2019-06-18 09:32:39

by Ard Biesheuvel

[permalink] [raw]
Subject: [PATCH 0/2] net: fastopen: follow-up tweaks for SipHash switch

A pair of tweaks for issues spotted by Eric Biggers. Patch #1 is
mostly cosmetic, since the error check it adds is unreachable in
practice, and the other changes are syntactic cleanups. Patch #2
adds endian swabbing of the SipHash output for big endian systems
so that the in-memory representation is the same as on little
endian systems.

cc: Eric Biggers <[email protected]>
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]

Ard Biesheuvel (2):
net: fastopen: make key handling more robust against future changes
net: fastopen: use endianness agnostic representation of the cookie

include/linux/tcp.h | 2 +-
include/net/tcp.h | 5 +--
net/ipv4/tcp_fastopen.c | 34 +++++++++++---------
3 files changed, 23 insertions(+), 18 deletions(-)

--
2.17.1


2019-06-18 09:32:44

by Ard Biesheuvel

[permalink] [raw]
Subject: [PATCH 2/2] net: fastopen: use endianness agnostic representation of the cookie

Use an explicit little endian representation of the fastopen
cookie, so that the value no longer depends on the endianness
of the system. This fixes a theoretical issue only, since
fastopen keys are unlikely to be shared across load balancing
server farms that are mixed in endiannes, but it might pop up
in validation/selftests as well, so let's just settle on little
endian across the board.

Note that this change only affects big endian systems.

Signed-off-by: Ard Biesheuvel <[email protected]>
---
include/linux/tcp.h | 2 +-
net/ipv4/tcp_fastopen.c | 16 ++++++++--------
2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 3d3659c638a6..f3a85a7fb4b1 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -58,7 +58,7 @@ static inline unsigned int tcp_optlen(const struct sk_buff *skb)

/* TCP Fast Open Cookie as stored in memory */
struct tcp_fastopen_cookie {
- u64 val[DIV_ROUND_UP(TCP_FASTOPEN_COOKIE_MAX, sizeof(u64))];
+ __le64 val[DIV_ROUND_UP(TCP_FASTOPEN_COOKIE_MAX, sizeof(u64))];
s8 len;
bool exp; /* In RFC6994 experimental option format */
};
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index 61c15c3d3584..2704441a0bb0 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -123,10 +123,10 @@ static bool __tcp_fastopen_cookie_gen_cipher(struct request_sock *req,
if (req->rsk_ops->family == AF_INET) {
const struct iphdr *iph = ip_hdr(syn);

- foc->val[0] = siphash(&iph->saddr,
- sizeof(iph->saddr) +
- sizeof(iph->daddr),
- key);
+ foc->val[0] = cpu_to_le64(siphash(&iph->saddr,
+ sizeof(iph->saddr) +
+ sizeof(iph->daddr),
+ key));
foc->len = TCP_FASTOPEN_COOKIE_SIZE;
return true;
}
@@ -134,10 +134,10 @@ static bool __tcp_fastopen_cookie_gen_cipher(struct request_sock *req,
if (req->rsk_ops->family == AF_INET6) {
const struct ipv6hdr *ip6h = ipv6_hdr(syn);

- foc->val[0] = siphash(&ip6h->saddr,
- sizeof(ip6h->saddr) +
- sizeof(ip6h->daddr),
- key);
+ foc->val[0] = cpu_to_le64(siphash(&ip6h->saddr,
+ sizeof(ip6h->saddr) +
+ sizeof(ip6h->daddr),
+ key));
foc->len = TCP_FASTOPEN_COOKIE_SIZE;
return true;
}
--
2.17.1

2019-06-18 09:32:45

by Ard Biesheuvel

[permalink] [raw]
Subject: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

Some changes to the TCP fastopen code to make it more robust
against future changes in the choice of key/cookie size, etc.

