2018-07-16 11:58:28

by Ursula Braun

Subject: [PATCH net V2 1/1] net/smc: take sock lock in smc_ioctl()

From: Ursula Braun <[email protected]>

SMC ioctl processing requires the sock lock to work properly in
all thinkable scenarios.
Problem has been found with RaceFuzzer and fixes:
KASAN: null-ptr-deref Read in smc_ioctl

Reported-by: Byoungyoung Lee <[email protected]>
Reported-by: [email protected]
Signed-off-by: Ursula Braun <[email protected]>
---
net/smc/af_smc.c | 19 +++++++++++++++----
1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 5334157f5065..c12a7fc18f56 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -1524,10 +1524,13 @@ static int smc_ioctl(struct socket *sock, unsigned int cmd,
return -EBADF;
return smc->clcsock->ops->ioctl(smc->clcsock, cmd, arg);
}
+ lock_sock(&smc->sk);
switch (cmd) {
case SIOCINQ: /* same as FIONREAD */
- if (smc->sk.sk_state == SMC_LISTEN)
+ if (smc->sk.sk_state == SMC_LISTEN) {
+ release_sock(&smc->sk);
return -EINVAL;
+ }
if (smc->sk.sk_state == SMC_INIT ||
smc->sk.sk_state == SMC_CLOSED)
answ = 0;
@@ -1536,8 +1539,10 @@ static int smc_ioctl(struct socket *sock, unsigned int cmd,
break;
case SIOCOUTQ:
/* output queue size (not send + not acked) */
- if (smc->sk.sk_state == SMC_LISTEN)
+ if (smc->sk.sk_state == SMC_LISTEN) {
+ release_sock(&smc->sk);
return -EINVAL;
+ }
if (smc->sk.sk_state == SMC_INIT ||
smc->sk.sk_state == SMC_CLOSED)
answ = 0;
@@ -1547,8 +1552,10 @@ static int smc_ioctl(struct socket *sock, unsigned int cmd,
break;
case SIOCOUTQNSD:
/* output queue size (not send only) */
- if (smc->sk.sk_state == SMC_LISTEN)
+ if (smc->sk.sk_state == SMC_LISTEN) {
+ release_sock(&smc->sk);
return -EINVAL;
+ }
if (smc->sk.sk_state == SMC_INIT ||
smc->sk.sk_state == SMC_CLOSED)
answ = 0;
@@ -1556,8 +1563,10 @@ static int smc_ioctl(struct socket *sock, unsigned int cmd,
answ = smc_tx_prepared_sends(&smc->conn);
break;
case SIOCATMARK:
- if (smc->sk.sk_state == SMC_LISTEN)
+ if (smc->sk.sk_state == SMC_LISTEN) {
+ release_sock(&smc->sk);
return -EINVAL;
+ }
if (smc->sk.sk_state == SMC_INIT ||
smc->sk.sk_state == SMC_CLOSED) {
answ = 0;
@@ -1573,8 +1582,10 @@ static int smc_ioctl(struct socket *sock, unsigned int cmd,
}
break;
default:
+ release_sock(&smc->sk);
return -ENOIOCTLCMD;
}
+ release_sock(&smc->sk);

return put_user(answ, (int __user *)arg);
}
--
2.16.4

2018-07-16 13:15:44

by Stefano Brivio

Subject: Re: [PATCH net V2 1/1] net/smc: take sock lock in smc_ioctl()

On Mon, 16 Jul 2018 13:56:52 +0200
Ursula Braun <[email protected]> wrote:

> From: Ursula Braun <[email protected]>
>
> SMC ioctl processing requires the sock lock to work properly in
> all thinkable scenarios.
> Problem has been found with RaceFuzzer and fixes:
> KASAN: null-ptr-deref Read in smc_ioctl
>
> Reported-by: Byoungyoung Lee <[email protected]>
> Reported-by: [email protected]
> Signed-off-by: Ursula Braun <[email protected]>

Reviewed-by: Stefano Brivio <[email protected]>

--
Stefano

2018-07-16 21:46:48

by David Miller

Subject: Re: [PATCH net V2 1/1] net/smc: take sock lock in smc_ioctl()

From: Ursula Braun <[email protected]>
Date: Mon, 16 Jul 2018 13:56:52 +0200

> From: Ursula Braun <[email protected]>
>
> SMC ioctl processing requires the sock lock to work properly in
> all thinkable scenarios.
> Problem has been found with RaceFuzzer and fixes:
> KASAN: null-ptr-deref Read in smc_ioctl
>
> Reported-by: Byoungyoung Lee <[email protected]>
> Reported-by: [email protected]
> Signed-off-by: Ursula Braun <[email protected]>

Applied and queued up for -stable.