[PATCH trace-cmd v2 0/3] trace-cmd filtering

[PATCH trace-cmd v2 0/3] trace-cmd filtering

[Gabriel Krisman Bertazi]

Hi Steve,

This implements my fix to restore single input filtering behavior
(bz217038) and adds support for global filters, as you requested.

On another topic, a current behavior that seems weird in my opinion is
that the following negates the second filter as well.

  "trace-cmd -i trace.dat.1 -v -F tp1 -i trace.dat.2 -F tp2"

I'd prefer that -v would apply only to the following -F. It'd allow me
to do:

  "trace-cmd -v -F 'mm_page_alloc' -i trace.dat.1 -i trace.dat.2 ... \
             -i trace.dat.10 -F 'mm_page_alloc:order==1'"

It obviously breaks the interface, so I didn't implement it here.  Would
like to hear your input, though.  We could have a new syntax:

  "trace-cmd ! -F tp1 -F tp2 ! -F tp3"

Thanks,

Gabriel Krisman Bertazi (3):
  trace-cmd report: Ensure filter is applied to single input file
  trace-cmd-report: Support global filters
  documentation: trace-cmd-report: Document filter scope

 .../trace-cmd/trace-cmd-report.1.txt          |  8 +++-
 tracecmd/trace-read.c                         | 45 ++++++++++++-------
 2 files changed, 34 insertions(+), 19 deletions(-)

-- 
2.40.0

[PATCH trace-cmd v2 1/3] trace-cmd report: Ensure filter is applied to single input file

[Gabriel Krisman Bertazi]

Since 955d05fc7aee ("trace-cmd report: Make filter arguments match their
files"), the -F filtering is silently ignored when a trace file is
provided with -i and the filter comes after -i .  The reason is that the
filter is now associated with input_files and not saved to the global
list only in this case, but process_filters still only checks the global
list when handles->input_file is not set.

Avoid this by checking last_input_file first, which always contains a
pointer to the correct filter in this case.

This was only lightly tested, using a single trace file.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=217038
Fixes: 955d05fc7aee ("trace-cmd report: Make filter arguments match their files")
Signed-off-by: Gabriel krisman Bertazi <krisman@suse.de>

---
Since v1:
  - use rev. xmas tree for variable declaration (steve)
---
 tracecmd/trace-read.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tracecmd/trace-read.c b/tracecmd/trace-read.c
index 52ba818..6a872d7 100644
--- a/tracecmd/trace-read.c
+++ b/tracecmd/trace-read.c
@@ -564,14 +564,15 @@ static void make_pid_filter(struct tracecmd_input *handle,
 
 static void process_filters(struct handle_list *handles)
 {
+	struct input_files *input_file = handles->input_file ?: last_input_file;
 	struct tracecmd_filter *trace_filter;
 	struct filter_str *filter;
 	int filters = 0;
 
-	make_pid_filter(handles->handle, handles->input_file);
+	make_pid_filter(handles->handle, input_file);
 
-	if (handles->input_file)
-		filter = handles->input_file->filter_str;
+	if (input_file)
+		filter = input_file->filter_str;
 	else
 		filter = filter_strings;
 
-- 
2.40.0

[PATCH trace-cmd v2 2/3] trace-cmd-report: Support global filters

[Gabriel Krisman Bertazi]

When a filter is provided before an input file, consider it globally
applying to all traces:

i.e., the following will trace kernel_stack out of both traces instead
of only the first:

  $ trace-cmd report -v -F kernel_stack -i trace1.dat -i trace2.dat

Filter order matters due to the order of processing. Always process
global filters first, which are likely more generic.

Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 tracecmd/trace-read.c | 42 ++++++++++++++++++++++++++----------------
 1 file changed, 26 insertions(+), 16 deletions(-)

diff --git a/tracecmd/trace-read.c b/tracecmd/trace-read.c
index 6a872d7..f645c05 100644
--- a/tracecmd/trace-read.c
+++ b/tracecmd/trace-read.c
@@ -562,30 +562,40 @@ static void make_pid_filter(struct tracecmd_input *handle,
 	}
 }
 
+static int __process_filters(struct tracecmd_input *handle,
+			     struct filter_str *filters)
+{
+	struct tracecmd_filter *trace_filter;
+
+	for (; filters; filters = filters->next) {
+		trace_filter = tracecmd_filter_add(handle,
+						   filters->filter,
+						   filters->neg);
+		if (!trace_filter)
+			die("Failed to create event filter: %s",
+			    filters->filter);
+	}
+
+	return !!filters;
+}
+
 static void process_filters(struct handle_list *handles)
 {
 	struct input_files *input_file = handles->input_file ?: last_input_file;
-	struct tracecmd_filter *trace_filter;
-	struct filter_str *filter;
-	int filters = 0;
+	int added = 0;
 
 	make_pid_filter(handles->handle, input_file);
 
+	/*
+	 * Order of filter processing matters. Apply the global filters
+	 * before file-specific ones.
+	 */
+	added += __process_filters(handles->handle, filter_strings);
 	if (input_file)
-		filter = input_file->filter_str;
-	else
-		filter = filter_strings;
-
-	for (; filter; filter = filter->next) {
-		trace_filter = tracecmd_filter_add(handles->handle,
-						   filter->filter,
-						   filter->neg);
-		if (!trace_filter)
-			die("Failed to create event filter: %s", filter->filter);
+		added += __process_filters(handles->handle,
+					   input_file->filter_str);
 
-		filters++;
-	}
-	if (filters && test_filters_mode)
+	if (added && test_filters_mode)
 		exit(0);
 }
 
-- 
2.40.0

[PATCH trace-cmd v2 3/3] documentation: trace-cmd-report: Document filter scope

[Gabriel Krisman Bertazi]

Document how filters provided before any input files are global while
filters coming after a -i only apply to that filter.

Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 Documentation/trace-cmd/trace-cmd-report.1.txt | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/Documentation/trace-cmd/trace-cmd-report.1.txt b/Documentation/trace-cmd/trace-cmd-report.1.txt
index 71df9a8..ddbde81 100644
--- a/Documentation/trace-cmd/trace-cmd-report.1.txt
+++ b/Documentation/trace-cmd/trace-cmd-report.1.txt
@@ -69,8 +69,12 @@ OPTIONS
     is only to the microsecond. To see the full timestamp, add the *-t* option.
 
 *-F* 'filter'::
-    Add a filter to limit what events are displayed. The format of the filter
-    is:
+    Add a filter to limit what events are displayed.  Filters defined
+    after an input file (specified with *-i*) only apply to that
+    input file. Filters provided before any input file is given are
+    considered global and apply to all input files.
+
+    The format of the filter is:
 
 [source,bison]
 ----
-- 
2.40.0

Re: [PATCH trace-cmd v2 0/3] trace-cmd filtering

[Gabriel Krisman Bertazi]

Gabriel Krisman Bertazi <krisman@suse.de> writes:

> Hi Steve,

I wanted to ping you regarding this patchset.  Do you intend to take
this as-is?

Thanks!

>
> This implements my fix to restore single input filtering behavior
> (bz217038) and adds support for global filters, as you requested.
>
> On another topic, a current behavior that seems weird in my opinion is
> that the following negates the second filter as well.
>
>   "trace-cmd -i trace.dat.1 -v -F tp1 -i trace.dat.2 -F tp2"
>
> I'd prefer that -v would apply only to the following -F. It'd allow me
> to do:
>
>   "trace-cmd -v -F 'mm_page_alloc' -i trace.dat.1 -i trace.dat.2 ... \
>              -i trace.dat.10 -F 'mm_page_alloc:order==1'"
>
> It obviously breaks the interface, so I didn't implement it here.  Would
> like to hear your input, though.  We could have a new syntax:
>
>   "trace-cmd ! -F tp1 -F tp2 ! -F tp3"
>
> Thanks,
>
> Gabriel Krisman Bertazi (3):
>   trace-cmd report: Ensure filter is applied to single input file
>   trace-cmd-report: Support global filters
>   documentation: trace-cmd-report: Document filter scope
>
>  .../trace-cmd/trace-cmd-report.1.txt          |  8 +++-
>  tracecmd/trace-read.c                         | 45 ++++++++++++-------
>  2 files changed, 34 insertions(+), 19 deletions(-)

-- 
Gabriel Krisman Bertazi

Re: [PATCH trace-cmd v2 0/3] trace-cmd filtering

[Steven Rostedt]

On Sat, 06 May 2023 18:35:34 -0400
Gabriel Krisman Bertazi <krisman@suse.de> wrote:

> Gabriel Krisman Bertazi <krisman@suse.de> writes:
> 
> > Hi Steve,  
> 
> I wanted to ping you regarding this patchset.  Do you intend to take
> this as-is?
> 

Apologies, I've been mostly focused on the kernel the last few weeks,
but as I'm giving a talk on the tracing libraries on Thursday, I'll be
focusing on them now ;-)

   https://ossna2023.sched.com/event/1K58e

So I'll reply in a few days.

-- Steve

Re: [PATCH trace-cmd v2 0/3] trace-cmd filtering

[Steven Rostedt]

On Thu, 30 Mar 2023 15:52:08 -0300
Gabriel Krisman Bertazi <krisman@suse.de> wrote:

> Hi Steve,

Hi Gabriel,

Sorry for taking so long. I finally got time to start looking at the
libtrace* trace-cmd code, and I queued this up for the next push.

Thanks Gabriel!

-- Steve


> 
> This implements my fix to restore single input filtering behavior
> (bz217038) and adds support for global filters, as you requested.
> 
> On another topic, a current behavior that seems weird in my opinion is
> that the following negates the second filter as well.
> 
>   "trace-cmd -i trace.dat.1 -v -F tp1 -i trace.dat.2 -F tp2"
> 
> I'd prefer that -v would apply only to the following -F. It'd allow me
> to do:
> 
>   "trace-cmd -v -F 'mm_page_alloc' -i trace.dat.1 -i trace.dat.2 ... \
>              -i trace.dat.10 -F 'mm_page_alloc:order==1'"
> 
> It obviously breaks the interface, so I didn't implement it here.  Would
> like to hear your input, though.  We could have a new syntax:
> 
>   "trace-cmd ! -F tp1 -F tp2 ! -F tp3"
> 
> Thanks,
> 
> Gabriel Krisman Bertazi (3):
>   trace-cmd report: Ensure filter is applied to single input file
>   trace-cmd-report: Support global filters
>   documentation: trace-cmd-report: Document filter scope
> 
>  .../trace-cmd/trace-cmd-report.1.txt          |  8 +++-
>  tracecmd/trace-read.c                         | 45 ++++++++++++-------
>  2 files changed, 34 insertions(+), 19 deletions(-)
>