> This specific hunk is only evaluated if UPF_CONS_FLOW is set; > currently, there are only 3 in-tree users of UPF_CONS_FLOW and all of them > are samsung UARTs. Furthermore, UPF_CONS_FLOW cannot be set from userspace > with ioctl(TIOCSSERIAL). Now it makes sense. There were two mistakes I made: 1. When originally writing this, I assumed UPF_CONS_FLOW was actually something important, and 2. When responding, I managed to get confused between this and the earlier loop (the UPF_CONS_FLOW part didn't make it into the 3 lines of context) which *is* the important part. >> If I understand what you mean by "the actual shifter", i.e. the difference >> between THRE and TEMT, then that's always 8 bits (+ start + stop). > I doubt that's true for every uart design supported by this driver. But > I do think the number is small, and I agree that it's almost certainly < 16 bits > for 16550 work-alikes. Perhaps I haven't developed the necessary cynicism regarding hardware designers. :-) A question I might ask you is whether you think it's necessary to wait for TEMT in serial8250_put_poll_char at all. Yes, a panic just as the baud rate is being changed or a break is being sent could mangle the last character, but such a race is a serious risk of losing those characters *anyway*. -- To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
