DF-5573 — ppb_MS_init_msq(): instruction index ins is never bounds-checked — controlled OOB kernel write of up to ~8 KB from the microsequence array

Severity: Low · Confidence: certain · File: sys/bus/ppbus/ppb_msq.c (pass 2)

Reproduce

1. sh build.sh — compiles the verbatim production ppb_msq.c into msqhx.ko
2. sh run.sh — case 0 (non-fatal)
3. Observe dmesg (dmesg | grep MSQHX) / the serial console.

Expected

BASELINE (unfixed): 'OOB ins=60 ... CONTROLLED WRITE LANDED' and 'OOB ins=200 ... CONTROLLED WRITE LANDED' — kernel memory outside the caller's array is written with fully controlled 4/8-byte values. PATCHED: in-bounds init still works; ins=200 (>= PPB_MS_MAXLEN) fails fast with 'panic: ppb_MS_init_msq: instruction out of range (200)' before any write.

Harness (shared by DF-5573..DF-5577)

msqhx.c + Makefile compile the verbatim production /usr/src/sys/bus/ppbus/ppb_msq.c (sha256 223e7f9184c81ab83f49228ae7d1f99a 6e3936b9a452b01c6bb6176b150dda67, identical to the repo sys/ copy) into a KLD, together with a fake ppbus controller on nexus whose ppbus_exec_microseq method implements the exact dispatch semantics of ppc_exec_microseq() (sys/dev/misc/ppc/ppc.c:1332-1542) for the exercised opcodes. The interpreter target device's softc is a real struct ppb_data and its child's ivars are a real struct ppb_device, so mode2xfer(), ppb_MS_init(), ppb_MS_exec(), ppb_MS_microseq(), ppb_MS_loop() and ppb_MS_init_msq() all run their production code paths. Evidence class: "KLD-harness-with-production-code" (no parallel port in the QEMU rig) — family standard, cf. DF-5418.

Build

# inside the guest, from this directory
sh build.sh          # -> /tmp/msqhx-build/msqhx.ko

Run

sh run.sh            # runs this pack's case via sysctl hw.msqhx_case
sh run.sh <case>     # 0=suite(5575+5573) 2=5574 4=5576a 5=5576b 6=5577

Cases 2/4/5 PANIC the guest on unfixed production code (that is the finding); the panics land on the serial console (dfbsd-qemu/vm.sh log).

Fix validation

fix.diff (sha256 8e3dfc90e6049ce1704c3cef833cba0904b6404f11360bdd299e37980e9327bb) patches sys/bus/ppbus/ppb_msq.c + sys/bus/ppbus/ppbconf.h (all five findings; never applied to the repo's read-only sys/ tree). Validation procedure: copy the fixed files into the guest's /usr/src/sys/bus/ppbus/, sh build.sh again (recompiles the production source), re-run the same case. Baseline vs patched outputs are in run.log / fix_run.log (plus panic*.txt / fix_run.log for the fatal cases).

DF-5573 — verdict narrative

Finding

ppb_MS_init_msq(): instruction index ins is never bounds-checked — controlled OOB kernel write of up to ~8 KB from the microsequence array

Bottom line (status: reproduced, impact: none)

Reproduced on the QEMU guest (DragonFly 6.5-DEVELOPMENT x86_64) via the KLD-harness-with-production-code evidence class: the unmodified /usr/src/sys/bus/ppbus/ppb_msq.c (sha256 223e7f91..., identical to the repo sys/ copy) was compiled with -Werror into msqhx.ko and driven against a fake ppbus controller implementing the ppc_exec_microseq() contract. ppb_MS_init_msq() extracts the instruction position with MS_INS(param) (8 bits wide, ppb_msq.h:38) but validates only the argument index (arg >= PPB_MS_MAXARGS panic at ppb_msq.c:206-208); ins is used directly as msq[ins].arg[arg] at ppb_msq.c:218/222/226/230 with no bound. Harness case 0 proves the primitive on a private kmalloc'd 256-instruction buffer: MS_PARAM(60,0,MS_TYP_INT) writes 0x41424344 at +1792 bytes past a 4-instruction caller array and MS_PARAM(200,2,MS_TYP_PTR) writes a pointer value at +6416 bytes (max reachable: ins=255, arg=2, MS_TYP_PTR/FUN -> +8184 bytes, 8-byte value). A buggy or third-party KLD driver that computes a descriptor from runtime data (instead of the compile-time MS_PARAM constants all in-tree callers use) gets silent kernel memory corruption rather than the panic the half-sibling arg check provides. Impact ceiling: with no unprivileged input path into descriptors (ppi's user-reachable microseq code is #if 0'd, ppi.c:419-442; all other callers pass static descriptors) this is a kernel-API hardening defect (driver-misuse -> corruption), not a direct privilege boundary; severity Low, bucket memcorrupt-class primitive. Fix validated: ins >= PPB_MS_MAXLEN(64) now panics fail-fast ('instruction out of range (200)', production backtrace ppb_MS_init_msq+0x1f9 in fix_run.log) and the in-bounds regression (ins=2) still initializes correctly. Residual: the function cannot know the caller's actual array length, so ins<64 remains trusting; a complete fix would pass the array length (KPI change).

