On 17/11/15 11:49, Steve Capper wrote:
> On 17 November 2015 at 11:25, Marc Zyngier <marc.zyng...@arm.com> wrote:
>> On 17/11/15 11:13, Steve Capper wrote:
>>> On 16 November 2015 at 13:11, Marc Zyngier <marc.zyng...@arm.com> wrote:
>>>> Implement the fpsimd save restore, keeping the lazy part in
>>>> assembler (as returning to C would be overkill).
>>>>
>>>> Signed-off-by: Marc Zyngier <marc.zyng...@arm.com>
>>>> ---
>>>> arch/arm64/kvm/hyp/Makefile | 1 +
>>>> arch/arm64/kvm/hyp/entry.S | 32 +++++++++++++++++++++++++++++++-
>>>> arch/arm64/kvm/hyp/fpsimd.S | 33 +++++++++++++++++++++++++++++++++
>>>> arch/arm64/kvm/hyp/hyp.h | 3 +++
>>>> arch/arm64/kvm/hyp/switch.c | 8 ++++++++
>>>> 5 files changed, 76 insertions(+), 1 deletion(-)
>>>> create mode 100644 arch/arm64/kvm/hyp/fpsimd.S
>>>>
>>>> diff --git a/arch/arm64/kvm/hyp/Makefile b/arch/arm64/kvm/hyp/Makefile
>>>> index 9c11b0f..56238d0 100644
>>>> --- a/arch/arm64/kvm/hyp/Makefile
>>>> +++ b/arch/arm64/kvm/hyp/Makefile
>>>> @@ -9,3 +9,4 @@ obj-$(CONFIG_KVM_ARM_HOST) += sysreg-sr.o
>>>> obj-$(CONFIG_KVM_ARM_HOST) += debug-sr.o
>>>> obj-$(CONFIG_KVM_ARM_HOST) += entry.o
>>>> obj-$(CONFIG_KVM_ARM_HOST) += switch.o
>>>> +obj-$(CONFIG_KVM_ARM_HOST) += fpsimd.o
>>>> diff --git a/arch/arm64/kvm/hyp/entry.S b/arch/arm64/kvm/hyp/entry.S
>>>> index 2c4449a..7552922 100644
>>>> --- a/arch/arm64/kvm/hyp/entry.S
>>>> +++ b/arch/arm64/kvm/hyp/entry.S
>>>> @@ -27,6 +27,7 @@
>>>>
>>>> #define CPU_GP_REG_OFFSET(x) (CPU_GP_REGS + x)
>>>> #define CPU_XREG_OFFSET(x) CPU_GP_REG_OFFSET(CPU_USER_PT_REGS + 8*x)
>>>> +#define CPU_SYSREG_OFFSET(x) (CPU_SYSREGS + 8*x)
>>>>
>>>> .text
>>>> .pushsection .hyp.text, "ax"
>>>> @@ -152,4 +153,33 @@ ENTRY(__guest_exit)
>>>> ret
>>>> ENDPROC(__guest_exit)
>>>>
>>>> - /* Insert fault handling here */
>>>> +ENTRY(__fpsimd_guest_restore)
>>>> + push x4, lr
>>>> +
>>>> + mrs x2, cptr_el2
>>>> + bic x2, x2, #CPTR_EL2_TFP
>>>> + msr cptr_el2, x2
>>>> + isb
>>>> +
>>>> + mrs x3, tpidr_el2
>>>> +
>>>> + ldr x0, [x3, #VCPU_HOST_CONTEXT]
>>>> + kern_hyp_va x0
>>>> + add x0, x0, #CPU_GP_REG_OFFSET(CPU_FP_REGS)
>>>> + bl __fpsimd_save_state
>>>> +
>>>> + add x2, x3, #VCPU_CONTEXT
>>>> + add x0, x2, #CPU_GP_REG_OFFSET(CPU_FP_REGS)
>>>> + bl __fpsimd_restore_state
>>>> +
>>>> + mrs x1, hcr_el2
>>>> + tbnz x1, #HCR_RW_SHIFT, 1f
>>>> + ldr x4, [x2, #CPU_SYSREG_OFFSET(FPEXC32_EL2)]
>>>> + msr fpexc32_el2, x4
>>>> +1:
>>>> + pop x4, lr
>>>> + pop x2, x3
>>>> + pop x0, x1
>>>> +
>>>> + eret
>>>> +ENDPROC(__fpsimd_guest_restore)
>>>> diff --git a/arch/arm64/kvm/hyp/fpsimd.S b/arch/arm64/kvm/hyp/fpsimd.S
>>>> new file mode 100644
>>>> index 0000000..da3f22c
>>>> --- /dev/null
>>>> +++ b/arch/arm64/kvm/hyp/fpsimd.S
>>>> @@ -0,0 +1,33 @@
>>>> +/*
>>>> + * 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/linkage.h>
>>>> +
>>>> +#include <asm/fpsimdmacros.h>
>>>> +
>>>> + .text
>>>> + .pushsection .hyp.text, "ax"
>>>> +
>>>> +ENTRY(__fpsimd_save_state)
>>>> + fpsimd_save x0, 1
>>>> + ret
>>>> +ENDPROC(__fpsimd_save_state)
>>>> +
>>>> +ENTRY(__fpsimd_restore_state)
>>>> + fpsimd_restore x0, 1
>>>> + ret
