iommu_pages = zdev->iommu_size >> PAGE_SHIFT;
	zdev->iommu_bitmap = vzalloc(zdev->iommu_pages / 8);
	if (!zdev->iommu_bitmap) {
		rc = -ENOMEM;
		goto free_dma_table;
	}

	rc = zpci_register_ioat(zdev, 0, zdev->start_dma, zdev->end_dma,
				(u64) zdev->dma_table);
	if (rc)
		goto free_bitmap;

	return 0;
free_bitmap:
	vfree(zdev->iommu_bitmap);
	zdev->iommu_bitmap = NULL;
free_dma_table:
	dma_free_cpu_table(zdev->dma_table);
	zdev->dma_table = NULL;
out:
	return rc;
}

void zpci_dma_exit_device(struct zpci_dev *zdev)
{
	/*
	 * At this point, if the device is part of an IOMMU domain, this would
	 * be a strong hint towards a bug in the IOMMU API (common) code and/or
	 * simultaneous access via IOMMU and DMA API. So let's issue a warning.
	 */
	WARN_ON(zdev->s390_domain);

	zpci_unregister_ioat(zdev, 0);
	dma_cleanup_tables(zdev->dma_table);
	zdev->dma_table = NULL;
	vfree(zdev->iommu_bitmap);
	zdev->iommu_bitmap = NULL;
	zdev->next_bit = 0;
}

static int __init dma_alloc_cpu_table_caches(void)
{
	dma_region_table_cache = kmem_cache_create("PCI_DMA_region_tables",
					ZPCI_TABLE_SIZE, ZPCI_TABLE_ALIGN,
					0, NULL);
	if (!dma_region_table_cache)
		return -ENOMEM;

	dma_page_table_cache = kmem_cache_create("PCI_DMA_page_tables",
					ZPCI_PT_SIZE, ZPCI_PT_ALIGN,
					0, NULL);
	if (!dma_page_table_cache) {
		kmem_cache_destroy(dma_region_table_cache);
		return -ENOMEM;
	}
	return 0;
}

int __init zpci_dma_init(void)
{
	return dma_alloc_cpu_table_caches();
}

void zpci_dma_exit(void)
{
	kmem_cache_destroy(dma_page_table_cache);
	kmem_cache_destroy(dma_region_table_cache);
}

#define PREALLOC_DMA_DEBUG_ENTRIES	(1 << 16)

static int __init dma_debug_do_init(void)
{
	dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
	return 0;
}
fs_initcall(dma_debug_do_init);

struct dma_map_ops s390_pci_dma_ops = {
	.alloc		= s390_dma_alloc,
	.free		= s390_dma_free,
	.map_sg		= s390_dma_map_sg,
	.unmap_sg	= s390_dma_unmap_sg,
	.map_page	= s390_dma_map_pages,
	.unmap_page	= s390_dma_unmap_pages,
	/* if we support direct DMA this must be conditional */
	.is_phys	= 0,
	/* dma_supported is unconditionally true without a callback */
};
EXPORT_SYMBOL_GPL(s390_pci_dma_ops);

static int __init s390_iommu_setup(char *str)
{
	if (!strncmp(str, "strict", 6))
		s390_iommu_strict = 1;
	return 0;
}

