In the Linux kernel, the following vulnerability has been resolved: fs: dlm: fix race in lowcomms This patch fixes a race between queue_work() in _dlm_lowcomms_commit_msg() and srcu_read_unlock(). The queue_work() can take the final reference of a dlm_msg and so msg->idx can contain garbage which is signaled by the following warning: [ 676.237050] ------------[ cut here ]------------ [ 676.237052] WARNING: CPU: 0 PID: 1060 at include/linux/srcu.h:189 dlm_lowcomms_commit_msg+0x41/0x50 [ 676.238945] Modules linked in: dlm_locktorture torture rpcsec_gss_krb5 intel_rapl_msr intel_rapl_common iTCO_wdt iTCO_vendor_support qxl kvm_intel drm_ttm_helper vmw_vsock_virtio_transport kvm vmw_vsock_virtio_transport_common ttm irqbypass crc32_pclmul joydev crc32c_intel serio_raw drm_kms_helper vsock virtio_scsi virtio_console virtio_balloon snd_pcm drm syscopyarea sysfillrect sysimgblt snd_timer fb_sys_fops i2c_i801 lpc_ich snd i2c_smbus soundcore pcspkr [ 676.244227] CPU: 0 PID: 1060 Comm: lock_torture_wr Not tainted 5.19.0-rc3+ #1546 [ 676.245216] Hardware name: Red Hat KVM/RHEL-AV, BIOS 1.16.0-2.module+el8.7.0+15506+033991b0 04/01/2014 [ 676.246460] RIP: 0010:dlm_lowcomms_commit_msg+0x41/0x50 [ 676.247132] Code: fe ff ff ff 75 24 48 c7 c6 bd 0f 49 bb 48 c7 c7 38 7c 01 bd e8 00 e7 ca ff 89 de 48 c7 c7 60 78 01 bd e8 42 3d cd ff 5b 5d c3 <0f> 0b eb d8 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 55 48 [ 676.249253] RSP: 0018:ffffa401c18ffc68 EFLAGS: 00010282 [ 676.249855] RAX: 0000000000000001 RBX: 00000000ffff8b76 RCX: 0000000000000006 [ 676.250713] RDX: 0000000000000000 RSI: ffffffffbccf3a10 RDI: ffffffffbcc7b62e [ 676.251610] RBP: ffffa401c18ffc70 R08: 0000000000000001 R09: 0000000000000001 [ 676.252481] R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000005 [ 676.253421] R13: ffff8b76786ec370 R14: ffff8b76786ec370 R15: ffff8b76786ec480 [ 676.254257] FS: 0000000000000000(0000) GS:ffff8b7777800000(0000) knlGS:0000000000000000 [ 676.255239] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 676.255897] CR2: 00005590205d88b8 CR3: 000000017656c003 CR4: 0000000000770ee0 [ 676.256734] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 676.257567] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 676.258397] PKRU: 55555554 [ 676.258729] Call Trace: [ 676.259063] <TASK> [ 676.259354] dlm_midcomms_commit_mhandle+0xcc/0x110 [ 676.259964] queue_bast+0x8b/0xb0 [ 676.260423] grant_pending_locks+0x166/0x1b0 [ 676.261007] _unlock_lock+0x75/0x90 [ 676.261469] unlock_lock.isra.57+0x62/0xa0 [ 676.262009] dlm_unlock+0x21e/0x330 [ 676.262457] ? lock_torture_stats+0x80/0x80 [dlm_locktorture] [ 676.263183] torture_unlock+0x5a/0x90 [dlm_locktorture] [ 676.263815] ? preempt_count_sub+0xba/0x100 [ 676.264361] ? complete+0x1d/0x60 [ 676.264777] lock_torture_writer+0xb8/0x150 [dlm_locktorture] [ 676.265555] kthread+0x10a/0x130 [ 676.266007] ? kthread_complete_and_exit+0x20/0x20 [ 676.266616] ret_from_fork+0x22/0x30 [ 676.267097] </TASK> [ 676.267381] irq event stamp: 9579855 [ 676.267824] hardirqs last enabled at (9579863): [<ffffffffbb14e6f8>] __up_console_sem+0x58/0x60 [ 676.268896] hardirqs last disabled at (9579872): [<ffffffffbb14e6dd>] __up_console_sem+0x3d/0x60 [ 676.270008] softirqs last enabled at (9579798): [<ffffffffbc200349>] __do_softirq+0x349/0x4c7 [ 676.271438] softirqs last disabled at (9579897): [<ffffffffbb0d54c0>] irq_exit_rcu+0xb0/0xf0 [ 676.272796] ---[ end trace 0000000000000000 ]--- I reproduced this warning with dlm_locktorture test which is currently not upstream. However this patch fix the issue by make a additional refcount between dlm_lowcomms_new_msg() and dlm_lowcomms_commit_msg(). In case of the race the kref_put() in dlm_lowcomms_commit_msg() will be the final put.

In the Linux kernel, the following vulnerability has been resolved: scsi: target: iscsi: Fix a race condition between login_work and the login thread In case a malicious initiator sends some random data immediately after a login PDU; the iscsi_target_sk_data_ready() callback will schedule the login_work and, at the same time, the negotiation may end without clearing the LOGIN_FLAGS_INITIAL_PDU flag (because no additional PDU exchanges are required to complete the login). The login has been completed but the login_work function will find the LOGIN_FLAGS_INITIAL_PDU flag set and will never stop from rescheduling itself; at this point, if the initiator drops the connection, the iscsit_conn structure will be freed, login_work will dereference a released socket structure and the kernel crashes. BUG: kernel NULL pointer dereference, address: 0000000000000230 PF: supervisor write access in kernel mode PF: error_code(0x0002) - not-present page Workqueue: events iscsi_target_do_login_rx [iscsi_target_mod] RIP: 0010:_raw_read_lock_bh+0x15/0x30 Call trace: iscsi_target_do_login_rx+0x75/0x3f0 [iscsi_target_mod] process_one_work+0x1e8/0x3c0 Fix this bug by forcing login_work to stop after the login has been completed and the socket callbacks have been restored. Add a comment to clearify the return values of iscsi_target_do_login()

