ARM: KVM: VGIC distributor handling

Add the GIC distributor emulation code. A number of the GIC features
are simply ignored as they are not required to boot a Linux guest.

Reviewed-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
This commit is contained in:
Marc Zyngier 2013-01-21 19:36:14 -05:00
parent 330690cdce
commit b47ef92af8
2 changed files with 677 additions and 1 deletions

View file

@ -19,12 +19,94 @@
#ifndef __ASM_ARM_KVM_VGIC_H
#define __ASM_ARM_KVM_VGIC_H
#include <linux/kernel.h>
#include <linux/kvm.h>
#include <linux/kvm_host.h>
#include <linux/irqreturn.h>
#include <linux/spinlock.h>
#include <linux/types.h>
#include <linux/irqchip/arm-gic.h>
#define VGIC_NR_IRQS 128
#define VGIC_NR_SGIS 16
#define VGIC_NR_PPIS 16
#define VGIC_NR_PRIVATE_IRQS (VGIC_NR_SGIS + VGIC_NR_PPIS)
#define VGIC_NR_SHARED_IRQS (VGIC_NR_IRQS - VGIC_NR_PRIVATE_IRQS)
#define VGIC_MAX_CPUS KVM_MAX_VCPUS
/* Sanity checks... */
#if (VGIC_MAX_CPUS > 8)
#error Invalid number of CPU interfaces
#endif
#if (VGIC_NR_IRQS & 31)
#error "VGIC_NR_IRQS must be a multiple of 32"
#endif
#if (VGIC_NR_IRQS > 1024)
#error "VGIC_NR_IRQS must be <= 1024"
#endif
/*
* The GIC distributor registers describing interrupts have two parts:
* - 32 per-CPU interrupts (SGI + PPI)
* - a bunch of shared interrupts (SPI)
*/
struct vgic_bitmap {
union {
u32 reg[VGIC_NR_PRIVATE_IRQS / 32];
DECLARE_BITMAP(reg_ul, VGIC_NR_PRIVATE_IRQS);
} percpu[VGIC_MAX_CPUS];
union {
u32 reg[VGIC_NR_SHARED_IRQS / 32];
DECLARE_BITMAP(reg_ul, VGIC_NR_SHARED_IRQS);
} shared;
};
struct vgic_bytemap {
u32 percpu[VGIC_MAX_CPUS][VGIC_NR_PRIVATE_IRQS / 4];
u32 shared[VGIC_NR_SHARED_IRQS / 4];
};
struct vgic_dist {
#ifdef CONFIG_KVM_ARM_VGIC
spinlock_t lock;
/* Virtual control interface mapping */
void __iomem *vctrl_base;
/* Distributor and vcpu interface mapping in the guest */
phys_addr_t vgic_dist_base;
phys_addr_t vgic_cpu_base;
/* Distributor enabled */
u32 enabled;
/* Interrupt enabled (one bit per IRQ) */
struct vgic_bitmap irq_enabled;
/* Interrupt 'pin' level */
struct vgic_bitmap irq_state;
/* Level-triggered interrupt in progress */
struct vgic_bitmap irq_active;
/* Interrupt priority. Not used yet. */
struct vgic_bytemap irq_priority;
/* Level/edge triggered */
struct vgic_bitmap irq_cfg;
/* Source CPU per SGI and target CPU */
u8 irq_sgi_sources[VGIC_MAX_CPUS][VGIC_NR_SGIS];
/* Target CPU for each IRQ */
u8 irq_spi_cpu[VGIC_NR_SHARED_IRQS];
struct vgic_bitmap irq_spi_target[VGIC_MAX_CPUS];
/* Bitmap indicating which CPU has something pending */
unsigned long irq_pending_on_cpu;
#endif
};
struct vgic_cpu {

View file

@ -22,6 +22,43 @@
#include <linux/io.h>
#include <asm/kvm_emulate.h>
/*
* How the whole thing works (courtesy of Christoffer Dall):
*
* - At any time, the dist->irq_pending_on_cpu is the oracle that knows if
* something is pending
* - VGIC pending interrupts are stored on the vgic.irq_state vgic
* bitmap (this bitmap is updated by both user land ioctls and guest
* mmio ops, and other in-kernel peripherals such as the
* arch. timers) and indicate the 'wire' state.
