sit: convert configuration to RCU and lockless fill_info
Subsystems (3)
cover sit: convert configuration to RCU and lockless fill_info
SIT (IPv6-in-IPv4) tunnel configuration and status reporting have historically relied on the RTNL lock for synchronization. Consequently, netlink dumps via ipip6_fill_info() had to run with RTNL held, adding contention during network device dumps.
At the same time, the transmit path (dev->lltx == true), tunnel lookups, and error handling run locklessly and can race with configuration updates. This can result in torn reads of multi-word fields (such as the 128-bit 6RD IPv6 prefix) or transiently zeroed encapsulation parameters. Furthermore, ipip6_tunnel_update() currently unhashes, re-hashes, and calls synchronize_net() unconditionally, even when the tunnel endpoint addresses (saddr and daddr) have not changed.
This patch series addresses PRL issues, modernizes SIT parameter management to use RCU protection, optimizes tunnel updates, and removes the RTNL requirement from ipip6_fill_info():
- Patch 1 fixes a pre-existing UAF in PRL (Potential Router List) deletion where call_rcu() was invoked before unlinking t->prl.
- Patch 2 adds GFP_KERNEL_ACCOUNT to struct ip_tunnel_prl_entry allocations in ipip6_tunnel_add_prl().
- Patch 3 removes the unsafe in-place memset() in ip_tunnel_encap_setup() and uses WRITE_ONCE() to prevent lockless readers from observing transiently zeroed or torn fields.
- Patch 4 annotates data races on tunnel->fwmark with READ_ONCE() and WRITE_ONCE().
- Patch 5 converts 6RD configuration (tunnel->ip6rd) to an RCU-protected pointer, preventing torn reads on the 128-bit IPv6 prefix.
- Patch 6 implements a dedicated ipip6_get_iflink() callback to decouple SIT parameter handling from generic ip_tunnel.
- Patch 7 dynamically allocates struct ip_tunnel_parm_kern (sit_parms) as a preparatory step.
- Patch 8 converts tunnel->sit_parms to full RCU protection. Updates publish new parameters via rcu_assign_pointer() and free the old ones via kfree_rcu(). When saddr and daddr do not change, unhashing, re-hashing, and synchronize_net() are completely bypassed.
- Patch 9 wraps attribute serialization in ipip6_fill_info() under rcu_read_lock(), eliminating the reliance on the RTNL lock.
Eric Dumazet (9):
sit: fix UAF in ipip6_tunnel_del_prl()
sit: charge ip_tunnel_prl_entry allocations to memcg
ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup
sit: annotate data-races around tunnel->fwmark
sit: convert 6RD configuration to RCU protection
sit: implement ipip6_get_iflink()
sit: dynamically allocate struct ip_tunnel_parm_kern
sit: convert configuration to RCU protection
sit: no longer rely on RTNL in ipip6_fill_info()
include/net/ip_tunnels.h | 5 +- net/ipv4/ip_tunnel.c | 14 +- net/ipv6/sit.c | 477 +++++++++++++++++++++++++++------------ 3 files changed, 339 insertions(+), 157 deletions(-)
Signature · 1 line
2.55.0.979.g7e5102b832-goog
PATCH 1/9 sit: fix UAF in ipip6_tunnel_del_prl()
When flushing all potential router list (PRL) entries, ipip6_tunnel_del_prl() called call_rcu(&x->rcu_head, prl_list_destroy_rcu) before clearing t->prl.
A concurrent reader in isatap_chksrc() could enter an RCU read-side critical section after call_rcu() but before t->prl is set to NULL, allowing prl_list_destroy_rcu() to free nodes while the reader traverses them.
Clear t->prl using RCU_INIT_POINTER() before invoking call_rcu(). Also use rcu_assign_pointer() when unlinking individual entries.
--- net/ipv6/sit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
net/ipv6/sit.c +2 -2
@@ -436,7 +436,7 @@ ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) (x =rtnl_dereference (*p)) !=NULL ; p = &x->next) {if (x->addr == a->addr) { - *p = x->next; +rcu_assign_pointer (*p,rtnl_dereference (x->next));kfree_rcu (x, rcu_head); t->prl_count--; goto out;
@@ -447,8 +447,8 @@ ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) x =rtnl_dereference (t->prl);if (x) { t->prl_count =0 ; +RCU_INIT_POINTER (t->prl,NULL );call_rcu (&x->rcu_head, prl_list_destroy_rcu); - t->prl =NULL ; } } out:
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 1/9] sit: fix UAF in ipip6_tunnel_del_prl()
Quoted 9 lines
When flushing all potential router list (PRL) entries,
ipip6_tunnel_del_prl()calledcall_rcu(&x->rcu_head, prl_list_destroy_rcu)before clearingt->prl.A concurrent reader in
isatap_chksrc()could enter an RCU read-side critical section aftercall_rcu()but beforet->prlis set to NULL, allowingprl_list_destroy_rcu()to free nodes while the reader traverses them.
Quoted 5 lines
Clear
t->prlusingRCU_INIT_POINTER()before invokingcall_rcu(). Also usercu_assign_pointer()when unlinking individual entries.Fixes ef9a9d1183b3 ("ipv6 sit: RCU conversion phase I")Signed-off-by Eric Dumazet <[email protected]>
Acked-by Lorenzo Bianconi <[email protected]>
Quoted 30 lines
--- net/ipv6/sit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)net/ipv6/sit.c +2
-2@@ -436,7 +436,7 @@ ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) (x =rtnl_dereference (*p)) !=NULL ; p = &x->next) {if (x->addr == a->addr) { - *p = x->next; +rcu_assign_pointer (*p,rtnl_dereference (x->next));kfree_rcu (x, rcu_head); t->prl_count--; goto out;@@ -447,8 +447,8 @@ ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) x =rtnl_dereference (t->prl);if (x) { t->prl_count =0 ; +RCU_INIT_POINTER (t->prl,NULL );call_rcu (&x->rcu_head, prl_list_destroy_rcu); - t->prl =NULL ; } } out:Signature · 1 line
2.55.0.979.g7e5102b832-goog
PATCH 2/9 sit: charge ip_tunnel_prl_entry allocations to memcg
In reply to Eric Dumazetipip6_tunnel_add_prl() allocates struct ip_tunnel_prl_entry for each potential router entry added via SIOCADDPRL.
An unprivileged user with CAP_NET_ADMIN in a user namespace can add an unbounded number of entries, consuming uncharged kernel memory.
Commit 1b51d8271973 ("memcg: ipv6/sit: account and don't WARN on ip_tunnel_prl structs allocation") added accounting to the temporary buffer in ipip6_tunnel_get_prl(), but missed the entry allocations in ipip6_tunnel_add_prl().
Use GFP_KERNEL_ACCOUNT when allocating struct ip_tunnel_prl_entry.
--- net/ipv6/sit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
net/ipv6/sit.c +1 -1
@@ -395,7 +395,7 @@ ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) goto out; } - p = kzalloc_obj(struct ip_tunnel_prl_entry); + p = kzalloc_obj(struct ip_tunnel_prl_entry, GFP_KERNEL_ACCOUNT);if (!p) { err = -ENOBUFS ; goto out;
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 2/9] sit: charge ip_tunnel_prl_entry allocations to memcg
Quoted 10 lines
ipip6_tunnel_add_prl()allocates structip_tunnel_prl_entryfor each potential router entry added via SIOCADDPRL.
An unprivileged user withCAP_NET_ADMINin a user namespace can add an unbounded number of entries, consuming uncharged kernel memory.Commit
1b51d8271973("memcg: ipv6/sit: account and don't WARN onip_tunnel_prlstructs allocation") added accounting to the temporary buffer inipip6_tunnel_get_prl(), but missed the entry allocations inipip6_tunnel_add_prl().
Quoted 4 lines
Use
GFP_KERNEL_ACCOUNTwhen allocating structip_tunnel_prl_entry.Fixes 1b51d8271973 ("memcg: ipv6/sit: account and don't WARN on ip_tunnel_prl structs allocation")Signed-off-by Eric Dumazet <[email protected]>
Quoted 20 lines
--- net/ipv6/sit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)net/ipv6/sit.c +1
-1@@ -395,7 +395,7 @@ ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) goto out; } - p = kzalloc_obj(struct ip_tunnel_prl_entry); + p = kzalloc_obj(struct ip_tunnel_prl_entry, GFP_KERNEL_ACCOUNT);if (!p) { err = -ENOBUFS ; goto out;Signature · 1 line
2.55.0.979.g7e5102b832-goog
PATCH 3/9 ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup
In reply to Eric DumazetUpdate ip_tunnel_encap_setup() to use WRITE_ONCE() when writing to encap fields (type, sport, dport, flags) and hlen fields. This ensures that concurrent lockless readers (like fill_info) do not see torn writes.
Also remove the unsafe memset() on t->encap which could cause concurrent readers to transiently see zeroed fields. Removing it also fixes a bug where t->encap was left cleared even if ip_encap_hlen() failed, resulting in partial configuration.
--- net/ipv4/ip_tunnel.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-)
net/ipv4/ip_tunnel.c +6 -8
@@ -491,19 +491,17 @@ int ip_tunnel_encap_setup(struct ip_tunnel *t, {int hlen; -memset (&t->encap,0 ,sizeof (t->encap)); - hlen =ip_encap_hlen (ipencap);if (hlen <0 )return hlen; - t->encap.type = ipencap->type; - t->encap.sport = ipencap->sport; - t->encap.dport = ipencap->dport; - t->encap.flags = ipencap->flags; +WRITE_ONCE (t->encap.type, ipencap->type); +WRITE_ONCE (t->encap.sport, ipencap->sport); +WRITE_ONCE (t->encap.dport, ipencap->dport); +WRITE_ONCE (t->encap.flags, ipencap->flags); - t->encap_hlen = hlen; - t->hlen = t->encap_hlen + t->tun_hlen; +WRITE_ONCE (t->encap_hlen, hlen); +WRITE_ONCE (t->hlen, hlen + t->tun_hlen);return 0 ; }
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 3/9] ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup
Quoted 5 lines
Update
ip_tunnel_encap_setup()to useWRITE_ONCE()when writing to encap fields (type, sport, dport, flags) and hlen fields. This ensures that concurrent lockless readers (likefill_info) do not see torn writes.
Quoted 7 lines
Also remove the unsafe
memset()ont->encapwhich could cause concurrent readers to transiently see zeroed fields. Removing it also fixes a bug wheret->encapwas left cleared even ifip_encap_hlen()failed, resulting in partial configuration.Fixes 56328486539d ("net: Changes to ip_tunnel to support foo-over-udp encapsulation")Signed-off-by Eric Dumazet <[email protected]>
Quoted 37 lines
--- net/ipv4/ip_tunnel.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-)net/ipv4/ip_tunnel.c +6
-8@@ -491,19 +491,17 @@ int ip_tunnel_encap_setup(struct ip_tunnel *t, {int hlen; -memset (&t->encap,0 ,sizeof (t->encap)); - hlen =ip_encap_hlen (ipencap);if (hlen <0 )return hlen; - t->encap.type = ipencap->type; - t->encap.sport = ipencap->sport; - t->encap.dport = ipencap->dport; - t->encap.flags = ipencap->flags; +WRITE_ONCE (t->encap.type, ipencap->type); +WRITE_ONCE (t->encap.sport, ipencap->sport); +WRITE_ONCE (t->encap.dport, ipencap->dport); +WRITE_ONCE (t->encap.flags, ipencap->flags); - t->encap_hlen = hlen; - t->hlen = t->encap_hlen + t->tun_hlen; +WRITE_ONCE (t->encap_hlen, hlen); +WRITE_ONCE (t->hlen, hlen + t->tun_hlen);return 0 ; }Signature · 1 line
2.55.0.979.g7e5102b832-goog
[email protected]
AI reviewRe: [PATCH net-next 3/9] ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup
In reply to Eric DumazetThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Critical: 0 · High: 0 · Medium: 1 (2) · Low: 0
- [Medium] The mirror helper ip6_tnl_encap_setup() (net/ipv6/ip6_tunnel.c:1816) is left unconverted, so the exact bug…
Pre-existing issues:
- [Medium] ip_tunnel_encap_setup() recomputes t->hlen as 'hlen + t->tun_hlen' (net/ipv4/ip_tunnel.c:504), which disagrees with the…
- [Medium] The six encap/hlen fields are published as independent WRITE_ONCE() stores with no seqcount, RCU pointer swap or release…
--
ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup
This converts the encap field and hlen stores in ip_tunnel_encap_setup() to WRITE_ONCE() so lockless readers such as fill_info cannot observe torn values. It also drops the memset() of t->encap, which both removed a transient all-zero window for readers and stopped leaving t->encap cleared when ip_encap_hlen() fails.
No Link: tags present in the commit header.
net/ipv4/ip_tunnel.c int ip_tunnel_encap_setup(struct ip_tunnel *t, · line 496
Quoted 13 lines
net/ipv4/ip_tunnel.c +0
-2@@ -491,19 +491,17 @@ int ip_tunnel_encap_setup(struct ip_tunnel *t, {int hlen; -memset (&t->encap,0 ,sizeof (t->encap)); - hlen =ip_encap_hlen (ipencap);if (hlen <0 )return hlen;
[Severity: Medium]
Should the IPv6 twin get the same treatment? ip6_tnl_encap_setup() in net/ipv6/ip6_tunnel.c still has the original shape at this commit:
net/ipv6/ip6_tunnel.c:ip6_tnl_encap_setup() {
memset(&t->encap, 0, sizeof(t->encap));
hlen = ip6_encap_hlen(ipencap);
if (hlen < 0)
return hlen;
t->encap.type = ipencap->type;
t->encap.sport = ipencap->sport;
t->encap.dport = ipencap->dport;
t->encap.flags = ipencap->flags;
t->encap_hlen = hlen;
t->hlen = t->encap_hlen + t->tun_hlen;
}So the "t->encap was left cleared even if ip_encap_hlen() failed" case described in the changelog is still reachable for ip6tnl and ip6gre, on already configured tunnels:
ip6_tnl_changelink()
ip6_tnl_encap_setup(t, &ipencap) /* t is live */
memset(&t->encap, 0, ...)
ip6_encap_hlen() returns -EINVAL /* e.g. fou6 ops not registered,
or type >= MAX_IPTUN_ENCAP_OPS */
return hlen
ip6gre_changelink_common() reaches the same helper the same way. The netlink command fails, but the live tunnel's encapsulation configuration has already been wiped.
