freebsd-nq/sys/dev/nvdimm/nvdimm_spa.c
Alexander Motin 024932aae9 Use atomic for start_count in devstat_start_transaction().
Combined with earlier nstart/nend removal it allows to remove several locks
from request path of GEOM and few other places.  It would be cool if we had
more SMP-friendly statistics, but this helps too.

Sponsored by:	iXsystems, Inc.
2019-12-30 03:13:38 +00:00

623 lines
16 KiB
C

/*-
* Copyright (c) 2017, 2018 The FreeBSD Foundation
* All rights reserved.
* Copyright (c) 2018, 2019 Intel Corporation
*
* This software was developed by Konstantin Belousov <kib@FreeBSD.org>
* under sponsorship from the FreeBSD Foundation.
*
* 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 THE AUTHOR AND CONTRIBUTORS ``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 THE AUTHOR 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_acpi.h"
#include "opt_ddb.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/devicestat.h>
#include <sys/disk.h>
#include <sys/efi.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/rwlock.h>
#include <sys/sglist.h>
#include <sys/uio.h>
#include <sys/uuid.h>
#include <geom/geom.h>
#include <geom/geom_int.h>
#include <machine/vmparam.h>
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_pager.h>
#include <contrib/dev/acpica/include/acpi.h>
#include <contrib/dev/acpica/include/accommon.h>
#include <contrib/dev/acpica/include/acuuid.h>
#include <dev/acpica/acpivar.h>
#include <dev/nvdimm/nvdimm_var.h>
#define UUID_INITIALIZER_VOLATILE_MEMORY \
{0x7305944f,0xfdda,0x44e3,0xb1,0x6c,{0x3f,0x22,0xd2,0x52,0xe5,0xd0}}
#define UUID_INITIALIZER_PERSISTENT_MEMORY \
{0x66f0d379,0xb4f3,0x4074,0xac,0x43,{0x0d,0x33,0x18,0xb7,0x8c,0xdb}}
#define UUID_INITIALIZER_CONTROL_REGION \
{0x92f701f6,0x13b4,0x405d,0x91,0x0b,{0x29,0x93,0x67,0xe8,0x23,0x4c}}
#define UUID_INITIALIZER_DATA_REGION \
{0x91af0530,0x5d86,0x470e,0xa6,0xb0,{0x0a,0x2d,0xb9,0x40,0x82,0x49}}
#define UUID_INITIALIZER_VOLATILE_VIRTUAL_DISK \
{0x77ab535a,0x45fc,0x624b,0x55,0x60,{0xf7,0xb2,0x81,0xd1,0xf9,0x6e}}
#define UUID_INITIALIZER_VOLATILE_VIRTUAL_CD \
{0x3d5abd30,0x4175,0x87ce,0x6d,0x64,{0xd2,0xad,0xe5,0x23,0xc4,0xbb}}
#define UUID_INITIALIZER_PERSISTENT_VIRTUAL_DISK \
{0x5cea02c9,0x4d07,0x69d3,0x26,0x9f,{0x44,0x96,0xfb,0xe0,0x96,0xf9}}
#define UUID_INITIALIZER_PERSISTENT_VIRTUAL_CD \
{0x08018188,0x42cd,0xbb48,0x10,0x0f,{0x53,0x87,0xd5,0x3d,0xed,0x3d}}
static struct nvdimm_SPA_uuid_list_elm {
const char *u_name;
struct uuid u_id;
const bool u_usr_acc;
} nvdimm_SPA_uuid_list[] = {
[SPA_TYPE_VOLATILE_MEMORY] = {
.u_name = "VOLA MEM ",
.u_id = UUID_INITIALIZER_VOLATILE_MEMORY,
.u_usr_acc = true,
},
[SPA_TYPE_PERSISTENT_MEMORY] = {
.u_name = "PERS MEM",
.u_id = UUID_INITIALIZER_PERSISTENT_MEMORY,
.u_usr_acc = true,
},
[SPA_TYPE_CONTROL_REGION] = {
.u_name = "CTRL RG ",
.u_id = UUID_INITIALIZER_CONTROL_REGION,
.u_usr_acc = false,
},
[SPA_TYPE_DATA_REGION] = {
