On Wed, Feb 13, 2019 at 09:03:58PM +0900, Takao Indoh wrote: > Ok, let me think about how I should change this patch.
Just drop it. > I'm thinking that the problem of this patch is adding processor specific > code into NVMe common driver, is this correct? Or another problem? It > would be great if you could give me a hint to improve this patch. For one it is processor / root port specific. But even more importantly it is incorrect. If the device wants to issue a relaxed order transaction it should do so all by itself, not by hacking around it from the kernel.