[dpdk-dev] [PATCH 0/2] prevent out of bounds read with checksum

[dpdk-dev] [PATCH 0/2] prevent out of bounds read with checksum

[Bruce Richardson]

The functions for checksumming the packet payload don't perform bounds
checks, and are used by the TAP driver which does not do any bounds checks
on the incoming packet either. This means a packet received with an
incorrect IP header can read beyond the end of the mbuf.

In the worst case, where the length is specified as being smaller than the
IPv4 header, 32-bit wrap-around on subtraction occurs, meaning that approx
4GB of memory will be read.

To fix this, we can introduce a sanity check into the ipv4 function to
ensure that underflow does not occur. Since the checksum function does not
take the mbuf length as a parameter, we cannot check for overflow there,
so we instead perform the checks in the TAP driver directly.

Ideally, in a future release, all checksum functions should be modified to
take a max buffer length parameter to fix this issue globally.

NOTE: It appears that the dpaa driver also uses these functions, but from
what I can see there, they are only used on TX, which means that there
should be less need for parameter length checking, as the data does not
come from an untrusted source. Perhaps maintainers, Hemant and Shreyansh,
can confirm?

CC: Hemant Agrawal <hemant.agrawal@nxp.com>
CC: Shreyansh Jain <shreyansh.jain@nxp.com>

Bruce Richardson (2):
  net: fix underflow for checksum of invalid IPv4 packets
  net/tap: add buffer overflow checks before checksum

 drivers/net/tap/rte_eth_tap.c | 14 ++++++++++++++
 lib/librte_net/rte_ip.h       | 12 ++++++++----
 2 files changed, 22 insertions(+), 4 deletions(-)

-- 
2.19.2

[dpdk-dev] [PATCH 1/2] net: fix underflow for checksum of invalid IPv4 packets

[Bruce Richardson]

If we receive a packet with an invalid IP header, where the total packet
length is reported as less than the IP header length, we would end up
getting an underflow in the length subtraction. This could cause us to
checksum e.g. 4GB of data in the case where the result of the subtraction
was -1. We fix this by having the function return 0 - an invalid sum - when
the length is less than the header length.

CC: stable@dpdk.org
Fixes: af75078fece3 ("first public release")
Fixes: 6006818cfb26 ("net: new checksum functions")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 lib/librte_net/rte_ip.h | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/lib/librte_net/rte_ip.h b/lib/librte_net/rte_ip.h
index f2a8904a2..f9b909090 100644
--- a/lib/librte_net/rte_ip.h
+++ b/lib/librte_net/rte_ip.h
@@ -310,16 +310,20 @@ rte_ipv4_phdr_cksum(const struct ipv4_hdr *ipv4_hdr, uint64_t ol_flags)
  * @param l4_hdr
  *   The pointer to the beginning of the L4 header.
  * @return
- *   The complemented checksum to set in the IP packet.
+ *   The complemented checksum to set in the IP packet
+ *   or 0 on error
  */
 static inline uint16_t
 rte_ipv4_udptcp_cksum(const struct ipv4_hdr *ipv4_hdr, const void *l4_hdr)
 {
 	uint32_t cksum;
-	uint32_t l4_len;
+	uint32_t l3_len, l4_len;
+
+	l3_len = rte_be_to_cpu_16(ipv4_hdr->total_length);
+	if (l3_len < sizeof(struct ipv4_hdr))
+		return 0;
 
-	l4_len = (uint32_t)(rte_be_to_cpu_16(ipv4_hdr->total_length) -
-		sizeof(struct ipv4_hdr));
+	l4_len = l3_len - sizeof(struct ipv4_hdr);
 
 	cksum = rte_raw_cksum(l4_hdr, l4_len);
 	cksum += rte_ipv4_phdr_cksum(ipv4_hdr, 0);
-- 
2.19.2

[dpdk-dev] [PATCH 2/2] net/tap: add buffer overflow checks before checksum

[Bruce Richardson]

The checksum calculation APIs take only the packet headers pointers as
parameters, so they assume that the lengths reported in those headers are
correct. However, a malicious packet could claim to be far larger than it
is, so we need to check the header lengths in the driver before calling
the checksum API.

A better fix would be to allow the lengths to be passed into the API
function, but that would be an API break, so fixing in TAP driver for
now.

