Information Disclosure

Use-after-free in Linux kernel KVM x86 MMIO emulation allows local authenticated users with low privileges to potentially execute arbitrary code, escalate privileges, or cause denial of service. The flaw occurs when KVM's emulator initiates MMIO writes using on-stack variables that cross page boundaries between two MMIO pages, creating dangling pointers when fragments are processed across separate KVM_RUN calls, especially when different tasks handle subsequent runs. EPSS exploitation probability is very low (0.02%, 5th percentile), and vendor patches are available for kernel versions 6.12.83, 6.18.24, 6.19.14, and 7.0.1. No active exploitation or public POC identified at time of analysis.

Use-after-free in Linux kernel q6apm audio driver allows local authenticated attackers with low privileges to achieve arbitrary code execution, denial of service, or information disclosure with high impact to confidentiality, integrity, and availability. The flaw affects Qualcomm ASoC q6apm component registration code used in devices like Lenovo 21N2ZC5PUS laptops. Vendor-released patches are available across multiple kernel version branches (6.12.83, 6.18.24, 6.19.14, 7.0.1). EPSS score of 0.02% (5th percentile) indicates low probability of mass exploitation despite high CVSS 7.8, with no confirmed active exploitation or public POC identified at time of analysis.

In the Linux kernel, the following vulnerability has been resolved: media: vidtv: fix nfeeds state corruption on start_streaming failure syzbot reported a memory leak in vidtv_psi_service_desc_init [1]. When vidtv_start_streaming() fails inside vidtv_start_feed(), the nfeeds counter is left incremented even though no feed was actually started. This corrupts the driver state: subsequent start_feed calls see nfeeds > 1 and skip starting the mux, while stop_feed calls eventually try to stop a non-existent stream. This state corruption can also lead to memory leaks, since the mux and channel resources may be partially allocated during a failed start_streaming but never cleaned up, as the stop path finds dvb->streaming == false and returns early. Fix by decrementing nfeeds back when start_streaming fails, keeping the counter in sync with the actual number of active feeds. [1] BUG: memory leak unreferenced object 0xffff888145b50820 (size 32): comm "syz.0.17", pid 6068, jiffies 4294944486 backtrace (crc 90a0c7d4): vidtv_psi_service_desc_init+0x74/0x1b0 drivers/media/test-drivers/vidtv/vidtv_psi.c:288 vidtv_channel_s302m_init+0xb1/0x2a0 drivers/media/test-drivers/vidtv/vidtv_channel.c:83 vidtv_channels_init+0x1b/0x40 drivers/media/test-drivers/vidtv/vidtv_channel.c:524 vidtv_mux_init+0x516/0xbe0 drivers/media/test-drivers/vidtv/vidtv_mux.c:518 vidtv_start_streaming drivers/media/test-drivers/vidtv/vidtv_bridge.c:194 [inline] vidtv_start_feed+0x33e/0x4d0 drivers/media/test-drivers/vidtv/vidtv_bridge.c:239

Use-after-free in Linux kernel MediaTek video encoder allows local authenticated users to corrupt memory and potentially execute arbitrary code. The flaw affects the vcodec driver's encoder release path where ctx memory is freed before canceling scheduled workqueue tasks, enabling race conditions between cleanup and worker threads that may dereference freed memory. KASAN-confirmed exploitation requires local access with low privileges (CVSS AV:L/PR:L). Patches available for kernel versions 6.12.83, 6.18.24, 6.19.14, and 7.0.1. EPSS score of 0.02% (5th percentile) indicates very low probability of automated exploitation, with no public exploit identified at time of analysis.

In the Linux kernel, the following vulnerability has been resolved: hwmon: (powerz) Fix use-after-free on USB disconnect After powerz_disconnect() frees the URB and releases the mutex, a subsequent powerz_read() call can acquire the mutex and call powerz_read_data(), which dereferences the freed URB pointer. Fix by: - Setting priv->urb to NULL in powerz_disconnect() so that powerz_read_data() can detect the disconnected state. - Adding a !priv->urb check at the start of powerz_read_data() to return -ENODEV on a disconnected device. - Moving usb_set_intfdata() before hwmon registration so the disconnect handler can always find the priv pointer.

In the Linux kernel, the following vulnerability has been resolved: ALSA: 6fire: fix use-after-free on disconnect In usb6fire_chip_abort(), the chip struct is allocated as the card's private data (via snd_card_new with sizeof(struct sfire_chip)). When snd_card_free_when_closed() is called and no file handles are open, the card and embedded chip are freed synchronously. The subsequent chip->card = NULL write then hits freed slab memory. Call trace: usb6fire_chip_abort sound/usb/6fire/chip.c:59 [inline] usb6fire_chip_disconnect+0x348/0x358 sound/usb/6fire/chip.c:182 usb_unbind_interface+0x1a8/0x88c drivers/usb/core/driver.c:458 ... hub_event+0x1a04/0x4518 drivers/usb/core/hub.c:5953 Fix by moving the card lifecycle out of usb6fire_chip_abort() and into usb6fire_chip_disconnect(). The card pointer is saved in a local before any teardown, snd_card_disconnect() is called first to prevent new opens, URBs are aborted while chip is still valid, and snd_card_free_when_closed() is called last so chip is never accessed after the card may be freed.

In the Linux kernel, the following vulnerability has been resolved: media: as102: fix to not free memory after the device is registered in as102_usb_probe() In as102_usb driver, the following race condition occurs: ``` CPU0 CPU1 as102_usb_probe() kzalloc(); // alloc as102_dev_t .... usb_register_dev(); fd = sys_open("/path/to/dev"); // open as102 fd .... usb_deregister_dev(); .... kfree(); // free as102_dev_t .... sys_close(fd); as102_release() // UAF!! as102_usb_release() kfree(); // DFB!! ``` When a USB character device registered with usb_register_dev() is later unregistered (via usb_deregister_dev() or disconnect), the device node is removed so new open() calls fail. However, file descriptors that are already open do not go away immediately: they remain valid until the last reference is dropped and the driver's .release() is invoked. In as102, as102_usb_probe() calls usb_register_dev() and then, on an error path, does usb_deregister_dev() and frees as102_dev_t right away. If userspace raced a successful open() before the deregistration, that open FD will later hit as102_release() --> as102_usb_release() and access or free as102_dev_t again, occur a race to use-after-free and double-free vuln. The fix is to never kfree(as102_dev_t) directly once usb_register_dev() has succeeded. After deregistration, defer freeing memory to .release(). In other words, let release() perform the last kfree when the final open FD is closed.

