linux/drivers/acpi/nfit.c

1873 lines
49 KiB
C
Raw Normal View History

/*
* Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of version 2 of the GNU General Public License as
* published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*/
#include <linux/list_sort.h>
#include <linux/libnvdimm.h>
#include <linux/module.h>
#include <linux/mutex.h>
2015-06-09 02:27:06 +08:00
#include <linux/ndctl.h>
#include <linux/list.h>
#include <linux/acpi.h>
2015-05-02 01:11:27 +08:00
#include <linux/sort.h>
#include <linux/pmem.h>
#include <linux/io.h>
x86, pmem: clarify that ARCH_HAS_PMEM_API implies PMEM mapped WB Given that a write-back (WB) mapping plus non-temporal stores is expected to be the most efficient way to access PMEM, update the definition of ARCH_HAS_PMEM_API to imply arch support for WB-mapped-PMEM. This is needed as a pre-requisite for adding PMEM to the direct map and mapping it with struct page. The above clarification for X86_64 means that memcpy_to_pmem() is permitted to use the non-temporal arch_memcpy_to_pmem() rather than needlessly fall back to default_memcpy_to_pmem() when the pcommit instruction is not available. When arch_memcpy_to_pmem() is not guaranteed to flush writes out of cache, i.e. on older X86_32 implementations where non-temporal stores may just dirty cache, ARCH_HAS_PMEM_API is simply disabled. The default fall back for persistent memory handling remains. Namely, map it with the WT (write-through) cache-type and hope for the best. arch_has_pmem_api() is updated to only indicate whether the arch provides the proper helpers to meet the minimum "writes are visible outside the cache hierarchy after memcpy_to_pmem() + wmb_pmem()". Code that cares whether wmb_pmem() actually flushes writes to pmem must now call arch_has_wmb_pmem() directly. Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> [hch: set ARCH_HAS_PMEM_API=n on x86_32] Reviewed-by: Christoph Hellwig <hch@lst.de> [toshi: x86_32 compile fixes] Signed-off-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-25 06:29:38 +08:00
#include <asm/cacheflush.h>
#include "nfit.h"
/*
* For readq() and writeq() on 32-bit builds, the hi-lo, lo-hi order is
* irrelevant.
*/
#include <linux/io-64-nonatomic-hi-lo.h>
static bool force_enable_dimms;
module_param(force_enable_dimms, bool, S_IRUGO|S_IWUSR);
MODULE_PARM_DESC(force_enable_dimms, "Ignore _STA (ACPI DIMM device) status");
struct nfit_table_prev {
struct list_head spas;
struct list_head memdevs;
struct list_head dcrs;
struct list_head bdws;
struct list_head idts;
struct list_head flushes;
};
static u8 nfit_uuid[NFIT_UUID_MAX][16];
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
const u8 *to_nfit_uuid(enum nfit_uuids id)
{
return nfit_uuid[id];
}
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
EXPORT_SYMBOL(to_nfit_uuid);
2015-06-09 02:27:06 +08:00
static struct acpi_nfit_desc *to_acpi_nfit_desc(
struct nvdimm_bus_descriptor *nd_desc)
{
return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
}
static struct acpi_device *to_acpi_dev(struct acpi_nfit_desc *acpi_desc)
{
struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
/*
* If provider == 'ACPI.NFIT' we can assume 'dev' is a struct
* acpi_device.
*/
if (!nd_desc->provider_name
|| strcmp(nd_desc->provider_name, "ACPI.NFIT") != 0)
return NULL;
return to_acpi_device(acpi_desc->dev);
}
static int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc,
struct nvdimm *nvdimm, unsigned int cmd, void *buf,
unsigned int buf_len)
{
2015-06-09 02:27:06 +08:00
struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
const struct nd_cmd_desc *desc = NULL;
union acpi_object in_obj, in_buf, *out_obj;
struct device *dev = acpi_desc->dev;
const char *cmd_name, *dimm_name;
unsigned long dsm_mask;
acpi_handle handle;
const u8 *uuid;
u32 offset;
int rc, i;
if (nvdimm) {
struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
struct acpi_device *adev = nfit_mem->adev;
if (!adev)
return -ENOTTY;
dimm_name = nvdimm_name(nvdimm);
2015-06-09 02:27:06 +08:00
cmd_name = nvdimm_cmd_name(cmd);
dsm_mask = nfit_mem->dsm_mask;
desc = nd_cmd_dimm_desc(cmd);
uuid = to_nfit_uuid(NFIT_DEV_DIMM);
handle = adev->handle;
} else {
struct acpi_device *adev = to_acpi_dev(acpi_desc);
cmd_name = nvdimm_bus_cmd_name(cmd);
dsm_mask = nd_desc->dsm_mask;
desc = nd_cmd_bus_desc(cmd);
uuid = to_nfit_uuid(NFIT_DEV_BUS);
handle = adev->handle;
dimm_name = "bus";
}
if (!desc || (cmd && (desc->out_num + desc->in_num == 0)))
return -ENOTTY;
if (!test_bit(cmd, &dsm_mask))
return -ENOTTY;
in_obj.type = ACPI_TYPE_PACKAGE;
in_obj.package.count = 1;
in_obj.package.elements = &in_buf;
in_buf.type = ACPI_TYPE_BUFFER;
in_buf.buffer.pointer = buf;
in_buf.buffer.length = 0;
/* libnvdimm has already validated the input envelope */
for (i = 0; i < desc->in_num; i++)
in_buf.buffer.length += nd_cmd_in_size(nvdimm, cmd, desc,
i, buf);
if (IS_ENABLED(CONFIG_ACPI_NFIT_DEBUG)) {
dev_dbg(dev, "%s:%s cmd: %s input length: %d\n", __func__,
dimm_name, cmd_name, in_buf.buffer.length);
print_hex_dump_debug(cmd_name, DUMP_PREFIX_OFFSET, 4,
4, in_buf.buffer.pointer, min_t(u32, 128,
in_buf.buffer.length), true);
}
out_obj = acpi_evaluate_dsm(handle, uuid, 1, cmd, &in_obj);
if (!out_obj) {
dev_dbg(dev, "%s:%s _DSM failed cmd: %s\n", __func__, dimm_name,
cmd_name);
return -EINVAL;
}
if (out_obj->package.type != ACPI_TYPE_BUFFER) {
dev_dbg(dev, "%s:%s unexpected output object type cmd: %s type: %d\n",
__func__, dimm_name, cmd_name, out_obj->type);
rc = -EINVAL;
goto out;
}
if (IS_ENABLED(CONFIG_ACPI_NFIT_DEBUG)) {
dev_dbg(dev, "%s:%s cmd: %s output length: %d\n", __func__,
dimm_name, cmd_name, out_obj->buffer.length);
print_hex_dump_debug(cmd_name, DUMP_PREFIX_OFFSET, 4,
4, out_obj->buffer.pointer, min_t(u32, 128,
out_obj->buffer.length), true);
}
for (i = 0, offset = 0; i < desc->out_num; i++) {
u32 out_size = nd_cmd_out_size(nvdimm, cmd, desc, i, buf,
(u32 *) out_obj->buffer.pointer);
if (offset + out_size > out_obj->buffer.length) {
dev_dbg(dev, "%s:%s output object underflow cmd: %s field: %d\n",
__func__, dimm_name, cmd_name, i);
break;
}
if (in_buf.buffer.length + offset + out_size > buf_len) {
dev_dbg(dev, "%s:%s output overrun cmd: %s field: %d\n",
__func__, dimm_name, cmd_name, i);
rc = -ENXIO;
goto out;
}
memcpy(buf + in_buf.buffer.length + offset,
out_obj->buffer.pointer + offset, out_size);
offset += out_size;
}
if (offset + in_buf.buffer.length < buf_len) {
if (i >= 1) {
/*
* status valid, return the number of bytes left
* unfilled in the output buffer
*/
rc = buf_len - offset - in_buf.buffer.length;
} else {
dev_err(dev, "%s:%s underrun cmd: %s buf_len: %d out_len: %d\n",
__func__, dimm_name, cmd_name, buf_len,
offset);
rc = -ENXIO;
}
} else
rc = 0;
out:
ACPI_FREE(out_obj);
return rc;
}
static const char *spa_type_name(u16 type)
{
static const char *to_name[] = {
[NFIT_SPA_VOLATILE] = "volatile",
[NFIT_SPA_PM] = "pmem",
[NFIT_SPA_DCR] = "dimm-control-region",
[NFIT_SPA_BDW] = "block-data-window",
[NFIT_SPA_VDISK] = "volatile-disk",
[NFIT_SPA_VCD] = "volatile-cd",
