1
0
Fork 0
alistair23-linux/drivers/nvdimm/bus.c

870 lines
20 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.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
2015-06-08 12:27:06 -06:00
#include <linux/vmalloc.h>
#include <linux/uaccess.h>
#include <linux/module.h>
#include <linux/blkdev.h>
#include <linux/fcntl.h>
#include <linux/async.h>
#include <linux/genhd.h>
2015-06-08 12:27:06 -06:00
#include <linux/ndctl.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/io.h>
2015-06-08 12:27:06 -06:00
#include <linux/mm.h>
#include <linux/nd.h>
#include "nd-core.h"
#include "nd.h"
2015-06-08 12:27:06 -06:00
int nvdimm_major;
static int nvdimm_bus_major;
static struct class *nd_class;
static int to_nd_device_type(struct device *dev)
{
if (is_nvdimm(dev))
return ND_DEVICE_DIMM;
else if (is_nd_pmem(dev))
return ND_DEVICE_REGION_PMEM;
else if (is_nd_blk(dev))
return ND_DEVICE_REGION_BLK;
else if (is_nd_dax(dev))
return ND_DEVICE_DAX_PMEM;
else if (is_nd_pmem(dev->parent) || is_nd_blk(dev->parent))
return nd_region_to_nstype(to_nd_region(dev->parent));
return 0;
}
static int nvdimm_bus_uevent(struct device *dev, struct kobj_uevent_env *env)
{
/*
* Ensure that region devices always have their numa node set as
* early as possible.
*/
if (is_nd_pmem(dev) || is_nd_blk(dev))
set_dev_node(dev, to_nd_region(dev)->numa_node);
return add_uevent_var(env, "MODALIAS=" ND_DEVICE_MODALIAS_FMT,
to_nd_device_type(dev));
}
static int nvdimm_bus_match(struct device *dev, struct device_driver *drv)
{
struct nd_device_driver *nd_drv = to_nd_device_driver(drv);
return !!test_bit(to_nd_device_type(dev), &nd_drv->type);
}
static struct module *to_bus_provider(struct device *dev)
{
/* pin bus providers while regions are enabled */
if (is_nd_pmem(dev) || is_nd_blk(dev)) {
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
return nvdimm_bus->module;
}
return NULL;
}
2015-05-01 11:11:27 -06:00
static void nvdimm_bus_probe_start(struct nvdimm_bus *nvdimm_bus)
{
nvdimm_bus_lock(&nvdimm_bus->dev);
nvdimm_bus->probe_active++;
nvdimm_bus_unlock(&nvdimm_bus->dev);
}
static void nvdimm_bus_probe_end(struct nvdimm_bus *nvdimm_bus)
{
nvdimm_bus_lock(&nvdimm_bus->dev);
if (--nvdimm_bus->probe_active == 0)
wake_up(&nvdimm_bus->probe_wait);
nvdimm_bus_unlock(&nvdimm_bus->dev);
}
static int nvdimm_bus_probe(struct device *dev)
{
struct nd_device_driver *nd_drv = to_nd_device_driver(dev->driver);
struct module *provider = to_bus_provider(dev);
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
int rc;
if (!try_module_get(provider))
return -ENXIO;
2015-05-01 11:11:27 -06:00
nvdimm_bus_probe_start(nvdimm_bus);
rc = nd_drv->probe(dev);
2015-05-01 11:11:27 -06:00
if (rc == 0)
nd_region_probe_success(nvdimm_bus, dev);
else
nd_region_disable(nvdimm_bus, dev);
2015-05-01 11:11:27 -06:00
nvdimm_bus_probe_end(nvdimm_bus);
dev_dbg(&nvdimm_bus->dev, "%s.probe(%s) = %d\n", dev->driver->name,
dev_name(dev), rc);
if (rc != 0)
module_put(provider);
return rc;
}
static int nvdimm_bus_remove(struct device *dev)
{
