Loïc Minier wrote: > Arnaud Patard looked a bit into this today; we both reproduce in QEMU. Awesome. I just recovered my nslu2 by flashing firmware from an installation of stable. Not ideal, but at least my internet gateway in the cabin is back to using 3 watts rather than the temporary laptop I had sucking down my battery bank. My nslu2 had the same run-init related problem: ��rEXT3-fs: mounted filesystem with ordered data mode. �i done. Begin: Running /scripts/init-bottom ... �303: run-init: Unknowo error 582636 �tempted to kill init! _ 19.580000] [<c002af98>] (unwind_backtrace+0x0/0xdc) from [<c025b180>] (panic+0x34/0x128) /�4/0x128) from [<c0045b58>] (do_exit+0x6c/0x670) [ 19.600000] [<c0045b58>] (do_exit+0x6c/0x670) from [<c00461e4>] (do_group_exit+0x88/0xbc) [ 19.610000] [<c00461e4>] (do_group_exit+0x88/0xbc) from [<c004622c>] (sys_exit_group+0x14/0x24) [ 19.610000] [<c004622c>] (sys_exit_group+0x14/0x24) from [<c0024ec0>] (ret_fast_syscall+0x0/0x28) > It seems to be an issue with busybox or the toolchain; replacing > /bin/sh in the initrd with bash works fine. Starting from your initrd > and a bash binary for armel do: > mkdir tmp > cd tmp > gunzip -c < ../initrd.img-2.6.38-1-ixp4xx | cpio -i > mv bin/sh bin/sh-broken > cp ../bash bin/sh > find . | cpio -o -H newc | gzip -c >../initrd.img-2.6.38-1-ixp4xx-new > > If I spawn a busybox shell with exec /bin/sh </dev/console > >/dev/console, I will get random results when running "exec run-init > /root /sbin/init", it even booted once. So, run-init is a klibc binary, not busybox. Somehow busybox shell is failing to run it correctly, so that run-init later fails? > busybox was uploaded multiple times recently, so it could be fixed with > latest busybox and toolchain. I'll check that out. Possibly.. My nslu2 had busybox 1:1.18.3-1. -- see shy jo
Attachment:
signature.asc
Description: Digital signature