[RFC PATCH 0/2] Add a test to catch unprobed Devicetree devices

[RFC PATCH 0/2] Add a test to catch unprobed Devicetree devices

[Nícolas F. R. A. Prado]

Regressions that cause a device to no longer be probed by a driver can
have a big impact on the platform's functionality, and despite being
relatively common there isn't currently any generic test to detect them.
As an example, bootrr [1] does test for device probe, but it requires
defining the expected probed devices for each platform.

Given that the Devicetree already provides a static description of
devices on the system, it is a good basis for building such a test on
top.

This series introduces a test to catch regressions that prevent devices
from probing.

Patch 1 introduces a script to parse the kernel source using Coccinelle
and extract all compatibles that can be matched by a Devicetree node to
a driver. Patch 2 adds a kselftest that walks over the Devicetree nodes
on the current platform and compares the compatibles to the ones on the
list, and on an ignore list, to point out devices that failed to be
probed.

A compatible list is needed because not all compatibles that can show up
in a Devicetree node can be used to match to a driver, for example the
code for that compatible might use "OF_DECLARE" type macros and avoid
the driver framework, or the node might be controlled by a driver that
was bound to a different node.

An ignore list is needed for the few cases where it's common for a
driver to match a device but not probe, like for the "simple-mfd"
compatible, where the driver only probes if that compatible is the
node's first compatible.

Even though there's already scripts/dtc/dt-extract-compatibles that does
a similar job, it didn't seem to find all compatibles, returning ~3k,
while Coccinelle found ~11k. Besides that, Coccinelle actually parses
the C files, so it should be a more robust solution than relying on
regexes.

The reason for parsing the kernel source instead of relying on
information exposed by the kernel at runtime (say, looking at modaliases
or introducing some other mechanism), is to be able to catch issues
where a config was renamed or a driver moved across configs, and the
.config used by the kernel not updated accordingly. We need to parse the
source to find all compatibles present in the kernel independent of the
current config being run.

Feedback is very much welcome.

Thanks,
Nícolas

[1] https://github.com/kernelci/bootrr


Nícolas F. R. A. Prado (2):
  scripts/dtc: Add script to extract matchable DT compatibles
  kselftest: Add Devicetree unprobed devices test

 scripts/dtc/extract-matchable-dt-compatibles  | 33 +++++++++++
 scripts/dtc/matchable_dt_compatibles.cocci    | 58 +++++++++++++++++++
 tools/testing/selftests/Makefile              |  1 +
 tools/testing/selftests/dt/.gitignore         |  1 +
 tools/testing/selftests/dt/Makefile           | 17 ++++++
 .../selftests/dt/compatible_ignore_list       |  3 +
 .../selftests/dt/test_unprobed_devices.sh     | 58 +++++++++++++++++++
 7 files changed, 171 insertions(+)
 create mode 100755 scripts/dtc/extract-matchable-dt-compatibles
 create mode 100644 scripts/dtc/matchable_dt_compatibles.cocci
 create mode 100644 tools/testing/selftests/dt/.gitignore
 create mode 100644 tools/testing/selftests/dt/Makefile
 create mode 100644 tools/testing/selftests/dt/compatible_ignore_list
 create mode 100755 tools/testing/selftests/dt/test_unprobed_devices.sh

-- 
2.41.0

[RFC PATCH 1/2] scripts/dtc: Add script to extract matchable DT compatibles

[Nícolas F. R. A. Prado]

Introduce a script to extract all compatibles that can match a device
described on the Devicetree to a driver.

The compatible extraction is done by running Coccinelle with a newly
introduced Coccinelle file that detects compatibles listed inside a
of_device_id table which is referenced by the of_match_table of a
driver struct.

Signed-off-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>

---

 scripts/dtc/extract-matchable-dt-compatibles | 33 +++++++++++
 scripts/dtc/matchable_dt_compatibles.cocci   | 58 ++++++++++++++++++++
 2 files changed, 91 insertions(+)
 create mode 100755 scripts/dtc/extract-matchable-dt-compatibles
 create mode 100644 scripts/dtc/matchable_dt_compatibles.cocci

diff --git a/scripts/dtc/extract-matchable-dt-compatibles b/scripts/dtc/extract-matchable-dt-compatibles
new file mode 100755
index 000000000000..b5e96c7ba42e
--- /dev/null
+++ b/scripts/dtc/extract-matchable-dt-compatibles
@@ -0,0 +1,33 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Copyright (c) 2023 Collabora Ltd
+#
+# Based on coccicheck script.
+#
+# This script uses Coccinelle to extract all compatibles that can be matched by
+# Devicetree devices from the kernel source.
+
+DIR="$(dirname $(readlink -f $0))"
+SPATCH=`which ${SPATCH:=spatch}`
+COCCI_FILE=$DIR/matchable_dt_compatibles.cocci
+
+if [ ! -x "$SPATCH" ]; then
+    echo 'spatch is part of the Coccinelle project and is available at http://coccinelle.lip6.fr/'
+    exit 1
+fi
+
+# Use only one thread per core by default if hyperthreading is enabled
+THREADS_PER_CORE=$(LANG=C lscpu | grep "Thread(s) per core: " | tr -cd "[:digit:]")
+if [ -z "$J" ]; then
+	NPROC=$(getconf _NPROCESSORS_ONLN)
+	if [ $THREADS_PER_CORE -gt 1 -a $NPROC -gt 4 ] ; then
+		NPROC=$((NPROC/2))
+	fi
+else
+	NPROC="$J"
+fi
+
+"$SPATCH" --cocci-file "$COCCI_FILE" --dir .  --jobs "$NPROC" --chunksize 1 \
+	--no-show-diff --no-includes --include-headers --very-quiet \
+	| grep '"' | sed -e 's/"//g;s/^ *\([^ ]*\) *$/\1/'
diff --git a/scripts/dtc/matchable_dt_compatibles.cocci b/scripts/dtc/matchable_dt_compatibles.cocci
new file mode 100644
index 000000000000..ecd3705681aa
--- /dev/null
+++ b/scripts/dtc/matchable_dt_compatibles.cocci
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: GPL-2.0-only
+// Copyright: (C) 2023 Collabora Ltd
+@rule1@
+identifier table;
+type drv;
+identifier drvname;
+identifier drvfield;
+identifier drvfield1;
+identifier drvfield2;
+@@
+
+(
+drv drvname = {
+	.drvfield = {
+		.of_match_table = table,
+	},
+};
+|
+drv drvname = {
+	.drvfield = {
+		.of_match_table = of_match_ptr(table),
+	},
+};
+|
+// Accounts for mdio and spimem drivers
+drv drvname = {
+	.drvfield1 = {
+		.drvfield2 = {
+			.of_match_table = table,
+		},
+	},
+};
+|
+drv drvname = {
+	.drvfield1 = {
+		.drvfield2 = {
+			.of_match_table = of_match_ptr(table),
+		},
+	},
+};
+)
+
+@rule2@
+identifier rule1.table;
+expression compat;
+@@
+
+struct of_device_id table[] = {
+...,
+{ .compatible = compat, },
+...
+};
+
+@script:python@
+compat << rule2.compat;
+@@
+
+print("%s" % compat)
-- 
2.41.0

Re: [RFC PATCH 0/2] Add a test to catch unprobed Devicetree devices

[Rob Herring]

On Thu, Aug 10, 2023 at 04:23:49PM -0400, Nícolas F. R. A. Prado wrote:
> 
> Regressions that cause a device to no longer be probed by a driver can
> have a big impact on the platform's functionality, and despite being
> relatively common there isn't currently any generic test to detect them.
> As an example, bootrr [1] does test for device probe, but it requires
> defining the expected probed devices for each platform.
> 
> Given that the Devicetree already provides a static description of
> devices on the system, it is a good basis for building such a test on
> top.
> 
> This series introduces a test to catch regressions that prevent devices
> from probing.
> 
> Patch 1 introduces a script to parse the kernel source using Coccinelle
> and extract all compatibles that can be matched by a Devicetree node to
> a driver. Patch 2 adds a kselftest that walks over the Devicetree nodes
> on the current platform and compares the compatibles to the ones on the
> list, and on an ignore list, to point out devices that failed to be
> probed.
> 
> A compatible list is needed because not all compatibles that can show up
> in a Devicetree node can be used to match to a driver, for example the
> code for that compatible might use "OF_DECLARE" type macros and avoid
> the driver framework, or the node might be controlled by a driver that
> was bound to a different node.
> 
> An ignore list is needed for the few cases where it's common for a
> driver to match a device but not probe, like for the "simple-mfd"
> compatible, where the driver only probes if that compatible is the
> node's first compatible.
> 
> Even though there's already scripts/dtc/dt-extract-compatibles that does
> a similar job, it didn't seem to find all compatibles, returning ~3k,
> while Coccinelle found ~11k. Besides that, Coccinelle actually parses
> the C files, so it should be a more robust solution than relying on
> regexes.

I just sent a patch[1] last week fixing missing a bunch. I only looked 
at the change in count of undocumented (by schema) though.

In any case, I'm happy if we have a better solution, but really we 
should only have 1. So your script would need to replace the existing 
one.

I'd be interested in a performance comparison. IME, coccinelle is 
fairly slow. Slower is okay to a point though.

> 
> The reason for parsing the kernel source instead of relying on
> information exposed by the kernel at runtime (say, looking at modaliases
> or introducing some other mechanism), is to be able to catch issues
> where a config was renamed or a driver moved across configs, and the
> .config used by the kernel not updated accordingly. We need to parse the
> source to find all compatibles present in the kernel independent of the
> current config being run.

I've been down this route. I had another implementation using gdb to 
extract all of_device_id objects from a built kernel, but besides the 
build time, it was really slow.

Rob

[1] https://lore.kernel.org/all/20230804190130.1936566-1-robh@kernel.org/

Re: [RFC PATCH 0/2] Add a test to catch unprobed Devicetree devices

[Rob Herring]

On Thu, Aug 10, 2023 at 03:43:09PM -0600, Rob Herring wrote:
> On Thu, Aug 10, 2023 at 04:23:49PM -0400, Nícolas F. R. A. Prado wrote:
> > 
> > Regressions that cause a device to no longer be probed by a driver can
> > have a big impact on the platform's functionality, and despite being
> > relatively common there isn't currently any generic test to detect them.
> > As an example, bootrr [1] does test for device probe, but it requires
> > defining the expected probed devices for each platform.
> > 
> > Given that the Devicetree already provides a static description of
> > devices on the system, it is a good basis for building such a test on
> > top.
> > 
> > This series introduces a test to catch regressions that prevent devices
> > from probing.
> > 
> > Patch 1 introduces a script to parse the kernel source using Coccinelle
> > and extract all compatibles that can be matched by a Devicetree node to
> > a driver. Patch 2 adds a kselftest that walks over the Devicetree nodes
> > on the current platform and compares the compatibles to the ones on the
> > list, and on an ignore list, to point out devices that failed to be
> > probed.
> > 
> > A compatible list is needed because not all compatibles that can show up
> > in a Devicetree node can be used to match to a driver, for example the
> > code for that compatible might use "OF_DECLARE" type macros and avoid
> > the driver framework, or the node might be controlled by a driver that
> > was bound to a different node.
> > 
> > An ignore list is needed for the few cases where it's common for a
> > driver to match a device but not probe, like for the "simple-mfd"
> > compatible, where the driver only probes if that compatible is the
> > node's first compatible.
> > 
> > Even though there's already scripts/dtc/dt-extract-compatibles that does
> > a similar job, it didn't seem to find all compatibles, returning ~3k,
> > while Coccinelle found ~11k. Besides that, Coccinelle actually parses
> > the C files, so it should be a more robust solution than relying on
> > regexes.
> 
> I just sent a patch[1] last week fixing missing a bunch. I only looked 
> at the change in count of undocumented (by schema) though.

