I am trying to compile an example of "hello world" Kernel Module,
problems found on ubuntu 11.04, kernel 3.2.6, gcc 4.5.2 and fedora 16, kernel 3.2.7, gcc 4.6.7.
code:
#include <linux/module.h>
#include <linux/init.h>
MODULE_LICENSE("GPL");
static int __init hello_init (void)
{
printk("Hello module init\n");
return 0;
}
static void __exit hello_exit (void)
{
printk("Hello module exit\n");
}
module_init(hello_init);
module_exit(hello_exit);
compiled with:
gcc -D__KERNEL__ -I /usr/src/linux/include/ -DMODULE -Wall -O2 -c hello.c -o hello.o
error:
In file included from /usr/src/linux/include/linux/kernel.h:13:0,
from /usr/src/linux/include/linux/cache.h:4,
from /usr/src/linux/include/linux/time.h:7,
from /usr/src/linux/include/linux/stat.h:60,
from /usr/src/linux/include/linux/module.h:10,
from hello.c:1: /usr/src/linux/include/linux/linkage.h:5:25: fatal error:
asm/linkage.h: file not found
then I found in /usr/src/linux/include/ there is no folder named 'asm' but 'asm-generic';
so I made a soft link 'asm' to 'asm-generic', and compiled agail:
this time the error was:
In file included from /usr/src/linux/include/linux/preempt.h:9:0,
from /usr/src/linux/include/linux/spinlock.h:50,
from /usr/src/linux/include/linux/seqlock.h:29,
from /usr/src/linux/include/linux/time.h:8,
from /usr/src/linux/include/linux/stat.h:60,
from /usr/src/linux/include/linux/module.h:10,
from hello.c:1: /usr/src/linux/include/linux/thread_info.h:53:29: fatal error:
asm/thread_info.h: file not found
So I realized I was wrong, but why ? T_T
obj-m += hello.o
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
is a proper way to build modules see kbuild documentation
And to see difference beetween your compiler invocation you could
cat /lib/modules/$(shell uname -r)/build/Makefile
And analyze an output
obj-m += hello.o
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
Here hello.c is your kernel source file. just use make to build your hello.ko module.
asm should be a link to the actual architecture you're compiling for, not to asm-generic.
You can't compile a generic kernel module, that would work on a generic architecture. You have to compile it for the particular architecture you're going to use.
I don't know why the asm didn't exist. It should be created as part of the configuration process.
You might get other errors later, if configuration is incomplete in other ways.
The asm includes (such as linkage.h) are architecture specific. There should be a set of directories under:
/usr/src/kernels/(kernel version goes here)/arch
that provide specific includes for the specific CPU architecture you are targeting your code to be compiled for.
Try adding this to your Makefile:
KVERSION :=R(shell uname -r)
and add the kernel and architecture (x86 in this example):
INCDIRS = -I./include -I/usr/src/kernels/$(KVERSION)/include -I/usr/src/kernels/$(KVERSION)/arch/x86
module compiling : asm/linkage.h file not found
This means this particular file was not found in specified DIR, which gets specified when we use -I option with make.
We can either link that asm-generic to asm, if all headers are present in asm-generic, or we can use make utility.
Make utility is preferred in case of building kernel modules.
Create a 'Makefile' in working DIR.
obj-m += hello.o
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
Use of -C option will change to DIR specified before reading the makefiles or doing anything else.
So to avoid this error, use -C option with DIR/lib/modules/$(shell uname -r)/build
By this your program will be able to find required files, you will get hello.ko file.
You can add this to kernel modules by
sudo insmod hello.ko
Similarly you can remove by
sudo rmmod hello
Related
I would like to define a macro for my kernel module by using the -D flag, but I can't figure out how to do it in a custom loadable kernel module.
Just to be clear, to set the macro TEST to 1 I usually do something like:
cc -D TEST=1 file.c -o file
And inside the file.c I have
#if TEST
//do something
#endif
Now, having the same code in a kernel module, how can I set TEST to 1 without touching the code?
This is my the Makefile:
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
Since the -C flag it's recursively calling multiple makefiles, adding -D TEST=1 does not work, I get the following error:
make: invalid option -- 'D'
Anybody knows how to solve this problem?
Thanks in advance.
As suggested by #n.m. in the comments, the solution is to use the EXTRA_CFLAGS. So in my case it would be something like this:
all:
make EXTRA_CFLAGS=-DTEST=2 -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
or simply
EXTRA_CFLAGS:= -D TEST=2
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
I'm trying to build a kernel module in part of a larger project that uses premake5 to generate the Makefiles. However, there's no documentation on using premake to build a kernel module. Is this possible in premake? I'd like to stay consistent with using Premake in the project, if possible.
Thank you in advance for any insights.
