Re: [Qemu-devel] [PULL 00/47] ppc-for-2.10 queue 20170424

[Cédric Le Goater]

On 04/24/2017 04:38 PM, Cédric Le Goater wrote:
> On 04/24/2017 03:12 PM, Peter Maydell wrote:
>> On 24 April 2017 at 02:58, David Gibson <address@hidden> wrote:
>>> The following changes since commit 32c7e0ab755745e961f1772e95cac381cc68769d:
>>>
>>>   Merge remote-tracking branch 
>>> 'remotes/juanquintela/tags/migration/20170421' into staging (2017-04-21 
>>> 15:59:27 +0100)
>>>
>>> are available in the git repository at:
>>>
>>>   git://github.com/dgibson/qemu.git tags/ppc-for-2.10-20170424
>>>
>>> for you to fetch changes up to 4cab48942a1c5353f0a314fab1aa85a5f0a61461:
>>>
>>>   target/ppc: Style fixes (2017-04-24 08:56:19 +1000)
>>>
>>> ----------------------------------------------------------------
>>> ppc patch queue 2017-04-24
>>>
>>> Here's my first pull request for qemu-2.10, consisting of assorted
>>> patches which have accumulated while qemu-2.9 stabilized.  Highlights
>>> are:
>>>     * Rework / cleanup of the XICS interrupt controller
>>>     * Substantial improvement to the 'powernv' machine type
>>>         - Includes an MMIO XICS version
>>>     * POWER9 support improvements
>>>         - POWER9 guests with KVM
>>>         - Partial support for POWER9 guests with TCG
>>>     * IOMMU and VFIO improvements
>>>     * Assorted minor changes
>>>
>>> There are several IPMI patches here that aren't usually in my area of
>>> maintenance, but there isn't a regular maintainer and these patches
>>> are for the benefit of the powernv machine type.
>>
>>
>> Hi -- I'm afraid this pullreq generates a new clang sanitizer warning
>> running 'make check':
>>
>> /home/petmay01/linaro/qemu-for-merges/hw/core/loader.c:67:15: runtime
>> error: null pointer passed as argument 1, which is declared to never
>> be null
>>
>> It looks like we try to call open(NULL, ...).
>>
>> Affected tests are in check-qtest-i386:
>> tests/bios-tables-test
>>   /i386/acpi/piix4/ipmi
>>   /i386/acpi/q35/ipmi
>> tests/ipmi-kcs-test
>>
>> Here's a backtrace:
>>
>> #0  0x00005555567e6197 in get_image_size (filename=0x0) at
>> /home/petmay01/linaro/qemu-for-merges/hw/core/loader.c:67
>> #1  0x00005555568ead3c in ipmi_fru_init (fru=0x55555ab98588)
>>     at /home/petmay01/linaro/qemu-for-merges/hw/ipmi/ipmi_bmc_sim.c:1902
>> #2  0x00005555568e91ac in ipmi_sim_realize (dev=<optimised out>,
>> errp=<optimised out>)
>>     at /home/petmay01/linaro/qemu-for-merges/hw/ipmi/ipmi_bmc_sim.c:1943
>> #3  0x00005555567d6df6 in device_set_realized (obj=<optimised out>,
>> value=<optimised out>, errp=<optimised out>)
>>     at /home/petmay01/linaro/qemu-for-merges/hw/core/qdev.c:905
>> #4  0x0000555556d7b1f8 in property_set_bool (obj=0x55555ab93c90,
>> v=<optimised out>, name=<optimised out>, opaque=<optimised out>,
>> errp=0x7fffffffde90) at
>> /home/petmay01/linaro/qemu-for-merges/qom/object.c:1860
>> #5  0x0000555556d7dd1d in object_property_set_qobject (obj=0x0,
>> value=<optimised out>, name=0x58fdf801 <error: Cannot access memory at
>> address 0x58fdf801>, errp=0x1) at
>> /home/petmay01/linaro/qemu-for-merges/qom/qom-qobject.c:27
>> #6  0x0000555556d78cb2 in object_property_set_bool
>> (obj=0x55555ab93c90, value=<optimised out>, name=0x55555710e087
>> "realized", errp=0x7fffffffde90) at
>> /home/petmay01/linaro/qemu-for-merges/qom/object.c:1163
>> #7  0x000055555661d80d in qdev_device_add (opts=<optimised out>,
>> errp=<optimised out>)
>>     at /home/petmay01/linaro/qemu-for-merges/qdev-monitor.c:630
>> #8  0x00005555566374cb in device_init_func (opaque=<optimised out>,
>> opts=0x0, errp=0x58fdf801)
>>     at /home/petmay01/linaro/qemu-for-merges/vl.c:2305
>> #9  0x000055555701dc81 in qemu_opts_foreach (list=<optimised out>,
>> func=<optimised out>, opaque=<optimised out>, errp=<optimised out>) at
>> /home/petmay01/linaro/qemu-for-merges/util/qemu-option.c:1114
>> #10 0x0000555556633289 in main (argc=<optimised out>, argv=<optimised
>> out>, envp=<optimised out>)
>>     at /home/petmay01/linaro/qemu-for-merges/vl.c:4583
>>
>> It looks like the device doesn't handle not having its fru.filename
>> property set -- this should either cause an error in realize or the
>> rest of the device code should handle NULL.
> 
> I missed that one beacause the IPMI tests are not run on ppc64. I will 
> work on a fix and include the IPMI tests under ppc64 also.

I could not reproduce the error with the clang environment I am 
using (f24). Nevertheless, I have sent a couple of patches which 
should fix the problem.

Thanks,

C.