System vulnerabilities

In the Linux kernel, the following vulnerability has been resolved: net/mlx5e: Fix deadlocks between devlink and netdev instance locks In the mentioned "Fixes" commit, various work tasks triggering devlink health reporter recovery were switched to use netdev_trylock to protect against concurrent tear down of the channels being recovered. But this had the side effect of introducing potential deadlocks because of incorrect lock ordering. The correct lock order is described by the init flow: probe_one -> mlx5_init_one (acquires devlink lock) -> mlx5_init_one_devl_locked -> mlx5_register_device -> mlx5_rescan_drivers_locked -...-> mlx5e_probe -> _mlx5e_probe -> register_netdev (acquires rtnl lock) -> register_netdevice (acquires netdev lock) => devlink lock -> rtnl lock -> netdev lock. But in the current recovery flow, the order is wrong: mlx5e_tx_err_cqe_work (acquires netdev lock) -> mlx5e_reporter_tx_err_cqe -> mlx5e_health_report -> devlink_health_report (acquires devlink lock => boom!) -> devlink_health_reporter_recover -> mlx5e_tx_reporter_recover -> mlx5e_tx_reporter_recover_from_ctx -> mlx5e_tx_reporter_err_cqe_recover The same pattern exists in: mlx5e_reporter_rx_timeout mlx5e_reporter_tx_ptpsq_unhealthy mlx5e_reporter_tx_timeout Fix these by moving the netdev_trylock calls from the work handlers lower in the call stack, in the respective recovery functions, where they are actually necessary.

In the Linux kernel, the following vulnerability has been resolved: power: supply: pf1550: Fix use-after-free in power_supply_changed() Using the `devm_` variant for requesting IRQ _before_ the `devm_` variant for allocating/registering the `power_supply` handle, means that the `power_supply` handle will be deallocated/unregistered _before_ the interrupt handler (since `devm_` naturally deallocates in reverse allocation order). This means that during removal, there is a race condition where an interrupt can fire just _after_ the `power_supply` handle has been freed, *but* just _before_ the corresponding unregistration of the IRQ handler has run. This will lead to the IRQ handler calling `power_supply_changed()` with a freed `power_supply` handle. Which usually crashes the system or otherwise silently corrupts the memory... Note that there is a similar situation which can also happen during `probe()`; the possibility of an interrupt firing _before_ registering the `power_supply` handle. This would then lead to the nasty situation of using the `power_supply` handle *uninitialized* in `power_supply_changed()`. Fix this racy use-after-free by making sure the IRQ is requested _after_ the registration of the `power_supply` handle.

In the Linux kernel, the following vulnerability has been resolved: xfrm: fix ip_rt_bug race in icmp_route_lookup reverse path icmp_route_lookup() performs multiple route lookups to find a suitable route for sending ICMP error messages, with special handling for XFRM (IPsec) policies. The lookup sequence is: 1. First, lookup output route for ICMP reply (dst = original src) 2. Pass through xfrm_lookup() for policy check 3. If blocked (-EPERM) or dst is not local, enter "reverse path" 4. In reverse path, call xfrm_decode_session_reverse() to get fl4_dec which reverses the original packet's flow (saddr<->daddr swapped) 5. If fl4_dec.saddr is local (we are the original destination), use __ip_route_output_key() for output route lookup 6. If fl4_dec.saddr is NOT local (we are a forwarding node), use ip_route_input() to simulate the reverse packet's input path 7. Finally, pass rt2 through xfrm_lookup() with XFRM_LOOKUP_ICMP flag The bug occurs in step 6: ip_route_input() is called with fl4_dec.daddr (original packet's source) as destination. If this address becomes local between the initial check and ip_route_input() call (e.g., due to concurrent "ip addr add"), ip_route_input() returns a LOCAL route with dst.output set to ip_rt_bug. This route is then used for ICMP output, causing dst_output() to call ip_rt_bug(), triggering a WARN_ON: ------------[ cut here ]------------ WARNING: net/ipv4/route.c:1275 at ip_rt_bug+0x21/0x30, CPU#1 Call Trace: <TASK> ip_push_pending_frames+0x202/0x240 icmp_push_reply+0x30d/0x430 __icmp_send+0x1149/0x24f0 ip_options_compile+0xa2/0xd0 ip_rcv_finish_core+0x829/0x1950 ip_rcv+0x2d7/0x420 __netif_receive_skb_one_core+0x185/0x1f0 netif_receive_skb+0x90/0x450 tun_get_user+0x3413/0x3fb0 tun_chr_write_iter+0xe4/0x220 ... Fix this by checking rt2->rt_type after ip_route_input(). If it's RTN_LOCAL, the route cannot be used for output, so treat it as an error. The reproducer requires kernel modification to widen the race window, making it unsuitable as a selftest. It is available at: https://gist.github.com/mrpre/eae853b72ac6a750f5d45d64ddac1e81

