* [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests
@ 2016-12-09 2:23 David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type David Gibson
` (4 more replies)
0 siblings, 5 replies; 18+ messages in thread
From: David Gibson @ 2016-12-09 2:23 UTC (permalink / raw)
To: paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth, lvivier, David Gibson
This series implements the host side of the PAPR ACR to allow runtime
resizing of the Hashed Page Table (HPT) for pseries guests.
Exercising this feature requires a guest OS which is also aware of it.
Patches to implement the guest side in Linux have just been submitted
upstream:
https://lists.ozlabs.org/pipermail/linuxppc-dev/2016-December/152164.html
Availability of the feature is controlled by a new 'resize-hpt'
machine option: it can be set to "disabled", "enabled" or "required".
The last means that qemu will refuse to boot a guest which is not
aware of the HPT resizing feature (it will instead quit during feature
negotiation). This is potentially useful because guests which don't
support resizing will need an HPT sized for their maximum possible
RAM, which can be very wasteful of host resources.
So far this is only tested for TCG guests. PR KVM guests should work
in theory, but I haven't been able to test yet due to (probably
unrelated) problems booting a suitable guest kernel. HV KVM guests
will require substantial extra work in the host kernel; development is
in progress.
David Gibson (5):
pseries: Add pseries-2.9 machine type
pseries: Stubs for HPT resizing
pseries: Implement HPT resizing
pseries: Enable HPT resizing for 2.9
pseries: Use smaller default hash page tables when guest can resize
hw/ppc/spapr.c | 117 +++++++++++--
hw/ppc/spapr_hcall.c | 408 ++++++++++++++++++++++++++++++++++++++++++++
hw/ppc/trace-events | 2 +
include/hw/compat.h | 3 +
include/hw/ppc/spapr.h | 19 +++
include/hw/ppc/spapr_ovec.h | 1 +
target-ppc/kvm.c | 26 +++
target-ppc/kvm_ppc.h | 5 +
target-ppc/mmu-hash64.h | 4 +
9 files changed, 575 insertions(+), 10 deletions(-)
--
2.9.3
^ permalink raw reply [flat|nested] 18+ messages in thread
* [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type
2016-12-09 2:23 [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests David Gibson
@ 2016-12-09 2:23 ` David Gibson
2016-12-09 7:59 ` Thomas Huth
2016-12-09 13:25 ` Laurent Vivier
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing David Gibson
` (3 subsequent siblings)
4 siblings, 2 replies; 18+ messages in thread
From: David Gibson @ 2016-12-09 2:23 UTC (permalink / raw)
To: paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth, lvivier, David Gibson
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
hw/ppc/spapr.c | 23 +++++++++++++++++++++--
include/hw/compat.h | 3 +++
2 files changed, 24 insertions(+), 2 deletions(-)
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index cfadc46..0f25e83 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -2761,18 +2761,37 @@ static const TypeInfo spapr_machine_info = {
type_init(spapr_machine_register_##suffix)
/*
+ * pseries-2.9
+ */
+static void spapr_machine_2_9_instance_options(MachineState *machine)
+{
+}
+
+static void spapr_machine_2_9_class_options(MachineClass *mc)
+{
+ /* Defaults for the latest behaviour inherited from the base class */
+}
+
+DEFINE_SPAPR_MACHINE(2_9, "2.9", true);
+
+/*
* pseries-2.8
*/
+#define SPAPR_COMPAT_2_8 \
+ HW_COMPAT_2_8
+
static void spapr_machine_2_8_instance_options(MachineState *machine)
{
+ spapr_machine_2_9_instance_options(machine);
}
static void spapr_machine_2_8_class_options(MachineClass *mc)
{
- /* Defaults for the latest behaviour inherited from the base class */
+ spapr_machine_2_9_class_options(mc);
+ SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8);
}
-DEFINE_SPAPR_MACHINE(2_8, "2.8", true);
+DEFINE_SPAPR_MACHINE(2_8, "2.8", false);
/*
* pseries-2.7
diff --git a/include/hw/compat.h b/include/hw/compat.h
index 0f06e11..f8b8354 100644
--- a/include/hw/compat.h
+++ b/include/hw/compat.h
@@ -1,6 +1,9 @@
#ifndef HW_COMPAT_H
#define HW_COMPAT_H
+#define HW_COMPAT_2_8 \
+ /* empty */
+
#define HW_COMPAT_2_7 \
{\
.driver = "virtio-pci",\
--
2.9.3
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 2:23 [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type David Gibson
@ 2016-12-09 2:23 ` David Gibson
2016-12-09 8:18 ` Thomas Huth
2016-12-09 17:08 ` Michael Roth
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 3/5] pseries: Implement " David Gibson
` (2 subsequent siblings)
4 siblings, 2 replies; 18+ messages in thread
From: David Gibson @ 2016-12-09 2:23 UTC (permalink / raw)
To: paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth, lvivier, David Gibson
This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
extension to allow run time resizing of a guest's hash page table. It
also adds a new machine property for controlling whether this new
facility is available, and logic to check that against availability
with KVM (only supported with KVM PR for now).
Finally, it adds a new string to the hypertas property in the device
tree, advertising to the guest the availability of the HPT resizing
hypercalls. This is a tentative suggested value, and would need to be
standardized by PAPR before being merged.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
hw/ppc/trace-events | 2 ++
include/hw/ppc/spapr.h | 11 +++++++++
target-ppc/kvm.c | 26 ++++++++++++++++++++
target-ppc/kvm_ppc.h | 5 ++++
6 files changed, 146 insertions(+)
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index 0f25e83..ecb0822 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
add_str(hypertas, "hcall-multi-tce");
}
+
+ if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
+ add_str(hypertas, "hcall-hpt-resize");
+ }
+
_FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
hypertas->str, hypertas->len));
g_string_free(hypertas, TRUE);
@@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
long load_limit, fw_size;
char *filename;
int smt = kvmppc_smt_threads();
+ Error *resize_hpt_err = NULL;
msi_nonbroken = true;
QLIST_INIT(&spapr->phbs);
+ /* Check HPT resizing availability */
+ kvmppc_check_papr_resize_hpt(&resize_hpt_err);
+ if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
+ if (resize_hpt_err) {
+ spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
+ error_free(resize_hpt_err);
+ resize_hpt_err = NULL;
+ } else {
+ spapr->resize_hpt = smc->resize_hpt_default;
+ }
+ }
+
+ assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
+
+ if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
+ error_report_err(resize_hpt_err);
+ exit(1);
+ }
+
/* Allocate RMA if necessary */
rma_alloc_size = kvmppc_alloc_rma(&rma);
@@ -2236,6 +2261,40 @@ static void spapr_set_modern_hotplug_events(Object *obj, bool value,
spapr->use_hotplug_event_source = value;
}
+static char *spapr_get_resize_hpt(Object *obj, Error **errp)
+{
+ sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
+
+ switch (spapr->resize_hpt) {
+ case SPAPR_RESIZE_HPT_DEFAULT:
+ return g_strdup("default");
+ case SPAPR_RESIZE_HPT_DISABLED:
+ return g_strdup("disabled");
+ case SPAPR_RESIZE_HPT_ENABLED:
+ return g_strdup("enabled");
+ case SPAPR_RESIZE_HPT_REQUIRED:
+ return g_strdup("required");
+ }
+ assert(0);
+}
+
+static void spapr_set_resize_hpt(Object *obj, const char *value, Error **errp)
+{
+ sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
+
+ if (strcmp(value, "default") == 0) {
+ spapr->resize_hpt = SPAPR_RESIZE_HPT_DEFAULT;
+ } else if (strcmp(value, "disabled") == 0) {
+ spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
+ } else if (strcmp(value, "enabled") == 0) {
+ spapr->resize_hpt = SPAPR_RESIZE_HPT_ENABLED;
+ } else if (strcmp(value, "required") == 0) {
+ spapr->resize_hpt = SPAPR_RESIZE_HPT_REQUIRED;
+ } else {
+ error_setg(errp, "Bad value for \"resize-hpt\" property");
+ }
+}
+
static void spapr_machine_initfn(Object *obj)
{
sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
@@ -2256,6 +2315,12 @@ static void spapr_machine_initfn(Object *obj)
" place of standard EPOW events when possible"
" (required for memory hot-unplug support)",
NULL);
+
+ object_property_add_str(obj, "resize-hpt",
+ spapr_get_resize_hpt, spapr_set_resize_hpt, NULL);
+ object_property_set_description(obj, "resize-hpt",
+ "Resizing of the Hash Page Table (enabled, disabled, required)",
+ NULL);
}
static void spapr_machine_finalizefn(Object *obj)
@@ -2707,6 +2772,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
smc->dr_lmb_enabled = true;
smc->tcg_default_cpu = "POWER8";
+ smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
fwc->get_dev_path = spapr_get_fw_dev_path;
nc->nmi_monitor_handler = spapr_nmi;
diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
index fd9f1d4..72a9c4d 100644
--- a/hw/ppc/spapr_hcall.c
+++ b/hw/ppc/spapr_hcall.c
@@ -355,6 +355,38 @@ static target_ulong h_read(PowerPCCPU *cpu, sPAPRMachineState *spapr,
return H_SUCCESS;
}
+static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
+ sPAPRMachineState *spapr,
+ target_ulong opcode,
+ target_ulong *args)
+{
+ target_ulong flags = args[0];
+ target_ulong shift = args[1];
+
+ if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
+ return H_AUTHORITY;
+ }
+
+ trace_spapr_h_resize_hpt_prepare(flags, shift);
+ return H_HARDWARE;
+}
+
+static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
+ sPAPRMachineState *spapr,
+ target_ulong opcode,
+ target_ulong *args)
+{
+ target_ulong flags = args[0];
+ target_ulong shift = args[1];
+
+ if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
+ return H_AUTHORITY;
+ }
+
+ trace_spapr_h_resize_hpt_commit(flags, shift);
+ return H_HARDWARE;
+}
+
static target_ulong h_set_sprg0(PowerPCCPU *cpu, sPAPRMachineState *spapr,
target_ulong opcode, target_ulong *args)
{
@@ -1134,6 +1166,10 @@ static void hypercall_register_types(void)
/* hcall-bulk */
spapr_register_hypercall(H_BULK_REMOVE, h_bulk_remove);
+ /* hcall-hpt-resize */
+ spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
+ spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
+
/* hcall-splpar */
spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
spapr_register_hypercall(H_CEDE, h_cede);
diff --git a/hw/ppc/trace-events b/hw/ppc/trace-events
index 2297ead..bf59a8f 100644
--- a/hw/ppc/trace-events
+++ b/hw/ppc/trace-events
@@ -16,6 +16,8 @@ spapr_cas_continue(unsigned long n) "Copy changes to the guest: %ld bytes"
# hw/ppc/spapr_hcall.c
spapr_cas_pvr_try(uint32_t pvr) "%x"
spapr_cas_pvr(uint32_t cur_pvr, bool cpu_match, uint32_t new_pvr, uint64_t pcr) "current=%x, cpu_match=%u, new=%x, compat flags=%"PRIx64
+spapr_h_resize_hpt_prepare(uint64_t flags, uint64_t shift) "flags=0x%"PRIx64", shift=%"PRIu64
+spapr_h_resize_hpt_commit(uint64_t flags, uint64_t shift) "flags=0x%"PRIx64", shift=%"PRIu64
# hw/ppc/spapr_iommu.c
spapr_iommu_put(uint64_t liobn, uint64_t ioba, uint64_t tce, uint64_t ret) "liobn=%"PRIx64" ioba=0x%"PRIx64" tce=0x%"PRIx64" ret=%"PRId64
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index a2d8964..d2c758b 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -31,6 +31,13 @@ typedef struct sPAPRMachineState sPAPRMachineState;
#define SPAPR_MACHINE_CLASS(klass) \
OBJECT_CLASS_CHECK(sPAPRMachineClass, klass, TYPE_SPAPR_MACHINE)
+typedef enum {
+ SPAPR_RESIZE_HPT_DEFAULT = 0,
+ SPAPR_RESIZE_HPT_DISABLED,
+ SPAPR_RESIZE_HPT_ENABLED,
+ SPAPR_RESIZE_HPT_REQUIRED,
+} sPAPRResizeHPT;
+
/**
* sPAPRMachineClass:
*/
@@ -46,6 +53,7 @@ struct sPAPRMachineClass {
uint64_t *buid, hwaddr *pio,
hwaddr *mmio32, hwaddr *mmio64,
unsigned n_dma, uint32_t *liobns, Error **errp);
+ sPAPRResizeHPT resize_hpt_default;
};
/**
@@ -61,6 +69,7 @@ struct sPAPRMachineState {
XICSState *xics;
DeviceState *rtc;
+ sPAPRResizeHPT resize_hpt;
void *htab;
uint32_t htab_shift;
hwaddr rma_size;
@@ -347,6 +356,8 @@ struct sPAPRMachineState {
#define H_XIRR_X 0x2FC
#define H_RANDOM 0x300
#define H_SET_MODE 0x31C
+#define H_RESIZE_HPT_PREPARE 0x36C
+#define H_RESIZE_HPT_COMMIT 0x370
#define H_SIGNAL_SYS_RESET 0x380
#define MAX_HCALL_OPCODE H_SIGNAL_SYS_RESET
diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
index 15e12f3..dcd8cef 100644
--- a/target-ppc/kvm.c
+++ b/target-ppc/kvm.c
@@ -22,6 +22,7 @@
#include <linux/kvm.h>
#include "qemu-common.h"
+#include "qapi/error.h"
#include "qemu/error-report.h"
#include "cpu.h"
#include "qemu/timer.h"
@@ -2269,6 +2270,7 @@ int kvmppc_reset_htab(int shift_hint)
/* Full emulation, tell caller to allocate htab itself */
return 0;
}
+
if (kvm_check_extension(kvm_state, KVM_CAP_PPC_ALLOC_HTAB)) {
int ret;
ret = kvm_vm_ioctl(kvm_state, KVM_PPC_ALLOCATE_HTAB, &shift);
@@ -2672,3 +2674,27 @@ int kvmppc_enable_hwrng(void)
return kvmppc_enable_hcall(kvm_state, H_RANDOM);
}
+
+void kvmppc_check_papr_resize_hpt(Error **errp)
+{
+ if (!kvm_enabled()) {
+ return;
+ }
+
+ /* TODO: Check specific capabilities for HPT resize aware host kernels */
+
+ /*
+ * It's tempting to try to check directly if the HPT is under our
+ * control or KVM's, which is what's really relevant here.
+ * Unfortunately, in order to correctly size the HPT, we need to
+ * know if we can do resizing, _before_ we attempt to allocate it
+ * with KVM. Before that point, we don't officially know whether
+ * we'll control the HPT or not. So we have to use a fallback
+ * test for PR vs HV KVM to predict that.
+ */
+ if (kvmppc_is_pr(kvm_state)) {
+ return;
+ }
+
+ error_setg(errp, "Hash page table resizing not available with this KVM version");
+}
diff --git a/target-ppc/kvm_ppc.h b/target-ppc/kvm_ppc.h
index 841a29b..3e852ba 100644
--- a/target-ppc/kvm_ppc.h
+++ b/target-ppc/kvm_ppc.h
@@ -59,6 +59,7 @@ bool kvmppc_has_cap_htm(void);
int kvmppc_enable_hwrng(void);
int kvmppc_put_books_sregs(PowerPCCPU *cpu);
PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
+void kvmppc_check_papr_resize_hpt(Error **errp);
#else
@@ -270,6 +271,10 @@ static inline PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void)
return NULL;
}
+static inline void kvmppc_check_papr_resize_hpt(Error **errp)
+{
+ return;
+}
#endif
#ifndef CONFIG_KVM
--
2.9.3
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [Qemu-devel] [for-2.9 3/5] pseries: Implement HPT resizing
2016-12-09 2:23 [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing David Gibson
@ 2016-12-09 2:23 ` David Gibson
2016-12-09 8:36 ` Thomas Huth
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 4/5] pseries: Enable HPT resizing for 2.9 David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 5/5] pseries: Use smaller default hash page tables when guest can resize David Gibson
4 siblings, 1 reply; 18+ messages in thread
From: David Gibson @ 2016-12-09 2:23 UTC (permalink / raw)
To: paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth, lvivier, David Gibson
This patch implements hypercalls allowing a PAPR guest to resize its own
hash page table. This will eventually allow for more flexible memory
hotplug.
The implementation is partially asynchronous, handled in a special thread
running the hpt_prepare_thread() function. The state of a pending resize
is stored in SPAPR_MACHINE->pending_hpt.
The H_RESIZE_HPT_PREPARE hypercall will kick off creation of a new HPT, or,
if one is already in progress, monitor it for completion. If there is an
existing HPT resize in progress that doesn't match the size specified in
the call, it will cancel it, replacing it with a new one matching the
given size.
The H_RESIZE_HPT_COMMIT completes transition to a resized HPT, and can only
be called successfully once H_RESIZE_HPT_PREPARE has successfully
completed initialization of a new HPT. The guest must ensure that there
are no concurrent accesses to the existing HPT while this is called (this
effectively means stop_machine() for Linux guests).
For now H_RESIZE_HPT_COMMIT goes through the whole old HPT, rehashing each
HPTE into the new HPT. This can have quite high latency, but it seems to
be of the order of typical migration downtime latencies for HPTs of size
up to ~2GiB (which would be used in a 256GiB guest).
In future we probably want to move more of the rehashing to the "prepare"
phase, by having H_ENTER and other hcalls update both current and
pending HPTs. That's a project for another day, but should be possible
without any changes to the guest interface.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
hw/ppc/spapr.c | 4 +-
hw/ppc/spapr_hcall.c | 346 +++++++++++++++++++++++++++++++++++++++++++++++-
include/hw/ppc/spapr.h | 6 +
target-ppc/mmu-hash64.h | 4 +
4 files changed, 354 insertions(+), 6 deletions(-)
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index ecb0822..f3b74dc 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -93,8 +93,6 @@
#define PHANDLE_XICP 0x00001111
-#define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift))
-
static XICSState *try_create_xics(const char *type, int nr_servers,
int nr_irqs, Error **errp)
{
@@ -1055,7 +1053,7 @@ static void close_htab_fd(sPAPRMachineState *spapr)
spapr->htab_fd = -1;
}
-static int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
+int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
{
int shift;
diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
index 72a9c4d..9f421c1 100644
--- a/hw/ppc/spapr_hcall.c
+++ b/hw/ppc/spapr_hcall.c
@@ -2,6 +2,7 @@
#include "qapi/error.h"
#include "sysemu/sysemu.h"
#include "qemu/log.h"
+#include "qemu/error-report.h"
#include "cpu.h"
#include "exec/exec-all.h"
#include "helper_regs.h"
@@ -355,20 +356,320 @@ static target_ulong h_read(PowerPCCPU *cpu, sPAPRMachineState *spapr,
return H_SUCCESS;
}
+struct sPAPRPendingHPT {
+ /* These fields are read-only after initialization */
+ int shift;
+ QemuThread thread;
+
+ /* These fields are protected by the BQL */
+ bool complete;
+
+ /* These fields are private to the preparation thread if
+ * !complete, otherwise protected by the BQL */
+ int ret;
+ void *hpt;
+};
+
+static void free_pending_hpt(sPAPRPendingHPT *pending)
+{
+ if (pending->hpt) {
+ qemu_vfree(pending->hpt);
+ }
+
+ g_free(pending);
+}
+
+static void *hpt_prepare_thread(void *opaque)
+{
+ sPAPRPendingHPT *pending = opaque;
+ size_t size = 1ULL << pending->shift;
+
+ pending->hpt = qemu_memalign(size, size);
+ if (pending->hpt) {
+ memset(pending->hpt, 0, size);
+ pending->ret = H_SUCCESS;
+ } else {
+ pending->ret = H_NO_MEM;
+ }
+
+ qemu_mutex_lock_iothread();
+
+ if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt != pending) {
+ /* We've been cancelled, clean ourselves up */
+ free_pending_hpt(pending);
+ goto out;
+ }
+
+ pending->complete = true;
+
+out:
+ qemu_mutex_unlock_iothread();
+ return NULL;
+}
+
+/* Must be called with BQL held */
+static void cancel_hpt_prepare(sPAPRMachineState *spapr)
+{
+ sPAPRPendingHPT *pending = spapr->pending_hpt;
+
+ /* Let the thread know it's cancelled */
+ spapr->pending_hpt = NULL;
+
+ if (!pending) {
+ /* Nothing to do */
+ return;
+ }
+
+ if (!pending->complete) {
+ /* thread will clean itself up */
+ return;
+ }
+
+ free_pending_hpt(pending);
+}
+
+static int build_dimm_list(Object *obj, void *opaque)
+{
+ GSList **list = opaque;
+
+ if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
+ DeviceState *dev = DEVICE(obj);
+ if (dev->realized) { /* only realized DIMMs matter */
+ *list = g_slist_prepend(*list, dev);
+ }
+ }
+
+ object_child_foreach(obj, build_dimm_list, opaque);
+ return 0;
+}
+
+static ram_addr_t get_current_ram_size(void)
+{
+ GSList *list = NULL, *item;
+ ram_addr_t size = ram_size;
+
+ build_dimm_list(qdev_get_machine(), &list);
+ for (item = list; item; item = g_slist_next(item)) {
+ Object *obj = OBJECT(item->data);
+ if (!strcmp(object_get_typename(obj), TYPE_PC_DIMM)) {
+ size += object_property_get_int(obj, PC_DIMM_SIZE_PROP,
+ &error_abort);
+ }
+ }
+ g_slist_free(list);
+
+ return size;
+}
+
static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
sPAPRMachineState *spapr,
target_ulong opcode,
target_ulong *args)
{
target_ulong flags = args[0];
- target_ulong shift = args[1];
+ int shift = args[1];
+ sPAPRPendingHPT *pending = spapr->pending_hpt;
if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
return H_AUTHORITY;
}
trace_spapr_h_resize_hpt_prepare(flags, shift);
- return H_HARDWARE;
+
+ if (flags != 0) {
+ return H_PARAMETER;
+ }
+
+ if (shift && ((shift < 18) || (shift > 46))) {
+ return H_PARAMETER;
+ }
+
+ if (pending) {
+ /* something already in progress */
+ if (pending->shift == shift) {
+ /* and it's suitable */
+ if (pending->complete) {
+ return pending->ret;
+ } else {
+ return H_LONG_BUSY_ORDER_100_MSEC;
+ }
+ }
+
+ /* not suitable, cancel and replace */
+ cancel_hpt_prepare(spapr);
+ }
+
+ if (!shift) {
+ /* nothing to do */
+ return H_SUCCESS;
+ }
+
+ /* start new prepare */
+
+ /* We only allow the guest to allocate an HPT one order above what
+ * we'd normally give them (to stop a small guest claiming a huge
+ * chunk of resources in the HPT */
+ if (shift > (spapr_hpt_shift_for_ramsize(get_current_ram_size()) + 1)) {
+ return H_RESOURCE;
+ }
+
+ pending = g_malloc0(sizeof(*pending));
+ pending->shift = shift;
+ pending->ret = H_HARDWARE;
+
+ qemu_thread_create(&pending->thread, "sPAPR HPT prepare",
+ hpt_prepare_thread, pending, QEMU_THREAD_DETACHED);
+
+ spapr->pending_hpt = pending;
+
+ /* In theory we could estimate the time more accurately based on
+ * the new size, but there's not much point */
+ return H_LONG_BUSY_ORDER_100_MSEC;
+}
+
+static uint64_t new_hpte_load0(void *htab, uint64_t pteg, int slot)
+{
+ uint8_t *addr = htab;
+
+ addr += pteg * HASH_PTEG_SIZE_64;
+ addr += slot * HASH_PTE_SIZE_64;
+ return ldq_p(addr);
+}
+
+static void new_hpte_store(void *htab, uint64_t pteg, int slot,
+ uint64_t pte0, uint64_t pte1)
+{
+ uint8_t *addr = htab;
+
+ addr += pteg * HASH_PTEG_SIZE_64;
+ addr += slot * HASH_PTE_SIZE_64;
+
+ stq_p(addr, pte0);
+ stq_p(addr + HASH_PTE_SIZE_64/2, pte1);
+}
+
+static int rehash_hpte(PowerPCCPU *cpu, uint64_t token,
+ void *old, uint64_t oldsize,
+ void *new, uint64_t newsize,
+ uint64_t pteg, int slot)
+{
+ uint64_t old_hash_mask = (oldsize >> 7) - 1;
+ uint64_t new_hash_mask = (newsize >> 7) - 1;
+ target_ulong pte0 = ppc_hash64_load_hpte0(cpu, token, slot);
+ target_ulong pte1;
+ uint64_t avpn;
+ unsigned shift;
+ uint64_t hash, new_pteg, replace_pte0;
+
+ if (!(pte0 & HPTE64_V_VALID) || !(pte0 & HPTE64_V_BOLTED)) {
+ return H_SUCCESS;
+ }
+
+ pte1 = ppc_hash64_load_hpte1(cpu, token, slot);
+
+ shift = ppc_hash64_hpte_page_shift_noslb(cpu, pte0, pte1);
+ assert(shift); /* H_ENTER should never have allowed a bad encoding */
+ avpn = HPTE64_V_AVPN_VAL(pte0) & ~(((1ULL << shift) - 1) >> 23);
+
+ if (pte0 & HPTE64_V_SECONDARY) {
+ pteg = ~pteg;
+ }
+
+ if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_256M) {
+ uint64_t offset, vsid;
+
+ /* We only have 28 - 23 bits of offset in avpn */
+ offset = (avpn & 0x1f) << 23;
+ vsid = avpn >> 5;
+ /* We can find more bits from the pteg value */
+ if (shift < 23) {
+ offset |= ((vsid ^ pteg) & old_hash_mask) << shift;
+ }
+
+ hash = vsid ^ (offset >> shift);
+ } else if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_1T) {
+ uint64_t offset, vsid;
+
+ /* We only have 40 - 23 bits of seg_off in avpn */
+ offset = (avpn & 0x1ffff) << 23;
+ vsid = avpn >> 17;
+ if (shift < 23) {
+ offset |= ((vsid ^ (vsid << 25) ^ pteg) & old_hash_mask) << shift;
+ }
+
+ hash = vsid ^ (vsid << 25) ^ (offset >> shift);
+ } else {
+ error_report("rehash_pte: Bad segment size in HPTE");
+ return H_HARDWARE;
+ }
+
+ new_pteg = hash & new_hash_mask;
+ if (pte0 & HPTE64_V_SECONDARY) {
+ assert(~pteg == (hash & old_hash_mask));
+ new_pteg = ~new_pteg;
+ } else {
+ assert(pteg == (hash & old_hash_mask));
+ }
+ assert((oldsize != newsize) || (pteg == new_pteg));
+ replace_pte0 = new_hpte_load0(new, new_pteg, slot);
+ if (replace_pte0 & HPTE64_V_VALID) {
+ assert(newsize < oldsize);
+ if (replace_pte0 & HPTE64_V_BOLTED) {
+ if (pte0 & HPTE64_V_BOLTED) {
+ /* Bolted collision, nothing we can do */
+ return H_PTEG_FULL;
+ } else {
+ /* Discard this hpte */
+ return H_SUCCESS;
+ }
+ }
+ }
+
+ new_hpte_store(new, new_pteg, slot, pte0, pte1);
+ return H_SUCCESS;
+}
+
+static int rehash_hpt(PowerPCCPU *cpu,
+ void *old, uint64_t oldsize,
+ void *new, uint64_t newsize)
+{
+ CPUPPCState *env = &cpu->env;
+ uint64_t n_ptegs = oldsize >> 7;
+ uint64_t pteg;
+ int slot;
+ int rc;
+
+ assert(env->external_htab == old);
+
+ for (pteg = 0; pteg < n_ptegs; pteg++) {
+ uint64_t token = ppc_hash64_start_access(cpu, pteg * HPTES_PER_GROUP);
+
+ if (!token) {
+ return H_HARDWARE;
+ }
+
+ for (slot = 0; slot < HPTES_PER_GROUP; slot++) {
+ rc = rehash_hpte(cpu, token, old, oldsize, new, newsize,
+ pteg, slot);
+ if (rc != H_SUCCESS) {
+ ppc_hash64_stop_access(cpu, token);
+ return rc;
+ }
+ }
+ ppc_hash64_stop_access(cpu, token);
+ }
+
+ return H_SUCCESS;
+}
+
+static void pivot_hpt(CPUState *cs, run_on_cpu_data arg)
+{
+ sPAPRMachineState *spapr = SPAPR_MACHINE(arg.host_ptr);
+ PowerPCCPU *cpu = POWERPC_CPU(cs);
+
+ cpu_synchronize_state(cs);
+ ppc_hash64_set_external_hpt(cpu, spapr->htab, spapr->htab_shift,
+ &error_fatal);
}
static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
@@ -378,13 +679,52 @@ static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
{
target_ulong flags = args[0];
target_ulong shift = args[1];
+ sPAPRPendingHPT *pending = spapr->pending_hpt;
+ int rc;
+ size_t newsize;
if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
return H_AUTHORITY;
}
trace_spapr_h_resize_hpt_commit(flags, shift);
- return H_HARDWARE;
+
+ if (flags != 0) {
+ return H_PARAMETER;
+ }
+
+ if (!pending || (pending->shift != shift)) {
+ /* no matching prepare */
+ return H_CLOSED;
+ }
+
+ if (!pending->complete) {
+ /* prepare has not completed */
+ return H_BUSY;
+ }
+
+ newsize = 1ULL << pending->shift;
+ rc = rehash_hpt(cpu, spapr->htab, HTAB_SIZE(spapr),
+ pending->hpt, newsize);
+ if (rc == H_SUCCESS) {
+ CPUState *cs;
+
+ qemu_vfree(spapr->htab);
+ spapr->htab = pending->hpt;
+ spapr->htab_shift = pending->shift;
+
+ CPU_FOREACH(cs) {
+ run_on_cpu(cs, pivot_hpt, RUN_ON_CPU_HOST_PTR(spapr));
+ }
+
+ pending->hpt = NULL; /* so it's not free()d */
+ }
+
+ /* Clean up */
+ spapr->pending_hpt = NULL;
+ free_pending_hpt(pending);
+
+ return rc;
}
static target_ulong h_set_sprg0(PowerPCCPU *cpu, sPAPRMachineState *spapr,
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index d2c758b..954bada 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -14,6 +14,7 @@ struct sPAPRNVRAM;
typedef struct sPAPRConfigureConnectorState sPAPRConfigureConnectorState;
typedef struct sPAPREventLogEntry sPAPREventLogEntry;
typedef struct sPAPREventSource sPAPREventSource;
+typedef struct sPAPRPendingHPT sPAPRPendingHPT;
#define HPTE64_V_HPTE_DIRTY 0x0000000000000040ULL
#define SPAPR_ENTRY_POINT 0x100
@@ -72,6 +73,8 @@ struct sPAPRMachineState {
sPAPRResizeHPT resize_hpt;
void *htab;
uint32_t htab_shift;
+ sPAPRPendingHPT *pending_hpt; /* in-progress resize */
+
hwaddr rma_size;
int vrma_adjust;
ssize_t rtas_size;
@@ -627,6 +630,7 @@ void spapr_hotplug_req_remove_by_count_indexed(sPAPRDRConnectorType drc_type,
uint32_t count, uint32_t index);
void *spapr_populate_hotplug_cpu_dt(CPUState *cs, int *fdt_offset,
sPAPRMachineState *spapr);
+int spapr_hpt_shift_for_ramsize(uint64_t ramsize);
/* rtas-configure-connector state */
struct sPAPRConfigureConnectorState {
@@ -674,4 +678,6 @@ int spapr_rng_populate_dt(void *fdt);
void spapr_do_system_reset_on_cpu(CPUState *cs, run_on_cpu_data arg);
+#define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift))
+
#endif /* HW_SPAPR_H */
diff --git a/target-ppc/mmu-hash64.h b/target-ppc/mmu-hash64.h
index ab5d347..4a1b781 100644
--- a/target-ppc/mmu-hash64.h
+++ b/target-ppc/mmu-hash64.h
@@ -60,11 +60,15 @@ void ppc_hash64_update_rmls(CPUPPCState *env);
#define HASH_PTE_SIZE_64 16
#define HASH_PTEG_SIZE_64 (HASH_PTE_SIZE_64 * HPTES_PER_GROUP)
+#define HPTE64_V_SSIZE SLB_VSID_B
+#define HPTE64_V_SSIZE_256M SLB_VSID_B_256M
+#define HPTE64_V_SSIZE_1T SLB_VSID_B_1T
#define HPTE64_V_SSIZE_SHIFT 62
#define HPTE64_V_AVPN_SHIFT 7
#define HPTE64_V_AVPN 0x3fffffffffffff80ULL
#define HPTE64_V_AVPN_VAL(x) (((x) & HPTE64_V_AVPN) >> HPTE64_V_AVPN_SHIFT)
#define HPTE64_V_COMPARE(x, y) (!(((x) ^ (y)) & 0xffffffffffffff83ULL))
+#define HPTE64_V_BOLTED 0x0000000000000010ULL
#define HPTE64_V_LARGE 0x0000000000000004ULL
#define HPTE64_V_SECONDARY 0x0000000000000002ULL
#define HPTE64_V_VALID 0x0000000000000001ULL
--
2.9.3
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [Qemu-devel] [for-2.9 4/5] pseries: Enable HPT resizing for 2.9
2016-12-09 2:23 [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests David Gibson
` (2 preceding siblings ...)
