|
@@ -1,6 +1,4 @@
|
|
/*
|
|
/*
|
|
- * Copyright (C) 2001, 2002, 2003 Broadcom Corporation
|
|
|
|
- *
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU General Public License
|
|
* modify it under the terms of the GNU General Public License
|
|
* as published by the Free Software Foundation; either version 2
|
|
* as published by the Free Software Foundation; either version 2
|
|
@@ -14,10 +12,16 @@
|
|
* You should have received a copy of the GNU General Public License
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write to the Free Software
|
|
* along with this program; if not, write to the Free Software
|
|
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
|
|
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
|
|
|
|
+ *
|
|
|
|
+ * Copyright (C) 2001, 2002, 2003 Broadcom Corporation
|
|
|
|
+ * Copyright (C) 2007 Ralf Baechle <ralf@linux-mips.org>
|
|
|
|
+ * Copyright (C) 2007 MIPS Technologies, Inc.
|
|
|
|
+ * written by Ralf Baechle <ralf@linux-mips.org>
|
|
*/
|
|
*/
|
|
|
|
|
|
-#define SBPROF_TB_DEBUG 0
|
|
|
|
|
|
+#undef DEBUG
|
|
|
|
|
|
|
|
+#include <linux/device.h>
|
|
#include <linux/module.h>
|
|
#include <linux/module.h>
|
|
#include <linux/kernel.h>
|
|
#include <linux/kernel.h>
|
|
#include <linux/types.h>
|
|
#include <linux/types.h>
|
|
@@ -27,24 +31,98 @@
|
|
#include <linux/vmalloc.h>
|
|
#include <linux/vmalloc.h>
|
|
#include <linux/fs.h>
|
|
#include <linux/fs.h>
|
|
#include <linux/errno.h>
|
|
#include <linux/errno.h>
|
|
-#include <linux/reboot.h>
|
|
|
|
-#include <linux/smp_lock.h>
|
|
|
|
|
|
+#include <linux/types.h>
|
|
#include <linux/wait.h>
|
|
#include <linux/wait.h>
|
|
-#include <asm/uaccess.h>
|
|
|
|
|
|
+
|
|
#include <asm/io.h>
|
|
#include <asm/io.h>
|
|
#include <asm/sibyte/sb1250.h>
|
|
#include <asm/sibyte/sb1250.h>
|
|
#include <asm/sibyte/sb1250_regs.h>
|
|
#include <asm/sibyte/sb1250_regs.h>
|
|
#include <asm/sibyte/sb1250_scd.h>
|
|
#include <asm/sibyte/sb1250_scd.h>
|
|
#include <asm/sibyte/sb1250_int.h>
|
|
#include <asm/sibyte/sb1250_int.h>
|
|
-#include <asm/sibyte/trace_prof.h>
|
|
|
|
|
|
+#include <asm/system.h>
|
|
|
|
+#include <asm/uaccess.h>
|
|
|
|
|
|
-#define DEVNAME "bcm1250_tbprof"
|
|
|
|
|
|
+#define SBPROF_TB_MAJOR 240
|
|
|
|
+
|
|
|
|
+typedef u64 tb_sample_t[6*256];
|
|
|
|
+
|
|
|
|
+enum open_status {
|
|
|
|
+ SB_CLOSED,
|
|
|
|
+ SB_OPENING,
|
|
|
|
+ SB_OPEN
|
|
|
|
+};
|
|
|
|
+
|
|
|
|
+struct sbprof_tb {
|
|
|
|
+ wait_queue_head_t tb_sync;
|
|
|
|
+ wait_queue_head_t tb_read;
|
|
|
|
+ struct mutex lock;
|
|
|
|
+ enum open_status open;
|
|
|
|
+ tb_sample_t *sbprof_tbbuf;
|
|
|
|
+ int next_tb_sample;
|
|
|
|
+
|
|
|
|
+ volatile int tb_enable;
|
|
|
|
+ volatile int tb_armed;
|
|
|
|
+
|
|
|
|
+};
|
|
|
|
|
|
static struct sbprof_tb sbp;
|
|
static struct sbprof_tb sbp;
|
|
|
|
|
|
|
|
+#define MAX_SAMPLE_BYTES (24*1024*1024)
|
|
|
|
+#define MAX_TBSAMPLE_BYTES (12*1024*1024)
|
|
|
|
+
|
|
|
|
+#define MAX_SAMPLES (MAX_SAMPLE_BYTES/sizeof(u_int32_t))
|
|
|
|
+#define TB_SAMPLE_SIZE (sizeof(tb_sample_t))
|
|
|
|
+#define MAX_TB_SAMPLES (MAX_TBSAMPLE_BYTES/TB_SAMPLE_SIZE)
|
|
|
|
+
|
|
|
|
+/* ioctls */
|
|
|
|
+#define SBPROF_ZBSTART _IOW('s', 0, int)
|
|
|
|
+#define SBPROF_ZBSTOP _IOW('s', 1, int)
|
|
|
|
+#define SBPROF_ZBWAITFULL _IOW('s', 2, int)
|
|
|
|
+
|
|
|
|
+/*
|
|
|
|
+ * Routines for using 40-bit SCD cycle counter
|
|
|
|
+ *
|
|
|
|
+ * Client responsible for either handling interrupts or making sure
|
|
|
|
+ * the cycles counter never saturates, e.g., by doing
|
|
|
|
+ * zclk_timer_init(0) at least every 2^40 - 1 ZCLKs.