CC: stable@dpdk.org
Fixes: 8ae3023387e9 ("net/tap: add Rx/Tx checksum offload support")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/tap/rte_eth_tap.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/drivers/net/tap/rte_eth_tap.c b/drivers/net/tap/rte_eth_tap.c
index 49afd38dd..0ec030bef 100644
--- a/drivers/net/tap/rte_eth_tap.c
+++ b/drivers/net/tap/rte_eth_tap.c
@@ -281,13 +281,27 @@ tap_verify_csum(struct rte_mbuf *mbuf)
 		l3_len = 4 * (iph->version_ihl & 0xf);
 		if (unlikely(l2_len + l3_len > rte_pktmbuf_data_len(mbuf)))
 			return;
+		/* check that the total length reported by header is not
+		 * greater than the total received size
+		 */
+		if (l2_len + rte_be_to_cpu_16(iph->total_length) >
+				rte_pktmbuf_data_len(mbuf))
+			return;
 
 		cksum = ~rte_raw_cksum(iph, l3_len);
 		mbuf->ol_flags |= cksum ?
 			PKT_RX_IP_CKSUM_BAD :
 			PKT_RX_IP_CKSUM_GOOD;
 	} else if (l3 == RTE_PTYPE_L3_IPV6) {
+		struct ipv6_hdr *iph = l3_hdr;
+
 		l3_len = sizeof(struct ipv6_hdr);
+		/* check that the total length reported by header is not
+		 * greater than the total received size
+		 */
+		if (l2_len + l3_len + rte_be_to_cpu_16(iph->payload_len) >
+				rte_pktmbuf_data_len(mbuf))
+			return;
 	} else {
 		/* IPv6 extensions are not supported */
 		return;
-- 
2.19.2

Re: [dpdk-dev] [PATCH 0/2] prevent out of bounds read with checksum

[Hemant Agrawal]

HI Bruce,

On 17-Dec-18 9:20 PM, Bruce Richardson wrote:
> The functions for checksumming the packet payload don't perform bounds
> checks, and are used by the TAP driver which does not do any bounds checks
> on the incoming packet either. This means a packet received with an
> incorrect IP header can read beyond the end of the mbuf.
>
> In the worst case, where the length is specified as being smaller than the
> IPv4 header, 32-bit wrap-around on subtraction occurs, meaning that approx
> 4GB of memory will be read.
>
> To fix this, we can introduce a sanity check into the ipv4 function to
> ensure that underflow does not occur. Since the checksum function does not
> take the mbuf length as a parameter, we cannot check for overflow there,
> so we instead perform the checks in the TAP driver directly.
>
> Ideally, in a future release, all checksum functions should be modified to
> take a max buffer length parameter to fix this issue globally.
>
> NOTE: It appears that the dpaa driver also uses these functions, but from
> what I can see there, they are only used on TX, which means that there
> should be less need for parameter length checking, as the data does not
> come from an untrusted source. Perhaps maintainers, Hemant and Shreyansh,
> can confirm?

In DPAA, we are using software based checksum calculation for self 
generated packets largely.

They are mostly trust worthy unless someone is deliberately or 
mistakenly trying to send a corrupt packet.

We will check, if we can also add some checks in DPAA driver in these 
legs without making performance impact for self generated packets.


Regards,

Hemant

>
> CC: Hemant Agrawal <hemant.agrawal@nxp.com>
> CC: Shreyansh Jain <shreyansh.jain@nxp.com>
>
> Bruce Richardson (2):
>    net: fix underflow for checksum of invalid IPv4 packets
>    net/tap: add buffer overflow checks before checksum
>
>   drivers/net/tap/rte_eth_tap.c | 14 ++++++++++++++
>   lib/librte_net/rte_ip.h       | 12 ++++++++----
>   2 files changed, 22 insertions(+), 4 deletions(-)
>

Re: [dpdk-dev] [PATCH 0/2] prevent out of bounds read with checksum

[Richardson, Bruce]

> -----Original Message-----
> From: Hemant Agrawal [mailto:hemant.agrawal@nxp.com]
> Sent: Tuesday, December 18, 2018 12:50 PM
> To: Richardson, Bruce <bruce.richardson@intel.com>; Olivier Matz
> <olivier.matz@6wind.com>; Wiles, Keith <keith.wiles@intel.com>
> Cc: dev@dpdk.org; Shreyansh Jain <shreyansh.jain@nxp.com>
> Subject: Re: [dpdk-dev] [PATCH 0/2] prevent out of bounds read with
> checksum
> 
> HI Bruce,
> 
> On 17-Dec-18 9:20 PM, Bruce Richardson wrote:
> > The functions for checksumming the packet payload don't perform bounds
> > checks, and are used by the TAP driver which does not do any bounds
> > checks on the incoming packet either. This means a packet received
> > with an incorrect IP header can read beyond the end of the mbuf.
> >
> > In the worst case, where the length is specified as being smaller than
> > the
> > IPv4 header, 32-bit wrap-around on subtraction occurs, meaning that
> > approx 4GB of memory will be read.
> >
> > To fix this, we can introduce a sanity check into the ipv4 function to
> > ensure that underflow does not occur. Since the checksum function does
> > not take the mbuf length as a parameter, we cannot check for overflow
> > there, so we instead perform the checks in the TAP driver directly.
> >
> > Ideally, in a future release, all checksum functions should be
> > modified to take a max buffer length parameter to fix this issue
> globally.
> >
> > NOTE: It appears that the dpaa driver also uses these functions, but
> > from what I can see there, they are only used on TX, which means that
> > there should be less need for parameter length checking, as the data
> > does not come from an untrusted source. Perhaps maintainers, Hemant
> > and Shreyansh, can confirm?
> 
> In DPAA, we are using software based checksum calculation for self
> generated packets largely.
> 
> They are mostly trust worthy unless someone is deliberately or mistakenly
> trying to send a corrupt packet.
> 
> We will check, if we can also add some checks in DPAA driver in these legs
> without making performance impact for self generated packets.
> 

