diff options
| -rw-r--r-- | sys/amd64/conf/GENERIC | 5 | ||||
| -rw-r--r-- | sys/amd64/conf/NOTES | 7 | ||||
| -rw-r--r-- | sys/conf/files.amd64 | 24 | ||||
| -rw-r--r-- | sys/dev/ixl/README | 410 | ||||
| -rw-r--r-- | sys/dev/ixl/i40e_osdep.c | 34 | ||||
| -rw-r--r-- | sys/dev/ixl/if_ixl.c | 216 | ||||
| -rw-r--r-- | sys/dev/ixl/if_ixlv.c | 2862 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl.h | 38 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_debug.h | 23 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_pf.h | 39 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_pf_iov.c | 195 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_pf_iov.h | 12 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_pf_main.c | 554 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_pf_qmgr.c | 20 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_pf_qmgr.h | 16 | ||||
| -rw-r--r-- | sys/dev/ixl/ixl_txrx.c | 249 | ||||
| -rw-r--r-- | sys/dev/ixl/ixlv.h | 151 | ||||
| -rw-r--r-- | sys/dev/ixl/ixlv_vc_mgr.h | 76 | ||||
| -rw-r--r-- | sys/dev/ixl/ixlvc.c | 614 | ||||
| -rw-r--r-- | sys/modules/Makefile | 3 | ||||
| -rw-r--r-- | sys/modules/iavf/Makefile (renamed from sys/modules/ixlv/Makefile) | 6 | ||||
| -rw-r--r-- | sys/modules/ixl/Makefile | 10 | ||||
| -rw-r--r-- | sys/net/iflib.c | 124 | ||||
| -rw-r--r-- | sys/net/iflib.h | 4 | ||||
| -rw-r--r-- | sys/net/iflib_private.h | 1 |
25 files changed, 2158 insertions, 3535 deletions
diff --git a/sys/amd64/conf/GENERIC b/sys/amd64/conf/GENERIC index da3d200fcec7..32b6db7dec61 100644 --- a/sys/amd64/conf/GENERIC +++ b/sys/amd64/conf/GENERIC @@ -240,9 +240,8 @@ device de # DEC/Intel DC21x4x (``Tulip'') device em # Intel PRO/1000 Gigabit Ethernet Family device ix # Intel PRO/10GbE PCIE PF Ethernet device ixv # Intel PRO/10GbE PCIE VF Ethernet -device ixl # Intel XL710 40Gbe PCIE Ethernet -#options IXL_IW # Enable iWARP Client Interface in ixl(4) -#device ixlv # Intel XL710 40Gbe VF PCIE Ethernet +device ixl # Intel 700 Series Physical Function +device iavf # Intel Adaptive Virtual Function device le # AMD Am7900 LANCE and Am79C9xx PCnet device ti # Alteon Networks Tigon I/II gigabit Ethernet device txp # 3Com 3cR990 (``Typhoon'') diff --git a/sys/amd64/conf/NOTES b/sys/amd64/conf/NOTES index e11120fbdd58..c56b13592a8a 100644 --- a/sys/amd64/conf/NOTES +++ b/sys/amd64/conf/NOTES @@ -313,8 +313,6 @@ options DRM_DEBUG # Include debug printfs (slow) # iwn: Intel Wireless WiFi Link 1000/105/135/2000/4965/5000/6000/6050 abgn # 802.11 network adapters # Requires the iwn firmware module -# ixl: Intel XL710 40Gbe PCIE Ethernet -# ixlv: Intel XL710 40Gbe VF PCIE Ethernet # mthca: Mellanox HCA InfiniBand # mlx4ib: Mellanox ConnectX HCA InfiniBand # mlx4en: Mellanox ConnectX HCA Ethernet @@ -332,9 +330,8 @@ options ED_SIC device ipw # Intel 2100 wireless NICs. device iwi # Intel 2200BG/2225BG/2915ABG wireless NICs. device iwn # Intel 4965/1000/5000/6000 wireless NICs. -device ixl # Intel XL710 40Gbe PCIE Ethernet -#options IXL_IW # Enable iWARP Client Interface in ixl(4) -#device ixlv # Intel XL710 40Gbe VF PCIE Ethernet +device ixl # Intel 700 Series Physical Function +device iavf # Intel Adaptive Virtual Function device mthca # Mellanox HCA InfiniBand device mlx4 # Shared code module between IB and Ethernet device mlx4ib # Mellanox ConnectX HCA InfiniBand diff --git a/sys/conf/files.amd64 b/sys/conf/files.amd64 index 0407dd30c7f1..3d0c60487b66 100644 --- a/sys/conf/files.amd64 +++ b/sys/conf/files.amd64 @@ -271,25 +271,23 @@ dev/ixl/ixl_pf_iov.c optional ixl pci pci_iov \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/ixl_pf_i2c.c optional ixl pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -#dev/ixl/ixl_iw.c optional ixl pci \ -# compile-with "${NORMAL_C} -I$S/dev/ixl" -#dev/ixl/if_ixlv.c optional ixlv pci \ -# compile-with "${NORMAL_C} -I$S/dev/ixl" -#dev/ixl/ixlvc.c optional ixlv pci \ -# compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/ixl_txrx.c optional ixl pci | ixlv pci \ +dev/ixl/if_ixlv.c optional iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/i40e_osdep.c optional ixl pci | ixlv pci \ +dev/ixl/ixlvc.c optional iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/i40e_lan_hmc.c optional ixl pci | ixlv pci \ +dev/ixl/ixl_txrx.c optional ixl pci | iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/i40e_hmc.c optional ixl pci | ixlv pci \ +dev/ixl/i40e_osdep.c optional ixl pci | iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/i40e_common.c optional ixl pci | ixlv pci \ +dev/ixl/i40e_lan_hmc.c optional ixl pci | iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/i40e_nvm.c optional ixl pci | ixlv pci \ +dev/ixl/i40e_hmc.c optional ixl pci | iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" -dev/ixl/i40e_adminq.c optional ixl pci | ixlv pci \ +dev/ixl/i40e_common.c optional ixl pci | iavf pci \ + compile-with "${NORMAL_C} -I$S/dev/ixl" +dev/ixl/i40e_nvm.c optional ixl pci | iavf pci \ + compile-with "${NORMAL_C} -I$S/dev/ixl" +dev/ixl/i40e_adminq.c optional ixl pci | iavf pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_dcb.c optional ixl pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" diff --git a/sys/dev/ixl/README b/sys/dev/ixl/README deleted file mode 100644 index dc0149ce623f..000000000000 --- a/sys/dev/ixl/README +++ /dev/null @@ -1,410 +0,0 @@ - ixl FreeBSD* Base Driver and ixlv VF Driver for the - Intel XL710 Ethernet Controller Family - -/*$FreeBSD$*/ -================================================================ - -August 26, 2014 - - -Contents -======== - -- Overview -- Supported Adapters -- The VF Driver -- Building and Installation -- Additional Configurations -- Known Limitations - - -Overview -======== - -This file describes the IXL FreeBSD* Base driver and the IXLV VF Driver -for the XL710 Ethernet Family of Adapters. The Driver has been developed -for use with FreeBSD 10.0 or later, but should be compatible with any -supported release. - -For questions related to hardware requirements, refer to the documentation -supplied with your Intel XL710 adapter. All hardware requirements listed -apply for use with FreeBSD. - - -Supported Adapters -================== - -The drivers in this release are compatible with XL710 and X710-based -Intel Ethernet Network Connections. - - -SFP+ Devices with Pluggable Optics ----------------------------------- - -SR Modules ----------- - Intel DUAL RATE 1G/10G SFP+ SR (bailed) FTLX8571D3BCV-IT - Intel DUAL RATE 1G/10G SFP+ SR (bailed) AFBR-703SDZ-IN2 - -LR Modules ----------- - Intel DUAL RATE 1G/10G SFP+ LR (bailed) FTLX1471D3BCV-IT - Intel DUAL RATE 1G/10G SFP+ LR (bailed) AFCT-701SDZ-IN2 - -QSFP+ Modules -------------- - Intel TRIPLE RATE 1G/10G/40G QSFP+ SR (bailed) E40GQSFPSR - Intel TRIPLE RATE 1G/10G/40G QSFP+ LR (bailed) E40GQSFPLR - QSFP+ 1G speed is not supported on XL710 based devices. - -X710/XL710 Based SFP+ adapters support all passive and active limiting direct -attach cables that comply with SFF-8431 v4.1 and SFF-8472 v10.4 specifications. - -The VF Driver -================== -The VF driver is normally used in a virtualized environment where a host -driver manages SRIOV, and provides a VF device to the guest. With this -first release the only host environment tested was using Linux QEMU/KVM. -Support is planned for Xen and VMWare hosts at a later time. - -In the FreeBSD guest the IXLV driver would be loaded and will function -using the VF device assigned to it. - -The VF driver provides most of the same functionality as the CORE driver, -but is actually a slave to the Host, access to many controls are actually -accomplished by a request to the Host via what is called the "Admin queue". -These are startup and initialization events however, once in operation -the device is self-contained and should achieve near native performance. - -Some notable limitations of the VF environment: for security reasons -the driver is never permitted to be promiscuous, therefore a tcpdump -will not behave the same with the interface. Second, media info is not -available from the PF, so it will always appear as auto. - -Tarball Building and Installation -========================= - -NOTE: You must have kernel sources installed to compile the driver tarball. - -These instructions assume a standalone driver tarball, building the driver -already in the kernel source is simply a matter of adding the device entry -to the kernel config file, or building in the ixl or ixlv module directory. - -In the instructions below, x.x.x is the driver version -as indicated in the name of the driver tarball. The example is -for ixl, the same procedure applies for ixlv. - -1. Move the base driver tar file to the directory of your choice. - For example, use /home/username/ixl or /usr/local/src/ixl. - -2. Untar/unzip the archive: - tar xfz ixl-x.x.x.tar.gz - -3. To install man page: - cd ixl-x.x.x - gzip -c ixl.4 > /usr/share/man/man4/ixl.4.gz - -4. To load the driver onto a running system: - cd ixl-x.x.x/src - make load - -5. To assign an IP address to the interface, enter the following: - ifconfig ixl<interface_num> <IP_address> - -6. Verify that the interface works. Enter the following, where <IP_address> - is the IP address for another machine on the same subnet as the interface - that is being tested: - - ping <IP_address> - -7. If you want the driver to load automatically when the system is booted: - - cd ixl-x.x.x/src - make - make install - - Edit /boot/loader.conf, and add the following line: - if_ixl_load="YES" - - Edit /etc/rc.conf, and create the appropriate - ifconfig_ixl<interface_num> entry: - - ifconfig_ixl<interface_num>="<ifconfig_settings>" - - Example usage: - - ifconfig_ixl0="inet 192.168.10.1 netmask 255.255.255.0" - - NOTE: For assistance, see the ifconfig man page. - - - -Configuration and Tuning -========================= - -Both drivers supports Transmit/Receive Checksum Offload for IPv4 and IPv6, -TSO forIPv4 and IPv6, LRO, and Jumbo Frames on all 40 Gigabit adapters. - - Jumbo Frames - ------------ - To enable Jumbo Frames, use the ifconfig utility to increase - the MTU beyond 1500 bytes. - - - The Jumbo Frames setting on the switch must be set to at least - 22 byteslarger than that of the adapter. - - - The maximum MTU setting for Jumbo Frames is 9706. This value - coincides with the maximum jumbo frames size of 9728. - To modify the setting, enter the following: - - ifconfig ixl<interface_num> <hostname or IP address> mtu 9000 - - - To confirm an interface's MTU value, use the ifconfig command. - To confirm the MTU used between two specific devices, use: - - route get <destination_IP_address> - - VLANs - ----- - To create a new VLAN pseudo-interface: - - ifconfig <vlan_name> create - - To associate the VLAN pseudo-interface with a physical interface - and assign a VLAN ID, IP address, and netmask: - - ifconfig <vlan_name> <ip_address> netmask <subnet_mask> vlan - <vlan_id> vlandev <physical_interface> - - Example: - - ifconfig vlan10 10.0.0.1 netmask 255.255.255.0 vlan 10 vlandev ixl0 - - In this example, all packets will be marked on egress with - 802.1Q VLAN tags, specifying a VLAN ID of 10. - - To remove a VLAN pseudo-interface: - - ifconfig <vlan_name> destroy - - - Checksum Offload - ---------------- - - Checksum offloading supports IPv4 and IPv6 with TCP and UDP packets - and is supported for both transmit and receive. Checksum offloading - for transmit and recieve is enabled by default for both IPv4 and IPv6. - - Checksum offloading can be enabled or disabled using ifconfig. - Transmit and receive offloading for IPv4 and Ipv6 are enabled - and disabled seperately. - - NOTE: TSO requires Tx checksum, so when Tx checksum - is disabled, TSO will also be disabled. - - To enable Tx checksum offloading for ipv4: - - ifconfig ixl<interface_num> txcsum4 - - To disable Tx checksum offloading for ipv4: - - ifconfig ixl<interface_num> -txcsum4 - (NOTE: This will disable TSO4) - - To enable Rx checksum offloading for ipv6: - - ifconfig ixl<interface_num> rxcsum6 - - To disable Rx checksum offloading for ipv6: - - ifconfig ixl<interface_num> -rxcsum6 - (NOTE: This will disable TSO6) - - - To confirm the current settings: - - ifconfig ixl<interface_num> - - - TSO - --- - - TSO supports both IPv4 and IPv6 and is enabled by default. TSO can - be disabled and enabled using the ifconfig utility. - - NOTE: TSO requires Tx checksum, so when Tx checksum is - disabled, TSO will also be disabled. - - To disable TSO IPv4: - - ifconfig ixl<interface_num> -tso4 - - To enable TSO IPv4: - - ifconfig ixl<interface_num> tso4 - - To disable TSO IPv6: - - ifconfig ixl<interface_num> -tso6 - - To enable TSO IPv6: - - ifconfig ixl<interface_num> tso6 - - To disable BOTH TSO IPv4 and IPv6: - - ifconfig ixl<interface_num> -tso - - To enable BOTH TSO IPv4 and IPv6: - - ifconfig ixl<interface_num> tso - - - LRO - --- - - Large Receive Offload is enabled by default. It can be enabled - or disabled by using the ifconfig utility. - - NOTE: LRO should be disabled when forwarding packets. - - To disable LRO: - - ifconfig ixl<interface_num> -lro - - To enable LRO: - - ifconfig ixl<interface_num> lro - - -Flow Control (IXL only) ------------- -Flow control is disabled by default. To change flow control settings use sysctl. - -To enable flow control to Rx pause frames: - - sysctl dev.ixl.<interface_num>.fc=1 - -To enable flow control to Tx pause frames: - - sysctl dev.ixl.<interface_num>.fc=2 - -To enable flow control to Rx and Tx pause frames: - - sysctl dev.ixl.<interface_num>.fc=3 - -To disable flow control: - - sysctl dev.ixl.<interface_num>.fc=0 - - -NOTE: You must have a flow control capable link partner. - -NOTE: The VF driver does not have access to flow control, it must be - managed from the host side. - - - Important system configuration changes: - ======================================= - --Change the file /etc/sysctl.conf, and add the line: - - hw.intr_storm_threshold: 0 (the default is 1000) - --Best throughput results are seen with a large MTU; use 9706 if possible. - --The default number of descriptors per ring is 1024, increasing this may -improve performance depending on the use case. - --The VF driver uses a relatively large buf ring, this was found to eliminate - UDP transmit errors, it is a tuneable, and if no UDP traffic is used it can - be reduced. It is memory used per queue. - - -Known Limitations -================= - -Network Memory Buffer allocation --------------------------------- - FreeBSD may have a low number of network memory buffers (mbufs) by default. -If your mbuf value is too low, it may cause the driver to fail to initialize -and/or cause the system to become unresponsive. You can check to see if the -system is mbuf-starved by running 'netstat -m'. Increase the number of mbufs -by editing the lines below in /etc/sysctl.conf: - - kern.ipc.nmbclusters - kern.ipc.nmbjumbop - kern.ipc.nmbjumbo9 - kern.ipc.nmbjumbo16 - kern.ipc.nmbufs - -The amount of memory that you allocate is system specific, and may -require some trial and error. - -Also, increasing the follwing in /etc/sysctl.conf could help increase -network performance: - - kern.ipc.maxsockbuf - net.inet.tcp.sendspace - net.inet.tcp.recvspace - net.inet.udp.maxdgram - net.inet.udp.recvspace - - -UDP Stress Test Dropped Packet Issue ------------------------------------- -Under small packet UDP stress test with the ixl driver, the FreeBSD system -may drop UDP packets due to the fullness of socket buffers. You may want to -change the driver's Flow Control variables to the minimum value for -controlling packet reception. - - -Disable LRO when routing/bridging ---------------------------------- -LRO must be turned off when forwarding traffic. - - -Lower than expected performance -------------------------------- -Some PCIe x8 slots are actually configured as x4 slots. These slots have -insufficient bandwidth for full line rate with dual port and quad port -devices. - -In addition, if you put a PCIe Generation 3-capable adapter into a PCIe -Generation 2 slot, you cannot get full bandwidth. The driver detects this -situation and writes the following message in the system log: - - "PCI-Express bandwidth available for this card is not sufficient for - optimal performance. For optimal performance a x8 PCI-Express slot - is required." - -If this error occurs, moving your adapter to a true PCIe Generation 3 x8 -slot will resolve the issue. - - -Support -======= - -For general information and support, go to the Intel support website at: - - http://support.intel.com - -If an issue is identified with the released source code on the supported kernel -with a supported adapter, email the specific information related to the issue -to freebsdnic@mailbox.intel.com. - - -License -======= - -This software program is released under the terms of a license agreement -between you ('Licensee') and Intel. Do not use or load this software or any -associated materials (collectively, the 'Software') until you have carefully -read the full terms and conditions of the LICENSE located in this software -package. By loadingor using the Software, you agree to the terms of this -Agreement. If you do not agree with the terms of this Agreement, do not -install or use the Software. - -* Other names and brands may be claimed as the property of others. - - diff --git a/sys/dev/ixl/i40e_osdep.c b/sys/dev/ixl/i40e_osdep.c index 3dad1f76ce40..ee8ca26c5d7a 100644 --- a/sys/dev/ixl/i40e_osdep.c +++ b/sys/dev/ixl/i40e_osdep.c @@ -161,27 +161,25 @@ i40e_destroy_spinlock(struct i40e_spinlock *lock) mtx_destroy(&lock->mutex); } +static inline int +ixl_ms_scale(int x) +{ + if (hz == 1000) + return (x); + else if (hz > 1000) + return (x*(hz/1000)); + else + return (max(1, x/(1000/hz))); +} + void i40e_msec_pause(int msecs) { - int ticks_to_pause = (msecs * hz) / 1000; - int start_ticks = ticks; - - if (cold || SCHEDULER_STOPPED()) { + if (cold || SCHEDULER_STOPPED()) i40e_msec_delay(msecs); - return; - } - - while (1) { - kern_yield(PRI_USER); - int yielded_ticks = ticks - start_ticks; - if (yielded_ticks > ticks_to_pause) - break; - else if (yielded_ticks < 0 - && (yielded_ticks + INT_MAX + 1 > ticks_to_pause)) { - break; - } - } + else + // ERJ: (msecs * hz) could overflow + pause("ixl", ixl_ms_scale(msecs)); } /* @@ -272,7 +270,5 @@ i40e_write_pci_cfg(struct i40e_hw *hw, u32 reg, u16 value) { pci_write_config(((struct i40e_osdep *)hw->back)->dev, reg, value, 2); - - return; } diff --git a/sys/dev/ixl/if_ixl.c b/sys/dev/ixl/if_ixl.c index 23c9d13c695d..99b14b52e782 100644 --- a/sys/dev/ixl/if_ixl.c +++ b/sys/dev/ixl/if_ixl.c @@ -48,7 +48,7 @@ * Driver version *********************************************************************/ #define IXL_DRIVER_VERSION_MAJOR 2 -#define IXL_DRIVER_VERSION_MINOR 0 +#define IXL_DRIVER_VERSION_MINOR 1 #define IXL_DRIVER_VERSION_BUILD 0 #define IXL_DRIVER_VERSION_STRING \ @@ -115,10 +115,11 @@ static void ixl_if_timer(if_ctx_t ctx, uint16_t qid); static void ixl_if_vlan_register(if_ctx_t ctx, u16 vtag); static void ixl_if_vlan_unregister(if_ctx_t ctx, u16 vtag); static uint64_t ixl_if_get_counter(if_ctx_t ctx, ift_counter cnt); -static void ixl_if_vflr_handle(if_ctx_t ctx); -// static void ixl_if_link_intr_enable(if_ctx_t ctx); static int ixl_if_i2c_req(if_ctx_t ctx, struct ifi2creq *req); static int ixl_if_priv_ioctl(if_ctx_t ctx, u_long command, caddr_t data); +#ifdef PCI_IOV +static void ixl_if_vflr_handle(if_ctx_t ctx); +#endif /*** Other ***/ static int ixl_mc_filter_apply(void *arg, struct ifmultiaddr *ifma, int); @@ -137,9 +138,9 @@ static device_method_t ixl_methods[] = { DEVMETHOD(device_detach, iflib_device_detach), DEVMETHOD(device_shutdown, iflib_device_shutdown), #ifdef PCI_IOV - DEVMETHOD(pci_iov_init, ixl_iov_init), - DEVMETHOD(pci_iov_uninit, ixl_iov_uninit), - DEVMETHOD(pci_iov_add_vf, ixl_add_vf), + DEVMETHOD(pci_iov_init, iflib_device_iov_init), + DEVMETHOD(pci_iov_uninit, iflib_device_iov_uninit), + DEVMETHOD(pci_iov_add_vf, iflib_device_iov_add_vf), #endif DEVMETHOD_END }; @@ -169,7 +170,6 @@ static device_method_t ixl_if_methods[] = { DEVMETHOD(ifdi_msix_intr_assign, ixl_if_msix_intr_assign), DEVMETHOD(ifdi_intr_enable, ixl_if_enable_intr), DEVMETHOD(ifdi_intr_disable, ixl_if_disable_intr), - //DEVMETHOD(ifdi_link_intr_enable, ixl_if_link_intr_enable), DEVMETHOD(ifdi_rx_queue_intr_enable, ixl_if_rx_queue_intr_enable), DEVMETHOD(ifdi_tx_queue_intr_enable, ixl_if_tx_queue_intr_enable), DEVMETHOD(ifdi_tx_queues_alloc, ixl_if_tx_queues_alloc), @@ -185,9 +185,14 @@ static device_method_t ixl_if_methods[] = { DEVMETHOD(ifdi_vlan_register, ixl_if_vlan_register), DEVMETHOD(ifdi_vlan_unregister, ixl_if_vlan_unregister), DEVMETHOD(ifdi_get_counter, ixl_if_get_counter), - DEVMETHOD(ifdi_vflr_handle, ixl_if_vflr_handle), DEVMETHOD(ifdi_i2c_req, ixl_if_i2c_req), DEVMETHOD(ifdi_priv_ioctl, ixl_if_priv_ioctl), +#ifdef PCI_IOV + DEVMETHOD(ifdi_iov_init, ixl_if_iov_init), + DEVMETHOD(ifdi_iov_uninit, ixl_if_iov_uninit), + DEVMETHOD(ifdi_iov_vf_add, ixl_if_iov_vf_add), + DEVMETHOD(ifdi_vflr_handle, ixl_if_vflr_handle), +#endif // ifdi_led_func // ifdi_debug DEVMETHOD_END @@ -202,7 +207,7 @@ static driver_t ixl_if_driver = { */ static SYSCTL_NODE(_hw, OID_AUTO, ixl, CTLFLAG_RD, 0, - "IXL driver parameters"); + "ixl driver parameters"); /* * Leave this on unless you need to send flow control @@ -222,6 +227,13 @@ SYSCTL_INT(_hw_ixl, OID_AUTO, i2c_access_method, CTLFLAG_RDTUN, &ixl_i2c_access_method, 0, IXL_SYSCTL_HELP_I2C_METHOD); +static int ixl_enable_vf_loopback = 1; +TUNABLE_INT("hw.ixl.enable_vf_loopback", + &ixl_enable_vf_loopback); +SYSCTL_INT(_hw_ixl, OID_AUTO, enable_vf_loopback, CTLFLAG_RDTUN, + &ixl_enable_vf_loopback, 0, + IXL_SYSCTL_HELP_VF_LOOPBACK); + /* * Different method for processing TX descriptor * completion. @@ -333,9 +345,9 @@ ixl_register(device_t dev) static int ixl_allocate_pci_resources(struct ixl_pf *pf) { - int rid; - struct i40e_hw *hw = &pf->hw; device_t dev = iflib_get_dev(pf->vsi.ctx); + struct i40e_hw *hw = &pf->hw; + int rid; /* Map BAR0 */ rid = PCIR_BAR(0); @@ -386,21 +398,17 @@ ixl_if_attach_pre(if_ctx_t ctx) enum i40e_status_code status; int error = 0; - INIT_DEBUGOUT("ixl_if_attach_pre: begin"); + INIT_DBG_DEV(dev, "begin"); - /* Allocate, clear, and link in our primary soft structure */ dev = iflib_get_dev(ctx); pf = iflib_get_softc(ctx); + vsi = &pf->vsi; vsi->back = pf; pf->dev = dev; hw = &pf->hw; - /* - ** Note this assumes we have a single embedded VSI, - ** this could be enhanced later to allocate multiple - */ - //vsi->dev = pf->dev; + vsi->dev = dev; vsi->hw = &pf->hw; vsi->id = 0; vsi->num_vlans = 0; @@ -545,6 +553,7 @@ ixl_if_attach_pre(if_ctx_t ctx) * sizeof(struct i40e_tx_desc), DBA_ALIGN); scctx->isc_txrx = &ixl_txrx_dwb; } + scctx->isc_txrx->ift_legacy_intr = ixl_intr; scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union i40e_32byte_rx_desc), DBA_ALIGN); scctx->isc_msix_bar = PCIR_BAR(IXL_MSIX_BAR); @@ -556,7 +565,7 @@ ixl_if_attach_pre(if_ctx_t ctx) scctx->isc_tx_csum_flags = CSUM_OFFLOAD; scctx->isc_capabilities = scctx->isc_capenable = IXL_CAPS; - INIT_DEBUGOUT("ixl_if_attach_pre: end"); + INIT_DBG_DEV(dev, "end"); return (0); err_mac_hmc: @@ -579,7 +588,7 @@ ixl_if_attach_post(if_ctx_t ctx) int error = 0; enum i40e_status_code status; - INIT_DEBUGOUT("ixl_if_attach_post: begin"); + INIT_DBG_DEV(dev, "begin"); dev = iflib_get_dev(ctx); pf = iflib_get_softc(ctx); @@ -587,6 +596,10 @@ ixl_if_attach_post(if_ctx_t ctx) vsi->ifp = iflib_get_ifp(ctx); hw = &pf->hw; + /* Save off determined number of queues for interface */ + vsi->num_rx_queues = vsi->shared->isc_nrxqsets; + vsi->num_tx_queues = vsi->shared->isc_ntxqsets; + /* Setup OS network interface / ifnet */ if (ixl_setup_interface(dev, pf)) { device_printf(dev, "interface setup failed!\n"); @@ -694,6 +707,10 @@ err: return (error); } +/** + * XXX: iflib always ignores the return value of detach() + * -> This means that this isn't allowed to fail + */ static int ixl_if_detach(if_ctx_t ctx) { @@ -702,7 +719,7 @@ ixl_if_detach(if_ctx_t ctx) struct i40e_hw *hw = &pf->hw; device_t dev = pf->dev; enum i40e_status_code status; -#if defined(PCI_IOV) || defined(IXL_IW) +#ifdef IXL_IW int error; #endif @@ -713,17 +730,10 @@ ixl_if_detach(if_ctx_t ctx) error = ixl_iw_pf_detach(pf); if (error == EBUSY) { device_printf(dev, "iwarp in use; stop it first.\n"); - return (error); + //return (error); } } #endif -#ifdef PCI_IOV - error = pci_iov_detach(dev); - if (error != 0) { - device_printf(dev, "SR-IOV in use; detach first.\n"); - return (error); - } -#endif /* Remove all previously allocated media types */ ifmedia_removeall(vsi->media); @@ -751,7 +761,6 @@ ixl_if_detach(if_ctx_t ctx) return (0); } -/* TODO: Do shutdown-specific stuff here */ static int ixl_if_shutdown(if_ctx_t ctx) { @@ -796,43 +805,13 @@ ixl_if_resume(if_ctx_t ctx) return (0); } -/* Set Report Status queue fields to 0 */ -static void -ixl_init_tx_rsqs(struct ixl_vsi *vsi) -{ - if_softc_ctx_t scctx = vsi->shared; - struct ixl_tx_queue *tx_que; - int i, j; - - for (i = 0, tx_que = vsi->tx_queues; i < vsi->num_tx_queues; i++, tx_que++) { - struct tx_ring *txr = &tx_que->txr; - - txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0; - - for (j = 0; j < scctx->isc_ntxd[0]; j++) - txr->tx_rsq[j] = QIDX_INVALID; - } -} - -static void -ixl_init_tx_cidx(struct ixl_vsi *vsi) -{ - struct ixl_tx_queue *tx_que; - int i; - - for (i = 0, tx_que = vsi->tx_queues; i < vsi->num_tx_queues; i++, tx_que++) { - struct tx_ring *txr = &tx_que->txr; - - txr->tx_cidx_processed = 0; - } -} - void ixl_if_init(if_ctx_t ctx) { struct ixl_pf *pf = iflib_get_softc(ctx); struct ixl_vsi *vsi = &pf->vsi; struct i40e_hw *hw = &pf->hw; + struct ifnet *ifp = iflib_get_ifp(ctx); device_t dev = iflib_get_dev(ctx); u8 tmpaddr[ETHER_ADDR_LEN]; int ret; @@ -840,12 +819,12 @@ ixl_if_init(if_ctx_t ctx) /* * If the aq is dead here, it probably means something outside of the driver * did something to the adapter, like a PF reset. - * So rebuild the driver's state here if that occurs. + * So, rebuild the driver's state here if that occurs. */ if (!i40e_check_asq_alive(&pf->hw)) { device_printf(dev, "Admin Queue is down; resetting...\n"); ixl_teardown_hw_structs(pf); - ixl_reset(pf); + ixl_rebuild_hw_structs_after_reset(pf); } /* Get the latest mac address... User might use a LAA */ @@ -872,8 +851,7 @@ ixl_if_init(if_ctx_t ctx) return; } - // TODO: Call iflib setup multicast filters here? - // It's called in ixgbe in D5213 + /* Reconfigure multicast filters in HW */ ixl_if_multi_set(ctx); /* Set up RSS */ @@ -895,8 +873,12 @@ ixl_if_init(if_ctx_t ctx) i40e_aq_set_default_vsi(hw, vsi->seid, NULL); + /* Re-add configure filters to HW */ ixl_reconfigure_filters(vsi); + /* Configure promiscuous mode */ + ixl_if_promisc_set(ctx, if_getflags(ifp)); + #ifdef IXL_IW if (ixl_enable_iwarp && pf->iw_enabled) { ret = ixl_iw_pf_init(pf); @@ -923,7 +905,7 @@ ixl_if_stop(if_ctx_t ctx) #endif ixl_disable_rings_intr(vsi); - ixl_disable_rings(vsi); + ixl_disable_rings(pf, vsi, &pf->qtag); } static int @@ -936,6 +918,9 @@ ixl_if_msix_intr_assign(if_ctx_t ctx, int msix) int err, i, rid, vector = 0; char buf[16]; + MPASS(vsi->shared->isc_nrxqsets > 0); + MPASS(vsi->shared->isc_ntxqsets > 0); + /* Admin Que must use vector 0*/ rid = vector + 1; err = iflib_irq_alloc_generic(ctx, &vsi->irq, rid, IFLIB_INTR_ADMIN, @@ -943,14 +928,14 @@ ixl_if_msix_intr_assign(if_ctx_t ctx, int msix) if (err) { iflib_irq_free(ctx, &vsi->irq); device_printf(iflib_get_dev(ctx), - "Failed to register Admin que handler"); + "Failed to register Admin Que handler"); return (err); } - // TODO: Re-enable this at some point - // iflib_softirq_alloc_generic(ctx, rid, IFLIB_INTR_IOV, pf, 0, "ixl_iov"); + /* Create soft IRQ for handling VFLRs */ + iflib_softirq_alloc_generic(ctx, &pf->iov_irq, IFLIB_INTR_IOV, pf, 0, "iov"); /* Now set up the stations */ - for (i = 0, vector = 1; i < vsi->num_rx_queues; i++, vector++, rx_que++) { + for (i = 0, vector = 1; i < vsi->shared->isc_nrxqsets; i++, vector++, rx_que++) { rid = vector + 1; snprintf(buf, sizeof(buf), "rxq%d", i); @@ -960,7 +945,7 @@ ixl_if_msix_intr_assign(if_ctx_t ctx, int msix) * what's expected in the iflib context? */ if (err) { device_printf(iflib_get_dev(ctx), - "Failed to allocate q int %d err: %d", i, err); + "Failed to allocate queue RX int vector %d, err: %d\n", i, err); vsi->num_rx_queues = i + 1; goto fail; } @@ -969,16 +954,16 @@ ixl_if_msix_intr_assign(if_ctx_t ctx, int msix) bzero(buf, sizeof(buf)); - for (i = 0; i < vsi->num_tx_queues; i++, tx_que++) { + for (i = 0; i < vsi->shared->isc_ntxqsets; i++, tx_que++) { snprintf(buf, sizeof(buf), "txq%d", i); iflib_softirq_alloc_generic(ctx, - &vsi->rx_queues[i % vsi->num_rx_queues].que_irq, + &vsi->rx_queues[i % vsi->shared->isc_nrxqsets].que_irq, IFLIB_INTR_TX, tx_que, tx_que->txr.me, buf); /* TODO: Maybe call a strategy function for this to figure out which * interrupts to map Tx queues to. I don't know if there's an immediately * better way than this other than a user-supplied map, though. */ - tx_que->msix = (i % vsi->num_rx_queues) + 1; + tx_que->msix = (i % vsi->shared->isc_nrxqsets) + 1; } return (0); @@ -1051,11 +1036,10 @@ ixl_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid) { struct ixl_pf *pf = iflib_get_softc(ctx); struct ixl_vsi *vsi = &pf->vsi; - struct i40e_hw *hw = vsi->hw; - struct ixl_tx_queue *tx_que = &vsi->tx_queues[txqid]; + struct i40e_hw *hw = vsi->hw; + struct ixl_tx_queue *tx_que = &vsi->tx_queues[txqid]; ixl_enable_queue(hw, tx_que->msix - 1); - return (0); } @@ -1066,12 +1050,11 @@ ixl_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxq struct ixl_vsi *vsi = &pf->vsi; if_softc_ctx_t scctx = vsi->shared; struct ixl_tx_queue *que; - // int i; int i, j, error = 0; - MPASS(vsi->num_tx_queues > 0); + MPASS(scctx->isc_ntxqsets > 0); MPASS(ntxqs == 1); - MPASS(vsi->num_tx_queues == ntxqsets); + MPASS(scctx->isc_ntxqsets == ntxqsets); /* Allocate queue structure memory */ if (!(vsi->tx_queues = @@ -1118,9 +1101,12 @@ ixl_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxq struct ixl_rx_queue *que; int i, error = 0; - MPASS(vsi->num_rx_queues > 0); +#ifdef INVARIANTS + if_softc_ctx_t scctx = vsi->shared; + MPASS(scctx->isc_nrxqsets > 0); MPASS(nrxqs == 1); - MPASS(vsi->num_rx_queues == nrxqsets); + MPASS(scctx->isc_nrxqsets == nrxqsets); +#endif /* Allocate queue structure memory */ if (!