struct nd_device_driver *nd_drv = to_nd_device_driver(dev->driver);
struct module *provider = to_bus_provider(dev);
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
int rc = 0;
if (nd_drv->remove)
rc = nd_drv->remove(dev);
2015-05-01 11:11:27 -06:00
nd_region_disable(nvdimm_bus, dev);
dev_dbg(&nvdimm_bus->dev, "%s.remove(%s) = %d\n", dev->driver->name,
dev_name(dev), rc);
module_put(provider);
return rc;
}
void nd_device_notify(struct device *dev, enum nvdimm_event event)
{
device_lock(dev);
if (dev->driver) {
struct nd_device_driver *nd_drv;
nd_drv = to_nd_device_driver(dev->driver);
if (nd_drv->notify)
nd_drv->notify(dev, event);
}
device_unlock(dev);
}
EXPORT_SYMBOL(nd_device_notify);
void nvdimm_region_notify(struct nd_region *nd_region, enum nvdimm_event event)
{
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(&nd_region->dev);
if (!nvdimm_bus)
return;
/* caller is responsible for holding a reference on the device */
nd_device_notify(&nd_region->dev, event);
}
EXPORT_SYMBOL_GPL(nvdimm_region_notify);
long nvdimm_clear_poison(struct device *dev, phys_addr_t phys,
unsigned int len)
{
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
struct nvdimm_bus_descriptor *nd_desc;
struct nd_cmd_clear_error clear_err;
struct nd_cmd_ars_cap ars_cap;
u32 clear_err_unit, mask;
int cmd_rc, rc;
if (!nvdimm_bus)
return -ENXIO;
nd_desc = nvdimm_bus->nd_desc;
if (!nd_desc->ndctl)
return -ENXIO;
memset(&ars_cap, 0, sizeof(ars_cap));
ars_cap.address = phys;
ars_cap.length = len;
rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_CAP, &ars_cap,
sizeof(ars_cap), &cmd_rc);
if (rc < 0)
return rc;
if (cmd_rc < 0)
return cmd_rc;
clear_err_unit = ars_cap.clear_err_unit;
if (!clear_err_unit || !is_power_of_2(clear_err_unit))
return -ENXIO;
mask = clear_err_unit - 1;
if ((phys | len) & mask)
return -ENXIO;
memset(&clear_err, 0, sizeof(clear_err));
clear_err.address = phys;
clear_err.length = len;
rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_CLEAR_ERROR, &clear_err,
sizeof(clear_err), &cmd_rc);
if (rc < 0)
return rc;
if (cmd_rc < 0)
return cmd_rc;
return clear_err.cleared;
}
EXPORT_SYMBOL_GPL(nvdimm_clear_poison);
static struct bus_type nvdimm_bus_type = {
.name = "nd",
.uevent = nvdimm_bus_uevent,
.match = nvdimm_bus_match,
.probe = nvdimm_bus_probe,
.remove = nvdimm_bus_remove,
};
static ASYNC_DOMAIN_EXCLUSIVE(nd_async_domain);
void nd_synchronize(void)
{
async_synchronize_full_domain(&nd_async_domain);
}
EXPORT_SYMBOL_GPL(nd_synchronize);
static void nd_async_device_register(void *d, async_cookie_t cookie)
{
struct device *dev = d;
if (device_add(dev) != 0) {
dev_err(dev, "%s: failed\n", __func__);
put_device(dev);
}
put_device(dev);
}
static void nd_async_device_unregister(void *d, async_cookie_t cookie)
{
struct device *dev = d;
/* flush bus operations before delete */
nvdimm_bus_lock(dev);
nvdimm_bus_unlock(dev);
device_unregister(dev);
put_device(dev);
}
void __nd_device_register(struct device *dev)
{
if (!dev)
return;
dev->bus = &nvdimm_bus_type;
get_device(dev);
async_schedule_domain(nd_async_device_register, dev,
