Unreal, you mean someone broke it? On that particular build, if I take out the obj-m reference it seems to work, but I was back here because now my driver itself has link failures. The driver has about 12 libraries amounting to 100's modules and 10's if not 100's of thousands of lines of code. Some are specific to driver support (that build in the start of this thread), others are just built for several environments, from Windows Drivers, Linux Drivers, BIOS, UEFI, etc.. >From encryption, compression, complex data handling, etc.. The driver requires all that, no way I could just try to include the source somehow. I'll look at the patch and if I want to move to this broken V5 version or just stick with 4.x forever (I already tried moving a block driver over to blk-mq, one I'm trying to build now to test, that was a surprise too, broken because old method taken out, not sure if I should still queue requests to my worker thread or now that is done for me?). Thanks for feedback. On Mon, Feb 24, 2020 at 6:57 PM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote: > > On Tue, Feb 25, 2020 at 7:02 AM David F. <df7729@xxxxxxxxx> wrote: > > > > Hi, > > > > I have a private custom module I build via various scripts since I > > setup since I knew I'd have no clue how to do it in the future. It's > > been working great for years, but now it's broken with the move from > > 4.19.xx to 5.4.22 so the future is here and don't remember much about > > how I set it all up. But it does look to me that the failure has to > > do with a support library I have it build. Can the kbuild experts > > tell me what changed with V5 from V4.19 that breaks building my custom > > module. > > > > Thanks!! > > > > Here's the information (I'm writing this in Windows so my dir output > > will be from windows): > > > > The input/source files: > > > > 01/25/2016 06:52 PM 1,501 linuxdriverc.c > > 01/24/2016 06:59 PM 284 linuxdrivercpp.cpp > > 01/22/2016 07:20 PM 1,124 linuxdriverdumpbuf.c > > 09/22/2019 12:17 AM 4,985 linuxdriverfile.c > > 01/22/2016 07:20 PM 2,142 linuxdriverprintk.c > > > > When the 4.x version (this was 4.19.97) finishes the dir contains: > > > > 01/22/2020 12:43 AM 338 .junk.ko.cmd > > 01/22/2020 12:43 AM 29,743 .junk.mod.o.cmd > > 01/22/2020 12:43 AM 220 .junk.o.cmd > > 01/22/2020 12:43 AM 581 .lib.a.cmd > > 01/22/2020 12:43 AM 29,795 .linuxdriverc.o.cmd > > 01/22/2020 12:43 AM 30,248 .linuxdriverdumpbuf.o.cmd > > 01/22/2020 12:43 AM 36,098 .linuxdriverfile.o.cmd > > 01/22/2020 12:43 AM 30,235 .linuxdriverprintk.o.cmd > > 01/22/2020 12:43 AM 1,908 junk.ko > > 01/22/2020 12:43 AM 650 junk.mod.c > > 01/22/2020 12:43 AM 1,792 junk.mod.o > > 01/22/2020 12:43 AM 260 junk.o > > 01/22/2020 12:43 AM 1,084 lib.a > > 01/22/2020 12:43 AM 29 linuxdriverc.c > > 01/22/2020 12:43 AM 1,040 linuxdriverc.o > > 01/22/2020 12:43 AM 33 linuxdrivercpp.cpp > > 01/22/2020 12:43 AM 976 linuxdrivercpp.o > > 01/22/2020 12:43 AM 35 linuxdriverdumpbuf.c > > 01/22/2020 12:43 AM 1,320 linuxdriverdumpbuf.o > > 01/22/2020 12:43 AM 32 linuxdriverfile.c > > 01/22/2020 12:43 AM 1,448 linuxdriverfile.o > > 01/22/2020 12:43 AM 34 linuxdriverprintk.c > > 01/22/2020 12:43 AM 1,564 linuxdriverprintk.o > > 01/22/2020 12:43 AM 1,084 linuxdriversup.a > > 01/26/2016 05:16 PM 2,188 Makefile > > 01/22/2020 12:43 AM 0 Module.symvers > > 01/22/2020 12:43 AM 71 modules.order > > > > For the 5.4.22 the files are: > > > > 02/24/2020 12:04 PM 573 .lib.a.cmd > > 02/24/2020 12:04 PM 30,740 .linuxdriverc.o.cmd > > 02/24/2020 12:04 PM 31,278 .linuxdriverdumpbuf.o.cmd > > 02/24/2020 12:04 PM 37,708 .linuxdriverfile.o.cmd > > 02/24/2020 12:04 PM 31,265 .linuxdriverprintk.o.cmd > > 02/24/2020 12:04 PM 1,080 lib.a > > 01/25/2016 06:52 PM 29 linuxdriverc.c > > 02/24/2020 12:04 PM 1,040 linuxdriverc.o > > 01/24/2016 06:59 PM 33 linuxdrivercpp.cpp > > 02/24/2020 