aboutsummaryrefslogtreecommitdiff
path: root/drivers/net/ethernet/intel/ice
AgeCommit message (Collapse)AuthorFilesLines
2020-05-31ice: support adding 16 unicast/multicast filter on untrusted VFPaul Greenwalt1-1/+4
Allow untrusted VF to add 16 unicast/multicast filters. VF uses 1 filter for the default/perm_addr/LAA MAC, 1 for broadcast, and 16 additional unicast/multicast filters. Signed-off-by: Paul Greenwalt <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-31ice: allow host to clear administratively set VF MACBrett Creeley1-8/+14
Currently a user is not allowed to clear a VF's administratively set MAC on the PF. Fix this by allowing an all zero MAC address via "ip link set ${pf_eth} vf ${vf_id} mac 00:00:00:00:00:00". An example use case for this would be issuing a "virsh shutdown" command on a VM. The call to iproute mentioned above is part of this flow. Without this change the driver incorrectly rejects clearing the VF's administratively set MAC and prints unhelpful log messages. Also, improve the comments surrounding this change. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Refactor VF VSI release and setup functionsBrett Creeley1-31/+55
Currently when a VF VSI calls ice_vsi_release() and ice_vsi_setup() it subsequently clears/sets the VF cached variables for lan_vsi_idx and lan_vsi_num. This works fine, but can be improved by handling this in the VF specific VSI release and setup functions. Also, when a VF VSI is setup too many parameters are passed that can be derived from the VF. Fix this by only calling VF VSI setup with the bare minimum parameters. Also, add functionality to invalidate a VF's VSI when it's released and/or setup fails. This will make it so a VF VSI cannot be accessed via its cached vsi_idx/vsi_num in these cases. Finally when a VF's VSI is invalidated set the lan_vsi_idx and lan_vsi_num to ICE_NO_VSI to clearly show that there is no valid VSI associated with this VF. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Refactor VF resetBrett Creeley2-187/+130
Currently VF VSI are being reset twice during a PFR or greater. This is causing reset, specifically resetting all VFs, to take too long. This is causing various issues with VF drivers not being able to gracefully handle the VF reset timeout. Fix this by refactoring how VF reset is handled for the case mentioned previously and for the VFR/VFLR case. The refactor was done by doing the following: 1. Removing the call to ice_vsi_rebuild_by_type for ICE_VSI_VF VSI, which was causing the initial VSI rebuild. 2. Adding functions for pre/post VSI rebuild functions that can be called in both the reset all VFs case and reset individual VF case. 3. Adding VSI rebuild functions that are specific for the reset all VFs case and adding functions that are specific for the reset individual VF case. 4. Calling the pre-rebuild function, then the specific VSI rebuild function based on the reset type, and then calling the post-rebuild function to handle VF resets. This patch series makes some assumptions about how VSI are handling by FW during reset: 1. During a PFR or greater all VSI in FW will be cleared. 2. During a VFR/VFLR the VSI rebuild responsibility is in the hands of the PF software. 3. There is code in the ice_reset_all_vfs() case to amortize operations if possible. This was left intact. 4. PF software should not be replaying VSI based filters that were added other than host configured, PF software configured, or the VF's default/LAA MAC. This is the VF drivers job after it has been reset. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: remove VM/VF disable command on CORER/GLOBR resetPaul Greenwalt1-11/+0
Remove VM/VF disable AQC (opcode 0x0C31) when resetting all VFs. This is not required for CORER/GLOBR reset. Signed-off-by: Paul Greenwalt <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Add functions to rebuild host VLAN/MAC config for a VFBrett Creeley1-32/+89
When resetting a VF the VLAN and MAC filter configurations need to be replayed. Add helper functions for this purpose. Signed-off-by: Brett Creeley <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Add function to set trust mode bit on resetBrett Creeley1-4/+13
As the title says, use a function to set trust mode bit on reset. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Renaming and simplification in VF init pathBrett Creeley1-29/+54
Some function names weren't very clear and some portions of VF creation could be moved into functions for clarity. Fix this by renaming some functions and move pieces of code into clearly name functions. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Separate VF VSI initialization/creation from reset flowBrett Creeley1-4/+106
Currently the same flow is used for VF VSI initialization/creation and VF VSI reset. This makes the initialization/creation flow unnecessarily complicated. Fix this by separating the initialization/creation of the VF VSI from the reset flow. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Add helper function for clearing VPGEN_VFRTRIGBrett Creeley1-11/+20
Create a helper function for clearing VPGEN_VFRTRIG as this needs to be done on reset to notify the VF that we are done resetting it. Also, it needs to be done on SR-IOV initialization/creation in case it was left in a bad state after SR-IOV tear down. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Simplify ice_sriov_configureBrett Creeley1-24/+48
Add a new function for checking if SR-IOV can be configured based on the PF and/or device's state/capabilities. Also, simplify the flow in ice_sriov_configure(). Signed-off-by: Brett Creeley <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Refactor ice_ena_vf_mappings to split MSIX and queue mappingsBrett Creeley1-32/+59
Currently ice_ena_vf_mappings() does all of the VF's MSIX and queue mapping in one function. This makes it hard to digest. Fix this by creating a new function for enabling MSIX mappings and one for enabling queue mappings. Also, rename some variables in the functions for clarity. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Declare functions staticTony Nguyen2-6/+3
ice_get_pfa_module_tlv() and ice_read_sr_word() are not being called outside of their file. Declare them as static. Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: fix kernel BUG if register_netdev failsJacob Keller1-2/+4
If register_netdev() fails, the driver will attempt to cleanup the q_vectors and inadvertently trigger a kernel BUG due to a NULL pointer dereference. This occurs because cleaning up q_vectors attempts to call netif_napi_del on napi_structs which were never initialized. Resolve this by releasing the netdev in ice_cfg_netdev and setting vsi->netdev to NULL. This ensures that after ice_cfg_netdev fails the state is rewound to match as if ice_cfg_netdev was never called. Signed-off-by: Jacob Keller <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: fix potential double free in probe unrollingJacob Keller1-1/+2
If ice_init_interrupt_scheme fails, ice_probe will jump to clearing up the interrupts. This can lead to some static analysis tools such as the compiler sanitizers complaining about double free problems. Since ice_init_interrupt_scheme already unrolls internally on failure, there is no need to call ice_clear_interrupt_scheme when it fails. Add a new unroll label and use that instead. Signed-off-by: Jacob Keller <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: cleanup VSI context initializationJacob Keller1-1/+0
Remove an unnecessary copy of vsi->info into ctxt->info in ice_vsi_init. This line is essentially a no-op because ice_set_dflt_vsi_ctx performs a memset to clear the info from the context structure. Signed-off-by: Jacob Keller <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-28ice: Poll for reset completion when DDP load failsAnirudh Venkataramanan1-0/+3
There are certain cases where the DDP load fails and the FW issues a core reset. For these cases, wait for reset to complete before proceeding with reset of the driver init. Signed-off-by: Anirudh Venkataramanan <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Check UMEM FQ size when allocating bufsKrzysztof Kazimierczak1-9/+21
If a UMEM is present on a queue when an interface/queue pair is being enabled, the driver will try to prepare the Rx buffers in advance to improve performance. However, if fill queue is shorter than HW Rx ring, the driver will report failure after getting the last address from the fill queue. This still lets the driver process the packets correctly during the NAPI poll, but leads to a constant NAPI rescheduling. Not allocating the buffers in advance would result in a potential performance decrease. Commit d57d76428ae9 ("xsk: Add API to check for available entries in FQ") provides an API that lets drivers check the number of addresses that the fill queue holds. Notify the user if fill queue is not long enough to prepare all buffers before packet processing starts, and allocate the buffers during the NAPI poll. If the fill queue size is sufficient, prepare Rx buffers in advance. Signed-off-by: Krzysztof Kazimierczak <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Refactor Rx checksum checksAnirudh Venkataramanan1-16/+11
We don't need both rx_status and rx_error parameters, as the latter is a subset of the former. Remove rx_error completely and check the right bit in rx_status. Rename rx_status to rx_status0, and rx_status_err1 to rx_status1. This naming more closely reflects the specification. Signed-off-by: Anirudh Venkataramanan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: avoid undefined behaviorBruce Allan4-6/+16
When writing the driver's struct ice_tlan_ctx structure, do not write the 8-bit element int_q_state with the associated internal-to-hardware field which is 122-bits, otherwise the helper function ice_write_byte() will use undefined behavior when setting the mask used for that write. This should not cause any functional change and will avoid use of undefined behavior. Also, update a comment to highlight this structure element is not written. Signed-off-by: Bruce Allan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Change number of XDP Tx queues to match number of Rx queuesMarta Plantykow2-2/+2
In current implementation number of XDP Tx queues is the same as the number of transmit queues, which is not always true. This patch changes this number to match the number of receive queues. XDP programs are running on Rx rings, so what we actually need to provide is the XDP Tx ring per each Rx ring so that the whole XDP ecosystem is functional, e.g. if the result of XDP prog is XDP_TX then you have the need to access the XDP Tx ring. Signed-off-by: Marta Plantykow <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Add XDP Tx to VSI ring statsMarta Plantykow1-9/+33
When XDP Tx program is loaded and packets are sent from interface, VSI statistics are not updated. This patch adds packets sent on Tx XDP ring to VSI ring stats. Signed-off-by: Marta Plantykow <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Change number of XDP TxQ to 0 when destroying ringsMarta Plantykow1-0/+3
When XDP Tx rings are destroyed the number of XDP Tx queues is not changing. This patch is changing this number to 0. Signed-off-by: Marta Plantykow <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Handle critical FW error during admin queue initializationEvan Swanson5-54/+80
A race condition between FW and SW can occur between admin queue setup and the first command sent. A link event may occur and FW attempts to notify a non-existent queue. FW will set the critical error bit and disable the queue. When this happens retry queue setup. Signed-off-by: Evan Swanson <[email protected]> Signed-off-by: Anirudh Venkataramanan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Don't allow VLAN stripping change when pvid setBrett Creeley1-0/+6
Currently, if the PVID is set in the VLAN handling section of the VSI context the driver still allows VLAN stripping to be enabled/disabled. VLAN stripping should only be modifiable when the PVID is not set. Fix this by preventing VLAN stripping modification when PVID is set. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Add more Rx errors to netdev's rx_error counterBrett Creeley1-1/+7
Currently we are only including illegal_bytes and rx_crc_errors in the PF netdev's rx_error counter. There are many more causes of Rx errors that the device supports and reports via Ethtool. Accumulate all Rx errors in the PF netdev's rx_error counter. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Fix for memory leaks and modify ICE_FREE_CQ_BUFSSurabhi Boob1-21/+28
Handle memory leaks during control queue initialization and buffer allocation failures. The macro ICE_FREE_CQ_BUFS is modified to re-use for this fix. Signed-off-by: Surabhi Boob <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: Fix memory leakSurabhi Boob1-1/+7
Handle memory leak on filter management initialization failure. Signed-off-by: Surabhi Boob <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: fix MAC write commandJesse Brandeburg2-10/+5
The manage MAC write command was implemented in an overly complex way that actually didn't work, as it wasn't symmetric to the manage MAC read command, and was feeding bytes out of order to the firmware. Fix the implementation by just using a simple array to represent the MAC address when it is being written via firmware command. Signed-off-by: Jesse Brandeburg <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: set VF default LAN addressPaul Greenwalt1-3/+5
Remove is_zero_ether_add() check when setting the VF default LAN address. This check assumed that the address had been delete and zeroed before calling ice_vc_add_mac_addr(). Now the default LAN address will be set to the last unicast MAC address added by the VF. The default LAN address is reported by the PF via ndo_get_vf_config. Signed-off-by: Paul Greenwalt <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: remove unused macroJesse Brandeburg1-2/+0
The driver had an unused define that can be removed. Found by compiler -Werror=unused-macros check. Signed-off-by: Jesse Brandeburg <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-27ice: fix signed vs unsigned comparisonsJesse Brandeburg3-9/+10
Fix the remaining signed vs unsigned issues, which appear when compiling with -Werror=sign-compare. Many of these are because there is an external interface that is passing an int to us (which we can't change) but that we (rightfully) store and compare against as an unsigned in our data structures. Signed-off-by: Jesse Brandeburg <[email protected]> Signed-off-by: Bruce Allan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-23Merge branch '100GbE' of ↵David S. Miller32-113/+5402
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue Jeff Kirsher says: ==================== 100GbE Intel Wired LAN Driver Updates 2020-05-22 This series contains updates to virtchnl and the ice driver. Geert Uytterhoeven fixes a data structure alignment issue in the virtchnl structures. Henry adds Flow Director support which allows for the redirection on ntuple rules over six patches. Initially Henry adds the initial infrastructure for Flow Director, and then later adds IPv4 and IPv6 support, as well as being able to display the ntuple rules. Bret add Accelerated Receive Flow Steering (aRFS) support which is used to steer receive flows to a specific queue. Fixes a transmit timeout when the VF link transitions from up/down/up because the transmit and receive queue interrupts are not enabled as part of VF's link up. Fixed an issue when the default VF LAN address is changed and after reset the PF will attempt to add the new MAC, which fails because it already exists. This causes the VF to be disabled completely until it is removed and enabled via sysfs. Anirudh (Ani) makes a fix where the ice driver needs to call set_mac_cfg to enable jumbo frames, so ensure it gets called during initialization and after reset. Fix bad register reads during a register dump in ethtool by removing the bad registers. Paul fixes an issue where the receive Malicious Driver Detection (MDD) auto reset message was not being logged because it occurred after the VF reset. Victor adds a check for compatibility between the Dynamic Device Personalization (DDP) package and the NIC firmware to ensure that everything aligns. Jesse fixes a administrative queue string call with the appropriate error reporting variable. Also fixed the loop variables that are comparing or assigning signed against unsigned values. ==================== Signed-off-by: David S. Miller <[email protected]>
2020-05-22ice: cleanup unsigned loopsJesse Brandeburg4-13/+14
Fix loop variables that are comparing or assigning signed against unsigned values, mostly by declaring loop counters as unsigned. Signed-off-by: Jesse Brandeburg <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: fix usage of incorrect variableJesse Brandeburg2-5/+5
The driver was using rq_last_status where it should have been using sq_last_status. Fix the string to be using the correct error reporting variable. Signed-off-by: Jesse Brandeburg <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Fix bad register readsAnirudh Venkataramanan2-6/+0
The "ethtool -d" handler reads registers in the ice_regs_dump_list array and returns read values back to the userspace. The register offsets PFINT0_ITR* are not valid as per the specification and reading these causes a "unable to handle kernel paging request" bug in the driver. Remove these registers from ice_regs_dump_list. Signed-off-by: Anirudh Venkataramanan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: check for compatibility between DDP package and firmwareVictor Raj6-30/+102
Require the Dynamic Device Personalization (DDP) file to have the same major version number and the same or older minor number than the firmware version major and minor, respectively. Check the OS and NVM package versions before downloading the package. If the OS package version is not compatible with NVM then return an appropriate error. Split the 32-byte segment name into a 28-byte segment name and a 4-byte Track-ID. Older packages will still work with this change because no package has a name that will take up more than 28 bytes; in this case the Track-ID will be 0. Note that the driver will store the segment name as 32-bytes in the ice_hw structure, in order to normalize the length of the various package name strings that it uses. Also add section ID and structure for the segment metadata section. Signed-off-by: Victor Raj <[email protected]> Signed-off-by: Dan Nowlin <[email protected]> Signed-off-by: Bruce Allan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Check if unicast MAC exists before setting VF MACBrett Creeley1-0/+43
Currently if a unicast MAC is set via ndo_set_vf_mac, the PF driver will set the VF's dflt_lan_addr.addr once some basic checks have passed. The VF is then reset. During reset the PF driver will attempt to program the VF's MAC from the dflt_lan_addr.addr field. This fails when the MAC already exists on the PF's switch. This is causing the VF to be completely disabled until removing/enabling any VFs via sysfs. Fix this by checking if the unicast MAC exists before triggering a VF reset directly in ndo_set_vf_mac. Also, add a check if the unicast MAC is set to the same value as before and return 0 if that is the case. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Fix Tx timeout when link is toggled on a VF's interfaceBrett Creeley1-0/+48
Currently if the iavf is loaded and a VF link transitions from up to down to up again a Tx timeout will be triggered. This happens because Tx/Rx queue interrupts are only enabled when receiving the VIRTCHNL_OP_CONFIG_MAP_IRQ message, which happens on reset or initial iavf driver load, but not when bringing link up. This is problematic because they are disabled on the VIRTCHNL_OP_DISABLE_QUEUES message, which is part of bringing a VF's link down. However, they are not enabled on the VIRTCHNL_OP_ENABLE_QUEUES message, which is part of bringing a VF's link up. Fix this by re-enabling the VF's Rx and Tx queue interrupts when they were previously configured. This is done by first checking to make sure the previous value in QINT_[R|T]QCTL.MSIX_INDX is not 0, which is used to represent the OICR in the VF's interrupt space. If the MSIX_INDX is non-zero then enable the interrupt by setting the QINT_[R|T]CTL.CAUSE_ENA bit to 1. Signed-off-by: Brett Creeley <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: print Rx MDD auto reset message before VF resetPaul Greenwalt3-7/+27
Rx MDD auto reset message was not being logged because logging occurred after the VF reset and the VF MDD data was reinitialized. Log the Rx MDD auto reset message before triggering the VF reset. Signed-off-by: Paul Greenwalt <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Call ice_aq_set_mac_cfgAnirudh Venkataramanan5-0/+103
As per the specification, the driver needs to call set_mac_cfg (opcode 0x0603) to be able to exercise jumbo frames. Call the function during initialization and the post reset rebuild flow. Signed-off-by: Anirudh Venkataramanan <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Implement aRFSBrett Creeley7-12/+805
Enable accelerated Receive Flow Steering (aRFS). It is used to steer Rx flows to a specific queue. This functionality is triggered by the network stack through ndo_rx_flow_steer and requires Flow Director (ntuple on) to function. The fltr_info is used to add/remove/update flow rules in the HW, the fltr_state is used to determine what to do with the filter with respect to HW and/or SW, and the flow_id is used in co-ordination with the network stack. The work for aRFS is split into two paths: the ndo_rx_flow_steer operation and the ice_service_task. The former is where the kernel hands us an Rx SKB among other items to setup aRFS and the latter is where the driver adds/updates/removes filter rules from HW and updates filter state. In the Rx path the following things can happen: 1. New aRFS entries are added to the hash table and the state is set to ICE_ARFS_INACTIVE so the filter can be updated in HW by the ice_service_task path. 2. aRFS entries have their Rx Queue updated if we receive a pre-existing flow_id and the filter state is ICE_ARFS_ACTIVE. The state is set to ICE_ARFS_INACTIVE so the filter can be updated in HW by the ice_service_task path. 3. aRFS entries marked as ICE_ARFS_TODEL are deleted In the ice_service_task path the following things can happen: 1. New aRFS entries marked as ICE_ARFS_INACTIVE are added or updated in HW. and their state is updated to ICE_ARFS_ACTIVE. 2. aRFS entries are deleted from HW and their state is updated to ICE_ARFS_TODEL. Signed-off-by: Brett Creeley <[email protected]> Signed-off-by: Madhu Chittim <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Restore filters following resetHenry Tieman4-6/+109
Following a reset, Flow Director filters are cleared from the hardware. Rebuild the filters using the software structures containing the filter rules. Signed-off-by: Henry Tieman <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Enable flex-bytes supportHenry Tieman6-2/+265
Flex-bytes allows for packet matching based on an offset and value. This is supported via the ethtool user-def option. It is specified by providing an offset followed by a 2 byte match value. Offset is measured from the start of the MAC address. The following restrictions apply to flex-bytes. The specified offset must be an even number and be smaller than 0x1fe. Example usage: ethtool -N eth0 flow-type tcp4 src-ip 192.168.0.55 dst-ip 172.16.0.55 \ src-port 12 dst-port 13 user-def 0x10ffff action 32 Signed-off-by: Henry Tieman <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Support IPv6 Flow Director filtersHenry Tieman5-81/+642
Extend supported filters to allow for IPv6 filters. Supported fields are: src-ip, dst-ip, src-port, and dst-port Supported flow-types are: tcp6, udp6, sctp6, ip6 Example usage: ethtool -N eth0 flow-type tcp6 src-port 12 dst-port 13 \ src-ip fce0::1:34 dst-ip fce0::1:35 action 32 Signed-off-by: Henry Tieman <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Support IPv4 Flow Director filtersHenry Tieman12-0/+1493
Support the addition and deletion of IPv4 filters. Supported fields are: src-ip, dst-ip, src-port, and dst-port Supported flow-types are: tcp4, udp4, sctp4, ip4 Example usage: ethtool -N eth0 flow-type tcp4 src-ip 192.168.0.55 dst-ip 172.16.0.55 \ src-port 16 dst-port 12 action 32 Signed-off-by: Henry Tieman <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Support displaying ntuple rulesHenry Tieman8-0/+280
Add functionality for ethtool --show-ntuple, allowing for filters to be displayed when set functionality is added. Add statistics related to Flow Director matches and status. Signed-off-by: Henry Tieman <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22ice: Initialize Flow Director resourcesHenry Tieman22-38/+1553
Flow Director allows for redirection based on ntuple rules. Rules are programmed using the ethtool set-ntuple interface. Supported actions are redirect to queue and drop. Setup the initial framework to process Flow Director filters. Create and allocate resources to manage and program filters to the hardware. Filters are processed via a sideband interface; a control VSI is created to manage communication and process requests through the sideband. Upon allocation of resources, update the hardware tables to accept perfect filters. Signed-off-by: Henry Tieman <[email protected]> Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
2020-05-22Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-nextDavid S. Miller4-360/+55
Daniel Borkmann says: ==================== pull-request: bpf-next 2020-05-23 The following pull-request contains BPF updates for your *net-next* tree. We've added 50 non-merge commits during the last 8 day(s) which contain a total of 109 files changed, 2776 insertions(+), 2887 deletions(-). The main changes are: 1) Add a new AF_XDP buffer allocation API to the core in order to help lowering the bar for drivers adopting AF_XDP support. i40e, ice, ixgbe as well as mlx5 have been moved over to the new API and also gained a small improvement in performance, from Björn Töpel and Magnus Karlsson. 2) Add getpeername()/getsockname() attach types for BPF sock_addr programs in order to allow for e.g. reverse translation of load-balancer backend to service address/port tuple from a connected peer, from Daniel Borkmann. 3) Improve the BPF verifier is_branch_taken() logic to evaluate pointers being non-NULL, e.g. if after an initial test another non-NULL test on that pointer follows in a given path, then it can be pruned right away, from John Fastabend. 4) Larger rework of BPF sockmap selftests to make output easier to understand and to reduce overall runtime as well as adding new BPF kTLS selftests that run in combination with sockmap, also from John Fastabend. 5) Batch of misc updates to BPF selftests including fixing up test_align to match verifier output again and moving it under test_progs, allowing bpf_iter selftest to compile on machines with older vmlinux.h, and updating config options for lirc and v6 segment routing helpers, from Stanislav Fomichev, Andrii Nakryiko and Alan Maguire. 6) Conversion of BPF tracing samples outdated internal BPF loader to use libbpf API instead, from Daniel T. Lee. 7) Follow-up to BPF kernel test infrastructure in order to fix a flake in the XDP selftests, from Jesper Dangaard Brouer. 8) Minor improvements to libbpf's internal hashmap implementation, from Ian Rogers. ==================== Signed-off-by: David S. Miller <[email protected]>
2020-05-21ice: Rename build_ctob to ice_build_ctobTony Nguyen4-10/+11
To make the function easier to identify as being part of the ice driver, prepend ice to the function name. Signed-off-by: Tony Nguyen <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>