On Tue, Jun 04, 2024 at 09:08:25AM -0700, Dave Hansen wrote: > On 5/28/24 02:55, Kirill A. Shutemov wrote: > > Keep track of the number of shared pages. This will allow for > > cross-checking against the shared information in the direct mapping > > and reporting if the shared bit is lost. > > It's probably also worth mentioning that conversions are slow and > relatively rare and even though a global atomic isn't really scalable, > it also isn't worth doing anything fancier. Okay, will do. > > diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c > > index 26fa47db5782..979891e97d83 100644 > > --- a/arch/x86/coco/tdx/tdx.c > > +++ b/arch/x86/coco/tdx/tdx.c > > @@ -38,6 +38,8 @@ > > > > #define TDREPORT_SUBTYPE_0 0 > > > > +static atomic_long_t nr_shared; > > Doesn't this technically need to be: > > static atomic_long_t nr_shared = ATOMIC_LONG_INIT(0); > > ? I thought we had some architectures where the 0 logical value wasn't > actually all 0's. Hm. I am not aware of such requirement. I see plenty uninitilized atomic_long_t in generic code. For instance, invalid_kread_bytes. And I doubt TDX will ever be built for non-x86 :P -- Kiryl Shutsemau / Kirill A. Shutemov