In the Linux kernel, the following vulnerability has been resolved: ath11k: fix netdev open race Make sure to allocate resources needed before registering the device. This specifically avoids having a racing open() trigger a BUG_ON() in mod_timer() when ath11k_mac_op_start() is called before the mon_reap_timer as been set up. I did not see this issue with next-20220310, but I hit it on every probe with next-20220511. Perhaps some timing changed in between. Here's the backtrace: [ 51.346947] kernel BUG at kernel/time/timer.c:990! [ 51.346958] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ... [ 51.578225] Call trace: [ 51.583293] __mod_timer+0x298/0x390 [ 51.589518] mod_timer+0x14/0x20 [ 51.595368] ath11k_mac_op_start+0x41c/0x4a0 [ath11k] [ 51.603165] drv_start+0x38/0x60 [mac80211] [ 51.610110] ieee80211_do_open+0x29c/0x7d0 [mac80211] [ 51.617945] ieee80211_open+0x60/0xb0 [mac80211] [ 51.625311] __dev_open+0x100/0x1c0 [ 51.631420] __dev_change_flags+0x194/0x210 [ 51.638214] dev_change_flags+0x24/0x70 [ 51.644646] do_setlink+0x228/0xdb0 [ 51.650723] __rtnl_newlink+0x460/0x830 [ 51.657162] rtnl_newlink+0x4c/0x80 [ 51.663229] rtnetlink_rcv_msg+0x124/0x390 [ 51.669917] netlink_rcv_skb+0x58/0x130 [ 51.676314] rtnetlink_rcv+0x18/0x30 [ 51.682460] netlink_unicast+0x250/0x310 [ 51.688960] netlink_sendmsg+0x19c/0x3e0 [ 51.695458] ____sys_sendmsg+0x220/0x290 [ 51.701938] ___sys_sendmsg+0x7c/0xc0 [ 51.708148] __sys_sendmsg+0x68/0xd0 [ 51.714254] __arm64_sys_sendmsg+0x28/0x40 [ 51.720900] invoke_syscall+0x48/0x120 Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3

In the Linux kernel, the following vulnerability has been resolved: sysctl: Fix data races in proc_douintvec(). A sysctl variable is accessed concurrently, and there is always a chance of data-race. So, all readers and writers need some basic protection to avoid load/store-tearing. This patch changes proc_douintvec() to use READ_ONCE() and WRITE_ONCE() internally to fix data-races on the sysctl side. For now, proc_douintvec() itself is tolerant to a data-race, but we still need to add annotations on the other subsystem's side.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix data-races around sysctl_tcp_max_reordering. While reading sysctl_tcp_max_reordering, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: kcm: close race conditions on sk_receive_queue sk->sk_receive_queue is protected by skb queue lock, but for KCM sockets its RX path takes mux->rx_lock to protect more than just skb queue. However, kcm_recvmsg() still only grabs the skb queue lock, so race conditions still exist. We can teach kcm_recvmsg() to grab mux->rx_lock too but this would introduce a potential performance regression as struct kcm_mux can be shared by multiple KCM sockets. So we have to enforce skb queue lock in requeue_rx_msgs() and handle skb peek case carefully in kcm_wait_data(). Fortunately, skb_recv_datagram() already handles it nicely and is widely used by other sockets, we can just switch to skb_recv_datagram() after getting rid of the unnecessary sock lock in kcm_recvmsg() and kcm_splice_read(). Side note: SOCK_DONE is not used by KCM sockets, so it is safe to get rid of this check too. I ran the original syzbot reproducer for 30 min without seeing any issue.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix data-races around sysctl_tcp_fastopen. While reading sysctl_tcp_fastopen, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix a data-race around sysctl_tcp_ecn_fallback. While reading sysctl_tcp_ecn_fallback, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: icmp: Fix a data-race around sysctl_icmp_errors_use_inbound_ifaddr. While reading sysctl_icmp_errors_use_inbound_ifaddr, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: ieee802154/adf7242: defer destroy_workqueue call There is a possible race condition (use-after-free) like below (FREE) | (USE) adf7242_remove | adf7242_channel cancel_delayed_work_sync | destroy_workqueue (1) | adf7242_cmd_rx | mod_delayed_work (2) | The root cause for this race is that the upper layer (ieee802154) is unaware of this detaching event and the function adf7242_channel can be called without any checks. To fix this, we can add a flag write at the beginning of adf7242_remove and add flag check in adf7242_channel. Or we can just defer the destructive operation like other commit 3e0588c291d6 ("hamradio: defer ax25 kfree after unregister_netdev") which let the ieee802154_unregister_hw() to handle the synchronization. This patch takes the second option. runs")

