[PATCH 1/5] OProfile update

[PATCH 1/5] OProfile update

[John Levon]

The five patches here change the following files :

 Documentation/kernel-parameters.txt |    3 ++
 drivers/oprofile/buffer_sync.c      |   49 ++++++++++++++++++++++++++----------
 drivers/oprofile/cpu_buffer.c       |    6 +---
 drivers/oprofile/event_buffer.h     |    1
 drivers/oprofile/oprof.c            |   25 +++++++++++++-----
 fs/dcookies.c                       |    6 ++++
 6 files changed, 67 insertions(+), 23 deletions(-)

please apply
john


My previous fix was incomplete, we could get the same thing happening
on the init-failure path. Fix that.

diff -Naur -X dontdiff linux-cvs/drivers/oprofile/buffer_sync.c linux-me/drivers/oprofile/buffer_sync.c
--- linux-cvs/drivers/oprofile/buffer_sync.c	2003-05-26 03:20:20.000000000 +0100
+++ linux-me/drivers/oprofile/buffer_sync.c	2003-05-26 04:25:15.000000000 +0100
@@ -127,6 +127,14 @@
 };
 
  
+static void end_sync_timer(void)
+{
+	del_timer_sync(&sync_timer);
+	/* timer might have queued work, make sure it's completed. */
+	flush_scheduled_work();
+}
+
+
 int sync_start(void)
 {
 	int err;
@@ -158,7 +166,7 @@
 out2:
 	profile_event_unregister(EXIT_TASK, &exit_task_nb);
 out1:
-	del_timer_sync(&sync_timer);
+	end_sync_timer();
 	goto out;
 }
 
@@ -169,9 +177,7 @@
 	profile_event_unregister(EXIT_TASK, &exit_task_nb);
 	profile_event_unregister(EXIT_MMAP, &exit_mmap_nb);
 	profile_event_unregister(EXEC_UNMAP, &exec_unmap_nb);
-	del_timer_sync(&sync_timer);
-	/* timer might have queued work, make sure it's completed. */
-	flush_scheduled_work();
+	end_sync_timer();
 }
 
  

[PATCH 2/5] OProfile update

[John Levon]

The code that attempts to reset last_task and in_kernel has a race
against samples appearing during the handling of the buffers, that
causes a small number of mis-attribution of samples. Closing the
window is non-obvious, and not worth it, so we just make it smaller.
Even without the patch, there seem to be few such "bad" samples
because its effects are mitigated on a switch into userspace or
a task switch.

diff -Naur -X dontdiff linux-cvs/drivers/oprofile/buffer_sync.c linux-me/drivers/oprofile/buffer_sync.c
--- linux-cvs/drivers/oprofile/buffer_sync.c	2003-05-26 03:20:20.000000000 +0100
+++ linux-me/drivers/oprofile/buffer_sync.c	2003-05-26 04:25:15.000000000 +0100
@@ -366,12 +377,26 @@
 }
  
 
-/* compute number of filled slots in cpu_buffer queue */
-static unsigned long nr_filled_slots(struct oprofile_cpu_buffer * b)
+/* "acquire" as many cpu buffer slots as we can */
+static unsigned long get_slots(struct oprofile_cpu_buffer * b)
 {
 	unsigned long head = b->head_pos;
 	unsigned long tail = b->tail_pos;
 
+	/*
+	 * Subtle. This resets the persistent last_task
+	 * and in_kernel values used for switching notes.
+	 * BUT, there is a small window between reading
+	 * head_pos, and this call, that means samples
+	 * can appear at the new head position, but not
+	 * be prefixed with the notes for switching
+	 * kernel mode or a task switch. This small hole
+	 * can lead to mis-attribution or samples where
+	 * we don't know if it's in the kernel or not,
+	 * at the start of an event buffer.
+	 */
+	cpu_buffer_reset(b);
+
 	if (head >= tail)
 		return head - tail;
 
@@ -408,9 +433,9 @@
  
 	/* Remember, only we can modify tail_pos */
 
-	unsigned long const available_elements = nr_filled_slots(cpu_buf);
+	unsigned long const available = get_slots(cpu_buf);
   
-	for (i=0; i < available_elements; ++i) {
+	for (i=0; i < available; ++i) {
 		struct op_sample * s = &cpu_buf->buffer[cpu_buf->tail_pos];
  
 		if (is_ctx_switch(s->eip)) {
@@ -435,8 +460,6 @@
 		increment_tail(cpu_buf);
 	}
 	release_mm(mm);
-
-	cpu_buffer_reset(cpu_buf);
 }
  
  

[PATCH 5/5] OProfile update

[John Levon]

d_path() can return -ENAMETOOLONG these days. Pass it upstream.

