2017-09-11 18:18:29

by Bo Yan

Subject: [PATCH] tracing: erase irqsoff trace with empty write

One convenient way to erase trace is "echo > trace". However, this
is currently broken if the current tracer is irqsoff tracer. This
is because irqsoff tracer use max_buffer as the default trace
buffer.

Set the max_buffer as the one to be cleared when it's the trace
buffer currently in use.

Signed-off-by: Bo Yan <[email protected]>
---
kernel/trace/trace.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index 5360b7aec57a..1f91bdd9365b 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -4020,11 +4020,17 @@ static int tracing_open(struct inode *inode, struct file *file)
/* If this file was open for write, then erase contents */
if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
int cpu = tracing_get_cpu(inode);
+ struct trace_buffer *tr_buf = &tr->trace_buffer;
+
+#ifdef CONFIG_TRACER_MAX_TRACE
+ if (tr->current_trace->print_max)
+ tr_buf = &tr->max_buffer;
+#endif

if (cpu == RING_BUFFER_ALL_CPUS)
- tracing_reset_online_cpus(&tr->trace_buffer);
+ tracing_reset_online_cpus(tr_buf);
else
- tracing_reset(&tr->trace_buffer, cpu);
+ tracing_reset(tr_buf, cpu);
}

if (file->f_mode & FMODE_READ) {
--
2.7.4

2017-09-18 14:38:44

by Steven Rostedt

Subject: Re: [PATCH] tracing: erase irqsoff trace with empty write

On Mon, 11 Sep 2017 11:16:35 -0700
Bo Yan <[email protected]> wrote:

> One convenient way to erase trace is "echo > trace". However, this
> is currently broken if the current tracer is irqsoff tracer. This
> is because irqsoff tracer use max_buffer as the default trace
> buffer.
>
> Set the max_buffer as the one to be cleared when it's the trace
> buffer currently in use.

Hi Bo,

I have no problems with the logic of the patch. But just one nit. The
tr->trace_buffer has been called "trace_buf" elsewhere in the file. Can
you resend with using the variable name "trace_buf" instead of "tr_buf"
just to keep it consistent in the file. It's also been called 'buf' but
trace_buf would probably be more appropriate.

Thanks!

-- Steve

>
> Signed-off-by: Bo Yan <[email protected]>
> ---
> kernel/trace/trace.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index 5360b7aec57a..1f91bdd9365b 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -4020,11 +4020,17 @@ static int tracing_open(struct inode *inode, struct file *file)
> /* If this file was open for write, then erase contents */
> if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
> int cpu = tracing_get_cpu(inode);
> + struct trace_buffer *tr_buf = &tr->trace_buffer;
> +
> +#ifdef CONFIG_TRACER_MAX_TRACE
> + if (tr->current_trace->print_max)
> + tr_buf = &tr->max_buffer;
> +#endif
>
> if (cpu == RING_BUFFER_ALL_CPUS)
> - tracing_reset_online_cpus(&tr->trace_buffer);
> + tracing_reset_online_cpus(tr_buf);
> else
> - tracing_reset(&tr->trace_buffer, cpu);
> + tracing_reset(tr_buf, cpu);
> }
>
> if (file->f_mode & FMODE_READ) {

2017-09-19 17:43:28

by Bo Yan

Subject: Re: [PATCH] tracing: erase irqsoff trace with empty write

Steven,

Thanks for review. I have posted the following update

https://lkml.org/lkml/2017/9/18/715

Bo

On 09/18/2017 07:38 AM, Steven Rostedt wrote:
> On Mon, 11 Sep 2017 11:16:35 -0700
> Bo Yan <[email protected]> wrote:
>
>> One convenient way to erase trace is "echo > trace". However, this
>> is currently broken if the current tracer is irqsoff tracer. This
>> is because irqsoff tracer use max_buffer as the default trace
>> buffer.
>>
>> Set the max_buffer as the one to be cleared when it's the trace
>> buffer currently in use.
>
> Hi Bo,
>
> I have no problems with the logic of the patch. But just one nit. The
> tr->trace_buffer has been called "trace_buf" elsewhere in the file. Can
> you resend with using the variable name "trace_buf" instead of "tr_buf"
> just to keep it consistent in the file. It's also been called 'buf' but
> trace_buf would probably be more appropriate.
>
> Thanks!
>
> -- Steve
>
>>
>> Signed-off-by: Bo Yan <[email protected]>
>> ---
>> kernel/trace/trace.c | 10 ++++++++--
>> 1 file changed, 8 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
>> index 5360b7aec57a..1f91bdd9365b 100644
>> --- a/kernel/trace/trace.c
>> +++ b/kernel/trace/trace.c
>> @@ -4020,11 +4020,17 @@ static int tracing_open(struct inode *inode, struct file *file)
>> /* If this file was open for write, then erase contents */
>> if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
>> int cpu = tracing_get_cpu(inode);
>> + struct trace_buffer *tr_buf = &tr->trace_buffer;
>> +
>> +#ifdef CONFIG_TRACER_MAX_TRACE
>> + if (tr->current_trace->print_max)
>> + tr_buf = &tr->max_buffer;
>> +#endif
>>
>> if (cpu == RING_BUFFER_ALL_CPUS)
>> - tracing_reset_online_cpus(&tr->trace_buffer);
>> + tracing_reset_online_cpus(tr_buf);
>> else
>> - tracing_reset(&tr->trace_buffer, cpu);
>> + tracing_reset(tr_buf, cpu);
>> }
>>
>> if (file->f_mode & FMODE_READ) {
>

2017-09-19 17:46:34

by Steven Rostedt

Subject: Re: [PATCH] tracing: erase irqsoff trace with empty write

On Tue, 19 Sep 2017 10:41:34 -0700
Bo Yan <[email protected]> wrote:

> Steven,
>
> Thanks for review. I have posted the following update
>
> https://lkml.org/lkml/2017/9/18/715
>

Yes, I actually applied it. It's pushed to my ftrace/urgent branch, and
I'm currently testing it. If/when it passes I'll be pushing it off to
Linus.

Thanks!

-- Steve