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
A multithreaded namespace creation/destruction stress test currently
deadlocks with the following lockup signature:
INFO: task ndctl:2924 blocked for more than 122 seconds.
Tainted: G OE 5.2.0-rc4+ #3382
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
ndctl D 0 2924 1176 0x00000000
Call Trace:
? __schedule+0x27e/0x780
schedule+0x30/0xb0
wait_nvdimm_bus_probe_idle+0x8a/0xd0 [libnvdimm]
? finish_wait+0x80/0x80
uuid_store+0xe6/0x2e0 [libnvdimm]
kernfs_fop_write+0xf0/0x1a0
vfs_write+0xb7/0x1b0
ksys_write+0x5c/0xd0
do_syscall_64+0x60/0x240
INFO: task ndctl:2923 blocked for more than 122 seconds.
Tainted: G OE 5.2.0-rc4+ #3382
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
ndctl D 0 2923 1175 0x00000000
Call Trace:
? __schedule+0x27e/0x780
? __mutex_lock+0x489/0x910
schedule+0x30/0xb0
schedule_preempt_disabled+0x11/0x20
__mutex_lock+0x48e/0x910
? nvdimm_namespace_common_probe+0x95/0x4d0 [libnvdimm]
? __lock_acquire+0x23f/0x1710
? nvdimm_namespace_common_probe+0x95/0x4d0 [libnvdimm]
nvdimm_namespace_common_probe+0x95/0x4d0 [libnvdimm]
__dax_pmem_probe+0x5e/0x210 [dax_pmem_core]
? nvdimm_bus_probe+0x1d0/0x2c0 [libnvdimm]
dax_pmem_probe+0xc/0x20 [dax_pmem]
nvdimm_bus_probe+0x90/0x2c0 [libnvdimm]
really_probe+0xef/0x390
driver_probe_device+0xb4/0x100
In this sequence an 'nd_dax' device is being probed and trying to take
the lock on its backing namespace to validate that the 'nd_dax' device
indeed has exclusive access to the backing namespace. Meanwhile, another
thread is trying to update the uuid property of that same backing
namespace. So one thread is in the probe path trying to acquire the
lock, and the other thread has acquired the lock and tries to flush the
probe path.
Fix this deadlock by not holding the namespace device_lock over the
wait_nvdimm_bus_probe_idle() synchronization step. In turn this requires
the device_lock to be held on entry to wait_nvdimm_bus_probe_idle() and
subsequently dropped internally to wait_nvdimm_bus_probe_idle().
Cc: <stable@vger.kernel.org>
Fixes: bf9bccc14c ("libnvdimm: pmem label sets and namespace instantiation")
Cc: Vishal Verma <vishal.l.verma@intel.com>
Tested-by: Jane Chu <jane.chu@oracle.com>
Link: https://lore.kernel.org/r/156341210094.292348.2384694131126767789.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In preparation for fixing a deadlock between wait_for_bus_probe_idle()
and the nvdimm_bus_list_mutex arrange for __nd_ioctl() without
nvdimm_bus_list_mutex held. This also unifies the 'dimm' and 'bus' level
ioctls into a common nd_ioctl() preamble implementation.
Marked for -stable as it is a pre-requisite for a follow-on fix.
Cc: <stable@vger.kernel.org>
Fixes: bf9bccc14c ("libnvdimm: pmem label sets and namespace instantiation")
Cc: Vishal Verma <vishal.l.verma@intel.com>
Tested-by: Jane Chu <jane.chu@oracle.com>
Link: https://lore.kernel.org/r/156341209518.292348.7183897251740665198.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In preparation for not holding a lock over the execution of nd_ioctl(),
update the implementation to allow multiple threads to be attempting
ioctls at the same time. The bus lock still prevents multiple in-flight
->ndctl() invocations from corrupting each other's state, but static
global staging buffers are moved to the heap.
