| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| kin-openapi is a Go project for handling OpenAPI files. Prior to 0.148.0, the OpenAPI 3.1 validation path passes YAML NaN and infinity numeric values to santhosh-tekuri/jsonschema v6.0.2 without applying kin-openapi's existing finite-number guards. A remote unauthenticated client can send an application/yaml or application/x-yaml request body containing .nan, .inf, or -.inf to an operation with a numeric constraint, causing the dependency to create and dereference a nil big.Rat during ValidateRequest. Standard net/http servers recover the panic per connection, while other hosts or application goroutines may terminate the process. This issue is fixed in version 0.148.0. |
| In the Linux kernel, the following vulnerability has been resolved:
bpf: Avoid soft lockup in __htab_map_lookup_and_delete_batch()
__htab_map_lookup_and_delete_batch() has no rescheduling point. The
batch count bounds how many entries are copied out, not how many
buckets are visited, so one BPF_MAP_LOOKUP_BATCH call can walk the
map end to end. The empty-bucket fast path is worse: it stays inside
a single rcu_read_lock() / bpf_disable_instrumentation() section for
any run of consecutive empty buckets.
That holds up on small maps, but it falls apart at scale. On a
144-CPU arm64 host running a CONFIG_PREEMPT_NONE kernel, periodic
BPF_MAP_LOOKUP_BATCH calls against an LRU hash map with 16,777,216
buckets held a CPU inside the batch op for 77+ seconds and triggered
the soft lockup watchdog.
Commit 75134f16e7dd ("bpf: Add schedule points in batch ops") fixed this
same problem in the generic batch ops, but not in this htab-native path,
which every htab-based hash map variant uses for its lookup[_and_delete]
batch ops.
Complete that fix here. Leave the critical section after 64 consecutive
empty buckets, call cond_resched_tasks_rcu_qs(), and resume at the saved
bucket cursor. No locks are held at that point, and resuming from the
cursor is already the function's behavior for non-empty buckets. Add the
same call to the per-bucket loop after copy_to_user(), where every lock
has been dropped. cond_resched_rcu() is not enough here: sleeping with
bpf_prog_active elevated makes tracing programs on that CPU silently
skip their invocations.
Plain cond_resched() is not enough either. It is a no-op under PREEMPT
and PREEMPT_LAZY, the only models arm64 and x86 have offered since
commit 7dadeaa6e851 ("sched: Further restrict the preemption modes").
It is also never a Tasks RCU quiescent state, in any model: the
reschedule counts as a preemption. The walking task stays a holdout and
stalls every synchronize_rcu_tasks() caller, ftrace and BPF trampoline
teardown included, until the syscall returns [1].
cond_resched_tasks_rcu_qs() is the usual tool for that [2]. It reports
the quiescent state at each yield and still reschedules as
cond_resched() does on PREEMPT_NONE and PREEMPT_VOLUNTARY kernels. |
| In the Linux kernel, the following vulnerability has been resolved:
HID: winwing: fix use-after-free in force feedback teardown
winwing_init_ff() passes the driver's private data, allocated with
devm_kzalloc() in winwing_probe(), as the effect context to
input_ff_create_memless(). The memoryless force-feedback core takes
ownership of that pointer and frees it with kfree() from
input_ff_destroy() (ml_ff_destroy()) when the input device is
destroyed.
Freeing a devm-managed allocation with kfree() is an invalid free, and
the same object is then released again by devres when the HID device is
torn down, a double free. As the allocation also embeds the LED class
devices, their timers and work item live on freed memory and the slab
gets corrupted. This triggers on unbind, rmmod, hot-unplug and on system
suspend, where the firmware cache walks the now-corrupt devres list.
KASAN reports:
BUG: KASAN: invalid-free in input_ff_destroy
Allocated by task N:
winwing_probe
Pass NULL as the memless context instead and fetch the driver data from
the input device in winwing_play_effect(): the HID core already stores
the hid_device as the input device's drvdata. The force-feedback core
then owns nothing that it must not free. |
| Subscriber PHP Object Injection in Dynamic User Directory <= 2.4 versions. |
| In the Linux kernel, the following vulnerability has been resolved:
Bluetooth: btintel_pcie: validate device-supplied DMA indices
In btintel_pcie_msix_rx_handle(), the driver processes RX completion
descriptors (urbd1) written by the PCIe device into DMA-coherent memory.
urbd1->frbd_tag (a 16-bit field fully controlled by the device firmware
via DMA) is used directly as an array index into rxq->bufs[] without any
bounds check. rxq->bufs[] has only BTINTEL_PCIE_RX_DESCS_COUNT (64)
entries, while frbd_tag can be any value 0-65535. A malicious or
malfunctioning device can write an out-of-range frbd_tag, causing the
driver to dereference an out-of-bounds data_buf pointer.
Additionally, cr_hia is read from a DMA-shared index array also writable
by the device; if the device sets cr_hia >= rxq->count, the while-loop
never terminates because cr_tia is wrapped via modulo rxq->count and can
never equal an out-of-range cr_hia.
