| Age | Commit message (Collapse) | Author |
|---|
|
Move the acpi_processor_ffh_lpi_probe() call from
acpi_processor_setup_cpuidle_dev(), where its return value is
ignored, to acpi_processor_get_power_info(), which can take the
return value of that function into account in a meaningful way
and generally is a more suitable place for calling it.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
[ rjw: Message adjustment, subject and changelog edits ]
Link: https://patch.msgid.link/20260120112258.1595164-4-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The callers of acpi_processor_setup_cpuidle_dev() ignore its return
value, so convert it to a void function.
No intentional functional impact.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
[ rjw: Subject and changelog edits ]
Link: https://patch.msgid.link/20260120112258.1595164-3-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Since all callers of acpi_processor_setup_cpuidle_states() ignore its
return value and it simply passes the acpi_processor_setup_lpi_states()
return value to its callers, convert both of these functions to void.
No intentional functional impact.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
[ rjw: Subject and changelog edits ]
Link: https://patch.msgid.link/20260120112258.1595164-2-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
To probe an IWB in an ACPI based system it is required:
- to implement the IORT functions handling the IWB IORT node and create
functions to retrieve IWB firmware information
- to augment the driver to match the DSDT ACPI "ARMH0003" device and
retrieve the IWB wire and trigger mask from the GSI interrupt descriptor
in the IWB msi_domain_ops.msi_translate() function
Make the required driver changes to enable IWB probing in ACPI systems.
The GICv5 GSI format requires special handling for IWB routed IRQs.
Add IWB GSI detection to the top level driver gic_v5_get_gsi_domain_id()
function so that the correct IRQ domain for a GSI can be detected by
parsing the GSI and check whether it is an IWB-backed IRQ or not.
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Acked-by: Thomas Gleixner <tglx@kernel.org>
Link: https://patch.msgid.link/20260115-gicv5-host-acpi-v3-6-c13a9a150388@kernel.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
On ACPI ARM64 systems the GICv5 ITS configuration and translate frames
are described in the MADT table.
Refactor the current GICv5 ITS driver code to share common functions
between ACPI and OF and implement ACPI probing in the GICv5 ITS driver.
Add iort_msi_xlate() to map a device ID and retrieve an MSI controller
fwnode node for ACPI systems and update pci_msi_map_rid_ctlr_node() to
use it in its ACPI code path.
Add the required functions to IORT code for deviceID retrieval and IRQ
domain registration and look-up so that the GICv5 ITS driver in an
ACPI based system can be successfully probed.
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Acked-by: Thomas Gleixner <tglx@kernel.org>
Link: https://patch.msgid.link/20260115-gicv5-host-acpi-v3-5-c13a9a150388@kernel.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
On ARM64 ACPI systems GICv5 IRSes are described in MADT sub-entries.
Add the required plumbing to parse MADT IRS firmware table entries and
probe the IRS components in ACPI.
Augment the irqdomain_ops.translate() for PPI and SPI IRQs in order to
provide support for their ACPI based firmware translation.
Implement an irqchip ACPI based callback to initialize the global GSI
domain upon an MADT IRS detection.
The IRQCHIP_ACPI_DECLARE() entry in the top level GICv5 driver is only used
to trigger the IRS probing (ie the global GSI domain is initialized once on
the first call on multi-IRS systems); IRS probing takes place by calling
acpi_table_parse_madt() in the IRS sub-driver, that probes all IRSes
in sequence.
Add a new ACPI interrupt model so that it can be detected at runtime and
distinguished from previous GIC architecture models.
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Acked-by: Thomas Gleixner <tglx@kernel.org>
Link: https://patch.msgid.link/20260115-gicv5-host-acpi-v3-4-c13a9a150388@kernel.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm
Pull CPUFreq Arm updates for 7.0 from Viresh Kumar:
"- Update cpufreq-dt-platdev list for tegra, qcom, TI (Aaron Kling,
Dhruva Gole, and Konrad Dybcio).
- Minor improvements to the cpufreq / cpumask rust implementation
(Alexandre Courbot, Alice Ryhl, Tamir Duberstein, and Yilin Chen).