Reported-by: Vishal Verma <vishal.l.verma@intel.com>
Reviewed-by: Vishal Verma <vishal.l.verma@intel.com>
Tested-by: Vishal Verma <vishal.l.verma@intel.com>
Link: https://lore.kernel.org/r/156341208947.292348.10560140326807607481.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
A multithreaded namespace creation/destruction stress test currently
fails with signatures like the following:
sysfs group 'power' not found for kobject 'dax1.1'
RIP: 0010:sysfs_remove_group+0x76/0x80
Call Trace:
device_del+0x73/0x370
device_unregister+0x16/0x50
nd_async_device_unregister+0x1e/0x30 [libnvdimm]
async_run_entry_fn+0x39/0x160
process_one_work+0x23c/0x5e0
worker_thread+0x3c/0x390
BUG: kernel NULL pointer dereference, address: 0000000000000020
RIP: 0010:klist_put+0x1b/0x6c
Call Trace:
klist_del+0xe/0x10
device_del+0x8a/0x2c9
? __switch_to_asm+0x34/0x70
? __switch_to_asm+0x40/0x70
device_unregister+0x44/0x4f
nd_async_device_unregister+0x22/0x2d [libnvdimm]
async_run_entry_fn+0x47/0x15a
process_one_work+0x1a2/0x2eb
worker_thread+0x1b8/0x26e
Use the kill_device() helper to atomically resolve the race of multiple
threads issuing kill, device_unregister(), requests.
Reported-by: Jane Chu <jane.chu@oracle.com>
Reported-by: Erwin Tsaur <erwin.tsaur@oracle.com>
Fixes: 4d88a97aa9 ("libnvdimm, nvdimm: dimm driver and base libnvdimm device-driver...")
Cc: <stable@vger.kernel.org>
Link: https://github.com/pmem/ndctl/issues/96
Tested-by: Tested-by: Jane Chu <jane.chu@oracle.com>
Link: https://lore.kernel.org/r/156341207846.292348.10435719262819764054.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Based on 1 normalized pattern(s):
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
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 64 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Alexios Zavras <alexios.zavras@intel.com>
Reviewed-by: Allison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190529141901.894819585@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Several places (dimm_devs.c, core.c etc) include label.h but only
label.c uses NSINDEX_SIGNATURE, so move its definition to label.c
instead.
In file included from drivers/nvdimm/dimm_devs.c:23:
drivers/nvdimm/label.h:41:19: warning: 'NSINDEX_SIGNATURE' defined but
not used [-Wunused-const-variable=]
Also, some places abuse "/**" which is only reserved for the kernel-doc.
drivers/nvdimm/bus.c:648: warning: cannot understand function prototype:
'struct attribute_group nd_device_attribute_group = '
drivers/nvdimm/bus.c:677: warning: cannot understand function prototype:
'struct attribute_group nd_numa_attribute_group = '
Those are just some member assignments for the "struct attribute_group"
instances and it can't be expressed in the kernel-doc.
Reviewed-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Force the device registration for nvdimm devices to be closer to the actual
device. This is achieved by using either the NUMA node ID of the region, or
of the parent. By doing this we can have everything above the region based
on the region, and everything below the region based on the nvdimm bus.
By guaranteeing NUMA locality I see an improvement of as high as 25% for
per-node init of a system with 12TB of persistent memory.
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* Use common helpers, bitmap_zalloc() and kstrndup(), to replace open
coded versions.
* Clarify the comments around hotplug vs initial init case for the nfit
driver.
* Cleanup the libnvdimm init path.
Add support for the NVDIMM_FAMILY_INTEL "ovewrite" capability as
described by the Intel DSM spec v1.7. This will allow triggering of
overwrite on Intel NVDIMMs. The overwrite operation can take tens of
minutes. When the overwrite DSM is issued successfully, the NVDIMMs will
be unaccessible. The kernel will do backoff polling to detect when the
overwrite process is completed. According to the DSM spec v1.7, the 128G
NVDIMMs can take up to 15mins to perform overwrite and larger DIMMs will
take longer.
Given that overwrite puts the DIMM in an indeterminate state until it
completes introduce the NDD_SECURITY_OVERWRITE flag to prevent other
operations from executing when overwrite is happening. The
NDD_WORK_PENDING flag is added to denote that there is a device reference
on the nvdimm device for an async workqueue thread context.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Some NVDIMMs, like the ones defined by the NVDIMM_FAMILY_INTEL command
set, expose a security capability to lock the DIMMs at poweroff and
require a passphrase to unlock them. The security model is derived from
ATA security. In anticipation of other DIMMs implementing a similar
scheme, and to abstract the core security implementation away from the
device-specific details, introduce nvdimm_security_ops.
Initially only a status retrieval operation, ->state(), is defined,
along with the base infrastructure and definitions for future
operations.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Co-developed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The id check was not executed immediately following ida_simple_get. Just
change the codes position, without function change.