__setup("s390_iommu=", s390_iommu_setup);
Commit	Line	Data
828b35f6 JG	1	/*
	2	* Copyright IBM Corp. 2012
	3	*
	4	* Author(s):
	5	* Jan Glauber <jang@linux.vnet.ibm.com>
	6	*/
	7
	8	#include <linux/kernel.h>
	9	#include <linux/slab.h>
	10	#include <linux/export.h>
	11	#include <linux/iommu-helper.h>
	12	#include <linux/dma-mapping.h>
22459321	13	#include <linux/vmalloc.h>
828b35f6 JG	14	#include <linux/pci.h>
	15	#include <asm/pci_dma.h>
	16
828b35f6 JG	17	static struct kmem_cache *dma_region_table_cache;
828b35f6 JG	18	static struct kmem_cache *dma_page_table_cache;
c60d1ae4 GS	19	static int s390_iommu_strict;
	20
	21	static int zpci_refresh_global(struct zpci_dev *zdev)
	22	{
	23	return zpci_refresh_trans((u64) zdev->fh << 32, zdev->start_dma,
	24	zdev->iommu_pages * PAGE_SIZE);
	25	}
828b35f6	26
8128f23c	27	unsigned long *dma_alloc_cpu_table(void)
828b35f6 JG	28	{
	29	unsigned long table, entry;
	30
	31	table = kmem_cache_alloc(dma_region_table_cache, GFP_ATOMIC);
	32	if (!table)
	33	return NULL;
	34
	35	for (entry = table; entry < table + ZPCI_TABLE_ENTRIES; entry++)
4d5a6b72	36	*entry = ZPCI_TABLE_INVALID;
828b35f6 JG	37	return table;
	38	}
	39
	40	static void dma_free_cpu_table(void *table)
	41	{
	42	kmem_cache_free(dma_region_table_cache, table);
	43	}
	44
	45	static unsigned long *dma_alloc_page_table(void)
	46	{
	47	unsigned long table, entry;
	48
	49	table = kmem_cache_alloc(dma_page_table_cache, GFP_ATOMIC);
	50	if (!table)
	51	return NULL;
	52
	53	for (entry = table; entry < table + ZPCI_PT_ENTRIES; entry++)
4d5a6b72	54	*entry = ZPCI_PTE_INVALID;
828b35f6 JG	55	return table;
	56	}
	57
	58	static void dma_free_page_table(void *table)
	59	{
	60	kmem_cache_free(dma_page_table_cache, table);
	61	}
	62
	63	static unsigned long dma_get_seg_table_origin(unsigned long entry)
	64	{
	65	unsigned long *sto;
	66
	67	if (reg_entry_isvalid(*entry))
	68	sto = get_rt_sto(*entry);
	69	else {
	70	sto = dma_alloc_cpu_table();
	71	if (!sto)
	72	return NULL;
	73
	74	set_rt_sto(entry, sto);
	75	validate_rt_entry(entry);
	76	entry_clr_protected(entry);
	77	}
	78	return sto;
	79	}
	80
	81	static unsigned long dma_get_page_table_origin(unsigned long entry)
	82	{
	83	unsigned long *pto;
	84
	85	if (reg_entry_isvalid(*entry))
	86	pto = get_st_pto(*entry);
	87	else {
	88	pto = dma_alloc_page_table();
	89	if (!pto)
	90	return NULL;
	91	set_st_pto(entry, pto);
	92	validate_st_entry(entry);
	93	entry_clr_protected(entry);
	94	}
	95	return pto;
	96	}
	97
66728eee	98	unsigned long dma_walk_cpu_trans(unsigned long rto, dma_addr_t dma_addr)
828b35f6 JG	99	{
	100	unsigned long sto, pto;
	101	unsigned int rtx, sx, px;
	102
	103	rtx = calc_rtx(dma_addr);
	104	sto = dma_get_seg_table_origin(&rto[rtx]);
	105	if (!sto)
	106	return NULL;
	107
	108	sx = calc_sx(dma_addr);
	109	pto = dma_get_page_table_origin(&sto[sx]);
	110	if (!pto)
	111	return NULL;
	112
	113	px = calc_px(dma_addr);
	114	return &pto[px];
	115	}
	116
66728eee	117	void dma_update_cpu_trans(unsigned long entry, void page_addr, int flags)
828b35f6	118	{