diff -Naur -X dontdiff linux-cvs/fs/dcookies.c linux-me/fs/dcookies.c
--- linux-cvs/fs/dcookies.c	2003-05-19 19:57:28.000000000 +0100
+++ linux-me/fs/dcookies.c	2003-05-19 20:05:16.000000000 +0100
@@ -175,6 +175,11 @@
 	/* FIXME: (deleted) ? */
 	path = d_path(dcs->dentry, dcs->vfsmnt, kbuf, PAGE_SIZE);
 
+	if (IS_ERR(path)) {
+		err = PTR_ERR(path);
+		goto out_free;
+	}
+
 	err = -ERANGE;
  
 	pathlen = kbuf + PAGE_SIZE - path;
@@ -184,6 +189,7 @@
 			err = -EFAULT;
 	}
 
+out_free:
 	kfree(kbuf);
 out:
 	up(&dcookie_sem);

[PATCH 4/5] OProfile update

[John Levon]

Fix a stale comment.

diff -Naur -X dontdiff linux-cvs/drivers/oprofile/cpu_buffer.c linux-me/drivers/oprofile/cpu_buffer.c
--- linux-cvs/drivers/oprofile/cpu_buffer.c	2003-05-26 03:20:20.000000000 +0100
+++ linux-me/drivers/oprofile/cpu_buffer.c	2003-05-26 03:24:38.000000000 +0100
@@ -168,10 +168,8 @@
 }
 
 
-/* resets the cpu buffer to a sane state - should be called with 
- * cpu_buf->int_lock held
- */
-void cpu_buffer_reset(struct oprofile_cpu_buffer *cpu_buf)
+/* Resets the cpu buffer to a sane state. */
+void cpu_buffer_reset(struct oprofile_cpu_buffer * cpu_buf)
 {
 	/* reset these to invalid values; the next sample
 	 * collected will populate the buffer with proper

[PATCH 3/5] OProfile update

[John Levon]

A patch mostly by Will Cohen, adding a parameter to OProfile to
over-ride use of the perfctr hardware. Useful for testing and
a host of other things.

diff -Naur -X dontdiff linux-cvs/Documentation/kernel-parameters.txt linux-me/Documentation/kernel-parameters.txt
--- linux-cvs/Documentation/kernel-parameters.txt	2003-03-30 21:12:29.000000000 +0100
+++ linux-me/Documentation/kernel-parameters.txt	2003-05-26 04:53:57.000000000 +0100
@@ -652,6 +652,9 @@
 	opl3sa2=	[HW,OSS]
 			Format: <io>,<irq>,<dma>,<dma2>,<mss_io>,<mpu_io>,<ymode>,<loopback>[,<isapnp>,<multiple]
  
+	oprofile.timer=	[HW]
+			Use timer interrupt instead of performance counters
+
 	optcd=		[HW,CD]
 			Format: <io>
 
diff -Naur -X dontdiff linux-cvs/drivers/oprofile/oprof.c linux-me/drivers/oprofile/oprof.c
--- linux-cvs/drivers/oprofile/oprof.c	2003-05-10 17:30:28.000000000 +0100
+++ linux-me/drivers/oprofile/oprof.c	2003-05-26 04:52:51.000000000 +0100
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/oprofile.h>
+#include <linux/moduleparam.h>
 #include <asm/semaphore.h>
 
 #include "oprof.h"
@@ -24,6 +25,12 @@
 static unsigned long is_setup;
 static DECLARE_MUTEX(start_sem);
 
+/* timer
+   0 - use performance monitoring hardware if available
+   1 - use the timer int mechanism regardless
+ */
+static int timer = 0;
+
 int oprofile_setup(void)
 {
 	int err;
@@ -124,13 +131,16 @@
 
 static int __init oprofile_init(void)
 {
-	int err;
+	int err = -ENODEV;
+
+	if (!timer) {
+		/* Architecture must fill in the interrupt ops and the
+		 * logical CPU type, or we can fall back to the timer
+		 * interrupt profiler.
+		 */
+		err = oprofile_arch_init(&oprofile_ops);
+	}
 
-	/* Architecture must fill in the interrupt ops and the
-	 * logical CPU type, or we can fall back to the timer
-	 * interrupt profiler.
-	 */
-	err = oprofile_arch_init(&oprofile_ops);
 	if (err == -ENODEV) {
 		timer_init(&oprofile_ops);
 		err = 0;
@@ -163,6 +173,9 @@
  
 module_init(oprofile_init);
 module_exit(oprofile_exit);
+
+module_param_named(timer, timer, int, 0644);
+MODULE_PARM_DESC(timer, "force use of timer interrupt");
  
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("John Levon <levon@movementarian.org>");