Signed-off-by: Ocean He <hehy1@lenovo.com>
Reviewed-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Add command definition for security commands defined in Intel DSM
specification v1.8 [1]. This includes "get security state", "set
passphrase", "unlock unit", "freeze lock", "secure erase", "overwrite",
"overwrite query", "master passphrase enable/disable", and "master
erase", . Since this adds several Intel definitions, move the relevant
bits to their own header.
These commands mutate physical data, but that manipulation is not cache
coherent. The requirement to flush and invalidate caches makes these
commands unsuitable to be called from userspace, so extra logic is added
to detect and block these commands from being submitted via the ioctl
command submission path.
Lastly, the commands may contain sensitive key material that should not
be dumped in a standard debug session. Update the nvdimm-command
payload-dump facility to move security command payloads behind a
default-off compile time switch.
[1]: http://pmem.io/documents/NVDIMM_DSM_Interface-V1.8.pdf
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
This change makes it so that we don't repeatedly overwrite the device node
for nvdimm regions. The earliest we can set the node is immediately after
calling device init, so I have moved the code there so we can avoid
rewriting the node with each uevent.
Signed-off-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Unlike asynchronous initialization in the core we have not yet associated
the device with the parent, and as such the device doesn't hold a reference
to the parent.
In order to resolve that we should be holding a reference on the parent
until the asynchronous initialization has completed.
Cc: <stable@vger.kernel.org>
Fixes: 4d88a97aa9 ("libnvdimm: ...base ... infrastructure")
Signed-off-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Commit efda1b5d87 ("acpi, nfit, libnvdimm: fix / harden ars_status output length handling")
Introduced additional hardening for ambiguity in the ACPI spec for
ars_status output sizing. However, it had a couple of cases mixed up.
Where it should have been checking for (and returning) "out_field[1] -
4" it was using "out_field[1] - 8" and vice versa.
This caused a four byte discrepancy in the buffer size passed on to
the command handler, and in some cases, this caused memory corruption
like:
./daxdev-errors.sh: line 76: 24104 Aborted (core dumped) ./daxdev-errors $busdev $region
malloc(): memory corruption
Program received signal SIGABRT, Aborted.
[...]
#5 0x00007ffff7865a2e in calloc () from /lib64/libc.so.6
#6 0x00007ffff7bc2970 in ndctl_bus_cmd_new_ars_status (ars_cap=ars_cap@entry=0x6153b0) at ars.c:136
#7 0x0000000000401644 in check_ars_status (check=0x7fffffffdeb0, bus=0x604c20) at daxdev-errors.c:144
#8 test_daxdev_clear_error (region_name=<optimized out>, bus_name=<optimized out>)
at daxdev-errors.c:332
Cc: <stable@vger.kernel.org>
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Lukasz Dorau <lukasz.dorau@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Fixes: efda1b5d87 ("acpi, nfit, libnvdimm: fix / harden ars_status output length handling")
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-of-by: Dave Jiang <dave.jiang@intel.com>
Instrument nvdimm_bus_probe() to emit timestamps for the start and end
of libnvdimm device probing. This is useful for identifying sources of
libnvdimm sub-system initialization latency.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The pmem driver does not honor a forced read-only setting for very long:
$ blockdev --setro /dev/pmem0
$ blockdev --getro /dev/pmem0
1
followed by various commands like these:
$ blockdev --rereadpt /dev/pmem0
or
$ mkfs.ext4 /dev/pmem0
results in this in the kernel serial log:
nd_pmem namespace0.0: region0 read-write, marking pmem0 read-write
with the read-only setting lost:
$ blockdev --getro /dev/pmem0
0
That's from bus.c nvdimm_revalidate_disk(), which always applies the
setting from nd_region (which is initially based on the ACPI NFIT
NVDIMM state flags not_armed bit).
In contrast, commit 20bd1d026a ("scsi: sd: Keep disk read-only when
re-reading partition") fixed this issue for SCSI devices to preserve
the previous setting if it was set to read-only.
This patch modifies bus.c to preserve any previous read-only setting.
It also eliminates the kernel serial log print except for cases where
read-write is changed to read-only, so it doesn't print read-only to
read-only non-changes.
