ContextForge is an AI gateway, registry, and proxy that provides centralized discovery, guardrails, and management for MCP, A2A, and REST or gRPC APIs. Prior to 1.0.3, the /admin/gateways/test call site in mcpgateway/admin.py calls validate_gateway_test_url() in mcpgateway/common/validators.py to resolve and reject private, loopback, link-local, and cloud-metadata addresses, but ResilientHttpClient later resolves the original hostname again without binding the validated address. When MCPGATEWAY_ADMIN_API_ENABLED is enabled, an attacker with a database-backed role containing explicit gateways.read permission can use DNS rebinding to return a public address during validation and a private or metadata address during connection, bypassing ssrf_blocked_networks and ssrf_dns_fail_closed because those controls apply only to the validation-time result. The endpoint's allow_admin_bypass=False setting means a bootstrap-only virtual platform-admin identity without a database role is not sufficient. Successful exploitation can reach internal services and cloud metadata, expose cloud credentials, access internal APIs, or probe internal network ports. This issue is fixed in version 1.0.3.
A use-after-free vulnerability in the Linux kernel's at91_udc USB gadget driver was resolved. The issue occurs because a polled-VBUS timer and work cycle continue to reference a device context (udc) after it has been freed during driver removal or probe failure. This can lead to dereferencing freed memory, causing potential system instability or compromise. The fix involves properly shutting down the timer and work cycle during driver removal and probe failure paths to prevent use-after-free conditions.
A vulnerability in the Linux kernel's serial driver for imx UART ports was resolved. The issue involved improper serialization of the imx_uart_ports[] array lifetime, where a dangling pointer could remain after a failed port addition or removal. This allowed a sibling probe to register a shared console through a stale entry, potentially causing inconsistent or unintended behavior. The fix ensures the entry remains valid through removal and protects port addition and removal operations to prevent interleaving. Occupied slots are now rejected rather than overwritten during duplicate-line probes.
In the Linux kernel, the following vulnerability has been resolved: power: supply: charger-manager: register regulators before exposing sysfs charger_manager_remove() and the err_reg_extcon probe error path free each charger regulator with regulator_put() before tearing down the power_supply sysfs entries (power_supply_unregister()). charger_manager_remove() also calls try_charger_enable(cm, false) after the regulator_put() loop. A concurrent write to a charger's externally_control sysfs attribute that lands between regulator_put() and power_supply_unregister() can run charger_externally_control_store() and call try_charger_enable(), which, when charging is enabled, dereferences the already-freed consumer handle. When charging is enabled, try_charger_enable(cm, false) in .remove() also dereferences the freed handles directly. Both leave use-after-free windows. Symmetrically, probe registers the sysfs entries (power_supply_register) before acquiring the regulators (regulator_get, inside charger_manager_register_extcon), so userspace can reach externally_control before the regulators are available. Split charger_manager_register_extcon() on the sync/async boundary: charger_manager_get_regulators() (regulator_get only, no async producer) now runs before power_supply_register() so sysfs is not live before regulators are available, and charger_manager_register_extcon() keeps only the extcon notifier/work setup, still after power_supply_register() so a power_supply_register() failure cannot reach extcon setup. This keeps the sysfs setup/teardown ordering symmetric without introducing an asynchronous producer on the earlier probe-error path. Move power_supply_unregister() and try_charger_enable(cm, false) ahead of the regulator_put() loop on both teardown paths, and adjust err_reg_extcon (power_supply_unregister() then fall through err_regulator for regulator_put(); get_regulators self-rolls back on its own failure). This does not address the separate extcon-notifier-driven deref of the same handles, which needs its own synchronization design. Found by an in-house static analysis tool.
A high-severity vulnerability in the Linux kernel's msm_iommu_probe() function causes improper cleanup on registration failure, leading to dangling list entries and potential kernel memory corruption. This occurs when the IOMMU sysfs device addition or registration fails, leaving stale references that may be dereferenced later. The issue has been resolved by unwinding the sysfs device and global list entries in reverse order on failure paths.
In the Linux kernel, the following vulnerability has been resolved:
platform/x86: panasonic-laptop: Fix sentinel write past pcc->sinf[]
acpi_pcc_retrieve_biosdata() rejects SINF packages only when
pcc->num_sifr is strictly less than hkey->package.count, then
unconditionally writes a trailing sentinel at
pcc->sinf[hkey->package.count]. But pcc->sinf[] is allocated with
exactly pcc->num_sifr elements (valid indices 0..num_sifr-1), so that
write needs num_sifr strictly greater than package.count to stay in
bounds -- num_sifr == package.count passes the existing check but
still overflows by one element.
This is exactly the case probe()'s existing num_sifr++ workaround
("Some DSDT-s have an off-by-one bug where the SINF package count is
one higher than the SQTY reported value") is written to accommodate:
when a DSDT's SINF package count equals SQTY+1, the workaround makes
num_sifr equal to package.count, which is precisely the boundary that
overflows here. Found via UBSan (array-index-out-of-bounds) on
hardware where HKEY.SQTY returns 37 and HKEY.SINF()'s package has 38
elements: num_sifr becomes 38 after the += 1 workaround, the loop
correctly fills indices 0..37, and the sentinel write then targets
index 38, one past the end -- a silent 4-byte heap overflow on kernels
without CONFIG_UBSAN.
Tightening the rejection check to num_sifr <= package.count would
avoid the overflow but breaks probe() entirely on exactly this
hardware, since num_sifr == package.count is the case the off-by-one
workaround exists to support. Nothing else in the driver reads this
sentinel value back, so simply skip the write when there is no room
for it instead.
A vulnerability in the Linux kernel's amd_pmc driver causes resource leaks and list corruption when the STB initialization fails during probe. This leads to stale registrations and potential kernel crashes or use of torn-down driver state during s2idle transitions.
In the Linux kernel, the following vulnerability has been resolved: wifi: rtl8xxxu: fix use-after-free from rx_urb_wq on stop rtl8xxxu arms rx_urb_wq from the RX completion path: rtl8xxxu_rx_complete() hands the URB to rtl8xxxu_queue_rx_urb(), which queues it on rx_urb_pending_list and, once the list grows past RTL8XXXU_RX_URB_PENDING_WATER, schedules rx_urb_wq. The worker rtl8xxxu_rx_urb_work() drains rx_urb_pending_list, recovers priv through container_of, and resubmits each URB through rtl8xxxu_submit_rx_urb(), which anchors it on rx_anchor and dereferences priv->udev. rtl8xxxu_stop() cancels the sibling work items (c2hcmd_work, ra_watchdog, update_beacon_work) but never cancels rx_urb_wq, so a worker armed during the last burst of RX traffic can run rtl8xxxu_rx_urb_work() after rtl8xxxu_disconnect() has called ieee80211_free_hw(), which frees priv, producing a use-after-free. The window opens under active RX traffic (pending count above the watermark) followed by a disconnect. There are two teardown races to close: * rtl8xxxu_queue_rx_urb() decided whether to enqueue under rx_urb_lock but called schedule_work() after dropping the lock. A completion that observed shutdown == false and released the lock could then call schedule_work() after rtl8xxxu_stop() had set shutdown and cancel_work_sync() had already returned, arming the worker to run after the teardown. Move schedule_work() under the same !shutdown branch so the arming decision is atomic with the shutdown check. * rtl8xxxu_rx_urb_work() anchors every URB it drained back onto rx_anchor through rtl8xxxu_submit_rx_urb(). A worker still running when usb_kill_anchored_urbs(&priv->rx_anchor) returned would submit a URB that escaped the kill. In rtl8xxxu_stop(), call cancel_work_sync(&priv->rx_urb_wq) before the kill so the worker is drained first. After priv->shutdown is set under rx_urb_lock, completions can no longer queue rx_urb_wq. cancel_work_sync() then drains the last queued or running worker, and the following usb_kill_anchored_urbs() kills the URBs it may have submitted. rtl8xxxu_disconnect() is covered because ieee80211_unregister_hw() guarantees .stop() runs for a live interface before ieee80211_free_hw() frees priv. The probe error path needs no cancel: rx_urb_wq is INIT_WORK()'d there but cannot have been scheduled, since no URB is submitted before ieee80211_register_hw() succeeds. This bug was found by static analysis.
CVE-2026-89623 is a medium-severity vulnerability in the Linux kernel's HID driver for the MCP2221 device. It involves a race condition where device IO is not properly stopped before hardware teardown during probe failure or device removal, potentially leading to a use-after-free condition. The issue is addressed by quiescing device IO at the start of the cleanup callback and guarding the call to stop device IO with a flag to prevent spurious warnings and ensure proper synchronization.
A reference leak vulnerability in the Linux kernel power supply driver for bq25890 was fixed. The issue involved the bq25890_fw_probe() function acquiring a reference to a secondary charger without releasing it on probe failures or driver detach, potentially causing resource leaks. The fix registers a device-managed cleanup action to ensure proper release of the reference in all failure scenarios.
In the Linux kernel, the following vulnerability has been resolved: PCI: plda: Fix use-after-free of event IRQs during teardown plda_pcie_irq_domain_deinit() removes pcie->event_domain via irq_domain_remove(), but the per-event IRQs mapped from that domain are requested with devm_request_irq() in plda_init_interrupts(). The actual free_irq() for a devm-managed IRQ is deferred by devres until after the calling probe()/remove() function returns. This means irq_domain_remove() can free the domain's internal data before the deferred free_irq() for IRQs still mapped into it has run. When devres later processes that deferred cleanup, it can end up dereferencing the already-freed domain. Free each event IRQ explicitly with devm_free_irq() before removing the domain. This triggers the free immediately and removes the IRQ from the devres tracking list, so devres will not attempt to free it a second time later. Also dispose of the event, INTx, and MSI IRQ mappings with irq_dispose_mapping() before their owning domains are removed. Finally, guard the calls to irq_set_chained_handler_and_data() for pcie->irq, pcie->msi_irq, and pcie->intx_irq so they only run when those fields hold a valid (>0) IRQ number. This is a pre-existing issue, flagged by automated review during work on an earlier, unrelated patch to this driver. Build-tested and boot-tested on StarFive VisionFive v1.2A board
CVE-2026-89440 is a high-severity vulnerability in the Linux kernel's mmc via-sdmmc driver. It occurs when the card-detect interrupt remains registered after a probe failure, leading to use-after-free conditions where interrupt handlers access freed memory. This can cause system instability or crashes. The issue was identified through static analysis and manual review.
In the Linux kernel, the following vulnerability has been resolved: net: dsa: realtek: use gpiod_set_value_cansleep for reset GPIO rtl83xx_reset_assert() and rtl83xx_reset_deassert() are only called from the probe path, which may sleep and is not timing-critical. When the reset GPIO is provided by a sleeping controller such as an I2C I/O expander, gpiod_set_value() warns: WARNING: drivers/gpio/gpiolib.c:4030 at gpiod_set_value+0x44/0x80, CPU#1: kworker/u16:4/61 Hardware name: B&O MAP CA33 Rev f (UNKNOWN) (DT) Workqueue: events_unbound deferred_probe_work_func pc : gpiod_set_value+0x44/0x80 lr : rtl83xx_probe+0x1d8/0x3a0 Call trace: gpiod_set_value+0x44/0x80 (P) rtl83xx_probe+0x1d8/0x3a0 realtek_mdio_probe+0x24/0xa0 mdio_probe+0x38/0x78 really_probe+0xc4/0x3e0 __driver_probe_device+0x15c/0x1b8 driver_probe_device+0xb4/0x120 __device_attach_driver+0xb8/0x1a0 bus_for_each_drv+0x88/0xf0 __device_attach+0xa0/0x1d8 device_initial_probe+0x54/0x68 bus_probe_device+0x38/0xa0 deferred_probe_work_func+0xb8/0x120 process_one_work+0x184/0x4e8 worker_thread+0x188/0x308 kthread+0x130/0x150 ret_from_fork+0x10/0x20 Switch both helpers to gpiod_set_value_cansleep() so such a reset GPIO can be used without triggering the warning. The reset GPIO has been driven with the non-sleeping gpiod_set_value() since the driver was added in v4.19. The call has since been refactored across several files - from realtek-smi.c / realtek-mdio.c into the common rtl83xx.c module and then into the rtl83xx_reset_assert() and rtl83xx_reset_deassert() helpers (both in v6.9). This patch therefore applies as-is only to kernels that carry those helpers (v6.9+); older stable kernels need the same gpiod_set_value_cansleep() conversion at the corresponding open-coded call sites.
In the Linux kernel, the following vulnerability has been resolved: ALSA: 6fire: bound the MIDI event length from the device usb6fire_comm_receiver_handler() forwards a MIDI event using a length byte the device supplies, with no bound and no check that the transfer delivered that many bytes: if (!urb->status) { if (rt->receiver_buffer[0] == 0x10) /* midi in event */ if (midi_rt) midi_rt->in_received(midi_rt, rt->receiver_buffer + 2, rt->receiver_buffer[1]); } receiver_buffer is a 64-byte kzalloc() buffer (COMM_RECEIVER_BUFSIZE), so only 62 bytes follow the two-byte header. receiver_buffer[1] is a u8 the device chooses, so a device that answers with 0x10 and a length of 0xFF makes snd_rawmidi_receive() read 255 bytes starting two bytes into a 64-byte object. The bytes past the buffer are handed to userspace through the rawmidi read path. urb->actual_length is not consulted either, so a short transfer leaves both the type byte and the length byte at their previous values and the handler acts on stale data. The receiver URB is submitted from usb6fire_comm_init() at probe, so the read happens on plug with no user action; forwarding to userspace also needs a MIDI input substream open, since usb6fire_midi_in_received() only calls snd_rawmidi_receive() when rt->in is set. KASAN on 7.2.0-rc5 (arm64), single packet from an emulated device: BUG: KASAN: slab-out-of-bounds in snd_rawmidi_receive Read of size 255 at addr ffff000009f64682 by task bash/183 __asan_memcpy snd_rawmidi_receive usb6fire_midi_in_received [snd_usb_6fire] usb6fire_comm_receiver_handler [snd_usb_6fire] Allocated by task 11: usb6fire_comm_init [snd_usb_6fire] usb6fire_chip_probe [snd_usb_6fire] The buggy address is located 2 bytes inside of allocated 64-byte region [ffff000009f64680, ffff000009f646c0) Reject the event when the length exceeds the bytes that follow the header, and require the transfer to have delivered the header plus that many bytes. The receiver URB is submitted with a 64-byte transfer_buffer_length, so a genuine device cannot deliver an event longer than those 62 bytes and nothing valid is dropped. Discovered by XBOW, triaged by Baul Lee <baul.lee@xbow.com>
In the Linux kernel, the following vulnerability has been resolved: ALSA: aloop: Check card index validity at probe aloop driver blindly trusts that the given devptr->id value is within the proper card index range at probe. It's OK for the devices the driver itself creates at the module probe time, but if the device is bound manually via sysfs interface, this could be -1 as "none", and this leads to OOB access for index[] and other parameters. Add a sanity check for the card index and warn/correct it if it's a value out of the range.
In the Linux kernel, the following vulnerability has been resolved: ALSA: mpu401: Check card index validity at probe mpu401 driver blindly trusts that the given devptr->id value is within the proper card index range at probe. It's OK for the devices the driver itself creates at the module probe time, but if the device is bound manually via sysfs interface, this could be -1 as "none", and this leads to OOB access for index[] and other parameters. Add a sanity check for the card index and warn/correct it if it's a value out of the range.
In the Linux kernel, the following vulnerability has been resolved: ALSA: mts64: Check card index validity at probe Although mts64 driver has a check of the given devptr->id value, it doesn't check for a negative id, which is often given as "none" or such value when bound via sysfs. This may lead to OOB access for index[] and other parameters. Add a sanity check for the card index and warn/correct it if it's a value out of the range.
In the Linux kernel, the following vulnerability has been resolved: ALSA: portman2x4: Check card index validity at probe Although portman2x4 driver has a check of the given devptr->id value, it doesn't check for a negative id, which is often given as "none" or such value when bound via sysfs. This may lead to OOB access for index[] and other parameters. Add a sanity check for the card index and warn/correct it if it's a value out of the range.
In the Linux kernel, the following vulnerability has been resolved: ALSA: serial-u16550: Check card index validity at probe serial-u16550 driver blindly trusts that the given devptr->id value is within the proper card index range at probe. It's OK for the devices the driver itself creates at the module probe time, but if the device is bound manually via sysfs interface, this could be -1 as "none", and this leads to OOB access for index[] and other parameters. Add a sanity check for the card index and warn/correct it if it's a value out of the range.