Skip to content

Commit 287c988

Browse files
author
Sabrina Dubroca
committed
xfrm: state: fix out-of-bounds read during lookup
JIRA: https://issues.redhat.com/browse/RHEL-84576 Conflicts: missing the "xfrm: per cpu xfrm state handling" series - skip the changes in xfrm_input_state_lookup (introduced upstream in 81a331a ("xfrm: Add an inbound percpu state cache.")) - context changes xfrm_state_find (from 0045e3d ("xfrm: Cache used outbound xfrm states at the policy.")) commit e952837 Author: Florian Westphal <fw@strlen.de> Date: Thu Nov 28 15:26:25 2024 +0100 xfrm: state: fix out-of-bounds read during lookup lookup and resize can run in parallel. The xfrm_state_hash_generation seqlock ensures a retry, but the hash functions can observe a hmask value that is too large for the new hlist array. rehash does: rcu_assign_pointer(net->xfrm.state_bydst, ndst) [..] net->xfrm.state_hmask = nhashmask; While state lookup does: h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family); hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) { This is only safe in case the update to state_bydst is larger than net->xfrm.xfrm_state_hmask (or if the lookup function gets serialized via state spinlock again). Fix this by prefetching state_hmask and the associated pointers. The xfrm_state_hash_generation seqlock retry will ensure that the pointer and the hmask will be consistent. The existing helpers, like xfrm_dst_hash(), are now unsafe for RCU side, add lockdep assertions to document that they are only safe for insert side. xfrm_state_lookup_byaddr() uses the spinlock rather than RCU. AFAICS this is an oversight from back when state lookup was converted to RCU, this lock should be replaced with RCU in a future patch. Reported-by: syzbot+5f9f31cb7d985f584d8e@syzkaller.appspotmail.com Closes: https://lore.kernel.org/netdev/CACT4Y+azwfrE3uz6A5ZErov5YN2LYBN5KrsymBerT36VU8qzBA@mail.gmail.com/ Diagnosed-by: Dmitry Vyukov <dvyukov@google.com> Fixes: c2f672f ("xfrm: state lookup can be lockless") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com> Signed-off-by: Sabrina Dubroca <sdubroca@redhat.com>
1 parent 1f46989 commit 287c988

File tree

1 file changed

+66
-18
lines changed

1 file changed

+66
-18
lines changed

‎net/xfrm/xfrm_state.c‎

Lines changed: 66 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -34,6 +34,8 @@
3434

3535
#define xfrm_state_deref_prot(table, net) \
3636
rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
37+
#define xfrm_state_deref_check(table, net) \
38+
rcu_dereference_check((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
3739

3840
static void xfrm_state_gc_task(struct work_struct *work);
3941

@@ -62,6 +64,8 @@ static inline unsigned int xfrm_dst_hash(struct net *net,
6264
u32 reqid,
6365
unsigned short family)
6466
{
67+
lockdep_assert_held(&net->xfrm.xfrm_state_lock);
68+
6569
return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
6670
}
6771

@@ -70,18 +74,24 @@ static inline unsigned int xfrm_src_hash(struct net *net,
7074
const xfrm_address_t *saddr,
7175
unsigned short family)
7276
{
77+
lockdep_assert_held(&net->xfrm.xfrm_state_lock);
78+
7379
return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
7480
}
7581

7682
static inline unsigned int
7783
xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
7884
__be32 spi, u8 proto, unsigned short family)
7985
{
86+
lockdep_assert_held(&net->xfrm.xfrm_state_lock);
87+
8088
return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
8189
}
8290

8391
static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
8492
{
93+
lockdep_assert_held(&net->xfrm.xfrm_state_lock);
94+
8595
return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
8696
}
8797

@@ -1032,16 +1042,38 @@ xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
10321042
x->props.family = tmpl->encap_family;
10331043
}
10341044

