Date: Fri, 1 Mar 2019 02:16:37 +0000 (UTC) From: Ben Widawsky <bwidawsk@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-12@freebsd.org Subject: svn commit: r344680 - in stable/12/sys: dev/nvdimm modules/nvdimm Message-ID: <201903010216.x212GbqH058994@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: bwidawsk Date: Fri Mar 1 02:16:37 2019 New Revision: 344680 URL: https://svnweb.freebsd.org/changeset/base/344680 Log: MFC r344474-r344476 r344474: nvdimm: split spa dev into a separate entity Separate code for exposing a device backed by a system physical address range away from the NVDIMM spa code. This will allow a future patch to add support for NVDIMM namespaces while using the same device code. Submitted by: D Scott Phillips <d.scott.phillips@intel.com> Reviewed by: bwidawsk Sponsored by: Intel Corporation Differential Revision: https://reviews.freebsd.org/D18736 r344475: nvdimm: Read NVDIMM namespace labels When attaching to NVDIMM devices, read and verify the namespace labels from the special namespace label storage area. A later change will expose NVDIMM namespaces derived from this label data. Submitted by: D Scott Phillips <d.scott.phillips@intel.com> Discussed with: kib Sponsored by: Intel Corporation Differential Revision: https://reviews.freebsd.org/D18735 r344476: nvdimm: Simple namespace support Add support for simple NVDIMM v1.2 namespaces from the UEFI version 2.7 specification. The combination of NVDIMM regions and labels can lead to a wide variety of namespace layouts. Here we support a simple subset of namespaces where each NVDIMM SPA range is composed of a single region per member dimm. Submitted by: D Scott Phillips <d.scott.phillips@intel.com> Discussed with: kib Sponsored by: Intel Corporation Differential Revision: https://reviews.freebsd.org/D18736 Added: stable/12/sys/dev/nvdimm/nvdimm_ns.c - copied unchanged from r344476, head/sys/dev/nvdimm/nvdimm_ns.c Modified: stable/12/sys/dev/nvdimm/nvdimm.c stable/12/sys/dev/nvdimm/nvdimm_spa.c stable/12/sys/dev/nvdimm/nvdimm_var.h stable/12/sys/modules/nvdimm/Makefile Directory Properties: stable/12/ (props changed) Modified: stable/12/sys/dev/nvdimm/nvdimm.c ============================================================================== --- stable/12/sys/dev/nvdimm/nvdimm.c Fri Mar 1 02:08:12 2019 (r344679) +++ stable/12/sys/dev/nvdimm/nvdimm.c Fri Mar 1 02:16:37 2019 (r344680) @@ -36,6 +36,7 @@ __FBSDID("$FreeBSD$"); #include <sys/param.h> #include <sys/bio.h> +#include <sys/bitstring.h> #include <sys/bus.h> #include <sys/kernel.h> #include <sys/lock.h> @@ -51,10 +52,240 @@ __FBSDID("$FreeBSD$"); #define _COMPONENT ACPI_OEM ACPI_MODULE_NAME("NVDIMM") +static struct uuid intel_nvdimm_dsm_uuid = + {0x4309AC30,0x0D11,0x11E4,0x91,0x91,{0x08,0x00,0x20,0x0C,0x9A,0x66}}; +#define INTEL_NVDIMM_DSM_REV 1 +#define INTEL_NVDIMM_DSM_GET_LABEL_SIZE 4 +#define INTEL_NVDIMM_DSM_GET_LABEL_DATA 5 + static devclass_t nvdimm_devclass; static devclass_t nvdimm_root_devclass; MALLOC_DEFINE(M_NVDIMM, "nvdimm", "NVDIMM driver memory"); +static int +read_label_area_size(struct nvdimm_dev *nv) +{ + ACPI_OBJECT *result_buffer; + ACPI_HANDLE handle; + ACPI_STATUS status; + ACPI_BUFFER result; + uint32_t *out; + int error; + + handle = nvdimm_root_get_acpi_handle(nv->nv_dev); + if (handle == NULL) + return (ENODEV); + result.Length = ACPI_ALLOCATE_BUFFER; + result.Pointer = NULL; + status = acpi_EvaluateDSM(handle, (uint8_t *)&intel_nvdimm_dsm_uuid, + INTEL_NVDIMM_DSM_REV, INTEL_NVDIMM_DSM_GET_LABEL_SIZE, NULL, + &result); + error = ENXIO; + if (ACPI_SUCCESS(status) && result.Pointer != NULL && + result.Length >= sizeof(ACPI_OBJECT)) { + result_buffer = result.Pointer; + if (result_buffer->Type == ACPI_TYPE_BUFFER && + result_buffer->Buffer.Length >= 12) { + out = (uint32_t *)result_buffer->Buffer.Pointer; + nv->label_area_size = out[1]; + nv->max_label_xfer = out[2]; + error = 0; + } + } + if (result.Pointer != NULL) + AcpiOsFree(result.Pointer); + return (error); +} + +static int +read_label_area(struct nvdimm_dev *nv, uint8_t *dest, off_t offset, + off_t length) +{ + ACPI_BUFFER result; + ACPI_HANDLE handle; + ACPI_OBJECT params_pkg, params_buf, *result_buf; + ACPI_STATUS status; + uint32_t params[2]; + off_t to_read; + int error; + + error = 0; + handle = nvdimm_root_get_acpi_handle(nv->nv_dev); + if (offset < 0 || length <= 0 || + offset + length > nv->label_area_size || + handle == NULL) + return (ENODEV); + params_pkg.Type = ACPI_TYPE_PACKAGE; + params_pkg.Package.Count = 1; + params_pkg.Package.Elements = ¶ms_buf; + params_buf.Type = ACPI_TYPE_BUFFER; + params_buf.Buffer.Length = sizeof(params); + params_buf.Buffer.Pointer = (UINT8 *)params; + while (length > 0) { + to_read = MIN(length, nv->max_label_xfer); + params[0] = offset; + params[1] = to_read; + result.Length = ACPI_ALLOCATE_BUFFER; + result.Pointer = NULL; + status = acpi_EvaluateDSM(handle, + (uint8_t *)&intel_nvdimm_dsm_uuid, INTEL_NVDIMM_DSM_REV, + INTEL_NVDIMM_DSM_GET_LABEL_DATA, ¶ms_pkg, &result); + if (ACPI_FAILURE(status) || + result.Length < sizeof(ACPI_OBJECT) || + result.Pointer == NULL) { + error = ENXIO; + break; + } + result_buf = (ACPI_OBJECT *)result.Pointer; + if (result_buf->Type != ACPI_TYPE_BUFFER || + result_buf->Buffer.Pointer == NULL || + result_buf->Buffer.Length != 4 + to_read || + ((uint16_t *)result_buf->Buffer.Pointer)[0] != 0) { + error = ENXIO; + break; + } + bcopy(result_buf->Buffer.Pointer + 4, dest, to_read); + dest += to_read; + offset += to_read; + length -= to_read; + if (result.Pointer != NULL) { + AcpiOsFree(result.Pointer); + result.Pointer = NULL; + } + } + if (result.Pointer != NULL) + AcpiOsFree(result.Pointer); + return (error); +} + +static uint64_t +fletcher64(const void *data, size_t length) +{ + size_t i; + uint32_t a, b; + const uint32_t *d; + + a = 0; + b = 0; + d = (const uint32_t *)data; + length = length / sizeof(uint32_t); + for (i = 0; i < length; i++) { + a += d[i]; + b += a; + } + return ((uint64_t)b << 32 | a); +} + +static bool +label_index_is_valid(struct nvdimm_label_index *index, uint32_t max_labels, + size_t size, size_t offset) +{ + uint64_t checksum; + + index = (struct nvdimm_label_index *)((uint8_t *)index + offset); + if (strcmp(index->signature, NVDIMM_INDEX_BLOCK_SIGNATURE) != 0) + return false; + checksum = index->checksum; + index->checksum = 0; + if (checksum != fletcher64(index, size) || + index->this_offset != size * offset || index->this_size != size || + index->other_offset != size * (offset == 0 ? 