{"id":"CVE-2026-98615","kind":"vulnerability","aliases":["CVE-2026-98615"],"related":[],"title":"Linux kernel: fbnic firmware mailbox flush can leave requests stuck","description":"In the Linux kernel, the following vulnerability has been resolved:\n\neth: fbnic: Set AW_FLUSH_MODE alongside AW_FLUSH when flushing the mailbox\n\nWhen tearing down the FW mailbox Rx ring, fbnic_mbx_reset_desc_ring()\nwrites AW_CFG with FLUSH set and everything else, BME included, cleared.\nClearing BME halts the device's writes to the host but leaves the staged\nrequests parked in the PUL write pipeline rather than draining them, so\non the write path FLUSH alone never terminates the outstanding requests\nand the flush the firmware waits on never completes.\n\nAdd the FLUSH_MODE definition and set both bits so the staged writes\ndrain out of the pipeline on their own. BME stays cleared, so nothing\nlands on the host; it is restored later in fbnic_mbx_init_desc_ring()\nwhen the ring is rebuilt, once the outstanding writes are gone.\n\nThe read path is unaffected. AR_CFG has no equivalent mode bit and\nAR_FLUSH terminates the outstanding reads by itself, so it is left as\nis.\n\nBoth writes remain plain stores rather than read-modify-writes. That is\ndeliberate: the matching write in fbnic_mbx_init_desc_ring() restores\nBME and the TLP attributes, and clears both flush bits as a side effect.","summary":"The fbnic network driver in the Linux kernel did not fully drain staged firmware mailbox writes when tearing down the mailbox Rx ring, so the firmware flush could never complete. The sources describe a stall in the write path; they do not describe code execution or data exposure.","severity":"unknown","cvss":null,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["fbnic","linux-kernel","package","self-hosted"],"not_affected_if":["no-fbnic-hardware"],"how_to_check":"- Run `uname -r` and check whether the running kernel includes the fbnic AW_FLUSH_MODE fix (unknown: the sources do not give the fixed version)\n- Check whether the host uses Meta fbnic network hardware (unknown: the sources do not say how to identify it)","fix":"- Update the Linux kernel to a release that includes the upstream fbnic AW_FLUSH_MODE commit (unknown: the sources do not list the fixed version)","status":"open","published_at":"2026-10-18T16:17:14+00:00","first_seen_at":"2026-10-23T21:57:16+00:00","triaged_at":"2026-10-25T05:21:21+00:00","updated_at":"2026-10-25T05:21:21+00:00","category":"package","confidence":0.5,"triage":{"at":"2026-10-25T05:21:21Z","model":"claude-haiku-5-5","output":{"fix":"- Update the Linux kernel to a release that includes the upstream fbnic AW_FLUSH_MODE commit (unknown: the sources do not list the fixed version)","reach":"unknown","title":"Linux kernel: fbnic firmware mailbox flush can leave requests stuck","impact":"low","summary":"The fbnic network driver in the Linux kernel did not fully drain staged firmware mailbox writes when tearing down the mailbox Rx ring, so the firmware flush could never complete. The sources describe a stall in the write path; they do not describe code execution or data exposure.","category":"package","nicknames":[],"who_cares":["linux-kernel","fbnic","self-hosted"],"confidence":0.5,"needs_auth":"unknown","worst_case":"firmware mailbox flush stalls, device teardown hangs","analyst_note":"Confirm the fixed kernel version and the upstream commit from the NVD entry and the kernel stable tree; the sources give no version or severity.","how_to_check":"- Run `uname -r` and check whether the running kernel includes the fbnic AW_FLUSH_MODE fix (unknown: the sources do not give the fixed version)\n- Check whether the host uses Meta fbnic network hardware (unknown: the sources do not say how to identify it)","needs_analyst":true,"not_affected_if":["no-fbnic-hardware"],"who_is_affected":"Only hosts that run the fbnic Ethernet driver, which is used with Meta fbnic network adapters. Systems with other NICs are not affected by this driver flaw. The sources do not describe the trigger or the effect beyond a stalled firmware flush.","user_interaction":"no","needs_special_setup":"unknown","severity_if_unknown":"unknown"},"cost_usd":0.000672,"input_tokens":3689,"output_tokens":606,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'-2026':12A,15A '-98615':13A,16A 'add':167C 'alon':152C 'alongsid':89C 'ar':223C,231C 'attribut':273C 'aw':86C,90C,110C 'bit':175C,229C,278C 'bme':118C,122C,188C,269C 'cfg':111C,224C 'clear':120C,121C,190C,275C 'code':68B 'complet':54B,166C 'could':52B 'cve':11A,14A 'data':71B 'definit':171C 'deliber':258C 'desc':107C,205C,266C 'describ':57B,67B 'devic':125C 'drain':36B,144C,180C 'driver':28B 'effect':282C 'els':117C 'equival':227C 'eth':83C 'everyth':116C 'execut':69B 'exposur':72B 'fbnic':3A,17B,26B,84C,104C,202C,263C 'firmwar':4A,38B,50B,162C 'flush':6A,51B,87C,91C,93C,113C,151C,160C,169C,232C,277C 'follow':78C 'fulli':35B 'fw':100C 'gone':217C 'halt':123C 'host':24B,130C,196C 'includ':119C 'init':204C,265C 'kernel':2A,20B,32B,76C 'land':193C 'later':200C 'leav':8A,132C 'left':242C 'linux':1A,19B,31B,75C 'linux-kernel':18B 'mailbox':5A,39B,45B,95C,101C 'match':260C 'mbx':105C,203C,264C 'mode':88C,170C,228C 'modifi':254C 'network':27B 'never':53B,153C,165C 'noth':192C 'outstand':156C,214C,235C 'packag':21B 'park':136C 'path':63B,150C,220C 'pipelin':141C,184C 'plain':248C 'pul':139C 'rather':142C,250C 'read':219C,236C,253C 'read-modify-writ':252C 'rebuilt':211C 'remain':247C 'request':9A,135C,157C 'reset':106C 'resolv':82C 'restor':199C,268C 'ring':47B,103C,108C,206C,209C,267C 'rx':46B,102C 'self':23B 'self-host':22B 'set':85C,114C,173C 'side':281C 'sourc':56B 'stage':37B,134C,178C 'stall':59B 'stay':189C 'store':249C 'stuck':10A 'tear':42B,97C 'termin':154C,233C 'tlp':272C 'unaffect':222C 'vulner':79C 'wait':163C 'write':40B,62B,109C,127C,140C,149C,179C,215C,246C,255C,261C","nicknames":[],"mitigations":[],"who_affected":"Only hosts that run the fbnic Ethernet driver, which is used with Meta fbnic network adapters. Systems with other NICs are not affected by this driver flaw. The sources do not describe the trigger or the effect beyond a stalled firmware flush.","tier":"C","verified":{},"analyst_revisit_at":null,"first_public_at":"2026-10-18T16:17:14+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-98615","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-98615","severity":null,"published_at":"2026-10-18T16:17:14+00:00","modified_at":"2026-10-19T10:19:25+00:00","first_seen_at":"2026-10-23T21:57:16+00:00","fetched_at":"2026-10-23T22:28:56+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"}