- Instead of keeping the SipHash key in an untyped u8[] buffer
and casting it to the right type upon use, use the correct
siphash_key_t type directly. This ensures that the key will
appear at the correct alignment if we ever change the way
these data structures are allocated. (Currently, they are
only allocated via kmalloc so they always appear at the
correct alignment)

- Use DIV_ROUND_UP when sizing the u64[] array to hold the
cookie, so it is always of sufficient size, even when
TCP_FASTOPEN_COOKIE_MAX is no longer a multiple of 8.

- Add a key length check to tcp_fastopen_reset_cipher(). No
callers exist currently that fail this check (they all pass
compile constant values that equal TCP_FASTOPEN_KEY_LENGTH),
but future changes might create problems, e.g., by leaving part
of the key uninitialized, or overflowing the key buffers.

Note that none of these are functional changes wrt the current
state of the code.

Signed-off-by: Ard Biesheuvel <[email protected]>
---
include/linux/tcp.h | 2 +-
include/net/tcp.h | 5 +++--
net/ipv4/tcp_fastopen.c | 22 ++++++++++++--------
3 files changed, 17 insertions(+), 12 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2689b0b0b68a..3d3659c638a6 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -58,7 +58,7 @@ static inline unsigned int tcp_optlen(const struct sk_buff *skb)

/* TCP Fast Open Cookie as stored in memory */
struct tcp_fastopen_cookie {
- u64 val[TCP_FASTOPEN_COOKIE_MAX / sizeof(u64)];
+ u64 val[DIV_ROUND_UP(TCP_FASTOPEN_COOKIE_MAX, sizeof(u64))];
s8 len;
bool exp; /* In RFC6994 experimental option format */
};
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 573c9e9b0d72..9456b0834e21 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -43,6 +43,7 @@
#include <linux/seq_file.h>
#include <linux/memcontrol.h>
#include <linux/bpf-cgroup.h>
+#include <linux/siphash.h>

extern struct inet_hashinfo tcp_hashinfo;

@@ -1623,14 +1624,14 @@ void tcp_fastopen_init_key_once(struct net *net);
bool tcp_fastopen_cookie_check(struct sock *sk, u16 *mss,
struct tcp_fastopen_cookie *cookie);
bool tcp_fastopen_defer_connect(struct sock *sk, int *err);
-#define TCP_FASTOPEN_KEY_LENGTH 16
+#define TCP_FASTOPEN_KEY_LENGTH sizeof(siphash_key_t)
#define TCP_FASTOPEN_KEY_MAX 2
#define TCP_FASTOPEN_KEY_BUF_LENGTH \
(TCP_FASTOPEN_KEY_LENGTH * TCP_FASTOPEN_KEY_MAX)

/* Fastopen key context */
struct tcp_fastopen_context {
- __u8 key[TCP_FASTOPEN_KEY_MAX][TCP_FASTOPEN_KEY_LENGTH];
+ siphash_key_t key[TCP_FASTOPEN_KEY_MAX];
int num;
struct rcu_head rcu;
};
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index 46b67128e1ca..61c15c3d3584 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -7,7 +7,6 @@
#include <linux/tcp.h>
#include <linux/rcupdate.h>
#include <linux/rculist.h>
-#include <linux/siphash.h>
#include <net/inetpeer.h>
#include <net/tcp.h>

@@ -81,9 +80,15 @@ int tcp_fastopen_reset_cipher(struct net *net, struct sock *sk,
goto out;
}

