Embedded-Intelligence

Learning from Linux Kernel

2010-04-19T08:21:00.000-07:00

1. container_of: Being passed a pointer, it returns the pointer to the structure that contains this.

#define container_of(ptr, type, member) ({                      \
        const typeof( ((type *)0)->member ) *__mptr = (ptr);    \
        (type *)( (char *)__mptr - offsetof(type,member) );})

where

#define offsetof(type, member) \
    ((size_t) ( (char *)&((type *)(0))->member - (char *)0 ))

My Favourite Quotes

2010-04-05T13:15:00.000-07:00

Pascal : "If I had more time, I would written you a shorter letter "

Einstein : "God doesn't play dice with the Universe"

Neils Bohr : "Prediction is very difficult, especially if it's about the future"

Profiling Tools : Valgrind & co

2010-03-11T06:16:00.000-08:00

VALGRIND
============
We will use valgrind with following tools :-

> memcheck [default]
> callgrind
> cachegrind
> massif
> helgrind

valgrind --tool=memcheck [prog-name] [prog-argumements]
a simple example is valgrind --tool=memcheck ls -al
===================================================
nemesis@nemesis-laptop:~$ valgrind --tool=memcheck ls -al
==3497== Memcheck, a memory error detector
==3497== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.
==3497== Using Valgrind-3.5.0-Debian and LibVEX; rerun with -h for copyright info
==3497== Command: ls -al
==3497==
total 13300
==3497==
==3497== HEAP SUMMARY:
==3497==     in use at exit: 14,671 bytes in 95 blocks
==3497==   total heap usage: 1,902 allocs, 1,807 frees, 150,632 bytes allocated
==3497==
==3497== LEAK SUMMARY:
==3497==    definitely lost: 200 bytes in 3 blocks
==3497==    indirectly lost: 240 bytes in 20 blocks
==3497==      possibly lost: 0 bytes in 0 blocks
==3497==    still reachable: 14,231 bytes in 72 blocks
==3497==         suppressed: 0 bytes in 0 blocks
==3497== Rerun with --leak-check=full to see details of leaked memory
==3497==
==3497== For counts of detected and suppressed errors, rerun with: -v
==3497== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 33 from 10)
===================================================
[NOTE]: In the above trace "3497" which precedes every line of Valgrind output is the pid of the program being run under valgrind i.e. ls -al in this case.

PRE-REQUISITES
==============

1. Strongly recommended to run the program with -g (with debug symbols) option. It is generally true of all profiling tools.

2. Use no optimization or as a compromise -0

Bitfields

2010-03-10T11:25:00.000-08:00

-- Used for saving space
-- NOT portable across platforms
-- bitfield declaration cannot use const and volatile qualifiers
-- Max bit field length is 64 bits. Using more than 32 bits is likely to be non portable

-- CANNOT take the address of a bitfield
-- CANNOT have a pointer to a bitfield
-- CANNOT define an array of bitfields

-- A bitfield of size 0 will force alignment to the nearest word boundary [depends actually on compiler behaviour -- some compilers may pad to the size of the base type...some to the word boundary]

-- A structure containing bitfields is suitably/appropriately padded.

Cool Compile time check

2010-03-09T12:11:00.000-08:00

If only we could evaluate a condition at compile time instead of link or run time. Well there is a way ...
Picked up this nifty trick from the linux kernel code.

#define BUILD_BUG_ON(condition) ((void)sizeof(char[1 - 2 *!!(condition)]))

Awesome piece of code ....It's a beauty...

Concept : Array of negative size is illegal. Remember array of size 0 is possible
Therefore if a condition evaluates to true we have :-
void(sizeof(NEGATIVE ARRAY) which would trigger a compile time error

If the condition is false, we have :-

void (sizeof(char[1 - 2 * !!(0))])
==> void(sizeof(char[1-2* !1])
====>void(sizeof(char[1-0])
======>void(sizeof(char[1])
========> void(const number) ====> This has no impact on C code although it is a valid expression.
you could have 1; anywhere in the code but to be "gcc warning free" it is better to have (void) 1 .

Here is a simple example, don't worry about the logic as I just needed a condition and do NOT mean to say that size of int should be smaller than that of a char :)

#define ADY_BUILD_BUG_ON(condition) ((void)sizeof(char[1 - 2 * !!(condition)]))

int main(int argc, char *argv [])
{
/* will not compile */
ADY_BUILD_BUG_ON(sizeof(char) < sizeof(int));
}

and here is what we get :--

nemesis@nemesis-laptop:~/test_code$ gcc build_bug_on_example.c
build_bug_on_example.c: In function ‘main’:
build_bug_on_example.c:9: error: size of array ‘type name’ is negative

Mod Utils

2010-03-09T11:29:00.000-08:00

Package name : module-init-tools
sudo apt-get install module-init-tools

Use :-
lsmod
modprobe
depmod
rmmod

Modprobe : Intelligent enough to figure out the dependencies.
[it is just a depmod -a combined with insmod of the necessary modules]

Default location of the modules :
/lib/modules/`uname -r`

Module dependency
File : /lib/modules/`uname -r`/modules.dep

e.g from the modules.dep file :-
-------------x--------------------------x---------------------------x--------------------------------

kernel/fs/freevxfs/freevxfs.ko:
kernel/fs/nfs/nfs.ko: kernel/fs/lockd/lockd.ko kernel/fs/nfs_common/nfs_acl.ko kernel/net/sunrpc/auth_gss/auth_rpcgss.ko kernel/net/sunrpc/sunrpc.ko
kernel/fs/exportfs/exportfs.ko:
kernel/fs/nfsd/nfsd.ko: kernel/fs/lockd/lockd.ko kernel/fs/nfs_common/nfs_acl.ko kernel/net/sunrpc/auth_gss/auth_rpcgss.ko kernel/net/sunrpc/sunrpc.ko kernel/fs/exportfs/exp
ortfs.ko
kernel/fs/lockd/lockd.ko: kernel/net/sunrpc/sunrpc.ko

-------------x--------------------------x---------------------------x--------------------------------

Therefore, nfs.ko requires nfs_acl.ko to be already installed. This dependency is build by depmod -a

qsort: Using array

2010-03-08T10:44:00.000-08:00

1. The realloc operation can be expensive if the the operation is carried out for a large number of times AND/OR if the memory is already fragmented.

