Vulnerabilities

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> exfat: fix potential use-after-free in exfat_find_dir_entry()<br /> <br /> In exfat_find_dir_entry(), the buffer_head obtained from<br /> exfat_get_dentry() is released with brelse(bh) before the fall-through<br /> TYPE_EXTEND branch reads the directory entry through ep (which points<br /> into bh-&gt;b_data):<br /> <br /> brelse(bh);<br /> if (entry_type == TYPE_EXTEND) {<br /> ...<br /> len = exfat_extract_uni_name(ep, entry_uniname);<br /> ...<br /> }<br /> <br /> After brelse() drops our reference, nothing guarantees that the<br /> underlying page backing bh-&gt;b_data remains valid for the subsequent<br /> exfat_extract_uni_name() read. This is the same pattern fixed in<br /> commit fc961522ddbd ("exfat: Fix potential use after free in<br /> exfat_load_upcase_table()").<br /> <br /> Move brelse(bh) so it runs after ep is no longer dereferenced on<br /> each branch.<br /> <br /> Confirmed on QEMU x86_64 with CONFIG_KASAN=y + CONFIG_DEBUG_PAGEALLOC=y<br /> + CONFIG_PAGE_POISONING=y on linux-next, using a crafted exFAT image<br /> (long filename with same-hash collisions forcing the TYPE_EXTEND path).<br /> With a debug-only invalidate_bdev() inserted between brelse(bh) and<br /> the ep read to make the stale-deref window deterministic, the<br /> unpatched kernel faults:<br /> <br /> BUG: KASAN: use-after-free in exfat_find_dir_entry+0x133b/0x15a0<br /> BUG: unable to handle page fault for address: ffff88801a5fa0c2<br /> Oops: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN NOPTI<br /> RIP: 0010:exfat_find_dir_entry+0x1188/0x15a0<br /> <br /> With this patch applied, the same instrumented harness completes<br /> cleanly under the same sanitizer stack. I have not reproduced a<br /> crash on an uninstrumented kernel under ordinary reclaim; the<br /> instrumented A/B establishes the lifetime violation and that the<br /> patch closes it, not an unaided triggerability claim.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> bpf: use kvfree() for replaced sysctl write buffer<br /> <br /> proc_sys_call_handler() allocates its temporary sysctl buffer with<br /> kvzalloc() and passes it to __cgroup_bpf_run_filter_sysctl(). Since<br /> kvzalloc() may fall back to vmalloc() for large allocations, freeing<br /> that buffer with kfree() is wrong and can corrupt memory.<br /> <br /> Use kvfree() to safely handle both kmalloc and kvzalloc()/vmalloc<br /> allocations.<br /> <br /> The bug was first flagged by an experimental analysis tool we are<br /> developing for kernel memory-management bugs while analyzing<br /> v6.13-rc1. The tool is still under development and is not yet publicly<br /> available. Manual inspection confirms that the bug is still<br /> present in v7.1-rc5.