- memcpy(ctx->key[0], primary_key, len);
+ if (unlikely(len != TCP_FASTOPEN_KEY_LENGTH)) {
+ pr_err("TCP: TFO key length %u invalid\n", len);
+ err = -EINVAL;
+ goto out;
+ }
+
+ memcpy(&ctx->key[0], primary_key, len);
if (backup_key) {
- memcpy(ctx->key[1], backup_key, len);
+ memcpy(&ctx->key[1], backup_key, len);
ctx->num = 2;
} else {
ctx->num = 1;
@@ -110,10 +115,9 @@ int tcp_fastopen_reset_cipher(struct net *net, struct sock *sk,

static bool __tcp_fastopen_cookie_gen_cipher(struct request_sock *req,
struct sk_buff *syn,
- const u8 *key,
+ const siphash_key_t *key,
struct tcp_fastopen_cookie *foc)
{
- BUILD_BUG_ON(TCP_FASTOPEN_KEY_LENGTH != sizeof(siphash_key_t));
BUILD_BUG_ON(TCP_FASTOPEN_COOKIE_SIZE != sizeof(u64));

if (req->rsk_ops->family == AF_INET) {
@@ -122,7 +126,7 @@ static bool __tcp_fastopen_cookie_gen_cipher(struct request_sock *req,
foc->val[0] = siphash(&iph->saddr,
sizeof(iph->saddr) +
sizeof(iph->daddr),
- (const siphash_key_t *)key);
+ key);
foc->len = TCP_FASTOPEN_COOKIE_SIZE;
return true;
}
@@ -133,7 +137,7 @@ static bool __tcp_fastopen_cookie_gen_cipher(struct request_sock *req,
foc->val[0] = siphash(&ip6h->saddr,
sizeof(ip6h->saddr) +
sizeof(ip6h->daddr),
- (const siphash_key_t *)key);
+ key);
foc->len = TCP_FASTOPEN_COOKIE_SIZE;
return true;
}
@@ -154,7 +158,7 @@ static void tcp_fastopen_cookie_gen(struct sock *sk,
rcu_read_lock();
ctx = tcp_fastopen_get_ctx(sk);
if (ctx)
- __tcp_fastopen_cookie_gen_cipher(req, syn, ctx->key[0], foc);
+ __tcp_fastopen_cookie_gen_cipher(req, syn, &ctx->key[0], foc);
rcu_read_unlock();
}

@@ -218,7 +222,7 @@ static int tcp_fastopen_cookie_gen_check(struct sock *sk,
if (!ctx)
goto out;
for (i = 0; i < tcp_fastopen_context_len(ctx); i++) {
- __tcp_fastopen_cookie_gen_cipher(req, syn, ctx->key[i], foc);
+ __tcp_fastopen_cookie_gen_cipher(req, syn, &ctx->key[i], foc);
if (tcp_fastopen_cookie_match(foc, orig)) {
ret = i + 1;
goto out;
--
2.17.1

2019-06-18 09:37:49

by Eric Dumazet

[permalink] [raw]
Subject: Re: [PATCH 0/2] net: fastopen: follow-up tweaks for SipHash switch

On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
<[email protected]> wrote:
>
> A pair of tweaks for issues spotted by Eric Biggers. Patch #1 is
> mostly cosmetic, since the error check it adds is unreachable in
> practice, and the other changes are syntactic cleanups. Patch #2
> adds endian swabbing of the SipHash output for big endian systems
> so that the in-memory representation is the same as on little
> endian systems.
>

Please always add net or net-next in your patches for netdev@

( Documentation/networking/netdev-FAQ.rst )

Thanks.

2019-06-18 09:39:31

by Eric Dumazet

[permalink] [raw]
Subject: Re: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
<[email protected]> wrote:
>
> Some changes to the TCP fastopen code to make it more robust
> against future changes in the choice of key/cookie size, etc.
>
> - Instead of keeping the SipHash key in an untyped u8[] buffer
> and casting it to the right type upon use, use the correct
> siphash_key_t type directly. This ensures that the key will
> appear at the correct alignment if we ever change the way
> these data structures are allocated. (Currently, they are
> only allocated via kmalloc so they always appear at the
> correct alignment)
>
> - Use DIV_ROUND_UP when sizing the u64[] array to hold the
> cookie, so it is always of sufficient size, even when
> TCP_FASTOPEN_COOKIE_MAX is no longer a multiple of 8.
>
> - Add a key length check to tcp_fastopen_reset_cipher(). No
> callers exist currently that fail this check (they all pass
> compile constant values that equal TCP_FASTOPEN_KEY_LENGTH),
> but future changes might create problems, e.g., by leaving part
> of the key uninitialized, or overflowing the key buffers.
>
> Note that none of these are functional changes wrt the current
> state of the code.
>
...

