1. Patchew
  2. linux
  3. [PATCH v7 00/16] New perf ilist app
  4. View patch

[PATCH v7 01/16] perf python: Add more exceptions on error paths

Ian Rogers posted 16 patches 2 months, 3 weeks ago
There is a newer version of this series
[PATCH v7 01/16] perf python: Add more exceptions on error paths
Posted by Ian Rogers 2 months, 3 weeks ago 
Returning NULL will cause the python interpreter to fail but not
report an error. If none wants to be returned then Py_None needs
returning. Set the error for the cases returning NULL so that more
meaningful interpreter behavior is had.

Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/util/python.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
index 2f28f71325a8..02544387f39d 100644
--- a/tools/perf/util/python.c
+++ b/tools/perf/util/python.c
@@ -475,13 +475,19 @@ static PyObject *pyrf_event__new(const union perf_event *event)
 	if ((event->header.type < PERF_RECORD_MMAP ||
 	     event->header.type > PERF_RECORD_SAMPLE) &&
 	    !(event->header.type == PERF_RECORD_SWITCH ||
-	      event->header.type == PERF_RECORD_SWITCH_CPU_WIDE))
+	      event->header.type == PERF_RECORD_SWITCH_CPU_WIDE)){
+		PyErr_Format(PyExc_TypeError, "Unexpected header type %u",
+			     event->header.type);
 		return NULL;
+	}
 
 	// FIXME this better be dynamic or we need to parse everything
 	// before calling perf_mmap__consume(), including tracepoint fields.
-	if (sizeof(pevent->event) < event->header.size)
+	if (sizeof(pevent->event) < event->header.size) {
+		PyErr_Format(PyExc_TypeError, "Unexpected event version: %zd < %u",
+			     sizeof(pevent->event), event->header.size);
 		return NULL;
+	}
 
 	ptype = pyrf_event__type[event->header.type];
 	pevent = PyObject_New(struct pyrf_event, ptype);
@@ -1199,8 +1205,10 @@ static PyObject *pyrf_evlist__read_on_cpu(struct pyrf_evlist *pevlist,
 		return NULL;
 
 	md = get_md(evlist, cpu);
-	if (!md)
+	if (!md) {
+		PyErr_Format(PyExc_TypeError, "Unknown CPU '%d'", cpu);
 		return NULL;
+	}
 
 	if (perf_mmap__read_init(&md->core) < 0)
 		goto end;
-- 
2.50.0.727.gbf7dc18ff4-goog
Re: [PATCH v7 01/16] perf python: Add more exceptions on error paths
Posted by Namhyung Kim 2 months, 2 weeks ago 
On Mon, Jul 14, 2025 at 09:43:49AM -0700, Ian Rogers wrote:
> Returning NULL will cause the python interpreter to fail but not
> report an error. If none wants to be returned then Py_None needs
> returning. Set the error for the cases returning NULL so that more
> meaningful interpreter behavior is had.

It looks like you are adding error messages for the failure cases, not
adding new exceptions, right?  IIUC returning NULL in pyrf_event__new()
ends up having PyErr_NoMemory().  Then now it has different messages?

> 
> Signed-off-by: Ian Rogers <irogers@google.com>
> ---
>  tools/perf/util/python.c | 14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
> index 2f28f71325a8..02544387f39d 100644
> --- a/tools/perf/util/python.c
> +++ b/tools/perf/util/python.c
> @@ -475,13 +475,19 @@ static PyObject *pyrf_event__new(const union perf_event *event)
>  	if ((event->header.type < PERF_RECORD_MMAP ||
>  	     event->header.type > PERF_RECORD_SAMPLE) &&
>  	    !(event->header.type == PERF_RECORD_SWITCH ||
> -	      event->header.type == PERF_RECORD_SWITCH_CPU_WIDE))
> +	      event->header.type == PERF_RECORD_SWITCH_CPU_WIDE)){
> +		PyErr_Format(PyExc_TypeError, "Unexpected header type %u",
> +			     event->header.type);
>  		return NULL;
> +	}
>  
>  	// FIXME this better be dynamic or we need to parse everything
>  	// before calling perf_mmap__consume(), including tracepoint fields.
> -	if (sizeof(pevent->event) < event->header.size)
> +	if (sizeof(pevent->event) < event->header.size) {
> +		PyErr_Format(PyExc_TypeError, "Unexpected event version: %zd < %u",

Maybe "Unexpected event size" instead?

