[qemu.git] / hw / msix.c

/*
 * MSI-X device support
 *
 * This module includes support for MSI-X in pci devices.
 *
 * Author: Michael S. Tsirkin <mst@redhat.com>
 *
 *  Copyright (c) 2009, Red Hat Inc, Michael S. Tsirkin (mst@redhat.com)
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 */

#include "hw.h"
#include "msix.h"
#include "pci.h"
#include "range.h"

/* MSI-X capability structure */
#define MSIX_TABLE_OFFSET 4
#define MSIX_PBA_OFFSET 8
#define MSIX_CAP_LENGTH 12

/* MSI enable bit and maskall bit are in byte 1 in FLAGS register */
#define MSIX_CONTROL_OFFSET (PCI_MSIX_FLAGS + 1)
#define MSIX_ENABLE_MASK (PCI_MSIX_FLAGS_ENABLE >> 8)
#define MSIX_MASKALL_MASK (PCI_MSIX_FLAGS_MASKALL >> 8)

/* MSI-X table format */
#define MSIX_MSG_ADDR 0
#define MSIX_MSG_UPPER_ADDR 4
#define MSIX_MSG_DATA 8
#define MSIX_VECTOR_CTRL 12
#define MSIX_ENTRY_SIZE 16
#define MSIX_VECTOR_MASK 0x1

/* How much space does an MSIX table need. */
/* The spec requires giving the table structure
 * a 4K aligned region all by itself. */
#define MSIX_PAGE_SIZE 0x1000
/* Reserve second half of the page for pending bits */
#define MSIX_PAGE_PENDING (MSIX_PAGE_SIZE / 2)
#define MSIX_MAX_ENTRIES 32


/* Flag for interrupt controller to declare MSI-X support */
int msix_supported;

/* Add MSI-X capability to the config space for the device. */
/* Given a bar and its size, add MSI-X table on top of it
 * and fill MSI-X capability in the config space.
 * Original bar size must be a power of 2 or 0.
 * New bar size is returned. */
static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,
                           unsigned bar_nr, unsigned bar_size)
{
    int config_offset;
    uint8_t *config;
    uint32_t new_size;

    if (nentries < 1 || nentries > PCI_MSIX_FLAGS_QSIZE + 1)
        return -EINVAL;
    if (bar_size > 0x80000000)
        return -ENOSPC;

    /* Add space for MSI-X structures */
    if (!bar_size) {
        new_size = MSIX_PAGE_SIZE;
    } else if (bar_size < MSIX_PAGE_SIZE) {
        bar_size = MSIX_PAGE_SIZE;
        new_size = MSIX_PAGE_SIZE * 2;
    } else {
        new_size = bar_size * 2;
    }

    pdev->msix_bar_size = new_size;
    config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
                                       0, MSIX_CAP_LENGTH);
    if (config_offset < 0)
        return config_offset;
    config = pdev->config + config_offset;

    pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
    /* Table on top of BAR */
    pci_set_long(config + MSIX_TABLE_OFFSET, bar_size | bar_nr);
    /* Pending bits on top of that */
    pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) |
                 bar_nr);
    pdev->msix_cap = config_offset;
    /* Make flags bit writeable. */
    pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
	    MSIX_MASKALL_MASK;
    return 0;
}

static uint32_t msix_mmio_readl(void *opaque, target_phys_addr_t addr)
{
    PCIDevice *dev = opaque;
    unsigned int offset = addr & (MSIX_PAGE_SIZE - 1) & ~0x3;
    void *page = dev->msix_table_page;

    return pci_get_long(page + offset);
}

static uint32_t msix_mmio_read_unallowed(void *opaque, target_phys_addr_t addr)
{
    fprintf(stderr, "MSI-X: only dword read is allowed!\n");
    return 0;
}

static uint8_t msix_pending_mask(int vector)
{
    return 1 << (vector % 8);
}

static uint8_t *msix_pending_byte(PCIDevice *dev, int vector)
{
    return dev->msix_table_page + MSIX_PAGE_PENDING + vector / 8;
}

static int msix_is_pending(PCIDevice *dev, int vector)
{
    return *msix_pending_byte(dev, vector) & msix_pending_mask(vector);
}