.u_name = "DATA RG ",
.u_id = UUID_INITIALIZER_DATA_REGION,
.u_usr_acc = true,
},
[SPA_TYPE_VOLATILE_VIRTUAL_DISK] = {
.u_name = "VIRT DSK",
.u_id = UUID_INITIALIZER_VOLATILE_VIRTUAL_DISK,
.u_usr_acc = true,
},
[SPA_TYPE_VOLATILE_VIRTUAL_CD] = {
.u_name = "VIRT CD ",
.u_id = UUID_INITIALIZER_VOLATILE_VIRTUAL_CD,
.u_usr_acc = true,
},
[SPA_TYPE_PERSISTENT_VIRTUAL_DISK] = {
.u_name = "PV DSK ",
.u_id = UUID_INITIALIZER_PERSISTENT_VIRTUAL_DISK,
.u_usr_acc = true,
},
[SPA_TYPE_PERSISTENT_VIRTUAL_CD] = {
.u_name = "PV CD ",
.u_id = UUID_INITIALIZER_PERSISTENT_VIRTUAL_CD,
.u_usr_acc = true,
},
};
enum SPA_mapping_type
nvdimm_spa_type_from_name(const char *name)
{
int j;
for (j = 0; j < nitems(nvdimm_SPA_uuid_list); j++) {
if (strcmp(name, nvdimm_SPA_uuid_list[j].u_name) != 0)
continue;
return (j);
}
return (SPA_TYPE_UNKNOWN);
}
enum SPA_mapping_type
nvdimm_spa_type_from_uuid(struct uuid *uuid)
{
int j;
for (j = 0; j < nitems(nvdimm_SPA_uuid_list); j++) {
if (uuidcmp(uuid, &nvdimm_SPA_uuid_list[j].u_id) != 0)
continue;
return (j);
}
return (SPA_TYPE_UNKNOWN);
}
bool
nvdimm_spa_type_user_accessible(enum SPA_mapping_type spa_type)
{
if ((int)spa_type < 0 || spa_type >= nitems(nvdimm_SPA_uuid_list))
return (false);
return (nvdimm_SPA_uuid_list[spa_type].u_usr_acc);
}
static vm_memattr_t
nvdimm_spa_memattr(uint64_t efi_mem_flags)
{
vm_memattr_t mode;
if ((efi_mem_flags & EFI_MD_ATTR_WB) != 0)
mode = VM_MEMATTR_WRITE_BACK;
else if ((efi_mem_flags & EFI_MD_ATTR_WT) != 0)
mode = VM_MEMATTR_WRITE_THROUGH;
else if ((efi_mem_flags & EFI_MD_ATTR_WC) != 0)
mode = VM_MEMATTR_WRITE_COMBINING;
else if ((efi_mem_flags & EFI_MD_ATTR_WP) != 0)
mode = VM_MEMATTR_WRITE_PROTECTED;
else if ((efi_mem_flags & EFI_MD_ATTR_UC) != 0)
mode = VM_MEMATTR_UNCACHEABLE;
else {
if (bootverbose)
printf("SPA mapping attr %#lx unsupported\n",
efi_mem_flags);
mode = VM_MEMATTR_UNCACHEABLE;
}
return (mode);
}
static int
nvdimm_spa_uio(struct nvdimm_spa_dev *dev, struct uio *uio)
{
struct vm_page m, *ma;
off_t off;
vm_memattr_t mattr;
int error, n;
error = 0;
if (dev->spa_kva == NULL) {
mattr = dev->spa_memattr;
bzero(&m, sizeof(m));
vm_page_initfake(&m, 0, mattr);
ma = &m;
while (uio->uio_resid > 0) {
if (uio->uio_offset >= dev->spa_len)
break;
off = dev->spa_phys_base + uio->uio_offset;
vm_page_updatefake(&m, trunc_page(off), mattr);
n = PAGE_SIZE;
if (n > uio->uio_resid)
n = uio->uio_resid;
error = uiomove_fromphys(&ma, off & PAGE_MASK, n, uio);
if (error != 0)
break;
}
} else {
while (uio->uio_resid > 0) {
if (uio->uio_offset >= dev->spa_len)
break;
n = INT_MAX;
if (n > uio->uio_resid)
n = uio->uio_resid;
if (uio->uio_offset + n > dev->spa_len)
n = dev->spa_len - uio->uio_offset;
error = uiomove((char *)dev->spa_kva + uio->uio_offset,
n, uio);
if (error != 0)
break;
}
}
return (error);
}
static int
nvdimm_spa_rw(struct cdev *dev, struct uio *uio, int ioflag)
{
return (nvdimm_spa_uio(dev->si_drv1, uio));
