CVE Feed

In the Linux kernel, the following vulnerability has been resolved: KVM: nVMX: Service local TLB flushes on failed nested VM-Enter KVM services local TLB flushes on "full" nested VM-Exits (through __nested_vmx_vmexit()), but not if a nested VM-Enter fails (e.g. due to failed VMCS checks in nested_vmx_enter_non_root_mode()). However, it is possible that KVM had queued TLB flushes that need to be performed, even if the nested VM-Enter was not successful. For example, if VPID is disabled for L2 (via nested_vmx_transition_tlb_flush(), or if via the MSR load lists, as the SDM says: If any MSR is being loaded in such a way that would architecturally require a TLB flush, the TLBs are updated so that, after VM entry, the logical processor will not use any translations that were cached before the transition. The SDM is unclear about when the TLB flush should occur, and whether or not a failed VM entry would flush the TLB, so it is safer to always do the TLB flush in this case. More concretely, KVM also updates the last VPID L1 used for L2 in nested_vmx_transition_tlb_flush() (i.e. last_vpid), even if the VM entry ultimately fails. With the current code, KVM could miss a TLB flush if L1 changes L2's VPID, then does a failed VM entry followed by a successful one, as the failed VM entry would update last_vpid but not actually flush the TLB. Servicing local TLB flushes on failed VM entries makes sure that the TLB is always flushed when last_vpid is updated.

In the Linux kernel, the following vulnerability has been resolved: KVM: nVM: Ensure INVVPID is emulated on the correct physical CPU When emulating INVVPID, KVM executes INVVPID on the physical CPU using vpid02 (instead of the L1 assigned VPID), after doing some validations on the operands. However, it is possible that the physical CPU KVM executes INVVPID on is different from the CPU L2 is running on. For example, in the following scenario: - L2 runs on CPU #1 and exits to L1 (vmx->nested.vmcs02.cpu=1) - L1 migrates to CPU #2 and executes INVVPID - KVM executes INVVPID on CPU #2 - L1 migrates back to CPU #1 and runs L2 (vmx->nested.vmcs02.cpu=1) The TLB entries on CPU #1 are never invalidated, because INVVPID was executed on CPU #2, and vmcs02 never ran on a different pCPU (i.e. vmx_vcpu_load_vmcs() will *not* request KVM_REQ_TLB_FLUSH). Ensure that INVVPID is being executed on the same pCPU that L2 last ran on, and if not, fallback to clearing last_vpid=0 to trigger a full VPID flush on the next nested VM-Enter (as KVM will detect L1 using a different VPID for L2). If L2 ends up running on a different pCPU, KVM will flush the TLB anyway through vmx_vcpu_load_vmcs().

In the Linux kernel, the following vulnerability has been resolved: KVM: x86/mmu: Consume the locked rmap value in the lockless rmap walk __kvm_rmap_lock() deliberately elides the rmap lock when it observes an empty rmap. In that case kvm_rmap_lock_readonly() also re-enables preemption and returns zero, so the caller holds neither the rmap lock nor a preemption reference. The elision documents the invariant it relies on: * Elide the lock if the rmap is empty, as lockless walkers (read-only * mode) don't need to (and can't) walk an empty rmap, nor can they add * entries to the rmap. I.e. the only paths that process empty rmaps * do so while holding mmu_lock for write, and are mutually exclusive. kvm_rmap_age_gfn_range() ignores the returned value and unconditionally enters for_each_rmap_spte_lockless(). The iterator started with rmap_get_first(), which re-reads rmap_head->val rather than using the value returned by the lock. If a writer populates the rmap between the lock's read and the iterator's re-read, the aging path walks the newly installed rmap without holding its lock. For a KVM_RMAP_MANY rmap this leaves the walker following a pte_list_desc chain that it never locked. A writer holding mmu_lock for write may free that chain (e.g. kvm_zap_all_rmap_sptes() on the recycle path, or any rmap zap) via kmem_cache_free() while the walk is in progress, giving a slab use-after-free. Nothing serialises the two: the aging path runs without mmu_lock when CONFIG_KVM_MMU_LOCKLESS_AGING=y, and the rmap lock that would otherwise exclude the writer was elided. Because the empty path re-enables preemption, the interval between the two reads can span an arbitrary scheduling delay. Fix the class of bug by having the lockless walk consume the value returned by the lock instead of re-reading the rmap. Split rmap_get_first() into __rmap_get_first(), which starts an iterator from an already-read rmap value, and make for_each_rmap_spte_lockless() take that value and call __rmap_get_first() directly. kvm_rmap_age_gfn_range() passes the value returned by kvm_rmap_lock_readonly(): when the lock was elided the value is zero, __rmap_get_first() returns NULL, and the walk is skipped. No lockless walker re-reads the rmap, so the lock-elision invariant cannot be violated, and no lock()-without-paired-unlock() path is added to the aging code.