(vsi->rx_queues = @@ -1156,7 +1142,7 @@ ixl_if_queues_free(if_ctx_t ctx) struct ixl_pf *pf = iflib_get_softc(ctx); struct ixl_vsi *vsi = &pf->vsi; - if (vsi->enable_head_writeback) { + if (!vsi->enable_head_writeback) { struct ixl_tx_queue *que; int i = 0; @@ -1208,6 +1194,20 @@ ixl_update_link_status(struct ixl_pf *pf) } } +static void +ixl_handle_lan_overflow_event(struct ixl_pf *pf, struct i40e_arq_event_info *e) +{ + device_t dev = pf->dev; + u32 rxq_idx, qtx_ctl; + + rxq_idx = (e->desc.params.external.param0 & I40E_PRTDCB_RUPTQ_RXQNUM_MASK) >> + I40E_PRTDCB_RUPTQ_RXQNUM_SHIFT; + qtx_ctl = e->desc.params.external.param1; + + device_printf(dev, "LAN overflow event: global rxq_idx %d\n", rxq_idx); + device_printf(dev, "LAN overflow event: QTX_CTL 0x%08x\n", qtx_ctl); +} + static int ixl_process_adminq(struct ixl_pf *pf, u16 *pending) { @@ -1248,7 +1248,7 @@ ixl_process_adminq(struct ixl_pf *pf, u16 *pending) * aren't currently configured. */ case i40e_aqc_opc_event_lan_overflow: - device_printf(dev, "LAN overflow event\n"); + ixl_handle_lan_overflow_event(pf, &event); break; default: break; @@ -1278,13 +1278,9 @@ ixl_if_update_admin_status(if_ctx_t ctx) if (pf->state & IXL_PF_STATE_MDD_PENDING) ixl_handle_mdd_event(pf); -#ifdef PCI_IOV - if (pf->state & IXL_PF_STATE_VF_RESET_REQ) - iflib_iov_intr_deferred(ctx); -#endif - ixl_process_adminq(pf, &pending); ixl_update_link_status(pf); + ixl_update_stats_counters(pf); /* * If there are still messages to process, reschedule ourselves. @@ -1301,14 +1297,16 @@ ixl_if_multi_set(if_ctx_t ctx) { struct ixl_pf *pf = iflib_get_softc(ctx); struct ixl_vsi *vsi = &pf->vsi; - struct i40e_hw *hw = vsi->hw; - int mcnt = 0, flags; + struct i40e_hw *hw = vsi->hw; + int mcnt = 0, flags; + int del_mcnt; IOCTL_DEBUGOUT("ixl_if_multi_set: begin"); mcnt = if_multiaddr_count(iflib_get_ifp(ctx), MAX_MULTICAST_ADDR); - /* delete existing MC filters */ - ixl_del_multi(vsi); + /* Delete filters for removed multicast addresses */ + del_mcnt = ixl_del_multi(vsi); + vsi->num_macs -= del_mcnt; if (__predict_false(mcnt == MAX_MULTICAST_ADDR)) { i40e_aq_set_vsi_multicast_promiscuous(hw, @@ -1316,13 +1314,17 @@ ixl_if_multi_set(if_ctx_t ctx) return; } /* (re-)install filters for all mcast addresses */ + /* XXX: This bypasses filter count tracking code! */ mcnt = if_multi_apply(iflib_get_ifp(ctx), ixl_mc_filter_apply, vsi); if (mcnt > 0) { + vsi->num_macs += mcnt; flags = (IXL_FILTER_ADD | IXL_FILTER_USED | IXL_FILTER_MC); ixl_add_hw_filters(vsi, flags, mcnt); } + ixl_dbg_filter(pf, "%s: filter mac total: %d\n", + __func__, vsi->num_macs); IOCTL_DEBUGOUT("ixl_if_multi_set: end"); } @@ -1518,32 +1520,11 @@ ixl_if_promisc_set(if_ctx_t ctx, int flags) static void ixl_if_timer(if_ctx_t ctx, uint16_t qid) { - struct ixl_pf *pf = iflib_get_softc(ctx); - //struct i40e_hw *hw = &pf->hw; - //struct ixl_tx_queue *que = &vsi->tx_queues[qid]; - #if 0 - u32 mask; - - /* - ** Check status of the queues - */ - mask = (I40E_PFINT_DYN_CTLN_INTENA_MASK | - I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK); - - /* If queue param has outstanding work, trigger sw irq */ - // TODO: TX queues in iflib don't use HW interrupts; does this do anything? - if (que->busy) - wr32(hw, I40E_PFINT_DYN_CTLN(que->txr.me), mask); -#endif - if (qid != 0) return; /* Fire off the adminq task */ iflib_admin_intr_deferred(ctx); - - /* Update stats */ - ixl_update_stats_counters(pf); } static void @@ -1612,13 +1593,15 @@ ixl_if_get_counter(if_ctx_t ctx, ift_counter cnt) } } +#ifdef PCI_IOV static void ixl_if_vflr_handle(if_ctx_t ctx) { - IXL_DEV_ERR(iflib_get_dev(ctx), ""); + struct ixl_pf *pf = iflib_get_softc(ctx); - // TODO: call ixl_handle_vflr() + ixl_handle_vflr(pf); } +#endif static int ixl_if_i2c_req(if_ctx_t ctx, struct ifi2creq *req) @@ -1676,6 +1659,7 @@ ixl_save_pf_tunables(struct ixl_pf *pf) pf->dbg_mask = ixl_core_debug_mask; pf->hw.debug_mask = ixl_shared_debug_mask; pf->vsi.enable_head_writeback = !!(ixl_enable_head_writeback); + pf->enable_vf_loopback = !!(ixl_enable_vf_loopback); #if 0 pf->dynamic_rx_itr = ixl_dynamic_rx_itr; pf->dynamic_tx_itr = ixl_dynamic_tx_itr; diff --git a/sys/dev/ixl/if_ixlv.c b/sys/dev/ixl/if_ixlv.c index dd72d52ab187..9d3495e3b1ed 100644 --- a/sys/dev/ixl/if_ixlv.c +++ b/sys/dev/ixl/if_ixlv.c @@ -32,19 +32,19 @@ ******************************************************************************/ /*$FreeBSD$*/ -#include "ixl.h" #include "ixlv.h" /********************************************************************* * Driver version *********************************************************************/ -#define IXLV_DRIVER_VERSION_MAJOR 1 -#define IXLV_DRIVER_VERSION_MINOR 5 -#define IXLV_DRIVER_VERSION_BUILD 4 +#define IAVF_DRIVER_VERSION_MAJOR 2 +#define IAVF_DRIVER_VERSION_MINOR 0 +#define IAVF_DRIVER_VERSION_BUILD 0 -char ixlv_driver_version[] = __XSTRING(IXLV_DRIVER_VERSION_MAJOR) "." - __XSTRING(IXLV_DRIVER_VERSION_MINOR) "." - __XSTRING(IXLV_DRIVER_VERSION_BUILD) "-iflib-k"; +#define IAVF_DRIVER_VERSION_STRING \ + __XSTRING(IAVF_DRIVER_VERSION_MAJOR) "." \ + __XSTRING(IAVF_DRIVER_VERSION_MINOR) "." \ + __XSTRING(IAVF_DRIVER_VERSION_BUILD) "-k" /********************************************************************* * PCI Device ID Table @@ -56,9 +56,9 @@ char ixlv_driver_version[] = __XSTRING(IXLV_DRIVER_VERSION_MAJOR) "." static pci_vendor_info_t ixlv_vendor_info_array[] = { - {I40E_INTEL_VENDOR_ID, I40E_DEV_ID_VF, 0, 0, 0}, - {I40E_INTEL_VENDOR_ID, I40E_DEV_ID_X722_VF, 0, 0, 0}, - {I40E_INTEL_VENDOR_ID, I40E_DEV_ID_ADAPTIVE_VF, 0, 0, 0}, + PVID(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_VF, "Intel(R) Ethernet Virtual Function 700 Series"), + PVID(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_X722_VF, "Intel(R) Ethernet Virtual Function 700 Series (X722)"), + PVID(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_ADAPTIVE_VF, "Intel(R) Ethernet Adaptive Virtual Function"), /* required last entry */ PVID_END }; @@ -66,7 +66,7 @@ static pci_vendor_info_t ixlv_vendor_info_array[] = /********************************************************************* * Function prototypes *********************************************************************/ -static void *ixlv_register(device_t dev); +static void *ixlv_register(device_t dev); static int ixlv_if_attach_pre(if_ctx_t ctx); static int ixlv_if_attach_post(if_ctx_t ctx); static int ixlv_if_detach(if_ctx_t ctx); @@ -76,7 +76,8 @@ static int ixlv_if_resume(if_ctx_t ctx); static int ixlv_if_msix_intr_assign(if_ctx_t ctx, int msix); static void ixlv_if_enable_intr(if_ctx_t ctx); static void ixlv_if_disable_intr(if_ctx_t ctx); -static int ixlv_if_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid); +static int ixlv_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid); +static int ixlv_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid); static int ixlv_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets); static int ixlv_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nqs, int nqsets); static void ixlv_if_queues_free(if_ctx_t ctx); @@ -100,8 +101,8 @@ static int ixlv_vf_config(struct ixlv_sc *); static void ixlv_init_filters(struct ixlv_sc *); static void ixlv_free_pci_resources(struct ixlv_sc *); static void ixlv_free_filters(struct ixlv_sc *); -static void ixlv_setup_interface(device_t, struct ixl_vsi *); -static void ixlv_add_sysctls(struct ixlv_sc *); +static void ixlv_setup_interface(device_t, struct ixlv_sc *); +static void ixlv_add_device_sysctls(struct ixlv_sc *); static void ixlv_enable_adminq_irq(struct i40e_hw *); static void ixlv_disable_adminq_irq(struct i40e_hw *); static void ixlv_enable_queue_irq(struct i40e_hw *, int); @@ -113,21 +114,24 @@ static int ixlv_add_mac_filter(struct ixlv_sc *, u8 *, u16); static int ixlv_del_mac_filter(struct ixlv_sc *sc, u8 *macaddr); static int ixlv_msix_que(void *); static int ixlv_msix_adminq(void *); -static void ixlv_do_adminq_locked(struct ixlv_sc *sc); -static void ixl_init_cmd_complete(struct ixl_vc_cmd *, void *, - enum i40e_status_code); -static void ixlv_configure_itr(struct ixlv_sc *); - -static void ixlv_setup_vlan_filters(struct ixlv_sc *); - -static char *ixlv_vc_speed_to_string(enum virtchnl_link_speed link_speed); -static int ixlv_sysctl_current_speed(SYSCTL_HANDLER_ARGS); - -// static void ixlv_add_sysctls(struct ixlv_sc *); -#ifdef IXL_DEBUG -static int ixlv_sysctl_qtx_tail_handler(SYSCTL_HANDLER_ARGS); -static int ixlv_sysctl_qrx_tail_handler(SYSCTL_HANDLER_ARGS); -#endif +//static void ixlv_del_multi(struct ixlv_sc *sc); +static void ixlv_init_multi(struct ixlv_sc *sc); +static void ixlv_configure_itr(struct ixlv_sc *sc); + +static int ixlv_sysctl_rx_itr(SYSCTL_HANDLER_ARGS); +static int ixlv_sysctl_tx_itr(SYSCTL_HANDLER_ARGS); +static int ixlv_sysctl_current_speed(SYSCTL_HANDLER_ARGS); +static int ixlv_sysctl_sw_filter_list(SYSCTL_HANDLER_ARGS); +static int ixlv_sysctl_queue_interrupt_table(SYSCTL_HANDLER_ARGS); +static int ixlv_sysctl_vf_reset(SYSCTL_HANDLER_ARGS); +static int ixlv_sysctl_vflr_reset(SYSCTL_HANDLER_ARGS); + +char *ixlv_vc_speed_to_string(enum virtchnl_link_speed link_speed); +static void ixlv_save_tunables(struct ixlv_sc *); +static enum i40e_status_code + ixlv_process_adminq(struct ixlv_sc *, u16 *); +static int ixlv_send_vc_msg(struct ixlv_sc *sc, u32 op); +static int ixlv_send_vc_msg_sleep(struct ixlv_sc *sc, u32 op); /********************************************************************* * FreeBSD Device Interface Entry Points @@ -144,17 +148,21 @@ static device_method_t ixlv_methods[] = { }; static driver_t ixlv_driver = { - "ixlv", ixlv_methods, sizeof(struct ixlv_sc), + "iavf", ixlv_methods, sizeof(struct ixlv_sc), }; devclass_t ixlv_devclass; -DRIVER_MODULE(ixlv, pci, ixlv_driver, ixlv_devclass, 0, 0); +DRIVER_MODULE(iavf, pci, ixlv_driver, ixlv_devclass, 0, 0); MODULE_PNP_INFO("U32:vendor;U32:device;U32:subvendor;U32:subdevice;U32:revision", - pci, ixlv, ixlv_vendor_info_array, + pci, iavf, ixlv_vendor_info_array, nitems(ixlv_vendor_info_array) - 1); -MODULE_DEPEND(ixlv, pci, 1, 1, 1); -MODULE_DEPEND(ixlv, ether, 1, 1, 1); -MODULE_DEPEND(ixlv, iflib, 1, 1, 1); +MODULE_VERSION(iavf, 1); + +MODULE_DEPEND(iavf, pci, 1, 1, 1); +MODULE_DEPEND(iavf, ether, 1, 1, 1); +MODULE_DEPEND(iavf, iflib, 1, 1, 1); + +MALLOC_DEFINE(M_IXLV, "iavf", "iavf driver allocations"); static device_method_t ixlv_if_methods[] = { DEVMETHOD(ifdi_attach_pre, ixlv_if_attach_pre), @@ -168,14 +176,14 @@ static device_method_t ixlv_if_methods[] = { DEVMETHOD(ifdi_msix_intr_assign, ixlv_if_msix_intr_assign), DEVMETHOD(ifdi_intr_enable, ixlv_if_enable_intr), DEVMETHOD(ifdi_intr_disable, ixlv_if_disable_intr), - DEVMETHOD(ifdi_queue_intr_enable, ixlv_if_queue_intr_enable), + DEVMETHOD(ifdi_rx_queue_intr_enable, ixlv_if_rx_queue_intr_enable), + DEVMETHOD(ifdi_tx_queue_intr_enable, ixlv_if_tx_queue_intr_enable), DEVMETHOD(ifdi_tx_queues_alloc, ixlv_if_tx_queues_alloc), DEVMETHOD(ifdi_rx_queues_alloc, ixlv_if_rx_queues_alloc), DEVMETHOD(ifdi_queues_free, ixlv_if_queues_free), DEVMETHOD(ifdi_update_admin_status, ixlv_if_update_admin_status), DEVMETHOD(ifdi_multi_set, ixlv_if_multi_set), DEVMETHOD(ifdi_mtu_set, ixlv_if_mtu_set), - // DEVMETHOD(ifdi_crcstrip_set, ixlv_if_crcstrip_set), DEVMETHOD(ifdi_media_status, ixlv_if_media_status), DEVMETHOD(ifdi_media_change, ixlv_if_media_change), DEVMETHOD(ifdi_promisc_set, ixlv_if_promisc_set), @@ -187,95 +195,73 @@ static device_method_t ixlv_if_methods[] = { }; static driver_t ixlv_if_driver = { - "ixlv_if", ixlv_if_methods, sizeof(struct ixlv_sc) + "iavf_if", ixlv_if_methods, sizeof(struct ixlv_sc) }; /* ** TUNEABLE PARAMETERS: */ -static SYSCTL_NODE(_hw, OID_AUTO, ixlv, CTLFLAG_RD, 0, - "IXLV driver parameters"); - -/* -** Number of descriptors per ring: -** - TX and RX sizes are independently configurable -*/ -static int ixlv_tx_ring_size = IXL_DEFAULT_RING; -TUNABLE_INT("hw.ixlv.tx_ring_size", &ixlv_tx_ring_size); -SYSCTL_INT(_hw_ixlv, OID_AUTO, tx_ring_size, CTLFLAG_RDTUN, - &ixlv_tx_ring_size, 0, "TX Descriptor Ring Size"); - -static int ixlv_rx_ring_size = IXL_DEFAULT_RING; -TUNABLE_INT("hw.ixlv.rx_ring_size", &ixlv_rx_ring_size); -SYSCTL_INT(_hw_ixlv, OID_AUTO, rx_ring_size, CTLFLAG_RDTUN, - &ixlv_rx_ring_size, 0, "TX Descriptor Ring Size"); - -/* Set to zero to auto calculate */ -int ixlv_max_queues = 0; -TUNABLE_INT("hw.ixlv.max_queues", &ixlv_max_queues); -SYSCTL_INT(_hw_ixlv, OID_AUTO, max_queues, CTLFLAG_RDTUN, - &ixlv_max_queues, 0, "Number of Queues"); +static SYSCTL_NODE(_hw, OID_AUTO, iavf, CTLFLAG_RD, 0, + "iavf driver parameters"); /* * Different method for processing TX descriptor * completion. */ static int ixlv_enable_head_writeback = 0; -TUNABLE_INT("hw.ixlv.enable_head_writeback", +TUNABLE_INT("hw.iavf.enable_head_writeback", &ixlv_enable_head_writeback); -SYSCTL_INT(_hw_ixlv, OID_AUTO, enable_head_writeback, CTLFLAG_RDTUN, +SYSCTL_INT(_hw_iavf, OID_AUTO, enable_head_writeback, CTLFLAG_RDTUN, &ixlv_enable_head_writeback, 0, "For detecting last completed TX descriptor by hardware, use value written by HW instead of checking descriptors"); -/* -** Controls for Interrupt Throttling -** - true/false for dynamic adjustment -** - default values for static ITR -*/ -int ixlv_dynamic_rx_itr = 0; -TUNABLE_INT("hw.ixlv.dynamic_rx_itr", &ixlv_dynamic_rx_itr); -SYSCTL_INT(_hw_ixlv, OID_AUTO, dynamic_rx_itr, CTLFLAG_RDTUN, - &ixlv_dynamic_rx_itr, 0, "Dynamic RX Interrupt Rate"); +static int ixlv_core_debug_mask = 0; +TUNABLE_INT("hw.iavf.core_debug_mask", + &ixlv_core_debug_mask); +SYSCTL_INT(_hw_iavf, OID_AUTO, core_debug_mask, CTLFLAG_RDTUN, + &ixlv_core_debug_mask, 0, + "Display debug statements that are printed in non-shared code"); -int ixlv_dynamic_tx_itr = 0; -TUNABLE_INT("hw.ixlv.dynamic_tx_itr", &ixlv_dynamic_tx_itr); -SYSCTL_INT(_hw_ixlv, OID_AUTO, dynamic_tx_itr, CTLFLAG_RDTUN, - &ixlv_dynamic_tx_itr, 0, "Dynamic TX Interrupt Rate"); +static int ixlv_shared_debug_mask = 0; +TUNABLE_INT("hw.iavf.shared_debug_mask", + &ixlv_shared_debug_mask); +SYSCTL_INT(_hw_iavf, OID_AUTO, shared_debug_mask, CTLFLAG_RDTUN, + &ixlv_shared_debug_mask, 0, + "Display debug statements that are printed in shared code"); int ixlv_rx_itr = IXL_ITR_8K; -TUNABLE_INT("hw.ixlv.rx_itr", &ixlv_rx_itr); -SYSCTL_INT(_hw_ixlv, OID_AUTO, rx_itr, CTLFLAG_RDTUN, +TUNABLE_INT("hw.iavf.rx_itr", &ixlv_rx_itr); +SYSCTL_INT(_hw_iavf, OID_AUTO, rx_itr, CTLFLAG_RDTUN, &ixlv_rx_itr, 0, "RX Interrupt Rate"); int ixlv_tx_itr = IXL_ITR_4K; -TUNABLE_INT("hw.ixlv.tx_itr", &ixlv_tx_itr); -SYSCTL_INT(_hw_ixlv, OID_AUTO, tx_itr, CTLFLAG_RDTUN, +TUNABLE_INT("hw.iavf.tx_itr", &ixlv_tx_itr); +SYSCTL_INT(_hw_iavf, OID_AUTO, tx_itr, CTLFLAG_RDTUN, &ixlv_tx_itr, 0, "TX Interrupt Rate"); -extern struct if_txrx ixl_txrx; +extern struct if_txrx ixl_txrx_hwb; +extern struct if_txrx ixl_txrx_dwb; static struct if_shared_ctx ixlv_sctx_init = { .isc_magic = IFLIB_MAGIC, .isc_q_align = PAGE_SIZE,/* max(DBA_ALIGN, PAGE_SIZE) */ .isc_tx_maxsize = IXL_TSO_SIZE + sizeof(struct ether_vlan_header), - .isc_tx_maxsegsize = PAGE_SIZE, + .isc_tx_maxsegsize = IXL_MAX_DMA_SEG_SIZE, .isc_tso_maxsize = IXL_TSO_SIZE + sizeof(struct ether_vlan_header), - .isc_tso_maxsegsize = PAGE_SIZE, - // TODO: Review the rx_maxsize and rx_maxsegsize params - // Where are they used in iflib? + .isc_tso_maxsegsize = IXL_MAX_DMA_SEG_SIZE, .isc_rx_maxsize = 16384, - .isc_rx_nsegments = 1, - .isc_rx_maxsegsize = 16384, - // TODO: What is isc_nfl for? + .isc_rx_nsegments = IXL_MAX_RX_SEGS, + .isc_rx_maxsegsize = IXL_MAX_DMA_SEG_SIZE, .isc_nfl = 1, .isc_ntxqs = 1, .isc_nrxqs = 1, .isc_admin_intrcnt = 1, .isc_vendor_info = ixlv_vendor_info_array, - .isc_driver_version = ixlv_driver_version, + .isc_driver_version = IAVF_DRIVER_VERSION_STRING, .isc_driver = &ixlv_if_driver, + .isc_flags = IFLIB_NEED_SCRATCH | IFLIB_NEED_ZERO_CSUM | IFLIB_IS_VF, .isc_nrxd_min = {IXL_MIN_RING}, .isc_ntxd_min = {IXL_MIN_RING}, @@ -288,64 +274,82 @@ static struct if_shared_ctx ixlv_sctx_init = { if_shared_ctx_t ixlv_sctx = &ixlv_sctx_init; /*** Functions ***/ - static void * ixlv_register(device_t dev) { return (ixlv_sctx); - } +} + +static int +ixlv_allocate_pci_resources(struct ixlv_sc *sc) +{ + struct i40e_hw *hw = &sc->hw; + device_t dev = iflib_get_dev(sc->vsi.ctx); + int rid; + + /* Map BAR0 */ + rid = PCIR_BAR(0); + sc->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, + &rid, RF_ACTIVE); + + if (!(sc->pci_mem)) { + device_printf(dev, "Unable to allocate bus resource: PCI memory\n"); + return (ENXIO); + } + + /* Save off the PCI information */ + hw->vendor_id = pci_get_vendor(dev); + hw->device_id = pci_get_device(dev); + hw->revision_id = pci_read_config(dev, PCIR_REVID, 1); + hw->subsystem_vendor_id = + pci_read_config(dev, PCIR_SUBVEND_0, 2); + hw->subsystem_device_id = + pci_read_config(dev, PCIR_SUBDEV_0, 2); + + hw->bus.device = pci_get_slot(dev); + hw->bus.func = pci_get_function(dev); + + /* Save off register access information */ + sc->osdep.mem_bus_space_tag = + rman_get_bustag(sc->pci_mem); + sc->osdep.mem_bus_space_handle = + rman_get_bushandle(sc->pci_mem); + sc->osdep.mem_bus_space_size = rman_get_size(sc->pci_mem); + sc->osdep.flush_reg = I40E_VFGEN_RSTAT; + sc->osdep.dev = dev; + + sc->hw.hw_addr = (u8 *) &sc->osdep.mem_bus_space_handle; + sc->hw.back = &sc->osdep; + + return (0); +} static int ixlv_if_attach_pre(if_ctx_t ctx) { device_t dev; - struct ixlv_sc *sc; - struct i40e_hw *hw; - struct ixl_vsi *vsi; + struct ixlv_sc *sc; + struct i40e_hw *hw; + struct ixl_vsi *vsi; if_softc_ctx_t scctx; int error = 0; - INIT_DBG_DEV(dev, "begin"); - dev = iflib_get_dev(ctx); sc = iflib_get_softc(ctx); - hw = &sc->hw; - /* - ** Note this assumes we have a single embedded VSI, - ** this could be enhanced later to allocate multiple - */ + vsi = &sc->vsi; - vsi->dev = dev; vsi->back = sc; + sc->dev = dev; + hw = &sc->hw; + + vsi->dev = dev; vsi->hw = &sc->hw; - // vsi->id = 0; vsi->num_vlans = 0; vsi->ctx = ctx; vsi->media = iflib_get_media(ctx); vsi->shared = scctx = iflib_get_softc_ctx(ctx); - sc->dev = dev; - /* Initialize hw struct */ - ixlv_init_hw(sc); - /* - * These are the same across all current ixl models - */ - vsi->shared->isc_tx_nsegments = IXL_MAX_TX_SEGS; - vsi->shared->isc_msix_bar = PCIR_BAR(IXL_MSIX_BAR); - vsi->shared->isc_tx_tso_segments_max = IXL_MAX_TSO_SEGS; - vsi->shared->isc_tx_tso_size_max = IXL_TSO_SIZE; - vsi->shared->isc_tx_tso_segsize_max = PAGE_SIZE; - - /* Save this tunable */ - vsi->enable_head_writeback = ixlv_enable_head_writeback; - - scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] - * sizeof(struct i40e_tx_desc) + sizeof(u32), DBA_ALIGN); - scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] - * sizeof(union i40e_32byte_rx_desc), DBA_ALIGN); - /* XXX: No idea what this does */ - /* TODO: This value may depend on resources received */ - scctx->isc_max_txqsets = scctx->isc_max_rxqsets = 16; + ixlv_save_tunables(sc); /* Do PCI setup - map BAR0, etc */ if (ixlv_allocate_pci_resources(sc)) { @@ -355,9 +359,12 @@ ixlv_if_attach_pre(if_ctx_t ctx) goto err_early; } - INIT_DBG_DEV(dev, "Allocated PCI resources and MSIX vectors"); + ixlv_dbg_init(sc, "Allocated PCI resources and MSIX vectors\n"); - /* XXX: This is called by init_shared_code in the PF driver */ + /* + * XXX: This is called by init_shared_code in the PF driver, + * but the rest of that function does not support VFs. + */ error = i40e_set_mac_type(hw); if (error) { device_printf(dev, "%s: set_mac_type failed: %d\n", @@ -372,7 +379,7 @@ ixlv_if_attach_pre(if_ctx_t ctx) goto err_pci_res; } - INIT_DBG_DEV(dev, "VF Device is ready for configuration"); + ixlv_dbg_init(sc, "VF Device is ready for configuration\n"); /* Sets up Admin Queue */ error = ixlv_setup_vc(sc); @@ -382,7 +389,7 @@ ixlv_if_attach_pre(if_ctx_t ctx) goto err_pci_res; } - INIT_DBG_DEV(dev, "PF API version verified"); + ixlv_dbg_init(sc, "PF API version verified\n"); /* Need API version before sending reset message */ error = ixlv_reset(sc); @@ -391,7 +398,7 @@ ixlv_if_attach_pre(if_ctx_t ctx) goto err_aq; } - INIT_DBG_DEV(dev, "VF reset complete"); + ixlv_dbg_init(sc, "VF reset complete\n"); /* Ask for VF config from PF */ error = ixlv_vf_config(sc); @@ -407,13 +414,12 @@ ixlv_if_attach_pre(if_ctx_t ctx) sc->vf_res->max_vectors, sc->vf_res->rss_key_size, sc->vf_res->rss_lut_size); -#ifdef IXL_DEBUG - device_printf(dev, "Offload flags: 0x%b\n", - sc->vf_res->vf_offload_flags, IXLV_PRINTF_VF_OFFLOAD_FLAGS); -#endif + ixlv_dbg_info(sc, "Capabilities=%b\n", + sc->vf_res->vf_cap_flags, IXLV_PRINTF_VF_OFFLOAD_FLAGS); /* got VF config message back from PF, now we can parse it */ for (int i = 0; i < sc->vf_res->num_vsis; i++) { + /* XXX: We only use the first VSI we find */ if (sc->vf_res->vsi_res[i].vsi_type == I40E_VSI_SRIOV) sc->vsi_res = &sc->vf_res->vsi_res[i]; } @@ -424,7 +430,7 @@ ixlv_if_attach_pre(if_ctx_t ctx) } vsi->id = sc->vsi_res->vsi_id; - INIT_DBG_DEV(dev, "Resource Acquisition complete"); + ixlv_dbg_init(sc, "Resource Acquisition complete\n"); /* If no mac address was assigned just make a random one */ if (!ixlv_check_ether_addr(hw->mac.addr)) { @@ -437,30 +443,41 @@ ixlv_if_attach_pre(if_ctx_t ctx) bcopy(hw->mac.addr, hw->mac.perm_addr, ETHER_ADDR_LEN); iflib_set_mac(ctx, hw->mac.addr); - // TODO: Is this still safe to call? - // ixl_vsi_setup_rings_size(vsi, ixlv_tx_ring_size, ixlv_rx_ring_size); - /* Allocate filter lists */ ixlv_init_filters(sc); /* Fill out more iflib parameters */ - scctx->isc_txrx = &ixl_txrx; - // TODO: Probably needs changing - vsi->shared->isc_rss_table_size = sc->hw.func_caps.rss_table_size; + scctx->isc_ntxqsets_max = scctx->isc_nrxqsets_max = + sc->vsi_res->num_queue_pairs; + if (vsi->enable_head_writeback) { + scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] + * sizeof(struct i40e_tx_desc) + sizeof(u32), DBA_ALIGN); + scctx->isc_txrx = &ixl_txrx_hwb; + } else { + scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] + * sizeof(struct i40e_tx_desc), DBA_ALIGN); + scctx->isc_txrx = &ixl_txrx_dwb; + } + scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] + * sizeof(union i40e_32byte_rx_desc), DBA_ALIGN); + scctx->isc_msix_bar = PCIR_BAR(IXL_MSIX_BAR); + scctx->isc_tx_nsegments = IXL_MAX_TX_SEGS; + scctx->isc_tx_tso_segments_max = IXL_MAX_TSO_SEGS; + scctx->isc_tx_tso_size_max = IXL_TSO_SIZE; + scctx->isc_tx_tso_segsize_max = IXL_MAX_DMA_SEG_SIZE; + scctx->isc_rss_table_size = IXL_RSS_VSI_LUT_SIZE; scctx->isc_tx_csum_flags = CSUM_OFFLOAD; scctx->isc_capabilities = scctx->isc_capenable = IXL_CAPS; - INIT_DBG_DEV(dev, "end"); return (0); + err_res_buf: - free(sc->vf_res, M_DEVBUF); + free(sc->vf_res, M_IXLV); err_aq: i40e_shutdown_adminq(hw); err_pci_res: ixlv_free_pci_resources(sc); err_early: - ixlv_free_filters(sc); - INIT_DBG_DEV(dev, "end: error %d", error); return (error); } @@ -476,57 +493,52 @@ ixlv_if_attach_post(if_ctx_t ctx) INIT_DBG_DEV(dev, "begin"); dev = iflib_get_dev(ctx); - vsi = iflib_get_softc(ctx); + sc = iflib_get_softc(ctx); + vsi = &sc->vsi; vsi->ifp = iflib_get_ifp(ctx); - sc = (struct ixlv_sc *)vsi->back; hw = &sc->hw; + /* Save off determined number of queues for interface */ + vsi->num_rx_queues = vsi->shared->isc_nrxqsets; + vsi->num_tx_queues = vsi->shared->isc_ntxqsets; + /* Setup the stack interface */ - if (ixlv_setup_interface(dev, sc) != 0) { - device_printf(dev, "%s: setup interface failed!\n", - __func__); - error = EIO; - goto out; - } + ixlv_setup_interface(dev, sc); INIT_DBG_DEV(dev, "Interface setup complete"); /* Initialize statistics & add sysctls */ bzero(&sc->vsi.eth_stats, sizeof(struct i40e_eth_stats)); - ixlv_add_sysctls(sc); + ixlv_add_device_sysctls(sc); + + sc->init_state = IXLV_INIT_READY; + atomic_store_rel_32(&sc->queues_enabled, 0); - /* We want AQ enabled early */ + /* We want AQ enabled early for init */ ixlv_enable_adminq_irq(hw); + INIT_DBG_DEV(dev, "end"); + return (error); -// TODO: Check if any failures can happen above -#if 0 -out: - free(sc->vf_res, M_DEVBUF); - i40e_shutdown_adminq(hw); - ixlv_free_pci_resources(sc); - ixlv_free_filters(sc); - INIT_DBG_DEV(dev, "end: error %d", error); - return (error); -#endif } +/** + * XXX: iflib always ignores the return value of detach() + * -> This means that this isn't allowed to fail + */ static int ixlv_if_detach(if_ctx_t ctx) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - struct ixlv_sc *sc = vsi->back; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; struct i40e_hw *hw = &sc->hw; - device_t dev = sc->dev; + device_t dev = sc->dev; enum i40e_status_code status; INIT_DBG_DEV(dev, "begin"); /* Remove all the media and link information */ - ifmedia_removeall(&sc->media); - - /* Drain VC mgr */ - callout_drain(&sc->vc_mgr.callout); + ifmedia_removeall(vsi->media); ixlv_disable_adminq_irq(hw); status = i40e_shutdown_adminq(&sc->hw); @@ -536,7 +548,7 @@ ixlv_if_detach(if_ctx_t ctx) i40e_stat_str(hw, status)); } - free(sc->vf_res, M_DEVBUF); + free(sc->vf_res, M_IXLV); ixlv_free_pci_resources(sc); ixlv_free_filters(sc); @@ -544,411 +556,157 @@ ixlv_if_detach(if_ctx_t ctx) return (0); } -/* TODO: Do shutdown-specific stuff here */ static int ixlv_if_shutdown(if_ctx_t ctx) { - int error = 0; - - INIT_DBG_DEV(dev, "begin"); - - /* TODO: Call ixl_if_stop()? */ - - return (error); + return (0); } -/* TODO: What is a VF supposed to do in suspend/resume? */ static int ixlv_if_suspend(if_ctx_t ctx) { - int error = 0; - - INIT_DBG_DEV(dev, "begin"); - - /* TODO: Call ixl_if_stop()? */ - - return (error); + return (0); } static int ixlv_if_resume(if_ctx_t ctx) { - struct ifnet *ifp = iflib_get_ifp(ctx); - - INIT_DBG_DEV(dev, "begin"); - - /* Read & clear wake-up registers */ - - /* Required after D3->D0 transition */ - if (ifp->if_flags & IFF_UP) - ixlv_if_init(ctx); - return (0); } -#if 0 static int -ixlv_ioctl(struct ifnet *ifp, u_long command, caddr_t data) -{ - struct ixl_vsi *vsi = ifp->if_softc; - struct ixlv_sc *sc = vsi->back; - struct ifreq *ifr = (struct ifreq *)data; -#if defined(INET) || defined(INET6) - struct ifaddr *ifa = (struct ifaddr *)data; - bool avoid_reset = FALSE; -#endif - int error = 0; - - - switch (command) { - - case SIOCSIFADDR: -#ifdef INET - if (ifa->ifa_addr->sa_family == AF_INET) - avoid_reset = TRUE; -#endif -#ifdef INET6 - if (ifa->ifa_addr->sa_family == AF_INET6) - avoid_reset = TRUE; -#endif -#if defined(INET) || defined(INET6) - /* - ** Calling init results in link renegotiation, - ** so we avoid doing it when possible. - */ - if (avoid_reset) { - ifp->if_flags |= IFF_UP; - if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) - ixlv_init(vsi); -#ifdef INET - if (!(ifp->if_flags & IFF_NOARP)) - arp_ifinit(ifp, ifa); -#endif - } else - error = ether_ioctl(ifp, command, data); - break; -#endif - case SIOCSIFMTU: - IOCTL_DBG_IF2(ifp, "SIOCSIFMTU (Set Interface MTU)"); - mtx_lock(&sc->mtx); - if (ifr->ifr_mtu > IXL_MAX_FRAME - - ETHER_HDR_LEN - ETHER_CRC_LEN - ETHER_VLAN_ENCAP_LEN) { - error = EINVAL; - IOCTL_DBG_IF(ifp, "mtu too large"); - } else { - IOCTL_DBG_IF2(ifp, "mtu: %lu -> %d", (u_long)ifp->if_mtu, ifr->ifr_mtu); - // ERJ: Interestingly enough, these types don't match - ifp->if_mtu = (u_long)ifr->ifr_mtu; - vsi->max_frame_size = - ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN - + ETHER_VLAN_ENCAP_LEN; - if (ifp->if_drv_flags & IFF_DRV_RUNNING) - ixlv_init_locked(sc); - } - mtx_unlock(&sc->mtx); - break; - case SIOCSIFFLAGS: - IOCTL_DBG_IF2(ifp, "SIOCSIFFLAGS (Set Interface Flags)"); - mtx_lock(&sc->mtx); - if (ifp->if_flags & IFF_UP) { - if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) - ixlv_init_locked(sc); - } else - if (ifp->if_drv_flags & IFF_DRV_RUNNING) - ixlv_stop(sc); - sc->if_flags = ifp->if_flags; - mtx_unlock(&sc->mtx); - break; - case SIOCADDMULTI: - IOCTL_DBG_IF2(ifp, "SIOCADDMULTI"); - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - mtx_lock(&sc->mtx); - ixlv_disable_intr(vsi); - ixlv_add_multi(vsi); - ixlv_enable_intr(vsi); - mtx_unlock(&sc->mtx); - } - break; - case SIOCDELMULTI: - IOCTL_DBG_IF2(ifp, "SIOCDELMULTI"); - if (sc->init_state == IXLV_RUNNING) { - mtx_lock(&sc->mtx); - ixlv_disable_intr(vsi); - ixlv_del_multi(vsi); - ixlv_enable_intr(vsi); - mtx_unlock(&sc->mtx); - } - break; - case SIOCSIFMEDIA: - case SIOCGIFMEDIA: - IOCTL_DBG_IF2(ifp, "SIOCxIFMEDIA (Get/Set Interface Media)"); - error = ifmedia_ioctl(ifp, ifr, &sc->media, command); - break; - case SIOCSIFCAP: - { - int mask = ifr->ifr_reqcap ^ ifp->if_capenable; - IOCTL_DBG_IF2(ifp, "SIOCSIFCAP (Set Capabilities)"); - - ixlv_cap_txcsum_tso(vsi, ifp, mask); - - if (mask & IFCAP_RXCSUM) - ifp->if_capenable ^= IFCAP_RXCSUM; - if (mask & IFCAP_RXCSUM_IPV6) - ifp->if_capenable ^= IFCAP_RXCSUM_IPV6; - if (mask & IFCAP_LRO) - ifp->if_capenable ^= IFCAP_LRO; - if (mask & IFCAP_VLAN_HWTAGGING) - ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; - if (mask & IFCAP_VLAN_HWFILTER) - ifp->if_capenable ^= IFCAP_VLAN_HWFILTER; - if (mask & IFCAP_VLAN_HWTSO) - ifp->if_capenable ^= IFCAP_VLAN_HWTSO; - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - ixlv_init(vsi); - } - VLAN_CAPABILITIES(ifp); +ixlv_send_vc_msg_sleep(struct ixlv_sc *sc, u32 op) +{ + int error = 0; + if_ctx_t ctx = sc->vsi.ctx; - break; + error = ixl_vc_send_cmd(sc, op); + if (error != 0) { + ixlv_dbg_vc(sc, "Error sending %b: %d\n", op, IXLV_FLAGS, error); + return (error); } - default: - IOCTL_DBG_IF2(ifp, "UNKNOWN (0x%X)", (int)command); - error = ether_ioctl(ifp, command, data); - break; + /* Don't wait for a response if the device is being detached. */ + if (!iflib_in_detach(ctx)) { + ixlv_dbg_vc(sc, "Sleeping for op %b\n", op, IXLV_FLAGS); + error = sx_sleep(ixl_vc_get_op_chan(sc, op), + iflib_ctx_lock_get(ctx), PRI_MAX, "ixlvc", IXLV_AQ_TIMEOUT); + + if (error == EWOULDBLOCK) + device_printf(sc->dev, "%b timed out\n", op, IXLV_FLAGS); } return (error); } -#endif -/* -** To do a reinit on the VF is unfortunately more complicated -** than a physical device, we must have the PF more or less -** completely recreate our memory, so many things that were -** done only once at attach in traditional drivers now must be -** redone at each reinitialization. This function does that -** 'prelude' so we can then call the normal locked init code. -*/ -int -ixlv_reinit_locked(struct ixlv_sc *sc) +static int +ixlv_send_vc_msg(struct ixlv_sc *sc, u32 op) { - struct i40e_hw *hw = &sc->hw; - struct ixl_vsi *vsi = &sc->vsi; - struct ifnet *ifp = vsi->ifp; - struct ixlv_mac_filter *mf, *mf_temp; - struct ixlv_vlan_filter *vf; - int error = 0; - - INIT_DBG_IF(ifp, "begin"); - - if (ifp->if_drv_flags & IFF_DRV_RUNNING) - ixlv_stop(sc); - - error = ixlv_reset(sc); - - INIT_DBG_IF(ifp, "VF was reset"); - - /* set the state in case we went thru RESET */ - sc->init_state = IXLV_RUNNING; - - /* - ** Resetting the VF drops all filters from hardware; - ** we need to mark them to be re-added in init. - */ - SLIST_FOREACH_SAFE(mf, sc->mac_filters, next, mf_temp) { - if (mf->flags & IXL_FILTER_DEL) { - SLIST_REMOVE(sc->mac_filters, mf, - ixlv_mac_filter, next); - free(mf, M_DEVBUF); - } else - mf->flags |= IXL_FILTER_ADD; - } - if (vsi->num_vlans != 0) - SLIST_FOREACH(vf, sc->vlan_filters, next) - vf->flags = IXL_FILTER_ADD; - else { /* clean any stale filters */ - while (!SLIST_EMPTY(sc->vlan_filters)) { - vf = SLIST_FIRST(sc->vlan_filters); - SLIST_REMOVE_HEAD(sc->vlan_filters, next); - free(vf, M_DEVBUF); - } - } + int error = 0; - ixlv_enable_adminq_irq(hw); - ixl_vc_flush(&sc->vc_mgr); + error = ixl_vc_send_cmd(sc, op); + if (error != 0) + ixlv_dbg_vc(sc, "Error sending %b: %d\n", op, IXLV_FLAGS, error); - INIT_DBG_IF(ifp, "end"); return (error); } static void -ixl_init_cmd_complete(struct ixl_vc_cmd *cmd, void *arg, - enum i40e_status_code code) +ixlv_init_queues(struct ixl_vsi *vsi) { - struct ixlv_sc *sc; + if_softc_ctx_t scctx = vsi->shared; + struct ixl_tx_queue *tx_que = vsi->tx_queues; + struct ixl_rx_queue *rx_que = vsi->rx_queues; + struct rx_ring *rxr; - sc = arg; + for (int i = 0; i < vsi->num_tx_queues; i++, tx_que++) + ixl_init_tx_ring(vsi, tx_que); - /* - * Ignore "Adapter Stopped" message as that happens if an ifconfig down - * happens while a command is in progress, so we don't print an error - * in that case. - */ - if (code != I40E_SUCCESS && code != I40E_ERR_ADAPTER_STOPPED) { - if_printf(sc->vsi.ifp, - "Error %s waiting for PF to complete operation %d\n", - i40e_stat_str(&sc->hw, code), cmd->request); + for (int i = 0; i < vsi->num_rx_queues; i++, rx_que++) { + rxr = &rx_que->rxr; + + if (scctx->isc_max_frame_size <= MCLBYTES) + rxr->mbuf_sz = MCLBYTES; + else + rxr->mbuf_sz = MJUMPAGESIZE; + + wr32(vsi->hw, rxr->tail, 0); } } void ixlv_if_init(if_ctx_t ctx) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - if_softc_ctx_t scctx = vsi->shared; - struct ixlv_sc *sc = vsi->back; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; struct i40e_hw *hw = &sc->hw; struct ifnet *ifp = iflib_get_ifp(ctx); - struct ixl_tx_queue *tx_que = vsi->tx_queues; - struct ixl_rx_queue *rx_que = vsi->rx_queues; - + u8 tmpaddr[ETHER_ADDR_LEN]; int error = 0; INIT_DBG_IF(ifp, "begin"); - IXLV_CORE_LOCK_ASSERT(sc); + MPASS(sx_xlocked(iflib_ctx_lock_get(ctx))); - /* Do a reinit first if an init has already been done */ - if ((sc->init_state == IXLV_RUNNING) || - (sc->init_state == IXLV_RESET_REQUIRED) || - (sc->init_state == IXLV_RESET_PENDING)) - error = ixlv_reinit_locked(sc); - /* Don't bother with init if we failed reinit */ - if (error) - goto init_done; + error = ixlv_reset_complete(hw); + if (error) { + device_printf(sc->dev, "%s: VF reset failed\n", + __func__); + } - /* Remove existing MAC filter if new MAC addr is set */ - if (bcmp(IF_LLADDR(ifp), hw->mac.addr, ETHER_ADDR_LEN) != 0) { - error = ixlv_del_mac_filter(sc, hw->mac.addr); - if (error == 0) - ixl_vc_enqueue(&sc->vc_mgr, &sc->del_mac_cmd, - IXLV_FLAG_AQ_DEL_MAC_FILTER, ixl_init_cmd_complete, - sc); + if (!i40e_check_asq_alive(hw)) { + ixlv_dbg_info(sc, "ASQ is not alive, re-initializing AQ\n"); + pci_enable_busmaster(sc->dev); + i40e_shutdown_adminq(hw); + i40e_init_adminq(hw); } - /* Check for an LAA mac address... */ - bcopy(IF_LLADDR(ifp), hw->mac.addr, ETHER_ADDR_LEN); + /* Make sure queues are disabled */ + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_DISABLE_QUEUES); - /* Add mac filter for this VF to PF */ - if (i40e_validate_mac_addr(hw->mac.addr) == I40E_SUCCESS) { - error = ixlv_add_mac_filter(sc, hw->mac.addr, 0); - if (!error || error == EEXIST) - ixl_vc_enqueue(&sc->vc_mgr, &sc->add_mac_cmd, - IXLV_FLAG_AQ_ADD_MAC_FILTER, ixl_init_cmd_complete, - sc); + bcopy(IF_LLADDR(ifp), tmpaddr, ETHER_ADDR_LEN); + if (!cmp_etheraddr(hw->mac.addr, tmpaddr) && + (i40e_validate_mac_addr(tmpaddr) == I40E_SUCCESS)) { + error = ixlv_del_mac_filter(sc, hw->mac.addr); + if (error == 0) + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_DEL_MAC_FILTER); + + bcopy(tmpaddr, hw->mac.addr, ETH_ALEN); } - /* Setup vlan's if needed */ - ixlv_setup_vlan_filters(sc); + error = ixlv_add_mac_filter(sc, hw->mac.addr, 0); + if (!error || error == EEXIST) + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_ADD_MAC_FILTER); + iflib_set_mac(ctx, hw->mac.addr); - // TODO: Functionize /* Prepare the queues for operation */ - for (int i = 0; i < vsi->num_tx_queues; i++, tx_que++) { - // TODO: Necessary? Correct? - ixl_init_tx_ring(vsi, tx_que); - } - for (int i = 0; i < vsi->num_rx_queues; i++, rx_que++) { - struct rx_ring *rxr = &rx_que->rxr; - - if (scctx->isc_max_frame_size <= MCLBYTES) - rxr->mbuf_sz = MCLBYTES; - else - rxr->mbuf_sz = MJUMPAGESIZE; - } + ixlv_init_queues(vsi); /* Set initial ITR values */ ixlv_configure_itr(sc); - /* Configure queues */ - ixl_vc_enqueue(&sc->vc_mgr, &sc->config_queues_cmd, - IXLV_FLAG_AQ_CONFIGURE_QUEUES, ixl_init_cmd_complete, sc); + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_CONFIGURE_QUEUES); /* Set up RSS */ ixlv_config_rss(sc); /* Map vectors */ - ixl_vc_enqueue(&sc->vc_mgr, &sc->map_vectors_cmd, - IXLV_FLAG_AQ_MAP_VECTORS, ixl_init_cmd_complete, sc); - - /* Enable queues */ - ixl_vc_enqueue(&sc->vc_mgr, &sc->enable_queues_cmd, - IXLV_FLAG_AQ_ENABLE_QUEUES, ixl_init_cmd_complete, sc); - - sc->init_state = IXLV_RUNNING; + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_MAP_VECTORS); -init_done: - INIT_DBG_IF(ifp, "end"); - return; -} - -#if 0 -void -ixlv_init(void *arg) -{ - struct ixl_vsi *vsi = (struct ixl_vsi *)arg; - struct ixlv_sc *sc = vsi->back; - int retries = 0; - - /* Prevent init from running again while waiting for AQ calls - * made in init_locked() to complete. */ - mtx_lock(&sc->mtx); - if (sc->init_in_progress) { - mtx_unlock(&sc->mtx); - return; - } else - sc->init_in_progress = true; - - ixlv_init_locked(sc); - mtx_unlock(&sc->mtx); - - /* Wait for init_locked to finish */ - while (!(vsi->ifp->if_drv_flags & IFF_DRV_RUNNING) - && ++retries < IXLV_MAX_INIT_WAIT) { - i40e_msec_pause(25); - } - if (retries >= IXLV_MAX_INIT_WAIT) { - if_printf(vsi->ifp, - "Init failed to complete in allotted time!\n"); - } + /* Init SW TX ring indices */ + if (vsi->enable_head_writeback) + ixl_init_tx_cidx(vsi); + else + ixl_init_tx_rsqs(vsi); - mtx_lock(&sc->mtx); - sc->init_in_progress = false; - mtx_unlock(&sc->mtx); -} + /* Configure promiscuous mode */ + ixlv_if_promisc_set(ctx, if_getflags(ifp)); -/* - * ixlv_attach() helper function; gathers information about - * the (virtual) hardware for use elsewhere in the driver. - */ -static void -ixlv_init_hw(struct ixlv_sc *sc) -{ - struct i40e_hw *hw = &sc->hw; - device_t dev = sc->dev; - - /* Save off the information about this board */ - hw->vendor_id = pci_get_vendor(dev); - hw->device_id = pci_get_device(dev); - hw->revision_id = pci_read_config(dev, PCIR_REVID, 1); - hw->subsystem_vendor_id = - pci_read_config(dev, PCIR_SUBVEND_0, 2); - hw->subsystem_device_id = - pci_read_config(dev, PCIR_SUBDEV_0, 2); + /* Enable queues */ + ixlv_send_vc_msg_sleep(sc, IXLV_FLAG_AQ_ENABLE_QUEUES); - hw->bus.device = pci_get_slot(dev); - hw->bus.func = pci_get_function(dev); + sc->init_state = IXLV_RUNNING; } -#endif /* * ixlv_attach() helper function; initalizes the admin queue @@ -980,7 +738,7 @@ ixlv_setup_vc(struct ixlv_sc *sc) continue; } - INIT_DBG_DEV(dev, "Initialized Admin Queue; starting" + ixlv_dbg_init(sc, "Initialized Admin Queue; starting" " send_api_ver attempt %d", i+1); retry_send: @@ -1009,7 +767,7 @@ retry_send: if (asq_retries > IXLV_AQ_MAX_ERR) continue; - INIT_DBG_DEV(dev, "Sent API version message to PF"); + ixlv_dbg_init(sc, "Sent API version message to PF"); /* Verify that the VF accepts the PF's API version */ error = ixlv_verify_api_ver(sc); @@ -1076,13 +834,13 @@ retry_config: i40e_msec_pause(10); } - INIT_DBG_DEV(dev, "Sent VF config message to PF, attempt %d", + ixlv_dbg_init(sc, "Sent VF config message to PF, attempt %d\n", retried + 1); if (!sc->vf_res) { bufsz = sizeof(struct virtchnl_vf_resource) + (I40E_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource)); - sc->vf_res = malloc(bufsz, M_DEVBUF, M_NOWAIT); + sc->vf_res = malloc(bufsz, M_IXLV, M_NOWAIT); if (!sc->vf_res) { device_printf(dev, "%s: Unable to allocate memory for VF configuration" @@ -1113,7 +871,7 @@ retry_config: goto done; fail: - free(sc->vf_res, M_DEVBUF); + free(sc->vf_res, M_IXLV); done: return (ret_error); } @@ -1121,53 +879,65 @@ done: static int ixlv_if_msix_intr_assign(if_ctx_t ctx, int msix) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - struct ixlv_sc *sc = vsi->back; - struct ixl_rx_queue *que = vsi->rx_queues; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + struct ixl_rx_queue *rx_que = vsi->rx_queues; struct ixl_tx_queue *tx_que = vsi->tx_queues; int err, i, rid, vector = 0; char buf[16]; + MPASS(vsi->shared->isc_nrxqsets > 0); + MPASS(vsi->shared->isc_ntxqsets > 0); + /* Admin Que is vector 0*/ rid = vector + 1; - err = iflib_irq_alloc_generic(ctx, &vsi->irq, rid, IFLIB_INTR_ADMIN, - ixlv_msix_adminq, sc, 0, "aq"); + ixlv_msix_adminq, sc, 0, "aq"); if (err) { iflib_irq_free(ctx, &vsi->irq); - device_printf(iflib_get_dev(ctx), "Failed to register Admin que handler"); + device_printf(iflib_get_dev(ctx), + "Failed to register Admin Que handler"); return (err); } - sc->admvec = vector; - ++vector; /* Now set up the stations */ - for (i = 0; i < vsi->num_rx_queues; i++, vector++, que++) { + for (i = 0, vector = 1; i < vsi->shared->isc_nrxqsets; i++, vector++, rx_que++) { rid = vector + 1; snprintf(buf, sizeof(buf), "rxq%d", i); - err = iflib_irq_alloc_generic(ctx, &que->que_irq, rid, IFLIB_INTR_RX, - ixlv_msix_que, que, que->rxr.me, buf); + err = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, + IFLIB_INTR_RX, ixlv_msix_que, rx_que, rx_que->rxr.me, buf); + /* XXX: Does the driver work as expected if there are fewer num_rx_queues than + * what's expected in the iflib context? */ if (err) { - device_printf(iflib_get_dev(ctx), "Failed to allocate q int %d err: %d", i, err); + device_printf(iflib_get_dev(ctx), + "Failed to allocate queue RX int vector %d, err: %d\n", i, err); vsi->num_rx_queues = i + 1; goto fail; } - que->msix = vector; + rx_que->msix = vector; } - for (i = 0, tx_que = vsi->tx_queues; i < vsi->num_tx_queues; i++, tx_que++) { + bzero(buf, sizeof(buf)); + + for (i = 0; i < vsi->shared->isc_ntxqsets; i++, tx_que++) { snprintf(buf, sizeof(buf), "txq%d", i); - rid = que->msix + 1; - iflib_softirq_alloc_generic(ctx, rid, IFLIB_INTR_TX, tx_que, tx_que->txr.me, buf); + iflib_softirq_alloc_generic(ctx, + &vsi->rx_queues[i % vsi->shared->isc_nrxqsets].que_irq, + IFLIB_INTR_TX, tx_que, tx_que->txr.me, buf); + + /* TODO: Maybe call a strategy function for this to figure out which + * interrupts to map Tx queues to. I don't know if there's an immediately + * better way than this other than a user-supplied map, though. */ + tx_que->msix = (i % vsi->shared->isc_nrxqsets) + 1; } return (0); fail: iflib_irq_free(ctx, &vsi->irq); - que = vsi->rx_queues; - for (int i = 0; i < vsi->num_rx_queues; i++, que++) - iflib_irq_free(ctx, &que->que_irq); + rx_que = vsi->rx_queues; + for (int i = 0; i < vsi->num_rx_queues; i++, rx_que++) + iflib_irq_free(ctx, &rx_que->que_irq); return (err); } @@ -1175,7 +945,8 @@ fail: static void ixlv_if_enable_intr(if_ctx_t ctx) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; ixlv_enable_intr(vsi); } @@ -1184,34 +955,48 @@ ixlv_if_enable_intr(if_ctx_t ctx) static void ixlv_if_disable_intr(if_ctx_t ctx) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; ixlv_disable_intr(vsi); } -/* Enable queue interrupt */ static int -ixlv_if_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid) +ixlv_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - struct i40e_hw *hw = vsi->hw; - struct ixl_rx_queue *que = &vsi->rx_queues[rxqid]; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + struct i40e_hw *hw = vsi->hw; + struct ixl_rx_queue *rx_que = &vsi->rx_queues[rxqid]; + + ixlv_enable_queue_irq(hw, rx_que->msix - 1); + return (0); +} - ixlv_enable_queue_irq(hw, que->rxr.me); +static int +ixlv_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid) +{ + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + struct i40e_hw *hw = vsi->hw; + struct ixl_tx_queue *tx_que = &vsi->tx_queues[txqid]; + ixlv_enable_queue_irq(hw, tx_que->msix - 1); return (0); } static int ixlv_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + if_softc_ctx_t scctx = vsi->shared; struct ixl_tx_queue *que; - int i; + int i, j, error = 0; - MPASS(vsi->num_tx_queues > 0); + MPASS(scctx->isc_ntxqsets > 0); MPASS(ntxqs == 1); - MPASS(vsi->num_tx_queues == ntxqsets); + MPASS(scctx->isc_ntxqsets == ntxqsets); /* Allocate queue structure memory */ if (!(vsi->tx_queues = @@ -1222,40 +1007,56 @@ ixlv_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntx for (i = 0, que = vsi->tx_queues; i < ntxqsets; i++, que++) { struct tx_ring *txr = &que->txr; + txr->me = i; que->vsi = vsi; + if (!vsi->enable_head_writeback) { + /* Allocate report status array */ + if (!(txr->tx_rsq = malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_IXLV, M_NOWAIT))) { + device_printf(iflib_get_dev(ctx), "failed to allocate tx_rsq memory\n"); + error = ENOMEM; + goto fail; + } + /* Init report status array */ + for (j = 0; j < scctx->isc_ntxd[0]; j++) + txr->tx_rsq[j] = QIDX_INVALID; + } /* get the virtual and physical address of the hardware queues */ txr->tail = I40E_QTX_TAIL1(txr->me); - txr->tx_base = (struct i40e_tx_desc *)vaddrs[i]; - txr->tx_paddr = paddrs[i]; + txr->tx_base = (struct i40e_tx_desc *)vaddrs[i * ntxqs]; + txr->tx_paddr = paddrs[i * ntxqs]; txr->que = que; } - - // TODO: Do a config_gtask_init for admin queue here? - // iflib_config_gtask_init(ctx, &adapter->mod_task, ixgbe_handle_mod, "mod_task"); - device_printf(iflib_get_dev(ctx), "%s: allocated for %d txqs\n", __func__, vsi->num_tx_queues); return (0); +fail: + ixlv_if_queues_free(ctx); + return (error); } static int ixlv_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; struct ixl_rx_queue *que; - int i; + int i, error = 0; - MPASS(vsi->num_rx_queues > 0); +#ifdef INVARIANTS + if_softc_ctx_t scctx = vsi->shared; + MPASS(scctx->isc_nrxqsets > 0); MPASS(nrxqs == 1); - MPASS(vsi->num_rx_queues == nrxqsets); + MPASS(scctx->isc_nrxqsets == nrxqsets); +#endif /* Allocate queue structure memory */ if (!(vsi->rx_queues = (struct ixl_rx_queue *) malloc(sizeof(struct ixl_rx_queue) * nrxqsets, M_IXLV, M_NOWAIT | M_ZERO))) { device_printf(iflib_get_dev(ctx), "Unable to allocate RX ring memory\n"); - return (ENOMEM); + error = ENOMEM; + goto fail; } for (i = 0, que = vsi->rx_queues; i < nrxqsets; i++, que++) { @@ -1266,19 +1067,35 @@ ixlv_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrx /* get the virtual and physical address of the hardware queues */ rxr->tail = I40E_QRX_TAIL1(rxr->me); - rxr->rx_base = (union i40e_rx_desc *)vaddrs[i]; - rxr->rx_paddr = paddrs[i]; + rxr->rx_base = (union i40e_rx_desc *)vaddrs[i * nrxqs]; + rxr->rx_paddr = paddrs[i * nrxqs]; rxr->que = que; } - device_printf(iflib_get_dev(ctx), "%s: allocated for %d rxqs\n", __func__, vsi->num_rx_queues); return (0); +fail: + ixlv_if_queues_free(ctx); + return (error); } static void ixlv_if_queues_free(if_ctx_t ctx) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + + if (!vsi->enable_head_writeback) { + struct ixl_tx_queue *que; + int i = 0; + + for (i = 0, que = vsi->tx_queues; i < vsi->shared->isc_ntxqsets; i++, que++) { + struct tx_ring *txr = &que->txr; + if (txr->tx_rsq != NULL) { + free(txr->tx_rsq, M_IXLV); + txr->tx_rsq = NULL; + } + } + } if (vsi->tx_queues != NULL) { free(vsi->tx_queues, M_IXLV); @@ -1290,274 +1107,237 @@ ixlv_if_queues_free(if_ctx_t ctx) } } -// TODO: Implement -static void -ixlv_if_update_admin_status(if_ctx_t ctx) +static int +ixlv_check_aq_errors(struct ixlv_sc *sc) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - //struct ixlv_sc *sc = vsi->back; - //struct i40e_hw *hw = &sc->hw; - //struct i40e_arq_event_info event; - //i40e_status ret; - //u32 loop = 0; - //u16 opcode - u16 result = 0; - //u64 baudrate; - - /* TODO: Split up - * - Update admin queue stuff - * - Update link status - * - Enqueue aq task - * - Re-enable admin intr - */ + struct i40e_hw *hw = &sc->hw; + device_t dev = sc->dev; + u32 reg, oldreg; + u8 aq_error = false; -/* TODO: Does VF reset need to be handled here? */ -#if 0 - if (pf->state & IXL_PF_STATE_EMPR_RESETTING) { - /* Flag cleared at end of this function */ - ixl_handle_empr_reset(pf); - return; + /* check for Admin queue errors */ + oldreg = reg = rd32(hw, hw->aq.arq.len); + if (reg & I40E_VF_ARQLEN1_ARQVFE_MASK) { + device_printf(dev, "ARQ VF Error detected\n"); + reg &= ~I40E_VF_ARQLEN1_ARQVFE_MASK; + aq_error = true; } -#endif + if (reg & I40E_VF_ARQLEN1_ARQOVFL_MASK) { + device_printf(dev, "ARQ Overflow Error detected\n"); + reg &= ~I40E_VF_ARQLEN1_ARQOVFL_MASK; + aq_error = true; + } + if (reg & I40E_VF_ARQLEN1_ARQCRIT_MASK) { + device_printf(dev, "ARQ Critical Error detected\n"); + reg &= ~I40E_VF_ARQLEN1_ARQCRIT_MASK; + aq_error = true; + } + if (oldreg != reg) + wr32(hw, hw->aq.arq.len, reg); -#if 0 - event.buf_len = IXL_AQ_BUF_SZ; - event.msg_buf = malloc(event.buf_len, - M_IXLV, M_NOWAIT | M_ZERO); - if (!event.msg_buf) { - device_printf(pf->dev, "%s: Unable to allocate memory for Admin" - " Queue event!\n", __func__); - return; + oldreg = reg = rd32(hw, hw->aq.asq.len); + if (reg & I40E_VF_ATQLEN1_ATQVFE_MASK) { + device_printf(dev, "ASQ VF Error detected\n"); + reg &= ~I40E_VF_ATQLEN1_ATQVFE_MASK; + aq_error = true; + } + if (reg & I40E_VF_ATQLEN1_ATQOVFL_MASK) { + device_printf(dev, "ASQ Overflow Error detected\n"); + reg &= ~I40E_VF_ATQLEN1_ATQOVFL_MASK; + aq_error = true; + } + if (reg & I40E_VF_ATQLEN1_ATQCRIT_MASK) { + device_printf(dev, "ASQ Critical Error detected\n"); + reg &= ~I40E_VF_ATQLEN1_ATQCRIT_MASK; + aq_error = true; + } + if (oldreg != reg) + wr32(hw, hw->aq.asq.len, reg); + + if (aq_error) { + device_printf(dev, "WARNING: Stopping VF!\n"); + /* + * A VF reset might not be enough to fix a problem here; + * a PF reset could be required. + */ + sc->init_state = IXLV_RESET_REQUIRED; + ixlv_stop(sc); + ixlv_request_reset(sc); } + return (aq_error ? EIO : 0); +} + +static enum i40e_status_code +ixlv_process_adminq(struct ixlv_sc *sc, u16 *pending) +{ + enum i40e_status_code status = I40E_SUCCESS; + struct i40e_arq_event_info event; + struct i40e_hw *hw = &sc->hw; + struct virtchnl_msg *v_msg; + int error = 0, loop = 0; + u32 reg; + + error = ixlv_check_aq_errors(sc); + if (error) + return (I40E_ERR_ADMIN_QUEUE_CRITICAL_ERROR); + + event.buf_len = IXL_AQ_BUF_SZ; + event.msg_buf = sc->aq_buffer; + bzero(event.msg_buf, IXL_AQ_BUF_SZ); + v_msg = (struct virtchnl_msg *)&event.desc; + /* clean and process any events */ do { - ret = i40e_clean_arq_element(hw, &event, &result); - if (ret) - break; - opcode = LE16_TO_CPU(event.desc.opcode); - ixl_dbg(pf, IXL_DBG_AQ, - "Admin Queue event: %#06x\n", opcode); - switch (opcode) { - case i40e_aqc_opc_get_link_status: - ixl_link_event(pf, &event); - break; - case i40e_aqc_opc_send_msg_to_pf: -#ifdef PCI_IOV - ixl_handle_vf_msg(pf, &event); -#endif - break; - case i40e_aqc_opc_event_lan_overflow: - break; - default: -#ifdef IXL_DEBUG - printf("AdminQ unknown event %x\n", opcode); -#endif + status = i40e_clean_arq_element(hw, &event, pending); + /* + * Also covers normal case when i40e_clean_arq_element() + * returns "I40E_ERR_ADMIN_QUEUE_NO_WORK" + */ + if (status) break; - } + ixlv_vc_completion(sc, v_msg->v_opcode, + v_msg->v_retval, event.msg_buf, event.msg_len); + bzero(event.msg_buf, IXL_AQ_BUF_SZ); + } while (*pending && (loop++ < IXL_ADM_LIMIT)); - } while (result && (loop++ < IXL_ADM_LIMIT)); + /* Re-enable admin queue interrupt cause */ + reg = rd32(hw, I40E_VFINT_ICR0_ENA1); + reg |= I40E_VFINT_ICR0_ENA1_ADMINQ_MASK; + wr32(hw, I40E_VFINT_ICR0_ENA1, reg); - free(event.msg_buf, M_IXLV); -#endif + return (status); +} -#if 0 - /* XXX: This updates the link status */ - if (pf->link_up) { - if (vsi->link_active == FALSE) { - vsi->link_active = TRUE; - baudrate = ixl_max_aq_speed_to_value(pf->link_speed); - iflib_link_state_change(ctx, LINK_STATE_UP, baudrate); - ixl_link_up_msg(pf); - // ixl_ping_all_vfs(adapter); - } - } else { /* Link down */ - if (vsi->link_active == TRUE) { - vsi->link_active = FALSE; - iflib_link_state_change(ctx, LINK_STATE_DOWN, 0); - // ixl_ping_all_vfs(adapter); - } - } -#endif +static void +ixlv_if_update_admin_status(if_ctx_t ctx) +{ + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct i40e_hw *hw = &sc->hw; + u16 pending; + + ixlv_process_adminq(sc, &pending); + ixlv_update_link_status(sc); /* - * If there are still messages to process, reschedule ourselves. - * Otherwise, re-enable our interrupt and go to sleep. + * If there are still messages to process, reschedule. + * Otherwise, re-enable the Admin Queue interrupt. */ - if (result > 0) + if (pending > 0) iflib_admin_intr_deferred(ctx); else - /* TODO: Link/adminq interrupt should be re-enabled in IFDI_LINK_INTR_ENABLE */ - ixlv_enable_intr(vsi); + ixlv_enable_adminq_irq(hw); +} + +static int +ixlv_mc_filter_apply(void *arg, struct ifmultiaddr *ifma, int count __unused) +{ + struct ixlv_sc *sc = arg; + int error = 0; + + if (ifma->ifma_addr->sa_family != AF_LINK) + return (0); + error = ixlv_add_mac_filter(sc, + (u8*)LLADDR((struct sockaddr_dl *) ifma->ifma_addr), + IXL_FILTER_MC); + + return (!error); } static void ixlv_if_multi_set(if_ctx_t ctx) { - // struct ixl_vsi *vsi = iflib_get_softc(ctx); - // struct i40e_hw *hw = vsi->hw; - // struct ixlv_sc *sc = vsi->back; - // int mcnt = 0, flags; + struct ixlv_sc *sc = iflib_get_softc(ctx); + int mcnt = 0; - IOCTL_DEBUGOUT("ixl_if_multi_set: begin"); + IOCTL_DEBUGOUT("ixlv_if_multi_set: begin"); - // TODO: Implement -#if 0 mcnt = if_multiaddr_count(iflib_get_ifp(ctx), MAX_MULTICAST_ADDR); - /* delete existing MC filters */ - ixlv_del_multi(vsi); - if (__predict_false(mcnt == MAX_MULTICAST_ADDR)) { - // Set promiscuous mode (multicast) - // TODO: This needs to get handled somehow -#if 0 - ixl_vc_enqueue(&sc->vc_mgr, &sc->add_vlan_cmd, - IXLV_FLAG_AQ_CONFIGURE_PROMISC, ixl_init_cmd_complete, sc); -#endif + /* Delete MC filters and enable mulitcast promisc instead */ + ixlv_init_multi(sc); + sc->promisc_flags |= FLAG_VF_MULTICAST_PROMISC; + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_CONFIGURE_PROMISC); return; } - /* (re-)install filters for all mcast addresses */ - mcnt = if_multi_apply(iflib_get_ifp(ctx), ixl_mc_filter_apply, vsi); - - if (mcnt > 0) { - flags = (IXL_FILTER_ADD | IXL_FILTER_USED | IXL_FILTER_MC); - ixlv_add_hw_filters(vsi, flags, mcnt); - } -#endif - IOCTL_DEBUGOUT("ixl_if_multi_set: end"); + /* If there aren't too many filters, delete existing MC filters */ + ixlv_init_multi(sc); + + /* And (re-)install filters for all mcast addresses */ + mcnt = if_multi_apply(iflib_get_ifp(ctx), ixlv_mc_filter_apply, sc); + + if (mcnt > 0) + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_ADD_MAC_FILTER); +} + +static int +ixlv_if_mtu_set(if_ctx_t ctx, uint32_t mtu) +{ + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + + IOCTL_DEBUGOUT("ioctl: SIOCSIFMTU (Set Interface MTU)"); + if (mtu > IXL_MAX_FRAME - ETHER_HDR_LEN - ETHER_CRC_LEN - + ETHER_VLAN_ENCAP_LEN) + return (EINVAL); + + vsi->shared->isc_max_frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + + ETHER_VLAN_ENCAP_LEN; + + return (0); } static void ixlv_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - struct ixlv_sc *sc = (struct ixlv_sc *)vsi->back; - struct i40e_hw *hw = &sc->hw; +#ifdef IXL_DEBUG + struct ifnet *ifp = iflib_get_ifp(ctx); +#endif + struct ixlv_sc *sc = iflib_get_softc(ctx); - INIT_DEBUGOUT("ixl_media_status: begin"); + INIT_DBG_IF(ifp, "begin"); - hw->phy.get_link_info = TRUE; - i40e_get_link_status(hw, &sc->link_up); + ixlv_update_link_status(sc); ifmr->ifm_status = IFM_AVALID; ifmr->ifm_active = IFM_ETHER; - if (!sc->link_up) { + if (!sc->link_up) return; - } ifmr->ifm_status |= IFM_ACTIVE; /* Hardware is always full-duplex */ ifmr->ifm_active |= IFM_FDX; - // TODO: Check another variable to get link speed -#if 0 - switch (hw->phy.link_info.phy_type) { - /* 100 M */ - case I40E_PHY_TYPE_100BASE_TX: - ifmr->ifm_active |= IFM_100_TX; - break; - /* 1 G */ - case I40E_PHY_TYPE_1000BASE_T: - ifmr->ifm_active |= IFM_1000_T; - break; - case I40E_PHY_TYPE_1000BASE_SX: - ifmr->ifm_active |= IFM_1000_SX; - break; - case I40E_PHY_TYPE_1000BASE_LX: - ifmr->ifm_active |= IFM_1000_LX; - break; - case I40E_PHY_TYPE_1000BASE_T_OPTICAL: - ifmr->ifm_active |= IFM_OTHER; - break; - /* 10 G */ - case I40E_PHY_TYPE_10GBASE_SFPP_CU: - ifmr->ifm_active |= IFM_10G_TWINAX; - break; - case I40E_PHY_TYPE_10GBASE_SR: - ifmr->ifm_active |= IFM_10G_SR; - break; - case I40E_PHY_TYPE_10GBASE_LR: - ifmr->ifm_active |= IFM_10G_LR; - break; - case I40E_PHY_TYPE_10GBASE_T: - ifmr->ifm_active |= IFM_10G_T; - break; - case I40E_PHY_TYPE_XAUI: - case I40E_PHY_TYPE_XFI: - case I40E_PHY_TYPE_10GBASE_AOC: - ifmr->ifm_active |= IFM_OTHER; - break; - /* 25 G */ - case I40E_PHY_TYPE_25GBASE_KR: - ifmr->ifm_active |= IFM_25G_KR; - break; - case I40E_PHY_TYPE_25GBASE_CR: - ifmr->ifm_active |= IFM_25G_CR; - break; - case I40E_PHY_TYPE_25GBASE_SR: - ifmr->ifm_active |= IFM_25G_SR; - break; - case I40E_PHY_TYPE_25GBASE_LR: - ifmr->ifm_active |= IFM_UNKNOWN; - break; - /* 40 G */ - case I40E_PHY_TYPE_40GBASE_CR4: - case I40E_PHY_TYPE_40GBASE_CR4_CU: - ifmr->ifm_active |= IFM_40G_CR4; - break; - case I40E_PHY_TYPE_40GBASE_SR4: - ifmr->ifm_active |= IFM_40G_SR4; - break; - case I40E_PHY_TYPE_40GBASE_LR4: - ifmr->ifm_active |= IFM_40G_LR4; - break; - case I40E_PHY_TYPE_XLAUI: - ifmr->ifm_active |= IFM_OTHER; - break; - case I40E_PHY_TYPE_1000BASE_KX: - ifmr->ifm_active |= IFM_1000_KX; - break; - case I40E_PHY_TYPE_SGMII: - ifmr->ifm_active |= IFM_1000_SGMII; - break; - /* ERJ: What's the difference between these? */ - case I40E_PHY_TYPE_10GBASE_CR1_CU: - case I40E_PHY_TYPE_10GBASE_CR1: - ifmr->ifm_active |= IFM_10G_CR1; - break; - case I40E_PHY_TYPE_10GBASE_KX4: - ifmr->ifm_active |= IFM_10G_KX4; - break; - case I40E_PHY_TYPE_10GBASE_KR: - ifmr->ifm_active |= IFM_10G_KR; - break; - case I40E_PHY_TYPE_SFI: - ifmr->ifm_active |= IFM_10G_SFI; - break; - /* Our single 20G media type */ - case I40E_PHY_TYPE_20GBASE_KR2: - ifmr->ifm_active |= IFM_20G_KR2; - break; - case I40E_PHY_TYPE_40GBASE_KR4: - ifmr->ifm_active |= IFM_40G_KR4; - break; - case I40E_PHY_TYPE_XLPPI: - case I40E_PHY_TYPE_40GBASE_AOC: - ifmr->ifm_active |= IFM_40G_XLPPI; - break; - /* Unknown to driver */ - default: - ifmr->ifm_active |= IFM_UNKNOWN; - break; + /* Based on the link speed reported by the PF over the AdminQ, choose a + * PHY type to report. This isn't 100% correct since we don't really + * know the underlying PHY type of the PF, but at least we can report + * a valid link speed... + */ + switch (sc->link_speed) { + case VIRTCHNL_LINK_SPEED_100MB: + ifmr->ifm_active |= IFM_100_TX; + break; + case VIRTCHNL_LINK_SPEED_1GB: + ifmr->ifm_active |= IFM_1000_T; + break; + case VIRTCHNL_LINK_SPEED_10GB: + ifmr->ifm_active |= IFM_10G_SR; + break; + case VIRTCHNL_LINK_SPEED_20GB: + case VIRTCHNL_LINK_SPEED_25GB: + ifmr->ifm_active |= IFM_25G_SR; + break; + case VIRTCHNL_LINK_SPEED_40GB: + ifmr->ifm_active |= IFM_40G_SR4; + break; + default: + ifmr->ifm_active |= IFM_UNKNOWN; + break; } - /* Report flow control status as well */ - if (hw->phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX) - ifmr->ifm_active |= IFM_ETH_TXPAUSE; - if (hw->phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX) - ifmr->ifm_active |= IFM_ETH_RXPAUSE; - #endif + + INIT_DBG_IF(ifp, "end"); } static int @@ -1574,57 +1354,44 @@ ixlv_if_media_change(if_ctx_t ctx) return (ENODEV); } -// TODO: Rework static int ixlv_if_promisc_set(if_ctx_t ctx, int flags) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); struct ifnet *ifp = iflib_get_ifp(ctx); - struct i40e_hw *hw = vsi->hw; - int err; - bool uni = FALSE, multi = FALSE; + + sc->promisc_flags = 0; if (flags & IFF_ALLMULTI || if_multiaddr_count(ifp, MAX_MULTICAST_ADDR) == MAX_MULTICAST_ADDR) - multi = TRUE; + sc->promisc_flags |= FLAG_VF_MULTICAST_PROMISC; if (flags & IFF_PROMISC) - uni = TRUE; + sc->promisc_flags |= FLAG_VF_UNICAST_PROMISC; - err = i40e_aq_set_vsi_unicast_promiscuous(hw, - vsi->seid, uni, NULL, false); - if (err) - return (err); - err = i40e_aq_set_vsi_multicast_promiscuous(hw, - vsi->seid, multi, NULL); - return (err); + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_CONFIGURE_PROMISC); + + return (0); } static void ixlv_if_timer(if_ctx_t ctx, uint16_t qid) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - struct ixlv_sc *sc = vsi->back; - //struct i40e_hw *hw = &sc->hw; - //struct ixl_tx_queue *que = &vsi->tx_queues[qid]; - //u32 mask; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct i40e_hw *hw = &sc->hw; + u32 val; -#if 0 - /* - ** Check status of the queues - */ - mask = (I40E_PFINT_DYN_CTLN_INTENA_MASK | - I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK); - - /* If queue param has outstanding work, trigger sw irq */ - // TODO: TX queues in iflib don't use HW interrupts; does this do anything? - if (que->busy) - wr32(hw, I40E_PFINT_DYN_CTLN(que->txr.me), mask); - #endif - - // XXX: Is this timer per-queue? if (qid != 0) return; + /* Check for when PF triggers a VF reset */ + val = rd32(hw, I40E_VFGEN_RSTAT) & + I40E_VFGEN_RSTAT_VFR_STATE_MASK; + if (val != VIRTCHNL_VFR_VFACTIVE + && val != VIRTCHNL_VFR_COMPLETED) { + ixlv_dbg_info(sc, "reset in progress! (%d)\n", val); + return; + } + /* Fire off the adminq task */ iflib_admin_intr_deferred(ctx); @@ -1635,35 +1402,49 @@ ixlv_if_timer(if_ctx_t ctx, uint16_t qid) static void ixlv_if_vlan_register(if_ctx_t ctx, u16 vtag) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - //struct i40e_hw *hw = vsi->hw; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + struct ixlv_vlan_filter *v; if ((vtag == 0) || (vtag > 4095)) /* Invalid */ return; ++vsi->num_vlans; - // TODO: Redo - // ixlv_add_filter(vsi, hw->mac.addr, vtag); + v = malloc(sizeof(struct ixlv_vlan_filter), M_IXLV, M_WAITOK | M_ZERO); + SLIST_INSERT_HEAD(sc->vlan_filters, v, next); + v->vlan = vtag; + v->flags = IXL_FILTER_ADD; + + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_ADD_VLAN_FILTER); } static void ixlv_if_vlan_unregister(if_ctx_t ctx, u16 vtag) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); - //struct i40e_hw *hw = vsi->hw; + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; + struct ixlv_vlan_filter *v; + int i = 0; if ((vtag == 0) || (vtag > 4095)) /* Invalid */ return; - --vsi->num_vlans; - // TODO: Redo - // ixlv_del_filter(vsi, hw->mac.addr, vtag); + SLIST_FOREACH(v, sc->vlan_filters, next) { + if (v->vlan == vtag) { + v->flags = IXL_FILTER_DEL; + ++i; + --vsi->num_vlans; + } + } + if (i) + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_DEL_VLAN_FILTER); } static uint64_t ixlv_if_get_counter(if_ctx_t ctx, ift_counter cnt) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); + struct ixl_vsi *vsi = &sc->vsi; if_t ifp = iflib_get_ifp(ctx); switch (cnt) { @@ -1697,53 +1478,6 @@ ixlv_if_get_counter(if_ctx_t ctx, ift_counter cnt) } } -static int -ixlv_allocate_pci_resources(struct ixlv_sc *sc) -{ - struct i40e_hw *hw = &sc->hw; - device_t dev = iflib_get_dev(sc->vsi.ctx); - int rid; - - /* Map BAR0 */ - rid = PCIR_BAR(0); - sc->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, - &rid, RF_ACTIVE); - - if (!