netdev - Re: [PATCH net-next v2] net: Add TCP_FORCE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3e780f88-41df-413c-7a81-6a63fd750605@gmail.com>
Date:   Thu, 23 Apr 2020 06:40:54 -0700
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Cambda Zhu <cambda@...ux.alibaba.com>,
        netdev <netdev@...r.kernel.org>,
        Eric Dumazet <eric.dumazet@...il.com>
Cc:     Dust Li <dust.li@...ux.alibaba.com>,
        Tony Lu <tonylu@...ux.alibaba.com>
Subject: Re: [PATCH net-next v2] net: Add TCP_FORCE_LINGER2 to TCP setsockopt



On 4/23/20 12:35 AM, Cambda Zhu wrote:
> This patch adds a new TCP socket option named TCP_FORCE_LINGER2. The
> option has same behavior as TCP_LINGER2, except the tp->linger2 value
> can be greater than sysctl_tcp_fin_timeout if the user_ns is capable
> with CAP_NET_ADMIN.
> 
> As a server, different sockets may need different FIN-WAIT timeout and
> in most cases the system default value will be used. The timeout can
> be adjusted by setting TCP_LINGER2 but cannot be greater than the
> system default value. If one socket needs a timeout greater than the
> default, we have to adjust the sysctl which affects all sockets using
> the system default value. And if we want to adjust it for just one
> socket and keep the original value for others, all the other sockets
> have to set TCP_LINGER2. But with TCP_FORCE_LINGER2, the net admin can
> set greater tp->linger2 than the default for one socket and keep
> the sysctl_tcp_fin_timeout unchanged.
> 
> Signed-off-by: Cambda Zhu <cambda@...ux.alibaba.com>
> ---
>  Changes in v2:
>    - Add int overflow check.
> 
>  include/uapi/linux/capability.h |  1 +
>  include/uapi/linux/tcp.h        |  1 +
>  net/ipv4/tcp.c                  | 11 +++++++++++
>  3 files changed, 13 insertions(+)
> 
> diff --git a/include/uapi/linux/capability.h b/include/uapi/linux/capability.h
> index 272dc69fa080..0e30c9756a04 100644
> --- a/include/uapi/linux/capability.h
> +++ b/include/uapi/linux/capability.h
> @@ -199,6 +199,7 @@ struct vfs_ns_cap_data {
>  /* Allow multicasting */
>  /* Allow read/write of device-specific registers */
>  /* Allow activation of ATM control sockets */
> +/* Allow setting TCP_LINGER2 regardless of sysctl_tcp_fin_timeout */
>  
>  #define CAP_NET_ADMIN        12
>  
> diff --git a/include/uapi/linux/tcp.h b/include/uapi/linux/tcp.h
> index f2acb2566333..e21e0ce98ca1 100644
> --- a/include/uapi/linux/tcp.h
> +++ b/include/uapi/linux/tcp.h
> @@ -128,6 +128,7 @@ enum {
>  #define TCP_CM_INQ		TCP_INQ
>  
>  #define TCP_TX_DELAY		37	/* delay outgoing packets by XX usec */
> +#define TCP_FORCE_LINGER2	38	/* Set TCP_LINGER2 regardless of sysctl_tcp_fin_timeout */
>  
>  
>  #define TCP_REPAIR_ON		1
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 6d87de434377..d8cd1fd66bc1 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -3149,6 +3149,17 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
>  			tcp_enable_tx_delay();
>  		tp->tcp_tx_delay = val;
>  		break;
> +	case TCP_FORCE_LINGER2:
> +		if (val < 0)
> +			tp->linger2 = -1;
> +		else if (val > INT_MAX / HZ)
> +			err = -EINVAL;
> +		else if (val > net->ipv4.sysctl_tcp_fin_timeout / HZ &&
> +			 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
> +			tp->linger2 = 0;
> +		else
> +			tp->linger2 = val * HZ;
> +		break;
>  	default:
>  		err = -ENOPROTOOPT;
>  		break;
> 

INT_MAX looks quite 

Anyway, I do not think we need a new socket option, since really it will need documentation and add more confusion.

net->ipv4.sysctl_tcp_fin_timeout is the default value for sockets which have tp->linger2 cleared.

Fact that it has been used to cap TCP_LINGER2 was probably a mistake.

What about adding a new define and simply let TCP_LINGER2 use it ?

Really there is no point trying to allow hours or even days for FIN timeout,
and no point limiting a socket from having a value between net->ipv4.sysctl_tcp_fin_timeout and 2 minutes,
at least from security perspective, these values seem legal as far as TCP specs are concerned.



diff --git a/include/net/tcp.h b/include/net/tcp.h
index dcf9a72eeaa6912202e8a1ca6cf800f7401bf517..8dceea9ae87712613243a48edddd83d9ac629295 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -126,6 +126,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
                                  * to combine FIN-WAIT-2 timeout with
                                  * TIME-WAIT timer.
                                  */
+#define TCP_FIN_TIMEOUT_MAX (120 * HZ) /* max TCP_LINGER2 value (two minutes) */
 
 #define TCP_DELACK_MAX ((unsigned)(HZ/5))      /* maximal time to delay before sending an ACK */
 #if HZ >= 100
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 6d87de434377e3741314772e5fd866de1c599108..a723fec8704ba4dff235818622e52d67ec9ef489 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -3035,7 +3035,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
        case TCP_LINGER2:
                if (val < 0)
                        tp->linger2 = -1;
-               else if (val > net->ipv4.sysctl_tcp_fin_timeout / HZ)
+               else if (val > TCP_FIN_TIMEOUT_MAX / HZ)
                        tp->linger2 = 0;
                else
                        tp->linger2 = val * HZ;