In the Linux kernel, the following vulnerability has been resolved: netdev: fix double-free in netdev_nl_bind_rx_doit() Sashiko flags that genlmsg_reply() always consumes the skb. The error path calls nlmsg_free(rsp) so we can't jump directly to it. Let's not unbind, just propagate the error to the user. This is the typical way of handling genlmsg_reply() failures. They shouldn't happen unless user does something silly like calling the kernel with an already-full rcvbuf.

In the Linux kernel, the following vulnerability has been resolved: fhandle: fix UAF due to unlocked ->mnt_ns read in may_decode_fh() may_decode_fh() accesses mount::mnt_ns without holding any locks; that means the mount can concurrently be unmounted, and the mnt_namespace can concurrently be freed after an RCU grace period. This race can happens as follows, assuming that the mount point was created by open_tree(..., OPEN_TREE_CLONE): thread 1 thread 2 RCU __do_sys_open_by_handle_at do_handle_open handle_to_path may_decode_fh is_mounted [mount::mnt_ns access] [mount::mnt_ns access] __do_sys_close fput_close_sync __fput dissolve_on_fput umount_tree class_namespace_excl_destructor namespace_unlock free_mnt_ns mnt_ns_tree_remove call_rcu(mnt_ns_release_rcu) mnt_ns_release_rcu mnt_ns_release kfree [mnt_namespace::user_ns access] **UAF** Fix it by taking rcu_read_lock() around the mount::mnt_ns access, like in __prepend_path(). Additionally, document the semantics of mount::mnt_ns, and use WRITE_ONCE() for writers that can race with lockless readers. This bug is unreachable unless one of the following is set: - CONFIG_PREEMPTION - CONFIG_RCU_STRICT_GRACE_PERIOD because it requires an RCU grace period to happen during a syscall without an explicit preemption. This doesn't seem to have interesting security impact; worst-case, it could leak the result of an integer comparison to userspace (from the level check in cap_capable()), cause an endless loop, or crash the kernel by dereferencing an invalid address.

In the Linux kernel, the following vulnerability has been resolved: ipc: limit next_id allocation to the valid ID range The checkpoint/restore sysctl path can request the next SysV IPC id through ids->next_id. ipc_idr_alloc() currently forwards that request to idr_alloc() with an open-ended upper bound. If the valid tail of the SysV IPC id space is full, the allocation can spill beyond ipc_mni. The returned SysV IPC id still uses the normal index encoding, so later lookup and removal can target the wrong slot. This leaves the real IDR entry behind and breaks the IDR state for the object. The bug is in ipc_idr_alloc() in the checkpoint/restore path. 1. ids->next_id is passed to: idr_alloc(&ids->ipcs_idr, new, ipcid_to_idx(next_id), 0, ...) 2. The zero upper bound makes the allocation effectively open-ended. Once the valid SysV IPC tail is occupied, idr_alloc() can spill past ipc_mni and allocate an entry beyond the valid IPC id range. 3. The new object id is still encoded with the narrower SysV IPC index width: new->id = (new->seq << ipcmni_seq_shift()) + idx 4. Later removal goes through ipc_rmid(), which uses: ipcid_to_idx(ipcp->id) That truncates the real IDR index. An object actually stored at a high index can then be removed as if it lived at a low in-range index. 5. For shared memory, shm_destroy() frees the current object anyway, but the real high IDR slot is left behind as a dangling pointer. 6. A subsequent walk of /proc/sysvipc/shm reaches the stale IDR entry and dereferences freed memory. Prevent this by bounding the requested allocation to ipc_mni so the checkpoint/restore path fails once the valid range is exhausted.

In the Linux kernel, the following vulnerability has been resolved: drm/nouveau: avoid a use-after-free when BO init fails nouveau_bo_init() is backed by ttm_bo_init() and ferries its return code back to the caller. On failures, ttm_bo_init() invokes the provided destructor which should de-initialize and free the memory. Thus, when nouveau_bo_init() returns an error the gem object has already been released and the memory freed by nouveau_bo_del_ttm().

In the Linux kernel, the following vulnerability has been resolved: futex: Drop CLONE_THREAD requirement for private default hash alloc Currently need_futex_hash_allocate_default() depends on strict pthread semantics, abusing CLONE_THREAD. This breaks the non-concurrency assumptions when doing the mm->futex_ref pcpu allocations, leading to bugs[0] when sharing the mm in other ways; ie: BUG: KASAN: slab-use-after-free in futex_hash_put ... where the +1 bias can end up on a percpu counter that mm->futex_ref no longer points at. Loosen the check to cover any CLONE_VM clone, except vfork(). Excluding vfork keeps the existing paths untouched (no overhead), and we can't race in the first place: either the parent is suspended and the child runs alone, or mm->futex_ref is already allocated from an earlier CLONE_VM.

In the Linux kernel, the following vulnerability has been resolved: greybus: raw: fix use-after-free if write is called after disconnect If a user writes to the chardev after disconnect has been called, the kernel panics with the following trace (with CONFIG_INIT_ON_FREE_DEFAULT_ON=y): BUG: kernel NULL pointer dereference, address: 0000000000000218 ... Call Trace: <TASK> gb_operation_create_common+0x61/0x180 gb_operation_create_flags+0x28/0xa0 gb_operation_sync_timeout+0x6f/0x100 raw_write+0x7b/0xc7 [gb_raw] vfs_write+0xcf/0x420 ? task_mm_cid_work+0x136/0x220 ksys_write+0x63/0xe0 do_syscall_64+0xa4/0x290 entry_SYSCALL_64_after_hwframe+0x77/0x7f Disconnect calls gb_connection_destroy, which ends up freeing the connection object. When gb_operation_sync is called in the write file operations, its gets a freed connection as parameter and the kernel panics. The gb_connection_destroy cannot be moved out of the disconnect function, as the Greybus subsystem expect all connections belonging to a bundle to be destroyed when disconnect returns. To prevent this bug, use a rw lock to synchronize access between write and disconnect. This guarantees that the write function doesn't try to use a disconnected connection.

In the Linux kernel, the following vulnerability has been resolved: media: atomisp: Fix use after free in atomisp_alloc_css_stat_bufs() The "s3a_buf" is freed along with all the other items on the "asd->s3a_stats" list. It leads to a double free and a use after free.

