On Wed, Apr 06, 2022 at 02:23:09PM -0700, Ira Weiny wrote: > On Wed, Apr 06, 2022 at 11:11:10AM +0100, Jonathan Cameron wrote: > > On Tue, 5 Apr 2022 16:22:11 -0700 > > Ira Weiny <ira.weiny@xxxxxxxxx> wrote: > > > On Sat, Apr 02, 2022 at 04:48:45PM +0200, Lukas Wunner wrote: > > > > On Thu, Mar 31, 2022 at 08:19:56AM -0700, Ira Weiny wrote: > > > > > On Wed, Mar 30, 2022 at 10:50:31PM -0700, Christoph Hellwig wrote: > > > > > > On Wed, Mar 30, 2022 at 04:59:13PM -0700, ira.weiny@xxxxxxxxx wrote: > > > > > > > Introduced in a PCI v6.0[1], DOE provides a config space based mailbox > > > > > > > with standard protocol discovery. Each mailbox is accessed through a > > > > > > > DOE Extended Capability. > > > > > > > > > > > > I really don't think this should be built unconditionally and bloat > > > > > > every single kernel built with PCI support. > > > > > > > > > > I can add a Kconfig. > > > > > > > > Ideally, that config option should live in the pcie/ subdirectory, > > > > i.e. in drivers/pci/pcie/Kconfig, alongside drivers/pci/pcie/doe.c, > > > > as we try to consolidate PCIe-specific features there and reserve > > > > core code in drivers/pci/*.c for functionality that also applies > > > > to Conventional PCI. > > > > > > Thanks for letting me know about this direction. I was unaware of this. > > > > We had this in the pcie directory, but Bjorn asked us to move it to the pci > > directory as there isn't anything specific to PCIe about DOE. You could > > implement it on pci-x (maybe 2.0?) I think even though it's in the PCIe specification. > > > > https://lore.kernel.org/linux-pci/20210413194927.GA2241331@bjorn-Precision-5520/ > > :-/ > > I'd forgotten that far back. > > Bjorn? I would still prefer it in drivers/pci because I don't think there's enough value to justify the pcie/ subdirectory. We have ats.c, ecam.c, iov.c, pci-pf-stub.c, and vc.c in drivers/pci even though they're PCIe-specific. Other files in drivers/pci like access.c, pci-acpi.c, pci.c, probe.c, etc also have some PCIe content. Bjorn