}
static int
nvdimm_spa_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
struct thread *td)
{
struct nvdimm_spa_dev *dev;
int error;
dev = cdev->si_drv1;
error = 0;
switch (cmd) {
case DIOCGSECTORSIZE:
*(u_int *)data = DEV_BSIZE;
break;
case DIOCGMEDIASIZE:
*(off_t *)data = dev->spa_len;
break;
default:
error = ENOTTY;
break;
}
return (error);
}
static int
nvdimm_spa_mmap_single(struct cdev *cdev, vm_ooffset_t *offset, vm_size_t size,
vm_object_t *objp, int nprot)
{
struct nvdimm_spa_dev *dev;
dev = cdev->si_drv1;
if (dev->spa_obj == NULL)
return (ENXIO);
if (*offset >= dev->spa_len || *offset + size < *offset ||
*offset + size > dev->spa_len)
return (EINVAL);
vm_object_reference(dev->spa_obj);
*objp = dev->spa_obj;
return (0);
}
static struct cdevsw spa_cdevsw = {
.d_version = D_VERSION,
.d_flags = D_DISK,
.d_name = "nvdimm_spa",
.d_read = nvdimm_spa_rw,
.d_write = nvdimm_spa_rw,
.d_ioctl = nvdimm_spa_ioctl,
.d_mmap_single = nvdimm_spa_mmap_single,
};
static void
nvdimm_spa_g_all_unmapped(struct nvdimm_spa_dev *dev, struct bio *bp, int rw)
{
struct vm_page maa[bp->bio_ma_n];
vm_page_t ma[bp->bio_ma_n];
vm_memattr_t mattr;
int i;
mattr = dev->spa_memattr;
for (i = 0; i < nitems(ma); i++) {
bzero(&maa[i], sizeof(maa[i]));
vm_page_initfake(&maa[i], dev->spa_phys_base +
trunc_page(bp->bio_offset) + PAGE_SIZE * i, mattr);
ma[i] = &maa[i];
}
if (rw == BIO_READ)
pmap_copy_pages(ma, bp->bio_offset & PAGE_MASK, bp->bio_ma,
bp->bio_ma_offset, bp->bio_length);
else
pmap_copy_pages(bp->bio_ma, bp->bio_ma_offset, ma,
bp->bio_offset & PAGE_MASK, bp->bio_length);
}
static void
nvdimm_spa_g_thread(void *arg)
{
struct g_spa *sc;
struct bio *bp;
struct uio auio;
struct iovec aiovec;
int error;
sc = arg;
for (;;) {
mtx_lock(&sc->spa_g_mtx);
for (;;) {
bp = bioq_takefirst(&sc->spa_g_queue);
if (bp != NULL)
break;
msleep(&sc->spa_g_queue, &sc->spa_g_mtx, PRIBIO,
"spa_g", 0);
if (!sc->spa_g_proc_run) {
sc->spa_g_proc_exiting = true;
wakeup(&sc->spa_g_queue);
mtx_unlock(&sc->spa_g_mtx);
kproc_exit(0);
}
continue;
}
mtx_unlock(&sc->spa_g_mtx);
if (bp->bio_cmd != BIO_READ && bp->bio_cmd != BIO_WRITE &&
bp->bio_cmd != BIO_FLUSH) {
error = EOPNOTSUPP;
goto completed;
}
error = 0;
if (bp->bio_cmd == BIO_FLUSH) {
if (sc->dev->spa_kva != NULL) {
pmap_large_map_wb(sc->dev->spa_kva,
sc->dev->spa_len);
} else {
pmap_flush_cache_phys_range(
(vm_paddr_t)sc->dev->spa_phys_base,
(vm_paddr_t)sc->dev->spa_phys_base +
sc->dev->spa_len, sc->dev->spa_memattr);
}
/*
* XXX flush IMC
*/
goto completed;
}
if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
if (sc->dev->spa_kva != NULL) {
aiovec.iov_base = (char *)sc->dev->spa_kva +
bp->bio_offset;
aiovec.iov_len = bp->bio_length;
auio.uio_iov = &aiovec;
auio.uio_iovcnt = 1;
auio.uio_resid = bp->bio_length;
auio.uio_offset = bp->bio_offset;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = bp->bio_cmd == BIO_READ ?
UIO_WRITE : UIO_READ;
auio.uio_td = curthread;
error = uiomove_fromphys(bp->bio_ma,
bp->bio_ma_offset, bp->bio_length, &auio);
bp->bio_resid = auio.uio_resid;
} else {
nvdimm_spa_g_all_unmapped(sc->dev, bp,
bp->bio_cmd);
bp->bio_resid = bp->bio_length;
error = 0;
}
} else {
aiovec.iov_base = bp->bio_data;
aiovec.iov_len = bp->bio_length;
auio.uio_iov = &aiovec;
auio.uio_iovcnt = 1;
auio.uio_resid = bp->bio_length;
auio.uio_offset = bp->bio_offset;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = bp->bio_cmd == BIO_READ ? UIO_READ :
UIO_WRITE;
auio.uio_td = curthread;
error = nvdimm_spa_uio(sc->dev, &auio);
bp->bio_resid = auio.uio_resid;
}
bp->bio_bcount = bp->bio_length;
devstat_end_transaction_bio(sc->spa_g_devstat, bp);
completed:
bp->bio_completed = bp->bio_length;
g_io_deliver(bp, error);
}
}
static void
nvdimm_spa_g_start(struct bio *bp)
{
struct g_spa *sc;
sc = bp->bio_to->geom->softc;
if (bp->bio_cmd == BIO_READ || bp->bio_cmd == BIO_WRITE) {
devstat_start_transaction_bio(sc->spa_g_devstat, bp);
}
mtx_lock(&sc->spa_g_mtx);
bioq_disksort(&sc->spa_g_queue, bp);
wakeup(&sc->spa_g_queue);
mtx_unlock(&sc->spa_g_mtx);
}
static int
nvdimm_spa_g_access(struct g_provider *pp, int r, int w, int e)
{
return (0);
}
static struct g_geom * nvdimm_spa_g_create(struct nvdimm_spa_dev *dev,
const char *name);
static g_ctl_destroy_geom_t nvdimm_spa_g_destroy_geom;
struct g_class nvdimm_spa_g_class = {
.name = "SPA",
.version = G_VERSION,
.start = nvdimm_spa_g_start,
.access = nvdimm_spa_g_access,
.destroy_geom = nvdimm_spa_g_destroy_geom,
};
DECLARE_GEOM_CLASS(nvdimm_spa_g_class, g_spa);
int
nvdimm_spa_init(struct SPA_mapping *spa, ACPI_NFIT_SYSTEM_ADDRESS *nfitaddr,
enum SPA_mapping_type spa_type)
{
char *name;
int error;
spa->spa_type = spa_type;
spa->spa_nfit_idx = nfitaddr->RangeIndex;
spa->dev.spa_domain =
((nfitaddr->Flags & ACPI_NFIT_PROXIMITY_VALID) != 0) ?
nfitaddr->ProximityDomain : -1;
spa->dev.spa_phys_base = nfitaddr->Address;
spa->dev.spa_len = nfitaddr->Length;
spa->dev.spa_efi_mem_flags = nfitaddr->MemoryMapping;
if (bootverbose) {
printf("NVDIMM SPA%d base %#016jx len %#016jx %s fl %#jx\n",
spa->spa_nfit_idx,
(uintmax_t)spa->dev.spa_phys_base,
(uintmax_t)spa->dev.spa_len,
nvdimm_SPA_uuid_list[spa_type].u_name,
spa->dev.spa_efi_mem_flags);
}
spa->dev.spa_memattr = nvdimm_spa_memattr(nfitaddr->MemoryMapping);
if (!nvdimm_SPA_uuid_list[spa_type].u_usr_acc)
return (0);
asprintf(&name, M_NVDIMM, "spa%d", spa->spa_nfit_idx);
error = nvdimm_spa_dev_init(&spa->dev, name, spa->spa_nfit_idx);
free(name, M_NVDIMM);
return (error);
}
int
nvdimm_spa_dev_init(struct nvdimm_spa_dev *dev, const char *name, int unit)
{
struct make_dev_args mda;
struct sglist *spa_sg;
char *devname;
int error, error1;
error1 = pmap_large_map(dev->spa_phys_base, dev->spa_len,
&dev->spa_kva, dev->spa_memattr);
if (error1 != 0) {
printf("NVDIMM %s cannot map into KVA, error %d\n", name,
error1);
dev->spa_kva = NULL;
}
spa_sg = sglist_alloc(1, M_WAITOK);
error = sglist_append_phys(spa_sg, dev->spa_phys_base,
dev->spa_len);