The plain stores after the memset in ip6_tnl_encap_setup() also leave the transient all-zero encap visible to ip6_tnl_fill_info()/ip6gre_fill_info(), which run without RTNL since RTM_GETLINK is registered with RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED in net/core/rtnetlink.c, and to the ip6 xmit path.
net/ipv4/ip_tunnel.c int ip_tunnel_encap_setup(struct ip_tunnel *t, · line 501
Quoted 9 lines
net/ipv4/ip_tunnel.c +4
-4- t->encap.type = ipencap->type; - t->encap.sport = ipencap->sport; - t->encap.dport = ipencap->dport; - t->encap.flags = ipencap->flags; +WRITE_ONCE (t->encap.type, ipencap->type); +WRITE_ONCE (t->encap.sport, ipencap->sport); +WRITE_ONCE (t->encap.dport, ipencap->dport); +WRITE_ONCE (t->encap.flags, ipencap->flags);
[Severity: Medium]
This is a pre-existing issue, not something this patch introduces, but is it worth noting in the changelog that WRITE_ONCE() only removes torn reads and does not give readers a coherent snapshot?
encap.type, the field that gates encapsulation, is still published first, with no seqcount or RCU pointer swap, so a reader can pair the new type with the stale sport/dport/flags or with the stale (smaller) encap_hlen and hlen. The same ordering existed before the patch (with zeroes instead of stale values), so nothing gets worse here.
On the transmit side the two reads are separated:
net/ipv6/sit.c:ipip6_tunnel_xmit() {
int t_hlen = tunnel->hlen + sizeof(struct iphdr);
...
max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
if (skb_headroom(skb) < max_headroom || ...)
new_skb = skb_realloc_headroom(skb, max_headroom);
...
if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
}The headroom comes from the earlier hlen snapshot while ip_tunnel_encap() reads the live tunnel->encap, and fou_build_udp()/__gue_build_header() use skb_push(). I could not show that the reserved headroom is actually exceeded, since skb_realloc_headroom()/__skb_cow() round expansions up, so this part is speculative.
net/ipv4/ip_tunnel.c int ip_tunnel_encap_setup(struct ip_tunnel *t, · line 504
Quoted 5 lines
net/ipv4/ip_tunnel.c +2
-2- t->encap_hlen = hlen; - t->hlen = t->encap_hlen + t->tun_hlen; +WRITE_ONCE (t->encap_hlen, hlen); +WRITE_ONCE (t->hlen, hlen + t->tun_hlen);
[Severity: Medium]
This isn't a bug introduced by this patch, since the formula is unchanged (t->encap_hlen + t->tun_hlen before, hlen + t->tun_hlen after), but does this hlen formula disagree with the ERSPAN one?
net/ipv4/ip_gre.c:erspan_tunnel_init() {
tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
erspan_hdr_len(tunnel->erspan_ver);
}erspan_changelink() reaches this helper on an already initialised device:
erspan_changelink()
ipgre_newlink_encap_setup()
ip_tunnel_encap_setup() /* last writer of t->hlen */ndo_init does not run again on that path, so an
"ip link set <erspan> type erspan encap ..." command appears to drop erspan_hdr_len() (12 bytes for v1, 16 for v2) from t->hlen. ip_tunnel_bind_dev() then derives both the MTU and
dev->needed_headroom = ip_tunnel_limit_headroom(t_hlen + hlen);
from the smaller value, while erspan_xmit() only does skb_cow_head(skb, dev->needed_headroom) before pushing the erspan header, the GRE header, the encap header and the outer IPv4 header.
Quoted 3 lines
net/ipv4/ip_tunnel.c +0
-0return 0 ; }
Signature · 1 line
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260907075846.2913645-1-edumazet%40google.com
PATCH 4/9 sit: annotate data-races around tunnel->fwmark
In reply to Eric Dumazetipip6_tunnel_xmit() runs locklessly (dev->lltx == true) and reads tunnel->fwmark.
In preparation for converting ipip6_fill_info() to run without RTNL, add READ_ONCE() and WRITE_ONCE() annotations around tunnel->fwmark.
--- net/ipv6/sit.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
net/ipv6/sit.c +3 -3
@@ -929,7 +929,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,if (!dst && !ipip6_tunnel_dst_find (skb, &dst, false)) goto tx_error; -flowi4_init_output (&fl4, tunnel->parms.link, tunnel->fwmark, +flowi4_init_output (&fl4, tunnel->parms.link,READ_ONCE (tunnel->fwmark), tos &INET_DSCP_MASK ,RT_SCOPE_UNIVERSE ,IPPROTO_IPV6 ,0 , dst, tiph->saddr,0 ,0 ,sock_net_uid (tunnel->net,NULL ));
@@ -1153,7 +1153,7 @@ static void ipip6_tunnel_update(struct ip_tunnel *t, t->parms.iph.frag_off = p->iph.frag_off;if (t->parms.link != p->link || t->fwmark != fwmark) { t->parms.link = p->link; - t->fwmark = fwmark; +WRITE_ONCE (t->fwmark, fwmark);ipip6_tunnel_bind_dev (t->dev); }dst_cache_reset (&t->dst_cache);
@@ -1706,7 +1706,7 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)nla_put_u8 (skb,IFLA_IPTUN_PROTO , parm->iph.protocol) ||nla_put_be16 (skb,IFLA_IPTUN_FLAGS ,ip_tunnel_flags_to_be16 (parm->i_flags)) || - nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark)) + nla_put_u32(skb, IFLA_IPTUN_FWMARK, READ_ONCE(tunnel->fwmark))) goto nla_put_failure;#ifdef CONFIG_IPV6_SIT_6RD
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 4/9] sit: annotate data-races around tunnel->fwmark
Quoted 5 lines
ipip6_tunnel_xmit()runs locklessly (dev->lltx== true) and readstunnel->fwmark.In preparation for converting
ipip6_fill_info()to run without RTNL, addREAD_ONCE()andWRITE_ONCE()annotations aroundtunnel->fwmark.
Quoted 40 lines
Signed-off-by Eric Dumazet <[email protected]>--- net/ipv6/sit.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)net/ipv6/sit.c +3
-3@@ -929,7 +929,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,if (!dst && !ipip6_tunnel_dst_find (skb, &dst, false)) goto tx_error; -flowi4_init_output (&fl4, tunnel->parms.link, tunnel->fwmark, +flowi4_init_output (&fl4, tunnel->parms.link,READ_ONCE (tunnel->fwmark), tos &INET_DSCP_MASK ,RT_SCOPE_UNIVERSE ,IPPROTO_IPV6 ,0 , dst, tiph->saddr,0 ,0 ,sock_net_uid (tunnel->net,NULL ));@@ -1153,7 +1153,7 @@ static void ipip6_tunnel_update(struct ip_tunnel *t, t->parms.iph.frag_off = p->iph.frag_off;if (t->parms.link != p->link || t->fwmark != fwmark) { t->parms.link = p->link; - t->fwmark = fwmark; +WRITE_ONCE (t->fwmark, fwmark);ipip6_tunnel_bind_dev (t->dev); }dst_cache_reset (&t->dst_cache);@@ -1706,7 +1706,7 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)nla_put_u8 (skb,IFLA_IPTUN_PROTO , parm->iph.protocol) ||nla_put_be16 (skb,IFLA_IPTUN_FLAGS ,ip_tunnel_flags_to_be16 (parm->i_flags)) || - nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark)) + nla_put_u32(skb, IFLA_IPTUN_FWMARK, READ_ONCE(tunnel->fwmark))) goto nla_put_failure;#ifdef CONFIG_IPV6_SIT_6RD Signature · 1 line
2.55.0.979.g7e5102b832-goog
[email protected]
AI reviewRe: [PATCH net-next 4/9] sit: annotate data-races around tunnel->fwmark
In reply to Eric DumazetThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Critical: 0 · High: 0 · Medium: 0 · Low: 1
- [Low] Incomplete fwmark annotation for sit devices:
ipip6_tunnel_update() now publishes tunnel->fwmark with WRITE_ONCE()…
--
sit: annotate data-races around tunnel->fwmark
This adds READ_ONCE()/WRITE_ONCE() annotations around tunnel->fwmark in the sit driver, since ipip6_tunnel_xmit() runs locklessly with dev->lltx == true. It is preparation for running ipip6_fill_info() without RTNL.
Quoted 13 lines
net/ipv6/sit.c +1
-1@@ -929,7 +929,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,if (!dst && !ipip6_tunnel_dst_find (skb, &dst, false)) goto tx_error; -flowi4_init_output (&fl4, tunnel->parms.link, tunnel->fwmark, +flowi4_init_output (&fl4, tunnel->parms.link,READ_ONCE (tunnel->fwmark), tos &INET_DSCP_MASK ,RT_SCOPE_UNIVERSE ,IPPROTO_IPV6 ,0 , dst, tiph->saddr,0 ,0 ,sock_net_uid (tunnel->net,NULL ));
[ ... ]
net/ipv6/sit.c static void ipip6_tunnel_update(struct ip_tunnel *t, · line 1156
Quoted 6 lines
net/ipv6/sit.c +1
-1@@ -1153,7 +1153,7 @@ static void ipip6_tunnel_update(struct ip_tunnel *t, t->parms.iph.frag_off = p->iph.frag_off;if (t->parms.link != p->link || t->fwmark != fwmark) { t->parms.link = p->link; - t->fwmark = fwmark; +WRITE_ONCE (t->fwmark, fwmark);
[Severity: Low]
Should the other lockless transmit path of the same sit device also be annotated?
ipip6_tunnel_xmit() is only one of the two xmit paths. For IPv4 and MPLS payloads, sit_tunnel_xmit() dispatches elsewhere:
net/ipv6/sit.c:sit_tunnel_xmit() {
switch (skb->protocol) {
case htons(ETH_P_IP):
sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
break;
case htons(ETH_P_IPV6):
ipip6_tunnel_xmit(skb, dev);
break;
#if IS_ENABLED(CONFIG_MPLS)
case htons(ETH_P_MPLS_UC):
sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
break;
#endif
...
}and sit_tunnel_xmit__() -> ip_tunnel_xmit() reads the same field with a plain load, right next to an already annotated read:
net/ipv4/ip_tunnel.c:ip_tunnel_xmit() {
...
ip_tunnel_init_flow(&fl4, protocol, dst, tnl_params->saddr,
tunnel->parms.o_key, tos & INET_DSCP_MASK,
tunnel->net, READ_ONCE(tunnel->parms.link),
tunnel->fwmark, skb_get_hash(skb), 0);
...
}So for a sit tunnel carrying IPv4 or MPLS frames, an RTNL writer running ipip6_changelink()/ioctl -> ipip6_tunnel_update() -> WRITE_ONCE(t->fwmark, fwmark) still races with an unmarked read, and KCSAN can still report it. The field is a naturally aligned u32, so the read cannot tear; the visible effects are the KCSAN report plus one route lookup using the pre-update mark.
Note that the generic writer in net/ipv4/ip_tunnel.c:ip_tunnel_update() is also still a plain store to t->fwmark, so closing this would mean touching the generic ip_tunnel code as well, which may be outside the intended scope of this series. It was checked at the end of the series and the plain read in ip_tunnel_xmit() is still there.
[ ... ]
Signature · 1 line
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260907075846.2913645-1-edumazet%40google.com
PATCH 5/9 sit: convert 6RD configuration to RCU protection
In reply to Eric DumazetIn order to allow lockless readers in future patches, convert 'tunnel->ip6rd' to an RCU protected pointer.
Updating 6RD configuration via ipip6_tunnel_update_6rd() or ipip6_tunnel_clone_6rd() now allocates a struct ip_tunnel_6rd_parm and uses rcu_assign_pointer() to publish it, freeing the previous parameters with kfree_rcu().
Readers in check_6rd() and only_dnatted() use rcu_dereference() under existing RCU read lock, preventing torn reads on the 128-bit IPv6 prefix.
--- include/net/ip_tunnels.h | 3 +- net/ipv6/sit.c | 160 ++++++++++++++++++++++++++++----------- 2 files changed, 117 insertions(+), 46 deletions(-)
include/net/ip_tunnels.h +2 -1
@@ -127,6 +127,7 @@ struct ip_tunnel_6rd_parm {__be32 relay_prefix;u16 prefixlen;u16 relay_prefixlen; + struct rcu_head rcu; };#endif
@@ -185,7 +186,7 @@ struct ip_tunnel {/* for SIT */ #ifdef CONFIG_IPV6_SIT_6RD -struct ip_tunnel_6rd_parm ip6rd; +struct ip_tunnel_6rd_parm __rcu *ip6rd;#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */
net/ipv6/sit.c +115 -45
294 lines not on this page · raw message
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 5/9] sit: convert 6RD configuration to RCU protection
Quoted 8 lines
In order to allow lockless readers in future patches, convert '
tunnel->ip6rd' to an RCU protected pointer.Updating 6RD configuration via
ipip6_tunnel_update_6rd()oripip6_tunnel_clone_6rd()now allocates a structip_tunnel_6rd_parmand usesrcu_assign_pointer()to publish it, freeing the previous parameters withkfree_rcu().
Quoted 5 lines
Readers in
check_6rd()andonly_dnatted()usercu_dereference()under existing RCU read lock, preventing torn reads on the 128-bit IPv6 prefix.Signed-off-by Eric Dumazet <[email protected]>
Hi Eric,
just few nits inline.
Regards,
Lorenzo
Quoted 31 lines
--- include/net/ip_tunnels.h | 3 +- net/ipv6/sit.c | 160 ++++++++++++++++++++++++++++----------- 2 files changed, 117 insertions(+), 46 deletions(-)include/net/ip_tunnels.h +2
-1@@ -127,6 +127,7 @@ struct ip_tunnel_6rd_parm {__be32 relay_prefix;u16 prefixlen;u16 relay_prefixlen; + struct rcu_head rcu; };#endif @@ -185,7 +186,7 @@ struct ip_tunnel {/* for SIT */ #ifdef CONFIG_IPV6_SIT_6RD -struct ip_tunnel_6rd_parm ip6rd; +struct ip_tunnel_6rd_parm __rcu *ip6rd;#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */ net/ipv6/sit.c +0
-0
net/ipv6/sit.c static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) · line 190
Quoted 9 lines
net/ipv6/sit.c +2
-1@@ -183,21 +183,47 @@ static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)rcu_assign_pointer (*tp, t); } -static void ipip6_tunnel_clone_6rd (struct net_device *dev,struct sit_net *sitn) +static int ipip6_tunnel_clone_6rd (struct net_device *dev,struct sit_net *sitn) {#ifdef CONFIG_IPV6_SIT_6RD struct ip_tunnel *t =netdev_priv (dev); +struct ip_tunnel_6rd_parm *new_6rd, *old_6rd;
RCT
net/ipv6/sit.c static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) · line 192
Quoted 2 lines
net/ipv6/sit.c +2
-0+ + new_6rd =kmalloc_obj (*new_6rd);
what about using kzalloc_obj() and get rid of the 0 initialization? This is just control path, so I guess it is fine.