With the existing script, I get 11761 compatibles and 12916 with my fix. 
So how are you getting only 3k?

Rob

Re: [RFC PATCH 0/2] Add a test to catch unprobed Devicetree devices

[Nícolas F. R. A. Prado]

On Thu, Aug 10, 2023 at 03:43:09PM -0600, Rob Herring wrote:
> On Thu, Aug 10, 2023 at 04:23:49PM -0400, Nícolas F. R. A. Prado wrote:
> > 
> > Regressions that cause a device to no longer be probed by a driver can
> > have a big impact on the platform's functionality, and despite being
> > relatively common there isn't currently any generic test to detect them.
> > As an example, bootrr [1] does test for device probe, but it requires
> > defining the expected probed devices for each platform.
> > 
> > Given that the Devicetree already provides a static description of
> > devices on the system, it is a good basis for building such a test on
> > top.
> > 
> > This series introduces a test to catch regressions that prevent devices
> > from probing.
> > 
> > Patch 1 introduces a script to parse the kernel source using Coccinelle
> > and extract all compatibles that can be matched by a Devicetree node to
> > a driver. Patch 2 adds a kselftest that walks over the Devicetree nodes
> > on the current platform and compares the compatibles to the ones on the
> > list, and on an ignore list, to point out devices that failed to be
> > probed.
> > 
> > A compatible list is needed because not all compatibles that can show up
> > in a Devicetree node can be used to match to a driver, for example the
> > code for that compatible might use "OF_DECLARE" type macros and avoid
> > the driver framework, or the node might be controlled by a driver that
> > was bound to a different node.
> > 
> > An ignore list is needed for the few cases where it's common for a
> > driver to match a device but not probe, like for the "simple-mfd"
> > compatible, where the driver only probes if that compatible is the
> > node's first compatible.
> > 
> > Even though there's already scripts/dtc/dt-extract-compatibles that does
> > a similar job, it didn't seem to find all compatibles, returning ~3k,
> > while Coccinelle found ~11k. Besides that, Coccinelle actually parses
> > the C files, so it should be a more robust solution than relying on
> > regexes.
> 
> I just sent a patch[1] last week fixing missing a bunch. I only looked 
> at the change in count of undocumented (by schema) though.

Ah, looks like I mixed up the output from the dt-extract-compatibles script and
the output from the make dt_compatible_check. The python script does list
practically (*) all compatibles that Coccinelle found. So I'll look into
extending it for the purposes of this test next.

(*) it misses 3 compatibles in .h files, and fsl,mpc5200-gpt-gpio because the
comment above it has ';'. Those are easy to fix though, either on the regex or
on the driver's code.

> 
> In any case, I'm happy if we have a better solution, but really we 
> should only have 1. So your script would need to replace the existing 
> one.

Agreed.

> 
> I'd be interested in a performance comparison. IME, coccinelle is 
> fairly slow. Slower is okay to a point though.

Yes, Coccinelle seems to be quite a bit slower. I can provide a comparison after
I've tweaked the python script to get the same matches as Coccinelle so it is a
fair comparison.

> 
> > 
> > The reason for parsing the kernel source instead of relying on
> > information exposed by the kernel at runtime (say, looking at modaliases
> > or introducing some other mechanism), is to be able to catch issues
> > where a config was renamed or a driver moved across configs, and the
> > .config used by the kernel not updated accordingly. We need to parse the
> > source to find all compatibles present in the kernel independent of the
> > current config being run.
> 
> I've been down this route. I had another implementation using gdb to 
> extract all of_device_id objects from a built kernel, but besides the 
> build time, it was really slow.

Interesting to know, that's another option that I'd considered.

Thanks,
Nícolas