In the Linux kernel, the following vulnerability has been resolved: KVM: x86: hyper-v: Clamp stimer deadline to avoid livelock Fix an issue where userspace or the guest can program an Hyper-V synthetic timer to have a deadline in the past via integer overflow, preventing the CPU from making progress and triggering an RCU stall. Hyper-V's SynIC exposes 4 per-vCPU synthetic timers to the guest, which are emulated by KVM. Each is programmed through the HV_X64_MSR_STIMERi_CONFIG and HV_X64_MSR_STIMERi_COUNT MSRs. Depending on CONFIG, COUNT represents either the absolute expiration time or the period of a periodic timer, both expressed in 100ns ticks. These timers may be set both by the guest (WRMSR) and the host (KVM_SET_MSRS). When the timer is enabled, stimer_start() translates COUNT to an absolute monotonic deadline and arms an hrtimer. If COUNT is set to a value close to U64_MAX, the deadline calculation can overflow. ktime_add_ns(ktime_now, 100 * (stimer->exp_time - time_now)) This can result in a CPU livelock. stimer_start() arms the timer via hrtimer_start() with a deadline in the past, which causes it to immediately fire. The stimer callback then raises KVM_RQ_HV_STIMER, with the intention of causing KVM to deliver a synthetic interrupt on the next vCPU guest enter. Then, once userspace issues KVM_RUN, vcpu_enter_guest() consumes the request, calling kvm_hv_process_stimers(). This would normally disable the timer via stimer_expiration() once the deadline is in the past. However, the deadline comparison is done between the KVM reference counter and stime->exp_time, which is a big value close to U64_MAX, so this never happens for a few thousand years. kvm_hv_process_timers() then re-arms the timer via stimer_start(), since it was not disabled, which again fires immediately. Before entering the guest, kvm_vcpu_exit_request() checks kvm_request_pending(), which returns true due to the newly raised KVM_REQ_HV_STIMER. Then vcpu_enter_guest() aborts the guest entry, returning early into vcpu_run(), which loops back again into vcpu_enter_guest(), restarting the cycle. Since there are no manual yields in this loop, a task with SCHED_FIFO may starve RCU grace-period kthreads, which exposes the stalls found by syzcaller: rcu: INFO: rcu_preempt detected stalls on CPUs/tasks: rcu: (detected by 1, t=10502 jiffies, g=14269, q=1142 ncpus=2) rcu: All QSes seen, last rcu_preempt kthread activity 10500 (4294965239-4294954739), jiffies_till_next_fqs=1, root ->qsmask 0x0 rcu: rcu_preempt kthread starved for 10500 jiffies! g14269 f0x2 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=0 rcu: Unless rcu_preempt kthread gets sufficient CPU time, OOM is now expected behavior. ( ... ) Call Trace: <IRQ> __run_hrtimer kernel/time/hrtimer.c:1773 [inline] __hrtimer_run_queues+0x408/0xc30 kernel/time/hrtimer.c:1841 hrtimer_interrupt+0x45b/0xaa0 kernel/time/hrtimer.c:1903 local_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1045 [inline] __sysvec_apic_timer_interrupt+0x102/0x3e0 arch/x86/kernel/apic/apic.c:1062 instr_sysvec_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1056 [inline] sysvec_apic_timer_interrupt+0xa1/0xc0 arch/x86/kernel/apic/apic.c:1056 </IRQ> <TASK> asm_sysvec_apic_timer_interrupt+0x1a/0x20 arch/x86/include/asm/idtentry.h:697 RIP: 0010:__raw_spin_unlock_irqrestore include/linux/spinlock_api_smp.h:152 [inline] RIP: 0010:_raw_spin_unlock_irqrestore+0xa8/0x110 kernel/locking/spinlock.c:194 Code: 74 05 e8 0b f4 5f f6 48 c7 44 24 20 00 00 00 00 9c 8f 44 24 20 f6 44 24 21 02 75 4f f7 c3 00 02 00 00 74 01 fb bf 01 00 00 00 <e8> 23 6b 27 f6 65 8b 05 7c 60 5a 07 85 c0 74 40 48 c7 04 24 0e 36 RSP: 0018:ffffc900040a7320 EFLAGS: 00000206 RAX: 5de15cb931505900 RBX: 0000000000000a06 RCX: 5de15cb931505900 RDX: 0000000000000007 RSI: ffffffff8daa9dc3 RDI: 0000000000000001 RBP: ffffc900040a73b0 R08: ffffffff8fc3d0 ---truncated---