In the Linux kernel, the following vulnerability has been resolved: powerpc/eeh: fix recursive pci_lock_rescan_remove locking in EEH event handling The recent commit 1010b4c012b0 ("powerpc/eeh: Make EEH driver device hotplug safe") restructured the EEH driver to improve synchronization with the PCI hotplug layer. However, it inadvertently moved pci_lock_rescan_remove() outside its intended scope in eeh_handle_normal_event(), leading to broken PCI error reporting and improper EEH event triggering. Specifically, eeh_handle_normal_event() acquired pci_lock_rescan_remove() before calling eeh_pe_bus_get(), but eeh_pe_bus_get() itself attempts to acquire the same lock internally, causing nested locking and disrupting normal EEH event handling paths. This patch adds a boolean parameter do_lock to _eeh_pe_bus_get(), with two public wrappers: eeh_pe_bus_get() with locking enabled. eeh_pe_bus_get_nolock() that skips locking. Callers that already hold pci_lock_rescan_remove() now use eeh_pe_bus_get_nolock() to avoid recursive lock acquisition. Additionally, pci_lock_rescan_remove() calls are restored to the correct position—after eeh_pe_bus_get() and immediately before iterating affected PEs and devices. This ensures EEH-triggered PCI removes occur under proper bus rescan locking without recursive lock contention. The eeh_pe_loc_get() function has been split into two functions: eeh_pe_loc_get(struct eeh_pe *pe) which retrieves the loc for given PE. eeh_pe_loc_get_bus(struct pci_bus *bus) which retrieves the location code for given bus. This resolves lockdep warnings such as: <snip> [ 84.964298] [ T928] ============================================ [ 84.964304] [ T928] WARNING: possible recursive locking detected [ 84.964311] [ T928] 6.18.0-rc3 #51 Not tainted [ 84.964315] [ T928] -------------------------------------------- [ 84.964320] [ T928] eehd/928 is trying to acquire lock: [ 84.964324] [ T928] c000000003b29d58 (pci_rescan_remove_lock){+.+.}-{3:3}, at: pci_lock_rescan_remove+0x28/0x40 [ 84.964342] [ T928] but task is already holding lock: [ 84.964347] [ T928] c000000003b29d58 (pci_rescan_remove_lock){+.+.}-{3:3}, at: pci_lock_rescan_remove+0x28/0x40 [ 84.964357] [ T928] other info that might help us debug this: [ 84.964363] [ T928] Possible unsafe locking scenario: [ 84.964367] [ T928] CPU0 [ 84.964370] [ T928] ---- [ 84.964373] [ T928] lock(pci_rescan_remove_lock); [ 84.964378] [ T928] lock(pci_rescan_remove_lock); [ 84.964383] [ T928] *** DEADLOCK *** [ 84.964388] [ T928] May be due to missing lock nesting notation [ 84.964393] [ T928] 1 lock held by eehd/928: [ 84.964397] [ T928] #0: c000000003b29d58 (pci_rescan_remove_lock){+.+.}-{3:3}, at: pci_lock_rescan_remove+0x28/0x40 [ 84.964408] [ T928] stack backtrace: [ 84.964414] [ T928] CPU: 2 UID: 0 PID: 928 Comm: eehd Not tainted 6.18.0-rc3 #51 VOLUNTARY [ 84.964417] [ T928] Hardware name: IBM,9080-HEX POWER10 (architected) 0x800200 0xf000006 of:IBM,FW1060.00 (NH1060_022) hv:phyp pSeries [ 84.964419] [ T928] Call Trace: [ 84.964420] [ T928] [c0000011a7157990] [c000000001705de4] dump_stack_lvl+0xc8/0x130 (unreliable) [ 84.964424] [ T928] [c0000011a71579d0] [c0000000002f66e0] print_deadlock_bug+0x430/0x440 [ 84.964428] [ T928] [c0000011a7157a70] [c0000000002fd0c0] __lock_acquire+0x1530/0x2d80 [ 84.964431] [ T928] [c0000011a7157ba0] [c0000000002fea54] lock_acquire+0x144/0x410 [ 84.964433] [ T928] [c0000011a7157cb0] [c0000011a7157cb0] __mutex_lock+0xf4/0x1050 [ 84.964436] [ T928] [c0000011a7157e00] [c000000000de21d8] pci_lock_rescan_remove+0x28/0x40 [ 84.964439] [ T928] [c0000011a7157e20] [c00000000004ed98] eeh_pe_bus_get+0x48/0xc0 [ 84.964442] [ T928] [c0000011a7157e50] [c00000 ---truncated---

In the Linux kernel, the following vulnerability has been resolved: bpf: Fix memory access flags in helper prototypes After commit 37cce22dbd51 ("bpf: verifier: Refactor helper access type tracking"), the verifier started relying on the access type flags in helper function prototypes to perform memory access optimizations. Currently, several helper functions utilizing ARG_PTR_TO_MEM lack the corresponding MEM_RDONLY or MEM_WRITE flags. This omission causes the verifier to incorrectly assume that the buffer contents are unchanged across the helper call. Consequently, the verifier may optimize away subsequent reads based on this wrong assumption, leading to correctness issues. For bpf_get_stack_proto_raw_tp, the original MEM_RDONLY was incorrect since the helper writes to the buffer. Change it to ARG_PTR_TO_UNINIT_MEM which correctly indicates write access to potentially uninitialized memory. Similar issues were recently addressed for specific helpers in commit ac44dcc788b9 ("bpf: Fix verifier assumptions of bpf_d_path's output buffer") and commit 2eb7648558a7 ("bpf: Specify access type of bpf_sysctl_get_name args"). Fix these prototypes by adding the correct memory access flags.

In the Linux kernel, the following vulnerability has been resolved: power: supply: bq256xx: Fix use-after-free in power_supply_changed() Using the `devm_` variant for requesting IRQ _before_ the `devm_` variant for allocating/registering the `power_supply` handle, means that the `power_supply` handle will be deallocated/unregistered _before_ the interrupt handler (since `devm_` naturally deallocates in reverse allocation order). This means that during removal, there is a race condition where an interrupt can fire just _after_ the `power_supply` handle has been freed, *but* just _before_ the corresponding unregistration of the IRQ handler has run. This will lead to the IRQ handler calling `power_supply_changed()` with a freed `power_supply` handle. Which usually crashes the system or otherwise silently corrupts the memory... Note that there is a similar situation which can also happen during `probe()`; the possibility of an interrupt firing _before_ registering the `power_supply` handle. This would then lead to the nasty situation of using the `power_supply` handle *uninitialized* in `power_supply_changed()`. Fix this racy use-after-free by making sure the IRQ is requested _after_ the registration of the `power_supply` handle.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nf_tables: revert commit_mutex usage in reset path It causes circular lock dependency between commit_mutex, nfnl_subsys_ipset and nlk_cb_mutex when nft reset, ipset list, and iptables-nft with '-m set' rule run at the same time. Previous patches made it safe to run individual reset handlers concurrently so commit_mutex is no longer required to prevent this.

In the Linux kernel, the following vulnerability has been resolved: crypto: caam - fix netdev memory leak in dpaa2_caam_probe When commit 0e1a4d427f58 ("crypto: caam: Unembed net_dev structure in dpaa2") converted embedded net_device to dynamically allocated pointers, it added cleanup in dpaa2_dpseci_disable() but missed adding cleanup in dpaa2_dpseci_free() for error paths. This causes memory leaks when dpaa2_dpseci_dpio_setup() fails during probe due to DPIO devices not being ready yet. The kernel's deferred probe mechanism handles the retry successfully, but the netdevs allocated during the failed probe attempt are never freed, resulting in kmemleak reports showing multiple leaked netdev-related allocations all traced back to dpaa2_caam_probe(). Fix this by preserving the CPU mask of allocated netdevs during setup and using it for cleanup in dpaa2_dpseci_free(). This approach ensures that only the CPUs that actually had netdevs allocated will be cleaned up, avoiding potential issues with CPU hotplug scenarios.

