- 15 Mar, 2013 8 commits
-
-
Li RongQing authored
replace ip_fast_csum with csum_replace2 to save cpu cycles Signed-off-by:
Li RongQing <roy.qing.li@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Li RongQing authored
The default value is 128, not 256 #grep gc_thresh1 net/ -rI net/decnet/dn_neigh.c: .gc_thresh1 = 128, net/ipv6/ndisc.c: .gc_thresh1 = 128, net/ipv4/arp.c: .gc_thresh1 = 128, Signed-off-by:
-
Joe Perches authored
I believe these error messages are already logged on allocation failure by warn_alloc_failed and so get a dump_stack on OOM. Remove the unnecessary additional error logging. Around these deletions: o Alignment neatening. o Remove unnecessary casts of dma_alloc_coherent. o Hoist assigns from ifs. Signed-off-by:
Joe Perches <joe@perches.com> Signed-off-by:
-
Somnath Kotur authored
This is needed as the earlier method of manipulating this register via PCI Config space is disallowed by certain Hypervisors. Signed-off-by:
Somnath Kotur <somnath.kotur@emulex.com> Signed-off-by:
-
Somnath Kotur authored
As the NIC PCI function may be used by other protocols, the chip interrupts must be enabled in be_probe() itself rather than be_open(). Signed-off-by:
Sathya Perla <sathya.perla@emulex.com> Signed-off-by:
-
Fernando Luis Vazquez Cao authored
COMPAT_NET_DEV_OPS was removed a while back and with it the definition of netdev_resync_ops() went away. Let's finish the clean-up. Signed-off-by:
Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp> Signed-off-by:
-
Reilly Grant authored
The resource ID used for VM socket control packets (0) is already used for the VMCI_GET_CONTEXT_ID hypercall so a new ID (15) must be used when the guest sends these datagrams to the hypervisor. The hypervisor context ID must also be removed from the internal blacklist. Signed-off-by:
Reilly Grant <grantr@vmware.com> Acked-by:
Andy King <acking@vmware.com> Signed-off-by:
-
nikolay@redhat.com authored
This patch allows LRO aggregation on bonded devices that contain an NX3031 device. It also adds a for_each_netdev_in_bond_rcu(bond, slave) macro which executes for each slave that has bond as master. V3: After testing and discussing this with Rajesh, I decided to keep the vlan ip cache and just rename it to ip_cache since it will store bond ip addresses too. A new master flag has been added to the ip cache to denote that the address has been added because of a master device. I've taken care of the enslave/release cases by checking for various combinations of events and flags (e.g. netxen has a master, it's a bond master and it's not marked as a slave means it is being enslaved and is dev_open()ed in bond_enslave). I've changed netxen_free_ip_list() to have a "master" parameter which causes all IP addresses marked as master to be deleted (used when a netxen is being released). I've made the patch use the new upper device API as well. The following cases were tested: - bond -> netxen - vlan -> netxen - vlan -> bond -> netxen V2: Remove local ip caching, retrieve addresses dynamically and restore them if necessary. Note: Tested with NX3031 adapter. Tested-by:Rajesh Borundia <rajesh.borundia@qlogic.com> Signed-off-by:
Andy Gospodarek <agospoda@redhat.com> Signed-off-by:
Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by:
-
- 14 Mar, 2013 23 commits
-
-
git://git.open-mesh.org/linux-mergeDavid S. Miller authored
Included changes: - introduction of the new Network Coding component. This new mechanism aims to increase throughput by fusing multiple packets in one transmission. - minor cleanups Signed-off-by: -
Arvind Bhushan authored
This patch removes chip specific operations from the common hardware paths, as well as the Makefile change to accomodate the new files. Signed-off-by:
Arvind Bhushan <arvindb@chelsio.com> Signed-off-by:
Naresh Kumar Inna <naresh@chelsio.com> Signed-off-by:
-
Arvind Bhushan authored
This patch defines the common operations to support multiple chips. It includes common header file modifications to support the current chips (T4 and T5). It also includes the following bug fixes: - reconfirms the rnode state after an implicit logo. - corrects the stats array size. - sets up and checks flags correctly when coming up as master and finding the card initialized Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Arvind Bhushan authored
This patch creates a new file for T5 adapter operations. Signed-off-by:
-
Arvind Bhushan authored
This patch separates T4 adapter operations into a new file. Signed-off-by:
-
Vipul Pandya authored
T5 adapter does not support onchip queue memory. Present logic fails to allocate QP for T5 and returns an error. Also, if module parameter ocqp_support is zero then we are unable to allocate QP which should not be the case. Ideally if ocqp_support parameter is 0 or onchip queue support is disable then host QP should be allocated before returning an error. Signed-off-by:
Vipul Pandya <vipul@chelsio.com> Signed-off-by:
-
Vipul Pandya authored
Always bump the tcam_full stat. Also, bump wr reply timeout to 30 seconds. Signed-off-by:
-
Vipul Pandya authored
Signed-off-by:
-
Vipul Pandya authored
It enables direct DMA by HW to memory region PBL arrays and fast register PBL arrays from host memory, vs the T4 way of passing these arrays in the WR itself. The result is lower latency for memory registration, and larger PBL array support for fast register operations. This patch also updates ULP_TX_MEM_WRITE command fields for T5. Ordering bit of ULP_TX_MEM_WRITE is at bit position 22 in T5 and at 23 in T4. Signed-off-by:
-
Vipul Pandya authored
Both DB Flow-Control and DB Coalescing are disabled by default on T5 Signed-off-by:
-
Vipul Pandya authored
Signed-off-by:
-
Vipul Pandya authored
Adds support for Chelsio T5 adapter. Enables T5's Write Combining feature. Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
Santosh Rastapur <santosh@chelsio.com> Signed-off-by:
-
Santosh Rastapur authored
All T5 adapters will only support VFs on PF0-3 despite the ability of the hardware to support them on PF4-7. This keeps our T4 and T5 adapters more similar which simplifies host driver software. Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
This patch implements a low latency Write Combining (aka Write Coalescing) work request path. PCIE maps User Space Doorbell BAR2 region writes to the new interface to SGE. SGE pulls a new message from PCIE new interface and if its a coalesced write work request then pushes it for processing. This patch copies coalesced work request to memory mapped BAR2 space. Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
Santosh Rastapur authored
Signed-off-by:
-
- 13 Mar, 2013 9 commits
-
-
Martin Hundebøll authored
When receiving a network coded packet, the decoding buffer is searched for a packet to use for decoding. The source, destination, and crc32 from the coded packet is used to identify the wanted packet. The decoded packet is passed to the usual unicast receiver function, as had it never been network coded. Signed-off-by:
Martin Hundebøll <martin@hundeboll.net> Signed-off-by:
Marek Lindner <lindner_marek@yahoo.de> Signed-off-by:
Antonio Quartulli <ordex@autistici.org>
-
Martin Hundebøll authored
To be able to decode a network coded packet, a node must already know one of the two coded packets. This is done by buffering skbs before transmission and buffering packets sniffed with promiscuous mode from other hosts. Packets are kept in a buffer similar to the one with forward-skbs: A hash table, where each entry, which corresponds to a src-dst pair, has a linked list packets. Signed-off-by:
-
Martin Hundebøll authored
Before adding forward-skbs to the coding buffer, the buffer is searched for a potential coding opportunity. If one is found, the two packets are network coded and transmitted right away. If not, the forward-skb is added to the buffer. Network coded packets are transmitted with information about the two receivers and the two coded packets. The first receiver is given by the MAC header, while the second is given in the payload/bat-header. The second receiver uses promiscuous mode to receive the packet and check the second destination. Signed-off-by:
-
Martin Hundebøll authored
Two be able to network code two packets, one packet must be buffered until the next is available. This is done in a "coding buffer", which is essentially a hash table with lists of packets. Each entry in the hash table corresponds to a specific src-dst pair, which has a linked list of packets that are buffered. This patch adds skbs to the buffer just before forwarding them. The buffer is traversed every 10 ms, where timed skbs are removed from the buffer and transmitted. To allow experiments with the network coding scheme, the timeout is tunable through a file in debugfs. Signed-off-by:
-
Martin Hundebøll authored
To use network coding efficiently, a relay must know when neighbor nodes are likely to have enough information to be able to decode a network coded packet. This is detected by using OGMs from batman-adv to discover when one neighbor is in range of another neighbor. The relay check the TLL to detect when an OGM is forwarded from one neighbor by another neighbor, and thereby knows that the two neighbors are in range and thus overhear packets sent by each other. This information is saved in the orig_node struct to be used when searching for coding opportunities. Two lists are added to the orig_node struct: One for neighbors that can hear the orig_node (outgoing nc_nodes) and one for neighbors that the orig_node can hear (incoming nc_nodes). Information about nc_nodes is kept for 10 seconds and is available through debugfs in batman_adv/nc_nodes to use when debugging network coding. Signed-off-by:
-
Martin Hundebøll authored
Network coding exploits the 802.11 shared medium to allow multiple packets to be sent in a single transmission. In brief, a relay can XOR two packets, and send the coded packet to two destinations. The receivers can decode one of the original packets by XOR'ing the coded packet with the other original packet. This will lead to increased throughput in topologies where two packets cross one relay. In a simple topology with three nodes, it takes four transmissions without network coding to get one packet from Node A to Node B and one from Node B to Node A: 1. Node A ---- p1 ---> Node R Node B 2. Node A Node R <--- p2 ---- Node B 3. Node A <--- p2 ---- Node R Node B 4. Node A Node R ---- p1 ---> Node B With network coding, the relay only needs one transmission, which saves us one slot of valuable airtime: 1. Node A ---- p1 ---> Node R Node B 2. Node A Node R <--- p2 ---- Node B 3. Node A <- p1 x p2 - Node R - p1 x p2 -> Node B The same principle holds for a topology including five nodes. Here the packets from Node A and Node B are overheard by Node C and Node D, respectively. This allows Node R to send a network coded packet to save one transmission: Node A Node B | \ / | | p1 p2 | | \ / | p1 > Node R < p2 | | | / \ | | p1 x p2 p1 x p2 | v / \ v / \ Node C < > Node D More information is available on the open-mesh.org wiki[1]. This patch adds the initial code to support network coding in batman-adv. It sets up a worker thread to do house keeping and adds a sysfs file to enable/disable network coding. The feature is disabled by default, as it requires a wifi-driver with working promiscuous mode, and also because it adds a small delay at each hop. [1] http://www.open-mesh.org/projects/batman-adv/wiki/CatwomanSigned-off-by: -
Antonio Quartulli authored
In C standard any expression different from 0 will be converted to 'true' when casting to bool (whatever is the length of the value). Therefore all the "!!" conversions can be removed. Signed-off-by:
-
Martin Hundebøll authored
batadv_check_unicast_packet() is changed to return a value based on the reason to drop the packet, which will be useful information for future users of batadv_check_unicast_packet(). Signed-off-by:
-
Marek Lindner authored
The batadv_priv struct carries a pointer to its own interface struct. Therefore, it is not necessary to retrieve the soft_iface via the primary interface. Signed-off-by:
-