|
|
|
|
+ */
|
|
|
|
+
|
|
|
|
+/*
|
|
|
|
+ * Configures SCD counter 0 to count ZCLKs starting from val;
|
|
|
|
+ * Configures SCD counters1,2,3 to count nothing.
|
|
|
|
+ * Must not be called while gathering ZBbus profiles.
|
|
|
|
+ */
|
|
|
|
+
|
|
|
|
+#define zclk_timer_init(val) \
|
|
|
|
+ __asm__ __volatile__ (".set push;" \
|
|
|
|
+ ".set mips64;" \
|
|
|
|
+ "la $8, 0xb00204c0;" /* SCD perf_cnt_cfg */ \
|
|
|
|
+ "sd %0, 0x10($8);" /* write val to counter0 */ \
|
|
|
|
+ "sd %1, 0($8);" /* config counter0 for zclks*/ \
|
|
|
|
+ ".set pop" \
|
|
|
|
+ : /* no outputs */ \
|
|
|
|
+ /* enable, counter0 */ \
|
|
|
|
+ : /* inputs */ "r"(val), "r" ((1ULL << 33) | 1ULL) \
|
|
|
|
+ : /* modifies */ "$8" )
|
|
|
|
+
|
|
|
|
+
|
|
|
|
+/* Reads SCD counter 0 and puts result in value
|
|
|
|
+ unsigned long long val; */
|
|
|
|
+#define zclk_get(val) \
|
|
|
|
+ __asm__ __volatile__ (".set push;" \
|
|
|
|
+ ".set mips64;" \
|
|
|
|
+ "la $8, 0xb00204c0;" /* SCD perf_cnt_cfg */ \
|
|
|
|
+ "ld %0, 0x10($8);" /* write val to counter0 */ \
|
|
|
|
+ ".set pop" \
|
|
|
|
+ : /* outputs */ "=r"(val) \
|
|
|
|
+ : /* inputs */ \
|
|
|
|
+ : /* modifies */ "$8" )
|
|
|
|
+
|
|
|
|
+#define DEVNAME "bcm1250_tbprof"
|
|
|
|
+
|
|
#define TB_FULL (sbp.next_tb_sample == MAX_TB_SAMPLES)
|
|
#define TB_FULL (sbp.next_tb_sample == MAX_TB_SAMPLES)
|
|
|
|
|
|
-/************************************************************************
|
|
|
|
|
|
+/*
|
|
* Support for ZBbus sampling using the trace buffer
|
|
* Support for ZBbus sampling using the trace buffer
|
|
*
|
|
*
|
|
* We use the SCD performance counter interrupt, caused by a Zclk counter
|
|
* We use the SCD performance counter interrupt, caused by a Zclk counter
|
|
@@ -54,30 +132,36 @@ static struct sbprof_tb sbp;
|
|
* overflow.
|
|
* overflow.
|
|
*
|
|
*
|
|
* We map the interrupt for trace_buffer_freeze to handle it on CPU 0.
|
|
* We map the interrupt for trace_buffer_freeze to handle it on CPU 0.
|
|
- *
|
|
|
|
- ************************************************************************/
|
|
|
|
|
|
+ */
|
|
|
|
|
|
-static u_int64_t tb_period;
|
|
|
|
|
|
+static u64 tb_period;
|
|
|
|
|
|
static void arm_tb(void)
|
|
static void arm_tb(void)
|
|
{
|
|
{
|
|
- u_int64_t scdperfcnt;
|
|
|
|
- u_int64_t next = (1ULL << 40) - tb_period;
|
|
|
|
- u_int64_t tb_options = M_SCD_TRACE_CFG_FREEZE_FULL;
|
|
|
|
- /* Generate an SCD_PERFCNT interrupt in TB_PERIOD Zclks to
|
|
|
|
- trigger start of trace. XXX vary sampling period */
|
|
|
|
|
|
+ u64 scdperfcnt;
|
|
|
|
+ u64 next = (1ULL << 40) - tb_period;
|
|
|
|
+ u64 tb_options = M_SCD_TRACE_CFG_FREEZE_FULL;
|
|
|
|
+
|
|
|
|
+ /*
|
|
|
|
+ * Generate an SCD_PERFCNT interrupt in TB_PERIOD Zclks to trigger
|
|
|
|
+ *start of trace. XXX vary sampling period
|
|
|
|
+ */
|
|
__raw_writeq(0, IOADDR(A_SCD_PERF_CNT_1));
|
|
__raw_writeq(0, IOADDR(A_SCD_PERF_CNT_1));
|
|
scdperfcnt = __raw_readq(IOADDR(A_SCD_PERF_CNT_CFG));
|
|
scdperfcnt = __raw_readq(IOADDR(A_SCD_PERF_CNT_CFG));
|
|
- /* Unfortunately, in Pass 2 we must clear all counters to knock down
|
|
|
|
- a previous interrupt request. This means that bus profiling
|
|
|
|
- requires ALL of the SCD perf counters. */
|
|
|
|
|
|
+
|
|
|
|
+ /*
|
|
|
|
+ * Unfortunately, in Pass 2 we must clear all counters to knock down a
|
|
|
|
+ * previous interrupt request. This means that bus profiling requires
|
|
|
|
+ * ALL of the SCD perf counters.
