kbuild: add header check facility as a manually run static analyzer
PATCH 1/1 kbuild: add header check facility as a manually run static analyzer
There have been various attempts at adding a header test or check mechanism in the kernel build system. The header check primarily consists of ensuring headers are self-contained, have include guards, and, in some cases, pass kernel-doc.
The main problems have been:
- The dependency tracking creates undesirable artefacts (infamously also known as disgusting turds) in the build directory.
- Gating the feature behind a kconfig option is complicated due to allyesconfig builds. It's possible, but requires a verbose and confusing negative proxy config option.
- Naming the dependency tracking files with a dot prefix or placing them in a dot prefixed subdirectory in the build directory to hide them has been exceedingly difficult to achieve. (In part due to some Makefiles building files in subdirectory hierarchies.)
- The debate which headers, if any, should really be self-contained is virtually open-ended.
Approach the problem from a slightly different angle. Add a top level "headercheck" static analysis target which you have to explicitly run, and where you have to explicitly state which headers to check:
$ make HEADER_CHECK="(headers|dirs)" headercheck
For example:
$ make HEADER_CHECK="include/drm drivers/gpu/drm/drm_draw_internal.h" headercheck
You can specify multiple path/to/header.h or path/to/dir, relative to $(srctree), where path/to/dir is recursively scanned for any .h files.
Add a generic %.header-check rule in scripts/Makefile.build to support this. Note that this rule should *not* be used in, say, driver or subsystem makefile targets, precisely because of the problems listed above.
While this still generates the .header-check dependency tracking artefacts in the build directory, you will only get them as a result of manually running 'make headercheck', not as part of the regular build.
This approach also allows anyone to check any headers in the tree, with no makefile modification or kconfig changes required. What you do with the results is up to you, and the header check facility does not enforce anything like it would as part of the regular build.
--- .gitignore | 1 + Makefile | 25 ++++++++++++++++++++++++- scripts/Makefile.build | 21 +++++++++++++++++++++ 3 files changed, 46 insertions(+), 1 deletion(-)
.gitignore +1 -0
@@ -28,6 +28,7 @@ *.gcno *.gcda *.gz +*.header-check *.i *.ko *.lex.c
Makefile +24 -1
@@ -301,7 +301,7 @@ no-dot-config-targets := $(clean-targets) \ run-command no-sync-config-targets := $(no-dot-config-targets) %install modules_sign kernelrelease \ image_name -single-targets := %.a %.i %.ko %.lds %.ll %.lst %.mod %.o %.rsi %.s %/ +single-targets := %.a %.i %.ko %.lds %.ll %.lst %.mod %.o %.rsi %.s %.header-check %/ config-build := mixed-build :=
@@ -1540,6 +1540,26 @@ PHONY += scripts_gen_packed_field_checks scripts_gen_packed_field_checks: scripts_basic $(Q)$(MAKE) $(build)=scripts scripts/gen_packed_field_checks +# --------------------------------------------------------------------------- +# Header check + +# Check the headers in HEADER_CHECK=(headers|dirs) +PHONY += headercheck + +ifneq ($(HEADER_CHECK),) + +header-check-dirs := $(filter-out %.h,$(HEADER_CHECK)) +header-check-files := $(filter %.h,$(HEADER_CHECK)) $(if $(header-check-dirs),$(shell cd $(srctree) && find $(header-check-dirs) -name '*.h' 2>/dev/null)) +header-check-targets := $(patsubst %.h,%.header-check,$(sort $(header-check-files))) + +headercheck: + $(if $(header-check-targets),,$(error $@ found no headers in HEADER_CHECK="$(HEADER_CHECK)")) + $(Q)$(MAKE) $(header-check-targets) +else +headercheck: + $(error $@ requires HEADER_CHECK=(headers|dirs)) +endif + # --------------------------------------------------------------------------- # Install
