Re: [PATCH v3 13/14] PCI/MSI: Document the various ways of ending up with NO_MSI

From: Jeremy Linton
Date: Tue May 18 2021 - 00:29:02 EST


On 3/30/21 10:11 AM, Marc Zyngier wrote:
We have now three ways of ending up with NO_MSI being set.
Document them.

Acked-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
drivers/pci/msi.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
index d9c73c173c14..217dc9f0231f 100644
--- a/drivers/pci/msi.c
+++ b/drivers/pci/msi.c
@@ -871,8 +871,15 @@ static int pci_msi_supported(struct pci_dev *dev, int nvec)
* Any bridge which does NOT route MSI transactions from its
* secondary bus to its primary bus must set NO_MSI flag on
* the secondary pci_bus.
- * We expect only arch-specific PCI host bus controller driver
- * or quirks for specific PCI bridges to be setting NO_MSI.
+ *
+ * The NO_MSI flag can either be set directly by:
+ * - arch-specific PCI host bus controller drivers (deprecated)
+ * - quirks for specific PCI bridges
+ *
+ * or indirectly by platform-specific PCI host bridge drivers by
+ * advertising the 'msi_domain' property, which results in
+ * the NO_MSI flag when no MSI domain is found for this bridge
+ * at probe time.

I have an ACPI machine with a gicv2 (no m), and a MSI region that isn't described by ACPI because its non-standard. In the past this tended to work because PCIe device drivers would fall back to legacy pci intx silently. But, with 5.13, it seems this series now triggers the WARN_ON_ONCE() in arch_setup_msi_irq, because duh, no MSI support.

Everything of course continues to work, it just gets this ugly splat on bootup telling me basically the machine doesn't support MSIs. So, I considered a few patches, including just basically setting nomsi if gicv2 && acpi, or eek a host bridge quirk.

None of these seem great, so how can this be fixed?


for (bus = dev->bus; bus; bus = bus->parent)
if (bus->bus_flags & PCI_BUS_FLAGS_NO_MSI)