On 03/29/2016 10:04 PM, Christian Seiler wrote:
> Unfortunately, apt-get update fails with illegal instruction in this
> image (but other binaries such as bash work). QEMU_CPU is set
> correctly, just in case you were wondering.
Yeah, I just tried to reproduce it and it affects the latest
qemu git snapshot as well.
> (sid-powerpcspe-sbuild)root@...:/# apt-get --help
> apt 1.0.9.9 for powerpcspe compiled on Apr 28 2015 23:27:12
> Usage: apt-get [options] command
> apt-get [options] install|remove pkg1 [pkg2 ...]
> apt-get [options] source pkg1 [pkg2 ...]
> [...]
> This APT has Super Cow Powers.
>
> (This means that apt-get in principle works.)
And it means that the CPU type is set correctly. Otherwise you
wouldn't have been able to chroot into the base system.
> If I run qemu-ppc-static with -strace on that binary, I get the following
> before the invalid instruction message:
>
> 2540 execve("/usr/lib/apt/methods/http",{"/usr/lib/apt/methods/http",NULL}) = 0x00000001 ([5,],[],[],NULL)
> 2531 read(5,0xf6fbe68c,64000) = 64
> 2531 stat64("/var/lib/apt/lists/192.168.178.15:3142_debian-ports_dists_sid_InRelease",0xf6fce060) = -1 errno=2 (No such file or directory)
> 2531 stat64("/var/lib/apt/lists/partial/192.168.178.15:3142_debian-ports_dists_sid_InRelease.reverify",0xf6fce060) = -1 errno=2 (No such file or directory)
> 2531 stat64("/var/lib/apt/lists/192.168.178.15:3142_debian_dists_sid_InRelease",0xf6fce060) = -1 errno=2 (No such file or directory)
> 2531 stat64("/var/lib/apt/lists/partial/192.168.178.15:3142_debian_dists_sid_InRelease.reverify",0xf6fce060) = -1 errno=2 (No such file or directory)
> 2531 gettimeofday(-151198432,0,0,1919236640,1695156746,2) = 0
> 2531 gettimeofday(-151198424,0,0,1919236640,1695156746,2) = 0
> 2531 _newselect(9,[5,],[8,],[],{0,500000}) = 0x00000001 ([],[8,],[],{0,499998})
> 2531 write(8,0xf669df8c,4979) = 4979
> 2531 gettimeofday(-151201560,0,-160863132,0,0,0) = 0
> Invalid instruction
That's where I suspected it to fail.
> Don't know if that could help or not. I only changed the APT mirror
> to my local apt-cacher-ng instance, and copied in qemu-ppc-static to
> /usr/bin, nothing else, the rest of the image is identical to your
> tarball. Of course, this could be a bug in QEMU... (I'm using the
> version from jessie-backports btw.)
Yeah, all that is according to documentation, so it really seems like
a bug to me. It's actually possible that the instruction pointer
contains non-sense and points to an address which actually doesn't
contain a valid instruction. Or the opcode table might have been
corrupted when running a multi-threaded application.
We have seen similar issues with the m68k emulation in qemu where
a bug caused the opcode table to be destroyed when a second
thread was started, so it's absolutely possible that qemu has
a bug here.
> Maybe the best thing is to just wait a few weeks until more stuff is
> compiled in sid and try bootstrapping again... I'm very sorry, but
> unless you can give me a tarball with gcc5 that works with qemu-user
> or an image (with gcc5 + ssh) + a kernel that works with qemu-system,
> I'm afraid that I don't think I'm using my time productively.
No worries. There isn't much you can do at the moment. The bug
in qemu needs to be traced down and fixed first.
> I'll gladly help with porting the package itself once I have a
> working base setup for this.
Thanks.
> PS: Just a heads-up: I had submitted a porterbox request to nm@d.o
> a while back for dietlibc, for a couple of platforms, but ppcspe
> was among them (I only started this thread later on after I was told
> in IRC that ppcspe may possible not have any porterboxes yet). myon
> forwarded my request to DSA today, so if they contact you w.r.t. it
> don't be surprised.
Ok, thanks for the heads up! So far, I didn't receive any request.
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@debian.org
`. `' Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
Attachment:
signature.asc
Description: OpenPGP digital signature