| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargs_data field, cache identifiers in the features.mm_hashes field, ranges in the features.mm_placeholders field, and wire-selected multimodal field processors without rebinding them to the active model renderer contract. Forged grid geometry, field types, or non-positive placeholder lengths can terminate the shared EngineCore; when an attacker knows or can induce a victim's content hash, forged cache hashes can poison or retrieve cross-request encoder-cache state; and dropped sparse placeholder masks can alter replayed transport semantics. This issue is fixed in version 0.30.0. |
| Python's elementtree C accelerator failed to initialise Expat's hash salt during initialization. This could make it easy to conduct denial of service attacks against Expat by constructing an XML document that would cause pathological hash collisions in Expat's internal data structures, consuming large amounts CPU and RAM. The vulnerability exists in Python versions 3.7.0, 3.6.0 through 3.6.6, 3.5.0 through 3.5.6, 3.4.0 through 3.4.9, 2.7.0 through 2.7.15. |
| Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.83.0 until 2.131.0, the KServeV2OcrModel class defined in docling/models/stages/ocr/kserve_v2_ocr_model.py sends page images to its configured endpoint without checking the pipeline_options.enable_remote_services setting, even when the caller sets that policy control to false. The StandardPdfPipeline._make_ocr_model method also fails to pass the flag into the OCR factory, allowing remote OCR processing in configurations that rely on remote services being disabled. The destination is configured by the caller rather than selected by an attacker. This issue is fixed in 2.131.0. |
| MISP contains a defect in its event save workflow that prevents the correlation engine from recalculating correlations when an event's distribution level or sharing group is modified.
When a user edits an existing event and changes its distribution or sharing_group_id, the internal before-save hook stored the incoming (new) data rather than the previously persisted values. As a result, the after-save comparison that determines whether a correlation refresh is needed never detected the change, and stale correlations persisted.
Security impact:
- Stale correlations may continue to expose event data to users in a broader sharing group after the event has been moved to a more restrictive group, resulting in unintended information disclosure.
- Conversely, newly relevant correlations may not appear after a distribution widening, degrading the completeness of threat intelligence sharing.
Preconditions:
- An authenticated user with write access to at least one MISP event.
- The user modifies the event's distribution or sharing_group_id field.
Affected versions: <2.5.48 |
| In the Linux kernel, the following vulnerability has been resolved:
net: psp: avoid conflicts with skb->decrypted and sk_validate_xmit_skb()
PSP conflicts with TLS ULP in its usage of both skb->decrypted and
sk->sk_validate_xmit_skb().
Make PSP mutually exclusive with TLS ULP, the only other user of either
of these. As other users of skb->decrypted come along, they can be added
to sk_has_decrypt_user(). It would make sense to also assert that
sk->sk_validate_xmit_skb() is also NULL in both of these setup paths for
similar future proofing, but the PSP listener/sk_clone() path is still
broken and it could be seen as a regression to not allow rx assoc to run
on a child of a listener socket with PSP tx assoc state.
Include all TCP ULPs in the sk_has_decrypt_user() check, even though TLS
is the only one that conflicts with PSP via the decrypted bit. This is
intentional because PSP was not designed to be used with ULPs. It is
best to close off surface area that may make bugs reachable, until
someone wishes to design and test an actual user of PSP with ULPs. |
| BuildKit may be tricked into performing file actions with special file inodes where regular files are expected. Special files may block operations or, on rootful workers, allow unintended host device access. |
| Pexip Infinity before 38.2, plus 39.0, 39.1, and 40.0, is affected by improper input validation within an internal Pexip Infinity service that allows an attacker with local access to escalate privileges to root. Exploitation requires an attacker to be able to run arbitrary code on a node by either achieving remote code execution via some other vulnerability or having administrative access to the operating system. |
| Nezha Dashboard from 1.8.0 before 2.3.13 contains an improper locking vulnerability where a non-deferred mutex unlock leaks on a nil-map panic path. Any authenticated non-admin member can issue four notification API calls to permanently deadlock the alerting subsystem, then exhaust memory with blocking requests. |
| Improper initialization in an ASUS certain motherboard allows an physically proximate user to read or write arbitrary memory by inserting a specially crafted device.
