Pull libnvdimm fix from Dan Williams:
"This contains a regression fix for a problem that was introduced in
v4.7-rc6.
In 4.7-rc1 we introduced auto-probing for the ACPI DSM (device-
specific-method) format that the platform firmware implements for
nvdimm devices. We initially fixed a regression in probing the QEMU
DSM implementation by making acpi_check_dsm() tolerant of the way QEMU
reports the "0 DSMs supported" condition.
However, that broke HPE platforms since that tolerance caused the
driver to mistakenly match the 1-zero-byte response those platforms
give to "unknown" commands. Instead, we simply make the driver
tolerant of not finding any supported DSMs. This has been tested to
work with both QEMU and HPE platforms.
This commit has appeared in a -next release with no reported issues"
* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
nfit: make DIMM DSMs optional
Commit 4995734e97 "acpi, nfit: fix acpi_check_dsm() vs zero functions
implemented" attempted to fix a QEMU regression by supporting its usage
of a zero-mask as a valid response to a DSM-family probe request.
However, this behavior breaks HP platforms that return a zero-mask by
default causing the probe to misidentify the DSM-family.
Instead, the QEMU regression can be fixed by simply not requiring the DSM
family to be identified.
This effectively reverts commit 4995734e97, and removes the DSM
requirement from the init path.
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Cc: Linda Knippers <linda.knippers@hpe.com>
Fixes: 4995734e97 ("acpi, nfit: fix acpi_check_dsm() vs zero functions implemented")
Reported-by: Jerry Hoemann <jerry.hoemann@hpe.com>
Tested-by: Jerry Hoemann <jerry.hoemann@hpe.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Revert commit 2f38b1b16d (ACPICA: Namespace: Fix deadlock triggered by
MLC support in dynamic table loading) that attempted to fix a deadlock
issue introduced by a previous commit, but it led to a lock ordering
inconsistency that caused further problems to appear.
Fixes: 2f38b1b16d (ACPICA: Namespace: Fix deadlock triggered by MLC support in dynamic table loading)
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There is an order issue in ec_remove_handlers() that acpi_ec_stop()
is called before removing the operation region handler. That is
incorrect, because the operation region handler removal triggers
_REG(DISCONNECT) which may result in new EC transactions to carry
out.
That existing issue has been triggered by the following commit:
Commit: dcf15cbded
Subject: ACPI / EC: Fix a boot EC regresion by restoring boot EC
which changed the driver to call ec_remove_handlers() after invoking
_REG(CONNECT), so the issue has become visible.
Fixes: dcf15cbded (ACPI / EC: Fix a boot EC regresion by restoring boot EC)
Link: https://bugzilla.kernel.org/show_bug.cgi?id=102421
Reported-and-tested-by: Wolfram Sang <wsa@the-dreams.de>
Reported-by: Nicholas <nkudriavtsev@gmail.com>
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
[ rjw: Changelog ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The FIFO unlocking mechanism in acpi_dbg has been broken by the
following commit:
Commit: 287980e49f
Subject: remove lots of IS_ERR_VALUE abuses
It converted !IS_ERR_VALUE(ret) into !ret which was not entirely
correct. Fix the regression by taking ret > 0 into account too as
appropriate.
Fixes: 287980e49f (remove lots of IS_ERR_VALUE abuses)
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
[ rjw: Simplifications, changelog & subject massage ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There is a lock order issue in acpi_load_tables(). The namespace lock
is held before holding the interpreter lock.
With ACPI_MUTEX_DEBUG enabled in the kernel, this is printed to the
log during boot:
[ 0.885699] ACPI Error: Invalid acquire order: Thread 405884224 owns [ACPI_MTX_Namespace], wants [ACPI_MTX_Interpreter] (20160422/utmutex-263)
[ 0.885881] ACPI Error: Could not acquire AML Interpreter mutex (20160422/exutils-95)
[ 0.893846] ACPI Error: Mutex [0x0] is not acquired, cannot release (20160422/utmutex-326)
[ 0.894019] ACPI Error: Could not release AML Interpreter mutex (20160422/exutils-133)
The issue has been introduced by the following commit:
Commit: 2f38b1b16d
ACPICA Commit: bfe03ffcde8ed56a7eae38ea0b188aeb12f9c52e
Subject: ACPICA: Namespace: Fix a regression that MLC support triggers
dead lock in dynamic table loading
Which fixed a deadlock issue for acpi_ns_load_table() in
acpi_ex_add_table() but didn't take care of the lock order in
acpi_ns_load_table() correctly.