In the Linux kernel, the following vulnerability has been resolved: net/sched: taprio: fix use-after-free in advance_sched() on schedule switch In advance_sched(), when should_change_schedules() returns true, switch_schedules() is called to promote the admin schedule to oper. switch_schedules() queues the old oper schedule for RCU freeing via call_rcu(), but 'next' still points into an entry of the old oper schedule. The subsequent 'next->end_time = end_time' and rcu_assign_pointer(q->current_entry, next) are use-after-free. Fix this by selecting 'next' from the new oper schedule immediately after switch_schedules(), and using its pre-calculated end_time. setup_first_end_time() sets the first entry's end_time to base_time + interval when the schedule is installed, so the value is already correct. The deleted 'end_time = sched_base_time(admin)' assignment was also harmful independently: it would overwrite the new first entry's pre-calculated end_time with just base_time.

In the Linux kernel, the following vulnerability has been resolved: wifi: nl80211: reject oversized EMA RNR lists nl80211_parse_rnr_elems() stores the parsed element count in a u8-backed cfg80211_rnr_elems::cnt field and uses that count to size the flexible array allocation. Reject nested NL80211_ATTR_EMA_RNR_ELEMS input once the count reaches 255, before incrementing it again. This keeps the parser aligned with the data structure it fills and matches the existing bound check used by nl80211_parse_mbssid_elems().

In the Linux kernel, the following vulnerability has been resolved: io_uring/poll: fix signed comparison in io_poll_get_ownership() io_poll_get_ownership() uses a signed comparison to check whether poll_refs has reached the threshold for the slowpath: if (unlikely(atomic_read(&req->poll_refs) >= IO_POLL_REF_BIAS)) atomic_read() returns int (signed). When IO_POLL_CANCEL_FLAG (BIT(31)) is set in poll_refs, the value becomes negative in signed arithmetic, so the >= 128 comparison always evaluates to false and the slowpath is never taken. Fix this by casting the atomic_read() result to unsigned int before the comparison, so that the cancel flag is treated as a large positive value and correctly triggers the slowpath.

In the Linux kernel, the following vulnerability has been resolved: ovl: keep err zero after successful ovl_cache_get() ovl_iterate_merged() stores PTR_ERR(cache) in err before checking IS_ERR(cache). On success err holds the truncated cache pointer and can be returned as a bogus non-zero error. The syzbot reproducer reaches this through overlay-on-overlay readdir: getdents64 iterate_dir(outer overlay file) ovl_iterate_merged() ovl_cache_get() ovl_dir_read_merged() ovl_dir_read() iterate_dir(inner overlay file) ovl_iterate_merged() Only compute PTR_ERR(cache) on the error path.

In the Linux kernel, the following vulnerability has been resolved: ip6mr: fix UAF issue in ip6mr_sk_done() when addrconf_init_net() failed If the initialization fails in calling addrconf_init_net(), devconf_all is the pointer that has been released. Then ip6mr_sk_done() is called to release the net, accessing devconf->mc_forwarding directly causes invalid pointer access. The process is as follows: setup_net() ops_init() addrconf_init_net() all = kmemdup(...) ---> alloc "all" ... net->ipv6.devconf_all = all; __addrconf_sysctl_register() ---> failed ... kfree(all); ---> ipv6.devconf_all invalid ... ops_exit_list() ... ip6mr_sk_done() devconf = net->ipv6.devconf_all; //devconf is invalid pointer if (!devconf || !atomic_read(&devconf->mc_forwarding)) The following is the Call Trace information: BUG: KASAN: use-after-free in ip6mr_sk_done+0x112/0x3a0 Read of size 4 at addr ffff888075508e88 by task ip/14554 Call Trace: <TASK> dump_stack_lvl+0x8e/0xd1 print_report+0x155/0x454 kasan_report+0xba/0x1f0 kasan_check_range+0x35/0x1b0 ip6mr_sk_done+0x112/0x3a0 rawv6_close+0x48/0x70 inet_release+0x109/0x230 inet6_release+0x4c/0x70 sock_release+0x87/0x1b0 igmp6_net_exit+0x6b/0x170 ops_exit_list+0xb0/0x170 setup_net+0x7ac/0xbd0 copy_net_ns+0x2e6/0x6b0 create_new_namespaces+0x382/0xa50 unshare_nsproxy_namespaces+0xa6/0x1c0 ksys_unshare+0x3a4/0x7e0 __x64_sys_unshare+0x2d/0x40 do_syscall_64+0x35/0x80 entry_SYSCALL_64_after_hwframe+0x46/0xb0 RIP: 0033:0x7f7963322547 </TASK> Allocated by task 14554: kasan_save_stack+0x1e/0x40 kasan_set_track+0x21/0x30 __kasan_kmalloc+0xa1/0xb0 __kmalloc_node_track_caller+0x4a/0xb0 kmemdup+0x28/0x60 addrconf_init_net+0x1be/0x840 ops_init+0xa5/0x410 setup_net+0x5aa/0xbd0 copy_net_ns+0x2e6/0x6b0 create_new_namespaces+0x382/0xa50 unshare_nsproxy_namespaces+0xa6/0x1c0 ksys_unshare+0x3a4/0x7e0 __x64_sys_unshare+0x2d/0x40 do_syscall_64+0x35/0x80 entry_SYSCALL_64_after_hwframe+0x46/0xb0 Freed by task 14554: kasan_save_stack+0x1e/0x40 kasan_set_track+0x21/0x30 kasan_save_free_info+0x2a/0x40 ____kasan_slab_free+0x155/0x1b0 slab_free_freelist_hook+0x11b/0x220 __kmem_cache_free+0xa4/0x360 addrconf_init_net+0x623/0x840 ops_init+0xa5/0x410 setup_net+0x5aa/0xbd0 copy_net_ns+0x2e6/0x6b0 create_new_namespaces+0x382/0xa50 unshare_nsproxy_namespaces+0xa6/0x1c0 ksys_unshare+0x3a4/0x7e0 __x64_sys_unshare+0x2d/0x40 do_syscall_64+0x35/0x80 entry_SYSCALL_64_after_hwframe+0x46/0xb0

In the Linux kernel, the following vulnerability has been resolved: parport: Fix race between port and client registration The parport subsystem registers port devices before they are fully initialised, resulting in a race condition where client drivers such as lp can attach to ports that are not completely initialised or even being torn down. When the port and client drivers are built as modules and loaded around the same time during boot, this occasionally results in a crash. I was able to make this happen reliably in a VM with a PC-style parallel port by patching parport_pc to fail probing: > --- a/drivers/parport/parport_pc.c > +++ b/drivers/parport/parport_pc.c > @@ -2069,7 +2069,7 @@ static struct parport *__parport_pc_probe_port(unsigned long int base, > if (!p) > goto out3; > > - base_res = request_region(base, 3, p->name); > + base_res = NULL; > if (!base_res) > goto out4; > and then running: while true; do modprobe lp & modprobe parport_pc wait rmmod lp parport_pc done for a few seconds. In the long term I think port registration should be changed to put the call to device_add() inside parport_announce_port(), but since the latter currently cannot fail this will require changing all port drivers. For now, add a flag to indicate whether a port has been "announced" and only try to attach client drivers to ports when the flag is set.

In the Linux kernel, the following vulnerability has been resolved: Bluetooth: btmtk: avoid UAF in btmtk_process_coredump hci_devcd_append may lead to the release of the skb, so it cannot be accessed once it is called. ================================================================== BUG: KASAN: slab-use-after-free in btmtk_process_coredump+0x2a7/0x2d0 [btmtk] Read of size 4 at addr ffff888033cfabb0 by task kworker/0:3/82 CPU: 0 PID: 82 Comm: kworker/0:3 Tainted: G U 6.6.40-lockdep-03464-g1d8b4eb3060e #1 b0b3c1cc0c842735643fb411799d97921d1f688c Hardware name: Google Yaviks_Ufs/Yaviks_Ufs, BIOS Google_Yaviks_Ufs.15217.552.0 05/07/2024 Workqueue: events btusb_rx_work [btusb] Call Trace: <TASK> dump_stack_lvl+0xfd/0x150 print_report+0x131/0x780 kasan_report+0x177/0x1c0 btmtk_process_coredump+0x2a7/0x2d0 [btmtk 03edd567dd71a65958807c95a65db31d433e1d01] btusb_recv_acl_mtk+0x11c/0x1a0 [btusb 675430d1e87c4f24d0c1f80efe600757a0f32bec] btusb_rx_work+0x9e/0xe0 [btusb 675430d1e87c4f24d0c1f80efe600757a0f32bec] worker_thread+0xe44/0x2cc0 kthread+0x2ff/0x3a0 ret_from_fork+0x51/0x80 ret_from_fork_asm+0x1b/0x30 </TASK> Allocated by task 82: stack_trace_save+0xdc/0x190 kasan_set_track+0x4e/0x80 __kasan_slab_alloc+0x4e/0x60 kmem_cache_alloc+0x19f/0x360 skb_clone+0x132/0xf70 btusb_recv_acl_mtk+0x104/0x1a0 [btusb] btusb_rx_work+0x9e/0xe0 [btusb] worker_thread+0xe44/0x2cc0 kthread+0x2ff/0x3a0 ret_from_fork+0x51/0x80 ret_from_fork_asm+0x1b/0x30 Freed by task 1733: stack_trace_save+0xdc/0x190 kasan_set_track+0x4e/0x80 kasan_save_free_info+0x28/0xb0 ____kasan_slab_free+0xfd/0x170 kmem_cache_free+0x183/0x3f0 hci_devcd_rx+0x91a/0x2060 [bluetooth] worker_thread+0xe44/0x2cc0 kthread+0x2ff/0x3a0 ret_from_fork+0x51/0x80 ret_from_fork_asm+0x1b/0x30 The buggy address belongs to the object at ffff888033cfab40 which belongs to the cache skbuff_head_cache of size 232 The buggy address is located 112 bytes inside of freed 232-byte region [ffff888033cfab40, ffff888033cfac28) The buggy address belongs to the physical page: page:00000000a174ba93 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x33cfa head:00000000a174ba93 order:1 entire_mapcount:0 nr_pages_mapped:0 pincount:0 anon flags: 0x4000000000000840(slab|head|zone=1) page_type: 0xffffffff() raw: 4000000000000840 ffff888100848a00 0000000000000000 0000000000000001 raw: 0000000000000000 0000000080190019 00000001ffffffff 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff888033cfaa80: fb fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc ffff888033cfab00: fc fc fc fc fc fc fc fc fa fb fb fb fb fb fb fb >ffff888033cfab80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff888033cfac00: fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc fc ffff888033cfac80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== Check if we need to call hci_devcd_complete before calling hci_devcd_append. That requires that we check data->cd_info.cnt >= MTK_COREDUMP_NUM instead of data->cd_info.cnt > MTK_COREDUMP_NUM, as we increment data->cd_info.cnt only once the call to hci_devcd_append succeeds.

In the Linux kernel, the following vulnerability has been resolved: virtiofs: fix UAF on submount umount iput() called from fuse_release_end() can Oops if the super block has already been destroyed. Normally this is prevented by waiting for num_waiting to go down to zero before commencing with super block shutdown. This only works, however, for the last submount instance, as the wait counter is per connection, not per superblock. Revert to using synchronous release requests for the auto_submounts case, which is virtiofs only at this time.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nft_ct: bail out on template ct in get eval I noticed this issue while looking at a historic syzbot report [1]. A rule like the one below is enough to trigger the bug: table ip t { chain pre { type filter hook prerouting priority raw; ct zone set 1 ct original saddr 1.2.3.4 accept } } The first expression attaches a per-cpu template ct via nft_ct_set_zone_eval() (nf_ct_tmpl_alloc -> kzalloc, tuple is all zero, nf_ct_l3num(ct) == 0). The next expression then calls nft_ct_get_eval() on the same skb, treats the template as a real ct and hits the 16-byte memcpy path. With dreg at NFT_REG32_15 this overflows past struct nft_regs on the kernel stack; with smaller dreg values it silently clobbers adjacent registers. Reject template ct at the eval entry and in nft_ct_get_fast_eval(), mirroring the check nft_ct_set_eval() already has. Additionally, bound the address copy in NFT_CT_SRC / NFT_CT_DST by priv->len instead of by nf_ct_l3num(ct): nf_ct_get_tuple() zeroes the tuple before pkt_to_tuple() fills in only the protocol-relevant leading bytes, so the trailing bytes of tuple->{src,dst}.u3.all are well-defined zero. priv->len is validated at rule load, so the copy size is now bounded by the destination register rather than by an untrusted field on the conntrack. [1]: https://syzkaller.appspot.com/bug?id=389cf09cb72926114fce90dc85a2c3231dcb647c