> - memcpy(ctx->key[0], primary_key, len);
> + if (unlikely(len != TCP_FASTOPEN_KEY_LENGTH)) {
> + pr_err("TCP: TFO key length %u invalid\n", len);
> + err = -EINVAL;
> + goto out;
> + }


Why a pr_err() is there ?

Can unpriv users flood the syslog ?

2019-06-18 09:40:14

by Ard Biesheuvel

[permalink] [raw]
Subject: Re: [PATCH 0/2] net: fastopen: follow-up tweaks for SipHash switch

On Tue, 18 Jun 2019 at 11:37, Eric Dumazet <[email protected]> wrote:
>
> On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
> <[email protected]> wrote:
> >
> > A pair of tweaks for issues spotted by Eric Biggers. Patch #1 is
> > mostly cosmetic, since the error check it adds is unreachable in
> > practice, and the other changes are syntactic cleanups. Patch #2
> > adds endian swabbing of the SipHash output for big endian systems
> > so that the in-memory representation is the same as on little
> > endian systems.
> >
>
> Please always add net or net-next in your patches for netdev@
>
> ( Documentation/networking/netdev-FAQ.rst )
>

Apologies. These patches are intended for net-next

2019-06-18 09:42:03

by Ard Biesheuvel

[permalink] [raw]
Subject: Re: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

On Tue, 18 Jun 2019 at 11:39, Eric Dumazet <[email protected]> wrote:
>
> On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
> <[email protected]> wrote:
> >
> > Some changes to the TCP fastopen code to make it more robust
> > against future changes in the choice of key/cookie size, etc.
> >
> > - Instead of keeping the SipHash key in an untyped u8[] buffer
> > and casting it to the right type upon use, use the correct
> > siphash_key_t type directly. This ensures that the key will
> > appear at the correct alignment if we ever change the way
> > these data structures are allocated. (Currently, they are
> > only allocated via kmalloc so they always appear at the
> > correct alignment)
> >
> > - Use DIV_ROUND_UP when sizing the u64[] array to hold the
> > cookie, so it is always of sufficient size, even when
> > TCP_FASTOPEN_COOKIE_MAX is no longer a multiple of 8.
> >
> > - Add a key length check to tcp_fastopen_reset_cipher(). No
> > callers exist currently that fail this check (they all pass
> > compile constant values that equal TCP_FASTOPEN_KEY_LENGTH),
> > but future changes might create problems, e.g., by leaving part
> > of the key uninitialized, or overflowing the key buffers.
> >
> > Note that none of these are functional changes wrt the current
> > state of the code.
> >
> ...
>
> > - memcpy(ctx->key[0], primary_key, len);
> > + if (unlikely(len != TCP_FASTOPEN_KEY_LENGTH)) {
> > + pr_err("TCP: TFO key length %u invalid\n", len);
> > + err = -EINVAL;
> > + goto out;
> > + }
>
>
> Why a pr_err() is there ?
>
> Can unpriv users flood the syslog ?

They can if they could do so before: there was a call to
crypto_cipher_setkey() in the original pre-SipHash code which would
also result in a pr_err() on an invalid key length. That call got
removed along with the AES cipher handling, and this basically
reinstates it, as suggested by EricB.