2. A malloc operation could be considered but it would require a flag to indicate the first instance of executing the function

if ( FLAG == TRUE)
   Memory already allocated. Don't allocate any memory
else { /* Flag = FALSE */
   Allocate Memory
   Set Flag = TRUE
}

--------------------------------------x------------------x-------------------x-----------------------------
#define ARRAY_SIZE 10

typedef struct adi_data_t_ {
   unsigned int number;
} adi_data_t;

typedef struct adi_delete_list_t_ {
   adi_data_t ** first;
   unsigned int count;
} adi_delete_list_t;

int adi_compare_data (const void * a, const void *b)
{
   const adi_data_t ** one = (const adi_data_t **) a;
   const adi_data_t ** two = (const adi_data_t **) b;

   printf("Number one : [%d] Number two : [%d] ret [%d]\n", (*one)->number, (*two)->number, (*one)->number - (*two)->number);
   //return (*one)->number - (*two)->number ;
   return (*two)->number - (*one)->number ;

}

int adi_delete_walker(adi_delete_list_t * delete, adi_data_t * data)
{
   static adi_data_t* array[ARRAY_SIZE];
   void *temp = NULL;

   delete->first = array;
   memcpy(&array[delete->count], &data, sizeof(adi_data_t *));
   delete->count++;

}

int main (int argc, char* argv[])
{
   adi_delete_list_t delete;
   adi_data_t data_array[ARRAY_SIZE];
   unsigned int i = 0;

   memset(&delete, 0, sizeof(adi_delete_list_t));

   /* Fill the array and pass it to the delete function */
   for (i=0; i < ARRAY_SIZE ; i++) {
   data_array[i].number = rand()/1000000;
   printf(" Data Array filled with %d \n", data_array[i].number);
   adi_delete_walker(&delete, &data_array[i]);
   }

   printf("qsort parameters: first [%p] count [%d]", delete.first, delete.count);
   /* qsort the array */
   qsort(delete.first, delete.count, sizeof(adi_data_t *), adi_compare_data);

   for (i=0; i < ARRAY_SIZE ; i++) {
   printf(" POST SORT: Data Array filled with %d \n", (*(delete.first)[i]).number);
   }
}

qsort Example

2010-03-03T10:30:00.001-08:00

1. This example will demonstrate how to sort a given list in ascending or descending order

2. Many a times the structures are quite big and allocating such structures could be resource intensive exercise. Therefore I allocate a list of pointers to pointers (double pointer) to the original data structure.

3. QSORT sorts only arrays or data stored in contiguous memory locations. Therefore can't be directly used to sort linked lists.

Workaround : Walk the Linked List
   Store the pointers to these nodes in an array/allocated memory location
   Call Qsort to sort this array based on a comparison function that you define.

###include files : stdio.h -- stdlib.h -- string .h ####

-- This function will read the contents of a linked-list/array and copy into an "array" of pointers to the data. Qsort will sort this array.

#define ARRAY_SIZE 10

typedef struct adi_data_t_ {
   unsigned int number;
} adi_data_t;

typedef struct adi_delete_list_t_ {
   adi_data_t ** first;
   unsigned int count;
} adi_delete_list_t;

/* compare function to sort in ASCENDING order */
int adi_compare_data (const void * a, const void *b)
{
   const adi_data_t ** one = (const adi_data_t **) a;
   const adi_data_t ** two = (const adi_data_t **) b;

   printf("Number one : [%d] Number two : [%d] ret [%d]\n", (*one)->number, (*two)->number, (*one)->number - (*two)->number);
   return (*one)->number - (*two)->number ;

}

/* compare function to sort in DESCENDING order */

int adi_compare_data (const void * a, const void *b)

{

const adi_data_t ** one = (const adi_data_t **) a;

const adi_data_t ** two = (const adi_data_t **) b;

printf("Number one : [%d] Number two : [%d] ret [%d]\n", (*one)->number, (*two)->number, (*one)->number - (*two)->number);

return (*two)->number - (*one)->number ;

}

int adi_delete_walker(adi_delete_list_t * delete, adi_data_t * data)
{
   static adi_data_t ** list = NULL;
   void *temp = NULL;

   /* allocate memory for the list */
   temp = realloc(list, (delete->count + 1) * sizeof (adi_data_t *));
   if(temp == NULL) {
   printf("###!Ady: Hell Freezes over: Realloc failed \n");
   } else {
   list = (adi_data_t **) temp;
   delete->first = list;
   memcpy(&list[delete->count], &data, sizeof(adi_data_t *));
   printf(" First [%p] List [%p] Current element:[%p] \n", delete->first, list, &list[delete->count]);
   delete->count++;
   }
}

int main (int argc, char* argv[])

{

adi_delete_list_t delete;

adi_data_t data_array[10];

unsigned int i = 0;

memset(&delete, 0, sizeof(adi_delete_list_t));

/* Fill the array and pass it to the delete function */

for (i=0; i < 10 ; i++) {

data_array[i].number = rand()/1000000;

printf(" Data Array filled with %d \n", data_array[i].number);

adi_delete_walker(&delete, &data_array[i]);

}

printf("qsort parameters: first [%p] count [%d]", delete.first, delete.count);

/* qsort the array */

qsort(delete.first, delete.count, sizeof(adi_data_t *), adi_compare_data);

for (i=0; i < 10 ; i++) {

printf(" POST SORT: Data Array filled with %d \n", (*(delete.first)[i]).number);

}

free(delete.first);

}

Miscellaneous Debugging tips

2009-12-18T09:33:00.000-08:00

LTRACE
=========================================
from the Man page
ltrace is a program that simply runs the specified command until it exits. It intercepts and records the dynamic library calls which are called by the executed process and the signals which are received by that        process. It can also intercept and print the system calls executed by the program.

STRACE
===========================================
In the simplest case strace runs the specified command until it exits. It intercepts and records the system calls which are called by a process and the signals which are received by a process. The name of each system call, its arguments and its return value are printed on standard error or to the file specified with the -o    option.

nemesis@nemesis-laptop:~/test_code$ strace ./crashcourse
execve("./crashcourse", ["./crashcourse"], [/* 42 vars */]) = 0
brk(0)                                  = 0x80ba000
access("/etc/ld.so.nohwcap", F_OK)      = -1 ENOENT (No such file or directory)
mmap2(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7846000
access("/etc/ld.so.preload", R_OK)      = -1 ENOENT (No such file or directory)
open("/etc/ld.so.cache", O_RDONLY)      = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=72786, ...}) = 0
mmap2(NULL, 72786, PROT_READ, MAP_PRIVATE, 3, 0) = 0xb7834000
close(3)                                = 0
access("/etc/ld.so.nohwcap", F_OK)      = -1 ENOENT (No such file or directory)
open("/lib/tls/i686/cmov/libc.so.6", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\260l\1\0004\0\0\0"..., 512) = 512
fstat64(3, {st_mode=S_IFREG|0755, st_size=1319364, ...}) = 0
mmap2(NULL, 1325416, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x626000
mmap2(0x764000, 12288, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x13e) = 0x764000
mmap2(0x767000, 10600, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x767000
close(3)                                = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7833000
set_thread_area({entry_number:-1 -> 6, base_addr:0xb78336c0, limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, seg_not_present:0, useable:1}) = 0
mprotect(0x764000, 8192, PROT_READ)     = 0
mprotect(0x8049000, 4096, PROT_READ)    = 0
mprotect(0x3d0000, 4096, PROT_READ)     = 0
munmap(0xb7834000, 72786)               = 0
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV +++

MTRACE
=============================================
nemesis@nemesis-laptop:~/test_code$ mtrace ./crashcourse
No memory leaks

LibC

2009-12-18T08:45:00.001-08:00

nemesis@nemesis-laptop:/lib$ ./libc.so.6
GNU C Library (EGLIBC) stable release version 2.10.1, by Roland McGrath et al.
Copyright (C) 2009 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.
Compiled by GNU CC version 4.4.1.
Compiled on a Linux >>2.6.24-23-server<< system on 2009-10-07.
Available extensions:
        crypt add-on version 2.1 by Michael Glad and others
        GNU Libidn by Simon Josefsson
        Native POSIX Threads Library by Ulrich Drepper et al
        BIND-8.2.3-T5B
For bug reporting instructions, please see:
.

Core Dump : how to enable & Debug

2009-12-18T08:32:00.000-08:00

1. find out if core will be dumped.

nemesis@nemesis-laptop:~/test_code$ ulimit -c
0

This means that core will NOT be dumped. ulimit specifies the maximum size of the core file.

2. Now set the value of ulimit to whatever you want it to be

ulimit -c [size]

3. you can also specify the directory in which core files will be placed
root@nemesis-laptop:/home/nemesis/test_code# echo "/tmp/corefiles/core" > /proc/sys/kernel/core_pattern

4. small program to cause the segmentation fault

###########################################
#include

void func2(void)
{
    int *p = NULL;
    *p = 0xdeadcafe;
}

void func1(void)
{
    func2();
}

int main (int argc, char **argv)
{
    func1();
    int *p = NULL;
    *p = 0xdeadcafe;

}

##############################################

nemesis@nemesis-laptop:~/test_code$ gcc -g -o crashcourse crash_core.c

5. now generate the core file
nemesis@nemesis-laptop:~/test_code$ ./corefile
Segmentation fault (core dumped)

6. a bit more details from the file command [Tells you the program which generated the core dump]

nemesis@nemesis-laptop:~/test_code$ file /tmp/corefiles/core
/tmp/corefiles/core: ELF 32-bit LSB core file Intel 80386, version 1 (SYSV), SVR4-style, from './corefile'

7. gdb [executable_file] [core_file]

##################################################
nemesis@nemesis-laptop:~/test_code$ gdb crashcourse /tmp/corefiles/core
GNU gdb (GDB) 7.0-ubuntu
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i486-linux-gnu".
For bug reporting instructions, please see:
...
Reading symbols from /home/nemesis/test_code/crashcourse...done.

warning: Can't read pathname for load map: Input/output error.
Reading symbols from /lib/tls/i686/cmov/libc.so.6...(no debugging symbols found)...done.
Loaded symbols for /lib/tls/i686/cmov/libc.so.6
Reading symbols from /lib/ld-linux.so.2...(no debugging symbols found)...done.
Loaded symbols for /lib/ld-linux.so.2
Core was generated by `./crashcourse'.
Program terminated with signal 11, Segmentation fault.
#0 0x080483c4 in func2 () at crash_core.c:6
6           *p = 0xdeadcafe;

##################################################

8. Get the backtrace

(gdb) bt
#0 0x080483c4 in func2 () at crash_core.c:6
#1 0x080483d4 in func1 () at crash_core.c:11
#2 0x080483e1 in main (argc=1, argv=0xbfe81b44) at crash_core.c:17

9. use the "up" and "down" commands to go through the code. (you can't run the code remember :) )

==============================================
UP
==============================================
(gdb) up
#1 0x080483d4 in func1 () at crash_core.c:11
11          func2();
(gdb) up
#2 0x080483e1 in main (argc=1, argv=0xbfe81b44) at crash_core.c:17
17          func1();
(gdb) up
Initial frame selected; you cannot go up.

