Skip to contentSkip to conversation

scripts/config: honor $KBUILD_OUTPUT by default

PATCH Superseded Dmitry Voytik 2 participants 5 messages Sep 27 – Oct 1 Kbuild

v3 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 scripts/config: honor $KBUILD_OUTPUT by default

If $KBUILD_OUTPUT is set, use it by default if --file is not passed.

Fix the issue when the script silently updates a stale .config in the root directory, while the user expects ${KBUILD_OUTPUT}/.config to be updated.
The problem can be worked around with --file, but this is easy to miss, and is inconsistent with how the Makefile and other tools work (e.g., scripts/diffconfig).

How to reproduce the issue:

	rm .config
	export KBUILD_OUTPUT=.out
	make defconfig
	./scripts/config \
	       -e CONFIG_UBSAN
	grep: .config: No such file or directory

---
v3:
- Added an example in the commit message how to reproduce the issue
v2:
- Addressed comments by sashiko:
- https://sashiko.dev/#/patchset/20260920064350.17999-1-voytikd%40gmail.com
- use basename when --refresh is used
- remove echo which can be misleading when --file is passed
v1:
- https://lore.kernel.org/all/[email protected]/

---
 scripts/config | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
scripts/config +5 -1
index 32428ea909c2..cdee67157fe1 100755
@@ -130,6 +130,10 @@ on_exit() {
 trap on_exit EXIT
 
 FN=.config
+if [ -n "${KBUILD_OUTPUT}" ]; then
+	FN=${KBUILD_OUTPUT}/${FN}
+fi
+
 CMDS=()
 while [[ $# -gt 0 ]]; do
 	if [ "$1" = "--file" ]; then
@@ -229,7 +233,7 @@ while [ "$1" != "" ] ; do
 		;;
 
 	--refresh)
-		yes "" | make oldconfig KCONFIG_CONFIG=$FN
+		yes "" | make oldconfig KCONFIG_CONFIG=$(basename $FN)
 		;;
 
 	*)
Signature · 1 line
2.55.0
maintainer
#1 lore raw
Quoted 48 lines · Dmitry Voytik

If $KBUILD_OUTPUT is set, use it by default if --file is not passed.

Fix the issue when the script silently updates a stale .config in the root directory, while the user expects ${KBUILD_OUTPUT}/.config to be updated.
The problem can be worked around with --file, but this is easy to miss, and is inconsistent with how the Makefile and other tools work (e.g., scripts/diffconfig).

How to reproduce the issue:

	rm .config
	export KBUILD_OUTPUT=.out
	make defconfig
	./scripts/config \
	       -e CONFIG_UBSAN
	grep: .config: No such file or directory

---
v3:
- Added an example in the commit message how to reproduce the issue
v2:
- Addressed comments by sashiko:
- https://sashiko.dev/#/patchset/20260920064350.17999-1-voytikd%40gmail.com
- use basename when --refresh is used
- remove echo which can be misleading when --file is passed
v1:
- https://lore.kernel.org/all/[email protected]/

---
 scripts/config | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
scripts/config +4 -0
index 32428ea909c2..cdee67157fe1 100755
@@ -130,6 +130,10 @@ on_exit() {
 trap on_exit EXIT
 
 FN=.config
+if [ -n "${KBUILD_OUTPUT}" ]; then
+	FN=${KBUILD_OUTPUT}/${FN}
+fi
+
 CMDS=()
 while [[ $# -gt 0 ]]; do
 	if [ "$1" = "--file" ]; then

scripts/config while [ "$1" != "" ] ; do · line 236

Quoted 5 lines
scripts/config +1 -1
@@ -229,7 +233,7 @@ while [ "$1" != "" ] ; do
 		;;
 
 	--refresh)
-		yes "" | make oldconfig KCONFIG_CONFIG=$FN
+		yes "" | make oldconfig KCONFIG_CONFIG=$(basename $FN)

Thanks for attempting to fix the '--refresh' option. But this way, this may lead to very unexpected behaviour, as it regardlessly modifies the path to the config file, even if it was specified via --file.

If --file is given, it should take precedence over any KBUILD_OUTPUT (or KCONFIG_CONFIG) from the environment and not be changed unrequested.

It would be a tiny bit nicer to just strip the KBUILD_OUTPUT path from FN, e.g. (untested):

		yes "" | make oldconfig KCONFIG_CONFIG="${FN#${KBUILD_OUTPUT}/}"

But I am still not feeling happy with that.