&nd_async_domain);
}
void nd_device_register(struct device *dev)
{
device_initialize(dev);
__nd_device_register(dev);
}
EXPORT_SYMBOL(nd_device_register);
void nd_device_unregister(struct device *dev, enum nd_async_mode mode)
{
switch (mode) {
case ND_ASYNC:
get_device(dev);
async_schedule_domain(nd_async_device_unregister, dev,
&nd_async_domain);
break;
case ND_SYNC:
nd_synchronize();
device_unregister(dev);
break;
}
}
EXPORT_SYMBOL(nd_device_unregister);
/**
* __nd_driver_register() - register a region or a namespace driver
* @nd_drv: driver to register
* @owner: automatically set by nd_driver_register() macro
* @mod_name: automatically set by nd_driver_register() macro
*/
int __nd_driver_register(struct nd_device_driver *nd_drv, struct module *owner,
const char *mod_name)
{
struct device_driver *drv = &nd_drv->drv;
if (!nd_drv->type) {
pr_debug("driver type bitmask not set (%pf)\n",
__builtin_return_address(0));
return -EINVAL;
}
if (!nd_drv->probe) {
pr_debug("%s ->probe() must be specified\n", mod_name);
return -EINVAL;
}
drv->bus = &nvdimm_bus_type;
drv->owner = owner;
drv->mod_name = mod_name;
return driver_register(drv);
}
EXPORT_SYMBOL(__nd_driver_register);
int nvdimm_revalidate_disk(struct gendisk *disk)
{
struct device *dev = disk->driverfs_dev;
struct nd_region *nd_region = to_nd_region(dev->parent);
const char *pol = nd_region->ro ? "only" : "write";
if (nd_region->ro == get_disk_ro(disk))
return 0;
dev_info(dev, "%s read-%s, marking %s read-%s\n",
dev_name(&nd_region->dev), pol, disk->disk_name, pol);
set_disk_ro(disk, nd_region->ro);
return 0;
}
EXPORT_SYMBOL(nvdimm_revalidate_disk);
static ssize_t modalias_show(struct device *dev, struct device_attribute *attr,
char *buf)
{
return sprintf(buf, ND_DEVICE_MODALIAS_FMT "\n",
to_nd_device_type(dev));
}
static DEVICE_ATTR_RO(modalias);
static ssize_t devtype_show(struct device *dev, struct device_attribute *attr,
char *buf)
{
return sprintf(buf, "%s\n", dev->type->name);
}
static DEVICE_ATTR_RO(devtype);
static struct attribute *nd_device_attributes[] = {
&dev_attr_modalias.attr,
&dev_attr_devtype.attr,
NULL,
};
/**
* nd_device_attribute_group - generic attributes for all devices on an nd bus
*/
struct attribute_group nd_device_attribute_group = {
.attrs = nd_device_attributes,
};
EXPORT_SYMBOL_GPL(nd_device_attribute_group);
static ssize_t numa_node_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
return sprintf(buf, "%d\n", dev_to_node(dev));
}
static DEVICE_ATTR_RO(numa_node);
static struct attribute *nd_numa_attributes[] = {
&dev_attr_numa_node.attr,
NULL,
};
static umode_t nd_numa_attr_visible(struct kobject *kobj, struct attribute *a,
int n)
{
if (!IS_ENABLED(CONFIG_NUMA))
return 0;
return a->mode;
}
/**
* nd_numa_attribute_group - NUMA attributes for all devices on an nd bus
*/
struct attribute_group nd_numa_attribute_group = {
.attrs = nd_numa_attributes,
.is_visible = nd_numa_attr_visible,
};
EXPORT_SYMBOL_GPL(nd_numa_attribute_group);
int nvdimm_bus_create_ndctl(struct nvdimm_bus *nvdimm_bus)
{
