diff mbox series

[nf] netfilter: ensure rcu_read_lock() in ipv4_find_option()

Message ID 20200224185529.50530-1-mcroce@redhat.com
State Awaiting Upstream
Delegated to: David Miller
Headers show
Series [nf] netfilter: ensure rcu_read_lock() in ipv4_find_option() | expand

Commit Message

Matteo Croce Feb. 24, 2020, 6:55 p.m. UTC
As in commit c543cb4a5f07 ("ipv4: ensure rcu_read_lock() in ipv4_link_failure()")
and commit 3e72dfdf8227 ("ipv4: ensure rcu_read_lock() in cipso_v4_error()"),
__ip_options_compile() must be called under rcu protection.

Fixes: dbb5281a1f84 ("netfilter: nf_tables: add support for matching IPv4 options")
Signed-off-by: Matteo Croce <mcroce@redhat.com>
---
 net/netfilter/nft_exthdr.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

Comments

Florian Westphal Feb. 24, 2020, 7:11 p.m. UTC | #1
Matteo Croce <mcroce@redhat.com> wrote:
> As in commit c543cb4a5f07 ("ipv4: ensure rcu_read_lock() in ipv4_link_failure()")
> and commit 3e72dfdf8227 ("ipv4: ensure rcu_read_lock() in cipso_v4_error()"),
> __ip_options_compile() must be called under rcu protection.

This is not needed, all netfilter hooks run with rcu_read_lock held.
Matteo Croce Feb. 24, 2020, 7:42 p.m. UTC | #2
On Mon, Feb 24, 2020 at 8:12 PM Florian Westphal <fw@strlen.de> wrote:
>
> Matteo Croce <mcroce@redhat.com> wrote:
> > As in commit c543cb4a5f07 ("ipv4: ensure rcu_read_lock() in ipv4_link_failure()")
> > and commit 3e72dfdf8227 ("ipv4: ensure rcu_read_lock() in cipso_v4_error()"),
> > __ip_options_compile() must be called under rcu protection.
>
> This is not needed, all netfilter hooks run with rcu_read_lock held.
>

Ok, so let's drop it, thanks.
Matteo Croce Feb. 24, 2020, 7:47 p.m. UTC | #3
On Mon, Feb 24, 2020 at 8:42 PM Matteo Croce <mcroce@redhat.com> wrote:
>
> On Mon, Feb 24, 2020 at 8:12 PM Florian Westphal <fw@strlen.de> wrote:
> >
> > Matteo Croce <mcroce@redhat.com> wrote:
> > > As in commit c543cb4a5f07 ("ipv4: ensure rcu_read_lock() in ipv4_link_failure()")
> > > and commit 3e72dfdf8227 ("ipv4: ensure rcu_read_lock() in cipso_v4_error()"),
> > > __ip_options_compile() must be called under rcu protection.
> >
> > This is not needed, all netfilter hooks run with rcu_read_lock held.
> >
>
> Ok, so let's drop it, thanks.

What about adding a RCU_LOCKDEP_WARN() in __ip_options_compile() to
protect against future errors? Something like:

----------------------------------%<-------------------------------------
@@ -262,6 +262,9 @@ int __ip_options_compile(struct net *net,
  unsigned char *iph;
  int optlen, l;

+ RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
+ __FUNC__ " needs rcu_read_lock() protection");
+
  if (skb) {
  rt = skb_rtable(skb);
  optptr = (unsigned char *)&(ip_hdr(skb)[1]);
---------------------------------->%-------------------------------------

Bye,
diff mbox series

Patch

diff --git a/net/netfilter/nft_exthdr.c b/net/netfilter/nft_exthdr.c
index a5e8469859e3..752264b3043a 100644
--- a/net/netfilter/nft_exthdr.c
+++ b/net/netfilter/nft_exthdr.c
@@ -77,6 +77,7 @@  static int ipv4_find_option(struct net *net, struct sk_buff *skb,
 	bool found = false;
 	__be32 info;
 	int optlen;
+	int ret;
 
 	iph = skb_header_pointer(skb, 0, sizeof(_iph), &_iph);
 	if (!iph)
@@ -95,7 +96,11 @@  static int ipv4_find_option(struct net *net, struct sk_buff *skb,
 		return -EBADMSG;
 	opt->optlen = optlen;
 
-	if (__ip_options_compile(net, opt, NULL, &info))
+	rcu_read_lock();
+	ret = __ip_options_compile(net, opt, NULL, &info);
+	rcu_read_unlock();
+
+	if (ret)
 		return -EBADMSG;
 
 	switch (target) {