linux/drivers/nvdimm/dimm_devs.c

1014 lines
25 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/moduleparam.h>
#include <linux/vmalloc.h>
#include <linux/device.h>
2015-06-08 18:27:06 +00:00
#include <linux/ndctl.h>
#include <linux/slab.h>
#include <linux/io.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include "nd-core.h"
#include "label.h"
#include "pmem.h"
#include "nd.h"
static DEFINE_IDA(dimm_ida);
static bool noblk;
module_param(noblk, bool, 0444);
MODULE_PARM_DESC(noblk, "force disable BLK / local alias support");
/*
* Retrieve bus and dimm handle and return if this bus supports
* get_config_data commands
*/
int nvdimm_check_config_data(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
if (!nvdimm->cmd_mask ||
!test_bit(ND_CMD_GET_CONFIG_DATA, &nvdimm->cmd_mask)) {
if (test_bit(NDD_LABELING, &nvdimm->flags))
return -ENXIO;
else
return -ENOTTY;
}
return 0;
}
static int validate_dimm(struct nvdimm_drvdata *ndd)
{
int rc;
if (!ndd)
return -EINVAL;
rc = nvdimm_check_config_data(ndd->dev);
if (rc)
2019-03-25 19:32:28 +00:00
dev_dbg(ndd->dev, "%ps: %s error: %d\n",
__builtin_return_address(0), __func__, rc);
return rc;
}
/**
* nvdimm_init_nsarea - determine the geometry of a dimm's namespace area
* @nvdimm: dimm to initialize
*/
int nvdimm_init_nsarea(struct nvdimm_drvdata *ndd)
{
struct nd_cmd_get_config_size *cmd = &ndd->nsarea;
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
struct nvdimm_bus_descriptor *nd_desc;
int rc = validate_dimm(ndd);
int cmd_rc = 0;
if (rc)
return rc;
if (cmd->config_size)
return 0; /* already valid */
memset(cmd, 0, sizeof(*cmd));
nd_desc = nvdimm_bus->nd_desc;
rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev),
ND_CMD_GET_CONFIG_SIZE, cmd, sizeof(*cmd), &cmd_rc);
if (rc < 0)
return rc;
return cmd_rc;
}
int nvdimm_get_config_data(struct nvdimm_drvdata *ndd, void *buf,
size_t offset, size_t len)
{
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
int rc = validate_dimm(ndd), cmd_rc = 0;
struct nd_cmd_get_config_data_hdr *cmd;
size_t max_cmd_size, buf_offset;
if (rc)
return rc;
if (offset + len > ndd->nsarea.config_size)
return -ENXIO;
max_cmd_size = min_t(u32, len, ndd->nsarea.max_xfer);
cmd = kvzalloc(max_cmd_size + sizeof(*cmd), GFP_KERNEL);
if (!cmd)
return -ENOMEM;
for (buf_offset = 0; len;
len -= cmd->in_length, buf_offset += cmd->in_length) {
size_t cmd_size;
cmd->in_offset = offset + buf_offset;
cmd->in_length = min(max_cmd_size, len);
cmd_size = sizeof(*cmd) + cmd->in_length;
rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev),
ND_CMD_GET_CONFIG_DATA, cmd, cmd_size, &cmd_rc);
if (rc < 0)
break;
if (cmd_rc < 0) {
rc = cmd_rc;
break;
}
/* out_buf should be valid, copy it into our output buffer */
memcpy(buf + buf_offset, cmd->out_buf, cmd->in_length);
}
kvfree(cmd);
return rc;
}
int nvdimm_set_config_data(struct nvdimm_drvdata *ndd, size_t offset,
void *buf, size_t len)
{
size_t max_cmd_size, buf_offset;
struct nd_cmd_set_config_hdr *cmd;
int rc = validate_dimm(ndd), cmd_rc = 0;
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
if (rc)
return rc;
if (offset + len > ndd->nsarea.config_size)
return -ENXIO;
max_cmd_size = min_t(u32, len, ndd->nsarea.max_xfer);
cmd = kvzalloc(max_cmd_size + sizeof(*cmd) + sizeof(u32), GFP_KERNEL);
if (!cmd)
return -ENOMEM;
for (buf_offset = 0; len; len -= cmd->in_length,
buf_offset += cmd->in_length) {
size_t cmd_size;
cmd->in_offset = offset + buf_offset;
cmd->in_length = min(max_cmd_size, len);
memcpy(cmd->in_buf, buf + buf_offset, cmd->in_length);
/* status is output in the last 4-bytes of the command buffer */
cmd_size = sizeof(*cmd) + cmd->in_length + sizeof(u32);
rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev),
ND_CMD_SET_CONFIG_DATA, cmd, cmd_size, &cmd_rc);
if (rc < 0)
break;
if (cmd_rc < 0) {
rc = cmd_rc;
break;
}
}
kvfree(cmd);
return rc;
}
void nvdimm_set_labeling(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
set_bit(NDD_LABELING, &nvdimm->flags);
}
void nvdimm_set_locked(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
set_bit(NDD_LOCKED, &nvdimm->flags);
}
void nvdimm_clear_locked(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
clear_bit(NDD_LOCKED, &nvdimm->flags);
}
static void nvdimm_release(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
ida_simple_remove(&dimm_ida, nvdimm->id);
kfree(nvdimm);
}
struct nvdimm *to_nvdimm(struct device *dev)
{
struct nvdimm *nvdimm = container_of(dev, struct nvdimm, dev);
WARN_ON(!is_nvdimm(dev));
return nvdimm;
}
EXPORT_SYMBOL_GPL(to_nvdimm);
struct nvdimm *nd_blk_region_to_dimm(struct nd_blk_region *ndbr)
{
struct nd_region *nd_region = &ndbr->nd_region;
struct nd_mapping *nd_mapping = &nd_region->mapping[0];
return nd_mapping->nvdimm;
}
EXPORT_SYMBOL_GPL(nd_blk_region_to_dimm);
unsigned long nd_blk_memremap_flags(struct nd_blk_region *ndbr)
{
/* pmem mapping properties are private to libnvdimm */
return ARCH_MEMREMAP_PMEM;
}
EXPORT_SYMBOL_GPL(nd_blk_memremap_flags);
struct nvdimm_drvdata *to_ndd(struct nd_mapping *nd_mapping)
{
struct nvdimm *nvdimm = nd_mapping->nvdimm;
WARN_ON_ONCE(!is_nvdimm_bus_locked(&nvdimm->dev));
return dev_get_drvdata(&nvdimm->dev);
}
EXPORT_SYMBOL(to_ndd);
void nvdimm_drvdata_release(struct kref *kref)
{
struct nvdimm_drvdata *ndd = container_of(kref, typeof(*ndd), kref);
struct device *dev = ndd->dev;
struct resource *res, *_r;
dev_dbg(dev, "trace\n");
nvdimm_bus_lock(dev);
for_each_dpa_resource_safe(ndd, res, _r)
nvdimm_free_dpa(ndd, res);
nvdimm_bus_unlock(dev);
kvfree(ndd->data);
kfree(ndd);
put_device(dev);
}
void get_ndd(struct nvdimm_drvdata *ndd)
{
kref_get(&ndd->kref);
}
void put_ndd(struct nvdimm_drvdata *ndd)
{
if (ndd)
kref_put(&ndd->kref, nvdimm_drvdata_release);
}
const char *nvdimm_name(struct nvdimm *nvdimm)
{
return dev_name(&nvdimm->dev);
}
EXPORT_SYMBOL_GPL(nvdimm_name);
struct kobject *nvdimm_kobj(struct nvdimm *nvdimm)
{
return &nvdimm->dev.kobj;
}
EXPORT_SYMBOL_GPL(nvdimm_kobj);
unsigned long nvdimm_cmd_mask(struct nvdimm *nvdimm)
{
return nvdimm->cmd_mask;
}
EXPORT_SYMBOL_GPL(nvdimm_cmd_mask);
void *nvdimm_provider_data(struct nvdimm *nvdimm)
{
2015-06-08 18:27:06 +00:00
if (nvdimm)
return nvdimm->provider_data;
return NULL;
}
EXPORT_SYMBOL_GPL(nvdimm_provider_data);
2015-06-08 18:27:06 +00:00
static ssize_t commands_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
int cmd, len = 0;
if (!nvdimm->cmd_mask)
2015-06-08 18:27:06 +00:00
return sprintf(buf, "\n");
for_each_set_bit(cmd, &nvdimm->cmd_mask, BITS_PER_LONG)
2015-06-08 18:27:06 +00:00
len += sprintf(buf + len, "%s ", nvdimm_cmd_name(cmd));
len += sprintf(buf + len, "\n");
return len;
}
static DEVICE_ATTR_RO(commands);
static ssize_t flags_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
return sprintf(buf, "%s%s%s\n",
test_bit(NDD_ALIASING, &nvdimm->flags) ? "alias " : "",
test_bit(NDD_LABELING, &nvdimm->flags) ? "label " : "",
test_bit(NDD_LOCKED, &nvdimm->flags) ? "lock " : "");
}
static DEVICE_ATTR_RO(flags);
2015-05-01 17:11:27 +00:00
static ssize_t state_show(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
/*
* The state may be in the process of changing, userspace should
* quiesce probing if it wants a static answer
*/
nvdimm_bus_lock(dev);
nvdimm_bus_unlock(dev);
return sprintf(buf, "%s\n", atomic_read(&nvdimm->busy)
? "active" : "idle");
}
static DEVICE_ATTR_RO(state);
static ssize_t available_slots_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nvdimm_drvdata *ndd = dev_get_drvdata(dev);
ssize_t rc;
u32 nfree;
if (!ndd)
return -ENXIO;
nvdimm_bus_lock(dev);
nfree = nd_label_nfree(ndd);
if (nfree - 1 > nfree) {
dev_WARN_ONCE(dev, 1, "we ate our last label?\n");
nfree = 0;
} else
nfree--;
rc = sprintf(buf, "%d\n", nfree);
nvdimm_bus_unlock(dev);
return rc;
}
static DEVICE_ATTR_RO(available_slots);
__weak ssize_t security_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
if (test_bit(NVDIMM_SECURITY_OVERWRITE, &nvdimm->sec.flags))
return sprintf(buf, "overwrite\n");
if (test_bit(NVDIMM_SECURITY_DISABLED, &nvdimm->sec.flags))
return sprintf(buf, "disabled\n");
if (test_bit(NVDIMM_SECURITY_UNLOCKED, &nvdimm->sec.flags))
return sprintf(buf, "unlocked\n");
if (test_bit(NVDIMM_SECURITY_LOCKED, &nvdimm->sec.flags))
return sprintf(buf, "locked\n");
return -ENOTTY;
}
static ssize_t frozen_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
return sprintf(buf, "%d\n", test_bit(NVDIMM_SECURITY_FROZEN,
&nvdimm->sec.flags));
}
static DEVICE_ATTR_RO(frozen);
static ssize_t security_store(struct device *dev,
struct device_attribute *attr, const char *buf, size_t len)
{
ssize_t rc;
/*
* Require all userspace triggered security management to be
* done while probing is idle and the DIMM is not in active use
* in any region.