In the Linux kernel, the following vulnerability has been resolved: media: hackrf: fix to not free memory after the device is registered in hackrf_probe() In hackrf driver, the following race condition occurs: ``` CPU0 CPU1 hackrf_probe() kzalloc(); // alloc hackrf_dev .... v4l2_device_register(); .... fd = sys_open("/path/to/dev"); // open hackrf fd .... v4l2_device_unregister(); .... kfree(); // free hackrf_dev .... sys_ioctl(fd, ...); v4l2_ioctl(); video_is_registered() // UAF!! .... sys_close(fd); v4l2_release() // UAF!! hackrf_video_release() kfree(); // DFB!! ``` When a V4L2 or video device is unregistered, the device node is removed so new open() calls are blocked. However, file descriptors that are already open-and any in-flight I/O-do not terminate immediately; they remain valid until the last reference is dropped and the driver's release() is invoked. Therefore, freeing device memory on the error path after hackrf_probe() has registered dev it will lead to a race to use-after-free vuln, since those already-open handles haven't been released yet. And since release() free memory too, race to use-after-free and double-free vuln occur. To prevent this, if device is registered from probe(), it should be modified to free memory only through release() rather than calling kfree() directly.

In the Linux kernel, the following vulnerability has been resolved: mm/userfaultfd: fix hugetlb fault mutex hash calculation In mfill_atomic_hugetlb(), linear_page_index() is used to calculate the page index for hugetlb_fault_mutex_hash(). However, linear_page_index() returns the index in PAGE_SIZE units, while hugetlb_fault_mutex_hash() expects the index in huge page units. This mismatch means that different addresses within the same huge page can produce different hash values, leading to the use of different mutexes for the same huge page. This can cause races between faulting threads, which can corrupt the reservation map and trigger the BUG_ON in resv_map_release(). Fix this by introducing hugetlb_linear_page_index(), which returns the page index in huge page granularity, and using it in place of linear_page_index().

In the Linux kernel, the following vulnerability has been resolved: clockevents: Add missing resets of the next_event_forced flag The prevention mechanism against timer interrupt starvation missed to reset the next_event_forced flag in a couple of places: - When the clock event state changes. That can cause the flag to be stale over a shutdown/startup sequence - When a non-forced event is armed, which then prevents rearming before that event. If that event is far out in the future this will cause missed timer interrupts. - In the suspend wakeup handler. That led to stalls which have been reported by several people. Add the missing resets, which fixes the problems for the reporters.

In the Linux kernel, the following vulnerability has been resolved: media: verisilicon: Fix kernel panic due to __initconst misuse Fix a kernel panic when probing the driver as a module: Unable to handle kernel paging request at virtual address ffffd9c18eb05000 of_find_matching_node_and_match+0x5c/0x1a0 hantro_probe+0x2f4/0x7d0 [hantro_vpu] The imx8mq_vpu_shared_resources array is referenced by variant structures through their shared_devices field. When built as a module, __initconst causes this data to be freed after module init, but it's later accessed during probe, causing a page fault. The imx8mq_vpu_shared_resources is referenced from non-init code, so keeping __initconst or __initconst_or_module here is wrong. Drop the __initconst annotation and let it live in the normal .rodata section. A bug of __initconst called from regular non-init probe code leading to bugs during probe deferrals or during unbind-bind cycles.

In the Linux kernel, the following vulnerability has been resolved: i2c: designware: amdisp: Fix resume-probe race condition issue Identified resume-probe race condition in kernel v7.0 with the commit 38fa29b01a6a ("i2c: designware: Combine the init functions"),but this issue existed from the beginning though not detected. The amdisp i2c device requires ISP to be in power-on state for probe to succeed. To meet this requirement, this device is added to genpd to control ISP power using runtime PM. The pm_runtime_get_sync() called before i2c_dw_probe() triggers PM resume, which powers on ISP and also invokes the amdisp i2c runtime resume before the probe completes resulting in this race condition and a NULL dereferencing issue in v7.0 Fix this race condition by using the genpd APIs directly during probe: - Call dev_pm_genpd_resume() to Power ON ISP before probe - Call dev_pm_genpd_suspend() to Power OFF ISP after probe - Set the device to suspended state with pm_runtime_set_suspended() - Enable runtime PM only after the device is fully initialized

In the Linux kernel, the following vulnerability has been resolved: drm/i915: Unlink NV12 planes earlier unlink_nv12_plane() will clobber parts of the plane state potentially already set up by plane_atomic_check(), so we must make sure not to call the two in the wrong order. The problem happens when a plane previously selected as a Y plane is now configured as a normal plane by user space. plane_atomic_check() will first compute the proper plane state based on the userspace request, and unlink_nv12_plane() later clears some of the state. This used to work on account of unlink_nv12_plane() skipping the state clearing based on the plane visibility. But I removed that check, thinking it was an impossible situation. Now when that situation happens unlink_nv12_plane() will just WARN and proceed to clobber the state. Rather than reverting to the old way of doing things, I think it's more clear if we unlink the NV12 planes before we even compute the new plane state. (cherry picked from commit 017ecd04985573eeeb0745fa2c23896fb22ee0cc)

Out-of-bounds heap write in Linux kernel CAN gateway CRC8 checksum processing allows adjacent network attackers to corrupt kernel memory and potentially achieve code execution. The cgw_csum_crc8_rel() function in the CAN gateway subsystem uses raw negative index values instead of bounds-checked variables when accessing canfd_frame data, enabling writes up to 56 bytes before the heap object. Exploitation requires CAP_NET_ADMIN capability to configure CAN gateway CRC8 checksums. EPSS exploitation probability is very low (0.02%, 7th percentile) and no active exploitation has been reported. Vendor patches available across multiple kernel versions (5.10.253, 5.15.203, 6.1.168, 6.6.131, 6.12.80, 6.18.21, 6.19.11, 7.0).

