Fixes for header length truncation
Subsystems (3)
Willem de Bruijn
maintainercover Fixes for header length truncation
From: Willem de Bruijn <[email protected]>
Packets should not enter the stack with truncated link layer headers and link layer headers should always be stored in the skb linear segment.
Patch 1 ensures the first for PF_PACKET sockets
Patch 2 ensures the second for PF_PACKET GSO sockets without tx_ring
Willem de Bruijn (2):
net: introduce device min_header_len
packet: round up linear to header len
drivers/net/loopback.c | 1 + include/linux/netdevice.h | 4 ++++ net/ethernet/eth.c | 1 + net/packet/af_packet.c | 7 ++++--- 4 files changed, 10 insertions(+), 3 deletions(-)
Signature · 1 line
2.11.0.483.g087da7b7c-goog
Willem de Bruijn
maintainerPATCH 1/2 net: introduce device min_header_len
From: Willem de Bruijn <[email protected]>
The stack must not pass packets to device drivers that are shorter than the minimum link layer header length.
Previously, packet sockets would drop packets smaller than or equal to dev->hard_header_len, but this has false positives. Zero length payload is used over Ethernet. Other link layer protocols support variable length headers. Support for validation of these protocols removed the min length check for all protocols.
Introduce an explicit dev->min_header_len parameter and drop all packets below this value. Initially, set it to non-zero only for Ethernet and loopback. Other protocols can follow in a patch to net-next.
--- drivers/net/loopback.c | 1 + include/linux/netdevice.h | 4 ++++ net/ethernet/eth.c | 1 + 3 files changed, 6 insertions(+)
drivers/net/loopback.c +1 -0
@@ -164,6 +164,7 @@ static void loopback_setup(struct net_device *dev) { dev->mtu =64 *1024 ; dev->hard_header_len =ETH_HLEN ;/* 14 */ + dev->min_header_len =ETH_HLEN ;/* 14 */ dev->addr_len =ETH_ALEN ;/* 6 */ dev->type =ARPHRD_LOOPBACK ;/* 0x0001*/ dev->flags =IFF_LOOPBACK ;
include/linux/netdevice.h +4 -0
@@ -1511,6 +1511,7 @@ enum netdev_priv_flags {* @max_mtu: Interface Maximum MTU value * @type: Interface hardware type * @hard_header_len: Maximum hardware header length. +* @min_header_len: Minimum hardware header length * * @needed_headroom: Extra headroom the hardware may need, but not in all * cases can this be guaranteed
@@ -1728,6 +1729,7 @@ struct net_device {unsigned int max_mtu;unsigned short type;unsigned short hard_header_len; +unsigned short min_header_len;unsigned short needed_headroom;unsigned short needed_tailroom;
@@ -2694,6 +2696,8 @@ static inline bool dev_validate_header(const struct net_device *dev, {if (likely (len >= dev->hard_header_len))return true; +if (len < dev->min_header_len) +return false;if (capable (CAP_SYS_RAWIO )) {memset (ll_header + len,0 , dev->hard_header_len - len);
net/ethernet/eth.c +1 -0
@@ -356,6 +356,7 @@ void ether_setup(struct net_device *dev) dev->header_ops = ð_header_ops; dev->type =ARPHRD_ETHER ; dev->hard_header_len =ETH_HLEN ; + dev->min_header_len =ETH_HLEN ; dev->mtu =ETH_DATA_LEN ; dev->min_mtu =ETH_MIN_MTU ; dev->max_mtu =ETH_DATA_LEN ;
Signature · 1 line
2.11.0.483.g087da7b7c-goog
Re: [PATCH net 1/2] net: introduce device min_header_len
Quoted 16 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
The stack must not pass packets to device drivers that are shorter than the minimum link layer header length.
Previously, packet sockets would drop packets smaller than or equal to
dev->hard_header_len, but this has false positives. Zero length payload is used over Ethernet. Other link layer protocols support variable length headers. Support for validation of these protocols removed the min length check for all protocols.Introduce an explicit
dev->min_header_lenparameter and drop all packets below this value. Initially, set it to non-zero only for Ethernet and loopback. Other protocols can follow in a patch to net-next.
