On Thursday 10 March 2016, Vineet Gupta wrote: > On Wednesday 09 March 2016 10:51 PM, Lada Trimasova wrote: > > Memory access primitives should use cpu_to_le16, cpu_to_le32, le16_to_cpu > > and le32_to_cpu because it is not really guaranteed that drivers handles > > any ordering themselves. > > That is the driver issue. readxx as API simply returns data in native endianness. > We've had EZChip running big endian and so far and they didn't need this change. Most drivers using readl() or readl_relaxed() expect those to perform byte swaps on big-endian architectures, as the registers tend to be fixed endian, so the change seems reasonable. It depends a little bit on how endian mode is implemented in the CPU: do you follow the normal model of swapping byte order in the load/store instructions of the CPU when running big-endian, or is the CPU always running little-endian but the bus addresses get mangled on byte accesses to give the illusion of a big-endian system? Arnd
