5 files changed, 324 insertions, 0 deletions

diff --git a/arch/ia64/oprofile/Kconfig b/arch/ia64/oprofile/Kconfig
new file mode 100644
index 00000000000..56e6f614b04
--- /dev/null
+++ b/arch/ia64/oprofile/Kconfig
@@ -0,0 +1,26 @@
+
+menu "Profiling support"
+	depends on EXPERIMENTAL
+
+config PROFILING
+	bool "Profiling support (EXPERIMENTAL)"
+	help
+	  Say Y here to enable the extended profiling support mechanisms used
+	  by profilers such as OProfile.
+
+config OPROFILE
+	tristate "OProfile system profiling (EXPERIMENTAL)"
+	depends on PROFILING
+	help
+	  OProfile is a profiling system capable of profiling the
+	  whole system, include the kernel, kernel modules, libraries,
+	  and applications.
+
+	  Due to firmware bugs, you may need to use the "nohalt" boot
+	  option if you're using OProfile with the hardware performance
+	  counters.
+
+	  If unsure, say N.
+
+endmenu
+
diff --git a/arch/ia64/oprofile/Makefile b/arch/ia64/oprofile/Makefile
new file mode 100644
index 00000000000..aad27a718ee
--- /dev/null
+++ b/arch/ia64/oprofile/Makefile
@@ -0,0 +1,10 @@
+obj-$(CONFIG_OPROFILE) += oprofile.o
+
+DRIVER_OBJS := $(addprefix ../../../drivers/oprofile/, \
+		oprof.o cpu_buffer.o buffer_sync.o \
+		event_buffer.o oprofile_files.o \
+		oprofilefs.o oprofile_stats.o \
+		timer_int.o )
+
+oprofile-y := $(DRIVER_OBJS) init.o backtrace.o
+oprofile-$(CONFIG_PERFMON) += perfmon.o
diff --git a/arch/ia64/oprofile/backtrace.c b/arch/ia64/oprofile/backtrace.c
new file mode 100644
index 00000000000..b7dabbfb0d6
--- /dev/null
+++ b/arch/ia64/oprofile/backtrace.c
@@ -0,0 +1,150 @@
+/**
+ * @file backtrace.c
+ *
+ * @remark Copyright 2004 Silicon Graphics Inc.  All Rights Reserved.
+ * @remark Read the file COPYING
+ *
+ * @author Greg Banks <gnb@melbourne.sgi.com>
+ * @author Keith Owens <kaos@melbourne.sgi.com>
+ * Based on work done for the ia64 port of the SGI kernprof patch, which is
+ *    Copyright (c) 2003-2004 Silicon Graphics Inc.  All Rights Reserved.
+ */
+
+#include <linux/oprofile.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <asm/ptrace.h>
+#include <asm/system.h>
+
+/*
+ * For IA64 we need to perform a complex little dance to get both
+ * the struct pt_regs and a synthetic struct switch_stack in place
+ * to allow the unwind code to work.  This dance requires our unwind
+ * using code to be called from a function called from unw_init_running().
+ * There we only get a single void* data pointer, so use this struct
+ * to hold all the data we need during the unwind.
+ */
+typedef struct
+{
+	unsigned int depth;
+	struct pt_regs *regs;
+	struct unw_frame_info frame;
+	u64 *prev_pfs_loc;	/* state for WAR for old spinlock ool code */
+} ia64_backtrace_t;
+
+#if __GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 3)
+/*
+ * Returns non-zero if the PC is in the spinlock contention out-of-line code
+ * with non-standard calling sequence (on older compilers).
+ */
+static __inline__ int in_old_ool_spinlock_code(unsigned long pc)
+{
+	extern const char ia64_spinlock_contention_pre3_4[] __attribute__ ((weak));
+	extern const char ia64_spinlock_contention_pre3_4_end[] __attribute__ ((weak));
+	unsigned long sc_start = (unsigned long)ia64_spinlock_contention_pre3_4;
+	unsigned long sc_end = (unsigned long)ia64_spinlock_contention_pre3_4_end;
+	return (sc_start && sc_end && pc >= sc_start && pc < sc_end);
+}
+#else
+/* Newer spinlock code does a proper br.call and works fine with the unwinder */
+#define in_old_ool_spinlock_code(pc)	0
+#endif
+
+/* Returns non-zero if the PC is in the Interrupt Vector Table */
+static __inline__ int in_ivt_code(unsigned long pc)
+{
+	extern char ia64_ivt[];
+	return (pc >= (u_long)ia64_ivt && pc < (u_long)ia64_ivt+32768);
+}
+
+/*
+ * Unwind to next stack frame.
