CVE-2024-53153

What it actually does

Race condition in qcom_pcie_perst_assert causes dw_pcie_ep_cleanup and pci_epc_deinit_notify to execute after host disables refclk, resulting in unclocked hardware register accesses that crash the endpoint controller.

Why it matters to the business

Endpoint devices using affected Qualcomm SoCs can experience sudden crashes and loss of PCIe connectivity during normal host link resets, leading to unplanned downtime for storage, networking, or other endpoint functions that rely on the controller.

How an attack unfolds

An attacker with control of the host side asserts PERST#; the endpoint immediately begins cleanup routines that require register access, but refclk is already removed, triggering a hardware-level crash of the endpoint SoC before any further code can run.

How to fix it

Update the Linux kernel to commit e03b5f1615c84f4139cb53ef8659f4cdb8d6a563 or later stable releases that move the cleanup routines into qcom_pcie_perst_deassert after refclk is guaranteed active; rebuild and deploy the updated kernel on all affected endpoint systems.

How to confirm the fix worked

Confirm the running kernel contains commit e03b5f1615c84f4139cb53ef8659f4cdb8d6a563; trigger a controlled PERST# assert/deassert cycle from the host and verify the endpoint remains operational without register-access crashes.

Typical attack path

1. Attacker gains ability to assert PERST# on the host-to-endpoint link
2. Host performs link reset sequence that disables refclk shortly after PERST#
3. Endpoint executes qcom_pcie_perst_assert which invokes register-dependent cleanup
4. Unclocked register access causes immediate endpoint controller crash

How to detect it

Review kernel logs on endpoint systems for sudden crashes or register access faults immediately following PCIe link reset events; correlate host-side PERST# assertions with endpoint watchdog or panic timestamps.

What your security team should do

• Prioritize kernel upgrade on all Qualcomm endpoint platforms within standard change windows
• Test the fixed kernel under realistic host reset workloads before broad rollout
• Monitor vendor stable branches for additional backports of the same commit
• Document endpoint crash rates during link resets as a pre-patch health metric