In the Linux kernel, the following vulnerability has been resolved: bpf: Use RCU-safe iteration in dev_map_redirect_multi() SKB path The DEVMAP_HASH branch in dev_map_redirect_multi() uses hlist_for_each_entry_safe() to iterate hash buckets, but this function runs under RCU protection (called from xdp_do_generic_redirect_map() in softirq context). Concurrent writers (__dev_map_hash_update_elem, dev_map_hash_delete_elem) modify the list using RCU primitives (hlist_add_head_rcu, hlist_del_rcu). hlist_for_each_entry_safe() performs plain pointer dereferences without rcu_dereference(), missing the acquire barrier needed to pair with writers' rcu_assign_pointer(). On weakly-ordered architectures (ARM64, POWER), a reader can observe a partially-constructed node. It also defeats CONFIG_PROVE_RCU lockdep validation and KCSAN data-race detection. Replace with hlist_for_each_entry_rcu() using rcu_read_lock_bh_held() as the lockdep condition, consistent with the rcu_dereference_check() used in the DEVMAP (non-hash) branch of the same functions. Also fix the same incorrect lockdep_is_held(&dtab->index_lock) condition in dev_map_enqueue_multi(), where the lock is not held either.

In the Linux kernel, the following vulnerability has been resolved: mm/vma: do not try to unmap a VMA if mmap_prepare() invoked from mmap() The mmap_prepare hook functionality includes the ability to invoke mmap_prepare() from the mmap() hook of existing 'stacked' drivers, that is ones which are capable of calling the mmap hooks of other drivers/file systems (e.g. overlayfs, shm). As part of the mmap_prepare action functionality, we deal with errors by unmapping the VMA should one arise. This works in the usual mmap_prepare case, as we invoke this action at the last moment, when the VMA is established in the maple tree. However, the mmap() hook passes a not-fully-established VMA pointer to the caller (which is the motivation behind the mmap_prepare() work), which is detached. So attempting to unmap a VMA in this state will be problematic, with the most obvious symptom being a warning in vma_mark_detached(), because the VMA is already detached. It's also unncessary - the mmap() handler will clean up the VMA on error. So to fix this issue, this patch propagates whether or not an mmap action is being completed via the compatibility layer or directly. If the former, then we do not attempt VMA cleanup, if the latter, then we do. This patch also updates the userland VMA tests to reflect the change.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nf_queue: hold bridge skb->dev while queued br_pass_frame_up() rewrites skb->dev from the ingress port to the bridge master before queueing bridge LOCAL_IN packets. NFQUEUE only holds references on state.in/out and bridge physdevs, so a queued bridge packet can retain a freed bridge master in skb->dev until reinjection. When the verdict is reinjected later, br_netif_receive_skb() re-enters the receive path with skb->dev still pointing at the freed bridge master, triggering a use-after-free. Store skb->dev in the queue entry, hold a reference on it for the queue lifetime, and use the saved device when dropping queued packets during NETDEV_DOWN handling.

In the Linux kernel, the following vulnerability has been resolved: drm/xe/dma-buf: handle empty bo and UAF races There look to be some nasty races here when triggering the invalidate_mappings hook: 1) We do xe_bo_alloc() followed by the attach, before the actual full bo init step in xe_dma_buf_init_obj(). However the bo is visible on the attachments list after the attach. This is bad since exporter driver, say amdgpu, can at any time call back into our invalidate_mappings hook, with an empty/bogus bo, leading to potential bugs/crashes. 2) Similar to 1) but here we get a UAF, when the invalidate_mappings hook is triggered. For example, we get as far as xe_bo_init_locked() but this fails in some way. But here the bo will be freed on error, but we still have it attached from dma-buf pov, so if the invalidate_mappings is now triggered then the bo we access is gone and we trigger UAF and more bugs/crashes. To fix this, move the attach step until after we actually have a fully set up buffer object. Note that the bo is not published to userspace until later, so not sure what the comment "Don't publish the bo until we have a valid attachment", is referring to. We have at least two different customers reporting hitting a NULL ptr deref in evict_flags when importing something from amdgpu, followed by triggering the evict flow. Hit rate is also pretty low, which would hint at some kind of race, so something like 1) or 2) might explain this. v2: - Shuffle the order of the ops slightly (no functional change) - Improve the comment to better explain the ordering (Matt B) (cherry picked from commit af1f2ad0c59fe4e2f924c526f66e968289d77971)

In the Linux kernel, the following vulnerability has been resolved: netfilter: nat: use kfree_rcu to release ops Florian Westphal says: "Historically this is not an issue, even for normal base hooks: the data path doesn't use the original nf_hook_ops that are used to register the callbacks. However, in v5.14 I added the ability to dump the active netfilter hooks from userspace. This code will peek back into the nf_hook_ops that are available at the tail of the pointer-array blob used by the datapath. The nat hooks are special, because they are called indirectly from the central nat dispatcher hook. They are currently invisible to the nfnl hook dump subsystem though. But once that changes the nat ops structures have to be deferred too." Update nf_nat_register_fn() to deal with partial exposition of the hooks from error path which can be also an issue for nfnetlink_hook.

In the Linux kernel, the following vulnerability has been resolved: s390/bpf: Zero-extend bpf prog return values and kfunc arguments s390x ABI requires callers to zero-extend unsigned arguments and sign-extend signed arguments, and callees to zero-extend unsigned return values and sign-extend signed return values. s390 BPF JIT currently implements only sign extension. Fix this omission and implement zero extension too.

In the Linux kernel, the following vulnerability has been resolved: accel/ethosu: fix IFM region index out-of-bounds in command stream parser NPU_SET_IFM_REGION extracts the region index with param & 0x7f, giving a maximum value of 127. However region_size[] and output_region[] in struct ethosu_validated_cmdstream_info are both sized to NPU_BASEP_REGION_MAX (8), giving valid indices [0..7]. Every other region assignment in the same switch uses param & 0x7: NPU_SET_OFM_REGION: st.ofm.region = param & 0x7; NPU_SET_IFM2_REGION: st.ifm2.region = param & 0x7; NPU_SET_WEIGHT_REGION: st.weight[0].region = param & 0x7; NPU_SET_SCALE_REGION: st.scale[0].region = param & 0x7; The 0x7f mask on IFM is inconsistent and appears to be a typo. feat_matrix_length() and calc_sizes() use the region index directly as an array subscript into the kzalloc'd info struct: info->region_size[fm->region] = max(...); A userspace caller supplying NPU_SET_IFM_REGION with param > 7 causes a write up to 127*8 = 1016 bytes past the start of region_size[], corrupting adjacent kernel heap data. Fix by applying the same & 0x7 mask used by all other region assignments.

