Re: [Qemu-devel] [PATCH 3/5] spapr_cpu_core: add missing rollback on realization path

[Greg Kurz]

On Fri, 15 Jun 2018 16:29:15 +1000
David Gibson <address@hidden> wrote:

> On Fri, Jun 15, 2018 at 07:58:05AM +0200, Greg Kurz wrote:
> > On Fri, 15 Jun 2018 10:14:31 +1000
> > David Gibson <address@hidden> wrote:
> >   
> > > On Fri, Jun 15, 2018 at 10:02:25AM +1000, David Gibson wrote:  
> > > > On Thu, Jun 14, 2018 at 11:50:42PM +0200, Greg Kurz wrote:    
> > > > > The spapr_realize_vcpu() function doesn't rollback in case of error.
> > > > > This isn't a problem with coldplugged CPUs because the machine won't
> > > > > start and QEMU will exit. Hotplug is a different story though: the
> > > > > CPU thread is started under object_property_set_bool() and it assumes
> > > > > it can access the CPU object.
> > > > > 
> > > > > If icp_create() fails, we return an error without unregistering the
> > > > > reset handler for this CPU, and we let the underlying QEMU thread for
> > > > > this CPU alive. Since spapr_cpu_core_realize() doesn't care to 
> > > > > unrealize
> > > > > already realized CPUs either, but happily frees all of them anyway, 
> > > > > the
> > > > > CPU thread crashes instantly:
> > > > > 
> > > > > (qemu) device_add host-spapr-cpu-core,core-id=1,id=gku
> > > > > GKU: failing icp_create (cpu 0x11497fd0)
> > > > >                              ^^^^^^^^^^
> > > > > Program received signal SIGSEGV, Segmentation fault.
> > > > > [Switching to Thread 0x7fffee3feaa0 (LWP 24725)]
> > > > > 0x00000000104c8374 in object_dynamic_cast_assert (obj=0x11497fd0,
> > > > >                                                   ^^^^^^^^^^^^^^
> > > > >                                              pointer to the CPU object
> > > > > 623         trace_object_dynamic_cast_assert(obj ? 
> > > > > obj->class->type->name
> > > > > (gdb) p obj->class->type
> > > > > $1 = (Type) 0x0
> > > > > (gdb) p * obj
> > > > > $2 = {class = 0x10ea9c10, free = 0x11244620,
> > > > >                                  ^^^^^^^^^^
> > > > >                               should be g_free
> > > > > (gdb) p g_free
> > > > > $3 = {<text variable, no debug info>} 0x7ffff282bef0 <g_free>
> > > > > 
> > > > > obj is a dangling pointer to the CPU that was just destroyed in
> > > > > spapr_cpu_core_realize().
> > > > > 
> > > > > This patch adds proper rollback to both spapr_realize_vcpu() and
> > > > > spapr_cpu_core_realize().
> > > > > 
> > > > > Signed-off-by: Greg Kurz <address@hidden>    
> > > > 
> > > > Applied to ppc-for-3.0, since it definitely looks to fix some
> > > > problems.    
> > > 
> > > Uh.. actually it has a definite bug - the first exit point will call
> > > g_free() on an uninitialized spapr_cpu.  I fixed it up with a NULL
> > > initialization in my tree.  
> > 
> > Ah... as said in the cover letter, all the series is based on machine_data
> > being set before the call to object_property_set_bool()... Maybe I should
> > have made that explicit with a preparatory patch... Sorry.  
> 
> Ah, that makes sense.
> 
> So, I ended up having to rework a little differently, after I yanked
> by intc -> machine_data patch because it broke things for clg.  I
> think I've fixed it up correctly now - if you can check the latest
> ppc-for-3.0 I pushed out, that would be great.
> 

I'll do this ASAP.

pgpyu7mG_pNu6.pgp
Description: OpenPGP digital signature