Quoted 22 lines
net/ipv6/sit.c +13
-5+if (!new_6rd) +return -ENOMEM ;if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) { -ipv6_addr_set (&t->ip6rd.prefix,htonl (0x20020000 ),0 ,0 ,0 ); - t->ip6rd.relay_prefix =0 ; - t->ip6rd.prefixlen =16 ; - t->ip6rd.relay_prefixlen =0 ; +ipv6_addr_set (&new_6rd->prefix,htonl (0x20020000 ),0 ,0 ,0 ); + new_6rd->relay_prefix =0 ; + new_6rd->prefixlen =16 ; + new_6rd->relay_prefixlen =0 ; }else {struct ip_tunnel *t0 =netdev_priv (sitn->fb_tunnel_dev); -memcpy (&t->ip6rd, &t0->ip6rd,sizeof (t->ip6rd)); +struct ip_tunnel_6rd_parm *t0_6rd; + + t0_6rd =rtnl_dereference (t0->ip6rd); +if (t0_6rd) { + *new_6rd = *t0_6rd; + }else { +ipv6_addr_set (&new_6rd->prefix,htonl (0x20020000 ),0 ,0 ,0 );
net/ipv6/sit.c static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) · line 219
Quoted 10 lines
net/ipv6/sit.c +10
-0+ new_6rd->relay_prefix =0 ; + new_6rd->prefixlen =16 ; + new_6rd->relay_prefixlen =0 ; + } + } + + old_6rd =rcu_dereference_protected (t->ip6rd, +lockdep_rtnl_is_held () || + dev->reg_state ==NETREG_UNINITIALIZED ); +rcu_assign_pointer (t->ip6rd, new_6rd);
what about using rcu_replace_pointer()?
Quoted 11 lines
net/ipv6/sit.c +5
-0+if (old_6rd) { +dst_cache_reset (&t->dst_cache); +netdev_state_change (t->dev); +kfree_rcu (old_6rd, rcu); }#endif +return 0 ; }static int ipip6_tunnel_create (struct net_device *dev)
net/ipv6/sit.c static int ipip6_tunnel_create(struct net_device *dev) · line 237
Quoted 6 lines
net/ipv6/sit.c +3
-0@@ -206,6 +232,10 @@ static int ipip6_tunnel_create(struct net_device *dev)struct sit_net *sitn =net_generic (t->net, sit_net_id);int err; + err =ipip6_tunnel_clone_6rd (dev, sitn); +if (err <0 ) + goto out;
I guess you can just return err here.
Quoted 91 lines
net/ipv6/sit.c +22
-16+__dev_addr_set (dev, &t->parms.iph.saddr,4 );memcpy (dev->broadcast, &t->parms.iph.daddr,4 );@@ -218,8 +248,6 @@ static int ipip6_tunnel_create(struct net_device *dev)if (err <0 ) goto out; -ipip6_tunnel_clone_6rd (dev, sitn); -ipip6_tunnel_link (sitn, t);return 0 ;@@ -281,6 +309,7 @@ static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,return nt; failed_free: +ipip6_dev_free (dev);free_netdev (dev); failed:return NULL ;@@ -631,8 +660,13 @@ static bool only_dnatted(const struct ip_tunnel *tunnel,int prefix_len;#ifdef CONFIG_IPV6_SIT_6RD - prefix_len = tunnel->ip6rd.prefixlen +32 - - tunnel->ip6rd.relay_prefixlen; +const struct ip_tunnel_6rd_parm *ip6rd; + + ip6rd =rcu_dereference (tunnel->ip6rd); +if (ip6rd) + prefix_len = ip6rd->prefixlen +32 - ip6rd->relay_prefixlen; +else + prefix_len =48 ;#else prefix_len =48 ;#endif @@ -810,25 +844,28 @@ static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, __be32 *v4dst) {#ifdef CONFIG_IPV6_SIT_6RD -if (ipv6_prefix_equal (v6dst, &tunnel->ip6rd.prefix, - tunnel->ip6rd.prefixlen)) { +const struct ip_tunnel_6rd_parm *ip6rd; + + ip6rd =rcu_dereference (tunnel->ip6rd); +if (ip6rd &&ipv6_prefix_equal (v6dst, &ip6rd->prefix, + ip6rd->prefixlen)) {unsigned int pbw0, pbi0;int pbi1;u32 d; - pbw0 = tunnel->ip6rd.prefixlen >>5 ; - pbi0 = tunnel->ip6rd.prefixlen &0x1f ; + pbw0 = ip6rd->prefixlen >>5 ; + pbi0 = ip6rd->prefixlen &0x1f ; - d = tunnel->ip6rd.relay_prefixlen <32 ? + d = ip6rd->relay_prefixlen <32 ? (ntohl (v6dst->s6_addr32[pbw0]) << pbi0) >> - tunnel->ip6rd.relay_prefixlen :0 ; + ip6rd->relay_prefixlen :0 ; - pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; + pbi1 = pbi0 - ip6rd->relay_prefixlen;if (pbi1 >0 ) d |=ntohl (v6dst->s6_addr32[pbw0 +1 ]) >> (32 - pbi1); - *v4dst = tunnel->ip6rd.relay_prefix |htonl (d); + *v4dst = ip6rd->relay_prefix |htonl (d);return true; }#else @@ -1164,6 +1201,7 @@ static void ipip6_tunnel_update(struct ip_tunnel *t,static int ipip6_tunnel_update_6rd (struct ip_tunnel *t,struct ip_tunnel_6rd *ip6rd) { +struct ip_tunnel_6rd_parm *new_6rd, *old_6rd;struct in6_addr prefix;__be32 relay_prefix;@@ -1183,10 +1221,20 @@ static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,if (relay_prefix != ip6rd->relay_prefix)return -EINVAL ; - t->ip6rd.prefix = prefix; - t->ip6rd.relay_prefix = relay_prefix; - t->ip6rd.prefixlen = ip6rd->prefixlen; - t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen; + new_6rd =kmalloc_obj (*new_6rd);
net/ipv6/sit.c static int ipip6_tunnel_update_6rd(struct ip_tunnel *t, · line 1234
Quoted 10 lines
net/ipv6/sit.c +10
-0+if (!new_6rd) +return -ENOMEM ; + + new_6rd->prefix = prefix; + new_6rd->relay_prefix = relay_prefix; + new_6rd->prefixlen = ip6rd->prefixlen; + new_6rd->relay_prefixlen = ip6rd->relay_prefixlen; + + old_6rd =rtnl_dereference (t->ip6rd); +rcu_assign_pointer (t->ip6rd, new_6rd);
rcu_replace_pointer()?
Quoted 151 lines
net/ipv6/sit.c +48
-23+if (old_6rd) +kfree_rcu (old_6rd, rcu); +dst_cache_reset (&t->dst_cache);netdev_state_change (t->dev);return 0 ;@@ -1195,6 +1243,7 @@ static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,static int ipip6_tunnel_get6rd (struct net_device *dev,struct ip_tunnel_parm __user *data) { +const struct ip_tunnel_6rd_parm *ip6rd_parm;struct ip_tunnel *t =netdev_priv (dev);struct ip_tunnel_parm_kern p;struct ip_tunnel_6rd ip6rd;@@ -1207,10 +1256,15 @@ ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)if (!t) t =netdev_priv (dev); - ip6rd.prefix = t->ip6rd.prefix; - ip6rd.relay_prefix = t->ip6rd.relay_prefix; - ip6rd.prefixlen = t->ip6rd.prefixlen; - ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; + ip6rd_parm =rtnl_dereference (t->ip6rd); +if (ip6rd_parm) { + ip6rd.prefix = ip6rd_parm->prefix; + ip6rd.relay_prefix = ip6rd_parm->relay_prefix; + ip6rd.prefixlen = ip6rd_parm->prefixlen; + ip6rd.relay_prefixlen = ip6rd_parm->relay_prefixlen; + }else { +memset (&ip6rd,0 ,sizeof (ip6rd)); + }if (copy_to_user (data, &ip6rd,sizeof (ip6rd)))return -EFAULT ;return 0 ;@@ -1222,20 +1276,16 @@ ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data, {struct ip_tunnel *t =netdev_priv (dev);struct ip_tunnel_6rd ip6rd; -int err;if (!ns_capable (t->net->user_ns,CAP_NET_ADMIN ))return -EPERM ;if (copy_from_user (&ip6rd, data,sizeof (ip6rd)))return -EFAULT ; -if (cmd !=SIOCDEL6RD ) { - err =ipip6_tunnel_update_6rd (t, &ip6rd); -if (err <0 ) -return err; - }else -ipip6_tunnel_clone_6rd (dev,dev_to_sit_net (dev)); -return 0 ; +if (cmd !=SIOCDEL6RD ) +return ipip6_tunnel_update_6rd (t, &ip6rd); + +return ipip6_tunnel_clone_6rd (dev,dev_to_sit_net (dev)); }#endif /* CONFIG_IPV6_SIT_6RD */ @@ -1406,8 +1456,17 @@ static const struct net_device_ops ipip6_netdev_ops = {static void ipip6_dev_free (struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); +#ifdef CONFIG_IPV6_SIT_6RD +struct ip_tunnel_6rd_parm *ip6rd; -dst_cache_destroy (&tunnel->dst_cache); + ip6rd =rcu_dereference_protected (tunnel->ip6rd,1 ); +RCU_INIT_POINTER (tunnel->ip6rd,NULL ); +kfree (ip6rd); +#endif +if (tunnel->dst_cache.cache) { +dst_cache_destroy (&tunnel->dst_cache); + tunnel->dst_cache.cache =NULL ; + } }#define SIT_FEATURES (NETIF_F_SG | \@@ -1576,8 +1635,10 @@ static int ipip6_newlink(struct net_device *dev,return -EEXIST ; err =ipip6_tunnel_create (dev); -if (err <0 ) +if (err <0 ) { +ipip6_dev_free (dev);return err; + }if (tb[IFLA_MTU ]) {u32 mtu =nla_get_u32 (tb[IFLA_MTU ]);@@ -1695,6 +1756,9 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev);struct ip_tunnel_parm_kern *parm = &tunnel->parms; +#ifdef CONFIG_IPV6_SIT_6RD +const struct ip_tunnel_6rd_parm *ip6rd; +#endif if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||@@ -1710,14 +1774,16 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) goto nla_put_failure;#ifdef CONFIG_IPV6_SIT_6RD -if (nla_put_in6_addr (skb,IFLA_IPTUN_6RD_PREFIX , - &tunnel->ip6rd.prefix) || -nla_put_in_addr (skb,IFLA_IPTUN_6RD_RELAY_PREFIX , - tunnel->ip6rd.relay_prefix) || -nla_put_u16 (skb,IFLA_IPTUN_6RD_PREFIXLEN , - tunnel->ip6rd.prefixlen) || -nla_put_u16 (skb,IFLA_IPTUN_6RD_RELAY_PREFIXLEN , - tunnel->ip6rd.relay_prefixlen)) + ip6rd =rcu_dereference_rtnl (tunnel->ip6rd); +if (ip6rd && + (nla_put_in6_addr (skb,IFLA_IPTUN_6RD_PREFIX , + &ip6rd->prefix) || +nla_put_in_addr (skb,IFLA_IPTUN_6RD_RELAY_PREFIX , + ip6rd->relay_prefix) || +nla_put_u16 (skb,IFLA_IPTUN_6RD_PREFIXLEN , + ip6rd->prefixlen) || +nla_put_u16 (skb,IFLA_IPTUN_6RD_RELAY_PREFIXLEN , + ip6rd->relay_prefixlen))) goto nla_put_failure;#endif @@ -1863,17 +1929,21 @@ static int __net_init sit_init_net(struct net *net) t =netdev_priv (sitn->fb_tunnel_dev); t->net = net; + err =ipip6_tunnel_clone_6rd (sitn->fb_tunnel_dev, sitn); +if (err <0 ) + goto err_reg_dev; + err =register_netdev (sitn->fb_tunnel_dev);if (err) goto err_reg_dev; -ipip6_tunnel_clone_6rd (sitn->fb_tunnel_dev, sitn);ipip6_fb_tunnel_init (sitn->fb_tunnel_dev);strscpy (t->parms.name, sitn->fb_tunnel_dev->name);return 0 ; err_reg_dev: +ipip6_dev_free (sitn->fb_tunnel_dev);free_netdev (sitn->fb_tunnel_dev); err_alloc_dev:return err;Signature · 1 line
2.55.0.979.g7e5102b832-goog
[email protected]
AI reviewRe: [PATCH net-next 5/9] sit: convert 6RD configuration to RCU protection
In reply to Eric DumazetThank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Critical: 0 · High: 0 · Medium: 1 · Low: 1
- [Medium] check_6rd() now performs ip6rd = rcu_dereference(tunnel->ip6rd) (net/ipv6/sit.c:849), but check_6rd()…
- [Low] The patch changes SIOCDEL6RD behaviour beyond the advertised RCU
conversion: ipip6_tunnel_clone_6rd() now, when an old 6RD…
--
sit: convert 6RD configuration to RCU protection
This converts tunnel->ip6rd from an embedded struct into an RCU protected pointer, so updates via ipip6_tunnel_update_6rd() and ipip6_tunnel_clone_6rd() allocate a new struct ip_tunnel_6rd_parm, publish it with rcu_assign_pointer() and free the old one with kfree_rcu(). Readers in check_6rd(), only_dnatted(), ipip6_tunnel_get6rd() and ipip6_fill_info() are updated accordingly.
Quoted 46 lines
net/ipv6/sit.c +25
-6@@ -183,21 +183,47 @@ static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)rcu_assign_pointer (*tp, t); } -static void ipip6_tunnel_clone_6rd (struct net_device *dev,struct sit_net *sitn) +static int ipip6_tunnel_clone_6rd (struct net_device *dev,struct sit_net *sitn) {#ifdef CONFIG_IPV6_SIT_6RD struct ip_tunnel *t =netdev_priv (dev); +struct ip_tunnel_6rd_parm *new_6rd, *old_6rd; + + new_6rd =kmalloc_obj (*new_6rd); +if (!new_6rd) +return -ENOMEM ;if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) { -ipv6_addr_set (&t->ip6rd.prefix,htonl (0x20020000 ),0 ,0 ,0 ); - t->ip6rd.relay_prefix =0 ; - t->ip6rd.prefixlen =16 ; - t->ip6rd.relay_prefixlen =0 ; +ipv6_addr_set (&new_6rd->prefix,htonl (0x20020000 ),0 ,0 ,0 ); + new_6rd->relay_prefix =0 ; + new_6rd->prefixlen =16 ; + new_6rd->relay_prefixlen =0 ; }else {struct ip_tunnel *t0 =netdev_priv (sitn->fb_tunnel_dev); -memcpy (&t->ip6rd, &t0->ip6rd,sizeof (t->ip6rd)); +struct ip_tunnel_6rd_parm *t0_6rd; + + t0_6rd =rtnl_dereference (t0->ip6rd); +if (t0_6rd) { + *new_6rd = *t0_6rd; + }else { +ipv6_addr_set (&new_6rd->prefix,htonl (0x20020000 ),0 ,0 ,0 ); + new_6rd->relay_prefix =0 ; + new_6rd->prefixlen =16 ; + new_6rd->relay_prefixlen =0 ; + } + } + + old_6rd = rcu_dereference_protected(t->ip6rd, + lockdep_rtnl_is_held() ||
net/ipv6/sit.c static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) · line 227
Quoted 10 lines
net/ipv6/sit.c +7
-0+ dev->reg_state == NETREG_UNINITIALIZED); +rcu_assign_pointer (t->ip6rd, new_6rd); +if (old_6rd) { +dst_cache_reset (&t->dst_cache); +netdev_state_change (t->dev); +kfree_rcu (old_6rd, rcu); }#endif +return 0 ; }
[Severity: Low]
Is the new dst_cache_reset() plus netdev_state_change() in ipip6_tunnel_clone_6rd() intended as part of this conversion?
Before this patch, the SIOCDEL6RD path did neither:
ipip6_tunnel_6rdctl()
ipip6_tunnel_clone_6rd() /* just overwrote t->ip6rd fields */
Only ipip6_tunnel_update_6rd() invalidated the destination cache and notified userspace. Since ipip6_tunnel_xmit() derives the IPv4 destination from the 6RD configuration via try_6rd() and then consults dst_cache_get_ip4(&tunnel->dst_cache, ...), the added reset changes where packets go after a 6RD deletion, and netdev_state_change() -> netif_state_change() emits an extra NETDEV_CHANGE plus RTM_NEWLINK that link watchers will see.
Would it make sense to split that out into its own patch, with a Fixes:
tag if it is fixing stale routing after SIOCDEL6RD? The changelog currently only describes the allocation and publish conversion.
[ ... ]
net/ipv6/sit.c static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, · line 851
Quoted 11 lines
net/ipv6/sit.c +5
-2@@ -810,25 +844,28 @@ static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, __be32 *v4dst) {#ifdef CONFIG_IPV6_SIT_6RD -if (ipv6_prefix_equal (v6dst, &tunnel->ip6rd.prefix, - tunnel->ip6rd.prefixlen)) { +const struct ip_tunnel_6rd_parm *ip6rd; + + ip6rd =rcu_dereference (tunnel->ip6rd); +if (ip6rd &&ipv6_prefix_equal (v6dst, &ip6rd->prefix, + ip6rd->prefixlen)) {
[Severity: Medium]
Is plain rcu_dereference() the right accessor here? The changelog says:
"Readers in check_6rd() and only_dnatted() use rcu_dereference() under
existing RCU read lock"
That holds for the receive path, but check_6rd() is also reached from transmit:
netdev_start_xmit()
sit_tunnel_xmit()
ipip6_tunnel_xmit()
try_6rd()
check_6rd()
rcu_dereference(tunnel->ip6rd)and the xmit callers do not hold rcu_read_lock(). __dev_queue_xmit() takes rcu_read_lock_bh(), which acquires rcu_bh_lock_map, and the AF_PACKET qdisc bypass path __dev_direct_xmit() only does local_bh_disable():
net/core/dev.c:__dev_direct_xmit() {
...
local_bh_disable();
dev_xmit_recursion_inc();
HARD_TX_LOCK(dev, txq, smp_processor_id());
...
}rcu_dereference() expands to rcu_dereference_check(p, 0), whose condition is rcu_read_lock_held():
kernel/rcu/update.c:rcu_read_lock_held() {
...
return lock_is_held(&rcu_lock_map);
}so on a CONFIG_PROVE_RCU kernel, would sending a packet on a sit device from these contexts produce a "suspicious rcu_dereference_check() usage" splat? Object lifetime looks fine, since BH-disabled regions are RCU read-side sections, but the annotation seems off.
Would rcu_dereference_bh() (or rcu_dereference_check() with rcu_read_lock_bh_held()) be more accurate, or alternatively an explicit rcu_read_lock() in ipip6_tunnel_xmit(), like vxlan_xmit() does before its RCU lookups? The same question applies to the new rcu_dereference() in only_dnatted().
[ ... ]
Quoted 13 lines
net/ipv6/sit.c +0
-3@@ -1222,20 +1276,16 @@ ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data, {struct ip_tunnel *t =netdev_priv (dev);struct ip_tunnel_6rd ip6rd; -int err;if (!ns_capable (t->net->user_ns,CAP_NET_ADMIN ))return -EPERM ;if (copy_from_user (&ip6rd, data,sizeof (ip6rd)))return -EFAULT ; -if (cmd !=SIOCDEL6RD ) { - err =ipip6_tunnel_update_6rd (t, &ip6rd);
net/ipv6/sit.c ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data, · line 1289
Quoted 10 lines
net/ipv6/sit.c +4
-5-if (err <0 ) -return err; - }else -ipip6_tunnel_clone_6rd (dev,dev_to_sit_net (dev)); -return 0 ; +if (cmd !=SIOCDEL6RD ) +return ipip6_tunnel_update_6rd (t, &ip6rd); + +return ipip6_tunnel_clone_6rd (dev,dev_to_sit_net (dev)); }
Related to the comment above: SIOCDEL6RD can now also fail with -ENOMEM, where it previously always returned 0. Is that intentional?
Signature · 1 line
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260907075846.2913645-1-edumazet%40google.com
PATCH 6/9 sit: implement ipip6_get_iflink()
In reply to Eric DumazetSIT currently uses ip_tunnel_get_iflink() which reads tunnel->parms.link. In order to decouple SIT parameter handling from ip_tunnel, implement its own ndo_get_iflink callback.
--- net/ipv6/sit.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
net/ipv6/sit.c +8 -1
@@ -1444,12 +1444,19 @@ ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr, } } +static int ipip6_get_iflink (const struct net_device *dev) +{ +struct ip_tunnel *tunnel =netdev_priv (dev); + +return READ_ONCE (tunnel->parms.link); +} +static const struct net_device_ops ipip6_netdev_ops = { .ndo_init = ipip6_tunnel_init, .ndo_uninit = ipip6_tunnel_uninit, .ndo_start_xmit = sit_tunnel_xmit, .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate, - .ndo_get_iflink = ip_tunnel_get_iflink, + .ndo_get_iflink = ipip6_get_iflink, .ndo_tunnel_ctl = ipip6_tunnel_ctl, };
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 6/9] sit: implement ipip6_get_iflink()
Quoted 5 lines
SIT currently uses
ip_tunnel_get_iflink()which readstunnel->parms.link. In order to decouple SIT parameter handling fromip_tunnel, implement its ownndo_get_iflinkcallback.Signed-off-by Eric Dumazet <[email protected]>
Quoted 32 lines
--- net/ipv6/sit.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)net/ipv6/sit.c +8
-1@@ -1444,12 +1444,19 @@ ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr, } } +static int ipip6_get_iflink (const struct net_device *dev) +{ +struct ip_tunnel *tunnel =netdev_priv (dev); + +return READ_ONCE (tunnel->parms.link); +} +static const struct net_device_ops ipip6_netdev_ops = { .ndo_init = ipip6_tunnel_init, .ndo_uninit = ipip6_tunnel_uninit, .ndo_start_xmit = sit_tunnel_xmit, .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate, - .ndo_get_iflink = ip_tunnel_get_iflink, + .ndo_get_iflink = ipip6_get_iflink, .ndo_tunnel_ctl = ipip6_tunnel_ctl, };Signature · 1 line
2.55.0.979.g7e5102b832-goog
PATCH 7/9 sit: dynamically allocate struct ip_tunnel_parm_kern
In reply to Eric DumazetIn preparation for converting SIT configuration parameters to RCU, dynamically allocate struct ip_tunnel_parm_kern as sit_parms in struct ip_tunnel.
--- include/net/ip_tunnels.h | 1 + net/ipv6/sit.c | 152 +++++++++++++++++++++++---------------- 2 files changed, 92 insertions(+), 61 deletions(-)
include/net/ip_tunnels.h +1 -0
@@ -190,6 +190,7 @@ struct ip_tunnel {#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */ +struct ip_tunnel_parm_kern *sit_parms;unsigned int ip_tnl_net_id;struct gro_cells gro_cells;__u32 fwmark;
net/ipv6/sit.c +91 -61
362 lines not on this page · raw message
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 7/9] sit: dynamically allocate struct ip_tunnel_parm_kern
Quoted 5 lines
In preparation for converting SIT configuration parameters to RCU, dynamically allocate struct
ip_tunnel_parm_kernassit_parmsin structip_tunnel.Signed-off-by Eric Dumazet <[email protected]>
Quoted 415 lines
--- include/net/ip_tunnels.h | 1 + net/ipv6/sit.c | 152 +++++++++++++++++++++++---------------- 2 files changed, 92 insertions(+), 61 deletions(-)include/net/ip_tunnels.h +1
-0@@ -190,6 +190,7 @@ struct ip_tunnel {#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */ +struct ip_tunnel_parm_kern *sit_parms;unsigned int ip_tnl_net_id;struct gro_cells gro_cells;__u32 fwmark;net/ipv6/sit.c +91
-61@@ -108,24 +108,24 @@ static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,int ifindex = dev ? dev->ifindex :0 ;for_each_ip_tunnel_rcu (t, sitn->tunnels_r_l[h0 ^ h1]) { -if (local == t->parms.iph.saddr && - remote == t->parms.iph.daddr && - (!dev || !t->parms.link || ifindex == t->parms.link || - sifindex == t->parms.link) && +if (local == t->sit_parms->iph.saddr && + remote == t->sit_parms->iph.daddr && + (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || + sifindex == t->sit_parms->link) && (t->dev->flags &IFF_UP ))return t; }for_each_ip_tunnel_rcu (t, sitn->tunnels_r[h0]) { -if (remote == t->parms.iph.daddr && - (!dev || !t->parms.link || ifindex == t->parms.link || - sifindex == t->parms.link) && +if (remote == t->sit_parms->iph.daddr && + (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || + sifindex == t->sit_parms->link) && (t->dev->flags &IFF_UP ))return t; }for_each_ip_tunnel_rcu (t, sitn->tunnels_l[h1]) { -if (local == t->parms.iph.saddr && - (!dev || !t->parms.link || ifindex == t->parms.link || - sifindex == t->parms.link) && +if (local == t->sit_parms->iph.saddr && + (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || + sifindex == t->sit_parms->link) && (t->dev->flags &IFF_UP ))return t; }@@ -157,7 +157,7 @@ __ipip6_bucket(struct sit_net *sitn, struct ip_tunnel_parm_kern *parms)static inline struct ip_tunnel __rcu **ipip6_bucket (struct sit_net *sitn,struct ip_tunnel *t) { -return __ipip6_bucket (sitn, &t->parms); +return __ipip6_bucket (sitn, t->sit_parms); }static void ipip6_tunnel_unlink (struct sit_net *sitn,struct ip_tunnel *t)@@ -236,10 +236,11 @@ static int ipip6_tunnel_create(struct net_device *dev)if (err <0 ) goto out; -__dev_addr_set (dev, &t->parms.iph.saddr,4 ); -memcpy (dev->broadcast, &t->parms.iph.daddr,4 ); + t->parms = *t->sit_parms; +__dev_addr_set (dev, &t->sit_parms->iph.saddr,4 ); +memcpy (dev->broadcast, &t->sit_parms->iph.daddr,4 ); -if (test_bit (IP_TUNNEL_SIT_ISATAP_BIT , t->parms.i_flags)) +if (test_bit (IP_TUNNEL_SIT_ISATAP_BIT , t->sit_parms->i_flags)) dev->priv_flags |=IFF_ISATAP ; dev->rtnl_link_ops = &sit_link_ops;@@ -270,9 +271,9 @@ static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,for (tp =__ipip6_bucket (sitn, parms); (t =rtnl_dereference (*tp)) !=NULL ; tp = &t->next) { -if (local == t->parms.iph.saddr && - remote == t->parms.iph.daddr && - parms->link == t->parms.link) { +if (local == t->sit_parms->iph.saddr && + remote == t->sit_parms->iph.daddr && + parms->link == t->sit_parms->link) {if (create)return NULL ;else @@ -299,7 +300,10 @@ static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, nt =netdev_priv (dev); nt->net = net; - nt->parms = *parms; + nt->sit_parms =kmalloc_obj (*nt->sit_parms); +if (!nt->sit_parms) + goto failed_free; + *nt->sit_parms = *parms;if (ipip6_tunnel_create (dev) <0 ) goto failed_free;@@ -602,12 +606,12 @@ static int ipip6_err(struct sk_buff *skb, u32 info)if (type ==ICMP_DEST_UNREACH && code ==ICMP_FRAG_NEEDED ) {ipv4_update_pmtu (skb,dev_net (skb->dev), info, - t->parms.link, iph->protocol); + t->sit_parms->link, iph->protocol); err =0 ; goto out; }if (type ==ICMP_REDIRECT ) { -ipv4_redirect (skb,dev_net (skb->dev), t->parms.link, +ipv4_redirect (skb,dev_net (skb->dev), t->sit_parms->link, iph->protocol); err =0 ; goto out;@@ -618,10 +622,10 @@ static int ipip6_err(struct sk_buff *skb, u32 info) !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len)) goto out; -if (t->parms.iph.daddr ==0 ) +if (t->sit_parms->iph.daddr ==0 ) goto out; -if (t->parms.iph.ttl ==0 && type ==ICMP_TIME_EXCEEDED ) +if (t->sit_parms->iph.ttl ==0 && type ==ICMP_TIME_EXCEEDED ) goto out;if (time_before (jiffies,READ_ONCE (t->err_time) +IPTUNNEL_ERR_TIMEO ))@@ -722,8 +726,8 @@ static int ipip6_rcv(struct sk_buff *skb) tunnel =ipip6_tunnel_lookup (dev_net (skb->dev), skb->dev, iph->saddr, iph->daddr, sifindex);if (tunnel) { -if (tunnel->parms.iph.protocol !=IPPROTO_IPV6 && - tunnel->parms.iph.protocol !=0 ) +if (tunnel->sit_parms->iph.protocol !=IPPROTO_IPV6 && + tunnel->sit_parms->iph.protocol !