On November 6, 2024 10:15:13 AM PST, Jakub Jelinek <ja...@redhat.com> wrote: >On Wed, Nov 06, 2024 at 10:03:25AM -0800, H. Peter Anvin wrote: >> The issue is that we want the frame pointer chain to be maintained, even >> across alternatives. > >If the current function doesn't have frame pointer set up yet (or is in the >epilogue after it got restored already), then the chain is still maintained, >just one frame will be missing from it. I don't see why that would be a big >deal. > >> This isn't to quiet a spurious warning, rather >> objtool is throwing a legitimate warning that a level environment >> invariant is being violated. > >Anyway, do you have a testcase where you saw problematic inline asm being >moved into the prologue or after the epilogue or is that just theoretical? >So that we can test if "redzone" clobber prevents that or if "=m" (dummy) >does or similar? >The rsp register var in output is something we should warn about at least, >it isn't really something kernel should keep using. > > Jakub >
Not theoretical. I don't have an example *right now* because I'm about to step out the door for an internal trip.
