Re: [Qemu-devel] PPC* and Sparc32 crash

[Blue Swirl]

On Thu, Aug 25, 2011 at 7:31 PM, Avi Kivity <address@hidden> wrote:
> On 08/25/2011 08:33 PM, Blue Swirl wrote:
>>
>> (gdb) bt
>> #0  0x00007ffff3da2165 in *__GI_raise (sig=<value optimized out>)
>>     at ../nptl/sysdeps/unix/sysv/linux/raise.c:64
>> #1  0x00007ffff3da4f70 in *__GI_abort () at abort.c:92
>> #2  0x00007ffff3d9b2b1 in *__GI___assert_fail (
>>     assertion=0x5c7390 "info->qdev.size>= sizeof(SysBusDevice)",
>>     file=<value optimized out>, line=0x9c,
>>     function=0x5c73c0 "sysbus_register_withprop") at assert.c:81
>> #3  0x000000000052e000 in sysbus_register_withprop (info=<value optimized
>> out>)
>>     at /src/qemu/hw/sysbus.c:156
>> #4  0x000000000041d9c3 in module_call_init (type=<value optimized out>)
>>     at /src/qemu/module.c:78
>> #5  0x0000000000518b64 in main (argc=<value optimized out>,
>>     argv=<value optimized out>, envp=<value optimized out>)
>>     at /src/qemu/vl.c:3241
>> (gdb) up 3
>> #3  0x000000000052e000 in sysbus_register_withprop (info=<value optimized
>> out>)
>>     at /src/qemu/hw/sysbus.c:156
>> 156         assert(info->qdev.size>= sizeof(SysBusDevice));
>> (gdb) p *info
>> Cannot access memory at address 0x0
>> (gdb) p info
>> $1 =<value optimized out>
>
> -O1 fixes that.
>
>> (gdb) up
>> #4  0x000000000041d9c3 in module_call_init (type=<value optimized out>)
>>     at /src/qemu/module.c:78
>> 78              e->init();
>> (gdb) p *e
>> $2 = {type = MODULE_INIT_BLOCK,
>>   init = 0x53cf50<sun4c_intctl_register_devices>, node = {
>>     tqe_next = 0x1051130, tqe_prev = 0x10510b0}}
>
> static SysBusDeviceInfo sun4c_intctl_info = {
>    .init = sun4c_intctl_init1,
>    .qdev.name  = "sun4c_intctl",
>    .qdev.size  = sizeof(Sun4c_INTCTLState),
>    .qdev.vmsd  = &vmstate_sun4c_intctl,
>    .qdev.reset = sun4c_intctl_reset,
> };
>
> typedef struct Sun4c_INTCTLState {
>    SysBusDevice busdev;
>    ...
> }
>
> so the code looks fine.  Can you dig deeper?  Did something stomp on
> qdev.size?

Bah, I run 'make clean', recompiled and the problem disappeared. Sorry
for the trouble and noise. Bugs with Makefile dependencies?