- Add support for AM62L3 SoC to ti-cpufreq driver (Dhruva Gole).
- Update FIE arch_freq_scale in ticks for non-PCC regs (Jie Zhan).
- Other minor cleanups / improvements (Felix Gu, Juan Martinez, Luca
Weiss, and Sergey Shtylyov)."
* tag 'cpufreq-arm-updates-7.0-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm:
cpufreq: scmi: Fix device_node reference leak in scmi_cpu_domain_id()
cpufreq: ti-cpufreq: add support for AM62L3 SoC
cpufreq: dt-platdev: Add ti,am62l3 to blocklist
cpufreq/amd-pstate: Add comment explaining nominal_perf usage for performance policy
cpufreq: scmi: correct SCMI explanation
cpufreq: dt-platdev: Block the driver from probing on more QC platforms
rust: cpumask: rename methods of Cpumask for clarity and consistency
cpufreq: CPPC: Update FIE arch_freq_scale in ticks for non-PCC regs
cpufreq: CPPC: Factor out cppc_fie_kworker_init()
ACPI: CPPC: Factor out and export per-cpu cppc_perf_ctrs_in_pcc_cpu()
rust: cpufreq: replace `kernel::c_str!` with C-Strings
cpufreq: Add Tegra186 and Tegra194 to cpufreq-dt-platdev blocklist
dt-bindings: cpufreq: qcom-hw: document Milos CPUFREQ Hardware
rust: cpufreq: add __rust_helper to helpers
rust: cpufreq: always inline functions using build_assert with arguments
|
|
Factor out cppc_perf_ctrs_in_pcc_cpu() for checking whether per-cpu CPC
regs are defined in PCC channels, and export it out for further use.
Reviewed-by: Lifeng Zheng <zhenglifeng1@huawei.com>
Reviewed-by: Pierre Gondois <pierre.gondois@arm.com>
Signed-off-by: Jie Zhan <zhanjie9@hisilicon.com>
Acked-by: Rafael J. Wysocki (Intel) <rafael@kernel.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
Now we have a helper so there's no need to open-code.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/20260121085105.2282380-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Now we have a helper so there's no need to open-code.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://patch.msgid.link/20260121084654.2227037-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The debug messages added by commit f7599be2bb76 ("ACPI: PM: postpone
bringing devices to D0 unless we need them") in acpi_subsys_resume_early()
and acpi_subsys_resume() are not quite accurate because what is postponed
is not just a transition to D0, but also an adjustment of the device's
wakeup setting (which may involve disabling a GPE among other things).
Moreover, these messages don't even mention ACPI.
Rephrase them and adjust the style to match other messages in device_pm.c.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Link: https://patch.msgid.link/5969819.DvuYhMxLoT@rafael.j.wysocki
|
|
The rev_id is always 0 for AMD since commit
e32d546483a2 ("ACPI: x86: Drop quirk for HP Elitebook"),
so this condition will never be true.
Remove the dead code.
Signed-off-by: Gergo Koteles <soyer@irl.hu>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Link: https://patch.msgid.link/20260109001619.37532-1-soyer@irl.hu
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
|
|
A processor voltage domain is an identifier that specifies the voltage
plane associated with a given group of processors.
Refer to section 6.2.10. _VDM (Voltage Domain) of ACPI 6.6 specification
for more information.
Link: https://github.com/acpica/acpica/commit/d0dbb157646d
Signed-off-by: Pawel Chmielewski <pawel.chmielewski@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/1921526.atdPhlSkOF@rafael.j.wysocki
|
|
Some asltests test the behavior of the Fatal() opcode and thus require
that said opcode does not return an error when called.
Introduce a compile-time option called ACPI_CONTINUE_ON_FATAL to
instruct the executor to continue the execution of AML bytecode when
encountering a Fatal() opcode. Also update the asltest to use this
new option.
Fixes: ("Abort AML bytecode execution when executing AML_FATAL_OP")
Link: https://github.com/acpica/acpica/commit/428b3410c490
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/2052065.usQuhbGJ8B@rafael.j.wysocki
|
|
The ACPI specification states that when executing AML_FATAL_OP,
the OS should log the fatal error event and shutdown in a timely
fashion.