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Fix length check __import_wp_info() struct kvm_hw_breakpoint::len is a __u64 that is fully controlled by user space. This is then assigned to wp_info->len, which is an int. The bounds check is done on the truncated value while the allocation uses the untruncated one: wp_info->len = bp_data->len; [...] if (wp_info->len < 0 || wp_info->len > MAX_WP_SIZE) return -EINVAL; wp_info->old_data = kmalloc(bp_data->len, GFP_KERNEL_ACCOUNT); Use the validated value for the allocation as intended. Without this fix userspace can trigger >4GB allocations which will fail and result in a WARN due to MAX_PAGE_ORDER.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Fix memory leak in guest debug handling bp_data is freed only for the error case by kfree(bp_data). Every successful KVM_SET_GUEST_DEBUG will leak bp_data.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Fix old_data leak in guest debug error path __import_wp_info() allocates a per-watchpoint old_data buffer to back up the original guest memory contents. If a later watchpoint of the same KVM_SET_GUEST_DEBUG request fails to import, kvm_s390_import_bp_data() jumps to the error label, which frees the wp_info array but not the old_data buffers of the entries that were imported successfully. Up to MAX_BP_COUNT - 1 buffers of up to MAX_WP_SIZE bytes are leaked per failed request, and the request can be repeated. Create error handling for cleaning up all created old_data memory areas.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Free guest debug data on vcpu destroy kvm_s390_clear_bp_data() is only called from kvm_arch_vcpu_ioctl_set_guest_debug(), i.e. when user space changes or disables debugging. A vCPU that is destroyed while hardware breakpoints are still armed - the normal case when the VMM just exits or crashes - leaks hw_bp_info, hw_wp_info and all old_data buffers, since generic KVM frees the vCPU right after kvm_arch_vcpu_destroy(). That is bounded by MAX_BP_COUNT entries, so roughly 8 KiB per vCPU, but it is unbounded over VM lifetimes. The allocations are GFP_KERNEL_ACCOUNT, so the charge also outlives the exiting process and pins dying memcgs. Fix by clearing the debug data on vCPU destruction. Calling it unconditionally is fine: struct kvm_vcpu is zero allocated, so for a vCPU that never enabled debugging the counters are 0 and the pointers NULL.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Take srcu when importing watchpoint data __import_wp_info() backs up the original guest memory contents of a watchpoint with read_guest_abs(), which is kvm_read_guest() and therefore resolves the memslot via __kvm_memslots(). That requires kvm->srcu (or kvm->slots_lock) to be held, otherwise a concurrent memslot update can free the memslots array under us once its SRCU grace period has elapsed. As this is not fast path, following lock ordering (mutex first, then srcu) take the big hammer and hold the srcu for the full import.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Zero initialize data structures for inject_pfault_token __kvm_inject_pfault_token() only sets .type and .u.ext.ext_params2 of the on-stack struct kvm_s390_irq but the full ext substructure is copied into the cpu local variable on inject. ext_params and pad contain stale stack values. Interrupt delivery only uses ext_params2, so nothing leaks to the guest, but a host user can use the migration ioctls to get to the data. Fix by zero-initializing the irq struct. Do the same for the inti data structure.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: Fix memory corruption by not reinjecting CK machine checks Channel-subsystem damage machine checks are for the host channel subsystem. The guest channel subsystem is emulated in the userspace VMM. There is no point in forwarding such machine checks into the guest. This also simplifies the machine check reinjection and avoids kfree of a stack variable as reported by sashiko. There might be still machine checks that have the ck bit set with another bit (like instruction damage), mask out the CK bit in s390_backup_mcck_info(), like the CP and ED bits already are.

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: keyop: use mmu_lock to read gmap->asce Every other dat_* consumer in this file (kvm_s390_get_skeys, set_skeys, get_cmma_bits, set_cmma_bits, MEM_CLR_CMMA, kvm_s390_fixup_prefix, kvm_test_age_gfn, kvm_age_gfn) reads kvm->arch.gmap->asce *inside* the mmu_lock read-side. keyop is the only outlier. gmap->asce is mutated under write_lock(mmu_lock) by gmap_set_limit() and keyop might use a stale asce value for walking as KVM_S390_KEYOP and KVM_S390_VM_MEM_LIMIT_SIZE can run concurrently. This can result in memory corruption.

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Correctly handle end of VA space TLBI invalidation Our TLB invalidation by VA code is based on comparing two ranges, one defined by the TLB, and one defined by the TLBI instruction. Each range is defined by a start and a size. However, the way the comparison is done doesn't account for address rollover, as it compares an address with (base + size). This works nicely until this expression represent the last page/block in the TTBR1 VA space, as the result is a big fat 0. And a failed TLB invalidation. Rewrite the comparison in a way that is immune to the address rollover (making the end address inclusive instead of exclusive), and move this into a common helper that is used by both VA and IPA invalidations, as suggested by Hyunwoo Kim (although the IPA version didn't suffer from this particular problem, obviously).

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Handle VNCR TLB invalidation race with vcpu_put() VNCR unmapping While VNCR TLB invalidation always occurs under the MMU lock, vcpu_put() doesn't, while it unmaps the VNCR page. The problem is that the invalidation evaluates vncr_tlb::cpu to decide whether an unmapping needs to take place (cpu != -1) before performing it. On the other hand, this_cpu_reset_vncr_fixmap() unconditionally unmaps if L1_VNCR_MAPPED is set. These two obviously can race, with a TOCTOU pattern on the TLBI path, and a BUG_ON() on the vcpu_put() path. And the two can end-up calling vncr_fixmap(-1), with extra lethal effects. Move the reset of vncr_tlb::cpu to -1 to a common function, and make this update atomic so that only a single thread can reset the field and perform the corresponding unmap. The vcpu_put() still need to unconditionally unmap the current VNCR to close another ugly race. Finally, the assignment of vncr_tlb::cpu is moved to be kept in sync with the actual mapping, similar to L1_VNCR_MAPPED being set.

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Make VNCR invalidation participate in MMU invalidation retry A VNCR TLB invalidation can occur on one vcpu while another vcpu is faulting in this same page. Without correctly handling this, we can end up with the following scenario: - vcpu A walks the PTs to translate VNCR - before vcpu A is able to grab the MMU lock to insert the TLB, vcpu B updates the S1 PTs with an invalid entry, and issues a TLBI S1E2 for this VA - vcpu A inserts the TLB for something that is now invalid This isn't a new problem, and we manage S2 by having the MMU notifier to bump up mmu_invalidate_seq on invalidation so that the fault can be replayed. We can perform something similar here, and extend invalidate_vncr_va() to update the same counter, clearly indicating that the context has changed under our feet. This is safe as the invalidation always happen while holding the MMU lock for write, and that we sample the sequence number before walking S1.

