KVM: ARM: World-switch implementation

Provides complete world-switch implementation to switch to other guests
running in non-secure modes. Includes Hyp exception handlers that
capture necessary exception information and stores the information on
the VCPU and KVM structures.

The following Hyp-ABI is also documented in the code:

Hyp-ABI: Calling HYP-mode functions from host (in SVC mode):
   Switching to Hyp mode is done through a simple HVC #0 instruction. The
   exception vector code will check that the HVC comes from VMID==0 and if
   so will push the necessary state (SPSR, lr_usr) on the Hyp stack.
   - r0 contains a pointer to a HYP function
   - r1, r2, and r3 contain arguments to the above function.
   - The HYP function will be called with its arguments in r0, r1 and r2.
   On HYP function return, we return directly to SVC.

A call to a function executing in Hyp mode is performed like the following:

        <svc code>
        ldr     r0, =BSYM(my_hyp_fn)
        ldr     r1, =my_param
        hvc #0  ; Call my_hyp_fn(my_param) from HYP mode
        <svc code>

Otherwise, the world-switch is pretty straight-forward. All state that
can be modified by the guest is first backed up on the Hyp stack and the
VCPU values is loaded onto the hardware. State, which is not loaded, but
theoretically modifiable by the guest is protected through the
virtualiation features to generate a trap and cause software emulation.
Upon guest returns, all state is restored from hardware onto the VCPU
struct and the original state is restored from the Hyp-stack onto the
hardware.

SMP support using the VMPIDR calculated on the basis of the host MPIDR
and overriding the low bits with KVM vcpu_id contributed by Marc Zyngier.

Reuse of VMIDs has been implemented by Antonios Motakis and adapated from
a separate patch into the appropriate patches introducing the
functionality. Note that the VMIDs are stored per VM as required by the ARM
architecture reference manual.

To support VFP/NEON we trap those instructions using the HPCTR. When
we trap, we switch the FPU.  After a guest exit, the VFP state is
returned to the host.  When disabling access to floating point
instructions, we also mask FPEXC_EN in order to avoid the guest
receiving Undefined instruction exceptions before we have a chance to
switch back the floating point state.  We are reusing vfp_hard_struct,
so we depend on VFPv3 being enabled in the host kernel, if not, we still
trap cp10 and cp11 in order to inject an undefined instruction exception
whenever the guest tries to use VFP/NEON. VFP/NEON developed by
Antionios Motakis and Rusty Russell.

Aborts that are permission faults, and not stage-1 page table walk, do
not report the faulting address in the HPFAR.  We have to resolve the
IPA, and store it just like the HPFAR register on the VCPU struct. If
the IPA cannot be resolved, it means another CPU is playing with the
page tables, and we simply restart the guest.  This quirk was fixed by
Marc Zyngier.

Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Antonios Motakis <a.motakis@virtualopensystems.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>

1 files changed, 393 insertions, 3 deletions

diff --git a/arch/arm/kvm/interrupts.S b/arch/arm/kvm/interrupts.S
index f701aff31e44..c5400d2e97ca 100644
--- a/arch/arm/kvm/interrupts.S
+++ b/arch/arm/kvm/interrupts.S
@@ -20,9 +20,12 @@
 #include <linux/const.h>
 #include <asm/unified.h>
 #include <asm/page.h>
+#include <asm/ptrace.h>
 #include <asm/asm-offsets.h>
 #include <asm/kvm_asm.h>
 #include <asm/kvm_arm.h>
+#include <asm/vfpmacros.h>
+#include "interrupts_head.S"
 
 	.text
 
@@ -31,23 +34,164 @@ __kvm_hyp_code_start:
 
