[jnordwick]

It isn't the cache side channel that is the important part of meltdown. Flush-reload is just out the information is retrived and ask acceptable problem.

The problem is data hitting the cache when it comes from an unreadable page. Meltdown looks like a bug in the cache hit logic because the page information is already in the tlb, and the fix is probably fairly trivial.

If the unreadabld page doesnt hit cache or is never mapped there in the first place, spectre can only read its own process.