Signature · 1 line
Nicolas
#2 lore raw

Hi Nicolas,

Thank you for taking your time and reviewing the patch.

Quoted 60 lines · Nicolas Schier
Quoted 48 lines · Dmitry Voytik

If $KBUILD_OUTPUT is set, use it by default if --file is not passed.

Fix the issue when the script silently updates a stale .config in the root directory, while the user expects ${KBUILD_OUTPUT}/.config to be updated.
The problem can be worked around with --file, but this is easy to miss, and is inconsistent with how the Makefile and other tools work (e.g., scripts/diffconfig).

How to reproduce the issue:
rm .config
export KBUILD_OUTPUT=.out
make defconfig
./scripts/config \
-e CONFIG_UBSAN
grep: .config: No such file or directory

---
v3:
- Added an example in the commit message how to reproduce the issue
v2:
- Addressed comments by sashiko:
- https://sashiko.dev/#/patchset/20260920064350.17999-1-voytikd%40gmail.com
- use basename when --refresh is used
- remove echo which can be misleading when --file is passed
v1:
- https://lore.kernel.org/all/[email protected]/

---
 scripts/config | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
scripts/config +4 -0
index 32428ea909c2..cdee67157fe1 100755
@@ -130,6 +130,10 @@ on_exit() {
 trap on_exit EXIT
 
 FN=.config
+if [ -n "${KBUILD_OUTPUT}" ]; then
+     FN=${KBUILD_OUTPUT}/${FN}
+fi
+
 CMDS=()
 while [[ $# -gt 0 ]]; do
      if [ "$1" = "--file" ]; then

scripts/config while [ "$1" != "" ] ; do · line 236

Quoted 5 lines
scripts/config +1 -1
@@ -229,7 +233,7 @@ while [ "$1" != "" ] ; do
              ;;
 
      --refresh)
-             yes "" | make oldconfig KCONFIG_CONFIG=$FN
+             yes "" | make oldconfig KCONFIG_CONFIG=$(basename $FN)

Thanks for attempting to fix the '--refresh' option. But this way, this may lead to very unexpected behaviour, as it regardlessly modifies the path to the config file, even if it was specified via --file.

Quoted 2 lines

If --file is given, it should take precedence over any KBUILD_OUTPUT (or KCONFIG_CONFIG) from the environment and not be changed unrequested.

Thanks. I misunderstood what KBUILD_OUTPUT really means. For some reason, I thought it should be a filename and not a full path. The documentation entry is a bit vague on it.

Quoted 6 lines

It would be a tiny bit nicer to just strip the KBUILD_OUTPUT path from FN, e.g. (untested):

                yes "" | make oldconfig KCONFIG_CONFIG="${FN#${KBUILD_OUTPUT}/}"

But I am still not feeling happy with that.

Thanks for the suggestion! So, what I did - I just completely dropped the second change hunk, i.e. kept it as is:
yes "" | make oldconfig KCONFIG_CONFIG=$FN

I tested the following three scenarios and it looks like it works as expected:

make mrproper
export KBUILD_OUTPUT=.out
make defconfig
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config

# test basic use case:
./scripts/config \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config
./scripts/config -d CONFIG_UBSAN
# test --refresh picks up correct config file by using $KBUILD_OUTPUT
./scripts/config --refresh \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config
./scripts/config -d CONFIG_UBSAN
# test --file takes precedence over $KBUILD_OUTPUT
cp .out/.config .my_config
grep "CONFIG_UBSAN=y" .my_config
./scripts/config --file $(pwd)/.my_config --refresh \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" .my_config

If you are fine with the above, I can send v4 with adding the above testing steps into the commit message.
Thanks!

Quoted 2 lines

--
Nicolas

Signature · 2 lines
Best Regards,
Dmitry Voytik.
maintainer
#3 lore raw
Quoted 102 lines · Dmitry Voytik

Hi Nicolas,

Thank you for taking your time and reviewing the patch.

Quoted 60 lines · Nicolas Schier
Quoted 48 lines · Dmitry Voytik

If $KBUILD_OUTPUT is set, use it by default if --file is not passed.

Fix the issue when the script silently updates a stale .config in the root directory, while the user expects ${KBUILD_OUTPUT}/.config to be updated.
The problem can be worked around with --file, but this is easy to miss, and is inconsistent with how the Makefile and other tools work (e.g., scripts/diffconfig).