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 3/5] pseries: Implement " David Gibson
@ 2016-12-09 2:23 ` David Gibson
2016-12-09 13:26 ` Laurent Vivier
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 5/5] pseries: Use smaller default hash page tables when guest can resize David Gibson
4 siblings, 1 reply; 18+ messages in thread
From: David Gibson @ 2016-12-09 2:23 UTC (permalink / raw)
To: paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth, lvivier, David Gibson
We've now implemented a PAPR extensions which allows PAPR guests (i.e.
"pseries" machine type) to resize their hash page table during runtime.
However, that extension is only enabled if explicitly chosen on the
command line. This patch enables it by default for spapr-2.9, but leaves
it disabled (by default) for older machine types.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
hw/ppc/spapr.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index f3b74dc..25468f6 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -2770,7 +2770,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
smc->dr_lmb_enabled = true;
smc->tcg_default_cpu = "POWER8";
- smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
+ smc->resize_hpt_default = SPAPR_RESIZE_HPT_ENABLED;
mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
fwc->get_dev_path = spapr_get_fw_dev_path;
nc->nmi_monitor_handler = spapr_nmi;
@@ -2851,8 +2851,11 @@ static void spapr_machine_2_8_instance_options(MachineState *machine)
static void spapr_machine_2_8_class_options(MachineClass *mc)
{
+ sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
+
spapr_machine_2_9_class_options(mc);
SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8);
+ smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
}
DEFINE_SPAPR_MACHINE(2_8, "2.8", false);
--
2.9.3
^ permalink raw reply related [flat|nested] 18+ messages in thread
* [Qemu-devel] [for-2.9 5/5] pseries: Use smaller default hash page tables when guest can resize
2016-12-09 2:23 [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests David Gibson
` (3 preceding siblings ...)
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 4/5] pseries: Enable HPT resizing for 2.9 David Gibson
@ 2016-12-09 2:23 ` David Gibson
4 siblings, 0 replies; 18+ messages in thread
From: David Gibson @ 2016-12-09 2:23 UTC (permalink / raw)
To: paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth, lvivier, David Gibson
We've now implemented a PAPR extension allowing PAPR guest to resize
their hash page table (HPT) during runtime.
This patch makes use of that facility to allocate smaller HPTs by default.
Specifically when a guest is aware of the HPT resize facility, qemu sizes
the HPT to the initial memory size, rather than the maximum memory size on
the assumption that the guest will resize its HPT if necessary for hot
plugged memory.
When the initial memory size is much smaller than the maximum memory size
(a common configuration with e.g. oVirt / RHEV) then this can save
significant memory on the HPT.
If the guest does *not* advertise HPT resize awareness when it makes the
ibm,client-architecture-support call, qemu resizes the HPT for maxmimum
memory size (unless it's been configured not to allow such guests at all).
For now we make that reallocation assuming the guest has not yet used the
HPT at all. That's true in practice, but not, strictly, an architectural
or PAPR requirement. If we need to in future we can fix this by having
the client-architecture-support call reboot the guest with the revised
HPT size (the client-architecture-support call is explicitly permitted to
trigger a reboot in this way).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
hw/ppc/spapr.c | 21 ++++++++++++++++-----
hw/ppc/spapr_hcall.c | 32 ++++++++++++++++++++++++++++++++
include/hw/ppc/spapr.h | 2 ++
include/hw/ppc/spapr_ovec.h | 1 +
4 files changed, 51 insertions(+), 5 deletions(-)
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index 25468f6..48c5cc0 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -1066,8 +1066,8 @@ int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
return shift;
}
-static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
- Error **errp)
+void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
+ Error **errp)
{
long rc;
@@ -1140,14 +1140,20 @@ static void ppc_spapr_reset(void)
hwaddr rtas_addr, fdt_addr;
void *fdt;
int rc;
+ int hpt_shift;
/* Check for unknown sysbus devices */
foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
/* Allocate and/or reset the hash page table */
- spapr_reallocate_hpt(spapr,
- spapr_hpt_shift_for_ramsize(machine->maxram_size),
- &error_fatal);
+ if ((spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED)
+ || (spapr->cas_reboot
+ && !spapr_ovec_test(spapr->ov5_cas, OV5_HPT_RESIZE))) {
+ hpt_shift = spapr_hpt_shift_for_ramsize(machine->maxram_size);
+ } else {
+ hpt_shift = spapr_hpt_shift_for_ramsize(machine->ram_size);
+ }
+ spapr_reallocate_hpt(spapr, hpt_shift, &error_fatal);
/* Update the RMA size if necessary */
if (spapr->vrma_adjust) {
@@ -1932,6 +1938,11 @@ static void ppc_spapr_init(MachineState *machine)
spapr_ovec_set(spapr->ov5, OV5_HP_EVT);
}
+ /* advertise support for HPT resizing */
+ if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
+ spapr_ovec_set(spapr->ov5, OV5_HPT_RESIZE);
+ }
+
/* init CPUs */
if (machine->cpu_model == NULL) {
machine->cpu_model = kvm_enabled() ? "host" : smc->tcg_default_cpu;
diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
index 9f421c1..fbd1e9f 100644
--- a/hw/ppc/spapr_hcall.c
+++ b/hw/ppc/spapr_hcall.c
@@ -1416,6 +1416,38 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu_,
ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
+ /*
+ * HPT resizing is a bit of a special case, because when enabled
+ * we assume the guest will support it until it says it doesn't,
+ * instead of assuming it won't support it until it says it does.
+ * Strictly speaking that approach could break for guests which
+ * don't make a CAS call, but those are so old we don't care about
+ * them. Without that assumption we'd have to make at least a
+ * temporary allocation of an HPT sized for max memory, which
+ * could be impossibly difficult under KVM HV if maxram is large.
+ */
+ if (!spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
+ int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
+
+ if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
+ error_report(
+ "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
+ exit(1);
+ }
+
+ if (spapr->htab_shift < maxshift) {
+ CPUState *cs;
+ /* Guest doesn't know about HPT resizing, so we
+ * pre-emptively resize for the maximum permitted RAM. At
+ * the point this is called, nothing should have been
+ * entered into the existing HPT */
+ spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
+ CPU_FOREACH(cs) {
+ run_on_cpu(cs, pivot_hpt, RUN_ON_CPU_HOST_PTR(spapr));
+ }
+ }
+ }
+
/* NOTE: there are actually a number of ov5 bits where input from the
* guest is always zero, and the platform/QEMU enables them independently
* of guest input. To model these properly we'd want some sort of mask,
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index 954bada..ff3bf87 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -631,6 +631,8 @@ void spapr_hotplug_req_remove_by_count_indexed(sPAPRDRConnectorType drc_type,
void *spapr_populate_hotplug_cpu_dt(CPUState *cs, int *fdt_offset,
sPAPRMachineState *spapr);
int spapr_hpt_shift_for_ramsize(uint64_t ramsize);
+void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
+ Error **errp);
/* rtas-configure-connector state */
struct sPAPRConfigureConnectorState {
diff --git a/include/hw/ppc/spapr_ovec.h b/include/hw/ppc/spapr_ovec.h
index 355a344..f5fed87 100644
--- a/include/hw/ppc/spapr_ovec.h
+++ b/include/hw/ppc/spapr_ovec.h
@@ -47,6 +47,7 @@ typedef struct sPAPROptionVector sPAPROptionVector;
#define OV5_DRCONF_MEMORY OV_BIT(2, 2)
#define OV5_FORM1_AFFINITY OV_BIT(5, 0)
#define OV5_HP_EVT OV_BIT(6, 5)
+#define OV5_HPT_RESIZE OV_BIT(6, 7)
/* interfaces */
sPAPROptionVector *spapr_ovec_new(void);
--
2.9.3
^ permalink raw reply related [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type David Gibson
@ 2016-12-09 7:59 ` Thomas Huth
2016-12-09 13:25 ` Laurent Vivier
1 sibling, 0 replies; 18+ messages in thread
From: Thomas Huth @ 2016-12-09 7:59 UTC (permalink / raw)
To: David Gibson, paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, lvivier
On 09.12.2016 03:23, David Gibson wrote:
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/ppc/spapr.c | 23 +++++++++++++++++++++--
> include/hw/compat.h | 3 +++
> 2 files changed, 24 insertions(+), 2 deletions(-)
Reviewed-by: Thomas Huth <thuth@redhat.com>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing David Gibson
@ 2016-12-09 8:18 ` Thomas Huth
2016-12-09 9:09 ` David Gibson
2016-12-09 17:08 ` Michael Roth
1 sibling, 1 reply; 18+ messages in thread
From: Thomas Huth @ 2016-12-09 8:18 UTC (permalink / raw)
To: David Gibson, paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, lvivier
On 09.12.2016 03:23, David Gibson wrote:
> This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
> H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
> extension to allow run time resizing of a guest's hash page table. It
> also adds a new machine property for controlling whether this new
> facility is available, and logic to check that against availability
> with KVM (only supported with KVM PR for now).
>
> Finally, it adds a new string to the hypertas property in the device
> tree, advertising to the guest the availability of the HPT resizing
> hypercalls. This is a tentative suggested value, and would need to be
> standardized by PAPR before being merged.
>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
> hw/ppc/trace-events | 2 ++
> include/hw/ppc/spapr.h | 11 +++++++++
> target-ppc/kvm.c | 26 ++++++++++++++++++++
> target-ppc/kvm_ppc.h | 5 ++++
> 6 files changed, 146 insertions(+)
>
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index 0f25e83..ecb0822 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
> if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
> add_str(hypertas, "hcall-multi-tce");
> }
> +
> + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
> + add_str(hypertas, "hcall-hpt-resize");
> + }
> +
> _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
> hypertas->str, hypertas->len));
> g_string_free(hypertas, TRUE);
> @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
> long load_limit, fw_size;
> char *filename;
> int smt = kvmppc_smt_threads();
> + Error *resize_hpt_err = NULL;
>
> msi_nonbroken = true;
>
> QLIST_INIT(&spapr->phbs);
>
> + /* Check HPT resizing availability */
> + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
> + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
> + if (resize_hpt_err) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> + error_free(resize_hpt_err);
> + resize_hpt_err = NULL;
> + } else {
> + spapr->resize_hpt = smc->resize_hpt_default;
> + }
> + }
> +
> + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
> +
> + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
> + error_report_err(resize_hpt_err);
> + exit(1);
> + }
The logic here is IMHO a little bit hard to understand. Why do you need
the SPAPR_RESIZE_HPT_DEFAULT state here at all? Wouldn't it be
sufficient to start with SPAPR_RESIZE_HPT_DISABLED by default and to
only enable it when the extension is available? ... at least some
explaining comments in the code about SPAPR_RESIZE_HPT_DEFAULT could
help here.
> @@ -2236,6 +2261,40 @@ static void spapr_set_modern_hotplug_events(Object *obj, bool value,
> spapr->use_hotplug_event_source = value;
> }
>
> +static char *spapr_get_resize_hpt(Object *obj, Error **errp)
> +{
> + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> +
> + switch (spapr->resize_hpt) {
> + case SPAPR_RESIZE_HPT_DEFAULT:
> + return g_strdup("default");
> + case SPAPR_RESIZE_HPT_DISABLED:
> + return g_strdup("disabled");
> + case SPAPR_RESIZE_HPT_ENABLED:
> + return g_strdup("enabled");
> + case SPAPR_RESIZE_HPT_REQUIRED:
> + return g_strdup("required");
> + }
> + assert(0);
> +}
> +
> +static void spapr_set_resize_hpt(Object *obj, const char *value, Error **errp)
> +{
> + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> +
> + if (strcmp(value, "default") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DEFAULT;
> + } else if (strcmp(value, "disabled") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> + } else if (strcmp(value, "enabled") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_ENABLED;
> + } else if (strcmp(value, "required") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_REQUIRED;
> + } else {
> + error_setg(errp, "Bad value for \"resize-hpt\" property");
> + }
> +}
> +
> static void spapr_machine_initfn(Object *obj)
> {
> sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> @@ -2256,6 +2315,12 @@ static void spapr_machine_initfn(Object *obj)
> " place of standard EPOW events when possible"
> " (required for memory hot-unplug support)",
> NULL);
> +
> + object_property_add_str(obj, "resize-hpt",
> + spapr_get_resize_hpt, spapr_set_resize_hpt, NULL);
> + object_property_set_description(obj, "resize-hpt",
> + "Resizing of the Hash Page Table (enabled, disabled, required)",
> + NULL);
... and "default" ?
> }
>
> static void spapr_machine_finalizefn(Object *obj)
> @@ -2707,6 +2772,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
>
> smc->dr_lmb_enabled = true;
> smc->tcg_default_cpu = "POWER8";
> + smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
> mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
> fwc->get_dev_path = spapr_get_fw_dev_path;
> nc->nmi_monitor_handler = spapr_nmi;
...
> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> index 15e12f3..dcd8cef 100644
> --- a/target-ppc/kvm.c
> +++ b/target-ppc/kvm.c
> @@ -22,6 +22,7 @@
> #include <linux/kvm.h>
>
> #include "qemu-common.h"
> +#include "qapi/error.h"
> #include "qemu/error-report.h"
> #include "cpu.h"
> #include "qemu/timer.h"
> @@ -2269,6 +2270,7 @@ int kvmppc_reset_htab(int shift_hint)
> /* Full emulation, tell caller to allocate htab itself */
> return 0;
> }
> +
Unnecessary white space change.
> if (kvm_check_extension(kvm_state, KVM_CAP_PPC_ALLOC_HTAB)) {
> int ret;
> ret = kvm_vm_ioctl(kvm_state, KVM_PPC_ALLOCATE_HTAB, &shift);
> @@ -2672,3 +2674,27 @@ int kvmppc_enable_hwrng(void)
>
> return kvmppc_enable_hcall(kvm_state, H_RANDOM);
> }
> +
> +void kvmppc_check_papr_resize_hpt(Error **errp)
> +{
> + if (!kvm_enabled()) {
> + return;
> + }
> +
> + /* TODO: Check specific capabilities for HPT resize aware host kernels */
> +
> + /*
> + * It's tempting to try to check directly if the HPT is under our
> + * control or KVM's, which is what's really relevant here.
> + * Unfortunately, in order to correctly size the HPT, we need to
> + * know if we can do resizing, _before_ we attempt to allocate it
> + * with KVM. Before that point, we don't officially know whether
> + * we'll control the HPT or not. So we have to use a fallback
> + * test for PR vs HV KVM to predict that.
> + */
> + if (kvmppc_is_pr(kvm_state)) {
> + return;
> + }
Couldn't we get HPT resizing on KVM-PR one day, too?
> + error_setg(errp, "Hash page table resizing not available with this KVM version");
> +}
> diff --git a/target-ppc/kvm_ppc.h b/target-ppc/kvm_ppc.h
> index 841a29b..3e852ba 100644
> --- a/target-ppc/kvm_ppc.h
> +++ b/target-ppc/kvm_ppc.h
> @@ -59,6 +59,7 @@ bool kvmppc_has_cap_htm(void);
> int kvmppc_enable_hwrng(void);
> int kvmppc_put_books_sregs(PowerPCCPU *cpu);
> PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
> +void kvmppc_check_papr_resize_hpt(Error **errp);
>
> #else
>
> @@ -270,6 +271,10 @@ static inline PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void)
> return NULL;
> }
>
> +static inline void kvmppc_check_papr_resize_hpt(Error **errp)
> +{
> + return;
> +}
> #endif
Thomas
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 3/5] pseries: Implement HPT resizing
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 3/5] pseries: Implement " David Gibson
@ 2016-12-09 8:36 ` Thomas Huth
2016-12-09 9:19 ` David Gibson
0 siblings, 1 reply; 18+ messages in thread
From: Thomas Huth @ 2016-12-09 8:36 UTC (permalink / raw)
To: David Gibson, paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, lvivier
On 09.12.2016 03:23, David Gibson wrote:
> This patch implements hypercalls allowing a PAPR guest to resize its own
> hash page table. This will eventually allow for more flexible memory
> hotplug.
>
> The implementation is partially asynchronous, handled in a special thread
> running the hpt_prepare_thread() function. The state of a pending resize
> is stored in SPAPR_MACHINE->pending_hpt.