static void msix_set_pending(PCIDevice *dev, int vector)
{
    *msix_pending_byte(dev, vector) |= msix_pending_mask(vector);
}

static void msix_clr_pending(PCIDevice *dev, int vector)
{
    *msix_pending_byte(dev, vector) &= ~msix_pending_mask(vector);
}

static int msix_function_masked(PCIDevice *dev)
{
    return dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] & MSIX_MASKALL_MASK;
}

static int msix_is_masked(PCIDevice *dev, int vector)
{
    unsigned offset = vector * MSIX_ENTRY_SIZE + MSIX_VECTOR_CTRL;
    return msix_function_masked(dev) ||
	   dev->msix_table_page[offset] & MSIX_VECTOR_MASK;
}

static void msix_handle_mask_update(PCIDevice *dev, int vector)
{
    if (!msix_is_masked(dev, vector) && msix_is_pending(dev, vector)) {
        msix_clr_pending(dev, vector);
        msix_notify(dev, vector);
    }
}

/* Handle MSI-X capability config write. */
void msix_write_config(PCIDevice *dev, uint32_t addr,
                       uint32_t val, int len)
{
    unsigned enable_pos = dev->msix_cap + MSIX_CONTROL_OFFSET;
    int vector;
    int i;

    if (!range_covers_byte(addr, len, enable_pos)) {
        return;
    }

    if (!msix_enabled(dev)) {
        return;
    }

    for (i = 0; i < PCI_NUM_PINS; ++i) {
        qemu_set_irq(dev->irq[i], 0);
    }

    if (msix_function_masked(dev)) {
        return;
    }

    for (vector = 0; vector < dev->msix_entries_nr; ++vector) {
        msix_handle_mask_update(dev, vector);
    }
}

static void msix_mmio_writel(void *opaque, target_phys_addr_t addr,
                             uint32_t val)
{
    PCIDevice *dev = opaque;
    unsigned int offset = addr & (MSIX_PAGE_SIZE - 1) & ~0x3;
    int vector = offset / MSIX_ENTRY_SIZE;
    pci_set_long(dev->msix_table_page + offset, val);
    msix_handle_mask_update(dev, vector);
}

static void msix_mmio_write_unallowed(void *opaque, target_phys_addr_t addr,
                                      uint32_t val)
{
    fprintf(stderr, "MSI-X: only dword write is allowed!\n");
}

static CPUWriteMemoryFunc * const msix_mmio_write[] = {
    msix_mmio_write_unallowed, msix_mmio_write_unallowed, msix_mmio_writel
};

static CPUReadMemoryFunc * const msix_mmio_read[] = {
    msix_mmio_read_unallowed, msix_mmio_read_unallowed, msix_mmio_readl
};

/* Should be called from device's map method. */
void msix_mmio_map(PCIDevice *d, int region_num,
                   pcibus_t addr, pcibus_t size, int type)
{
    uint8_t *config = d->config + d->msix_cap;
    uint32_t table = pci_get_long(config + MSIX_TABLE_OFFSET);
    uint32_t offset = table & ~(MSIX_PAGE_SIZE - 1);
    /* TODO: for assigned devices, we'll want to make it possible to map
     * pending bits separately in case they are in a separate bar. */
    int table_bir = table & PCI_MSIX_FLAGS_BIRMASK;

    if (table_bir != region_num)
        return;
    if (size <= offset)
        return;
    cpu_register_physical_memory(addr + offset, size - offset,
                                 d->msix_mmio_index);
}

static void msix_mask_all(struct PCIDevice *dev, unsigned nentries)
{
    int vector;
    for (vector = 0; vector < nentries; ++vector) {
        unsigned offset = vector * MSIX_ENTRY_SIZE + MSIX_VECTOR_CTRL;
        dev->msix_table_page[offset] |= MSIX_VECTOR_MASK;
    }
}

