Skip to contentSkip to conversation

tracing: Updates for 4.5 (and some stuff going into 4.4)

PATCH In review Steven Rostedt 2 participants 17 messages Nov 25, 2015 – Feb 18, 2016

Subsystems (4)
lore
Share & tools
lore

Steven Rostedt

author maintainer
#0 lore raw

cover tracing: Updates for 4.5 (and some stuff going into 4.4)

git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace.git
for-next

Head SHA1: ffaea3b8472da0f0e5a2b4fc9ccefb1ff460e824

Steven Rostedt (Red Hat) (6):
ring-buffer: Update read stamp with first real commit on page
ring-buffer: Put back the length if crossed page with add_timestamp
ftracetest: Add instance create and delete test
ring-buffer: Use READ_ONCE() for most tail_page access
ring-buffer: Remove redundant update of page timestamp
ring-buffer: Process commits whenever moving to a new page.

Tom Zanussi (1):
tracing: Update cond flag when enabling or disabling a trigger

----

 kernel/module.c                                    | 18 ++++-
 kernel/trace/ring_buffer.c                         | 74 +++++++++---------
 kernel/trace/trace_events_trigger.c                | 10 ++-
 .../selftests/ftrace/test.d/instances/instance.tc  | 90 ++++++++++++++++++++++
 4 files changed, 147 insertions(+), 45 deletions(-)
 create mode 100644 tools/testing/selftests/ftrace/test.d/instances/instance.tc
author maintainer
#1 lore raw

PATCH 1/7 ring-buffer: Update read stamp with first real commit on page

From: "Steven Rostedt (Red Hat)" <[email protected]>

Do not update the read stamp after swapping out the reader page from the write buffer. If the reader page is swapped out of the buffer before an event is written to it, then the read_stamp may get an out of date timestamp, as the page timestamp is updated on the first commit to that page.

rb_get_reader_page() only returns a page if it has an event on it, otherwise it will return NULL. At that point, check if the page being returned has events and has not been read yet. Then at that point update the read_stamp to match the time stamp of the reader page.

---
 kernel/trace/ring_buffer.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)
kernel/trace/ring_buffer.c +5 -7
index 75f1d05ea82d..4dd6d5bc4e11 100644
@@ -1887,12 +1887,6 @@ rb_event_index(struct ring_buffer_event *event)
 	return (addr & ~PAGE_MASK) - BUF_PAGE_HDR_SIZE;
 }
 
-static void rb_reset_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
-{
-	cpu_buffer->read_stamp = cpu_buffer->reader_page->page->time_stamp;
-	cpu_buffer->reader_page->read = 0;
-}
-
 static void rb_inc_iter(struct ring_buffer_iter *iter)
 {
 	struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
@@ -3626,7 +3620,7 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 
 	/* Finally update the reader page to the new head */
 	cpu_buffer->reader_page = reader;
-	rb_reset_reader_page(cpu_buffer);
+	cpu_buffer->reader_page->read = 0;
 
 	if (overwrite != cpu_buffer->last_overrun) {
 		cpu_buffer->lost_events = overwrite - cpu_buffer->last_overrun;
@@ -3636,6 +3630,10 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 	goto again;
 
  out:
+	/* Update the read_stamp on the first event */
+	if (reader && reader->read == 0)
+		cpu_buffer->read_stamp = reader->page->time_stamp;
+
 	arch_spin_unlock(&cpu_buffer->lock);
 	local_irq_restore(flags);
 
Signature · 1 line
2.6.1
author maintainer
#2 lore raw

PATCH 2/7 ring-buffer: Put back the length if crossed page with add_timestamp

In reply to Steven Rostedt

From: "Steven Rostedt (Red Hat)" <[email protected]>

Commit fcc742eaad7c "ring-buffer: Add event descriptor to simplify passing data" added a descriptor that holds various data instead of passing around several variables through parameters. The problem was that one of the parameters was modified in a function and the code was designed not to have an effect on that modified parameter. Now that the parameter is a descriptor and any modifications to it are non-volatile, the size of the data could be unnecessarily expanded.

Remove the extra space added if a timestamp was added and the event went across the page.