Out-of-bounds memory access in Linux Kernel's KVM subsystem for LoongArch architecture allows local authenticated attackers with low privileges to read limited kernel memory and cause system crashes. The vulnerability stems from improper handling of empty EIOINTC coremap values in eiointc_update_sw_coremap(), resulting in invalid array indexing into kvm_arch::phyid_map::phys_map[]. While CVSS rates this 7.3 HIGH, the EPSS score of 0.02% (4th percentile) indicates minimal real-world exploitation activity. No active exploitation (not in CISA KEV) or public POC has been identified. Vendor patches are available across multiple stable kernel branches (6.18.21, 6.19.11, 7.0, and mainline).

In the Linux kernel, the following vulnerability has been resolved: s390/mm: Add missing secure storage access fixups for donated memory There are special cases where secure storage access exceptions happen in a kernel context for pages that don't have the PG_arch_1 bit set. That bit is set for non-exported guest secure storage (memory) but is absent on storage donated to the Ultravisor since the kernel isn't allowed to export donated pages. Prior to this patch we would try to export the page by calling arch_make_folio_accessible() which would instantly return since the arch bit is absent signifying that the page was already exported and no further action is necessary. This leads to secure storage access exception loops which can never be resolved. With this patch we unconditionally try to export and if that fails we fixup.

In the Linux kernel, the following vulnerability has been resolved: PM: sleep: Drop spurious WARN_ON() from pm_restore_gfp_mask() Commit 35e4a69b2003f ("PM: sleep: Allow pm_restrict_gfp_mask() stacking") introduced refcount-based GFP mask management that warns when pm_restore_gfp_mask() is called with saved_gfp_count == 0. Some hibernation paths call pm_restore_gfp_mask() defensively where the GFP mask may or may not be restricted depending on the execution path. For example, the uswsusp interface invokes it in SNAPSHOT_CREATE_IMAGE, SNAPSHOT_UNFREEZE, and snapshot_release(). Before the stacking change this was a silent no-op; it now triggers a spurious WARNING. Remove the WARN_ON() wrapper from the !saved_gfp_count check while retaining the check itself, so that defensive calls remain harmless without producing false warnings. [ rjw: Subject tweak ]

Use-after-free in Linux kernel AMD GPU driver allows local authenticated users to potentially execute arbitrary code, escalate privileges, or cause denial of service. The amdgpu_amdkfd_submit_ib() function in the AMD KFD (Kernel Fusion Driver) prematurely releases a DMA fence reference before waiting on it, creating a race condition where the fence memory may be freed before use. Vendor-released patches are available for multiple stable kernel branches (6.1.168, 6.6.131, 6.12.80, 6.18.21, 6.19.11, 7.0). EPSS exploitation probability is very low at 0.02% (7th percentile), and no public exploit or active exploitation has been identified at time of analysis.

In the Linux kernel, the following vulnerability has been resolved: RDMA/irdma: Fix deadlock during netdev reset with active connections Resolve deadlock that occurs when user executes netdev reset while RDMA applications (e.g., rping) are active. The netdev reset causes ice driver to remove irdma auxiliary driver, triggering device_delete and subsequent client removal. During client removal, uverbs_client waits for QP reference count to reach zero while cma_client holds the final reference, creating circular dependency and indefinite wait in iWARP mode. Skip QP reference count wait during device reset to prevent deadlock.

In the Linux kernel, the following vulnerability has been resolved: LoongArch: KVM: Fix base address calculation in kvm_eiointc_regs_access() In function kvm_eiointc_regs_access(), the register base address is caculated from array base address plus offset, the offset is absolute value from the base address. The data type of array base address is u64, it should be converted into the "void *" type and then plus the offset.

Memory management flaw in Linux kernel's Cadence macb network driver causes kernel warning and potential denial of service. Specifically affects the macb Ethernet driver on ARM64 ZynqMP platforms (kernel versions 6.1+ containing commit 6bc8a5098bf4). The vulnerability stems from calling napi_consume_skb() with IRQs disabled during TX packet cleanup, violating kernel API contracts and potentially causing system instability under network load. EPSS exploitation probability is very low (0.02%, 7th percentile) with vendor-released patches available across all stable kernel branches (6.1.168, 6.6.131, 6.12.80, 6.18.21, 6.19.11, 7.0). No active exploitation or public exploit code identified at time of analysis.

Out-of-bounds array access in Linux kernel KVM subsystem on LoongArch allows local authenticated attackers with low privileges to execute arbitrary code, escalate privileges, or cause denial of service by passing negative cpuid values to kvm_get_vcpu_by_cpuid(). The function lacks bounds checking before indexing phyid_map::phys_map[], enabling read/write beyond array boundaries with container escape potential (CVSS scope change). Vendor patches available across multiple stable kernel branches (6.12.80, 6.18.21, 6.19.11). EPSS score of 0.02% indicates low automated exploitation likelihood, with no confirmed active exploitation or public POC at time of analysis.

A workqueue deadlock in Linux kernel NVMe-over-Fabrics target (nvmet) allows remote denial of service via recursive locking during controller disconnect. The nvmet subsystem's async event handler can trigger reentrant workqueue completion when nvmet_ctrl_free() flushes work on the same queue (nvmet-wq) that invoked it, causing a lockdep-detected recursive lock scenario. EPSS score of 0.02% indicates very low probability of exploitation in the wild. Patches available for kernel versions 6.12.80, 6.18.21, 6.19.11, and mainline 7.0 via upstream commits.

In the Linux kernel, the following vulnerability has been resolved: xfs: scrub: unlock dquot before early return in quota scrub xchk_quota_item can return early after calling xchk_fblock_process_error. When that helper returns false, the function returned immediately without dropping dq->q_qlock, which can leave the dquot lock held and risk lock leaks or deadlocks in later quota operations. Fix this by unlocking dq->q_qlock before the early return.

