Skip to contentSkip to conversation

updates for 2.6.31

GIT PULL In review Steven Rostedt 2 participants 10 messages Jun 15, 2009

Subsystems (2)
lore
Share & tools
lore

Steven Rostedt

author maintainer
#0 lore raw

Hi Ingo,

This patch series are fixes that should get into 2.6.31 (maybe some of them even into stable?).

Please pull the latest tracing/ftrace tree, which can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git
tracing/ftrace

Li Zefan (4):
tracing: fix a typo in tracing_cpumask_write()
tracing: replace a GFP_ATOMIC with GFP_KERNEL allocation tracing/filters: operand can be negative tracing/filters: strloc should be unsigned short

Rusty Russell (1):
cpumask: use new operators in kernel/trace

Steven Rostedt (1):
ring-buffer: prevent adding write in discarded area

Wu Zhangjin (1):
tracing: fix undeclared 'PAGE_SIZE' in include/linux/trace_seq.h

----

 include/linux/trace_seq.h          |    2 +
 kernel/trace/kmemtrace.c           |    2 +-
 kernel/trace/ring_buffer.c         |   72 +++++++++++++++++++++++++++++-------
 kernel/trace/trace.c               |    8 ++--
 kernel/trace/trace_events_filter.c |    9 +++-
 5 files changed, 72 insertions(+), 21 deletions(-)
author maintainer
#1 lore raw

PATCH 1/7 cpumask: use new operators in kernel/trace

From: Rusty Russell <[email protected]>

LKML-Reference: <[email protected]>

---
 kernel/trace/kmemtrace.c   |    2 +-
 kernel/trace/ring_buffer.c |    4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)
kernel/trace/kmemtrace.c +1 -1
index 86cdf67..1edaa95 100644
@@ -186,7 +186,7 @@ static int kmem_trace_init(struct trace_array *tr)
 	int cpu;
 	kmemtrace_array = tr;
 
-	for_each_cpu_mask(cpu, cpu_possible_map)
+	for_each_cpu(cpu, cpu_possible_mask)
 		tracing_reset(tr, cpu);
 
 	kmemtrace_start_probes();
kernel/trace/ring_buffer.c +2 -2
index 2e642b2..9c31c9f 100644
@@ -3105,7 +3105,7 @@ static int rb_cpu_notify(struct notifier_block *self,
 	switch (action) {
 	case CPU_UP_PREPARE:
 	case CPU_UP_PREPARE_FROZEN:
-		if (cpu_isset(cpu, *buffer->cpumask))
+		if (cpumask_test_cpu(cpu, buffer->cpumask))
 			return NOTIFY_OK;
 
 		buffer->buffers[cpu] =
@@ -3116,7 +3116,7 @@ static int rb_cpu_notify(struct notifier_block *self,
 			return NOTIFY_OK;
 		}
 		smp_wmb();
-		cpu_set(cpu, *buffer->cpumask);
+		cpumask_set_cpu(cpu, buffer->cpumask);
 		break;
 	case CPU_DOWN_PREPARE:
 	case CPU_DOWN_PREPARE_FROZEN:
Signature · 3 lines
1.6.3.1

-- 
author maintainer
#2 lore raw

PATCH 2/7 tracing: fix undeclared PAGE_SIZE in include/linux/trace_seq.h

In reply to Steven Rostedt

From: Wu Zhangjin <[email protected]>

when compiling linux-mips with kmemtrace enabled, there will be an error:

include/linux/trace_seq.h:12: error: 'PAGE_SIZE' undeclared here (not in
a function)

I checked the source code and found trace_seq.h used PAGE_SIZE but not included the relative header file, so, fix it via adding the header file <asm/page.h>

LKML-Reference: <[email protected]>

---
 include/linux/trace_seq.h |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/trace_seq.h +2 -0
index c68bccb..c134dd1 100644
@@ -3,6 +3,8 @@
 
 #include <linux/fs.h>
 
+#include <asm/page.h>
+
 /*
  * Trace sequences are used to allow a function to call several other functions
  * to create a string of data to use (up to a max of PAGE_SIZE.
Signature · 3 lines
1.6.3.1

-- 
author maintainer
#3 lore raw

PATCH 3/7 tracing: fix a typo in tracing_cpumask_write()

In reply to Steven Rostedt

From: Li Zefan <[email protected]>

It's tracing_cpumask_new that should be kfree()ed.