@@ -1886,6 +1906,8 @@ help: @echo ' versioncheck - Sanity check on version.h usage' @echo ' includecheck - Check for duplicate included header files' @echo ' headerdep - Detect inclusion cycles in headers' + @echo ' headercheck - Check headers in HEADER_CHECK=(headers|dirs) are' + @echo ' self-contained, have header guards, and pass kernel-doc.' @echo ' coccicheck - Check with Coccinelle' @echo ' kconfig-sym-check - Check for dangling Kconfig symbol references' @echo ' clang-analyzer - Check with clang static analyzer'
@@ -2251,6 +2273,7 @@ clean: $(clean-dirs) -o -name '*.ll' \ -o -name '*.gcno' \ -o -name '*.long-type-*.txt' \ + -o -name '*.header-check' \ \) -type f -print \ -o -name '.tmp_*' -print \ | xargs rm -rf
scripts/Makefile.build +21 -0
@@ -306,6 +306,27 @@ quiet_cmd_cc_lst_c = MKLST $@ $(obj)/%.lst: $(obj)/%.c FORCE $(call if_changed_dep,cc_lst_c) +# Compile C headers (.h) for header check +# --------------------------------------------------------------------------- + +# Check a header is self-contained, has header guards, and passes kernel-doc. +# +# This is for 'make HEADER_CHECK=(headers|dirs) headercheck'. Not to be confused +# with CONFIG_UAPI_HEADER_TEST. +# +# Please do *not* plug %.header-check into any configurable targets, unless you +# also figure out a way to completely hide the %.header-check artefacts. + +quiet_cmd_header_check = HDRCHK $(patsubst %.header-check,%.h,$@) + cmd_header_check = \ + $(CC) $(c_flags) -fsyntax-only -x c /dev/null -include $< -include $<; \ + PYTHONDONTWRITEBYTECODE=1 $(PYTHON3) $(KERNELDOC) -none $(KDOCFLAGS) \ + $(if $(findstring e, $(KBUILD_EXTRA_WARN))$(CONFIG_WERROR),-Werror) $<; \ + touch $@ + +$(obj)/%.header-check: $(obj)/%.h FORCE + $(call if_changed_dep,header_check) + # Compile Rust sources (.rs) # ---------------------------------------------------------------------------
Signature · 1 line
2.47.3
Hi Jani,
Quoted 14 lines · Jani Nikula
There have been various attempts at adding a header test or check mechanism in the kernel build system. The header check primarily consists of ensuring headers are self-contained, have include guards, and, in some cases, pass kernel-doc.
The main problems have been:
- The dependency tracking creates undesirable artefacts (infamously also known as disgusting turds) in the build directory.
- Gating the feature behind a kconfig option is complicated due to allyesconfig builds. It's possible, but requires a verbose and confusing negative proxy config option.
Quoted 7 lines
- Naming the dependency tracking files with a dot prefix or placing them in a dot prefixed subdirectory in the build directory to hide them has been exceedingly difficult to achieve. (In part due to some Makefiles building files in subdirectory hierarchies.)
- The debate which headers, if any, should really be self-contained is virtually open-ended.
I would expect that headers in include/uapi/*.h should be self-contained, but apparently that's just a pipe dream on my part, or maybe it's just a maintainer option.
With 'make HEADER_CHECK="include/uapi/" headercheck'
I see over 70 errors (mostly typedefs or defined constants, but not only those), such as:
In file included from <command-line>:./../include/uapi/linux/hdlc/ioctl.h:74:21: error: ‘IFNAMSIZ’ undeclared here (not in a function)
74 | char master[IFNAMSIZ]; /* Name of master FRAD device */
| ^~~~~~~~
In file included from <command-line>:
./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef]
29 | #if (__BITS_PER_LONG != 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__)
| ^~~~~~~~~~~~~~~
./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’
34 | __kernel_ulong_t __sec;
| ^~~~~~~~~~~~~~~~In file included from ./../include/uapi/linux/papr_pdsm.h:14,
from <command-line>:../include/linux/ndctl.h:19:33: error: ‘PAGE_SIZE’ undeclared here (not in a function)
19 | ND_MIN_NAMESPACE_SIZE = PAGE_SIZE,
| ^~~~~~~~~
In file included from <command-line>:./../include/uapi/linux/patchkey.h:15:2: error: #error "patchkey.h included directly"
15 | #error "patchkey.h included directly"
| ^~~~~
In file included from <command-line>:
include/uapi/xen/gntdev.h:159:25: error: unknown type name ‘grant_ref_t’
159 | grant_ref_t ref;
| ^~~~~~~~~~~
include/uapi/xen/gntdev.h:161:25: error: unknown type name ‘domid_t’
161 | domid_t domid;
| ^~~~~~~I like it; I think it's useful.
