- 15 Jul, 2024 40 commits
-
-
Jakub Kicinski authored
First part of "net: Make timestamping selectable" from Kory Maincent. Change the driver-facing type already to lower rebasing pain. Link: https://lore.kernel.org/20240709-feature_ptp_netnext-v17-0-b5317f50df2a@bootlin.com/Signed-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Kory Maincent authored
In prevision to add new UAPI for hwtstamp we will be limited to the struct ethtool_ts_info that is currently passed in fixed binary format through the ETHTOOL_GET_TS_INFO ethtool ioctl. It would be good if new kernel code already started operating on an extensible kernel variant of that structure, similar in concept to struct kernel_hwtstamp_config vs struct hwtstamp_config. Since struct ethtool_ts_info is in include/uapi/linux/ethtool.h, here we introduce the kernel-only structure in include/linux/ethtool.h. The manual copy is then made in the function called by ETHTOOL_GET_TS_INFO. Acked-by:
Shannon Nelson <shannon.nelson@amd.com> Acked-by:
Alexandra Winter <wintera@linux.ibm.com> Signed-off-by:
Kory Maincent <kory.maincent@bootlin.com> Link: https://patch.msgid.link/20240709-feature_ptp_netnext-v17-6-b5317f50df2a@bootlin.comSigned-off-by:
-
Kory Maincent authored
Prepare for future support of saving hwtstamp source in PTP xarray by introducing HWTSTAMP_SOURCE_UNSPEC to hwtstamp_source enum, setting it to 0 to match old behavior of no source defined. Signed-off-by:
-
Kory Maincent authored
Change the API to select MAC default time stamping instead of the PHY. Indeed the PHY is closer to the wire therefore theoretically it has less delay than the MAC timestamping but the reality is different. Due to lower time stamping clock frequency, latency in the MDIO bus and no PHC hardware synchronization between different PHY, the PHY PTP is often less precise than the MAC. The exception is for PHY designed specially for PTP case but these devices are not very widespread. For not breaking the compatibility default_timestamp flag has been introduced in phy_device that is set by the phy driver to know we are using the old API behavior. Reviewed-by:
Rahul Rameshbabu <rrameshbabu@nvidia.com> Signed-off-by:
-
Kory Maincent authored
Timestamping software or hardware flags are often used as a group, therefore adding these masks will ease future use. I did not use SOF_TIMESTAMPING_SYS_HARDWARE flag as it is deprecated and not used at all. Reviewed-by:
Willem de Bruijn <willemb@google.com> Reviewed-by:
Florian Fainelli <florian.fainelli@broadcom.com> Signed-off-by:
-
Aleksander Jan Bajkowski authored
The same name is set when allocating the netdevice structure in the alloc_etherdev_mq()->alloc_etherrdev_mqs() function. Therefore, there is no need to manually set it. This fixes CheckPatch warnings: WARNING: Prefer strscpy over strcpy - see: https://github.com/KSPP/linux/issues/88 strcpy(dev->name, "eth%d"); Signed-off-by:
Aleksander Jan Bajkowski <olek2@wp.pl> Link: https://patch.msgid.link/20240713170920.863171-1-olek2@wp.plSigned-off-by:
-
Jakub Kicinski authored
Pawel Dembicki says: ==================== net: dsa: vsc73xx: Implement VLAN operations This patch series is a result of splitting a larger patch series [0], where some parts was merged before. The first patch implements port state configuration, which is required for bridge functionality. STP frames are not forwarded at this moment. BPDU frames are only forwarded from/to the PI/SI interface. For more information, see chapter 2.7.1 (CPU Forwarding) in the datasheet. Patches 2, 7-9 and 11 provide a basic implementation of tag_8021q functionality with QinQ support, without VLAN filtering in the bridge and simple VLAN awareness in VLAN filtering mode. Patches 3-6 