+ */
+static __inline__ int next_frame(ia64_backtrace_t *bt)
+{
+	/*
+	 * Avoid unsightly console message from unw_unwind() when attempting
+	 * to unwind through the Interrupt Vector Table which has no unwind
+	 * information.
+	 */
+	if (in_ivt_code(bt->frame.ip))
+		return 0;
+
+	/*
+	 * WAR for spinlock contention from leaf functions.  ia64_spinlock_contention_pre3_4
+	 * has ar.pfs == r0.  Leaf functions do not modify ar.pfs so ar.pfs remains
+	 * as 0, stopping the backtrace.  Record the previous ar.pfs when the current
+	 * IP is in ia64_spinlock_contention_pre3_4 then unwind, if pfs_loc has not changed
+	 * after unwind then use pt_regs.ar_pfs which is where the real ar.pfs is for
+	 * leaf functions.
+	 */
+	if (bt->prev_pfs_loc && bt->regs && bt->frame.pfs_loc == bt->prev_pfs_loc)
+		bt->frame.pfs_loc = &bt->regs->ar_pfs;
+	bt->prev_pfs_loc = (in_old_ool_spinlock_code(bt->frame.ip) ? bt->frame.pfs_loc : NULL);
+
+	return unw_unwind(&bt->frame) == 0;
+}
+
+
+static void do_ia64_backtrace(struct unw_frame_info *info, void *vdata)
+{
+	ia64_backtrace_t *bt = vdata;
+	struct switch_stack *sw;
+	int count = 0;
+	u_long pc, sp;
+
+	sw = (struct switch_stack *)(info+1);
+	/* padding from unw_init_running */
+	sw = (struct switch_stack *)(((unsigned long)sw + 15) & ~15);
+
+	unw_init_frame_info(&bt->frame, current, sw);
+
+	/* skip over interrupt frame and oprofile calls */
+	do {
+		unw_get_sp(&bt->frame, &sp);
+		if (sp >= (u_long)bt->regs)
+			break;
+		if (!next_frame(bt))
+			return;
+	} while (count++ < 200);
+
+	/* finally, grab the actual sample */
+	while (bt->depth-- && next_frame(bt)) {
+		unw_get_ip(&bt->frame, &pc);
+		oprofile_add_trace(pc);
+		if (unw_is_intr_frame(&bt->frame)) {
+			/*
+			 * Interrupt received on kernel stack; this can
+			 * happen when timer interrupt fires while processing
+			 * a softirq from the tail end of a hardware interrupt
+			 * which interrupted a system call.  Don't laugh, it
+			 * happens!  Splice the backtrace into two parts to
+			 * avoid spurious cycles in the gprof output.
+			 */
+			/* TODO: split rather than drop the 2nd half */
+			break;
+		}
+	}
+}
+
+void
+ia64_backtrace(struct pt_regs * const regs, unsigned int depth)
+{
+	ia64_backtrace_t bt;
+	unsigned long flags;
+
+	/*
+	 * On IA64 there is little hope of getting backtraces from
+	 * user space programs -- the problems of getting the unwind
+	 * information from arbitrary user programs are extreme.