2019-06-18 09:54:42

by Eric Dumazet

[permalink] [raw]
Subject: Re: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

On Tue, Jun 18, 2019 at 2:41 AM Ard Biesheuvel
<[email protected]> wrote:
>
> On Tue, 18 Jun 2019 at 11:39, Eric Dumazet <[email protected]> wrote:
> >
> > On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
> > <[email protected]> wrote:
> > >
> > > Some changes to the TCP fastopen code to make it more robust
> > > against future changes in the choice of key/cookie size, etc.
> > >
> > > - Instead of keeping the SipHash key in an untyped u8[] buffer
> > > and casting it to the right type upon use, use the correct
> > > siphash_key_t type directly. This ensures that the key will
> > > appear at the correct alignment if we ever change the way
> > > these data structures are allocated. (Currently, they are
> > > only allocated via kmalloc so they always appear at the
> > > correct alignment)
> > >
> > > - Use DIV_ROUND_UP when sizing the u64[] array to hold the
> > > cookie, so it is always of sufficient size, even when
> > > TCP_FASTOPEN_COOKIE_MAX is no longer a multiple of 8.
> > >
> > > - Add a key length check to tcp_fastopen_reset_cipher(). No
> > > callers exist currently that fail this check (they all pass
> > > compile constant values that equal TCP_FASTOPEN_KEY_LENGTH),
> > > but future changes might create problems, e.g., by leaving part
> > > of the key uninitialized, or overflowing the key buffers.
> > >
> > > Note that none of these are functional changes wrt the current
> > > state of the code.
> > >
> > ...
> >
> > > - memcpy(ctx->key[0], primary_key, len);
> > > + if (unlikely(len != TCP_FASTOPEN_KEY_LENGTH)) {
> > > + pr_err("TCP: TFO key length %u invalid\n", len);
> > > + err = -EINVAL;
> > > + goto out;
> > > + }
> >
> >
> > Why a pr_err() is there ?
> >
> > Can unpriv users flood the syslog ?
>
> They can if they could do so before: there was a call to
> crypto_cipher_setkey() in the original pre-SipHash code which would
> also result in a pr_err() on an invalid key length. That call got
> removed along with the AES cipher handling, and this basically
> reinstates it, as suggested by EricB.

This tcp_fastopen_reset_cipher() function is internal to TCP stack, all callers
always pass the correct length.

We could add checks all over the place, and end up having a TCP stack
full of defensive
checks and 10,000 additional lines of code :/

I would prefer not reinstating this.

2019-06-18 10:03:50

by Ard Biesheuvel

[permalink] [raw]
Subject: Re: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

On Tue, 18 Jun 2019 at 11:53, Eric Dumazet <[email protected]> wrote:
>
> On Tue, Jun 18, 2019 at 2:41 AM Ard Biesheuvel
> <[email protected]> wrote:
> >
> > On Tue, 18 Jun 2019 at 11:39, Eric Dumazet <[email protected]> wrote:
> > >
> > > On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
> > > <[email protected]> wrote:
> > > >
> > > > Some changes to the TCP fastopen code to make it more robust
> > > > against future changes in the choice of key/cookie size, etc.
> > > >
> > > > - Instead of keeping the SipHash key in an untyped u8[] buffer
> > > > and casting it to the right type upon use, use the correct
> > > > siphash_key_t type directly. This ensures that the key will
> > > > appear at the correct alignment if we ever change the way
> > > > these data structures are allocated. (Currently, they are
> > > > only allocated via kmalloc so they always appear at the
> > > > correct alignment)
> > > >
> > > > - Use DIV_ROUND_UP when sizing the u64[] array to hold the
> > > > cookie, so it is always of sufficient size, even when
> > > > TCP_FASTOPEN_COOKIE_MAX is no longer a multiple of 8.
> > > >
> > > > - Add a key length check to tcp_fastopen_reset_cipher(). No
> > > > callers exist currently that fail this check (they all pass
> > > > compile constant values that equal TCP_FASTOPEN_KEY_LENGTH),
> > > > but future changes might create problems, e.g., by leaving part
> > > > of the key uninitialized, or overflowing the key buffers.
> > > >
> > > > Note that none of these are functional changes wrt the current
> > > > state of the code.
> > > >
> > > ...
> > >
> > > > - memcpy(ctx->key[0], primary_key, len);
> > > > + if (unlikely(len != TCP_FASTOPEN_KEY_LENGTH)) {
> > > > + pr_err("TCP: TFO key length %u invalid\n", len);
> > > > + err = -EINVAL;
> > > > + goto out;
> > > > + }
> > >
> > >
> > > Why a pr_err() is there ?
> > >
> > > Can unpriv users flood the syslog ?
> >
> > They can if they could do so before: there was a call to
> > crypto_cipher_setkey() in the original pre-SipHash code which would
> > also result in a pr_err() on an invalid key length. That call got
> > removed along with the AES cipher handling, and this basically
> > reinstates it, as suggested by EricB.
>
> This tcp_fastopen_reset_cipher() function is internal to TCP stack, all callers
> always pass the correct length.
>
> We could add checks all over the place, and end up having a TCP stack
> full of defensive
> checks and 10,000 additional lines of code :/
>
> I would prefer not reinstating this.

Fair enough.

2019-06-18 18:18:25

by Eric Biggers

[permalink] [raw]
Subject: Re: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

On Tue, Jun 18, 2019 at 02:53:05AM -0700, Eric Dumazet wrote:
> On Tue, Jun 18, 2019 at 2:41 AM Ard Biesheuvel
> <[email protected]> wrote:
> >
> > On Tue, 18 Jun 2019 at 11:39, Eric Dumazet <[email protected]> wrote:
> > >
> > > On Tue, Jun 18, 2019 at 2:32 AM Ard Biesheuvel
> > > <[email protected]> wrote:
> > > >
> > > > Some changes to the TCP fastopen code to make it more robust
> > > > against future changes in the choice of key/cookie size, etc.
> > > >
> > > > - Instead of keeping the SipHash key in an untyped u8[] buffer
> > > > and casting it to the right type upon use, use the correct
> > > > siphash_key_t type directly. This ensures that the key will
> > > > appear at the correct alignment if we ever change the way
> > > > these data structures are allocated. (Currently, they are
> > > > only allocated via kmalloc so they always appear at the
> > > > correct alignment)
> > > >
> > > > - Use DIV_ROUND_UP when sizing the u64[] array to hold the
> > > > cookie, so it is always of sufficient size, even when
> > > > TCP_FASTOPEN_COOKIE_MAX is no longer a multiple of 8.
> > > >
> > > > - Add a key length check to tcp_fastopen_reset_cipher(). No
> > > > callers exist currently that fail this check (they all pass
> > > > compile constant values that equal TCP_FASTOPEN_KEY_LENGTH),
> > > > but future changes might create problems, e.g., by leaving part
> > > > of the key uninitialized, or overflowing the key buffers.
> > > >
> > > > Note that none of these are functional changes wrt the current
> > > > state of the code.
> > > >
> > > ...
> > >
> > > > - memcpy(ctx->key[0], primary_key, len);
> > > > + if (unlikely(len != TCP_FASTOPEN_KEY_LENGTH)) {
> > > > + pr_err("TCP: TFO key length %u invalid\n", len);
> > > > + err = -EINVAL;
> > > > + goto out;
> > > > + }
> > >
> > >
> > > Why a pr_err() is there ?
> > >
> > > Can unpriv users flood the syslog ?
> >
> > They can if they could do so before: there was a call to
> > crypto_cipher_setkey() in the original pre-SipHash code which would
> > also result in a pr_err() on an invalid key length. That call got
> > removed along with the AES cipher handling, and this basically
> > reinstates it, as suggested by EricB.
>
> This tcp_fastopen_reset_cipher() function is internal to TCP stack, all callers
> always pass the correct length.
>
> We could add checks all over the place, and end up having a TCP stack
> full of defensive
> checks and 10,000 additional lines of code :/
>
> I would prefer not reinstating this.