came from Vladimir Oltean. They prepare for making tag8021q more common. VSC73XX uses very similar tag recognition, and some code from tag_sja1105 could be moved to tag_8021q for common use. Patch 10 is preparation for use tag_8021q bridge functions as generic implementation of the 'ds->ops->port_bridge_*()'. Patch 12 is required to avoid problem with learning on standalone ports. [0] https://patchwork.kernel.org/project/netdevbpf/list/?series=841034&state=%2A&archive=both ==================== Link: https://patch.msgid.link/20240713211620.1125910-1-paweldembicki@gmail.comSigned-off-by:
-
Pawel Dembicki authored
This patch implements .port_pre_bridge_flags() and .port_bridge_flags(), which are required for properly treating the BR_LEARNING flag. Also, .port_stp_state_set() is tweaked and now disables learning for standalone ports. Disabling learning for standalone ports is required to avoid situations where one port sees traffic originating from another, which could cause packet drops. Signed-off-by:
Pawel Dembicki <paweldembicki@gmail.com> Acked-by:
Linus Walleij <linus.walleij@linaro.org> Reviewed-by:
Vladimir Oltean <olteanv@gmail.com> Link: https://patch.msgid.link/20240713211620.1125910-13-paweldembicki@gmail.comSigned-off-by:
-
Pawel Dembicki authored
This patch adds bridge support for the vsc73xx driver. The vsc73xx requires minimal operations and ithe generic dsa_tag_8021q_bridge_* API is sufficient. The forwarding matrix is managed by vsc73xx_port_stp_state_set() -> vsc73xx_refresh_fwd_map()i routine, which is called immediately after .port_bridge_join() and .port_bridge_leave(). Reviewed-by:
-
Pawel Dembicki authored
The 'dsa_tag_8021q_bridge_join' could be used as a generic implementation of the 'ds->ops->port_bridge_join()' function. However, it is necessary to synchronize their arguments. This patch also moves the 'tx_fwd_offload' flag configuration line into 'dsa_tag_8021q_bridge_join' body. Currently, every (sja1105) driver sets it, and the future vsc73xx implementation will also need it for simplification. Suggested-by:
-
Pawel Dembicki authored
Max number of bridges in tag8021q implementation is strictly limited by VBID size: 3 bits. But zero is reserved and only 7 values can be used. This patch adds define which describe maximum possible value. Suggested-by:
-
Pawel Dembicki authored
This patch is a simple implementation of 802.1q tagging in the vsc73xx driver. Currently, devices with DSA_TAG_PROTO_NONE are not functional. The VSC73XX family doesn't provide any tag support for external Ethernet ports. The only option available is VLAN-based tagging, which requires constant hardware VLAN filtering. While the VSC73XX family supports provider bridging, it only supports QinQ without full implementation of 802.1AD. This means it only allows the doubled 0x8100 TPID. In the simple port mode, QinQ is enabled to preserve forwarding of VLAN-tagged frames. Signed-off-by:
-
Pawel Dembicki authored
This commit introduces a new tagger based on 802.1q tagging. It's designed for the vsc73xx driver. The VSC73xx family doesn't have any tag support for the RGMII port, but it could be based on VLANs. Reviewed-by:
-
Vladimir Oltean authored
A new tagging protocol implementation based on tag_8021q is on the horizon, and it appears that it also has to open-code the complicated logic of finding a source port based on a VLAN header. Create a single dsa_tag_8021q_find_user() and make sja1105 call it. Signed-off-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by:
-
Vladimir Oltean authored
