On Wed, 30 Jan 2019 01:17:06 -0800 Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > On Tue, Jan 29, 2019 at 04:24:45PM +0100, Thomas Bogendoerfer wrote: > > > From an abstraction point of view this doesn't really belong into > > > a bridge driver as it is a global exported function. I guess we can > > > keep it here with a fixme comment, but we should probably move this > > > into a method call instead. > > > > or put the nodeid into the bus struct ? > > Doesn't sound to bad to me, you'll just have to update a fair > amount of arch implementations. and it's already there:-) Each struct device has a field numa_node and pci_bus has contains a struct device. arm64 is already using it only not so nice part is the usage of pcibios_root_bridge_prepare() to set the numa_node for the root bus. > > I'm all for it. I looked at the examples for using dma_pfn_offset and the > > only one coming close to usefull for me is arch/sh/drivers/pci/pcie-sh7786.c > > It overloads pcibios_bus_add_device() to set dma_pfn_offset, which doesn't > > look much nicer. What about having a dma_pfn_offset in struct pci_bus > > which all device inherit from ? > > Or add a add_dev callback, similar to what I did for a previous series > that we didn't end up needing after all: > > http://git.infradead.org/users/hch/misc.git/commitdiff/06d9b4fc7deed336edc1292fe2e661729e98ec39 that's exactly what I'm looking for. Should I add the patch for my patchset or are you going to submit it after having a use case ? Thomas. -- SUSE Linux GmbH GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg)
