{"id":"CVE-2026-99091","kind":"vulnerability","aliases":["CVE-2026-99091"],"related":[],"title":"Linux kernel: Bluetooth flushed commands report success instead of errors","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nBluetooth: MGMT: Fix status of pending commands flushed on power off\n\ncmd_complete_rsp() receives a struct cmd_lookup and, for a pending\ncommand that has no cmd_complete callback, falls through to\ncmd_status_rsp(cmd, data). cmd_status_rsp() reads data as a u8 *, so\nthe status sent comes from the first byte of match->sk's\nrepresentation, not from match->mgmt_status.\n\ncommit f53e1c9c726d (\"Bluetooth: MGMT: Fix possible crash on\nmgmt_index_removed\") changed the callers' data from &status to &match\nand updated the cmd_complete branch to match->mgmt_status, but left\nthe fall-through passing data unchanged.\n\nIn mgmt_index_removed() match->sk is NULL, so every such command is\nanswered with status 0x00 (Success) instead of\nMGMT_STATUS_INVALID_INDEX. In __mgmt_power_off() it is NULL too,\nunless a Set Powered command was pending: then settings_rsp() has\nstored that command's socket in match->sk, and the byte sent comes\nfrom the socket pointer. Either way the status the caller set,\nMGMT_STATUS_NOT_POWERED (or MGMT_STATUS_INVALID_INDEX when the device\nis being unregistered), is lost.\n\nPass the status the callers set.\n\nThis was seen on a QCA9377 laptop: bluetoothd's Start Discovery,\npending when the adapter was powered off, was answered with Command\nStatus 0x00 and no parameters, and bluetoothd 5.72 crashed on the\nzero-length success. BlueZ master has since added the length check\non that path (\"adapter: Fix crash on short start discovery reply\",\nBlueZ a734b06059cb), but userspace should still be told Not Powered,\nnot Success.\n\nTested on Ubuntu 7.0.0-31 with a virtual controller (hci_vhci):\npower off is held in HCI Write Scan Enable while a Start Discovery is\nsubmitted, and the flush answers it with status 0x00 before this\nchange and 0x0f (Not Powered) after.","summary":"When the Bluetooth adapter is powered off, pending management commands such as Start Discovery are answered with status 0x00 (Success) instead of Not Powered or Invalid Index, because the wrong value is read from a pointer. The result is a wrong status to userspace, which can crash bluetoothd on a zero-length success reply.","severity":"unknown","cvss":null,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["bluetooth","bluez","infrastructure","linux"],"not_affected_if":["no-bluetooth-adapter"],"how_to_check":"- unknown","fix":"- Apply the upstream Linux kernel fix that passes the caller's status through cmd_complete_rsp (commit f53e1c9c726d follow-up), and update to a kernel release containing it.\n- Update BlueZ to a release with the short start discovery reply length check (BlueZ a734b06059cb).","status":"open","published_at":"2026-10-26T15:16:47+00:00","first_seen_at":"2026-10-26T16:02:26+00:00","triaged_at":"2026-10-26T16:04:10+00:00","updated_at":"2026-10-26T16:04:10+00:00","category":"infrastructure","confidence":0.5,"triage":{"at":"2026-10-26T16:04:10Z","model":"claude-haiku-5-5","output":{"fix":"- Apply the upstream Linux kernel fix that passes the caller's status through cmd_complete_rsp (commit f53e1c9c726d follow-up), and update to a kernel release containing it.