CVE-2026-90099
UnknownSummary
A vulnerability in the Linux kernel's net/sched classifier mechanism allows unprivileged users to pin kernel memory outside memcg control by using GFP_KERNEL allocations without __GFP_ACCOUNT in classifier *_change() paths and the shared tcf_exts_init_ex() action array. Additionally, an error path in cls_basic leaves a dangling pointer in the IDR.
Risk Assessment
The risk is that an unprivileged user can increase kernel memory usage beyond cgroup limits, potentially leading to resource exhaustion (DoS) and bypassing memory limits, which could impact other processes on the system.
Recommendation
Apply the official Linux kernel patch that adds GFP_KERNEL_ACCOUNT to classifier allocations and fixes the error path in cls_basic. Additionally, consider restricting unprivileged users' ability to create tc filters.
Other vulnerabilities in Linux kernel
See all- CVE-2026-98163Unknown
A vulnerability was found in the Linux kernel's cgroup subsystem, involving iteration over dying tasks with a zero reference count. This leads to a race condition where the iterator may access a task after it has been freed, resulting in a use-after-free.
- CVE-2026-98162Unknown
In the Linux kernel, the smb2_tree_connect() function of the SMB server (ksmbd) leaks a tree connection. When ksmbd_iov_pin_rsp() fails, the newly created tree connection is not disconnected, leading to a resource leak.
- CVE-2026-98161Unknown
In the Linux kernel, the nvdimm (pmem) pmem_submit_bio() function records a REQ_PREFLUSH error but continues to copy bio data and can later overwrite the error with a successful REQ_FUA flush. This allows data writes to run after a failed preflush and can complete the bio successfully despite the failed ordering barrier.
- CVE-2026-98160Unknown
In the Linux kernel, the staging rtl8723bs driver's rtw_sdio_if1_init() function frees padapter->HalData with kfree(), even though it was allocated via vzalloc(). Using kfree() to release a vmalloc-backed buffer can lead to memory corruption.
- CVE-2026-100079Unknown
In the Linux kernel, the USB Type-C (ucsi) subsystem's ucsi_register() creates per-instance debugfs entries, but ucsi_unregister() keeps them until ucsi_destroy(). Drivers like ucsi_glink that unregister/register the same UCSI instance across remoteproc restart then try to create an already existing debugfs directory.
- CVE-2026-100078Unknown
In the Linux kernel, the iwlwifi (mei) driver's iwl_mei_write_cyclic_buf() function receives an incorrect first argument — the q_head pointer is passed instead of cldev. The bug has been fixed.
- CVE-2026-100077Unknown
In the Linux kernel, the drm/msm driver does not safely retire a hung submit before GPU recovery completes. Retiring the submit triggers BO free, which can result in GPU pagefaults since the GPU may be actively accessing those BOs.
- CVE-2026-100074Unknown
In the Linux kernel, a bug was fixed where the BPF_REFCOUNT field was not marked as unique, although it should be. The fix addresses this oversight.
- CVE-2026-100073Unknown
In the Linux kernel, a bug in the ext4 filesystem related to transaction overflow during writeback was fixed. A previous fix was too eager in reducing reserved transaction credits, leading to insufficient reservation in some corner cases. The fix uses ext4_meta_trans_blocks() for a correct upper bound estimate.
- CVE-2026-100072Unknown
In the Linux kernel, a problem in the ACPI subsystem was fixed where the use of acpi_get_first_physical_node() in acpi_platform_fill_resource() and acpi_create_platform_device() was unsafe because the returned device could be freed at any time. The fix replaces it with acpi_bus_get_primary_device() and adjusts the code to call it only once.
Original NVD description (English source)
In the Linux kernel, the following vulnerability has been resolved: net/sched: account classifier filter allocations to memcg Allocations in the tc classifier *_change() paths (filter objects, per-CPU counters, and per-filter aux data) use plain GFP_KERNEL without __GFP_ACCOUNT, allowing unprivileged users to pin kernel memory outside memcg charging. The shared tcf_exts_init_ex() action array allocation in cls_api.c was also uncharged; this patch closes it along with the per-classifier filter-object/percpu/aux allocations that remain unaccounted. Add GFP_KERNEL_ACCOUNT to: - the shared tcf_exts_init_ex() action array (cls_api.c), common to every filter of every classifier (32 pointers, 256 bytes); - the filter-object, per-CPU-counter, and per-filter aux allocations in cls_basic, cls_bpf, cls_cgroup, cls_flow, cls_flower, cls_fw, cls_matchall, cls_route and cls_u32; - the u32_init_knode() replace-path knode allocation (cls_u32.c), which allocates the same struct tc_u_knode + sel.keys on every replace of an existing knode and was missed by the create-path-only conversion. Also fix the cls_basic error path: basic_change() inserts fnew into the IDR before allocating the per-CPU counter. If alloc_percpu() fails the errout path kfree'd fnew without idr_remove, leaving a dangling pointer in the IDR. With GFP_KERNEL_ACCOUNT the percpu alloc becomes failable on demand (memcg at memory.max), making the dead path attacker-reachable and burning the handle permanently. Add the idr_remove on the percpu failure path, matching the basic_set_parms failure-path pattern. Note: [email protected] provided a poc for basic_cls, but it was easy to extend to the other classifiers. Conditions to recreate the bug: - CONFIG_NET_SCHED, CONFIG_NET_CLS_* (the classifier being used), CONFIG_NET_CLS_ACT, CONFIG_MEMCG, CONFIG_USER_NS, CONFIG_NET_NS. - Unprivileged user in a fresh user+network namespace (unshare -Urn), or root with CAP_NET_ADMIN. - Create a large number of tc filters (e.g. tc filter add dev lo ingress ... <classifier> ...) while watching a memcg-limited cgroup: system slab grows far faster than memory.current, pinning kernel memory outside memcg charging.
Vulnerability data from NVD (NIST) · CISA KEV · EPSS