On Mon, Sep 05, 2016 at 11:58:24AM +0200, Laura Garcia wrote:
> On Mon, Sep 05, 2016 at 11:10:28AM +0200, Pablo Neira Ayuso wrote:
> > On Mon, Sep 05, 2016 at 10:36:57AM +0200, Laura Garcia Liebana wrote:
> > > Add support to pass through an offset to the hash value. With this
> > > feature, the sysadmin is able to generate a hash with a given
> > > offset value.
> > >
> > > Example:
> > >
> > > meta mark set jhash ip saddr mod 2 seed 0xabcd sum 100
> > >
> > > This option generates marks according to the source address from 100 to
> > > 101.
> > >
> > > Signed-off-by: Laura Garcia Liebana <nevola@xxxxxxxxx>
> > > ---
> > > include/uapi/linux/netfilter/nf_tables.h | 2 ++
> > > net/netfilter/nft_hash.c | 13 +++++++++++--
> > > 2 files changed, 13 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h
> > > index 4dbeeed..8026684 100644
> > > --- a/include/uapi/linux/netfilter/nf_tables.h
> > > +++ b/include/uapi/linux/netfilter/nf_tables.h
> > > @@ -764,6 +764,7 @@ enum nft_meta_keys {
> > > * @NFTA_HASH_LEN: source data length (NLA_U32)
> > > * @NFTA_HASH_MODULUS: modulus value (NLA_U32)
> > > * @NFTA_HASH_SEED: seed value (NLA_U32)
> > > + * @NFTA_HASH_SUM: Hash offset value (NLA_U32)
> > > */
> > > enum nft_hash_attributes {
> > > NFTA_HASH_UNSPEC,
> > > @@ -772,6 +773,7 @@ enum nft_hash_attributes {
> > > NFTA_HASH_LEN,
> > > NFTA_HASH_MODULUS,
> > > NFTA_HASH_SEED,
> > > + NFTA_HASH_SUM,
> > > __NFTA_HASH_MAX,
> > > };
> > > #define NFTA_HASH_MAX (__NFTA_HASH_MAX - 1)
> > > diff --git a/net/netfilter/nft_hash.c b/net/netfilter/nft_hash.c
> > > index b7e3b40..8ab04d9 100644
> > > --- a/net/netfilter/nft_hash.c
> > > +++ b/net/netfilter/nft_hash.c
> > > @@ -23,6 +23,7 @@ struct nft_hash {
> > > u8 len;
> > > u32 modulus;
> > > u32 seed;
> > > + u32 sum;
> > > };
> > >
> > > static void nft_hash_eval(const struct nft_expr *expr,
> > > @@ -35,7 +36,7 @@ static void nft_hash_eval(const struct nft_expr *expr,
> > >
> > > h = reciprocal_scale(jhash(data, priv->len, priv->seed), priv->modulus);
> > >
> > > - regs->data[priv->dreg] = h;
> > > + regs->data[priv->dreg] = priv->sum + h;
> > > }
> > >
> > > const struct nla_policy nft_hash_policy[NFTA_HASH_MAX + 1] = {
> > > @@ -44,6 +45,7 @@ const struct nla_policy nft_hash_policy[NFTA_HASH_MAX + 1] = {
> > > [NFTA_HASH_LEN] = { .type = NLA_U32 },
> > > [NFTA_HASH_MODULUS] = { .type = NLA_U32 },
> > > [NFTA_HASH_SEED] = { .type = NLA_U32 },
> > > + [NFTA_HASH_SUM] = { .type = NLA_U32 },
> > > };
> > >
> > > static int nft_hash_init(const struct nft_ctx *ctx,
> > > @@ -60,6 +62,11 @@ static int nft_hash_init(const struct nft_ctx *ctx,
> > > !tb[NFTA_HASH_MODULUS])
> > > return -EINVAL;
> > >
> > > + if (tb[NFTA_HASH_SUM])
> > > + priv->sum = ntohl(nla_get_be32(tb[NFTA_HASH_SUM]));
> > > + else
> > > + priv->sum = 0;
> >
> > There is a corner case that we should reject from the kernel, I think
> > this is:
> >
> > if (priv->sum + priv->modulus - 1 < priv->sum)
> > return -EOVERFLOW;
> >
> > We'll handle this from userspace anyway too, but I think it's easy to
> > reject this crazy this.
>
> Such case shouldn't happen cause the modulus must be > 1. The init()
> provides:
>
> priv->modulus = ntohl(nla_get_be32(tb[NFTA_HASH_MODULUS]));
> if (priv->modulus <= 1)
> return -ERANGE;
I don't see how this is preventing an overflow of hash(x) + sum.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
