02.Building, Configuring kernel for Device Driver Development

spatlou

Go to /lib/modules/$(uname -r)/build

at this path you should have Makefile (atleast)

if not the follow. http://www.emblogic.com/members/document_view/1258

If you face any issue, error, dough or want any clarifications. Post it here.

dhannjayv1996

after giving command "make menuconfig" got following error...

<command-line>: fatal error: curses.h: No such file or directory

compilation terminated.

make[1]: *** [scripts/Makefile.host:108: scripts/kconfig/mconf.o] Error 1

make: *** [Makefile:540: menuconfig] Error 2

neharajput3193

@page { margin: 2cm }
p { margin-bottom: 0.25cm; line-height: 120% }

When
launching make command I obtain the following warning:

CC [M]
/home/neharajput/pd/neha/kernal/registermodule/init1_func.o

CC [M]
/home/neharajput/pd/neha/kernal/registermodule/cleanup_func.o

LD [M]
/home/neharajput/pd/neha/kernal/registermodule/test.o

Building modules, stage 2.

MODPOST 1 modules
WARNING:
"majarno"
[/home/neharajput/pd/neha/kernal/registermodule/test.ko] undefined!
I
think this is a linking problem, but I do not know how to fix it. Can
anyone help me?

hitesh.arora1997

whenever i try to insert my module in kernel , it do get insert but some error comes when i do dmesg

loading out-of-tree module taints kernel.

[ 2734.820564] newker: module license 'unspecified' taints kernel.

[ 2734.820566] Disabling lock debugging due to kernel taint

[ 2734.820634] newker: module verification failed: signature and/or required key missing - tainting kernel

tanejasakshi1999

CHK     include/linux/version.h
CHK     include/generated/utsrelease.h
CC      kernel/bounds.s
In file included from include/linux/compiler.h:48:0,
                 from include/linux/stddef.h:4,
                 from include/linux/posix_types.h:4,
                 from include/linux/types.h:17,
                 from include/linux/page-flags.h:8,
                 from kernel/bounds.c:9:
include/linux/compiler-gcc.h:90:1: fatal error: linux/compiler-gcc7.h: No such file or directory
#include gcc_header(__GNUC__)

^~~

compilation terminated.
make[1]: *** [/usr/src/kernels/linux-2.6.38/./Kbuild:36: kernel/bounds.s] Error 1
make: *** [Makefile:955: prepare0] Error 2

ERROR --> for kernel 2.6.38

tyagivaibhav92

1. Having System Model - HP Laptop 15-da0xxx

2. Unfortunately, Linux doesn’t include a WiFi driver for my device. I need to disable secure boot and then install these drivers from the link below:

https://we.tl/t-elG77ZwF2X

steps to install :-

sudo apt update && sudo apt install git dkms

cd Downloads

git clone -b extended --single-branch https://github.com/lwfinger/rtlwifi_new.git unofficial

cd rtlwifi_new

/////////////////////////////////////////////////////////////////////////////////////////////////

Turn off secure boot in booting options

install fedora 29 on pc

cp -rf unofficial/ rtlwifi_new/

cd rtlwifi_new/

yum update

reboot

now, install kernel headers and kernel development tools for the current kernel version.

sudo yum install "kernel-devel-uname-r == $(uname -r)"

yum install make

yum install dkms

/////////////////////////////////////////////////////////////////////////////////////////////////

Now, either you can run:

make

sudo make install

/////////////////////////////////////////////////////////////////////////////////////////////////

sudo dkms add ../rtlwifi_new

sudo dkms build rtlwifi-new/0.6

sudo dkms install rtlwifi-new/0.6

sudo modprobe -v rtl8723de ant_sel=2

////////////////////////////////////////////////////////////////////////////////////////////////

3. Now tried to build kernel linux- 5.10.8

By make install got following error

sh ./arch/x86/boot/install.sh 5.10.8 arch/x86/boot/bzImage \ system.map "/boot"

Error! Bad return status for module build on kernel : 5.10.8 (x86_64)

consult /var/lib/dkms/rtlwifi - new/0.6/build/make.log for more information.

4.Got log in make.log file

Didn't able to get wifi drivers during build can someone suggest something ?

tyagivaibhav92

kvanshaj

I am facing this error after "make install" command :

dracut-install: No SOURCE argument given
Usage: dracut-install -D DESTROOTDIR [-r SYSROOTDIR] [OPTION]... -a SOURCE...
or: dracut-install -D DESTROOTDIR [-r SYSROOTDIR] [OPTION]... SOURCE DEST
or: dracut-install -D DESTROOTDIR [-r SYSROOTDIR] [OPTION]... -m KERNELMODULE [KERNELMODULE …]

Install SOURCE (from rootfs or SYSROOTDIR) to DEST in DESTROOTDIR with all needed dependencies.

KERNELMODULE can have the format:
     <absolute path> with a leading /
     =<kernel subdir>[/<kernel subdir>…] like '=drivers/hid'
     <module name>

Can anyone help me with this

rahul

Kernel compression Mode

Note

During kernel compile

Select ZSTD kernel compression.