1 : 0) || + index->seq == 0 || index->seq > 3 || index->slot_cnt > max_labels || + index->label_size != 1) + return false; + return true; +} + +static int +read_label(struct nvdimm_dev *nv, int num) +{ + struct nvdimm_label_entry *entry, *i, *next; + uint64_t checksum; + off_t offset; + int error; + + offset = nv->label_index->label_offset + + num * (128 << nv->label_index->label_size); + entry = malloc(sizeof(*entry), M_NVDIMM, M_WAITOK); + error = read_label_area(nv, (uint8_t *)&entry->label, offset, + sizeof(struct nvdimm_label)); + if (error != 0) { + free(entry, M_NVDIMM); + return (error); + } + checksum = entry->label.checksum; + entry->label.checksum = 0; + if (checksum != fletcher64(&entry->label, sizeof(entry->label)) || + entry->label.slot != num) { + free(entry, M_NVDIMM); + return (ENXIO); + } + + /* Insertion ordered by dimm_phys_addr */ + if (SLIST_EMPTY(&nv->labels) || + entry->label.dimm_phys_addr <= + SLIST_FIRST(&nv->labels)->label.dimm_phys_addr) { + SLIST_INSERT_HEAD(&nv->labels, entry, link); + return (0); + } + SLIST_FOREACH_SAFE(i, &nv->labels, link, next) { + if (next == NULL || + entry->label.dimm_phys_addr <= next->label.dimm_phys_addr) { + SLIST_INSERT_AFTER(i, entry, link); + return (0); + } + } + __unreachable(); +} + +static int +read_labels(struct nvdimm_dev *nv) +{ + struct nvdimm_label_index *indices; + size_t bitfield_size, index_size, num_labels; + int error, n; + bool index_0_valid, index_1_valid; + + for (index_size = 256; ; index_size += 256) { + num_labels = 8 * (index_size - + sizeof(struct nvdimm_label_index)); + if (index_size + num_labels * sizeof(struct nvdimm_label) >= + nv->label_area_size) + break; + } + num_labels = (nv->label_area_size - index_size) / + sizeof(struct nvdimm_label); + bitfield_size = roundup2(num_labels, 8) / 8; + indices = malloc(2 * index_size, M_NVDIMM, M_WAITOK); + error = read_label_area(nv, (void *)indices, 0, 2 * index_size); + if (error != 0) { + free(indices, M_NVDIMM); + return (error); + } + index_0_valid = label_index_is_valid(indices, num_labels, index_size, + 0); + index_1_valid = label_index_is_valid(indices, num_labels, index_size, + 1); + if (!index_0_valid && !index_1_valid) { + free(indices, M_NVDIMM); + return (ENXIO); + } + if (index_0_valid && index_1_valid && + (indices[1].seq > indices[0].seq || + (indices[1].seq == 1 && indices[0].seq == 3))) + index_0_valid = false; + nv->label_index = malloc(index_size, M_NVDIMM, M_WAITOK); + bcopy(indices + (index_0_valid ? 0 : 1), nv->label_index, index_size); + free(indices, M_NVDIMM); + for (bit_ffc_at((bitstr_t *)nv->label_index->free, 0, num_labels, &n); + n >= 0; + bit_ffc_at((bitstr_t *)nv->label_index->free, n + 1, num_labels, + &n)) { + read_label(nv, n); + } + return (0); +} + struct nvdimm_dev * nvdimm_find_by_handle(nfit_handle_t nv_handle) { @@ -90,6 +321,7 @@ nvdimm_attach(device_t dev) ACPI_TABLE_NFIT *nfitbl; ACPI_HANDLE handle; ACPI_STATUS status; + int error; nv = device_get_softc(dev); handle = nvdimm_root_get_acpi_handle(dev); @@ -107,6 +339,14 @@ nvdimm_attach(device_t dev) acpi_nfit_get_flush_addrs(nfitbl, nv->nv_handle, &nv->nv_flush_addr, &nv->nv_flush_addr_cnt); AcpiPutTable(&nfitbl->Header); + error = read_label_area_size(nv); + if (error == 0) { + /* + * Ignoring errors reading labels. Not all NVDIMMs + * support labels and namespaces. + */ + read_labels(nv); + } return (0); } @@ -114,9 +354,15 @@ static int nvdimm_detach(device_t dev) { struct nvdimm_dev *nv; + struct nvdimm_label_entry *label, *next; nv = device_get_softc(dev); free(nv->nv_flush_addr, M_NVDIMM); + free(nv->label_index, M_NVDIMM); + SLIST_FOREACH_SAFE(label, &nv->labels, link, next) { + SLIST_REMOVE_HEAD(&nv->labels, link); + free(label, M_NVDIMM); + } return (0); } @@ -216,6 +462,7 @@ nvdimm_root_create_spas(struct nvdimm_root_dev *dev, A free(spa, M_NVDIMM); break; } + nvdimm_create_namespaces(spa_mapping, nfitbl); SLIST_INSERT_HEAD(&dev->spas, spa_mapping, link); } free(spas, M_NVDIMM); @@ -273,6 +520,7 @@ nvdimm_root_detach(device_t dev) root = device_get_softc(dev); SLIST_FOREACH_SAFE(spa, &root->spas, link, next) { + nvdimm_destroy_namespaces(spa); nvdimm_spa_fini(spa); SLIST_REMOVE_HEAD(&root->spas, link); free(spa, M_NVDIMM); Copied: stable/12/sys/dev/nvdimm/nvdimm_ns.c (from r344476, head/sys/dev/nvdimm/nvdimm_ns.c) ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ stable/12/sys/dev/nvdimm/nvdimm_ns.c Fri Mar 1 02:16:37 2019 (r344680, copy of r344476, head/sys/dev/nvdimm/nvdimm_ns.c) @@ -0,0 +1,97 @@ +/*- + * Copyright (c) 2018 Intel Corporation + * 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 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 <sys/param.h> +#include <sys/bio.h> +#include <sys/bus.h> +#include <sys/malloc.h> +#include <sys/uuid.h> + +#include <contrib/dev/acpica/include/acpi.h> +#include <dev/acpica/acpivar.h> +#include <dev/nvdimm/nvdimm_var.h> + +int +nvdimm_create_namespaces(struct SPA_mapping *spa, ACPI_TABLE_NFIT *nfitbl) +{ + ACPI_NFIT_MEMORY_MAP **regions; + struct nvdimm_dev *nv; + struct nvdimm_label_entry *e; + struct nvdimm_namespace *ns; + nfit_handle_t dimm_handle; + char *name; + int i, error, num_regions; + + acpi_nfit_get_region_mappings_by_spa_range(nfitbl, spa->spa_nfit_idx, + ®ions, &num_regions); + if (num_regions == 0 || num_regions != regions[0]->InterleaveWays) { + free(regions, M_NVDIMM); + return (ENXIO); + } + dimm_handle = regions[0]->DeviceHandle; + nv = nvdimm_find_by_handle(dimm_handle); + if (nv == NULL) { + free(regions, M_NVDIMM); + return (ENXIO); + } + i = 0; + error = 0; + SLIST_FOREACH(e, &nv->labels, link) { + ns = malloc(sizeof(struct nvdimm_namespace), M_NVDIMM, + M_WAITOK | M_ZERO); + ns->dev.spa_domain = spa->dev.spa_domain; + ns->dev.spa_phys_base = spa->dev.spa_phys_base + + regions[0]->RegionOffset + + num_regions * + (e->label.dimm_phys_addr - regions[0]->Address); + ns->dev.spa_len = num_regions * e->label.raw_size; + ns->dev.spa_efi_mem_flags = spa->dev.spa_efi_mem_flags; + asprintf(&name, M_NVDIMM, "spa%dns%d", spa->spa_nfit_idx, i); + error = nvdimm_spa_dev_init(&ns->dev, name); + free(name, M_NVDIMM); + if (error != 0) + break; + SLIST_INSERT_HEAD(&spa->namespaces, ns, link); + i++; + } + free(regions, M_NVDIMM); + return (error); +} + +void +nvdimm_destroy_namespaces(struct SPA_mapping *spa) +{ + struct nvdimm_namespace *ns, *next; + + SLIST_FOREACH_SAFE(ns, &spa->namespaces, link, next) { + SLIST_REMOVE_HEAD(&spa->namespaces, link); + nvdimm_spa_dev_fini(&ns->dev); + free(ns, M_NVDIMM); + } +} Modified: stable/12/sys/dev/nvdimm/nvdimm_spa.c ============================================================================== --- stable/12/sys/dev/nvdimm/nvdimm_spa.c Fri Mar 1 02:08:12 2019 (r344679) +++ stable/12/sys/dev/nvdimm/nvdimm_spa.c Fri Mar 1 02:16:37 2019 (r344680) @@ -143,45 +143,45 @@ nvdimm_spa_type_from_uuid(struct uuid *uuid) } static vm_memattr_t -nvdimm_spa_memattr(struct SPA_mapping *spa) +nvdimm_spa_memattr(struct nvdimm_spa_dev *dev) { vm_memattr_t mode; - if ((spa->spa_efi_mem_flags & EFI_MD_ATTR_WB) != 0) + if ((dev->spa_efi_mem_flags & EFI_MD_ATTR_WB) != 0) mode = VM_MEMATTR_WRITE_BACK; - else if ((spa->spa_efi_mem_flags & EFI_MD_ATTR_WT) != 0) + else if ((dev->spa_efi_mem_flags & EFI_MD_ATTR_WT) != 0) mode = VM_MEMATTR_WRITE_THROUGH; - else if ((spa->spa_efi_mem_flags & EFI_MD_ATTR_WC) != 0) + else if ((dev->spa_efi_mem_flags & EFI_MD_ATTR_WC) != 0) mode = VM_MEMATTR_WRITE_COMBINING; - else if ((spa->spa_efi_mem_flags & EFI_MD_ATTR_WP) != 0) + else if ((dev->spa_efi_mem_flags & EFI_MD_ATTR_WP) != 0) mode = VM_MEMATTR_WRITE_PROTECTED; - else if ((spa->spa_efi_mem_flags & EFI_MD_ATTR_UC) != 0) + else if ((dev->spa_efi_mem_flags & EFI_MD_ATTR_UC) != 0) mode = VM_MEMATTR_UNCACHEABLE; else { if (bootverbose) - printf("SPA%d mapping attr unsupported\n", - spa->spa_nfit_idx); + printf("SPA mapping attr %#lx unsupported\n", + dev->spa_efi_mem_flags); mode = VM_MEMATTR_UNCACHEABLE; } return (mode); } static int -nvdimm_spa_uio(struct SPA_mapping *spa, struct uio *uio) +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; - if (spa->spa_kva == NULL) { - mattr = nvdimm_spa_memattr(spa); + if (dev->spa_kva == NULL) { + mattr = nvdimm_spa_memattr(dev); vm_page_initfake(&m, 0, mattr); ma = &m; while (uio->uio_resid > 0) { - if (uio->uio_offset >= spa->spa_len) + if (uio->uio_offset >= dev->spa_len) break; - off = spa->spa_phys_base + uio->uio_offset; + off = dev->spa_phys_base + uio->uio_offset; vm_page_updatefake(&m, trunc_page(off), mattr); n = PAGE_SIZE; if (n > uio->uio_resid) @@ -192,14 +192,14 @@ nvdimm_spa_uio(struct SPA_mapping *spa, struct uio *ui } } else { while (uio->uio_resid > 0) { - if (uio->uio_offset >= spa->spa_len) + 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 > spa->spa_len) - n = spa->spa_len - uio->uio_offset; - error = uiomove((char *)spa->spa_kva + uio->uio_offset, + 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; @@ -216,20 +216,20 @@ nvdimm_spa_rw(struct cdev *dev, struct uio *uio, int i } static int -nvdimm_spa_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag, +nvdimm_spa_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag, struct thread *td) { - struct SPA_mapping *spa; + struct nvdimm_spa_dev *dev; int error; - spa = dev->si_drv1; + dev = cdev->si_drv1; error = 0; switch (cmd) { case DIOCGSECTORSIZE: *(u_int *)data = DEV_BSIZE; break; case DIOCGMEDIASIZE: - *(off_t *)data = spa->spa_len; + *(off_t *)data = dev->spa_len; break; default: error = ENOTTY; @@ -239,19 +239,19 @@ nvdimm_spa_ioctl(struct cdev *dev, u_long cmd, caddr_t } static int -nvdimm_spa_mmap_single(struct cdev *dev, vm_ooffset_t *offset, vm_size_t size, +nvdimm_spa_mmap_single(struct cdev *cdev, vm_ooffset_t *offset, vm_size_t size, vm_object_t *objp, int nprot) { - struct SPA_mapping *spa; + struct nvdimm_spa_dev *dev; - spa = dev->si_drv1; - if (spa->spa_obj == NULL) + dev = cdev->si_drv1; + if (dev->spa_obj == NULL) return (ENXIO); - if (*offset >= spa->spa_len || *offset + size < *offset || - *offset + size > spa->spa_len) + if (*offset >= dev->spa_len || *offset + size < *offset || + *offset + size > dev->spa_len) return (EINVAL); - vm_object_reference(spa->spa_obj); - *objp = spa->spa_obj; + vm_object_reference(dev->spa_obj); + *objp = dev->spa_obj; return (0); } @@ -266,18 +266,17 @@ static struct cdevsw spa_cdevsw = { }; static void -nvdimm_spa_g_all_unmapped(struct