Cc: <stable@vger.kernel.org>
Fixes: 5813882094 ("libnvdimm, nfit: handle unarmed dimms, mark namespaces read-only")
Signed-off-by: Robert Elliott <elliott@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
We want to be able to cross reference the region and bus devices
with the device tree node that they were spawned from. libNVDIMM
handles creating the actual devices for these internally, so we
need to pass in a pointer to the relevant node in the descriptor.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Dynamic debug can be instructed to add the function name to the debug
output using the +f switch, so there is no need for the libnvdimm
modules to do it again. If a user decides to add the +f switch for
libnvdimm's dynamic debug this results in double prints of the function
name.
Reported-by: Johannes Thumshirn <jthumshirn@suse.de>
Reported-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The kernel's ND_IOCTL_SMART_THRESHOLD command is based on a payload
definition that has become broken / out-of-sync with recent versions of
the NVDIMM_FAMILY_INTEL definition. Deprecate the use of the
ND_IOCTL_SMART_THRESHOLD command in favor of the ND_CMD_CALL approach
taken by NVDIMM_FAMILY_{HPE,MSFT}, where we can manage the per-vendor
variance in userspace.
In a couple years, when the new scheme is widely deployed in userspace
packages, the ND_IOCTL_SMART_THRESHOLD support can be removed. For now
we prevent new binaries from compiling against the kernel header
definitions, but kernel still compatible with old binaries. The
libndctl.h [1] header is now the authoritative interface definition for
NVDIMM SMART.
[1]: https://github.com/pmem/ndctl
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
nfit_test needs to use the poison list manipulation code as well. Make
it more generic and in the process rename poison to badrange, and move
all the related helpers to a new file.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
[vishal: Add badrange.o to nfit_test's Kbuild]
[vishal: add a missed include in bus.c for the new badrange functions]
[vishal: rename all instances of 'be' to 'bre']
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Delay the check of nd_reserved2 to the actual endpoint (acpi_nfit_ctl)
that uses it, as a prevention of a potential double-fetch bug.
While examining the kernel source code, I found a dangerous operation that
could turn into a double-fetch situation (a race condition bug) where
the same userspace memory region are fetched twice into kernel with sanity
checks after the first fetch while missing checks after the second fetch.
In the case of _IOC_NR(ioctl_cmd) == ND_CMD_CALL:
1. The first fetch happens in line 935 copy_from_user(&pkg, p, sizeof(pkg)
2. subsequently `pkg.nd_reserved2` is asserted to be all zeroes
(line 984 to 986).
3. The second fetch happens in line 1022 copy_from_user(buf, p, buf_len)
4. Given that `p` can be fully controlled in userspace, an attacker can
race condition to override the header part of `p`, say,
`((struct nd_cmd_pkg *)p)->nd_reserved2` to arbitrary value
(say nine 0xFFFFFFFF for `nd_reserved2`) after the first fetch but before the
second fetch. The changed value will be copied to `buf`.
5. There is no checks on the second fetches until the use of it in
line 1034: nd_cmd_clear_to_send(nvdimm_bus, nvdimm, cmd, buf) and
line 1038: nd_desc->ndctl(nd_desc, nvdimm, cmd, buf, buf_len, &cmd_rc)
which means that the assumed relation, `p->nd_reserved2` are all zeroes might
not hold after the second fetch. And once the control goes to these functions
we lose the context to assert the assumed relation.
6. Based on my manual analysis, `p->nd_reserved2` is not used in function
`nd_cmd_clear_to_send` and potential implementations of `nd_desc->ndctl`
so there is no working exploit against it right now. However, this could
easily turns to an exploitable one if careless developers start to use
`p->nd_reserved2` later and assume that they are all zeroes.
Move the validation of the nd_reserved2 field to the ->ndctl()
implementation where it has a stable buffer to evaluate.
Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Dan reports:
The patch 62232e45f4: "libnvdimm: control (ioctl) messages for
nvdimm_bus and nvdimm devices" from Jun 8, 2015, leads to the
following static checker warning:
drivers/nvdimm/bus.c:1018 __nd_ioctl()
warn: integer overflows 'buf_len'
From a casual review, this seems like it might be a real bug. On
the first iteration we load some data into in_env[]. On the second
iteration we read a use controlled "in_size" from nd_cmd_in_size().
It can go up to UINT_MAX - 1. A high number means we will fill the
whole in_env[] buffer. But we potentially keep looping and adding
more to in_len so now it can be any value.
It simple enough to change, but it feels weird that we keep looping
even though in_env is totally full. Shouldn't we just return an
error if we don't have space for desc->in_num.
We keep looping because the size of the total input is allowed to be
bigger than the 'envelope' which is a subset of the payload that tells
us how much data to expect. For safety explicitly check that buf_len
does not overflow which is what the checker flagged.