Windows complies with this requirement by immediatly entering a
Bso_d, effectively aborting the execution of the AML bytecode in
question.
ACPICA however might continue with the AML bytecode execution
should acpi_os_signal() simply return AE_OK. This will cause issues
because ACPI BIOS implementations might assume that the Fatal()
operator does not return.
Fix this by aborting the AML bytecode execution in such a case
by returning AE_ERROR. Also turn struct acpi_signal_fatal_info into a
local variable because of its small size (12 bytes) and to ensure
that acpi_os_signal() always receives valid information about the
fatal ACPI BIOS error.
Link: https://github.com/acpica/acpica/commit/d516c7758ba6
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/3325491.5fSG56mABF@rafael.j.wysocki
|
|
As per [1].
Link: https://learn.microsoft.com/en-us/windows-hardware/drivers/display/automatic-display-switch [1]
Link: https://github.com/acpica/acpica/commit/28b644211ff2
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/10790566.nUPlyArG6x@rafael.j.wysocki
|
|
Cover a missed execution path with a new check.
Fixes: 0acf24ad7e10 ("ACPICA: Add support for PCC Opregion special context data")
Link: https://github.com/acpica/acpica/commit/f421dd9dd897
Signed-off-by: Alexey Simakov <bigalex934@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/3030574.e9J7NaK4W3@rafael.j.wysocki
|
|
acpi_ut_create_internal_object() may allocate memory from a slab cache
via kmem_cache_zalloc(), but the code currently frees it with ACPI_FREE(),
which calls kfree().
This mismatch prevents the object from being released properly and may
lead to memory leaks or other issues.
Fix this by replacing ACPI_FREE() with acpi_ut_delete_object_desc(),
which matches the allocation method used for internal objects.
Link: https://github.com/acpica/acpica/commit/a1c55dfea194
Signed-off-by: Zilin Guan <zilin@seu.edu.cn>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/4710853.LvFx2qVVIh@rafael.j.wysocki
|
|
Use LIST_HEAD() for initializing an on-stack list head in two places and
remove an empty code line added by mistake.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Link: https://patch.msgid.link/12825056.O9o76ZdvQC@rafael.j.wysocki
|
|
After starting to use platform devices for representing buttons
enumerated via ACPI, acpi_mark_gpe_for_wake() is insufficient for
preparing their GPEs to wake up the system from sleep because it
does not change the "dispatch type" of the given GPE to
ACPI_GPE_DISPATCH_NOTIFY. Subsequently, this causes acpi_enable_gpe()
in __acpi_device_wakeup_enable() to fail and system suspend transitions
to be aborted.
Address this by updating acpi_wakeup_gpe_init() to use
acpi_setup_gpe_for_wake() for buttons like for any other devices.
This allows acpi_setup_gpe_for_wake() to be simplified further because
buttons are not a special case in it any more, so do that as well.
Fixes: 52d864019636 ("ACPI: button: Convert the driver to a platform one")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Xi Pardee <xi.pardee@linux.intel.com>
Link: https://patch.msgid.link/2259694.irdbgypaU6@rafael.j.wysocki
|
|
It is pointless to attach the generic ACPI PM domain to devices whose
ACPI companions don't support ACPI power management and don't have a
wakeup GPE, so update acpi_dev_pm_attach() to skip such devices.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/5050298.GXAFRqVoOG@rafael.j.wysocki
|
|
While binding drivers directly to struct acpi_device objects allows
basic functionality to be provided, at least in the majority of cases,
there are some problems with it, related to general consistency, sysfs
layout, power management operation ordering, and code cleanliness.
Overall, it is better to bind drivers to platform devices than to their
ACPI companions, so convert the ACPI video driver to a platform one.
While this is not expected to alter functionality, it changes sysfs
layout and so it will be visible to user space.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Armin Wolf <W_Armin@gmx.de>
Link: https://patch.msgid.link/1957556.tdWV9SEqCh@rafael.j.wysocki
|
|
Adjust acpi_video_bus_notify() to cast its "data" argument to a struct
acpi_video_bus pointer instead of a struct acpi_device one, which allows
the use of acpi_driver_data() to be limited and will facilitate
subsequent changes.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Armin Wolf <W_Armin@gmx.de>
Link: https://patch.msgid.link/2409089.ElGaqSPkdT@rafael.j.wysocki
|
|
ACPI device objects associated with backlight interfaces are special
because they are ACPI companions of PCI devices (GPUs), but the
interfaces exposed by them resemble platform device one.