Now that dsa_8021q_rcv() handles better the case where we don't overwrite the precise source information if it comes from an external (non-tag_8021q) source, we can now unify the call sequence between sja1105_rcv() and sja1110_rcv(). This is a preparatory change for creating a higher-level wrapper for the entire sequence which will live in tag_8021q. Signed-off-by:
-
Vladimir Oltean authored
tag_sja1105 has a wrapper over dsa_8021q_rcv(): sja1105_vlan_rcv(), which determines whether the packet came from a bridge with vlan_filtering=1 (the case resolved via dsa_find_designated_bridge_port_by_vid()), or if it contains a tag_8021q header. Looking at a new tagger implementation for vsc73xx, based also on tag_8021q, it is becoming clear that the logic is needed there as well. So instead of forcing each tagger to wrap around dsa_8021q_rcv(), let's merge the logic into the core. Signed-off-by:
-
Vladimir Oltean authored
In both sja1105_rcv() and sja1110_rcv(), we may have precise source port information coming from parallel hardware mechanisms, in addition to the tag_8021q header. Only sja1105_rcv() has extra logic to not overwrite that precise info with what's present in the VLAN tag. This is because sja1110_rcv() gets by, by having a reversed set of checks when assigning skb->dev. When the source port is imprecise (vbid >=1), source_port and switch_id will be set to zeroes by dsa_8021q_rcv(), which might be problematic. But by checking for vbid >= 1 first, sja1110_rcv() fends that off. We would like to make more code common between sja1105_rcv() and sja1110_rcv(), and for that, we need to make sure that sja1110_rcv() also goes through the precise source port preservation logic. Signed-off-by:
-
Pawel Dembicki authored
This patch implements VLAN filtering for the vsc73xx driver. After starting VLAN filtering, the switch is reconfigured from QinQ to a simple VLAN aware mode. This is required because VSC73XX chips do not support inner VLAN tag filtering. Signed-off-by:
-
Pawel Dembicki authored
This isn't a fully functional implementation of 802.1D, but port_stp_state_set is required for a future tag8021q operations. This implementation handles properly all states, but vsc73xx doesn't forward STP packets. Reviewed-by:
-
MD Danish Anwar authored
icssg_prueth.c and icssg_prueth_sr1.c drivers use multiple common .c files. These common objects are getting added to multiple modules. As a result when both drivers are enabled in .config, below warning is seen. drivers/net/ethernet/ti/Makefile: icssg/icssg_common.o is added to multiple modules: icssg-prueth icssg-prueth-sr1 drivers/net/ethernet/ti/Makefile: icssg/icssg_classifier.o is added to multiple modules: icssg-prueth icssg-prueth-sr1 drivers/net/ethernet/ti/Makefile: icssg/icssg_config.o is added to multiple modules: icssg-prueth icssg-prueth-sr1 drivers/net/ethernet/ti/Makefile: icssg/icssg_mii_cfg.o is added to multiple modules: icssg-prueth icssg-prueth-sr1 drivers/net/ethernet/ti/Makefile: icssg/icssg_stats.o is added to multiple modules: icssg-prueth icssg-prueth-sr1 drivers/net/ethernet/ti/Makefile: icssg/icssg_ethtool.o is added to multiple modules: icssg-prueth icssg-prueth-sr1 Fix this by building a new module (icssg.o) for all the common objects. Both the driver can then depend on this common module. Some APIs being exported have emac_ as the prefix which may result into confusion with other existing APIs with emac_ prefix, to avoid confusion, rename the APIs being exported with emac_ to icssg_ prefix. This also fixes below error seen when both drivers are built. ERROR: modpost: "icssg_queue_pop" [drivers/net/ethernet/ti/icssg-prueth-sr1.ko] undefined! ERROR: modpost: "icssg_queue_push" [drivers/net/ethernet/ti/icssg-prueth-sr1.ko] undefined! Reported-and-tested-by:
Thorsten Leemhuis <linux@leemhuis.info> Closes: https://lore.kernel.org/oe-kbuild-all/202405182038.ncf1mL7Z-lkp@intel.com/ Fixes: 487f7323 ("net: ti: icssg-prueth: Add helper functions to configure FDB") Reviewed-by:
Roger Quadros <rogerq@kernel.org> Signed-off-by:
MD Danish Anwar <danishanwar@ti.com> Reviewed-by:
Sai Krishna <saikrishnag@marvell.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Breno Leitao authored
After the commit bdacf3e3 ("net: Use nested-BH locking for napi_alloc_cache.") was merged, the following warning began to appear: WARNING: CPU: 5 PID: 1 at net/core/skbuff.c:1451 napi_skb_cache_put+0x82/0x4b0 __warn+0x12f/0x340 napi_skb_cache_put+0x82/0x4b0 napi_skb_cache_put+0x82/0x4b0 report_bug+0x165/0x370 handle_bug+0x3d/0x80 exc_invalid_op+0x1a/0x50 asm_exc_invalid_op+0x1a/0x20 __free_old_xmit+0x1c8/0x510 napi_skb_cache_put+0x82/0x4b0 __free_old_xmit+0x1c8/0x510 __free_old_xmit+0x1c8/0x510 __pfx___free_old_xmit+0x10/0x10 The issue arises because virtio is assuming it's running in NAPI context even when it's not, such as in the netpoll case. To resolve this, modify virtnet_poll_tx() to only set NAPI when budget is available. Same for virtnet_poll_cleantx(), which always assumed that it was in a NAPI context. Fixes: df133f3f ("virtio_net: bulk free tx skbs") Suggested-by:
Breno Leitao <leitao@debian.org> Reviewed-by:
Michael S. Tsirkin <mst@redhat.com> Acked-by:
Jason Wang <jasowang@redhat.com> Reviewed-by:
Heng Qi <hengqi@linux.alibaba.com> Link: https://patch.msgid.link/20240712115325.54175-1-leitao@debian.orgSigned-off-by:
-
Jakub Kicinski authored
Kamil Horák says: ==================== net: phy: bcm5481x: add support for BroadR-Reach mode PATCH 1 - Add the 10baseT1BRR_Full link mode PATCH 2 - Add the definitions of LRE registers, necessary to use BroadR-Reach modes on the BCM5481x PHY PATCH 3 - Add brr-mode flag to switch between IEEE802.3 and BroadR-Reach PATCH 4 - Implementation of the BroadR-Reach modes for the Broadcom PHYs ==================== Link: https://patch.msgid.link/20240712150709.3134474-1-kamilh@axis.comSigned-off-by:
-
Kamil Horák (2N) authored
Implement single-pair BroadR-Reach modes on bcm5481x PHY by Broadcom. Create set of functions alternative to IEEE 802.3 to handle configuration of these modes on compatible Broadcom PHYs. There is only subset of capabilities supported because of limited collection of hardware available for the development. For BroadR-Reach capable PHYs, the LRE (Long Reach Ethernet) alternative register set is handled. Only bcm54811 PHY is verified, for bcm54810, there is some support possible but untested. There is no auto-negotiation of the link parameters (called LDS in the Broadcom terminology, Long-Distance Signaling) for bcm54811. It should be possible to enable LDS for bcm54810. Signed-off-by:
Kamil Horák (2N) <kamilh@axis.com> Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Reviewed-by:
-
Kamil Horák (2N) authored
There is a group of PHY chips supporting BroadR-Reach link modes in a manner allowing for more or less identical register usage as standard Clause 22 PHY. These chips support standard Ethernet link modes as well, however, the circuitry is mutually exclusive and cannot be auto-detected. The link modes in question are 100Base-T1 as defined in IEEE802.3bw, based on Broadcom's 1BR-100 link mode, and newly defined 10Base-T1BRR (1BR-10 in Broadcom documents). Add optional brr-mode flag to switch the PHY to BroadR-Reach mode. Signed-off-by:
Rob Herring (Arm) <robh@kernel.org> Reviewed-by:
-
Kamil Horák (2N) authored
Add the definitions of LRE registers for Broadcom BCM5481x PHY Signed-off-by:
-
Kamil Horák (2N) authored