Exploit chain

n/a (kernel-internal API; no unprivileged descriptor input exists in-tree — primitive characterized on a private buffer instead)

Evidence

run.log lines 'OOB ins=60 ... CONTROLLED WRITE LANDED' / 'OOB ins=200 ... CONTROLLED WRITE LANDED' (baseline, pristine production code); fix_run.log (in-bounds ok + 'panic: ppb_MS_init_msq: instruction out of range (200)' with production backtrace); build.log ('-c /usr/src/sys/bus/ppbus/ppb_msq.c' with -Werror, sha256 223e7f91...); VERDICT.md

Kernel references

• sys/bus/ppbus/ppb_msq.c:201
• sys/bus/ppbus/ppb_msq.c:206
• sys/bus/ppbus/ppb_msq.c:218
• sys/bus/ppbus/ppb_msq.c:222
• sys/bus/ppbus/ppb_msq.c:226
• sys/bus/ppbus/ppb_msq.c:230
• sys/bus/ppbus/ppb_msq.h:38
• sys/bus/ppbus/ppbconf.h:135

Fix validation (fixed)

Applied 'ins >= PPB_MS_MAXLEN -> fail-fast panic' to the guest's /usr/src copy and rebuilt the harness: the ins=200 OOB write is now blocked by 'panic: ppb_MS_init_msq: instruction out of range (200)' (production frame ppb_MS_init_msq+0x1f9) and the in-bounds ins=2 initialization still works. Residual documented: ins<64 cannot be checked against the caller's real array length without a KPI change.

fix.diff (sha256 8e3dfc90e6049ce1704c3cef833cba0904b6404f11360bdd299e37980e9327bb) carries all five ppb_msq.c fixes for this pass; the hunks relevant to this finding are marked with its DF id in the comments. Fix validation followed the family standard (cf. DF-5418): the fixed sources were installed into the guest's /usr/src/sys/bus/ppbus/, the harness was rebuilt (recompiling the production file), and the exact same case was re-run — baseline vs patched outputs are in run.log / fix_run.log (+ panic*.txt). The repo's sys/ tree was never modified; the guest was vm.sh reset with-src afterwards (sources restored+verified by snapshot).

PoC changes vs seed

No seed PoC existed; harness written fresh. Iterations: (1) load-time kenv tunable replaced by sysctl hw.msqhx_case after the first module could not be unloaded (drivers initially lacked detach methods); (2) diagnostic reads of the fix-added ppb_in_xfer field removed so the same harness compiles against pristine and patched headers. Production ppb_msq.c compiled clean on first attempt.

Guest: DragonFly dfbsd 6.5-DEVELOPMENT DragonFly 6.5-DEVELOPMENT #0: Thu Jul 2 06:02:54 UTC 2026 root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC x86_64 Harness sha256: 4b728315b40157934f62c99e847f386b30202eef77083dc1a5c065b7b0448ded

Fix status (batch validation)

Applied with the 36 sibling fix.diffs of this batch in findings/poc/_batch41/all_fixes.patch; single in-guest make -j6 nativekernel KERNCONF=X86_64_GENERIC completed RC=0 with zero errors under -Werror (full log: findings/poc/_batch41/fix_build8.log). Behavior A/B intentionally skipped for this Info-tier item.