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Remove VM-wide VNCR mapping counter The global VNCR mapping counter is used to decide whether an L1 provided VNCR page is mapped in L0 on any CPU at the point of dealing with a TLB invalidation. It is incremented when a mapping is made in the fixmap, and decremented when unmapped. As it turns out, this tracking has several flaws: - we are trying to invalidate TLBs, and the mapping is only an opportunistic consequence of the TLB. Checking this counter to decide whether a TLB needs to be invalidated may result in missed invalidations. - an L1 vcpu invalidating its own TLB (a very likely case) will not succeed in invalidating the VNCR pseudo TLB because that page is not mapped in L0 at this stage. Given that this tracking fails at delivering the minimum guarantees that are required and is only a performance optimisation, remove it completely.

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Sign-extend VA for range-based TLBI invalidation When the decode_range_tlbi() helper was moved to be used for S1 TLBIs, the required sign extension was omitted. Add it. As a result, special care must be taken to not overflow PA bits when this is used for S2 invalidation.

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: vgic-v3: take an LPI reference in vgic_v3_save_pending_tables vgic_v3_save_pending_tables() iterates dist->lpi_xa using xa_for_each() and dereferences the returned struct vgic_irq in the loop body without holding a reference on the LPI. The xarray iterator only provides temporary RCU coverage while looking up the current entry. That is not sufficient for this loop body, which reads fields from struct vgic_irq and performs guest memory accesses before the iteration completes. A concurrent path can trigger this race: the irqfd cached injection path (vgic_its_inject_cached_translation) obtains a transient LPI reference via vgic_its_check_cache() without holding kvm->lock, vcpu->mutex, config_lock, or its_lock. If guest ITS DISCARD then drops the cache and ITE references under its_lock, the transient inject reference may become the final one. When vgic_put_irq() drops it, the LPI is erased from lpi_xa and freed via kfree_rcu(). Meanwhile, vgic_v3_save_pending_tables() may still hold a stale pointer obtained from the xarray iterator and dereference it after the RCU grace period completes. Fix this by re-fetching each iterated LPI via vgic_get_irq(), which takes a stable reference, and dropping it with vgic_put_irq() on all paths. This matches the pattern already used by other lpi_xa iterators in the vgic ITS code.

In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Correctly cap TLBI Range to the architural limit TLB Invalidation by Range has a fairly powerful way of encoding pretty large ranges in a small number of bits. This range can be based on an arbitrary VA, which means it is pretty easy for a guest to generate an overflow should the hypervisor be naive enough to add the range to the base... Make sure the range is capped to the limit dictated by the address bit that determines the VA range. For an IPA invalidation, this is further corrected down the line to ignore the upper range.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: KVM: Fix uninitialized stack variable issue with dmsintc Variable vector[] is declared on stack in function dmsintc_inject_irq() and sometimes it is used without initialized. Here fix this issue.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: KVM: Free init resources if kvm_init() fails kvm_loongarch_init() calls kvm_loongarch_env_init() to allocate the per-CPU kvm_context (vmcs) and kvm_loongarch_ops and to register the perf callbacks, and then calls kvm_init(). If kvm_init() fails its result is returned directly, but since module_init() does not run the module_exit() stuff on failure, so kvm_loongarch_env_exit() is never called and those resources are leaked. So call kvm_loongarch_env_exit() when kvm_init() fails, matching the teardown-on-failure pattern used by riscv_kvm_init().