Cc: <stable@vger.kernel.org>
Fixes: 62232e45f4: "libnvdimm: control (ioctl) messages for nvdimm_bus..."
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
With the ACPI NFIT 'DSM' methods, acpi can be called from IO paths.
Specifically, the DSM to clear media errors is called during writes, so
that we can provide a writes-fix-errors model.
However it is easy to imagine a scenario like:
-> write through the nvdimm driver
-> acpi allocation
-> writeback, causes more IO through the nvdimm driver
-> deadlock
Fix this by using memalloc_noio_{save,restore}, which sets the GFP_NOIO
flag for the current scope when issuing commands/IOs that are expected
to clear errors.
Cc: <linux-acpi@vger.kernel.org>
Cc: <linux-nvdimm@lists.01.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Robert Moore <robert.moore@intel.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Have dsm functions called via the pass thru mechanism also
be checked against clear to send.
Signed-off-by: Jerry Hoemann <jerry.hoemann@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Allow volatile nfit ranges to participate in all the same infrastructure
provided for persistent memory regions. A resulting resulting namespace
device will still be called "pmem", but the parent region type will be
"nd_volatile". This is in preparation for disabling the dax ->flush()
operation in the pmem driver when it is hosted on a volatile range.
Cc: Jan Kara <jack@suse.cz>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Sysfs "badblocks" information may be updated during run-time that:
- MCE, SCI, and sysfs "scrub" may add new bad blocks
- Writes and ioctl() may clear bad blocks
Add support to send sysfs notifications to sysfs "badblocks" file
under region and pmem directories when their badblocks information
is re-evaluated (but is not necessarily changed) during run-time.
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Cc: Linda Knippers <linda.knippers@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Toshi noticed that the new support for a region-level badblocks missed
the case where errors are cleared due to BTT I/O.
An initial attempt to fix this ran into a "sleeping while atomic"
warning due to taking the nvdimm_bus_lock() in the BTT I/O path to
satisfy the locking requirements of __nvdimm_bus_badblocks_clear().
However, that lock is not needed since we are not acting on any data that
is subject to change under that lock. The badblocks instance has its own
internal lock to handle mutations of the error list.
So, in order to make it clear that we are just acting on region devices,
rename __nvdimm_bus_badblocks_clear() to nvdimm_clear_badblocks_regions().
Eliminate the lock and consolidate all support routines for the new
nvdimm_account_cleared_poison() in drivers/nvdimm/bus.c. Finally, to the
opportunity to cleanup to some unnecessary casts, make the calling
convention of nvdimm_clear_badblocks_regions() clearer by replacing struct
resource with the minimal struct clear_badblocks_context, and use the
DEVICE_ATTR macro.
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Reported-by: Toshi Kani <toshi.kani@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
ND_CMD_CLEAR_ERROR command returns 'clear_err.cleared', the length
of error actually cleared, which may be smaller than its requested
'len'.
Change nvdimm_clear_poison() to call nvdimm_forget_poison() with
'clear_err.cleared' when this value is valid.
Cc: <stable@vger.kernel.org>
Fixes: e046114af5 ("libnvdimm: clear the internal poison_list when clearing badblocks")
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The following warning results from holding a lane spinlock,
preempt_disable(), or the btt map spinlock and then trying to take the
reconfig_mutex to walk the poison list and potentially add new entries.
BUG: sleeping function called from invalid context at kernel/locking/mutex.
c:747
in_atomic(): 1, irqs_disabled(): 0, pid: 17159, name: dd
[..]
Call Trace:
dump_stack+0x85/0xc8
___might_sleep+0x184/0x250
__might_sleep+0x4a/0x90
__mutex_lock+0x58/0x9b0
? nvdimm_bus_lock+0x21/0x30 [libnvdimm]
? __nvdimm_bus_badblocks_clear+0x2f/0x60 [libnvdimm]
? acpi_nfit_forget_poison+0x79/0x80 [nfit]
? _raw_spin_unlock+0x27/0x40
mutex_lock_nested+0x1b/0x20
nvdimm_bus_lock+0x21/0x30 [libnvdimm]
nvdimm_forget_poison+0x25/0x50 [libnvdimm]
nvdimm_clear_poison+0x106/0x140 [libnvdimm]
nsio_rw_bytes+0x164/0x270 [libnvdimm]
btt_write_pg+0x1de/0x3e0 [nd_btt]
? blk_queue_enter+0x30/0x290
btt_make_request+0x11a/0x310 [nd_btt]
? blk_queue_enter+0xb7/0x290
? blk_queue_enter+0x30/0x290
generic_make_request+0x118/0x3b0
A spinlock is introduced to protect the poison list. This allows us to not
having to acquire the reconfig_mutex for touching the poison list. The
add_poison() function has been broken out into two helper functions. One to
allocate the poison entry and the other to apppend the entry. This allows us
to unlock the poison_lock in non-I/O path and continue to be able to allocate
the poison entry with GFP_KERNEL. We will use GFP_NOWAIT in the I/O path in
order to satisfy being in atomic context.
Reviewed-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Providing mechanism to clear poison list via the ndctl ND_CMD_CLEAR_ERROR
call. We will update the poison list and also the badblocks at region level
if the region is in dax mode or in pmem mode and not active. In other
words we force badblocks to be cleared through write requests if the
address is currently accessed through a block device, otherwise it can
only be done via the ioctl+dsm path.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Holding the reconfig_mutex over a potential userspace fault sets up a
lockdep dependency chain between filesystem-DAX and the libnvdimm ioctl
path. Move the user access outside of the lock.
[ INFO: possible circular locking dependency detected ]
4.11.0-rc3+ #13 Tainted: G W O
-------------------------------------------------------
fallocate/16656 is trying to acquire lock:
(&nvdimm_bus->reconfig_mutex){+.+.+.}, at: [<ffffffffa00080b1>] nvdimm_bus_lock+0x21/0x30 [libnvdimm]
but task is already holding lock:
(jbd2_handle){++++..}, at: [<ffffffff813b4944>] start_this_handle+0x104/0x460
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (jbd2_handle){++++..}:
lock_acquire+0xbd/0x200
start_this_handle+0x16a/0x460
jbd2__journal_start+0xe9/0x2d0
__ext4_journal_start_sb+0x89/0x1c0
ext4_dirty_inode+0x32/0x70
__mark_inode_dirty+0x235/0x670
generic_update_time+0x87/0xd0
touch_atime+0xa9/0xd0
ext4_file_mmap+0x90/0xb0
mmap_region+0x370/0x5b0
do_mmap+0x415/0x4f0
vm_mmap_pgoff+0xd7/0x120
SyS_mmap_pgoff+0x1c5/0x290
SyS_mmap+0x22/0x30
entry_SYSCALL_64_fastpath+0x1f/0xc2
-> #1 (&mm->mmap_sem){++++++}:
lock_acquire+0xbd/0x200
__might_fault+0x70/0xa0
__nd_ioctl+0x683/0x720 [libnvdimm]
nvdimm_ioctl+0x8b/0xe0 [libnvdimm]
do_vfs_ioctl+0xa8/0x740
SyS_ioctl+0x79/0x90
do_syscall_64+0x6c/0x200
return_from_SYSCALL_64+0x0/0x7a
-> #0 (&nvdimm_bus->reconfig_mutex){+.+.+.}:
__lock_acquire+0x16b6/0x1730
lock_acquire+0xbd/0x200
__mutex_lock+0x88/0x9b0
mutex_lock_nested+0x1b/0x20
nvdimm_bus_lock+0x21/0x30 [libnvdimm]
nvdimm_forget_poison+0x25/0x50 [libnvdimm]
nvdimm_clear_poison+0x106/0x140 [libnvdimm]
pmem_do_bvec+0x1c2/0x2b0 [nd_pmem]
pmem_make_request+0xf9/0x270 [nd_pmem]
generic_make_request+0x118/0x3b0
submit_bio+0x75/0x150
Cc: <stable@vger.kernel.org>
Fixes: 62232e45f4 ("libnvdimm: control (ioctl) messages for nvdimm_bus and nvdimm devices")
Cc: Dave Jiang <dave.jiang@intel.com>
Reported-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Given ambiguities in the ACPI 6.1 definition of the "Output (Size)"
field of the ARS (Address Range Scrub) Status command, a firmware
implementation may in practice return 0, 4, or 8 to indicate that there
is no output payload to process.
The specification states "Size of Output Buffer in bytes, including this
field.". However, 'Output Buffer' is also the name of the entire
payload, and earlier in the specification it states "Max Query ARS
Status Output Buffer Size: Maximum size of buffer (including the Status
and Extended Status fields)".