 /********************************************************************
  * Flush per-VMID TLBs
+ *
+ * void __kvm_tlb_flush_vmid(struct kvm *kvm);
+ *
+ * We rely on the hardware to broadcast the TLB invalidation to all CPUs
+ * inside the inner-shareable domain (which is the case for all v7
+ * implementations).  If we come across a non-IS SMP implementation, we'll
+ * have to use an IPI based mechanism. Until then, we stick to the simple
+ * hardware assisted version.
  */
 ENTRY(__kvm_tlb_flush_vmid)
+	push	{r2, r3}
+
+	add	r0, r0, #KVM_VTTBR
+	ldrd	r2, r3, [r0]
+	mcrr	p15, 6, r2, r3, c2	@ Write VTTBR
+	isb
+	mcr     p15, 0, r0, c8, c3, 0	@ TLBIALLIS (rt ignored)
+	dsb
+	isb
+	mov	r2, #0
+	mov	r3, #0
+	mcrr	p15, 6, r2, r3, c2	@ Back to VMID #0
+	isb				@ Not necessary if followed by eret
+
+	pop	{r2, r3}
 	bx	lr
 ENDPROC(__kvm_tlb_flush_vmid)
 
 /********************************************************************
- * Flush TLBs and instruction caches of current CPU for all VMIDs
+ * Flush TLBs and instruction caches of all CPUs inside the inner-shareable
+ * domain, for all VMIDs
+ *
+ * void __kvm_flush_vm_context(void);
  */
 ENTRY(__kvm_flush_vm_context)
+	mov	r0, #0			@ rn parameter for c15 flushes is SBZ
+
+	/* Invalidate NS Non-Hyp TLB Inner Shareable (TLBIALLNSNHIS) */
+	mcr     p15, 4, r0, c8, c3, 4
+	/* Invalidate instruction caches Inner Shareable (ICIALLUIS) */
+	mcr     p15, 0, r0, c7, c1, 0
+	dsb
+	isb				@ Not necessary if followed by eret
+
 	bx	lr
 ENDPROC(__kvm_flush_vm_context)
 
+
 /********************************************************************
  *  Hypervisor world-switch code
+ *
+ *
+ * int __kvm_vcpu_run(struct kvm_vcpu *vcpu)
  */
 ENTRY(__kvm_vcpu_run)
-	bx	lr
+	@ Save the vcpu pointer
+	mcr	p15, 4, vcpu, c13, c0, 2	@ HTPIDR
+
+	save_host_regs
+
+	@ Store hardware CP15 state and load guest state
+	read_cp15_state store_to_vcpu = 0
+	write_cp15_state read_from_vcpu = 1
+
+	@ If the host kernel has not been configured with VFPv3 support,
+	@ then it is safer if we deny guests from using it as well.
+#ifdef CONFIG_VFPv3
+	@ Set FPEXC_EN so the guest doesn't trap floating point instructions
+	VFPFMRX r2, FPEXC		@ VMRS
+	push	{r2}
+	orr	r2, r2, #FPEXC_EN
+	VFPFMXR FPEXC, r2		@ VMSR
+#endif
+
+	@ Configure Hyp-role
+	configure_hyp_role vmentry
+
+	@ Trap coprocessor CRx accesses
+	set_hstr vmentry
+	set_hcptr vmentry, (HCPTR_TTA | HCPTR_TCP(10) | HCPTR_TCP(11))
+	set_hdcr vmentry
+
+	@ Write configured ID register into MIDR alias
+	ldr	r1, [vcpu, #VCPU_MIDR]
+	mcr	p15, 4, r1, c0, c0, 0
+
+	@ Write guest view of MPIDR into VMPIDR
+	ldr	r1, [vcpu, #CP15_OFFSET(c0_MPIDR)]
+	mcr	p15, 4, r1, c0, c0, 5
+
+	@ Set up guest memory translation
+	ldr	r1, [vcpu, #VCPU_KVM]
+	add	r1, r1, #KVM_VTTBR
+	ldrd	r2, r3, [r1]
+	mcrr	p15, 6, r2, r3, c2	@ Write VTTBR
+
+	@ We're all done, just restore the GPRs and go to the guest
+	restore_guest_regs
+	clrex				@ Clear exclusive monitor
+	eret
+
+__kvm_vcpu_return:
+	/*
+	 * return convention:
+	 * guest r0, r1, r2 saved on the stack
+	 * r0: vcpu pointer
+	 * r1: exception code
+	 */
+	save_guest_regs
+
+	@ Set VMID == 0
+	mov	r2, #0
+	mov	r3, #0
+	mcrr	p15, 6, r2, r3, c2	@ Write VTTBR
+
+	@ Don't trap coprocessor accesses for host kernel
+	set_hstr vmexit
+	set_hdcr vmexit
+	set_hcptr vmexit, (HCPTR_TTA | HCPTR_TCP(10) | HCPTR_TCP(11))
+
+#ifdef CONFIG_VFPv3
+	@ Save floating point registers we if let guest use them.
+	tst	r2, #(HCPTR_TCP(10) | HCPTR_TCP(11))
+	bne	after_vfp_restore
+
+	@ Switch VFP/NEON hardware state to the host's
+	add	r7, vcpu, #VCPU_VFP_GUEST
+	store_vfp_state r7
+	add	r7, vcpu, #VCPU_VFP_HOST
+	ldr	r7, [r7]
+	restore_vfp_state r7
+
+after_vfp_restore:
+	@ Restore FPEXC_EN which we clobbered on entry
+	pop	{r2}
+	VFPFMXR FPEXC, r2
+#endif
+
+	@ Reset Hyp-role
+	configure_hyp_role vmexit
+
+	@ Let host read hardware MIDR
+	mrc	p15, 0, r2, c0, c0, 0
+	mcr	p15, 4, r2, c0, c0, 0
+
+	@ Back to hardware MPIDR
+	mrc	p15, 0, r2, c0, c0, 5
+	mcr	p15, 4, r2, c0, c0, 5
+
+	@ Store guest CP15 state and restore host state
+	read_cp15_state store_to_vcpu = 1
+	write_cp15_state read_from_vcpu = 0
+
+	restore_host_regs
+	clrex				@ Clear exclusive monitor
+	mov	r0, r1			@ Return the return code
+	mov	r1, #0			@ Clear upper bits in return value
+	bx	lr			@ return to IOCTL
 