Currently, the ACPI video driver binds to them with the help of a
special "synthetic" device ID regardless of the pairing with the PCI
devices, but since it is generally better to use platform drivers for
handling such interfaces, the plan is to convert that drviver into a
platform one.
However, for this purpose, platform devices corresponding to the
ACPI backlight device objects need to be registered, so update
acpi_bus_attach() to apply the default ACPI enumeration to them
and modify acpi_create_platform_device() to avoid bailing out early
if a "physical" device is already attached to a backlight ACPI device
object.
In addition, update acpi_companion_match() to return a valid struct
acpi_device pointer if the ACPI companion of the given device is a
backlight ACPI device object, which will facilitate driver matching
for platform devices corresponding to those objects.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Armin Wolf <W_Armin@gmx.de>
Link: https://patch.msgid.link/5081593.31r3eYUQgx@rafael.j.wysocki
|
|
According to ACPI spec, entry method in LPI sub-package must be a
buffer or an integer.
The driver will disable the state whose the entry method is invalid
by zeroing flags in struct acpi_lpi_state.
The entry method is very key in cpuidle. A debug log is very useful
for developers.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
[ rjw: Subject and changelog edits, changed "illegal" to "invalid" ]
Link: https://patch.msgid.link/20251125064702.3666149-1-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
In acpi_processor_errata_piix4(), the pointer dev is first assigned an IDE
device and then reassigned an ISA device:
dev = pci_get_subsys(..., PCI_DEVICE_ID_INTEL_82371AB, ...);
dev = pci_get_subsys(..., PCI_DEVICE_ID_INTEL_82371AB_0, ...);
If the first lookup succeeds but the second fails, dev becomes NULL. This
leads to a potential null-pointer dereference when dev_dbg() is called:
if (errata.piix4.bmisx)
dev_dbg(&dev->dev, ...);
To prevent this, use two temporary pointers and retrieve each device
independently, avoiding overwriting dev with a possible NULL value.
Signed-off-by: Tuo Li <islituo@gmail.com>
[ rjw: Subject adjustment, added an empty code line ]
Link: https://patch.msgid.link/20260111163214.202262-1-islituo@gmail.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Like the JWIPC JVC9100 has its serial IRQ (10 and 11) described
as ActiveLow in the DSDT, which the kernel overrides to EdgeHigh which
breaks the serial.
irq 10, level, active-low, shared, skip-override
irq 11, level, active-low, shared, skip-override
Add the JVC9100 to the irq1_level_low_skip_override[] quirk table to fix
this.
Signed-off-by: Ai Chao <aichao@kylinos.cn>
Link: https://patch.msgid.link/20260113072719.4154485-1-aichao@kylinos.cn
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
During pre-production development, drivers may provide both ACPI and OF
match tables while a formal ACPI HID for the device is not yet
allocated. Such devices are enumerated via PRP0001. In this case,
acpi_device_get_match_data() consults only the driver’s ACPI match table
and returns NULL, even though the device was successfully matched via
PRP0001.
This behavior also risks breaking existing PRP0001 setups if a driver
later gains an ACPI HID, as the presence of an ACPI match table changes
the match-data lookup path.
Make acpi_device_get_match_data() use the same precedence as driver
matching by using __acpi_match_device(). Return match data from the
acpi_id or of_id that was actually matched.
Remove now-unused acpi_of_device_get_match_data().
Signed-off-by: Kartik Rajput <kkartik@nvidia.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Link: https://patch.msgid.link/20260114082306.48119-1-kkartik@nvidia.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
When Firmware First is enabled, BIOS handles errors first and then it
makes them available to the kernel via the Common Platform Error Record
(CPER) sections (UEFI 2.11 Appendix N.2.13). Linux parses the CPER
sections via one of two similar paths, either ELOG or GHES. The errors
managed by ELOG are signaled to the BIOS by the I/O Machine Check
Architecture (I/O MCA).
