On Mon, Jun 05, 2023 at 12:18:21PM -0700, Dave Hansen wrote:
> On 6/2/23 07:26, Tom Lendacky wrote:
> >> So this is a change in this version. If tdx_accept_memory() fails,
> >> you'll report unknown platform. Wouldn't it be better to have an error
> >> message that indicates a failure in the accept path?
> >>
> >
> > Maybe you can keep it similar to the v12 version with just a new error
> > message, something like:
> >
> > if (early_is_tdx_guest()) {
> > if (!tdx_accept_memory(start, end))
> > error("TDX error accepting memory\n");
> > } else {
> > error("Cannot accept memory: unknown platform\n");
> > }
>
> In the end, these errors aren't plumbed out to the page allocator. They
> *need* to succeed or we are dead anyway. Should we just send a fatal
> error up to the TDX module when we fail to accept memory? It's
> _slightly_ less opaque than plowing into an unaccepted page.
This is decompressor's error()s which are fatal.
arch_accept_memory() in the main kernel uses panic() in the same spot.
--
Kiryl Shutsemau / Kirill A. Shutemov
