- 13 Sep, 2023 22 commits
-
-
Alan Brady authored
Initialize all the ethtool ops that are supported by the driver and add the necessary support for the ethtool callbacks. Also add asynchronous link notification virtchnl support where the device Control Plane sends the link status and link speed as an asynchronous event message. Driver report the link speed on ethtool .idpf_get_link_ksettings query. Introduce soft reset function which is used by some of the ethtool callbacks such as .set_channels, .set_ringparam etc. to change the existing queue configuration. It deletes the existing queues by sending delete queues virtchnl message to the CP and calls the 'vport_stop' flow which disables the queues, vport etc. New set of queues are requested to the CP and reconfigure the queue context by calling the 'vport_open' flow. Soft reset flow also adjusts the number of vectors associated to a vport if .set_channels is called. Signed-off-by:
Alan Brady <alan.brady@intel.com> Co-developed-by:
Alice Michael <alice.michael@intel.com> Signed-off-by:
Joshua Hay <joshua.a.hay@intel.com> Signed-off-by:
Phani Burra <phani.r.burra@intel.com> Signed-off-by:
Sridhar Samudrala <sridhar.samudrala@intel.com> Reviewed-by:
Willem de Bruijn <willemb@google.com> Co-developed-by:
Pavan Kumar Linga <pavan.kumar.linga@intel.com> Signed-off-by:
Tony Nguyen <anthony.l.nguyen@intel.com>
-
Joshua Hay authored
Add the start_xmit, TX and RX napi poll support for the single queue model. Unlike split queue model, single queue uses same queue to post buffer descriptors and completed descriptors. Signed-off-by:
Madhu Chittim <madhu.chittim@intel.com> Signed-off-by:
-
Alan Brady authored
Add support to handle interrupts for the RX completion queue and RX buffer queue. When the interrupt fires on RX completion queue, process the RX descriptors that are received. Allocate and prepare the SKB with the RX packet info, for both data and header buffer. IDPF uses software maintained refill queues to manage buffers between RX queue producer and the buffer queue consumer. They are required in order to maintain a lockless buffer management system and are strictly software only constructs. Instead of updating the RX buffer queue tail with available buffers right after the clean routine, it posts the buffer ids to the refill queues, only to post them to the HW later. If the generic receive offload (GRO) is enabled in the capabilities and turned on by default or via ethtool, then HW performs the packet coalescing if certain criteria are met by the incoming packets and updates the RX descriptor. Similar to GRO, if generic checksum is enabled, HW computes the checksum and updates the respective fields in the descriptor. Add support to update the SKB fields with the GRO and the generic checksum received. Signed-off-by:
Jakub Kicinski <kuba@kernel.org> Co-developed-by:
-
Joshua Hay authored
Add support to handle the interrupts for the TX completion queue and process the various completion types. In the flow scheduling mode, the driver processes primarily buffer completions as well as descriptor completions occasionally. This mode supports out of order TX completions. To do so, HW generates one buffer completion per packet. Each of those completions contains the unique tag provided during the TX encoding which is used to locate the packet either on the TX buffer ring or in a hash table. The hash table is used to track TX buffer information so the descriptor(s) for a given packet can be reused while the driver is still waiting on the buffer completion(s). Packets end up in the hash table in one of 2 ways: 1) a packet was stashed during descriptor completion cleaning, or 2) because an out of order buffer completion was processed. A descriptor completion arrives only every so often and is primarily used to guarantee the TX descriptor ring can be reused without having to wait on the individual buffer completions. E.g. a descriptor completion for N+16 guarantees HW read all of the descriptors for packets N through N+15, therefore all of the buffers for packets N through N+15 are stashed into the hash table and the descriptors can be reused for more TX packets. Similarly, a packet can be stashed in the hash table because an out an order buffer completion was processed. E.g. processing a buffer completion for packet N+3 implies that HW read all of the descriptors for packets N through N+3 and they can be reused. However, the HW did not do the DMA yet. The buffers for packets N through N+2 cannot be freed, so they are stashed in the hash table. In either case, the buffer completions will eventually be processed for all of the stashed packets, and all of the buffers will be cleaned from the hash table. In queue based scheduling mode, the driver processes primarily descriptor completions and cleans the TX ring the conventional way. Finally, the driver triggers a TX queue drain after sending the disable queues virtchnl message. When the HW completes the queue draining, it sends the driver a queue marker packet completion. The driver determines when all TX queues have been drained and proceeds with the disable flow. With this, the driver can send TX packets and clean up the resources properly. Signed-off-by:
-
Joshua Hay authored
Add start_xmit support for split queue model. To start with, add the necessary checks to linearize the skb if it uses more number of buffers than the hardware supported limit. Stop the transmit queue if there are no enough descriptors available for the skb to use or if there we're going to potentially overrun the completion queue. Finally prepare the descriptor with all the required information and update the tail. Signed-off-by:
-
Pavan Kumar Linga authored
To further continue 'vport open', initialize all the resources required for the interrupts. To start with, initialize the queue vector indices with the ones received from the device Control Plane. Now that all the TX and RX queues are initialized, map the RX descriptor and buffer queues as well as TX completion queues to the allocated vectors. Initialize and enable the napi handler for the napi polling. Finally, request the IRQs for the interrupt vectors from the stack and setup the interrupt handler. Once the interrupt init is done, send 'map queue vector', 'enable queues' and 'enable vport' virtchnl messages to the CP to complete the 'vport open' flow. Co-developed-by:
-
Alan Brady authored
Similar to the TX, RX also supports both single and split queue models. In single queue model, the same descriptor queue is used by SW to post buffer descriptors to HW and by HW to post completed descriptors to SW. In split queue model, "RX buffer queues" are used to pass descriptor buffers from SW to HW whereas "RX queues" are used to post the descriptor completions i.e. descriptors that point to completed buffers, from HW to SW. "RX queue group" is a set of RX queues grouped together and will be serviced by a "RX buffer queue group". IDPF supports 2 buffer queues i.e. large buffer (4KB) queue and small buffer (2KB) queue per buffer queue group. HW uses large buffers for 'hardware gro' feature and also if the packet size is more than 2KB, if not 2KB buffers are used. Add all the resources required for the RX queues initialization. Allocate memory for the RX queue and RX buffer queue groups. Initialize the software maintained refill queues for buffer management algorithm. Same like the TX queues, initialize the queue parameters for the RX queues and send the config RX queue virtchnl message to the device Control Plane. Signed-off-by:
-
Alan Brady authored