 /********************************************************************
  *  Call function in Hyp mode
@@ -77,12 +221,258 @@ ENTRY(kvm_call_hyp)
 
 /********************************************************************
  * Hypervisor exception vector and handlers
+ *
+ *
+ * The KVM/ARM Hypervisor ABI is defined as follows:
+ *
+ * Entry to Hyp mode from the host kernel will happen _only_ when an HVC
+ * instruction is issued since all traps are disabled when running the host
+ * kernel as per the Hyp-mode initialization at boot time.
+ *
+ * HVC instructions cause a trap to the vector page + offset 0x18 (see hyp_hvc
+ * below) when the HVC instruction is called from SVC mode (i.e. a guest or the
+ * host kernel) and they cause a trap to the vector page + offset 0xc when HVC
+ * instructions are called from within Hyp-mode.
+ *
+ * Hyp-ABI: Calling HYP-mode functions from host (in SVC mode):
+ *    Switching to Hyp mode is done through a simple HVC #0 instruction. The
+ *    exception vector code will check that the HVC comes from VMID==0 and if
+ *    so will push the necessary state (SPSR, lr_usr) on the Hyp stack.
+ *    - r0 contains a pointer to a HYP function
+ *    - r1, r2, and r3 contain arguments to the above function.
+ *    - The HYP function will be called with its arguments in r0, r1 and r2.
+ *    On HYP function return, we return directly to SVC.
+ *
+ * Note that the above is used to execute code in Hyp-mode from a host-kernel
+ * point of view, and is a different concept from performing a world-switch and
+ * executing guest code SVC mode (with a VMID != 0).
  */
 