This causes tracing_cpumask to be freed due to the typo:

 # echo z > tracing_cpumask
 bash: echo: write error: Invalid argument

And subsequent reads/writes to tracing_cpuamsk will access this already-freed tracing_cpumask, thus may lead to crash.

[ Impact: fix leak and crash when writing invalid val to tracing_cpumask ]

LKML-Reference: <[email protected]>

---
 kernel/trace/trace.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)
kernel/trace/trace.c +3 -3
index 8acd9b8..7355a38 100644
@@ -2191,11 +2191,12 @@ tracing_cpumask_write(struct file *filp, const char __user *ubuf,
 	if (!alloc_cpumask_var(&tracing_cpumask_new, GFP_KERNEL))
 		return -ENOMEM;
 
-	mutex_lock(&tracing_cpumask_update_lock);
 	err = cpumask_parse_user(ubuf, count, tracing_cpumask_new);
 	if (err)
 		goto err_unlock;
 
+	mutex_lock(&tracing_cpumask_update_lock);
+
 	local_irq_disable();
 	__raw_spin_lock(&ftrace_max_lock);
 	for_each_tracing_cpu(cpu) {
@@ -2223,8 +2224,7 @@ tracing_cpumask_write(struct file *filp, const char __user *ubuf,
 	return count;
 
 err_unlock:
-	mutex_unlock(&tracing_cpumask_update_lock);
-	free_cpumask_var(tracing_cpumask);
+	free_cpumask_var(tracing_cpumask_new);
 
 	return err;
 }
Signature · 3 lines
1.6.3.1

-- 
author maintainer
#4 lore raw

PATCH 4/7 tracing: replace a GFP_ATOMIC with GFP_KERNEL allocation

In reply to Steven Rostedt

From: Li Zefan <[email protected]>

Atomic allocation is not needed here.

[ Impact: clean up of memory alloction type ]

LKML-Reference: <[email protected]>

---
 kernel/trace/trace.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
kernel/trace/trace.c +1 -1
index 7355a38..0f57f1b 100644
@@ -3627,7 +3627,7 @@ tracing_stats_read(struct file *filp, char __user *ubuf,
 	struct trace_seq *s;
 	unsigned long cnt;
 
-	s = kmalloc(sizeof(*s), GFP_ATOMIC);
+	s = kmalloc(sizeof(*s), GFP_KERNEL);
 	if (!s)
 		return ENOMEM;
 
Signature · 3 lines
1.6.3.1

-- 
#5 lore raw

Re: [PATCH 4/7] tracing: replace a GFP_ATOMIC with GFP_KERNEL allocation

Quoted 5 lines · Steven Rostedt

From: Li Zefan <[email protected]>

Atomic allocation is not needed here.

[ Impact: clean up of memory alloction type ]

Note, the impact line was probably not needed here - and if we add one then the correct one would be:

[ Impact: use less failure-prone (non-atomic) allocations ]

Ingo

author maintainer
#6 lore raw

PATCH 5/7 tracing/filters: operand can be negative

In reply to Steven Rostedt

From: Li Zefan <[email protected]>

This should be a bug:

 # cat format
 name: foo_bar
 ID: 71
 format:
	 ...
         field:int bar;  offset:24;      size:4;
 # echo 'bar < 0' > filter
 # echo 'bar < -1' > filter
 bash: echo: write error: Invalid argument

[ Impact: fix to allow negative operand in filer expr ]