Refer to the '
Security Update for Multiple ASUS Motherboards ' section on the ASUS Security Advisory for more information. |
| In the Linux kernel, the following vulnerability has been resolved:
nvme-fc: Do not cancel requests in io target before it is initialized
A new nvme-fc controller in CONNECTING state sees admin request timeout
schedules ctrl->ioerr_work to abort inflight requests. This ends up
calling __nvme_fc_abort_outstanding_ios() which aborts requests in both
admin and io tagsets. In case fc_ctrl->tag_set was not initialized we
see the warning below. This is because ctrl.queue_count is initialized
early in nvme_fc_alloc_ctrl().
nvme nvme0: NVME-FC{0}: starting error recovery Connectivity Loss
INFO: trying to register non-static key.
The code is fine but needs lockdep annotation, or maybe
lpfc 0000:ab:00.0: queue 0 connect admin queue failed (-6).
you didn't initialize this object before use?
turning off the locking correctness validator.
Workqueue: nvme-reset-wq nvme_fc_ctrl_ioerr_work [nvme_fc]
Call Trace:
<TASK>
dump_stack_lvl+0x57/0x80
register_lock_class+0x567/0x580
__lock_acquire+0x330/0xb90
lock_acquire.part.0+0xad/0x210
blk_mq_tagset_busy_iter+0xf9/0xc00
__nvme_fc_abort_outstanding_ios+0x23f/0x320 [nvme_fc]
nvme_fc_ctrl_ioerr_work+0x172/0x210 [nvme_fc]
process_one_work+0x82c/0x1450
worker_thread+0x5ee/0xfd0
kthread+0x3a0/0x750
ret_from_fork+0x439/0x670
ret_from_fork_asm+0x1a/0x30
</TASK>
Update the check in __nvme_fc_abort_outstanding_ios() confirm that io
tagset was created before iterating over busy requests. Also make sure
to cancel ctrl->ioerr_work before removing io tagset. |
| In the Linux kernel, the following vulnerability has been resolved:
usb: xhci: bail out of setup if the controller is inaccessible
xhci_gen_setup() locates the operational registers using the capability
length read from the very first register:
xhci->op_regs = hcd->regs +
HC_LENGTH(readl(&xhci->cap_regs->hc_capbase));
If the controller is dead or has dropped off the bus, that read returns
~0, HC_LENGTH() truncates it to 0xff, and op_regs ends up 0xff bytes
past the page-aligned MMIO base, i.e. unaligned. The first access
through it, xhci_halt() -> xhci_handshake() reading op_regs->status, is
then an unaligned readl() on device memory. arm64 faults on unaligned
device accesses, so instead of xhci_handshake() catching the all-ones
value and returning -ENODEV, setup oopses:
xhci-pci-renesas 0005:08:00.0: Unable to change power state from D3cold to D0, device inaccessible
xhci-pci-renesas 0005:08:00.0: xHCI Host Controller
xhci-pci-renesas 0005:08:00.0: new USB bus registered, assigned bus number 1
Unable to handle kernel paging request at virtual address ffff80030a770103
ESR = 0x0000000096000021
FSC = 0x21: alignment fault
Internal error: Oops: 0000000096000021 [#1] SMP
pc : xhci_halt [xhci_hcd]
Call trace:
xhci_halt
xhci_gen_setup
xhci_pci_setup
usb_add_hcd
usb_hcd_pci_probe
xhci_pci_common_probe
xhci_pci_renesas_probe
This was hit with a Renesas uPD720201 that failed to power up ("Unable
to change power state from D3cold to D0, device inaccessible") yet still
reached the HCD probe path.
