Re: [PATCH ima-evm-utils] evmctl: fix memory leak in get_password

[Bruno Meneguele <bmeneg@redhat.com>]

On Wed, Aug 11, 2021 at 02:28:37PM -0400, Ken Goldman wrote:
> On 8/11/2021 1:31 PM, Mimi Zohar wrote:
> > On Wed, 2021-08-11 at 13:51 -0300, Bruno Meneguele wrote:
> > > On Wed, Aug 11, 2021 at 10:52:00AM -0400, Mimi Zohar wrote:
> > > 
> > > > > 
> > > > > -	return pwd;
> > > > > +	return password;
> > > > 
> > > > Wouldn't a simpler fix be to test "pwd" here?
> > > >          if (!pwd)
> > > >                  free(password);
> > > >          return pwd;
> > > > 
> > > 
> > > The problem is on success, when 'pwd' is actually not NULL.
> > > With that, I can't free(password). I would need to asprintf(pwd, ...) or
> > > strndup(password). Because of that, I thought it would be cleaner to
> > > remove 'password' completely.
> > 
> > I see.  So instead of "return pwd" as suggested above,
> > 
> >          if (!pwd) {
> >                  free(password);
> >                  password = NULL;  <== set or return NULL
> >          }
> > 
> >          return password;
> 
> That looks cleaner to me.
> 
> My style would be
> 
> 	if (pwd == NULL)
> 
> which compiles to the same binary, but it less prone to error.
> 
> In addition, since this is reading from stdin
> 
> 1 - Do you want the newline to be part of the password?

I would say 'yes'. AFAIK OpenSSL preserves the newline if it's present
in the input from <stdin>:

"The returned string is always NUL-terminated and the '\n' is preserved
if present in the input data" (BIO_gets() manpage from OpenSSL)

Also, if the user passed the password to the PEM file creation through
the arguments list (no newline) it can also do the same with evmctl.

> 2 = Why is an empty password an error?
> 

Considering the item 1, I don't think we have an empty string in this
case.

-- 
bmeneg 
PGP Key: http://bmeneg.com/pubkey.txt