Vulnerabilities

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> net: mana: Fix double destroy_workqueue on service rescan PCI path<br /> <br /> While testing corner cases in the driver, a use-after-free crash<br /> was found on the service rescan PCI path.<br /> <br /> When mana_serv_reset() calls mana_gd_suspend(), mana_gd_cleanup()<br /> destroys gc-&gt;service_wq. If the subsequent mana_gd_resume() fails<br /> with -ETIMEDOUT or -EPROTO, the code falls through to<br /> mana_serv_rescan() which triggers pci_stop_and_remove_bus_device().<br /> This invokes the PCI .remove callback (mana_gd_remove), which calls<br /> mana_gd_cleanup() a second time, attempting to destroy the already-<br /> freed workqueue. Fix this by NULL-checking gc-&gt;service_wq in<br /> mana_gd_cleanup() and setting it to NULL after destruction.<br /> <br /> Call stack of issue for reference:<br /> [Sat Feb 21 18:53:48 2026] Call Trace:<br /> [Sat Feb 21 18:53:48 2026] <br /> [Sat Feb 21 18:53:48 2026] mana_gd_cleanup+0x33/0x70 [mana]<br /> [Sat Feb 21 18:53:48 2026] mana_gd_remove+0x3a/0xc0 [mana]<br /> [Sat Feb 21 18:53:48 2026] pci_device_remove+0x41/0xb0<br /> [Sat Feb 21 18:53:48 2026] device_remove+0x46/0x70<br /> [Sat Feb 21 18:53:48 2026] device_release_driver_internal+0x1e3/0x250<br /> [Sat Feb 21 18:53:48 2026] device_release_driver+0x12/0x20<br /> [Sat Feb 21 18:53:48 2026] pci_stop_bus_device+0x6a/0x90<br /> [Sat Feb 21 18:53:48 2026] pci_stop_and_remove_bus_device+0x13/0x30<br /> [Sat Feb 21 18:53:48 2026] mana_do_service+0x180/0x290 [mana]<br /> [Sat Feb 21 18:53:48 2026] mana_serv_func+0x24/0x50 [mana]<br /> [Sat Feb 21 18:53:48 2026] process_one_work+0x190/0x3d0<br /> [Sat Feb 21 18:53:48 2026] worker_thread+0x16e/0x2e0<br /> [Sat Feb 21 18:53:48 2026] kthread+0xf7/0x130<br /> [Sat Feb 21 18:53:48 2026] ? __pfx_worker_thread+0x10/0x10<br /> [Sat Feb 21 18:53:48 2026] ? __pfx_kthread+0x10/0x10<br /> [Sat Feb 21 18:53:48 2026] ret_from_fork+0x269/0x350<br /> [Sat Feb 21 18:53:48 2026] ? __pfx_kthread+0x10/0x10<br /> [Sat Feb 21 18:53:48 2026] ret_from_fork_asm+0x1a/0x30<br /> [Sat Feb 21 18:53:48 2026]

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> drm/atmel-hlcdc: fix memory leak from the atomic_destroy_state callback<br /> <br /> After several commits, the slab memory increases. Some drm_crtc_commit<br /> objects are not freed. The atomic_destroy_state callback only put the<br /> framebuffer. Use the __drm_atomic_helper_plane_destroy_state() function<br /> to put all the objects that are no longer needed.<br /> <br /> It has been seen after hours of usage of a graphics application or using<br /> kmemleak:<br /> <br /> unreferenced object 0xc63a6580 (size 64):<br /> comm "egt_basic", pid 171, jiffies 4294940784<br /> hex dump (first 32 bytes):<br /> 40 50 34 c5 01 00 00 00 ff ff ff ff 8c 65 3a c6 @P4..........e:.<br /> 8c 65 3a c6 ff ff ff ff 98 65 3a c6 98 65 3a c6 .e:......e:..e:.<br /> backtrace (crc c25aa925):<br /> kmemleak_alloc+0x34/0x3c<br /> __kmalloc_cache_noprof+0x150/0x1a4<br /> drm_atomic_helper_setup_commit+0x1e8/0x7bc<br /> drm_atomic_helper_commit+0x3c/0x15c<br /> drm_atomic_commit+0xc0/0xf4<br /> drm_atomic_helper_set_config+0x84/0xb8<br /> drm_mode_setcrtc+0x32c/0x810<br /> drm_ioctl+0x20c/0x488<br /> sys_ioctl+0x14c/0xc20<br /> ret_fast_syscall+0x0/0x54