===============================================
DOWN
===============================================
(gdb) down
#1 0x080483d4 in func1 () at crash_core.c:11
11          func2();
(gdb) down
#0 0x080483c4 in func2 () at crash_core.c:6
6           *p = 0xdeadcafe;

=========================================
NOTE: What if somebody handed you a core file and you don't the crashing process ? Happens a lot of times when the test team would pass you the core file and wash their hands off the issue altogether :)

Answer:
1. from the (file) following command

file [CORE-FILE]

nemesis@nemesis-laptop:~/test_code$ file /tmp/corefiles/core

/tmp/corefiles/core: ELF 32-bit LSB core file Intel 80386, version 1 (SYSV), SVR4-style, from './corefile'

2. just load the core file in gdb with "any" program name.

nemesis@nemesis-laptop:~/test_code$ gdb /bin/ps /tmp/corefiles/core
GNU gdb (GDB) 7.0-ubuntu
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i486-linux-gnu".
For bug reporting instructions, please see:
...
Reading symbols from /bin/ps...(no debugging symbols found)...done.

warning: core file may not match specified executable file.
Core was generated by `./crashcourse'.
Program terminated with signal 11, Segmentation fault.
#0 0x080483c4 in ?? ()

GDB is intelligent enough to catch your bluff but the good that comes out of it is that it tells you the correct name of the program that core-d. (GDB can't tell if your bluffing or plain ignorant :))

GIT Guide --- update

2009-12-13T13:00:00.000-08:00

Before your commits make sure you have these variables defined. :)
==================================================
nemesis@nemesis-laptop:~/qemu/qemu$ git config --global user.name "nemesisofstate"
nemesis@nemesis-laptop:~/qemu/qemu$ git config --global user.email "nemesisofstate@gmail.com"

for checkin

========

edit [file]

git add [file]

git commit -m "your-message-here" -s

-s

--signoff

Add Signed-off-by line by the committer at the end of the commit log message.

======================

see your patch

git format-patch origin/master

your messages

===============

git show

gitk ===> graphic tool for git

Send out your patch

================

I use gmail to send my patches ...so u need msmtp for it [http://msmtp.sourceforge.net]. Install the file and configure your ~/.msmtprc file

git-config --global sendemail.smtpserver /usr/local/bin/msmtp

sudo apt-get install git-email

git email-send

e.g

nemesis@nemesis-laptop:~/qemu/qemu$ git send-email 0001-correcting-ARM-CPSR-register-bit-position-comment.patch

HOWTO: Get specific linux kernel headers

2009-12-13T08:04:00.000-08:00

On UBUNTU

sudo apt-get install linux-headers-$(uname -r)

QEMU + U-BOOT

2009-12-12T14:32:00.000-08:00

1. Get the U-Boot sources

2. sudo apt-get install qemu-kvm-extras [for qemu-system-arm]

We will compile for ARM (needless to say) target---ARM versatile board (needless again)

[NOTE]: [u-boot-dir]

/include/configs contains the default configs for most of the platforms. It almost corresponds to the one in linux kernel sources.

STEPS
=======

1. make versatile_config
2. make ARCH=arm CROSS_COMPILE=arm-none-linux-gnueabi-
make ....make....make...

[NOTE]: Uboot needs the linux with uboot specific headers in order to boot it. This u-boot specific image can be generated using "make uImage".

However this would need mkimage. Where do we get it from---

Easy:)

---- after running make you can find the mkimage in [u-boot-dir]/tools directory.

---- Or get it from sudo apt-get install uboot-mkimage

3. Launch
sudo qemu-system-arm -nographic -m 256 -M versatileab -kernel u-boot.bin

The above command is good if your testing u-boot as a standalone app. However it would be wonderful to have networking between host and target to boot a linux kernel

sudo qemu-system-arm -M versatileab -kernel u-boot -net nic -net tap -nographic -serial stdio

Don't worry about the error message we will get to it.
===================================================
nemesis@nemesis-laptop:~/u-boot-2009.08$ sudo qemu-system-arm -nographic -m 256 -M versatileab -kernel u-boot.bin

U-Boot 2009.08 (Dec 13 2009 - 00:28:50)

DRAM: 0 kB
## Unknown FLASH on Bank 1 - Size = 0x00000000 = 0 MB
Flash: 0 kB
*** Warning - bad CRC, using default environment