If you forgot to select this mode. Then bzimage will not be made. So, After make. Install ZSTD ( yum install zstd / dnf install zstd )

then do make again you get your bzimage in 10 min

Then coontinue your install process

ritde1

How to go about solving the "Failed to load init script" ?

jagjeet.singh9650

I am trying to install kernel and tried to run make command , it just gives an error in between about openssl , that unable to run command openssl , i tried installing openssl-devel , it says already installed but i dont get worksation on boot up.

jagjeet.singh9650

i dont understand when i try to install openssl it says already installed but when i run make command it gives error about openssl

ritde1

Try reinstalling openssl

abdulrehmanjmi

when I am doing make command getting this error

root@fedora linux-5.10.8]# make
SYNC include/config/auto.conf.cmd
LEX scripts/kconfig/lexer.lex.c
/bin/sh: line 1: flex: command not found
make[2]: *** [scripts/Makefile.host:9: scripts/kconfig/lexer.lex.c] Error 127
make[1]: *** [Makefile:602: syncconfig] Error 2
Makefile:688: include/config/auto.conf.cmd: No such file or directory
make: *** [Makefile:710: include/config/auto.conf.cmd] Error 2

ritde1

It's logically correct to compile linux kernel 2.6.38 with gcc 11 but why are there so many errors and warnings?

jagjeet.singh9650

Getting error while compiling linux 2.6.36 , fatal error no such file gcc.h in include/linux/compiler-gcc.h , also I dont see if it is any gcc compiler version issue.

ritde1

Can't use 'defined(@array)' (Maybe you should just omit the defined()?) at kernel/timeconst.pl line 373.

make[1]: *** [/usr/src/linux-2.6.38/kernel/Makefile:140: kernel/timeconst.h] Error 255

make: *** [Makefile:923: kernel] Error 2

rahul

I am compiling kernel 2.6.38 with gcc 10.

It gives error #include gcc_header(__GNUC__)

I checked files and folders od kernel. I found that kernel have dependencies of gcc version 3 or version 4.

That's why i am having error.

Now i want to know how to bypass gcc4 dependecies and compile it with gcc10.

rahul

I tried to install gcc 4.9.3 on fedora 33

I got error given below.

How to solve it

cfns.gperf:101:1: error: ‘const char* libc_name_p(const char*, unsigned int)’ redeclared inline with ‘gnu_inline’ attribute

cfns.gperf:26:14: note: ‘const char* libc_name_p(const char*, unsigned int)’ previously declared here

cfns.gperf:26:14: warning: inline function ‘const char* libc_name_p(const char*, unsigned int)’ used but never defined

spatlou

Follow these resources for Linux Workspace Setup.

https://www.emblogic.com/member/course/module/11/151

deepak4u106

While trying to boot kernel my screen got stuck on :->

Loading Linux 5.14.8 ...

Loading intial ramdisk ...

_

deepak4u106

if i am trying to boot with 5.10.14

My screen got blank after kernel loading

deepak4u106

Everything is installed perfectly

But when i try to boot with 5.10.14 and 5.1.0 i got an error :->

kobject_add_internal failed for wmi_bus-PNP0C14:01 with -EEXISTM, dont try to register same thing with the same name in the same directory.

misc_register failed.

AND SCREEN GOT BLANK

swatiupendra1999

kernel version :5.9.9

error:../../grub-core/kern/efi/sb.c:150:bad shim signature.

error:../../grub-core/loader/i386/efi/linux/linux.c:208:you need to load the kernel first

lksharma906

Fedora Kernel 5.14.8

Issue is as below and reported when we do make.

BTF: .tmp_vmlinux.btf: pahole (pahole) is not available

Failed to generate BTF for vmlinux

Try to disable CONFIG_DEBUG_INFO_BTF

make: *** [Makefile:1183: vmlinux] Error 1

It can be fixed using below command :

sudo dnf install dwarves

Another Issue which came on my machine was realted to zstd, issue was like below:

arch/x86/tools/insn_decoder_test: success: Decoded and checked 7223888 instructions

TEST posttest

arch/x86/tools/insn_sanity: Success: decoded and checked 1000000 random instructions with 0 errors (seed:0xdfcd279d)

ZSTD22 arch/x86/boot/compressed/vmlinux.bin.zst

/bin/sh: line 1: zstd: command not found

make[2]: *** [arch/x86/boot/compressed/Makefile:136: arch/x86/boot/compressed/vmlinux.bin.zst] Error 127

make[2]: *** Deleting file 'arch/x86/boot/compressed/vmlinux.bin.zst'

make[1]: *** [arch/x86/boot/Makefile:115: arch/x86/boot/compressed/vmlinux] Error 2

make: *** [arch/x86/Makefile:270: bzImage] Error 2

[lokesh@fedora linux-5.14.8]$

To fix this : - sudo dnf install zstd

Pravjotsir

update-grub

...

The update-grub command was created to make things easier - it is simply a shell script in /usr/sbin/:

#!/bin/sh
set -e
exec grub-mkconfig -o /boot/grub/grub.cfg "$@"

If it is not there, you can make your own. To do that, run this, and paste in the above script:

sudo nano /usr/sbin/update-grub

Save with Ctrl+O, and exit with Ctrl+X.