Quoted 34 lines
Approach the problem from a slightly different angle. Add a top level "headercheck" static analysis target which you have to explicitly run, and where you have to explicitly state which headers to check:
$ make HEADER_CHECK="(headers|dirs)" headercheckFor example:
$ make HEADER_CHECK="include/drm drivers/gpu/drm/drm_draw_internal.h" headercheckYou can specify multiple
path/to/header.hor path/to/dir, relative to $(srctree), where path/to/dir is recursively scanned for any .h files.Add a generic %.header-check rule in
scripts/Makefile.buildto support this. Note that this rule should *not* be used in, say, driver or subsystem makefile targets, precisely because of the problems listed above.While this still generates the .header-check dependency tracking artefacts in the build directory, you will only get them as a result of manually running 'make headercheck', not as part of the regular build.
This approach also allows anyone to check any headers in the tree, with no makefile modification or kconfig changes required. What you do with the results is up to you, and the header check facility does not enforce anything like it would as part of the regular build.
Cc Linus Torvalds <[email protected]>Cc Nathan Chancellor <[email protected]>Cc Nicolas Schier <[email protected]>Cc Jason Gunthorpe <[email protected]>Cc Masahiro Yamada <[email protected]>Signed-off-by Jani Nikula <[email protected]>
Quoted 5 lines
--- .gitignore | 1 + Makefile | 25 ++++++++++++++++++++++++- scripts/Makefile.build | 21 +++++++++++++++++++++ 3 files changed, 46 insertions(+), 1 deletion(-)
thanks.
Signature · 1 line
~Randy
Nathan Chancellor
maintainerQuoted 50 lines · Randy Dunlap
Quoted 14 lines · Jani Nikula
There have been various attempts at adding a header test or check mechanism in the kernel build system. The header check primarily consists of ensuring headers are self-contained, have include guards, and, in some cases, pass kernel-doc.
The main problems have been:
- The dependency tracking creates undesirable artefacts (infamously also known as disgusting turds) in the build directory.
- Gating the feature behind a kconfig option is complicated due to allyesconfig builds. It's possible, but requires a verbose and confusing negative proxy config option.
Quoted 7 lines
- Naming the dependency tracking files with a dot prefix or placing them in a dot prefixed subdirectory in the build directory to hide them has been exceedingly difficult to achieve. (In part due to some Makefiles building files in subdirectory hierarchies.)
- The debate which headers, if any, should really be self-contained is virtually open-ended.
I would expect that headers in
include/uapi/*.hshould be self-contained, but apparently that's just a pipe dream on my part, or maybe it's just a maintainer option.With 'make HEADER_CHECK="include/uapi/" headercheck'
I see over 70 errors (mostly typedefs or defined constants, but not only those), such as:In file included from <command-line>:
./../include/uapi/linux/hdlc/ioctl.h:74:21: error: ‘IFNAMSIZ’ undeclared here (not in a function)
74 | char master[IFNAMSIZ]; /* Name of master FRAD device */| ^~~~~~~~ In file included from <command-line>: ./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef] 29 | #if (__BITS_PER_LONG != 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__) | ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~In file included from
./../include/uapi/linux/papr_pdsm.h:14,
from <command-line>:../include/linux/ndctl.h:19:33: error: ‘PAGE_SIZE’ undeclared here (not in a function)
19 |ND_MIN_NAMESPACE_SIZE=PAGE_SIZE,
| ^~~~~~~~~
In file included from <command-line>:./../include/uapi/linux/patchkey.h:15:2: error: #error "patchkey.hincluded directly"
15 | #error "patchkey.hincluded directly"
Quoted 8 lines
| ^~~~~
In file included from <command-line>:
include/uapi/xen/gntdev.h:159:25: error: unknown type name ‘grant_ref_t’
159 | grant_ref_t ref;
| ^~~~~~~~~~~
include/uapi/xen/gntdev.h:161:25: error: unknown type name ‘domid_t’
161 | domid_t domid;
| ^~~~~~~include/uapi already has its own header checking infrastructure under CONFIG_UAPI_HEADER_TEST and usr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbid HEADER_CHECK from including 'include/uapi' and refer people to use CONFIG_UAPI_HEADER_TEST instead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.
