On Wed, Jun 27, 2012 at 11:35 AM, Marc Zyngier <marc.zyngier at arm.com> wrote: > On 27/06/12 15:54, Christoffer Dall wrote: >> On Wed, Jun 27, 2012 at 6:04 AM, Marc Zyngier <marc.zyngier at arm.com> wrote: >>> On 27/06/12 10:47, Alexander Graf wrote: >>>> >>>> >>>> On 27.06.2012, at 11:23, Marc Zyngier <marc.zyngier at arm.com> wrote: >>>> >>>>> Hi Christoffer, >>>>> >>>>> On 26/06/12 23:24, Christoffer Dall wrote: >>>>>> Signed-off-by: Christoffer Dall <c.dall at virtualopensystems.com> >>>>>> --- >>>>>> arch/arm/kvm/interrupts.S | ? ?6 +++--- >>>>>> 1 file changed, 3 insertions(+), 3 deletions(-) >>>>>> >>>>>> diff --git a/arch/arm/kvm/interrupts.S b/arch/arm/kvm/interrupts.S >>>>>> index 625ba6c..d1a9b75 100644 >>>>>> --- a/arch/arm/kvm/interrupts.S >>>>>> +++ b/arch/arm/kvm/interrupts.S >>>>>> @@ -271,9 +271,9 @@ ENDPROC(__kvm_flush_vm_context) >>>>>> ?* (hardware reset value is 0) */ >>>>>> .macro set_hstr entry >>>>>> ? ?mrc ? ?p15, 4, r2, c1, c1, 3 >>>>>> - ? ?ldr ? ?r3, =0x8e00 >>>>>> + ? ?ldr ? ?r3, =0x00108e00 >>>>> >>>>> Where does this value comes from? In my copy of the ARM ARM, bits >>>>> 31:18,14,4 of HSTR are reserved. >> >> yes, complete brain damage, thanks for pointing this out. >> >>>> >>>> Speaking of which, any reason to not use proper constant #defines here, so that this number makes any sense at all? :) >>> >>> So you really want maintainable code? Heretic! ;-) >>> >> >> a little unsure what we should do with the CPACR (are there >> potentially other coprocessors that we need to take care of and >> switch, or...?) and correspondingly what to do with TCPAC, TASE, and >> TCPn (except for the VFP patch already in the works). If we trap all >> the coproc accesses, TASE, or TCPAC the guest dies on undefined >> exceptions (after the latest patch). Marc? > > I don't see why we should trap the access to CPACR, unless we want to > lie to the guest about which coprocessors are implemented, so I would be > inclined to leave TCPAC to 0. > > I don't immediately see why you'd die on an UNDEF, though. You should > end up in handle_exit() and hit the BUG() there... > that's what I really meant ;) >> what I really meant was, of course, this: > > Looks much better to me. > > ? ? ? ?M. > -- > Jazz is not dead. It just smells funny... >
