kernel.git/drivers/net/virtio_net.c, branch linux-6.2.y

virtio_net: suppress cpu stall when free_unused_bufs

2023-05-17T11:58:51Z

[ Upstream commit f8bb5104394560e29017c25bcade4c6b7aabd108 ] For multi-queue and large ring-size use case, the following error occurred when free_unused_bufs: rcu: INFO: rcu_sched self-detected stall on CPU. Fixes: 986a4f4d452d ("virtio_net: multiqueue support") Signed-off-by: Wenliang Wang Acked-by: Michael S. Tsirkin Signed-off-by: David S. Miller Signed-off-by: Sasha Levin

virtio_net: bugfix overflow inside xdp_linearize_page()

2023-04-26T12:29:56Z

[ Upstream commit 853618d5886bf94812f31228091cd37d308230f7 ] Here we copy the data from the original buf to the new page. But we not check that it may be overflow. As long as the size received(including vnethdr) is greater than 3840 (PAGE_SIZE -VIRTIO_XDP_HEADROOM). Then the memcpy will overflow. And this is completely possible, as long as the MTU is large, such as 4096. In our test environment, this will cause crash. Since crash is caused by the written memory, it is meaningless, so I do not include it. Fixes: 72979a6c3590 ("virtio_net: xdp, add slowpath case for non contiguous buffers") Signed-off-by: Xuan Zhuo Acked-by: Jason Wang Acked-by: Michael S. Tsirkin Signed-off-by: David S. Miller Signed-off-by: Sasha Levin

virtio-net: Keep stop() to follow mirror sequence of open()

2023-02-02T18:10:53Z

Cited commit in fixes tag frees rxq xdp info while RQ NAPI is still enabled and packet processing may be ongoing. Follow the mirror sequence of open() in the stop() callback. This ensures that when rxq info is unregistered, no rx packet processing is ongoing. Fixes: 754b8a21a96d ("virtio_net: setup xdp_rxq_info") Acked-by: Michael S. Tsirkin Reviewed-by: Jiri Pirko Signed-off-by: Parav Pandit Link: https://lore.kernel.org/r/20230202163516.12559-1-parav@nvidia.com Signed-off-by: Jakub Kicinski

virtio-net: execute xdp_do_flush() before napi_complete_done()

2023-01-28T06:27:59Z

Make sure that xdp_do_flush() is always executed before napi_complete_done(). This is important for two reasons. First, a redirect to an XSKMAP assumes that a call to xdp_do_redirect() from napi context X on CPU Y will be followed by a xdp_do_flush() from the same napi context and CPU. This is not guaranteed if the napi_complete_done() is executed before xdp_do_flush(), as it tells the napi logic that it is fine to schedule napi context X on another CPU. Details from a production system triggering this bug using the veth driver can be found following the first link below. The second reason is that the XDP_REDIRECT logic in itself relies on being inside a single NAPI instance through to the xdp_do_flush() call for RCU protection of all in-kernel data structures. Details can be found in the second link below. Fixes: 186b3c998c50 ("virtio-net: support XDP_REDIRECT") Signed-off-by: Magnus Karlsson Acked-by: Toke Høiland-Jørgensen Link: https://lore.kernel.org/r/20221220185903.1105011-1-sbohrer@cloudflare.com Link: https://lore.kernel.org/all/20210624160609.292325-1-toke@redhat.com/ Acked-by: Michael S. Tsirkin Signed-off-by: Jakub Kicinski

virtio-net: correctly enable callback during start_xmit

2023-01-18T14:28:56Z

Commit a7766ef18b33("virtio_net: disable cb aggressively") enables virtqueue callback via the following statement: do { if (use_napi) virtqueue_disable_cb(sq->vq); free_old_xmit_skbs(sq, false); } while (use_napi && kick && unlikely(!virtqueue_enable_cb_delayed(sq->vq))); When NAPI is used and kick is false, the callback won't be enabled here. And when the virtqueue is about to be full, the tx will be disabled, but we still don't enable tx interrupt which will cause a TX hang. This could be observed when using pktgen with burst enabled. TO be consistent with the logic that tries to disable cb only for NAPI, fixing this by trying to enable delayed callback only when NAPI is enabled when the queue is about to be full. Fixes: a7766ef18b33 ("virtio_net: disable cb aggressively") Signed-off-by: Jason Wang Tested-by: Laurent Vivier Signed-off-by: David S. Miller

drivers/net/virtio_net.c: Added USO support.

2022-12-12T09:29:56Z

Now, it possible to enable GSO_UDP_L4("tx-udp-segmentation") for VirtioNet. Signed-off-by: Andrew Melnychenko Signed-off-by: David S. Miller

Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

2022-11-29T21:04:52Z

tools/lib/bpf/ringbuf.c 927cbb478adf ("libbpf: Handle size overflow for ringbuf mmap") b486d19a0ab0 ("libbpf: checkpatch: Fixed code alignments in ringbuf.c") https://lore.kernel.org/all/20221121122707.44d1446a@canb.auug.org.au/ Signed-off-by: Jakub Kicinski

virtio_net: Fix probe failed when modprobe virtio_net

2022-11-24T08:26:35Z

When doing the following test steps, an error was found: step 1: modprobe virtio_net succeeded # modprobe virtio_net <-- OK step 2: fault injection in register_netdevice() # modprobe -r virtio_net <-- OK # ... FAULT_INJECTION: forcing a failure. name failslab, interval 1, probability 0, space 0, times 0 CPU: 0 PID: 3521 Comm: modprobe Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), Call Trace: ... should_failslab+0xa/0x20 ... dev_set_name+0xc0/0x100 netdev_register_kobject+0xc2/0x340 register_netdevice+0xbb9/0x1320 virtnet_probe+0x1d72/0x2658 [virtio_net] ... virtio_net: probe of virtio0 failed with error -22 step 3: modprobe virtio_net failed # modprobe virtio_net <-- failed virtio_net: probe of virtio0 failed with error -2 The root cause of the problem is that the queues are not disable on the error handling path when register_netdevice() fails in virtnet_probe(), resulting in an error "-ENOENT" returned in the next modprobe call in setup_vq(). virtio_pci_modern_device uses virtqueues to send or receive message, and "queue_enable" records whether the queues are available. In vp_modern_find_vqs(), all queues will be selected and activated, but once queues are enabled there is no way to go back except reset. Fix it by reset virtio device on error handling path. This makes error handling follow the same order as normal device cleanup in virtnet_remove() which does: unregister, destroy failover, then reset. And that flow is better tested than error handling so we can be reasonably sure it works well. Fixes: 024655555021 ("virtio_net: fix use after free on allocation failure") Signed-off-by: Li Zetao Acked-by: Michael S. Tsirkin Link: https://lore.kernel.org/r/20221122150046.3910638-1-lizetao1@huawei.com Signed-off-by: Paolo Abeni

net: Remove the obsolte u64_stats_fetch_*_irq() users (drivers).

2022-10-29T03:13:54Z

Now that the 32bit UP oddity is gone and 32bit uses always a sequence count, there is no need for the fetch_irq() variants anymore. Convert to the regular interface. Signed-off-by: Thomas Gleixner Signed-off-by: Sebastian Andrzej Siewior Acked-by: Peter Zijlstra (Intel) Signed-off-by: Jakub Kicinski

Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost

2022-10-10T21:02:53Z

Pull virtio updates from Michael Tsirkin: - 9k mtu perf improvements - vdpa feature provisioning - virtio blk SECURE ERASE support - fixes and cleanups all over the place * tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: virtio_pci: don't try to use intxif pin is zero vDPA: conditionally read MTU and MAC in dev cfg space vDPA: fix spars cast warning in vdpa_dev_net_mq_config_fill vDPA: check virtio device features to detect MQ vDPA: check VIRTIO_NET_F_RSS for max_virtqueue_paris's presence vDPA: only report driver features if FEATURES_OK is set vDPA: allow userspace to query features of a vDPA device virtio_blk: add SECURE ERASE command support vp_vdpa: support feature provisioning vdpa_sim_net: support feature provisioning vdpa: device feature provisioning virtio-net: use mtu size as buffer length for big packets virtio-net: introduce and use helper function for guest gso support checks virtio: drop vp_legacy_set_queue_size virtio_ring: make vring_alloc_queue_packed prettier virtio_ring: split: Operators use unified style vhost: add __init/__exit annotations to module init/exit funcs