LKML-Reference: <[email protected]>

---
 kernel/trace/trace_events_filter.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)
kernel/trace/trace_events_filter.c +6 -1
index db6e54b..1d81923 100644
@@ -546,6 +546,7 @@ static int filter_add_pred(struct filter_parse_state *ps,
 	filter_pred_fn_t fn;
 	unsigned long long val;
 	int string_type;
+	int ret;
 
 	pred->fn = filter_pred_none;
 
@@ -581,7 +582,11 @@ static int filter_add_pred(struct filter_parse_state *ps,
 			pred->not = 1;
 		return filter_add_pred_fn(ps, call, pred, fn);
 	} else {
-		if (strict_strtoull(pred->str_val, 0, &val)) {
+		if (field->is_signed)
+			ret = strict_strtoll(pred->str_val, 0, &val);
+		else
+			ret = strict_strtoull(pred->str_val, 0, &val);
+		if (ret) {
 			parse_error(ps, FILT_ERR_ILLEGAL_INTVAL, 0);
 			return -EINVAL;
 		}
Signature · 3 lines
1.6.3.1

-- 
author maintainer
#7 lore raw

PATCH 6/7 tracing/filters: strloc should be unsigned short

In reply to Steven Rostedt

From: Li Zefan <[email protected]>

I forgot to update filter code accordingly in "tracing/events: change the type of __str_loc_item to unsigned short" (commt b0aae68cc5508f3c2fbf728988c954db4c8b8a53)

It can cause system crash:

 # echo 1 > tracing/events/irq/irq_handler_entry/enable
 # echo 'name == eth0' > tracing/events/irq/irq_handler_entry/filter

[ Impact: fix crash while filtering on __string() field ]

LKML-Reference: <[email protected]>

---
 kernel/trace/trace_events_filter.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
kernel/trace/trace_events_filter.c +1 -1
index 1d81923..b24ab0e 100644
@@ -178,7 +178,7 @@ static int filter_pred_string(struct filter_pred *pred, void *event,
 static int filter_pred_strloc(struct filter_pred *pred, void *event,
 			      int val1, int val2)
 {
-	int str_loc = *(int *)(event + pred->offset);
+	unsigned short str_loc = *(unsigned short *)(event + pred->offset);
 	char *addr = (char *)(event + str_loc);
 	int cmp, match;
 
Signature · 3 lines
1.6.3.1

-- 
author maintainer
#8 lore raw

PATCH 7/7 ring-buffer: prevent adding write in discarded area

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

This a very tight race where an interrupt could come in and not have enough data to put into the end of a buffer page, and that it would fail to write and need to go to the next page.

But if this happened when another writer was about to reserver their data, and that writer has smaller data to reserve, then it could succeed even though the interrupt moved the tail page.

To pervent that, if we fail to store data, and by subtracting the amount we reserved we still have room for smaller data, we need to fill that space with "discarded" data.

[ Impact: prevent race were buffer data may be lost ]

---
 kernel/trace/ring_buffer.c |   68 ++++++++++++++++++++++++++++++++++++--------
 1 files changed, 56 insertions(+), 12 deletions(-)
kernel/trace/ring_buffer.c +56 -12
index 9c31c9f..dbc0f93 100644
@@ -205,6 +205,7 @@ EXPORT_SYMBOL_GPL(tracing_is_on);
 #define RB_EVNT_HDR_SIZE (offsetof(struct ring_buffer_event, array))
 #define RB_ALIGNMENT		4U
 #define RB_MAX_SMALL_DATA	(RB_ALIGNMENT * RINGBUF_TYPE_DATA_TYPE_LEN_MAX)
+#define RB_EVNT_MIN_SIZE	8U	/* two 32bit words */
 
 /* define RINGBUF_TYPE_DATA for 'case RINGBUF_TYPE_DATA:' */
 #define RINGBUF_TYPE_DATA 0 ... RINGBUF_TYPE_DATA_TYPE_LEN_MAX
@@ -1170,6 +1171,59 @@ static unsigned rb_calculate_event_length(unsigned length)
 	return length;
 }
 