Originally (before the above commit), ACPICA used the
namespace/interpreter locks in the following 2 key code
paths:
1. Table loading:
acpi_ns_load_table
L(Namespace)
acpi_ns_parse_table
acpi_ns_one_complete_parse
U(Namespace)
2. Object evaluation:
acpi_ns_evaluate
L(Interpreter)
acpi_ps_execute_method
U(Interpreter)
acpi_ns_load_table
L(Namespace)
U(Namespace)
acpi_ev_initialize_region
L(Namespace)
U(Namespace)
address_space.setup
L(Namespace)
U(Namespace)
address_space.handler
L(Namespace)
U(Namespace)
acpi_os_wait_semaphore
acpi_os_acquire_mutex
acpi_os_sleep
L(Interpreter)
U(Interpreter)
During runtime, while acpi_ns_evaluate is called, the lock order is
always Interpreter -> Namespace.
In turn, the problematic commit acquires the locks in the following
order:
3. Table loading:
acpi_ns_load_table
L(Namespace)
acpi_ns_parse_table
L(Interpreter)
acpi_ns_one_complete_parse
U(Interpreter)
U(Namespace)
To fix the lock order issue, move the interpreter lock to
acpi_ns_load_table() to ensure the lock order correctness:
4. Table loading:
acpi_ns_load_table
L(Interpreter)
L(Namespace)
acpi_ns_parse_table
acpi_ns_one_complete_parse
U(Namespace)
U(Interpreter)
However, this doesn't fix the current design issues related to the
namespace lock. For example, we can notice that in acpi_ns_evaluate(),
outside of acpi_ns_load_table(), the namespace objects may be created
by the named object creation control methods. And the creation of
the method-owned namespace objects are not locked by the namespace
lock. This patch doesn't try to fix such kind of existing issues.
Fixes: 2f38b1b16d (ACPICA: Namespace: Fix a regression that MLC support triggers dead lock in dynamic table loading)
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since commit 103544d869 (ACPI,PCI,IRQ: reduce resource requirements)
the penalty values are calculated on the fly rather than at boot time.
This works fine for PCI interrupts but not so well for ISA interrupts.
The information on whether or not an ISA interrupt is in use is not
available to the pci_link.c code directly. That information is
obtained from the outside via acpi_penalize_isa_irq(). [If its
"active" argument is true, then the IRQ is in use by ISA.]
Since the current code relies on PCI Link objects for determination
of penalties, we are factoring in the PCI penalty twice after
acpi_penalize_isa_irq() function is called.
To avoid that, limit the newly added functionality to just PCI
interrupts so that old behavior is still maintained.
Fixes: 103544d869 (ACPI,PCI,IRQ: reduce resource requirements)
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Tested-by: Wim Osterholt <wim@djo.tudelft.nl>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Trying to make the ISA and PCI init functionality common turned out
to be a bad idea, because the ISA path depends on external
functionality.
Restore the previous behavior and limit the refactoring to PCI
interrupts only.
Fixes: 1fcb6a813c "ACPI,PCI,IRQ: remove redundant code in acpi_irq_penalty_init()"
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Tested-by: Wim Osterholt <wim@djo.tudelft.nl>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The change introduced in commit 103544d869 (ACPI,PCI,IRQ: reduce
resource requirements) omitted the initially applied PCI_POSSIBLE
penalty when the IRQ is active.
Incorrect calculation of the penalty leads the ACPI code to assigning
a wrong interrupt number to a PCI INTx interrupt.
This would not be as bad as it sounds in theory. It would just cause
the interrupts to be shared and result in performance penalty.
However, some drivers (like the parallel port driver) don't like
interrupt sharing and in the above case they will causes all of
the PCI drivers wanting to share the interrupt to be unable to
request it.
