[CVE-2026-98737][MODERATE 7.0] Bluetooth: hci_sync: Fix inquiry cache use-after-free

[CVE-2026-98737][MODERATE 7.0] Bluetooth: hci_sync: Fix inquiry cache use-after-free

[AL-KERNEL]

CVE: CVE-2026-98737
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: Bluetooth: hci_sync: Fix inquiry cache use-after-free
Commit: 52edf1c0e59dea37b8368c4a2559dca96b2fdfea
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=52edf1c0e59dea37b8368c4a2559dca96b2fdfea
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98737
Analysis date: Mon, 19 Oct 2026 02:56:26 -0400
ActionableScore: 5
ActionableScore lower bound: 4
Actionable bucket: Actionable Moderate / Strong Important candidate requiring manual review
Manual review required: YES

Summary:
A race between the Bluetooth HCI command worker and a concurrent HCIINQUIRY cache flush causes a confirmed slab use-after-free whose stale inquiry fields feed active connection state and HCI commands, requiring manual review for delegated-interface reachability and post-reclaim leverage.

======================================================================
ABOUT THIS REPORT
======================================================================

The original Linux kernel CVE announcement for CVE-2026-98737 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98737

The original announcement does not normally provide a security severity
estimate, CVSS assessment, or enough information to determine whether the
reported kernel bug represents a practically relevant security issue.

This report was generated by AL-KERNEL, an AI-assisted Linux kernel
vulnerability analysis system developed by Alexander Larkin. It combines
an autonomous classifier with LLM-assisted technical analysis and a
separate ActionableScore mechanism.

The purpose of this report is to prioritize Linux kernel CVEs before
manual review, identify cases that require prompt investigation, and
support automatic closure of issues that are unlikely to have meaningful
security impact.

Published priority for this report: MODERATE 7.0
Manual review required: YES

A detailed explanation of the methodology and priority rules is included
at the end of this message.

======================================================================
AL-KERNEL CLASSIFICATION RESULT
======================================================================

CVE-2026-98737	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-98737 CHECK WITH IMPACT FROM ORIG NN IMPORTANT | AEGIS IMPACT MODERATE_KPANIC AEGIS_AS=4 LEGACY=MODERATE_KPANIC RESULT=SAME (final reconciliation agrees; DECISION: both MODERATE_KPANIC)	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H';*CWE-416;CWE-362;CWE-667;Other CVSS 'AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H';BEST CVSS score: '7';DESCR 'A race in the Bluetooth HCI inquiry cache can cause a slab use after free because synchronous command workers look up an inquiry entry and read its fields without holding hdev lock. A concurrent local HCIINQUIRY ioctl can flush the cache and free the entry before hci_acl_create_conn_sync or hci_stop_discovery_sync finishes using it. KASAN confirms reads from freed slab memory and the immediate demonstrated consequence is a kernel crash. For the CVSS the PR:L reflects that a local process needs access to Bluetooth HCI control operations which may require delegated CAP_NET_RAW or CAP_NET_ADMIN but does not necessarily require unrestricted root access. AC:H applies because reliable triggering depends on winning a race window between cache lookup and cache removal. The issue is not directly reachable by a remote network peer alone. A nearby Bluetooth device may affect inquiry cache contents but local control is still needed to flush the cache and drive the competing synchronous path. Impact is at least a local denial of service via kernel crash and in the paranoid worst case may include confidentiality or integrity impact due to reuse of freed slab memory. Practical privilege escalation is not demonstrated and requires manual review of object reuse and attacker control. Automated impact decision path. The original SimpleNet classifier selected IMPORTANT. The primary LLM estimated paranoid CVSS 7 and the production ActionableScore returned 5 with recommendation Actionable Moderate / Strong Important candidate requiring manual review. The production rules then evaluated CVSS thresholds and ActionableScore thresholds. Recorded rule transitions were KP and FD specialist checks stage IMPORTANT to MODERATE_KPANIC because FD false positive review allowed a downgrade. The legacy KP specialist produced a review artifact and its decision was processed by the applicable KP rules. The legacy FD or DH downgrade specialist produced a review artifact and its decision was processed by the applicable downgrade and preservation rules. Exact triggered rule markers were DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFDH. The independent weaker Aegis shadow evaluated CVSS 2.5 and ActionableScore 4 with primitive class CLASS_B and manual review REQUIRED. Its independently reconciled impact was MODERATE_KPANIC. The shadow agreed with the final production impact and aggregation did not change that impact. The final reported impact is MODERATE_KPANIC.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate / Strong Important candidate requiring manual review (with actual score 5)	MAYBE	REMOTE READ OOB LOCK DANGER RACE UAF NETWORK LINUS KPANIC  DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFDH	-	-	checked

======================================================================
ACTIONABLESCORE ANALYSIS
======================================================================

ActionableScore=5
ActionableScoreLower=4
PrimitiveClass=CLASS_B
RealUAF=YES
HostAdminRequired=UNKNOWN
NamespaceScopedPrivilege=UNKNOWN
ImportantCandidate=YES
ManualReview=REQUIRED
AutoDowngradeAllowed=NO
RequiresHostUID0=UNKNOWN
AuthorizationDecisionBypass=NO
ProtectedResourceImpact=NONE
B4PendingReview=NO
B4PendingConfirmed=NO

1. ActionableScore

* Raw conservative score: 4
* Conservative score: 4
* Raw paranoid score: 5
* Paranoid score: 5
* No floor or cap changed either score.
* Final recommended bucket: **Actionable Moderate / Strong Important candidate requiring manual review**::

2. Signal breakdown

Conservative signals:

* +2 Real UAF. The inquiry entry is freed by `hci_inquiry_cache_flush()` while the command worker retains and dereferences the same pointer.
* +1 Reliable kernel crash or strong DoS. KASAN confirms slab use-after-free in two active command-worker paths. An Oops or crash is a realistic consequence.
* +1 Attacker-influenced lifetime transition. The vulnerable sequence specifically involves a concurrent `HCIINQUIRY` ioctl flushing the cache while connection or discovery processing consumes an entry. The operation can be repeated by an actor with the required HCI access.

Paranoid-only signal:

* +1 Structured stale-object leverage. Freed inquiry metadata is consumed by active processing to populate connection state and prepare create-connection or remote-name-cancel commands. This includes copying `dev_class` into an active connection object and using a cached address in a privileged HCI command. Control over useful slab replacement is unresolved, so this remains Class B rather than Class A.

Not awarded:

* No remote-reachability bonus. A nearby Bluetooth peer may influence inquiry-result data, but it cannot independently perform the local cache-flush ioctl.
* No strong exploitation bonus. Controlled reclaim, attacker-selected replacement objects, arbitrary writes, callbacks, and refcount takeover are not demonstrated.
* No confidentiality or integrity consequence bonus. Stale metadata reaches active state, but disclosure, protected-state modification, or privilege escalation is not established.
* No extra availability point. The same potential kernel crash is counted only once.
* No race penalty. The timing window exists, but the supplied evidence does not establish a one-shot or externally uncontrollable race.
* No repeated-release signal. Neither vulnerable path performs a second put, free, or refcount mutation through the dangling entry.
* B5-R and B3-PENDING do not apply. There is no retry-accumulated OOB write or shared-backing write.
* B4-PENDING does not apply. No authorization decision changes from DENY to ALLOW.

Repeated-release checks:

* First lifetime operation can free the entry: yes, `hci_inquiry_cache_flush()` frees it.
* Exact reference survives into later use: yes.
* Later operation mutates its lifetime or refcount: no, it reads fields and copies data.
* Concrete repeated release or refcount +1 awarded: no.
* Paranoid repeated-refcount +1 awarded: no.
* Floor 5 applied: no. The paranoid score reaches 5 through active structured consumption, not a scoring floor.

3. Reachability analysis

The demonstrated trigger combines an active Bluetooth command-worker operation with a concurrent local `HCIINQUIRY` ioctl that flushes the inquiry cache. The complete HCI socket and ioctl permission chain is not shown. Depending on the kernel and socket channel, access may involve `CAP_NET_RAW`, `CAP_NET_ADMIN`, or a privileged Bluetooth service.

Below-host-admin access through a delegated Bluetooth service or reduced capability is credible but not established for the complete sequence. Therefore `HostAdminRequired` and `RequiresHostUID0` remain `UNKNOWN`. The evidence does not establish that all required HCI operations are evaluated in an attacker-created user or network namespace, so `NamespaceScopedPrivilege` is also `UNKNOWN`.

A Bluetooth controller and an active inquiry cache are prerequisites. These are common Bluetooth conditions rather than rare hardware requirements. A nearby peer can influence inquiry-result metadata, but local HCI control is still required to flush the cache and drive the competing command path. Repeated attempts appear feasible once the required local HCI access is available, so no hard-race deduction is applied.

4. Primitive and severity interpretation

The exact primitive is a race-driven slab use-after-free of `struct inquiry_entry`. This is active structured storage, not a fixed DMA buffer or a cleanup-only pointer. The stale fields participate in ongoing connection and discovery processing.

In `hci_acl_create_conn_sync()`, inquiry metadata is read and `dev_class` is copied into an active connection object. In `hci_stop_discovery_sync()`, the stale address is passed into remote-name cancellation. The privileged downstream consumers are the kernel HCI command worker and Bluetooth controller command path. The unresolved boundary is whether a below-host-admin actor with delegated HCI access can reliably combine cache flushing, inquiry-data influence, and command scheduling.

The demonstrated consequence is an invalid post-free read with realistic kernel crash or DoS. The paranoid concern is that reclaimed slab contents could influence active connection metadata or command parameters. Controlled replacement, arbitrary kernel read or write, privilege escalation, and a purely remote trigger are not established.

This is `CLASS_B` because a real active-object lifetime violation exists but no strong Class A exploitation primitive is proven. `ImportantCandidate=YES` preserves the unresolved active-state and delegated-interface question for review. `AutoDowngradeAllowed=NO` because automatic crash-only classification would discard the concrete active structured UAF and its downstream state consumption. This is a review candidate, not proof of Important severity.

5. One-sentence report phrase

A race between the Bluetooth HCI command worker and a concurrent HCIINQUIRY cache flush causes a confirmed slab use-after-free whose stale inquiry fields feed active connection state and HCI commands, requiring manual review for delegated-interface reachability and post-reclaim leverage.

6. Manual review recommendation

MANUAL CHECK REQUIRED

Confirm the exact HCI socket capability checks, whether a delegated Bluetooth client can drive both sides of the race, and whether inquiry-entry slab reuse can controllably influence `conn->dev_class`, connection parameters, or remote-name cancellation.

7. Call-site confidence

Call-site confidence: high

======================================================================
UPSTREAM PATCH SUMMARY
======================================================================

Patch: Bluetooth: hci_sync: Fix inquiry cache use-after-free
Commit: 52edf1c0e59dea37b8368c4a2559dca96b2fdfea
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=52edf1c0e59dea37b8368c4a2559dca96b2fdfea

Commit description:

The sync command worker holds hdev->req_lock, but inquiry-cache updates
and flushes use hdev->lock. Both hci_acl_create_conn_sync() and
hci_stop_discovery_sync() look up entries and read their fields without
taking hdev->lock.

After either lookup returns, a concurrent HCIINQUIRY ioctl can acquire
hdev->lock and flush the cache, freeing the entry. The worker then reads
the freed entry while preparing a create-connection or remote-name-cancel
command. The list traversal also races with cache updates and removal.

KASAN reported these accesses:

  BUG: KASAN: slab-use-after-free in hci_acl_create_conn_sync+0x5f1/0x650
  Workqueue: hci0 hci_cmd_sync_work
  Call Trace:
   hci_acl_create_conn_sync+0x5f1/0x650
   hci_cmd_sync_work+0x13c/0x290

  Allocated by task 90:
   hci_inquiry_cache_update+0x3e6/0x7d0
   hci_inquiry_result_evt+0x3cb/0x560

  Freed by task 93:
   hci_inquiry_cache_flush+0x111/0x2b0
   hci_inquiry+0x2f2/0x780
   hci_sock_ioctl+0x269/0x5f0

  BUG: KASAN: slab-use-after-free in hci_stop_discovery_sync+0x3b1/0x3c0
  Workqueue: hci0 hci_cmd_sync_work
  Call Trace:
   hci_stop_discovery_sync+0x3b1/0x3c0
   hci_cmd_sync_work+0x173/0x300

  Allocated by task 86:
   hci_inquiry_cache_update+0x483/0x940
   hci_inquiry_result_evt+0x3cb/0x560

  Freed by task 91:
   hci_inquiry_cache_flush+0x13e/0x2f0
   hci_inquiry+0x2f2/0x780
   hci_sock_ioctl+0x269/0x5f0

Hold hdev->lock across each lookup and all reads from its result. Copy the
remote address before unlocking so discovery cancellation does not retain
a cache entry pointer. Release the lock before sending synchronous HCI
commands, since their completion handlers may need the same lock.

Fixes: cf75ad8 ("Bluetooth: hci_sync: Convert MGMT_SET_POWERED")
Fixes: 4534009 ("Bluetooth: hci_conn: Only do ACL connections sequentially")
Cc: stable@vger.kernel.org
Signed-off-by: Chengfeng Ye <nicoyip.dev@gmail.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Changed files:
  net/bluetooth/hci_sync.c

Diff excerpt:

Not included in this email. See the upstream URL for the full patch.