In: serial
Out: serial
Err: serial

===================================================

Build linux kernel Image for U-boot

nemesis@nemesis-laptop:~/linux_kernel/linux-2.6.32$ make uImage
CHK include/linux/version.h
make[1]: `include/asm-arm/mach-types.h' is up to date.
CHK include/linux/utsrelease.h
SYMLINK include/asm -> include/asm-arm
CALL scripts/checksyscalls.sh
MKELF scripts/mod/elfconfig.h
HOSTCC scripts/mod/file2alias.o
HOSTCC scripts/mod/modpost.o
scripts/mod/modpost.c: In function ‘get_markers’:
scripts/mod/modpost.c:1562: warning: ignoring return value of ‘asprintf’, declared with attribute warn_unused_result
scripts/mod/modpost.c: In function ‘add_marker’:
scripts/mod/modpost.c:1982: warning: ignoring return value of ‘asprintf’, declared with attribute warn_unused_result
HOSTCC scripts/mod/sumversion.o
HOSTLD scripts/mod/modpost
CHK include/linux/compile.h
GEN lib/crc32table.h
CC lib/crc32.o
LD lib/built-in.o
LD vmlinux.o
MODPOST vmlinux.o
GEN .version
CHK include/linux/compile.h
UPD include/linux/compile.h
CC init/version.o
LD init/built-in.o
LD .tmp_vmlinux1
KSYM .tmp_kallsyms1.S
AS .tmp_kallsyms1.o
LD .tmp_vmlinux2
KSYM .tmp_kallsyms2.S
AS .tmp_kallsyms2.o
LD vmlinux
SYSMAP System.map
SYSMAP .tmp_System.map
OBJCOPY arch/arm/boot/Image
Kernel: arch/arm/boot/Image is ready
GZIP arch/arm/boot/compressed/piggy.gz
AS arch/arm/boot/compressed/piggy.o
LD arch/arm/boot/compressed/vmlinux
OBJCOPY arch/arm/boot/zImage
Kernel: arch/arm/boot/zImage is ready
UIMAGE arch/arm/boot/uImage
Image Name: Linux-2.6.32
Created: Sun Dec 13 02:29:45 2009
Image Type: ARM Linux Kernel Image (uncompressed)
Data Size: 1535396 Bytes = 1499.41 kB = 1.46 MB
Load Address: 00008000
Entry Point: 00008000
Image arch/arm/boot/uImage is ready

GET READY TO BOOT YOUR LINUX KERNEL
======================================

VersatilePB # setenv ipaddr 10.0.2.16
VersatilePB # setenv serverip 10.0.2.15
VersatilePB # ping 10.0.2.15
SMC91111: PHY auto-negotiate timed out
Using MAC Address 52:54:00:12:34:56
host 10.0.2.15 is alive
VersatilePB # tftp 8000 uImage
SMC91111: PHY auto-negotiate timed out
Using MAC Address 52:54:00:12:34:56
TFTP from server 10.0.2.15; our IP address is 10.0.2.16
Filename 'uImage'.
Load address: 0x8000
Loading: T T T T T T T #################################################################
###T #####################################
done
Bytes transferred = 1535460 (176de4 hex)

PROBLEM: There seems to be a bug in the sense that DRAM setting are not correct for versatile.

Switching to Integrator platform. [build uboot and linux image for integrator]
(uboot config =====> integratorcp_config)

nemesis@nemesis-laptop:~/u-boot-2009.08$ sudo qemu-system-arm -kernel u-boot -net nic -net tap -nographic -serial stdio
[sudo] password for nemesis:

U-Boot 2009.08 (Dec 13 2009 - 04:00:17)

DRAM: 128 MB
## Unknown FLASH on Bank 1 - Size = 0x00000000 = 0 MB
Flash: 0 kB
*** Warning - bad CRC, using default environment

In: serial
Out: serial
Err: serial
Hit any key to stop autoboot: 0
Integrator-CP # setenv serverip 10.0.2.15
Integrator-CP # setenv ipaddr 10.0.2.16
Integrator-CP # tftp 8000 uImage
SMC91111: PHY auto-negotiate timed out
Using MAC Address 52:54:00:12:34:56
TFTP from server 10.0.2.15; our IP address is 10.0.2.16
Filename 'uImage'.
Load address: 0x8000
Loading: T T #################################################################
################################
done
Bytes transferred = 1411276 (1588cc hex)
Integrator-CP # bootm
## Booting kernel from Legacy Image at 00008000 ...
Image Name: Linux-2.6.32
Image Type: ARM Linux Kernel Image (uncompressed)
Data Size: 1411212 Bytes = 1.3 MB
Load Address: 00008000
Entry Point: 00008000
XIP Kernel Image ... OK
OK

Starting kernel ...

qemu: fatal: Trying to execute code outside RAM or ROM at 0xfffff1c2

R00=00000000 R01=00000107 R02=00000100 R03=00000002
R04=00fddfc0 R05=00000000 R06=00008000 R07=00000113
R08=00fddfdc R09=010172c0 R10=010176ac R11=00fddced
R12=0100f750 R13=00000000 R14=00008008 R15=fffff1c2
PSR=600001db -ZC- A und32
Aborted
nemesis@nemesis-laptop:~/u-boot-2009.08$

============================================

So far so good...Uboot passes control to linux kernel but the kernel boot does a crash and burn. Will need to solve this one now....

Busybox

2009-12-06T02:33:00.000-08:00

1. Download the source and patches from the busybox website.

http://www.busybox.net/download.html

The version at the time of writing this is 1.14.1.

2. untar the sources to the directory where you will run the build.

3. patch the sources

patch -p0 <>.patch
e.g patch -p0 <>

4. config the busybox using your favourite config method. I prefer to use xconfig.