*/
driver-core, libnvdimm: Let device subsystems add local lockdep coverage For good reason, the standard device_lock() is marked lockdep_set_novalidate_class() because there is simply no sane way to describe the myriad ways the device_lock() ordered with other locks. However, that leaves subsystems that know their own local device_lock() ordering rules to find lock ordering mistakes manually. Instead, introduce an optional / additional lockdep-enabled lock that a subsystem can acquire in all the same paths that the device_lock() is acquired. A conversion of the NFIT driver and NVDIMM subsystem to a lockdep-validate device_lock() scheme is included. The debug_nvdimm_lock() implementation implements the correct lock-class and stacking order for the libnvdimm device topology hierarchy. Yes, this is a hack, but hopefully it is a useful hack for other subsystems device_lock() debug sessions. Quoting Greg: "Yeah, it feels a bit hacky but it's really up to a subsystem to mess up using it as much as anything else, so user beware :) I don't object to it if it makes things easier for you to debug." Cc: Ingo Molnar <mingo@redhat.com> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Will Deacon <will.deacon@arm.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Vishal Verma <vishal.l.verma@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Link: https://lore.kernel.org/r/156341210661.292348.7014034644265455704.stgit@dwillia2-desk3.amr.corp.intel.com
2019-07-18 01:08:26 +00:00
nd_device_lock(dev);
nvdimm_bus_lock(dev);
wait_nvdimm_bus_probe_idle(dev);
rc = nvdimm_security_store(dev, buf, len);
nvdimm_bus_unlock(dev);
driver-core, libnvdimm: Let device subsystems add local lockdep coverage For good reason, the standard device_lock() is marked lockdep_set_novalidate_class() because there is simply no sane way to describe the myriad ways the device_lock() ordered with other locks. However, that leaves subsystems that know their own local device_lock() ordering rules to find lock ordering mistakes manually. Instead, introduce an optional / additional lockdep-enabled lock that a subsystem can acquire in all the same paths that the device_lock() is acquired. A conversion of the NFIT driver and NVDIMM subsystem to a lockdep-validate device_lock() scheme is included. The debug_nvdimm_lock() implementation implements the correct lock-class and stacking order for the libnvdimm device topology hierarchy. Yes, this is a hack, but hopefully it is a useful hack for other subsystems device_lock() debug sessions. Quoting Greg: "Yeah, it feels a bit hacky but it's really up to a subsystem to mess up using it as much as anything else, so user beware :) I don't object to it if it makes things easier for you to debug." Cc: Ingo Molnar <mingo@redhat.com> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Will Deacon <will.deacon@arm.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Vishal Verma <vishal.l.verma@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Link: https://lore.kernel.org/r/156341210661.292348.7014034644265455704.stgit@dwillia2-desk3.amr.corp.intel.com
2019-07-18 01:08:26 +00:00
nd_device_unlock(dev);
return rc;
}
static DEVICE_ATTR_RW(security);
2015-06-08 18:27:06 +00:00
static struct attribute *nvdimm_attributes[] = {
2015-05-01 17:11:27 +00:00
&dev_attr_state.attr,
&dev_attr_flags.attr,
2015-06-08 18:27:06 +00:00
&dev_attr_commands.attr,
&dev_attr_available_slots.attr,
&dev_attr_security.attr,
&dev_attr_frozen.attr,
2015-06-08 18:27:06 +00:00
NULL,
};
static umode_t nvdimm_visible(struct kobject *kobj, struct attribute *a, int n)
{
struct device *dev = container_of(kobj, typeof(*dev), kobj);
struct nvdimm *nvdimm = to_nvdimm(dev);
if (a != &dev_attr_security.attr && a != &dev_attr_frozen.attr)
return a->mode;
if (!nvdimm->sec.flags)
return 0;
if (a == &dev_attr_security.attr) {
/* Are there any state mutation ops (make writable)? */
if (nvdimm->sec.ops->freeze || nvdimm->sec.ops->disable
|| nvdimm->sec.ops->change_key
|| nvdimm->sec.ops->erase
|| nvdimm->sec.ops->overwrite)
return a->mode;
return 0444;
}
if (nvdimm->sec.ops->freeze)
return a->mode;
return 0;
}
static const struct attribute_group nvdimm_attribute_group = {
2015-06-08 18:27:06 +00:00
.attrs = nvdimm_attributes,
.is_visible = nvdimm_visible,
2015-06-08 18:27:06 +00:00
};
PM, libnvdimm: Add runtime firmware activation support Abstract platform specific mechanics for nvdimm firmware activation behind a handful of generic ops. At the bus level ->activate_state() indicates the unified state (idle, busy, armed) of all DIMMs on the bus, and ->capability() indicates the system state expectations for activate. At the DIMM level ->activate_state() indicates the per-DIMM state, ->activate_result() indicates the outcome of the last activation attempt, and ->arm() attempts to transition the DIMM from 'idle' to 'armed'. A new hibernate_quiet_exec() facility is added to support firmware activation in an OS defined system quiesce state. It leverages the fact that the hibernate-freeze state wants to assert that a memory hibernation snapshot can be taken. This is in contrast to a platform firmware defined quiesce state that may forcefully quiet the memory controller independent of whether an individual device-driver properly supports hibernate-freeze. The libnvdimm sysfs interface is extended to support detection of a firmware activate capability. The mechanism supports enumeration and triggering of firmware activate, optionally in the hibernate_quiet_exec() context. [rafael: hibernate_quiet_exec() proposal] [vishal: fix up sparse warning, grammar in Documentation/] Cc: Pavel Machek <pavel@ucw.cz> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Len Brown <len.brown@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Vishal Verma <vishal.l.verma@intel.com> Reported-by: kernel test robot <lkp@intel.com> Co-developed-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Signed-off-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
