2009-07-01 08:50:28

by Xiao Guangrong

Subject: [PATCH] tracing: fix trace_print_seq()

We will lose something if trace_seq->buffer[0] is 0, because the copy length
is calculated by strlen() in seq_puts(), so using seq_write() instead of
seq_puts().

There have a example:
after reboot:
# echo kmemtrace > current_tracer
# echo 0 > options/kmem_minimalistic
# cat trace
# tracer: kmemtrace
#
#
Nothing is exported, because the first byte of trace_seq->buffer[ ]
is KMEMTRACE_USER_ALLOC.
( the value of KMEMTRACE_USER_ALLOC is zero, seeing
kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)

Signed-off-by: Xiao Guangrong <[email protected]>
---
kernel/trace/trace_output.c | 3 +--
1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index 7938f3a..e0c2545 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -27,8 +27,7 @@ void trace_print_seq(struct seq_file *m, struct trace_seq *s)
{
int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;

- s->buffer[len] = 0;
- seq_puts(m, s->buffer);
+ seq_write(m, s->buffer, len);

trace_seq_init(s);
}
--
1.6.1.2

2009-07-01 14:01:30

by Frederic Weisbecker

Subject: Re: [PATCH] tracing: fix trace_print_seq()

On Wed, Jul 01, 2009 at 04:50:25PM +0800, Xiao Guangrong wrote:
> We will lose something if trace_seq->buffer[0] is 0, because the copy length
> is calculated by strlen() in seq_puts(), so using seq_write() instead of
> seq_puts().
>
> There have a example:
> after reboot:
> # echo kmemtrace > current_tracer
> # echo 0 > options/kmem_minimalistic
> # cat trace
> # tracer: kmemtrace
> #
> #
> Nothing is exported, because the first byte of trace_seq->buffer[ ]
> is KMEMTRACE_USER_ALLOC.
> ( the value of KMEMTRACE_USER_ALLOC is zero, seeing
> kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)
>
> Signed-off-by: Xiao Guangrong <[email protected]>
> ---
> kernel/trace/trace_output.c | 3 +--
> 1 files changed, 1 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
> index 7938f3a..e0c2545 100644
> --- a/kernel/trace/trace_output.c
> +++ b/kernel/trace/trace_output.c
> @@ -27,8 +27,7 @@ void trace_print_seq(struct seq_file *m, struct trace_seq *s)
> {
> int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;
>
> - s->buffer[len] = 0;
> - seq_puts(m, s->buffer);
> + seq_write(m, s->buffer, len);
>
> trace_seq_init(s);
> }
> --
> 1.6.1.2
>

Looks good, thanks.

Acked-by: Frederic Weisbecker <[email protected]>

But having a '\0' in a non binary trace looks weird. Why do
we have such plain KMEMTRACE_USER_ALLOC whereas binary tracing
is not set, as your example shows it?

2009-07-01 14:07:28

by Xiao Guangrong

Subject: [tip:tracing/urgent] tracing: Fix trace_print_seq()

Commit-ID: 5c37f525657e879a17a27c83021602cd55837531
Gitweb: http://git.kernel.org/tip/5c37f525657e879a17a27c83021602cd55837531
Author: Xiao Guangrong <[email protected]>
AuthorDate: Wed, 1 Jul 2009 16:50:25 +0800
Committer: Ingo Molnar <[email protected]>
CommitDate: Wed, 1 Jul 2009 16:03:50 +0200

tracing: Fix trace_print_seq()

We will lose something if trace_seq->buffer[0] is 0, because the copy length
is calculated by strlen() in seq_puts(), so using seq_write() instead of
seq_puts().

There have a example:
after reboot:

# echo kmemtrace > current_tracer
# echo 0 > options/kmem_minimalistic
# cat trace
# tracer: kmemtrace
#
#

Nothing is exported, because the first byte of trace_seq->buffer[ ]
is KMEMTRACE_USER_ALLOC.

( the value of KMEMTRACE_USER_ALLOC is zero, seeing
kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)

Signed-off-by: Xiao Guangrong <[email protected]>
Acked-by: Frederic Weisbecker <[email protected]>
Cc: Eduard - Gabriel Munteanu <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: Pekka Enberg <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>


---
kernel/trace/trace_output.c | 3 +--
1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index 7938f3a..e0c2545 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -27,8 +27,7 @@ void trace_print_seq(struct seq_file *m, struct trace_seq *s)
{
int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;

- s->buffer[len] = 0;
- seq_puts(m, s->buffer);
+ seq_write(m, s->buffer, len);

trace_seq_init(s);
}

2009-07-01 14:16:12

by Pekka Enberg

Subject: Re: [tip:tracing/urgent] tracing: Fix trace_print_seq()

On Wed, 2009-07-01 at 14:06 +0000, tip-bot for Xiao Guangrong wrote:
> Commit-ID: 5c37f525657e879a17a27c83021602cd55837531
> Gitweb: http://git.kernel.org/tip/5c37f525657e879a17a27c83021602cd55837531
> Author: Xiao Guangrong <[email protected]>
> AuthorDate: Wed, 1 Jul 2009 16:50:25 +0800
> Committer: Ingo Molnar <[email protected]>
> CommitDate: Wed, 1 Jul 2009 16:03:50 +0200
>
> tracing: Fix trace_print_seq()
>
> We will lose something if trace_seq->buffer[0] is 0, because the copy length
> is calculated by strlen() in seq_puts(), so using seq_write() instead of
> seq_puts().
>
> There have a example:
> after reboot:
>
> # echo kmemtrace > current_tracer
> # echo 0 > options/kmem_minimalistic
> # cat trace
> # tracer: kmemtrace
> #
> #
>
> Nothing is exported, because the first byte of trace_seq->buffer[ ]
> is KMEMTRACE_USER_ALLOC.
>
> ( the value of KMEMTRACE_USER_ALLOC is zero, seeing
> kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)
>
> Signed-off-by: Xiao Guangrong <[email protected]>
> Acked-by: Frederic Weisbecker <[email protected]>
> Cc: Eduard - Gabriel Munteanu <[email protected]>
> Cc: Steven Rostedt <[email protected]>
> Cc: Pekka Enberg <[email protected]>
> LKML-Reference: <[email protected]>
> Signed-off-by: Ingo Molnar <[email protected]>

Nice catch!

