{"id":"CVE-2026-98521","kind":"vulnerability","aliases":["CVE-2026-98521"],"related":[],"title":"Linux kernel i915 GPU driver: RCU teardown order causes endless loop","description":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/i915: fix incorrect RCU teardown order\n\ni915_gem_busy_ioctl uses dma_resv_for_each_fence_unlocked() to iterate\nover the fences in an GEM object without holding a reference but only\nthe RCU read side lock.\n\nWhat can happen here is that the GEM object is destroyed concurrently\nwhile i915_gem_busy_ioctl is still running. This won't free the GEM\nobjects memory, but still drops all the dma_fence references.\n\nNow when dma_resv_for_each_fence_unlocked() sees a destroyed dma_fence it\nassumes that a new fence list was installed and re-starts the loop.\n\nBut in the case of a destroyed GEM object a new fence list is never\ninstalled, only the old one freed and therefore the iteration never\nfinishes resulting in an endless loop.\n\nThe solution is to drop the fence references only after the RCU grace\nperiod.\n\nThe fixes tag is not necessary the patch introducing the problem, but the\none making it so worse that we need to address it.\n\nThis problem was pointed out by Sashiko-bot.\n\n(cherry picked from commit 5113479556025093bf8133bb2dcaa33be2d50921)","summary":"In the Linux i915 graphics driver, the busy ioctl can loop forever when a GEM object is destroyed while the ioctl is iterating its fences. The sources describe a denial-of-service style hang on the kernel side, not code execution.","severity":"unknown","cvss":null,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["i915","infrastructure","intel-graphics","kernel","linux"],"not_affected_if":["i915-not-loaded","no-i915-hardware"],"how_to_check":"- Check the running kernel with `uname -r` and compare against the fixed stable kernel releases listed in the upstream commit 5113479556025093bf8133bb2dcaa33be2d50921\n- Confirm the i915 driver is loaded with `lsmod | grep i915`","fix":"- Update to a Linux kernel stable release that includes upstream commit 5113479556025093bf8133bb2dcaa33be2d50921 (exact fixed version unknown from sources)\n- Reboot after the kernel update so the fixed i915 module is loaded","status":"open","published_at":"2026-10-18T16:17:03+00:00","first_seen_at":"2026-10-23T21:57:15+00:00","triaged_at":"2026-10-25T05:25:12+00:00","updated_at":"2026-10-25T05:25:12+00:00","category":"infrastructure","confidence":0.5,"triage":{"at":"2026-10-25T05:25:12Z","model":"claude-haiku-5-5","output":{"fix":"- Update to a Linux kernel stable release that includes upstream commit 5113479556025093bf8133bb2dcaa33be2d50921 (exact fixed version unknown from sources)\n- Reboot after the kernel update so the fixed i915 module is loaded","reach":"local","title":"Linux kernel i915 GPU driver: RCU teardown order causes endless loop","impact":"low","summary":"In the Linux i915 graphics driver, the busy ioctl can loop forever when a GEM object is destroyed while the ioctl is iterating its fences. The sources describe a denial-of-service style hang on the kernel side, not code execution.","category":"infrastructure","nicknames":[],"who_cares":["linux","kernel","intel-graphics","i915"],"confidence":0.5,"needs_auth":"unknown","worst_case":"local user can hang a kernel thread indefinitely","analyst_note":"Confirm fixed kernel versions and whether unprivileged local users can reach the busy ioctl, using the NVD entry and the upstream kernel stable commit.","how_to_check":"- Check the running kernel with `uname -r` and compare against the fixed stable kernel releases listed in the upstream commit 5113479556025093bf8133bb2dcaa33be2d50921\n- Confirm the i915 driver is loaded with `lsmod | grep i915`","needs_analyst":true,"not_affected_if":["no-i915-hardware","i915-not-loaded"],"who_is_affected":"Systems running a Linux kernel with the i915 Intel graphics driver, typically desktops, workstations and laptops with Intel integrated graphics. Systems without i915 hardware or with the module not loaded are not exposed; the sources do not say how an unprivileged user reaches the ioctl.","user_interaction":"no","needs_special_setup":"unknown","severity_if_unknown":"unknown"},"cost_usd":0.000686,"input_tokens":3618,"output_tokens":648,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'-2026':13A,16A '-98521':14A,17A '5113479556025093bf8133bb2dcaa33be2d50921':261C 'address':246C 'assum':164C 'bot':256C 'busi':32B,85C,129C 'case':181C 'caus':9A 'cherri':257C 'code':65B 'commit':260C 'concurr':125C 'cve':12A,15A 'denial':55B 'denial-of-servic':54B 'describ':52B 'destroy':42B,124C,160C,184C 'dma':88C,147C,152C,161C 'driver':5A,30B 'drm/i915':77C 'drop':144C,214C 'endless':10A,208C 'execut':66B 'fenc':49B,92C,98C,148C,156C,162C,168C,189C,216C 'finish':204C 'fix':78C,225C 'follow':72C 'forev':36B 'free':137C 'freed':198C 'gem':39B,84C,101C,121C,128C,139C,185C 'gpu':4A 'grace':222C 'graphic':22B,29B 'hang':59B 'happen':116C 'hold':104C 'i915':3A,18B,28B,83C,127C 'incorrect':79C 'infrastructur':19B 'instal':171C,193C 'intel':21B 'intel-graph':20B 'introduc':232C 'ioctl':33B,45B,86C,130C 'iter':47B,95C,202C 'kernel':2A,23B,62B,70C 'linux':1A,24B,27B,69C 'list':169C,190C 'lock':113C 'loop':11A,35B,177C,209C 'make':238C 'memori':141C 'necessari':229C 'need':244C 'never':192C,203C 'new':167C,188C 'object':40B,102C,122C,140C,186C 'old':196C 'one':197C,237C 'order':8A,82C 'patch':231C 'period':223C 'pick':258C 'point':251C 'problem':234C,249C 'rcu':6A,80C,110C,221C 're':174C 're-start':173C 'read':111C 'refer':106C,149C,217C 'resolv':76C 'result':205C 'resv':89C,153C 'run':133C 'sashiko':255C 'sashiko-bot':254C 'see':158C 'servic':57B 'side':63B,112C 'solut':211C 'sourc':51B 'start':175C 'still':132C,143C 'style':58B 'tag':226C 'teardown':7A,81C 'therefor':200C 'unlock':93C,157C 'use':87C 'vulner':73C 'without':103C 'won':135C 'wors':241C","nicknames":[],"mitigations":[],"who_affected":"Systems running a Linux kernel with the i915 Intel graphics driver, typically desktops, workstations and laptops with Intel integrated graphics. Systems without i915 hardware or with the module not loaded are not exposed; the sources do not say how an unprivileged user reaches the ioctl.","tier":"C","verified":{},"analyst_revisit_at":null,"first_public_at":"2026-10-18T16:17:03+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-98521","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-98521","severity":null,"published_at":"2026-10-18T16:17:03+00:00","modified_at":"2026-10-19T10:19:06+00:00","first_seen_at":"2026-10-23T21:57:15+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"}