5. make your day :) -------- run make

6. once the make has run successful , you would want to install the busybox in a seperate directory which you can later ready to use as the rootfs.

make CONFIG_PREFIX=~/busybox/target install

where I am installing busybox in the /home/nemesis/busybox/target directory

7 cd to the install directory

cd ~/busybox/target

8. There isn't a great deal there

ls -al would return

sbin, bin , usr and linuxrc (a symbolic link to /bin/busybox)

9. Create the dev, dev/pts, etc, etc/init.d, lib, mnt, opt, proc, root, sys, tmp, var, and var/log directories. Also create the device node for the initial console

mkdir dev etc lib mnt opt proc root sys tmp var debug
mkdir etc/init.d
mkdir var/log

sudo mknod dev/console c 5 1
sudo mknod dev/null c 1 3

10. Mount proc and /dev/pts at boot time
===========================

edit /etc/fstab for these filesystems

11. Shared library and library loader
==========================
We still need the
---------- shared libraries (from your toolchain)
---------- library loader (e.g ld-linux.so.3 in the listing below)

nemesis@nemesis-laptop:~/codesourcery/arm-2009q3/arm-none-linux-gnueabi/libc/lib$ ls
ld-2.10.1.so libdl-2.10.1.so libnss_dns.so.2 libresolv-2.10.1.so
ld-linux.so.3 libdl.so.2 libnss_files-2.10.1.so libresolv.so.2
libanl-2.10.1.so libgcc_s.so libnss_files.so.2 librt-2.10.1.so
libanl.so.1 libgcc_s.so.1 libnss_hesiod-2.10.1.so librt.so.1
libBrokenLocale-2.10.1.so libm-2.10.1.so libnss_hesiod.so.2 libSegFault.so
libBrokenLocale.so.1 libmemusage.so libnss_nis-2.10.1.so libthread_db-1.0.so
libc-2.10.1.so libm.so.6 libnss_nisplus-2.10.1.so libthread_db.so.1
libcidn-2.10.1.so libnsl-2.10.1.so libnss_nisplus.so.2 libutil-2.10.1.so
libcidn.so.1 libnsl.so.1 libnss_nis.so.2 libutil.so.1
libcrypt-2.10.1.so libnss_compat-2.10.1.so libpcprofile.so
libcrypt.so.1 libnss_compat.so.2 libpthread-2.10.1.so
libc.so.6 libnss_dns-2.10.1.so libpthread.so.0

--- that's an aweful lot. Copying them indiscriminately will cause our rootfs to be bloated (something we don't want). So just pick what you need. And And And ofcourse strip these libraries :)

nemesis@nemesis-laptop:~/busybox/target/lib$ ~/codesourcery/arm-2009q3/bin/arm-none-linux-gnueabi-strip *

Thats the minimal set for the rootfs

12. Kernel Modules
=================
Now it is time to get our dear modules.

--- go to the kernel build directory
--- issue the following command : make modules

after it has completed successfully , we need to install the modules in our rootfs. Elementary my dear watson!

make INSTALL_MOD_PATH=/path/to/rootfs modules_install

Once we are out of the debugging phase, we would like to strip the modules to slim down.

cd /path/to/rootfs/lib/modules
arm-none-linux-gnueabi-strip `find . –name “*.ko”`

Root File System

2009-12-06T00:06:00.000-08:00

Umpteen ways to get this done :-

1. NFS Mounted. [One of the most popular methods ]

2. Initial RAMDISK (Initrd) ====> Will cover this in more detail in a later post

3. JFFS/JFFS2

4. CRAMFS (Read-only)

We will use busybox -- the veritable "Swiss Army Knife" for embedded systems.

Each of these will be explained in seperate posts, starting with Busybox

ARM on QEMU

2009-12-05T11:53:00.000-08:00

Goal : Run linux on a verstaline board (simulated by QEMU)

why : Because it is not always possible to have the real board available for development.

Prerequisites:
==============

1. QEMU : ofcourse
on debian systems : sudo apt-get install qemu
on RH systems : sudo yum install qemu

2. This one depends on what kind of communication mechanism is used between the host (your development) system and the guest sytem on QEMU. I prefer to use networking.

---- Have DHCP running so that guest system on qemu can use it to get an IP address.
---- host TAP devices used to connect to QEMU. The kernel config of the host should have TAP/TUN networking enabled. The switch for this is CONFIG_TUN.

grep CONFIG_TUN= /boot/config-`uname -r`

This should return you the state of the switch.

nemesis@nemesis-laptop:~$ grep CONFIG_TUN= /boot/config-`uname -r`
CONFIG_TUN=m

nemesis@nemesis-laptop:~$ ls -al /dev/net/tun
crw-rw-rw- 1 root root 10, 200 2009-11-03 11:46 /dev/net/tun

This verifies that the device exists indeed. If not, create it manually.

mknod /dev/net/tun c 10 200

Configure bridge-control
==================
Add the following to a script file

#!/bin/sh
/usr/sbin/brctl addbr br0 ### Add a bridge
/sbin/ifconfig eth0 0.0.0.0 promisc up ### set up eth0
/usr/sbin/brctl addif br0 eth0 ### add an intf(eth0) to the bridge (br0)
/sbin/dhclient br0 ### enable dhcp
/sbin/iptables -F FORWARD

Build kernel
=========

The quickest way is to use the default config file for the versatile platform.
Simply copy the config file from configs directory for ARM arch and rename it as .config

cp /arch/arm/configs/versatile_defconfig /.config

Config the Linux kernel (xconfig, menuconfig or simply using vim to edit..have it your way)

1. Enable ARM EABI support : switch CONFIG_AEABI (part of kernel features)
2. Enable DHCP : Switch CONFIG_IP_PNP_DHCP
3. Enable Universal TUN/TAP driver : switch CONFIG_TUN
4. Enable tmpfs : Switch CONFIG_TMPFS

Now we are ready to make (my day)
Ofcourse ARCH = arm
and CROSS_COMPILE = your_cross_compiler (I use codesourcery and for me it is
arm-none-linux-gnueabi-)

What about the Root File System ?
=======================

We will NFS mount the root filesystem. Build your rootfs.

1. edit /etc/exports ===> Add the following line

/mnt/arm-linux-rootfs / *(rw,sync,no_root_squash)

Here the untarred rootfs is at /mnt/arm-linux-rootfs and it is accessible as rw filesystem from any systems (indicated by *). If static ip-addresses are assigned, use the ip-address of the peer here.

