On Tue, Mar 06, 2012 at 04:50:21AM -0800, Eric Dumazet wrote:
> On Tue, 2012-03-06 at 12:22 +0100, pablo@xxxxxxxxxxxxx wrote:
> > From: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx>
> >
> > All ctnetlink operations are invoked inside rcu_read_lock
> > (see net/netfilter/nfnetlink.c).
> >
> > Allocations have to be atomic, as RCU requires.
> >
> > Signed-off-by: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx>
> > ---
> > net/netfilter/nf_conntrack_netlink.c | 4 ++--
> > 1 files changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c
> > index 1068769..867843f 100644
> > --- a/net/netfilter/nf_conntrack_netlink.c
> > +++ b/net/netfilter/nf_conntrack_netlink.c
> > @@ -1002,7 +1002,7 @@ ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
> > ct = nf_ct_tuplehash_to_ctrack(h);
> >
> > err = -ENOMEM;
> > - skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
> > + skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
> > if (skb2 == NULL) {
> > nf_ct_put(ct);
> > return -ENOMEM;
> > @@ -1865,7 +1865,7 @@ ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
> > }
> >
> > err = -ENOMEM;
> > - skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
> > + skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
> > if (skb2 == NULL) {
> > nf_ct_expect_put(exp);
> > goto out;
>
> This cant be right.
>
> Really this must be kept as GFP_KERNEL allocations.
>
> Only if .call_rcu member is used in place of .call rcu_read_lock() is
> held instead of nfnl_lock().
I thought we couldn't sleep while holding rcu_read_lock.
> You should take a look at all GFP_ATOMIC uses in
> net/netfilter/nf_conntrack_netlink.c and check if they can be GFP_KERNEL
> instead.
David, can you take all patches except this one?
I'll have to rebase my tree after this, sorry.
--
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
