- 30 Sep, 2014 21 commits
-
-
Antoine Ténart authored
Add the device tree support to the pxa168_eth driver. Signed-off-by:
Antoine Tenart <antoine.tenart@free-electrons.com> Acked-by:
Arnd Bergmann <arnd@arndb.de> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Antoine Ténart authored
Clean up a bit the pxa168_eth driver before adding the device tree support. Signed-off-by:
-
David S. Miller authored
Or Gerlitz says: ==================== mlx4_core driver updates A series from Jack and Co of low-level fixes for the mlx4_core driver ==================== Signed-off-by: -
Jack Morgenstein authored
ConnectX2 HCAs have max_mtu=4k and max_vl=8 vls. However, if you specify a 4K mtu, the max_vl supported for 4K is 4 vls. The driver at startup attempts to set a 4K mtu using the max_vl value obtained from QUERY_PORT. Since the max_vl value is 8 vls (which is supported up to 2K mtu size), the first attempt to set the mtl/vl port value will fail, generating the following error message in the log: mlx4_core 0000:06:00.0: command 0xc failed: fw status = 0x40 The driver then tries again, using mtu=4k, vls=4, and this succeeds. Since we do not want to have this error message always displayed at driver start when there are ConnectX2 HCAs on the host, we deprecate the error message for this specific command/input_modifier/opcode_modifier/fw-status to be debug. Signed-off-by:
Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by:
Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by:
-
Jack Morgenstein authored
The function mlx4_QUERY_PORT_wrapper implements only the QUERY_PORT "general" case (opcode modifier = 0). Verify that the opcode modifier is zero, and also that the input modifier contains only the port number in bits 0..7 (all other bits should be zero). Signed-off-by:
-
Majd Dibbiny authored
In the new flow, we separate the pci initialization and teardown from the initialization and teardown of the other resources. __mlx4_init_one handles the pci resources initialization. It then calls mlx4_load_one to initialize the remainder of the resources. When removing a device, mlx4_remove_one is invoked. However, now mlx4_remove_one calls mlx4_unload_one to free all the resources except the pci resources. When mlx4_unload_one returns, mlx4_remove_one then frees the pci resources. The above separation will allow us to implement 'reset flow' in the future. It will also enable more EQs for VFs and is a pre-step to the modern API to enable/disable SRIOV. Also added nvfs; an integer array of size MLX4_MAX_PORTS + 1; to the mlx4_dev struct. This new field is used to avoid parsing the num_vfs module parameter each time the mlx4_restart_one is called. Signed-off-by:
Majd Dibbiny <majd@mellanox.com> Signed-off-by:
-
Jack Morgenstein authored
When unloading the host driver while there are VFs active on VMs, the PF driver disabled sriov anyway, causing kernel crashes. We now leave SRIOV enabled, to avoid that. When the driver is reloaded, __mlx4_init_one is invoked on the PF. It now checks to see if SRIOV is already enabled on the PF -- and if so does not enable sriov again. Signed-off-by:
Tal Alon <talal@mellanox.com> Signed-off-by:
-
Florian Westphal authored
Eric reports build failure with CONFIG_BRIDGE_NETFILTER=n We insist to build br_nf_core.o unconditionally, but we must only do so if br_netfilter was enabled, else it fails to build due to functions being defined to empty stubs (and some structure members being defined out). Also, BRIDGE_NETFILTER=y|m makes no sense when BRIDGE=n. Fixes: 34666d46 (netfilter: bridge: move br_netfilter out of the core) Reported-by:
Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by:
Florian Westphal <fw@strlen.de> Signed-off-by:
Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by:
-
David S. Miller authored
Markus Pargmann says: ==================== net: cpsw: Support for am335x chip MACIDs This series adds support to the cpsw driver to read the MACIDs of the am335x chip and use them as fallback. These addresses are only used if there are no mac addresses in the devicetree, for example set by a bootloader. ==================== Acked-by:
Mugunthan V N <mugunthanvnm@gmail.com> Signed-off-by:
-
Markus Pargmann authored
There are 2 MACIDs stored in the control module of the am33xx. These are read by the cpsw driver if no valid MACID was found in the devicetree. Signed-off-by:
Markus Pargmann <mpa@pengutronix.de> Reviewed-by:
Wolfram Sang <wsa@the-dreams.de> Acked-by:
Tony Lindgren <tony@atomide.com> Signed-off-by:
-
Markus Pargmann authored
Signed-off-by:
-
Markus Pargmann authored
This patch adds a function to get the MACIDs from the am33xx SoC control module registers which hold unique vendor MACIDs. This is only used if of_get_mac_address() fails to get a valid mac address. Signed-off-by:
Steven Rostedt <rostedt@goodmis.org> Acked-by:
-
Markus Pargmann authored
Use dev_err instead of pr_err. Signed-off-by:
-
Markus Pargmann authored
"MII_BUS_ID_SIZE" is defined in linux/phy.h which is not included in the cpsw.h file. Signed-off-by:
-
Markus Pargmann authored
ret is set 0 at this point, so jumping to that error label would result in a return value of 0. Set ret to -ENOMEM to return a proper error value. Signed-off-by:
-
Markus Pargmann authored
mac-address is an optional property. If no mac-address is set, a random mac-address will be generated. Signed-off-by:
-
Andy Gospodarek authored
As the code stands today, bonding stats are based simply on the stats from the member interfaces. If a member was to be removed from a bond, the stats would instantly drop. This would be confusing to an admin would would suddonly see interface stats drop while traffic is still flowing. In addition to preventing the stats drops mentioned above, new members will now be added to the bond and only traffic received after the member was added to the bond will be counted as part of bonding stats. Bonding counters will also be updated when any slaves are dropped to make sure the reported stats are reliable. v2: Changes suggested by Nik to properly allocate/free stats memory. v3: Properly destroy workqueue and fix netlink configuration path. v4: Moved cached stats into bonding and slave structs as there does not seem to be a complexity/performance benefit to using alloc'd memory vs in-struct memory. Signed-off-by:
Andy Gospodarek <gospo@cumulusnetworks.com> Signed-off-by:
-
John Fastabend authored
After previous patches to simplify qstats the qstats can be made per cpu with a packed union in Qdisc struct. Signed-off-by:
John Fastabend <john.r.fastabend@intel.com> Signed-off-by:
-
John Fastabend authored
This removes the use of qstats->qlen variable from the classifiers and makes it an explicit argument to gnet_stats_copy_queue(). The qlen represents the qdisc queue length and is packed into the qstats at the last moment before passnig to user space. By handling it explicitely we avoid, in the percpu stats case, having to figure out which per_cpu variable to put it in. It would probably be best to remove it from qstats completely but qstats is a user space ABI and can't be broken. A future patch could make an internal only qstats structure that would avoid having to allocate an additional u32 variable on the Qdisc struct. This would make the qstats struct 128bits instead of 128+32. Signed-off-by:
-
John Fastabend authored
This adds helpers to manipulate qstats logic and replaces locations that touch the counters directly. This simplifies future patches to push qstats onto per cpu counters. Signed-off-by:
-
John Fastabend authored
In order to run qdisc's without locking statistics and estimators need to be handled correctly. To resolve bstats make the statistics per cpu. And because this is only needed for qdiscs that are running without locks which is not the case for most qdiscs in the near future only create percpu stats when qdiscs set the TCQ_F_CPUSTATS flag. Next because estimators use the bstats to calculate packets per second and bytes per second the estimator code paths are updated to use the per cpu statistics. Signed-off-by:
-
- 29 Sep, 2014 19 commits
-
-
Michael Braun authored
This patch adds a new mode of operation to macvlan, called "source". It allows one to set a list of allowed mac address, which is used to match against source mac address from received frames on underlying interface. This enables creating mac based VLAN associations, instead of standard port or tag based. The feature is useful to deploy 802.1x mac based behavior, where drivers of underlying interfaces doesn't allows that. Configuration is done through the netlink interface using e.g.: ip link add link eth0 name macvlan0 type macvlan mode source ip link add link eth0 name macvlan1 type macvlan mode source ip link set link dev macvlan0 type macvlan macaddr add 00:11:11:11:11:11 ip link set link dev macvlan0 type macvlan macaddr add 00:22:22:22:22:22 ip link set link dev macvlan0 type macvlan macaddr add 00:33:33:33:33:33 ip link set link dev macvlan1 type macvlan macaddr add 00:33:33:33:33:33 ip link set link dev macvlan1 type macvlan macaddr add 00:44:44:44:44:44 This allows clients with MAC addresses 00:11:11:11:11:11, 00:22:22:22:22:22 to be part of only VLAN associated with macvlan0 interface. Clients with MAC addresses 00:44:44:44:44:44 with only VLAN associated with macvlan1 interface. And client with MAC address 00:33:33:33:33:33 to be associated with both VLANs. Based on work of Stefan Gula <steweg@gmail.com> v8: last version of Stefan Gula for Kernel 3.2.1 v9: rework onto linux-next 2014-03-12 by Michael Braun add MACADDR_SET command, enable to configure mac for source mode while creating interface v10: - reduce indention level - rename source_list to source_entry - use aligned 64bit ether address - use hash_64 instead of addr[5] v11: - rebase for 3.14 / linux-next 20.04.2014 v12 - rebase for linux-next 2014-09-25 Signed-off-by:Michael Braun <michael-dev@fami-braun.de> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-nextDavid S. Miller authored
Pablo Neira Ayuso says: ==================== pull request: netfilter/ipvs updates for net-next The following patchset contains Netfilter/IPVS updates for net-next, most relevantly they are: 1) Four patches to make the new nf_tables masquerading support independent of the x_tables infrastructure. This also resolves a compilation breakage if the masquerade target is disabled but the nf_tables masq expression is enabled. 2) ipset updates via Jozsef Kadlecsik. This includes the addition of the skbinfo extension that allows you to store packet metainformation in the elements. This can be used to fetch and restore this to the packets through the iptables SET target, patches from Anton Danilov. 3) Add the hash:mac set type to ipset, from Jozsef Kadlecsick. 4) Add simple weighted fail-over scheduler via Simon Horman. This provides a fail-over IPVS scheduler (unlike existing load balancing schedulers). Connections are directed to the appropriate server based solely on highest weight value and server availability, patch from Kenny Mathis. 5) Support IPv6 real servers in IPv4 virtual-services and vice versa. Simon Horman informs that the motivation for this is to allow more flexibility in the choice of IP version offered by both virtual-servers and real-servers as they no longer need to match: An IPv4 connection from an end-user may be forwarded to a real-server using IPv6 and vice versa. No ip_vs_sync support yet though. Patches from Alex Gartrell and Julian Anastasov. 6) Add global generation ID to the nf_tables ruleset. When dumping from several different object lists, we need a way to identify that an update has ocurred so userspace knows that it needs to refresh its lists. This also includes a new command to obtain the 32-bits generation ID. The less significant 16-bits of this ID is also exposed through res_id field in the nfnetlink header to quickly detect the interference and retry when there is no risk of ID wraparound. 7) Move br_netfilter out of the bridge core. The br_netfilter code is built in the bridge core by default. This causes problems of different kind to people that don't want this: Jesper reported performance drop due to the inconditional hook registration and I remember to have read complains on netdev from people regarding the unexpected behaviour of our bridging stack when br_netfilter is enabled (fragmentation handling, layer 3 and upper inspection). People that still need this should easily undo the damage by modprobing the new br_netfilter module. 