In the Linux kernel, the following vulnerability has been resolved: LoongArch: KVM: Preserve memslot arch flags on KVM_MR_FLAGS_ONLY kvm_arch_prepare_memory_region() computes new->arch.flags, i.e. whether a memslot is KVM_MEM_HUGEPAGE_CAPABLE or KVM_MEM_HUGEPAGE_INCAPABLE, only for KVM_MR_CREATE and KVM_MR_MOVE, and returns early for every other change. But the generic code allocates a zeroed memslot for every change and never copies old->arch, so after a KVM_MR_FLAGS_ONLY update, e.g. toggling KVM_MEM_LOG_DIRTY_PAGES for live migration, the active memslot has arch.flags == 0. With both flags clear, fault_supports_huge_mapping() falls through to the alignment check on the HVA range alone, which no longer verifies that the GPA and HVA have the same offset within a PMD. A memslot that was marked KVM_MEM_HUGEPAGE_INCAPABLE because of a GPA/HVA offset mismatch can then be mapped with PMD entries on read faults, and since kvm_map_page() aligns the gfn and the pfn independently, the guest ends up accessing the wrong host pages, exactly the "d -> f, e -> g" case described in the comment above the check. Carry the arch flags over from the old memslot for KVM_MR_FLAGS_ONLY, as the GPA, HVA and size are guaranteed to be unchanged for that case.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: KVM: Validate MSI data before routing it to EIOINTC pch_msi_set_irq() passes e->msi.data straight into eiointc_set_irq() as the irq number. The MSI data comes from userspace, that either via a KVM_IRQ_ROUTING_MSI entry set with KVM_SET_GSI_ROUTING (used by irqfd and KVM_IRQ_LINE) or directly via KVM_SIGNAL_MSI, and is never checked against EIOINTC_IRQS. eiointc_set_irq() uses the value with __set_bit()/__clear_bit() on the 256-bit isr bitmap, eiointc_update_irq() then indexes sw_coremap[] and the per-cpu coreisr/sw_coreisr bitmaps with it. Therefore a data value >= 256 reads and writes memory past the end of those arrays, i.e. any process holding a VM fd can corrupt kernel memory beyond the allocation of loongarch_eiointc. Reject MSI data that doesn't fit in the EIOINTC irq space. The DMSINTC path is unaffected as it decodes the vector from the address and masks it.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: BPF: Refactor jump offset calculation in tail call The old macro-based jmp_offset calculation derives the jump distance from a stale prior-pass code stride, which can lead to wrong branch offsets and soft lockups under extra JIT passes. Fix this by calculating the offset directly on the absolute target: "ctx->offset[insn + 1] - ctx->idx". To avoid a false 16-bit range check abort during size estimation, add a "ctx->image == NULL" guard to inject a safe dummy offset.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: BPF: Move arena register slot below TCC context Currently, the stack layout places the optional arena register slot above the tail call counter context. When arena_vm_start is dynamically enabled, it shifts the relative offset of the tcc_ptr slot within the stack frame, causing hardcoded tracking macros to mismatch and leading to memory misalignment or corruption potentially. To fix this, move the arena register save and restore sequences below the tail call counter context slots in both build_prologue() and the epilogue. Update __build_epilogue() to insert a proper offset decrement to safely skip the unneeded tcc_ptr reading block while accurately aligning with the relocated arena slot at the very bottom. With this patch, the tcc_ptr slot is always positioned at a fixed distance directly underneath the base callee-saved registers that is independent of whether the arena features are on.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: Fix acpi_package_ids[] array overflow With LoongArch virt machine, a typical setting is one core per socket, there will max 256 sockets (packages) on one VM. With PPTT acpi table, array acpi_package_ids[] will be overflowed. Here change the array size of acpi_package_ids[] with the max value of MAX_PACKAGES and KVM_MAX_VCPUS.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: Do not save/restore percpu base register in rethook trampoline The rethook trampoline saves $r21 ($u0), the percpu base, into its frame at entry and restores it at exit. Inbetween rethook_trampoline_handler() may schedule via preempt_enable_notrace(). If the task migrates to another CPU, the frame's $r21 holds the old CPU's percpu base, and restoring it poisons $r21 on the new CPU. Until the next user->kernel transition heals $r21, all this_cpu_*() accesses (runqueues, RCU per-CPU data, timer tick programming, FPU ownership) hit the wrong CPU's percpu area. Under kretprobe-heavy preemptible load this can corrupt scheduler and timer state: scheduling-while-atomic splats, wrong-CPU RCU warnings, WARN_ON_ONCE(rq != this_rq()) in nohz_balance_exit_idle(), and CPUs parking in the idle loop with the constant timer never re-armed (hard lockup). Reproduces on a Loongson-3A6000 with kretprobes on VFS paths plus heavy file churn (OS install / unsquashfs). By convention $r21 always holds the current CPU's percpu base in kernel mode: SAVE_SOME() at exception entry reloads it only when coming from user mode, and RESTORE_SOME() restores it only when returning to user mode; the context-switch path never writes it. Therefore the live $r21 at trampoline exit is already correct, and nothing inbetween can change it legitimately (kernel C code cannot write a global register variable). The same flaw existed even in the pre-rethook kretprobe trampoline since v6.3; it was carried over when rethook replaced it. Drop both the save and the restore here. Drop the restore is enough to solve the issue, and drop the save is to keep the code tidy and no need to clear it.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: Avoid preempt count underflow without probe LoongArch uses break 11 for the breakpoint placed after an instruction that Kprobes executes out of line. Since userspace can issue the same break instruction, do_bp() can reach kprobe_singlestep_handler() when there is no current probe. The handler actually returns false in this case, but it first calls preempt_enable_no_resched(). The corresponding preempt_disable() is done by kprobe_breakpoint_handler() on a real Kprobe hit, so it has not run here. As a result, an ordinary userspace breakpoint (code 11) underflows the current task's preempt count. This also makes in_interrupt() return true until the task schedules. One visible consequence is the socket cgroup attribution: cgroup_sk_alloc() treats the allocation as interrupt context and assigns the socket to the root cgroup. A socket opened from the SIGTRAP handler can then avoid a BPF_CGROUP_INET_SOCK_CREATE policy attached to the task's own cgroup. Return as soon as kprobe_running() reports no active probe. The same check has appeared in [PATCH v10 2/4] of the original LoongArch Kprobes series, but was dropped before the feature reached mainline.

In the Linux kernel, the following vulnerability has been resolved: media: airspy: use vb2_video_unregister_device() on disconnect to fix NULL deref airspy_disconnect() clears s->udev under v4l2_lock, but airspy_stop_streaming() unconditionally calls airspy_ctrl_msg() and airspy_free_stream_bufs() afterwards. If a streaming user closes the device after disconnect, stop_streaming() runs and dereferences the NULL s->udev: airspy_stop_streaming() airspy_ctrl_msg(s, CMD_RECEIVER_MODE, 0, 0, NULL, 0) usb_sndctrlpipe(s->udev, 0) /* NULL deref */ airspy_free_stream_bufs(s) usb_free_coherent(s->udev, ...) /* NULL deref */ The airspy driver uses vb2_fop_release() in its file_operations, so replace video_unregister_device(&s->vdev) with vb2_video_unregister_device(&s->vdev) and move it before clearing s->udev. vb2_video_unregister_device() releases the vb2 queue, which synchronously runs airspy_stop_streaming() if streaming is active, so the URBs, coherent DMA stream buffers and the hardware stop control message all execute while s->udev is still valid. vb2_video_unregister_device() locks vdev->queue->lock (vb_queue_lock) internally, and stop_streaming() locks v4l2_lock, so the previous outer mutex_lock(&s->vb_queue_lock) / mutex_lock(&s->v4l2_lock) pair around the unregister sequence would self-deadlock and has been removed. A short v4l2_lock critical section around s->udev = NULL remains so any ioctl path that still holds the file descriptor sees coherent state. Issue identified by automated review of the INV-003 series at https://sashiko.dev/

