Re: [PATCH 1/3] PCI: hv: Set irq_retrigger callback for the Hyper-V PCI MSI irqchip

[Wei Liu <wei.liu@kernel.org>]

On Mon, Aug 24, 2026 at 11:57:56AM +0530, Naman Jain wrote:
> 
> 
> On 8/24/2026 5:44 AM, Wei Liu wrote:
> > On Thu, Jul 30, 2026 at 04:56:27PM +0530, Naman Jain wrote:
> > > 
> > > 
> > > On 7/28/2026 7:34 PM, sashiko-bot@kernel.org wrote:
> > > > Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> > > > 
> > > > Pre-existing issues:
> > > > - [High] The fix for dropped pending interrupts during CPU hot-unplug was missed in the sister Hyper-V root partition MSI domain.
> > > > --
> > > > 
> > > > --- Patch [1]: [PATCH 1/3] PCI: hv: Set irq_retrigger callback for the Hyper-V PCI MSI irqchip ---
> > > > commit b8b1c843c457dd145f0d360b44f4b1bf34bd8cf5
> > > > Author: Naman Jain <namjain@linux.microsoft.com>
> > > > 
> > > > PCI: hv: Set irq_retrigger callback for the Hyper-V PCI MSI irqchip
> > > > 
> > > > The Hyper-V vPCI MSI irqchip never installs an irq_retrigger() callback.
> > > > On CPU hot-unplug fixup_irqs() migrates the interrupts which are affine to
> > > > the outgoing CPU to a new target. As the Hyper-V PCI/MSI chip does not
> > > > provide that callback, the pending interrupt is silently dropped, which
> > > > can result in lost interrupts, stalls and "No irq handler for vector"
> > > > messages.
> > > > 
> > > > Fixes: 4daace0d8ce85 ("PCI: hv: Add paravirtual PCI front-end for Microsoft Hyper-V VMs")
> > > > Cc: stable@vger.kernel.org
> > > > Suggested-by: Long Li <longli@microsoft.com>
> > > > Suggested-by: Thomas Gleixner <tglx@kernel.org>
> > > > Signed-off-by: Naman Jain <namjain@linux.microsoft.com>
> > > > 
> > > > > diff --git a/drivers/pci/controller/pci-hyperv.c b/drivers/pci/controller/pci-hyperv.c
> > > > > index cfc8fa403dad6..89816a2bd7cd3 100644
> > > > > --- a/drivers/pci/controller/pci-hyperv.c
> > > > > +++ b/drivers/pci/controller/pci-hyperv.c
> > > > > @@ -2119,6 +2119,7 @@ static bool hv_pcie_init_dev_msi_info(struct device *dev, struct irq_domain *dom
> > > > >    	info->ops->msi_prepare = hv_msi_prepare;
> > > > >    	chip->irq_set_affinity = irq_chip_set_affinity_parent;
> > > > > +	chip->irq_retrigger = irq_chip_retrigger_hierarchy;
> > > > 
> > > > [Severity: High]
> > > > This isn't a bug introduced by this patch, but does the sister Hyper-V root
> > > > partition MSI domain in arch/x86/hyperv/irqdomain.c also need this fix?
> > > > 
> > > > Looking at hv_init_dev_msi_info(), it appears to be missing the exact same
> > > > callback initialization for the root partition's top-level irq_chip:
> > > > 
> > > > arch/x86/hyperv/irqdomain.c:hv_init_dev_msi_info() {
> > > > 	...
> > > > 	if (!msi_lib_init_dev_msi_info(dev, domain, real_parent, info))
> > > > 		return false;
> > > > 
> > > > 	chip->flags |= IRQCHIP_SKIP_SET_WAKE | IRQCHIP_MOVE_DEFERRED;
> > > > 
> > > > 	info->ops->msi_prepare = pci_msi_prepare;
> > > > 
> > > > 	return true;
> > > > }
> > > > 
> > > > If a CPU is hot-unplugged while a device MSI interrupt is pending in its
> > > > local APIC IRR, could fixup_irqs() fail to resend the pending interrupt
> > > > to the new target CPU because this callback is NULL?
> > > > 
> > > > This could potentially lead to the same lost interrupts and I/O stalls
> > > > described in this commit message, just in the root partition MSI domain
> > > > instead of the guest paravirtual MSI domain.
> > > > 
> > > > >    	if (IS_ENABLED(CONFIG_X86))
> > > > >    		chip->flags |= IRQCHIP_MOVE_DEFERRED;
> > > > > 
> > > > 
> > > 
> > > I am not planning to include this in this series, since this issue is not
> > > seen in root partition yet and I am not sure if it is applicable here.
> > > Second patch would anyways fix it going forward, but I am OK adding the
> > > change for this, if we need.
> > 
> > What's the downside of fixing the root partition irq domain?
> 
> No downside as such. It would be a latent bug, exposed only if we exercise
> hotplug in root partition, which seems like we don't. After patch 2 of this
> series, that would go away anyways. If you tell me, I can add a change
> similar to patch 1, which is backportable, but then I would need to remove
> it similarly in patch 3.

Yes, please fix that for the root partition, too. Being consistent
across different use cases is going to save us some headache in the
future.

Wei