|
|
|
|
+ */
|
|
__raw_writeq((scdperfcnt & ~M_SPC_CFG_SRC1) |
|
|
__raw_writeq((scdperfcnt & ~M_SPC_CFG_SRC1) |
|
|
- // keep counters 0,2,3 as is
|
|
|
|
- M_SPC_CFG_ENABLE | // enable counting
|
|
|
|
- M_SPC_CFG_CLEAR | // clear all counters
|
|
|
|
- V_SPC_CFG_SRC1(1), // counter 1 counts cycles
|
|
|
|
|
|
+ /* keep counters 0,2,3 as is */
|
|
|
|
+ M_SPC_CFG_ENABLE | /* enable counting */
|
|
|
|
+ M_SPC_CFG_CLEAR | /* clear all counters */
|
|
|
|
+ V_SPC_CFG_SRC1(1), /* counter 1 counts cycles */
|
|
IOADDR(A_SCD_PERF_CNT_CFG));
|
|
IOADDR(A_SCD_PERF_CNT_CFG));
|
|
__raw_writeq(next, IOADDR(A_SCD_PERF_CNT_1));
|
|
__raw_writeq(next, IOADDR(A_SCD_PERF_CNT_1));
|
|
|
|
+
|
|
/* Reset the trace buffer */
|
|
/* Reset the trace buffer */
|
|
__raw_writeq(M_SCD_TRACE_CFG_RESET, IOADDR(A_SCD_TRACE_CFG));
|
|
__raw_writeq(M_SCD_TRACE_CFG_RESET, IOADDR(A_SCD_TRACE_CFG));
|
|
#if 0 && defined(M_SCD_TRACE_CFG_FORCECNT)
|
|
#if 0 && defined(M_SCD_TRACE_CFG_FORCECNT)
|
|
@@ -91,43 +175,45 @@ static void arm_tb(void)
|
|
static irqreturn_t sbprof_tb_intr(int irq, void *dev_id)
|
|
static irqreturn_t sbprof_tb_intr(int irq, void *dev_id)
|
|
{
|
|
{
|
|
int i;
|
|
int i;
|
|
- DBG(printk(DEVNAME ": tb_intr\n"));
|
|
|
|
|
|
+
|
|
|
|
+ pr_debug(DEVNAME ": tb_intr\n");
|
|
|
|
+
|
|
if (sbp.next_tb_sample < MAX_TB_SAMPLES) {
|
|
if (sbp.next_tb_sample < MAX_TB_SAMPLES) {
|
|
/* XXX should use XKPHYS to make writes bypass L2 */
|
|
/* XXX should use XKPHYS to make writes bypass L2 */
|
|
- u_int64_t *p = sbp.sbprof_tbbuf[sbp.next_tb_sample++];
|
|
|
|
|
|
+ u64 *p = sbp.sbprof_tbbuf[sbp.next_tb_sample++];
|
|
/* Read out trace */
|
|
/* Read out trace */
|
|
__raw_writeq(M_SCD_TRACE_CFG_START_READ,
|
|
__raw_writeq(M_SCD_TRACE_CFG_START_READ,
|
|
IOADDR(A_SCD_TRACE_CFG));
|
|
IOADDR(A_SCD_TRACE_CFG));
|
|
__asm__ __volatile__ ("sync" : : : "memory");
|
|
__asm__ __volatile__ ("sync" : : : "memory");
|
|
/* Loop runs backwards because bundles are read out in reverse order */
|
|
/* Loop runs backwards because bundles are read out in reverse order */
|
|
for (i = 256 * 6; i > 0; i -= 6) {
|
|
for (i = 256 * 6; i > 0; i -= 6) {
|
|
- // Subscripts decrease to put bundle in the order
|
|
|
|
- // t0 lo, t0 hi, t1 lo, t1 hi, t2 lo, t2 hi
|
|
|
|
|
|
+ /* Subscripts decrease to put bundle in the order */
|
|
|
|
+ /* t0 lo, t0 hi, t1 lo, t1 hi, t2 lo, t2 hi */
|
|
p[i - 1] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
p[i - 1] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
- // read t2 hi
|
|
|
|
|
|
+ /* read t2 hi */
|
|
p[i - 2] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
p[i - 2] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
- // read t2 lo
|
|
|
|
|
|
+ /* read t2 lo */
|
|
p[i - 3] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
p[i - 3] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
- // read t1 hi
|
|
|
|
|
|
+ /* read t1 hi */
|
|
p[i - 4] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
p[i - 4] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
- // read t1 lo
|
|
|
|
|
|
+ /* read t1 lo */
|
|
p[i - 5] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
p[i - 5] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
- // read t0 hi
|
|
|
|
|
|
+ /* read t0 hi */
|
|
p[i - 6] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
p[i - 6] = __raw_readq(IOADDR(A_SCD_TRACE_READ));
|
|
- // read t0 lo
|
|
|
|
|
|
+ /* read t0 lo */
|
|
}
|
|
}
|
|
if (!sbp.tb_enable) {
|
|
