On Mon, Oct 19, 2020 at 12:11:36PM -0500, Tom Lendacky wrote: > > Hum, I assume it is broken also. Actually quite a swath of drivers > > and devices will be broken under this :\ > > Not sure what you mean by the last statement - in general or when running > under VFIO/DPDK? In general, traditional in kernel drivers work just fine > under SME without any changes. Split user/kernel drivers are common enough. Looks like maybe ~50 drivers in the kernel potentially are mmaping IO memory so would be broken here. Looking for pgprot_noncached() or pgprot_writecombine() around VMA mappings is a pretty good clue it is working on IO memory. I checked through the infiniband ones and they seem to be using io_remap vs remap properly, but other places may need fixing. Jason