Introduce a new link mode necessary for 10 MBit single-pair connection in BroadR-Reach mode on bcm5481x PHY by Broadcom. This new link mode, 10baseT1BRR, is known as 1BR10 in the Broadcom terminology. Another link mode to be used is 1BR100 and it is already present as 100baseT1, because Broadcom's 1BR100 became 100baseT1 (IEEE 802.3bw). Signed-off-by:
-
Jakub Kicinski authored
Xuan Zhuo says: ==================== virtio-net: support AF_XDP zero copy v5: http://lore.kernel.org/all/20240611114147.31320-1-xuanzhuo@linux.alibaba.com XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero copy feature of xsk (XDP socket) needs to be supported by the driver. The performance of zero copy is very good. mlx5 and intel ixgbe already support this feature, This patch set allows virtio-net to support xsk's zerocopy xmit feature. At present, we have completed some preparation: 1. vq-reset (virtio spec and kernel code) 2. virtio-core premapped dma 3. virtio-net xdp refactor So it is time for Virtio-Net to complete the support for the XDP Socket Zerocopy. Virtio-net can not increase the queue num at will, so xsk shares the queue with kernel. On the other hand, Virtio-Net does not support generate interrupt from driver manually, so when we wakeup tx xmit, we used some tips. If the CPU run by TX NAPI last time is other CPUs, use IPI to wake up NAPI on the remote CPU. If it is also the local CPU, then we wake up napi directly. This patch set includes some refactor to the virtio-net to let that to support AF_XDP. Because there are too many commits, the work of virtio net supporting af-xdp is split to rx part and tx part. This patch set is for rx part. So the flag NETDEV_XDP_ACT_XSK_ZEROCOPY is not added, if someone want to test for af-xdp rx, the flag needs to be adding locally. ENV: Qemu with vhost-user(polling mode). Host CPU: Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz testpmd> show port stats all ######################## NIC statistics for port 0 ######################## RX-packets: 19531092064 RX-missed: 0 RX-bytes: 1093741155584 RX-errors: 0 RX-nombuf: 0 TX-packets: 5959955552 TX-errors: 0 TX-bytes: 371030645664 Throughput (since last show) Rx-pps: 8861574 Rx-bps: 3969985208 Tx-pps: 8861493 Tx-bps: 3969962736 ############################################################################ testpmd> show port stats all ######################## NIC statistics for port 0 ######################## RX-packets: 68152727 RX-missed: 0 RX-bytes: 3816552712 RX-errors: 0 RX-nombuf: 0 TX-packets: 68114967 TX-errors: 33216 TX-bytes: 3814438152 Throughput (since last show) Rx-pps: 6333196 Rx-bps: 2837272088 Tx-pps: 6333227 Tx-bps: 2837285936 ############################################################################ But AF_XDP consumes more CPU for tx and rx napi(100% and 86%). ==================== Link: https://patch.msgid.link/20240708112537.96291-1-xuanzhuo@linux.alibaba.comSigned-off-by:
-
Xuan Zhuo authored
Support AF-XDP for merge mode. Signed-off-by:
Xuan Zhuo <xuanzhuo@linux.alibaba.com> Acked-by:
-
Xuan Zhuo authored
In the process: 1. We may need to copy data to create skb for XDP_PASS. 2. We may need to call xsk_buff_free() to release the buffer. 3. The handle for xdp_buff is difference from the buffer. If we pushed this logic into existing receive handle(merge and small), we would have to maintain code scattered inside merge and small (and big). So I think it is a good choice for us to put the xsk code into an independent function. Signed-off-by:
-
Xuan Zhuo authored
Implement the logic of filling rq with XSK buffers. Signed-off-by:
-
Xuan Zhuo authored
xsk wakeup is used to trigger the logic for xsk xmit by xsk framework or user. Virtio-net does not support to actively generate an interruption, so it tries to trigger tx NAPI on the local cpu. Signed-off-by:
-
Xuan Zhuo authored