Fixes fcc742eaad7c "ring-buffer: Add event descriptor to simplify passing data"
---
 kernel/trace/ring_buffer.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/trace/ring_buffer.c +4 -1
index 4dd6d5bc4e11..9c6045a27ba3 100644
@@ -2797,8 +2797,11 @@ rb_reserve_next_event(struct ring_buffer *buffer,
 
 	event = __rb_reserve_next(cpu_buffer, &info);
 
-	if (unlikely(PTR_ERR(event) == -EAGAIN))
+	if (unlikely(PTR_ERR(event) == -EAGAIN)) {
+		if (info.add_timestamp)
+			info.length -= RB_LEN_TIME_EXTEND;
 		goto again;
+	}
 
 	if (!event)
 		goto out_fail;
Signature · 1 line
2.6.1
author maintainer
#3 lore raw

PATCH 3/7 ftracetest: Add instance create and delete test

In reply to Steven Rostedt

From: "Steven Rostedt (Red Hat)" <[email protected]>

Create a test to test instance creation and deletion. Several tasks are created that create 3 directories and delete them. The tasks all create the same directories. This places a stress on the code that creates and deletes instances.

---
 .../selftests/ftrace/test.d/instances/instance.tc  | 90 ++++++++++++++++++++++
 1 file changed, 90 insertions(+)
 create mode 100644 tools/testing/selftests/ftrace/test.d/instances/instance.tc
tools/testing/selftests/ftrace/test.d/instances/instance.tc new file +90 -0
new file mode 100644
index 000000000000..773e276ff90b
@@ -0,0 +1,90 @@
+#!/bin/sh
+# description: Test creation and deletion of trace instances
+
+if [ ! -d instances ] ; then
+    echo "no instance directory with this kernel"
+    exit_unsupported;
+fi
+
+fail() { # mesg
+    rmdir x y z 2>/dev/null
+    echo $1
+    set -e
+    exit $FAIL
+}
+
+cd instances
+
+# we don't want to fail on error
+set +e
+
+mkdir x
+rmdir x
+result=$?
+
+if [ $result -ne 0 ]; then
+    echo "instance rmdir not supported"
+    exit_unsupported
+fi
+
+instance_slam() {
+    while :; do
+	mkdir x
+	mkdir y
+	mkdir z
+	rmdir x
+	rmdir y
+	rmdir z
+    done 2>/dev/null
+}
+
+instance_slam &
+x=`jobs -l`
+p1=`echo $x | cut -d' ' -f2`
+echo $p1
+
+instance_slam &
+x=`jobs -l | tail -1`
+p2=`echo $x | cut -d' ' -f2`
+echo $p2
+
+instance_slam &
+x=`jobs -l | tail -1`
+p3=`echo $x | cut -d' ' -f2`
+echo $p3
+
+instance_slam &
+x=`jobs -l | tail -1`
+p4=`echo $x | cut -d' ' -f2`
+echo $p4
+
+instance_slam &
+x=`jobs -l | tail -1`
+p5=`echo $x | cut -d' ' -f2`
+echo $p5
+
+ls -lR >/dev/null
+sleep 1
+
+kill -1 $p1
+kill -1 $p2
+kill -1 $p3
+kill -1 $p4
+kill -1 $p5
+
+echo "Wait for processes to finish"
+wait $p1 $p2 $p3 $p4 $p5
+echo "all processes finished, wait for cleanup"
+
+mkdir x y z
+ls x y z
+rmdir x y z
+for d in x y z; do
+        if [ -d $d ]; then
+                fail "instance $d still exists"
+        fi
+done
+
+set -e
+
+exit 0
Signature · 1 line
2.6.1
author maintainer
#4 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

Shuah,

Sorry, I forgot to add you to the Cc of this patch. I have another test to add, but I forgot to commit it into git ;-)

-- Steve

Quoted 109 lines · On Wed, 25 Nov 2015 11:08:41 -0500 Steven Rostedt <[email protected]> wrote:

From: "Steven Rostedt (Red Hat)" <[email protected]>

Create a test to test instance creation and deletion. Several tasks are created that create 3 directories and delete them. The tasks all create the same directories. This places a stress on the code that creates and deletes instances.