8) Dump the set policy nf_tables that allows set parameterization. So userspace can keep user-defined preferences when saving the ruleset. From Arturo Borrero. 9) Use __seq_open_private() helper function to reduce boiler plate code in x_tables, From Rob Jones. 10) Safer default behaviour in case that you forget to load the protocol tracker. Daniel Borkmann and Florian Westphal detected that if your ruleset is stateful, you allow traffic to at least one single SCTP port and the SCTP protocol tracker is not loaded, then any SCTP traffic may be pass through unfiltered. After this patch, the connection tracking classifies SCTP/DCCP/UDPlite/GRE packets as invalid if your kernel has been compiled with support for these modules. ==================== Trivially resolved conflict in include/linux/skbuff.h, Eric moved some netfilter skbuff members around, and the netfilter tree adjusted the ifdef guards for the bridging info pointer. Signed-off-by: -
Florian Westphal authored
Suggested by Stephen. Also drop inline keyword and let compiler decide. gcc 4.7.3 decides to no longer inline tcp_ecn_check_ce, so split it up. The actual evaluation is not inlined anymore while the ECN_OK test is. Suggested-by:
Stephen Hemminger <stephen@networkplumber.org> Signed-off-by:
Daniel Borkmann <dborkman@redhat.com> Signed-off-by:
-
Florian Westphal authored
After Octavian Purdilas tcp ipv4/ipv6 unification work this helper only has a single callsite. While at it, convert name to lowercase, suggested by Stephen. Suggested-by:
-
David S. Miller authored
Michael Grzeschik says: ==================== ARCNET: add support for EAE multi interfac card this series adds support for the PLX Bridge based multi interface pci cards and adds support to change device address on com200xx chips during runtime. This series is based on v3.17-rc7. It is fixed for build against com20020_cs. ==================== Signed-off-by: -
Michael Grzeschik authored
This patch adds support for the EAE arcnet cards which has two Interfaces. Signed-off-by:
Michael Grzeschik <m.grzeschik@pengutronix.de> Signed-off-by:
-
Michael Grzeschik authored
The com20020-pci driver is currently designed to instance one netdev with one pci device. This patch adds support to instance many cards with one pci device, depending on the device data in the private data. Signed-off-by:
-
Michael Grzeschik authored
This patch adds metadata for the com20020 to prepare for devices with multiple io address areas with multi card interfaces. Signed-off-by:
-
Michael Grzeschik authored
This patch adds com20020_set_hwaddr to make it possible to change the hwaddr on runtime. Signed-off-by:
-
Michael Grzeschik authored
The interrupt handler needs to return IRQ_NONE in case two devices are used with the shared interrupt handler. Otherwise it could steal interrupts from the other interface. Signed-off-by:
-
Li RongQing authored
This variable i is overwritten to 0 by following code Signed-off-by:
Li RongQing <roy.qing.li@gmail.com> Signed-off-by:
-
Eric Dumazet authored
With proliferation of bit fields in sk_buff, __copy_skb_header() became quite expensive, showing as the most expensive function in a GSO workload. __copy_skb_header() performance is also critical for non GSO TCP operations, as it is used from skb_clone() This patch carefully moves all the fields that were not copied in a separate zone : cloned, nohdr, fclone, peeked, head_frag, xmit_more Then I moved all other fields and all other copied fields in a section delimited by headers_start[0]/headers_end[0] section so that we can use a single memcpy() call, inlined by compiler using long word load/stores. I also tried to make all copies in the natural orders of sk_buff, to help hardware prefetching. I made sure sk_buff size did not change. Signed-off-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
-