In the Linux kernel, the following vulnerability has been resolved: futex: Clear stale exiting pointer in futex_lock_pi() retry path Fuzzying/stressing futexes triggered: WARNING: kernel/futex/core.c:825 at wait_for_owner_exiting+0x7a/0x80, CPU#11: futex_lock_pi_s/524 When futex_lock_pi_atomic() sees the owner is exiting, it returns -EBUSY and stores a refcounted task pointer in 'exiting'. After wait_for_owner_exiting() consumes that reference, the local pointer is never reset to nil. Upon a retry, if futex_lock_pi_atomic() returns a different error, the bogus pointer is passed to wait_for_owner_exiting(). CPU0 CPU1 CPU2 futex_lock_pi(uaddr) // acquires the PI futex exit() futex_cleanup_begin() futex_state = EXITING; futex_lock_pi(uaddr) futex_lock_pi_atomic() attach_to_pi_owner() // observes EXITING *exiting = owner; // takes ref return -EBUSY wait_for_owner_exiting(-EBUSY, owner) put_task_struct(); // drops ref // exiting still points to owner goto retry; futex_lock_pi_atomic() lock_pi_update_atomic() cmpxchg(uaddr) *uaddr ^= WAITERS // whatever // value changed return -EAGAIN; wait_for_owner_exiting(-EAGAIN, exiting) // stale WARN_ON_ONCE(exiting) Fix this by resetting upon retry, essentially aligning it with requeue_pi.

Use-after-free in Linux kernel futex subsystem allows local authenticated attackers to achieve code execution, privilege escalation, or denial of service via sys_futex_requeue() with mismatched flags. Discovered through automated LLM analysis by Nicholas, this affects kernel versions 6.7 through 6.19.x, with patches available in 6.12.80, 6.18.21, 6.19.11, and 7.0. EPSS score of 0.02% (5th percentile) indicates low observed exploitation probability, and no active exploitation or public POC has been identified. The vulnerability requires local access with low-privilege authenticated user credentials (PR:L), making it a post-compromise escalation vector rather than a remote entry point.

Address calculation error in Linux kernel KVM on ARM64 allows local authenticated attackers with low privileges to corrupt memory descriptors, potentially enabling container escape or privilege escalation to compromise host integrity and confidentiality. The vulnerability affects KVM's stage-1/stage-2 page table descriptor swapping logic where pointer arithmetic incorrectly multiplies the offset by 8, causing writes to unintended memory locations. Vendor patches available for Linux 6.19.11 and mainline with EPSS exploitation probability at 5th percentile, indicating low observed exploitation despite high CVSS severity.

In the Linux kernel, the following vulnerability has been resolved: wifi: mac80211: Fix static_branch_dec() underflow for aql_disable. syzbot reported static_branch_dec() underflow in aql_enable_write(). [0] The problem is that aql_enable_write() does not serialise concurrent write()s to the debugfs. aql_enable_write() checks static_key_false(&aql_disable.key) and later calls static_branch_inc() or static_branch_dec(), but the state may change between the two calls. aql_disable does not need to track inc/dec. Let's use static_branch_enable() and static_branch_disable(). [0]: val == 0 WARNING: kernel/jump_label.c:311 at __static_key_slow_dec_cpuslocked.part.0+0x107/0x120 kernel/jump_label.c:311, CPU#0: syz.1.3155/20288 Modules linked in: CPU: 0 UID: 0 PID: 20288 Comm: syz.1.3155 Tainted: G U L syzkaller #0 PREEMPT(full) Tainted: [U]=USER, [L]=SOFTLOCKUP Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/24/2026 RIP: 0010:__static_key_slow_dec_cpuslocked.part.0+0x107/0x120 kernel/jump_label.c:311 Code: f2 c9 ff 5b 5d c3 cc cc cc cc e8 54 f2 c9 ff 48 89 df e8 ac f9 ff ff eb ad e8 45 f2 c9 ff 90 0f 0b 90 eb a2 e8 3a f2 c9 ff 90 <0f> 0b 90 eb 97 48 89 df e8 5c 4b 33 00 e9 36 ff ff ff 0f 1f 80 00 RSP: 0018:ffffc9000b9f7c10 EFLAGS: 00010293 RAX: 0000000000000000 RBX: ffffffff9b3e5d40 RCX: ffffffff823c57b4 RDX: ffff8880285a0000 RSI: ffffffff823c5846 RDI: ffff8880285a0000 RBP: 0000000000000000 R08: 0000000000000005 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000000 R12: 000000000000000a R13: 1ffff9200173ef88 R14: 0000000000000001 R15: ffffc9000b9f7e98 FS: 00007f530dd726c0(0000) GS:ffff8881245e3000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000200000001140 CR3: 000000007cc4a000 CR4: 00000000003526f0 Call Trace: <TASK> __static_key_slow_dec_cpuslocked kernel/jump_label.c:297 [inline] __static_key_slow_dec kernel/jump_label.c:321 [inline] static_key_slow_dec+0x7c/0xc0 kernel/jump_label.c:336 aql_enable_write+0x2b2/0x310 net/mac80211/debugfs.c:343 short_proxy_write+0x133/0x1a0 fs/debugfs/file.c:383 vfs_write+0x2aa/0x1070 fs/read_write.c:684 ksys_pwrite64 fs/read_write.c:793 [inline] __do_sys_pwrite64 fs/read_write.c:801 [inline] __se_sys_pwrite64 fs/read_write.c:798 [inline] __x64_sys_pwrite64+0x1eb/0x250 fs/read_write.c:798 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline] do_syscall_64+0xc9/0xf80 arch/x86/entry/syscall_64.c:94 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7f530cf9aeb9 Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 e8 ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007f530dd72028 EFLAGS: 00000246 ORIG_RAX: 0000000000000012 RAX: ffffffffffffffda RBX: 00007f530d215fa0 RCX: 00007f530cf9aeb9 RDX: 0000000000000003 RSI: 0000000000000000 RDI: 0000000000000010 RBP: 00007f530d008c1f R08: 0000000000000000 R09: 0000000000000000 R10: 4200000000000005 R11: 0000000000000246 R12: 0000000000000000 R13: 00007f530d216038 R14: 00007f530d215fa0 R15: 00007ffde89fb978 </TASK>

In the Linux kernel, the following vulnerability has been resolved: pmdomain: bcm: bcm2835-power: Increase ASB control timeout The bcm2835_asb_control() function uses a tight polling loop to wait for the ASB bridge to acknowledge a request. During intensive workloads, this handshake intermittently fails for V3D's master ASB on BCM2711, resulting in "Failed to disable ASB master for v3d" errors during runtime PM suspend. As a consequence, the failed power-off leaves V3D in a broken state, leading to bus faults or system hangs on later accesses. As the timeout is insufficient in some scenarios, increase the polling timeout from 1us to 5us, which is still negligible in the context of a power domain transition. Also, replace the open-coded ktime_get_ns()/ cpu_relax() polling loop with readl_poll_timeout_atomic().