Add bounds validation for cr_hia and frbd_tag in the RX path, and cr_hia
in the TX path. Log invalid values with bt_dev_err before returning. |
| In the Linux kernel, the following vulnerability has been resolved:
Bluetooth: bnep: fix out-of-bounds reads on short RX/TX frames and control fallthrough
Fix multiple out-of-bounds reads in Bluetooth BNEP frame processing:
1. In bnep_rx_frame() and bnep_ctrl_frame() (net/bluetooth/bnep/core.c),
use pskb_may_pull() to verify the BNEP header, control type byte,
filter count, and extension headers exist before reading them, and
return 0 after handling BNEP_CONTROL instead of falling through to
Ethernet frame submission when no extension headers follow.
2. In bnep_net_xmit() (net/bluetooth/bnep/netdev.c), verify skb->len >=
ETH_HLEN with pskb_may_pull() before reading the 14-byte Ethernet
header to prevent an out-of-bounds heap read and infoleak on short
AF_PACKET TX frames. |
| In the Linux kernel, the following vulnerability has been resolved:
netfilter: nft_ct: bail out on template ct in get eval
I noticed this issue while looking at a historic syzbot report [1].
A rule like the one below is enough to trigger the bug:
table ip t {
chain pre {
type filter hook prerouting priority raw;
ct zone set 1
ct original saddr 1.2.3.4 accept
}
}
The first expression attaches a per-cpu template ct via
nft_ct_set_zone_eval() (nf_ct_tmpl_alloc -> kzalloc, tuple is all
zero, nf_ct_l3num(ct) == 0). The next expression then calls
nft_ct_get_eval() on the same skb, treats the template as a real ct
and hits the 16-byte memcpy path. With dreg at NFT_REG32_15 this
overflows past struct nft_regs on the kernel stack; with smaller
dreg values it silently clobbers adjacent registers.
Reject template ct at the eval entry and in nft_ct_get_fast_eval(),
mirroring the check nft_ct_set_eval() already has. Additionally,
bound the address copy in NFT_CT_SRC / NFT_CT_DST by priv->len
instead of by nf_ct_l3num(ct): nf_ct_get_tuple() zeroes the tuple
before pkt_to_tuple() fills in only the protocol-relevant leading
bytes, so the trailing bytes of tuple->{src,dst}.u3.all are
well-defined zero. priv->len is validated at rule load, so the
copy size is now bounded by the destination register rather than
by an untrusted field on the conntrack.
[1]: https://syzkaller.appspot.com/bug?id=389cf09cb72926114fce90dc85a2c3231dcb647c |
| In the Linux kernel, the following vulnerability has been resolved:
arm64: errata: Mitigate TLBI errata on various Arm CPUs
A number of CPUs developed by Arm suffer from errata whereby a broadcast
TLBI;DSB sequence may complete before the global observation of writes
which are translated by an affected TLB entry.
These errata ONLY affect the completion of memory accesses which have
been translated by an invalidated TLB entry, and these errata DO NOT
affect the actual invalidation of TLB entries. TLB entries are removed
correctly.
This issue has been assigned CVE ID CVE-2025-10263.
To mitigate this issue, Arm recommends that software follows any
affected TLBI;DSB sequence with an additional TLBI;DSB, which will
ensure that all memory write effects affected by the first TLBI have
been globally observed. The additional TLBI can use any operation that
is broadcast to affected CPUs, and the additional DSB can use any option
that is sufficient to complete the additional TLBI.
The ARM64_WORKAROUND_REPEAT_TLBI workaround is sufficient to mitigate
the issue. Enable this workaround for affected CPUs, and update the
silicon errata documentation accordingly.
Note that due to the manner in which Arm develops IP and tracks errata,
some CPUs share a common erratum number. |
| The WP Photo Album Plus plugin for WordPress is vulnerable to Stored Cross-Site Scripting via REQUEST_URI Session History in all versions up to, and including, 9.3.03.002 due to insufficient input sanitization and output escaping. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. The bypass works because esc_url_raw() strips literal angle brackets but retains HTML entities, which wppaEntityDecode() silently converts back to live HTML tags before jQuery('#wppa-modal-container').html() renders them. |
| In the Linux kernel, the following vulnerability has been resolved:
net/sched: act_ct: don't WARN on benign flow_offload_alloc() failure
flow_offload_alloc() returns NULL when the conntrack entry is dying
(e.g. raced with a conntrack flush) or when the GFP_ATOMIC allocation
fails; both are expected under load and neither is a kernel bug. This
path runs from softirq on every committed packet, so with
panic_on_warn=1 an unprivileged user can panic the box just by racing
a conntrack flush against a `tc ... action ct commit` classifier.
