Fighting with some potential race in ivshmem-net, I noticed that we lack
the right memory barriers before sending ivshmem doorbell events. On
x86, this even revealed some potential issue of Linux [1].

This series cleanes up our dirty corners, adding the right IPI barrier
for x2APIC mode and changing the spec and implementation of doorbell
writes to include a memory barrier. There are also some ARM cleanups
along this.

Please review carefully!

Thanks,
Jan

[1] https://lore.kernel.org/lkml/[email protected]/

Jan Kiszka (6):
  x86: Add proper fence before sending IPIs
  arm-common: Add memory barrier before ivshmem interrupt submission
  arm-common: Get rid of arm_cpu_kick
  core, arm-common: Clarify role of spin_unlock before event submission
  arm-common: Factor out irqchip_trigger_external_irq
  Documentation: ivshmem: Require that doorbell writes act as memory
    barriers

 Documentation/ivshmem-v2-specification.md        |  4 ++++
 hypervisor/arch/arm-common/control.c             |  3 ++-
 hypervisor/arch/arm-common/include/asm/control.h |  6 ------
 hypervisor/arch/arm-common/include/asm/irqchip.h |  2 ++
 hypervisor/arch/arm-common/irqchip.c             | 14 +++++++-------
 hypervisor/arch/arm-common/ivshmem.c             | 11 +++++++++--
 hypervisor/arch/arm-common/psci.c                |  7 ++++++-
 hypervisor/arch/x86/apic.c                       | 11 +++++++++++
 hypervisor/arch/x86/control.c                    |  3 ---
 hypervisor/arch/x86/include/asm/control.h        |  2 --
 hypervisor/control.c                             |  4 ++++
 hypervisor/include/jailhouse/control.h           | 15 +++++++++++++++
 12 files changed, 60 insertions(+), 22 deletions(-)

-- 
2.16.4

-- 
You received this message because you are subscribed to the Google Groups 
"Jailhouse" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jailhouse-dev/cover.1583516039.git.jan.kiszka%40siemens.com.

Reply via email to