In the Linux kernel, the following vulnerability has been resolved: media: cec: disable delayed work before freeing an interrupted transmit cec_transmit_msg_fh() drops adap->lock to wait for a blocking transmit in wait_for_completion_killable(). If that wait is interrupted by a signal, cancel_delayed_work_sync() can run before the CEC kthread arms the reply timeout via schedule_delayed_work(&data->work) in cec_transmit_done_ts(). The work is then armed after the cancel, and the data is freed with its delayed_work still pending: ODEBUG: free active (active state 0) object: ... hint: cec_wait_timeout Use disable_delayed_work_sync(): it cancels the work and disables it, so the later schedule_delayed_work() becomes a no-op and the work cannot be re-armed. The data is freed right after, so it need not be re-enabled.

In the Linux kernel, the following vulnerability has been resolved: media: cec: extron-da-hd-4k-plus: add sanity check Add check to prevent overflowing msg.msg[] in case the incoming data is malformed.

In the Linux kernel, the following vulnerability has been resolved: media: cec: Serialize exclusive follower delivery cec_receive_notify() reads the exclusive follower pointer without the adapter lock. Serialize the no-follower check and message delivery against mode changes and release.

In the Linux kernel, the following vulnerability has been resolved: media: cedrus: fix memory leak in cedrus_init_ctrls() In cedrus_init_ctrls(), the V4L2 control handler is initialized before allocating memory for ctx->ctrls. If this allocation fails, the function returns -ENOMEM without freeing the previously allocated handler resources, leading to a memory leak. Fix this by calling v4l2_ctrl_handler_free() on the ctx->ctrls allocation failure path. The bug was first flagged by an experimental analysis tool we are developing for kernel memory-management bugs while analyzing v6.13-rc1. The tool is still under development and is not yet publicly available. Manual inspection confirms that the bug is still present in v7.1.1. An x86_64 allyesconfig build showed no new warnings. As we do not have an Allwinner SoC or board with a Cedrus VPU available to test with, no runtime testing was able to be performed.

In the Linux kernel, the following vulnerability has been resolved: media: cobalt: Avoid freeing ALSA private data twice snd_cobalt_card_create() stores cobsc in sc->private_data and installs snd_cobalt_card_private_free() as sc->private_free. From that point, snd_card_free(sc) releases cobsc through the ALSA card cleanup path. If cobalt_alsa_init() fails after snd_cobalt_card_create(), the err_exit_free path calls snd_card_free(sc) and then kfree(cobsc). That second free releases the same object again. Remove the explicit kfree(cobsc) and leave ownership with the ALSA card. This issue was found by a static analysis checker and confirmed by manual source review.

In the Linux kernel, the following vulnerability has been resolved: media: cx231xx: reject geometry changes while the VBI queue is busy vidioc_s_fmt_vid_cap() and vidioc_s_std() change the device-wide dev->width / dev->norm but only refuse the change when the *video* queue (dev->vidq) is busy. The VBI queue (dev->vbiq) shares that same geometry: cx231xx_init_vbi_isoc() latches dma_q->lines_per_field from dev->norm, the VBI videobuf2 plane is sized from dev->width / dev->norm in vbi_queue_setup() and vbi_buf_prepare(), and cx231xx_do_vbi_copy() then recomputes the destination offset from the *live* dev->width and the latched lines_per_field on every URB completion: offset = lines_completed * (dev->width << 1) + ...; if (dma_q->current_field == 2) offset += dev->width * 2 * dma_q->lines_per_field; memcpy(plane + offset, p_buffer, lencopy); Because the VBI node shares video_ioctl_ops with the video node, an application can size a small VBI plane (REQBUFS/QBUF with a small width, or with the NTSC standard), then enlarge dev->width (or switch dev->norm to PAL) through the video node while the VBI stream is running -- the change is allowed because only dev->vidq is checked -- and let the device deliver a field-2 VBI payload. cx231xx_do_vbi_copy() now computes the offset with the larger geometry and memcpy()s past the end of the smaller plane that was already allocated, a heap out-of-bounds write whose offset is attacker-chosen and whose contents come from the device. The per-field guard in cx231xx_copy_vbi_line() does not help: it bounds the copy against the latched lines_per_field, not the plane's real capacity, and vb2 does not re-run buf_prepare() for an already prepared buffer. Refuse the format/standard change when the VBI queue is busy as well, so the geometry cannot change underneath an allocated VBI buffer.

In the Linux kernel, the following vulnerability has been resolved: media: cx23885: cancel NetUP CI work before teardown netup_ci_exit() frees a netup_ci_state while its work item, netup_read_ci_status(), may still be pending or running on the system workqueue. The worker obtains the state with container_of() and dereferences it, so it must not outlive the state. netup_ci_init() queues the initial status read, and CI GPIO interrupts subsequently queue the same work from netup_ci_slot_status(). During remove, cx23885_finidev() calls free_irq() before the CI device is unregistered. free_irq() prevents further IRQ handlers from running, but does not drain work queued previously, so the worker can run after netup_ci_exit() frees the state. Call cancel_work_sync() before dvb_ca_en50221_release() and kfree(). This issue was found by an in-house static analysis tool.

In the Linux kernel, the following vulnerability has been resolved: media: em28xx: defer audio-only extension registration The audio-only path registers extensions while probing the primary device. For a dual-TS board, this happens before dev_next is created. The duplicate device inherits is_audio_only and is then independently inserted into em28xx_devlist. The list is intended to contain only primary devices: extension operations reach the secondary device through dev_next. The independently linked secondary can be freed during disconnect while its list node remains reachable, resulting in a use-after-free. Defer audio-only extension registration to the module-request work item. It runs only after probing has completed construction of the optional secondary device, so only the primary is registered and extension callbacks reach the secondary through dev_next.

In the Linux kernel, the following vulnerability has been resolved: media: em28xx: fix use-after-free of dev_next->devlist on disconnect When a device with has_dual_ts=1 is probed and the is_audio_only path is taken, both dev and dev->dev_next are added to the global em28xx_devlist via em28xx_init_extension(). However, during disconnect, em28xx_close_extension(dev) only calls list_del(&dev->devlist), leaving dev->dev_next->devlist still linked in the global list. When dev_next is subsequently freed via kref_put(), its devlist entry becomes a dangling pointer in em28xx_devlist. The next device probe that calls em28xx_init_extension() triggers a list corruption BUG when list_add_tail detects the freed node. This bug was exposed by commit a368ecde8a50 ("USB: core: Fix duplicate endpoint bug by clearing reserved bits in the descriptor") which clears reserved bits in bEndpointAddress during endpoint parsing. This causes fuzzed endpoint addresses like 0xf3 to be normalized to 0x83, which em28xx interprets as a vendor audio endpoint, enabling the is_audio_only + has_dual_ts code path that was previously unreachable with such descriptors. Fix this by removing dev->dev_next->devlist from the global list in em28xx_close_extension() before the device is freed.

In the Linux kernel, the following vulnerability has been resolved: media: go7007: defer the ALSA v4l2 put until card release go7007_snd_init() already takes a v4l2_device reference for the ALSA side, but go7007_snd_remove() drops it immediately after calling snd_card_free_when_closed(). That is too early when a userspace process still has the capture PCM open. The ALSA card and its PCM callbacks remain alive until the last file is closed, so the release path can still reach struct go7007 through pcm->private_data and call go7007_snd_hw_free() after the V4L2 release path has freed the object. Move the matching v4l2_device_put() to the ALSA card private_free callback so the existing ALSA reference covers the whole deferred card lifetime.

In the Linux kernel, the following vulnerability has been resolved: media: i2c: imx415: Release runtime PM reference on VBLANK error The VBLANK path returned immediately when programming VMAX failed after pm_runtime_get_if_in_use() had taken a runtime PM reference. Break out of the switch instead so the common pm_runtime_put() path is used.

In the Linux kernel, the following vulnerability has been resolved: media: i2c: ov02a10: fix endpoint parsing use-after-free The ov02a10_check_hwcfg() function calls fwnode_handle_put(ep) immediately after allocating and parsing the endpoint. However, it subsequently calls fwnode_property_read_u32() using the same 'ep' handle, leading to a potential use-after-free. Additionally, reading the optional 'ovti,mipi-clock-voltage' property used to overwrite the 'ret' variable. If the property was missing, 'ret' would become negative, and this failure code would be incorrectly returned at the end of the function, causing probe to fail entirely. Fix the use-after-free by moving fwnode_property_read_u32() before the endpoint is parsed and freed. Avoid the error leak by not assigning the result of fwnode_property_read_u32() to 'ret'.

In the Linux kernel, the following vulnerability has been resolved: media: i2c: ov7740: fix use-after-destroy in remove The ov7740_remove() function had a severe teardown order bug where it destroyed the driver's mutex before freeing the V4L2 control handler which relies on that mutex, leading to a use-after-destroy kernel panic. Furthermore, the driver explicitly called v4l2_ctrl_handler_free() and mutex_destroy() sequentially, but then called ov7740_free_controls() which invokes both of them a second time, resulting in a double-free. This patch fixes the issue by unregistering the subdevice first, and relying exclusively on ov7740_free_controls() to safely tear down the mutex and control handler in the correct order.

In the Linux kernel, the following vulnerability has been resolved: media: platform: mtk-mdp3: Fix SCP device refcounting mdp_probe() first tries to get the SCP handle with scp_get(). When that fails, it falls back to looking up the SCP platform device with __get_pdev_by_id() and then reads its driver data. The fallback lookup returns the platform device with a reference, just like scp_get() does. However, the fallback path currently drops that reference immediately after platform_get_drvdata(). The driver later still calls scp_put(mdp->scp) unconditionally from the probe error path and from mdp_video_device_release(), which drops the SCP device reference again. Keep the fallback reference until the existing scp_put() call, so that the fallback path follows the same ownership rules as the scp_get() path.

In the Linux kernel, the following vulnerability has been resolved: media: platform: mtk-mdp3: fix NULL deref on failed SCP lookup Add the missing sanity check after looking up the SCP to avoid dereferencing a NULL-pointer in case its driver has not yet been bound.

In the Linux kernel, the following vulnerability has been resolved: media: rc: sunxi-cir: Unregister rc device on probe failure After rc_register_device() succeeds, later probe failures must undo the registration with rc_unregister_device(). The current error path jumps to the allocation cleanup label and only calls rc_free_device(), leaving the rc device registration and resources created by rc_register_device() behind. Add a registered-device unwind label for the IRQ lookup, IRQ request, and hardware initialization failure paths. Keep rc_free_device() for failures before rc_register_device() succeeds.