SPA_mapping *spa, struct bio *bp, - int rw) +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 = nvdimm_spa_memattr(spa); + mattr = nvdimm_spa_memattr(dev); for (i = 0; i < nitems(ma); i++) { maa[i].flags = 0; - vm_page_initfake(&maa[i], spa->spa_phys_base + + vm_page_initfake(&maa[i], dev->spa_phys_base + trunc_page(bp->bio_offset) + PAGE_SIZE * i, mattr); ma[i] = &maa[i]; } @@ -292,30 +291,30 @@ nvdimm_spa_g_all_unmapped(struct SPA_mapping *spa, str static void nvdimm_spa_g_thread(void *arg) { - struct SPA_mapping *spa; + struct g_spa *sc; struct bio *bp; struct uio auio; struct iovec aiovec; int error; - spa = arg; + sc = arg; for (;;) { - mtx_lock(&spa->spa_g_mtx); + mtx_lock(&sc->spa_g_mtx); for (;;) { - bp = bioq_takefirst(&spa->spa_g_queue); + bp = bioq_takefirst(&sc->spa_g_queue); if (bp != NULL) break; - msleep(&spa->spa_g_queue, &spa->spa_g_mtx, PRIBIO, + msleep(&sc->spa_g_queue, &sc->spa_g_mtx, PRIBIO, "spa_g", 0); - if (!spa->spa_g_proc_run) { - spa->spa_g_proc_exiting = true; - wakeup(&spa->spa_g_queue); - mtx_unlock(&spa->spa_g_mtx); + 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(&spa->spa_g_mtx); + mtx_unlock(&sc->spa_g_mtx); if (bp->bio_cmd != BIO_READ && bp->bio_cmd != BIO_WRITE && bp->bio_cmd != BIO_FLUSH) { error = EOPNOTSUPP; @@ -324,13 +323,15 @@ nvdimm_spa_g_thread(void *arg) error = 0; if (bp->bio_cmd == BIO_FLUSH) { - if (spa->spa_kva != NULL) { - pmap_large_map_wb(spa->spa_kva, spa->spa_len); + 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)spa->spa_phys_base, - (vm_paddr_t)spa->spa_phys_base + - spa->spa_len, nvdimm_spa_memattr(spa)); + (vm_paddr_t)sc->dev->spa_phys_base, + (vm_paddr_t)sc->dev->spa_phys_base + + sc->dev->spa_len, + nvdimm_spa_memattr(sc->dev)); } /* * XXX flush IMC @@ -339,8 +340,8 @@ nvdimm_spa_g_thread(void *arg) } if ((bp->bio_flags & BIO_UNMAPPED) != 0) { - if (spa->spa_kva != NULL) { - aiovec.iov_base = (char *)spa->spa_kva + + 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; @@ -355,7 +356,8 @@ nvdimm_spa_g_thread(void *arg) bp->bio_ma_offset, bp->bio_length, &auio); bp->bio_resid = auio.uio_resid; } else { - nvdimm_spa_g_all_unmapped(spa, bp, bp->bio_cmd); + nvdimm_spa_g_all_unmapped(sc->dev, bp, + bp->bio_cmd); bp->bio_resid = bp->bio_length; error = 0; } @@ -370,11 +372,11 @@ nvdimm_spa_g_thread(void *arg) auio.uio_rw = bp->bio_cmd == BIO_READ ? UIO_READ : UIO_WRITE; auio.uio_td = curthread; - error = nvdimm_spa_uio(spa, &auio); + error = nvdimm_spa_uio(sc->dev, &auio); bp->bio_resid = auio.uio_resid; } bp->bio_bcount = bp->bio_length; - devstat_end_transaction_bio(spa->spa_g_devstat, bp); + devstat_end_transaction_bio(sc->spa_g_devstat, bp); completed: bp->bio_completed = bp->bio_length; g_io_deliver(bp, error); @@ -384,18 +386,18 @@ completed: static void nvdimm_spa_g_start(struct bio *bp) { - struct SPA_mapping *spa; + struct g_spa *sc; - spa = bp->bio_to->geom->softc; + sc = bp->bio_to->geom->softc; if (bp->bio_cmd == BIO_READ || bp->bio_cmd == BIO_WRITE) { - mtx_lock(&spa->spa_g_stat_mtx); - devstat_start_transaction_bio(spa->spa_g_devstat, bp); - mtx_unlock(&spa->spa_g_stat_mtx); + mtx_lock(&sc->spa_g_stat_mtx); + devstat_start_transaction_bio(sc->spa_g_devstat, bp); + mtx_unlock(&sc->spa_g_stat_mtx); } - mtx_lock(&spa->spa_g_mtx); - bioq_disksort(&spa->spa_g_queue, bp); - wakeup(&spa->spa_g_queue); - mtx_unlock(&spa->spa_g_mtx); + 