It would appear to me that the init process is running before the edits are made to the /etc/inittab file thus starting the shells attached to tty1 through tty6 but not starting the shell for ttyS0. When booting from the 2004.3 (or 2004.3-r1) livecd using a serial console, i.e. boot: gentoo-nofb console=ttyS0,115200 no shell is spawned for the serial console. A shell is started attached to the graphics console, and there is a line in /etc/inittab which should start the serial console shell c1:12345:respawn:/sbin/agetty -nl /bin/bashlogin 115200 ttyS0 vt100 but no shell is started. If the agetty command from the /etc/inittab file is executed, like this livecd root # /sbin/agetty -nl /bin/bashlogin 115200 ttyS0 vt100 the shell on the serial console is started correctly. I also noted that there are bash shells running attached to tty1 through tty6 even though the lines which would start those shells are commented out in /etc/inittab. # End of /etc/inittab #c1:12345:respawn:/sbin/agetty -nl /bin/bashlogin 38400 tty1 linux #c2:12345:respawn:/sbin/agetty -nl /bin/bashlogin 38400 tty2 linux #c3:12345:respawn:/sbin/agetty -nl /bin/bashlogin 38400 tty3 linux #c4:12345:respawn:/sbin/agetty -nl /bin/bashlogin 38400 tty4 linux #c5:12345:respawn:/sbin/agetty -nl /bin/bashlogin 38400 tty5 linux #c6:12345:respawn:/sbin/agetty -nl /bin/bashlogin 38400 tty6 linux c1:12345:respawn:/sbin/agetty -nl /bin/bashlogin 115200 ttyS0 vt100 Booting from the 2004.2 livecd behaves as expected. Reproducible: Always Steps to Reproduce: 1. boot from livecd 2. boot: gentoo-nofb console=ttyS0,115200 3. Actual Results: shell prompt appears on graphics console but not on serial console Expected Results: shell prompt should appear on serial console Dell Poweredge 2650 console redirection enabled console attached to first serial port
Hrrrmn... I'll try to see what is up with this and get a resolution. As for the comments, don't worry about that. We hacked up some stuff to get that going.
PING Does anyone know if this has been addressed for 2005.0?
It has not been addressed yet. If you have any ideas on how to resolve this, I'm definitely open for suggestions.
I'm afraid I do not know nearly enough about the LiveCD boot process to have any real suggestions. I did poke around enough to make the observations that were in the original bug report. Namely, it looks as though the /etc/inittab file is being edited correctly to start the serial console shell yet the init process is not starting it. I presume that either the init process is not paying attention to the file which is located at /etc/inittab after the system is running or that the init process is being started before the edits to the file are being made.
Just an update to note that the bug still exists in 2005.0
I dug a little deeper into this today using a 2005.0 boot CD. It is the /sbin/rc script which edits the /etc/inittab file. /sbin/rc calls the function livecd_fix_inittab which is defined in the file /sbin/livecd-functions.sh. The function livecd_fix_inittab adds a line to /etc/inittab to start a shell on the serial console. I noticed that if I commented out the line with id "c1" which livecd_fix_inittab added, ran "telinit q", and then un-commented the line, the console was terminated on /dev/vc/1 and started on the serial console as it should have at boot time. I also noticed that if I replaced the "c1" id field with "c9" on the line that livecd_fix_inittab added and then ran "telinit q" the serial console shell was also started. So there must be some confusion within init about the line which livecd_fix_inittab adds having the id field "c1". I would suggest changing livecd_fix_inittab so that the label on the line which starts the serial console is "c9" or even "c999" rather than "c1". Granted this is a work-around but it should get serial consoles working again when booting off of the live CDs until someone can figure out why init is confused about the line which live_cd_inttab adds with the "c1" id field.
Just curious (I have nothing with serial to verify) but what if it is c0 rather than c9? It shouldn't make a difference, correct? The reason why I ask, is I now have control over livecd-functions.sh in livecd-tools, so I can make changes without baselayout bumps. I've currently got it using c0 for the serial consoles, which should work. I beleive that Project Dolphin should be using this updated version of livecd-tools, but I can also understand your reluctance to download such a large ISO, if you have not already. If you're willing to help with more testing (you've been awesome already) then I can build a newer ISO image with the newer livecd-tools to see if this resolves the issue.
No response...
