Fuzzers and static checkers might not detect when tcp_sk() is used with
a non tcp_sock structure.
This kind of mistake already happened a few times with MPTCP: when
wrongly using TCP-specific helpers with mptcp_sock pointers. On the
other hand, there are many 'tcp_xxx()' helpers that are taking a 'struct
sock' pointer as arguments, and some of them are only looking at fields
from 'struct sock', and nothing from 'struct tcp_sock'. It is then
tempting to use them with a 'struct mptcp_sock'.
So a new simple check is done when CONFIG_DEBUG_NET is enabled to tell
kernel devs when a non-TCP socket is being used as a TCP one. 'tcp_sk()'
macro is then re-defined to add a WARN when an unexpected socket is
being used.
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
Notes:
- v2:
- Move from include/linux/tcp.h to net/mptcp/protocol.h: specific
to TCP (Paolo)
- Use a macro instead of an inlined function (Paolo)
- Adapt the commit message after the recent changes.
- v3:
- add parenthesis around 'ptr' (checkpatch)
- there is still this check from checkpatch but I guess that's fine:
Macro argument reuse 'ptr' - possible side-effects?
---
net/mptcp/protocol.h | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index eefd1397106d..6b62a7f35dd9 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -348,6 +348,15 @@ static inline void msk_owned_by_me(const struct mptcp_sock *msk)
sock_owned_by_me((const struct sock *)msk);
}
+#ifdef CONFIG_DEBUG_NET
+/* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
+#undef tcp_sk
+#define tcp_sk(ptr) ({ \
+ WARN_ON((ptr)->sk_protocol != IPPROTO_TCP); \
+ container_of_const(ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \
+})
+#endif
+
#define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
/* the msk socket don't use the backlog, also account for the bulk
--
2.43.0
On Wed, 7 Feb 2024, Matthieu Baerts (NGI0) wrote: > Fuzzers and static checkers might not detect when tcp_sk() is used with > a non tcp_sock structure. > > This kind of mistake already happened a few times with MPTCP: when > wrongly using TCP-specific helpers with mptcp_sock pointers. On the > other hand, there are many 'tcp_xxx()' helpers that are taking a 'struct > sock' pointer as arguments, and some of them are only looking at fields > from 'struct sock', and nothing from 'struct tcp_sock'. It is then > tempting to use them with a 'struct mptcp_sock'. > > So a new simple check is done when CONFIG_DEBUG_NET is enabled to tell > kernel devs when a non-TCP socket is being used as a TCP one. 'tcp_sk()' > macro is then re-defined to add a WARN when an unexpected socket is > being used. > > Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> > --- > Notes: > - v2: > - Move from include/linux/tcp.h to net/mptcp/protocol.h: specific > to TCP (Paolo) > - Use a macro instead of an inlined function (Paolo) > - Adapt the commit message after the recent changes. > - v3: > - add parenthesis around 'ptr' (checkpatch) > - there is still this check from checkpatch but I guess that's fine: > Macro argument reuse 'ptr' - possible side-effects? > --- > net/mptcp/protocol.h | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h > index eefd1397106d..6b62a7f35dd9 100644 > --- a/net/mptcp/protocol.h > +++ b/net/mptcp/protocol.h > @@ -348,6 +348,15 @@ static inline void msk_owned_by_me(const struct mptcp_sock *msk) > sock_owned_by_me((const struct sock *)msk); > } > > +#ifdef CONFIG_DEBUG_NET > +/* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ > +#undef tcp_sk > +#define tcp_sk(ptr) ({ \ > + WARN_ON((ptr)->sk_protocol != IPPROTO_TCP); \ > + container_of_const(ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ > +}) Hi Matthieu - Could make checkpatch happy with: #define tcp_sk(ptr) ({ \ typeof(ptr) _ptr = (ptr); \ WARN_ON((_ptr)->sk_protocol != IPPROTO_TCP); \ container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ }) (same idea for patch 3) - Mat > +#endif > + > #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) > > /* the msk socket don't use the backlog, also account for the bulk > > -- > 2.43.0 > > >
Hi Mat, On 14/02/2024 20:03, Mat Martineau wrote: > On Wed, 7 Feb 2024, Matthieu Baerts (NGI0) wrote: > >> Fuzzers and static checkers might not detect when tcp_sk() is used with >> a non tcp_sock structure. >> >> This kind of mistake already happened a few times with MPTCP: when >> wrongly using TCP-specific helpers with mptcp_sock pointers. On the >> other hand, there are many 'tcp_xxx()' helpers that are taking a 'struct >> sock' pointer as arguments, and some of them are only looking at fields >> from 'struct sock', and nothing from 'struct tcp_sock'. It is then >> tempting to use them with a 'struct mptcp_sock'. >> >> So a new simple check is done when CONFIG_DEBUG_NET is enabled to tell >> kernel devs when a non-TCP socket is being used as a TCP one. 'tcp_sk()' >> macro is then re-defined to add a WARN when an unexpected socket is >> being used. >> >> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> >> --- >> Notes: >> - v2: >> - Move from include/linux/tcp.h to net/mptcp/protocol.h: specific >> to TCP (Paolo) >> - Use a macro instead of an inlined function (Paolo) >> - Adapt the commit message after the recent changes. >> - v3: >> - add parenthesis around 'ptr' (checkpatch) >> - there is still this check from checkpatch but I guess that's fine: >> Macro argument reuse 'ptr' - possible side-effects? >> --- >> net/mptcp/protocol.h | 9 +++++++++ >> 1 file changed, 9 insertions(+) >> >> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h >> index eefd1397106d..6b62a7f35dd9 100644 >> --- a/net/mptcp/protocol.h >> +++ b/net/mptcp/protocol.h >> @@ -348,6 +348,15 @@ static inline void msk_owned_by_me(const struct >> mptcp_sock *msk) >> sock_owned_by_me((const struct sock *)msk); >> } >> >> +#ifdef CONFIG_DEBUG_NET >> +/* MPTCP-specific: we might (indirectly) call this helper with the >> wrong sk */ >> +#undef tcp_sk >> +#define tcp_sk(ptr) ({ \ >> + WARN_ON((ptr)->sk_protocol != IPPROTO_TCP); \ >> + container_of_const(ptr, struct tcp_sock, >> inet_conn.icsk_inet.sk); \ >> +}) > > Hi Matthieu - > > Could make checkpatch happy with: > > #define tcp_sk(ptr) ({ \ > typeof(ptr) _ptr = (ptr); \ > WARN_ON((_ptr)->sk_protocol != IPPROTO_TCP); \ > container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ > }) > > (same idea for patch 3) Good idea! Better be safe than sorry :) I just sent a v4. While at it, I also removed the extra parenthesis around '_ptr' in 'WARN_ON((_ptr)...)': they are no longer needed now. Cheers, Matt -- Sponsored by the NGI0 Core fund.
© 2016 - 2024 Red Hat, Inc.