[PATCH] powerpc: G4+ oprofile support
This patch adds oprofile support for the 7450 and all its multitudinous
derivatives.
* Added 7450 (and derivatives) support for oprofile
* Changed e500 cputable to have oprofile model and cpu_type fields
* Added support for classic 32-bit performance monitor interrupt
* Cleaned up common powerpc oprofile code to be as common as possible
* Cleaned up oprofile_impl.h to reflect 32 bit classic code
* Added 32-bit MMCRx bitfield definitions and SPR numbers
Signed-off-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
diff --git a/arch/powerpc/kernel/cputable.c b/arch/powerpc/kernel/cputable.c
index 1d85ced..f7f2a83 100644
--- a/arch/powerpc/kernel/cputable.c
+++ b/arch/powerpc/kernel/cputable.c
@@ -545,7 +545,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7450 2.1 */
.pvr_mask = 0xffffffff,
@@ -556,7 +560,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7450 2.3 and newer */
.pvr_mask = 0xffff0000,
@@ -567,7 +575,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7455 rev 1.x */
.pvr_mask = 0xffffff00,
@@ -578,7 +590,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7455 rev 2.0 */
.pvr_mask = 0xffffffff,
@@ -589,7 +605,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7455 others */
.pvr_mask = 0xffff0000,
@@ -600,7 +620,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7447/7457 Rev 1.0 */
.pvr_mask = 0xffffffff,
@@ -611,7 +635,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7447/7457 Rev 1.1 */
.pvr_mask = 0xffffffff,
@@ -622,7 +650,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7447/7457 Rev 1.2 and later */
.pvr_mask = 0xffff0000,
@@ -633,7 +665,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7447A */
.pvr_mask = 0xffff0000,
@@ -644,7 +680,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 7448 */
.pvr_mask = 0xffff0000,
@@ -655,7 +695,11 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 6,
- .cpu_setup = __setup_cpu_745x
+ .cpu_setup = __setup_cpu_745x,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/7450",
+ .oprofile_model = &op_model_7450,
+#endif
},
{ /* 82xx (8240, 8245, 8260 are all 603e cores) */
.pvr_mask = 0x7fff0000,
@@ -979,6 +1023,10 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 4,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/e500",
+ .oprofile_model = &op_model_fsl_booke,
+#endif
},
{ /* e500v2 */
.pvr_mask = 0xffff0000,
@@ -992,6 +1040,10 @@
.icache_bsize = 32,
.dcache_bsize = 32,
.num_pmcs = 4,
+#ifdef CONFIG_OPROFILE
+ .oprofile_cpu_type = "ppc/e500",
+ .oprofile_model = &op_model_fsl_booke,
+#endif
},
#endif
#if !CLASSIC_PPC
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index 6359e36..bf37ef2 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -466,16 +466,11 @@
* by executing an altivec instruction.
*/
. = 0xf00
- b Trap_0f
+ b PerformanceMonitor
. = 0xf20
b AltiVecUnavailable
-Trap_0f:
- EXCEPTION_PROLOG
- addi r3,r1,STACK_FRAME_OVERHEAD
- EXC_XFER_EE(0xf00, unknown_exception)
-
/*
* Handle TLB miss for instruction on 603/603e.
* Note: we get an alternate set of r0 - r3 to use automatically.
@@ -719,6 +714,11 @@
#endif /* CONFIG_ALTIVEC */
EXC_XFER_EE_LITE(0xf20, altivec_unavailable_exception)
+PerformanceMonitor:
+ EXCEPTION_PROLOG
+ addi r3,r1,STACK_FRAME_OVERHEAD
+ EXC_XFER_STD(0xf00, performance_monitor_exception)
+
#ifdef CONFIG_ALTIVEC