* - Every time the bitmap changes, the irq_pending_on_cpu oracle is
* recalculated
* - To calculate the oracle, we need info for each cpu from
* compute_pending_for_cpu, which considers:
* - PPI: dist->irq_state & dist->irq_enable
* - SPI: dist->irq_state & dist->irq_enable & dist->irq_spi_target
* - irq_spi_target is a 'formatted' version of the GICD_ICFGR
* registers, stored on each vcpu. We only keep one bit of
* information per interrupt, making sure that only one vcpu can
* accept the interrupt.
* - The same is true when injecting an interrupt, except that we only
* consider a single interrupt at a time. The irq_spi_cpu array
* contains the target CPU for each SPI.
*
* The handling of level interrupts adds some extra complexity. We
* need to track when the interrupt has been EOIed, so we can sample
* the 'line' again. This is achieved as such:
*
* - When a level interrupt is moved onto a vcpu, the corresponding
* bit in irq_active is set. As long as this bit is set, the line
* will be ignored for further interrupts. The interrupt is injected
* into the vcpu with the GICH_LR_EOI bit set (generate a
* maintenance interrupt on EOI).
* - When the interrupt is EOIed, the maintenance interrupt fires,
* and clears the corresponding bit in irq_active. This allow the
* interrupt line to be sampled again.
*/
#define VGIC_ADDR_UNDEF (-1)
#define IS_VGIC_ADDR_UNDEF(_x) ((_x) == VGIC_ADDR_UNDEF)
@ -34,6 +71,119 @@
#define ACCESS_WRITE_VALUE (3 << 1)
#define ACCESS_WRITE_MASK(x) ((x) & (3 << 1))
static void vgic_update_state(struct kvm *kvm);
static void vgic_dispatch_sgi(struct kvm_vcpu *vcpu, u32 reg);
static u32 *vgic_bitmap_get_reg(struct vgic_bitmap *x,
int cpuid, u32 offset)
{
offset >>= 2;
if (!offset)
return x->percpu[cpuid].reg;
else
return x->shared.reg + offset - 1;
}
static int vgic_bitmap_get_irq_val(struct vgic_bitmap *x,
int cpuid, int irq)
{
if (irq < VGIC_NR_PRIVATE_IRQS)
return test_bit(irq, x->percpu[cpuid].reg_ul);
return test_bit(irq - VGIC_NR_PRIVATE_IRQS, x->shared.reg_ul);
}
static void vgic_bitmap_set_irq_val(struct vgic_bitmap *x, int cpuid,
int irq, int val)
{
unsigned long *reg;
if (irq < VGIC_NR_PRIVATE_IRQS) {
reg = x->percpu[cpuid].reg_ul;
} else {
reg = x->shared.reg_ul;
irq -= VGIC_NR_PRIVATE_IRQS;
}
if (val)
set_bit(irq, reg);
else
clear_bit(irq, reg);
}
static unsigned long *vgic_bitmap_get_cpu_map(struct vgic_bitmap *x, int cpuid)
{
if (unlikely(cpuid >= VGIC_MAX_CPUS))
return NULL;
return x->percpu[cpuid].reg_ul;
}
static unsigned long *vgic_bitmap_get_shared_map(struct vgic_bitmap *x)
{
return x->shared.reg_ul;
}
static u32 *vgic_bytemap_get_reg(struct vgic_bytemap *x, int cpuid, u32 offset)
{
offset >>= 2;
BUG_ON(offset > (VGIC_NR_IRQS / 4));
if (offset < 4)
return x->percpu[cpuid] + offset;
else
return x->shared + offset - 8;
}
#define VGIC_CFG_LEVEL 0
#define VGIC_CFG_EDGE 1
static bool vgic_irq_is_edge(struct kvm_vcpu *vcpu, int irq)
{
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
int irq_val;
irq_val = vgic_bitmap_get_irq_val(&dist->irq_cfg, vcpu->vcpu_id, irq);
return irq_val == VGIC_CFG_EDGE;