\n- Update BlueZ to a release with the short start discovery reply length check (BlueZ a734b06059cb).","reach":"local","title":"Linux kernel: Bluetooth flushed commands report success instead of errors","impact":"low","summary":"When the Bluetooth adapter is powered off, pending management commands such as Start Discovery are answered with status 0x00 (Success) instead of Not Powered or Invalid Index, because the wrong value is read from a pointer. The result is a wrong status to userspace, which can crash bluetoothd on a zero-length success reply.","category":"infrastructure","nicknames":[],"who_cares":["linux","bluetooth","bluez"],"confidence":0.5,"needs_auth":"unknown","worst_case":"bluetoothd crashes and clients receive wrong command status","analyst_note":"Check the kernel.org stable advisory and the affected kernel version range for CVE-2026-99091, and confirm the patched version.","how_to_check":"- unknown","needs_analyst":true,"not_affected_if":["no-bluetooth-adapter"],"who_is_affected":"Systems running Linux with Bluetooth adapters that use the mgmt interface, such as desktops and laptops running bluetoothd. The sources describe a crash in userspace on a zero-length success reply, and the sources do not say which kernel versions are affected or whether remote attackers can trigger it.","user_interaction":"unknown","needs_special_setup":"unknown","severity_if_unknown":"low"},"cost_usd":0.000686,"input_tokens":3950,"output_tokens":582,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'-2026':12A,15A '-31':351C '-99091':13A,16A '0x00':39B,205C,302C,380C '0x0f':385C '5.72':308C '7.0.0':350C 'a734b06059cb':336C 'adapt':24B,293C,327C 'add':320C 'answer':36B,202C,298C,376C 'bluetooth':3A,17B,23B,86C,153C 'bluetoothd':68B,286C,307C 'bluez':18B,316C,335C 'branch':175C 'byte':140C,242C 'callback':115C 'caller':164C,254C,277C 'chang':162C,383C 'check':323C 'cmd':97C,103C,113C,119C,122C,124C,173C 'come':136C,244C 'command':5A,30B,92C,109C,200C,225C,234C,300C 'commit':151C 'complet':98C,114C,174C 'control':355C 'crash':67B,157C,309C,329C 'cve':11A,14A 'data':123C,128C,165C,187C 'devic':267C 'discoveri':34B,289C,333C,370C 'either':249C 'enabl':366C 'error':10A 'everi':198C 'f53e1c9c726d':152C 'fall':116C,184C 'fall-through':183C 'first':139C 'fix':88C,155C,328C 'flush':4A,93C,375C 'follow':81C 'hci':356C,363C 'held':361C 'index':47B,160C,191C,212C,264C 'infrastructur':19B 'instead':8A,41B,207C 'invalid':46B,211C,263C 'kernel':2A,79C 'laptop':285C 'left':181C 'length':73B,314C,322C 'linux':1A,20B,78C 'lookup':104C 'lost':272C 'manag':29B 'master':317C 'match':142C,148C,169C,177C,193C,238C 'mgmt':87C,149C,154C,159C,178C,190C,209C,214C,256C,261C 'null':196C,219C 'paramet':305C 'pass':186C,273C 'path':326C 'pend':28B,91C,108C,227C,290C 'pointer':56B,248C 'possibl':156C 'power':26B,44B,95C,215C,224C,259C,295C,344C,358C,387C 'qca9377':284C 'read':53B,127C 'receiv':100C 'remov':161C,192C 'repli':75B,334C 'report':6A 'represent':145C 'resolv':85C 'result':58B 'rsp':99C,121C,126C,230C 'scan':365C 'seen':281C 'sent':135C,243C 'set':223C,229C,255C,278C 'short':331C 'sinc':319C 'sk':143C,194C,239C 'socket':236C,247C 'start':33B,288C,332C,369C 'status':38B,62B,89C,120C,125C,134C,150C,167C,179C,204C,210C,252C,257C,262C,275C,301C,379C 'still':340C 'store':232C 'struct':102C 'submit':372C 'success':7A,40B,74B,206C,315C,346C 'test':347C 'told':342C 'u8':131C 'ubuntu':349C 'unchang':188C 'unless':221C 'unregist':270C 'updat':171C 'userspac':64B,338C 'valu':51B 'vhci':357C 'virtual':354C 'vulner':82C 'way':250C 'write':364C 'wrong':50B,61B 'zero':72B,313C 'zero-length':71B,312C","nicknames":[],"mitigations":[],"who_affected":"Systems running Linux with Bluetooth adapters that use the mgmt interface, such as desktops and laptops running bluetoothd. The sources describe a crash in userspace on a zero-length success reply, and the sources do not say which kernel versions are affected or whether remote attackers can trigger it.","tier":"C","verified":{},"analyst_revisit_at":null,"first_public_at":"2026-10-26T15:16:47+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-99091","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-99091","severity":null,"published_at":"2026-10-26T15:16:47+00:00","modified_at":"2026-10-26T15:16:47+00:00","first_seen_at":"2026-10-26T16:02:26+00:00","fetched_at":"2026-10-26T16:02:26+00:00","note":null}],"documents":[],"notice":"Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai"}