Currently, ELOG and GHES show some inconsistencies in how they report to
userspace via trace events.
Therefore, make the two mentioned paths act similarly by tracing the CPER
CXL Protocol Error Section.
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com>
Link: https://patch.msgid.link/20260114101543.85926-6-fabio.m.de.francesco@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Make a helper out of cxl_cper_post_prot_err() that checks the CXL agent
type and copy the CPER CXL protocol errors information to a work data
structure.
Export the new symbol for reuse by ELOG.
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com>
[ rjw: Subject tweak ]
Link: https://patch.msgid.link/20260114101543.85926-5-fabio.m.de.francesco@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Move the CPER CXL protocol errors validity check out of
cxl_cper_post_prot_err() to new cxl_cper_sec_prot_err_valid() and limit
the serial number check only to CXL agents that are CXL devices (UEFI
v2.10, Appendix N.2.13).
Export the new symbol for reuse by ELOG.
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com>
[ rjw: Subject tweak ]
Link: https://patch.msgid.link/20260114101543.85926-4-fabio.m.de.francesco@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
I/O Machine Check Architecture events may signal failing PCIe components
or links. The AER event contains details on what was happening on the wire
when the error was signaled.
Trace the CPER PCIe Error section (UEFI v2.11, Appendix N.2.7) reported
by the I/O MCA.
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com>
Link: https://patch.msgid.link/20260114101543.85926-3-fabio.m.de.francesco@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
ghes_do_proc() has a catch-all for unknown or unhandled CPER formats
(UEFI v2.11 Appendix N 2.3), extlog_print() does not. This gap was
noticed by a RAS test that injected CXL protocol errors which were
notified to extlog_print() via the IOMCA (I/O Machine Check
Architecture) mechanism. Bring parity to the extlog_print() path by
including a similar log_non_standard_event().
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Reviewed-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Reviewed-by: Shuai Xue <xueshuai@linux.alibaba.com>
Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com>
Link: https://patch.msgid.link/20260114101543.85926-2-fabio.m.de.francesco@linux.intel.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Performance testing on ARMv8 systems shows significant overhead in error
status handling in SEA error handling.
- ghes_peek_estatus(): 8,138.3 ns (21,160 cycles).
- ghes_clear_estatus(): 2,038.3 ns (5,300 cycles).
Apply the same optimization used in ghes_notify_nmi() to
ghes_notify_sea() by checking for active errors before processing,
Tested-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Shuai Xue <xueshuai@linux.alibaba.com>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
Link: https://patch.msgid.link/20260112032239.30023-4-xueshuai@linux.alibaba.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Refactors the GHES driver by extracting common functionality into
reusable helper functions:
1. ghes_has_active_errors() - Checks if any error sources in a given list
have active errors
2. ghes_map_error_status() - Maps error status address to virtual address
3. ghes_unmap_error_status() - Unmaps error status virtual address
4. Use `guard(rcu)()` instead of explicit `rcu_read_lock()`/`rcu_read_unlock()`.
These helpers eliminate code duplication in the NMI path and prepare for
similar usage in the SEA path in a subsequent patch.
No functional change intended.
Tested-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Shuai Xue <xueshuai@linux.alibaba.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
Link: https://patch.msgid.link/20260112032239.30023-3-xueshuai@linux.alibaba.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
ghes_notify_nmi() is called for every NMI and must check whether the NMI was
generated because an error was signalled by platform firmware.
This check is very expensive as for each registered GHES NMI source it reads
from the acpi generic address attached to this error source to get the physical
address of the acpi_hest_generic_status block. It then checks the "block_status"
to see if an error was logged.
The ACPI/APEI code must create virtual mappings for each of those physical
addresses, and tear them down afterwards. On an Icelake system this takes around
15,000 TSC cycles. Enough to disturb efforts to profile system performance.
If that were not bad enough, there are some atomic accesses in the code path
that will cause cache line bounces between CPUs. A problem that gets worse as
the core count increases.