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 @@ -405,11 +407,16 @@ nvdimm_spa_g_access(struct g_provider *pp, int r, int 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); @@ -417,49 +424,63 @@ int nvdimm_spa_init(struct SPA_mapping *spa, ACPI_NFIT_SYSTEM_ADDRESS *nfitaddr, enum SPA_mapping_type spa_type) { - struct make_dev_args mda; - struct sglist *spa_sg; - int error, error1; + char *name; + int error; spa->spa_type = spa_type; - spa->spa_domain = ((nfitaddr->Flags & ACPI_NFIT_PROXIMITY_VALID) != 0) ? - nfitaddr->ProximityDomain : -1; spa->spa_nfit_idx = nfitaddr->RangeIndex; - spa->spa_phys_base = nfitaddr->Address; - spa->spa_len = nfitaddr->Length; - spa->spa_efi_mem_flags = nfitaddr->MemoryMapping; + 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->spa_phys_base, (uintmax_t)spa->spa_len, + (uintmax_t)spa->dev.spa_phys_base, + (uintmax_t)spa->dev.spa_len, nvdimm_SPA_uuid_list[spa_type].u_name, - spa->spa_efi_mem_flags); + spa->dev.spa_efi_mem_flags); } if (!nvdimm_SPA_uuid_list[spa_type].u_usr_acc) return (0); - error1 = pmap_large_map(spa->spa_phys_base, spa->spa_len, - &spa->spa_kva, nvdimm_spa_memattr(spa)); + asprintf(&name, M_NVDIMM, "spa%d", spa->spa_nfit_idx); + error = nvdimm_spa_dev_init(&spa->dev, name); + free(name, M_NVDIMM); + return (error); +} + +int +nvdimm_spa_dev_init(struct nvdimm_spa_dev *dev, const char *name) +{ + 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, nvdimm_spa_memattr(dev)); if (error1 != 0) { - printf("NVDIMM SPA%d cannot map into KVA, error %d\n", - spa->spa_nfit_idx, error1); - spa->spa_kva = NULL; + 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, spa->spa_phys_base, - spa->spa_len); + error = sglist_append_phys(spa_sg, dev->spa_phys_base, + dev->spa_len); if (error == 0) { - spa->spa_obj = vm_pager_allocate(OBJT_SG, spa_sg, spa->spa_len, + dev->spa_obj = vm_pager_allocate(OBJT_SG, spa_sg, dev->spa_len, VM_PROT_ALL, 0, NULL); - if (spa->spa_obj == NULL) { - printf("NVDIMM SPA%d failed to alloc vm object", - spa->spa_nfit_idx); + if (dev->spa_obj == NULL) { + printf("NVDIMM %s failed to alloc vm object", name); sglist_free(spa_sg); } } else { - printf("NVDIMM SPA%d failed to init sglist, error %d", - spa->spa_nfit_idx, error); + printf("NVDIMM %s failed to init sglist, error %d", name, + error); sglist_free(spa_sg); } @@ -470,78 +491,112 @@ nvdimm_spa_init(struct SPA_mapping *spa, ACPI_NFIT_SYS mda.mda_uid = UID_ROOT; mda.mda_gid = GID_OPERATOR; mda.mda_mode = 0660; - mda.mda_si_drv1 = spa; - error = make_dev_s(&mda, &spa->spa_dev, "nvdimm_spa%d", - spa->spa_nfit_idx); + mda.mda_si_drv1 = dev; + 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 SPA%d cannot create devfs node, error %d\n", - spa->spa_nfit_idx, error); + 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); +} - bioq_init(&spa->spa_g_queue); - mtx_init(&spa->spa_g_mtx, "spag", NULL, MTX_DEF); - mtx_init(&spa->spa_g_stat_mtx, "spagst", NULL, MTX_DEF); - spa->spa_g_proc_run = true; - spa->spa_g_proc_exiting = false; - error = kproc_create(nvdimm_spa_g_thread, spa, &spa->spa_g_proc, 0, 0, - "g_spa%d", spa->spa_nfit_idx); +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); + mtx_init(&sc->spa_g_stat_mtx, "spagst", 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) { - printf("NVDIMM SPA%d cannot create geom worker, error %d\n", - spa->spa_nfit_idx, error); - if (error1 == 0) - error1 = error; + mtx_destroy(&sc->spa_g_mtx); + mtx_destroy(&sc->spa_g_stat_mtx); + free(sc, M_NVDIMM); + printf("NVDIMM %s cannot create geom worker, error %d\n", name, + error); } else { g_topology_lock(); - spa->spa_g = g_new_geomf(&nvdimm_spa_g_class, "spa%d", - spa->spa_nfit_idx); - spa->spa_g->softc = spa; - spa->spa_p = g_new_providerf(spa->spa_g, "spa%d", - spa->spa_nfit_idx); - spa->spa_p->mediasize = spa->spa_len; - spa->spa_p->sectorsize = DEV_BSIZE; - spa->spa_p->flags |= G_PF_DIRECT_SEND | G_PF_DIRECT_RECEIVE | + 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(spa->spa_p, 0); - spa->spa_g_devstat = devstat_new_entry("spa", spa->spa_nfit_idx, - DEV_BSIZE, DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT, + 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 (error1); + return (gp); } void nvdimm_spa_fini(struct SPA_mapping *spa) { - mtx_lock(&spa->spa_g_mtx); - spa->spa_g_proc_run = false; - wakeup(&spa->spa_g_queue); - while (!spa->spa_g_proc_exiting) - msleep(&spa->spa_g_queue, &spa->spa_g_mtx, PRIBIO, "spa_e", 0); - mtx_unlock(&spa->spa_g_mtx); - if (spa->spa_g != NULL) { + nvdimm_spa_dev_fini(&spa->dev); +} + +void +nvdimm_spa_dev_fini(struct nvdimm_spa_dev *dev) +{ + + if (dev->spa_g != NULL) { g_topology_lock(); - g_wither_geom(spa->spa_g, ENXIO); + nvdimm_spa_g_destroy_geom(NULL, dev->spa_g->class, dev->spa_g); g_topology_unlock(); - spa->spa_g = NULL; - spa->spa_p = NULL; } - if (spa->spa_g_devstat != NULL) { - devstat_remove_entry(spa->spa_g_devstat); - spa->spa_g_devstat = NULL; + if (dev->spa_dev != NULL) { + destroy_dev(dev->spa_dev); + dev->spa_dev = NULL; } - if (spa->spa_dev != NULL) { - destroy_dev(spa->spa_dev); - spa->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; } - vm_object_deallocate(spa->spa_obj); - if (spa->spa_kva != NULL) { - pmap_large_unmap(spa->spa_kva, spa->spa_len); - spa->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(&spa->spa_g_mtx); - mtx_destroy(&spa->spa_g_stat_mtx); + mtx_destroy(&sc->spa_g_mtx); + mtx_destroy(&sc->spa_g_stat_mtx); + free(sc, M_NVDIMM); + return (0); } Modified: stable/12/sys/dev/nvdimm/nvdimm_var.h ============================================================================== --- stable/12/sys/dev/nvdimm/nvdimm_var.h Fri Mar 1 02:08:12 2019 (r344679) +++ stable/12/sys/dev/nvdimm/nvdimm_var.h Fri Mar 1 02:16:37 2019 (r344680) @@ -33,6 +33,51 @@ #ifndef __DEV_NVDIMM_VAR_H__ #define __DEV_NVDIMM_VAR_H__ +#define NVDIMM_INDEX_BLOCK_SIGNATURE "NAMESPACE_INDEX" + +struct nvdimm_label_index { + char signature[16]; + uint8_t flags[3]; + uint8_t label_size; + uint32_t seq; + uint64_t this_offset; + uint64_t this_size; + uint64_t other_offset; + uint64_t label_offset; + uint32_t slot_cnt; + uint16_t rev_major; + uint16_t rev_minor; + uint64_t checksum; + uint8_t free[0]; +}; + +struct nvdimm_label { + struct uuid uuid; + char name[64]; + uint32_t flags; + uint16_t nlabel; + uint16_t position; + uint64_t set_cookie; + uint64_t lba_size; + uint64_t dimm_phys_addr; + uint64_t raw_size; + uint32_t slot; + uint8_t alignment; + uint8_t reserved[3]; + struct uuid type_guid; + struct uuid address_abstraction_guid; + uint8_t reserved1[88]; + uint64_t checksum; +}; + +struct nvdimm_label_entry { + SLIST_ENTRY(nvdimm_label_entry) link; + struct nvdimm_label label; +}; *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201903010216.x212GbqH058994>