This patch implement the logic of bind/unbind xsk pool to rq. Signed-off-by:
-
Xuan Zhuo authored
This commit separates the function receive_mergeable(), put the logic of appending frag to the skb as an independent function. The subsequent commit will reuse it. Signed-off-by:
-
Xuan Zhuo authored
This commit separates the function receive_buf(), then we wrap the logic of handling the skb to an independent function virtnet_receive_done(). The subsequent commit will reuse it. Signed-off-by:
-
Xuan Zhuo authored
This patch separates two sub-functions from virtnet_tx_resize(): * virtnet_tx_pause * virtnet_tx_resume Then the subsequent virtnet_tx_reset() can share these two functions. Signed-off-by:
-
Xuan Zhuo authored
This patch separates two sub-functions from virtnet_rx_resize(): * virtnet_rx_pause * virtnet_rx_resume Then the subsequent reset rx for xsk can share these two functions. Signed-off-by:
-
Xuan Zhuo authored
virtio net has VIRTIO_XDP_HEADROOM that is equal to XDP_PACKET_HEADROOM to calculate the headroom for xdp. But here we should use the macro XDP_PACKET_HEADROOM from bpf.h to calculate the headroom for xdp. So here we remove the VIRTIO_XDP_HEADROOM, and use the XDP_PACKET_HEADROOM to replace it. Signed-off-by:
-
Jakub Kicinski authored
Vladimir Oltean says: ==================== Eliminate CONFIG_NR_CPUS dependency in dpaa-eth and enable COMPILE_TEST in fsl_qbman Breno's previous attempt at enabling COMPILE_TEST for the fsl_qbman driver (now included here as patch 5/5) triggered compilation warnings for large CONFIG_NR_CPUS values: https://lore.kernel.org/all/202406261920.l5pzM1rj-lkp@intel.com/ Patch 1/5 switches two NR_CPUS arrays in the dpaa-eth driver to dynamic allocation to avoid that warning. There is more NR_CPUS usage in the fsl-qbman driver, but that looks relatively harmless and I couldn't find a good reason to change it. I noticed, while testing, that the driver doesn't actually work properly with high CONFIG_NR_CPUS values, and patch 2/5 addresses that. During code analysis, I have identified two places which treat conditions that can never happen. Patches 3/5 and 4/5 simplify the probing code - dpaa_fq_setup() - just a little bit. Finally we have at 5/5 the patch that triggered all of this. There is an okay from Herbert to take it via netdev, despite it being on soc/qbman: https://lore.kernel.org/all/Zns%2FeVVBc7pdv0yM@gondor.apana.org.au/ Link to v1: https://lore.kernel.org/netdev/20240710230025.46487-1-vladimir.oltean@nxp.com/ ==================== Link: https://patch.msgid.link/20240713225336.1746343-1-vladimir.oltean@nxp.comSigned-off-by:
-
Breno Leitao authored
As most of the drivers that depend on ARCH_LAYERSCAPE, make FSL_DPAA depend on COMPILE_TEST for compilation and testing. # grep -r depends.\*ARCH_LAYERSCAPE.\*COMPILE_TEST | wc -l 29 Signed-off-by:
Madalin Bucur <madalin.bucur@oss.nxp.com> Link: https://patch.msgid.link/20240713225336.1746343-6-vladimir.oltean@nxp.comSigned-off-by:
-
Vladimir Oltean authored
dpaa_fq_setup() iterates through the &priv->dpaa_fq_list elements allocated by dpaa_alloc_all_fqs(). This includes a call to: if (!dpaa_fq_alloc(dev, 0, dpaa_max_num_txqs(), list, FQ_TYPE_TX)) goto fq_alloc_failed; which gives us dpaa_max_num_txqs() elements of FQ_TYPE_TX type. The code block which we are deleting runs after an earlier iteration through &priv->dpaa_fq_list. So at the end of this iteration (for which there is no early break), egress_cnt will be unconditionally equal to dpaa_max_num_txqs(). In other words, dpaa_alloc_all_fqs() has already allocated TX queues for all possible CPUs and the maximal number of traffic classes, and we've already iterated once through them all. The while() condition is dead code, remove it. Signed-off-by:
-