if (error == 0) {
dev->spa_obj = vm_pager_allocate(OBJT_SG, spa_sg, dev->spa_len,
VM_PROT_ALL, 0, NULL);
if (dev->spa_obj == NULL) {
printf("NVDIMM %s failed to alloc vm object", name);
sglist_free(spa_sg);
}
} else {
printf("NVDIMM %s failed to init sglist, error %d", name,
error);
sglist_free(spa_sg);
}
make_dev_args_init(&mda);
mda.mda_flags = MAKEDEV_WAITOK | MAKEDEV_CHECKNAME;
mda.mda_devsw = &spa_cdevsw;
mda.mda_cr = NULL;
mda.mda_uid = UID_ROOT;
mda.mda_gid = GID_OPERATOR;
mda.mda_mode = 0660;
mda.mda_si_drv1 = dev;
mda.mda_unit = unit;
asprintf(&devname, M_NVDIMM, "nvdimm_%s", name);
error = make_dev_s(&mda, &dev->spa_dev, "%s", devname);
free(devname, M_NVDIMM);
if (error != 0) {
printf("NVDIMM %s cannot create devfs node, error %d\n", name,
error);
if (error1 == 0)
error1 = error;
}
dev->spa_g = nvdimm_spa_g_create(dev, name);
if (dev->spa_g == NULL && error1 == 0)
error1 = ENXIO;
return (error1);
}
static struct g_geom *
nvdimm_spa_g_create(struct nvdimm_spa_dev *dev, const char *name)
{
struct g_geom *gp;
struct g_spa *sc;
int error;
gp = NULL;
sc = malloc(sizeof(struct g_spa), M_NVDIMM, M_WAITOK | M_ZERO);
sc->dev = dev;
bioq_init(&sc->spa_g_queue);
mtx_init(&sc->spa_g_mtx, "spag", NULL, MTX_DEF);
sc->spa_g_proc_run = true;
sc->spa_g_proc_exiting = false;
error = kproc_create(nvdimm_spa_g_thread, sc, &sc->spa_g_proc, 0, 0,
"g_spa");
if (error != 0) {
mtx_destroy(&sc->spa_g_mtx);
free(sc, M_NVDIMM);
printf("NVDIMM %s cannot create geom worker, error %d\n", name,
error);
} else {
g_topology_lock();
gp = g_new_geomf(&nvdimm_spa_g_class, "%s", name);
gp->softc = sc;
sc->spa_p = g_new_providerf(gp, "%s", name);
sc->spa_p->mediasize = dev->spa_len;
sc->spa_p->sectorsize = DEV_BSIZE;
sc->spa_p->flags |= G_PF_DIRECT_SEND | G_PF_DIRECT_RECEIVE |
G_PF_ACCEPT_UNMAPPED;
g_error_provider(sc->spa_p, 0);
sc->spa_g_devstat = devstat_new_entry("spa", -1, DEV_BSIZE,
DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT,
DEVSTAT_PRIORITY_MAX);
g_topology_unlock();
}
return (gp);
}
void
nvdimm_spa_fini(struct SPA_mapping *spa)
{
nvdimm_spa_dev_fini(&spa->dev);
}
void
nvdimm_spa_dev_fini(struct nvdimm_spa_dev *dev)
{
if (dev->spa_g != NULL) {
g_topology_lock();
nvdimm_spa_g_destroy_geom(NULL, dev->spa_g->class, dev->spa_g);
g_topology_unlock();
}
if (dev->spa_dev != NULL) {
destroy_dev(dev->spa_dev);
dev->spa_dev = NULL;
}
vm_object_deallocate(dev->spa_obj);
if (dev->spa_kva != NULL) {
pmap_large_unmap(dev->spa_kva, dev->spa_len);
dev->spa_kva = NULL;
}
}
static int
nvdimm_spa_g_destroy_geom(struct gctl_req *req, struct g_class *cp,
struct g_geom *gp)
{
struct g_spa *sc;
sc = gp->softc;
mtx_lock(&sc->spa_g_mtx);
sc->spa_g_proc_run = false;
wakeup(&sc->spa_g_queue);
while (!sc->spa_g_proc_exiting)
msleep(&sc->spa_g_queue, &sc->spa_g_mtx, PRIBIO, "spa_e", 0);
mtx_unlock(&sc->spa_g_mtx);
g_topology_assert();
g_wither_geom(gp, ENXIO);
sc->spa_p = NULL;
if (sc->spa_g_devstat != NULL) {
devstat_remove_entry(sc->spa_g_devstat);
sc->spa_g_devstat = NULL;
}
mtx_destroy(&sc->spa_g_mtx);
free(sc, M_NVDIMM);
return (0);
}