How to reproduce the issue:
rm .config
export KBUILD_OUTPUT=.out
make defconfig
./scripts/config \
-e CONFIG_UBSAN
grep: .config: No such file or directory

---
v3:
- Added an example in the commit message how to reproduce the issue
v2:
- Addressed comments by sashiko:
- https://sashiko.dev/#/patchset/20260920064350.17999-1-voytikd%40gmail.com
- use basename when --refresh is used
- remove echo which can be misleading when --file is passed
v1:
- https://lore.kernel.org/all/[email protected]/

---
 scripts/config | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
scripts/config +4 -0
index 32428ea909c2..cdee67157fe1 100755
@@ -130,6 +130,10 @@ on_exit() {
 trap on_exit EXIT
 
 FN=.config
+if [ -n "${KBUILD_OUTPUT}" ]; then
+     FN=${KBUILD_OUTPUT}/${FN}
+fi
+
 CMDS=()
 while [[ $# -gt 0 ]]; do
      if [ "$1" = "--file" ]; then

scripts/config while [ "$1" != "" ] ; do · line 236

Quoted 5 lines
scripts/config +1 -1
@@ -229,7 +233,7 @@ while [ "$1" != "" ] ; do
              ;;
 
      --refresh)
-             yes "" | make oldconfig KCONFIG_CONFIG=$FN
+             yes "" | make oldconfig KCONFIG_CONFIG=$(basename $FN)

Thanks for attempting to fix the '--refresh' option. But this way, this may lead to very unexpected behaviour, as it regardlessly modifies the path to the config file, even if it was specified via --file.

Quoted 2 lines

If --file is given, it should take precedence over any KBUILD_OUTPUT (or KCONFIG_CONFIG) from the environment and not be changed unrequested.

Thanks. I misunderstood what KBUILD_OUTPUT really means. For some reason, I thought it should be a filename and not a full path. The documentation entry is a bit vague on it.

Quoted 6 lines

It would be a tiny bit nicer to just strip the KBUILD_OUTPUT path from FN, e.g. (untested):

                yes "" | make oldconfig KCONFIG_CONFIG="${FN#${KBUILD_OUTPUT}/}"

But I am still not feeling happy with that.

Thanks for the suggestion! So, what I did - I just completely dropped the second change hunk, i.e. kept it as is:
yes "" | make oldconfig KCONFIG_CONFIG=$FN

I tested the following three scenarios and it looks like it works as expected:

make mrproper
export KBUILD_OUTPUT=.out
make defconfig
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config

# test basic use case:
./scripts/config \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config
./scripts/config -d CONFIG_UBSAN
# test --refresh picks up correct config file by using $KBUILD_OUTPUT
./scripts/config --refresh \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config
./scripts/config -d CONFIG_UBSAN
Quoted 6 lines
# test --file takes precedence over $KBUILD_OUTPUT
cp .out/.config .my_config
grep "CONFIG_UBSAN=y" .my_config
./scripts/config --file $(pwd)/.my_config --refresh \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" .my_config

This one shows one of the problems with scripts/config: while the script itself handles relative paths to --file FILE from the current working directory, a call of Kbuild changes the working directory to $(KBUILD_OUTOUT) and realtive paths in KCONFIG_CONFIG are resolved from there.

I don't have a solution in mind, that fits it all. But I think it's important to not make things worse (or more unexpected) than before, by accident.

Quoted 2 lines

If you are fine with the above, I can send v4 with adding the above testing steps into the commit message.

yeah, please send a v4 and then let's see what we'll do with it, ok?

Thanks and kind regards,
Nicolas

#4 lore raw

Hi Nicolas,

Thanks for the review!

Quoted 111 lines · Nicolas Schier
Quoted 102 lines · Dmitry Voytik

Hi Nicolas,

Thank you for taking your time and reviewing the patch.

Quoted 60 lines · Nicolas Schier
Quoted 48 lines · Dmitry Voytik

If $KBUILD_OUTPUT is set, use it by default if --file is not passed.

Fix the issue when the script silently updates a stale .config in the root directory, while the user expects ${KBUILD_OUTPUT}/.config to be updated.
The problem can be worked around with --file, but this is easy to miss, and is inconsistent with how the Makefile and other tools work (e.g., scripts/diffconfig).

