VxLAN offloading is not functional if the NIC is running in multichannel mode (UMC, FLEX-10, VNIC...). Enabling this additionally kills whole connectivity through the NIC and the device needs to be down and up to restore it. The firmware should take care about it and does not allow the conversion of interface to tunnel type (be_cmd_manage_iface) or should support VxLAN offloading if multichannel config is enabled. I have tested this on the latest available firmware (10.6.144.21). Result: [root@sm-04 ~]# ip link set enp5s0f0 up[root@sm-04 ~]# ip addr add 172.30.10.50/24 dev enp5s0f0 [root@sm-04 ~]# ping -c 3 172.30.10.254PING 172.30.10.254 (172.30.10.254) 56(84) bytes of data. 64 bytes from 172.30.10.254: icmp_seq=1 ttl=64 time=0.317 ms 64 bytes from 172.30.10.254: icmp_seq=2 ttl=64 time=0.187 ms 64 bytes from 172.30.10.254: icmp_seq=3 ttl=64 time=0.188 ms --- 172.30.10.254 ping statistics --- 3 packets transmitted, 3 received, 0% packet loss, time 2000ms rtt min/avg/max/mdev = 0.187/0.230/0.317/0.063 ms [root@sm-04 ~]# ip link add link enp5s0f0 vxlan10 type vxlan id 10 remote 172.30.10.60 dstport 4789 [root@sm-04 ~]# ip link set vxlan10 up [ 7900.442811] be2net 0000:05:00.0: Enabled VxLAN offloads for UDP port 4789 [ 7900.455722] be2net 0000:05:00.1: Enabled VxLAN offloads for UDP port 4789 [ 7900.468635] be2net 0000:05:00.2: Enabled VxLAN offloads for UDP port 4789 [ 7900.481553] be2net 0000:05:00.3: Enabled VxLAN offloads for UDP port 4789 [root@sm-04 ~]# ping -c 3 172.30.10.254 PING 172.30.10.254 (172.30.10.254) 56(84) bytes of data. --- 172.30.10.254 ping statistics --- 3 packets transmitted, 0 received, 100% packet loss, time 1999ms [root@sm-04 ~]# ip link set vxlan10 down [ 7959.434093] be2net 0000:05:00.0: Disabled VxLAN offloads for UDP port 4789 [ 7959.444792] be2net 0000:05:00.1: Disabled VxLAN offloads for UDP port 4789 [ 7959.455592] be2net 0000:05:00.2: Disabled VxLAN offloads for UDP port 4789 [ 7959.466416] be2net 0000:05:00.3: Disabled VxLAN offloads for UDP port 4789 [root@sm-04 ~]# ip link del vxlan10 [root@sm-04 ~]# ping -c 3 172.30.10.254 PING 172.30.10.254 (172.30.10.254) 56(84) bytes of data. --- 172.30.10.254 ping statistics --- 3 packets transmitted, 0 received, 100% packet loss, time 1999ms [root@sm-04 ~]# ip link set enp5s0f0 down [root@sm-04 ~]# ip link set enp5s0f0 up [ 8071.019003] be2net 0000:05:00.0 enp5s0f0: Link is Up [root@sm-04 ~]# ping -c 3 172.30.10.254 PING 172.30.10.254 (172.30.10.254) 56(84) bytes of data. 64 bytes from 172.30.10.254: icmp_seq=1 ttl=64 time=0.318 ms 64 bytes from 172.30.10.254: icmp_seq=2 ttl=64 time=0.196 ms 64 bytes from 172.30.10.254: icmp_seq=3 ttl=64 time=0.194 ms --- 172.30.10.254 ping statistics --- 3 packets transmitted, 3 received, 0% packet loss, time 2000ms rtt min/avg/max/mdev = 0.194/0.236/0.318/0.057 ms Cc: Sathya Perla <sathya.perla@avagotech.com> Cc: Ajit Khaparde <ajit.khaparde@avagotech.com> Cc: Padmanabh Ratnakar <padmanabh.ratnakar@avagotech.com> Cc: Sriharsha Basavapatna <sriharsha.basavapatna@avagotech.com> Signed-off-by: Ivan Vecera <ivecera@redhat.com> Acked-by: Ajit Khaparde <ajit.khaparde@avagotech.com> Signed-off-by: David S. Miller <davem@davemloft.net>
There are paths in the driver such as an unrecoverable error (UE) detection followed by a driver unload wherein be_clear() is invoked twice. Individual data structures are reset so that they are not cleaned/freed twice. This patch does the same for eqo->affinity_mask. It is freed only if EQs haven't yet been destroyed. This fixes a possible crash when affinity_mask is freed twice. Signed-off-by: Kalesh AP <kalesh.purayil@avagotech.com> Signed-off-by: Sathya Perla <sathya.perla@avagotech.com> Signed-off-by: David S. Miller <davem@davemloft.net>
An RX stall issue was seen on Lancer adapters, when RXQs are destroyed while they are in an "out of buffer" state. This patch fixes this issue by posting 64 buffers to each RXQ before destroying them in the close path. This is done after ensuring that no more new packets are selected for transfer to the RXQs by disabling interface filters. Signed-off-by: Kalesh AP <kalesh.purayil@avagotech.com> Signed-off-by: Sathya Perla <sathya.perla@avagotech.com> Signed-off-by: David S. Miller <davem@davemloft.net>