12:04 PM 976 linuxdrivercpp.o > > 01/22/2016 07:20 PM 35 linuxdriverdumpbuf.c > > 02/24/2020 12:04 PM 1,320 linuxdriverdumpbuf.o > > 09/22/2019 12:17 AM 32 linuxdriverfile.c > > 02/24/2020 12:04 PM 1,448 linuxdriverfile.o > > 01/22/2016 07:20 PM 34 linuxdriverprintk.c > > 02/24/2020 12:04 PM 1,564 linuxdriverprintk.o > > 01/26/2016 05:16 PM 2,188 Makefile > > > > There are no error messages other than it can't find junk.o for LD: > > > > Starting KBuild make... > > make[3]: Entering directory '/usr/src/linux-headers-5.4.22-686-iflnet' > > LD [M] /media/sf_SRC_DRIVE/libs/linux_drivers/5.4.22-686-iflnet/junk.o > > ld: no input files > > scripts/Makefile.build:446: recipe for target > > '/media/sf_SRC_DRIVE/libs/linux_drivers/5.4.22-686-iflnet/junk.o' > > failed > > make[4]: *** [/media/sf_SRC_DRIVE/libs/linux_drivers/5.4.22-686-iflnet/junk.o] > > Error 1 > > Makefile:1652: recipe for target > > '/media/sf_SRC_DRIVE/libs/linux_drivers/5.4.22-686-iflnet' failed > > make[3]: *** [/media/sf_SRC_DRIVE/libs/linux_drivers/5.4.22-686-iflnet] Error 2 > > make[3]: Leaving directory '/usr/src/linux-headers-5.4.22-686-iflnet' > > > > > > The makefile used is: > > > > # You should only need to change the items marked ** fill in ** > > # everything else should be automatic. > > > > # the version of the kernel the module is being compiled for > > KVER?=$(shell uname -r) > > # the directory of the kernel build system > > KDIR:=/lib/modules/$(KVER)/build > > # fill in the verbose level you want > > V:=0 > > > > ################################# > > # fill in the name of the module ** fill in ** > > ################################# > > mymodlib_name:=linuxdriversup > > > > ################################# > > # unique cpp build flags ** fill in ** > > ################################# > > # CC_FLAGS+=-I. > > > > # flags file for building CPP versions > > FLAGS:=/media/sf_SRC_DRIVE/libs/linux_drivers/buildopts/$(KVER)/buildopts.cfg > > > > ################################# > > # extra C flags to use ** fill in ** > > ################################# > > EXTRA_CFLAGS+=-I. > > #EXTRA_CFLAGS+=-Werror > > #EXTRA_CFLAGS += -std=gnu99 > > #EXTRA_CFLAGS += -Wno-declaration-after-statement > > > > ############## > > # code start # > > ############## > > ifeq ($(V),1) > > Q= > > else > > Q=@ > > endif > > > > # this was the only way I figured out how to get the > > # kbuild envrionment to build the library for me. the > > # cpp item doesn't really need to be built in this > > # envrionment but we do it since it works. > > > > # Setup module name for kbuild > > obj-m:=junk.o > > junk-objs:=lib.a > > > Perhaps, this line is the cause of the failure. > > Kbuild does not support linking *.a to a module. > > There exists a workaround patch, but it was rejected. > https://patchwork.kernel.org/patch/11318691/#23088533 > > > > Masahiro Yamada > > > > > > # build sources in the target directory > > SOURCES_ALL:=$(filter-out %.mod.c,$(shell find $(M) -maxdepth 1 -name > > "*.cpp" -or -name "*.c")) > > > > # add the sources and to the kbuild > > lib-y+=$(addsuffix .o,$(notdir $(basename $(SOURCES_ALL)))) > > > > > > # pattern for compiling the c++ parts > > %.o: %.cpp $(FLAGS) > > $(info doing [$@]) > > $(Q)g++ `cat $(FLAGS)` $(CC_FLAGS) -fno-exceptions -fno-rtti -Wall > > -c -o $@ $< > > > > .PHONY: all > > all: > > $(Q)echo Run make from main directory > > > > .PHONY: renamelib > > renamelib: > > # the -p is not working to preserve timestamps but touch works > > @cp -p lib.a $(mymodlib_name).a ; \ > > touch -r lib.a $(mymodlib_name).a; \ > > rm -fr .tmp_versions > > # rm -f junk.* .*.cmd ; removing either of these trigger full build > > > > > > .PHONY: clean > > clean: > > # note this deletes .a .o etc files in subdirs too! > > $(Q)$(MAKE) -C $(KDIR) M=$(CURDIR) V=$(V) clean > > > > -- > Best Regards > Masahiro Yamada