=0 ) goto out; skb->mac_header = skb->network_header;@@ -798,8 +802,8 @@ static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)if (tunnel) {const struct tnl_ptk_info *tpi; -if (tunnel->parms.iph.protocol != ipproto && - tunnel->parms.iph.protocol !=0 ) +if (tunnel->sit_parms->iph.protocol != ipproto && + tunnel->sit_parms->iph.protocol !=0 ) goto drop;if (!xfrm4_policy_check (NULL ,XFRM_POLICY_IN , skb))@@ -938,9 +942,9 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->parms.iph; +const struct iphdr *tiph = &tunnel->sit_parms->iph;const struct ipv6hdr *iph6 =ipv6_hdr (skb); -u8 tos = tunnel->parms.iph.tos; +u8 tos = tunnel->sit_parms->iph.tos;__be16 df = tiph->frag_off;struct rtable *rt;/* Route to the other host */ struct net_device *tdev;/* Device to other host */ @@ -966,7 +970,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,if (!dst && !ipip6_tunnel_dst_find (skb, &dst, false)) goto tx_error; -flowi4_init_output (&fl4, tunnel->parms.link,READ_ONCE (tunnel->fwmark), +flowi4_init_output (&fl4, tunnel->sit_parms->link,READ_ONCE (tunnel->fwmark), tos &INET_DSCP_MASK ,RT_SCOPE_UNIVERSE ,IPPROTO_IPV6 ,0 , dst, tiph->saddr,0 ,0 ,sock_net_uid (tunnel->net,NULL ));@@ -1014,7 +1018,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, df =0 ; } -if (tunnel->parms.iph.daddr) +if (tunnel->sit_parms->iph.daddr)skb_dst_update_pmtu_no_confirm (skb, mtu);if (skb->len > mtu && !skb_is_gso (skb)) {@@ -1083,7 +1087,7 @@ static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, struct net_device *dev, u8 ipproto) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->parms.iph; +const struct iphdr *tiph = &tunnel->sit_parms->iph;if (iptunnel_handle_offloads (skb,SKB_GSO_IPXIP4 )) goto tx_error;@@ -1138,7 +1142,7 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev)const struct iphdr *iph;struct flowi4 fl4; - iph = &tunnel->parms.iph; + iph = &tunnel->sit_parms->iph;if (iph->daddr) { struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,@@ -1147,7 +1151,7 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev)0 ,0 ,IPPROTO_IPV6 , iph->tos &INET_DSCP_MASK , - tunnel->parms.link); + tunnel->sit_parms->link);if (!IS_ERR (rt)) { tdev = rt->dst.dev;@@ -1156,8 +1160,8 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev) dev->flags |=IFF_POINTOPOINT ; } -if (!tdev && tunnel->parms.link) - tdev =__dev_get_by_index (tunnel->net, tunnel->parms.link); +if (!tdev && tunnel->sit_parms->link) + tdev =__dev_get_by_index (tunnel->net, tunnel->sit_parms->link);if (tdev && !netif_is_l3_master (tdev)) {int mtu;@@ -1171,30 +1175,41 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev) dev->needed_headroom =ip_tunnel_limit_headroom (t_hlen + hlen); } -static void ipip6_tunnel_update (struct ip_tunnel *t, -struct ip_tunnel_parm_kern *p, -__u32 fwmark) +static int ipip6_tunnel_update (struct ip_tunnel *t, +struct ip_tunnel_parm_kern *p, +__u32 fwmark) {struct net *net = t->net;struct sit_net *sitn =net_generic (net, sit_net_id); +struct ip_tunnel_parm_kern *new_p, *old_p; + old_p = t->sit_parms; + new_p =kmalloc_obj (*new_p); +if (!new_p) +return -ENOMEM ; + *new_p = *old_p; + new_p->iph.saddr = p->iph.saddr; + new_p->iph.daddr = p->iph.daddr; + new_p->iph.ttl = p->iph.ttl; + new_p->iph.tos = p->iph.tos; + new_p->iph.frag_off = p->iph.frag_off; + new_p->link = p->link;ipip6_tunnel_unlink (sitn, t);synchronize_net (); - t->parms.iph.saddr = p->iph.saddr; - t->parms.iph.daddr = p->iph.daddr; + t->sit_parms = new_p; + t->parms.iph = new_p->iph; +WRITE_ONCE (t->parms.link, new_p->link);__dev_addr_set (t->dev, &p->iph.saddr,4 );memcpy (t->dev->broadcast, &p->iph.daddr,4 );ipip6_tunnel_link (sitn, t); - t->parms.iph.ttl = p->iph.ttl; - t->parms.iph.tos = p->iph.tos; - t->parms.iph.frag_off = p->iph.frag_off; -if (t->parms.link != p->link || t->fwmark != fwmark) { - t->parms.link = p->link; +if (old_p->link != p->link || t->fwmark != fwmark) {WRITE_ONCE (t->fwmark, fwmark);ipip6_tunnel_bind_dev (t->dev); }dst_cache_reset (&t->dst_cache);netdev_state_change (t->dev); +kfree (old_p); +return 0 ; }#ifdef CONFIG_IPV6_SIT_6RD @@ -1326,7 +1341,7 @@ ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm_kern *p) t =ipip6_tunnel_locate (t->net, p,0 );if (!t) t =netdev_priv (dev); -memcpy (p, &t->parms,sizeof (*p)); +memcpy (p, t->sit_parms,sizeof (*p));return 0 ; }@@ -1371,7 +1386,9 @@ ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm_kern *p) t =netdev_priv (dev); } -ipip6_tunnel_update (t, p, t->fwmark); + err =ipip6_tunnel_update (t, p, t->fwmark); +if (err) +return err; }return 0 ;@@ -1448,7 +1465,7 @@ static int ipip6_get_iflink(const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -return READ_ONCE (tunnel->parms.link); +return READ_ONCE (tunnel->sit_parms->link); }static const struct net_device_ops ipip6_netdev_ops = {@@ -1470,6 +1487,8 @@ static void ipip6_dev_free(struct net_device *dev)RCU_INIT_POINTER (tunnel->ip6rd,NULL );kfree (ip6rd);#endif +kfree (tunnel->sit_parms); + tunnel->sit_parms =NULL ;if (tunnel->dst_cache.cache) {dst_cache_destroy (&tunnel->dst_cache); tunnel->dst_cache.cache =NULL ;@@ -1512,7 +1531,7 @@ static int ipip6_tunnel_init(struct net_device *dev)int err; tunnel->dev = dev; -strscpy (tunnel->parms.name, dev->name); +strscpy (tunnel->sit_parms->name, dev->name);ipip6_tunnel_bind_dev (dev);@@ -1528,15 +1547,9 @@ static int ipip6_tunnel_init(struct net_device *dev)static void __net_init ipip6_fb_tunnel_init (struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -struct iphdr *iph = &tunnel->parms.iph;struct net *net =dev_net (dev);struct sit_net *sitn =net_generic (net, sit_net_id); - iph->version =4 ; - iph->protocol =IPPROTO_IPV6 ; - iph->ihl =5 ; - iph->ttl =64 ; -rcu_assign_pointer (sitn->tunnels_wc[0 ], tunnel); }@@ -1623,6 +1636,7 @@ static int ipip6_newlink(struct net_device *dev,#ifdef CONFIG_IPV6_SIT_6RD struct ip_tunnel_6rd ip6rd;#endif +struct ip_tunnel_parm_kern p;struct net *net;int err;@@ -1636,11 +1650,16 @@ static int ipip6_newlink(struct net_device *dev,return err; } -ipip6_netlink_parms (data, &nt->parms, &nt->fwmark); +ipip6_netlink_parms (data, &p, &nt->fwmark); -if (ipip6_tunnel_locate (net, &nt->parms,0 )) +if (ipip6_tunnel_locate (net, &p,0 ))return -EEXIST ; + nt->sit_parms =kmalloc_obj (*nt->sit_parms); +if (!nt->sit_parms) +return -ENOMEM ; + *nt->sit_parms = p; + err =ipip6_tunnel_create (dev);if (err <0 ) {ipip6_dev_free (dev);@@ -1707,7 +1726,9 @@ static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[], }else t =netdev_priv (dev); -ipip6_tunnel_update (t, &p, fwmark); + err =ipip6_tunnel_update (t, &p, fwmark); +if (err) +return err;#ifdef CONFIG_IPV6_SIT_6RD if (ipip6_netlink_6rd_parms (data, &ip6rd))@@ -1762,7 +1783,7 @@ static size_t ipip6_get_size(const struct net_device *dev)static int ipip6_fill_info (struct sk_buff *skb,const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -struct ip_tunnel_parm_kern *parm = &tunnel->parms; +struct ip_tunnel_parm_kern *parm = tunnel->sit_parms;#ifdef CONFIG_IPV6_SIT_6RD const struct ip_tunnel_6rd_parm *ip6rd;#endif @@ -1935,6 +1956,17 @@ static int __net_init sit_init_net(struct net *net) t =netdev_priv (sitn->fb_tunnel_dev); t->net = net; + t->sit_parms =kzalloc_obj (*t->sit_parms); +if (!t->sit_parms) { + err = -ENOMEM ; + goto err_reg_dev; + } + t->sit_parms->iph.version =4 ; + t->sit_parms->iph.protocol =IPPROTO_IPV6 ; + t->sit_parms->iph.ihl =5 ; + t->sit_parms->iph.ttl =64 ; +strscpy (t->sit_parms->name, sitn->fb_tunnel_dev->name); + t->parms = *t->sit_parms; err =ipip6_tunnel_clone_6rd (sitn->fb_tunnel_dev, sitn);if (err <0 )@@ -1945,8 +1977,6 @@ static int __net_init sit_init_net(struct net *net) goto err_reg_dev;ipip6_fb_tunnel_init (sitn->fb_tunnel_dev); - -strscpy (t->parms.name, sitn->fb_tunnel_dev->name);return 0 ; err_reg_dev:Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 7/9] sit: dynamically allocate struct ip_tunnel_parm_kern
In reply to Eric DumazetQuoted 5 lines · Eric Dumazet
net/ipv6/sit.c +1
-2ipip6_tunnel_unlink (sitn, t);synchronize_net (); - t->parms.iph.saddr = p->iph.saddr; - t->parms.iph.daddr = p->iph.daddr; + t->sit_parms = new_p;
[...]
net/ipv6/sit.c static void ipip6_tunnel_bind_dev(struct net_device *dev) · line 1211
Quoted 1 line
net/ipv6/sit.c +1
-0+kfree (old_p);
At this point in the series old_p is freed with no grace period: the synchronize_net() runs before the pointer switch and only covers the hash readers. ipip6_tunnel_xmit() doesn't go through the hash and keeps tiph = &tunnel->sit_parms->iph across the route lookup, and ipip6_get_iflink() is reached from iflink_show() with no lock at all. 8/9 turns this into kfree_rcu(), so the end result is fine, but this commit on its own is a UAF and breaks bisection. Moving the rcu_head and kfree_rcu() here (or squashing 7 and 8) would keep each step safe.
Smaller thing of the same kind in 5/9: check_6rd() gets rcu_dereference() but the rcu_read_lock() around the xmit path only arrives in 8/9.
Artem
Re: [PATCH net-next 7/9] sit: dynamically allocate struct ip_tunnel_parm_kern
Quoted 19 lines · Artem Lytkin
Quoted 5 lines · Eric Dumazet
ipip6_tunnel_unlink(sitn, t); synchronize_net();net/ipv6/sit.c +1
-2- t->parms.iph.saddr = p->iph.saddr; - t->parms.iph.daddr = p->iph.daddr; + t->sit_parms = new_p;[...]
Quoted 1 line
+
kfree(old_p);At this point in the series
old_pis freed with no grace period: thesynchronize_net()runs before the pointer switch and only covers the hash readers.ipip6_tunnel_xmit()doesn't go through the hash and keeps tiph =&tunnel->sit_parms->iphacross the route lookup, andipip6_get_iflink()is reached fromiflink_show()with no lock at all. 8/9 turns this intokfree_rcu(), so the end result is fine, but this commit on its own is a UAF and breaks bisection. Moving thercu_headandkfree_rcu()here (or squashing 7 and 8) would keep each step safe.
Quoted 3 lines
Smaller thing of the same kind in 5/9:
check_6rd()getsrcu_dereference()but thercu_read_lock()around the xmit path only arrives in 8/9.
Yeah, I had to split my original patch into three parts because our sashiko 5,000,000 token limit was hit.
[email protected]
AI reviewRe: [PATCH net-next 7/9] sit: dynamically allocate struct ip_tunnel_parm_kern
In reply to Eric DumazetThank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Critical: 0 · High: 0 · Medium: 1 · Low: 1
- [Medium] Intermediate-state use-after-free in net/ipv6/sit.c:
ipip6_tunnel_update() publishes the freshly allocated parameter block…
- [Low] Missing publication barrier / unannotated lockless pointer: in this commit sit_parms is a plain (non-__rcu) pointer in struct…
--
sit: dynamically allocate struct ip_tunnel_parm_kern
This moves the SIT configuration parameters out of the embedded ip_tunnel.parms into a separately allocated struct ip_tunnel_parm_kern reachable through the new ip_tunnel.sit_parms pointer, in preparation for converting them to RCU. All readers and writers in net/ipv6/sit.c are switched over to the new pointer.
A couple of questions about the intermediate state this commit leaves the tree in.
include/net/ip_tunnels.h struct ip_tunnel { · line 193
Quoted 9 lines
include/net/ip_tunnels.h +1
-0@@ -190,6 +190,7 @@ struct ip_tunnel {#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */ +struct ip_tunnel_parm_kern *sit_parms;
^^^^
[Severity: Low]
Should this member already carry the __rcu annotation in this commit? It is read from softirq packet paths (ipip6_tunnel_lookup(), ipip6_rcv(), sit_tunnel_rcv(), ipip6_tunnel_xmit(), sit_tunnel_xmit__()) without RTNL, while ipip6_tunnel_update() replaces it with a plain store. As a plain pointer, sparse cannot flag the unprotected dereferences, the publish has no release semantics, and the loads are plain data races for KCSAN.
The following patch in the series, "sit: convert configuration to RCU protection", changes this to struct ip_tunnel_parm_kern __rcu *sit_parms and converts publication to rcu_assign_pointer(), so the window is limited to this single commit. Would it be worth folding the annotation in here?
Quoted 9 lines
include/net/ip_tunnels.h +0
-0unsigned int ip_tnl_net_id;struct gro_cells gro_cells;__u32 fwmark;net/ipv6/sit.c +0
-0@@ -108,24 +108,24 @@ static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,int ifindex = dev ? dev->ifindex :0 ;
net/ipv6/sit.c static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net, · line 114
Quoted 10 lines
net/ipv6/sit.c +4
-4for_each_ip_tunnel_rcu (t, sitn->tunnels_r_l[h0 ^ h1]) { - if (local == t->parms.iph.saddr && - remote == t->parms.iph.daddr && - (!dev || !t->parms.link || ifindex == t->parms.link || - sifindex == t->parms.link) && + if (local == t->sit_parms->iph.saddr && + remote == t->sit_parms->iph.daddr && + (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || + sifindex == t->sit_parms->link) &&
^^^^
[Severity: Low]
Can this mix fields from two different parameter blocks? t->sit_parms is reloaded up to four times inside one match decision here, and ipip6_tunnel_update() swaps the pointer concurrently, so saddr/daddr could come from the old block and link from the new one.