The length parameter makes no sense if it's not checked, though. Either it
should exist and be checked, or it should be removed and the length should be
implicitly TCP_FASTOPEN_KEY_LENGTH.

- Eric

2019-06-18 18:19:39

by Eric Dumazet

[permalink] [raw]
Subject: Re: [PATCH 1/2] net: fastopen: make key handling more robust against future changes

On Tue, Jun 18, 2019 at 11:18 AM Eric Biggers <[email protected]> wrote:

> The length parameter makes no sense if it's not checked, though. Either it
> should exist and be checked, or it should be removed and the length should be
> implicitly TCP_FASTOPEN_KEY_LENGTH.

Sure, please send a patch.

2019-06-18 18:23:14

by Eric Biggers

[permalink] [raw]
Subject: Re: [PATCH 2/2] net: fastopen: use endianness agnostic representation of the cookie

On Tue, Jun 18, 2019 at 11:32:07AM +0200, Ard Biesheuvel wrote:
> Use an explicit little endian representation of the fastopen
> cookie, so that the value no longer depends on the endianness
> of the system. This fixes a theoretical issue only, since
> fastopen keys are unlikely to be shared across load balancing
> server farms that are mixed in endiannes, but it might pop up
> in validation/selftests as well, so let's just settle on little
> endian across the board.
>
> Note that this change only affects big endian systems.
>
> Signed-off-by: Ard Biesheuvel <[email protected]>
> ---
> include/linux/tcp.h | 2 +-
> net/ipv4/tcp_fastopen.c | 16 ++++++++--------
> 2 files changed, 9 insertions(+), 9 deletions(-)
>

What about the TCP_FASTOPEN_KEY option for setsockopt and getsockopt? Those
APIs treat the key as an array of bytes (let's say it's little endian), so
doesn't it need to be converted to/from the CPU endianness of siphash_key_t?

- Eric

2019-06-18 18:40:53

by Ard Biesheuvel

[permalink] [raw]
Subject: Re: [PATCH 2/2] net: fastopen: use endianness agnostic representation of the cookie

On Tue, 18 Jun 2019 at 20:22, Eric Biggers <[email protected]> wrote:
>
> On Tue, Jun 18, 2019 at 11:32:07AM +0200, Ard Biesheuvel wrote:
> > Use an explicit little endian representation of the fastopen
> > cookie, so that the value no longer depends on the endianness
> > of the system. This fixes a theoretical issue only, since
> > fastopen keys are unlikely to be shared across load balancing
> > server farms that are mixed in endiannes, but it might pop up
> > in validation/selftests as well, so let's just settle on little
> > endian across the board.
> >
> > Note that this change only affects big endian systems.
> >
> > Signed-off-by: Ard Biesheuvel <[email protected]>
> > ---
> > include/linux/tcp.h | 2 +-
> > net/ipv4/tcp_fastopen.c | 16 ++++++++--------
> > 2 files changed, 9 insertions(+), 9 deletions(-)
> >
>
> What about the TCP_FASTOPEN_KEY option for setsockopt and getsockopt? Those
> APIs treat the key as an array of bytes (let's say it's little endian), so
> doesn't it need to be converted to/from the CPU endianness of siphash_key_t?
>

Yeah, good point.

Can we first agree on whether we care about this or not? If so, i can spin a v2.

2019-06-18 22:41:15

by David Miller

[permalink] [raw]
Subject: Re: [PATCH 2/2] net: fastopen: use endianness agnostic representation of the cookie

From: Ard Biesheuvel <[email protected]>
Date: Tue, 18 Jun 2019 20:40:18 +0200

> Can we first agree on whether we care about this or not? If so, i
> can spin a v2.

Well, how can it possibly work otherwise in deployment scenerios involving
both big and little endian hosts?