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> md-cluster: fix NULL pointer dereference in process_metadata_update<br /> <br /> The function process_metadata_update() blindly dereferences the &amp;#39;thread&amp;#39;<br /> pointer (acquired via rcu_dereference_protected) within the wait_event()<br /> macro.<br /> <br /> While the code comment states "daemon thread must exist", there is a valid<br /> race condition window during the MD array startup sequence (md_run):<br /> <br /> 1. bitmap_load() is called, which invokes md_cluster_ops-&gt;join().<br /> 2. join() starts the "cluster_recv" thread (recv_daemon).<br /> 3. At this point, recv_daemon is active and processing messages.<br /> 4. However, mddev-&gt;thread (the main MD thread) is not initialized until<br /> later in md_run().<br /> <br /> If a METADATA_UPDATED message is received from a remote node during this<br /> specific window, process_metadata_update() will be called while<br /> mddev-&gt;thread is still NULL, leading to a kernel panic.<br /> <br /> To fix this, we must validate the &amp;#39;thread&amp;#39; pointer. If it is NULL, we<br /> release the held lock (no_new_dev_lockres) and return early, safely<br /> ignoring the update request as the array is not yet fully ready to<br /> process it.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> media: mtk-mdp: Fix a reference leak bug in mtk_mdp_remove()<br /> <br /> In mtk_mdp_probe(), vpu_get_plat_device() increases the reference<br /> count of the returned platform device. Add platform_device_put()<br /> to prevent reference leak.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> ring-buffer: Fix possible dereference of uninitialized pointer<br /> <br /> There is a pointer head_page in rb_meta_validate_events() which is not<br /> initialized at the beginning of a function. This pointer can be dereferenced<br /> if there is a failure during reader page validation. In this case the control<br /> is passed to "invalid" label where the pointer is dereferenced in a loop.<br /> <br /> To fix the issue initialize orig_head and head_page before calling<br /> rb_validate_buffer.<br /> <br /> Found by Linux Verification Center (linuxtesting.org) with SVACE.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> ceph: supply snapshot context in ceph_zero_partial_object()<br /> <br /> The ceph_zero_partial_object function was missing proper snapshot<br /> context for its OSD write operations, which could lead to data<br /> inconsistencies in snapshots.<br /> <br /> Reproducer:<br /> ../src/vstart.sh --new -x --localhost --bluestore<br /> ./bin/ceph auth caps client.fs_a mds &amp;#39;allow rwps fsname=a&amp;#39; mon &amp;#39;allow r fsname=a&amp;#39; osd &amp;#39;allow rw tag cephfs data=a&amp;#39;<br /> mount -t ceph fs_a@.a=/ /mnt/mycephfs/ -o conf=./ceph.conf<br /> dd if=/dev/urandom of=/mnt/mycephfs/foo bs=64K count=1<br /> mkdir /mnt/mycephfs/.snap/snap1<br /> md5sum /mnt/mycephfs/.snap/snap1/foo<br /> fallocate -p -o 0 -l 4096 /mnt/mycephfs/foo<br /> echo 3 &gt; /proc/sys/vm/drop/caches<br /> md5sum /mnt/mycephfs/.snap/snap1/foo # get different md5sum!!