2020-07-20 22:08:18 +00:00
static ssize_t result_show(struct device *dev, struct device_attribute *attr, char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
enum nvdimm_fwa_result result;
if (!nvdimm->fw_ops)
return -EOPNOTSUPP;
nvdimm_bus_lock(dev);
result = nvdimm->fw_ops->activate_result(nvdimm);
nvdimm_bus_unlock(dev);
switch (result) {
case NVDIMM_FWA_RESULT_NONE:
return sprintf(buf, "none\n");
case NVDIMM_FWA_RESULT_SUCCESS:
return sprintf(buf, "success\n");
case NVDIMM_FWA_RESULT_FAIL:
return sprintf(buf, "fail\n");
case NVDIMM_FWA_RESULT_NOTSTAGED:
return sprintf(buf, "not_staged\n");
case NVDIMM_FWA_RESULT_NEEDRESET:
return sprintf(buf, "need_reset\n");
default:
return -ENXIO;
}
}
static DEVICE_ATTR_ADMIN_RO(result);
static ssize_t activate_show(struct device *dev, struct device_attribute *attr, char *buf)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
enum nvdimm_fwa_state state;
if (!nvdimm->fw_ops)
return -EOPNOTSUPP;
nvdimm_bus_lock(dev);
state = nvdimm->fw_ops->activate_state(nvdimm);
nvdimm_bus_unlock(dev);
switch (state) {
case NVDIMM_FWA_IDLE:
return sprintf(buf, "idle\n");
case NVDIMM_FWA_BUSY:
return sprintf(buf, "busy\n");
case NVDIMM_FWA_ARMED:
return sprintf(buf, "armed\n");
default:
return -ENXIO;
}
}
static ssize_t activate_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t len)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
enum nvdimm_fwa_trigger arg;
int rc;
if (!nvdimm->fw_ops)
return -EOPNOTSUPP;
if (sysfs_streq(buf, "arm"))
arg = NVDIMM_FWA_ARM;
else if (sysfs_streq(buf, "disarm"))
arg = NVDIMM_FWA_DISARM;
else
return -EINVAL;
nvdimm_bus_lock(dev);
rc = nvdimm->fw_ops->arm(nvdimm, arg);
nvdimm_bus_unlock(dev);
if (rc < 0)
return rc;
return len;
}
static DEVICE_ATTR_ADMIN_RW(activate);
static struct attribute *nvdimm_firmware_attributes[] = {
&dev_attr_activate.attr,
&dev_attr_result.attr,
libnvdimm: KASAN: global-out-of-bounds Read in internal_create_group Because the last member of the "nvdimm_firmware_attributes" array was not assigned a null ptr, when traversal of "grp->attrs" array is out of bounds in "create_files" func. func: create_files: ->for (i = 0, attr = grp->attrs; *attr && !error; i++, attr++) ->.... BUG: KASAN: global-out-of-bounds in create_files fs/sysfs/group.c:43 [inline] BUG: KASAN: global-out-of-bounds in internal_create_group+0x9d8/0xb20 fs/sysfs/group.c:149 Read of size 8 at addr ffffffff8a2e4cf0 by task kworker/u17:10/959 CPU: 2 PID: 959 Comm: kworker/u17:10 Not tainted 5.8.0-syzkaller #0 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014 Workqueue: events_unbound async_run_entry_fn Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x18f/0x20d lib/dump_stack.c:118 print_address_description.constprop.0.cold+0x5/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 create_files fs/sysfs/group.c:43 [inline] internal_create_group+0x9d8/0xb20 fs/sysfs/group.c:149 internal_create_groups.part.0+0x90/0x140 fs/sysfs/group.c:189 internal_create_groups fs/sysfs/group.c:185 [inline] sysfs_create_groups+0x25/0x50 fs/sysfs/group.c:215 device_add_groups drivers/base/core.c:2024 [inline] device_add_attrs drivers/base/core.c:2178 [inline] device_add+0x7fd/0x1c40 drivers/base/core.c:2881 nd_async_device_register+0x12/0x80 drivers/nvdimm/bus.c:506 async_run_entry_fn+0x121/0x530 kernel/async.c:123 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the variable: nvdimm_firmware_attributes+0x10/0x40 Link: https://lore.kernel.org/r/20200812085501.30963-1-qiang.zhang@windriver.com Link: https://lore.kernel.org/r/20200814150509.225615-1-vaibhav@linux.ibm.com Fixes: 48001ea50d17f ("PM, libnvdimm: Add runtime firmware activation support") Reported-by: syzbot+1cf0ffe61aecf46f588f@syzkaller.appspotmail.com Reported-by: Sandipan Das <sandipan@linux.ibm.com> Reported-by: Vaibhav Jain <vaibhav@linux.ibm.com> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Zqiang <qiang.zhang@windriver.com> Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
2020-08-12 08:55:01 +00:00
NULL,
