[Qemu-devel] qemu-fast 0.6.0 hanging...

[Damien Mascord]

Hi,

I am trying to get qemu 0.6.0 running on Debian testing, running vanilla 
2.4.26 kernel.

If I use the qemu 0.6.0 binaries from the qemu website, qemu-fast and 
qemu binaries do in fact load the kernel as expected, though both hang 
upon shutdown.

Though, if I compile the 0.6.0 from source on the host itself, I see the 
following behaviour:

If I run qemu without the "-fast" affix, then it will load the 
linux-test image (as provided on the qemu website), though on shutdown, 
it hangs and uses 99-100% CPU.

If I run qemu-fast, then it hangs on startup, and uses 99-100% CPU 
straight away.

I am using gcc (GCC) 3.3.4 (Debian 1:3.3.4-3), with libc6 2.3.2.

I am compiling with the following options:

CFLAGS="-Wall -g -O2" ./configure --target-list="i386-user i386 
i386-softmmu arm-user sparc-user ppc-user ppc-softmmu" --prefix=/usr 
--enable-slirp
Install prefix    /usr
BIOS directory    /usr/share/qemu
binary directory  /usr/bin
Manual directory  /usr/share/man
ELF interp prefix /usr/gnemul/qemu-%M
Source path       /home/tusker/qemu-0.6.0
C compiler        gcc
make              make
host CPU          i386
host big endian   no
target list       i386-user i386 i386-softmmu arm-user sparc-user 
ppc-user ppc-softmmu
gprof enabled     no
static build      no
SDL support       yes
SDL static link   yes
mingw32 support   no

If I change the -O2 to -O0, it fails to compile.  If I change to -O1 it 
compiles correctly, though still behaves like -O2. (ie, hangs on startup 
for qemu-fast, hangs on shutdown for qemu.)

The compile error for -O0 is as follows:

gcc -Wall -g -O0 -fno-strict-aliasing -fomit-frame-pointer -I. 
-I/home/tusker/qemu-0.6.0/target-i386 -I/home/tusker/qemu-0.6.0 
-I/home/tusker/qemu-0.6.0/linux-user 
-I/home/tusker/qemu-0.6.0/linux-user/i386 -D_GNU_SOURCE 
-I/home/tusker/qemu-0.6.0/slirp -c -o cpu-exec.o 
/home/tusker/qemu-0.6.0/cpu-exec.c
/home/tusker/qemu-0.6.0/exec-all.h: In function `testandset':
/home/tusker/qemu-0.6.0/exec-all.h:413: error: can't find a register in 
class `GENERAL_REGS' while reloading `asm'
make[2]: *** [cpu-exec.o] Error 1
make[2]: Leaving directory `/home/tusker/qemu-0.6.0/i386-user'
make[1]: *** [all] Error 1
make[1]: Leaving directory `/home/tusker/qemu-0.6.0'
make: *** [build-stamp] Error 2

Is there an easy way to debug what is going wrong here?

According to the user documention, attaching with gdb, and then issueing 
a few commands through gdb should help, but I'm not very familiar with 
gdb... :)

For interest sake, here is the following when normal qemu is shutting down:

1) On the console it says:

flushing ide devices: hda hdc
Restarting system.

2) gdb 'info reg' shows:
eax            0x909d   37021
ecx            0x30001  196609
edx            0x80     128
ebx            0x80     128
esp            0x8ff8   0x8ff8
ebp            0x0      0x0
esi            0x9d     157
edi            0xffe4   65508
eip            0xff53   0xff53
eflags         0x86     134
cs             0xf000   61440
ss             0x9000   36864
ds             0x9000   36864
es             0x9000   36864
fs             0x9000   36864
gs             0x9000   36864

3) gdb 'x/10i $eip' shows:
0xff53: (bad)
0xff54: jbe    0xff55
0xff56: add    %al,(%eax)
0xff58: aad    $0x3
0xff5a: outsb  %ds:(%esi),(%dx)
0xff5b: call   *%ebp
0xff5d: aas
0xff5e: aad    $0x3
0xff60: push   %eax
0xff61: add    %dl,%ah

For the qemu-fast binary, gdb can't seem to attach to anything, and the 
following message comes up:

Interrupted while waiting for the program.
Give up (and stop debugging it)? (y or n) n


Is there anything I can do to work out what is going on ?

Damien

--
Damien Mascord (tusker at tusker dot org)
GPG key 2CB181BE / 93B2 EF21 0C7C F022 F467  7966 219E 92B3 2CB1 81BE