- 20 Aug, 2017 10 commits
-
-
Martin KaFai Lau authored
This patch makes the needed changes to allow each process of the INNER_LRU_HASH_PREALLOC test to provide its numa node id when creating the lru map. Signed-off-by:
Martin KaFai Lau <kafai@fb.com> Acked-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
Alexei Starovoitov <ast@fb.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Martin KaFai Lau authored
The current map creation API does not allow to provide the numa-node preference. The memory usually comes from where the map-creation-process is running. The performance is not ideal if the bpf_prog is known to always run in a numa node different from the map-creation-process. One of the use case is sharding on CPU to different LRU maps (i.e. an array of LRU maps). Here is the test result of map_perf_test on the INNER_LRU_HASH_PREALLOC test if we force the lru map used by CPU0 to be allocated from a remote numa node: [ The machine has 20 cores. CPU0-9 at node 0. CPU10-19 at node 1 ] ># taskset -c 10 ./map_perf_test 512 8 1260000 8000000 5:inner_lru_hash_map_perf pre-alloc 1628380 events per sec 4:inner_lru_hash_map_perf pre-alloc 1626396 events per sec 3:inner_lru_hash_map_perf pre-alloc 1626144 events per sec 6:inner_lru_hash_map_perf pre-alloc 1621657 events per sec 2:inner_lru_hash_map_perf pre-alloc 1621534 events per sec 1:inner_lru_hash_map_perf pre-alloc 1620292 events per sec 7:inner_lru_hash_map_perf pre-alloc 1613305 events per sec 0:inner_lru_hash_map_perf pre-alloc 1239150 events per sec #<<< After specifying numa node: ># taskset -c 10 ./map_perf_test 512 8 1260000 8000000 5:inner_lru_hash_map_perf pre-alloc 1629627 events per sec 3:inner_lru_hash_map_perf pre-alloc 1628057 events per sec 1:inner_lru_hash_map_perf pre-alloc 1623054 events per sec 6:inner_lru_hash_map_perf pre-alloc 1616033 events per sec 2:inner_lru_hash_map_perf pre-alloc 1614630 events per sec 4:inner_lru_hash_map_perf pre-alloc 1612651 events per sec 7:inner_lru_hash_map_perf pre-alloc 1609337 events per sec 0:inner_lru_hash_map_perf pre-alloc 1619340 events per sec #<<< This patch adds one field, numa_node, to the bpf_attr. Since numa node 0 is a valid node, a new flag BPF_F_NUMA_NODE is also added. The numa_node field is honored if and only if the BPF_F_NUMA_NODE flag is set. Numa node selection is not supported for percpu map. This patch does not change all the kmalloc. F.e. 'htab = kzalloc()' is not changed since the object is small enough to stay in the cache. Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in a netdev_info message Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Colin Ian King authored
Trival fix to spelling mistakes: firware -> firmware invald -> invalid mutilcast -> multicast Signed-off-by:
-
David S. Miller authored
Arvind Yadav says: ==================== constify net eisa_device_id eisa_device_id are not supposed to change at runtime. All functions working with eisa_device_id provided by <linux/eisa.h> work with const eisa_device_id. So mark the non-const structs as const. ==================== Signed-off-by: -
Arvind Yadav authored
eisa_device_id are not supposed to change at runtime. All functions working with eisa_device_id provided by <linux/eisa.h> work with const eisa_device_id. So mark the non-const structs as const. Signed-off-by:
Arvind Yadav <arvind.yadav.cs@gmail.com> Reviewed-by:
Maciej W. Rozycki <macro@linux-mips.org> Signed-off-by:
-
Arvind Yadav authored
eisa_device_id are not supposed to change at runtime. All functions working with eisa_device_id provided by <linux/eisa.h> work with const eisa_device_id. So mark the non-const structs as const. Signed-off-by:
-
Arvind Yadav authored
eisa_device_id are not supposed to change at runtime. All functions working with eisa_device_id provided by <linux/eisa.h> work with const eisa_device_id. So mark the non-const structs as const. Signed-off-by:
-
Arvind Yadav authored
eisa_device_id are not supposed to change at runtime. All functions working with eisa_device_id provided by <linux/eisa.h> work with const eisa_device_id. So mark the non-const structs as const. Signed-off-by:
-
Arvind Yadav authored
eisa_device_id are not supposed to change at runtime. All functions working with eisa_device_id provided by <linux/eisa.h> work with const eisa_device_id. So mark the non-const structs as const. Signed-off-by:
-
- 19 Aug, 2017 24 commits
-
-
David S. Miller authored
Jakub Kicinski says: ==================== nfp: add basic ethtool callbacks to representors This set extends the basic ethtool functionality to representor netdevs. I start with providing link state via ethtool and then move on to functions such as driver information, statistics and FW log dump. The series contains a number of clean ups to the ethtool stats code too, some of the logic is simplified by making better use of the nfp_port abstraction. The stats we expose on representors are only the PCIe and MAC port statistics firmware maintains for us. ==================== Signed-off-by: -
Jakub Kicinski authored
We were reusing skb pointer when reading page frag, since ring entries contain a union of a skb and frag pointer. This can be confusing to people reading the code. Refactor the code to read frag pointer directly. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
Simon Horman <simon.horman@netronome.com> Signed-off-by:
-
Jakub Kicinski authored
Data and control vNICs currently use the same area name and error message. This could lead to confusion. Make sure the error message says "ctrl" in case of control and the data area is called "nfp.bar0". Signed-off-by:
-
Jakub Kicinski authored
Representors may be associated with both VFs or more importantly with physical ports. Allow vNIC and MAC statistics to be read with ethtool -S on representors. In case of vNICs we reuse the vNIC statistic helper, we just need to swap RX and TX to give statistics the "switch perspective." Signed-off-by:
-
Jakub Kicinski authored
Simplify the statistics handling code by keeping pointer to vNIC's config memory in nfp_port. Note that this is referring to the representor side of vNICs, vNIC side has the pointer in nfp_net. Signed-off-by:
-
Jakub Kicinski authored
Add reporting of MAC statistics in ethtool. MAC statistics are read out from the MAC IP and accumulated by application FW, therefore their presence depends on the application FW. Add missing defines and string names for the statistics and dump them in ethtool -S. Signed-off-by:
-
Jakub Kicinski authored
Store pointer to device memory containing MAC statistics in nfp_port. This simplifies representor code and will be used to dump those statistics in ethtool as well. Signed-off-by:
-
Jakub Kicinski authored
In preparation for reporting vNIC HW stats on representors split handling of the SW and HW stats in ethtool -S. Representors don't have SW stats (since vNIC is assigned to the VM). Remove the questionable defines which assume nn variable exists in the scope. Signed-off-by:
-
Jakub Kicinski authored
Add a helper for printing ethtool strings and advancing the pointer correctly. Signed-off-by:
-
Jakub Kicinski authored
We have been recently called out as a bad example for reporting standard netdev statistics as part of ethtool. Fix that :) Removing standard statistics allows us to simplify the structure holding definitions since we no longer have to mux different types of statistics. Signed-off-by:
-
Jakub Kicinski authored
Users should be able to dump the management FW logs on any of the driver's netdevs. Make the code only depend on the nfp_app and share it between vNICs and representors. Storing the dump flag is simply dropped for now, since we only support the argument being set to 0. Signed-off-by:
-
Jakub Kicinski authored
Extend representors' ethtool ops to show basic info like firmware version, driver version, and driver name. While at it don't set drvinfo.n_stats and drvinfo.regdump_len, core will invoke appropriate handlers to get those. A helper is added to turn a netdev into nfp_app for convenience. Signed-off-by:
-
Jakub Kicinski authored
Start linking ethtool ops to representors. Begin by adding a separate ops structure and providing link state. Next patches will convert appropriate functions to only use nfp_port, which will make them reusable on representors. Signed-off-by:
-
David S. Miller authored
Stephen Hemminger says: ==================== net: sysfs related cleanups Network sysfs infrastructure changes. Mostly related to using ro_after_init to make function tables immutable. ==================== Signed-off-by: -
stephen hemminger authored
Make code closer to current style. Mostly whitespace changes. Signed-off-by:
Stephen Hemminger <sthemmin@microsoft.com> Signed-off-by:
-
stephen hemminger authored
Signed-off-by:
-
stephen hemminger authored
The XPS queue attributes can be ro_after_init. Also use __ATTR_RX macros to simplify initialization. Signed-off-by:
-
stephen hemminger authored
Also fix macro to not have ; at end. Signed-off-by:
-
stephen hemminger authored
The show and store functions don't need/use the attribute. Signed-off-by:
-
stephen hemminger authored
The attributes of net devices are immutable. Ideally, attribute groups would contain const attributes but there are too many places that do modifications of list during startup (in other code) to allow that. Signed-off-by:
-
stephen hemminger authored
This can be const. Signed-off-by:
-
stephen hemminger authored
The net_class in sysfs is only modified on init. Signed-off-by:
-
stephen hemminger authored
These functions are wrapper arount class_create_file which can take a const attribute. Signed-off-by:
-
stephen hemminger authored
If kobject_init_and_add failed, then the failure path would decrement the reference count of the queue kobject whose reference count was already zero. Fixes: 114cf580 ("bql: Byte queue limits") Signed-off-by:
-
- 18 Aug, 2017 6 commits
-
-
David S. Miller authored
Tom Lendacky says: ==================== amd-xgbe: AMD XGBE driver updates 2017-08-17 The following updates are included in this driver update series: - Set the MDIO mode to clause 45 for the 10GBase-T configuration - Set the MII control width to 8-bits for speeds less than 1Gbps - Fix an issue to related to module removal when the devices are up - Fix ethtool statistics related to packet counting of TSO packets - Add support for device renaming - Add additional dynamic debug output for the PCS window calculation - Optimize reading of DMA channel interrupt enablement register - Add additional dynamic debug output about the hardware features - Add per queue Tx and Rx ethtool statistics - Add a macro to clear ethtool_link_ksettings modes - Convert the driver to use the ethtool_link_ksettings - Add support for VXLAN offload capabilities - Add additional ethtool statistics related to VXLAN This patch series is based on net-next. ==================== Signed-off-by: -
Lendacky, Thomas authored
Add some additional statistics for tracking VXLAN packets and checksum errors. Signed-off-by:
Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by:
-
Lendacky, Thomas authored
The hardware has the capability to perform checksum offload support (both Tx and Rx) and TSO support for VXLAN packets. Add the support required to enable this. The hardware can only support a single VXLAN port for offload. If more than one VXLAN port is added then the offload capabilities have to be disabled and can no longer be advertised. Signed-off-by:
-
Lendacky, Thomas authored
Convert from using the old u32 supported, advertising, etc. link settings to the new link mode settings that support bit positions / settings greater than 32 bits. Signed-off-by:
-
Lendacky, Thomas authored
There are currently macros to set and test an ETHTOOL_LINK_MODE_ setting, but not to clear one. Add a macro to clear an ETHTOOL_LINK_MODE_ setting. Signed-off-by:
-
Lendacky, Thomas authored
Add per queue Tx and Rx packet and byte counts. Signed-off-by:
-
