[lkml]   [2005]   [Jul]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectIs this a bug in linux-2.6.12 ipsec code function xfrm4_rcv_encap ??
I see a possible race in linux-2.6.12 ipsec code function xfrm4_rcv_encap.
I want to double check with the group.
The issue is with SMP(mostly) or Preemptible Kernels.
The race comes when someone flushes the SA's
(setkey -Fexecuting on another processor )
while xfrm_rcv_encap is executing one processor.

Below is the function code.
I am putting comments in the code where probably the race comes.
correct me if I am wrong.

int xfrm4_rcv_encap(struct sk_buff *skb, __u16 encap_type)
int err;
u32 spi, seq;
struct sec_decap_state xfrm_vec[XFRM_MAX_DEPTH];
struct xfrm_state *x;
int xfrm_nr = 0;
int decaps = 0;

if ((err = xfrm4_parse_spi(skb, skb->nh.iph->protocol, &spi, &seq)) != 0)
goto drop;

do {
struct iphdr *iph = skb->nh.iph;

if (xfrm_nr == XFRM_MAX_DEPTH)
goto drop;

x = xfrm_state_lookup((xfrm_address_t *)&iph->daddr, spi,
iph->protocol, AF_INET);

First Race here . Check is being done without x being locked. What if
x becomes null because
of SA FLUSH (setkey -F) after the check.
if (x == NULL)
goto drop;

if (unlikely(x->km.state != XFRM_STATE_VALID))
goto drop_unlock;

if (x->props.replay_window && xfrm_replay_check(x, seq))
goto drop_unlock;

if (xfrm_state_check_expire(x))
goto drop_unlock;

xfrm_vec[xfrm_nr].decap.decap_type = encap_type;
if (x->type->input(x, &(xfrm_vec[xfrm_nr].decap), skb))
goto drop_unlock;

/* only the first xfrm gets the encap type */
encap_type = 0;

if (x->props.replay_window)
xfrm_replay_advance(x, seq);

x->curlft.bytes += skb->len;


Second Race Here. Note the above line unlock already called.
xfrm_vec[xfrm_nr++].xvec = x;

iph = skb->nh.iph;

Third Race Here . Again the Check is without Lock
if (x->props.mode) {
if (iph->protocol != IPPROTO_IPIP)
goto drop;
if (!pskb_may_pull(skb, sizeof(struct iphdr)))
goto drop;
if (skb_cloned(skb) &&
pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
goto drop;
if (x->props.flags & XFRM_STATE_DECAP_DSCP)
ipv4_copy_dscp(iph, skb->h.ipiph);
if (!(x->props.flags & XFRM_STATE_NOECN))
skb->mac.raw = memmove(skb->data - skb->mac_len,
skb->mac.raw, skb->mac_len);
skb->nh.raw = skb->data;
memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
decaps = 1;

if ((err = xfrm_parse_spi(skb, skb->nh.iph->protocol, &spi, &seq)) < 0)
goto drop;
} while (!err);

/* Allocate new secpath or COW existing one. */

if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
struct sec_path *sp;
sp = secpath_dup(skb->sp);
if (!sp)
goto drop;
if (skb->sp)
skb->sp = sp;
if (xfrm_nr + skb->sp->len > XFRM_MAX_DEPTH)
goto drop;

memcpy(skb->sp->x+skb->sp->len, xfrm_vec, xfrm_nr*sizeof(struct
skb->sp->len += xfrm_nr;

if (decaps) {
if (!(skb->dev->flags&IFF_LOOPBACK)) {
skb->dst = NULL;
return 0;
} else {
return -skb->nh.iph->protocol;

while (--xfrm_nr >= 0)

return 0;

I am just guessing.
If I am wrong I request anyone to give me a reason why it is not a bug ?
I haven't checked the IPv6 front and the IPSec outbound side.
Once this proves to be a bug I will check them.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-07-23 03:11    [W:0.025 / U:0.192 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site