- 18 Feb, 2011 1 commit
-
-
Vasanthy Kolluri authored
We believe that our earlier patch for supporting multiple hardware receive queues per enic device requires more internal testing. At this point, we think that it's best to disable the use of multiple receive queues. The current patch provides an effective means for the same. Also, we continue to disallow multiple hardware transmit queues per device. But change the way we enforce this in order to maintain consistency with the way receive queues are handled. Signed-off-by:
Christian Benvenuti <benve@cisco.com> Signed-off-by:
Danny Guo <dannguo@cisco.com> Signed-off-by:
Vasanthy Kolluri <vkolluri@cisco.com> Signed-off-by:
Roopa Prabhu <roprabhu@cisco.com> Signed-off-by:
David Wang <dwang2@cisco.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
- 17 Feb, 2011 19 commits
-
-
David S. Miller authored
The only troublesome bit here is __mkroute_output which wants to override res->fi and res->type, compute those in local variables instead. Signed-off-by: -
David S. Miller authored
Signed-off-by: -
David S. Miller authored
GCC emits all kinds of crazy zero extensions when we go from signed int, to unsigned short, etc. etc. This transformation has to be legal because: 1) In tkey_extract_bits() in mask_pfx(), the values are used to perform shifts, on which negative values are undefined by C. 2) In fib_table_lookup() we perform comparisons with unsigned values, constants, and additions. None of which should encounter negative values. Signed-off-by: -
David S. Miller authored
This allows avoiding multiple writes to the initial __refcnt. The most simplest cases of wanting an initial reference of "1" in ipv4 and ipv6 have been converted, the rest have been left along and kept at the existing "0". Signed-off-by: -
David S. Miller authored
This also allows us to combine all the dst->flags settings and avoid read/modify/write sequences to this struct member. Signed-off-by: -
David S. Miller authored
Simplifies tail of __ip_route_output_key(). Signed-off-by: -
David S. Miller authored
There's a lot of redundancy and unnecessary stack frames in the output route creation path. 1) Make __mkroute_output() return error pointers. 2) Eliminate ip_mkroute_output() entirely, made possible by #1. 3) Call __mkroute_output() directly and handling the returning error pointers in ip_route_output_slow(). Signed-off-by: -
Michał Mirosław authored
This also enables TSOv6, TSO-ECN, and UFO as loopback clearly can handle them. Signed-off-by:
Michał Mirosław <mirq-linux@rere.qmqm.pl> Signed-off-by:
-
Michał Mirosław authored
Introduce NETIF_F_RXCSUM to replace device-private flags for RX checksum offload. Integrate it with ndo_fix_features. ethtool_op_get_rx_csum() is removed altogether as nothing in-tree uses it. Signed-off-by:
-
Michał Mirosław authored
Signed-off-by:
-
Michał Mirosław authored
Signed-off-by:
-
Michał Mirosław authored
This introduces a new framework to handle device features setting. It consists of: - new fields in struct net_device: + hw_features - features that hw/driver supports toggling + wanted_features - features that user wants enabled, when possible - new netdev_ops: + feat = ndo_fix_features(dev, feat) - API checking constraints for enabling features or their combinations + ndo_set_features(dev) - API updating hardware state to match changed dev->features - new ethtool commands: + ETHTOOL_GFEATURES/ETHTOOL_SFEATURES: get/set dev->wanted_features and trigger device reconfiguration if resulting dev->features changed + ETHTOOL_GSTRINGS(ETH_SS_FEATURES): get feature bits names (meaning) Signed-off-by:
-
Michał Mirosław authored
This allows to enable GRO even if RX csum is disabled. GRO will not be used for packets without hardware checksum anyway. Signed-off-by:
-
Michał Mirosław authored
This is needed for unified offloads patch. Signed-off-by:
-
Michał Mirosław authored
Signed-off-by:
-
Michał Mirosław authored
Signed-off-by:
-
Vasanthy Kolluri authored
Signed-off-by:
-
Vasanthy Kolluri authored
During a device reset, clear the counter for the no. of unicast addresses registered. Also, rename the routines that update unicast and multicast address lists. Signed-off-by:
-
Tom Herbert authored
Support fetching and retrieving RX indirection table via ethtool. Signed-off-by:
Tom Herbert <therbert@google.com> Acked-by:
Eric Dumazet <eric.dumazet@gmail.com> Acked-by:
Eilon Greenstein <eilong@broadcom.com> Signed-off-by:
-
- 15 Feb, 2011 20 commits
-
-
-
-
Ben Hutchings authored
Implement the ndo_setup_tc() operation with 2 traffic classes. Current Solarstorm controllers do not implement TX queue priority, but they do allow queues to be 'paced' with an enforced delay between packets. Paced and unpaced queues are scheduled in round-robin within two separate hardware bins (paced queues with a large delay may be placed into a third bin temporarily, but we won't use that). If there are queues in both bins, the TX scheduler will alternate between them. If we make high-priority queues unpaced and best-effort queues paced, and high-priority queues are mostly empty, a single high-priority queue can then instantly take 50% of the packet rate regardless of how many of the best-effort queues have descriptors outstanding. We do not actually want an enforced delay between packets on best- effort queues, so we set the pace value to a reserved value that actually results in a delay of 0. Signed-off-by:Ben Hutchings <bhutchings@solarflare.com>
-
Ben Hutchings authored
efx_channel_get_{rx,tx}_queue() currently return NULL if the channel isn't used for traffic in that direction. In most cases this is a bug, but some callers rely on it as an existence test. Add existence test functions efx_channel_has_{rx_queue,tx_queues}() and use them as appropriate. Change efx_channel_get_{rx,tx}_queue() to assert that the requested queue exists. Remove now-redundant initialisation from efx_set_channels(). Signed-off-by: -
Ben Hutchings authored
efx_hard_start_xmit() needs to implement a mapping which is the inverse of tx_queue::core_txq. Move the initialisation of tx_queue::core_txq next to efx_hard_start_xmit() to make the connection more obvious. Signed-off-by: -
Ben Hutchings authored
If the root qdisc for a net device is mqprio, and the driver's ndo_setup_tc() operation dynamically adds and remvoes TX queues, netif_set_real_num_tx_queues() will be called during device unregistration to remove the extra TX queues when the qdisc is destroyed. Currently this causes the corresponding kobjects to be leaked, and the device's reference count never drops to 0. Signed-off-by: -
Uwe Kleine-König authored
Also optimize not to reread the value written to FEC_R_CNTRL. Signed-off-by:Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
-
Uwe Kleine-König authored
Reported-by:
Lothar Waßmann <LW@KARO-electronics.de> Signed-off-by:
-
Uwe Kleine-König authored
This fixes warnings when CONFIG_DMA_API_DEBUG=y: NULL NULL: DMA-API: device driver tries to free DMA memory it has not allocated [device address=0x000000004781a020] [size=64 bytes] net eth0: DMA-API: device driver frees DMA memory with different size [device address=0x000000004781a020] [map size=2048 bytes] [unmap size=64 bytes] Moreover pass the platform device to dma_{,un}map_single which makes more sense because the logical network device doesn't know anything about dma. Passing the platform device was a suggestion by Lothar Waßmann. Signed-off-by: -
Uwe Kleine-König authored
Signed-off-by: -
Uwe Kleine-König authored
A few of these were found and reported by Lothar Waßmann. Signed-off-by: -
Uwe Kleine-König authored
A variable named "dev" usually (usually subjective) points to a struct device. Signed-off-by: -
Uwe Kleine-König authored
This undoes the effects of phy_start in fec_enet_open. Reported-by:
-
Uwe Kleine-König authored
Moreover stop listing all i.MX platforms featuring a FEC, and use the platform's config symbol that selects registration of a fec device instead. This might make it easier to add new platforms. Set default = y for ARMs having a fec to reduce defconfig sizes. Signed-off-by: -
Uwe Kleine-König authored
Saving it first into struct net_device->base_addr (which is an unsigned long) is pointless and only needs to use more casts than necessary. Reported-by:
-
Uwe Kleine-König authored
alloc_etherdev internally uses kzalloc, so the private data is already zerod out. Reported-by:
-
Uwe Kleine-König authored
dev_set_drvdata is called unconditionally in the probe function and so it cannot be NULL. Reported-by:
-
Uwe Kleine-König authored
Reported-by:
Lothar Waßmann <LW@KARO-elektronics.de> Signed-off-by:
-
Uwe Kleine-König authored
Reported-by:
-
Uwe Kleine-König authored
memcpy takes a const void * as 2nd argument. So the argument is converted automatically to void * anyhow. Signed-off-by:
-
