From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: From: Richard Weinberger Date: Mon, 4 Jun 2012 22:27:31 +0200 Message-Id: <1338841657-30358-1-git-send-email-richard@nod.at> Sender: linux-kernel-owner@vger.kernel.org Subject: um: TTY fixes (?) To: jslaby@suse.cz Cc: user-mode-linux-devel@lists.sourceforge.net, gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org, viro@zeniv.linux.org.uk, alan@linux.intel.com List-ID: This patch set moves the UML console driver to the new TTY port interface. It does ref counting and uses the tty_port_*-helpers. Please note, it's not yet UML mconsole safe! Anyway, I see some really strange things and I'm not sure whether my patch is sane or not... If I implement tty_operations->hangup() the following happens: FC12: Login on tty0 works fine. On all other ttys login works but bash dies because of of -EIO. After vhangup() the tty returns -EIO upon read()/write(). FC16: Login broken on all ttys (bash dies with EIO like on FC12). If I start UML with rootfs read-only login works on tty0. - WTF?! Debian 6.0: Login works perfectly fine on all ttys Without tty_operations->hangup() the following happens: FC12: Login on tty0 works fine. mingetty is unable to start on anything else than tty0. It exits after a few seconds. FC16: Unable to start any mingetty (like on FC12 it exits after a few seconds) With read-only rootfs mingetty starts at least on tty0 and login works. (Again, WTF?) Debian 6.0 Login works perfectly fine on all ttys. I have no idea what's the root cause of this, there seems to be a lot of black magic involved. Alan, do you think the issues are caused by Fedora's broken user space? How can we fix this? Thanks, //richard (wearing a voodoo priests robe) [PATCH 1/6] TTY: um/line, add tty_port [PATCH 2/6] TTY: um/line, use tty from tty_port [PATCH 3/6] um: remove line_ioctl() [PATCH 4/6] um: Remove dead code [PATCH 5/6] um: fully use tty_port [PATCH 6/6] um: remove count_lock