In the Linux kernel, the following vulnerability has been resolved: ext4: drop extent cache when splitting extent fails When the split extent fails, we might leave some extents still being processed and return an error directly, which will result in stale extent entries remaining in the extent status tree. So drop all of the remaining potentially stale extents if the splitting fails.

In the Linux kernel, the following vulnerability has been resolved: RDMA/iwcm: Fix workqueue list corruption by removing work_list The commit e1168f0 ("RDMA/iwcm: Simplify cm_event_handler()") changed the work submission logic to unconditionally call queue_work() with the expectation that queue_work() would have no effect if work was already pending. The problem is that a free list of struct iwcm_work is used (for which struct work_struct is embedded), so each call to queue_work() is basically unique and therefore does indeed queue the work. This causes a problem in the work handler which walks the work_list until it's empty to process entries. This means that a single run of the work handler could process item N+1 and release it back to the free list while the actual workqueue entry is still queued. It could then get reused (INIT_WORK...) and lead to list corruption in the workqueue logic. Fix this by just removing the work_list. The workqueue already does this for us. This fixes the following error that was observed when stress testing with ucmatose on an Intel E830 in iWARP mode: [ 151.465780] list_del corruption. next->prev should be ffff9f0915c69c08, but was ffff9f0a1116be08. (next=ffff9f0a15b11c08) [ 151.466639] ------------[ cut here ]------------ [ 151.466986] kernel BUG at lib/list_debug.c:67! [ 151.467349] Oops: invalid opcode: 0000 [#1] SMP NOPTI [ 151.467753] CPU: 14 UID: 0 PID: 2306 Comm: kworker/u64:18 Not tainted 6.19.0-rc4+ #1 PREEMPT(voluntary) [ 151.468466] Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [ 151.469192] Workqueue: 0x0 (iw_cm_wq) [ 151.469478] RIP: 0010:__list_del_entry_valid_or_report+0xf0/0x100 [ 151.469942] Code: c7 58 5f 4c b2 e8 10 50 aa ff 0f 0b 48 89 ef e8 36 57 cb ff 48 8b 55 08 48 89 e9 48 89 de 48 c7 c7 a8 5f 4c b2 e8 f0 4f aa ff <0f> 0b 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 90 90 90 90 90 90 [ 151.471323] RSP: 0000:ffffb15644e7bd68 EFLAGS: 00010046 [ 151.471712] RAX: 000000000000006d RBX: ffff9f0915c69c08 RCX: 0000000000000027 [ 151.472243] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff9f0a37d9c600 [ 151.472768] RBP: ffff9f0a15b11c08 R08: 0000000000000000 R09: c0000000ffff7fff [ 151.473294] R10: 0000000000000001 R11: ffffb15644e7bba8 R12: ffff9f092339ee68 [ 151.473817] R13: ffff9f0900059c28 R14: ffff9f092339ee78 R15: 0000000000000000 [ 151.474344] FS: 0000000000000000(0000) GS:ffff9f0a847b5000(0000) knlGS:0000000000000000 [ 151.474934] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 151.475362] CR2: 0000559e233a9088 CR3: 000000020296b004 CR4: 0000000000770ef0 [ 151.475895] PKRU: 55555554 [ 151.476118] Call Trace: [ 151.476331] <TASK> [ 151.476497] move_linked_works+0x49/0xa0 [ 151.476792] __pwq_activate_work.isra.46+0x2f/0xa0 [ 151.477151] pwq_dec_nr_in_flight+0x1e0/0x2f0 [ 151.477479] process_scheduled_works+0x1c8/0x410 [ 151.477823] worker_thread+0x125/0x260 [ 151.478108] ? __pfx_worker_thread+0x10/0x10 [ 151.478430] kthread+0xfe/0x240 [ 151.478671] ? __pfx_kthread+0x10/0x10 [ 151.478955] ? __pfx_kthread+0x10/0x10 [ 151.479240] ret_from_fork+0x208/0x270 [ 151.479523] ? __pfx_kthread+0x10/0x10 [ 151.479806] ret_from_fork_asm+0x1a/0x30 [ 151.480103] </TASK>

In the Linux kernel, the following vulnerability has been resolved: netfilter: nft_counter: serialize reset with spinlock Add a global static spinlock to serialize counter fetch+reset operations, preventing concurrent dump-and-reset from underrunning values. The lock is taken before fetching the total so that two parallel resets cannot both read the same counter values and then both subtract them. A global lock is used for simplicity since resets are infrequent. If this becomes a bottleneck, it can be replaced with a per-net lock later.

In the Linux kernel, the following vulnerability has been resolved: mtd: intel-dg: Fix accessing regions before setting nregions The regions array is counted by nregions, but it's set only after accessing it: [] UBSAN: array-index-out-of-bounds in drivers/mtd/devices/mtd_intel_dg.c:750:15 [] index 0 is out of range for type '<unknown> [*]' Fix it by also fixing an undesired behavior: the loop silently ignores ENOMEM and continues setting the other entries.

In the Linux kernel, the following vulnerability has been resolved: quota: fix livelock between quotactl and freeze_super When a filesystem is frozen, quotactl_block() enters a retry loop waiting for the filesystem to thaw. It acquires s_umount, checks the freeze state, drops s_umount and uses sb_start_write() - sb_end_write() pair to wait for the unfreeze. However, this retry loop can trigger a livelock issue, specifically on kernels with preemption disabled. The mechanism is as follows: 1. freeze_super() sets SB_FREEZE_WRITE and calls sb_wait_write(). 2. sb_wait_write() calls percpu_down_write(), which initiates synchronize_rcu(). 3. Simultaneously, quotactl_block() spins in its retry loop, immediately executing the sb_start_write() - sb_end_write() pair. 4. Because the kernel is non-preemptible and the loop contains no scheduling points, quotactl_block() never yields the CPU. This prevents that CPU from reaching an RCU quiescent state. 5. synchronize_rcu() in the freezer thread waits indefinitely for the quotactl_block() CPU to report a quiescent state. 6. quotactl_block() spins indefinitely waiting for the freezer to advance, which it cannot do as it is blocked on the RCU sync. This results in a hang of the freezer process and 100% CPU usage by the quota process. While this can occur intermittently on multi-core systems, it is reliably reproducing on a node with the following script, running both the freezer and the quota toggle on the same CPU: # mkfs.ext4 -O quota /dev/sda 2g && mkdir a_mount # mount /dev/sda -o quota,usrquota,grpquota a_mount # taskset -c 3 bash -c "while true; do xfs_freeze -f a_mount; \ xfs_freeze -u a_mount; done" & # taskset -c 3 bash -c "while true; do quotaon a_mount; \ quotaoff a_mount; done" & Adding cond_resched() to the retry loop fixes the issue. It acts as an RCU quiescent state, allowing synchronize_rcu() in percpu_down_write() to complete.