HW issues were observed on Lancer adapters if IFACE filters (flags, mac addrs etc) are enabled *before* creating RXQs. This patch changes the driver design by enabling filters in be_open() -- instead of be_setup() -- after RXQs are created and buffers posted. Two new wrapper functions, be_enable_if_filters() and be_disable_if_filters() are introduced to enable/disable IFACE filters in be_open()/be_close() respectively. In be_setup() the IFACE is now created only with the RSS flag. Signed-off-by: Kalesh AP <kalesh.purayil@avagotech.com> Signed-off-by: Sathya Perla <sathya.perla@avagotech.com> Signed-off-by: David S. Miller <davem@davemloft.net>
One more missing piece of the puzzle. Add vlan dump support to switchdev port's bridge_getlink. iproute2 "bridge vlan show" cmd already knows how to show the vlans installed on the bridge and the device , but (until now) no one implemented the port vlan part of the netlink PF_BRIDGE:RTM_GETLINK msg. Before this patch, "bridge vlan show": $ bridge -c vlan show port vlan ids sw1p1 30-34 << bridge side vlans 57 sw1p1 << device side vlans (missing) sw1p2 57 sw1p2 sw1p3 sw1p4 br0 None (When the port is bridged, the output repeats the vlan list for the vlans on the bridge side of the port and the vlans on the device side of the port. The listing above show no vlans for the device side even though they are installed). After this patch: $ bridge -c vlan show port vlan ids sw1p1 30-34 << bridge side vlan 57 sw1p1 30-34 << device side vlans 57 3840 PVID sw1p2 57 sw1p2 57 3840 PVID sw1p3 3842 PVID sw1p4 3843 PVID br0 None I re-used ndo_dflt_bridge_getlink to add vlan fill call-back func. switchdev support adds an obj dump for VLAN objects, using the same call-back scheme as FDB dump. Support included for both compressed and un-compressed vlan dumps. Signed-off-by: Scott Feldman <sfeldma@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
This howto made sense in the 1990s when users had to manually configure ISA cards with jumpers or vendor utilities, but with the implementation of PCI it became increasingly less and less relevant, to the point where it has been well over a decade since I last updated it. And there is no value in anyone else taking over updating it either. However the references to it continue to spread as boiler plate text from one Kconfig file into the next. We are not doing end users any favours by pointing them at this old document, so lets kill it with fire, once and for all, to hopefully stop any further spread. No code is changed in this commit, just Kconfig help text. Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
There are several places in the driver (all in control paths) where coherent dma memory is being allocated using either dma_alloc_coherent() or the deprecated pci_alloc_consistent(). All these calls should be changed to use dma_zalloc_coherent() to avoid uninitialized fields in data structures backed by this memory. Reported-by: Joerg Roedel <jroedel@suse.de> Tested-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@avagotech.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts: drivers/net/phy/amd-xgbe-phy.c drivers/net/wireless/iwlwifi/Kconfig include/net/mac80211.h iwlwifi/Kconfig and mac80211.h were both trivial overlapping changes. The drivers/net/phy/amd-xgbe-phy.c file got removed in 'net-next' and the bug fix that happened on the 'net' side is already integrated into the rest of the amd-xgbe driver. Signed-off-by: David S. Miller <davem@davemloft.net>
da91309e0a7e (cpumask: Utility function to set n'th cpu...) created a genuinely weird function. I never saw it before, it went through DaveM. (He only does this to make us other maintainers feel better about our own mistakes.) cpumask_set_cpu_local_first's purpose is say "I need to spread things across N online cpus, choose the ones on this numa node first"; you call it in a loop. It can fail. One of the two callers ignores this, the other aborts and fails the device open. It can fail in two ways: allocating the off-stack cpumask, or through a convoluted codepath which AFAICT can only occur if cpu_online_mask changes. Which shouldn't happen, because if cpu_online_mask can change while you call this, it could return a now-offline cpu anyway. It contains a nonsensical test "!cpumask_of_node(numa_node)". This was drawn to my attention by Geert, who said this causes a warning on Sparc. It sets a single bit in a cpumask instead of returning a cpu number, because that's what the callers want. It could be made more efficient by passing the previous cpu rather than an index, but that would be more invasive to the callers. Fixes: da91309e0a7e8966d916a74cce42ed170fde06bf Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (then rebased) Tested-by: Amir Vadai <amirv@mellanox.com> Acked-by: Amir Vadai <amirv@mellanox.com> Acked-by: David S. Miller <davem@davemloft.net>