828b35f6 JG	119	if (flags & ZPCI_PTE_INVALID) {
828b35f6 JG	120	invalidate_pt_entry(entry);
828b35f6 JG	121	} else {
	122	set_pt_pfaa(entry, page_addr);
	123	validate_pt_entry(entry);
	124	}
	125
	126	if (flags & ZPCI_TABLE_PROTECTED)
	127	entry_set_protected(entry);
	128	else
	129	entry_clr_protected(entry);
	130	}
	131
	132	static int dma_update_trans(struct zpci_dev *zdev, unsigned long pa,
	133	dma_addr_t dma_addr, size_t size, int flags)
	134	{
	135	unsigned int nr_pages = PAGE_ALIGN(size) >> PAGE_SHIFT;
	136	u8 page_addr = (u8 ) (pa & PAGE_MASK);
	137	dma_addr_t start_dma_addr = dma_addr;
	138	unsigned long irq_flags;
66728eee	139	unsigned long *entry;
828b35f6 JG	140	int i, rc = 0;
	141
	142	if (!nr_pages)
	143	return -EINVAL;
	144
	145	spin_lock_irqsave(&zdev->dma_table_lock, irq_flags);
66728eee SO	146	if (!zdev->dma_table) {
66728eee SO	147	rc = -EINVAL;
828b35f6	148	goto no_refresh;
66728eee	149	}
828b35f6 JG	150
828b35f6 JG	151	for (i = 0; i < nr_pages; i++) {
66728eee SO	152	entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr);
	153	if (!entry) {
	154	rc = -ENOMEM;
	155	goto undo_cpu_trans;
	156	}
	157	dma_update_cpu_trans(entry, page_addr, flags);
828b35f6 JG	158	page_addr += PAGE_SIZE;
	159	dma_addr += PAGE_SIZE;
	160	}
	161
	162	/*
c60d1ae4 GS	163	* With zdev->tlb_refresh == 0, rpcit is not required to establish new
	164	* translations when previously invalid translation-table entries are
	165	* validated. With lazy unmap, it also is skipped for previously valid
	166	* entries, but a global rpcit is then required before any address can
	167	* be re-used, i.e. after each iommu bitmap wrap-around.
828b35f6 JG	168	*/
828b35f6 JG	169	if (!zdev->tlb_refresh &&
c60d1ae4 GS	170	(!s390_iommu_strict \|\|
c60d1ae4 GS	171	((flags & ZPCI_PTE_VALID_MASK) == ZPCI_PTE_VALID)))
828b35f6	172	goto no_refresh;
b2a9e87d	173
9389339f MS	174	rc = zpci_refresh_trans((u64) zdev->fh << 32, start_dma_addr,
9389339f MS	175	nr_pages * PAGE_SIZE);
66728eee SO	176	undo_cpu_trans:
	177	if (rc && ((flags & ZPCI_PTE_VALID_MASK) == ZPCI_PTE_VALID)) {
	178	flags = ZPCI_PTE_INVALID;
	179	while (i-- > 0) {
	180	page_addr -= PAGE_SIZE;
	181	dma_addr -= PAGE_SIZE;
	182	entry = dma_walk_cpu_trans(zdev->dma_table, dma_addr);
	183	if (!entry)
	184	break;
	185	dma_update_cpu_trans(entry, page_addr, flags);
	186	}
	187	}
828b35f6 JG	188
	189	no_refresh:
	190	spin_unlock_irqrestore(&zdev->dma_table_lock, irq_flags);
	191	return rc;
	192	}
	193
8128f23c	194	void dma_free_seg_table(unsigned long entry)
828b35f6 JG	195	{
	196	unsigned long *sto = get_rt_sto(entry);
	197	int sx;
	198
	199	for (sx = 0; sx < ZPCI_TABLE_ENTRIES; sx++)
	200	if (reg_entry_isvalid(sto[sx]))
	201	dma_free_page_table(get_st_pto(sto[sx]));
	202
	203	dma_free_cpu_table(sto);
	204	}
	205
8128f23c	206	void dma_cleanup_tables(unsigned long *table)
828b35f6	207	{
828b35f6 JG	208	int rtx;
828b35f6 JG	209
8128f23c	210	if (!table)
828b35f6 JG	211	return;
	212
	213	for (rtx = 0; rtx < ZPCI_TABLE_ENTRIES; rtx++)
	214	if (reg_entry_isvalid(table[rtx]))
	215	dma_free_seg_table(table[rtx]);
	216
	217	dma_free_cpu_table(table);
828b35f6 JG	218	}
828b35f6 JG	219
9a99649f	220	static unsigned long __dma_alloc_iommu(struct device *dev,
5ec6d491	221	unsigned long start, int size)
828b35f6	222	{
9a99649f	223	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
5ec6d491	224	unsigned long boundary_size;
828b35f6	225
9a99649f	226	boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
5ec6d491	227	PAGE_SIZE) >> PAGE_SHIFT;
828b35f6	228	return iommu_area_alloc(zdev->iommu_bitmap, zdev->iommu_pages,
8ee2db3c SO	229	start, size, zdev->start_dma >> PAGE_SHIFT,
8ee2db3c SO	230	boundary_size, 0);
828b35f6 JG	231	}
828b35f6 JG	232
9a99649f	233	static unsigned long dma_alloc_iommu(struct device *dev, int size)
828b35f6	234	{
9a99649f	235	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
828b35f6	236	unsigned long offset, flags;
c60d1ae4	237	int wrap = 0;
828b35f6 JG	238
828b35f6 JG	239	spin_lock_irqsave(&zdev->iommu_bitmap_lock, flags);
9a99649f	240	offset = __dma_alloc_iommu(dev, zdev->next_bit, size);
c60d1ae4 GS	241	if (offset == -1) {
c60d1ae4 GS	242	/* wrap-around */
9a99649f	243	offset = __dma_alloc_iommu(dev, 0, size);
c60d1ae4 GS	244	wrap = 1;
c60d1ae4 GS	245	}
828b35f6 JG	246
	247	if (offset != -1) {
	248	zdev->next_bit = offset + size;
c60d1ae4 GS	249	if (!zdev->tlb_refresh && !s390_iommu_strict && wrap)
	250	/* global flush after wrap-around with lazy unmap */
	251	zpci_refresh_global(zdev);
828b35f6 JG	252	}
	253	spin_unlock_irqrestore(&zdev->iommu_bitmap_lock, flags);
	254	return offset;
	255	}
	256
9a99649f	257	static void dma_free_iommu(struct device *dev, unsigned long offset, int size)
828b35f6	258	{
9a99649f	259	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
828b35f6 JG	260	unsigned long flags;
	261
	262	spin_lock_irqsave(&zdev->iommu_bitmap_lock, flags);
	263	if (!zdev->iommu_bitmap)
	264	goto out;
	265	bitmap_clear(zdev->iommu_bitmap, offset, size);
c60d1ae4 GS	266	/*
	267	* Lazy flush for unmap: need to move next_bit to avoid address re-use
	268	* until wrap-around.
	269	*/
	270	if (!s390_iommu_strict && offset >= zdev->next_bit)
828b35f6 JG	271	zdev->next_bit = offset + size;
	272	out:
	273	spin_unlock_irqrestore(&zdev->iommu_bitmap_lock, flags);
	274	}
	275
52d43d81 SO	276	static inline void zpci_err_dma(unsigned long rc, unsigned long addr)
	277	{
	278	struct {
	279	unsigned long rc;
	280	unsigned long addr;
	281	} __packed data = {rc, addr};
	282
	283	zpci_err_hex(&data, sizeof(data));
	284	}
	285
828b35f6 JG	286	static dma_addr_t s390_dma_map_pages(struct device dev, struct page page,
	287	unsigned long offset, size_t size,
	288	enum dma_data_direction direction,
00085f1e	289	unsigned long attrs)
828b35f6	290	{
198a5278	291	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
828b35f6 JG	292	unsigned long nr_pages, iommu_page_index;
	293	unsigned long pa = page_to_phys(page) + offset;
	294	int flags = ZPCI_PTE_VALID;
	295	dma_addr_t dma_addr;
52d43d81	296	int ret;
828b35f6	297
828b35f6 JG	298	/* This rounds up number of pages based on size and offset */