That said, I think the overall idea seems fine and relatively clean, at least from my Kbuild perspective, as it is completely opt in, so the previous objection to CONFIG_HEADER_CHECK_DISABLE and widely exposing this to builds is pretty much moot.
Signature · 2 lines
Cheers, Nathan
Quoted 60 lines · Nathan Chancellor
Quoted 50 lines · Randy Dunlap
Quoted 14 lines · Jani Nikula
There have been various attempts at adding a header test or check mechanism in the kernel build system. The header check primarily consists of ensuring headers are self-contained, have include guards, and, in some cases, pass kernel-doc.
The main problems have been:
- The dependency tracking creates undesirable artefacts (infamously also known as disgusting turds) in the build directory.
- Gating the feature behind a kconfig option is complicated due to allyesconfig builds. It's possible, but requires a verbose and confusing negative proxy config option.
Quoted 7 lines
- Naming the dependency tracking files with a dot prefix or placing them in a dot prefixed subdirectory in the build directory to hide them has been exceedingly difficult to achieve. (In part due to some Makefiles building files in subdirectory hierarchies.)
- The debate which headers, if any, should really be self-contained is virtually open-ended.
I would expect that headers in
include/uapi/*.hshould be self-contained, but apparently that's just a pipe dream on my part, or maybe it's just a maintainer option.With 'make HEADER_CHECK="include/uapi/" headercheck'
I see over 70 errors (mostly typedefs or defined constants, but not only those), such as:In file included from <command-line>:
./../include/uapi/linux/hdlc/ioctl.h:74:21: error: ‘IFNAMSIZ’ undeclared here (not in a function)
74 | char master[IFNAMSIZ]; /* Name of master FRAD device */| ^~~~~~~~ In file included from <command-line>: ./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef] 29 | #if (__BITS_PER_LONG != 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__) | ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~In file included from
./../include/uapi/linux/papr_pdsm.h:14,
from <command-line>:../include/linux/ndctl.h:19:33: error: ‘PAGE_SIZE’ undeclared here (not in a function)
19 |ND_MIN_NAMESPACE_SIZE=PAGE_SIZE,
| ^~~~~~~~~
In file included from <command-line>:./../include/uapi/linux/patchkey.h:15:2: error: #error "patchkey.hincluded directly"
15 | #error "patchkey.hincluded directly"Quoted 8 lines
| ^~~~~ In file included from <command-line>: include/uapi/xen/gntdev.h:159:25: error: unknown type name ‘grant_ref_t’ 159 | grant_ref_t ref; | ^~~~~~~~~~~ include/uapi/xen/gntdev.h:161:25: error: unknown type name ‘domid_t’ 161 | domid_t domid; | ^~~~~~~
Quoted 9 lines
include/uapialready has its own header checking infrastructure underCONFIG_UAPI_HEADER_TESTandusr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbidHEADER_CHECKfrom including 'include/uapi' and refer people to useCONFIG_UAPI_HEADER_TESTinstead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.
Thanks for the reminder. I'll limit any checking I do to non-uapi files.
Quoted 5 lines
That said, I think the overall idea seems fine and relatively clean, at least from my
Kbuildperspective, as it is completely opt in, so the previous objection toCONFIG_HEADER_CHECK_DISABLEand widely exposing this to builds is pretty much moot.
Signature · 1 line
~Randy
Quoted 60 lines · Nathan Chancellor
Quoted 50 lines · Randy Dunlap
Quoted 14 lines · Jani Nikula
There have been various attempts at adding a header test or check mechanism in the kernel build system. The header check primarily consists of ensuring headers are self-contained, have include guards, and, in some cases, pass kernel-doc.
The main problems have been:
- The dependency tracking creates undesirable artefacts (infamously also known as disgusting turds) in the build directory.