In the Linux kernel, the following vulnerability has been resolved: ibmvnic: fix race between xmit and reset There is a race between reset and the transmit paths that can lead to ibmvnic_xmit() accessing an scrq after it has been freed in the reset path. It can result in a crash like: Kernel attempted to read user page (0) - exploit attempt? (uid: 0) BUG: Kernel NULL pointer dereference on read at 0x00000000 Faulting instruction address: 0xc0080000016189f8 Oops: Kernel access of bad area, sig: 11 [#1] ... NIP [c0080000016189f8] ibmvnic_xmit+0x60/0xb60 [ibmvnic] LR [c000000000c0046c] dev_hard_start_xmit+0x11c/0x280 Call Trace: [c008000001618f08] ibmvnic_xmit+0x570/0xb60 [ibmvnic] (unreliable) [c000000000c0046c] dev_hard_start_xmit+0x11c/0x280 [c000000000c9cfcc] sch_direct_xmit+0xec/0x330 [c000000000bfe640] __dev_xmit_skb+0x3a0/0x9d0 [c000000000c00ad4] __dev_queue_xmit+0x394/0x730 [c008000002db813c] __bond_start_xmit+0x254/0x450 [bonding] [c008000002db8378] bond_start_xmit+0x40/0xc0 [bonding] [c000000000c0046c] dev_hard_start_xmit+0x11c/0x280 [c000000000c00ca4] __dev_queue_xmit+0x564/0x730 [c000000000cf97e0] neigh_hh_output+0xd0/0x180 [c000000000cfa69c] ip_finish_output2+0x31c/0x5c0 [c000000000cfd244] __ip_queue_xmit+0x194/0x4f0 [c000000000d2a3c4] __tcp_transmit_skb+0x434/0x9b0 [c000000000d2d1e0] __tcp_retransmit_skb+0x1d0/0x6a0 [c000000000d2d984] tcp_retransmit_skb+0x34/0x130 [c000000000d310e8] tcp_retransmit_timer+0x388/0x6d0 [c000000000d315ec] tcp_write_timer_handler+0x1bc/0x330 [c000000000d317bc] tcp_write_timer+0x5c/0x200 [c000000000243270] call_timer_fn+0x50/0x1c0 [c000000000243704] __run_timers.part.0+0x324/0x460 [c000000000243894] run_timer_softirq+0x54/0xa0 [c000000000ea713c] __do_softirq+0x15c/0x3e0 [c000000000166258] __irq_exit_rcu+0x158/0x190 [c000000000166420] irq_exit+0x20/0x40 [c00000000002853c] timer_interrupt+0x14c/0x2b0 [c000000000009a00] decrementer_common_virt+0x210/0x220 --- interrupt: 900 at plpar_hcall_norets_notrace+0x18/0x2c The immediate cause of the crash is the access of tx_scrq in the following snippet during a reset, where the tx_scrq can be either NULL or an address that will soon be invalid: ibmvnic_xmit() { ... tx_scrq = adapter->tx_scrq[queue_num]; txq = netdev_get_tx_queue(netdev, queue_num); ind_bufp = &tx_scrq->ind_buf; if (test_bit(0, &adapter->resetting)) { ... } But beyond that, the call to ibmvnic_xmit() itself is not safe during a reset and the reset path attempts to avoid this by stopping the queue in ibmvnic_cleanup(). However just after the queue was stopped, an in-flight ibmvnic_complete_tx() could have restarted the queue even as the reset is progressing. Since the queue was restarted we could get a call to ibmvnic_xmit() which can then access the bad tx_scrq (or other fields). We cannot however simply have ibmvnic_complete_tx() check the ->resetting bit and skip starting the queue. This can race at the "back-end" of a good reset which just restarted the queue but has not cleared the ->resetting bit yet. If we skip restarting the queue due to ->resetting being true, the queue would remain stopped indefinitely potentially leading to transmit timeouts. IOW ->resetting is too broad for this purpose. Instead use a new flag that indicates whether or not the queues are active. Only the open/ reset paths control when the queues are active. ibmvnic_complete_tx() and others wake up the queue only if the queue is marked active. So we will have: A. reset/open thread in ibmvnic_cleanup() and __ibmvnic_open() ->resetting = true ->tx_queues_active = false disable tx queues ... ->tx_queues_active = true start tx queues B. Tx interrupt in ibmvnic_complete_tx(): if (->tx_queues_active) netif_wake_subqueue(); To ensure that ->tx_queues_active and state of the queues are consistent, we need a lock which: - must also be taken in the interrupt path (ibmvnic_complete_tx()) - shared across the multiple ---truncated---

In the Linux kernel, the following vulnerability has been resolved: tcp/dccp: Fix a data-race around sysctl_tcp_fwmark_accept. While reading sysctl_tcp_fwmark_accept, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: ipv4: Fix a data-race around sysctl_fib_sync_mem. While reading sysctl_fib_sync_mem, it can be changed concurrently. So, we need to add READ_ONCE() to avoid a data-race.

In the Linux kernel, the following vulnerability has been resolved: dm ioctl: fix misbehavior if list_versions races with module loading __list_versions will first estimate the required space using the "dm_target_iterate(list_version_get_needed, &needed)" call and then will fill the space using the "dm_target_iterate(list_version_get_info, &iter_info)" call. Each of these calls locks the targets using the "down_read(&_lock)" and "up_read(&_lock)" calls, however between the first and second "dm_target_iterate" there is no lock held and the target modules can be loaded at this point, so the second "dm_target_iterate" call may need more space than what was the first "dm_target_iterate" returned. The code tries to handle this overflow (see the beginning of list_version_get_info), however this handling is incorrect. The code sets "param->data_size = param->data_start + needed" and "iter_info.end = (char *)vers+len" - "needed" is the size returned by the first dm_target_iterate call; "len" is the size of the buffer allocated by userspace. "len" may be greater than "needed"; in this case, the code will write up to "len" bytes into the buffer, however param->data_size is set to "needed", so it may write data past the param->data_size value. The ioctl interface copies only up to param->data_size into userspace, thus part of the result will be truncated. Fix this bug by setting "iter_info.end = (char *)vers + needed;" - this guarantees that the second "dm_target_iterate" call will write only up to the "needed" buffer and it will exit with "DM_BUFFER_FULL_FLAG" if it overflows the "needed" space - in this case, userspace will allocate a larger buffer and retry. Note that there is also a bug in list_version_get_needed - we need to add "strlen(tt->name) + 1" to the needed size, not "strlen(tt->name)".

In the Linux kernel, the following vulnerability has been resolved: perf/x86/amd: Fix crash due to race between amd_pmu_enable_all, perf NMI and throttling amd_pmu_enable_all() does: if (!test_bit(idx, cpuc->active_mask)) continue; amd_pmu_enable_event(cpuc->events[idx]); A perf NMI of another event can come between these two steps. Perf NMI handler internally disables and enables _all_ events, including the one which nmi-intercepted amd_pmu_enable_all() was in process of enabling. If that unintentionally enabled event has very low sampling period and causes immediate successive NMI, causing the event to be throttled, cpuc->events[idx] and cpuc->active_mask gets cleared by x86_pmu_stop(). This will result in amd_pmu_enable_event() getting called with event=NULL when amd_pmu_enable_all() resumes after handling the NMIs. This causes a kernel crash: BUG: kernel NULL pointer dereference, address: 0000000000000198 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page [...] Call Trace: <TASK> amd_pmu_enable_all+0x68/0xb0 ctx_resched+0xd9/0x150 event_function+0xb8/0x130 ? hrtimer_start_range_ns+0x141/0x4a0 ? perf_duration_warn+0x30/0x30 remote_function+0x4d/0x60 __flush_smp_call_function_queue+0xc4/0x500 flush_smp_call_function_queue+0x11d/0x1b0 do_idle+0x18f/0x2d0 cpu_startup_entry+0x19/0x20 start_secondary+0x121/0x160 secondary_startup_64_no_verify+0xe5/0xeb </TASK> amd_pmu_disable_all()/amd_pmu_enable_all() calls inside perf NMI handler were recently added as part of BRS enablement but I'm not sure whether we really need them. We can just disable BRS in the beginning and enable it back while returning from NMI. This will solve the issue by not enabling those events whose active_masks are set but are not yet enabled in hw pmu.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix data-races around sysctl_tcp_l3mdev_accept. While reading sysctl_tcp_l3mdev_accept, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix a data-race around sysctl_tcp_notsent_lowat. While reading sysctl_tcp_notsent_lowat, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: icmp: Fix data-races around sysctl. While reading icmp sysctl variables, they can be changed concurrently. So, we need to add READ_ONCE() to avoid data-races.

In the Linux kernel, the following vulnerability has been resolved: ipv4: Fix a data-race around sysctl_fib_multipath_use_neigh. While reading sysctl_fib_multipath_use_neigh, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: ip: Fix data-races around sysctl_ip_fwd_use_pmtu. While reading sysctl_ip_fwd_use_pmtu, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: cipso: Fix data-races around sysctl. While reading cipso sysctl variables, they can be changed concurrently. So, we need to add READ_ONCE() to avoid data-races.

In the Linux kernel, the following vulnerability has been resolved: nbd: fix race between nbd_alloc_config() and module removal When nbd module is being removing, nbd_alloc_config() may be called concurrently by nbd_genl_connect(), although try_module_get() will return false, but nbd_alloc_config() doesn't handle it. The race may lead to the leak of nbd_config and its related resources (e.g, recv_workq) and oops in nbd_read_stat() due to the unload of nbd module as shown below: BUG: kernel NULL pointer dereference, address: 0000000000000040 Oops: 0000 [#1] SMP PTI CPU: 5 PID: 13840 Comm: kworker/u17:33 Not tainted 5.14.0+ #1 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996) Workqueue: knbd16-recv recv_work [nbd] RIP: 0010:nbd_read_stat.cold+0x130/0x1a4 [nbd] Call Trace: recv_work+0x3b/0xb0 [nbd] process_one_work+0x1ed/0x390 worker_thread+0x4a/0x3d0 kthread+0x12a/0x150 ret_from_fork+0x22/0x30 Fixing it by checking the return value of try_module_get() in nbd_alloc_config(). As nbd_alloc_config() may return ERR_PTR(-ENODEV), assign nbd->config only when nbd_alloc_config() succeeds to ensure the value of nbd->config is binary (valid or NULL). Also adding a debug message to check the reference counter of nbd_config during module removal.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix data-races around sysctl_tcp_recovery. While reading sysctl_tcp_recovery, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: igmp: Fix data-races around sysctl_igmp_qrv. While reading sysctl_igmp_qrv, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers. This test can be packed into a helper, so such changes will be in the follow-up series after net is merged into net-next. qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);

In the Linux kernel, the following vulnerability has been resolved: rxrpc: Fix call timer start racing with call destruction The rxrpc_call struct has a timer used to handle various timed events relating to a call. This timer can get started from the packet input routines that are run in softirq mode with just the RCU read lock held. Unfortunately, because only the RCU read lock is held - and neither ref or other lock is taken - the call can start getting destroyed at the same time a packet comes in addressed to that call. This causes the timer - which was already stopped - to get restarted. Later, the timer dispatch code may then oops if the timer got deallocated first. Fix this by trying to take a ref on the rxrpc_call struct and, if successful, passing that ref along to the timer. If the timer was already running, the ref is discarded. The timer completion routine can then pass the ref along to the call's work item when it queues it. If the timer or work item where already queued/running, the extra ref is discarded.

In the Linux kernel, the following vulnerability has been resolved: ip: Fix a data-race around sysctl_fwmark_reflect. While reading sysctl_fwmark_reflect, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix a data-race around sysctl_tcp_early_retrans. While reading sysctl_tcp_early_retrans, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: list: fix a data-race around ep->rdllist ep_poll() first calls ep_events_available() with no lock held and checks if ep->rdllist is empty by list_empty_careful(), which reads rdllist->prev. Thus all accesses to it need some protection to avoid store/load-tearing. Note INIT_LIST_HEAD_RCU() already has the annotation for both prev and next. Commit bf3b9f6372c4 ("epoll: Add busy poll support to epoll with socket fds.") added the first lockless ep_events_available(), and commit c5a282e9635e ("fs/epoll: reduce the scope of wq lock in epoll_wait()") made some ep_events_available() calls lockless and added single call under a lock, finally commit e59d3c64cba6 ("epoll: eliminate unnecessary lock for zero timeout") made the last ep_events_available() lockless. BUG: KCSAN: data-race in do_epoll_wait / do_epoll_wait write to 0xffff88810480c7d8 of 8 bytes by task 1802 on cpu 0: INIT_LIST_HEAD include/linux/list.h:38 [inline] list_splice_init include/linux/list.h:492 [inline] ep_start_scan fs/eventpoll.c:622 [inline] ep_send_events fs/eventpoll.c:1656 [inline] ep_poll fs/eventpoll.c:1806 [inline] do_epoll_wait+0x4eb/0xf40 fs/eventpoll.c:2234 do_epoll_pwait fs/eventpoll.c:2268 [inline] __do_sys_epoll_pwait fs/eventpoll.c:2281 [inline] __se_sys_epoll_pwait+0x12b/0x240 fs/eventpoll.c:2275 __x64_sys_epoll_pwait+0x74/0x80 fs/eventpoll.c:2275 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae read to 0xffff88810480c7d8 of 8 bytes by task 1799 on cpu 1: list_empty_careful include/linux/list.h:329 [inline] ep_events_available fs/eventpoll.c:381 [inline] ep_poll fs/eventpoll.c:1797 [inline] do_epoll_wait+0x279/0xf40 fs/eventpoll.c:2234 do_epoll_pwait fs/eventpoll.c:2268 [inline] __do_sys_epoll_pwait fs/eventpoll.c:2281 [inline] __se_sys_epoll_pwait+0x12b/0x240 fs/eventpoll.c:2275 __x64_sys_epoll_pwait+0x74/0x80 fs/eventpoll.c:2275 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae value changed: 0xffff88810480c7d0 -> 0xffff888103c15098 Reported by Kernel Concurrency Sanitizer on: CPU: 1 PID: 1799 Comm: syz-fuzzer Tainted: G W 5.17.0-rc7-syzkaller-dirty #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011

In the Linux kernel, the following vulnerability has been resolved: icmp: Fix data-races around sysctl_icmp_echo_enable_probe. While reading sysctl_icmp_echo_enable_probe, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: sysctl: Fix data races in proc_douintvec_minmax(). A sysctl variable is accessed concurrently, and there is always a chance of data-race. So, all readers and writers need some basic protection to avoid load/store-tearing. This patch changes proc_douintvec_minmax() to use READ_ONCE() and WRITE_ONCE() internally to fix data-races on the sysctl side. For now, proc_douintvec_minmax() itself is tolerant to a data-race, but we still need to add annotations on the other subsystem's side.

