forked from luck/tmp_suning_uos_patched
18f2190d79
Add PPU event-based and cycle-based profiling support to Oprofile for Cell. Oprofile is expected to collect data on all CPUs simultaneously. However, there is one set of performance counters per node. There are two hardware threads or virtual CPUs on each node. Hence, OProfile must multiplex in time the performance counter collection on the two virtual CPUs. The multiplexing of the performance counters is done by a virtual counter routine. Initially, the counters are configured to collect data on the even CPUs in the system, one CPU per node. In order to capture the PC for the virtual CPU when the performance counter interrupt occurs (the specified number of events between samples has occurred), the even processors are configured to handle the performance counter interrupts for their node. The virtual counter routine is called via a kernel timer after the virtual sample time. The routine stops the counters, saves the current counts, loads the last counts for the other virtual CPU on the node, sets interrupts to be handled by the other virtual CPU and restarts the counters, the virtual timer routine is scheduled to run again. The virtual sample time is kept relatively small to make sure sampling occurs on both CPUs on the node with a relatively small granularity. Whenever the counters overflow, the performance counter interrupt is called to collect the PC for the CPU where data is being collected. The oprofile driver relies on a firmware RTAS call to setup the debug bus to route the desired signals to the performance counter hardware to be counted. The RTAS call must set the routing registers appropriately in each of the islands to pass the signals down the debug bus as well as routing the signals from a particular island onto the bus. There is a second firmware RTAS call to reset the debug bus to the non pass thru state when the counters are not in use. Signed-off-by: Carl Love <carll@us.ibm.com> Signed-off-by: Maynard Johnson <mpjohn@us.ibm.com> Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
192 lines
4.7 KiB
C
192 lines
4.7 KiB
C
/*
|
|
* cbe_regs.c
|
|
*
|
|
* Accessor routines for the various MMIO register blocks of the CBE
|
|
*
|
|
* (c) 2006 Benjamin Herrenschmidt <benh@kernel.crashing.org>, IBM Corp.
|
|
*/
|
|
|
|
#include <linux/percpu.h>
|
|
#include <linux/types.h>
|
|
#include <linux/module.h>
|
|
|
|
#include <asm/io.h>
|
|
#include <asm/pgtable.h>
|
|
#include <asm/prom.h>
|
|
#include <asm/ptrace.h>
|
|
|
|
#include "cbe_regs.h"
|
|
|
|
/*
|
|
* Current implementation uses "cpu" nodes. We build our own mapping
|
|
* array of cpu numbers to cpu nodes locally for now to allow interrupt
|
|
* time code to have a fast path rather than call of_get_cpu_node(). If
|
|
* we implement cpu hotplug, we'll have to install an appropriate norifier
|
|
* in order to release references to the cpu going away
|
|
*/
|
|
static struct cbe_regs_map
|
|
{
|
|
struct device_node *cpu_node;
|
|
struct cbe_pmd_regs __iomem *pmd_regs;
|
|
struct cbe_iic_regs __iomem *iic_regs;
|
|
struct cbe_mic_tm_regs __iomem *mic_tm_regs;
|
|
struct cbe_pmd_shadow_regs pmd_shadow_regs;
|
|
} cbe_regs_maps[MAX_CBE];
|
|
static int cbe_regs_map_count;
|
|
|
|
static struct cbe_thread_map
|
|
{
|
|
struct device_node *cpu_node;
|
|
struct cbe_regs_map *regs;
|
|
} cbe_thread_map[NR_CPUS];
|
|
|
|
static struct cbe_regs_map *cbe_find_map(struct device_node *np)
|
|
{
|
|
int i;
|
|
struct device_node *tmp_np;
|
|
|
|
if (strcasecmp(np->type, "spe") == 0) {
|
|
if (np->data == NULL) {
|
|
/* walk up path until cpu node was found */
|
|
tmp_np = np->parent;
|
|
while (tmp_np != NULL && strcasecmp(tmp_np->type, "cpu") != 0)
|
|