Without this fix if the BIOS happens to return 0 it causes memory
corruption as evidenced by this result from the acpi_nfit_ctl() unit
test.
ars_status00000000: 00020000 00000000 ........
BUG: stack guard page was hit at ffffc90001750000 (stack is ffffc9000174c000..ffffc9000174ffff)
kernel stack overflow (page fault): 0000 [#1] SMP DEBUG_PAGEALLOC
task: ffff8803332d2ec0 task.stack: ffffc9000174c000
RIP: 0010:[<ffffffff814cfe72>] [<ffffffff814cfe72>] __memcpy+0x12/0x20
RSP: 0018:ffffc9000174f9a8 EFLAGS: 00010246
RAX: ffffc9000174fab8 RBX: 0000000000000000 RCX: 000000001fffff56
RDX: 0000000000000000 RSI: ffff8803231f5a08 RDI: ffffc90001750000
RBP: ffffc9000174fa88 R08: ffffc9000174fab0 R09: ffff8803231f54b8
R10: 0000000000000008 R11: 0000000000000001 R12: 0000000000000000
R13: 0000000000000000 R14: 0000000000000003 R15: ffff8803231f54a0
FS: 00007f3a611af640(0000) GS:ffff88033ed00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffc90001750000 CR3: 0000000325b20000 CR4: 00000000000406e0
Stack:
ffffffffa00bc60d 0000000000000008 ffffc90000000001 ffffc9000174faac
0000000000000292 ffffffffa00c24e4 ffffffffa00c2914 0000000000000000
0000000000000000 ffffffff00000003 ffff880331ae8ad0 0000000800000246
Call Trace:
[<ffffffffa00bc60d>] ? acpi_nfit_ctl+0x49d/0x750 [nfit]
[<ffffffffa01f4fe0>] nfit_test_probe+0x670/0xb1b [nfit_test]
Cc: <stable@vger.kernel.org>
Fixes: 747ffe11b4 ("libnvdimm, tools/testing/nvdimm: fix 'ars_status' output buffer sizing")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
nvdimm_clear_poison cleared the user-visible badblocks, and sent
commands to the NVDIMM to clear the areas marked as 'poison', but it
neglected to clear the same areas from the internal poison_list which is
used to marshal ARS results before sorting them by namespace. As a
result, once on-demand ARS functionality was added:
37b137f nfit, libnvdimm: allow an ARS scrub to be triggered on demand
A scrub triggered from either sysfs or an MCE was found to be adding
stale entries that had been cleared from gendisk->badblocks, but were
still present in nvdimm_bus->poison_list. Additionally, the stale entries
could be triggered into producing stale disk->badblocks by simply disabling
and re-enabling the namespace or region.
This adds the missing step of clearing poison_list entries when clearing
poison, so that it is always in sync with badblocks.
Fixes: 37b137f ("nfit, libnvdimm: allow an ARS scrub to be triggered on demand")
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Bad blocks can be injected via /sys/block/pmemN/badblocks. In a situation
where legacy pmem is being used or a pmem region created by using memmap
kernel parameter, the injected bad blocks are not cleared due to
nvdimm_clear_poison() failing from lack of ndctl function pointer. In
this case we need to just return as handled and allow the bad blocks to
be cleared rather than fail.
Reviewed-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
1/ Replace pcommit with ADR / directed-flushing:
The pcommit instruction, which has not shipped on any product, is
deprecated. Instead, the requirement is that platforms implement either
ADR, or provide one or more flush addresses per nvdimm. ADR
(Asynchronous DRAM Refresh) flushes data in posted write buffers to the
memory controller on a power-fail event. Flush addresses are defined in
ACPI 6.x as an NVDIMM Firmware Interface Table (NFIT) sub-structure:
"Flush Hint Address Structure". A flush hint is an mmio address that
when written and fenced assures that all previous posted writes
targeting a given dimm have been flushed to media.
2/ On-demand ARS (address range scrub):
Linux uses the results of the ACPI ARS commands to track bad blocks
in pmem devices. When latent errors are detected we re-scrub the media
to refresh the bad block list, userspace can also request a re-scrub at
any time.
3/ Support for the Microsoft DSM (device specific method) command format.
4/ Support for EDK2/OVMF virtual disk device memory ranges.