Then run these:

sudo chown root:root /usr/sbin/update-grub
sudo chmod 755 /usr/sbin/update-grub

And you should now be able to run update-grub. :-)

Pravjotsir

Path: /etc/default

Update this file:

Filename: grub

1 GRUB_TIMEOUT=5

2 GRUB_DISTRIBUTOR="$(sed 's, release .*$,,g' /etc/system-release)"

3 GRUB_DEFAULT=saved

4 GRUB_DISABLE_SUBMENU=true

5 GRUB_TERMINAL_OUTPUT="console"

6 GRUB_CMDLINE_LINUX="rhgb quiet"

7 GRUB_DISABLE_RECOVERY="true"

8 GRUB_ENABLE_BLSCFG=true

9 GRUB_TIMEOUT_STYLE=menu

bhatnagarashish1998

While booting the new kernel, the kernel is not booting up correctly it first shows below error:

[FAILED] Failed to start Load Kernel Modules

And after that the screen will be blackout but the power of the machine is still active.

jyotipandey9813

@abdulrehmanjmi

Please install flex by using command : yum install flex .

jyotipandey9813

@swatiupendra1999

Please disable secure boot . if you boot your system with secure boot disabled then this error while booting will not occur and you will be able to boot successfully .

lksharma906

Just FYI :- Sir wrote yum install flex, but we can use dnf also as dnf is updated and latest version of yum as with dnf we get a little extra edge that if the enabled repository is not there it will skip that and look for installation from other available repo's however yum may not work in that case, but i read it somewhere that dnf can also remove kernel repo's which cannot be done by yum. I am not really sure on this one so yum looks safe inherently.

rakeshsingh59432

compile the kernel and compress the botImage

[root@localhost linux-5.16.8]# cp arch/x86/configs/i386_defconfig.config

cp: missing destination file operand after 'arch/x86/configs/i386_defconfig.config'

Try 'cp --help' for more information.

rakeshsingh59432

i reslove the above issue go to cp --help

rakeshsingh59432

when i need to make bootloder about the arrival .

cp arch/x86/boot/bzImage/boot/vmlinuz

cp: missing destination file operand after 'arch/x86/boot/bzImage/boot/vmlinuz'

Try 'cp --help' for more information.

nirmalpg124

KERNEL INSTALLATION USING SEQUENCE OF MAKE COMMANDS INVOLVING TWO HRS: -

Result is that MORE MODULES ARE BUILD INTO THE KERNEL FOR PRODUCTION ENV as compared to simple usb key insatallation that takes only 15 mins.

using linux kernel ver 6.2.9-300.fc38.x86_64(Fedora 38) & trying to install latest 6.4.2 version

getting following error while running make menuconfig

make[1]: *** [scripts/kconfig/Makefile:215: scripts/kconfig/mconf-libs] Error 1
make: *** [Makefile:692: menuconfig] Error 2

sol:

for menuconfig installed lex bison libncurses using yum install __ & menuconfig ran.

other errors are:

/home/nirmal/linux-6.4.2/tools/objtool/include/objtool/elf.h:10:10: fatal error: gelf.h: No such file or directory

10 | #include <gelf.h>

sol

yum install -y elfutils-libelf-devel

/bin/sh: line 1: openssl: command not found

make[2]: *** [certs/Makefile:53: certs/signing_key.pem] Error 127

make[1]: *** [scripts/Makefile.build:494: certs] Error 2

make: *** [Makefile:2026: .] Error 2

sol

yum install openssl

BTF: .tmp_vmlinux.btf: pahole (pahole) is not available

Failed to generate BTF for vmlinux

Try to disable CONFIG_DEBUG_INFO_BTF

make[1]: *** [scripts/Makefile.vmlinux:35: vmlinux] Error 1

make: *** [Makefile:1250: vmlinux] Error 2

sol:

[root@fedora linux-6.4.2]# dnf install dwarves

nirmalpg124

make[2]: *** [arch/x86/boot/compressed/Makefile:146: arch/x86/boot/compressed/vmlinux.bin.zst] Error 127

make[2]: *** Deleting file 'arch/x86/boot/compressed/vmlinux.bin.zst'

make[1]: *** [arch/x86/boot/Makefile:116: arch/x86/boot/compressed/vmlinux] Error 2

make: *** [arch/x86/Makefile:282: bzImage] Error 2

sol:

[root@fedora linux-6.4.2]# yum install zstd

gaurav133232

error: ../../grub-core/kern/efl/sb.c:182:bad shim signature.

error: ../../grub-core/loader/I386/lefl/inux.c:258: you need to load the kernel first.

press any key to continue ...

Facing this error during boot?

Solution:

Turn secure boot off

preetamverma9076

CC [M] kernel/kheaders.o
debian/rules:6: recipe for target 'build' failed
make[2]: *** [build] Error 2
dpkg-buildpackage: error: debian/rules build subprocess returned exit status 2
scripts/Makefile.package:73: recipe for target 'deb-pkg' failed
make[1]: *** [deb-pkg] Error 2
Makefile:1469: recipe fpr target 'deb-pkg' failed
make: *** [deb-pkg] Error

nirmalpg124

DRIVER INIT & CLEANUP FUNCTIONS

getting following error

/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:6:29: error: expected ‘)’ before ‘int’

6 | module_param(size_of_device, int, S_IRUGO);

| ^~~~

| )

/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:7:24: error: expected ‘)’ before ‘int’

7 | module_param(no_of_reg, int, S_IRUGO);

| ^~~~

| )

/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:8:26: error: expected ‘)’ before ‘int’