- Gating the feature behind a kconfig option is complicated due to allyesconfig builds. It's possible, but requires a verbose and confusing negative proxy config option.
Quoted 7 lines
- Naming the dependency tracking files with a dot prefix or placing them in a dot prefixed subdirectory in the build directory to hide them has been exceedingly difficult to achieve. (In part due to some Makefiles building files in subdirectory hierarchies.)
- The debate which headers, if any, should really be self-contained is virtually open-ended.
I would expect that headers in
include/uapi/*.hshould be self-contained, but apparently that's just a pipe dream on my part, or maybe it's just a maintainer option.With 'make HEADER_CHECK="include/uapi/" headercheck'
I see over 70 errors (mostly typedefs or defined constants, but not only those), such as:In file included from <command-line>:
./../include/uapi/linux/hdlc/ioctl.h:74:21: error: ‘IFNAMSIZ’ undeclared here (not in a function)
74 | char master[IFNAMSIZ]; /* Name of master FRAD device */| ^~~~~~~~ In file included from <command-line>: ./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef] 29 | #if (__BITS_PER_LONG != 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__) | ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~In file included from
./../include/uapi/linux/papr_pdsm.h:14,
from <command-line>:../include/linux/ndctl.h:19:33: error: ‘PAGE_SIZE’ undeclared here (not in a function)
19 |ND_MIN_NAMESPACE_SIZE=PAGE_SIZE,
| ^~~~~~~~~
In file included from <command-line>:./../include/uapi/linux/patchkey.h:15:2: error: #error "patchkey.hincluded directly"
15 | #error "patchkey.hincluded directly"Quoted 8 lines
| ^~~~~ In file included from <command-line>: include/uapi/xen/gntdev.h:159:25: error: unknown type name ‘grant_ref_t’ 159 | grant_ref_t ref; | ^~~~~~~~~~~ include/uapi/xen/gntdev.h:161:25: error: unknown type name ‘domid_t’ 161 | domid_t domid; | ^~~~~~~
Quoted 8 lines
include/uapialready has its own header checking infrastructure underCONFIG_UAPI_HEADER_TESTandusr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbidHEADER_CHECKfrom including 'include/uapi' and refer people to useCONFIG_UAPI_HEADER_TESTinstead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.
Something like this on top would fail if there are any include/uapi headers in there:
Makefile +1 -0
@@ -1554,6 +1554,7 @@ header-check-targets := $(patsubst %.h,%.header-check,$(sort $(header-check-file headercheck: $(if $(header-check-targets),,$(error $@ found no headers in HEADER_CHECK="$(HEADER_CHECK)")) + $(if $(filter include/uapi/%,$(header-check-targets)),$(error $@ found include/uapi headers in HEADER_CHECK="$(HEADER_CHECK)")) $(Q)$(MAKE) $(header-check-targets) else headercheck:
I'll wait a bit for more feedback before sending a v2.
Quoted 4 lines
That said, I think the overall idea seems fine and relatively clean, at least from my
Kbuildperspective, as it is completely opt in, so the previous objection toCONFIG_HEADER_CHECK_DISABLEand widely exposing this to builds is pretty much moot.
Thanks Nathan and Randy, this feels encouraging. :)
BR,
Jani.
Signature · 1 line
Jani Nikula, Intel
Quoted 60 lines · Nathan Chancellor
Quoted 50 lines · Randy Dunlap
Quoted 14 lines · Jani Nikula
There have been various attempts at adding a header test or check mechanism in the kernel build system. The header check primarily consists of ensuring headers are self-contained, have include guards, and, in some cases, pass kernel-doc.
The main problems have been:
- The dependency tracking creates undesirable artefacts (infamously also known as disgusting turds) in the build directory.
- Gating the feature behind a kconfig option is complicated due to allyesconfig builds. It's possible, but requires a verbose and confusing negative proxy config option.
Quoted 7 lines
- Naming the dependency tracking files with a dot prefix or placing them in a dot prefixed subdirectory in the build directory to hide them has been exceedingly difficult to achieve. (In part due to some Makefiles building files in subdirectory hierarchies.)
- The debate which headers, if any, should really be self-contained is virtually open-ended.
