On Tue, Oct 15, 2024 at 01:25:29PM +0100, Joey Gouly wrote:
On Tue, Oct 15, 2024 at 12:41:16PM +0100, Will Deacon wrote:
if (system_supports_poe() && err == 0 && user->poe_offset) { ...
which gives the wrong impression that the POR is somehow optional, even if the CPU supports POE. So we should drop that check of 'user->poe_offset' as it cannot be NULL here.
That was cargo culted (by me) from the rest of the function (apart from TPIDR2 and FPMR). I think Kevin is planning on sending his signal changes still, but is on holiday, maybe he can remove the last part of the condition as part of his series.
That's there because the decisions about "should we save this thing" are taken in setup_sigframe_layout() and for a bunch of the extensions we suppress the saving if they're in some sort of default state (eg, when we don't have TIF_SVE set we don't output the SVE sigframe).