[NFIT_SPA_PDISK] = "persistent-disk",
[NFIT_SPA_PCD] = "persistent-cd",
};
if (type > NFIT_SPA_PCD)
return "unknown";
return to_name[type];
}
static int nfit_spa_type(struct acpi_nfit_system_address *spa)
{
int i;
for (i = 0; i < NFIT_UUID_MAX; i++)
if (memcmp(to_nfit_uuid(i), spa->range_guid, 16) == 0)
return i;
return -1;
}
static bool add_spa(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev,
struct acpi_nfit_system_address *spa)
{
struct device *dev = acpi_desc->dev;
struct nfit_spa *nfit_spa;
list_for_each_entry(nfit_spa, &prev->spas, list) {
if (memcmp(nfit_spa->spa, spa, sizeof(*spa)) == 0) {
list_move_tail(&nfit_spa->list, &acpi_desc->spas);
return true;
}
}
nfit_spa = devm_kzalloc(dev, sizeof(*nfit_spa), GFP_KERNEL);
if (!nfit_spa)
return false;
INIT_LIST_HEAD(&nfit_spa->list);
nfit_spa->spa = spa;
list_add_tail(&nfit_spa->list, &acpi_desc->spas);
dev_dbg(dev, "%s: spa index: %d type: %s\n", __func__,
spa->range_index,
spa_type_name(nfit_spa_type(spa)));
return true;
}
static bool add_memdev(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev,
struct acpi_nfit_memory_map *memdev)
{
struct device *dev = acpi_desc->dev;
struct nfit_memdev *nfit_memdev;
list_for_each_entry(nfit_memdev, &prev->memdevs, list)
if (memcmp(nfit_memdev->memdev, memdev, sizeof(*memdev)) == 0) {
list_move_tail(&nfit_memdev->list, &acpi_desc->memdevs);
return true;
}
nfit_memdev = devm_kzalloc(dev, sizeof(*nfit_memdev), GFP_KERNEL);
if (!nfit_memdev)
return false;
INIT_LIST_HEAD(&nfit_memdev->list);
nfit_memdev->memdev = memdev;
list_add_tail(&nfit_memdev->list, &acpi_desc->memdevs);
dev_dbg(dev, "%s: memdev handle: %#x spa: %d dcr: %d\n",
__func__, memdev->device_handle, memdev->range_index,
memdev->region_index);
return true;
}
static bool add_dcr(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev,
struct acpi_nfit_control_region *dcr)
{
struct device *dev = acpi_desc->dev;
struct nfit_dcr *nfit_dcr;
list_for_each_entry(nfit_dcr, &prev->dcrs, list)
if (memcmp(nfit_dcr->dcr, dcr, sizeof(*dcr)) == 0) {
list_move_tail(&nfit_dcr->list, &acpi_desc->dcrs);
return true;
}
nfit_dcr = devm_kzalloc(dev, sizeof(*nfit_dcr), GFP_KERNEL);
if (!nfit_dcr)
return false;
INIT_LIST_HEAD(&nfit_dcr->list);
nfit_dcr->dcr = dcr;
list_add_tail(&nfit_dcr->list, &acpi_desc->dcrs);
dev_dbg(dev, "%s: dcr index: %d windows: %d\n", __func__,
dcr->region_index, dcr->windows);
return true;
}
static bool add_bdw(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev,
struct acpi_nfit_data_region *bdw)
{
struct device *dev = acpi_desc->dev;
struct nfit_bdw *nfit_bdw;
list_for_each_entry(nfit_bdw, &prev->bdws, list)
if (memcmp(nfit_bdw->bdw, bdw, sizeof(*bdw)) == 0) {
list_move_tail(&nfit_bdw->list, &acpi_desc->bdws);
return true;
}
nfit_bdw = devm_kzalloc(dev, sizeof(*nfit_bdw), GFP_KERNEL);
if (!nfit_bdw)
return false;
INIT_LIST_HEAD(&nfit_bdw->list);
nfit_bdw->bdw = bdw;
list_add_tail(&nfit_bdw->list, &acpi_desc->bdws);
dev_dbg(dev, "%s: bdw dcr: %d windows: %d\n", __func__,
bdw->region_index, bdw->windows);
return true;
}
static bool add_idt(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev,
struct acpi_nfit_interleave *idt)
{
struct device *dev = acpi_desc->dev;
struct nfit_idt *nfit_idt;
list_for_each_entry(nfit_idt, &prev->idts, list)
if (memcmp(nfit_idt->idt, idt, sizeof(*idt)) == 0) {
list_move_tail(&nfit_idt->list, &acpi_desc->idts);
return true;
}
nfit_idt = devm_kzalloc(dev, sizeof(*nfit_idt), GFP_KERNEL);
if (!nfit_idt)
return false;
INIT_LIST_HEAD(&nfit_idt->list);
nfit_idt->idt = idt;
list_add_tail(&nfit_idt->list, &acpi_desc->idts);
dev_dbg(dev, "%s: idt index: %d num_lines: %d\n", __func__,
idt->interleave_index, idt->line_count);
return true;
}
static bool add_flush(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev,
struct acpi_nfit_flush_address *flush)
{
struct device *dev = acpi_desc->dev;
struct nfit_flush *nfit_flush;
list_for_each_entry(nfit_flush, &prev->flushes, list)
if (memcmp(nfit_flush->flush, flush, sizeof(*flush)) == 0) {
list_move_tail(&nfit_flush->list, &acpi_desc->flushes);
return true;
}
nfit_flush = devm_kzalloc(dev, sizeof(*nfit_flush), GFP_KERNEL);
if (!nfit_flush)
return false;
INIT_LIST_HEAD(&nfit_flush->list);
nfit_flush->flush = flush;
list_add_tail(&nfit_flush->list, &acpi_desc->flushes);
dev_dbg(dev, "%s: nfit_flush handle: %d hint_count: %d\n", __func__,
flush->device_handle, flush->hint_count);
return true;
}
static void *add_table(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev, void *table, const void *end)
{
struct device *dev = acpi_desc->dev;
struct acpi_nfit_header *hdr;
void *err = ERR_PTR(-ENOMEM);
if (table >= end)
return NULL;
hdr = table;
if (!hdr->length) {
dev_warn(dev, "found a zero length table '%d' parsing nfit\n",
hdr->type);
return NULL;
}
switch (hdr->type) {
case ACPI_NFIT_TYPE_SYSTEM_ADDRESS:
if (!add_spa(acpi_desc, prev, table))
return err;
break;
case ACPI_NFIT_TYPE_MEMORY_MAP:
if (!add_memdev(acpi_desc, prev, table))
return err;
break;
case ACPI_NFIT_TYPE_CONTROL_REGION:
if (!add_dcr(acpi_desc, prev, table))
return err;
break;
case ACPI_NFIT_TYPE_DATA_REGION:
if (!add_bdw(acpi_desc, prev, table))
return err;
break;
case ACPI_NFIT_TYPE_INTERLEAVE:
if (!add_idt(acpi_desc, prev, table))
return err;
break;
case ACPI_NFIT_TYPE_FLUSH_ADDRESS:
if (!add_flush(acpi_desc, prev, table))
return err;
break;
case ACPI_NFIT_TYPE_SMBIOS:
dev_dbg(dev, "%s: smbios\n", __func__);
break;
default:
dev_err(dev, "unknown table '%d' parsing nfit\n", hdr->type);
break;
}
return table + hdr->length;
}
static void nfit_mem_find_spa_bdw(struct acpi_nfit_desc *acpi_desc,
struct nfit_mem *nfit_mem)
{
u32 device_handle = __to_nfit_memdev(nfit_mem)->device_handle;
u16 dcr = nfit_mem->dcr->region_index;
struct nfit_spa *nfit_spa;
list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
u16 range_index = nfit_spa->spa->range_index;
int type = nfit_spa_type(nfit_spa->spa);
struct nfit_memdev *nfit_memdev;
if (type != NFIT_SPA_BDW)
continue;
list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
if (nfit_memdev->memdev->range_index != range_index)
continue;
if (nfit_memdev->memdev->device_handle != device_handle)
continue;
if (nfit_memdev->memdev->region_index != dcr)
continue;
nfit_mem->spa_bdw = nfit_spa->spa;
return;
}
}
dev_dbg(acpi_desc->dev, "SPA-BDW not found for SPA-DCR %d\n",
nfit_mem->spa_dcr->range_index);
nfit_mem->bdw = NULL;
}
static int nfit_mem_add(struct acpi_nfit_desc *acpi_desc,
struct nfit_mem *nfit_mem, struct acpi_nfit_system_address *spa)
{
u16 dcr = __to_nfit_memdev(nfit_mem)->region_index;
struct nfit_memdev *nfit_memdev;
struct nfit_flush *nfit_flush;
struct nfit_dcr *nfit_dcr;
struct nfit_bdw *nfit_bdw;
struct nfit_idt *nfit_idt;
u16 idt_idx, range_index;
list_for_each_entry(nfit_dcr, &acpi_desc->dcrs, list) {
if (nfit_dcr->dcr->region_index != dcr)
continue;
nfit_mem->dcr = nfit_dcr->dcr;
break;
}
if (!nfit_mem->dcr) {
dev_dbg(acpi_desc->dev, "SPA %d missing:%s%s\n",
spa->range_index, __to_nfit_memdev(nfit_mem)
? "" : " MEMDEV", nfit_mem->dcr ? "" : " DCR");
