* [PATCH net v4] gtp: serialize PDP context updates
@ 2026-08-14 3:15 Qing Ming
0 siblings, 0 replies; only message in thread
From: Qing Ming @ 2026-08-14 3:15 UTC (permalink / raw)
To: Pablo Neira Ayuso, Harald Welte
Cc: Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, horms, kuniyu, osmocom-net-gprs, netdev,
linux-kernel, Qing Ming, stable
PDP contexts can be deleted through GTP_CMD_DELPDP or while the GTP
network device is being unregistered. The latter is serialized by RTNL,
but the generic-netlink delete path only holds RCU.
Running both paths concurrently can therefore make both paths delete the
same PDP context. The issue was found through static analysis and
reproduced on a KASAN-enabled kernel by a simple two-thread program
racing GTP_CMD_DELPDP against RTM_DELLINK:
Oops: general protection fault, probably for non-canonical address
KASAN: maybe wild-memory-access in range
[0xdead000000000120-0xdead000000000127]
RIP: gtp_genl_del_pdp+0x1c1/0x420 [gtp]
RBP: dead000000000122
The second deletion dereferenced the poisoned hlist pprev pointer.
Serialize gtp_pdp_add(), gtp_genl_del_pdp(), and gtp_dellink() with a
shared mutex. With this change, the same reproducer completed 50 races
on a KASAN- and lockdep-enabled kernel without a KASAN report, oops, or
lockdep warning.
Fixes: 459aa660eb1d ("gtp: add initial driver for datapath of GPRS Tunneling Protocol (GTP-U)")
Cc: stable@vger.kernel.org
Assisted-by: Codex:gpt-5
Signed-off-by: Qing Ming <a0yami@mailbox.org>
---
v4:
- Replace guard() with plain mutex_lock()/mutex_unlock() in the
GTP_CMD_NEWPDP path.
- Narrow the RCU read-side section in gtp_genl_del_pdp() to
gtp_find_pdp(); gtp_find_dev() uses dev_get_by_index_rcu().
- Add mutex-aware lockdep conditions to the shared PDP lookup helpers.
v3: https://lore.kernel.org/netdev/20260811022012.5416-1-a0yami@mailbox.org/
drivers/net/gtp.c | 25 +++++++++++++++++++------
1 file changed, 19 insertions(+), 6 deletions(-)
diff --git a/drivers/net/gtp.c b/drivers/net/gtp.c
index 9a12cc53da00..ea83a6b6aa1a 100644
--- a/drivers/net/gtp.c
+++ b/drivers/net/gtp.c
@@ -12,6 +12,7 @@
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/module.h>
+#include <linux/mutex.h>
#include <linux/skbuff.h>
#include <linux/udp.h>
#include <linux/rculist.h>
@@ -108,6 +109,7 @@ struct gtp_net {
};
static u32 gtp_h_initval;
+static DEFINE_MUTEX(gtp_pdp_lock);
static struct genl_family gtp_genl_family;
@@ -151,7 +153,8 @@ static struct pdp_ctx *gtp0_pdp_find(struct gtp_dev *gtp, u64 tid, u16 family)
head = >p->tid_hash[gtp0_hashfn(tid) % gtp->hash_size];
- hlist_for_each_entry_rcu(pdp, head, hlist_tid) {
+ hlist_for_each_entry_rcu(pdp, head, hlist_tid,
+ lockdep_is_held(>p_pdp_lock)) {
if (pdp->af == family &&
pdp->gtp_version == GTP_V0 &&
pdp->u.v0.tid == tid)
@@ -168,7 +171,8 @@ static struct pdp_ctx *gtp1_pdp_find(struct gtp_dev *gtp, u32 tid, u16 family)
head = >p->tid_hash[gtp1u_hashfn(tid) % gtp->hash_size];
- hlist_for_each_entry_rcu(pdp, head, hlist_tid) {
+ hlist_for_each_entry_rcu(pdp, head, hlist_tid,
+ lockdep_is_held(>p_pdp_lock)) {
if (pdp->af == family &&
pdp->gtp_version == GTP_V1 &&
pdp->u.v1.i_tei == tid)
@@ -185,7 +189,8 @@ static struct pdp_ctx *ipv4_pdp_find(struct gtp_dev *gtp, __be32 ms_addr)
head = >p->addr_hash[ipv4_hashfn(ms_addr) % gtp->hash_size];
- hlist_for_each_entry_rcu(pdp, head, hlist_addr) {
+ hlist_for_each_entry_rcu(pdp, head, hlist_addr,
+ lockdep_is_held(>p_pdp_lock)) {
if (pdp->af == AF_INET &&
pdp->ms.addr.s_addr == ms_addr)
return pdp;
@@ -220,7 +225,8 @@ static struct pdp_ctx *ipv6_pdp_find(struct gtp_dev *gtp,
head = >p->addr_hash[ipv6_hashfn(ms_addr) % gtp->hash_size];
- hlist_for_each_entry_rcu(pdp, head, hlist_addr) {
+ hlist_for_each_entry_rcu(pdp, head, hlist_addr,
+ lockdep_is_held(>p_pdp_lock)) {
if (pdp->af == AF_INET6 &&
ipv6_pdp_addr_equal(&pdp->ms.addr6, ms_addr))
return pdp;
@@ -1555,9 +1561,11 @@ static void gtp_dellink(struct net_device *dev, struct list_head *head)
struct pdp_ctx *pctx;
int i;
+ mutex_lock(>p_pdp_lock);
for (i = 0; i < gtp->hash_size; i++)
hlist_for_each_entry_safe(pctx, next, >p->tid_hash[i], hlist_tid)
pdp_context_delete(pctx);
+ mutex_unlock(>p_pdp_lock);
list_del(>p->list);
unregister_netdevice_queue(dev, head);
@@ -2053,7 +2061,9 @@ static int gtp_genl_new_pdp(struct sk_buff *skb, struct genl_info *info)
goto out_unlock;
}
+ mutex_lock(>p_pdp_lock);
pctx = gtp_pdp_add(gtp, sk, info);
+ mutex_unlock(>p_pdp_lock);
if (IS_ERR(pctx)) {
err = PTR_ERR(pctx);
} else {
@@ -2134,9 +2144,12 @@ static int gtp_genl_del_pdp(struct sk_buff *skb, struct genl_info *info)
if (!info->attrs[GTPA_VERSION])
return -EINVAL;
- rcu_read_lock();
+ mutex_lock(>p_pdp_lock);
+ /* gtp_find_dev() uses dev_get_by_index_rcu(). */
+ rcu_read_lock();
pctx = gtp_find_pdp(sock_net(skb->sk), info->attrs);
+ rcu_read_unlock();
if (IS_ERR(pctx)) {
err = PTR_ERR(pctx);
goto out_unlock;
@@ -2153,7 +2166,7 @@ static int gtp_genl_del_pdp(struct sk_buff *skb, struct genl_info *info)
pdp_context_delete(pctx);
out_unlock:
- rcu_read_unlock();
+ mutex_unlock(>p_pdp_lock);
return err;
}
base-commit: 24ef02f934eeb48830cff6b739abc3c62b1d107b
--
2.53.0
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2026-08-14 3:15 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-08-14 3:15 [PATCH net v4] gtp: serialize PDP context updates Qing Ming
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox