$RANDOM in linux bash will alway be the same upon bootup

Question

I ran into this problem and I'm so confused.

I am using $RANDOM in bash as a test string at the very beginning of the linux bootup process. Very interestingly, I observed that $RANDOM will be the same every time linux booted up. In my case, it's 18869.

I was simply echoing $RANDOM to a file. I observed the same number in every boot up. After that, $RANDOM seems to return "real" random numbers.

echo "$RANDOM is a test string" >> /tmp/test

Can anyone explain the reason in this?

Look inside `bash` source code. You'll probably find out that `$RANDOM` is some *pseudo* random number generator (so in fact it is deterministic)! — Basile Starynkevitch, Aug 29 '13 at 21:36

score 5 · Answer 1 · edited Oct 23 '20 at 16:33

Here is the initial seed algorithm from variables.c in the Bash 4.2 source code:

static void                                                                                               
seedrand ()                                                                                               
{                                                                                                         
  struct timeval tv;                                                                                      
                                                                                                          
  gettimeofday (&tv, NULL);                                                                               
  sbrand (tv.tv_sec ^ tv.tv_usec ^ getpid ());                                                            
}

In other words, it generates the seed based on pid and current time in seconds and microseconds (with a system accuracy of ~1/100th second).

If you check $RANDOM on a small, predictable system, before the system clock is initialized, you're likely to see the same value every time.

Thanks for the information. In this scenario where the system clock is not initialized on an embedded system, what do you suggest to do to generate a random number? — iglory, Aug 29 '13 at 21:47
Is this embedded system managing /dev/urandom like larger linux distros do? If so, you could ensure your script runs just after /dev/urandom is initialized, and set RANDOM from a few bytes of that data. — sjnarv, Aug 29 '13 at 21:57
Some embedded systems like Raspberry Pi has a hardware number generator device you could use. You can also write a file periodically or on shutdown that writes `date +%s >> /tmp/seed` and then `RANDOM=$( — that other guy, Aug 29 '13 at 22:29

score 0 · Answer 2 · answered Aug 29 '13 at 21:46

Because the seed is always the same at the startup. You can always assign a value

RANDOM=1000

to seed the RANDOM manually, perhaps with the PID

RANDOM=$$

Trying to find your seed, I wrote this small script

for ((i=-150000; i<150000 ;i=i+1)); do
  RANDOM=$i;
  echo "$i: $RANDOM $RANDOM $RANDOM";
done | grep "18869[^:]"

and found:

-118657: 8608 18869 19673
-108446: 18869 18855 15976
-78471: 694 18869 19296
-75678: 18869 23165 7967
-42910: 18869 27476 32727
-38285: 25548 18869 18919
-10142: 18869 31786 24719
27971: 18869 6258 29786
33149: 13707 18869 8396
60739: 18869 10568 21777
73335: 5793 18869 8019
93507: 18869 14878 13769
113521: 30647 18869 7642
126275: 18869 19188 5761
137004: 18869 3793 18604

See that this script has lots of $RANDOM but still is fully deterministic, it'll result always the same if you call it. This link has more information about $RANDOM and this other link has information on how to use /dev/random and /dev/urandom, as an alternative. Another way to generate a random number without the seed issue is to get some information about the actual time in nanoseconds.

date +%N

$RANDOM in linux bash will alway be the same upon bootup

2 Answers2

Linked