/* Initialize the MSI-X structures. Note: if MSI-X is supported, BAR size is
 * modified, it should be retrieved with msix_bar_size. */
int msix_init(struct PCIDevice *dev, unsigned short nentries,
              unsigned bar_nr, unsigned bar_size)
{
    int ret;
    /* Nothing to do if MSI is not supported by interrupt controller */
    if (!msix_supported)
        return -ENOTSUP;

    if (nentries > MSIX_MAX_ENTRIES)
        return -EINVAL;

    dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES *
                                        sizeof *dev->msix_entry_used);

    dev->msix_table_page = qemu_mallocz(MSIX_PAGE_SIZE);
    msix_mask_all(dev, nentries);

    dev->msix_mmio_index = cpu_register_io_memory(msix_mmio_read,
                                                  msix_mmio_write, dev);
    if (dev->msix_mmio_index == -1) {
        ret = -EBUSY;
        goto err_index;
    }

    dev->msix_entries_nr = nentries;
    ret = msix_add_config(dev, nentries, bar_nr, bar_size);
    if (ret)
        goto err_config;

    dev->cap_present |= QEMU_PCI_CAP_MSIX;
    return 0;

err_config:
    dev->msix_entries_nr = 0;
    cpu_unregister_io_memory(dev->msix_mmio_index);
err_index:
    qemu_free(dev->msix_table_page);
    dev->msix_table_page = NULL;
    qemu_free(dev->msix_entry_used);
    dev->msix_entry_used = NULL;
    return ret;
}

static void msix_free_irq_entries(PCIDevice *dev)
{
    int vector;

    for (vector = 0; vector < dev->msix_entries_nr; ++vector) {
        dev->msix_entry_used[vector] = 0;
        msix_clr_pending(dev, vector);
    }
}

/* Clean up resources for the device. */
int msix_uninit(PCIDevice *dev)
{
    if (!(dev->cap_present & QEMU_PCI_CAP_MSIX))
        return 0;
    pci_del_capability(dev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
    dev->msix_cap = 0;
    msix_free_irq_entries(dev);
    dev->msix_entries_nr = 0;
    cpu_unregister_io_memory(dev->msix_mmio_index);
    qemu_free(dev->msix_table_page);
    dev->msix_table_page = NULL;
    qemu_free(dev->msix_entry_used);
    dev->msix_entry_used = NULL;
    dev->cap_present &= ~QEMU_PCI_CAP_MSIX;
    return 0;
}

void msix_save(PCIDevice *dev, QEMUFile *f)
{
    unsigned n = dev->msix_entries_nr;

    if (!(dev->cap_present & QEMU_PCI_CAP_MSIX)) {
        return;
    }

    qemu_put_buffer(f, dev->msix_table_page, n * MSIX_ENTRY_SIZE);
    qemu_put_buffer(f, dev->msix_table_page + MSIX_PAGE_PENDING, (n + 7) / 8);
}

/* Should be called after restoring the config space. */
void msix_load(PCIDevice *dev, QEMUFile *f)
{
    unsigned n = dev->msix_entries_nr;

    if (!(dev->cap_present & QEMU_PCI_CAP_MSIX)) {
        return;
    }

    msix_free_irq_entries(dev);
    qemu_get_buffer(f, dev->msix_table_page, n * MSIX_ENTRY_SIZE);
    qemu_get_buffer(f, dev->msix_table_page + MSIX_PAGE_PENDING, (n + 7) / 8);
}

/* Does device support MSI-X? */
int msix_present(PCIDevice *dev)
{
    return dev->cap_present & QEMU_PCI_CAP_MSIX;
}

/* Is MSI-X enabled? */
int msix_enabled(PCIDevice *dev)
{
    return (dev->cap_present & QEMU_PCI_CAP_MSIX) &&
        (dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
         MSIX_ENABLE_MASK);
}

/* Size of bar where MSI-X table resides, or 0 if MSI-X not supported. */
uint32_t msix_bar_size(PCIDevice *dev)
{
    return (dev->cap_present & QEMU_PCI_CAP_MSIX) ?
        dev->msix_bar_size : 0;
}

/* Send an MSI-X message */
void msix_notify(PCIDevice *dev, unsigned vector)
{
    uint8_t *table_entry = dev->msix_table_page + vector * MSIX_ENTRY_SIZE;
    uint64_t address;
    uint32_t data;

    if (vector >= dev->msix_entries_nr || !dev->msix_entry_used[vector])
        return;
    if (msix_is_masked(dev, vector)) {
        msix_set_pending(dev, vector);
        return;
    }

    address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR);
    address = (address << 32) | pci_get_long(table_entry + MSIX_MSG_ADDR);
    data = pci_get_long(table_entry + MSIX_MSG_DATA);
    stl_phys(address, data);
}

void msix_reset(PCIDevice *dev)
{
    if (!(dev->cap_present & QEMU_PCI_CAP_MSIX))
        return;
    msix_free_irq_entries(dev);
    dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &=
	    ~dev->wmask[dev->msix_cap + MSIX_CONTROL_OFFSET];
    memset(dev->msix_table_page, 0, MSIX_PAGE_SIZE);
    msix_mask_all(dev, dev->msix_entries_nr);
}

