AMD K6-2 166Mhz, 512MB ram, desktop - TC11.1 runs fine TC12 fails.

[gadget42]

AMD K6-2 166Mhz, 512MB ram, desktop -  TC11.1 runs fine TC12 fails.

TC11.1 uses 5.4.3 kernel (i586)
TC12.0 uses 5.10.3 kernel (i686)

Have searched the forum but didn't see anything specific to this issue.

This machine is the oldest in the collection and the "canary in the coal mine" regarding operating systems.

What do I need to do/get to compile the newer kernel to still be usable with our older machines? Thanks!

[Rich]

Hi gadget42
TC12 fails isn't much to go on. Perhaps if you provide a more thorough description of what is happening we can
figure out the problem.

What type of media are you booting from?
If you are booting from a USB device, did you use the tc-install utility or a 3rd party program? Did you include the waitusb
boot code?
Are you seeing any error messages?
Does it get to a command prompt?

[gadget42]

attempting to boot TC12 using a proven cd-r and get this on screen:

Loading /boot/vmlinuz.......
Loading /boot/core.gz............
Wrong EFI loader signature.
early console in extract_kernel
input_data: 0x0071d0ae
input_len: 0x0049a1d1
output: 0x00100000
output_len: 0x009be788
kernel_total_size: 0x00ad4000
needed_size: 0x00ad4000

Decompressing Linux... Parsing ELF... No relocation needed... done.
Booting the kernel.
Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004
Kernel Offset: disabled
Rebooting in 60 seconds..
_

*and it stops there until the power switch is turned off...

during this testing the hard drive with Win3.1/Win3.11/TC11.1 is removed.

thanks!

[Rich]

Hi gadget42
A couple of other forum members have had the same issue:
http://forum.tinycorelinux.net/index.php/topic,25011.0.html
http://forum.tinycorelinux.net/index.php/topic,25081.0.html

Try the following boot codes one at a time to see if they make things any better:

Code: [Select]

acpi=off  pci=noacpi  acpi=noirq  pnpacpi=off  noapic  nolapic

[gadget42]

Hi Rich, read those forum posts and tried the boot codes(yes, one at a time) but nothing has changed so far.

adding "debug" boot code gave much additional "verbosity" and browsing through the paused digital camera
video footage revealed the following traps:

traps: modprobe[84] trap invalid opcode ip:b7f34081 sp:bf94447c error:0 in ld-2.32.so[b7f1d000+18000]

traps: modprobe[85] trap invalid opcode ip:b7f53081 sp:bfadf3ec error:0 in ld-2.32.so[b7f3c000+18000]

traps: init[1] trap invalid opcode ip:b7ef5081 sp:bfa66acc error:0 in ld-2.32.so[b7ede000+18000]

also read about seemingly similar issues others were having:

https://forums.gentoo.org/viewtopic-t-1089838-start-0.html

https://unix.stackexchange.com/questions/448937/is-ld-so-an-executable

http://forum.tinycorelinux.net/index.php?topic=21272.0

Thanks!

[curaga]

Oh, so libc in 12.x needs more than a 686. I believe it's getting harder to get a proper 486 build.

[gadget42]

Oh, so libc in 12.x needs more than a 686. I believe it's getting harder to get a proper 486 build.

hi curaga,
it would appear so. i was going to look for differences between TC11.1 libc and TC12 libc but haven't yet. i also seem to recall reading about other distros that have started compiling kernels without support for the "antique" hardware. at least it will force me to learn more so that i can continue to enjoy these machines.

any suggestions and/or pointers? thanks!

[Juanito]

glibc was compiled for tc-12.x with:

Code: [Select]

$ echo "CFLAGS += -fomit-frame-pointer -march=i486 -mtune=i686 -Os -pipe" > configparms
..and for tc-11.x with

Code: [Select]

$ echo "CFLAGS += -march=i486 -mtune=i686 -Os -pipe" > configparms
However, in addition, for glibc in tc-11.x, uname_hack was used to force i486, but could not be used for glibc in tc-12.x because it does not compile for linux-5.x kernels.

[andyj]

I've seen this in some extensions too. Mariadb and dhcp require -march=i686 or they won't compile.

[Juanito]

I think we’re looking at the opposite problem - that it was asked to compile for i486 and actually compiled for something else without giving any error or warning.

[curaga]

gadget42, you could try compiling the newer glibc on TC 11 on that old box. Then replacing the files incl. ld-so inside the TC 12 initrd with the ones you built. It may take many hours with an old cpu.

Instructions for doing this should be under the TC 12 web tree, release/x86/src.

However, the K6 has 3dnow. If the glibc build tunes to the host cpu, libc built on a k6 may not run on newer processors (since they lack 3dnow).

[gadget42]

Huge Thanks to Rich, curaga, Juanito, and andyj for all your helpful insights!

We have decided to re-evaluate our criteria for the machines we maintain.

If it boots TC12 it stays and if it fails then it goes. So this machine will go.

[Rich]

Hi gadget42
You could use an older version of Tinycore and turn the machine into a file server. Install a couple of 320 Gig hard drives
for storage. Boot from a CD to install to a small boot partition (1 or 2 Gig) on one of the drives. Tell the BIOS the first
drive is only 1 or 2 Gig and the second drive does not exist. When the kernel boots, it will detect all of the drive space
since it figures it out instead of relying on the BIOS.

[Juanito]

I played around with uname_hack and glibc to see if it made a difference - you could try to boot 12.x after replacing the existing rootfs.gz with the file here:

http://tinycorelinux.net/12.x/x86/release/src/toolchain/test_i486/

[gadget42]

I played around with uname_hack and glibc to see if it made a difference - you could try to boot 12.x after replacing the existing rootfs.gz with the file here:

http://tinycorelinux.net/12.x/x86/release/src/toolchain/test_i486/

Thanks for your efforts! I did make a new boot cd with your modified rootfs.gz and it boots in the next machine in line(a compaq presario 566mhz machine),
but for some reason the amd k6-2 machine has become flaky with reading some cd disks but not others. I disassembled it earlier today and will reuse some
of the components and the case to freshen up another machine. Again, thanks!