On Thu, 2 Apr 2020 14:25:30 +0100
Anthony PERARD <anthony.per...@citrix.com> wrote:

> On Thu, Apr 02, 2020 at 02:29:25PM +0200, Igor Mammedov wrote:
> > On Mon, 30 Mar 2020 17:52:48 +0100
> > Anthony PERARD <anthony.per...@citrix.com> wrote:
> >   
> > > On Fri, Mar 27, 2020 at 06:48:28AM -0400, Igor Mammedov wrote:  
> > > > Since bd457782b3b0 ("x86/pc: use memdev for RAM") Xen
> > > > machine fails to start with:
> > > >    qemu-system-i386: xen: failed to populate ram at 0
> > > > 
> > > > The reason is that xen_ram_alloc() which is called by
> > > > memory_region_init_ram(), compares memory region with
> > > > statically allocated 'global' ram_memory memory region
> > > > that it uses for RAM, and does nothing in case it matches.
> > > > 
> > > > While it's possible feed machine->ram to xen_ram_alloc()
> > > > in the same manner to keep that hack working, I'd prefer
> > > > not to keep that circular dependency and try to untangle that.
> > > > 
> > > > However it doesn't look trivial to fix, so as temporary
> > > > fixup opt out Xen machine from memdev based RAM allocation,
> > > > and let xen_ram_alloc() do its trick for now.
> > > > 
> > > > Reported-by: Anthony PERARD <anthony.per...@citrix.com>
> > > > Signed-off-by: Igor Mammedov <imamm...@redhat.com>    
> > > 
> > > Reviewed-by: Anthony PERARD <anthony.per...@citrix.com>
> > > 
> > > That should work on most configs. But we also sometime use the "pc"
> > > machine with accel=xen, to run without the "xen-platform" pci device,
> > > but that would be less common.  
> > 
> > does following work for you in case of pc machine?
> > 
> > diff --git a/hw/xen/xen-common.c b/hw/xen/xen-common.c
> > index 15650d7f6a..f19c0883ae 100644
> > --- a/hw/xen/xen-common.c
> > +++ b/hw/xen/xen-common.c
> > @@ -151,6 +151,8 @@ static void xen_setup_post(MachineState *ms, AccelState 
> > *accel)
> >  
> >  static int xen_init(MachineState *ms)
> >  {
> > +    MachineClass *mc = MACHINE_GET_CLASS(ms);
> > +
> >      xen_xc = xc_interface_open(0, 0, 0);
> >      if (xen_xc == NULL) {
> >          xen_pv_printf(NULL, 0, "can't open xen interface\n");
> > @@ -170,6 +172,10 @@ static int xen_init(MachineState *ms)
> >          return -1;
> >      }
> >      qemu_add_vm_change_state_handler(xen_change_state_handler, NULL);
> > +    /*
> > +     * opt out of system RAM being allocated by generic code
> > +     */
> > +    m->default_ram_id = NULL;
> >      return 0;  
> 
> After fixing the build issues, it does work, yes. I've tested both "xenfv"
> and "pc,accel=xen".

thanks, I'll post a formal patch

however it's all workarounds,
we need to fix ram allocation properly later on
so far I only have questions, hope you can help with clarifying them

1. why xen uses memory_region_init_ram() which does not allocate anything
   can it use plain memory_region_init() instead?
2. how main ram is allocated?
3. code has
           /*                                                                   
    
         * Xen does not allocate the memory continuously, it keeps a            
 
         * hole of the size computed above or passed in.                        
 
         */                                                                     
 
        block_len = (1ULL << 32) + x86ms->above_4g_mem_size; 
   which fixes up size ram memory region
   can we allocate 1 memory region of ram_size and then
   alias lower and upper memory instead of that?
4. how RAM migration works in case of xen?

> 
> Build issue:
> - I've added #include "hw/boards.h"
> - and s/m->/mc->/
ccccccgndiej> gngfiehlivuil

> Thanks!nbchihheikhjlnervve
> 


Reply via email to