CPU_SCAN_METHOD was processing insert events first and only if insert event was
not present then it would check remove event.
Normally it's not an issue as it doesn't make much sense tho hotplug and
immediately unplug it. In this corner case, which can be reproduced with:
qemu -smp 1,maxcpus=2 -cpu host -monitor stdio \
-drive if=pflash,format=raw,readonly,file=edk2-x86_64-code.fd
* boot till GRUB prompt and pause guest (either via monitor or stop GRUB
from automatic boot)
* at monitor prompt add CPU:
device_add host-x86_64-cpu,socket-id=0,core-id=1,thread-id=0,id=foo
* let guest OS boot completely, and unplug CPU from monitor prompt:
device_del foo
which triggers GPE event that leads to CPU_SCAN_METHOD on guest side
as result of above cpu 'foo' will not be hotunplugged, since QEMU sees
insert event and ignores remove event (leaving it in pending state) for
the GPE event.
Any follow up CPU hotplug/unplug action from QEMU side will handle
previously ignored event, so as workaround user can repeat device_del.
Fix this corner-case by queuing remove events independently from insert
events, aka the same way as we do with insert events. And then go over remove
queue to send eject notify events to OSPM within the same GPE event.
PS:
Process remove queue after the cpu add queue has been processed 1st
to ensure that OSPM gets hotadd evets after hotremove ones.
PS2:
Case where it's still borken happens when guest OS is Linux and
device_del happens before guest OS initializes ACPI subsystem.
Culprit in this case though is the guest kernel, which mangles GPE.sts
(by clearing them up) and thus pending SCI turns to NOP leaving
insert/remove events in pending state.
That is the guest bug and should be fixed there.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reported-by: Eric Mackay <eric.mackay@oracle.com>
Message-Id: <20241210163945.3422623-3-imammedo@redhat.com>
Tested-by: Eric Mackay <eric.mackay@oracle.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Headers in include/sysemu/ are not only related to system
*emulation*, they are also used by virtualization. Rename
as system/ which is clearer.
Files renamed manually then mechanical change using sed tool.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Message-Id: <20241203172445.28576-1-philmd@linaro.org>
This reverts commit 2d6cfbaf17.
The patch is supposed to be part of ARM CPU hotplug series and has not value
on its own without it. The series however is still in RFC stage and outside
of scope 9.2 release.
On top of that it introduces not needed callback that pokes directly into
CPU state without any need for that. Instead properties and AML generator
option should be used to configure static platform depended vCPU presence
state.
Drop the patch so that corrected version could be posted along with
ARM CPU hotplug series and properly reviewed in relevant context.
That also helps us to keep history cleaner with new patch being
against original code vs a string of fixups on top of current mess.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20241112170258.2996640-4-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reflect the QOM vCPUs ACPI CPU hotplug states in the `_STA.Present` and
and `_STA.Enabled` bits when the guest kernel evaluates the ACPI
`_STA` method during initialization, as well as when vCPUs are
hot-plugged or hot-unplugged. If the CPU is present then the its
`enabled` status can be fetched using architecture-specific code [1].
Reference:
[1] Example implementation of architecture-specific hook to fetch CPU
`enabled status
Link: https://github.com/salil-mehta/qemu/commit/c0b416b11e5af6505e558866f0eb6c9f3709173e
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Message-Id: <20241103102419.202225-4-salil.mehta@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
On most architectures, during vCPU hot-plug and hot-unplug actions, the
firmware or VMM/QEMU can update the OS on vCPU status by toggling the
ACPI method `_STA.Present` bit. However, certain CPU architectures
prohibit [1] modifications to a CPU’s `presence` status after the kernel
has booted.
This limitation [2][3] exists because many per-CPU components, such as
interrupt controllers and various per-CPU features tightly integrated
with CPUs, may not support reconfiguration once the kernel is
initialized. Often, these components cannot be powered down, as they may
belong to an `always-on` power domain. As a result, some architectures
require all CPUs to remain `_STA.Present` after system initialization.
Therefore, it is essential to mirror the exact QOM vCPU status through
ACPI for the Guest kernel. For this, we should determine—via
architecture-specific code[4]—whether vCPUs must always remain present
and whether the associated `AcpiCpuStatus::cpu` object should remain
valid, even following a vCPU hot-unplug operation.
References:
[1] Check comment 5 in the bugzilla entry
Link: https://bugzilla.tianocore.org/show_bug.cgi?id=4481#c5
[2] KVMForum 2023 Presentation: Challenges Revisited in Supporting Virt CPU Hotplug on
architectures that don’t Support CPU Hotplug (like ARM64)
a. Kernel Link: https://kvm-forum.qemu.org/2023/KVM-forum-cpu-hotplug_7OJ1YyJ.pdf
b. Qemu Link: https://kvm-forum.qemu.org/2023/Challenges_Revisited_in_Supporting_Virt_CPU_Hotplug_-__ii0iNb3.pdf
[3] KVMForum 2020 Presentation: Challenges in Supporting Virtual CPU Hotplug on
SoC Based Systems (like ARM64)
Link: https://kvmforum2020.sched.com/event/eE4m
[4] Example implementation of architecture-specific CPU persistence hook
Link: https://github.com/salil-mehta/qemu/commit/c0b416b11e5af6505e558866f0eb6c9f3709173e
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Message-Id: <20241103102419.202225-2-salil.mehta@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The "hw/boards.h" is unused since the previous commit. Since its removal
requires include fixes in various unrelated files to keep the code compiling it
has been split in a dedicated commit.