/* PCI spec suggests that devices make it possible for software to configure
 * less vectors than supported by the device, but does not specify a standard
 * mechanism for devices to do so.
 *
 * We support this by asking devices to declare vectors software is going to
 * actually use, and checking this on the notification path. Devices that
 * don't want to follow the spec suggestion can declare all vectors as used. */

/* Mark vector as used. */
int msix_vector_use(PCIDevice *dev, unsigned vector)
{
    if (vector >= dev->msix_entries_nr)
        return -EINVAL;
    dev->msix_entry_used[vector]++;
    return 0;
}

/* Mark vector as unused. */
void msix_vector_unuse(PCIDevice *dev, unsigned vector)
{
    if (vector >= dev->msix_entries_nr || !dev->msix_entry_used[vector]) {
        return;
    }
    if (--dev->msix_entry_used[vector]) {
        return;
    }
    msix_clr_pending(dev, vector);
}

void msix_unuse_all_vectors(PCIDevice *dev)
{
    if (!(dev->cap_present & QEMU_PCI_CAP_MSIX))
        return;
    msix_free_irq_entries(dev);
}
Commit	Line	Data
02eb84d0 MT	1	/*
	2	* MSI-X device support
	3	*
	4	* This module includes support for MSI-X in pci devices.
	5	*
	6	* Author: Michael S. Tsirkin <mst@redhat.com>
	7	*
	8	* Copyright (c) 2009, Red Hat Inc, Michael S. Tsirkin (mst@redhat.com)
	9	*
	10	* This work is licensed under the terms of the GNU GPL, version 2. See
	11	* the COPYING file in the top-level directory.
	12	*/
	13
	14	#include "hw.h"
	15	#include "msix.h"
	16	#include "pci.h"
bf1b0071	17	#include "range.h"
02eb84d0	18
02eb84d0 MT	19	/* MSI-X capability structure */
	20	#define MSIX_TABLE_OFFSET 4
	21	#define MSIX_PBA_OFFSET 8
	22	#define MSIX_CAP_LENGTH 12
	23
2760952b MT	24	/* MSI enable bit and maskall bit are in byte 1 in FLAGS register */
2760952b MT	25	#define MSIX_CONTROL_OFFSET (PCI_MSIX_FLAGS + 1)
02eb84d0	26	#define MSIX_ENABLE_MASK (PCI_MSIX_FLAGS_ENABLE >> 8)
5b5cb086	27	#define MSIX_MASKALL_MASK (PCI_MSIX_FLAGS_MASKALL >> 8)
02eb84d0 MT	28
	29	/* MSI-X table format */
	30	#define MSIX_MSG_ADDR 0
	31	#define MSIX_MSG_UPPER_ADDR 4
	32	#define MSIX_MSG_DATA 8
	33	#define MSIX_VECTOR_CTRL 12
	34	#define MSIX_ENTRY_SIZE 16
	35	#define MSIX_VECTOR_MASK 0x1
5a1fc5e8 MT	36
	37	/* How much space does an MSIX table need. */
	38	/* The spec requires giving the table structure
	39	* a 4K aligned region all by itself. */
	40	#define MSIX_PAGE_SIZE 0x1000
	41	/* Reserve second half of the page for pending bits */
	42	#define MSIX_PAGE_PENDING (MSIX_PAGE_SIZE / 2)
02eb84d0 MT	43	#define MSIX_MAX_ENTRIES 32
	44
	45
02eb84d0 MT	46	/* Flag for interrupt controller to declare MSI-X support */
	47	int msix_supported;
	48
	49	/* Add MSI-X capability to the config space for the device. */
	50	/* Given a bar and its size, add MSI-X table on top of it
	51	* and fill MSI-X capability in the config space.
	52	* Original bar size must be a power of 2 or 0.
	53	* New bar size is returned. */
	54	static int msix_add_config(struct PCIDevice *pdev, unsigned short nentries,
