qemu/hw/ppc/spapr_irq.c
Greg Kurz c55bcb1f47 spapr: Clarify error and documentation for broken KVM XICS
When starting an L2 KVM guest with `ic-mode=dual,kernel-irqchip=on`,
QEMU fails with:

KVM is too old to support ic-mode=dual,kernel-irqchip=on

This error message was introduced to detect older KVM versions that
didn't allow destruction and re-creation of the XICS KVM device that
we do at reboot. But it is actually the same issue that we get with
nested guests : when running under pseries, KVM currently provides
a genuine XICS device (not the XICS-on-XIVE device that we get
under powernv) which doesn't support destruction/re-creation.

This will eventually be fixed in KVM but in the meantime, update
the error message and documentation to mention the nested case.
While here, mention that in "No XIVE support in KVM" section that
this can also happen with "guest OSes supporting XIVE" since
we check this at init time before starting the guest.

Reported-by: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
Buglink: https://bugs.launchpad.net/qemu/+bug/1890290
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <159664243614.622889.18307368735989783528.stgit@bahia.lan>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
2020-08-12 13:16:27 +10:00

600 lines
16 KiB
C

/*
* QEMU PowerPC sPAPR IRQ interface
*
* Copyright (c) 2018, IBM Corporation.
*
* This code is licensed under the GPL version 2 or later. See the
* COPYING file in the top-level directory.
*/
#include "qemu/osdep.h"
#include "qemu/log.h"
#include "qemu/error-report.h"
#include "qapi/error.h"
#include "hw/irq.h"
#include "hw/ppc/spapr.h"
#include "hw/ppc/spapr_cpu_core.h"
#include "hw/ppc/spapr_xive.h"
#include "hw/ppc/xics.h"
#include "hw/ppc/xics_spapr.h"
#include "hw/qdev-properties.h"
#include "cpu-models.h"
#include "sysemu/kvm.h"
#include "trace.h"
static const TypeInfo spapr_intc_info = {
.name = TYPE_SPAPR_INTC,
.parent = TYPE_INTERFACE,
.class_size = sizeof(SpaprInterruptControllerClass),
};
static void spapr_irq_msi_init(SpaprMachineState *spapr)
{
if (SPAPR_MACHINE_GET_CLASS(spapr)->legacy_irq_allocation) {
/* Legacy mode doesn't use this allocator */
return;
}
spapr->irq_map_nr = spapr_irq_nr_msis(spapr);
spapr->irq_map = bitmap_new(spapr->irq_map_nr);
}
int spapr_irq_msi_alloc(SpaprMachineState *spapr, uint32_t num, bool align,
Error **errp)
{
int irq;
/*
* The 'align_mask' parameter of bitmap_find_next_zero_area()
* should be one less than a power of 2; 0 means no
* alignment. Adapt the 'align' value of the former allocator
* to fit the requirements of bitmap_find_next_zero_area()
*/
align -= 1;
irq = bitmap_find_next_zero_area(spapr->irq_map, spapr->irq_map_nr, 0, num,
align);
if (irq == spapr->irq_map_nr) {
error_setg(errp, "can't find a free %d-IRQ block", num);
return -1;
}
bitmap_set(spapr->irq_map, irq, num);
return irq + SPAPR_IRQ_MSI;
}
void spapr_irq_msi_free(SpaprMachineState *spapr, int irq, uint32_t num)
{
bitmap_clear(spapr->irq_map, irq - SPAPR_IRQ_MSI, num);
}
int spapr_irq_init_kvm(SpaprInterruptControllerInitKvm fn,
SpaprInterruptController *intc,
uint32_t nr_servers,
Error **errp)
{
Error *local_err = NULL;
if (kvm_enabled() && kvm_kernel_irqchip_allowed()) {
if (fn(intc, nr_servers, &local_err) < 0) {
if (kvm_kernel_irqchip_required()) {
error_prepend(&local_err,
"kernel_irqchip requested but unavailable: ");
error_propagate(errp, local_err);
return -1;
}
/*
* We failed to initialize the KVM device, fallback to
* emulated mode
*/
error_prepend(&local_err,
"kernel_irqchip allowed but unavailable: ");
error_append_hint(&local_err,
"Falling back to kernel-irqchip=off\n");
warn_report_err(local_err);
}
}
return 0;
}
/*
* XICS IRQ backend.
*/
SpaprIrq spapr_irq_xics = {
.xics = true,
.xive = false,
};
/*
* XIVE IRQ backend.
*/
SpaprIrq spapr_irq_xive = {
.xics = false,
.xive = true,
};
/*
* Dual XIVE and XICS IRQ backend.