PM, libnvdimm: Add runtime firmware activation support Abstract platform specific mechanics for nvdimm firmware activation behind a handful of generic ops. At the bus level ->activate_state() indicates the unified state (idle, busy, armed) of all DIMMs on the bus, and ->capability() indicates the system state expectations for activate. At the DIMM level ->activate_state() indicates the per-DIMM state, ->activate_result() indicates the outcome of the last activation attempt, and ->arm() attempts to transition the DIMM from 'idle' to 'armed'. A new hibernate_quiet_exec() facility is added to support firmware activation in an OS defined system quiesce state. It leverages the fact that the hibernate-freeze state wants to assert that a memory hibernation snapshot can be taken. This is in contrast to a platform firmware defined quiesce state that may forcefully quiet the memory controller independent of whether an individual device-driver properly supports hibernate-freeze. The libnvdimm sysfs interface is extended to support detection of a firmware activate capability. The mechanism supports enumeration and triggering of firmware activate, optionally in the hibernate_quiet_exec() context. [rafael: hibernate_quiet_exec() proposal] [vishal: fix up sparse warning, grammar in Documentation/] Cc: Pavel Machek <pavel@ucw.cz> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Len Brown <len.brown@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Vishal Verma <vishal.l.verma@intel.com> Reported-by: kernel test robot <lkp@intel.com> Co-developed-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Signed-off-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
2020-07-20 22:08:18 +00:00
};
static umode_t nvdimm_firmware_visible(struct kobject *kobj, struct attribute *a, int n)
{
struct device *dev = container_of(kobj, typeof(*dev), kobj);
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
struct nvdimm *nvdimm = to_nvdimm(dev);
enum nvdimm_fwa_capability cap;
if (!nd_desc->fw_ops)
return 0;
if (!nvdimm->fw_ops)
return 0;
nvdimm_bus_lock(dev);
cap = nd_desc->fw_ops->capability(nd_desc);
nvdimm_bus_unlock(dev);
if (cap < NVDIMM_FWA_CAP_QUIESCE)
return 0;
return a->mode;
}
static const struct attribute_group nvdimm_firmware_attribute_group = {
.name = "firmware",
.attrs = nvdimm_firmware_attributes,
.is_visible = nvdimm_firmware_visible,
};
static const struct attribute_group *nvdimm_attribute_groups[] = {
&nd_device_attribute_group,
&nvdimm_attribute_group,
PM, libnvdimm: Add runtime firmware activation support Abstract platform specific mechanics for nvdimm firmware activation behind a handful of generic ops. At the bus level ->activate_state() indicates the unified state (idle, busy, armed) of all DIMMs on the bus, and ->capability() indicates the system state expectations for activate. At the DIMM level ->activate_state() indicates the per-DIMM state, ->activate_result() indicates the outcome of the last activation attempt, and ->arm() attempts to transition the DIMM from 'idle' to 'armed'. A new hibernate_quiet_exec() facility is added to support firmware activation in an OS defined system quiesce state. It leverages the fact that the hibernate-freeze state wants to assert that a memory hibernation snapshot can be taken. This is in contrast to a platform firmware defined quiesce state that may forcefully quiet the memory controller independent of whether an individual device-driver properly supports hibernate-freeze. The libnvdimm sysfs interface is extended to support detection of a firmware activate capability. The mechanism supports enumeration and triggering of firmware activate, optionally in the hibernate_quiet_exec() context. [rafael: hibernate_quiet_exec() proposal] [vishal: fix up sparse warning, grammar in Documentation/] Cc: Pavel Machek <pavel@ucw.cz> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Len Brown <len.brown@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Vishal Verma <vishal.l.verma@intel.com> Reported-by: kernel test robot <lkp@intel.com> Co-developed-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Signed-off-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
2020-07-20 22:08:18 +00:00
&nvdimm_firmware_attribute_group,
NULL,
};
static const struct device_type nvdimm_device_type = {
.name = "nvdimm",
.release = nvdimm_release,
.groups = nvdimm_attribute_groups,
};
bool is_nvdimm(struct device *dev)
{
return dev->type == &nvdimm_device_type;
}
2015-06-08 18:27:06 +00:00
struct nvdimm *__nvdimm_create(struct nvdimm_bus *nvdimm_bus,
void *provider_data, const struct attribute_group **groups,
unsigned long flags, unsigned long cmd_mask, int num_flush,
struct resource *flush_wpq, const char *dimm_id,
const struct nvdimm_security_ops *sec_ops,
const struct nvdimm_fw_ops *fw_ops)
{
struct nvdimm *nvdimm = kzalloc(sizeof(*nvdimm), GFP_KERNEL);
struct device *dev;
if (!nvdimm)
return NULL;
nvdimm->id = ida_simple_get(&dimm_ida, 0, 0, GFP_KERNEL);
if (nvdimm->id < 0) {
kfree(nvdimm);
return NULL;
}
nvdimm->dimm_id = dimm_id;
nvdimm->provider_data = provider_data;
if (noblk)
flags |= 1 << NDD_NOBLK;
nvdimm->flags = flags;
nvdimm->cmd_mask = cmd_mask;
nvdimm->num_flush = num_flush;
nvdimm->flush_wpq = flush_wpq;
2015-05-01 17:11:27 +00:00
atomic_set(&nvdimm->busy, 0);
dev = &nvdimm->dev;
dev_set_name(dev, "nmem%d", nvdimm->id);
dev->parent = &nvdimm_bus->dev;
dev->type = &nvdimm_device_type;
2015-06-08 18:27:06 +00:00
dev->devt = MKDEV(nvdimm_major, nvdimm->id);
dev->groups = groups;
nvdimm->sec.ops = sec_ops;
nvdimm->fw_ops = fw_ops;
nvdimm->sec.overwrite_tmo = 0;
INIT_DELAYED_WORK(&nvdimm->dwork, nvdimm_security_overwrite_query);
/*
* Security state must be initialized before device_add() for
* attribute visibility.