828b35f6 JG	299	nr_pages = iommu_num_pages(pa, size, PAGE_SIZE);
9a99649f	300	iommu_page_index = dma_alloc_iommu(dev, nr_pages);
52d43d81 SO	301	if (iommu_page_index == -1) {
52d43d81 SO	302	ret = -ENOSPC;
828b35f6	303	goto out_err;
52d43d81	304	}
828b35f6 JG	305
	306	/* Use rounded up size */
	307	size = nr_pages * PAGE_SIZE;
	308
	309	dma_addr = zdev->start_dma + iommu_page_index * PAGE_SIZE;
52d43d81 SO	310	if (dma_addr + size > zdev->end_dma) {
52d43d81 SO	311	ret = -ERANGE;
828b35f6	312	goto out_free;
52d43d81	313	}
828b35f6 JG	314
	315	if (direction == DMA_NONE \|\| direction == DMA_TO_DEVICE)
	316	flags \|= ZPCI_TABLE_PROTECTED;
	317
52d43d81 SO	318	ret = dma_update_trans(zdev, pa, dma_addr, size, flags);
	319	if (ret)
	320	goto out_free;
	321
	322	atomic64_add(nr_pages, &zdev->mapped_pages);
	323	return dma_addr + (offset & ~PAGE_MASK);
828b35f6 JG	324
828b35f6 JG	325	out_free:
9a99649f	326	dma_free_iommu(dev, iommu_page_index, nr_pages);
828b35f6	327	out_err:
1f1dcbd4	328	zpci_err("map error:\n");
52d43d81	329	zpci_err_dma(ret, pa);
828b35f6 JG	330	return DMA_ERROR_CODE;
	331	}
	332
	333	static void s390_dma_unmap_pages(struct device *dev, dma_addr_t dma_addr,
	334	size_t size, enum dma_data_direction direction,
00085f1e	335	unsigned long attrs)
828b35f6	336	{
198a5278	337	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
828b35f6	338	unsigned long iommu_page_index;
52d43d81	339	int npages, ret;
828b35f6 JG	340
	341	npages = iommu_num_pages(dma_addr, size, PAGE_SIZE);
	342	dma_addr = dma_addr & PAGE_MASK;
52d43d81 SO	343	ret = dma_update_trans(zdev, 0, dma_addr, npages * PAGE_SIZE,
	344	ZPCI_PTE_INVALID);
	345	if (ret) {
1f1dcbd4	346	zpci_err("unmap error:\n");
52d43d81 SO	347	zpci_err_dma(ret, dma_addr);
52d43d81 SO	348	return;
1f1dcbd4	349	}
828b35f6	350
6001018a	351	atomic64_add(npages, &zdev->unmapped_pages);
828b35f6	352	iommu_page_index = (dma_addr - zdev->start_dma) >> PAGE_SHIFT;
9a99649f	353	dma_free_iommu(dev, iommu_page_index, npages);
828b35f6 JG	354	}
	355
	356	static void s390_dma_alloc(struct device dev, size_t size,
	357	dma_addr_t *dma_handle, gfp_t flag,
00085f1e	358	unsigned long attrs)
828b35f6	359	{
198a5278	360	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
828b35f6 JG	361	struct page *page;
	362	unsigned long pa;
	363	dma_addr_t map;
	364
	365	size = PAGE_ALIGN(size);
	366	page = alloc_pages(flag, get_order(size));
	367	if (!page)
	368	return NULL;
d0b08853	369
828b35f6 JG	370	pa = page_to_phys(page);
	371	memset((void *) pa, 0, size);
	372
00085f1e	373	map = s390_dma_map_pages(dev, page, 0, size, DMA_BIDIRECTIONAL, 0);
828b35f6 JG	374	if (dma_mapping_error(dev, map)) {
	375	free_pages(pa, get_order(size));
	376	return NULL;
	377	}
	378
6001018a	379	atomic64_add(size / PAGE_SIZE, &zdev->allocated_pages);
828b35f6 JG	380	if (dma_handle)
	381	*dma_handle = map;
	382	return (void *) pa;
	383	}
	384
	385	static void s390_dma_free(struct device *dev, size_t size,
	386	void *pa, dma_addr_t dma_handle,
00085f1e	387	unsigned long attrs)