return -ENODEV;
}
/*
* We've found enough to create an nvdimm, optionally
* find an associated BDW
*/
list_add(&nfit_mem->list, &acpi_desc->dimms);
list_for_each_entry(nfit_bdw, &acpi_desc->bdws, list) {
if (nfit_bdw->bdw->region_index != dcr)
continue;
nfit_mem->bdw = nfit_bdw->bdw;
break;
}
if (!nfit_mem->bdw)
return 0;
nfit_mem_find_spa_bdw(acpi_desc, nfit_mem);
if (!nfit_mem->spa_bdw)
return 0;
range_index = nfit_mem->spa_bdw->range_index;
list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
if (nfit_memdev->memdev->range_index != range_index ||
nfit_memdev->memdev->region_index != dcr)
continue;
nfit_mem->memdev_bdw = nfit_memdev->memdev;
idt_idx = nfit_memdev->memdev->interleave_index;
list_for_each_entry(nfit_idt, &acpi_desc->idts, list) {
if (nfit_idt->idt->interleave_index != idt_idx)
continue;
nfit_mem->idt_bdw = nfit_idt->idt;
break;
}
list_for_each_entry(nfit_flush, &acpi_desc->flushes, list) {
if (nfit_flush->flush->device_handle !=
nfit_memdev->memdev->device_handle)
continue;
nfit_mem->nfit_flush = nfit_flush;
break;
}
break;
}
return 0;
}
static int nfit_mem_dcr_init(struct acpi_nfit_desc *acpi_desc,
struct acpi_nfit_system_address *spa)
{
struct nfit_mem *nfit_mem, *found;
struct nfit_memdev *nfit_memdev;
int type = nfit_spa_type(spa);
u16 dcr;
switch (type) {
case NFIT_SPA_DCR:
case NFIT_SPA_PM:
break;
default:
return 0;
}
list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
int rc;
if (nfit_memdev->memdev->range_index != spa->range_index)
continue;
found = NULL;
dcr = nfit_memdev->memdev->region_index;
list_for_each_entry(nfit_mem, &acpi_desc->dimms, list)
if (__to_nfit_memdev(nfit_mem)->region_index == dcr) {
found = nfit_mem;
break;
}
if (found)
nfit_mem = found;
else {
nfit_mem = devm_kzalloc(acpi_desc->dev,
sizeof(*nfit_mem), GFP_KERNEL);
if (!nfit_mem)
return -ENOMEM;
INIT_LIST_HEAD(&nfit_mem->list);
}
if (type == NFIT_SPA_DCR) {
struct nfit_idt *nfit_idt;
u16 idt_idx;
/* multiple dimms may share a SPA when interleaved */
nfit_mem->spa_dcr = spa;
nfit_mem->memdev_dcr = nfit_memdev->memdev;
idt_idx = nfit_memdev->memdev->interleave_index;
list_for_each_entry(nfit_idt, &acpi_desc->idts, list) {
if (nfit_idt->idt->interleave_index != idt_idx)
continue;
nfit_mem->idt_dcr = nfit_idt->idt;
break;
}
} else {
/*
* A single dimm may belong to multiple SPA-PM
* ranges, record at least one in addition to
* any SPA-DCR range.
*/
nfit_mem->memdev_pmem = nfit_memdev->memdev;
}
if (found)
continue;
rc = nfit_mem_add(acpi_desc, nfit_mem, spa);
if (rc)
return rc;
}
return 0;
}
static int nfit_mem_cmp(void *priv, struct list_head *_a, struct list_head *_b)
{
struct nfit_mem *a = container_of(_a, typeof(*a), list);
struct nfit_mem *b = container_of(_b, typeof(*b), list);
u32 handleA, handleB;
handleA = __to_nfit_memdev(a)->device_handle;
handleB = __to_nfit_memdev(b)->device_handle;
if (handleA < handleB)
return -1;
else if (handleA > handleB)
return 1;
return 0;
}
static int nfit_mem_init(struct acpi_nfit_desc *acpi_desc)
{
struct nfit_spa *nfit_spa;
/*
* For each SPA-DCR or SPA-PMEM address range find its
* corresponding MEMDEV(s). From each MEMDEV find the
* corresponding DCR. Then, if we're operating on a SPA-DCR,
* try to find a SPA-BDW and a corresponding BDW that references
* the DCR. Throw it all into an nfit_mem object. Note, that
* BDWs are optional.
*/
list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
int rc;
rc = nfit_mem_dcr_init(acpi_desc, nfit_spa->spa);
if (rc)
return rc;
}
list_sort(NULL, &acpi_desc->dimms, nfit_mem_cmp);
return 0;
}
static ssize_t revision_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
return sprintf(buf, "%d\n", acpi_desc->nfit->header.revision);
}
static DEVICE_ATTR_RO(revision);
static struct attribute *acpi_nfit_attributes[] = {
&dev_attr_revision.attr,
NULL,
};
static struct attribute_group acpi_nfit_attribute_group = {
.name = "nfit",
.attrs = acpi_nfit_attributes,
};
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
const struct attribute_group *acpi_nfit_attribute_groups[] = {
&nvdimm_bus_attribute_group,
&acpi_nfit_attribute_group,
NULL,
};
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
EXPORT_SYMBOL_GPL(acpi_nfit_attribute_groups);
static struct acpi_nfit_memory_map *to_nfit_memdev(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
return __to_nfit_memdev(nfit_mem);
}
static struct acpi_nfit_control_region *to_nfit_dcr(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
return nfit_mem->dcr;
}
static ssize_t handle_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_memory_map *memdev = to_nfit_memdev(dev);
return sprintf(buf, "%#x\n", memdev->device_handle);
}
static DEVICE_ATTR_RO(handle);
static ssize_t phys_id_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_memory_map *memdev = to_nfit_memdev(dev);
return sprintf(buf, "%#x\n", memdev->physical_id);
}
static DEVICE_ATTR_RO(phys_id);
static ssize_t vendor_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
return sprintf(buf, "%#x\n", dcr->vendor_id);
}
static DEVICE_ATTR_RO(vendor);
static ssize_t rev_id_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
return sprintf(buf, "%#x\n", dcr->revision_id);
}
static DEVICE_ATTR_RO(rev_id);
static ssize_t device_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
return sprintf(buf, "%#x\n", dcr->device_id);
}
static DEVICE_ATTR_RO(device);
static ssize_t format_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
return sprintf(buf, "%#x\n", dcr->code);
}
static DEVICE_ATTR_RO(format);
static ssize_t serial_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
return sprintf(buf, "%#x\n", dcr->serial_number);
}
static DEVICE_ATTR_RO(serial);
static ssize_t flags_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
u16 flags = to_nfit_memdev(dev)->flags;
return sprintf(buf, "%s%s%s%s%s\n",
2015-08-27 00:20:23 +08:00
flags & ACPI_NFIT_MEM_SAVE_FAILED ? "save_fail " : "",
flags & ACPI_NFIT_MEM_RESTORE_FAILED ? "restore_fail " : "",
flags & ACPI_NFIT_MEM_FLUSH_FAILED ? "flush_fail " : "",
flags & ACPI_NFIT_MEM_NOT_ARMED ? "not_armed " : "",
2015-08-27 00:20:23 +08:00
flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event " : "");
}
static DEVICE_ATTR_RO(flags);
static struct attribute *acpi_nfit_dimm_attributes[] = {
&dev_attr_handle.attr,
&dev_attr_phys_id.attr,
&dev_attr_vendor.attr,
&dev_attr_device.attr,
&dev_attr_format.attr,
&dev_attr_serial.attr,
&dev_attr_rev_id.attr,
&dev_attr_flags.attr,
NULL,
};
static umode_t acpi_nfit_dimm_attr_visible(struct kobject *kobj,
struct attribute *a, int n)
{
struct device *dev = container_of(kobj, struct device, kobj);
if (to_nfit_dcr(dev))
return a->mode;
else
return 0;
}
static struct attribute_group acpi_nfit_dimm_attribute_group = {
.name = "nfit",
.attrs = acpi_nfit_dimm_attributes,
.is_visible = acpi_nfit_dimm_attr_visible,
};
static const struct attribute_group *acpi_nfit_dimm_attribute_groups[] = {
2015-06-09 02:27:06 +08:00
&nvdimm_attribute_group,
&nd_device_attribute_group,
&acpi_nfit_dimm_attribute_group,
NULL,
};