<br /> <br /> Reproduced the bug based on v7.1-rc4 in a QEMU x86_64 guest booted with<br /> KASAN and CONFIG_FAILSLAB enabled. To exercise the replacement path, the<br /> test tree also included the accompanying fix for the stale ret == 1<br /> check in __cgroup_bpf_run_filter_sysctl(). The reproducer confines<br /> failslab injections to the proc_sys_call_handler() range, uses<br /> stacktrace-depth=32, and injects fail-nth=1 while writing 8191 bytes to<br /> /proc/sys/kernel/domainname from a task in the target cgroup. Under<br /> that setup, fail-nth=1 triggered the fault:<br /> <br /> BUG: unable to handle page fault for address: ffffeb0200024d48<br /> #PF: supervisor read access in kernel mode<br /> #PF: error_code(0x0000) - not-present page<br /> PGD 0 P4D 0<br /> Oops: Oops: 0000 SMP KASAN NOPTI<br /> CPU: 2 UID: 0 PID: 209 Comm: repro_proc_sys_ Not tainted 7.1.0-rc4-00686-g97625979a5d4 PREEMPT(lazy)<br /> Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014<br /> RIP: 0010:kfree+0x6e/0x510<br /> ...<br /> Call Trace:<br /> <br /> ? __cgroup_bpf_run_filter_sysctl+0x626/0xc30<br /> __cgroup_bpf_run_filter_sysctl+0x74d/0xc30<br /> ? __pfx___cgroup_bpf_run_filter_sysctl+0x10/0x10<br /> ? srso_return_thunk+0x5/0x5f<br /> ? __kvmalloc_node_noprof+0x345/0x870<br /> ? proc_sys_call_handler+0x250/0x480<br /> ? srso_return_thunk+0x5/0x5f<br /> proc_sys_call_handler+0x3a2/0x480<br /> ? __pfx_proc_sys_call_handler+0x10/0x10<br /> ? srso_return_thunk+0x5/0x5f<br /> ? selinux_file_permission+0x39f/0x500<br /> ? srso_return_thunk+0x5/0x5f<br /> ? lock_is_held_type+0x9e/0x120<br /> vfs_write+0x98e/0x1000<br /> ...<br /> <br /> <br /> With this fix applied on top of the same test setup, rerunning the<br /> reproducer with fail-nth=1 yields no corresponding Oops reports.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> irqchip/imgpdc: Fix resource leak, add missing chained handler cleanup on remove<br /> <br /> The driver allocates domain generic chips using<br /> irq_alloc_domain_generic_chips() during probe and sets up chained<br /> handlers using irq_set_chained_handler_and_data(). However, on driver<br /> removal, the generic chips are not freed and the chained handlers are<br /> not removed.<br /> <br /> The generic chips remain on the global gc_list and may later be accessed by<br /> generic interrupt chip suspend, resume, or shutdown callbacks after the<br /> driver has been removed, potentially resulting in a use-after-free and<br /> kernel crash.<br /> <br /> The chained handlers that were installed in probe for peripheral and<br /> syswake interrupts are also left dangling, which can lead to spurious<br /> interrupts accessing freed memory.<br /> <br /> Fix these issues by:<br /> <br /> - Setting IRQ_DOMAIN_FLAG_DESTROY_GC flag in domain-&gt;flags, so the<br /> core code automatically removes generic chips when irq_domain_remove()<br /> is called<br /> <br /> - Clearing all chained handlers with NULL in pdc_intc_remove()

