| /openbmc/linux/net/ipv6/ |
| H A D | udp_impl.h | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| H A D | udplite.c | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| H A D | af_inet6.c | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| H A D | udp.c | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| /openbmc/linux/include/net/ |
| H A D | udplite.h | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| H A D | udp.h | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| H A D | ipv6.h | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| /openbmc/linux/net/ipv4/ |
| H A D | udplite.c | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
| H A D | udp.c | d38afeec Thu Oct 06 13:53:47 CDT 2022 Kuniyuki Iwashima <kuniyu@amazon.com> tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct().
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|