IDPF supports two queue models i.e. single queue which is a traditional queueing model as well as split queue model. In single queue model, the same descriptor queue is used by SW to post descriptors to the HW, HW to post completed descriptors to SW. In split queue model, "TX Queues" are used to pass buffers from SW to HW and "TX Completion Queues" are used to post descriptor completions from HW to SW. Device supports asymmetric ratio of TX queues to TX completion queues. Considering this, queue group mechanism is used i.e. some TX queues are grouped together which will be serviced by only one TX completion queue per TX queue group. Add all the resources required for the TX queues initialization. To start with, allocate memory for the TX queue groups, TX queues and TX completion queues. Then, allocate the descriptors for both TX and TX completion queues, and bookkeeping buffers for TX queues alone. Also, allocate queue vectors for the vport and initialize the TX queue related fields for each queue vector. Initialize the queue parameters such as q_id, q_type and tail register offset with the info received from the device control plane (CP). Once all the TX queues are configured, send config TX queue virtchnl message to the CP with all the TX queue context information. Signed-off-by:
-
Pavan Kumar Linga authored
Add the virtchnl support to request the packet types. Parse the responses received from CP and based on the protocol headers, populate the packet type structure with necessary information. Initialize the MAC address and add the virtchnl support to add and del MAC address. Co-developed-by:
Shailendra Bhatnagar <shailendra.bhatnagar@intel.com> Signed-off-by:
-
Pavan Kumar Linga authored
Add the required support to create a vport by spawning the init task. Once the vport is created, initialize and allocate the resources needed for it. Configure and register a netdev for each vport with all the features supported by the device based on the capabilities received from the device Control Plane. Spawn the init task till all the default vports are created. Co-developed-by:
-
Pavan Kumar Linga authored
As the mailbox is setup, add the necessary send and receive mailbox message framework to support the virtchnl communication between the driver and device Control Plane (CP). Add the core initialization. To start with, driver confirms the virtchnl version with the CP. Once that is done, it requests and gets the required capabilities and resources needed such as max vectors, queues etc. Based on the vector information received in 'VIRTCHNL2_OP_GET_CAPS', request the stack to allocate the required vectors. Finally add the interrupt handling mechanism for the mailbox queue and enable the interrupt. Note: Checkpatch issues a warning about IDPF_FOREACH_VPORT_VC_STATE and IDPF_GEN_STRING being complex macros and should be enclosed in parentheses but it's not the case. They are never used as a statement and instead only used to define the enum and array. Co-developed-by:
Emil Tantilov <emil.s.tantilov@intel.com> Signed-off-by:
-
Joshua Hay authored
At the end of the probe, initialize and schedule the event workqueue. It calls the hard reset function where reset checks are done to find if the device is out of the reset. Control queue initialization and the necessary control queue support is added. Introduce function pointers for the register operations which are different between PF and VF devices. Signed-off-by:
-
Phani Burra authored
Add the required support to register IDPF PCI driver, as well as probe and remove call backs. Enable the PCI device and request the kernel to reserve the memory resources that will be used by the driver. Finally map the BAR0 address space. Signed-off-by:
-
Pavan Kumar Linga authored
Virtchnl version 1 is an interface used by the current generation of foundational NICs to negotiate the capabilities and configure the HW resources such as queues, vectors, RSS LUT, etc between the PF and VF drivers. It is not extensible to enable new features supported in the next generation of NICs/IPUs and to negotiate descriptor types, packet types and register offsets. To overcome the limitations of the existing interface, introduce the virtchnl version 2 and add the necessary opcodes, structures, definitions, and descriptor formats. The driver also learns the data queue and other register offsets to use instead of hardcoding them. The advantage of this approach is that it gives the flexibility to modify the register offsets if needed, restrict the use of certain descriptor types and negotiate the supported packet types. Co-developed-by:
-
David S. Miller authored
Pedro Tammela says: ==================== selftests/tc-testing: add tests covering classid Patches 1-3 add missing tests covering classid behaviour on tdc for cls_fw, cls_route and cls_fw. This behaviour was recently fixed by valis[0]. Patch 4 comes from the development done in the previous patches as it turns out cls_route never returns meaningful errors. [0] https://lore.kernel.org/all/20230729123202.72406-1-jhs@mojatatu.com/ v2->v3: https://lore.kernel.org/all/20230825155148.659895-1-pctammela@mojatatu.com/ - Added changes that were left in the working tree (Jakub) - Fixed two typos in commit message titles - Added Victor tags v1->v2: https://lore.kernel.org/all/20230818163544.351104-1-pctammela@mojatatu.com/ - Drop u32 updates ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Pedro Tammela authored
Use netlink extended ack and parsing policies to return more meaningful errors instead of the relying solely on errnos. Reviewed-by:
Victor Nogueira <victor@mojatatu.com> Signed-off-by:
Pedro Tammela <pctammela@mojatatu.com> Signed-off-by:
-
Pedro Tammela authored
As discussed in '3044b16e', cls_u32 was handling the use of classid incorrectly. Add a test to check if it's conforming to the correct behaviour. Reviewed-by:
-
Pedro Tammela authored
As discussed in 'b80b829e', cls_route was handling the use of classid incorrectly. Add a test to check if it's conforming to the correct behaviour. Reviewed-by:
-
Pedro Tammela authored
As discussed in '76e42ae8', cls_fw was handling the use of classid incorrectly. Add a few tests to check if it's conforming to the correct behaviour. Reviewed-by:
-
Andy Gospodarek authored
Signed-off-by:
Andy Gospodarek <gospo@broadcom.com> Signed-off-by: Pavan Chebbi pavan.chebbi@broadcom.com Signed-off-by:
Siva Reddy Kallam <siva.kallam@broadcom.com> Signed-off-by:
Prashant Sreedharan <prashant.sreedharan@broadcom.com> Reviewed-by:
Michael Chan <mchan@broadcom.com> Signed-off-by:
-
Christophe JAILLET authored
Use devm_kasprintf() instead of hand writing it. This is less verbose and less error prone. Signed-off-by:
Christophe JAILLET <christophe.jaillet@wanadoo.fr> Reviewed-by:
Simon Horman <horms@kernel.org> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queueDavid S. Miller authored
Tony Nguyen says: ==================== Intel Wired LAN Driver Updates 2023-09-11 (i40e, iavf) This series contains updates to i40e and iavf drivers. Andrii ensures all VSIs are cleaned up for remove in i40e. Brett reworks logic for setting promiscuous mode that can, currently, cause incorrect states on iavf. --- v2: - Remove redundant i40e_vsi_free_q_vectors() and kfree() calls (patch 1) v1: https://lore.kernel.org/netdev/20230905180521.887861-1-anthony.l.nguyen@intel.com/ ==================== Signed-off-by:
-
- 12 Sep, 2023 10 commits
-
-
Paolo Abeni authored
Eric Dumazet says: ==================== tcp: backlog processing optims First patches are mostly preparing the ground for the last one. Last patch of the series implements sort of ACK reduction only for the cases a TCP receiver is under high stress, which happens for high throughput flows. This gives us a ~20% increase of single TCP flow (100Gbit -> 120Gbit) ==================== Link: https://lore.kernel.org/r/20230911170531.828100-1-edumazet@google.comSigned-off-by:
Paolo Abeni <pabeni@redhat.com>
-
Eric Dumazet authored