if (!sbp.tb_enable) {
|
|
- DBG(printk(DEVNAME ": tb_intr shutdown\n"));
|
|
|
|
|
|
+ pr_debug(DEVNAME ": tb_intr shutdown\n");
|
|
__raw_writeq(M_SCD_TRACE_CFG_RESET,
|
|
__raw_writeq(M_SCD_TRACE_CFG_RESET,
|
|
IOADDR(A_SCD_TRACE_CFG));
|
|
IOADDR(A_SCD_TRACE_CFG));
|
|
sbp.tb_armed = 0;
|
|
sbp.tb_armed = 0;
|
|
wake_up(&sbp.tb_sync);
|
|
wake_up(&sbp.tb_sync);
|
|
} else {
|
|
} else {
|
|
- arm_tb(); // knock down current interrupt and get another one later
|
|
|
|
|
|
+ arm_tb(); /* knock down current interrupt and get another one later */
|
|
}
|
|
}
|
|
} else {
|
|
} else {
|
|
/* No more trace buffer samples */
|
|
/* No more trace buffer samples */
|
|
- DBG(printk(DEVNAME ": tb_intr full\n"));
|
|
|
|
|
|
+ pr_debug(DEVNAME ": tb_intr full\n");
|
|
__raw_writeq(M_SCD_TRACE_CFG_RESET, IOADDR(A_SCD_TRACE_CFG));
|
|
__raw_writeq(M_SCD_TRACE_CFG_RESET, IOADDR(A_SCD_TRACE_CFG));
|
|
sbp.tb_armed = 0;
|
|
sbp.tb_armed = 0;
|
|
if (!sbp.tb_enable) {
|
|
if (!sbp.tb_enable) {
|
|
@@ -135,6 +221,7 @@ static irqreturn_t sbprof_tb_intr(int irq, void *dev_id)
|
|
}
|
|
}
|
|
wake_up(&sbp.tb_read);
|
|
wake_up(&sbp.tb_read);
|
|
}
|
|
}
|
|
|
|
+
|
|
return IRQ_HANDLED;
|
|
return IRQ_HANDLED;
|
|
}
|
|
}
|
|
|
|
|
|
@@ -144,23 +231,30 @@ static irqreturn_t sbprof_pc_intr(int irq, void *dev_id)
|
|
return IRQ_NONE;
|
|
return IRQ_NONE;
|
|
}
|
|
}
|
|
|
|
|
|
-int sbprof_zbprof_start(struct file *filp)
|
|
|
|
|
|
+/*
|
|
|
|
+ * Requires: Already called zclk_timer_init with a value that won't
|
|
|
|
+ * saturate 40 bits. No subsequent use of SCD performance counters
|
|
|
|
+ * or trace buffer.
|
|
|
|
+ */
|
|
|
|
+
|
|
|
|
+static int sbprof_zbprof_start(struct file *filp)
|
|
{
|
|
{
|
|
- u_int64_t scdperfcnt;
|
|
|
|
|
|
+ u64 scdperfcnt;
|
|
|
|
+ int err;
|
|
|
|
|
|
- if (sbp.tb_enable)
|
|
|
|
|
|
+ if (xchg(&sbp.tb_enable, 1))
|
|
return -EBUSY;
|
|
return -EBUSY;
|
|
|
|
|
|
- DBG(printk(DEVNAME ": starting\n"));
|
|
|
|
|
|
+ pr_debug(DEVNAME ": starting\n");
|
|
|
|
|
|
- sbp.tb_enable = 1;
|
|
|
|
sbp.next_tb_sample = 0;
|
|
sbp.next_tb_sample = 0;
|
|
filp->f_pos = 0;
|
|
filp->f_pos = 0;
|
|
|
|
|
|
- if (request_irq
|
|
|
|
- (K_INT_TRACE_FREEZE, sbprof_tb_intr, 0, DEVNAME " trace freeze", &sbp)) {
|
|
|
|
|
|
+ err = request_irq(K_INT_TRACE_FREEZE, sbprof_tb_intr, 0,
|
|
|
|
+ DEVNAME " trace freeze", &sbp);
|
|
|
|
+ if (err)
|
|
return -EBUSY;
|
|
return -EBUSY;
|
|
- }
|
|
|
|
|
|
+
|
|
/* Make sure there isn't a perf-cnt interrupt waiting */
|
|
/* Make sure there isn't a perf-cnt interrupt waiting */
|
|
scdperfcnt = __raw_readq(IOADDR(A_SCD_PERF_CNT_CFG));
|
|
scdperfcnt = __raw_readq(IOADDR(A_SCD_PERF_CNT_CFG));
|
|
/* Disable and clear counters, override SRC_1 */
|
|
/* Disable and clear counters, override SRC_1 */
|
|
@@ -168,18 +262,21 @@ int sbprof_zbprof_start(struct file *filp)
|
|
M_SPC_CFG_ENABLE | M_SPC_CFG_CLEAR | V_SPC_CFG_SRC1(1),
|
|
M_SPC_CFG_ENABLE | M_SPC_CFG_CLEAR | V_SPC_CFG_SRC1(1),
|
|
IOADDR(A_SCD_PERF_CNT_CFG));
|
|
IOADDR(A_SCD_PERF_CNT_CFG));
|
|
|
|
|
|
- /* We grab this interrupt to prevent others from trying to use
|
|
|
|
- it, even though we don't want to service the interrupts
|
|
|
|
- (they only feed into the trace-on-interrupt mechanism) */
|
|
|
|
- if (request_irq
|
|
|
|
- (K_INT_PERF_CNT, sbprof_pc_intr, 0, DEVNAME " scd perfcnt", &sbp)) {
|
|
|
|
- free_irq(K_INT_TRACE_FREEZE, &sbp);
|
|
|
|
- return -EBUSY;
|
|
|
|
- }
|
|
|
|
-
|
|
|
|
