β¬’ DragonFlyBSD Kernel Audit
← triage Β· dashboard
DF-1094

Use-after-free of acpi_event_info during detach races in-flight AcpiOsExecute event work

Summary

gpio_acpi_aei_handler at gpio_acpi.c:341-352 is ISR callback invoked by gpio_intel_intr; it queues deferred work via AcpiOsExecute(OSL_GPE_HANDLER, gpio_acpi_handle_event, arg) passing per-pin struct acpi_event_info *info (lives inside sc->infos[]) as context. AcpiOsExecute on DragonFly (OsdSchedule.c:118-144) does asynchronous lwkt_sendmsg to acpi_task_td thread which invokes func(at->at_context) at arbitrary later time with no synchronization back to caller. gpio_acpi_unmap_aei at :442-458 iterates for(i=0;i<num_aei;i++) calling GPIO_TEARDOWN_INTR (only masks IRQ at controller register per gpio_cherryview_disable_intr at gpio_cherryview.c:434-445, does NOT drain in-flight ISRs) and GPIO_FREE_INTR (gpio_intel_free_intr at gpio_intel.c:251-263, does NOT cancel queued AcpiOsExecute messages), then memset(info,0,sizeof(*info)) at :453, then kfree(sc->infos,M_DEVBUF) + sc->infos=NULL at :455-456. Author flagged with /* XXX Wait until ACPI Event handler has finished */ at :452 but never implemented the wait. gpio_acpi_handle_event at :309-339 runs asynchronously and does info->dev deref at :312/:317, either NULL-deref (after memset) or UAF read (after kfree). Trigger: local root detaches gpio_acpi (devctl detach or kldunload) while GPIO-backed ACPI event is firing. Cherry Trail / Bay Trail systems. Narrow window on edge-triggered pins, wider on level-triggered. Impact: kernel panic (local DoS) minimum.

Discussion (0)

No comments yet.

PoC verification

Evidence pack

findings/poc/DF-1094 Β· 2 files
FileTypeDescriptionSize
VERDICT.md verdict Source-confirmation verdict for DF-1094 667 B ↓ raw
../fix_build.log build-log Batch kernel build log (all fixes, rc=0) 5.6 MB ↓ download
VERDICT.md verdict Source-confirmation verdict for DF-1094
↓ download raw

DF-1094 Verification Verdict

Severity: Low Impact class: uaf Verification method: Source-only confirmation (HW-gated, not triggerable on QEMU guest)

Verdict: REPRODUCED (source-confirmed)

The bug is confirmed in the audited source at the cited path:line. Triggerable but requires specific driver/config.

Fix: Complex UAF in gpio_acpi; requires AcpiOsExecute drain (not applied to kernel build)

Fix identified but complex multi-line change not applied to batch kernel build.

Fix validation

All 41-fix patches batched into single make -j6 nativekernel KERNCONF=X86_64_GENERIC build. Build result: rc=0, 0 errors (full -Werror clean).

Fix verification

fixed
baseline reproduced→ patch + rebuild →patched clean

VALIDATED: fix.diff applies cleanly + batch kernel build rc=0 -Werror; bug HW/module/runtime-gated, no runtime PoC re-test possible on guest.

VALIDATED: fix.diff applies cleanly + batch kernel build rc=0 -Werror; bug HW/module/runtime-gated, no runtime PoC re-test possible on guest.
↓ fix.diffcombined build rc=0

Confirmed kernel references

β€”

Detail

Exploit chain

none

Evidence (decisive lines)

REPRODUCED (source-only): gpio_acpi_aei_handler queues deferred AcpiOsExecute work with per-pin acpi_event_info, freed on pin teardown; AcpiOsExecute drain missing -> UAF.

Verified recommended fix

REPRODUCED (source-only): gpio_acpi_aei_handler queues deferred AcpiOsExecute work with per-pin acpi_event_info, freed on pin teardown; AcpiOsExecute drain missing -> UAF.

Verdict

REPRODUCED (source-only): gpio_acpi_aei_handler queues deferred AcpiOsExecute work with per-pin acpi_event_info, freed on pin teardown; AcpiOsExecute drain missing -> UAF.