After this change, in 'safe' mode we get the correct MPS. The kernel log with DEBUG defined for drivers/pci shows: pci 0000:03:00.0: Dev MPS 128 MPSS 256 MRRS 512 pci 0000:03:00.0: Dev MPS 256 MPSS 256 MRRS 512 pci 0000:03:00.1: Dev MPS 128 MPSS 256 MRRS 512 pci 0000:03:00.1: Dev MPS 256 MPSS 256 MRRS 512 but in 'performance' mode the result is incorrect: pci 0000:03:00.0: Dev MPS 128 MPSS 256 MRRS 512 pci_bus 0000:03: Bus MPSS 512 pci 0000:03:00.0: MPS configured higher than maximum supported by the device. If a bus issue occurs, try running with pci=pcie_bus_safe. pci 0000:03:00.0: Dev MPS 512 MPSS 512 MRRS 512 pci 0000:03:00.1: Dev MPS 128 MPSS 256 MRRS 512 pci_bus 0000:03: Bus MPSS 512 pci 0000:03:00.1: MPS configured higher than maximum supported by the device. If a bus issue occurs, try running with pci=pcie_bus_safe. pci 0000:03:00.1: Dev MPS 512 MPSS 512 MRRS 512 I thought the idea was to let bridges have higher MPS than some of their descendants while setting MRRS on the descendants as an extra constraint on payload size. But this is instead increasing MPS and MRRS (and the nominal MPSS!) on the descendant, which is crazy. However, Jon's latest fixes seem to produce the correct result: pci 0000:03:00.0: Dev MPS 128 MPSS 256 MRRS 512 pci 0000:03:00.0: Dev MPS 256 MPSS 256 MRRS 256 pci 0000:03:00.1: Dev MPS 128 MPSS 256 MRRS 512 pci 0000:03:00.1: Dev MPS 256 MPSS 256 MRRS 256 Ben. -- Ben Hutchings, Staff Engineer, Solarflare Not speaking for my employer; that's the marketing department's job. They asked us to note that Solarflare product names are trademarked. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