In the Linux kernel, the following vulnerability has been resolved: drm/xe: Fix missing runtime PM reference in ccs_mode_store ccs_mode_store() calls xe_gt_reset() which internally invokes xe_pm_runtime_get_noresume(). That function requires the caller to already hold an outer runtime PM reference and warns if none is held: [46.891177] xe 0000:03:00.0: [drm] Missing outer runtime PM protection [46.891178] WARNING: drivers/gpu/drm/xe/xe_pm.c:885 at xe_pm_runtime_get_noresume+0x8b/0xc0 Fix this by protecting xe_gt_reset() with the scope-based guard(xe_pm_runtime)(xe), which is the preferred form when the reference lifetime matches a single scope. v2: - Use scope-based guard(xe_pm_runtime)(xe) (Shuicheng) - Update commit message accordingly (cherry picked from commit 7937ea733f79b3f25e802a0c8360bf7423856f36)

In the Linux kernel, the following vulnerability has been resolved: NFC: nxp-nci: allow GPIOs to sleep Allow the firmware and enable GPIOs to sleep. This fixes a `WARN_ON' and allows the driver to operate GPIOs which are connected to I2C GPIO expanders. -- >8 -- kernel: WARNING: CPU: 3 PID: 2636 at drivers/gpio/gpiolib.c:3880 gpiod_set_value+0x88/0x98 -- >8 --

Denial of service in Linux kernel x86/platform/uv module when sockets are deconfigured, causing kernel panic during UV hub info structure allocation on systems with SGI UV architecture. Affects authenticated local attackers with standard user privileges. Vendor-released patches available for multiple kernel versions (6.6.130, 6.12.78, 6.18.20, 6.19.10 and others). EPSS score of 0.02% indicates exploitation is unlikely in typical environments despite low CVSS score barrier.

Use-after-free in Linux kernel tracing subsystem allows local authenticated attackers to achieve arbitrary code execution, privilege escalation, or denial of service. The vulnerability occurs when deleting tracing instances with copy_trace_marker enabled, where improper RCU synchronization leaves freed memory accessible. Exploitation requires local access with low privileges to manipulate kernel tracing facilities. EPSS score of 0.02% (4th percentile) indicates low observed exploitation probability. Vendor patches available across multiple stable kernel versions (6.18.20, 6.19.10, 7.0).

Race condition in Linux kernel SMB Direct receive credit management allows remote denial of service against SMB3 network storage services. The flaw enables remote unauthenticated attackers to exhaust receive buffer credits through timing exploitation of the gap between hardware packet reception and completion processing, causing service disruption. EPSS exploitation probability is low (0.02%, 4th percentile), and patches are available from kernel.org for versions 6.18.x, 6.19.x, and 7.0. This affects only systems using SMB Direct (RDMA-enabled SMB3), not standard SMB implementations.

Denial of service in Linux kernel SMB server (ksmbd) affects versions 6.18 through 7.0-rc via race condition in SMBDirect receive credit management. Remote unauthenticated attackers can trigger resource exhaustion through crafted SMB packets exploiting the window between hardware reception and completion processing. Vendor patches released for stable branches 6.18.11, 6.19.1, and mainline 7.0. Low EPSS score (0.02%) indicates limited exploitation interest despite network attack vector and no authentication requirement.

Denial of service in Linux kernel SMB server implementation allows local authenticated users to crash the system by triggering data stream corruption through improper credit management in smbdirect socket operations. The vulnerability affects kernel versions prior to 6.18.11, 6.19.1, and 7.0, and requires local access with limited privileges to exploit.

Use-after-free in Linux kernel SMB server (ksmbd) RDMA handling allows remote unauthenticated attackers to execute arbitrary code, escalate privileges, or crash the system via crafted SMB Direct connections. The vulnerability arises when batched RDMA send operations without IB_SEND_SIGNALED flags are prematurely freed during connection failures, causing memory corruption. Vendor patches are available for kernel versions 6.18.11, 6.19.1, and 7.0. EPSS score of 0.02% suggests low observed exploitation probability, and no active exploitation or public POC is confirmed at time of analysis, though the critical CVSS score (9.8) reflects severe potential impact if the SMB Direct feature is enabled.

A race condition in the Linux kernel SMB client's recv_io credit management allows local authenticated users to cause a denial of service through timing-sensitive credit accounting between incoming data reception and completion processing. The vulnerability affects SMBDirect socket credit handling where credits may be granted to peers before corresponding recv buffers are actually posted, creating a window where credit accounting becomes inconsistent. Exploitation requires local access and moderate complexity but is not confirmed as actively exploited (not listed in CISA KEV).

In the Linux kernel, the following vulnerability has been resolved: smb: client: let send_done handle a completion without IB_SEND_SIGNALED With smbdirect_send_batch processing we likely have requests without IB_SEND_SIGNALED, which will be destroyed in the final request that has IB_SEND_SIGNALED set. If the connection is broken all requests are signaled even without explicit IB_SEND_SIGNALED.

Apache Airflow versions prior to 3.2.1 allow authenticated users with read access to at least one directed acyclic graph (DAG) to enumerate and discover the names and existence of all other DAGs and assets in the deployment, regardless of their assigned permissions. This information disclosure vulnerability enables privilege escalation reconnaissance by revealing the complete asset topology to users with limited scope authorization. The vulnerability requires valid user credentials but no elevated privileges, and has no known public exploit code at time of analysis.

Apache Airflow versions prior to 3.2.1 fail to enforce per-DAG access control on the /ui/dags endpoint, allowing authenticated users with read access to at least one DAG to retrieve Human-in-the-Loop prompts and full TaskInstance details for DAGs outside their authorized scope. This information disclosure bypasses the intended per-DAG RBAC boundary, exposing operator parameters and task context data to all authenticated users regardless of their assigned DAG permissions.

Privilege escalation in Azure IoT Central enables authenticated attackers to gain unauthorized access to sensitive information and elevate their permissions across tenant boundaries. An attacker with low-privilege credentials can exploit exposed sensitive data over the network to compromise confidentiality, integrity, and availability of other tenant resources. Microsoft has published security guidance, but no independent confirmation of patch availability exists at time of analysis.

Heap over-read in Open Virtual Network (OVN) DHCPv6 client ID processing allows remote unauthenticated attackers to extract sensitive memory contents across network boundaries. The vulnerability affects OVN's DHCPv6 implementation and carries a CVSS score of 8.6 with scope change, enabling cross-tenant information disclosure in multi-tenant virtualized environments. Public advisory released via oss-security mailing list on 2026-04-20, though no confirmed active exploitation or public POC identified at time of analysis.

Heap over-read in OVN's ICMP error response generation allows remote attackers to leak sensitive memory contents, causing information disclosure and potential denial of service. The vulnerability affects OVN versions prior to the 2026 security update, exploitable over the network without authentication or user interaction via crafted ICMP packets. No public exploit code has been identified, but the attack vector is network-accessible with high complexity requirements.

Out-of-bounds read in libXpm's xpmNextWord() parser function can be triggered by a local attacker with low privileges, crashing any X11 application that processes a maliciously crafted XPM image file, resulting in a denial of service. The vulnerability was disclosed via the oss-security mailing list on 2026-04-21 by X.Org and is tracked under CWE-125. No public exploit code or CISA KEV listing has been identified at time of analysis, and the EPSS score was not provided in available intelligence.

Liaison Site Prober plugin for WordPress allows unauthenticated attackers to retrieve sensitive audit log data including IP addresses, user IDs, usernames, and login events through an improperly secured REST API endpoint (/wp-json/site-prober/v1/logs) in all versions up to 1.2.1. The vulnerability stems from a permission callback that unconditionally returns true without validating user capabilities, enabling information disclosure with network-level access and no authentication required. No public exploit code or active exploitation has been identified at time of analysis.

HubSpot All-In-One Marketing plugin for WordPress (versions up to 11.3.32) exposes sensitive information via the class-adminconstants.php file, allowing authenticated users with Contributor-level access or higher to retrieve a complete list of installed plugins and their versions. This information disclosure enables reconnaissance for follow-on attacks targeting vulnerable plugins, though exploitation requires valid WordPress authentication and contributor-level privileges.

Remote code execution in Delta Electronics AS320T industrial automation server allows unauthenticated network attackers to trigger memory corruption via malformed GET/PUT requests to the web service. The incorrect buffer size calculation (CWE-131) enables stack-based overflow attacks against network-exposed management interfaces. With CVSS 9.8 (AV:N/AC:L/PR:N/UI:N) indicating trivial exploitation conditions and CRITICAL severity, this vulnerability represents an immediate risk to industrial control systems deploying this Delta OT product, though no public exploit or active exploitation confirmed at time of analysis.

Kyverno's apiCall feature automatically attaches the admission controller's ServiceAccount token to HTTP requests without validating the destination URL, enabling authenticated attackers to exfiltrate tokens to attacker-controlled servers and achieve full cluster compromise through webhook configuration tampering. Affects Kyverno versions prior to 1.18.0-rc1, 1.17.2-rc1, and 1.16.4. Vendor-released patches available across all three affected version branches. EPSS data not provided, but the vulnerability enables privilege escalation from low-privilege Kubernetes user to cluster admin via token theft, representing critical risk in multi-tenant environments.

Out-of-bounds read and write in OP-TEE OS PKCS#11 Trusted Application (versions 3.13.0-4.10.0) allows authenticated local attackers with low privileges to read up to 7 bytes beyond heap boundaries and write arbitrary attribute values outside allocated buffers, potentially compromising the integrity and confidentiality of the Trusted Execution Environment. The vulnerability affects Arm TrustZone-based TEE implementations running alongside Linux kernels on Cortex-A cores. Patches available in three upstream commits targeting version 4.11.0. EPSS data not provided; no CISA KEV status indicating targeted rather than widespread exploitation. CVSS 8.7 reflects high confidentiality/integrity impact with scope change, representing potential TEE compromise from the normal world.

Improper session lifetime enforcement in SenseLive X3050's web management interface allows attackers with access to a previously authenticated session to maintain administrative access without re-authentication, potentially enabling unauthorized configuration changes or information disclosure. The vulnerability affects the product's session management mechanism, permitting extended session validity beyond legitimate user activity windows. CVSS 6.9 indicates moderate risk; exploitation requires prior session compromise but no special configuration.

Information disclosure in Canon production printers and office/small office multifunction printers allows authenticated administrators to access sensitive device information through crafted requests to the browser-based remote management interface. The vulnerability affects multiple printer models and requires high-privilege administrative access; no active exploitation has been confirmed at time of analysis, though the remote network vector and low attack complexity indicate practical exploitability by privileged internal users.

SenseLive X3050 web management interface transmits all administrative communication including authentication credentials and configuration data over unencrypted HTTP, allowing network-adjacent attackers to intercept sensitive operational information without authentication or user interaction. The vulnerability affects all versions of the X3050 and is classified as information disclosure with confirmed CISA ICS advisory coverage.

Authentication bypass in SenseLive X3050 web management interface allows remote unauthenticated attackers to gain administrative access using default or previously-set credentials. After factory restore via SenseLive Config 2.0 tool, password updates fail to propagate correctly - the interface falsely reports success while the backend continues accepting old credentials. CISA ICS-CERT has issued an advisory (ICSA-26-111-12), indicating this affects industrial control system deployments. With CVSS 9.3 (AV:N/AC:L/PR:N) and CWE-522 (Insufficiently Protected Credentials), this represents critical risk for remotely accessible devices where administrators believe credentials have been changed but remain exploitable.

OpenClaw before 2026.4.2 contains an approval integrity vulnerability in pnpm dlx that fails to bind local script operands consistently with pnpm exec flows. Attackers can replace approved local scripts before execution without invalidating the approval plan, allowing execution of modified script contents.

OpenClaw before 2026.3.31 contains an environment variable leakage vulnerability in SSH-based sandbox backends that pass unsanitized process.env to child processes. Attackers can exploit this by leveraging non-default SSH environment forwarding configurations to leak sensitive environment variables from parent processes to SSH child processes.

OpenClaw before 2026.4.2 contains an insufficient scope vulnerability in Zalo webhook replay dedupe keys that allows legitimate events from different conversations or senders to collide. Attackers can exploit weak deduplication scoping to cause silent message suppression and disrupt bot workflows across chat sessions.