2. restart the nfs service so that our changes come into effect.

/etc/init.d/nfs restart

Boot into Linux on QEMU
======================

qemu-system-arm -M versatilepb -kernel zImage-versatile -append root="/dev/nfs nfsroot=:/mnt/arm-linux-rootfs rw ip=dhcp" \
-net nic,vlan=0 -net tap,vlan=0,ifname=tap0,script=./qemu-ifup

qemu-ifup
=============================

#!/bin/sh

echo "Executing /etc/qemu-ifup"
echo "Bringing up $1 for bridged mode..."
sudo /sbin/ifconfig $1 0.0.0.0 promisc up
echo "Adding $1 to br0..."
sudo /usr/sbin/brctl addif br0 $1
sleep 2

Source : http://fedoraproject.org/wiki/Architectures/ARM/HowToQemu
http://www.aurel32.net/info/debian_arm_qemu.php

Cross Toolchains : PART2 (Kernel Headers)

2009-11-29T04:24:00.000-08:00

1. Download the kernel sources from kernel.org

2. cd to the sources root directory

3. run the following command to extract the headers for ARM

make headers_install ARCH=arm INSTALL_HDR_PATH=/usr/include/arm-linux-headers

Where ARCH specifies the architecture for which the headers are to be generated. INSTALL_HDR_PATH is the path on the local system where the header files will be placed. The default is /usr/include

To see a list of all supported architectures

ls -d include/asm-* | sed 's/.*-//'

make headers_install_all exports headers for all architectures simultaneously. Not very useful for the average embedded programmer.

#Decribe the API for the user space programs wishing to use kernel services
#Used by system's C library (glibc or uClibc) to define available system calls as well as constants and structures to be used with these system calls.
#Kernel headers are backward compatible.

[Sources : 1. Documentation/make/headers_install.txt ]

nemesis@nemesis-laptop:~/arm-linux-kernel/linux-2.6.31.6$ sudo make headers_install ARCH=arm INSTALL_HDR_PATH=/usr/include/arm-linux
CHK     include/linux/version.h                                                                                                   
make[1]: `scripts/unifdef' is up to date.                                                                                           
INSTALL include/asm-generic (34 files)                                                                                            
INSTALL include/drm (12 files)                                                                                                    
INSTALL include/linux/byteorder (2 files)                                                                                         
INSTALL include/linux/can (4 files)                                                                                               
INSTALL include/linux/dvb (8 files)                                                                                               
INSTALL include/linux/hdlc (1 file)                                                                                               
INSTALL include/linux/isdn (1 file)                                                                                               
INSTALL include/linux/netfilter (58 files)                                                                                        
INSTALL include/linux/netfilter_arp (2 files)                                                                                     
INSTALL include/linux/netfilter_bridge (16 files)                                                                                 
INSTALL include/linux/netfilter_ipv4 (46 files)                                                                                   
INSTALL include/linux/netfilter_ipv6 (21 files)                                                                                   
INSTALL include/linux/nfsd (6 files)                                                                                              
INSTALL include/linux/raid (2 files)                                                                                              
INSTALL include/linux/spi (1 file)                                                                                                
INSTALL include/linux/sunrpc (1 file)                                                                                             
INSTALL include/linux/tc_act (6 files)                                                                                            
INSTALL include/linux/tc_ematch (4 files)                                                                                         
INSTALL include/linux/usb (8 files)                                                                                               
INSTALL include/linux/wimax (1 file)                                                                                              
INSTALL include/linux (352 files)                                                                                                 
INSTALL include/mtd (5 files)                                                                                                     
INSTALL include/rdma (1 file)                                                                                                     
INSTALL include/scsi (4 files)                                                                                                    
INSTALL include/sound (9 files)                                                                                                   
INSTALL include/video (3 files)                                                                                                   
INSTALL include/xen (1 file)                                                                                                      
INSTALL include (0 file)                                                                                                          
INSTALL include/asm (32 files)

Cross Toolchains : PART 1 (Binutils)

2009-11-29T03:02:00.001-08:00

Components of crosstoolchain
===================

1. Binutils
2. Kernel header
3. C library
4. Compiler [e.g gcc]

optional : gdb debugger etc

Binutils
=====
The target architecture needs to have its own set of tools to generate binary files for it. Also it would be nice to have some tools to enable examining these.

ld ---> linker

strip --> to strip the object file of a particular section. e.g strip the comments to reduce the code size.

      strip [ -F BFDNAME | --target=BFDNAME ]
      [ -I BFDNAME | --input-target=BFDNAME ]
      [ -O BFDNAME | --output-target=BFDNAME ]
      [ -s | --strip-all ] [ -S | -g | --strip-debug ]
      [ -K SYMBOLNAME | --keep-symbol=SYMBOLNAME ]
      [ -N SYMBOLNAME | --strip-symbol=SYMBOLNAME ]
      [ -x | --discard-all ] [ -X | --discard-locals ]
      [ -R SECTIONNAME | --remove-section=SECTIONNAME ]
      [ -o FILE ] [ -p | --preserve-dates ]
      [ -v | --verbose ]  [ -V | --version ]  [ --help ]
      OBJFILE...

e.g strip --remove section=.comment

as --> assembler
objdump
objcopy
readelf: very useful tool. I invariably use it with -a switch to retrieve the max info.

nm : lists all the symbols in an object file.

