[dpdk-dev,v3] bond: inherit maximum rx packet length
Commit Message
Instead of a hard-coded maximum receive length, allow the bond interface
to inherit this limit from the slave interfaces. This allows
an application that uses jumbo frames to pass realistic values to
rte_eth_dev_configure without causing an error.
Before the bond interface is configured, allow slaves with any
max_rx_pktlen to be added and remember the lowest of these values as
a candidate value. During dev_configure, set the bond device's
max_rx_pktlen to the candidate value. After this point only slaves
with a max_rx_pktlen greater or equal to that of the bonding device
can be added.
If all slaves are removed, the bond device's pktlen is cleared.
Signed-off-by: Eric Kinzie <ehkinzie@gmail.com>
---
drivers/net/bonding/rte_eth_bond_api.c | 18 +++++++++++++++++-
drivers/net/bonding/rte_eth_bond_pmd.c | 6 +++++-
drivers/net/bonding/rte_eth_bond_private.h | 3 +++
3 files changed, 25 insertions(+), 2 deletions(-)
Comments
On 07/05/16 04:45, Eric Kinzie wrote:
> Instead of a hard-coded maximum receive length, allow the bond interface
> to inherit this limit from the slave interfaces. This allows
> an application that uses jumbo frames to pass realistic values to
> rte_eth_dev_configure without causing an error.
>
> Before the bond interface is configured, allow slaves with any
> max_rx_pktlen to be added and remember the lowest of these values as
> a candidate value. During dev_configure, set the bond device's
> max_rx_pktlen to the candidate value. After this point only slaves
> with a max_rx_pktlen greater or equal to that of the bonding device
> can be added.
>
> If all slaves are removed, the bond device's pktlen is cleared.
>
> Signed-off-by: Eric Kinzie <ehkinzie@gmail.com>
> ---
>
....
>
Acked-by: Declan Dohetry <declan.doherty@intel.com>
On Fri, Jun 17, 2016 at 03:30:03PM +0100, Declan Doherty wrote:
> On 07/05/16 04:45, Eric Kinzie wrote:
> > Instead of a hard-coded maximum receive length, allow the bond interface
> > to inherit this limit from the slave interfaces. This allows
> > an application that uses jumbo frames to pass realistic values to
> > rte_eth_dev_configure without causing an error.
> >
> > Before the bond interface is configured, allow slaves with any
> > max_rx_pktlen to be added and remember the lowest of these values as
> > a candidate value. During dev_configure, set the bond device's
> > max_rx_pktlen to the candidate value. After this point only slaves
> > with a max_rx_pktlen greater or equal to that of the bonding device
> > can be added.
> >
> > If all slaves are removed, the bond device's pktlen is cleared.
> >
> >Signed-off-by: Eric Kinzie <ehkinzie@gmail.com>
> >---
> >
> ....
> >
>
> Acked-by: Declan Dohetry <declan.doherty@intel.com>
Applied to dpdk-next-net/rel_16_07, with one checkpatch fix and with title
corrected to pass check with check-git-log.sh
/Bruce
@@ -247,6 +247,8 @@ rte_eth_bond_create(const char *name, uint8_t mode, uint8_t socket_id)
internals->active_slave_count = 0;
internals->rx_offload_capa = 0;
internals->tx_offload_capa = 0;
+ internals->candidate_max_rx_pktlen = 0;
+ internals->max_rx_pktlen = 0;
/* Initially allow to choose any offload type */
internals->flow_type_rss_offloads = ETH_RSS_PROTO_MASK;
@@ -331,9 +333,15 @@ __eth_bond_slave_add_lock_free(uint8_t bonded_port_id, uint8_t slave_port_id)
/* Add slave details to bonded device */
slave_eth_dev->data->dev_flags |= RTE_ETH_DEV_BONDED_SLAVE;
- slave_add(internals, slave_eth_dev);
rte_eth_dev_info_get(slave_port_id, &dev_info);
+ if (dev_info.max_rx_pktlen < internals->max_rx_pktlen) {
+ RTE_BOND_LOG(ERR, "Slave (port %u) max_rx_pktlen too small",
+ slave_port_id);
+ return -1;
+ }
+
+ slave_add(internals, slave_eth_dev);
/* We need to store slaves reta_size to be able to synchronize RETA for all
* slave devices even if its sizes are different.
@@ -365,6 +373,9 @@ __eth_bond_slave_add_lock_free(uint8_t bonded_port_id, uint8_t slave_port_id)
internals->tx_offload_capa = dev_info.tx_offload_capa;
internals->flow_type_rss_offloads = dev_info.flow_type_rss_offloads;
+ /* Inherit first slave's max rx packet size */
+ internals->candidate_max_rx_pktlen = dev_info.max_rx_pktlen;
+
} else {
/* Check slave link properties are supported if props are set,
* all slaves must be the same */
@@ -391,6 +402,9 @@ __eth_bond_slave_add_lock_free(uint8_t bonded_port_id, uint8_t slave_port_id)
if (internals->reta_size > dev_info.reta_size)
internals->reta_size = dev_info.reta_size;
+ if (!internals->max_rx_pktlen
+ && dev_info.max_rx_pktlen < internals->candidate_max_rx_pktlen)
+ internals->candidate_max_rx_pktlen = dev_info.max_rx_pktlen;
}
bonded_eth_dev->data->dev_conf.rx_adv_conf.rss_conf.rss_hf &=
@@ -536,6 +550,8 @@ __eth_bond_slave_remove_lock_free(uint8_t bonded_port_id, uint8_t slave_port_id)
internals->tx_offload_capa = 0;
internals->flow_type_rss_offloads = ETH_RSS_PROTO_MASK;
internals->reta_size = 0;
+ internals->candidate_max_rx_pktlen = 0;
+ internals->max_rx_pktlen = 0;
}
return 0;
}
@@ -1650,7 +1650,8 @@ bond_ethdev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
dev_info->max_mac_addrs = 1;
- dev_info->max_rx_pktlen = (uint32_t)2048;
+ dev_info->max_rx_pktlen = internals->candidate_max_rx_pktlen ?
+ internals->candidate_max_rx_pktlen : 2048;
dev_info->max_rx_queues = (uint16_t)128;
dev_info->max_tx_queues = (uint16_t)512;
@@ -2294,6 +2295,9 @@ bond_ethdev_configure(struct rte_eth_dev *dev)
}
}
+ /* set the max_rx_pktlen */
+ internals->max_rx_pktlen = internals->candidate_max_rx_pktlen;
+
/*
* if no kvlist, it means that this bonded device has been created
* through the bonding api.
@@ -169,6 +169,9 @@ struct bond_dev_private {
struct rte_kvargs *kvlist;
uint8_t slave_update_idx;
+
+ uint32_t candidate_max_rx_pktlen;
+ uint32_t max_rx_pktlen;
};
extern const struct eth_dev_ops default_dev_ops;