kern/164402: [pf] pf crashes with a particular set of rules when first matching packet arrives

[Gleb Smirnoff]

On Sun, Apr 15, 2012 at 12:00:21PM +0000, Gleb Smirnoff wrote:
T>  On Sun, Apr 15, 2012 at 11:10:03AM +0000, Gleb Smirnoff wrote:
T>  T>    I have a vague suspicion on what is happening. Your description of
T>  T>  the problem looks like if a packet processing in the kernel has entered
T>  T>  an endless loop.
T>  T>  
T>  T>    Looking at pf_route() I see such possibility. From OpenBSD we have
T>  T>  this protection against endless looping:
T>  T>  
T>  T>          if ((*m)->m_pkthdr.pf.routed++ > 3) {
T>  T>                  m0 = *m;
T>  T>                  *m = NULL;
T>  T>                  goto bad;
T>  T>          }
T>  T>  
T>  T>  In our code this transforms to:
T>  T>  
T>  T>          if (pd->pf_mtag->routed++ > 3) {
T>  T>                  m0 = *m;
T>  T>                  *m = NULL;
T>  T>                  goto bad;
T>  T>          }
T>  T>  
T>  T>  The root difference between storing the tag on mbuf and on pfdesc
T>  T>  is that we lose pfdesc, and thus the tag, when we enter pf_test()
T>  T>  recursively. And pf_route() does this recursion:
T>  T>  
T>  T>          if (oifp != ifp) {
T>  T>                  if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) {
T>  T>                          goto bad;
T>  T>  	....
T>  
T>  On second look I see that my suspicion may not be true. In the
T>  beginning of pf_test() we do pf_get_mtag() which preserves already
T>  present tag if there is one.

Further investigation showed that problem exist when route applied
ends in lo0, and packet passes to if_simloop(). There all mtags are
stripped from the mbuf, including the pf mtag. Then packet is again
processed by ip_input() again entering pf(4), if it again matches
a routing rule, then we got an endless loop.

We can try to fix this applying MTAG_PERSISTENT to the pf(4) tag id.

-- 
Totus tuus, Glebius.