- /* I need the core to mask these, but the interrupt mapper to
|
|
|
|
- pass them through. I am exploiting my knowledge that
|
|
|
|
- cp0_status masks out IP[5]. krw */
|
|
|
|
|
|
+ /*
|
|
|
|
+ * We grab this interrupt to prevent others from trying to use it, even
|
|
|
|
+ * though we don't want to service the interrupts (they only feed into
|
|
|
|
+ * the trace-on-interrupt mechanism)
|
|
|
|
+ */
|
|
|
|
+ err = request_irq(K_INT_PERF_CNT, sbprof_pc_intr, 0,
|
|
|
|
+ DEVNAME " scd perfcnt", &sbp);
|
|
|
|
+ if (err)
|
|
|
|
+ goto out_free_irq;
|
|
|
|
+
|
|
|
|
+ /*
|
|
|
|
+ * I need the core to mask these, but the interrupt mapper to pass them
|
|
|
|
+ * through. I am exploiting my knowledge that cp0_status masks out
|
|
|
|
+ * IP[5]. krw
|
|
|
|
+ */
|
|
__raw_writeq(K_INT_MAP_I3,
|
|
__raw_writeq(K_INT_MAP_I3,
|
|
IOADDR(A_IMR_REGISTER(0, R_IMR_INTERRUPT_MAP_BASE) +
|
|
IOADDR(A_IMR_REGISTER(0, R_IMR_INTERRUPT_MAP_BASE) +
|
|
(K_INT_PERF_CNT << 3)));
|
|
(K_INT_PERF_CNT << 3)));
|
|
@@ -201,7 +298,7 @@ int sbprof_zbprof_start(struct file *filp)
|
|
__raw_writeq(0, IOADDR(A_ADDR_TRAP_CFG_3));
|
|
__raw_writeq(0, IOADDR(A_ADDR_TRAP_CFG_3));
|
|
|
|
|
|
/* Initialize Trace Event 0-7 */
|
|
/* Initialize Trace Event 0-7 */
|
|
- // when interrupt
|
|
|
|
|
|
+ /* when interrupt */
|
|
__raw_writeq(M_SCD_TREVT_INTERRUPT, IOADDR(A_SCD_TRACE_EVENT_0));
|
|
__raw_writeq(M_SCD_TREVT_INTERRUPT, IOADDR(A_SCD_TRACE_EVENT_0));
|
|
__raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_1));
|
|
__raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_1));
|
|
__raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_2));
|
|
__raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_2));
|
|
@@ -212,10 +309,10 @@ int sbprof_zbprof_start(struct file *filp)
|
|
__raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_7));
|
|
__raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_7));
|
|
|
|
|
|
/* Initialize Trace Sequence 0-7 */
|
|
/* Initialize Trace Sequence 0-7 */
|
|
- // Start on event 0 (interrupt)
|
|
|
|
|
|
+ /* Start on event 0 (interrupt) */
|
|
__raw_writeq(V_SCD_TRSEQ_FUNC_START | 0x0fff,
|
|
__raw_writeq(V_SCD_TRSEQ_FUNC_START | 0x0fff,
|
|
IOADDR(A_SCD_TRACE_SEQUENCE_0));
|
|
IOADDR(A_SCD_TRACE_SEQUENCE_0));
|
|
- // dsamp when d used | asamp when a used
|
|
|
|
|
|
+ /* dsamp when d used | asamp when a used */
|
|
__raw_writeq(M_SCD_TRSEQ_ASAMPLE | M_SCD_TRSEQ_DSAMPLE |
|
|
__raw_writeq(M_SCD_TRSEQ_ASAMPLE | M_SCD_TRSEQ_DSAMPLE |
|
|
K_SCD_TRSEQ_TRIGGER_ALL,
|
|
K_SCD_TRSEQ_TRIGGER_ALL,
|
|
IOADDR(A_SCD_TRACE_SEQUENCE_1));
|
|
IOADDR(A_SCD_TRACE_SEQUENCE_1));
|
|
@@ -232,33 +329,41 @@ int sbprof_zbprof_start(struct file *filp)
|
|
|
|
|
|
arm_tb();
|
|
arm_tb();
|
|
|
|
|
|
- DBG(printk(DEVNAME ": done starting\n"));
|
|
|
|
|
|
+ pr_debug(DEVNAME ": done starting\n");
|
|
|
|
|
|
return 0;
|
|
return 0;
|
|
|
|
+
|
|
|
|
+out_free_irq:
|
|
|
|
+ free_irq(K_INT_TRACE_FREEZE, &sbp);
|
|
|
|
+
|
|
|
|
+ return err;
|
|
}
|
|
}
|
|
|
|
|
|
-int sbprof_zbprof_stop(void)
|
|
|
|
|
|
+static int sbprof_zbprof_stop(void)
|
|
{
|
|
{
|
|
- DEFINE_WAIT(wait);
|
|
|
|
- DBG(printk(DEVNAME ": stopping\n"));
|
|
|
|
|
|
+ int err;
|
|
|
|
+
|
|
|
|
+ pr_debug(DEVNAME ": stopping\n");
|
|
|
|
|
|
if (sbp.tb_enable) {
|
|
if (sbp.tb_enable) {
|
|
|
|
+ /*
|
|
|
|
+ * XXXKW there is a window here where the intr handler may run,
|
|
|
|
+ * see the disable, and do the wake_up before this sleep
|
|
|
|
+ * happens.