(sc->pci_mem)) { - device_printf(dev, "Unable to allocate bus resource: PCI memory\n"); - return (ENXIO); - } - - /* Save off the PCI information */ - hw->vendor_id = pci_get_vendor(dev); - hw->device_id = pci_get_device(dev); - hw->revision_id = pci_read_config(dev, PCIR_REVID, 1); - hw->subsystem_vendor_id = - pci_read_config(dev, PCIR_SUBVEND_0, 2); - hw->subsystem_device_id = - pci_read_config(dev, PCIR_SUBDEV_0, 2); - - hw->bus.device = pci_get_slot(dev); - hw->bus.func = pci_get_function(dev); - - /* Save off register access information */ - sc->osdep.mem_bus_space_tag = - rman_get_bustag(sc->pci_mem); - sc->osdep.mem_bus_space_handle = - rman_get_bushandle(sc->pci_mem); - sc->osdep.mem_bus_space_size = rman_get_size(sc->pci_mem); - sc->osdep.flush_reg = I40E_VFGEN_RSTAT; - sc->osdep.dev = dev; - - sc->hw.hw_addr = (u8 *) &sc->osdep.mem_bus_space_handle; - sc->hw.back = &sc->osdep; - - /* Disable adminq interrupts (just in case) */ - /* TODO: Probably not necessary */ - // ixlv_disable_adminq_irq(&sc->hw); - - return (0); - } static void ixlv_free_pci_resources(struct ixlv_sc *sc) @@ -1753,13 +1487,10 @@ ixlv_free_pci_resources(struct ixlv_sc *sc) device_t dev = sc->dev; /* We may get here before stations are setup */ - // TODO: Check if we can still check against sc->msix - if ((sc->msix > 0) || (rx_que == NULL)) + if (rx_que == NULL) goto early; - /* - ** Release all msix VSI resources: - */ + /* Release all interrupts */ iflib_irq_free(vsi->ctx, &vsi->irq); for (int i = 0; i < vsi->num_rx_queues; i++, rx_que++) @@ -1795,6 +1526,7 @@ ixlv_reset(struct ixlv_sc *sc) __func__); return (error); } + pci_enable_busmaster(dev); error = i40e_shutdown_adminq(hw); if (error) { @@ -1807,9 +1539,10 @@ ixlv_reset(struct ixlv_sc *sc) if (error) { device_printf(dev, "%s: init_adminq failed: %d\n", __func__, error); - return(error); + return (error); } + ixlv_enable_adminq_irq(hw); return (0); } @@ -1833,17 +1566,14 @@ ixlv_reset_complete(struct i40e_hw *hw) } static void -ixlv_setup_interface(device_t dev, struct ixl_vsi *vsi) +ixlv_setup_interface(device_t dev, struct ixlv_sc *sc) { + struct ixl_vsi *vsi = &sc->vsi; if_ctx_t ctx = vsi->ctx; - struct ixlv_sc *sc = vsi->back; struct ifnet *ifp = iflib_get_ifp(ctx); - uint64_t cap; - //struct ixl_queue *que = vsi->queues; INIT_DBG_DEV(dev, "begin"); - /* TODO: Remove VLAN_ENCAP_LEN? */ vsi->shared->isc_max_frame_size = ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN; @@ -1853,255 +1583,9 @@ ixlv_setup_interface(device_t dev, struct ixl_vsi *vsi) if_initbaudrate(ifp, IF_Gbps(40)); #endif - /* Media types based on reported link speed over AdminQ */ - ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX, 0, NULL); - ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_T, 0, NULL); - ifmedia_add(&sc->media, IFM_ETHER | IFM_10G_SR, 0, NULL); - ifmedia_add(&sc->media, IFM_ETHER | IFM_25G_SR, 0, NULL); - ifmedia_add(&sc->media, IFM_ETHER | IFM_40G_SR4, 0, NULL); - - ifmedia_add(&sc->media, IFM_ETHER | IFM_AUTO, 0, NULL); - ifmedia_set(&sc->media, IFM_ETHER | IFM_AUTO); - - INIT_DBG_DEV(dev, "end"); - return (0); -} -#if 0 - -/* -** Allocate and setup a single queue -*/ -static int -ixlv_setup_queue(struct ixlv_sc *sc, struct ixl_queue *que) -{ - device_t dev = sc->dev; - struct tx_ring *txr; - struct rx_ring *rxr; - int rsize, tsize; - int error = I40E_SUCCESS; - - txr = &que->txr; - txr->que = que; - txr->tail = I40E_QTX_TAIL1(que->me); - /* Initialize the TX lock */ - snprintf(txr->mtx_name, sizeof(txr->mtx_name), "%s:tx(%d)", - device_get_nameunit(dev), que->me); - mtx_init(&txr->mtx, txr->mtx_name, NULL, MTX_DEF); - /* - * Create the TX descriptor ring - * - * In Head Writeback mode, the descriptor ring is one bigger - * than the number of descriptors for space for the HW to - * write back index of last completed descriptor. - */ - if (sc->vsi.enable_head_writeback) { - tsize = roundup2((que->num_tx_desc * - sizeof(struct i40e_tx_desc)) + - sizeof(u32), DBA_ALIGN); - } else { - tsize = roundup2((que->num_tx_desc * - sizeof(struct i40e_tx_desc)), DBA_ALIGN); - } - if (i40e_allocate_dma_mem(&sc->hw, - &txr->dma, i40e_mem_reserved, tsize, DBA_ALIGN)) { - device_printf(dev, - "Unable to allocate TX Descriptor memory\n"); - error = ENOMEM; - goto err_destroy_tx_mtx; - } - txr->base = (struct i40e_tx_desc *)txr->dma.va; - bzero((void *)txr->base, tsize); - /* Now allocate transmit soft structs for the ring */ - if (ixl_allocate_tx_data(que)) { - device_printf(dev, - "Critical Failure setting up TX structures\n"); - error = ENOMEM; - goto err_free_tx_dma; - } - /* Allocate a buf ring */ - txr->br = buf_ring_alloc(ixlv_txbrsz, M_DEVBUF, - M_WAITOK, &txr->mtx); - if (txr->br == NULL) { - device_printf(dev, - "Critical Failure setting up TX buf ring\n"); - error = ENOMEM; - goto err_free_tx_data; - } - - /* - * Next the RX queues... - */ - rsize = roundup2(que->num_rx_desc * - sizeof(union i40e_rx_desc), DBA_ALIGN); - rxr = &que->rxr; - rxr->que = que; - rxr->tail = I40E_QRX_TAIL1(que->me); - - /* Initialize the RX side lock */ - snprintf(rxr->mtx_name, sizeof(rxr->mtx_name), "%s:rx(%d)", - device_get_nameunit(dev), que->me); - mtx_init(&rxr->mtx, rxr->mtx_name, NULL, MTX_DEF); - - if (i40e_allocate_dma_mem(&sc->hw, - &rxr->dma, i40e_mem_reserved, rsize, 4096)) { //JFV - should this be DBA? - device_printf(dev, - "Unable to allocate RX Descriptor memory\n"); - error = ENOMEM; - goto err_destroy_rx_mtx; - } - rxr->base = (union i40e_rx_desc *)rxr->dma.va; - bzero((void *)rxr->base, rsize); - - /* Allocate receive soft structs for the ring */ - if (ixl_allocate_rx_data(que)) { - device_printf(dev, - "Critical Failure setting up receive structs\n"); - error = ENOMEM; - goto err_free_rx_dma; - } - - return (0); - -err_free_rx_dma: - i40e_free_dma_mem(&sc->hw, &rxr->dma); -err_destroy_rx_mtx: - mtx_destroy(&rxr->mtx); - /* err_free_tx_buf_ring */ - buf_ring_free(txr->br, M_DEVBUF); -err_free_tx_data: - ixl_free_que_tx(que); -err_free_tx_dma: - i40e_free_dma_mem(&sc->hw, &txr->dma); -err_destroy_tx_mtx: - mtx_destroy(&txr->mtx); - - return (error); -} -#endif - -/* -** Allocate and setup the interface queues -*/ -static int -ixlv_setup_queues(struct ixlv_sc *sc) -{ - device_t dev = sc->dev; - struct ixl_vsi *vsi; - struct ixl_queue *que; - int i; - int error = I40E_SUCCESS; - - vsi = &sc->vsi; - vsi->back = (void *)sc; - vsi->hw = &sc->hw; - vsi->num_vlans = 0; - - /* Get memory for the station queues */ - if (!(vsi->queues = - (struct ixl_queue *) malloc(sizeof(struct ixl_queue) * - vsi->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { - device_printf(dev, "Unable to allocate queue memory\n"); - return ENOMEM; - } - - for (i = 0; i < vsi->num_queues; i++) { - que = &vsi->queues[i]; - que->num_tx_desc = vsi->num_tx_desc; - que->num_rx_desc = vsi->num_rx_desc; - que->me = i; - que->vsi = vsi; - - if (ixlv_setup_queue(sc, que)) { - error = ENOMEM; - goto err_free_queues; - } - } - - return (0); - -err_free_queues: - while (i--) - ixlv_free_queue(sc, &vsi->queues[i]); - - free(vsi->queues, M_DEVBUF); - - return (error); -} - -#if 0 -/* -** This routine is run via an vlan config EVENT, -** it enables us to use the HW Filter table since -** we can get the vlan id. This just creates the -** entry in the soft version of the VFTA, init will -** repopulate the real table. -*/ -static void -ixlv_register_vlan(void *arg, struct ifnet *ifp, u16 vtag) -{ - struct ixl_vsi *vsi = arg; - struct ixlv_sc *sc = vsi->back; - struct ixlv_vlan_filter *v; - - - if (ifp->if_softc != arg) /* Not our event */ - return; - - if ((vtag == 0) || (vtag > 4095)) /* Invalid */ - return; - - /* Sanity check - make sure it doesn't already exist */ - SLIST_FOREACH(v, sc->vlan_filters, next) { - if (v->vlan == vtag) - return; - } - - mtx_lock(&sc->mtx); - ++vsi->num_vlans; - v = malloc(sizeof(struct ixlv_vlan_filter), M_DEVBUF, M_NOWAIT | M_ZERO); - SLIST_INSERT_HEAD(sc->vlan_filters, v, next); - v->vlan = vtag; - v->flags = IXL_FILTER_ADD; - ixl_vc_enqueue(&sc->vc_mgr, &sc->add_vlan_cmd, - IXLV_FLAG_AQ_ADD_VLAN_FILTER, ixl_init_cmd_complete, sc); - mtx_unlock(&sc->mtx); - return; -} - -/* -** This routine is run via an vlan -** unconfig EVENT, remove our entry -** in the soft vfta. -*/ -static void -ixlv_unregister_vlan(void *arg, struct ifnet *ifp, u16 vtag) -{ - struct ixl_vsi *vsi = arg; - struct ixlv_sc *sc = vsi->back; - struct ixlv_vlan_filter *v; - int i = 0; - - if (ifp->if_softc != arg) - return; - - if ((vtag == 0) || (vtag > 4095)) /* Invalid */ - return; - - mtx_lock(&sc->mtx); - SLIST_FOREACH(v, sc->vlan_filters, next) { - if (v->vlan == vtag) { - v->flags = IXL_FILTER_DEL; - ++i; - --vsi->num_vlans; - } - } - if (i) - ixl_vc_enqueue(&sc->vc_mgr, &sc->del_vlan_cmd, - IXLV_FLAG_AQ_DEL_VLAN_FILTER, ixl_init_cmd_complete, sc); - mtx_unlock(&sc->mtx); - return; + ifmedia_add(vsi->media, IFM_ETHER | IFM_AUTO, 0, NULL); + ifmedia_set(vsi->media, IFM_ETHER | IFM_AUTO); } -#endif /* ** Get a new filter and add it to the mac filter list. @@ -2112,7 +1596,7 @@ ixlv_get_mac_filter(struct ixlv_sc *sc) struct ixlv_mac_filter *f; f = malloc(sizeof(struct ixlv_mac_filter), - M_DEVBUF, M_NOWAIT | M_ZERO); + M_IXLV, M_NOWAIT | M_ZERO); if (f) SLIST_INSERT_HEAD(sc->mac_filters, f, next); @@ -2126,7 +1610,7 @@ static struct ixlv_mac_filter * ixlv_find_mac_filter(struct ixlv_sc *sc, u8 *macaddr) { struct ixlv_mac_filter *f; - bool match = FALSE; + bool match = FALSE; SLIST_FOREACH(f, sc->mac_filters, next) { if (cmp_etheraddr(f->macaddr, macaddr)) { @@ -2148,36 +1632,38 @@ ixlv_msix_adminq(void *arg) { struct ixlv_sc *sc = arg; struct i40e_hw *hw = &sc->hw; - // device_t dev = sc->dev; - u32 reg; + u32 reg, mask; bool do_task = FALSE; ++sc->admin_irq; reg = rd32(hw, I40E_VFINT_ICR01); + /* + * For masking off interrupt causes that need to be handled before + * they can be re-enabled + */ mask = rd32(hw, I40E_VFINT_ICR0_ENA1); - reg = rd32(hw, I40E_VFINT_DYN_CTL01); - reg |= I40E_VFINT_DYN_CTL01_CLEARPBA_MASK; - wr32(hw, I40E_VFINT_DYN_CTL01, reg); - /* Check on the cause */ - if (reg & I40E_VFINT_ICR0_ADMINQ_MASK) + if (reg & I40E_VFINT_ICR0_ADMINQ_MASK) { + mask &= ~I40E_VFINT_ICR0_ENA_ADMINQ_MASK; do_task = TRUE; + } + + wr32(hw, I40E_VFINT_ICR0_ENA1, mask); + ixlv_enable_adminq_irq(hw); if (do_task) - iflib_admin_intr_deferred(sc->vsi.ctx); + return (FILTER_SCHEDULE_THREAD); else - ixlv_enable_adminq_irq(hw); - - return (FILTER_HANDLED); + return (FILTER_HANDLED); } void ixlv_enable_intr(struct ixl_vsi *vsi) { - struct i40e_hw *hw = vsi->hw; - struct ixl_rx_queue *que = vsi->rx_queues; + struct i40e_hw *hw = vsi->hw; + struct ixl_rx_queue *que = vsi->rx_queues; ixlv_enable_adminq_irq(hw); for (int i = 0; i < vsi->num_rx_queues; i++, que++) @@ -2187,10 +1673,9 @@ ixlv_enable_intr(struct ixl_vsi *vsi) void ixlv_disable_intr(struct ixl_vsi *vsi) { - struct i40e_hw *hw = vsi->hw; - struct ixl_rx_queue *que = vsi->rx_queues; + struct i40e_hw *hw = vsi->hw; + struct ixl_rx_queue *que = vsi->rx_queues; - ixlv_disable_adminq_irq(hw); for (int i = 0; i < vsi->num_rx_queues; i++, que++) ixlv_disable_queue_irq(hw, que->rxr.me); } @@ -2232,344 +1717,103 @@ ixlv_disable_queue_irq(struct i40e_hw *hw, int id) wr32(hw, I40E_VFINT_DYN_CTLN1(id), I40E_VFINT_DYN_CTLN1_ITR_INDX_MASK); rd32(hw, I40E_VFGEN_RSTAT); - return; } -/* - * Get initial ITR values from tunable values. - */ static void -ixlv_configure_itr(struct ixlv_sc *sc) +ixlv_configure_tx_itr(struct ixlv_sc *sc) { struct i40e_hw *hw = &sc->hw; struct ixl_vsi *vsi = &sc->vsi; - struct ixl_rx_queue *rx_que = vsi->rx_queues; + struct ixl_tx_queue *que = vsi->tx_queues; - vsi->rx_itr_setting = ixlv_rx_itr; - //vsi->tx_itr_setting = ixlv_tx_itr; + vsi->tx_itr_setting = sc->tx_itr; - for (int i = 0; i < vsi->num_rx_queues; i++, rx_que++) { - struct rx_ring *rxr = &rx_que->rxr; - - wr32(hw, I40E_VFINT_ITRN1(IXL_RX_ITR, i), - vsi->rx_itr_setting); - rxr->itr = vsi->rx_itr_setting; - rxr->latency = IXL_AVE_LATENCY; - -#if 0 + for (int i = 0; i < vsi->num_tx_queues; i++, que++) { struct tx_ring *txr = &que->txr; + wr32(hw, I40E_VFINT_ITRN1(IXL_TX_ITR, i), vsi->tx_itr_setting); txr->itr = vsi->tx_itr_setting; txr->latency = IXL_AVE_LATENCY; -#endif } } -/* -** Provide a update to the queue RX -** interrupt moderation value. -*/ static void -ixlv_set_queue_rx_itr(struct ixl_rx_queue *que) +ixlv_configure_rx_itr(struct ixlv_sc *sc) { - struct ixl_vsi *vsi = que->vsi; - struct i40e_hw *hw = vsi->hw; - struct rx_ring *rxr = &que->rxr; - u16 rx_itr; - u16 rx_latency = 0; - int rx_bytes; - - - /* Idle, do nothing */ - if (rxr->bytes == 0) - return; - - if (ixlv_dynamic_rx_itr) { - rx_bytes = rxr->bytes/rxr->itr; - rx_itr = rxr->itr; + struct i40e_hw *hw = &sc->hw; + struct ixl_vsi *vsi = &sc->vsi; + struct ixl_rx_queue *que = vsi->rx_queues; - /* Adjust latency range */ - switch (rxr->latency) { - case IXL_LOW_LATENCY: - if (rx_bytes > 10) { - rx_latency = IXL_AVE_LATENCY; - rx_itr = IXL_ITR_20K; - } - break; - case IXL_AVE_LATENCY: - if (rx_bytes > 20) { - rx_latency = IXL_BULK_LATENCY; - rx_itr = IXL_ITR_8K; - } else if (rx_bytes <= 10) { - rx_latency = IXL_LOW_LATENCY; - rx_itr = IXL_ITR_100K; - } - break; - case IXL_BULK_LATENCY: - if (rx_bytes <= 20) { - rx_latency = IXL_AVE_LATENCY; - rx_itr = IXL_ITR_20K; - } - break; - } + vsi->rx_itr_setting = sc->rx_itr; - rxr->latency = rx_latency; + for (int i = 0; i < vsi->num_rx_queues; i++, que++) { + struct rx_ring *rxr = &que->rxr; - if (rx_itr != rxr->itr) { - /* do an exponential smoothing */ - rx_itr = (10 * rx_itr * rxr->itr) / - ((9 * rx_itr) + rxr->itr); - rxr->itr = min(rx_itr, IXL_MAX_ITR); - wr32(hw, I40E_VFINT_ITRN1(IXL_RX_ITR, - que->rxr.me), rxr->itr); - } - } else { /* We may have have toggled to non-dynamic */ - if (vsi->rx_itr_setting & IXL_ITR_DYNAMIC) - vsi->rx_itr_setting = ixlv_rx_itr; - /* Update the hardware if needed */ - if (rxr->itr != vsi->rx_itr_setting) { - rxr->itr = vsi->rx_itr_setting; - wr32(hw, I40E_VFINT_ITRN1(IXL_RX_ITR, - que->rxr.me), rxr->itr); - } + wr32(hw, I40E_VFINT_ITRN1(IXL_RX_ITR, i), + vsi->rx_itr_setting); + rxr->itr = vsi->rx_itr_setting; + rxr->latency = IXL_AVE_LATENCY; } - rxr->bytes = 0; - rxr->packets = 0; - return; } +/* + * Get initial ITR values from tunable values. + */ +static void +ixlv_configure_itr(struct ixlv_sc *sc) +{ + ixlv_configure_tx_itr(sc); + ixlv_configure_rx_itr(sc); +} /* -** Provide a update to the queue TX +** Provide a update to the queue RX ** interrupt moderation value. */ static void -ixlv_set_queue_tx_itr(struct ixl_tx_queue *que) +ixlv_set_queue_rx_itr(struct ixl_rx_queue *que) { struct ixl_vsi *vsi = que->vsi; struct i40e_hw *hw = vsi->hw; - struct tx_ring *txr = &que->txr; - u16 tx_itr; - u16 tx_latency = 0; - int tx_bytes; - + struct rx_ring *rxr = &que->rxr; /* Idle, do nothing */ - if (txr->bytes == 0) + if (rxr->bytes == 0) return; - if (ixlv_dynamic_tx_itr) { - tx_bytes = txr->bytes/txr->itr; - tx_itr = txr->itr; - - switch (txr->latency) { - case IXL_LOW_LATENCY: - if (tx_bytes > 10) { - tx_latency = IXL_AVE_LATENCY; - tx_itr = IXL_ITR_20K; - } - break; - case IXL_AVE_LATENCY: - if (tx_bytes > 20) { - tx_latency = IXL_BULK_LATENCY; - tx_itr = IXL_ITR_8K; - } else if (tx_bytes <= 10) { - tx_latency = IXL_LOW_LATENCY; - tx_itr = IXL_ITR_100K; - } - break; - case IXL_BULK_LATENCY: - if (tx_bytes <= 20) { - tx_latency = IXL_AVE_LATENCY; - tx_itr = IXL_ITR_20K; - } - break; - } - - txr->latency = tx_latency; - - if (tx_itr != txr->itr) { - /* do an exponential smoothing */ - tx_itr = (10 * tx_itr * txr->itr) / - ((9 * tx_itr) + txr->itr); - txr->itr = min(tx_itr, IXL_MAX_ITR); - wr32(hw, I40E_VFINT_ITRN1(IXL_TX_ITR, - que->txr.me), txr->itr); - } - - } else { /* We may have have toggled to non-dynamic */ - if (vsi->tx_itr_setting & IXL_ITR_DYNAMIC) - vsi->tx_itr_setting = ixlv_tx_itr; - /* Update the hardware if needed */ - if (txr->itr != vsi->tx_itr_setting) { - txr->itr = vsi->tx_itr_setting; - wr32(hw, I40E_VFINT_ITRN1(IXL_TX_ITR, - que->txr.me), txr->itr); - } - } - txr->bytes = 0; - txr->packets = 0; - return; -} - -#if 0 -/* -** -** MSIX Interrupt Handlers and Tasklets -** -*/ -static void -ixlv_handle_que(void *context, int pending) -{ - struct ixl_queue *que = context; - struct ixl_vsi *vsi = que->vsi; - struct i40e_hw *hw = vsi->hw; - struct tx_ring *txr = &que->txr; - struct ifnet *ifp = vsi->ifp; - bool more; - - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - more = ixl_rxeof(que, IXL_RX_LIMIT); - mtx_lock(&txr->mtx); - ixl_txeof(que); - if (!drbr_empty(ifp, txr->br)) - ixl_mq_start_locked(ifp, txr); - mtx_unlock(&txr->mtx); - if (more) { - taskqueue_enqueue(que->tq, &que->task); - return; - } + /* Update the hardware if needed */ + if (rxr->itr != vsi->rx_itr_setting) { + rxr->itr = vsi->rx_itr_setting; + wr32(hw, I40E_VFINT_ITRN1(IXL_RX_ITR, + que->rxr.me), rxr->itr); } - - /* Reenable this interrupt - hmmm */ - ixlv_enable_queue_irq(hw, que->me); - return; } -#endif - static int ixlv_msix_que(void *arg) { - struct ixl_rx_queue *que = arg; + struct ixl_rx_queue *rx_que = arg; - ++que->irqs; + ++rx_que->irqs; - ixlv_set_queue_rx_itr(que); - ixlv_set_queue_tx_itr(que); + ixlv_set_queue_rx_itr(rx_que); + // ixlv_set_queue_tx_itr(que); return (FILTER_SCHEDULE_THREAD); } - -/********************************************************************* - * - * Media Ioctl callback - * - * This routine is called whenever the user queries the status of - * the interface using ifconfig. - * - **********************************************************************/ -static void -ixlv_media_status(struct ifnet * ifp, struct ifmediareq * ifmr) -{ - struct ixl_vsi *vsi = ifp->if_softc; - struct ixlv_sc *sc = vsi->back; - - INIT_DBG_IF(ifp, "begin"); - - mtx_lock(&sc->mtx); - - ixlv_update_link_status(sc); - - ifmr->ifm_status = IFM_AVALID; - ifmr->ifm_active = IFM_ETHER; - - if (!sc->link_up) { - mtx_unlock(&sc->mtx); - INIT_DBG_IF(ifp, "end: link not up"); - return; - } - - ifmr->ifm_status |= IFM_ACTIVE; - /* Hardware is always full-duplex */ - ifmr->ifm_active |= IFM_FDX; - - /* Based on the link speed reported by the PF over the AdminQ, choose a - * PHY type to report. This isn't 100% correct since we don't really - * know the underlying PHY type of the PF, but at least we can report - * a valid link speed... - */ - switch (sc->link_speed) { - case VIRTCHNL_LINK_SPEED_100MB: - ifmr->ifm_active |= IFM_100_TX; - break; - case VIRTCHNL_LINK_SPEED_1GB: - ifmr->ifm_active |= IFM_1000_T; - break; - case VIRTCHNL_LINK_SPEED_10GB: - ifmr->ifm_active |= IFM_10G_SR; - break; - case VIRTCHNL_LINK_SPEED_20GB: - case VIRTCHNL_LINK_SPEED_25GB: - ifmr->ifm_active |= IFM_25G_SR; - break; - case VIRTCHNL_LINK_SPEED_40GB: - ifmr->ifm_active |= IFM_40G_SR4; - break; - default: - ifmr->ifm_active |= IFM_UNKNOWN; - break; - } - - mtx_unlock(&sc->mtx); - INIT_DBG_IF(ifp, "end"); - return; -} - -/********************************************************************* - * - * Media Ioctl callback - * - * This routine is called when the user changes speed/duplex using - * media/mediopt option with ifconfig. - * - **********************************************************************/ -static int -ixlv_media_change(struct ifnet * ifp) -{ - struct ixl_vsi *vsi = ifp->if_softc; - struct ifmedia *ifm = &vsi->media; - - INIT_DBG_IF(ifp, "begin"); - - if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) - return (EINVAL); - - if_printf(ifp, "Changing speed is not supported\n"); - - INIT_DBG_IF(ifp, "end"); - return (ENODEV); -} - - -#if 0 /********************************************************************* * Multicast Initialization * * This routine is called by init to reset a fresh state. * **********************************************************************/ - static void -ixlv_init_multi(struct ixl_vsi *vsi) +ixlv_init_multi(struct ixlv_sc *sc) { struct ixlv_mac_filter *f; - struct ixlv_sc *sc = vsi->back; - int mcnt = 0; - - IOCTL_DBG_IF(vsi->ifp, "begin"); + int mcnt = 0; /* First clear any multicast filters */ SLIST_FOREACH(f, sc->mac_filters, next) { @@ -2580,155 +1824,7 @@ ixlv_init_multi(struct ixl_vsi *vsi) } } if (mcnt > 0) - ixl_vc_enqueue(&sc->vc_mgr, &sc->del_multi_cmd, - IXLV_FLAG_AQ_DEL_MAC_FILTER, ixl_init_cmd_complete, - sc); - - IOCTL_DBG_IF(vsi->ifp, "end"); -} - -static void -ixlv_add_multi(struct ixl_vsi *vsi) -{ - struct ifmultiaddr *ifma; - struct ifnet *ifp = vsi->ifp; - struct ixlv_sc *sc = vsi->back; - int mcnt = 0; - - IOCTL_DBG_IF(ifp, "begin"); - - if_maddr_rlock(ifp); - /* - ** Get a count, to decide if we - ** simply use multicast promiscuous. - */ - CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { - if (ifma->ifma_addr->sa_family != AF_LINK) - continue; - mcnt++; - } - if_maddr_runlock(ifp); - - /* TODO: Remove -- cannot set promiscuous mode in a VF */ - if (__predict_false(mcnt >= MAX_MULTICAST_ADDR)) { - /* delete all multicast filters */ - ixlv_init_multi(vsi); - sc->promiscuous_flags |= FLAG_VF_MULTICAST_PROMISC; - ixl_vc_enqueue(&sc->vc_mgr, &sc->add_multi_cmd, - IXLV_FLAG_AQ_CONFIGURE_PROMISC, ixl_init_cmd_complete, - sc); - IOCTL_DEBUGOUT("%s: end: too many filters", __func__); - return; - } - - mcnt = 0; - if_maddr_rlock(ifp); - CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { - if (ifma->ifma_addr->sa_family != AF_LINK) - continue; - if (!ixlv_add_mac_filter(sc, - (u8*)LLADDR((struct sockaddr_dl *) ifma->ifma_addr), - IXL_FILTER_MC)) - mcnt++; - } - if_maddr_runlock(ifp); - /* - ** Notify AQ task that sw filters need to be - ** added to hw list - */ - if (mcnt > 0) - ixl_vc_enqueue(&sc->vc_mgr, &sc->add_multi_cmd, - IXLV_FLAG_AQ_ADD_MAC_FILTER, ixl_init_cmd_complete, - sc); - - IOCTL_DBG_IF(ifp, "end"); -} - -static void -ixlv_del_multi(struct ixl_vsi *vsi) -{ - struct ixlv_mac_filter *f; - struct ifmultiaddr *ifma; - struct ifnet *ifp = vsi->ifp; - struct ixlv_sc *sc = vsi->back; - int mcnt = 0; - bool match = FALSE; - - IOCTL_DBG_IF(ifp, "begin"); - - /* Search for removed multicast addresses */ - if_maddr_rlock(ifp); - SLIST_FOREACH(f, sc->mac_filters, next) { - if ((f->flags & IXL_FILTER_USED) - && (f->flags & IXL_FILTER_MC)) { - /* check if mac address in filter is in sc's list */ - match = FALSE; - CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { - if (ifma->ifma_addr->sa_family != AF_LINK) - continue; - u8 *mc_addr = - (u8 *)LLADDR((struct sockaddr_dl *)ifma->ifma_addr); - if (cmp_etheraddr(f->macaddr, mc_addr)) { - match = TRUE; - break; - } - } - /* if this filter is not in the sc's list, remove it */ - if (match == FALSE && !(f->flags & IXL_FILTER_DEL)) { - f->flags |= IXL_FILTER_DEL; - mcnt++; - IOCTL_DBG_IF(ifp, "marked: " MAC_FORMAT, - MAC_FORMAT_ARGS(f->macaddr)); - } - else if (match == FALSE) - IOCTL_DBG_IF(ifp, "exists: " MAC_FORMAT, - MAC_FORMAT_ARGS(f->macaddr)); - } - } - if_maddr_runlock(ifp); - - if (mcnt > 0) - ixl_vc_enqueue(&sc->vc_mgr, &sc->del_multi_cmd, - IXLV_FLAG_AQ_DEL_MAC_FILTER, ixl_init_cmd_complete, - sc); - - IOCTL_DBG_IF(ifp, "end"); -} - -static void -ixlv_local_timer(void *arg) -{ - struct ixlv_sc *sc = arg; - struct i40e_hw *hw = &sc->hw; - struct ixl_vsi *vsi = &sc->vsi; - u32 val; - - IXLV_CORE_LOCK_ASSERT(sc); - - /* If Reset is in progress just bail */ - if (sc->init_state == IXLV_RESET_PENDING) - return; - - /* Check for when PF triggers a VF reset */ - val = rd32(hw, I40E_VFGEN_RSTAT) & - I40E_VFGEN_RSTAT_VFR_STATE_MASK; - - if (val != VIRTCHNL_VFR_VFACTIVE - && val != VIRTCHNL_VFR_COMPLETED) { - DDPRINTF(sc->dev, "reset in progress! (%d)", val); - return; - } - - ixlv_request_stats(sc); - - /* clean and process any events */ - taskqueue_enqueue(sc->tq, &sc->aq_irq); - - /* Increment stat when a queue shows hung */ - if (ixl_queue_hang_check(vsi)) - sc->watchdog_events++; - - callout_reset(&sc->timer, hz, ixlv_local_timer, sc); + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_DEL_MAC_FILTER); } /* @@ -2739,29 +1835,23 @@ ixlv_local_timer(void *arg) void ixlv_update_link_status(struct ixlv_sc *sc) { - struct ixl_vsi *vsi = &sc->vsi; - struct ifnet *ifp = vsi->ifp; + struct ixl_vsi *vsi = &sc->vsi; + u64 baudrate; if (sc->link_up){ if (vsi->link_active == FALSE) { - if (bootverbose) - if_printf(ifp,"Link is Up, %s\n", - ixlv_vc_speed_to_string(sc->link_speed)); vsi->link_active = TRUE; - if_link_state_change(ifp, LINK_STATE_UP); + baudrate = ixl_max_vc_speed_to_value(sc->link_speed); + ixlv_dbg_info(sc, "baudrate: %lu\n", baudrate); + iflib_link_state_change(vsi->ctx, LINK_STATE_UP, baudrate); } } else { /* Link down */ if (vsi->link_active == TRUE) { - if (bootverbose) - if_printf(ifp,"Link is Down\n"); - if_link_state_change(ifp, LINK_STATE_DOWN); vsi->link_active = FALSE; + iflib_link_state_change(vsi->ctx, LINK_STATE_DOWN, 0); } } - - return; } -#endif /********************************************************************* * @@ -2774,29 +1864,19 @@ static void ixlv_stop(struct ixlv_sc *sc) { struct ifnet *ifp; - int start; ifp = sc->vsi.ifp; - INIT_DBG_IF(ifp, "begin"); - - ixl_vc_flush(&sc->vc_mgr); - ixlv_disable_queues(sc); - start = ticks; - while ((ifp->if_drv_flags & IFF_DRV_RUNNING) && - ((ticks - start) < hz/10)) - ixlv_do_adminq_locked(sc); + ixlv_disable_intr(&sc->vsi); - /* Stop the local timer */ - callout_stop(&sc->timer); - - INIT_DBG_IF(ifp, "end"); + if (atomic_load_acq_32(&sc->queues_enabled)) + ixlv_send_vc_msg_sleep(sc, IXLV_FLAG_AQ_DISABLE_QUEUES); } static void ixlv_if_stop(if_ctx_t ctx) { - struct ixl_vsi *vsi = iflib_get_softc(ctx); + struct ixlv_sc *sc = iflib_get_softc(ctx); ixlv_stop(sc); } @@ -2888,14 +1968,11 @@ ixlv_config_rss_reg(struct ixlv_sc *sc) static void ixlv_config_rss_pf(struct ixlv_sc *sc) { - ixl_vc_enqueue(&sc->vc_mgr, &sc->config_rss_key_cmd, - IXLV_FLAG_AQ_CONFIG_RSS_KEY, ixl_init_cmd_complete, sc); + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_CONFIG_RSS_KEY); - ixl_vc_enqueue(&sc->vc_mgr, &sc->set_rss_hena_cmd, - IXLV_FLAG_AQ_SET_RSS_HENA, ixl_init_cmd_complete, sc); + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_SET_RSS_HENA); - ixl_vc_enqueue(&sc->vc_mgr, &sc->config_rss_lut_cmd, - IXLV_FLAG_AQ_CONFIG_RSS_LUT, ixl_init_cmd_complete, sc); + ixlv_send_vc_msg(sc, IXLV_FLAG_AQ_CONFIG_RSS_LUT); } /* @@ -2907,42 +1984,16 @@ static void ixlv_config_rss(struct ixlv_sc *sc) { if (sc->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_REG) { - DDPRINTF(sc->dev, "Setting up RSS using VF registers..."); + ixlv_dbg_info(sc, "Setting up RSS using VF registers..."); ixlv_config_rss_reg(sc); } else if (sc->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) { - DDPRINTF(sc->dev, "Setting up RSS using messages to PF..."); + ixlv_dbg_info(sc, "Setting up RSS using messages to PF..."); ixlv_config_rss_pf(sc); } else device_printf(sc->dev, "VF does not support RSS capability sent by PF.\n"); } /* -** This routine refreshes vlan filters, called by init -** it scans the filter table and then updates the AQ -*/ -static void -ixlv_setup_vlan_filters(struct ixlv_sc *sc) -{ - struct ixl_vsi *vsi = &sc->vsi; - struct ixlv_vlan_filter *f; - int cnt = 0; - - if (vsi->num_vlans == 0) - return; - /* - ** Scan the filter table for vlan entries, - ** and if found call for the AQ update. - */ - SLIST_FOREACH(f, sc->vlan_filters, next) - if (f->flags & IXL_FILTER_ADD) - cnt++; - if (cnt > 0) - ixl_vc_enqueue(&sc->vc_mgr, &sc->add_vlan_cmd, - IXLV_FLAG_AQ_ADD_VLAN_FILTER, ixl_init_cmd_complete, sc); -} - - -/* ** This routine adds new MAC filters to the sc's list; ** these are later added in hardware by sending a virtual ** channel message. @@ -2955,7 +2006,7 @@ ixlv_add_mac_filter(struct ixlv_sc *sc, u8 *macaddr, u16 flags) /* Does one already exist? */ f = ixlv_find_mac_filter(sc, macaddr); if (f != NULL) { - IDPRINTF(sc->vsi.ifp, "exists: " MAC_FORMAT, + ixlv_dbg_filter(sc, "exists: " MAC_FORMAT "\n", MAC_FORMAT_ARGS(macaddr)); return (EEXIST); } @@ -2963,12 +2014,12 @@ ixlv_add_mac_filter(struct ixlv_sc *sc, u8 *macaddr, u16 flags) /* If not, get a new empty filter */ f = ixlv_get_mac_filter(sc); if (f == NULL) { - if_printf(sc->vsi.ifp, "%s: no filters available!!\n", + device_printf(sc->dev, "%s: no filters available!!\n", __func__); return (ENOMEM); } - IDPRINTF(sc->vsi.ifp, "marked: " MAC_FORMAT, + ixlv_dbg_filter(sc, "marked: " MAC_FORMAT "\n", MAC_FORMAT_ARGS(macaddr)); bcopy(macaddr, f->macaddr, ETHER_ADDR_LEN); @@ -2993,236 +2044,79 @@ ixlv_del_mac_filter(struct ixlv_sc *sc, u8 *macaddr) return (0); } +/* + * Re-uses the name from the PF driver. + */ static void -ixlv_do_adminq_locked(struct ixlv_sc *sc) +ixlv_add_device_sysctls(struct ixlv_sc *sc) { - struct i40e_hw *hw = &sc->hw; - struct i40e_arq_event_info event; - struct virtchnl_msg *v_msg; - device_t dev = sc->dev; - u16 result = 0; - u32 reg, oldreg; - i40e_status ret; - bool aq_error = false; + struct ixl_vsi *vsi = &sc->vsi; + device_t dev = sc->dev; - event.buf_len = IXL_AQ_BUF_SZ; - event.msg_buf = sc->aq_buffer; - v_msg = (struct virtchnl_msg *)&event.desc; + struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); + struct sysctl_oid_list *ctx_list = + SYSCTL_CHILDREN(device_get_sysctl_tree(dev)); + struct sysctl_oid *debug_node; + struct sysctl_oid_list *debug_list; - do { - ret = i40e_clean_arq_element(hw, &event, &result); - if (ret) - break; - ixlv_vc_completion(sc, v_msg->v_opcode, - v_msg->v_retval, event.msg_buf, event.msg_len); - if (result != 0) - bzero(event.msg_buf, IXL_AQ_BUF_SZ); - } while (result); + SYSCTL_ADD_PROC(ctx, ctx_list, + OID_AUTO, "current_speed", CTLTYPE_STRING | CTLFLAG_RD, + sc, 0, ixlv_sysctl_current_speed, "A", "Current Port Speed"); - /* check for Admin queue errors */ - oldreg = reg = rd32(hw, hw->aq.arq.len); - if (reg & I40E_VF_ARQLEN1_ARQVFE_MASK) { - device_printf(dev, "ARQ VF Error detected\n"); - reg &= ~I40E_VF_ARQLEN1_ARQVFE_MASK; - aq_error = true; - } - if (reg & I40E_VF_ARQLEN1_ARQOVFL_MASK) { - device_printf(dev, "ARQ Overflow Error detected\n"); - reg &= ~I40E_VF_ARQLEN1_ARQOVFL_MASK; - aq_error = true; - } - if (reg & I40E_VF_ARQLEN1_ARQCRIT_MASK) { - device_printf(dev, "ARQ Critical Error detected\n"); - reg &= ~I40E_VF_ARQLEN1_ARQCRIT_MASK; - aq_error = true; - } - if (oldreg != reg) - wr32(hw, hw->aq.arq.len, reg); + SYSCTL_ADD_PROC(ctx, ctx_list, + OID_AUTO, "tx_itr", CTLTYPE_INT | CTLFLAG_RW, + sc, 0, ixlv_sysctl_tx_itr, "I", + "Immediately set TX ITR value for all queues"); - oldreg = reg = rd32(hw, hw->aq.asq.len); - if (reg & I40E_VF_ATQLEN1_ATQVFE_MASK) { - device_printf(dev, "ASQ VF Error detected\n"); - reg &= ~I40E_VF_ATQLEN1_ATQVFE_MASK; - aq_error = true; - } - if (reg & I40E_VF_ATQLEN1_ATQOVFL_MASK) { - device_printf(dev, "ASQ Overflow Error detected\n"); - reg &= ~I40E_VF_ATQLEN1_ATQOVFL_MASK; - aq_error = true; - } - if (reg & I40E_VF_ATQLEN1_ATQCRIT_MASK) { - device_printf(dev, "ASQ Critical Error detected\n"); - reg &= ~I40E_VF_ATQLEN1_ATQCRIT_MASK; - aq_error = true; - } - if (oldreg != reg) - wr32(hw, hw->aq.asq.len, reg); + SYSCTL_ADD_PROC(ctx, ctx_list, + OID_AUTO, "rx_itr", CTLTYPE_INT | CTLFLAG_RW, + sc, 0, ixlv_sysctl_rx_itr, "I", + "Immediately set RX ITR value for all queues"); - if (aq_error) { - /* Need to reset adapter */ - device_printf(dev, "WARNING: Resetting!\n"); - sc->init_state = IXLV_RESET_REQUIRED; - ixlv_stop(sc); - // TODO: Make stop/init calls match - ixlv_if_init(sc->vsi.ctx); - } - ixlv_enable_adminq_irq(hw); -} + /* Add sysctls meant to print debug information, but don't list them + * in "sysctl -a" output. */ + debug_node = SYSCTL_ADD_NODE(ctx, ctx_list, + OID_AUTO, "debug", CTLFLAG_RD | CTLFLAG_SKIP, NULL, "Debug Sysctls"); + debug_list = SYSCTL_CHILDREN(debug_node); -static void -ixlv_add_sysctls(struct ixlv_sc *sc) -{ - device_t dev = sc->dev; - struct ixl_vsi *vsi = &sc->vsi; - struct i40e_eth_stats *es = &vsi->eth_stats; + SYSCTL_ADD_UINT(ctx, debug_list, + OID_AUTO, "shared_debug_mask", CTLFLAG_RW, + &sc->hw.debug_mask, 0, "Shared code debug message level"); - struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); - struct sysctl_oid *tree = device_get_sysctl_tree(dev); - struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree); + SYSCTL_ADD_UINT(ctx, debug_list, + OID_AUTO, "core_debug_mask", CTLFLAG_RW, + &sc->dbg_mask, 0, "Non-shared code debug message level"); - struct sysctl_oid *vsi_node; // *queue_node; - struct sysctl_oid_list *vsi_list; // *queue_list; + SYSCTL_ADD_PROC(ctx, debug_list, + OID_AUTO, "filter_list", CTLTYPE_STRING | CTLFLAG_RD, + sc, 0, ixlv_sysctl_sw_filter_list, "A", "SW Filter List"); -#define QUEUE_NAME_LEN 32 - //char queue_namebuf[QUEUE_NAME_LEN]; + SYSCTL_ADD_PROC(ctx, debug_list, + OID_AUTO, "queue_interrupt_table", CTLTYPE_STRING | CTLFLAG_RD, + sc, 0, ixlv_sysctl_queue_interrupt_table, "A", "View MSI-X indices for TX/RX queues"); -#if 0 - struct ixl_queue *queues = vsi->queues; - struct tX_ring *txr; - struct rx_ring *rxr; -#endif + SYSCTL_ADD_PROC(ctx, debug_list, + OID_AUTO, "do_vf_reset", CTLTYPE_INT | CTLFLAG_WR, + sc, 0, ixlv_sysctl_vf_reset, "A", "Request a VF reset from PF"); - /* Driver statistics sysctls */ - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "watchdog_events", - CTLFLAG_RD, &sc->watchdog_events, - "Watchdog timeouts"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "admin_irq", - CTLFLAG_RD, &sc->admin_irq, - "Admin Queue IRQ Handled"); - - SYSCTL_ADD_INT(ctx, child, OID_AUTO, "tx_ring_size", - CTLFLAG_RD, &vsi->num_tx_desc, 0, - "TX ring size"); - SYSCTL_ADD_INT(ctx, child, OID_AUTO, "rx_ring_size", - CTLFLAG_RD, &vsi->num_rx_desc, 0, - "RX ring size"); - - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "current_speed", - CTLTYPE_STRING | CTLFLAG_RD, - sc, 0, ixlv_sysctl_current_speed, - "A", "Current Port Speed"); - - /* VSI statistics sysctls */ - vsi_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "vsi", - CTLFLAG_RD, NULL, "VSI-specific statistics"); - vsi_list = SYSCTL_CHILDREN(vsi_node); - - struct ixl_sysctl_info ctls[] = - { - {&es->rx_bytes, "good_octets_rcvd", "Good Octets Received"}, - {&es->rx_unicast, "ucast_pkts_rcvd", - "Unicast Packets Received"}, - {&es->rx_multicast, "mcast_pkts_rcvd", - "Multicast Packets Received"}, - {&es->rx_broadcast, "bcast_pkts_rcvd", - "Broadcast Packets Received"}, - {&es->rx_discards, "rx_discards", "Discarded RX packets"}, - {&es->rx_unknown_protocol, "rx_unknown_proto", "RX unknown protocol packets"}, - {&es->tx_bytes, "good_octets_txd", "Good Octets Transmitted"}, - {&es->tx_unicast, "ucast_pkts_txd", "Unicast Packets Transmitted"}, - {&es->tx_multicast, "mcast_pkts_txd", - "Multicast Packets Transmitted"}, - {&es->tx_broadcast, "bcast_pkts_txd", - "Broadcast Packets Transmitted"}, - {&es->tx_errors, "tx_errors", "TX packet errors"}, - // end - {0,0,0} - }; - struct ixl_sysctl_info *entry = ctls; - while (entry->stat != NULL) - { - SYSCTL_ADD_QUAD(ctx, child, OID_AUTO, entry->name, - CTLFLAG_RD, entry->stat, - entry->description); - entry++; - } - -#if 0 - /* Queue sysctls */ - for (int q = 0; q < vsi->num_queues; q++) { - snprintf(queue_namebuf, QUEUE_NAME_LEN, "que%d", q); - queue_node = SYSCTL_ADD_NODE(ctx, vsi_list, OID_AUTO, queue_namebuf, - CTLFLAG_RD, NULL, "Queue Name"); - queue_list = SYSCTL_CHILDREN(queue_node); - - txr = &(queues[q].txr); - rxr = &(queues[q].rxr); - - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "mbuf_defrag_failed", - CTLFLAG_RD, &(queues[q].mbuf_defrag_failed), - "m_defrag() failed"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "dropped", - CTLFLAG_RD, &(queues[q].dropped_pkts), - "Driver dropped packets"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "irqs", - CTLFLAG_RD, &(queues[q].irqs), - "irqs on this queue"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "tso_tx", - CTLFLAG_RD, &(queues[q].tso), - "TSO"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "tx_dmamap_failed", - CTLFLAG_RD, &(queues[q].tx_dmamap_failed), - "Driver tx dma failure in xmit"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "no_desc_avail", - CTLFLAG_RD, &(txr->no_desc), - "Queue No Descriptor Available"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "tx_packets", - CTLFLAG_RD, &(txr->total_packets), - "Queue Packets Transmitted"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "tx_bytes", - CTLFLAG_RD, &(txr->tx_bytes), - "Queue Bytes Transmitted"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "rx_packets", - CTLFLAG_RD, &(rxr->rx_packets), - "Queue Packets Received"); - SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "rx_bytes", - CTLFLAG_RD, &(rxr->rx_bytes), - "Queue Bytes Received"); - SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "rx_itr", - CTLFLAG_RD, &(rxr->itr), 0, - "Queue Rx ITR Interval"); - SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "tx_itr", - CTLFLAG_RD, &(txr->itr), 0, - "Queue Tx ITR Interval"); + SYSCTL_ADD_PROC(ctx, debug_list, + OID_AUTO, "do_vflr_reset", CTLTYPE_INT | CTLFLAG_WR, + sc, 0, ixlv_sysctl_vflr_reset, "A", "Request a VFLR reset from HW"); + + /* Add stats sysctls */ + ixl_add_vsi_sysctls(dev, vsi, ctx, "vsi"); + ixl_add_queues_sysctls(dev, vsi); -#ifdef IXL_DEBUG - /* Examine queue state */ - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "qtx_head", - CTLTYPE_UINT | CTLFLAG_RD, &queues[q], - sizeof(struct ixl_queue), - ixlv_sysctl_qtx_tail_handler, "IU", - "Queue Transmit Descriptor Tail"); - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "qrx_head", - CTLTYPE_UINT | CTLFLAG_RD, &queues[q], - sizeof(struct ixl_queue), - ixlv_sysctl_qrx_tail_handler, "IU", - "Queue Receive Descriptor Tail"); - SYSCTL_ADD_INT(ctx, queue_list, OID_AUTO, "watchdog_timer", - CTLFLAG_RD, &(txr.watchdog_timer), 0, - "Ticks before watchdog event is triggered"); -#endif - } -#endif } static void ixlv_init_filters(struct ixlv_sc *sc) { - sc->mac_filters = malloc(sizeof(struct ixlv_mac_filter), - M_DEVBUF, M_NOWAIT | M_ZERO); + sc->mac_filters = malloc(sizeof(struct mac_list), + M_IXLV, M_WAITOK | M_ZERO); SLIST_INIT(sc->mac_filters); - sc->vlan_filters = malloc(sizeof(struct ixlv_vlan_filter), - M_DEVBUF, M_NOWAIT | M_ZERO); + sc->vlan_filters = malloc(sizeof(struct vlan_list), + M_IXLV, M_WAITOK | M_ZERO); SLIST_INIT(sc->vlan_filters); } @@ -3235,18 +2129,18 @@ ixlv_free_filters(struct ixlv_sc *sc) while (!SLIST_EMPTY(sc->mac_filters)) { f = SLIST_FIRST(sc->mac_filters); SLIST_REMOVE_HEAD(sc->mac_filters, next); - free(f, M_DEVBUF); + free(f, M_IXLV); } - free(sc->mac_filters, M_DEVBUF); + free(sc->mac_filters, M_IXLV); while (!SLIST_EMPTY(sc->vlan_filters)) { v = SLIST_FIRST(sc->vlan_filters); SLIST_REMOVE_HEAD(sc->vlan_filters, next); - free(v, M_DEVBUF); + free(v, M_IXLV); } - free(sc->vlan_filters, M_DEVBUF); + free(sc->vlan_filters, M_IXLV); } -static char * +char * ixlv_vc_speed_to_string(enum virtchnl_link_speed link_speed) { int index; @@ -3301,49 +2195,241 @@ ixlv_sysctl_current_speed(SYSCTL_HANDLER_ARGS) return (error); } -#ifdef IXL_DEBUG -/** - * ixlv_sysctl_qtx_tail_handler - * Retrieves I40E_QTX_TAIL1 value from hardware - * for a sysctl. +/* + * Sanity check and save off tunable values. */ -static int -ixlv_sysctl_qtx_tail_handler(SYSCTL_HANDLER_ARGS) +static void +ixlv_save_tunables(struct ixlv_sc *sc) { - struct ixl_queue *que; - int error; - u32 val; + device_t dev = sc->dev; - que = ((struct ixl_queue *)oidp->oid_arg1); - if (!que) return 0; + /* Save tunable information */ + sc->dbg_mask = ixlv_core_debug_mask; + sc->hw.debug_mask = ixlv_shared_debug_mask; + sc->vsi.enable_head_writeback = !!(ixlv_enable_head_writeback); + + if (ixlv_tx_itr < 0 || ixlv_tx_itr > IXL_MAX_ITR) { + device_printf(dev, "Invalid tx_itr value of %d set!\n", + ixlv_tx_itr); + device_printf(dev, "tx_itr must be between %d and %d, " + "inclusive\n", + 0, IXL_MAX_ITR); + device_printf(dev, "Using default value of %d instead\n", + IXL_ITR_4K); + sc->tx_itr = IXL_ITR_4K; + } else + sc->tx_itr = ixlv_tx_itr; + + if (ixlv_rx_itr < 0 || ixlv_rx_itr > IXL_MAX_ITR) { + device_printf(dev, "Invalid rx_itr value of %d set!\n", + ixlv_rx_itr); + device_printf(dev, "rx_itr must be between %d and %d, " + "inclusive\n", + 0, IXL_MAX_ITR); + device_printf(dev, "Using default value of %d instead\n", + IXL_ITR_8K); + sc->rx_itr = IXL_ITR_8K; + } else + sc->rx_itr = ixlv_rx_itr; +} - val = rd32(que->vsi->hw, que->txr.tail); - error = sysctl_handle_int(oidp, &val, 0, req); - if (error || !req->newptr) - return error; - return (0); +/* + * Used to set the Tx ITR value for all of the VF's queues. + * Writes to the ITR registers immediately. + */ +static int +ixlv_sysctl_tx_itr(SYSCTL_HANDLER_ARGS) +{ + struct ixlv_sc *sc = (struct ixlv_sc *)arg1; + device_t dev = sc->dev; + int requested_tx_itr; + int error = 0; + + requested_tx_itr = sc->tx_itr; + error = sysctl_handle_int(oidp, &requested_tx_itr, 0, req); + if ((error) || (req->newptr == NULL)) + return (error); + if (requested_tx_itr < 0 || requested_tx_itr > IXL_MAX_ITR) { + device_printf(dev, + "Invalid TX itr value; value must be between 0 and %d\n", + IXL_MAX_ITR); + return (EINVAL); + } + + sc->tx_itr = requested_tx_itr; + ixlv_configure_tx_itr(sc); + + return (error); } -/** - * ixlv_sysctl_qrx_tail_handler - * Retrieves I40E_QRX_TAIL1 value from hardware - * for a sysctl. +/* + * Used to set the Rx ITR value for all of the VF's queues. + * Writes to the ITR registers immediately. */ -static int -ixlv_sysctl_qrx_tail_handler(SYSCTL_HANDLER_ARGS) +static int +ixlv_sysctl_rx_itr(SYSCTL_HANDLER_ARGS) { - struct ixl_queue *que; - int error; - u32 val; + struct ixlv_sc *sc = (struct ixlv_sc *)arg1; + device_t dev = sc->dev; + int requested_rx_itr; + int error = 0; + + requested_rx_itr = sc->rx_itr; + error = sysctl_handle_int(oidp, &requested_rx_itr, 0, req); + if ((error) || (req->newptr == NULL)) + return (error); + if (requested_rx_itr < 0 || requested_rx_itr > IXL_MAX_ITR) { + device_printf(dev, + "Invalid RX itr value; value must be between 0 and %d\n", + IXL_MAX_ITR); + return (EINVAL); + } - que = ((struct ixl_queue *)oidp->oid_arg1); - if (!que) return 0; + sc->rx_itr = requested_rx_itr; + ixlv_configure_rx_itr(sc); - val = rd32(que->vsi->hw, que->rxr.tail); - error = sysctl_handle_int(oidp, &val, 0, req); - if (error || !req->newptr) - return error; - return (0); + return (error); } -#endif +static int +ixlv_sysctl_sw_filter_list(SYSCTL_HANDLER_ARGS) +{ + struct ixlv_sc *sc = (struct ixlv_sc *)arg1; + struct ixlv_mac_filter *f; + struct ixlv_vlan_filter *v; + device_t dev = sc->dev; + int ftl_len, ftl_counter = 0, error = 0; + struct sbuf *buf; + + buf = sbuf_new_for_sysctl(NULL, NULL, 128, req); + if (!buf) { + device_printf(dev, "Could not allocate sbuf for output.\n"); + return (ENOMEM); + } + + sbuf_printf(buf, "\n"); + + /* Print MAC filters */ + sbuf_printf(buf, "MAC Filters:\n"); + ftl_len = 0; + SLIST_FOREACH(f, sc->mac_filters, next) + ftl_len++; + if (ftl_len < 1) + sbuf_printf(buf, "(none)\n"); + else { + SLIST_FOREACH(f, sc->mac_filters, next) { + sbuf_printf(buf, + MAC_FORMAT ", flags %#06x\n", + MAC_FORMAT_ARGS(f->macaddr), f->flags); + } + } + + /* Print VLAN filters */ + sbuf_printf(buf, "VLAN Filters:\n"); + ftl_len = 0; + SLIST_FOREACH(v, sc->vlan_filters, next) + ftl_len++; + if (ftl_len < 1) + sbuf_printf(buf, "(none)"); + else { + SLIST_FOREACH(v, sc->vlan_filters, next) { + sbuf_printf(buf, + "%d, flags %#06x", + v->vlan, v->flags); + /* don't print '\n' for last entry */ + if (++ftl_counter != ftl_len) + sbuf_printf(buf, "\n"); + } + } + + error = sbuf_finish(buf); + if (error) + device_printf(dev, "Error finishing sbuf: %d\n", error); + + sbuf_delete(buf); + return (error); +} + +/* + * Print out mapping of TX queue indexes and Rx queue indexes + * to MSI-X vectors. + */ +static int +ixlv_sysctl_queue_interrupt_table(SYSCTL_HANDLER_ARGS) +{ + struct ixlv_sc *sc = (struct ixlv_sc *)arg1; + struct ixl_vsi *vsi = &sc->vsi; + device_t dev = sc->dev; + struct sbuf *buf; + int error = 0; + + struct ixl_rx_queue *rx_que = vsi->rx_queues; + struct ixl_tx_queue *tx_que = vsi->tx_queues; + + buf = sbuf_new_for_sysctl(NULL, NULL, 128, req); + if (!buf) { + device_printf(dev, "Could not allocate sbuf for output.\n"); + return (ENOMEM); + } + + sbuf_cat(buf, "\n"); + for (int i = 0; i < vsi->num_rx_queues; i++) { + rx_que = &vsi->rx_queues[i]; + sbuf_printf(buf, "(rxq %3d): %d\n", i, rx_que->msix); + } + for (int i = 0; i < vsi->num_tx_queues; i++) { + tx_que = &vsi->tx_queues[i]; + sbuf_printf(buf, "(txq %3d): %d\n", i, tx_que->msix); + } + + error = sbuf_finish(buf); + if (error) + device_printf(dev, "Error finishing sbuf: %d\n", error); + sbuf_delete(buf); + + return (error); +} + +#define CTX_ACTIVE(ctx) ((if_getdrvflags(iflib_get_ifp(ctx)) & IFF_DRV_RUNNING)) +static int +ixlv_sysctl_vf_reset(SYSCTL_HANDLER_ARGS) +{ + struct ixlv_sc *sc = (struct ixlv_sc *)arg1; + int do_reset = 0, error = 0; + + error = sysctl_handle_int(oidp, &do_reset, 0, req); + if ((error) || (req->newptr == NULL)) + return (error); + + if (do_reset == 1) { + ixlv_reset(sc); + if (CTX_ACTIVE(sc->vsi.ctx)) + iflib_request_reset(sc->vsi.ctx); + } + + return (error); +} + +static int +ixlv_sysctl_vflr_reset(SYSCTL_HANDLER_ARGS) +{ + struct ixlv_sc *sc = (struct ixlv_sc *)arg1; + device_t dev = sc->dev; + int do_reset = 0, error = 0; + + error = sysctl_handle_int(oidp, &do_reset, 0, req); + if ((error) || (req->newptr == NULL)) + return (error); + + if (do_reset == 1) { + if (!pcie_flr(dev, max(pcie_get_max_completion_timeout(dev) / 1000, 10), true)) { + device_printf(dev, "PCIE FLR failed\n"); + error = EIO; + } + else if (CTX_ACTIVE(sc->vsi.ctx)) + iflib_request_reset(sc->vsi.ctx); + } + + return (error); +} +#undef CTX_ACTIVE diff --git a/sys/dev/ixl/ixl.h b/sys/dev/ixl/ixl.h index 90edfa2e4cba..e3147fc8939d 100644 --- a/sys/dev/ixl/ixl.h +++ b/sys/dev/ixl/ixl.h @@ -32,7 +32,6 @@ ******************************************************************************/ /*$FreeBSD$*/ - #ifndef _IXL_H_ #define _IXL_H_ @@ -136,8 +135,6 @@ #define IXL_MSIX_BAR 3 #define IXL_ADM_LIMIT 2 -// TODO: Find out which TSO_SIZE to use -//#define IXL_TSO_SIZE 65535 #define IXL_TSO_SIZE ((255*1024)-1) #define IXL_TX_BUF_SZ ((u32) 1514) #define IXL_AQ_BUF_SZ ((u32) 4096) @@ -145,12 +142,14 @@ #define IXL_TX_ITR 1 #define IXL_ITR_NONE 3 #define IXL_QUEUE_EOL 0x7FF +#define IXL_MIN_FRAME 17 #define IXL_MAX_FRAME 9728 #define IXL_MAX_TX_SEGS 8 #define IXL_MAX_RX_SEGS 5 #define IXL_MAX_TSO_SEGS 128 #define IXL_SPARSE_CHAIN 7 #define IXL_MIN_TSO_MSS 64 +#define IXL_MAX_TSO_MSS 9668 #define IXL_MAX_DMA_SEG_SIZE ((16 * 1024) - 1) #define IXL_RSS_KEY_SIZE_REG 13 @@ -210,16 +209,6 @@ #define IXL_RX_CTX_BASE_UNITS 128 #define IXL_TX_CTX_BASE_UNITS 128 -#if 0 -#define IXL_VPINT_LNKLSTN_REG(hw, vector, vf_num) \ - I40E_VPINT_LNKLSTN(((vector) - 1) + \ - (((hw)->func_caps.num_msix_vectors_vf - 1) * (vf_num))) - -#define IXL_VFINT_DYN_CTLN_REG(hw, vector, vf_num) \ - I40E_VFINT_DYN_CTLN(((vector) - 1) + \ - (((hw)->func_caps.num_msix_vectors_vf - 1) * (vf_num))) -#endif - #define IXL_PF_PCI_CIAA_VF_DEVICE_STATUS 0xAA #define IXL_PF_PCI_CIAD_VF_TRANS_PENDING_MASK 0x20 @@ -299,6 +288,9 @@ #define IXL_SET_NOPROTO(vsi, count) (vsi)->noproto = (count) #endif +/* For stats sysctl naming */ +#define QUEUE_NAME_LEN 32 + #define IXL_DEV_ERR(_dev, _format, ...) \ device_printf(_dev, "%s: " _format " (%s:%d)\n", __func__, ##__VA_ARGS__, __FILE__, __LINE__) @@ -415,16 +407,15 @@ struct ixl_vsi { if_ctx_t ctx; if_softc_ctx_t shared; struct ifnet *ifp; - //device_t dev; + device_t dev; struct i40e_hw *hw; struct ifmedia *media; -#define num_rx_queues shared->isc_nrxqsets -#define num_tx_queues shared->isc_ntxqsets + + int num_rx_queues; + int num_tx_queues; void *back; enum i40e_vsi_type type; - // TODO: Remove? - u64 que_mask; int id; u32 rx_itr_setting; u32 tx_itr_setting; @@ -541,9 +532,18 @@ struct ixl_sysctl_info { extern const uint8_t ixl_bcast_addr[ETHER_ADDR_LEN]; /* Common function prototypes between PF/VF driver */ +void ixl_debug_core(device_t dev, u32 enabled_mask, u32 mask, char *fmt, ...); void ixl_init_tx_ring(struct ixl_vsi *vsi, struct ixl_tx_queue *que); void ixl_get_default_rss_key(u32 *); const char * i40e_vc_stat_str(struct i40e_hw *hw, enum virtchnl_status_code stat_err); -u64 ixl_max_aq_speed_to_value(u8); +void ixl_init_tx_rsqs(struct ixl_vsi *vsi); +void ixl_init_tx_cidx(struct ixl_vsi *vsi); +u64 ixl_max_vc_speed_to_value(u8 link_speeds); +void ixl_add_vsi_sysctls(device_t dev, struct ixl_vsi *vsi, + struct sysctl_ctx_list *ctx, const char *sysctl_name); +void ixl_add_sysctls_eth_stats(struct sysctl_ctx_list *ctx, + struct sysctl_oid_list *child, + struct i40e_eth_stats *eth_stats); +void ixl_add_queues_sysctls(device_t dev, struct ixl_vsi *vsi); #endif /* _IXL_H_ */ diff --git a/sys/dev/ixl/ixl_debug.h b/sys/dev/ixl/ixl_debug.h index c003a9a60a3e..b97f2b87372d 100644 --- a/sys/dev/ixl/ixl_debug.h +++ b/sys/dev/ixl/ixl_debug.h @@ -91,12 +91,9 @@ enum ixl_dbg_mask { IXL_DBG_EN_DIS = 0x00000002, IXL_DBG_AQ = 0x00000004, IXL_DBG_NVMUPD = 0x00000008, + IXL_DBG_FILTER = 0x00000010, - IXL_DBG_IOCTL_KNOWN = 0x00000010, - IXL_DBG_IOCTL_UNKNOWN = 0x00000020, - IXL_DBG_IOCTL_ALL = 0x00000030, - - I40E_DEBUG_RSS = 0x00000100, + IXL_DEBUG_RSS = 0x00000100, IXL_DBG_IOV = 0x00001000, IXL_DBG_IOV_VC = 0x00002000, @@ -107,4 +104,20 @@ enum ixl_dbg_mask { IXL_DBG_ALL = 0xFFFFFFFF }; +enum ixlv_dbg_mask { + IXLV_DBG_INFO = 0x00000001, + IXLV_DBG_EN_DIS = 0x00000002, + IXLV_DBG_AQ = 0x00000004, + IXLV_DBG_INIT = 0x00000008, + IXLV_DBG_FILTER = 0x00000010, + + IXLV_DEBUG_RSS = 0x00000100, + + IXLV_DBG_VC = 0x00001000, + + IXLV_DBG_SWITCH_INFO = 0x00010000, + + IXLV_DBG_ALL = 0xFFFFFFFF +}; + #endif /* _IXL_DEBUG_H_ */ diff --git a/sys/dev/ixl/ixl_pf.h b/sys/dev/ixl/ixl_pf.h index d2f74dd424a0..35d1cf4679fa 100644 --- a/sys/dev/ixl/ixl_pf.h +++ b/sys/dev/ixl/ixl_pf.h @@ -87,10 +87,6 @@ struct ixl_vf { /* Physical controller structure */ struct ixl_pf { - /* - * This is first so that iflib_get_softc can return - * either the VSI or the PF structures. - */ struct ixl_vsi vsi; struct i40e_hw hw; @@ -103,7 +99,6 @@ struct ixl_pf { int iw_msix; bool iw_enabled; #endif - int if_flags; u32 state; u8 supported_speeds; @@ -111,13 +106,12 @@ struct ixl_pf { struct ixl_pf_qtag qtag; /* Tunable values */ - bool enable_msix; - int max_queues; bool enable_tx_fc_filter; int dynamic_rx_itr; int dynamic_tx_itr; int tx_itr; int rx_itr; + int enable_vf_loopback; bool link_up; int advertised_speed; @@ -126,7 +120,6 @@ struct ixl_pf { bool has_i2c; /* Misc stats maintained by the driver */ - u64 watchdog_events; u64 admin_irq; /* Statistics from hw */ @@ -145,8 +138,7 @@ struct ixl_pf { struct ixl_vf *vfs; int num_vfs; uint16_t veb_seid; - struct task vflr_task; - int vc_debug_lvl; + struct if_irq iov_irq; }; /* @@ -226,6 +218,12 @@ struct ixl_pf { "\t3 - Use Admin Queue command (best)\n" \ "Using the Admin Queue is only supported on 710 devices with FW version 1.7 or higher" +#define IXL_SYSCTL_HELP_VF_LOOPBACK \ +"\nDetermines mode that embedded device switch will use when SR-IOV is initialized:\n" \ +"\t0 - Disable (VEPA)\n" \ +"\t1 - Enable (VEB)\n" \ +"Enabling this will allow VFs in separate VMs to communicate over the hardware bridge." + extern const char * const ixl_fc_string[6]; MALLOC_DECLARE(M_IXL); @@ -242,14 +240,10 @@ MALLOC_DECLARE(M_IXL); ixl_send_vf_nack_msg((pf), (vf), (op), (st), __FILE__, __LINE__) /* Debug printing */ -#define ixl_dbg(p, m, s, ...) ixl_debug_core(p, m, s, ##__VA_ARGS__) -void ixl_debug_core(struct ixl_pf *, enum ixl_dbg_mask, char *, ...); - -/* For stats sysctl naming */ -#define QUEUE_NAME_LEN 32 - -/* For netmap(4) compatibility */ -#define ixl_disable_intr(vsi) ixl_disable_rings_intr(vsi) +#define ixl_dbg(pf, m, s, ...) ixl_debug_core(pf->dev, pf->dbg_mask, m, s, ##__VA_ARGS__) +#define ixl_dbg_info(pf, s, ...) ixl_debug_core(pf->dev, pf->dbg_mask, IXL_DBG_INFO, s, ##__VA_ARGS__) +#define ixl_dbg_filter(pf, s, ...) ixl_debug_core(pf->dev, pf->dbg_mask, IXL_DBG_FILTER, s, ##__VA_ARGS__) +#define ixl_dbg_iov(pf, s, ...) ixl_debug_core(pf->dev, pf->dbg_mask, IXL_DBG_IOV, s, ##__VA_ARGS__) /* PF-only function declarations */ int ixl_setup_interface(device_t, struct ixl_pf *); @@ -292,7 +286,6 @@ void ixl_stat_update32(struct i40e_hw *, u32, bool, u64 *, u64 *); void ixl_stop(struct ixl_pf *); -void ixl_add_vsi_sysctls(struct ixl_pf *pf, struct ixl_vsi *vsi, struct sysctl_ctx_list *ctx, const char *sysctl_name); int ixl_get_hw_capabilities(struct ixl_pf *); void ixl_link_up_msg(struct ixl_pf *); void ixl_update_link_status(struct ixl_pf *); @@ -333,7 +326,7 @@ int ixl_aq_get_link_status(struct ixl_pf *, int ixl_handle_nvmupd_cmd(struct ixl_pf *, struct ifdrv *); void ixl_handle_empr_reset(struct ixl_pf *); int ixl_prepare_for_reset(struct ixl_pf *pf, bool is_up); -int ixl_rebuild_hw_structs_after_reset(struct ixl_pf *, bool is_up); +int ixl_rebuild_hw_structs_after_reset(struct ixl_pf *); void ixl_set_queue_rx_itr(struct ixl_rx_queue *); void ixl_set_queue_tx_itr(struct ixl_tx_queue *); @@ -342,7 +335,7 @@ void ixl_add_filter(struct ixl_vsi *, const u8 *, s16 vlan); void ixl_del_filter(struct ixl_vsi *, const u8 *, s16 vlan); void ixl_reconfigure_filters(struct ixl_vsi *vsi); -int ixl_disable_rings(struct ixl_vsi *); +int ixl_disable_rings(struct ixl_pf *, struct ixl_vsi *, struct ixl_pf_qtag *); int ixl_disable_tx_ring(struct ixl_pf *, struct ixl_pf_qtag *, u16); int ixl_disable_rx_ring(struct ixl_pf *, struct ixl_pf_qtag *, u16); int ixl_disable_ring(struct ixl_pf *pf, struct ixl_pf_qtag *, u16); @@ -364,11 +357,12 @@ void ixl_enable_intr(struct ixl_vsi *); void ixl_disable_rings_intr(struct ixl_vsi *); void ixl_set_promisc(struct ixl_vsi *); void ixl_add_multi(struct ixl_vsi *); -void ixl_del_multi(struct ixl_vsi *); +int ixl_del_multi(struct ixl_vsi *); void ixl_setup_vlan_filters(struct ixl_vsi *); void ixl_init_filters(struct ixl_vsi *); void ixl_add_hw_filters(struct ixl_vsi *, int, int); void ixl_del_hw_filters(struct ixl_vsi *, int); +void ixl_del_default_hw_filters(struct ixl_vsi *); struct ixl_mac_filter * ixl_find_filter(struct ixl_vsi *, const u8 *, s16); void ixl_add_mc_filter(struct ixl_vsi *, u8 *); @@ -400,5 +394,6 @@ s32 ixl_write_i2c_byte_aq(struct ixl_pf *pf, u8 byte_offset, int ixl_get_fw_lldp_status(struct ixl_pf *pf); int ixl_attach_get_link_status(struct ixl_pf *); +u64 ixl_max_aq_speed_to_value(u8); #endif /* _IXL_PF_H_ */ diff --git a/sys/dev/ixl/ixl_pf_iov.c b/sys/dev/ixl/ixl_pf_iov.c index b7ae4cc8e17c..4353545a1a5d 100644 --- a/sys/dev/ixl/ixl_pf_iov.c +++ b/sys/dev/ixl/ixl_pf_iov.c @@ -77,14 +77,21 @@ static void ixl_vf_del_vlan_msg(struct ixl_pf *pf, struct ixl_vf *vf, void *msg, static void ixl_vf_config_promisc_msg(struct ixl_pf *pf, struct ixl_vf *vf, void *msg, uint16_t msg_size); static void ixl_vf_get_stats_msg(struct ixl_pf *pf, struct ixl_vf *vf, void *msg, uint16_t msg_size); static int ixl_vf_reserve_queues(struct ixl_pf *pf, struct ixl_vf *vf, int num_queues); +static int ixl_config_pf_vsi_loopback(struct ixl_pf *pf, bool enable); static int ixl_adminq_err_to_errno(enum i40e_admin_queue_err err); +/* + * TODO: Move pieces of this into iflib and call the rest in a handler? + * + * e.g. ixl_if_iov_set_schema + * + * It's odd to do pci_iov_detach() there while doing pci_iov_attach() + * in the driver. + */ void ixl_initialize_sriov(struct ixl_pf *pf) { - return; -#if 0 device_t dev = pf->dev; struct i40e_hw *hw = &pf->hw; nvlist_t *pf_schema, *vf_schema; @@ -101,7 +108,7 @@ ixl_initialize_sriov(struct ixl_pf *pf) IOV_SCHEMA_HASDEFAULT, FALSE); pci_iov_schema_add_uint16(vf_schema, "num-queues", IOV_SCHEMA_HASDEFAULT, - max(1, hw->func_caps.num_msix_vectors_vf - 1) % IXLV_MAX_QUEUES); + max(1, min(hw->func_caps.num_msix_vectors_vf - 1, IXLV_MAX_QUEUES))); iov_error = pci_iov_attach(dev, pf_schema, vf_schema); if (iov_error != 0) { @@ -110,9 +117,6 @@ ixl_initialize_sriov(struct ixl_pf *pf) iov_error); } else device_printf(dev, "SR-IOV ready\n"); - - pf->vc_debug_lvl = 1; -#endif } @@ -142,7 +146,9 @@ ixl_vf_alloc_vsi(struct ixl_pf *pf, struct ixl_vf *vf) bzero(&vsi_ctx.info, sizeof(vsi_ctx.info)); vsi_ctx.info.valid_sections = htole16(I40E_AQ_VSI_PROP_SWITCH_VALID); - vsi_ctx.info.switch_id = htole16(0); + if (pf->enable_vf_loopback) + vsi_ctx.info.switch_id = + htole16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB); vsi_ctx.info.valid_sections |= htole16(I40E_AQ_VSI_PROP_SECURITY_VALID); vsi_ctx.info.sec_flags = 0; @@ -157,7 +163,7 @@ ixl_vf_alloc_vsi(struct ixl_pf *pf, struct ixl_vf *vf) htole16(I40E_AQ_VSI_PROP_QUEUE_MAP_VALID); vsi_ctx.info.mapping_flags = htole16(I40E_AQ_VSI_QUE_MAP_NONCONTIG); - /* ERJ: Only scattered allocation is supported for VFs right now */ + /* XXX: Only scattered allocation is supported for VFs right now */ for (i = 0; i < vf->qtag.num_active; i++) vsi_ctx.info.queue_mapping[i] = vf->qtag.qidx[i]; for (; i < nitems(vsi_ctx.info.queue_mapping); i++) @@ -172,8 +178,6 @@ ixl_vf_alloc_vsi(struct ixl_pf *pf, struct ixl_vf *vf) return (ixl_adminq_err_to_errno(hw->aq.asq_last_status)); vf->vsi.seid = vsi_ctx.seid; vf->vsi.vsi_num = vsi_ctx.vsi_number; - // TODO: How to deal with num tx queues / num rx queues split? - // I don't think just assigning this variable is going to work vf->vsi.num_rx_queues = vf->qtag.num_active; vf->vsi.num_tx_queues = vf->qtag.num_active; @@ -204,10 +208,15 @@ ixl_vf_setup_vsi(struct ixl_pf *pf, struct ixl_vf *vf) if (error != 0) return (error); + /* Let VF receive broadcast Ethernet frames */ + error = i40e_aq_set_vsi_broadcast(hw, vf->vsi.seid, TRUE, NULL); + if (error) + device_printf(pf->dev, "Error configuring VF VSI for broadcast promiscuous\n"); + /* Re-add VF's MAC/VLAN filters to its VSI */ + ixl_reconfigure_filters(&vf->vsi); + /* Reset stats? */ vf->vsi.hw_filters_add = 0; vf->vsi.hw_filters_del = 0; - // ixl_add_filter(&vf->vsi, ixl_bcast_addr, IXL_VLAN_ANY); - ixl_reconfigure_filters(&vf->vsi); return (0); } @@ -372,12 +381,16 @@ ixl_reset_vf(struct ixl_pf *pf, struct ixl_vf *vf) hw = &pf->hw; + ixl_dbg_iov(pf, "Resetting VF-%d\n", vf->vf_num); + vfrtrig = rd32(hw, I40E_VPGEN_VFRTRIG(vf->vf_num)); vfrtrig |= I40E_VPGEN_VFRTRIG_VFSWR_MASK; wr32(hw, I40E_VPGEN_VFRTRIG(vf->vf_num), vfrtrig); ixl_flush(hw); ixl_reinit_vf(pf, vf); + + ixl_dbg_iov(pf, "Resetting VF-%d done.\n", vf->vf_num); } static void @@ -413,7 +426,8 @@ ixl_reinit_vf(struct ixl_pf *pf, struct ixl_vf *vf) wr32(hw, I40E_VPGEN_VFRTRIG(vf->vf_num), vfrtrig); if (vf->vsi.seid != 0) - ixl_disable_rings(&vf->vsi); + ixl_disable_rings(pf, &vf->vsi, &vf->qtag); + ixl_pf_qmgr_clear_queue_flags(&vf->qtag); ixl_vf_release_resources(pf, vf); ixl_vf_setup_vsi(pf, vf); @@ -649,7 +663,7 @@ ixl_vf_config_rx_queue(struct ixl_pf *pf, struct ixl_vf *vf, rxq.tphwdesc_ena = 1; rxq.tphdata_ena = 1; rxq.tphhead_ena = 1; - rxq.lrxqthresh = 2; + rxq.lrxqthresh = 1; rxq.prefena = 1; status = i40e_set_lan_rx_queue_context(hw, global_queue_num, &rxq); @@ -924,7 +938,7 @@ ixl_vf_enable_queues_msg(struct ixl_pf *pf, struct ixl_vf *vf, void *msg, continue; /* Warn if this queue is already marked as enabled */ if (ixl_pf_qmgr_is_queue_enabled(&vf->qtag, i, true)) - device_printf(pf->dev, "VF %d: TX ring %d is already enabled!\n", + ixl_dbg_iov(pf, "VF %d: TX ring %d is already enabled!\n", vf->vf_num, i); error = ixl_enable_tx_ring(pf, &vf->qtag, i); @@ -949,7 +963,7 @@ ixl_vf_enable_queues_msg(struct ixl_pf *pf, struct ixl_vf *vf, void *msg, continue; /* Warn if this queue is already marked as enabled */ if (ixl_pf_qmgr_is_queue_enabled(&vf->qtag, i, false)) - device_printf(pf->dev, "VF %d: RX ring %d is already enabled!\n", + ixl_dbg_iov(pf, "VF %d: RX ring %d is already enabled!\n", vf->vf_num, i); error = ixl_enable_rx_ring(pf, &vf->qtag, i); if (error) @@ -1003,7 +1017,7 @@ ixl_vf_disable_queues_msg(struct ixl_pf *pf, struct ixl_vf *vf, continue; /* Warn if this queue is already marked as disabled */ if (!ixl_pf_qmgr_is_queue_enabled(&vf->qtag, i, true)) { - device_printf(pf->dev, "VF %d: TX ring %d is already disabled!\n", + ixl_dbg_iov(pf, "VF %d: TX ring %d is already disabled!\n", vf->vf_num, i); continue; } @@ -1029,7 +1043,7 @@ ixl_vf_disable_queues_msg(struct ixl_pf *pf, struct ixl_vf *vf, continue; /* Warn if this queue is already marked as disabled */ if (!ixl_pf_qmgr_is_queue_enabled(&vf->qtag, i, false)) { - device_printf(pf->dev, "VF %d: RX ring %d is already disabled!\n", + ixl_dbg_iov(pf, "VF %d: RX ring %d is already disabled!\n", vf->vf_num, i); continue; } @@ -1292,6 +1306,7 @@ ixl_vf_config_promisc_msg(struct ixl_pf *pf, struct ixl_vf *vf, void *msg, uint16_t msg_size) { struct virtchnl_promisc_info *info; + struct i40e_hw *hw = &pf->hw; enum i40e_status_code code; if (msg_size != sizeof(*info)) { @@ -1301,8 +1316,11 @@ ixl_vf_config_promisc_msg(struct ixl_pf *pf, struct ixl_vf *vf, } if (!(vf->vf_flags & VF_FLAG_PROMISC_CAP)) { - i40e_send_vf_nack(pf, vf, - VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, I40E_ERR_PARAM); + /* + * Do the same thing as the Linux PF driver -- lie to the VF + */ + ixl_send_vf_ack(pf, vf, + VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE); return; } @@ -1313,19 +1331,25 @@ ixl_vf_config_promisc_msg(struct ixl_pf *pf, struct ixl_vf *vf, return; } - code = i40e_aq_set_vsi_unicast_promiscuous(&pf->hw, info->vsi_id, + code = i40e_aq_set_vsi_unicast_promiscuous(hw, vf->vsi.seid, info->flags & FLAG_VF_UNICAST_PROMISC, NULL, TRUE); if (code != I40E_SUCCESS) { + device_printf(pf->dev, "i40e_aq_set_vsi_unicast_promiscuous (seid %d) failed: status %s," + " error %s\n", vf->vsi.seid, i40e_stat_str(hw, code), + i40e_aq_str(hw, hw->aq.asq_last_status)); i40e_send_vf_nack(pf, vf, - VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, code); + VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, I40E_ERR_PARAM); return; } - code = i40e_aq_set_vsi_multicast_promiscuous(&pf->hw, info->vsi_id, + code = i40e_aq_set_vsi_multicast_promiscuous(hw, vf->vsi.seid, info->flags & FLAG_VF_MULTICAST_PROMISC, NULL); if (code != I40E_SUCCESS) { + device_printf(pf->dev, "i40e_aq_set_vsi_multicast_promiscuous (seid %d) failed: status %s," + " error %s\n", vf->vsi.seid, i40e_stat_str(hw, code), + i40e_aq_str(hw, hw->aq.asq_last_status)); i40e_send_vf_nack(pf, vf, - VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, code); + VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, I40E_ERR_PARAM); return; } @@ -1627,19 +1651,25 @@ ixl_handle_vf_msg(struct ixl_pf *pf, struct i40e_arq_event_info *event) /* Handle any VFs that have reset themselves via a Function Level Reset(FLR). */ void -ixl_handle_vflr(void *arg, int pending) +ixl_handle_vflr(struct ixl_pf *pf) { - struct ixl_pf *pf; struct ixl_vf *vf; struct i40e_hw *hw; uint16_t global_vf_num; uint32_t vflrstat_index, vflrstat_mask, vflrstat, icr0; int i; - pf = arg; hw = &pf->hw; - /* TODO: May need to lock this */ + ixl_dbg_iov(pf, "%s: begin\n", __func__); + + /* Re-enable VFLR interrupt cause so driver doesn't miss a + * reset interrupt for another VF */ + icr0 = rd32(hw, I40E_PFINT_ICR0_ENA); + icr0 |= I40E_PFINT_ICR0_ENA_VFLR_MASK; + wr32(hw, I40E_PFINT_ICR0_ENA, icr0); + ixl_flush(hw); + for (i = 0; i < pf->num_vfs; i++) { global_vf_num = hw->func_caps.vf_base_id + i; @@ -1654,17 +1684,12 @@ ixl_handle_vflr(void *arg, int pending) wr32(hw, I40E_GLGEN_VFLRSTAT(vflrstat_index), vflrstat_mask); + ixl_dbg_iov(pf, "Reinitializing VF-%d\n", i); ixl_reinit_vf(pf, vf); + ixl_dbg_iov(pf, "Reinitializing VF-%d done\n", i); } } - atomic_clear_32(&pf->state, IXL_PF_STATE_VF_RESET_REQ); - icr0 = rd32(hw, I40E_PFINT_ICR0_ENA); - icr0 |= I40E_PFINT_ICR0_ENA_VFLR_MASK; - wr32(hw, I40E_PFINT_ICR0_ENA, icr0); - ixl_flush(hw); - - // IXL_PF_UNLOCK() } static int @@ -1721,23 +1746,52 @@ ixl_adminq_err_to_errno(enum i40e_admin_queue_err err) } } +static int +ixl_config_pf_vsi_loopback(struct ixl_pf *pf, bool enable) +{ + struct i40e_hw *hw = &pf->hw; + device_t dev = pf->dev; + struct ixl_vsi *vsi = &pf->vsi; + struct i40e_vsi_context ctxt; + int error; + + memset(&ctxt, 0, sizeof(ctxt)); + + ctxt.seid = vsi->seid; + if (pf->veb_seid != 0) + ctxt.uplink_seid = pf->veb_seid; + ctxt.pf_num = hw->pf_id; + ctxt.connection_type = IXL_VSI_DATA_PORT; + + ctxt.info.valid_sections = htole16(I40E_AQ_VSI_PROP_SWITCH_VALID); + ctxt.info.switch_id = (enable) ? + htole16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB) : 0; + + /* error is set to 0 on success */ + error = i40e_aq_update_vsi_params(hw, &ctxt, NULL); + if (error) { + device_printf(dev, "i40e_aq_update_vsi_params() failed, error %d," + " aq_error %d\n", error, hw->aq.asq_last_status); + } + + return (error); +} + int -ixl_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params) +ixl_if_iov_init(if_ctx_t ctx, uint16_t num_vfs, const nvlist_t *params) { - struct ixl_pf *pf; + struct ixl_pf *pf = iflib_get_softc(ctx); + device_t dev = iflib_get_dev(ctx); struct i40e_hw *hw; struct ixl_vsi *pf_vsi; enum i40e_status_code ret; int i, error; - pf = device_get_softc(dev); hw = &pf->hw; pf_vsi = &pf->vsi; - //IXL_PF_LOCK(pf); pf->vfs = malloc(sizeof(struct ixl_vf) * num_vfs, M_IXL, M_NOWAIT | M_ZERO); - if (pf->vfs == NULL) { error = ENOMEM; goto fail; @@ -1746,65 +1800,77 @@ ixl_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params) for (i = 0; i < num_vfs; i++) sysctl_ctx_init(&pf->vfs[i].ctx); + /* + * Add the VEB and ... + * - do nothing: VEPA mode + * - enable loopback mode on connected VSIs: VEB mode + */ ret = i40e_aq_add_veb(hw, pf_vsi->uplink_seid, pf_vsi->seid, 1, FALSE, &pf->veb_seid, FALSE, NULL); if (ret != I40E_SUCCESS) { - error = ixl_adminq_err_to_errno(hw->aq.asq_last_status); - device_printf(dev, "add_veb failed; code=%d error=%d", ret, - error); + error = hw->aq.asq_last_status; + device_printf(dev, "i40e_aq_add_veb failed; status %s error %s", + i40e_stat_str(hw, ret), i40e_aq_str(hw, error)); goto fail; } + if (pf->enable_vf_loopback) + ixl_config_pf_vsi_loopback(pf, true); + + /* + * Adding a VEB brings back the default MAC filter(s). Remove them, + * and let the driver add the proper filters back. + */ + ixl_del_default_hw_filters(pf_vsi); + ixl_reconfigure_filters(pf_vsi); pf->num_vfs = num_vfs; - //IXL_PF_UNLOCK(pf); return (0); fail: free(pf->vfs, M_IXL); pf->vfs = NULL; - //IXL_PF_UNLOCK(pf); return (error); } void -ixl_iov_uninit(device_t dev) +ixl_if_iov_uninit(if_ctx_t ctx) { - struct ixl_pf *pf; + struct ixl_pf *pf = iflib_get_softc(ctx); struct i40e_hw *hw; struct ixl_vsi *vsi; struct ifnet *ifp; struct ixl_vf *vfs; int i, num_vfs; - pf = device_get_softc(dev); hw = &pf->hw; vsi = &pf->vsi; ifp = vsi->ifp; - //IXL_PF_LOCK(pf); for (i = 0; i < pf->num_vfs; i++) { if (pf->vfs[i].vsi.seid != 0) i40e_aq_delete_element(hw, pf->vfs[i].vsi.seid, NULL); ixl_pf_qmgr_release(&pf->qmgr, &pf->vfs[i].qtag); ixl_free_mac_filters(&pf->vfs[i].vsi); - DDPRINTF(dev, "VF %d: %d released\n", + ixl_dbg_iov(pf, "VF %d: %d released\n", i, pf->vfs[i].qtag.num_allocated); - DDPRINTF(dev, "Unallocated total: %d\n", ixl_pf_qmgr_get_num_free(&pf->qmgr)); + ixl_dbg_iov(pf, "Unallocated total: %d\n", ixl_pf_qmgr_get_num_free(&pf->qmgr)); } if (pf->veb_seid != 0) { i40e_aq_delete_element(hw, pf->veb_seid, NULL); pf->veb_seid = 0; } + /* Reset PF VSI loopback mode */ + if (pf->enable_vf_loopback) + ixl_config_pf_vsi_loopback(pf, false); vfs = pf->vfs; num_vfs = pf->num_vfs; pf->vfs = NULL; pf->num_vfs = 0; - //IXL_PF_UNLOCK(pf); - /* Do this after the unlock as sysctl_ctx_free might sleep. */ + /* sysctl_ctx_free might sleep, but this func is called w/ an sx lock */ for (i = 0; i < num_vfs; i++) sysctl_ctx_free(&vfs[i].ctx); free(vfs, M_IXL); @@ -1823,9 +1889,9 @@ ixl_vf_reserve_queues(struct ixl_pf *pf, struct ixl_vf *vf, int num_queues) if (num_queues < 1) { device_printf(dev, "Setting VF %d num-queues to 1\n", vf->vf_num); num_queues = 1; - } else if (num_queues > 16) { - device_printf(dev, "Setting VF %d num-queues to 16\n", vf->vf_num); - num_queues = 16; + } else if (num_queues > IXLV_MAX_QUEUES) { + device_printf(dev, "Setting VF %d num-queues to %d\n", vf->vf_num, IXLV_MAX_QUEUES); + num_queues = IXLV_MAX_QUEUES; } error = ixl_pf_qmgr_alloc_scattered(&pf->qmgr, num_queues, &vf->qtag); if (error) { @@ -1834,30 +1900,27 @@ ixl_vf_reserve_queues(struct ixl_pf *pf, struct ixl_vf *vf, int num_queues) return (ENOSPC); } - DDPRINTF(dev, "VF %d: %d allocated, %d active", + ixl_dbg_iov(pf, "VF %d: %d allocated, %d active\n", vf->vf_num, vf->qtag.num_allocated, vf->qtag.num_active); - DDPRINTF(dev, "Unallocated total: %d", ixl_pf_qmgr_get_num_free(&pf->qmgr)); + ixl_dbg_iov(pf, "Unallocated total: %d\n", ixl_pf_qmgr_get_num_free(&pf->qmgr)); return (0); } int -ixl_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params) +ixl_if_iov_vf_add(if_ctx_t ctx, uint16_t vfnum, const nvlist_t *params) { + struct ixl_pf *pf = iflib_get_softc(ctx); + device_t dev = pf->dev; char sysctl_name[QUEUE_NAME_LEN]; - struct ixl_pf *pf; struct ixl_vf *vf; const void *mac; size_t size; int error; int vf_num_queues; - pf = device_get_softc(dev); vf = &pf->vfs[vfnum]; - - //IXL_PF_LOCK(pf); vf->vf_num = vfnum; - vf->vsi.back = pf; vf->vf_flags = VF_FLAG_ENABLED; SLIST_INIT(&vf->vsi.ftl); @@ -1893,12 +1956,12 @@ ixl_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params) vf->vf_flags |= VF_FLAG_VLAN_CAP; + /* VF needs to be reset before it can be used */ ixl_reset_vf(pf, vf); out: - //IXL_PF_UNLOCK(pf); if (error == 0) { snprintf(sysctl_name, sizeof(sysctl_name), "vf%d", vfnum); - ixl_add_vsi_sysctls(pf, &vf->vsi, &vf->ctx, sysctl_name); + ixl_add_vsi_sysctls(dev, &vf->vsi, &vf->ctx, sysctl_name); } return (error); diff --git a/sys/dev/ixl/ixl_pf_iov.h b/sys/dev/ixl/ixl_pf_iov.h index edc4ae44676b..e412df86ae6b 100644 --- a/sys/dev/ixl/ixl_pf_iov.h +++ b/sys/dev/ixl/ixl_pf_iov.h @@ -45,19 +45,19 @@ /* Public functions */ /* - * These three are DEVMETHODs required for SR-IOV PF support. + * These three are DEVMETHODs required for SR-IOV PF support in iflib. */ -int ixl_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params); -void ixl_iov_uninit(device_t dev); -int ixl_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params); +int ixl_if_iov_init(if_ctx_t ctx, uint16_t num_vfs, const nvlist_t *params); +void ixl_if_iov_uninit(if_ctx_t ctx); +int ixl_if_iov_vf_add(if_ctx_t ctx, uint16_t vfnum, const nvlist_t *params); /* - * The standard PF driver needs to call these during normal execution when + * The base PF driver needs to call these during normal execution when * SR-IOV mode is active. */ void ixl_initialize_sriov(struct ixl_pf *pf); void ixl_handle_vf_msg(struct ixl_pf *pf, struct i40e_arq_event_info *event); -void ixl_handle_vflr(void *arg, int pending); +void ixl_handle_vflr(struct ixl_pf *pf); void ixl_broadcast_link_state(struct ixl_pf *pf); #endif /* _IXL_PF_IOV_H_ */ diff --git a/sys/dev/ixl/ixl_pf_main.c b/sys/dev/ixl/ixl_pf_main.c index 77ab5b7043d3..f775947fdc4c 100644 --- a/sys/dev/ixl/ixl_pf_main.c +++ b/sys/dev/ixl/ixl_pf_main.c @@ -46,7 +46,6 @@ static u8 ixl_convert_sysctl_aq_link_speed(u8, bool); static void ixl_sbuf_print_bytes(struct sbuf *, u8 *, int, int, bool); -static void ixl_del_default_hw_filters(struct ixl_vsi *); /* Sysctls */ static int ixl_sysctl_set_flowcntl(SYSCTL_HANDLER_ARGS); @@ -113,21 +112,6 @@ static char *ixl_fec_string[3] = { MALLOC_DEFINE(M_IXL, "ixl", "ixl driver allocations"); -void -ixl_debug_core(struct ixl_pf *pf, enum ixl_dbg_mask mask, char *fmt, ...) -{ - va_list args; - - if (!(mask & pf->dbg_mask)) - return; - - /* Re-implement device_printf() */ - device_print_prettyname(pf->dev); - va_start(args, fmt); - vprintf(fmt, args); - va_end(args); -} - /* ** Put the FW, API, NVM, EEtrackID, and OEM version information into a string */ @@ -354,6 +338,7 @@ ixl_teardown_hw_structs(struct ixl_pf *pf) "init: Admin Queue shutdown failure; status %s\n", i40e_stat_str(hw, status)); + ixl_pf_qmgr_release(&pf->qmgr, &pf->qtag); err_out: return (status); } @@ -458,8 +443,7 @@ ixl_reset(struct ixl_pf *pf) err_out: return (error); #endif - // TODO: Fix second parameter - ixl_rebuild_hw_structs_after_reset(pf, false); + ixl_rebuild_hw_structs_after_reset(pf); /* The PF reset should have cleared any critical errors */ atomic_clear_32(&pf->state, IXL_PF_STATE_PF_CRIT_ERR); @@ -527,11 +511,11 @@ ixl_intr(void *arg) int ixl_msix_que(void *arg) { - struct ixl_rx_queue *que = arg; + struct ixl_rx_queue *rx_que = arg; - ++que->irqs; + ++rx_que->irqs; - ixl_set_queue_rx_itr(que); + ixl_set_queue_rx_itr(rx_que); // ixl_set_queue_tx_itr(que); return (FILTER_SCHEDULE_THREAD); @@ -557,8 +541,10 @@ ixl_msix_adminq(void *arg) ++pf->admin_irq; reg = rd32(hw, I40E_PFINT_ICR0); - // For masking off interrupt causes that need to be handled before - // they can be re-enabled + /* + * For masking off interrupt causes that need to be handled before + * they can be re-enabled + */ mask = rd32(hw, I40E_PFINT_ICR0_ENA); /* Check on the cause */ @@ -637,11 +623,12 @@ ixl_msix_adminq(void *arg) #ifdef PCI_IOV if (reg & I40E_PFINT_ICR0_VFLR_MASK) { mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK; - atomic_set_32(&pf->state, IXL_PF_STATE_VF_RESET_REQ); - do_task = TRUE; + iflib_iov_intr_deferred(pf->vsi.ctx); } #endif + wr32(hw, I40E_PFINT_ICR0_ENA, mask); + ixl_enable_intr0(hw); if (do_task) return (FILTER_SCHEDULE_THREAD); @@ -703,7 +690,7 @@ ixl_add_multi(struct ixl_vsi *vsi) IOCTL_DEBUGOUT("ixl_add_multi: end"); } -void +int ixl_del_multi(struct ixl_vsi *vsi) { struct ifnet *ifp = vsi->ifp; @@ -738,6 +725,8 @@ ixl_del_multi(struct ixl_vsi *vsi) if (mcnt > 0) ixl_del_hw_filters(vsi, mcnt); + + return (mcnt); } void @@ -1028,7 +1017,6 @@ ixl_setup_interface(device_t dev, struct ixl_pf *pf) INIT_DBG_DEV(dev, "begin"); - /* TODO: Remove VLAN_ENCAP_LEN? */ vsi->shared->isc_max_frame_size = ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN; @@ -1068,6 +1056,29 @@ ixl_setup_interface(device_t dev, struct ixl_pf *pf) } /* + * Input: bitmap of enum i40e_aq_link_speed + */ +u64 +ixl_max_aq_speed_to_value(u8 link_speeds) +{ + if (link_speeds & I40E_LINK_SPEED_40GB) + return IF_Gbps(40); + if (link_speeds & I40E_LINK_SPEED_25GB) + return IF_Gbps(25); + if (link_speeds & I40E_LINK_SPEED_20GB) + return IF_Gbps(20); + if (link_speeds & I40E_LINK_SPEED_10GB) + return IF_Gbps(10); + if (link_speeds & I40E_LINK_SPEED_1GB) + return IF_Gbps(1); + if (link_speeds & I40E_LINK_SPEED_100MB) + return IF_Mbps(100); + else + /* Minimum supported link speed */ + return IF_Mbps(100); +} + +/* ** Run when the Admin Queue gets a link state change interrupt. */ void @@ -1194,7 +1205,7 @@ ixl_initialize_vsi(struct ixl_vsi *vsi) * the driver may not use all of them). */ tc_queues = fls(pf->qtag.num_allocated) - 1; - ctxt.info.tc_mapping[0] = ((0 << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) + ctxt.info.tc_mapping[0] = ((pf->qtag.first_qidx << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) & I40E_AQ_VSI_TC_QUE_OFFSET_MASK) | ((tc_queues << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT) & I40E_AQ_VSI_TC_QUE_NUMBER_MASK); @@ -1493,23 +1504,6 @@ ixl_set_queue_tx_itr(struct ixl_tx_queue *que) return; } -void -ixl_add_vsi_sysctls(struct ixl_pf *pf, struct ixl_vsi *vsi, - struct sysctl_ctx_list *ctx, const char *sysctl_name) -{ - struct sysctl_oid *tree; - struct sysctl_oid_list *child; - struct sysctl_oid_list *vsi_list; - - tree = device_get_sysctl_tree(pf->dev); - child = SYSCTL_CHILDREN(tree); - vsi->vsi_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, sysctl_name, - CTLFLAG_RD, NULL, "VSI Number"); - vsi_list = SYSCTL_CHILDREN(vsi->vsi_node); - - ixl_add_sysctls_eth_stats(ctx, vsi_list, &vsi->eth_stats); -} - #ifdef IXL_DEBUG /** * ixl_sysctl_qtx_tail_handler @@ -1634,131 +1628,17 @@ ixl_add_hw_stats(struct ixl_pf *pf) struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); struct sysctl_oid *tree = device_get_sysctl_tree(dev); struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree); - struct sysctl_oid_list *vsi_list, *queue_list; - struct sysctl_oid *queue_node; - char queue_namebuf[32]; - - struct ixl_rx_queue *rx_que; - struct ixl_tx_queue *tx_que; - struct tx_ring *txr; - struct rx_ring *rxr; /* Driver statistics */ - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "watchdog_events", - CTLFLAG_RD, &pf->watchdog_events, - "Watchdog timeouts"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "admin_irq", CTLFLAG_RD, &pf->admin_irq, - "Admin Queue IRQ Handled"); - - ixl_add_vsi_sysctls(pf, &pf->vsi, ctx, "pf"); - vsi_list = SYSCTL_CHILDREN(pf->vsi.vsi_node); - - /* Queue statistics */ - for (int q = 0; q < vsi->num_rx_queues; q++) { - snprintf(queue_namebuf, QUEUE_NAME_LEN, "rxq%02d", q); - queue_node = SYSCTL_ADD_NODE(ctx, vsi_list, - OID_AUTO, queue_namebuf, CTLFLAG_RD, NULL, "RX Queue #"); - queue_list = SYSCTL_CHILDREN(queue_node); - - rx_que = &(vsi->rx_queues[q]); - rxr = &(rx_que->rxr); - - - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs", - CTLFLAG_RD, &(rx_que->irqs), - "irqs on this queue (both Tx and Rx)"); - - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "packets", - CTLFLAG_RD, &(rxr->rx_packets), - "Queue Packets Received"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "bytes", - CTLFLAG_RD, &(rxr->rx_bytes), - "Queue Bytes Received"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "desc_err", - CTLFLAG_RD, &(rxr->desc_errs), - "Queue Rx Descriptor Errors"); - SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "itr", - CTLFLAG_RD, &(rxr->itr), 0, - "Queue Rx ITR Interval"); -#ifdef IXL_DEBUG - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "qrx_tail", - CTLTYPE_UINT | CTLFLAG_RD, rx_que, - sizeof(struct ixl_rx_queue), - ixl_sysctl_qrx_tail_handler, "IU", - "Queue Receive Descriptor Tail"); -#endif - } - for (int q = 0; q < vsi->num_tx_queues; q++) { - snprintf(queue_namebuf, QUEUE_NAME_LEN, "txq%02d", q); - queue_node = SYSCTL_ADD_NODE(ctx, vsi_list, - OID_AUTO, queue_namebuf, CTLFLAG_RD, NULL, "TX Queue #"); - queue_list = SYSCTL_CHILDREN(queue_node); - - tx_que = &(vsi->tx_queues[q]); - txr = &(tx_que->txr); - - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tso", - CTLFLAG_RD, &(tx_que->tso), - "TSO"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "mss_too_small", - CTLFLAG_RD, &(txr->mss_too_small), - "TSO sends with an MSS less than 64"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "packets", - CTLFLAG_RD, &(txr->tx_packets), - "Queue Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "bytes", - CTLFLAG_RD, &(txr->tx_bytes), - "Queue Bytes Transmitted"); - SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "itr", - CTLFLAG_RD, &(txr->itr), 0, - "Queue Tx ITR Interval"); -#ifdef IXL_DEBUG - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "qtx_tail", - CTLTYPE_UINT | CTLFLAG_RD, tx_que, - sizeof(struct ixl_tx_queue), - ixl_sysctl_qtx_tail_handler, "IU", - "Queue Transmit Descriptor Tail"); -#endif - } + "Admin Queue IRQs received"); - /* MAC stats */ - ixl_add_sysctls_mac_stats(ctx, child, pf_stats); -} + ixl_add_vsi_sysctls(dev, vsi, ctx, "pf"); -void -ixl_add_sysctls_eth_stats(struct sysctl_ctx_list *ctx, - struct sysctl_oid_list *child, - struct i40e_eth_stats *eth_stats) -{ - struct ixl_sysctl_info ctls[] = - { - {ð_stats->rx_bytes, "good_octets_rcvd", "Good Octets Received"}, - {ð_stats->rx_unicast, "ucast_pkts_rcvd", - "Unicast Packets Received"}, - {ð_stats->rx_multicast, "mcast_pkts_rcvd", - "Multicast Packets Received"}, - {ð_stats->rx_broadcast, "bcast_pkts_rcvd", - "Broadcast Packets Received"}, - {ð_stats->rx_discards, "rx_discards", "Discarded RX packets"}, - {ð_stats->tx_bytes, "good_octets_txd", "Good Octets Transmitted"}, - {ð_stats->tx_unicast, "ucast_pkts_txd", "Unicast Packets Transmitted"}, - {ð_stats->tx_multicast, "mcast_pkts_txd", - "Multicast Packets Transmitted"}, - {ð_stats->tx_broadcast, "bcast_pkts_txd", - "Broadcast Packets Transmitted"}, - // end - {0,0,0} - }; + ixl_add_queues_sysctls(dev, vsi); - struct ixl_sysctl_info *entry = ctls; - while (entry->stat != 0) - { - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, entry->name, - CTLFLAG_RD, entry->stat, - entry->description); - entry++; - } + ixl_add_sysctls_mac_stats(ctx, child, pf_stats); } void @@ -1986,7 +1866,7 @@ ixl_setup_vlan_filters(struct ixl_vsi *vsi) * configured which interferes with filters managed by driver. * Make sure it's removed. */ -static void +void ixl_del_default_hw_filters(struct ixl_vsi *vsi) { struct i40e_aqc_remove_macvlan_element_data e; @@ -2052,8 +1932,6 @@ ixl_add_mc_filter(struct ixl_vsi *vsi, u8 *macaddr) f->flags |= IXL_FILTER_MC; else printf("WARNING: no filter available!!\n"); - - return; } void @@ -2063,8 +1941,10 @@ ixl_reconfigure_filters(struct ixl_vsi *vsi) } /* -** This routine adds macvlan filters -*/ + * This routine adds a MAC/VLAN filter to the software filter + * list, then adds that new filter to the HW if it doesn't already + * exist in the SW filter list. + */ void ixl_add_filter(struct ixl_vsi *vsi, const u8 *macaddr, s16 vlan) { @@ -2104,8 +1984,8 @@ ixl_add_filter(struct ixl_vsi *vsi, const u8 *macaddr, s16 vlan) else vsi->num_macs++; + f->flags |= IXL_FILTER_USED; ixl_add_hw_filters(vsi, f->flags, 1); - return; } void @@ -2165,12 +2045,15 @@ ixl_add_hw_filters(struct ixl_vsi *vsi, int flags, int cnt) enum i40e_status_code status; int j = 0; - MPASS(cnt > 0); - pf = vsi->back; - dev = iflib_get_dev(vsi->ctx); + dev = vsi->dev; hw = &pf->hw; + if (cnt < 1) { + ixl_dbg_info(pf, "ixl_add_hw_filters: cnt == 0\n"); + return; + } + a = malloc(sizeof(struct i40e_aqc_add_macvlan_element_data) * cnt, M_DEVBUF, M_NOWAIT | M_ZERO); if (a == NULL) { @@ -2197,6 +2080,9 @@ ixl_add_hw_filters(struct ixl_vsi *vsi, int flags, int cnt) b->flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH; f->flags &= ~IXL_FILTER_ADD; j++; + + ixl_dbg_filter(pf, "ADD: " MAC_FORMAT "\n", + MAC_FORMAT_ARGS(f->macaddr)); } if (j == cnt) break; @@ -2232,7 +2118,7 @@ ixl_del_hw_filters(struct ixl_vsi *vsi, int cnt) pf = vsi->back; hw = &pf->hw; - dev = iflib_get_dev(vsi->ctx); + dev = vsi->dev; d = malloc(sizeof(struct i40e_aqc_remove_macvlan_element_data) * cnt, M_DEVBUF, M_NOWAIT | M_ZERO); @@ -2252,6 +2138,10 @@ ixl_del_hw_filters(struct ixl_vsi *vsi, int cnt) } else { e->vlan_tag = f->vlan; } + + ixl_dbg_filter(pf, "DEL: " MAC_FORMAT "\n", + MAC_FORMAT_ARGS(f->macaddr)); + /* delete entry from vsi list */ SLIST_REMOVE(&vsi->ftl, f, ixl_mac_filter, next); free(f, M_DEVBUF); @@ -2456,30 +2346,22 @@ ixl_disable_ring(struct ixl_pf *pf, struct ixl_pf_qtag *qtag, u16 vsi_qidx) return (error); } -/* For PF VSI only */ int -ixl_disable_rings(struct ixl_vsi *vsi) +ixl_disable_rings(struct ixl_pf *pf, struct ixl_vsi *vsi, struct ixl_pf_qtag *qtag) { - struct ixl_pf *pf = vsi->back; - int error = 0; + int error = 0; for (int i = 0; i < vsi->num_tx_queues; i++) - error = ixl_disable_tx_ring(pf, &pf->qtag, i); + error = ixl_disable_tx_ring(pf, qtag, i); for (int i = 0; i < vsi->num_rx_queues; i++) - error = ixl_disable_rx_ring(pf, &pf->qtag, i); + error = ixl_disable_rx_ring(pf, qtag, i); return (error); } -/** - * ixl_handle_mdd_event - * - * Called from interrupt handler to identify possibly malicious vfs - * (But also detects events from the PF, as well) - **/ -void -ixl_handle_mdd_event(struct ixl_pf *pf) +static void +ixl_handle_tx_mdd_event(struct ixl_pf *pf) { struct i40e_hw *hw = &pf->hw; device_t dev = pf->dev; @@ -2487,88 +2369,163 @@ ixl_handle_mdd_event(struct ixl_pf *pf) bool mdd_detected = false; bool pf_mdd_detected = false; bool vf_mdd_detected = false; + u16 vf_num, queue; + u8 pf_num, event; + u8 pf_mdet_num, vp_mdet_num; u32 reg; /* find what triggered the MDD event */ reg = rd32(hw, I40E_GL_MDET_TX); if (reg & I40E_GL_MDET_TX_VALID_MASK) { - u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >> - I40E_GL_MDET_TX_PF_NUM_SHIFT; - u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >> - I40E_GL_MDET_TX_EVENT_SHIFT; - u16 queue = (reg & I40E_GL_MDET_TX_QUEUE_MASK) >> - I40E_GL_MDET_TX_QUEUE_SHIFT; - device_printf(dev, - "Malicious Driver Detection event %d" - " on TX queue %d, pf number %d\n", - event, queue, pf_num); + pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >> + I40E_GL_MDET_TX_PF_NUM_SHIFT; + vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >> + I40E_GL_MDET_TX_VF_NUM_SHIFT; + event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >> + I40E_GL_MDET_TX_EVENT_SHIFT; + queue = (reg & I40E_GL_MDET_TX_QUEUE_MASK) >> + I40E_GL_MDET_TX_QUEUE_SHIFT; wr32(hw, I40E_GL_MDET_TX, 0xffffffff); mdd_detected = true; } - reg = rd32(hw, I40E_GL_MDET_RX); - if (reg & I40E_GL_MDET_RX_VALID_MASK) { - u8 pf_num = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >> - I40E_GL_MDET_RX_FUNCTION_SHIFT; - u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >> - I40E_GL_MDET_RX_EVENT_SHIFT; - u16 queue = (reg & I40E_GL_MDET_RX_QUEUE_MASK) >> - I40E_GL_MDET_RX_QUEUE_SHIFT; - device_printf(dev, - "Malicious Driver Detection event %d" - " on RX queue %d, pf number %d\n", - event, queue, pf_num); - wr32(hw, I40E_GL_MDET_RX, 0xffffffff); - mdd_detected = true; - } - if (mdd_detected) { - reg = rd32(hw, I40E_PF_MDET_TX); - if (reg & I40E_PF_MDET_TX_VALID_MASK) { - wr32(hw, I40E_PF_MDET_TX, 0xFFFF); - device_printf(dev, - "MDD TX event is for this function!\n"); - pf_mdd_detected = true; - } - reg = rd32(hw, I40E_PF_MDET_RX); - if (reg & I40E_PF_MDET_RX_VALID_MASK) { - wr32(hw, I40E_PF_MDET_RX, 0xFFFF); - device_printf(dev, - "MDD RX event is for this function!\n"); - pf_mdd_detected = true; - } - } + if (!mdd_detected) + return; - if (pf_mdd_detected) { - atomic_set_32(&pf->state, IXL_PF_STATE_PF_RESET_REQ); - goto end; + reg = rd32(hw, I40E_PF_MDET_TX); + if (reg & I40E_PF_MDET_TX_VALID_MASK) { + wr32(hw, I40E_PF_MDET_TX, 0xFFFF); + pf_mdet_num = hw->pf_id; + pf_mdd_detected = true; } - // Handle VF detection - for (int i = 0; i < pf->num_vfs && mdd_detected; i++) { + /* Check if MDD was caused by a VF */ + for (int i = 0; i < pf->num_vfs; i++) { vf = &(pf->vfs[i]); reg = rd32(hw, I40E_VP_MDET_TX(i)); if (reg & I40E_VP_MDET_TX_VALID_MASK) { wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF); + vp_mdet_num = i; vf->num_mdd_events++; - device_printf(dev, "MDD TX event is for VF %d\n", i); vf_mdd_detected = true; } + } + + /* Print out an error message */ + if (vf_mdd_detected && pf_mdd_detected) + device_printf(dev, + "Malicious Driver Detection event %d" + " on TX queue %d, pf number %d (PF-%d), vf number %d (VF-%d)\n", + event, queue, pf_num, pf_mdet_num, vf_num, vp_mdet_num); + else if (vf_mdd_detected && !pf_mdd_detected) + device_printf(dev, + "Malicious Driver Detection event %d" + " on TX queue %d, pf number %d, vf number %d (VF-%d)\n", + event, queue, pf_num, vf_num, vp_mdet_num); + else if (!vf_mdd_detected && pf_mdd_detected) + device_printf(dev, + "Malicious Driver Detection event %d" + " on TX queue %d, pf number %d (PF-%d)\n", + event, queue, pf_num, pf_mdet_num); + /* Theoretically shouldn't happen */ + else + device_printf(dev, + "TX Malicious Driver Detection event (unknown)\n"); +} + +static void +ixl_handle_rx_mdd_event(struct ixl_pf *pf) +{ + struct i40e_hw *hw = &pf->hw; + device_t dev = pf->dev; + struct ixl_vf *vf; + bool mdd_detected = false; + bool pf_mdd_detected = false; + bool vf_mdd_detected = false; + u16 queue; + u8 pf_num, event; + u8 pf_mdet_num, vp_mdet_num; + u32 reg; + + /* + * GL_MDET_RX doesn't contain VF number information, unlike + * GL_MDET_TX. + */ + reg = rd32(hw, I40E_GL_MDET_RX); + if (reg & I40E_GL_MDET_RX_VALID_MASK) { + pf_num = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >> + I40E_GL_MDET_RX_FUNCTION_SHIFT; + event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >> + I40E_GL_MDET_RX_EVENT_SHIFT; + queue = (reg & I40E_GL_MDET_RX_QUEUE_MASK) >> + I40E_GL_MDET_RX_QUEUE_SHIFT; + wr32(hw, I40E_GL_MDET_RX, 0xffffffff); + mdd_detected = true; + } + + if (!mdd_detected) + return; + + reg = rd32(hw, I40E_PF_MDET_RX); + if (reg & I40E_PF_MDET_RX_VALID_MASK) { + wr32(hw, I40E_PF_MDET_RX, 0xFFFF); + pf_mdet_num = hw->pf_id; + pf_mdd_detected = true; + } + /* Check if MDD was caused by a VF */ + for (int i = 0; i < pf->num_vfs; i++) { + vf = &(pf->vfs[i]); reg = rd32(hw, I40E_VP_MDET_RX(i)); if (reg & I40E_VP_MDET_RX_VALID_MASK) { wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF); + vp_mdet_num = i; vf->num_mdd_events++; - device_printf(dev, "MDD RX event is for VF %d\n", i); vf_mdd_detected = true; } - - // TODO: Disable VF if there are too many MDD events from it } - if (vf_mdd_detected) - atomic_set_32(&pf->state, IXL_PF_STATE_VF_RESET_REQ); + /* Print out an error message */ + if (vf_mdd_detected && pf_mdd_detected) + device_printf(dev, + "Malicious Driver Detection event %d" + " on RX queue %d, pf number %d (PF-%d), (VF-%d)\n", + event, queue, pf_num, pf_mdet_num, vp_mdet_num); + else if (vf_mdd_detected && !pf_mdd_detected) + device_printf(dev, + "Malicious Driver Detection event %d" + " on RX queue %d, pf number %d, (VF-%d)\n", + event, queue, pf_num, vp_mdet_num); + else if (!vf_mdd_detected && pf_mdd_detected) + device_printf(dev, + "Malicious Driver Detection event %d" + " on RX queue %d, pf number %d (PF-%d)\n", + event, queue, pf_num, pf_mdet_num); + /* Theoretically shouldn't happen */ + else + device_printf(dev, + "RX Malicious Driver Detection event (unknown)\n"); +} + +/** + * ixl_handle_mdd_event + * + * Called from interrupt handler to identify possibly malicious vfs + * (But also detects events from the PF, as well) + **/ +void +ixl_handle_mdd_event(struct ixl_pf *pf) +{ + struct i40e_hw *hw = &pf->hw; + u32 reg; + + /* + * Handle both TX/RX because it's possible they could + * both trigger in the same interrupt. + */ + ixl_handle_tx_mdd_event(pf); + ixl_handle_rx_mdd_event(pf); -end: atomic_clear_32(&pf->state, IXL_PF_STATE_MDD_PENDING); /* re-enable mdd interrupt cause */ @@ -2578,14 +2535,12 @@ end: ixl_flush(hw); } -/* This only enables HW interrupts for the RX queues */ void ixl_enable_intr(struct ixl_vsi *vsi) { struct i40e_hw *hw = vsi->hw; struct ixl_rx_queue *que = vsi->rx_queues; - // TODO: Check iflib interrupt mode instead? if (vsi->shared->isc_intr == IFLIB_INTR_MSIX) { for (int i = 0; i < vsi->num_rx_queues; i++, que++) ixl_enable_queue(hw, que->rxr.me); @@ -2845,7 +2800,7 @@ ixl_prepare_for_reset(struct ixl_pf *pf, bool is_up) } int -ixl_rebuild_hw_structs_after_reset(struct ixl_pf *pf, bool is_up) +ixl_rebuild_hw_structs_after_reset(struct ixl_pf *pf) { struct i40e_hw *hw = &pf->hw; struct ixl_vsi *vsi = &pf->vsi; @@ -2902,6 +2857,25 @@ ixl_rebuild_hw_structs_after_reset(struct ixl_pf *pf, bool is_up) if (error) { device_printf(dev, "ixl_rebuild_hw_structs_after_reset: ixl_switch_config() failed: %d\n", error); + error = EIO; + goto ixl_rebuild_hw_structs_after_reset_err; + } + + error = i40e_aq_set_phy_int_mask(hw, IXL_DEFAULT_PHY_INT_MASK, + NULL); + if (error) { + device_printf(dev, "init: i40e_aq_set_phy_mask() failed: err %d," + " aq_err %d\n", error, hw->aq.asq_last_status); + error = EIO; + goto ixl_rebuild_hw_structs_after_reset_err; + } + + u8 set_fc_err_mask; + error = i40e_set_fc(hw, &set_fc_err_mask, true); + if (error) { + device_printf(dev, "init: setting link flow control failed; retcode %d," + " fc_err_mask 0x%02x\n", error, set_fc_err_mask); + error = EIO; goto ixl_rebuild_hw_structs_after_reset_err; } @@ -2954,7 +2928,7 @@ ixl_handle_empr_reset(struct ixl_pf *pf) ixl_dbg(pf, IXL_DBG_INFO, "Reset wait count: %d\n", count); - ixl_rebuild_hw_structs_after_reset(pf, is_up); + ixl_rebuild_hw_structs_after_reset(pf); atomic_clear_int(&pf->state, IXL_PF_STATE_ADAPTER_RESETTING); } @@ -3314,12 +3288,6 @@ ixl_add_device_sysctls(struct ixl_pf *pf) OID_AUTO, "read_i2c_diag_data", CTLTYPE_STRING | CTLFLAG_RD, pf, 0, ixl_sysctl_read_i2c_diag_data, "A", "Dump selected diagnostic data from FW"); } - -#ifdef PCI_IOV - SYSCTL_ADD_UINT(ctx, debug_list, - OID_AUTO, "vc_debug_level", CTLFLAG_RW, &pf->vc_debug_lvl, - 0, "PF/VF Virtual Channel debug level"); -#endif } /* @@ -3332,9 +3300,7 @@ ixl_sysctl_unallocated_queues(SYSCTL_HANDLER_ARGS) struct ixl_pf *pf = (struct ixl_pf *)arg1; int queues; - //IXL_PF_LOCK(pf); queues = (int)ixl_pf_qmgr_get_num_free(&pf->qmgr); - //IXL_PF_UNLOCK(pf); return sysctl_handle_int(oidp, NULL, queues, req); } @@ -3998,44 +3964,72 @@ ixl_sysctl_sw_filter_list(SYSCTL_HANDLER_ARGS) struct ixl_pf *pf = (struct ixl_pf *)arg1; struct ixl_vsi *vsi = &pf->vsi; struct ixl_mac_filter *f; - char *buf, *buf_i; + device_t dev = pf->dev; + int error = 0, ftl_len = 0, ftl_counter = 0; - int error = 0; - int ftl_len = 0; - int ftl_counter = 0; - int buf_len = 0; - int entry_len = 42; + struct sbuf *buf; - SLIST_FOREACH(f, &vsi->ftl, next) { - ftl_len++; + buf = sbuf_new_for_sysctl(NULL, NULL, 128, req); + if (!buf) { + device_printf(dev, "Could not allocate sbuf for output.\n"); + return (ENOMEM); } - if (ftl_len < 1) { - sysctl_handle_string(oidp, "(none)", 6, req); - return (0); - } + sbuf_printf(buf, "\n"); - buf_len = sizeof(char) * (entry_len + 1) * ftl_len + 2; - buf = buf_i = malloc(buf_len, M_DEVBUF, M_WAITOK); + /* Print MAC filters */ + sbuf_printf(buf, "PF Filters:\n"); + SLIST_FOREACH(f, &vsi->ftl, next) + ftl_len++; - sprintf(buf_i++, "\n"); - SLIST_FOREACH(f, &vsi->ftl, next) { - sprintf(buf_i, - MAC_FORMAT ", vlan %4d, flags %#06x", - MAC_FORMAT_ARGS(f->macaddr), f->vlan, f->flags); - buf_i += entry_len; - /* don't print '\n' for last entry */ - if (++ftl_counter != ftl_len) { - sprintf(buf_i, "\n"); - buf_i++; + if (ftl_len < 1) + sbuf_printf(buf, "(none)\n"); + else { + SLIST_FOREACH(f, &vsi->ftl, next) { + sbuf_printf(buf, + MAC_FORMAT ", vlan %4d, flags %#06x", + MAC_FORMAT_ARGS(f->macaddr), f->vlan, f->flags); + /* don't print '\n' for last entry */ + if (++ftl_counter != ftl_len) + sbuf_printf(buf, "\n"); } } - error = sysctl_handle_string(oidp, buf, strlen(buf), req); +#ifdef PCI_IOV + /* TODO: Give each VF its own filter list sysctl */ + struct ixl_vf *vf; + if (pf->num_vfs > 0) { + sbuf_printf(buf, "\n\n"); + for (int i = 0; i < pf->num_vfs; i++) { + vf = &pf->vfs[i]; + if (!(vf->vf_flags & VF_FLAG_ENABLED)) + continue; + + vsi = &vf->vsi; + ftl_len = 0, ftl_counter = 0; + sbuf_printf(buf, "VF-%d Filters:\n", vf->vf_num); + SLIST_FOREACH(f, &vsi->ftl, next) + ftl_len++; + + if (ftl_len < 1) + sbuf_printf(buf, "(none)\n"); + else { + SLIST_FOREACH(f, &vsi->ftl, next) { + sbuf_printf(buf, + MAC_FORMAT ", vlan %4d, flags %#06x\n", + MAC_FORMAT_ARGS(f->macaddr), f->vlan, f->flags); + } + } + } + } +#endif + + error = sbuf_finish(buf); if (error) - printf("sysctl error: %d\n", error); - free(buf, M_DEVBUF); - return error; + device_printf(dev, "Error finishing sbuf: %d\n", error); + sbuf_delete(buf); + + return (error); } #define IXL_SW_RES_SIZE 0x14 diff --git a/sys/dev/ixl/ixl_pf_qmgr.c b/sys/dev/ixl/ixl_pf_qmgr.c index 1871a99e3252..92ce8d61bd13 100644 --- a/sys/dev/ixl/ixl_pf_qmgr.c +++ b/sys/dev/ixl/ixl_pf_qmgr.c @@ -45,7 +45,7 @@ ixl_pf_qmgr_init(struct ixl_pf_qmgr *qmgr, u16 num_queues) qmgr->num_queues = num_queues; qmgr->qinfo = malloc(num_queues * sizeof(struct ixl_pf_qmgr_qinfo), - M_IXL, M_ZERO | M_WAITOK); + M_IXL, M_ZERO | M_NOWAIT); if (qmgr->qinfo == NULL) return ENOMEM; @@ -266,13 +266,29 @@ ixl_pf_qmgr_is_queue_configured(struct ixl_pf_qtag *qtag, u16 vsi_qidx, bool tx) return (qmgr->qinfo[pf_qidx].rx_configured); } +void +ixl_pf_qmgr_clear_queue_flags(struct ixl_pf_qtag *qtag) +{ + MPASS(qtag != NULL); + + struct ixl_pf_qmgr *qmgr = qtag->qmgr; + for (u16 i = 0; i < qtag->num_allocated; i++) { + u16 pf_qidx = ixl_pf_qidx_from_vsi_qidx(qtag, i); + + qmgr->qinfo[pf_qidx].tx_configured = 0; + qmgr->qinfo[pf_qidx].rx_configured = 0; + qmgr->qinfo[pf_qidx].rx_enabled = 0; + qmgr->qinfo[pf_qidx].tx_enabled = 0; + } +} + u16 ixl_pf_qidx_from_vsi_qidx(struct ixl_pf_qtag *qtag, u16 index) { MPASS(index < qtag->num_allocated); if (qtag->type == IXL_PF_QALLOC_CONTIGUOUS) - return qtag->qidx[0] + index; + return qtag->first_qidx + index; else return qtag->qidx[index]; } diff --git a/sys/dev/ixl/ixl_pf_qmgr.h b/sys/dev/ixl/ixl_pf_qmgr.h index e399afc76105..670a277ff6a5 100644 --- a/sys/dev/ixl/ixl_pf_qmgr.h +++ b/sys/dev/ixl/ixl_pf_qmgr.h @@ -53,11 +53,11 @@ /* Manager */ struct ixl_pf_qmgr_qinfo { - bool allocated; - bool tx_enabled; - bool rx_enabled; - bool tx_configured; - bool rx_configured; + u8 allocated; + u8 tx_enabled; + u8 rx_enabled; + u8 tx_configured; + u8 rx_configured; }; struct ixl_pf_qmgr { @@ -74,7 +74,10 @@ enum ixl_pf_qmgr_qalloc_type { struct ixl_pf_qtag { struct ixl_pf_qmgr *qmgr; enum ixl_pf_qmgr_qalloc_type type; - u16 qidx[IXL_MAX_SCATTERED_QUEUES]; + union { + u16 qidx[IXL_MAX_SCATTERED_QUEUES]; + u16 first_qidx; + }; u16 num_allocated; u16 num_active; }; @@ -101,6 +104,7 @@ void ixl_pf_qmgr_mark_queue_disabled(struct ixl_pf_qtag *qtag, u16 vsi_qidx, boo void ixl_pf_qmgr_mark_queue_configured(struct ixl_pf_qtag *qtag, u16 vsi_qidx, bool tx); bool ixl_pf_qmgr_is_queue_enabled(struct ixl_pf_qtag *qtag, u16 vsi_qidx, bool tx); bool ixl_pf_qmgr_is_queue_configured(struct ixl_pf_qtag *qtag, u16 vsi_qidx, bool tx); +void ixl_pf_qmgr_clear_queue_flags(struct ixl_pf_qtag *qtag); /* Public tag functions */ u16 ixl_pf_qidx_from_vsi_qidx(struct ixl_pf_qtag *qtag, u16 index); diff --git a/sys/dev/ixl/ixl_txrx.c b/sys/dev/ixl/ixl_txrx.c index 3994e16a3d16..4966725426de 100644 --- a/sys/dev/ixl/ixl_txrx.c +++ b/sys/dev/ixl/ixl_txrx.c @@ -65,8 +65,6 @@ static int ixl_isc_rxd_available(void *arg, uint16_t rxqid, qidx_t idx, qidx_t budget); static int ixl_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri); -extern int ixl_intr(void *arg); - struct if_txrx ixl_txrx_hwb = { ixl_isc_txd_encap, ixl_isc_txd_flush, @@ -75,7 +73,7 @@ struct if_txrx ixl_txrx_hwb = { ixl_isc_rxd_pkt_get, ixl_isc_rxd_refill, ixl_isc_rxd_flush, - ixl_intr + NULL }; struct if_txrx ixl_txrx_dwb = { @@ -86,7 +84,7 @@ struct if_txrx ixl_txrx_dwb = { ixl_isc_rxd_pkt_get, ixl_isc_rxd_refill, ixl_isc_rxd_flush, - ixl_intr + NULL }; /* @@ -133,6 +131,21 @@ i40e_vc_stat_str(struct i40e_hw *hw, enum virtchnl_status_code stat_err) return hw->err_str; } +void +ixl_debug_core(device_t dev, u32 enabled_mask, u32 mask, char *fmt, ...) +{ + va_list args; + + if (!(mask & enabled_mask)) + return; + + /* Re-implement device_printf() */ + device_print_prettyname(dev); + va_start(args, fmt); + vprintf(fmt, args); + va_end(args); +} + static bool ixl_is_tx_desc_done(struct tx_ring *txr, int idx) { @@ -236,6 +249,8 @@ ixl_tx_setup_offload(struct ixl_tx_queue *que, *cmd |= I40E_TX_DESC_CMD_L4T_EOFT_TCP; *off |= (pi->ipi_tcp_hlen >> 2) << I40E_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; + /* Check for NO_HEAD MDD event */ + MPASS(pi->ipi_tcp_hlen != 0); } break; case IPPROTO_UDP: @@ -268,23 +283,37 @@ ixl_tso_setup(struct tx_ring *txr, if_pkt_info_t pi) if_softc_ctx_t scctx; struct i40e_tx_context_desc *TXD; u32 cmd, mss, type, tsolen; - int idx; + int idx, total_hdr_len; u64 type_cmd_tso_mss; idx = pi->ipi_pidx; TXD = (struct i40e_tx_context_desc *) &txr->tx_base[idx]; - tsolen = pi->ipi_len - (pi->ipi_ehdrlen + pi->ipi_ip_hlen + pi->ipi_tcp_hlen); + total_hdr_len = pi->ipi_ehdrlen + pi->ipi_ip_hlen + pi->ipi_tcp_hlen; + tsolen = pi->ipi_len - total_hdr_len; scctx = txr->que->vsi->shared; type = I40E_TX_DESC_DTYPE_CONTEXT; cmd = I40E_TX_CTX_DESC_TSO; - /* TSO MSS must not be less than 64 */ + /* + * TSO MSS must not be less than 64; this prevents a + * BAD_LSO_MSS MDD event when the MSS is too small. + */ if (pi->ipi_tso_segsz < IXL_MIN_TSO_MSS) { txr->mss_too_small++; pi->ipi_tso_segsz = IXL_MIN_TSO_MSS; } mss = pi->ipi_tso_segsz; + /* Check for BAD_LS0_MSS MDD event (mss too large) */ + MPASS(mss <= IXL_MAX_TSO_MSS); + /* Check for NO_HEAD MDD event (header lengths are 0) */ + MPASS(pi->ipi_ehdrlen != 0); + MPASS(pi->ipi_ip_hlen != 0); + /* Partial check for BAD_LSO_LEN MDD event */ + MPASS(tsolen != 0); + /* Partial check for WRONG_SIZE MDD event (during TSO) */ + MPASS(total_hdr_len + mss <= IXL_MAX_FRAME); + type_cmd_tso_mss = ((u64)type << I40E_TXD_CTX_QW1_DTYPE_SHIFT) | ((u64)cmd << I40E_TXD_CTX_QW1_CMD_SHIFT) | ((u64)tsolen << I40E_TXD_CTX_QW1_TSO_LEN_SHIFT) | @@ -319,20 +348,16 @@ ixl_isc_txd_encap(void *arg, if_pkt_info_t pi) int i, j, mask, pidx_last; u32 cmd, off, tx_intr; - // device_printf(iflib_get_dev(vsi->ctx), "%s: begin\n", __func__); - cmd = off = 0; i = pi->ipi_pidx; tx_intr = (pi->ipi_flags & IPI_TX_INTR); -#if 0 - device_printf(iflib_get_dev(vsi->ctx), "%s: tx_intr %d\n", __func__, tx_intr); -#endif /* Set up the TSO/CSUM offload */ if (pi->ipi_csum_flags & CSUM_OFFLOAD) { /* Set up the TSO context descriptor if required */ if (pi->ipi_csum_flags & CSUM_TSO) { + /* Prevent MAX_BUFF MDD event (for TSO) */ if (ixl_tso_detect_sparse(segs, nsegs, pi)) return (EFBIG); i = ixl_tso_setup(txr, pi); @@ -344,12 +369,21 @@ ixl_isc_txd_encap(void *arg, if_pkt_info_t pi) cmd |= I40E_TX_DESC_CMD_ICRC; mask = scctx->isc_ntxd[0] - 1; + /* Check for WRONG_SIZE MDD event */ + MPASS(pi->ipi_len >= IXL_MIN_FRAME); +#ifdef INVARIANTS + if (!(pi->ipi_csum_flags & CSUM_TSO)) + MPASS(pi->ipi_len <= IXL_MAX_FRAME); +#endif for (j = 0; j < nsegs; j++) { bus_size_t seglen; txd = &txr->tx_base[i]; seglen = segs[j].ds_len; + /* Check for ZERO_BSIZE MDD event */ + MPASS(seglen != 0); + txd->buffer_addr = htole64(segs[j].ds_addr); txd->cmd_type_offset_bsz = htole64(I40E_TX_DESC_DTYPE_DATA @@ -387,6 +421,8 @@ ixl_isc_txd_flush(void *arg, uint16_t txqid, qidx_t pidx) * Advance the Transmit Descriptor Tail (Tdt), this tells the * hardware that this frame is available to transmit. */ + /* Check for ENDLESS_TX MDD event */ + MPASS(pidx < vsi->shared->isc_ntxd[0]); wr32(vsi->hw, txr->tail, pidx); } @@ -406,9 +442,7 @@ ixl_init_tx_ring(struct ixl_vsi *vsi, struct ixl_tx_queue *que) (sizeof(struct i40e_tx_desc)) * (vsi->shared->isc_ntxd[0] + (vsi->enable_head_writeback ? 1 : 0))); - // TODO: Write max descriptor index instead of 0? wr32(vsi->hw, txr->tail, 0); - wr32(vsi->hw, I40E_QTX_HEAD(txr->me), 0); } /* @@ -470,9 +504,14 @@ ixl_isc_txd_credits_update_dwb(void *arg, uint16_t txqid, bool clear) MPASS(cur != QIDX_INVALID); is_done = ixl_is_tx_desc_done(txr, cur); - if (clear == false || !is_done) + if (!is_done) return (0); + /* If clear is false just let caller know that there + * are descriptors to reclaim */ + if (!clear) + return (1); + prev = txr->tx_cidx_processed; ntxd = scctx->isc_ntxd[0]; do { @@ -547,14 +586,6 @@ ixl_isc_rxd_available(void *arg, uint16_t rxqid, qidx_t idx, qidx_t budget) nrxd = vsi->shared->isc_nrxd[0]; - if (budget == 1) { - rxd = &rxr->rx_base[idx]; - qword = le64toh(rxd->wb.qword1.status_error_len); - status = (qword & I40E_RXD_QW1_STATUS_MASK) - >> I40E_RXD_QW1_STATUS_SHIFT; - return !!(status & (1 << I40E_RX_DESC_STATUS_DD_SHIFT)); - } - for (cnt = 0, i = idx; cnt < nrxd - 1 && cnt <= budget;) { rxd = &rxr->rx_base[i]; qword = le64toh(rxd->wb.qword1.status_error_len); @@ -657,7 +688,7 @@ ixl_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri) MPASS((status & (1 << I40E_RX_DESC_STATUS_DD_SHIFT)) != 0); ri->iri_len += plen; - rxr->bytes += plen; + rxr->rx_bytes += plen; cur->wb.qword1.status_error_len = 0; eop = (status & (1 << I40E_RX_DESC_STATUS_EOF_SHIFT)); @@ -745,25 +776,179 @@ ixl_rx_checksum(if_rxd_info_t ri, u32 status, u32 error, u8 ptype) ri->iri_csum_data |= htons(0xffff); } +/* Set Report Status queue fields to 0 */ +void +ixl_init_tx_rsqs(struct ixl_vsi *vsi) +{ + if_softc_ctx_t scctx = vsi->shared; + struct ixl_tx_queue *tx_que; + int i, j; + + for (i = 0, tx_que = vsi->tx_queues; i < vsi->num_tx_queues; i++, tx_que++) { + struct tx_ring *txr = &tx_que->txr; + + txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0; + + for (j = 0; j < scctx->isc_ntxd[0]; j++) + txr->tx_rsq[j] = QIDX_INVALID; + } +} + +void +ixl_init_tx_cidx(struct ixl_vsi *vsi) +{ + struct ixl_tx_queue *tx_que; + int i; + + for (i = 0, tx_que = vsi->tx_queues; i < vsi->num_tx_queues; i++, tx_que++) { + struct tx_ring *txr = &tx_que->txr; + + txr->tx_cidx_processed = 0; + } +} + /* - * Input: bitmap of enum i40e_aq_link_speed + * Input: bitmap of enum virtchnl_link_speed */ u64 -ixl_max_aq_speed_to_value(u8 link_speeds) +ixl_max_vc_speed_to_value(u8 link_speeds) { - if (link_speeds & I40E_LINK_SPEED_40GB) + if (link_speeds & VIRTCHNL_LINK_SPEED_40GB) return IF_Gbps(40); - if (link_speeds & I40E_LINK_SPEED_25GB) + if (link_speeds & VIRTCHNL_LINK_SPEED_25GB) return IF_Gbps(25); - if (link_speeds & I40E_LINK_SPEED_20GB) + if (link_speeds & VIRTCHNL_LINK_SPEED_20GB) return IF_Gbps(20); - if (link_speeds & I40E_LINK_SPEED_10GB) + if (link_speeds & VIRTCHNL_LINK_SPEED_10GB) return IF_Gbps(10); - if (link_speeds & I40E_LINK_SPEED_1GB) + if (link_speeds & VIRTCHNL_LINK_SPEED_1GB) return IF_Gbps(1); - if (link_speeds & I40E_LINK_SPEED_100MB) + if (link_speeds & VIRTCHNL_LINK_SPEED_100MB) return IF_Mbps(100); else /* Minimum supported link speed */ return IF_Mbps(100); } + +void +ixl_add_vsi_sysctls(device_t dev, struct ixl_vsi *vsi, + struct sysctl_ctx_list *ctx, const char *sysctl_name) +{ + struct sysctl_oid *tree; + struct sysctl_oid_list *child; + struct sysctl_oid_list *vsi_list; + + tree = device_get_sysctl_tree(dev); + child = SYSCTL_CHILDREN(tree); + vsi->vsi_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, sysctl_name, + CTLFLAG_RD, NULL, "VSI Number"); + vsi_list = SYSCTL_CHILDREN(vsi->vsi_node); + + ixl_add_sysctls_eth_stats(ctx, vsi_list, &vsi->eth_stats); +} + +void +ixl_add_sysctls_eth_stats(struct sysctl_ctx_list *ctx, + struct sysctl_oid_list *child, + struct i40e_eth_stats *eth_stats) +{ + struct ixl_sysctl_info ctls[] = + { + {ð_stats->rx_bytes, "good_octets_rcvd", "Good Octets Received"}, + {ð_stats->rx_unicast, "ucast_pkts_rcvd", + "Unicast Packets Received"}, + {ð_stats->rx_multicast, "mcast_pkts_rcvd", + "Multicast Packets Received"}, + {ð_stats->rx_broadcast, "bcast_pkts_rcvd", + "Broadcast Packets Received"}, + {ð_stats->rx_discards, "rx_discards", "Discarded RX packets"}, + {ð_stats->tx_bytes, "good_octets_txd", "Good Octets Transmitted"}, + {ð_stats->tx_unicast, "ucast_pkts_txd", "Unicast Packets Transmitted"}, + {ð_stats->tx_multicast, "mcast_pkts_txd", + "Multicast Packets Transmitted"}, + {ð_stats->tx_broadcast, "bcast_pkts_txd", + "Broadcast Packets Transmitted"}, + // end + {0,0,0} + }; + + struct ixl_sysctl_info *entry = ctls; + while (entry->stat != 0) + { + SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, entry->name, + CTLFLAG_RD, entry->stat, + entry->description); + entry++; + } +} + +void +ixl_add_queues_sysctls(device_t dev, struct ixl_vsi *vsi) +{ + struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); + struct sysctl_oid_list *vsi_list, *queue_list; + struct sysctl_oid *queue_node; + char queue_namebuf[32]; + + struct ixl_rx_queue *rx_que; + struct ixl_tx_queue *tx_que; + struct tx_ring *txr; + struct rx_ring *rxr; + + vsi_list = SYSCTL_CHILDREN(vsi->vsi_node); + + /* Queue statistics */ + for (int q = 0; q < vsi->num_rx_queues; q++) { + bzero(queue_namebuf, sizeof(queue_namebuf)); + snprintf(queue_namebuf, QUEUE_NAME_LEN, "rxq%02d", q); + queue_node = SYSCTL_ADD_NODE(ctx, vsi_list, + OID_AUTO, queue_namebuf, CTLFLAG_RD, NULL, "RX Queue #"); + queue_list = SYSCTL_CHILDREN(queue_node); + + rx_que = &(vsi->rx_queues[q]); + rxr = &(rx_que->rxr); + + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs", + CTLFLAG_RD, &(rx_que->irqs), + "irqs on this queue (both Tx and Rx)"); + + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "packets", + CTLFLAG_RD, &(rxr->rx_packets), + "Queue Packets Received"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "bytes", + CTLFLAG_RD, &(rxr->rx_bytes), + "Queue Bytes Received"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "desc_err", + CTLFLAG_RD, &(rxr->desc_errs), + "Queue Rx Descriptor Errors"); + SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "itr", + CTLFLAG_RD, &(rxr->itr), 0, + "Queue Rx ITR Interval"); + } + for (int q = 0; q < vsi->num_tx_queues; q++) { + bzero(queue_namebuf, sizeof(queue_namebuf)); + snprintf(queue_namebuf, QUEUE_NAME_LEN, "txq%02d", q); + queue_node = SYSCTL_ADD_NODE(ctx, vsi_list, + OID_AUTO, queue_namebuf, CTLFLAG_RD, NULL, "TX Queue #"); + queue_list = SYSCTL_CHILDREN(queue_node); + + tx_que = &(vsi->tx_queues[q]); + txr = &(tx_que->txr); + + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tso", + CTLFLAG_RD, &(tx_que->tso), + "TSO"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "mss_too_small", + CTLFLAG_RD, &(txr->mss_too_small), + "TSO sends with an MSS less than 64"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "packets", + CTLFLAG_RD, &(txr->tx_packets), + "Queue Packets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "bytes", + CTLFLAG_RD, &(txr->tx_bytes), + "Queue Bytes Transmitted"); + SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "itr", + CTLFLAG_RD, &(txr->itr), 0, + "Queue Tx ITR Interval"); + } +} diff --git a/sys/dev/ixl/ixlv.h b/sys/dev/ixl/ixlv.h index 916b887e87d0..9f6f1cb8eefb 100644 --- a/sys/dev/ixl/ixlv.h +++ b/sys/dev/ixl/ixlv.h @@ -36,13 +36,12 @@ #ifndef _IXLV_H_ #define _IXLV_H_ -#include "ixlv_vc_mgr.h" +#include "ixl.h" #define IXLV_AQ_MAX_ERR 200 #define IXLV_MAX_FILTERS 128 #define IXLV_MAX_QUEUES 16 #define IXLV_AQ_TIMEOUT (1 * hz) -#define IXLV_CALLOUT_TIMO (hz / 50) /* 20 msec */ #define IXLV_FLAG_AQ_ENABLE_QUEUES (u32)(1 << 0) #define IXLV_FLAG_AQ_DISABLE_QUEUES (u32)(1 << 1) @@ -58,42 +57,38 @@ #define IXLV_FLAG_AQ_CONFIG_RSS_KEY (u32)(1 << 11) #define IXLV_FLAG_AQ_SET_RSS_HENA (u32)(1 << 12) #define IXLV_FLAG_AQ_GET_RSS_HENA_CAPS (u32)(1 << 13) -#define IXLV_FLAG_AQ_CONFIG_RSS_LUT (u32)(1 << 14) +#define IXLV_FLAG_AQ_CONFIG_RSS_LUT (u32)(1 << 14) -/* printf %b arg */ +/* printf %b flag args */ #define IXLV_FLAGS \ "\20\1ENABLE_QUEUES\2DISABLE_QUEUES\3ADD_MAC_FILTER" \ "\4ADD_VLAN_FILTER\5DEL_MAC_FILTER\6DEL_VLAN_FILTER" \ "\7CONFIGURE_QUEUES\10MAP_VECTORS\11HANDLE_RESET" \ - "\12CONFIGURE_PROMISC\13GET_STATS" + "\12CONFIGURE_PROMISC\13GET_STATS\14CONFIG_RSS_KEY" \ + "\15SET_RSS_HENA\16GET_RSS_HENA_CAPS\17CONFIG_RSS_LUT" #define IXLV_PRINTF_VF_OFFLOAD_FLAGS \ - "\20\1I40E_VIRTCHNL_VF_OFFLOAD_L2" \ - "\2I40E_VIRTCHNL_VF_OFFLOAD_IWARP" \ - "\3I40E_VIRTCHNL_VF_OFFLOAD_FCOE" \ - "\4I40E_VIRTCHNL_VF_OFFLOAD_RSS_AQ" \ - "\5I40E_VIRTCHNL_VF_OFFLOAD_RSS_REG" \ - "\6I40E_VIRTCHNL_VF_OFFLOAD_WB_ON_ITR" \ - "\21I40E_VIRTCHNL_VF_OFFLOAD_VLAN" \ - "\22I40E_VIRTCHNL_VF_OFFLOAD_RX_POLLING" \ - "\23I40E_VIRTCHNL_VF_OFFLOAD_RSS_PCTYPE_V2" \ - "\24I40E_VIRTCHNL_VF_OFFLOAD_RSS_PF" - -static MALLOC_DEFINE(M_IXLV, "ixlv", "ixlv driver allocations"); + "\20\1L2" \ + "\2IWARP" \ + "\3RSVD" \ + "\4RSS_AQ" \ + "\5RSS_REG" \ + "\6WB_ON_ITR" \ + "\7REQ_QUEUES" \ + "\21VLAN" \ + "\22RX_POLLING" \ + "\23RSS_PCTYPE_V2" \ + "\24RSS_PF" \ + "\25ENCAP" \ + "\26ENCAP_CSUM" \ + "\27RX_ENCAP_CSUM" + +MALLOC_DECLARE(M_IXLV); /* Driver state */ enum ixlv_state_t { - IXLV_START, - IXLV_FAILED, IXLV_RESET_REQUIRED, IXLV_RESET_PENDING, - IXLV_VERSION_CHECK, - IXLV_GET_RESOURCES, IXLV_INIT_READY, - IXLV_INIT_START, - IXLV_INIT_CONFIG, - IXLV_INIT_MAPPING, - IXLV_INIT_ENABLE, - IXLV_INIT_COMPLETE, IXLV_RUNNING, }; @@ -115,77 +110,48 @@ SLIST_HEAD(vlan_list, ixlv_vlan_filter); /* Software controller structure */ struct ixlv_sc { + struct ixl_vsi vsi; + struct i40e_hw hw; struct i40e_osdep osdep; device_t dev; struct resource *pci_mem; - struct resource *msix_mem; enum ixlv_state_t init_state; - int init_in_progress; - - /* - * Interrupt resources - */ - void *tag; - struct resource *res; /* For the AQ */ struct ifmedia media; - struct callout timer; - int msix; - int pf_version; - int if_flags; + struct virtchnl_version_info version; + enum ixl_dbg_mask dbg_mask; + u16 promisc_flags; bool link_up; enum virtchnl_link_speed link_speed; - struct mtx mtx; - - u32 qbase; - u32 admvec; - struct timeout_task timeout; -#ifdef notyet - struct task aq_irq; - struct task aq_sched; -#endif - - struct ixl_vsi vsi; + /* Tunable settings */ + int tx_itr; + int rx_itr; + int dynamic_tx_itr; + int dynamic_rx_itr; /* Filter lists */ struct mac_list *mac_filters; struct vlan_list *vlan_filters; - /* Promiscuous mode */ - u32 promiscuous_flags; - - /* Admin queue task flags */ - u32 aq_wait_count; - - struct ixl_vc_mgr vc_mgr; - struct ixl_vc_cmd add_mac_cmd; - struct ixl_vc_cmd del_mac_cmd; - struct ixl_vc_cmd config_queues_cmd; - struct ixl_vc_cmd map_vectors_cmd; - struct ixl_vc_cmd enable_queues_cmd; - struct ixl_vc_cmd add_vlan_cmd; - struct ixl_vc_cmd del_vlan_cmd; - struct ixl_vc_cmd add_multi_cmd; - struct ixl_vc_cmd del_multi_cmd; - struct ixl_vc_cmd config_rss_key_cmd; - struct ixl_vc_cmd get_rss_hena_caps_cmd; - struct ixl_vc_cmd set_rss_hena_cmd; - struct ixl_vc_cmd config_rss_lut_cmd; - /* Virtual comm channel */ struct virtchnl_vf_resource *vf_res; struct virtchnl_vsi_resource *vsi_res; /* Misc stats maintained by the driver */ - u64 watchdog_events; u64 admin_irq; + /* Buffer used for reading AQ responses */ u8 aq_buffer[IXL_AQ_BUF_SZ]; + + /* State flag used in init/stop */ + u32 queues_enabled; + u8 enable_queues_chan; + u8 disable_queues_chan; }; /* @@ -203,6 +169,13 @@ ixlv_check_ether_addr(u8 *addr) return (status); } +/* Debug printing */ +#define ixlv_dbg(sc, m, s, ...) ixl_debug_core(sc->dev, sc->dbg_mask, m, s, ##__VA_ARGS__) +#define ixlv_dbg_init(sc, s, ...) ixl_debug_core(sc->dev, sc->dbg_mask, IXLV_DBG_INIT, s, ##__VA_ARGS__) +#define ixlv_dbg_info(sc, s, ...) ixl_debug_core(sc->dev, sc->dbg_mask, IXLV_DBG_INFO, s, ##__VA_ARGS__) +#define ixlv_dbg_vc(sc, s, ...) ixl_debug_core(sc->dev, sc->dbg_mask, IXLV_DBG_VC, s, ##__VA_ARGS__) +#define ixlv_dbg_filter(sc, s, ...) ixl_debug_core(sc->dev, sc->dbg_mask, IXLV_DBG_FILTER, s, ##__VA_ARGS__) + /* ** VF Common function prototypes */ @@ -214,28 +187,32 @@ int ixlv_send_vf_config_msg(struct ixlv_sc *); int ixlv_get_vf_config(struct ixlv_sc *); void ixlv_init(void *); int ixlv_reinit_locked(struct ixlv_sc *); -void ixlv_configure_queues(struct ixlv_sc *); -void ixlv_enable_queues(struct ixlv_sc *); -void ixlv_disable_queues(struct ixlv_sc *); -void ixlv_map_queues(struct ixlv_sc *); +int ixlv_configure_queues(struct ixlv_sc *); +int ixlv_enable_queues(struct ixlv_sc *); +int ixlv_disable_queues(struct ixlv_sc *); +int ixlv_map_queues(struct ixlv_sc *); void ixlv_enable_intr(struct ixl_vsi *); void ixlv_disable_intr(struct ixl_vsi *); -void ixlv_add_ether_filters(struct ixlv_sc *); -void ixlv_del_ether_filters(struct ixlv_sc *); -void ixlv_request_stats(struct ixlv_sc *); -void ixlv_request_reset(struct ixlv_sc *); +int ixlv_add_ether_filters(struct ixlv_sc *); +int ixlv_del_ether_filters(struct ixlv_sc *); +int ixlv_request_stats(struct ixlv_sc *); +int ixlv_request_reset(struct ixlv_sc *); void ixlv_vc_completion(struct ixlv_sc *, enum virtchnl_ops, enum virtchnl_status_code, u8 *, u16); -void ixlv_add_ether_filter(struct ixlv_sc *); -void ixlv_add_vlans(struct ixlv_sc *); -void ixlv_del_vlans(struct ixlv_sc *); +int ixlv_add_ether_filter(struct ixlv_sc *); +int ixlv_add_vlans(struct ixlv_sc *); +int ixlv_del_vlans(struct ixlv_sc *); void ixlv_update_stats_counters(struct ixlv_sc *, struct i40e_eth_stats *); void ixlv_update_link_status(struct ixlv_sc *); -void ixlv_get_default_rss_key(u32 *, bool); -void ixlv_config_rss_key(struct ixlv_sc *); -void ixlv_set_rss_hena(struct ixlv_sc *); -void ixlv_config_rss_lut(struct ixlv_sc *); - +int ixlv_get_default_rss_key(u32 *, bool); +int ixlv_config_rss_key(struct ixlv_sc *); +int ixlv_set_rss_hena(struct ixlv_sc *); +int ixlv_config_rss_lut(struct ixlv_sc *); +int ixlv_config_promisc_mode(struct ixlv_sc *); + +int ixl_vc_send_cmd(struct ixlv_sc *sc, uint32_t request); +char *ixlv_vc_speed_to_string(enum virtchnl_link_speed link_speed); +void *ixl_vc_get_op_chan(struct ixlv_sc *sc, uint32_t request); #endif /* _IXLV_H_ */ diff --git a/sys/dev/ixl/ixlv_vc_mgr.h b/sys/dev/ixl/ixlv_vc_mgr.h deleted file mode 100644 index 424348b5b595..000000000000 --- a/sys/dev/ixl/ixlv_vc_mgr.h +++ /dev/null @@ -1,76 +0,0 @@ -/****************************************************************************** - - Copyright (c) 2013-2018, Intel Corporation - All rights reserved. - - Redistribution and use in source and binary forms, with or without - modification, are permitted provided that the following conditions are met: - - 1. Redistributions of source code must retain the above copyright notice, - this list of conditions and the following disclaimer. - - 2. Redistributions in binary form must reproduce the above copyright - notice, this list of conditions and the following disclaimer in the - documentation and/or other materials provided with the distribution. - - 3. Neither the name of the Intel Corporation nor the names of its - contributors may be used to endorse or promote products derived from - this software without specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" - AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE - IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE - ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE - LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR - CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF - SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS - INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN - CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) - ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE - POSSIBILITY OF SUCH DAMAGE. - -******************************************************************************/ -/*$FreeBSD$*/ - -#ifndef _IXLV_VC_MGR_H_ -#define _IXLV_VC_MGR_H_ - -#include <sys/queue.h> - -struct ixl_vc_cmd; - -typedef void ixl_vc_callback_t(struct ixl_vc_cmd *, void *, - enum i40e_status_code); - - -#define IXLV_VC_CMD_FLAG_BUSY 0x0001 - -struct ixl_vc_cmd -{ - uint32_t request; - uint32_t flags; - - ixl_vc_callback_t *callback; - void *arg; - - TAILQ_ENTRY(ixl_vc_cmd) next; -}; - -struct ixl_vc_mgr -{ - struct ixlv_sc *sc; - struct ixl_vc_cmd *current; - struct callout callout; - - TAILQ_HEAD(, ixl_vc_cmd) pending; -}; - -#define IXLV_VC_TIMEOUT (2 * hz) - -void ixl_vc_init_mgr(struct ixlv_sc *, struct ixl_vc_mgr *); -void ixl_vc_enqueue(struct ixl_vc_mgr *, struct ixl_vc_cmd *, - uint32_t, ixl_vc_callback_t *, void *); -void ixl_vc_flush(struct ixl_vc_mgr *mgr); - -#endif - diff --git a/sys/dev/ixl/ixlvc.c b/sys/dev/ixl/ixlvc.c index 12a21ba35933..9a33310f0e75 100644 --- a/sys/dev/ixl/ixlvc.c +++ b/sys/dev/ixl/ixlvc.c @@ -40,118 +40,11 @@ #include "ixl.h" #include "ixlv.h" -#include "i40e_prototype.h" - /* busy wait delay in msec */ #define IXLV_BUSY_WAIT_DELAY 10 #define IXLV_BUSY_WAIT_COUNT 50 -static void ixl_vc_process_resp(struct ixl_vc_mgr *, uint32_t, - enum virtchnl_status_code); -static void ixl_vc_process_next(struct ixl_vc_mgr *mgr); -static void ixl_vc_schedule_retry(struct ixl_vc_mgr *mgr); -static void ixl_vc_send_current(struct ixl_vc_mgr *mgr); - -#ifdef IXL_DEBUG -/* -** Validate VF messages -*/ -static int ixl_vc_validate_vf_msg(struct ixlv_sc *sc, u32 v_opcode, - u8 *msg, u16 msglen) -{ - bool err_msg_format = false; - int valid_len; - - /* Validate message length. */ - switch (v_opcode) { - case VIRTCHNL_OP_VERSION: - valid_len = sizeof(struct virtchnl_version_info); - break; - case VIRTCHNL_OP_RESET_VF: - valid_len = 0; - break; - case VIRTCHNL_OP_GET_VF_RESOURCES: - /* Valid length in api v1.0 is 0, v1.1 is 4 */ - valid_len = 4; - break; - case VIRTCHNL_OP_CONFIG_TX_QUEUE: - valid_len = sizeof(struct virtchnl_txq_info); - break; - case VIRTCHNL_OP_CONFIG_RX_QUEUE: - valid_len = sizeof(struct virtchnl_rxq_info); - break; - case VIRTCHNL_OP_CONFIG_VSI_QUEUES: - valid_len = sizeof(struct virtchnl_vsi_queue_config_info); - if (msglen >= valid_len) { - struct virtchnl_vsi_queue_config_info *vqc = - (struct virtchnl_vsi_queue_config_info *)msg; - valid_len += (vqc->num_queue_pairs * - sizeof(struct - virtchnl_queue_pair_info)); - if (vqc->num_queue_pairs == 0) - err_msg_format = true; - } - break; - case VIRTCHNL_OP_CONFIG_IRQ_MAP: - valid_len = sizeof(struct virtchnl_irq_map_info); - if (msglen >= valid_len) { - struct virtchnl_irq_map_info *vimi = - (struct virtchnl_irq_map_info *)msg; - valid_len += (vimi->num_vectors * - sizeof(struct virtchnl_vector_map)); - if (vimi->num_vectors == 0) - err_msg_format = true; - } - break; - case VIRTCHNL_OP_ENABLE_QUEUES: - case VIRTCHNL_OP_DISABLE_QUEUES: - valid_len = sizeof(struct virtchnl_queue_select); - break; - case VIRTCHNL_OP_ADD_ETH_ADDR: - case VIRTCHNL_OP_DEL_ETH_ADDR: - valid_len = sizeof(struct virtchnl_ether_addr_list); - if (msglen >= valid_len) { - struct virtchnl_ether_addr_list *veal = - (struct virtchnl_ether_addr_list *)msg; - valid_len += veal->num_elements * - sizeof(struct virtchnl_ether_addr); - if (veal->num_elements == 0) - err_msg_format = true; - } - break; - case VIRTCHNL_OP_ADD_VLAN: - case VIRTCHNL_OP_DEL_VLAN: - valid_len = sizeof(struct virtchnl_vlan_filter_list); - if (msglen >= valid_len) { - struct virtchnl_vlan_filter_list *vfl = - (struct virtchnl_vlan_filter_list *)msg; - valid_len += vfl->num_elements * sizeof(u16); - if (vfl->num_elements == 0) - err_msg_format = true; - } - break; - case VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE: - valid_len = sizeof(struct virtchnl_promisc_info); - break; - case VIRTCHNL_OP_GET_STATS: - valid_len = sizeof(struct virtchnl_queue_select); - break; - /* These are always errors coming from the VF. */ - case VIRTCHNL_OP_EVENT: - case VIRTCHNL_OP_UNKNOWN: - default: - return EPERM; - break; - } - /* few more checks */ - if ((valid_len != msglen) || (err_msg_format)) - return EINVAL; - else - return 0; -} -#endif - /* ** ixlv_send_pf_msg ** @@ -161,31 +54,39 @@ static int ixlv_send_pf_msg(struct ixlv_sc *sc, enum virtchnl_ops op, u8 *msg, u16 len) { - struct i40e_hw *hw = &sc->hw; - device_t dev = sc->dev; - i40e_status err; - -#ifdef IXL_DEBUG - /* - ** Pre-validating messages to the PF - */ + struct i40e_hw *hw = &sc->hw; + device_t dev = sc->dev; + i40e_status status; int val_err; - val_err = ixl_vc_validate_vf_msg(sc, op, msg, len); + + /* Validating message before sending it to the PF */ + val_err = virtchnl_vc_validate_vf_msg(&sc->version, op, msg, len); if (val_err) device_printf(dev, "Error validating msg to PF for op %d," " msglen %d: error %d\n", op, len, val_err); -#endif - err = i40e_aq_send_msg_to_pf(hw, op, I40E_SUCCESS, msg, len, NULL); - if (err) + if (!i40e_check_asq_alive(hw)) { + if (op != VIRTCHNL_OP_GET_STATS) + device_printf(dev, "Unable to send opcode %s to PF, " + "ASQ is not alive\n", ixl_vc_opcode_str(op)); + return (0); + } + + if (op != VIRTCHNL_OP_GET_STATS) + ixlv_dbg_vc(sc, + "Sending msg (op=%s[%d]) to PF\n", + ixl_vc_opcode_str(op), op); + + status = i40e_aq_send_msg_to_pf(hw, op, I40E_SUCCESS, msg, len, NULL); + if (status && op != VIRTCHNL_OP_GET_STATS) device_printf(dev, "Unable to send opcode %s to PF, " "status %s, aq error %s\n", ixl_vc_opcode_str(op), - i40e_stat_str(hw, err), + i40e_stat_str(hw, status), i40e_aq_str(hw, hw->aq.asq_last_status)); - return err; -} + return (status); +} /* ** ixlv_send_api_ver @@ -224,11 +125,7 @@ ixlv_verify_api_ver(struct ixlv_sc *sc) int retries = 0; event.buf_len = IXL_AQ_BUF_SZ; - event.msg_buf = malloc(event.buf_len, M_DEVBUF, M_NOWAIT); - if (!event.msg_buf) { - err = ENOMEM; - goto out; - } + event.msg_buf = malloc(event.buf_len, M_IXLV, M_WAITOK); for (;;) { if (++retries > IXLV_AQ_MAX_ERR) @@ -266,8 +163,10 @@ ixlv_verify_api_ver(struct ixlv_sc *sc) (pf_vvi->minor > VIRTCHNL_VERSION_MINOR))) { device_printf(dev, "Critical PF/VF API version mismatch!\n"); err = EIO; - } else - sc->pf_version = pf_vvi->minor; + } else { + sc->version.major = pf_vvi->major; + sc->version.minor = pf_vvi->minor; + } /* Log PF/VF api versions */ device_printf(dev, "PF API %d.%d / VF API %d.