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> ocfs2: reject oversized group bitmap descriptors<br /> <br /> ocfs2_validate_gd_parent() only bounds bg_bits against the parent<br /> allocator&amp;#39;s chain geometry. A malicious descriptor can still claim a<br /> bg_size/bg_bits pair that exceeds the bitmap bytes that physically fit in<br /> the group descriptor block, so later bitmap scans and bit updates can run<br /> past bg_bitmap.<br /> <br /> Add a physical-cap check based on ocfs2_group_bitmap_size() for the parent<br /> allocator type and reject descriptors whose bg_size or bg_bits exceed that<br /> capacity. Keep the existing chain geometry check so both the on-disk<br /> bitmap layout and the allocator metadata must agree before the descriptor<br /> is used.<br /> <br /> Validation reproduced this kernel report:<br /> KASAN use-after-free in _find_next_bit+0x7f/0xc0<br /> Read of size 8<br /> Call trace:<br /> dump_stack_lvl+0x66/0xa0 (?:?)<br /> print_report+0xd0/0x630 (?:?)<br /> _find_next_bit+0x7f/0xc0 (?:?)<br /> srso_alias_return_thunk+0x5/0xfbef5 (?:?)<br /> __virt_addr_valid+0x188/0x2f0 (?:?)<br /> kasan_report+0xe4/0x120 (?:?)<br /> ocfs2_find_max_contig_free_bits+0x35/0x70 (fs/ocfs2/suballoc.c:1375)<br /> ocfs2_block_group_set_bits+0x472/0x4b0 (fs/ocfs2/suballoc.c:1457)<br /> ocfs2_cluster_group_search+0x16b/0x440 (fs/ocfs2/suballoc.c:86)<br /> ocfs2_bg_discontig_fix_result+0x1ef/0x230 (fs/ocfs2/suballoc.c:1786)<br /> ocfs2_search_chain+0x8f8/0x10a0 (fs/ocfs2/suballoc.c:1886)<br /> get_page_from_freelist+0x70e/0x2370 (?:?)<br /> lock_release+0xc6/0x290 (?:?)<br /> do_raw_spin_unlock+0x9a/0x100 (?:?)<br /> kasan_unpoison+0x27/0x60 (?:?)<br /> __bfs+0x147/0x240 (?:?)<br /> get_page_from_freelist+0x83d/0x2370 (?:?)<br /> ocfs2_claim_suballoc_bits+0x38c/0xe70 (fs/ocfs2/suballoc.c:96)<br /> sched_domains_numa_masks_clear+0x70/0xd0 (?:?)<br /> check_irq_usage+0xe8/0xb70 (?:?)<br /> __ocfs2_claim_clusters+0x18d/0x4c0 (fs/ocfs2/suballoc.c:2497)<br /> check_path+0x24/0x50 (?:?)<br /> rcu_is_watching+0x20/0x50 (?:?)<br /> check_prev_add+0xfd/0xd00 (?:?)<br /> ocfs2_add_clusters_in_btree+0x17d/0x810 (fs/ocfs2/suballoc.c:?)<br /> __folio_batch_add_and_move+0x1f5/0x3d0 (?:?)<br /> ocfs2_add_inode_data+0xd9/0x120 (fs/ocfs2/suballoc.c:?)<br /> filemap_add_folio+0x105/0x1f0 (?:?)<br /> ocfs2_write_begin_nolock+0x29f7/0x2f80 (fs/ocfs2/suballoc.c:3043)<br /> ocfs2_read_inode_block+0xb5/0x110 (fs/ocfs2/suballoc.c:?)<br /> down_write+0xf5/0x180 (?:?)<br /> ocfs2_write_begin+0x180/0x240 (fs/ocfs2/suballoc.c:?)<br /> __mark_inode_dirty+0x758/0x9a0 (?:?)<br /> inode_to_bdi+0x41/0x90 (?:?)<br /> balance_dirty_pages_ratelimited_flags+0xf8/0x1d0 (?:?)<br /> generic_perform_write+0x252/0x440 (?:?)<br /> mnt_put_write_access_file+0x16/0x70 (?:?)<br /> file_update_time_flags+0xe4/0x200 (?:?)<br /> ocfs2_file_write_iter+0x80a/0x1320 (fs/ocfs2/suballoc.c:?)<br /> lock_acquire+0x184/0x2f0 (?:?)<br /> ksys_write+0xd2/0x170 (?:?)<br /> apparmor_file_permission+0xf5/0x310 (?:?)<br /> read_zero+0x8d/0x140 (?:?)<br /> lock_is_held_type+0x8f/0x100 (?:?)