Quoted 4 lines
Fixes 9ed988cd5915 ("packet: validate variable length ll headers")Reported-by Sowmini Varadhan <[email protected]>Signed-off-by Willem de Bruijn <[email protected]>---
Re: [PATCH net 1/2] net: introduce device min_header_len
In reply to Willem de BruijnQuoted 5 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
The stack must not pass packets to device drivers that are shorter than the minimum link layer header length.
Willem de Bruijn
maintainerPATCH 2/2 packet: round up linear to header len
In reply to Willem de BruijnFrom: Willem de Bruijn <[email protected]>
Link layer protocols may unconditionally pull headers, as Ethernet does in eth_type_trans. Ensure that the entire link layer header always lies in the skb linear segment. tpacket_snd has such a check. Extend this to packet_snd.
Variable length link layer headers complicate the computation somewhat. Here skb->len may be smaller than dev->hard_header_len.
Round up the linear length to be at least as long as the smallest of the two.
--- net/packet/af_packet.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
net/packet/af_packet.c +4 -3
@@ -2755,7 +2755,7 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len)struct virtio_net_hdr vnet_hdr = {0 };int offset =0 ;struct packet_sock *po =pkt_sk (sk); -int hlen, tlen; +int hlen, tlen, linear;int extra_len =0 ;/*
@@ -2816,8 +2816,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) err = -ENOBUFS ; hlen =LL_RESERVED_SPACE (dev); tlen = dev->needed_tailroom; - skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, -__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len), + linear =__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len); + linear =max (linear,min_t (int, len, dev->hard_header_len)); + skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, linear, msg->msg_flags &MSG_DONTWAIT , &err);if (skb ==NULL ) goto out_unlock;
Signature · 1 line
2.11.0.483.g087da7b7c-goog
Re: [PATCH net 2/2] packet: round up linear to header len
Quoted 10 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
Link layer protocols may unconditionally pull headers, as Ethernet does in
eth_type_trans. Ensure that the entire link layer header always lies in the skb linear segment.tpacket_sndhas such a check. Extend this topacket_snd.Variable length link layer headers complicate the computation somewhat. Here
skb->lenmay be smaller thandev->hard_header_len.
Quoted 6 lines
Round up the linear length to be at least as long as the smallest of the two.
Reported-by Dmitry Vyukov <[email protected]>Signed-off-by Willem de Bruijn <[email protected]>---
Re: [PATCH net 2/2] packet: round up linear to header len
In reply to Willem de Bruijnnet/packet/af_packet.c static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) · line 2822
Quoted 10 lines · Willem de Bruijn
net/packet/af_packet.c +3
-2@@ -2816,8 +2816,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) err = -ENOBUFS ; hlen =LL_RESERVED_SPACE (dev); tlen = dev->needed_tailroom; - skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, -__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len), + linear =__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len); + linear =max (linear,min_t (int, len, dev->hard_header_len)); + skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, linear, msg->msg_flags &MSG_DONTWAIT , &err);
do we need a similar check in packet_sendsmg_spkt (even if it is deprecated, would be better to get it to align with packet_snd?)
Also tpacket_fill_skb should ensure that copylen is set up like the code above?
--Sowmini
Willem de Bruijn
maintainerRe: [PATCH net 2/2] packet: round up linear to header len
Quoted 14 lines · Sowmini Varadhan
net/packet/af_packet.c static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) · line 2822
Quoted 10 lines · Willem de Bruijn
net/packet/af_packet.c +3
-2@@ -2816,8 +2816,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) err = -ENOBUFS ; hlen =LL_RESERVED_SPACE (dev); tlen = dev->needed_tailroom; - skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, -__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len), + linear =__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len); + linear =max (linear,min_t (int, len, dev->hard_header_len)); + skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, linear, msg->msg_flags &MSG_DONTWAIT , &err);do we need a similar check in
packet_sendsmg_spkt(even if it is deprecated, would be better to get it to align withpacket_snd?)
That path does not need this check, because it does not support large packets and always allocates the entire packet as linear.