/* Note that the AltiVec support is closely modeled after the FP
* support. Changes to one are likely to be applicable to the
diff --git a/arch/powerpc/kernel/pmc.c b/arch/powerpc/kernel/pmc.c
index 2d333cc..e6fb194 100644
--- a/arch/powerpc/kernel/pmc.c
+++ b/arch/powerpc/kernel/pmc.c
@@ -43,8 +43,13 @@
mtspr(SPRN_MMCR0, mmcr0);
}
#else
+/* Ensure exceptions are disabled */
static void dummy_perf(struct pt_regs *regs)
{
+ unsigned int mmcr0 = mfspr(SPRN_MMCR0);
+
+ mmcr0 &= ~(MMCR0_PMXE);
+ mtspr(SPRN_MMCR0, mmcr0);
}
#endif
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index 76b579c..6c79346 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -901,12 +901,10 @@
die("Unrecoverable VMX/Altivec Unavailable Exception", regs, SIGABRT);
}
-#if defined(CONFIG_PPC64) || defined(CONFIG_E500)
void performance_monitor_exception(struct pt_regs *regs)
{
perf_irq(regs);
}
-#endif
#ifdef CONFIG_8xx
void SoftwareEmulation(struct pt_regs *regs)
diff --git a/arch/powerpc/oprofile/Makefile b/arch/powerpc/oprofile/Makefile
index 0782d0c..554cd7c 100644
--- a/arch/powerpc/oprofile/Makefile
+++ b/arch/powerpc/oprofile/Makefile
@@ -9,3 +9,4 @@
oprofile-y := $(DRIVER_OBJS) common.o
oprofile-$(CONFIG_PPC64) += op_model_rs64.o op_model_power4.o
oprofile-$(CONFIG_FSL_BOOKE) += op_model_fsl_booke.o
+oprofile-$(CONFIG_PPC32) += op_model_7450.o
diff --git a/arch/powerpc/oprofile/common.c b/arch/powerpc/oprofile/common.c
index af2c05d..a370778 100644
--- a/arch/powerpc/oprofile/common.c
+++ b/arch/powerpc/oprofile/common.c
@@ -14,9 +14,6 @@
*/
#include <linux/oprofile.h>
-#ifndef __powerpc64__
-#include <linux/slab.h>
-#endif /* ! __powerpc64__ */
#include <linux/init.h>
#include <linux/smp.h>
#include <linux/errno.h>
@@ -31,10 +28,6 @@
static struct op_counter_config ctr[OP_MAX_COUNTER];
static struct op_system_config sys;
-#ifndef __powerpc64__
-static char *cpu_type;
-#endif /* ! __powerpc64__ */
-
static void op_handle_interrupt(struct pt_regs *regs)
{
model->handle_interrupt(regs, ctr);
@@ -53,14 +46,7 @@
model->reg_setup(ctr, &sys, model->num_counters);
/* Configure the registers on all cpus. */
-#ifdef __powerpc64__
on_each_cpu(model->cpu_setup, NULL, 0, 1);
-#else /* __powerpc64__ */
-#if 0
- /* FIXME: Make multi-cpu work */
- on_each_cpu(model->reg_setup, NULL, 0, 1);
-#endif
-#endif /* __powerpc64__ */
return 0;
}
@@ -95,7 +81,7 @@
{
int i;
-#ifdef __powerpc64__
+#ifdef CONFIG_PPC64
/*
* There is one mmcr0, mmcr1 and mmcra for setting the events for
* all of the counters.
@@ -103,7 +89,7 @@
oprofilefs_create_ulong(sb, root, "mmcr0", &sys.mmcr0);
oprofilefs_create_ulong(sb, root, "mmcr1", &sys.mmcr1);
oprofilefs_create_ulong(sb, root, "mmcra", &sys.mmcra);
-#endif /* __powerpc64__ */
+#endif
for (i = 0; i < model->num_counters; ++i) {
struct dentry *dir;
@@ -115,65 +101,46 @@
oprofilefs_create_ulong(sb, dir, "enabled", &ctr[i].enabled);
oprofilefs_create_ulong(sb, dir, "event", &ctr[i].event);
oprofilefs_create_ulong(sb, dir, "count", &ctr[i].count);
-#ifdef __powerpc64__
+
/*
- * We dont support per counter user/kernel selection, but
- * we leave the entries because userspace expects them
+ * Classic PowerPC doesn't support per-counter
+ * control like this, but the options are
+ * expected, so they remain. For Freescale
+ * Book-E style performance monitors, we do
+ * support them.