In the Linux kernel, the following vulnerability has been resolved: iommu/vt-d: Clear Present bit before tearing down PASID entry The Intel VT-d Scalable Mode PASID table entry consists of 512 bits (64 bytes). When tearing down an entry, the current implementation zeros the entire 64-byte structure immediately using multiple 64-bit writes. Since the IOMMU hardware may fetch these 64 bytes using multiple internal transactions (e.g., four 128-bit bursts), updating or zeroing the entire entry while it is active (P=1) risks a "torn" read. If a hardware fetch occurs simultaneously with the CPU zeroing the entry, the hardware could observe an inconsistent state, leading to unpredictable behavior or spurious faults. Follow the "Guidance to Software for Invalidations" in the VT-d spec (Section 6.5.3.3) by implementing the recommended ownership handshake: 1. Clear only the 'Present' (P) bit of the PASID entry. 2. Use a dma_wmb() to ensure the cleared bit is visible to hardware before proceeding. 3. Execute the required invalidation sequence (PASID cache, IOTLB, and Device-TLB flush) to ensure the hardware has released all cached references. 4. Only after the flushes are complete, zero out the remaining fields of the PASID entry. Also, add a dma_wmb() in pasid_set_present() to ensure that all other fields of the PASID entry are visible to the hardware before the Present bit is set.

In the Linux kernel, the following vulnerability has been resolved: apparmor: Fix & Optimize table creation from possibly unaligned memory Source blob may come from userspace and might be unaligned. Try to optize the copying process by avoiding unaligned memory accesses. - Added Fixes tag - Added "Fix &" to description as this doesn't just optimize but fixes a potential unaligned memory access [jj: remove duplicate word "convert" in comment trigger checkpatch warning]

In the Linux kernel, the following vulnerability has been resolved: ext4: drop extent cache after doing PARTIAL_VALID1 zeroout When splitting an unwritten extent in the middle and converting it to initialized in ext4_split_extent() with the EXT4_EXT_MAY_ZEROOUT and EXT4_EXT_DATA_VALID2 flags set, it could leave a stale unwritten extent. Assume we have an unwritten file and buffered write in the middle of it without dioread_nolock enabled, it will allocate blocks as written extent. 0 A B N [UUUUUUUUUUUU] on-disk extent U: unwritten extent [UUUUUUUUUUUU] extent status tree [--DDDDDDDD--] D: valid data |<- ->| ----> this range needs to be initialized ext4_split_extent() first try to split this extent at B with EXT4_EXT_DATA_PARTIAL_VALID1 and EXT4_EXT_MAY_ZEROOUT flag set, but ext4_split_extent_at() failed to split this extent due to temporary lack of space. It zeroout B to N and leave the entire extent as unwritten. 0 A B N [UUUUUUUUUUUU] on-disk extent [UUUUUUUUUUUU] extent status tree [--DDDDDDDDZZ] Z: zeroed data ext4_split_extent() then try to split this extent at A with EXT4_EXT_DATA_VALID2 flag set. This time, it split successfully and leave an written extent from A to N. 0 A B N [UUWWWWWWWWWW] on-disk extent W: written extent [UUUUUUUUUUUU] extent status tree [--DDDDDDDDZZ] Finally ext4_map_create_blocks() only insert extent A to B to the extent status tree, and leave an stale unwritten extent in the status tree. 0 A B N [UUWWWWWWWWWW] on-disk extent W: written extent [UUWWWWWWWWUU] extent status tree [--DDDDDDDDZZ] Fix this issue by always cached extent status entry after zeroing out the second part.

In the Linux kernel, the following vulnerability has been resolved: net: hns3: fix double free issue for tx spare buffer In hns3_set_ringparam(), a temporary copy (tmp_rings) of the ring structure is created for rollback. However, the tx_spare pointer in the original ring handle is incorrectly left pointing to the old backup memory. Later, if memory allocation fails in hns3_init_all_ring() during the setup, the error path attempts to free all newly allocated rings. Since tx_spare contains a stale (non-NULL) pointer from the backup, it is mistaken for a newly allocated buffer and is erroneously freed, leading to a double-free of the backup memory. The root cause is that the tx_spare field was not cleared after its value was saved in tmp_rings, leaving a dangling pointer. Fix this by setting tx_spare to NULL in the original ring structure when the creation of the new `tx_spare` fails. This ensures the error cleanup path only frees genuinely newly allocated buffers.

In the Linux kernel, the following vulnerability has been resolved: xen-netback: reject zero-queue configuration from guest A malicious or buggy Xen guest can write "0" to the xenbus key "multi-queue-num-queues". The connect() function in the backend only validates the upper bound (requested_num_queues > xenvif_max_queues) but not zero, allowing requested_num_queues=0 to reach vzalloc(array_size(0, sizeof(struct xenvif_queue))), which triggers WARN_ON_ONCE(!size) in __vmalloc_node_range(). On systems with panic_on_warn=1, this allows a guest-to-host denial of service. The Xen network interface specification requires the queue count to be "greater than zero". Add a zero check to match the validation already present in xen-blkback, which has included this guard since its multi-queue support was added.

In the Linux kernel, the following vulnerability has been resolved: mptcp: do not account for OoO in mptcp_rcvbuf_grow() MPTCP-level OoOs are physiological when multiple subflows are active concurrently and will not cause retransmissions nor are caused by drops. Accounting for them in mptcp_rcvbuf_grow() causes the rcvbuf slowly drifting towards tcp_rmem[2]. Remove such accounting. Note that subflows will still account for TCP-level OoO when the MPTCP-level rcvbuf is propagated. This also closes a subtle and very unlikely race condition with rcvspace init; active sockets with user-space holding the msk-level socket lock, could complete such initialization in the receive callback, after that the first OoO data reaches the rcvbuf and potentially triggering a divide by zero Oops.

