I have a file with binary data and I need to replace a few bytes in a certain position. I've come up with the following to direct bash to the offset and show me that it found the place I want:
dd bs=1 if=file iseek=24 conv=block cbs=2 | hexdump
Now, to use "file" as the output:
echo anInteger | dd bs=1 of=hextest.txt oseek=24 conv=block cbs=2
This seems to work just fine, I can review the changes made in a hex editor. Problem is, "anInteger" will be written as the ASCII representation of that integer (which makes sense) but I need to write the binary representation.
I want to use bash for this and the script should run on as many systems as possible (I don't know if the target system will have python or whatever installed).
How do I tell the command to convert the input to binary (possibly from a hex)?
You can use echo to emit specific bytes using hex or octal. For example:
echo -n -e \\x30
will print ascii 0 (0x30)
(-n remove trailing newline)
printf
is more portable than echo
. This function takes a decimal integer and outputs a byte with that value:
echobyte () {
if (( $1 >= 0 && $1 <= 255 ))
then
printf "\\x$(printf "%x" $1)"
else
printf "Invalid value\n" >&2
return 1
fi
}
$ echobyte 97
a
$ for i in {0..15}; do echobyte $i; done | hd
00000000 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f |................|
00000010
xxd
is the better way. xxd -r infile outfile
will take ascii hex-value in infile to patch outfile, and you can specify the specific position in infile by this: 1FE:55AA
Worked like a treat. I used the following code to replace 4 bytes at byte 24 in little endian with two integers (1032 and 1920). The code does not truncate the file.
echo -e \\x08\\x04\\x80\\x07 | dd of=<file> obs=1 oseek=24 conv=block,notrunc cbs=4
Thanks again.
If you're willing to rely on bc (which is fairly common)
echo -e "ibase=16\n obase=2 \n A1" | bc -q
might help.
You might put the desired input into a file and use the "if=" option to dd to insert exactly the input you desire.
I have a function to do this:
# number representation from 0 to 255 (one char long)
function chr() { printf "\\$(printf '%03o' "$1")" ; return 0 ; }
# from 0 to 65535 (two char long)
function word_litleendian() { chr $(($1 / 256)) ; chr $(($1 % 256)) ; return 0 ; }
function word_bigendian() { chr $(($1 % 256)) ; chr $(($1 / 256)) ; return 0 ; }
# from 0 to 4294967295 (four char long)
function dword_litleendian() { word_lilteendian $(($1 / 65536)) ; word_litleendian $(($1 % 65536)) ; return 0 ; }
function dword_bigendian() { word_bigendian $(($1 / 65536)) ; word_bigendian $(($1 % 65536)) ; return 0 ; }
You can use piping or redirection to catch the result.
With bash, "printf" has the "-v" option, and all shell has logical operators.
So here is simplier form in bash :
int2bin() {
local i=$1
local f
printf -v f '\\x%02x\\x%02x\\x%02x\\x%02x' $((i&255)) $((i >> 8 & 255)) $((i >> 16 & 255)) $((i >> 24 & 255))
printf "$f"
}
In my case, I needed to go from a decimal numeric argument to the actual unsigned 16-bit big endian value. This is probably not the most efficient way, but it works:
# $1 is whatever number (0 to 65535) the caller specifies
DECVAL=$1
HEXSTR=`printf "%04x" "$DECVAL"`
BYTEONE=`echo -n "$HEXSTR" | cut -c 1-2`
BYTETWO=`echo -n "$HEXSTR" | cut -c 3-4`
echo -ne "\x$BYTEONE\x$BYTETWO" | dd of="$FILENAME" bs=1 seek=$((0xdeadbeef)) conv=notrunc