>
> The H_RESIZE_HPT_PREPARE hypercall will kick off creation of a new HPT, or,
> if one is already in progress, monitor it for completion. If there is an
> existing HPT resize in progress that doesn't match the size specified in
> the call, it will cancel it, replacing it with a new one matching the
> given size.
>
> The H_RESIZE_HPT_COMMIT completes transition to a resized HPT, and can only
> be called successfully once H_RESIZE_HPT_PREPARE has successfully
> completed initialization of a new HPT. The guest must ensure that there
> are no concurrent accesses to the existing HPT while this is called (this
> effectively means stop_machine() for Linux guests).
>
> For now H_RESIZE_HPT_COMMIT goes through the whole old HPT, rehashing each
> HPTE into the new HPT. This can have quite high latency, but it seems to
> be of the order of typical migration downtime latencies for HPTs of size
> up to ~2GiB (which would be used in a 256GiB guest).
>
> In future we probably want to move more of the rehashing to the "prepare"
> phase, by having H_ENTER and other hcalls update both current and
> pending HPTs. That's a project for another day, but should be possible
> without any changes to the guest interface.
>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/ppc/spapr.c | 4 +-
> hw/ppc/spapr_hcall.c | 346 +++++++++++++++++++++++++++++++++++++++++++++++-
I wonder whether it makes sense to put all the hpt related code into a
separate new file?
> include/hw/ppc/spapr.h | 6 +
> target-ppc/mmu-hash64.h | 4 +
> 4 files changed, 354 insertions(+), 6 deletions(-)
>
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index ecb0822..f3b74dc 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -93,8 +93,6 @@
>
> #define PHANDLE_XICP 0x00001111
>
> -#define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift))
> -
> static XICSState *try_create_xics(const char *type, int nr_servers,
> int nr_irqs, Error **errp)
> {
> @@ -1055,7 +1053,7 @@ static void close_htab_fd(sPAPRMachineState *spapr)
> spapr->htab_fd = -1;
> }
>
> -static int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
> +int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
> {
> int shift;
>
> diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
> index 72a9c4d..9f421c1 100644
> --- a/hw/ppc/spapr_hcall.c
> +++ b/hw/ppc/spapr_hcall.c
> @@ -2,6 +2,7 @@
> #include "qapi/error.h"
> #include "sysemu/sysemu.h"
> #include "qemu/log.h"
> +#include "qemu/error-report.h"
> #include "cpu.h"
> #include "exec/exec-all.h"
> #include "helper_regs.h"
> @@ -355,20 +356,320 @@ static target_ulong h_read(PowerPCCPU *cpu, sPAPRMachineState *spapr,
> return H_SUCCESS;
> }
>
> +struct sPAPRPendingHPT {
> + /* These fields are read-only after initialization */
> + int shift;
> + QemuThread thread;
> +
> + /* These fields are protected by the BQL */
> + bool complete;
> +
> + /* These fields are private to the preparation thread if
> + * !complete, otherwise protected by the BQL */
> + int ret;
> + void *hpt;
> +};
> +
> +static void free_pending_hpt(sPAPRPendingHPT *pending)
> +{
> + if (pending->hpt) {
> + qemu_vfree(pending->hpt);
> + }
> +
> + g_free(pending);
> +}
> +
> +static void *hpt_prepare_thread(void *opaque)
> +{
> + sPAPRPendingHPT *pending = opaque;
> + size_t size = 1ULL << pending->shift;
> +
> + pending->hpt = qemu_memalign(size, size);
> + if (pending->hpt) {
> + memset(pending->hpt, 0, size);
> + pending->ret = H_SUCCESS;
> + } else {
> + pending->ret = H_NO_MEM;
> + }
> +
> + qemu_mutex_lock_iothread();
> +
> + if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt != pending) {
> + /* We've been cancelled, clean ourselves up */
> + free_pending_hpt(pending);
> + goto out;
> + }
> +
> + pending->complete = true;
> +
> +out:
You could easily avoid the goto here:
if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt != pending) {
/* We've been cancelled, clean ourselves up */
free_pending_hpt(pending);
} else {
pending->complete = true;
}
?
> + qemu_mutex_unlock_iothread();
> + return NULL;
> +}
> +
> +/* Must be called with BQL held */
> +static void cancel_hpt_prepare(sPAPRMachineState *spapr)
> +{
> + sPAPRPendingHPT *pending = spapr->pending_hpt;
> +
> + /* Let the thread know it's cancelled */
> + spapr->pending_hpt = NULL;
> +
> + if (!pending) {
> + /* Nothing to do */
> + return;
> + }
> +
> + if (!pending->complete) {
> + /* thread will clean itself up */
> + return;
> + }
> +
> + free_pending_hpt(pending);
> +}
> +
> +static int build_dimm_list(Object *obj, void *opaque)
> +{
> + GSList **list = opaque;
> +
> + if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
> + DeviceState *dev = DEVICE(obj);
> + if (dev->realized) { /* only realized DIMMs matter */
> + *list = g_slist_prepend(*list, dev);
> + }
> + }
> +
> + object_child_foreach(obj, build_dimm_list, opaque);
> + return 0;
> +}
> +
> +static ram_addr_t get_current_ram_size(void)
> +{
> + GSList *list = NULL, *item;
> + ram_addr_t size = ram_size;
> +
> + build_dimm_list(qdev_get_machine(), &list);
> + for (item = list; item; item = g_slist_next(item)) {
> + Object *obj = OBJECT(item->data);
> + if (!strcmp(object_get_typename(obj), TYPE_PC_DIMM)) {
> + size += object_property_get_int(obj, PC_DIMM_SIZE_PROP,
> + &error_abort);
> + }
> + }
> + g_slist_free(list);
> +
> + return size;
> +}
> +
> static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
> sPAPRMachineState *spapr,
> target_ulong opcode,
> target_ulong *args)
> {
> target_ulong flags = args[0];
> - target_ulong shift = args[1];
> + int shift = args[1];
> + sPAPRPendingHPT *pending = spapr->pending_hpt;
>
> if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> return H_AUTHORITY;
> }
>
> trace_spapr_h_resize_hpt_prepare(flags, shift);
> - return H_HARDWARE;
> +
> + if (flags != 0) {
> + return H_PARAMETER;
> + }
> +
> + if (shift && ((shift < 18) || (shift > 46))) {
> + return H_PARAMETER;
> + }
> +
> + if (pending) {
> + /* something already in progress */
> + if (pending->shift == shift) {
> + /* and it's suitable */
> + if (pending->complete) {
> + return pending->ret;
> + } else {
> + return H_LONG_BUSY_ORDER_100_MSEC;
> + }
> + }
> +
> + /* not suitable, cancel and replace */
> + cancel_hpt_prepare(spapr);
> + }
> +
> + if (!shift) {
> + /* nothing to do */
> + return H_SUCCESS;
> + }
> +
> + /* start new prepare */
> +
> + /* We only allow the guest to allocate an HPT one order above what
> + * we'd normally give them (to stop a small guest claiming a huge
> + * chunk of resources in the HPT */
> + if (shift > (spapr_hpt_shift_for_ramsize(get_current_ram_size()) + 1)) {
> + return H_RESOURCE;
> + }
> +
> + pending = g_malloc0(sizeof(*pending));
Maybe use g_new0() instead?
> + pending->shift = shift;
> + pending->ret = H_HARDWARE;
> +
> + qemu_thread_create(&pending->thread, "sPAPR HPT prepare",
> + hpt_prepare_thread, pending, QEMU_THREAD_DETACHED);
> +
> + spapr->pending_hpt = pending;
> +
> + /* In theory we could estimate the time more accurately based on
> + * the new size, but there's not much point */
> + return H_LONG_BUSY_ORDER_100_MSEC;
> +}
> +
> +static uint64_t new_hpte_load0(void *htab, uint64_t pteg, int slot)
> +{
> + uint8_t *addr = htab;
> +
> + addr += pteg * HASH_PTEG_SIZE_64;
> + addr += slot * HASH_PTE_SIZE_64;
> + return ldq_p(addr);
> +}
> +
> +static void new_hpte_store(void *htab, uint64_t pteg, int slot,
> + uint64_t pte0, uint64_t pte1)
> +{
> + uint8_t *addr = htab;
> +
> + addr += pteg * HASH_PTEG_SIZE_64;
> + addr += slot * HASH_PTE_SIZE_64;
> +
> + stq_p(addr, pte0);
> + stq_p(addr + HASH_PTE_SIZE_64/2, pte1);
> +}
> +
> +static int rehash_hpte(PowerPCCPU *cpu, uint64_t token,
> + void *old, uint64_t oldsize,
> + void *new, uint64_t newsize,
> + uint64_t pteg, int slot)
> +{
> + uint64_t old_hash_mask = (oldsize >> 7) - 1;
> + uint64_t new_hash_mask = (newsize >> 7) - 1;
> + target_ulong pte0 = ppc_hash64_load_hpte0(cpu, token, slot);
> + target_ulong pte1;
> + uint64_t avpn;
> + unsigned shift;
> + uint64_t hash, new_pteg, replace_pte0;
> +
> + if (!(pte0 & HPTE64_V_VALID) || !(pte0 & HPTE64_V_BOLTED)) {
> + return H_SUCCESS;
> + }
> +
> + pte1 = ppc_hash64_load_hpte1(cpu, token, slot);
> +
> + shift = ppc_hash64_hpte_page_shift_noslb(cpu, pte0, pte1);
> + assert(shift); /* H_ENTER should never have allowed a bad encoding */
> + avpn = HPTE64_V_AVPN_VAL(pte0) & ~(((1ULL << shift) - 1) >> 23);
> +
> + if (pte0 & HPTE64_V_SECONDARY) {
> + pteg = ~pteg;
> + }
> +
> + if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_256M) {
> + uint64_t offset, vsid;
> +
> + /* We only have 28 - 23 bits of offset in avpn */
> + offset = (avpn & 0x1f) << 23;
> + vsid = avpn >> 5;
> + /* We can find more bits from the pteg value */
> + if (shift < 23) {
> + offset |= ((vsid ^ pteg) & old_hash_mask) << shift;
> + }
> +
> + hash = vsid ^ (offset >> shift);
> + } else if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_1T) {
> + uint64_t offset, vsid;
> +
> + /* We only have 40 - 23 bits of seg_off in avpn */
> + offset = (avpn & 0x1ffff) << 23;
> + vsid = avpn >> 17;
> + if (shift < 23) {
> + offset |= ((vsid ^ (vsid << 25) ^ pteg) & old_hash_mask) << shift;
> + }
> +
> + hash = vsid ^ (vsid << 25) ^ (offset >> shift);
> + } else {
> + error_report("rehash_pte: Bad segment size in HPTE");
> + return H_HARDWARE;
> + }
> +
> + new_pteg = hash & new_hash_mask;
> + if (pte0 & HPTE64_V_SECONDARY) {
> + assert(~pteg == (hash & old_hash_mask));
> + new_pteg = ~new_pteg;
> + } else {
> + assert(pteg == (hash & old_hash_mask));
> + }
> + assert((oldsize != newsize) || (pteg == new_pteg));
> + replace_pte0 = new_hpte_load0(new, new_pteg, slot);
> + if (replace_pte0 & HPTE64_V_VALID) {
> + assert(newsize < oldsize);
> + if (replace_pte0 & HPTE64_V_BOLTED) {
> + if (pte0 & HPTE64_V_BOLTED) {
> + /* Bolted collision, nothing we can do */
> + return H_PTEG_FULL;
> + } else {
> + /* Discard this hpte */
> + return H_SUCCESS;
> + }
> + }
> + }
> +
> + new_hpte_store(new, new_pteg, slot, pte0, pte1);
> + return H_SUCCESS;
> +}
> +
> +static int rehash_hpt(PowerPCCPU *cpu,
> + void *old, uint64_t oldsize,
> + void *new, uint64_t newsize)
> +{
> + CPUPPCState *env = &cpu->env;
> + uint64_t n_ptegs = oldsize >> 7;
> + uint64_t pteg;
> + int slot;
> + int rc;
> +
> + assert(env->external_htab == old);
> +
> + for (pteg = 0; pteg < n_ptegs; pteg++) {
> + uint64_t token = ppc_hash64_start_access(cpu, pteg * HPTES_PER_GROUP);
> +
> + if (!token) {
> + return H_HARDWARE;
> + }
> +
> + for (slot = 0; slot < HPTES_PER_GROUP; slot++) {
> + rc = rehash_hpte(cpu, token, old, oldsize, new, newsize,
> + pteg, slot);
> + if (rc != H_SUCCESS) {
> + ppc_hash64_stop_access(cpu, token);
> + return rc;
> + }
> + }
> + ppc_hash64_stop_access(cpu, token);
> + }
> +
> + return H_SUCCESS;
> +}
> +
> +static void pivot_hpt(CPUState *cs, run_on_cpu_data arg)
> +{
> + sPAPRMachineState *spapr = SPAPR_MACHINE(arg.host_ptr);
> + PowerPCCPU *cpu = POWERPC_CPU(cs);
> +
> + cpu_synchronize_state(cs);
> + ppc_hash64_set_external_hpt(cpu, spapr->htab, spapr->htab_shift,
> + &error_fatal);
> }
>
> static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
> @@ -378,13 +679,52 @@ static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
> {
> target_ulong flags = args[0];
> target_ulong shift = args[1];
> + sPAPRPendingHPT *pending = spapr->pending_hpt;
> + int rc;
> + size_t newsize;
Why size_t here? size_t could be 32-bit on some systems, and don't you
rather want to always have 64-bit (or at least target_ulong) here?
> if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> return H_AUTHORITY;
> }
>
> trace_spapr_h_resize_hpt_commit(flags, shift);
> - return H_HARDWARE;
> +
> + if (flags != 0) {
> + return H_PARAMETER;
> + }
> +
> + if (!pending || (pending->shift != shift)) {
> + /* no matching prepare */
> + return H_CLOSED;
> + }
> +
> + if (!pending->complete) {
> + /* prepare has not completed */
> + return H_BUSY;
> + }
> +
> + newsize = 1ULL << pending->shift;
> + rc = rehash_hpt(cpu, spapr->htab, HTAB_SIZE(spapr),
> + pending->hpt, newsize);
> + if (rc == H_SUCCESS) {
> + CPUState *cs;
> +
> + qemu_vfree(spapr->htab);
> + spapr->htab = pending->hpt;
> + spapr->htab_shift = pending->shift;
> +
> + CPU_FOREACH(cs) {
> + run_on_cpu(cs, pivot_hpt, RUN_ON_CPU_HOST_PTR(spapr));
> + }
> +
> + pending->hpt = NULL; /* so it's not free()d */
> + }
> +
> + /* Clean up */
> + spapr->pending_hpt = NULL;
> + free_pending_hpt(pending);
> +
> + return rc;
> }
Thomas
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 8:18 ` Thomas Huth
@ 2016-12-09 9:09 ` David Gibson
2016-12-09 9:18 ` Thomas Huth
0 siblings, 1 reply; 18+ messages in thread
From: David Gibson @ 2016-12-09 9:09 UTC (permalink / raw)
To: Thomas Huth
Cc: paulus, sjitindarsingh, agraf, mdroth, qemu-ppc, qemu-devel, lvivier
[-- Attachment #1: Type: text/plain, Size: 9290 bytes --]
On Fri, Dec 09, 2016 at 09:18:51AM +0100, Thomas Huth wrote:
> On 09.12.2016 03:23, David Gibson wrote:
> > This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
> > H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
> > extension to allow run time resizing of a guest's hash page table. It
> > also adds a new machine property for controlling whether this new
> > facility is available, and logic to check that against availability
> > with KVM (only supported with KVM PR for now).
> >
> > Finally, it adds a new string to the hypertas property in the device
> > tree, advertising to the guest the availability of the HPT resizing
> > hypercalls. This is a tentative suggested value, and would need to be
> > standardized by PAPR before being merged.