*
* Both interrupt mode, XIVE and XICS, objects are created but the
* machine starts in legacy interrupt mode (XICS). It can be changed
* by the CAS negotiation process and, in that case, the new mode is
* activated after an extra machine reset.
*/
/*
* Define values in sync with the XIVE and XICS backend
*/
SpaprIrq spapr_irq_dual = {
.xics = true,
.xive = true,
};
static int spapr_irq_check(SpaprMachineState *spapr, Error **errp)
{
ERRP_GUARD();
MachineState *machine = MACHINE(spapr);
/*
* Sanity checks on non-P9 machines. On these, XIVE is not
* advertised, see spapr_dt_ov5_platform_support()
*/
if (!ppc_type_check_compat(machine->cpu_type, CPU_POWERPC_LOGICAL_3_00,
0, spapr->max_compat_pvr)) {
/*
* If the 'dual' interrupt mode is selected, force XICS as CAS
* negotiation is useless.
*/
if (spapr->irq == &spapr_irq_dual) {
spapr->irq = &spapr_irq_xics;
return 0;
}
/*
* Non-P9 machines using only XIVE is a bogus setup. We have two
* scenarios to take into account because of the compat mode:
*
* 1. POWER7/8 machines should fail to init later on when creating
* the XIVE interrupt presenters because a POWER9 exception
* model is required.
* 2. POWER9 machines using the POWER8 compat mode won't fail and
* will let the OS boot with a partial XIVE setup : DT
* properties but no hcalls.
*
* To cover both and not confuse the OS, add an early failure in
* QEMU.
*/
if (spapr->irq == &spapr_irq_xive) {
error_setg(errp, "XIVE-only machines require a POWER9 CPU");
return -1;
}
}
/*
* On a POWER9 host, some older KVM XICS devices cannot be destroyed and
* re-created. Same happens with KVM nested guests. Detect that early to
* avoid QEMU to exit later when the guest reboots.
*/
if (kvm_enabled() &&
spapr->irq == &spapr_irq_dual &&
kvm_kernel_irqchip_required() &&
xics_kvm_has_broken_disconnect(spapr)) {
error_setg(errp,
"KVM is incompatible with ic-mode=dual,kernel-irqchip=on");
error_append_hint(errp,
"This can happen with an old KVM or in a KVM nested guest.\n");
error_append_hint(errp,
"Try without kernel-irqchip or with kernel-irqchip=off.\n");
return -1;
}
return 0;
}
/*
* sPAPR IRQ frontend routines for devices
*/
#define ALL_INTCS(spapr_) \
{ SPAPR_INTC((spapr_)->ics), SPAPR_INTC((spapr_)->xive), }
int spapr_irq_cpu_intc_create(SpaprMachineState *spapr,
PowerPCCPU *cpu, Error **errp)
{
SpaprInterruptController *intcs[] = ALL_INTCS(spapr);
int i;
int rc;
for (i = 0; i < ARRAY_SIZE(intcs); i++) {
SpaprInterruptController *intc = intcs[i];
if (intc) {
SpaprInterruptControllerClass *sicc = SPAPR_INTC_GET_CLASS(intc);
rc = sicc->cpu_intc_create(intc, cpu, errp);
if (rc < 0) {
return rc;
}
}
}
return 0;
}
void spapr_irq_cpu_intc_reset(SpaprMachineState *spapr, PowerPCCPU *cpu)
{
SpaprInterruptController *intcs[] = ALL_INTCS(spapr);
int i;
for (i = 0; i < ARRAY_SIZE(intcs); i++) {
SpaprInterruptController *intc = intcs[i];
if (intc) {
SpaprInterruptControllerClass *sicc = SPAPR_INTC_GET_CLASS(intc);
sicc->cpu_intc_reset(intc, cpu);
}
}
}
void spapr_irq_cpu_intc_destroy(SpaprMachineState *spapr, PowerPCCPU *cpu)
{
SpaprInterruptController *intcs[] = ALL_INTCS(spapr);
int i;
for (i = 0; i < ARRAY_SIZE(intcs); i++) {
SpaprInterruptController *intc = intcs[i];
if (intc) {
SpaprInterruptControllerClass *sicc = SPAPR_INTC_GET_CLASS(intc);
sicc->cpu_intc_destroy(intc, cpu);
}
}
}
static void spapr_set_irq(void *opaque, int irq, int level)
{
SpaprMachineState *spapr = SPAPR_MACHINE(opaque);
SpaprInterruptControllerClass *sicc
= SPAPR_INTC_GET_CLASS(spapr->active_intc);
sicc->set_irq(spapr->active_intc, irq, level);
}
void spapr_irq_print_info(SpaprMachineState *spapr, Monitor *mon)
{
SpaprInterruptControllerClass *sicc
= SPAPR_INTC_GET_CLASS(spapr->active_intc);
sicc->print_info(spapr->active_intc, mon);
}