*/
-#endif /* __powerpc64__ */
oprofilefs_create_ulong(sb, dir, "kernel", &ctr[i].kernel);
oprofilefs_create_ulong(sb, dir, "user", &ctr[i].user);
-#ifndef __powerpc64__
- /* FIXME: Not sure if this is used */
-#endif /* ! __powerpc64__ */
oprofilefs_create_ulong(sb, dir, "unit_mask", &ctr[i].unit_mask);
}
oprofilefs_create_ulong(sb, root, "enable_kernel", &sys.enable_kernel);
oprofilefs_create_ulong(sb, root, "enable_user", &sys.enable_user);
-#ifdef __powerpc64__
+#ifdef CONFIG_PPC64
oprofilefs_create_ulong(sb, root, "backtrace_spinlocks",
&sys.backtrace_spinlocks);
-#endif /* __powerpc64__ */
+#endif
/* Default to tracing both kernel and user */
sys.enable_kernel = 1;
sys.enable_user = 1;
-#ifdef __powerpc64__
+#ifdef CONFIG_PPC64
/* Turn on backtracing through spinlocks by default */
sys.backtrace_spinlocks = 1;
-#endif /* __powerpc64__ */
+#endif
return 0;
}
int __init oprofile_arch_init(struct oprofile_operations *ops)
{
-#ifndef __powerpc64__
-#ifdef CONFIG_FSL_BOOKE
- model = &op_model_fsl_booke;
-#else
- return -ENODEV;
-#endif
-
- cpu_type = kmalloc(32, GFP_KERNEL);
- if (NULL == cpu_type)
- return -ENOMEM;
-
- sprintf(cpu_type, "ppc/%s", cur_cpu_spec->cpu_name);
-
- model->num_counters = cur_cpu_spec->num_pmcs;
-
- ops->cpu_type = cpu_type;
-#else /* __powerpc64__ */
if (!cur_cpu_spec->oprofile_model || !cur_cpu_spec->oprofile_cpu_type)
return -ENODEV;
model = cur_cpu_spec->oprofile_model;
model->num_counters = cur_cpu_spec->num_pmcs;
ops->cpu_type = cur_cpu_spec->oprofile_cpu_type;
-#endif /* __powerpc64__ */
ops->create_files = op_powerpc_create_files;
ops->setup = op_powerpc_setup;
ops->shutdown = op_powerpc_shutdown;
@@ -188,8 +155,4 @@
void oprofile_arch_exit(void)
{
-#ifndef __powerpc64__
- kfree(cpu_type);
- cpu_type = NULL;
-#endif /* ! __powerpc64__ */
}
diff --git a/arch/powerpc/oprofile/op_model_7450.c b/arch/powerpc/oprofile/op_model_7450.c
new file mode 100644
index 0000000..32abfdb
--- /dev/null
+++ b/arch/powerpc/oprofile/op_model_7450.c
@@ -0,0 +1,206 @@
+/*
+ * oprofile/op_model_7450.c
+ *
+ * Freescale 745x/744x oprofile support, based on fsl_booke support
+ * Copyright (C) 2004 Anton Blanchard <anton@au.ibm.com>, IBM
+ *
+ * Copyright (c) 2004 Freescale Semiconductor, Inc
+ *
+ * Author: Andy Fleming
+ * Maintainer: Kumar Gala <galak@kernel.crashing.org>
+ *
+ * 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 (at your option) any later version.
+ */
+
+#include <linux/oprofile.h>
+#include <linux/init.h>
+#include <linux/smp.h>
+#include <asm/ptrace.h>
+#include <asm/system.h>
+#include <asm/processor.h>
+#include <asm/cputable.h>
+#include <asm/page.h>
+#include <asm/pmc.h>
+#include <asm/oprofile_impl.h>
+
+static unsigned long reset_value[OP_MAX_COUNTER];
+
+static int oprofile_running;
+static u32 mmcr0_val, mmcr1_val, mmcr2_val;
+
+#define MMCR0_PMC1_SHIFT 6
+#define MMCR0_PMC2_SHIFT 0
+#define MMCR1_PMC3_SHIFT 27
+#define MMCR1_PMC4_SHIFT 22
+#define MMCR1_PMC5_SHIFT 17
+#define MMCR1_PMC6_SHIFT 11
+
+#define mmcr0_event1(event) \
+ ((event << MMCR0_PMC1_SHIFT) & MMCR0_PMC1SEL)
+#define mmcr0_event2(event) \
+ ((event << MMCR0_PMC2_SHIFT) & MMCR0_PMC2SEL)
+
+#define mmcr1_event3(event) \
+ ((event << MMCR1_PMC3_SHIFT) & MMCR1_PMC3SEL)
+#define mmcr1_event4(event) \
+ ((event << MMCR1_PMC4_SHIFT) & MMCR1_PMC4SEL)
+#define mmcr1_event5(event) \
+ ((event << MMCR1_PMC5_SHIFT) & MMCR1_PMC5SEL)
+#define mmcr1_event6(event) \
+ ((event << MMCR1_PMC6_SHIFT) & MMCR1_PMC6SEL)
+
+#define MMCR0_INIT (MMCR0_FC | MMCR0_FCS | MMCR0_FCP | MMCR0_FCM1 | MMCR0_FCM0)
+
+/* Unfreezes the counters on this CPU, enables the interrupt,
+ * enables the counters to trigger the interrupt, and sets the
+ * counters to only count when the mark bit is not set.