/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:11:1: error: unknown type name ‘dev_t’

sol : typo errors sorted out

nirmalpg124

/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:8:26: error: expected ‘)’ before ‘int’
/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:11:1: error: unknown type name ‘dev_t’
11 | dev_t devid, devno;
| ^~~~~
/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:13:1: error: unknown type name ‘Dev’
13 | Dev *dev;
| ^~~
/home/nirmal/Documents/Documents_09_08_23/01_open_release/init.c:17:19: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘initFunc’
17 | static int __init initFunc(void)
| ^~~~~~~~

sol : typo errors sorted out

nirmalpg124

GENERAL ERROR REPORTING BY COMPILER FROM KCONFIG KBUILD MAKE LANGUAGE special errors:

make[2]: *** [scripts/Makefile.build:252: /home/nirmal/Documents/Documents_09_08_23/01_open_release/readDevice.o] Error 1

make[1]: *** [Makefile:2026: /home/nirmal/Documents/Documents_09_08_23/01_open_release] Error 2

make[1]: Leaving directory '/home/nirmal/linux-6.4.2'

make: *** [Makefile:15: all] Error 2

going through func codes ...

& found

error 1 is for not updated files issues & error 2 is for real errors, my .ko & my mod.c files were not getting generated inside modules folder. The other .o files were there. There was issue with struct file_operations fops used for mapping open, release, read & write application system calls vis a vis to corresponding kernel routines.

So, I included this once in init file and externally included via declarations.h in all other files. The prob solved.

nirmalpg124

MATCHING FUNC PARAMETERS STRICTLY IN DECLARATIONS WITH FUNC OPERATIONS IN APPLICATION & KERNEL PROGR.

getting following error:

In file included from /home/nirmal/Documents/Documents_09_09_23/01_open_release/init.c:4:

/home/nirmal/Documents/Documents_09_09_23/01_open_release/operations.h:5:18: error: initialization of ‘ssize_t (*)(struct file *, char *, size_t, loff_t *)’ {aka ‘long int (*)(struct file *, char *, long unsigned int, long long int *)’} from incompatible pointer type ‘int (*)(struct file *, const char *, size_t, off_t *)’ {aka ‘int (*)(struct file *, const char *, long unsigned int, long int *)’} [-Werror=incompatible-pointer-types]

5 | read :readDevice,

| ^~~~~~~~~~

/home/nirmal/Documents/Documents_09_09_23/01_open_release/operations.h:5:18: note: (near initialization for ‘fops.read’)

/home/nirmal/Documents/Documents_09_09_23/01_open_release/operations.h:6:18: error: initialization of ‘ssize_t (*)(struct file *, const char *, size_t, loff_t *)’ {aka ‘long int (*)(struct file *, const char *, long unsigned int, long long int *)’} from incompatible pointer type ‘int (*)(struct file *, const char *, size_t, off_t *)’ {aka ‘int (*)(struct file *, const char *, long unsigned int, long int *)’} [-Werror=incompatible-pointer-types]

6 | write :writeDevice

sol: defined the createScull func in declarations.h that was missing

nirmalpg124

MKNOD

got error while calling mknod for creating an inode file. This inode file works as an intermediary for the application programs and kernel programs.

error: open : No such file or folder exists.

while calling mknod parameter were used as c (for character) major no( 1) minor no (0)

& file was created by the root user.

sol: - during dmesg noted down the major no provided to the device driver by the kernel as the same was fed in the calling of mknod for majorno parameter during runtime. that solved.

secondly file ownership was changed to current user using chown

nirmalpg124

got error of kernel NULL POINTER DEREFERENCE error:

[10562.394012] initFunc: Device Registration Successful. Minor No: 19
[10562.394014] initFunc: End
[10592.557848] Begin :
:, _func_
[10592.557853] Begin:trimDevice
[10592.557860] trimDevice: 20:
[10592.557863] trimDevice: Nothing to trim.
[10592.557866] trimDevice: End
[10592.557868] openDevice:End
[10600.156197] writeDevice: Begin
[10600.156212] createScull: Begin
[10600.156229] BUG: kernel NULL pointer dereference, address: 0000000000000008
[10600.156236] #PF: supervisor read access in kernel mode
[10600.156241] #PF: error_code(0x0000) - not-present page
[10600.156246] PGD 800000000745c067 P4D 800000000745c067 PUD 745d067 PMD 0
[10600.156261] Oops: 0000 [#2] PREEMPT SMP PTI

APPROACH: -

error in this instruction pointed by RIP:

[10600.156280] RIP: 0010:createScull+0x169/0x270 [mycdd]
[10600.156296] Code: 01 83 d8 ff 89 04 24 85 c0 7e 67 45 31 ff 45 31 f6 eb 17 8b 0d a4 25 00 00 41 83 c6 01 41 83 c7 01 44 39 3c 24 74 4b 48 63 f9 <48> 8b 43 08 4d 63 ee be c0 0c 00 00 4e 8d 24 e8 e8 b2 7b 3c f5 49
[10600.156301] RSP: 0018:ffffa57cc2fd3d38 EFLAGS: 00010246
[10600.156305] RAX: 0000000000000013 RBX: 0000000000000000 RCX: 0000000000000002
[10600.156308] RDX: 0000000000000001 RSI: 0000000000000000 RDI: 0000000000000002

sol:

crash> bt
PID: 8296 TASK: ffff8aa6be7fd180 CPU: 2 COMMAND: "crash"
(active)
crash> dis -l createScull+0x169
0xffffffffc0fae239 <createScull+361>: mov 0x8(%rbx),%rax
crash> dis -l createScull+0x270
dis: invalid expression: createScull+0x270 evaluates to: __pfx_writeDevice+0

nirmalpg124

REMEDY for NULL POINTER DEREFERENCE error upto some extent :

crash>dis -r 0xffffffffc0fae239

OxFFFffffcofaelee <createscull+286>: cmp %rldd,%r13d
OXFFFFffffcofaelfl <createscull+289>: jne OXFFffffffcofaelae <createscull+208:
OxFFFFffffcofaelf3 <createscull+291>: mov 0x25d3(%rip) ,%ecx # OXFFFFFfffcafboTce <size of reg:
OXFFFFffffcofaelfo <createscull+297>: mov
OxFFFFfffcofaelfd <createscull+301>: xor
OxFFFFffffcofaelff <createscull+303>: movslg
OXFFFFffffcofae202 <createscull+306>: mov %rbx,0x25b7(%rip) # OXFFFFFfffcafbarce <first:
OXFFFFffffcofae200 <createscull+313>: div %rdi
OxFFFFffffcofaedoc <createscull+316>: cmp $6xL,%rdx
OXFFFFffffcofae210 <createscull+320>: sbb SOxFFFFFFFf,%eax
OxFFFFffffcofae213 <createscull+323>: mov eax, (%rsp.
OxFFFFffffcofae216 <createscull+326>: test f%eax,%eax
OXFFFFffffcofae21s <createscull+328>: jle OXFFFfffffcafae2sl <createscull+433:
OxFFFFfffcofaela <createscull+330>: xor
OxFFFFffffcofae2ld <createscull+333>: xor % %
OXFFFFffffcofae220 <createscull+336>: jmp OXFFFfffffcafae239 <createscull+36l:
OxFFFFffffcofae22) <createscull+338>: mov 0x25a4(%rip) ,%ecx # OXFFFFFfffcafboTce <size of reg:
OxFFFFfffcofae22s <createscull+3ad>: add $ex1,%rlad
OxFFFFffffcofae2ac <createscull+3as>: add $ex1,%rlsd
OxFFFFfffcofae230 <createscull+3s2>: cmp %r1sd, (%rsp
OXFFFFffffcofae23a <createscull+3s6>: je OXFffffffcafae2sl <createscull+433:
OxFFFfffcofae236 <createscull+3s8>: movslq %ecx,%rd
OXFFFFffffcofae230 <createscull+361>: mov OX8(%rbx),%rax

execution halted at the last intruction OXFFFFffffcofae230 producing kernel null pointer dereference error

got following error prone section of code in createScull by analysing assembly instr. :

createScull.c :

for( l=i=0; l<tnoq ; l++ )        // alloting space for all quantums of memory
        {

            printk(KERN_INFO "i=%d, l=%d \n", i, l);     // DEBUG
        itemPtr->data[i] = kmalloc(size_of_reg , GFP_KERNEL ) ;
        totalmemall +=2;

}

which was not running .. & further it was not running due to a faulty code:

first = itemPtr ;

--> after traversing through the whole set of nodes, itemPtr needed to be intialised to first instead as:

itemPtr = first;

When corrected, the error was removed.

nirmalpg124

THOUGHT PROCESS BEHIND IOCTLS in character device development: -

nr is ordinal no or sequence no. of command we r defining in our custom ioctl.h for our device

magic no is any unique no. for a file that uniquely defines that file

we r defining GETREGSIZE as a macro which is further defined or expanded as _IOR() .., .. ..) , when we give GETREGSIZE

in applicatinon this MACRO gets expanded to further definition as described in our custom ioctl.h

MAGIC + ordinal no uniquely identify each ioctl func

size of data is 3rd argument which is to be gicen for ioctl func with read write or read write parameters

this is optional

nirmalpg124

IOCTL COMMANDS: KEY OBSERV. : -

'Write' and 'read' are from the user's point of view, just like the

system calls 'write' and 'read'. For example, a SET_FOO ioctl would

be _IOW, although the kernel would actually read data from user space;

a GET_FOO ioctl would be _IOR, although the kernel would actually write

data to user space.

ref: ioctl-numers.rst

nirmalpg124

CHARACTER DEVICE DEVELOPMENT INTRICACIES WITH IOCTL :

while inserting the module we are assigning values to the global variables of size_of_reg, no_of_reg, etc via module_param func in kernel

using these global variables we are initialising all no . of devices with said values in kernel init.h

when using defaults.h in kernel we are also assigning default values to these variables

when using ioctl.h via GETREGSIZE, SETREGSIZE etc we are getting n setting values to these variables for particular device

another obs: - when we use same device driver for usb multiple times while inserting & removing usb the kernel assigns a different device no. to the device each time

nirmalpg124

ACCESS_OK

validity of user space pointer if valid or not : -

applying if( access_ok ) for all get and set operations in ioctl.. as VERIFY_READ for get & VERIFY_WRITE for set

createScull is a type of space of allocated memory in itself in which data is allocated in three rounds of malloc allocation .. 1 for item, 2 for data pointer, & 3 for the quantums or registers..

access_ok not accepting 3 arguments ... error description: func passing 3 arguments takes 2 arguements

access_ok ( 3 args) defined as __ access_ok( 2 args ) in uaccess.h

if( access_ok( ldev , sizeof( Dev ) ) )

{

#ifdef DEBUG

printk( KERN_INFO "Permission denied\n" );

#endif

}

nirmalpg124

KERNEL ASSIGNING DIFFERENT DEVICE NO EACH TIME DRIVER INSERTED

when inserting the character device driver into kernl it should be taken care of while assigning device node no

when i had installed ver 6.4.2 then it gave device node no 237 when i reinstalled ver 6.4.2 after some software failure it gave node no 235

so the code was not running at first then when given the node no 235 it ran as previously successfully

nirmalpg124

WITH LSEEK OPERATION IN READ

seeking from SEEK_SET & giving offset 3

code is reading well

in Application:

openDev: file is opened in writing mode with fd : 3
APP:writeDev: Begin
writeDev: Mode is writing with fd = 3
writeDev: Please enter the string you want to write
NirmalPatel2
APP:writeDev: Wrote 12 bytes successfully
APP:writeDev: End

openDev: file is opened in reading mode with fd: 4
readDev: Begin
readDev: Please enter origin for reading
readDev: 1. From beginning
1
readDev.c: Please enter the offset
3
readDev.c: pos: 3
readDev.c: New Position after lseek: 3
readDev: buff: 0x7fa59cd75e40
readDev: buff+pos: 0x7fa59cd75e43
APP:readDev: Read 9 bytes successfully as: malPatel2
APP:readDev: End

in kernel :

Begin: lseekDevice
[ 4072.268569] lseekDevice: Before f_pos: 0
[ 4072.268572] lseekDevice: Before offset:3
[ 4072.268574] SEEK_SET
[ 4072.268575] lseekDevice: Afterf_pos: 3
[ 4072.268577] lseekDevice: After offset: 3
[ 4072.268578] End: lseekDevice
[ 4072.268596] readDevice: Begin
[ 4072.268599] readDevice: Read buffer size : 100
[ 4072.268602] readDevice: loff: 3
[ 4072.268604] readDevice: size of data: 12
[ 4072.268605] readDevice: f_pos: 3
[ 4072.268607] readDevice:*loff= 3 , size_of_reg = 2, oiq: 1
[ 4072.268609] l = *loff/ldev->size_of_reg=1
[ 4072.268611] readDevice: before copy_to_user noq = 5noctr= 1, nocsr = 0, i= 1, oiq=1
[ 4072.268614] readDevice: not = 0, curr->data[i] = rm;[\x91\xef\x89ܚV;[\x91\xef\x89\xd4102
[ 4072.268616] readDevice: after copy_to_user noq = 5noctr= 8, nocsr = 1, i= 1, oiq=1
[ 4072.268619] readDevice: before copy_to_user noq = 5noctr= 2, nocsr = 1, i= 2, oiq=0
[ 4072.268621] readDevice: not = 0, curr->data[i] = al;[\x91
[ 4072.268622] readDevice: after copy_to_user noq = 5noctr= 6, nocsr = 3, i= 2, oiq=0
[ 4072.268625] readDevice: before copy_to_user noq = 5noctr= 2, nocsr = 3, i= 3, oiq=0
[ 4072.268627] readDevice: not = 0, curr->data[i] = Pa;[\x91\xef\x88\x14
[ 4072.268628] readDevice: after copy_to_user noq = 5noctr= 4, nocsr = 5, i= 3, oiq=0
[ 4072.268630] readDevice: before copy_to_user noq = 5noctr= 2, nocsr = 5, i= 0, oiq=0
[ 4072.268633] readDevice: not = 0, curr->data[i] = te;[\x91\xef\x8ed
[ 4072.268634] readDevice: after copy_to_user noq = 5noctr= 2, nocsr = 7, i= 0, oiq=0
[ 4072.268636] readDevice: before copy_to_user noq = 5noctr= 2, nocsr = 7, i= 1, oiq=0
[ 4072.268638] readDevice: not = 0, curr->data[i] = l2;[\x91\xef\x82\xfcjY;[\x91\xef\x82\xf4RZ;[\x91\xef\x82\xecI
[ 4072.268640] readDevice: after copy_to_user noq = 5noctr= 0, nocsr = 9, i= 1, oiq=0
[ 4072.268642] End :readDevice

nirmalpg124

but when offset given as 4 with SEEK_SET

getting following error:

in Application:

readDev.c: New Position after lseek: 4
readDev: buff: 0x7f10ae5f5e40
readDev: buff+pos: 0x7f10ae5f5e44
APP:readDev: Read 6 bytes successfully as: alPate
APP:readDev: End
readDev: End

in kernel:

readDevice: loff: 4
[ 5791.976996] readDevice: size of data: 12
[ 5791.976998] readDevice: f_pos: 4
[ 5791.977001] readDevice:*loff= 4 , size_of_reg = 2, oiq: 0
[ 5791.977005] l = *loff/ldev->size_of_reg=2
[ 5791.977007] readDevice: before copy_to_user noq = 4noctr= 2, nocsr = 0, i= 2, oiq=0
[ 5791.977012] readDevice: not = 0, curr->data[i] = alfR\x98\xb2&\x8c\xaa\xfffR\x98\xb2&\x84\xf2\xfdfR\x98\xb2&\xbc\xfa\xfafR\x98\xb2&\xb4io.max
[ 5791.977014] readDevice: after copy_to_user noq = 4noctr= 6, nocsr = 2, i= 2, oiq=0
[ 5791.977018] readDevice: before copy_to_user noq = 4noctr= 2, nocsr = 2, i= 3, oiq=0
[ 5791.977022] readDevice: not = 0, curr->data[i] = PafR\x98\xb2(\x84B\xfffR\x98\xb2(\xbc
[ 5791.977024] readDevice: after copy_to_user noq = 4noctr= 4, nocsr = 4, i= 3, oiq=0
[ 5791.977028] readDevice: before copy_to_user noq = 4noctr= 2, nocsr = 4, i= 0, oiq=0
[ 5791.977032] readDevice: not = 0, curr->data[i] = tefR\x98\xb2+|\xc2\xf1fR\x98\xb2+tZ\xf7fR\x98\xb2+l\xfa\xf5fR\x98\xb2+d\xca\xf9fR\x98\xb2,\x9c\xba\xf1fR\x98\xb2,\x94\xc2\xfffR\x98\xb2,\x8c*\xfbfR\x98\xb2,\x842\xf7fR\x98\xb2,\xbcrmfR\x98\xb2,\xb4\x92\xf9fR\x98\xb2,\xaclo
[ 5791.977033] readDevice: after copy_to_user noq = 4noctr= 2, nocsr = 6, i= 0, oiq=0
[ 5791.977037] End :readDevice

nirmalpg124

when code corrected to

*loff = *loff -1;

&

l = ( *loff )/ (ldev->size_of_reg);

results were:

String: NirmalPatel2

offset -3

read - rmalPatel2

offset -4

read - malPatel2

for offset - 5

read - alPate // code not working current pointer not reading last register

nirmalpg124

again code corrected to :

*loff = *loff - 1;

        if( ldev->size_of_data < size - *loff)
        {
                lsize = noctr = ldev->size_of_data - *loff;
                noq = ldev->size_of_data/ldev->size_of_reg;

        }
        else
        {
                lsize = noctr = size - *loff;
                noq = size/ldev->size_of_reg;
        }

then for size_of_data = 12

code running for offset = 3, 5, 7 , 4, 6 , 8

code not running for offset = 9

nirmalpg124

ERROR DETECTION & ARRIVING TO THE COMPLICATED ERROR

for offset = 9

error:

readDevice: Begin
[ 2302.541734] readDevice: Read buffer size : 100
[ 2302.541737] readDevice: loff: 9
[ 2302.541739] readDevice: size of data: 12
[ 2302.541741] readDevice: f_pos: 9
[ 2302.541745] readDevice: noq: 6
[ 2302.541747] readDevice: lsize=4, ldev->size_of_data=12, size=100, *loff=8
[ 2302.541750] readDevice:*loff= 8 , size_of_reg = 2, oiq: 0
[ 2302.541753] l = *loff/ldev->size_of_reg=4
[ 2302.541756] readDevice: before copy_to_user l = 4, noq = 6noctr= 2, nocsr = 0, i= 4, oiq=0
[ 2302.541772] general protection fault, probably for non-canonical address 0x732e79726f6d656d: 0000 [#5] PREEMPT SMP PTI
[ 2302.541779] CPU: 3 PID: 15509 Comm: app Tainted: G D OE 6.4.2 #2
[ 2302.541784] Hardware name: To be filled by O.E.M. To be filled by O.E.M./H61, BIOS 4.6.5 10/22/2020
[ 2302.541788] RIP: 0010:string+0x48/0xe0

nirmalpg124

FAULT HUNTING

for offset = 9 the fault was that

since curr pointer starts with the first item & keeps on proceeding with curr pointing to curr->next

here offset = 9, then noctr ( no of char to read )= 12-( 9-1 ) = 4

but code is:

for( i=l; l<=noq; l++ )
{

}

so printing values in kernel gives: -

[ 3205.879479] readDevice: noq: 5
[ 3205.879480] readDevice: lsize=4, ldev->size_of_data=12, size=100, *loff=8
[ 3205.879483] l = *loff/ldev->size_of_reg=4
[ 3205.879484] readDevice: before copy_to_user l = 4, noq = 5noctr= 2, nocsr = 0, i= 4, oiq=0
[ 3205.879487] readDevice: i=4, not = 0, curr->data[i]+oiq =
[ 3205.879488] readDevice: after copy_to_user noq = 5noctr= 2, nocsr = 2, i= 4, oiq=0
[ 3205.879490] readDevice: before copy_to_user l = 5, noq = 5noctr= 2, nocsr = 2, i= 5, oiq=0
[ 3205.879492] readDevice: i=5, not = 0, curr->data[i]+oiq =
[ 3205.879494] readDevice: after copy_to_user noq = 5noctr= 0, nocsr = 4, i= 5, oiq=0
[ 3205.879495] End :readDevice

though l ( variable L) keeps on counting till noq ( from 0 to 5 )

i ( variable i has value 0 to 3 then again 0 to 3 & so on.. )

here value of i is 4 that doesn't exist

so code not reading beyond offset >9

nirmalpg124

HIGHER DEGREE RESOLUTION: -

again code corrected for offset > 8 upto 12

        for( i=l; l<=noq; l++ )
        {

          if( noctr >= ldev->size_of_reg )
                        noctr = ldev->size_of_reg - (int)oiq;


                if( i> ldev->no_of_reg-1 )
                {
                        curr= curr->next;
                        i=0;

                }

                not = copy_to_user( ubuff+nocsr, curr->data[i] +oiq , noctr );

......

