On 06.03.2018 18:28, Mark Cave-Ayland wrote:
> On 06/03/18 16:48, Thomas Huth wrote:
>
>> Something in the recent commits introduced a new way to cause unexpected
>> aborts of QEMU:
>>
>> $ ppc64-softmmu/qemu-system-ppc64 -monitor stdio
>> QEMU 2.11.50 monitor - type 'help' for more information
>> (qemu) device_add macio-newworld
>> Unexpected error in qemu_chr_fe_init() at
>> /home/thuth/devel/qemu/chardev/char-fe.c:222:
>> Device 'serial0' is in use
>> Aborted (core dumped)
>>
>> Of course it does not make sense to add a macio-newworld device on the
>> pseries machine, but QEMU should not abort in this case - it should just
>> print an error message and continue afterwards. Any ideas how to fix
>> this?
>
> So the backtrace from git master looks like this:
>
> Thread 1 "qemu-system-ppc" received signal SIGABRT, Aborted.
> __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
>
>
> 51 ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
>
>
> (gdb) bt
>
>
> #0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
>
> #1 0x00007fffdbd6e3fa in __GI_abort () at abort.c:89
>
>
> #2 0x0000555555de6d86 in error_handle_fatal (errp=0x555556bdfb90
> <error_abort>, err=0x555556ef5a00) at util/error.c:38
>
> #3 0x0000555555de6eb6 in error_setv (errp=0x555556bdfb90 <error_abort>,
> src=0x555556031ad0 "chardev/char-fe.c", line=222, func=0x555556031c50
> <__func__.18713> "qemu_chr_fe_init", err_class=ERROR_CLASS_GENERIC_ERROR,
> fmt=0x555556031b50 "Device '%s' is in use", ap=0x7fffffffd010,
> suffix=0x0) at util/error.c:71
>
> #4 0x0000555555de7097 in error_setg_internal (errp=0x555556bdfb90
> <error_abort>, src=0x555556031ad0 "chardev/char-fe.c", line=222,
> func=0x555556031c50 <__func__.18713> "qemu_chr_fe_init",
> fmt=0x555556031b50 "Device '%s' is in use") at util/error.c:95
>
>
> #5 0x0000555555d673fc in qemu_chr_fe_init (b=0x555557aef590,
> s=0x555556de2290, errp=0x555556bdfb90 <error_abort>) at
> chardev/char-fe.c:222
> #6 0x0000555555af5467 in set_chr (obj=0x555557aeef80, v=0x555557960c20,
> name=0x555555f65fc5 "chrA", opaque=0x555556658410 <escc_properties+432>,
> errp=0x555556bdfb90 <error_abort>) at hw/core/qdev-properties-system.c:216
> #7 0x0000555555cb326a in object_property_set (obj=0x555557aeef80,
> v=0x555557960c20, name=0x555555f65fc5 "chrA", errp=0x555556bdfb90
> <error_abort>) at qom/object.c:1109
> #8 0x0000555555cb6232 in object_property_set_qobject
> (obj=0x555557aeef80, value=0x555557960bf0, name=0x555555f65fc5 "chrA",
> errp=0x555556bdfb90 <error_abort>) at qom/qom-qobject.c:27
> #9 0x0000555555cb32af in object_property_set_str (obj=0x555557aeef80,
> value=0x555556de23c0 "serial0", name=0x555555f65fc5 "chrA",
> errp=0x555556bdfb90 <error_abort>) at qom/object.c:1117
> #10 0x0000555555af5d94 in qdev_prop_set_chr (dev=0x555557aeef80,
> name=0x555555f65fc5 "chrA", value=0x555556de2290) at
> hw/core/qdev-properties-system.c:427
> #11 0x0000555555b350a8 in macio_instance_init (obj=0x555557aec3c0) at
> hw/misc/macio/macio.c:347
>
>
> It seems that the error is being raised when setting the property rather
> than during realize so I'm not sure what I can do to handle this. Any
> thoughts?
Does the device need to be hot-pluggable or even user_creatable at all?
It seems like it is also using serial_hds[] directly, so that is a good
indication that it is *not* user creatable. So maybe the easiest fix is
to simply set
dc->user_creatable = false;
in macio_class_init() ?
Thomas
