Re: [Xen-devel] [PATCH for-next 07/16] xen/arm: Introduce copy_to_guest_phys_flush_dcache

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Thu, 23 Nov 2017, Julien Grall wrote:
> Hi Andrew,
> 
> On 23/11/17 18:49, Andrew Cooper wrote:
> > On 23/11/17 18:32, Julien Grall wrote:
> > > This new function will be used in a follow-up patch to copy data to the
> > > guest
> > > using the IPA (aka guest physical address) and then clean the cache.
> > > 
> > > Signed-off-by: Julien Grall <julien.grall@xxxxxxxxxx>
> > > ---
> > >   xen/arch/arm/guestcopy.c           | 10 ++++++++++
> > >   xen/include/asm-arm/guest_access.h |  6 ++++++
> > >   2 files changed, 16 insertions(+)
> > > 
> > > diff --git a/xen/arch/arm/guestcopy.c b/xen/arch/arm/guestcopy.c
> > > index be53bee559..7958663970 100644
> > > --- a/xen/arch/arm/guestcopy.c
> > > +++ b/xen/arch/arm/guestcopy.c
> > > @@ -110,6 +110,16 @@ unsigned long raw_copy_from_guest(void *to, const
> > > void __user *from, unsigned le
> > >                         COPY_from_guest | COPY_linear);
> > >   }
> > >   +unsigned long copy_to_guest_phys_flush_dcache(struct domain *d,
> > > +                                              paddr_t gpa,
> > > +                                              void *buf,
> > > +                                              unsigned int len)
> > > +{
> > > +    /* P2M is shared between all vCPUs, so the vCPU used does not matter.
> > > */
> > 
> > Be very careful with this line of thinking.  It is only works after
> > DOMCTL_max_vcpus has succeeded, and before that point, it is a latent
> > NULL pointer dereference.
> 
> I really don't expect that function been used before DOMCT_max_vcpus is set.
> It is only used for hardware emulation or Xen loading image into the hardware
> domain memory. I could add a check d->vcpus to be safe.
> 
> > 
> > Also, what about vcpus configured with alternative views?
> 
> It is not important because the underlying call is get_page_from_gfn that does
> not care about the alternative view (that function take a domain in
> parameter). I can update the comment.
 
Since this is a new function, would it make sense to take a struct
vcpu* as parameter, instead of a struct domain* ?

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel