Skip to contentSkip to conversation

patch for ia64 oprofile support for 2.5.67 kernel

Linux-ia64 Will Cohen 3 participants 4 messages May 5, 2003 – May 20, 2003

lore
Share & tools
lore

Will Cohen

author
#0 lore raw

I started with the most recent 2.5 kernel for ia64 I could find, the stock 2.5.67 kernel from kernel.org with the linuxia64.org patch (linux-2.5.67-ia64-030416.diff.bz2). Attached is a patch that provides oprofile kernel support that works oprofile 0.5.2. It is built on top of the perfmon support. For the oprofile to use the performance monitoring hardware CONFIG_PERFMON also needs to be enabled in .config. If perfmon support is not enabled, oprofile will just supply the basic TIMER_INT implementation.

The patch applies cleanly to the 2.5.67 kernel with the linuxia64.org patchs. I have built the kernel, with required code compiled in (no modules) and booted the resulting kernel successfully on an Intanium 2 SMP machine. I am sure some revison is required for the patch for the 2.5.69 kernel. There was some refactoring of the code TIMER_INT code.

Comments?

-Will

arch/ia64/kernel/time.c renamed from arch/ia64/kernel/time.c.orig +3 -31
@@ -17,6 +17,7 @@
 #include <linux/time.h>
 #include <linux/interrupt.h>
 #include <linux/efi.h>
+#include <linux/profile.h>
 
 #include <asm/delay.h>
 #include <asm/hw_irq.h>
@@ -35,30 +36,6 @@
 
 #endif
 
-static void
-do_profile (unsigned long ip)
-{
-	extern unsigned long prof_cpu_mask;
-	extern char _stext;
-
-	if (!prof_buffer)
-		return;
-
-	if (!((1UL << smp_processor_id()) & prof_cpu_mask))
-		return;
-
-	ip -= (unsigned long) &_stext;
-	ip >>= prof_shift;
-	/*
-	 * Don't ignore out-of-bounds IP values silently, put them into the last
-	 * histogram slot, so if present, they will show up as a sharp peak.
-	 */
-	if (ip > prof_len - 1)
-		ip = prof_len - 1;
-
-	atomic_inc((atomic_t *) &prof_buffer[ip]);
-}
-
 /*
  * Return the number of nano-seconds that elapsed since the last update to jiffy.  The
  * xtime_lock must be at least read-locked when calling this routine.
@@ -186,14 +163,9 @@
 		printk(KERN_ERR "Oops: timer tick before it's due (itc=%lx,itm=%lx)\n",
 		       ia64_get_itc(), new_itm);
 
+	ia64_do_profile(regs);
+
 	while (1) {
-		/*
-		 * Do kernel PC profiling here.  We multiply the instruction number by
-		 * four so that we can use a prof_shift of 2 to get instruction-level
-		 * instead of just bundle-level accuracy.
-		 */
-		if (!user_mode(regs))
-			do_profile(regs->cr_iip + 4*ia64_psr(regs)->ri);
 
 #ifdef CONFIG_SMP
 		smp_do_timer(regs);
arch/ia64/kernel/perfmon.c renamed from arch/ia64/kernel/perfmon.c.orig +0 -3
@@ -132,9 +132,6 @@
 
 #define PFM_REG_RETFLAG_SET(flags, val)	do { flags &= ~PFM_REG_RETFL_MASK; flags |= (val); } while(0)
 
-#define PFM_CPUINFO_CLEAR(v)	__get_cpu_var(pfm_syst_info) &= ~(v)
-#define PFM_CPUINFO_SET(v)	__get_cpu_var(pfm_syst_info) |= (v)
-
 /*
  * debugging
  */
arch/ia64/Makefile renamed from arch/ia64/Makefile.orig +2 -0
@@ -56,6 +56,8 @@
 drivers-$(CONFIG_IA64_HP_SIM)	+= arch/ia64/hp/sim/
 drivers-$(CONFIG_IA64_HP_ZX1)	+= arch/ia64/hp/common/ arch/ia64/hp/zx1/
 drivers-$(CONFIG_IA64_GENERIC)	+= arch/ia64/hp/common/ arch/ia64/hp/zx1/ arch/ia64/hp/sim/
+# must be linked after kernel/
+drivers-$(CONFIG_OPROFILE)	+= arch/ia64/oprofile/
 
 boot := arch/ia64/boot
 tools := arch/ia64/tools
arch/ia64/oprofile/Kconfig renamed from arch/ia64/oprofile/Kconfig.orig +23 -0
@@ -0,0 +1,23 @@
+
+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.
+
+	  If unsure, say N.
+
+endmenu
+
arch/ia64/oprofile/init.c renamed from arch/ia64/oprofile/init.c.orig +29 -0
@@ -0,0 +1,29 @@
+/**
+ * @file init.c
+ *
+ * @remark Copyright 2002-2003 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon <[email protected]>
+ * @author Will Cohen <[email protected]>
+ */
+
+#include <linux/kernel.h>
+#include <linux/oprofile.h>
+#include <linux/init.h>
+ 
+/* We support CPUs that have performance counters like the IA64
+ * with irq mode samples.
+ */
+ 
+extern int irq_init(struct oprofile_operations ** ops);
+extern void timer_init(struct oprofile_operations ** ops);
+
+int __init oprofile_arch_init(struct oprofile_operations ** ops)
+{
+#ifdef CONFIG_PERFMON
+	if (!irq_init(ops))
+#endif
+		timer_init(ops);
+	return 0;
+}
arch/ia64/oprofile/irq_int.c renamed from arch/ia64/oprofile/irq_int.c.orig +253 -0

253 lines not on this page · raw message

arch/ia64/oprofile/Makefile renamed from arch/ia64/oprofile/Makefile.orig +11 -0
@@ -0,0 +1,11 @@
+# arch/ia64/oprofile/Makefile
+
+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 )
+
+oprofile-y				:= $(DRIVER_OBJS) init.o timer_int.o
+oprofile-$(CONFIG_PERFMON)	 	+= irq_int.o op_model_ia64.o
arch/ia64/oprofile/op_counter.h renamed from arch/ia64/oprofile/op_counter.h.orig +29 -0
@@ -0,0 +1,29 @@
+/**
+ * @file op_counter.h
+ *
+ * @remark Copyright 2002 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon
+ */
+ 
+#ifndef OP_COUNTER_H
+#define OP_COUNTER_H
+ 
+#define OP_MAX_COUNTER 8
+ 
+/* Per-perfctr configuration as set via
+ * oprofilefs.
+ */
+struct op_counter_config {
+        unsigned long count;
+        unsigned long enabled;
+        unsigned long event;
+        unsigned long kernel;
+        unsigned long user;
+        unsigned long unit_mask;
+};
+
+extern struct op_counter_config counter_config[];
+
+#endif /* OP_COUNTER_H */
arch/ia64/oprofile/op_ia64_model.h renamed from arch/ia64/oprofile/op_ia64_model.h.orig +80 -0

80 lines not on this page · raw message

arch/ia64/oprofile/op_model_ia64.c renamed from arch/ia64/oprofile/op_model_ia64.c.orig +210 -0

210 lines not on this page · raw message

arch/ia64/oprofile/timer_int.c renamed from arch/ia64/oprofile/timer_int.c.orig +57 -0

57 lines not on this page · raw message

arch/ia64/Kconfig renamed from arch/ia64/Kconfig.orig +1 -0
@@ -782,6 +782,7 @@
 
 source "arch/ia64/hp/sim/Kconfig"
 
+source "arch/ia64/oprofile/Kconfig"
 
 menu "Kernel hacking"
 
include/asm-ia64/hw_irq.h renamed from include/asm-ia64/hw_irq.h.orig +43 -0

53 lines not on this page · raw message

include/asm-ia64/perfmon.h renamed from include/asm-ia64/perfmon.h.orig +3 -0
@@ -199,6 +199,9 @@
 #define PFM_CPUINFO_EXCL_IDLE	0x4	/* the system wide session excludes the idle task */
 
 
+#define PFM_CPUINFO_CLEAR(v)	__get_cpu_var(pfm_syst_info) &= ~(v)
+#define PFM_CPUINFO_SET(v)	__get_cpu_var(pfm_syst_info) |= (v)
+
 #endif /* __KERNEL__ */
 
 #endif /* _ASM_IA64_PERFMON_H */
include/asm-ia64/ptrace.h renamed from include/asm-ia64/ptrace.h.orig +1 -0
@@ -214,6 +214,7 @@
 };
 
 #ifdef __KERNEL__
+#define instruction_pointer(regs) ((regs)->cr_iip)
   /* given a pointer to a task_struct, return the user's pt_regs */
 # define ia64_task_regs(t)		(((struct pt_regs *) ((char *) (t) + IA64_STK_OFFSET)) - 1)
 # define ia64_psr(regs)			((struct ia64_psr *) &(regs)->cr_ipsr)
#1 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line · Will Cohen

Bill> Comments?

The patch looks fine for 2.4 purposes, but for 2.5/2.6, I'd much prefer to see all PMU-related manipulation done in user-space (through the perfmon syscall and/or pfmlib). I really don't want to have to maintain _two_ subsystems that need to be updated with each new processor.

I talked to Stephane about this and he said he should have the "custom sampling buffer format" patch ready in about a week's time. Once that is in, you should be able to hook oprofile to perfmon with a trivial glue "module".

	--david
author
#2 lore raw
In reply to Will Cohen

How are things going on the "custom sampling buffer format" patch? Is there somthing availble I can look at this time?

-Will

Quoted 17 lines · David Mosberger
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 1 line · Will Cohen

Bill> Comments?

The patch looks fine for 2.4 purposes, but for 2.5/2.6, I'd much prefer to see all PMU-related manipulation done in user-space (through the perfmon syscall and/or pfmlib). I really don't want to have to maintain _two_ subsystems that need to be updated with each new processor.

I talked to Stephane about this and he said he should have the "custom sampling buffer format" patch ready in about a week's time. Once that is in, you should be able to hook oprofile to perfmon with a trivial glue "module".

	--david
#3 lore raw
In reply to Will Cohen

Will,

Quoted 3 lines · Will Cohen

How are things going on the "custom sampling buffer format" patch? Is there somthing availble I can look at this time?

It is in fairly good shape for 2.5.69. I hope to get this to David for the next kernel patch (2.5.70 or higher).

Can you wait that long?

Signature · 1 line
-Stephane