}

Now, the code is running for all offsets from 1 to 12 for size_of_data =12.

nirmalpg124

RESULT :

writeDev: Please enter the string you want to write
HelloKernels
APP:writeDev: Wrote 12 bytes successfully

openDev : Please enter the file mode to open. (reading, writing, IOCTL)
analysing
openDev: ERROR :Invalid mode for opening the file.

readDev.c: Please enter the offset
0
readDev.c: pos: 0
readDev.c: New Position after lseek: 0
readDev: buff: 0x7fbba87d1e40
readDev: buff+pos: 0x7fbba87d1e40
APP:readDev: Read 12 bytes successfully as: HelloKernels

readDev.c: pos: 2
readDev.c: New Position after lseek: 2
readDev: buff: 0x7fbba87d1e40
readDev: buff+pos: 0x7fbba87d1e42
APP:readDev: Read 11 bytes successfully as: elloKernels

readDev.c: pos: 3
readDev.c: New Position after lseek: 3
readDev: buff: 0x7fcc40bbfe40
readDev: buff+pos: 0x7fcc40bbfe43
APP:readDev: Read 10 bytes successfully as: lloKernels

readDev.c: pos: 4
readDev.c: New Position after lseek: 4
readDev: buff: 0x7fcc40bbfe40
readDev: buff+pos: 0x7fcc40bbfe44
APP:readDev: Read 9 bytes successfully as: loKernels

readDev.c: pos: 7
readDev.c: New Position after lseek: 7
readDev: buff: 0x7fcc40bbfe40
readDev: buff+pos: 0x7fcc40bbfe47
APP:readDev: Read 6 bytes successfully as: ernels

nirmalpg124

TIMERS :

kenel :

read routine jiffies

[42321.128053] readDevice jiffies consumed on reading with j1 =4308423298, j2= 4308423298 are :0
[42321.128056] End :readDevice

time_t has been redefined as __kernel_old_time_t & susecond_t as __kernel_suseconds_t in <linux/time.h> kernel ver 6.4.2

[ 215.582656] readDevice jiffies consumed on reading with j1 =4294882780, j2= 4294882780 are :0
[ 215.582658] readDevice jiffies consumed on reading in microseconds us1 =4210451296, us2=4210451296 are: 0
[ 215.582659] readDevice jiffies consumed on reading in nanoseconds ns1 =4210451296000, ns2=4210451296000 are:0
[ 215.582661] readDevice no of cycles consumed on reading with nocyc1= 813335760528, nocyc2=813335879064 are : 118536

jiffies_to_timeval func is not available in my reinstalled ker ver in linux/jiffies.h

i used jiffies_to_usecs & jiffies_to_nsecs func

for converting jiffies to micro & nano seconds

mohdmohsinkhanee

To facilitate a smoother build process for the kernel, ensure the installation of the following packages:

dnf install ncurses-devel

dnf install pkg-config

dnf install openssl

dnf install openssl-devel

dnf install flex

dnf install flex bison

dnf install dwarves

dnf install zstd

dnf install -y elfutils-libelf-devel

These packages provide essential dependencies and tools necessary for an efficient kernel compilation.

sahirmanzoorbrz

[root@sahir linux-6.6.6]# make install

INSTALL /boot

grub2-editenv: error: invalid environment block.

Getting the above error

sahirmanzoorbrz

Tried four times kernal buildup now with careful consideration knowing my system limits as

In a Process of kernal buildup I have done these steps:

1.

Downloaded the new version of kernal 6.6.6 from this link:

wget https://mirrors.edge.kernel.org/pub/linux/kernel/v6.x/linux-6.6.6.tar.gz

2.

extracting the source code using:

tar -xvf linux-6.6.6.tar.gz

3.for compiling the new kernal I need to configure it so i used this command to have the old configuration working in my system .

cp /boot/config-5.14.18-100.fc33.x86_64 .config

4. for additional configuration I used:

make menuconfig

5. The following GUI Options in fedora I enabled:

went inside > 1.Device Drivers >character devices > enabled every missing module except serial drivers and saved it.

>Below the Device Drivers > File Systems > DOS/FAT/EXFAT/NT Filesystems ---> enabled every missing module in it and saved it.

6. Now compiling the kernal using:

make modules_install

7. after successfully installing the make_modules commmand

8.

make

7.After 24 hours I am able to see my kernal buildup failed with this error :

while Building my kernal using make I am getting this error during the link process

CALL scripts/checksyscalls.sh

DESCEND objtool

INSTALL libsubcmd_headers

DESCEND bpf/resolve_btfids

INSTALL libsubcmd_headers

CHK kernel/kheaders_data.tar.xz

LD vmlinux.o

Terminated

make[2]: *** [scripts/Makefile.vmlinux_o:62: vmlinux.o] Error 143

make[2]: *** Deleting file 'vmlinux.o'

make[1]: *** [/usr/src/kernels/linux-6.6.6/Makefile:1146: vmlinux_o] Error 2

make: *** [Makefile:234: __sub-make] Error 2

02.Building, Configuring kernel for Device Driver Development

Howdy, Stranger!

Categories