But BIOS changes neither the acpi generic address nor the physical address of
the acpi_hest_generic_status block. So this walk can be done once when the NMI is
registered to save the virtual address (unmapping if the NMI is ever unregistered).
The "block_status" can be checked directly in the NMI handler. This can be done
without any atomic accesses.
Resulting time to check that there is not an error record is around 900 cycles.
Reported-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tested-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Shuai Xue <xueshuai@linux.alibaba.com>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
Link: https://patch.msgid.link/20260112032239.30023-2-xueshuai@linux.alibaba.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The logic at ghes_new() prevents allocating too large records, by
checking if they're bigger than GHES_ESTATUS_MAX_SIZE (currently, 64KB).
Yet, the allocation is done with the actual number of pages from the
CPER bios table location, which can be smaller.
Yet, a bad firmware could send data with a different size, which might
be bigger than the allocated memory, causing an OOPS:
Unable to handle kernel paging request at virtual address fff00000f9b40000
Mem abort info:
ESR = 0x0000000096000007
EC = 0x25: DABT (current EL), IL = 32 bits
SET = 0, FnV = 0
EA = 0, S1PTW = 0
FSC = 0x07: level 3 translation fault
Data abort info:
ISV = 0, ISS = 0x00000007, ISS2 = 0x00000000
CM = 0, WnR = 0, TnD = 0, TagAccess = 0
GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
swapper pgtable: 4k pages, 52-bit VAs, pgdp=000000008ba16000
[fff00000f9b40000] pgd=180000013ffff403, p4d=180000013fffe403, pud=180000013f85b403, pmd=180000013f68d403, pte=0000000000000000
Internal error: Oops: 0000000096000007 [#1] SMP
Modules linked in:
CPU: 0 UID: 0 PID: 303 Comm: kworker/0:1 Not tainted 6.19.0-rc1-00002-gda407d200220 #34 PREEMPT
Hardware name: QEMU QEMU Virtual Machine, BIOS unknown 02/02/2022
Workqueue: kacpi_notify acpi_os_execute_deferred
pstate: 214020c5 (nzCv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
pc : hex_dump_to_buffer+0x30c/0x4a0
lr : hex_dump_to_buffer+0x328/0x4a0
sp : ffff800080e13880
x29: ffff800080e13880 x28: ffffac9aba86f6a8 x27: 0000000000000083
x26: fff00000f9b3fffc x25: 0000000000000004 x24: 0000000000000004
x23: ffff800080e13905 x22: 0000000000000010 x21: 0000000000000083
x20: 0000000000000001 x19: 0000000000000008 x18: 0000000000000010
x17: 0000000000000001 x16: 00000007c7f20fec x15: 0000000000000020
x14: 0000000000000008 x13: 0000000000081020 x12: 0000000000000008
x11: ffff800080e13905 x10: ffff800080e13988 x9 : 0000000000000000
x8 : 0000000000000000 x7 : 0000000000000001 x6 : 0000000000000020
x5 : 0000000000000030 x4 : 00000000fffffffe x3 : 0000000000000000
x2 : ffffac9aba78c1c8 x1 : ffffac9aba76d0a8 x0 : 0000000000000008
Call trace:
hex_dump_to_buffer+0x30c/0x4a0 (P)
print_hex_dump+0xac/0x170
cper_estatus_print_section+0x90c/0x968
cper_estatus_print+0xf0/0x158
__ghes_print_estatus+0xa0/0x148
ghes_proc+0x1bc/0x220
ghes_notify_hed+0x5c/0xb8
notifier_call_chain+0x78/0x148
blocking_notifier_call_chain+0x4c/0x80
acpi_hed_notify+0x28/0x40
acpi_ev_notify_dispatch+0x50/0x80
acpi_os_execute_deferred+0x24/0x48
process_one_work+0x15c/0x3b0
worker_thread+0x2d0/0x400
kthread+0x148/0x228
ret_from_fork+0x10/0x20
Code: 6b14033f 540001ad a94707e2 f100029f (b8747b44)
---[ end trace 0000000000000000 ]---
Prevent that by taking the actual allocated are into account when
checking for CPER length.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
[ rjw: Subject tweaks ]
Link: https://patch.msgid.link/4e70310a816577fabf37d94ed36cde4ad62b1e0a.1767871950.git.mchehab+huawei@kernel.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
If the BIOS generates a very small ARM Processor Error, or
an incomplete one, the current logic will fail to deferrence
err->section_length
and
ctx_info->size
Add checks to avoid that. With such changes, such GHESv2
records won't cause OOPSes like this:
[ 1.492129] Internal error: Oops: 0000000096000005 [#1] SMP
[ 1.495449] Modules linked in:
[ 1.495820] CPU: 0 UID: 0 PID: 9 Comm: kworker/0:0 Not tainted 6.18.0-rc1-00017-gabadcc3553dd-dirty #18 PREEMPT
[ 1.496125] Hardware name: QEMU QEMU Virtual Machine, BIOS unknown 02/02/2022
[ 1.496433] Workqueue: kacpi_notify acpi_os_execute_deferred
[ 1.496967] pstate: 814000c5 (Nzcv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
[ 1.497199] pc : log_arm_hw_error+0x5c/0x200
[ 1.497380] lr : ghes_handle_arm_hw_error+0x94/0x220
0xffff8000811c5324 is in log_arm_hw_error (../drivers/ras/ras.c:75).