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> rpmsg: char: Fix use-after-free on probe error path<br /> <br /> rpmsg_chrdev_probe() stores the newly allocated eptdev in the default<br /> endpoint&amp;#39;s priv pointer before calling rpmsg_chrdev_eptdev_add(). If<br /> rpmsg_chrdev_eptdev_add() then fails, its error path frees eptdev while<br /> the default endpoint may still dispatch callbacks with the stale priv<br /> pointer.<br /> <br /> Avoid publishing eptdev through the default endpoint until<br /> rpmsg_chrdev_eptdev_add() succeeds. Messages received before the priv<br /> pointer is published should be ignored by rpmsg_ept_cb(). Flow-control<br /> updates can hit rpmsg_ept_flow_cb() in the same window, so make both<br /> callbacks return success when priv is NULL.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> sched/mmcid: Fix OOB clear_bit when CID is MM_CID_UNSET in fixup path<br /> <br /> In mm_cid_fixup_cpus_to_tasks(), when rq-&gt;curr has the target mm and<br /> mm_cid.active is set, the CID is checked with cid_in_transit() before<br /> setting the transition bit. In per-CPU mode a newly forked or exec&amp;#39;d<br /> task can be running with mm_cid.cid == MM_CID_UNSET because CIDs are<br /> assigned lazily on schedule-in. With cid_in_transit() the guard passes<br /> for MM_CID_UNSET (no transit bit), converts it to MM_CID_UNSET |<br /> MM_CID_TRANSIT and stores it back; later mm_cid_schedout() feeds this<br /> to clear_bit() with MM_CID_UNSET as the bit number, triggering an<br /> out-of-bounds write.<br /> <br /> Symptoms: this is genuine memory corruption, but a bounded out-of-bounds<br /> write, not an arbitrary one. MM_CID_UNSET is the fixed sentinel BIT(31),<br /> so once the bad value reaches mm_cid_schedout() the cid_from_transit_cid()<br /> strip leaves MM_CID_UNSET, which fails the "cid clear_bit(MM_CID_UNSET,<br /> mm_cidmask(mm)). The cid bitmap is embedded in the mm_struct slab object<br /> (after cpu_bitmap and mm_cpus_allowed) and is only num_possible_cpus()<br /> bits wide, so clearing bit 31 is a deterministic OOB bit-clear at a<br /> fixed offset of 2^31 / 8 == 256 MiB past the bitmap base. The address is<br /> not attacker-influenced (fixed sentinel -&gt; fixed offset) and the op only<br /> clears a single bit; what sits 256 MiB further along the direct map is<br /> whatever kernel object happens to live there, so this corrupts one bit of<br /> unpredictable kernel memory -- it is not an arbitrary-address or<br /> arbitrary-value write.<br /> <br /> It triggers only in per-CPU CID mode, when a CPU is running an active<br /> task of the target mm whose cid is still MM_CID_UNSET -- the<br /> fork()/execve() window before that task&amp;#39;s next schedule-in assigns it a<br /> real CID -- and a per-CPU -&gt; per-task fixup walks over it (the mode<br /> fallback driven by a thread exit, sched_mm_cid_exit(), or by the deferred<br /> max_cids recompute in mm_cid_work_fn()).<br /> <br /> In practice syzkaller surfaced it as a KASAN use-after-free reported in<br /> __schedule -&gt; mm_cid_switch_to, where the offending clear_bit() is inlined<br /> via mm_cid_schedout() -&gt; mm_drop_cid().<br /> <br /> Guard the transition-bit assignment against MM_CID_UNSET, in addition to<br /> the existing cid_in_transit() check, so the bit is only set on a genuine<br /> task-owned CID. A CPU-owned (MM_CID_ONCPU) CID of a running active task<br /> is handled by the cid_on_cpu(pcp-&gt;cid) branch above and never reaches<br /> this path, so excluding MM_CID_UNSET (and the already-transitioning case)<br /> is sufficient.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> pNFS: Fix use-after-free in pnfs_update_layout()<br /> <br /> When hitting the NFS_LAYOUT_RETURN branch in pnfs_update_layout(),<br /> the code calls pnfs_prepare_to_retry_layoutget(lo). If it succeeds,<br /> pnfs_put_layout_hdr(lo) is called before trace_pnfs_update_layout(),<br /> which still references &amp;#39;lo&amp;#39;. This results in a use-after-free when the<br /> tracepoint accesses lo&amp;#39;s fields.<br /> <br /> Fix this by moving the tracepoint call before pnfs_put_layout_hdr(lo).

