qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Qemu-devel] [PATCH v4 00/10] pc: do not create invalid MADT.LAPIC/Proce


From: Igor Mammedov
Subject: [Qemu-devel] [PATCH v4 00/10] pc: do not create invalid MADT.LAPIC/Processor entries
Date: Fri, 26 Feb 2016 14:59:18 +0100

Changes since v3:
 - replace possible-cpus QOM interface with
   MachineClass.possible_cpu_arch_ids() hook
     Eduardo Habkost <address@hidden>
 - add ACPI tables test case for sparse topology
Changes since v2:
 - build possible CPUs list only once.
     Marcel Apfelbaum <address@hidden>
 - replace MachineClass possible_cpu_arch_ids() hook with
   QOM interface, so only targets that need it would implement it
 - fix ^2 times present CPU lookup for initial CPUs
    Eduardo Habkost <address@hidden>
 - drop found_cpus bitmap altogether

Changes since v1:
 - rebased on top of PCI tree that contains
   Eduardo's guest_info removel series
 - fix ^2 times present CPU lookup when creating CPON
   package (spotted-by: Eduardo Habkost <address@hidden>)

It's mostly clean up series that removes invalid CPU
entries from MADT/DSDT/SRAT tables when APIC IDs are
sparse distributed*.
Series also removes intermediate present CPUs bitmap
in ACPI tables generation code, replacing it with
machine reported presence status. That should
help later for consolidating and sharing CPU hotplug
codebase and extending supported CPU count above 256
on ACPI side, where I'm going to replace current
"not scalable" bitmap based CPU hotplug MMIO interface
with memory-hotplug like one, which could easily
scale and provide additional info for ACPI CPU device
objects.

Tested hoptlug with: RHEL72 and WS2003 / WS2012R2.
Git tree for testing:
https://github.com/imammedo/qemu.git pc_madt_dsdt_lapic_cleanups_v4

* example topology with sparse APIC IDs:
  -smp X,sockets=2,cores=3,maxcpus=6

* it's not possible to remove notion of apic_ad_limit
  since guest visible interfaces like CPU hoptlug MMIO
  (CPON array in ACPI + corresponding MMIO in QEMU) and
  FWCFG should stay the same for compat reasons with
  current setups and legacy SeaBIOS.

Igor Mammedov (10):
  tests: pc: acpi: piix4: add sparse CPU hotplug case
  pc: init pcms->apic_id_limit once and use it throughout pc.c
  machine: introduce MachineClass.possible_cpu_arch_ids() hook
  pc: acpi: cleanup qdev_get_machine() calls
  pc: acpi: SRAT: create only valid processor lapic entries
  pc: acpi: create MADT.lapic entries only for valid lapics
  pc: acpi: create Processor and Notify objects only for valid lapics
  pc: acpi: drop cpu->found_cpus bitmap
  pc: acpi: clarify why possible LAPIC entries must be present in MADT
  tests: update ACPI tables blobs for cpuhp_sparse case

 hw/i386/acpi-build.c                      | 147 +++++++++++++++---------------
 hw/i386/pc.c                              |  89 ++++++++++++------
 include/hw/boards.h                       |  26 ++++++
 include/hw/i386/pc.h                      |   1 +
 tests/acpi-test-data/pc/APIC.cpuhp_sparse | Bin 0 -> 160 bytes
 tests/acpi-test-data/pc/DSDT.cpuhp_sparse | Bin 0 -> 5821 bytes
 tests/acpi-test-data/pc/SRAT.cpuhp_sparse | Bin 0 -> 264 bytes
 tests/bios-tables-test.c                  |  16 ++++
 8 files changed, 175 insertions(+), 104 deletions(-)
 create mode 100644 tests/acpi-test-data/pc/APIC.cpuhp_sparse
 create mode 100644 tests/acpi-test-data/pc/DSDT.cpuhp_sparse
 create mode 100644 tests/acpi-test-data/pc/SRAT.cpuhp_sparse

-- 
1.8.3.1




reply via email to

[Prev in Thread] Current Thread [Next in Thread]