On Tue, Dec 01, 2015 at 03:01:16PM +0000, Marc Zyngier wrote:
> On 01/12/15 14:47, Christoffer Dall wrote:
> > On Tue, Dec 01, 2015 at 01:06:31PM +0000, Marc Zyngier wrote:
> >> On 01/12/15 12:56, Christoffer Dall wrote:
> >>> On Fri, Nov 27, 2015 at 06:50:02PM +0000, Marc Zyngier wrote:
> >>>> Implement the debug save restore as a direct translation of
> >>>> the assembly code version.
> >>>>
> >>>> Signed-off-by: Marc Zyngier <marc.zyng...@arm.com>
> >>>> ---
> >>>> arch/arm64/kvm/hyp/Makefile | 1 +
> >>>> arch/arm64/kvm/hyp/debug-sr.c | 130
> >>>> ++++++++++++++++++++++++++++++++++++++++++
> >>>> arch/arm64/kvm/hyp/hyp.h | 9 +++
> >>>> 3 files changed, 140 insertions(+)
> >>>> create mode 100644 arch/arm64/kvm/hyp/debug-sr.c
> >>>>
> >>>> diff --git a/arch/arm64/kvm/hyp/Makefile b/arch/arm64/kvm/hyp/Makefile
> >>>> index ec94200..ec14cac 100644
> >>>> --- a/arch/arm64/kvm/hyp/Makefile
> >>>> +++ b/arch/arm64/kvm/hyp/Makefile
> >>>> @@ -6,3 +6,4 @@ obj-$(CONFIG_KVM_ARM_HOST) += vgic-v2-sr.o
> >>>> obj-$(CONFIG_KVM_ARM_HOST) += vgic-v3-sr.o
> >>>> obj-$(CONFIG_KVM_ARM_HOST) += timer-sr.o
> >>>> obj-$(CONFIG_KVM_ARM_HOST) += sysreg-sr.o
> >>>> +obj-$(CONFIG_KVM_ARM_HOST) += debug-sr.o
> >>>> diff --git a/arch/arm64/kvm/hyp/debug-sr.c
> >>>> b/arch/arm64/kvm/hyp/debug-sr.c
> >>>> new file mode 100644
> >>>> index 0000000..a0b2b99
> >>>> --- /dev/null
> >>>> +++ b/arch/arm64/kvm/hyp/debug-sr.c
> >>>> @@ -0,0 +1,130 @@
> >>>> +/*
> >>>> + * Copyright (C) 2015 - ARM Ltd
> >>>> + * Author: Marc Zyngier <marc.zyng...@arm.com>
> >>>> + *
> >>>> + * This program is free software; you can redistribute it and/or modify
> >>>> + * it under the terms of the GNU General Public License version 2 as
> >>>> + * published by the Free Software Foundation.
> >>>> + *
> >>>> + * This program is distributed in the hope that it will be useful,
> >>>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> >>>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> >>>> + * GNU General Public License for more details.
> >>>> + *
> >>>> + * You should have received a copy of the GNU General Public License
> >>>> + * along with this program. If not, see <http://www.gnu.org/licenses/>.