---
 .../selftests/ftrace/test.d/instances/instance.tc  | 90 ++++++++++++++++++++++
 1 file changed, 90 insertions(+)
 create mode 100644 tools/testing/selftests/ftrace/test.d/instances/instance.tc
tools/testing/selftests/ftrace/test.d/instances/instance.tc new file +90 -0
new file mode 100644
index 000000000000..773e276ff90b
@@ -0,0 +1,90 @@
+#!/bin/sh
+# description: Test creation and deletion of trace instances
+
+if [ ! -d instances ] ; then
+    echo "no instance directory with this kernel"
+    exit_unsupported;
+fi
+
+fail() { # mesg
+    rmdir x y z 2>/dev/null
+    echo $1
+    set -e
+    exit $FAIL
+}
+
+cd instances
+
+# we don't want to fail on error
+set +e
+
+mkdir x
+rmdir x
+result=$?
+
+if [ $result -ne 0 ]; then
+    echo "instance rmdir not supported"
+    exit_unsupported
+fi
+
+instance_slam() {
+    while :; do
+	mkdir x
+	mkdir y
+	mkdir z
+	rmdir x
+	rmdir y
+	rmdir z
+    done 2>/dev/null
+}
+
+instance_slam &
+x=`jobs -l`
+p1=`echo $x | cut -d' ' -f2`
+echo $p1
+
+instance_slam &
+x=`jobs -l | tail -1`
+p2=`echo $x | cut -d' ' -f2`
+echo $p2
+
+instance_slam &
+x=`jobs -l | tail -1`
+p3=`echo $x | cut -d' ' -f2`
+echo $p3
+
+instance_slam &
+x=`jobs -l | tail -1`
+p4=`echo $x | cut -d' ' -f2`
+echo $p4
+
+instance_slam &
+x=`jobs -l | tail -1`
+p5=`echo $x | cut -d' ' -f2`
+echo $p5
+
+ls -lR >/dev/null
+sleep 1
+
+kill -1 $p1
+kill -1 $p2
+kill -1 $p3
+kill -1 $p4
+kill -1 $p5
+
+echo "Wait for processes to finish"
+wait $p1 $p2 $p3 $p4 $p5
+echo "all processes finished, wait for cleanup"
+
+mkdir x y z
+ls x y z
+rmdir x y z
+for d in x y z; do
+        if [ -d $d ]; then
+                fail "instance $d still exists"
+        fi
+done
+
+set -e
+
+exit 0
#5 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

In reply to Steven Rostedt
Quoted 13 lines · Steven Rostedt

From: "Steven Rostedt (Red Hat)" <[email protected]>

Create a test to test instance creation and deletion. Several tasks are created that create 3 directories and delete them. The tasks all create the same directories. This places a stress on the code that creates and deletes instances.

tools/testing/selftests/ftrace/test.d/instances/instance.tc new file +0 -0
new file mode 100644
index 000000000000..773e276ff90b
@@ -0,0 +1,90 @@

<snip>

tools/testing/selftests/ftrace/test.d/instances/instance.tc line 73

Quoted 13 lines
tools/testing/selftests/ftrace/test.d/instances/instance.tc new file +13 -0
+instance_slam &
+x=`jobs -l | tail -1`
+p5=`echo $x | cut -d' ' -f2`
+echo $p5
+
+ls -lR >/dev/null
+sleep 1
+
+kill -1 $p1
+kill -1 $p2
+kill -1 $p3
+kill -1 $p4
+kill -1 $p5

Tells me:

./ftracetest: 72: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or kill -l [exitstatus]

And the test hangs forever.

I suspect because my /bin/sh is dash?

# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]

cheers

author maintainer
#6 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

Quoted 7 lines · On Wed, 09 Dec 2015 16:09:38 +1100 Michael Ellerman <[email protected]> wrote:

I suspect because my /bin/sh is dash?

# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]

I did some digging and found a better way to do this. Does the belowe patch fix things for you?