In the Linux kernel, the following vulnerability has been resolved: mm/huge_memory: update file PMD counter before folio_put() __split_huge_pmd_locked() updates the file/shmem RSS counter after dropping the PMD mapping's folio reference. If folio_put() drops the last reference, mm_counter_file() can later read freed folio state via folio_test_swapbacked(). Move the counter update before folio_put().

In the Linux kernel, the following vulnerability has been resolved: USB: serial: kl5kusb105: fix bulk-out buffer overflow klsi_105_prepare_write_buffer() is called by the generic write path with the bulk-out buffer and its size (bulk_out_size, 64 bytes). It stores a two-byte length header at the start of the buffer and copies the payload from the write fifo starting at buf + KLSI_HDR_LEN, but passes the full buffer size as the number of bytes to copy: count = kfifo_out_locked(&port->write_fifo, buf + KLSI_HDR_LEN, size, &port->lock); When the fifo holds at least size bytes, size bytes are copied starting two bytes into the size-byte buffer, writing KLSI_HDR_LEN bytes past its end. Copy at most size - KLSI_HDR_LEN bytes instead, leaving room for the header as safe_serial already does. Writing bulk_out_size or more bytes to the tty triggers a slab out-of-bounds write, observed with KASAN by emulating the device with dummy_hcd and raw-gadget: BUG: KASAN: slab-out-of-bounds in kfifo_copy_out+0x83/0xc0 Write of size 64 at addr ffff888112c62202 by task python3 kfifo_copy_out klsi_105_prepare_write_buffer [kl5kusb105] usb_serial_generic_write_start [usbserial] Allocated by task 139: usb_serial_probe [usbserial] The buggy address is located 2 bytes inside of allocated 64-byte region The out-of-bounds write no longer occurs with this change applied.

In the Linux kernel, the following vulnerability has been resolved: Bluetooth: ISO: Fix a use-after-free of the hci_conn pointer In iso_sock_rebind_bc(), the bis pointer is cached, then the socket lock is dropped: bis = iso_pi(sk)->conn->hcon; /* Release the socket before lookups since that requires hci_dev_lock * which shall not be acquired while holding sock_lock for proper * ordering. */ release_sock(sk); hci_dev_lock(bis->hdev); During the unlocked window, could a concurrent close() destroy the connection and free the bis structure, causing hci_dev_lock(bis->hdev) to access memory after it is freed, fix this by using the hdev reference which was safely acquired via iso_conn_get_hdev().

In the Linux kernel, the following vulnerability has been resolved: idpf: fix double free and use-after-free in aux device error paths When auxiliary_device_add() fails in idpf_plug_vport_aux_dev() or idpf_plug_core_aux_dev(), the err_aux_dev_add label calls auxiliary_device_uninit() and falls through to err_aux_dev_init. The uninit call will trigger put_device(), which invokes the release callback (idpf_vport_adev_release / idpf_core_adev_release) that frees iadev. The fall-through then reads adev->id from the freed iadev for ida_free() and double-frees iadev with kfree(). Free the IDA slot and clear the back-pointer before uninit, while adev is still valid, then return immediately. Commit 65637c3a1811 ("idpf: fix UAF in RDMA core aux dev deinitialization") fixed the same use-after-free in the matching unplug path in this file but missed both probe error paths.

In the Linux kernel, the following vulnerability has been resolved: net: ibm: emac: Fix use-after-free during device removal The driver was using devm_register_netdev() which causes unregister_netdev() to be deferred until the devres cleanup phase, which runs after emac_remove() returns. This creates a use-after-free window where: 1. emac_remove() is called, which tears down hardware (cancels work, detaches modules, unregisters from MAL) 2. emac_remove() returns 3. devres cleanup runs and finally calls unregister_netdev() During step 3, the network stack might still process packets, triggering emac_irq(), emac_poll(), or other handlers that access now-freed hardware resources (dev->emacp, dev->mal, etc.). Fix this by replacing devm_register_netdev() with manual register_netdev() and calling unregister_netdev() at the beginning of emac_remove(), before any hardware teardown. This ensures the network device is fully stopped and unregistered before hardware resources are released. The change is safe because: - dev->ndev is assigned very early in probe (before any error paths that could bypass emac_remove) - platform_set_drvdata() is only called after successful registration, so emac_remove() only runs for fully registered devices - unregister_netdev() is idempotent and safe to call on any registered device

In the Linux kernel, the following vulnerability has been resolved: RDMA: During rereg_mr ensure that REREG_ACCESS is compatible If IB_MR_REREG_ACCESS changes from RO to RW then the umem has to be re-evaluated to ensure it is properly pinned as RW. Since the umem is hidden inside each driver's mr struct add a ib_umem_check_rereg() function that each driver has to call before processing IB_MR_REREG_ACCESS. mlx4 has to retain its duplicate ib_access_writable check because it implements IB_MR_REREG_ACCESS | IB_MR_REREG_TRANS by changing both items in place sequentially while the MR is live, so it will continue to not support this combination.

In the Linux kernel, the following vulnerability has been resolved: drm/xe: Fix error cleanup in xe_exec_queue_create_ioctl() Two error handling issues exist in xe_exec_queue_create_ioctl(): 1. When xe_hw_engine_group_add_exec_queue() fails, the error path jumps to put_exec_queue which skips xe_exec_queue_kill(). If the VM is in preempt fence mode, xe_vm_add_compute_exec_queue() has already added the queue to the VM's compute exec queue list. Skipping the kill leaves the queue on that list, leading to a dangling pointer after the queue is freed. 2. When xa_alloc() fails after xe_hw_engine_group_add_exec_queue() has succeeded, the error path does not call xe_hw_engine_group_del_exec_queue() to remove the queue from the hw engine group list. The queue is then freed while still linked into the hw engine group, causing a use-after-free. Fix both by: - Changing the xe_hw_engine_group_add_exec_queue() failure path to jump to kill_exec_queue so that xe_exec_queue_kill() properly removes the queue from the VM's compute list. - Adding a del_hw_engine_group label before kill_exec_queue for the xa_alloc() failure path, which removes the queue from the hw engine group before proceeding with the rest of the cleanup. (cherry picked from commit 37c831f401746a45d510b312b0ed7a77b1e06ec8)

In the Linux kernel, the following vulnerability has been resolved: drm/amd/display: Clamp VBIOS HDMI retimer register count to array size [Why & How] The VBIOS integrated info tables (v1_11 and v2_1) contain HdmiRegNum and Hdmi6GRegNum fields that are used as loop bounds when copying retimer I2C register settings into fixed-size arrays (dp*_ext_hdmi_reg_settings[9] and dp*_ext_hdmi_6g_reg_settings[3]). These u8 fields are not validated before use, so a malformed VBIOS can specify values up to 255, causing an out-of-bounds heap write during driver probe. Clamp each register count to the destination array size using min_t() before the copy loops, in both get_integrated_info_v11() and get_integrated_info_v2_1(). (cherry picked from commit 5a7f0ef90195940c54b0f5bb85b87da55f038c69)

In the Linux kernel, the following vulnerability has been resolved: xfrm: policy: fix use-after-free on inexact bin in xfrm_policy_bysel_ctx() Fix the race by pruning the bin while still holding xfrm_policy_lock, before dropping it. Use __xfrm_policy_inexact_prune_bin() directly since the lock is already held. The wrapper xfrm_policy_inexact_prune_bin() becomes unused and is removed. Race: CPU0 (XFRM_MSG_DELPOLICY) CPU1 (XFRM_MSG_NEWSPDINFO) ========================== ========================== xfrm_policy_bysel_ctx(): spin_lock_bh(xfrm_policy_lock) bin = xfrm_policy_inexact_lookup() __xfrm_policy_unlink(pol) spin_unlock_bh(xfrm_policy_lock) xfrm_policy_kill(ret) // wide window, lock not held xfrm_hash_rebuild(): spin_lock_bh(xfrm_policy_lock) __xfrm_policy_inexact_flush(): kfree_rcu(bin) // bin freed spin_unlock_bh(xfrm_policy_lock) xfrm_policy_inexact_prune_bin(bin) // UAF: bin is freed

In the Linux kernel, the following vulnerability has been resolved: mailbox: mailbox-test: don't free the reused channel The RX channel can be aliased to the TX channel if it has a different MMIO. This special case needs to be handled when freeing the channels otherwise a double-free occurs.

In the Linux kernel, the following vulnerability has been resolved: batman-adv: fix tp_meter counter underflow during shutdown batadv_tp_sender_shutdown() unconditionally decrements the "sending" atomic counter. If multiple paths (e.g. timeout, user cancel, and normal finish) call this function, the counter can underflow to -1. Since the sender logic treats any non-zero value as "still sending", a negative value causes the sender kthread to loop indefinitely. This leads to a use-after-free when the interface is removed while the zombie thread is still active. Fix this by using atomic_xchg() to ensure the counter only transitions from 1 to 0 once. [sven: added missing change in batadv_tp_send]

In the Linux kernel, the following vulnerability has been resolved: netfilter: nft_tunnel: fix use-after-free on object destroy nft_tunnel_obj_destroy() calls metadata_dst_free() which directly kfree()s the metadata_dst, ignoring the dst_entry refcount. Packets that took a reference via dst_hold() in nft_tunnel_obj_eval() and are still queued (e.g. in a netem qdisc) are left with a dangling pointer. When these packets are eventually dequeued, dst_release() operates on freed memory. Replace metadata_dst_free() with dst_release() so the metadata_dst is freed only after all references are dropped. The dst subsystem already handles metadata_dst cleanup in dst_destroy() when DST_METADATA is set.

In the Linux kernel, the following vulnerability has been resolved: ipv4: account for fraggap on the paged allocation path In __ip_append_data(), when the paged-allocation branch is taken, alloclen and pagedlen are computed as alloclen = fragheaderlen + transhdrlen; pagedlen = datalen - transhdrlen; datalen already includes fraggap, but the fraggap bytes carried over from the previous skb are copied into the new skb's linear area at offset transhdrlen by the subsequent skb_copy_and_csum_bits(). The linear area is therefore undersized by fraggap bytes while pagedlen is overstated by the same amount. The non-paged branch sets alloclen to fraglen, which already accounts for fraggap because datalen does. Bring the paged branch in line by adding fraggap to alloclen and subtracting it from pagedlen. After this adjustment, copy no longer collapses to -fraggap on the paged path, so remove the stale comment describing that old arithmetic.

In the Linux kernel, the following vulnerability has been resolved: accel/ivpu: Fix signed integer truncation in IPC receive Fix potential buffer overflow where firmware-supplied data_size is cast to signed int before being used in min_t(). Large unsigned values (>= 0x80000000) become negative, causing unsigned wraparound and oversized memcpy operations that can overflow the stack buffer. Change min_t(int, ...) to min() as both values are unsigned and can be handled by min() without explicit cast.

In the Linux kernel, the following vulnerability has been resolved: misc: fastrpc: Fix copy buffer page size For non-registered buffer, fastrpc driver copies the buffer and pass it to the remote subsystem. There is a problem with current implementation of page size calculation which is not considering the offset in the calculation. This might lead to passing of improper and out-of-bounds page size which could result in memory issue. Calculate page start and page end using the offset adjusted address instead of absolute address.

In the Linux kernel, the following vulnerability has been resolved: net: skbuff: fix missing zerocopy reference in pskb_carve helpers pskb_carve_inside_header() and pskb_carve_inside_nonlinear() both copy the old skb_shared_info header into a new buffer via memcpy(), which includes the destructor_arg pointer (uarg) for MSG_ZEROCOPY skbs. Neither function calls net_zcopy_get() for the new shinfo, creating an unaccounted holder: every skb_shared_info with destructor_arg set will call skb_zcopy_clear() once when freed, but the corresponding net_zcopy_get() was never called for the new copy. Repeated calls drive uarg->refcnt to zero prematurely, freeing ubuf_info_msgzc while TX skbs still hold live destructor_arg pointers. KASAN reports use-after-free on a freed ubuf_info_msgzc: BUG: KASAN: slab-use-after-free in skb_release_data+0x77b/0x810 Read of size 8 at addr ffff88801574d3e8 by task poc/220 Call Trace: skb_release_data+0x77b/0x810 kfree_skb_list_reason+0x13e/0x610 skb_release_data+0x4cd/0x810 sk_skb_reason_drop+0xf3/0x340 skb_queue_purge_reason+0x282/0x440 rds_tcp_inc_free+0x1e/0x30 rds_recvmsg+0x354/0x1780 __sys_recvmsg+0xdf/0x180 Allocated by task 219: msg_zerocopy_realloc+0x157/0x7b0 tcp_sendmsg_locked+0x2892/0x3ba0 Freed by task 219: ip_recv_error+0x74a/0xb10 tcp_recvmsg+0x475/0x530 The skb consuming the late access still referenced the same uarg via shinfo->destructor_arg copied by pskb_carve_inside_nonlinear() without a refcount bump. This has been verified to be reliably exploitable: a working proof-of-concept achieves full root privilege escalation from an unprivileged local user on a default kernel configuration. The fix follows the pattern of pskb_expand_head() which has the same memcpy/cloned structure. For pskb_carve_inside_header(), net_zcopy_get() is placed after skb_orphan_frags() succeeds, so the orphan error path needs no cleanup. For pskb_carve_inside_nonlinear(), net_zcopy_get() is placed after all failure points and just before skb_release_data(), so no error path needs cleanup at all -- matching pskb_expand_head() more closely and avoiding the need for a balancing net_zcopy_put().