> >
> > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> > hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> > hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
> > hw/ppc/trace-events | 2 ++
> > include/hw/ppc/spapr.h | 11 +++++++++
> > target-ppc/kvm.c | 26 ++++++++++++++++++++
> > target-ppc/kvm_ppc.h | 5 ++++
> > 6 files changed, 146 insertions(+)
> >
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index 0f25e83..ecb0822 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
> > if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
> > add_str(hypertas, "hcall-multi-tce");
> > }
> > +
> > + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
> > + add_str(hypertas, "hcall-hpt-resize");
> > + }
> > +
> > _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
> > hypertas->str, hypertas->len));
> > g_string_free(hypertas, TRUE);
> > @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
> > long load_limit, fw_size;
> > char *filename;
> > int smt = kvmppc_smt_threads();
> > + Error *resize_hpt_err = NULL;
> >
> > msi_nonbroken = true;
> >
> > QLIST_INIT(&spapr->phbs);
> >
> > + /* Check HPT resizing availability */
> > + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
> > + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
> > + if (resize_hpt_err) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> > + error_free(resize_hpt_err);
> > + resize_hpt_err = NULL;
> > + } else {
> > + spapr->resize_hpt = smc->resize_hpt_default;
> > + }
> > + }
> > +
> > + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
> > +
> > + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
> > + error_report_err(resize_hpt_err);
> > + exit(1);
> > + }
>
> The logic here is IMHO a little bit hard to understand. Why do you need
> the SPAPR_RESIZE_HPT_DEFAULT state here at all? Wouldn't it be
> sufficient to start with SPAPR_RESIZE_HPT_DISABLED by default and to
> only enable it when the extension is available?
Because if the user explicitly sets something, we should never
override itin the logic here. In order to determine whether the user
has set something explicitly, we need another value that means "not
set explicitly".
> ... at least some
> explaining comments in the code about SPAPR_RESIZE_HPT_DEFAULT could
> help here.
Hm, maybe.
> > @@ -2236,6 +2261,40 @@ static void spapr_set_modern_hotplug_events(Object *obj, bool value,
> > spapr->use_hotplug_event_source = value;
> > }
> >
> > +static char *spapr_get_resize_hpt(Object *obj, Error **errp)
> > +{
> > + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> > +
> > + switch (spapr->resize_hpt) {
> > + case SPAPR_RESIZE_HPT_DEFAULT:
> > + return g_strdup("default");
> > + case SPAPR_RESIZE_HPT_DISABLED:
> > + return g_strdup("disabled");
> > + case SPAPR_RESIZE_HPT_ENABLED:
> > + return g_strdup("enabled");
> > + case SPAPR_RESIZE_HPT_REQUIRED:
> > + return g_strdup("required");
> > + }
> > + assert(0);
> > +}
> > +
> > +static void spapr_set_resize_hpt(Object *obj, const char *value, Error **errp)
> > +{
> > + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> > +
> > + if (strcmp(value, "default") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DEFAULT;
> > + } else if (strcmp(value, "disabled") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> > + } else if (strcmp(value, "enabled") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_ENABLED;
> > + } else if (strcmp(value, "required") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_REQUIRED;
> > + } else {
> > + error_setg(errp, "Bad value for \"resize-hpt\" property");
> > + }
> > +}
> > +
> > static void spapr_machine_initfn(Object *obj)
> > {
> > sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> > @@ -2256,6 +2315,12 @@ static void spapr_machine_initfn(Object *obj)
> > " place of standard EPOW events when possible"
> > " (required for memory hot-unplug support)",
> > NULL);
> > +
> > + object_property_add_str(obj, "resize-hpt",
> > + spapr_get_resize_hpt, spapr_set_resize_hpt, NULL);
> > + object_property_set_description(obj, "resize-hpt",
> > + "Resizing of the Hash Page Table (enabled, disabled, required)",
> > + NULL);
>
> ... and "default" ?
>
> > }
> >
> > static void spapr_machine_finalizefn(Object *obj)
> > @@ -2707,6 +2772,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
> >
> > smc->dr_lmb_enabled = true;
> > smc->tcg_default_cpu = "POWER8";
> > + smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
> > mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
> > fwc->get_dev_path = spapr_get_fw_dev_path;
> > nc->nmi_monitor_handler = spapr_nmi;
> ...
> > diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> > index 15e12f3..dcd8cef 100644
> > --- a/target-ppc/kvm.c
> > +++ b/target-ppc/kvm.c
> > @@ -22,6 +22,7 @@
> > #include <linux/kvm.h>
> >
> > #include "qemu-common.h"
> > +#include "qapi/error.h"
> > #include "qemu/error-report.h"
> > #include "cpu.h"
> > #include "qemu/timer.h"
> > @@ -2269,6 +2270,7 @@ int kvmppc_reset_htab(int shift_hint)
> > /* Full emulation, tell caller to allocate htab itself */
> > return 0;
> > }
> > +
>
> Unnecessary white space change.
>
> > if (kvm_check_extension(kvm_state, KVM_CAP_PPC_ALLOC_HTAB)) {
> > int ret;
> > ret = kvm_vm_ioctl(kvm_state, KVM_PPC_ALLOCATE_HTAB, &shift);
> > @@ -2672,3 +2674,27 @@ int kvmppc_enable_hwrng(void)
> >
> > return kvmppc_enable_hcall(kvm_state, H_RANDOM);
> > }
> > +
> > +void kvmppc_check_papr_resize_hpt(Error **errp)
> > +{
> > + if (!kvm_enabled()) {
> > + return;
> > + }
> > +
> > + /* TODO: Check specific capabilities for HPT resize aware host kernels */
> > +
> > + /*
> > + * It's tempting to try to check directly if the HPT is under our
> > + * control or KVM's, which is what's really relevant here.
> > + * Unfortunately, in order to correctly size the HPT, we need to
> > + * know if we can do resizing, _before_ we attempt to allocate it
> > + * with KVM. Before that point, we don't officially know whether
> > + * we'll control the HPT or not. So we have to use a fallback
> > + * test for PR vs HV KVM to predict that.
> > + */
> > + if (kvmppc_is_pr(kvm_state)) {
> > + return;
> > + }
>
> Couldn't we get HPT resizing on KVM-PR one day, too?
I guess I need to reword that. HPT resizing should work on PR right
now (but I haven't been able to test it yet) - because qemu still
allocates and manages the HPT on PR, the qemu side logic is
sufficient.
> > + error_setg(errp, "Hash page table resizing not available with this KVM version");
> > +}
> > diff --git a/target-ppc/kvm_ppc.h b/target-ppc/kvm_ppc.h
> > index 841a29b..3e852ba 100644
> > --- a/target-ppc/kvm_ppc.h
> > +++ b/target-ppc/kvm_ppc.h
> > @@ -59,6 +59,7 @@ bool kvmppc_has_cap_htm(void);
> > int kvmppc_enable_hwrng(void);
> > int kvmppc_put_books_sregs(PowerPCCPU *cpu);
> > PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
> > +void kvmppc_check_papr_resize_hpt(Error **errp);
> >
> > #else
> >
> > @@ -270,6 +271,10 @@ static inline PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void)
> > return NULL;
> > }
> >
> > +static inline void kvmppc_check_papr_resize_hpt(Error **errp)
> > +{
> > + return;
> > +}
> > #endif
>
> Thomas
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 9:09 ` David Gibson
@ 2016-12-09 9:18 ` Thomas Huth
2016-12-12 1:06 ` David Gibson
0 siblings, 1 reply; 18+ messages in thread
From: Thomas Huth @ 2016-12-09 9:18 UTC (permalink / raw)
To: David Gibson
Cc: paulus, sjitindarsingh, agraf, mdroth, qemu-ppc, qemu-devel, lvivier
[-- Attachment #1: Type: text/plain, Size: 3703 bytes --]
On 09.12.2016 10:09, David Gibson wrote:
> On Fri, Dec 09, 2016 at 09:18:51AM +0100, Thomas Huth wrote:
>> On 09.12.2016 03:23, David Gibson wrote:
>>> This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
>>> H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
>>> extension to allow run time resizing of a guest's hash page table. It
>>> also adds a new machine property for controlling whether this new
>>> facility is available, and logic to check that against availability
>>> with KVM (only supported with KVM PR for now).
>>>
>>> Finally, it adds a new string to the hypertas property in the device
>>> tree, advertising to the guest the availability of the HPT resizing
>>> hypercalls. This is a tentative suggested value, and would need to be
>>> standardized by PAPR before being merged.
>>>
>>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
>>> ---
>>> hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
>>> hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
>>> hw/ppc/trace-events | 2 ++
>>> include/hw/ppc/spapr.h | 11 +++++++++
>>> target-ppc/kvm.c | 26 ++++++++++++++++++++
>>> target-ppc/kvm_ppc.h | 5 ++++
>>> 6 files changed, 146 insertions(+)
>>>
>>> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
>>> index 0f25e83..ecb0822 100644
>>> --- a/hw/ppc/spapr.c
>>> +++ b/hw/ppc/spapr.c
>>> @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
>>> if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
>>> add_str(hypertas, "hcall-multi-tce");
>>> }
>>> +
>>> + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
>>> + add_str(hypertas, "hcall-hpt-resize");
>>> + }
>>> +
>>> _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
>>> hypertas->str, hypertas->len));
>>> g_string_free(hypertas, TRUE);
>>> @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
>>> long load_limit, fw_size;
>>> char *filename;
>>> int smt = kvmppc_smt_threads();
>>> + Error *resize_hpt_err = NULL;
>>>
>>> msi_nonbroken = true;
>>>
>>> QLIST_INIT(&spapr->phbs);
>>>
>>> + /* Check HPT resizing availability */
>>> + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
>>> + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
>>> + if (resize_hpt_err) {
>>> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
>>> + error_free(resize_hpt_err);
>>> + resize_hpt_err = NULL;
>>> + } else {
>>> + spapr->resize_hpt = smc->resize_hpt_default;
>>> + }
>>> + }
>>> +
>>> + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
>>> +
>>> + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
>>> + error_report_err(resize_hpt_err);
>>> + exit(1);
>>> + }
>>
>> The logic here is IMHO a little bit hard to understand. Why do you need
>> the SPAPR_RESIZE_HPT_DEFAULT state here at all? Wouldn't it be
>> sufficient to start with SPAPR_RESIZE_HPT_DISABLED by default and to
>> only enable it when the extension is available?
>
> Because if the user explicitly sets something, we should never
> override itin the logic here. In order to determine whether the user
> has set something explicitly, we need another value that means "not
> set explicitly".
Hmm, ok, I think I slowly get it. Another question: Does this also take
care of the handling if the user migrates from a machine that supports
HPT resizing to a machine that does not support this feature?
Thomas
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 3/5] pseries: Implement HPT resizing
2016-12-09 8:36 ` Thomas Huth
@ 2016-12-09 9:19 ` David Gibson
0 siblings, 0 replies; 18+ messages in thread
From: David Gibson @ 2016-12-09 9:19 UTC (permalink / raw)
To: Thomas Huth
Cc: paulus, sjitindarsingh, agraf, mdroth, qemu-ppc, qemu-devel, lvivier
[-- Attachment #1: Type: text/plain, Size: 16732 bytes --]
On Fri, Dec 09, 2016 at 09:36:17AM +0100, Thomas Huth wrote:
> On 09.12.2016 03:23, David Gibson wrote:
> > This patch implements hypercalls allowing a PAPR guest to resize its own
> > hash page table. This will eventually allow for more flexible memory
> > hotplug.
> >
> > The implementation is partially asynchronous, handled in a special thread
> > running the hpt_prepare_thread() function. The state of a pending resize
> > is stored in SPAPR_MACHINE->pending_hpt.
> >
> > The H_RESIZE_HPT_PREPARE hypercall will kick off creation of a new HPT, or,
> > if one is already in progress, monitor it for completion. If there is an
> > existing HPT resize in progress that doesn't match the size specified in
> > the call, it will cancel it, replacing it with a new one matching the
> > given size.
> >
> > The H_RESIZE_HPT_COMMIT completes transition to a resized HPT, and can only
> > be called successfully once H_RESIZE_HPT_PREPARE has successfully
> > completed initialization of a new HPT. The guest must ensure that there
> > are no concurrent accesses to the existing HPT while this is called (this
> > effectively means stop_machine() for Linux guests).
> >
> > For now H_RESIZE_HPT_COMMIT goes through the whole old HPT, rehashing each
> > HPTE into the new HPT. This can have quite high latency, but it seems to
> > be of the order of typical migration downtime latencies for HPTs of size
> > up to ~2GiB (which would be used in a 256GiB guest).
> >
> > In future we probably want to move more of the rehashing to the "prepare"
> > phase, by having H_ENTER and other hcalls update both current and
> > pending HPTs. That's a project for another day, but should be possible
> > without any changes to the guest interface.
> >
> > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> > hw/ppc/spapr.c | 4 +-
> > hw/ppc/spapr_hcall.c | 346 +++++++++++++++++++++++++++++++++++++++++++++++-
>
> I wonder whether it makes sense to put all the hpt related code into a
> separate new file?
Maybe, but not really in scope here.
> > include/hw/ppc/spapr.h | 6 +
> > target-ppc/mmu-hash64.h | 4 +
> > 4 files changed, 354 insertions(+), 6 deletions(-)
> >
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index ecb0822..f3b74dc 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -93,8 +93,6 @@
> >
> > #define PHANDLE_XICP 0x00001111
> >
> > -#define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift))
> > -
> > static XICSState *try_create_xics(const char *type, int nr_servers,
> > int nr_irqs, Error **errp)
> > {
> > @@ -1055,7 +1053,7 @@ static void close_htab_fd(sPAPRMachineState *spapr)
> > spapr->htab_fd = -1;
> > }
> >
> > -static int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
> > +int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
> > {
> > int shift;
> >
> > diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
> > index 72a9c4d..9f421c1 100644
> > --- a/hw/ppc/spapr_hcall.c
> > +++ b/hw/ppc/spapr_hcall.c
> > @@ -2,6 +2,7 @@
> > #include "qapi/error.h"
> > #include "sysemu/sysemu.h"
> > #include "qemu/log.h"
> > +#include "qemu/error-report.h"
> > #include "cpu.h"
> > #include "exec/exec-all.h"
> > #include "helper_regs.h"
> > @@ -355,20 +356,320 @@ static target_ulong h_read(PowerPCCPU *cpu, sPAPRMachineState *spapr,
> > return H_SUCCESS;
> > }
> >
> > +struct sPAPRPendingHPT {
> > + /* These fields are read-only after initialization */
> > + int shift;
> > + QemuThread thread;
> > +
> > + /* These fields are protected by the BQL */
> > + bool complete;
> > +
> > + /* These fields are private to the preparation thread if
> > + * !complete, otherwise protected by the BQL */
> > + int ret;
> > + void *hpt;
> > +};
> > +
> > +static void free_pending_hpt(sPAPRPendingHPT *pending)
> > +{
> > + if (pending->hpt) {
> > + qemu_vfree(pending->hpt);
> > + }
> > +
> > + g_free(pending);
> > +}
> > +
> > +static void *hpt_prepare_thread(void *opaque)
> > +{
> > + sPAPRPendingHPT *pending = opaque;
> > + size_t size = 1ULL << pending->shift;
> > +
> > + pending->hpt = qemu_memalign(size, size);
> > + if (pending->hpt) {
> > + memset(pending->hpt, 0, size);
> > + pending->ret = H_SUCCESS;
> > + } else {
> > + pending->ret = H_NO_MEM;
> > + }
> > +
> > + qemu_mutex_lock_iothread();
> > +
> > + if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt != pending) {
> > + /* We've been cancelled, clean ourselves up */
> > + free_pending_hpt(pending);
> > + goto out;
> > + }
> > +
> > + pending->complete = true;
> > +
> > +out:
>
> You could easily avoid the goto here:
>
> if (SPAPR_MACHINE(qdev_get_machine())->pending_hpt != pending) {
> /* We've been cancelled, clean ourselves up */
> free_pending_hpt(pending);
> } else {
> pending->complete = true;
> }
>
> ?
Ah, yeah, I guess so. I think I had more cases that used the same
bailout path during development.
