- 10 Sep, 2019 6 commits
-
-
David S. Miller authored
Vivien Didelot says: ==================== net: dsa: mv88e6xxx: add PCL support This small series implements the ethtool RXNFC operations in the mv88e6xxx DSA driver to configure a port's Layer 2 Policy Control List (PCL) supported by models such as 88E6352 and 88E6390 and equivalent. This allows to configure a port to discard frames based on a configured destination or source MAC address and an optional VLAN, with e.g.: # ethtool --config-nfc lan1 flow-type ether src 00:11:22:33:44:55 action -1 ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
-
Vivien Didelot authored
Implement the .get_rxnfc and .set_rxnfc DSA operations to configure a port's Layer 2 Policy Control List (PCL) via ethtool. Currently only dropping frames based on MAC Destination or Source Address (including the option VLAN parameter) is supported. Signed-off-by:
Vivien Didelot <vivien.didelot@gmail.com> Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Vivien Didelot authored
Introduce a new .port_set_policy operation to configure a port's Policy Control List, based on mapping such as DA, SA, Etype and so on. Models similar to 88E6352 and 88E6390 are supported at the moment. Signed-off-by:
-
Vivien Didelot authored
Marvell has different values for the state of a MAC address, depending on its multicast bit. This patch completes the definitions for these states. At the same time, use 0 which is intuitive enough and simplifies the code a bit, instead of the UC or MC unused value. Signed-off-by:
-
Colin Ian King authored
Don't populate the array modes on the stack but instead make it static const. Makes the object code smaller by 303 bytes. Before: text data bss dec hex filename 51240 5008 1312 57560 e0d8 mellanox/mlx4/en_ethtool.o After: text data bss dec hex filename 50937 5008 1312 57257 dfa9 mellanox/mlx4/en_ethtool.o (gcc version 9.2.1, amd64) Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queueDavid S. Miller authored
Jeff Kirsher says: ==================== Intel Wired LAN Driver Updates 2019-09-09 This series contains a variety of cold and hot savoury changes to Intel drivers. Some of the fixes could be considered for stable even though the author did not request it. Hulk Robert cleans up (i.e. removes) a function that has no caller for the iavf driver. Radoslaw fixes an issue when there is no link in the VM after the hypervisor is restored from a low-power state due to the driver not properly restoring features in the device that had been disabled during the suspension for ixgbevf. Kai-Heng Feng modified e1000e to use mod_delayed_work() to help resolve a hot plug speed detection issue by adding a deterministic 1 second delay before running watchdog task after an interrupt. Sasha moves functions around to avoid forward declarations, since the forward declarations are not necessary for these static functions in igc. Also added a check for igc during driver probe to validate the NVM checksum. Cleaned up code defines that were not being used in the igc driver. Adds support for IP generic transmit checksum offload in the igc driver. Updated the iavf kernel documentation by a developer with no life. Jake provides another fm10k update to a local variable for ease of code readability. Mitch fixes the iavf driver to allow the VF to override the MAC address set by the host, if the VF is in "trusted" mode. Mauro S. M. Rodrigues provides several changes for i40e driver, first with resolving hw_dbg usage and referencing a i40e_hw attribute. Also implemented a debug macro using pr_debug, since the use of netdev_dbg could cause a NULL pointer dereference during probe. Finally cleaned up code that is no longer used or needed. Firo Yang provides a change in the ixgbe driver to ensure we sync the first fragment unconditionally to help resolve an issue seen in the XEN environment when the upper network stack could receive an incomplete network packet. Mariusz adds a missing device to the i40e PCI table in the driver. v2: Mauro S. M. Rodrigues updated patches 10 & 11 of the series based on feedback from Jakub Kicinski. Also updated patch 13 description so that the "Fixes:" tag was no wrapped. ==================== Signed-off-by:
-
- 09 Sep, 2019 15 commits
-
-
Mariusz Stachura authored
Add I40E_DEV_ID_10G_BASE_T_BC to i40e_pci_tbl Signed-off-by:
Mariusz Stachura <mariusz.stachura@intel.com> Tested-by:
Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
-
Sasha Neftin authored
Add IP generic TX checksum offload functionality. Signed-off-by:
Sasha Neftin <sasha.neftin@intel.com> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
-
Firo Yang authored
In Xen environment, if Xen-swiotlb is enabled, ixgbe driver could possibly allocate a page, DMA memory buffer, for the first fragment which is not suitable for Xen-swiotlb to do DMA operations. Xen-swiotlb have to internally allocate another page for doing DMA operations. This mechanism requires syncing the data from the internal page to the page which ixgbe sends to upper network stack. However, since commit f3213d93 ("ixgbe: Update driver to make use of DMA attributes in Rx path"), the unmap operation is performed with DMA_ATTR_SKIP_CPU_SYNC. As a result, the sync is not performed. Since the sync isn't performed, the upper network stack could receive a incomplete network packet. By incomplete, it means the linear data on the first fragment(between skb->head and skb->end) is invalid. So we have to copy the data from the internal xen-swiotlb page to the page which ixgbe sends to upper network stack through the sync operation. More details from Alexander Duyck: Specifically since we are mapping the frame with DMA_ATTR_SKIP_CPU_SYNC we have to unmap with that as well. As a result a sync is not performed on an unmap and must be done manually as we skipped it for the first frag. As such we need to always sync before possibly performing a page unmap operation. Fixes: f3213d93 ("ixgbe: Update driver to make use of DMA attributes in Rx path") Signed-off-by:
Firo Yang <firo.yang@suse.com> Reviewed-by:
Alexander Duyck <alexander.h.duyck@linux.intel.com> Tested-by:
-
Mauro S. M. Rodrigues authored
Since commit '5098850c ("i40e/i40evf: i40e_register.h updates")' it is no longer possible to trigger an EMP Reset from debugfs, but it's possible to request it either way, to end up with a bad reset request: echo empr > /sys/kernel/debug/i40e/0002\:01\:00.1/command i40e 0002:01:00.1: debugfs: forcing EMPR i40e 0002:01:00.1: bad reset request 0x00010000 So let's remove this piece of code and show the available valid commands as it is when any invalid command is issued. Signed-off-by:
"Mauro S. M. Rodrigues" <maurosr@linux.vnet.ibm.com> Tested-by:
-
Mauro S. M. Rodrigues authored
There are several uses of hw_dbg in the code, producing no output. This patch implements it using dev_debug. Initially the intention was to implement it using netdev_dbg, analogously to what is done in ixgbe for instance. That approach was avoided due to some early usages of hw_dbg, like i40e_pf_reset, before the VSI structure initialization causing NULL pointer dereference during the driver probe if the debug messages were turned on as soon as the module is probed. v2: - Use dev_dbg instead of pr_debug, and take advantage of dev_name instead of crafting pretty much the same device name locally as suggested by Jakub Kicinski. Signed-off-by:
-
Mauro S. M. Rodrigues authored
The mentioned function references a i40e_hw attribute, as parameter for hw_dbg, but it doesn't exist in the function scope. Fixes it by changing parameters from i40e_hmc_info to i40e_hw which can retrieve the necessary i40e_hmc_info. v2: - Fixed reverse xmas tree code style issue as suggested by Jakub Kicinski Signed-off-by:
-
Sasha Neftin authored
PCIe device control 2 defines does not use internally. This patch comes to clean up those. Signed-off-by:
-
Mitch Williams authored
Allow the VF to override the "permanent" MAC address set by the host. This allows bonding to work in the case where the administrator has set the VF MAC. Note that the VF must still be set to Trusted on the host if this change is to be accepted by the PF driver. Signed-off-by:
Mitch Williams <mitch.a.williams@intel.com> Tested-by:
-
Sasha Neftin authored
Add NVM checksum validation during probe functionality. Signed-off-by:
-
Jacob Keller authored
In the function fm10k_xmit_frame_ring, we recently switched to using the skb_frag_size accessor instead of directly using the size member of the skb fragment. This made the for loop slightly harder to read because it created a very long line that is difficult to split up. Avoid this by using a local variable in the for loop, so that we do not have to break the line on an open parenthesis. Signed-off-by:
Jacob Keller <jacob.e.keller@intel.com> Tested-by:
-
Jeff Kirsher authored
Update the LAN driver documentation to include the latest feature implementation and driver capabilities. Signed-off-by:
-
Sasha Neftin authored
Move igc_phy_setup_autoneg, igc_wait_autoneg and igc_set_fc_watermarks up to avoid forward declaration. It is not necessary to forward declare these static methods. Signed-off-by:
-
Kai-Heng Feng authored
After hot plugging an 1Gbps Ethernet cable with 1Gbps link partner, the MII_BMSR may report 10Mbps, renders the network rather slow. The issue has much lower fail rate after commit 59653e64 ("e1000e: Make watchdog use delayed work"), which essentially introduces some delay before running the watchdog task. But there's still a chance that the hot plugging event and the queued watchdog task gets run at the same time, then the original issue can be observed once again. So let's use mod_delayed_work() to add a deterministic 1 second delay before running watchdog task, after an interrupt. Signed-off-by:
Kai-Heng Feng <kai.heng.feng@canonical.com> Tested-by:
-
Radoslaw Tyl authored
This patch fixed issue in VM which shows no link when hypervisor is restored from low-power state. The driver is responsible for re-enabling any features of the device that had been disabled during suspend calls, such as IRQs and bus mastering. Signed-off-by:
Radoslaw Tyl <radoslawx.tyl@intel.com> Tested-by:
-
YueHaibing authored
There is no caller of function iavf_debug_d() in tree since commit 75051ce4 ("iavf: Fix up debug print macro"), so it can be removed. Reported-by:
Hulk Robot <hulkci@huawei.com> Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Tested-by:
-
- 07 Sep, 2019 19 commits
-
-
David S. Miller authored
Jakub Kicinski says: ==================== net/tls: small TX offload optimizations This set brings small TLS TX device optimizations. The biggest gain comes from fixing a misuse of non temporal copy instructions. On a synthetic workload modelled after customer's RFC application I see 3-5% percent gain. ==================== Signed-off-by: -
Jakub Kicinski authored
Unlike normal TCP code TLS has to touch the cache lines it copies into to fill header info. On memory-heavy workloads having non temporal stores and normal accesses targeting the same cache line leads to significant overhead. Measured 3% overhead running 3600 round robin connections with additional memory heavy workload. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
Dirk van der Merwe <dirk.vandermerwe@netronome.com> Signed-off-by:
-
Jakub Kicinski authored
For TLS device offload the tag/message authentication code are filled in by the device. The kernel merely reserves space for them. Because device overwrites it, the contents of the tag make do no matter. Current code tries to save space by reusing the header as the tag. This, however, leads to an additional frag being created and defeats buffer coalescing (which trickles all the way down to the drivers). Remove this optimization, and try to allocate the space for the tag in the usual way, leave the memory uninitialized. If memory allocation fails rewind the record pointer so that we use the already copied user data as tag. Note that the optimization was actually buggy, as the tag for TLS 1.2 is 16 bytes, but header is just 13, so the reuse may had looked past the end of the page.. Signed-off-by:
-
Jakub Kicinski authored
All modifications to TLS record list happen under the socket lock. Since records form an ordered queue readers are only concerned about elements being removed, additions can happen concurrently. Use RCU primitives to ensure the correct access types (READ_ONCE/WRITE_ONCE). Signed-off-by:
-
Jakub Kicinski authored
It's generally more cache friendly to walk arrays in order, especially those which are likely not in cache. Signed-off-by:
-
David S. Miller authored
Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next Johan Hedberg says: ==================== pull request: bluetooth-next 2019-09-06 Here's the main bluetooth-next pull request for the 5.4 kernel. - Cleanups & fixes to btrtl driver - Fixes for Realtek devices in btusb, e.g. for suspend handling - Firmware loading support for BCM4345C5 - hidp_send_message() return value handling fixes - Added support for utilizing Fast Advertising Interval - Various other minor cleanups & fixes Please let me know if there are any issues pulling. Thanks. ==================== Signed-off-by:
-
Colin Ian King authored
Don't populate the array spec_opcode on the stack but instead make it static const. Makes the object code smaller by 48 bytes. Before: text data bss dec hex filename 6914 1040 128 8082 1f92 hns3/hns3vf/hclgevf_cmd.o After: text data bss dec hex filename 6866 1040 128 8034 1f62 hns3/hns3vf/hclgevf_cmd.o (gcc version 9.2.1, amd64) Signed-off-by:
-
Colin Ian King authored
Don't populate the arrays on the stack but instead make them static const. Makes the object code smaller by 281 bytes. Before: text data bss dec hex filename 87553 5672 0 93225 16c29 benet/be_cmds.o After: text data bss dec hex filename 87112 5832 0 92944 16b10 benet/be_cmds.o (gcc version 9.2.1, amd64) Signed-off-by:
-
YueHaibing authored
Remove including <linux/version.h> that don't need it. Signed-off-by:
Shannon Nelson <snelson@pensando.io> Signed-off-by:
-
David S. Miller authored
Jose Abreu says: ==================== net: stmmac: Improvements and fixes for -next Improvements and fixes for recently introduced features. All for -next tree. More info in commit logs. ==================== Signed-off-by: -
Jose Abreu authored
We may have some SoCs that can't achieve XGMAC max speed. Limit it if asked to. Signed-off-by:
Jose Abreu <joabreu@synopsys.com> Signed-off-by:
-
Jose Abreu authored
Add a test to validate that Split Header feature is working correctly. It works by using the rececently introduced counter that increments each time a packet with split header is received. Signed-off-by:
-
Jose Abreu authored
We are already doing it by default in the TX path so we can also enable Jumbo Frame support in the RX path independently of MTU value. Signed-off-by:
-
Jose Abreu authored
We need to set the RX tail pointer so that RX engine starts working again after finishing the Flow Control test. Signed-off-by:
-
Jose Abreu authored
Add checks for support of Source Address Insertion/Replacement before running the test. Signed-off-by:
-
Hangbin Liu authored
This is a re-post of previous patch wrote by David Miller[1]. Phil Karn reported[2] that on busy networks with lots of unresolved multicast routing entries, the creation of new multicast group routes can be extremely slow and unreliable. The reason is we hard-coded multicast route entries with unresolved source addresses(cache_resolve_queue_len) to 10. If some multicast route never resolves and the unresolved source addresses increased, there will be no ability to create new multicast route cache. To resolve this issue, we need either add a sysctl entry to make the cache_resolve_queue_len configurable, or just remove cache_resolve_queue_len limit directly, as we already have the socket receive queue limits of mrouted socket, pointed by David. >From my side, I'd perfer to remove the cache_resolve_queue_len limit instead of creating two more(IPv4 and IPv6 version) sysctl entry. [1] https://lkml.org/lkml/2018/7/22/11 [2] https://lkml.org/lkml/2018/7/21/343 v3: instead of remove cache_resolve_queue_len totally, let's only remove the hard code limit when allocate the unresolved cache, as Eric Dumazet suggested, so we don't need to re-count it in other places. v2: hold the mfc_unres_lock while walking the unresolved list in queue_count(), as Nikolay Aleksandrov remind. Reported-by:
Phil Karn <karn@ka9q.net> Signed-off-by:
Hangbin Liu <liuhangbin@gmail.com> Reviewed-by:
Nikolay Aleksandrov <nikolay@cumulusnetworks.com> Signed-off-by:
-
David S. Miller authored
Haiyang Zhang says: ==================== hv_netvsc: Enable sg as tunable, sync offload settings to VF NIC This patch set fixes an issue in SG tuning, and sync offload settings from synthetic NIC to VF NIC. ==================== Signed-off-by: -
Haiyang Zhang authored
VF NIC may go down then come up during host servicing events. This causes the VF NIC offloading feature settings to roll back to the defaults. This patch can synchronize features from synthetic NIC to the VF NIC during ndo_set_features (ethtool -K), and netvsc_register_vf when VF comes back after host events. Signed-off-by:
Haiyang Zhang <haiyangz@microsoft.com> Cc: Mark Bloch <markb@mellanox.com> Signed-off-by:
-
Haiyang Zhang authored
In a previous patch, the NETIF_F_SG was missing after the code changes. That caused the SG feature to be "fixed". This patch includes it into hw_features, so it is tunable again. Fixes: 23312a3b ("netvsc: negotiate checksum and segmentation parameters") Signed-off-by:
-