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230908084234.17642-5-shentey@gmail.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
build_cpus_aml() is architecture independent but needs to create architecture-
specific CPU AML. So far this was achieved by using a virtual method from
TYPE_ACPI_DEVICE_IF. However, build_cpus_aml() would resolve this interface from
global (!) state. This makes it quite incomprehensible where this interface
comes from (TYPE_PIIX4_PM?, TYPE_ICH9_LPC_DEVICE?, TYPE_ACPI_GED_X86?) an can
lead to crashes when the generic code is ported to new architectures.
So far, build_cpus_aml() is only called in architecture-specific code -- and
only in x86. We can therefore simply pass pc_madt_cpu_entry() as callback to
build_cpus_aml(). This is the same callback that would be used through
TYPE_ACPI_DEVICE_IF.
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230908084234.17642-3-shentey@gmail.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The has_FOO for pointer-valued FOO are redundant, except for arrays.
They are also a nuisance to work with. Recent commit "qapi: Start to
elide redundant has_FOO in generated C" provided the means to elide
them step by step. This is the step for qapi/acpi.py.
Said commit explains the transformation in more detail. The invariant
violations mentioned there do not occur here.
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Ani Sinha <ani@anisinha.ca>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20221104160712.3005652-7-armbru@redhat.com>
The migration code will not look at a VMStateDescription's
minimum_version_id_old field unless that VMSD has set the
load_state_old field to something non-NULL. (The purpose of
minimum_version_id_old is to specify what migration version is needed
for the code in the function pointed to by load_state_old to be able
to handle it on incoming migration.)
We have exactly one VMSD which still has a load_state_old,
in the PPC CPU; every other VMSD which sets minimum_version_id_old
is doing so unnecessarily. Delete all the unnecessary ones.
Commit created with:
sed -i '/\.minimum_version_id_old/d' $(git grep -l '\.minimum_version_id_old')
with the one legitimate use then hand-edited back in.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
It missed vmstate_ppc_cpu.
Instead of composing disabled _MAT entry and then later on
patching it to enabled for hotpluggbale CPUs in DSDT,
set it to enabled at the time _MAT entry is built.
It will allow to drop usage of packed structures in
following patches when build_madt() is switched to use
build_append_int_noprefix() API.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-Id: <20210924122802.1455362-24-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
if firmware and QEMU negotiated CPU hotunplug support, generate
_EJ0 method so that it will mark CPU for removal by firmware and
pass control to it by triggering SMI.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20201207140739.3829993-6-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Adds bit #4 to status/control field of CPU hotplug MMIO interface.
New bit will be used OSPM to mark CPUs as pending for removal by firmware,
when it calls _EJ0 method on CPU device node. Later on, when firmware
sees this bit set, it will perform CPU eject which will clear bit #4
as well.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20201207140739.3829993-3-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Only qemu-system-FOO and qemu-storage-daemon provide QMP
monitors, therefore such declarations and definitions are
irrelevant for user-mode emulation.
Extracting the ACPI commands to their own schema reduces the size of
the qapi-misc* headers generated, and pulls less QAPI-generated code
into user-mode.
Suggested-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Igor Mammedov <imammedo@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20200913195348.1064154-8-philmd@redhat.com>
[Commit message tweaked]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
In case firmware has negotiated CPU hotplug SMI feature, generate
AML to describe SMI IO port region and send SMI to firmware
on each CPU hotplug SCI in case new CPUs were hotplugged.
Since new CPUs can be hotplugged while CPU_SCAN_METHOD is running
we can't send SMI before new CPUs are fetched from QEMU as it
could cause sending Notify to a CPU that firmware hasn't seen yet.
So fetch new CPUs into local cache first, then send SMI and
after that send Notify events to cached CPUs. This should ensure
that Notify is sent only to CPUs which were processed by firmware
first.
Any CPUs that were hotplugged after caching will be processed
by the next CPU_SCAN_METHOD, when pending SCI is handled.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20200923094650.1301166-10-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Firmware can enumerate present at boot APs by broadcasting wakeup IPI,
so that woken up secondary CPUs could register them-selves.
However in CPU hotplug case, it would need to know architecture
specific CPU IDs for possible and hotplugged CPUs so it could
prepare environment for and wake hotplugged AP.
Reuse and extend existing CPU hotplug interface to return architecture
specific ID for currently selected CPU in 2 registers:
- lower 32 bits in ACPI_CPU_CMD_DATA_OFFSET_RW
- upper 32 bits in ACPI_CPU_CMD_DATA2_OFFSET_R
On x86, firmware will use CPHP_GET_CPU_ID_CMD for fetching the APIC ID
when handling hotplug SMI.
Later, CPHP_GET_CPU_ID_CMD will be used on ARM to retrieve MPIDR,
which serves the similar to APIC ID purpose.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1575896942-331151-10-git-send-email-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
No functional change in practice, patch only aims to properly
document (in spec and code) intended usage of the reserved space.
The new field is to be used for 2 purposes:
- detection of modern CPU hotplug interface using
CPHP_GET_NEXT_CPU_WITH_EVENT_CMD command.
procedure will be described in follow up patch:
"acpi: cpuhp: spec: add typical usecases"
- for returning upper 32 bits of architecture specific CPU ID,
for new CPHP_GET_CPU_ID_CMD command added by follow up patch:
"acpi: cpuhp: add CPHP_GET_CPU_ID_CMD command"
Change is backward compatible with 4.2 and older machines, as field was
unconditionally reserved and always returned 0x0 if modern CPU hotplug
interface was enabled.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1575896942-331151-8-git-send-email-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