My attempt thus far:
project "mymod"
kind "ConsoleApp"
language "C"
location "./"
files { "**.h", "**.c" }
if _ACTION == "clean" then
os.rmdir("obj")
end
buildcommands {
"make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules %{cfg.buildcfg}"
}
rebuildcommands {
"make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules %{cfg.buildcfg} rebuild"
}
cleancommands {
"make clean %{cfg.buildcfg}"
}
buildoptions{ "-Wno-unused-value", "-Wno-sign-compare",
"-Wno-unused-parameter", "-Wno-unused-function",
"-Wno-missing-field-initializers"}
configuration {"Release"}
defines { "NDEBUG" }
flags { "Optimize", "ExtraWarnings", "FatalWarnings", "FloatStrict", "EnableSSE2" }
What follows is an example of a working Makefile for building a kernel module. Granted, it's short and probably doesn't seem worth the trouble of using Premake, but that's because it is missing several object files from the listing, which currently have to be entered manually.
# Note that multiple KBUILD_EXTRA_SYMBOLS Entries are space-separated
KBUILD_EXTRA_SYMBOLS=$(ETHERCAT_TOPDIR)/Module.symvers \
$(HOME_DIR)/shmdrv/Module.symvers \
# the output kernel module
obj-m := corert.o
# the object files used
corert-objs := ctsrt_main.o globals.o ProcessMain.o \
CtsSharedData.o ProcessData.o \
ProcessSdo.o ProcessSimulate.o \
FsmWrite.o
ETHERCAT_TOPDIR = /usr/local/src/ethercat
EXTRA_CFLAGS = -I$(ETHERCAT_TOPDIR)/include -ffast-math -mhard-float
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) \
modules
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
I made Linux external module and because my module need Linux kernel symbol,
I set the Modversion config and make it.(also did make modules_install)
My module was made exactly in lib/modules/(my linux)/extra/
But when I insert my module, shell prints error message
insmod: ERROR: could not insert module oxen_aggregator_module.ko: Invalid module format
This is dmesg contents
[ 341.458351] oxen_aggregator_module: version magic '3.17.8-gentoo-r1 SMP mod_unload modversions ' should be '3.17.8-gentoo-r1 SMP mod_unload '
But I need modversions flag because if I didn't set that flag, Module.symvers have just 0x00000000 addresses.
Could you help me? How can I solve?
My Make file
SRCS = oxen_aggregator_module.c
OBJS = $(SRCS:.c=.o)
obj-m += $(OBJS)
KBUILD_EXTRA_SYMBOLS={/usr/src/linux-$(shell unamr -r)/Module.symvers}
all:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) modules_install
clean:
make -C /lib/modules/$(shell uname -r)/build M=$(PWD) clean
$(RM) Module.markers modules.order
You attempt to load module into kernel, for which it(module) hasn't be built. Or you build module for one kernel, but use KBUILD_EXTRA_SYMBOLS with Module.symvers file, created for another kernel.
If you need to use KBUILD_EXTRA_SYMBOLS with kernel's Module.symvers file, then it is definitely mess with different kernels somewhere.
I have a basic linux device driver module :
#include <linux/init.h>
#include <linux/module.h>
#include <linux/kernel.h>
static int __init hello_init(void)
{
printk(KERN_ALERT "Hello, world \n");
return 0;
}
static void __exit hello_exit(void)
{
printk(KERN_ALERT "Goodbye, world \n");
}
module_init(hello_init);
module_exit(hello_exit);
I am able to compile this module in traditional way which is by using a simple Makefile which uses obj-m , but I want to compile this using command line gcc. This is because I can use gcc -save-temps flag to see the intermediate generated files(this can be particularly helpful to understand as Linux kernel uses lot of preprocessor stuff).
So is there a way to compile using command line gcc ??
EDIT Attaching the Makefile I have used
ifeq ($(KERNELRELEASE),)
KERNELDIR ?= /lib/modules/$(shell uname -r)/build
PWD := $(shell pwd)
EXTRA_CFLAGS+= -save-temps
modules:
$(MAKE) -C $(KERNELDIR) M=$(PWD) modules
modules_install:
$(MAKE) -C $(KERNELDIR) M=$(PWD) modules_install
clean:
rm -rf *.o *~ core .depend .*.cmd *.ko *.mod.c .tmp_versions
.PHONY: modules modules_install clean
else
obj-m := hello.o
endif
Could you try to add "EXTRA_CFLAGS" in your module's Makefile?
such as EXTRA_CFLAGS += -save-temps
Hope it help you!
I don't know how to to that directly in Makefile, but you can generate your .i file by file. From the root directory of the Linux kernel source:
make drivers/media/pci/sta2x11/sta2x11_vip.i
This will generate the .i file. To generate your module source file (which is outside the kernel tree) just use the absolute path to it:
make /path/to/hello.i
It should work
I am trying to run a hello world kernel module but its showing module.h is no present. but i have module.h in /usr/src/linux.2.xx.xx/includes/.Please help me how to set this path?
Try the following in your shell in the directory with the source of your module:
export KDIR=/usr/src/linux.2.xx.xx
make -C $KDIR M=`pwd`
That header should be used via #include <linux/module.h>
try to touch a new makefile which coding like below
ifneq ($(KERNELRELEASE),)
obj-m := hello.o
else
KERNELDIR ?= /lib/modules/$(shell uname -r)/build
PWD := $(shell pwd)
default:
$(MAKE) -C $(KERNELDIR) M=$(PWD) modules
endif
this makefile should name "Makefile"
put it in the same directory with the hello.c file