Reproduced with a custom repro under QEMU: a small, fixed set of UDP
flows through `tc filter ... action ct commit` on lo, raced against
threads flooding bare ctnetlink CT_DELETE (flush) requests. Hits
WARNING: net/sched/act_ct.c:437 (tcf_ct_flow_table_add(), inlined
into tcf_ct_act() in this build) within ~15s on the unpatched kernel;
same setup is clean on the patched kernel. The fix itself is
behavior-preserving: both branches already did `goto err_alloc`
before and after, only the WARN is removed. |
| `GET /api/v1/user/subscriptions` returned the metadata of every repository a user was watching without checking that the user still had access to it. A user who watched a private repository while having access, and later lost it, for example through a team permission downgrade, could still read its name, description, clone URLs, owner and counters. The issue was fixed in 1.27.0 and 28.0.0 and disclosed with 28.2.0. |
| F´ (F Prime) is a framework that enables development and deployment of spaceflight and other embedded software applications. Prior to 4.3.0, Svc::FileDownlink SendPartial range validation is vulnerable because unchecked startOffset plus length arithmetic bypasses the range check and stores a wrapped m_endOffset that is inconsistent with m_byteOffset. When an existing source file is larger than a valid startOffset and a SendPartial command supplies a length whose U32 sum with that offset wraps below the file size, under standard F Prime assertion handling, the first data-packet transition reaches FW_ASSERT and aborts or resets the process, causing denial of service. This issue is fixed in version 4.3.0. |
| F´ (F Prime) is a framework that enables development and deployment of spaceflight and other embedded software applications. Prior to 4.2.0, Svc::GenericHub fromBufferDriver_handler deserialization is vulnerable because the receive handler uses FW_ASSERT for externally influenced deserialization status and type validation and permits a command payload size subtraction to underflow. When an unauthenticated inter-processor hub peer sends a truncated buffer, an invalid HubType, or a header-only command-dispatch payload, malformed hub traffic aborts or resets the process, causing denial of service. This issue is fixed in version 4.2.0. |
| Discourse is an open-source discussion platform. Prior to 2026.9.0, 2026.8.1, 2026.7.3, and 2026.1.9, oEmbed processing treated an entire attacker-controlled HTML fragment as trusted when the fragment contained an iframe whose source matched the configured allowlist. A user with posting permission who controlled a public webpage and its oEmbed response could combine an allowlisted iframe with unrelated sibling HTML or CSS, causing stored visual overlays or other styled content to appear in post content and enabling misleading or fraudulent user interfaces. Exploitation requires a posting account, control of the fetched webpage and oEmbed response, and an allowlisted iframe source. The issue affects content integrity but does not expose restricted data, disrupt availability, or require interaction with Discourse Content Security Policy. This issue is fixed in versions 2026.9.0, 2026.8.1, 2026.7.3, and 2026.1.9. |
| Discourse is an open-source discussion platform. Prior to 2026.9.0, 2026.8.1, 2026.7.3, and 2026.1.9, sites with secure uploads enabled did not reconcile existing upload security metadata when a previously public category became read-restricted or when category permissions changed during a bulk import. Media and attachments uploaded while the category was public retained a public accessibility flag and remained reachable by anonymous users through their short upload URLs, including uploads associated with deleted topics. This issue is fixed in versions 2026.9.0, 2026.8.1, 2026.7.3, and 2026.1.9. |
| A flaw was found in DPDK lib/vhost vhost-crypto. Write-back chain cleanup can follow an uninitialized next pointer when mapping a guest descriptor fails, causing the host vhost-crypto process to hang or crash. |
| A flaw was found in DPDK lib/vhost. Memory size is not adequately validated before mmap during vhost memory-region setup, which can lead to a host crash or memory corruption. |
| Gogs is an open source self-hosted Git service. Prior to 0.15.0+dev, an unauthenticated attacker can provide a backslash-containing target to the /redirect endpoint, and similar post-action redirect flows are also affected. internal/urlx/urlx.go IsSameSite validates a normalized form of the target while cmd/gogs/internal/web/web.go getRedirect and other callers use the original value in the Location response. This backslash normalization mismatch lets browsers interpret the Location value as a scheme-relative external URL, allowing links on a trusted Gogs domain to redirect victims to attacker-controlled content for phishing. This vulnerability is fixed in 0.15.0+dev. |
| KEDA is a Kubernetes-based Event Driven Autoscaling component. In version 2.20.2 and earlier, a namespace administrator who can create TriggerAuthentication and ScaledObject resources can set hashiCorpVault.address to a tenant-controlled endpoint and select Kubernetes authentication without an explicit credential. KEDA then uses the operator pod's projected ServiceAccount token and sends it to that endpoint as a Vault login credential. The JWT-shape filter from the prior CVE-2025-68476 remediation accepts the operator token because its subject is a Kubernetes ServiceAccount, so the attacker can obtain the operator's cluster-wide permissions, including broad Secret reads and workload or admission-related operations. This vulnerability is fixed in 2.21.0. |
| Mongoose is an embedded web server and network library. Prior to 7.23, the built-in TLS certificate parser selected by MG_TLS_BUILTIN fails to verify that a long-form DER element length fits within the remaining certificate buffer. In mg_der_to_tlv() and mg_der_next(), a peer-controlled length can advance the parser beyond the certificate and underflow the remaining uint32_t length, causing mg_tls_parse_cert_der() to continue walking an invalid pointer. An unauthenticated malicious TLS server, or a client presenting a certificate to a server configured for client certificates, can trigger an out-of-bounds read and denial of service during the handshake. This issue is fixed in version 7.23. |
