> ...instead of taking and dropping a reference in both uart_insert_char > and uart_push_chars there. But it may be as well an overkill, as this > stuff is not time critical. So any opinions? (If not I will proceed > without ticketing having tty_port_tty_get/put in every call. Like I > have it now.) The cure for this is different (and a spot more invasive) We need to move the tty_buffer objects into the tty_port. At that point the lifetime becomes create device (and tty_port struct) request_irq lifetime of physical port (ttys come and go) free_irq free tty_port done and the locking goes away. On the flush_to_ldisc side we know the port physically exists because the work queue for the ldisc is killed and we wait for it when we kill off the port (we may need a spot more checking there) This needs every tty of all kinds to have a tty_port even if isn't used for anything else. That also cleans a ton of stuff up because then we can always put the tty_port in tty->private_data which means we can clean up some of the glue further. Alan -- 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
