Re: NanoNote debootstrap debugging
The nano note's CPU does NOT have a fp. So it needs the kernel to
emulate the fp instruction.
yajin
http://vm-kernel.org
2010/3/26 Phil Endecott <spam_from_debian_mips@chezphil.org>:
> Hello again,
>
> I'm still trying to debootstrap my NanoNote. I have the feeling that there
> is something fairly fundamental that is preventing this from working. Maybe
> there is a MIPS expert out there who understands this properly?
>
> The NanoNote ships with OpenWRT. I have a debootstrapped tree on a microSD
> card.
>
> The symptom I have is that executables from my Debian chroot just hang using
> 100% CPU when run. When I run them under gdb I see something like this:
>
> # gdb /usr/sbin/chroot
> (gdb) run /card /bin/ls
> {100% CPU observed in top in another terminal}
> ^C
> Program received signal SIGINT, Interrupt.
> 0x2aac2d00 in ?? ()
> (gdb) disas 0x2aac2cf0 0x2aac2d10
> Dump of assembler code from 0x2aac2cf0 to 0x2aac2d10:
> 0x2aac2cf0: lui gp,0x2
> 0x2aac2cf4: addiu gp,gp,-7392
> 0x2aac2cf8: addu gp,gp,t9
> 0x2aac2cfc: move v0,a0
> 0x2aac2d00: swc1 $f20,56(a0)
> 0x2aac2d04: swc1 $f21,60(a0)
> 0x2aac2d08: swc1 $f22,64(a0)
> 0x2aac2d0c: swc1 $f23,68(a0)
>
> It always hangs at the same address in all the programs I've tried (ls,
> bash, echo, ...), except for ldconfig, which I believe is the only
> statically-linked program; that stops at a different address which
> disassembles to:
>
> 0x00441530 in ?? ()
> (gdb) disas 0x00441520 0x00441540
> Dump of assembler code from 0x441520 to 0x441540:
> 0x00441520: lui v0,0x1c3
> 0x00441524: ori v0,v0,0xffff
> 0x00441528: and a0,a0,v0
> 0x0044152c: lui v0,0xfe3c
> 0x00441530: cfc1 v1,$31
> 0x00441534: nop
> 0x00441538: and v1,v1,v0
> 0x0044153c: or v1,v1,a0
>
> That "swc1 $f20" looks like a floating-point instruction. Similarly "cfc1
> v1,$31".
>
> My guess is that this is the first FP instruction that it finds, and that
> there is either a kernel or hardware incompatibility issue. Does the CPU in
> my NanoNote have FP? Is there some sort of kernel emulation or other
> support requirement?
>
>
> Thanks for any suggestions.
>
>
> Phil.
>
>
>
>
>
> --
> To UNSUBSCRIBE, email to debian-mips-REQUEST@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact
> listmaster@lists.debian.org
> Archive:
> 1269603366089@dmwebmail.dmwebmail.chezphil.org">http://lists.debian.org/1269603366089@dmwebmail.dmwebmail.chezphil.org
>
>
Reply to: