{"resultsPerPage":1,"startIndex":0,"totalResults":1,"format":"NVD_CVE","version":"2.0","timestamp":"2026-04-22T18:57:09.095","vulnerabilities":[{"cve":{"id":"CVE-2023-54149","sourceIdentifier":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","published":"2025-12-24T13:16:16.910","lastModified":"2026-04-15T00:35:42.020","vulnStatus":"Deferred","cveTags":[],"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nnet: dsa: avoid suspicious RCU usage for synced VLAN-aware MAC addresses\n\nWhen using the felix driver (the only one which supports UC filtering\nand MC filtering) as a DSA master for a random other DSA switch, one can\nsee the following stack trace when the downstream switch ports join a\nVLAN-aware bridge:\n\n=============================\nWARNING: suspicious RCU usage\n-----------------------------\nnet/8021q/vlan_core.c:238 suspicious rcu_dereference_protected() usage!\n\nstack backtrace:\nWorkqueue: dsa_ordered dsa_slave_switchdev_event_work\nCall trace:\n lockdep_rcu_suspicious+0x170/0x210\n vlan_for_each+0x8c/0x188\n dsa_slave_sync_uc+0x128/0x178\n __hw_addr_sync_dev+0x138/0x158\n dsa_slave_set_rx_mode+0x58/0x70\n __dev_set_rx_mode+0x88/0xa8\n dev_uc_add+0x74/0xa0\n dsa_port_bridge_host_fdb_add+0xec/0x180\n dsa_slave_switchdev_event_work+0x7c/0x1c8\n process_one_work+0x290/0x568\n\nWhat it's saying is that vlan_for_each() expects rtnl_lock() context and\nit's not getting it, when it's called from the DSA master's ndo_set_rx_mode().\n\nThe caller of that - dsa_slave_set_rx_mode() - is the slave DSA\ninterface's dsa_port_bridge_host_fdb_add() which comes from the deferred\ndsa_slave_switchdev_event_work().\n\nWe went to great lengths to avoid the rtnl_lock() context in that call\npath in commit 0faf890fc519 (\"net: dsa: drop rtnl_lock from\ndsa_slave_switchdev_event_work\"), and calling rtnl_lock() is simply not\nan option due to the possibility of deadlocking when calling\ndsa_flush_workqueue() from the call paths that do hold rtnl_lock() -\nbasically all of them.\n\nSo, when the DSA master calls vlan_for_each() from its ndo_set_rx_mode(),\nthe state of the 8021q driver on this device is really not protected\nfrom concurrent access by anything.\n\nLooking at net/8021q/, I don't think that vlan_info->vid_list was\nparticularly designed with RCU traversal in mind, so introducing an RCU\nread-side form of vlan_for_each() - vlan_for_each_rcu() - won't be so\neasy, and it also wouldn't be exactly what we need anyway.\n\nIn general I believe that the solution isn't in net/8021q/ anyway;\nvlan_for_each() is not cut out for this task. DSA doesn't need rtnl_lock()\nto be held per se - since it's not a netdev state change that we're\nblocking, but rather, just concurrent additions/removals to a VLAN list.\nWe don't even need sleepable context - the callback of vlan_for_each()\njust schedules deferred work.\n\nThe proposed escape is to remove the dependency on vlan_for_each() and\nto open-code a non-sleepable, rtnl-free alternative to that, based on\ncopies of the VLAN list modified from .ndo_vlan_rx_add_vid() and\n.ndo_vlan_rx_kill_vid()."}],"metrics":{},"references":[{"url":"https://git.kernel.org/stable/c/3948c69b3837fec2ee5a90fbc911c343199be0ac","source":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"},{"url":"https://git.kernel.org/stable/c/3f9e79f31e51b7d5bf95c617540deb6cf2816a3f","source":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"},{"url":"https://git.kernel.org/stable/c/d06f925f13976ab82167c93467c70a337a0a3cda","source":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"}]}}]}