On Fri, Oct 28, 2022 at 08:21:54PM +0100, Al Viro wrote: > I wonder if we ought to simply add "depends on BROKEN" for CONFIG_BOUNCE... ;-/ Depends on BROKEN does not help anyone, we might better remove it. Users are: paride drivers drivers/mmc/ for a few drivers 3 SCSI LLDDs (two of those parport attached, one ISA) usb-storage The thing that most matters is usb-storage that can't be used on highmem systems, and maybe one or two of the mmc drivers might exist on arm highmem systems. In general nothing should use these, for MMIO/PIO based drivers they should kmap, and for DMA drivers nothing should use the kernel virtual address, except that the USB interfaces are a mess that seems to partially require one.