Florian Westphal authored
Given following iptables ruleset: -P FORWARD DROP -A FORWARD -m sctp --dport 9 -j ACCEPT -A FORWARD -p tcp --dport 80 -j ACCEPT -A FORWARD -p tcp -m conntrack -m state ESTABLISHED,RELATED -j ACCEPT One would assume that this allows SCTP on port 9 and TCP on port 80. Unfortunately, if the SCTP conntrack module is not loaded, this allows *all* SCTP communication, to pass though, i.e. -p sctp -j ACCEPT, which we think is a security issue. This is because on the first SCTP packet on port 9, we create a dummy "generic l4" conntrack entry without any port information (since conntrack doesn't know how to extract this information). All subsequent packets that are unknown will then be in established state since they will fallback to proto_generic and will match the 'generic' entry. Our originally proposed version [1] completely disabled generic protocol tracking, but Jozsef suggests to not track protocols for which a more suitable helper is available, hence we now mitigate the issue for in tree known ct protocol helpers only, so that at least NAT and direction information will still be preserved for others. [1] http://www.spinics.net/lists/netfilter-devel/msg33430.html Joint work with Daniel Borkmann. Signed-off-by:
Jozsef Kadlecsik <kadlec@blackhole.kfki.hu> Signed-off-by:
-
Arturo Borrero authored
We want to know in which cases the user explicitly sets the policy options. In that case, we also want to dump back the info. Signed-off-by:
Arturo Borrero Gonzalez <arturo.borrero.glez@gmail.com> Signed-off-by:
-
David S. Miller authored
Stefan Wahren says: ==================== add Qualcomm QCA7000 ethernet driver This patch series adds support for the Qualcomm QCA7000 Homeplug GreenPHY. The QCA7000 is serial-to-powerline bridge with two interfaces: UART and SPI. These patches handles only the last one, with an Ethernet over SPI protocol driver. This driver based on the Qualcomm code [1], but contains a lot of changes since last year: * devicetree support * DebugFS support * ethtool support * better error handling * performance improvements * code cleanup * some bugfixes The code has been tested only on Freescale i.MX28 boards, but should work on other platforms. [1] - https://github.com/IoE/qca7000 Changes in V3: - Use ether_addr_copy instead of memcpy - Remove qcaspi_set_mac_address - Improve DT parsing - replace OF_GPIO dependancy with OF - fix compile error caused by SET_ETHTOOL_OPS - fix possible endless loop when spi read fails - fix DT documentation - fix coding style - fix sparse warnings Changes in V2: - replace in DT the SPI intr GPIO with pure interrupt - make legacy mode a boolean DT property and remove it as module parameter - make burst length a module parameter instead of DT property - make pluggable a module parameter instead of DT property - improve DT documentation - replace debugFS register dump with ethtool function - replace debugFS stats with ethtool function - implement function to get ring parameter via ethtool - implement function to set TX ring count via ethtool - fix TX ring state in debugFS - optimize tx ring flush - add byte limit for TX ring to avoid bufferbloat - fix TX queue full and write buffer miss counter - fix SPI clk speed module parameter - fix possible packet loss - fix possible race during transmit ==================== Signed-off-by:
-
Stefan Wahren authored
This patch adds the Ethernet over SPI driver for the Qualcomm QCA7000 HomePlug GreenPHY. Signed-off-by:
Stefan Wahren <stefan.wahren@i2se.com> Signed-off-by:
-
Stefan Wahren authored
This patch adds the Device tree bindings for the Ethernet over SPI protocol driver of the Qualcomm QCA7000 HomePlug GreenPHY. Signed-off-by:
-
David S. Miller authored
