LinuxQuestions.org - [SOLVED] Kernel module unexpected exit when using insmod

- Linux - Newbie (https://www.linuxquestions.org/questions/linux-newbie-8/)

- - Kernel module unexpected exit when using insmod (https://www.linuxquestions.org/questions/linux-newbie-8/kernel-module-unexpected-exit-when-using-insmod-4175625109/)

Kernel module unexpected exit when using insmod

Hi,
I have written a simple kernel driver program. It has simple module entry and exit functions.

Code:

static int __init driver_entry(void){

 printk(KERN_INFO "module entry");

 return 0;

}

static void __exit driver_exit(void){

 printk(KERN_ALERT "module exited");

} 

module_init(driver_entry);

module_exit(driver_exit);

I created the standard makefile required for compiling module as follows:

Code:

CONFIG_MODULE_SIG=n

obj-m := testmod.o



KERNEL_DIR = /usr/src/linux-headers-$(shell uname -r)



all:

    $(MAKE) -C $(KERNEL_DIR) SUBDIRS=$(PWD) modules

clean:

    rm -rf *.o *.ko *.mod.* *.symvers *.order *~

The module compiles without error.I previously received a module verification error due to signature while insmod hence added the CONFIG_MODULE_SIG=n.
I am using the virtualbox environment running Ubuntu 16.04. However, when I insert the module using insmod, the "module exited" message appears highlighted with red background in dmesg log even before the "module entry" message. The function driver_exit gets executed automatically with the insmod call. Not able to understand this phenomenon?

Thank you.

Is that the previous copy of your driver being removed when you insmod a new one? Did you check with lsmod?

@smallpond, I have checked with lsmod and the module was not loaded before insmod was used. I changed the Makefile to compile with /lib/modules/$(shell uname -r)/build instead of /usr/src/linux-headers-$(shell uname -r).I am posting the full code below:

module1.c

Code:

#include <linux/module.h>

#include <linux/kernel.h>

#include <linux/fs.h>

#include <linux/cdev.h>

#include <linux/semaphore.h>

#include <asm/uaccess.h>

#include <linux/init.h>



static int __init driver_entry(void)

{

        printk(KERN_ALERT "module1: module entered");

        return 0;

}

static void __exit driver_exit(void)

{

        printk(KERN_ALERT "module1: module exit");

}

module_init(driver_entry);

module_exit(driver_exit);



MODULE_AUTHOR("Abhi");

MODULE_DESCRIPTION("Simple driver");

MODULE_LICENSE("GPL");

Makefile:

Code:

CONFIG_MODULE_SIG=n

obj-m := module1.o



KERNEL_DIR = /lib/modules/$(shell uname -r)/build



all:

        $(MAKE) -C $(KERNEL_DIR) M=$(PWD) modules



clean:

        $(MAKE) -C $(KERNEL_DIR) M=$(PWD) clean

The module throws compile error if I try to compile with "sudo make", but compiles successfully without the use of "sudo". I am attaching the log below:

Code:

abhinav@abhinav-VirtualBox:~/tuts/ch1$ ls

Makefile  module1.c



abhinav@abhinav-VirtualBox:~/tuts/ch1$ dmesg



abhinav@abhinav-VirtualBox:~/tuts/ch1$ make 

make -C /lib/modules/4.13.0-36-generic/build M=/home/abhinav/tuts/ch1 modules

make[1]: Entering directory '/usr/src/linux-headers-4.13.0-36-generic'

  CC [M]  /home/abhinav/tuts/ch1/module1.o

  Building modules, stage 2.

  MODPOST 1 modules

  CC      /home/abhinav/tuts/ch1/module1.mod.o

  LD [M]  /home/abhinav/tuts/ch1/module1.ko

make[1]: Leaving directory '/usr/src/linux-headers-4.13.0-36-generic'



abhinav@abhinav-VirtualBox:~/tuts/ch1$ sudo insmod module1.ko



abhinav@abhinav-VirtualBox:~/tuts/ch1$ dmesg

[ 1420.222942] module1: module exit



abhinav@abhinav-VirtualBox:~/tuts/ch1$ lsmod

Module                  Size  Used by

module1                16384  0

snd_intel8x0          40960  2

snd_ac97_codec        131072  1 snd_intel8x0

intel_powerclamp      16384  0

ac97_bus              16384  1 snd_ac97_codec

snd_pcm                98304  2 snd_ac97_codec,snd_intel8x0

crct10dif_pclmul      16384  0

crc32_pclmul          16384  0

ghash_clmulni_intel    16384  0

cryptd                24576  1 ghash_clmulni_intel

joydev                20480  0

intel_rapl_perf        16384  0

snd_seq_midi          16384  0

input_leds            16384  0

snd_seq_midi_event    16384  1 snd_seq_midi

snd_rawmidi            32768  1 snd_seq_midi

snd_seq                65536  2 snd_seq_midi_event,snd_seq_midi

serio_raw              16384  0

snd_seq_device        16384  3 snd_seq,snd_rawmidi,snd_seq_midi

vboxvideo              53248  2

ttm                    94208  1 vboxvideo

drm_kms_helper        167936  1 vboxvideo

snd_timer              32768  2 snd_seq,snd_pcm

drm                  360448  5 vboxvideo,ttm,drm_kms_helper

fb_sys_fops            16384  1 drm_kms_helper

syscopyarea            16384  2 vboxvideo,drm_kms_helper

snd                    81920  11 snd_seq,snd_ac97_codec,snd_timer,snd_rawmidi,snd_intel8x0,snd_seq_device,snd_pcm