Acked-by: Pekka Enberg <[email protected]>

2009-07-01 14:43:34

by Xiao Guangrong

Subject: [tip:tracing/urgent] tracing: Fix trace_print_seq()

Commit-ID: 393ab9a2a6cf958c9549e35bd88c03b9b3eae64e
Gitweb: http://git.kernel.org/tip/393ab9a2a6cf958c9549e35bd88c03b9b3eae64e
Author: Xiao Guangrong <[email protected]>
AuthorDate: Wed, 1 Jul 2009 16:50:25 +0800
Committer: Ingo Molnar <[email protected]>
CommitDate: Wed, 1 Jul 2009 16:41:31 +0200

tracing: Fix trace_print_seq()

We will lose something if trace_seq->buffer[0] is 0, because the copy length
is calculated by strlen() in seq_puts(), so using seq_write() instead of
seq_puts().

There have a example:
after reboot:

# echo kmemtrace > current_tracer
# echo 0 > options/kmem_minimalistic
# cat trace
# tracer: kmemtrace
#
#

Nothing is exported, because the first byte of trace_seq->buffer[ ]
is KMEMTRACE_USER_ALLOC.

( the value of KMEMTRACE_USER_ALLOC is zero, seeing
kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)

Signed-off-by: Xiao Guangrong <[email protected]>
Acked-by: Frederic Weisbecker <[email protected]>
Acked-by: Pekka Enberg <[email protected]>
Cc: Eduard - Gabriel Munteanu <[email protected]>
Cc: Steven Rostedt <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>


---
kernel/trace/trace_output.c | 3 +--
1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index 7938f3a..e0c2545 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -27,8 +27,7 @@ void trace_print_seq(struct seq_file *m, struct trace_seq *s)
{
int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;

- s->buffer[len] = 0;
- seq_puts(m, s->buffer);
+ seq_write(m, s->buffer, len);

trace_seq_init(s);
}

2009-07-02 05:56:17

by Li Zefan

Subject: Re: [PATCH] tracing: fix trace_print_seq()

>> There have a example:
>> after reboot:
>> # echo kmemtrace > current_tracer
>> # echo 0 > options/kmem_minimalistic
>> # cat trace
>> # tracer: kmemtrace
>> #
>> #
>> Nothing is exported, because the first byte of trace_seq->buffer[ ]
>> is KMEMTRACE_USER_ALLOC.
>> ( the value of KMEMTRACE_USER_ALLOC is zero, seeing
>> kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)
>>
...
> Looks good, thanks.
>
> Acked-by: Frederic Weisbecker <[email protected]>
>
> But having a '\0' in a non binary trace looks weird. Why do
> we have such plain KMEMTRACE_USER_ALLOC whereas binary tracing
> is not set, as your example shows it?
>

Because of this commit:

| commit 42af9054c0eeed09ec58d13ec8bf52d225ebcfcc
| Author: Eduard - Gabriel Munteanu <[email protected]>
| Date: Mon Mar 23 15:12:26 2009 +0200
|
| kmemtrace: restore original tracing data binary format, improve ABI

I think we should change it to use trace_event->binary() interface for
binary trace.

2009-07-02 06:56:20

by Xiao Guangrong

Subject: [tip:tracing/urgent] tracing: Fix trace_print_seq()

Commit-ID: e1af3aec3e2e7d21d4b3054323779d478c19d907
Gitweb: http://git.kernel.org/tip/e1af3aec3e2e7d21d4b3054323779d478c19d907
Author: Xiao Guangrong <[email protected]>
AuthorDate: Wed, 1 Jul 2009 16:50:25 +0800
Committer: Ingo Molnar <[email protected]>
CommitDate: Thu, 2 Jul 2009 08:51:13 +0200

tracing: Fix trace_print_seq()

We will lose something if trace_seq->buffer[0] is 0, because the copy length
is calculated by strlen() in seq_puts(), so using seq_write() instead of
seq_puts().

There have a example:
after reboot:

# echo kmemtrace > current_tracer
# echo 0 > options/kmem_minimalistic
# cat trace
# tracer: kmemtrace
#
#

Nothing is exported, because the first byte of trace_seq->buffer[ ]
is KMEMTRACE_USER_ALLOC.

( the value of KMEMTRACE_USER_ALLOC is zero, seeing
kmemtrace_print_alloc_user() in kernel/trace/kmemtrace.c)

Signed-off-by: Xiao Guangrong <[email protected]>
Acked-by: Frederic Weisbecker <[email protected]>
Acked-by: Pekka Enberg <[email protected]>
Acked-by: Eduard - Gabriel Munteanu <[email protected]>
Cc: Steven Rostedt <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>


---
kernel/trace/trace_output.c | 3 +--
1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index 7938f3a..e0c2545 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -27,8 +27,7 @@ void trace_print_seq(struct seq_file *m, struct trace_seq *s)
{
int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;

- s->buffer[len] = 0;
- seq_puts(m, s->buffer);
+ seq_write(m, s->buffer, len);

trace_seq_init(s);
}