In the Linux kernel, the following vulnerability has been resolved: md/raid1: fix memory leak in raid1_run() raid1_run() calls setup_conf() which registers a thread via md_register_thread(). If raid1_set_limits() fails, the previously registered thread is not unregistered, resulting in a memory leak of the md_thread structure and the thread resource itself. Add md_unregister_thread() to the error path to properly cleanup the thread, which aligns with the error handling logic of other paths in this function. Compile tested only. Issue found using a prototype static analysis tool and code review.

In the Linux kernel, the following vulnerability has been resolved: af_unix: Fix memleak of newsk in unix_stream_connect(). When prepare_peercred() fails in unix_stream_connect(), unix_release_sock() is not called for newsk, and the memory is leaked. Let's move prepare_peercred() before unix_create1().

In the Linux kernel, the following vulnerability has been resolved: bpf: Fix bpf_xdp_store_bytes proto for read-only arg While making some maps in Cilium read-only from the BPF side, we noticed that the bpf_xdp_store_bytes proto is incorrect. In particular, the verifier was throwing the following error: ; ret = ctx_store_bytes(ctx, l3_off + offsetof(struct iphdr, saddr), &nat->address, 4, 0); 635: (79) r1 = *(u64 *)(r10 -144) ; R1=ctx() R10=fp0 fp-144=ctx() 636: (b4) w2 = 26 ; R2=26 637: (b4) w4 = 4 ; R4=4 638: (b4) w5 = 0 ; R5=0 639: (85) call bpf_xdp_store_bytes#190 write into map forbidden, value_size=6 off=0 size=4 nat comes from a BPF_F_RDONLY_PROG map, so R3 is a PTR_TO_MAP_VALUE. The verifier checks the helper's memory access to R3 in check_mem_size_reg, as it reaches ARG_CONST_SIZE argument. The third argument has expected type ARG_PTR_TO_UNINIT_MEM, which includes the MEM_WRITE flag. The verifier thus checks for a BPF_WRITE access on R3. Given R3 points to a read-only map, the check fails. Conversely, ARG_PTR_TO_UNINIT_MEM can also lead to the helper reading from uninitialized memory. This patch simply fixes the expected argument type to match that of bpf_skb_store_bytes.

In the Linux kernel, the following vulnerability has been resolved: power: supply: cpcap-battery: Fix use-after-free in power_supply_changed() Using the `devm_` variant for requesting IRQ _before_ the `devm_` variant for allocating/registering the `power_supply` handle, means that the `power_supply` handle will be deallocated/unregistered _before_ the interrupt handler (since `devm_` naturally deallocates in reverse allocation order). This means that during removal, there is a race condition where an interrupt can fire just _after_ the `power_supply` handle has been freed, *but* just _before_ the corresponding unregistration of the IRQ handler has run. This will lead to the IRQ handler calling `power_supply_changed()` with a freed `power_supply` handle. Which usually crashes the system or otherwise silently corrupts the memory... Note that there is a similar situation which can also happen during `probe()`; the possibility of an interrupt firing _before_ registering the `power_supply` handle. This would then lead to the nasty situation of using the `power_supply` handle *uninitialized* in `power_supply_changed()`. Fix this racy use-after-free by making sure the IRQ is requested _after_ the registration of the `power_supply` handle.

In the Linux kernel, the following vulnerability has been resolved: apparmor: avoid per-cpu hold underflow in aa_get_buffer When aa_get_buffer() pulls from the per-cpu list it unconditionally decrements cache->hold. If hold reaches 0 while count is still non-zero, the unsigned decrement wraps to UINT_MAX. This keeps hold non-zero for a very long time, so aa_put_buffer() never returns buffers to the global list, which can starve other CPUs and force repeated kmalloc(aa_g_path_max) allocations. Guard the decrement so hold never underflows.

In the Linux kernel, the following vulnerability has been resolved: iio: sca3000: Fix a resource leak in sca3000_probe() spi->irq from request_threaded_irq() not released when iio_device_register() fails. Add an return value check and jump to a common error handler when iio_device_register() fails.

In the Linux kernel, the following vulnerability has been resolved: power: supply: pm8916_bms_vm: Fix use-after-free in power_supply_changed() Using the `devm_` variant for requesting IRQ _before_ the `devm_` variant for allocating/registering the `power_supply` handle, means that the `power_supply` handle will be deallocated/unregistered _before_ the interrupt handler (since `devm_` naturally deallocates in reverse allocation order). This means that during removal, there is a race condition where an interrupt can fire just _after_ the `power_supply` handle has been freed, *but* just _before_ the corresponding unregistration of the IRQ handler has run. This will lead to the IRQ handler calling `power_supply_changed()` with a freed `power_supply` handle. Which usually crashes the system or otherwise silently corrupts the memory... Note that there is a similar situation which can also happen during `probe()`; the possibility of an interrupt firing _before_ registering the `power_supply` handle. This would then lead to the nasty situation of using the `power_supply` handle *uninitialized* in `power_supply_changed()`. Fix this racy use-after-free by making sure the IRQ is requested _after_ the registration of the `power_supply` handle.

In the Linux kernel, the following vulnerability has been resolved: soc: mediatek: svs: Fix memory leak in svs_enable_debug_write() In svs_enable_debug_write(), the buf allocated by memdup_user_nul() is leaked if kstrtoint() fails. Fix this by using __free(kfree) to automatically free buf, eliminating the need for explicit kfree() calls and preventing leaks. [Angelo: Added missing cleanup.h inclusion]

In the Linux kernel, the following vulnerability has been resolved: PCI/P2PDMA: Release per-CPU pgmap ref when vm_insert_page() fails When vm_insert_page() fails in p2pmem_alloc_mmap(), p2pmem_alloc_mmap() doesn't invoke percpu_ref_put() to free the per-CPU ref of pgmap acquired after gen_pool_alloc_owner(), and memunmap_pages() will hang forever when trying to remove the PCI device. Fix it by adding the missed percpu_ref_put().