}
static int vgic_irq_is_enabled(struct kvm_vcpu *vcpu, int irq)
{
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
return vgic_bitmap_get_irq_val(&dist->irq_enabled, vcpu->vcpu_id, irq);
}
static void vgic_dist_irq_set(struct kvm_vcpu *vcpu, int irq)
{
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
vgic_bitmap_set_irq_val(&dist->irq_state, vcpu->vcpu_id, irq, 1);
}
static void vgic_dist_irq_clear(struct kvm_vcpu *vcpu, int irq)
{
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
vgic_bitmap_set_irq_val(&dist->irq_state, vcpu->vcpu_id, irq, 0);
}
static void vgic_cpu_irq_set(struct kvm_vcpu *vcpu, int irq)
{
if (irq < VGIC_NR_PRIVATE_IRQS)
set_bit(irq, vcpu->arch.vgic_cpu.pending_percpu);
else
set_bit(irq - VGIC_NR_PRIVATE_IRQS,
vcpu->arch.vgic_cpu.pending_shared);
}
static void vgic_cpu_irq_clear(struct kvm_vcpu *vcpu, int irq)
{
if (irq < VGIC_NR_PRIVATE_IRQS)
clear_bit(irq, vcpu->arch.vgic_cpu.pending_percpu);
else
clear_bit(irq - VGIC_NR_PRIVATE_IRQS,
vcpu->arch.vgic_cpu.pending_shared);
}
static u32 mmio_data_read(struct kvm_exit_mmio *mmio, u32 mask)
{
return *((u32 *)mmio->data) & mask;
@ -105,6 +255,291 @@ static void vgic_reg_access(struct kvm_exit_mmio *mmio, u32 *reg,
}
}
static bool handle_mmio_misc(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio, phys_addr_t offset)
{
u32 reg;
u32 word_offset = offset & 3;
switch (offset & ~3) {
case 0: /* CTLR */
reg = vcpu->kvm->arch.vgic.enabled;
vgic_reg_access(mmio, &reg, word_offset,
ACCESS_READ_VALUE | ACCESS_WRITE_VALUE);
if (mmio->is_write) {
vcpu->kvm->arch.vgic.enabled = reg & 1;
vgic_update_state(vcpu->kvm);
return true;
}
break;
case 4: /* TYPER */
reg = (atomic_read(&vcpu->kvm->online_vcpus) - 1) << 5;
reg |= (VGIC_NR_IRQS >> 5) - 1;
vgic_reg_access(mmio, &reg, word_offset,
ACCESS_READ_VALUE | ACCESS_WRITE_IGNORED);
break;
case 8: /* IIDR */
reg = 0x4B00043B;
vgic_reg_access(mmio, &reg, word_offset,
ACCESS_READ_VALUE | ACCESS_WRITE_IGNORED);
break;
}
return false;
}
static bool handle_mmio_raz_wi(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio, phys_addr_t offset)
{
vgic_reg_access(mmio, NULL, offset,
ACCESS_READ_RAZ | ACCESS_WRITE_IGNORED);
return false;
}
static bool handle_mmio_set_enable_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio,
phys_addr_t offset)
{
u32 *reg = vgic_bitmap_get_reg(&vcpu->kvm->arch.vgic.irq_enabled,
vcpu->vcpu_id, offset);
vgic_reg_access(mmio, reg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_SETBIT);
if (mmio->is_write) {
vgic_update_state(vcpu->kvm);
return true;
}
return false;
}
static bool handle_mmio_clear_enable_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio,
phys_addr_t offset)
{
u32 *reg = vgic_bitmap_get_reg(&vcpu->kvm->arch.vgic.irq_enabled,
vcpu->vcpu_id, offset);
vgic_reg_access(mmio, reg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_CLEARBIT);
if (mmio->is_write) {
if (offset < 4) /* Force SGI enabled */
*reg |= 0xffff;
vgic_update_state(vcpu->kvm);
return true;
}
return false;
}
static bool handle_mmio_set_pending_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio,