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> media: chips-media: wave5: Fix Null reference while testing fluster<br /> <br /> When multi instances are created/destroyed, many interrupts happens<br /> and structures for decoder are removed.<br /> "struct vpu_instance" this structure is shared for all flow in the decoder,<br /> so if the structure is not protected by lock, Null dereference<br /> could happens sometimes.<br /> IRQ Handler was spilt to two phases and Lock was added as well.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> fbdev: of: display_timing: fix refcount leak in of_get_display_timings()<br /> <br /> of_parse_phandle() returns a device_node with refcount incremented,<br /> which is stored in &amp;#39;entry&amp;#39; and then copied to &amp;#39;native_mode&amp;#39;. When the<br /> error paths at lines 184 or 192 jump to &amp;#39;entryfail&amp;#39;, native_mode&amp;#39;s<br /> refcount is not decremented, causing a refcount leak.<br /> <br /> Fix this by changing the goto target from &amp;#39;entryfail&amp;#39; to &amp;#39;timingfail&amp;#39;,<br /> which properly calls of_node_put(native_mode) before cleanup.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> KVM: x86: Ignore -EBUSY when checking nested events from vcpu_block()<br /> <br /> Ignore -EBUSY when checking nested events after exiting a blocking state<br /> while L2 is active, as exiting to userspace will generate a spurious<br /> userspace exit, usually with KVM_EXIT_UNKNOWN, and likely lead to the VM&amp;#39;s<br /> demise. Continuing with the wakeup isn&amp;#39;t perfect either, as *something*<br /> has gone sideways if a vCPU is awakened in L2 with an injected event (or<br /> worse, a nested run pending), but continuing on gives the VM a decent<br /> chance of surviving without any major side effects.<br /> <br /> As explained in the Fixes commits, it _should_ be impossible for a vCPU to<br /> be put into a blocking state with an already-injected event (exception,<br /> IRQ, or NMI). Unfortunately, userspace can stuff MP_STATE and/or injected<br /> events, and thus put the vCPU into what should be an impossible state.<br /> <br /> Don&amp;#39;t bother trying to preserve the WARN, e.g. with an anti-syzkaller<br /> Kconfig, as WARNs can (hopefully) be added in paths where _KVM_ would be<br /> violating x86 architecture, e.g. by WARNing if KVM attempts to inject an<br /> exception or interrupt while the vCPU isn&amp;#39;t running.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> arm64: Add support for TSV110 Spectre-BHB mitigation<br /> <br /> The TSV110 processor is vulnerable to the Spectre-BHB (Branch History<br /> Buffer) attack, which can be exploited to leak information through<br /> branch prediction side channels. This commit adds the MIDR of TSV110<br /> to the list for software mitigation.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> gfs2: fiemap page fault fix<br /> <br /> In gfs2_fiemap(), we are calling iomap_fiemap() while holding the inode<br /> glock. This can lead to recursive glock taking if the fiemap buffer is<br /> memory mapped to the same inode and accessing it triggers a page fault.<br /> <br /> Fix by disabling page faults for iomap_fiemap() and faulting in the<br /> buffer by hand if necessary.<br /> <br /> Fixes xfstest generic/742.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> EFI/CPER: don&amp;#39;t go past the ARM processor CPER record buffer<br /> <br /> There&amp;#39;s a logic inside GHES/CPER to detect if the section_length<br /> is too small, but it doesn&amp;#39;t detect if it is too big.<br /> <br /> Currently, if the firmware receives an ARM processor CPER record<br /> stating that a section length is big, kernel will blindly trust<br /> section_length, producing a very long dump. For instance, a 67<br /> bytes record with ERR_INFO_NUM set 46198 and section length<br /> set to 854918320 would dump a lot of data going a way past the<br /> firmware memory-mapped area.<br /> <br /> Fix it by adding a logic to prevent it to go past the buffer<br /> if ERR_INFO_NUM is too big, making it report instead:<br /> <br /> [Hardware Error]: Hardware error from APEI Generic Hardware Error Source: 1<br /> [Hardware Error]: event severity: recoverable<br /> [Hardware Error]: Error 0, type: recoverable<br /> [Hardware Error]: section_type: ARM processor error<br /> [Hardware Error]: MIDR: 0xff304b2f8476870a<br /> [Hardware Error]: section length: 854918320, CPER size: 67<br /> [Hardware Error]: section length is too big<br /> [Hardware Error]: firmware-generated error record is incorrect<br /> [Hardware Error]: ERR_INFO_NUM is 46198<br /> <br /> [ rjw: Subject and changelog tweaks ]