+/* Handle undef, svc, pabt, or dabt by crashing with a user notice */
+.macro bad_exception exception_code, panic_str
+	push	{r0-r2}
+	mrrc	p15, 6, r0, r1, c2	@ Read VTTBR
+	lsr	r1, r1, #16
+	ands	r1, r1, #0xff
+	beq	99f
+
+	load_vcpu			@ Load VCPU pointer
+	.if \exception_code == ARM_EXCEPTION_DATA_ABORT
+	mrc	p15, 4, r2, c5, c2, 0	@ HSR
+	mrc	p15, 4, r1, c6, c0, 0	@ HDFAR
+	str	r2, [vcpu, #VCPU_HSR]
+	str	r1, [vcpu, #VCPU_HxFAR]
+	.endif
+	.if \exception_code == ARM_EXCEPTION_PREF_ABORT
+	mrc	p15, 4, r2, c5, c2, 0	@ HSR
+	mrc	p15, 4, r1, c6, c0, 2	@ HIFAR
+	str	r2, [vcpu, #VCPU_HSR]
+	str	r1, [vcpu, #VCPU_HxFAR]
+	.endif
+	mov	r1, #\exception_code
+	b	__kvm_vcpu_return
+
+	@ We were in the host already. Let's craft a panic-ing return to SVC.
+99:	mrs	r2, cpsr
+	bic	r2, r2, #MODE_MASK
+	orr	r2, r2, #SVC_MODE
+THUMB(	orr	r2, r2, #PSR_T_BIT	)
+	msr	spsr_cxsf, r2
+	mrs	r1, ELR_hyp
+	ldr	r2, =BSYM(panic)
+	msr	ELR_hyp, r2
+	ldr	r0, =\panic_str
+	eret
+.endm
+
+	.text
+
 	.align 5
 __kvm_hyp_vector:
 	.globl __kvm_hyp_vector
-	nop
+
+	@ Hyp-mode exception vector
+	W(b)	hyp_reset
+	W(b)	hyp_undef
+	W(b)	hyp_svc
+	W(b)	hyp_pabt
+	W(b)	hyp_dabt
+	W(b)	hyp_hvc
+	W(b)	hyp_irq
+	W(b)	hyp_fiq
+
+	.align
+hyp_reset:
+	b	hyp_reset
+
+	.align
+hyp_undef:
+	bad_exception ARM_EXCEPTION_UNDEFINED, und_die_str
+
+	.align
+hyp_svc:
+	bad_exception ARM_EXCEPTION_HVC, svc_die_str
+
+	.align
+hyp_pabt:
+	bad_exception ARM_EXCEPTION_PREF_ABORT, pabt_die_str
+
+	.align
+hyp_dabt:
+	bad_exception ARM_EXCEPTION_DATA_ABORT, dabt_die_str
+
+	.align
+hyp_hvc:
+	/*
+	 * Getting here is either becuase of a trap from a guest or from calling
+	 * HVC from the host kernel, which means "switch to Hyp mode".
+	 */
+	push	{r0, r1, r2}
+
+	@ Check syndrome register
+	mrc	p15, 4, r1, c5, c2, 0	@ HSR
+	lsr	r0, r1, #HSR_EC_SHIFT
+#ifdef CONFIG_VFPv3
+	cmp	r0, #HSR_EC_CP_0_13
+	beq	switch_to_guest_vfp
+#endif
+	cmp	r0, #HSR_EC_HVC
+	bne	guest_trap		@ Not HVC instr.
+
+	/*
+	 * Let's check if the HVC came from VMID 0 and allow simple
+	 * switch to Hyp mode
+	 */
+	mrrc    p15, 6, r0, r2, c2
+	lsr     r2, r2, #16
+	and     r2, r2, #0xff
+	cmp     r2, #0
+	bne	guest_trap		@ Guest called HVC
+
+host_switch_to_hyp:
+	pop	{r0, r1, r2}
+
+	push	{lr}
+	mrs	lr, SPSR
+	push	{lr}
+
+	mov	lr, r0
+	mov	r0, r1
+	mov	r1, r2
+	mov	r2, r3
+
+THUMB(	orr	lr, #1)
+	blx	lr			@ Call the HYP function
+
+	pop	{lr}
+	msr	SPSR_csxf, lr
+	pop	{lr}
+	eret
+
+guest_trap:
+	load_vcpu			@ Load VCPU pointer to r0
+	str	r1, [vcpu, #VCPU_HSR]
+
+	@ Check if we need the fault information