phys_addr_t offset)
{
u32 *reg = vgic_bitmap_get_reg(&vcpu->kvm->arch.vgic.irq_state,
vcpu->vcpu_id, offset);
vgic_reg_access(mmio, reg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_SETBIT);
if (mmio->is_write) {
vgic_update_state(vcpu->kvm);
return true;
}
return false;
}
static bool handle_mmio_clear_pending_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio,
phys_addr_t offset)
{
u32 *reg = vgic_bitmap_get_reg(&vcpu->kvm->arch.vgic.irq_state,
vcpu->vcpu_id, offset);
vgic_reg_access(mmio, reg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_CLEARBIT);
if (mmio->is_write) {
vgic_update_state(vcpu->kvm);
return true;
}
return false;
}
static bool handle_mmio_priority_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio,
phys_addr_t offset)
{
u32 *reg = vgic_bytemap_get_reg(&vcpu->kvm->arch.vgic.irq_priority,
vcpu->vcpu_id, offset);
vgic_reg_access(mmio, reg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_VALUE);
return false;
}
#define GICD_ITARGETSR_SIZE 32
#define GICD_CPUTARGETS_BITS 8
#define GICD_IRQS_PER_ITARGETSR (GICD_ITARGETSR_SIZE / GICD_CPUTARGETS_BITS)
static u32 vgic_get_target_reg(struct kvm *kvm, int irq)
{
struct vgic_dist *dist = &kvm->arch.vgic;
struct kvm_vcpu *vcpu;
int i, c;
unsigned long *bmap;
u32 val = 0;
irq -= VGIC_NR_PRIVATE_IRQS;
kvm_for_each_vcpu(c, vcpu, kvm) {
bmap = vgic_bitmap_get_shared_map(&dist->irq_spi_target[c]);
for (i = 0; i < GICD_IRQS_PER_ITARGETSR; i++)
if (test_bit(irq + i, bmap))
val |= 1 << (c + i * 8);
}
return val;
}
static void vgic_set_target_reg(struct kvm *kvm, u32 val, int irq)
{
struct vgic_dist *dist = &kvm->arch.vgic;
struct kvm_vcpu *vcpu;
int i, c;
unsigned long *bmap;
u32 target;
irq -= VGIC_NR_PRIVATE_IRQS;
/*
* Pick the LSB in each byte. This ensures we target exactly
* one vcpu per IRQ. If the byte is null, assume we target
* CPU0.
*/
for (i = 0; i < GICD_IRQS_PER_ITARGETSR; i++) {
int shift = i * GICD_CPUTARGETS_BITS;
target = ffs((val >> shift) & 0xffU);
target = target ? (target - 1) : 0;
dist->irq_spi_cpu[irq + i] = target;
kvm_for_each_vcpu(c, vcpu, kvm) {
bmap = vgic_bitmap_get_shared_map(&dist->irq_spi_target[c]);
if (c == target)
set_bit(irq + i, bmap);
else
clear_bit(irq + i, bmap);
}
}
}
static bool handle_mmio_target_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio,
phys_addr_t offset)
{
u32 reg;
/* We treat the banked interrupts targets as read-only */
if (offset < 32) {
u32 roreg = 1 << vcpu->vcpu_id;
roreg |= roreg << 8;
roreg |= roreg << 16;
vgic_reg_access(mmio, &roreg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_IGNORED);
return false;
}
reg = vgic_get_target_reg(vcpu->kvm, offset & ~3U);
vgic_reg_access(mmio, &reg, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_VALUE);
if (mmio->is_write) {
vgic_set_target_reg(vcpu->kvm, reg, offset & ~3U);
vgic_update_state(vcpu->kvm);
return true;
}
return false;
}
static u32 vgic_cfg_expand(u16 val)
{
u32 res = 0;
int i;
/*
* Turn a 16bit value like abcd...mnop into a 32bit word
* a0b0c0d0...m0n0o0p0, which is what the HW cfg register is.