static struct nvdimm *acpi_nfit_dimm_by_handle(struct acpi_nfit_desc *acpi_desc,
u32 device_handle)
{
struct nfit_mem *nfit_mem;
list_for_each_entry(nfit_mem, &acpi_desc->dimms, list)
if (__to_nfit_memdev(nfit_mem)->device_handle == device_handle)
return nfit_mem->nvdimm;
return NULL;
}
2015-06-09 02:27:06 +08:00
static int acpi_nfit_add_dimm(struct acpi_nfit_desc *acpi_desc,
struct nfit_mem *nfit_mem, u32 device_handle)
{
struct acpi_device *adev, *adev_dimm;
struct device *dev = acpi_desc->dev;
const u8 *uuid = to_nfit_uuid(NFIT_DEV_DIMM);
int i;
2015-06-09 02:27:06 +08:00
nfit_mem->dsm_mask = acpi_desc->dimm_dsm_force_en;
adev = to_acpi_dev(acpi_desc);
if (!adev)
return 0;
adev_dimm = acpi_find_child_device(adev, device_handle, false);
nfit_mem->adev = adev_dimm;
if (!adev_dimm) {
dev_err(dev, "no ACPI.NFIT device with _ADR %#x, disabling...\n",
device_handle);
return force_enable_dimms ? 0 : -ENODEV;
2015-06-09 02:27:06 +08:00
}
for (i = ND_CMD_SMART; i <= ND_CMD_VENDOR; i++)
if (acpi_check_dsm(adev_dimm->handle, uuid, 1, 1ULL << i))
set_bit(i, &nfit_mem->dsm_mask);
return 0;
2015-06-09 02:27:06 +08:00
}
static int acpi_nfit_register_dimms(struct acpi_nfit_desc *acpi_desc)
{
struct nfit_mem *nfit_mem;
int dimm_count = 0;
list_for_each_entry(nfit_mem, &acpi_desc->dimms, list) {
struct nvdimm *nvdimm;
unsigned long flags = 0;
u32 device_handle;
u16 mem_flags;
2015-06-09 02:27:06 +08:00
int rc;
device_handle = __to_nfit_memdev(nfit_mem)->device_handle;
nvdimm = acpi_nfit_dimm_by_handle(acpi_desc, device_handle);
if (nvdimm) {
dimm_count++;
continue;
}
if (nfit_mem->bdw && nfit_mem->memdev_pmem)
flags |= NDD_ALIASING;
mem_flags = __to_nfit_memdev(nfit_mem)->flags;
if (mem_flags & ACPI_NFIT_MEM_NOT_ARMED)
flags |= NDD_UNARMED;
2015-06-09 02:27:06 +08:00
rc = acpi_nfit_add_dimm(acpi_desc, nfit_mem, device_handle);
if (rc)
continue;
nvdimm = nvdimm_create(acpi_desc->nvdimm_bus, nfit_mem,
2015-06-09 02:27:06 +08:00
acpi_nfit_dimm_attribute_groups,
flags, &nfit_mem->dsm_mask);
if (!nvdimm)
return -ENOMEM;
nfit_mem->nvdimm = nvdimm;
dimm_count++;
if ((mem_flags & ACPI_NFIT_MEM_FAILED_MASK) == 0)
continue;
2015-08-27 00:20:23 +08:00
dev_info(acpi_desc->dev, "%s flags:%s%s%s%s\n",
nvdimm_name(nvdimm),
2015-08-27 00:20:23 +08:00
mem_flags & ACPI_NFIT_MEM_SAVE_FAILED ? " save_fail" : "",
mem_flags & ACPI_NFIT_MEM_RESTORE_FAILED ? " restore_fail":"",
mem_flags & ACPI_NFIT_MEM_FLUSH_FAILED ? " flush_fail" : "",
mem_flags & ACPI_NFIT_MEM_NOT_ARMED ? " not_armed" : "");
}
return nvdimm_bus_check_dimm_count(acpi_desc->nvdimm_bus, dimm_count);
}
2015-06-09 02:27:06 +08:00
static void acpi_nfit_init_dsms(struct acpi_nfit_desc *acpi_desc)
{
struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
const u8 *uuid = to_nfit_uuid(NFIT_DEV_BUS);
struct acpi_device *adev;
int i;
nd_desc->dsm_mask = acpi_desc->bus_dsm_force_en;
2015-06-09 02:27:06 +08:00
adev = to_acpi_dev(acpi_desc);
if (!adev)
return;
for (i = ND_CMD_ARS_CAP; i <= ND_CMD_ARS_STATUS; i++)
if (acpi_check_dsm(adev->handle, uuid, 1, 1ULL << i))
set_bit(i, &nd_desc->dsm_mask);
}
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
static ssize_t range_index_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nd_region *nd_region = to_nd_region(dev);
struct nfit_spa *nfit_spa = nd_region_provider_data(nd_region);
return sprintf(buf, "%d\n", nfit_spa->spa->range_index);
}
static DEVICE_ATTR_RO(range_index);
static struct attribute *acpi_nfit_region_attributes[] = {
&dev_attr_range_index.attr,
NULL,
};
static struct attribute_group acpi_nfit_region_attribute_group = {
.name = "nfit",
.attrs = acpi_nfit_region_attributes,
};
static const struct attribute_group *acpi_nfit_region_attribute_groups[] = {
&nd_region_attribute_group,
&nd_mapping_attribute_group,
&nd_device_attribute_group,
&nd_numa_attribute_group,
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
&acpi_nfit_region_attribute_group,
NULL,
};
2015-05-02 01:11:27 +08:00
/* enough info to uniquely specify an interleave set */
struct nfit_set_info {
struct nfit_set_info_map {
u64 region_offset;
u32 serial_number;
u32 pad;
} mapping[0];
};
static size_t sizeof_nfit_set_info(int num_mappings)
{
return sizeof(struct nfit_set_info)
+ num_mappings * sizeof(struct nfit_set_info_map);
}
static int cmp_map(const void *m0, const void *m1)
{
const struct nfit_set_info_map *map0 = m0;
const struct nfit_set_info_map *map1 = m1;
return memcmp(&map0->region_offset, &map1->region_offset,
sizeof(u64));
}
/* Retrieve the nth entry referencing this spa */
static struct acpi_nfit_memory_map *memdev_from_spa(
struct acpi_nfit_desc *acpi_desc, u16 range_index, int n)
{
struct nfit_memdev *nfit_memdev;
list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list)
if (nfit_memdev->memdev->range_index == range_index)
if (n-- == 0)
return nfit_memdev->memdev;
return NULL;
}
static int acpi_nfit_init_interleave_set(struct acpi_nfit_desc *acpi_desc,
struct nd_region_desc *ndr_desc,
struct acpi_nfit_system_address *spa)
{
int i, spa_type = nfit_spa_type(spa);
struct device *dev = acpi_desc->dev;
struct nd_interleave_set *nd_set;
u16 nr = ndr_desc->num_mappings;
struct nfit_set_info *info;
if (spa_type == NFIT_SPA_PM || spa_type == NFIT_SPA_VOLATILE)
/* pass */;
else
return 0;
nd_set = devm_kzalloc(dev, sizeof(*nd_set), GFP_KERNEL);
if (!nd_set)
return -ENOMEM;
info = devm_kzalloc(dev, sizeof_nfit_set_info(nr), GFP_KERNEL);
if (!info)
return -ENOMEM;
for (i = 0; i < nr; i++) {
struct nd_mapping *nd_mapping = &ndr_desc->nd_mapping[i];
struct nfit_set_info_map *map = &info->mapping[i];
struct nvdimm *nvdimm = nd_mapping->nvdimm;
struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
struct acpi_nfit_memory_map *memdev = memdev_from_spa(acpi_desc,
spa->range_index, i);
if (!memdev || !nfit_mem->dcr) {
dev_err(dev, "%s: failed to find DCR\n", __func__);
return -ENODEV;
}
map->region_offset = memdev->region_offset;
map->serial_number = nfit_mem->dcr->serial_number;
}
sort(&info->mapping[0], nr, sizeof(struct nfit_set_info_map),
cmp_map, NULL);
nd_set->cookie = nd_fletcher64(info, sizeof_nfit_set_info(nr), 0);
ndr_desc->nd_set = nd_set;
devm_kfree(dev, info);
return 0;
}
static u64 to_interleave_offset(u64 offset, struct nfit_blk_mmio *mmio)
{
struct acpi_nfit_interleave *idt = mmio->idt;
u32 sub_line_offset, line_index, line_offset;
u64 line_no, table_skip_count, table_offset;
line_no = div_u64_rem(offset, mmio->line_size, &sub_line_offset);
table_skip_count = div_u64_rem(line_no, mmio->num_lines, &line_index);
line_offset = idt->line_offset[line_index]
* mmio->line_size;
table_offset = table_skip_count * mmio->table_size;
return mmio->base_offset + line_offset + table_offset + sub_line_offset;
}
static void wmb_blk(struct nfit_blk *nfit_blk)
{
if (nfit_blk->nvdimm_flush) {
/*
* The first wmb() is needed to 'sfence' all previous writes
* such that they are architecturally visible for the platform
* buffer flush. Note that we've already arranged for pmem
* writes to avoid the cache via arch_memcpy_to_pmem(). The
* final wmb() ensures ordering for the NVDIMM flush write.
