Re: [PATCH 2/2] hyperv/synic: Allocate as ram_device

[Michael S. Tsirkin]

On Thu, Jan 09, 2020 at 08:28:00AM -0500, Michael S. Tsirkin wrote:
> On Thu, Jan 09, 2020 at 01:22:42PM +0000, Dr. David Alan Gilbert wrote:
> > * Michael S. Tsirkin (address@hidden) wrote:
> > > On Thu, Jan 09, 2020 at 12:22:37PM +0000, Dr. David Alan Gilbert wrote:
> > > > * Michael S. Tsirkin (address@hidden) wrote:
> > > > > On Thu, Jan 09, 2020 at 12:08:20PM +0000, Dr. David Alan Gilbert 
> > > > > wrote:
> > > > > > * Michael S. Tsirkin (address@hidden) wrote:
> > > > > > > On Wed, Jan 08, 2020 at 01:53:53PM +0000, Dr. David Alan Gilbert 
> > > > > > > (git) wrote:
> > > > > > > > From: "Dr. David Alan Gilbert" <address@hidden>
> > > > > > > > 
> > > > > > > > Mark the synic pages as ram_device so that they won't be visible
> > > > > > > > to vhost.
> > > > > > > > 
> > > > > > > > Signed-off-by: Dr. David Alan Gilbert <address@hidden>
> > > > > > > 
> > > > > > > 
> > > > > > > I think I disagree with this one.
> > > > > > >  * A RAM device represents a mapping to a physical device, such 
> > > > > > > as to a PCI
> > > > > > >  * MMIO BAR of an vfio-pci assigned device.  The memory region 
> > > > > > > may be mapped
> > > > > > >  * into the VM address space and access to the region will modify 
> > > > > > > memory
> > > > > > >  * directly.  However, the memory region should not be included 
> > > > > > > in a memory
> > > > > > >  * dump (device may not be enabled/mapped at the time of the 
> > > > > > > dump), and
> > > > > > >  * operations incompatible with manipulating MMIO should be 
> > > > > > > avoided.  Replaces
> > > > > > >  * skip_dump flag.
> > > > > > > 
> > > > > > > Looks like an abuse of notation.
> > > > > > 
> > > > > > OK, it did feel a bit like that - any suggestions of another way to 
> > > > > > do
> > > > > > it?
> > > > > >   This clearly isn't normal RAM.
> > > > > > 
> > > > > > Dave
> > > > > 
> > > > > If it's just an optimization for vhost/postcopy/etc, then I think
> > > > 
> > > > Note it's not an optimisation; postcopy fails unless you can aggregate
> > > > the members of the hugepage.
> > > > And I think vhost-user will fail if you have too many sections - and
> > > > the 16 sections from synic I think will blow the slots available.
> > > 
> > > Right, so both are internal reasons.
> > > 
> > > > > an API that says how this isn't normal ram would be ok.
> > > > > E.g. it's not DMA'd into? Then maybe _nodma?
> > > > 
> > > > Do we want a new memory_region_init for that or just to be able to add
> > > > a flag?
> > > > 
> > > > Dave
> > > 
> > > I think a flag API is preferable since this can apply to any kind of
> > > region. But can go either way, Paolo's the maintainer there.
> > 
> > (Copying Paolo in)
> > So what exactly does this flag mean; to me it's 'no vhost' - but is it
> > actually more general?
> > 
> > Dave
> 
> I think it's also handy for VFIO, that should skip it too.

BTW if it's "per cpu" then that is another way to put it.
Neither vfio nor vhost have a concept of cpu so neither
can support accessing per cpu things.

> 
> > > > > > > 
> > > > > > > 
> > > > > > > > ---
> > > > > > > >  hw/hyperv/hyperv.c | 14 ++++++++------
> > > > > > > >  1 file changed, 8 insertions(+), 6 deletions(-)
> > > > > > > > 
> > > > > > > > diff --git a/hw/hyperv/hyperv.c b/hw/hyperv/hyperv.c
> > > > > > > > index da8ce82725..4de3ec411d 100644
> > > > > > > > --- a/hw/hyperv/hyperv.c
> > > > > > > > +++ b/hw/hyperv/hyperv.c
> > > > > > > > @@ -95,12 +95,14 @@ static void synic_realize(DeviceState *dev, 
> > > > > > > > Error **errp)
> > > > > > > >      msgp_name = g_strdup_printf("synic-%u-msg-page", vp_index);
> > > > > > > >      eventp_name = g_strdup_printf("synic-%u-event-page", 
> > > > > > > > vp_index);
> > > > > > > >  
> > > > > > > > -    memory_region_init_ram(&synic->msg_page_mr, obj, msgp_name,
> > > > > > > > -                           sizeof(*synic->msg_page), 
> > > > > > > > &error_abort);
> > > > > > > > -    memory_region_init_ram(&synic->event_page_mr, obj, 
> > > > > > > > eventp_name,
> > > > > > > > -                           sizeof(*synic->event_page), 
> > > > > > > > &error_abort);
> > > > > > > > -    synic->msg_page = 
> > > > > > > > memory_region_get_ram_ptr(&synic->msg_page_mr);
> > > > > > > > -    synic->event_page = 
> > > > > > > > memory_region_get_ram_ptr(&synic->event_page_mr);
> > > > > > > > +    synic->msg_page = qemu_memalign(qemu_real_host_page_size,
> > > > > > > > +                                    sizeof(*synic->msg_page));
> > > > > > > > +    synic->event_page = qemu_memalign(qemu_real_host_page_size,
> > > > > > > > +                                      
> > > > > > > > sizeof(*synic->event_page));
> > > > > > > > +    memory_region_init_ram_device_ptr(&synic->msg_page_mr, 
> > > > > > > > obj, msgp_name,
> > > > > > > > +                           sizeof(*synic->msg_page), 
> > > > > > > > synic->msg_page);
> > > > > > > > +    memory_region_init_ram_device_ptr(&synic->event_page_mr, 
> > > > > > > > obj, eventp_name,
> > > > > > > > +                           sizeof(*synic->event_page), 
> > > > > > > > synic->event_page);
> > > > > > > >  
> > > > > > > >      g_free(msgp_name);
> > > > > > > >      g_free(eventp_name);
> > > > > > > > -- 
> > > > > > > > 2.24.1
> > > > > > > 
> > > > > > --
> > > > > > Dr. David Alan Gilbert / address@hidden / Manchester, UK
> > > > > 
> > > > --
> > > > Dr. David Alan Gilbert / address@hidden / Manchester, UK
> > > 
> > --
> > Dr. David Alan Gilbert / address@hidden / Manchester, UK