On 29 April 2016 at 09:19, Ni, Ruiyu <[email protected]> wrote: > Ard, > At first glance of this patch, I immediately checked the PI spec > SubResources() description, > it says when a root bridge doesn't require any resource, "a zero-length > resource request > must explicitly be submitted." > "zero-length resource" can be interpreted in two ways: > 1. only END descriptor > 2. MEM resource with 0-length, plus IO resource with 0-length, plus END > descriptor > > I believe that both ways should be accepted by PciHostBridge driver. > But without your patch, the zero descriptor cannot be accepted by > PciHostBridge driver. > I think it fails in the check: > > case ACPI_ADDRESS_SPACE_TYPE_MEM /* 0 */: > if (Descriptor->AddrSpaceGranularity != 32 && > Descriptor->AddrSpaceGranularity != 64) { > return EFI_INVALID_PARAMETER; > } >
I agree with your analysis. The NULL entry is misidentified as a ACPI_ADDRESS_SPACE_TYPE_MEM resource descriptor because the constant is defined as 0 > I agree with your patch. > > Reviewed-by: Ruiyu Ni <[email protected]> > Thank you. _______________________________________________ edk2-devel mailing list [email protected] https://lists.01.org/mailman/listinfo/edk2-devel