Right. Thanks for confirming it's not on RX path which would be the main risk.
I would assume that data coming from the app should be trusted, unless the
app is deliberately trying to crash itself. :-) (I didn't look to try and fix this
in DPAA because of that assumption, but glad you are looking into it.)

Re: [dpdk-dev] [PATCH 1/2] net: fix underflow for checksum of invalid IPv4 packets

[Hemant Agrawal]

Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com<mailto:hemant.agrawal@nxp.com>>

Re: [dpdk-dev] [PATCH 1/2] net: fix underflow for checksum of invalid IPv4 packets

[Hemant Agrawal]

After fixing my mail client issues.

Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com><mailto:hemant.agrawal@nxp.com>


On 18-Dec-18 6:45 PM, Hemant Agrawal wrote:

Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com<mailto:hemant.agrawal@nxp.com><mailto:hemant.agrawal@nxp.com><mailto:hemant.agrawal@nxp.com>>

Re: [dpdk-dev] [PATCH 2/2] net/tap: add buffer overflow checks before checksum

[Ferruh Yigit]

On 12/17/2018 3:50 PM, Bruce Richardson wrote:
> The checksum calculation APIs take only the packet headers pointers as
> parameters, so they assume that the lengths reported in those headers are
> correct. However, a malicious packet could claim to be far larger than it
> is, so we need to check the header lengths in the driver before calling
> the checksum API.
> 
> A better fix would be to allow the lengths to be passed into the API
> function, but that would be an API break, so fixing in TAP driver for
> now.
> 
> CC: stable@dpdk.org
> Fixes: 8ae3023387e9 ("net/tap: add Rx/Tx checksum offload support")
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

Re: [dpdk-dev] [PATCH 0/2] prevent out of bounds read with checksum

[Ferruh Yigit]

On 12/17/2018 3:50 PM, Bruce Richardson wrote:
> The functions for checksumming the packet payload don't perform bounds
> checks, and are used by the TAP driver which does not do any bounds checks
> on the incoming packet either. This means a packet received with an
> incorrect IP header can read beyond the end of the mbuf.
> 
> In the worst case, where the length is specified as being smaller than the
> IPv4 header, 32-bit wrap-around on subtraction occurs, meaning that approx
> 4GB of memory will be read.
> 
> To fix this, we can introduce a sanity check into the ipv4 function to
> ensure that underflow does not occur. Since the checksum function does not
> take the mbuf length as a parameter, we cannot check for overflow there,
> so we instead perform the checks in the TAP driver directly.
> 
> Ideally, in a future release, all checksum functions should be modified to
> take a max buffer length parameter to fix this issue globally.
> 
> NOTE: It appears that the dpaa driver also uses these functions, but from
> what I can see there, they are only used on TX, which means that there
> should be less need for parameter length checking, as the data does not
> come from an untrusted source. Perhaps maintainers, Hemant and Shreyansh,
> can confirm?
> 
> CC: Hemant Agrawal <hemant.agrawal@nxp.com>
> CC: Shreyansh Jain <shreyansh.jain@nxp.com>
> 
> Bruce Richardson (2):
>   net: fix underflow for checksum of invalid IPv4 packets
>   net/tap: add buffer overflow checks before checksum

Series applied to dpdk-next-net/master, thanks.

Re: [dpdk-dev] [PATCH 2/2] net/tap: add buffer overflow checks before checksum

[Wiles, Keith]

> On Dec 17, 2018, at 9:50 AM, Richardson, Bruce <bruce.richardson@intel.com> wrote:
> 
> The checksum calculation APIs take only the packet headers pointers as
> parameters, so they assume that the lengths reported in those headers are
> correct. However, a malicious packet could claim to be far larger than it
> is, so we need to check the header lengths in the driver before calling
> the checksum API.
> 
> A better fix would be to allow the lengths to be passed into the API
> function, but that would be an API break, so fixing in TAP driver for
> now.
> 
> CC: stable@dpdk.org
> Fixes: 8ae3023387e9 ("net/tap: add Rx/Tx checksum offload support")
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> —

Acked-by: Keith Wiles <keith.wiles@intel.com>

Regards,
Keith