motorhead/qemu
1
0
Fork 0
mirror of https://github.com/Motorhead1991/qemu.git synced 2025-08-04 00:03:54 -06:00
Code Issues Projects Releases Packages Wiki Activity Actions

spapr: nested: Add support for reporting Hostwide state counter

Browse source 
Add support for reporting Hostwide state counters for nested KVM pseries
guests running with 'cap-nested-papr' on Qemu-TCG acting as
L0-hypervisor. The Hostwide state counters are statistics about state that
L0-hypervisor maintains for the L2-guests and represent the state of all
L2-guests, not just a specific one.

These stats counters are exposed to L1-Hypervisor by the L0-Hypervisor via a
new bit-flag named 'getHostWideState' for the H_GUEST_GET_STATE hcall which
is documented at [1]. Once this flag is set the hcall should populate the
Guest-State-Elements in the requested GSB with the stat counter
values. Currently following five counters are supported:

* l0_guest_heap_size_inuse
* l0_guest_heap_size_max
* l0_guest_pagetable_size_inuse
* l0_guest_pagetable_size_max
* l0_guest_pagetable_reclaimed

At the moment '0' is being reported for all these counters as these
counters doesn't align with how L0-Qemu manages Guest memory.

The patch implements support for these counters by adding new members to
the 'struct SpaprMachineStateNested'. These new members are then plugged
into the existing 'guest_state_element_types[]' with the help of a new
macro 'GSBE_NESTED_MACHINE_DW' together with a new helper
'get_machine_ptr()'. guest_state_request_check() is updated to ensure
correctness of the requested GSB and finally h_guest_getset_state() is
updated to handle the newly introduced flag
'GUEST_STATE_REQUEST_HOST_WIDE'.

This patch is tested with the proposed linux-kernel implementation to
expose these stat-counter as perf-events at [2].

[1]
https://lore.kernel.org/all/20241222140247.174998-2-vaibhav@linux.ibm.com

[2]
https://lore.kernel.org/all/20241222140247.174998-1-vaibhav@linux.ibm.com

Signed-off-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Message-ID: <20250221155449.530645-1-vaibhav@linux.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
This commit is contained in:
Vaibhav Jain 2025-02-21 21:24:48 +05:30 • committed by Nicholas Piggin
parent 5f361ea187
commit 5f7d861e65
2 changed files with 147 additions and 39 deletions
Download patch file Download diff file Expand all files Collapse all files

119
hw/ppc/spapr_nested.c
View file

@ -65,10 +65,9 @@ static
SpaprMachineStateNestedGuest *spapr_get_nested_guest(SpaprMachineState *spapr,
target_ulong guestid)
{
SpaprMachineStateNestedGuest *guest;
guest = g_hash_table_lookup(spapr->nested.guests, GINT_TO_POINTER(guestid));
return guest;
return spapr->nested.guests ?
g_hash_table_lookup(spapr->nested.guests,
GINT_TO_POINTER(guestid)) : NULL;
}
bool spapr_get_pate_nested_papr(SpaprMachineState *spapr, PowerPCCPU *cpu,
@ -594,26 +593,37 @@ static bool spapr_nested_vcpu_check(SpaprMachineStateNestedGuest *guest,
return false;
}
static void *get_vcpu_state_ptr(SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid)
static void *get_vcpu_state_ptr(SpaprMachineState *spapr,
SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid)
{
assert(spapr_nested_vcpu_check(guest, vcpuid, false));
return &guest->vcpus[vcpuid].state;
}
static void *get_vcpu_ptr(SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid)
static void *get_vcpu_ptr(SpaprMachineState *spapr,
SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid)
{
assert(spapr_nested_vcpu_check(guest, vcpuid, false));
return &guest->vcpus[vcpuid];
}
static void *get_guest_ptr(SpaprMachineStateNestedGuest *guest,
static void *get_guest_ptr(SpaprMachineState *spapr,
SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid)
{
return guest; /* for GSBE_NESTED */
}
static void *get_machine_ptr(SpaprMachineState *spapr,
SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid)
{
/* ignore guest and vcpuid for this */
return &spapr->nested;
}
/*
* set=1 means the L1 is trying to set some state
* set=0 means the L1 is trying to get some state
@ -1013,7 +1023,15 @@ struct guest_state_element_type guest_state_element_types[] = {
GSBE_NESTED_VCPU(GSB_VCPU_OUT_BUFFER, 0x10, runbufout, copy_state_runbuf),
GSBE_NESTED_VCPU(GSB_VCPU_OUT_BUF_MIN_SZ, 0x8, runbufout, out_buf_min_size),
GSBE_NESTED_VCPU(GSB_VCPU_HDEC_EXPIRY_TB, 0x8, hdecr_expiry_tb,
copy_state_hdecr)
copy_state_hdecr),
GSBE_NESTED_MACHINE_DW(GSB_L0_GUEST_HEAP_INUSE, l0_guest_heap_inuse),
GSBE_NESTED_MACHINE_DW(GSB_L0_GUEST_HEAP_MAX, l0_guest_heap_max),
GSBE_NESTED_MACHINE_DW(GSB_L0_GUEST_PGTABLE_SIZE_INUSE,
l0_guest_pgtable_size_inuse),
GSBE_NESTED_MACHINE_DW(GSB_L0_GUEST_PGTABLE_SIZE_MAX,
l0_guest_pgtable_size_max),
GSBE_NESTED_MACHINE_DW(GSB_L0_GUEST_PGTABLE_RECLAIMED,
l0_guest_pgtable_reclaimed),
};
void spapr_nested_gsb_init(void)
@ -1031,8 +1049,13 @@ void spapr_nested_gsb_init(void)
else if (type->id >= GSB_VCPU_IN_BUFFER)
/* 0x0c00 - 0xf000 Thread + RW */
type->flags = 0;
else if (type->id >= GSB_L0_GUEST_HEAP_INUSE)
/*0x0800 - 0x0804 Hostwide Counters + RO */
type->flags = GUEST_STATE_ELEMENT_TYPE_FLAG_HOST_WIDE |
GUEST_STATE_ELEMENT_TYPE_FLAG_READ_ONLY;
else if (type->id >= GSB_VCPU_LPVR)
/* 0x0003 - 0x0bff Guest + RW */
/* 0x0003 - 0x07ff Guest + RW */
type->flags = GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE;
else if (type->id >= GSB_HV_VCPU_STATE_SIZE)
/* 0x0001 - 0x0002 Guest + RO */
@ -1139,18 +1162,26 @@ static bool guest_state_request_check(struct guest_state_request *gsr)
return false;
}
if (type->flags & GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE) {
if (type->flags & GUEST_STATE_ELEMENT_TYPE_FLAG_HOST_WIDE) {
/* Hostwide elements cant be clubbed with other types */
if (!(gsr->flags & GUEST_STATE_REQUEST_HOST_WIDE)) {
qemu_log_mask(LOG_GUEST_ERROR, "trying to get/set a host wide "
"Element ID:%04x.\n", id);
return false;
}
} else if (type->flags & GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE) {
/* guest wide element type */
if (!(gsr->flags & GUEST_STATE_REQUEST_GUEST_WIDE)) {
qemu_log_mask(LOG_GUEST_ERROR, "trying to set a guest wide "
qemu_log_mask(LOG_GUEST_ERROR, "trying to get/set a guest wide "
"Element ID:%04x.\n", id);
return false;
}
} else {
/* thread wide element type */
if (gsr->flags & GUEST_STATE_REQUEST_GUEST_WIDE) {
qemu_log_mask(LOG_GUEST_ERROR, "trying to set a thread wide "
"Element ID:%04x.\n", id);
if (gsr->flags & (GUEST_STATE_REQUEST_GUEST_WIDE |
GUEST_STATE_REQUEST_HOST_WIDE)) {
qemu_log_mask(LOG_GUEST_ERROR, "trying to get/set a thread wide"
" Element ID:%04x.\n", id);
return false;
}
}
@ -1419,7 +1450,8 @@ static target_ulong h_guest_create_vcpu(PowerPCCPU *cpu,
return H_SUCCESS;
}
static target_ulong getset_state(SpaprMachineStateNestedGuest *guest,
static target_ulong getset_state(SpaprMachineState *spapr,
SpaprMachineStateNestedGuest *guest,
uint64_t vcpuid,
struct guest_state_request *gsr)
{
@ -1452,7 +1484,7 @@ static target_ulong getset_state(SpaprMachineStateNestedGuest *guest,
/* Get pointer to guest data to get/set */
if (type->location && type->copy) {
ptr = type->location(guest, vcpuid);
ptr = type->location(spapr, guest, vcpuid);
assert(ptr);
if (!~(type->mask) && is_gsr_invalid(gsr, element, type)) {
return H_INVALID_ELEMENT_VALUE;
@ -1469,6 +1501,7 @@ next_element:
}
static target_ulong map_and_getset_state(PowerPCCPU *cpu,
SpaprMachineState *spapr,
SpaprMachineStateNestedGuest *guest,
uint64_t vcpuid,
struct guest_state_request *gsr)
@ -1492,7 +1525,7 @@ static target_ulong map_and_getset_state(PowerPCCPU *cpu,
goto out1;
}
rc = getset_state(guest, vcpuid, gsr);
rc = getset_state(spapr, guest, vcpuid, gsr);
out1:
address_space_unmap(CPU(cpu)->as, gsr->gsb, len, is_write, len);
@ -1510,27 +1543,46 @@ static target_ulong h_guest_getset_state(PowerPCCPU *cpu,
target_ulong buf = args[3];
target_ulong buflen = args[4];
struct guest_state_request gsr;
SpaprMachineStateNestedGuest *guest;
SpaprMachineStateNestedGuest *guest = NULL;
guest = spapr_get_nested_guest(spapr, lpid);
if (!guest) {
return H_P2;
}
gsr.buf = buf;
assert(buflen <= GSB_MAX_BUF_SIZE);
gsr.len = buflen;
gsr.flags = 0;
if (flags & H_GUEST_GETSET_STATE_FLAG_GUEST_WIDE) {
/* Works for both get/set state */
if ((flags & H_GUEST_GET_STATE_FLAGS_GUEST_WIDE) ||
(flags & H_GUEST_SET_STATE_FLAGS_GUEST_WIDE)) {
gsr.flags |= GUEST_STATE_REQUEST_GUEST_WIDE;
}
if (flags & ~H_GUEST_GETSET_STATE_FLAG_GUEST_WIDE) {
return H_PARAMETER; /* flag not supported yet */
}
if (set) {
if (flags & ~H_GUEST_SET_STATE_FLAGS_MASK) {
return H_PARAMETER;
}
gsr.flags |= GUEST_STATE_REQUEST_SET;
} else {
/*
* No reserved fields to be set in flags nor both
* GUEST/HOST wide bits
*/
if ((flags & ~H_GUEST_GET_STATE_FLAGS_MASK) ||
(flags == H_GUEST_GET_STATE_FLAGS_MASK)) {
return H_PARAMETER;
}
if (flags & H_GUEST_GET_STATE_FLAGS_HOST_WIDE) {
gsr.flags |= GUEST_STATE_REQUEST_HOST_WIDE;
}
}
return map_and_getset_state(cpu, guest, vcpuid, &gsr);
if (!(gsr.flags & GUEST_STATE_REQUEST_HOST_WIDE)) {
guest = spapr_get_nested_guest(spapr, lpid);
if (!guest) {
return H_P2;
}
}
return map_and_getset_state(cpu, spapr, guest, vcpuid, &gsr);
}
static target_ulong h_guest_set_state(PowerPCCPU *cpu,
@ -1641,7 +1693,8 @@ static int get_exit_ids(uint64_t srr0, uint16_t ids[16])
return nr;
}
static void exit_process_output_buffer(PowerPCCPU *cpu,
static void exit_process_output_buffer(SpaprMachineState *spapr,
PowerPCCPU *cpu,
SpaprMachineStateNestedGuest *guest,
target_ulong vcpuid,
target_ulong *r3)
@ -1679,7 +1732,7 @@ static void exit_process_output_buffer(PowerPCCPU *cpu,
gsr.gsb = gsb;
gsr.len = VCPU_OUT_BUF_MIN_SZ;
gsr.flags = 0; /* get + never guest wide */
getset_state(guest, vcpuid, &gsr);
getset_state(spapr, guest, vcpuid, &gsr);
address_space_unmap(CPU(cpu)->as, gsb, len, true, len);
return;
@ -1705,7 +1758,7 @@ void spapr_exit_nested_papr(SpaprMachineState *spapr, PowerPCCPU *cpu, int excp)
exit_nested_store_l2(cpu, excp, vcpu);
/* do the output buffer for run_vcpu*/
exit_process_output_buffer(cpu, guest, vcpuid, &r3_return);
exit_process_output_buffer(spapr, cpu, guest, vcpuid, &r3_return);
assert(env->spr[SPR_LPIDR] != 0);
nested_load_state(cpu, spapr_cpu->nested_host_state);
@ -1820,7 +1873,7 @@ static target_ulong h_guest_run_vcpu(PowerPCCPU *cpu,
gsr.buf = vcpu->runbufin.addr;
gsr.len = vcpu->runbufin.size;
gsr.flags = GUEST_STATE_REQUEST_SET; /* Thread wide + writing */
rc = map_and_getset_state(cpu, guest, vcpuid, &gsr);
rc = map_and_getset_state(cpu, spapr, guest, vcpuid, &gsr);
if (rc == H_SUCCESS) {
nested_papr_run_vcpu(cpu, lpid, vcpu);
} else {

67
include/hw/ppc/spapr_nested.h
View file

@ -11,7 +11,13 @@
#define GSB_TB_OFFSET 0x0004 /* Timebase Offset */
#define GSB_PART_SCOPED_PAGETBL 0x0005 /* Partition Scoped Page Table */
#define GSB_PROCESS_TBL 0x0006 /* Process Table */
/* RESERVED 0x0007 - 0x0BFF */
/* RESERVED 0x0007 - 0x07FF */
#define GSB_L0_GUEST_HEAP_INUSE 0x0800 /* Guest Management Heap Size */
#define GSB_L0_GUEST_HEAP_MAX 0x0801 /* Guest Management Heap Max Size */
#define GSB_L0_GUEST_PGTABLE_SIZE_INUSE 0x0802 /* Guest Pagetable Size */
#define GSB_L0_GUEST_PGTABLE_SIZE_MAX 0x0803 /* Guest Pagetable Max Size */
#define GSB_L0_GUEST_PGTABLE_RECLAIMED 0x0804 /* Pagetable Reclaim in bytes */
/* RESERVED 0x0805 - 0xBFF */
#define GSB_VCPU_IN_BUFFER 0x0C00 /* Run VCPU Input Buffer */
#define GSB_VCPU_OUT_BUFFER 0x0C01 /* Run VCPU Out Buffer */
#define GSB_VCPU_VPA 0x0C02 /* HRA to Guest VCPU VPA */
@ -196,6 +202,38 @@ typedef struct SpaprMachineStateNested {
#define NESTED_API_PAPR 2
bool capabilities_set;
uint32_t pvr_base;
/**
* l0_guest_heap_inuse: The currently used bytes in the Hypervisor's Guest
* Management Space associated with the Host Partition.
**/
uint64_t l0_guest_heap_inuse;
/**
* host_heap_max: The maximum bytes available in the Hypervisor's Guest
* Management Space associated with the Host Partition.
**/
uint64_t l0_guest_heap_max;
/**
* host_pagetable: The currently used bytes in the Hypervisor's Guest
* Page Table Management Space associated with the Host Partition.
**/
uint64_t l0_guest_pgtable_size_inuse;
/**
* host_pagetable_max: The maximum bytes available in the Hypervisor's Guest
* Page Table Management Space associated with the Host Partition.
**/
uint64_t l0_guest_pgtable_size_max;
/**
* host_pagetable_reclaim: The amount of space in bytes that has been
* reclaimed due to overcommit in the Hypervisor's Guest Page Table
* Management Space associated with the Host Partition.
**/
uint64_t l0_guest_pgtable_reclaimed;
GHashTable *guests;
} SpaprMachineStateNested;
@ -229,9 +267,15 @@ typedef struct SpaprMachineStateNestedGuest {
#define HVMASK_HDEXCR 0x00000000FFFFFFFF
#define HVMASK_TB_OFFSET 0x000000FFFFFFFFFF
#define GSB_MAX_BUF_SIZE (1024 * 1024)
#define H_GUEST_GETSET_STATE_FLAG_GUEST_WIDE 0x8000000000000000
#define GUEST_STATE_REQUEST_GUEST_WIDE 0x1
#define GUEST_STATE_REQUEST_SET 0x2
#define H_GUEST_GET_STATE_FLAGS_MASK 0xC000000000000000ULL
#define H_GUEST_SET_STATE_FLAGS_MASK 0x8000000000000000ULL
#define H_GUEST_SET_STATE_FLAGS_GUEST_WIDE 0x8000000000000000ULL
#define H_GUEST_GET_STATE_FLAGS_GUEST_WIDE 0x8000000000000000ULL
#define H_GUEST_GET_STATE_FLAGS_HOST_WIDE 0x4000000000000000ULL
#define GUEST_STATE_REQUEST_GUEST_WIDE 0x1
#define GUEST_STATE_REQUEST_HOST_WIDE 0x2
#define GUEST_STATE_REQUEST_SET 0x4
/*
* As per ISA v3.1B, following bits are reserved:
@ -251,6 +295,15 @@ typedef struct SpaprMachineStateNestedGuest {
.copy = (c) \
}
#define GSBE_NESTED_MACHINE_DW(i, f) { \
.id = (i), \
.size = 8, \
.location = get_machine_ptr, \
.offset = offsetof(struct SpaprMachineStateNested, f), \
.copy = copy_state_8to8, \
.mask = HVMASK_DEFAULT \
}
#define GSBE_NESTED(i, sz, f, c) { \
.id = (i), \
.size = (sz), \
@ -509,9 +562,11 @@ struct guest_state_element_type {
uint16_t id;
int size;
#define GUEST_STATE_ELEMENT_TYPE_FLAG_GUEST_WIDE 0x1
#define GUEST_STATE_ELEMENT_TYPE_FLAG_READ_ONLY 0x2
#define GUEST_STATE_ELEMENT_TYPE_FLAG_HOST_WIDE 0x2
#define GUEST_STATE_ELEMENT_TYPE_FLAG_READ_ONLY 0x4
uint16_t flags;
void *(*location)(SpaprMachineStateNestedGuest *, target_ulong);
void *(*location)(struct SpaprMachineState *, SpaprMachineStateNestedGuest *,
target_ulong);
size_t offset;
void (*copy)(void *, void *, bool);
uint64_t mask;