Marek Vasut wrote:

So what alignment problems do you observe ? If you copy using the CPU only, why do you need the bounce buffer at all ? I don't quite get it.

Sorry for not explaining it good enough: I don't observe any alignment problems. mach-socfpga can do unaligned accesses as well. This driver does CPU based copy, but since it transfers words, not bytes, I guess on other platforms, the CPU might fail when trying to read a word from an unaligned pointer.

Vignesh added this about a year ago and from that commit message, it seems like he was observing alignment errors on his TI platform. Also, those commits have a reviewed-by tag from you and Jagan.

For me, reverting these patches would be OK as well, but I guess Vinesh's TI platform kind of needs them?

Simon