dev_t devt = MKDEV(nvdimm_bus_major, nvdimm_bus->id);
struct device *dev;
dev = device_create(nd_class, &nvdimm_bus->dev, devt, nvdimm_bus,
"ndctl%d", nvdimm_bus->id);
if (IS_ERR(dev))
dev_dbg(&nvdimm_bus->dev, "failed to register ndctl%d: %ld\n",
nvdimm_bus->id, PTR_ERR(dev));
return PTR_ERR_OR_ZERO(dev);
}
void nvdimm_bus_destroy_ndctl(struct nvdimm_bus *nvdimm_bus)
{
device_destroy(nd_class, MKDEV(nvdimm_bus_major, nvdimm_bus->id));
}
2015-06-08 12:27:06 -06:00
static const struct nd_cmd_desc __nd_cmd_dimm_descs[] = {
[ND_CMD_IMPLEMENTED] = { },
[ND_CMD_SMART] = {
.out_num = 2,
.out_sizes = { 4, 128, },
2015-06-08 12:27:06 -06:00
},
[ND_CMD_SMART_THRESHOLD] = {
.out_num = 2,
.out_sizes = { 4, 8, },
},
[ND_CMD_DIMM_FLAGS] = {
.out_num = 2,
.out_sizes = { 4, 4 },
},
[ND_CMD_GET_CONFIG_SIZE] = {
.out_num = 3,
.out_sizes = { 4, 4, 4, },
},
[ND_CMD_GET_CONFIG_DATA] = {
.in_num = 2,
.in_sizes = { 4, 4, },
.out_num = 2,
.out_sizes = { 4, UINT_MAX, },
},
[ND_CMD_SET_CONFIG_DATA] = {
.in_num = 3,
.in_sizes = { 4, 4, UINT_MAX, },
.out_num = 1,
.out_sizes = { 4, },
},
[ND_CMD_VENDOR] = {
.in_num = 3,
.in_sizes = { 4, 4, UINT_MAX, },
.out_num = 3,
.out_sizes = { 4, 4, UINT_MAX, },
},
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
[ND_CMD_CALL] = {
.in_num = 2,
.in_sizes = { sizeof(struct nd_cmd_pkg), UINT_MAX, },
.out_num = 1,
.out_sizes = { UINT_MAX, },
},
2015-06-08 12:27:06 -06:00
};
const struct nd_cmd_desc *nd_cmd_dimm_desc(int cmd)
{
if (cmd < ARRAY_SIZE(__nd_cmd_dimm_descs))
return &__nd_cmd_dimm_descs[cmd];
return NULL;
}
EXPORT_SYMBOL_GPL(nd_cmd_dimm_desc);
static const struct nd_cmd_desc __nd_cmd_bus_descs[] = {
[ND_CMD_IMPLEMENTED] = { },
[ND_CMD_ARS_CAP] = {
.in_num = 2,
.in_sizes = { 8, 8, },
.out_num = 4,
.out_sizes = { 4, 4, 4, 4, },
2015-06-08 12:27:06 -06:00
},
[ND_CMD_ARS_START] = {
.in_num = 5,
.in_sizes = { 8, 8, 2, 1, 5, },
.out_num = 2,
.out_sizes = { 4, 4, },
2015-06-08 12:27:06 -06:00
},
[ND_CMD_ARS_STATUS] = {
.out_num = 3,
.out_sizes = { 4, 4, UINT_MAX, },
2015-06-08 12:27:06 -06:00
},
[ND_CMD_CLEAR_ERROR] = {
.in_num = 2,
.in_sizes = { 8, 8, },
.out_num = 3,
.out_sizes = { 4, 4, 8, },
},
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
[ND_CMD_CALL] = {
.in_num = 2,
.in_sizes = { sizeof(struct nd_cmd_pkg), UINT_MAX, },
.out_num = 1,
.out_sizes = { UINT_MAX, },
},
2015-06-08 12:27:06 -06:00
};
const struct nd_cmd_desc *nd_cmd_bus_desc(int cmd)
{
if (cmd < ARRAY_SIZE(__nd_cmd_bus_descs))
return &__nd_cmd_bus_descs[cmd];
return NULL;
}
EXPORT_SYMBOL_GPL(nd_cmd_bus_desc);
u32 nd_cmd_in_size(struct nvdimm *nvdimm, int cmd,
const struct nd_cmd_desc *desc, int idx, void *buf)
{
if (idx >= desc->in_num)
return UINT_MAX;
if (desc->in_sizes[idx] < UINT_MAX)
return desc->in_sizes[idx];
if (nvdimm && cmd == ND_CMD_SET_CONFIG_DATA && idx == 2) {
struct nd_cmd_set_config_hdr *hdr = buf;
return hdr->in_length;
} else if (nvdimm && cmd == ND_CMD_VENDOR && idx == 2) {
struct nd_cmd_vendor_hdr *hdr = buf;
return hdr->in_length;
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
} else if (cmd == ND_CMD_CALL) {
struct nd_cmd_pkg *pkg = buf;
return pkg->nd_size_in;
2015-06-08 12:27:06 -06:00
}
return UINT_MAX;
}
EXPORT_SYMBOL_GPL(nd_cmd_in_size);
u32 nd_cmd_out_size(struct nvdimm *nvdimm, int cmd,
const struct nd_cmd_desc *desc, int idx, const u32 *in_field,
const u32 *out_field)
{
if (idx >= desc->out_num)
return UINT_MAX;
if (desc->out_sizes[idx] < UINT_MAX)
return desc->out_sizes[idx];
if (nvdimm && cmd == ND_CMD_GET_CONFIG_DATA && idx == 1)
return in_field[1];
else if (nvdimm && cmd == ND_CMD_VENDOR && idx == 2)
return out_field[1];
else if (!nvdimm && cmd == ND_CMD_ARS_STATUS && idx == 2)
return out_field[1] - 8;
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
else if (cmd == ND_CMD_CALL) {
struct nd_cmd_pkg *pkg = (struct nd_cmd_pkg *) in_field;
return pkg->nd_size_out;
}
2015-06-08 12:27:06 -06:00
return UINT_MAX;
}
EXPORT_SYMBOL_GPL(nd_cmd_out_size);
void wait_nvdimm_bus_probe_idle(struct device *dev)
2015-05-01 11:11:27 -06:00
{
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
2015-05-01 11:11:27 -06:00
do {
if (nvdimm_bus->probe_active == 0)
break;
nvdimm_bus_unlock(&nvdimm_bus->dev);
wait_event(nvdimm_bus->probe_wait,
nvdimm_bus->probe_active == 0);
nvdimm_bus_lock(&nvdimm_bus->dev);
} while (true);
}
static int pmem_active(struct device *dev, void *data)
{
if (is_nd_pmem(dev) && dev->driver)
return -EBUSY;
return 0;
}
2015-05-01 11:11:27 -06:00
/* set_config requires an idle interleave set */
static int nd_cmd_clear_to_send(struct nvdimm_bus *nvdimm_bus,
struct nvdimm *nvdimm, unsigned int cmd)
2015-05-01 11:11:27 -06:00
{
struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
/* ask the bus provider if it would like to block this request */
if (nd_desc->clear_to_send) {
int rc = nd_desc->clear_to_send(nd_desc, nvdimm, cmd);
if (rc)
return rc;
}
2015-05-01 11:11:27 -06:00
/* require clear error to go through the pmem driver */
if (!nvdimm && cmd == ND_CMD_CLEAR_ERROR)
return device_for_each_child(&nvdimm_bus->dev, NULL,
pmem_active);
2015-05-01 11:11:27 -06:00
if (!nvdimm || cmd != ND_CMD_SET_CONFIG_DATA)
return 0;
/* prevent label manipulation while the kernel owns label updates */
wait_nvdimm_bus_probe_idle(&nvdimm_bus->dev);
2015-05-01 11:11:27 -06:00
if (atomic_read(&nvdimm->busy))
return -EBUSY;
return 0;
}
2015-06-08 12:27:06 -06:00
static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
int read_only, unsigned int ioctl_cmd, unsigned long arg)
{
struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
size_t buf_len = 0, in_len = 0, out_len = 0;
static char out_env[ND_CMD_MAX_ENVELOPE];
static char in_env[ND_CMD_MAX_ENVELOPE];
const struct nd_cmd_desc *desc = NULL;
unsigned int cmd = _IOC_NR(ioctl_cmd);
void __user *p = (void __user *) arg;
struct device *dev = &nvdimm_bus->dev;
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
struct nd_cmd_pkg pkg;
2015-06-08 12:27:06 -06:00
const char *cmd_name, *dimm_name;
unsigned long cmd_mask;
2015-06-08 12:27:06 -06:00
void *buf;
int rc, i;
if (nvdimm) {
desc = nd_cmd_dimm_desc(cmd);
cmd_name = nvdimm_cmd_name(cmd);