Daniel Borkmann says: ==================== net: tcp: DCTCP congestion control algorithm This patch series adds support for the DataCenter TCP (DCTCP) congestion control algorithm. Please see individual patches for the details. The last patch adds DCTCP as a congestion control module, and previous ones add needed infrastructure to extend the congestion control framework. Joint work between Florian Westphal, Daniel Borkmann and Glenn Judd. v3 -> v2: - No changes anywhere, just a resend as requested by Dave - Added Stephen's ACK v1 -> v2: - Rebased to latest net-next - Addressed Eric's feedback, thanks! - Update stale comment wrt. DCTCP ECN usage - Don't call INET_ECN_xmit for every packet - Add dctcp ss/inetdiag support to expose internal stats to userspace ==================== Signed-off-by: -
Daniel Borkmann authored
This work adds the DataCenter TCP (DCTCP) congestion control algorithm [1], which has been first published at SIGCOMM 2010 [2], resp. follow-up analysis at SIGMETRICS 2011 [3] (and also, more recently as an informational IETF draft available at [4]). DCTCP is an enhancement to the TCP congestion control algorithm for data center networks. Typical data center workloads are i.e. i) partition/aggregate (queries; bursty, delay sensitive), ii) short messages e.g. 50KB-1MB (for coordination and control state; delay sensitive), and iii) large flows e.g. 1MB-100MB (data update; throughput sensitive). DCTCP has therefore been designed for such environments to provide/achieve the following three requirements: * High burst tolerance (incast due to partition/aggregate) * Low latency (short flows, queries) * High throughput (continuous data updates, large file transfers) with commodity, shallow buffered switches The basic idea of its design consists of two fundamentals: i) on the switch side, packets are being marked when its internal queue length > threshold K (K is chosen so that a large enough headroom for marked traffic is still available in the switch queue); ii) the sender/host side maintains a moving average of the fraction of marked packets, so each RTT, F is being updated as follows: F := X / Y, where X is # of marked ACKs, Y is total # of ACKs alpha := (1 - g) * alpha + g * F, where g is a smoothing constant The resulting alpha (iow: probability that switch queue is congested) is then being used in order to adaptively decrease the congestion window W: W := (1 - (alpha / 2)) * W The means for receiving marked packets resp. marking them on switch side in DCTCP is the use of ECN. RFC3168 describes a mechanism for using Explicit Congestion Notification from the switch for early detection of congestion, rather than waiting for segment loss to occur. However, this method only detects the presence of congestion, not the *extent*. In the presence of mild congestion, it reduces the TCP congestion window too aggressively and unnecessarily affects the throughput of long flows [4]. DCTCP, as mentioned, enhances Explicit Congestion Notification (ECN) processing to estimate the fraction of bytes that encounter congestion, rather than simply detecting that some congestion has occurred. DCTCP then scales the TCP congestion window based on this estimate [4], thus it can derive multibit feedback from the information present in the single-bit sequence of marks in its control law. And thus act in *proportion* to the extent of congestion, not its *presence*. Switches therefore set the Congestion Experienced (CE) codepoint in packets when internal queue lengths exceed threshold K. Resulting, DCTCP delivers the same or better throughput than normal TCP, while using 90% less buffer space. It was found in [2] that DCTCP enables the applications to handle 10x the current background traffic, without impacting foreground traffic. Moreover, a 10x increase in foreground traffic did not cause any timeouts, and thus largely eliminates TCP incast collapse problems. The algorithm itself has already seen deployments in large production data centers since then. We did a long-term stress-test and analysis in a data center, short summary of our TCP incast tests with iperf compared to cubic: This test measured DCTCP throughput and latency and compared it with CUBIC throughput and latency for an incast scenario. In this test, 19 senders sent at maximum rate to a single receiver. The receiver simply ran iperf -s. The senders ran iperf -c <receiver> -t 30. All senders started simultaneously (using local clocks synchronized by ntp). This test was repeated multiple times. Below shows the results from a single test. Other tests are similar. (DCTCP results were extremely consistent, CUBIC results show some variance induced by the TCP timeouts that CUBIC encountered.) For this test, we report statistics on the number of TCP timeouts, flow throughput, and traffic latency. 1) Timeouts (total over all flows, and per flow summaries): CUBIC DCTCP Total 3227 25 Mean 169.842 1.316 Median 183 1 Max 207 5 Min 123 0 Stddev 28.991 1.600 Timeout data is taken by measuring the net change in netstat -s "other TCP timeouts" reported. As a result, the timeout measurements above are not restricted to the test traffic, and we believe that it is likely that all of the "DCTCP timeouts" are actually timeouts for non-test traffic. We report them nevertheless. CUBIC will also include some non-test timeouts, but they are drawfed by bona fide test traffic timeouts for CUBIC. Clearly DCTCP does an excellent job of preventing TCP timeouts. DCTCP reduces timeouts by at least two orders of magnitude and may well have eliminated them in this scenario. 2) Throughput (per flow in Mbps): CUBIC DCTCP Mean 521.684 521.895 Median 464 523 Max 776 527 Min 403 519 Stddev 105.891 2.601 Fairness 0.962 0.999 Throughput data was simply the average throughput for each flow reported by iperf. By avoiding TCP timeouts, DCTCP is able to achieve much better per-flow results. In CUBIC, many flows experience TCP timeouts which makes flow throughput unpredictable and unfair. DCTCP, on the other hand, provides very clean predictable throughput without incurring TCP timeouts. Thus, the standard deviation of CUBIC throughput is dramatically higher than the standard deviation of DCTCP throughput. Mean throughput is nearly identical because even though cubic flows suffer TCP timeouts, other flows will step in and fill the unused bandwidth. Note that this test is something of a best case scenario for incast under CUBIC: it allows other flows to fill in for flows experiencing a timeout. Under situations where the receiver is issuing requests and then waiting for all flows to complete, flows cannot fill in for timed out flows and throughput will drop dramatically. 3) Latency (in ms): CUBIC DCTCP Mean 4.0088 0.04219 Median 4.055 0.0395 Max 4.2 0.085 Min 3.32 0.028 Stddev 0.1666 0.01064 Latency for each protocol was computed by running "ping -i 0.2 <receiver>" from a single sender to the receiver during the incast test. For DCTCP, "ping -Q 0x6 -i 0.2 <receiver>" was used to ensure that traffic traversed the DCTCP queue and was not dropped when the queue size was greater than the marking threshold. The summary statistics above are over all ping metrics measured between the single sender, receiver pair. The latency results for this test show a dramatic difference between CUBIC and DCTCP. CUBIC intentionally overflows the switch buffer which incurs the maximum queue latency (more buffer memory will lead to high latency.) DCTCP, on the other hand, deliberately attempts to keep queue occupancy low. The result is a two orders of magnitude reduction of latency with DCTCP - even with a switch with relatively little RAM. Switches with larger amounts of RAM will incur increasing amounts of latency for CUBIC, but not for DCTCP. 4) Convergence and stability test: This test measured the time that DCTCP took to fairly redistribute bandwidth when a new flow commences. It also measured DCTCP's ability to remain stable at a fair bandwidth distribution. DCTCP is compared with CUBIC for this test. At the commencement of this test, a single flow is sending at maximum rate (near 10 Gbps) to a single receiver. One second after that first flow commences, a new flow from a distinct server begins sending to the same receiver as the first flow. After the second flow has sent data for 10 seconds, the second flow is terminated. The first flow sends for an additional second. Ideally, the bandwidth would be evenly shared as soon as the second flow starts, and recover as soon as it stops. The results of this test are shown below. Note that the