> > + qemu_mutex_unlock_iothread();
> > + return NULL;
> > +}
> > +
> > +/* Must be called with BQL held */
> > +static void cancel_hpt_prepare(sPAPRMachineState *spapr)
> > +{
> > + sPAPRPendingHPT *pending = spapr->pending_hpt;
> > +
> > + /* Let the thread know it's cancelled */
> > + spapr->pending_hpt = NULL;
> > +
> > + if (!pending) {
> > + /* Nothing to do */
> > + return;
> > + }
> > +
> > + if (!pending->complete) {
> > + /* thread will clean itself up */
> > + return;
> > + }
> > +
> > + free_pending_hpt(pending);
> > +}
> > +
> > +static int build_dimm_list(Object *obj, void *opaque)
> > +{
> > + GSList **list = opaque;
> > +
> > + if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
> > + DeviceState *dev = DEVICE(obj);
> > + if (dev->realized) { /* only realized DIMMs matter */
> > + *list = g_slist_prepend(*list, dev);
> > + }
> > + }
> > +
> > + object_child_foreach(obj, build_dimm_list, opaque);
> > + return 0;
> > +}
> > +
> > +static ram_addr_t get_current_ram_size(void)
> > +{
> > + GSList *list = NULL, *item;
> > + ram_addr_t size = ram_size;
> > +
> > + build_dimm_list(qdev_get_machine(), &list);
> > + for (item = list; item; item = g_slist_next(item)) {
> > + Object *obj = OBJECT(item->data);
> > + if (!strcmp(object_get_typename(obj), TYPE_PC_DIMM)) {
> > + size += object_property_get_int(obj, PC_DIMM_SIZE_PROP,
> > + &error_abort);
> > + }
> > + }
> > + g_slist_free(list);
> > +
> > + return size;
> > +}
> > +
> > static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
> > sPAPRMachineState *spapr,
> > target_ulong opcode,
> > target_ulong *args)
> > {
> > target_ulong flags = args[0];
> > - target_ulong shift = args[1];
> > + int shift = args[1];
> > + sPAPRPendingHPT *pending = spapr->pending_hpt;
> >
> > if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> > return H_AUTHORITY;
> > }
> >
> > trace_spapr_h_resize_hpt_prepare(flags, shift);
> > - return H_HARDWARE;
> > +
> > + if (flags != 0) {
> > + return H_PARAMETER;
> > + }
> > +
> > + if (shift && ((shift < 18) || (shift > 46))) {
> > + return H_PARAMETER;
> > + }
> > +
> > + if (pending) {
> > + /* something already in progress */
> > + if (pending->shift == shift) {
> > + /* and it's suitable */
> > + if (pending->complete) {
> > + return pending->ret;
> > + } else {
> > + return H_LONG_BUSY_ORDER_100_MSEC;
> > + }
> > + }
> > +
> > + /* not suitable, cancel and replace */
> > + cancel_hpt_prepare(spapr);
> > + }
> > +
> > + if (!shift) {
> > + /* nothing to do */
> > + return H_SUCCESS;
> > + }
> > +
> > + /* start new prepare */
> > +
> > + /* We only allow the guest to allocate an HPT one order above what
> > + * we'd normally give them (to stop a small guest claiming a huge
> > + * chunk of resources in the HPT */
> > + if (shift > (spapr_hpt_shift_for_ramsize(get_current_ram_size()) + 1)) {
> > + return H_RESOURCE;
> > + }
> > +
> > + pending = g_malloc0(sizeof(*pending));
>
> Maybe use g_new0() instead?
Ah, good idea.
> > + pending->shift = shift;
> > + pending->ret = H_HARDWARE;
> > +
> > + qemu_thread_create(&pending->thread, "sPAPR HPT prepare",
> > + hpt_prepare_thread, pending, QEMU_THREAD_DETACHED);
> > +
> > + spapr->pending_hpt = pending;
> > +
> > + /* In theory we could estimate the time more accurately based on
> > + * the new size, but there's not much point */
> > + return H_LONG_BUSY_ORDER_100_MSEC;
> > +}
> > +
> > +static uint64_t new_hpte_load0(void *htab, uint64_t pteg, int slot)
> > +{
> > + uint8_t *addr = htab;
> > +
> > + addr += pteg * HASH_PTEG_SIZE_64;
> > + addr += slot * HASH_PTE_SIZE_64;
> > + return ldq_p(addr);
> > +}
> > +
> > +static void new_hpte_store(void *htab, uint64_t pteg, int slot,
> > + uint64_t pte0, uint64_t pte1)
> > +{
> > + uint8_t *addr = htab;
> > +
> > + addr += pteg * HASH_PTEG_SIZE_64;
> > + addr += slot * HASH_PTE_SIZE_64;
> > +
> > + stq_p(addr, pte0);
> > + stq_p(addr + HASH_PTE_SIZE_64/2, pte1);
> > +}
> > +
> > +static int rehash_hpte(PowerPCCPU *cpu, uint64_t token,
> > + void *old, uint64_t oldsize,
> > + void *new, uint64_t newsize,
> > + uint64_t pteg, int slot)
> > +{
> > + uint64_t old_hash_mask = (oldsize >> 7) - 1;
> > + uint64_t new_hash_mask = (newsize >> 7) - 1;
> > + target_ulong pte0 = ppc_hash64_load_hpte0(cpu, token, slot);
> > + target_ulong pte1;
> > + uint64_t avpn;
> > + unsigned shift;
> > + uint64_t hash, new_pteg, replace_pte0;
> > +
> > + if (!(pte0 & HPTE64_V_VALID) || !(pte0 & HPTE64_V_BOLTED)) {
> > + return H_SUCCESS;
> > + }
> > +
> > + pte1 = ppc_hash64_load_hpte1(cpu, token, slot);
> > +
> > + shift = ppc_hash64_hpte_page_shift_noslb(cpu, pte0, pte1);
> > + assert(shift); /* H_ENTER should never have allowed a bad encoding */
> > + avpn = HPTE64_V_AVPN_VAL(pte0) & ~(((1ULL << shift) - 1) >> 23);
> > +
> > + if (pte0 & HPTE64_V_SECONDARY) {
> > + pteg = ~pteg;
> > + }
> > +
> > + if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_256M) {
> > + uint64_t offset, vsid;
> > +
> > + /* We only have 28 - 23 bits of offset in avpn */
> > + offset = (avpn & 0x1f) << 23;
> > + vsid = avpn >> 5;
> > + /* We can find more bits from the pteg value */
> > + if (shift < 23) {
> > + offset |= ((vsid ^ pteg) & old_hash_mask) << shift;
> > + }
> > +
> > + hash = vsid ^ (offset >> shift);
> > + } else if ((pte0 & HPTE64_V_SSIZE) == HPTE64_V_SSIZE_1T) {
> > + uint64_t offset, vsid;
> > +
> > + /* We only have 40 - 23 bits of seg_off in avpn */
> > + offset = (avpn & 0x1ffff) << 23;
> > + vsid = avpn >> 17;
> > + if (shift < 23) {
> > + offset |= ((vsid ^ (vsid << 25) ^ pteg) & old_hash_mask) << shift;
> > + }
> > +
> > + hash = vsid ^ (vsid << 25) ^ (offset >> shift);
> > + } else {
> > + error_report("rehash_pte: Bad segment size in HPTE");
> > + return H_HARDWARE;
> > + }
> > +
> > + new_pteg = hash & new_hash_mask;
> > + if (pte0 & HPTE64_V_SECONDARY) {
> > + assert(~pteg == (hash & old_hash_mask));
> > + new_pteg = ~new_pteg;
> > + } else {
> > + assert(pteg == (hash & old_hash_mask));
> > + }
> > + assert((oldsize != newsize) || (pteg == new_pteg));
> > + replace_pte0 = new_hpte_load0(new, new_pteg, slot);
> > + if (replace_pte0 & HPTE64_V_VALID) {
> > + assert(newsize < oldsize);
> > + if (replace_pte0 & HPTE64_V_BOLTED) {
> > + if (pte0 & HPTE64_V_BOLTED) {
> > + /* Bolted collision, nothing we can do */
> > + return H_PTEG_FULL;
> > + } else {
> > + /* Discard this hpte */
> > + return H_SUCCESS;
> > + }
> > + }
> > + }
> > +
> > + new_hpte_store(new, new_pteg, slot, pte0, pte1);
> > + return H_SUCCESS;
> > +}
> > +
> > +static int rehash_hpt(PowerPCCPU *cpu,
> > + void *old, uint64_t oldsize,
> > + void *new, uint64_t newsize)
> > +{
> > + CPUPPCState *env = &cpu->env;
> > + uint64_t n_ptegs = oldsize >> 7;
> > + uint64_t pteg;
> > + int slot;
> > + int rc;
> > +
> > + assert(env->external_htab == old);
> > +
> > + for (pteg = 0; pteg < n_ptegs; pteg++) {
> > + uint64_t token = ppc_hash64_start_access(cpu, pteg * HPTES_PER_GROUP);
> > +
> > + if (!token) {
> > + return H_HARDWARE;
> > + }
> > +
> > + for (slot = 0; slot < HPTES_PER_GROUP; slot++) {
> > + rc = rehash_hpte(cpu, token, old, oldsize, new, newsize,
> > + pteg, slot);
> > + if (rc != H_SUCCESS) {
> > + ppc_hash64_stop_access(cpu, token);
> > + return rc;
> > + }
> > + }
> > + ppc_hash64_stop_access(cpu, token);
> > + }
> > +
> > + return H_SUCCESS;
> > +}
> > +
> > +static void pivot_hpt(CPUState *cs, run_on_cpu_data arg)
> > +{
> > + sPAPRMachineState *spapr = SPAPR_MACHINE(arg.host_ptr);
> > + PowerPCCPU *cpu = POWERPC_CPU(cs);
> > +
> > + cpu_synchronize_state(cs);
> > + ppc_hash64_set_external_hpt(cpu, spapr->htab, spapr->htab_shift,
> > + &error_fatal);
> > }
> >
> > static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
> > @@ -378,13 +679,52 @@ static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
> > {
> > target_ulong flags = args[0];
> > target_ulong shift = args[1];
> > + sPAPRPendingHPT *pending = spapr->pending_hpt;
> > + int rc;
> > + size_t newsize;
>
> Why size_t here? size_t could be 32-bit on some systems, and don't you
> rather want to always have 64-bit (or at least target_ulong) here?
In that case (and if newsize would be > 4G) we're stuffed anyway - we
allocate the htab with qemu_memalign(), which only takes a size_t, and
even if it didn't there's nowhere we could actually allocate the hash table.
> > if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> > return H_AUTHORITY;
> > }
> >
> > trace_spapr_h_resize_hpt_commit(flags, shift);
> > - return H_HARDWARE;
> > +
> > + if (flags != 0) {
> > + return H_PARAMETER;
> > + }
> > +
> > + if (!pending || (pending->shift != shift)) {
> > + /* no matching prepare */
> > + return H_CLOSED;
> > + }
> > +
> > + if (!pending->complete) {
> > + /* prepare has not completed */
> > + return H_BUSY;
> > + }
> > +
> > + newsize = 1ULL << pending->shift;
> > + rc = rehash_hpt(cpu, spapr->htab, HTAB_SIZE(spapr),
> > + pending->hpt, newsize);
> > + if (rc == H_SUCCESS) {
> > + CPUState *cs;
> > +
> > + qemu_vfree(spapr->htab);
> > + spapr->htab = pending->hpt;
> > + spapr->htab_shift = pending->shift;
> > +
> > + CPU_FOREACH(cs) {
> > + run_on_cpu(cs, pivot_hpt, RUN_ON_CPU_HOST_PTR(spapr));
> > + }
> > +
> > + pending->hpt = NULL; /* so it's not free()d */
> > + }
> > +
> > + /* Clean up */
> > + spapr->pending_hpt = NULL;
> > + free_pending_hpt(pending);
> > +
> > + return rc;
> > }
>
> Thomas
>
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type David Gibson
2016-12-09 7:59 ` Thomas Huth
@ 2016-12-09 13:25 ` Laurent Vivier
1 sibling, 0 replies; 18+ messages in thread
From: Laurent Vivier @ 2016-12-09 13:25 UTC (permalink / raw)
To: David Gibson, paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth
On 09/12/2016 03:23, David Gibson wrote:
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/ppc/spapr.c | 23 +++++++++++++++++++++--
> include/hw/compat.h | 3 +++
> 2 files changed, 24 insertions(+), 2 deletions(-)
>
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index cfadc46..0f25e83 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -2761,18 +2761,37 @@ static const TypeInfo spapr_machine_info = {
> type_init(spapr_machine_register_##suffix)
>
> /*
> + * pseries-2.9
> + */
> +static void spapr_machine_2_9_instance_options(MachineState *machine)
> +{
> +}
> +
> +static void spapr_machine_2_9_class_options(MachineClass *mc)
> +{
> + /* Defaults for the latest behaviour inherited from the base class */
> +}
> +
> +DEFINE_SPAPR_MACHINE(2_9, "2.9", true);
> +
> +/*
> * pseries-2.8
> */
> +#define SPAPR_COMPAT_2_8 \
> + HW_COMPAT_2_8
> +
> static void spapr_machine_2_8_instance_options(MachineState *machine)
> {
> + spapr_machine_2_9_instance_options(machine);
> }
>
> static void spapr_machine_2_8_class_options(MachineClass *mc)
> {
> - /* Defaults for the latest behaviour inherited from the base class */
> + spapr_machine_2_9_class_options(mc);
> + SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8);
> }
>
> -DEFINE_SPAPR_MACHINE(2_8, "2.8", true);
> +DEFINE_SPAPR_MACHINE(2_8, "2.8", false);
>
> /*
> * pseries-2.7
> diff --git a/include/hw/compat.h b/include/hw/compat.h
> index 0f06e11..f8b8354 100644
> --- a/include/hw/compat.h
> +++ b/include/hw/compat.h
> @@ -1,6 +1,9 @@
> #ifndef HW_COMPAT_H
> #define HW_COMPAT_H
>
> +#define HW_COMPAT_2_8 \
> + /* empty */
> +
> #define HW_COMPAT_2_7 \
> {\
> .driver = "virtio-pci",\
>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 4/5] pseries: Enable HPT resizing for 2.9
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 4/5] pseries: Enable HPT resizing for 2.9 David Gibson
@ 2016-12-09 13:26 ` Laurent Vivier
0 siblings, 0 replies; 18+ messages in thread
From: Laurent Vivier @ 2016-12-09 13:26 UTC (permalink / raw)
To: David Gibson, paulus, sjitindarsingh
Cc: agraf, mdroth, qemu-ppc, qemu-devel, thuth
On 09/12/2016 03:23, David Gibson wrote:
> We've now implemented a PAPR extensions which allows PAPR guests (i.e.
> "pseries" machine type) to resize their hash page table during runtime.
>
> However, that extension is only enabled if explicitly chosen on the
> command line. This patch enables it by default for spapr-2.9, but leaves
> it disabled (by default) for older machine types.
>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/ppc/spapr.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index f3b74dc..25468f6 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -2770,7 +2770,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
>
> smc->dr_lmb_enabled = true;
> smc->tcg_default_cpu = "POWER8";
> - smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
> + smc->resize_hpt_default = SPAPR_RESIZE_HPT_ENABLED;
> mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
> fwc->get_dev_path = spapr_get_fw_dev_path;
> nc->nmi_monitor_handler = spapr_nmi;
> @@ -2851,8 +2851,11 @@ static void spapr_machine_2_8_instance_options(MachineState *machine)
>
> static void spapr_machine_2_8_class_options(MachineClass *mc)
> {
> + sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
> +
> spapr_machine_2_9_class_options(mc);
> SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8);
> + smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
> }
>
> DEFINE_SPAPR_MACHINE(2_8, "2.8", false);
>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing David Gibson
2016-12-09 8:18 ` Thomas Huth
@ 2016-12-09 17:08 ` Michael Roth
2016-12-10 9:10 ` David Gibson
2016-12-12 0:40 ` David Gibson
1 sibling, 2 replies; 18+ messages in thread
From: Michael Roth @ 2016-12-09 17:08 UTC (permalink / raw)
To: David Gibson, paulus, sjitindarsingh
Cc: agraf, qemu-ppc, qemu-devel, thuth, lvivier
Quoting David Gibson (2016-12-08 20:23:11)
> This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
> H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
> extension to allow run time resizing of a guest's hash page table. It
> also adds a new machine property for controlling whether this new
> facility is available, and logic to check that against availability
> with KVM (only supported with KVM PR for now).
>
> Finally, it adds a new string to the hypertas property in the device
> tree, advertising to the guest the availability of the HPT resizing
> hypercalls. This is a tentative suggested value, and would need to be
> standardized by PAPR before being merged.
>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
> hw/ppc/trace-events | 2 ++
> include/hw/ppc/spapr.h | 11 +++++++++
> target-ppc/kvm.c | 26 ++++++++++++++++++++
> target-ppc/kvm_ppc.h | 5 ++++
> 6 files changed, 146 insertions(+)
>
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index 0f25e83..ecb0822 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
> if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
> add_str(hypertas, "hcall-multi-tce");
> }
> +
> + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
> + add_str(hypertas, "hcall-hpt-resize");
> + }
> +
> _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
> hypertas->str, hypertas->len));
> g_string_free(hypertas, TRUE);
> @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
> long load_limit, fw_size;
> char *filename;
> int smt = kvmppc_smt_threads();
> + Error *resize_hpt_err = NULL;
>
> msi_nonbroken = true;
>
> QLIST_INIT(&spapr->phbs);
>
> + /* Check HPT resizing availability */
> + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
> + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
> + if (resize_hpt_err) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> + error_free(resize_hpt_err);
> + resize_hpt_err = NULL;
> + } else {
> + spapr->resize_hpt = smc->resize_hpt_default;
> + }
> + }
> +
> + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
> +
> + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
> + error_report_err(resize_hpt_err);
> + exit(1);
> + }
I'm also a bit confused by this. Aren't semantics that HPT_ENABLED will allow
a fallback to non-resizable HPTs if the feature isn't available (whereas
HPT_REQUIRED won't)? I don't understand how such a fallback can ever be reached
with this check in place.