In the Linux kernel, the following vulnerability has been resolved: zsmalloc: fix races between asynchronous zspage free and page migration The asynchronous zspage free worker tries to lock a zspage's entire page list without defending against page migration. Since pages which haven't yet been locked can concurrently migrate off the zspage page list while lock_zspage() churns away, lock_zspage() can suffer from a few different lethal races. It can lock a page which no longer belongs to the zspage and unsafely dereference page_private(), it can unsafely dereference a torn pointer to the next page (since there's a data race), and it can observe a spurious NULL pointer to the next page and thus not lock all of the zspage's pages (since a single page migration will reconstruct the entire page list, and create_page_chain() unconditionally zeroes out each list pointer in the process). Fix the races by using migrate_read_lock() in lock_zspage() to synchronize with page migration.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix a data-race around sysctl_tcp_mtu_probe_floor. While reading sysctl_tcp_mtu_probe_floor, it can be changed concurrently. Thus, we need to add READ_ONCE() to its reader.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix data-races around sysctl_tcp_fastopen_blackhole_timeout. While reading sysctl_tcp_fastopen_blackhole_timeout, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: af_unix: Fix a data-race in unix_dgram_peer_wake_me(). unix_dgram_poll() calls unix_dgram_peer_wake_me() without `other`'s lock held and check if its receive queue is full. Here we need to use unix_recvq_full_lockless() instead of unix_recvq_full(), otherwise KCSAN will report a data-race.

In the Linux kernel, the following vulnerability has been resolved: tcp: Fix data-races around sysctl_tcp_min_snd_mss. While reading sysctl_tcp_min_snd_mss, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: ipv4: Fix data-races around sysctl_fib_multipath_hash_fields. While reading sysctl_fib_multipath_hash_fields, it can be changed concurrently. Thus, we need to add READ_ONCE() to its readers.

In the Linux kernel, the following vulnerability has been resolved: ice: fix concurrent reset and removal of VFs Commit c503e63200c6 ("ice: Stop processing VF messages during teardown") introduced a driver state flag, ICE_VF_DEINIT_IN_PROGRESS, which is intended to prevent some issues with concurrently handling messages from VFs while tearing down the VFs. This change was motivated by crashes caused while tearing down and bringing up VFs in rapid succession. It turns out that the fix actually introduces issues with the VF driver caused because the PF no longer responds to any messages sent by the VF during its .remove routine. This results in the VF potentially removing its DMA memory before the PF has shut down the device queues. Additionally, the fix doesn't actually resolve concurrency issues within the ice driver. It is possible for a VF to initiate a reset just prior to the ice driver removing VFs. This can result in the remove task concurrently operating while the VF is being reset. This results in similar memory corruption and panics purportedly fixed by that commit. Fix this concurrency at its root by protecting both the reset and removal flows using the existing VF cfg_lock. This ensures that we cannot remove the VF while any outstanding critical tasks such as a virtchnl message or a reset are occurring. This locking change also fixes the root cause originally fixed by commit c503e63200c6 ("ice: Stop processing VF messages during teardown"), so we can simply revert it. Note that I kept these two changes together because simply reverting the original commit alone would leave the driver vulnerable to worse race conditions.

In the Linux kernel, the following vulnerability has been resolved: fscache: Fix oops due to race with cookie_lru and use_cookie If a cookie expires from the LRU and the LRU_DISCARD flag is set, but the state machine has not run yet, it's possible another thread can call fscache_use_cookie and begin to use it. When the cookie_worker finally runs, it will see the LRU_DISCARD flag set, transition the cookie->state to LRU_DISCARDING, which will then withdraw the cookie. Once the cookie is withdrawn the object is removed the below oops will occur because the object associated with the cookie is now NULL. Fix the oops by clearing the LRU_DISCARD bit if another thread uses the cookie before the cookie_worker runs. BUG: kernel NULL pointer dereference, address: 0000000000000008 ... CPU: 31 PID: 44773 Comm: kworker/u130:1 Tainted: G E 6.0.0-5.dneg.x86_64 #1 Hardware name: Google Compute Engine/Google Compute Engine, BIOS Google 08/26/2022 Workqueue: events_unbound netfs_rreq_write_to_cache_work [netfs] RIP: 0010:cachefiles_prepare_write+0x28/0x90 [cachefiles] ... Call Trace: netfs_rreq_write_to_cache_work+0x11c/0x320 [netfs] process_one_work+0x217/0x3e0 worker_thread+0x4a/0x3b0 kthread+0xd6/0x100

In the Linux kernel, the following vulnerability has been resolved: configfs: fix a race in configfs_{,un}register_subsystem() When configfs_register_subsystem() or configfs_unregister_subsystem() is executing link_group() or unlink_group(), it is possible that two processes add or delete list concurrently. Some unfortunate interleavings of them can cause kernel panic. One of cases is: A --> B --> C --> D A <-- B <-- C <-- D delete list_head *B | delete list_head *C --------------------------------|----------------------------------- configfs_unregister_subsystem | configfs_unregister_subsystem unlink_group | unlink_group unlink_obj | unlink_obj list_del_init | list_del_init __list_del_entry | __list_del_entry __list_del | __list_del // next == C | next->prev = prev | | next->prev = prev prev->next = next | | // prev == B | prev->next = next Fix this by adding mutex when calling link_group() or unlink_group(), but parent configfs_subsystem is NULL when config_item is root. So I create a mutex configfs_subsystem_mutex.

In the Linux kernel, the following vulnerability has been resolved: ice: Fix race condition during interface enslave Commit 5dbbbd01cbba83 ("ice: Avoid RTNL lock when re-creating auxiliary device") changes a process of re-creation of aux device so ice_plug_aux_dev() is called from ice_service_task() context. This unfortunately opens a race window that can result in dead-lock when interface has left LAG and immediately enters LAG again. Reproducer: ``` #!/bin/sh ip link add lag0 type bond mode 1 miimon 100 ip link set lag0 for n in {1..10}; do echo Cycle: $n ip link set ens7f0 master lag0 sleep 1 ip link set ens7f0 nomaster done ``` This results in: [20976.208697] Workqueue: ice ice_service_task [ice] [20976.213422] Call Trace: [20976.215871] __schedule+0x2d1/0x830 [20976.219364] schedule+0x35/0xa0 [20976.222510] schedule_preempt_disabled+0xa/0x10 [20976.227043] __mutex_lock.isra.7+0x310/0x420 [20976.235071] enum_all_gids_of_dev_cb+0x1c/0x100 [ib_core] [20976.251215] ib_enum_roce_netdev+0xa4/0xe0 [ib_core] [20976.256192] ib_cache_setup_one+0x33/0xa0 [ib_core] [20976.261079] ib_register_device+0x40d/0x580 [ib_core] [20976.266139] irdma_ib_register_device+0x129/0x250 [irdma] [20976.281409] irdma_probe+0x2c1/0x360 [irdma] [20976.285691] auxiliary_bus_probe+0x45/0x70 [20976.289790] really_probe+0x1f2/0x480 [20976.298509] driver_probe_device+0x49/0xc0 [20976.302609] bus_for_each_drv+0x79/0xc0 [20976.306448] __device_attach+0xdc/0x160 [20976.310286] bus_probe_device+0x9d/0xb0 [20976.314128] device_add+0x43c/0x890 [20976.321287] __auxiliary_device_add+0x43/0x60 [20976.325644] ice_plug_aux_dev+0xb2/0x100 [ice] [20976.330109] ice_service_task+0xd0c/0xed0 [ice] [20976.342591] process_one_work+0x1a7/0x360 [20976.350536] worker_thread+0x30/0x390 [20976.358128] kthread+0x10a/0x120 [20976.365547] ret_from_fork+0x1f/0x40 ... [20976.438030] task:ip state:D stack: 0 pid:213658 ppid:213627 flags:0x00004084 [20976.446469] Call Trace: [20976.448921] __schedule+0x2d1/0x830 [20976.452414] schedule+0x35/0xa0 [20976.455559] schedule_preempt_disabled+0xa/0x10 [20976.460090] __mutex_lock.isra.7+0x310/0x420 [20976.464364] device_del+0x36/0x3c0 [20976.467772] ice_unplug_aux_dev+0x1a/0x40 [ice] [20976.472313] ice_lag_event_handler+0x2a2/0x520 [ice] [20976.477288] notifier_call_chain+0x47/0x70 [20976.481386] __netdev_upper_dev_link+0x18b/0x280 [20976.489845] bond_enslave+0xe05/0x1790 [bonding] [20976.494475] do_setlink+0x336/0xf50 [20976.502517] __rtnl_newlink+0x529/0x8b0 [20976.543441] rtnl_newlink+0x43/0x60 [20976.546934] rtnetlink_rcv_msg+0x2b1/0x360 [20976.559238] netlink_rcv_skb+0x4c/0x120 [20976.563079] netlink_unicast+0x196/0x230 [20976.567005] netlink_sendmsg+0x204/0x3d0 [20976.570930] sock_sendmsg+0x4c/0x50 [20976.574423] ____sys_sendmsg+0x1eb/0x250 [20976.586807] ___sys_sendmsg+0x7c/0xc0 [20976.606353] __sys_sendmsg+0x57/0xa0 [20976.609930] do_syscall_64+0x5b/0x1a0 [20976.613598] entry_SYSCALL_64_after_hwframe+0x65/0xca 1. Command 'ip link ... set nomaster' causes that ice_plug_aux_dev() is called from ice_service_task() context, aux device is created and associated device->lock is taken. 2. Command 'ip link ... set master...' calls ice's notifier under RTNL lock and that notifier calls ice_unplug_aux_dev(). That function tries to take aux device->lock but this is already taken by ice_plug_aux_dev() in step 1 3. Later ice_plug_aux_dev() tries to take RTNL lock but this is already taken in step 2 4. Dead-lock The patch fixes this issue by following changes: - Bit ICE_FLAG_PLUG_AUX_DEV is kept to be set during ice_plug_aux_dev() call in ice_service_task() - The bit is checked in ice_clear_rdma_cap() and only if it is not set then ice_unplug_aux_dev() is called. If it is set (in other words plugging of aux device was requested and ice_plug_aux_dev() is potentially running) then the function only clears the ---truncated---

In the Linux kernel, the following vulnerability has been resolved: cfg80211: fix race in netlink owner interface destruction My previous fix here to fix the deadlock left a race where the exact same deadlock (see the original commit referenced below) can still happen if cfg80211_destroy_ifaces() already runs while nl80211_netlink_notify() is still marking some interfaces as nl_owner_dead. The race happens because we have two loops here - first we dev_close() all the netdevs, and then we destroy them. If we also have two netdevs (first one need only be a wdev though) then we can find one during the first iteration, close it, and go to the second iteration -- but then find two, and try to destroy also the one we didn't close yet. Fix this by only iterating once.

A use-after-free vulnerability was found in network namespaces code affecting the Linux kernel before 4.14.11. The function get_net_ns_by_id() in net/core/net_namespace.c does not check for the net::count value after it has found a peer network in netns_ids idr, which could lead to double free and memory corruption. This vulnerability could allow an unprivileged local user to induce kernel memory corruption on the system, leading to a crash. Due to the nature of the flaw, privilege escalation cannot be fully ruled out, although it is thought to be unlikely.

An issue was discovered in the Linux kernel through 5.19.8. drivers/firmware/efi/capsule-loader.c has a race condition with a resultant use-after-free.

In the Linux kernel, the following vulnerability has been resolved: net/mlx5e: Check for NOT_READY flag state after locking Currently the check for NOT_READY flag is performed before obtaining the necessary lock. This opens a possibility for race condition when the flow is concurrently removed from unready_flows list by the workqueue task, which causes a double-removal from the list and a crash[0]. Fix the issue by moving the flag check inside the section protected by uplink_priv->unready_flows_lock mutex. [0]: [44376.389654] general protection fault, probably for non-canonical address 0xdead000000000108: 0000 [#1] SMP [44376.391665] CPU: 7 PID: 59123 Comm: tc Not tainted 6.4.0-rc4+ #1 [44376.392984] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 [44376.395342] RIP: 0010:mlx5e_tc_del_fdb_flow+0xb3/0x340 [mlx5_core] [44376.396857] Code: 00 48 8b b8 68 ce 02 00 e8 8a 4d 02 00 4c 8d a8 a8 01 00 00 4c 89 ef e8 8b 79 88 e1 48 8b 83 98 06 00 00 48 8b 93 90 06 00 00 <48> 89 42 08 48 89 10 48 b8 00 01 00 00 00 00 ad de 48 89 83 90 06 [44376.399167] RSP: 0018:ffff88812cc97570 EFLAGS: 00010246 [44376.399680] RAX: dead000000000122 RBX: ffff8881088e3800 RCX: ffff8881881bac00 [44376.400337] RDX: dead000000000100 RSI: ffff88812cc97500 RDI: ffff8881242f71b0 [44376.401001] RBP: ffff88811cbb0940 R08: 0000000000000400 R09: 0000000000000001 [44376.401663] R10: 0000000000000001 R11: 0000000000000000 R12: ffff88812c944000 [44376.402342] R13: ffff8881242f71a8 R14: ffff8881222b4000 R15: 0000000000000000 [44376.402999] FS: 00007f0451104800(0000) GS:ffff88852cb80000(0000) knlGS:0000000000000000 [44376.403787] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [44376.404343] CR2: 0000000000489108 CR3: 0000000123a79003 CR4: 0000000000370ea0 [44376.405004] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [44376.405665] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [44376.406339] Call Trace: [44376.406651] <TASK> [44376.406939] ? die_addr+0x33/0x90 [44376.407311] ? exc_general_protection+0x192/0x390 [44376.407795] ? asm_exc_general_protection+0x22/0x30 [44376.408292] ? mlx5e_tc_del_fdb_flow+0xb3/0x340 [mlx5_core] [44376.408876] __mlx5e_tc_del_fdb_peer_flow+0xbc/0xe0 [mlx5_core] [44376.409482] mlx5e_tc_del_flow+0x42/0x210 [mlx5_core] [44376.410055] mlx5e_flow_put+0x25/0x50 [mlx5_core] [44376.410529] mlx5e_delete_flower+0x24b/0x350 [mlx5_core] [44376.411043] tc_setup_cb_reoffload+0x22/0x80 [44376.411462] fl_reoffload+0x261/0x2f0 [cls_flower] [44376.411907] ? mlx5e_rep_indr_setup_ft_cb+0x160/0x160 [mlx5_core] [44376.412481] ? mlx5e_rep_indr_setup_ft_cb+0x160/0x160 [mlx5_core] [44376.413044] tcf_block_playback_offloads+0x76/0x170 [44376.413497] tcf_block_unbind+0x7b/0xd0 [44376.413881] tcf_block_setup+0x17d/0x1c0 [44376.414269] tcf_block_offload_cmd.isra.0+0xf1/0x130 [44376.414725] tcf_block_offload_unbind+0x43/0x70 [44376.415153] __tcf_block_put+0x82/0x150 [44376.415532] ingress_destroy+0x22/0x30 [sch_ingress] [44376.415986] qdisc_destroy+0x3b/0xd0 [44376.416343] qdisc_graft+0x4d0/0x620 [44376.416706] tc_get_qdisc+0x1c9/0x3b0 [44376.417074] rtnetlink_rcv_msg+0x29c/0x390 [44376.419978] ? rep_movs_alternative+0x3a/0xa0 [44376.420399] ? rtnl_calcit.isra.0+0x120/0x120 [44376.420813] netlink_rcv_skb+0x54/0x100 [44376.421192] netlink_unicast+0x1f6/0x2c0 [44376.421573] netlink_sendmsg+0x232/0x4a0 [44376.421980] sock_sendmsg+0x38/0x60 [44376.422328] ____sys_sendmsg+0x1d0/0x1e0 [44376.422709] ? copy_msghdr_from_user+0x6d/0xa0 [44376.423127] ___sys_sendmsg+0x80/0xc0 [44376.423495] ? ___sys_recvmsg+0x8b/0xc0 [44376.423869] __sys_sendmsg+0x51/0x90 [44376.424226] do_syscall_64+0x3d/0x90 [44376.424587] entry_SYSCALL_64_after_hwframe+0x46/0xb0 [44376.425046] RIP: 0033:0x7f045134f887 [44376.425403] Code: 0a 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b9 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 2e 00 ---truncated---

In the Linux kernel, the following vulnerability has been resolved: power: supply: bq25890: Fix external_power_changed race bq25890_charger_external_power_changed() dereferences bq->charger, which gets sets in bq25890_power_supply_init() like this: bq->charger = devm_power_supply_register(bq->dev, &bq->desc, &psy_cfg); As soon as devm_power_supply_register() has called device_add() the external_power_changed callback can get called. So there is a window where bq25890_charger_external_power_changed() may get called while bq->charger has not been set yet leading to a NULL pointer dereference. This race hits during boot sometimes on a Lenovo Yoga Book 1 yb1-x90f when the cht_wcove_pwrsrc (extcon) power_supply is done with detecting the connected charger-type which happens to exactly hit the small window: BUG: kernel NULL pointer dereference, address: 0000000000000018 <snip> RIP: 0010:__power_supply_is_supplied_by+0xb/0xb0 <snip> Call Trace: <TASK> __power_supply_get_supplier_property+0x19/0x50 class_for_each_device+0xb1/0xe0 power_supply_get_property_from_supplier+0x2e/0x50 bq25890_charger_external_power_changed+0x38/0x1b0 [bq25890_charger] __power_supply_changed_work+0x30/0x40 class_for_each_device+0xb1/0xe0 power_supply_changed_work+0x5f/0xe0 <snip> Fixing this is easy. The external_power_changed callback gets passed the power_supply which will eventually get stored in bq->charger, so bq25890_charger_external_power_changed() can simply directly use the passed in psy argument which is always valid.

A race condition flaw was found in the Linux kernel sound subsystem due to improper locking. It could lead to a NULL pointer dereference while handling the SNDCTL_DSP_SYNC ioctl. A privileged local user (root or member of the audio group) could use this flaw to crash the system, resulting in a denial of service condition

In the Linux kernel, the following vulnerability has been resolved: accel/ivpu: Fix race condition when unbinding BOs Fix 'Memory manager not clean during takedown' warning that occurs when ivpu_gem_bo_free() removes the BO from the BOs list before it gets unmapped. Then file_priv_unbind() triggers a warning in drm_mm_takedown() during context teardown. Protect the unmapping sequence with bo_list_lock to ensure the BO is always fully unmapped when removed from the list. This ensures the BO is either fully unmapped at context teardown time or present on the list and unmapped by file_priv_unbind().

In the Linux kernel, the following vulnerability has been resolved: timers: Fix NULL function pointer race in timer_shutdown_sync() There is a race condition between timer_shutdown_sync() and timer expiration that can lead to hitting a WARN_ON in expire_timers(). The issue occurs when timer_shutdown_sync() clears the timer function to NULL while the timer is still running on another CPU. The race scenario looks like this: CPU0 CPU1 <SOFTIRQ> lock_timer_base() expire_timers() base->running_timer = timer; unlock_timer_base() [call_timer_fn enter] mod_timer() ... timer_shutdown_sync() lock_timer_base() // For now, will not detach the timer but only clear its function to NULL if (base->running_timer != timer) ret = detach_if_pending(timer, base, true); if (shutdown) timer->function = NULL; unlock_timer_base() [call_timer_fn exit] lock_timer_base() base->running_timer = NULL; unlock_timer_base() ... // Now timer is pending while its function set to NULL. // next timer trigger <SOFTIRQ> expire_timers() WARN_ON_ONCE(!fn) // hit ... lock_timer_base() // Now timer will detach if (base->running_timer != timer) ret = detach_if_pending(timer, base, true); if (shutdown) timer->function = NULL; unlock_timer_base() The problem is that timer_shutdown_sync() clears the timer function regardless of whether the timer is currently running. This can leave a pending timer with a NULL function pointer, which triggers the WARN_ON_ONCE(!fn) check in expire_timers(). Fix this by only clearing the timer function when actually detaching the timer. If the timer is running, leave the function pointer intact, which is safe because the timer will be properly detached when it finishes running.

In the Linux kernel, the following vulnerability has been resolved: iommu/amd/pgtbl: Fix possible race while increase page table level The AMD IOMMU host page table implementation supports dynamic page table levels (up to 6 levels), starting with a 3-level configuration that expands based on IOVA address. The kernel maintains a root pointer and current page table level to enable proper page table walks in alloc_pte()/fetch_pte() operations. The IOMMU IOVA allocator initially starts with 32-bit address and onces its exhuasted it switches to 64-bit address (max address is determined based on IOMMU and device DMA capability). To support larger IOVA, AMD IOMMU driver increases page table level. But in unmap path (iommu_v1_unmap_pages()), fetch_pte() reads pgtable->[root/mode] without lock. So its possible that in exteme corner case, when increase_address_space() is updating pgtable->[root/mode], fetch_pte() reads wrong page table level (pgtable->mode). It does compare the value with level encoded in page table and returns NULL. This will result is iommu_unmap ops to fail and upper layer may retry/log WARN_ON. CPU 0 CPU 1 ------ ------ map pages unmap pages alloc_pte() -> increase_address_space() iommu_v1_unmap_pages() -> fetch_pte() pgtable->root = pte (new root value) READ pgtable->[mode/root] Reads new root, old mode Updates mode (pgtable->mode += 1) Since Page table level updates are infrequent and already synchronized with a spinlock, implement seqcount to enable lock-free read operations on the read path.

In the Linux kernel, the following vulnerability has been resolved: net: bridge: switchdev: Skip MDB replays of deferred events on offload Before this change, generation of the list of MDB events to replay would race against the creation of new group memberships, either from the IGMP/MLD snooping logic or from user configuration. While new memberships are immediately visible to walkers of br->mdb_list, the notification of their existence to switchdev event subscribers is deferred until a later point in time. So if a replay list was generated during a time that overlapped with such a window, it would also contain a replay of the not-yet-delivered event. The driver would thus receive two copies of what the bridge internally considered to be one single event. On destruction of the bridge, only a single membership deletion event was therefore sent. As a consequence of this, drivers which reference count memberships (at least DSA), would be left with orphan groups in their hardware database when the bridge was destroyed. This is only an issue when replaying additions. While deletion events may still be pending on the deferred queue, they will already have been removed from br->mdb_list, so no duplicates can be generated in that scenario. To a user this meant that old group memberships, from a bridge in which a port was previously attached, could be reanimated (in hardware) when the port joined a new bridge, without the new bridge's knowledge. For example, on an mv88e6xxx system, create a snooping bridge and immediately add a port to it: root@infix-06-0b-00:~$ ip link add dev br0 up type bridge mcast_snooping 1 && \ > ip link set dev x3 up master br0 And then destroy the bridge: root@infix-06-0b-00:~$ ip link del dev br0 root@infix-06-0b-00:~$ mvls atu ADDRESS FID STATE Q F 0 1 2 3 4 5 6 7 8 9 a DEV:0 Marvell 88E6393X 33:33:00:00:00:6a 1 static - - 0 . . . . . . . . . . 33:33:ff:87:e4:3f 1 static - - 0 . . . . . . . . . . ff:ff:ff:ff:ff:ff 1 static - - 0 1 2 3 4 5 6 7 8 9 a root@infix-06-0b-00:~$ The two IPv6 groups remain in the hardware database because the port (x3) is notified of the host's membership twice: once via the original event and once via a replay. Since only a single delete notification is sent, the count remains at 1 when the bridge is destroyed. Then add the same port (or another port belonging to the same hardware domain) to a new bridge, this time with snooping disabled: root@infix-06-0b-00:~$ ip link add dev br1 up type bridge mcast_snooping 0 && \ > ip link set dev x3 up master br1 All multicast, including the two IPv6 groups from br0, should now be flooded, according to the policy of br1. But instead the old memberships are still active in the hardware database, causing the switch to only forward traffic to those groups towards the CPU (port 0). Eliminate the race in two steps: 1. Grab the write-side lock of the MDB while generating the replay list. This prevents new memberships from showing up while we are generating the replay list. But it leaves the scenario in which a deferred event was already generated, but not delivered, before we grabbed the lock. Therefore: 2. Make sure that no deferred version of a replay event is already enqueued to the switchdev deferred queue, before adding it to the replay list, when replaying additions.