In the Linux kernel, the following vulnerability has been resolved: RDMA/umem: Fix truncation for block sizes >= 4G When the iommu is used the linearization of the mapping can give a single block that is very large split across multiple SG entries. When __rdma_block_iter_next() reassembles the split SG entries it is overflowing the 32 bit stack values and computed the wrong DMA addresses for blocks after the truncation. Use the right types to hold DMA addresses.

In the Linux kernel, the following vulnerability has been resolved: drm/amdkfd: Fix buffer overflow in SDMA queue checkpoint/restore on GFX11 The v11 MQD manager incorrectly assigned the CP-compute variants of checkpoint_mqd/restore_mqd for KFD_MQD_TYPE_SDMA queues. These functions use sizeof(struct v11_compute_mqd) (2048 bytes) instead of sizeof(struct v11_sdma_mqd) (512 bytes), causing a 1536-byte overflow. During CRIU checkpoint of an SDMA queue on Navi3x: - checkpoint_mqd() reads 2048 bytes from a 512-byte SDMA MQD buffer, leaking 1536 bytes of adjacent GTT memory to userspace During CRIU restore: - restore_mqd() writes 2048 bytes into a 512-byte SDMA MQD buffer, corrupting 1536 bytes of adjacent GTT memory (often the ring buffer or neighboring MQDs) This is a copy-paste regression unique to v11. All other ASIC backends (cik, vi, v9, v10, v12) correctly use the SDMA-specific variants. Add checkpoint_mqd_sdma() and restore_mqd_sdma() functions that properly handle the smaller v11_sdma_mqd structure, matching the pattern used in other MQD managers. (cherry picked from commit 6fa41db7ffdec97d62433adf03b7b9b759af8c2c)

In the Linux kernel, the following vulnerability has been resolved: net: phonet: free phonet_device after RCU grace period phonet_device_destroy() removes a phonet_device from the per-net device list with list_del_rcu(), but frees it immediately. RCU readers walking the same list can still hold a pointer to the object after it has been removed, leading to a slab-use-after-free. Use kfree_rcu(), matching the lifetime rule already used by phonet_address_del() for the same object type.

In the Linux kernel, the following vulnerability has been resolved: ALSA: PCM: Fix wait queue list corruption in snd_pcm_drain() on linked streams snd_pcm_drain() uses init_waitqueue_entry which does not clear entry.prev/next, and add_wait_queue with a conditional remove_wait_queue that is skipped when to_check is no longer in the group after concurrent UNLINK. The orphaned wait entry remains on the unlinked substream sleep queue. On the next drain iteration, add_wait_queue adds the entry to a new queue while still linked on the old one, corrupting both lists. A subsequent wake_up dereferences NULL at the func pointer (mapped from the spinlock at offset 0 of the misinterpreted wait_queue_head_t), causing a kernel panic. Replace init_waitqueue_entry/add_wait_queue/conditional remove_wait_queue with init_wait_entry/prepare_to_wait/ finish_wait. init_wait_entry clears prev/next via INIT_LIST_HEAD on each iteration and sets autoremove_wake_function which auto-removes the entry on wake-up. finish_wait safely handles both the already-removed and still-queued cases.

In the Linux kernel, the following vulnerability has been resolved: zram: fix use-after-free in zram_bvec_write_partial() zram_read_page() picks the sync or async backing device read path based on whether the parent bio is NULL. zram_bvec_write_partial() passes its parent bio down, so for ZRAM_WB slots the read is dispatched asynchronously and zram_read_page() returns 0 while the bio is still in flight. The caller then runs memcpy_from_bvec(), zram_write_page() and __free_page() on the buffer, leaving the async read to write into a freed page. zram_bvec_read_partial() was switched to NULL in commit 4e3c87b9421d ("zram: fix synchronous reads") for the same reason; the write_partial counterpart was missed.

In the Linux kernel, the following vulnerability has been resolved: pfifo_tail_enqueue: Drop new packet when sch->limit == 0 Expected behaviour: In case we reach scheduler's limit, pfifo_tail_enqueue() will drop a packet in scheduler's queue and decrease scheduler's qlen by one. Then, pfifo_tail_enqueue() enqueue new packet and increase scheduler's qlen by one. Finally, pfifo_tail_enqueue() return `NET_XMIT_CN` status code. Weird behaviour: In case we set `sch->limit == 0` and trigger pfifo_tail_enqueue() on a scheduler that has no packet, the 'drop a packet' step will do nothing. This means the scheduler's qlen still has value equal 0. Then, we continue to enqueue new packet and increase scheduler's qlen by one. In summary, we can leverage pfifo_tail_enqueue() to increase qlen by one and return `NET_XMIT_CN` status code. The problem is: Let's say we have two qdiscs: Qdisc_A and Qdisc_B. - Qdisc_A's type must have '->graft()' function to create parent/child relationship. Let's say Qdisc_A's type is `hfsc`. Enqueue packet to this qdisc will trigger `hfsc_enqueue`. - Qdisc_B's type is pfifo_head_drop. Enqueue packet to this qdisc will trigger `pfifo_tail_enqueue`. - Qdisc_B is configured to have `sch->limit == 0`. - Qdisc_A is configured to route the enqueued's packet to Qdisc_B. Enqueue packet through Qdisc_A will lead to: - hfsc_enqueue(Qdisc_A) -> pfifo_tail_enqueue(Qdisc_B) - Qdisc_B->q.qlen += 1 - pfifo_tail_enqueue() return `NET_XMIT_CN` - hfsc_enqueue() check for `NET_XMIT_SUCCESS` and see `NET_XMIT_CN` => hfsc_enqueue() don't increase qlen of Qdisc_A. The whole process lead to a situation where Qdisc_A->q.qlen == 0 and Qdisc_B->q.qlen == 1. Replace 'hfsc' with other type (for example: 'drr') still lead to the same problem. This violate the design where parent's qlen should equal to the sum of its childrens'qlen. Bug impact: This issue can be used for user->kernel privilege escalation when it is reachable.