	55	unsigned bar_nr, unsigned bar_size)
	56	{
	57	int config_offset;
	58	uint8_t *config;
	59	uint32_t new_size;
	60
	61	if (nentries < 1 \|\| nentries > PCI_MSIX_FLAGS_QSIZE + 1)
	62	return -EINVAL;
	63	if (bar_size > 0x80000000)
	64	return -ENOSPC;
	65
	66	/* Add space for MSI-X structures */
5e520a7d	67	if (!bar_size) {
5a1fc5e8 MT	68	new_size = MSIX_PAGE_SIZE;
	69	} else if (bar_size < MSIX_PAGE_SIZE) {
	70	bar_size = MSIX_PAGE_SIZE;
	71	new_size = MSIX_PAGE_SIZE * 2;
	72	} else {
02eb84d0	73	new_size = bar_size * 2;
5a1fc5e8	74	}
02eb84d0 MT	75
02eb84d0 MT	76	pdev->msix_bar_size = new_size;
ca77089d IY	77	config_offset = pci_add_capability(pdev, PCI_CAP_ID_MSIX,
ca77089d IY	78	0, MSIX_CAP_LENGTH);
02eb84d0 MT	79	if (config_offset < 0)
	80	return config_offset;
	81	config = pdev->config + config_offset;
	82
	83	pci_set_word(config + PCI_MSIX_FLAGS, nentries - 1);
	84	/* Table on top of BAR */
	85	pci_set_long(config + MSIX_TABLE_OFFSET, bar_size \| bar_nr);
	86	/* Pending bits on top of that */
5a1fc5e8 MT	87	pci_set_long(config + MSIX_PBA_OFFSET, (bar_size + MSIX_PAGE_PENDING) \|
5a1fc5e8 MT	88	bar_nr);
02eb84d0 MT	89	pdev->msix_cap = config_offset;
02eb84d0 MT	90	/* Make flags bit writeable. */
5b5cb086 MT	91	pdev->wmask[config_offset + MSIX_CONTROL_OFFSET] \|= MSIX_ENABLE_MASK \|
5b5cb086 MT	92	MSIX_MASKALL_MASK;
02eb84d0 MT	93	return 0;
	94	}
	95
c227f099	96	static uint32_t msix_mmio_readl(void *opaque, target_phys_addr_t addr)
02eb84d0 MT	97	{
02eb84d0 MT	98	PCIDevice *dev = opaque;
76f5159d	99	unsigned int offset = addr & (MSIX_PAGE_SIZE - 1) & ~0x3;
02eb84d0	100	void *page = dev->msix_table_page;
02eb84d0	101
76f5159d	102	return pci_get_long(page + offset);
02eb84d0 MT	103	}
02eb84d0 MT	104
c227f099	105	static uint32_t msix_mmio_read_unallowed(void *opaque, target_phys_addr_t addr)
02eb84d0 MT	106	{
	107	fprintf(stderr, "MSI-X: only dword read is allowed!\n");
	108	return 0;
	109	}
	110
	111	static uint8_t msix_pending_mask(int vector)
	112	{
	113	return 1 << (vector % 8);
	114	}
	115
	116	static uint8_t msix_pending_byte(PCIDevice dev, int vector)
	117	{
5a1fc5e8	118	return dev->msix_table_page + MSIX_PAGE_PENDING + vector / 8;
02eb84d0 MT	119	}
	120
	121	static int msix_is_pending(PCIDevice *dev, int vector)
	122	{
	123	return *msix_pending_byte(dev, vector) & msix_pending_mask(vector);
	124	}
	125
	126	static void msix_set_pending(PCIDevice *dev, int vector)
	127	{
	128	*msix_pending_byte(dev, vector) \|= msix_pending_mask(vector);
	129	}
	130
	131	static void msix_clr_pending(PCIDevice *dev, int vector)
	132	{
	133	*msix_pending_byte(dev, vector) &= ~msix_pending_mask(vector);
	134	}
	135
5b5cb086 MT	136	static int msix_function_masked(PCIDevice *dev)
	137	{
	138	return dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] & MSIX_MASKALL_MASK;
	139	}
	140
02eb84d0 MT	141	static int msix_is_masked(PCIDevice *dev, int vector)
	142	{
	143	unsigned offset = vector * MSIX_ENTRY_SIZE + MSIX_VECTOR_CTRL;
5b5cb086 MT	144	return msix_function_masked(dev) \|\|
