# 08:33:43 |
smartin |
joins #crosstool-ng |
# 16:11:48 |
alan_o |
joins #crosstool-ng |
# 16:15:21 |
alan_o |
Hi, I'm having a problem with a ct-ng-based system. It's x86 and PC-104, so it should be easy, but I'm stuck. I can build and run the Linux kernel, but any program I try to run as /init fails with general protection. I know that it's a vague description of the problem, but I'm not sure what else is relevant. I'm running ct-ng 1.14.0 and have selected all the packages as the latest (except for the kernel). I'm mostly right now looki |
# 16:15:21 |
alan_o |
ng for someone to point me in the right direction as to what I can do to solve this. I've googled for this problem and haven't found much, but I could be just searching for the wrong things. |
# 16:15:29 |
alan_o |
anyway, any help is appreciated. |
# 16:16:05 |
alan_o |
I'm note even sure ct-ng is the right place to be asking |
# 16:19:41 |
alan_o |
Actually, I just tried a re-build of the kernel, and I think I may have it.... |
# 16:20:14 |
alan_o |
I rebuilt the kernel (3.1) after changing the kernel version in ct-ng to be an older one than that (3.0.x) |
# 16:20:27 |
alan_o |
ok now to build busybox again..... |
# 17:49:48 |
alan_o |
looks like it all works now. |
# 17:50:44 |
alan_o |
It would appear that when I build the kernel, I have to have CT-NG configured for that kernel version or older, or else programs built with that same CT-NG will not run under the built kernel. |
# 17:52:28 |
y_morin |
joins #crosstool-ng |
# 18:26:46 |
y_morin |
is now known as: y_morin|away |
# 18:36:52 |
y_morin|away |
is now known as: y_morin |
# 19:26:22 |
y_morin |
is now known as: y_morin|away |
# 19:34:23 |
kos_tom |
alan_o: can you pastebin your ct-ng .config file ? |
# 19:44:32 |
smartin |
quits |
# 19:50:57 |
alan_o |
kos_tom: sure |
# 19:53:27 |
alan_o |
http://pastebin.com/u4TwrLpU |
# 19:59:51 |
kos_tom |
do you pass special flags to gcc when building your userspace side ? |
# 20:16:35 |
y_morin |
joins #crosstool-ng |
# 20:19:42 |
y_morin|away |
quits : Ping timeout: 252 seconds |
# 21:37:43 |
y_morin |
alan_o: NPTL will not work for i486. NPTL requires instructions that only came with i686 (maybe i586, but I'm not sure). |
# 21:37:54 |
y_morin |
kos_tom: hello! How was FOSDEM? |
# 21:38:18 |
y_morin |
alan_o: you will have to use LinuxThreads for anything i486. |
# 21:39:21 |
y_morin |
alan_o: As for the kernel version, the running kernel must be supported by the glibc: |
# 21:39:49 |
y_morin |
So, either one of these three conditions is enough: |
# 21:39:57 |
y_morin |
1) the same kernel headers are used |
# 21:40:09 |
y_morin |
2) a more recent kernel is running |
# 21:40:25 |
y_morin |
2b) or an older kernel is used to build the toolchain |
# 21:40:37 |
y_morin |
3) the glibc is ocnfigured with back-ward compatibility shims |
# 21:47:32 |
kos_tom |
y_morin: hello! FOSDEM was too short, as usual. |
# 21:47:46 |
kos_tom |
and the way back to Toulouse was a bit complicated by the snow in Toulouse. |
# 21:48:00 |
y_morin |
kos_tom: too short, I can understand! ;-) |
# 21:48:03 |
kos_tom |
our flight on Sunday evening was canceled, we had to wait Monday morning... but we had 10 customers waiting in Toulouse for a training session... |
# 21:48:44 |
y_morin |
Ooopss.. Hopefully, they understood you were doing some kind of "technology survey", hence your beinf late! ;-) |
# 21:48:53 |
y_morin |
*being |
# 21:50:19 |
y_morin |
kos_tom: I ca never properly remember the requirements for NPTL. Was the minimum CPU supported i486, i586, or i686? |
# 21:50:29 |
y_morin |
*can never... |
# 21:54:03 |
kos_tom |
I also never remember. |
# 21:54:27 |
y_morin |
Sigh... The only docs I could find wrt NPTL design are dated back to 2003-2005... |
# 21:54:37 |
kos_tom |
I think it uses cmpxchg |
# 21:54:42 |
kos_tom |
so I would be tempted to say i486 |
# 21:54:44 |
kos_tom |
but I'm unsure. |
# 21:54:53 |
y_morin |
kos_tom: Ah, yes, that's the instruction! Thanks! |
# 21:55:22 |
kos_tom |
http://sourceware.org/bugzilla/show_bug.cgi?id=849#c4 |
# 21:55:32 |
kos_tom |
seems to confirm that >= i486 is the requirement for nptl |
# 21:56:32 |
y_morin |
kos_tom: Thanks! It would indeed yield a toolchain compile error, not a runtime error. |
# 21:57:44 |
y_morin |
kos_tom: further assesed by the first sentence there: https://en.wikipedia.org/wiki/Intel_486#Improvements |
# 21:59:18 |
y_morin |
remembers that time, when a *fast* CPU would run at a mere 33MHz clock... Sigh, he's getting old... :-] |
# 22:35:52 |
y_morin |
is now known as: y_morin|away |
# 22:47:01 |
y_morin|away |
is now known as: y_morin |
# 23:48:34 |
y_morin |
quits : Quit: Night! |
# 23:49:10 |
y_morin |
joins #crosstool-ng |
# 23:57:17 |
y_morin |
quits : Quit: Night! (yes, for real, this time! ;-) |