A recent discussion about the PCI/MSI management for virtio unearthed a
violation of the MSI-X specification vs. writing the MSI-X message: under
certain circumstances the entry is written without being masked.
While looking at that and the related violation of the x86 non-remapped
interrupt affinity mechanism a few other issues were discovered by
inspection.
The following series addresses these.
Note this does not fix the virtio issue, but while staring at the above
problems I came up with a plan to address this. I'm still trying to
convince myself that I can get away without sprinkling locking all over the
place, so don't hold your breath that this will materialize tomorrow.
The series is also available from git:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git irq/msi
V1 can be found here:
https://lore.kernel.org/r/20210721191126.274946280@xxxxxxxxxxxxx
Changes vs. V1:
- Identified and addressed more inconsistencies, especially the lack of
serialization for multi-MSI masking
- Removed the extra vector masking in S390
- Addressed review comments and picked up tags where applicable
- Clean up of the naming of msi_desc::masked as discussed in the V1
thread
- Consolidation of the mask/unmask functions
Thanks,
tglx
---
arch/s390/pci/pci_irq.c | 4
arch/x86/kernel/apic/io_apic.c | 6
arch/x86/kernel/apic/msi.c | 11 +
arch/x86/kernel/hpet.c | 2
drivers/base/core.c | 1
drivers/pci/msi.c | 274 ++++++++++++++++++++++-------------------
include/linux/device.h | 1
include/linux/irq.h | 2
include/linux/msi.h | 10 -
kernel/irq/chip.c | 5
10 files changed, 178 insertions(+), 138 deletions(-)