	145	dev->msix_table_page[offset] & MSIX_VECTOR_MASK;
	146	}
	147
	148	static void msix_handle_mask_update(PCIDevice *dev, int vector)
	149	{
	150	if (!msix_is_masked(dev, vector) && msix_is_pending(dev, vector)) {
	151	msix_clr_pending(dev, vector);
	152	msix_notify(dev, vector);
	153	}
	154	}
	155
	156	/* Handle MSI-X capability config write. */
	157	void msix_write_config(PCIDevice *dev, uint32_t addr,
	158	uint32_t val, int len)
	159	{
	160	unsigned enable_pos = dev->msix_cap + MSIX_CONTROL_OFFSET;
	161	int vector;
57c6db2e	162	int i;
5b5cb086	163
98a3cb02	164	if (!range_covers_byte(addr, len, enable_pos)) {
5b5cb086 MT	165	return;
	166	}
	167
	168	if (!msix_enabled(dev)) {
	169	return;
	170	}
	171
57c6db2e IY	172	for (i = 0; i < PCI_NUM_PINS; ++i) {
	173	qemu_set_irq(dev->irq[i], 0);
	174	}
5b5cb086 MT	175
	176	if (msix_function_masked(dev)) {
	177	return;
	178	}
	179
	180	for (vector = 0; vector < dev->msix_entries_nr; ++vector) {
	181	msix_handle_mask_update(dev, vector);
	182	}
02eb84d0 MT	183	}
02eb84d0 MT	184
c227f099	185	static void msix_mmio_writel(void *opaque, target_phys_addr_t addr,
02eb84d0 MT	186	uint32_t val)
	187	{
	188	PCIDevice *dev = opaque;
76f5159d	189	unsigned int offset = addr & (MSIX_PAGE_SIZE - 1) & ~0x3;
02eb84d0	190	int vector = offset / MSIX_ENTRY_SIZE;
76f5159d	191	pci_set_long(dev->msix_table_page + offset, val);
5b5cb086	192	msix_handle_mask_update(dev, vector);
02eb84d0 MT	193	}
02eb84d0 MT	194
c227f099	195	static void msix_mmio_write_unallowed(void *opaque, target_phys_addr_t addr,
02eb84d0 MT	196	uint32_t val)
	197	{
	198	fprintf(stderr, "MSI-X: only dword write is allowed!\n");
	199	}
	200
d60efc6b	201	static CPUWriteMemoryFunc * const msix_mmio_write[] = {
02eb84d0 MT	202	msix_mmio_write_unallowed, msix_mmio_write_unallowed, msix_mmio_writel
	203	};
	204
d60efc6b	205	static CPUReadMemoryFunc * const msix_mmio_read[] = {
02eb84d0 MT	206	msix_mmio_read_unallowed, msix_mmio_read_unallowed, msix_mmio_readl
	207	};
	208
	209	/* Should be called from device's map method. */
	210	void msix_mmio_map(PCIDevice *d, int region_num,
6e355d90	211	pcibus_t addr, pcibus_t size, int type)
02eb84d0 MT	212	{
	213	uint8_t *config = d->config + d->msix_cap;
	214	uint32_t table = pci_get_long(config + MSIX_TABLE_OFFSET);
5a1fc5e8	215	uint32_t offset = table & ~(MSIX_PAGE_SIZE - 1);
02eb84d0 MT	216	/* TODO: for assigned devices, we'll want to make it possible to map
	217	* pending bits separately in case they are in a separate bar. */
	218	int table_bir = table & PCI_MSIX_FLAGS_BIRMASK;
	219
	220	if (table_bir != region_num)
	221	return;
	222	if (size <= offset)
	223	return;
	224	cpu_register_physical_memory(addr + offset, size - offset,
	225	d->msix_mmio_index);
	226	}
	227
ae1be0bb MT	228	static void msix_mask_all(struct PCIDevice *dev, unsigned nentries)
	229	{
	230	int vector;
	231	for (vector = 0; vector < nentries; ++vector) {
	232	unsigned offset = vector * MSIX_ENTRY_SIZE + MSIX_VECTOR_CTRL;
	233	dev->msix_table_page[offset] \|= MSIX_VECTOR_MASK;
	234	}