> +
> /* Allocate RMA if necessary */
> rma_alloc_size = kvmppc_alloc_rma(&rma);
>
> @@ -2236,6 +2261,40 @@ static void spapr_set_modern_hotplug_events(Object *obj, bool value,
> spapr->use_hotplug_event_source = value;
> }
>
> +static char *spapr_get_resize_hpt(Object *obj, Error **errp)
> +{
> + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> +
> + switch (spapr->resize_hpt) {
> + case SPAPR_RESIZE_HPT_DEFAULT:
> + return g_strdup("default");
> + case SPAPR_RESIZE_HPT_DISABLED:
> + return g_strdup("disabled");
> + case SPAPR_RESIZE_HPT_ENABLED:
> + return g_strdup("enabled");
> + case SPAPR_RESIZE_HPT_REQUIRED:
> + return g_strdup("required");
> + }
> + assert(0);
> +}
> +
> +static void spapr_set_resize_hpt(Object *obj, const char *value, Error **errp)
> +{
> + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> +
> + if (strcmp(value, "default") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DEFAULT;
> + } else if (strcmp(value, "disabled") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> + } else if (strcmp(value, "enabled") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_ENABLED;
> + } else if (strcmp(value, "required") == 0) {
> + spapr->resize_hpt = SPAPR_RESIZE_HPT_REQUIRED;
> + } else {
> + error_setg(errp, "Bad value for \"resize-hpt\" property");
> + }
> +}
> +
> static void spapr_machine_initfn(Object *obj)
> {
> sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> @@ -2256,6 +2315,12 @@ static void spapr_machine_initfn(Object *obj)
> " place of standard EPOW events when possible"
> " (required for memory hot-unplug support)",
> NULL);
> +
> + object_property_add_str(obj, "resize-hpt",
> + spapr_get_resize_hpt, spapr_set_resize_hpt, NULL);
> + object_property_set_description(obj, "resize-hpt",
> + "Resizing of the Hash Page Table (enabled, disabled, required)",
> + NULL);
> }
>
> static void spapr_machine_finalizefn(Object *obj)
> @@ -2707,6 +2772,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
>
> smc->dr_lmb_enabled = true;
> smc->tcg_default_cpu = "POWER8";
> + smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
> mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
> fwc->get_dev_path = spapr_get_fw_dev_path;
> nc->nmi_monitor_handler = spapr_nmi;
> diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
> index fd9f1d4..72a9c4d 100644
> --- a/hw/ppc/spapr_hcall.c
> +++ b/hw/ppc/spapr_hcall.c
> @@ -355,6 +355,38 @@ static target_ulong h_read(PowerPCCPU *cpu, sPAPRMachineState *spapr,
> return H_SUCCESS;
> }
>
> +static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
> + sPAPRMachineState *spapr,
> + target_ulong opcode,
> + target_ulong *args)
> +{
> + target_ulong flags = args[0];
> + target_ulong shift = args[1];
> +
> + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> + return H_AUTHORITY;
> + }
> +
> + trace_spapr_h_resize_hpt_prepare(flags, shift);
> + return H_HARDWARE;
> +}
> +
> +static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
> + sPAPRMachineState *spapr,
> + target_ulong opcode,
> + target_ulong *args)
> +{
> + target_ulong flags = args[0];
> + target_ulong shift = args[1];
> +
> + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> + return H_AUTHORITY;
> + }
> +
> + trace_spapr_h_resize_hpt_commit(flags, shift);
> + return H_HARDWARE;
> +}
> +
> static target_ulong h_set_sprg0(PowerPCCPU *cpu, sPAPRMachineState *spapr,
> target_ulong opcode, target_ulong *args)
> {
> @@ -1134,6 +1166,10 @@ static void hypercall_register_types(void)
> /* hcall-bulk */
> spapr_register_hypercall(H_BULK_REMOVE, h_bulk_remove);
>
> + /* hcall-hpt-resize */
> + spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
> + spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
> +
> /* hcall-splpar */
> spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
> spapr_register_hypercall(H_CEDE, h_cede);
> diff --git a/hw/ppc/trace-events b/hw/ppc/trace-events
> index 2297ead..bf59a8f 100644
> --- a/hw/ppc/trace-events
> +++ b/hw/ppc/trace-events
> @@ -16,6 +16,8 @@ spapr_cas_continue(unsigned long n) "Copy changes to the guest: %ld bytes"
> # hw/ppc/spapr_hcall.c
> spapr_cas_pvr_try(uint32_t pvr) "%x"
> spapr_cas_pvr(uint32_t cur_pvr, bool cpu_match, uint32_t new_pvr, uint64_t pcr) "current=%x, cpu_match=%u, new=%x, compat flags=%"PRIx64
> +spapr_h_resize_hpt_prepare(uint64_t flags, uint64_t shift) "flags=0x%"PRIx64", shift=%"PRIu64
> +spapr_h_resize_hpt_commit(uint64_t flags, uint64_t shift) "flags=0x%"PRIx64", shift=%"PRIu64
>
> # hw/ppc/spapr_iommu.c
> spapr_iommu_put(uint64_t liobn, uint64_t ioba, uint64_t tce, uint64_t ret) "liobn=%"PRIx64" ioba=0x%"PRIx64" tce=0x%"PRIx64" ret=%"PRId64
> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
> index a2d8964..d2c758b 100644
> --- a/include/hw/ppc/spapr.h
> +++ b/include/hw/ppc/spapr.h
> @@ -31,6 +31,13 @@ typedef struct sPAPRMachineState sPAPRMachineState;
> #define SPAPR_MACHINE_CLASS(klass) \
> OBJECT_CLASS_CHECK(sPAPRMachineClass, klass, TYPE_SPAPR_MACHINE)
>
> +typedef enum {
> + SPAPR_RESIZE_HPT_DEFAULT = 0,
> + SPAPR_RESIZE_HPT_DISABLED,
> + SPAPR_RESIZE_HPT_ENABLED,
> + SPAPR_RESIZE_HPT_REQUIRED,
> +} sPAPRResizeHPT;
> +
> /**
> * sPAPRMachineClass:
> */
> @@ -46,6 +53,7 @@ struct sPAPRMachineClass {
> uint64_t *buid, hwaddr *pio,
> hwaddr *mmio32, hwaddr *mmio64,
> unsigned n_dma, uint32_t *liobns, Error **errp);
> + sPAPRResizeHPT resize_hpt_default;
> };
>
> /**
> @@ -61,6 +69,7 @@ struct sPAPRMachineState {
> XICSState *xics;
> DeviceState *rtc;
>
> + sPAPRResizeHPT resize_hpt;
> void *htab;
> uint32_t htab_shift;
> hwaddr rma_size;
> @@ -347,6 +356,8 @@ struct sPAPRMachineState {
> #define H_XIRR_X 0x2FC
> #define H_RANDOM 0x300
> #define H_SET_MODE 0x31C
> +#define H_RESIZE_HPT_PREPARE 0x36C
> +#define H_RESIZE_HPT_COMMIT 0x370
> #define H_SIGNAL_SYS_RESET 0x380
> #define MAX_HCALL_OPCODE H_SIGNAL_SYS_RESET
>
> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> index 15e12f3..dcd8cef 100644
> --- a/target-ppc/kvm.c
> +++ b/target-ppc/kvm.c
> @@ -22,6 +22,7 @@
> #include <linux/kvm.h>
>
> #include "qemu-common.h"
> +#include "qapi/error.h"
> #include "qemu/error-report.h"
> #include "cpu.h"
> #include "qemu/timer.h"
> @@ -2269,6 +2270,7 @@ int kvmppc_reset_htab(int shift_hint)
> /* Full emulation, tell caller to allocate htab itself */
> return 0;
> }
> +
> if (kvm_check_extension(kvm_state, KVM_CAP_PPC_ALLOC_HTAB)) {
> int ret;
> ret = kvm_vm_ioctl(kvm_state, KVM_PPC_ALLOCATE_HTAB, &shift);
> @@ -2672,3 +2674,27 @@ int kvmppc_enable_hwrng(void)
>
> return kvmppc_enable_hcall(kvm_state, H_RANDOM);
> }
> +
> +void kvmppc_check_papr_resize_hpt(Error **errp)
> +{
> + if (!kvm_enabled()) {
> + return;
> + }
> +
> + /* TODO: Check specific capabilities for HPT resize aware host kernels */
> +
> + /*
> + * It's tempting to try to check directly if the HPT is under our
> + * control or KVM's, which is what's really relevant here.
> + * Unfortunately, in order to correctly size the HPT, we need to
> + * know if we can do resizing, _before_ we attempt to allocate it
> + * with KVM. Before that point, we don't officially know whether
> + * we'll control the HPT or not. So we have to use a fallback
> + * test for PR vs HV KVM to predict that.
> + */
> + if (kvmppc_is_pr(kvm_state)) {
> + return;
> + }
> +
> + error_setg(errp, "Hash page table resizing not available with this KVM version");
> +}
> diff --git a/target-ppc/kvm_ppc.h b/target-ppc/kvm_ppc.h
> index 841a29b..3e852ba 100644
> --- a/target-ppc/kvm_ppc.h
> +++ b/target-ppc/kvm_ppc.h
> @@ -59,6 +59,7 @@ bool kvmppc_has_cap_htm(void);
> int kvmppc_enable_hwrng(void);
> int kvmppc_put_books_sregs(PowerPCCPU *cpu);
> PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
> +void kvmppc_check_papr_resize_hpt(Error **errp);
>
> #else
>
> @@ -270,6 +271,10 @@ static inline PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void)
> return NULL;
> }
>
> +static inline void kvmppc_check_papr_resize_hpt(Error **errp)
> +{
> + return;
> +}
> #endif
>
> #ifndef CONFIG_KVM
> --
> 2.9.3
>
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 17:08 ` Michael Roth
@ 2016-12-10 9:10 ` David Gibson
2016-12-12 0:40 ` David Gibson
1 sibling, 0 replies; 18+ messages in thread
From: David Gibson @ 2016-12-10 9:10 UTC (permalink / raw)
To: Michael Roth
Cc: paulus, sjitindarsingh, agraf, qemu-ppc, qemu-devel, thuth, lvivier
[-- Attachment #1: Type: text/plain, Size: 13638 bytes --]
On Fri, Dec 09, 2016 at 11:08:02AM -0600, Michael Roth wrote:
> Quoting David Gibson (2016-12-08 20:23:11)
> > This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
> > H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
> > extension to allow run time resizing of a guest's hash page table. It
> > also adds a new machine property for controlling whether this new
> > facility is available, and logic to check that against availability
> > with KVM (only supported with KVM PR for now).
> >
> > Finally, it adds a new string to the hypertas property in the device
> > tree, advertising to the guest the availability of the HPT resizing
> > hypercalls. This is a tentative suggested value, and would need to be
> > standardized by PAPR before being merged.
> >
> > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> > hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> > hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
> > hw/ppc/trace-events | 2 ++
> > include/hw/ppc/spapr.h | 11 +++++++++
> > target-ppc/kvm.c | 26 ++++++++++++++++++++
> > target-ppc/kvm_ppc.h | 5 ++++
> > 6 files changed, 146 insertions(+)
> >
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index 0f25e83..ecb0822 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
> > if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
> > add_str(hypertas, "hcall-multi-tce");
> > }
> > +
> > + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
> > + add_str(hypertas, "hcall-hpt-resize");
> > + }
> > +
> > _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
> > hypertas->str, hypertas->len));
> > g_string_free(hypertas, TRUE);
> > @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
> > long load_limit, fw_size;
> > char *filename;
> > int smt = kvmppc_smt_threads();
> > + Error *resize_hpt_err = NULL;
> >
> > msi_nonbroken = true;
> >
> > QLIST_INIT(&spapr->phbs);
> >
> > + /* Check HPT resizing availability */
> > + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
> > + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
> > + if (resize_hpt_err) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> > + error_free(resize_hpt_err);
> > + resize_hpt_err = NULL;
> > + } else {
> > + spapr->resize_hpt = smc->resize_hpt_default;
> > + }
> > + }
> > +
> > + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
> > +
> > + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
> > + error_report_err(resize_hpt_err);
> > + exit(1);
> > + }
>
> I'm also a bit confused by this. Aren't semantics that HPT_ENABLED will allow
> a fallback to non-resizable HPTs if the feature isn't available (whereas
> HPT_REQUIRED won't)? I don't understand how such a fallback can ever be reached
> with this check in place.
resize-hpt=enabled means we'll fall back if the guest doesn't support
it. An error here means the _host_ can't support it. I think not
falling back is the better option in this case - if you've explicitly
tried to enable it from it should be available or fail. In particular
this will make it safer for migrations where the source is already
using hpt resizing.