>>>> +ENDPROC(__fpsimd_restore_state)
>>>> diff --git a/arch/arm64/kvm/hyp/hyp.h b/arch/arm64/kvm/hyp/hyp.h
>>>> index bf13238..240fb79 100644
>>>> --- a/arch/arm64/kvm/hyp/hyp.h
>>>> +++ b/arch/arm64/kvm/hyp/hyp.h
>>>> @@ -70,6 +70,9 @@ void __debug_clear_restore_state(struct kvm_vcpu *vcpu,
>>>> struct kvm_guest_debug_arch *dbg,
>>>> struct kvm_cpu_context *ctxt);
>>>>
>>>> +void __fpsimd_save_state(struct user_fpsimd_state *fp_regs);
>>>> +void __fpsimd_restore_state(struct user_fpsimd_state *fp_regs);
>>>> +
>>>> u64 __guest_enter(struct kvm_vcpu *vcpu, struct kvm_cpu_context
>>>> *host_ctxt);
>>>>
>>>> #endif /* __ARM64_KVM_HYP_H__ */
>>>> diff --git a/arch/arm64/kvm/hyp/switch.c b/arch/arm64/kvm/hyp/switch.c
>>>> index a3af81a..06d3e20 100644
>>>> --- a/arch/arm64/kvm/hyp/switch.c
>>>> +++ b/arch/arm64/kvm/hyp/switch.c
>>>> @@ -88,6 +88,7 @@ int __hyp_text __guest_run(struct kvm_vcpu *vcpu)
>>>> {
>>>> struct kvm_cpu_context *host_ctxt;
>>>> struct kvm_cpu_context *guest_ctxt;
>>>> + bool fp_enabled;
>>>> u64 exit_code;
>>>>
>>>> vcpu = kern_hyp_va(vcpu);
>>>> @@ -117,6 +118,8 @@ int __hyp_text __guest_run(struct kvm_vcpu *vcpu)
>>>> exit_code = __guest_enter(vcpu, host_ctxt);
>>>> /* And we're baaack! */
>>>>
>>>> + fp_enabled = !!(read_sysreg(cptr_el2) & CPTR_EL2_TFP);
>>>
>>> Should this not be a single logical not?
>>> If CPTR_EL2_TFP is set then the floating point will trap in the guest,
>>> thus floating point hasn't been used by the guest?
>>>
>>> I am chasing some issues when running kvmtool on my system with this
>>> patch set applied.
>>> The above change stops NetworkManager crashing on my host.
>>>
>>> Unfortunately I am still chasing some problems with the guest (illegal
>>> instructions).
>>
>> Yeah, it looks like I screwed up in the base code, and somewhat fixed it
>> in the VHE branch... And as I'm testing the full setup, I didn't spot
>> it, sorry.
>>
>> Can you please give the kvm-arm64/vhe-wsinc branch a spin and let me
>> know if it works for you?
>
> Thanks Marc,
> kvm-arm64/vhe-wsinc works well on my test system.
Right. So for the record (and my karma), the actual fix is:
diff --git a/arch/arm64/kvm/hyp/switch.c b/arch/arm64/kvm/hyp/switch.c
index a2885f5..6241785 100644
--- a/arch/arm64/kvm/hyp/switch.c
+++ b/arch/arm64/kvm/hyp/switch.c
@@ -29,7 +29,7 @@ static void __hyp_text __activate_traps(struct kvm_vcpu *vcpu)
* we set FPEXC.EN to prevent traps to EL1, when setting the TFP bit.
*/
val = vcpu->arch.hcr_el2;
- if (val & HCR_RW) {
+ if (!(val & HCR_RW)) {
write_sysreg(1 << 30, fpexc32_el2);
isb();
}
@@ -118,7 +118,7 @@ static int __hyp_text __guest_run(struct kvm_vcpu *vcpu)
exit_code = __guest_enter(vcpu, host_ctxt);
/* And we're baaack! */
- fp_enabled = !!(read_sysreg(cptr_el2) & CPTR_EL2_TFP);
+ fp_enabled = !(read_sysreg(cptr_el2) & CPTR_EL2_TFP);
__sysreg_save_state(guest_ctxt);
__sysreg32_save_state(vcpu);
I'll push out both branches again later today, with these fixes (and a
couple more).
Many thanks for testing this Steve.
M.
--
Jazz is not dead. It just smells funny...
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