+	 */
+	if (user_mode(regs))
+		return;
+
+	bt.depth = depth;
+	bt.regs = regs;
+	bt.prev_pfs_loc = NULL;
+	local_irq_save(flags);
+	unw_init_running(do_ia64_backtrace, &bt);
+	local_irq_restore(flags);
+}
diff --git a/arch/ia64/oprofile/init.c b/arch/ia64/oprofile/init.c
new file mode 100644
index 00000000000..125a602a660
--- /dev/null
+++ b/arch/ia64/oprofile/init.c
@@ -0,0 +1,38 @@
+/**
+ * @file init.c
+ *
+ * @remark Copyright 2002 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon <levon@movementarian.org>
+ */
+
+#include <linux/kernel.h>
+#include <linux/oprofile.h>
+#include <linux/init.h>
+#include <linux/errno.h>
+ 
+extern int perfmon_init(struct oprofile_operations * ops);
+extern void perfmon_exit(void);
+extern void ia64_backtrace(struct pt_regs * const regs, unsigned int depth);
+
+int __init oprofile_arch_init(struct oprofile_operations * ops)
+{
+	int ret = -ENODEV;
+
+#ifdef CONFIG_PERFMON
+	/* perfmon_init() can fail, but we have no way to report it */
+	ret = perfmon_init(ops);
+#endif
+	ops->backtrace = ia64_backtrace;
+
+	return ret;
+}
+
+
+void oprofile_arch_exit(void)
+{
+#ifdef CONFIG_PERFMON
+	perfmon_exit();
+#endif
+}
diff --git a/arch/ia64/oprofile/perfmon.c b/arch/ia64/oprofile/perfmon.c
new file mode 100644
index 00000000000..b7975a469fb
--- /dev/null
+++ b/arch/ia64/oprofile/perfmon.c
@@ -0,0 +1,100 @@
+/**
+ * @file perfmon.c
+ *
+ * @remark Copyright 2003 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon <levon@movementarian.org>
+ */
+
+#include <linux/kernel.h>
+#include <linux/config.h>
+#include <linux/oprofile.h>
+#include <linux/sched.h>
+#include <asm/perfmon.h>
+#include <asm/ptrace.h>
+#include <asm/errno.h>
+
+static int allow_ints;
+
+static int
+perfmon_handler(struct task_struct *task, void *buf, pfm_ovfl_arg_t *arg,
+                struct pt_regs *regs, unsigned long stamp)
+{
+	int event = arg->pmd_eventid;
+ 
+	arg->ovfl_ctrl.bits.reset_ovfl_pmds = 1;
+
+	/* the owner of the oprofile event buffer may have exited
+	 * without perfmon being shutdown (e.g. SIGSEGV)
+	 */
+	if (allow_ints)
+		oprofile_add_sample(regs, event);
+	return 0;
+}
+
+
+static int perfmon_start(void)
+{
+	allow_ints = 1;
+	return 0;
+}
+
+
+static void perfmon_stop(void)
+{
+	allow_ints = 0;
+}
+
+
+#define OPROFILE_FMT_UUID { \
+	0x77, 0x7a, 0x6e, 0x61, 0x20, 0x65, 0x73, 0x69, 0x74, 0x6e, 0x72, 0x20, 0x61, 0x65, 0x0a, 0x6c }
+
+static pfm_buffer_fmt_t oprofile_fmt = {
+ 	.fmt_name 	    = "oprofile_format",
+ 	.fmt_uuid	    = OPROFILE_FMT_UUID,
+ 	.fmt_handler	    = perfmon_handler,
+};
+
+
+static char * get_cpu_type(void)
+{
+	__u8 family = local_cpu_data->family;
+
+	switch (family) {
+		case 0x07:
+			return "ia64/itanium";
+		case 0x1f:
+			return "ia64/itanium2";
+		default:
+			return "ia64/ia64";
+	}
+}
+
+
+/* all the ops are handled via userspace for IA64 perfmon */
+
+static int using_perfmon;
+
+int perfmon_init(struct oprofile_operations * ops)
+{
+	int ret = pfm_register_buffer_fmt(&oprofile_fmt);
+	if (ret)
+		return -ENODEV;
+
+	ops->cpu_type = get_cpu_type();
+	ops->start = perfmon_start;
+	ops->stop = perfmon_stop;
+	using_perfmon = 1;
+	printk(KERN_INFO "oprofile: using perfmon.\n");
+	return 0;
+}
+
+
+void perfmon_exit(void)
+{
+	if (!using_perfmon)
+		return;
+
+	pfm_unregister_buffer_fmt(oprofile_fmt.fmt_uuid);
+}