Bug 2502274 (CVE-2026-63874)

Summary: CVE-2026-63874 kernel: net: mctp: usb: fix race between urb completion and rx_retry cancellation
Product: [Other] Security Response Reporter: OSIDB Bzimport <bzimport>
Component: vulnerabilityAssignee: Product Security <prodsec-ir-bot>
Status: NEW --- QA Contact:
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: unspecifiedCC: rhel-process-autobot, watson-tool-maintainers
Target Milestone: ---Keywords: Security
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: ---
Doc Text:
A flaw was found in the Linux kernel's USB Management Component Transport Protocol (MCTP) driver. A race condition exists between the completion of a USB Request Block (URB) and the cancellation of retry operations during the driver's shutdown process. This timing issue can cause a URB to be re-queued, potentially leading to unexpected system behavior or resource management issues. This vulnerability could impact system stability.
Story Points: ---
Clone Of: Environment:
Last Closed: Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description OSIDB Bzimport 2026-07-19 15:01:26 UTC
In the Linux kernel, the following vulnerability has been resolved:

net: mctp: usb: fix race between urb completion and rx_retry cancellation

It's possible that sequencing between setting ->stopped and cancelling
the rx_retry work (in ndo_stop) could leave us with an urb queued:

    T1: ndo_stop                  T2: rx_retry_work
    ------------                  ----------------
                                  LD: ->stopped => false
    ST: ->stopped <= true
    usb_kill_urb()
                                  mctp_usb_rx_queue()
                                    usb_submit_urb()
    cancel_delayed_work_sync()

That urb completion can then re-schedule rx_retry_work.

Strenghen the sequencing between the stop (preventing another requeue)
and the cancel by updating both atomically under a new rx lock. After
setting ->rx_stopped, and cancelling pending work, we know that the
requeue cannot occur, so all that's left is killing any pending urb.