{"id":"AZL-109431","summary":"CVE-2026-98473 affecting package kernel 6.6.157.1-1","details":"In the Linux kernel, the following vulnerability has been resolved:\n\ns390/vfio-ap: fix KVM GISC and page leak when queue removed from host config\n\nThree related problems exist in the handling of KVM interrupt and page\nresources when a queue is removed from the host's AP configuration\nwhile assigned to a mediated device (mdev).\n\nProblem 1:\n~~~~~~~~~\nAP_RESPONSE_Q_NOT_AVAIL not handled in vfio_ap_mdev_reset_queue()\n\nWhen the AP bus removes a queue device whose adapter or domain has\nbeen removed from the host's AP configuration,\nvfio_ap_mdev_remove_queue() is called. If the queue is still in the\nhost's AP configuration at that point, it calls\nvfio_ap_mdev_reset_queue(), which issues a PQAP(ZAPQ). Since the\nadapter is already gone from the host configuration, ap_zapq() returns\nAP_RESPONSE_Q_NOT_AVAIL (0x01). This response code is not handled in\nvfio_ap_mdev_reset_queue()'s switch statement and falls through to\nthe default case, which issues a WARN but does not call\nvfio_ap_free_aqic_resources(). As a result, if IRQ handling was\nenabled for the queue by the guest, the KVM GISC registration and\nthe pinned guest page holding the notification indicator byte (NIB)\nare both leaked.\n\nThis is fixed by adding AP_RESPONSE_Q_NOT_AVAIL to the same case as\nAP_RESPONSE_DECONFIGURED and AP_RESPONSE_CHECKSTOPPED in\nvfio_ap_mdev_reset_queue(). Like those response codes, Q_NOT_AVAIL\nindicates the queue is not operational and no further reset attempts\nare possible; the correct action is to free the IRQ resources\nimmediately.\n\nProblem 2:\n~~~~~~~~~\n AP_RESPONSE_Q_NOT_AVAIL not handled in apq_status_check()\n\nIn vfio_ap_mdev_reset_queue(), there are four cases that indicate a queue\nreset has not yet completed, in which case apq_reset_check() is queued to\na work queue to verify completion of the reset operation. This function\nuses the PQAP(TAPQ) function to get the queue's status and calls\napq_status_check() to verify whether the reset has completed, failed or\nneeds to be executed again. As described in Problem #1 above,\napq_reset_check() does not specifically check for AP_RESPONSE_Q_NOT_AVAIL,\nthereby potentially leaking KVM GISC registration and the pinned guest page\nholding the NIB.\n\nThis is fixed by adding a case statement for AP_RESPONSE_Q_NOT_AVAIL to\napq_status_check() and returning -ENODEV for that case. The caller,\napq_reset_check() will then check for this return code and call\nvfio_ap_free_aqic_resources() to prevent the leak.\n\nProblem 3:\n~~~~~~~~~\nvfio_ap_free_aqic_resources() leaks saved_isc when kvm is NULL\n\nvfio_ap_free_aqic_resources() guards the call to\nkvm_s390_gisc_unregister() with:\n\n    if (q-\u003esaved_isc != VFIO_AP_ISC_INVALID &&\n        !WARN_ON(!(q-\u003ematrix_mdev && q-\u003ematrix_mdev-\u003ekvm)))\n\nIf matrix_mdev-\u003ekvm is NULL -- which can happen when\nvfio_ap_mdev_unset_kvm() has already run and cleared kvm before a\nsubsequent cleanup path reaches this function -- the WARN_ON fires\nand the entire block is skipped. This leaves q-\u003esaved_isc set to a\nnon-invalid value, creating a potential double-free on any subsequent\ncall to this function.\n\nWhen kvm is NULL the KVM guest is already torn down, so\nkvm_s390_gisc_unregister() need not and cannot be called; however,\nq-\u003esaved_isc must always be cleared. Fix this by separating the\nkvm_s390_gisc_unregister() call from the q-\u003esaved_isc reset. The\nWARN_ON now guards only the genuinely impossible case of matrix_mdev\nbeing NULL. A NULL kvm is handled gracefully by skipping only the\nunregister call, and q-\u003esaved_isc = VFIO_AP_ISC_INVALID is set\nunconditionally whenever saved_isc was not already invalid.\n\nAdditionally, add an else clause to the host-config check in\nvfio_ap_mdev_remove_queue() to call vfio_ap_free_aqic_resources()\ndirectly when the queue is not in the host's AP configuration. This\nserves as a backstop: when the AP bus fires the driver .remove\ncallback after an adapter is removed from the host config, the queue\nis by definition no longer addressable, so vfio_ap_mdev_reset_queue\n---truncated---","modified":"2026-10-19T14:16:50.136345404Z","published":"2026-10-18T16:16:57Z","upstream":["CVE-2026-98473"],"references":[{"type":"WEB","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-98473"}],"affected":[{"package":{"name":"kernel","ecosystem":"Azure Linux:3","purl":"pkg:rpm/azure-linux/kernel"},"ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"last_affected":"6.6.157.1-1"}]}],"database_specific":{"source":"https://github.com/microsoft/AzureLinuxVulnerabilityData/blob/main/osv/AZL-109431.json"}}],"schema_version":"1.9.0"}