The issue has not been caught in testing because the behavior is
platform-specific and depends on the peripherals ending up sharing
the IRQ and their drivers.
Before the above commit the code would add the PCI_POSSIBLE value
divided by the number of possible IRQ users to the IRQ penalty
during initialization.
Later in that code path, if the IRQ is chosen as the active IRQ or
if it is used by ISA; additional penalties are added.
Fixes: 103544d869 (ACPI,PCI,IRQ: reduce resource requirements)
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Tested-by: Wim Osterholt <wim@djo.tudelft.nl>
[ rjw: Changelog ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Pull ACPI fix from Rafael Wysocki:
"Fix an expression in the ACPI PCI IRQ management code added by a
recent commit that overlooked missing parens in it, so the result of
the computation is incorrect in some cases (Sinan Kaya)"
* tag 'acpi-4.7-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
ACPI,PCI,IRQ: correct operator precedence
Pull libnvdimm fixes from Dan Williams:
"1/ Two regression fixes since v4.6: one for the byte order of a sysfs
attribute (bz121161) and another for QEMU 2.6's NVDIMM _DSM (ACPI
Device Specific Method) implementation that gets tripped up by new
auto-probing behavior in the NFIT driver.
2/ A fix tagged for -stable that stops the kernel from
clobbering/ignoring changes to the configuration of a 'pfn'
instance ("struct page" driver). For example changing the
alignment from 2M to 1G may silently revert to 2M if that value is
currently stored on media.
3/ A fix from Eric for an xfstests failure in dax. It is not
currently tagged for -stable since it requires an 8-exabyte file
system to trigger, and there appear to be no user visible side
effects"
* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
nfit: fix format interface code byte order
dax: fix offset overflow in dax_io
acpi, nfit: fix acpi_check_dsm() vs zero functions implemented
libnvdimm, pfn, dax: fix initialization vs autodetect for mode + alignment
The omitted parenthesis prevents the addition operation when
acpi_penalize_isa_irq function is called.
Fixes: 103544d869 (ACPI,PCI,IRQ: reduce resource requirements)
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Per JEDEC Annex L Release 3 the SPD data is:
Bits 9~5 00 000 = Function Undefined
00 001 = Byte addressable energy backed
00 010 = Block addressed
00 011 = Byte addressable, no energy backed
All other codes reserved
Bits 4~0 0 0000 = Proprietary interface
0 0001 = Standard interface 1
All other codes reserved; see Definitions of Functions
...and per the ACPI 6.1 spec:
byte0: Bits 4~0 (0 or 1)
byte1: Bits 9~5 (1, 2, or 3)
...so a format interface code displayed as 0x301 should be stored in the
nfit as (0x1, 0x3), little-endian.
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Cc: Robert Moore <robert.moore@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=121161
Fixes: 30ec5fd464 ("nfit: fix format interface code byte order per ACPI6.1")
Fixes: 5ad9a7fde0 ("acpi/nfit: Update nfit driver to comply with ACPI 6.1")
Reported-by: Kristin Jacque <kristin.jacque@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
QEMU 2.6 implements nascent support for nvdimm DSMs. Depending on
configuration it may only implement the function0 dsm to indicate that
no other DSMs are available. Commit 31eca76ba2 "nfit, libnvdimm:
limited/whitelisted dimm command marshaling mechanism" breaks QEMU, but
QEMU is spec compliant. Per the spec the way to indicate that no
functions are supported is:
If Function Index is zero, the return is a buffer containing one bit
for each function index, starting with zero. Bit 0 indicates whether
there is support for any functions other than function 0 for the
specified UUID and Revision ID. If set to zero, no functions are
supported (other than function zero) for the specified UUID and
Revision ID.
Update the nfit driver to determine the family (interface UUID) without
requiring the implementation to define any other functions, i.e.
short-circuit acpi_check_dsm() to succeed per the spec. The nfit driver
appears to be the only user passing funcs==0 to acpi_check_dsm(), so
this behavior change of the common routine should be limited to the
probing done by the nfit driver.