In the Linux kernel, the following vulnerability has been resolved: power: supply: bq25980: Fix use-after-free in power_supply_changed() Using the `devm_` variant for requesting IRQ _before_ the `devm_` variant for allocating/registering the `power_supply` handle, means that the `power_supply` handle will be deallocated/unregistered _before_ the interrupt handler (since `devm_` naturally deallocates in reverse allocation order). This means that during removal, there is a race condition where an interrupt can fire just _after_ the `power_supply` handle has been freed, *but* just _before_ the corresponding unregistration of the IRQ handler has run. This will lead to the IRQ handler calling `power_supply_changed()` with a freed `power_supply` handle. Which usually crashes the system or otherwise silently corrupts the memory... Note that there is a similar situation which can also happen during `probe()`; the possibility of an interrupt firing _before_ registering the `power_supply` handle. This would then lead to the nasty situation of using the `power_supply` handle *uninitialized* in `power_supply_changed()`. Fix this racy use-after-free by making sure the IRQ is requested _after_ the registration of the `power_supply` handle.

In the Linux kernel, the following vulnerability has been resolved: drm/amdkfd: Fix watch_id bounds checking in debug address watch v2 The address watch clear code receives watch_id as an unsigned value (u32), but some helper functions were using a signed int and checked bits by shifting with watch_id. If a very large watch_id is passed from userspace, it can be converted to a negative value. This can cause invalid shifts and may access memory outside the watch_points array. drm/amdkfd: Fix watch_id bounds checking in debug address watch v2 Fix this by checking that watch_id is within MAX_WATCH_ADDRESSES before using it. Also use BIT(watch_id) to test and clear bits safely. This keeps the behavior unchanged for valid watch IDs and avoids undefined behavior for invalid ones. Fixes the below: drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_debug.c:448 kfd_dbg_trap_clear_dev_address_watch() error: buffer overflow 'pdd->watch_points' 4 <= u32max user_rl='0-3,2147483648-u32max' uncapped drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_debug.c 433 int kfd_dbg_trap_clear_dev_address_watch(struct kfd_process_device *pdd, 434 uint32_t watch_id) 435 { 436 int r; 437 438 if (!kfd_dbg_owns_dev_watch_id(pdd, watch_id)) kfd_dbg_owns_dev_watch_id() doesn't check for negative values so if watch_id is larger than INT_MAX it leads to a buffer overflow. (Negative shifts are undefined). 439 return -EINVAL; 440 441 if (!pdd->dev->kfd->shared_resources.enable_mes) { 442 r = debug_lock_and_unmap(pdd->dev->dqm); 443 if (r) 444 return r; 445 } 446 447 amdgpu_gfx_off_ctrl(pdd->dev->adev, false); --> 448 pdd->watch_points[watch_id] = pdd->dev->kfd2kgd->clear_address_watch( 449 pdd->dev->adev, 450 watch_id); v2: (as per, Jonathan Kim) - Add early watch_id >= MAX_WATCH_ADDRESSES validation in the set path to match the clear path. - Drop the redundant bounds check in kfd_dbg_owns_dev_watch_id().

In the Linux kernel, the following vulnerability has been resolved: HID: intel-ish-hid: fix NULL-ptr-deref in ishtp_bus_remove_all_clients During a warm reset flow, the cl->device pointer may be NULL if the reset occurs while clients are still being enumerated. Accessing cl->device->reference_count without a NULL check leads to a kernel panic. This issue was identified during multi-unit warm reboot stress clycles. Add a defensive NULL check for cl->device to ensure stability under such intensive testing conditions. KASAN: null-ptr-deref in range [0000000000000000-0000000000000007] Workqueue: ish_fw_update_wq fw_reset_work_fn Call Trace: ishtp_bus_remove_all_clients+0xbe/0x130 [intel_ishtp] ishtp_reset_handler+0x85/0x1a0 [intel_ishtp] fw_reset_work_fn+0x8a/0xc0 [intel_ish_ipc]

In the Linux kernel, the following vulnerability has been resolved: arm64/gcs: Fix error handling in arch_set_shadow_stack_status() alloc_gcs() returns an error-encoded pointer on failure, which comes from do_mmap(), not NULL. The current NULL check fails to detect errors, which could lead to using an invalid GCS address. Use IS_ERR_VALUE() to properly detect errors, consistent with the check in gcs_alloc_thread_stack().

In the Linux kernel, the following vulnerability has been resolved: mfd: arizona: Fix regulator resource leak on wm5102_clear_write_sequencer() failure The wm5102_clear_write_sequencer() helper may return an error and just return, bypassing the cleanup sequence and causing regulators to remain enabled, leading to a resource leak. Change the direct return to jump to the err_reset label to properly free the resources.

In the Linux kernel, the following vulnerability has been resolved: phy: freescale: imx8qm-hsio: fix NULL pointer dereference During the probe the refclk_pad pointer is set to NULL if the 'fsl,refclk-pad-mode' property is not defined in the devicetree node. But in imx_hsio_configure_clk_pad() this pointer is unconditionally used which could result in a NULL pointer dereference. So check the pointer before to use it.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nft_set_rbtree: check for partial overlaps in anonymous sets Userspace provides an optimized representation in case intervals are adjacent, where the end element is omitted. The existing partial overlap detection logic skips anonymous set checks on start elements for this reason. However, it is possible to add intervals that overlap to this anonymous where two start elements with the same, eg. A-B, A-C where C < B. start end A B start end A C Restore the check on overlapping start elements to report an overlap.

In the Linux kernel, the following vulnerability has been resolved: scsi: smartpqi: Fix memory leak in pqi_report_phys_luns() pqi_report_phys_luns() fails to release the rpl_list buffer when encountering an unsupported data format or when the allocation for rpl_16byte_wwid_list fails. These early returns bypass the cleanup logic, leading to memory leaks. Consolidate the error handling by adding an out_free_rpl_list label and use goto statements to ensure rpl_list is consistently freed on failure. Compile tested only. Issue found using a prototype static analysis tool and code review.

In the Linux kernel, the following vulnerability has been resolved: tpm: st33zp24: Fix missing cleanup on get_burstcount() error get_burstcount() can return -EBUSY on timeout. When this happens, st33zp24_send() returns directly without releasing the locality acquired earlier. Use goto out_err to ensure proper cleanup when get_burstcount() fails.

