Re: [PATCH v6 03/14] KVM: X86: Don't track dirty for KVM_SET_[TSS_ADDR|IDENTITY_MAP_ADDR]

[Peter Xu]

On Wed, Mar 11, 2020 at 10:09:40AM -0700, Sean Christopherson wrote:
> On Wed, Mar 11, 2020 at 12:39:06PM -0400, Peter Xu wrote:
> > On Wed, Mar 11, 2020 at 09:10:04AM +0800, kbuild test robot wrote:
> > > Hi Peter,
> > > 
> > > Thank you for the patch! Perhaps something to improve:
> > > 
> > > [auto build test WARNING on tip/auto-latest]
> > > [also build test WARNING on vhost/linux-next linus/master v5.6-rc5 next-20200310]
> > > [cannot apply to kvm/linux-next linux/master]
> > > [if your patch is applied to the wrong git tree, please drop us a note to help
> > > improve the system. BTW, we also suggest to use '--base' option to specify the
> > > base tree in git format-patch, please see https://stackoverflow.com/a/37406982]
> > > 
> > > url:    https://github.com/0day-ci/linux/commits/Peter-Xu/KVM-Dirty-ring-interface/20200310-070637
> > > base:   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 12481c76713078054f2d043b3ce946e4814ac29f
> > > reproduce:
> > >         # apt-get install sparse
> > >         # sparse version: v0.6.1-174-g094d5a94-dirty
> > >         make ARCH=x86_64 allmodconfig
> > >         make C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__'
> > > 
> > > If you fix the issue, kindly add following tag
> > > Reported-by: kbuild test robot <lkp@xxxxxxxxx>
> > > 
> > > 
> > > sparse warnings: (new ones prefixed by >>)
> > > 
> > >    arch/x86/kvm/x86.c:2599:38: sparse: sparse: incorrect type in argument 1 (different address spaces) @@    expected void const [noderef] <asn:1> * @@    got  const [noderef] <asn:1> * @@
> > >    arch/x86/kvm/x86.c:2599:38: sparse:    expected void const [noderef] <asn:1> *
> > >    arch/x86/kvm/x86.c:2599:38: sparse:    got unsigned char [usertype] *
> > >    arch/x86/kvm/x86.c:7501:15: sparse: sparse: incompatible types in comparison expression (different address spaces):
> > >    arch/x86/kvm/x86.c:7501:15: sparse:    struct kvm_apic_map [noderef] <asn:4> *
> > >    arch/x86/kvm/x86.c:7501:15: sparse:    struct kvm_apic_map *
> > > >> arch/x86/kvm/x86.c:9794:31: sparse: sparse: incorrect type in return expression (different address spaces) @@    expected void [noderef] <asn:1> * @@    got n:1> * @@
> > 
> > I'm not sure on how I can reproduce this locally, and also I'm not
> > very sure I understand this warning.  I'd be glad to know if anyone
> > knows...
> > 
> > If without further hints, I'll try to remove the __user for
> > __x86_set_memory_region() and use a cast on the callers next.
> 
> Ah, it's complaining that the ERR_PTR() returns in __x86_set_memory_region()
> aren't explicitly casting to a __user pointer.
> 
> Part of me wonders if something along the lines of your original approach
> of keeping the "int" return and passing a "void __user **p_hva" would be
> cleaner overall, as opposed to having to cast everywhere.  The diff would
> certainly be smaller.  E.g.
> 
> int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size,
> 			    void __user **p_hva)
> {
> 	...
> 
> 	if (p_hva)
> 		*p_hva = (void __user *)hva;
> 
>         return 0;
> }

Returning an adress still has some advantage on one less param.  To
avoid going back and forth, I defined ERR_PTR_USR() and used there to
fix the sparse warnings.  Thanks,

-- 
Peter Xu