Patchwork lnf_conntrack: nfct_cmp NFCT_CMP_TIMEOUT_* flags not supported?

login
register
mail settings
Submitter Florian Westphal
Date Nov. 28, 2012, 12:59 p.m.
Message ID <20121128125930.GF14156@breakpoint.cc>
Download mbox | patch
Permalink /patch/202444/
State RFC
Headers show

Comments

Florian Westphal - Nov. 28, 2012, 12:59 p.m.
Hi.

I added api_tests for the various nfct_cmp timeout flags.
And guess what: They don't work 8-}

It fails on the 2nd assert below:
assert(nfct_cmp(ct, ct2, NFCT_CMP_TIMEOUT_EQ) == 1);
nfct_set_attr_u32(ct2, ATTR_TIMEOUT, nfct_get_attr_u32(ct, ATTR_TIMEOUT) + 1);
assert(nfct_cmp(ct2, ct, NFCT_CMP_TIMEOUT_EQ) == 0);

The reason is that __compare() doesn't know about NFCT_CMP_TIMEOUT*
flags and returns 1 unconditionally.

So, my question is:
How are the NFCT_CMP_TIMEOUT flags supposed to be used?

From the documentation it appears as if they should be used
together with _ALL, _ORIG, _REPLY, or even standalone, i.e.
__compare needs to check for these, too:


With the above change the new tests pass.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 28, 2012, 3:16 p.m.
Hi Florian,

On Wed, Nov 28, 2012 at 01:59:30PM +0100, Florian Westphal wrote:
> Hi.
> 
> I added api_tests for the various nfct_cmp timeout flags.
> And guess what: They don't work 8-}
> 
> It fails on the 2nd assert below:
> assert(nfct_cmp(ct, ct2, NFCT_CMP_TIMEOUT_EQ) == 1);
> nfct_set_attr_u32(ct2, ATTR_TIMEOUT, nfct_get_attr_u32(ct, ATTR_TIMEOUT) + 1);
> assert(nfct_cmp(ct2, ct, NFCT_CMP_TIMEOUT_EQ) == 0);
> 
> The reason is that __compare() doesn't know about NFCT_CMP_TIMEOUT*
> flags and returns 1 unconditionally.
> 
> So, my question is:
> How are the NFCT_CMP_TIMEOUT flags supposed to be used?

They planned to be used by the conntrack utility. To obtain timers
that are over/under some given timeout. But that was never
implemented, so that code has remain untested there so far until
someone has come to show some interest on it ;-).

> From the documentation it appears as if they should be used
> together with _ALL, _ORIG, _REPLY, or even standalone, i.e.
> __compare needs to check for these, too:

I think standalone if the way to go, I think they deserve special
treatment. Note that I'm using nfct_cmp in conntrackd to look up for
entries in the internal cache hashtable, so enabling that comparison
with _ALL, _ORIG and _REPLY would resulting in mismatching.

> diff --git a/src/conntrack/compare.c b/src/conntrack/compare.c
> index b18f7fc..7cd28e7 100644
> --- a/src/conntrack/compare.c
> +++ b/src/conntrack/compare.c
> @@ -407,5 +407,8 @@ int __compare(const struct nf_conntrack *ct1,
>         if (flags & NFCT_CMP_REPL && !cmp_repl(ct1, ct2, flags))
>                 return 0;
>  
> +       if (flags & (NFCT_CMP_TIMEOUT_GT|NFCT_CMP_TIMEOUT_LE))
> +               return cmp_meta(ct1, ct2, flags);
> +
>         return 1;
>  }
> 
> With the above change the new tests pass.
> --
> To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Florian Westphal - Nov. 28, 2012, 3:33 p.m.
Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > So, my question is:
> > How are the NFCT_CMP_TIMEOUT flags supposed to be used?
> 
> They planned to be used by the conntrack utility. To obtain timers
> that are over/under some given timeout. But that was never
> implemented, so that code has remain untested there so far until
> someone has come to show some interest on it ;-).

Alright, I'll hold off with this change, then.
Problem is that this change:

> > diff --git a/src/conntrack/compare.c b/src/conntrack/compare.c
> > index b18f7fc..7cd28e7 100644
> > --- a/src/conntrack/compare.c
> > +++ b/src/conntrack/compare.c
> > @@ -407,5 +407,8 @@ int __compare(const struct nf_conntrack *ct1,
> >         if (flags & NFCT_CMP_REPL && !cmp_repl(ct1, ct2, flags))
> >                 return 0;
> >  
> > +       if (flags & (NFCT_CMP_TIMEOUT_GT|NFCT_CMP_TIMEOUT_LE))
> > +               return cmp_meta(ct1, ct2, flags);
> > +
> >         return 1;
> >  }

... is wrong after all.
IMO e.g. NFCT_CMP_TIMEOUT_EQ means that the comparision is valid
when the timeout is equal.  But above code means that its
valid when the timeout is equal AND all the other meta flags are
equal, too.  And thats very un-intuitive.

Regards,
Florian
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/src/conntrack/compare.c b/src/conntrack/compare.c
index b18f7fc..7cd28e7 100644
--- a/src/conntrack/compare.c
+++ b/src/conntrack/compare.c
@@ -407,5 +407,8 @@  int __compare(const struct nf_conntrack *ct1,
        if (flags & NFCT_CMP_REPL && !cmp_repl(ct1, ct2, flags))
                return 0;
 
+       if (flags & (NFCT_CMP_TIMEOUT_GT|NFCT_CMP_TIMEOUT_LE))
+               return cmp_meta(ct1, ct2, flags);
+
        return 1;
 }