===========
Extract from IBM's website

objdump & nm

The commands objdump and nm both display information about object files. If a crash occurs and a corefile is produced, these commands help you analyze the file.

objdump: Use this command to disassemble shared objects and libraries. After you have discovered which library or object has caused the problem, use objdump to locate the method in which the problem originates. To invoke objdump, type: objdump
nm: This command lists symbol names from object files. These symbol names can be either functions, global variables, or static variables. For each symbol, the value, symbol type, and symbol name are displayed. Lower case symbol types mean the symbol is local, while upper case means the symbol is global or external. To use this tool, type: nm

You can see a complete list of options by typing objdump -H. The -d option disassembles contents of executable sections

Run these commands on the same machine as the one that produced the core files to get the most accurate symbolic information available. This output (together with the core file, if small enough) is used by the IBM^® support team for Java™ to diagnose a problem.

=================================

addr2line : very important for debugging. E.g lets say you get a core dump with a backtrace. You can use the backtrace info to go to the code directly

void test_funk()
{
printf("#NEMESIS: testing funky func \n");
printf("#NEMESIS: address of test_funk [%p]", &test_funk);
}

Output: in the output the addess printed is 0x845c. Pass this as an input to the addr2line.]

nemesis@nemesis-laptop:~/test_code$ arm-none-linux-gnueabi-addr2line 0x845c
/home/adikumar/test_code/test_3.c:5 ===> Lo and behold.You have the line in the code.

strings: Lists all string in an object file. use it with grep to find a particular string

nemesis@nemesis-laptop:~/codesourcery/arm-2009q3/bin$ arm-none-linux-gnueabi-strings test2
/lib/ld-linux.so.3
libgcc_s.so.1
__aeabi_unwind_cpp_pr0
__gmon_start__
_Jv_RegisterClasses
__aeabi_unwind_cpp_pr1
libc.so.6
abort
printf
__libc_start_main
GCC_3.5
GLIBC_2.4
###!Nemesis: LittleEndian =========> my prints
###!Nemesis: BigEndian =========> my prints
0x%x

Whats your Endian-ness ?

2009-11-25T03:50:00.000-08:00

int i= 0x1;

*((char*)&i) ? printf("###!Nemesis: LittleEndian\n"):printf("###!Nemesis: BigEndian\n");

printf("0x%x \n", *((char*)&i));

Forthcoming Embedded Articles

2009-11-18T06:43:00.000-08:00

1. Building cross compiled linux kernel [With debug support]. Explain various configuration options.

2. Building a customized root filesystem

3. initial Ram disk

4. Uboot Bootloader: Internals, porting

5. Linux kernel start code walk through [ARM perspective]

6. Busybox

7. Using emulators [Qemu]

8. Linux device driver Serial : ARM Primecell UART

9. Linux Network driver

10. Linux Serial Driver

11. Gnu Tool chain [ gdb, gcc etc ]. binutils Making your own toolchain

12. ucLinux , ucLibc

13. NFS mounting

14. Remote debugging with gdbserver

15. Integrating Eclipse into your embedded development system

16. Misc tools : valgrind, profiling tools etc

17. kernel boot parameters. Explain nfs boot [ ip assigned by DHCP]

18. Various kernel image formats...zImage, uImage, vmlinux etc [ explain the linking and image components]

19. uClibC vs glibc

20. different filesystems (esp squashfs )

21. scratchbox, Buildroot

Difference between "su" and "su -"

2009-07-08T02:29:00.000-07:00

A lot of people don't know the difference [and there is a day light of difference :) ]

==================

1. ENVIRONMENT : The environment is the same as that of the user's from which you entered root. In the following example, the environment would be the same as that of user nemesis.

e.g nemesis@tantra:~/device-drivers$su

================================

su -

================================

1. shell reads a relevant startup files and simulates a login

2. ENVIRONMENT: The enviroment is that of the superuser/root and NOT that of the account from you you invoked "su".

For more detailed information, please refer to the man pages.

Strings: Single Quote Vs Double Quotes

2009-07-06T01:55:00.001-07:00

In UNIX Shell

============

Double Quotes : In (Most ) UNIX shells and other scripting languages like perl, the interpreter would parse through the string inorder to perform substitutions for any variables in the string.

Single Quotes: Interpreter leaves the string alone. It is treated as a string literal.

In C

===========

Double Quotes: For strings

Single Quotes : For single character literals.

e.g

char c[];

c = 'c'; ==> SINGLE Quotes

c = "char"; ==> DOUBLE quotes

Minimal GIT guide

2009-01-13T01:57:00.000-08:00

1. Create a new branch on the remote git repository

git push origin origin:refs/heads/

e.g git push origin origin:/refs/heads/nemesis-debug

creates a branch called nemesis-debug on the remote repository.

Confirm this by typing git branch -r. Your new respository should appear there.

2. create a local branch to track the newly created remote branch.

git checkout --track -b

e.g git checkout --track -b nemesis-debug origin/nemesis-debug

3. make all necessary changes.

4. commit your changes

a.add files/directories you want to commit

git add

git add . # adds all the files in the directory recursively

b. commit

git commit -m "commit-message"

c. push all your changes to the repository

git push origin

e.g git push origin nemesis-debug

5. In case you screwed up and want to delete the remote branch

git push origin :heads/

e.g git push origin :heads/nemesis-debug

Memory Optimizations: Kernel Programming

2009-01-06T00:52:00.000-08:00

UNLIKE User-space code/data, kernel code/data reside PERMANENTLY in memory. They are NOT swapped Out.

>>> module_init : RUN Time Optimization. Executed when the associated module is loaded. If the module is statically included in the kernel, the kernel can free the module_init routine right at the boot time, after it runs.

>>> module_exit: LINK Time Optimization. It is never executed if the modules are statically included in the kernel. In such cases, it is not needed to include module_exit routines into the kernel image and are discarded at Link Time.