linux-stable.git/drivers/acpi, branch v4.4.220

x86/mm: split vmalloc_sync_all()

2020-04-02T17:02:31+00:00

commit 763802b53a427ed3cbd419dbba255c414fdd9e7c upstream.

Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in
__purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in
the vunmap() code-path.  While this change was necessary to maintain
correctness on x86-32-pae kernels, it also adds additional cycles for
architectures that don't need it.

Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported
severe performance regressions in micro-benchmarks because it now also
calls the x86-64 implementation of vmalloc_sync_all() on vunmap().  But
the vmalloc_sync_all() implementation on x86-64 is only needed for newly
created mappings.

To avoid the unnecessary work on x86-64 and to gain the performance
back, split up vmalloc_sync_all() into two functions:

	* vmalloc_sync_mappings(), and
	* vmalloc_sync_unmappings()

Most call-sites to vmalloc_sync_all() only care about new mappings being
synchronized.  The only exception is the new call-site added in the
above mentioned commit.

Shile Zhang directed us to a report of an 80% regression in reaim
throughput.

Fixes: 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()")
Reported-by: kernel test robot 
Reported-by: Shile Zhang 
Signed-off-by: Joerg Roedel 
Signed-off-by: Andrew Morton 
Tested-by: Borislav Petkov 
Acked-by: Rafael J. Wysocki 	[GHES]
Cc: Dave Hansen 
Cc: Andy Lutomirski 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: Ingo Molnar 
Cc: 
Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org
Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2KEPC6KGKS6J25AIDB/
Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.com
Signed-off-by: Linus Torvalds 
Signed-off-by: Greg Kroah-Hartman

ACPICA: Disassembler: create buffer fields in ACPI_PARSE_LOAD_PASS1

2020-02-28T14:38:59+00:00

[ Upstream commit 5ddbd77181dfca61b16d2e2222382ea65637f1b9 ]

ACPICA commit 29cc8dbc5463a93625bed87d7550a8bed8913bf4

create_buffer_field is a deferred op that is typically processed in
load pass 2. However, disassembly of control method contents walk the
parse tree with ACPI_PARSE_LOAD_PASS1 and AML_CREATE operators are
processed in a later walk. This is a problem when there is a control
method that has the same name as the AML_CREATE object. In this case,
any use of the name segment will be detected as a method call rather
than a reference to a buffer field. If this is detected as a method
call, it can result in a mal-formed parse tree if the control methods
have parameters.

This change in processing AML_CREATE ops earlier solves this issue by
inserting the named object in the ACPI namespace so that references
to this name would be detected as a name string rather than a method
call.

Link: https://github.com/acpica/acpica/commit/29cc8dbc
Reported-by: Elia Geretto 
Tested-by: Elia Geretto 
Signed-off-by: Bob Moore 
Signed-off-by: Erik Kaneda 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin

ACPI: PM: Avoid attaching ACPI PM domain to certain devices

2019-12-21T09:35:28+00:00

commit b9ea0bae260f6aae546db224daa6ac1bd9d94b91 upstream.

Certain ACPI-enumerated devices represented as platform devices in
Linux, like fans, require special low-level power management handling
implemented by their drivers that is not in agreement with the ACPI
PM domain behavior.  That leads to problems with managing ACPI fans
during system-wide suspend and resume.

For this reason, make acpi_dev_pm_attach() skip the affected devices
by adding a list of device IDs to avoid to it and putting the IDs of
the affected devices into that list.

Fixes: e5cc8ef31267 (ACPI / PM: Provide ACPI PM callback routines for subsystems)
Reported-by: Zhang Rui 
Tested-by: Todd Brandt 
Cc: 3.10+  # 3.10+
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Greg Kroah-Hartman

ACPI: bus: Fix NULL pointer check in acpi_bus_get_private_data()

2019-12-21T09:35:28+00:00

commit 627ead724eff33673597216f5020b72118827de4 upstream.

kmemleak reported backtrace:
    [] kmem_cache_alloc_trace+0x128/0x260
    [<6677f215>] i2c_acpi_install_space_handler+0x4b/0xe0
    [<1180f4fc>] i2c_register_adapter+0x186/0x400
    [<6083baf7>] i2c_add_adapter+0x4e/0x70
    [] intel_gmbus_setup+0x1a2/0x2c0 [i915]
    [<84cb69ae>] i915_driver_probe+0x8d8/0x13a0 [i915]
    [<81911d4b>] i915_pci_probe+0x48/0x160 [i915]
    [<4b159af1>] pci_device_probe+0xdc/0x160
    [] really_probe+0x1ee/0x450
    [] driver_probe_device+0x142/0x1b0
    [] device_driver_attach+0x49/0x50
    [] __driver_attach+0xc9/0x150
    [] bus_for_each_dev+0x56/0xa0
    [<80089bba>] driver_attach+0x19/0x20
    [] bus_add_driver+0x177/0x220
    [<7b29d8c7>] driver_register+0x56/0xf0

In i2c_acpi_remove_space_handler(), a leak occurs whenever the
"data" parameter is initialized to 0 before being passed to
acpi_bus_get_private_data().

