Skip to contentSkip to conversation

bpf: Fix NULL pointer dereference in bpf_sk_storage_clone and diag paths

PATCH Superseded Weiming Shi 4 participants 4 messages Apr 21 – Apr 22

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 bpf: Fix NULL pointer dereference in bpf_sk_storage_clone and diag paths

bpf_selem_unlink_nofail() sets SDATA(selem)->smap to NULL before removing the selem from the storage hlist. A concurrent RCU reader in bpf_sk_storage_clone() can observe the selem still on the list with smap already NULL, causing a NULL pointer dereference.

 general protection fault, probably for non-canonical address 0xdffffc000000000a:
 KASAN: null-ptr-deref in range [0x0000000000000050-0x0000000000000057]
 RIP: 0010:bpf_sk_storage_clone+0x1cd/0xaa0 net/core/bpf_sk_storage.c:174
 Call Trace:
  <IRQ>
  sk_clone+0xfed/0x1980 net/core/sock.c:2591
  inet_csk_clone_lock+0x30/0x760 net/ipv4/inet_connection_sock.c:1222
  tcp_create_openreq_child+0x35/0x2680 net/ipv4/tcp_minisocks.c:571
  tcp_v4_syn_recv_sock+0x123/0xf90 net/ipv4/tcp_ipv4.c:1729
  tcp_check_req+0x8e1/0x2580 include/net/tcp.h:855
  tcp_v4_rcv+0x1845/0x3b80 net/ipv4/tcp_ipv4.c:2347

Add a NULL check for smap in bpf_sk_storage_clone(). bpf_sk_storage_diag_put_all() has the same unprotected dereference pattern, add a NULL check there as well.

Fixes 5d800f87d0a5 ("bpf: Support lockless unlink when freeing map or local storage")

---
v2:
drop the NULL check in diag_get(); The caller already checks smap for
NULL.

 net/core/bpf_sk_storage.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
net/core/bpf_sk_storage.c +3 -1
index f8338acebf077..67901de5b9c65 100644
@@ -172,7 +172,7 @@ int bpf_sk_storage_clone(const struct sock *sk, struct sock *newsk)
 		struct bpf_map *map;
 
 		smap = rcu_dereference(SDATA(selem)->smap);
-		if (!(smap->map.map_flags & BPF_F_CLONE))
+		if (!smap || !(smap->map.map_flags & BPF_F_CLONE))
 			continue;
 
 		/* Note that for lockless listeners adding new element
@@ -599,6 +599,8 @@ static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb,
 	saved_len = skb->len;
 	hlist_for_each_entry_rcu(selem, &sk_storage->list, snode) {
 		smap = rcu_dereference(SDATA(selem)->smap);
+		if (!smap)
+			continue;
 		diag_size += nla_value_size(smap->map.value_size);
 
 		if (nla_stgs && diag_get(SDATA(selem), skb))
Signature · 1 line
2.43.0
#1 lore raw
Martin KaFai Lau pointed out in the v1 review that the NULL check for smap doesn't fully address the race condition. The issue is that diag_get() will re-dereference sdata->smap independently:
Quoted 4 lines
net/core/bpf_sk_storage.c +0 -0
index 14eb7812bda4..b2cd2b0292cb 100644

[ ... ]

net/core/bpf_sk_storage.c static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb, · line 603

Quoted 9 lines
net/core/bpf_sk_storage.c +2 -0
@@ -596,6 +596,8 @@ static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb,
 	saved_len = skb->len;
 	hlist_for_each_entry_rcu(selem, &sk_storage->list, snode) {
 		smap = rcu_dereference(SDATA(selem)->smap);
+		if (!smap)
+			continue;
 		diag_size += nla_value_size(smap->map.value_size);
 
 		if (nla_stgs && diag_get(SDATA(selem), skb))

Martin KaFai Lau pointed out in the v1 review that the NULL check for smap doesn't fully address the race condition. The issue is that diag_get() will re-dereference sdata->smap independently:

net/core/bpf_sk_storage.c:diag_get() {
    ...
    smap = rcu_dereference(sdata->smap);
    ...
}

Between the NULL check above and the call to diag_get(), smap could become NULL again, leading to the same race condition.

Martin suggested passing smap as a parameter to diag_get() to avoid the second dereference. This would ensure diag_get() uses the same smap pointer that was already checked.

Could the fix be updated to pass smap to diag_get() as Martin suggested?

Reference: https://lore.kernel.org/bpf/[email protected]/

---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

CI run summary: https://github.com/kernel-patches/bpf/actions/runs/24732581685

#2 lore raw
Quoted 31 lines · On Tue, Apr 21, 2026 at 9:24 AM <[email protected]> wrote:
Quoted 5 lines
net/core/bpf_sk_storage.c +0 -0
index 14eb7812bda4..b2cd2b0292cb 100644

[ ... ]

net/core/bpf_sk_storage.c static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb, · line 603

Quoted 9 lines
net/core/bpf_sk_storage.c +2 -0
@@ -596,6 +596,8 @@ static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb,
      saved_len = skb->len;
      hlist_for_each_entry_rcu(selem, &sk_storage->list, snode) {
              smap = rcu_dereference(SDATA(selem)->smap);
+             if (!smap)
+                     continue;
              diag_size += nla_value_size(smap->map.value_size);
 
              if (nla_stgs && diag_get(SDATA(selem), skb))

Martin KaFai Lau pointed out in the v1 review that the NULL check for smap doesn't fully address the race condition. The issue is that diag_get() will re-dereference sdata->smap independently:

net/core/bpf_sk_storage.c:diag_get() {
    ...
    smap = rcu_dereference(sdata->smap);
    ...
}

Between the NULL check above and the call to diag_get(), smap could become NULL again, leading to the same race condition.

Quoted 9 lines

Martin suggested passing smap as a parameter to diag_get() to avoid the second dereference. This would ensure diag_get() uses the same smap pointer that was already checked.

Could the fix be updated to pass smap to diag_get() as Martin suggested?

Reference: https://lore.kernel.org/bpf/[email protected]/

CI is right. Please pass smap to diag_get() as Martin suggested. It will be the caller's responsibility to do rcu_dereference() and null-check it. Note that there are more than one caller to diag_get().

Quoted 6 lines

---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

CI run summary: https://github.com/kernel-patches/bpf/actions/runs/24732581685

#3 lore raw
In reply to Weiming Shi
Sashiko AI review found 1 potential issue(s): - [High] The NULL pointer dereference fix in the diag paths is incomplete due to a Time-Of-Check to Time-Of-Use (TOCTOU) race condition in `diag_get()`.…

Sashiko AI review found 1 potential issue(s):
- [High] The NULL pointer dereference fix in the diag paths is incomplete due to a Time-Of-Check to Time-Of-Use (TOCTOU) race condition in diag_get(). --

commit 89ba4092c0928afed16ec4db32d2b5ffaa88889b
Author: Weiming Shi <[email protected]>

bpf: Fix NULL pointer dereference in bpf_sk_storage_clone and diag paths

This commit adds NULL checks for the storage map pointer in bpf_sk_storage_clone() and bpf_sk_storage_diag_put_all() to prevent a NULL pointer dereference when racing with a concurrent storage map removal.

Quoted 4 lines
net/core/bpf_sk_storage.c +0 -0
index 14eb7812bda4a..b2cd2b0292cb6 100644

[ ... ]

net/core/bpf_sk_storage.c static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb, · line 603

Quoted 9 lines
net/core/bpf_sk_storage.c +2 -0
@@ -596,6 +596,8 @@ static int bpf_sk_storage_diag_put_all(struct sock *sk, struct sk_buff *skb,
 	saved_len = skb->len;
 	hlist_for_each_entry_rcu(selem, &sk_storage->list, snode) {
 		smap = rcu_dereference(SDATA(selem)->smap);
+		if (!smap)
+			continue;
 		diag_size += nla_value_size(smap->map.value_size);
 
 		if (nla_stgs && diag_get(SDATA(selem), skb))

Can this still trigger a NULL pointer dereference due to a race condition?

If a concurrent unlinking operation sets the storage map pointer to NULL after the check above, but before diag_get() executes, it looks like diag_get() could still read the NULL pointer:

net/core/bpf_sk_storage.c:diag_get() {
    ...
    smap = rcu_dereference(sdata->smap);
    if (nla_put_u32(skb, SK_DIAG_BPF_STORAGE_MAP_ID, smap->map.id))
    ...
}

Because rcu_dereference() re-fetches the pointer from memory, it could now return NULL, leading to a crash on the smap->map.id access.

Would it be safer to pass the already-validated smap pointer directly to diag_get() as an argument, or add a NULL check inside diag_get() itself?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1