In the Linux kernel, the following vulnerability has been resolved: thunderbolt: Clamp XDomain response data copy to allocation size tb_xdp_properties_request() derives the per-packet copy length from the response header without checking that it fits in the previously allocated data buffer. A malicious peer can set its length field larger than the declared data_length, causing memcpy to write past the kcalloc allocation. Clamp the per-packet copy length so that the cumulative offset never exceeds data_len.

In the Linux kernel, the following vulnerability has been resolved: drm/gem: Try to fix change_handle ioctl, attempt 4 [airlied: just added some comments on how to reenable] On-list because the cat is out of the bag and we're clearly not good enough to figure this out in private. The story thus far: 5e28b7b94408 ("drm: Set old handle to NULL before prime swap in change_handle") tried to fix a race condition between the gem_close and gem_change_handle ioctls, but got a few things wrong: - There's a confusion with the local variable handle, which is actually the new handle, and so the two-stage trick was actually applied to the wrong idr slot. 7164d78559b0 ("drm/gem: fix race between change_handle and handle_delete") tried to fix that by adding yet another code block, but forgot to add the error handling. Which meant we now have two paths, both kinda wrong. - dc366607c41c ("drm: Replace old pointer to new idr") tried to apply another fix, but inconsistently, again because of the handle confusion - this would be the right fix (kinda, somewhat, it's a mess) if we'd do the two-stage approach for the new handle. Except that wasn't the intent of the original fix. We also didn't have an igt merged for the original ioctl, which is a big no-go. This was attempted to address off-list in the original bugfix, and amd QA people claimed the bug was fixed now. Very clearly that's not the case. Here's my attempt to sort this out: - Rename the local variable to new_handle, the old aliasing with args->handle is just too dangerously confusing. - Merge the gem obj lookup with the two-stage idr_replace so that we avoid getting ourselves confused there. - This means we don't have a surplus temporary reference anymore, only an inherited from the idr. A concurrent gem_close on the new_handle could steal that. Fix that with the same two-stage approach create_tail uses. This is a bit overkill as documented in the comment, but I also don't trust my ability to understand this all correctly, so go with the established pattern we have from other ioctls instead for maximum paranoia. - Adjust error paths. I've tried to make the error and success paths common, because they are identical except for which handle is removed and on which we call idr_replace to (re)install the object again. But that made things messier to read, so I've left it at the more verbose version, which unfortunately hides the symmetry in the entire code flow a bit. - While at it, also replace the 7 space indent with 1 tab. And finally, because I flat out don't trust my abilities here at all anymore: - Disable the ioctl until we have the igt situation and everything else sorted out on-list and with full consensus. v2: Sashiko noticed that I didn't handle the error path for idr_replace correctly, it must be checked with IS_ERR_OR_NULL like in gem_handle_delete. So yeah, definitely should just the existing paths 1:1 because this is endless amounts of tricky. Also add the Fixes: line for the original ioctl, I forgot that too.

In the Linux kernel, the following vulnerability has been resolved: net/rds: Restrict use of RDS/IB to the initial network namespace Prevent using RDS/IB in network namespaces other than the initial one. The existing RDS/IB code will not work properly in non-initial network namespaces.

In the Linux kernel, the following vulnerability has been resolved: virt: sev-guest: Do not use host-controlled page order in cleanup path When issuing an extended guest request (SVM_VMGEXIT_EXT_GUEST_REQUEST), get_ext_report() allocates a buffer to retrieve a certificate blob from the host, keeping track of its size in report_req->certs_len. However, the host may return SNP_GUEST_VMM_ERR_INVALID_LEN, indicating an invalid buffer size, as well as the expected length of such buffer. get_ext_report() subsequently updates report_req->certs_len with the host-controlled value, and cleans up the buffer by computing a page order from such value. This is incorrect, as the host-provided length may not match the page order of the original allocation, potentially resulting in corruption in the page allocator. Fix this by using alloc_pages_exact() instead, and reusing @npages to compute the size passed to free_pages_exact(). For consistency, also use @npages to compute the size when allocating the pages, even though this last change has no functional effect.

In the Linux kernel, the following vulnerability has been resolved: KVM: Reject wrapped offset in kvm_reset_dirty_gfn() kvm_reset_dirty_gfn() guards the gfn range with if (!memslot || (offset + __fls(mask)) >= memslot->npages) return; but offset is u64 and the addition is unchecked. The check can be silently bypassed by a u64 wrap. The dirty ring backing those entries is MAP_SHARED at KVM_DIRTY_LOG_PAGE_OFFSET of the vcpu fd, so the VMM can rewrite the slot and offset fields of any entry between when the kernel pushes them and when KVM_RESET_DIRTY_RINGS consumes them. On reset, kvm_dirty_ring_reset() re-reads the values via READ_ONCE() and feeds them straight back into this check; only the flags handshake is treated as the handover, the slot/offset payload is taken on trust. Crafting two entries entry[i].offset = 0xffffffffffffffc1 entry[i+1].offset = 0 makes the coalescing loop in kvm_dirty_ring_reset() compute delta = (s64)(0 - 0xffffffffffffffc1) = 63 which falls in [0, BITS_PER_LONG), so it folds entry[i+1] into the existing mask by setting bit 63. The trailing kvm_reset_dirty_gfn() call then sees offset = 0xffffffffffffffc1 and __fls(mask) = 63; the sum is 0 in u64 and the bounds check passes. That offset propagates into kvm_arch_mmu_enable_log_dirty_pt_masked() unchanged. On the legacy MMU path -- kvm_memslots_have_rmaps() == true, i.e. shadow paging, any VM that has allocated shadow roots, or a write-tracked slot -- it reaches gfn_to_rmap(), which indexes slot->arch.rmap[0][] with a near-U64_MAX gfn. That is an out-of-bounds load of a kvm_rmap_head, followed by a conditional clear of PT_WRITABLE_MASK in whatever the loaded pointer points at. The path is reachable from any process holding /dev/kvm. Range-check offset on its own first, so the addition cannot wrap. memslot->npages is bounded well below U64_MAX, so once offset < npages holds, offset + __fls(mask) (with __fls(mask) < BITS_PER_LONG) stays in range.