This idea came after a particular workload requested the quickack attribute set on routes, and a performance drop was noticed for large bulk transfers. For high throughput flows, it is best to use one cpu running the user thread issuing socket system calls, and a separate cpu to process incoming packets from BH context. (With TSO/GRO, bottleneck is usually the 'user' cpu) Problem is the user thread can spend a lot of time while holding the socket lock, forcing BH handler to queue most of incoming packets in the socket backlog. Whenever the user thread releases the socket lock, it must first process all accumulated packets in the backlog, potentially adding latency spikes. Due to flood mitigation, having too many packets in the backlog increases chance of unexpected drops. Backlog processing unfortunately shifts a fair amount of cpu cycles from the BH cpu to the 'user' cpu, thus reducing max throughput. This patch takes advantage of the backlog processing, and the fact that ACK are mostly cumulative. The idea is to detect we are in the backlog processing and defer all eligible ACK into a single one, sent from tcp_release_cb(). This saves cpu cycles on both sides, and network resources. Performance of a single TCP flow on a 200Gbit NIC: - Throughput is increased by 20% (100Gbit -> 120Gbit). - Number of generated ACK per second shrinks from 240,000 to 40,000. - Number of backlog drops per second shrinks from 230 to 0. Benchmark context: - Regular netperf TCP_STREAM (no zerocopy) - Intel(R) Xeon(R) Platinum 8481C (Saphire Rapids) - MAX_SKB_FRAGS = 17 (~60KB per GRO packet) This feature is guarded by a new sysctl, and enabled by default: /proc/sys/net/ipv4/tcp_backlog_ack_defer Signed-off-by:
Eric Dumazet <edumazet@google.com> Acked-by:
Yuchung Cheng <ycheng@google.com> Acked-by:
Neal Cardwell <ncardwell@google.com> Acked-by:
Soheil Hassas Yeganeh <soheil@google.com> Acked-by:
Dave Taht <dave.taht@gmail.com> Signed-off-by:
-
Eric Dumazet authored
__sk_flush_backlog() / sk_flush_backlog() are used when TCP recvmsg()/sendmsg() process large chunks, to not let packets in the backlog too long. It makes sense to call tcp_release_cb() to also process actions held in sk->sk_tsq_flags for smoother scheduling. Signed-off-by:
-
Eric Dumazet authored
sock_release_ownership() should only be called by user owning the socket lock. After prior commit, we can remove one condition. Signed-off-by:
-
Eric Dumazet authored
This partially reverts c3f9b018 ("tcp: tcp_release_cb() should release socket ownership"). prequeue has been removed by Florian in commit e7942d06 ("tcp: remove prequeue support") __tcp_checksum_complete_user() being gone, we no longer have to release socket ownership in tcp_release_cb(). This is a prereq for third patch in the series ("net: call prot->release_cb() when processing backlog"). Signed-off-by:
-
Andy Shevchenko authored
Use bitmap_zalloc() and bitmap_free() instead of hand-writing them. It is less verbose and it improves the type checking and semantic. While at it, add missing header inclusion (should be bitops.h, but with the above change it becomes bitmap.h). Reviewed-by:
Loic Poulain <loic.poulain@linaro.org> Reviewed-by:
Sergey Ryazanov <ryazanov.s.a@gmail.com> Signed-off-by:
Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20230911131618.4159437-1-andriy.shevchenko@linux.intel.comSigned-off-by:
-
Zhengchao Shao authored
Since commit 1202cdd6("Remove DECnet support from kernel") has been merged, all callers pass in the initial_ref value of 1 when they call dst_alloc(). Therefore, remove initial_ref when the dst_alloc() is declared and replace initial_ref with 1 in dst_alloc(). Also when all callers call dst_init(), the value of initial_ref is 1. Therefore, remove the input parameter initial_ref of the dst_init() and replace initial_ref with the value 1 in dst_init. Signed-off-by:
Zhengchao Shao <shaozhengchao@huawei.com> Link: https://lore.kernel.org/r/20230911125045.346390-1-shaozhengchao@huawei.comSigned-off-by:
-
Paolo Abeni authored
MD Danish Anwar says: ==================== Add support for ICSSG on AM64x EVM This series adds support for ICSSG driver on AM64x EVM. First patch of the series adds compatible for AM64x EVM in icssg-prueth dt binding. Second patch adds support for AM64x compatible in the ICSSG driver. This series addresses comments on [v1] (which was posted as RFC). This series is based on the latest net-next/main. This series has no dependency. Changes from v1 to v2: *) Made the compatible list in patch 1 alphanumerically ordered as asked by Krzysztof. *) Dropped the RFC tag. *) Added RB tags of Andrew and Roger. [v1] https://lore.kernel.org/all/20230830113724.1228624-1-danishanwar@ti.com/ ==================== Link: https://lore.kernel.org/r/20230911054308.2163076-1-danishanwar@ti.comSigned-off-by:
-
MD Danish Anwar authored
Add AM64x ICSSG support which is similar to am65x SR2.0, but required: - all ring configured in exposed ring mode - always fill both original and buffer fields in cppi5 desc Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Reviewed-by:
Roger Quadros <rogerq@kernel.org> Signed-off-by:
MD Danish Anwar <danishanwar@ti.com> Signed-off-by:
-
MD Danish Anwar authored
Add compatible for AM64x in icssg-prueth dt bindings. AM64x supports ICSSG similar to AM65x SR2.0. Acked-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Reviewed-by:
-
- 11 Sep, 2023 8 commits
-
-
Brett Creeley authored
Currently when configuring promiscuous mode on the AVF we detect a change in the netdev->flags. We use IFF_PROMISC and IFF_ALLMULTI to determine whether or not we need to request/release promiscuous mode and/or multicast promiscuous mode. The problem is that the AQ calls for setting/clearing promiscuous/multicast mode are treated separately. This leads to a case where we can trigger two promiscuous mode AQ calls in a row with the incorrect state. To fix this make a few changes. Use IAVF_FLAG_AQ_CONFIGURE_PROMISC_MODE instead of the previous IAVF_FLAG_AQ_[REQUEST|RELEASE]_[PROMISC|ALLMULTI] flags. In iavf_set_rx_mode() detect if there is a change in the netdev->flags in comparison with adapter->flags and set the IAVF_FLAG_AQ_CONFIGURE_PROMISC_MODE aq_required bit. Then in iavf_process_aq_command() only check for IAVF_FLAG_CONFIGURE_PROMISC_MODE and call iavf_set_promiscuous() if it's set. In iavf_set_promiscuous() check again to see which (if any) promiscuous mode bits have changed when comparing the netdev->flags with the adapter->flags. Use this to set the flags which get sent to the PF driver. Add a spinlock that is used for updating current_netdev_promisc_flags and only allows one promiscuous mode AQ at a time. [1] Fixes the fact that we will only have one AQ call in the aq_required queue at any one time. [2] Streamlines the change in promiscuous mode to only set one AQ required bit. [3] This allows us to keep track of the current state of the flags and also makes it so we can take the most recent netdev->flags promiscuous mode state. [4] This fixes the problem where a change in the netdev->flags can cause IAVF_FLAG_AQ_CONFIGURE_PROMISC_MODE to be set in iavf_set_rx_mode(), but cleared in iavf_set_promiscuous() before the change is ever made via AQ call. Fixes: 47d34839 ("i40evf: Add driver support for promiscuous mode") Signed-off-by:
Brett Creeley <brett.creeley@intel.com> Signed-off-by:
Ahmed Zaki <ahmed.zaki@intel.com> Tested-by:
Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by:
-
Andrii Staikov authored
Instead of freeing memory of a single VSI, make sure the memory for all VSIs is cleared before releasing VSIs. Add releasing of their resources in a loop with the iteration number equal to the number of allocated VSIs. Fixes: 41c445ff ("i40e: main driver core") Signed-off-by:
Andrii Staikov <andrii.staikov@intel.com> Signed-off-by:
Aleksandr Loktionov <aleksandr.loktionov@intel.com> Reviewed-by:
-
Lorenzo Bianconi authored
MT7988 SoC support 3 NICs. Fix pse_port configuration in mtk_flow_set_output_device routine if the traffic is offloaded to eth2. Rely on mtk_pse_port definitions. Fixes: 88efedf5 ("net: ethernet: mtk_eth_soc: enable nft hw flowtable_offload for MT7988 SoC") Signed-off-by:
Lorenzo Bianconi <lorenzo@kernel.org> Signed-off-by:
-
Daniel Golle authored
Variable dma_addr in function mtk_poll_rx can be uninitialized on some of the error paths. In practise this doesn't matter, even random data present in uninitialized stack memory can safely be used in the way it happens in the error path. However, in order to make Smatch happy make sure the variable is always initialized. Signed-off-by:
Daniel Golle <daniel@makrotopia.org> Signed-off-by:
-
Shigeru Yoshida authored
syzbot reported a memory leak like below: BUG: memory leak unreferenced object 0xffff88810b088c00 (size 240): comm "syz-executor186", pid 5012, jiffies 4294943306 (age 13.680s) hex dump (first 32 bytes): 00 89 08 0b 81 88 ff ff 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<ffffffff83e5d5ff>] __alloc_skb+0x1ef/0x230 net/core/skbuff.c:634 [<ffffffff84606e59>] alloc_skb include/linux/skbuff.h:1289 [inline] [<ffffffff84606e59>] kcm_sendmsg+0x269/0x1050 net/kcm/kcmsock.c:815 [<ffffffff83e479c6>] sock_sendmsg_nosec net/socket.c:725 [inline] [<ffffffff83e479c6>] sock_sendmsg+0x56/0xb0 net/socket.c:748 [<ffffffff83e47f55>] ____sys_sendmsg+0x365/0x470 net/socket.c:2494 [<ffffffff83e4c389>] ___sys_sendmsg+0xc9/0x130 net/socket.c:2548 [<ffffffff83e4c536>] __sys_sendmsg+0xa6/0x120 net/socket.c:2577 [<ffffffff84ad7bb8>] do_syscall_x64 arch/x86/entry/common.c:50 [inline] [<ffffffff84ad7bb8>] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 [<ffffffff84c0008b>] entry_SYSCALL_64_after_hwframe+0x63/0xcd In kcm_sendmsg(), kcm_tx_msg(head)->last_skb is used as a cursor to append newly allocated skbs to 'head'. If some bytes are copied, an error occurred, and jumped to out_error label, 'last_skb' is left unmodified. A later kcm_sendmsg() will use an obsoleted 'last_skb' reference, corrupting the 'head' frag_list and causing the leak. This patch fixes this issue by properly updating the last allocated skb in 'last_skb'. Fixes: ab7ac4eb ("kcm: Kernel Connection Multiplexor module") Reported-and-tested-by: syzbot+6f98de741f7dbbfc4ccb@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=6f98de741f7dbbfc4ccbSigned-off-by:Shigeru Yoshida <syoshida@redhat.com> Signed-off-by:
-
Hayes Wang authored
According to the document of napi, there is no rx process when the budget is 0. Therefore, r8152_poll() has to return 0 directly when the budget is equal to 0. Fixes: d2187f8e ("r8152: divide the tx and rx bottom functions") Signed-off-by:
Hayes Wang <hayeswang@realtek.com> Signed-off-by:
-
David S. Miller authored
Vladimir Oltean says: ==================== Fixes for SJA1105 DSA FDB regressions A report by Yanan Yang has prompted an investigation into the sja1105 driver's behavior w.r.t. multicast. The report states that when adding multicast L2 addresses with "bridge mdb add", only the most recently added address works - the others seem to be overwritten. This is solved by patch 3/5 (with patch 2/5 as a dependency for it). Patches 4/5 and 5/5 fix a series of race conditions introduced during the same patch set as the bug above, namely this one: https://patchwork.kernel.org/project/netdevbpf/cover/20211024171757.3753288-1-vladimir.oltean@nxp.com/ Finally, patch 1/5 fixes an issue found ever since the introduction of multicast forwarding offload in sja1105, which is that the multicast addresses are visible (with the "self" flag) in "bridge fdb show". ==================== Signed-off-by:
-
Vladimir Oltean authored
Currently, when we add the first sja1105 port to a bridge with vlan_filtering 1, then we sometimes see this output: sja1105 spi2.2: port 4 failed to read back entry for be:79:b4:9e:9e:96 vid 3088: -ENOENT sja1105 spi2.2: Reset switch and programmed static config. Reason: VLAN filtering sja1105 spi2.2: port 0 failed to add be:79:b4:9e:9e:96 vid 0 to fdb: -2 It is because sja1105_fdb_add() runs from the dsa_owq which is no longer serialized with switch resets since it dropped the rtnl_lock() in the blamed commit. Either performing the FDB accesses before the reset, or after the reset, is equally fine, because sja1105_static_fdb_change() backs up those changes in the static config, but FDB access during reset isn't ok. Make sja1105_static_config_reload() take the fdb_lock to fix that. Fixes: 0faf890f ("net: dsa: drop rtnl_lock from dsa_slave_switchdev_event_work") Signed-off-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by:
-