I would expect that headers in
include/uapi/*.hshould be self-contained, but apparently that's just a pipe dream on my part, or maybe it's just a maintainer option.With 'make HEADER_CHECK="include/uapi/" headercheck'
I see over 70 errors (mostly typedefs or defined constants, but not only those), such as:In file included from <command-line>:
./../include/uapi/linux/hdlc/ioctl.h:74:21: error: ‘IFNAMSIZ’ undeclared here (not in a function)
74 | char master[IFNAMSIZ]; /* Name of master FRAD device */| ^~~~~~~~ In file included from <command-line>: ./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef] 29 | #if (__BITS_PER_LONG != 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__) | ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~In file included from
./../include/uapi/linux/papr_pdsm.h:14,
from <command-line>:../include/linux/ndctl.h:19:33: error: ‘PAGE_SIZE’ undeclared here (not in a function)
19 |ND_MIN_NAMESPACE_SIZE=PAGE_SIZE,
| ^~~~~~~~~
In file included from <command-line>:./../include/uapi/linux/patchkey.h:15:2: error: #error "patchkey.hincluded directly"
15 | #error "patchkey.hincluded directly"Quoted 8 lines
| ^~~~~ In file included from <command-line>: include/uapi/xen/gntdev.h:159:25: error: unknown type name ‘grant_ref_t’ 159 | grant_ref_t ref; | ^~~~~~~~~~~ include/uapi/xen/gntdev.h:161:25: error: unknown type name ‘domid_t’ 161 | domid_t domid; | ^~~~~~~
Quoted 8 lines
include/uapialready has its own header checking infrastructure underCONFIG_UAPI_HEADER_TESTandusr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbidHEADER_CHECKfrom including 'include/uapi' and refer people to useCONFIG_UAPI_HEADER_TESTinstead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.
Testing the UAPI headers here too would still be valuable. CONFIG_UAPI_HEADER_TEST tests the headers from the perspective of userspace after they have undergone processing. The kernel build might see the same headers quite differently.
For instance the example from above:
./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef]
29 | #if (__BITS_PER_LONG != 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__)
| ^~~~~~~~~~~~~~~
./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’
34 | __kernel_ulong_t __sec;
| ^~~~~~~~~~~~~~~~These are legitimate issues, the inclusion of the necessary header is gated behind #ifndef __KERNEL__ although it should not be.
(...)
Thomas
Quoted 11 lines · Thomas Weißschuh
For instance the example from above:
./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef]
29 | #if (__BITS_PER_LONG!= 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__)| ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~These are legitimate issues, the inclusion of the necessary header is gated behind #ifndef
__KERNEL__although it should not be.
Yeah, if the header is included by a kernel file then this makefile test should evaluate it under those conditions, and it should still be self-contained within the kernel environment.. Wasn't that one of the problems with earlier draft that it did not figure out the right compilation options for the context the header is included in?
FWIW I think this is a reasonable compromise too
Jason
Nathan Chancellor
maintainerQuoted 19 lines · Thomas Weißschuh
Quoted 8 lines · Nathan Chancellor
include/uapialready has its own header checking infrastructure underCONFIG_UAPI_HEADER_TESTandusr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbidHEADER_CHECKfrom including 'include/uapi' and refer people to useCONFIG_UAPI_HEADER_TESTinstead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.Testing the UAPI headers here too would still be valuable.
CONFIG_UAPI_HEADER_TESTtests the headers from the perspective of userspace after they have undergone processing. The kernel build might see the same headers quite differently.For instance the example from above:
./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef]
29 | #if (__BITS_PER_LONG!= 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__)
Quoted 7 lines
| ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~These are legitimate issues, the inclusion of the necessary header is gated behind #ifndef
__KERNEL__although it should not be.
Hmmm, fair enough. I guess we can leave this functionality in place for UAPI headers then but I just worry about people doing what Randy did and reporting these issues when they are already known and handled in some other way. Maybe the UAPI header test could grow an "internal kernel usage" check similar to the other checks that it has at some point then we could forbid HEADER_CHECK=include/uapi but it is fine for now.
Signature · 2 lines
Cheers, Nathan
Quoted 28 lines · Nathan Chancellor
Quoted 19 lines · Thomas Weißschuh
Quoted 8 lines · Nathan Chancellor
include/uapialready has its own header checking infrastructure underCONFIG_UAPI_HEADER_TESTandusr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbidHEADER_CHECKfrom including 'include/uapi' and refer people to useCONFIG_UAPI_HEADER_TESTinstead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.Testing the UAPI headers here too would still be valuable.
CONFIG_UAPI_HEADER_TESTtests the headers from the perspective of userspace after they have undergone processing. The kernel build might see the same headers quite differently.For instance the example from above:
./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef]
29 | #if (__BITS_PER_LONG!= 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__)Quoted 7 lines
| ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~These are legitimate issues, the inclusion of the necessary header is gated behind #ifndef
__KERNEL__although it should not be.
Quoted 6 lines
Hmmm, fair enough. I guess we can leave this functionality in place for UAPI headers then but I just worry about people doing what Randy did and reporting these issues when they are already known and handled in some other way. Maybe the UAPI header test could grow an "internal kernel usage" check similar to the other checks that it has at some point then we could forbid HEADER_CHECK=include/uapi but it is fine for now.
Do you think the patch at hand is good to go, then? Anything further you need from me?
BR,
Jani.
Signature · 1 line
Jani Nikula, Intel
Nathan Chancellor
maintainerQuoted 36 lines · Jani Nikula
Quoted 28 lines · Nathan Chancellor
Quoted 19 lines · Thomas Weißschuh
Quoted 8 lines · Nathan Chancellor
include/uapialready has its own header checking infrastructure underCONFIG_UAPI_HEADER_TESTandusr/include/Makefile, which avoids this with a no-header-test list that includes many of the files listed in these messages. To be honest, we should probably forbidHEADER_CHECKfrom including 'include/uapi' and refer people to useCONFIG_UAPI_HEADER_TESTinstead, as there are other differences like being built under a different C standard or C++ and such that the existing infrastructure handles.Testing the UAPI headers here too would still be valuable.
CONFIG_UAPI_HEADER_TESTtests the headers from the perspective of userspace after they have undergone processing. The kernel build might see the same headers quite differently.For instance the example from above:
./../include/uapi/linux/input.h:29:6: warning: ‘__BITS_PER_LONG’ is not defined, evaluates to ‘0’ [-Wundef]
29 | #if (__BITS_PER_LONG!= 32 || !defined(__USE_TIME_BITS64)) && !defined(__KERNEL__)Quoted 7 lines
| ^~~~~~~~~~~~~~~ ./../include/uapi/linux/input.h:34:9: error: unknown type name ‘__kernel_ulong_t’ 34 | __kernel_ulong_t __sec; | ^~~~~~~~~~~~~~~~These are legitimate issues, the inclusion of the necessary header is gated behind #ifndef
__KERNEL__although it should not be.Quoted 6 lines
Hmmm, fair enough. I guess we can leave this functionality in place for UAPI headers then but I just worry about people doing what Randy did and reporting these issues when they are already known and handled in some other way. Maybe the UAPI header test could grow an "internal kernel usage" check similar to the other checks that it has at some point then we could forbid HEADER_CHECK=include/uapi but it is fine for now.
Quoted 2 lines
Do you think the patch at hand is good to go, then? Anything further you need from me?
I don't have any other comments or objections. I will give a few days for other folks to comment on it before I apply it to the kbuild tree.
Signature · 2 lines
Cheers, Nathan
Nathan Chancellor
maintainerQuoted 1 line · Jani Nikula
kbuild: add header check facility as a manually run static analyzer
Applied to
https://git.kernel.org/pub/scm/linux/kernel/git/kbuild/linux.git kbuild-next-unstable
Thanks!
[1/1] kbuild: add header check facility as a manually run static analyzer
https://git.kernel.org/kbuild/c/26d665bef44fa
Please look out for regression or issue reports or other follow up comments, as they may result in the patch/series getting dropped or reverted. Patches applied to an "unstable" branch are accepted pending wider testing in -next and any post-commit review; they will generally be moved to the main branch in a week if no issues are found.
Best regards,
Signature · 2 lines
Cheers, Nathan