{"id":"CVE-2026-64586","aliases":[],"url":"https://o3.security/vulnerability/CVE-2026-64586","summary":"wifi: brcmfmac: drain bus_reset work on device removal","details":"In the Linux kernel, the following vulnerability has been resolved:\n\nwifi: brcmfmac: drain bus_reset work on device removal\n\nbrcmf_fw_crashed() and the debugfs \"reset\" entry both schedule\ndrvr->bus_reset, whose callback recovers drvr through container_of()\nand dereferences it.  The removal path frees drvr (brcmf_free ->\nwiphy_free) without draining the work, so a bus_reset callback pending\nor running during removal can outlive drvr.\n\nCancellation cannot live in brcmf_detach() or brcmf_free(): the work\ncallback reaches teardown through the bus .reset op (PCIe\nbrcmf_pcie_reset -> brcmf_detach; SDIO brcmf_sdio_bus_reset ->\nbrcmf_sdiod_remove -> brcmf_free), so cancelling there would wait for\nthe running work and deadlock.\n\nAdd a per-bus mutex (bus_reset_lock) and route all arming through\nbrcmf_bus_schedule_reset(), which under the lock skips when the bus is\nmarked removing.  Each bus remove entry calls\nbrcmf_bus_cancel_reset_work(), which under the same lock sets removing\nand cancels the work.  Holding the mutex across cancel_work_sync() makes\nthe set-removing + drain step atomic.  Every producer reaches the arming\npath from process context -- the PCIe firmware-halt notification runs in\nthe threaded IRQ handler (brcmf_pcie_isr_thread) and the SDIO hostmail\npath runs from the data workqueue -- so the mutex is taken only in\nsleepable contexts.  Where applicable the remove entry first stops the\nfirmware-crash producer: on PCIe mask the mailbox and synchronize_irq;\non SDIO unregister the bus interrupt and cancel the data worker, which\nalso reports firmware halts through brcmf_fw_crashed().  The mutex is\ninitialized at bus allocation.  The SDIO suspend power-off path frees\ndrvr through the same brcmf_sdiod_remove() and takes the same lock;\nresume re-allows the work only on a successful re-probe.\n\nAlso guard brcmf_fw_crashed() against a NULL bus_if/drvr: it can fire\nbefore brcmf_attach() wires up drvr, and it dereferences drvr\n(bphy_err/brcmf_dev_coredump) before reaching the arming gate.\n\nThe bus_reset work is shared across buses, so the drain is applied to\nevery remove path: PCIe (the .reset op introduced by the Fixes commit),\nSDIO (arms the same work through brcmf_fw_crashed()), and USB (via the\ndebugfs \"reset\" entry).  cancel_work_sync() drains a running or pending\nbus_reset work item before removal frees drvr, and patch 1/2 makes the\nscratch-buffer release safe when reset teardown has already released\nthose DMA buffers.\n\nThis patch fixes the lifetime of the bus_reset work item itself.  It does\nnot attempt to address the separate, pre-existing lifetime of the\nasynchronous firmware completion started by the PCIe reset path.  That\ncallback needs its own lifetime/ownership protocol and is being tracked\nseparately.\n\nThis issue was found by an in-house static analysis tool.","published":"2026-08-06T07:06:27.158Z","modified":"2026-08-08T03:32:28.963006876Z","cvss":null,"epss":null,"cisaKev":null,"exploitsKnown":0,"affectedPackages":[{"ecosystem":"Linux","name":"Kernel","fixedVersion":"7.1.6"}],"fix":null,"references":[{"type":"WEB","url":"https://git.kernel.org/stable/c/177a25be1195f8bdc6160ba5f1a5699f7041c985"},{"type":"WEB","url":"https://git.kernel.org/stable/c/43b25879f004c98defa2776bedc6ca4763c51945"},{"type":"ADVISORY","url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/64xxx/CVE-2026-64586.json"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-64586"},{"type":"PACKAGE","url":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git"}],"provenance":{"sources":["OSV.dev","FIRST.org (EPSS)"],"lastVerified":"2026-08-08T03:32:28.963006876Z"}}