*/
wmb();
writeq(1, nfit_blk->nvdimm_flush);
wmb();
} else
wmb_pmem();
}
static u32 read_blk_stat(struct nfit_blk *nfit_blk, unsigned int bw)
{
struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
u64 offset = nfit_blk->stat_offset + mmio->size * bw;
if (mmio->num_lines)
offset = to_interleave_offset(offset, mmio);
libnvdimm for 4.3: 1/ Introduce ZONE_DEVICE and devm_memremap_pages() as a generic mechanism for adding device-driver-discovered memory regions to the kernel's direct map. This facility is used by the pmem driver to enable pfn_to_page() operations on the page frames returned by DAX ('direct_access' in 'struct block_device_operations'). For now, the 'memmap' allocation for these "device" pages comes from "System RAM". Support for allocating the memmap from device memory will arrive in a later kernel. 2/ Introduce memremap() to replace usages of ioremap_cache() and ioremap_wt(). memremap() drops the __iomem annotation for these mappings to memory that do not have i/o side effects. The replacement of ioremap_cache() with memremap() is limited to the pmem driver to ease merging the api change in v4.3. Completion of the conversion is targeted for v4.4. 3/ Similar to the usage of memcpy_to_pmem() + wmb_pmem() in the pmem driver, update the VFS DAX implementation and PMEM api to provide persistence guarantees for kernel operations on a DAX mapping. 4/ Convert the ACPI NFIT 'BLK' driver to map the block apertures as cacheable to improve performance. 5/ Miscellaneous updates and fixes to libnvdimm including support for issuing "address range scrub" commands, clarifying the optimal 'sector size' of pmem devices, a clarification of the usage of the ACPI '_STA' (status) property for DIMM devices, and other minor fixes. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1 iQIcBAABAgAGBQJV6Nx7AAoJEB7SkWpmfYgCWyYQAI5ju6Gvw27RNFtPovHcZUf5 JGnxXejI6/AqeTQ+IulgprxtEUCrXOHjCDA5dkjr1qvsoqK1qxug+vJHOZLgeW0R OwDtmdW4Qrgeqm+CPoxETkorJ8wDOc8mol81kTiMgeV3UqbYeeHIiTAmwe7VzZ0C nNdCRDm5g8dHCjTKcvK3rvozgyoNoWeBiHkPe76EbnxDICxCB5dak7XsVKNMIVFQ NuYlnw6IYN7+rMHgpgpRux38NtIW8VlYPWTmHExejc2mlioWMNBG/bmtwLyJ6M3e zliz4/cnonTMUaizZaVozyinTa65m7wcnpjK+vlyGV2deDZPJpDRvSOtB0lH30bR 1gy+qrKzuGKpaN6thOISxFLLjmEeYwzYd7SvC9n118r32qShz+opN9XX0WmWSFlA sajE1ehm4M7s5pkMoa/dRnAyR8RUPu4RNINdQ/Z9jFfAOx+Q26rLdQXwf9+uqbEb bIeSQwOteK5vYYCstvpAcHSMlJAglzIX5UfZBvtEIJN7rlb0VhmGWfxAnTu+ktG1 o9cqAt+J4146xHaFwj5duTsyKhWb8BL9+xqbKPNpXEp+PbLsrnE/+WkDLFD67jxz dgIoK60mGnVXp+16I2uMqYYDgAyO5zUdmM4OygOMnZNa1mxesjbDJC6Wat1Wsndn slsw6DkrWT60CRE42nbK =o57/ -----END PGP SIGNATURE----- Merge tag 'libnvdimm-for-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm Pull libnvdimm updates from Dan Williams: "This update has successfully completed a 0day-kbuild run and has appeared in a linux-next release. The changes outside of the typical drivers/nvdimm/ and drivers/acpi/nfit.[ch] paths are related to the removal of IORESOURCE_CACHEABLE, the introduction of memremap(), and the introduction of ZONE_DEVICE + devm_memremap_pages(). Summary: - Introduce ZONE_DEVICE and devm_memremap_pages() as a generic mechanism for adding device-driver-discovered memory regions to the kernel's direct map. This facility is used by the pmem driver to enable pfn_to_page() operations on the page frames returned by DAX ('direct_access' in 'struct block_device_operations'). For now, the 'memmap' allocation for these "device" pages comes from "System RAM". Support for allocating the memmap from device memory will arrive in a later kernel. - Introduce memremap() to replace usages of ioremap_cache() and ioremap_wt(). memremap() drops the __iomem annotation for these mappings to memory that do not have i/o side effects. The replacement of ioremap_cache() with memremap() is limited to the pmem driver to ease merging the api change in v4.3. Completion of the conversion is targeted for v4.4. - Similar to the usage of memcpy_to_pmem() + wmb_pmem() in the pmem driver, update the VFS DAX implementation and PMEM api to provide persistence guarantees for kernel operations on a DAX mapping. - Convert the ACPI NFIT 'BLK' driver to map the block apertures as cacheable to improve performance. - Miscellaneous updates and fixes to libnvdimm including support for issuing "address range scrub" commands, clarifying the optimal 'sector size' of pmem devices, a clarification of the usage of the ACPI '_STA' (status) property for DIMM devices, and other minor fixes" * tag 'libnvdimm-for-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: (34 commits) libnvdimm, pmem: direct map legacy pmem by default libnvdimm, pmem: 'struct page' for pmem libnvdimm, pfn: 'struct page' provider infrastructure x86, pmem: clarify that ARCH_HAS_PMEM_API implies PMEM mapped WB add devm_memremap_pages mm: ZONE_DEVICE for "device memory" mm: move __phys_to_pfn and __pfn_to_phys to asm/generic/memory_model.h dax: drop size parameter to ->direct_access() nd_blk: change aperture mapping from WC to WB nvdimm: change to use generic kvfree() pmem, dax: have direct_access use __pmem annotation dax: update I/O path to do proper PMEM flushing pmem: add copy_from_iter_pmem() and clear_pmem() pmem, x86: clean up conditional pmem includes pmem: remove layer when calling arch_has_wmb_pmem() pmem, x86: move x86 PMEM API to new pmem.h header libnvdimm, e820: make CONFIG_X86_PMEM_LEGACY a tristate option pmem: switch to devm_ allocations devres: add devm_memremap libnvdimm, btt: write and validate parent_uuid ...
2015-09-09 05:35:59 +08:00
return readl(mmio->addr.base + offset);
}
static void write_blk_ctl(struct nfit_blk *nfit_blk, unsigned int bw,
resource_size_t dpa, unsigned int len, unsigned int write)
{
u64 cmd, offset;
struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
enum {
BCW_OFFSET_MASK = (1ULL << 48)-1,
BCW_LEN_SHIFT = 48,
BCW_LEN_MASK = (1ULL << 8) - 1,
BCW_CMD_SHIFT = 56,
};
cmd = (dpa >> L1_CACHE_SHIFT) & BCW_OFFSET_MASK;
len = len >> L1_CACHE_SHIFT;
cmd |= ((u64) len & BCW_LEN_MASK) << BCW_LEN_SHIFT;
cmd |= ((u64) write) << BCW_CMD_SHIFT;
offset = nfit_blk->cmd_offset + mmio->size * bw;
if (mmio->num_lines)
offset = to_interleave_offset(offset, mmio);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
writeq(cmd, mmio->addr.base + offset);
wmb_blk(nfit_blk);
if (nfit_blk->dimm_flags & ND_BLK_DCR_LATCH)
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
readq(mmio->addr.base + offset);
}
static int acpi_nfit_blk_single_io(struct nfit_blk *nfit_blk,
resource_size_t dpa, void *iobuf, size_t len, int rw,
unsigned int lane)
{
struct nfit_blk_mmio *mmio = &nfit_blk->mmio[BDW];
unsigned int copied = 0;
u64 base_offset;
int rc;
base_offset = nfit_blk->bdw_offset + dpa % L1_CACHE_BYTES
+ lane * mmio->size;
write_blk_ctl(nfit_blk, lane, dpa, len, rw);
while (len) {
unsigned int c;
u64 offset;
if (mmio->num_lines) {
u32 line_offset;
offset = to_interleave_offset(base_offset + copied,
mmio);
div_u64_rem(offset, mmio->line_size, &line_offset);
c = min_t(size_t, len, mmio->line_size - line_offset);
} else {
offset = base_offset + nfit_blk->bdw_offset;
c = len;
}
if (rw)
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
memcpy_to_pmem(mmio->addr.aperture + offset,
iobuf + copied, c);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
else {
if (nfit_blk->dimm_flags & ND_BLK_READ_FLUSH)
mmio_flush_range((void __force *)
mmio->addr.aperture + offset, c);
memcpy_from_pmem(iobuf + copied,