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> tipc: fix slab-use-after-free Read in tipc_aead_decrypt_done<br /> <br /> tipc_aead_decrypt() goes straight from tipc_bearer_hold(b) to<br /> crypto_aead_decrypt(req) without taking a reference on the netns, unlike<br /> the encrypt path. When crypto_aead_decrypt() is offloaded asynchronously<br /> (e.g. the SIMD aead wrapper queuing to cryptd), the cryptd worker runs<br /> tipc_aead_decrypt_done() later. If the bearer&amp;#39;s netns is torn down in the<br /> meantime, cleanup_net() -&gt; tipc_exit_net() -&gt; tipc_crypto_stop() frees the<br /> per-netns tipc_crypto, and the completion then reads it:<br /> tipc_aead_decrypt_done() dereferences aead-&gt;crypto-&gt;stats and<br /> aead-&gt;crypto-&gt;net, and tipc_crypto_rcv_complete() dereferences<br /> aead-&gt;crypto-&gt;aead[] and the node table -- reading freed memory.<br /> <br /> Decoded KASAN splat (v7.1-rc7, CONFIG_KASAN_INLINE + TIPC + TIPC_CRYPTO):<br /> <br /> BUG: KASAN: slab-use-after-free in tipc_aead_decrypt_done (net/tipc/crypto.c:999)<br /> Read of size 8 at addr ffff8881056258a8 by task kworker/u16:2/51<br /> Workqueue: events_unbound<br /> Call Trace:<br /> tipc_aead_decrypt_done (net/tipc/crypto.c:999)<br /> process_one_work (kernel/workqueue.c:3314)<br /> worker_thread (kernel/workqueue.c:3397 kernel/workqueue.c:3478)<br /> kthread (kernel/kthread.c:436)<br /> ret_from_fork (arch/x86/kernel/process.c:158)<br /> ret_from_fork_asm (arch/x86/entry/entry_64.S:245)<br /> <br /> Allocated by task 169:<br /> __kasan_kmalloc (mm/kasan/common.c:398 mm/kasan/common.c:415)<br /> tipc_crypto_start (net/tipc/crypto.c:1502)<br /> tipc_init_net (net/tipc/core.c:72)<br /> ops_init (net/core/net_namespace.c:137)<br /> setup_net (net/core/net_namespace.c:446)<br /> copy_net_ns (net/core/net_namespace.c:579)<br /> create_new_namespaces (kernel/nsproxy.c:132)<br /> __x64_sys_unshare (kernel/fork.c:3316)<br /> do_syscall_64 (arch/x86/entry/syscall_64.c:63)<br /> entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:121)<br /> <br /> Freed by task 8:<br /> kfree (mm/slub.c:6566)<br /> tipc_exit_net (net/tipc/core.c:119)<br /> cleanup_net (net/core/net_namespace.c:704)<br /> process_one_work (kernel/workqueue.c:3314)<br /> kthread (kernel/kthread.c:436)<br /> <br /> This is the same class of bug that commit e279024617134 ("net/tipc: fix<br /> slab-use-after-free Read in tipc_aead_encrypt_done") fixed for the encrypt<br /> side. The encrypt path takes maybe_get_net(aead-&gt;crypto-&gt;net) before<br /> crypto_aead_encrypt() and drops it with put_net() on the synchronous<br /> return paths and in tipc_aead_encrypt_done(); the -EINPROGRESS/-EBUSY<br /> return keeps the reference for the async callback to release. The decrypt<br /> path was left without the equivalent guard.<br /> <br /> Mirror the encrypt-side fix on the decrypt path: take a net reference<br /> before crypto_aead_decrypt() (failing with -ENODEV and the matching<br /> bearer put if it cannot be acquired), keep it across the<br /> -EINPROGRESS/-EBUSY async return, and drop it with put_net() on the<br /> synchronous success/error return and at the end of<br /> tipc_aead_decrypt_done().<br /> <br /> Reproduced under KASAN on v7.1-rc7: a UDP bearer with a cluster key is<br /> flooded with crafted encrypted frames from an unknown peer (driving the<br /> cluster-key decrypt path) while the bearer&amp;#39;s netns is repeatedly torn<br /> down. The completion must run asynchronously to outlive<br /> tipc_crypto_stop(); on x86 the stock aesni gcm(aes) now decrypts<br /> synchronously, so the async path was exercised via cryptd offload. The<br /> unguarded aead-&gt;crypto dereference in tipc_aead_decrypt_done() is the<br /> unpatched upstream path; tipc_aead_decrypt() still lacks<br /> maybe_get_net(aead-&gt;crypto-&gt;net), so the completion can outlive the free<br /> on any config where crypto_aead_decrypt() goes async.<br /> <br /> Found by 0sec automated security-research tooling (https://0sec.ai).

