| Message ID | 1443746068-33050-1-git-send-email-drinkcat@chromium.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Nicolas Boichat <drinkcat@chromium.org> writes: > The section mismatch warning can be easy to miss during the kernel build > process. Allow it to be marked as fatal to be easily caught and prevent > bugs from slipping in. > > Setting CONFIG_SECTION_MISMATCH_WARNING=y causes these warnings to be > non-fatal, since there are a number of section mismatches when using > allmodconfig on some architectures, and we do not want to break these > builds by default. > > Signed-off-by: Nicolas Boichat <drinkcat@chromium.org> > > Change-Id: Ic346706e3297c9f0d790e3552aa94e5cff9897a6 > --- > > I'm trying to revive this old patch. When it was first submitted [1], > Jonathan got the following feedback: > - The logic of the option should be inverted (i.e. SECTION_MISMATCH_WARNING), > so that is not not enabled in allmodconfig for some architectures that do > have section mismatches. I've seen some failures (namely, on arm64), so I > did that. > - CONFIG_DEBUG_SECTION_MISMATCH should be removed and warnings should always > be shown verbosely. This option does 3 things: > 1. Enable -fno-inline-functions-called-onc > 2. Run the section mismatch analysis for each module/built-in.o > 3. Enable verbose reporting from modpost > We definitely do not want 1 by default, so I think we should keep the option. > If we enable 2 & 3 by default, which I think would be reasonable, then the > option name does not make much sense anymore, and I'm not sure what to do > with the documentation that is currently provided in the Kconfig description. > > Tested on x86-64 allmodconfig, setting the option to =n, and creating a > section mismatch by running: > sed -i -e 's/\(ssize_t soc_codec_reg_show\)/__init \1/' sound/soc/soc-core.c Minor feedback: 1) Please rename to CONFIG_SECTION_MISMATCH_WARN_ONLY. 2) -f means force, perhaps -E for error? Otherwise, quite nice. Thanks, Rusty. -- To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index d190d44..74314bb 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -311,6 +311,15 @@ config DEBUG_SECTION_MISMATCH - Enable verbose reporting from modpost in order to help resolve the section mismatches that are reported. +config SECTION_MISMATCH_WARNING + bool "Make section mismatch errors non-fatal" + default y + help + If you say N here, the build process will fail if there are any + section mismatch, instead of just throwing warnings. + + If unsure, say Y. + # # Select this config option from the architecture Kconfig, if it # is preferred to always offer frame pointers as a config diff --git a/scripts/Makefile.modpost b/scripts/Makefile.modpost index 9c40dae..d2cbd83 100644 --- a/scripts/Makefile.modpost +++ b/scripts/Makefile.modpost @@ -77,6 +77,7 @@ modpost = scripts/mod/modpost \ $(if $(KBUILD_EXTRA_SYMBOLS), $(patsubst %, -e %,$(KBUILD_EXTRA_SYMBOLS))) \ $(if $(KBUILD_EXTMOD),-o $(modulesymfile)) \ $(if $(CONFIG_DEBUG_SECTION_MISMATCH),,-S) \ + $(if $(CONFIG_SECTION_MISMATCH_WARNING),,-f) \ $(if $(KBUILD_EXTMOD)$(KBUILD_MODPOST_WARN),-w) \ $(if $(CONFIG_LTO),-w) diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c index d583c98..bc729b8 100644 --- a/scripts/mod/modpost.c +++ b/scripts/mod/modpost.c @@ -38,6 +38,7 @@ static int warn_unresolved = 0; /* How a symbol is exported */ static int sec_mismatch_count = 0; static int sec_mismatch_verbose = 1; +static int sec_mismatch_fatal = 0; /* ignore missing files */ static int ignore_missing_files; @@ -2385,7 +2386,7 @@ int main(int argc, char **argv) struct ext_sym_list *extsym_iter; struct ext_sym_list *extsym_start = NULL; - while ((opt = getopt(argc, argv, "i:I:e:mnsST:o:awM:K:")) != -1) { + while ((opt = getopt(argc, argv, "i:I:e:mnsST:o:awM:K:f")) != -1) { switch (opt) { case 'i': kernel_read = optarg; @@ -2426,6 +2427,9 @@ int main(int argc, char **argv) case 'w': warn_unresolved = 1; break; + case 'f': + sec_mismatch_fatal = 1; + break; default: exit(1); } @@ -2475,14 +2479,20 @@ int main(int argc, char **argv) sprintf(fname, "%s.mod.c", mod->name); write_if_changed(&buf, fname); } - if (dump_write) write_dump(dump_write); - if (sec_mismatch_count && !sec_mismatch_verbose) - warn("modpost: Found %d section mismatch(es).\n" - "To see full details build your kernel with:\n" - "'make CONFIG_DEBUG_SECTION_MISMATCH=y'\n", - sec_mismatch_count); + if (sec_mismatch_count) { + if (!sec_mismatch_verbose) { + warn("modpost: Found %d section mismatch(es).\n" + "To see full details build your kernel with:\n" + "'make CONFIG_DEBUG_SECTION_MISMATCH=y'\n", + sec_mismatch_count); + } + if (sec_mismatch_fatal) { + fatal("modpost: Section mismatches detected.\n" + "Set CONFIG_SECTION_MISMATCH_WARNING=y to allow them.\n"); + } + } return err; }
The section mismatch warning can be easy to miss during the kernel build process. Allow it to be marked as fatal to be easily caught and prevent bugs from slipping in. Setting CONFIG_SECTION_MISMATCH_WARNING=y causes these warnings to be non-fatal, since there are a number of section mismatches when using allmodconfig on some architectures, and we do not want to break these builds by default. Signed-off-by: Nicolas Boichat <drinkcat@chromium.org> Change-Id: Ic346706e3297c9f0d790e3552aa94e5cff9897a6 --- I'm trying to revive this old patch. When it was first submitted [1], Jonathan got the following feedback: - The logic of the option should be inverted (i.e. SECTION_MISMATCH_WARNING), so that is not not enabled in allmodconfig for some architectures that do have section mismatches. I've seen some failures (namely, on arm64), so I did that. - CONFIG_DEBUG_SECTION_MISMATCH should be removed and warnings should always be shown verbosely. This option does 3 things: 1. Enable -fno-inline-functions-called-onc 2. Run the section mismatch analysis for each module/built-in.o 3. Enable verbose reporting from modpost We definitely do not want 1 by default, so I think we should keep the option. If we enable 2 & 3 by default, which I think would be reasonable, then the option name does not make much sense anymore, and I'm not sure what to do with the documentation that is currently provided in the Kconfig description. Tested on x86-64 allmodconfig, setting the option to =n, and creating a section mismatch by running: sed -i -e 's/\(ssize_t soc_codec_reg_show\)/__init \1/' sound/soc/soc-core.c Applies on linux-next/20151001 [1] http://thread.gmane.org/gmane.linux.kbuild.devel/9044 lib/Kconfig.debug | 9 +++++++++ scripts/Makefile.modpost | 1 + scripts/mod/modpost.c | 24 +++++++++++++++++------- 3 files changed, 27 insertions(+), 7 deletions(-)