freebsd-dev/usr.sbin/bhyve/pci_virtio_block.c
Alexander Motin 066a8f1411 Rewrite virtio block device driver to work asynchronously and use the block
I/O interface.

Asynchronous operation, based on r280026 change, allows to not block virtual
CPU during I/O processing, that on slow/busy storage can take seconds.
Use of recently improved block I/O interface allows to process multiple
requests same time, that improves random I/O performance on wide storages.

Benchmarks of virtual disk, backed by ZVOL on RAID10 pool of 4 HDDs, show
~3.5 times random read performance improvements, while no degradation on
linear I/O.  Guest CPU usage during test dropped from 100% to almost zero.

MFC after:	2 weeks
2015-03-15 14:57:11 +00:00

409 lines
11 KiB
C

/*-
* Copyright (c) 2011 NetApp, Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/linker_set.h>
#include <sys/stat.h>
#include <sys/uio.h>
#include <sys/ioctl.h>
#include <sys/disk.h>
#include <errno.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <strings.h>
#include <unistd.h>
#include <assert.h>
#include <pthread.h>
#include <md5.h>
#include "bhyverun.h"
#include "pci_emul.h"
#include "virtio.h"
#include "block_if.h"
#define VTBLK_RINGSZ 64
#define VTBLK_MAXSEGS 32
#define VTBLK_S_OK 0
#define VTBLK_S_IOERR 1
#define VTBLK_S_UNSUPP 2
#define VTBLK_BLK_ID_BYTES 20
/* Capability bits */
#define VTBLK_F_SEG_MAX (1 << 2) /* Maximum request segments */
#define VTBLK_F_BLK_SIZE (1 << 6) /* cfg block size valid */
#define VTBLK_F_TOPOLOGY (1 << 10) /* Optimal I/O alignment */
/*
* Host capabilities
*/
#define VTBLK_S_HOSTCAPS \
( VTBLK_F_SEG_MAX | \
VTBLK_F_BLK_SIZE | \
VTBLK_F_TOPOLOGY | \
VIRTIO_RING_F_INDIRECT_DESC ) /* indirect descriptors */
/*
* Config space "registers"
*/
struct vtblk_config {
uint64_t vbc_capacity;
uint32_t vbc_size_max;
uint32_t vbc_seg_max;
struct {
uint16_t cylinders;
uint8_t heads;
uint8_t sectors;
} vbc_geometry;
uint32_t vbc_blk_size;
struct {
uint8_t physical_block_exp;
uint8_t alignment_offset;
uint16_t min_io_size;
uint32_t opt_io_size;
} vbc_topology;
uint8_t vbc_writeback;
} __packed;
/*
* Fixed-size block header
*/
struct virtio_blk_hdr {
#define VBH_OP_READ 0
#define VBH_OP_WRITE 1
#define VBH_OP_FLUSH 4
#define VBH_OP_FLUSH_OUT 5
#define VBH_OP_IDENT 8
#define VBH_FLAG_BARRIER 0x80000000 /* OR'ed into vbh_type */
uint32_t vbh_type;
uint32_t vbh_ioprio;
uint64_t vbh_sector;
} __packed;
/*
* Debug printf
*/
static int pci_vtblk_debug;
#define DPRINTF(params) if (pci_vtblk_debug) printf params
#define WPRINTF(params) printf params
struct pci_vtblk_ioreq {
struct blockif_req io_req;
struct pci_vtblk_softc *io_sc;
uint8_t *io_status;
uint16_t io_idx;
};
/*
* Per-device softc
*/
struct pci_vtblk_softc {
struct virtio_softc vbsc_vs;
pthread_mutex_t vsc_mtx;
struct vqueue_info vbsc_vq;
struct vtblk_config vbsc_cfg;
struct blockif_ctxt *bc;
char vbsc_ident[VTBLK_BLK_ID_BYTES];
struct pci_vtblk_ioreq vbsc_ios[VTBLK_RINGSZ];
};
static void pci_vtblk_reset(void *);
static void pci_vtblk_notify(void *, struct vqueue_info *);
static int pci_vtblk_cfgread(void *, int, int, uint32_t *);
static int pci_vtblk_cfgwrite(void *, int, int, uint32_t);
static struct virtio_consts vtblk_vi_consts = {
"vtblk", /* our name */
1, /* we support 1 virtqueue */