Quoted 2 lines
Also
tpacket_fill_skbshould ensure that copylen is set up like the code above?
Do you mean the difference that it unconditionally pulls hard_header_len, optionally with zero padding, whereas this new path can check against new min_header_len and thus allows packets shorter than hard_header_len?
Re: [PATCH net 2/2] packet: round up linear to header len
Quoted 8 lines · Willem de Bruijn
Date: Wed, 8 Feb 2017 08:37:19 -0800
From: Willem de Bruijn <[email protected]>
To: Sowmini Varadhan <[email protected]>
Cc: Network Development <[email protected]>, David Miller
<[email protected]>, Eric Dumazet <[email protected]>, Dmitry Vyukov <[email protected]>, Willem de Bruijn <[email protected]>
Subject: Re: [PATCH net 2/2] packet: round up linear to header len
Quoted 14 lines
Quoted 12 lines · Sowmini Varadhan
Quoted 10 lines · Willem de Bruijn
net/packet/af_packet.c +3
-2@@ -2816,8 +2816,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) err = -ENOBUFS ; hlen =LL_RESERVED_SPACE (dev); tlen = dev->needed_tailroom; - skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, -__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len), + linear =__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len); + linear =max (linear,min_t (int, len, dev->hard_header_len)); + skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, linear, msg->msg_flags &MSG_DONTWAIT , &err);
:
Quoted 4 lines
Do you mean the difference that it unconditionally pulls
hard_header_len, optionally with zero padding, whereas this new path can check against newmin_header_lenand thus allows packets shorter thanhard_header_len?
yes, maybe it doesnt matter, becaues hard_header_len >= min_header_len at all times
--Sowmini
Willem de Bruijn
maintainerRe: [PATCH net 2/2] packet: round up linear to header len
Quoted 23 lines · Sowmini Varadhan
Quoted 8 lines · Willem de Bruijn
Date: Wed, 8 Feb 2017 08:37:19 -0800
From: Willem de Bruijn <[email protected]>
To: Sowmini Varadhan <[email protected]>
Cc: Network Development <[email protected]>, David Miller
<[email protected]>, Eric Dumazet <[email protected]>, Dmitry Vyukov <[email protected]>, Willem de Bruijn <[email protected]>
Subject: Re: [PATCH net 2/2] packet: round up linear to header lenQuoted 14 lines
Quoted 12 lines · Sowmini Varadhan
Quoted 10 lines · Willem de Bruijn
net/packet/af_packet.c +3
-2@@ -2816,8 +2816,9 @@ static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) err = -ENOBUFS ; hlen =LL_RESERVED_SPACE (dev); tlen = dev->needed_tailroom; - skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, -__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len), + linear =__virtio16_to_cpu (vio_le (), vnet_hdr.hdr_len); + linear =max (linear,min_t (int, len, dev->hard_header_len)); + skb =packet_alloc_skb (sk, hlen + tlen, hlen, len, linear, msg->msg_flags &MSG_DONTWAIT , &err);
Quoted 8 lines
:
Quoted 4 lines
Do you mean the difference that it unconditionally pulls
hard_header_len, optionally with zero padding, whereas this new path can check against newmin_header_lenand thus allows packets shorter thanhard_header_len?yes, maybe it doesnt matter, becaues
hard_header_len>=min_header_lenat all times
The code is not subject to this bug, so I'd rather not touch it in this fix for stable.
But you raise a good point. This logic is subtle and fragile. It will be good to deduplicate across packet_snd and tpacket_snd in a follow-up to net-next.
David Miller
maintainerFrom: Willem de Bruijn <[email protected]>
Date: Tue, 7 Feb 2017 15:57:19 -0500
Quoted 8 lines
From: Willem de Bruijn <[email protected]>
Packets should not enter the stack with truncated link layer headers and link layer headers should always be stored in the skb linear segment.
Patch 1 ensures the first for
PF_PACKETsockets
Patch 2 ensures the second forPF_PACKETGSO sockets withouttx_ring
Nice work, I'll apply this series and queue it up for -stable.
Thanks!