> From: Alex Williamson <alex.williamson@xxxxxxxxxx> > Sent: Friday, March 8, 2024 4:21 AM > > On Thu, 7 Mar 2024 08:37:53 +0000 > "Tian, Kevin" <kevin.tian@xxxxxxxxx> wrote: > > > > From: Alex Williamson <alex.williamson@xxxxxxxxxx> > > > Sent: Thursday, March 7, 2024 5:15 AM > > > > > > Mask operations through config space changes to DisINTx may race INTx > > > configuration changes via ioctl. Create wrappers that add locking for > > > paths outside of the core interrupt code. > > > > > > In particular, irq_type is updated holding igate, therefore testing > > > is_intx() requires holding igate. For example clearing DisINTx from > > > config space can otherwise race changes of the interrupt configuration. > > > > > > > Looks the suspend path still checks irq_type w/o holding igate: > > > > vdev->pm_intx_masked = ((vdev->irq_type == > VFIO_PCI_INTX_IRQ_INDEX) && > > vfio_pci_intx_mask(vdev)); > > > > Is it with assumption that no change of configuration is possible at > > that point? > > Yes, I believe this is relatively safe because userspace is frozen at > this point. That's not however to claim that irq_type is absolutely > used consistently after this series. I just didn't see the other > violations rise to the same level as the fixes in this series and > wanted to avoid the distraction. I've stashed a number of patches that > I'd eventually like to post as follow-ups to this series. Thanks, > Reviewed-by: Kevin Tian <kevin.tian@xxxxxxxxx>