Read the capability register once, and if it reads back the all-ones
value (as xhci_handshake() and xhci_reset() already test for), abort
setup with -ENODEV before op_regs is derived from it. Reading it once
also avoids re-reading a register that may change under a concurrent
hot-removal. |
| In the Linux kernel, the following vulnerability has been resolved:
wifi: iwlwifi: mei: pass correct argument to function
The first argument to iwl_mei_write_cyclic_buf() should be the cldev
but the q_head pointer is passed instead. Fix it. |
| NVIDIA vGPU Virtual GPU Manager for Linux contains a vulnerability where an attacker could cause incorrect resource transfer between spheres. A successful exploit of this vulnerability might lead to code execution, denial of service, escalation of privileges, information disclosure, and data tampering. |
| NVIDIA GPU Display Driver for Windows contains a vulnerability in the kernel module where a user could cause improper locking. A successful exploit of this vulnerability might lead to denial of service. |
| Werkzeug is a comprehensive WSGI web application library. Prior to 3.1.9, the safe_join function used by send_from_directory can allow a NUL: special-device path because safe_join checks the Windows device name without first removing an empty NTFS ADS marker. The trigger is that an application runs on Windows with NTFS and serves a user-specified path ending in a special device name such as NUL:. The attack mechanism is that a requested path ends in a Windows special device name with an empty ADS marker. The impact is that the special device opens successfully and the file read hangs indefinitely. This issue is fixed in version 3.1.9. |
| In the Linux kernel, the following vulnerability has been resolved:
bpf: Cancel special fields when recycling rhtab elements
rhtab_map_update_existing() and rhtab_delete_elem() call
bpf_obj_free_fields() when replacing or deleting a value. These map
operations can run from BPF programs in NMI context, where releasing a
referenced kptr or another complex field is not generally safe.
Array and hash maps avoid that problem by cancelling only the asynchronous
fields which can be stopped safely in the caller context. Other ownership
state remains attached to the allocation until its memory allocator
destructor performs the final cleanup.
Use bpf_obj_cancel_fields() for the corresponding rhtab paths as well. This
cancels timers, workqueues, and task work while allowing rhtab_mem_dtor() to
release referenced kptrs when the allocation is eventually destroyed.
[ kkd: Rebased, used direct helper calls, and rewrote the commit log ] |
| Multiple BinSoft products are vulnerable to Remote Command Execution. An authorized user with access to the application and direct access to the backend database can achieve system command execution by uploading an attachment and modifying its storage path in the database to reference an attacker-controlled remote network resource. Alternatively, it is possible to use a previously uploaded file and change its reference. When the application processes the attachment, and a user tries to open it, the referenced resource is executed by the system.
Critically, this vulnerability can be exploited by any unauthenticated attacker by chaining it with CVE-2026-40550 and CVE-2026-40551, which allows obtaining database access, and logging onto any account.
The described issue affects all published versions. The vendor stated that this issue is a direct result of the architecture model in which the software is distributed, and that it will be mitigated with a corrected installation manual. |
| Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. From 8.0.0 until 8.0.6, the SMTP MIME parser in rust/src/mime/smtp.rs does not fully reset state when processing Content-Type: message/rfc822 encapsulation. An outer MIME part's encoding or filename state can leak into the inner message, allowing crafted mail to evade detections based on file.data, file.name, or extracted URLs when SMTP MIME decoding is enabled. This issue is fixed in version 8.0.6. |
| Ghidra before 12.1.4 fails to validate the TYPE_COL byte in OptionsDB.createUnregisteredOption(), causing an ArrayIndexOutOfBoundsException that leaves domain objects permanently locked. Attackers can craft a malicious program database file that, when imported, causes the application to stall and prevents resource cleanup or graceful shutdown. |
| Improper Exposure of Resource to Wrong Sphere in the host file helper (gofer) in Google gVisor prior to commit 573a9e73cf844f on Linux platforms with CUSE enabled allows a local attacker with container image deployment privileges to achieve root code execution on the host system. By including a /dev/cuse character device node in a container image, opening the device passes through to the host, allowing the sandboxed attacker to register a host device and exploit CUSE unrestricted ioctl handling to overwrite root udev helper memory. |