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> blk-cgroup: fix UAF in __blkcg_rstat_flush()<br /> <br /> When multiple blkgs in the same blkcg are released concurrently,<br /> a use-after-free can occur. The race happens when one blkg&amp;#39;s<br /> __blkcg_rstat_flush() removes another blkg&amp;#39;s iostat entries via<br /> llist_del_all(). The second blkg sees an empty list and proceeds<br /> to free itself while the first is still iterating over its entries.<br /> <br /> Move the flush from __blkg_release() (RCU callback) to blkg_release()<br /> (before call_rcu). This ensures the RCU grace period waits for any<br /> concurrent flush&amp;#39;s rcu_read_lock() section to complete before freeing.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> hdlc_ppp: sync per-proto timers before freeing hdlc state<br /> <br /> Each PPP control protocol (LCP/IPCP/IPV6CP) embedded in struct ppp<br /> registers a timer via timer_setup(). That struct ppp is the<br /> hdlc-&gt;state allocation, which detach_hdlc_protocol() frees with kfree()<br /> in both teardown paths: unregister_hdlc_device() and the re-attach inside<br /> attach_hdlc_protocol().<br /> <br /> The ppp proto never registered a .detach callback, so<br /> detach_hdlc_protocol() performs no timer synchronization before the<br /> kfree(). The only cancel, timer_delete(&amp;proto-&gt;timer) in ppp_cp_event(),<br /> is partial (it does not wait for a running callback) and only runs on the<br /> -&gt;CLOSED transition; ppp_stop()/ppp_close() do not sync either. A<br /> ppp_timer callback already executing (blocked on ppp-&gt;lock) survives the<br /> kfree and then dereferences proto-&gt;state / ppp-&gt;lock in freed memory,<br /> leading to a use-after-free.<br /> <br /> Fix this by adding a .detach helper that calls timer_shutdown_sync() on<br /> every per-proto timer. detach_hdlc_protocol() invokes proto-&gt;detach(dev)<br /> before kfree(hdlc-&gt;state), so timer_shutdown_sync()<br /> now runs on both free paths.<br /> timer_shutdown_sync() is used instead of timer_delete_sync() because the<br /> keepalive path re-arms the timer through add_timer()/mod_timer() and<br /> shutdown blocks any re-activation during teardown.<br /> <br /> Initialize the per-protocol timers in ppp_ioctl() when the protocol is<br /> attached, and remove the now-redundant timer_setup() from ppp_start(), so<br /> that the timers are initialized exactly once at attach time and<br /> ppp_timer_release() never operates on uninitialized timer_list<br /> structures. attach_hdlc_protocol() uses kmalloc() (not kzalloc), so<br /> struct ppp&amp;#39;s protos[i].timer is uninitialized garbage until the first<br /> timer_setup(); without this init-at-attach, attaching the PPP protocol<br /> without ever bringing the device up would leave timer_shutdown_sync()<br /> operating on uninitialized memory in .detach. Moving the init out of<br /> ppp_start() (which only runs on NETDEV_UP) into the attach path makes the<br /> initialization unconditional and avoids initializing the same timer_list<br /> twice.<br /> <br /> This bug was found by static analysis.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> KVM: SVM: Fix page overflow in sev_dbg_crypt() for ENCRYPT path<br /> <br /> In sev_dbg_crypt(), the per-iteration transfer length is bounded by<br /> the source page offset (PAGE_SIZE - s_off) but not by the destination<br /> page offset (PAGE_SIZE - d_off). When d_off &gt; s_off, the encrypt<br /> path (__sev_dbg_encrypt_user) performs a read-modify-write using a<br /> single-page intermediate buffer (dst_tpage):<br /> <br /> 1. __sev_dbg_decrypt() expands the size to round_up(len + (d_off &amp; 15), 16)<br /> before issuing the PSP command. If len + (d_off &amp; 15) &gt; PAGE_SIZE,<br /> the PSP writes beyond the end of the 4096-byte dst_tpage allocation.