-- Steve

tools/testing/selftests/ftrace/test.d/instances/instance.tc +5 -10
index 773e276ff90b..1e1abe0ad354 100644
@@ -39,28 +39,23 @@ instance_slam() {
 }
 
 instance_slam &
-x=`jobs -l`
-p1=`echo $x | cut -d' ' -f2`
+p1=$!
 echo $p1
 
 instance_slam &
-x=`jobs -l | tail -1`
-p2=`echo $x | cut -d' ' -f2`
+p2=$!
 echo $p2
 
 instance_slam &
-x=`jobs -l | tail -1`
-p3=`echo $x | cut -d' ' -f2`
+p3=$!
 echo $p3
 
 instance_slam &
-x=`jobs -l | tail -1`
-p4=`echo $x | cut -d' ' -f2`
+p4=$!
 echo $p4
 
 instance_slam &
-x=`jobs -l | tail -1`
-p5=`echo $x | cut -d' ' -f2`
+p5=$!
 echo $p5
 
 ls -lR >/dev/null
#7 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

Quoted 5 lines · Steven Rostedt
Quoted 2 lines · On Wed, 09 Dec 2015 16:09:38 +1100 Michael Ellerman <[email protected]> wrote:

I suspect because my /bin/sh is dash?

Quoted 8 lines
Quoted 5 lines
# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]

I did some digging and found a better way to do this. Does the belowe patch fix things for you?

Yeah that works, thanks.

cheers

author maintainer
#8 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

Quoted 12 lines · On Fri, 15 Jan 2016 16:08:41 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 11 lines · Steven Rostedt
Quoted 7 lines · On Wed, 09 Dec 2015 16:09:38 +1100 Michael Ellerman <[email protected]> wrote:

I suspect because my /bin/sh is dash?

# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]
Quoted 5 lines
Quoted 2 lines

I did some digging and found a better way to do this. Does the belowe patch fix things for you?

Yeah that works, thanks.

Crap, I was waiting for a response, and forgot about this patch. I'll go ahead and add this to my queue of fixes.

Thanks!

-- Steve

#9 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

Quoted 15 lines · Steven Rostedt
Quoted 12 lines · On Fri, 15 Jan 2016 16:08:41 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 11 lines · Steven Rostedt
Quoted 7 lines · On Wed, 09 Dec 2015 16:09:38 +1100 Michael Ellerman <[email protected]> wrote:

I suspect because my /bin/sh is dash?

# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]
Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

I did some digging and found a better way to do this. Does the belowe patch fix things for you?

Yeah that works, thanks.

Crap, I was waiting for a response, and forgot about this patch. I'll go ahead and add this to my queue of fixes.

No worries, I was a bit slow replying :}

cheers

#10 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

In reply to Steven Rostedt
Quoted 15 lines · Steven Rostedt
Quoted 12 lines · On Fri, 15 Jan 2016 16:08:41 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 11 lines · Steven Rostedt
Quoted 7 lines · On Wed, 09 Dec 2015 16:09:38 +1100 Michael Ellerman <[email protected]> wrote:

I suspect because my /bin/sh is dash?

# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

I did some digging and found a better way to do this. Does the belowe patch fix things for you?

Yeah that works, thanks.

Crap, I was waiting for a response, and forgot about this patch. I'll go ahead and add this to my queue of fixes.

Hi Steve,

Looks like this got lost?

cheers

author maintainer
#11 lore raw

Re: [for-next][PATCH 3/7] ftracetest: Add instance create and delete test

Quoted 22 lines · On Thu, 18 Feb 2016 23:37:36 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 21 lines · Steven Rostedt
Quoted 17 lines · On Fri, 15 Jan 2016 16:08:41 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 5 lines · Steven Rostedt
Quoted 2 lines · On Wed, 09 Dec 2015 16:09:38 +1100 Michael Ellerman <[email protected]> wrote:

I suspect because my /bin/sh is dash?

Quoted 8 lines
Quoted 5 lines
# /bin/dash
# kill -1
/bin/dash: 1: kill: Usage: kill [-s sigspec | -signum | -sigspec] [pid | job]... or
kill -l [exitstatus]

I did some digging and found a better way to do this. Does the belowe patch fix things for you?

Yeah that works, thanks.

Quoted 6 lines
Quoted 2 lines

Crap, I was waiting for a response, and forgot about this patch. I'll go ahead and add this to my queue of fixes.

Hi Steve,

Looks like this got lost?

No, not lost, just pushed down the priority list. I got inundated by higher priority items since then :-/

I still need to do this though. Thanks for the reminder.

-- Steve

author maintainer
#12 lore raw

PATCH 4/7 ring-buffer: Use READ_ONCE() for most tail_page access

In reply to Steven Rostedt

From: "Steven Rostedt (Red Hat)" <[email protected]>

As cpu_buffer->tail_page may be modified by interrupts at almost any time, the flow of logic is very important. Do not let gcc get smart with re-reading cpu_buffer->tail_page by adding READ_ONCE() around most of its accesses.

---
 kernel/trace/ring_buffer.c | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)
kernel/trace/ring_buffer.c +11 -7
index 9c6045a27ba3..ab102e6259bc 100644
@@ -1036,7 +1036,7 @@ static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer,
 	 * it is, then it is up to us to update the tail
 	 * pointer.
 	 */
-	if (tail_page == cpu_buffer->tail_page) {
+	if (tail_page == READ_ONCE(cpu_buffer->tail_page)) {
 		/* Zero the write counter */
 		unsigned long val = old_write & ~RB_WRITE_MASK;
 		unsigned long eval = old_entries & ~RB_WRITE_MASK;
@@ -2036,12 +2036,15 @@ rb_handle_head_page(struct ring_buffer_per_cpu *cpu_buffer,
 	 * the tail page would have moved.
 	 */
 	if (ret == RB_PAGE_NORMAL) {
+		struct buffer_page *buffer_tail_page;
+
+		buffer_tail_page = READ_ONCE(cpu_buffer->tail_page);
 		/*
 		 * If the tail had moved passed next, then we need
 		 * to reset the pointer.
 		 */
-		if (cpu_buffer->tail_page != tail_page &&
-		    cpu_buffer->tail_page != next_page)
+		if (buffer_tail_page != tail_page &&
+		    buffer_tail_page != next_page)
 			rb_head_page_set_normal(cpu_buffer, new_head,
 						next_page,
 						RB_PAGE_HEAD);
@@ -2362,7 +2365,7 @@ rb_try_to_discard(struct ring_buffer_per_cpu *cpu_buffer,
 	addr = (unsigned long)event;
 	addr &= PAGE_MASK;
 
-	bpage = cpu_buffer->tail_page;
+	bpage = READ_ONCE(cpu_buffer->tail_page);
 
 	if (bpage->page == (void *)addr && rb_page_write(bpage) == old_index) {
 		unsigned long write_mask =
@@ -2410,7 +2413,7 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer)
  again:
 	max_count = cpu_buffer->nr_pages * 100;
 
-	while (cpu_buffer->commit_page != cpu_buffer->tail_page) {
+	while (cpu_buffer->commit_page != READ_ONCE(cpu_buffer->tail_page)) {
 		if (RB_WARN_ON(cpu_buffer, !(--max_count)))
 			return;
 		if (RB_WARN_ON(cpu_buffer,
@@ -2443,7 +2446,7 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer)
 	 * and pushed the tail page forward, we will be left with
 	 * a dangling commit that will never go forward.
 	 */
-	if (unlikely(cpu_buffer->commit_page != cpu_buffer->tail_page))
+	if (unlikely(cpu_buffer->commit_page != READ_ONCE(cpu_buffer->tail_page)))
 		goto again;
 }
 
@@ -2699,7 +2702,8 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 	if (unlikely(info->add_timestamp))
 		info->length += RB_LEN_TIME_EXTEND;
 
-	tail_page = info->tail_page = cpu_buffer->tail_page;
+	/* Don't let the compiler play games with cpu_buffer->tail_page */
+	tail_page = info->tail_page = READ_ONCE(cpu_buffer->tail_page);
 	write = local_add_return(info->length, &tail_page->write);
 
 	/* set write to only the index of the write */
Signature · 1 line
2.6.1
author maintainer
#13 lore raw

PATCH 5/7 ring-buffer: Remove redundant update of page timestamp

In reply to Steven Rostedt

From: "Steven Rostedt (Red Hat)" <[email protected]>

The first commit of a buffer page updates the timestamp of that page. No need to have the update to the next page add the timestamp too. It will only be replaced by the first commit on that page anyway.

Only update to a page if it contains an event.