> >>>> + */
> >>>> +
> >>>> +#include <linux/compiler.h>
> >>>> +#include <linux/kvm_host.h>
> >>>> +
> >>>> +#include <asm/kvm_mmu.h>
> >>>> +
> >>>> +#include "hyp.h"
> >>>> +
> >>>> +#define read_debug(r,n) read_sysreg(r##n##_el1)
> >>>> +#define write_debug(v,r,n) write_sysreg(v, r##n##_el1)
> >>>> +
> >>>> +#define save_debug(ptr,reg,nr)
> >>>> \
> >>>> + switch (nr) {
> >>>> \
> >>>> + case 15: ptr[15] = read_debug(reg, 15);
> >>>> \
> >>>> + case 14: ptr[14] = read_debug(reg, 14);
> >>>> \
> >>>> + case 13: ptr[13] = read_debug(reg, 13);
> >>>> \
> >>>> + case 12: ptr[12] = read_debug(reg, 12);
> >>>> \
> >>>> + case 11: ptr[11] = read_debug(reg, 11);
> >>>> \
> >>>> + case 10: ptr[10] = read_debug(reg, 10);
> >>>> \
> >>>> + case 9: ptr[9] = read_debug(reg, 9);
> >>>> \
> >>>> + case 8: ptr[8] = read_debug(reg, 8);
> >>>> \
> >>>> + case 7: ptr[7] = read_debug(reg, 7);
> >>>> \
> >>>> + case 6: ptr[6] = read_debug(reg, 6);
> >>>> \
> >>>> + case 5: ptr[5] = read_debug(reg, 5);
> >>>> \
> >>>> + case 4: ptr[4] = read_debug(reg, 4);
> >>>> \
> >>>> + case 3: ptr[3] = read_debug(reg, 3);
> >>>> \
> >>>> + case 2: ptr[2] = read_debug(reg, 2);
> >>>> \
> >>>> + case 1: ptr[1] = read_debug(reg, 1);
> >>>> \
> >>>> + default: ptr[0] = read_debug(reg, 0);
> >>>> \
> >>>> + }
> >>>> +
> >>>> +#define restore_debug(ptr,reg,nr)
> >>>> \
> >>>> + switch (nr) {
> >>>> \
> >>>> + case 15: write_debug(ptr[15], reg, 15);
> >>>> \
> >>>> + case 14: write_debug(ptr[14], reg, 14);
> >>>> \
> >>>> + case 13: write_debug(ptr[13], reg, 13);
> >>>> \
> >>>> + case 12: write_debug(ptr[12], reg, 12);
> >>>> \
> >>>> + case 11: write_debug(ptr[11], reg, 11);
> >>>> \
> >>>> + case 10: write_debug(ptr[10], reg, 10);
> >>>> \
> >>>> + case 9: write_debug(ptr[9], reg, 9);
> >>>> \
> >>>> + case 8: write_debug(ptr[8], reg, 8);
> >>>> \
> >>>> + case 7: write_debug(ptr[7], reg, 7);
> >>>> \
> >>>> + case 6: write_debug(ptr[6], reg, 6);
> >>>> \
> >>>> + case 5: write_debug(ptr[5], reg, 5);
> >>>> \
> >>>> + case 4: write_debug(ptr[4], reg, 4);
> >>>> \
> >>>> + case 3: write_debug(ptr[3], reg, 3);
> >>>> \
> >>>> + case 2: write_debug(ptr[2], reg, 2);
> >>>> \
> >>>> + case 1: write_debug(ptr[1], reg, 1);
> >>>> \
> >>>> + default: write_debug(ptr[0], reg, 0);
> >>>> \
> >>>> + }
> >>>> +
> >>>> +void __hyp_text __debug_save_state(struct kvm_vcpu *vcpu,
> >>>> + struct kvm_guest_debug_arch *dbg,
> >>>> + struct kvm_cpu_context *ctxt)
> >>>> +{
> >>>> + if (vcpu->arch.debug_flags & KVM_ARM64_DEBUG_DIRTY) {
> >>>> + u64 aa64dfr0 = read_sysreg(id_aa64dfr0_el1);
> >>>> + int brps, wrps;
> >>>> +
> >>>> + brps = (aa64dfr0 >> 12) & 0xf;
> >>>> + wrps = (aa64dfr0 >> 20) & 0xf;
> >>>> +
> >>>> + save_debug(dbg->dbg_bcr, dbgbcr, brps);
> >>>> + save_debug(dbg->dbg_bvr, dbgbvr, brps);
> >>>> + save_debug(dbg->dbg_wcr, dbgwcr, wrps);
> >>>> + save_debug(dbg->dbg_wvr, dbgwvr, wrps);
> >>>> +
> >>>> + ctxt->sys_regs[MDCCINT_EL1] = read_sysreg(mdccint_el1);
> >>>> + }
> >>>> +}
> >>>> +
> >>>> +void __hyp_text __debug_restore_state(struct kvm_vcpu *vcpu,
> >>>> + struct kvm_guest_debug_arch *dbg,