The next patch caches a single rcu_dereference(t->sit_parms) in a local parms variable per candidate, which removes this; could that be done here instead?
Quoted 3 lines
net/ipv6/sit.c +0
-0(t->dev->flags & IFF_UP))return t; }
[ ... ]
net/ipv6/sit.c static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, · line 947
Quoted 9 lines
net/ipv6/sit.c +2
-2@@ -938,9 +942,9 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->parms.iph; +const struct iphdr *tiph = &tunnel->sit_parms->iph;const struct ipv6hdr *iph6 =ipv6_hdr (skb); -u8 tos = tunnel->parms.iph.tos; +u8 tos = tunnel->sit_parms->iph.tos;
^^^^
[Severity: Medium]
Can this read freed memory? ipip6_tunnel_xmit() reaches the tunnel via netdev_priv(dev), never through the sitn->tunnels[] buckets, and it keeps the interior pointer tiph across the route lookup, the PMTU handling, the headroom reallocation and iptunnel_xmit().
Meanwhile ipip6_tunnel_update() frees the block the pointer points into:
net/ipv6/sit.c:ipip6_tunnel_update() {
...
ipip6_tunnel_unlink(sitn, t);
synchronize_net();
t->sit_parms = new_p;
...
kfree(old_p);
return 0;
}The synchronize_net() runs before the new pointer is published and only drains readers that came in through the hash buckets, so it does not cover this path. dev->lltx is set for sit, so nothing else serializes a SIOCCHGTUNNEL or RTM_NEWLINK change against traffic on the same device. The values read out of the freed block (tiph->saddr, tos, ttl, frag_off) end up in the outer IPv4 header.
The immediately following patch, "sit: convert configuration to RCU protection", wraps the transmit path in rcu_read_lock() and uses kfree_rcu(old_p, rcu), so this hazard exists only between the two commits (a concern for bisection and partial backports).
Quoted 3 lines
net/ipv6/sit.c +0
-0__be16 df = tiph->frag_off;struct rtable *rt;/* Route to the other host */ struct net_device *tdev;/* Device to other host */
[ ... ]
net/ipv6/sit.c static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, · line 1090
Quoted 6 lines
net/ipv6/sit.c +1
-1@@ -1083,7 +1087,7 @@ static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, struct net_device *dev, u8 ipproto) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->parms.iph; +const struct iphdr *tiph = &tunnel->sit_parms->iph;
^^^^
[Severity: Medium]
Same question here: this hands an interior pointer of the freeable parameter block to ip_tunnel_xmit() with no reference and no RCU section held, so a concurrent ipip6_tunnel_update() can kfree() it underneath.
Quoted 3 lines
net/ipv6/sit.c +0
-0if (iptunnel_handle_offloads (skb,SKB_GSO_IPXIP4 )) goto tx_error;
[ ... ]
Quoted 21 lines
net/ipv6/sit.c +10
-3@@ -1171,30 +1175,41 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev) dev->needed_headroom =ip_tunnel_limit_headroom (t_hlen + hlen); } -static void ipip6_tunnel_update (struct ip_tunnel *t, -struct ip_tunnel_parm_kern *p, -__u32 fwmark) +static int ipip6_tunnel_update (struct ip_tunnel *t, +struct ip_tunnel_parm_kern *p, +__u32 fwmark) {struct net *net = t->net;struct sit_net *sitn =net_generic (net, sit_net_id); +struct ip_tunnel_parm_kern *new_p, *old_p; + old_p = t->sit_parms; + new_p =kmalloc_obj (*new_p); +if (!new_p) +return -ENOMEM ; + *new_p = *old_p; + new_p->iph.saddr = p->iph.saddr;
net/ipv6/sit.c static void ipip6_tunnel_bind_dev(struct net_device *dev) · line 1199
Quoted 10 lines
net/ipv6/sit.c +6
-2+ new_p->iph.daddr = p->iph.daddr; + new_p->iph.ttl = p->iph.ttl; + new_p->iph.tos = p->iph.tos; + new_p->iph.frag_off = p->iph.frag_off; + new_p->link = p->link;ipip6_tunnel_unlink (sitn, t);synchronize_net (); - t->parms.iph.saddr = p->iph.saddr; - t->parms.iph.daddr = p->iph.daddr; + t->sit_parms = new_p;
^^^^
[Severity: Medium]
Is the ordering here intentional? The grace period happens before the new pointer is published, so nothing separates the store of new_p from the kfree(old_p) below for readers that already loaded the old pointer.
Quoted 7 lines
net/ipv6/sit.c +2
-2+ t->parms.iph = new_p->iph; +WRITE_ONCE (t->parms.link, new_p->link);__dev_addr_set (t->dev, &p->iph.saddr,4 );memcpy (t->dev->broadcast, &p->iph.daddr,4 );ipip6_tunnel_link (sitn, t); - t->parms.iph.ttl = p->iph.ttl; - t->parms.iph.tos = p->iph.tos;
net/ipv6/sit.c static void ipip6_tunnel_bind_dev(struct net_device *dev) · line 1211
Quoted 10 lines
net/ipv6/sit.c +2
-3- t->parms.iph.frag_off = p->iph.frag_off; -if (t->parms.link != p->link || t->fwmark != fwmark) { - t->parms.link = p->link; +if (old_p->link != p->link || t->fwmark != fwmark) {WRITE_ONCE (t->fwmark, fwmark);ipip6_tunnel_bind_dev (t->dev); }dst_cache_reset (&t->dst_cache);netdev_state_change (t->dev); +kfree (old_p);
^^^^
[Severity: Medium]
Would kfree_rcu() be needed here, given that ipip6_tunnel_xmit(), sit_tunnel_xmit__() and ipip6_get_iflink() all reach sit_parms through netdev_priv(dev) rather than through the hash buckets?
The next patch in the series adds a struct rcu_head to ip_tunnel_parm_kern and turns this into kfree_rcu(old_p, rcu).
Quoted 2 lines
net/ipv6/sit.c +1
-0+return 0 ; }
[ ... ]
net/ipv6/sit.c static int ipip6_get_iflink(const struct net_device *dev) · line 1468
Quoted 6 lines
net/ipv6/sit.c +1
-1@@ -1448,7 +1465,7 @@ static int ipip6_get_iflink(const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -return READ_ONCE (tunnel->parms.link); +return READ_ONCE (tunnel->sit_parms->link);
^^^^
[Severity: Medium]
Here the READ_ONCE() now covers only the link field, not the pointer load itself. Since this accessor can run without RTNL, can it dereference the block that ipip6_tunnel_update() is about to kfree()? The following patch adds an rcu_read_lock() section and an rcu_dereference() here.
Quoted 3 lines
net/ipv6/sit.c +0
-0}static const struct net_device_ops ipip6_netdev_ops = {
[ ... ]
Signature · 1 line
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260907075846.2913645-1-edumazet%40google.com
PATCH 8/9 sit: convert configuration to RCU protection
In reply to Eric DumazetNow that SIT parameters are dynamically allocated, convert tunnel->sit_parms to an RCU-protected pointer.
Updates in ipip6_tunnel_update() allocate a new parameter block, publish it using rcu_assign_pointer(), and free the old one via kfree_rcu().
We only need to unlink and re-link the tunnel in the hash table if either saddr or daddr changed. When neither address changes, the unhash/re-hash and synchronize_net() can be completely skipped.
Readers in ipip6_tunnel_lookup(), ipip6_tunnel_xmit(), ipip6_err(), and ipip6_rcv() now safely dereference tunnel->sit_parms under RCU.
--- include/net/ip_tunnels.h | 3 +- net/ipv6/sit.c | 245 +++++++++++++++++++++++++-------------- 2 files changed, 157 insertions(+), 91 deletions(-)
include/net/ip_tunnels.h +2 -1
@@ -149,6 +149,7 @@ struct ip_tunnel_parm_kern {__be32 o_key;int link;struct iphdr iph; +struct rcu_head rcu; };struct ip_tunnel {
@@ -190,7 +191,7 @@ struct ip_tunnel {#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */ -struct ip_tunnel_parm_kern *sit_parms; +struct ip_tunnel_parm_kern __rcu *sit_parms;unsigned int ip_tnl_net_id;struct gro_cells gro_cells;__u32 fwmark;
net/ipv6/sit.c +155 -90
508 lines not on this page · raw message
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 8/9] sit: convert configuration to RCU protection
Quoted 7 lines
Now that SIT parameters are dynamically allocated, convert
tunnel->sit_parmsto an RCU-protected pointer.Updates in
ipip6_tunnel_update()allocate a new parameter block, publish it usingrcu_assign_pointer(), and free the old one viakfree_rcu().
Quoted 6 lines
We only need to unlink and re-link the tunnel in the hash table if either saddr or daddr changed. When neither address changes, the unhash/re-hash and
synchronize_net()can be completely skipped.Readers in
ipip6_tunnel_lookup(),ipip6_tunnel_xmit(),ipip6_err(), andipip6_rcv()now safely dereferencetunnel->sit_parmsunder RCU.
I think this patch is fine, I am just wondering if we can use more generic name with respect to 'sit_parms' since I guess we have the same issue for IPIP and IP6IP6 tunnels. Do you prefer to have dedicated pointers for them?
Regards,
Lorenzo
Quoted 574 lines
Signed-off-by Eric Dumazet <[email protected]>--- include/net/ip_tunnels.h | 3 +- net/ipv6/sit.c | 245 +++++++++++++++++++++++++-------------- 2 files changed, 157 insertions(+), 91 deletions(-)include/net/ip_tunnels.h +2
-1@@ -149,6 +149,7 @@ struct ip_tunnel_parm_kern {__be32 o_key;int link;struct iphdr iph; +struct rcu_head rcu; };struct ip_tunnel {@@ -190,7 +191,7 @@ struct ip_tunnel {#endif struct ip_tunnel_prl_entry __rcu *prl;/* potential router list */ unsigned int prl_count;/* # of entries in PRL */ -struct ip_tunnel_parm_kern *sit_parms; +struct ip_tunnel_parm_kern __rcu *sit_parms;unsigned int ip_tnl_net_id;struct gro_cells gro_cells;__u32 fwmark;net/ipv6/sit.c +155
-90@@ -108,24 +108,33 @@ static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,int ifindex = dev ? dev->ifindex :0 ;for_each_ip_tunnel_rcu (t, sitn->tunnels_r_l[h0 ^ h1]) { -if (local == t->sit_parms->iph.saddr && - remote == t->sit_parms->iph.daddr && - (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || - sifindex == t->sit_parms->link) && +const struct ip_tunnel_parm_kern *parms; + + parms =rcu_dereference (t->sit_parms); +if (local == parms->iph.saddr && + remote == parms->iph.daddr && + (!dev || !parms->link || ifindex == parms->link || + sifindex == parms->link) && (t->dev->flags &IFF_UP ))return t; }for_each_ip_tunnel_rcu (t, sitn->tunnels_r[h0]) { -if (remote == t->sit_parms->iph.daddr && - (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || - sifindex == t->sit_parms->link) && +const struct ip_tunnel_parm_kern *parms; + + parms =rcu_dereference (t->sit_parms); +if (remote == parms->iph.daddr && + (!dev || !parms->link || ifindex == parms->link || + sifindex == parms->link) && (t->dev->flags &IFF_UP ))return t; }for_each_ip_tunnel_rcu (t, sitn->tunnels_l[h1]) { -if (local == t->sit_parms->iph.saddr && - (!dev || !t->sit_parms->link || ifindex == t->sit_parms->link || - sifindex == t->sit_parms->link) && +const struct ip_tunnel_parm_kern *parms; + + parms =rcu_dereference (t->sit_parms); +if (local == parms->iph.saddr && + (!dev || !parms->link || ifindex == parms->link || + sifindex == parms->link) && (t->dev->flags &IFF_UP ))return t; }@@ -157,7 +166,7 @@ __ipip6_bucket(struct sit_net *sitn, struct ip_tunnel_parm_kern *parms)static inline struct ip_tunnel __rcu **ipip6_bucket (struct sit_net *sitn,struct ip_tunnel *t) { -return __ipip6_bucket (sitn, t->sit_parms); +return __ipip6_bucket (sitn,rtnl_dereference (t->sit_parms)); }static void ipip6_tunnel_unlink (struct sit_net *sitn,struct ip_tunnel *t)@@ -230,17 +239,19 @@ static int ipip6_tunnel_create(struct net_device *dev) {struct ip_tunnel *t =netdev_priv (dev);struct sit_net *sitn =net_generic (t->net, sit_net_id); +struct ip_tunnel_parm_kern *parms;int err; err =ipip6_tunnel_clone_6rd (dev, sitn);if (err <0 ) goto out; - t->parms = *t->sit_parms; -__dev_addr_set (dev, &t->sit_parms->iph.saddr,4 ); -memcpy (dev->broadcast, &t->sit_parms->iph.daddr,4 ); + parms =rtnl_dereference (t->sit_parms); + t->parms = *parms; +__dev_addr_set (dev, &parms->iph.saddr,4 ); +memcpy (dev->broadcast, &parms->iph.daddr,4 ); -if (test_bit (IP_TUNNEL_SIT_ISATAP_BIT , t->sit_parms->i_flags)) +if (test_bit (IP_TUNNEL_SIT_ISATAP_BIT , parms->i_flags)) dev->priv_flags |=IFF_ISATAP ; dev->rtnl_link_ops = &sit_link_ops;@@ -264,6 +275,7 @@ static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,__be32 local = parms->iph.saddr;struct ip_tunnel *t, *nt;struct ip_tunnel __rcu **tp; +struct ip_tunnel_parm_kern *nt_parms;struct net_device *dev;char name[IFNAMSIZ ];struct sit_net *sitn =net_generic (net, sit_net_id);@@ -271,9 +283,12 @@ static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,for (tp =__ipip6_bucket (sitn, parms); (t =rtnl_dereference (*tp)) !