---
 kernel/trace/ring_buffer.c | 32 ++++++++------------------------
 1 file changed, 8 insertions(+), 24 deletions(-)
kernel/trace/ring_buffer.c +8 -24
index ab102e6259bc..631541a53baf 100644
@@ -1001,17 +1001,13 @@ static int rb_head_page_replace(struct buffer_page *old,
 
 /*
  * rb_tail_page_update - move the tail page forward
- *
- * Returns 1 if moved tail page, 0 if someone else did.
  */
-static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer,
+static void rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer,
 			       struct buffer_page *tail_page,
 			       struct buffer_page *next_page)
 {
-	struct buffer_page *old_tail;
 	unsigned long old_entries;
 	unsigned long old_write;
-	int ret = 0;
 
 	/*
 	 * The tail page now needs to be moved forward.
@@ -1061,14 +1057,9 @@ static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer,
 		 */
 		local_set(&next_page->page->commit, 0);
 
-		old_tail = cmpxchg(&cpu_buffer->tail_page,
-				   tail_page, next_page);
-
-		if (old_tail == tail_page)
-			ret = 1;
+		/* Again, either we update tail_page or an interrupt does */
+		(void)cmpxchg(&cpu_buffer->tail_page, tail_page, next_page);
 	}
-
-	return ret;
 }
 
 static int rb_check_bpage(struct ring_buffer_per_cpu *cpu_buffer,
@@ -2150,7 +2141,6 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	struct ring_buffer *buffer = cpu_buffer->buffer;
 	struct buffer_page *next_page;
 	int ret;
-	u64 ts;
 
 	next_page = tail_page;
 
@@ -2224,15 +2214,7 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 		}
 	}
 
-	ret = rb_tail_page_update(cpu_buffer, tail_page, next_page);
-	if (ret) {
-		/*
-		 * Nested commits always have zero deltas, so
-		 * just reread the time stamp
-		 */
-		ts = rb_time_stamp(buffer);
-		next_page->page->time_stamp = ts;
-	}
+	rb_tail_page_update(cpu_buffer, tail_page, next_page);
 
  out_again:
 
@@ -2422,8 +2404,10 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer)
 		local_set(&cpu_buffer->commit_page->page->commit,
 			  rb_page_write(cpu_buffer->commit_page));
 		rb_inc_page(cpu_buffer, &cpu_buffer->commit_page);
-		cpu_buffer->write_stamp =
-			cpu_buffer->commit_page->page->time_stamp;
+		/* Only update the write stamp if the page has an event */
+		if (rb_page_write(cpu_buffer->commit_page))
+			cpu_buffer->write_stamp =
+				cpu_buffer->commit_page->page->time_stamp;
 		/* add barrier to keep gcc from optimizing too much */
 		barrier();
 	}
Signature · 1 line
2.6.1
author maintainer
#14 lore raw

PATCH 6/7 ring-buffer: Process commits whenever moving to a new page.

In reply to Steven Rostedt

From: "Steven Rostedt (Red Hat)" <[email protected]>

When crossing over to a new page, commit the current work. This will allow readers to get data with less latency, and also simplifies the work to get timestamps working for interrupted events.

---
 kernel/module.c            | 18 ++++++++++++++++--
 kernel/trace/ring_buffer.c |  7 +++++++
 2 files changed, 23 insertions(+), 2 deletions(-)
kernel/module.c +16 -2
index 8f051a106676..c0ae7c18c2fa 100644
@@ -284,11 +284,25 @@ static void module_assert_mutex(void)
 static void module_assert_mutex_or_preempt(void)
 {
 #ifdef CONFIG_LOCKDEP
+	static int once;
+
 	if (unlikely(!debug_locks))
 		return;
 
-	WARN_ON(!rcu_read_lock_sched_held() &&
-		!lockdep_is_held(&module_mutex));
+	/*
+	 * Would be nice to use WARN_ON_ONCE(), but the warning
+	 * that causes a stack trace may call __module_address()
+	 * which may call here, and we trigger the warning again,
+	 * before the WARN_ON_ONCE() updates its flag.
+	 * To prevent the recursion, we need to open code the
+	 * once logic.
+	 */
+	if (!once &&
+	    unlikely(!rcu_read_lock_sched_held() &&
+		     !lockdep_is_held(&module_mutex))) {
+		once++;
+		WARN_ON(1);
+	}
 #endif
 }
 
kernel/trace/ring_buffer.c +7 -0
index 631541a53baf..95181e36891a 100644
@@ -2129,6 +2129,8 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	local_sub(length, &tail_page->write);
 }
 
+static inline void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer);
+
 /*
  * This is the slow path, force gcc not to inline it.
  */
@@ -2220,6 +2222,11 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 
 	rb_reset_tail(cpu_buffer, tail, info);
 
+	/* Commit what we have for now. */
+	rb_end_commit(cpu_buffer);
+	/* rb_end_commit() decs committing */
+	local_inc(&cpu_buffer->committing);
+
 	/* fail and let the caller try again */
 	return ERR_PTR(-EAGAIN);
 
Signature · 1 line
2.6.1
author maintainer
#15 lore raw

Re: [for-next][PATCH 6/7] ring-buffer: Process commits whenever moving to a new page.

Quoted 24 lines
kernel/module.c +9 -2
index 8f051a106676..c0ae7c18c2fa 100644
@@ -284,11 +284,25 @@ static void module_assert_mutex(void)
 static void module_assert_mutex_or_preempt(void)
 {
 #ifdef CONFIG_LOCKDEP
+	static int once;
+
 	if (unlikely(!debug_locks))
 		return;
 
-	WARN_ON(!rcu_read_lock_sched_held() &&
-		!lockdep_is_held(&module_mutex));
+	/*
+	 * Would be nice to use WARN_ON_ONCE(), but the warning
+	 * that causes a stack trace may call __module_address()
+	 * which may call here, and we trigger the warning again,
+	 * before the WARN_ON_ONCE() updates its flag.
+	 * To prevent the recursion, we need to open code the
+	 * once logic.

kernel/module.c static void module_assert_mutex(void) · line 308

Quoted 10 lines
kernel/module.c +7 -0
+	 */
+	if (!once &&
+	    unlikely(!rcu_read_lock_sched_held() &&
+		     !lockdep_is_held(&module_mutex))) {
+		once++;
+		WARN_ON(1);
+	}
 #endif
 }
 

Damn it. I added this via quilt to debug something (as this was causing a crash), and forgot to pop it when doing my "git commit -a -s".

I'll have to nuke this part of the patch and rebase my tree :-(

-- Steve

author maintainer
#16 lore raw

PATCH 7/7 tracing: Update cond flag when enabling or disabling a trigger

In reply to Steven Rostedt

From: Tom Zanussi <[email protected]>

When a trigger is enabled, the cond flag should be set beforehand, otherwise a trigger that's expecting to process a trace record (e.g. one with post_trigger set) could be invoked without one.

Likewise a trigger's cond flag should be reset after it's disabled, not before.

---
 kernel/trace/trace_events_trigger.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
kernel/trace/trace_events_trigger.c +6 -4
index 42a4009fd75a..4d2f3ccc56f6 100644
@@ -543,11 +543,12 @@ static int register_trigger(char *glob, struct event_trigger_ops *ops,
 	list_add_rcu(&data->list, &file->triggers);
 	ret++;
 
+	update_cond_flag(file);
 	if (trace_event_trigger_enable_disable(file, 1) < 0) {
 		list_del_rcu(&data->list);
+		update_cond_flag(file);
 		ret--;
 	}
-	update_cond_flag(file);
 out:
 	return ret;
 }
@@ -575,8 +576,8 @@ static void unregister_trigger(char *glob, struct event_trigger_ops *ops,
 		if (data->cmd_ops->trigger_type == test->cmd_ops->trigger_type) {
 			unregistered = true;
 			list_del_rcu(&data->list);
-			update_cond_flag(file);
 			trace_event_trigger_enable_disable(file, 0);
+			update_cond_flag(file);
 			break;
 		}
 	}
@@ -1319,11 +1320,12 @@ static int event_enable_register_trigger(char *glob,
 	list_add_rcu(&data->list, &file->triggers);
 	ret++;
 
+	update_cond_flag(file);
 	if (trace_event_trigger_enable_disable(file, 1) < 0) {
 		list_del_rcu(&data->list);
+		update_cond_flag(file);
 		ret--;
 	}
-	update_cond_flag(file);
 out:
 	return ret;
 }
@@ -1344,8 +1346,8 @@ static void event_enable_unregister_trigger(char *glob,
 		    (enable_data->file == test_enable_data->file)) {
 			unregistered = true;
 			list_del_rcu(&data->list);
-			update_cond_flag(file);
 			trace_event_trigger_enable_disable(file, 0);
+			update_cond_flag(file);
 			break;
 		}
 	}
Signature · 1 line
2.6.1