	235	}
	236
02eb84d0 MT	237	/* Initialize the MSI-X structures. Note: if MSI-X is supported, BAR size is
	238	* modified, it should be retrieved with msix_bar_size. */
	239	int msix_init(struct PCIDevice *dev, unsigned short nentries,
5a1fc5e8	240	unsigned bar_nr, unsigned bar_size)
02eb84d0 MT	241	{
	242	int ret;
	243	/* Nothing to do if MSI is not supported by interrupt controller */
	244	if (!msix_supported)
	245	return -ENOTSUP;
	246
	247	if (nentries > MSIX_MAX_ENTRIES)
	248	return -EINVAL;
	249
	250	dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES *
	251	sizeof *dev->msix_entry_used);
	252
5a1fc5e8	253	dev->msix_table_page = qemu_mallocz(MSIX_PAGE_SIZE);
ae1be0bb	254	msix_mask_all(dev, nentries);
02eb84d0 MT	255
	256	dev->msix_mmio_index = cpu_register_io_memory(msix_mmio_read,
	257	msix_mmio_write, dev);
	258	if (dev->msix_mmio_index == -1) {
	259	ret = -EBUSY;
	260	goto err_index;
	261	}
	262
	263	dev->msix_entries_nr = nentries;
	264	ret = msix_add_config(dev, nentries, bar_nr, bar_size);
	265	if (ret)
	266	goto err_config;
	267
	268	dev->cap_present \|= QEMU_PCI_CAP_MSIX;
	269	return 0;
	270
	271	err_config:
3174ecd1	272	dev->msix_entries_nr = 0;
02eb84d0 MT	273	cpu_unregister_io_memory(dev->msix_mmio_index);
	274	err_index:
	275	qemu_free(dev->msix_table_page);
	276	dev->msix_table_page = NULL;
	277	qemu_free(dev->msix_entry_used);
	278	dev->msix_entry_used = NULL;
	279	return ret;
	280	}
	281
98304c84 MT	282	static void msix_free_irq_entries(PCIDevice *dev)
	283	{
	284	int vector;
	285
	286	for (vector = 0; vector < dev->msix_entries_nr; ++vector) {
	287	dev->msix_entry_used[vector] = 0;
	288	msix_clr_pending(dev, vector);
	289	}
	290	}
	291
02eb84d0 MT	292	/* Clean up resources for the device. */
	293	int msix_uninit(PCIDevice *dev)
	294	{
	295	if (!(dev->cap_present & QEMU_PCI_CAP_MSIX))
	296	return 0;
	297	pci_del_capability(dev, PCI_CAP_ID_MSIX, MSIX_CAP_LENGTH);
	298	dev->msix_cap = 0;
	299	msix_free_irq_entries(dev);
	300	dev->msix_entries_nr = 0;
	301	cpu_unregister_io_memory(dev->msix_mmio_index);
	302	qemu_free(dev->msix_table_page);
	303	dev->msix_table_page = NULL;
	304	qemu_free(dev->msix_entry_used);
	305	dev->msix_entry_used = NULL;
	306	dev->cap_present &= ~QEMU_PCI_CAP_MSIX;
	307	return 0;
	308	}
	309
	310	void msix_save(PCIDevice dev, QEMUFile f)
	311	{
9a3e12c8 MT	312	unsigned n = dev->msix_entries_nr;
9a3e12c8 MT	313
72755a70	314	if (!(dev->cap_present & QEMU_PCI_CAP_MSIX)) {
9a3e12c8	315	return;
72755a70	316	}
9a3e12c8 MT	317
9a3e12c8 MT	318	qemu_put_buffer(f, dev->msix_table_page, n * MSIX_ENTRY_SIZE);
5a1fc5e8	319	qemu_put_buffer(f, dev->msix_table_page + MSIX_PAGE_PENDING, (n + 7) / 8);
02eb84d0 MT	320	}
	321
	322	/* Should be called after restoring the config space. */
	323	void msix_load(PCIDevice dev, QEMUFile f)
	324	{
	325	unsigned n = dev->msix_entries_nr;
	326
98846d73	327	if (!(dev->cap_present & QEMU_PCI_CAP_MSIX)) {
02eb84d0	328	return;
98846d73	329	}
02eb84d0	330
4bfd1712	331	msix_free_irq_entries(dev);