=NULL ; tp = &t->next) { -if (local == t->sit_parms->iph.saddr && - remote == t->sit_parms->iph.daddr && - parms->link == t->sit_parms->link) { +const struct ip_tunnel_parm_kern *tparms; + + tparms =rtnl_dereference (t->sit_parms); +if (local == tparms->iph.saddr && + remote == tparms->iph.daddr && + parms->link == tparms->link) {if (create)return NULL ;else @@ -300,10 +315,11 @@ static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, nt =netdev_priv (dev); nt->net = net; - nt->sit_parms =kmalloc_obj (*nt->sit_parms); -if (!nt->sit_parms) + nt_parms =kmalloc_obj (*nt_parms); +if (!nt_parms) goto failed_free; - *nt->sit_parms = *parms; + *nt_parms = *parms; +rcu_assign_pointer (nt->sit_parms, nt_parms);if (ipip6_tunnel_create (dev) <0 ) goto failed_free;@@ -599,41 +615,45 @@ static int ipip6_err(struct sk_buff *skb, u32 info) err = -ENOENT ; sifindex =netif_is_l3_master (skb->dev) ?IPCB (skb)->iif :0 ; +rcu_read_lock (); t =ipip6_tunnel_lookup (dev_net (skb->dev), skb->dev, iph->daddr, iph->saddr, sifindex); -if (!t) - goto out; +if (t) { +const struct ip_tunnel_parm_kern *parms; -if (type ==ICMP_DEST_UNREACH && code ==ICMP_FRAG_NEEDED ) { -ipv4_update_pmtu (skb,dev_net (skb->dev), info, - t->sit_parms->link, iph->protocol); - err =0 ; - goto out; - } -if (type ==ICMP_REDIRECT ) { -ipv4_redirect (skb,dev_net (skb->dev), t->sit_parms->link, - iph->protocol); - err =0 ; - goto out; - } + parms =rcu_dereference (t->sit_parms); +if (type ==ICMP_DEST_UNREACH && code ==ICMP_FRAG_NEEDED ) { +ipv4_update_pmtu (skb,dev_net (skb->dev), info, + parms->link, iph->protocol); + err =0 ; + goto out; + } +if (type ==ICMP_REDIRECT ) { +ipv4_redirect (skb,dev_net (skb->dev), parms->link, + iph->protocol); + err =0 ; + goto out; + } - err =0 ; -if (__in6_dev_get (skb->dev) && - !ip6_err_gen_icmpv6_unreach (skb, iph->ihl *4 , type, data_len)) - goto out; + err =0 ; +if (__in6_dev_get (skb->dev) && + !ip6_err_gen_icmpv6_unreach (skb, iph->ihl *4 , type, data_len)) + goto out; -if (t->sit_parms->iph.daddr ==0 ) - goto out; +if (parms->iph.daddr ==0 ) + goto out; -if (t->sit_parms->iph.ttl ==0 && type ==ICMP_TIME_EXCEEDED ) - goto out; +if (parms->iph.ttl ==0 && type ==ICMP_TIME_EXCEEDED ) + goto out; -if (time_before (jiffies,READ_ONCE (t->err_time) +IPTUNNEL_ERR_TIMEO )) -WRITE_ONCE (t->err_count,READ_ONCE (t->err_count) +1 ); -else -WRITE_ONCE (t->err_count,1 ); -WRITE_ONCE (t->err_time, jiffies); +if (time_before (jiffies,READ_ONCE (t->err_time) +IPTUNNEL_ERR_TIMEO )) +WRITE_ONCE (t->err_count,READ_ONCE (t->err_count) +1 ); +else +WRITE_ONCE (t->err_count,1 ); +WRITE_ONCE (t->err_time, jiffies); + } out: +rcu_read_unlock ();return err; }@@ -726,8 +746,11 @@ static int ipip6_rcv(struct sk_buff *skb) tunnel =ipip6_tunnel_lookup (dev_net (skb->dev), skb->dev, iph->saddr, iph->daddr, sifindex);if (tunnel) { -if (tunnel->sit_parms->iph.protocol !=IPPROTO_IPV6 && - tunnel->sit_parms->iph.protocol !=0 ) +const struct ip_tunnel_parm_kern *parms; + + parms =rcu_dereference (tunnel->sit_parms); +if (parms->iph.protocol !=IPPROTO_IPV6 && + parms->iph.protocol !=0 ) goto out; skb->mac_header = skb->network_header;@@ -800,10 +823,12 @@ static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto) tunnel =ipip6_tunnel_lookup (dev_net (skb->dev), skb->dev, iph->saddr, iph->daddr, sifindex);if (tunnel) { +const struct ip_tunnel_parm_kern *parms;const struct tnl_ptk_info *tpi; -if (tunnel->sit_parms->iph.protocol != ipproto && - tunnel->sit_parms->iph.protocol !=0 ) + parms =rcu_dereference (tunnel->sit_parms); +if (parms->iph.protocol != ipproto && + parms->iph.protocol !=0 ) goto drop;if (!xfrm4_policy_check (NULL ,XFRM_POLICY_IN , skb))@@ -942,20 +967,27 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->sit_parms->iph; +const struct ip_tunnel_parm_kern *parms; +const struct iphdr *tiph;const struct ipv6hdr *iph6 =ipv6_hdr (skb); -u8 tos = tunnel->sit_parms->iph.tos; -__be16 df = tiph->frag_off; +u8 tos; +__be16 df;struct rtable *rt;/* Route to the other host */ struct net_device *tdev;/* Device to other host */ unsigned int max_headroom;/* The extra header space needed */ -__be32 dst = tiph->daddr; +__be32 dst;int err_count, mtu;struct flowi4 fl4;u8 ttl;u8 protocol =IPPROTO_IPV6 ;int t_hlen = tunnel->hlen +sizeof (struct iphdr ); + parms =rcu_dereference (tunnel->sit_parms); + tiph = &parms->iph; + tos = parms->iph.tos; + df = tiph->frag_off; + dst = tiph->daddr; +if (tos ==1 ) tos =ipv6_get_dsfield (iph6);@@ -970,7 +1002,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,if (!dst && !ipip6_tunnel_dst_find (skb, &dst, false)) goto tx_error; -flowi4_init_output (&fl4, tunnel->sit_parms->link,READ_ONCE (tunnel->fwmark), +flowi4_init_output (&fl4, parms->link,READ_ONCE (tunnel->fwmark), tos &INET_DSCP_MASK ,RT_SCOPE_UNIVERSE ,IPPROTO_IPV6 ,0 , dst, tiph->saddr,0 ,0 ,sock_net_uid (tunnel->net,NULL ));@@ -1018,7 +1050,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, df =0 ; } -if (tunnel->sit_parms->iph.daddr) +if (parms->iph.daddr)skb_dst_update_pmtu_no_confirm (skb, mtu);if (skb->len > mtu && !skb_is_gso (skb)) {@@ -1087,13 +1119,17 @@ static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, struct net_device *dev, u8 ipproto) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->sit_parms->iph; +const struct ip_tunnel_parm_kern *parms; +const struct iphdr *tiph;if (iptunnel_handle_offloads (skb,SKB_GSO_IPXIP4 )) goto tx_error;skb_set_inner_ipproto (skb, ipproto); + parms =rcu_dereference (tunnel->sit_parms); + tiph = &parms->iph; +ip_tunnel_xmit (skb, dev, tiph, ipproto);return NETDEV_TX_OK ; tx_error:@@ -1108,6 +1144,7 @@ static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,if (!pskb_inet_may_pull (skb)) goto tx_err; +rcu_read_lock ();switch (skb->protocol) {case htons (ETH_P_IP ):sit_tunnel_xmit__ (skb, dev,IPPROTO_IPIP );@@ -1121,8 +1158,10 @@ static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,break ;#endif default : +rcu_read_unlock (); goto tx_err; } +rcu_read_unlock ();return NETDEV_TX_OK ;@@ -1130,19 +1169,20 @@ static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,DEV_STATS_INC (dev, tx_errors);kfree_skb (skb);return NETDEV_TX_OK ; - }static void ipip6_tunnel_bind_dev (struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev);int t_hlen = tunnel->hlen +sizeof (struct iphdr ); +const struct ip_tunnel_parm_kern *parms;struct net_device *tdev =NULL ;int hlen =LL_MAX_HEADER ;const struct iphdr *iph;struct flowi4 fl4; - iph = &tunnel->sit_parms->iph; + parms =rtnl_dereference (tunnel->sit_parms); + iph = &parms->iph;if (iph->daddr) { struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,@@ -1151,7 +1191,7 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev)0 ,0 ,IPPROTO_IPV6 , iph->tos &INET_DSCP_MASK , - tunnel->sit_parms->link); + parms->link);if (!IS_ERR (rt)) { tdev = rt->dst.dev;@@ -1160,8 +1200,8 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev) dev->flags |=IFF_POINTOPOINT ; } -if (!tdev && tunnel->sit_parms->link) - tdev =__dev_get_by_index (tunnel->net, tunnel->sit_parms->link); +if (!tdev && parms->link) + tdev =__dev_get_by_index (tunnel->net, parms->link);if (tdev && !netif_is_l3_master (tdev)) {int mtu;@@ -1182,8 +1222,9 @@ static int ipip6_tunnel_update(struct ip_tunnel *t,struct net *net = t->net;struct sit_net *sitn =net_generic (net, sit_net_id);struct ip_tunnel_parm_kern *new_p, *old_p; +bool move; - old_p = t->sit_parms; + old_p =rtnl_dereference (t->sit_parms); new_p =kmalloc_obj (*new_p);if (!new_p)return -ENOMEM ;@@ -1194,21 +1235,29 @@ static int ipip6_tunnel_update(struct ip_tunnel *t, new_p->iph.tos = p->iph.tos; new_p->iph.frag_off = p->iph.frag_off; new_p->link = p->link; -ipip6_tunnel_unlink (sitn, t); -synchronize_net (); - t->sit_parms = new_p; + move = old_p->iph.saddr != p->iph.saddr || + old_p->iph.daddr != p->iph.daddr; + +if (move) +ipip6_tunnel_unlink (sitn, t); + t->parms.iph = new_p->iph;WRITE_ONCE (t->parms.link, new_p->link); -__dev_addr_set (t->dev, &p->iph.saddr,4 ); -memcpy (t->dev->broadcast, &p->iph.daddr,4 ); -ipip6_tunnel_link (sitn, t); +rcu_assign_pointer (t->sit_parms, new_p); + +if (move) { +synchronize_net (); +__dev_addr_set (t->dev, &p->iph.saddr,4 ); +memcpy (t->dev->broadcast, &p->iph.daddr,4 ); +ipip6_tunnel_link (sitn, t); + }if (old_p->link != p->link || t->fwmark != fwmark) {WRITE_ONCE (t->fwmark, fwmark);ipip6_tunnel_bind_dev (t->dev); }dst_cache_reset (&t->dst_cache);netdev_state_change (t->dev); -kfree (old_p); +kfree_rcu (old_p, rcu);return 0 ; }@@ -1336,12 +1385,14 @@ static int ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm_kern *p) { struct ip_tunnel *t = netdev_priv(dev); +const struct ip_tunnel_parm_kern *parms;if (dev ==dev_to_sit_net (dev)->fb_tunnel_dev) t =ipip6_tunnel_locate (t->net, p,0 );if (!t) t =netdev_priv (dev); -memcpy (p, t->sit_parms,sizeof (*p)); + parms =rtnl_dereference (t->sit_parms); +memcpy (p, parms,sizeof (*p));return 0 ; }@@ -1464,8 +1515,15 @@ ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,static int ipip6_get_iflink (const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); +const struct ip_tunnel_parm_kern *parms; +int link; -return READ_ONCE (tunnel->sit_parms->link); +rcu_read_lock (); + parms =rcu_dereference (tunnel->sit_parms); + link = parms ? parms->link :0 ; +rcu_read_unlock (); + +return link; }static const struct net_device_ops ipip6_netdev_ops = {@@ -1480,6 +1538,7 @@ static const struct net_device_ops ipip6_netdev_ops = {static void ipip6_dev_free (struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); +struct ip_tunnel_parm_kern *parms;#ifdef CONFIG_IPV6_SIT_6RD struct ip_tunnel_6rd_parm *ip6rd;@@ -1487,8 +1546,9 @@ static void ipip6_dev_free(struct net_device *dev)RCU_INIT_POINTER (tunnel->ip6rd,NULL );kfree (ip6rd);#endif -kfree (tunnel->sit_parms); - tunnel->sit_parms =NULL ; + parms =rcu_dereference_protected (tunnel->sit_parms,1 ); +RCU_INIT_POINTER (tunnel->sit_parms,NULL ); +kfree (parms);if (tunnel->dst_cache.cache) {dst_cache_destroy (&tunnel->dst_cache); tunnel->dst_cache.cache =NULL ;@@ -1528,10 +1588,12 @@ static void ipip6_tunnel_setup(struct net_device *dev)static int ipip6_tunnel_init (struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); +struct ip_tunnel_parm_kern *parms;int err; tunnel->dev = dev; -strscpy (tunnel->sit_parms->name, dev->name); + parms =rtnl_dereference (tunnel->sit_parms); +strscpy (parms->name, dev->name);ipip6_tunnel_bind_dev (dev);@@ -1549,7 +1611,6 @@ static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)struct ip_tunnel *tunnel =netdev_priv (dev);struct net *net =dev_net (dev);struct sit_net *sitn =net_generic (net, sit_net_id); -rcu_assign_pointer (sitn->tunnels_wc[0 ], tunnel); }@@ -1636,6 +1697,7 @@ static int ipip6_newlink(struct net_device *dev,#ifdef CONFIG_IPV6_SIT_6RD struct ip_tunnel_6rd ip6rd;#endif +struct ip_tunnel_parm_kern *nt_parms;struct ip_tunnel_parm_kern p;struct net *net;int err;@@ -1655,10 +1717,11 @@ static int ipip6_newlink(struct net_device *dev,if (ipip6_tunnel_locate (net, &p,0 ))return -EEXIST ; - nt->sit_parms =kmalloc_obj (*nt->sit_parms); -if (!nt->sit_parms) + nt_parms =kmalloc_obj (*nt_parms); +if (!nt_parms)return -ENOMEM ; - *nt->sit_parms = p; + *nt_parms = p; +rcu_assign_pointer (nt->sit_parms, nt_parms); err =ipip6_tunnel_create (dev);if (err <0 ) {@@ -1783,7 +1846,7 @@ static size_t ipip6_get_size(const struct net_device *dev)static int ipip6_fill_info (struct sk_buff *skb,const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -struct ip_tunnel_parm_kern *parm = tunnel->sit_parms; +const struct ip_tunnel_parm_kern *parm =rtnl_dereference (tunnel->sit_parms);#ifdef CONFIG_IPV6_SIT_6RD const struct ip_tunnel_6rd_parm *ip6rd;#endif @@ -1929,6 +1992,7 @@ static void __net_exit sit_exit_rtnl_net(struct net *net, struct list_head *headstatic int __net_init sit_init_net (struct net *net) {struct sit_net *sitn =net_generic (net, sit_net_id); +struct ip_tunnel_parm_kern *nt_parms;struct ip_tunnel *t;int err;@@ -1956,17 +2020,18 @@ static int __net_init sit_init_net(struct net *net) t =netdev_priv (sitn->fb_tunnel_dev); t->net = net; - t->sit_parms =kzalloc_obj (*t->sit_parms); -if (!t->sit_parms) { + nt_parms =kzalloc_obj (*nt_parms); +if (!nt_parms) { err = -ENOMEM ; goto err_reg_dev; } - t->sit_parms->iph.version =4 ; - t->sit_parms->iph.protocol =IPPROTO_IPV6 ; - t->sit_parms->iph.ihl =5 ; - t->sit_parms->iph.ttl =64 ; -strscpy (t->sit_parms->name, sitn->fb_tunnel_dev->name); - t->parms = *t->sit_parms; + nt_parms->iph.version =4 ; + nt_parms->iph.protocol =IPPROTO_IPV6 ; + nt_parms->iph.ihl =5 ; + nt_parms->iph.ttl =64 ; +strscpy (nt_parms->name, sitn->fb_tunnel_dev->name); + t->parms = *nt_parms; +rcu_assign_pointer (t->sit_parms, nt_parms); err =ipip6_tunnel_clone_6rd (sitn->fb_tunnel_dev, sitn);if (err <0 )Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 8/9] sit: convert configuration to RCU protection
Quoted 12 lines · Lorenzo Bianconi
Quoted 12 lines
Now that SIT parameters are dynamically allocated, convert
tunnel->sit_parmsto an RCU-protected pointer.Updates in
ipip6_tunnel_update()allocate a new parameter block, publish it usingrcu_assign_pointer(), and free the old one viakfree_rcu().We only need to unlink and re-link the tunnel in the hash table if either saddr or daddr changed. When neither address changes, the unhash/re-hash and
synchronize_net()can be completely skipped.