%d\n", @@ -275,8 +174,7 @@ ixlv_verify_api_ver(struct ixlv_sc *sc) VIRTCHNL_VERSION_MAJOR, VIRTCHNL_VERSION_MINOR); out_alloc: - free(event.msg_buf, M_DEVBUF); -out: + free(event.msg_buf, M_IXLV); return (err); } @@ -296,7 +194,10 @@ ixlv_send_vf_config_msg(struct ixlv_sc *sc) VIRTCHNL_VF_OFFLOAD_RSS_PF | VIRTCHNL_VF_OFFLOAD_VLAN; - if (sc->pf_version == VIRTCHNL_VERSION_MINOR_NO_VF_CAPS) + ixlv_dbg_info(sc, "Sending offload flags: 0x%b\n", + caps, IXLV_PRINTF_VF_OFFLOAD_FLAGS); + + if (sc->version.minor == VIRTCHNL_VERSION_MINOR_NO_VF_CAPS) return ixlv_send_pf_msg(sc, VIRTCHNL_OP_GET_VF_RESOURCES, NULL, 0); else @@ -326,11 +227,7 @@ ixlv_get_vf_config(struct ixlv_sc *sc) len = sizeof(struct virtchnl_vf_resource) + sizeof(struct virtchnl_vsi_resource); event.buf_len = len; - event.msg_buf = malloc(event.buf_len, M_DEVBUF, M_NOWAIT); - if (!event.msg_buf) { - err = ENOMEM; - goto out; - } + event.msg_buf = malloc(event.buf_len, M_IXLV, M_WAITOK); for (;;) { err = i40e_clean_arq_element(hw, &event, NULL); @@ -371,8 +268,7 @@ ixlv_get_vf_config(struct ixlv_sc *sc) i40e_vf_parse_hw_config(hw, sc->vf_res); out_alloc: - free(event.msg_buf, M_DEVBUF); -out: + free(event.msg_buf, M_IXLV); return err; } @@ -381,7 +277,7 @@ out: ** ** Request that the PF set up our queues. */ -void +int ixlv_configure_queues(struct ixlv_sc *sc) { device_t dev = sc->dev; @@ -401,11 +297,10 @@ ixlv_configure_queues(struct ixlv_sc *sc) pairs = max(vsi->num_tx_queues, vsi->num_rx_queues); len = sizeof(struct virtchnl_vsi_queue_config_info) + (sizeof(struct virtchnl_queue_pair_info) * pairs); - vqci = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO); + vqci = malloc(len, M_IXLV, M_NOWAIT | M_ZERO); if (!vqci) { device_printf(dev, "%s: unable to allocate memory\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (ENOMEM); } vqci->vsi_id = sc->vsi_res->vsi_id; vqci->num_queue_pairs = pairs; @@ -413,6 +308,7 @@ ixlv_configure_queues(struct ixlv_sc *sc) /* Size check is not needed here - HW max is 16 queue pairs, and we * can fit info for 31 of them into the AQ buffer before it overflows. */ + // TODO: the above is wrong now; X722 VFs can have 256 queues for (int i = 0; i < pairs; i++, tx_que++, rx_que++, vqpi++) { txr = &tx_que->txr; rxr = &rx_que->rxr; @@ -422,22 +318,29 @@ ixlv_configure_queues(struct ixlv_sc *sc) vqpi->txq.ring_len = scctx->isc_ntxd[0]; vqpi->txq.dma_ring_addr = txr->tx_paddr; /* Enable Head writeback */ - vqpi->txq.headwb_enabled = 0; - vqpi->txq.dma_headwb_addr = 0; + if (!vsi->enable_head_writeback) { + vqpi->txq.headwb_enabled = 0; + vqpi->txq.dma_headwb_addr = 0; + } else { + vqpi->txq.headwb_enabled = 1; + vqpi->txq.dma_headwb_addr = txr->tx_paddr + + sizeof(struct i40e_tx_desc) * scctx->isc_ntxd[0]; + } vqpi->rxq.vsi_id = vqci->vsi_id; vqpi->rxq.queue_id = i; vqpi->rxq.ring_len = scctx->isc_nrxd[0]; vqpi->rxq.dma_ring_addr = rxr->rx_paddr; vqpi->rxq.max_pkt_size = scctx->isc_max_frame_size; - // TODO: Get this value from iflib, somehow vqpi->rxq.databuffer_size = rxr->mbuf_sz; vqpi->rxq.splithdr_enabled = 0; } ixlv_send_pf_msg(sc, VIRTCHNL_OP_CONFIG_VSI_QUEUES, (u8 *)vqci, len); - free(vqci, M_DEVBUF); + free(vqci, M_IXLV); + + return (0); } /* @@ -445,7 +348,7 @@ ixlv_configure_queues(struct ixlv_sc *sc) ** ** Request that the PF enable all of our queues. */ -void +int ixlv_enable_queues(struct ixlv_sc *sc) { struct virtchnl_queue_select vqs; @@ -453,10 +356,11 @@ ixlv_enable_queues(struct ixlv_sc *sc) vqs.vsi_id = sc->vsi_res->vsi_id; /* XXX: In Linux PF, as long as neither of these is 0, * every queue in VF VSI is enabled. */ - vqs.tx_queues = (1 << sc->vsi_res->num_queue_pairs) - 1; + vqs.tx_queues = (1 << sc->vsi.num_tx_queues) - 1; vqs.rx_queues = vqs.tx_queues; ixlv_send_pf_msg(sc, VIRTCHNL_OP_ENABLE_QUEUES, (u8 *)&vqs, sizeof(vqs)); + return (0); } /* @@ -464,7 +368,7 @@ ixlv_enable_queues(struct ixlv_sc *sc) ** ** Request that the PF disable all of our queues. */ -void +int ixlv_disable_queues(struct ixlv_sc *sc) { struct virtchnl_queue_select vqs; @@ -472,10 +376,11 @@ ixlv_disable_queues(struct ixlv_sc *sc) vqs.vsi_id = sc->vsi_res->vsi_id; /* XXX: In Linux PF, as long as neither of these is 0, * every queue in VF VSI is disabled. */ - vqs.tx_queues = (1 << sc->vsi_res->num_queue_pairs) - 1; + vqs.tx_queues = (1 << sc->vsi.num_tx_queues) - 1; vqs.rx_queues = vqs.tx_queues; ixlv_send_pf_msg(sc, VIRTCHNL_OP_DISABLE_QUEUES, (u8 *)&vqs, sizeof(vqs)); + return (0); } /* @@ -484,7 +389,7 @@ ixlv_disable_queues(struct ixlv_sc *sc) ** Request that the PF map queues to interrupt vectors. Misc causes, including ** admin queue, are always mapped to vector 0. */ -void +int ixlv_map_queues(struct ixlv_sc *sc) { struct virtchnl_irq_map_info *vm; @@ -502,12 +407,11 @@ ixlv_map_queues(struct ixlv_sc *sc) q = scctx->isc_vectors - 1; len = sizeof(struct virtchnl_irq_map_info) + - (scctx->isc_vectors * sizeof(struct i40e_virtchnl_vector_map)); - vm = malloc(len, M_DEVBUF, M_NOWAIT); + (scctx->isc_vectors * sizeof(struct virtchnl_vector_map)); + vm = malloc(len, M_IXLV, M_NOWAIT); if (!vm) { device_printf(dev, "%s: unable to allocate memory\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (ENOMEM); } vm->num_vectors = scctx->isc_vectors; @@ -515,7 +419,8 @@ ixlv_map_queues(struct ixlv_sc *sc) for (i = 0; i < q; i++, rx_que++) { vm->vecmap[i].vsi_id = sc->vsi_res->vsi_id; vm->vecmap[i].vector_id = i + 1; /* first is adminq */ - // vm->vecmap[i].txq_map = (1 << que->me); + // TODO: Re-examine this + vm->vecmap[i].txq_map = (1 << rx_que->rxr.me); vm->vecmap[i].rxq_map = (1 << rx_que->rxr.me); vm->vecmap[i].rxitr_idx = 0; vm->vecmap[i].txitr_idx = 1; @@ -531,7 +436,9 @@ ixlv_map_queues(struct ixlv_sc *sc) ixlv_send_pf_msg(sc, VIRTCHNL_OP_CONFIG_IRQ_MAP, (u8 *)vm, len); - free(vm, M_DEVBUF); + free(vm, M_IXLV); + + return (0); } /* @@ -539,10 +446,10 @@ ixlv_map_queues(struct ixlv_sc *sc) ** to be added, then create the data to hand to the AQ ** for handling. */ -void +int ixlv_add_vlans(struct ixlv_sc *sc) { - struct virtchnl_vlan_filter_list *v; + struct virtchnl_vlan_filter_list *v; struct ixlv_vlan_filter *f, *ftmp; device_t dev = sc->dev; int len, i = 0, cnt = 0; @@ -553,11 +460,8 @@ ixlv_add_vlans(struct ixlv_sc *sc) cnt++; } - if (!cnt) { /* no work... */ - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_ADD_VLAN_FILTER, - VIRTCHNL_STATUS_SUCCESS); - return; - } + if (!cnt) /* no work... */ + return (ENOENT); len = sizeof(struct virtchnl_vlan_filter_list) + (cnt * sizeof(u16)); @@ -565,16 +469,14 @@ ixlv_add_vlans(struct ixlv_sc *sc) if (len > IXL_AQ_BUF_SZ) { device_printf(dev, "%s: Exceeded Max AQ Buf size\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (EFBIG); } - v = malloc(len, M_DEVBUF, M_NOWAIT); + v = malloc(len, M_IXLV, M_NOWAIT); if (!v) { device_printf(dev, "%s: unable to allocate memory\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (ENOMEM); } v->vsi_id = sc->vsi_res->vsi_id; @@ -592,8 +494,9 @@ ixlv_add_vlans(struct ixlv_sc *sc) } ixlv_send_pf_msg(sc, VIRTCHNL_OP_ADD_VLAN, (u8 *)v, len); - free(v, M_DEVBUF); + free(v, M_IXLV); /* add stats? */ + return (0); } /* @@ -601,12 +504,12 @@ ixlv_add_vlans(struct ixlv_sc *sc) ** to be removed, then create the data to hand to the AQ ** for handling. */ -void +int ixlv_del_vlans(struct ixlv_sc *sc) { - device_t dev = sc->dev; struct virtchnl_vlan_filter_list *v; struct ixlv_vlan_filter *f, *ftmp; + device_t dev = sc->dev; int len, i = 0, cnt = 0; /* Get count of VLAN filters to delete */ @@ -615,11 +518,8 @@ ixlv_del_vlans(struct ixlv_sc *sc) cnt++; } - if (!cnt) { /* no work... */ - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_DEL_VLAN_FILTER, - VIRTCHNL_STATUS_SUCCESS); - return; - } + if (!cnt) /* no work... */ + return (ENOENT); len = sizeof(struct virtchnl_vlan_filter_list) + (cnt * sizeof(u16)); @@ -627,16 +527,14 @@ ixlv_del_vlans(struct ixlv_sc *sc) if (len > IXL_AQ_BUF_SZ) { device_printf(dev, "%s: Exceeded Max AQ Buf size\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (EFBIG); } - v = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO); + v = malloc(len, M_IXLV, M_NOWAIT | M_ZERO); if (!v) { device_printf(dev, "%s: unable to allocate memory\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (ENOMEM); } v->vsi_id = sc->vsi_res->vsi_id; @@ -648,15 +546,16 @@ ixlv_del_vlans(struct ixlv_sc *sc) bcopy(&f->vlan, &v->vlan_id[i], sizeof(u16)); i++; SLIST_REMOVE(sc->vlan_filters, f, ixlv_vlan_filter, next); - free(f, M_DEVBUF); + free(f, M_IXLV); } if (i == cnt) break; } ixlv_send_pf_msg(sc, VIRTCHNL_OP_DEL_VLAN, (u8 *)v, len); - free(v, M_DEVBUF); + free(v, M_IXLV); /* add stats? */ + return (0); } @@ -665,13 +564,14 @@ ixlv_del_vlans(struct ixlv_sc *sc) ** table and creates an Admin Queue call to create ** the filters in the hardware. */ -void +int ixlv_add_ether_filters(struct ixlv_sc *sc) { struct virtchnl_ether_addr_list *a; struct ixlv_mac_filter *f; - device_t dev = sc->dev; - int len, j = 0, cnt = 0; + device_t dev = sc->dev; + int len, j = 0, cnt = 0; + enum i40e_status_code status; /* Get count of MAC addresses to add */ SLIST_FOREACH(f, sc->mac_filters, next) { @@ -679,21 +579,18 @@ ixlv_add_ether_filters(struct ixlv_sc *sc) cnt++; } if (cnt == 0) { /* Should not happen... */ - DDPRINTF(dev, "cnt == 0, exiting..."); - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_ADD_MAC_FILTER, - VIRTCHNL_STATUS_SUCCESS); - return; + ixlv_dbg_vc(sc, "%s: cnt == 0, exiting...\n", __func__); + return (ENOENT); } len = sizeof(struct virtchnl_ether_addr_list) + (cnt * sizeof(struct virtchnl_ether_addr)); - a = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO); + a = malloc(len, M_IXLV, M_NOWAIT | M_ZERO); if (a == NULL) { device_printf(dev, "%s: Failed to get memory for " "virtchnl_ether_addr_list\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (ENOMEM); } a->vsi_id = sc->vsi.id; a->num_elements = cnt; @@ -705,7 +602,7 @@ ixlv_add_ether_filters(struct ixlv_sc *sc) f->flags &= ~IXL_FILTER_ADD; j++; - DDPRINTF(dev, "ADD: " MAC_FORMAT, + ixlv_dbg_vc(sc, "ADD: " MAC_FORMAT "\n", MAC_FORMAT_ARGS(f->macaddr)); } if (j == cnt) @@ -713,11 +610,12 @@ ixlv_add_ether_filters(struct ixlv_sc *sc) } DDPRINTF(dev, "len %d, j %d, cnt %d", len, j, cnt); - ixlv_send_pf_msg(sc, + + status = ixlv_send_pf_msg(sc, VIRTCHNL_OP_ADD_ETH_ADDR, (u8 *)a, len); /* add stats? */ - free(a, M_DEVBUF); - return; + free(a, M_IXLV); + return (status); } /* @@ -725,13 +623,13 @@ ixlv_add_ether_filters(struct ixlv_sc *sc) ** sc MAC filter list and creates an Admin Queue call ** to delete those filters in the hardware. */ -void +int ixlv_del_ether_filters(struct ixlv_sc *sc) { struct virtchnl_ether_addr_list *d; - device_t dev = sc->dev; - struct ixlv_mac_filter *f, *f_temp; - int len, j = 0, cnt = 0; + struct ixlv_mac_filter *f, *f_temp; + device_t dev = sc->dev; + int len, j = 0, cnt = 0; /* Get count of MAC addresses to delete */ SLIST_FOREACH(f, sc->mac_filters, next) { @@ -739,21 +637,18 @@ ixlv_del_ether_filters(struct ixlv_sc *sc) cnt++; } if (cnt == 0) { - DDPRINTF(dev, "cnt == 0, exiting..."); - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_DEL_MAC_FILTER, - VIRTCHNL_STATUS_SUCCESS); - return; + ixlv_dbg_vc(sc, "%s: cnt == 0, exiting...\n", __func__); + return (ENOENT); } len = sizeof(struct virtchnl_ether_addr_list) + (cnt * sizeof(struct virtchnl_ether_addr)); - d = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO); + d = malloc(len, M_IXLV, M_NOWAIT | M_ZERO); if (d == NULL) { device_printf(dev, "%s: Failed to get memory for " "virtchnl_ether_addr_list\n", __func__); - ixl_vc_schedule_retry(&sc->vc_mgr); - return; + return (ENOMEM); } d->vsi_id = sc->vsi.id; d->num_elements = cnt; @@ -762,11 +657,11 @@ ixlv_del_ether_filters(struct ixlv_sc *sc) SLIST_FOREACH_SAFE(f, sc->mac_filters, next, f_temp) { if (f->flags & IXL_FILTER_DEL) { bcopy(f->macaddr, d->list[j].addr, ETHER_ADDR_LEN); - DDPRINTF(dev, "DEL: " MAC_FORMAT, + ixlv_dbg_vc(sc, "DEL: " MAC_FORMAT "\n", MAC_FORMAT_ARGS(f->macaddr)); j++; SLIST_REMOVE(sc->mac_filters, f, ixlv_mac_filter, next); - free(f, M_DEVBUF); + free(f, M_IXLV); } if (j == cnt) break; @@ -774,15 +669,15 @@ ixlv_del_ether_filters(struct ixlv_sc *sc) ixlv_send_pf_msg(sc, VIRTCHNL_OP_DEL_ETH_ADDR, (u8 *)d, len); /* add stats? */ - free(d, M_DEVBUF); - return; + free(d, M_IXLV); + return (0); } /* ** ixlv_request_reset ** Request that the PF reset this VF. No response is expected. */ -void +int ixlv_request_reset(struct ixlv_sc *sc) { /* @@ -792,13 +687,14 @@ ixlv_request_reset(struct ixlv_sc *sc) */ wr32(&sc->hw, I40E_VFGEN_RSTAT, VIRTCHNL_VFR_INPROGRESS); ixlv_send_pf_msg(sc, VIRTCHNL_OP_RESET_VF, NULL, 0); + return (0); } /* ** ixlv_request_stats ** Request the statistics for this VF's VSI from PF. */ -void +int ixlv_request_stats(struct ixlv_sc *sc) { struct virtchnl_queue_select vqs; @@ -808,10 +704,10 @@ ixlv_request_stats(struct ixlv_sc *sc) /* Low priority, we don't need to error check */ error = ixlv_send_pf_msg(sc, VIRTCHNL_OP_GET_STATS, (u8 *)&vqs, sizeof(vqs)); -#ifdef IXL_DEBUG if (error) device_printf(sc->dev, "Error sending stats request to PF: %d\n", error); -#endif + + return (0); } /* @@ -850,7 +746,7 @@ ixlv_update_stats_counters(struct ixlv_sc *sc, struct i40e_eth_stats *es) vsi->eth_stats = *es; } -void +int ixlv_config_rss_key(struct ixlv_sc *sc) { struct virtchnl_rss_key *rss_key_msg; @@ -867,26 +763,27 @@ ixlv_config_rss_key(struct ixlv_sc *sc) /* Send the fetched key */ key_length = IXL_RSS_KEY_SIZE; msg_len = sizeof(struct virtchnl_rss_key) + (sizeof(u8) * key_length) - 1; - rss_key_msg = malloc(msg_len, M_DEVBUF, M_NOWAIT | M_ZERO); + rss_key_msg = malloc(msg_len, M_IXLV, M_NOWAIT | M_ZERO); if (rss_key_msg == NULL) { device_printf(sc->dev, "Unable to allocate msg memory for RSS key msg.\n"); - return; + return (ENOMEM); } rss_key_msg->vsi_id = sc->vsi_res->vsi_id; rss_key_msg->key_len = key_length; bcopy(rss_seed, &rss_key_msg->key[0], key_length); - DDPRINTF(sc->dev, "config_rss: vsi_id %d, key_len %d", + ixlv_dbg_vc(sc, "config_rss: vsi_id %d, key_len %d\n", rss_key_msg->vsi_id, rss_key_msg->key_len); ixlv_send_pf_msg(sc, VIRTCHNL_OP_CONFIG_RSS_KEY, (u8 *)rss_key_msg, msg_len); - free(rss_key_msg, M_DEVBUF); + free(rss_key_msg, M_IXLV); + return (0); } -void +int ixlv_set_rss_hena(struct ixlv_sc *sc) { struct virtchnl_rss_hena hena; @@ -899,9 +796,10 @@ ixlv_set_rss_hena(struct ixlv_sc *sc) ixlv_send_pf_msg(sc, VIRTCHNL_OP_SET_RSS_HENA, (u8 *)&hena, sizeof(hena)); + return (0); } -void +int ixlv_config_rss_lut(struct ixlv_sc *sc) { struct virtchnl_rss_lut *rss_lut_msg; @@ -912,10 +810,10 @@ ixlv_config_rss_lut(struct ixlv_sc *sc) lut_length = IXL_RSS_VSI_LUT_SIZE; msg_len = sizeof(struct virtchnl_rss_lut) + (lut_length * sizeof(u8)) - 1; - rss_lut_msg = malloc(msg_len, M_DEVBUF, M_NOWAIT | M_ZERO); + rss_lut_msg = malloc(msg_len, M_IXLV, M_NOWAIT | M_ZERO); if (rss_lut_msg == NULL) { device_printf(sc->dev, "Unable to allocate msg memory for RSS lut msg.\n"); - return; + return (ENOMEM); } rss_lut_msg->vsi_id = sc->vsi_res->vsi_id; @@ -942,7 +840,21 @@ ixlv_config_rss_lut(struct ixlv_sc *sc) ixlv_send_pf_msg(sc, VIRTCHNL_OP_CONFIG_RSS_LUT, (u8 *)rss_lut_msg, msg_len); - free(rss_lut_msg, M_DEVBUF); + free(rss_lut_msg, M_IXLV); + return (0); +} + +int +ixlv_config_promisc_mode(struct ixlv_sc *sc) +{ + struct virtchnl_promisc_info pinfo; + + pinfo.vsi_id = sc->vsi_res->vsi_id; + pinfo.flags = sc->promisc_flags; + + ixlv_send_pf_msg(sc, VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, + (u8 *)&pinfo, sizeof(pinfo)); + return (0); } /* @@ -958,7 +870,10 @@ ixlv_vc_completion(struct ixlv_sc *sc, enum virtchnl_status_code v_retval, u8 *msg, u16 msglen) { device_t dev = sc->dev; - struct ixl_vsi *vsi = &sc->vsi; + + if (v_opcode != VIRTCHNL_OP_GET_STATS) + ixlv_dbg_vc(sc, "%s: opcode %s\n", __func__, + ixl_vc_opcode_str(v_opcode)); if (v_opcode == VIRTCHNL_OP_EVENT) { struct virtchnl_pf_event *vpe = @@ -966,11 +881,9 @@ ixlv_vc_completion(struct ixlv_sc *sc, switch (vpe->event) { case VIRTCHNL_EVENT_LINK_CHANGE: -#ifdef IXL_DEBUG - device_printf(dev, "Link change: status %d, speed %d\n", + ixlv_dbg_vc(sc, "Link change: status %d, speed %s\n", vpe->event_data.link_event.link_status, - vpe->event_data.link_event.link_speed); -#endif + ixlv_vc_speed_to_string(vpe->event_data.link_event.link_speed)); sc->link_up = vpe->event_data.link_event.link_status; sc->link_speed = @@ -983,8 +896,8 @@ ixlv_vc_completion(struct ixlv_sc *sc, ixlv_if_init(sc->vsi.ctx); break; default: - device_printf(dev, "%s: Unknown event %d from AQ\n", - __func__, vpe->event); + ixlv_dbg_vc(sc, "Unknown event %d from AQ\n", + vpe->event); break; } @@ -998,273 +911,104 @@ ixlv_vc_completion(struct ixlv_sc *sc, __func__, i40e_vc_stat_str(&sc->hw, v_retval), ixl_vc_opcode_str(v_opcode)); } -#ifdef IXL_DEBUG - if (v_opcode != VIRTCHNL_OP_GET_STATS) - DDPRINTF(dev, "opcode %d", v_opcode); -#endif - switch (v_opcode) { case VIRTCHNL_OP_GET_STATS: ixlv_update_stats_counters(sc, (struct i40e_eth_stats *)msg); break; case VIRTCHNL_OP_ADD_ETH_ADDR: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_ADD_MAC_FILTER, - v_retval); if (v_retval) { device_printf(dev, "WARNING: Error adding VF mac filter!\n"); device_printf(dev, "WARNING: Device may not receive traffic!\n"); } break; case VIRTCHNL_OP_DEL_ETH_ADDR: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_DEL_MAC_FILTER, - v_retval); break; case VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_CONFIGURE_PROMISC, - v_retval); break; case VIRTCHNL_OP_ADD_VLAN: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_ADD_VLAN_FILTER, - v_retval); break; case VIRTCHNL_OP_DEL_VLAN: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_DEL_VLAN_FILTER, - v_retval); break; case VIRTCHNL_OP_ENABLE_QUEUES: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_ENABLE_QUEUES, - v_retval); - if (v_retval == 0) { - /* Update link status */ - ixlv_update_link_status(sc); - /* Turn on all interrupts */ - ixlv_enable_intr(vsi); - /* And inform the stack we're ready */ - // vsi->ifp->if_drv_flags |= IFF_DRV_RUNNING; - /* TODO: Clear a state flag, so we know we're ready to run init again */ - } + atomic_store_rel_32(&sc->queues_enabled, 1); + wakeup_one(&sc->enable_queues_chan); break; case VIRTCHNL_OP_DISABLE_QUEUES: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_DISABLE_QUEUES, - v_retval); - if (v_retval == 0) { - /* Turn off all interrupts */ - ixlv_disable_intr(vsi); - /* Tell the stack that the interface is no longer active */ - vsi->ifp->if_drv_flags &= ~(IFF_DRV_RUNNING); - } + atomic_store_rel_32(&sc->queues_enabled, 0); + wakeup_one(&sc->disable_queues_chan); break; case VIRTCHNL_OP_CONFIG_VSI_QUEUES: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_CONFIGURE_QUEUES, - v_retval); break; case VIRTCHNL_OP_CONFIG_IRQ_MAP: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_MAP_VECTORS, - v_retval); break; case VIRTCHNL_OP_CONFIG_RSS_KEY: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_CONFIG_RSS_KEY, - v_retval); break; case VIRTCHNL_OP_SET_RSS_HENA: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_SET_RSS_HENA, - v_retval); break; case VIRTCHNL_OP_CONFIG_RSS_LUT: - ixl_vc_process_resp(&sc->vc_mgr, IXLV_FLAG_AQ_CONFIG_RSS_LUT, - v_retval); break; default: -#ifdef IXL_DEBUG - device_printf(dev, - "%s: Received unexpected message %s from PF.\n", - __func__, ixl_vc_opcode_str(v_opcode)); -#endif + ixlv_dbg_vc(sc, + "Received unexpected message %s from PF.\n", + ixl_vc_opcode_str(v_opcode)); break; } - return; } -static void +int ixl_vc_send_cmd(struct ixlv_sc *sc, uint32_t request) { switch (request) { case IXLV_FLAG_AQ_MAP_VECTORS: - ixlv_map_queues(sc); - break; + return ixlv_map_queues(sc); case IXLV_FLAG_AQ_ADD_MAC_FILTER: - ixlv_add_ether_filters(sc); - break; + return ixlv_add_ether_filters(sc); case IXLV_FLAG_AQ_ADD_VLAN_FILTER: - ixlv_add_vlans(sc); - break; + return ixlv_add_vlans(sc); case IXLV_FLAG_AQ_DEL_MAC_FILTER: - ixlv_del_ether_filters(sc); - break; + return ixlv_del_ether_filters(sc); case IXLV_FLAG_AQ_DEL_VLAN_FILTER: - ixlv_del_vlans(sc); - break; + return ixlv_del_vlans(sc); case IXLV_FLAG_AQ_CONFIGURE_QUEUES: - ixlv_configure_queues(sc); - break; + return ixlv_configure_queues(sc); case IXLV_FLAG_AQ_DISABLE_QUEUES: - ixlv_disable_queues(sc); - break; + return ixlv_disable_queues(sc); case IXLV_FLAG_AQ_ENABLE_QUEUES: - ixlv_enable_queues(sc); - break; + return ixlv_enable_queues(sc); case IXLV_FLAG_AQ_CONFIG_RSS_KEY: - ixlv_config_rss_key(sc); - break; + return ixlv_config_rss_key(sc); case IXLV_FLAG_AQ_SET_RSS_HENA: - ixlv_set_rss_hena(sc); - break; + return ixlv_set_rss_hena(sc); case IXLV_FLAG_AQ_CONFIG_RSS_LUT: - ixlv_config_rss_lut(sc); - break; - } -} - -void -ixl_vc_init_mgr(struct ixlv_sc *sc, struct ixl_vc_mgr *mgr) -{ - mgr->sc = sc; - mgr->current = NULL; - TAILQ_INIT(&mgr->pending); - callout_init_mtx(&mgr->callout, &sc->mtx, 0); -} + return ixlv_config_rss_lut(sc); -static void -ixl_vc_process_completion(struct ixl_vc_mgr *mgr, enum i40e_status_code err) -{ - struct ixl_vc_cmd *cmd; - - cmd = mgr->current; - mgr->current = NULL; - cmd->flags &= ~IXLV_VC_CMD_FLAG_BUSY; - - cmd->callback(cmd, cmd->arg, err); - ixl_vc_process_next(mgr); -} - -static void -ixl_vc_process_resp(struct ixl_vc_mgr *mgr, uint32_t request, - enum virtchnl_status_code err) -{ - struct ixl_vc_cmd *cmd; - - cmd = mgr->current; - if (cmd == NULL || cmd->request != request) - return; - - callout_stop(&mgr->callout); - /* ATM, the virtchnl codes map to i40e ones directly */ - ixl_vc_process_completion(mgr, (enum i40e_status_code)err); -} - -static void -ixl_vc_cmd_timeout(void *arg) -{ - struct ixl_vc_mgr *mgr = (struct ixl_vc_mgr *)arg; - - ixl_vc_process_completion(mgr, I40E_ERR_TIMEOUT); -} - -static void -ixl_vc_cmd_retry(void *arg) -{ - struct ixl_vc_mgr *mgr = (struct ixl_vc_mgr *)arg; - - ixl_vc_send_current(mgr); -} - -static void -ixl_vc_send_current(struct ixl_vc_mgr *mgr) -{ - struct ixl_vc_cmd *cmd; - - cmd = mgr->current; - ixl_vc_send_cmd(mgr->sc, cmd->request); - callout_reset(&mgr->callout, IXLV_VC_TIMEOUT, ixl_vc_cmd_timeout, mgr); -} - -static void -ixl_vc_process_next(struct ixl_vc_mgr *mgr) -{ - struct ixl_vc_cmd *cmd; - - if (mgr->current != NULL) - return; - - if (TAILQ_EMPTY(&mgr->pending)) - return; - - cmd = TAILQ_FIRST(&mgr->pending); - TAILQ_REMOVE(&mgr->pending, cmd, next); - - mgr->current = cmd; - ixl_vc_send_current(mgr); -} - -static void -ixl_vc_schedule_retry(struct ixl_vc_mgr *mgr) -{ - - callout_reset(&mgr->callout, howmany(hz, 100), ixl_vc_cmd_retry, mgr); -} - -void -ixl_vc_enqueue(struct ixl_vc_mgr *mgr, struct ixl_vc_cmd *cmd, - uint32_t req, ixl_vc_callback_t *callback, void *arg) -{ - if (cmd->flags & IXLV_VC_CMD_FLAG_BUSY) { - if (mgr->current == cmd) - mgr->current = NULL; - else - TAILQ_REMOVE(&mgr->pending, cmd, next); + case IXLV_FLAG_AQ_CONFIGURE_PROMISC: + return ixlv_config_promisc_mode(sc); } - cmd->request = req; - cmd->callback = callback; - cmd->arg = arg; - cmd->flags |= IXLV_VC_CMD_FLAG_BUSY; - TAILQ_INSERT_TAIL(&mgr->pending, cmd, next); - - ixl_vc_process_next(mgr); + return (0); } -void -ixl_vc_flush(struct ixl_vc_mgr *mgr) +void * +ixl_vc_get_op_chan(struct ixlv_sc *sc, uint32_t request) { - struct ixl_vc_cmd *cmd; - - KASSERT(TAILQ_EMPTY(&mgr->pending) || mgr->current != NULL, - ("ixlv: pending commands waiting but no command in progress")); - - cmd = mgr->current; - if (cmd != NULL) { - mgr->current = NULL; - cmd->flags &= ~IXLV_VC_CMD_FLAG_BUSY; - cmd->callback(cmd, cmd->arg, I40E_ERR_ADAPTER_STOPPED); - } - - while ((cmd = TAILQ_FIRST(&mgr->pending)) != NULL) { - TAILQ_REMOVE(&mgr->pending, cmd, next); - cmd->flags &= ~IXLV_VC_CMD_FLAG_BUSY; - cmd->callback(cmd, cmd->arg, I40E_ERR_ADAPTER_STOPPED); + switch (request) { + case IXLV_FLAG_AQ_ENABLE_QUEUES: + return (&sc->enable_queues_chan); + case IXLV_FLAG_AQ_DISABLE_QUEUES: + return (&sc->disable_queues_chan); + default: + return (NULL); } - - callout_stop(&mgr->callout); } - diff --git a/sys/modules/Makefile b/sys/modules/Makefile index 726144753c8b..33dad3f6a145 100644 --- a/sys/modules/Makefile +++ b/sys/modules/Makefile @@ -153,6 +153,7 @@ SUBDIR= \ ${_hwpmc_mips74k} \ ${_hyperv} \ i2c \ + ${_iavf} \ ${_ibcore} \ ${_ibcs2} \ ${_ichwd} \ @@ -717,9 +718,9 @@ _x86bios= x86bios .if ${MACHINE_CPUARCH} == "amd64" _ccp= ccp _efirt= efirt +_iavf= iavf _ioat= ioat _ixl= ixl -_ixlv= ixlv _linux64= linux64 _linux_common= linux_common _ntb= ntb diff --git a/sys/modules/ixlv/Makefile b/sys/modules/iavf/Makefile index 36b8eb4171a1..e1fd0083be99 100644 --- a/sys/modules/ixlv/Makefile +++ b/sys/modules/iavf/Makefile @@ -2,9 +2,9 @@ .PATH: ${SRCTOP}/sys/dev/ixl -KMOD = if_ixlv +KMOD = if_iavf SRCS = device_if.h bus_if.h pci_if.h ifdi_if.h -SRCS += opt_inet.h opt_inet6.h opt_rss.h opt_ixl.h opt_iflib.h +SRCS += opt_inet.h opt_inet6.h opt_rss.h opt_ixl.h opt_iflib.h opt_global.h SRCS += if_ixlv.c ixlvc.c ixl_txrx.c i40e_osdep.c # Shared source @@ -12,5 +12,7 @@ SRCS += i40e_common.c i40e_nvm.c i40e_adminq.c # Debug messages / sysctls # CFLAGS += -DIXL_DEBUG +# Enable asserts and other debugging facilities +# CFLAGS += -DINVARIANTS -DINVARIANTS_SUPPORT -DWITNESS .include <bsd.kmod.mk> diff --git a/sys/modules/ixl/Makefile b/sys/modules/ixl/Makefile index ac199617077e..beeb8f9a5564 100644 --- a/sys/modules/ixl/Makefile +++ b/sys/modules/ixl/Makefile @@ -6,7 +6,7 @@ KMOD = if_ixl SRCS = device_if.h bus_if.h pci_if.h ifdi_if.h SRCS += opt_inet.h opt_inet6.h opt_rss.h opt_ixl.h opt_iflib.h SRCS += if_ixl.c ixl_pf_main.c ixl_pf_qmgr.c ixl_txrx.c ixl_pf_i2c.c i40e_osdep.c -SRCS.PCI_IOV = pci_iov_if.h ixl_pf_iov.c +SRCS.PCI_IOV += pci_iov_if.h ixl_pf_iov.c # Shared source SRCS += i40e_common.c i40e_nvm.c i40e_adminq.c i40e_lan_hmc.c i40e_hmc.c i40e_dcb.c @@ -14,7 +14,11 @@ SRCS += i40e_common.c i40e_nvm.c i40e_adminq.c i40e_lan_hmc.c i40e_hmc.c i40e # Debug messages / sysctls # CFLAGS += -DIXL_DEBUG -#CFLAGS += -DIXL_IW -#SRCS += ixl_iw.c +# Enable asserts and other debugging facilities +# CFLAGS += -DINVARIANTS -DINVARIANTS_SUPPORT -DWITNESS + +# Enable iWARP client interface +# CFLAGS += -DIXL_IW +# SRCS += ixl_iw.c .include <bsd.kmod.mk> diff --git a/sys/net/iflib.c b/sys/net/iflib.c index d9a6e3b962d7..594d50395d82 100644 --- a/sys/net/iflib.c +++ b/sys/net/iflib.c @@ -101,6 +101,10 @@ __FBSDID("$FreeBSD$"); #include <x86/iommu/busdma_dmar.h> #endif +#ifdef PCI_IOV +#include <dev/pci/pci_iov.h> +#endif + #include <sys/bitstring.h> /* * enable accounting of every mbuf as it comes in to and goes out of @@ -157,9 +161,9 @@ typedef struct iflib_filter_info { struct iflib_ctx { KOBJ_FIELDS; - /* - * Pointer to hardware driver's softc - */ + /* + * Pointer to hardware driver's softc + */ void *ifc_softc; device_t ifc_dev; if_t ifc_ifp; @@ -178,7 +182,6 @@ struct iflib_ctx { uint32_t ifc_if_flags; uint32_t ifc_flags; uint32_t ifc_max_fl_buf_size; - int ifc_in_detach; int ifc_link_state; int ifc_link_irq; @@ -254,12 +257,6 @@ iflib_get_flags(if_ctx_t ctx) } void -iflib_set_detach(if_ctx_t ctx) -{ - ctx->ifc_in_detach = 1; -} - -void iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN]) { @@ -571,6 +568,13 @@ rxd_info_zero(if_rxd_info_t ri) #define CALLOUT_LOCK(txq) mtx_lock(&txq->ift_mtx) #define CALLOUT_UNLOCK(txq) mtx_unlock(&txq->ift_mtx) +void +iflib_set_detach(if_ctx_t ctx) +{ + STATE_LOCK(ctx); + ctx->ifc_flags |= IFC_IN_DETACH; + STATE_UNLOCK(ctx); +} /* Our boot-time initialization hook */ static int iflib_module_event_handler(module_t, int, void *); @@ -738,6 +742,7 @@ static void iflib_add_device_sysctl_post(if_ctx_t ctx); static void iflib_ifmp_purge(iflib_txq_t txq); static void _iflib_pre_assert(if_softc_ctx_t scctx); static void iflib_if_init_locked(if_ctx_t ctx); +static void iflib_free_intr_mem(if_ctx_t ctx); #ifndef __NO_STRICT_ALIGNMENT static struct mbuf * iflib_fixup_rx(struct mbuf *m); #endif @@ -2072,6 +2077,16 @@ __iflib_fl_refill_lt(if_ctx_t ctx, iflib_fl_t fl, int max) _iflib_fl_refill(ctx, fl, min(max, reclaimable)); } +uint8_t +iflib_in_detach(if_ctx_t ctx) +{ + bool in_detach; + STATE_LOCK(ctx); + in_detach = !!(ctx->ifc_flags & IFC_IN_DETACH); + STATE_UNLOCK(ctx); + return (in_detach); +} + static void iflib_fl_bufs_free(iflib_fl_t fl) { @@ -2087,7 +2102,8 @@ iflib_fl_bufs_free(iflib_fl_t fl) if (fl->ifl_sds.ifsd_map != NULL) { bus_dmamap_t sd_map = fl->ifl_sds.ifsd_map[i]; bus_dmamap_unload(fl->ifl_desc_tag, sd_map); - if (fl->ifl_rxq->ifr_ctx->ifc_in_detach) + // XXX: Should this get moved out? + if (iflib_in_detach(fl->ifl_rxq->ifr_ctx)) bus_dmamap_destroy(fl->ifl_desc_tag, sd_map); } if (*sd_m != NULL) { @@ -3842,7 +3858,7 @@ _task_fn_admin(void *context) if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; iflib_txq_t txq; int i; - bool oactive, running, do_reset, do_watchdog; + bool oactive, running, do_reset, do_watchdog, in_detach; uint32_t reset_on = hz / 2; STATE_LOCK(ctx); @@ -3850,11 +3866,13 @@ _task_fn_admin(void *context) oactive = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE); do_reset = (ctx->ifc_flags & IFC_DO_RESET); do_watchdog = (ctx->ifc_flags & IFC_DO_WATCHDOG); + in_detach = (ctx->ifc_flags & IFC_IN_DETACH); ctx->ifc_flags &= ~(IFC_DO_RESET|IFC_DO_WATCHDOG); STATE_UNLOCK(ctx); - if ((!running & !oactive) && - !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN)) + if ((!running && !oactive) && !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN)) + return; + if (in_detach) return; CTX_LOCK(ctx); @@ -3893,7 +3911,8 @@ _task_fn_iov(void *context) { if_ctx_t ctx = context; - if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)) + if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING) && + !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN)) return; CTX_LOCK(ctx); @@ -4680,17 +4699,18 @@ iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ct ctx->ifc_flags |= IFC_INIT_DONE; CTX_UNLOCK(ctx); return (0); + fail_detach: ether_ifdetach(ctx->ifc_ifp); fail_intr_free: - if (scctx->isc_intr == IFLIB_INTR_MSIX || scctx->isc_intr == IFLIB_INTR_MSI) - pci_release_msi(ctx->ifc_dev); fail_queues: iflib_tx_structures_free(ctx); iflib_rx_structures_free(ctx); fail: + iflib_free_intr_mem(ctx); IFDI_DETACH(ctx); CTX_UNLOCK(ctx); + return (err); } @@ -4975,12 +4995,21 @@ iflib_device_deregister(if_ctx_t ctx) /* Make sure VLANS are not using driver */ if (if_vlantrunkinuse(ifp)) { - device_printf(dev,"Vlan in use, detach first\n"); + device_printf(dev, "Vlan in use, detach first\n"); return (EBUSY); } +#ifdef PCI_IOV + if (!CTX_IS_VF(ctx) && pci_iov_detach(dev) != 0) { + device_printf(dev, "SR-IOV in use; detach first.\n"); + return (EBUSY); + } +#endif + + STATE_LOCK(ctx); + ctx->ifc_flags |= IFC_IN_DETACH; + STATE_UNLOCK(ctx); CTX_LOCK(ctx); - ctx->ifc_in_detach = 1; iflib_stop(ctx); CTX_UNLOCK(ctx); @@ -5021,17 +5050,7 @@ iflib_device_deregister(if_ctx_t ctx) /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/ CTX_LOCK_DESTROY(ctx); device_set_softc(ctx->ifc_dev, NULL); - if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_LEGACY) { - pci_release_msi(dev); - } - if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_MSIX) { - iflib_irq_free(ctx, &ctx->ifc_legacy_irq); - } - if (ctx->ifc_msix_mem != NULL) { - bus_release_resource(ctx->ifc_dev, SYS_RES_MEMORY, - ctx->ifc_softc_ctx.isc_msix_bar, ctx->ifc_msix_mem); - ctx->ifc_msix_mem = NULL; - } + iflib_free_intr_mem(ctx); bus_generic_detach(dev); if_free(ifp); @@ -5040,10 +5059,27 @@ iflib_device_deregister(if_ctx_t ctx) iflib_rx_structures_free(ctx); if (ctx->ifc_flags & IFC_SC_ALLOCATED) free(ctx->ifc_softc, M_IFLIB); + STATE_LOCK_DESTROY(ctx); free(ctx, M_IFLIB); return (0); } +static void +iflib_free_intr_mem(if_ctx_t ctx) +{ + + if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_LEGACY) { + pci_release_msi(ctx->ifc_dev); + } + if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_MSIX) { + iflib_irq_free(ctx, &ctx->ifc_legacy_irq); + } + if (ctx->ifc_msix_mem != NULL) { + bus_release_resource(ctx->ifc_dev, SYS_RES_MEMORY, + ctx->ifc_softc_ctx.isc_msix_bar, ctx->ifc_msix_mem); + ctx->ifc_msix_mem = NULL; + } +} int iflib_device_detach(device_t dev) @@ -5215,7 +5251,7 @@ iflib_register(if_ctx_t ctx) CTX_LOCK_INIT(ctx); STATE_LOCK_INIT(ctx, device_get_nameunit(ctx->ifc_dev)); - ifp = ctx->ifc_ifp = if_gethandle(IFT_ETHER); + ifp = ctx->ifc_ifp = if_alloc(IFT_ETHER); if (ifp == NULL) { device_printf(dev, "can not allocate ifnet structure\n"); return (ENOMEM); @@ -5399,7 +5435,7 @@ iflib_queues_alloc(if_ctx_t ctx) fl[j].ifl_ifdi = &rxq->ifr_ifdi[j + rxq->ifr_fl_offset]; fl[j].ifl_rxd_size = scctx->isc_rxd_size[j]; } - /* Allocate receive buffers for the ring*/ + /* Allocate receive buffers for the ring */ if (iflib_rxsd_alloc(rxq)) { device_printf(dev, "Critical Failure setting up receive buffers\n"); @@ -5554,6 +5590,8 @@ iflib_rx_structures_free(if_ctx_t ctx) for (int i = 0; i < ctx->ifc_softc_ctx.isc_nrxqsets; i++, rxq++) { iflib_rx_sds_free(rxq); } + free(ctx->ifc_rxqs, M_IFLIB); + ctx->ifc_rxqs = NULL; } static int @@ -5814,7 +5852,7 @@ iflib_irq_alloc_generic(if_ctx_t ctx, if_irq_t irq, int rid, } void -iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, void *arg, int qid, const char *name) +iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, void *arg, int qid, const char *name) { struct grouptask *gtask; struct taskqgroup *tqg; @@ -6132,8 +6170,9 @@ iflib_msix_init(if_ctx_t ctx) if (ctx->ifc_sysctl_qs_eq_override == 0) { #ifdef INVARIANTS if (tx_queues != rx_queues) - device_printf(dev, "queue equality override not set, capping rx_queues at %d and tx_queues at %d\n", - min(rx_queues, tx_queues), min(rx_queues, tx_queues)); + device_printf(dev, + "queue equality override not set, capping rx_queues at %d and tx_queues at %d\n", + min(rx_queues, tx_queues), min(rx_queues, tx_queues)); #endif tx_queues = min(rx_queues, tx_queues); rx_queues = min(rx_queues, tx_queues); @@ -6143,8 +6182,7 @@ iflib_msix_init(if_ctx_t ctx) vectors = rx_queues + admincnt; if ((err = pci_alloc_msix(dev, &vectors)) == 0) { - device_printf(dev, - "Using MSIX interrupts with %d vectors\n", vectors); + device_printf(dev, "Using MSIX interrupts with %d vectors\n", vectors); scctx->isc_vectors = vectors; scctx->isc_nrxqsets = rx_queues; scctx->isc_ntxqsets = tx_queues; @@ -6152,7 +6190,8 @@ iflib_msix_init(if_ctx_t ctx) return (vectors); } else { - device_printf(dev, "failed to allocate %d msix vectors, err: %d - using MSI\n", vectors, err); + device_printf(dev, + "failed to allocate %d msix vectors, err: %d - using MSI\n", vectors, err); bus_release_resource(dev, SYS_RES_MEMORY, bar, ctx->ifc_msix_mem); ctx->ifc_msix_mem = NULL; @@ -6463,6 +6502,15 @@ iflib_add_device_sysctl_post(if_ctx_t ctx) } +void +iflib_request_reset(if_ctx_t ctx) +{ + + STATE_LOCK(ctx); + ctx->ifc_flags |= IFC_DO_RESET; + STATE_UNLOCK(ctx); +} + #ifndef __NO_STRICT_ALIGNMENT static struct mbuf * iflib_fixup_rx(struct mbuf *m) diff --git a/sys/net/iflib.h b/sys/net/iflib.h index bda5ad45dfc8..8c2be41b3f33 100644 --- a/sys/net/iflib.h +++ b/sys/net/iflib.h @@ -246,7 +246,7 @@ struct if_shared_ctx { /* fields necessary for probe */ pci_vendor_info_t *isc_vendor_info; char *isc_driver_version; -/* optional function to transform the read values to match the table*/ + /* optional function to transform the read values to match the table*/ void (*isc_parse_devinfo) (uint16_t *device_id, uint16_t *subvendor_id, uint16_t *subdevice_id, uint16_t *rev_id); int isc_nrxd_min[8]; @@ -375,6 +375,8 @@ if_softc_ctx_t iflib_get_softc_ctx(if_ctx_t ctx); if_shared_ctx_t iflib_get_sctx(if_ctx_t ctx); void iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN]); +void iflib_request_reset(if_ctx_t ctx); +uint8_t iflib_in_detach(if_ctx_t ctx); /* * If the driver can plug cleanly in to newbus use these diff --git a/sys/net/iflib_private.h b/sys/net/iflib_private.h index 1f10cf9d6903..508a327c03dc 100644 --- a/sys/net/iflib_private.h +++ b/sys/net/iflib_private.h @@ -42,6 +42,7 @@ #define IFC_DO_WATCHDOG 0x100 #define IFC_CHECK_HUNG 0x200 #define IFC_PSEUDO 0x400 +#define IFC_IN_DETACH 0x800 #define IFC_NETMAP_TX_IRQ 0x80000000 |