*/
/* get security state and extended (master) state */
nvdimm->sec.flags = nvdimm_security_flags(nvdimm, NVDIMM_USER);
nvdimm->sec.ext_flags = nvdimm_security_flags(nvdimm, NVDIMM_MASTER);
nd_device_register(dev);
return nvdimm;
}
EXPORT_SYMBOL_GPL(__nvdimm_create);
static void shutdown_security_notify(void *data)
{
struct nvdimm *nvdimm = data;
sysfs_put(nvdimm->sec.overwrite_state);
}
int nvdimm_security_setup_events(struct device *dev)
{
struct nvdimm *nvdimm = to_nvdimm(dev);
if (!nvdimm->sec.flags || !nvdimm->sec.ops
|| !nvdimm->sec.ops->overwrite)
return 0;
nvdimm->sec.overwrite_state = sysfs_get_dirent(dev->kobj.sd, "security");
if (!nvdimm->sec.overwrite_state)
return -ENOMEM;
return devm_add_action_or_reset(dev, shutdown_security_notify, nvdimm);
}
EXPORT_SYMBOL_GPL(nvdimm_security_setup_events);
int nvdimm_in_overwrite(struct nvdimm *nvdimm)
{
return test_bit(NDD_SECURITY_OVERWRITE, &nvdimm->flags);
}
EXPORT_SYMBOL_GPL(nvdimm_in_overwrite);
int nvdimm_security_freeze(struct nvdimm *nvdimm)
{
int rc;
WARN_ON_ONCE(!is_nvdimm_bus_locked(&nvdimm->dev));
if (!nvdimm->sec.ops || !nvdimm->sec.ops->freeze)
return -EOPNOTSUPP;
if (!nvdimm->sec.flags)
return -EIO;
if (test_bit(NDD_SECURITY_OVERWRITE, &nvdimm->flags)) {
dev_warn(&nvdimm->dev, "Overwrite operation in progress.\n");
return -EBUSY;
}
rc = nvdimm->sec.ops->freeze(nvdimm);
nvdimm->sec.flags = nvdimm_security_flags(nvdimm, NVDIMM_USER);
return rc;
}
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
static unsigned long dpa_align(struct nd_region *nd_region)
{
struct device *dev = &nd_region->dev;
if (dev_WARN_ONCE(dev, !is_nvdimm_bus_locked(dev),
"bus lock required for capacity provision\n"))
return 0;
if (dev_WARN_ONCE(dev, !nd_region->ndr_mappings || nd_region->align
% nd_region->ndr_mappings,
"invalid region align %#lx mappings: %d\n",
nd_region->align, nd_region->ndr_mappings))
return 0;
return nd_region->align / nd_region->ndr_mappings;
}
int alias_dpa_busy(struct device *dev, void *data)
{
libnvdimm: fix blk free space accounting Commit a1f3e4d6a0c3 "libnvdimm, region: update nd_region_available_dpa() for multi-pmem support" reworked blk dpa (DIMM Physical Address) accounting to comprehend multiple pmem namespace allocations aliasing with a given blk-dpa range. The following call trace is a result of failing to account for allocated blk capacity. WARNING: CPU: 1 PID: 2433 at tools/testing/nvdimm/../../../drivers/nvdimm/names 4 size_store+0x6f3/0x930 [libnvdimm] nd_region region5: allocation underrun: 0x0 of 0x1000000 bytes [..] Call Trace: dump_stack+0x86/0xc3 __warn+0xcb/0xf0 warn_slowpath_fmt+0x5f/0x80 size_store+0x6f3/0x930 [libnvdimm] dev_attr_store+0x18/0x30 If a given blk-dpa allocation does not alias with any pmem ranges then the full allocation should be accounted as busy space, not the size of the current pmem contribution to the region. The thinkos that led to this confusion was not realizing that the struct resource management is already guaranteeing no collisions between pmem allocations and blk allocations on the same dimm. Also, we do not try to support blk allocations in aliased pmem holes. This patch also fixes a case where the available blk goes negative. Cc: <stable@vger.kernel.org> Fixes: a1f3e4d6a0c3 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support"). Reported-by: Dariusz Dokupil <dariusz.dokupil@intel.com> Reported-by: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-04-04 22:08:36 +00:00
resource_size_t map_end, blk_start, new;
struct blk_alloc_info *info = data;
struct nd_mapping *nd_mapping;
struct nd_region *nd_region;
struct nvdimm_drvdata *ndd;
struct resource *res;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
unsigned long align;
int i;
if (!is_memory(dev))
return 0;
nd_region = to_nd_region(dev);
for (i = 0; i < nd_region->ndr_mappings; i++) {
nd_mapping = &nd_region->mapping[i];
if (nd_mapping->nvdimm == info->nd_mapping->nvdimm)
break;
}
if (i >= nd_region->ndr_mappings)
return 0;
ndd = to_ndd(nd_mapping);
map_end = nd_mapping->start + nd_mapping->size - 1;
blk_start = nd_mapping->start;
/*
* In the allocation case ->res is set to free space that we are
* looking to validate against PMEM aliasing collision rules
* (i.e. BLK is allocated after all aliased PMEM).
*/
if (info->res) {
if (info->res->start >= nd_mapping->start
&& info->res->start < map_end)
/* pass */;
else
return 0;
}
retry:
/*
* Find the free dpa from the end of the last pmem allocation to
libnvdimm: fix blk free space accounting Commit a1f3e4d6a0c3 "libnvdimm, region: update nd_region_available_dpa() for multi-pmem support" reworked blk dpa (DIMM Physical Address) accounting to comprehend multiple pmem namespace allocations aliasing with a given blk-dpa range. The following call trace is a result of failing to account for allocated blk capacity. WARNING: CPU: 1 PID: 2433 at tools/testing/nvdimm/../../../drivers/nvdimm/names 4 size_store+0x6f3/0x930 [libnvdimm] nd_region region5: allocation underrun: 0x0 of 0x1000000 bytes [..] Call Trace: dump_stack+0x86/0xc3 __warn+0xcb/0xf0 warn_slowpath_fmt+0x5f/0x80 size_store+0x6f3/0x930 [libnvdimm] dev_attr_store+0x18/0x30 If a given blk-dpa allocation does not alias with any pmem ranges then the full allocation should be accounted as busy space, not the size of the current pmem contribution to the region. The thinkos that led to this confusion was not realizing that the struct resource management is already guaranteeing no collisions between pmem allocations and blk allocations on the same dimm. Also, we do not try to support blk allocations in aliased pmem holes. This patch also fixes a case where the available blk goes negative. Cc: <stable@vger.kernel.org> Fixes: a1f3e4d6a0c3 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support"). Reported-by: Dariusz Dokupil <dariusz.dokupil@intel.com> Reported-by: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-04-04 22:08:36 +00:00
* the end of the interleave-set mapping.