Quoted 6 lines
Quoted 2 lines
Readers in
ipip6_tunnel_lookup(),ipip6_tunnel_xmit(),ipip6_err(), andipip6_rcv()now safely dereferencetunnel->sit_parmsunder RCU.I think this patch is fine, I am just wondering if we can use more generic name with respect to '
sit_parms' since I guess we have the same issue for IPIP and IP6IP6 tunnels. Do you prefer to have dedicated pointers for them?
Name could be more generic indeed.
I was thinking of using a union for future tunnel conversions.
Quoted 5 lines
Acked-by Lorenzo Bianconi <[email protected]>Regards,
Lorenzo
[email protected]
AI reviewRe: [PATCH net-next 8/9] sit: convert configuration to RCU protection
In reply to Eric DumazetThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Critical: 0 · High: 0 · Medium: 1 · Low: 0
- [Medium] SIT tunnel parameters are now read from two objects that live in different synchronization domains, and ipip6_tunnel_update()…
--
sit: convert configuration to RCU protection
This converts tunnel->sit_parms into an RCU-protected pointer, so that ipip6_tunnel_update() publishes a freshly allocated parameter block with rcu_assign_pointer() and frees the old one with kfree_rcu(). The unlink / re-link plus synchronize_net() are now skipped when neither saddr nor daddr changed.
Quoted 4 lines
net/ipv6/sit.c +0
-0
[ ... ]
Quoted 19 lines
net/ipv6/sit.c +5
-1@@ -1087,13 +1119,17 @@ static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, struct net_device *dev, u8 ipproto) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct iphdr *tiph = &tunnel->sit_parms->iph; +const struct ip_tunnel_parm_kern *parms; +const struct iphdr *tiph;if (iptunnel_handle_offloads (skb,SKB_GSO_IPXIP4 )) goto tx_error;skb_set_inner_ipproto (skb, ipproto); + parms =rcu_dereference (tunnel->sit_parms); + tiph = &parms->iph; +ip_tunnel_xmit (skb, dev, tiph, ipproto);return NETDEV_TX_OK ; tx_error:
[ ... ]
Quoted 9 lines
net/ipv6/sit.c +2
-3@@ -1194,21 +1235,29 @@ static int ipip6_tunnel_update(struct ip_tunnel *t, new_p->iph.tos = p->iph.tos; new_p->iph.frag_off = p->iph.frag_off; new_p->link = p->link; -ipip6_tunnel_unlink (sitn, t); -synchronize_net (); - t->sit_parms = new_p; + move = old_p->iph.saddr != p->iph.saddr || + old_p->iph.daddr != p->iph.daddr;
net/ipv6/sit.c static int ipip6_tunnel_update(struct ip_tunnel *t, · line 1246
Quoted 10 lines
net/ipv6/sit.c +5
-3+ +if (move) +ipip6_tunnel_unlink (sitn, t); + t->parms.iph = new_p->iph;WRITE_ONCE (t->parms.link, new_p->link); -__dev_addr_set (t->dev, &p->iph.saddr,4 ); -memcpy (t->dev->broadcast, &p->iph.daddr,4 ); -ipip6_tunnel_link (sitn, t); +rcu_assign_pointer (t->sit_parms, new_p);
[Severity: Medium]
After this change, are the SIT transmit parameters read from two objects that are no longer updated together?
ipip6_tunnel_update() now performs two separate stores here:
t->parms.iph = new_p->iph; /* plain struct copy */ WRITE_ONCE(t->parms.link, new_p->link); rcu_assign_pointer(t->sit_parms, new_p); /* published */
On the transmit side, sit_tunnel_xmit__() takes the header out of the RCU-published block:
parms = rcu_dereference(tunnel->sit_parms); tiph = &parms->iph; ip_tunnel_xmit(skb, dev, tiph, ipproto);
but ip_tunnel_xmit() in net/ipv4/ip_tunnel.c mixes that with the shadow copy that is not RCU protected:
connected = (tunnel->parms.iph.daddr != 0); ... dst = tnl_params->daddr; ... rt = connected ? dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr) : NULL; ... else if (!md && connected) dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
Since sit_tunnel_xmit() runs locklessly, can it observe the two objects from different generations, e.g. sit_parms with daddr == 0 (so dst is derived per packet from the skb/neighbour) while tunnel->parms.iph.daddr already holds the new address, making connected true? In that case ip_tunnel_xmit() would consult and populate tunnel->dst_cache for a destination-specific route on a tunnel that the header says is NBMA, so a burst of packets could be routed via the cached entry. The reverse pairing (old shadow daddr, new sit_parms daddr) looks possible too where the plain tunnel->parms.iph.daddr load can be reordered against the rcu_dereference().
Independently of the ordering question, is the plain struct assignment t->parms.iph = new_p->iph a data race? It writes roughly twenty bytes without any annotation while the xmit path is reading ttl, tos, frag_off and daddr from tunnel->parms, which KCSAN should flag and which can hand the transmit path a mixture of old and new field values.
The synchronize_net() in the move case only runs after publication, so it bounds but does not close the window for a reader that entered right at the grace period boundary.
Would it make sense for ip_tunnel_xmit() to derive connected and link from the same parameter block that tiph came from, so that only one object is consulted per packet?
Quoted 17 lines
net/ipv6/sit.c +7
-1+if (move) { +synchronize_net (); +__dev_addr_set (t->dev, &p->iph.saddr,4 ); +memcpy (t->dev->broadcast, &p->iph.daddr,4 ); +ipip6_tunnel_link (sitn, t); + }if (old_p->link != p->link || t->fwmark != fwmark) {WRITE_ONCE (t->fwmark, fwmark);ipip6_tunnel_bind_dev (t->dev); }dst_cache_reset (&t->dst_cache);netdev_state_change (t->dev); -kfree (old_p); +kfree_rcu (old_p, rcu);return 0 ; }
[ ... ]
Signature · 1 line
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260907075846.2913645-1-edumazet%40google.com
PATCH 9/9 sit: no longer rely on RTNL in ipip6_fill_info()
In reply to Eric DumazetNow that SIT parameters and 6RD parameters are RCU-protected, and fwmark/encap are annotated with READ_ONCE()/WRITE_ONCE(), ipip6_fill_info() no longer requires RTNL to be held.
Wrap the attribute serialization in rcu_read_lock() and rcu_read_unlock().
--- net/ipv6/sit.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-)
net/ipv6/sit.c +15 -6
@@ -1846,11 +1846,18 @@ static size_t ipip6_get_size(const struct net_device *dev)static int ipip6_fill_info (struct sk_buff *skb,const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct ip_tunnel_parm_kern *parm =rtnl_dereference (tunnel->sit_parms); +const struct ip_tunnel_parm_kern *parm;#ifdef CONFIG_IPV6_SIT_6RD const struct ip_tunnel_6rd_parm *ip6rd;#endif +rcu_read_lock (); + parm =rcu_dereference (tunnel->sit_parms); +if (!parm) { +rcu_read_unlock (); +return -ENODEV ; + } + if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) || nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
@@ -1865,7 +1872,7 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) goto nla_put_failure;#ifdef CONFIG_IPV6_SIT_6RD - ip6rd =rcu_dereference_rtnl (tunnel->ip6rd); + ip6rd =rcu_dereference (tunnel->ip6rd); if (ip6rd && (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX, &ip6rd->prefix) ||
@@ -1879,18 +1886,20 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)#endif if (nla_put_u16 (skb,IFLA_IPTUN_ENCAP_TYPE , - tunnel->encap.type) || +READ_ONCE (tunnel->encap.type)) ||nla_put_be16 (skb,IFLA_IPTUN_ENCAP_SPORT , - tunnel->encap.sport) || +READ_ONCE (tunnel->encap.sport)) ||nla_put_be16 (skb,IFLA_IPTUN_ENCAP_DPORT , - tunnel->encap.dport) || +READ_ONCE (tunnel->encap.dport)) ||nla_put_u16 (skb,IFLA_IPTUN_ENCAP_FLAGS , - tunnel->encap.flags)) +READ_ONCE (tunnel->encap.flags))) goto nla_put_failure; +rcu_read_unlock ();return 0 ; nla_put_failure: +rcu_read_unlock ();return -EMSGSIZE ; }
Signature · 1 line
2.55.0.979.g7e5102b832-goog
Re: [PATCH net-next 9/9] sit: no longer rely on RTNL in ipip6_fill_info()
Quoted 5 lines
Now that SIT parameters and 6RD parameters are RCU-protected, and fwmark/encap are annotated with READ_ONCE()/WRITE_ONCE(),
ipip6_fill_info()no longer requires RTNL to be held.Wrap the attribute serialization in
rcu_read_lock()andrcu_read_unlock().
Quoted 67 lines
Signed-off-by Eric Dumazet <[email protected]>--- net/ipv6/sit.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-)net/ipv6/sit.c +15
-6@@ -1846,11 +1846,18 @@ static size_t ipip6_get_size(const struct net_device *dev)static int ipip6_fill_info (struct sk_buff *skb,const struct net_device *dev) {struct ip_tunnel *tunnel =netdev_priv (dev); -const struct ip_tunnel_parm_kern *parm =rtnl_dereference (tunnel->sit_parms); +const struct ip_tunnel_parm_kern *parm;#ifdef CONFIG_IPV6_SIT_6RD const struct ip_tunnel_6rd_parm *ip6rd;#endif +rcu_read_lock (); + parm =rcu_dereference (tunnel->sit_parms); +if (!parm) { +rcu_read_unlock (); +return -ENODEV ; + } + if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) || nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||@@ -1865,7 +1872,7 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) goto nla_put_failure;#ifdef CONFIG_IPV6_SIT_6RD - ip6rd =rcu_dereference_rtnl (tunnel->ip6rd); + ip6rd =rcu_dereference (tunnel->ip6rd); if (ip6rd && (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX, &ip6rd->prefix) ||@@ -1879,18 +1886,20 @@ static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)#endif if (nla_put_u16 (skb,IFLA_IPTUN_ENCAP_TYPE , - tunnel->encap.type) || +READ_ONCE (tunnel->encap.type)) ||nla_put_be16 (skb,IFLA_IPTUN_ENCAP_SPORT , - tunnel->encap.sport) || +READ_ONCE (tunnel->encap.sport)) ||nla_put_be16 (skb,IFLA_IPTUN_ENCAP_DPORT , - tunnel->encap.dport) || +READ_ONCE (tunnel->encap.dport)) ||nla_put_u16 (skb,IFLA_IPTUN_ENCAP_FLAGS , - tunnel->encap.flags)) +READ_ONCE (tunnel->encap.flags))) goto nla_put_failure; +rcu_read_unlock ();return 0 ; nla_put_failure: +rcu_read_unlock ();return -EMSGSIZE ; }Signature · 1 line
2.55.0.979.g7e5102b832-goog
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 13 lines · On Mon, 7 Sep 2026 07:58:37 +0000 you wrote:
SIT (IPv6-in-IPv4) tunnel configuration and status reporting have historically relied on the RTNL lock for synchronization. Consequently, netlink dumps via
ipip6_fill_info()had to run with RTNL held, adding contention during network device dumps.At the same time, the transmit path (
dev->lltx== true), tunnel lookups, and error handling run locklessly and can race with configuration updates. This can result in torn reads of multi-word fields (such as the 128-bit 6RD IPv6 prefix) or transiently zeroed encapsulation parameters. Furthermore,ipip6_tunnel_update()currently unhashes, re-hashes, and callssynchronize_net()unconditionally, even when the tunnel endpoint addresses (saddr and daddr) have not changed.
Quoted 1 line
[...]
Here is the summary with links:
- [net-next,1/9] sit: fix UAF in ipip6_tunnel_del_prl()
https://git.kernel.org/netdev/net-next/c/5145bb6a2c26 - [net-next,2/9] sit: charge ip_tunnel_prl_entry allocations to memcg https://git.kernel.org/netdev/net-next/c/40a2b90f5190 - [net-next,3/9] ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup https://git.kernel.org/netdev/net-next/c/88b84cae6b94 - [net-next,4/9] sit: annotate data-races around tunnel->fwmark https://git.kernel.org/netdev/net-next/c/b5e8eadb1c8a - [net-next,5/9] sit: convert 6RD configuration to RCU protection https://git.kernel.org/netdev/net-next/c/eaa2098a94cd - [net-next,6/9] sit: implement ipip6_get_iflink() https://git.kernel.org/netdev/net-next/c/e3434672b7ad - [net-next,7/9] sit: dynamically allocate struct ip_tunnel_parm_kern https://git.kernel.org/netdev/net-next/c/3cd52b7c0bbf - [net-next,8/9] sit: convert configuration to RCU protection https://git.kernel.org/netdev/net-next/c/0301127e4d98 - [net-next,9/9] sit: no longer rely on RTNL in ipip6_fill_info() https://git.kernel.org/netdev/net-next/c/f712bf6f17d8
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html