In the Linux kernel, the following vulnerability has been resolved: SUNRPC: auth_gss: fix memory leaks in XDR decoding error paths The gssx_dec_ctx(), gssx_dec_status(), and gssx_dec_name() functions allocate memory via gssx_dec_buffer(), which calls kmemdup(). When a subsequent decode operation fails, these functions return immediately without freeing previously allocated buffers, causing memory leaks. The leak in gssx_dec_ctx() is particularly relevant because the caller (gssp_accept_sec_context_upcall) initializes several buffer length fields to non-zero values, resulting in memory allocation: struct gssx_ctx rctxh = { .exported_context_token.len = GSSX_max_output_handle_sz, .mech.len = GSS_OID_MAX_LEN, .src_name.display_name.len = GSSX_max_princ_sz, .targ_name.display_name.len = GSSX_max_princ_sz }; If, for example, gssx_dec_name() succeeds for src_name but fails for targ_name, the memory allocated for exported_context_token, mech, and src_name.display_name remains unreferenced and cannot be reclaimed. Add error handling with goto-based cleanup to free any previously allocated buffers before returning an error.

In the Linux kernel, the following vulnerability has been resolved: power: supply: wm97xx: Fix NULL pointer dereference in power_supply_changed() In `probe()`, `request_irq()` is called before allocating/registering a `power_supply` handle. If an interrupt is fired between the call to `request_irq()` and `power_supply_register()`, the `power_supply` handle will be used uninitialized in `power_supply_changed()` in `wm97xx_bat_update()` (triggered from the interrupt handler). This will lead to a `NULL` pointer dereference since Fix this racy `NULL` pointer dereference by making sure the IRQ is requested _after_ the registration of the `power_supply` handle. Since the IRQ is the last thing requests in the `probe()` now, remove the error path for freeing it. Instead add one for unregistering the `power_supply` handle when IRQ request fails.

In the Linux kernel, the following vulnerability has been resolved: pinctrl: single: fix refcount leak in pcs_add_gpio_func() of_parse_phandle_with_args() returns a device_node pointer with refcount incremented in gpiospec.np. The loop iterates through all phandles but never releases the reference, causing a refcount leak on each iteration. Add of_node_put() calls to release the reference after extracting the needed arguments and on the error path when devm_kzalloc() fails. This bug was detected by our static analysis tool and verified by my code review.

In the Linux kernel, the following vulnerability has been resolved: power: supply: act8945a: Fix use-after-free in power_supply_changed() Using the `devm_` variant for requesting IRQ _before_ the `devm_` variant for allocating/registering the `power_supply` handle, means that the `power_supply` handle will be deallocated/unregistered _before_ the interrupt handler (since `devm_` naturally deallocates in reverse allocation order). This means that during removal, there is a race condition where an interrupt can fire just _after_ the `power_supply` handle has been freed, *but* just _before_ the corresponding unregistration of the IRQ handler has run. This will lead to the IRQ handler calling `power_supply_changed()` with a freed `power_supply` handle. Which usually crashes the system or otherwise silently corrupts the memory... Note that there is a similar situation which can also happen during `probe()`; the possibility of an interrupt firing _before_ registering the `power_supply` handle. This would then lead to the nasty situation of using the `power_supply` handle *uninitialized* in `power_supply_changed()`. Fix this racy use-after-free by making sure the IRQ is requested _after_ the registration of the `power_supply` handle.

In the Linux kernel, the following vulnerability has been resolved: serial: caif: fix use-after-free in caif_serial ldisc_close() There is a use-after-free bug in caif_serial where handle_tx() may access ser->tty after the tty has been freed. The race condition occurs between ldisc_close() and packet transmission: CPU 0 (close) CPU 1 (xmit) ------------- ------------ ldisc_close() tty_kref_put(ser->tty) [tty may be freed here] <-- race window --> caif_xmit() handle_tx() tty = ser->tty // dangling ptr tty->ops->write() // UAF! schedule_work() ser_release() unregister_netdevice() The root cause is that tty_kref_put() is called in ldisc_close() while the network device is still active and can receive packets. Since ser and tty have a 1:1 binding relationship with consistent lifecycles (ser is allocated in ldisc_open and freed in ser_release via unregister_netdevice, and each ser binds exactly one tty), we can safely defer the tty reference release to ser_release() where the network device is unregistered. Fix this by moving tty_kref_put() from ldisc_close() to ser_release(), after unregister_netdevice(). This ensures the tty reference is held as long as the network device exists, preventing the UAF. Note: We save ser->tty before unregister_netdevice() because ser is embedded in netdev's private data and will be freed along with netdev (needs_free_netdev = true). How to reproduce: Add mdelay(500) at the beginning of ldisc_close() to widen the race window, then run the reproducer program [1]. Note: There is a separate deadloop issue in handle_tx() when using PORT_UNKNOWN serial ports (e.g., /dev/ttyS3 in QEMU without proper serial backend). This deadloop exists even without this patch, and is likely caused by inconsistency between uart_write_room() and uart_write() in serial core. It has been addressed in a separate patch [2]. KASAN report: ================================================================== BUG: KASAN: slab-use-after-free in handle_tx+0x5d1/0x620 Read of size 1 at addr ffff8881131e1490 by task caif_uaf_trigge/9929 Call Trace: <TASK> dump_stack_lvl+0x10e/0x1f0 print_report+0xd0/0x630 kasan_report+0xe4/0x120 handle_tx+0x5d1/0x620 dev_hard_start_xmit+0x9d/0x6c0 __dev_queue_xmit+0x6e2/0x4410 packet_xmit+0x243/0x360 packet_sendmsg+0x26cf/0x5500 __sys_sendto+0x4a3/0x520 __x64_sys_sendto+0xe0/0x1c0 do_syscall_64+0xc9/0xf80 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7f615df2c0d7 Allocated by task 9930: Freed by task 64: Last potentially related work creation: The buggy address belongs to the object at ffff8881131e1000 which belongs to the cache kmalloc-cg-2k of size 2048 The buggy address is located 1168 bytes inside of freed 2048-byte region [ffff8881131e1000, ffff8881131e1800) The buggy address belongs to the physical page: page_owner tracks the page as allocated page last free pid 9778 tgid 9778 stack trace: Memory state around the buggy address: ffff8881131e1380: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8881131e1400: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8881131e1480: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8881131e1500: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8881131e1580: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== [1]: https://gist.github.com/mrpre/f683f244544f7b11e7fa87df9e6c2eeb [2]: https://lore.kernel.org/linux-serial/20260204074327.226165-1-jiayuan.chen@linux.dev/T/#u

In the Linux kernel, the following vulnerability has been resolved: mctp i2c: initialise event handler read bytes Set a 0xff value for i2c reads of an mctp-i2c device. Otherwise reads will return "val" from the i2c bus driver. For i2c-aspeed and i2c-npcm7xx that is a stack uninitialised u8. Tested with "i2ctransfer -y 1 r10@0x34" where 0x34 is a mctp-i2c instance, now it returns all 0xff.