70 err_info = (struct cper_arm_err_info *)(err + 1);
71 ctx_info = (struct cper_arm_ctx_info *)(err_info + err->err_info_num);
72 ctx_err = (u8 *)ctx_info;
73
74 for (n = 0; n < err->context_info_num; n++) {
75 sz = sizeof(struct cper_arm_ctx_info) + ctx_info->size;
76 ctx_info = (struct cper_arm_ctx_info *)((long)ctx_info + sz);
77 ctx_len += sz;
78 }
79
and similar ones while trying to access section_length on an
error dump with too small size.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
[ rjw: Subject tweaks ]
Link: https://patch.msgid.link/7fd9f38413be05ee2d7cfdb0dc31ea2274cf1a54.1767871950.git.mchehab+huawei@kernel.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Commit 32ece31db4df ("ACPI: PM: s2idle: Only retrieve constraints when
needed") attempted to avoid useless evaluation of LPS0 _DSM Function 1
in lps0_device_attach() because pm_debug_messages_on might never be set
(and that is the case on production systems most of the time), but it
turns out that LPS0 _DSM Function 1 is generally problematic on some
platforms and causes suspend issues to occur when pm_debug_messages_on
is set now.
In Linux, LPS0 _DSM Function 1 is only useful for diagnostics and only
in the cases when the system does not reach the deepest platform idle
state during suspend-to-idle for some reason. If such diagnostics is
not necessary, evaluating it is a loss of time, so using it along with
the other pm_debug_messages_on diagnostics is questionable because the
latter is expected to be suitable for collecting debug information even
during production use of system suspend.
For this reason, add a module parameter called check_lps0_constraints
to control whether or not the list of LPS0 constraints will be checked
in acpi_s2idle_prepare_late_lps0() and so whether or not to evaluate
LPS0 _DSM Function 1 (once) in acpi_s2idle_begin_lps0().
Fixes: 32ece31db4df ("ACPI: PM: s2idle: Only retrieve constraints when needed")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Link: https://patch.msgid.link/2827214.mvXUDI8C0e@rafael.j.wysocki
|
|
Commit 32ece31db4df ("ACPI: PM: s2idle: Only retrieve constraints
when needed"), that attempted to avoid useless evaluation of LPS0 _DSM
Function 1 in lps0_device_attach(), forgot to add checks for
lps0_device_handle and sleep_no_lps0 to acpi_s2idle_begin_lps0()
where they should be done before calling lpi_device_get_constraints()
or lpi_device_get_constraints_amd().
Add the missing checks.
Fixes: 32ece31db4df ("ACPI: PM: s2idle: Only retrieve constraints when needed")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Link: https://patch.msgid.link/2818730.mvXUDI8C0e@rafael.j.wysocki
|
|
Move the cpuidle driver check from __acpi_processor_start() to
acpi_processor_power_init() which allows variable acpi_idle_driver to
become static.