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
mmio->addr.aperture + offset, c);
}
copied += c;
len -= c;
}
if (rw)
wmb_blk(nfit_blk);
rc = read_blk_stat(nfit_blk, lane) ? -EIO : 0;
return rc;
}
static int acpi_nfit_blk_region_do_io(struct nd_blk_region *ndbr,
resource_size_t dpa, void *iobuf, u64 len, int rw)
{
struct nfit_blk *nfit_blk = nd_blk_region_provider_data(ndbr);
struct nfit_blk_mmio *mmio = &nfit_blk->mmio[BDW];
struct nd_region *nd_region = nfit_blk->nd_region;
unsigned int lane, copied = 0;
int rc = 0;
lane = nd_region_acquire_lane(nd_region);
while (len) {
u64 c = min(len, mmio->size);
rc = acpi_nfit_blk_single_io(nfit_blk, dpa + copied,
iobuf + copied, c, rw, lane);
if (rc)
break;
copied += c;
len -= c;
}
nd_region_release_lane(nd_region, lane);
return rc;
}
static void nfit_spa_mapping_release(struct kref *kref)
{
struct nfit_spa_mapping *spa_map = to_spa_map(kref);
struct acpi_nfit_system_address *spa = spa_map->spa;
struct acpi_nfit_desc *acpi_desc = spa_map->acpi_desc;
WARN_ON(!mutex_is_locked(&acpi_desc->spa_map_mutex));
dev_dbg(acpi_desc->dev, "%s: SPA%d\n", __func__, spa->range_index);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
if (spa_map->type == SPA_MAP_APERTURE)
memunmap((void __force *)spa_map->addr.aperture);
else
iounmap(spa_map->addr.base);
release_mem_region(spa->address, spa->length);
list_del(&spa_map->list);
kfree(spa_map);
}
static struct nfit_spa_mapping *find_spa_mapping(
struct acpi_nfit_desc *acpi_desc,
struct acpi_nfit_system_address *spa)
{
struct nfit_spa_mapping *spa_map;
WARN_ON(!mutex_is_locked(&acpi_desc->spa_map_mutex));
list_for_each_entry(spa_map, &acpi_desc->spa_maps, list)
if (spa_map->spa == spa)
return spa_map;
return NULL;
}
static void nfit_spa_unmap(struct acpi_nfit_desc *acpi_desc,
struct acpi_nfit_system_address *spa)
{
struct nfit_spa_mapping *spa_map;
mutex_lock(&acpi_desc->spa_map_mutex);
spa_map = find_spa_mapping(acpi_desc, spa);
if (spa_map)
kref_put(&spa_map->kref, nfit_spa_mapping_release);
mutex_unlock(&acpi_desc->spa_map_mutex);
}
static void __iomem *__nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
struct acpi_nfit_system_address *spa, enum spa_map_type type)
{
resource_size_t start = spa->address;
resource_size_t n = spa->length;
struct nfit_spa_mapping *spa_map;
struct resource *res;
WARN_ON(!mutex_is_locked(&acpi_desc->spa_map_mutex));
spa_map = find_spa_mapping(acpi_desc, spa);
if (spa_map) {
kref_get(&spa_map->kref);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
return spa_map->addr.base;
}
spa_map = kzalloc(sizeof(*spa_map), GFP_KERNEL);
if (!spa_map)
return NULL;
INIT_LIST_HEAD(&spa_map->list);
spa_map->spa = spa;
kref_init(&spa_map->kref);
spa_map->acpi_desc = acpi_desc;
res = request_mem_region(start, n, dev_name(acpi_desc->dev));
if (!res)
goto err_mem;
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
spa_map->type = type;
if (type == SPA_MAP_APERTURE)
spa_map->addr.aperture = (void __pmem *)memremap(start, n,
ARCH_MEMREMAP_PMEM);
else
spa_map->addr.base = ioremap_nocache(start, n);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
if (!spa_map->addr.base)
goto err_map;
list_add_tail(&spa_map->list, &acpi_desc->spa_maps);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
return spa_map->addr.base;
err_map:
release_mem_region(start, n);
err_mem:
kfree(spa_map);
return NULL;
}
/**
* nfit_spa_map - interleave-aware managed-mappings of acpi_nfit_system_address ranges
* @nvdimm_bus: NFIT-bus that provided the spa table entry
* @nfit_spa: spa table to map
* @type: aperture or control region
*
* In the case where block-data-window apertures and
* dimm-control-regions are interleaved they will end up sharing a
* single request_mem_region() + ioremap() for the address range. In
* the style of devm nfit_spa_map() mappings are automatically dropped
* when all region devices referencing the same mapping are disabled /
* unbound.
*/
static void __iomem *nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
struct acpi_nfit_system_address *spa, enum spa_map_type type)
{
void __iomem *iomem;
mutex_lock(&acpi_desc->spa_map_mutex);
iomem = __nfit_spa_map(acpi_desc, spa, type);
mutex_unlock(&acpi_desc->spa_map_mutex);
return iomem;
}
static int nfit_blk_init_interleave(struct nfit_blk_mmio *mmio,
struct acpi_nfit_interleave *idt, u16 interleave_ways)
{
if (idt) {
mmio->num_lines = idt->line_count;
mmio->line_size = idt->line_size;
if (interleave_ways == 0)
return -ENXIO;
mmio->table_size = mmio->num_lines * interleave_ways
* mmio->line_size;
}
return 0;
}
static int acpi_nfit_blk_get_flags(struct nvdimm_bus_descriptor *nd_desc,
struct nvdimm *nvdimm, struct nfit_blk *nfit_blk)
{
struct nd_cmd_dimm_flags flags;
int rc;
memset(&flags, 0, sizeof(flags));
rc = nd_desc->ndctl(nd_desc, nvdimm, ND_CMD_DIMM_FLAGS, &flags,
sizeof(flags));
if (rc >= 0 && flags.status == 0)
nfit_blk->dimm_flags = flags.flags;
else if (rc == -ENOTTY) {
/* fall back to a conservative default */
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
nfit_blk->dimm_flags = ND_BLK_DCR_LATCH | ND_BLK_READ_FLUSH;
rc = 0;
} else
rc = -ENXIO;
return rc;
}
static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
struct device *dev)
{
struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
struct nd_blk_region *ndbr = to_nd_blk_region(dev);
struct nfit_flush *nfit_flush;
struct nfit_blk_mmio *mmio;
struct nfit_blk *nfit_blk;
struct nfit_mem *nfit_mem;
struct nvdimm *nvdimm;
int rc;
nvdimm = nd_blk_region_to_dimm(ndbr);
nfit_mem = nvdimm_provider_data(nvdimm);
if (!nfit_mem || !nfit_mem->dcr || !nfit_mem->bdw) {
dev_dbg(dev, "%s: missing%s%s%s\n", __func__,
nfit_mem ? "" : " nfit_mem",
(nfit_mem && nfit_mem->dcr) ? "" : " dcr",
(nfit_mem && nfit_mem->bdw) ? "" : " bdw");
return -ENXIO;
}
nfit_blk = devm_kzalloc(dev, sizeof(*nfit_blk), GFP_KERNEL);
if (!nfit_blk)
return -ENOMEM;
nd_blk_region_set_provider_data(ndbr, nfit_blk);
nfit_blk->nd_region = to_nd_region(dev);
/* map block aperture memory */
nfit_blk->bdw_offset = nfit_mem->bdw->offset;
mmio = &nfit_blk->mmio[BDW];
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
mmio->addr.base = nfit_spa_map(acpi_desc, nfit_mem->spa_bdw,
SPA_MAP_APERTURE);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
if (!mmio->addr.base) {
dev_dbg(dev, "%s: %s failed to map bdw\n", __func__,
nvdimm_name(nvdimm));
return -ENOMEM;
}
mmio->size = nfit_mem->bdw->size;
mmio->base_offset = nfit_mem->memdev_bdw->region_offset;
mmio->idt = nfit_mem->idt_bdw;
mmio->spa = nfit_mem->spa_bdw;
rc = nfit_blk_init_interleave(mmio, nfit_mem->idt_bdw,
nfit_mem->memdev_bdw->interleave_ways);
if (rc) {
dev_dbg(dev, "%s: %s failed to init bdw interleave\n",
__func__, nvdimm_name(nvdimm));
return rc;
}
/* map block control memory */
nfit_blk->cmd_offset = nfit_mem->dcr->command_offset;
nfit_blk->stat_offset = nfit_mem->dcr->status_offset;
mmio = &nfit_blk->mmio[DCR];
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
mmio->addr.base = nfit_spa_map(acpi_desc, nfit_mem->spa_dcr,
SPA_MAP_CONTROL);
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
if (!mmio->addr.base) {
dev_dbg(dev, "%s: %s failed to map dcr\n", __func__,
nvdimm_name(nvdimm));
return -ENOMEM;
}
mmio->size = nfit_mem->dcr->window_size;
mmio->base_offset = nfit_mem->memdev_dcr->region_offset;
mmio->idt = nfit_mem->idt_dcr;
mmio->spa = nfit_mem->spa_dcr;
rc = nfit_blk_init_interleave(mmio, nfit_mem->idt_dcr,
nfit_mem->memdev_dcr->interleave_ways);
if (rc) {
