Here is a simple patch to allow using a fixed size random file source with
shred.  Shred has the limitation that if --random-source=<file> is
specified, the source must be at least as large as the destination, which
may not be feasible in many cases.  Without this patch, shred will exit
prematurely when it reaches the end of the source file rather than just
completing the job using as many copies of the given data as is necessary.

This is very helpful if you are not concerned with having blocks of random
data duplicated as much as significantly speeding up the process over using
/dev/urandom, or if you want to overwrite with a specific pattern.

===== cut here =====
*** lib/randread.c.orig 2008-01-17 03:14:51.000000000 -0600
--- lib/randread.c  2009-09-30 19:38:10.000000000 -0500
***************
*** 217,222 ****
--- 217,226 ----
         size -= inbytes;
         if (size == 0)
         break;
+       if (feof(s->source)) {
+         rewind(s->source);
+         continue;
+       }
         errno = (ferror (s->source) ? fread_errno : 0);
         s->handler (s->handler_arg);
       }
===== cut here =====

Thanks.

Evan