tmp_np = tmp_np->parent;
|
|
|
|
np->data = cbe_find_map(tmp_np);
|
|
}
|
|
return np->data;
|
|
}
|
|
|
|
for (i = 0; i < cbe_regs_map_count; i++)
|
|
if (cbe_regs_maps[i].cpu_node == np)
|
|
return &cbe_regs_maps[i];
|
|
return NULL;
|
|
}
|
|
|
|
struct cbe_pmd_regs __iomem *cbe_get_pmd_regs(struct device_node *np)
|
|
{
|
|
struct cbe_regs_map *map = cbe_find_map(np);
|
|
if (map == NULL)
|
|
return NULL;
|
|
return map->pmd_regs;
|
|
}
|
|
EXPORT_SYMBOL_GPL(cbe_get_pmd_regs);
|
|
|
|
struct cbe_pmd_regs __iomem *cbe_get_cpu_pmd_regs(int cpu)
|
|
{
|
|
struct cbe_regs_map *map = cbe_thread_map[cpu].regs;
|
|
if (map == NULL)
|
|
return NULL;
|
|
return map->pmd_regs;
|
|
}
|
|
EXPORT_SYMBOL_GPL(cbe_get_cpu_pmd_regs);
|
|
|
|
struct cbe_pmd_shadow_regs *cbe_get_pmd_shadow_regs(struct device_node *np)
|
|
{
|
|
struct cbe_regs_map *map = cbe_find_map(np);
|
|
if (map == NULL)
|
|
return NULL;
|
|
return &map->pmd_shadow_regs;
|
|
}
|
|
|
|
struct cbe_pmd_shadow_regs *cbe_get_cpu_pmd_shadow_regs(int cpu)
|
|
{
|
|
struct cbe_regs_map *map = cbe_thread_map[cpu].regs;
|
|
if (map == NULL)
|
|
return NULL;
|
|
return &map->pmd_shadow_regs;
|
|
}
|
|
|
|
struct cbe_iic_regs __iomem *cbe_get_iic_regs(struct device_node *np)
|
|
{
|
|
struct cbe_regs_map *map = cbe_find_map(np);
|
|
if (map == NULL)
|
|
return NULL;
|
|
return map->iic_regs;
|
|
}
|
|
|
|
struct cbe_iic_regs __iomem *cbe_get_cpu_iic_regs(int cpu)
|
|
{
|
|
struct cbe_regs_map *map = cbe_thread_map[cpu].regs;
|
|
if (map == NULL)
|
|
return NULL;
|
|
return map->iic_regs;
|
|
}
|
|
|
|
struct cbe_mic_tm_regs __iomem *cbe_get_mic_tm_regs(struct device_node *np)
|
|
{
|
|
struct cbe_regs_map *map = cbe_find_map(np);
|
|
if (map == NULL)
|
|
return NULL;
|
|
return map->mic_tm_regs;
|
|
}
|
|
|
|
struct cbe_mic_tm_regs __iomem *cbe_get_cpu_mic_tm_regs(int cpu)
|
|
{
|
|
struct cbe_regs_map *map = cbe_thread_map[cpu].regs;
|
|
if (map == NULL)
|
|
return NULL;
|
|
return map->mic_tm_regs;
|
|
}
|
|
EXPORT_SYMBOL_GPL(cbe_get_cpu_mic_tm_regs);
|
|
|
|
/* FIXME
|
|
* This is little more than a stub at the moment. It should be
|
|
* fleshed out so that it works for both SMT and non-SMT, no
|
|
* matter if the passed cpu is odd or even.
|
|
* For SMT enabled, returns 0 for even-numbered cpu; otherwise 1.
|
|
* For SMT disabled, returns 0 for all cpus.
|
|
*/
|
|
u32 cbe_get_hw_thread_id(int cpu)
|
|
{
|
|
return (cpu & 1);
|
|
}
|
|
EXPORT_SYMBOL_GPL(cbe_get_hw_thread_id);
|
|
|
|
void __init cbe_regs_init(void)
|
|
{
|
|
int i;
|
|
struct device_node *cpu;
|
|
|
|
/* Build local fast map of CPUs */
|
|
for_each_possible_cpu(i)
|
|
cbe_thread_map[i].cpu_node = of_get_cpu_node(i, NULL);
|
|
|
|
/* Find maps for each device tree CPU */
|
|
for_each_node_by_type(cpu, "cpu") {
|
|
struct cbe_regs_map *map = &cbe_regs_maps[cbe_regs_map_count++];
|
|
|
|
/* That hack must die die die ! */
|
|
const struct address_prop {
|
|
unsigned long address;
|
|
unsigned int len;
|
|
} __attribute__((packed)) *prop;
|
|
|
|
|
|
if (cbe_regs_map_count > MAX_CBE) {
|
|
printk(KERN_ERR "cbe_regs: More BE chips than supported"
|
|
"!\n");
|
|
cbe_regs_map_count--;
|
|
return;
|
|
}
|
|
map->cpu_node = cpu;
|
|
for_each_possible_cpu(i)
|
|
if (cbe_thread_map[i].cpu_node == cpu)
|
|
cbe_thread_map[i].regs = map;
|
|
|
|
prop = get_property(cpu, "pervasive", NULL);
|
|
if (prop != NULL)
|
|
map->pmd_regs = ioremap(prop->address, prop->len);
|
|
|
|
prop = get_property(cpu, "iic", NULL);
|
|
if (prop != NULL)
|
|
map->iic_regs = ioremap(prop->address, prop->len);
|
|
|
|
prop = (struct address_prop *)get_property(cpu, "mic-tm",
|
|
NULL);
|
|
if (prop != NULL)
|
|
map->mic_tm_regs = ioremap(prop->address, prop->len);
|
|
}
|
|
}
|
|
|