> > +
> > /* Allocate RMA if necessary */
> > rma_alloc_size = kvmppc_alloc_rma(&rma);
> >
> > @@ -2236,6 +2261,40 @@ static void spapr_set_modern_hotplug_events(Object *obj, bool value,
> > spapr->use_hotplug_event_source = value;
> > }
> >
> > +static char *spapr_get_resize_hpt(Object *obj, Error **errp)
> > +{
> > + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> > +
> > + switch (spapr->resize_hpt) {
> > + case SPAPR_RESIZE_HPT_DEFAULT:
> > + return g_strdup("default");
> > + case SPAPR_RESIZE_HPT_DISABLED:
> > + return g_strdup("disabled");
> > + case SPAPR_RESIZE_HPT_ENABLED:
> > + return g_strdup("enabled");
> > + case SPAPR_RESIZE_HPT_REQUIRED:
> > + return g_strdup("required");
> > + }
> > + assert(0);
> > +}
> > +
> > +static void spapr_set_resize_hpt(Object *obj, const char *value, Error **errp)
> > +{
> > + sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> > +
> > + if (strcmp(value, "default") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DEFAULT;
> > + } else if (strcmp(value, "disabled") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> > + } else if (strcmp(value, "enabled") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_ENABLED;
> > + } else if (strcmp(value, "required") == 0) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_REQUIRED;
> > + } else {
> > + error_setg(errp, "Bad value for \"resize-hpt\" property");
> > + }
> > +}
> > +
> > static void spapr_machine_initfn(Object *obj)
> > {
> > sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
> > @@ -2256,6 +2315,12 @@ static void spapr_machine_initfn(Object *obj)
> > " place of standard EPOW events when possible"
> > " (required for memory hot-unplug support)",
> > NULL);
> > +
> > + object_property_add_str(obj, "resize-hpt",
> > + spapr_get_resize_hpt, spapr_set_resize_hpt, NULL);
> > + object_property_set_description(obj, "resize-hpt",
> > + "Resizing of the Hash Page Table (enabled, disabled, required)",
> > + NULL);
> > }
> >
> > static void spapr_machine_finalizefn(Object *obj)
> > @@ -2707,6 +2772,7 @@ static void spapr_machine_class_init(ObjectClass *oc, void *data)
> >
> > smc->dr_lmb_enabled = true;
> > smc->tcg_default_cpu = "POWER8";
> > + smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED;
> > mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
> > fwc->get_dev_path = spapr_get_fw_dev_path;
> > nc->nmi_monitor_handler = spapr_nmi;
> > diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
> > index fd9f1d4..72a9c4d 100644
> > --- a/hw/ppc/spapr_hcall.c
> > +++ b/hw/ppc/spapr_hcall.c
> > @@ -355,6 +355,38 @@ static target_ulong h_read(PowerPCCPU *cpu, sPAPRMachineState *spapr,
> > return H_SUCCESS;
> > }
> >
> > +static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
> > + sPAPRMachineState *spapr,
> > + target_ulong opcode,
> > + target_ulong *args)
> > +{
> > + target_ulong flags = args[0];
> > + target_ulong shift = args[1];
> > +
> > + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> > + return H_AUTHORITY;
> > + }
> > +
> > + trace_spapr_h_resize_hpt_prepare(flags, shift);
> > + return H_HARDWARE;
> > +}
> > +
> > +static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
> > + sPAPRMachineState *spapr,
> > + target_ulong opcode,
> > + target_ulong *args)
> > +{
> > + target_ulong flags = args[0];
> > + target_ulong shift = args[1];
> > +
> > + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
> > + return H_AUTHORITY;
> > + }
> > +
> > + trace_spapr_h_resize_hpt_commit(flags, shift);
> > + return H_HARDWARE;
> > +}
> > +
> > static target_ulong h_set_sprg0(PowerPCCPU *cpu, sPAPRMachineState *spapr,
> > target_ulong opcode, target_ulong *args)
> > {
> > @@ -1134,6 +1166,10 @@ static void hypercall_register_types(void)
> > /* hcall-bulk */
> > spapr_register_hypercall(H_BULK_REMOVE, h_bulk_remove);
> >
> > + /* hcall-hpt-resize */
> > + spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
> > + spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
> > +
> > /* hcall-splpar */
> > spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
> > spapr_register_hypercall(H_CEDE, h_cede);
> > diff --git a/hw/ppc/trace-events b/hw/ppc/trace-events
> > index 2297ead..bf59a8f 100644
> > --- a/hw/ppc/trace-events
> > +++ b/hw/ppc/trace-events
> > @@ -16,6 +16,8 @@ spapr_cas_continue(unsigned long n) "Copy changes to the guest: %ld bytes"
> > # hw/ppc/spapr_hcall.c
> > spapr_cas_pvr_try(uint32_t pvr) "%x"
> > spapr_cas_pvr(uint32_t cur_pvr, bool cpu_match, uint32_t new_pvr, uint64_t pcr) "current=%x, cpu_match=%u, new=%x, compat flags=%"PRIx64
> > +spapr_h_resize_hpt_prepare(uint64_t flags, uint64_t shift) "flags=0x%"PRIx64", shift=%"PRIu64
> > +spapr_h_resize_hpt_commit(uint64_t flags, uint64_t shift) "flags=0x%"PRIx64", shift=%"PRIu64
> >
> > # hw/ppc/spapr_iommu.c
> > spapr_iommu_put(uint64_t liobn, uint64_t ioba, uint64_t tce, uint64_t ret) "liobn=%"PRIx64" ioba=0x%"PRIx64" tce=0x%"PRIx64" ret=%"PRId64
> > diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
> > index a2d8964..d2c758b 100644
> > --- a/include/hw/ppc/spapr.h
> > +++ b/include/hw/ppc/spapr.h
> > @@ -31,6 +31,13 @@ typedef struct sPAPRMachineState sPAPRMachineState;
> > #define SPAPR_MACHINE_CLASS(klass) \
> > OBJECT_CLASS_CHECK(sPAPRMachineClass, klass, TYPE_SPAPR_MACHINE)
> >
> > +typedef enum {
> > + SPAPR_RESIZE_HPT_DEFAULT = 0,
> > + SPAPR_RESIZE_HPT_DISABLED,
> > + SPAPR_RESIZE_HPT_ENABLED,
> > + SPAPR_RESIZE_HPT_REQUIRED,
> > +} sPAPRResizeHPT;
> > +
> > /**
> > * sPAPRMachineClass:
> > */
> > @@ -46,6 +53,7 @@ struct sPAPRMachineClass {
> > uint64_t *buid, hwaddr *pio,
> > hwaddr *mmio32, hwaddr *mmio64,
> > unsigned n_dma, uint32_t *liobns, Error **errp);
> > + sPAPRResizeHPT resize_hpt_default;
> > };
> >
> > /**
> > @@ -61,6 +69,7 @@ struct sPAPRMachineState {
> > XICSState *xics;
> > DeviceState *rtc;
> >
> > + sPAPRResizeHPT resize_hpt;
> > void *htab;
> > uint32_t htab_shift;
> > hwaddr rma_size;
> > @@ -347,6 +356,8 @@ struct sPAPRMachineState {
> > #define H_XIRR_X 0x2FC
> > #define H_RANDOM 0x300
> > #define H_SET_MODE 0x31C
> > +#define H_RESIZE_HPT_PREPARE 0x36C
> > +#define H_RESIZE_HPT_COMMIT 0x370
> > #define H_SIGNAL_SYS_RESET 0x380
> > #define MAX_HCALL_OPCODE H_SIGNAL_SYS_RESET
> >
> > diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> > index 15e12f3..dcd8cef 100644
> > --- a/target-ppc/kvm.c
> > +++ b/target-ppc/kvm.c
> > @@ -22,6 +22,7 @@
> > #include <linux/kvm.h>
> >
> > #include "qemu-common.h"
> > +#include "qapi/error.h"
> > #include "qemu/error-report.h"
> > #include "cpu.h"
> > #include "qemu/timer.h"
> > @@ -2269,6 +2270,7 @@ int kvmppc_reset_htab(int shift_hint)
> > /* Full emulation, tell caller to allocate htab itself */
> > return 0;
> > }
> > +
> > if (kvm_check_extension(kvm_state, KVM_CAP_PPC_ALLOC_HTAB)) {
> > int ret;
> > ret = kvm_vm_ioctl(kvm_state, KVM_PPC_ALLOCATE_HTAB, &shift);
> > @@ -2672,3 +2674,27 @@ int kvmppc_enable_hwrng(void)
> >
> > return kvmppc_enable_hcall(kvm_state, H_RANDOM);
> > }
> > +
> > +void kvmppc_check_papr_resize_hpt(Error **errp)
> > +{
> > + if (!kvm_enabled()) {
> > + return;
> > + }
> > +
> > + /* TODO: Check specific capabilities for HPT resize aware host kernels */
> > +
> > + /*
> > + * It's tempting to try to check directly if the HPT is under our
> > + * control or KVM's, which is what's really relevant here.
> > + * Unfortunately, in order to correctly size the HPT, we need to
> > + * know if we can do resizing, _before_ we attempt to allocate it
> > + * with KVM. Before that point, we don't officially know whether
> > + * we'll control the HPT or not. So we have to use a fallback
> > + * test for PR vs HV KVM to predict that.
> > + */
> > + if (kvmppc_is_pr(kvm_state)) {
> > + return;
> > + }
> > +
> > + error_setg(errp, "Hash page table resizing not available with this KVM version");
> > +}
> > diff --git a/target-ppc/kvm_ppc.h b/target-ppc/kvm_ppc.h
> > index 841a29b..3e852ba 100644
> > --- a/target-ppc/kvm_ppc.h
> > +++ b/target-ppc/kvm_ppc.h
> > @@ -59,6 +59,7 @@ bool kvmppc_has_cap_htm(void);
> > int kvmppc_enable_hwrng(void);
> > int kvmppc_put_books_sregs(PowerPCCPU *cpu);
> > PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
> > +void kvmppc_check_papr_resize_hpt(Error **errp);
> >
> > #else
> >
> > @@ -270,6 +271,10 @@ static inline PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void)
> > return NULL;
> > }
> >
> > +static inline void kvmppc_check_papr_resize_hpt(Error **errp)
> > +{
> > + return;
> > +}
> > #endif
> >
> > #ifndef CONFIG_KVM
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 17:08 ` Michael Roth
2016-12-10 9:10 ` David Gibson
@ 2016-12-12 0:40 ` David Gibson
1 sibling, 0 replies; 18+ messages in thread
From: David Gibson @ 2016-12-12 0:40 UTC (permalink / raw)
To: Michael Roth
Cc: paulus, sjitindarsingh, agraf, qemu-ppc, qemu-devel, thuth, lvivier
[-- Attachment #1: Type: text/plain, Size: 3601 bytes --]
On Fri, Dec 09, 2016 at 11:08:02AM -0600, Michael Roth wrote:
> Quoting David Gibson (2016-12-08 20:23:11)
> > This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
> > H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
> > extension to allow run time resizing of a guest's hash page table. It
> > also adds a new machine property for controlling whether this new
> > facility is available, and logic to check that against availability
> > with KVM (only supported with KVM PR for now).
> >
> > Finally, it adds a new string to the hypertas property in the device
> > tree, advertising to the guest the availability of the HPT resizing
> > hypercalls. This is a tentative suggested value, and would need to be
> > standardized by PAPR before being merged.
> >
> > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> > hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> > hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
> > hw/ppc/trace-events | 2 ++
> > include/hw/ppc/spapr.h | 11 +++++++++
> > target-ppc/kvm.c | 26 ++++++++++++++++++++
> > target-ppc/kvm_ppc.h | 5 ++++
> > 6 files changed, 146 insertions(+)
> >
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index 0f25e83..ecb0822 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
> > if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
> > add_str(hypertas, "hcall-multi-tce");
> > }
> > +
> > + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
> > + add_str(hypertas, "hcall-hpt-resize");
> > + }
> > +
> > _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
> > hypertas->str, hypertas->len));
> > g_string_free(hypertas, TRUE);
> > @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
> > long load_limit, fw_size;
> > char *filename;
> > int smt = kvmppc_smt_threads();
> > + Error *resize_hpt_err = NULL;
> >
> > msi_nonbroken = true;
> >
> > QLIST_INIT(&spapr->phbs);
> >
> > + /* Check HPT resizing availability */
> > + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
> > + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
> > + if (resize_hpt_err) {
> > + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> > + error_free(resize_hpt_err);
> > + resize_hpt_err = NULL;
> > + } else {
> > + spapr->resize_hpt = smc->resize_hpt_default;
> > + }
> > + }
> > +
> > + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
> > +
> > + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
> > + error_report_err(resize_hpt_err);
> > + exit(1);
> > + }
>
> I'm also a bit confused by this. Aren't semantics that HPT_ENABLED will allow
> a fallback to non-resizable HPTs if the feature isn't available (whereas
> HPT_REQUIRED won't)? I don't understand how such a fallback can ever be reached
> with this check in place.
We fall back if the _guest_ doesn't support HPT resizing. This error
path is if the _host_ can't support HPT resizing. Some extra
information about this in the reply I'll shortly send to Thomas.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing
2016-12-09 9:18 ` Thomas Huth
@ 2016-12-12 1:06 ` David Gibson
0 siblings, 0 replies; 18+ messages in thread
From: David Gibson @ 2016-12-12 1:06 UTC (permalink / raw)
To: Thomas Huth
Cc: paulus, sjitindarsingh, agraf, mdroth, qemu-ppc, qemu-devel, lvivier
[-- Attachment #1: Type: text/plain, Size: 6205 bytes --]
On Fri, Dec 09, 2016 at 10:18:59AM +0100, Thomas Huth wrote:
> On 09.12.2016 10:09, David Gibson wrote:
> > On Fri, Dec 09, 2016 at 09:18:51AM +0100, Thomas Huth wrote:
> >> On 09.12.2016 03:23, David Gibson wrote:
> >>> This introduces stub implementations of the H_RESIZE_HPT_PREPARE and
> >>> H_RESIZE_HPT_COMMIT hypercalls which we hope to add in a PAPR
> >>> extension to allow run time resizing of a guest's hash page table. It
> >>> also adds a new machine property for controlling whether this new
> >>> facility is available, and logic to check that against availability
> >>> with KVM (only supported with KVM PR for now).
> >>>
> >>> Finally, it adds a new string to the hypertas property in the device
> >>> tree, advertising to the guest the availability of the HPT resizing
> >>> hypercalls. This is a tentative suggested value, and would need to be
> >>> standardized by PAPR before being merged.
> >>>
> >>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> >>> ---
> >>> hw/ppc/spapr.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++
> >>> hw/ppc/spapr_hcall.c | 36 +++++++++++++++++++++++++++
> >>> hw/ppc/trace-events | 2 ++
> >>> include/hw/ppc/spapr.h | 11 +++++++++
> >>> target-ppc/kvm.c | 26 ++++++++++++++++++++
> >>> target-ppc/kvm_ppc.h | 5 ++++
> >>> 6 files changed, 146 insertions(+)
> >>>
> >>> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> >>> index 0f25e83..ecb0822 100644
> >>> --- a/hw/ppc/spapr.c
> >>> +++ b/hw/ppc/spapr.c
> >>> @@ -760,6 +760,11 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
> >>> if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
> >>> add_str(hypertas, "hcall-multi-tce");
> >>> }
> >>> +
> >>> + if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) {
> >>> + add_str(hypertas, "hcall-hpt-resize");
> >>> + }
> >>> +
> >>> _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions",
> >>> hypertas->str, hypertas->len));
> >>> g_string_free(hypertas, TRUE);
> >>> @@ -1839,11 +1844,31 @@ static void ppc_spapr_init(MachineState *machine)
> >>> long load_limit, fw_size;
> >>> char *filename;
> >>> int smt = kvmppc_smt_threads();
> >>> + Error *resize_hpt_err = NULL;
> >>>
> >>> msi_nonbroken = true;
> >>>
> >>> QLIST_INIT(&spapr->phbs);
> >>>
> >>> + /* Check HPT resizing availability */
> >>> + kvmppc_check_papr_resize_hpt(&resize_hpt_err);
> >>> + if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) {
> >>> + if (resize_hpt_err) {
> >>> + spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED;
> >>> + error_free(resize_hpt_err);
> >>> + resize_hpt_err = NULL;
> >>> + } else {
> >>> + spapr->resize_hpt = smc->resize_hpt_default;
> >>> + }
> >>> + }
> >>> +
> >>> + assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT);
> >>> +
> >>> + if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) {
> >>> + error_report_err(resize_hpt_err);
> >>> + exit(1);
> >>> + }
> >>
> >> The logic here is IMHO a little bit hard to understand. Why do you need
> >> the SPAPR_RESIZE_HPT_DEFAULT state here at all? Wouldn't it be
> >> sufficient to start with SPAPR_RESIZE_HPT_DISABLED by default and to
> >> only enable it when the extension is available?
> >
> > Because if the user explicitly sets something, we should never
> > override itin the logic here. In order to determine whether the user
> > has set something explicitly, we need another value that means "not
> > set explicitly".
>
> Hmm, ok, I think I slowly get it. Another question: Does this also take
> care of the handling if the user migrates from a machine that supports
> HPT resizing to a machine that does not support this feature?
Ok, so there are a fair few cases here. The key point, though is that
even before now, we helpefully transferred the size of the HPT in the
migration stream before sending its contents (that was done so we
could update the sizing heuristics without breaking migration).
For TCG and KVM PR:
Here the ability to HPT resize depends purely on the qemu version.
Assuming this gets merged for qemu-2.9, pseries-2.9 TCG or PR guests
should always support resizing so that should be fine.
If you explicitly enabled HPT resizing for a pseries-2.8 guest running
under qemu-2.9 to an older qemu, then things could get odd. The
destination guest would get an HPT of the size it had at the source at
the point of migration. Any further attempts to resize would fail.
Given you have to explicitly set up this poor situation and the
failure mode isn't _that_ bad, I think that's ok.
For KVM HV:
Assuming we're migrating from a host kernel which can support resize
to one which can't:
If the resize-hpt is (explicitly) set differently on the the two ends
then, well, don't do that. So assume it's the same on each end:
If resize-hpt=disabled, things are fine, obviously.
If resize=hpt=enabled|required, then qemu will fail to start on the
destination because the host kernel doesn't support it. So you should
know something's wrong before even attempting migration, which is the
best we can do.
If resize-hpt=default is the messiest case. In theory we could get a
failure case to the one mentioned above for TCG guests. In practice,
however, the migration will fail before completing: kernels without
the HPT resizing support in fact don't even allow the guest HPT to be
resized at reset time. The initial setup of the destination will
initialize the in-kernel HPT based on maxmem size (because resize is
unavailable on this end). When the migration stream comes in, we'll
try to reallocate the in-kernel HPT to the incoming size and fail
because the host kernel won't allow us to.
I think this behaviour is as good as we can reasonably get.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
end of thread, other threads:[~2016-12-12 3:03 UTC | newest]
Thread overview: 18+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-12-09 2:23 [Qemu-devel] [for-2.9 0/5] Hash Page Table resizing for TCG pseries guests David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 1/5] pseries: Add pseries-2.9 machine type David Gibson
2016-12-09 7:59 ` Thomas Huth
2016-12-09 13:25 ` Laurent Vivier
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 2/5] pseries: Stubs for HPT resizing David Gibson
2016-12-09 8:18 ` Thomas Huth
2016-12-09 9:09 ` David Gibson
2016-12-09 9:18 ` Thomas Huth
2016-12-12 1:06 ` David Gibson
2016-12-09 17:08 ` Michael Roth
2016-12-10 9:10 ` David Gibson
2016-12-12 0:40 ` David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 3/5] pseries: Implement " David Gibson
2016-12-09 8:36 ` Thomas Huth
2016-12-09 9:19 ` David Gibson
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 4/5] pseries: Enable HPT resizing for 2.9 David Gibson
2016-12-09 13:26 ` Laurent Vivier
2016-12-09 2:23 ` [Qemu-devel] [for-2.9 5/5] pseries: Use smaller default hash page tables when guest can resize David Gibson
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.