On Thu, Apr 01 2021 at 09:37, Will Deacon wrote: > On Wed, Mar 31, 2021 at 01:54:52PM -0700, Kees Cook wrote: >> Hi Will (and Mark and Catalin), >> >> Can you take this via the arm64 tree for v5.13 please? Thomas has added >> his Reviewed-by, so it only leaves arm64's. :) > > Sorry, these got mixed up in my inbox so I just replied to v7 and v8 and > then noticed v9. Argh! > > However, I think the comments still apply: namely that the dummy "=m" looks > dangerous to me > https://lore.kernel.org/r/20210401083034.GA8554@willie-the-truck Hrmpf, didn't think about that. > and I think you're accessing pcpu variables with preemption enabled > for the arm64 part: That's my fault. On x86 this is invoked right after coming up into C code and _before_ enabling interrupts, which I why I suggested not to use the wrapped one. raw_cpu_read() should be fine everywhere. Thanks, tglx
