A classic and standard C quoting bug
I recently read Evan Martin's entry on quoting and escaping, which brought to mind one of those classic C and Unix programming mistakes which hopefully is not done very much any more.
Suppose that ustr
is a string that comes from user input. The classic
form of this particular quoting bug is to write printf(ustr)
(or
perhaps fprintf(out, ustr)
); a much obvious version of this is
syslog(pri, ustr)
.
(Some people might think that the printf()
version is silly since
there are much more nominally efficient ways to print plain strings, but
in practice plenty of people find it much more efficient on programmer
time to use printf()
or fprintf()
to output everything.)
The problem with all of these is that the first argument to printf()
and syslog()
is not a plain string to be printed; it is a format.
Like many quoting bugs, this goes undetected for much of the time
because a true plain string is also a valid format that formats to
itself. However, if someone supplies a 'plain string' that includes
printf formatting directives, things rapidly go off the rails; if
you are lucky the program crashes right away so you can immediately
notice the problem.
(If you are sufficiently unlucky, this is an exploitable security vulnerability. And yes, people have written code like this that made it into important programs.)
The right solution is of course to quote the user supplied string. The
simple way to do this is to supply your own simple formatting string:
syslog(pri, "%s", ustr)
(or the equivalent with printf()
et
al). Of course at this point you might want to think about other bad
characters that could appear in ustr
and how you want to display the
message to make it clear that the string comes from the user, not from
your program's internals.
This bug can happen with any function that takes a format string as an argument and in any language, not just C. C is just a more dangerous language to have it happen in because C generally has no check for the wrong number of arguments being supplied to a function.
(I have opinions on how languages can avoid this entire class of bugs, but that's something for another entry.)
(Perhaps this is not strictly a quoting bug. It is in my mind, but I may have a somewhat odd view of what constitutes one.)
Comments on this page:
|
|