02eb84d0	332	qemu_get_buffer(f, dev->msix_table_page, n * MSIX_ENTRY_SIZE);
5a1fc5e8	333	qemu_get_buffer(f, dev->msix_table_page + MSIX_PAGE_PENDING, (n + 7) / 8);
02eb84d0 MT	334	}
	335
	336	/* Does device support MSI-X? */
	337	int msix_present(PCIDevice *dev)
	338	{
	339	return dev->cap_present & QEMU_PCI_CAP_MSIX;
	340	}
	341
	342	/* Is MSI-X enabled? */
	343	int msix_enabled(PCIDevice *dev)
	344	{
	345	return (dev->cap_present & QEMU_PCI_CAP_MSIX) &&
2760952b	346	(dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
02eb84d0 MT	347	MSIX_ENABLE_MASK);
	348	}
	349
	350	/* Size of bar where MSI-X table resides, or 0 if MSI-X not supported. */
	351	uint32_t msix_bar_size(PCIDevice *dev)
	352	{
	353	return (dev->cap_present & QEMU_PCI_CAP_MSIX) ?
	354	dev->msix_bar_size : 0;
	355	}
	356
	357	/* Send an MSI-X message */
	358	void msix_notify(PCIDevice *dev, unsigned vector)
	359	{
	360	uint8_t table_entry = dev->msix_table_page + vector MSIX_ENTRY_SIZE;
	361	uint64_t address;
	362	uint32_t data;
	363
	364	if (vector >= dev->msix_entries_nr \|\| !dev->msix_entry_used[vector])
	365	return;
	366	if (msix_is_masked(dev, vector)) {
	367	msix_set_pending(dev, vector);
	368	return;
	369	}
	370
	371	address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR);
	372	address = (address << 32) \| pci_get_long(table_entry + MSIX_MSG_ADDR);
	373	data = pci_get_long(table_entry + MSIX_MSG_DATA);
	374	stl_phys(address, data);
	375	}
	376
	377	void msix_reset(PCIDevice *dev)
	378	{
	379	if (!(dev->cap_present & QEMU_PCI_CAP_MSIX))
	380	return;
	381	msix_free_irq_entries(dev);
2760952b MT	382	dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &=
2760952b MT	383	~dev->wmask[dev->msix_cap + MSIX_CONTROL_OFFSET];
5a1fc5e8	384	memset(dev->msix_table_page, 0, MSIX_PAGE_SIZE);
ae1be0bb	385	msix_mask_all(dev, dev->msix_entries_nr);
02eb84d0 MT	386	}
	387
	388	/* PCI spec suggests that devices make it possible for software to configure
	389	* less vectors than supported by the device, but does not specify a standard
	390	* mechanism for devices to do so.
	391	*
	392	* We support this by asking devices to declare vectors software is going to
	393	* actually use, and checking this on the notification path. Devices that
	394	* don't want to follow the spec suggestion can declare all vectors as used. */
	395
	396	/* Mark vector as used. */
	397	int msix_vector_use(PCIDevice *dev, unsigned vector)
	398	{
	399	if (vector >= dev->msix_entries_nr)
	400	return -EINVAL;
	401	dev->msix_entry_used[vector]++;
	402	return 0;
	403	}
	404
	405	/* Mark vector as unused. */
	406	void msix_vector_unuse(PCIDevice *dev, unsigned vector)
	407	{
98304c84 MT	408	if (vector >= dev->msix_entries_nr \|\| !dev->msix_entry_used[vector]) {
	409	return;
	410	}
	411	if (--dev->msix_entry_used[vector]) {
	412	return;
	413	}
	414	msix_clr_pending(dev, vector);
02eb84d0	415	}
b5f28bca MT	416
	417	void msix_unuse_all_vectors(PCIDevice *dev)
	418	{
	419	if (!(dev->cap_present & QEMU_PCI_CAP_MSIX))
	420	return;
	421	msix_free_irq_entries(dev);
	422	}