On 01/12/15 15:41, Christoffer Dall wrote:
> 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.
Fair enough. I'll rework it that way.
Thanks,
M.
--
Jazz is not dead. It just smells funny...
--
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