|
|
|
|
+ */
|
|
|
|
+ pr_debug(DEVNAME ": wait for disarm\n");
|
|
|
|
+ err = wait_event_interruptible(sbp.tb_sync, !sbp.tb_armed);
|
|
|
|
+ pr_debug(DEVNAME ": disarm complete, stat %d\n", err);
|
|
|
|
+
|
|
|
|
+ if (err)
|
|
|
|
+ return err;
|
|
|
|
+
|
|
sbp.tb_enable = 0;
|
|
sbp.tb_enable = 0;
|
|
- /* XXXKW there is a window here where the intr handler
|
|
|
|
- may run, see the disable, and do the wake_up before
|
|
|
|
- this sleep happens. */
|
|
|
|
- if (sbp.tb_armed) {
|
|
|
|
- DBG(printk(DEVNAME ": wait for disarm\n"));
|
|
|
|
- prepare_to_wait(&sbp.tb_sync, &wait, TASK_INTERRUPTIBLE);
|
|
|
|
- schedule();
|
|
|
|
- finish_wait(&sbp.tb_sync, &wait);
|
|
|
|
- DBG(printk(DEVNAME ": disarm complete\n"));
|
|
|
|
- }
|
|
|
|
free_irq(K_INT_TRACE_FREEZE, &sbp);
|
|
free_irq(K_INT_TRACE_FREEZE, &sbp);
|
|
free_irq(K_INT_PERF_CNT, &sbp);
|
|
free_irq(K_INT_PERF_CNT, &sbp);
|
|
}
|
|
}
|
|
|
|
|
|
- DBG(printk(DEVNAME ": done stopping\n"));
|
|
|
|
|
|
+ pr_debug(DEVNAME ": done stopping\n");
|
|
|
|
|
|
return 0;
|
|
return 0;
|
|
}
|
|
}
|
|
@@ -268,42 +373,45 @@ static int sbprof_tb_open(struct inode *inode, struct file *filp)
|
|
int minor;
|
|
int minor;
|
|
|
|
|
|
minor = iminor(inode);
|
|
minor = iminor(inode);
|
|
- if (minor != 0) {
|
|
|
|
|
|
+ if (minor != 0)
|
|
return -ENODEV;
|
|
return -ENODEV;
|
|
- }
|
|
|
|
- if (sbp.open) {
|
|
|
|
|
|
+
|
|
|
|
+ if (xchg(&sbp.open, SB_OPENING) != SB_CLOSED)
|
|
return -EBUSY;
|
|
return -EBUSY;
|
|
- }
|
|
|
|
|
|
|
|
memset(&sbp, 0, sizeof(struct sbprof_tb));
|
|
memset(&sbp, 0, sizeof(struct sbprof_tb));
|
|
|
|
+
|
|
sbp.sbprof_tbbuf = vmalloc(MAX_TBSAMPLE_BYTES);
|
|
sbp.sbprof_tbbuf = vmalloc(MAX_TBSAMPLE_BYTES);
|
|
- if (!sbp.sbprof_tbbuf) {
|
|
|
|
|
|
+ if (!sbp.sbprof_tbbuf)
|
|
return -ENOMEM;
|
|
return -ENOMEM;
|
|
- }
|
|
|
|
|
|
+
|
|
memset(sbp.sbprof_tbbuf, 0, MAX_TBSAMPLE_BYTES);
|
|
memset(sbp.sbprof_tbbuf, 0, MAX_TBSAMPLE_BYTES);
|
|
init_waitqueue_head(&sbp.tb_sync);
|
|
init_waitqueue_head(&sbp.tb_sync);
|
|
init_waitqueue_head(&sbp.tb_read);
|
|
init_waitqueue_head(&sbp.tb_read);
|
|
- sbp.open = 1;
|
|
|
|
|
|
+ mutex_init(&sbp.lock);
|
|
|
|
+
|
|
|
|
+ sbp.open = SB_OPEN;
|
|
|
|
|
|
return 0;
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
|
|
static int sbprof_tb_release(struct inode *inode, struct file *filp)
|
|
static int sbprof_tb_release(struct inode *inode, struct file *filp)
|
|
{
|
|
{
|
|
- int minor;
|
|
|
|
|
|
+ int minor = iminor(inode);
|
|
|
|
|
|
- minor = iminor(inode);
|
|
|
|
- if (minor != 0 || !sbp.open) {
|
|
|
|
|
|
+ if (minor != 0 || !sbp.open)
|
|
return -ENODEV;
|
|
return -ENODEV;
|
|
- }
|
|
|
|
|
|
|
|
- if (sbp.tb_armed || sbp.tb_enable) {
|
|
|
|
|
|
+ mutex_lock(&sbp.lock);
|
|
|
|
+
|
|
|
|
+ if (sbp.tb_armed || sbp.tb_enable)
|
|
sbprof_zbprof_stop();
|
|
sbprof_zbprof_stop();
|
|
- }
|
|
|
|
|
|
|
|
vfree(sbp.sbprof_tbbuf);
|
|
vfree(sbp.sbprof_tbbuf);
|
|
sbp.open = 0;
|
|
sbp.open = 0;
|
|
|
|
|
|
|
|
+ mutex_unlock(&sbp.lock);
|
|
|
|
+
|
|
return 0;
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
|
|
@@ -311,21 +419,35 @@ static ssize_t sbprof_tb_read(struct file *filp, char *buf,
|
|
size_t size, loff_t *offp)
|
|
size_t size, loff_t *offp)
|
|
{
|
|
{
|
|
int cur_sample, sample_off, cur_count, sample_left;
|
|
int cur_sample, sample_off, cur_count, sample_left;
|
|
- char *src;
|
|
|
|
- int count = 0;
|
|
|
|
- char *dest = buf;
|
|
|
|
long cur_off = *offp;
|
|
long cur_off = *offp;
|
|
|
|
+ char *dest = buf;
|
|
|
|
+ int count = 0;
|
|
|
|
+ char *src;
|
|
|
|
+
|
|
|
|
+ if (!access_ok(VERIFY_WRITE, buf, size))
|
|
|
|
+ return -EFAULT;
|
|
|
|
+
|
|
|
|
+ mutex_lock(&sbp.lock);
|
|
|
|
|
|
count = 0;
|
|
count = 0;
|
|
cur_sample = cur_off / TB_SAMPLE_SIZE;
|
|
cur_sample = cur_off / TB_SAMPLE_SIZE;
|
|
sample_off = cur_off % TB_SAMPLE_SIZE;
|
|
sample_off = cur_off % TB_SAMPLE_SIZE;
|
|
sample_left = TB_SAMPLE_SIZE - sample_off;
|
|
sample_left = TB_SAMPLE_SIZE - sample_off;
|
|
|
|
+
|
|
while (size && (cur_sample < sbp.next_tb_sample)) {
|
|
while (size && (cur_sample < sbp.next_tb_sample)) {
|
|
|
|
+ int err;
|
|
|
|
+
|
|
cur_count = size < sample_left ? size : sample_left;
|
|
cur_count = size < sample_left ? size : sample_left;
|
|
src = (char *)(((long)sbp.sbprof_tbbuf[cur_sample])+sample_off);
|
|
src = (char *)(((long)sbp.sbprof_tbbuf[cur_sample])+sample_off);
|
|
- copy_to_user(dest, src, cur_count);
|
|
|
|
- DBG(printk(DEVNAME ": read from sample %d, %d bytes\n",
|
|
|
|
- cur_sample, cur_count));
|
|
|
|
|
|
+ err = __copy_to_user(dest, src, cur_count);
|
|
|
|
+ if (err) {
|
|
|
|
+ *offp = cur_off + cur_count - err;
|
|
|
|
+ mutex_unlock(&sbp.lock);
|
|
|
|
+ return err;
|
|
|
|
+ }
|
|
|
|
+
|
|
|
|
+ pr_debug(DEVNAME ": read from sample %d, %d bytes\n",
|
|
|
|
+ cur_sample, cur_count);
|
|
size -= cur_count;
|
|
size -= cur_count;
|
|
sample_left -= cur_count;
|
|
sample_left -= cur_count;
|
|
if (!sample_left) {
|
|
if (!sample_left) {
|
|
@@ -339,37 +461,43 @@ static ssize_t sbprof_tb_read(struct file *filp, char *buf,
|
|
dest += cur_count;
|
|
dest += cur_count;
|
|
count += cur_count;
|
|
count += cur_count;
|
|
}
|
|
}
|
|
|
|
+
|
|
*offp = cur_off;
|
|
*offp = cur_off;
|
|
|
|
+ mutex_unlock(&sbp.lock);
|
|
|
|
|
|
return count;
|
|
return count;
|
|
}
|
|
}
|
|
|
|
|
|
-static long sbprof_tb_ioctl(struct file *filp,
|
|
|
|
- unsigned int command,
|
|
|
|
- unsigned long arg)
|
|
|
|
|
|
+static long sbprof_tb_ioctl(struct file *filp, unsigned int command,
|
|
|
|
+ unsigned long arg)
|
|
{
|
|
{
|
|
int error = 0;
|
|
int error = 0;
|
|
|
|
|
|
- lock_kernel();
|
|
|
|
switch (command) {
|
|
switch (command) {
|
|
case SBPROF_ZBSTART:
|
|
case SBPROF_ZBSTART:
|
|
|
|
+ mutex_lock(&sbp.lock);
|
|
error = sbprof_zbprof_start(filp);
|
|
error = sbprof_zbprof_start(filp);
|
|
|
|
+ mutex_unlock(&sbp.lock);
|
|
break;
|
|
break;
|
|
|
|
+
|
|
case SBPROF_ZBSTOP:
|
|
case SBPROF_ZBSTOP:
|
|
|
|
+ mutex_lock(&sbp.lock);
|
|
error = sbprof_zbprof_stop();
|
|
error = sbprof_zbprof_stop();
|
|
|
|
+ mutex_unlock(&sbp.lock);
