- 29 Aug, 2016 26 commits
-
-
Jacob Keller authored
When the PF assigns a new MAC address to a VF it uses the base address registers to store the MAC address. This allows a VF which loads after this setup the ability to get the initial address without having to wait for a mailbox message. Unfortunately to do this, the PF must take queue ownership away from the VF, which can cause fault errors when there is already an active VF driver. This queue ownership assignment causes race condition between the PF and the VF such that potentially a VF can cause FUM fault errors due to normal PF/VF driver behavior. It is not safe to simply allow the PF to write the base address registers without taking queue ownership back as the PF must also disable the queues, and this would impact active VF use. The current code is safe because the queue ownership will prevent the VF from actually writing but does trigger the FUM fault. We can do better by simply avoiding the register write process when a mailbox message suffices. If the message can be sent over the mailbox, then we will not perform the queue ownership assignment and we won't update the base address to be the same as the MAC address. We do still have to write the TXQCTL registers in order to update the VID of the queue. This is necessary because the TXQCTL register is read-only from the VF, and thus the VF cannot do this for itself. This register does not need to wait for the Tx queue to be disabled and is safe for the PF to write during normal VF operation, so we move this write to the top of the function above the mailbox message. Without this, the TXQCTL register would be misconfigured and cause the VF to Tx hang. Signed-off-by:
Jacob Keller <jacob.e.keller@intel.com> Tested-by:
Krishneil Singh <Krishneil.k.singh@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
-
Jacob Keller authored
Ensure that other bits in the RXQCTL register do not get cleared. This ensures that bits related to queue ownership are maintained. Signed-off-by:
-
Jacob Keller authored
Signed-off-by:
-
Jacob Keller authored
Similar to how we handle VXLAN offload, enable support for a single Geneve tunnel. Signed-off-by:
-
Jacob Keller authored
In preparation for adding Geneve Rx offload support, refactor the current VXLAN offload flow to be a bit more generic so that it will be easier to add the new Geneve code. The fm10k hardware supports one VXLAN and one Geneve tunnel, so we will eventually treat the VXLAN and Geneve tunnels identically. To this end, factor out the code that handles the current list so that we can use the generic flow for both tunnels in the next patch. Signed-off-by:
-
Jacob Keller authored
In the event of a surprise remove, we expect the driver to go down, which includes calling .stop_hw(). However, this function will return an error because the queues won't appear to cleanly disable. Prevent this and avoid the unnecessary checks by just returning when FM10K_REMOVED(hw->hw_addr) is true. Signed-off-by:
-
Jacob Keller authored
In the event of an uncorrectable AER error occurring when the driver has not loaded, the recovery routines are not done. This is done because future loads of the driver may not be aware of the IO state and may not be able to recover at all. In this case, when we next load the driver it fails due to what appears to be a surprise remove event. Instead, add a check to ensure that the device is in the normal IO state before continuing to probe. This allows us to give a more descriptive message of what is wrong. Without this change, the driver will attempt to probe up to our first call of .reset_hw() which will be unable to read registers and act as if a surprise remove event occurred. Signed-off-by:
-
Jacob Keller authored
Signed-off-by:
-
Jacob Keller authored
When fm10k_poll fully cleans rings it returns 0. This is incorrect as it messes up the budget accounting in the core NAPI code. Fix this by returning actual work done, capped at budget - 1 since the core doesn't expect a return of the full budget when the driver modifies the NAPI status. Cc: Paolo Abeni <pabeni@redhat.com> Cc: Venkatesh Srinivas <venkateshs@google.com> Signed-off-by:
Paolo Abeni <pabeni@redhat.com> Tested-by:
-
Jacob Keller authored
While technically not needed, as all our uses of ACCESS_ONCE are scalar types, we already use READ_ONCE in a few places, and for code readability we can swap all the uses of the older ACCESS_ONCE into READ_ONCE. Signed-off-by:
-
Jacob Keller authored
The function is only used in fm10k_ethtool.c, so make it static. Signed-off-by:
-
Jacob Keller authored
Signed-off-by:
-
Jacob Keller authored
A previous patch added support to check for hardware Tx pending in the fm10k_down routine. This support was intended to ensure that we accurately check what the hardware state is. However, checking for Tx hangs in this manor during the hotpath results in a large performance hit. Avoid this by making the hotpath check use the SW counters instead. Fixes: a0f53cf49cb0 ("fm10k: use actual hardware registers when checking for pending Tx", 2016-06-08) Signed-off-by: -
Jacob Keller authored
A previous patch removed the pci_disable_device() call in .io_error_detected. This call corresponded to a pci_enable_device_mem() call within .io_slot_reset handler. Change the call here to a pci_reenable_device() so that it does not increment and leak the enable_cnt reference count for the device. Without this change, VF devices may fail during an unbind/bind, and we'll never zero the reference counter for the pci_dev structure. Signed-off-by:
-
Arnd Bergmann authored
The addition of VLAN support caused a possible use of uninitialized data if we encounter a zero TCA_FLOWER_KEY_ETH_TYPE key, as pointed out by "gcc -Wmaybe-uninitialized": net/sched/cls_flower.c: In function 'fl_change': net/sched/cls_flower.c:366:22: error: 'ethertype' may be used uninitialized in this function [-Werror=maybe-uninitialized] This changes the code to only set the ethertype field if it was nonzero, as before the patch. Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Fixes: 9399ae9a ("net_sched: flower: Add vlan support") Cc: Hadar Hen Zion <hadarh@mellanox.com> Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Arnd Bergmann authored
The newly added reset logic uses helper functions for the MMIO that may fail. However, when the read operation fails, we end up writing back uninitialized data to the register, as gcc warns: drivers/net/ethernet/apm/xgene/xgene_enet_xgmac.c: In function 'xgene_enet_link_state': drivers/net/ethernet/apm/xgene/xgene_enet_xgmac.c:213:2: error: 'data' may be used uninitialized in this function [-Werror=maybe-uninitialized] drivers/net/ethernet/apm/xgene/xgene_enet_xgmac.c:209:6: note: 'data' was declared here u32 data; We already print a warning to the console log if that happens, the best alternative that I can see is skip the rest of the reset sequence if the register value cannot be read: Most likely the write would fail as well, and if it succeeded, worse things could happen. Signed-off-by:
-
Vidya Sagar Ravipati authored
This patch enhances ethtool link mode bitmap to include missing interface modes for 1G/10G speeds Changes: 1000baseX is the mode introduced to cover all 1G Fiber cases. All modes under 1000BaseX i.e. 1000BASE-SX, 1000BASE-LX, 1000BASE-LX10 and 1000BASE-BX10 are not explicitly defined at this moment. 10G CR,SR,LR and ER link modes are included for 10G speed.. Issue: ethtool on 1G/10G SFP port reports Base-T as this port supports 1000baseX,10G CR, SR and LR modes. root@tor-02$ ethtool swp1 Settings for swp1: Supported ports: [ FIBRE ] Supported link modes: 1000baseT/Full 10000baseT/Full Supported pause frame use: Symmetric Receive-only Supports auto-negotiation: Yes Advertised link modes: 1000baseT/Full Advertised pause frame use: No Advertised auto-negotiation: No Speed: 10000Mb/s Duplex: Full Port: FIBRE PHYAD: 0 Transceiver: external Auto-negotiation: off Current message level: 0x00000000 (0) Link detected: yes After fix: root@tor-02$ ethtool swp1 Settings for swp1: Supported ports: [ FIBRE ] Supported link modes: 1000baseX/Full 10000baseCR/Full 10000baseSR/Full 10000baseLR/Full 10000baseER/Full Supported pause frame use: Symmetric Receive-only Supports auto-negotiation: Yes Advertised link modes: 1000baseT/Full Advertised pause frame use: No Advertised auto-negotiation: No Speed: 10000Mb/s Duplex: Full Port: FIBRE PHYAD: 0 Transceiver: external Auto-negotiation: off Current message level: 0x00000000 (0) Link detected: yes Signed-off-by:Vidya Sagar Ravipati <vidya@cumulusnetworks.com> Signed-off-by:
-
James Morse authored
After resume from hibernate on arm64, any amd-xgbe devices that were running when we hibernated are reported as down, even when it is not. Re-plugging the cables does not cause the interface to come back, the link must be marked as down then up via 'ip set link' using the serial console. This happens because the device has been power-cycled and possibly re-initialised by firmware, whereas the driver's memory structures have been restored from the hibernate image and the two do not agree. Schedule a restart of the device after powerup in case the world changed while we were asleep. Signed-off-by:
James Morse <james.morse@arm.com> Signed-off-by:
-
Eric Dumazet authored
When TCP operates in lossy environments (between 1 and 10 % packet losses), many SACK blocks can be exchanged, and I noticed we could drop them on busy senders, if these SACK blocks have to be queued into the socket backlog. While the main cause is the poor performance of RACK/SACK processing, we can try to avoid these drops of valuable information that can lead to spurious timeouts and retransmits. Cause of the drops is the skb->truesize overestimation caused by : - drivers allocating ~2048 (or more) bytes as a fragment to hold an Ethernet frame. - various pskb_may_pull() calls bringing the headers into skb->head might have pulled all the frame content, but skb->truesize could not be lowered, as the stack has no idea of each fragment truesize. The backlog drops are also more visible on bidirectional flows, since their sk_rmem_alloc can be quite big. Let's add some room for the backlog, as only the socket owner can selectively take action to lower memory needs, like collapsing receive queues or partial ofo pruning. Signed-off-by:
Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Cc: Neal Cardwell <ncardwell@google.com> Acked-by:
Neal Cardwell <ncardwell@google.com> Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in dev_warn message. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in dev_warn message. Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in dev_err message. Signed-off-by:
-
David S. Miller authored
Tom Herbert says: ==================== strp: Generalize stream parser to work with other socket types Add a read_sock protocol operation function that allows something like tcp_read_sock to be called for other protocol types. Specific changes in this patch set: - Add read_sock function to proto_ops. This has the same signature as tcp_read_sock. sk_read_actor_t is also defined in net.h. - Set peek_len and read_sock proto_op functions for TCPv4 and TCPv6 stream ops. - Remove references to tcp in strparser. - Call peek_len and read_sock operations from strparser instead of calling TCP specific functions. ==================== Signed-off-by: -
Tom Herbert authored
kcm and strparser need to work with any type of stream socket not just TCP. Eliminate references to TCP and call generic proto_ops functions of read_sock and peek_len. Also in strp_init check if the socket support the proto_ops read_sock and peek_len. Signed-off-by:
Tom Herbert <tom@herbertland.com> Signed-off-by:
-
Tom Herbert authored
In inet_stream_ops we set read_sock to tcp_read_sock and peek_len to tcp_peek_len (which is just a stub function that calls tcp_inq). Signed-off-by:
-
Tom Herbert authored
Add new function in proto_ops structure. This includes moving the typedef got sk_read_actor into net.h and removing the definition from tcp.h. Signed-off-by:
-
- 27 Aug, 2016 14 commits
-
-
Wei Yongjun authored
Fix to return a negative error code from the cpsw_fill_rx_channels() error handling case instead of 0, as done elsewhere in this function. Fixes: ce52c744 ("net: ethernet: ti: cpsw: add ethtool channels support") Signed-off-by:
Wei Yongjun <weiyongjun1@huawei.com> Signed-off-by:
-
Zhu Yanjun authored
This header file is not used in vxlan.c file. Signed-off-by:
Zhu Yanjun <zyjzyj2000@gmail.com> Signed-off-by:
-
Wei Yongjun authored
Fixes the following sparse warning: drivers/crypto/chelsio/chcr_algo.c:593:5: warning: symbol 'cxgb4_is_crypto_q_full' was not declared. Should it be static? Signed-off-by:
Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by:
-
Jon Cooper authored
MC_CMD_TRIGGER_INTERRUPT does not work on the SFC9140, as used in the sfn7x42q and sfn7x24f. Check for this using the MCDI workaround mechanism. The command is only used during self test. If it's not supported, skip the interrupt test. Signed-off-by:
Edward Cree <ecree@solarflare.com> Signed-off-by:
-
Edward Cree authored
nic_data was already initialised to the right thing, no need to assign it again. Signed-off-by:
-
Edward Cree authored
TX size bins were not supported on the 7000's 40G MAC, but the 8000 series does support them and the MCPU advertises that via a new capability bit. Signed-off-by:
-
David S. Miller authored
Richard Alpe says: ==================== tipc: introduce UDP replicast This series introduces UDP replicast. A concept where we emulate multicast by sending multiple unicast messages to configured peers. This allows TIPC to be used in environments where IP multicast is disabled. There is a corresponding patch series for the tipc user space tool that allows a user to add remote addresses to the replicast list. ==================== Signed-off-by: -
Richard Alpe authored
When using replicast a UDP bearer can have an arbitrary amount of remote ip addresses associated with it. This means we cannot simply add all remote ip addresses to an existing bearer data message as it might fill the message, leaving us with a truncated message that we can't safely resume. To handle this we introduce the new netlink command TIPC_NL_UDP_GET_REMOTEIP. This command is intended to be called when the bearer data message has the TIPC_NLA_UDP_MULTI_REMOTEIP flag set, indicating there are more than one remote ip (replicast). Signed-off-by:
Richard Alpe <richard.alpe@ericsson.com> Reviewed-by:
Jon Maloy <jon.maloy@ericsson.com> Signed-off-by:
-
Richard Alpe authored
Add UDP bearer options to netlink bearer get message. This is used by the tipc user space tool to display UDP options. The UDP bearer information is passed using either a sockaddr_in or sockaddr_in6 structs. This means the user space receiver should intermediately store the retrieved data in a large enough struct (sockaddr_strage) before casting to the proper IP version type. Signed-off-by:
Ying Xue <ying.xue@windriver.com> Signed-off-by:
-
Richard Alpe authored
Automatically learn UDP remote IP addresses of communicating peers by looking at the source IP address of incoming TIPC link configuration messages (neighbor discovery). This makes configuration slightly easier and removes the problematic scenario where a node receives directly addressed neighbor discovery messages sent using replicast which the node cannot "reply" to using mutlicast, leaving the link FSM in a limbo state. Signed-off-by:
-
Richard Alpe authored
This patch introduces UDP replicast. A concept where we emulate multicast by sending multiple unicast messages to configured peers. The purpose of replicast is mainly to be able to use TIPC in cloud environments where IP multicast is disabled. Using replicas to unicast multicast messages is costly as we have to copy each skb and send the copies individually. Signed-off-by:
-
Richard Alpe authored
Add a function to check if a tipc UDP media address is a multicast address or not. This is a purely cosmetic change. Signed-off-by:
-
Richard Alpe authored
Split the UDP send function into two. One callback that prepares the skb and one transmit function that sends the skb. This will come in handy in later patches, when we introduce UDP replicast. Signed-off-by:
-
Richard Alpe authored
Split the UDP netlink parse function so that it only parses one netlink attribute at the time. This makes the parse function more generic and allow future UDP API functions to use it for parsing. Signed-off-by:
-
