[Qemu-devel] [PATCH v4 0/7] pc: Ensure APIC ID limits before aborting or corrupting memory

[Eduardo Habkost <ehabkost@redhat.com>]

Changes v3 -> v4:
 * Commit message update on patch 5/7
 * Small comment change (s/should/shall/) on patch 6/7

Changes v2 -> v3:
 * Don't use MAX_CPUMASK_BITS on acpi-build.c, use ACPI_CPU_HOTPLUG_ID_LIMIT;
 * Rename MAX_CPUMASK_BITS to MAX_CPUS, and document it;
 * Use MAX_CPUS when checking max_cpus limit on vl.c.

Changes v1 -> v2:
 * None. v1 was tagged locally by mistake and never submitted to qemu-devel.

This series adds checks for APIC ID limits on initialization and CPU hotplug
code.  This fixes multiple issues:

1) Assertion failure when -smp parameter results in a too large APIC ID. e.g.:

    $ ./install/bin/qemu-system-x86_64 -S -smp 254,cores=17,threads=17,sockets=17,maxcpus=254 -nographic
    **
    ERROR:hw/acpi/cpu_hotplug.c:58:AcpiCpuHotplug_init: assertion failed: ((id / 8) < ACPI_GPE_PROC_LEN)
    Aborted (core dumped)

2) Memory corruption on AcpiCpuHotplug_add() when APIC ID is too large (similar
   to the case above, but on CPU hotplug).

Eduardo Habkost (7):
  acpi: Add ACPI_CPU_HOTPLUG_ID_LIMIT macro
  pc: Refuse CPU hotplug if the resulting APIC ID is too large
  acpi: Assert sts array limit on AcpiCpuHotplug_add()
  acpi: Don't use MAX_CPUMASK_BITS for APIC ID bitmap
  pc: Refuse max_cpus if it results in too large APIC ID
  vl.c: Rename MAX_CPUMASK_BITS to MAX_CPUS
  vl.c: Use MAX_CPUS macro instead of hardcoded constant

 hw/acpi/cpu_hotplug.c              |  1 +
 hw/i386/acpi-build.c               |  4 ++--
 hw/i386/pc.c                       | 16 ++++++++++++++++
 include/hw/acpi/cpu_hotplug_defs.h |  8 ++++++++
 include/sysemu/sysemu.h            |  9 ++++++++-
 vl.c                               | 12 ++++++------
 6 files changed, 41 insertions(+), 9 deletions(-)

-- 
1.8.5.3