On Fri, 18 May 2018 19:28:36 +0800 Baoquan He <b...@redhat.com> wrote:
> > Note that it's not KASLR specific: if we had some other kernel feature that > > tried > > to allocate a piece of memory from what appears to be perfectly usable > > generic RAM > > we'd have the same problems! > > Hmm, this may not be the situation for 1GB huge pages. For 1GB huge > pages, the bug is that on KVM guest with 4GB ram, when user adds > 'default_hugepagesz=1G hugepagesz=1G hugepages=1' to kernel > command-line, if 'nokaslr' is specified, they can get 1GB huge page > allocated successfully. If remove 'nokaslr', namely KASLR is enabled, > the 1GB huge page allocation failed. Let me clarify that this issue is not specific to KVM in any way. The same issue happens on bare-metal, but if you have lots of memory you'll hardly notice it. On the other hand, it's common to create KVM guests with a few GBs of memory. In those guests, you may not be able to get a 1GB hugepage at all if kaslr is enabled. This series is a simple fix for this bug. It hooks up into already existing KASLR code that scans memory regions to be avoided. The memory hotplug issue is left for another day. Now, if I understand what Ingo is saying is that he wants to see all problems solved with a generic solution vs. a specific solution for each problem.