cmd_mask = nvdimm->cmd_mask;
2015-06-08 12:27:06 -06:00
dimm_name = dev_name(&nvdimm->dev);
} else {
desc = nd_cmd_bus_desc(cmd);
cmd_name = nvdimm_bus_cmd_name(cmd);
cmd_mask = nd_desc->cmd_mask;
2015-06-08 12:27:06 -06:00
dimm_name = "bus";
}
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
if (cmd == ND_CMD_CALL) {
if (copy_from_user(&pkg, p, sizeof(pkg)))
return -EFAULT;
}
2015-06-08 12:27:06 -06:00
if (!desc || (desc->out_num + desc->in_num == 0) ||
!test_bit(cmd, &cmd_mask))
2015-06-08 12:27:06 -06:00
return -ENOTTY;
/* fail write commands (when read-only) */
if (read_only)
switch (cmd) {
case ND_CMD_VENDOR:
case ND_CMD_SET_CONFIG_DATA:
case ND_CMD_ARS_START:
case ND_CMD_CLEAR_ERROR:
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
case ND_CMD_CALL:
2015-06-08 12:27:06 -06:00
dev_dbg(&nvdimm_bus->dev, "'%s' command while read-only.\n",
nvdimm ? nvdimm_cmd_name(cmd)
: nvdimm_bus_cmd_name(cmd));
return -EPERM;
default:
break;
}
/* process an input envelope */
for (i = 0; i < desc->in_num; i++) {
u32 in_size, copy;
in_size = nd_cmd_in_size(nvdimm, cmd, desc, i, in_env);
if (in_size == UINT_MAX) {
dev_err(dev, "%s:%s unknown input size cmd: %s field: %d\n",
__func__, dimm_name, cmd_name, i);
return -ENXIO;
}
if (in_len < sizeof(in_env))
copy = min_t(u32, sizeof(in_env) - in_len, in_size);
else
copy = 0;
if (copy && copy_from_user(&in_env[in_len], p + in_len, copy))
return -EFAULT;
in_len += in_size;
}
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-28 17:23:43 -06:00
if (cmd == ND_CMD_CALL) {
dev_dbg(dev, "%s:%s, idx: %llu, in: %zu, out: %zu, len %zu\n",
__func__, dimm_name, pkg.nd_command,
in_len, out_len, buf_len);
for (i = 0; i < ARRAY_SIZE(pkg.nd_reserved2); i++)
if (pkg.nd_reserved2[i])
return -EINVAL;
}
2015-06-08 12:27:06 -06:00
/* process an output envelope */
for (i = 0; i < desc->out_num; i++) {
u32 out_size = nd_cmd_out_size(nvdimm, cmd, desc, i,
(u32 *) in_env, (u32 *) out_env);
u32 copy;
if (out_size == UINT_MAX) {
dev_dbg(dev, "%s:%s unknown output size cmd: %s field: %d\n",
__func__, dimm_name, cmd_name, i);
return -EFAULT;
}
if (out_len < sizeof(out_env))
copy = min_t(u32, sizeof(out_env) - out_len, out_size);
else
copy = 0;
if (copy && copy_from_user(&out_env[out_len],
p + in_len + out_len, copy))
return -EFAULT;
out_len += out_size;
}
buf_len = out_len + in_len;
if (buf_len > ND_IOCTL_MAX_BUFLEN) {
dev_dbg(dev, "%s:%s cmd: %s buf_len: %zu > %d\n", __func__,
dimm_name, cmd_name, buf_len,
ND_IOCTL_MAX_BUFLEN);
return -EINVAL;
}
buf = vmalloc(buf_len);
if (!buf)
return -ENOMEM;
if (copy_from_user(buf, p, buf_len)) {
rc = -EFAULT;
goto out;
}
2015-05-01 11:11:27 -06:00
nvdimm_bus_lock(&nvdimm_bus->dev);
rc = nd_cmd_clear_to_send(nvdimm_bus, nvdimm, cmd);
2015-05-01 11:11:27 -06:00
if (rc)
goto out_unlock;
rc = nd_desc->ndctl(nd_desc, nvdimm, cmd, buf, buf_len, NULL);
2015-06-08 12:27:06 -06:00
if (rc < 0)
2015-05-01 11:11:27 -06:00
goto out_unlock;
2015-06-08 12:27:06 -06:00
if (copy_to_user(p, buf, buf_len))
rc = -EFAULT;
2015-05-01 11:11:27 -06:00