1035-
static struct xfrm_state *__xfrm_state_lookup_all(struct net *net, u32 mark,
1045+
struct xfrm_hash_state_ptrs {
1046+
const struct hlist_head *bydst;
1047+
const struct hlist_head *bysrc;
1048+
const struct hlist_head *byspi;
1049+
unsigned int hmask;
1050+
};
1051+
1052+
static void xfrm_hash_ptrs_get(const struct net *net, struct xfrm_hash_state_ptrs *ptrs)
1053+
{
1054+
unsigned int sequence;
1055+
1056+
do {
1057+
sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1058+
1059+
ptrs->bydst = xfrm_state_deref_check(net->xfrm.state_bydst, net);
1060+
ptrs->bysrc = xfrm_state_deref_check(net->xfrm.state_bysrc, net);
1061+
ptrs->byspi = xfrm_state_deref_check(net->xfrm.state_byspi, net);
1062+
ptrs->hmask = net->xfrm.state_hmask;
1063+
} while (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence));
1064+
}
1065+
1066+
static struct xfrm_state *__xfrm_state_lookup_all(const struct xfrm_hash_state_ptrs *state_ptrs,
1067+
u32 mark,
10361068
const xfrm_address_t *daddr,
10371069
__be32 spi, u8 proto,
10381070
unsigned short family,
10391071
struct xfrm_dev_offload *xdo)
10401072
{
1041-
unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
1073+
unsigned int h = __xfrm_spi_hash(daddr, spi, proto, family, state_ptrs->hmask);
10421074
struct xfrm_state *x;
10431075

1044-
hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
1076+
hlist_for_each_entry_rcu(x, state_ptrs->byspi + h, byspi) {
10451077
#ifdef CONFIG_XFRM_OFFLOAD
10461078
if (xdo->type == XFRM_DEV_OFFLOAD_PACKET) {
10471079
if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
@@ -1075,15 +1107,16 @@ static struct xfrm_state *__xfrm_state_lookup_all(struct net *net, u32 mark,
10751107
return NULL;
10761108
}
10771109

1078-
static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
1110+
static struct xfrm_state *__xfrm_state_lookup(const struct xfrm_hash_state_ptrs *state_ptrs,
1111+
u32 mark,
10791112
const xfrm_address_t *daddr,
10801113
__be32 spi, u8 proto,
10811114
unsigned short family)
10821115
{
1083-
unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
1116+
unsigned int h = __xfrm_spi_hash(daddr, spi, proto, family, state_ptrs->hmask);
10841117
struct xfrm_state *x;
10851118

1086-
hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
1119+
hlist_for_each_entry_rcu(x, state_ptrs->byspi + h, byspi) {
10871120
if (x->props.family != family ||
10881121
x->id.spi != spi ||
10891122
x->id.proto != proto ||
@@ -1100,15 +1133,16 @@ static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
11001133
return NULL;
11011134
}
11021135

1103-
static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1136+
static struct xfrm_state *__xfrm_state_lookup_byaddr(const struct xfrm_hash_state_ptrs *state_ptrs,
1137+
u32 mark,
11041138
const xfrm_address_t *daddr,
11051139
const xfrm_address_t *saddr,
11061140
u8 proto, unsigned short family)
11071141
{
1108-
unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
1142+
unsigned int h = __xfrm_src_hash(daddr, saddr, family, state_ptrs->hmask);
11091143
struct xfrm_state *x;
11101144

1111-
hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
1145+
hlist_for_each_entry_rcu(x, state_ptrs->bysrc + h, bysrc) {
11121146
if (x->props.family != family ||
11131147
x->id.proto != proto ||
11141148
!xfrm_addr_equal(&x->id.daddr, daddr, family) ||
@@ -1128,14 +1162,17 @@ static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
11281162
static inline struct xfrm_state *
11291163
__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
11301164
{
1165+
struct xfrm_hash_state_ptrs state_ptrs;
11311166
struct net *net = xs_net(x);
11321167
u32 mark = x->mark.v & x->mark.m;
11331168

1169+
xfrm_hash_ptrs_get(net, &state_ptrs);
1170+
11341171
if (use_spi)
1135-
return __xfrm_state_lookup(net, mark, &x->id.daddr,
1172+
return __xfrm_state_lookup(&state_ptrs, mark, &x->id.daddr,
11361173
x->id.spi, x->id.proto, family);
11371174
else
1138-
return __xfrm_state_lookup_byaddr(net, mark,
1175+
return __xfrm_state_lookup_byaddr(&state_ptrs, mark,
11391176
&x->id.daddr,
11401177
&x->props.saddr,
11411178
x->id.proto, family);
@@ -1198,6 +1235,7 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
11981235
unsigned short family, u32 if_id)
11991236
{
12001237
static xfrm_address_t saddr_wildcard = { };
1238+
struct xfrm_hash_state_ptrs state_ptrs;
12011239
struct net *net = xp_net(pol);
12021240
unsigned int h, h_wildcard;
12031241
struct xfrm_state *x, *x0, *to_put;
@@ -1214,8 +1252,10 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
12141252
sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
12151253

12161254
rcu_read_lock();
1217-
h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1218-
hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1255+
xfrm_hash_ptrs_get(net, &state_ptrs);
1256+
1257+
h = __xfrm_dst_hash(daddr, saddr, tmpl->reqid, encap_family, state_ptrs.hmask);
1258+
hlist_for_each_entry_rcu(x, state_ptrs.bydst + h, bydst) {
12191259
#ifdef CONFIG_XFRM_OFFLOAD
12201260
if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
12211261
if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
@@ -1248,8 +1288,9 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
12481288
if (best || acquire_in_progress)
12491289
goto found;
12501290

1251-
h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1252-
hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1291+
h_wildcard = __xfrm_dst_hash(daddr, &saddr_wildcard, tmpl->reqid,
1292+
encap_family, state_ptrs.hmask);
1293+
hlist_for_each_entry_rcu(x, state_ptrs.bydst + h_wildcard, bydst) {
12531294
#ifdef CONFIG_XFRM_OFFLOAD
12541295
if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
12551296
if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
@@ -1284,7 +1325,7 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
12841325
x = best;
12851326
if (!x && !error && !acquire_in_progress) {
12861327
if (tmpl->id.spi &&
1287-
(x0 = __xfrm_state_lookup_all(net, mark, daddr,
1328+
(x0 = __xfrm_state_lookup_all(&state_ptrs, mark, daddr,
12881329
tmpl->id.spi, tmpl->id.proto,
12891330
encap_family,
12901331
&pol->xdo)) != NULL) {
@@ -2030,10 +2071,13 @@ struct xfrm_state *
20302071
xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
20312072
u8 proto, unsigned short family)
20322073
{
2074+
struct xfrm_hash_state_ptrs state_ptrs;
20332075
struct xfrm_state *x;
20342076

20352077
rcu_read_lock();
2036-
x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
2078+
xfrm_hash_ptrs_get(net, &state_ptrs);
2079+
2080+
x = __xfrm_state_lookup(&state_ptrs, mark, daddr, spi, proto, family);
20372081
rcu_read_unlock();
20382082
return x;
20392083
}
@@ -2044,10 +2088,14 @@ xfrm_state_lookup_byaddr(struct net *net, u32 mark,
20442088
const xfrm_address_t *daddr, const xfrm_address_t *saddr,
20452089
u8 proto, unsigned short family)
20462090
{
2091+
struct xfrm_hash_state_ptrs state_ptrs;
20472092
struct xfrm_state *x;
20482093

20492094
spin_lock_bh(&net->xfrm.xfrm_state_lock);
2050-
x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
2095+
2096+
xfrm_hash_ptrs_get(net, &state_ptrs);
2097+
2098+
x = __xfrm_state_lookup_byaddr(&state_ptrs, mark, daddr, saddr, proto, family);
20512099
spin_unlock_bh(&net->xfrm.xfrm_state_lock);
20522100
return x;
20532101
}

0 commit comments

Comments
 (0)