flow bandwidth for the two flows was measured near the same time, but not simultaneously. DCTCP performs nearly perfectly within the measurement limitations of this test: bandwidth is quickly distributed fairly between the two flows, remains stable throughout the duration of the test, and recovers quickly. CUBIC, in contrast, is slow to divide the bandwidth fairly, and has trouble remaining stable. CUBIC DCTCP Seconds Flow 1 Flow 2 Seconds Flow 1 Flow 2 0 9.93 0 0 9.92 0 0.5 9.87 0 0.5 9.86 0 1 8.73 2.25 1 6.46 4.88 1.5 7.29 2.8 1.5 4.9 4.99 2 6.96 3.1 2 4.92 4.94 2.5 6.67 3.34 2.5 4.93 5 3 6.39 3.57 3 4.92 4.99 3.5 6.24 3.75 3.5 4.94 4.74 4 6 3.94 4 5.34 4.71 4.5 5.88 4.09 4.5 4.99 4.97 5 5.27 4.98 5 4.83 5.01 5.5 4.93 5.04 5.5 4.89 4.99 6 4.9 4.99 6 4.92 5.04 6.5 4.93 5.1 6.5 4.91 4.97 7 4.28 5.8 7 4.97 4.97 7.5 4.62 4.91 7.5 4.99 4.82 8 5.05 4.45 8 5.16 4.76 8.5 5.93 4.09 8.5 4.94 4.98 9 5.73 4.2 9 4.92 5.02 9.5 5.62 4.32 9.5 4.87 5.03 10 6.12 3.2 10 4.91 5.01 10.5 6.91 3.11 10.5 4.87 5.04 11 8.48 0 11 8.49 4.94 11.5 9.87 0 11.5 9.9 0 SYN/ACK ECT test: This test demonstrates the importance of ECT on SYN and SYN-ACK packets by measuring the connection probability in the presence of competing flows for a DCTCP connection attempt *without* ECT in the SYN packet. The test was repeated five times for each number of competing flows. Competing Flows 1 | 2 | 4 | 8 | 16 ------------------------------ Mean Connection Probability 1 | 0.67 | 0.45 | 0.28 | 0 Median Connection Probability 1 | 0.65 | 0.45 | 0.25 | 0 As the number of competing flows moves beyond 1, the connection probability drops rapidly. Enabling DCTCP with this patch requires the following steps: DCTCP must be running both on the sender and receiver side in your data center, i.e.: sysctl -w net.ipv4.tcp_congestion_control=dctcp Also, ECN functionality must be enabled on all switches in your data center for DCTCP to work. The default ECN marking threshold (K) heuristic on the switch for DCTCP is e.g., 20 packets (30KB) at 1Gbps, and 65 packets (~100KB) at 10Gbps (K > 1/7 * C * RTT, [4]). In above tests, for each switch port, traffic was segregated into two queues. For any packet with a DSCP of 0x01 - or equivalently a TOS of 0x04 - the packet was placed into the DCTCP queue. All other packets were placed into the default drop-tail queue. For the DCTCP queue, RED/ECN marking was enabled, here, with a marking threshold of 75 KB. More details however, we refer you to the paper [2] under section 3). There are no code changes required to applications running in user space. DCTCP has been implemented in full *isolation* of the rest of the TCP code as its own congestion control module, so that it can run without a need to expose code to the core of the TCP stack, and thus nothing changes for non-DCTCP users. Changes in the CA framework code are minimal, and DCTCP algorithm operates on mechanisms that are already available in most Silicon. The gain (dctcp_shift_g) is currently a fixed constant (1/16) from the paper, but we leave the option that it can be chosen carefully to a different value by the user. In case DCTCP is being used and ECN support on peer site is off, DCTCP falls back after 3WHS to operate in normal TCP Reno mode. ss {-4,-6} -t -i diag interface: ... dctcp wscale:7,7 rto:203 rtt:2.349/0.026 mss:1448 cwnd:2054 ssthresh:1102 ce_state 0 alpha 15 ab_ecn 0 ab_tot 735584 send 10129.2Mbps pacing_rate 20254.1Mbps unacked:1822 retrans:0/15 reordering:101 rcv_space:29200 ... dctcp-reno wscale:7,7 rto:201 rtt:0.711/1.327 ato:40 mss:1448 cwnd:10 ssthresh:1102 fallback_mode send 162.9Mbps pacing_rate 325.5Mbps rcv_rtt:1.5 rcv_space:29200 More information about DCTCP can be found in [1-4]. [1] http://simula.stanford.edu/~alizade/Site/DCTCP.html [2] http://simula.stanford.edu/~alizade/Site/DCTCP_files/dctcp-final.pdf [3] http://simula.stanford.edu/~alizade/Site/DCTCP_files/dctcp_analysis-full.pdf [4] http://tools.ietf.org/html/draft-bensley-tcpm-dctcp-00 Joint work with Florian Westphal and Glenn Judd. Signed-off-by:Glenn Judd <glenn.judd@morganstanley.com> Acked-by:
-