In the Linux kernel, the following vulnerability has been resolved: media: rkvdec: hevc: guard INTER_REF_PIC_SET_PRED index underflow st_ref_pic_set_prediction() computes the reference RPS index as st_rps_idx - (delta_idx_minus1 + 1) per HEVC spec equation 7-59. Both operands are u8, so when delta_idx_minus1 + 1 exceeds the current index the subtraction wraps and the subsequent array access at calculated_rps_st_sets[ref_rps_idx] reads far out of bounds. A userspace V4L2 client that can open the RKVDEC m2m decoder can submit an EXT_SPS_ST_RPS control with INTER_REF_PIC_SET_PRED set and delta_idx_minus1 crafted to trigger the underflow. Reject the entry early when the reference index would underflow.

In the Linux kernel, the following vulnerability has been resolved: media: rtl2832_sdr: use vb2_video_unregister_device() on remove to fix DMA leak rtl2832_sdr_remove() runs on USB disconnect and clears dev->udev to NULL before any pending streaming teardown has run. When user space later closes its file descriptor, vb2 calls rtl2832_sdr_stop_streaming() which in turn calls rtl2832_sdr_free_stream_bufs(). That helper releases each coherent buffer with: usb_free_coherent(dev->udev, dev->buf_size, dev->buf_list[dev->buf_num], dev->dma_addr[dev->buf_num]); usb_free_coherent() returns immediately when its dev argument is NULL, so every DMA stream buffer that was live at disconnect is silently leaked. The URBs allocated in rtl2832_sdr_alloc_urbs() outlive the device for the same reason. The rtl2832_sdr driver uses vb2_fop_release() in its file_operations, so replace video_unregister_device(&dev->vdev) with vb2_video_unregister_device(&dev->vdev) and move it before clearing dev->udev. vb2_video_unregister_device() releases the vb2 queue, which synchronously runs rtl2832_sdr_stop_streaming() if streaming is active, so URBs and coherent DMA stream buffers are freed while dev->udev is still valid. vb2_video_unregister_device() locks vdev->queue->lock (vb_queue_lock) internally, and stop_streaming() locks v4l2_lock, so the previous outer mutex_lock(&dev->vb_queue_lock) / mutex_lock(&dev->v4l2_lock) pair around the unregister sequence would self-deadlock and has been removed. A short v4l2_lock critical section around dev->udev = NULL remains so any ioctl path that still holds the file descriptor sees coherent state. Issue identified by automated review of the INV-003 series at https://sashiko.dev/

In the Linux kernel, the following vulnerability has been resolved: media: rtl2832_sdr: release URBs and stream buffers on start_streaming() failure rtl2832_sdr_start_streaming() calls rtl2832_sdr_alloc_stream_bufs(), rtl2832_sdr_alloc_urbs() and rtl2832_sdr_submit_urbs() in sequence and shares a single err: label that only unlocks the mutex and returns. When alloc_urbs() succeeds but submit_urbs() fails, or when alloc_urbs() itself returns -ENOMEM after alloc_stream_bufs() has already succeeded, the URBs and/or the coherent DMA stream buffers stay allocated while streaming reports failure to vb2. Two latent defects follow on the next VIDIOC_STREAMON: 1) rtl2832_sdr_alloc_stream_bufs() unconditionally resets dev->buf_num to 0 and overwrites dev->buf_list[]/dev->dma_addr[], permanently leaking the coherent DMA memory allocated by the previous attempt. 2) rtl2832_sdr_alloc_urbs() never resets dev->urbs_initialized and only increments it. After a second successful pass urbs_initialized can exceed MAX_BULK_BUFS, so the subsequent rtl2832_sdr_free_urbs() walks from urbs_initialized - 1 down to 0 and reads past the end of dev->urb_list[], passing garbage pointers to usb_free_urb(). Mirror the teardown that stop_streaming() already performs: on the error path call rtl2832_sdr_free_urbs() and rtl2832_sdr_free_stream_bufs() before unlocking. Both helpers are idempotent (free_urbs kills and zeros urbs_initialized; free_stream_bufs is gated on URB_BUF and clears the buf_num counter), so partial-failure paths and the no-allocation paths remain safe. Issue identified by automated review of the INV-003 series at https://sashiko.dev/

In the Linux kernel, the following vulnerability has been resolved: media: s2255: bound JPEG frame size before copying into the buffer s2255_fillbuff() memcpy()s vc->jpg_size bytes of a captured JPEG/MJPEG frame into the vb2 plane. vc->jpg_size is taken verbatim from the S2255_MARKER_FRAME header the device sends (pdword[4] in save_frame()) and, unlike the frame payload length just above it, is never bounded: payload = le32_to_cpu(pdword[3]); if (payload > vc->req_image_size) /* payload is checked ... */ return -EINVAL; vc->pkt_size = payload; vc->jpg_size = le32_to_cpu(pdword[4]); /* ... jpg_size is not */ A malicious or malfunctioning device can therefore report a jpg_size larger than the destination vb2 plane, and the memcpy() writes past it. jpg_size is a signed int, so a value with the top bit set also turns into a huge length. Reject a frame whose jpg_size is negative or exceeds the plane size before copying it.

In the Linux kernel, the following vulnerability has been resolved: media: s2255: check firmware size before reading trailing marker s2255_probe() reads a 4-byte marker and version from the last 8 bytes of the firmware blob (fw->data[fw_size - 8] and [fw_size - 4]). If the firmware file is shorter than 8 bytes, fw_size - 8 underflows and the access reads out of bounds. Validate the firmware size before indexing.