> On Apr 25, 2019, at 10:17 AM, Lipiec, Herakliusz
> <herakliusz.lip...@intel.com> wrote:
>
> When secondary to primary process synchronization occours
> there is no check for number of fds which could cause buffer overrun.
>
> Bugzilla ID: 252
> Fixes: c9aa56edec8e ("net/tap: access primary process queues from secondary")
> Cc: rasl...@mellanox.com
> Cc: sta...@dpdk.org
>
> Signed-off-by: Herakliusz Lipiec <herakliusz.lip...@intel.com>
> ---
> drivers/net/tap/rte_eth_tap.c | 13 +++++++++++--
> 1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/tap/rte_eth_tap.c b/drivers/net/tap/rte_eth_tap.c
> index e9fda8cf6..4a2ef5ce7 100644
> --- a/drivers/net/tap/rte_eth_tap.c
> +++ b/drivers/net/tap/rte_eth_tap.c
> @@ -2111,6 +2111,10 @@ tap_mp_attach_queues(const char *port_name, struct
> rte_eth_dev *dev)
> TAP_LOG(DEBUG, "Received IPC reply for %s", reply_param->port_name);
>
> /* Attach the queues from received file descriptors */
> + if (reply_param->rxq_count + reply_param->txq_count != reply->num_fds) {
> + TAP_LOG(ERR, "Unexpected number of fds received");
> + return -1;
> + }
This check is reasonable, but why is this being done on the receive side and
not checked on the send side. There may need to be a check for num_fds being
zero or greater than 8 as that is the limit to the number of FDs that can be
handle by the IPC.
In a different thread for Bug-258 we need to return an indicator that the
receive side detected an error by returning 0 for num_fds and I have patch for
that one.
https://bugs.dpdk.org/show_bug.cgi?id=258
I would have expected the sender to make sure they match and then this test is
not needed, but a test for num_fds being zero or > 8 is needed if you want to
detect the failure here or not if you do not care as long as nb_[r/t]x_queues
is zero too.
> dev->data->nb_rx_queues = reply_param->rxq_count;
> dev->data->nb_tx_queues = reply_param->txq_count;
> fd_iterator = 0;
> @@ -2151,12 +2155,16 @@ tap_mp_sync_queues(const struct rte_mp_msg *request,
> const void *peer)
> /* Fill file descriptors for all queues */
> reply.num_fds = 0;
> reply_param->rxq_count = 0;
> + if (dev->data->nb_rx_queues + dev->data->nb_tx_queues >
> + RTE_MP_MAX_FD_NUM){
> + TAP_LOG(ERR, "Number of rx/tx queues exceeds max number of
> fds");
> + return -1;
> + }
> for (queue = 0; queue < dev->data->nb_rx_queues; queue++) {
> reply.fds[reply.num_fds++] = process_private->rxq_fds[queue];
> reply_param->rxq_count++;
> }
> RTE_ASSERT(reply_param->rxq_count == dev->data->nb_rx_queues);
> - RTE_ASSERT(reply_param->txq_count == dev->data->nb_tx_queues);
> RTE_ASSERT(reply.num_fds <= RTE_MP_MAX_FD_NUM);
>
> reply_param->txq_count = 0;
> @@ -2164,7 +2172,8 @@ tap_mp_sync_queues(const struct rte_mp_msg *request,
> const void *peer)
> reply.fds[reply.num_fds++] = process_private->txq_fds[queue];
> reply_param->txq_count++;
> }
> -
> + RTE_ASSERT(reply_param->txq_count == dev->data->nb_tx_queues);
> + RTE_ASSERT(reply.num_fds <= RTE_MP_MAX_FD_NUM);
> /* Send reply */
> strlcpy(reply.name, request->name, sizeof(reply.name));
> strlcpy(reply_param->port_name, request_param->port_name,
> --
> 2.17.2
>
Regards,
Keith