sizeof(struct vtblk_config), /* config reg size */
pci_vtblk_reset, /* reset */
pci_vtblk_notify, /* device-wide qnotify */
pci_vtblk_cfgread, /* read PCI config */
pci_vtblk_cfgwrite, /* write PCI config */
NULL, /* apply negotiated features */
VTBLK_S_HOSTCAPS, /* our capabilities */
};
static void
pci_vtblk_reset(void *vsc)
{
struct pci_vtblk_softc *sc = vsc;
DPRINTF(("vtblk: device reset requested !\n"));
vi_reset_dev(&sc->vbsc_vs);
}
static void
pci_vtblk_done(struct blockif_req *br, int err)
{
struct pci_vtblk_ioreq *io = br->br_param;
struct pci_vtblk_softc *sc = io->io_sc;
/* convert errno into a virtio block error return */
if (err == EOPNOTSUPP || err == ENOSYS)
*io->io_status = VTBLK_S_UNSUPP;
else if (err != 0)
*io->io_status = VTBLK_S_IOERR;
else
*io->io_status = VTBLK_S_OK;
/*
* Return the descriptor back to the host.
* We wrote 1 byte (our status) to host.
*/
pthread_mutex_lock(&sc->vsc_mtx);
vq_relchain(&sc->vbsc_vq, io->io_idx, 1);
vq_endchains(&sc->vbsc_vq, 0);
pthread_mutex_unlock(&sc->vsc_mtx);
}
static void
pci_vtblk_proc(struct pci_vtblk_softc *sc, struct vqueue_info *vq)
{
struct virtio_blk_hdr *vbh;
struct pci_vtblk_ioreq *io;
int i, n;
int err;
int iolen;
int writeop, type;
off_t offset;
struct iovec iov[VTBLK_MAXSEGS + 2];
uint16_t idx, flags[VTBLK_MAXSEGS + 2];
n = vq_getchain(vq, &idx, iov, VTBLK_MAXSEGS + 2, flags);
/*
* The first descriptor will be the read-only fixed header,
* and the last is for status (hence +2 above and below).
* The remaining iov's are the actual data I/O vectors.
*
* XXX - note - this fails on crash dump, which does a
* VIRTIO_BLK_T_FLUSH with a zero transfer length
*/
assert(n >= 2 && n <= VTBLK_MAXSEGS + 2);
io = &sc->vbsc_ios[idx];
assert((flags[0] & VRING_DESC_F_WRITE) == 0);
assert(iov[0].iov_len == sizeof(struct virtio_blk_hdr));
vbh = iov[0].iov_base;
memcpy(&io->io_req.br_iov, &iov[1], sizeof(struct iovec) * (n - 2));
io->io_req.br_iovcnt = n - 2;
io->io_req.br_offset = vbh->vbh_sector * DEV_BSIZE;
io->io_status = iov[--n].iov_base;
assert(iov[n].iov_len == 1);
assert(flags[n] & VRING_DESC_F_WRITE);
/*
* XXX
* The guest should not be setting the BARRIER flag because
* we don't advertise the capability.
*/
type = vbh->vbh_type & ~VBH_FLAG_BARRIER;
writeop = (type == VBH_OP_WRITE);
iolen = 0;
for (i = 1; i < n; i++) {
/*
* - write op implies read-only descriptor,
* - read/ident op implies write-only descriptor,
* therefore test the inverse of the descriptor bit
* to the op.
*/
assert(((flags[i] & VRING_DESC_F_WRITE) == 0) == writeop);
iolen += iov[i].iov_len;
}
DPRINTF(("virtio-block: %s op, %d bytes, %d segs, offset %ld\n\r",
writeop ? "write" : "read/ident", iolen, i - 1, offset));
switch (type) {
case VBH_OP_READ:
err = blockif_read(sc->bc, &io->io_req);
break;
case VBH_OP_WRITE:
err = blockif_write(sc->bc, &io->io_req);
break;
case VBH_OP_FLUSH:
case VBH_OP_FLUSH_OUT:
err = blockif_flush(sc->bc, &io->io_req);
break;
case VBH_OP_IDENT:
/* Assume a single buffer */
strlcpy(iov[1].iov_base, sc->vbsc_ident,
MIN(iov[1].iov_len, sizeof(sc->vbsc_ident)));
pci_vtblk_done(&io->io_req, 0);
return;
default:
pci_vtblk_done(&io->io_req, EOPNOTSUPP);
return;
}
assert(err == 0);
}
static void
pci_vtblk_notify(void *vsc, struct vqueue_info *vq)
{
struct pci_vtblk_softc *sc = vsc;
while (vq_has_descs(vq))