<br /> <br /> 2. The subsequent memcpy()/copy_from_user() into<br /> page_address(dst_tpage) + (d_off &amp; 15) of &amp;#39;len&amp;#39; bytes overflows<br /> by up to 15 bytes under the same condition.<br /> <br /> Trigger example: s_off = 0, d_off = 1, debug.len = PAGE_SIZE -<br /> the PSP is instructed to write round_up(4097, 16) = 4112 bytes to<br /> a 4096-byte buffer.<br /> <br /> Fix by also bounding len by (PAGE_SIZE - d_off), the same check that<br /> sev_send_update_data() already performs for its single-page guest<br /> region.<br /> <br /> ==================================================================<br /> BUG: KASAN: slab-use-after-free in sev_dbg_crypt+0x993/0xd10 [kvm_amd]<br /> Write of size 4095 at addr ff110062293bb009 by task sev_dbg_test/228214<br /> <br /> CPU: 96 UID: 0 PID: 228214 Comm: sev_dbg_test Tainted: G U W 7.0.0-smp--5ce9b0c48211-dbg #156 PREEMPTLAZY<br /> Tainted: [U]=USER, [W]=WARN<br /> Hardware name: Google Astoria/astoria, BIOS 0.20250817.1-0 08/25/2025<br /> Call Trace:<br /> <br /> dump_stack_lvl+0x54/0x70<br /> print_report+0xbc/0x260<br /> kasan_report+0xa2/0xd0<br /> kasan_check_range+0x25f/0x2c0<br /> __asan_memcpy+0x40/0x70<br /> sev_dbg_crypt+0x993/0xd10 [kvm_amd]<br /> sev_mem_enc_ioctl+0x33c/0x450 [kvm_amd]<br /> kvm_vm_ioctl+0x65d/0x6d0 [kvm]<br /> __se_sys_ioctl+0xb2/0x100<br /> do_syscall_64+0xe8/0x870<br /> entry_SYSCALL_64_after_hwframe+0x4b/0x53<br /> <br /> <br /> The buggy address belongs to the physical page:<br /> page: refcount:1 mapcount:0 mapping:0000000000000000 index:0x7fe72b6a0 pfn:0x62293bb<br /> memcg:ff11000112827d82<br /> flags: 0x1400000000000000(node=1|zone=1)<br /> raw: 1400000000000000 0000000000000000 dead000000000122 0000000000000000<br /> raw: 00000007fe72b6a0 0000000000000000 00000001ffffffff ff11000112827d82<br /> page dumped because: kasan: bad access detected<br /> <br /> Memory state around the buggy address:<br /> ff110062293bbf00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00<br /> ff110062293bbf80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00<br /> &gt;ff110062293bc000: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc<br /> ^<br /> ff110062293bc080: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc<br /> ff110062293bc100: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc<br /> ==================================================================<br /> Disabling lock debugging due to kernel taint<br /> <br /> [sean: add sample KASAN splat, Fixes, and stable@]

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> fbdev: Fix fb_new_modelist to prevent null-ptr-deref in fb_videomode_to_var<br /> <br /> info-&gt;var, a framebuffer&amp;#39;s current mode, is expected to have a matching<br /> entry in info-&gt;modelist. var_to_display() relies on this and treats a<br /> failed fb_match_mode() as "This should not happen". fb_set_var() keeps it<br /> true by adding the mode to the list on every change, and<br /> do_register_framebuffer() does the same at registration.<br /> <br /> store_modes() replaces the modelist from userspace. fb_new_modelist()<br /> validates the new modes but does not check that info-&gt;var still has a<br /> match. It relies on fbcon_new_modelist() to re-point consoles, but that<br /> only handles consoles mapped to the framebuffer. With fbcon unbound there<br /> are none, so info-&gt;var is left describing a mode that is no longer in the<br /> list.<br /> <br /> A later console takeover runs var_to_display(), where fb_match_mode()<br /> returns NULL and leaves fb_display[i].mode NULL. fbcon_switch() passes it<br /> to display_to_var(), and fb_videomode_to_var() dereferences the NULL mode.<br /> <br /> Keep the current mode in the list in fb_new_modelist(), the same way<br /> fb_set_var() does.