Re: [PULL 02/10] pci-bridge/cxl_downstream: Add a CXL switch downstream port

[Jonathan Cameron via <qemu-devel@nongnu.org>]

On Sun, 4 Dec 2022 08:23:55 +0100
Thomas Huth <thuth@redhat.com> wrote:

> On 04/11/2022 07.47, Thomas Huth wrote:
> > On 16/06/2022 18.57, Michael S. Tsirkin wrote:  
> >> From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> >>
> >> Emulation of a simple CXL Switch downstream port.
> >> The Device ID has been allocated for this use.
> >>
> >> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> >> Message-Id: <20220616145126.8002-3-Jonathan.Cameron@huawei.com>
> >> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> >> Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
> >> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> >> ---
> >>   hw/cxl/cxl-host.c              |  43 +++++-
> >>   hw/pci-bridge/cxl_downstream.c | 249 +++++++++++++++++++++++++++++++++
> >>   hw/pci-bridge/meson.build      |   2 +-
> >>   3 files changed, 291 insertions(+), 3 deletions(-)
> >>   create mode 100644 hw/pci-bridge/cxl_downstream.c  
> > 
> >   Hi!
> > 
> > There is a memory problem somewhere in this new device. I can make QEMU 
> > crash by running something like this:
> > 
> > $ MALLOC_PERTURB_=59 ./qemu-system-x86_64 -M x-remote \
> >      -display none -monitor stdio
> > QEMU 7.1.50 monitor - type 'help' for more information
> > (qemu) device_add cxl-downstream
> > ./qemu/qom/object.c:1188:5: runtime error: member access within misaligned 
> > address 0x3b3b3b3b3b3b3b3b for type 'struct Object', which requires 8 byte 
> > alignment
> > 0x3b3b3b3b3b3b3b3b: note: pointer points here
> > <memory cannot be printed>
> > Bus error (core dumped)
> > 
> > Could you have a look if you've got some spare minutes?  
> 
> Ping! Jonathan, Michael, any news on this bug?
> 
> (this breaks one of my local tests, that's why it's annoying for me)
Sorry, my email filters ate your earlier message.

Looking into this now. I'll note that it also happens on
device_add xio3130-downstream so not specific to this new device.

So far all I've managed to do is track it to something rcu related
as failing in a call to drain_call_rcu() in qmp_device_add()

Will continue digging.

Jonathan


> 
>   Thomas
>