OpenClaw before 2026.3.31 contains a credential exposure vulnerability in media download functionality that forwards Authorization headers across cross-origin redirects. Attackers can exploit this by crafting malicious cross-origin redirect chains to intercept sensitive authorization credentials intended for legitimate requests.

OpenClaw before 2026.3.31 lacks a shared pre-auth concurrency budget on the public LINE webhook path, allowing attackers to cause transient availability loss. Remote attackers can flood the webhook endpoint with concurrent requests before signature verification to exhaust resources and degrade service availability.

OpenClaw before 2026.4.2 exposes configPath and stateDir metadata in Gateway connect success snapshots to non-admin authenticated clients. Non-admin clients can recover host-specific filesystem paths and deployment details, enabling host fingerprinting and facilitating chained attacks.

OpenClaw before 2026.3.31 contains a callback origin mutation vulnerability in Plivo voice-call replay that allows attackers to mutate in-process callback origin before replay rejection. Attackers with captured valid callbacks for live calls can exploit this to manipulate callback origins during the replay process.

OpenClaw before 2026.3.31 contains an information disclosure vulnerability in the Control Interface bootstrap JSON that exposes version and assistant agent identifiers. Attackers can extract sensitive fingerprinting information from the Control UI bootstrap payload to identify system versions and agent configurations.

OpenClaw before 2026.3.28 contains an environment variable sanitization vulnerability where GIT_TEMPLATE_DIR and AWS_CONFIG_FILE are not blocked in the host-env blocklist. Attackers can exploit approved exec requests to redirect git or AWS CLI behavior through attacker-controlled configuration files to execute untrusted code or load malicious credentials.

A request smuggling vulnerability exists in libsoup's HTTP/1 header parsing logic. The soup_message_headers_append_common() function in libsoup/soup-message-headers.c unconditionally appends each header value without validating for duplicate or conflicting Content-Length fields. This allows an attacker to send HTTP requests containing multiple Content-Length headers with differing values.

Flowise is a drag & drop user interface to build a customized large language model flow. Prior to 3.1.0, the GET /api/v1/public-chatflows/:id endpoint returns the full chatflow object without sanitization for public chatflows. Docker validation revealed this is worse than initially assessed: the sanitizeFlowDataForPublicEndpoint function does NOT exist in the released v3.0.13 Docker image. Both public-chatflows AND public-chatbotConfig return completely raw flowData including credential IDs, plaintext API keys, and password-type fields. This vulnerability is fixed in 3.1.0.

A weakness in SpiceJet’s public booking retrieval page permits full passenger booking details to be accessed using only a PNR and last name, with no authentication or verification mechanisms. This results in exposure of extensive personal, travel, and booking metadata to any unauthenticated user who can obtain or guess those basic inputs. The issue arises from improper access control on a sensitive data retrieval function.

Mastodon is a free, open-source social network server based on ActivityPub. Prior to v4.5.9, v4.4.16, and v4.3.22, Mastodon allows restricting new user sign-up based on e-mail domain names, and performs basic validation on e-mail addresses, but fails to restrict characters that are interpreted differently by some mailing servers. This vulnerability is fixed in v4.5.9, v4.4.16, and v4.3.22.

Flowise is a drag & drop user interface to build a customized large language model flow. Prior to 3.1.0, /api/v1/public-chatbotConfig/:id ep exposes sensitive data including API keys, HTTP authorization headers and internal configuration without any authentication. An attacker with knowledge just of a chatflow UUID can retrieve credentials stored in password type fields and HTTP headers, leading to credential theft and more. This vulnerability is fixed in 3.1.0.

@node-oauth/oauth2-server is a module for implementing an OAuth2 server in Node.js. The token exchange path accepts RFC7636-invalid code_verifier values (including one-character strings) for S256 PKCE flows. Because short/weak verifiers are accepted and failed verifier attempts do not consume the authorization code, an attacker who intercepts an authorization code can brute-force code_verifier guesses online until token issuance succeeds.

OpenTelemetry dotnet is a dotnet telemetry framework. From 1.13.1 to before 1.15.2, When exporting telemetry to a back-end/collector over gRPC or HTTP using OpenTelemetry Protocol format (OTLP), if the request results in a unsuccessful request (i.e. HTTP 4xx or 5xx), the response is read into memory with no upper-bound on the number of bytes consumed. This could cause memory exhaustion in the consuming application if the configured back-end/collector endpoint is attacker-controlled (or a network attacker can MitM the connection) and an extremely large body is returned by the response. This vulnerability is fixed in 1.15.2.

Race in GPU in Google Chrome on Windows prior to 147.0.7727.117 allowed a remote attacker to potentially perform a sandbox escape via a crafted video file. (Chromium security severity: Medium)

Out of bounds read in GPU in Google Chrome on Android prior to 147.0.7727.117 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

In the Linux kernel, the following vulnerability has been resolved: net/tls: fix use-after-free in -EBUSY error path of tls_do_encryption The -EBUSY handling in tls_do_encryption(), introduced by commit 859054147318 ("net: tls: handle backlogging of crypto requests"), has a use-after-free due to double cleanup of encrypt_pending and the scatterlist entry. When crypto_aead_encrypt() returns -EBUSY, the request is enqueued to the cryptd backlog and the async callback tls_encrypt_done() will be invoked upon completion. That callback unconditionally restores the scatterlist entry (sge->offset, sge->length) and decrements ctx->encrypt_pending. However, if tls_encrypt_async_wait() returns an error, the synchronous error path in tls_do_encryption() performs the same cleanup again, double-decrementing encrypt_pending and double-restoring the scatterlist. The double-decrement corrupts the encrypt_pending sentinel (initialized to 1), making tls_encrypt_async_wait() permanently skip the wait for pending async callbacks. A subsequent sendmsg can then free the tls_rec via bpf_exec_tx_verdict() while a cryptd callback is still pending, resulting in a use-after-free when the callback fires on the freed record. Fix this by skipping the synchronous cleanup when the -EBUSY async wait returns an error, since the callback has already handled encrypt_pending and sge restoration.

A flaw was found in the X.Org X server's XKB key types request validation. A local attacker could send a specially crafted request to the X server, leading to an out-of-bounds memory access vulnerability. This could result in the disclosure of sensitive information or cause the server to crash, leading to a Denial of Service (DoS). In certain configurations, higher impact outcomes may be possible.

Cache poisoning in @astrojs/node versions 9.4.4 and earlier allows unauthenticated remote attackers to poison CDN caches by sending malformed if-match headers to static asset endpoints, causing the server to return 500 errors with immutable one-year cache directives instead of the correct 412 Precondition Failed response. This vulnerability affects all subsequent requests to poisoned assets until the cache expires, breaking application functionality for legitimate users. The vulnerability is not actively exploited in the wild, but proof-of-concept exploitation is straightforward and requires only a single crafted HTTP request.

n8n-mcp v2.47.10 and earlier in HTTP transport mode logs sensitive authentication credentials and request metadata regardless of authentication outcome, allowing disclosure of bearer tokens, API keys, and JSON-RPC payloads to any system with access to server logs. While access control correctly rejects unauthenticated requests with 401 responses, the sensitive data from those rejected requests is persisted in logs before authentication is enforced, creating an information disclosure vulnerability (CWE-532) with CVSS 5.3 (low confidentiality impact). No public exploit code or active exploitation is documented; patch is available in v2.47.11.

An unauthenticated remote attacker is able to exhaust all available TCP connections in the CODESYS EtherNet/IP adapter stack, preventing legitimate clients from establishing new connections.

Uninitialized variable usage in OpenSC's libopensc library enables information disclosure and denial of service when processing specially crafted responses from malicious USB devices or smart cards. Attackers must physically present a crafted USB or smart card device to trigger the vulnerability, which reads uninitialized memory from the stack or heap, potentially exposing sensitive data or causing application crashes. No public exploit code has been identified at time of analysis.

Use-after-free in Linux kernel CAN raw socket implementation allows local authenticated attackers to corrupt memory and potentially achieve code execution. The vulnerability stems from premature deallocation of percpu uniq storage in raw_release() while raw_rcv() may still access it via deferred RCU callbacks. Patches available for kernel versions 6.12.83, 6.18.24, 6.19.14, and 7.0.1. EPSS exploitation probability remains low (0.02%, 5th percentile) with no active exploitation confirmed at time of analysis.

Denial of service in Linux kernel nexthop query handling allows local authenticated attackers to crash the kernel by querying nexthop objects with large Equal-Cost Multi-Path groups via the RTM_GETNEXTHOP netlink message. The vulnerability stems from fixed-size socket buffer allocation that overflows when processing nexthop groups exceeding approximately 512 members, triggering a kernel warning and potential system instability.

Information disclosure in Google BigQuery materialized view refresh allows authenticated users to extract sensitive data via crafted views that generate error messages containing confidential information. Google Cloud Platform patched this server-side vulnerability on 29 January 2026 with automatic remediation requiring no customer action. The CVSS 4.0 score of 7.1 reflects high confidentiality impact with network-accessible attack vector and low attack complexity, though no public exploit or CISA KEV listing exists at time of analysis.

PRNG state reuse across forked processes in CryptX for Perl allows remote attackers to recover private signing keys through cryptographic nonce-reuse attacks. When Crypt::PK objects are created before fork() in preforking web servers like Starman, every child process inherits identical PRNG state, causing duplicate randomness in cryptographic operations. Two ECDSA or DSA signatures generated by different worker processes are sufficient to mathematically recover the private key. EPSS exploitation probability is low (0.02%), but CISA SSVC framework confirms proof-of-concept availability and automatable exploitation. Vendor patch released in CryptX 0.088.

The HT Mega Addons for Elementor WordPress plugin before 3.0.7 contains an unauthenticated AJAX action returning some PII (such as full name, city, state and country) of customers who placed orders in the last 7 days

UUID library versions before 14.0.0 make unexpected writes to external output buffers when generating UUID versions 3, 5, or 6, potentially corrupting adjacent memory. UUID version 4, the most commonly deployed variant, is unaffected. The vulnerability requires local access and non-default buffer configuration to exploit, resulting in integrity compromise rather than code execution or availability impact.

Reliance on reverse DNS resolution in ericc-ch copilot-api up to version 0.7.0 allows authenticated remote attackers to manipulate the Host header in the /token endpoint, leading to information disclosure. The vulnerability affects the Header Handler component and has been publicly disclosed with exploit code available; the vendor did not respond to early disclosure notification.

IBM Security Verify Access and Verify Identity Access products versions 10.0 through 10.0.9.1 and 11.0 through 11.0.2 use cryptographic algorithms weaker than expected, allowing authenticated network attackers to decrypt highly sensitive information. The vulnerability affects both containerized and non-containerized deployments across multiple major versions. CVSS 6.5 reflects high confidentiality impact with low attack complexity, though authenticated access is required.

IBM Guardium Data Protection 12.0 through 12.2 contains a security misconfiguration in the user access control panel that allows high-privilege administrators to modify integrity settings without proper authorization constraints. The vulnerability is remotely accessible and requires existing administrative credentials, resulting in limited integrity impact with no confidentiality or availability effect. CVSS score of 2.7 reflects the low risk profile due to required administrative authentication and minimal scope of impact.

LangSmith Client SDKs in JavaScript (prior to 0.5.19) and Python (prior to 0.7.31) fail to apply output redaction controls to streaming token events, allowing sensitive LLM-generated content to leak into LangSmith platform storage despite hideOutputs/hide_outputs being enabled. Unauthenticated remote attackers can intercept or access unredacted streamed tokens if they gain visibility into run events, bypassing the intended confidentiality controls.

DLL hijacking in LiveOn Meet Client and Canon Network Camera Plugin installers allows local attackers to execute arbitrary code with installer privileges when users run vulnerable installer executables from directories containing malicious DLLs. The flaw affects four installer executables (Downloader5Installer.exe, Downloader5InstallerForAdmin.exe, CanonNWCamPlugin.exe, CanonNWCamPluginForAdmin.exe) version 1.0.0.0. No public exploit identified at time of analysis, though the attack technique is well-documented. CVSS 8.4 (High) reflects significant impact contingent on user interaction and attacker's ability to place malicious files in the installer directory before execution - real-world risk depends heavily on organizational download practices and endpoint controls preventing untrusted DLL placement.

Sandbox bypass in ByteDance verl up to version 0.7.0 allows remote attackers to achieve limited information disclosure through manipulation of the math_equal function in prime_math/grader.py. The vulnerability requires high attack complexity and has been publicly documented with exploit code available, though the vendor has not responded to early disclosure attempts.