*/
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
align = dpa_align(nd_region);
if (!align)
return 0;
for_each_dpa_resource(ndd, res) {
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
resource_size_t start, end;
libnvdimm: fix blk free space accounting Commit a1f3e4d6a0c3 "libnvdimm, region: update nd_region_available_dpa() for multi-pmem support" reworked blk dpa (DIMM Physical Address) accounting to comprehend multiple pmem namespace allocations aliasing with a given blk-dpa range. The following call trace is a result of failing to account for allocated blk capacity. WARNING: CPU: 1 PID: 2433 at tools/testing/nvdimm/../../../drivers/nvdimm/names 4 size_store+0x6f3/0x930 [libnvdimm] nd_region region5: allocation underrun: 0x0 of 0x1000000 bytes [..] Call Trace: dump_stack+0x86/0xc3 __warn+0xcb/0xf0 warn_slowpath_fmt+0x5f/0x80 size_store+0x6f3/0x930 [libnvdimm] dev_attr_store+0x18/0x30 If a given blk-dpa allocation does not alias with any pmem ranges then the full allocation should be accounted as busy space, not the size of the current pmem contribution to the region. The thinkos that led to this confusion was not realizing that the struct resource management is already guaranteeing no collisions between pmem allocations and blk allocations on the same dimm. Also, we do not try to support blk allocations in aliased pmem holes. This patch also fixes a case where the available blk goes negative. Cc: <stable@vger.kernel.org> Fixes: a1f3e4d6a0c3 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support"). Reported-by: Dariusz Dokupil <dariusz.dokupil@intel.com> Reported-by: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-04-04 22:08:36 +00:00
if (strncmp(res->name, "pmem", 4) != 0)
continue;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
start = ALIGN_DOWN(res->start, align);
end = ALIGN(res->end + 1, align) - 1;
if ((start >= blk_start && start < map_end)
|| (end >= blk_start && end <= map_end)) {
new = max(blk_start, min(map_end, end) + 1);
libnvdimm: fix blk free space accounting Commit a1f3e4d6a0c3 "libnvdimm, region: update nd_region_available_dpa() for multi-pmem support" reworked blk dpa (DIMM Physical Address) accounting to comprehend multiple pmem namespace allocations aliasing with a given blk-dpa range. The following call trace is a result of failing to account for allocated blk capacity. WARNING: CPU: 1 PID: 2433 at tools/testing/nvdimm/../../../drivers/nvdimm/names 4 size_store+0x6f3/0x930 [libnvdimm] nd_region region5: allocation underrun: 0x0 of 0x1000000 bytes [..] Call Trace: dump_stack+0x86/0xc3 __warn+0xcb/0xf0 warn_slowpath_fmt+0x5f/0x80 size_store+0x6f3/0x930 [libnvdimm] dev_attr_store+0x18/0x30 If a given blk-dpa allocation does not alias with any pmem ranges then the full allocation should be accounted as busy space, not the size of the current pmem contribution to the region. The thinkos that led to this confusion was not realizing that the struct resource management is already guaranteeing no collisions between pmem allocations and blk allocations on the same dimm. Also, we do not try to support blk allocations in aliased pmem holes. This patch also fixes a case where the available blk goes negative. Cc: <stable@vger.kernel.org> Fixes: a1f3e4d6a0c3 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support"). Reported-by: Dariusz Dokupil <dariusz.dokupil@intel.com> Reported-by: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-04-04 22:08:36 +00:00
if (new != blk_start) {
blk_start = new;
goto retry;
}
}
}
/* update the free space range with the probed blk_start */
if (info->res && blk_start > info->res->start) {
info->res->start = max(info->res->start, blk_start);
if (info->res->start > info->res->end)
info->res->end = info->res->start - 1;
return 1;
}
libnvdimm: fix blk free space accounting Commit a1f3e4d6a0c3 "libnvdimm, region: update nd_region_available_dpa() for multi-pmem support" reworked blk dpa (DIMM Physical Address) accounting to comprehend multiple pmem namespace allocations aliasing with a given blk-dpa range. The following call trace is a result of failing to account for allocated blk capacity. WARNING: CPU: 1 PID: 2433 at tools/testing/nvdimm/../../../drivers/nvdimm/names 4 size_store+0x6f3/0x930 [libnvdimm] nd_region region5: allocation underrun: 0x0 of 0x1000000 bytes [..] Call Trace: dump_stack+0x86/0xc3 __warn+0xcb/0xf0 warn_slowpath_fmt+0x5f/0x80 size_store+0x6f3/0x930 [libnvdimm] dev_attr_store+0x18/0x30 If a given blk-dpa allocation does not alias with any pmem ranges then the full allocation should be accounted as busy space, not the size of the current pmem contribution to the region. The thinkos that led to this confusion was not realizing that the struct resource management is already guaranteeing no collisions between pmem allocations and blk allocations on the same dimm. Also, we do not try to support blk allocations in aliased pmem holes. This patch also fixes a case where the available blk goes negative. Cc: <stable@vger.kernel.org> Fixes: a1f3e4d6a0c3 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support"). Reported-by: Dariusz Dokupil <dariusz.dokupil@intel.com> Reported-by: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2017-04-04 22:08:36 +00:00
info->available -= blk_start - nd_mapping->start;
return 0;
}
/**
* nd_blk_available_dpa - account the unused dpa of BLK region
* @nd_mapping: container of dpa-resource-root + labels
*
* Unlike PMEM, BLK namespaces can occupy discontiguous DPA ranges, but
* we arrange for them to never start at an lower dpa than the last
* PMEM allocation in an aliased region.
*/
resource_size_t nd_blk_available_dpa(struct nd_region *nd_region)
{
struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(&nd_region->dev);
struct nd_mapping *nd_mapping = &nd_region->mapping[0];
struct nvdimm_drvdata *ndd = to_ndd(nd_mapping);
struct blk_alloc_info info = {
.nd_mapping = nd_mapping,
.available = nd_mapping->size,
.res = NULL,
};
struct resource *res;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
unsigned long align;
if (!ndd)
return 0;
device_for_each_child(&nvdimm_bus->dev, &info, alias_dpa_busy);
/* now account for busy blk allocations in unaliased dpa */
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
align = dpa_align(nd_region);
if (!align)
return 0;
for_each_dpa_resource(ndd, res) {
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
resource_size_t start, end, size;
if (strncmp(res->name, "blk", 3) != 0)
continue;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
start = ALIGN_DOWN(res->start, align);
end = ALIGN(res->end + 1, align) - 1;
size = end - start + 1;
if (size >= info.available)
return 0;
info.available -= size;
}
return info.available;
}
/**
* nd_pmem_max_contiguous_dpa - For the given dimm+region, return the max
* contiguous unallocated dpa range.