5/ Various fixes and cleanups across the subsystem.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=xCBG
-----END PGP SIGNATURE-----
Merge tag 'libnvdimm-for-4.8' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm
Pull libnvdimm updates from Dan Williams:
- Replace pcommit with ADR / directed-flushing.
The pcommit instruction, which has not shipped on any product, is
deprecated. Instead, the requirement is that platforms implement
either ADR, or provide one or more flush addresses per nvdimm.
ADR (Asynchronous DRAM Refresh) flushes data in posted write buffers
to the memory controller on a power-fail event.
Flush addresses are defined in ACPI 6.x as an NVDIMM Firmware
Interface Table (NFIT) sub-structure: "Flush Hint Address Structure".
A flush hint is an mmio address that when written and fenced assures
that all previous posted writes targeting a given dimm have been
flushed to media.
- On-demand ARS (address range scrub).
Linux uses the results of the ACPI ARS commands to track bad blocks
in pmem devices. When latent errors are detected we re-scrub the
media to refresh the bad block list, userspace can also request a
re-scrub at any time.
- Support for the Microsoft DSM (device specific method) command
format.
- Support for EDK2/OVMF virtual disk device memory ranges.
- Various fixes and cleanups across the subsystem.
* tag 'libnvdimm-for-4.8' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: (41 commits)
libnvdimm-btt: Delete an unnecessary check before the function call "__nd_device_register"
nfit: do an ARS scrub on hitting a latent media error
nfit: move to nfit/ sub-directory
nfit, libnvdimm: allow an ARS scrub to be triggered on demand
libnvdimm: register nvdimm_bus devices with an nd_bus driver
pmem: clarify a debug print in pmem_clear_poison
x86/insn: remove pcommit
Revert "KVM: x86: add pcommit support"
nfit, tools/testing/nvdimm/: unify shutdown paths
libnvdimm: move ->module to struct nvdimm_bus_descriptor
nfit: cleanup acpi_nfit_init calling convention
nfit: fix _FIT evaluation memory leak + use after free
tools/testing/nvdimm: add manufacturing_{date|location} dimm properties
tools/testing/nvdimm: add virtual ramdisk range
acpi, nfit: treat virtual ramdisk SPA as pmem region
pmem: kill __pmem address space
pmem: kill wmb_pmem()
libnvdimm, pmem: use nvdimm_flush() for namespace I/O writes
fs/dax: remove wmb_pmem()
libnvdimm, pmem: flush posted-write queues on shutdown
...
A recent effort to add a new nvdimm bus provider attribute highlighted a
race between interrogating nvdimm_bus->nd_desc and nvdimm_bus tear down.
The typical way to handle these races is to take the device_lock() in
the attribute method and validate that the device is still active. In
order for a device to be 'active' it needs to be associated with a
driver. So, we create the small boilerplate for a driver and register
nvdimm_bus devices on the 'nvdimm_bus_type' bus.
A result of this change is that ndbusX devices now appear under
/sys/bus/nd/devices. In fact this makes /sys/class/nd somewhat
redundant, but removing that will need to take a long deprecation period
given its use by ndctl binaries in the field.
This change naturally pulls code from drivers/nvdimm/core.c to
drivers/nvdimm/bus.c, so it is a nice code organization clean-up as
well.
Cc: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Let the provider module be explicitly passed in rather than implicitly
assumed by the module that calls nvdimm_bus_register(). This is in
preparation for unifying the nfit and nfit_test driver teardown paths.
Reviewed-by: Lee, Chun-Yi <jlee@suse.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Now that all drivers that specify a ->driverfs_dev have been converted
to device_add_disk(), the pointer can be removed from struct gendisk.
Cc: Jens Axboe <axboe@fb.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Prompted by commit 287980e49f "remove lots of IS_ERR_VALUE abuses", I
ran make coccicheck against drivers/nvdimm/ and found that:
if (IS_ERR(x))
return PTR_ERR(x);
return 0;
...can be replaced with PTR_ERR_OR_ZERO().
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The dax_pmem driver was implementing an empty ->remove() method to
satisfy the nvdimm bus driver that unconditionally calls ->remove().
Teach the core bus driver to check if ->remove() is NULL to remove that
requirement.
Reported-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Device DAX is the device-centric analogue of Filesystem DAX
(CONFIG_FS_DAX). It allows persistent memory ranges to be allocated and
mapped without need of an intervening file system. This initial
infrastructure arranges for a libnvdimm pfn-device to be represented as
a different device-type so that it can be attached to a driver other
than the pmem driver.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