No intentional functional impact.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Tested-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://patch.msgid.link/20251223100914.2407069-7-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Notice that acpi_processor_power_init() and acpi_processor_power_exit()
don't need to return any values because their callers don't check them
anyway, so redefine those functions as void.
While at it, rearrange the code in acpi_processor_power_init() to
reduce the indentation level, get rid of a redundant local variable
in that function, and rephrase a code comment in it.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Tested-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://patch.msgid.link/20251223100914.2407069-5-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Commit 7a8c994cbb2d ("ACPI: processor: idle: Optimize ACPI idle
driver registration") moved the ACPI idle driver registration to
acpi_processor_driver_init() and acpi_processor_power_init() does
not register an idle driver any more.
Accordingly, the cpuidle driver check in __acpi_processor_start() needs
to be updated to avoid calling acpi_processor_power_init() without a
cpuidle driver, in which case the registration of the cpuidle device
in that function would lead to a NULL pointer dereference in
__cpuidle_register_device().
Fixes: 7a8c994cbb2d ("ACPI: processor: idle: Optimize ACPI idle driver registration")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Tested-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://patch.msgid.link/20251223100914.2407069-4-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Currently, the ACPI idle driver is registered from within a CPU
hotplug callback. Although this didn't cause any functional issues,
this is questionable and confusing. And it is better to register
the cpuidle driver when all of the CPUs have been brought up.
So add a new function to initialize acpi_idle_driver based on the
power management information of an available CPU and register cpuidle
driver in acpi_processor_driver_init().
This commit has four changes under the commit 7a8c994cbb2d (ACPI:
processor: idle: Optimize ACPI idle driver registration):
1) move acpi_processor_register_idle_driver() ahead of the
driver_register().
2) add acpi_processor_cstate_first_run_checks() before calling
acpi_processor_get_power_info().
3) squash the commit 9d68320b2bca (ACPI: processor: idle: Fix
function defined but not used warning) into this change.
4) use for_each_possible_cpu(cpu) to scan all possible cpus.
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Tested-by: Borislav Petkov (AMD) <bp@alien8.de>
[ rjw: New comment edits, changelog tweak ]
Link: https://patch.msgid.link/20251223100914.2407069-2-lihuisong@huawei.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
AGDI has two types of signaling modes: SDEI and interrupt.
Currently, the AGDI driver only supports SDEI.
Therefore, add support for interrupt signaling mode.
The interrupt vector is retrieved from the AGDI table, and call panic
function when an interrupt occurs.
Acked-by: Hanjun Guo <guohanjun@huawei.com>
Reviewed-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
Signed-off-by: Kazuhiro Abe <fj1078ii@aa.jp.fujitsu.com>
Signed-off-by: Will Deacon <will@kernel.org>
|
|
Don't populate the read-only array non_mmio_desc on the stack at run
time, instead make it static const.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Link: https://patch.msgid.link/20251219215900.494211-1-colin.i.king@gmail.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Replace the separate declaration of 'resource_list' and subsequent
INIT_LIST_HEAD() call with LIST_HEAD(), which declares and initializes
the list head in one idiomatic step. This reduces code verbosity and
aligns with common kernel coding patterns, without functional change.
Signed-off-by: Can Peng <pengcan@kylinos.cn>
[ rjw: Subject edits ]
Link: https://patch.msgid.link/20251209072757.3110467-1-pengcan@kylinos.cn
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
In ACPI Global System Interrupts (GSIs) are described using a 32-bit
value.
ACPI/PCI legacy interrupts (INTx) parsing code treats GSIs as 'int',
which poses issues if the GSI interrupt value is a 32-bit value with the
MSB set (as required in some interrupt configurations - eg ARM64 GICv5
systems) because acpi_pci_link_allocate_irq() treats a negative gsi
return value as a failed GSI allocation (and acpi_irq_get_penalty()
would trigger an out-of-bounds array dereference if the 'irq' param is
a negative value).
Fix ACPI/PCI legacy INTx parsing by converting variables representing
GSIs from 'int' to 'u32' bringing the code in line with the ACPI
specification and fixing the current parsing issue.
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/20260105101705.36703-1-lpieralisi@kernel.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