828b35f6	388	{
198a5278	389	struct zpci_dev *zdev = to_zpci(to_pci_dev(dev));
f7038b7c SO	390
f7038b7c SO	391	size = PAGE_ALIGN(size);
6001018a	392	atomic64_sub(size / PAGE_SIZE, &zdev->allocated_pages);
00085f1e	393	s390_dma_unmap_pages(dev, dma_handle, size, DMA_BIDIRECTIONAL, 0);
828b35f6 JG	394	free_pages((unsigned long) pa, get_order(size));
	395	}
	396
	397	static int s390_dma_map_sg(struct device dev, struct scatterlist sg,
	398	int nr_elements, enum dma_data_direction dir,
00085f1e	399	unsigned long attrs)
828b35f6 JG	400	{
	401	int mapped_elements = 0;
	402	struct scatterlist *s;
	403	int i;
	404
	405	for_each_sg(sg, s, nr_elements, i) {
	406	struct page *page = sg_page(s);
	407	s->dma_address = s390_dma_map_pages(dev, page, s->offset,
00085f1e	408	s->length, dir, 0);
828b35f6 JG	409	if (!dma_mapping_error(dev, s->dma_address)) {
	410	s->dma_length = s->length;
	411	mapped_elements++;
	412	} else
	413	goto unmap;
	414	}
	415	out:
	416	return mapped_elements;
	417
	418	unmap:
	419	for_each_sg(sg, s, mapped_elements, i) {
	420	if (s->dma_address)
	421	s390_dma_unmap_pages(dev, s->dma_address, s->dma_length,
00085f1e	422	dir, 0);
828b35f6 JG	423	s->dma_address = 0;
	424	s->dma_length = 0;
	425	}
	426	mapped_elements = 0;
	427	goto out;
	428	}
	429
	430	static void s390_dma_unmap_sg(struct device dev, struct scatterlist sg,
	431	int nr_elements, enum dma_data_direction dir,
00085f1e	432	unsigned long attrs)
828b35f6 JG	433	{
	434	struct scatterlist *s;
	435	int i;
	436
	437	for_each_sg(sg, s, nr_elements, i) {
00085f1e KK	438	s390_dma_unmap_pages(dev, s->dma_address, s->dma_length, dir,
00085f1e KK	439	0);
828b35f6 JG	440	s->dma_address = 0;
	441	s->dma_length = 0;
	442	}
	443	}
	444
	445	int zpci_dma_init_device(struct zpci_dev *zdev)
	446	{
828b35f6 JG	447	int rc;
828b35f6 JG	448
8128f23c GS	449	/*
	450	* At this point, if the device is part of an IOMMU domain, this would
	451	* be a strong hint towards a bug in the IOMMU API (common) code and/or
	452	* simultaneous access via IOMMU and DMA API. So let's issue a warning.
	453	*/
	454	WARN_ON(zdev->s390_domain);
	455
828b35f6 JG	456	spin_lock_init(&zdev->iommu_bitmap_lock);
	457	spin_lock_init(&zdev->dma_table_lock);
	458
	459	zdev->dma_table = dma_alloc_cpu_table();
	460	if (!zdev->dma_table) {
	461	rc = -ENOMEM;
dba59909	462	goto out;
828b35f6 JG	463	}
828b35f6 JG	464
69eea95c GS	465	/*
	466	* Restrict the iommu bitmap size to the minimum of the following:
	467	* - main memory size
	468	* - 3-level pagetable address limit minus start_dma offset
	469	* - DMA address range allowed by the hardware (clp query pci fn)
	470	*
	471	* Also set zdev->end_dma to the actual end address of the usable
	472	* range, instead of the theoretical maximum as reported by hardware.
	473	*/
53b1bc9a	474	zdev->start_dma = PAGE_ALIGN(zdev->start_dma);
69eea95c GS	475	zdev->iommu_size = min3((u64) high_memory,
	476	ZPCI_TABLE_SIZE_RT - zdev->start_dma,
	477	zdev->end_dma - zdev->start_dma + 1);
	478	zdev->end_dma = zdev->start_dma + zdev->iommu_size - 1;
828b35f6	479	zdev->iommu_pages = zdev->iommu_size >> PAGE_SHIFT;
22459321	480	zdev->iommu_bitmap = vzalloc(zdev->iommu_pages / 8);
828b35f6 JG	481	if (!zdev->iommu_bitmap) {
828b35f6 JG	482	rc = -ENOMEM;
dba59909	483	goto free_dma_table;
828b35f6 JG	484	}
828b35f6 JG	485
69eea95c	486	rc = zpci_register_ioat(zdev, 0, zdev->start_dma, zdev->end_dma,
828b35f6 JG	487	(u64) zdev->dma_table);
828b35f6 JG	488	if (rc)
dba59909	489	goto free_bitmap;
828b35f6	490
dba59909 SO	491	return 0;
	492	free_bitmap:
	493	vfree(zdev->iommu_bitmap);
	494	zdev->iommu_bitmap = NULL;
	495	free_dma_table:
828b35f6	496	dma_free_cpu_table(zdev->dma_table);
dba59909 SO	497	zdev->dma_table = NULL;
dba59909 SO	498	out:
828b35f6 JG	499	return rc;
	500	}
	501
	502	void zpci_dma_exit_device(struct zpci_dev *zdev)
	503	{
8128f23c GS	504	/*
	505	* At this point, if the device is part of an IOMMU domain, this would
	506	* be a strong hint towards a bug in the IOMMU API (common) code and/or
	507	* simultaneous access via IOMMU and DMA API. So let's issue a warning.
	508	*/
	509	WARN_ON(zdev->s390_domain);
	510
828b35f6	511	zpci_unregister_ioat(zdev, 0);
8128f23c GS	512	dma_cleanup_tables(zdev->dma_table);
8128f23c GS	513	zdev->dma_table = NULL;
22459321	514	vfree(zdev->iommu_bitmap);
828b35f6 JG	515	zdev->iommu_bitmap = NULL;
	516	zdev->next_bit = 0;
	517	}
	518
	519	static int __init dma_alloc_cpu_table_caches(void)
	520	{
	521	dma_region_table_cache = kmem_cache_create("PCI_DMA_region_tables",
	522	ZPCI_TABLE_SIZE, ZPCI_TABLE_ALIGN,
	523	0, NULL);
	524	if (!dma_region_table_cache)
	525	return -ENOMEM;
	526
	527	dma_page_table_cache = kmem_cache_create("PCI_DMA_page_tables",
	528	ZPCI_PT_SIZE, ZPCI_PT_ALIGN,
	529	0, NULL);
	530	if (!dma_page_table_cache) {
	531	kmem_cache_destroy(dma_region_table_cache);
	532	return -ENOMEM;
	533	}
	534	return 0;
	535	}
	536
	537	int __init zpci_dma_init(void)
	538	{
	539	return dma_alloc_cpu_table_caches();
	540	}
	541
	542	void zpci_dma_exit(void)
	543	{
	544	kmem_cache_destroy(dma_page_table_cache);
	545	kmem_cache_destroy(dma_region_table_cache);
	546	}
	547
	548	#define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16)
	549
	550	static int __init dma_debug_do_init(void)
	551	{
	552	dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
	553	return 0;
	554	}
	555	fs_initcall(dma_debug_do_init);
	556
e82becfc	557	struct dma_map_ops s390_pci_dma_ops = {
828b35f6 JG	558	.alloc = s390_dma_alloc,
	559	.free = s390_dma_free,
	560	.map_sg = s390_dma_map_sg,
	561	.unmap_sg = s390_dma_unmap_sg,
	562	.map_page = s390_dma_map_pages,
	563	.unmap_page = s390_dma_unmap_pages,
	564	/* if we support direct DMA this must be conditional */
	565	.is_phys = 0,
	566	/* dma_supported is unconditionally true without a callback */
	567	};
e82becfc	568	EXPORT_SYMBOL_GPL(s390_pci_dma_ops);
c60d1ae4 GS	569
	570	static int __init s390_iommu_setup(char *str)
	571	{
	572	if (!strncmp(str, "strict", 6))
	573	s390_iommu_strict = 1;
	574	return 0;
	575	}
	576
	577	__setup("s390_iommu=", s390_iommu_setup);