In the Linux kernel, the following vulnerability has been resolved: fs/ntfs3: prevent infinite loops caused by the next valid being the same When processing valid within the range [valid : pos), if valid cannot be retrieved correctly, for example, if the retrieved valid value is always the same, this can trigger a potential infinite loop, similar to the hung problem reported by syzbot [1]. Adding a check for the valid value within the loop body, and terminating the loop and returning -EINVAL if the value is the same as the current value, can prevent this. [1] INFO: task syz.4.21:6056 blocked for more than 143 seconds. Call Trace: rwbase_write_lock+0x14f/0x750 kernel/locking/rwbase_rt.c:244 inode_lock include/linux/fs.h:1027 [inline] ntfs_file_write_iter+0xe6/0x870 fs/ntfs3/file.c:1284

In the Linux kernel, the following vulnerability has been resolved: i3c: dw: Fix memory leak in dw_i3c_master_i2c_xfers() The dw_i3c_master_i2c_xfers() function allocates memory for the xfer structure using dw_i3c_master_alloc_xfer(). If pm_runtime_resume_and_get() fails, the function returns without freeing the allocated xfer, resulting in a memory leak. Add a dw_i3c_master_free_xfer() call to the error path to ensure the allocated memory is properly freed. Compile tested only. Issue found using a prototype static analysis tool and code review.

In the Linux kernel, the following vulnerability has been resolved: iommu/vt-d: Flush cache for PASID table before using it When writing the address of a freshly allocated zero-initialized PASID table to a PASID directory entry, do that after the CPU cache flush for this PASID table, not before it, to avoid the time window when this PASID table may be already used by non-coherent IOMMU hardware while its contents in RAM is still some random old data, not zero-initialized.

In the Linux kernel, the following vulnerability has been resolved: gfs2: Fix slab-use-after-free in qd_put Commit a475c5dd16e5 ("gfs2: Free quota data objects synchronously") started freeing quota data objects during filesystem shutdown instead of putting them back onto the LRU list, but it failed to remove these objects from the LRU list, causing LRU list corruption. This caused use-after-free when the shrinker (gfs2_qd_shrink_scan) tried to access already-freed objects on the LRU list. Fix this by removing qd objects from the LRU list before freeing them in qd_put(). Initial fix from Deepanshu Kartikey <kartikey406@gmail.com>.

In the Linux kernel, the following vulnerability has been resolved: netfilter: nf_conncount: increase the connection clean up limit to 64 After the optimization to only perform one GC per jiffy, a new problem was introduced. If more than 8 new connections are tracked per jiffy the list won't be cleaned up fast enough possibly reaching the limit wrongly. In order to prevent this issue, only skip the GC if it was already triggered during the same jiffy and the increment is lower than the clean up limit. In addition, increase the clean up limit to 64 connections to avoid triggering GC too often and do more effective GCs. This has been tested using a HTTP server and several performance tools while having nft_connlimit/xt_connlimit or OVS limit configured. Output of slowhttptest + OVS limit at 52000 connections: slow HTTP test status on 340th second: initializing: 0 pending: 432 connected: 51998 error: 0 closed: 0 service available: YES

In the Linux kernel, the following vulnerability has been resolved: netfilter: nfnetlink_queue: do shared-unconfirmed check before segmentation Ulrich reports a regression with nfqueue: If an application did not set the 'F_GSO' capability flag and a gso packet with an unconfirmed nf_conn entry is received all packets are now dropped instead of queued, because the check happens after skb_gso_segment(). In that case, we did have exclusive ownership of the skb and its associated conntrack entry. The elevated use count is due to skb_clone happening via skb_gso_segment(). Move the check so that its peformed vs. the aggregated packet. Then, annotate the individual segments except the first one so we can do a 2nd check at reinject time. For the normal case, where userspace does in-order reinjects, this avoids packet drops: first reinjected segment continues traversal and confirms entry, remaining segments observe the confirmed entry. While at it, simplify nf_ct_drop_unconfirmed(): We only care about unconfirmed entries with a refcnt > 1, there is no need to special-case dying entries. This only happens with UDP. With TCP, the only unconfirmed packet will be the TCP SYN, those aren't aggregated by GRO. Next patch adds a udpgro test case to cover this scenario.

In the Linux kernel, the following vulnerability has been resolved: ext4: don't zero the entire extent if EXT4_EXT_DATA_PARTIAL_VALID1 When allocating initialized blocks from a large unwritten extent, or when splitting an unwritten extent during end I/O and converting it to initialized, there is currently a potential issue of stale data if the extent needs to be split in the middle. 0 A B N [UUUUUUUUUUUU] U: unwritten extent [--DDDDDDDD--] D: valid data |<- ->| ----> this range needs to be initialized ext4_split_extent() first try to split this extent at B with EXT4_EXT_DATA_ENTIRE_VALID1 and EXT4_EXT_MAY_ZEROOUT flag set, but ext4_split_extent_at() failed to split this extent due to temporary lack of space. It zeroout B to N and mark the entire extent from 0 to N as written. 0 A B N [WWWWWWWWWWWW] W: written extent [SSDDDDDDDDZZ] Z: zeroed, S: stale data ext4_split_extent() then try to split this extent at A with EXT4_EXT_DATA_VALID2 flag set. This time, it split successfully and left a stale written extent from 0 to A. 0 A B N [WW|WWWWWWWWWW] [SS|DDDDDDDDZZ] Fix this by pass EXT4_EXT_DATA_PARTIAL_VALID1 to ext4_split_extent_at() when splitting at B, don't convert the entire extent to written and left it as unwritten after zeroing out B to N. The remaining work is just like the standard two-part split. ext4_split_extent() will pass the EXT4_EXT_DATA_VALID2 flag when it calls ext4_split_extent_at() for the second time, allowing it to properly handle the split. If the split is successful, it will keep extent from 0 to A as unwritten.