out_unlock:
nvdimm_bus_unlock(&nvdimm_bus->dev);
2015-06-08 12:27:06 -06:00
out:
vfree(buf);
return rc;
}
static long nd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
2015-06-08 12:27:06 -06:00
long id = (long) file->private_data;
int rc = -ENXIO, ro;
2015-06-08 12:27:06 -06:00
struct nvdimm_bus *nvdimm_bus;
ro = ((file->f_flags & O_ACCMODE) == O_RDONLY);
2015-06-08 12:27:06 -06:00
mutex_lock(&nvdimm_bus_list_mutex);
list_for_each_entry(nvdimm_bus, &nvdimm_bus_list, list) {
if (nvdimm_bus->id == id) {
rc = __nd_ioctl(nvdimm_bus, NULL, ro, cmd, arg);
2015-06-08 12:27:06 -06:00
break;
}
}
mutex_unlock(&nvdimm_bus_list_mutex);
return rc;
}
static int match_dimm(struct device *dev, void *data)
{
long id = (long) data;
if (is_nvdimm(dev)) {
struct nvdimm *nvdimm = to_nvdimm(dev);
return nvdimm->id == id;
}
return 0;
}
static long nvdimm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
int rc = -ENXIO, ro;
2015-06-08 12:27:06 -06:00
struct nvdimm_bus *nvdimm_bus;
ro = ((file->f_flags & O_ACCMODE) == O_RDONLY);
2015-06-08 12:27:06 -06:00
mutex_lock(&nvdimm_bus_list_mutex);
list_for_each_entry(nvdimm_bus, &nvdimm_bus_list, list) {
struct device *dev = device_find_child(&nvdimm_bus->dev,
file->private_data, match_dimm);
struct nvdimm *nvdimm;
if (!dev)
continue;
nvdimm = to_nvdimm(dev);
rc = __nd_ioctl(nvdimm_bus, nvdimm, ro, cmd, arg);
2015-06-08 12:27:06 -06:00
put_device(dev);
break;
}
mutex_unlock(&nvdimm_bus_list_mutex);
return rc;
}
static int nd_open(struct inode *inode, struct file *file)
{
long minor = iminor(inode);
file->private_data = (void *) minor;
return 0;
}
static const struct file_operations nvdimm_bus_fops = {
.owner = THIS_MODULE,
2015-06-08 12:27:06 -06:00
.open = nd_open,
.unlocked_ioctl = nd_ioctl,
.compat_ioctl = nd_ioctl,
.llseek = noop_llseek,
};
2015-06-08 12:27:06 -06:00
static const struct file_operations nvdimm_fops = {
.owner = THIS_MODULE,
.open = nd_open,
.unlocked_ioctl = nvdimm_ioctl,
.compat_ioctl = nvdimm_ioctl,
.llseek = noop_llseek,
};
int __init nvdimm_bus_init(void)
{
int rc;
BUILD_BUG_ON(sizeof(struct nd_smart_payload) != 128);
BUILD_BUG_ON(sizeof(struct nd_smart_threshold_payload) != 8);
rc = bus_register(&nvdimm_bus_type);
if (rc)
return rc;
rc = register_chrdev(0, "ndctl", &nvdimm_bus_fops);
if (rc < 0)
2015-06-08 12:27:06 -06:00
goto err_bus_chrdev;
nvdimm_bus_major = rc;
2015-06-08 12:27:06 -06:00
rc = register_chrdev(0, "dimmctl", &nvdimm_fops);
if (rc < 0)
goto err_dimm_chrdev;
nvdimm_major = rc;
nd_class = class_create(THIS_MODULE, "nd");
if (IS_ERR(nd_class)) {
rc = PTR_ERR(nd_class);
goto err_class;
}
return 0;
err_class:
2015-06-08 12:27:06 -06:00
unregister_chrdev(nvdimm_major, "dimmctl");
err_dimm_chrdev:
unregister_chrdev(nvdimm_bus_major, "ndctl");
2015-06-08 12:27:06 -06:00
err_bus_chrdev:
bus_unregister(&nvdimm_bus_type);
return rc;
}
void nvdimm_bus_exit(void)
{
class_destroy(nd_class);
unregister_chrdev(nvdimm_bus_major, "ndctl");
2015-06-08 12:27:06 -06:00
unregister_chrdev(nvdimm_major, "dimmctl");
bus_unregister(&nvdimm_bus_type);
}