* @nd_region: constrain available space check to this reference region
* @nd_mapping: container of dpa-resource-root + labels
*/
resource_size_t nd_pmem_max_contiguous_dpa(struct nd_region *nd_region,
struct nd_mapping *nd_mapping)
{
struct nvdimm_drvdata *ndd = to_ndd(nd_mapping);
struct nvdimm_bus *nvdimm_bus;
resource_size_t max = 0;
struct resource *res;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
unsigned long align;
/* if a dimm is disabled the available capacity is zero */
if (!ndd)
return 0;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
align = dpa_align(nd_region);
if (!align)
return 0;
nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
if (__reserve_free_pmem(&nd_region->dev, nd_mapping->nvdimm))
return 0;
for_each_dpa_resource(ndd, res) {
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
resource_size_t start, end;
if (strcmp(res->name, "pmem-reserve") != 0)
continue;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
/* trim free space relative to current alignment setting */
start = ALIGN(res->start, align);
end = ALIGN_DOWN(res->end + 1, align) - 1;
if (end < start)
continue;
if (end - start + 1 > max)
max = end - start + 1;
}
release_free_pmem(nvdimm_bus, nd_mapping);
return max;
}
/**
* nd_pmem_available_dpa - for the given dimm+region account unallocated dpa
* @nd_mapping: container of dpa-resource-root + labels
* @nd_region: constrain available space check to this reference region
* @overlap: calculate available space assuming this level of overlap
*
* Validate that a PMEM label, if present, aligns with the start of an
* interleave set and truncate the available size at the lowest BLK
* overlap point.
*
* The expectation is that this routine is called multiple times as it
* probes for the largest BLK encroachment for any single member DIMM of
* the interleave set. Once that value is determined the PMEM-limit for
* the set can be established.
*/
resource_size_t nd_pmem_available_dpa(struct nd_region *nd_region,
struct nd_mapping *nd_mapping, resource_size_t *overlap)
{
resource_size_t map_start, map_end, busy = 0, available, blk_start;
struct nvdimm_drvdata *ndd = to_ndd(nd_mapping);
struct resource *res;
const char *reason;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
unsigned long align;
if (!ndd)
return 0;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
align = dpa_align(nd_region);
if (!align)
return 0;
map_start = nd_mapping->start;
map_end = map_start + nd_mapping->size - 1;
blk_start = max(map_start, map_end + 1 - *overlap);
for_each_dpa_resource(ndd, res) {
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
resource_size_t start, end;
start = ALIGN_DOWN(res->start, align);
end = ALIGN(res->end + 1, align) - 1;
if (start >= map_start && start < map_end) {
if (strncmp(res->name, "blk", 3) == 0)
blk_start = min(blk_start,
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
max(map_start, start));
else if (end > map_end) {
reason = "misaligned to iset";
goto err;
} else
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
busy += end - start + 1;
} else if (end >= map_start && end <= map_end) {
if (strncmp(res->name, "blk", 3) == 0) {
/*
* If a BLK allocation overlaps the start of
* PMEM the entire interleave set may now only
* be used for BLK.
*/
blk_start = map_start;
} else
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
busy += end - start + 1;
} else if (map_start > start && map_start < end) {
/* total eclipse of the mapping */
busy += nd_mapping->size;
blk_start = map_start;
}
}
*overlap = map_end + 1 - blk_start;
available = blk_start - map_start;
if (busy < available)
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-30 20:06:23 +00:00
return ALIGN_DOWN(available - busy, align);
return 0;
err:
nd_dbg_dpa(nd_region, ndd, res, "%s\n", reason);
return 0;
}
void nvdimm_free_dpa(struct nvdimm_drvdata *ndd, struct resource *res)
{
WARN_ON_ONCE(!is_nvdimm_bus_locked(ndd->dev));
kfree(res->name);
__release_region(&ndd->dpa, res->start, resource_size(res));
}
struct resource *nvdimm_allocate_dpa(struct nvdimm_drvdata *ndd,
struct nd_label_id *label_id, resource_size_t start,
resource_size_t n)
{
char *name = kmemdup(label_id, sizeof(*label_id), GFP_KERNEL);
struct resource *res;
if (!name)
return NULL;
WARN_ON_ONCE(!is_nvdimm_bus_locked(ndd->dev));
res = __request_region(&ndd->dpa, start, n, name, 0);
if (!res)
kfree(name);
return res;
}
/**
* nvdimm_allocated_dpa - sum up the dpa currently allocated to this label_id
* @nvdimm: container of dpa-resource-root + labels
* @label_id: dpa resource name of the form {pmem|blk}-<human readable uuid>
*/
resource_size_t nvdimm_allocated_dpa(struct nvdimm_drvdata *ndd,
struct nd_label_id *label_id)
{
resource_size_t allocated = 0;
struct resource *res;
for_each_dpa_resource(ndd, res)
if (strcmp(res->name, label_id->id) == 0)
allocated += resource_size(res);
return allocated;
}
static int count_dimms(struct device *dev, void *c)
{
int *count = c;
if (is_nvdimm(dev))
(*count)++;
return 0;
}
int nvdimm_bus_check_dimm_count(struct nvdimm_bus *nvdimm_bus, int dimm_count)
{
int count = 0;
/* Flush any possible dimm registration failures */
nd_synchronize();
device_for_each_child(&nvdimm_bus->dev, &count, count_dimms);
dev_dbg(&nvdimm_bus->dev, "count: %d\n", count);
if (count != dimm_count)
return -ENXIO;
return 0;
}
EXPORT_SYMBOL_GPL(nvdimm_bus_check_dimm_count);
void __exit nvdimm_devs_exit(void)
{
ida_destroy(&dimm_ida);
}