dev_dbg(dev, "%s: %s failed to init dcr interleave\n",
__func__, nvdimm_name(nvdimm));
return rc;
}
rc = acpi_nfit_blk_get_flags(nd_desc, nvdimm, nfit_blk);
if (rc < 0) {
dev_dbg(dev, "%s: %s failed get DIMM flags\n",
__func__, nvdimm_name(nvdimm));
return rc;
}
nfit_flush = nfit_mem->nfit_flush;
if (nfit_flush && nfit_flush->flush->hint_count != 0) {
nfit_blk->nvdimm_flush = devm_ioremap_nocache(dev,
nfit_flush->flush->hint_address[0], 8);
if (!nfit_blk->nvdimm_flush)
return -ENOMEM;
}
x86, pmem: clarify that ARCH_HAS_PMEM_API implies PMEM mapped WB Given that a write-back (WB) mapping plus non-temporal stores is expected to be the most efficient way to access PMEM, update the definition of ARCH_HAS_PMEM_API to imply arch support for WB-mapped-PMEM. This is needed as a pre-requisite for adding PMEM to the direct map and mapping it with struct page. The above clarification for X86_64 means that memcpy_to_pmem() is permitted to use the non-temporal arch_memcpy_to_pmem() rather than needlessly fall back to default_memcpy_to_pmem() when the pcommit instruction is not available. When arch_memcpy_to_pmem() is not guaranteed to flush writes out of cache, i.e. on older X86_32 implementations where non-temporal stores may just dirty cache, ARCH_HAS_PMEM_API is simply disabled. The default fall back for persistent memory handling remains. Namely, map it with the WT (write-through) cache-type and hope for the best. arch_has_pmem_api() is updated to only indicate whether the arch provides the proper helpers to meet the minimum "writes are visible outside the cache hierarchy after memcpy_to_pmem() + wmb_pmem()". Code that cares whether wmb_pmem() actually flushes writes to pmem must now call arch_has_wmb_pmem() directly. Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> [hch: set ARCH_HAS_PMEM_API=n on x86_32] Reviewed-by: Christoph Hellwig <hch@lst.de> [toshi: x86_32 compile fixes] Signed-off-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-25 06:29:38 +08:00
if (!arch_has_wmb_pmem() && !nfit_blk->nvdimm_flush)
dev_warn(dev, "unable to guarantee persistence of writes\n");
if (mmio->line_size == 0)
return 0;
if ((u32) nfit_blk->cmd_offset % mmio->line_size
+ 8 > mmio->line_size) {
dev_dbg(dev, "cmd_offset crosses interleave boundary\n");
return -ENXIO;
} else if ((u32) nfit_blk->stat_offset % mmio->line_size
+ 8 > mmio->line_size) {
dev_dbg(dev, "stat_offset crosses interleave boundary\n");
return -ENXIO;
}
return 0;
}
static void acpi_nfit_blk_region_disable(struct nvdimm_bus *nvdimm_bus,
struct device *dev)
{
struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
struct nd_blk_region *ndbr = to_nd_blk_region(dev);
struct nfit_blk *nfit_blk = nd_blk_region_provider_data(ndbr);
int i;
if (!nfit_blk)
return; /* never enabled */
/* auto-free BLK spa mappings */
for (i = 0; i < 2; i++) {
struct nfit_blk_mmio *mmio = &nfit_blk->mmio[i];
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
if (mmio->addr.base)
nfit_spa_unmap(acpi_desc, mmio->spa);
}
nd_blk_region_set_provider_data(ndbr, NULL);
/* devm will free nfit_blk */
}
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
static int acpi_nfit_init_mapping(struct acpi_nfit_desc *acpi_desc,
struct nd_mapping *nd_mapping, struct nd_region_desc *ndr_desc,
struct acpi_nfit_memory_map *memdev,
struct acpi_nfit_system_address *spa)
{
struct nvdimm *nvdimm = acpi_nfit_dimm_by_handle(acpi_desc,
memdev->device_handle);
struct nd_blk_region_desc *ndbr_desc;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
struct nfit_mem *nfit_mem;
int blk_valid = 0;
if (!nvdimm) {
dev_err(acpi_desc->dev, "spa%d dimm: %#x not found\n",
spa->range_index, memdev->device_handle);
return -ENODEV;
}
nd_mapping->nvdimm = nvdimm;
switch (nfit_spa_type(spa)) {
case NFIT_SPA_PM:
case NFIT_SPA_VOLATILE:
nd_mapping->start = memdev->address;
nd_mapping->size = memdev->region_size;
break;
case NFIT_SPA_DCR:
nfit_mem = nvdimm_provider_data(nvdimm);
if (!nfit_mem || !nfit_mem->bdw) {
dev_dbg(acpi_desc->dev, "spa%d %s missing bdw\n",
spa->range_index, nvdimm_name(nvdimm));
} else {
nd_mapping->size = nfit_mem->bdw->capacity;
nd_mapping->start = nfit_mem->bdw->start_address;
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
ndr_desc->num_lanes = nfit_mem->bdw->windows;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
blk_valid = 1;
}
ndr_desc->nd_mapping = nd_mapping;
ndr_desc->num_mappings = blk_valid;
ndbr_desc = to_blk_region_desc(ndr_desc);
ndbr_desc->enable = acpi_nfit_blk_region_enable;
ndbr_desc->disable = acpi_nfit_blk_region_disable;
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
ndbr_desc->do_io = acpi_desc->blk_do_io;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
if (!nvdimm_blk_region_create(acpi_desc->nvdimm_bus, ndr_desc))
return -ENOMEM;
break;
}
return 0;
}
static int acpi_nfit_register_region(struct acpi_nfit_desc *acpi_desc,
struct nfit_spa *nfit_spa)
{
static struct nd_mapping nd_mappings[ND_MAX_MAPPINGS];
struct acpi_nfit_system_address *spa = nfit_spa->spa;
struct nd_blk_region_desc ndbr_desc;
struct nd_region_desc *ndr_desc;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
struct nfit_memdev *nfit_memdev;
struct nvdimm_bus *nvdimm_bus;
struct resource res;
2015-05-02 01:11:27 +08:00
int count = 0, rc;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
if (nfit_spa->is_registered)
return 0;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
if (spa->range_index == 0) {
dev_dbg(acpi_desc->dev, "%s: detected invalid spa index\n",
__func__);
return 0;
}
memset(&res, 0, sizeof(res));
memset(&nd_mappings, 0, sizeof(nd_mappings));
memset(&ndbr_desc, 0, sizeof(ndbr_desc));
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
res.start = spa->address;
res.end = res.start + spa->length - 1;
ndr_desc = &ndbr_desc.ndr_desc;
ndr_desc->res = &res;
ndr_desc->provider_data = nfit_spa;
ndr_desc->attr_groups = acpi_nfit_region_attribute_groups;
if (spa->flags & ACPI_NFIT_PROXIMITY_VALID)
ndr_desc->numa_node = acpi_map_pxm_to_online_node(
spa->proximity_domain);
else
ndr_desc->numa_node = NUMA_NO_NODE;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
struct nd_mapping *nd_mapping;
if (memdev->range_index != spa->range_index)
continue;
if (count >= ND_MAX_MAPPINGS) {
dev_err(acpi_desc->dev, "spa%d exceeds max mappings %d\n",
spa->range_index, ND_MAX_MAPPINGS);
return -ENXIO;
}
nd_mapping = &nd_mappings[count++];
rc = acpi_nfit_init_mapping(acpi_desc, nd_mapping, ndr_desc,
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
memdev, spa);
if (rc)
return rc;
}
ndr_desc->nd_mapping = nd_mappings;
ndr_desc->num_mappings = count;
rc = acpi_nfit_init_interleave_set(acpi_desc, ndr_desc, spa);
2015-05-02 01:11:27 +08:00
if (rc)
return rc;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
nvdimm_bus = acpi_desc->nvdimm_bus;
if (nfit_spa_type(spa) == NFIT_SPA_PM) {
if (!nvdimm_pmem_region_create(nvdimm_bus, ndr_desc))
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
return -ENOMEM;
} else if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE) {
if (!nvdimm_volatile_region_create(nvdimm_bus, ndr_desc))
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
return -ENOMEM;
}
nfit_spa->is_registered = 1;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
return 0;
}
static int acpi_nfit_register_regions(struct acpi_nfit_desc *acpi_desc)
{
struct nfit_spa *nfit_spa;
list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
int rc = acpi_nfit_register_region(acpi_desc, nfit_spa);
if (rc)
return rc;
}
return 0;
}
static int acpi_nfit_check_deletions(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev)
{
struct device *dev = acpi_desc->dev;
if (!list_empty(&prev->spas) ||
!list_empty(&prev->memdevs) ||
!list_empty(&prev->dcrs) ||
!list_empty(&prev->bdws) ||
!list_empty(&prev->idts) ||
!list_empty(&prev->flushes)) {
dev_err(dev, "new nfit deletes entries (unsupported)\n");
return -ENXIO;
}
return 0;
}
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, acpi_size sz)
{
struct device *dev = acpi_desc->dev;
struct nfit_table_prev prev;
const void *end;
u8 *data;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
int rc;
mutex_lock(&acpi_desc->init_mutex);
INIT_LIST_HEAD(&prev.spas);
INIT_LIST_HEAD(&prev.memdevs);
INIT_LIST_HEAD(&prev.dcrs);
INIT_LIST_HEAD(&prev.bdws);
INIT_LIST_HEAD(&prev.idts);
INIT_LIST_HEAD(&prev.flushes);
list_cut_position(&prev.spas, &acpi_desc->spas,
acpi_desc->spas.prev);
list_cut_position(&prev.memdevs, &acpi_desc->memdevs,
acpi_desc->memdevs.prev);
list_cut_position(&prev.dcrs, &acpi_desc->dcrs,
acpi_desc->dcrs.prev);
list_cut_position(&prev.bdws, &acpi_desc->bdws,
acpi_desc->bdws.prev);
list_cut_position(&prev.idts, &acpi_desc->idts,
acpi_desc->idts.prev);
list_cut_position(&prev.flushes, &acpi_desc->flushes,
acpi_desc->flushes.prev);
data = (u8 *) acpi_desc->nfit;
end = data + sz;
data += sizeof(struct acpi_table_nfit);
while (!IS_ERR_OR_NULL(data))
data = add_table(acpi_desc, &prev, data, end);
if (IS_ERR(data)) {
dev_dbg(dev, "%s: nfit table parsing error: %ld\n", __func__,
PTR_ERR(data));
rc = PTR_ERR(data);
goto out_unlock;
}
rc = acpi_nfit_check_deletions(acpi_desc, &prev);
if (rc)
goto out_unlock;
if (nfit_mem_init(acpi_desc) != 0) {
rc = -ENOMEM;
goto out_unlock;
}
2015-06-09 02:27:06 +08:00
acpi_nfit_init_dsms(acpi_desc);
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
rc = acpi_nfit_register_dimms(acpi_desc);
if (rc)
goto out_unlock;
rc = acpi_nfit_register_regions(acpi_desc);
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
out_unlock:
mutex_unlock(&acpi_desc->init_mutex);
return rc;
}
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
EXPORT_SYMBOL_GPL(acpi_nfit_init);
static struct acpi_nfit_desc *acpi_nfit_desc_init(struct acpi_device *adev)
{
struct nvdimm_bus_descriptor *nd_desc;
struct acpi_nfit_desc *acpi_desc;
struct device *dev = &adev->dev;
acpi_desc = devm_kzalloc(dev, sizeof(*acpi_desc), GFP_KERNEL);
if (!acpi_desc)
return ERR_PTR(-ENOMEM);
dev_set_drvdata(dev, acpi_desc);
acpi_desc->dev = dev;
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
acpi_desc->blk_do_io = acpi_nfit_blk_region_do_io;
nd_desc = &acpi_desc->nd_desc;
nd_desc->provider_name = "ACPI.NFIT";
nd_desc->ndctl = acpi_nfit_ctl;
nd_desc->attr_groups = acpi_nfit_attribute_groups;
acpi_desc->nvdimm_bus = nvdimm_bus_register(dev, nd_desc);
if (!acpi_desc->nvdimm_bus) {
devm_kfree(dev, acpi_desc);
return ERR_PTR(-ENXIO);
}
INIT_LIST_HEAD(&acpi_desc->spa_maps);
INIT_LIST_HEAD(&acpi_desc->spas);
INIT_LIST_HEAD(&acpi_desc->dcrs);
INIT_LIST_HEAD(&acpi_desc->bdws);
INIT_LIST_HEAD(&acpi_desc->idts);
INIT_LIST_HEAD(&acpi_desc->flushes);
INIT_LIST_HEAD(&acpi_desc->memdevs);
INIT_LIST_HEAD(&acpi_desc->dimms);
mutex_init(&acpi_desc->spa_map_mutex);
mutex_init(&acpi_desc->init_mutex);
return acpi_desc;
}
static int acpi_nfit_add(struct acpi_device *adev)
{
struct acpi_buffer buf = { ACPI_ALLOCATE_BUFFER, NULL };
struct acpi_nfit_desc *acpi_desc;
struct device *dev = &adev->dev;
struct acpi_table_header *tbl;
acpi_status status = AE_OK;
acpi_size sz;
int rc;
status = acpi_get_table_with_size("NFIT", 0, &tbl, &sz);
if (ACPI_FAILURE(status)) {
/* This is ok, we could have an nvdimm hotplugged later */
dev_dbg(dev, "failed to find NFIT at startup\n");
return 0;
}
acpi_desc = acpi_nfit_desc_init(adev);
if (IS_ERR(acpi_desc)) {
dev_err(dev, "%s: error initializing acpi_desc: %ld\n",
__func__, PTR_ERR(acpi_desc));
return PTR_ERR(acpi_desc);
}
acpi_desc->nfit = (struct acpi_table_nfit *) tbl;
/* Evaluate _FIT and override with that if present */
status = acpi_evaluate_object(adev->handle, "_FIT", NULL, &buf);
if (ACPI_SUCCESS(status) && buf.length > 0) {
acpi_desc->nfit = (struct acpi_table_nfit *)buf.pointer;
sz = buf.length;
}
rc = acpi_nfit_init(acpi_desc, sz);
if (rc) {
nvdimm_bus_unregister(acpi_desc->nvdimm_bus);
return rc;
}
return 0;
}
static int acpi_nfit_remove(struct acpi_device *adev)
{
struct acpi_nfit_desc *acpi_desc = dev_get_drvdata(&adev->dev);
nvdimm_bus_unregister(acpi_desc->nvdimm_bus);
return 0;
}
static void acpi_nfit_notify(struct acpi_device *adev, u32 event)
{
struct acpi_nfit_desc *acpi_desc = dev_get_drvdata(&adev->dev);
struct acpi_buffer buf = { ACPI_ALLOCATE_BUFFER, NULL };
struct acpi_table_nfit *nfit_saved;
struct device *dev = &adev->dev;
acpi_status status;
int ret;
dev_dbg(dev, "%s: event: %d\n", __func__, event);
device_lock(dev);
if (!dev->driver) {
/* dev->driver may be null if we're being removed */
dev_dbg(dev, "%s: no driver found for dev\n", __func__);
return;
}
if (!acpi_desc) {
acpi_desc = acpi_nfit_desc_init(adev);
if (IS_ERR(acpi_desc)) {
dev_err(dev, "%s: error initializing acpi_desc: %ld\n",
__func__, PTR_ERR(acpi_desc));
goto out_unlock;
}
}
/* Evaluate _FIT */
status = acpi_evaluate_object(adev->handle, "_FIT", NULL, &buf);
if (ACPI_FAILURE(status)) {
dev_err(dev, "failed to evaluate _FIT\n");
goto out_unlock;
}
nfit_saved = acpi_desc->nfit;
acpi_desc->nfit = (struct acpi_table_nfit *)buf.pointer;
ret = acpi_nfit_init(acpi_desc, buf.length);
if (!ret) {
/* Merge failed, restore old nfit, and exit */
acpi_desc->nfit = nfit_saved;
dev_err(dev, "failed to merge updated NFIT\n");
}
kfree(buf.pointer);
out_unlock:
device_unlock(dev);
}
static const struct acpi_device_id acpi_nfit_ids[] = {
{ "ACPI0012", 0 },
{ "", 0 },
};
MODULE_DEVICE_TABLE(acpi, acpi_nfit_ids);
static struct acpi_driver acpi_nfit_driver = {
.name = KBUILD_MODNAME,
.ids = acpi_nfit_ids,
.ops = {
.add = acpi_nfit_add,
.remove = acpi_nfit_remove,
.notify = acpi_nfit_notify,
},
};
static __init int nfit_init(void)
{
BUILD_BUG_ON(sizeof(struct acpi_table_nfit) != 40);
BUILD_BUG_ON(sizeof(struct acpi_nfit_system_address) != 56);
BUILD_BUG_ON(sizeof(struct acpi_nfit_memory_map) != 48);
BUILD_BUG_ON(sizeof(struct acpi_nfit_interleave) != 20);
BUILD_BUG_ON(sizeof(struct acpi_nfit_smbios) != 9);
BUILD_BUG_ON(sizeof(struct acpi_nfit_control_region) != 80);
BUILD_BUG_ON(sizeof(struct acpi_nfit_data_region) != 40);
acpi_str_to_uuid(UUID_VOLATILE_MEMORY, nfit_uuid[NFIT_SPA_VOLATILE]);
acpi_str_to_uuid(UUID_PERSISTENT_MEMORY, nfit_uuid[NFIT_SPA_PM]);
acpi_str_to_uuid(UUID_CONTROL_REGION, nfit_uuid[NFIT_SPA_DCR]);
acpi_str_to_uuid(UUID_DATA_REGION, nfit_uuid[NFIT_SPA_BDW]);
acpi_str_to_uuid(UUID_VOLATILE_VIRTUAL_DISK, nfit_uuid[NFIT_SPA_VDISK]);
acpi_str_to_uuid(UUID_VOLATILE_VIRTUAL_CD, nfit_uuid[NFIT_SPA_VCD]);
acpi_str_to_uuid(UUID_PERSISTENT_VIRTUAL_DISK, nfit_uuid[NFIT_SPA_PDISK]);
acpi_str_to_uuid(UUID_PERSISTENT_VIRTUAL_CD, nfit_uuid[NFIT_SPA_PCD]);
acpi_str_to_uuid(UUID_NFIT_BUS, nfit_uuid[NFIT_DEV_BUS]);
acpi_str_to_uuid(UUID_NFIT_DIMM, nfit_uuid[NFIT_DEV_DIMM]);
return acpi_bus_register_driver(&acpi_nfit_driver);
}
static __exit void nfit_exit(void)
{
acpi_bus_unregister_driver(&acpi_nfit_driver);
}
module_init(nfit_init);
module_exit(nfit_exit);
MODULE_LICENSE("GPL v2");
MODULE_AUTHOR("Intel Corporation");