> >>>> + struct kvm_cpu_context *ctxt)
> >>>> +{
> >>>> + if (vcpu->arch.debug_flags & KVM_ARM64_DEBUG_DIRTY) {
> >>>> + u64 aa64dfr0 = read_sysreg(id_aa64dfr0_el1);
> >>>> + int brps, wrps;
> >>>> +
> >>>> + brps = (aa64dfr0 >> 12) & 0xf;
> >>>> + wrps = (aa64dfr0 >> 20) & 0xf;
> >>>> +
> >>>> + restore_debug(dbg->dbg_bcr, dbgbcr, brps);
> >>>> + restore_debug(dbg->dbg_bvr, dbgbvr, brps);
> >>>> + restore_debug(dbg->dbg_wcr, dbgwcr, wrps);
> >>>> + restore_debug(dbg->dbg_wvr, dbgwvr, wrps);
> >>>> +
> >>>> + write_sysreg(ctxt->sys_regs[MDCCINT_EL1], mdccint_el1);
> >>>> + }
> >>>> +}
> >>>> +
> >>>> +void __hyp_text __debug_cond_save_host_state(struct kvm_vcpu *vcpu)
> >>>> +{
> >>>> + if ((vcpu->arch.ctxt.sys_regs[MDSCR_EL1] & DBG_MDSCR_KDE) ||
> >>>> + (vcpu->arch.ctxt.sys_regs[MDSCR_EL1] & DBG_MDSCR_KDE))
> >>>> + vcpu->arch.debug_flags |= KVM_ARM64_DEBUG_DIRTY;
> >>>> +
> >>>> + __debug_save_state(vcpu, &vcpu->arch.host_debug_state,
> >>>> + kern_hyp_va(vcpu->arch.host_cpu_context));
> >>>
> >>> doesn't the assmebly code jump across saving this state neither bits are
> >>> set where this always saves the state?
> >>
> >> It doesn't. The save/restore functions are guarded by tests on
> >> KVM_ARM64_DEBUG_DIRTY, just like we have skip_debug_state on all actions
> >> involving the save/restore in the assembly version.
> >
> > I think the confusing part is that the save function unconditionally
> > calls __debug_save_state where the restore function only calls it when
> > the dirty flag is set. Plus I suck at reading assembly apparently.
>
> So the way I initially wrote it, I had the same 'if' statement as in the
> restore function, making them fairly symmetric. But it quickly became
> obvious that this double-if was a bit pointless.
>
> And actually, I wonder if I shouldn't drop it from the restore function,
> because it only save us a spurious clear of the dirty bit.
>
I would just move the __debug_restore_state call above the conditional,
then they look more symmetric. Does that work?
> >>> in any case, I feel some context is lost when this is moved away from
> >>> assembly and understanding this patch would be easier if the semantics
> >>> of these two _cond functions were documented.
> >>
> >> I can migrate the existing comments if you think that helps.
> >>
> > It just wasn't not quite clear to me exactly when
> > __debug_cond_save_host_state is called for example - is this going to be
> > called unconditionally on every entry - that's how I understand it now
> > anyway.
>
> On every entry, yes. I'm trying to have the guest_run function as simple
> as possible, with the various subsystems making their 'own' decisions.
>
> Not optimal (you get to branch for nothing), but clearer. At least for
> me, but I may be the odd duck out here. Any idea to make the flow look
> clearer?
>
For me, if you always make the call unconditionally on both paths and
then change the implementations to do
if (!(vcpu->arch.debug_flags & KVM_ARM64_DEBUG_DIRTY))
return;
then I think it's clear enough.
Thanks,
-Christoffer
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
