Bug 197220

This problem also appears on x86_64 machine running the 2.6.17-1.2356.fc6. Can
demonstrate the problem with installing the appropriate kernel-debuginfo and
systemtap rpms, then run a very simple systemtap probe from the commandline:

# /usr/bin/stap -vv -e 'probe begin { log ("hello world") }'

End up getting the following when trying to compile the resulting C code of the
systemtap module:


Running make -C "/lib/modules/2.6.17-1.2356.fc6/build" M="/tmp/stapfSW0Zy"
modules V=1
make: Entering directory `/usr/src/kernels/2.6.17-1.2356.fc6-x86_64'
Makefile:452: *** kernel configuration not valid - run 'make prepare' in
/usr/src/kernels/2.6.17-1.2356.fc6-x86_64 to update it.  Stop.
make: Leaving directory `/usr/src/kernels/2.6.17-1.2356.fc6-x86_64'
Pass 4: compiled C into "stap_20491.ko" in 190usr/230sys/412real ms.
Pass 4: compilation failed.  Try again with more '-v' (verbose) options.
Running rm -rf /tmp/stapfSW0Zy

Did some more searching to figure out what is going on. Newer kernel have the
following rule in the Makefile:

# If .config is newer than include/config/auto.conf, someone tinkered
# with it and forgot to run make oldconfig.
# if auto.conf.cmd is missing then we are probably in a cleaned tree so
# we execute the config step to be sure to catch updated Kconfig files
include/config/auto.conf: $(KCONFIG_CONFIG) include/config/auto.conf.cmd
ifeq ($(KBUILD_EXTMOD),)
	$(Q)$(MAKE) -f $(srctree)/Makefile silentoldconfig
else
	$(error kernel configuration not valid - run 'make prepare' in $(srctree) to
update it)
endif

The problem is the timestamps of the files for this rule:

$ ls -l include/config/auto.conf .config include/config/auto.conf.cmd
-rw-r--r-- 1 root root 61870 Jul  7 04:18 .config
-rw-r--r-- 1 root root 43721 Jul  7 03:59 include/config/auto.conf
-rw-r--r-- 1 root root  6923 Jul  7 03:59 include/config/auto.conf.cmd

I was able to do the following in the /usr/src/kernel/2.6.17-1.2358.fc6-x86_64
directory and this cleared up the problem:

 touch include/config/auto.conf

touching include/config/auto.conf makes the "make prepare" message go away alright.
However, depending on the kernel module I'm trying to build I'm quite often
seeing the following message:

 /bin/sh: line 1: 12162 Floating point exceptionscripts/basic/fixdep
/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/.i4l_idi.o.d
/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/i4l_idi.o 'gcc
-m32
-Wp,-MD,/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/.i4l_idi.o.d
 -nostdinc -isystem /usr/lib/gcc/i386-redhat-linux/4.1.1/include -D__KERNEL__
-Iinclude  -include include/linux/autoconf.h -Wall -Wundef -Wstrict-prototypes
-Wno-trigraphs -fno-strict-aliasing -fno-common -Wstrict-prototypes -Wundef
-Werror-implicit-function-declaration  -Os -fno-omit-frame-pointer
-fno-optimize-sibling-calls -fasynchronous-unwind-tables -g -pipe -msoft-float
-fno-builtin-sprintf -fno-builtin-log2 -fno-builtin-puts 
-mpreferred-stack-boundary=2  -march=i586 -mtune=generic -mregparm=3
-ffreestanding -Iinclude/asm-i386/mach-default -Wdeclaration-after-statement
-Wno-pointer-sign   -DMODULE -D"KBUILD_STR(s)=\#s"
-D"KBUILD_BASENAME=KBUILD_STR(i4l_idi)"  -D"KBUILD_MODNAME=KBUILD_STR(diva2i4l)"
-c -o
/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/.tmp_i4l_idi.o
/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/i4l_idi.c'
>/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/.i4l_idi.o.tmp
make[1]: ***
[/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_/i4l_idi.o]
Error 136
make: ***
[_module_/home/andreas/fedora/redhat/BUILD/diva2i4l-kmod-1.0/_kmod_build_] Error 2
make: Leaving directory `/usr/src/kernels/2.6.17-1.2405.fc6-i586'
error: Bad exit status from /var/tmp/rpm-tmp.78057 (%build)


I'm wondering if this is related to the changed makefile.

Any news on this bug? Is it fixed in rawhide yet?

You can't just touch include/config/auto.conf because that file isn't a current
copy of the .config - so building modules might well go awry.

I just committed a fix which copies the .config file into
include/config/auto.conf instead - that way they really are both the same.
Hopefully this fixes the issue and we can avoid a call to make prepare in the
spec file (that breaks some of the custom links setup for powerpc arches).

Jon.

I'm going to go ahead and close this BZ now since the test kernel I built
earlier (2.6.17-1.2517) works for me now - if necessary it can be reopened again
later.