pci_vtblk_proc(sc, vq);
}
static int
pci_vtblk_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts)
{
char bident[sizeof("XX:X:X")];
struct blockif_ctxt *bctxt;
MD5_CTX mdctx;
u_char digest[16];
struct pci_vtblk_softc *sc;
off_t size;
int i, sectsz, sts, sto;
if (opts == NULL) {
printf("virtio-block: backing device required\n");
return (1);
}
/*
* The supplied backing file has to exist
*/
snprintf(bident, sizeof(bident), "%d:%d", pi->pi_slot, pi->pi_func);
bctxt = blockif_open(opts, bident);
if (bctxt == NULL) {
perror("Could not open backing file");
return (1);
}
size = blockif_size(bctxt);
sectsz = blockif_sectsz(bctxt);
blockif_psectsz(bctxt, &sts, &sto);
sc = calloc(1, sizeof(struct pci_vtblk_softc));
sc->bc = bctxt;
for (i = 0; i < VTBLK_RINGSZ; i++) {
struct pci_vtblk_ioreq *io = &sc->vbsc_ios[i];
io->io_req.br_callback = pci_vtblk_done;
io->io_req.br_param = io;
io->io_sc = sc;
io->io_idx = i;
}
pthread_mutex_init(&sc->vsc_mtx, NULL);
/* init virtio softc and virtqueues */
vi_softc_linkup(&sc->vbsc_vs, &vtblk_vi_consts, sc, pi, &sc->vbsc_vq);
sc->vbsc_vs.vs_mtx = &sc->vsc_mtx;
sc->vbsc_vq.vq_qsize = VTBLK_RINGSZ;
/* sc->vbsc_vq.vq_notify = we have no per-queue notify */
/*
* Create an identifier for the backing file. Use parts of the
* md5 sum of the filename
*/
MD5Init(&mdctx);
MD5Update(&mdctx, opts, strlen(opts));
MD5Final(digest, &mdctx);
sprintf(sc->vbsc_ident, "BHYVE-%02X%02X-%02X%02X-%02X%02X",
digest[0], digest[1], digest[2], digest[3], digest[4], digest[5]);
/* setup virtio block config space */
sc->vbsc_cfg.vbc_capacity = size / DEV_BSIZE; /* 512-byte units */
sc->vbsc_cfg.vbc_size_max = 0; /* not negotiated */
sc->vbsc_cfg.vbc_seg_max = VTBLK_MAXSEGS;
sc->vbsc_cfg.vbc_geometry.cylinders = 0; /* no geometry */
sc->vbsc_cfg.vbc_geometry.heads = 0;
sc->vbsc_cfg.vbc_geometry.sectors = 0;
sc->vbsc_cfg.vbc_blk_size = sectsz;
sc->vbsc_cfg.vbc_topology.physical_block_exp =
(sts > sectsz) ? (ffsll(sts / sectsz) - 1) : 0;
sc->vbsc_cfg.vbc_topology.alignment_offset =
(sto != 0) ? ((sts - sto) / sectsz) : 0;
sc->vbsc_cfg.vbc_topology.min_io_size = 0;
sc->vbsc_cfg.vbc_topology.opt_io_size = 0;
sc->vbsc_cfg.vbc_writeback = 0;
/*
* Should we move some of this into virtio.c? Could
* have the device, class, and subdev_0 as fields in
* the virtio constants structure.
*/
pci_set_cfgdata16(pi, PCIR_DEVICE, VIRTIO_DEV_BLOCK);
pci_set_cfgdata16(pi, PCIR_VENDOR, VIRTIO_VENDOR);
pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE);
pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_BLOCK);
pci_lintr_request(pi);
if (vi_intr_init(&sc->vbsc_vs, 1, fbsdrun_virtio_msix())) {
blockif_close(sc->bc);
free(sc);
return (1);
}
vi_set_io_bar(&sc->vbsc_vs, 0);
return (0);
}
static int
pci_vtblk_cfgwrite(void *vsc, int offset, int size, uint32_t value)
{
DPRINTF(("vtblk: write to readonly reg %d\n\r", offset));
return (1);
}
static int
pci_vtblk_cfgread(void *vsc, int offset, int size, uint32_t *retval)
{
struct pci_vtblk_softc *sc = vsc;
void *ptr;
/* our caller has already verified offset and size */
ptr = (uint8_t *)&sc->vbsc_cfg + offset;
memcpy(retval, ptr, size);
return (0);
}
struct pci_devemu pci_de_vblk = {
.pe_emu = "virtio-blk",
.pe_init = pci_vtblk_init,
.pe_barwrite = vi_pci_write,
.pe_barread = vi_pci_read
};
PCI_EMUL_SET(pci_de_vblk);