void spapr_irq_dt(SpaprMachineState *spapr, uint32_t nr_servers,
void *fdt, uint32_t phandle)
{
SpaprInterruptControllerClass *sicc
= SPAPR_INTC_GET_CLASS(spapr->active_intc);
sicc->dt(spapr->active_intc, nr_servers, fdt, phandle);
}
uint32_t spapr_irq_nr_msis(SpaprMachineState *spapr)
{
SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
if (smc->legacy_irq_allocation) {
return smc->nr_xirqs;
} else {
return SPAPR_XIRQ_BASE + smc->nr_xirqs - SPAPR_IRQ_MSI;
}
}
void spapr_irq_init(SpaprMachineState *spapr, Error **errp)
{
SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
if (kvm_enabled() && kvm_kernel_irqchip_split()) {
error_setg(errp, "kernel_irqchip split mode not supported on pseries");
return;
}
if (spapr_irq_check(spapr, errp) < 0) {
return;
}
/* Initialize the MSI IRQ allocator. */
spapr_irq_msi_init(spapr);
if (spapr->irq->xics) {
Object *obj;
obj = object_new(TYPE_ICS_SPAPR);
object_property_add_child(OBJECT(spapr), "ics", obj);
object_property_set_link(obj, ICS_PROP_XICS, OBJECT(spapr),
&error_abort);
object_property_set_int(obj, "nr-irqs", smc->nr_xirqs, &error_abort);
if (!qdev_realize(DEVICE(obj), NULL, errp)) {
return;
}
spapr->ics = ICS_SPAPR(obj);
}
if (spapr->irq->xive) {
uint32_t nr_servers = spapr_max_server_number(spapr);
DeviceState *dev;
int i;
dev = qdev_new(TYPE_SPAPR_XIVE);
qdev_prop_set_uint32(dev, "nr-irqs", smc->nr_xirqs + SPAPR_XIRQ_BASE);
/*
* 8 XIVE END structures per CPU. One for each available
* priority
*/
qdev_prop_set_uint32(dev, "nr-ends", nr_servers << 3);
object_property_set_link(OBJECT(dev), "xive-fabric", OBJECT(spapr),
&error_abort);
sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
spapr->xive = SPAPR_XIVE(dev);
/* Enable the CPU IPIs */
for (i = 0; i < nr_servers; ++i) {
SpaprInterruptControllerClass *sicc
= SPAPR_INTC_GET_CLASS(spapr->xive);
if (sicc->claim_irq(SPAPR_INTC(spapr->xive), SPAPR_IRQ_IPI + i,
false, errp) < 0) {
return;
}
}
spapr_xive_hcall_init(spapr);
}
spapr->qirqs = qemu_allocate_irqs(spapr_set_irq, spapr,
smc->nr_xirqs + SPAPR_XIRQ_BASE);
/*
* Mostly we don't actually need this until reset, except that not
* having this set up can cause VFIO devices to issue a
* false-positive warning during realize(), because they don't yet
* have an in-kernel irq chip.
*/
spapr_irq_update_active_intc(spapr);
}
int spapr_irq_claim(SpaprMachineState *spapr, int irq, bool lsi, Error **errp)
{
SpaprInterruptController *intcs[] = ALL_INTCS(spapr);
int i;
SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
int rc;
assert(irq >= SPAPR_XIRQ_BASE);
assert(irq < (smc->nr_xirqs + SPAPR_XIRQ_BASE));
for (i = 0; i < ARRAY_SIZE(intcs); i++) {
SpaprInterruptController *intc = intcs[i];
if (intc) {
SpaprInterruptControllerClass *sicc = SPAPR_INTC_GET_CLASS(intc);
rc = sicc->claim_irq(intc, irq, lsi, errp);
if (rc < 0) {
return rc;
}
}
}
return 0;
}
void spapr_irq_free(SpaprMachineState *spapr, int irq, int num)
{
SpaprInterruptController *intcs[] = ALL_INTCS(spapr);
int i, j;
SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
assert(irq >= SPAPR_XIRQ_BASE);
assert((irq + num) <= (smc->nr_xirqs + SPAPR_XIRQ_BASE));
for (i = irq; i < (irq + num); i++) {
for (j = 0; j < ARRAY_SIZE(intcs); j++) {
SpaprInterruptController *intc = intcs[j];
if (intc) {
SpaprInterruptControllerClass *sicc
= SPAPR_INTC_GET_CLASS(intc);
sicc->free_irq(intc, i);
}
}
}
}
qemu_irq spapr_qirq(SpaprMachineState *spapr, int irq)
{
SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
/*
* This interface is basically for VIO and PHB devices to find the
* right qemu_irq to manipulate, so we only allow access to the
* external irqs for now. Currently anything which needs to
* access the IPIs most naturally gets there via the guest side
* interfaces, we can change this if we need to in future.
*/
assert(irq >= SPAPR_XIRQ_BASE);
assert(irq < (smc->nr_xirqs + SPAPR_XIRQ_BASE));
if (spapr->ics) {
assert(ics_valid_irq(spapr->ics, irq));
}
if (spapr->xive) {
assert(irq < spapr->xive->nr_irqs);
assert(xive_eas_is_valid(&spapr->xive->eat[irq]));
}
return spapr->qirqs[irq];
}
int spapr_irq_post_load(SpaprMachineState *spapr, int version_id)
{
SpaprInterruptControllerClass *sicc;
spapr_irq_update_active_intc(spapr);
sicc = SPAPR_INTC_GET_CLASS(spapr->active_intc);
return sicc->post_load(spapr->active_intc, version_id);
}
void spapr_irq_reset(SpaprMachineState *spapr, Error **errp)
{
assert(!spapr->irq_map || bitmap_empty(spapr->irq_map, spapr->irq_map_nr));
spapr_irq_update_active_intc(spapr);
}
int spapr_irq_get_phandle(SpaprMachineState *spapr, void *fdt, Error **errp)
{
const char *nodename = "interrupt-controller";
int offset, phandle;
offset = fdt_subnode_offset(fdt, 0, nodename);
if (offset < 0) {
error_setg(errp, "Can't find node \"%s\": %s",
nodename, fdt_strerror(offset));
return -1;
}
phandle = fdt_get_phandle(fdt, offset);
if (!phandle) {
error_setg(errp, "Can't get phandle of node \"%s\"", nodename);
return -1;
}
return phandle;
}
static void set_active_intc(SpaprMachineState *spapr,
SpaprInterruptController *new_intc)
{
SpaprInterruptControllerClass *sicc;
uint32_t nr_servers = spapr_max_server_number(spapr);
assert(new_intc);
if (new_intc == spapr->active_intc) {
/* Nothing to do */
return;
}
if (spapr->active_intc) {
sicc = SPAPR_INTC_GET_CLASS(spapr->active_intc);
if (sicc->deactivate) {
sicc->deactivate(spapr->active_intc);
}
}
sicc = SPAPR_INTC_GET_CLASS(new_intc);
if (sicc->activate) {
sicc->activate(new_intc, nr_servers, &error_fatal);
}
spapr->active_intc = new_intc;
/*
* We've changed the kernel irqchip, let VFIO devices know they
* need to readjust.
*/
kvm_irqchip_change_notify();
}
void spapr_irq_update_active_intc(SpaprMachineState *spapr)
{
SpaprInterruptController *new_intc;
if (!spapr->ics) {
/*
* XXX before we run CAS, ov5_cas is initialized empty, which
* indicates XICS, even if we have ic-mode=xive. TODO: clean
* up the CAS path so that we have a clearer way of handling
* this.
*/
new_intc = SPAPR_INTC(spapr->xive);
} else if (spapr->ov5_cas
&& spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
new_intc = SPAPR_INTC(spapr->xive);
} else {
new_intc = SPAPR_INTC(spapr->ics);
}
set_active_intc(spapr, new_intc);
}
/*
* XICS legacy routines - to deprecate one day
*/
static int ics_find_free_block(ICSState *ics, int num, int alignnum)
{
int first, i;
for (first = 0; first < ics->nr_irqs; first += alignnum) {
if (num > (ics->nr_irqs - first)) {
return -1;
}
for (i = first; i < first + num; ++i) {
if (!ics_irq_free(ics, i)) {
break;
}
}
if (i == (first + num)) {
return first;
}
}
return -1;
}
int spapr_irq_find(SpaprMachineState *spapr, int num, bool align, Error **errp)
{
ICSState *ics = spapr->ics;
int first = -1;
assert(ics);
/*
* MSIMesage::data is used for storing VIRQ so
* it has to be aligned to num to support multiple
* MSI vectors. MSI-X is not affected by this.
* The hint is used for the first IRQ, the rest should
* be allocated continuously.
*/
if (align) {
assert((num == 1) || (num == 2) || (num == 4) ||
(num == 8) || (num == 16) || (num == 32));
first = ics_find_free_block(ics, num, num);
} else {
first = ics_find_free_block(ics, num, 1);
}
if (first < 0) {
error_setg(errp, "can't find a free %d-IRQ block", num);
return -1;
}
return first + ics->offset;
}
SpaprIrq spapr_irq_xics_legacy = {
.xics = true,
.xive = false,
};
static void spapr_irq_register_types(void)
{
type_register_static(&spapr_intc_info);
}
type_init(spapr_irq_register_types)