On Mon, May 10, 2021 at 9:43 AM Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > On Sun, May 09, 2021 at 06:22:05PM -0700, Florian Fainelli wrote: > > This does not qualify as a regression in that it has never worked for > > the specific platform that I have shown above until your 3 commits came > > in and fixed that particular FDT placement. To me this qualifies as a > > bug fix, and given that the 3 (now 4) commits applied without hunks, it > > seems reasonable to me to back port those to stable. > > As this isn't a regression, why not just use 5.12 on these platforms? > Why is 5.4 and 5.10 needed? Actually I think it *is* a regression, but not a common one. The bug that Ard is fixing can appear when the kernel grows over a certain size. If a user compile in a new set of functionality and the kernel size reach a tripping point so that the DTB ends up just outside the 1:1 lowmem map, disaster strikes. This has been a long standing mysterious bug for people using attached device trees. Yours, Linus Walleij