How to reproduce the issue:
rm .config
export KBUILD_OUTPUT=.out
make defconfig
./scripts/config \
-e CONFIG_UBSAN
grep: .config: No such file or directory

---
v3:
- Added an example in the commit message how to reproduce the issue
v2:
- Addressed comments by sashiko:
- https://sashiko.dev/#/patchset/20260920064350.17999-1-voytikd%40gmail.com
- use basename when --refresh is used
- remove echo which can be misleading when --file is passed
v1:
- https://lore.kernel.org/all/[email protected]/

---
 scripts/config | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
scripts/config +4 -0
index 32428ea909c2..cdee67157fe1 100755
@@ -130,6 +130,10 @@ on_exit() {
 trap on_exit EXIT
 
 FN=.config
+if [ -n "${KBUILD_OUTPUT}" ]; then
+     FN=${KBUILD_OUTPUT}/${FN}
+fi
+
 CMDS=()
 while [[ $# -gt 0 ]]; do
      if [ "$1" = "--file" ]; then

scripts/config while [ "$1" != "" ] ; do · line 236

Quoted 5 lines
scripts/config +1 -1
@@ -229,7 +233,7 @@ while [ "$1" != "" ] ; do
              ;;
 
      --refresh)
-             yes "" | make oldconfig KCONFIG_CONFIG=$FN
+             yes "" | make oldconfig KCONFIG_CONFIG=$(basename $FN)

Thanks for attempting to fix the '--refresh' option. But this way, this may lead to very unexpected behaviour, as it regardlessly modifies the path to the config file, even if it was specified via --file.

Quoted 2 lines

If --file is given, it should take precedence over any KBUILD_OUTPUT (or KCONFIG_CONFIG) from the environment and not be changed unrequested.

Thanks. I misunderstood what KBUILD_OUTPUT really means. For some reason, I thought it should be a filename and not a full path. The documentation entry is a bit vague on it.

Quoted 6 lines

It would be a tiny bit nicer to just strip the KBUILD_OUTPUT path from FN, e.g. (untested):

                yes "" | make oldconfig KCONFIG_CONFIG="${FN#${KBUILD_OUTPUT}/}"

But I am still not feeling happy with that.

Thanks for the suggestion! So, what I did - I just completely dropped the second change hunk, i.e. kept it as is:
yes "" | make oldconfig KCONFIG_CONFIG=$FN

I tested the following three scenarios and it looks like it works as expected:

make mrproper
export KBUILD_OUTPUT=.out
make defconfig
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config

# test basic use case:
./scripts/config \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config
./scripts/config -d CONFIG_UBSAN
# test --refresh picks up correct config file by using $KBUILD_OUTPUT
./scripts/config --refresh \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" $KBUILD_OUTPUT/.config
./scripts/config -d CONFIG_UBSAN
Quoted 6 lines
# test --file takes precedence over $KBUILD_OUTPUT
cp .out/.config .my_config
grep "CONFIG_UBSAN=y" .my_config
./scripts/config --file $(pwd)/.my_config --refresh \
       -e CONFIG_UBSAN
grep "CONFIG_UBSAN=y" .my_config
Quoted 5 lines

This one shows one of the problems with scripts/config: while the script itself handles relative paths to --file FILE from the current working directory, a call of Kbuild changes the working directory to $(KBUILD_OUTOUT) and realtive paths in KCONFIG_CONFIG are resolved from there.

Indeed, with the patch, if I pass a relative path, it creates the config file in an unexpected place:

       make mrproper
       export KBUILD_OUTPUT=.out
       make defconfig
       mkdir .out2
       cp .out/.config .out2/.my_config
       ./scripts/config --file .out2/.my_config \
              --refresh \
              -e CONFIG_UBSAN
       ls .out/.out2/.my_config

Also, apparently, the following test case with --refresh misbehaves:

       ./scripts/config --refresh \
              -e CONFIG_UBSAN

It creates .out/.out/.config :)

I decided to always enforce converting $FN to the absolute path:

       diff --git a/scripts/config b/scripts/config
       index 4e29c3084fec..0b28fc498195 100755
       --- a/scripts/config
       +++ b/scripts/config
       @@ -147,6 +147,7 @@ while [[ $# -gt 0 ]]; do
       shift
       fi
       done
       +FN=$(realpath "$FN")

       set -- "${CMDS[@]}"
       if [ "$1" = "" ] ; then

With this change, all tests pass, no side effects as it seems.

Quoted 8 lines

I don't have a solution in mind, that fits it all. But I think it's important to not make things worse (or more unexpected) than before, by accident.

Quoted 2 lines

If you are fine with the above, I can send v4 with adding the above testing steps into the commit message.

yeah, please send a v4 and then let's see what we'll do with it, ok?

I'll send a v4 patch.
Thanks for the review!

Quoted 2 lines

Thanks and kind regards,
Nicolas

Best Regards,
Dmitry Voytik.