|
|
break;
|
|
break;
|
|
|
|
+
|
|
case SBPROF_ZBWAITFULL:
|
|
case SBPROF_ZBWAITFULL:
|
|
- DEFINE_WAIT(wait);
|
|
|
|
- prepare_to_wait(&sbp.tb_read, &wait, TASK_INTERRUPTIBLE);
|
|
|
|
- schedule();
|
|
|
|
- finish_wait(&sbp.tb_read, &wait);
|
|
|
|
- /* XXXKW check if interrupted? */
|
|
|
|
- return put_user(TB_FULL, (int *) arg);
|
|
|
|
|
|
+ error = wait_event_interruptible(sbp.tb_read, TB_FULL);
|
|
|
|
+ if (error)
|
|
|
|
+ break;
|
|
|
|
+
|
|
|
|
+ error = put_user(TB_FULL, (int *) arg);
|
|
|
|
+ break;
|
|
|
|
+
|
|
default:
|
|
default:
|
|
error = -EINVAL;
|
|
error = -EINVAL;
|
|
break;
|
|
break;
|
|
}
|
|
}
|
|
- unlock_kernel();
|
|
|
|
|
|
|
|
return error;
|
|
return error;
|
|
}
|
|
}
|
|
@@ -384,23 +512,60 @@ static const struct file_operations sbprof_tb_fops = {
|
|
.mmap = NULL,
|
|
.mmap = NULL,
|
|
};
|
|
};
|
|
|
|
|
|
|
|
+static struct class *tb_class;
|
|
|
|
+static struct device *tb_dev;
|
|
|
|
+
|
|
static int __init sbprof_tb_init(void)
|
|
static int __init sbprof_tb_init(void)
|
|
{
|
|
{
|
|
|
|
+ struct device *dev;
|
|
|
|
+ struct class *tbc;
|
|
|
|
+ int err;
|
|
|
|
+
|
|
if (register_chrdev(SBPROF_TB_MAJOR, DEVNAME, &sbprof_tb_fops)) {
|
|
if (register_chrdev(SBPROF_TB_MAJOR, DEVNAME, &sbprof_tb_fops)) {
|
|
printk(KERN_WARNING DEVNAME ": initialization failed (dev %d)\n",
|
|
printk(KERN_WARNING DEVNAME ": initialization failed (dev %d)\n",
|
|
SBPROF_TB_MAJOR);
|
|
SBPROF_TB_MAJOR);
|
|
return -EIO;
|
|
return -EIO;
|
|
}
|
|
}
|
|
|
|
+
|
|
|
|
+ tbc = class_create(THIS_MODULE, "sb_tracebuffer");
|
|
|
|
+ if (IS_ERR(tbc)) {
|
|
|
|
+ err = PTR_ERR(tbc);
|
|
|
|
+ goto out_chrdev;
|
|
|
|
+ }
|
|
|
|
+
|
|
|
|
+ tb_class = tbc;
|
|
|
|
+
|
|
|
|
+ dev = device_create(tbc, NULL, MKDEV(SBPROF_TB_MAJOR, 0), "tb");
|
|
|
|
+ if (IS_ERR(dev)) {
|
|
|
|
+ err = PTR_ERR(dev);
|
|
|
|
+ goto out_class;
|
|
|
|
+ }
|
|
|
|
+ tb_dev = dev;
|
|
|
|
+
|
|
sbp.open = 0;
|
|
sbp.open = 0;
|
|
tb_period = zbbus_mhz * 10000LL;
|
|
tb_period = zbbus_mhz * 10000LL;
|
|
- printk(KERN_INFO DEVNAME ": initialized - tb_period = %lld\n", tb_period);
|
|
|
|
|
|
+ pr_info(DEVNAME ": initialized - tb_period = %lld\n", tb_period);
|
|
|
|
+
|
|
return 0;
|
|
return 0;
|
|
|
|
+
|
|
|
|
+out_class:
|
|
|
|
+ class_destroy(tb_class);
|
|
|
|
+out_chrdev:
|
|
|
|
+ unregister_chrdev(SBPROF_TB_MAJOR, DEVNAME);
|
|
|
|
+
|
|
|
|
+ return err;
|
|
}
|
|
}
|
|
|
|
|
|
static void __exit sbprof_tb_cleanup(void)
|
|
static void __exit sbprof_tb_cleanup(void)
|
|
{
|
|
{
|
|
|
|
+ device_destroy(tb_class, MKDEV(SBPROF_TB_MAJOR, 0));
|
|
unregister_chrdev(SBPROF_TB_MAJOR, DEVNAME);
|
|
unregister_chrdev(SBPROF_TB_MAJOR, DEVNAME);
|
|
|
|
+ class_destroy(tb_class);
|
|
}
|
|
}
|
|
|
|
|
|
module_init(sbprof_tb_init);
|
|
module_init(sbprof_tb_init);
|
|
module_exit(sbprof_tb_cleanup);
|
|
module_exit(sbprof_tb_cleanup);
|
|
|
|
+
|
|
|
|
+MODULE_ALIAS_CHARDEV_MAJOR(SBPROF_TB_MAJOR);
|
|
|
|
+MODULE_AUTHOR("Ralf Baechle <ralf@linux-mips.org>");
|
|
|
|
+MODULE_LICENSE("GPL");
|