+static inline void
+rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
+	      struct buffer_page *tail_page,
+	      unsigned long tail, unsigned long length)
+{
+	struct ring_buffer_event *event;
+
+	/*
+	 * Only the event that crossed the page boundary
+	 * must fill the old tail_page with padding.
+	 */
+	if (tail >= BUF_PAGE_SIZE) {
+		local_sub(length, &tail_page->write);
+		return;
+	}
+
+	event = __rb_page_index(tail_page, tail);
+
+	/*
+	 * If this event is bigger than the minimum size, then
+	 * we need to be careful that we don't subtract the
+	 * write counter enough to allow another writer to slip
+	 * in on this page.
+	 * We put in a discarded commit instead, to make sure
+	 * that this space is not used again.
+	 *
+	 * If we are less than the minimum size, we don't need to
+	 * worry about it.
+	 */
+	if (tail > (BUF_PAGE_SIZE - RB_EVNT_MIN_SIZE)) {
+		/* No room for any events */
+
+		/* Mark the rest of the page with padding */
+		rb_event_set_padding(event);
+
+		/* Set the write back to the previous setting */
+		local_sub(length, &tail_page->write);
+		return;
+	}
+
+	/* Put in a discarded event */
+	event->array[0] = (BUF_PAGE_SIZE - tail) - RB_EVNT_HDR_SIZE;
+	event->type_len = RINGBUF_TYPE_PADDING;
+	/* time delta must be non zero */
+	event->time_delta = 1;
+	/* Account for this as an entry */
+	local_inc(&tail_page->entries);
+	local_inc(&cpu_buffer->entries);
+
+	/* Set write to end of buffer */
+	length = (tail + length) - BUF_PAGE_SIZE;
+	local_sub(length, &tail_page->write);
+}
 
 static struct ring_buffer_event *
 rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
@@ -1264,17 +1318,7 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 		cpu_buffer->tail_page->page->time_stamp = *ts;
 	}
 
-	/*
-	 * The actual tail page has moved forward.
-	 */
-	if (tail < BUF_PAGE_SIZE) {
-		/* Mark the rest of the page with padding */
-		event = __rb_page_index(tail_page, tail);
-		rb_event_set_padding(event);
-	}
-
-	/* Set the write back to the previous setting */
-	local_sub(length, &tail_page->write);
+	rb_reset_tail(cpu_buffer, tail_page, tail, length);
 
 	/*
 	 * If this was a commit entry that failed,
@@ -1293,7 +1337,7 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 
  out_reset:
 	/* reset write */
-	local_sub(length, &tail_page->write);
+	rb_reset_tail(cpu_buffer, tail_page, tail, length);
 
 	if (likely(lock_taken))
 		__raw_spin_unlock(&cpu_buffer->lock);
Signature · 3 lines
1.6.3.1

-- 
#9 lore raw
In reply to Steven Rostedt
Quoted 26 lines · Steven Rostedt

Hi Ingo,

This patch series are fixes that should get into 2.6.31 (maybe some of them even into stable?).

Please pull the latest tracing/ftrace tree, which can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git
tracing/ftrace

Li Zefan (4):
tracing: fix a typo in tracing_cpumask_write()
tracing: replace a GFP_ATOMIC with GFP_KERNEL allocation tracing/filters: operand can be negative tracing/filters: strloc should be unsigned short

Rusty Russell (1):
cpumask: use new operators in kernel/trace

Steven Rostedt (1):
ring-buffer: prevent adding write in discarded area

Wu Zhangjin (1):
tracing: fix undeclared 'PAGE_SIZE' in include/linux/trace_seq.h

Quoted 7 lines

----

 include/linux/trace_seq.h          |    2 +
 kernel/trace/kmemtrace.c           |    2 +-
 kernel/trace/ring_buffer.c         |   72 +++++++++++++++++++++++++++++-------
 kernel/trace/trace.c               |    8 ++--
 kernel/trace/trace_events_filter.c |    9 +++-
 5 files changed, 72 insertions(+), 21 deletions(-)

Pulled into tracing/urgent - thanks Steve!

Ingo