+ */
+static void pmc_start_ctrs(void)
+{
+ u32 mmcr0 = mfspr(SPRN_MMCR0);
+
+ mmcr0 &= ~(MMCR0_FC | MMCR0_FCM0);
+ mmcr0 |= (MMCR0_FCECE | MMCR0_PMC1CE | MMCR0_PMCnCE | MMCR0_PMXE);
+
+ mtspr(SPRN_MMCR0, mmcr0);
+}
+
+/* Disables the counters on this CPU, and freezes them */
+static void pmc_stop_ctrs(void)
+{
+ u32 mmcr0 = mfspr(SPRN_MMCR0);
+
+ mmcr0 |= MMCR0_FC;
+ mmcr0 &= ~(MMCR0_FCECE | MMCR0_PMC1CE | MMCR0_PMCnCE | MMCR0_PMXE);
+
+ mtspr(SPRN_MMCR0, mmcr0);
+}
+
+/* Configures the counters on this CPU based on the global
+ * settings */
+static void fsl7450_cpu_setup(void *unused)
+{
+ /* freeze all counters */
+ pmc_stop_ctrs();
+
+ mtspr(SPRN_MMCR0, mmcr0_val);
+ mtspr(SPRN_MMCR1, mmcr1_val);
+ mtspr(SPRN_MMCR2, mmcr2_val);
+}
+
+#define NUM_CTRS 6
+
+/* Configures the global settings for the countes on all CPUs. */
+static void fsl7450_reg_setup(struct op_counter_config *ctr,
+ struct op_system_config *sys,
+ int num_ctrs)
+{
+ int i;
+
+ /* Our counters count up, and "count" refers to
+ * how much before the next interrupt, and we interrupt
+ * on overflow. So we calculate the starting value
+ * which will give us "count" until overflow.
+ * Then we set the events on the enabled counters */
+ for (i = 0; i < NUM_CTRS; ++i)
+ reset_value[i] = 0x80000000UL - ctr[i].count;
+
+ /* Set events for Counters 1 & 2 */
+ mmcr0_val = MMCR0_INIT | mmcr0_event1(ctr[0].event)
+ | mmcr0_event2(ctr[1].event);
+
+ /* Setup user/kernel bits */
+ if (sys->enable_kernel)
+ mmcr0_val &= ~(MMCR0_FCS);
+
+ if (sys->enable_user)
+ mmcr0_val &= ~(MMCR0_FCP);
+
+ /* Set events for Counters 3-6 */
+ mmcr1_val = mmcr1_event3(ctr[2].event)
+ | mmcr1_event4(ctr[3].event)
+ | mmcr1_event5(ctr[4].event)
+ | mmcr1_event6(ctr[5].event);
+
+ mmcr2_val = 0;
+}
+
+/* Sets the counters on this CPU to the chosen values, and starts them */
+static void fsl7450_start(struct op_counter_config *ctr)
+{
+ int i;
+
+ mtmsr(mfmsr() | MSR_PMM);
+
+ for (i = 0; i < NUM_CTRS; ++i) {
+ if (ctr[i].enabled)
+ ctr_write(i, reset_value[i]);
+ else
+ ctr_write(i, 0);
+ }
+
+ /* Clear the freeze bit, and enable the interrupt.
+ * The counters won't actually start until the rfi clears
+ * the PMM bit */
+ pmc_start_ctrs();
+
+ oprofile_running = 1;
+}
+
+/* Stop the counters on this CPU */
+static void fsl7450_stop(void)
+{
+ /* freeze counters */
+ pmc_stop_ctrs();
+
+ oprofile_running = 0;
+
+ mb();
+}
+
+
+/* Handle the interrupt on this CPU, and log a sample for each
+ * event that triggered the interrupt */
+static void fsl7450_handle_interrupt(struct pt_regs *regs,
+ struct op_counter_config *ctr)
+{
+ unsigned long pc;
+ int is_kernel;
+ int val;
+ int i;
+
+ /* set the PMM bit (see comment below) */
+ mtmsr(mfmsr() | MSR_PMM);
+
+ pc = mfspr(SPRN_SIAR);
+ is_kernel = (pc >= KERNELBASE);
+
+ for (i = 0; i < NUM_CTRS; ++i) {
+ val = ctr_read(i);
+ if (val < 0) {
+ if (oprofile_running && ctr[i].enabled) {
+ oprofile_add_pc(pc, is_kernel, i);
+ ctr_write(i, reset_value[i]);
+ } else {
+ ctr_write(i, 0);
+ }
+ }
+ }
+
+ /* The freeze bit was set by the interrupt. */
+ /* Clear the freeze bit, and reenable the interrupt.
+ * The counters won't actually start until the rfi clears
+ * the PMM bit */
+ pmc_start_ctrs();
+}
+
+struct op_powerpc_model op_model_7450= {
+ .reg_setup = fsl7450_reg_setup,
+ .cpu_setup = fsl7450_cpu_setup,
+ .start = fsl7450_start,
+ .stop = fsl7450_stop,
+ .handle_interrupt = fsl7450_handle_interrupt,
+};