Thanks,
Namhyung


> +			     sizeof(pevent->event), event->header.size);
>  		return NULL;
> +	}
>  
>  	ptype = pyrf_event__type[event->header.type];
>  	pevent = PyObject_New(struct pyrf_event, ptype);
> @@ -1199,8 +1205,10 @@ static PyObject *pyrf_evlist__read_on_cpu(struct pyrf_evlist *pevlist,
>  		return NULL;
>  
>  	md = get_md(evlist, cpu);
> -	if (!md)
> +	if (!md) {
> +		PyErr_Format(PyExc_TypeError, "Unknown CPU '%d'", cpu);
>  		return NULL;
> +	}
>  
>  	if (perf_mmap__read_init(&md->core) < 0)
>  		goto end;
> -- 
> 2.50.0.727.gbf7dc18ff4-goog
>
Re: [PATCH v7 01/16] perf python: Add more exceptions on error paths
Posted by Ian Rogers 2 months, 2 weeks ago 
On Wed, Jul 23, 2025 at 11:13 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Mon, Jul 14, 2025 at 09:43:49AM -0700, Ian Rogers wrote:
> > Returning NULL will cause the python interpreter to fail but not
> > report an error. If none wants to be returned then Py_None needs
> > returning. Set the error for the cases returning NULL so that more
> > meaningful interpreter behavior is had.
>
> It looks like you are adding error messages for the failure cases, not
> adding new exceptions, right?  IIUC returning NULL in pyrf_event__new()
> ends up having PyErr_NoMemory().  Then now it has different messages?

This change doesn't alter the exception if there already is one. There
are lots of cases where NULL is being returned for an error but that
causes the python interpreter to fail/crash. This patch is just adding
the exceptions and still returning NULL, previously set exceptions are
left alone. With the exceptions returned the python interpreter does
something more useful than fail/crash :-)

> >
> > Signed-off-by: Ian Rogers <irogers@google.com>
> > ---
> >  tools/perf/util/python.c | 14 +++++++++++---
> >  1 file changed, 11 insertions(+), 3 deletions(-)
> >
> > diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
> > index 2f28f71325a8..02544387f39d 100644
> > --- a/tools/perf/util/python.c
> > +++ b/tools/perf/util/python.c
> > @@ -475,13 +475,19 @@ static PyObject *pyrf_event__new(const union perf_event *event)
> >       if ((event->header.type < PERF_RECORD_MMAP ||
> >            event->header.type > PERF_RECORD_SAMPLE) &&
> >           !(event->header.type == PERF_RECORD_SWITCH ||
> > -           event->header.type == PERF_RECORD_SWITCH_CPU_WIDE))
> > +           event->header.type == PERF_RECORD_SWITCH_CPU_WIDE)){
> > +             PyErr_Format(PyExc_TypeError, "Unexpected header type %u",
> > +                          event->header.type);
> >               return NULL;
> > +     }
> >
> >       // FIXME this better be dynamic or we need to parse everything
> >       // before calling perf_mmap__consume(), including tracepoint fields.
> > -     if (sizeof(pevent->event) < event->header.size)
> > +     if (sizeof(pevent->event) < event->header.size) {
> > +             PyErr_Format(PyExc_TypeError, "Unexpected event version: %zd < %u",
>
> Maybe "Unexpected event size" instead?

I think size is more accurate to the code, version is more useful to
the user. I believe there is existing use of the event size being used
as a quasi version number, so I think this description is consistent.
If you feel strongly I can change it.

Thanks,
Ian

> Thanks,
> Namhyung
>
>
> > +                          sizeof(pevent->event), event->header.size);
> >               return NULL;
> > +     }
> >
> >       ptype = pyrf_event__type[event->header.type];
> >       pevent = PyObject_New(struct pyrf_event, ptype);
> > @@ -1199,8 +1205,10 @@ static PyObject *pyrf_evlist__read_on_cpu(struct pyrf_evlist *pevlist,
> >               return NULL;
> >
> >       md = get_md(evlist, cpu);
> > -     if (!md)
> > +     if (!md) {
> > +             PyErr_Format(PyExc_TypeError, "Unknown CPU '%d'", cpu);
> >               return NULL;
> > +     }
> >
> >       if (perf_mmap__read_init(&md->core) < 0)
> >               goto end;
> > --
> > 2.50.0.727.gbf7dc18ff4-goog
> >
Re: [PATCH v7 01/16] perf python: Add more exceptions on error paths
Posted by Ian Rogers 2 months, 2 weeks ago 
On Wed, Jul 23, 2025 at 11:21 AM Ian Rogers <irogers@google.com> wrote:
>
> On Wed, Jul 23, 2025 at 11:13 AM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > On Mon, Jul 14, 2025 at 09:43:49AM -0700, Ian Rogers wrote:
> > > Returning NULL will cause the python interpreter to fail but not
> > > report an error. If none wants to be returned then Py_None needs
> > > returning. Set the error for the cases returning NULL so that more
> > > meaningful interpreter behavior is had.
> >
> > It looks like you are adding error messages for the failure cases, not
> > adding new exceptions, right?  IIUC returning NULL in pyrf_event__new()
> > ends up having PyErr_NoMemory().  Then now it has different messages?
>
> This change doesn't alter the exception if there already is one. There
> are lots of cases where NULL is being returned for an error but that
> causes the python interpreter to fail/crash. This patch is just adding
> the exceptions and still returning NULL, previously set exceptions are
> left alone. With the exceptions returned the python interpreter does
> something more useful than fail/crash :-)
>
> > >
> > > Signed-off-by: Ian Rogers <irogers@google.com>
> > > ---
> > >  tools/perf/util/python.c | 14 +++++++++++---
> > >  1 file changed, 11 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
> > > index 2f28f71325a8..02544387f39d 100644
> > > --- a/tools/perf/util/python.c
> > > +++ b/tools/perf/util/python.c
> > > @@ -475,13 +475,19 @@ static PyObject *pyrf_event__new(const union perf_event *event)
> > >       if ((event->header.type < PERF_RECORD_MMAP ||
> > >            event->header.type > PERF_RECORD_SAMPLE) &&
> > >           !(event->header.type == PERF_RECORD_SWITCH ||
> > > -           event->header.type == PERF_RECORD_SWITCH_CPU_WIDE))
> > > +           event->header.type == PERF_RECORD_SWITCH_CPU_WIDE)){
> > > +             PyErr_Format(PyExc_TypeError, "Unexpected header type %u",
> > > +                          event->header.type);
> > >               return NULL;
> > > +     }
> > >
> > >       // FIXME this better be dynamic or we need to parse everything
> > >       // before calling perf_mmap__consume(), including tracepoint fields.
> > > -     if (sizeof(pevent->event) < event->header.size)
> > > +     if (sizeof(pevent->event) < event->header.size) {
> > > +             PyErr_Format(PyExc_TypeError, "Unexpected event version: %zd < %u",
> >
> > Maybe "Unexpected event size" instead?
>
> I think size is more accurate to the code, version is more useful to
> the user. I believe there is existing use of the event size being used
> as a quasi version number, so I think this description is consistent.
> If you feel strongly I can change it.