Cc: Len Brown <lenb@kernel.org>
Cc: Jerry Hoemann <jerry.hoemann@hpe.com>
Acked-by: "Rafael J. Wysocki" <rafael@kernel.org>
Fixes: 31eca76ba2 ("nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism")
Reported-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Tested-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The new module-level code (MLC) approach invokes MLC on the per-table
basis, but the dynamic loading support of this is incorrect because
of the lock order:
acpi_ns_evaluate
acpi_ex_enter_intperter
acpi_ns_load_table (triggered by Load opcode)
acpi_ns_exec_module_code_list
acpi_ex_enter_intperter
The regression is introduced by the following commit:
Commit: 2785ce8d0d
ACPICA Commit: 071eff738c59eda1792ac24b3b688b61691d7e7c
Subject: ACPICA: Add per-table execution of module-level code
This patch fixes this regression by unlocking the interpreter lock
before invoking MLC. However, the unlocking is done to the
acpi_ns_load_table(), in which the interpreter lock should be locked
by acpi_ns_parse_table() but it wasn't.
Fixes: 2785ce8d0d (ACPICA: Add per-table execution of module-level code)
Reported-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
Cc: 4.5+ <stable@vger.kernel.org> # 4.5+
[ rjw : Subject ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Revert commit 66b1ed5aa8 "ACPICA: ACPI 2.0, Hardware: Add
access_width/bit_offset support for acpi_hw_write()" that is reported
to break suspend-to-RAM (ACPI S3) on one system.
The root cause of the failure is a wrong access width value for one of
the involved registers provided by the ACPI tables, but before commit
66b1ed5aa8 that value was not taken into account at all and things
worked.
Fixes: 66b1ed5aa8 "ACPICA: ACPI 2.0, Hardware: Add access_width/bit_offset support for acpi_hw_write()"
Reported-by: Andrey Skvortsov <andrej.skvortzov@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
According to the Windows probing result, during the table loading, the EC
device described in the ECDT should be used. And the ECDT EC is also
effective during the period the namespace objects are initialized (we can
see a separate process executing _STA/_INI on Windows before executing
other device specific control methods, for example, EC._REG). During the
device enumration, the EC device described in the DSDT should be used. But
there are differences between Linux and Windows around the device probing
order. Thus in Linux, we should enable the DSDT EC as early as possible
before enumerating devices in order not to trigger issues related to the
device enumeration order differences.
This patch thus converts acpi_boot_ec_enable() into acpi_ec_dsdt_probe() to
fix the gap. This also fixes a user reported regression triggered after we
switched the "table loading"/"ECDT support" to be ACPI spec 2.0 compliant.
Fixes: 59f0aa9480 (ACPI 2.0 / ECDT: Remove early namespace reference from EC)
Link: https://bugzilla.kernel.org/show_bug.cgi?id=119261
Reported-and-tested-by: Gabriele Mazzotta <gabriele.mzt@gmail.com>
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Roland Dreier reports that one of his systems cannot boot because of
the changes made by commit ac212b6980 (ACPI / processor: Use common
hotplug infrastructure).
The problematic part of it is the request_region() call in
acpi_processor_get_info() that used to run at module init time before
the above commit and now it runs much earlier. Unfortunately, the
region(s) reserved by it fall into a range the PCI subsystem attempts
to reserve for AHCI IO BARs. As a result, the PCI reservation fails
and AHCI doesn't work, while previously the PCI reservation would
be made before acpi_processor_get_info() and it would succeed.
That request_region() call, however, was overlooked by commit
ac212b6980, as it is not necessary for the enumeration of the
processors. It only is needed when the ACPI processor driver
actually attempts to handle them which doesn't happen before
loading the ACPI processor driver module. Therefore that call
should have been moved from acpi_processor_get_info() into that
module.
Address the problem by moving the request_region() call in question
out of acpi_processor_get_info() and use the observation that the
region reserved by it is only needed if the FADT-based CPU
throttling method is going to be used, which means that it should
be sufficient to invoke it from acpi_processor_get_throttling_fadt().
Fixes: ac212b6980 (ACPI / processor: Use common hotplug infrastructure)
Reported-by: Roland Dreier <roland@purestorage.com>
Tested-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Linus Torvalds
Dan Williams
Rafael J. Wysocki
Lv Zheng
Sinan Kaya