[PATCH v3 0/2] prevent incoherent can configuration in case of early return in the CAN netlink interface

struct can_priv has a set of flags (can_priv::ctrlmode) which are
correlated with the other fields of the structure. In
can_changelink(), those flags are set first and copied to can_priv. If
the function has to return early, for example due to an out of range
value provided by the user, then the global configuration might become
incoherent.

Example: the user provides an out of range dbitrate (e.g. 20
Mbps). The command fails (-EINVAL), however the FD flag was already
set resulting in a configuration where FD is on but the databittiming
parameters are empty.

* Illustration of above example *

| $ ip link set can0 type can bitrate 500000 dbitrate 20000000 fd on
| RTNETLINK answers: Invalid argument
| $ ip --details link show can0
| 1: can0: <NOARP,ECHO> mtu 72 qdisc noop state DOWN mode DEFAULT group default qlen 10
|     link/can  promiscuity 0 minmtu 0 maxmtu 0
|     can <FD> state STOPPED restart-ms 0
           ^^ FD flag is set without any of the databittiming parameters...
|         bitrate 500000 sample-point 0.875
|         tq 12 prop-seg 69 phase-seg1 70 phase-seg2 20 sjw 1
|         ES582.1/ES584.1: tseg1 2..256 tseg2 2..128 sjw 1..128 brp 1..512 brp-inc 1
|         ES582.1/ES584.1: dtseg1 2..32 dtseg2 1..16 dsjw 1..8 dbrp 1..32 dbrp-inc 1
|         clock 80000000 numtxqueues 1 numrxqueues 1 gso_max_size 65536 gso_max_segs 65535

To prevent this from happening, we do a local copy of can_priv, work
on it, an copy it at the very end of the function (i.e. only if all
previous checks succeeded).

Once this done, there is no more need to have a temporary variable for
a specific parameter. As such, the bittiming and data bittiming (bt
and dbt) are directly written to the temporary priv variable.


N.B. The temporary can_priv is too big to be allocated on the stack
because, on x86_64 sizeof(struct can_priv) is 448 and:

| $ objdump -d drivers/net/can/dev/netlink.o | ./scripts/checkstack.pl
| 0x00000000000002100 can_changelink []:            1200


Fixes: 9859ccd2c8be ("can: introduce the data bitrate configuration for CAN FD")
Signed-off-by: Vincent Mailhol <redacted>
---
 drivers/net/can/dev/netlink.c | 32 ++++++++++++++++++--------------
 1 file changed, 18 insertions(+), 14 deletions(-)

diff --git a/drivers/net/can/dev/netlink.c b/drivers/net/can/dev/netlink.c
index 80425636049d..21b76ca8cb22 100644
--- a/drivers/net/can/dev/netlink.c
+++ b/drivers/net/can/dev/netlink.c

@@ -58,14 +58,19 @@ static int can_changelink(struct net_device *dev, struct nlattr *tb[],
                          struct nlattr *data[],
                          struct netlink_ext_ack *extack)
 {
-       struct can_priv *priv = netdev_priv(dev);
+       /* Work on a local copy of priv to prevent inconsistent value
+        * in case of early return.
+        */
+       static struct can_priv *priv;
        int err;

        /* We need synchronization with dev->stop() */
        ASSERT_RTNL();

+       priv = kmemdup(netdev_priv(dev), sizeof(*priv), GFP_KERNEL);

struct can_priv has a set of flags (can_priv::ctrlmode) which are
correlated with the other fields of the structure. In
can_changelink(), those flags are set first and copied to can_priv. If
the function has to return early, for example due to an out of range
value provided by the user, then the global configuration might become
incoherent.

Example: the user provides an out of range dbitrate (e.g. 20
Mbps). The command fails (-EINVAL), however the FD flag was already
set resulting in a configuration where FD is on but the databittiming
parameters are empty.

* Illustration of above example *

| $ ip link set can0 type can bitrate 500000 dbitrate 20000000 fd on
| RTNETLINK answers: Invalid argument
| $ ip --details link show can0
| 1: can0: <NOARP,ECHO> mtu 72 qdisc noop state DOWN mode DEFAULT group default qlen 10
|     link/can  promiscuity 0 minmtu 0 maxmtu 0
|     can <FD> state STOPPED restart-ms 0
           ^^ FD flag is set without any of the databittiming parameters...
|         bitrate 500000 sample-point 0.875
|         tq 12 prop-seg 69 phase-seg1 70 phase-seg2 20 sjw 1
|         ES582.1/ES584.1: tseg1 2..256 tseg2 2..128 sjw 1..128 brp 1..512 brp-inc 1
|         ES582.1/ES584.1: dtseg1 2..32 dtseg2 1..16 dsjw 1..8 dbrp 1..32 dbrp-inc 1
|         clock 80000000 numtxqueues 1 numrxqueues 1 gso_max_size 65536 gso_max_segs 65535

To prevent this from happening, we do a local copy of can_priv, work
on it, an copy it at the very end of the function (i.e. only if all
previous checks succeeded).

Once this done, there is no more need to have a temporary variable for
a specific parameter. As such, the bittiming and data bittiming (bt
and dbt) are directly written to the temporary priv variable.


N.B. The temporary can_priv is too big to be allocated on the stack
because, on x86_64 sizeof(struct can_priv) is 448 and:

| $ objdump -d drivers/net/can/dev/netlink.o | ./scripts/checkstack.pl
| 0x00000000000002100 can_changelink []:            1200


Fixes: 9859ccd2c8be ("can: introduce the data bitrate configuration for CAN FD")
Signed-off-by: Vincent Mailhol <redacted>
---
 drivers/net/can/dev/netlink.c | 32 ++++++++++++++++++--------------
 1 file changed, 18 insertions(+), 14 deletions(-)

diff --git a/drivers/net/can/dev/netlink.c b/drivers/net/can/dev/netlink.c
index 80425636049d..21b76ca8cb22 100644
--- a/drivers/net/can/dev/netlink.c
+++ b/drivers/net/can/dev/netlink.c

@@ -58,14 +58,19 @@ static int can_changelink(struct net_device *dev, struct nlattr *tb[],
                          struct nlattr *data[],
                          struct netlink_ext_ack *extack)
 {
-       struct can_priv *priv = netdev_priv(dev);
+       /* Work on a local copy of priv to prevent inconsistent value
+        * in case of early return.
+        */
+       static struct can_priv *priv;
        int err;

        /* We need synchronization with dev->stop() */
        ASSERT_RTNL();

+       priv = kmemdup(netdev_priv(dev), sizeof(*priv), GFP_KERNEL);
+
        if (data[IFLA_CAN_BITTIMING]) {
-               struct can_bittiming bt;
+               struct can_bittiming *bt = &priv->bittiming;

                /* Do not allow changing bittiming while running */
                if (dev->flags & IFF_UP)

@@ -79,22 +84,20 @@ static int can_changelink(struct net_device *dev, struct nlattr *tb[],
                if (!priv->bittiming_const && !priv->do_set_bittiming)
                        return -EOPNOTSUPP;

-               memcpy(&bt, nla_data(data[IFLA_CAN_BITTIMING]), sizeof(bt));
-               err = can_get_bittiming(dev, &bt,
+               memcpy(bt, nla_data(data[IFLA_CAN_BITTIMING]), sizeof(*bt));
+               err = can_get_bittiming(dev, bt,
                                        priv->bittiming_const,
                                        priv->bitrate_const,
                                        priv->bitrate_const_cnt);
                if (err)
                        return err;

-               if (priv->bitrate_max && bt.bitrate > priv->bitrate_max) {
+               if (priv->bitrate_max && bt->bitrate > priv->bitrate_max) {
                        netdev_err(dev, "arbitration bitrate surpasses transceiver capabilities of %d bps\n",
                                   priv->bitrate_max);
                        return -EINVAL;
                }

-               memcpy(&priv->bittiming, &bt, sizeof(bt));
-
                if (priv->do_set_bittiming) {
                        /* Finally, set the bit-timing registers */
                        err = priv->do_set_bittiming(dev);