Full patch:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=52edf1c0e59dea37b8368c4a2559dca96b2fdfea

======================================================================
DETAILED REPORT METHODOLOGY
======================================================================

The original Linux kernel CVE announcement for CVE-2026-98737 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98737

The original CVE announcement normally does not include a security-level
estimate. In particular, it may not contain a CVSS assessment, an impact
level, or enough information to determine whether the reported bug is a
practically relevant security issue. One purpose of this parallel CVE list
is to provide that missing technical and prioritization information.

The original goal of the AL-KERNEL project was to prioritize Linux kernel
CVE analysis automatically before manual review. The system can also help
identify non-security issues that may be suitable for automatic closure.

This report was generated by AL-KERNEL, an AI-assisted Linux kernel
vulnerability analysis system developed by Alexander Larkin.

The first analysis stage combines an autonomous classifier with additional
LLM-based analysis. The autonomous classifier runs locally on a CPU and is
based on a backpropagation neural network. Together, these mechanisms
produce a technical vulnerability description, identify likely weakness
types, estimate CVSS severity, and provide input for ActionableScore.

Two CVSS estimates are retained because incomplete kernel vulnerability
information often permits more than one defensible interpretation:

  Conservative CVSS vector: AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H
  The Best / paranoid CVSS vector: AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H
  The Best / paranoid CVSS score: 7

The conservative vector represents a lower-impact interpretation.
The Best/paranoid vector intentionally represents a plausible upper-bound
interpretation and should not automatically be treated as demonstrated
real-world impact.

CVSS may also need to be adjusted for a particular Linux deployment,
because actual reachability, privileges, enabled kernel configuration,
hardware, namespaces, exposed device nodes, and other environmental
conditions can differ significantly between systems.

A separate ActionableScore mechanism evaluates practical remediation
urgency. Its analysis may include reachability, attack prerequisites,
subsystem exposure, memory-corruption characteristics, denial-of-service
reliability, and possible confidentiality, integrity, or
privilege-escalation impact.

  Conservative ActionableScore: 4
  Paranoid ActionableScore: 5

The final base severity is taken directly from the second tab-separated
field of the AL-KERNEL classification result. ActionableScore does not
replace or independently override that final AL-KERNEL decision, and
if ActionableScore adjusted impact level of ALKERNEL, then you would see
self-readable flags above like INCREASED_TO_HIGH_BASED_ON_ACTIONABLESCOREHIGHEREQTHAN7.

For an AL-KERNEL result of MODERATE, this report uses the following
additional presentation split:

  ActionableScore below 5   -> MODERATE REGULAR
  ActionableScore 5 or more -> MODERATE 7.0

The distinction between MODERATE REGULAR and MODERATE 7.0 makes it
possible to identify Moderate issues that should receive manual analysis
and fixes before lower-priority MODERATE REGULAR issues. In many cases,
MODERATE REGULAR fixes may wait for a later rebase or routine update.

There is one override in which MODERATE REGULAR becomes MODERATE 7.0
even when the ActionableScore is below 5. When the AL-KERNEL result
contains the KPANIC flag, a MODERATE result is always presented as
MODERATE 7.0. The KPANIC flag selected with few regexps without
usage of AI at all, so it helps to detect cases when Kernel Crash happens
and similar (to filter False-Negative results from the LLM usage).

KPANIC indicates that a reliable kernel crash, kernel panic, or similarly
serious kernel availability impact was identified by the classification
workflow.

AL-KERNEL base severity for this report: MODERATE
KPANIC detected for this report: YES
Published priority for this report (same as in Subject): MODERATE 7.0

These results are intended to support engineering triage. They are
machine-generated estimates, and cases marked for manual review should
be validated by a human security engineer before final disposition.
For more info read docs linked from here: https://kernelcve.org/
(and you can submit you own patch there to generate such a report
for non-existant CVE-id yet).

Note that in many cases this AI tool selects higher severity, than
real is (means you can expect Importants instead of Moderate 7.0 or
Moderates 7.0 instead of regular Moderates). If you see such cases,
please use reply email interface to add additional manual analyses
info to this particular CVE.
And please, please, let me know when you see Lows instead of Importants
or Important instead of Low (because particular for such cases I
need to tune this AI tool to make it better for this one and next similar).
My contact email for such notifications is alexanjelausa@gmail.com (and both
send reply to CVE record itself too and see "reply" button below for howto reply).