This is because the NULL pointer check in acpi_bus_get_private_data()
(condition->if(!*data)) returns EINVAL and, in consequence, memory is
never freed in i2c_acpi_remove_space_handler().

Fix the NULL pointer check in acpi_bus_get_private_data() to follow
the analogous check in acpi_get_data_full().

Signed-off-by: Vamshi K Sthambamkadi 
[ rjw: Subject & changelog ]
Cc: All applicable 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Greg Kroah-Hartman

ACPI: OSL: only free map once in osl.c

2019-12-21T09:35:27+00:00

commit 833a426cc471b6088011b3d67f1dc4e147614647 upstream.

acpi_os_map_cleanup checks map->refcount outside of acpi_ioremap_lock
before freeing the map. This creates a race condition the can result
in the map being freed more than once.
A panic can be caused by running

for ((i=0; i<10; i++))
do
        for ((j=0; j<100000; j++))
        do
                cat /sys/firmware/acpi/tables/data/BERT >/dev/null
        done &
done

This patch makes sure that only the process that drops the reference
to 0 does the freeing.

Fixes: b7c1fadd6c2e ("ACPI: Do not use krefs under a mutex in osl.c")
Signed-off-by: Francesco Ruggeri 
Reviewed-by: Dmitry Safonov <0x7f454c46@gmail.com>
Cc: All applicable 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Greg Kroah-Hartman

ACPI / APEI: Switch estatus pool to use vmalloc memory

2019-12-05T14:27:09+00:00

[ Upstream commit 0ac234be1a9497498e57d958f4251f5257b116b4 ]

The ghes code is careful to parse and round firmware's advertised
memory requirements for CPER records, up to a maximum of 64K.
However when ghes_estatus_pool_expand() does its work, it splits
the requested size into PAGE_SIZE granules.

This means if firmware generates 5K of CPER records, and correctly
describes this in the table, __process_error() will silently fail as it
is unable to allocate more than PAGE_SIZE.

Switch the estatus pool to vmalloc() memory. On x86 vmalloc() memory
may fault and be fixed up by vmalloc_fault(). To prevent this call
vmalloc_sync_all() before an NMI handler could discover the memory.

Signed-off-by: James Morse 
Reviewed-by: Borislav Petkov 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin

ACPI / LPSS: Ignore acpi_device_fix_up_power() return value

2019-12-05T14:26:41+00:00

[ Upstream commit 1a2fa02f7489dc4d746f2a15fb77b3ce1affade8 ]

Ignore acpi_device_fix_up_power() return value. If we return an error
we end up with acpi_default_enumeration() still creating a platform-
device for the device and we end up with the device still being used
but without the special LPSS related handling which is not useful.

Specicifically ignoring the error fixes the touchscreen no longer
working after a suspend/resume on a Prowise PT301 tablet.

This tablet has a broken _PS0 method on the touchscreen's I2C controller,
causing acpi_device_fix_up_power() to fail, causing fallback to standard
platform-dev handling and specifically causing acpi_lpss_save/restore_ctx
to not run.

The I2C controllers _PS0 method does actually turn on the device, but then
does some more nonsense which fails when run during early boot trying to
use I2C opregion handling on another not-yet registered I2C controller.

Signed-off-by: Hans de Goede 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin

ACPI / SBS: Fix rare oops when removing modules

2019-11-25T14:54:46+00:00

[ Upstream commit 757c968c442397f1249bb775a7c8c03842e3e0c7 ]

There was a small race when removing the sbshc module where
smbus_alarm() had queued acpi_smbus_callback() for deferred execution
but it hadn't been run yet, so that when it did run hc had been freed
and the module unloaded, resulting in an invalid paging request.

A similar race existed when removing the sbs module with regards to
acpi_sbs_callback() (which is called from acpi_smbus_callback()).

We therefore need to ensure no callbacks are pending or executing before
the cleanups are done and the modules are removed.

Signed-off-by: Ronald TschalÃ¤r 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin

PCI/ACPI: Correct error message for ASPM disabling

2019-11-25T14:54:00+00:00

[ Upstream commit 1ad61b612b95980a4d970c52022aa01dfc0f6068 ]

If _OSC execution fails today for platforms without an _OSC entry, code is
printing a misleading message saying disabling ASPM as follows:

  acpi PNP0A03:00: _OSC failed (AE_NOT_FOUND); disabling ASPM

We need to ensure that platform supports ASPM to begin with.

Reported-by: Michael Kelley 
Signed-off-by: Sinan Kaya 
Signed-off-by: Bjorn Helgaas 
Signed-off-by: Sasha Levin

ACPI: custom_method: fix memory leaks

2019-10-05T10:27:49+00:00

[ Upstream commit 03d1571d9513369c17e6848476763ebbd10ec2cb ]

In cm_write(), 'buf' is allocated through kzalloc(). In the following
execution, if an error occurs, 'buf' is not deallocated, leading to memory
leaks. To fix this issue, free 'buf' before returning the error.

Fixes: 526b4af47f44 ("ACPI: Split out custom_method functionality into an own driver")
Signed-off-by: Wenwen Wang 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin