openkylin
/
linux
mirror of https://gitee.com/openkylin/linux.git
9
0
Fork 0
Code Issues Projects Releases Wiki Activity
linux/arch/powerpc/perf/imc-pmu.c

1406 lines
35 KiB
C
Raw Blame History

/*
* In-Memory Collection (IMC) Performance Monitor counter support.
*
* Copyright (C) 2017 Madhavan Srinivasan, IBM Corporation.
* (C) 2017 Anju T Sudhakar, IBM Corporation.
* (C) 2017 Hemant K Shaw, IBM Corporation.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or later version.
*/
#include <linux/perf_event.h>
#include <linux/slab.h>
#include <asm/opal.h>
#include <asm/imc-pmu.h>
#include <asm/cputhreads.h>
#include <asm/smp.h>
#include <linux/string.h>
/* Nest IMC data structures and variables */
/*
* Used to avoid races in counting the nest-pmu units during hotplug
* register and unregister
*/
static DEFINE_MUTEX(nest_init_lock);
static DEFINE_PER_CPU(struct imc_pmu_ref *, local_nest_imc_refc);
static struct imc_pmu **per_nest_pmu_arr;
static cpumask_t nest_imc_cpumask;
static struct imc_pmu_ref *nest_imc_refc;
static int nest_pmus;
/* Core IMC data structures and variables */
static cpumask_t core_imc_cpumask;
static struct imc_pmu_ref *core_imc_refc;
static struct imc_pmu *core_imc_pmu;
/* Thread IMC data structures and variables */
static DEFINE_PER_CPU(u64 *, thread_imc_mem);
static struct imc_pmu *thread_imc_pmu;
static int thread_imc_mem_size;
static struct imc_pmu *imc_event_to_pmu(struct perf_event *event)
{
return container_of(event->pmu, struct imc_pmu, pmu);
}
PMU_FORMAT_ATTR(event, "config:0-40");
PMU_FORMAT_ATTR(offset, "config:0-31");
PMU_FORMAT_ATTR(rvalue, "config:32");
PMU_FORMAT_ATTR(mode, "config:33-40");
static struct attribute *imc_format_attrs[] = {
&format_attr_event.attr,
&format_attr_offset.attr,
&format_attr_rvalue.attr,
&format_attr_mode.attr,
NULL,
};
static struct attribute_group imc_format_group = {
.name = "format",
.attrs = imc_format_attrs,
};
/* Get the cpumask printed to a buffer "buf" */
static ssize_t imc_pmu_cpumask_get_attr(struct device *dev,
struct device_attribute *attr,
char *buf)
{
struct pmu *pmu = dev_get_drvdata(dev);
struct imc_pmu *imc_pmu = container_of(pmu, struct imc_pmu, pmu);
cpumask_t *active_mask;
switch(imc_pmu->domain){
case IMC_DOMAIN_NEST:
active_mask = &nest_imc_cpumask;
break;
case IMC_DOMAIN_CORE:
active_mask = &core_imc_cpumask;
break;
default:
return 0;
}
return cpumap_print_to_pagebuf(true, buf, active_mask);
}
static DEVICE_ATTR(cpumask, S_IRUGO, imc_pmu_cpumask_get_attr, NULL);
static struct attribute *imc_pmu_cpumask_attrs[] = {
&dev_attr_cpumask.attr,
NULL,
};
static struct attribute_group imc_pmu_cpumask_attr_group = {
.attrs = imc_pmu_cpumask_attrs,
};
/* device_str_attr_create : Populate event "name" and string "str" in attribute */
static struct attribute *device_str_attr_create(const char *name, const char *str)
{
struct perf_pmu_events_attr *attr;
attr = kzalloc(sizeof(*attr), GFP_KERNEL);
if (!attr)
return NULL;
sysfs_attr_init(&attr->attr.attr);
attr->event_str = str;
attr->attr.attr.name = name;
attr->attr.attr.mode = 0444;
attr->attr.show = perf_event_sysfs_show;
return &attr->attr.attr;
}
static int imc_parse_event(struct device_node *np, const char *scale,
const char *unit, const char *prefix,
u32 base, struct imc_events *event)
{
const char *s;
u32 reg;
if (of_property_read_u32(np, "reg", &reg))
goto error;
/* Add the base_reg value to the "reg" */
event->value = base + reg;
if (of_property_read_string(np, "event-name", &s))
goto error;
event->name = kasprintf(GFP_KERNEL, "%s%s", prefix, s);
if (!event->name)
goto error;
if (of_property_read_string(np, "scale", &s))
s = scale;
if (s) {
event->scale = kstrdup(s, GFP_KERNEL);
if (!event->scale)
goto error;
}
if (of_property_read_string(np, "unit", &s))
s = unit;
if (s) {
event->unit = kstrdup(s, GFP_KERNEL);
if (!event->unit)
goto error;
}
return 0;
error:
kfree(event->unit);
kfree(event->scale);
kfree(event->name);
return -EINVAL;
}
/*
* imc_free_events: Function to cleanup the events list, having
* "nr_entries".
*/
static void imc_free_events(struct imc_events *events, int nr_entries)
{
int i;
/* Nothing to clean, return */
if (!events)
return;
for (i = 0; i < nr_entries; i++) {
kfree(events[i].unit);
kfree(events[i].scale);
kfree(events[i].name);
}
kfree(events);
}
/*
* update_events_in_group: Update the "events" information in an attr_group
* and assign the attr_group to the pmu "pmu".
*/
static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu)
{
struct attribute_group *attr_group;
struct attribute **attrs, *dev_str;
struct device_node *np, *pmu_events;
u32 handle, base_reg;
int i = 0, j = 0, ct, ret;
const char *prefix, *g_scale, *g_unit;
const char *ev_val_str, *ev_scale_str, *ev_unit_str;
if (!of_property_read_u32(node, "events", &handle))
pmu_events = of_find_node_by_phandle(handle);
else
return 0;
/* Did not find any node with a given phandle */
if (!pmu_events)
return 0;
/* Get a count of number of child nodes */
ct = of_get_child_count(pmu_events);
/* Get the event prefix */
if (of_property_read_string(node, "events-prefix", &prefix))
return 0;
/* Get a global unit and scale data if available */
if (of_property_read_string(node, "scale", &g_scale))
g_scale = NULL;
if (of_property_read_string(node, "unit", &g_unit))
g_unit = NULL;
/* "reg" property gives out the base offset of the counters data */
of_property_read_u32(node, "reg", &base_reg);
/* Allocate memory for the events */
pmu->events = kcalloc(ct, sizeof(struct imc_events), GFP_KERNEL);
if (!pmu->events)
return -ENOMEM;
ct = 0;
/* Parse the events and update the struct */
for_each_child_of_node(pmu_events, np) {
ret = imc_parse_event(np, g_scale, g_unit, prefix, base_reg, &pmu->events[ct]);
if (!ret)
ct++;
}
/* Allocate memory for attribute group */
attr_group = kzalloc(sizeof(*attr_group), GFP_KERNEL);
if (!attr_group) {
imc_free_events(pmu->events, ct);
return -ENOMEM;
}
/*
* Allocate memory for attributes.
* Since we have count of events for this pmu, we also allocate
* memory for the scale and unit attribute for now.
* "ct" has the total event structs added from the events-parent node.
* So allocate three times the "ct" (this includes event, event_scale and
* event_unit).
*/
attrs = kcalloc(((ct * 3) + 1), sizeof(struct attribute *), GFP_KERNEL);
if (!attrs) {
kfree(attr_group);
imc_free_events(pmu->events, ct);
return -ENOMEM;
}
attr_group->name = "events";
attr_group->attrs = attrs;
do {
ev_val_str = kasprintf(GFP_KERNEL, "event=0x%x", pmu->events[i].value);
dev_str = device_str_attr_create(pmu->events[i].name, ev_val_str);
if (!dev_str)
continue;
attrs[j++] = dev_str;
if (pmu->events[i].scale) {
ev_scale_str = kasprintf(GFP_KERNEL, "%s.scale", pmu->events[i].name);
dev_str = device_str_attr_create(ev_scale_str, pmu->events[i].scale);
if (!dev_str)
continue;
attrs[j++] = dev_str;
}
if (pmu->events[i].unit) {
ev_unit_str = kasprintf(GFP_KERNEL, "%s.unit", pmu->events[i].name);
dev_str = device_str_attr_create(ev_unit_str, pmu->events[i].unit);
if (!dev_str)
continue;
attrs[j++] = dev_str;
}
} while (++i < ct);
/* Save the event attribute */
pmu->attr_groups[IMC_EVENT_ATTR] = attr_group;
return 0;
}
/* get_nest_pmu_ref: Return the imc_pmu_ref struct for the given node */
static struct imc_pmu_ref *get_nest_pmu_ref(int cpu)
{
return per_cpu(local_nest_imc_refc, cpu);
}
static void nest_change_cpu_context(int old_cpu, int new_cpu)
{
struct imc_pmu **pn = per_nest_pmu_arr;
if (old_cpu < 0 || new_cpu < 0)
return;
while (*pn) {
perf_pmu_migrate_context(&(*pn)->pmu, old_cpu, new_cpu);
pn++;
}
}
static int ppc_nest_imc_cpu_offline(unsigned int cpu)
{
int nid, target = -1;
const struct cpumask *l_cpumask;
struct imc_pmu_ref *ref;
/*
* Check in the designated list for this cpu. Dont bother
* if not one of them.
*/
if (!cpumask_test_and_clear_cpu(cpu, &nest_imc_cpumask))
return 0;
/*
* Check whether nest_imc is registered. We could end up here if the
* cpuhotplug callback registration fails. i.e, callback invokes the
* offline path for all successfully registered nodes. At this stage,
* nest_imc pmu will not be registered and we should return here.
*
* We return with a zero since this is not an offline failure. And
* cpuhp_setup_state() returns the actual failure reason to the caller,
* which in turn will call the cleanup routine.
*/
if (!nest_pmus)
return 0;
/*
* Now that this cpu is one of the designated,
* find a next cpu a) which is online and b) in same chip.
*/
nid = cpu_to_node(cpu);
l_cpumask = cpumask_of_node(nid);
target = cpumask_any_but(l_cpumask, cpu);
/*
* Update the cpumask with the target cpu and
* migrate the context if needed
*/
if (target >= 0 && target < nr_cpu_ids) {
cpumask_set_cpu(target, &nest_imc_cpumask);
nest_change_cpu_context(cpu, target);
} else {
opal_imc_counters_stop(OPAL_IMC_COUNTERS_NEST,
get_hard_smp_processor_id(cpu));
/*
* If this is the last cpu in this chip then, skip the reference
* count mutex lock and make the reference count on this chip zero.
*/
ref = get_nest_pmu_ref(cpu);
if (!ref)
return -EINVAL;
ref->refc = 0;
}
return 0;
}
static int ppc_nest_imc_cpu_online(unsigned int cpu)
{
const struct cpumask *l_cpumask;
static struct cpumask tmp_mask;
int res;
/* Get the cpumask of this node */
l_cpumask = cpumask_of_node(cpu_to_node(cpu));
/*
* If this is not the first online CPU on this node, then
* just return.
*/
if (cpumask_and(&tmp_mask, l_cpumask, &nest_imc_cpumask))
return 0;
/*
* If this is the first online cpu on this node
* disable the nest counters by making an OPAL call.
*/
res = opal_imc_counters_stop(OPAL_IMC_COUNTERS_NEST,
get_hard_smp_processor_id(cpu));
if (res)
return res;
/* Make this CPU the designated target for counter collection */
cpumask_set_cpu(cpu, &nest_imc_cpumask);
return 0;
}
static int nest_pmu_cpumask_init(void)
{
return cpuhp_setup_state(CPUHP_AP_PERF_POWERPC_NEST_IMC_ONLINE,
"perf/powerpc/imc:online",
ppc_nest_imc_cpu_online,
ppc_nest_imc_cpu_offline);
}
static void nest_imc_counters_release(struct perf_event *event)
{
int rc, node_id;
struct imc_pmu_ref *ref;
if (event->cpu < 0)
return;
node_id = cpu_to_node(event->cpu);
/*
* See if we need to disable the nest PMU.
* If no events are currently in use, then we have to take a
* mutex to ensure that we don't race with another task doing
* enable or disable the nest counters.
*/
ref = get_nest_pmu_ref(event->cpu);
if (!ref)
return;
/* Take the mutex lock for this node and then decrement the reference count */
mutex_lock(&ref->lock);
if (ref->refc == 0) {
/*
* The scenario where this is true is, when perf session is
* started, followed by offlining of all cpus in a given node.
*
* In the cpuhotplug offline path, ppc_nest_imc_cpu_offline()
* function set the ref->count to zero, if the cpu which is
* about to offline is the last cpu in a given node and make
* an OPAL call to disable the engine in that node.
*
*/
mutex_unlock(&ref->lock);
return;
}
ref->refc--;
if (ref->refc == 0) {
rc = opal_imc_counters_stop(OPAL_IMC_COUNTERS_NEST,
get_hard_smp_processor_id(event->cpu));
if (rc) {
mutex_unlock(&ref->lock);
pr_err("nest-imc: Unable to stop the counters for core %d\n", node_id);
return;
}
} else if (ref->refc < 0) {
WARN(1, "nest-imc: Invalid event reference count\n");
ref->refc = 0;
}
mutex_unlock(&ref->lock);
}
static int nest_imc_event_init(struct perf_event *event)
{
int chip_id, rc, node_id;
u32 l_config, config = event->attr.config;
struct imc_mem_info *pcni;
struct imc_pmu *pmu;
struct imc_pmu_ref *ref;
bool flag = false;
if (event->attr.type != event->pmu->type)
return -ENOENT;
/* Sampling not supported */
if (event->hw.sample_period)
return -EINVAL;
/* unsupported modes and filters */
if (event->attr.exclude_user ||
event->attr.exclude_kernel ||
event->attr.exclude_hv ||
event->attr.exclude_idle ||
event->attr.exclude_host ||
event->attr.exclude_guest)
return -EINVAL;
if (event->cpu < 0)
return -EINVAL;
pmu = imc_event_to_pmu(event);
/* Sanity check for config (event offset) */
if ((config & IMC_EVENT_OFFSET_MASK) > pmu->counter_mem_size)
return -EINVAL;
/*
* Nest HW counter memory resides in a per-chip reserve-memory (HOMER).
* Get the base memory addresss for this cpu.
*/
chip_id = cpu_to_chip_id(event->cpu);
pcni = pmu->mem_info;
do {
if (pcni->id == chip_id) {
flag = true;
break;
}
pcni++;
} while (pcni);
if (!flag)
return -ENODEV;
/*
* Add the event offset to the base address.
*/
l_config = config & IMC_EVENT_OFFSET_MASK;
event->hw.event_base = (u64)pcni->vbase + l_config;
node_id = cpu_to_node(event->cpu);
/*
* Get the imc_pmu_ref struct for this node.
* Take the mutex lock and then increment the count of nest pmu events
* inited.
*/
ref = get_nest_pmu_ref(event->cpu);
if (!ref)
return -EINVAL;
mutex_lock(&ref->lock);
if (ref->refc == 0) {
rc = opal_imc_counters_start(OPAL_IMC_COUNTERS_NEST,
get_hard_smp_processor_id(event->cpu));
if (rc) {
mutex_unlock(&ref->lock);
pr_err("nest-imc: Unable to start the counters for node %d\n",
node_id);
return rc;
}
}
++ref->refc;
mutex_unlock(&ref->lock);
event->destroy = nest_imc_counters_release;
return 0;
}
/*
* core_imc_mem_init : Initializes memory for the current core.
*
* Uses alloc_pages_node() and uses the returned address as an argument to
* an opal call to configure the pdbar. The address sent as an argument is
* converted to physical address before the opal call is made. This is the
* base address at which the core imc counters are populated.
*/
static int core_imc_mem_init(int cpu, int size)
{
int nid, rc = 0, core_id = (cpu / threads_per_core);
struct imc_mem_info *mem_info;
/*
* alloc_pages_node() will allocate memory for core in the
* local node only.
*/
nid = cpu_to_node(cpu);
mem_info = &core_imc_pmu->mem_info[core_id];
mem_info->id = core_id;
/* We need only vbase for core counters */
mem_info->vbase = page_address(alloc_pages_node(nid,
GFP_KERNEL | __GFP_ZERO | __GFP_THISNODE |
__GFP_NOWARN, get_order(size)));
if (!mem_info->vbase)
return -ENOMEM;
/* Init the mutex */
core_imc_refc[core_id].id = core_id;
mutex_init(&core_imc_refc[core_id].lock);
rc = opal_imc_counters_init(OPAL_IMC_COUNTERS_CORE,
__pa((void *)mem_info->vbase),
get_hard_smp_processor_id(cpu));
if (rc) {
free_pages((u64)mem_info->vbase, get_order(size));
mem_info->vbase = NULL;
}
return rc;
}
static bool is_core_imc_mem_inited(int cpu)
{
struct imc_mem_info *mem_info;
int core_id = (cpu / threads_per_core);
mem_info = &core_imc_pmu->mem_info[core_id];
if (!mem_info->vbase)
return false;
return true;
}
static int ppc_core_imc_cpu_online(unsigned int cpu)
{
const struct cpumask *l_cpumask;
static struct cpumask tmp_mask;
int ret = 0;
/* Get the cpumask for this core */
l_cpumask = cpu_sibling_mask(cpu);
/* If a cpu for this core is already set, then, don't do anything */
if (cpumask_and(&tmp_mask, l_cpumask, &core_imc_cpumask))
return 0;
if (!is_core_imc_mem_inited(cpu)) {
ret = core_imc_mem_init(cpu, core_imc_pmu->counter_mem_size);
if (ret) {
pr_info("core_imc memory allocation for cpu %d failed\n", cpu);
return ret;
}
}
/* set the cpu in the mask */
cpumask_set_cpu(cpu, &core_imc_cpumask);
return 0;
}
static int ppc_core_imc_cpu_offline(unsigned int cpu)
{
unsigned int core_id;
int ncpu;
struct imc_pmu_ref *ref;
/*
* clear this cpu out of the mask, if not present in the mask,
* don't bother doing anything.
*/
if (!cpumask_test_and_clear_cpu(cpu, &core_imc_cpumask))
return 0;
/*
* Check whether core_imc is registered. We could end up here
* if the cpuhotplug callback registration fails. i.e, callback
* invokes the offline path for all sucessfully registered cpus.
* At this stage, core_imc pmu will not be registered and we
* should return here.
*
* We return with a zero since this is not an offline failure.
* And cpuhp_setup_state() returns the actual failure reason
* to the caller, which inturn will call the cleanup routine.
*/
if (!core_imc_pmu->pmu.event_init)
return 0;
/* Find any online cpu in that core except the current "cpu" */
ncpu = cpumask_any_but(cpu_sibling_mask(cpu), cpu);
if (ncpu >= 0 && ncpu < nr_cpu_ids) {
cpumask_set_cpu(ncpu, &core_imc_cpumask);
perf_pmu_migrate_context(&core_imc_pmu->pmu, cpu, ncpu);
} else {
/*
* If this is the last cpu in this core then, skip taking refernce
* count mutex lock for this core and directly zero "refc" for
* this core.
*/
opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE,
get_hard_smp_processor_id(cpu));
core_id = cpu / threads_per_core;
ref = &core_imc_refc[core_id];
if (!ref)
return -EINVAL;
ref->refc = 0;
}
return 0;
}
static int core_imc_pmu_cpumask_init(void)
{
return cpuhp_setup_state(CPUHP_AP_PERF_POWERPC_CORE_IMC_ONLINE,
"perf/powerpc/imc_core:online",
ppc_core_imc_cpu_online,
ppc_core_imc_cpu_offline);
}
static void core_imc_counters_release(struct perf_event *event)
{
int rc, core_id;
struct imc_pmu_ref *ref;
if (event->cpu < 0)
return;
/*
* See if we need to disable the IMC PMU.
* If no events are currently in use, then we have to take a
* mutex to ensure that we don't race with another task doing
* enable or disable the core counters.
*/
core_id = event->cpu / threads_per_core;
/* Take the mutex lock and decrement the refernce count for this core */
ref = &core_imc_refc[core_id];
if (!ref)
return;
mutex_lock(&ref->lock);
if (ref->refc == 0) {
/*
* The scenario where this is true is, when perf session is
* started, followed by offlining of all cpus in a given core.
*
* In the cpuhotplug offline path, ppc_core_imc_cpu_offline()
* function set the ref->count to zero, if the cpu which is
* about to offline is the last cpu in a given core and make
* an OPAL call to disable the engine in that core.
*
*/
mutex_unlock(&ref->lock);
return;
}
ref->refc--;
if (ref->refc == 0) {
rc = opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE,
get_hard_smp_processor_id(event->cpu));
if (rc) {
mutex_unlock(&ref->lock);
pr_err("IMC: Unable to stop the counters for core %d\n", core_id);
return;
}
} else if (ref->refc < 0) {
WARN(1, "core-imc: Invalid event reference count\n");
ref->refc = 0;
}
mutex_unlock(&ref->lock);
}
static int core_imc_event_init(struct perf_event *event)
{
int core_id, rc;
u64 config = event->attr.config;
struct imc_mem_info *pcmi;
struct imc_pmu *pmu;
struct imc_pmu_ref *ref;
if (event->attr.type != event->pmu->type)
return -ENOENT;
/* Sampling not supported */
if (event->hw.sample_period)
return -EINVAL;
/* unsupported modes and filters */
if (event->attr.exclude_user ||
event->attr.exclude_kernel ||
event->attr.exclude_hv ||
event->attr.exclude_idle ||
event->attr.exclude_host ||
event->attr.exclude_guest)
return -EINVAL;
if (event->cpu < 0)
return -EINVAL;
event->hw.idx = -1;
pmu = imc_event_to_pmu(event);
/* Sanity check for config (event offset) */
if (((config & IMC_EVENT_OFFSET_MASK) > pmu->counter_mem_size))
return -EINVAL;
if (!is_core_imc_mem_inited(event->cpu))
return -ENODEV;
core_id = event->cpu / threads_per_core;
pcmi = &core_imc_pmu->mem_info[core_id];
if ((!pcmi->vbase))
return -ENODEV;
/* Get the core_imc mutex for this core */
ref = &core_imc_refc[core_id];
if (!ref)
return -EINVAL;
/*
* Core pmu units are enabled only when it is used.
* See if this is triggered for the first time.
* If yes, take the mutex lock and enable the core counters.
* If not, just increment the count in core_imc_refc struct.
*/
mutex_lock(&ref->lock);
if (ref->refc == 0) {
rc = opal_imc_counters_start(OPAL_IMC_COUNTERS_CORE,
get_hard_smp_processor_id(event->cpu));
if (rc) {
mutex_unlock(&ref->lock);
pr_err("core-imc: Unable to start the counters for core %d\n",
core_id);
return rc;
}
}
++ref->refc;
mutex_unlock(&ref->lock);
event->hw.event_base = (u64)pcmi->vbase + (config & IMC_EVENT_OFFSET_MASK);
event->destroy = core_imc_counters_release;
return 0;
}
/*
* Allocates a page of memory for each of the online cpus, and write the
* physical base address of that page to the LDBAR for that cpu.
*
* LDBAR Register Layout:
*
* 0 4 8 12 16 20 24 28
* | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - |
* | | [ ] [ Counter Address [8:50]
* | * Mode |
* | * PB Scope
* * Enable/Disable
*
* 32 36 40 44 48 52 56 60
* | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - |
* Counter Address [8:50] ]
*
*/
static int thread_imc_mem_alloc(int cpu_id, int size)
{
u64 ldbar_value, *local_mem = per_cpu(thread_imc_mem, cpu_id);
int nid = cpu_to_node(cpu_id);
if (!local_mem) {
/*
* This case could happen only once at start, since we dont
* free the memory in cpu offline path.
*/
local_mem = page_address(alloc_pages_node(nid,
GFP_KERNEL | __GFP_ZERO | __GFP_THISNODE |
__GFP_NOWARN, get_order(size)));
if (!local_mem)
return -ENOMEM;
per_cpu(thread_imc_mem, cpu_id) = local_mem;
}
ldbar_value = ((u64)local_mem & THREAD_IMC_LDBAR_MASK) | THREAD_IMC_ENABLE;
mtspr(SPRN_LDBAR, ldbar_value);
return 0;
}
static int ppc_thread_imc_cpu_online(unsigned int cpu)
{
return thread_imc_mem_alloc(cpu, thread_imc_mem_size);
}
static int ppc_thread_imc_cpu_offline(unsigned int cpu)
{
mtspr(SPRN_LDBAR, 0);
return 0;
}
static int thread_imc_cpu_init(void)
{
return cpuhp_setup_state(CPUHP_AP_PERF_POWERPC_THREAD_IMC_ONLINE,
"perf/powerpc/imc_thread:online",
ppc_thread_imc_cpu_online,
ppc_thread_imc_cpu_offline);
}
static int thread_imc_event_init(struct perf_event *event)
{
u32 config = event->attr.config;
struct task_struct *target;
struct imc_pmu *pmu;
if (event->attr.type != event->pmu->type)
return -ENOENT;
/* Sampling not supported */
if (event->hw.sample_period)
return -EINVAL;
event->hw.idx = -1;
pmu = imc_event_to_pmu(event);
/* Sanity check for config offset */
if (((config & IMC_EVENT_OFFSET_MASK) > pmu->counter_mem_size))
return -EINVAL;
target = event->hw.target;
if (!target)
return -EINVAL;
event->pmu->task_ctx_nr = perf_sw_context;
return 0;
}
static bool is_thread_imc_pmu(struct perf_event *event)
{
if (!strncmp(event->pmu->name, "thread_imc", strlen("thread_imc")))
return true;
return false;
}
static u64 * get_event_base_addr(struct perf_event *event)
{
u64 addr;
if (is_thread_imc_pmu(event)) {
addr = (u64)per_cpu(thread_imc_mem, smp_processor_id());
return (u64 *)(addr + (event->attr.config & IMC_EVENT_OFFSET_MASK));
}
return (u64 *)event->hw.event_base;
}
static void thread_imc_pmu_start_txn(struct pmu *pmu,
unsigned int txn_flags)
{
if (txn_flags & ~PERF_PMU_TXN_ADD)
return;
perf_pmu_disable(pmu);
}
static void thread_imc_pmu_cancel_txn(struct pmu *pmu)
{
perf_pmu_enable(pmu);
}
static int thread_imc_pmu_commit_txn(struct pmu *pmu)
{
perf_pmu_enable(pmu);
return 0;
}
static u64 imc_read_counter(struct perf_event *event)
{
u64 *addr, data;
/*
* In-Memory Collection (IMC) counters are free flowing counters.
* So we take a snapshot of the counter value on enable and save it
* to calculate the delta at later stage to present the event counter
* value.
*/
addr = get_event_base_addr(event);
data = be64_to_cpu(READ_ONCE(*addr));
local64_set(&event->hw.prev_count, data);
return data;
}
static void imc_event_update(struct perf_event *event)
{
u64 counter_prev, counter_new, final_count;
counter_prev = local64_read(&event->hw.prev_count);
counter_new = imc_read_counter(event);
final_count = counter_new - counter_prev;
/* Update the delta to the event count */
local64_add(final_count, &event->count);
}
static void imc_event_start(struct perf_event *event, int flags)
{
/*
* In Memory Counters are free flowing counters. HW or the microcode
* keeps adding to the counter offset in memory. To get event
* counter value, we snapshot the value here and we calculate
* delta at later point.
*/
imc_read_counter(event);
}
static void imc_event_stop(struct perf_event *event, int flags)
{
/*
* Take a snapshot and calculate the delta and update
* the event counter values.
*/
imc_event_update(event);
}
static int imc_event_add(struct perf_event *event, int flags)
{
if (flags & PERF_EF_START)
imc_event_start(event, flags);
return 0;
}
static int thread_imc_event_add(struct perf_event *event, int flags)
{
int core_id;
struct imc_pmu_ref *ref;
if (flags & PERF_EF_START)
imc_event_start(event, flags);
if (!is_core_imc_mem_inited(smp_processor_id()))
return -EINVAL;
core_id = smp_processor_id() / threads_per_core;
/*
* imc pmus are enabled only when it is used.
* See if this is triggered for the first time.
* If yes, take the mutex lock and enable the counters.
* If not, just increment the count in ref count struct.
*/
ref = &core_imc_refc[core_id];
if (!ref)
return -EINVAL;
mutex_lock(&ref->lock);
if (ref->refc == 0) {
if (opal_imc_counters_start(OPAL_IMC_COUNTERS_CORE,
get_hard_smp_processor_id(smp_processor_id()))) {
mutex_unlock(&ref->lock);
pr_err("thread-imc: Unable to start the counter\
for core %d\n", core_id);
return -EINVAL;
}
}
++ref->refc;
mutex_unlock(&ref->lock);
return 0;
}
static void thread_imc_event_del(struct perf_event *event, int flags)
{
int core_id;
struct imc_pmu_ref *ref;
/*
* Take a snapshot and calculate the delta and update
* the event counter values.
*/
imc_event_update(event);
core_id = smp_processor_id() / threads_per_core;
ref = &core_imc_refc[core_id];
mutex_lock(&ref->lock);
ref->refc--;
if (ref->refc == 0) {
if (opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE,
get_hard_smp_processor_id(smp_processor_id()))) {
mutex_unlock(&ref->lock);
pr_err("thread-imc: Unable to stop the counters\
for core %d\n", core_id);
return;
}
} else if (ref->refc < 0) {
ref->refc = 0;
}
mutex_unlock(&ref->lock);
}
/* update_pmu_ops : Populate the appropriate operations for "pmu" */
static int update_pmu_ops(struct imc_pmu *pmu)
{
pmu->pmu.task_ctx_nr = perf_invalid_context;
pmu->pmu.add = imc_event_add;
pmu->pmu.del = imc_event_stop;
pmu->pmu.start = imc_event_start;
pmu->pmu.stop = imc_event_stop;
pmu->pmu.read = imc_event_update;
pmu->pmu.attr_groups = pmu->attr_groups;
pmu->attr_groups[IMC_FORMAT_ATTR] = &imc_format_group;
switch (pmu->domain) {
case IMC_DOMAIN_NEST:
pmu->pmu.event_init = nest_imc_event_init;
pmu->attr_groups[IMC_CPUMASK_ATTR] = &imc_pmu_cpumask_attr_group;
break;
case IMC_DOMAIN_CORE:
pmu->pmu.event_init = core_imc_event_init;
pmu->attr_groups[IMC_CPUMASK_ATTR] = &imc_pmu_cpumask_attr_group;
break;
case IMC_DOMAIN_THREAD:
pmu->pmu.event_init = thread_imc_event_init;
pmu->pmu.add = thread_imc_event_add;
pmu->pmu.del = thread_imc_event_del;
pmu->pmu.start_txn = thread_imc_pmu_start_txn;
pmu->pmu.cancel_txn = thread_imc_pmu_cancel_txn;
pmu->pmu.commit_txn = thread_imc_pmu_commit_txn;
break;
default:
break;
}
return 0;
}
/* init_nest_pmu_ref: Initialize the imc_pmu_ref struct for all the nodes */
static int init_nest_pmu_ref(void)
{
int nid, i, cpu;
nest_imc_refc = kcalloc(num_possible_nodes(), sizeof(*nest_imc_refc),
GFP_KERNEL);
if (!nest_imc_refc)
return -ENOMEM;
i = 0;
for_each_node(nid) {
/*
* Mutex lock to avoid races while tracking the number of
* sessions using the chip's nest pmu units.
*/
mutex_init(&nest_imc_refc[i].lock);
/*
* Loop to init the "id" with the node_id. Variable "i" initialized to
* 0 and will be used as index to the array. "i" will not go off the
* end of the array since the "for_each_node" loops for "N_POSSIBLE"
* nodes only.
*/
nest_imc_refc[i++].id = nid;
}
/*
* Loop to init the per_cpu "local_nest_imc_refc" with the proper
* "nest_imc_refc" index. This makes get_nest_pmu_ref() alot simple.
*/
for_each_possible_cpu(cpu) {
nid = cpu_to_node(cpu);
for (i = 0; i < num_possible_nodes(); i++) {
if (nest_imc_refc[i].id == nid) {
per_cpu(local_nest_imc_refc, cpu) = &nest_imc_refc[i];
break;
}
}
}
return 0;
}
static void cleanup_all_core_imc_memory(void)
{
int i, nr_cores = DIV_ROUND_UP(num_possible_cpus(), threads_per_core);
struct imc_mem_info *ptr = core_imc_pmu->mem_info;
int size = core_imc_pmu->counter_mem_size;
/* mem_info will never be NULL */
for (i = 0; i < nr_cores; i++) {
if (ptr[i].vbase)
free_pages((u64)ptr[i].vbase, get_order(size));
}
kfree(ptr);
kfree(core_imc_refc);
}
static void thread_imc_ldbar_disable(void *dummy)
{
/*
* By Zeroing LDBAR, we disable thread-imc
* updates.
*/
mtspr(SPRN_LDBAR, 0);
}
void thread_imc_disable(void)
{
on_each_cpu(thread_imc_ldbar_disable, NULL, 1);
}
static void cleanup_all_thread_imc_memory(void)
{
int i, order = get_order(thread_imc_mem_size);
for_each_online_cpu(i) {
if (per_cpu(thread_imc_mem, i))
free_pages((u64)per_cpu(thread_imc_mem, i), order);
}
}
/* Function to free the attr_groups which are dynamically allocated */
static void imc_common_mem_free(struct imc_pmu *pmu_ptr)
{
if (pmu_ptr->attr_groups[IMC_EVENT_ATTR])
kfree(pmu_ptr->attr_groups[IMC_EVENT_ATTR]->attrs);
kfree(pmu_ptr->attr_groups[IMC_EVENT_ATTR]);
}
/*
* Common function to unregister cpu hotplug callback and
* free the memory.
* TODO: Need to handle pmu unregistering, which will be
* done in followup series.
*/
static void imc_common_cpuhp_mem_free(struct imc_pmu *pmu_ptr)
{
if (pmu_ptr->domain == IMC_DOMAIN_NEST) {
mutex_lock(&nest_init_lock);
if (nest_pmus == 1) {
cpuhp_remove_state(CPUHP_AP_PERF_POWERPC_NEST_IMC_ONLINE);
kfree(nest_imc_refc);
kfree(per_nest_pmu_arr);
per_nest_pmu_arr = NULL;
}
if (nest_pmus > 0)
nest_pmus--;
mutex_unlock(&nest_init_lock);
}
/* Free core_imc memory */
if (pmu_ptr->domain == IMC_DOMAIN_CORE) {
cpuhp_remove_state(CPUHP_AP_PERF_POWERPC_CORE_IMC_ONLINE);
cleanup_all_core_imc_memory();
}
/* Free thread_imc memory */
if (pmu_ptr->domain == IMC_DOMAIN_THREAD) {
cpuhp_remove_state(CPUHP_AP_PERF_POWERPC_THREAD_IMC_ONLINE);
cleanup_all_thread_imc_memory();
}
}
/*
* Function to unregister thread-imc if core-imc
* is not registered.
*/
void unregister_thread_imc(void)
{
imc_common_cpuhp_mem_free(thread_imc_pmu);
imc_common_mem_free(thread_imc_pmu);
perf_pmu_unregister(&thread_imc_pmu->pmu);
}
/*
* imc_mem_init : Function to support memory allocation for core imc.
*/
static int imc_mem_init(struct imc_pmu *pmu_ptr, struct device_node *parent,
int pmu_index)
{
const char *s;
int nr_cores, cpu, res = -ENOMEM;
if (of_property_read_string(parent, "name", &s))
return -ENODEV;
switch (pmu_ptr->domain) {
case IMC_DOMAIN_NEST:
/* Update the pmu name */
pmu_ptr->pmu.name = kasprintf(GFP_KERNEL, "%s%s_imc", "nest_", s);
if (!pmu_ptr->pmu.name)
goto err;
/* Needed for hotplug/migration */
if (!per_nest_pmu_arr) {
per_nest_pmu_arr = kcalloc(get_max_nest_dev() + 1,
sizeof(struct imc_pmu *),
GFP_KERNEL);
if (!per_nest_pmu_arr)
goto err;
}
per_nest_pmu_arr[pmu_index] = pmu_ptr;
break;
case IMC_DOMAIN_CORE:
/* Update the pmu name */
pmu_ptr->pmu.name = kasprintf(GFP_KERNEL, "%s%s", s, "_imc");
if (!pmu_ptr->pmu.name)
goto err;
nr_cores = DIV_ROUND_UP(num_possible_cpus(), threads_per_core);
pmu_ptr->mem_info = kcalloc(nr_cores, sizeof(struct imc_mem_info),
GFP_KERNEL);
if (!pmu_ptr->mem_info)
goto err;
core_imc_refc = kcalloc(nr_cores, sizeof(struct imc_pmu_ref),
GFP_KERNEL);
if (!core_imc_refc) {
kfree(pmu_ptr->mem_info);
goto err;
}
core_imc_pmu = pmu_ptr;
break;
case IMC_DOMAIN_THREAD:
/* Update the pmu name */
pmu_ptr->pmu.name = kasprintf(GFP_KERNEL, "%s%s", s, "_imc");
if (!pmu_ptr->pmu.name)
goto err;
thread_imc_mem_size = pmu_ptr->counter_mem_size;
for_each_online_cpu(cpu) {
res = thread_imc_mem_alloc(cpu, pmu_ptr->counter_mem_size);
if (res) {
cleanup_all_thread_imc_memory();
goto err;
}
}
thread_imc_pmu = pmu_ptr;
break;
default:
return -EINVAL;
}
return 0;
err:
return res;
}
/*
* init_imc_pmu : Setup and register the IMC pmu device.
*
* @parent: Device tree unit node
* @pmu_ptr: memory allocated for this pmu
* @pmu_idx: Count of nest pmc registered
*
* init_imc_pmu() setup pmu cpumask and registers for a cpu hotplug callback.
* Handles failure cases and accordingly frees memory.
*/
int init_imc_pmu(struct device_node *parent, struct imc_pmu *pmu_ptr, int pmu_idx)
{
int ret;
ret = imc_mem_init(pmu_ptr, parent, pmu_idx);
if (ret)
goto err_free_mem;
switch (pmu_ptr->domain) {
case IMC_DOMAIN_NEST:
/*
* Nest imc pmu need only one cpu per chip, we initialize the
* cpumask for the first nest imc pmu and use the same for the
* rest. To handle the cpuhotplug callback unregister, we track
* the number of nest pmus in "nest_pmus".
*/
mutex_lock(&nest_init_lock);
if (nest_pmus == 0) {
ret = init_nest_pmu_ref();
if (ret) {
mutex_unlock(&nest_init_lock);
kfree(per_nest_pmu_arr);
per_nest_pmu_arr = NULL;
goto err_free_mem;
}
/* Register for cpu hotplug notification. */
ret = nest_pmu_cpumask_init();
if (ret) {
mutex_unlock(&nest_init_lock);
kfree(nest_imc_refc);
kfree(per_nest_pmu_arr);
per_nest_pmu_arr = NULL;
goto err_free_mem;
}
}
nest_pmus++;
mutex_unlock(&nest_init_lock);
break;
case IMC_DOMAIN_CORE:
ret = core_imc_pmu_cpumask_init();
if (ret) {
cleanup_all_core_imc_memory();
goto err_free_mem;
}
break;
case IMC_DOMAIN_THREAD:
ret = thread_imc_cpu_init();
if (ret) {
cleanup_all_thread_imc_memory();
goto err_free_mem;
}
break;
default:
return -EINVAL; /* Unknown domain */
}
ret = update_events_in_group(parent, pmu_ptr);
if (ret)
goto err_free_cpuhp_mem;
ret = update_pmu_ops(pmu_ptr);
if (ret)
goto err_free_cpuhp_mem;
ret = perf_pmu_register(&pmu_ptr->pmu, pmu_ptr->pmu.name, -1);
if (ret)
goto err_free_cpuhp_mem;
pr_debug("%s performance monitor hardware support registered\n",
pmu_ptr->pmu.name);
return 0;
err_free_cpuhp_mem:
imc_common_cpuhp_mem_free(pmu_ptr);
err_free_mem:
imc_common_mem_free(pmu_ptr);
return ret;
}
Reference in New Issue View Git Blame Copy Permalink