On Sun, Dec 29, 2019 at 03:17:40PM +0000, Marc Zyngier wrote: > On Fri, 27 Dec 2019 13:01:57 +0000, > Alexandru Elisei <alexandru.elisei@xxxxxxx> wrote: > > On 12/20/19 3:05 PM, Mark Rutland wrote: > > > + // PSTATE.PAN is unchanged unless overridden by SCTLR_ELx.SPAN > > > + // See ARM DDI 0487E.a, page D5-2578. > > > + new |= (old & PSR_PAN_BIT); > > > + if (sctlr & SCTLR_EL1_SPAN) > > > + new |= PSR_PAN_BIT; > > > > On page D13-3264, it is stated that the PAN bit is set unconditionally if > > SCTLR_EL1.SPAN is clear, not set. > > Indeed. Given that when ARMv8.1-PAN is not implemented, SCTLR_EL1[23] > is RES1, it seems surprising to force PAN based on this bit being set. > > I've now dropped this series from my tree until Mark has a chance to > clarify this. Sorry for the mess; I'm fixing up the patches now and looking out for any similar mistakes. I'll try to have a v2 out by the end of today. Thanks, Mark.