+	lsr	r1, r1, #HSR_EC_SHIFT
+	cmp	r1, #HSR_EC_IABT
+	mrceq	p15, 4, r2, c6, c0, 2	@ HIFAR
+	beq	2f
+	cmp	r1, #HSR_EC_DABT
+	bne	1f
+	mrc	p15, 4, r2, c6, c0, 0	@ HDFAR
+
+2:	str	r2, [vcpu, #VCPU_HxFAR]
+
+	/*
+	 * B3.13.5 Reporting exceptions taken to the Non-secure PL2 mode:
+	 *
+	 * Abort on the stage 2 translation for a memory access from a
+	 * Non-secure PL1 or PL0 mode:
+	 *
+	 * For any Access flag fault or Translation fault, and also for any
+	 * Permission fault on the stage 2 translation of a memory access
+	 * made as part of a translation table walk for a stage 1 translation,
+	 * the HPFAR holds the IPA that caused the fault. Otherwise, the HPFAR
+	 * is UNKNOWN.
+	 */
+
+	/* Check for permission fault, and S1PTW */
+	mrc	p15, 4, r1, c5, c2, 0	@ HSR
+	and	r0, r1, #HSR_FSC_TYPE
+	cmp	r0, #FSC_PERM
+	tsteq	r1, #(1 << 7)		@ S1PTW
+	mrcne	p15, 4, r2, c6, c0, 4	@ HPFAR
+	bne	3f
+
+	/* Resolve IPA using the xFAR */
+	mcr	p15, 0, r2, c7, c8, 0	@ ATS1CPR
+	isb
+	mrrc	p15, 0, r0, r1, c7	@ PAR
+	tst	r0, #1
+	bne	4f			@ Failed translation
+	ubfx	r2, r0, #12, #20
+	lsl	r2, r2, #4
+	orr	r2, r2, r1, lsl #24
+
+3:	load_vcpu			@ Load VCPU pointer to r0
+	str	r2, [r0, #VCPU_HPFAR]
+
+1:	mov	r1, #ARM_EXCEPTION_HVC
+	b	__kvm_vcpu_return
+
+4:	pop	{r0, r1, r2}		@ Failed translation, return to guest
+	eret
+
+/*
+ * If VFPv3 support is not available, then we will not switch the VFP
+ * registers; however cp10 and cp11 accesses will still trap and fallback
+ * to the regular coprocessor emulation code, which currently will
+ * inject an undefined exception to the guest.
+ */
+#ifdef CONFIG_VFPv3
+switch_to_guest_vfp:
+	load_vcpu			@ Load VCPU pointer to r0
+	push	{r3-r7}
+
+	@ NEON/VFP used.  Turn on VFP access.
+	set_hcptr vmexit, (HCPTR_TCP(10) | HCPTR_TCP(11))
+
+	@ Switch VFP/NEON hardware state to the guest's
+	add	r7, r0, #VCPU_VFP_HOST
+	ldr	r7, [r7]
+	store_vfp_state r7
+	add	r7, r0, #VCPU_VFP_GUEST
+	restore_vfp_state r7
+
+	pop	{r3-r7}
+	pop	{r0-r2}
+	eret
+#endif
+
+	.align
+hyp_irq:
+	push	{r0, r1, r2}
+	mov	r1, #ARM_EXCEPTION_IRQ
+	load_vcpu			@ Load VCPU pointer to r0
+	b	__kvm_vcpu_return
+
+	.align
+hyp_fiq:
+	b	hyp_fiq
+
+	.ltorg
 
 __kvm_hyp_code_end:
 	.globl	__kvm_hyp_code_end
+
+	.section ".rodata"
+
+und_die_str:
+	.ascii	"unexpected undefined exception in Hyp mode at: %#08x"
+pabt_die_str:
+	.ascii	"unexpected prefetch abort in Hyp mode at: %#08x"
+dabt_die_str:
+	.ascii	"unexpected data abort in Hyp mode at: %#08x"
+svc_die_str:
+	.ascii	"unexpected HVC/SVC trap in Hyp mode at: %#08x"