*/
for (i = 0; i < 16; i++)
res |= ((val >> i) & VGIC_CFG_EDGE) << (2 * i + 1);
return res;
}
static u16 vgic_cfg_compress(u32 val)
{
u16 res = 0;
int i;
/*
* Turn a 32bit word a0b0c0d0...m0n0o0p0 into 16bit value like
* abcd...mnop which is what we really care about.
*/
for (i = 0; i < 16; i++)
res |= ((val >> (i * 2 + 1)) & VGIC_CFG_EDGE) << i;
return res;
}
/*
* The distributor uses 2 bits per IRQ for the CFG register, but the
* LSB is always 0. As such, we only keep the upper bit, and use the
* two above functions to compress/expand the bits
*/
static bool handle_mmio_cfg_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio, phys_addr_t offset)
{
u32 val;
u32 *reg = vgic_bitmap_get_reg(&vcpu->kvm->arch.vgic.irq_cfg,
vcpu->vcpu_id, offset >> 1);
if (offset & 2)
val = *reg >> 16;
else
val = *reg & 0xffff;
val = vgic_cfg_expand(val);
vgic_reg_access(mmio, &val, offset,
ACCESS_READ_VALUE | ACCESS_WRITE_VALUE);
if (mmio->is_write) {
if (offset < 4) {
*reg = ~0U; /* Force PPIs/SGIs to 1 */
return false;
}
val = vgic_cfg_compress(val);
if (offset & 2) {
*reg &= 0xffff;
*reg |= val << 16;
} else {
*reg &= 0xffff << 16;
*reg |= val;
}
}
return false;
}
static bool handle_mmio_sgi_reg(struct kvm_vcpu *vcpu,
struct kvm_exit_mmio *mmio, phys_addr_t offset)
{
u32 reg;
vgic_reg_access(mmio, &reg, offset,
ACCESS_READ_RAZ | ACCESS_WRITE_VALUE);
if (mmio->is_write) {
vgic_dispatch_sgi(vcpu, reg);
vgic_update_state(vcpu->kvm);
return true;
}
return false;
}
/*
* I would have liked to use the kvm_bus_io_*() API instead, but it
* cannot cope with banked registers (only the VM pointer is passed
@ -119,6 +554,66 @@ struct mmio_range {
};
static const struct mmio_range vgic_ranges[] = {
{
.base = GIC_DIST_CTRL,
.len = 12,
.handle_mmio = handle_mmio_misc,
},
{
.base = GIC_DIST_IGROUP,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_raz_wi,
},
{
.base = GIC_DIST_ENABLE_SET,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_set_enable_reg,
},
{
.base = GIC_DIST_ENABLE_CLEAR,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_clear_enable_reg,
},
{
.base = GIC_DIST_PENDING_SET,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_set_pending_reg,
},
{
.base = GIC_DIST_PENDING_CLEAR,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_clear_pending_reg,
},
{
.base = GIC_DIST_ACTIVE_SET,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_raz_wi,
},
{
.base = GIC_DIST_ACTIVE_CLEAR,
.len = VGIC_NR_IRQS / 8,
.handle_mmio = handle_mmio_raz_wi,
},
{
.base = GIC_DIST_PRI,
.len = VGIC_NR_IRQS,
.handle_mmio = handle_mmio_priority_reg,
},
{
.base = GIC_DIST_TARGET,
.len = VGIC_NR_IRQS,
.handle_mmio = handle_mmio_target_reg,
},