vboxguest            294912  1 vboxvideo

mac_hid                16384  0

sysfillrect            16384  2 vboxvideo,drm_kms_helper

i2c_piix4              24576  0

sysimgblt              16384  2 vboxvideo,drm_kms_helper

soundcore              16384  1 snd

parport_pc            32768  0

ppdev                  20480  0

lp                    20480  0

parport                49152  3 lp,parport_pc,ppdev

autofs4                40960  2

hid_generic            16384  0

usbhid                49152  0

hid                  118784  2 hid_generic,usbhid

ahci                  36864  2

e1000                143360  0

psmouse              147456  0

libahci                32768  1 ahci

pata_acpi              16384  0

video                  40960  0



abhinav@abhinav-VirtualBox:~/tuts/ch1$ sudo rmmod module1.ko



abhinav@abhinav-VirtualBox:~/tuts/ch1$ dmesg

[ 1420.222942] module1: module exit

[ 2283.046381] module1: module entered



abhinav@abhinav-VirtualBox:~/tuts/ch1$ lsmod

Module                  Size  Used by

snd_intel8x0          40960  2

snd_ac97_codec        131072  1 snd_intel8x0

intel_powerclamp      16384  0

ac97_bus              16384  1 snd_ac97_codec

snd_pcm                98304  2 snd_ac97_codec,snd_intel8x0

crct10dif_pclmul      16384  0

crc32_pclmul          16384  0

ghash_clmulni_intel    16384  0

cryptd                24576  1 ghash_clmulni_intel

joydev                20480  0

intel_rapl_perf        16384  0

snd_seq_midi          16384  0

input_leds            16384  0

snd_seq_midi_event    16384  1 snd_seq_midi

snd_rawmidi            32768  1 snd_seq_midi

snd_seq                65536  2 snd_seq_midi_event,snd_seq_midi

serio_raw              16384  0

snd_seq_device        16384  3 snd_seq,snd_rawmidi,snd_seq_midi

vboxvideo              53248  2

ttm                    94208  1 vboxvideo

drm_kms_helper        167936  1 vboxvideo

snd_timer              32768  2 snd_seq,snd_pcm

drm                  360448  5 vboxvideo,ttm,drm_kms_helper

fb_sys_fops            16384  1 drm_kms_helper

syscopyarea            16384  2 vboxvideo,drm_kms_helper

snd                    81920  11 snd_seq,snd_ac97_codec,snd_timer,snd_rawmidi,snd_intel8x0,snd_seq_device,snd_pcm

vboxguest            294912  1 vboxvideo

mac_hid                16384  0

sysfillrect            16384  2 vboxvideo,drm_kms_helper

i2c_piix4              24576  0

sysimgblt              16384  2 vboxvideo,drm_kms_helper

soundcore              16384  1 snd

parport_pc            32768  0

ppdev                  20480  0

lp                    20480  0

parport                49152  3 lp,parport_pc,ppdev

autofs4                40960  2

hid_generic            16384  0

usbhid                49152  0

hid                  118784  2 hid_generic,usbhid

ahci                  36864  2

e1000                143360  0

psmouse              147456  0

libahci                32768  1 ahci

pata_acpi              16384  0

video                  40960  0

Can clearly verify that module is loaded and unloaded by using lsmod, but not sure what is actually going behind the execution. It seems driver_exit function is triggered as soon as the module is instatiated by insmod. Could it be still the module verification error due to missing signature?

Is there some compelling reason you are using insmod instead of modprobe?

Quote:

Originally Posted by AwesomeMachine (Post 5828926)

Is there some compelling reason you are using insmod instead of modprobe?

Seems there a lot of work needs to be done with modprobe. It demands the module to be present in /lib/usr/$(shell uname -r) directory. Also I tried modules_install but it throws error that

Code:

cp: cannot stat './modules.order':No such file or directory

Also I found many methods to compile and load LKM, it has become slightly confusing:confused:

Modprobe is more trouble, but it also makes more checks, and it doesn't mysteriously fail like insmod.

Quote:

Originally Posted by AwesomeMachine (Post 5828949)

Modprobe is more trouble, but it also makes more checks, and it doesn't mysteriously fail like insmod.

I used modprobe this time by making a symbolic link and depmod. No change in the output though.
It seems that the driver_entry and driver_exit are behaving exactly opposite than intended. driver_exit is executed when the module is instantiated and driver_entry is executed when rmmod is used. The message indicating module is out-of-tree build and module verification failure appears initially every time I restart the system and load the module for the first time (even though CONFIG_MODULE_SIG=n in makefile). I also checked that CONFIG_MODULE_SIG_FORCE is not enabled.

Could it be that dmesg log is lagging behind or something similar?

Try tailing /var/log/messages and watch it as you try to load the module. It should give some clue. You could also enable kernel debugging.

Found the solution. It was the most stupid thing I could imagine, just added "\n" at the end of printk message. Guess the module was working properly but dmesg didn't show correct log until I added a newline at the end of the KERN_INFO & KERN_ALERT message.
But still wonder how this issue prompted "module exited" before "module entry" ?