On Sunday 27 November 2005 11:31, Blaisorblade wrote: > On Sunday 27 November 2005 05:48, murthy andukuri wrote: > > I can bring up a uml session successfully when > > invoking from command line. and can also bring up a > > uml session successfully from within a gdb (following > > instructions from the 'how to'). However, I am having > > trouble being able to input login info to the same uml > > when started thru umlgdb. All of these attempts are > > with the skas patch on the host. host and guest are > > both 2.6.12.1. > > > > can somebody please tell me what I am missing ? More > > details below. > > A workaround (for sure) is to make UML open its console not on fd:0,fd:1. > Either you move the login prompt away from tty0 and to ttyN, or you use > con0=xterm (or con0=your console channel).
I wanted to run a script that did UML with the output piped to "tee". Unfortunately, the standard "tee" will stop outputting to a target if it ever gets pipe full even once, and with the disk I/O problem you have when /tmp isn't tmpfs, that would usually happen to the xterm after about 5 minutes. So what I wound up doing was: (./build.sh > out.txt 2>&1 &) && tail -f out.txt Except that this didn't work because the detached ./build.sh would die as soon as it ran UML, because if UML gets EOF on stdin on its way up, it refuses to run. My solution? (cat /dev/zero | ./build.sh > out.txt 2>&1 &) && tail -f out.txt This works fine as long as nothing ever tries to read from stdin. (When I have a config file wrong and busybox/uclibc/linux runs "make oldconfig" it spits nulls and ctrl-g to the output file in a tight loop until it fills up the disk. But oh well...) Rob > I don't know for sure, but the problem is likely to be this: > - reading from stdin *requires* that you're passing UML a real tty, not a > normal fd. Nope, /dev/zero works fine for me. > - umlgdb redirects UML stdin to a pipe It probably closes stdin. > - things don't work. UML dies if stdin gives EOF on startup. (I vaguely remember catting /dev/null to it instead of /dev/zero and it didn't work. This was a while ago so I'm guessing...) > > Thru umlgdb, I am able to get the uml boot up to start > > in the right tty (set as a part of the initgdb method > > in umlgdb). But, if I type the user name at the login > > prompt, I get the response > > It's not succeeding in catching the input, but it's only throwing out > messages. UML's stdio console handling has always been fairly brittle. I banged on stdout, but never got around to poking at stdin because I could work around it. > > In the bootup log messages, I see a few 'TIOCSCTTY > > failed', errno =1. . . > > That's the "it requires a tty, not a pipe" thing. It does not. Try the /dev/null thing with init being some sort of script that echoes "hello world" or some such. > I guess it's parsing /etc/inittab, even if I don't understand how... Possibly the busybox inittab? The format for that is different... > anybody looking for /dev/ttyvc/0 is seriously confused... probably it's > splicing together /dev/tty and some field (first or last one) in > /etc/inittab (which is a bit buggy anyway). That's a mixture of the devfs and non-devfs syntaxes. Devfs has /dev/vc/0 and a real system has /dev/tty0. Rob -- Steve Ballmer: Innovation! Inigo Montoya: You keep using that word. I do not think it means what you think it means. ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click _______________________________________________ User-mode-linux-user mailing list User-mode-linux-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user