Actually I was thinking of the attribute size, let me fix this.

Thanks,
Ian

> Thanks,
> Ian
>
> > Thanks,
> > Namhyung
> >
> >
> > > +                          sizeof(pevent->event), event->header.size);
> > >               return NULL;
> > > +     }
> > >
> > >       ptype = pyrf_event__type[event->header.type];
> > >       pevent = PyObject_New(struct pyrf_event, ptype);
> > > @@ -1199,8 +1205,10 @@ static PyObject *pyrf_evlist__read_on_cpu(struct pyrf_evlist *pevlist,
> > >               return NULL;
> > >
> > >       md = get_md(evlist, cpu);
> > > -     if (!md)
> > > +     if (!md) {
> > > +             PyErr_Format(PyExc_TypeError, "Unknown CPU '%d'", cpu);
> > >               return NULL;
> > > +     }
> > >
> > >       if (perf_mmap__read_init(&md->core) < 0)
> > >               goto end;
> > > --
> > > 2.50.0.727.gbf7dc18ff4-goog
> > >
Re: [PATCH v7 01/16] perf python: Add more exceptions on error paths
Posted by Arnaldo Carvalho de Melo 2 months, 2 weeks ago 
On Mon, Jul 14, 2025 at 09:43:49AM -0700, Ian Rogers wrote:
> Returning NULL will cause the python interpreter to fail but not
> report an error. If none wants to be returned then Py_None needs
> returning. Set the error for the cases returning NULL so that more
> meaningful interpreter behavior is had.
> 
> Signed-off-by: Ian Rogers <irogers@google.com>
> ---
>  tools/perf/util/python.c | 14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
> index 2f28f71325a8..02544387f39d 100644
> --- a/tools/perf/util/python.c
> +++ b/tools/perf/util/python.c
> @@ -475,13 +475,19 @@ static PyObject *pyrf_event__new(const union perf_event *event)
>  	if ((event->header.type < PERF_RECORD_MMAP ||
>  	     event->header.type > PERF_RECORD_SAMPLE) &&
>  	    !(event->header.type == PERF_RECORD_SWITCH ||
> -	      event->header.type == PERF_RECORD_SWITCH_CPU_WIDE))
> +	      event->header.type == PERF_RECORD_SWITCH_CPU_WIDE)){

Nit, missing space before {

> +		PyErr_Format(PyExc_TypeError, "Unexpected header type %u",
> +			     event->header.type);
>  		return NULL;
> +	}
>  
>  	// FIXME this better be dynamic or we need to parse everything
>  	// before calling perf_mmap__consume(), including tracepoint fields.
> -	if (sizeof(pevent->event) < event->header.size)
> +	if (sizeof(pevent->event) < event->header.size) {
> +		PyErr_Format(PyExc_TypeError, "Unexpected event version: %zd < %u",
> +			     sizeof(pevent->event), event->header.size);
>  		return NULL;
> +	}
>  
>  	ptype = pyrf_event__type[event->header.type];
>  	pevent = PyObject_New(struct pyrf_event, ptype);
> @@ -1199,8 +1205,10 @@ static PyObject *pyrf_evlist__read_on_cpu(struct pyrf_evlist *pevlist,
>  		return NULL;
>  
>  	md = get_md(evlist, cpu);
> -	if (!md)
> +	if (!md) {
> +		PyErr_Format(PyExc_TypeError, "Unknown CPU '%d'", cpu);
>  		return NULL;
> +	}
>  
>  	if (perf_mmap__read_init(&md->core) < 0)
>  		goto end;
> -- 
> 2.50.0.727.gbf7dc18ff4-goog

Patchew 2.1-devel-b67e4c2

© 2016 - 2025 Red Hat, Inc.