{
.base = GIC_DIST_CONFIG,
.len = VGIC_NR_IRQS / 4,
.handle_mmio = handle_mmio_cfg_reg,
},
{
.base = GIC_DIST_SOFTINT,
.len = 4,
.handle_mmio = handle_mmio_sgi_reg,
},
{}
};
@ -152,7 +647,106 @@ struct mmio_range *find_matching_range(const struct mmio_range *ranges,
bool vgic_handle_mmio(struct kvm_vcpu *vcpu, struct kvm_run *run,
struct kvm_exit_mmio *mmio)
{
return KVM_EXIT_MMIO;
const struct mmio_range *range;
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
unsigned long base = dist->vgic_dist_base;
bool updated_state;
unsigned long offset;
if (!irqchip_in_kernel(vcpu->kvm) ||
mmio->phys_addr < base ||
(mmio->phys_addr + mmio->len) > (base + KVM_VGIC_V2_DIST_SIZE))
return false;
/* We don't support ldrd / strd or ldm / stm to the emulated vgic */
if (mmio->len > 4) {
kvm_inject_dabt(vcpu, mmio->phys_addr);
return true;
}
range = find_matching_range(vgic_ranges, mmio, base);
if (unlikely(!range || !range->handle_mmio)) {
pr_warn("Unhandled access %d %08llx %d\n",
mmio->is_write, mmio->phys_addr, mmio->len);
return false;
}
spin_lock(&vcpu->kvm->arch.vgic.lock);
offset = mmio->phys_addr - range->base - base;
updated_state = range->handle_mmio(vcpu, mmio, offset);
spin_unlock(&vcpu->kvm->arch.vgic.lock);
kvm_prepare_mmio(run, mmio);
kvm_handle_mmio_return(vcpu, run);
return true;
}
static void vgic_dispatch_sgi(struct kvm_vcpu *vcpu, u32 reg)
{
struct kvm *kvm = vcpu->kvm;
struct vgic_dist *dist = &kvm->arch.vgic;
int nrcpus = atomic_read(&kvm->online_vcpus);
u8 target_cpus;
int sgi, mode, c, vcpu_id;
vcpu_id = vcpu->vcpu_id;
sgi = reg & 0xf;
target_cpus = (reg >> 16) & 0xff;
mode = (reg >> 24) & 3;
switch (mode) {
case 0:
if (!target_cpus)
return;
case 1:
target_cpus = ((1 << nrcpus) - 1) & ~(1 << vcpu_id) & 0xff;
break;
case 2:
target_cpus = 1 << vcpu_id;
break;
}
kvm_for_each_vcpu(c, vcpu, kvm) {
if (target_cpus & 1) {
/* Flag the SGI as pending */
vgic_dist_irq_set(vcpu, sgi);
dist->irq_sgi_sources[c][sgi] |= 1 << vcpu_id;
kvm_debug("SGI%d from CPU%d to CPU%d\n", sgi, vcpu_id, c);
}
target_cpus >>= 1;
}
}
static int compute_pending_for_cpu(struct kvm_vcpu *vcpu)
{
return 0;
}
/*
* Update the interrupt state and determine which CPUs have pending
* interrupts. Must be called with distributor lock held.
*/
static void vgic_update_state(struct kvm *kvm)
{
struct vgic_dist *dist = &kvm->arch.vgic;
struct kvm_vcpu *vcpu;
int c;
if (!dist->enabled) {
set_bit(0, &dist->irq_pending_on_cpu);
return;
}
kvm_for_each_vcpu(c, vcpu, kvm) {
if (compute_pending_for_cpu(vcpu)) {
pr_debug("CPU%d has pending interrupts\n", c);
set_bit(c, &dist->irq_pending_on_cpu);
}
}
}
static bool vgic_ioaddr_overlap(struct kvm *kvm)