PROBLEM: kernel/init messages don't go to serial console

nickbrown

I have followed the widely available Remote-Serial-Console Howto and I can happily see the bootloader (lilo) messages over the serial connection, as the initial section of lilo.conf below has been setup correctly.
I can also succesfully login into the machine over the serial connection as I've added the appropriate agetty line to my inittab file (see below).
But even though I believe I have the correct append options for the kernel in my lilo.conf file (see below) and I have the correct kernel compile options (see below) I cannot see either the kernel boot messages or inits boot messages. "dmesg" confirms that the kernel received the "console" option (see below). It would appear however that it is not using this serial console as I still see the boot/init messages on the vga console.
Below is the output I see on the serial console. (lilo prompt, followed by load linux message then nothing until the login prompt)

Does anyone have any ideas what is wrong and how to solve this?

The problem;

BonezTheGoon · Posted: Fri Apr 25, 2003 12:36 pm Post subject:

Moving this thread to the duplicates forum and locking in favor of the previous topic posted here.

Please do not cross post one topic into multiple forums, please choose one forum to post it in. Later if you would like a moderator to move the thread into another forum please PM one of us and we will be happy to help.

Thanks,
BonezTheGoon

BonezTheGoon · Posted: Fri Apr 25, 2003 8:43 pm Post subject:

Per nickbrown's request I am moving this thread into Hardware and Laptops from the Duplicates forum and I will move the other thread linked above that is currently in Hardware and Laptops from there to Duplicate Threads. I am unlocking this thread and will lock the other given there is more complete and current information in this thread (according to nickbrown.)

Regards,
BonezTheGoon

SmegTheLight · Posted: Fri Apr 25, 2003 11:10 pm Post subject:

I also had a LOT of problem trying to get serial ONLY console to work.
There is a lot of conflicting info out there too.

It is now working for me - BUT many things cause it to not work.. Sometimes it works on some hardware, and not on others.

Here is what I suggest :
1) Get it working with a plain vanilla kernel WITHOUT VT console selected. This will force the kernel to put out to the serial port. Also make this the bare minimum that you need from the kernel to make your machine run (No ACPI, No PPP, No Netfilter, etc..)

This will prove that everything else your are doing is working (Lilo, etc.).
For me, patches to the Frame Buffer code in the Gentoo (and Wolk) kernels prevented the serial console from working.

2) On the kernel line (Grub, Lilo, etc.) I only needed "console=ttyS0,9600" (plus all your regular settings)

Some guides and papers have you put it in twice with an upper and lower case "console".. ie "CONSOLE=" and "console="..

I am using it now, and it is working with only the line above (plus regular settings, root=, etc..)

3) After you prove that it works, put back the VT Support

nickbrown · Posted: Sun Apr 27, 2003 1:36 am Post subject:

BonezTheGoon · Posted: Sun Apr 27, 2003 3:42 am Post subject:

Please post your bug report in this thread so that others can find it easily should they need to check on it or submit information regarding it.

Thanks,
BonezTheGoon

SmegTheLight · Posted: Sun Apr 27, 2003 7:45 am Post subject:

I am pretty sure that the bug is related to something related to frame buffer support patches not in the standard kernel (as of 2.4.20).

Well, "pretty sure" might be a bit strong.. I was very tired, and had be flying through unfamiliar kernel code at warp speed for an entire weekend trying to ressolve this problem. I didn't document my discoveries, and all that remains in my mind is a gut feeling that it was the Frame Buffer stuff. It is as if the console= iine is getting ignored.

When I compile a plain vanilla standard kernel, I can have both the VT and serial support turned on. Turning off the VT in the popular "Patched" kernels simply forces the kernel to have no other choice (and / or bypasses the patched buggy code)

With that I started adding in the patches I needed (Scheduler, O(1), MPPE/MPPC, IPSEC, etc). It's all good now. None of those patches messed it up. I now have my framebuffer graphics, plain old penguin logo, and serial ONLY kernel messages. I even have an agetty running on VT2 and VT3. VT1 is consumed by my GDK-Framebuffer app.

nickbrown

See bug report;
https://bugs.gentoo.org/show_bug.cgi?id=20025

pfeifer · Posted: Thu May 01, 2003 2:18 pm Post subject:

has nothing to do with framebuffer patches.

you can get the patch to fix from the bug report nickbrown lists above. i have attached the fix there. it'll be in the pfeifer/gentoo-sources in the next day. i will retro it to gentoo-sources-2.4.20-r2/r3 and pfeifer-sources-2.4.20_pre8 for now.

Regards,

Jay Pfeifer
Gentoo Linux Developer
Core System/Kernel

SmegTheLight · Posted: Fri May 02, 2003 1:18 am Post subject:

Yeah, I was watching the bugthread.

It was awhile ago for me when I was debugging it. It was probably a lot of coincidences all combining to lead me to belive it had to do with the fb patches.
(I was also having a lot of problems with fb logos, etc.. at the time).

I wonder what other code is/was failing because of -O3 ?